Cloning and Characterization of PBP 1C, a Third Member of the Multimodular Class A Penicillin-binding Proteins of Escherichia coli *

All proteins of Escherichia coli that covalently bind penicillin have been cloned except for the penicillin-binding protein (PBP) 1C. For a detailed understanding of the mode of action of β-lactam antibiotics, cloning of the gene encoding PBP1C was of major importance. Therefore, the structural gene was identified in the E. coli genomic λ library of Kohara and subcloned, and PBP1C was characterized biochemically. PBP1C is a close homologue to the bifunctional transpeptidases/transglycosylases PBP1A and PBP1B and likewise shows murein polymerizing activity, which can be blocked by the transglycosylase inhibitor moenomycin. Covalently linked to activated Sepharose, PBP1C specifically retained PBP1B and the transpeptidases PBP2 and -3 in addition to the murein hydrolase MltA. The specific interaction with these proteins suggests that PBP1C is assembled into a multienzyme complex consisting of both murein polymerases and hydrolases. Overexpression of PBP1C does not support growth of a PBP1Ats/PBP1B double mutant at the restrictive temperature, and PBP1C does not bind to the same variety of penicillin derivatives as PBPs 1A and 1B. Deletion of PBP1C resulted in an altered mode of murein synthesis. It is suggested that PBP1C functions in vivo as a transglycosylase only.

The high molecular weight penicillin-binding proteins (PBPs) 1 of Escherichia coli are known to be essential for the growth and division of the murein sacculus, the scaffolding structure of the bacterial cell envelope (1)(2)(3)(4)(5)(6)(7). PBP1A and PBP1B are bifunctional transglycosylases/transpeptidases and probably represent the major murein polymerizing enzymes (8 -12). At least one of these two proteins has to be active because a deletion in both genes, ponA and ponB, is known to be lethal (11,13,14). PBP2 and -3 are likely to be monofunctional DD-transpeptidases (7,15) with specificities for cell elongation and cell division, respectively (1). Consequently, specific inhibition of PBP2 results in the formation of spherical cells (16), and a block in PBP3 causes filamentation (17). All of the high molecular weight PBPs have been cloned and characterized in detail (4) with the only exception being PBP1C. This protein, which migrates in PBP assays between PBP1B and PBP2 at an apparent molecular mass of about 70 kDa, has not been studied so far (18), because, among other reasons, PBP1C is not visible in standard PBP assays unless certain ␤-lactam derivatives are applied to label the proteins. In addition to 125 I-labeled latamoxef (19), the only radioactive ␤-lactam currently known to label PBP1C reproducibly is the 125 I-labeled Bolton and Hunter derivative of ampicillin (18).
The molecular weight of PBP1C as well as its specific binding, together with PBPs 1A and 1B, to the immobilized transglycosylase inhibitor moenomycin (20) suggest that this PBP is a bifunctional transglycosylase/transpeptidase similar to PBP1A and -1B, which are the main lethal targets for penicillins (1,2,21). However, because a PBP1A ts /1B double mutant is not viable at the restrictive temperature, wild type levels of PBP1C are unable to substitute for the loss of PBPs 1A and 1B (13). This may indicate that PBP1C has a specific function different from 1A and 1B.
PBP1C attracted our attention when it turned out to be specifically retained together with PBPs 1B, 2, 3, and the DD-endopeptidases PBP4 and PBP7 by lytic transglycosylases coupled to activated Sepharose (20,22). This finding could indicate that PBP1C might play an important role in the formation of a multienzyme complex that has been speculated to combine murein synthases and hydrolases (20,23,24). Here we report on the cloning, purification, and biochemical characterization of PBP1C. In addition, it is shown that a deletion of the gene, although viable, greatly affected the type of murein cross-linkage.

Bacterial Strains, Plasmids, and Growth Conditions
The strains and plasmids used in this study are shown in Table I. Bacteria were grown aerobically in one of the following media at 37°C: Luria-Bertani (LB) medium (30); LB-maltose medium (LB supplemented with 0.2% maltose, 10 mM MgSO 4 , 10 mM CaCl 2 ); or LB plus 10 mM MgSO 4 . Growth was monitored by optical density readings at 578 nm (A 578 ). Antibiotics were added, when required, at the following concentrations: 50 g/ml kanamycin, 20 g/ml chloramphenicol, 40 g/ml spectinomycin, 30 g/ml streptomycin.

Preparation of Lysates of the Kohara E. coli Genomic Phage Library
The miniset Kohara E. coli genomic phage library (31), kindly provided by Y. Kohara, was delivered in microtiter plates with each individual clone poured in SM medium (50 mM Tris-HCl, 10 mM NaCl, 10 mM MgCl 2 , 0.01% gelatin, pH 7.5) containing 0.4% agarose and 30% (w/v) glycerol. The phages were eluted by overlaying each well with 50 l of SM medium before incubating the plates overnight at 4°C. Plaques from eluted phages were prepared as described (32) and were finally resuspended in 1 ml of SM medium containing one drop of chloroform. Fresh, high titer lysates were obtained according to the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: PBP, penicillin-binding protein; LB, Luria-Bertani; PAGE, polyacrylamide gel electrophoresis; IPTG, isopropyl-1-thio-␤-D-galactopyranoside; PCR, polymerase chain reaction; kb, kilobase. procedure described by Henderson et al. (33) with the following modifications. Resuspended plaques of selected members of the phage library (0.4 ml) were added to 75 l of a fresh overnight culture of E. coli W3110 grown in LB-maltose medium. Phage adsorption was allowed for 20 min at 37°C without shaking. LB-maltose medium (3 ml) was added to each tube, and samples were incubated overnight at 37°C in a shaking waterbath. Phage lysates were prepared as described (33).

Expression of Proteins Encoded by the Kohara E. coli Genomic Library
To reduce the background expression of PBPs upon infection of E. coli with selected members of freshly prepared phage lysates, the PBP4, -5, and -6 triple mutant D456 (25) was grown overnight in LB-maltose. LB supplemented with 10 mM MgSO 4 was inoculated at 1:250, and cells were grown to an A 578 of about 0.25. Then, 3 ml of culture were added to 0.5 ml of phage lysate. Samples were incubated with shaking for 45 min at 37°C. After centrifugation at 10,000 ϫ g for 2 min, cells were resuspended in 60 l of ice-cold 100 mM Tris-HCl, 10 mM MgSO 4 , 1% (w/v) Triton X-100, 0.02% NaN 3 , pH 8.0, prior to storage at Ϫ20°C. Aliquots (20 l) of each sample were thawed, and PBPs were analyzed as described below.

General DNA Techniques
Standard recombinant procedures were performed essentially as described by Sambrook et al. (32). Plasmid DNA was isolated according to Birnboim and Doly (34), chromosomal DNA was prepared as described by Murray and Thompson (35), and transformation with newly constructed plasmids was done following the procedure of Inoue et al. (36). Ligations were performed with the rapid DNA ligation kit (Roche Molecular Biochemicals). Labeling of the DNA probe for Southern blot hybridization was done with the DIG-High Prime kit (Roche Molecular Biochemicals).

PCR Amplification
Amplification of the 8-kb DNA fragment (37) covering the entire region between ndk and sseA was accomplished with the Expand Long Template PCR System (Roche Molecular Biochemicals). The reaction was performed in a volume of 50 l of 50 mM Tris-HCl (pH 9.2) containing 16 mM (NH 4 ) 2 SO 4 , 1.75 mM MgCl 2 , 350 M deoxyribonucleoside triphosphates, and 2.6 units of Taq and Pwo polymerase enzyme mix with 0.3 M concentrations of each of the primers NdkH3 (5Ј-TTTA-AGCTTTTTGCTACCGCGTTCGGTTTGA-3Ј) and SseAH3 (5Ј-TTTA-AGCTTATGTGCGGAAGCGGGGAAGTGT-3Ј) and 1 l of Kohara phage 429 lysate (approximately 1 ϫ 10 7 phage particles) as a template. Before cycling, the sample was incubated for 2 min at 93°C, followed by 10 cycles at 93°C for 10 s, 65°C for 30 s, and 68°C for 5 min. Another 22 cycles with increasing elongation times (additional 20 s/cycle) under the same conditions were performed. The final polymerization step was an incubation at 68°C for 7 min. The HindIII sites introduced by the primer pair NdkH3/SseAH3 at both ends of the PCR product allowed for cloning of the fragment into pBC SKϩ.

Construction of a Deletion in pbpC
To construct a deletion in pbpC, the method of Kulakauskas et al. (38) was followed. The 1-kb long upstream and downstream sequences flanking the gene were amplified by PCR using the following pairs of primers: for the upstream region 1CNa, 5Ј-TTTGGATCCGCCCGTCA-GCAAAAGGCT-3Ј and 1CNi, 5Ј-ATCAGGGACAAAACAGACTAGTCG-ACCAATCAGCAGATCTTCAG-3Ј; and for the downstream region 1CCa, 5Ј-TTTGGATCCCAGATGAACGCTGTTCTT-3Ј and 1CCi, 5Ј-G-GCTGAAGATCTGCTGATTGGTCGACTAGTCTGTTTTGTCCCTG-3Ј. The inner primers 1CNi and 1CCi were designed such that they were complementary to each other over the whole length and that a SalI site was introduced in their central regions (underlined characters). PCRs were performed in 100-l volumes containing 20 mM Tris-HCl (pH 8.55), 16 mM (NH 4 ) 2 SO 4 , 1.5 mM MgCl 2 , 200 M each of the four deoxyribonucleoside triphosphates, 4 units of Taq polymerase (AGS, Heidelberg, Germany), each primer at 300 nM, and 0.5 l of Kohara phage 429 lysate (approximately 5 ϫ 10 6 phage particles) as a template in a Peltier Thermal Cycler PTC-200 (MJ Research, Watertown, MA). Amplification was accomplished by incubation of the reaction mixture for 2 min at 94°C and cycling of the temperature 30 times (15 s at 94°C, 45 s at 43°C, and 90 s at 72°C) followed by a final 5-min incubation at 72°C. The PCR products were purified with the QIAquick PCR purification kit (Qiagen, Hilden, Germany) and eluted in 48 l of 10 mM Tris-HCl, 1 mM EDTA, pH 8.0. The purified PCR products were joined in a second PCR to create a single 2-kb fragment. Reaction conditions were as described above with the following exceptions: PCR was performed with 0.3 M of the outer primers 1CNa and 1CCa and 1 l each of the purified 1.0-kb PCR products as a template. After an initial denaturation for 2 min at 94°C, amplification was performed in 30 cycles at 94°C for 15 s, 52°C for 45 s, and 72°C for 3 min. The final polymerization step was an incubation at 72°C for 5 min. The purified DNA was restricted with BamHI and ligated with the linearized pBC SKϩ vector to yield plasmid pGS20. A kanamycin resistance cassette from plasmid pUC4K (29) was cloned into the introduced SalI restriction site of pGS20 yielding plasmid pGS33. The deletion of pbpC was transduced with the help of the Kohara phage 429 into the chromosome of E. coli MC1061. The deletion mutant in pbpC was named TG57.

Purification of PBP1C
E. coli TG4 harboring pGS21 was grown in 21 liters of LB medium at 28°C. To induce expression of PBP1C, 10 M IPTG was added at an A 578 of 0.4 and growth was continued until an A 578 of 0.5 was reached. The cells were cooled in an ice bath and harvested by centrifugation (24 g wet weight), resuspended in 189 ml (final volume) of 10 mM Tris- This study maleate buffer, pH 6.8, containing 10 mM MgCl 2 , 0.02% NaN 3 , 1 mM phenylmethylsulfonyl fluoride, and DNase I (10 g/ml; Roche Molecular Biochemicals). All of the following steps were performed at 4°C.
Step 1-Cells were broken by passage through a French pressure cell at 3,000 p.s.i. Membranes were separated by centrifugation (1 h at 70,000 ϫ g) and extracted twice with 1 M NaCl in 10 mM Tris-maleate, 10 mM MgCl 2 , 0.02% NaN 3 , pH 6.8. To solubilize membrane proteins, the membranes were treated overnight with 2% (w/v) Triton X-100 in 186 ml of the above mentioned buffer. The solubilized proteins were separated by centrifugation (see above) and dialyzed three times against 5 liters of 20 mM Tris-HCl, 5 mM MgCl 2 , 50 mM NaCl, 0.02% NaN 3 , pH 8.0. The Triton X-100 concentration in the dialyzed fraction (217 ml) was adjusted to 1% (w/v) by adding another 155 ml of dialysis buffer.
Step 3-The dialyzed material was applied at a flow rate of 61.6 ml/h onto an S-Sepharose Fast Flow column (33 ml, Amersham Pharmacia Biotech) that was equilibrated with dialysis buffer supplemented with 1% (w/v) Triton X-100. The column was washed with 1000 ml of equilibration buffer, and retained proteins were eluted at a flow rate of 14.6 ml/h with a linear gradient (330 ml) from 50 to 400 mM NaCl in 10 mM Tris-maleate, pH 6.0, containing 10 mM MgCl 2 , 1% (w/v) Triton X-100, and 0.02% NaN 3 . Fractions of 8.3 ml were collected. PBP1C eluted in fractions 15-25 between 140 and 240 mM NaCl.
Step 4 -The pool of 86 ml was directly loaded with a flow rate of 4.6 ml/h onto an equilibrated Green19-agarose column (5 ml, Sigma). The column was washed with 100 ml of equilibration buffer, and retained proteins were eluted in 7-ml fractions with two salt steps: 26 ml of 400 mM NaCl followed by 112 ml of 2 M NaCl in equilibration buffer. The SDS-PAGE of aliquots of the eluted fraction showed that PBP1C eluted with 2 M NaCl. A summary of the purification procedure is shown in Fig.  4. The comparison of the Coomassie Blue-stained gel with the autoradiography of a PBP assay clearly showed the identity of the purified 70-kDa protein with PBP1C (data not shown). The covalent binding of the ampicillin derivative also proves that the isolated protein was enzymatically active.

PBP1C-Sepharose Affinity Chromatography
Purified PBP1C (10 mg) was coupled to 3 ml of CNBr-activated Sepharose 4B (Amersham Pharmacia Biotech) following the instructions of the manufacturer and the protocol previously described (40). All buffers contained 1% (w/v) Triton X-100. As a control, activated Sepharose was treated identically except that no protein was added resulting in a material (Tris-Sepharose) with all functional groups blocked by Tris. A Triton X-100 membrane extract was prepared from a 2-liter culture of E. coli TG57(⌬pbpC) growing exponentially (A 578 , 0.5). Briefly, cells were broken in a French press, and membranes were separated by centrifugation (70,000 ϫ g, 1 h, 4°C). Membrane proteins were extracted overnight at 6°C with 10 ml of 10 mM Tris-maleate, 10 mM MgCl 2 , 150 mM NaCl, 0.02% NaN 3 , pH 6.8, containing 2% (w/v) Triton X-100. The solubilized proteins were separated from the membranes by centrifugation (see above), and prior to dialysis, the Triton X-100 concentration was reduced to 1% (w/v) by dilution with dialysis buffer (10 mM Tris-maleate, 10 mM MgCl 2 , 50 mM NaCl, 0.02% NaN 3 , pH 6.8).
Chromatography was performed in 3-ml columns as described previously (20). The samples (20 ml each) were applied at a flow rate of 1.1 ml/h onto the PBP1C-Sepharose and Tris-Sepharose columns, respectively, and equilibrated with 10 mM Tris-maleate, 10 mM MgCl 2 , 0.02% NaN 3 , pH 6.8, containing 50 mM NaCl and 1% (w/v) Triton X-100. After washing the columns at 4.5 ml/h with 65 ml of equilibration buffer containing only 0.05% (w/v) Triton X-100, elution of the retained proteins was achieved in two steps: first with 150 mM NaCl in equilibration buffer containing 0.05% (w/v) Triton X-100 (54 ml) and second with the same buffer but containing 1 M NaCl (54 ml). Aliquots of the applied sample, the flow-through, the wash, and the 150 mM NaCl and 1 M NaCl fractions were analyzed for PBPs or subjected to Western blot analysis.

PBP Assay
PBPs were analyzed with an 125 I-labeled Bolton and Hunter derivative of ampicillin following the protocol established by Schwarz et al. (18). Aliquots of the samples were incubated with 2 l (about 1 pmol) of the labeled ampicillin derivative for 30 min at 37°C. The PBP-penicilloyl complexes were separated by SDS-PAGE and analyzed by autoradiography as described below.

Determination of the Structure of Murein
The muropeptide structure of isolated murein sacculi was determined according to the procedure described by Glauner (41). Analysis of the length distribution of the glycan strands of murein sacculi was done as has been described previously (42).

Transglycosylase Assays
To measure transglycosylase activity in different transformants, two different methods were used. Incorporation of 14 C-N-acetylglucosamine from UDP-N-acetyl-D-[U-14 C]glucosamine (271 mCi/mmol, Amersham Pharmacia Biotech) into SDS-insoluble material was followed either in ether-permeabilized cells (43) or with crude membrane preparations (44).

General Methods
The protein concentration was measured by the bicinchoninic acid method of Smith et al. (45). SDS-PAGE was performed according to the procedure described by Lugtenberg et al. (39) using either 10 (w/v) or 12% (w/v) acrylamide. Autoradiography of 125 I-labeled PBPs was done with Hyperfilm-MP and two Hyperscreen intensifying screens (both from Amersham Pharmacia Biotech) at Ϫ70°C. Films were exposed for times between several hours and several days depending on the PBP concentrations of the samples. Transfer of proteins onto polyvinylidene difluoride membranes (Millipore) was done according to the method of Towbin et al. (46). Staining with Ponceau S (Sigma) confirmed that aliquot amounts of protein had been transferred to the membrane. For detection of the lytic transglycosylases, specific polyclonal antisera from rabbit were used. Immunoreactive bands were visualized with an alkaline phosphatase-coupled secondary antibody (Promega) as described previously (47).

Nucleotide Sequence Accession Number
The sequence of the pbpC gene has been deposited in GenBank™ under accession number U88571.

Screening of the Kohara Phage Library for the PBP1C
Structural Gene-Because of the apparent molecular mass, but also because of a specific binding to moenomycin (20), PBP1C was likely to be a bifunctional PBP similar to PBP1A and PBP1B. Homology searches, however, did not reveal the gene for PBP1C in the sequences that were present in the data bases in August 1996. This suggested to us that the gene was likely to map in regions not sequenced or not submitted to the data bases at that time. By using the 125 I-labeled Bolton and Hunter derivative of ampicillin to label PBPs, a highly sensitive and rather simple assay for PBP1C was available (18) that allowed screening of DNA libraries. We took advantage of the mapped Kohara E. coli genomic phage library (31) and decided to screen first all those 185 phages carrying nonsequenced (at that time) regions of the E. coli genome. The DNA sequences for a number of known PBPs were found using the PBP assay. Phage 109 resulted in overexpression of PBP3, phage 116 of PBP1B, phage 168 of PBP5, phage 169 of PBP2, and phage 209 of PBP6. Overproduction of PBP1C, as shown in Fig. 1, was observed with phage 429, which covers the region from 56.7-57.1 min of the E. coli chromosome. Phages 428 (data not shown) and 430, carrying the neighboring sequences, did not result in overexpression of PBP1C. Thus, the complete structural gene for PBP1C had to be present on phage 429.
Subcloning of pbpC-The sequence information that was available at that time for the 16-kb E. coli DNA fragment present in phage 429 is shown in Fig. 2A. The 8-kb stretch between ndk and sseA was amplified by PCR and cloned into pBC SKϩ as described under "Experimental Procedures." In addition, the E. coli DNA insert was restricted with EcoRI into four different fragments of 6.0, 4.7, 2.8, and 2.5 kb and ligated into pBC SKϩ. Transformants of E. coli XL1-Blue carrying the different constructs were analyzed in a PBP assay. The 8-kb fragment (pGS80) as well as the 6-kb EcoRI fragment (pGS60) directed overproduction of PBP1C. None of the other constructs resulted in overexpression of PBP1C (data not shown). On the basis of sequence information of the ndk-sseA intergenic region, which was kindly provided by F. R. Blattner's laboratory (University of Wisconsin, Madison, WI), a 2.4-kb Asp 718 /DraI fragment from pGS60 containing only one single open reading frame was cloned into the SmaI-site behind the tac promoter of pJFK118EH (28) and yielded plasmid pGS24. Induction of expression in E. coli XL1-Blue pGS24 by IPTG proved that the gene for PBP1C is encoded by the cloned fragment (Fig. 2B). The gene was named pbpC and is located at 56.9 min on the chromosome.
Sequence Analysis of PBP1C-The predicted amino acid sequence of PBP1C results in a calculated molecular mass of 85075.90 Da, which is 15 kDa larger than the apparent mass deduced from the behavior of the protein in SDS-PAGE (see Fig. 4). The protein consists of 38% hydrophobic, 20% polar, and 25% charged amino acids and has a calculated pI of 9.52. However, the percentage of identical amino acids is only 16.8 when compared with PBP1A and 10.9 when compared with PBP1B. The alignment of the sequences covering the transglycosylase as well as the transpeptidase domain of PBP1C with those of PBPs 1A and 1B according to the PIMA algorithm (48) reveals that PBP1C is clearly homologous to both of the other two bifunctional PBPs (Fig. 3). Transglycosylase as well as the transpeptidase motifs are present. However, when compared with recently published alignments of the transglycosylase domain of numerous multimodular class A PBPs as well as monofunctional transglycosylases (7,49), it is evident that in the transglycosylase domain of PBP1C the three otherwise highly conserved amino acid residues Arg 136 , Glu 140 , and Arg 218 are replaced by Gly, Gln, and Ala, respectively. In the transpeptidase domain the characteristic triads SXXK, SXN, and KTG are found unchanged. The hydrophobicity profile predicts an N-terminal transmembrane region (Gly 9 -Ala 28 ), which is also present in the PBPs 1A and 1B. Analogous to PBPs 1A and 1B, there is also no leader peptidase I recognition sequence in PBP1C. The exported protein, therefore, is predicted to be anchored to the cytoplasmic membrane via the N-terminal transmembrane region with the enzymatic domains present in the periplasmic space.
Purification of PBP1C-The degree of inducible overexpression of PBP1C from plasmid pGS24 was rather moderate (Fig.  2B); only a 10 -20-fold overexpression, as judged from band intensities in the PBP assay, could be achieved upon induction with 1 mM IPTG. Increasing the inducer concentration to as much as 10 mM did not further stimulate expression of PBP1C. The overexpressed PBP1C could not be visualized, however, by Coomassie Blue staining after SDS-PAGE of the membrane fraction of cells induced by 1 mM IPTG (data not shown). This might be attributed to the quite long distance of 97 nucleotides between the vector-encoded ribosome-binding site and the ATG start codon of pbpC. Therefore, to purify PBP1C at a preparative scale, an inducible overexpression system was constructed in which this distance was shortened to a length of 10 nucleotides. A 2.3-kb Eco57I/DraI fragment of pGS60 containing pbpC was cloned into the EcoRI site immediately behind the tac promoter of the pJFK118EH vector (28) that carries the kanamycin resistance cassette yielding plasmid pGS21. Before cloning, the overhanging bases were digested with mung bean endonuclease (New England Biolabs) to create blunt ends. Induction of expression from pGS21 by 1 mM IPTG resulted in a massive production of PBP1C that caused cell lysis about 30 -45 min after the addition of the inducer. When the same experiment was repeated in medium containing 12% (w/v) sucrose, lysis could not be prevented, suggesting that the accumulation of PBP1C affects the integrity of the membranes rather than destabilizing the murein sacculus. The level of overexpression under these conditions was so great that PBP1C could be visualized as the most prominent band in the Coomassie Blue stain after SDS-PAGE of the Triton X-100 solubilized membrane fraction of induced cells. PBP1C was the only PBP to be visualized in a PBP assay, because the overproduced protein titrated the total amount of added labeled ampicillin (data not shown). When the inducer concentration was reduced to 15 M IPTG and the growth temperature to 30°C, a moderate overproduction of PBP1C was achieved without inducing cell lysis. Nevertheless, PBP1C was still clearly visible in the Coomassie Blue stain (see also Fig. 4) and was still the only PBP detectable in the PBP assay. To avoid contamination with PBP1B, which shows biochemical features very similar to PBP1C, purification was done from the PBP1B deletion mutant TG4 transformed with pGS21. Solubilization of PBP1C was accomplished in the presence of 2% (w/v) Triton X-100 and 1 M NaCl at 6°C for 12 h. The details of the isolation protocol are described under "Experimental Procedures," and a summary of the purification is given in Fig. 4.
An anion exchange column was used as a first purification step to separate the basic protein PBP1C from the more acidic proteins that bind to Q-Sepharose. PBP1C was recovered in the flow-through from a Q-Sepharose Fast Flow column to which a Triton X-100 membrane extract from E. coli TG4/pGS21 had been applied. This purification step was followed by chromatography on the cationic exchange material S-Sepharose. As expected, PBP1C did bind to the column and could be eluted with a salt gradient of 50 -400 mM NaCl in elution buffer. PBP1C eluted at about 200 mM NaCl. A final purification was achieved by Green19-agarose affinity chromatography. PBP1C eluted from the column in a 2 M NaCl step. Analysis of the eluted material by SDS-PAGE showed that at least by Coomassie Blue staining no other proteins were visible (Fig. 4). The final yield of purified PBP1C was 1 mg of protein/liter of cell culture. The purified PBP1C was enzymatically active as shown by the covalent binding of 125 I-labeled Bolton and Hunter ampicillin derivative (data not shown).
Biochemical Characterization-The sequence of PBP1C predicts transpeptidase and transglycosylase activity. Although the covalent binding of the Bolton and Hunter derivative of ampicillin demonstrates the presence of an enzymatically active penicillin binding domain, the inherent transpeptidase activity could not be investigated because of difficulties in establishing an in vitro assay for murein transpeptidation (Refs 3, 9, and 15; see also "Discussion"). By contrast, transglycosylase activity can be determined quite easily by following the incorporation of labeled UDP-GlcNAc in the presence of UDP-MurNAc-pentapeptide into high molecular weight murein (43,44). Two different assays were employed that measure transglycosylase activity with the native membranebound enzyme present either in ether-permeabilized cells (43) or in crude membrane vesicles, obtained by shaking cells with glass beads (44). As shown in Table II induction of PBP1C from plasmid pGS24 by the addition of 1 mM IPTG in the ponB deletion mutant TG4 showed a greater than 3-fold increase in 14 C-UDP-GlcNAc incorporation into SDS-insoluble material in ether-treated cells. In agreement with the recently shown specific binding of PBP1C to moenomycin-agarose (20), the PBP1C-specific murein synthesizing activity was sensitive toward moenomycin, as is PBP1A-and -1B-driven murein synthesis (50). Deletion of PBP1C resulted in a decrease in murein synthesis by 75% as compared with the wild type.
When using the membrane system to measure murein syn- thesizing activity, the deletion of PBP1C resulted in a drop in overall activity by only less than 3%, whereas a deletion in PBP1B reduced the activity by more than 95%. Induction of PBP1C from plasmid pGS21 with 10 M IPTG in a ponB deletion background caused an increase in activity by a factor of about 7 as compared with the empty vector control. Murein synthesis was stimulated approximately 4-fold even in the absence of inducer because of the leakiness of PBP1C expression from pGS21 under those conditions (PBP assay not shown). As in the ether cell system, PBP1C-driven murein synthesis in crude membrane vesicles was sensitive toward moenomycin. Taken together, the data from both of the murein synthesizing systems demonstrated a correlation between the rate of murein synthesis and the amount of induced PBP1C.
Interaction of PBP1C with Other Murein-metabolizing Enzymes-Affinity chromatography with different murein-metabolizing enzymes as specific ligands revealed specific proteinprotein interactions between murein hydrolases and synthases including PBP1C (20,22,40). Therefore, purified PBP1C was covalently bound to CNBr-activated Sepharose, and different cell fractions were analyzed for a specific binding of proteins to the immobilized PBP1C. As shown in Fig. 5A, PBPs 1B, 3, and 4 were retained by the PBP1C-Sepharose column and could be eluted with increasing concentrations of NaCl (150 -1000 mM). PBP1C, which was present in all fractions of the eluent, was obviously bleeding from the column. As indicated by the control column (Tris-Sepharose), PBP4 interacts nonspecifically with the matrix, although the binding to the PBP1C column was significantly enhanced. An interesting observation was that PBP2 binds to PBP1C-Sepharose and is specifically eluted in the 1 M NaCl elution step only when a mixture of Triton X-100-solubilized membrane proteins and a periplasmic cell fraction is applied onto the column (data not shown). In the absence of the periplasmic fraction, PBP2 did not bind to PBP1C-Sepharose (Fig. 5A). Western blot analysis of the eluent of the PBP1C-Sepharose affinity chromatography (Fig. 5B) revealed that PBP1C specifically interacted with the membranebound lytic transglycosylase MltA but not with MltB or the soluble enzyme Slt70 (24).
Physiological Consequences of the Absence of PBP1C-A deletion in pbpC was constructed by PCR deletion mutagenesis. The upstream and downstream sequences (1 kb each) flanking the gene were amplified by PCR using the primer pairs 1CNa/ 1CNi and 1CCa/1CCi. The design of the inner primers was chosen such that they were complementary to each other over the whole length and that a SalI site was introduced in the central region. As a result, the products had complementary ends and could therefore be joined together in a second PCR using the outer primer pair to yield a single 2-kb fragment in which the pbpC gene was replaced by a SalI site. With the help of the BamHI sites that had been introduced at the ends, the fragment was inserted into pBC SKϩ. The resulting plasmid was named pGS20. A kanamycin resistance cassette from pUC4K was inserted into the SalI site of pGS20 to yield pGS33. Because the flanking regions of the inserted kan R gene had the same orientation as the ones of the pbpC gene in the chromosome, the mutation in pGS33 could be transduced with the help of the Kohara phage 429 into the E. coli chromosome of MC1061. The EcoRI-digested genomic DNA of three of the resulting kanamycin-resistant transductants was analyzed by Southern blotting. As a specific digoxigenin-labeled probe, the upstream 1-kb PCR product that was synthesized for the generation of the deletion (see above) was used. Fig. 6B shows that, as expected, the transductants hybridized with an EcoRI fragment that was 1 kb shorter than the fragment present in the wild type control sample. The absence of the gene product PBP1C was demonstrated in a PBP assay of one of the transductants, TG57, as shown in Fig. 6A.
Deletion of the pbpC gene in the strain TG57 did not affect the growth rate or growth yield at different temperatures and salt concentrations. Furthermore, the minimal inhibitory concentration values for different antibiotics, including mecillinam, aztreonam, penicillin G, ampicillin, cefsulodin, imipenem, and moenomycin, were not changed, and only a minor increase, by a factor of two in the case of ampicillin, was observed with TG57. Thus a deletion in pbpC shows no obvious phenotype.
Murein Structure of Multiple Deletion Mutants in ponB, pbpC, and mgt-To study the effect of a deletion in pbpC on the structure of the murein sacculus when, in addition, other murein polymerases were deleted, we constructed a series of isogenic mutants. Isogenic mutants were indicated because it is known that the genetic background significantly affects murein structure (51). E. coli MC1061 was chosen as a background to create multiple deletions in PBP1C, PBP1B, and the monofunctional murein transglycosylase Mgt (52) by P1 transduction (30). The ponB deletion was taken from SP1026 (14), the pbpC b IPTG was added at a concentration of 1 mM for pGS24 and at a concentration of 10 M for pGS21 and pJFK118EH. c Samples were preincubated with moenomycin at a final concentration of 1 g/ml for 5 min on ice. d Ether permeabilization was done as described previously (43). The incorporation of UDP-[ 14 C]GlcNAc into SDS-insoluble material during an incubation at 30°C for 78 min by wild type cells was 81,921 cpm/mg of protein and was set at 100%.
e Crude membranes were prepared as reported by Schaller et al. (44). The incorporation of UDP-[ 14 C]GlcNAc into material that remained at the origin of the paper chromatography after an incubation for 90 min at 30°C with membranes from wild type cells was 80,800 cpm/mg of protein and was set at 100%. deletion from TG57, and the mgt deletion from JE7975. 2 Because of the inherent streptomycin resistance of MC1061, the result of a mutation in rpsL, transductants were selected on agar plates containing 30 g/ml streptomycin, 60 g/ml spectinomycin, and in the case of TG57 an additional 50 g/ml kanamycin. PBP1C and -1B deletions were verified by a PBP assay and by Western blotting using polyclonal PBP1B antiserum; the deletion in mgt was checked by analytical PCR (data not shown). The isogenic mutants obtained are listed in Table I. The mutants had no obvious phenotype and grew with a generation time of 22 min in LB at 37°C.
Analysis of the muropeptide composition (41) of the different mutants showed no major differences in the overall muropeptide pattern, with one interesting exception, the dimeric muropeptide structure TetraPenta, in which two GlcNAc-␤-1.4-MurNAc disaccharide units are linked by the peptide bridge (shown in Structure 1). This muropeptide was increased between 5-and 10-fold or higher in the different pbpC, ponB, and mgt mutants when compared with the wild type level (Table  III). The highest increase in TetraPenta was observed with a double deletion in pbpC and mgt.

MurNAc-GlcNAc
MurNAc-GlcNAc P P An analysis of the length distribution of the glycan strands in the range from 1 to 27 disaccharide units in the murein (42) revealed no alterations in the mutant strains TG57(pbpC), TG72(mgt), and TG5772(pbpC mgt) in comparison to the isogenic wild type strain MC1061 (average length of the glycan chains was about 7.5 disaccharide units). In the PBP1B mutants TG4(ponB) and TG5704(pbpC ponB), a minor shift in the average chain length by 1 and 0.5 disaccharide unit, respectively, to longer glycan chains was observed (data not shown). In agreement with the result of the glycan chain analysis, the average chain length as determined by the relative amounts of anhydromuropeptides (see Table III) does not differ significantly for all mutant strains (data not shown).
Overexpression of PBP1C in a ponA ts /ponB Double Mutant-A loss of the functions of both PBP1A and PBP1B is known to be lethal despite the presence of wild type levels of PBP1C (13,14). It was of interest to ask whether overproduction of PBP1C may at least partially suppress the 1A/1B deletion phenotype. This experiment was suggested to us by R. D'Ari (CNRS, Université Paris, Paris). E. coli JE5615 (ponA ts ponB) transformed with pGS24 and as a control with pJFK118EH were grown at the permissive temperature (28°C) for several generations before 1 mM IPTG was added to induce expression of PBP1C for 55 min. The cultures were divided in two, and one-half was shifted to 42°C while the other was kept at 28°C. Inhibition of growth, which is followed by bacteriolysis, was not affected even by a 10 -20-fold overproduction of PBP1C when both PBP1A and 1B were absent and/or nonfunctional. The control cultures continued to grow with unchanged generation time, irrespective of whether they were induced for PBP1C overproduction. Growth of the two cultures at 42°C came to a halt, and cells began to lyse as judged from analyzing the cultures in the light microscope. Hence, overexpression of PBP1C did not suppress the autolysis phenotype of a ponA ts ponB mutant.

DISCUSSION
PBP1C has a modular structure combining a penicillin-binding domain with a transglycosylase domain. Consequently, both penicillin binding activity and transglycosylase activity could be shown. Because of the inavailability of isolated lipid II, transglycosylase activity of native membrane-bound PBP1C was determined indirectly in ether-permeabilized cells and in crude membrane vesicles. In both systems the rate of murein synthesis correlated with the amount of induced PBP1C. Moreover, the inducible murein synthesizing activity was sensitive toward the transglycosylase inhibitor moenomycin. The data presented cannot definitely rule out the possibility that PBP1C might not be a transglycosylase by itself but merely an activator of some secondary synthetic activity of PBP1A and/or -1B. This alternative, however, seems unlikely, because the sequence alignment with PBPs 1A and 1B (Fig. 3) clearly indicates the presence in PBP1C of a transglycosylase domain, only slightly altered by three amino acid substitutions. Importantly, PBP1C, like PBPs 1A and 1B, has also recently been shown to specifically interact with immobilized moenomycin, a phosphoglycolipid antibiotic that is supposed to act as substrate analogue of lipid II (50), indicating that PBPs 1A, 1B, and 1C share common structural features in their respective transglycosylase domains.
Interestingly, the relative murein synthesis activity in TG57(⌬pbpC) was highly dependent on the assay system. In the ether cell system, deletion of PBP1C resulted in a decrease in murein synthesis by 75% as compared with the wild type. This severe reduction in murein synthesis was surprising when compared with the minor decrease of about 3% found in the membrane system. Because murein synthesis in ether-treated cells is mainly catalyzed by PBP1B (43) and because we show here that even a 10 -20-fold overproduction of PBP1C in TG4(⌬ponB) can restore only 11-12% of the wild type activity, it seems unlikely that PBP1C is responsible for the bulk of murein synthesis in this system. One explanation could be that the murein synthesizing enzyme complex is maintained in the ether-treated cell system but not in the membrane system. Deletion of PBP1C may therefore result in a drastic reduction in overall murein synthesis in ether-treated cells by affecting the activity of the remaining enzymes in the complex, mainly PBP1B. The apparently low contribution of the wild type level of PBP1C to the overall murein synthesizing activity in the crude membrane vesicles may be due to a disruption of possible protein complexes, which may affect the activity of PBP1C in vivo.
The modular structure of PBP1C closely resembles that of PBPs 1A and 1B (6,7). Nevertheless, PBP1C cannot substitute for 1A or 1B, whereas 1A can substitute for 1B and vice versa. Even a 10 -20-fold overproduction of PBP1C did not rescue a PBP1A ts /1B double mutant at the restrictive temperature. We have to conclude that PBP1C has its own distinct function that differs from PBPs 1A and 1B. Another feature adds to the exceptional role of PBP1C among all PBPs; PBP1C does not bind to most of the ␤-lactams known to bind to the other binding proteins. Only latamoxef and the Bolton and Hunter derivative of ampicillin have been described so far as binding to PBP1C (18,19). This indicates that the penicillin-binding domain, which is also the catalytic transpeptidase site, must be different from those present in the homologues PBP1A and 1B, despite the fact that the characteristic signature triad is present. The different penicillin sensitivities and the finding that PBP1C cannot be a substitute for PBP1A or -1B gives rise to the speculation that the penicillin-binding domain in PBP1C may no longer function to catalyze transpeptidation in vivo. Moreover, preliminary experiments by M. Terrak in the laboratory of M. Nguyen-Disteche 3 failed to demonstrate transpeptidase activity in vitro under conditions where PBP1B showed activity. Therefore, PBP1C may function in vivo as monofunctional transglycosylase with its penicillin-binding domain conserved to allow for interactions with other proteins. Indeed, as shown here and in previous publications (20,22), PBP1C specifically interacts with PBP1B (PBP2 in the presence of periplasmic proteins), PBP3, and MltA. The formation of a multienzyme complex has been proposed to explain how the different enzymatic activities that are involved in the processes of growth and division of the murein sacculus can be coordi-3 M. Terrak and M. Nguyen-Disteche, personal communication. nated efficiently (23,24). The assembly of the multienzyme complex seems to depend on the presence of specific structural proteins such as the recently identified MipA protein, which binds both PBP1B and MltA (20). The penicillin-binding domain of PBP1C may have a similar structural function. This is reminiscent of the proposal that the non-penicillin-binding domain of the bimodular PBP3 may play a role in the association of the enzyme into multienzyme complexes (6,53). Interestingly, a deletion in pbpC and a double deletion in pbpC and mgt, which codes for a monofunctional transglycosylase (52), are viable. This finding indicates that PBP1C can be replaced functionally by other murein polymerases, including probably PBP1A, PBP1B, and Mgt. However, the analysis of the murein structure indicates that in all of the analyzed mutants murein synthesis takes place by a different mechanism. Surprisingly, also the single deletion in PBP1B showed a similar change. This may be the result, however, of a destabilization of PBP1C when PBP1B is missing in the multiprotein complex. An analogous effect of a deletion of PBP1C on the activity of PBP1B is likely to be the reason for the observed drop in in vitro murein synthesizing activity of ether-treated cells of a PBP1C deletion strain, as has been discussed above. The presence of pentapeptide moieties in the cross-bridges is evidence that cross-linkage took place between newly synthesized material, because only nascent murein carries pentapeptide side chains (24). Therefore, the observed increase in tetrapentapeptide cross-bridges indicates that in these mutants newly synthesized murein is cross-linked predominantly with itself, indicating a multistrand insertion of nascent murein into the sacculus (5,54). Such a mechanism is in contrast to the finding that in wild type E. coli during cell elongation newly synthesized strands are cross-linked with pre-existing, old murein strands because of a single-strand insertion mode into the murein sacculus (5). The altered mode of murein synthesis in mutants lacking either PBP1C or the monofunctional transglycosylase Mgt may be taken as additional indication that the in vivo function of PBP1C might be that of a monofunctional transglycosylase. However, it will be important to establish an in vivo system that allows us to unravel, in an unambiguous manner, the enzymatic function of PBP1C in a growing cell.