Tài liệu Báo cáo khoa học: Applications of diagonal chromatography for proteomewide characterization of protein modifications and activity-based analyses doc

This information is now available for a Keywords activity-based probe; ATP-binding proteins; COFRADIC; diagonal chromatography; N-terminal peptides; peptide sorting; protein N-glycosylat

Applications of diagonal chromatography for proteome-wide characterization of protein modifications and

activity-based analyses

Kris Gevaert1,2, Francis Impens1,2, Petra Van Damme1,2, Bart Ghesquie`re1,2, Xavier Hanoulle3 and Joe¨l Vandekerckhove1,2

1 Department of Medical Protein Research, VIB, Ghent, Belgium

2 Department of Biochemistry, Ghent University, Belgium

3 UMR 8576 CNRS ) University of Sciences and Technologies of Lille, Structural and Functional Glycobiology Unit, Villeneuve d’Ascq, France

Introduction

Proteomics refers to a qualitative, differential and

quantitative estimation of a proteome Proteomes can

be extremely complex, often encompassing more than

10 000 different components per cell Two-dimensional

gel electrophoresis [1] followed by electroblotting and

microsequencing [2–4] or in-gel digestion combined

with Edman sequencing [5] of the generated peptides

or peptide mass fingerprinting [6–10] have been the

methods of choice to reproducibly separate and iden-tify complex protein mixtures Although large-scale 2D gel electrophoresis separates thousands of proteins [11,12], probably no more than a few hundred different proteins have been identified from such gels To obtain better proteome coverage, alternative methods were introduced Groundbreaking methodologies became available when high-throughput genome sequencing started to cover the entire genetic information of several species This information is now available for a

Keywords

activity-based probe; ATP-binding proteins;

COFRADIC; diagonal chromatography;

N-terminal peptides; peptide sorting; protein

N-glycosylation; protein processing

Correspondence

K Gevaert, Department of Biochemistry,

Faculty of Medicine and Health Sciences,

Ghent University, A Baertsoenkaai 3,

B-9000 Ghent, Belgium

Fax: +32 92649496

Tel: +32 92649274

E-mail: kris.gevaert@ugent.be

Website: http://www.proteomics.be

(Received 24 April 2007, revised 10

Septem-ber 2007, accepted 17 OctoSeptem-ber 2007)

doi:10.1111/j.1742-4658.2007.06149.x

Numerous gel-free proteomics techniques have been reported over the past few years, introducing a move from proteins to peptides as bits of informa-tion in qualitative and quantitative proteome studies Many shotgun pro-teomics techniques randomly sample thousands of peptides in a qualitative and quantitative manner but overlook the vast majority of protein modifi-cations that are often crucial for proper protein structure and function Peptide-based proteomic approaches have thus been developed to profile a diverse set of modifications including, but not at all limited, to phosphory-lation, glycosylation and ubiquitination Typical here is that each modifica-tion needs a specific, tailor-made analytical procedure In this minireview,

we discuss how one technique) diagonal reverse-phase chromatogra-phy) is applied to study two different types of protein modification: pro-tein processing and propro-tein N-glycosylation Additionally, we discuss an activity-based proteome study in which purine-binding proteins were pro-filed by diagonal chromatography

Abbreviations

ABP, activity-based probe; COFRADIC, combined fractional diagonal chromatography; FSBA, 5¢-p-fluorosulfonylbenzoyladenosine; FSBG, 5¢-p-fluorosulfonylbenzoylguanosine; iTRAQ, isobaric tags for relative and absolute quantification; MudPIT, multidimensional protein

identification technology; PNGaseF, peptide N-glycosidase F; SB, sulfobenzoyl; SILAC, stable isotope labeling by amino acids in cell culture; Nbs2, 2,4,6-trinitrobenzenesulfonic acid.

large number of species, and it now suffices to generate

partial protein sequence information with which to

access entire (predicted) protein sequences stored in

expressed sequence tag, gene and protein sequence

databases

This brought the dawn of novel strategies for

tein identification Measured masses of peptides

pro-duced by cleaving a protein with a protease with

well-known specificity (e.g trypsin) were searched

against a database of peptide masses calculated from

protein sequences derived from genome sequences [6–

10] When these peptides are derived from a mixture

of proteins, they are subjected to MS⁄ MS

fragmenta-tion for identificafragmenta-tion [13] Recently, top-down

pro-tein sequencing combining ESI MS and highly

accurate FT MS [14] was shown to match proteins

larger than 200 kDa to sequences in databases [15]

Such strategies only became possible following the

availability of massive numbers of DNA sequences,

recent developments in MS, and bioinformatics tools

that link DNA and protein sequences to information

generated by different types of mass spectrometers

[16–18]

Recently, peptides have increasingly become the

center of analysis: protein mixtures, either partially

purified by prefractionation or as such, are digested

with trypsin, and the generated peptide mixture is

analyzed When cell or tissue lysates, or even isolated

organelles, are analyzed, the number of peptides

becomes so high that mass spectrometers can no

longer analyze all of the peptides This results in

poor sample coverage, generally referred to as

ran-dom sampling or undersampling [19], and it became

crucial to add peptide prefractionation before MS

analysis Yates’ group introduced separation of

pep-tides based on two parameters [20]) net charge and

hydrophobicity – and called their technique

multi-dimensional protein identification technology (MudPIT

[21]) MudPIT has since then been used in several

studies and has demonstrated its value, but it still

suffers from undersampling [19]

Selecting a lower number of peptides representative

of each protein originally present in the mixture may

alleviate this problem These so-called signature

pep-tides [22] are then the only analyzed components, and

in this way a less complex peptide mixture is presented

to the mass spectrometer The first reports using this

strategy were selective for cysteinyl peptides, allowed

quantification (differential analysis), and used biotin

tagging for consecutive capture by immobilized avidin

[23] Later on, affinity selection was used to isolate, for

instance, phosphopeptides [24], N-glycosylated peptides

[25], ubiquitinated peptides [26], and N-terminal pep-tides [27]

COFRADIC as a peptide-sorting tool Our peptide-centric proteome approach [28,29] sorts signature peptides and selects the part of a proteome containing the information of biological interest Our technique is based on diagonal chromatography [30,31] consisting of two repeated, identical peptide separa-tions with a specific modification reaction (sorting step) in between Peptides that remain unchanged elute

at the same position in the two chromatographic runs, whereas peptides that acquire a modification segregate from the unchanged peptides either in earlier or in later fractions To reduce the number of repetitive chromatographic runs, several fractions from the pri-mary run can be combined and subjected to the sort-ing reaction (Fig 1) For this reason, we call this adapted version of diagonal chromatography com-bined fractional diagonal chromatography

(COFRAD-IC [32])

It should be clear from the peptide-sorting principle that any chemical or enzymatic modification that is highly specific, is quantitative and produces a suffi-ciently large chromatographic shift can be imple-mented in COFRADIC This is illustrated by applications in which selection for methionyl or cyste-inyl peptides was carried out in tryptic digests of total cellular lysates [33,34], or where the N-terminal pep-tides of the proteins present in the mixture were selected [33–36] Similarly, as we can select peptides based on the specific chemical nature of their amino acid side chains, we can also select peptides carrying post-translational modifications, either by removing this modification (e.g by dephosphorylation of phos-phopeptides [37]), or by converting it into a moiety with altered properties (e.g by reducing nitrotyrosine to aminotyrosine [38]) An overview of the COFRADIC sorting protocols that have been developed is given in Table 1

We here concentrate on applications of COFRADIC

in studying selected post-translational modifica-tions) protein processing [34,36] and N-glycosylation [39] – and describe the use of COFRADIC for study-ing interactions between small molecules and proteins The latter is a particular application of ‘post-transla-tional COFRADIC’, by which a small molecule is covalently linked to a target protein and the corre-sponding modified tryptic peptide is then sorted using the principles of diagonal chromatography The exam-ple given here is a global activity-based proteome

analysis of purine-binding proteins in a total lysate of human Jurkat T-cells [40]

COFRADIC analysis of protein processing ) protease degradomics Protein processing introduces novel protein fragments that may be visualized on 2D polyacrylamide gels For example, Canals et al used the fluorescent 2D differ-ence gel electrophoresis technique [41] to catalog quan-titative differences in the protein composition of conditioned media of cells either expressing the metal-loproteinase ADAMTS1 at physiological levels or overexpressing it [42] The latter scenario led to an increase of fragments of proteins shed by ADAMTS1 into the medium that were picked by difference gel electrophoresis and identified by MS In fact, this study led to the identification of five potential ADAM-TS1 substrates, two of which (nidogens 1 and 2) were further validated Gel-free proteomic approaches have been introduced for ‘degradomics’ [43] research as well The group of Overall used isotope-coded affinity tag [23] combined with LC-MS⁄ MS to quantify the levels

of secreted extracellular matrix proteins in breast carcin-oma cell cultures overexpressing a membrane type 1 matrix metalloproteinase [44] and, more recently, they multiplexed their analyses using isobaric tags for relative and absolute quantification (iTRAQ) reagents for the identification of matrix metalloproteinase-2 substrates in fibroblasts [45]

Clearly, both gel-based and gel-free approaches point to potential protease substrates; however, at this stage it is important to note that the characterization

of the actual protein cleavage site has typically remained elusive Nonetheless, the latter information is highly valuable, as it can lead to more rational design

of protease inhibitors [46], it is vital for constructing precise algorithms that predict protease substrates [47], and, after all, protein processing is a post-translational modification that should preferably be characterized before any assumption concerning the protease sub-strate potential is made Protein processing produces a novel C-terminal peptide (from the N-terminal frag-ment of a substrate) and a novel N-terminal peptide (from the C-terminal fragment) Hence, identifying either one of these ‘reporter peptides’ directly points to the actual processing site As recently reviewed [29], in

a whole proteomic background, C-terminal peptides are only poorly isolated On the other hand, N-termi-nal COFRADIC [33], and the more or less ‘single-step’ isolations of N-terminal peptides by protein sequence tags [48] and positional proteomics [27] were shown to isolate N-terminal peptides from complex mixtures

min

mAU

0

200

400

600

800

1000

1200

1400

primary separation

combine primary fractions

COFRADIC sorting reaction

LC-MS/MS analysis

min

mAU

0

100

200

300

400

500

600

700

secondary separation

Fig 1 The COFRADIC peptide sorting scheme A peptide mixture

is first separated by RP-HPLC (the primary COFRADIC separation).

Here, the UV absorbance profile at 214 nm of a tryptic digest of a

proteome preparation from human Jurkat T-cells is shown Primary

fractions (indicated in light gray boxes) are combined ) here, four

primary fractions (each 1 min wide) that are separated by a 13 min

window – and undergo a chemical or enzymatic reaction (the

COFRADIC sorting reaction) In this particular case, the side chains

of methionines were oxidized by hydrogen peroxide, leading to the

formation of methione-sulfoxide During a second, identical

separa-tion (the secondary COFRADIC separasepara-tion), such oxidized methionyl

peptides undergo a hydrophilic shift and segregate away from the

bulk of nonmethionyl peptides Methionyl peptides are thus

col-lected (dark gray boxes) and analyzed by LC-MS ⁄ MS.

However, only the N-terminal COFRADIC approach

has thus far been applied to protease degradomics

research [36,38] and is discussed here (potential

draw-backs of the two affinity-based peptide isolation

proto-cols are discussed in Conclusions)

In essence, N-terminal COFRADIC segregates

pep-tides containing the protein N-termini from internal

peptides This is achieved following an initial

acetyla-tion or trideutero-acetylaacetyla-tion reacacetyla-tion on a complete

proteome prior to trypsin digestion This blocks all

free a-amines and e-amines and, further down,

distin-guishes between in vivo blocked (acetylated) and in vivo

free (trideutero-acetylated) protein N-termini Trypsin

no longer recognizes acetylated lysines, and,

conse-quently, upon digestion, Arg-C type peptides are

generated In fact, two types of peptides are now

apparent: N-terminal peptides with a blocked,

acety-lated⁄ trideutero-acetylated a-amine, and internal

pep-tides carrying a free a-amine This peptide mixture is

first separated by RP-HPLC and collected in a small

number of primary fractions Then, internal peptides

present in each fraction are reacted with

2,4,6-trinitro-benzenesulfonic acid, which is known to efficiently and

quantitatively modify primary amines [49] Internal

peptides thereby acquire a trinitrophenyl group at their

N-terminus and thus become very hydrophobic

Run-ning such TNBS-modified primary fractions a second

time on the same column and under identical

chro-matographic conditions will now segregate

TNBS-non-reactive N-terminal peptides (all their amino groups

were already blocked) from TNBS-reacted internal

peptides, which underwent a very strong hydrophobic

shift (Table 1) Following metabolic or postmetabolic

labeling, N-terminal peptides of two (or more)

proteo-mes can be weighed against each other and,

impor-tantly, neo-N-termini originating from protein

processing are readily distinguished [34,36]

The characterization of protease substrates by such

a differential N-terminal COFRADIC approach is illustrated in Fig 2 In an ongoing project, host cell substrates of the HIV-1 protease are catalogued in human Jurkat T-cells grown in stable isotope labeling

by amino acids in cell culture (SILAC) medium supple-mented with either natural, light 12C6-arginine or heavy 13C6-arginine [50] Arginine is clearly the essen-tial amino acid of choice, as all N-terminal peptides isolated by COFRADIC, by the nature of the process, will end on an arginine residue This metabolic labeling introduces a mass spacing of 6 Da between light and heavy N-terminal peptides

Cells are typically lysed by repeated freeze–thawing, and the lysate is either incubated with recombinant HIV-1 protease or left untreated (control) Following protease incubation, both proteomes are S-alkylated and acetylated, and equal amounts are then mixed and subjected to N-terminal COFRADIC In the setup depicted in Fig 2, neo-N-termini generated by the ret-roviral protease are expected in the ‘light proteome’ and will only be present in the light 12C6-arginine form Almost identical numbers of pre-existing N-ter-mini (i.e the N-terN-ter-mini of intact proteins), on the other hand, should appear as couples of light and heavy labeled peptides in ratios close to 1 : 1 This is illustrated by taking b-actin as an example: its acety-lated N-terminal peptide is present in a near 1 : 1 ratio (Fig 2B), whereas a second, now trideutero-acetylated peptide is only present in the light proteome (Fig 2C) Following MS⁄ MS analysis (Fig 2D), the latter peptide is identified as TEAPLNPKANR(106-116) and constitutes a neo-N-terminus indicative of HIV-1-mediated protein processing Processing of b-actin by the HIV-1 protease between Leu105 and Thr106 was already identified in previous studies [51], thereby vali-dating our findings

Table 1 Overview of the different COFRADIC procedures that have been developed The type of peptide, the sorting agent used in between the two consecutive RP-HPLC separation steps and the type of evoked shift are indicated References to our original papers, in which full technical details can be found, are given.

acid-modified cysteine

free a-amine peptides

Hydrophobic (internal peptides)

[33]

hydrophobic

[39]

A similar approach but now with postmetabolic,

trypsin-mediated 18O-labeling [52] was used to

charac-terize in vivo protein processing in Fas-induced

apopto-tic Jurkat cells [34] In this study, 93 cleavage sites in

71 different proteins were characterized in a ‘proteomic

background’ of more than 1800 proteins At the time

of reporting these results, the overall majority of the identified cleavage sites were uncharacterized An anal-ogous setup was used for an in vitro analysis of the substrates of the HrtA2⁄ Omi protease [36] In that

human Jurkat T-cells

SILAC medium

12C6-arginine

human Jurkat T-cells SILAC medium

13C6-arginine

freeze-thaw lysate

recombinant HIV-1 protease

A

B

D

C

control

combine N-terminal COFRADIC analysis

949.80 950.06

630.43; y5

744.46; y6

954.66; y8

347.24; b3 444.41; b4

557.42; b5

671.37; b6

1012.66; b9

Fig 2 HIV-1 protease processes b-actin in vitro at Leu105 The experimental route is sketched in (A) Following N-terminal COFRADIC, two different peptides from b-actin were identified Its N-terminal peptide, DDDIAALVVDNGSGMCKAGFAGDDAPR(2–28) (N-terminus acetylated, lysine trideuteroacetylated, methionine oxidized and cysteine carbamidomethylated) was present in both proteome digests [ion trap MS spectrum of triply charged precursor in (B)], whereas a second peptide was only present in the proteome treated with the HIV-1 protease [ion trap MS spectrum of doubly charged precursor in (C)] Following MS ⁄ MS analysis [(D), b and y fragment ions indicated), this peptide was identified as TEAPLNPKANR(106–116) (N-terminus and lysine were both trideuteroacetylated), pointing to a previously characterized cleavage site of the HIV-1 protease in b-actin].

study, we identified 50 different cleavage events in

15 human proteins, and further validated these Omi

substrates by nonproteomic methods Finally, as our

method directly points to the actual site of proteolytic

cleavage, data interpretation and the design of

follow-up analyses are straightforward

COFRADIC-based sorting of

N-glycosylated peptides

Glycosylation of asparagines in the Asn-Xaa-Ser⁄ Thr

acceptor motif [53] is a widespread protein

modifica-tion: a survey in the UniProtKB⁄ Swiss-Prot database

(release 52.2, 3 April 2007) indicates that 3694 human

protein entries (i.e about 23% of all human protein

entries) have at least one feature key pointing to an

N-glycosylation event

Different methods have been used to isolate and

identify N-glycosylated proteins and characterize their

glycosylation sites In general, glycosylated proteins

and peptides are affinity-isolated or chemically trapped

prior to further analysis Affinity-based isolation of

N-glycosylated proteins is rather simple and is based

on lectin-affinity chromatography Lectins are proteins

or glycoproteins that recognize oligosaccharides but

generally favor certain classes of oligosaccharides [54]

Thus, to increase the overall coverage of

N-glycosylat-ed proteins, several lectins were combinN-glycosylat-ed in

multilec-tin affinity chromatography [55,56] Alternatively, the

lectins’ glycan bias was exploited in a serial lectin

approach separating N-glycosylated (concanavalin A)

from O-glycosylated peptides (Jacalin) [57] Chemical

trapping and release of N-glycosylated peptides was

introduced by the group of Aebersold in 2003 [25] In

their approach, aldehydes are first introduced into the

glycan by periodate oxidation These aldehydes then

covalently bind to immobilized hydrazide groups by

which glycosylated proteins are retained and all

non-glycosylated proteins are removed Immobilized

gly-cosylated proteins are then further trimmed by trypsin

such that only tryptic peptides carrying glycans

remained fixed Such peptides are finally recovered by

peptide N-glycosidase F (PNGaseF), which efficiently

removes N-glycans from conjugated asparagines while

converting these to aspartic acids [58] The potential of

this chemical trapping approach is evident from recent

studies [59–62]; however, it requires several chemical

and enzymatic modification steps, and it is therefore

more complex than lectin-affinity methods; this could

potentially obstruct its widespread introduction in

proteomics laboratories

We recently showed that N-glycosylated peptides can

be isolated by diagonal chromatography [39] In our

approach, a protein mixture containing N-glycosylated proteins is digested with trypsin, and the resulting pep-tide mixture is separated by RP-HPLC N-glycosylated peptides are then specifically targeted by PNGaseF and thus deglycosylated (COFRADIC sorting step) When separated a second time by RP-HPLC, deglycosylated peptides shift out of the primary interval of nongly-cosylated peptides and are thereby isolated Impor-tantly, the shift evoked in this way can be both hydrophilic and hydrophobic, reflecting the nature of the glycan Indeed, N-glycans can contain negatively charged sugars such sialic acid [63] and sulfated carbo-hydrates [64], and removing such glycans with PNG-aseF evokes a hydrophobic shift analogous to that observed for dephosphorylated peptides [37] Following

MS⁄ MS analysis, former N-glycosylated asparagines in the Asn-Xaa-Ser⁄ Thr motif are deamidated to aspartic acids This mass signature in the consensus N-glycosyl-ation motif is used to distinguish deglycosylated pep-tides from artificially deamidated peppep-tides, especially in Asn-Gly and Asn-Ser motifs [65], undergoing small hydrophilic shifts [39]

Our COFRADIC procedure was applied to a trypsin digest of 10 lL of mouse serum depleted for its three most abundant proteins (albumin, IgGs and transfer-rin), and resulted in the characterization of 127 differ-ent N-glycosylation sites (comprising 10 novel sites) in

82 proteins estimated to span a concentration range of

at least five orders of magnitude [39] Several N-glyco-sylation sites of the large subunit of mouse carboxy-peptidase N (UniProtKB⁄ Swiss-Prot entry Q9DBB9) were identified in this study (Table 2) This protein binds to the catalytic subunit of carboxypeptidase N, which functions in protecting organisms from circulat-ing vasoactive and inflammatory peptides containcirculat-ing C-terminal arginine or lysine [66] The large subunit of this complex binds and stabilizes the catalytic subunit and thereby keeps the complex in circulation In silico predictions indicate that this protein potentially has nine different targets for N-glycosylation, six of which were identified in our study: the asparagines at posi-tions 74, 111, 119, 348, 359 and 367 (Table 2) Three other asparagines at positions 266, 311 and 520 were missed and, as is evident from the annotations in the UniProtKB⁄ Swiss-Prot database, have hitherto not been experimentally characterized A closer look at the sequences of the tryptic peptides harboring the poten-tial glycosylation sites at positions 266 and 311 clearly indicates that these peptides are very large (66 and 36 amino acids long, respectively) Therefore, they could have been missed either because they are insoluble or because our mass spectrometers, which have an empiri-cal upper mass limit close to 3000 Da for producing

MS⁄ MS spectra that are unambiguously identified by

mascot [67], could not detect them One obvious way

to overcome this is by using proteases with nontryptic

specificities such as proteinase K that generally

pro-duce smaller peptides [68] and thus increase the chance

that more glycopeptides will be finally identified

However, such protease digests sharply augment the

complexity of the analyte mixture

Activity-based proteome-wide profiling

of purine-binding proteins

In order to assign functions to the many

uncharacter-ized (hypothetical) proteins that genome sequencing

projects provide, several small compounds occupying

and modifying active sites of enzymes have recently

found their way into functional proteomics [69,70]

Quite a lot of these so-called activity-based probes

(ABPs) are natural protein-reactive products or

syn-thetic analogs [71] ABPs in functional proteome

stud-ies generally consist of four parts: a reactive group

targeting amino acids within the enzyme’s binding

pocket, a structural moiety that is recognized by this

binding pocket, a linker, and a tag for visualization

and⁄ or isolation of modified proteins [72] Different

classes of enzymes have already been studied using

activity-based proteomics Examples include

biotinyl-ated fluorophosphonates for monitoring serine

hydrolases [73], and biotinylated

a-bromobenzyl-phosphonates for detecting protein tyrosine

phosphata-ses [74]

ATP, ADP and AMP are important sensors for the

energy status of cells, interacting with and thereby

reg-ulating the activities of key enzymes in cellular

metab-olism In addition, ATP and, to a lesser extent, GTP

are known as carriers of high-energy phosphoryl

groups that can be covalently linked to proteins and

metabolites Here, kinases play a pivotal role and

transiently interact with triphospho derivatives before the b–c phosphodiester bond is cleaved To character-ize ATP-binding and GTP-binding proteins in cells,

we profiled purine-binding proteins on a proteome scale [40] For this purpose, we used 5¢-p-fluoro-sulfonylbenzoyladenosine (FSBA) (Fig 3B), a known reactive homolog of ATP (Fig 3A) that binds proteins

in their nucleotide-binding region and then covalently modifies nucleophilic amino acids (especially tyrosine and lysine) in its proximity [75] In the past, FSBA was mainly used to profile the ATP-binding features of selected, individual proteins However, in 2004, Moore

et al published a study in which FSBA and 5¢-p-fluoro-sulfonylbenzoylguanosine (FSBG) were used to profile ATP-binding and GTP-binding proteins, respectively,

in the proteomes of different lymphoid cells [76] In their approach, proteins were labeled with FSBA or FSBG in cell extracts, separated by 2D PAGE and electrotransferred onto a poly(vinylidene difuoride) membrane Subsequent treatment of sulfobenzoyl adenosine⁄ sulfobenzoyl guanosine (SBA ⁄ SBG)-labeled proteins with NaOH hydrolyzed the ester bond between the adenosine or guanosine and the sulfo-benzoyl (SB) group and exposed the latter Antibodies

to SB were then used to immunodetect FSBA-targeted

or FSBG-targeted proteins Overlaying an image of the immunoblot with the 2D pattern of silver-stained proteins pointed to candidate ATP-binding or GTP-binding proteins that were selected from the 2D gel and identified by MS In this way, 12 different proteins could be identified as FSBA-labeled proteins

Given the fact that a mild alkaline treatment as used

by Moore et al [76] hydrolyzes the rather unstable benzoate ester bond between the adenosine and the SB group, we recently developed a COFRADIC protocol sorting for SBA-labeled peptides [40] The central sort-ing reaction is shown in Fig 3C and consists of a

25 min incubation of SBA-labeled peptides in 50 mm

Table 2 N-glycosylation sites in the large subunit of mouse carboxypeptidase N Both N-glycosylation sites characterized in our study [39] and those that were missed are given Known or potential glycosylation sites are in bold type [M + H]+, mass of the singly protonated peptide ion.

Characterized N-glycosylation sites

Unidentified N-glycosylation sites

NaOH We first tested our protocol on a tryptic digest

of SBA-labeled recombinant Chinese hamster a-protein

kinase A, and noted that removing the adenosine group

in between the two COFRADIC separations resulted

in a strong hydrophilic shift of only one peptide that

was identified as HKETGNHYAMK*ILDK(62–76)

(K* indicates the SB-labeled lysine) This peptide

har-bors the site known to be involved in the catalytic

transfer of c-phosphate from ATP to protein kinase A

substrates [77]

When this sorting procedure was applied to a whole

proteome) here, a human Jurkat T-cell proteome

depleted of small compounds like ATP and GTP,

which could compete with FSBA) 185 sites in 132

proteins were identified Clearly, this is a significantly

higher number of proteins than were detected in

the previous gel-based study [76] Therefore, our

COFRADIC technique allows the functional

interpre-tation of a larger part of a sampled proteome More

importantly, our approach directly points to the actual

site that was modified and might thereby aid in

inter-preting structural features of ATP-binding proteins

As expected, the majority of FSBA-labeled Jurkat

proteins were known binders of small nucleotides,

cofactors, or DNA and RNA molecules However,

several proteins and sites were not readily explained by

the known affinity of FSBA for purine-binding

pock-ets Closer inspection revealed that at least 23 of such

unexplainable sites were previously characterized as

tyrosine phosphorylation sites Therefore, we assume

that when FSBA is recognized by an ATP-binding site,

there are two options for SBA labeling: either the

fluorosulfonyl group reacts with a target side chain

located on the protein carrying the ATP-binding site (homo-reaction), or, through lack of a suitable reac-tion partner, it may react with a side chain present on proteins that interact with the protein carrying the actual ATP-binding site (crossover reaction) An illus-tration of the second case is observed for kinases that can transfer the SBA group onto their substrate pro-teins by a pseudocatalytic mechanism In this way, the SBA group can be linked to proteins that have no ATP-binding site We have verified this assumption by incubating a Src substrate peptide with FSBA in the presence or absence of Src; it was shown that Src

‘catalyzed’ the labeling of the substrate peptides by a factor of more than 20 [40] Hence, we concluded that care must be taken when interpreting the results of activity-based proteome studies, as not all identified proteins will actually carry out the function that was assessed by the used ABP

Conclusions

As compared to other gel-free proteomics techniques [72], COFRADIC has a number of unique properties

As COFRADIC is essentially a peptide-sorting tech-nology by which only a set of peptides representative

of the proteomic problem is withdrawn from the com-plex analyte mixture, the sample-to-sample reproduc-ibility is much higher than in shotgun approaches [78] For instance, although MudPIT uses a powerful chromatographic technology combining two basic sep-aration principles (peptide net charge and hydro-phobicity), peptide separation still takes place on the entire, complex mixture COFRADIC eliminates a

O

HO OH

O

O

S

O

O

PEPTIDE

A

C

B

N N

NH 2

OH

O

S O

O PEPTIDE

N N

NH 2

O

HO OH

O P

O

P

O

P

O O O O O O

O

N N

O

HO OH

O

O

S O

O F

N N

NH 2

25 min @ 25°C

Fig 3 FSBA COFRADIC The structures of ATP and its reactive homolog FSBA are shown in (A) and (B), respectively The COFRADIC reaction sorting for SBA-labeled peptides is shown in (C).

large number of peptides that are irrelevant to the

biological problem under consideration, thereby

reduc-ing the complexity of the problem without losreduc-ing

much information Unlike targeted peptide-centric

approaches such as isotope-coded affinity tag [23],

COFRADIC is not based on affinity procedures,

which are limited at two levels: first, the chemistries

used to convert sets of peptides into affinity probes;

and second, the limitations of mass transfer that are

inherently to liquid–solid state chemistries [79] At the

first level, COFRADIC has a fundamental advantage

because its chemistries do not need to create different

affinity labels For instance, an affinity tag specific for

methionyl peptides is extremely difficult to establish; in

contrast, a simple oxidation step by hydrogen peroxide

will specifically produce methionyl-sulfoxide derivatives

showing significant hydrophilic shifts in diagonal

reverse-phase chromatography [32] At the second

level, affinity-based experiments [23,27,48] have

limita-tions either at the level of incomplete or variable

incor-poration of the tag (for example, linking a biotinyl

group to a specific set of peptides can be incomplete

and partly unspecific) or at the level of interactions of

tagged peptides with the affinity resin, where the

high-est affinities and avidities are not always reached In

contrast, using COFRADIC, we select subsets of

pep-tides related to the biological question under

investiga-tion For instance, for the study of the oxidation of

protein methionines during oxidative stress, cells can

be differentially labeled with [13C]methionine or

[12C]methionine With COFRADIC, we sort for

methi-onine-containing peptides only: thus, we select out of

the mixture only those peptides containing the

differ-ential information, while all other peptides, which are

of no relevance, are discarded

All kinds of peptide selections can be done

with-out, each time, modifying or adapting the sorting

apparatus itself The latter is, in principle, an

mated HPLC apparatus equipped with an

auto-sampler, and can be purchased from a variety of

companies; and, at least in our hands, HPLC solvent

gradients and flow rates can nowadays be controlled

such that the overall reproducibility of HPLC runs

is very high, allowing efficient peptide sorting by

COFRADIC The only parameter that needs

chang-ing is the nature of the COFRADIC sortchang-ing reaction,

which can be chemical or enzymatic, but should

under all circumstances be highly specific and

prefer-entially quantitative Together with a sufficiently large

chromatographic shift (to segregate altered and

unal-tered peptides to the highest degree), the specificity

and quantitative nature of the COFRADIC sorting

reaction are clearly crucial for efficient peptide

sort-ing Unspecific sorting reactions and only slight alter-ations in peptide column retention will yield ‘impure’ sorted peptides, whereas nonquantitative sorting reac-tions will lead to irreproducible peptide sorting In Table 1, the chemical or enzymatic modification reactions that have been used successfully in a COFRADIC-based approach are listed They cover sorting methods varying from modifications to spe-cific side chains, such as cysteinyl [80] and methionyl [32] moieties, to post-translational modifications by phosphatase [37] and PNGaseF [39] treatments In another application, peptides located at the N-termini

of proteins or of their fragments are sorted [33] In this way, we have successfully analyzed protein pro-cessing in highly complex proteomes by the target proteins and identified the exact cleavage site(s), cre-ating the basis for fundamental protease degradomics [34,36]

As mentioned above, it is also possible to set up spe-cific covalent interactions between proteins and small molecules such as drugs or mimetic molecules of natu-ral metabolites such that chromatographic shifts can

be evoked, thus allowing sorting of the conjugated peptides by COFRADIC The example shown relates

to a study with an ATP analog [40]; however, it can,

in principle, be extended to drugs that covalently inter-act with their target protein, either directly or after being metabolized in the tissue or organism to form reactive products

One of the drawbacks of COFRADIC relates to the segmentation of the peptide separation flow during the primary run: many peptides may end up in two con-secutive fractions for their secondary analyses When the same separation is repeated a second time or fur-ther times, peptides eluting at the boundaries of the primary selected time intervals may show a slight drift, thus ending up in different secondary fractions By the same effect, peptides that are differentially labeled by isotopes, such as hydrogen and deuterium, the deriva-tives of which display slightly different chromato-graphic properties, may be artificially enriched in certain fractions Such situations could impose a hin-drance on any type of quantitative differential analyses performed with COFRADIC In addition, if eluting peptide peaks are cut, the remaining material is diluted, imposing limitations on the overall sensitivity This sensitivity issue remains one of the limitations of the technology; however, it is very well compensated

by the efficiency of the other steps in the system: lim-ited losses during consecutive chromatographic separa-tions, more freedom in the selection of efficient reaction systems and better reaction conditions, which are performed in a homogeneous phase, and, finally,

the fact that at the end only a selection of peptides is

presented for analysis to mass spectrometers The

sta-bility of the peptide elution profile, particularly in the

primary runs of COFRADIC approaches, has not

been found to be a big problem as long as the buffer

conditions in which the sample has been prepared are

kept constant

One general drawback of signature peptides for

characterizing protein modifications is the fact that

only those peptides that can be separated by

RP-HPLC, ionize well in mass spectrometers and yield

informative MS⁄ MS spectra can be identified An

interesting, recent development is top-down protein

sequencing, which enables researchers to focus on an

increasing set of protein modifications [81,82] Such

top-down techniques focus on complete proteins, allow

detection of normally labile protein modifications, and

avoid several problems associated with signature

pep-tides (see above) However, proteins of interest need to

be rather pure (the number of contaminating proteins

should be low), which may currently hinder the routine

applicability of such approaches

This review has shown that the COFRADIC

tech-nology is extremely versatile and flexible and provides

profound insights into biological questions, often

much more than what could be obtained by

alterna-tive proteomics procedures such as 2D gels or shotgun

proteomics Its strong point is its high flexibility in

selecting specific chemistries or enzymatic

modifica-tions oriented towards the biological question(s) under

consideration It should be clear from the supporting

concepts that the repertoire of applications can only

be expected to grow in the future through the

develop-ment of specific chemical or enzymatic sorting

reac-tions that alter the chemical nature of a predetermined

set of peptides

Acknowledgements

F Impens is a research assistant of the Fund for

Scientific Research) Flanders (Belgium) The work in

this paper was supported by research grants from the

Fund for Scientific Research) Flanders (Belgium)

(project number G.0280.07) and the Inter University

Attraction Poles (IAP-Phase VI)

References

1 O’Farrell PH (1975) High resolution two-dimensional

electrophoresis of proteins J Biol Chem 250, 4007–4021

2 Vandekerckhove J, Bauw G, Puype M, Van Damme J

& Van Montagu M (1985) Protein-blotting on

Poly-brene-coated glass-fiber sheets A basis for acid

hydro-lysis and gas-phase sequencing of picomole quantities

of protein previously separated on sodium dodecyl sulfate⁄ polyacrylamide gel Eur J Biochem 152, 9–19

3 Aebersold RH, Teplow DB, Hood LE & Kent SB (1986) Electroblotting onto activated glass High effi-ciency preparation of proteins from analytical sodium dodecyl sulfate-polyacrylamide gels for direct sequence analysis J Biol Chem 261, 4229–4238

4 Bauw G, De Loose M, Inze D, Van Montagu M & Vandekerckhove J (1987) Alterations in the phenotype

of plant cells studied by NH(2)-terminal amino acid-sequence analysis of proteins electroblotted from two-dimensional gel-separated total extracts Proc Natl Acad Sci USA 84, 4806–4810

5 Rosenfeld J, Capdevielle J, Guillemot JC & Ferrara P (1992) In-gel digestion of proteins for internal sequence analysis after one- or two-dimensional gel electrophore-sis Anal Biochem 203, 173–179

6 Henzel WJ, Billeci TM, Stults JT, Wong SC, Grimley C

& Watanabe C (1993) Identifying proteins from two-dimensional gels by molecular mass searching of peptide fragments in protein sequence databases Proc Natl Acad Sci USA 90, 5011–5015

7 James P, Quadroni M, Carafoli E & Gonnet G (1993) Protein identification by mass profile fingerprinting Biochem Biophys Res Commun 195, 58–64

8 Mann M, Hojrup P & Roepstorff P (1993) Use of mass spectrometric molecular weight information to identify proteins in sequence databases Biol Mass Spectrometry

22, 338–345

9 Pappin DJ, Hojrup P & Bleasby AJ (1993) Rapid iden-tification of proteins by peptide-mass fingerprinting Curr Biol 3, 327–332

10 Yates JR 3rd, Speicher S, Griffin PR & Hunkapiller T (1993) Peptide mass maps: a highly informative approach to protein identification Anal Biochem 214, 397–408

11 Challapalli KK, Zabel C, Schuchhardt J, Kaindl AM, Klose J & Herzel H (2004) High reproducibility of large-gel two-dimensional electrophoresis Electrophore-sis 25, 3040–3047

12 Klose J (1999) Large-gel 2-D electrophoresis Methods Mol Biol 112, 147–172

13 Sadygov RG, Cociorva D, Yates JR 3rd (2004) Large-scale database searching using tandem mass spectra: looking up the answer in the back of the book Nat Methods 1, 195–202

14 Kelleher NL, Lin HY, Valaskovic GA, Aaserud DJ, Fridriksson EK & McLafferty FW (1999) Top down versus bottom up protein characterization by tandem high-resolution mass spectrometry J Am Chem Soc 121, 806–812

15 Han X, Jin M, Breuker K & McLafferty FW (2006) Extending top-down mass spectrometry to proteins with