Chlamydia co-opts the rod shape-determining proteins MreB and Pbp2 for cell division


E-mail; Tel. (+33) 14 438 9311; Fax (+33)14 061 3042.


Chlamydiae are obligate intracellular bacterial pathogens that have extensively reduced their genome in adapting to the intracellular environment. The chlamydial genome contains only three annotated cell division genes and lacks ftsZ. How this obligate intracellular pathogen divides is uncharacterized. Chlamydiae contain two high-molecular-weight (HMW) penicillin binding proteins (Pbp) implicated in peptidoglycan synthesis, Pbp2 and Pbp3/FtsI. We show here, using HMW Pbp-specific penicillin derivatives, that both Pbp2 and Pbp3 are essential for chlamydial cell division. Ultrastructural analyses of antibiotic-treated cultures revealed distinct phenotypes: Pbp2 inhibition induced internal cell bodies within a single outer membrane whereas Pbp3 inhibition induced elongated phenotypes with little internal division. Each HMW Pbp interacts with the Chlamydia cell division protein FtsK. Chlamydiae are coccoid yet contain MreB, a rod shape-determining protein linked to Pbp2 in bacilli. Using MreB-specific antibiotics, we show that MreB is essential for chlamydial growth and division. Importantly, co-treatment with MreB-specific and Pbp-specific antibiotics resulted in the MreB-inhibited phenotype, placing MreB upstream of Pbp function in chlamydial cell division. Finally, we showed that MreB also interacts with FtsK. We propose that, in Chlamydia, MreB acts as a central co-ordinator at the division site to substitute for the lack of FtsZ in this bacterium.


Chlamydia is a medically important, obligate intracellular pathogen causing a range of acute and chronic diseases in humans and animals (Schachter et al., 1973; Brunham et al., 1985; Grayston, 1992; Taylor-Robinson et al., 1992; Mabey et al., 2003). Chlamydia alternates between two morphologically and functionally distinct forms during its developmental cycle (see Abdel-Rahman and Belland, 2005 for review). The non-dividing elementary body (EB) mediates attachment to and uptake by susceptible host cells. The spore-like EB is adapted for extracellular survival and has a highly disulphide cross-linked outer membrane, histone-compacted DNA, little to no metabolic activity, and comes pre-packaged with the necessary type III effectors to effect its internalization. Once inside the cell, the EB begins differentiating to the reticulate body (RB) and remains within the pathogen-specified vacuole, called an inclusion, for the duration of its developmental cycle. The RB is metabolically active and capable of cell division but non-infectious. The bacteria continue to multiply within the inclusion until such time that they begin to asynchronously differentiate back to the infectious EBs, which are released from the cell to initiate a new round of infection. Due to this complex developmental program, there are no methods currently available for generating targeted mutations in these bacteria.

A unique feature of chlamydiae is the ability to enter a persistent growth state in response to stress (see Beatty et al., 1994 for review). Interferon-γ activation of infected host cells, iron or nutrient limitation, and penicillin treatment are all capable of inducing chlamydial persistence. A hallmark of persistence is the abrogation of cell division leading to abnormal RB forms that remain viable. Once the persistence-inducing stress is removed, the abnormal RBs begin to divide and re-enter the normal developmental cycle with the production of infectious EBs. It is generally assumed that persistence is a contributing factor in chronic chlamydial disease sequelae.

In adapting to obligate intracellular parasitism, Chlamydia has undergone extensive genome reduction (Stephens et al., 1998). These pathogens have eliminated genes from various functional pathways, mainly anabolic, as they rely on the host cell to provide nutrients and a stable environmental niche. Strikingly, Chlamydia has lost many genes encoding cell division (Fts) proteins that are usually highly conserved across the eubacterial lineage. More specifically, Chlamydia has lost FtsZ, a prokaryotic tubulin homologue that polymerizes around the future division site and acts as a scaffold for the recruitment of subsequent components of the cell division machinery (Bi and Lutkenhaus, 1991; Wang et al., 1997). Chlamydia has only three annotated cell division genes: ftsK, ftsI and ftsW (Stephens et al., 1998). How these bacteria divide is, therefore, an unanswered and fundamentally important question.

From the 1960s onwards, studies demonstrated that chlamydiae are sensitive to penicillin with the phenotypic effect being a block in cell division leading to enlarged, abnormal RBs (Bernkopf et al., 1962; Matsumoto and Manire, 1970). Later, using radiolabelled penicillin, the group of Caldwell (Barbour et al., 1982) showed that chlamydiae have three penicillin binding proteins (Pbp): two high-molecular-weight (HMW) Pbps and one low-molecular-weight (LMW) Pbp. The function of the HMW Pbps is to catalyse the polymerization of the PG subunits (Ghuysen and Goffin, 1999). The presence of Pbps and the sensitivity of chlamydiae to penicillin were surprising given that no peptidoglycan (PG) has yet been detected in the bacteria (Barbour et al., 1982). Moulder (1993) referred to this as the ‘Chlamydial Anomaly’. However, with the sequencing of the first chlamydial genome in 1998 (Stephens et al., 1998), a full PG biosynthetic pathway was revealed. This pathway is clearly important, and it is commonly assumed that chlamydiae synthesize PG strictly for cell division. Indeed, the group of Rockey (Brown and Rockey, 2000) isolated an antibody that recognized apparent division septa of chlamydiae, and the authors proposed that it was a non-proteinaceous antigen, i.e. peptidoglycan, that was being recognized.

Given that the key function of the cell division apparatus is arguably the synthesis of the division septum itself, a process requiring PG synthesis, we chose to analyse the role of the HMW Pbps in chlamydial cell division. In this regard, the cell division protein FtsI/Pbp3 catalyses the final steps of PG synthesis in the division septum (Pogliano et al., 1997). Escherichia coli temperature sensitive mutants of Fts proteins, including FtsI (Spratt, 1977a), become filamentous and are non-viable when cultured at non-permissive conditions. Chlamydia encodes a FtsI/Pbp3 homologue as well as another HMW Pbp, MrdA/Pbp2, that in E. coli is involved in side-wall PG synthesis (Tamaki et al., 1980), a so-called rod shape-determining protein. E. coli mutants of Pbp2 lose their rod shape and grow as cocci (Bendezú and de Boer, 2008). As Chlamydia is naturally coccoid, the function of Pbp2 in this organism is unclear. Moreover, Chlamydia has retained a number of rod shape-determining proteins in addition to Pbp2, specifically MreB, an actin-like protein that can also polymerize to form filaments (Jones et al., 2001). In rod-shape bacteria, MreB and other proteins of the Mre system recruit Pbp2 and direct its synthesis of cell wall (Divakaruni et al., 2005). The exact function of MreB in the coccoid bacterium Chlamydia is unknown. Here, we investigated the potential role of the HMW Pbps and MreB in chlamydial cell division. With the lack of a genetic system for generating targeted mutations, we chose a chemical genetic approach. Using Pbp- and MreB-specific inhibitors, we showed that both Pbp2 and FtsI/Pbp3 as well as MreB are required for normal chlamydial cell division. Our results thus suggest that Chlamydia has co-opted the function of the rod shape-specific PG synthesis machinery MreB/Pbp2 for cell division.


Bioinformatics analysis of the HMW Pbps of Chlamydia

Work from the laboratory of Caldwell demonstrated that Chlamydia possesses three Pbps (Barbour et al., 1982). The sequencing of the first chlamydial genome, C. trachomatis serovar D, identified two HMW Pbps annotated as Ct270 and Ct682 (Stephens et al., 1998). The genomic context of Ct270 strongly supports its designation as ftsI/pbp3. In E. coli, ftsI is downstream from mraW/rsmH and ftsL and upstream of murE, murF and mraY (Blattner et al., 1997) (Fig. 1A). In Chlamydia, Ct270 is downstream from Ct272/mraW and Ct271, a conserved Chlamydia-specific hypothetical protein, and upstream of Ct269/murE and Ct268/amiA, a predicted amidase (Fig. 1A). Ct756/murF and Ct757/mraY are associated with the mur operon on the opposite side of the chromosome. In E. coli, mrdA/pbp2 is upstream of its predicted lipid II flippase, mrdB/rodA (Blattner et al., 1997). Ct682 is a predicted monocistronic unit with Ct726/rodA located approximately 50 kb removed on the opposite strand.

Figure 1.

Bioinformatics analysis of the chlamydial HMW Pbps. A. Genomic context of the HMW Pbps in C. trachomatis and E. coli. Black arrows indicate genes within a predicted operon containing a HMW Pbp. Grey arrows indicate genes outside of such areas. Boxes indicate intergenic regions. B. Domain organization of the HMW Pbps spanning the indicated amino acids. In parentheses is the size of the indicated protein. TM, transmembrane; DD, dimerization domain; RR, repeat region; TP, transpeptidase. C. Alignment of chlamydial Pbp2 and Pbp3 versus the E. coli homologue with homology comparison below. The numbers in each field indicate: %identity/%similarity; number of homologous amino acids; significance. 100 = 100% identity. The asterisk ‘*’ indicates that removal of the repeat region from the analysis reduces the e value by 8 logs (i.e. is more homologous).

We further classified the structural domains of the HMW Pbps of Chlamydia and their E. coli homologues. Two transmembrane prediction programs, DAS and TMHMM (Krogh et al., 2001; Cservo et al., 2002), identified potential transmembrane domains within the first 50 amino acids for each Pbp. This is followed by a so-called Pbp dimerization domain (pfam03717) of approximately 200 amino acids and the C-terminal transpeptidase domain (pfam00905) (Fig. 1B). A bioinformatics analysis from Ghuysen and Goffin (1999) of the HMW Pbps of C. trachomatis suggested that both lack transglycosylase activity and are of the class B HMW Pbps. Chlamydial Ct682 is unique in containing an additional region between the dimerization domain and the transpeptidase domain that is common to all chlamydial species as assessed by BLAST. When blasted against non-chlamydial genomes, this region has no homologues. Comparing homologies between the HMW Pbps of Chlamydia and FtsI and Pbp2 of E. coli using the Align function of BLAST ( (Altschul et al., 1990), we observed that, indeed, Ct270 is most homologous to FtsI/Pbp3 and aligns well with it (Fig. 1C). However, there is little homology between Ct270 and E. coli FtsI/Pbp3 within the first 150 amino acids; rather, Ct270 displays homology to Pbp2 in its N-terminus. Interestingly, Ct682/Pbp2 is much larger than E. coli Pbp2 and contains duplicated segments within the additional region. When this region is excluded from the alignment, the e-value for homology between Ct682 and Pbp2/MrdA is reduced by 8 logs (i.e. is more homologous). We modelled each chlamydial HMW Pbp using Phyre2.0 (Kelley and Sternberg, 2009), and the structures can be compared in Fig. S1. Interestingly, Phyre predicts a large amount of structural disorder within the duplicated region of Pbp2, suggesting that additional factors may be required to stabilize the protein conformation (Dunker et al., 2008). We conclude that whereas Ct270 is clearly a FtsI homologue and Ct682 is clearly homologous to Pbp2, the changes outside of the enzymatic domain in each suggest that they may have additional or unique functionality within Chlamydia.

Chlamydiae are inhibited by Pbp-specific antibiotics

Although there has been recent progress in developing techniques for introducing DNA into Chlamydia (Wang et al., 2011), we do not currently have the tools to generate targeted mutations. Therefore, we employed a chemical genetic approach to assess the function of the HMW Pbps in this organism. Indeed, such a technique is preferred since it is unlikely that non-functional mutants in chlamydial cell division proteins would be isolated; in E. coli such mutants are typically conditional (e.g. temperature sensitive). We used two β-lactam penicillin-derivatives: mecillinam, which selectively blocks Pbp2 activity (Spratt, 1977b; Iwaya et al., 1978; Kitano and Tomasz, 1979; Tamaki et al., 1980; Canepari et al., 1984), and piperacillin, which selectively blocks Pbp3 activity (Kitano and Tomasz, 1979; Botta and Park, 1981; Pogliano et al., 1997; Eberhardt et al., 2003). The human epithelial cell line, HeLa, was infected with C. trachomatis serovar L2 and treated with varying doses of the different antibiotics. In untreated cells, robust chlamydial growth was observed as measured by inclusion forming units (IFUs) (Fig. 2A). Mature inclusions, with many RBs of approximately 1 µm lining the inclusion membrane, were easily visualized by immunofluorescence from duplicate samples (Fig. 2B). Chlamydiae were exquisitely sensitive to penicillin with growth abolished at doses as low as 0.05 µg ml−1, reflecting the likely synergistic action of targeting both the HMW and LMW Pbps. Mecillinam inhibited chlamydial growth at doses of 0.5 µg ml−1 whereas piperacillin was effective at 5 µg ml−1. Differences in the permeability of the host cell membrane, the inclusion membrane, and the bacterial outer membrane to each antibiotic, in the stability of the drugs, in the activity per mg of the compound, or in the expression level of the Pbps in Chlamydia may explain the differences in effective dose amongst these β-lactam antibiotics. Combination of sub-inhibitory doses of mecillinam (0.25 µg ml−1) and piperacillin (0.5 µg ml−1), which individually cause less than 10% inhibition, resulted in greater levels of inhibition (approximately 70%; data not shown).

Figure 2.

Effect of Pbp-specific penicillin derivatives on chlamydial growth and morphology. A. Quantification of recoverable inclusion forming units (IFUs) from C. trachomatis L2 infected HeLa cells treated with varying doses of antibiotics. Data are presented as an average of three experiments performed in triplicate with standard deviation. The black bar represents the number of IFUs recovered from an untreated control sample. Pen, penicillin (targets all Pbps); Mec, mecillinam (targets Pbp2); Pip, piperacillin (targets Pbp3). B. Immunofluorescent images of untreated (UTD) and antibiotic treated C. trachomatis L2 stained with antibodies directed against MOMP. Arrows indicate individual RBs of approximately 0.5–1 µm diameter lining the inclusion. ‘+’ indicates abnormal RB forms with a diameter of approximately 3–5 µm. Images were acquired at 630× magnification (scale bar = 10 µm).

Immunofluorescent microscopic analyses of the β-lactam-treated chlamydiae, using an antibody targeting the major outer membrane protein (MOMP), revealed that each drug induced a blockage in cell division as represented by the enlarged (5–10 µm), abnormal RB forms and reduced numbers of bacteria (3–5), in general, when compared with untreated chlamydial RBs (Fig. 2B). As others have shown (Skilton et al., 2009), removal of the β-lactam from the culture medium allows chlamydiae to recover normal growth characteristics (data not shown). Overall, these data indicate that both Pbp2 and Pbp3 are required for normal chlamydial development and that they likely function as cell division proteins.

The HMW Pbps interact with each other and with the cell division protein FtsK of Chlamydia in a bacterial two-hybrid assay

A key characteristic of cell division proteins is that they assemble in a multimolecular machinery at the bacterial septum (Adams and Errington, 2009). We, therefore, tested the ability of Pbp2 and Pbp3 to interact with each other and with the annotated chlamydial cell division proteins FtsW and FtsK as well as RodA, a FtsW homologue, using a bacterial two-hybrid (BACTH) system based on functional reconstitution of adenylate cyclase activity in E. coli (Karimova et al., 1998; 2005). This assay revealed the interaction between Pbp2 and Pbp3, the dimerization of Pbp3, and the association of both Pbp2 and Pbp3 with FtsK, a multifunctional cell division protein with chromosome segregation ATPase activity (Sherratt et al. 2010; Fig. 3). However, we did not detect interaction between the HMW Pbps and the putative lipid II transporters FtsW and RodA (data not shown). To show specificity of the interactions, we tested several unrelated transmembrane proteins from Chlamydia (such as GlnP, a putative glutamine transporter) and found no detectable interaction (Fig. 3). Our present findings with chlamydial Pbp2 suggest a potentially unique role for this protein in the cell division of this organism. In sum, these data show that the HMW Pbps may form a complex with FtsK in Chlamydia.

Figure 3.

BACTH analysis of interactions between chlamydial HMW Pbps and FtsK. E. coli DHT1 co-transformed with plasmids expressing the indicated full-length chlamydial proteins fused to either T25 or T18 were plated on selective minimal medium supplemented with X-gal and IPTG. Petri dishes were imaged under normal lighting conditions without magnification. Black colonies in the images are indicative of interactions between the expressed hybrid proteins. GlnP = predicted transmembrane protein from C. trachomatis used as a negative control. Neg = non-fused (i.e. empty) T18 vectors versus the indicated T25 fusions.

Chlamydial HMW Pbps preferentially localize to distinct puncta on chlamydiae

To determine the subcellular localization of the chlamydial HMW Pbps, we generated rabbit polyclonal antibodies directed against synthetic peptides derived from the transpeptidase domain of each Pbp. Each antibody preparation was affinity purified against a column to which the peptides had been covalently bound, and we verified their specificity by Western blotting extracts of E. coli expressing or not the cognate transpeptidase domain of each Pbp. As shown in Fig. 4A, the affinity purified antibodies specifically reacted with the cognate transpeptidase domains, which accumulated in the insoluble fraction when expressed in E. coli. Yet, these antibodies failed to detect the full-length Pbp proteins within C. trachomatis-infected cell lysates at either 24 or 48 h post infection (p.i.) – a time when dividing RBs or non-dividing EBs, respectively, should be present – likely because of low endogenous expression levels (as is the case for many cell division proteins in other bacteria).

Figure 4.

Anti-peptide antibody reactivity against the chlamydial HMW Pbps. A. Extracts of E. coli cells expressing the chlamydial HMW Pbp transpeptidase domains or lysates of HeLa cells infected or not with C. trachomatis were analysed by Western blot with antibodies raised against Pbp2 or Pbp3. The expected size of the full-length Pbp2 and Pbp3 from the infected cell lysates is 124 and 73 kDa, respectively, whereas the size of the transpeptidase domains expressed in E. coli is approximately 45 kDa. UI = uninduced, I = induced, Sol = soluble, Insol = insoluble: all refer to E. coli protein fractions. Uninf = uninfected HeLa cell lysate 24 h and 48 h indicate C. trachomatis-infected HeLa cell lysates collected at this time. B. Representative immunofluorescent images of C. trachomatis L2 infected HeLa cells either treated with piperacillin or left untreated and stained with antibodies against Pbp2 or Pbp3 (green) and the major outer membrane protein (MOMP-red). A schematic representation of the membrane structure for the piperacillin-treated culture is presented to the right of the merged colour image. The reader should note that the MOMP staining fails to uniformly label the membrane during antibiotic treatment. Pbp staining is indicated by arrows in the piperacillin treated cultures only for clarity. ‘+’ indicates abnormal RB forms of approximately 3–5 µm in diameter. Images were acquired at 630× magnification (scale bar = 5 µm). UTD = untreated; Pip = piperacillin. C. A schematic representation of the staining of Pbps in UTD and antibiotic treated cultures. In UTD cultures, normal RB forms (represented by small red circles) line the inclusion membrane (represented by a dashed circle-not visible in the images), and individual bacteria stain sporadically with the antibody (represented by ‘*’) reflecting that not all RBs divide synchronously. In antibiotic treated cultures, abnormal RBs (represented by larger red ovals) have multiple foci of Pbps distributed around the periphery of the individual organism.

We next performed an immunofluorescent analysis on C. trachomatis-infected cells, treated or not with the Pbp-specific antibiotics (Fig. 4B). A mouse monoclonal antibody against the MOMP was used to identify individual bacteria. In untreated cells, both Pbp antibodies gave a similar pattern of punctate staining that was present at no more than one copy per bacterial cell (Fig. 4B). Noticeably, that the Pbp punctate staining was not uniformly detected in all bacteria likely reflects the fact that not every bacterium is dividing simultaneously and that these HMW Pbps may only transiently assemble at the cell division site as observed in other bacterial species. The immunofluorescence detection of the Pbps was highly specific as (i) antibody labelling was abolished by pre-incubation with the cognate peptide-conjugated beads used for affinity purification and (ii) staining was specific to inclusions as the host cell did not exhibit any traces of labelling, consistent with the Western blot data (Fig. S2). Given the small size of chlamydiae when growing normally as well as their compaction within a defined space (i.e. the inclusion), the subcellular localization is difficult to determine for low abundance proteins, particularly for inner membrane proteins, and has not been previously addressed in Chlamydia. To overcome this limitation, we took advantage of the effects of the β-lactam antibiotics on chlamydiae vis-à-vis the enlarged, abnormal RB forms. As shown in Fig. 4B (and Fig. S3), the MOMP staining pattern becomes more irregular during β-lactam treatment and fails to uniformly label the bacterial membrane. Nevertheless, we were able to detect multiple puncta of Pbp staining at the periphery of each abnormal RB structure, in some cases colocalizing with MOMP (Figs 4B and S3). This peripheral labelling suggests that the Pbps are indeed membrane-associated proteins as this staining pattern is distinct from the homogenous labelling seen for cytoplasmic proteins such as Hsp60 (Fig. S4) or EUO (Ouellette et al., 2006). Moreover, if one assumes that the Pbp proteins do demarcate the cell division site, then the fact that Pbp proteins form distinct spots at the periphery of β-lactam-inhibited chlamydiae may indicate that multiple aborted cell division planes are generated during treatment. The immunofluorescent data is represented schematically in Fig. 4C, and we have included a sketch of the membrane structures next to the merged colour channel in Fig. 4B to aid the reader in identifying these structures. We conclude that, as predicted (Fig. 1), the chlamydial HMW Pbps are membrane-associated proteins.

Pbp-specific inhibition of Chlamydia results in distinct morphological changes

In E. coli, inhibition of Pbp3/FtsI induces filamentation (Pogliano et al., 1997) whereas inhibition of Pbp2 or other rod shape-determining proteins induces cell rounding (Bendezú and de Boer, 2008). Our immunofluorescence analysis clearly indicated that treatment with any of the β-lactam antibiotics induced abnormal morphology with multiple puncta of Pbp staining associated with each abnormal bacterium. To determine if these drugs induced specific alterations in chlamydial morphology, infected cell samples treated with each antibiotic were examined by electron microscopy. We prepared samples where chlamydiae were treated for 3 h only, beginning at 14 h p.i., or from the time of infection until fixation at 17 h p.i., to discriminate between short- and long-term effects of antibiotic exposure (Fig. 5).

Figure 5.

Electron microscopic analysis of C. trachomatis serovar L2 treated with Pbp-specific penicillin derivatives. HeLa cells were infected with chlamydiae and treated with antibiotics at the indicated times. All samples were fixed and processed at 17 h p.i. A representative RB is pictured in (A). In the top panel, infected cells were treated for 3 h only before fixation (B and D). In the bottom panel, infected cells were treated from the time of infection (E–G). The total number of inclusions and bacteria were quantified as well as the three abnormal phenotypes we observed as illustrated by the arrowheads: membrane blebs (B and E), fragmentation (C and F), and elongation (D and G). The number of individual bacteria displaying the abnormal phenotypes is indicated as well as the percentage of the population this represents. Some bacteria displaying more than one phenotype were counted in both fields. The width of each image represents 2 µm aside from (E) where the scale bar is 2 µm. A chi-squared test for independence reveals that all antibiotic-treated cultures are significantly different (P < 0.005) from untreated cultures, regardless of time point. Mec-treated samples are significantly different (P < 0.0001) from Pip or Pen whereas Pip and Pen are not significantly different (P < 0.25).

In untreated cells, normal chlamydial forms (Figs 5A and S5) were detected lining the periphery of the inclusion membrane, and various stages of the cell cycle could be visualized. We quantified three non-mutually exclusive phenotypes in the bacterial cells: (i) membrane blebbing, where a discrete gap could be seen between the inner and outer membranes (Fig. 5B and E); (ii) fragmentation, where multiple cell bodies were evident within a single outer membrane (Fig. 5C and F); and (iii) elongation, where the bacterial cell appeared more rod-shaped than coccoid (Fig. 5D and G). A low percentage of the indicated phenotypes was seen in untreated cells (Fig. 5) and likely represents background levels of abnormally developing bacteria.

After 3 h of treatment, all antibiotic-treated samples showed reduced cell division as represented by the number of bacteria per inclusion (untreated 14 h: 6.9; untreated 17 h: 12.9; penicillin: 6.1; mecillinam: 8.9; piperacillin: 5.8) (Fig. S5). Treatment with any of the antibiotics led to a large increase in the proportion of bacteria containing membrane blebs (Fig. 5). Each also showed a small percentage of fragmented bacteria. More interestingly, mecillinam-treated bacteria showed no elongation phenotype whereas penicillin and piperacillin more readily induced this phenotype. Moreover, the numbers of elongated forms in penicillin- or piperacillin-treated cultures may be under-represented due to the spherical shape constraint imposed on the inclusion by the turgor pressure of the host cell, which may prevent the bacteria from adopting a truly elongated phenotype. Indeed, many crescent-shaped bacteria are visible during these treatments as seen in Fig. 5E (see also Skilton et al., 2009).

In cells treated for 17 h (i.e. from the time of infection), chlamydial cell division is blocked regardless of the antibiotic used (average number of bacteria per inclusion is less than 2), and the bacteria are enlarged (Fig. S5). Strikingly, treatment with mecillinam from the time of infection resulted in massive fragmentation but no elongation and was statistically distinguishable from all other treatments (P < 0.0005). Piperacillin again showed an elongated phenotype and, as with the shorter treatment time, many bacteria contained membrane blebs, and this was more pronounced in the piperacillin-treated sample. Penicillin gave a mixed phenotypic profile but was not significantly different from piperacillin-treated cultures (P < 0.25). These data indicate that inhibition of Pbp2 or Pbp3 results in distinct phenotypes with the function of both being required for proper bacterial cell division. Furthermore, it supports the conclusion that mecillinam and piperacillin target discrete functions within Chlamydia.

MreB acts upstream of HMW Pbp function and interacts with FtsK

Chlamydiae contain another rod shape-determining protein, MreB, in addition to Pbp2 and RodA. To determine if this protein is important for normal chlamydial development, we treated C. trachomatis-infected cells with varying doses of the MreB-specific antibiotic A22 (Bean et al., 2009). As MreB is an actin homologue, we first tested whether the drug-affected actin structures in the host cell but failed to detect any demonstrable changes (Fig. S6) and nor did it appear toxic at the doses we tested. A22 efficiently inhibited the production of infectious particles in a dose-dependent manner (Fig. 6A) at concentrations that also effectively inhibit other MreB-containing bacteria (Takacs et al., 2010), indicating that MreB function is essential for chlamydial growth and development. Importantly, inhibition of MreB with A22 does not lead to death of the bacteria as removing the antibiotic from the medium allows the bacteria to recover (S.P. Ouellette, unpubl. obs.).

Figure 6.

Effect of MreB-inhibition on Chlamydia growth and division. A. Chlamydiae are sensitive to A22. Quantification of recoverable inclusion forming units (IFUs) from C. trachomatis L2 infected HeLa cells treated with varying doses of the MreB-inhibiting drug A22. B. MP265 and A22 both inhibit MreB. Quantification of IFUs from C. trachomatis L2 infected HeLa cells treated with sub-inhibitory doses of MP265 and/or A22. (C and D) MreB functions upstream of the HMW Pbps. Cells were infected with C. trachomatis serovar L2 and treated with the indicated antibiotics, individually (A22 at 75 µM; Mec and Pip at 5 µg ml−1) or in combination, at 6 h p.i. for 18 h. At 24 h p.i., cells were fixed and processed for electron microscopy. C. Representative images of chlamydiae treated with the indicated antibiotics are shown (scale bar = 2 µm). D. Quantification of inclusion diameter (± standard deviation), measured at the widest point, in µm during different treatments is presented below the electron micrographs. The number of inclusions analysed for each condition: 8 for UTD (untreated); 12 for A22; 10 for Pip (piperacillin); 16 for A22 + Pip; 21 for Mec (mecillinam); 11 for A22 + Mec. E. Chlamydial MreB interacts with FtsK. E. coli DHT1 co-transformed with plasmids expressing the indicated chlamydial proteins fused to either T25 or T18 were plated on selective medium supplemented with X-gal. Petri dishes were imaged under fluorescent lighting without magnification. FtsK indicates the full-length construct whereas FtsKN indicates the N-terminal 200 amino acids of the protein. Black colonies in the images are indicative of interactions between the expressed hybrid proteins. A test of MreB versus Pbp2 is shown as a negative result. The MreB versus MreB images are duplicated for presentation purposes.

We also tested the ability of MP265, another MreB-specific antibiotic, to block chlamydial growth. MP265 is considered more specific for MreB (Takacs et al., 2010), and indeed, we determined that, like A22, MP265 also blocks chlamydial growth (Fig. 6B). Interestingly, chlamydial growth was also efficiently blocked by combined treatment with lower, sub-inhibitory doses of either A22 or MP265 that alone were incapable of inhibition (Fig. 6B). We conclude that these drugs are specifically targeting and disrupting the function of MreB and that this results in an abrogation of chlamydial development.

Electron micrographic analysis revealed that A22 treatment resulted in the individual bacteria within the inclusion being enlarged (1–2 µm) compared with normal RBs (0.5–1 µm) and without the grossly abnormal bacterial phenotypes seen in β-lactam treated cultures (i.e. no fragmentation or elongation). Yet, most inclusions were reduced in size and contained only one or two bacteria (Fig. 6C). This suggests that MreB inhibition traps the bacteria in a pre-division state although we cannot exclude an additional or alternative function for MreB in general chlamydial growth.

In rod-shaped bacteria, MreB is a key element in nucleating and directing the PG machinery involved in cell elongation (i.e. Pbp2; Divakaruni et al., 2005). To probe the functional consequences of disrupting MreB in Chlamydia, we designed an experiment using co-treatment with both A22 and the Pbp-specific antibiotics. Such a chemical genetic approach would allow us to assign an epistatic relationship between these components. We hypothesized that if MreB were acting upstream of Pbp2 or Pbp3 in the cell division process, then treatment with both A22 and either mecillinam or piperacillin should result in the A22 phenotype. Indeed, when infected cells were treated with both A22 and a Pbp-specific β-lactam, chlamydiae displayed an enlarged coccoid morphology typical of the A22 phenotype (Fig. 6C). As a means of quantifying this effect, we measured the mean diameter of the inclusion (at the widest point) from different treatments. As shown in Fig. 6D, any sample treated with A22 resulted in small inclusions (with a diameter of roughly 3 µm) whereas untreated cultures or those treated with a Pbp-specific β-lactam showed inclusions with a diameter of approximately 10 µm. This effect is distinct from cessation of growth, for example as induced by chloramphenicol treatment where inclusions fail to develop and are mis-localized within the cell (Scidmore et al., 1996). Therefore, we conclude that MreB function is upstream of Pbp function and that both are essential for proper cell division.

We next determined if MreB could directly associate with the known cell division components of Chlamydia. Using the BACTH assay, we revealed an interaction between MreB and full-length FtsK but were unable to detect interactions between the HMW Pbps and MreB (Fig. 6E). We further showed that the N-terminal region of FtsK (FtsKN), spanning the first 200 amino acids of the protein and encoding the transmembrane segments of FtsK, could also interact with MreB (Fig. 6E). In sum, our results suggest that chlamydial MreB may play a critical role both in modulating Pbp function and in recruiting FtsK, and perhaps other as-yet unidentified cell division components, to the division site.


Chlamydia is among the rare bacteria that lack the critical cell division protein FtsZ, a tubulin homologue that polymerizes around the division plane and effects the recruitment of cell division components (Bi and Lutkenhaus, 1991; Wang et al., 1997). Thus, how Chlamydia divides in its absence and lacking other known mechanisms, for example the CDV system of archaebacteria (Lindås et al., 2008), is a fascinating and unresolved question. The only annotated protein that may serve a similar function to FtsZ in Chlamydia is MreB, an actin-like rod shape-determining protein that polymerizes and forms filaments. Recent work has suggested that, as opposed to forming helical structures under the cell wall (Jones et al., 2001), MreB may form circumferential, ring-like structures in association with the cell wall biosynthetic machinery (Domínguez-Escobar et al., 2011; Garner et al., 2011). We were intrigued by the presence of rod shape-determining genes in Chlamydia for two reasons: first, Chlamydia is coccoid, and second, the presence of these genes indicates an important function given the extensive genome reduction by which Chlamydia has adapted to its intracellular niche. In addition to mreB, Chlamydia encodes pbp2 and rodA homologues. Pbp2 is a HMW Pbp that functions in peptidoglycan (PG) synthesis. Given that the critical function for the cell division apparatus is to construct a cell wall (division septum) at the division site and that PG synthesis is important in this process, we initiated an investigation into the role that the HMW Pbps play in chlamydial cell division. We provide evidence indicating that both Pbp2 and Pbp3/FtsI are critical for chlamydial cell division, which results in distinct morphological changes when inhibited by Pbp-specific antibiotics. Further, these Pbps were shown to interact with cell division protein FtsK from Chlamydia and localize in discrete puncta within the membrane of the bacterium, further strengthening a role for these proteins in division. Finally, we provide evidence that MreB acts upstream of Pbp function and can itself interact with the cell division protein FtsK. We propose MreB may thus substitute for the role of FtsZ in Chlamydia in assembling the division apparatus.

We propose a model for the function of the chlamydial HMW Pbps and MreB in chlamydial cell division (Fig. 7), in which MreB/Pbp2 are necessary for identifying the division site and for subsequently generating a PG primer from which Pbp3 then extends into the division plane. In the absence of MreB function, no cell division occurs, and the bacteria are trapped in what appears to be a pre-division state. Inhibition of Pbp2 function with mecillinam results in massive fragmentation likely owing to the apparent dysregulated generation of division planes. This suggests that Pbp3 activity under these conditions is still capable of making division planes albeit in an unregulated manner. Conversely, inhibition of Pbp3 with piperacillin did not result in massive fragmentation as seen with mecillinam treatment but rather in a much larger proportion of bacteria exhibiting membrane blebs or elongated shape. This suggests that Pbp2 activity alone under these conditions is not capable of creating a division plane. Given the interaction between Pbp2 and Pbp3 (Fig. 3), we hypothesize Pbp2 may play a role in bringing Pbp3 to the correct division site either directly or through a link with FtsK or in stabilizing it once there. This would support the observation that the immunofluorescence labelling with each Pbp antibody gave similar patterns (Fig. 4). An antibody from a different species will be needed to test for colocalization as our antibodies were both prepared in rabbits. In the absence of Pbp2 function, multiple Pbp3 foci localize randomly on the membrane with apparent generation of division planes leading to fragmentation of the cell body. This implies a function for Pbp2 in coupling division site recognition with the divisome. In the absence of Pbp3 function, division planes are rarely visible, and the bacteria become elongated. Multiple Pbp2 foci are localized along the membrane suggesting that Pbp2 may continue to demarcate potential aborted cell division sites. However, we cannot exclude a potential role for all of these proteins in functions beyond cell division.

Figure 7.

Model for the role of MreB and the HMW Pbps in chlamydial cell division. A. Schematic organization of the core components of the chlamydial cell division machinery in the inner membrane (IM). Lacking FtsZ, Chlamydia uses MreB to demarcate the division plane. MreB recruits FtsK, which in turn may recruit Pbp2, Pbp3, and likely other unidentified proteins. B. Proposed pathway for chlamydial cell division. During normal growth, MreB designates the division plane and recruits downstream components. A22 blocks MreB ring formation and traps bacteria in a pre-division state at this early step. Mecillinam (Mec) and piperacillin (Pip) block cell division by inhibiting Pbp2 and Pbp3, respectively, leading to distinct phenotypes. For further details, see accompanying text. K, B, 2 and 3 are colour-coded to represent the proteins in (A). The elliptical rings around B represent the division septum whereas the others represent membranes.

In a previous study, the group of Rockey described an antibody that was generated against an adjuvant containing mycobacterial cell wall skeletons and that recognized a septum-localized antigen in Chlamydia (Brown and Rockey, 2000). The authors presented evidence that the antigen common to both Mycobacterium and Chlamydia and recognized by this antibody was not proteinaceous, raising the possibility that their antibody recognized PG. The absence of detectable PG in chlamydiae suggests that cell wall growth is not mediated by this component, and that PG synthesis is used strictly for cell division purposes. If this were accurate, then the most likely candidate to recruit the PG-synthesizing machinery would be, by analogy to other bacteria containing rod shape-determining proteins, MreB. In rod-shaped bacteria, MreB appears as a central co-ordinator of a large multi-protein PG synthesizing enzymatic complex that collaborates to co-ordinate cell elongation (Divakaruni et al., 2005). This complex includes the PG-synthesizing enzyme Pbp2, RodA, and many additional partners such as MreC, MreD, RodZ, etc, that are involved in the stabilization and/or regulation of this machinery. Similar or equivalent components may be present in Chlamydia. Indeed, bacterial two-hybrid assays did not reveal any direct interaction between the HMW Pbps and either MreB, RodA or FtsW, suggesting that some proteins should link them. Rather, we were able to demonstrate an interaction between MreB and FtsK, which in turn may recruit the HMW Pbps.

In summary, we provide here the first systematic analysis of chlamydial cell division. From our results, we propose that, during reductive evolution, Chlamydia has co-opted the rod shape-determining proteins Pbp2 and MreB for cell division and that MreB may substitute for FtsZ as the central co-ordinator of the cell division machinery. This study is the first step in unravelling one of the most intriguing microbiological conundrums: how Chlamydia divides.

Experimental procedures

Organisms and cell culture

Chlamydia trachomatis serovar L2 EBs were harvested from infected HeLa cell cultures at 37°C with 5% CO2, were purified by discontinuous density gradient centrifugation in Renografin (Bracco Diagnostics, Princeton, NJ), and were titred for infectivity by determining IFUs on fresh cell monolayers. HeLa cells were routinely cultivated at 37°C with 5% CO2 in DMEM containing glutamax, glucose and pyruvate (Invitrogen, Carlsbad, CA) and further supplemented with 10% FBS. The adenyate cyclase mutant strain (Δcya) DHT1 (Dautin et al., 2000) and XL1 strain (Stratagene, Santa Clara, CA) of E. coli were cultured in LB broth, and chemically competent cells were prepared following standard protocols. All antibiotics and chemicals used in the described experiments were purchased from Sigma (St. Louis, MO) unless otherwise indicated. IPTG, X-gal, and all restriction and other DNA-modifying enzymes were obtained from Fermentas (Thermo Fisher, Rockford, IL) whereas the components for the Gateway system were obtained from Invitrogen.

Infection of cells and quantification of inclusion forming units

HeLa cells were plated in six-well culture plates at a density of 1 × 106 cells per well. In a subset of wells, cells were plated onto glass coverslips for immunofluorescence microscopy. Approximately 18 h later, confluent cell monolayers were rinsed with PBS (Invitrogen), 106 IFUs of C. trachomatis (L2) were added directly to each well, and infected cells were incubated at 37°C with 5% CO2 in the presence or absence of the indicated concentrations of antibiotics. At 24 h p.i., medium was aspirated from the infected cells, the cells were rinsed with PBS, and 1 ml of SPG (sucrose/phosphate/glutamate) buffer was added to each well. Cells were scraped and collected from each well into a 1.5 ml microfuge tube with 3 glass beads. Samples were vortexed for 45 s and frozen at −80°C. Samples were titred for infectivity on fresh cell layers to quantify the number of IFU per well (Ouellette et al., 2006).


A list of primers and plasmids used in these studies can be viewed in Table S1. PCR was performed using DyNAzyme EXT DNA polymerase (Thermo Fisher) against purified genomic DNA from C. trachomatis serovar L2 following the manufacturer's guidelines. PCR products were purified using the PCR purification kit (Qiagen, Valencia, CA) and digested with the indicated restriction enzymes. Empty vector pKT25 and pUT18C plasmids from the BACTH system were similarly digested and treated with alkaline phosphatase before ligation (with T4 DNA ligase) to digested PCR products. Chemically competent XL1 cells were transformed and plated on selective antibiotics in the presence of 0.4% glucose to repress expression. Plasmids were prepared from colonies containing the correct construct. For Gateway constructs, full-length open reading frames for the genes of interest cloned into pDONR221 were obtained from the Pathogen Functional Genomic Resource Center ( The ORF was transferred to the Gateway-derivatized BACTH plasmids, pST25-DEST or pUT18C-DEST (S.P. Ouellette, E. Gauliard, Z. Antosova and D. Ladant, in preparation), using the LR reaction (Invitrogen) following the manufacturer's guidelines. All E. coli expression constructs are described in Karimova et al. (2005).

BACTH assays

Bacterial adenylate cyclase two-hybrid (BACTH) interactions were performed as previously described (Karimova et al., 2005) using the adenylate cyclase mutant (Δcya) strain of E. coli DHT1. Briefly, chemically competent DHT1 were co-transformed with each plasmid to be tested and plated on M63 minimal medium agar containing selective antibiotics, 40 µg ml−1 X-gal, 0.5 mM IPTG and 0.2% maltose (Karimova et al., 1998). Plates were incubated at room temperature for up to 15 days for interactions. Only cultures in which the adenylate cyclase activity is regenerated are able to support growth on minimal medium with maltose as the sole sugar source.

Preparation and affinity purification of polyclonal antibodies against the HMW Pbps

Anti-peptide polyclonal antibodies were prepared in rabbits by Genosphere Biotechnologies (Paris, France). An equimolar mixture of two peptides conjugated to KLH was used for vaccination. For Pbp2 the following peptides were used: CARVGLDRERGRMK and CIEKWEEIRKKSFS. For Pbp3 the following peptides were used: CIRAATKGYSSAGKT and CVSQLKLLYEEWNRK. Equimolar amounts of each peptide from each Pbp were mixed and conjugated to the N-hydroxysuccinimide ester agarose gel bead support Affigel-10 (Bio-Rad, Hercules, CA) according to the manufacturer's protocol. The pre-clarified rabbit serum was passed over the beads in a column format and washed extensively with PBS. Bound peptide-specific antibodies were eluted with 0.1 M glycine/0.15 M NaCl at pH 2.8 into 1 ml fractions containing 200 µl 1 M Tris pH 7.5 to neutralize the acidity. Purified antibodies from fractions containing elevated A280 readings were subsequently dialysed against PBS and concentrated. Antibody specificity was verified by Western blot (below) and by pre-adsorbing the antibodies with the peptide-conjugated beads before immunofluorescent detection (see Fig. S2). We also conjugated each Pbp2 peptide individually to the iodoacetyl-linked agarose support SulfoLink Coupling Resin (Thermo Fisher) to ensure the peptide was linked at its N-terminus as opposed to any free amine (e.g. lysines) as can occur with the Affigel-10 support. Antibody was purified following the same protocol. No differences were seen in the reactivity of the Pbp2 antibodies purified from either support, and peptide-conjugated beads blocked immunofluorescent labelling of peptide-specific antibodies regardless of the support from which they had been purified (Fig. S2).

Pbp transpeptidase expression and Western blot analysis

The transpeptidase domain from each HMW Pbp of C. trachomatis serovar L2 was PCR amplified and cloned into the heat-inducible expression plasmid, pBR-5 (Sotomayor Pérez et al., 2010). Bacterial cultures transformed with the expression plasmids were induced to produce protein by shifting the temperature of the culture to 42°C for 1.5 h. Uninduced duplicate cultures were maintained at 30°C. Bacteria were centrifuged and the pellet resuspended in B-PER lysis buffer following the manufacturer's guidelines (Thermo Fisher) to separate soluble and insoluble fractions. Samples were diluted in LDS sample buffer (Invitrogen) and incubated at 95°C for 5 min. Equal volumes were separated on a 4–12% Bis-Tris PAGE gel in MES buffer (Invitrogen) and transferred to nitrocellulose following standard techniques. The membrane was blocked in 5% milk with 0.05% Tween-20 for 1 h and incubated with the indicated anti-Pbp Ab overnight in the same buffer at room temperature. The blot was washed extensively with 0.05% Tween-20 in PBS before addition of the secondary donkey anti-rabbit HRP Ab (GE Healthcare, Piscataway, NJ) for 1.5 h. After again washing extensively, the blot was reacted with the ECL Detection System (GE Healthcare), exposed to film, and developed.


At corresponding times to the collection of IFU and EM samples, duplicate wells with cells plated on coverslips were fixed in methanol for 5 min at room temperature. A primary mouse monoclonal antibody (Ab) against the MOMP of C. trachomatis L2 was used followed by detection with a fluorescently labelled secondary goat anti-mouse Ab (Invitrogen). For visualization of the HMW Pbps, cells plated on coverslips were fixed in 4% paraformaldehyde diluted in PBS for 30 min at room temperature. The fixed cells were subsequently permeabilized in 0.1% Triton X-100 for 5 min at room temperature. Antibodies were added for 1 h at 30°C followed by significant washing in the following order: (1) rabbit polyclonal Ab against the HMW Pbp, (2) goat anti-rabbit Alexa488 Ab (Invitrogen), (3) mouse monoclonal Ab against MOMP, (4) goat anti-mouse Alexa594 Ab (Invitrogen). Images were acquired on an Axio observer Z1 microscope equipped with an ApoTome module (Zeiss, Germany) and a 63× Apochromat lens and using a Coolsnap HQ camera (Photometrics, Tucson, AZ) with the installed Axiovision software.

Electron microscopy

For one set of samples, cells were infected as above and treated with 5 µg ml−1 of indicated antibiotic from the time of infection. In another set of samples, 5 µg ml−1 of antibiotic was added at 14 h p.i. All samples were processed at 17 h p.i. where cells from infected cultures were collected and centrifuged for 5 min at 250 g. For co-treatment with A22 and Pbp-specific antibiotics, cells were infected and treated at 6 h p.i. to facilitate the detection of inclusions at 24 h p.i. when the samples were fixed. Cell pellets were resuspended in 2.5% EM-grade glutaraldehyde diluted in 0.1 M cacodylate buffer and centrifuged at 250 g for 5 min. Cells were subsequently processed for EM as described elsewhere (Ouellette et al., 2006) by the Imagopole service (PFMU) of the Institut Pasteur. Briefly, cells were washed three times with PBS then fixed for 1 h at room temperature in 1% osmium tetroxide diluted in PBS. Samples were dehydrated with a graded series of ethanol and embedded in Spurr's resin (Electron Microscopy Sciences, Ft. Washington, PA). Then 70–80 nm sections were cut, stained with uranyl acetate and lead citrate, and viewed on a JEOL 1200EX transmission electron microscope. The width of inclusions in Fig. 6 was measured with ImageJ software.


The authors would like to thank Dr Isabelle Bonne for technical assistance with EM sample preparation and Dr Agnes Ullmann for critical review of the manuscript. S.P.O. was supported by a fellowship from the Pasteur Foundation. The authors have no conflicts of interest to declare.