The N-terminal Nucleophile Serine of Cephalosporin Acylase Executes the Second Autoproteolytic Cleavage and Acylpeptide Hydrolysis*

Cephalosporin acylase (CA) precursor is translated as a single polypeptide chain and folds into a self-activating pre-protein. Activation requires two peptide bond cleavages that excise an internal spacer to form the mature αβ heterodimer. Using Q-TOF LC-MS, we located the second cleavage site between Glu159 and Gly160, and detected the corresponding 10-aa spacer 160GDPPDLADQG169 of CA mutants. The site of the second cleavage depended on Glu159: moving Glu into the spacer or removing 5–10 residues from the spacer sequence resulted in shorter spacers with the cleavage at the carboxylic side of Glu. The mutant E159D was cleaved more slowly than the wild-type, as were mutants G160A and G160L. This allowed kinetic measurements showing that the second cleavage reaction was a first-order, intra-molecular process. Glutaryl-7-aminocephalosporanic acid is the classic substrate of CA, in which the N-terminal Ser170 of the β-subunit, is the nucleophile. Glu and Asp resemble glutaryl, suggesting that CA might also remove N-terminal Glu or Asp from peptides. This was indeed the case, suggesting that the N-terminal nucleophile also performed the second proteolytic cleavage. We also found that CA is an acylpeptide hydrolase rather than a previously expected acylamino acid acylase. It only exhibited exopeptidase activity for the hydrolysis of an externally added peptide, supporting the intra-molecular interaction. We propose that the final CA activation is an intra-molecular process performed by an N-terminal nucleophile, during which large conformational changes in the α-subunit C-terminal region are required to bridge the gap between Glu159 and Ser170.

Catalytically active, mature CA consists of two subunits (␣, 161 aa, and ␤, 522 aa), which are generated by proteolytic cleavage from a single, properly folded 720-aa precursor containing a 27-aa signal peptide (19). The precursor CA (without signal peptide) is activated by two peptide bond cleavages, which excise an internal spacer (Fig. 1a). The first cleavage is known to be initiated by an intra-molecular nucleophilic attack by O␥ of Ser 170 on the adjacent peptide bond between Gly 169 and Ser 170 , via a triad catalytic mechanism assisted by His 192 and Glu 624 . Hydrolytic peptide bond cleavage generates the ␣Ј-subunit, which still contains the spacer attached to the C terminus, and the mature ␤-subunit (16, 20 -25). The C-terminal spacer is released from the ␣Ј-subunit by a second cleavage reaction, which generates the mature ␣-subunit. The spacer is expelled from its position in the CA, and has no further use. The mature ␣and ␤-subunits stay together and form the mature, active, heterodimeric CA (18,26).
It has been debated for years whether the second cleavage is inter-or intra-molecular, which amino acid residue catalyzes the reaction, and what determines the site of cleavage (25,(27)(28)(29). Matured CA from Pseudomonas sp. GK16 was considered to be an (␣␤) 2 heterotetramer complex before releasing a 9-aa spacer (21,30,31). Mutant S170C kept its capability for the first cleavage, which releases the ␤-subunit, but completely lost the activities for both second cleavage to generate the mature ␣-subunit, and substrate (Gl-7-ACA) hydrolysis. Based on these observations, it was proposed that the second cleavage is carried out by inter-molecular interaction using the same active site for substrate hydrolysis and released a 9-aa spacer for CA from Pseudomonas sp. GK16 or an 11-aa spacer for CA from Pseudomonas diminuta (21,22,25). However, the second cleavage of S170C was not performed when incubated with the mature enzyme (28). Although the lengths of spacers reported are different, Glu 159 was reported to be crucial for the second cleavage of both CA precursors from Pseudomonas GK16 and P. diminuta (25,27,28). Based on these observations, mutants containing an integral ␣Ј-subunit were crystallized to deduce the intra-molecular mechanism for the second cleavage. In the putative mechanism, acid Glu 159 , together with a H 2 O in the crystal structure were suggested to form a local active center where Glu 159 played a role as nucleophile that attacks the carbonyl carbon of Gly 160 and then released a 9-aa spacer (27,28).
Here, we precisely located the second cleavage site and detected the corresponding 10-aa spacer 160 GDPPD-LADQG 169 of CA mutants (Fig. 1). We found that the second cleavage site cleaved between Glu 159 and Gly 160 , even when the EG was moved into the spacer or when the spacer sequence was shortened. We also found that CA is an acylpeptide hydrolase, which exhibits only limited exopeptidase activity for N-glutamyl peptides only. These findings, together with the available crystal structure, that shows no general acid/base in the proximity of Glu 159 , suggested that the second cleavage is an intramolecular reaction catalyzed by the same active site Ser 170 that performs the substrate hydrolysis. For this to happen, the maturing CA must undergo a conformational change that cannot be observed by x-ray crystallography.

EXPERIMENTAL PROCEDURES
Strains, Plasmids, and Chemicals-Glutaric anhydride, succinate anhydride, glutaric acid, glutathione (GSH), 6-nitro-3phenylacetamidobenzoic acid, and other reagents were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai) and Aldrich. Cephalosporin C and 7-ACA were purchased from Zhejiang Haimen Pharmacy Co. Peptides and glutaryl peptides were synthesized by China Peptides (Shanghai). Gl-7-ACA was synthesized by acylation of 7-ACA with glutaric anhydride (32). The synthesized sample was identified as Gl-7-ACA by MS and NMR. Succinyl-7-ACA, adipyl-7-ACA, and other glutaryl compounds were synthesized similarly from the corresponding acid anhydrides and amino compounds. The CA gene was transferred from plasmid pKKCAI (19) to pET28a (Novagen) (NcoI and HindIII cleavage), containing a C-terminal His 6 tag. Escherichia coli DH10B was used for gene cloning and site-directed mutagenesis. Plasmid pDB3 carrying the gene coding for penicillin G acylase (PGA) and its mutant encoding S264C were used for protein expressions (33). E. coli BL21/DE3 (pLysE) (Novagen) was the host strain for protein expression.
Site-directed Mutagenesis-For mutagenesis of the CA spacer peptide region, a BsaI site was created upstream of the internal BamHI using PCR primers 5Ј-CAGCCATCATCAT-CATCATCAC-3Ј and 5Ј-TTGGATCCTTGATCGGCCAG-GTCCGGCGGGTCTC-3Ј, and a SmaI site was introduced downstream of the natural internal BamHI site using primers, 5Ј-AAGGATCCAACTCCTGGGCGGTGGCCCCGGGAA-3Ј and 5Ј-CACTATAGGGCGAATTGGGT-3Ј. Introduction of these two restriction sites did not change the CA aa sequence. The BsaI-SmaI fragment encodes Gly 160 to Ala 176 , including the spacer and 7 aa from the N terminus of the ␤-subunit. The mutations inside the spacer were constructed by replacing the BsaI-BamHI or BsaI-SmaI fragments with double-stranded oligos containing the desired base change(s). E159D and E159Q, which are outside the above BsaI-SmaI fragment, were constructed by PCR. All mutations were confirmed by DNA sequencing.
Enzyme Expression and Purification-Overnight cultures expressing CA and its mutant derivatives were inoculated into 500 ml of LB medium containing 0.1% glucose and incubated at 37°C and 220 rpm. At A 600 ϭ 0.6, 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside was added. The temperature was lowered to 30 or 25°C depending on protein stability of the mutants. Cells were harvested 8 h after isopropyl 1-thio-␤-D-galactopyranoside induction, and lysed by sonication. The C-terminal His 6 -tagged enzymes were recovered using the Novagen Hisbind kit at 4°C. The crude enzyme preparations of PGA and its mutant S264C were prepared as described (33). They were further purified by precipitation with 50% acetone on ice, and then applied to a Sepharose CM Fast Flow column (GE Healthcare) with a gradient of sodium acetate ranging from 0 to 500 mM (pH 5.4).
SDS-PAGE Analysis of Second Cleavage-Freshly prepared proteins were put on ice for further SDS-PAGE analysis. The time courses of in vitro second cleavage processing of CA and mutants were carried out in 100 mM PBS buffer (pH 8.0) and samples were immediately denatured by adding SDS-loading Vertical arrows indicate the tryptic cleavage site that generates the CTF TLGE of the ␣-subunit, and the first and second autocatalytic cleavage sites that excise the spacer. Brown, ␣-subunit; green, ␤-subunit; orange background, 10-aa spacer peptide; ␣Ј-subunit indicating the immature ␣-subunit still containing spacer peptide. b, structure of CA (PDB code 1ghd). Brown, ␣-subunit; green, ␤-subunit; stick pattern, Ser 170 and Glu 159 ; orange arrow, spacer peptide, which is removed from the mature CA; ruler, indicating the distance (22 Å) between the O␥ of Ser 170 and the carboxyl group of ␣-subunit C-terminal Glu 159 that was measured by software DS ViewerPro (Accelrys Co.). buffer and boiling for 2 min. The CA ␣and ␣Ј-subunits were then separated using 15% SDS-PAGE and imaged digitally using G:Box (Syngene) prior to densitometric analysis using GeneTools (Syngene).
Enzyme Activity and Kinetic Study for Substrate Hydrolysis-Enzyme activities of CA and PGA were determined colorimetrically by measuring the amount of product 7-ACA or 6-aminopenicillic acid, generated using p-dimethylaminobenzaldehyde (19,33). The peptide products and glutaric acid were quantified by Q-TOF LC-MS (Agilent 6530 Q-TOF LC-MS). The hydrolysis of 6-nitro-3-phenylacetamidobenzoic acid was followed by directly measuring the increase in absorbance at 405 nm. One unit of activity is that amount of enzyme that produces 1 mol of the nucleic product.
Steady-state kinetic parameters were determined by measuring initial velocities at different substrate concentrations. The kinetic constants K m and k cat were determined using Lineweaver-Burk plots. The inhibition kinetics of the hydrolysis of Gl-7-ACA employed the same procedures in the presence of 0.50, 1.0, and 2.0 mM glutarate, giving the disassociation constant K i for the competitive inhibition.
LC-MS and MS-MS Analysis of Oligopeptides-The Q-TOF LC-MS equipped with an Agilent Zorbax 300SB-C8 column (3.5 m 2.1 ϫ 150 mm) was used for oligopeptide separation, MS (4 GHz high resolution, 50 -1700 m/z mass range mode, positive mode), and MS-MS analysis (30 eV for fragmentation). Agilent MassHunter software was used to identify and quantification of peptide products.
To determine the C-terminal tryptic fragments (CTFs) of the ␣-subunits, the enzymes were denatured by incubating with 6 M urea at 65°C for 30 min. The samples were digested by trypsin (Promega) according to the protocol of the supplier. The gradient was 5-95% acetonitrile in 0.1% formic acid over 60 min, at a flow of 0.2 ml/min.
To identify the spacer released from CA mutants (E159D, E159Q, S170C, and m15), the purified enzyme was incubated in PBS buffer (pH 8) for 10 h at 30°C. The mature enzyme and the precursor with or without the first autocleavage were denatured by incubation at 95°C for 5 min and precipitated by centrifugation at 12,000 ϫ g for 10 min, the supernatant containing released spacer was analyzed by Q-TOF LC-MS. The mobile phase contained 0.1% formic acid and 5% acetonitrile for short peptides (up to 5 aa) separation. The gradient for long peptide separation was 5-95% acetonitrile in 0.1% formic acid over 30 min.

LC-MS Identification of the Second Cleavage Site-Previous
investigations reported spacers of four different lengths (8 -11 aa) in CA precursors from different species: 8 aa (PPDLADQG) from Pseudomonas C427 (34), 9 aa (DPPDLADQG) from Pseudomonas sp. GK16 (21,28), 10 aa (GDPPDLADQG) from Pseudomonas sp. 130 (6), and 11 aa (EGDPPDLADQG) from P. diminuta (29). This variation in the reported lengths of the spacer regions was surprising because these CA precursors have more than 98% aa identity (except for the CA from Pseudomonas C427, whose crystal structure is not available), and the available crystal structures are also very similar (supplemental Fig. S1, a and b). Thus, the different spacer lengths were difficult to explain, and we wanted to confirm the sequence of the spacer of CA from Pseudomonas sp. 130 using accurate mass Q-TOF LC-MS.
The first proteolytic step cleaves the CA precursor between Gly 169 and Ser 170 to generate the mature ␤-subunit with the catalytic N-terminal Ser 170 (or numbered S␤1 in the mature enzyme), and the ␣Ј-subunit with the spacer still attached (21).
To identify the C terminus of the mature CA ␣-subunit (after the spacer has been removed), CA from Pseudomonas sp. 130 was digested by trypsin, and analyzed by Q-TOF LC-MS. Of interest were peptides originating from near the C terminus of the CA ␣-subunit that did not end in Arg or Tyr. Such peptides were not generated by tryptic cleavage at the C-terminal end, and thus represented the likely C terminus of the mature CA ␣-subunit from which the spacer has been removed. A fragment, TLGE 159 , gave a high-resolution signal at m/z 419.2144, and matching b and y ions from the MS-MS spectrum confirmed the aa sequence ( Fig. 2a and c). Comparison with the CA precursor sequence predicted from the DNA sequence showed that TLGE 159 was followed by 160 GDPPDLADQG 169 , and then the sequence of the mature ␤-subunit starting with Ser 170 (Fig.  1). Therefore, the 10-aa sequence from Gly 160 -Gly 169 was the likely spacer, and the second activating cleavage hydrolyzed the bond between Glu 159 and Gly 160 . These results agree with the data of Li et al. (6), but differ from other research reported from almost identical CA orthologs.
Mutant CA That Performs the Second Cleavage Slowly, and Direct Observation of the Released 10-aa Spacer-The second cleavage that generates the mature CA happens very quickly after the first cleavage, which made it impossible to study the kinetics of the reaction. From studies with the P. diminuta CA, it was known that Glu 159 is important for hydrolysis of the spacer from the immature ␣Ј-subunit. Replacing Glu by Leu, Met, or Asn prevented the second cleavage reaction, without affecting the release of the ␤-subunit by the first autoproteolytic reaction (25). We created the P. sp. 130 mutant CA E159D, which was still active, but significantly retarded in the second cleavage, allowing the slow release of the spacer to be monitored in vitro (Fig. 2b).
The released spacer was detected by Q-TOF LC-MS, and tandem MS confirmed its 10-aa 160 GDPPDLADQG 169 sequence, which was in agreement with our above prediction from the analysis of the CA ␣-subunit (Fig. 2, b and d). Also the C terminus of the E159D ␣-subunit was confirmed to be TLGD 159 (supplemental Fig. S2a). Although it was reported that either S170C and E159Q completely lost the ability for the second cleavage (28), a small amount of released spacer was detected by Q-TOF LC-MS after a 10-h incubation at 30°C, giving similar tandem MS spectra as the spacer from E159D (supplemental Fig. S2, c and e). However, the rate of the spacer hydrolysis could not be accelerated by incubating S170C with the wild-type CA, which is consistent with previous reports (28) (supplemental Fig. S2c). Interestingly, a small amount of 8-aa spacer consisting of 162 PPDLADQG 169 was also detected from the preincubated mutant E159Q by LC-MS and MS-MS analysis (supplemental Fig. S2d). This result indicated that the second cleavage site may be changeable when Glu 159 was replaced with other aa residues, except for Asp, which structurally resembles Glu, and Asp 161 may play a role as Glu in this cleavage. We thus concluded that the Pseudomonas sp. 130 CA spacer must be the 10-aa sequence 160 GDPPDLADQG 169 .
Repositioning the Site for the Second Cleavage, and the Steric Hindrance Caused by Spacer Residues-If the peptide bond between Glu 159 and Gly 160 was the main determinant for the second cleavage, then it should be possible to reposition the site for the second cleavage. The proposed second cleavage may be affected by the nature of the aa side group at position 160. This was indeed the case because mutant proteins G160A and G160L containing larger side groups performed the second cleavage more slowly. The inhibition of the reaction increased with increasing bulk of the aa side group (Gly Ͻ Ala Ͻ Leu; Fig. 3a).
Mutants m15-m19 were constructed where EG was repositioned into the spacer (Table 1). These proteins performed the first cleavage completely, like the wild-type, but underwent the second cleavage at different rates (Fig. 3, b and c). Q-TOF LC-MS and MS-MS analyses of CTFs from the ␣-subunits of these mutants showed that in each case hydrolysis was between Glu and Gly (Fig. 3, c and d, and supplemental Fig. S3). From mutant m15 only, the main CTF was TLGDPDGE 163 G 164 LADQG, and the peptide TLGDPDGE 163 was present at a much lower concentration. After a 10-h incubation at 30°C, the expected 6-aa spacer, 164 GLADQG, was detected and confirmed by tandem MS (Fig. 3d). The CTF analysis showed that only a part of m15 had performed the second cleavage after such a long time incubation (supplemental Fig. S3, a and b). This significantly retarded second cleavage suggested that repositioning of EG might have created a steric hindrance caused by the changed spacer composition, resulting in the inefficient cleavage. Incomplete cleavage of the mutants, m18 and m19, which also contain 6-aa spacers ( 164 GDPPDG) but maintained the 5 original spacer residues (GDPPD) starting from the N terminus FIGURE 3. Autoproteolysis of CA mutants with aa changes in the spacer. a, SDS-PAGE analyses of the second cleavage of CA mutants G160A and G160L. The first cleavages of these mutants were complete in all cases, so that the ␤-subunits generated from this step were not shown in this figure. Lane 1, the mutant E159Q (second cleavage undetectable by SDS-PAGE), which had a full size ␣Ј-subunit containing an intact spacer was used as a negative control. Lanes 2 and 4, mutants G160A and G161L before incubation. Lanes 3 and 5, mutants G160A and G160L incubated at 30°C, showing slow processing of the ␣-subunits. b, SDS-PAGE analysis of the mutants with different size ␣-subunits. Lane 1, E159Q, which had a full size ␣Ј-subunit was used as the negative control. Lane 7, wild type was the positive control with complete removal of the spacer. Lanes 2-6, the ␣-subunits of mutants m15-m19 had intermediate sizes between E159Q and WT, indicating that different length spacer peptides were released. The additional unmarked bands between the ␤ and ␣Ј bands were caused by autocatalytic degradation of the proteins. This happened when the samples were denatured by boiling in SDS-PAGE loading buffer, and could not be avoided. c, extracted ion chromatogram for possible CTFs from the tryptic digestion of m15-m19. A15-A19, CTFs generated from the respective ␣-subunits; B15-B19, CTFs generated from the ␣Ј-subunits (C-terminal spacer peptide still attached). All A peaks except A15 were detected. B15, B18, and B19 were detected, but B16 and B17 were missing. The inset table shows the measured m/z values acquired by Q-TOF LC-MS that match the theoretical values within the error lower than 5 ppm (the CTF sequences are shown in Table 1). d, MS-MS of the spacer released from the slow processing mutant m15. After incubation at 30°C for 10 h, an extracted ion chromatogram signal appeared at m/z 560.2635, corresponding to [M ϩ 1] ϩ of the spacer, GLADQG (theoretical m/z 560.2675).

TABLE 1
Autocleavage of mutant spacers * "2", tryptic cleavage site (generates the CTF of the ␣-subunit); "͉", first cleavage site; inverted color: spacer detected by MS; underlined: mutant residues. ** and "ϩ", the rate of cleavage reaction was similar to the wild type; "Ϫ", no detectable cleavage; "/", no need for a second cleavage; "slow", cleavage significantly retarded as observed by SDS/PAGE; "very slow", small amount of cleavage detected by Q-TOF LC-MS but not on SDS-PAGE.
( Table 1), further supported this suggestion. Mutants m16 and m17, which contained a 4-and 2-aa spacer, respectively, performed the second cleavage as normal, suggesting that a shorter spacer may eliminate this composition effect. These findings provided strong evidence for the second cleavage occurring between Glu 159 and Gly 160 , and it showed that the specific cleavage site is movable in the spacer range, depending on the position of EG.
Cleavage of Shortened Spacers-Six CA mutants (m24 -m29) containing 5-10-aa deletions of the spacer region were created (Table 1). These mutants retained only 0 -5 aa of the original 10-aa spacer residues. These mutant proteins underwent the first cleavage releasing the ␤-subunit more slowly than the wild-type (Fig. 4a). All six mutants also produced fully processed ␣-subunits, indicating that the second cleavage occurred efficiently at the carboxylic side of Glu 159 as predicted. The CTF, TLGE 159 , was detected in the samples from all six mutants (m24 -m29) by Q-TOF LC-MS (Fig. 4b). Two additional mutants (m31 and m32) had shortened spacer regions and contained mutation E159D (additional G160A for m32), which retarded the second cleavage reaction (see above). Only in these two mutants could unprocessed ␣Ј-subunits be observed (Fig.  4, a and b). Mutant m24, from which the entire spacer has been deleted, is a special case because it required only one hydrolytic cleavage between Glu 159 and Ser 170 to produce the mature CA containing normal ␣and ␤-subunits. Remarkably, in mutant m25, where only a single Gly residue remained from the spacer, the second cleavage released this Gly residue from the ␣Ј-subunit. This again suggested that excision of a 9-aa spacer from CA was unlikely, and it indicated that the spacer composition was not critical for the second cleavage, unless it caused a serious steric hindrance. These findings again supported the hypothesis that the second cleavage of CA occurred between Glu 159 and Gly 160 , and that the site of cleavage was determined by the sequence EG.
All eight mutants with shortened spacer region may have caused some structural tension in the unprocessed precursors, resulting in the incomplete first cleavage reaction, and partial precipitation of the protein within 0.5-2 h at 30°C. The mutant m24 without the spacer sequence precipitated fastest and was least processed at 30°C, but remained active for several hours at 20°C (data not show).
First-order Reaction Kinetics of the Second Cleavage, and Inhibition by Glutarate-It was proposed that the Gl-7-ACA substrate binding at the active site occurred by the interaction of the glutaryl side chain with aa residues in the substrate side chain binding pocket (35). Amino acid changes affecting glutaryl side chain binding, such as S170C, Y205I or -S, R229K or -S, and F349T, simultaneously lost the activities for both the second cleavage and substrate hydrolysis partially or completely (25). Glutarate is a competitive inhibitor for the hydrolysis of Gl-7-ACA. For the CA from Pseudomonas sp. 130, the binding constant was measured by a double reciprocal plot, giving K i ϭ 11 mM (see "Experimental Procedures"). Similar inhibition on the second cleavage of E159D can also be seen in Fig. 2b. However, this mutant was not sufficiently stable for extended kinetics studies.
The kinetics of the second cleavage of mutants G160A or G160L was determined by measuring the relative amount of the ␣-subunit generated by incubation of newly purified proteins in PBS buffer (pH 8.0) at 37°C, where the enzymes were sufficiently stable for this assay (Fig. 5a). It is important to note that the time course at varying concentrations (2 and 4 mg/ml) fitted a single exponential equation, which denoted a firstorder process (Fig. 5b). For an unknown reason, the cleavage rate of the diluted mutant enzyme was not reduced but slightly increased, suggesting the time course was unlikely to fit a second-order equation. As expected, under the same conditions, the k obs of G160A (0.0062 min Ϫ1 ) was distinctly higher than that of G160L (0.0041 min Ϫ1 ) (Fig. 5b). Assuming that the spacer can form active and inactive conformations, like a HDV ribozyme (36), the end extent of the second cleavage fraction (F ∞ , Fig. 5b) for an active phase would be less than 100%. The F ∞ of G160A and G160L were 58 and 46%, respectively, suggesting that the increasing bulk of the aa side group not only reduced the cleavage rate but also decreased the end point F ∞ .
It was observed that glutarate had greater influence on mutant G160L rather than G160A (Fig. 5b). In the presence of 20 mM glutarate, the cleavage rates of both G160A and G160L  (Table 1) performed the second cleavage completely. The second cleavage of m32 and m31 was retarded because of the E159D aa substitution (additional G160A for m32). Only the ␣Ј but not the ␣ bands were visible on the gel. b, extracted ion chromatograms for the CTFs from mutants m24 -m29, m32, and m31. The ion chromatograms of A24 -A29, were extracted, giving a similar signal at 419.2136 Ϯ 5 ppm corresponding to TLGE. A very small amount of A31 at m/z 405.1998 was also detected, corresponding to TLGD. The inset shows the values measured by Q-TOF LC/MS. were decreased by a similar amount (about 20%), reflecting a similar competitive inhibition for second cleavage. However, the end extent second cleavage fraction (F ∞ ) of G160L was significantly decreased (from 46 to 33%), whereas G160A was slightly decreased (from 58 to 55%). This suggests that the binding of glutarate might partially be involved in the formation of an inactive conformer of G160L. These data support both hypotheses that the second cleavage is an intra-molecular interaction and catalyzed by the same active site for Gl-7-ACA substrate hydrolysis.
CA Is an Acylpeptide Hydrolase Rather than an Acylamino Acid Acylase-The specific but movable site for the second cleavage, which results in removal of the spacer from the ␣Ј-subunit suggested that CA may be a peptidase. In fact, derived from a nonribosomally synthesized peptide, 7-ACA, the core of the classic substrate of CA is a highly modified peptide produced by microbes (37). Here, we revealed that CA is actually an acylpeptide hydrolase, which catalyzes the hydrolysis of an N-acylpeptide, by investigation of the hydrolysis of a series of synthetic glutaryl peptides ( Table 2, supplemental Fig. S4).
Although it was described that an acylase did not hydrolyze acylpeptides but acylamino acids (38), we found that CA is a poor aminoacylase for the hydrolysis of glutaryl amino acids. Gl-GP and Gl-AP were cleaved by CA with an efficiency (k cat / K m ) that was about 10-fold higher than for Gl-G or Gl-W containing only a single aa. Remarkably, Gl-GDPP, which contains the same 4-aa sequence as the N-terminal end of the natural spacer peptide was hydrolyzed with the highest efficiency among all substrates tested in Table 2. Interestingly, the k cat value for hydrolysis of Gl-GDPPDLADQG, which contains the entire 10-aa spacer peptide sequence (6), was almost identical to that for Gl-GDPP hydrolysis, whereas the K m was increased ϳ2-fold. The highest k cat value (13.8 s Ϫ1 ) was more than twice that for Gl-7-ACA. This finding suggested that the spacer sequence may aid this hydrolytic activity. Unlike the specificity of the side chain of the substrate (the glutaryl group), the core of the substrate was varied from Gly to a long peptide. The reactivities of the substrates with the same glutaryl side chain were very variable and the k cat /K m values were between 0.01 and 40 s Ϫ1 mM Ϫ1 (Table 2). Slightly changing the spacer composition decreased the k cat /K m value for Gl-GDAADAADKG (changed residues are underlined) hydrolysis by 4-fold. A significant decrease of the k cat /K m value by large scale changes of the composition was observed for hydrolysis of two structurally comparable substrates, Gl-GGGGAA and Gl-GGGGGK. These results indicated an unexpected interaction between the peptide moiety and enzyme. Especially, inserting one Glu in the N-acylamide bond of Gl-GDPPDLADQG mainly resulted in an increase of the K m value for Gl-EGDPPDLADQG (Table 2), again suggesting that the spacer composition was important for the substrate hydrolysis.
Exopeptidase Activity to Remove the N-terminal Glu of the Peptide-We reasoned that the second proteolytic cleavage between Glu 159 and Gly 160 might be similar to the hydrolysis of Gl-7-ACA. Glu 159 would thus be equivalent to the glutaryl side FIGURE 5. Time course of the second cleavage of slow processing CA mutants. a, time course of the second cleavage. The time course of the second cleavage was carried out at 37°C in PBS buffer (pH 8.0) and the yield was quantified by gel densitometry as described (see "Experimental Procedures"). The data are in accordance with those from Q-TOF quantification (additional results from SDS-PAGE are shown in supplemental Fig. S5). b, the kinetics of the second cleavage of CA mutants G160A and G160L and its inhibition by Gl. The solid lines indicate the results obtained with CA mutants (4 mg/ml) with or without 20 mM Gl. The dashed line was obtained from the mutant G160A at lower concentration (2 mg/ml). The first-order rate constant, k, was obtained by fitting the data to the equation F ϭ F 0 ϩ F ϱ (1 Ϫ e Ϫkt ), where F is the yield at time t, F 0 is the initial F of the fresh protein, F ϱ is the latitude, representing the end extent cleavage fraction. Data analysis and graph preparation were done using KaleidaGraph (Synergy Software, PA). Independent experiments were also done by quantification with Q-TOF LC-MS.

TABLE 2 Kinetics for the hydrolysis of different substrates
Values are means of at least three independent experiments. Standard deviations were Յ15%. Initial velocities were determined at the linear region using different concentration of substrates at 37°C. The kinetic constants K m and k cat were determined from Lineweaver-Burk plots (see "Experimental Procedures"). chain of Gl-7-ACA (Fig. 6), and the spacer would be equivalent to 7-ACA. Succinyl-7-ACA was cleaved by CA much slower than Gl-7-ACA (Table 2). This was consistent with the above result where E159D detached the spacer slowly from the ␣Ј-subunit of CA, because Asp resembles the succinyl group (Fig. 6). CA exhibited an exopeptidase activity removing the N-terminal Glu from a peptide, although the k cat /K m value for E-GDPPDLADQG was 10-fold lower than that for Gl-GDPPD-LADQG, probably due to the highly specific substrate side chain interaction ("Discussion," Table 2). Compared with E-GDPP, the hydrolytic activity for D-GDPP was decreased 3-fold as expected, but GE-GDPP was decreased 13-fold (Table  3). Remarkably, the hydrolytic activity for TLGE-DGPP was decreased more than 100-fold. Although only the product GDPP was detected after hydrolysis of either GE-GDPP or TLGE-DGPP, this significantly decreased activity was caused by modification of Glu, together with the highly specific side chain requirement for substrate hydrolysis, suggesting that CA was unlikely to be an endopeptidase executing the proposed inter-molecular reaction. If it is the case, CA can only use an intra-molecular interaction for the second cleavage and the covalent binding of the spacer also becomes important for the second cleavage. This covalent binding, on one hand, can decrease the entropy favoring the cleavage reaction and, on the other hand, may allow the ␣-subunit C-terminal sequence to be placed in a specific place to minimize the potential steric hindrance caused by the binding of the internal Glu in the active center. These findings not only revealed that N-terminal Glu was essential for peptide hydrolysis when the specific substrate side chain binding was required, but it also indicated that for the second cleavage, both specific binding of Glu 159 and the covalent binding of the spacer were required.

DISCUSSION
The Second Autoproteolytic Cleavage of CA Is an Intra-molecular Interaction Catalyzed by Ntn-Due to the lack of information about conformational dynamics, the possibility that the second autoproteolytic cleavage of CA could be an intra-molecular interaction catalyzed by Ser 170 , the Ntn, has been neglected for years. Instead, two mechanisms, an intra-molecular interaction catalyzed by Glu 159 and an inter-molecular interaction catalyzed by Ntn, have been proposed for the reaction (21,25,28).
Our data provide extensive evidence suggesting that the second cleavage is performed by an intra-molecular interaction: the time course of the cleavage reaction fits a first-order equation, even in the presence of the competitive inhibitor glutarate. CA exhibited exopeptidase rather than endopeptidase activity for the hydrolysis of externally added peptides. This was in agreement with published results that the cleavage of mutant S170C ␣Ј-subunit was not catalyzed by CA wild-type (28). The hypothesis that the second cleavage uses the Ntn mechanism was substantiated by the observations on the cleavage specificity, the reversible inhibition by glutarate, and the correlation between the spacer cleavage and the acylpeptide hydrolysis, where the original spacer composition was essential for the high activity. The observation that the second cleavage site could be repositioned within the spacer removed the possibility that CA has an alternative active center, Glu 159 , as was suggested for the autoproteolysis of the CA from Pseudomonas sp. GK16 (28).
However, there was a 22-Å gap between Glu 159 and Ser 170 observed in the crystal structure of CA (PDB code 1ghd, Fig.  1b), indicating that a large scale conformational change would be required if the second cleavage was catalyzed by Ser 170 . The conformation of the active mutant with shortened spacer range, especially mutant m24, which contains no spacer, may reflect that CA is able to perform such a large conformational change, allowing the specific binding of Glu 159 to the active site. Thus, to further elucidate the Ntn mechanism involved in the second cleavage, several points below need to be seriously considered.
Specific Binding and Covalent Binding for the Second Cleavage-In the previous studies, besides CA, PGA and PvdQ were confirmed to be Ntn hydrolases that require two-step proteolysis for activation, releasing spacers consisting of 54, 22, or 23 aa, respectively (39,40) (Table 4). The long distances between the second cleavage sites and the Ntn Ser residues were also observed in the respective crystal structures (7,41). Although it was not confirmed, Flavobacterium glycosylasparaginase (GA) (42) may also contain an 8-aa spacer (deduced from the crystal structure available, PDB code 2gl9). The spacer of CA contains 3 aspartate residues (Asp 161 , Asp 164 , and Asp 167 ) separated by two other aa residues. This characterization was also observed in other Ntn members: multiple repeated Ala or Leu residues separated by 2 or 3 other aa residues in the long  Table 2 and compound 21 is listed in Table 3. spacer of PGA, 4 Gln residues in PvdQ, and 2 Asn residues in GA (Table 4). Supposing the spacer forms an ␣-helix, these repeated aa residues would align to reduce steric hindrance, and allow the spacer to insert smoothly into the catalytic pocket to trigger an efficient second cleavage. In this study, two binding interactions, the specific binding of Glu (equivalent to Gl of Gl-7-ACA) and the covalent binding of the spacer N terminus, were suggested to be important for the second cleavage of CA. However, CA might be a unique case that requires a specific residue for the cleavage, because a similar specific residue does not exist in the other Ntn members listed in Table 4. This might be consistent with the highly specific substrate side chain requirement of CA. A small change in the substrate side chain of Gl-7-ACA, such as prolonging or shortening by 1 carbon unit, or adding an ␣-amino group, usually caused a significant decrease of hydrolytic activity (Table 2). Other enzymes exhibited more flexibility in their requirements for substrate side chain structure. The substrate side chain of PGA can be an aromatic or aliphatic carboxyl group with 2-6 carbons (with k cat range from 4 to 15 s Ϫ1 , supplemental Table  S1), whereas that of PvdQ can be 4 -14 carbons long (40). Because these enzymes lack the special residue, which controlled the cleavage site shift in CA, covalent binding of the spacer became more important for the second cleavage. Nevertheless, some specific elements also exist in these enzymes, including the specific cleavage site and the repeated aa residues. Thus, further extensive investigations on the second cleavages of these enzymes should be carried out.
Substitution of the Ntn Serine Simultaneously Affected the Second Autoproteolytic Cleavage and Substrate Hydrolysis-Depending on the individual Ntn-amidohydrolase, Ser, Thr, or Cys functions as the nucleophile. Exchange of these residues normally has only a small effect on the first cleavage but it significantly inhibited both the second cleavage and substrate hydrolysis. Precise chemistry of the nucleophile that must optimize each individual active site structure was suggested for the Ntn superfamily (43).
Substitution of the Ntn Thr 152 of GA with Ser or Cys (T152S and T152C mutants) reduced k cat for substrate hydrolysis by 2 to 3 orders of magnitude (43). Autoproteolysis in the latter mutant was also very slow but it could be accelerated by hydroxylamine (44). This inactivation was possibly due to the slight deviation of the nucleophile group in the crystal structure (20). Although it was described that PGA mutant S264C (or named as S␤1C for mature PGA) can autocleave as normal but completely lost its substrate hydrolytic activity (39), we found that the mutant still kept a certain low activity for the hydrolysis of 6-nitro-3-phenylacetamidobenzoic acid (supplemental Table S1). Very interesting, the mutant T263G had abnormal first cleavage between Pro 261 and Tyr 260 , and failure of the second cleavage as seen in the mutant (45). This inactivation of the second cleavage may be dependent on the modification of Ser 264 N terminus by the tripeptide, Pro 261 -Thr 262 -Gly 263 . For CA mutant S170C, both second cleavage and substrate hydrolysis were thought to be completely inactivated as described (25). This inactivation might have been caused by the lack of bound water in the mutant crystal structure (28) that was suggested to be required for bridging the charge relay system in the active center (7). However, Cys 170 in the mutant may still function as a weak nucleophile, because trace amounts of both products from the second cleavage and the substrate hydrolysis were detected and the V max of the hydrolysis of Gl-GDPP was decreased by about 5 orders of magnitude (V max Ϸ 4.6Exp(-5), 5 mol/min mg) (supplemental Fig. S2, b and c). This comparison of the Ntn-substituted mutants of different enzymes suggested an intimate correlation between the two reactions, strongly supporting the hypothesis that the second cleavage and substrate hydrolysis were performed at the same active site.
General Acid Base Catalytic Mechanism for Different Enzymatic Reactions in Ntn Hydrolases-Most protein enzymes and ribozymes (46,47) use general acid-base catalysis to increase reaction rates. Ntn is a typical general acid-base catalytic mechanism involved in substrate hydrolysis, where the N-terminal amine group plays a role as base to deprotonate the hydroxyl group at the same residue (7). A triad charge relay system consisting of Ser 170 , His 192 , and Glu 624 was observed in CA, but this was only responsible for the first cleavage during the enzyme activation (48). For autoproteolysis of GA, a highly strained conformation at the scissile peptide bond had been identified and was hypothesized to be critical for driving the cleavage reaction through an N-O acyl shift (42). The general acid/base for the cleavage was identified as the aa pair Asp 151 and Thr 152 , according to the crystal structure of the mutant D151N that was not processed (49). Similar conformation constraints in the spacer can also be postulated from the crystal structure of the PGA mutant, T263G, in which the carbonyl group of Tyr 260 and the amide of Pro 261 are separated by 8 Å (45). It was revealed that both Thr 263 and Ser 264 were crucial for the first cleavage of PGA. Substitution of Ser 264 with Thr, Arg, or Gly resulted in the complete failure of enzyme activation (39). Substitution of Thr 263 with Ser or Cys had less effect on the first cleavage but substitution with Gly also resulted in the failure of the normal first cleavage between Thr 263 and Ser 264 . These observations may exhibit a novel feature of the Ser/Thr pair as the general acid/base catalytic center for the first cleavage of PGA.
A strained spacer may provide extra energy for the cleavage, allowing the enzyme to use a low efficiency catalytic (P)IVNIENHD Asn Not tested a Underlines represent repeated aa residues that are separated by 2 or 3 other aa residues in the spacer range. b Spacer composition was speculated from the crystal structure of GA (PDB code 2gl9).
pair, such as Asp/Thr or Ser/Thr, which is not widely adopted in enzymatic catalysis. However, similar conformation constraints may not exist during the second cleavage, because the spacer is free after the first cleavage. According to the crystal structure of CA, there were no residues containing active groups, such as -COOH, -OH, and -NH 2 or ϭNH in the proximity of Glu 159 (Fig. 7) (26), except for Asp 161 in the spacer (29). The cleavage site shift and spacer range shortening experiments confirmed that the spacer residues of CA did not participate in the first or second cleavage. Thus, there was no general acid/base pair system within 8 Å of the second cleavage site (Fig. 7).
In summary, we found that the second cleavage site of CA was between Glu 159 and Gly 160 , and that the cleavage site could be moved within the spacer by repositioning the sequence EG. CA is an acylpeptide hydrolase that exhibits only limited exopeptidase activity. These findings revealed an intra-molecular Ntn mechanism for the second cleavage during the enzyme activation. Together with the crystal structures, these findings also predicted a long distance movement of the spacer that is required to bridge the 22 Å gap between the cleavage site and the active center Ser 170 . Although only CA exhibited residue specificity for the ␣-subunit C terminus in the second cleavage, the belief is held that other Ntn hydrolases may adopt a similar mechanism to perform the second cleavage.