In rod-shaped bacteria, septal peptidoglycan synthesis involves the late recruitment of the ftsI gene product (PBP3 in Escherichia coli) to the FtsZ ring. We show that in Caulobacter crescentus, PBP3 accumulates at the new pole at the beginning of the cell cycle. Fluorescence recovery after photobleaching experiments reveal that polar PBP3 molecules are, constantly and independently of FtsZ, replaced by those present in the cellular pool, implying that polar PBP3 is not a remnant of the previous division. By the time cell constriction is initiated, all PBP3 polar accumulation has disappeared in favour of an FtsZ-dependent localization near midcell, consistent with PBP3 function in cell division. Kymograph analysis of time-lapse experiments shows that the recruitment of PBP3 to the FtsZ ring is progressive and initiated very early on, shortly after FtsZ ring formation and well before cell constriction starts. Accumulation of PBP3 near midcell is also highly dynamic with a rapid exchange of PBP3 molecules between midcell and cellular pools. Localization of PBP3 at both midcell and pole appears multifactorial, primarily requiring the catalytic site of PBP3. Collectively, our results suggest a role for PBP3 in pole morphogenesis and provide new insights into the process of peptidoglycan assembly during division.
Synthesis of septal PG depends on the assembly of the cytoskeletal tubulin homologue FtsZ at the future site of division (Errington et al., 2003; Lutkenhaus, 2007). FtsZ assembly occurs before cytokinesis is initiated, and the FtsZ ring contributes to pre-septal cell elongation by recruiting MurG, an enzyme involved in the synthesis of the PG precursor Lipid II, a key intermediate in PG biosynthesis (de Pedro et al., 1997; van Heijenoort, 1998; Rothfield, 2003; Aaron et al., 2007; Mohammadi et al., 2007). The contribution of FtsZ to cell elongation can be considerable in bacteria like Caulobacter crescentus where there is a large temporal gap between FtsZ ring formation and the initiation of cell constriction (Aaron et al., 2007). Later, FtsZ mediates cell division by providing a scaffold for the sequential recruitment of other cell division proteins (to form the so-called divisome) (Errington et al., 2003; Goehring et al., 2006). In Escherichia coli, the mature divisome includes FtsZ, ZipA/FtsA/ZapA, FtsE/X, FtsK, FtsQ, FtsB/L, FtsW, PBP3, FtsN, AmiC and EnvC (Goehring and Beckwith, 2005; den Blaauwen et al., 2008). Each protein requires upstream proteins to localize, reflecting a temporal sequence of recruitment. A current model proposes that the assembly and maturation of the divisome occurs in two defined steps (Aarsman et al., 2005). First, FtsZ assembles into a ring and recruits early-division proteins (such as ZipA, ZapA and FtsA); this step promotes a switch in the predominant mode of PG synthesis from cylindrical to medial, which contributes to pre-septal elongation. In the second step, which occurs after a certain delay, late cell division proteins assemble at the FtsZ ring to form the mature divisome, right around the onset of cell constriction.
Here we examine the temporal and spatial distribution of PBP3 in the Gram-negative bacterium C. crescentus. The ease of obtaining cell cycle-synchronized populations and the large delay between FtsZ ring formation and cell constriction make this bacterium a valuable model for such study. The C. crescentus cell cycle starts with the ‘swarmer’ cell, which carries a flagellum at the ‘old’ pole (i.e. pole opposite to the ‘new’ pole formed by the last division). The swarmer cell elongates through patchy or seemingly helical insertion of PG synthesis along the lateral walls (Aaron et al., 2007; Divakaruni et al., 2007). This is followed by the assembly of the FtsZ ring and redistribution of most PG synthesis at the site of the FtsZ ring (Kelly et al., 1998; Aaron et al., 2007). At this stage, a stalk (a thin extension of the cell envelope involved in nutrient uptake and cell attachment) forms at the old pole after flagellum ejection. Elongation of the stalked cell produces a pre-divisional cell, which builds a flagellum at the new pole. Cell division generates a swarmer daughter cell (which repeats the sequence described above) and a stalked daughter cell, which skips the swarmer cell stage and immediately assembles the FtsZ ring. By simultaneously visualizing PBP3 and FtsZ fusions to fluorescent proteins by time-lapse microscopy, we establish the precise temporal sequence of PBP3 and FtsZ distribution during the cell cycle. While there is a notable delay between FtsZ ring formation and PBP3 assembly at the divisome, the recruitment of PBP3 to the FtsZ ring is initiated well before the onset of cell constriction and occurs very gradually. After division, PBP3 localizes at the new pole in a dynamic and FtsZ-independent manner. We also found that the transpeptidase catalytic site of PBP3 is important for both polar and divisomal localization of PBP3. These findings provide new insights into the processes of divisome assembly and PG synthesis.
The essential cell division protein PBP3 changes its spatial distribution during the cell cycle
Consistent with a conserved function for PBP3 in cell division, Ohta et al. (1997) showed that C. crescentus CB15 cells carrying an uncharacterized temperature-sensitive (Ts) mutation in the ftsI gene [ftsI(Ts), therein referred to as divA305] formed long cell filaments when grown at the restrictive temperature. We transferred this ftsI(Ts) mutation (linked to a gentamicin resistance cassette) to the synchronizable CB15N background (creating strain CJW2141) where the Ts cell division phenotype was recapitulated (Fig. 1A). Sequencing the ftsI(Ts) allele revealed a mutation (A808T) that causes an amino acid substitution (Y270N) of a highly conserved residue within the transpeptidase domain of PBP3 (Goffin and Ghuysen, 1998).
In E. coli, polymerization of FtsZ at midcell is a prerequisite step for the recruitment of PBP3 to the divisome complex (Wang et al., 1998; Weiss et al., 1999). While FtsZ localization contributes both to pre-septal elongation and cell division, PBP3 is thought to act exclusively during cell division. Consistent with this idea, E. coli PBP3 localizes after FtsZ ring formation at a time corresponding approximately to the initiation of cell constriction. This is inferred from the observation that PBP3 localizes at midcell in only about 50% of exponential-phase E. coli cells (Weiss et al., 1999) whereas 90% of cells in similar growth phase display midcell FtsZ localization (Ma et al., 1996). Moreover, independent fluorescence microscopy experiments using the appearance of cell constriction as a temporal landmark suggest a delay of 14–21 min between FtsZ ring formation and midcell recruitment of PBP3 (Den Blaauwen et al., 1999; Aarsman et al., 2005). To gain temporal resolution, we wanted to visualize FtsZ and PBP3 simultaneously while the cells proceeded through the cell cycle. We reasoned that the large time gap between FtsZ assembly and cell constriction in C. crescentus (Kelly et al., 1998) should help in visualizing any temporal differences that may exist among FtsZ assembly, PBP3 localization and cell constriction. We therefore integrated a gfp–ftsI fusion under the control of the xylose-dependent promoter (Pxyl) at the xylX chromosomal locus in a strain carrying ftsZ–mcherry under the vanillate-inducible promoter (Pvan) at the vanA locus (creating strain CJW1823, which also carries ftsI and ftsZ at their native loci). The functionality of GFP–PBP3 was inferred from its ability to rescue the Ts cell division defect of ftsI(Ts) cells (strain CJW2235) in a xylose-dependent fashion (Fig. 1B). Immunoblot analysis showed that the full-length GFP–PBP3 fusion [together with a minor C-terminal truncated form that may correspond to C-terminal processing (Nakamura et al., 1983; Hara et al., 1989; Nagasawa et al., 1989)] accumulated under xylose induction but not when synthesis was blocked by substitution of xylose by glucose (Fig. S1). Biochemical fractionation followed by immunoblotting (Fig. 1C) confirmed that GFP–PBP3 was membrane-bound. The TM protein TipN and the cytoplasmic protein ParB were used to verify the quality of the fractionation.
Time-course experiments of synchronized CJW1823 cell populations (CB15N Pxyl::Pxyl–gfp–ftsI Pvan::Pvan–ftsZ–mcherry) producing both GFP–PBP3 and FtsZ–mCherry revealed that the GFP–PBP3 signal, in addition to a cellular distribution, accumulated at the new pole (i.e. previous site of division) of swarmer cells (Fig. 2A), just as FtsZ (Thanbichler and Shapiro, 2006). The polar accumulation of GFP–PBP3 was unlikely to be due to an overproduction effect, as this fusion protein also localized at the pole in cells (CJW2144; CB15N ftsI::gfp–ftsI) expressing gfp–ftsI as a single copy under the control of the ftsI native promoter (Fig. 3A). While FtsZ–mCherry moved from the new pole to form a band (ring) near midcell early in the cell cycle as previously reported (Kelly et al., 1998; Thanbichler and Shapiro, 2006), the polar GFP–PBP3 focus remained apparent for a longer period of time (Fig. 2A). By the time cell constriction was visible, the polar GFP–PBP3 focus had disappeared and GFP–PBP3 displayed a strong band-like localization at the FtsZ–mCherry ring location (Fig. 2A). The FtsZ–mCherry and GFP–PBP3 colocalization persisted throughout cell constriction, consistent with the role of PBP3 in cell division. Division yielded daughter cells with a PBP3 focus at the new pole. It should be noted that all cells, irrespective of their cell cycle stage, also displayed a diffuse cellular fluorescent signal (Fig. 2A). This signal (which was above background, data not shown) largely corresponded to GFP–PBP3 proteins diffusively distributed around the membrane based on biochemical fractionation (Fig. 1C). In C. crescentus, membranal localization of TM proteins produced in low abundance is often not evident by epifluorescence microscopy (Wheeler and Shapiro, 1999; Obuchowski and Jacobs-Wagner, 2008). This is because C. crescentus cells are quite thin (about 0.5 μm of width) relative to other model bacteria such as E. coli or Bacillus subtilis.
To determine the relative localization of PBP3 and FtsZ with greater temporal resolution, we acquired short time-interval sequences of FtsZ–mCherry and GFP–PBP3 localization during the cell cycle, starting with swarmer cells. Kymographs of these time-lapse recordings were constructed to display the maximum pixel intensity of FtsZ–mCherry and GFP–PBP3 fluorescence distribution along the cell length as a function of time (Fig. 2B; the kymographs were inverted to facilitate visualization of the fluorescence signal). They showed that both FtsZ–mCherry and GFP–PBP3 accumulate at the new pole during the swarmer cell stage. During the swarmer-to-stalked cell transition, FtsZ–mCherry signal moves from the pole and then exhibits a period of high mobility within the medial cell region before stabilizing at one position (Fig. 2B), in agreement with a previous report (Thanbichler and Shapiro, 2006). Surprisingly, kymograph analysis consistently showed that during a certain period of time (shown in light orange in Fig. 2B), the GFP–PBP3 signal is split between a medial localization and a polar accumulation. This split starts at 0.25 ± 0.07 cell cycle units (n = 17) shortly after FtsZ–mCherry stabilizes at one position at 0.15 ± 0.06 cell cycle units (n = 17). Over time, the medial GFP–PBP3 signal progressively increases at the expense of the polar signal. By the time cell constriction initiates (0.51 ± 0.10 cell cycle units, n = 17), all polar signal has disappeared and GFP–PBP3 exhibits a strong medial signal. (Note that the fluorescence signal corresponding to the GFP–PBP3 cellular pool was subtracted in the kymograph shown in Fig. 2B for better visualization of the overlapping period between polar and medial localization of GFP–PBP3.) The time overlap between polar and medial localization was also observed when GFP–PBP3 was expressed at endogenous levels under native conditions (Fig. 3B). Thus, a fraction of PBP3 molecules is recruited to the FtsZ ring in the elongation phase shortly after FtsZ assembly and well before cell constriction becomes visible. Accumulation of PBP3 at the FtsZ ring is clearly progressive and apparently at the expense of its polar localization.
FtsZ is required for the localization of PBP3 to the division site, but not for its maintenance at the pole
Next, we investigated the dependency of PBP3 localization on FtsZ. For this, we generated a strain (CJW2143, CB15N ftsI::gfp–ftsI ftsZ::Pxyl–ftsZ) that carries a gfp–ftsI fusion under native expression (as the only ftsI copy) and ftsZ under xylose-dependent expression to visualize GFP–PBP3 under conditions of FtsZ depletion. After growth in xylose-containing medium, swarmer cells of CJW2143 were obtained and re-suspended in liquid medium containing either xylose to maintain FtsZ synthesis or glucose to initiate immediate FtsZ depletion (Wang et al., 2001). Under FtsZ-depletion conditions (glucose), GFP–PBP3 failed to accumulate near midcell (Fig. 4A; representative time points are shown). Instead, GFP–PBP3 retained a polar localization for extended periods of time in the elongating FtsZ-depleted cells. Quantitative analysis (Fig. 4B) showed that the percentage of cells with a polar GFP–PBP3 focus remained relatively steady over time in the absence of FtsZ (glucose). Under condition of FtsZ synthesis (xylose), the cells exhibited a normal cell cycle profile of PBP3 localization with the polar accumulation being progressively replaced by a medial localization (Fig. 4C). The reappearance of polar GFP–PBP3 in some FtsZ-producing cells at the last time points (80 min; Fig. 4C) corresponded to cells that had divided (data not shown). These results show that FtsZ is required for PBP3 recruitment to the future division site, as expected. They also confirm that the medial localization of PBP3 comes at the expense of polar localization as polar localization persists when medial localization is prevented, consistent with a competitive relationship between the two localization sites.
PBP3 is dynamically recruited to the pole in an FtsZ-independent manner
A trivial explanation for the polar localization of PBP3 is that it corresponds to a remnant of PBP3 molecules from the previous divisome. If that were the case, polar PBP3 molecules would be stationary, and photobleaching the polar focus of GFP–PBP3 should lead to no fluorescence recovery, as there would be no exchange between polar GFP–PBP3 and the diffuse membranal pool of PBP3. However, fluorescence recovery after photobleaching (FRAP) experiments in CJW1823 cells (CB15N Pxyl::Pxyl–gfp–ftsI Pvan::Pvan–ftsZ–mcherry) showed that the bleached polar GFP–PBP3 signal recovered very rapidly with an average half-time of 25 ± 9 s (n = 38 cells) (Fig. 5A). We selectively bleached cells in which FtsZ–mCherry was no longer polar to show that fluorescence recovery occurred independently of FtsZ (Fig. 5A). Consistent with rapid exchange of GFP–PBP3 molecules between polar and membrane pools, the fluorescence intensity of unbleached cellular regions (after normalization to account for photobleaching from image acquisition) decreased concomitantly with the fluorescence recovery in the bleached polar region (Fig. 5A). We also observed rapid signal recovery at the pole after photobleaching in cells that did not produce an FtsZ–mCherry fusion (Fig. S2A; half-time = 19 ± 7 s, n = 83 cells) or that were treated with chloramphenicol (CAM) to block protein synthesis (Fig. S2B; 28 ± 6 s, n = 38 cells). This demonstrates that the fast fluorescence recovery was largely due to an exchange between polar and membranal pools, and not because of de novo synthesis of GFP–PBP3. It was not artificially caused by the overproduction of GFP–PBP3 or the presence of untagged PBP3 because we also obtained short half-time values of polar recovery (11 ± 6 s, n = 27 cells) with cells that produced GFP–PBP3 at native levels as the only source of PBP3 (Fig. S2C). Thus, the polar accumulation of PBP3 does not simply correspond to a remnant from the last division. Instead, PBP3 is actively recruited to the pole in a very dynamic manner independently of FtsZ.
GFP–PBP3 is dynamically recruited to the division site
It is often implicitly assumed that mature divisome complexes remain at the constriction site during the whole division process. However, the fact that PBP3 was found to be dynamically recruited to the cell pole led us to investigate whether PBP3 localization at the division complex is also a dynamic process. Similarly to polar PBP3, the medial GFP–PBP3 signal rapidly recovered after photobleaching with an average recovery half-time of 18 ± 5 s (n = 146 cells; Fig. 5B). When we measured the fluorescence intensity of unbleached regions, including only the ‘swarmer half’ or the ‘stalked half’ of pre-divisional cells, we observed that regions in both halves decreased similarly in intensity during the course of the FRAP experiment (Fig. 5B), showing no preference. Similar values of fluorescence recovery half-time were obtained in the presence of CAM (12 ± 7 s, n = 23 cells; Fig. S2D) or when gfp–ftsI was expressed at native levels as the only ftsI copy (16 ± 9 s, n = 37; Fig. S2E). Our results demonstrate that GFP–PBP3 molecules at the division site are rapidly exchanged with molecules from the diffuse membranal pool, indicating that no single population of PBP3 molecules continuously participates in septal PG synthesis for long periods of time.
The catalytic site of PBP3 is important for PBP3 localization
In E. coli, the N-terminal sequence of PBP3 and in particular its TM segment seem to contain the structural determinants of localization. An N-terminal 56-residue fragment encompassing the cytoplasmic sequence, the TM segment and nine downstream periplasmic residues recapitulates PBP3 septal localization (Piette et al., 2004). Point mutations in the PBP3 TM segment or TM sequence swapping with another protein (MalF) disrupt midcell localization (Weiss et al., 1999; 2005; Wissel and Weiss, 2004). Furthermore, a short 26-residue sequence containing only the TM segment is sufficient to mediate ring-like localization, albeit not as well as the full-length protein (Wissel et al., 2005). Surprisingly, the TM sequence of PBP3 is not well conserved among bacteria, suggesting that either PBP3 targeting sequence varies among species or the TM sequence has co-evolved with its target protein(s) (Wissel et al., 2005). We tested these ideas by generating a gfp fusion to the C. crescentus ftsI sequence encoding the N-terminal 90-residue sequence of C. crescentus PBP3. This fusion was designed to encode the cytoplasmic region (which is slightly longer than that of E. coli PBP3), the TM segment and the next 26 periplasmic residues of C. crescentus PBP3. This fusion displayed no discernible ring-like localization in the merodiploid strain carrying wild-type ftsI (Fig. 6A) or in the ftsI(Ts) mutant at either permissive or restrictive temperatures (Fig. 6B). Biochemical fractionation followed by immunoblotting (using GFP antibody) verified membrane attachment (Fig. 6C). The appearance of degradation products suggested that the GFP–PBP31−90 fusion was not fully stable. However, the products, including the smallest detectable one (asterisk, Fig. 6C), was membrane-bound, suggesting that the TM segment was present. The absence of ring-like localization (Fig. 6A and B) therefore argues that the TM sequence of PBP3 is not a major targeting determinant in C. crescentus.
A role for PBP3 catalytic activity in localization has been previously suggested based on the observation that treatment of E. coli with β-lactam antibiotics such as cephalexin and aztreonam that specifically block PBP3 transpeptidase activity affects PBP3 localization (Wang et al., 1998). However, later studies show that the localization of PBP3 is unaffected by similar drug treatments (Weiss et al., 1999; Den Blaauwen et al., 2003; Bertsche et al., 2006). We could not use cephalexin for our studies because this antibiotic displayed a broader spectrum of activity in C. crescentus, inhibiting binding of a fluorescent penicillin derivative (BOCILLIN-FL) to several PBPs at a similar level when added at concentrations that induce cell filamentation (Fig. S3A). This competitive binding assay was performed in a Δbla strain in which the β-lactamase-encoding gene is inactivated. Aztreonam was also useless, as even C. crescentusΔbla is resistant to it (data not shown). Instead, we used fosfomycin, which blocks PG precursors synthesis by inhibiting the MurA enzyme (Kahan et al., 1974). While this drug does not block PBP3 activity per se, it should ultimately reduce the availability of PBP3 substrates. CJW1823 cells (CB15N Pxyl::Pxyl–gfp–ftsI Pvan::Pvan–ftsZ–mCherry) producing GFP–PBP3 and FtsZ–mCherry were treated with 40 μg ml−1 of fosfomycin. Under these conditions, cell growth was arrested after 3 h of exposure (data not shown), suggesting a substantial depletion of PG precursors. The localization of GFP–PBP3 was quantified relative to FtsZ–mCherry rings, which served as a spatial reference. Without drug treatment, GFP–PBP3 accumulation at the pole and colocalization with FtsZ–mCherry were not impaired; however, after 3 h of fosfomycin treatment, the polar localization of PBP3 and its colocalization with FtsZ–mCherry were both appreciably reduced (Fig. S3B). It is likely that the pool of PG precursors was not completely depleted under the conditions used (which would be difficult to achieve without inducing significant cell lysis). Regardless, our results showed that reduction of PG precursors, which ultimately leads to a reduction of glycan strand incorporation and thus PBP3 substrates, is accompanied with a small but noticeable decrease in PBP3 localization.
A more conclusive way to address the role of transpeptidase activity in PBP3 localization was to disrupt the catalytic site of PBP3 by mutation. By sequence alignment, the active site of C. crescentus PBP3 was identified as serine-296, which corresponds to serine-307 in E. coli PBP3 (Houba-Herin et al., 1985; Nicholas et al., 1985). An ftsI allele creating a S296A substitution was fused to gfp and placed under Pxyl. Consistent with an inactivation of the catalytic site, production of GFP–PBP3S296A did not complement the ftsI(Ts) allele at 37°C in cells [CJW2236, CB15N ftsI(Ts) Pxyl::Pxyl–gfp–ftsIS296A] grown in xylose (Fig. 7A). Next, we examined the localization of inactive GFP–PBP3S296A in cells (CJW2237, CB15N Pxyl::Pxyl–gfp–ftsIS296A Pvan::Pvan–ftsZ–mCherry) that also produced FtsZ–mCherry (to serve as a spatial reference). We used synchronized swarmer and pre-divisional cell populations, which, for wild-type PBP3, predominantly exhibit a polar and medial localization respectively. While medial FtsZ–mCherry localization was normal, the percentage of swarmer and pre-divisional cells with polar and medial GFP–PBP3S296A localization was significantly reduced (Fig. 7B and C). PBP3 is thought to dimerize (Broome-Smith et al., 1985; Weiss et al., 1999; Marrec-Fairley et al., 2000; Wissel and Weiss, 2004), raising the possibility that the residual localization of GFP–PBP3S296A might be in part due to heterodimer formation between GFP–PBP3S296A and wild-type PBP3.
We therefore moved the ftsIS296A allele into the ftsI(Ts) mutant background [CJW2236, CB15N ftsI(Ts) Pxyl::Pxyl–gfp–ftsIS296A]. Immunoblot analysis revealed that at both 30°C and 37°C, GFP–PBP3S296A accumulated to a lower steady-state level than wild-type GFP–PBP3 under the same xylose-inducing conditions (Fig. S1, lanes 5 and 6 compared with lanes 1 and 3). This decrease in GFP–PBP3S296A levels was also observed when cells expressed a wild-type allele of ftsI[instead of ftsI(Ts)] (Fig. S1, lane 8 compared with lane 7), suggesting that the functionality of the catalytic site is important for protein stability. We do not believe that this defect affected our ability to examine the localization of GFP–PBP3S296A, because in all cases (irrespective of temperature or ftsI background), the levels of xylose-induced GFP–PBP3S296A were well above those of an endogenously expressed GFP–PBP3 (Fig. S1, lanes 5, 6 and 8 compared with lane 9) whose medial and polar localizations were readily detectable (Fig. 3). Consistent with this inference, synthesis of GFP–PBP3S296A in the ftsI(Ts) background at 30°C resulted in one or two strong fluorescent bands in all cells that were observed (100%, n = 168) (Fig. 7D and E). While this clearly indicates that the steady-state level of GFP–PBP3S296A is not a problem for visualizing localization, it was surprising that the localization of GFP–PBP3S296A was enhanced when the PBP3(Ts) mutant was present instead of wild-type PBP3. This was accompanied by a dominant effect on cell division as most cells were noticeably elongated (Fig. 7A).
At 37°C when the function of the PBP3(Ts) mutant was abrogated and all cells were filamentous, only about 20% and 3% of cells displayed one or two bands of GFP–PBP3S296A respectively (Fig. 7D and E). Thus, in total, GFP–PBP3S296A exhibited a band-like localization in only about 23% of ftsI(Ts) cells at 37°C. Even in these cases, the localization signal was considerably weaker than that at 30°C (arrow, Fig. 7D). The steady-state level of GFP–PBP3S296A at 37°C is similar to that at 30°C at which temperature, band localization is readily discernible in all cells (Fig. S1, lane 6 versus lane 5), indicating that the localization defects at 37°C are not caused by the cellular levels of GFP–PBP3S296A. Instead, they argue that the active site of PBP3 plays an important role in PBP3 localization.
This interpretation stands only if FtsZ rings are able to form in ftsI(Ts) cells at 37°C, as they do in E. coli (Addinall et al., 1996). Otherwise, impairment in FtsZ ring formation would explain the localization defect of GFP–PBP3S296A at 37°C. To verify that FtsZ rings form in the absence of functional PBP3 in C. crescentus, we examined FtsZ–YFP localization in ftsI(Ts) cells [CJW2577, CB15N ftsI(Ts) Pvan::Pvan–ftsZ–yfp] after incubation at 37°C. Over 90% of these filamentous ftsI(Ts) cells (n = 95) formed one or more FtsZ–YFP rings (Fig. 7E and F). About 29% of these cells had two rings per cell, while 10% and 6% of the cells displayed three and four (or more) rings respectively (data not shown). Thus, FtsZ can still form rings in the absence of functional PBP3. In contrast, abolition of functional PBP3 dramatically reduces the localization of the catalytically inactive PBP3S296A mutant, both in percentage of cells with band-like localization and in signal intensity (relative to its signal at 30°C). These observations suggest that in C. crescentus, the localization of PBP3 relies in large part on a functional PBP3 transpeptidase domain.
Temporal changes in PBP3 subpopulations
PBP3 localization varies during the cell cycle. At the beginning of the cell cycle, PBP3 accumulates at the new cell pole, creating a polar pool. Polar localization of PBP3 has also been reported in E. coli in one immunofluorescence microscopy study (Weiss et al., 1997). Later, using fluorescence microscopy and GFP tagging, polar localization of E. coli PBP3 was only observed when the protein was overproduced by 10-fold suggesting to the authors that previous observations of polar PBP3 may have been artificially produced (Weiss et al., 1999). In C. crescentus, polar accumulation is also observed when GFP–PBP3 is produced under native conditions (Fig. 3).
Later during the cell cycle, the polar accumulation of PBP3 disappears concomitantly with the re-localization of PBP3 to the FtsZ ring, generating a medial pool; this event is FtsZ-dependent. PBP3 and FtsZ remain colocalized throughout the constriction process that leads to the formation of new cell poles. Interestingly, there is also a portion of PBP3 molecules diffusely dispersed around the membrane at all times (i.e. during both cell elongation and division). The FRAP experiments demonstrate that this membranal pool of PBP3 is not static as it rapidly exchanges PBP3 molecules with the polar and medial pools, presumably through diffusion.
A role for a functional transpeptidase domain in PBP3 localization
A number of studies have suggested that PBP3 interactions with other divisomal proteins are important for PBP3 localization at midcell (den Blaauwen et al., 2008). Such interactions are likely responsible for the residual localization of the inactive PBP3S296A mutant in the filamentous ftsI(Ts) cells (Fig. 7D and E). This suggests that PBP3 localization is mediated by both protein–protein interaction and covalent substrate binding during transpeptidation, with the latter mechanism being responsible for the large localization defect of the catalytic site mutant. A similar substrate-driven mechanism of localization has been proposed for other PBPs in some Gram-positive bacteria (Morlot et al., 2004; Pinho and Errington, 2005). Alternatively, in C. crescentus, the interaction of PBP3 with its target protein(s) might be under allosteric regulation such that it might be favoured when the PBP3 enzyme is active. This does not appear to be the case in E. coli where a short N-terminal PBP3 fragment that includes the TM region is sufficient to mediate septal localization (Piette et al., 2004; Wissel et al., 2005). The observation that the TM sequence fails to localize in C. crescentus is consistent with its poor sequence conservation among bacteria and highlights the importance of studying homologues from different species.
Assembly of the mature divisome is thought to occur in two steps (Aarsman et al., 2005). First, FtsZ and early-division proteins assemble, and then, after a notable delay, PBP3 and other late-division proteins are simultaneously recruited to the FtsZ ring to form the mature divisome. Kymograph analysis of the relative spatial distribution of PBP3 and FtsZ during the C. crescentus cell cycle confirms the existence of a delay between FtsZ ring assembly and the start of PBP3 accumulation at the FtsZ ring. Interestingly, the initiation of PBP3 accumulation near midcell corresponds to about the time when the FtsZ ring stabilizes at one position.
The kymographs also demonstrate that the recruitment of PBP3 to the FtsZ ring is initiated well before cell constriction begins, and is very gradual, spanning a considerable part of the elongation phase (Figs 2B and 3B). What drives this slow and early accumulation of PBP3 at the FtsZ ring? The fact that this gradual accumulation is observed whether the GFP–PBP3 fusion is under native cell cycle-controlled synthesis (Fig. 3) or under constitutive production (Fig. 2) rules out the possibility that this is due to transcriptional control. The FRAP measurements of GFP–PBP3 dynamics clearly indicate that membrane diffusion is very rapid (Fig. 5 and Fig. S2), as previously demonstrated for other membrane-bound proteins (Deich et al., 2004; Mullineaux et al., 2006). Therefore, the progressive recruitment of PBP3 to the divisome is not limited by PBP3 diffusion. Instead, it may be caused by a slow and early accumulation of an upstream interacting division protein. Alternatively, it may be driven by the accumulation of PBP3 substrates. It is known that shortly after its formation, the FtsZ ring largely redirects PG synthesis near midcell (Aaron et al., 2007). PG synthesis at the FtsZ ring position may thus recruit PBP3 at this location by generating substrates for PBP3. Interactions with other divisomal proteins may then strengthen PBP3 localization. In this scenario, the gradual accumulation of PBP3 at the FtsZ ring would result from two competing sites that are rich in PBP3 substrates: the FtsZ ring location where new substrates are being produced, and the new pole where PBP3 substrates may remain from the previous division. Without repletion of PBP3 substrates, the polar site may slowly lose its ability to recruit PBP3 compared with the FtsZ ring site over time, causing a gradual accumulation of PBP3 at the FtsZ ring location.
Using short-lived complexes for PG synthesis
Our FRAP experiments reveal that PBP3 molecules at the division site are continuously being exchanged with PBP3 present in the membrane pool, and this exchange occurs within seconds. Compared with this time scale, the formation of poles during division is a very long process. This indicates that septal PG synthesis is not achieved by the same mature divisome complexes. Instead, the contribution of any one specific PBP3-containing complex to septal PG synthesis is very short-lived.
A role for PBP3 at the new pole?
The FRAP experiments also demonstrate that the accumulation of PBP3 at the new pole of newborn cells does not correspond to a remnant from the previous cell division given the rapid and FtsZ-independent exchange of PBP3 between polar and membrane pools. This might mean that after division, PBP3 continues performing transpeptidation (cross-linking) reactions at the new pole for a significant period of time. Division in C. crescentus generates hemispherical poles, which become gradually more pointed following division (Aaron et al., 2007). This morphological pole maturation event must involve PG synthesis or remodelling, and the polar PBP3 activity may contribute to this activity. We unsuccessfully attempted to test this hypothesis by using the ftsI(Ts) mutant (data not shown). The loss of PBP3 activity appeared to be too gradual and not immediate enough following temperature shift to provide useful information. Examining the role of PBP3 localization at the pole might require obtaining mutations or conditions that allow PBP3 targeting to the divisome while impairing its localization to the pole. As far as we can tell, there is no evidence to suggest that these two localization events use different mechanisms, suggesting that it might be difficult to uncouple them. Interestingly, it has been reported that a single amino acid substitution (N361S) in PBP3 generates E. coli cells with pointed poles (Taschner et al., 1988). We found that in this E. coli mutant, the formation of pointed poles occurs during cell division (Movie S1) and not following division, as is observed in wild-type C. crescentus (Aaron et al., 2007).
Bacterial strains, plasmids and growth conditions
Strains and plasmids are listed in Table 1 and their mode of construction is provided in Supporting information. Plasmids were constructed using the E. coli strains DH5α and S17-1, routinely grown in LB medium at 37°C with appropriate antibiotic selection. C. crescentus strains were grown at 30°C or 37°C [restrictive temperature for the ftsI(Ts) mutant] to log-phase in rich PYE medium or minimal M2G medium supplemented with 1% PYE (M2G+) (Ely, 1991). C. crescentus was genetically manipulated by transformation, conjugation and transduction as previously described (Ely, 1991). Synchronous cell populations were obtained as previously described (Evinger and Agabian, 1977). Expression from Pxyl or Pvan was achieved by adding 0.3% xylose or 0.5 mM vanillic acid to the medium when indicated.
KanR integrational vector carrying the sacB gene that allows plasmid remova after integration. Limus38 derivative
For time-course experiments, swarmer cells were incubated at 30°C in liquid PYE medium (containing xylose, glucose or vanillic acid as appropriate) to resume growth. At specific time intervals, an aliquot of the culture was taken. The cells were then washed in M2G+ medium and visualized on agarose-padded slides containing M2G+ medium.
For time-lapse experiments, cells were observed on slides padded with 1% agarose in M2G+ medium supplemented with 0.3% xylose and 0.5 mM vanillic acid. Time-lapse experiments with C. crescentus and E. coli were performed at room temperature and 37°C respectively. Kymographs were constructed using MetaMorph Software to display the maximum intensity values of fluorescent signals along the length of a 3-μm-wide region across the long axis of the cell as a function of time.
Samples were observed on a Nikon E1000 microscope equipped with Hamamatsu Orca-ER camera, differential interference contrast (DIC) and phase-contrast 100× objectives, and standard filter sets for visualization of GFP, YFP and mCherry. Image acquisition was performed using MetaMorph software (MDS Analytical Technologies). When appropriate, a heated objective was used. Quantitative analysis of protein localization was performed visually on still images.
Fluorescence recovery after photobleaching
FRAP experiments were performed using a Nikon Eclipse 80i microscope equipped with an Andor iXonEM+ (DU-897E) EMCCD camera and a galvanometer-controlled Photonic Instruments MicroPoint laser. Cells cultures immobilized on 1.5% agarose-padded slides containing M2G+ medium were observed with a DIC 100× objective, 2× magnification optivar and standard filter sets. Photobleaching was achieved using two pulses of 37 ms with 30% laser intensity, which was additionally attenuated by using external attenuator set to 2 and the attenuation plate set to 45, plus a ND16 filter. Images were collected before photobleaching (pre-bleach), immediately after bleaching and at 5–10 s intervals after bleaching for 2 min. Measurements of integrated fluorescence intensity in the bleached or unbleached regions, as well as whole-cell intensity, were obtained at each time point with MetaMorph and exported to Excel. For each measured region, background intensity was subtracted and the fluorescent signal normalized for photobleaching occurring during image acquisition (by measuring the fluorescence decay in unbleached neighbouring cells). The averaged pre-bleach fluorescence intensity was normalized to 100% by dividing the averaged intensity of all time points by the averaged pre-bleach intensity. All analysed samples were found to fit a normally distributed population with a 99% confidence interval. Recovery half-times were determined by performing a least-squared fit of the intensity of the bleached region over time to the single-exponential equation F(t) = A(1 − e−B*t), where F(t) is the fluorescence at time t, A is the maximum intensity and B the slope. Least-squared fits were performed in Excel using the Solver function. All obtained fitting curves showed r2 ≥ 0.95. Half-time recovery values were determined by the formula t1/2 = ln(0.5)/−B and the standard deviation was calculated using the SETXY formula of Excel. Data were collected for over 23 cells from at least two independent time-lapse experiments. After the completion of the FRAP experiments, control time-lapse experiments were performed to ensure that the photobleached cells grew and divided. To test recovery under conditions in which protein synthesis is blocked, gfp–ftsI expression was repressed by incubating the cells in 0.2% glucose for 30 min, then 20 μg ml−1 chloramphenicol (CAM) was added for 20 min, and then the cells were placed on M2G+ agarose-padded slides containing 20 μg ml−1 CAM. The whole-cell intensity was shown to remain constant for 60 min under these conditions.
After growth, cells were harvested and re-suspended in 50 mM sodium phosphate buffer (pH 7.4) containing 50 mM NaCl, 1 mM EDTA and 1× protease inhibitor cocktail (Roche). Whole-cell lysates were prepared by sonication and 0.3 ml of lysates were centrifuged at 14 000 r.p.m. for 10 min at 4°C or prepared by French pressed at 1500 psi and the cell lysate was centrifuged at 10 000 r.p.m. for 15 min at 4°C. The supernatant constituted the whole-cell lysate. All whole-cell lysates were then centrifuged at 120 000 r.p.m. in a TLA100.2 rotor for 1 h at 4°C using an Optima TLX ultracentrifuge (Beckman Coulter). The supernatant, corresponding to the cytoplasmic fraction, was collected while the pellet was re-suspended in 0.3 ml of the same phosphate buffer to produce the membrane fraction.
For immunoblot analysis of cell extracts, 3 ml of cultures grown under appropriate conditions were collected and the whole-cell lysates prepared by sonication. Thirty micrograms (15 μg for biochemical fractionation) of total protein was electrophoretically resolved on 12% SDS-PAGE gels, and the resolved proteins were electrotransferred to PVDF membranes, which were probed with anti-GFP (1:1000; Clontech), anti-ParB (1:4000) (Mohl and Gober, 1997) and anti-TipN (1:12 500) antibodies (Lam et al., 2006)
BOCILLIN-FL competition binding assays were performed from late-log-phase cultures (A660 = 0.5–0.8). Cells were first lysed by treatment with 100 μg ml−1 lysozyme in the presence of 10 U DNase I, and centrifuged. The membrane-enriched fraction (insoluble fraction) was re-suspended in a buffer (100 mM Tris-HCl pH 7.5, 1 mM EDTA pH 8.0) containing increasing amounts of cephalexin. Samples were incubated for 45 min at 30°C, washed and incubated under similar conditions with 10 μM BOCILLIN-FL, a fluorescent derivative of penicillin (Molecular Probes). Samples were then washed and the proteins separated by electrophoresis on 10% SDS-PAGE gels. BOCILLIN-FL-bound proteins were observed using a Typhoon imager Trio+ (Molecular Dynamics, Amersham Biosciences) under excitation and emission of 488 and 532 nm respectively. A large excess of penicillin G, which competes with BOCILLIN-FL, was used as a specificity control for the identification of PBPs.
We thank A.F. Jackson, N. Ohta, A. Newton, Y.V. Brun, P. Taschner, M. Thanbichler and L. Shapiro for providing strains, M. Mooseker for the use of his Ultracentrifuge, R. Breaker for the use of his Typhoon scanner, W. Vollmer and T. Emonet for valuable discussions, S. Poggio and P. Montero for their help with the sonication, and W. Vollmer and all members of the Jacobs-Wagner laboratory for critical reading of the manuscript. This work was supported by Anderson Postdoctoral Fellowships (to R.P. and T.C.), Fellowship SFRH/PBD/26232/2006 from ‘Fundação para a Ciência e a Tecnologia’ (to T.C.), National Institutes of Health Grants GM076698 and GM065835 (to C.J.-W.) and the Pew Charitable Trusts (to C.J.-W.). C.J.-W. is a Howard Hughes Medical Institute Investigator.