CMP-N-acetylneuraminic acid synthetase from Escherichia coli K1 is a bifunctional enzyme: identification of minimal catalytic domain for synthetase activity and novel functional domain for platelet-activating factor acetylhydrolase activity.

Escherichia coli CMP-NeuAc synthetase (EC 2.7.7.43) catalyzes the synthesis of CMP-NeuAc from CTP and NeuAc, which is essential for the formation of capsule polysialylate for strain K1. Alignment of the amino acid sequence of E. coli CMP-NeuAc synthetase with those from other bacterial species revealed that the conserved motifs were located in its N termini, whereas the C terminus appeared to be redundant. Based on this information, a series of deletions from the 3'-end of the CMPNeuAc synthetase coding region was constructed and expressed in E. coli. As a result, the catalytic domain required for CMP-NeuAc synthetase was found to be in the N-terminal half consisting of amino acids 1-229. Using the strategy of tertiary structure prediction based on the homologous search of the secondary structure, the C-terminal half was recognized as an alpha1-subunit of bovine brain platelet-activating factor acetylhydrolase isoform I. The biochemical analyses showed that the C-terminal half consisting of amino acids 228-418 exhibited platelet-activating factor acetylhydrolase activity. The enzyme properties and substrate specificity were similar to that of bovine brain alpha1-subunit. Although its physiological function is still unclear, it has been proposed that the alpha1-subunit-like domain of E. coli may be involved in the traversal of the blood-brain barrier.

Recently, CS from Neisseria meningitidis has been crystallized as a dimer, and the active site residues, mononucleotidebinding pocket, substrate-binding pocket, and dimerization domain have also been identified (28). The structural information obtained from N. meningitidis CS enables us to gain more insight into the structure and function relationships of its counterpart in E. coli. We have reported previously the cloning of the gene for CS from E. coli K1 strain 44277. By substitution of 6 bases in the initial region and stop codon of the wild type gene, we achieved a high level of expression of CS in E. coli. The recombinant CS protein consisted of 26% of the total bacterial proteins, and the activity was 850-fold higher than that of the wild strain (29). Alignment of the amino acid sequences of E. coli CS with its counterpart in N. meningitidis revealed that the highly conserved features, which include the phosphate binding loop, Mg 2ϩ , NeuAc, cytidine binding domain, and dimerization domain, mostly concentrate in the N-terminal half of E. coli CS. We therefore hypothesized that only the N-terminal half of E. coli CS was required for synthetase activity, whereas the C terminus appeared to be redundant. In order to verify our hypothesis, we constructed a series of enzyme mutants by truncating its C terminus. It turns out that the truncated enzymes were still active when more than 189 amino acid residues were removed from C terminus, and the instability of the mutants was also observed. Based on our observations, we concluded that C-terminal half might play a role in stabilizing of the catalytic domain. In an effort to better understand how the C-terminal half stabilizes the catalytic domain in the N-terminal half, we predicted the secondary structure of the C-terminal half and simulated the three-dimensional model by using the strategy of homologous search of the secondary structure. Strikingly, the C-terminal half was found to share a high three-dimensional structural homology with the ␣1-subunit of platelet-activating factor acetylhydrolase (PAF-AH) isoform I. Strong evidence has also been obtained to demonstrate that E. coli CS was a bifunctional enzyme that possessed both CS and PAF-AH activities. Here we present the first report that PAF-AH-like activity was recognized in a microorganism.
Microorganism and Plasmids-E. coli strain 44277 was obtained from the China General Microbiological Culture Collection Center. E. coli DH5␣ was purchased from Invitrogen. E. coli BL21(DE3)pLysS and plasmid pET-15b were obtained from Novagen. The plasmid pET-CS2 containing wild type CMP-NeuAc synthetase gene has been described previously (29).
Culture Conditions-Unless otherwise indicated, E. coli was routinely propagated at 37°C in LB broth contained 10 g of bactotryptone, 5 g of bactoyeast extract, and 10 g of NaCl per liter of culture medium. Growth medium was supplemented with 100 g/ml ampicillin when required.
Cloning and DNA Manipulations-All DNA manipulations were carried out using standard techniques described by Sambrook et al. (30) or according to the manufacturers' instructions.
Construction of Truncation Mutants-The C-terminal truncation mutants of CMP-NeuAc synthetase were generated by PCR with Pfu DNA polymerase. CMP-NeuAc synthetase mutant gene was amplified by using plasmid pET-CS2 as a template and a 5Ј-primer (5Ј-GTCCAT-GGATCTGTATCAGAACATTCATAATAGAATC) with an introduced NcoI site and a 3Ј-primer with a BamHI site. A stop codon was introduced into the 3Ј-primer so that translation could terminate at positions 395, 339, 283, 246, 229, 227, 224, 223, 221, 213, and 210. PCR was carried out by 30 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 1.5 min. The amplified fragments were inserted into the corresponding sites of pET-15b, and the resulting plasmids were designated as pET-CS 395 , pET-CS 339 , pET-CS 283 , pET-CS 246 , pET-CS 229 , pET-CS 227 , pET-CS 224 , pET-CS 223 , pET-CS 221 , pET-CS 213 , and pET-CS 210 , respectively.
As for N-terminal truncation mutants, the 3Ј-primer used was 5Ј-ATGGATCCTTATTTAACAATCTCCGCTAT, and a BamHI site was introduced. By introducing an NcoI site into 5Ј-primer, the translation of mutant construction could start at a chosen position with a start codon contained in the NcoI site. A mutant fused with a His tag was constructed by cloning the coding region for 230 -418 amino acid residues into pET-15b at the sites of NdeI/BamHI. Primers used in our investigation are summarized in Table I.
Protein Expression-The recombinant plasmid was transformed into E. coli BL21 (DE3) pLysS (Novagen). The resulting transformant was grown at 37°C in LB broth containing 100 g/ml ampicillin. When the A 600 values reach 0.6 -0.8, the culture was induced with 0.4 mM isopropyl-␤-D-thiogalactopyranoside for 10 h at 25°C. The cells were harvested at 5,000 rpm for 15 min and frozen at Ϫ70°C until needed.
Purification of the C-terminal Truncated CMP-NeuAc Synthetase-The cell pellets obtained from 1 liter of induced culture of transformed E. coli BL21 (DE3) pLysS was resuspended in 80 ml of 50 mM Tris-HCl (pH 8.0), and the cells were disrupted by sonication. The cellular debris was removed by centrifugation at 14,000 ϫ g for 15 min. The supernatant was precipitated by the addition of solid ammonium sulfate. The proteins precipitated between 35 and 60% were collected and dissolved in 50 ml of 50 mM Tris-HCl (pH 8.0). After dialysis against 3 liters of 50 mM Tris-HCl buffer (pH 7.6) containing 100 mM NaCl overnight, the enzyme solution was applied to a DEAE-cellulose column (2.5 ϫ 30 cm) pre-equilibrated with 50 mM Tris-HCl buffer (pH 7.6) containing 100 mM NaCl. The column was washed with 1 column volume of initial buffer to remove unbound proteins. The enzyme was eluted with 2 liters of a linear gradient of NaCl from 0.10 to 0.25 M. The enzyme fractions were pooled, dialyzed against the initial buffer, and concentrated with Immersible CX-10 TM (nominal molecular weight leakage (NMWL) 10,000-Da, Millipore). All purification steps were carried out at 4°C.  1-395  5Ј-GCCGGATCCTTATATTTTTTGTCTATTTTT  PET-CS 395  1-339  5Ј-GCCGGATCCTTACTTCGAGTTTATACCAGC  PET-CS 339  1-283  5Ј-CGCGGATCCTTATGATGCATCATTTTTCTT  PET-CS 283  1-246  5Ј-GCCGGATCCTTAATCAAATTCATTTCG  PET-CS 246  1-229  5Ј-GCGGGATCCTTAAGTATACATTTTATTTTAG  PET-CS 229  1-227  5Ј-GTGGATCCTTATTGTCTATTTTTTTTTTGCTG  PET-CS 227  1-224  5Ј-GTGGATCCTTATTTTTTTTGCTGAATGGTAAT  PET-CS 224  1-223  5Ј-GTGGATCCTTATTTTTGCTGAATGGTAATTGC  PET-CS 223  1-221  5Ј-AATGGATCCTTACTGAATGGTAATTGCAAG  PET-CS 221  1-213  5Ј-GCCGGATCCTTAATCCATTCTATCATCTAT  PET-CS  Purification of the N-terminal Truncated CMP-NeuAc Synthetase-Cells obtained from culture were resuspended in 50 mM Tris-HCl (pH 7.6) and disrupted by sonication. After centrifugation at 12,000 ϫ g for 15 min, the supernatant was collected. The supernatant was loaded onto a HiPrep 16/10-DEAE FF column (50 ml, flow rate 5 ml/min) pre-equilibrated with buffer A (20 mM Tris-HCl (pH 7.6)) and eluted with a linear gradient of 0 -0.3 M NaCl in buffer A. The fractions containing the target proteins were collected, dialyzed overnight in buffer A, and then loaded onto MonoQ HR 5/5 (20 ml, flow rate 0.5 ml/min), which had been equilibrated with buffer A (20 mM Tris-HCl (pH 7.6)). The mutant enzyme was eluted with a linear gradient of 0 -0.5 M NaCl in buffer A. The fractions containing the target proteins were loaded onto Superdex 200/75 10/30 (100 l, 0.5 ml/min) preequilibrated with buffer A containing 0.1 M NaCl. The fractions containing the target proteins were collected. All purification steps were carried out by using fast protein liquid chromatography system (Amersham Biosciences) at 4°C.
Enzyme and Protein Assays-The CMP-NeuAc synthetase activity was assayed using the thiobarbituric acid method (23). The enzyme was incubated in 250 l of buffer containing 5.5 mM CTP, 2.8 mM NeuAc, 0.2 M Tris, 20 mM MgCl 2 , and 0.2 mM dithiothreitol (pH 9.0). After the mixture was incubated at 37°C for 60 min, 50 l of 1.6 M NaBH 4 was added to destroy excess NeuAc at room temperature for 15 min. The mixture was then put on ice, and 50 l of 20 N H 3 PO 4 was added to decompose NaBH 4 . After standing at 0°C for 5 min, the mixture was then incubated at 37°C for 10 min to cleave the phosphoester bond of the formed CMP-NeuAc. The free NeuAc was oxidized with 50 l of 0.2 M NaIO 4 at room temperature for 10 min, and 400 l of 4% NaAsO 2 in 0.5 N HCl was added. The solution mixture was then transferred to a test tube containing 1 ml of 0.6% thiobarbituric acid in 0.5 M Na 2 SO 4 and heated in boiled water for 15 min. After the solution was cooled, 1 ml of the solution was taken out and mixed with 1 ml of cyclohexanone. The mixture was shaken and centrifuged, and the upper layer was subjected to absorbance measurement at 549 nm (⑀ ϭ 4.11 mM Ϫ1 cm Ϫ1 ). One unit is defined as the amount of enzyme that catalyzes the formation of 1 mol of CMP-NeuAc/min under the conditions of the assay.
The standard PAF-AH activity assay was carried out as described by Hattori et al. (31) with small modifications. The specific activity of [ 3 H]acetyl-PAF was adjusted to 200 dpm/nmol by dilution with unlabeled PAF. The standard incubation system for assaying consisted of 50 mM Tris-HCl (pH 7.4), 20 nmol of [ 3 H]acetyl-PAF, and the sample in a total volume of 250 l. The reaction was terminated by addition of 2.5 ml of chloroform/methanol (4:1, v/v) and 0.25 ml of water. The radioactivity of an aliquot (0.6 ml) of each upper phase was measured to determine the amount of acetate liberated. One unit is defined as the amount of enzyme that catalyzes the formation of 1 nmol of acetate/min.
The assay for serum (plasma) PAF-AH activity was carried out using Azwell Auto PAF-AH kit (Nesco, Japan) according to the manufacturer's instructions. The substrate used in this assay was 1-myristoyl-2-(4-nitrophenylsuccinyl) phosphatidylcholine.
Esterase activity assay was carried out at 25°C in 50 mM Tris-HCl (pH 7.0), containing 4 mM of p-nitrophenyl butyrate. An extinction coefficient for p-nitrophenol of 7.2 ml mmol Ϫ1 cm Ϫ1 at 400 nm was used. One unit is defined as the amount of enzyme catalyzing the appearance of 1 mol of p-nitrophenol per min.
Protein concentrations were determined by using the method described by Lowry et al. Bovine serum albumin was used as a standard.
TLC Analysis of Substrate Specificity-Substrate specificity was assessed by TLC as described by Grigg et al. (32) with small modifications. After incubation of 50 g of phospholipids/lipids with the enzyme for 3 h at 37°C, lipids were extracted by the Bligh-Dyer method (33), dried under a stream of N 2 , resuspended in methanol, and then applied to silica gel plates. A solvent system of chloroform/methanol/acetic acid/ water (50:25:8:4, by volume) was used to resolve the phospholipid samples and hexane/diethyl ether/acetic acid (70:30:1, v/v) to resolve the lipid samples (34). The plates were sprayed with either phosphomolybdic acid or rhodamine B, and the R F values of substrates and cleavage products were determined.
Computer Analysis-The multiple alignment of amino acid sequences was performed using Omiga (2.0). Blast search was performed using the program BLAST. The secondary structure was predicted using the method described by Pan (36). The prediction of tertiary structure of the protein was carried out using the strategy of homologous sequence search at website of Swiss-Model (www.expasy.org/ swissmod/SWISS-MODEL.html) or homologous secondary structure search in PDB Bank (www.rcsb.org/pdb/).

Identification of Minimal Functional Domain for CMP-NeuAc Synthetase Activity-Several
CSs have been isolated from both mammalian and bacterial sources (18 -26). According to their molecular size, these enzymes can be divided into two groups as follows: (i) those with molecular mass ranging 20 -30 kDa and possessing 200 -300 amino acids mostly come from bacterial species, such as 24.8 kDa from N. meningitidis; (ii) those with molecular mass ranging 40 -60 kDa and possessing 400 -500 amino acids mostly come from eukaryotic species. The second group has a molecular mass more than twice that of the first group. It is interesting to note, with a exception, that CSs from both E. coli and Helicobacter pylori belong to the second group. E. coli CS possesses 418 amino acid residues, which is almost twice of that of N. meningitidis (228 amino acids). Alignment of amino acid sequence of E. coli CS with that of N. meningitidis CS revealed that the highly conserved residues that are present in the active site, substrate-binding pocket, and dimerization domain could be found within the N-terminal half (1-210 amino acid residues) of E. coli CS (28). We therefore hypothesized that the remaining residues in the C-terminal half were redundant.
In order to verify our hypothesis, a series of mutant genes with deletion from 3Ј termini were generated and subcloned into NcoI/BamHI sites of pET-15b. When the resulting plasmid was transformed into BL21(DE3)pLysS cells, the truncated enzymes CS 210 , CS 213 , CS 221 , CS 229 , CS 246 , CS 283 , CS 339 , and CS 395 could be induced as soluble forms, which terminated at positions 210, 213, 221, 229, 246, 283, 339, and 395, respectively. The assay for CS activity showed that CS 229 , CS 246 , CS 283 , CS 339 , and CS 395 were active, whereas no activity was detected with CS 210 , CS 213 , or CS 221 . The wild type and active mutants were partially purified (Fig. 1). As shown in Table II, the activities of CS 395 , CS 339 , and CS 283 were 65, 38, and 31% of the wild type, respectively. CS 246 showed 85% loss of activity, while removal of 17 additional C-terminal amino acids to produce CS 229 resulted in a 3-fold increase of activity and showed 43% of activity of the wild type.

Discovery of PAF-AH Associated with E. coli CS
The highest activity for each active mutant occurred at 37°C. CS 229 even showed a slightly higher catalytic capacity than the wild type at 30°C (data no shown). The wild type enzyme exhibited the highest activity at pH 9.0; truncation of C terminus caused a shift of optimum pH values to 7-8 and showed a broader range than the wild type, with only one exception of CS 229 , which was similar to the wild type. Also, CS 229 was more active at pH 7-8 than the wild type. These results indicated that these mutants, especially CS 229 , became more flexible than the wild type enzyme.
The stability of mutant enzymes was determined to ascertain whether these deletions adversely affect the stability of the enzymes. As shown in Fig. 2, the wild type was stable at 37°C. Incubation at 40°C for 70 min caused about 30% of activity loss. As for mutants, only CS 229 showed 10% loss of activity at 37°C for 20 min; all others were stable. The CS 246 and CS 229 were less stable than the wild type at 40°C, and their activities dropped to about 60% in 20 min, whereas the wild type and other mutants remained at 80 -90% of activities.
Our results showed that the functional domain of E. coli CS was composed of residues 1-229. Removal of the C-terminal half (residues 230 -418) resulted in some increase of structural flexibility and therefore contributed to a slight increase in catalytic capacity under non-optimal conditions. It is reasonable to propose that the C-terminal half (residues 230 -418) played a role in stabilization by interacting with the N-terminal half of the enzyme.
Computer Modeling of CS Protein-Although CS 229 , CS 246 , CS 283 , CS 339 , and CS 395 remained active, a decrease in activity was observed, and a loss in stability was also documented for either CS 229 or CS 246 . These results, at least, indicated that the C-terminal half played a role in stability. The prediction of three-dimensional structure based on homologous sequence search revealed that CS 229 shared a three-dimensional structural identity with N. meningitidis CS (PDB code 1eyr:A) (28). N. meningitidis CS has been reported as a dimer, and the active site residues, mononucleotide-binding pocket, substratebinding pocket, and dimerization domain have also been identified (28). As shown in Fig. 3A, the first 220 amino acids of E. coli CS are needed for assembling all secondary structural units; however, no activity was detected with CS 221 . We also generated mutant CS 223 , CS 224 , and CS 227 with the translation terminated at position of 223, 224, and 227, respectively. However, all these three mutants were expressed mainly as inclusion bodies. Only CS 223 could express a small portion of soluble form. The soluble CS 223 exhibited 16% activity of the wild type. These observations suggested that the coil region formed by 222 QKKNRQKI 229 at the C terminus, especially the hydrophobic residue Ile 228 , could maintain the orientation of the Cterminal ␣-helix ( 211 RMDFELAITI 220 ) and therefore allow Mg 2ϩ -binding site (Arg 207 ) close enough to the active center. In our experiment, a decrease in stability of mutant CS 229 or CS 246 has been detected, whereas CS 283 , CS 339 , and CS 395 were as stable as the wild type. Also, we did not detect any dimer formed by either CS 229 or CS 246 on native PAGE (data not shown). Thus, the stability loss displayed by either CS 229 or CS 246 can be ascertained to the loss of interaction with the C-terminal half, at least, with residues 230 -283. In addition, the instability of CS 229 or CS 246 could also be explained by exposure of hydrophobic side chain of Ile 27 to the surface, in comparison with N. meningitidis CS (Fig. 3, B and C).
To understand how the C-terminal half stabilizes the Nterminal half and to explore the potential role of the C-terminal half besides its stabilizing role, we carried out homologous analysis by blasting the amino acid sequence of the C-terminal half (residues 230 -418) alone in GenBank TM , but no any homologous protein was retrieved. Then the secondary structure of the C-terminal half was predicted by using the method developed by Pan (36) (Fig. 4A). When the predicted secondary structure was used for computer modeling, to our surprise, the predicted tertiary structure of the C-terminal half was shown to share identity with that of ␣1-subunit of PAF-AH isoform I (PDB 1fxw:A) (Fig. 4C), a 29-kDa protein isolated from bovine brain soluble fraction (37-39). PAF-AH isoform I belongs to  serine esterase family and is an intracellular heterotrimeric enzyme composed of 29-(␣1), 30-(␣2), and 45-kDa (␤) subunits. Among these subunits, only ␣1-subunit served as the catalytic subunit. Also, ␣1 alone is enough to catalyze the removal of the acetyl moiety at the sn-2 position of PAF (1-O-alkyl-2-acetylsn-glycero-3-phospholine) to produce biologically inactive lyso-PAF. Ser 47 has been identified as the active center residue of ␣1. A large family of enzymes that have an essential serine residue possesses the consensus sequence, Gly-Xaa-Ser-Xaa-Gly. These include proteases, esterases, transacylases, and lipases (40). However, the second residue from the active Ser is Val in ␣1-subunit. In active center, Ser 47 , Asp 192 , and His 195 have been revealed to form a catalytic triad. In most serine esterases, the Asp and His are separated by more than 20 amino acid residues (41), whereas the distance in ␣1 is only 4 residues (Fig. 4). As revealed by the three-dimensional modeling of the C-terminal half of CS (Fig. 4D), Ser 257 , Asp 397 , and His 400 were recognized as a putative catalytic triad in catalytic site. The consensus sequence surrounding Ser 257 was GHSLF (Fig. 4B). Thus, we postulated that the C-terminal half (230 -418 residues) in CS was an "␣1-subunit-like" functional domain.
Evidence for PAF-AH Activity Associated with E. coli CS-Based on the information obtained from three-dimensional structure prediction, we first detected the PAF-AH activity with complete CS protein using [ 3 H]acetyl-PAF as substrate. As expected, CS could catalyze the removal of the acetyl group at the sn-2 position of PAF. Moreover, both CS and PAF-AH activities were assayed at each step during the purification procedure of complete CS protein. As shown in Table III, the increase of purification fold for PAF-AH activity was found to be almost the same as that of CS activity. Both PAF-AH and CS activity could be detected when the protein was purified to homogeneity at the final step (Fig. 5). These results clearly demonstrated that the PAF-AH activity was associated with CS protein.
We also constructed mutant enzyme CS 228 -418 and CS 230 -418 by truncating 227 and 229 residues at the N terminus, respectively. Both mutant enzymes were expressed in E. coli as soluble forms. The activity determination revealed that both mutants were devoid of CS activity; however, only CS 228 -418 was capable of hydrolyzing PAF, whereas CS 230 -418 had no PAF-AH activity. These results demonstrated that the PAF-AH functional domain was composed of amino acids 228-418 in the C-terminal half of CS. It should be pointed out that CS 228 -418 was not exactly initiated with Lys 228 at its N terminus. In order to truncate the N terminus of CS, CCAT-GGAT was introduced to the 5Ј-end of mutant gene so that the translation of mutant gene could start from ATG contained in the NcoI site. Thus, the N terminus of CS 228 -418 was Met-Gly-Lys 228 instead of Lys 228 . Due to the same reason as in CS 228 -418 , the CS 230 -418 translate was initiated with Met-Asp-Leu 230 but Leu 230 . Somehow it appeared that this modification had no effect on the TABLE III Purification of the wild-type enzyme from E. coli The cell pellets obtained from 1 liter of induced culture of E. coli harboring pET-CS2 were resuspended in 80 ml of 50 mM Tris-HCl (pH 8.0), and the cells were disrupted by sonication. The wild-type CS protein contained in the supernatant was collected and purified as described under ''Experimental Procedures.'' The assay for either CS or PAF-AH activity was carried out. Under the assay conditions described under ''Experimental Procedures,'' 1 unit of CS activity is defined as the amount of enzyme that catalyzes the formation of 1 mol of CMP-NeuAc/min, and 1 unit of PAF-AH is defined as the amount of enzyme that catalyzes the formation of 1 nmol of acetate/min.
Step  4. Prediction of the secondary structures and three-dimensional model of the C-terminal half of CMP-NeuAc synthetase. The three-dimensional structure (D) of the C-terminal half (residues 230 -418) of CS was predicted based on the homologous search of the secondary structure, which was predicted using the method described under "Experimental Procedures" (B). The secondary (A) and three-dimensional structure (C) of bovine brain PAF-AH I were retrieved from Protein Data Bank (PDB 1fxw:A). A and B, the ␣-helix is indicated as H, the ␤-sheet is shown as E, and the coil region as C. The serine esterase triad is marked in red.
PAF-AH activity, at least for the CS 228 -418 mutant.
General Properties of PAF-AH Functional Domain-To define the catalytic properties of the PAF-AH domain of CS protein, the mutant CS 228 -418 was purified. After purification on DEAE-Sepharose, Mono Q, and Superdex 75 columns, CS 228 -418 was purified to 19.5-fold with a specificity of 1.69 units/mg (Table  IV). The CS 228 -418 was purified to homogeneity, which was judged by SDS-PAGE stained with both Coomassie Brilliant Blue R-250 and silver staining (Fig. 6).
With purified wild type CS and CS 228 -418 , the effects of pH values and temperatures were determined. Both wild type and mutant were most active at 30°C. The wild type exhibited optimum pH of 7-8, whereas the optimum pH of mutant occurred at a pH value of 8 -9.
Substrate Specificities of PAF-AH Catalytic Domain-PAF-AHs are unique calcium-independent PLA 2 s, which hydrolyze the sn-2 ester linkage in PAF-like phospholipids with a marked preference for very short acyl chains, typically acetyl. As for ␣1 catalytic subunit of bovine brain PAF-AH isoform I, the side chains of Thr 103 , Leu 48 , and Leu 194 were identified to be involved in substrate recognition (42). These three residues were also conserved in the C-terminal half of CS as Thr 308 , Leu 258 , and Leu 399 (Fig. 4), which implies that the PAF-AH domain of CS shares the same substrate specificity with the ␣1-subunit. A previous investigation (35) has shown that the ␣1-subunit of PAF-AH I does not recognize oxidized phospholipids as substrates. By using purified wild type CS and mutant CS 228 -418 proteins, the abilities to hydrolyze PAF (1-O-alkyl-2-acetyl-snglycero-3-phosphocholine), PC (1-palmitoyl-2-arachidonoyl-snglycerol-3-phosphocholine), 1-myristoyl-2-(4-nitrophenylsuccinyl)-phosphatidylcholine, oxidized PC (1-palmitoyl-2-(5-oxovaleroyl)-sn-glycero-3-phosphocholine), glycerol triacetate (triacetin), and p-nitrophenyl butyrate were investigated. As shown in Fig. 7, neither the wild type CS nor CS 228 -418 had activity against PC, a substrate for PLA 2 , or oxidized PC, a substrate for PAF-AH isoform II. Both wild type CS and CS 228 -418 showed activities against glycerol triacetate, a substrate for lipase. When 6.1 g of wild type CS or 1.3 g of mutant CS 228 -418 was incubated with p-nitrophenyl butyrate, a substrate for esterase, the specific activity was determined as 75.6 and 319.8 units/mg, respectively. Although no activity was detected with either the wild type or CS 228 -418 when a substrate for plasma PAF-AH, 1-myristoyl-2-(4-nitrophenylsuccinyl)-phosphatidylcholine was used (data not shown). These results clearly demonstrated that both wild type and mutant enzymes exhibited activity against the sn-2 ester linkage in PAF-like phospholipids with short acetyl chains and also exhibited lipase-and esterase-like activities. No activity was detected toward PAF-like phospholipids with long acyl chains or oxidized phospholipids.
Effect of Various Compounds on FAF-AH Activity-Various compounds were tested for their effects on the PAF-AH activity of the wild type CS or CS 228 -418 . As summarized in Table V, p-bromophenacyl bromide, which has been shown to block various phospholipase A 2 activities by derivatizing the histidine residue at their active sites, inhibited the PAF-AH activity of the wild type CS or CS 228 -418 with a percentage of 70% at 4 mM. Phenylmethylsulfonyl fluoride, an inhibitor of serine esterase, inhibited the PAF-AH activity by 43-45% at 4 mM (Table V). On the other hand, 5,5Ј-dithiobis-(2-nitrobenzoic acid) and dithiothreitol could not inhibit the activity. These results suggested the presence of both essential histidine and serine residues at the active site, and the free sulfhydryl residue was not essential for the PAF-AH activity, which was consistent with the putative catalytic triad identified in threedimensional models. Moreover, Ca 2ϩ , Mg 2ϩ , and EDTA had no effect on the PAF-AH activity of wild type or CS 228 -418 (data no shown). It was interesting to note that the PAF-AH activity of wild type or CS 228 -418 was increased by 1-fold in the presence of 40% glycerol.
Kinetics of PAF-AH Activity-By using the purified wild type and CS 228 -418 mutant, the effect of PAF concentration on the rate of hydrolysis was determined. In our experiments, 22 g of the wild type CS or 11 g of CS 228 -418 was incubated with various amounts of PAF at 37°C for 30 min. The rate of PAF hydrolysis catalyzed by the wild type enzyme increased with increasing concentrations of PAF to about 10 M, whereas the rate of hydrolysis by mutant increased with increasing concentrations of PAF to about 20 M. The K m was estimated to be 6.7 M for the wild type and 4.3 M for CS 228 -418 , respectively (Fig.  8). Apparently, the substrate-binding affinity of CS 228 -418 is higher than that of the wild type.
Proteolysis of Wild Type Enzyme-The mutant analyses showed that CS activity and PAF-AH activity domains of CS protein could function separately, and we carried out proteolytic analysis of CS protein to explore other potential in vivo forms instead of full-length CS. When the CS was incubated with trypsin, after 1 h of incubation, a 22-kDa protein band became visible, whereas the density of CS protein band was found decreased on SDS-PAGE. A prolonged incubation resulted in a complete conversion of CS protein band into 22 kDa (Fig. 9). It is evident that the CS could be degraded by trypsin to produce a trypsin-resistant 22-kDa polypeptide. To define the product of trypsin treatment, the 22-kDa protein was trans-  1-7) or silver staining (lanes 8 and 9).

DISCUSSION
Central nervous system infection caused by bacterial, viral, fungal, or parasitic pathogens can lead to devastating neurological disability and death. It is likely that further improvements in outcome will require new therapeutic approaches based on a better understanding of molecular and cellular mechanisms of central nervous system infection. There are many steps that are required for microorganisms to cause central nervous system infection. The least understood step in the pathogenesis of central nervous system infection by microbes is how circulating microbes penetrate the BBB and enter cerebrospinal fluid. E. coli is one of the leading Gramnegative bacteria that cause neonatal meningitidis. It has been shown that the successful traversal of the BBB by circulating E. coli K1 is a complex process, requiring several bacterial and host factors and their interactions, such as a high degree of bacteremia, binding to and invasion of BMEC, BMEC actin cytoskeleton rearrangements and related signaling pathways, and traversal of BBB as live bacteria (10). Following the initial interaction between the host and E. coli, the bacteria are progressively drawn into the host by a "zipper mechanism," which utilizes the bacterium-induced actin filaments for bacterial uptake (43). K1 capsule, OmpA, Ibe proteins, AslA, TraJ, and CNF1 have been identified to be important virulence factors using in vivo and in vitro models (10). However, it is unclear how all of these E. coli determinants contribute to BMEC invasion in vitro and crossing of BBB in vivo. Previous investigations have determined that E. coli K1 capsule contributes to the high level of bacteremia, which was required for the development of meningitidis (12). It has been suggested that PSAs can mimic the host and escape detection by the immune system (44). Moreover, recent evidence suggests that the K1 capsule is not only necessary for invading BMECs but also responsible for maintaining the viability of the bacteria inside the BMECs (13). These observations strongly suggest an important role of K1 capsule in the traversal of BMEC.
The biosynthesis of K1 capsular ␣2,8-linked polysialic acids is involved in the activation NeuAc and polymerization of activated NeuAc. CMP-NeuAc synthetase, the enzyme that catalyzes the activation of NeuAc, has been extensively studied (23,45,46). Previously, we cloned a CS gene from E. coli K1 strain 44277 (29). It has been noticed that molecular weight of E. coli CS is almost twice that of other bacterial species, and all conserved structural motifs required for NeuAc activation were concentrated in the N-terminal half of E. coli CS. Thus, we hypothesized that the C termini of E. coli CS was not required for CS activity. Evidence we obtained definitely supports our hypothesis. The investigation of truncations at the C terminus revealed that the CS mutant (CS 229 ) was still active when up to 189 amino acids were removed from the C terminus; however, activity loss and less stability were observed. By taking advantage of structural information obtained from three-dimensional modeling, CS 229 was found to be the minimal truncation that could maintain the tertiary structure required for synthetase activity (Fig. 3). Our evidence also suggested that the C-terminal half (residues 230 -418) of E. coli CS could stabilize the CS catalytic domain. However, it is hard to understand why E. coli produces such a long peptide at the C terminus only for stabilization. The first idea was that the C-terminal half would interact with the putative dimerization domain in the N-terminal half and probably serve as another "subunit" as occurred in the dimer of N. meningitidis CS. It is possible that the C-terminal half is an inactive subunit derived from naturally occurring mutations of a previously functional subunit. Therefore, we carried out the three-dimensional structure prediction analysis based on the homologous sequence search of the Cterminal half (residues 230 -418). However, no homologous protein was found in GenBank TM . Then we predicted the threedimensional structure of the C-terminal half based on the homologous search of secondary structure in PDB. Most interesting, the secondary structure of the C terminus was homology with the ␣1-subunit of PAF-AH isoform I instead of N. meningitidis CS subunit-like domain.
PAF is one of the most potent lipid messengers involved in a variety of physiological events. The acetyl group at the sn-2 position of its glycerol backbone is well known for its biological activity, and its deacetylation induces loss of activity. The enzyme that catalyzes the deacetylation is PAF-AH, which was first reported in human sera (47) and belongs to calcium-independent phospholipase A 2 family. PAF-AH has been found widely distributed in mammalian tissues and blood so far. At least two types of PAF-AH exist in mammalian tissues, namely the intracellular (cytosolic) and extracellular (plasma) type. Arai et al. (48) found two types of intracellular PAF-AH from bovine brain, PAF-AH I and II. PAF-AH I is a heterotrimer composed of 29-(␣1), 30-(␣2), and 45-kDa (␤) subunits (35,38). The amino acid sequences of the three subunits show extremely high homologies among the mammalian species (31). Both ␣1 and ␣2 have a catalytic center (31). It has been shown that ␣1-␣2 complex can express the full level of activity (49). When ␣1 and ␣2 are purified or overexpressed in E. coli, they form catalytically competent homodimers (37). Although the dissociation of the dimeric protein into monomers leads to the inactivation and destabilization of protein (50), only one subunit is active in the heterodimer or homodimer (37,38). The biochemical differences of three possible catalytic dimers, ␣1/␣1, ␣1/␣2, and ␣2/␣2, have been studied. The ␣2/␣2 homodimer is significantly more active against PAF and 1-O-alkyl-2-acetyl-snglycerol-3-phosphoric acid than ␣1/␣1 and ␣1/␣2, whereas both homodimers have comparable catalytic efficiency against PAF (39,51). The ␤-subunit binds to all three catalytic dimers. The binding of ␤-subunit accelerates the hydrolytic rate of the ␣2/␣2 homodimer for PAF about 4 times, slightly suppresses the hydrolytic rate of the ␣1/␣1 homodimer, and has little effect on that of the ␣1/␣2 heterodimer (51). It is therefore proposed that the enzyme activity of PAF-AH I is regulated in multiple ways by altering the composition of catalytic subunit and by manipulating the ␤-subunit (48). Ser 47 has been identified as an active serine residue in ␣1. In the active center, Ser 47 , Asp 192 , and His 195 have been revealed to form a catalytic triad (37). By comparing the three-dimen-sional model of PAF-AH I ␣1 (PDB 1fxw:A), the putative catalytic serine was identified as Ser 257 , and the catalytic triad was composed of Ser 257 , Asp 397 , and His 400 (Fig. 4). Thus, we tested PAF-AH activity by using [ 3 H]acetyl-PAF as substrate. Our experimental results showed, as expected, that the wild type E. coli CS was able to release acetyl group from PAF. During the purification of CS protein, the purification fold of PAF-AH activity was almost the same as that of CS activity, which clearly demonstrated that the PAF-AH activity was associated with CS protein. In addition, the mutant analysis confirmed that the residues from 228 to 418 (CS 228 -418 ) could form a functional PAF-AH-like protein. With this evidence, we concluded that E. coli CS was a bifunctional enzyme possessing a CS activity in the N-terminal half and a PAF-AH activity in the C-terminal half.
Although we were unable to detect the dimer of CS 228 -418 on native PAGE (data not shown), the indirect evidence still suggested the existence of catalytic dimer formed by CS 228 -418 . The 48.6 kDa of pure CS protein exhibited a specific activity of 1.4 units/mg for PAF-AH activity (Table III). If the 22.5 kDa of pure CS 228 -418 protein existed as a monomer, then its theoretical specific activity should be at least twice that of CS (Figs. 5 and 6). However, according to our results, the specific activity of CS 228 -418 was 1.7 units/mg (Table IV). This unexpected lower specific activity of CS 228 -418 was not caused by instability of the mutant, because we found that CS 228 -418 was more stable than CS when we incubated either CS or CS 228 -418 at 37°C for 1 h. Therefore, the only plausible explanation is that CS 228 -418 functions as a dimer, and only one subunit exhibits activity. The slight increase of the specific activity displayed by dimeric CS 228 -418 can be explained by the difference in purity of CS and CS 228 -418 . The crystal structure of bovine ␣1-subunit reveals that the N-terminal ␣-helix (H1), together with the loop that follows, is part of the dimer-forming interface, which contains a significant number of charged and polar residues (38,42). As revealed by three-dimensional modeling of CS 228 -418 (Fig. 4), the putative H1 ␣-helix formed by the residues from 231-242 (YQNIHNRINEKR) could serve as part of dimer-forming interface. According to our results, both CS 230 -418 and CS 229 -418 were inactive. By fusing a His tag with the N terminus of inactive mutant CS 230 -418 , we were able to restore the PAF-AH activity (data not shown). These observations suggested that only H1 was not enough for dimer forming.
PAF-AH I has a strong specificity for the acetyl group attached to the glycerol backbone of PAF, whereas phospholipids having a head group other than choline are also recognized by the enzyme (51). This specificity is ascertained to Leu 48 , Leu 194 , and Thr 103 , which are responsible for the substrate specificity of ␣1 by forming contact with the methyl group of acetate (38). On the other hand, PAF-AH II shows substrate specificity similar to plasma PAF-AH. Both enzymes hydrolyze short chain diacylglycerols, triacylglycerols, and acetylated alkanols and display phospholipase A 1 activity. These two enzymes do not distinguish between an ester or an ether at the sn-1 position of PAF or PAF analogues, and both can hydrolyze phospholipids with short to medium length sn-2 acyl chains including truncated chains derived from oxidative cleavage of long chain polyunsaturated fatty acyl groups (52,53). As for the PAF-AH domain in E. coli CS, the specificity was predicted to be similar to PAF-AH I since the amino acid residues corresponding to Leu 48 , Leu 194 , and Thr 103 in ␣1-subunit of PAF-AH I were found to be conserved (identified as Leu 399 , Leu 258 , and Thr 308 ). The specificity analysis with the wild type CS or CS 229 -418 supported our prediction. Among the phospholipids tested, only PAF, glycerol triacetate, and p-nitrophenyl butyrate could be hydrolyzed. These results indicated that the PAF-AH domain of E. coli CS was specific for PAF and exhibited lipase-and esterase-like activities.
A large family of enzymes that have an essential serine residue possesses the consensus sequence, Gly-Xaa-Ser-Xaa-Gly. By comparing the amino acid sequences between ␣1-sub-unit of PAF-AH I and E. coli CS, it is interesting to note that the second Gly is substituted by Val in ␣1, which has been thought to contribute to the specificity against the short chain acyl moiety (37). In the PAF-AH domain of E. coli CS, the second Gly residue from the active serine is substituted by Phe. Thus, the preference of short chain acyl moiety for the Cterminal half of E. coli CS could be explained by this substitution. Another interesting fact is that the amino acid residue between the first Gly and the active serine residue in ␣1 is Asp, which is the same as those of most serine proteases but not those of serine esterase, in which the residue between the first Gly and the active serine residue is His, suggesting that PAF-AH I is closer to that of proteases other than that of lipases. In the PAF-AH domain of E. coli CS, the residue between the first Gly and the active serine residue is His, suggesting that the C-terminal half of E. coli CS is closer to that of lipases. As a matter of fact, the lipase activity of the PAF-AH domain of E. coli CS has been detected with glycerol triacetate, a substrate for lipase. Moreover, it has been shown that the substitution of Tyr 191 of ␣1 with Phe that presents in ␣2 gives the ␣1-subunit a substrate specificity similar to that of ␣2 (37). Tyr 191 was also found to be conserved in the C-terminal half of E. coli CS. By taking the structural features and substrate specificity analysis together, it can be concluded that the specificity of C-terminal half of E. coli CS is similar to that of the ␣1-subunit.
The evidence we present in this report clearly demonstrates that the C-terminal half of E. coli CS functions as ␣1-subunit of PAF-AH I. However, the physiological role of this functional domain has not been defined yet. It has been observed that the expression of ␣1-subunit is restricted in early developing neurons at embryonic stages of rat (54), and the highest expression level of ␣1 is also detected in the human fetal brain as well (55). In addition, certain PAF-AH inhibitors, which do not bind to PAF receptors, can inhibit migration of neurons in cultures (56). Thus, the ␣1-subunit has been suggested to play a role in neuron migration during development. Moreover, the significant levels of ␣1-subunit are expressed in the kidney, thymus, and colon of the adult human, in which cell migration remains prominent under normal physiological, as well as pathological, conditions. It is therefore thought that cells expressing the ␣1-subunit possess the ability to migrate in adult tissue. On the other hand, a recent report that ␣1 is not required for embryonic and early postnatal development in mouse makes the story of ␣1-subunit even more complicated (57,58). It is now recognized that the ␣-subunit is a member of an old family of hydrolytic proteins, identified some years ago as a family of lipolytic enzymes (59,60). Therefore, it is possible that the natural substrate of the presumed PAF-AH I could be an as yet unidentified molecule with an acetyl moiety attached via an ester bond (39). Obviously, the elucidation of endogenous substrate is of importance not only for the physiological function of mammalian ␣1-subunit but also for that of the E. coli ␣1subunit-like domain. Somehow, considering the correlation of mammalian ␣1-subunit to cell migration, it is possible that the E. coli ␣1-subunit-like domain may play a role in E. coli motility during the infection of BMEC.
In our investigation, we found that the CS activity of the wild type enzyme had a pH optimum of 9.0 and that the truncated CS 229 mutant was more active at pH 7-8. In contrast, we found that the PAF-AH activity of the wild type was optimal at 7-8, whereas the truncated mutant was most active at pH 9.0.
Apparently, it appears that the wild type enzyme mainly functions as CS at pH 9.0 and serves PAF-AH at pH 7-8. However, it is now hard to understand why the wild type enzyme exhibits its highest CS activity at pH 9.0 because it catalyzes the CMP-NeuAc synthesis in cytosol, which is a neutral environment. Although we did not detect PAF-AH domain alone in vivo, the proteolytic analysis showed that the PAF-AH could exist in vivo under some unknown conditions. Probably, in an environment of pH 9.0, the wild type enzyme or PAF-AH domain derived from proteolysis mainly functions as PAF-AH, which may be a clue to explore the physiological function.
In summary, we have demonstrated that the E. coli CS is a bifunctional enzyme with a CS activity in the N-terminal half and a PAF-AH-like activity in the C-terminal half. The tertiary structure of PAF-AH-like domain is significant homology with that of ␣1-subunit of bovine brain PAF-AH I. The biochemical properties of PAF-AH-like domain in E. coli CS are essentially similar to that of bovine brain ␣1-subunit. This protein is the first ␣1-subunit-like protein recognized in microorganisms. At present the in vivo substrate and the physiological function of this novel functional domain are still unclear. Our next challenge will be to identify the in vivo substrate for this domain and explore its physiological function.