Active-site Peptide “Fingerprinting” of Glycosidases in Complex Mixtures by Mass Spectrometry DISCOVERY OF A NOVEL RETAINING (cid:1) -1,4-GLYCANASE IN CELLULOMONAS FIMI *

New proteomics methods are required for targeting and identification of subsets of a proteome in an activity-based fashion. Here, we report the first gel-free, mass spectrometry-based strategy for mechanism-based profiling of retaining (cid:1) -endoglycosidases in complex proteomes. Using a biotinylated, cleavable 2-deoxy-2-flu-oroxylobioside inactivator, we have isolated and identified the active-site peptides of target retaining (cid:1) -1,4-glycanases in systems of increasing complexity: pure enzymes, artificial proteomes, and the secreted proteome of the aerobic mesophilic soil bacterium Cellulomonas fimi . The active-site peptide of a new C. fimi (cid:1) -1,4-gly-canase was identified in this manner, and the peptide sequence, which includes the catalytic nucleophile, is highly conserved among glycosidase family 10 members. The glycanase gene (GenBank TM accession number DQ146941) was cloned using inverse PCR techniques, and the protein was found to comprise a catalytic domain that shares (cid:1) 70% sequence identity with those of xylanases from Streptomyces sp.andafamily2bcarbohydrate-bindingmodule.The new glycanase hydrolyzes natural and artificial

With the completion of the genome sequences of many organisms, the field of proteomics faces several major tasks. One challenge is that of identification and assignment of structure/function to the tens of thousands of proteins encoded by prokaryotic and eukaryotic genomes. Another challenge is accurate quantitative analysis of changes in protein levels/activities that occur within a proteome as a response to biological perturbations that are due either to normal developmental and metabolic changes or to abnormalities associated with disease. Proteomic techniques such as comparative two-dimensional gel electrophoresis coupled with mass spectrometry, the isotope-coded affinity tagging approach (1), and variations of isotope-coded affinity tagging that identify sites of modification (e.g. glycosylation and phosphorylation) on proteins (2)(3)(4) fail to provide a direct assessment of protein function.
Recently, several chemical strategies for activity-based protein profiling (ABPP) 4 in complex proteomes that target several enzyme groups, including oxidoreductases (5,6), serine hydrolases (7,8), cysteine proteases (9,10), threonine proteases (11), metalloproteases (12), protein phosphatases (13), kinases (14), and exoglycosidases (15), have been employed. ABPP probes have two general features: 1) an active sitedirected (mechanism-based) inactivator or affinity label that reacts with a catalytic residue and forms a covalent adduct with the target enzyme(s) and 2) one or more reporter groups that enable rapid detection (e.g. a fluorophore) and/or affinity isolation (e.g. biotin) (16). As such, ABPP methods can provide direct information on post-translational forms of protein regulation (17). However, most of the ABPP research has been limited to detection of the target enzymes, usually by Western blotting or in-gel fluorescence. Recently, a gel-free, mass spectrometry-based ABPP (MS-ABPP) method was used in which the labeled active-site peptides of target enzymes were affinity-isolated from the proteolytic digest of the proteome and analyzed by liquid chromatography/mass spectrometry (LC/MS) and tandem mass spectrometry (MS/MS) procedures to determine their amino acid sequences and the labeled catalytic residue (6). However, the ABPP probe used in that attempt targeted several mechanistically distinct enzyme classes in a non-directed fashion (6). The MS-ABPP approach relies on the assumption that the active-site peptide, containing the labeled catalytic residue, is highly conserved among members of the same enzyme family and therefore provides sufficient information to decipher the enzyme family to which the target enzyme belongs (18).
We have demonstrated this MS-ABPP methodology at three levels of increasing complexity. First, as a test of the strategy, we used this approach to isolate and identify the labeled active-site peptides of an endoxylanase (Bcx from Bacillus circulans) and a mixed-function endoxylanase/cellulase (Cex from Cellulomonas fimi) from their peptic digests. As retaining endoxylanases, Bcx (GH family 11) and Cex (GH family 10) were ideal candidates for targeting by our probe and had been investigated extensively (21). Second, we applied the method to several model enzyme mixtures (artificial proteomes) including Bcx, Cex, or both as targets. Finally, we applied the method to a complex biological proteome, viz. the secreted proteome of the aerobic mesophilic xylanolytic/cellulolytic soil bacterium C. fimi. C. fimi secretes a complex array of xylanases and cellulases concurrently (25,26); however, only four C. fimi retaining ␤-1,4-glycanases have been characterized to date (see TABLE ONE) (21,25,(27)(28)(29)(30). Our MS-ABPP analysis, gene cloning, and kinetic characterization of the enzyme revealed that C. fimi secretes a previously unidentified family 10 glycanase with a carbohydrate-binding module (CBM) from CBM family 2b.  (35). The substrates 2,5-dinitrophenyl ␤-xylobioside (DNPX 2 ) and 2,4-dinitrophenyl ␤-cellobioside (DNPC) were prepared as described previously (20,30). Edman peptide sequencing, DNA sequencing, and synthesis of PCR primers were performed by the University of British Columbia Nucleic Acid and Protein Service Unit. The buffers used were as follows: buffer A, 0.020 M MES, 0.050 M NaCl (pH 6.0), and 0.1% bovine serum albumin (BSA); buffer B, 0.050 M NaH 2 PO 4 / Na 2 HPO 4 (pH 7.0) and 0.1% BSA; buffer C, 0.050 M NaH 2 PO 4 /H 3 PO 4 (pH 2.0); buffer D, 0.060 M Tris (pH 8.0); buffer E, 0.050 M NaH 2 PO 4 / Na 2 HPO 4 (pH 6.0); and buffer F, 0.050 M NaH 2 PO 4 /Na 2 HPO 4 (pH 6.5) and 0.1% BSA. Pepsin was from Roche Applied Science (Mannheim, Germany). Immunopure immobilized streptavidin and tris(2-carboxyethyl)phosphine (TCEP) were from Pierce. Pwo polymerase was from Roche Applied Science (Laval, Quebec, Canada). T4 polynucleotide kinase was from Fermentas (Burlington, Ontario). All other chemicals were obtained from Sigma and were reagent-grade. Synthesis of the ABPP probe DNP2FX 2 SSB will be reported elsewhere. 5 Inactivation Kinetics-All kinetic studies were performed using a Unicam 8700 UV-visible spectrophotometer equipped with a circulating water bath. The time-dependent inactivation of each enzyme or proteome by DNP2FX 2 SSB was monitored by measuring the residual xylanase/cellulase activity over time. This was accomplished by the addition of an aliquot of the inactivation mixture at appropriate time intervals to a solution of the substrate (DNPX 2 or DNPC) in the appropriate buffer. The initial rates were then measured under steady-state conditions by spectrophotometric monitoring of the release of the 2,4dinitrophenol(ate) or 2,5-dinitrophenol(ate) group at a wavelength of either 440 nm (DNPX 2 ) or 400 nm (DNPC). Pseudo first-order rate constants (k obs ϭ k i [I]/(K i ϩ [I])) for the decay of activity were determined in each case from fitting the decay curve to a single exponential decay equation using nonlinear regression with the program GraFit. For Bcx, the inactivation mixture contained 0.20 mg/ml Bcx and 0.500 mM DNP2FX 2 SSB in BSA-free buffer A. Both inactivation and control samples were incubated at 40°C, and aliquots were taken at appropriate time intervals and added to a cuvette containing DNPX 2 substrate so that the final assay contained 2.8 g/ml Bcx (active ϩ inactive) and 2.46 mM DNPX 2 in buffer A at 40°C. For Cex, the inactivation mixture contained 57 g/ml Cex and 22.9 M DNP2FX 2 SSB in BSA-free buffer B. Both inactivation and control samples were incubated at 37°C, and aliquots were taken at appropriate time intervals and added to a cuvette containing DNPC substrate so that the final assay contained 0.81 g/ml Cex (active ϩ inactive) and 0.493 mM DNPC in buffer B at 37°C. For the C. fimi secreted proteome, the inactivation mixture contained 5.5 mg/ml protein (from C. fimi culture medium) and 0.100 mM DNP2FX 2 SSB in BSA-free buffer B. Both inactivation and control samples were incubated at 37°C, and aliquots were taken at appropriate time intervals and added to the cuvette containing either DNPX 2 or DNPC substrate so that the final assay contained 0.11 mg/ml protein and 1.00 mM DNPX 2 or 2.00 mM DNPC in buffer B at 37°C. For enzyme inactivation prior to proteolysis for affinity isolation, the desired amount of enzyme or proteome was incubated in the presence of 0.100 -0.500 mM DNP2FX 2 SSB for 2-6 h at the temperatures and conditions mentioned above for each enzyme or proteome. Where necessary, samples were dialyzed at room temperature using a non-cellulose 5-kDa cutoff membrane centrifugal filter unit (Millipore Corp.) to exchange buffer A or B with buffer C and to remove excess inactivator. If no dialysis was necessary, enough buffer C was added to the sample to bring it to pH 2.0.

EXPERIMENTAL PROCEDURES
Proteolysis-For Bcx, a solution of inactivated Bcx in buffer C was first heat-denatured (boiling for 2 min) and then proteolyzed in the presence of pepsin (50:1 mass ratio) at room temperature for 3 h. For Cex, a solution of inactivated Cex in buffer C was proteolyzed in the presence of pepsin (100:1 mass ratio) at room temperature for 6 h. For model proteomes A-C, a solution of the inactivated enzyme mixture in buffer C was first heat-denatured (boiling for 2 min) and then proteolyzed in the presence of pepsin (50:1 mass ratio) at room temperature for 6 -8 h. For the C. fimi secreted proteome, a sample of the inactivated C. fimi secreted proteome in buffer C was first heat-denatured (boiling for 2 min) and then proteolyzed in the presence of pepsin (50:1 mass ratio) at room temperature for 8 -10 h. In all of the above cases, the proteolytic digest was either immediately subjected to the affinity isolation procedure or freeze-dried first.
Affinity Isolation of the Labeled Peptides-The peptic digest of the inactivated enzyme or proteome in buffer C was incubated with streptavidin resin at a ratio of 1.0 ml of streptavidin resin/60 nmol of biotinequivalent at room temperature for 1-2 h with shaking. The unlabeled peptides were washed away in buffer C using a spin filter (Novagen). The labeled peptides bound to the streptavidin resin were incubated with 30 mM TCEP in buffer D for 40 min at room temperature with shaking and eluted using a spin filter. The reduction process was repeated several times. The eluted peptide samples were then freezedried, redissolved in H 2 O, and analyzed by electrospray ionization (ESI) MS.
LC/ESI-MS and ESI-MS/MS Analyses-ESI-MS spectra were recorded on a PE-Sciex API 300 triple quadrupole mass spectrometer (Sciex, Thornhill, Ontario) equipped with an ion spray ion source. The ion source voltage was 5 kV, and the orifice voltage was 50 V. All spectra were obtained in the positive-ion single quadrupole scan mode (LC/ MS) or the tandem MS fragment ion scan mode (MS/MS). In the LC/MS mode, the quadrupole mass analyzer was scanned over an m/z range of 300 -2000 atomic mass units with a step size of 0.5 Da and a dwell time of 1.0 ms/step. In the MS/MS mode, the spectra were obtained by selectively introducing the precursor peptide with the m/z value of interest from the first quadrupole (Q1) into the collision cell (Q2) and observing the fragment ions in the third quadrupole (Q3). Q3 scan range was 50 -1900 atomic mass units with the same step size and dwell time as above. Protein or peptide samples were introduced by reverse-phase HPLC on an Ultrafast microprotein analyzer (Michrom Bioresources, Inc., Auburn, CA) directly interfaced with the mass spectrometer. For protein analysis, the sample was loaded onto a PLRP-S C 4 column (1.0 ϫ 50 mm; Michrom Bioresources Inc.) and eluted with a linear gradient of 5-90% solvent B over 3 min, followed by 90% solvent B over 5 min at a flow rate of 50 l/min (solvent A, 0.05% trifluoroacetic acid and 2% CH 3 CN in water; and solvent B, 0.045% trifluoroacetic acid and 80% CH 3 CN in water). For peptide analysis, the sample was loaded onto a Jupiter C 18 column (1.0 ϫ 150 mm; Phenomenex Inc., Torrance, CA) and eluted with a linear gradient of 0 -60% solvent B over 60 min at a flow rate of 50 l/min. C. fimi Culture Growth and Secreted Protein Preparation-C. fimi cells (ATCC 484) were grown in minimal medium containing 0.1% xylan (from birchwood) as the inducer at 30°C for 5 days. Cells were then removed by centrifugation at 4°C, and the culture supernatant was concentrated and exchanged with buffer E at 4°C under N 2 pressure in an Amicon ultrafiltration cell (Millipore, Beverly, MA) using a noncellulose ultrafiltration membrane disc (Pall Life Sciences, Ann Arbor, MI). The secreted protein preparation in buffer E was then filtered through a 0.45-m non-cellulose membrane (Millipore Corp.) and analyzed by SDS-PAGE. Protein concentration was determined using the Micro BCA TM protein assay reagent kit with BSA standards (Pierce). A measure of the xylanase and cellulase activities of the secreted protein preparation was obtained by assaying with DNPX 2 and DNPC as the substrates.
PCR Amplification of the Internal Gene Probe-The internal gene probe encoding the part of the new C. fimi glycanase (named Cfx) lying between the acid/base and nucleophile regions was PCR-amplified using 100 ng of NotI-digested C. fimi genomic DNA, 1 M each primer (Cfx-AB, 5Ј-TRS GAC GTS RTC AAC GAG-3Ј; and Cfx-Nu, 5Ј-GTC SAG CTC SGT GAT CTG SAC-3Ј), 20 mM Tris (pH 7.5), 0.2 mM each dNTP, 5% dimethyl sulfoxide, 4 mM MgSO 4 , and 2.5 units of Pwo polymerase in a 50-l reaction volume. Thirty PCR cycles (30 s at 94°C, 30 s at 60°C, and 45 s at 72°C) were performed in a thermal cycler (GeneAmp PCR System 2400, PerkinElmer Life Sciences). Both ends of the resulting PCR product were phosphorylated with T4 polynucleotide kinase, and then the DNA fragment was subcloned into SmaI-digested pUC18.
Inverse PCR-We used the inverse PCR method (iPCR) as described by Ochman et al. (36) to clone the entire sequence of the cfx gene. Genomic DNA was digested with BamHI, NotI, SacI, or XhoI, and then the self-ligated libraries were used as templates for iPCR. A total of three DNA fragments (Cfx-Sc, Cfx-Xh, and Cfx-Bm) were amplified using three primer sets as follows: a set of the iPCR-rev primer (5Ј-CCA GTC GTT GCC GGT GCG CTG CAG GTT GGA GTC-3Ј) and the iPCR-fw primer (5Ј-TGC GTC GGC TTC CAG GCG CAC TTC AAC TCC GGC-3Ј), a set of the iPCR-Sac-rev primer (5Ј-CCG AGC GGT CGC CGT CAG CCA TCT GCG AGT-3Ј) and the iPCR-rev primer, and a set of the iPCR-Bam-rev primer (5Ј-TCA GGG CGT GGT CAG GTC GAA GCT GCG GAC-3Ј) and the iPCR-Xho-fw primer (5Ј-GAC CGT CCG GTC GCC GCA GAA GAT CAT CGC-3Ј), respectively. After analysis of the sequences of the three fragments, primers for amplification of the full-length cfx gene were designed as follows: Cfx-TOP-Nc primer, 5Ј-GCG AGT GAC CAT GGC CAC GAA ACT CCA CGC GAC-3Ј; and Cfx-END-Nt primer, 5Ј-GTC ACG TGC GCG GCC GCA TGA CCC GCT GAC-3Ј. The products of PCR with these primers were subcloned into plasmid pR2TK, which has a B. subtilis ␣-amylase promoter (amyR2) from pAR2 (37) as well as multiple cloning sites, a sixhistidine tag sequence, and the T7 terminator from pET28a. The resulting plasmid was designated pR2Cfx-His 6 .
Expression of the cfx Gene in Escherichia coli-A single colony was cultured overnight in 50 ml of LB medium containing 20 g/ml kanamycin at 37°C. The pre-culture was then re-inoculated in 1 liter of medium and grown for an additional 15 h. The cell pellet was harvested and resuspended in 50 mM Tris-HCl (pH 8.0) containing 5 mM imidazole and 300 mM NaCl. The cell suspension was passed twice through a French press at 4°C and centrifuged at 10,000 ϫ g for 30 min, and the soluble extract was purified by nickel affinity chromatography. The protein was then dialyzed against 50 mM phosphate buffer (pH 7.0) and stored at 4°C. Protein concentration was determined using the Micro BCA TM protein assay reagent kit with BSA standards.
Cfx Kinetics-Initial rates of hydrolysis of p-nitrophenyl ␤-xyloside (pNPX), p-nitrophenyl ␤-glucoside (pNPG), p-nitrophenyl ␤-xylobioside (pNPX 2 ), and p-nitrophenyl ␤-cellobioside (pNPC) catalyzed by Cfx (the new xylanase from C. fimi) were determined at 37°C by monitoring the reactions spectrophotometrically at 400 nm, where ⌬⑀ 400 ϭ 7280 M Ϫ1 cm Ϫ1 (32). Assays were carried out in buffer F at 37°C, and reactions were initiated by the addition of enzyme. The substrate and enzyme concentrations in the assays were as follows: 1.0 -10 mM pNPX and 1.1 M Cfx; 1.0 -10 mM pNPG and 6.8 M Cfx; 0.02-2.0 mM pNPX 2 and 0.14 M Cfx; and 0.2-10 mM pNPC and 2.8 M Cfx. For pNPX 2 and pNPC, the plots of initial rates versus substrate concentrations were hyperbolic, and the values of K m and V max were determined by fitting the initial velocity curves to the standard Michaelis-Menten equation using nonlinear regression with the program GraFit. For pNPX and pNPG, no saturation occurred over the practical concentration ranges, and the values of V max /K m were determined by linear regression analysis of the data.
The initial rates of hydrolysis of birchwood xylan and Cm-cellulose catalyzed by Cfx were measured following the p-hydroxybenzoic acid hydrazide method (38), which quantitates the production of reducing sugar. Saturating solutions of xylan (Ͻ1% (w/v)) and Cm-cellulose (Ͻ2% (w/v)) in buffer F were prepared and preincubated at 37°C. Reactions were initiated by the addition of Cfx (0.14 M in xylan solution and 1.4 M in Cm-cellulose solution). Aliquots of 100 l were removed at appropriate time intervals and added to 900 l of 50 mM p-hydroxybenzoic acid hydrazide and 0.5 M NaOH in H 2 O in glass test tubes. Samples were then heated in a 75°C water bath for 30 min and cooled to room temperature. The absorbance at 420 nm was measured, and the amount of reducing sugar released was determined using a glucose standard curve generated under the same conditions. To ensure the validity of the assays for enzyme activity, enzyme concentrations were chosen so that the production of reducing sugar in the assay fell within the linear response range of the standard curve. Assuming that saturating solutions of xylan and Cm-cellulose were also saturating for the enzyme, the V max value was determined by linear regression analysis of the data in each case.
Cfx was found to be stable over a pH range of 4.0 -9.0 for several hours. The pH dependence of k cat /K m for Cfx was determined by the substrate depletion method. The time course of depletion of 0.1 mM pNPX 2 (ϽK m ) was followed for two or more half-lives in the presence of 0.11 M Cfx in 50 mM buffer at the appropriate pH at 37°C by monitoring the increase in A 400 with time. Reactions were initiated by the addition of Cfx. The buffers used were sodium citrate for pH 4.0 -6.0, buffer F for pH 6.5-7.8, and Tris for pH 8.0 -9.0. The depletion curve in each case was fitted to an exponential decay equation using the program GraFit, and the k cat /K m values were determined using the equation The effect of temperature on Cfx stability was determined by incubating aliquots of Cfx (0.16 mg/ml) in buffer F at 20, 30, 40, 50, 60, and 70°C for 10 h. Residual activities were then determined by measuring the initial rates as described above. The assays contained 1.0 mM pNPX 2 and 57 nM Cfx in buffer F at 37°C.

Detection of the Covalent Biotinylated Fluoroglycosyl-enzyme Intermediates and Demonstration of Reductive Cleavage of the Biotinylated Arm by ESI-MS-
The stoichiometry of full inactivation of Bcx and Cex by DNP2FX 2 SSB was determined using LC/ESI-MS analysis. Reconstructs of the ESI mass spectra of the native enzymes and those of the fully inactivated enzymes (Fig. 2B) showed that inactivation resulted in a mass increase of 885 Ϯ 2 Da, which corresponds to the molecular mass of 2FX 2 SSB. Therefore, in each case, one enzyme molecule reacted with one DNP2FX 2 SSB molecule, forming the biotinylated fluoroglycosylenzyme intermediate and releasing 2,4-dinitrophenol(ate) (Fig. 1B). Furthermore, we demonstrated that the biotinylated arm could be cleaved from the fluoroglycosyl-enzyme intermediate via reduction of the disulfide bond in the presence of the water-soluble phosphine TCEP without destroying the ester bond in the intermediate. Reconstructs of the ESI mass spectra of the inactivated enzymes that had been incubated with TCEP (Fig. 2B) showed that the reduction resulted in a mass decrease of 536 Ϯ 2 Da, which corresponds to the molecular mass of the reduced biotinylated arm (HSB). The choice of TCEP over dithiothreitol is based upon the fact that reduction by dithiothreitol is not effective at pH Ͻ7 (40) and also upon the known reaction of dithiothreitol with the ester linkage in the glycosyl-enzyme intermediate (41). TCEP was found to reduce the disulfide bond effectively at 5 Ͻ pH Ͻ 7 (42), in which range the fluoroglycosyl-enzyme intermediate has a longer lifetime (24) without cleavage of the ester linkage.

Detection of the Affinity-isolated, Labeled Active-site Peptides by ESI-MS and Their Identification by ESI-MS/MS-
The labeled activesite peptides of Bcx and Cex were isolated from the peptic digests of the pure labeled enzymes. The enzymes were first incubated with DNP2FX 2 SSB until fully inactivated; excess DNP2FX 2 SSB was removed; and samples were subjected to peptic proteolysis at pH 2.0. Samples were kept at acidic pH throughout the affinity isolation process  (24). The peptic digest was incubated with streptavidin resin; unbound peptides were washed away; and biotinylated, labeled activesite peptides were eluted via reduction of the disulfide bond with TCEP. The isolated peptides were then analyzed by LC/ESI-MS as shown in Fig. 2C. The m/z values of the isolated labeled peptides were exactly as expected based on previous studies with the non-biotinylated label (39,43). In the case of Cex, two additional fragments were also observed, most likely due to orifice collisions (Fig. 2C). The sequences of the isolated peptides were unambiguously confirmed by MS/MS fragmentation analysis (Fig. 2D) to be 69 YGWTRSPLIEY 79 and 229 VRITEL 234 for Bcx and Cex, respectively, containing the catalytic nucleophiles Glu 78 (Bcx) and Glu 233 (Cex) bearing the label as the ester (39,43).
Extension to Artificial Proteomes-To confirm the specificity of the ABPP probe and to demonstrate the mechanism-based nature of the strategy, studies were performed in which one or more xylanases were included in mixtures of other sugar-modifying enzymes. Three model proteomes were constructed: model proteome A included 10 nmol each native ␤-glucosidase from Agrobacterium sp. (Abg), Bcx, maltose-6Јphosphate glucosidase from B. subtilis (GlvA), HEWL, and lipopolysaccharyl-␣-galactosyltransferase C from N. meningitidis (LgtC); model proteome B included 10 nmol each native Abg, Cex, GlvA, LgtC, and mutant E78C Bcx; and model proteome C included 10 nmol each native Abg, Bcx, Cex, GlvA, and HEWL. Each of these mixtures was reacted with DNP2FX 2 SSB and treated exactly as described for the pure enzymes above. Fig. 3 (A-C) shows the ESI mass spectra of the peptides isolated in each case. As expected, the only labeled active-site peptide species isolated from model proteomes A and B were those of Bcx and Cex, respectively (Fig. 3, A and B). Notably, the active-site peptide of Bcx was not isolated from model proteome B (Fig. 3B), which contained mutant E78C Bcx (mutated catalytic nucleophile), in agreement with the previous finding that E78C Bcx is not capable of catalyzing the glycosylation reaction, the first step in the double-displacement mechanism (35). Model proteome C demonstrated the concurrent isolation of the active-site peptides of both target enzymes (Fig. 3C).
Analysis of the C. fimi Secreted Proteome-C. fimi cells (ATCC 484) were grown for 5 days at 30°C in minimal medium containing 0.1% xylan. The culture supernatant was concentrated, and a portion of the sample was then reacted with 0.10 mM DNP2FX 2 SSB. A measure of the xylanase and cellulase activities was obtained by assaying aliquots of the inactivation mixture with DNPX 2 and DNPC, respectively. As shown in Fig. 4A, the activity with DNPX 2 decreased toward zero in a time-de-pendent fashion according to pseudo first-order kinetics with k obs ϭ (3.7 Ϯ 0.1) ϫ 10 Ϫ4 s Ϫ1 . Interestingly, the activity with DNPC also decreased substantially, but only by ϳ84%, implying that some cellulases were not inactivated (Fig. 4A). The data fit well into a single exponential decay equation with offset: (initial rate) t ϭ ⌬(initial rate)(exp Ϫkobst ) ϩ (initial rate) tϭ∞ with k obs ϭ (6.4 Ϯ 0.8) ϫ 10 Ϫ4 s Ϫ1 .
The labeled C. fimi secreted proteome was then subjected to peptic proteolysis and streptavidin-based affinity purification as performed for the pure enzymes and model proteomes. Of the isolated peptides shown in Fig. 4B, the species at m/z 1084.0, 864.8, and 731.0 (ϩ1) clearly arose from Cex because these are the same peptides seen in Fig. 2C. The labeled peptide at m/z 1055.7 (ϩ1) was a completely new species and did not match the expected labeled active-site peptides of any of the four known retaining ␤-1,4-glycanases from C. fimi (TABLE ONE). The species at m/z 1055.7 (ϩ1) was therefore subjected to sequence analyses both by MS/MS fragmentation (Fig. 4C) and by Edman degradation. These analyses identified the sequence as VQITEL, with the label attached to the Glu residue. This sequence matches the highly conserved region of the primary amino acid sequence surrounding the catalytic nucleophile in GH family 10 enzymes.
Cloning the Gene Encoding the New Retaining ␤-1,4-Glycanase (Cfx)-To obtain partial information on the gene (cfx) encoding the new retaining ␤-1,4-glycanase (Cfx), we first amplified an internal gene fragment using two primers: the Cfx-AB primer, encoding peptide WDV(V/I)NE, a conserved region around the catalytic acid/base residue of GH family 10 enzymes; and the Cfx-Nu primer, encoding peptide VQITELD (VQITEL identified by the MS-ABPP method plus a conserved Asp residue). The nucleotide sequences of both primers were chosen according to the codon usage of C. fimi based on the 16 reported open reading frames. Restriction-digested genomic DNA (rather than the whole genomic DNA) was used as a template because C. fimi genes are Ͼ70% GC-rich. Indeed, 350-bp DNA fragments were PCR-amplified successfully in four reaction mixtures containing BamHI-, SacI-, NotI-, or XhoI-digested genomic DNA as a template. The PCR product obtained from the NotI-digested genomic DNA template (a NotI site exists in the corresponding gene fragment of Cex) was re-amplified under the same PCR conditions, and the amino acid sequence (deduced from the DNA sequence) confirmed that the PCR fragment encoded part of a GH family 10 xylanase. Two regions (iPCR-1 and iPCR-2) were designed for iPCR based on the deduced amino acid sequence showing no close similarity to Cex and C. fimi xylanase C. A PCR fragment from the region upstream of the cfx gene to the middle of the cfx gene was amplified from the SacI library only. The full nucleotide sequence was obtained by two more sequential iPCRs using the second primer set (iPCR-3 and iPCR-4) and the third primer set (iPCR-5 and iPCR-6) (Fig.  5). The cfx gene sequence has been deposited in the GenBank TM Data Bank (accession number DQ146941).
Kinetic Characterization of Cfx-To investigate the catalytic properties of Cfx, recombinant Cfx was produced in E. coli and purified by nickel affinity chromatography. Cfx stability was constant for 10 h over 20 -40°C, but rapidly decreased at higher temperatures (Fig. 6A). The steady-state kinetic parameters for hydrolysis of several aryl-␤-glyco-sides were determined: pNPX, pNPX 2 , pNPG, and pNPC (TABLE  TWO). Cfx had 10 3 -10 4 -fold higher specificity constants (k cat /K m ) for the p-nitrophenyl disaccharides than for the p-nitrophenyl monosaccharides. The enzyme showed 10 2 -fold higher k cat /K m values (10-fold lower K m and 10-fold higher k cat ) for p-nitrophenyl xylobioside than for p-nitrophenyl cellobioside. The k cat values were also estimated for hydrolysis of the natural substrates birchwood xylan and Cm-cellulose (TABLE TWO). Again, Cfx had a 60-fold higher k cat value for xylan than for Cm-cellulose. The pH dependence of k cat /K m was investigated using pNPX 2 as the substrate over a pH range of 4.0 -9.0. The k cat /K m -pH profile was found to be a classical bell-shaped curve (Fig. 6B), indicating two catalytically essential ionizable groups (pK a1 ϭ 4.5 Ϯ 0.1 and pK a2 ϭ 7.6 Ϯ 0.1).  (24,39,43).

DISCUSSION
In addition to being the first MS-based ABPP strategy applied to a biological proteome for glycosidase profiling, the approach described here is superior, in terms of specificity and efficacy, to those described previously for glycosidases (15, 46 -48). A previous method that made use of quinone methide chemistry produced cross-labeling when carried out in a mixture of proteins because, after initial reaction at the active site of a glycosidase, the glycoside that provided specificity was lost, and a reactive quinone methide was generated, which is known to leave the active site and to react with exposed nucleophilic residues anywhere on the protein or indeed on any protein in the mixture (46,47). A very recent attempt at proteomic analysis of exoglycosidases using a ligation strategy was more specific, but lacked the desired efficacy (15), with the tagged substrate reacting 10 4 times more slowly than the parent substrate. By contrast, our reagent (DNP2FX 2 SSB) reacted at rates ( Fig. 2A) that are almost identical to those of the parent armless inactivator (DNP2FX 2 ) (20,39). This indicates that the biotin arm, attached to the 4Ј-carbon of the xylobioside, does not sterically hinder its access into the active site to any significant extent. This is consistent with the fact that Bcx and Cex are both endoxylanases, with their active sites located in an open cleft (31,49,50). In contrast, exo-acting enzymes have a pocket-shaped active site that will not easily accommodate modified sugars, especially those with a bulky biotin arm (15,50).  Identification of a Previously Undiscovered GH Family 10 Glycosidase in the C. fimi Secreted Proteome-The xylanolytic/cellulolytic soil bacterium C. fimi, which produces and secretes a complex array of xylanases and cellulases (25,26), is an ideal system for the discovery of new retaining ␤-1,4-glycanases. To date, four such C. fimi retaining enzymes (TABLE ONE) have been cloned and characterized (21,25,(27)(28)(29)(30). However, given the nature of the genetic cloning strategies, it is conceivable that some enzymes were missed (26).
Interestingly, the hydrolytic activity of the C. fimi secreted proteome toward DNPX 2 was completely abolished in the presence of DNP2FX 2 SSB, whereas the hydrolytic activity toward DNPC plateaued at ϳ16% of its initial value (Fig. 4A). This is consistent with the fact that all known xylanases are retainers, whereas many cellulases are inverters (25). Indeed, many of the known C. fimi cellulases are inverting glycosidases (25), which operate via a single-displacement mechanism and lack a covalent glycosyl-enzyme intermediate (22). Furthermore, cellulases are not necessarily targeted by a xylobiose-based ABPP probe. Only two major labeled active-site peptides were affinity-isolated from this system (Fig. 4B) and identified by ESI-MS/MS (Fig. 4C), indicating that two major secreted enzymes are responsible for the observed retaining ␤-1,4-glycanase activity. One of those is indeed Cex, but the other enzyme with an active-site peptide sequence of VQITEL is a new GH family 10 ␤-1,4-glycanase previously unidentified in C. fimi, as further confirmed by cloning (Fig. 5). Of the other three known but undetected ␤-1,4-glycanases (TABLE ONE), CenD has extremely low xylanase activity and is not expected to react with DNP2FX 2 SSB (27), and XylC is likely intracellular because it lacks the leader sequence typical of secreted prokaryotic proteins (28). XylD is extracellular, but it is prob-ably secreted in minute amounts (below the detection limit of the ABPP methodology) under the growth conditions employed.
Characteristics of Cfx-The deduced amino acid sequence of the Cfx catalytic domain shares ϳ70% sequence identity with those of xylanases from Streptomyces sp., 52% with Cex, 32% with C. fimi XylC, and 28% with Cellvibrio japonicus XylA. Like Cex and many other GH family 10 enzymes, Cfx has a C-terminal CBM (see afmb.cnrs-mrs.fr/ϳcazy/   CAZY). Interestingly, the CBM of Cfx displays a high degree of identity to the xylan-binding modules from CBM family 2b, which have been previously found only in family 11 xylanases. Therefore, Cfx is the first example of a GH family 10 enzyme with CBM2b.
Cfx exhibits a substrate specificity expected of GH family 10 enzymes, viz. largely xylanolytic and, to a much lesser extent, also cellulolytic. This is in contrast to GH family 11 enzymes, which are exclusively xylanolytic. As shown in TABLE TWO, comparison of the kinetic parameters of Cfx with those of Cex and Streptomyces lividans Xyl10A shows that the kinetic behavior of Cfx and S. lividans Xyl10A is very similar and somewhat different from that of Cex. This is not surprising given the high amino acid sequence identity between Cfx and xylanases from Streptomyces sp., yet another testament to the fact that high primary sequence similarity often reflects structural and functional similarities. Finally, the k cat /K m for Cfx depends on two ionizable groups (in the free enzyme) of pK a1 ϭ 4.5 and pK a2 ϭ 7.6, very similar to those of Cex (pK a1 ϭ 4.1 and pK a2 ϭ 7.7) (30).
In summary, the specific MS-ABPP strategy described here is a particularly valuable approach for the identification of new ␤-1,4-glycanases from diverse proteomes or even from metaproteomes. ␤-1,4-Glycanases are of considerable utility in the pulp and paper industry (51), as additives for the baking industry, and in biomass conversion (52). More important, the MS-ABPP approach is also applicable to a wide range of retaining endoglycosidases.