Mass spectrometry of full-length integral membrane proteins to define functionally relevant structural features

Methods 46 (2008) 54–61

Contents lists available at ScienceDirect

Methods j o u r n a l h o m e p a g e : w w w . e l s e v i e r. c o m / l o c a t e / y m e t h

Mass spectrometry of full-length integral membrane proteins to define functionally relevant structural features Guillaume Gabant, Martine Cadene * Cen­tre de Bi­o­phy­sique Mol­écu­laire, CNRS UPR4301, Rue Charles Sa­dron, 45071 Orleans ce­dex 2, France

a r t i c l e

i n f o

Article history: Accepted 20 October 2008 Available online 29 October 2008  Key­words: Mass spec­trom­e­try Inte­gral mem­brane pro­tein Struc­ture–func­tion rela­tion­ship Lim­ited pro­te­ol­y­sis Pro­to­col

a b s t r a c t The crys­tal­li­za­tion and struc­ture deter­mi­na­tion of inte­gral mem­brane pro­teins remains a dif ­fi­cult task rely­ing on a good under­stand­ing of the behav­ior of the pro­tein for suc­cess. To date, mem­brane pro­tein struc­tures are still far out­num­bered by sol­ub ­ le pro­tein struc­tures. Mass spec­trom­e­try is a pow­er­ful and ver­sa­tile tool offer­ing deep insights into the state of the inte­gral mem­brane pro­tein the struc­tur­al­ist intends to crys­tal­lize. With appro­pri­ate sam­ple prep­a­ra­tion meth­ods, it pro­vides infor­ma­tion that can some­times prove crit­i­cal at var­i­ous stages of the struc­ture deter­mi­na­tion pro­cess, from pro­tein expres­ sion to model build­ing. More­over, valu­able knowl­edge is gained when the iden­ti­fied struc­tural fea­tures under­lie impor­tant func­tional aspects. Elec­tro­spray and matrix assisted laser desorp­tion ion­i­za­tion (MALDI) meth­ods, how­ever, face a par­tic­u­lar chal­lenge when deal­ing with inte­gral mem­brane pro­teins. A MALDI method spe­cif­i­cally opti­mized for mem­brane pro­tein anal­y­sis is pre­sented here, with detailed infor­ma­tion on the sam­ple prep­a­ra­tion and depo­si­tion, as well as guide­lines for domain deter­mi­na­tion by lim­ited pro­te­ol­y­sis. MALDI-time of flight mass spec­trom­e­try can be used to do a proper inven­tory of ini­ti­a­tion sites, to tai­lor a pro­tein to a sta­ble, well-folded form, and to eval­u­ate sele­no­me­thi­o­nine replace­ ment. These approaches are illus­trated with a few exam­ples drawn from the struc­tural biol­ogy of ion chan­nels. © 2008 Else­vier Inc. All rights reserved.

1. Intro­duc­tion When Rod­er­ick MacK­in­non and his team under­took the crys­tal­ li­za­tion and 3D-struc­ture deter­mi­na­tion of the potas­sium chan­nel circa 1996, there were coor­di­nates for only six inte­gral mem­brane pro­teins in the pro­tein data bank (PDB). With so few prece­dents for suc­cess, the task was con­sid­ered daunt­ing at the time. The work led to the crys­tal struc­ture of KcsA [1], a world-pre­miere for ion chan­nels in gen­eral and a land­mark for the under­stand­ing of the func­tion mech­a­nism of potas­sium chan­nels in par­tic­u­lar. To date, the struc­tures of about 90 inte­gral mem­brane pro­teins have been deter­mined. While these num­bers show that con­fid ­ ence in achiev­ ing this type of result is grow­ing, it is still a rel­a­tively small num­ber in com­par­i­son to sol­u­ble pro­tein struc­tures. Yet, mem­brane pro­ teins play a very impor­tant role in cells and account for about 50% of ther­ap ­ eu­tic tar­gets. Struc­tural infor­ma­tion on inte­gral mem­ brane pro­teins is thus as desir­able as ever. Here, we describe some tools that may be used in mass spec­trom­e­try (MS) in com­bi­na­tion with bio­chem­i­cal manip­u­la­tions to assist in get­ting homog­e­nous, sta­ble pro­tein prep­a­ra­tions for crys­tal­li­za­tion. These tools have also been used to derive preliminary infor­ma­tion about the pro­tein * Cor­re­spond­ing author. Fax: +33 238 63 15 17. E-mail address: cad­[email protected] (M. Cadene). 1046-2023/$ - see front matter © 2008 Else­vier Inc. All rights reserved. doi:10.1016/j.ymeth.2008.10.021

struc­ture, infor­ma­tion which has some­times proved to be precious at var­i­ous stages of the struc­ture deter­mi­na­tion pro­cess. Meth­ods for the mass spec­tro­met­ric anal­y­sis of mem­brane pro­ teins have evolved from requir­ing high amounts of mate­rial [2–5] to more rea­son­ably scaled approaches in the past few years [6,7]. Upon puri­fi­ca­tion, the mem­brane pro­tein finds itself embed­ded in a lipid-deter­gent shell arranged in a toroid shape. The basic issue for MS anal­y­sis is to smoothly trans­fer the pro­tein from this mem­ brane sur­ro­gate toroid to the sol­vents used in the anal­y­sis, with­out pro­vok­ing pre­cip­i­ta­tion or aggre­ga­tion. Meth­ods based on elec­tro­ spray ion­i­za­tion gen­er­ally still require a puri­fic­ a­tion step to rid the solu­tion of deter­gents, which are largely incom­pat­i­ble with this type of ion­i­za­tion (for an exam­ple, see [8]). MALDI ion­i­za­tion is ame­na­ble to direct anal­y­sis with­out prior removal of the deter­ gent, pro­vided the lat­ter has been selected to be com­pat­i­ble with the ion­i­za­tion and sam­ple depo­si­tion method. Ionic deter­gents are still largely incom­pat­i­ble with MALDI ion­i­za­tion, in which case one should con­sider buffer-exchang­ing the deter­gent for a zwit­ter­ionic or non-ionic one which will be com­pat­i­ble with MS. With the lat­ ter two types of deter­gents, the ultra­thin layer method is a good tool to oper­ate this smooth trans­fer of the pro­tein from the buffer solu­tion to the matrix–ana­lyte mix­ture solu­tion, then on to sam­ ple spot­ting and solid-state co-crys­tal arrange­ment on the sam­ ple plate. Fig. 1A shows the sig­nals obtained in the anal­y­sis of a



G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

55

A 25

Intensity (a.u.)

dried droplet

2+ 3+

1+

ultrathin layer

4+ 0

10000

20000

B

m/z

30000

40000

C

Fig. 1. (A) MALDI-TOF mass spec­tra of bac­te­rial Si­lic­ib­act­er pome­royi sodium chan­nel, 300 fmol/lL in 4HCCA/FWI (6 pmol/lL in buffer with 10 mM DM), see text for exper­i­ men­tal details. a.u., arbi­trary units. Spots were obtained using the dried drop­let (upper spec­trum) or the ultra­thin layer (lower spec­trum) meth­ods. The num­ber of charges is indi­cated on top of the peak for each charge state observed for the pro­tein. The upper spec­trum is off­set for dis­play pur­poses. (B) MALDI spots on gold-plated sam­ple plate, spot­ted by hand (from left to right): dried drop­let, ultra­thin layer. (C) Edge of the ultra­thin layer on the MALDI sam­ple stage, with the plate angled towards light.

sodium chan­nel using the dried drop­let [9] com­pared to the ultra­ thin layer method [6], while all other con­di­tions were kept iden­ti­ cal. The pho­to­graphs below (Fig. 1B and 1C) show the aspect of the spots and the ultra­thin layer itself. In the authors’ as well as oth­ers’ hands, sand­wich or reg­u­lar thin layer meth­ods [9,10] do not afford sim­i­lar jumps in sen­si­tiv­ity and res­o­lu­tion. Although the set of meth­ods pre­sented here has been applied to a wide range of mem­brane pro­tein types (b-bar­rel trans­port­ers, a-helix chan­nels and trans­port­ers [11,12], recep­tors such as GPCRs [6]), exam­ples given herein will focus on ion chan­nels. 2. Ultra­thin layer MALDI spot prep­a­ra­tion method The pro­to­col for the ultra­thin layer method was first pub­lished in 2000 [6], and was the object of a vid­eo­pub­li­ca­tion in 2007 [13]. It is derived from clas­si­cal thin layer approaches [9,10]. This opti­ mized var­ia ­ ­tion was found to be par­tic­u­larly effi­cient and robust for the anal­y­sis of inte­gral mem­brane pro­tein mix­tures in the pres­ ence of con­cen­tra­tions of deter­gents usu­ally found in sta­bil­iza­tion buf­fers, as well as rea­son­ably high con­cen­tra­tions of addi­tives used in the bio­chem­i­cal prep­ar­ a­tion of pro­teins, such as imid­az­ole, NaCl, even urea. Table 1 shows a list of com­mon buffer com­po­ nents as well as their use­able con­cen­tra­tions in buffer, pro­vided that the 1/20 dilu­tion in matrix solu­tion is respected. Inter­est­ingly, the method has showed great ver­sa­til­ity in that it can be used also for sol­u­ble pro­tein mix­tures in the ‘FWI’ sol­ vent described below. Sol­u­ble pro­teins how­ever do not fare well if one com­pletely removes water from this sol­vent (Ste­ven L. Cohen, per­sonal com­mu­ni­ca­tion). The abil­ity to ana­lyze sol­u­ble pro­teins con­com­i­tantly with mem­brane pro­teins is an impor­tant fea­ture for the detec­tion of sol­u­ble pro­tein con­tam­i­nants, sol­u­ble cleav­

age frag­ments, or sol­u­ble co-expressed domains. One of the main strengths of the method is that it lim­its the inten­sity bias in favor of smaller molec­u­lar ions observed in other meth­ods, so that a whole pro­tein can be ana­lyzed con­com­i­tantly with cleav­age prod­ucts of any size above 700 Da, whether they are hydro­pho­bic or not. This is an impor­tant fea­ture for appli­ca­tions such as lim­ited pro­te­ol­y­sis exper­i­ments. Finally, the rel­a­tive homo­ge­ne­ity of the sig­nal over the whole sam­ple spot greatly facil­i­tates data acqui­si­tion. The gen­eral pro­to­col flow is dis­played in Fig. 2. The appli­ca­tion of a very thin matrix layer all-over the sam­ple plate is usu­ally mas­ tered by oper­a­tors within 1–3 tri­als. It should be stressed that care and tim­ing are an impor­tant part of suc­cess in this pro­to­col. Con­se­ quently, the fol­low­ing steps should not be rushed through. Notes: Micro­tubes should never be sil­i­con­ized or made of col­ ored plas­tic. Unless otherwise stated, all sol­vents are high purity HPLC-grade, ‘ultra­pure’ water is prepared by a reverse osmo­sis/ ultra­fil­tra­tion sys­tem (resis­tiv­ity over 18 MX). Meth­a­noic acid 88% (v/v) is referred to as sim­ply “for­mic acid” in the text. The 4HCCA1 matrix is re­crys­tal­lized grade (for exam­ple from Applied Bio­sys­tems, or Bru­ker Dal­ton­ics) or reagent grade (Sigma–Aldrich) puri­fied by acid–base pre­cip­i­ta­tion in the lab­o­ra­tory, and is resus­ pended in ace­to­ni­trile for ali­quot­ing. Ali­quots are prepared by with­drawal from the stock with con­stant resus­pen­sion (using a cut micro­pi­pette tip), dried in a Speed-Vac and stored with caps firmly in place at room tem­per­a­ture in the dark. One gram of matrix can thus be ali­quoted in »100 tubes. Ali­quots are sta­ble for a few weeks at room tem­per­a­ture and at least one year at ¡80 °C. If ­fro­zen for 1 Abbre­vi­a­tions used: 4HCCA, 4-hydroxy cyano cin­namic acid (alpha-cyano 4-hydroxy cin­namic acid); CMC, crit­i­cal micel­lar con­cen­tra­tion; MALDI, matrixassisted laser desorp­tion ion­i­za­tion; TFA, tri­flu­o­ro­ace­tic acid; TOF: time-of-flight.

56

G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

Table 1 Tol­er­ances for com­mon mem­brane pro­tein buffer com­po­nents in ultra­thin layer MALDI-TOF anal­y­sis Name of com­po­nent

Max­i­mum final con­cen­tra­tionb Max­i­mum final con­cen­tra­tion in buffer, in matrix-ana­lyte mix­ture solu­tion, i.e. after i.e. before 1/20 dilu­tion into the sat­u­rated 1/20 dilu­tion into the sat­ur­ ated matrix matrix solu­tion solu­tion

Note

Imid­az­ole

50 mM

1M

Urea

80 mM

1.6 M

NaCl, KCl

15 mM

300 mM

Used in the puri­fi­ca­tion of over­ex­pres­sed His-tagged pro­teins Cha­o­tropic agent, used for dena­tur­ation/ rena­tur­ation Na+ and K+ adducts will be observed in MS when NaCl con­cen­tra­tion is above 50 mM in buffer before dilu­tion into matrix solu­tion

Tris-HCl or sim­i­lar organic buffer Glyc­erol, DMSO

5 mM 0.25 %

100 mM 5%

Phos­phate buffer, PBS

should be avoided alto­gether

Non-ionic deter­gents n-Dode­cyl-ß-D-mal­to­side DDM n-De­cyl-ß-D-mal­to­side DM n-Octyl-ß-D-gluco­py­ran­o­side ßOG

0.5 mM 0.5 mM 1.5 mM

10 mM 10 mM 30 mM

CMCa 0.17 mM 1.8 mM 19 mM

Zwit­ter­ionic deter­gents Lau­ryl­dim­eth­yl­amine oxide LDAO Zwit­ter­gent 3-12 N-dod­ecylphosph­och­o­line

0.5 mM 0.25 mM 0.25 mM

10 mM 5 mM 5 mM

CMCa 1.5 mM 2-4 mM 1.5mM

Ionic deter­gents

should be avoided alto­gether

Non-vol­a­tile com­pounds inter­fere with matrix crys­tal­li­za­tion Phos­phates inter­fere with ion­i­za­tion. Buffer exchange is rec­om­mended

Table expanded from Cad­ène and Chait [6]. a Crit­ic­ al micel­lar con­cen­tra­tion, CMC, val­ues pro­vided by Ana­trace, Inc. and Cal­bio­chem, Inc. b Con­cen­tra­tions exceed­ing these val­ues will pre­vent proper for­ma­tion of matrix-sam­ple co­crys­tals. If higher deter­gent con­cen­tra­tions are required (e.g. for bOG), the sam­ple dilu­tion step into matrix solu­tion should be raised pro­por­tion­ally.

stor­age, ali­quots have to be warmed up to room ­tem­per­a­ture (RT) before use to pre­vent hygro­scopic col­lec­tion of water onto the matrix. 2. A. Prep­a­ra­tion of sat­ur­ ated matrix solu­tion for sam­ple dilu­tion This solu­tion is used to dilute sam­ples imme­di­ately before spot­ ting onto the sam­ple plate. While this is only used after mak­ing the thin layer, the sol­vent used to make the sat­u­rated matrix solu­tion has to be prepared before­hand. A1: Prep­a­ra­tion of matrix sol­vent: The FWI sol­vent mix­ture used for mem­brane pro­teins (con­sist­ing of 3:1:2 for­mic acid:water:iso­ pro­pyl alco­hol (v/v)) has to be prepared at least 4 h in advance. The reason for this is unknown; how­ever, using freshly prepared FWI is the most com­mon cause for fail­ure. Once prepared, the solu­tion can be stored for weeks in a clean scin­til­la­tion vial, washed of all Tri­ton used in the man­u­fac­tur­ing pro­cess. Alu­min­ium-lined lids should

dilute sample protein 1/20 in a solution of 4HCCA saturated in FWI

sample plate

spot a 0.5 L aliquot wash

(0.1%TFA)

m/z

apply a very thin layer of 4HCCA matrix

MALDI-TOF MS

Fig. 2. Ultra­thin layer method depo­si­tion scheme for MALDI anal­y­sis.

not be used to avoid sol­vent con­tam­i­na­tion by cop­per. Because of the high for­mic acid con­tent, it is advis­able to store the FWI stock in a ven­ti­lated fume hood, with a tightly closed plas­tic lid. A2: Prep­a­ra­tion of sat­u­rated matrix solu­tion for sam­ple dilu­ tion: to a dried matrix ali­quot, »100–150 lL of FWI sol­vent is added. The solu­tion is pipet­ted up and down 10–15 times and vor­ texed at max­i­mum speed for 50–60 s. This ensures thor­ough mix­ ing, which is crit­i­cal. If the solu­tion is not clearly sat­u­rated, i.e. if all or nearly all the 4HCCA matrix has been sol­u­bi­lized in the sol­ vent, the solu­tion is pipet­ted back up and added to another ali­quot of dried matrix. The sus­pen­sion is mixed thor­oughly as above. A cloud of sus­pended matrix should be clearly vis­i­ble upon shak­ing the tube. A3: The sus­pen­sion prepared in A2 is cen­tri­fuged in a table-top micro­cen­tri­fuge for at least 6 min at 13,500 rpm (»14000g for a typ­i­cal 7 cm mean radius rotor). Care­ful cen­tri­fu­ga­tion ensures the removal of matrix micro-aggre­gates which read­ily pre­cip­i­ tate mem­brane pro­teins, lead­ing to a dra­matic loss in sen­si­tiv­ ity. Speed-start table mi­cro­fug­es are not appro­pri­ate for this step because the cen­trif­u­gal force is insuf ­fi­cient to pellet matrix microaggre­gates. After cen­tri­fu­ga­tion, tak­ing care not to dis­turb the matrix pellet, one gently pipettes up the super­na­tant and trans­fers it to a fresh tube. The tube is labeled with the matrix and sol­vent names (4HCCA/FWI). This solu­tion can only be used on the day it was prepared. 2. B. Prep­a­ra­tion of diluted matrix solu­tion for the plate B1: First, a sat­u­rated matrix solu­tion is prepared using a dried matrix ali­quot as described in step A, using »150 lL of TWA sol­ vent (TWA: 1:1 water:ace­to­ni­trile (v/v) with 0.1% final TFA). The sat­u­rated matrix solu­tion is trans­ferred to a fresh tube labeled 4HCCA/TWA. This solu­tion should be used in step B2 on the day it was prepared.



G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

B2. The sat­u­rated 4HCCA/TWA solu­tion is diluted 1/4 using iso­ pro­pyl alco­hol. This pro­vides the diluted matrix solu­tion for the plate. The sam­ple plate solu­tion is sta­ble for months in the dark at room tem­per­a­ture in a prop­erly sealed tube. 2. C. Sam­ple plate prep­a­ra­tion C1. It is rec­om­mended to pre­pare a clean work­ing sur­face on the bench for the sam­ple plate. The sam­ple plate can be a gold-plated or a stain­less steel plate. A gold sur­face sim­ply makes visu­al­i­za­ tion of ultra­thin matrix lay­ers eas­ier. Plates with recessed wells should not be used. Plates with lightly engraved spot rings or no engrav­ing at all are best suited for this method. This is the only step that should be under­taken under a fume hood, pref­er­a­bly wearing gloves. Wash steps are as fol­low: reagent grade meth­a­nol (MeOH) is squirted onto the top of the plate from a squeeze bot­tle, and wiped with a lint-free tis­sue wipe. Then the plate is washed by alter­nat­ ing deion­ized water with reagent grade MeOH, always wip­ing the sol­vent with a lint-free wipe and end­ing with a MeOH step until the plate is spot­less. Ace­tone can be used to dry cracks and holes on the back of the plate, but should never be used on the work­ing sur­face of a plate, as the strip­ping power of ace­tone cre­ates a slip­ pery sur­face which impairs the depo­si­tion of matrix solu­tions. C2: The plate is let to warm back to room tem­per­a­ture on the bench (sol­vent evap­o­ra­tion in step C1 causes the plate to cool down). If the plate has been warmed to speed up dry­ing, it is very impor­tant to let it cool back to RT on the bench before pro­ceed­ing to the next steps. In 5–10 min, the plate is both com­pletely dry and at RT. 2. D. Apply­ing the ultra­thin layer of matrix onto the plate D1: The work­ing area is prepared, with lint-free wipes and a pro­tein-gel pipette tip at hand. Reg­u­lar pipette tips can be used, pro­vided they are big enough and notch-free. As described below, a vol­ume of 20 lL of the sam­ple plate solu­ tion prepared in Sec­tion B will be suf ­fi­cient to cover roughly 20 cm2 of plate sur­face. If using stan­dard 384 spot plates (usable sur­face 10 £ 7 cm), the plate can be cov­ered in four quar­ters. Smaller plates (such as the Applied Bio­sys­tems 96-spots plate, usable sur­ face 4.5 £ 4.5 cm) can be cov­ered in one go. The outer 1 cm from the edge of the plate is left free of matrix to min­i­mize trans­fer of matrix to the instru­ment plate holder. This also guards from depos­it­ing spots in areas where edge effects can alter the qual­ity of mass cal­i­bra­tion. D2: About »20 lL of diluted matrix solu­tion is applied to the plate. The solu­tion is imme­di­ately spread onto to the sur­face of the plate using 2–3 large strokes. It is impor­tant not to “paint” the plate as this will result in non-homog­e­nous depos­its. The matrix solu­tion starts dry­ing imme­di­ately due to the high iso­pro­pyl alco­ hol con­tent. The “wip­ing tool” is prepared by curl­ing a wipe around an index fin­ger. A max­i­mum of one or two rounds is pref­er­a­ble for good fin­ger­tip sen­si­tiv­ity to sur­faces. The free wipe end can be trapped with the thumb. D3: When the matrix solu­tion on the plate is almost dry, the small amount of remain­ing mois­ture is gath­ered with the wipe around a fin­ger­tip and tapped over the whole matrix sur­face to redis­trib­ute mois­ture. Then the matrix is wiped off the plate using the wipe on the fin­ger, apply­ing fairly high pressure and going in 3–4 long strokes. The matrix layer should be wiped until the amount of matrix left on the plate is barely vis­i­ble as a yel­low­ish reflec­tion when the plate is angled towards the light. On stain­less steel, only the outer edge of the matrix layer is vis­i­ble. The goal is to pro­duce a homog­en ­ ous, ultra-thin layer of matrix, as this will be the seed­ing bed for matrix–ana­lyte co-crys­tals for­ma­tion.

57

The qual­ity of the ultra­thin layer can imme­di­ately be tested using a sat­u­rated solu­tion of matrix (for exam­ple, using the 4HCCA/FWI solu­tion prepared in step A). A whit­ish, uniform matrix spot should start form­ing within 15–30 s (Fig. 1B). In extreme atmo­spheric con­di­tions of tem­per­a­ture and/or humid­ity, the crys­tal­li­za­tion on the plate could take a little longer. As soon as the matrix crys­tal layer is uniform, the excess sol­vent drop­let is aspi­rated using a vac­uum line with a pipette tip on the end. The ultra­thin layer on the plate can be used for months, as long as sam­ ples are spot­ted using the same matrix (e.g., 4HCCA matrix–pro­ tein solu­tion onto 4HCCA layer). Sin­a­pic acid can also be used to make an ultra­thin layer, and sin­a­pic acid–pro­tein solu­tions spot­ted onto the layer using the same pro­to­col (higher sin­a­pic acid con­cen­ tra­tions are required to reach sat­u­ra­tion). Sin­a­pic acid matrix how­ ever may pro­duce matrix–pro­tein adducts, whereas 4HCCA matrix does not when using the ultra­thin method. 2. E. Sam­ple prep­a­ra­tion and depo­si­tion E1: Sam­ple prep­a­ra­tion. The method is rel­a­tively robust and tol­er­ant to con­tam­i­nants within the lim­its indi­cated in Table 1. How­ever, it is not com­pat­i­ble with ionic deter­gents, or very large con­cen­tra­tions of any given reagent. Buffer exchange (through gel per­me­ation–exclu­sion chro­ma­tog­ra­phy) should be con­sid­ered if con­tam­i­nant con­cen­tra­tions exceed the tol­er­ances of the method. The anal­y­sis of mem­brane pro­teins is achieved by keep­ing the pro­tein/con­tam­i­nants ratio to an opti­mum. On one hand, the deter­ gent con­cen­tra­tion in the sam­ple should be at least 1.5–2 times the CMC for pro­tein sta­bil­ity, and there is a practical limit to the pro­ tein con­cen­tra­tion. This sets an upper limit to the pro­tein/deter­ gent ratio in the sam­ple up until the anal­y­sis. On the other hand, the pro­tein con­cen­tra­tion after dilu­tion in the matrix solu­tion has to be high enough to achieve good sig­nal and the deter­gent con­ cen­tra­tion should be low enough to pre­vent inter­fer­ence. Since the pro­tein and con­tam­i­nants are diluted at the same time into the matrix solu­tion, there is an opti­mum fac­tor of dilu­tion to achieve good sen­si­tiv­ity with low con­tam­i­nant inter­fer­ence. In other words, the dilu­tion fac­tor has to be a com­pro­mise between low con­tam­ i­nant (one has to dilute enough) and high pro­tein con­cen­tra­tions (one can­not dilute too much). It is thus really impor­tant to dilute the pro­tein sam­ple 1/20 in sat­u­rated 4HCCA/FWI matrix solu­tion (as prepared in A). If high deter­gent con­cen­tra­tions are needed to reach the CMC, and buffer exchange is not fea­si­ble, then the pro­ tein sam­ple should be fur­ther diluted into the matrix solu­tion. The final pro­tein con­cen­tra­tion should remain above 300 fmol/lL for each pro­tein spe­cies in a mix­ture, mean­ing the stock pro­tein solu­ tion should have at least » 6 pmol/lL of each pro­tein spe­cies, or higher if the dilu­tion fac­tor needs to be increased. Typ­i­cally, 9.5 lL of 4HCCA/FWI matrix solu­tion is added to 0.5 mL micro­tubes (trans­par­ent, sil­i­cone-free) and the tube lids imme­di­ately closed to pre­vent evap­or­ a­tion. Then 0.5 lL of the pro­tein solu­tion to ana­ lyze is added to the 9.5 lL of matrix solu­tion, pipette-mix­ing 15–20 times into the solu­tion and the mix­ture is care­fully vor­texed at a low enough speed to pre­vent dis­per­sion of the mix­ture onto the micro­tube walls and lid. E2: Sam­ple depo­si­tion. Once the pro­tein–matrix mix­ture solu­tion is prepared, it should be depos­ited onto the sam­ple plate within ten min­utes. In our expe­ri­ence, longer expo­sure of the pro­tein to the for­mic acid in the 4HCCA/FWI mix­ture will result in ser­ine and thre­o­nine formy­la­tion. If the for­mic acid stock is low grade and con­tains per­ox­ides, oxi­da­tion may also be observed. This is usu­ally not a prob­lem with typ­i­cal, reagent grade for­mic acid. To spot the pro­tein sam­ple onto the plate, 0.5 lL of the pro­ tein–matrix mix­ture prepared above is depos­ited on the plate, tak­ ing care not to touch the plate with the pipette tip so as not to scrape the thin layer off the plate. The spot starts form­ing from the

58

G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

outside of the drop­let and reaches homo­ge­ne­ity within 15–30 s, and appears as whit­ish and opaque. In some cases, par­tic­u­larly if the deter­gent is highly con­cen­trated, a small void in the cen­ter of the spot can remain after the rest of the spot has crys­tal­lized prop­ erly. This does not pre­clude anal­y­sis. As soon as the spot has formed to homo­ge­ne­ity, the excess liquid is aspi­rated with a ­vac­uum line and the spot is allowed to air-dry. E3: Wash­ing. The sam­ples can be washed using room tem­per­a­ ture 0.1% aque­ous TFA (note: for hydro­philic pro­teins and pep­tides, washes should be avoided or per­formed with ice-cold 0.1% TFA). A 1–2 lL drop­let of 0.1% TFA is applied on top of the spot. After a few sec­onds (up to 15 s), the wash is aspi­rated with a vac­uum line. The spot is allowed to air-dry: it is now ready for mass spec­trom­e­try anal­y­sis. If the sam­ple spot was prepared with FWI, formy­la­tion may start to occur if the spot is not ana­lyzed within the hour. Note: if on first attempt, the spot does not pro­duce use­able sig­ nal in MALDI-TOF MS, a higher dilu­tion step in 4HCCA/FWI solu­ tion (e.g., 1/40 or 1/60) should be attempted. This has a far bet­ter rate of suc­cess than try­ing a lower dilu­tion fac­tor. Typ­ic­ ally, the con­tam­i­nants pre­clude ion­i­za­tion and thus greater dilu­tion will work bet­ter. If a highly con­cen­trated pro­tein can be prepared, a 1/100 dilu­tion can be employed with­out prob­lem. E4: Instru­men­tal setup. The ultra­thin layer method is use­ful for anal­y­sis of full-length pro­teins as well as mix­tures of pro­tein and cleav­age prod­ucts, with length vary­ing from a few amino acids to nearly full-length pro­tein. Ion­iz ­ a­tion can be per­formed with any instru­ment equipped with a MALDI source, although the most com­mon appli­ca­tion is with MALDI-TOF instru­ments. It per­forms well with MALDI-TOF-TOF, MALDI-ion trap and MALDI-Qq-TOF con­fig­u­ra­tions, pro­vided that the laser power is set to about 10–15% above the ion­i­za­tion thresh­old. The ultra­thin layer requires less laser power than the dried drop­let method. Since a given position on a sam­ple spot will be dug through within about 50 shots, it is

A

r­ ec­om­mended to stay with laser shot fre­quen­cies below 20 Hz, and to move the laser tar­get position every 25 shots. For whole pro­ teins, the best results are obtained with less than 5 Hz acqui­si­tions. Good sta­tis­tics is obtained with 150–200 laser hits. The method calls for rel­a­tively low laser strengths, which favors higher res­o­ lu­tion, lim­its the for­ma­tion of arte­fac­tual gas-phase dimers, and appar­ently elim­i­nates the for­ma­tion of 4HCCA matrix adducts (sin­a­pic acid adducts can still be observed when this matrix is used instead of 4HCCA). With care­ful inter­nal cal­i­bra­tion, masses can be mea­sured with sta­tis­ti­cal accu­ra­cies of 30–150 ppm in lin­ear TOF mode (using sev­eral charge states for full-length pro­tein or large fold­ing domains), and 5–10 ppm in reflec­tor TOF mode (for masses below 5000 Da). Finally, this method has been rou­tinely used in our group as well as sev­eral oth­ers for the anal­y­sis of pep­tide mix­tures in the con­text of pro­teo­mics pep­tide map­ping and mod­ifi ­ ed pep­tides char­ac­ter­iza­tion [14–20]. In this case, good results are often obtained using stan­dard WA (2:1 water:ace­to­ni­trile; v/v) or TWA (2:1 water:ace­to­ni­trile (v/v) with 0.1% final TFA) sol­vent solu­tions for the matrix solu­tion. 3. Role of ini­ti­a­tor methi­o­nine(s) in pro­tein sta­bil­ity and pro­tein func­tion In rou­tine struc­tural biol­ogy work, over-expressed pro­teins are first checked for proper primary sequence using SDS–PAGE. This pro­vides a rough esti­mate of the molec­u­lar mass, and indi­cates whether the pro­tein prep­a­ra­tion is homog­en ­ ous. If the pro­tein prep­a­ra­tion shows more than a sin­gle thin band, the exact nature of the het­er­og ­ e­ne­ity should be assessed using MS to pre­cisely iden­ tify the ori­gin of the het­er­og ­ e­ne­ity. Homog­en ­ ous pro­tein prep­ar­ a­ tions are desir­able to favor the for­ma­tion of crys­tals appro­pri­ate for X-ray dif­frac­tion. As the exam­ple below will show, a gel band at a lower appar­ent mass does not always trans­late into a pro­te­o­lytic

C 417

M240 (2+)

60

M240

Intensity (a.u.)

N 1

0

B

9500

(3+)

12500

m/z

N 1

pore

M107 M107

RCK domain

C 417

C 340 C 340

Fig. 3. Coex­pres­sion of potas­sium chan­nels with their own reg­ul­ a­tory C-ter­mi­nal domains. (A) MALDI-TOF mass spec­trum of the coex­pres­sed Esch­e­richia coli potas­sium chan­ nel pro­teins. a.u., arbi­trary units. (B) Model for chan­nel-RCK domain arrange­ment in the Meth­a­noc­coc­cus ther­mo­auto­troph­i­cum cal­cium-gated potas­sium chan­nel (adapted from Ji­ang et al. [19]). A sin­gle gene gives rise to two gene prod­ucts, and both prod­ucts are required in the assem­bly of a func­tional K+ chan­nel.



G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

cleav­age frag­ment. In fact, care­ful MS anal­y­sis can pro­vide a piece of infor­ma­tion on this extra band which is impor­tant not only from a struc­tural but also from a func­tional stand­point. The Esch­e­richia coli potas­sium chan­nel showed such a lower band on SDS–PAGE [21]. Mass spec­trom­e­try iden­ti­fied the ­full-length chan­nel, along with a lower mass spe­cies (Fig. 3A). This sec­ond spe­cies would have nat­u­rally been hypoth­e­sized to be a cleav­age frag­ment of the pro­tein. How­ever, no N-ter­mi­nal cleav­age prod­uct matched the observed mass within accept­able mea­sure­ ment error, when the accu­racy of the method with care­ful inter­nal cal­i­bra­tion placed the expected error at around 3 Da. For a C-ter­mi­ nus-con­tain­ing cleav­age prod­uct, we found a match for a sequence start­ing at methi­o­nine #240, sug­gest­ing a cleav­age after thre­o­nine #239, a rather unusual amino acid for pro­te­ol­y­sis, even if acci­den­ tal. This obser­va­tion prompted us to con­sider other inter­pre­ta­tions for the observed mass, in par­tic­u­lar the pres­ence of a sec­ond­ary site for the ini­ti­a­tion of trans­la­tion. If a sec­ond ini­ti­at­ ion site on the tran­script (or on a sec­ond tran­script) is used by the trans­la­tion machin­ery, then it is pos­si­ble to observe the simul­ta­neous expres­ sion, or co-expres­sion, of two dif­fer­ent pro­tein spe­cies start­ing at two dif­fer­ent me­thio­nines in the sequence. In the case of the E. coli potas­sium chan­nel, the iden­tity of the spe­cies start­ing at M240, appar­ently coel­ut­ing with the full-length chan­nel, was con­firmed by Ed­man sequenc­ing. Tak­ing this infor­ma­tion into account, this C-ter­mi­nal domain, cor­re­spond­ing to the RCK (reg­u­la­tor of chan­ nel con­duc­tance) domain of the chan­nel, was crys­tal­lized inde­ pen­dently from the full-length chan­nel [21]. The Meth­a­noc­coc­cus ther­mo­auto­troph­i­cum cal­cium-gated chan­nel exhib­ited sim­i­lar behav­ior: the gene appar­ently pos­sessed a sec­ond­ary trans­la­ tion ini­ti­a­tion site with co-expres­sion of the reg­u­la­tory domain. In an attempt to facil­i­tate crys­tal­li­za­tion by pro­duc­ing the most homog­en ­ ous pro­tein prep­a­ra­tion pos­si­ble, this sec­ond methi­o­ nine was mutated out. How­ever, the mutant pro­tein proved to be unsta­ble, prone to aggre­ga­tion, and unsuit­able for crys­tal­li­za­tion. Rod­er­ick Mack­in­non and his team finally achieved the res­o­lu­tion of the struc­ture of the chan­nel co-expressed with its own reg­u­la­ tory C-ter­mi­nal domain, ele­gantly pro­vid­ing the expla­na­tion for the behav­ior of the mutant (Fig. 3B) [22,23]. The RCK domain has to be pres­ent in stoi­chi­om­e­tric amounts to dimer­ize with the cor­ re­spond­ing domain in the full-length chan­nel. In its absence, the full-length chan­nel is left to dimer­ize with itself, a qua­ter­nary struc­ture which is incom­pat­i­ble with proper inser­tion of all the trans­mem­brane seg­ments into the mem­brane (Fig. 3B) and ulti­ mately leads to inac­tiv­ity. The iden­ti­fic­ a­tion of a sec­ond ini­ti­a­tor methi­o­nine was one of the key ele­ments in the elu­ci­da­tion of the chan­nel mech­a­nism with respect to gat­ing. The exam­ple above should thus serve to stress the impor­tance of a proper inven­tory of all func­tion­ally rel­e­vant trans­la­tion ini­ti­a­tion sites. 4. Lim­ited pro­te­ol­y­sis for delin­eat­ing folded pro­tein domains (domain deter­mi­na­tion) The use of lim­ited pro­te­ol­y­sis to delin­eate folded pro­tein domains is well doc­u­mented and has long been employed to assist in the design of pro­tein con­structs suit­able for crys­tal­li­za­tion (for a review of domain deter­mi­na­tion by lim­ited pro­te­ol­y­sis com­bined with mass spec­trom­e­try, see [24]). First, it shows whether a pro­ tein is cor­rectly folded when expressed accord­ing to the pre­dicted open read­ing frame. If the ini­tial sequence turns out to be prob­ lem­atic, domain deter­mi­na­tion is a great help in devis­ing strat­e­ gies for expres­sion of a suit­able con­struct. It should be noted that con­struct pro­duc­tion by auto­mated manip­u­la­tion of the gene sequence, while com­ple­men­tary to domain deter­mi­na­tion, does not yield the same detail of infor­ma­tion. For exam­ple, choos­ ing appro­pri­ate loca­tions for the begin­ning of the sequence may require know­ing nat­u­rally folded domains. A pro­tein sequence

59

Table 2 Sug­gested ini­tial enzyme:sub­strate ratios for set­ting up lim­ited pro­te­ol­y­sis exper­ i­ments Pro­tein­ase (sequenc­ing grade)

Pro­tein­ase: pro­tein ratio (w/w)

Endo­pro­tein­ase LysC Endo­pro­tein­ase ArgC Endo­pro­tein­ase AspN Endo­pro­tein­ase GluC (V8) Tryp­sin Chy­mo­tryp­sin Sub­til­i­sin Pro­tein­ase K

1:2000 1:500 1:1000 1:200 1:100 1:400 1:4000 1:2000

start­ing in the mid­dle of a folded domain is more likely to give rise to a mis­folded prod­uct. A loose C-ter­mi­nal stretch of sequence will be more sen­si­tive to adven­ti­tious pro­te­ol­y­sis dur­ing expres­sion and puri­fi­ca­tion. On the other hand, one some­times may decide to keep a floppy loop or extrem­ity in the sequence because it is impor­tant for expres­sion or pro­tein func­tion. Domain deter­mi­na­ tion pro­vides the struc­tural biol­og ­ ist with the level of infor­ma­tion nec­es­sary to make these deci­sions on a rational basis. Lim­ited pro­te­ol­y­sis is con­ducted in approx­i­mately sin­gle-hit sta­tis­tics con­di­tions so that, on aver­age, only one cleav­age occurs per pro­tein chain. This ensures that the gen­eral struc­ture of the pro­tein is con­served while pro­duc­ing a lim­ited num­ber of “nicks” on the pro­tein back­bone. The approach is akin to a foot­print­ing of sol­vent-acces­si­ble pep­tide bonds. The choice of enzymes work­ing in con­di­tions where the mem­ brane pro­tein is in its native state is usu­ally quite large. It is advis­ able to start exper­i­ments with pro­tein­ases with nar­row spec­i­fic­ity, i.e. with pro­te­o­lytic activ­ity against one or two amino acid types. The most com­mon are endo­pro­tein­ase LysC, ArgC, AspN, GluC, and tryp­sin. The infor­ma­tion obtained with spe­cific pro­te­ases can then be used to facil­i­tate the inter­pre­ta­tion of cleav­age results for lower spec­i­fic­ity pro­te­ases such as sub­til­i­sin or pro­tein­ase K. Table 2 shows enzyme: sub­strate ratios that have been opti­mized to be used as a start­ing point when design­ing lim­ited pro­te­ol­y­sis exper­i­ments. These ratios are based on numer­ous lim­ited pro­te­ol­ y­sis exper­i­ments on inte­gral mem­brane pro­teins, while the ratios sug­gested by man­u­fac­tur­ers are given for the pur­pose of total (not lim­ited) pro­te­ol­y­sis, and are aimed at pro­teins in gen­eral, not mem­brane pro­teins. The dif­fer­ent ratios also reflect the wide range of pro­tein­ases spe­cific activ­ity. All pro­te­ases should be sequenc­ ing grade. The mem­brane pro­tein sub­strate should be as homog­e­ nous as pos­si­ble. Het­er­o­ge­ne­ity in the start­ing mate­rial can lead to mis­in­ter­pre­ta­tion of the result as cleav­age prod­ucts can arise from the legit sub­strate or from a con­tam­i­nant spe­cies. The ­kinet­ics of ­cleav­age is mon­i­tored using MALDI-TOF MS, using the ultra­thin layer method described above. Cleav­age sites are then ranked accord­ing to how early they appear in the exper­i­ment, roughly divid­ing into “major” and “minor” sites, cor­re­spond­ing to “early” or “late” sites, respec­tively. A pro­tein­ase sen­si­tiv­ity map can then be drawn over the primary and/or sec­ond­ary struc­ture of the pro­tein. By repeat­ing pro­te­ol­y­sis exper­i­ments with var­i­ous pro­tein­ases, it is pos­si­ble to get a fairly detailed pic­ture of the folded domains. This approach was for exam­ple suc­cess­fully applied to the volt­agegated potas­sium chan­nel [25]. Lim­ited pro­te­ol­y­sis can also assist in the choice of the deter­gent for opti­mum pro­tein sta­bil­ity. Beyond the design of a sta­ble, prop­erly folded pro­tein con­struct, once dif­frac­tion data has been gath­ered, domain elu­ci­da­tion can pro­vide inter­est­ing clues when build­ing the struc­ture model. On the E. coli chlo­ride chan­nel, a num­ber of such exper­i­ments had been per­formed. Fig. 4A shows the cleav­age sites iden­ti­fied for a sub­set of two of the pro­tein­ases employed for this task. Topol­ogy pre­dic­tion graphs drawn at the time showed the puta­tive trans­ mem­brane heli­cal seg­ments, con­sist­ing of 12 ­heli­ces of roughly

60

G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

B

A

K2

K465

K2

K465

Fig. 4. Major cleav­age sites in a sub­set of lim­ited pro­te­ol­y­sis exper­i­ments on the Esch­e­richia coli ClC chlo­ride chan­nel. Green, endo­pro­tein­ase ArgC cleav­age. Orange, chy­mo­ tryp­sin. (A) Plot of major cleav­age sites onto the algo­rithm-pre­dicted helix arrange­ment. (B) Plot of major sites onto the helix arrange­ment as observed in the crys­tal struc­ ture. (For inter­pre­ta­tion of the ref­er­ences to col­our in this fig­ure leg­end, the reader is referred to the web ver­sion of this arti­cle.)

5. Time-depen­dent pro­te­ol­y­sis to obtain a homog­e­nous pro­tein seg­ment One of the main obsta­cles to obtain­ing highly dif­fract­ing pro­ tein crys­tals is the pres­ence of a micro-het­er­o­ge­ne­ity of sequence, usu­ally due to adven­ti­tious pro­te­ol­y­sis or a post-trans­la­tional mod­i­fi­ca­tion. Pro­tein­ases can be employed to tai­lor the pro­tein to a well-folded set of domains. In this case, opti­mum con­di­tions such as the choice of enzyme, the enzyme:sub­strate ratio, the tem­per­a­ture and dura­tion of cleav­age may be dif­fer­ent from the opti­mum con­di­tions found for domain deter­mi­na­tion by lim­ited pro­te­ol­y­sis (see Sec­tion 1.3). By def­i­ni­tion, one now aims at a com­plete cleav­age, lim­ited to one or two areas of the pro­tein, usu­ally close to the ter­mini. The kinet­ics of cleav­age is again mon­i­tored using MALDI-TOF MS and the ultra­thin layer method. In favor­able cases, it is pos­si­ble to find a set of con­di­tions which will pro­duce a homog­e­nous, well defined pro­tein spe­cies (for an exam­ple, see [26]). 6. Sele­no­me­thi­on ­ ine incor­po­ra­tion for MAD anal­y­sis For X-Ray struc­ture deter­mi­na­tion, phase deter­mi­na­tion can be achieved by iso­mor­phous replace­ment exper­i­ments using heavy atom deriv­at­ ives such as methyl mer­cury. How­ever, the

30

Intensity (a.u.)

equal length, sug­gest­ing neatly bunched seg­ments in rho­dop­ sin-like fash­ion. When plot­ting the observed cleav­ages onto the puta­tive helix arrange­ment, some odd results where imme­di­ately evi­dent, with unex­pected cleav­ages in the mid­dle of heli­ces, or no cleav­age within obvi­ously exposed loops (Fig. 4A, empha­sized in red). This appar­ent con­flict pro­vided clues that the topol­ogy pre­dic­tion plot was prob­a­bly mis­lead­ing. Fig. 4B shows the pro­ tein­ase tar­get sites sit­u­ated onto the actual helix arrange­ment as deter­mined by X-ray dif­frac­tion [26]. The struc­ture showed that heli­ces dif­fer in num­ber and length from the pre­dic­tion, and that some actu­ally lay across the mem­brane in near-hor­i­zon­tal direc­ tion [26]. The ori­en­ta­tion of polar­ized heli­ces is inte­gral to the chlo­ride chan­nel’s mech­a­nism of ion trans­port across the mem­ brane. Although sec­ond­ary struc­ture pre­dic­tion algo­rithms for mem­ brane pro­teins may have some­what improved in the past 6 years thanks to the accu­mu­la­tion of struc­ture data, they should still be used with cir­cum­spec­tion and cross-checked using lim­ited pro­te­ ol­y­sis.

Methionine

Seleno-Met.

0 19800

20000

20200

m/z Fig. 5. Sele­no­me­thi­o­nine replace­ment in Esch­e­richia coli potas­sium chan­nel RCK domain. The upper spec­trum is off­set for dis­play pur­poses. Sin­gly charged spe­cies are shown. a.u.: arbi­trary units. D(m/z) is the dif­fer­ence between the observed m/z ratios for the native and sele­no­me­thi­o­nine-replaced pro­tein spe­cies. It is used to cal­cu­late the dif­fer­ence in rel­a­tive molec­ul­ ar mass between the sele­no­me­thi­o­ nine- and methi­on ­ ine-con­tain­ing pro­tein spe­cies, tak­ing into account the pro­tein charge state z: D(m/z) £ z = DM = MSel­e­no­Met ¡ Mmet. The molec­u­lar mass dif­fer­ence DM is divided by 46.9 to cal­cu­late the num­ber of replace­ments. In this exam­ple, D(m/z) = 188.9 and z = 1, hence DM = 188.9 Da. Thus the num­ber of replace­ments is 188.9/46.9 = 4.03. This peak cor­re­sponds to a pro­tein pool where all four me­thio­ nines in the pro­tein domain have been suc­cess­fully replaced.

result­ing deriv­a­tized mem­brane pro­tein becomes even more hydro­pho­bic, ren­der­ing anal­y­sis of the pro­tein in solu­tion or of the dis­solved crys­tal quite dif ­fi­cult [8]. The MALDI-TOF MS method described above can some­times be suc­cess­ful. For MAD (Multi-wave­length Anom­a­lous Dis­per­sion) phase deter­mi­na­ tion exper­i­ments, me­thio­nines are replaced by sel­e­nom­ethi­o­ nines. Replace­ment of methi­on ­ ine by sele­no­me­thi­on ­ ine yields a mass shift of 46.9 Da, cor­re­spond­ing to the dif­fer­ence in aver­age atomic mass between sul­fur and sele­nium. MALDI-TOF MS can be used to check that met­a­bolic replace­ment is suf ­fi­ciently com­ plete. As an exam­ple, Fig. 5 shows the mass shift cor­re­spond­ing to the num­ber of replaced me­thio­nines for the E. coli potas­sium chan­nel RCK domain. The major peak shows all four me­thio­ nines in the pro­tein domain have been suc­cess­fully replaced. Note that a minor frac­tion with only three replace­ments can be observed, mean­ing that at every site a small per­cent­age of methi­ o­nine is pres­ent. This is because, although replace­ment occurs



G. Ga­bant, M. Cad­ene / Methods 46 (2008) 54–61

­ omo­ge­neously across the sequence, the ­sele­no­me­thi­o­nine used h to pro­duce the seleno-pro­tein always con­tains a small per­cent­age of methi­on ­ ine and met­a­bolic replace­ment may not be entirely com­plete. If p is the sele­no­me­thi­o­nine/methi­o­nine ratio at every site (also known as the occu­pancy), the pro­por­tion of par­tially replaced pro­tein can be pre­dicted using the bino­mial prob­a­bil­ ity equa­tion: P(k) = (n!/k!(n ¡ k)!)pk(1¡p)n¡k, where n is the total num­ber of me­thio­nines to be replaced, and k, the num­ber of sel­ e­nom­ethi­on ­ ines. In the exam­ple above, for p = 0.94, the pro­por­ tion of the pro­tein pool with 3 out of 4 replace­ments would be: P(3) = (4!/3!(4 ¡ 3)!). p3.(1 ¡ p)4¡3 = 0.2 (20%). ­Typ­i­cally, a p ratio (occu­pancy) over 90% is suf ­fi­cient for MAD anal­y­sis, in which case the major spe­cies in the spec­trum will be the fully replaced pro­tein. We sim­i­larly ana­lyzed the full-length E. coli chlo­ride chan­nel pro­tein using the above MALDI-TOF MS approach and found all 17 me­thio­nines to be replaced [26]. 7. Con­clu­sion Mass spec­trom­e­try is a pow­er­ful and ver­sa­tile tool offer­ing deep insights into the state of the inte­gral mem­brane pro­tein the struc­tur­al­ist intends to crys­tal­lize. With appro­pri­ate sam­ple prep­a­ ra­tion meth­ods, it pro­vides infor­ma­tion that can some­times prove crit­i­cal at var­i­ous stages of the struc­ture deter­mi­na­tion pro­cess, from pro­tein expres­sion to model build­ing. Acknowl­edge­ments MC wishes to thank all her col­lab­o­ra­tors, past and pres­ent, from the Rod­er­ick MacK­in­non lab­o­ra­tory at the Rocke­fel­ler Uni­ver­sity, par­tic­u­larly Youx­ing Ji­ang, Ray­mund Dut­zler, Alice MacK­in­non, Alex­an­der Pico, Va­nessa Ruta, Ernest Camp­b ell, Jac­que­line Gul­bis, Moto­hiko Nish­ida, Se­bast­ien Po­get and João Mo­rais Ca­bral. Many thanks to Ste­ven Cohen and Ju­lio Cesar Pa­do­van for use­ful dis­cus­sions. MC is also deeply grate­ful to Brian T. Chait and Rod­er­ick MacK­in­non for years of won­der­ful sci­ence.

61

Ref­er­ences [1] D.A. Doyle, J. Mo­rais Ca­bral, R.A. Pfu­etz­ner, A. Kuo, J.M. Gul­bis, S.L. Cohen, B.T. Chait, R. MacK­in­non, Sci­ence 280 (1998) 69–77. [2] K.L. Schey, D.I. Pa­pac, D.R. Knapp, R.K. Crouch, Bio­phys. J. 63 (1992) 1240–1243. [3] P.A. Schin­dler, A. Van Dors­sel­aer, A.M. Fa­lick, Anal. Bio­chem. 213 (1993) 256– 263. [4] V. Schna­ible, J. Mich­els, K. Zeth, J. Frei­gang, W. Welte, S. Buh­ler, M.O. Glock­er, M. Przybyl­ski, Int. J. Mass Spec­trom. Ion Pro­cess. 169–170 (1997) 165–177. [5] J.P. Whitel­egge, C.B. Gun­der­sen, K.F. Fau­ll, Pro­tein Sci. 7 (1998) 1423–1430. [6] M. Cad­ene, B.T. Chait, Anal. Chem. 72 (2000) 5655–5658. [7] Z. Ab­lo­nczy, R.K. Crouch, D.R. Knapp, J. Chro­ma­togr. B. Ana­lyt. Tech­nol. Bio­ med. Life Sci. 825 (2005) 169–175. [8] S.L. Cohen, J.C. Pa­do­van, B.T. Chait, Anal. Chem. 72 (2000) 574–579. [9] R.C. Be­a­vis, B.T. Chait, Meth­ods Enz­y­mol. 270 (1996) 519–551. [10] F. Xiang, R.C. Be­a­vis, Rapid Com­mun. Mass Spec­trom. 8 (1994) 199–204. [11] M. Saf­fer­ling, H. Grif ­fi th, J. Jin, J. Sharp, M. De Jesus, C. Ng, T.A. Krul­wich, D.N. Wang, Bio­chem­is­try 42 (2003) 13969–13976. [12] Y. Hu­ang, M.J. Lem­i­eux, J. Song, M. Auer, D.N. Wang, Sci­ence 301 (2003) 616– 620. [13] D. Fe­nyo, Q.J. Wang, J.A. De­Grasse, J.C. Pa­do­van, M. Cad­ene, B.T. Chait, J. Visu­ al­ized Exper­i­ments Cel­lu­lar Biol­ogy, Issue 3 (2007). http://www.jove.com/ index/details.stp?ID=192. [14] C. Buré, S. Gof ­fi­nont, A.F. Del­mas, M. Cad­ene, F. Cu­lard, J. Mol. Biol. 376 (2008) 120–130. [15] M. Trest­er-Zed­litz, A. Bur­lin­game, B. Ko­bilka, M. von Zas­trow, Bio­chem­is­try 44 (2005) 6133–6143. [16] C. Sei­bert, M. Cad­ene, A. San­fiz, B.T. Chait, T.P. Sak­mar, P.N.A.S. 99 (2002) 11031–11036. [17] R.W. He­pler, R. Kelly, T.B. McNe­ely, H.X. Fan, M.C. Los­ada, H.A. George, A. Woods, L.D. Cope, A. Ban­sal, J.C. Cook, G. Zang, S.L. Cohen, X.R. Wei, P.M. Kel­ler, E. Lef­fel, J.G. Joyce, L. Pitt, L.D. Schultz, K.U. Jan­sen, M. Ku­rtz, Vac­cine 24 (2006) 1501–1514. [18] W.M. Ab­ida, A. Niko­laev, W. Zhao, W. Zhang, W. Gu, J. Biol. Chem. 282 (2007) 1797–1804. [19] V. Bourge­aux, M. Cad­ene, F. Pil­ler, V. Pil­ler, Chem. Biol. Chem. 8 (2007) 37–40. [20] E.J. Chang, V. Ar­cha­mba­ult, D.T. McLach­lin, A.N. Krut­chin­sky, B.T. Chait, Anal. Chem. 76 (2004) 4472–4483. [21] Y. Ji­ang, A. Pico, M. Cad­ene, B.T. Chait, R. MacK­in­non, Neu­ron 29 (2001) 593– 601. [22] Y. Ji­ang, A. Lee, J. Chen, M. Cad­ene, B.T. Chait, R. MacK­in­non, Nature 417 (2002) 515–522. [23] Y. Ji­ang, A. Lee, J. Chen, M. Cad­ene, B.T. Chait, R. MacK­in­non, Nature 417 (2002) 523–526. [24] S.L. Cohen, B.T. Chait, Annu. Rev. Bio­phys. Bio­mol. Struct. 30 (2001) 67–85. [25] Y. Ji­ang, A. Lee, J. Chen, M. Cad­ene, B.T. Chait, R. MacK­in­non, Nature 423 (2003) 33–41. [26] R. Dut­zler, E.B. Camp­bell, M. Cad­ene, B.T. Chait, R. MacK­in­non, Nature 415 (2002) 287–294.