Essential role for FtsL in activation of septal PG synthesis

Spatiotemporal regulation of septal PG synthesis is achieved by coupling assembly and activation of the synthetic enzymes (FtsWI) to the Z ring, a cytoskeletal element required for division in most bacteria. In E. coli the recruitment of the FtsWI complex is dependent upon the cytoplasmic domain of FtsL, a component of the conserved FtsQLB complex. Once assembled, FtsWI is activated by the arrival of FtsN, which acts through FtsQLB and FtsA that are also essential for their recruitment. However, the mechanism of activation of FtsWI by FtsN is not clear. Here, we identify a region of FtsL that plays a key role in the activation of FtsWI which we designate AWI (Activation of FtsWI) and present evidence that FtsL acts through FtsI. Our results suggest that FtsN switches FtsQLB from a recruitment complex to an activator with FtsL interacting with FtsI to activate FtsW. Since FtsQLB and FtsWI are widely conserved in bacteria this mechanism is likely to be also widely conserved. Significance 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. Here we find an essential role for FtsL in activating septal PG synthesis and find that it acts on FtsI. Our results suggest a model where FtsWI is recruited in an inactive form by FtsQLB and upon FtsN arrival, FtsQLB undergoes a conformational change so that a region of FtsL, that we designate the AWI domain, becomes available to interact with FtsI and activate the FtsWI complex. This mechanism for activation of the divisome has similarities to activation of the elongasome and is likely to be widely conserved in bacteria.


Introduction 42
Bacterial cell division in most bacteria is carried out by a large protein complex called the 43 divisome or septal ring (1, 2). In E. coli it consists of 12 essential proteins and an ever-increasing 44 number of nonessential proteins. The essential proteins include FtsZ, which assembles into 45 treadmilling filaments that are tethered to the membrane by FtsA and ZipA (Z ring), and 9 46 additional proteins which display the following dependency for recruitment -FtsE/X < FtsK < FtsQ 47 ftsL E88K , ftsL L86F and ftsL E87K were not (Fig. S2A). If we assume that ftsL E88K mimics FtsN action and 130 switches FtsQLB to the ON state, it suggests that ftsL R61E and ftsL A90E are able to carry out steps 131 downstream of FtsN action. Based on these results we suspected overexpression of ftsN would 132 also rescue ftsL A90E but not ftsL E87K . As expected overexpression of ftsN rescued ftsL A90E but not 133 ftsL E87K (Fig. S2B). Since ftsL R61E and ftsL A90E can be rescued by enhancing the activation signal (by 134 introducing an ftsL activation mutation or ftsN overexpression), it suggests they favor the OFF 135 state but can carry out downstream events when activated. We therefore focused on ftsL L86F and 136 ftsL E87K since it is unclear if they are locked in the OFF state or are unable to produce a signal in 137 response to FtsN. 138 139

Dominant negative FtsL mutants are rescued by FtsW activation mutants 140
Based on our results we hypothesized that activation of FtsWI requires a signal from the 141 periplasmic domain of FtsL (AWI domain) which is made available by FtsN action or ftsL activation 142 mutations. Activation alleles of ftsW might rescue a strong dominant negative ftsL allele since 143 they require less input from FtsN. Two such ftsW alleles exist: ftsW M269I (which weakly bypasses 144 ftsN (12)) and ftsW E289G , which was isolated as described in the Materials and Methods and 145 bypasses ftsN. The same mutation was recently isolated using another approach and also shown 146 to bypass ftsN (biorivx 850073). 147 To see if these ftsW alleles could rescue ftsL L86F or ftsL E87K , a plasmid with these alleles 148 under an arabinose-inducible promoter (derivatives of pSD296 [Para::ftsL]), as well as a 149 compatible plasmid with ftsW alleles under an IPTG-inducible promoter (derivatives of pSEB429 150 [P204::ftsW]), were introduced into SD399 (ftsL::kan/pSD256 [repA ts ::ftsL]). The resultant strains 151 were spot tested at 37 o C to deplete WT ftsL and arabinose and IPTG were added to induce the 152 ftsL and ftsW alleles, respectively. Expression of ftsW M269I and ftsW E289G , but not ftsW, rescued 153 the dominant negative ftsL alleles (Fig. 3). These ftsW activation alleles still required the presence 154 of ftsL as they could not bypass it (Fig. 3, right panel). Also, ftsW M269I was able to rescue an allele 155 containing both mutations (ftsL L86F/E87K ) whereas overexpression of ftsN could not (Fig. S3A). 156 These results indicate that ftsL L86F/E87K cannot transmit the periplasmic signal in response to FtsN. 157 The above results demonstrate that the two dominant negative mutations (ftsL L86F or 158 ftsL E87K , alone or combined) block FtsN but do not distinguish between whether they lock FtsQLB 159 in the OFF state or prevent a downstream step (e.g, interaction with FtsWI). We suspect the latter 160 for the following reasons. To rescue ftsL L86F or ftsL E87K , ftsW E289G had to be overexpressed 161 whereas the chromosomal level of ftsW E289G is sufficient to bypass ftsN (expression of ftsW or the 162 activation alleles complement an ftsW depletion mutant in the absence of IPTG [ Fig. S4A] 163 whereas 15 to 30 µM is required to rescue ftsL L86F or ftsL E87K ). Consistent with this, expression of 164 ftsL E87K is toxic to a strain with ftsW M269I on the chromosome (Fig. S4B) highlighting that an active 165 ftsW allele cannot bypass the dominant negative ftsL mutation at the chromosomal level. 166 Therefore, the results suggest that the dominant negative ftsL mutants are defective in 167 interaction with FtsWI in the periplasm (lack of the periplasmic interaction may necessitate 168 overexpression of ftsW M269I ). Consistent with the ftsL mutations blocking a step downstream of 169 FtsN action, an active ftsB mutation, ftsB E56A , which can also bypass ftsN (10), cannot suppress 170 ftsL E87K (Fig. S3B). This conclusion is also consistent with an activation mutation in ftsL or 171 overexpression of ftsN being unable to rescue ftsL E87K (Fig. S3A). Also, all substitutions in ftsL E87 172 that remove the negative charge are dominant negative (Fig. S1C) (Fig. S5B). 185 Since FtsN is proposed to switch FtsQLB to the ON state to activate FtsWI (10, 11), we 186 speculated above that this switch involves a conformational change that exposes AWI to activate 187 FtsWI. If this is the case, then the activation mutations may compensate for the loss of cyto FtsL by 188 making the AWI domain available to recruit FtsWI as well as to activate it. As expected ftsL D1-30 189 failed to complement DftsL, however, ftsL D1-30 carrying two activation mutations (ftsL G92D and 190 ftsL E88K ) restored colony formation indicating that both recruitment and activation of FtsW were 191 restored (Fig. 4A). Further tests showed that both activation mutations were required for rescue 192 (Fig. S6A). The rescue was fairly effective as the average cell length of the strain expressing ftsL D1-193 30/G92D/E88K was only twice that of a strain expressing ftsL (Fig. S6B), whereas, the strain expressing 194 ftsL D1-30 was extremely filamentous. Combining these two activation mutations also eliminated the toxicity of the ftsL L24K/I28K allele (Fig. S6C) and rescued its ability to complement (Fig. S6D) of ftsW M269I , even at low levels of induction, rescued ftsL D1-30 and at higher levels of induction WT 219 ftsW also started to rescue (Fig. 4B). 220 Earlier we showed that overexpression of ftsW M269I and ftsW E289G but not ftsW rescued 221 ftsL carrying dominant negative mutations (Fig. 3). This result is consistent with these activated 222 mutants being recruited by the FtsL mutants (through cyto FtsL) but not requiring an activation 223 signal from the AWI domain (via FtsN) (12). In the absence of cyto FtsL, however, our results suggest 224 rescue is dependent upon a functional AWI in FtsL peri . If so, the dominant negative mutations 225 should be detrimental in this context. We found that the addition of either of two dominant 226 negative mutations (ftsL L86F or ftsL E87K ) to ftsL D1-30 prevented rescue by FtsW M269I (Fig. 4C)  to it more readily forming a complex with FtsI. 252 The above results indicate that the signal from FtsN via the AWI domain goes through 253 FtsI. As shown earlier, expression of activated alleles of ftsW suppresses ftsL L86F or ftsK E87K as they 254 no longer require the signal from AWI. In contrast, WT ftsW cannot suppress these alleles as it 255 still requires the AWI activation signal. Likewise, overexpression of ftsI should not rescue full 256 length FtsL carrying the dominant negative ftsL mutations. As expected, overexpression of ftsI 257 was unable to suppress ftsL L86F/E87K indicating the AWI signal was still required (Fig. 5B). 258 The possibility that AWI recruits and activates FtsWI by acting through FtsI was further 259 examined by testing FtsI mutants isolated by the Weiss lab (25). These mutants localize to the 260 division site but fail to complement and recruit FtsN. We thought it possible that some of these FtsI mutants were defective in relaying an activation signal from AWI to FtsW, and thus unable 262 to initiate a positive feedback loop leading to FtsN accumulation. Therefore, we compared the 263 rescue of these FtsI mutants by active mutants of FtsL and FtsW (FtsL G92D/E88K and FtsW M269I , 264 respectively) which are less dependent upon FtsN. The rationale was that if an active FtsL 265 converts FtsW to an active form then the rescue of FtsI mutants by FtsW M269I should be equal to 266 or greater than that by an active FtsL. On the other hand, if FtsL acts through FtsI to activate 267 FtsW, then FtsL G92D/E88K may be able to rescue FtsI mutants more efficiently than FtsW M269I . 268 Indeed, of seven FtsI mutants tested, two mutants (FtsI S61F and FtsI R210C ) were rescued by both 269 FtsW M269I and FtsL G92D/E88K ( Fig. 6 and S8). However, FtsL G92D/E88K rescued two additional mutants 270 (FtsI G57D and FtsI V86E ) not rescued by FtsW M269I . The rescue of these two mutants by an activated 271 FtsL (but not an activated FtsW) suggests that AWI acts through FtsI to activate FtsW rather than 272 acting directly on FtsW. 273

Interaction between FtsL and FtsWI 275
Our results are consistent with an interaction between the cytoplasmic domain of FtsL 276 and FtsW leading to recruitment of FtsWI and between the periplasmic domain of FtsL with FtsI 277 which is required for activation of FtsWI. As shown above, when the cyto FtsL-FtsW interaction is 278 eliminated activation mutations in ftsL and ftsW are able to rescue division which is abrogated 279 by the dominant negative mutations in ftsL. To obtain additional support for interactions 280 between the various proteins we tested the effect of these mutations using the bacterial two 281 hybrid (BTH) system. We observed strong interactions between FtsL and FtsW and between FtsL 282 and FtsI, however, which were eliminated when the cytoplasmic domain of FtsL was deleted, 283 consistent with cyto FtsL involved in recruiting FtsWI (FtsL D1-30 , Fig. 7A). This allowed us to use 284 FtsL D1-30 to assess the effects of the activation mutations in ftsL and ftsW. Although the ftsW 285 activation mutation had little effect, the addition of two ftsL activation mutations resulted in a 286 strong interaction between FtsL D1-30 and FtsI and a weaker interaction between FtsL D1-30 and FtsW 287 FtsL must be present for the fusion to displace FtsL from the FtsQLB complex and disrupt FtsW 300 recruitment. This is consistent with the transmembrane (TM) region of FtsL being unique (26) 301 and the TMs of FtsL and FtsB being required for these proteins to interact (16,18). Furthermore, 302 the MalF-FtsL fusion was unable to complement an ftsL depletion strain even if the strain carried 303 a ftsW M269I mutation and the ftsL construct carried the two activation mutations (Fig. S9B).

Since FtsQLB is required to recruit FtsWI and the MalF-FtsL fusion is insufficient to recruit 305
FtsW, we used an FtsW-cyto FtsK fusion which complements a ftsK deletion mutant, as well as a 306 ftsW deletion mutant, indicating it bypasses FtsQLB for recruitment (27) (and data not shown). 307 This MalF-FtsL fusion was unable to replace FtsL and rescue the growth of a strain containing 308 FtsW-cyto FtsK, even if the fusion carried activation both ftsL activation mutations (Fig. 8). This 309 inability to activate the FtsW-cyto FtsK fusion could be for a variety of reasons including that FtsQB 310 is uncoupled from FtsL and the FtsW-cyto FtsK likely competes with endogenous FtsW for FtsI. 311 However, this MalF-FtsL fusion with the two activation mutations was able to rescue a strain 312 containing the FtsW-cyto FtsK fusion with the ftsW M269I mutation. (Fig. 8). Even the MalF-FtsL fusion 313 without the ftsL activation mutations partially rescued growth at higher induction levels. These 314 results suggest that MalF-FtsL acts on FtsI associated with the FtsW M269I-cyto FtsK fusion that is 315 already at the Z ring to rescue growth. Since the activation mutations in ftsL potentiate MalF-316 FtsL, it suggests that in addition to making AWI available within the FtsQLB complex, they may 317 also alter the structure of AWI to enhance its interaction with FtsWI (since MalF-FtsL is not part 318 of the FtsQLB complex). 319 320

Discussion 321
Here we investigated how septal PG synthesis in the divisome is activated by FtsN and 322 identified a critical and unique role for FtsL. Our results are consistent with the recruitment of 323 FtsW requiring the cytoplasmic domain of FtsL and the activation of FtsWI being dependent upon 324 AWI in the periplasmic domain of FtsL. Based upon the seminal work by the de Boer lab, which is 325 supported by the work from the Bernhardt lab (10-11) and our results (12)  to support recruitment following the loss of cyto FtsL. In support of this, we could bypass ftsQ with 356 an activated allele of ftsA but were unable to bypass ftsB or ftsL (data not shown). 357 The dominant negative mutations in ftsL are less responsive to FtsN and most overlap the 358 CCD domain which was defined by hyperactive mutations that are less dependent upon FtsN (10, 359 11). Despite the overlap, the residues comprising each domain mostly lie on opposites sides of a 360 putative helix consistent (Fig. 2C). The dominant negative mutations appear to be unique to ftsL 361 as were unable to isolate any such mutations in ftsB. Previous studies suggested that FtsN induces 362 a change in FtsQLB from an OFF to ON conformation (10), however, it was not clear how this 363 switch led to activation of FtsWI. Here we identify the AWI domain of FtsL and suggest that the 364 function of the conformational switch is to make AWI available to interact with FtsWI. Since 365 FtsQLB is likely a dimer, the conformational change may involve disruption of this dimer which 366 makes AWI available, however, this will require further study (15,16,28).  and selecting Kan resistance and screening for arabinose dependency. PK4-1 (ftsL::kan/pKTP108 472 [Para::ftsL]) was generated by using the same procedure described above. Unless stated 473 otherwise, Luria-Broth (LB) medium containing 0.5% NaCl at indicated temperatures was used. 474 For selection on LB agar and growth in LB broth, the following antibiotics and reagents were 475 added at the indicated final concentrations as necessary (ampicillin, 100 μg/ml; spectinomycin, 476 50 μg/ml; kanamycin, 25 µg/ml; chloramphenicol, 10 µg/ml; tetracycline, 10 µg/ml; IPTG. 10-200 477 µM; glucose, 0.2%; and arabinose, 0.2%. 478 479

Random and Site-Directed Mutagenesis 511
To obtain the ftsL and ftsB mutant libraries (with a single missense mutation per ORF) an 512 optimal mutation rate (0.3-1 base/kb) for 1µg of template was adopted as recommended in 513 GeneMorph II Random Mutagenesis kit (Agilent Technologies). The PCR products were then 514 digested with EcoRI and HindIII and ligated into the pJF118EH vector using the same restriction 515 enzymes. A ligation pool of pJF118EH-ftsL or pJF118EH-ftsB containing putative mutations was 516 transformed into JS238 by electroporation and a library was prepared by isolating plasmids from 517 a pool of growing colonies. A dominant negative phenotype, largely characterized by flat colony 518 morphology was screened for by introducing the resulting plasmids into JS238. Specific point 519 mutations in ftsL, ftsL  and ftsW were introduced into some plasmids by using the 520 QuickChange site-directed mutagenesis kit according to the manufacturer's instruction (Agilent 521 Technologies). 522 523

Microscopy 553
The dominant negative effects of the FtsL mutants on cell division were assessed using 554 phase-contrast microscopy by monitoring the degree of filamentation. JS238 containing pKTP100 555 or derivatives carrying ftsL mutations was grown overnight in the presence of 100 μg/ml 556 ampicillin and 0.2% glucose and the cultures were diluted 1/200-1/500 in fresh LB medium 557 containing 100 μg/ml ampicillin. At OD540 ≈ 0.02, 50 µM IPTG was added and cell morphologies 558 were analyzed 2 hours later. 559 To visualize GFP-FtsI localization, SD285 [leu::Tn10 bla lacI q P207-gfp-ftsI] containing 560 pKTP106 (Para:: ftsL) or derivatives with the ftsL E87K or ftsL A90E mutations was grown overnight at 561 30°C in LB medium containing 50 μg/ml ampicillin and 10 μg/ml chloramphenicol. The overnight 562 cultures were diluted 1/200 ~ 1/500 in fresh LB medium containing the same antibiotics, 0.2% 563 arabinose, and 10-20 µM IPTG, and were incubated at 37°C until OD540≈ 0.4. Cells were 564 immobilized on an LB agarose pad and the localization of GFP-FtsI was recorded using a cooled 565 CCD camera and processed using Metamorph (Molecular Devices) and Adobe Photoshop. 566 567 Acknowledgements 568 569 We would like to thank Piet de Boer and David Weiss for strains and plasmids and Scott 570 Lovell for generating the model of the FtsL alpha helix. This study was supported by NIH grant 571 GM29746 to Joe Lutkenhaus. transformed into JS238. Colonies were picked and screened for sensitivity to IPTG. ftsL WT and 696 ftsL E88K (an activation allele) were included as controls and are not toxic. Several strong dominant 697 negative mutations (ftsL E87K , ftsL L86F and ftsL A90E ) and two weak mutations (ftsL R61C and ftsL E24K ) 698 were obtained in this way (Table 1). Additional mutations were obtained by site-directed 699 mutagenesis. B) Dominant negative mutants inhibit division. Phase contrast micrographs of JS238 700 expressing ftsL or ftsL E87K (derivatives of pKTP100 [Ptac::ftsL]) grown in liquid culture and induced 701 with 50 µM IPTG for two hours. Induction of the other alleles also inhibited division (Table 1)

. C) 702
FtsL, residues 54-99, was modelled (for illustration purposes) as an alpha helix since it is thought 703 to form a continuous alpha helix with the TM and this region is also thought to form a coiled coil 704 with FtsB. Altering residues in green leads to activation mutations, whereas altering those 705 residues in red are dominant negative. Altering residues in yellow had no effect. Note that the 706 activation mutations affect residues on one side of the helix whereas the dominant negative pKTP109 [Para::ftsI]). Transformants were spot tested on plates at 37 o C (to inactivate ftsI23 ts ) and 767 arabinose added to induce the ftsI alleles and increasing concentrations of IPTG to induce 768