Penicillin-binding protein 1B from Escherichia coli contains a membrane association site in addition to its transmembrane anchor.

A working structural model of penicillin-binding protein 1B (PBP 1B) from Escherichia coli derived from previous data consists of a highly charged aminoterminal cytoplasmic tail, a 23-amino-acid hydrophobic transmembrane anchor, and a 758-amino-acid periplasmic domain. Using an engineered thrombin cleavage site, we have investigated the solubility properties of the periplasmic domain of PBP 1B. Twelve amino acids, comprised of the consensus thrombin cleavage site (LVPR decreases GS) and flanking glycine residues, were inserted into PBP 1B just past its putative transmembrane segment. To aid in purification, a hexa-histidine tag was also inserted at its amino terminus, and the engineered protein (PBP 1B-GT/H6) was purified and characterized. Inclusion of the thrombin cleavage site had no effect on the protein's intrinsic tryptophan fluorescence and affinity for [14C]penicillin G, indicating that the protein structure was not significantly perturbed. PBP 1B-GT/H6 was readily cleaved by thrombin at low thrombin/protein ratios to a protein with properties consistent with the removal of its cytoplasmic tail and transmembrane regions. Cleavage of the protein was dependent upon the presence of the thrombin cleavage site, and the thrombin-cleaved protein (PBP 1Bper) displayed an identical affinity for [14C] penicillin G binding as wild-type PBP 1B and uncleaved PBP 1B-GT/H6. [14C]Penicillin G-labeled PBP 1Bper eluted from a gel filtration column in the presence but not in the absence of 0.7% 3-[(3-cholamidopropyl)dimethylammonio]-1- propanesulfonic acid, and PBP 1Bper was found entirely in the membrane fraction of a thrombin digest of membranes containing overproduced PBP 1B-GT/H6. To further characterize this unusual solubility behavior, purified PBP 1Bper was reconstituted into lipid vesicles, which were then floated on a sucrose gradient. Floated vesicles contained > 95% of total 125I-penicillin V binding, indicating that PBP 1Bper directly associates with lipid membranes. These results strongly suggest that the periplasmic domain of PBP 1B associates with membranes independent of its amino terminal transmembrane region.

Penicillin-binding proteins (PBPs)' are membrane-bound enzymes that catalyze the final stages of peptidoglycan synthesis in bacterial cells (1, 2). Penicillin, due to its structural similarity to the D-alanyl-D-alanine carboxyl terminus of the peptide moiety of peptidoglycan, exerts its lethal effect by binding to and covalently reacting with PBPs, thus preventing the cross-linking of the peptide chains of the peptidoglycan. In Escherichia coli there are at least seven PBPs, which can be divided into two groups based upon genetic and biochemical analysis. The high molecular weight PBPs, PBPs lA, lB, 2 and 3, are involved in diverse cellular processes such as cell elongation, cell shape, and cell division, whereas the low molecular weight PBPs, PBPs 4, 5 and 6, are not by themselves essential for cell growth, since each can be deleted without apparent deleterious effects (reviewed in Ref. 1). Recent evidence, however, suggests that the activity of either PBP 5 or 6 is involved in cell growth by regulating the availability of peptidoglycan precursors, especially when the activity of PBP 3 is limiting (3). All of the PBPs contain a serine residue in the active site of the penicillin-binding portion of the protein that forms a covalent bond with penicillin (4)(5)(6)(7). The roles of other specific amino acids in the catalytic cleft of the penicillin-binding region of PBPs has been the focus of several recent reports (8,9).
Several studies have focused on the function of PBP 1B. PBPs 1A and 1B are thought to be involved in cell elongation; mutants lacking either PBP 1A or lB, but not both, are viable (10, 11), and therefore each of these PBPs is able to compensate for the absence of the other. Mutants lacking PBP 1B have only about 10% of the peptidoglycan synthesis activity of wild-type cells, however, and thus PBP 1B is thought to be the major peptidoglycan synthetase in E. coli (12,13). Deletion of PBP 1B also results in a decreased growth rate and an increased sensitivity to lysis by the inactivation of PBP 2, either by specific antibiotics or by mutations that alter its stability (10,14). The peptidoglycan synthesized by PBP 1A alone is therefore likely to be partially defective, even though it is sufficient to support growth. Purified PBP 1B has been shown to catalyze both a penicillin-sensitive transpeptidase activity and a penicillin-insensitive transglycosylase activity (15,16). Several experiments with PBP 1B have localized these two activities in the primary structure of the protein.
Cleavage of PBP 1B with low amounts of trypsin results in the formation of a 50-kDa fragment that retains penicillin binding (16,17), which suggests that the penicillin-binding portion of the protein exists as a proteolytic domain. Since the active site serine residue of PBP 1B is at residue 510 (5), and the same residue in the related lower molecular weight PBPs (which are monofunctional) is only 40-50 amino acids in from the amino terminus, it follows that the penicillinbinding domain (and by inference the transpeptidase domain) is located in the carboxyl-terminal half of the PBP 1B sequence. The best evidence for the location of the transglycosylase domain comes from the measurement of transglycosylase activity in membranes from cells producing truncated forms of PBP lB, in which various portions of the carboxyl terminus of PBP 1B were removed by deleting the corresponding DNA in an expression plasmid. These studies indicated that deletion of residues 423-844 or 478-844 resulted in the retention of significant transglycosylase activity in the membrane fraction, but the new peptidoglycan synthesized by these membranes was not cross-linked (16). Thus, a working model emerges for PBP 1B in which the amino-terminal portion of the protein is the transglycosylase domain that polymerizes the glycan chains, and the carboxyl-terminal portion is the transpeptidase domain that catalyzes the crosslinking of the peptide chains.
On SDS-polyacrylamide gels, PBP 1B consists of at least three bands of varying intensity, termed a , @, and y. Genetic and biochemical analysis has determined that the a and y components represent two distinct translation start sites within the p o d gene that encodes PBP 1B and that @ is derived from CY by proteolysis (18,19). Each of the three components has been isolated from SDS-polyacrylamide gels, renatured, and shown to possess both transpeptidase and transglycosylase activities (20). Complementation studies in a PBP lAta PBP 1Bstrain have also shown that the y component alone is sufficient to rescue growth at the restrictive temperatures. These same studies have also shown that the last 64 amino acids at the carboxyl-terminal do not contribute to its function, as deletion of these amino acids still results in a functional protein as assayed by penicillin binding and complementation (18).
The topology of PBP 1B has also been investigated in some detail. Analysis of the hydropathy plot of PBP 1B predicts a single stretch of 25 hydrophobic amino acids at residues 64-87, indicating that this region may function as a both a signal sequence and transmembrane anchor. In an elegant set of experiments, the coding region for the mature form of TEM @-lactamase was fused in frame to random positions along the entire coding region of PBP lB, and the ability of these fusions to confer resistance to ampicillin was investigated (21). All p-lactamase fusions from residues 88 to 844 of PBP 1B resulted in high level resistance to ampicillin, which indicated that this region was periplasmic. These experiments reinforced the topology of PBP 1B suggested by its hydropathy plot and led to the model in which residues 1-63 are intracellular, 64-87 are transmembrane, and 88-844 are periplasmic. These results fit the generally accepted model that PBPs are essentially periplasmic enzymes that are anchored in the cytoplasmic membrane by a single transmembrane segment (22-25). This model therefore predicts that residues 88-844 would make a candidate for obtaining a soluble form suitable for crystallization studies of PBP 1B. However, in this study we report that a truncated protein comprising residues 90-844 is not soluble in the absence of detergents and interacts with membrane vesicles in a manner indistinguishable from native protein. We conclude from these experiments that PBP 1B interacts with the membrane at a region distinct from its transmembrane anchor region, although the nature of this interaction is as yet unknown. with M13 vectors, CJ236 (dut ung) for the production of uracilcontaining M13 phage, and NK5830 (F' lacP prolarg Ahc proxllr NalA rif recA-56 suo ara thy) for the overexpression of PBP 1B constructs. Luria-Bertani broth (LB) was used for bacterial growth and was supplemented with 50 pg/ml of kanamycin or 34 pg/ml of chloramphenicol when appropriate. Sequencing was performed with the Sequenase I1 kit from United States Biochemical.
Construction of PBP 1B Containing a Thrombin Cleavage Site and an Amino-terminal Hexahistidine Tag-The gene for PBP 1B-7 (hereafter referred to only as PBP lB), which was modified by mutagenesis to contain 34 bases 5' of the second translation start site of the ponB gene (27), was cloned into pTTQ18K (28) as an EcoRI-BamHI fragment to yield pRN1B-E. The EcoRI-SaZI fragment of pRN1B-E, which encodes the amino-terminal half of PBP 1B, was subcloned into M13mp19 and used to make uracil-containing singlestranded DNA temdate in CJ236 as described bv Kunkel et al. (29). The following oligonucleotide, 5"GGCGTTTATCTCGATCA-ATTCGTAGCCGT-3'. was used to loop-in the codons for 12 extra AGGTGGTCTGGTTCCGCGTGGATCCGGTGGTGGTGGTAAAamino acids (underscored; also see Fig: 1). Addition of these extra amino acids results in the formation of a thrombin cleavage site (LVPRGS) flanked by glycine residues just past the putative membrane-spanning region of PBP 1B (see Fig. 1A). Mutants containing a BamHI site (from the loop-in oligonucleotide) were identified and sequenced to ensure that correct insertion had occurred.
Mutant phage containing the thrombin cleavage site were then used to make uracil-containing single-stranded DNA terndate. and the following oligonucleotide, 5'-CCGATGCCGCGCAAACACCAT-CACCATCACCATGGTAAGGGCAAAGGC-3' was used to loop-in the codons for 6 histidine residues (underscored) along with a silent NcoI site. This same oligonucleotide was also used io loop-in the hexahistidine tag into the coding sequence of PBP 1B. The 6 histidine residues were located after the fourth amino acid of either wild-type PBP 1B or PBP 1B containing the thrombin cleavage site and provided a means to rapidly purify these proteins on immobilized Ni2+-chelated NTA resin (30). Positive phage were identified by the presence of the NcoI site and sequenced; to generate the expression plasmid for the modified PBP, the EcoRI-Sac11 fragment from one of the positive phage was used to replace the corresponding fragment in pRN1B-E, yielding either pRNlB-Hs or pRNIB-GT/H, (Fig. IB).
Expression and Purification of PBP lB-GT/H6-Cu~tures (6 liters) of NK5830/pRN1B-GT/H6 were grown to an ODGoo of 0.5, and IPTG was added to 300 p M to induce transcription from the tac promoter. After inducing for 2 h, the cells were harvested by centrifugation and the cell paste was frozen at -20 "C overnight. The next morning, the cells were resuspended in 40 ml of 50 mM sodium phosphate buffer, pH 7.0, 1 mM phenylmethylsulfonyl fluoride (lysis buffer), and lysed by three passes in a French pressure cell at 13,000 pounds/square inch. The cell lysate was then centrifuged at 100,000 X g for 1 h and the supernatant discarded. The membrane pellet was resuspended in lysis buffer, and the protein concentration was determined by the Lowry method (31). The membranes were diluted to 10 mg/ml with phosphate buffer, mixed with an equal volume of 2% CHAPS, 2 M NaCI, 50 mM Tris. HCI, pH 7.5, and incubated at room temperature for 1 h to extract the 1B proteins. Following extraction, the mixture was centrifuged at 120,000 X g for 1.5 h. The supernatant was mixed with an equal volume of 66% ammonium sulfate, incubated on ice for 1.5 h or overnight at 4 "C, and the precipitated PBP lB-H6 or PBP lB-GT/Hs was collected by centrifugation at 15,000 X g for 15 min. The precipitates were redissolved in 0.7% CHAPS, 0.5 M NaCI, 25 mM Tris.HC1, pH 8.0 (buffer A), and applied to an NTA column (3ml bed volume) equilibrated in the same buffer. The columns were washed with 10 volumes of buffer A containing 1 M NaC1, followed by 10 volumes of buffer A containing 10 mM imidazole. PBP 1B-H6 or PBP 1B-GT/H6 were eluted with buffer A containing 120 mM imidazole. In some experiments, the CHAPS detergent was exchanged for 8-octylglucoside by washing the bound PBP 1B-GT/H6 with 10 volumes of 0.7% 8-octylglucoside, 0.5 M NaC1, 25 mM Tris.HC1, pH 8.0, 10 mM imidazole, and then eluting with the same buffer contain-

Periplusmic Membrane Association of PBPl B
ing 120 mM imidazole. Fractions containing PBP 1B-H6 or PBP 1B-GT/& were identified by absorbance at 280 nm, pooled, and the imidazole was removed either by dialysis or by dilution/ultrafiltration. Fluorimetric Analysis of Purified PBP IB-H6 and PBP IB-GT/ &"The concentrations of the purified proteins were determined by the method of Lowry and by ODaso. The proteins were then diluted into either buffer A or 6 M guanidine. HCI, 25 mM Tris. HCI, pH 7.5, to make a 0.25 p M solution and their emission spectra analyzed on a Perkin-Elmer model MPF-2A Fluorimeter. The proteins were excited at 290 nm, and the emission spectra were recorded with a 12-nm slit width.
Thrombin Cleavage of PBP IB-GT/He and Purification of PBP IB,,-Purified PBP 1B-GT/Hs in buffer A was incubated at various thrombin/protein ratios for 1 h at 37 "C. Following incubation, the proteolyzed samples were subjected to SDS-polyacrylamide gel electrophoresis on an 8% gel. For large scale purification, PBP lB-GT/ HB (0.4 mg/ml) was digested with thrombin at a ratio of 1:250 (thrombin/protein) for 1 h at 37 "C, and then reapplied to a Ni2+-NTA column as described above. The eluant from the Ni2+-NTA column, which contained thrombin-cleaved PBP 1B-GT/H6 (PBP 1BpJ purified away from any undigested protein, was then used for subsequent experiments. Gel Filtration of PBP lB,, in the Presence and Absence of Detergents-Purified PBP lB,, in buffer A was incubated with ["C] penicillin G (40 pg/ml) for 10 min at 30 "C. The labeled protein was then layered onto a Sephacryl HR-200 column (1 X 50 cm) equilibrated in 25 mM Tris.HC1, pH 8.0, 500 mM NaCl, k 0.7% CHAPS, and eluted in the same buffer. Fractions (0.4 ml) were collected, and 140-pl aliquots were analyzed by liquid scintillation counting in 5 ml of ScintiVerse BD (Fisher). The column was standardized by running the following compounds under both elution conditions: blue dextran (void volume), bovine serum albumin (68 kDa), and 3H20 (included volume).
Reconstitution of PBP IB Proteins-Reconstitution of PBP 1B proteins into lipid vesicles and flotation on sucrose gradients was done essentially as described by Waxman and Strominger (32). An aliquot (3 mg/experiment) of asolectin lipid, purified as described by Kagawa and Racker (33) and stored at -20 "C in CHC13/MeOH (2:1, v/v), was dried under a stream of argon in a 15-ml polypropylene tube. The residue was twice redissolved in diethyl ether and reevaporated under argon. The lipid was redissolved and then sonicated (three 20-s bursts) in 0.2 mi of 4.4% 8-octylglucoside, 100 mM Tris. HC1, pH 8.0, followed by the addition of 60 pg of protein in 0.5 ml of 0.7% 0-octylglucoside, 25 mM Tris.HC1, pH 8.0, 500 mM NaCl. The clear solution was then dialyzed at 4 "C versus three changes of 500 ml of 25 mM Tris. HCI, pH 7.5, 30 mM sucrose, 0.5 M NaCl. Sucrose was omitted from the final dialysis buffer.
Flotation of Lipid Vesicles on a Sucrose Gradient-The slightly opaque samples of reconstituted lipid vesicles were spun at 750 X g for 10 min to remove insoluble material, and 0.25 g of sucrose was dissolved in 0.38 ml of the vesicle preparations. The samples were placed in a 5-ml centrifuge tube, layered with a linear gradient of 42 to 12% sucrose, and spun for 16 h at 37,000 rpm in a,SW 50.1 rotor (Beckman Instruments). Under these conditions, the vesicles floated to the top of the sucrose gradient. Fractions of 0.3 ml were collected, and the location of the lipid vesicles was ascertained by dilution of 50 pl into 450 pl of 25 mM Tris.HCI, pH 7.5, 0.5 M NaCl and determining the light scatter at 350 nm. The location of PBP proteins was determined by incubating 20 pl of each fraction with 30 pg/ml of ['261]penicillin V for 10 min at 30 "C followed by SDS-PAGE and silver staining to locate the protein bands. After drying the gel, the protein bands were cut out and counted in a gamma counter. In intermediate fractions where there was no visible silver staining, the area corresponding to the position of the protein was also removed and counted.
SDS-Polyacrylamide Gel Electrophoresis-Samples prepared as described above were submitted to electrophoresis on an 8% SDSpolyacrylamide gel (34). For samples incubated with either [3H] or ['4C]penicillin G, the gel was treated with EN3HANCE (Du Pont-New England Nuclear), dried, and exposed to Fuji RX film. For location of protein bands, the gel was stained with either silver or Coomassie Brilliant Blue R-250. For the estimation of band intensity on x-ray film, fluorographs were submitted to densitometry on a Hoeffer HS300 scanning densitometer hooked to an Apple11 computer.

RESULTS
Properties of a PBP 5-PBP 1B Fusion Protein-The apparently straightforward topology of PBP 1B as predicted by a hydropathy plot and 8-lactamase fusion experiments suggests that a soluble form could be constructed in which residues 1-87 were replaced by an exogenous signal sequence. This in fact was reported by Spratt and co-workers (35), in which the signal sequence and first 10 amino acids of mature PBP 5 were fused in frame with residues 88-844 of PBP 1B. Expression of this construct resulted in the appearance of a new PBP of the expected size, but the PBP was recovered only partly in the soluble fraction, and the biochemical characterization of the PBP was not pursued. Our laboratory also constructed a similar fusion protein, except that the signal sequence of PBP 5 was fused directly with residues 93-844. In our hands, the PBP 5-1B fusion protein was recovered mostly in the membrane fraction (>85% as determined by immunoblotting), with the remainder appearing in the supernatant (data not shown). When the two fractions were analyzed by [14C]penicillin G binding, however, the supernatant fraction appeared to be largely inactive. Because the majority of ['4C]penicillin G binding was present in membranes, we extracted the fusion protein from membranes with CHAPS detergent and purified it in reasonable quantities. When this protein was submitted to gel filtration in the presence 0.7% CHAPS, it eluted at a position expected of a protein it's size.
When the detergent was removed, however, no protein eluted from the column, suggesting that residues 93-844 were insoluble in the absence of detergent (data not shown). An alternative explanation of this result was that the signal sequence from PBP 5 was not cleaved during transport and mediated the insolubility of the protein due to its hydrophobic core. This latter possibility was tested by sequencing the purified protein. Unfortunately, no sequence could be obtained, and thus we were still not able to unequivocally distinguish between the two scenarios.
Construction of pRN1B-GT/H6-Because of some of the difficulties of growing the strain and purifying the protein from it, the unexpected results of the solubility of the fusion protein, and the possibility that the signal sequence was still attached, we decided to try a different approach for investigating the behavior of the periplasmic portion of PBP 1B. In developing this study, we reasoned that inclusion of several amino acids just on the carboxyl-terminal side of the putative membrane-spanning region of PBP 1B might be tolerated and still allow the protein to fold correctly. We therefore used mutagenesis to insert the codons for a thrombin cleavage site (Leu-Val Pro-Arg-Gly-Ser) into PBP 1B (PBP 1B-GT, Fig.  lA). Thrombin was chosen as a protease due to its availability and high specificity. Because of the possibility that thrombin would not be able to recognize this site so close to the folded structure of the periplasmic portion of PBP lB, we also introduced 2 glycine residues at the amino end and 4 glycine residues at the carboxyl end of the cleavage site. Previous results by Guan and Dixon (36) have shown that inclusion of glycine residues past the thrombin cleavage site of glutathione-S-transferase fusion proteins resulted in a much more efficient cleavage of the recombinant protein.
We also desired an easy method for purifying our PBP 1B proteins without using ampicillin affinity chromatography. The harsh conditions for elution (1 M NHzOH, p H 8.7, for 3 h) from the ampicillin resin and reports that PBP 1B purified in this manner exhibited very low stoichiometry of penicillin binding (16) made this purification procedure unattractive. tagged protein on an immobilized Ni2+-chelante resin (NTA) has been pioneered by Hochuli and colleagues (30). This method has several advantages: 1) it is a mild chromatographic procedure, 2) chromatography can be carried out in 0.5 M NaCl (purified PBP 1B precipitates in low salt buffers), and 3) it allows the purification of the tagged protein in a single chromatographic step. We therefore introduced codons for 6 histidine residues in between the fourth and fifth amino acid codons of the gene encoding both PBP 1B and PBP 1B-GT (Fig. 1A). The reconstructed expression plasmid containing the gene for PBP 1B with both the hexahistidine and thrombin cleavage site insertions ( pRNlB-GT/H6) is shown in Fig. 1B.
Expression and Purification of PBP 1 B-GT/Hs-NK5830 cells harboring pRNlB-GT/HG were grown to an ODso0 of 0.5, and the tac promoter was induced with 300 PM IPTG for 2 h. Fig. 2, lunes 2 and 3, shows an immunoblot of the expression of the protein before and after the addition of IPTG. To determine whether the expressed protein would bind radiolabeled antibiotic, the cell lysates were incubated with [''C] penicillin G, submitted to electrophoresis, and prepared for fluorography (Fig. 2, lanes 4 and 5). These results indicated that the expressed protein was produced in high amounts and was capable of binding [14C]penicillin G.
Purification of PBP 1B-GT/H6 (and PBP lB-H6) was accomplished as follows: membranes from IPTG-treated cells were extracted with 1% CHAPS and 1 M NaC1. This extraction resulted in the solubilization of 50-70% of PBP lB-GT/ H6, which was then precipitated from the detergent solution in 33% ammonium sulfate. The ammonium sulfate precipitation itself results in a significant purification step and allowed the protein to be dissolved in a small volume for purification on a Ni2'-NTA column. The precipitate was redissolved in 25 mM Tris.HC1, pH 8.0, 0.5 M NaC1, 0.7% CHAPS and applied to a Ni2+-NTA column, and after extensive washing to remove unbound proteins, PBP 1B-GT/H6 was eluted with the above buffer containing 120 mM imidazole. PBP lB-GT/HG at this stage was essentially a single band on a Coomassie-stained SDS-polyacrylamide gel (Fig. 3A, lane  6). A gel indicating the purity of PBP 1B-GT/H6 at various stages of the purification is shown in Fig. 3A.
The samples from steps along the purification were also assayed by ['Hlpenicillin G binding and fluorography (Fig.   3B, lanes 8-13). Interestingly, the intensity of the labeled PBP 1B-GT/H6 band greatly increased upon extraction from the membrane with 1% CHAPS, 1 M NaC1, which indicated that the expressed protein in membranes was not binding a stoichiometric amount of labeled antibiotic. This latency of antibiotic binding has also been observed for overexpressed PBP 1B and is not a result of the extra amino acids inserted into PBP 1B-GT/H6. We believe this behavior may be due to the sequestering of the overexpressed protein at membrane adhesion sites, which prevents the access of ['Hlpenicillin G to the binding sites of PBP lB-GT/Hs (see "Discussion").
Biochemical Analysis of P B P lB-GT/HG-Because of the possibility that the addition of the extra amino acids in PBP 1B-GT/H6 might alter the structure of the protein by causing it to fold incorrectly, we sought to quantitatively assess the integrity of the protein. To address the penicillin binding activities of the purified proteins, the affinities of PBP 1B-Hs and PBP 1B-GT/H6 for ['4C]penicillin G were determined as described under "Experimental Procedures.'' Because of the covalent nature of the interaction, the affinity of a PBP for radioactive antibiotic is defined as the concentration of that antibiotic that results in the covalent labeling of 50% of the PBP under defined conditions (in this case, 10 min at 30 "C; see Ref. 37 for further details). The saturation curves are shown in Fig. 4. From these data, it was determined that both proteins had an identical affinity for ['4C]penicillin G (-2.3 pg/ml or 6.2 nmol/ml). These values are very similar to the value reported for PBP 1B in E. coli membranes (3.0 pg/ ml; (38)).
We also sought to compare the overall structure of PBP lB-He with PBP lB-GT/Hs by using intrinsic tryptophan fluorescence. Both of these proteins have 13 tryptophan residues; 2 are in the transmembrane anchor, 3 are in the aminoterminal half, and 8 are in the carboxyl-terminal half of the protein. Equimolar solutions (0.25 p~) of PBP lB-Hs and PBP 1B-GT/H6 in 25 mM Tris. HC1, pH 8.0, 500 mM NaCl, 0.7% CHAPS were excited at 290 nm, and a scan of their respective emission spectra is shown in Fig. 5. Also shown in Fig. 5 are the emission spectra of the two proteins in 6 M guanidine. HCI, 25 mM Tris. HCl, pH 8.0. It is clear from these data that the intrinsic fluorescence of PBP lB-GT/He under native conditions is superimposable with PBP lB-Hs ( X , , , = 334 nm). Upon denaturation, the emission peak decreased by 50% and was shifted 10 nm to higher wavelength. Since the spectra for the two proteins are superimposable, it is very likely that the two proteins have nearly identical three-

FIG. 2. Expression of PBP lB-GT/Hs in NK5830 cells.
NK5830 cells harboring pRNlB-GT/HG were grown to 0.5 ODm and then induced for 2 h with 0.3 mM IPTG. Cells from 1 ml of culture were pelleted, resuspended in 100 pl of phosphate buffer, and subjected to two cycles of freeze/thaw. The mixture was then treated with RNase A and DNase I until the solution was no longer viscous. dimensional structures, and therefore the additional amino acids in PBP lB-GT/Hs have not altered the folding of the protein.

Cleavage of PBP lB-GT/He with Thrombin and Purification
of the Cleaved Protein-To determine whether thrombin would cleave PBP lB-GT/He and to assess its specificity, thrombin digests of both PBP lB-Hs and PBP lB-GT/Hs were submitted to electrophoresis on an 8% SDS-polyacrylamide gel, which easily separated undigested protein from cleaved product. The purified PBP 1B proteins were incubated with thrombin at either a 1:lOO or a 1:500 ratio (thrombin/protein, w/w) for 30 and 60 min at 37 "C ( Fig. 6). Digestion of PBP lB-GT/Hs by thrombin resulted in the appearance of a new band consistent with the removal of the cytoplasmic tail and membrane anchor of the protein and the loss of 6400 Da. Thrombin cleaved the PBP lB-GT/Hs to a significant extent even at a weight ratio of 1:500, and the digestion was essentially complete in 30 min at a 1:lOO ratio. Digestion of PBP lB-H6 by thrombin at the same ratios as above had no effect on the mobility of the protein (Fig. 6,  lanes 4-7), indicating that the digestion was dependent upon the presence of the engineered thrombin cleavage site. In other experiments, it was determined that at identical ratios of thrombin to PBP 1B-GT/H6 the extent of digestion was increased when the concentrations of the two proteins were increased.
One of the goals of this project was to ensure that the cleaved cytoplasmic tail/membrane anchor region of the protein was not still non-covalently attached to the remainder of the protein; if the periplasmic region of the protein remained insoluble in the absence of detergents, it would be important to show that it was not due to the continued non-covalent association of the hydrophobic peptide. Purified PBP 1B-GT/ He (80 pg/ml) was cleaved with thrombin at a weight ratio of 1:250 (thrombin/protein) for 1 h at 37 "C. At this concentration of PBP 1B-GT/H6 and thrombin/protein ratio, about 50% of PBP lB-GT/He is cleaved (Fig. 7, lane 4 ) . The digestion products were loaded onto a Ni2+-NTA column, and the flow-through and wash fractions were collected. The protein that remained bound to the column was eluted with 120 mM imidazole in buffer. SDS-PAGE analysis of the different pools indicated that thrombin-cleaved PBP 1B was present only in the flow-through/wash pool, whereas uncleaved PBP lB-GT/Hs was found only in the 120 mM imidazole pool (Fig.  7, lanes 5 and 6). These data convincingly demonstrate that the hexahistidine-containing cytoplasmic tail/membrane anchor portion of PBP lB-GT/Hs is not associated noncovalently with the protein after cleavage by thrombin, and also that the cleaved protein can be purified from any uncleaved protein still remaining in the digest.
Biochemical Analysis of Thrombin-cleaved PBP IB-Once it was clear that we could isolate the periplasmic portion of PBP 1B (PBP lB,J without contamination of the cytoplasmic tail/membrane anchor, we wanted to look at the biochemical and solubility properties of the protein. To address the question as to the integrity of the protein following thrombin cleavage, we determined the affinity of PBP lB,, for [14C]penicillin G as described above. As seen in Fig. 4, PBP 1B,, had an identical affinity for [14C]penicillin G as both PBP 1B-GT/H6 and PBP lB-H6. This suggests that the periplasmic domain of PBP 1B is stable (at least as defined by antibiotic affinity) and structurally similar to wild-type PBP 1B.
As mentioned above, structural models based on hydropathy analysis and 0-lactamase fusion experiments predict that PBP 1B,, should be freely soluble in aqueous solution. To  test this prediction, we performed gel filtration of [14C]penicillin G-prelabled PBP 1B,, in the presence and absence of 0.7% CHAPS. In the presence of 0.7% CHAPS, a radioactive peak corresponding to ['4C]penicilloyl-PBP lB,, eluted in the included volume of the column at a position expected of an 84-kDa protein (Fig. 8). However, when CHAPS was removed from the buffer and gel filtration was repeated, no eluting material could be identified. Since the binding of ["C] penicillin G to PBP 1B,, is done in CHAPS, this result cannot be due to the inability of the protein to bind antibiotic but must be related to its solubility in the absence of detergent. Indeed, SDS-PAGE analysis of the fractions from gel filtra- tion in the absence of CHAPS encompassing the elution position of PBP lR,,, failed to identify any protein by silver staining, which supports the idea that PBP lB,, is not soluble in the absence of detergents. These results were essentially identical to the results obtained with the PBP 5-1B fusion protein described earlier.
One possibility for the insolubility of PBP lB,, was that cell wall substrate, i.e. the lipid-linked disaccharide pentapeptide, was tightly bound in the active site of the transglycosy-  lase domain and mediated the insolubility of the protein in the absence of detergents. T o determine whether this possibility existed or whether the insolubility was inherent to the protein itself, PBP lB-GT/Hs was dialyzed extensively against guanidine hydrochloride (M, 12,000 cutoff), renatured by dialysis against 25 mM Tris. HC1, pH 8.0,0.5 M NaC1, 10% glycerol, and 1% CHAPS, and then concentrated and filtered. Approximately 26% of the input PBP lB-GT/Hs protein was recovered after renaturation, and the renatured protein was judged to bind near stoichiometric amounts of [3H]penicillin G by comparison of protein staining and fluorography after SDS-PAGE. When the renatured protein was cleaved by thrombin to produce PBP lBper, labeled with [3H]penicillin G, and run on gel filtration as described above, the protein eluted only in the presence and not in the absence of CHAPS (data not shown). These results strongly suggest that the insolubility is due to the structural properties of the protein itself and not from any tightly bound substrate.

Cleavage of PBP lB-GT/Hs in
Membranes-Analysis of the cleavage reaction of membranes from overexpressing cells indicated that at high thrombin to protein ratios, cleavage of PBP lB-GT/Hs to a form identical in mobility to PBP lB,,, occurred at reasonable levels even in the membrane preparation (Fig. 9, lane 8). This cleavage did not occur in membranes containing overexpressed PBP 1B, again indicating that the cleavage was dependent upon the presence of the thrombin cleavage site (Fig. 9, lane 4 ) . As another test of the solubility of PBP lBper, we fractionated the digest into soluble and membrane fractions. As expected from the gel filtration results, all of the cleaved material migrating at the position of PBP lB,,, (as determined by immunoblotting) fractionated with the membrane portion.
Interaction of PBP Proteins with Lipid Vesicles-As a further test of the solubility and membrane association properties of PBP lBper, we reconstituted PBP proteins into lipid vesicles and submitted the vesicles to flotation on sucrose gradients. Three proteins were chosen for this study: PBP lB-GT/Hs, PBP lB,,,, and soluble PBP 2 (sPBP 2) from Neisseria gonorrhoeae. N. gonorrhoeae sPBP 2 has been previously characterized in our laboratory and shown to be freely soluble in aqueous solutions in the absence of detergents, and therefore served as a control for the action of a soluble penicillin-binding protein (25). After reconstitution of the PBP proteins into asolectin vesicles and flotation of the vesicles on a sucrose gradient, the location of the vesicles and ['251]penicillin V binding activity of the fractions of the gradient were performed as described under "Experimental Procedures." As can be seen in Fig. 10, both PBP lB-GT/Hs and PBP lB,,, migrated with the lipid vesicles, whereas sPBP 2 was present entirely at the bottom of the gradient. We conclude from these results that PBP lB,,, associates directly with lipid membranes and that native PBP 1B must therefore interact with E. coli membranes at two or more positions.

DISCUSSION
We describe in this report the use of an engineered proteolytic cleavage site for generating the periplasmic domain of PBP 1B. When the periplasmic domain was purified and its properties investigated, it behaved in all instances like a membrane-associated protein. The properties of PBP lB,,, were identical to the properties of a PBP 5-1B fusion protein, in which the 29-amino-acid signal sequence of PBP 5 was fused directly with amino acid 93 of PBP 1B. It had been suggested by Edelman et al. (21) that one explanation for the unusual behavior of the fusion protein might be that it interacted with other membrane proteins. Because reconstitution of purified PBP lB,,, resulted in >95% of the protein associated with the liposomes, our results clearly indicate that no other proteins are required to mediate membrane association. This result was unexpected in that there are no other transmembrane regions or even hydrophobic stretches of amino acids in residues 88-844.
The rationale behind the addition of the thrombin cleavage site was to place it directly after the deduced end of the transmembrane region. In this way, we hypothesized that this region would not interfere with the normal folding of the protein since it would hypothetically represent an extension of the transmembrane anchor. The addition of 2 glycine residues before and 4 glycine residues after the six-aminoacid thrombin cleavage site (also referred to as a "glycine kinker") were included to assure that the cleavage site would be accessible to the added protease. Although it might be argued that the introduction of these 12 extra residues could cause the protein to fold aberrantly, and thus generate a pseudo membrane protein, we believe this is unlikely for many reasons. 1) The results obtained with purified PBP IB,,, were identical to the those obtained with a PBP 5-1B fusion protein. Active P B P 5-1B fusion protein was found almost entirely in the membrane fraction of a French press lysate, and the purified protein was insoluble in the absence, but not in the presence, of 0.7% CHAPS. These results were also obtained for purified P B P lBper. This strongly suggests that the addition of extra amino acids in PBP 1B-GT/H6 was not responsible for the insolubility of P B P lB,, in the absence of detergents but that this property was inherent in the protein itself. 2) The thrombin cleavage site was accessible to thrombin at low thrombin/protein ratios. We interpret this result as an indication that the engineered residues are protruding from the protein into solution and not interacting with another part of the protein. If these residues interrupted the proper folding of PBP lB-GT/Hs and thus interacted with other regions of the protein, one might not expect that the protease would cleave the protein at such low ratios. 3) The intrinsic fluorescence spectra of both PBP lB-H6 and PBP lB-GT/Hs were identical in both the native and denatured states. The superimposable spectra of the two proteins at equimolar concentrations indicates that the environment of the tryptophans in the molecule are not significantly altered. 4) The affinities of PBP lB-GT/Hs, PBP lB-Hs, and P B P lBp,, for [14C]penicillin G were identical. This indicates that the extra residues do not influence the binding of antibiotic and that the cleaved protein has an identical affinity as the uncleaved and wild-type proteins. 5) The ammonium sulfate concentrations needed to precipitate PBP lB-Hs, and PBP 1B-GT/Hs were identical (33%), and the two proteins behaved indistinguishably throughout the purification. Thus, our data indicate that PBP 1B contains a heretofore undescribed membrane association that is independent of its transmembrane anchor.
The presence of another site of membrane attachment in PBP 1B raises an important question: With which membrane does the periplasmic domain associate? Although we have no evidence to support either the cytoplasmic or outer membrane, other reports in the literature may give us a clue. Bayer et al. (39) have used immunoelectron microscopy of overexpressed PBP 1B to localize the protein at the cell surface. They observed significant immunolabeling of PBP 1B at the inner membrane and a cytoplasmic zone extending from the inner membrane. Increased PBP 1B was also observed in inner membrane-outer membrane contact areas or zones of membrane adhesion. We believe these results may explain at least one aspect of our work. It was shown that a latency of ["C] penicillin G binding was observed in membranes from cells overproducing PBP 1B constructs. Since Bayer et al. found a large portion of overexpressed PBP 1B sequestered in a cytoplasmic zone adjacent to the inner membrane, it is possible that this population of the protein is inaccessible to labeled antibiotic, even in membrane preparations. Only when these sequestered zones are disrupted with detergent would the PBP 1B molecules be able to interact normally with ["C] penicillin G. This phenomenon has been observed only when PBP 1B constructs have been highly overexpressed, and may represent (as they suggest) either a temporary accumulation of newly synthesized protein or the saturation of a limited number of membrane adhesion sites.
An even more important point raised by the work of Bayer et al. was that PBP 1B was often localized at inner membraneouter membrane adhesion sites. This could be explained with our results if the periplasmic domain of PBP 1B interacts with the outer membrane, thus bridging both the inner membrane (through the transmembrane anchor) and the outer membrane (through the periplasmic domain membrane association). Because it is known that the peptidoglycan layer in E. coli is intimately associated with the outer membrane via lipoprotein (40, 41), this explanation is intriguing. Thus, one might imagine that the major transpeptidase (PBP 1B) is associated with both the inner and outer membrane and synthesizes peptidoglycan by receiving the lipid-linked disaccharide pentapeptide substrate from the cytoplasmic membrane and polymerizing it onto peptidoglycan that is associated with the outer membrane. If this property is a common theme in all the major transpeptidases of Gram-negative bacteria, then PBP lA, which is able to compensate for the loss of PBP lB, might also exhibit this unusual membrane association. This idea is being actively pursued.
At the present time we do not have any information as to the location and nature of the periplasmic membrane association, but these experiments are in progress. It is very unlikely that another membrane-spanning region is responsible for this behavior since @-lactamase fusion experiments have shown that all fusions past amino acid 88 resulted in significant p-lactamase activity in the periplasm. It is possible that the association could be due to a hydrophobic region on the surface that is formed in the three-dimensional structure of the protein, or it could be due to post-translational modification, such as lipid modification. The latter explanation would be consistent with the known requirement of membrane translocation of PBP 1B in order to fold into an active protein (35,42). The properties of P B P 1B make it clearly different from other PBPs. For example, most of the other PBPs studied in E. coli and other organisms have a predictable membrane topology and can be made readily soluble using genetic techniques (22-25, 43, 44). Once the location and mechanism of the membrane association is identified, we can possibly modify the amino acids that are responsible for this property to make a true soluble derivative for growing crystals suitable for x-ray crystallography. Finally, if the periplasmic membrane association is due to post-translational modification, unraveling the mechanism of this modification may provide another avenue for antibiotic design and investigation.