Essential Role for FtsL in Activation of Septal Peptidoglycan Synthesis

A critical step in bacterial cytokinesis is the activation of septal peptidoglycan synthesis at the Z ring. Although FtsN is the trigger and acts through FtsQLB and FtsA to activate FtsWI the mechanism is unclear.

targets the FtsQLB complex to the Z ring in an FtsK-dependent fashion, and the cytoplasmic domain of FtsL is required to recruit FtsW (3) (Fig. 1). FtsL and FtsB form a multimer with interactions occurring between their alpha-helical transmembrane domains as well as their putative periplasmic coiled-coil domains (14)(15)(16)(17). They also interact with FtsQ through their C-terminal domains that lie beyond the coiled-coil domains forming a 1:1:1 complex which may dimerize (13,15,18). The structure of a peptide corresponding to the C-terminal region of FtsB bound to the periplasmic domain of FtsQ was recently determined (19,20).
Activation of FtsWI by FtsN requires two domains of FtsN; the cyto FtsN domain acts on FtsA, and the E FtsN domain, a short putative helical segment in the periplasm, likely acts on FtsQLB (10,21,22,36) (Fig. 1). In a proposed model, FtsN switches both FtsA and FtsQLB to an ON state which activates FtsWI (10,11). This regulatory model is based in part upon the isolation of "activation (superfission)" mutations (requiring less FtsN) in ftsL and ftsB which identified a short periplasmic region in both proteins, designated CCD for constriction control domain (10). The CCD connects the coiled-coil domain of each protein to its distal C-terminal region, which binds to FtsQ (13,18). It is not clear how these mutations work, but it is likely they mimic FtsN action, resulting in a change in conformation of the FtsQLB complex to the ON state that activates FtsWI (Fig. 1). Activation mutations have also been isolated in ftsA and ftsW (10,12). Such mutations in ftsA could cause it to act on FtsQLB or FtsW, whereas such mutations in ftsW could lead to an enzymatically active conformation. To address the mechanism of FtsWI activation, we set out to isolate dominant negative mutations in ftsL and ftsB. Such mutations should yield an FtsQLB complex that no longer activates FtsWI and yield information about the activation mechanism. By exploring the effect of the dominant negative mutations, as well as the activation mutations, on the recruitment and activation of FtsWI, we find an essential role for FtsL in the activation of FtsWI.

RESULTS
Isolation of dominant negative mutations in ftsL but not ftsB. To isolate dominant negative mutations in ftsL and ftsB, they were subjected to random mutagenesis, cloned into a plasmid downstream of an IPTG-inducible promoter, and introduced into a wild-type strain. Colonies were then picked, and dominant negative mutants were identified by screening for growth inhibition after streaking on plates containing increasing amounts of IPTG (isopropyl-b-D-thiogalactopyranoside). Three strong dominant negative mutations were obtained in ftsL (ftsL E87K , ftsL L86F , and ftsL A90E ) as well as two weak mutations (ftsL R61C and ftsL L24K ), but none were obtained in ftsB ( Fig. 2A and Table 1). Changing ftsL R61C to ftsL R61E resulted in a stronger dominant negative mutant (Table 1), while ftsL L24K is discussed later. Induction of the ftsL alleles in liquid culture resulted in filamentation ( Fig. 2B and Table 1). Complementation tests confirmed they were loss of function mutations, as they were unable to complement a DftsL strain (Fig. S1A, Table 1). Interestingly, three of these mutations overlapped the CCD domain, which was previously defined by activation mutations that decrease the dependency upon FtsN (10,11) (Fig. 2C). Using site-directed mutagenesis, we altered additional residues around the CCD and isolated three additional dominant negative mutations (ftsL R82E , ftsL N83K , and ftsL L84K ) ( Fig. 2C and Table 1). However, extending the mutagenesis to flanking regions as well as the C-terminal region of ftsL did not yield any additional dominant negative mutations ( Fig. 2C and Table 1). Although the residues we identified overlap the CCD, they are distinct from the residues involved in activation and lie mostly on the opposite side of a putative alpha helix. Since these mutations lead to a dominant negative effect, they behave as though they are nonresponsive to FtsN, just the opposite of activation mutations (Fig. 2D). We designate the region identified by the dominant negative residues as AWI (activation of FtsWI) based on the results described below.
Of residues composing the CCD domain of FtsL, residue E88 is the most conserved, and mutational analysis indicated that loss of the negative charge results in the activation phenotype (10). The neighboring residue E87 is even more conserved (Fig. S1B) and was altered in one of our dominant negative mutants. Additional analysis indicates that changing this residue to amino acids other than aspartate produces a dominant negative phenotype (Fig. S1C). Thus, the loss of the negative charge in two neighboring glutamate residues yields contrasting phenotypes. Since loss of the negative ftsL was subjected to random PCR mutagenesis, cloned downstream of the tac promoter in an expression vector containing an IPTG-inducible promoter (pJF118EH), and transformed into JS238. Transformants were screened for sensitivity to IPTG. ftsL WT and ftsL E88K (an activation allele) were included as controls and are not toxic. Several strong dominant negative mutations (ftsL E87K , ftsL L86F , and ftsL A90E ) and two weak mutations (ftsL R61C and ftsL E24K ) were obtained in this way (Table 1). Additional mutations were obtained by site-directed mutagenesis. (B) Dominant negative mutants inhibit division. Phase contrast micrographs of JS238 expressing ftsL or ftsL E87K (derivatives of pKTP100 [P tac ::ftsL]) grown in liquid culture and induced with 50 mM IPTG for 2 h. Induction of the other alleles also inhibited division (Table 1). (C) FtsL, residues 54 to 99, was modeled (for illustration purposes) as an alpha helix since it is thought to form a continuous alpha helix with the TM, and this region is also thought to form a coiled coil with FtsB. Altering the residues in green leads to activation mutations, whereas altering those residues in red results in dominant negative mutations. Altering the residues in yellow had no effect. Note that the activation mutations affect residues that lie mostly on one side of the helix, whereas the dominant negative mutations affect residues that lie mostly on the other side. The red residues (including L86 and E87) identify a region designated AWI (activation of FtsWI). The positions of residues 24 and 28 in the cytoplasmic domain are indicated along with the transmembrane (TM) domain. The cytoplasmic domain of FtsL is required to recruit FtsW, which in turn recruits FtsI. (D) Cartoons depicting the effect of various mutations on the activation of FtsWI according to the model. Top, FtsN action makes AWI available; middle, FtsL E88K is less dependent upon FtsN as the E88K substitution makes AWI available; bottom, FtsL E87K is resistant to FtsN action, and AWI does not become available or is defective in interaction with FtsWI.
FtsL Acts through FtsI ® charge in each case produced their respective phenotypes, it strongly suggests that these mutations disrupt rather than enhance interactions.
In our random mutagenesis screen, we did not isolate dominant negative mutations in ftsB; however, since six of the dominant negative mutations in ftsL overlapped the CCD, we used site-directed mutagenesis to alter the more conserved residues that overlap FtsB's CCD domain. Seven residues flanking the CCD domain were altered, but none produced a dominant negative phenotype ( Table 1). Six of these still complemented an ftsB deletion strain. This result suggests that the dominant negative mutations are unique to ftsL.
Dominant negative FtsL mutants are defective in activation of septal PG synthesis. A dominant negative phenotype could result from incorporation of an FtsL mutant into the FtsQLB complex that fails to (i) recruit downstream proteins (FtsWI), (ii) respond to FtsN (FtsQLB locked in OFF state), or (iii) generate an output signal in response to FtsN (ON state but failure to interact with a downstream partner). To test the first possibility, we assessed the localization of green fluorescent protein (GFP)-FtsI, which depends upon FtsW (3,23). It was present in crossbands within filamentous cells following expression of ftsL E87K or ftsL A90E , indicating recruitment to the Z ring (Fig. S1D). This result suggests that the ftsL mutations blocked either the response to FtsN or a downstream event such as interaction with FtsWI.
The dominant negative ftsL mutations were tested to see if they could be rescued by a strong activation mutation (ftsL E88K ) in cis. While ftsL R61E and ftsL A90E were readily rescued by ftsL E88K , ftsL L86F and ftsL E87K were not (Fig. S2A). If we assume that ftsL E88K mimics FtsN action and switches FtsQLB to the ON state, it suggests that ftsL R61E and ftsL A90E are able to carry out steps downstream of FtsN action. Based on these results, we suspected overexpression of ftsN would also rescue ftsL A90E and ftsL R61E but not ftsL E87K or ftsL L86F . This, in fact, was the case ( Fig. S2B and Table 1). Since ftsL R61E and ftsL A90E were rescued by enhancing the activation signal (by introducing an ftsL activation mutation or ftsN overexpression), it suggests they favor the OFF state (partially resistant to FtsN) but can carry out downstream events when activated. We therefore focused on ftsL L86F and ftsL E87K since it is unclear if they are locked in the OFF state or are unable to produce a signal in response to FtsN. Dominant negative FtsL mutants are rescued by FtsW activation mutants. Based on our results, we hypothesized that activation of FtsWI requires a signal from the periplasmic domain of FtsL (AWI domain) which is made available by FtsN action or ftsL activation mutations. We also hypothesized that activated alleles of ftsW might rescue a strong dominant negative ftsL allele since they require less input from FtsN. Two such ftsW alleles exist: ftsW M269I , which weakly bypasses ftsN (12), and ftsW E289G , which was isolated as described in Materials and Methods and bypasses ftsN. The latter mutation was also isolated using another approach and shown to bypass ftsN (24).
To see if these ftsW alleles could rescue ftsL L86F or ftsL E87K , a plasmid with these alleles under an arabinose-inducible promoter (derivatives of pSD296 [P ara ::ftsL]), as well as a compatible plasmid with ftsW alleles under an IPTG-inducible promoter (derivatives of pSEB429 [P 204 ::ftsW]), were introduced into SD399 (ftsL::kan/pSD256 [repA ts :: ftsL]). The resultant strains were tested on plates at 37°C to deplete wild-type (WT) ftsL, and arabinose and IPTG were added to induce the ftsL and ftsW alleles, respectively. Expression of ftsW M269I and ftsW E289G , but not ftsW, rescued the dominant negative ftsL alleles (Fig. 3). These ftsW activation alleles still required the presence of ftsL, as they could not bypass it (Fig. 3, right panel). Also, ftsW M269I was able to rescue an allele containing both mutations (ftsL L86F/E87K ), whereas overexpression of ftsN could not (Fig. S3A). These results indicate that ftsL L86F/E87K cannot transmit the periplasmic signal in response to FtsN.
Although the above-described results demonstrate that the two dominant negative mutations (ftsL L86F or ftsL E87K , alone or combined) block FtsN, they do not distinguish ftsL]) containing derivatives of pSD296 (P ara ::ftsL) with different alleles of ftsL was transformed with derivatives of pSEB429 (P 204 :: ftsW) carrying WT ftsW or either of two active alleles of ftsW. Transformants were spot tested at 37°C (to deplete WT FtsL) in the presence of arabinose (to induce the ftsL allele present on derivatives of pSD296) and increasing concentrations of IPTG to induce alleles of ftsW (ftsW, ftsW M269I or ftsW E289G ). The cartoons below depict the interpretation of the results. On the left, FtsWI is not recruited in the absence of FtsL; center, FtsWI is recruited but not activated in the presence of a dominant negative FtsL mutant; right, active FtsW mutants suppress dominant negative FtsL mutants in one of two ways (see the text).
FtsL Acts through FtsI ® between whether they lock FtsQLB in the OFF state (nonresponsive to FtsN) or prevent a downstream step (responsive to FtsN but failing to interact with FtsWI). We suspect the latter for the following reasons. To rescue ftsL L86F or ftsL E87K , ftsW E289G has to be overexpressed, whereas the chromosomal level of ftsW E289G was sufficient to bypass ftsN (expression of ftsW or the activation alleles from the plasmids complement an ftsW depletion mutant in the absence of IPTG [ Fig. S4A], whereas 15 to 30 mM is required to rescue ftsL L86F or ftsL E87K ). Consistent with this, expression of ftsL E87K is toxic to a strain with ftsW M269I on the chromosome (Fig. S4B), highlighting that an active ftsW allele cannot bypass the dominant negative ftsL mutation at the chromosomal level. These results suggest that the dominant negative ftsL mutants are defective in interaction with FtsWI in the periplasm (lack of the periplasmic interaction necessitates overexpression of an active ftsW). Consistent with the ftsL mutations blocking a step downstream of FtsN action, an active ftsB mutation, ftsB E56A , which can also bypass ftsN (10), cannot suppress ftsL E87K (Fig. S3B). This result is also consistent with an activation mutation in ftsL or overexpression of ftsN being unable to rescue ftsL E87K (Fig. S3A). Furthermore, all substitutions in ftsL E87 that remove the negative charge are dominant negative (Fig. S1C), suggesting they disrupt, rather than enhance, an interaction. Therefore, we favor the idea that these mutations in the AWI domain abrogate FtsL's interaction with FtsWI and that under physiological conditions, FtsWI is recruited by cyto FtsL and activated by FtsQLB when it is in the ON state (AWI available). Loss of cyto FtsL function rescued by activation mutations in the CCD domain of FtsL. One mutation from the random mutagenesis screen altered a residue in the cyto FtsL domain (ftsL L24K ). Although weak, adding a second mutation that altered a conserved residue in this domain (ftsL I28K ) yielded a stronger dominant negative phenotype (Fig. S5A). Since cyto FtsL is required for FtsW recruitment (13), it suggests that FtsL L24K , FtsL I28K , and the double mutant assemble into a complex with FtsQ and FtsB that poorly recruits FtsW. Consistent with this, deletion of the cytoplasmic domain of FtsL (FtsL D1-30 ) produced a strong dominant negative phenotype (Fig. S5B) resulting in filamentation and a failure to recruit FtsI (Fig. S5C).
Since FtsN is proposed to switch FtsQLB to the ON state to activate FtsWI (10, 11), we speculated above that this switch involves a conformational change that exposes AWI to activate FtsWI. If this is the case, the activation mutations may compensate for the loss of cyto FtsL by making the AWI domain available, which recruits FtsWI as well as activating it. As expected, ftsL D1-30 failed to complement DftsL; however, ftsL D1-30 carrying two activation mutations (ftsL G92D and ftsL E88K ) restored colony formation, indicating that both recruitment and activation of FtsW were restored (Fig. 4A). Further tests showed that both activation mutations were required for rescue (Fig. S6A). The rescue was fairly effective, as the average cell length of the strain expressing ftsL D1-30/G92D/E88K was only twice that of a strain expressing ftsL (Fig. S6B), whereas the strain expressing ftsL D1-30 was extremely filamentous. These two activation mutations also eliminated the toxicity of the ftsL L24K/I28K allele (Fig. S6C) and rescued its ability to complement (Fig. S6D). These results are consistent with a model in which the ftsL activation mutations cause a conformational change in FtsQLB that makes AWI available to recruit and activate FtsWI. It follows that under physiological conditions, the arrival of FtsN results in the exposure of AWI FtsL, which cooperates with cyto FtsL to recruit and activate FtsWI.
Since the ftsL activation mutations appear to mimic FtsN action, we expected that overexpression of ftsN would also rescue ftsL D1-30 . To test this, an ftsL depletion strain was transformed with a plasmid expressing ftsL D1-30 and a plasmid that overexpresses ftsN to a level that is sufficient to bypass zipA or ftsEX (21). The increased FtsN rescued ftsL D1-30 (Fig. S6E), suggesting that the excess FtsN caused AWI to be available to recruit and activate FtsWI, indicating that overexpression of ftsN is comparable to combining the two activation mutations (ftsL G92D and ftsL E88K ) in rescuing ftsL D1- 30 .
Dominant negative ftsL mutations negate rescue by activation mutations. If ftsL activation mutations rescue ftsL D1-30 by making AWI available to recruit and activate FtsWI, the dominant negative mutations should impair rescue by blocking the interaction. As seen in Fig. 4A, addition of ftsL E87K negated the rescue of ftsL D1-30 by the activation mutations, consistent with ftsL E87K blocking interaction between the AWI domain and FtsWI.
The FtsQLB complex probably exists in equilibrium between ON and OFF states, with the activation mutations and overexpression of FtsN favoring the ON state (AWI available). Overexpression of FtsW or FtsW M269I may also tip the equilibrium to the ON state and rescue ftsL D1-30 , as the increased level of FtsW may promote capture of the ON state. Indeed, expression of ftsW M269I , even at low levels of induction, rescued ftsL D1-30 , and at higher levels of induction, WT ftsW also started to rescue (Fig. 4B).
Earlier, we showed that overexpression of ftsW M269I and ftsW E289G , but not ftsW, rescued ftsL carrying dominant negative mutations (Fig. 3). This result is consistent with these activated mutants being recruited by the FtsL mutants (through cyto FtsL) but not requiring an activation signal from the AWI domain (via FtsN) (12). In the absence of cyto FtsL, however, our results suggest rescue requires a functional AWI in FtsL peri . If so, the dominant negative mutations should be detrimental in this context. As expected, the addition of either of two dominant negative mutations (ftsL L86F or ftsL E87K ) to ftsL D1-30 prevented rescue by FtsW M269I (Fig. 4C). These results are consistent with AWI being required to recruit FtsWI in the absence of cyto FtsL. It is worth noting that when either of two FtsL domains is nonfunctional (due to either inactivation of the cytoplasmic domain or the presence of the dominant negative mutations [such as L86F and E87K] in full-length FtsL), the active FtsW mutants must be overexpressed to rescue growth (see Discussion). FtsL Acts through FtsI ® Rescue of FtsL D1-30 by overexpression of FtsI. In the hierarchical assembly pathway, FtsW is recruited in a cyto FtsL-dependent manner followed by FtsI, which is recruited by interaction between FtsW and the transmembrane segment of FtsI (23). However, we considered the possibility that with FtsL D1-30 , the recruitment is reversed or FtsWI is recruited as a complex through interaction of AWI with FtsI. This thinking was driven in part by geometric constraints. The periplasmic domain of FtsL is thought to be a continuous alpha helix with its transmembrane domain such that the AWI domain would extend about ;45 Å away from the cytoplasmic membrane (15) (Fig. S7). In the RodA-PBP2 structure (homologous to FtsW-FtsI), the non-penicillin-binding (nPB) or pedestal domain of PBP2 sits on top of RodA and extends into the periplasm (25). Assuming FtsW-FtsI adopts a similar structure, FtsI could contact AWI in FtsL.
If FtsI interacts with the AWI domain, overexpression of ftsI may rescue FtsL D1-30 by enhancing the interaction with AWI FtsL and shifting the equilibrium of FtsQLB from OFF to ON through mass action. To test this, we compared the ability of the overexpression of ftsI and ftsW to rescue FtsL D1-30 . As shown in Fig. 5A, expression of ftsI was much more efficient than that of ftsW in rescuing FtsL D1-30 . The efficient rescue of FtsL D1-30 by FtsI suggests that it captures the transient ON state of FtsQLB (AWI exposed) and converts FtsQL D1-30 B into an active form similar to ftsL activation mutations (Fig. 4A). The rescue of FtsL D1-30 by overexpression of FtsW may involve the formation of an FtsWI complex that interacts with AWI, and the more efficient rescue of FtsL D1-30 by activated FtsW (compared to WT FtsW seen in Fig. 4B) may be due to it being active and more readily forming a complex with FtsI.

Park et al.
overexpression of ftsI would not be expected to rescue FtsL carrying the dominant negative ftsL mutations since the AWI activation signal would not be present. As expected, overexpression of ftsI was unable to suppress ftsL L86F/E87K , indicating the AWI signal was still required (Fig. 5B).
The possibility that AWI recruits and activates FtsWI by acting through FtsI was further examined by testing FtsI mutants isolated by the Weiss lab (26). These mutants localize to the division site but fail to complement a depletion strain and recruit FtsN. We reasoned that if an active FtsL acts directly on FtsW (to generate an active FtsW), an activated FtsL should have no more ability to rescue such mutants than an active FtsW mutant. However, if an activated FtsL acts on FtsI, it might have more ability to rescue FtsI mutants than an active FtsW. Therefore, each FtsI mutant was tested to see if it could be rescued by an active form of FtsL or FtsW (FtsL G92D/E88K and FtsW M269I , respectively). Of the seven FtsI mutants tested, two mutants (FtsI S61F and FtsI R210C ) were rescued by both FtsW M269I and FtsL G92D/E88K (Fig. 6 and Fig. S8). However, FtsL G92D/E88K rescued two additional mutants (FtsI G57D and FtsI V86E ; Fig. 6B, rows 5 and 9) not rescued by FtsW M269I (Fig. 6A, rows 3 and 5). The rescue of these two mutants by an activated FtsL (but not an activated FtsW) suggests that AWI acts through FtsI to activate FtsW rather than acting directly on FtsW. ). Transformants were spot tested on plates at 37°C (to inactivate ftsI23 ts ) with arabinose added to induce the ftsI alleles and increasing concentrations of IPTG to induce ftsW M269I . Note: additional alleles of ftsI were not rescued by ftsW M269I (Fig. S8). (B) Rescue of FtsI mutants by ftsL E88K/G92D . To test rescue of FtsI mutants by activated FtsL, MCI23 (ftsI23 ts recA:: spc) was transformed with compatible plasmids expressing an activated allele of ftsL (pKTP100* [P tac ::ftsL E88K/ Interaction between FtsL and FtsWI. Our results point to an interaction between the cytoplasmic domain of FtsL and FtsW required for recruitment of FtsWI and between the periplasmic domain of FtsL with FtsI, which is required for activation of FtsWI. To obtain additional support for interactions between the various proteins, we tested the effect of these mutations using the bacterial two-hybrid (BACTH) system. We observed strong interactions between FtsL and FtsW and between FtsL and FtsI, which were eliminated when the cytoplasmic domain of FtsL was deleted, consistent with cyto FtsL being required for recruiting FtsWI (FtsL D1-30 ; Fig. 7A). Elimination of these interactions allowed us to use FtsL D1-30 to assess the effects of the activation mutations in ftsL and ftsW on the interactions. Although the ftsW activation mutation had little effect, the addition of two ftsL activation mutations resulted in a strong interaction between FtsL D1-30 and FtsI and a weaker interaction between FtsL D1-30 and FtsW (Fig. 7B). The strong interaction with FtsI suggests it interacts with FtsL, where the weak interaction with FtsW suggests that FtsW is an intermediate. Importantly, the further addition of a dominant negative mutation (ftsL E87K ) eliminated the interaction conferred by the activation mutations. This FtsL variant with three amino acid substitutions was stable, as it interacted with FtsQ as well as the WT FtsL (Fig. S6F). These effects with FtsL D1-30 were also observed with FtsL L24K/I28K (Fig. S6G). The effects of these ftsL mutations in the BACTH system correlate with the effects these mutations have on the rescue of FtsL D1-30 and FtsL L24K/I28K ; the ftsL activation mutations promote rescue which is negated by an ftsL dominant negative mutation (Fig. 4A and Fig. S6D, respectively).
Rescue of DftsL by MalF-FtsL and FtsW-FtsK fusions. Next, we tested if the periplasmic portion of FtsL transported to the periplasm could activate FtsWI in the absence of full-length FtsL. To do this, a MalF-FtsL fusion was constructed under the control of an IPTG-inducible promoter in which the cytoplasmic and transmembrane (TM) domains of FtsL were replaced with the corresponding regions of MalF ( cyto/TM MalF-peri FtsL). In contrast to FtsL D1-30 , this MalF-FtsL fusion was not dominant negative (Fig. S9A), indicating that the TM region of FtsL must be present for the fusion to displace FtsL from the FtsQLB complex and disrupt FtsW recruitment. This is consistent with the TM region of FtsL being unique (27) and the TMs of FtsL and FtsB being required for these proteins to interact (16,18). Furthermore, the MalF-FtsL fusion was unable to complement an ftsL depletion strain even if the strain carried an ftsW M269I  (Fig. S9B). This was expected since FtsW would not be recruited.
Since the MalF-FtsL fusion cannot cooperate with FtsQB to recruit FtsW, we used an FtsW-cyto FtsK fusion which complements an ftsK deletion mutant, as well as a ftsW deletion mutant, indicating it is targeted directly to the Z ring and bypasses FtsQLB for recruitment (28 and data not shown). This MalF-FtsL fusion was unable to rescue the growth of a strain depleted for FtsL and containing FtsW-cyto FtsK, even if the fusion carried both ftsL mutations (Fig. 8, top panel). The inability to activate the FtsW-cyto FtsK fusion could be for a variety of reasons, including that FtsB is uncoupled from FtsL, and the FtsW-cyto FtsK likely competes with endogenous FtsW for FtsI. Nonetheless, the MalF-FtsL fusion with the two activation mutations was able to rescue an FtsL-depleted strain containing the FtsW-cyto FtsK fusion with the ftsW M269I mutation. (Fig. 8). Even the MalF-FtsL fusion without the ftsL activation mutations partially rescued growth at higher induction levels. These results suggest that MalF-FtsL acts on FtsI associated with the FtsW M269I-cyto FtsK fusion that is already at the Z ring to rescue growth. Since the activation mutations in ftsL potentiate MalF-FtsL activity, it suggests that in addition to making AWI available within the FtsQLB complex, they may also alter the structure of AWI to enhance its interaction with FtsWI.

DISCUSSION
Here, we investigated how septal PG synthesis in the divisome is activated by FtsN and identified a critical and unique role for FtsL. Our results are consistent with the recruitment of FtsW requiring the cytoplasmic domain of FtsL and the activation of FtsWI being dependent upon AWI in the periplasmic domain of FtsL. Based upon the seminal work by the de Boer lab, which is supported by the work from the Bernhardt lab (10, 11) and our results (12) and those here, we propose that the arrival of FtsN leads to a conformational change in the FtsQLB complex that makes the AWI domain of FtsL, as defined by the dominant negative ftsL mutations, available to activate FtsWI by acting through FtsI. Furthermore, activation mutations in the CCD domain of FtsL as well as those in FtsB mimic FtsN action to cause a conformational change in FtsQLB to FIG 8 Activation mutations allow a malF-ftsL fusion to complement DftsL in the presence of a ftsW-cyto ftsK fusion. Plasmid pKTP103 (P tac ::malF 1-37 -ftsL 58-121 -6xhis) was introduced into an ftsL depletion strain (SD439 ftsL::kan/pSD296 [P ara ::ftsL]) in the presence of a plasmid constitutively expressing a FtsW-FtsK cyto fusion without or with an activation mutation (pND16 [P ftsK ::ftsW-cyto ftsK] or pND16* [P ftsK ::ftsW M269I -cyto ftsK], respectively). The strains were spot tested on plates without arabinose (to deplete WT ftsL) and in the presence of IPTG (to induce malF-ftsL) (with or without the activation mutations [ftsL E88K and ftsL G92D ]). The cartoon to the right depicts the activity of the FtsL constructs.
FtsL Acts through FtsI ® expose the AWI domain. This model is supported by the ability of activation mutations in ftsL to rescue FtsL mutations (FtsL D1-30 and FtsL L24K/I28K ) deficient in FtsWI recruitment and by the dominant negative mutations in ftsL (ftsL L86F/E87K ) negating the rescue. The effects of these ftsL mutations (both activation and dominant negative) on the rescue of the FtsL mutants correlates with their effects on the observed interaction between FtsL and FtsWI in the BACTH system. The model is also supported by the ability of the expression of ftsI to rescue FtsL D1-30 more efficiently than ftsW. Furthermore, FtsL acting on FtsI to activate FtsW is supported by the ability of an active FtsL mutant to rescue FtsI mutants not rescued by an activated FtsW. Thus, we propose that as a result of E FtsN action, the AWI domain of FtsL becomes available to interact with FtsI within the FtsWI complex to activate FtsW and synergizes with cyto FtsL in stabilizing the FtsWI complex in the divisome. Thus, FtsL within the FtsQLB complex functions as a clamp to maintain FtsWI in the divisome.
The AWI domain. Altering seven residues in the periplasmic domain of FtsL produced a dominant negative phenotype. All, except for one, are clustered together around the CCD. We focused on L86 and E87 and believe these are central to the AWI domain. This suggestion is based upon the following: (i) L86 and E87 are relatively well conserved, and loss of the negative charge at E87 is sufficient to produce a dominant negative allele (suggesting disruption of an interaction); (ii) the ftsL L86F or ftsL E87K dominant negative mutations are not suppressed by activation mutations (ftsL E88K or ftsB E56A ) or ftsN overexpression; (iii) cyto FtsL mutants that fail to recruit FtsWI are rescued by the addition of two ftsL activation mutations (ftsL E88K/G92D ); (iv) the rescue of cyto FtsL mutants by ftsL activation mutations or overexpression of ftsW M269I is negated by adding dominant negative mutations (ftsL L86F or ftsL E87K ); and (v) the effects of these mutations on the interaction of FtsL with FtsWI in the BACTH system correlate well with the effects of these mutations on the rescue of FtsL D1-30 . It is likely that other regions of FtsL (and FtsB), such as the transmembrane domains (TM) and coiled coil domains, are also involved in interaction with FtsWI.
The dominant negative mutations in ftsL are less responsive to FtsN, and most overlap the CCD domain, which was defined by hyperactive mutations that are less dependent upon FtsN (10,11). Despite the overlap, the residues comprising each domain mostly lie on opposite sides of a putative helix (Fig. 2C). The dominant negative mutations appear to be unique to ftsL, as we were unable to isolate any such mutations in ftsB. Although previous studies suggested that FtsN induces a change in FtsQLB from an OFF to ON conformation (10), it was not clear how this switch led to activation of FtsWI. Here, we identify the AWI domain of FtsL and suggest that the function of the conformational switch is to make AWI available to interact with FtsWI. Since FtsQLB may be a dimer, the conformational change could involve disruption of this dimer which makes AWI available; however, this will require further study (15,16,29).
Additional evidence for the unique importance of the periplasmic domain of FtsL comes from the ability of the MalF-peri FtsL fusion to rescue an FtsW-cyto FtsK fusion when both are carrying activation mutations. The FtsW-cyto FtsK fusion is unable to support growth in the absence of FtsL even though it localizes. On the other hand, the MalF-peri FtsL fusion does not form a complex with FtsQB, so it is not recruited to the divisome. Nonetheless, the ability of the MalF-peri FtsL to collaborate with FtsW-cyto FtsK (when both are carrying activation mutations) to rescue growth suggests that the periplasmic domain of FtsL is able to act on FtsW-cyto FtsK complexed with FtsI.
While this paper was under review, Marmont and Bernhardt (30) reported that FtsLB was sufficient to activate PG synthesis by FtsWI in vitro, providing biochemical evidence for an activation model. They also isolated dominant negative mutations in ftsL which overlap those we isolated, even though their work was done in Pseudomonas aeruginosa and FtsL is not so highly conserved at the sequence level. Some, but not all, of the dominant negative mutants prevented activation in vitro. However, the in vitro system does not fully recapitulate the in vivo regulation, as FtsN was not required for activation.
Conditions that rescue FtsL D1-30 favor interaction between the AWI domain of FtsL and FtsI. Surprisingly, loss of the cytoplasmic domain of FtsL, which prevents recruitment of FtsWI and blocks cell division, could be rescued by activation mutations in the periplasmic domain of FtsL as well as by overexpression of FtsN. We reasoned that these activation conditions expose an interaction that normally occurs when the divisome is activated and that this interaction is able to compensate for the loss of cyto FtsL to recruit FtsWI. In support of this model, ftsL activation mutations in ftsL D1-30 promoted interaction between FtsL and both FtsW and FtsI. Also, these interactions were negated by the addition of a dominant negative mutation. These results suggest that FtsL within the FtsQLB complex functions as a transmembrane clamp (Fig. 1) to stabilize the active FtsWI complex within the divisome. The cytoplasmic domain of FtsL is required to recruit FtsW, which in turn recruits FtsI. FtsN action then frees the AWI domain to interact with FtsI and, as we have shown here, this domain, when freed, is able to rescue ftsL D1-30 , indicating FtsWI recruitment is restored.
Since it is likely FtsQLB exists in equilibrium between ON and OFF states, we reasoned that expression of the downstream partner might also rescue ftsL D1-30 by capturing the ON form and pulling the equilibrium in that direction. In fact, the active form of FtsW was effective in rescuing ftsL D1-30 , much more so than FtsW. However, expression of FtsI was very effective in rescuing ftsL D1-30 and much more so than overexpression of FtsW, which barely rescued at high overexpression. This (i) suggested that FtsI is the direct downstream target of AWI, (ii) suggested that rescue by expression of FtsW likely involves formation of an FtsWI complex recruited by AWI, and (iii) raises the possibility that the activated form of FtsW interacts more strongly with FtsI. Consistent with the rescue of ftsL D1-30 by expression of FtsI or activated FtsW being dependent upon the interaction of AWI with FtsWI in the periplasm, it was prevented by the addition of the dominant negative ftsL mutations. This is in stark contrast to the suppression of the dominant negative mutations in full-length ftsL by activated FtsW. When fulllength FtsL is present, an FtsW activated by mutation is recruited normally and no longer requires the activation signal so the dominant negative mutations do not prevent the rescue (although rescue is aided by overexpression of the activated FtsW). On the other hand, FtsW and FtsI are unable to rescue, as they still depend upon the AWI signal.
Our results suggest that FtsWI forms a dynamic complex, and it is this complex that is preferred by FtsL. If FtsWI formed a stable complex, then overexpression of FtsW would be toxic, as excess FtsW would titrate FtsI away from the division site inhibiting division. However, overexpression of ftsW is not toxic in WT cells and it only weakly rescued ftsL D1-30 . Also, when FtsQLB is overexpressed and purified, FtsW and FtsI only copurify efficiently if they are both expressed, indicating that the FtsWI complex interacts more stably with FtsQLB than FtsW or FtsI alone (31). Thus, overexpression of FtsW may favor complex formation with FtsI and septal localization to rescue ftsL D1-30 . More efficient rescue by an activated FtsW could be due to it favoring complex formation with FtsI. On the other hand, the rescue of ftsL D1-30 by FtsI expression is probably due to a direct interaction with AWI; otherwise, the rescue of ftsL D1-30 by FtsW and FtsI should be comparable, since overexpression of either should promote complex formation.
The product of FtsN action is an activated FtsWI complex in which both FtsW and FtsI are active. The ability of active FtsW mutants to suppress the dominant negative FtsL mutants (and bypass the periplasmic signal) indicates that an active FtsW leads to an active FtsWI complex. Among previously isolated FtsI mutants, we found some that were rescued by both an active FtsW mutant and an active FtsL mutant. However, an activated FtsL rescued two additional FtsI mutants that could not be rescued by an activated FtsW. This suggests that AWI acts on FtsI to activate FtsW and does not act directly on FtsW. In other words, the signal transmission from FtsN is from peri FtsL ! FtsI ! FtsW and not peri FtsL ! FtsW ! FtsI.
Although in vitro results suggest that FtsQLB acts as an inhibitor with FtsL inhibiting PBP1b and FtsQ inhibiting FtsI and therefore FtsW (31), our results are more compatible with a model in which AWI is sequestered within FtsQLB and becomes available upon FtsN action to activate FtsWI. The findings that ftsL activation mutations rescue FtsL D1-30 and promote interaction between FtsL D1-30 and FtsWI in the BACTH are consistent with the FtsL-FtsWI interaction activating FtsWI. This conclusion is also supported by the ftsL dominant negative mutations negating both of these activities.
Comparison of models for divisome and elongasome activation. It is interesting to compare our model for FtsWI activation with the model proposed for activation of the RodA-PBP2 pair that are part of the elongasome (homologous to FtsW-FtsI [PBP3]). That model is based upon (i) the structure of the MreC-PBP2 complex (32) and (ii) the finding that mutations that bypass mreC and activate RodA-PBP2 map to the nonpenicillin (nPD) or pedestal domain of PBP2 (33). It is thought that these mutations mimic the binding of MreC to PBP2, altering the conformation of PBP2, which results in the activation of RodA. In this way, the activity of RodA and PBP2 are coupled to ensure RodA only makes glycan strands when its cognate PBP is present. This is remarkably similar to our model for FtsW-FtsI (PBP3) activation with FtsL (with possibly a supporting role for FtsB) being analogous to MreC. The isolation of FtsW activation mutants that bypass FtsN suggests that an activated FtsW results in an active FtsI. Furthermore, an active FtsW mutant can rescue dominant negative FtsL mutants (i.e., bypass the signal from FtsN), indicating FtsI is also activated. Thus, we propose that FtsN action alters the conformation of FtsQLB so that AWI becomes available to interact with FtsI, leading to conformational change in FtsI that activates FtsWI's enzymatic activities.
Plasmids. The plasmids are listed in Table S1B. Genomic DNA extracted from the W3110 strain was used as a template to obtain PCR fragments to generate expression plasmids for ftsL. To construct the plasmids pKTP100 (P tac ::ftsL) and pKTP103 [P tac ::malF 1-37 ftsL 58-121 -6xhis], the ftsL open reading frame (ORF) was PCR amplified incorporating a strong ribosome binding site in the forward primers targeting ftsL, which included sequences for ftsL and malF 1-37 , respectively. The PCR fragments were digested with EcoRI and HindIII and ligated into the same sites in pJF118EH. Construction of pKTP104 (P T5 ::ftsL) and pKTP105 (P T5 ::ftsL 30-121 ) involved PCR amplification of the ftsL ORF, digestion with BamHI and HindIII, and ligation into the same sites in the pQE80L vector (Qiagen). The construction of pKTP108 [repA ts P syn135 :: ftsL] employed a similar approach to that used for pSD256 (12) except that a strong ribosome binding site was added and the XbaI site was used instead of EcoRI. To create pKTP109, the ftsI ORF was PCR amplified and digested with SacI and HindIII, followed by ligation into pBAD33 using sites with compatible overhangs. To generate plasmid pSD296 (P ara ::ftsL), the ftsL ORF and its flanking sequences (250 bp) were PCR-amplified, digested with XbaI and HindIII, and ligated into the same sites in the pBAD33 vector. Plasmids pKTP106 (P ara ::ftsL) and pKTP107 (P ara ::ftsL 30-121 ) were created by PCR amplification of ftsL and ftsL  , respectively, using the primers that contain the same ribosome binding site as in pKTP100. The two PCR fragments were cut with SacI and HindIII and cloned into sites in pBAD33 with compatible overhangs. To create pKTP101 (P tac ::ftsB), the ORF was PCR amplified and digested with EcoRI and HindIII followed by ligation into pJF118EH cut with the same enzymes. The pND16 [P ftsK ::ftsW-ftsK 179-1329 ] plasmid constitutively expresses the FtsW-FtsK C-terminal fusion protein, and pBL154 (repA TS P syn135 ::ftsN)