INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Cell division in bacteria culminates with the formation of the midcell septum followed by cell separation. In Escherichia coli, at least nine gene products, FtsZ, FtsA, ZipA, FtsQ, FtsL, FtsI, FtsK, FtsW, and FtsN, have been identified as being essential for septation, and the septal localization of all of them has been reported previously (1-3, 5, 8, 12, 20, 29, 30, 31). The nature of the mechanism leading to the selection of the midcell division site is still unknown and poses one of the most fundamental question in cell biology. One approach to this issue is to determine the nature of the septal localization signal that targets cell division proteins to the division site.

Recent progress indicates that only portions of certain Fts proteins are necessary for septal localization. In the case of FtsK, the N-terminal-15% portion of this protein (comprising the membrane-spanning segments) is sufficient for localization to the division site but not for complementation of an ftsK defect (16). For FtsN, the periplasmic domain promotes targeting of this protein to the septum (29). Studies with replacements of the cytoplasmic and transmembrane domain of the bitopic membrane protein FtsQ suggest that the cytoplasmic domain and the membrane-spanning segment have little or no role in FtsQ targeting and function at the division site (3, 5). In contrast, for FtsI, a bitopic membrane protein with a large periplasmic enzymatic domain involved in peptidoglycan biosynthesis, replacements of the cytoplasmic or transmembrane domain of FtsI failed to complement and localize at the division site. These results, along with the fact that no common targeting motif can be identified in these proteins, suggest that each cell division protein carries a distinct and specific localization signal.

In this paper, we identify the requirements for septal localization and function of the cell division protein FtsL. Although the function of FtsL is still unknown, its sequence is conserved in gram-negative and gram-positive bacteria, and it is an essential cell division protein in E. coli and Bacillus subtilis (7, 9). FtsL is a bitopic cytoplasmic membrane protein whose structure includes a small amino-terminal cytoplasmic domain (37 residues), a single transmembrane segment (20 residues), and a small periplasmic domain (64 residues). The sequence of the periplasmic domain of FtsL exhibits a stretch of five leucines separated from each other by seven residues which has features of an -helical leucine zipper. The presence of this motif raises the possibility that FtsL interacts with itself or other proteins to form homo- or heterodimers. Studies in which the three domains of FtsL were exchanged with those of other bitopic membrane proteins showed that all three domains of FtsL are required for its function (11). Different possible roles can be attributed to FtsL domains, including that of signaling localization to the cell septum. Thus, one possible consequence of exchanging FtsL domains might be to interfere with its localization.

We recently used fluorescence microscopy to demonstrate that E. coli FtsL (FtsL_Ecol) is localized to the division site and that FtsL localization depends on the prior recruitment of FtsZ, FtsA, ZipA, and FtsQ but not FtsI and FtsN. These results, along with similar experiments done with FtsI and FtsQ, led us to propose that FtsL is recruited after FtsZ, ZipA, FtsA, and FtsQ and before FtsI and FtsN (8).

In this paper, we use green fluorescent protein (GFP) fusion technology to analyze the septal localization and function of FtsL derivatives in which the cytoplasmic domain and/or membrane-spanning segment of FtsL is replaced with analogous parts of either MalF, a maltose transport protein with no role in cell division, or FtsL_Hinf, the Haemophilus influenzae FtsL_Ecol homologue. Our main finding is that both the membrane-spanning segment and the periplasmic domain of FtsL_Ecol, but not the cytoplasmic domain, are required for localization to the division site. Further, we found that mutagenesis of the leucine repeat heptad motif in the periplasmic domain of FtsL severely impaired both localization and function. Finally, we report that the FtsL periplasmic domain can drive formation of sodium dodecyl sulfate (SDS)-resistant multimers in vitro. These findings are consistent with the likely existence of a coiled-coil structure predicted for the periplasmic domain of all FtsL homologues, which may be involved in a homo- or heterodimerization process important in the mechanism of cell division.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and phage. Bacterial strains and plasmids used in this study are listed in Table 1 and Table 2, respectively. Strain construction was done by generalized transduction with P1 (24) or specialized transduction with InCh. InCh picks up plasmid-borne genes by homologous recombination. In this case, the genes were gfp-ftsL or gfp-ftsL derivatives, bla, and lacI^q from pDSW207-based plasmids (31). Chromosomal insertions made with InCh were subsequently stabilized by selecting for a deletion that removed most of the phage, including its killing functions (D. Boyd et al., unpublished data; see also http://beck1.med.harvard.edu/LambdaInCh.html).

Media. Media were NZY or Luria-Bertani (LB) medium (9, 27). L-Arabinose or D-glucose was added as indicated to modulate expression of genes under control of the P_BAD promoter (10). For localization studies, isopropyl--D-thiogalactoside (IPTG) was used to modulate expression of gfp fusions under the control of modified trc promoters: 2.5 µM for P₂₀₇-gfp-ftsL derivatives (8, 31). The IPTG-dependent plasmid pJMG64 was maintained with 5 mM IPTG. Antibiotics were used for selection at the following concentrations: ampicillin (AMP) at 200 µg/ml for plasmids and 25 µg/ml for chromosomal alleles, kanamycin (KAN) at 40 µg/ml, chloramphenicol at 30 µg/ml, tetracycline (TET) at 15 µg/ml, and spectinomycin (SPEC) at 25 µg/ml. In complementation experiments, pJMG64 was counterselected with streptomycin used at 1.5 mg/ml.

Molecular biological procedures. Standard procedures were used for cloning and analysis of DNA, PCR, electroporation, and transformation (27). Enzymes used to manipulate DNA were from New England Biolabs (Beverly, Mass.). Oligonucleotides were from Gibco BRL (Gaithersburg, Md.). DNA sequencing was performed by the Micro Core Facility in the Department of Microbiology and Molecular Genetics at Harvard Medical School.

Sequence analysis. Amino acid sequence comparisons were performed with the Clustal package program (14). Homology searches were performed with Blast 2.0 at the National Center for Biotechnology Information website (http: //www.ncbi.nlm.nih.gov/BLAST/unfinishedgenome.html). Hydropathy plots and DNA sequence manipulations were done with DNAStrider 1.3 (22). Coiled-coil propensity was determined with Coiled 2.2 (19) on the ISREC website (http: //www.isrec.isb-sib.ch/software/COILS_form.html).

Construction of gfp fusions. (i) gfp-ftsL and gfp-LLL. Plasmids for making gene fusions to gfp were based on pDSW207 (31). pDSW236 is described in reference 8. To make pDSW207-LLL, LLL was amplified by PCR with pLD45 (11) as template and oligonucleotides L1 (CGAGAATTCAACAACAACATGATCAGCAGACTGACAGAA) and L2 (TCGAAGCTTTTTATTTTTGCACTACGATATTTTC) as primers. The amplified DNA was digested with EcoRI and HindIII (sites underlined) and ligated into the same sites of pDSW207 to create pJMG453. The fusion protein encoded by pJMG453 has the linker sequence YKEFNNNMI, where YK are the last residues of GFP and MI are the first two residues of FtsL.

(ii) Fusions to FtsL-MalF swap proteins. To make gfp-FLL, gfp-LFL, and gfp-FFL, the FLL, LFL, and FFL genes were amplified by PCR with pLD63, pLD94, and pLD90 (11), respectively, as template and oligonucleotides F1 (CG AGAATTCAACAACAACATGGATGTCATTAAAAAGAAACATTGGTG GC) (FLL and FFL) or L1 (LFL) and pBADrev (ACCGCTTCTGCGTTCTGATT) as primers. The amplified DNA was digested with EcoRI (site underlined) and HindIII (site present downstream of the constructs in pLD63, pLD94, and pLD90) and ligated into the same sites of pDSW207 to create pJMG447, pJMG448, and pJMG449. The fusion protein encoded by pJMG448 has the same linker sequence as that in GFP-LLL. The fusion proteins encoded by pJMG447 and pJMG449 have the linker sequence YKEFNNNMD, where YK are the last two residues of GFP and MD are the first two residues of FLL and FFL, whose amino terminus is derived from malF.

(iii) Fusions to FtsL-FtsI swap proteins. To make gfp-LLI, the LLI gene was amplified by PCR with pLD116 (11) as template and oligonucleotides L1 and pBADrev. The amplified DNA was digested with EcoRI and HindIII and ligated into the same sites of pDSW207 to create pJMG474. The fusion protein encoded by pJMG474 has the same linker sequence as that of pDSW236.

(iv) Fusions to FtsL_Ecol-FtsL_Hinf swap proteins. ftsL_Hinf was amplified by PCR with pGHIIA44 (American Type Culture Collection [ATCC], Manassas, Va.) as template and oligonucleotides HiL-5 (CCCCGAATTCAGAGAGGACGAATGCATGTCTGAAAATAATAAGCCTCG and HiL-3 (CCCCAAGCTTTTATCCTTATTCTCTAATTTCAACTTCTTGC) as primers. In HiL-5, the sequence located downstream of the EcoRI site corresponds to the region placed upstream of E. coli FtsL (the ribosome binding site is in boldface) followed by the H. influenzae coding sequence (italicized). The amplified DNA was digested with EcoRI and HindIII (sites underlined) and ligated into the same sites of pBAD18 to create pJMG452.

(v) Fusions to FtsL_Ecol periplasmic domain mutants. LL_HinfL, LL_GCN4L, and LL_CEB/PL were constructed as follows. First, a silent mutation was introduced in the DNA region encoding the periplasmic domain of LLL by the megaprimer methods (28). The megaprimer was generated with pLD45 as template and oligonucleotides Lxho3mp (GAATGGCGCAACCTGATCCTCGAGGAGAATGC) and pBADfw as primers. LLxL was then amplified with pLD45 as template and the megaprimer and pBADrev as primers. The mutation modified the 86th and 87th codons of LLL (CTT GAA [Leu Glu]) to (CTCGAG [Leu Glu]) and introduced a unique XhoI site in the LLL sequence to create pJMG460.

Other constructs. pAM238 derivative plasmids described in Table 2 were created as follows: pBAD18-based constructs were amplified by PCR with oligonucleotides pBADfw and pBADrev as primers. The amplified DNA was digested with EcoRI and HindIII and ligated into the same sites of pAM238. LLL was constructed as follows: a stop linker (LzStop1 [TCGACCAAAAATAATACGTAA] and LzStop2 [ACGTTTACGTATTATTTTTGG]) was inserted at the XhoI site of LLxL in pJMG473. The resulting plasmid (pJMG155) presents a deletion of residues 87 to 119 in the FtsL coding sequence. For pJMG64, the SacI-HindIII fragment of pLMG180 was introduced into pBAD39 linearized with the same enzymes to create pJMG64.

Construct verification. All constructs were checked by PCR with specific and vector-based primers. Then each construct was sequenced in the pBAD18 or pAM238 context. For GFP fusions, the gene coding for the protein fused to GFP was sequenced from a PCR product obtained by amplification of chromosomal DNA. The sequence was checked up to the GFP-swap junction point. The constructs FFF and LLF were sequenced up to the junction point with the end of the region coding for the first membrane-spanning segment. The integrity of the rest of the malF portion was then checked by testing the ability of LLF (pJMG451) to complement a malF mutation in DHB4 on a MacConkey maltose plate.

Growth conditions. (i) GFP-fusion protein expression. (a) Wild-type background. Cells expressing GFP fusions ("gfp fusion in wild-type background" section of Table 1) were grown overnight at 30°C in NZY medium plus AMP, diluted 1:200 in fresh NZY medium (without AMP) containing 2.5 to 10 µM IPTG, and grown on a roller at 30°C until the optical density at 600 nm (OD₆₀₀) reached 0.25.

(b) Depletion experiments. To deplete FtsL from strains indicated in the "gfp fusion in depletion background" section of Table 1, strains were grown overnight in NZY with AMP, KAN, chloramphenicol, and 0.2% arabinose (which allowed high-level expression of FtsL from plasmid pJMG197). Cultures were diluted into fresh medium containing chloramphenicol and 0.02% arabinose and grown to an OD₆₀₀ of 0.5. Cells were washed and resuspended in NZY medium and then diluted 1:200 into NZY with 0.2% glucose (which shut off synthesis of FtsL from pJMG197) or NZY with 0.2% arabinose plus 2.5 to 5 µM IPTG. After about 3 to 5 h, an OD₆₀₀ of 0.25 was reached and cells had become filamentous in the glucose culture. Samples were harvested and prepared for microscopy and Western blot analysis.

(ii) FtsL derivative protein expression from pBAD18-based constructs. pBAD18-based FtsL derivatives were introduced into KS272, grown overnight at 37°C in LB medium plus AMP, diluted 1:200 in fresh LB medium, and grown on a roller at 37°C until the OD₆₀₀ reached 0.25. Then, arabinose or glucose (0.2%) was added and cultures were grown for 1 to 2 more h until the OD₆₀₀ reached 0.6 to 1.

Complementation of an ftsL null mutation. (i) Complementation of FtsL derivative as GFP fusions inserted on the E. coli chromosome. GFP-FtsL fusions were tested for complementation as follows. Strains corresponding to the "gfp fusion in depletion background" section in Table 1 were grown on low-AMP (25 µg/ml)-KAN-chloramphenicol-arabinose plates and were tested on plates containing low AMP, KAN, and 2.5 µM IPTG, supplemented with 0.2% glucose. Under these conditions, where the pJMG197-borne copy of ftsL is repressed, only strains expressing a functional GFP-FtsL derivative could grow.

(ii) Complementation of null mutant by counterselection. The functionality of FtsL derivatives in the pAM238 context was tested as follows. FtsL variants made in pBAD18 were subcloned in pAM238 (a plasmid with a pSC101 origin of replication) under plac promoter control and introduced in strain JMG33 containing pJMG64 (Table 1 and Table 2). pJMG64 is a pBAD39-derived plasmid expressing a wild-type allele of ftsL, whose induction (with arabinose) is required to complement the chromosomal ftsL-phoA null mutation. pBAD39 is a counterselectable pBAD18-derived plasmid, carrying a lacI gene, whose replication is IPTG dependent and which expresses rpsL, conferring a dominant sensitivity to streptomycin (1.5 mg/ml). The complete features of pBAD39 will be described in detail elsewhere (L. M. Guzman, unpublished data). The rationale of the counterselection-based complementation test is the following: without IPTG and AMP and in the presence of streptomycin, cells carrying pJMG64 are counterselected. The expression of genes carried by the pAM238-based plasmid is then constitutive in JMG33 deprived of pJMG64, as there is no longer a lacI gene in the cells. Any functional FtsL derivatives would then support the growth of JMG33 in the absence of pJMG64.

Microscopy and image analysis. Cells were grown and processed for protein localization experiments as described in reference 8. Microscopy and image analysis were performed as described in reference 8.

Western blotting. Western blotting was performed by standard techniques as described in reference 8. To determine the steady-state levels of the GFP fusion proteins, we harvested cells from the same cultures as those used for fluorescence microscopy. Cells were pelleted by centrifugation and resuspended with 10 µl of SDS sample buffer per OD₆₀₀ unit of 0.1. Samples were boiled for 3 min, and 10 µl of sample was loaded onto an SDS-polyacrylamide gel. To determine the multimerization state of the FtsL derivatives, cells were pelleted by centrifugation and resuspended with 10 µl of SDS sample buffer per OD₆₀₀ unit of 0.1. In most of the experiments, samples were not boiled and 10 µl of sample was loaded onto a 10% (GFP fusion) or 13% (FtsL swaps) SDS-polyacrylamide gel.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The membrane-spanning segment and the periplasmic domain of FtsL are both required for septal localization. In previous complementation tests, we used swap proteins where the three domains of FtsL were exchanged with topologically equivalent domains of MalF, FtsI, and FtsQ. This analysis indicated that all three domains were required for FtsL function (11). One simple interpretation of these data is that all three domains participate in septal localization. To test this hypothesis, we fused a bright variant of GFP (6) at the amino terminus (cytoplasmic domain) of several swap proteins: LLL, FLL, FFL, LFL, LLF, LLI, and FFF (Table 3). The three-letter names indicate the source of each domain in the swap proteins: the first letter indicates that of the cytoplasmic region, the second indicates that of the transmembrane segment, and the third indicates that of the periplasmic domain. For example, LFL refers to a protein in which the cytoplasmic and periplasmic domains are derived from FtsL and the membrane-spanning domain is derived from the first transmembrane segment from MalF. In LLI, I stands for the periplasmic domain of FtsI, another bitopic membrane protein involved in cell division. LLL and FFF constructs differ from wild-type FtsL and MalF in that two and four amino acids at the borders of the membrane-spanning segment have been altered due to introduction of restriction sites in corresponding regions of the ftsL and malF genes, respectively (11). We integrated the genes coding for GFP fused to the N terminus of LLL, FLL, FFL, LFL, LLF, LLI, and FFF to generate merodiploid strains similar to those used for localization of GFP-FtsL (5, 8, 31). Cells expressing GFP fusions to the FtsL-MalF swap proteins were fixed and examined by fluorescence microscopy. GFP-LLL localized as well as the control protein in which GFP was fused to the wild-type FtsL, while GFP-FFF did not localize to the septum. The GFP-FLL protein was the only swap protein to localize to the septum with a frequency similar to that of GFP-LLL (Table 3 and Fig. 1A).

View larger version (90K): [in this window] [in a new window]   FIG. 1.   Subcellular location of GFP-FtsL swap proteins in cells fixed during exponential growth. Arrows indicate rings at division sites. (A) Septal localization of FtsLEcol-MalF swap proteins fused to GFP. Strains used were as follows: GFP-FtsL, EC438; GFP-LLL, JMG287; and GFP-FLL, JMG469. (B) Septal localization of FtsLEcol-FtsLHinf swap proteins fused to GFP. Strains used were as follows: GFP-HLL, JMG371; GFP-LHL, JMG372; and GFP-HHL, JMG374. (C) Septal localization of FtsL periplasmic mutants fused to GFP. Strains used were as follows: GFP-LLL(L70H), JMG303; GFP-LLL(L84D), JMG304; and GFP-LLHinfL, JMG300.

View larger version (60K): [in this window] [in a new window]   FIG. 2.   Steady-state levels of GFP-FtsL-MalF swap proteins as determined by Western blotting with an anti-FtsL antibody. Molecular mass standards (kilodaltons) are indicated to the left of the blot, while the positions of GFP-FtsL fusions are shown to the right. The GFP fusion protein produced is indicated above each lane. Strains used were as follows: (A) GFP, EC452; GFP-FtsL, EC438; GFP-LLL, JMG287; GFP-LFL, JMG289; GFP-FFL, JMG292; GFP-FLL, JMG469; GFP-LHL, JMG372; GFP-HLL, JMG371; and GFP-HHL, JMG374; (B) GFP, EC452; GFP-FtsL, EC438; GFP-LLL, JMG287; GFP-LLL, JMG299; GFP-LLL(L70H), JMG303; GFP-LLL(L84D), JMG304; GFP-LLHinfL, JMG300; GFP-LLGCN4L, JMG301; and GFP-LLC/EBPL, JMG302.

The H. influenzae FtsL homologue neither complements nor localizes in E. coli. To further investigate the essential features of the FtsL localization signal, we asked whether a homologue of E. coli FtsL from another bacterial species could complement for FtsL function. Such an approach has proven to be informative in studies of the function of the cell division proteins FtsA and FtsZ (21). We chose the H. influenzae FtsL (FtsL_Hinf) as a good candidate to test for complementation and localization in E. coli. The FtsL_Hinf amino acid sequence is 40% identical to that of the FtsL of E. coli, and the chromosomal position and locus genetic organization of ftsL_Hinf are very similar to those of ftsL_Ecol. The homology between the two proteins is distributed evenly across the sequence except for the N terminus of FtsL_Hinf, which is shorter by 14 residues. FtsL_Ecol and FtsL_Hinf present the same predicted bitopic topology and have a conserved heptad motif in their periplasmic domain (Fig. 3). We cloned the ftsL_Hinf gene from the H. influenzae ATCC cosmid library and expressed it under the pBAD arabinose-inducible promoter with the ftsL_Ecol translation signals. Complementation tests showed that ftsL_Hinf did not complement an ftsL_Ecol null mutant. To verify the expression of FtsL_Hinf under these conditions, we generated fusions of FtsL_Hinf to alkaline phosphatase, where alkaline phosphatase is fused after the last residue of the FtsL_Hinf periplasmic domain. In this construct, the promoter and translation signals are the same as in constructs expressing FtsL_Ecol and FtsL_Hinf. PhoA fusions have been used to verify expression and proper topology of FtsL_Ecol (11). The hybrid protein FtsL_Hinf-PhoA gave high levels of alkaline phosphatase expression, as observed on indicator plates containing 5-bromo-3-chloro-indolylphosphate (data not shown), indicating that FtsL_Hinf, as FtsL_Hinf-PhoA, was expressed and assumed the predicted topology and arrangement in the membrane. Therefore, the lack of functionality of FtsL_Hinf does not appear to be due to lack of expression or proper insertion in the envelope.

View larger version (33K): [in this window] [in a new window]   FIG. 3.   Comparison between E. coli and H. influenzae FtsL. (A) Kyte and Doolittle hydropathy plot comparison between FtsLEcol and FtsLHinf. (B) Amino acid sequence comparison between FtsLEcol and FtsLHinf. The first and second numbers indicated above the sequences correspond to amino acid positions in FtsLEcol and FtsLHinf, respectively. FtsLEcol and FtsLHinf membrane-spanning segments are in oversized boldface. The shaded area corresponds to the region exchanged in the LLHinfL swap (see text). The a to f positions of the heptad repeat of the potential coiled-coil domain are indicated above the sequence of the periplasmic domain of FtsLEcol and FtsLHinf. The d position is indicated in boldface. Conserved residues are indicated with a star below the sequence. Conservative substitutions are indicated with dots.

Portions of FtsL_Hinf and FtsL_Ecol are functionally interchangeable. To test whether portions of the FtsL_Hinf protein could functionally substitute for the homologous regions of FtsL_Ecol, we replaced the first half of the periplasmic domain of the swap protein LLL (residues 58 to 85 [Fig. 3]) with its corresponding region from FtsL_Hinf (residues 45 to 72). This new protein, named LL_HinfL, is able to complement an ftsL null mutant, although less efficiently than LLL or wild-type FtsL (Table 4).

The integrity of the FtsL periplasmic domain is required for localization. Our data suggested that the periplasmic domain of the E. coli FtsL contains information essential for localization to the division site. This domain displays features of a potential leucine zipper with regular heptad repeats with Leu residues in most of the d position of the heptad (Fig. 3). This motif may promote the formation of a helical structure that encompasses at least the entire periplasmic domain (64 residues). We constructed FtsL mutants to test whether this potential structure is part of the septal localization signal. We first deleted portions of the FtsL periplasmic domain corresponding to the first three heptads: LLL had 27 residues (residues 58 to 85 were removed [Table 5]). This construct was not functional, nor did it localize as a GFP fusion. In order to determine the effect of the disruption of the potential leucine zipper motif, we mutagenized the second [Leu70 by a His in LLL_(L70H)] and fourth [Leu84 to an Asp in LLL_(L84D)] leucines located at the d positions of the heptad in the interaction face of the potential structural motif. The ability of these FtsL variants to complement for FtsL function was tested: LLL_(L70H) and LLL_(L84D) complemented weakly compared to the positive control LLL (Table 5). The poor functionality of LLL_(L70H) and LLL_(L84D) correlated with a reduced frequency of localization of GFP-LLL_(L70H) and GFP-LLL_(L84D) where GFP was fused to the N terminus of these proteins. In these cases, only about 20% of the cells displayed a septal localization signal (Table 5 and Fig. 1C). The consequences of the mutagenesis of leucines 70 and 84 on FtsL are consistent with a contribution of these leucines to a structural leucine zipper-type motif. To determine whether it would be sufficient for FtsL function to simply replace this region with a known leucine zipper, we exchanged the first three repeats of the periplasmic domain of FtsL (residues 58 to 85) with two well-characterized leucine zippers: the yeast GCN4 transcription activator and the mammalian C/EBP transcription activator (LL_GCN4L and LL_C/EBPL, respectively). We took great care in these constructs to keep unchanged the spacing of the leucines (Fig. 4). Neither construct complemented an FtsL defect or allowed localization to the division site (Table 5). Western blotting indicated that the stability and abundance of each FtsL variant were similar to or higher than that of FtsL or LLL (Fig. 2; see also Fig. 6).

View larger version (15K): [in this window] [in a new window]   FIG. 4.   Sequence alignment of the regions that are not common to LLL, LLGCN4L, and LLC/EBPL swap proteins. Dotted lines represent flanking FtsL sequence which is the same in the three proteins. Leu70 and Leu84 are mutated in LLL(L70H) and LLL(L84D).

The heptad motif of the periplasmic domain of E. coli FtsL is conserved in all identified FtsL homologues and has the characteristics of a coiled-coil domain. The results presented above suggested that the heptad repeats of the periplasmic domain of FtsL may be required for FtsL function and that they may be indicative of structural features of its periplasmic domain. Heptad repeat motifs such as those found in FtsL_Ecol and FtsL_Hinf have been reported to have a high propensity to form -helical multimeric coiled coils. Coiled coils are helical bundles of two to five helices that have a distinctive packing of amino acid side chains at helix-helix interfaces (18). This packing results in every seventh residue occupying an equivalent position on the helix surface. The seven positions are named a to g: a and d are usually occupied by hydrophobic residues and form the helix interface (core). Oppositely, charged residues in positions e and g flank the core and stabilize the protein through ionic interactions. The outer positions, b, c, and f, are not directly involved in the formation of the helical bundle and are free to interact with surrounding molecules. A wide variety of proteins involved in maintaining cell shape, in organizing the cytoplasm, and in cell movement and membrane fusion are coiled coils of two or more -helices (17).

View larger version (55K): [in this window] [in a new window]   FIG. 5.   Multiple alignment of FtsL homologues. Predicted membrane-spanning segments are boxed. The a to f positions of the heptad repeat of the potential coiled-coil domain are indicated above the sequence. The d position is indicated in boldface. The numbers above the alignment refer to the E. coli FtsL amino acid position. FtsLEcol, E. coli FtsL; FtsLYpes, Y. pestis FtsL; FtsLBaph, B. aphidicola FtsL; FtsLHinf, H. influenzae FtsL; FtsLBsub, B. subtilis FtsL; FtsLSaur, S. aureus FtsL; FtsLEfae, E. faecalis FtsL; FtsLSpne, S. pneumoniae FtsL.

The periplasmic domain of E. coli FtsL is a multimerization domain. The data presented above suggest that FtsL may adopt a coiled-coil structure that could encompass both the transmembrane and the periplasmic domain. Coiled-coil structures have been involved in the formation of multimers where two to five helices interact to form a superhelical bundle. The stability of the coiled-coil interactions is variable but may be high and in some cases thermostable and SDS resistant (13). To test whether potential FtsL multimers were stable enough to resist denaturation by SDS, we carried out an SDS-polyacrylamide gel electrophoresis analysis of a protein extract from a strain where FtsL was expressed from a multicopy plasmid under an inducible promoter. On SDS-polyacrylamide gels, two forms could be detected with anti-FtsL antibodies: a lower form that ran at the expected monomeric FtsL molecular mass (about 14 kDa) and an upper form running at 30 kDa that may correspond to an FtsL homodimer (Fig. 6). In some experiments, higher-order multimers could also be seen, although less consistently than the apparent dimer (data not shown). The abundance of the dimeric form was considerably reduced in boiled samples (data not shown). In order to determine whether the formation of this putative dimer is dependent on the presence of the periplasmic domain, cell extracts from strains expressing LLL, FLL, LFL, and FFL swap proteins as well as LLL_(L70H) and LLL_(L84D) were submitted to the same analysis. Samples from cells expressing LLL and FLL, but not LFL or FFL, displayed the putative dimer bands (Fig. 6A). Samples from strains expressing the two leucine mutants displayed a much-reduced ability to dimerize that could not be observed consistently (not shown in Fig. 6 and noted as +/ in Table 7). These results demonstrate that only constructs carrying the membrane-spanning segment and periplasmic domain of FtsL multimerize and that mutagenesis of two d positions [LLL_(L70H) and LLL_(L84D)] of the putative coiled-coil motif reduces the dimerization ability. To carry further this analysis, we analyzed cell extracts from strains expressing LLL, HLL, LHL, and HHL swap proteins as well as LL_HinfL and LL_GCN4L. Figure 6B shows that LLL, HLL, LHL, and LL_HinfL dimerize while HHL dimers could not consistently be observed (Fig. 6 and noted as +/ in Table 7). The levels of dimerization of HLL and LHL were found to be consistently lower than the one observed with FLL, although HLL and LHL were stable enough at the steady-state level. Samples from strains expressing LL_HinfL displayed about the same ability to dimerize as LLL. The introduction of a well-characterized dimerization motif in LL_GCN4L led to the efficient formation of dimers and even higher-order multimers (tetramers?) (Fig. 6). These results suggest that the FtsL_Hinf membrane-spanning segment cannot fully replace its E. coli homologue with regard to the multimerization phenotype although it fully complements the localization phenotype. It also shows that multimerization alone is not sufficient to fulfill the FtsL role in cell division.

View larger version (60K): [in this window] [in a new window]   FIG. 6.   Western blot of FtsL SDS-resistant multimers. The figure shows the multimerization state of FtsL swap proteins as determined by Western blotting with an anti-FtsL antibody. Molecular mass standards (kilodaltons) are indicated to the left of the blot, while the positions of FtsL monomers and dimers are shown to the right. The FtsL swap protein produced is indicated above each lane. Background strain used was KS272. Plasmids used were as follows: (A) control, pBAD18; LLL, pLD45; FLL, pLD63; LFL, pLD94; FFL, pLD90; LLL(L70H), pJMG438; and LLL(L84D), pJMG439; (B) control, pBAD18; LLL, pLD45; HLL, pJMG432; LHL, pJMG430; HHL, pJMG431; LLHinfL, pJMG427; and LLGCN4L, pJMG428.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this paper, we have analyzed the structure-function relationship of the essential cell division protein, FtsL. This analysis is facilitated by the simple domain structure of FtsL, a bitopic membrane protein composed of an N-terminal cytoplasmic domain, a transmembrane segment, and a periplasmic domain. We have studied the role of these domains by swapping them or portions of them with other cell division proteins or with an unrelated protein. In these studies, we have examined the role of these different domains in three properties of FtsL. These properties are (i) the ability to form dimers or even multimers on SDS gels, (ii) the ability to localize to the site of formation of the cell septum, and (iii) the ability to provide an essential function in the cell division process. The results of the swap studies (summarized in Table 7) show that the three functions of FtsL are separable.

Oligomerization of FtsL. We have shown that, on SDS gels, a substantial proportion of FtsL is found in a band corresponding to dimers. We believe that this oligomerization is made possible by the apparent coiled-coil structure that the periplasmic domain, with its heptad repeats of leucines, may assume, perhaps along with the two other domains of the protein (see below). We find that dimerization is correlated with the presence of the intact E. coli FtsL periplasmic and transmembrane domains; the presence of the periplasmic domain alone is not sufficient. Furthermore, mutations in either of two leucines of the heptad repeats in the periplasmic domain reduce the efficiency of dimerization. However, dimerization per se is not sufficient for the two other properties of FtsL: localization to the septum and function in cell division. This we showed by providing an alternative dimerization signal in the periplasmic domain, the leucine zipper region of the Saccharomyces cerevisiae GCN4 protein. In this case, the protein still forms dimers on gels but is nonfunctional for the other properties of FtsL. Thus, specific sequences in the periplasmic domain are required for the localization process, beyond those required for dimerization.

Localization to the septum. Localization of FtsL to the septal region of the E. coli cells also appears to be dependent on the presence of its periplasmic domain and on a transmembrane segment either from E. coli FtsL or from the H. influenzae homologue. The HLL, HHL, and LHL swap constructs all localize efficiently to the septum. Furthermore, the FLL construct, in which a cytoplasmic domain of the unrelated membrane protein, MalF, replaces the normal FtsL cytoplasmic domain, also localizes to the septum with a frequency comparable to that of the wild type. Thus, the cytoplasmic domain is needed neither for dimerization nor for localization. However, since the localization studies were done in a background that expresses wild-type FtsL from the chromosome, we cannot distinguish between the possibility that FLL localizes on its own to the septum and the possibility that it localizes via an interaction of the dimerization domain with an already localized wild-type FtsL.

Function in cell division. Finally, even certain swap constructs that localize very efficiently to the cell septum do not provide the complete set of functions required to complement an FtsL null mutation. The constructs FLL, HHL, and LHL all localize to the septum with the same efficiency as wild-type FtsL, but none of them complement. The finding with the FLL swap protein indicates that the cytoplasmic domain of FtsL, which is not essential for dimerization or septal localization, is, nevertheless, required for function. Complementation by the HLL construct indicates that the H. influenzae cytoplasmic domain of FtsL is sufficiently close in structure to that of the E. coli protein, despite a low degree of sequence conservation, that it can function properly. This is somewhat surprising since the FtsL_Hinf cytoplasmic domain is 14 residues shorter than the one of FtsL_Ecol. Thus, the essential region of the cytoplasmic domain of FtsL_Ecol may lie in the sequence that starts after residue 15 (Fig. 3). Two other gram-negative bacterial FtsL homologues, identified in Y. pestis and B. aphidicola, also display short cytoplasmic domains. Multiple alignments show that each of the three homologues has two conserved charged residues in common (R21 and D31 in FtsL_Ecol [Fig. 5]) that may be good mutagenesis targets for identifying critical residues in the FtsL cytoplasmic domain.

A potentially unbroken transmembrane and periplasmic coiled-coil motif is involved in localization to the division site. Localization and complementation analysis as well as coiled-coil predictions is compatible with a helical structure prolonged throughout the membrane-spanning segment and the cytoplasmic domain. This raises the possibility that the FtsL coiled coil could run essentially unbroken through the membrane and be involved in the formation of multimers. FtsL SDS-resistant multimers of higher order than dimers can be observed in some conditions, suggesting that, at least in vitro, interaction of more than two molecules may take place. Whereas the presence in vivo of such multimers remains to be demonstrated, preliminary data involving in vivo cross-linking between cysteines introduced in the periplasmic domain (Leu63 to Cys and Leu70 to Cys) suggest that the observed SDS-resistant multimers are likely to be homomultimers (J. M. Ghigo, unpublished results).