Structural Analysis of a Prokaryotic Ribosome Using a Novel Amidinating Cross-Linker and Mass Spectrometry

Transcription

1 pubs.as.org/jpr Strutural Analysis of a Prokaryoti Ribosome Using a Novel Amidinating Cross-Linker and Mass Spetrometry Matthew A. Lauber and James P. Reilly* Department of Chemistry, Indiana University, Bloomington, Indiana 47405, United States bs Supporting Information ABSTRACT: The struture of the Esherihia oli ribosome, a 2.5 MDa ribonuleoprotein omplex ontaining more than 50 proteins, was probed using the novel amidinating ross-linker diethyl suberthioimidate (DEST) and mass spetrometry. Peptide ross-links derived from this omplex struture were identified at high onfidene (FDR 0.8%) from preursor mass measurements and ollision-indued dissoiation (CID) fragmentation spetra. The aquired ross-linking data were found to be in exellent agreement with the rystal struture of the E. oli ribosome. DEST ross-links are partiularly amenable to strong ation exhange (SCX) hromatography, failitating a large-sale analysis. SCX enrihment and frationation were shown to inrease the number of ross-link spetra mathes in our analysis 10-fold. Evidene is presented that these tehniques an be used to study omplex interatomes. KEYWORDS: Protein ross-linking, mass spetrometry, amidination, diethyl suberthioimidate, DEST, ribosome, E. oli INTRODUCTION Protein protein interations are required for most biologial funtions. The abilities of RNA polymerase to transribe DNA, 1 the splieosome to proess gene transripts, 2 and the ribosome to translate mrna, 3 in partiular, require the orhestrated interations of many proteins. Unfortunately, the quaternary strutures that enable these interations have been hallenging to study. For example, although X-ray rystallography an provide atomi resolution, it is often very diffiult to produe and rystallize large maromoleular omplexes due to their dynami nature. 4 It has therefore beome apparent that omplementary methods are needed to best eluidate the strutures and interations of maromoleular omplexes. 4 Chemial ross-linking an be applied to study quaternary strutures and thus aid the development of hybrid struture models. It an apture dynami interations without being markedly hindered by sample strutural heterogeneity. Modern mass spetrometers have even made it possible to identify rosslinks and the exat residues involved in linkages with reasonably high onfidene. 5,6 In most ross-linking analyses, protein funtional groups are targeted for derivatization with a moleule that ontains two reative groups separated by a spaer arm of known length. Only funtional groups loser than the length of the spaer arm are apable of being linked. Identifiation of the residues involved in a ross-link thereby provides distane onstraints for strutural modeling. Despite these advantages, the potential of ross-linking studies an be limited by several analytial hallenges. Detetion of ross-linked peptides in the proteolyti digests of derivatized proteins is often impeded by the ombination of their low stoihiometri yield and the presene of other peptide speies. Enrihment tehniques have aordingly beome an intense fous for researhers in the field. 7 9 Due to the ombinatorial nature of ross-linking reations, mixtures ontaining ross-linked peptides an be extremely heterogeneous. Their frationation has therefore been neessary in the few large-sale ross-linking studies onduted to date. 10,11 Lastly, identifiation of ross-links typially requires mass auraies on the order of a few parts per million if large sequene databases are to be searhed. To failitate ross-linking studies, we previously developed the novel amidinating ross-linker, diethyl suberthioimidate (DEST). 12 This reagent is a water-soluble, eight-atom (11 Å) spaer arm ross-linker, similar to ommonly used and ommerially available ross-linkers, suh as bissulfosuinimidyl suberate (BS3). Unlike its analogs, however, DEST modifies primary amines at physiologial ph without sarifiing their native basiity. As a result, the use of DEST does not perturb the eletrostati properties of a protein and is thus unlikely to disrupt native protein struture. The fat that this reagent preserves the basiity of targeted amines upon modifiation also means that the ionization effiieny of the residue it modifies is not adversely affeted and that the ross-links it forms an be easily separated from other omponents of trypti digests using strong Reeived: Marh 21, 2011 Published: May 27, 2011 r 2011 Amerian Chemial Soiety 3604 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

2 Journal of Proteome Researh ation exhange (SCX) hromatography. We have previously employed this ross-linker to observe intraprotein ross-linking of ytohrome 12 and have subsequently beome interested in applying it to study very omplex strutures, as a step toward understanding their dynami interatomes. One suh omplex of interest is the Esherihia oli ribosome, the ellular mahine responsible for translating mrna into protein. It is a 2.5 MDa ribonuleoprotein omplex omposed of 53 unique proteins and is one of a few large maromoleular omplexes with a struture, albeit inomplete, determined by X-ray rystallography. The E. oli ribosome, as well as other prokaryoti ribosomes, is thus a rather unique system in whih to investigate the feasibility of large-sale mass spetrometri studies. In previous work, we have demonstrated the effetiveness of monofuntional thiomidates as hemial labeling reagents by probing the tertiary and quaternary strutures of the proteins in prokaryoti ribosomes and demonstrating that the labeling of proteins orrelates extremely well with preditions derived from rystal strutures The study of ribosome ross-linking is far more hallenging. Although there are only 678 unique modifiable amines in the sample, the ombinatorial nature of ross-linking means there are about potential modifiation produts with whih to be onerned. In the present work, we have probed the struture of the E. oli ribosome using the DEST ross-linker and mass spetrometry. This is one of the most omplex strutures to whih ross-linking ombined with mass spetrometry has been applied. Peptide ross-links derived from this struture were identified with high onfidene from single preursor mass measurements and ollision-indued dissoiation (CID) fragmentation spetra. Data ould be diretly ompared with the rystal struture of the E. oli ribosome. This study also demonstrates the ompatibility of DEST ross-linking with ation exhange hromatography. Finally, evidene of ribosome ribosome interations in the data suggests that these tehniques onstitute a viable method for studying omplex interatomes. EXPERIMENTAL PROCEDURES Materials Aetonitrile (ACN), glaial aeti aid, hydrohlori aid, sodium hydroxide, trifluoroaeti aid (TFA), and water were purhased from EMD Chemials (Gibbstown, NJ). Anhydrous diethyl ether was obtained from Fisher (Fair Lawn, NJ). Proteomis grade trypsin (T-6567), magnesium aetate tetrahydrate, Trizma base, 4-(2-hydroxyethyl)-1-piperazineethanesulfoni aid (HEPES), phenylmethylsulfonyl fluoride (PMSF), and surose were purhased from Sigma (St. Louis, MO). Ammonium hloride, formi aid (FA), magnesium hloride hexahydrate, 2-meraptoethanol, and suberonitrile were obtained from Aldrih (Milwaukee, WI). Calium hloride dihydrate, dihloromethane, (ethylenedinitrilo)tetraaeti aid (EDTA), sodium hloride, sodium phosphate monobasi, and type 3A moleular sieves were purhased from Mallinkrodt Baker (Phillipsburg, NJ). Ethanethiol was obtained from Aros Organis (Pittsburgh, PA). Hydrogen hloride (tehnial grade) was purhased from Matheson (Cuamonga, CA). Bato-tryptone and bato-yeast extrat were provided by Beton, Dikinson and Company (Sparks, MD). Complete EDTA-free Protease Inhibitor ( Mini ) Tablets were supplied by Rohe Applied Siene (Indianapolis, IN). Rapigest SF was purhased from Waters (Milford, MA). Synthesis of Diethyl Suberthioimidate (DEST) As desribed previously, 12 diethyl suberthioimidate (DEST) was prepared from ethanethiol and suberonitrile via the Pinner synthesis (Supporting Information, Sheme 1). 18 Suberonitrile (9 mmol) in anhydrous dihloromethane (1:3 v/v) was added to ie-old ethanethiol (90 mmol). The reation mixture, while being kept on ie and onstantly stirred, was sparged with hydrogen hloride gas for 1 h and subsequently kept at 4 C for an additional 16 h. Anhydrous diethyl ether was then added to aid preipitation, and the mixture was stored at 20 C until a solid had formed. This solid, after being washed several times with anhydrous diethyl ether, was stored in a vauum desiator at 4 C until needed. Preparation of E. oli Ribosomes E. oli K12 ells were grown for the preparation of ribosomes. Starter ultures were grown overnight at 37 C while being aerated at 180 rpm. Ribosomes were isolated from ells as previously desribed by Spedding. 19,20 Cross-Linking Conditions and Sample Preparation Solutions of ribosomes were exhanged into phosphate-buffered saline (20 mm sodium phosphate/150 mm sodium hloride, ph 7) ontaining 10 mm magnesium aetate using Amion Ultra 10K entrifugal filter devies (Millipore, Eshborn, Germany) prior to ross-linking. Reations for two biologial repliates, eah with two tehnial repliates, were arried out at a total protein onentration of 0.5 mg/ml (an approximately 0.7 μm ribosome solution assuming there is g of protein per mole of ribosomes) and DEST onentration of 2.5 mm. The protein onentration of the ribosome solutions was estimated by means of a Bradford assay, in whih bovine serum albumin was used as a standard. After proeeding at room temperature for 12 h, the reations were quenhed by addition of 0.5 M Tris to a final onentration of 50 mm. These are onditions that allow DEST to reat until it is near fully hydrolyzed; these are also onditions that yield only partial modifiation of proteins. 12 Shorter reation times may be neessary for studying less stable biologial systems. To preipitate the rrna, glaial aeti aid and 1 M magnesium hloride were added to the reation mixture suh that the final solution ontained 3:6:1 (v/v/v) reation mixture/glaial aeti aid/1 M magnesium hloride. The samples were vortexed and allowed to remain at room temperature for 40 min before the rrna preipitate was separated by entrifugation at g for 20 min. Supernatant from eah reation mixture, ontaining soluble ross-linked and unmodified ribosomal proteins, was desalted and leared of hydrolyzed reagent by exhange into 50 mm Tris using Amion Ultra 3K entrifugal filter devies (Millipore, Eshborn, Germany). The onentrate was then dried and stored at 20 C until later sample preparation for proteolyti digestion. Proteolyti Digestion Dried aliquots of DEST-modified proteins (40 μg) and proteomis grade trypsin (2 μg) were reonstituted in Rapigest (Waters, Milford, MA) ontaining solution, suh that the digestion was arried out in 18 μl of 100 mm Tris/10 mm alium hloride (ph 8) and 0.2% (w/v) Rapigest. Eah digest reation was allowed to proeed at 37 C for 24 h and subsequently quenhed by adding 10% TFA to a final onentration of 1%. To hydrolyze Rapigest, the quenhed digests were inubated for 3605 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

3 Journal of Proteome Researh 30 min at 37 C. Insoluble Rapigest byprodut was leared from the samples by entrifuging at g. SCX Enrihment and Frationation of DEST Cross-Links Strong ation exhange (SCX) hromatography was used in two ways to simplify proteolyti digests of ross-linked proteins prior to their analysis by nanolc-ms/ms. Interpeptide ross-links were enrihed from non-ross-linked peptide (linear peptide) speies via a low ph separation and then frationated via an intermediate ph separation (Sheme 2). Interpeptide ross-links were enrihed using SCX hromatography by exploiting the fat that they ontain more basi funtional groups, and thus positive harges at low ph, than other speies in a proteolyti digest. We have previously demonstrated the utility of this approah. 12 As before, trypti digest of the ross-linked sample was loaded on an SCX olumn (TSKgel SP-NPR, 4.6 mm 35 mm, Tosoh Biosiene, Montgomeryville, PA) equilibrated with 0.1% TFA in water (ph 2). Most nonross-linked (linear peptide) speies were eluted from the olumn with 300 mm NaCl mobile phase. Interpeptide rosslinks were then eluted from the SCX olumn onto a C18 trapping olumn (Thermo Hypersil-Keystone Javelin, 1.0 mm 20 mm, Bellefonte, PA) with mobile phase ontaining 1000 mm NaCl. After being desalted, the ontents of the C18 trapping olumn were eluted with organi mobile phase. This enrihed fration was dried under vauum and resuspended in 0.1% TFA for subsequent LC-MS/MS analysis. The amount of sample remaining after the 300 mm salt wash step of the enrihment was estimated via a miro-bca assay. 21,22 The A 562 was measured for the sample before it was loaded onto the SCX olumn as well as after it was proessed via the above proedure. Based on two biologial repliates, the amount of sample remaining after the salt wash proedure was estimated to be 15% of the original sample mass. DEST interpeptide ross-links were also subjeted to frationation using SCX hromatography. This hromatography was diretly integrated with the enrihment of DEST interpeptide ross-links. After the 300 mm salt wash of the enrihment, the mobile phases were hanged to a 20 mm sodium aetate, ph 5, buffer, and a 100 min gradient from 0 to 300 or 400 mm NaCl, followed with a 10 min isorati hold of 1000 mm NaCl, was implemented. DEST interpeptide ross-links eluting from the SCX olumn were olleted sequentially onto 10 C18 trapping olumns (Thermo Hypersil-Keystone Javelin, 1.0 mm 20 mm, Bellefonte, PA). An apparatus similar to one previously desribed was used to swith the flow path from one trapping olumn to the next every 10 min. 23 After being desalted with 0.1% TFA in water at 0.3 ml/min for 20 min, the ontents of the C18 trapping olumns were sequentially eluted with a 10 min isorati hold (flow rate 100 μl/min) of 5% aqueous mobile phase (0.1% TFA in water) and 95% organi mobile phase (0.1% TFA in ACN). The entire eluate from eah trap was dried under vauum. These frations of the SCX were then resuspended in 0.1% TFA in water for subsequent LC-MS/MS analysis. Capillary LC-ESI-MS/MS Capillary LC-ESI-MS/MS was performed using an IntegraFrit apillary trapping olumn paked with 1.5 m of C18 (150 μm 11 m, New Objetive, Woburn, MA; Magi C18, 5 μm, 200 Å, Mihrom BioResoures, Auburn, CA), a apillary analytial olumn paked with 15 m of C18 (75 μm 15 m, Magi C18, 5 μm, 100 Å, Mihrom BioResoures, Auburn, CA), an LTQ Orbitrap XL mass spetrometer (Thermo Eletron, Bremen, Germany), and a Dionex hromatography system (Ultimate 3000, Dionex, Sunnyvale, CA). In eah experiment, approximately 1 μg of trypti digest was injeted onto a trapping olumn to remove salts and ontaminants by flushing for 10 min with mobile phase A (0.1% FA in 97:3 water/acn) at a flow rate of 10 μl/min. The flow rate was then redued to 0.3 μl/min, effluent from the trapping olumn was direted to the apillary LC olumn, and a 100-min gradient between 0% and 35% mobile phase B (0.1% FA in ACN) was implemented. Eluting peptides were eletrosprayed into an LTQ-Orbitrap mass spetrometer operating in data-dependent mode to aquire a full MS san ( m/z) and subsequent MS/MS sans of the three most intense preursor ions. The AGC target value was set to for MS sans and for MS/MS sans. CID of the preursors ourred in the LTQ at 35% normalized ollision energy. Isolation width was set to 2 m/z and monoisotopi preursor seletion was enabled. Both the MS and MS/MS sans were aquired in the orbitrap with resolution set to and 7500, respetively. Dynami exlusion was employed with the following settings: a 90 s exlusion duration time, maximum exlusion list of 500, and one repeat ount. In addition, harge state rejetion was enabled for 1þ, 2þ, and unassigned harge states, exept when it was of interest to determine the hanges in sample omposition due to SCX enrihment. Beause rosslinked peptides formed in these experiments tend to have a higher harge state upon ESI than non-ross-linked (linear peptide) speies, rejetion of low harge states from MS/MS aquisition provides a useful bias for deteting ross-links. MS/ MS spetra were subjeted to data redution using Masot Distiller (version ), suh that preursor masses were redetermined via interpretation of isotopi distributions and MS/MS spetra were deisotoped. Proessed data were saved as.dta files then merged into.mgf files with syntax appropriate for database searhing. 11 xquest Database Searhing MS/MS data ontained in merged.mgf files were subsequently searhed against the sequenes of onstitutive proteins from the E. oli 30S and 50S ribosomal subunits for interpeptide ross-links using the web interfae of xquest (version 2.721). 11 This is an algorithm that an be used to searh for theoretial ross-links that have masses mathing measured preursor masses and to subsequently assign the fragment masses of MS/ MS spetra. In its interpretation of MS/MS spetra, xquest assumes that a ross-link preursor will fragment at only one peptide bond. For our searhes, xquest was used with default settings, exept that MS tolerane was set to 5 ppm, the MS/MS tolerane was set to 0.01 m/z, variable modifiation was set for methionine oxidation ( ), no fixed modifiations were used, and mass shifts for ross-linking produts were manually set to for the xlink mass-shift and (ADE), (TEDE), and (Tris dead-end) for the monolink mass-shifts. Mathes from all searhes were required to have preursor mass errors e4.3 ppm and g15% of the ion urrent in a given MS/MS spetrum assigned as b- and y-type ions. This value was found aeptable, sine xquest does not inlude ammonia and water loss fragment ions in its soring. MS/ MS mathes were also required to have normalized rossorrelation sores g0.052 for both ion series ontaining and not ontaining the ross-link mass shift. xquest mathes for intraprotein and interprotein rosslinks were required to meet different soring thresholds dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

4 Journal of Proteome Researh Intraprotein ross-link mathes were required to have xquest sores g25.5. Furthermore, a number of intraprotein ross-link mathes were manually inspeted to determine whether they indeed provided unambiguous identifiations. This was required for speial ases in whih the deteted ross-links were formed from two peptides, whih together omprised a ontiguous sequene from the protein. In this situation, a potential math ould also have been made to a dead-end modified peptide derived from the same sequene, beause it would have the exat same mass. These types of mathes were manually inspeted and disregarded if the fragmentation did not unambiguously indiate the rosslink struture. Interprotein ross-link mathes were required to have xquest sores g27.3 and were required to have xquest sores at least 2 units greater than their orresponding seond rank mathes. Moreover, interprotein ross-link mathes were disregarded unless there were three unique fragments assigned on eah peptide hain. Spetra mathes were assessed for this harateristi after assigning neutral losses and seond isotopi peaks. Fragments that ould be ambiguously mathed to the sequenes of either peptide hain in the ross-link were not inluded in this ount. The ross-link mathes were, in addition, required to have at least eight of these sequene-speifi fragments assigned. Finally, spetra that orresponded to ross-links, either intraprotein or interprotein, with multiple possible linkage patterns were manually inspeted. Exat linkages were only proposed when the two ross-linked residues ould be unequivoally defined by the observed fragmentation. False Disovery Rate Analysis The false disovery rate (FDR) of this analysis was first estimated by searhing files that produed mathes against randomized sequenes of the E. oli 30S and 50S ribosomal proteins. This yielded 2 intraprotein and 0 interprotein (non-intersubunit) deoy mathes, indiating that the FDR for our identified rosslink spetra mathes, whether intraprotein or interprotein, was less than 2%. It was lear, however, that this was not an aurate estimate of the FDR. When there are only a limited number of identified spetra mathes, a deoy database that is of equal size to a target database is inadequate for aurately defining a FDR. A solution to this problem is to use a deoy database that is larger than the target database and to sale the number of deoy mathes based on the differene in size between the deoy and target searh spaes. 24 We onstruted a large deoy database that retained the harateristis of the E. oli ribosomal proteins by onatenating and subsequently randomizing a database that ontained two opies of eah protein sequene. Searhes against this deoy database yielded 501 and 5 deoy mathes meeting the riteria for onfident intraprotein and interprotein mathes, respetively. FDRs for both the identifiation of intraprotein and interprotein ross-link mathes were alulated in two steps. The first step was to sale the number of deoy mathes by how many times larger the deoy searh spae was than the searh spae needed to identify either intraprotein or interprotein ross-links. The deoy searh spae ontained theoretial ross-linked peptides, while the searh spae needed to identify intraprotein ross-links ontained theoretial ross-linked peptides. Thus, the 501 deoy mathes orresponding to intraprotein ross-link mathes were saled to 2.0. Likewise, the searh spae needed to identify interprotein (nonintersubunit) ross-links ontained theoretial rosslinked peptides, so the 5 deoy mathes orresponding to the interprotein ross-link mathes were saled to 0.5. The FDRs were then obtained by dividing the saled number of deoy mathes by the number of mathes produed when searhing the experimental data against the ribosomal protein sequenes. Masot Database Searhing MS/MS data ontained in the.mgf files were searhed for peptide speies other than interpeptide ross-links (namely, linear peptide speies) using Masot (version 2.2.0) with a database ontaining the sequenes of E. oli ribosomal proteins. The MS and MS/MS toleranes for these searhes were 5 ppm and 0.01 m/z, respetively. Numerous variable modifiations for protein N-termini and lysine residues were inluded in these searhes, suh as the addition of amide dead ends (þ Da), thioester dead ends (þ ), Tris dead ends (þ ), and intrapeptide ross-links (þ Da). Other settings inluded a variable modifiation for methionine oxidation and an allowed number of missed trypti leavages equal to 6. Ion mathes with sores less than 17 (p > 0.005) were ignored. Protein Sequene Data The E. oli K12 proteome was obtained from The J. Craig Venter Institute (GenBank Aession. Version U ). Crystal Struture of the E. oli 70S Ribosome Distanes between R-arbons of lysine residues in the E. oli ribosome were determined using a program developed in-house and the oordinates from PDB files 2AW4 and 2AVY. 25 PyMOL v (DeLano Sientifi, was employed for the visualization and manipulation of rystal strutures. 26 RESULTS AND DISCUSSION Cross-Linking of the E. oli Ribosome The E. oli ribosome was subjeted to ross-linking with DEST, a novel bifuntional thioimidate reagent with an 11 Å spaer arm. Ribosomes prepared at less than intraellular onentrations (a. 1 μm versus 10 μm) 27,28 were modified in a physiologially relevant buffer with DEST present at a 5:1 reagent to protein amine ratio. DEST modifies the primary amines of proteins, thereby introduing a set of amidine-linked reation produts, inluding ross-links and so-alled dead ends. As noted in Sheme 1, ross-links between two amines introdue 136 Da mass shifts. Dead ends, whih form when one end of the reagent reats with an amine while the other end is hydrolyzed, introdue mass shifts of either 154 or 199 Da, depending on whether hydrolysis eliminates the thiol to form an amide (ADE) or eliminates ammonia to form a thioester (TEDE). To estimate the number of struturally possible DEST rosslinks supported by the E. oli ribosome, we investigated the distanes between modifiable residues in the previously reported rystal struture. 25 A speifi fous was made on lysine residues and not protein N-termini, given that lysine residues signifiantly out number protein N-termini and that the N-termini of several ribosomal proteins (L11, L16, L13, S5, S11, and S18) are not available for amidination due to post-translational methylation or aetylation. 15,20 The E. oli ribosome ontains 631 unique, modifiable lysine residues; 548 of these are modeled in the rystal struture. We alulated the distanes between R-arbons of these residues and ounted the number of values that were less than the maximum R-arbon to R-arbon ross-linking distane that DEST is apable of bridging. We defined this to be 24 Å, the 3607 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

5 Journal of Proteome Researh Sheme 1. Major Reation Produts of DEST Cross-Linking Sheme 2. SCX Enrihment and Frationation of DEST Interpeptide Cross-Links sum of DEST s spaer arm length and the length of two lysine side hains. This led to the estimate that there were at least 2200 lysine pairs within linkable distane in the ribosome sample. Of ourse, not all of the amines of lysine residues in the ribosome are solvent-aessible. In fat, judging by manual interpretation of the rystal struture, only 396 are fully aessible. 15 However, these fully aessible lysine residues are still apable of produing 1100 ross-links. A proteolyti digest of ross-linked ribosomal proteins was therefore expeted to be overwhelmingly omplex, beause it would ontain many peptides and many ross-links. Strategies for the Enrihment and Frationation of Cross- Links To redue the omplexity of suh a digest and failitate the detetion of ross-links, proteolyzed samples were subjeted to SCX hromatography. The fat that interpeptide ross-links formed by suinimidyl ester reagents have a onsiderable number of basi funtional groups has already been exploited by using SCX to partially separate ross-links from unmodified and deadend-modified peptides that happen to have fewer basi funtional groups. 10,11 DEST interpeptide ross-links should be better suited to suh physial enrihment, beause they ontain two additional basi funtional groups. DEST ross-links, if produed by trypti proteolysis, ontain at least six suh groups, while the undesired unmodified and dead-end-modified peptides of the digest tend to ontain only two or three. The tehnique, illustrated in Sheme 2, was simple to implement. 12 A trypti digest of DEST-modified ribosomal proteins was loaded onto an SCX olumn at low ph (ph 2), to ensure that all ionizable groups were protonated, and then washed with 300 mm sodium hloride. This salt onentration leared the sample of low harge speies without ausing signifiant loss of the more highly harged interpeptide ross-links. To further redue the omplexity of the digest before it was analyzed, we ombined this low ph 3608 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

6 Journal of Proteome Researh Figure 1. Intraprotein ross-link between K10 and K25 of ribosomal protein L25. (A) Orbitrap MS/MS spetrum of AANKFPAIIYGGK ross-linked to KEQGK. Peak assignments are shown in red if they orrespond to the R peptide, blue if they orrespond to the β peptide, and blak if they ould orrespond to either peptide hain. Peaks orresponding to fragmentation of the amidine bond of the ross-link, RþXL and βþxl, have been manually assigned. These fragmentation pathways, shown in Supporting Information Figure 1, have been disussed previously. 12 The β-nh 3 2þ ion does not ontain the ross-linker. Fragment ions have been assigned aording to the nomenlature proposed by Shilling and o-workers. 39 Peaks marked with asterisks orrespond to neutral losses of other assignments. (B) Strutural ontext of the linkage with respet to the 50S subunit (PDB 2AW4). The residues involved in the identified linkage are marked in red. enrihment with subsequent intermediate ph frationation. At intermediate phs, arboxyli aids beome deprotonated. Sine DEST ross-links, and any peptide speies for that matter, ontain differing numbers of suh groups, their net harges at intermediate ph will differ, thus enabling produtive frationation. While the sample enrihed for DEST interpeptide rosslinks was still adsorbed to the SCX olumn, the mobile phase ph was shifted and a salt gradient was implemented. Ten frations of the eluting DEST interpeptide ross-links were olleted. This sample preparation was ompleted for two biologial repliates, eah with two tehnial repliates. Beause we were interested in investigating the value of this sample handling proess, the effets of eah of the hromatographi steps were traed for one of the biologial repliates by preparing three additional types of samples, one in whih no SCX hromatography was ompleted, one in whih only SCX enrihment was ompleted, and one in whih only SCX frationation was ompleted. Mass Spetrometri Analysis of Cross-Linked Peptides Peptide mixtures produed by the differing sample preparations were analyzed via nanolc-ms/ms with an LTQ-Orbitrap high-resolution hybrid mass spetrometer using a strategy in whih both preursor and fragmentation mass spetra were aquired in the Orbitrap. These experiments produed aurate preursor and fragment ion masses that were searhed by the algorithm xquest for mathes to theoretial lysine to lysine interpeptide ross-links derived from the sequenes of the ribosomal proteins. Sore thresholds for ross-link spetra mathes were established through deoy database searhing, where the goal was to obtain a data set of identified linkages with an FDR <5%. We found that intraprotein ross-link mathes ould be aepted with lower xquest sores than interprotein ross-links mathes. This is easily explained by the fat that the size of the searh spae needed to identify intraprotein ross-links is muh smaller than the searh spae required to identify interprotein ross-links. Consequently, there is a lower likelihood of an intraprotein ross-link math being a false positive. For intraprotein ross-link mathes, use of an xquest sore threshold was suffiient for meeting our FDR goal. This was not the ase for interprotein ross-link mathes, whih were found to require manual validation to guarantee that they ontained fragmentation overage aross both peptides of the ross-links. An example of an intraprotein ross-link spetrum math onforming to these riteria is shown in Figure 1. This partiular math involves a ross-link between two peptides derived from ribosomal protein L25, AANKFPAIIYGGK and KEQGK dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

7 Journal of Proteome Researh Figure 2. Interprotein ross-link between K16 of ribosomal protein S19 and K46 of ribosomal protein S14. (A) Orbitrap MS/MS spetrum of KGPFIDLHLLKK ross-linked to WNAVLKLQTLPR. Peak assignments are shown in red if they orrespond to the R peptide and blue if they orrespond to the β peptide. A peak orresponding to fragmentation of the amidine bond of the ross-link, RþXL, has been manually assigned. This fragmentation pathway, shown in Supporting Information Figure 1, has been disussed previously. 12 Fragment ions have been assigned aording to the nomenlature proposed by Shilling and o-workers. 39 Peaks marked with asterisks orrespond to neutral losses of other assignments. (B) Strutural ontext of the linkage with respet to the 30S subunit (PDB 2AVY). The residues involved in the identified linkage are marked in red. Unlike the fragmentation spetrum of a single peptide, this spetrum ontains ions resulting from leavage of both peptide hains. To illustrate this, peak assignments are shown in red if they orrespond to fragmentation on the R peptide, and blue if they orrespond to the β peptide. Some fragment ions were observed that ould orrespond to either peptide hain, given that they share a C-terminal sequene of GK. This type of ambiguous ion assignment, suh as the peak labeled b 11R 3þ /b 3β 3þ, is shown in blak. As an be seen, the masses observed for preursor and fragment ions are in exellent agreement with theoretial masses. The majority (>90%) of the fragments ions assigned in this CID spetrum were in fat found to math theoretial masses to within 3 ppm. This impressive mass auray allowed xquest database searhing to be onduted with narrow mass toleranes, ensuring a small likelihood of obtaining false positives. Intraprotein ross-link mathes, suh as this, were identified at an FDR of 0.8% and, as noted earlier, did not require manual validation. In ontrast, interprotein ross-link mathes required additional onsiderations to obtain an equally low FDR value. Figure 2 illustrates an interprotein ross-link spetrum. The assignments made to the spetrum demonstrate that this is a ross-link between the S19 peptide KGPFIDLHLLKK and the S14 peptide WNA- VLKLQTLPR. Fragmentation of this preursor yielded a signifiant number of informative ions; a total of 16 unique fragment assignments ould be attributed to one speifi ross-link struture, notably one in whih the linkage between the peptides involved the 11th residue from the N-terminus of the S19 (R) peptide. It is important to note that this spetrum does not ontain any assignments eliminating the possibility of the linkage involving the C-terminal lysine residue of the S19 peptide. However, this residue is assumed to be unmodified, sine amidination is known to blok trypti leavage. 14,29 In this speifi spetrum, nine assignments ould be attributed to the R peptide sequene, and seven assignments ould be attributed to the β peptide sequene. The sequenes of both the R and β peptides were thus onfidently mathed. This was found to be a very strong indiator of a true positive. While analyzing deoy database searhes, we often observed false positive mathes in whih the assigned fragmentation on the peptides was limited or biased to only one of the peptide hains. We onsequently manually validated our interprotein ross-link mathes to ensure that at least three unique fragments were assigned to eah peptide hain and that there were a minimum of eight suh fragments assigned per ross-link struture. In doing this, we redued the FDR for interprotein ross-link mathes from approximately 10% to 0.8%. From our ross-linking analysis of the E. oli ribosome, we identified 325 ross-link spetra mathes that were assoiated with 71 different peptide linkages. Tables 1 and 2 summarize 3610 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

8 Journal of Proteome Researh Table 1. Intraprotein Cross-Link Mathes from the 30S Subunit and the 50S Subunit a 30S protein topology b distane xquest mathes 50S protein topology b distane xquest mathes S3 K44 K L1 K5 K13 4 S3 K48 K L1 K53 K166 4 S3 K78 K L1 K53 K204 2 S3 K107 K L3 K116 K159/K / S3 K107 K225 1 L4 K99 K S4 K82 K L4 K123 K S4 K150 K L4 K132 K S4 K155 K L9 K42 K S4 K155 K L10 K36 K104 1 S4 K166 K L14 K44 K S6 K35 K106 1 L14 K51 K S6 K56 K104 1 L14 K59 K S6 K56 K106 2 L15 K63 K S7 K148 K L16 K58 K S9 K99 K L17 K78 K S14 K18/K22 K / L18 K63 K S14 K18 K L19 K62 K S14 K22 K L20 K77 K S14 K75 K L22 K6 K S17 K29 K L22 K16 K S19 K16 K L24 K16 K S19 K20 K L24 K18/K20 K / S20 K33 K L25 K10 K S20 K68 K L25 K10 K total 118 L28 K61 K76 6 L29 Nterm/K2 K9 17.5/ L29 K4 K L33 K9 K L33 K9 K total 144 a Bold indiates that the linkage was observed in both biologial repliates. b Topologies shown with a slash separating residues indiate an ambiguous linkage. The distanes for either linkage are provided. Not appliable; one or both of the residues involved in the linkage are not present in the E. oli ribosome rystal struture. intraprotein and interprotein linkages, along with the number of spetra mathes supporting eah identifiation. There is learly a high degree of redundany as well as variability with whih linkages are represented in our data. For example, the linkage between K10 and K25 of ribosomal protein L25 was identified nearly 40 times more frequently than the linkage between K10 and K68 in the same protein. There are two primary fators that likely aount for this observation. Differing rates of ross-link formation among linkages is one. Cross-links are obviously more likely to form when residue reativities/solvent aessibilities, interresidue lengths, side hain orientations, and interresidue steri bulk are favorable for the reation to our. In the ase of the noted L25 intraprotein linkages, K10 and K25 are loser than K10 and K68. These onsiderations alone are not likely to fully explain the variability with whih linkages are represented in the data. A bias due to the methods used to identify the ross-links is ertainly another fator. Differenes in the ionization effiienies of the linked peptides and the fragmentation behavior of the ross-links, for example, should introdue an additional effet. The most signifiant onsequene of this variability and orresponding redundany is that the low FDR of 0.8% for spetra mathes inreases to 3% for linkage identifiations. Comparison of Cross-Linking and X-ray Crystallography Data Of the 71 identified linkages, 52 were intraprotein and 19 were interprotein. None of the latter, however, onneted the large and small ribosomal subunits. Crystal struture data indiated that only eight pairs of lysines were lose enough for DEST to bridge the 30S and 50S subunits, so this was relatively unsurprising. It is further noteworthy that only 20% of the identified linkages were interprotein, indiating a heavy bias toward the identifiation of intraprotein linkages. Nevertheless, our identifiations mirror the distribution of intra- and interprotein linkages deemed struturally possible in the rystal struture. Only about 20% of the lysine pairs bridging distanes less than 24 Å were interprotein, while 80% were intraprotein. This highlights the nature of the ribosome, a maromoleular omplex omposed of two-thirds ribonulei aid and one-third protein by mass. 30 Many ribosomal proteins are spatially separated due to the presene of RNA, limiting the number of interprotein rosslinks that an form. Distanes between the lysine residues in the observed linkages, as alulated from the rystal struture of the E. oli ribosome, are listed in Tables 1 and 2. In our results, there are 240 ross-link 3611 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

9 Journal of Proteome Researh Table 2. Interprotein Cross-Link Mathes a protein 1 protein 2 topology b (C R C R,Å) distane xquest mathes 30S S2 S8 K25/K27 K / S3 S14 K85 K S5 S8 K158 K S6 S18 K106 K29 15 S14 S19 K46 K S18 S21 K29 K S L2 L12 K67 K81 1 L6 L12 K28 K70 2 L6 L12 K85 K59 2 L6 L12 K85 K81 1 L9 L10 K57 K143 1 L9 L28 K42 K43 10 L9 L28 K42 K61 5 L9 L28 K57 K43 7 L11 L12 K99 K70 2 L16 L25 K127 K L17 L32 K121 K L18 L27 K17 K L22 L32 K28 K total 63 a Bold indiates that the linkage was observed in both biologial repliates. b Topologies shown with a slash separating residues indiate an ambiguous linkage. The distanes for either linkage are provided. Not appliable, one or both of the residues involved in the linkage are not present in the E. oli ribosome rystal struture. spetra mathes in whih linked residues are both unambiguously identified and present in the rystal struture. The strutural ontext of these data, and the 46 linkages they indiate, is displayed in Figure 3. In addition, Figure 4 displays the distribution of 240 spetral mathes as a funtion of the distane between the linked lysine residues. Note that all but 19 of the mathes orresponded to inter-residue distanes less than or equal to 24 Å, the maximum R-arbon to R-arbon ross-linking distane for DEST. Furthermore, all but two orresponded to interresidue distanes less than or equal to 25 Å, a distane that still demonstrates agreement given the antiipated error in the rystal struture. An estimate for the oordinate error is not provided in the PDB entry of the E. oli ribosome struture. Suh a value is, nevertheless, provided for the rystal struture of the T. thermophilus ribosome (PDB 2J00 and 2J01). 31 In this rystal struture of similar resolution, the estimated oordinate error is on the order of 1 Å. From our interpretation, then, 99% of the ross-link spetra mathes are onsistent with the rystal struture of the E. oli ribosome. DEST ross-linking thus provides reliable strutural information about maromoleular strutures. One of the two ross-link spetra mathes that orresponded to an interresidue distane greater than 25 Å indiates a linkage between ribosomal proteins S3 and S14. These two proteins reside at adjaent positions in the rystal struture. Still, the residues involved in the linkage are separated by 30.0 Å (Supporting Information Figure 2). Conformational flexibility or alternative onformations of these proteins in solution may aount for the observation of this linkage. The seond spetrum math that is inonsistent with the rystal struture defines an intraprotein linkage in ribosomal protein S9 between K99 and K114. These Figure 3. Strutural ontext of ross-linking data. Identified linkages involving residues present in the rystal struture of the E. oli ribosome (PDB 2AVY and 2AW4) 25 are shown. Ribosomal proteins are displayed in yan, intraprotein linkages in blue, and interprotein linkages in red. For larity, rrna are not shown. residues are separated by 36.6 Å (Supporting Information Figure 3). A disrepany as large as this is not likely to be aounted for by onformational flexibility of the protein. It is more likely that sample preparation or storage yielded a population of ribosomes with a strutural anomaly, suh as damaged RNA. Lysine 114 is part of an unstrutured domain anhored by RNA; damage to or leavage of this RNA, by endogenous RNase, for example, would release this protein domain, allowing K114 to ome within a linkable distane of K99. Alternatively, this linkage may simply be a false positive. The FDR determined for this analysis indiates that it is highly probable for our data to ontain at least one false positive. About one-fourth of the linkages identified in this study involved residues not present in the rystal struture of the E. oli ribosome. The majority of these linkages are onsistent with other strutural studies of prokaryoti ribosomes. For instane, although the loation of L28 is not properly defined in the E. oli rystal struture, three different linkages from our data demonstrate that it must make extensive ontats with ribosomal protein L9. The eletron density of L28 was properly refined in the T. thermophilus ribosome rystal struture, 31 and indeed from this struture, it an be seen that L28 diretly binds to L9. Nevertheless, a ouple of these linkages ould not be readily explained given what is urrently known about ribosome struture. These results are the subjet of further disussion below. Cross-Linking of the Stalk Complex Cross-linking of the proteins that onstitute the stalk omplex was partiularly intriguing, sine this prominent feature of the ribosome has not yet been fully refined in any rystal struture. The stalk omplex, formed by one opy of L10 and four opies of L12, has, nonetheless, been struturally eluidated in a hybrid model onstruted from the ombined use of 3612 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

10 Journal of Proteome Researh Figure 4. xquest ross-link mathes as a funtion of distane between lysine residues in the identified linkages. eletron mirosopy and X-ray rystallography. 32 This model demonstrates that the base of the omplex is situated adjaent to ribosomal proteins L6 and L11; our results apture this strutural feature of the ribosome, as we identified four different linkages between these two proteins and ribosomal protein L12. Two linkages involving the proteins of the stalk omplex (L10 and L12) that annot be explained by the ryoem/rystal struture model mentioned above were identified. These linkages orrespond to K67 of L2 ross-linked to K81 of L12 and K57 of L9 ross-linked to K143 of L10. Interestingly, L2 and L9 reside on the opposite side of the ribosome from the stalk omplex. Figure 5A displays the rystal struture of the E. oli ribosome and demonstrates the position of the stalk omplex with respet to the proteins to whih it was found ross-linked. In Figure 5B, the struture has been rotated 90 in either diretion, making the residues involved in these linkages visible. Sine we estimate the distane aross the ribosome to be at least 50 Å greater than the span of any of the proteins 25,32,33 and the ross-linker, these linkages are believed to orrespond to interations between ribosomes. A model for these interations is shown in Figure 6, where the stalk is depited as a large red irle. Our model shows that the stalk protein omplex may, in addition to other funtions, serve as a hub point between ribosomes and interat with ribosomal proteins L9 and L2 on an adjaent ribosome. Beause the reation onditions used here for ross-linking ribosomes have not been found to yield nonspeifi ross-linking of monomeri proteins, 12 we believe these ribosome ribosome interations are of funtional signifiane. Cross-linking between the stalk omplex and the L2 region on the opposite fae of the E. oli ribosome has been deteted before, albeit via different methods. Dey and o-workers previously inorporated a photoreative group into the struture of L12 as a means to ondut site-direted ross-linking and to identify the proteins with whih the stalk omplex interats. 34 Their interpretation for L12 being ross-linked to L2, made at a time when high-resolution strutures did not exist, was that an elongated struture of L12 might extend aross the body of the 50S subunit and that this interats with L2. In light of more reent work modeling the ribosome and the stalk omplex, 32 this seems unlikely. Even a fully extended L12 protein ould not span the distane between the base of the stalk omplex and ribosomal protein L2. Their results are onsistent with our suggestion that ribosome ribosome interations are ourring. A reent analysis of E. oli polysomes, lusters of ribosomes on an mrna, by ryoeletron tomography is onsistent with our ribosome ribosome interation hypothesis. 35 In this study, adjaent ribosomes were found to adopt preferential onfigurations, in whih 30S subunits aligned with orientations that differed by 180. Although the auray of doking the ribosome rystal struture 25 into their averaged 3D density of adjaent ribosomes was limited, it was suggested that in the most ommon onfiguration ontat between ribosomes extends from the L1 arm of one ribosome to a region near protein S4 in an adjaent ribosome. Furthermore, the doking showed that in this onfiguration ribosomal protein L9 extends out from one ribosome toward the large subunit of another. Our identifiation of a linkage between ribosomal protein L9 and the stalk omplex is onsistent with this preferred onfiguration of ribosome ribosome interation. However, it is unertain whether the ribosome ribosome interations identified in our analysis orrespond to polysomes or free ribosomes. Our preparation 3613 dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,

11 Journal of Proteome Researh Figure 5. Cross-linking of the stalk omplex. (A) The rystal struture of the E. oli ribosome (PDB 2AVY and 2AW4).25 The stalk omplex, positioned at its anonial binding site, is represented by a large red irle. The proteins found to ross-link to the stalk omplex, L2, L6, L9, and L11, are labeled. (B) Rotations of the E. oli ribosome. The L6/L11 (top) and L2/L9 faes (bottom) of the ribosome are shown. Residues involved in the linkages with the stalk omplex are highlighted in red. Figure 6. Model of ribosome ribosome interations based on DEST ross-linking data. of ribosomes, while expeted to ontain primarily free ribosomes, may ontain some polysomes. Further study involving purifiation of free ribosomes and polysomes is needed to distinguish the origin of our identifiations. Ribosome ribosome interations may be of funtional signifiane. Previous studies have indiated that L9, one of the proteins mentioned above, plays a role in enhaning the fidelity of translation by minimizing ribosome slippage.36,37 Among the hypotheses for how L9 is able to do so, it has been suggested that L9 may funtion as a strut38 to ouple ribosomes within a polysome.36 This oupling ould prevent the downstream ribosome from sliding relative to the mrna. Our results lend redene to this hypothesis by demonstrating the involvement of L9 in ribosome ribosome interations. Further study is needed to better haraterize ribosome ribosome interations and their funtional signifiane. Amenability of DEST Cross-Links to SCX Enrihment and Frationation It was of interest to assess the extent to whih SCX hromatography improved the analysis of DEST ross-links. The number of spetra mathes identified per LC-MS/MS experiment has been ataloged as a funtion of sample preparation for one biologial repliate in Figure 7. Without prior sample preparation, an average of only 5.5 ross-link spetra mathes was obtained per LC-MS/MS experiment. After SCX enrihment, however, the average number of ross-link spetra mathes per analysis was inreased 2.5-fold to This an be attributed to the fat that the enrihment physially altered the omposition of the digest, making it more abundant in DEST interpeptide rosslinks and less abundant in linear peptide speies. To support that this was the ase, the above samples were analyzed by LC-MS/ MS experiments in whih, unlike before, there was no harge rejetion to bias detetion against linear peptides, and the resulting data were searhed for both interpeptide ross-links and linear peptide speies. As antiipated, the ratio of interpeptide ross-link spetra mathes to linear peptide spetra mathes was found to inrease from to 0.34, a 20-fold hange, due to the enrihment. It is worth noting that these ratios of spetra mathes are not expeted to aurately translate to the perent omposition of the sample, beause ross-linked peptides are signifiantly more hallenging to identify with onfidene than linear peptides dx.doi.org/ /pr200260n J. Proteome Res. 2011, 10,