The localization of penicillin-binding protein 2 (PBP2) in Escherichia coli has been studied using a functional green fluorescent protein (GFP)–PBP2 fusion protein. PBP2 localized in the bacterial envelope in a spot-like pattern and also at mid-cell during cell division. PBP2 disappeared from mid-cell just before separation of the two daughter cells. It localized with a preference for the cylindrical part of the bacterium in comparison with the old cell poles, which are known to be inert with respect to peptidoglycan synthesis. In contrast to subunits of the divisome, PBP2 failed to localize at mid-cell when PBP3 was inhibited by the specific antibiotic aztreonam. Therefore, despite its dependency on active PBP3 for localization at mid-cell, it seems not to be an integral part of the divisome. Cells grown for approximately half a mass doubling time in the presence of the PBP2 inhibitor mecillinam synthesized nascent cell poles with an increased diameter, indicating that PBP2 is required for the maintenance of the correct diameter of the new cell pole.
In the classic study by Spratt (1975), a relationship was proposed between the function of a particular high-molecular-weight penicillin-binding protein (PBP) and the shaping of Escherichia coli. The phenotypes of thermosensitive mutants of pbpA and ftsI, encoding for PBP2 and PBP3 (FtsI), respectively, appeared to be most clear-cut. In the first case, cells became spherical and, in the latter case, filaments were produced at the non-permissive temperature. Thus, PBP2 has been suggested to be involved in cell elongation, whereas PBP3 would play a specific role during cytokinesis (Spratt, 1975).
In the case of PBP3, this was further corroborated by the finding that filaments were produced when PBP3-specific antibiotics, such as aztreonam, piperacillin, furazlocillin, cephalothin or cephalexin, were applied to wild-type E. coli (Burdett and Murray, 1974; Iida et al., 1978; Botta and Park, 1980; Olijhoek et al., 1982; Pogliano et al., 1997). Remarkably, these filaments were not smooth (like the phenotype of temperature-sensitive ftsZ mutant strains), but they showed blunt aborted constrictions (Burdett and Murray, 1974; Iida et al., 1978; Olijhoek et al., 1982). The blunt constrictions appeared to be inert with respect to [3H]-diaminopimelic acid incorporation in contrast to nascent ones (Wientjes and Nanninga, 1989). This has led to the suggestion that PBP3-independent peptidoglycan synthesis (PIPS) would be needed for the first stage of the constriction process and that PBP3 would take over later on (Wientjes and Nanninga, 1989). This interpretation has been confirmed by studying the segregation pattern of fluoresescence-immunolabelled sacculi derived from ftsI filaments and proteinA–gold-immunolabelled sacculi derived from filaments after inhibition of PBP3 by aztreonam (De Pedro et al., 1997). In these experiments, no label was found in initiating constrictions. So far, the identity of PIPS has not been disclosed. Eventually, PBP3 has been localized to the site of division by immunofluorescence microscopy (Weiss et al., 1999).
The cellular location of PBP2 has not been demonstrated by microscopy so far. In this paper, we report on the localization of green fluorescent protein (GFP)–PBP2. As expected, we detected GFP–PBP2 as fluorescent patches at the circumference of the cell. However, quite unexpectedly, we also found GFP–PBP2 at the site of division, but preferentially not at the old cell poles. This raises the question whether PBP2 is a genuine component of the divisome. Remarkably, GFP–PBP2 could not localize to the cell centre in the presence of aztreonam, an antibiotic that binds to PBP3 with high specificity (Sykes and Bonner, 1985). This suggests that PBP2 localization at the divisome is dependent on active PBP3. Inhibition of PBP2 by mecillinam results in an increase in the cell diameter of the nascent cell pole. We suggest that PBP2, probably in combination with PBP1b and/or PBP1a, is required for lateral peptidoglycan synthesis and maintenance of the correct diameter during lateral and centripetal growth.
GFP–PBP2 complements PBP2ts at the restricted temperature
To study the localization of PBP2, a GFP–PBP2 fusion protein was constructed. Like the majority of the proteins involved in envelope growth, PBP2 is a protein of low abundance with only 58 ± 8 molecules per average MC4100 cell grown in M9 minimal medium at 35°C (Dougherty et al., 1996), and it is toxic when overproduced (Stoker et al., 1983). To mimic the wild-type situation as closely as possible, we used the pTRC99A expression vector with a promoter weakened by site-directed mutagenesis, as described by Chen et al. (1999). As GFP cannot fold in the periplasmic space, the GFPmut2 (Cormack et al., 1996) was fused via a small linker of three asparagines to the amino-terminus of PBP2. For comparison, the cell division protein FtsQ also occurs in about 20–40 copies per average cell (Guzmán et al., 1992), and a similar construct expressing GFPmut2–FtsQ caused a slight inhibition of cell division (average cell length 2.6 µm for wild-type LMC500 and 3.0 µm for LMC500pGFP–FtsQ; T. den Blaauwen, unpublished results). A GFP–PBP2 wild-type protein fusion (pTB017) was constructed and transformed to the wild-type strain LMC500 and to strain LMC582, which harbours the chromosomal pbpa137ts mutant (Woldringh et al., 1988). The latter strain grows as spheres at 28°C and, at 42°C in minimal glucose medium (GB1) and in TY at 42°C, it eventually dies (Woldringh et al., 1988). GFP–PBP2 was expressed constitutively without induction at a low level in strain LMC582 and in LMC500.
In the presence of GFP–PBP2, the LMC582 cells showed normal rod-shaped morphology in GB1 and in TY medium at 28°C (Table 1). This suggests that the GFP–PBP2 either completely complements or somehow sustains the activity of the temperature-sensitive PBP2 in these cells. After two mass doublings at 42°C in GB1, cell division in the LMC582 strain seems to be slightly impaired in the presence of GFP–PBP2. The average cell length of the LMC582 expressing GFP–PBP2 was 4.3 µm compared with 2.5 µm for wild-type LMC500 cells grown at 42°C (Table 1). In addition, the cells showed slightly elongated cell poles. After four mass doublings in TY at 42°C, LMC582 was completely spherical, whereas LMC582 expressing GFP–PBP2 with a average length of 4.2 µm was somewhat longer compared with the average length of 3.2 µm in the wild-type strain MC4100lysA (Table 1). After eight mass doublings at 42°C in TY, LMC582 died, whereas LMC582pGFP–PBP2 continued to grow normally. In conclusion, cells expressing GFP–PBP2 showed a slightly elongated but otherwise normal cell morphology.
Table 1. . Morphological parameters of E. coli strains expressing GFP–PBP2 constitutively, growing in minimal glucose (GB1) medium at 28°C and after two mass doublings at 42°C, and growing in rich (TY) medium at 28°C and after four mass doublings at 42°C.
. The cells looked like very large spheres and double spheres, which were difficult to measure with an automatic computer program.
PBP2 localizes in a confined pattern in the cytoplasmic membrane with a preference for the cylindrical part of the cell and at the site of division, in comparison with the old cell pole
In the pbpa137ts strain LMC582 as well as in the wild-type PBP2 strain LMC500, the GFP–PBP2 protein localizes as spots in the cylindrical part of the cytoplasmic membrane (Fig. 1). Remarkably, GFP–PBP2 seemed to have a preference for the lateral wall, because it was absent from the cell poles in the majority of the cells (Fig. 1).
To our surprise, GFP–PBP2 was found not only in the cylindrical part of the cell periphery but also at the site of constriction (Fig. 1). To validate this surprising result further, we analysed the frequency of occurrence of a PBP2 signal at the site of division and the presence of a constriction in cells grown to steady state in GB1 medium at 28°C with a generation time of 95 min. It appeared that the PBP2 signal occurs at mid-cell more or less simultaneously with the appearance of a constriction, as visible by phase-contrast microscopy (Fig. 2). The signal seems to disappear just before the two daughter cells separate (Fig. 2).
To exclude the possibility that (part of) the localization pattern could be caused by deposition of excess material, the GFP–PBP2 gene was integrated in the chromosome of LMC500 using the λInCh system (Boyd et al., 2000) to ensure single-copy gene expression. The presence of the integrated gene was confirmed by colony polymerase chain reaction (PCR), and the new strain LMC1840 was grown to steady state in GB1 at 28°C with a mass doubling time of 80 min. The morphology of the cell was identical to that of the parental strain (Table 1), and the chromosomal-encoded GFP–PBP2 localization pattern was virtually identical to that of the plasmid-encoded GFP–PBP2 (Fig. 3). Because the fluorescent signal in LMC1840 was very weak in comparison with that of the plasmid-encoded signal, all experiments were performed with the plasmid-encoded system.
PBP2 is active at the site of division
If PBP2 is somehow involved in cell division, addition of the PBP2 inhibitor mecillinam (Park and Burman, 1973) should affect the rate of septum formation. Therefore, the percentage of constrictions, the average cell length and diameter in a culture grown to steady state in minimal glucose medium at 28°C was determined after 0 min, 20, 30 and 50 min of growth in the presence of mecillinam. The generation time of the culture was 80 min and, at each time sample, at least 400 cells were analysed. Initially, the percentage of constrictions increases from 23.6% to 28.5% to become stabilized at 26% after 50 min, indicating a slight delay in cell division. The average diameter of the cells increased 6 ± 0.6% during the same period. The average diameter had been measured using a sliding window over the length axis of the cell. However, when inspecting the phase-contrast images of the cells, the old cell poles seemed to be thinner than the newly synthesized cell poles. To substantiate this observation, we determined the difference in diameter of the two cell poles at a number of positions in the cell poles of all cells in the untreated culture and in the 50 min-treated culture (Fig. 4). When the diameter was determined at two positions within the cylindrical part of the cell (Fig. 4, near measure position 1), the difference was the same for both cultures. However, if the diameters were measured near the transition from the cylindrical part to the spherical part of the cells (Fig. 4, measurement position 0.5), the mecillinam-treated culture had a much larger (26%) difference between the diameter of the old and the nascent poles than the control culture. Therefore, the initial effect of PBP2 inhibition seems to be an increase in the diameter of the nascent cell pole.
PBP2 seems not to be a subunit of the divisome
To determine whether PBP2 could be a stable component of the divisome, we compared its localization behaviour with that of other cell division proteins in aztreonam-treated cells. As mentioned before, the β-lactam antibiotic aztreonam can be used as a PBP3-specific antibiotic in studies of septation because, in competitive binding assays, aztreonam binds most strongly to PBP3 (Sykes and Bonner, 1985) and is also very slowly deacylated with a rate constant of 5 × 10−5 s−1 (Adam et al., 1991). PBP3 is not needed for the assembly of FtsZ at mid-cell to form the division-initiating FtsZ ring (Weiss et al., 1999). In a steady-state minimal glucose medium culture, about 53% of the cells (n = 500) have an FtsZ ring at mid-cell, as detected by immunofluorescence microscopy using a monoclonal antibody against FtsZ (den Blaauwen et al., 1999). After one mass doubling in the presence of aztreonam, this number increases to 94% (n = 445) (Figs 5 and 6) and, after two mass doublings in the presence of the PBP3 inhibitor, 94% of the cells (n = 471) still have an FtsZ at mid-cell (Fig. 5). Thus, the FtsZ ring at the mid-cell position seemed to be stabilized and unable to depolymerize for at least two generations. This stabilization is characteristic not only of FtsZ, but also of FtsW and PBP3 if not for all the other known cell division proteins that are part of the protein machine involved in cell division (T. den Blaauwen, unpublished results). If PBP2 were a stable component of the divisome during cell division, it would be expected to localize at mid-cell for more than one mass doubling in the presence of aztreonam. Therefore, LMC582pGFP–PBP2 and LMC500pGFP–PBP2 cells were grown in the presence of aztreonam in minimal glucose medium at 28°C for zero, one and two mass doublings. An example of PBP2 and PBP3 localization after growth in the presence of aztreonam for two mass doublings is shown in Fig. 7. The number of cells with a fluorescent PBP2 signal at mid-cell was determined to be 24.5% ± 2% (n = 4), 13% and 0% (n = 2) respectively. This deviates considerably from the percentages found for true divisomal proteins as outlined above. Therefore, it can be concluded that the localization behaviour of PBP2 is clearly different from that of the majority of the divisomal proteins, but seems to be dependent on, at least, an active PBP3.
The location of GFP–PBP2 at the site of division might seem to be a surprise. However, there are several observations that point to a(n) (in)direct role for PBP2 in the division process. (i) Early observations by James et al. (1975) revealed that mecillinam (then called FL 1060) inhibits the division process. (ii) Inhibition of PBP2 eventually results in the inhibition of cell division, which can be relieved by overexpression of FtsZ (Vinella et al., 1993). However, this is probably an indirect effect resulting from the change in surface to volume ratio that decreases the FtsZ concentration. (iii) Addition of cephalexin (inhibiting PBP3) or mecillinam to synchronized cultures showed a dramatic effect on peptidoglycan synthesis in dividing cells compared with elongating cells in the case of cephalexin, whereas there was little difference between the two types of cells in the case of mecillinam (Wientjes and Nanninga, 1991). It was therefore concluded that PBP2 activity continues during the division process. The initial effect of inhibition of PBP2 by mecillinam is a diameter increase in the nascent cell pole in comparison with the old cell pole (Fig. 4). This increase in diameter of the newly synthesized cell pole could indicate a local increase in peptidoglycan hydrolytic activity, resulting in an unstable murein network prone to osmotic swelling. Murein sacculi isolated from cells that have been grown for one mass doubling in the presence of mecillinam also have a newly synthesized fat pole and a thin old pole (De Pedro et al., 2001). This likewise suggests that PBP2 is active at the site of division and that it is required for the maintenance of the diameter of the new cell pole in particular.
Is PBP2 a stable part of the divisome?
The behaviour of PBP2 with respect to localization at putative division sites deviates from the behaviour of the divisomal proteins such as FtsZ, FtsW and PBP3 in filaments obtained by the addition of aztreonam. In the absence of septal peptidoglycan synthesis, PBP2 fails to localize, whereas other cell division proteins such as FtsZ and FtsW still do. Therefore, PBP2 does not seem to be a stable part of the divisome. In principle, authentic divisome components are the so-called essential cell division genes (reviewed by Margolin, 2000). However, as peptidoglycan synthesis is an indisputable agent of the E. coli cytokinetic process (reviewed by Nanninga, 1991; Höltje, 1998), numerous additional gene products involved in peptidoglycan metabolism (Van Heijenoort, 2001) are likely to have a role in the divisome's function. It is not known whether there is, say, a division-specific lipid II component. If not, one could envisage a general peptidoglycan assembly housekeeping function on which specific enzymatic functions are superimposed (Wientjes and Nanninga, 1991). PBP2, which can be found in the lateral wall and at mid-cell during cell division, might thus have a housekeeping role. In contrast, PBP3 most probably represents a genuine cytokinesis protein. In fact, one might consider lateral wall synthesis as the default pathway of peptidoglycan assembly.
Why has GFP–PBP2 a preference for the lateral wall compared with the old cell pole?
PBP2 hardly localizes at the old cell pole, which is inert with respect to peptidoglycan synthesis (De Pedro et al., 1997). In contrast, GFP fusions to other bitopic membrane proteins such as FtsQ are evenly distributed over the cytoplasmic membrane of non-dividing cells if produced in sufficient amounts (T. den Blaauwen, unpublished results). The integral membrane protein GFP–FtsW is likewise evenly distributed over the cytoplasmic membrane in non-dividing cells (T. den Blaauwen, unpublished results). This suggests that the GFP–PBP2 protein must somehow be diffusion restricted. It can be envisaged that PBP2 localizes at sites where it encounters substrate, i.e. possibly near the translocation sites in the cytoplasmic membrane of the peptidoclycan precursors and/or the transglycosylation essential proteins PBP1b or PBP1a. This implies that these proteins also might not be able to diffuse freely in the cytoplasmic membrane. A tentative explanation could be that, like cell division proteins in the divisome, the lateral cell envelope-synthesizing proteins could also form a complex. Similar to recruiting function of the cell division proteins of the FtsZ ring, a scaffolding structure would be required to assemble the lateral envelope-synthesizing proteins. A putative candidate for such a structure would be the actin homologue MreB (Van den Ent et al., 2001), which has been shown to polymerize as a helix underneath the cytoplasmic membrane in Bacillus. subtilis (Jones et al., 2001) and in E. coli (T. den Blaauwen, unpublished results). As a speculation, it might be suggested that a rotational movement of the MreB helix, enforced by growth of the polymer, would ensure a regular insertion of new peptidoglycan building blocks in the existing layer.
Bacterial strains and growth conditions
Escherichia coli K-12 cells were grown to steady state in glucose minimal medium containing 6.33 g of K 2HPO4.3H2O, 2.95 g of KH2PO4, 1.05 g of (NH4)2, 0.10 g of MgSO4.7H2O, 0.28 mg of FeSO4.7H2O, 7.1 mg of Ca(NO3)2.4H2O, 4 mg of thiamine, 4 g of glucose and 50 µg of lysine per litre, pH 7.0, at 28°C. Filamentation was achieved by specific inhibition of PBP3 (Sykes and Bonner, 1985) by a 1:1000 dilution with 1 mg l−1 aztreonam freshly resolved in a saturated Na2CO3 solution at an optical density of 0.025 at 450 nm and further growth in the presence of the antibiotic for two mass doublings before harvesting the cells. Spherical growth was achieved by the addition of 2 µg ml−1 mecillinam that specifically inhibits PBP2 (Park, 1987).
All E. coli strains were grown at 28°C, 37°C or 42°C in rich medium containing 50 g of bacto tryptone, 25 g of yeast extract, 15 mmol of NaOH per litre (TY). When required (Table 2), 25 µg ml−1 kanamycin, 200 µg ml−1 ampicillin or 5 µg ml−1 tetracycline was added to the medium. Absorbance was measured at 450 nm with a 300-T-1 spectrophotometer (Gilford Instrument Laboratories). Cell numbers were monitored using an electronic particle counter (orifice 30 µm). Cultures were considered to be in a steady state of growth if the ratio between optical density and number of cells remained constant over time (Fishov et al., 1995).
pTHV037 digested with NcoI–EcoRI ligated with BspHI–EcoRI gfpmut2 PCR fragment
pTHV038 with EcoRI–HindIIl PCR fragment of pbpa encoding PBP2
Expression of GFP fusion proteins
For the localization studies, E. coli PBP2 proteins were expressed as N-terminal GFP fusion proteins. A linker consisting of three asparagines separated the GFP and the protein of interest. Genomic DNA was PCR amplified using Pfu DNA polymerase and the following primers: PBP2F (5′-GGC GCC GAC GAA TTC AAC AAC AAC ATG AAA CTA CAG AAC TCT TTT CGC GAC TAT ACG GCA GAG TCC-3′) and PBP2R (5′-CGC GCC AAG CTT AAT GGT CCT CCG CTG CGG CAA C-3′). PCR fragments were digested with EcoRI–HindIII and ligated into cleaved pTHV038 vector. This vector is a pTRC99A derivative with a weakened promoter as described by Chen et al. (1999), in which GFPmut2 (Cormack et al., 1996) has been cloned (Table 2). Cells from various strains were transformed and produced GFP–PBP2 constitutively without induction with IPTG. The GFP–PBP2 gene was integrated into the chromosome using the λInCh method and strain DHB6500 as the lambda phage donor (Boyd et al., 2000).
Fixation and permeabilization
Cells were fixed in 2.8% formaldehyde (FA) and 0.04% glutaraldehyde (GA) in growth medium for 15 min at room temperature. The cells were collected at 8000 g, for 5 min, washed once in PBS (140 mM NaCl, 27 mM KCl, 10 mM Na2HPO4.2H2O, 2 mM KH2PO4, pH 7.2) and subsequently incubated in 0.1% Triton X-100 in PBS for 45 min at room temperature. All subsequent centrifugations were performed at 4500 g. The cells were washed three times in PBS and incubated in PBS containing 100 µg ml−1 lysozyme and 5 mM EDTA for 45 min at room temperature. Finally, the cells were washed three times in PBS.
Non-specific binding sites were blocked by incubating the cells in 0.5% (w/v) blocking reagents (Boehringer) in PBS for 30 min at 37°C. Incubation with F168-12 Fab fragments conjugated with AlexaTM 488 against FtsZ, diluted in blocking buffer, was carried out for 60 min at 37°C. The cells were washed three times with PBS containing 0.05% (v/v) Tween-20. The cells were washed once in H2O and resuspended in H2O.
Microscopy and image analysis
Cells were immobilized on 1% agarose in water slab-coated object glasses as described by van Helvoort and Woldringh (1994) and photographed with a CoolSnap-fx (Photometrics) CCD camera mounted on an Olympus BX-60 fluorescence microscope through a UPLANFl 100×/1.3 oil objective (Japan). Images were taken using the public domain program object-image 2.08 by N. Vischer (University of Amsterdam; http:simon.bio.uva.nlobject-image.html), which is based on NIH image by W. Rasband. In all experiments, the cells were first photographed in the phase-contrast mode, then with the AlexaTM 546 filter (U-MNG, excitation 530–550 nm) or with the AlexaTM 488 filter (U-MNB, excitation 470–490 nm). The two photographs were stacked, and the length and diameter of the cells were determined from the phase-contrast images, and the localization and intensity of the fluorescence signal was analysed in the fluorescence images of bacteria. First, full automatic pass identified filamentous objects in the thresholded phase-contrast image, where a ‘filament’ was assumed to be a straight or curved tubular shape with round caps. The axis of an object was marked with a segmented line, the vertices of which were found by travelling along the object and obtaining new mid-points by sampling perpendicular to the current travel direction. Then, the diameter (d) was derived from the area and the axis length, assuming the ‘filament’ shape described above.
Shape matching was performed as follows: the length axis was shortened at both ends by d/2, then each axis segment with length L was symmetrically overlaid with a rectangle L × d and, around each vertex of the length axis, a circle with diameter d was overlaid. The union of all rectangles and circles resulted in the modelled filament shape, which was compared against the real object. The acceptance threshold was adjusted so that slight shape deviations such as constrictions were tolerated.
In the second pass, the operator inspected each cell and added information about constriction and presence of spots in the fluorescence image, or rejected the cell where artifacts had occurred. All marking was performed with non-destructive vector overlay with the program object-image, which can be downloaded free from http:simon.bio.uva.nl.
The authors thank Drs T. M. F. Vinkenvleugel and M. L. G. F. Côrte for technical assistance, D. S. Weiss for the gift of strain EC522, and D. H. Boyd for the gift of strain DHB6500. This work was supported by a Vernieuwingsimpuls Grant (T. den Blaauwen) from the Netherlands Organization for Scientific Research (NWO).