Localization of FtsI (PBP3) to the Septal Ring Requires Its Membrane Anchor, the Z Ring, FtsA, FtsQ, and FtsL

Abstract

Assembly of the division septum in bacteria is mediated by several proteins that localize to the division site. One of these, FtsI (also called penicillin-binding protein 3) of Escherichia coli, consists of a short cytoplasmic domain, a single membrane-spanning segment, and a large periplasmic domain that encodes a transpeptidase activity involved in synthesis of septal peptidoglycan. We have constructed a merodiploid strain with a wild-type copy of ftsI at the normal chromosomal locus and a genetic fusion of ftsI to the green fluorescent protein (gfp) at the lambda attachment site. gfp-ftsI was expressed at physiologically appropriate levels under control of a regulatable promoter. Consistent with previous results based on immunofluorescence microscopy GFP-FtsI localized to the division site during the later stages of cell growth and throughout septation. Localization of GFP-FtsI to the cell pole(s) was not observed unless the protein was overproduced about 10-fold. Membrane anchor alterations shown previously to impair division but not membrane insertion or transpeptidase activity were found to interfere with localization of GFP-FtsI to the division site. In contrast, GFP-FtsI localized well in the presence of β-lactam antibiotics that inhibit the transpeptidase activity of FtsI. Septal localization depended upon every other division protein tested (FtsZ, FtsA, FtsQ, and FtsL). We conclude that FtsI is a late recruit to the division site, and that its localization depends on an intact membrane anchor.

How the division septum is formed and how its formation is spatially and temporally regulated are among the most fundamental unanswered questions in prokaryotic cell biology. Studies with Escherichia coli and Bacillus subtilis indicate that septum assembly is mediated by a large number of proteins that localize to the division site, where they are postulated to form a multiprotein complex sometimes referred to as the septalsome or divisome (for recent reviews, see references 12, 37, and 43). Among the proteins known to localize to this site in E. coli are FtsZ, FtsA, FtsI, FtsN, FtsK, FtsW, and ZipA (3, 5, 8, 29, 38, 53, 57, 59). In B. subtilis, three division proteins have been shown to localize to the division site: FtsZ, the FtsQ-homologue DivIB, and DivIC, which has no obvious homologue in E. coli (30, 34, 35, 55). These findings allow us to rephrase the questions raised at the start of this paragraph: how do the division proteins localize to the division site? And what do they do once they get there?

In this article, we identify some of the requirements for septal localization of FtsI of E. coli. FtsI (also called penicillin-binding protein 3 [PBP3]) is a bitopic membrane protein with a large periplasmic domain that encodes an enzymatic activity (transpeptidase) involved in peptidoglycan synthesis (1, 10, 42). Genetic and biochemical evidence indicate that FtsI is required specifically for synthesis of peptidoglycan at the division septum, while penicillin-binding protein 2 (PBP2), a homologue of FtsI, appears to be the primary transpeptidase for cell elongation (9, 50, 51, 58).

Previously, we used immunofluorescence microscopy (IFM) to show that FtsI is localized to the division site during the later stages of cell growth and throughout septation (57). In these studies, FtsI was at the division site in about 50% of the cells. We also observed polar localization in 10 to 20% of the cells. Polar FtsI was not expected, and we offered several potential explanations, including that it might be an artifact. In our hands, FtsI proved very difficult to detect consistently by IFM, presumably owing to its low abundance (∼100 molecules/cell) (19, 57). Obtaining strong FtsI signals required high concentrations of primary antibody, inviting problems with spurious cross-reaction. In addition, detection of FtsI was very sensitive to the level of fixation and degree to which cells had been digested with lysozyme as part of the permeabilization procedure necessary for antibodies to gain access to FtsI. These problems, which were discussed in our initial report and which we have encountered to a lesser but still significant degree with other Fts proteins, prevented us from extending our studies to look at localization of FtsI in various fts mutants.

Wang et al. (53) have also studied FtsI localization using IFM. These authors reported that FtsI localizes to the division site in an FtsZ- and FtsA-dependent manner. Interestingly, localization of FtsI was not observed when cells were treated with furazlocillin, a β-lactam antibiotic that inactivates the transpeptidase activity of FtsI. This observation implies that FtsI requires its catalytic activity to localize to potential division sites. Wang et al. (53) did not observe polar localization of FtsI. One caveat that pertains to the studies of Wang et al. (53) is that the FtsI signals reported were extremely weak under conditions when the protein localizes well, suggesting that FtsI could have been overlooked in filaments or at the poles even if were present.

We have now achieved both high sensitivity and high specificity by fusing FtsI to a bright variant of the green fluorescent protein (GFP) of Aequorea victoria (14, 16). To express gfp-ftsI at physiologically appropriate levels and to avoid potential problems arising from the high and variable copy number of plasmids, we placed gfp-ftsI under control of an isopropyl-β-d-thiogalactoside (IPTG)-regulatable promoter and recombined the fusion into the E. coli chromosome at the lambda attachment site (attB). These genetic manipulations were facilitated by two new tools that we think will be generally useful for applying GFP to the study of protein localization in E. coli. The first tool is a set of plasmid vectors that allow one to make N-terminal or C-terminal fusions of a target protein to GFP and to express those fusions under the control of an IPTG-regulatable promoter of weak or moderate strength. The second and more unusual tool is a lambda phage that we call lambda InCh for in chromosome. This phage, which will be described elsewhere (11), is similar to a lambda vector used previously in our lab (21). Lambda InCh can be used to pick up plasmid-borne gene(s) in vivo by homologous recombination and to put them onto the chromosome by specialized transduction. The resulting strain is a merodiploid, which has a wild-type copy of the target gene at the normal chromosomal locus and a gfp fusion copy at attB. The power of these tools is that all of the molecular biology can be done in plasmids, while all of the steps involving lambda are done in vivo by growing the appropriate strains under the appropriate conditions of temperature and antibiotic selection. Using this system, we made chromosomal gfp fusions to several essential division genes and mutant derivatives thereof. Here we use some of these fusions to study localization of FtsI to the division site.

MATERIALS AND METHODS

Bacterial strains, plasmids, and phage.

Bacterial strains and plasmids used in this study are listed in Table 1. Strain construction was by generalized transduction with P1 (40) or specialized transduction with lambda InCh (11).

TABLE 1.

Strains and plasmids

Strain or plasmid	Relevant genetic markers or features	Construction	Source or reference
Strains without gfp fusions
BL21(DE3)	F−ompT hsdSB(rB− mB−) gal dcm (DE3)		Novagen, Madison, Wis.
KS272	F− ΔlacX74 galE galK thi rpsL ΔphoA(PvuII)		52
MC4100	F−araD139 ΔlacU169 ΔrelA1 rpsL150 thi mot flb5301 deoC7 ptsF25 rbsR		Lab collection
JP313	MC4100 Δara714		Joe Pogliano
LMG64	KS272 ftsI23(Ts) leu::Tn10 recA::cat		28
EC433	MG1655 ftsQ1 leu::Tn10		D. Weiss
MM61	F−araD139 ΔlacU169 strrftsA12(Ts) leu::Tn10		Lab collection
DRC14	MC4100 ftsZ84(Ts) leu::Tn10		D. RayChaudhuri
MJC431	KS272 ftsI::TnphoAI173ΔIS50R(Kanr)/pLMG173		28
EC548	KS272 ftsI::TnphoAI173ΔIS50R(Kanr)/pDSW262	From MJC431, replace pLMG173 with pDSW262	This study
JOE170	KS272 ftsQ::TnphoA80(Kanr)/pJC10		15
LMG145	KS272 ftsL::TnphoAL81ΔIS50R(Kanr)/pLMG180		26
SM551	F′ Δlac(MS562) λ− λsmel− nalArsupF58(=suIII+)		Susan Michaelis
DHB6521	SM551 λInCh1 (Kanr)		11
DHB6504	SM551 λInCh1 · pSX102 (Ampr)		11
gfp fusions in “wild-type” backgrounds
EC436	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp-ftsI	From pDSW234 using λInCh1	This study
EC447	MC4100 Δ(λattL-lom)::bla lacIqP210-ftsA-gfp	From pDSW233 using λInCh1	This study
EC448	MC4100 Δ(λattL-lom)::bla lacIqP208-ftsZ-gfp	From pDSW230 using λInCh1	This study
EC450	MC4100 Δ(λattL-lom)::bla lacIqP208-zipA-gfp	From pDSW242 using λInCh1	This study
EC452	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp	From pDSW207 using λInCh1	This study
EC454	P207-gfp-ftsI leu::Tn10	P1(DRC14) × EC436→select Tetr, screen Tia	This study
EC479	P210-ftsA-gfp leu::Tn10	P1(DRC14) × EC447→select Tetr, screen Ti	This study
EC484	P208-ftsZ-gfp leu::Tn10	P1(DRC14) × EC448→select Tetr, screen Ti	This study
EC489	P208-zipA-gfp leu::Tn10	P1(DRC14) × EC450→select Tetr, screen Ti	This study
EC505	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp-III	From pDSW246 using λInCh1	This study
EC507	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp-FFI	From pDSW247 using λInCh1	This study
EC509	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp-FII	From pDSW248 using λInCh1	This study
EC511	MC4100 Δ(λattL-lom)::bla lacIqP207-gfp-IFI	From pDSW249 using λInCh1	This study
EC522	MC4100 Δ(λattL-lom)::kan lacIqP207-gfp-ftsI	From pDSW254 using λInCh1 · pSX102	This study
EC530	JP313 Δ(λattL-lom)::kan lacIqP207-gfp-ftsI	From pDSW234 using λInCh1	This study
gfp fusions in Ts backgrounds
EC455	P207-gfp-ftsI leu::Tn10 ftsA12(Ts)	P1(MM61) × EC436→select Tetr, screen Ts	This study
EC457	P207-gfp-ftsI leu::Tn10 ftsQ1(Ts)	P1(EC433) × EC436→select Tetr, screen Ts	This study
EC458	P207-gfp-ftsI leu::Tn10 ftsZ84(Ts)	P1(DRC14) × EC436→select Tetr, screen Ts	This study
EC480	P210-ftsA-gfp leu::Tn10 ftsA12(Ts)	P1(MM61) × EC447→select Tetr, screen Ts	This study
EC481	P210-ftsA-gfp leu::Tn10 ftsI23(Ts)	P1(LMG64) × EC447→select Tetr, screen Ts	This study
EC486	P208-ftsZ-gfp leu::Tn10 ftsI23(Ts)	P1(LMG64) × EC448→select Tetr, screen Ts	This study
EC488	P208-ftsZ-gfp leu::Tn10 ftsZ84(Ts)	P1(DRC14) × EC448→select Tetr, screen Ts	This study
EC491	P208-zipA-gfp leu::Tn10 ftsI23(Ts)	P1(LMG64) × EC450→select Tetr, screen Ts	This study
gfp fusions in deple-tion backgrounds
EC550	ftsI::TnphoA P207-gfp/pDSW262	P1(EC452) × EC548→select Ampr	This study
EC552	ftsI::TnphoA P207-gfp-ftsI/pDSW262	P1(EC436) × EC548→select Ampr	This study
EC556	ftsI::TnphoA P207-gfp-III/pDSW262	P1(EC505) × EC548→select Ampr	This study
EC538	ftsQ::TnphoA P207-gfp-ftsI/pJC10	P1(EC436) × JOE170→select Ampr	This study
EC607	ftsL::TnphoA P207-gfp-ftsI/pJMG197	P1(LMG145) × EC530/pJMG197→select Kanr	This study
Plasmids
pBAD18	Arabinose regulation, Ampr		27
pBAD18-Kan	Arabinose regulation, Kanr		27
pBAD33	Arabinose regulation, Cmr		27
pET-26b(+)	Expression vector with T7lac promoter, pelB leader sequence, C-terminal His-tag, Kanr		Novagen
pTrc99A	Expression vector, Ampr		Pharmacia
pGFPmt2	Source of bright variant of gfp		16
pGC165-sfiA+	pUC8-sfiA		33
pZAQ	pBR322-ftsQAZ		56
pLMG173	pBAD18-ftsI		28
pLD30	pBAD18-III		28
pLD43	pBAD18-FFI		28
pLD57	pBAD18-FII		28
pLD118	pBAD18-IFI		28
pDSW163	pET-26b(+)-ftsI periplasmic domain		This study
pJC3	pBAD18-zipA		J. Chen
pJC10	pBAD33-ftsQ		15
pJMG197	pBAD33-ftsL		23
pDSW204	Promoter down mutation in −35 of pTrc99A		This study
pDSW206	Promoter down mutations in −35 and −10 of pTrc99A		This study
pDSW207	PDSW204-gfp-MCS (fusion vector)		This study
pDSW208	PDSW204-MCS-gfp (fusion vector)		This study
pDSW209	PDSW206-gfp-MCS (fusion vector)		This study
pDSW210	pDSW206-MCS-gfp (fusion vector)		This study
pDSW228	pDSW204-ftsZ		This study
pDSW230	pDSW208-ftsZ		This study
pDSW233	pDSW208-ftsA		This study
pDSW234	pDSW207-ftsI		This study
pDSW235	pDSW209-ftsI		This study
pDSW242	pDSW208-zipA		This study
pDSW246	pDSW207-III		This study
pDSW247	pDSW207-FFI		This study
pDSW248	pDSW207-FII		This study
pDSW249	pDSW207-IFI		This study
pDSW254	pDSW255-ftsI		This study
pDSW255	Kanr derivative of pDSW207		This study
pDSW259	pBAD18-Kan-sfiA		D. Weiss
pDSW262	pBAD33-ftsI		This study

^aTi, temperature independent (screened for growth at 42°C). 

Media.

Media were NZY, Luria broth (LB), and LB with 0% NaCl. l-Arabinose or d-glucose was added as indicated to modulate expression of genes under control of the P_BAD promoter (27). For localization studies, IPTG was used as follows to modulate expression of gfp fusions under control of modified trc promoters: 2.5 μM for P₂₀₈-ftsZ-gfp and P₂₀₇-gfp-ftsI; 100 μM for P₂₁₀-ftsA-gfp; and 50 μM for P₂₀₈-zipA-gfp. Antibiotics were used for selection at the following concentrations: ampicillin at 200 μg/ml for plasmids and 25 μg/ml for chromosomal alleles; kanamycin at 40 μg/ml; chloramphenicol at 30 μg/ml; tetracycline at 15 μg/ml. FtsI-specific antibiotics were used to induce filamentation at the following concentrations: cephalexin, 10 μg/ml; piperacillin, 2 μg/ml; furazlocillin, 1 μg/ml.

Molecular biological procedures.

Standard procedures for cloning and analysis of DNA, PCR, electroporation, and transformation were used (48). Enzymes used to manipulate DNA were from New England Biolabs (Beverly, Mass.). Oligonucleotides were from Genosys Biotechnologies (Woodlands, Tex.) or 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.

Construction of plasmids for making gene fusions to gfp.

Plasmids for making gene fusions to gfp were based on pTrc99A (Pharmacia, Piscataway, N.J.), a pBR-related vector that confers ampicillin resistance and allows for protein overproduction under control of a strong trc promoter. Transcription from P_trc is regulated by the Lac repressor, supplied by a copy of lacI^q on the plasmid. Downstream of P_trc is a Shine-Dalgarno sequence and a polylinker embedded in an open reading frame to facilitate making translational fusions.

To achieve physiologically appropriate levels of expression, we used site-directed mutagenesis to weaken the trc promoter. One promoter variant, in pDSW204, has a base change in the −35 region: TTGACA→TTTACA. The other variant, in pDSW206, has the mutation in the −35 region and two changes into the −10 region: TATAAT→CATTAT. These mutations were made using the QuikChange site-directed mutagenesis method (Stratagene, La Jolla, Calif.) and the following oligonucleotides (5′→3′; mutations underlined): −35 (top), GGCAAATATTCTGAAATGAGCTGTTTACAATTAATCATCCGG; −35 (bottom), CCGGATGATTAATTGTAAACAGCTCATTTCAGAATATTTG CC; −10 (top), CATCCGGCTCGCATTATGTGTGGAATTGTGAGCG; −10 (bottom), CGCTCACAATTCCACACATAATGCGAGCCGGATG.

To make plasmids that allow fusion of gfp to the N termini of target proteins, a bright allele of gfp was amplified by PCR with pGFPmt2 (16) as template and ACGATCATGAGTAAAGGAGAAGAACTTTTCAC plus CGTGAATTCTTTGTATAGTTCATCCATGCC as primers. The first primer hybridizes to the 5′ end of gfp and contains a BspHI restriction site (underlined) overlapping the start codon. The second primer hybridizes to the 3′ end of gfp and contains an EcoRI site (underlined) immediately in front of the stop codon which is not encoded by the primer. The amplified DNA was digested with BspHI and EcoRI and ligated into pDSW204 and pDSW206 that had been cut with NcoI (compatible with BspHI) and EcoRI to create plasmids pDSW207 and pDSW209.

Plasmids that allow fusion of gfp to the C termini of target proteins were constructed similarly. The primers used to amplify gfp were CCAGCTGCAGATGAGTAAAGGAGAAGAACTTTTC plus CCTGAAGCTTATTTGTATAGTTCATCCATGCC. The first primer hybridizes to the 5′ end of gfp and contains a PstI site immediately preceding the start codon. The second primer hybridizes to the 3′ end of gfp and contains a HindIII site (underlined) directly after the stop codon. The amplified DNA was digested with PstI and HindIII and ligated into the same sites of pDSW204 and pDSW206 to create pDSW208 and pDSW210, respectively.

Construction of gfp fusions. (i) ftsZ-gfp.

First, ftsZ was amplified by PCR with pZAQ (56) as template and CAGACCATGGCAGAACCAATGGAACTTACCAAT and TGGTCTGCAGGTTGTTGTTGTTATCAGCTTGCTTACGCAGGAATG as primers. The amplified product was digested with NcoI and PstI (sites underlined) and ligated into the same sites of pDSW204 to create pDSW228. Second, gfp was obtained on a PstI-HindIII restriction fragment from pDSW208 and ligated into the same sites of pDSW228 to create pDSW230. Inclusion of the NcoI site changes the N terminus of FtsZ from MFE to MAE. The linker sequence is ADNNNLQMS, where AD are the last two residues of FtsZ and MS are the first two residues of GFP.

(ii) ftsA-gfp.

ftsA was amplified by PCR with pZAQ as template and ATGGAATTCATCAAGGCGACGGACAGAAAACTG plus TGGTCTGCAGGTTGTTGTTAAACTCTTTTCGCAGCCAACT as primers. The amplified DNA fragment was digested with EcoRI and PstI (sites underlined), and ligated into the same sites of pDSW210 to create pDSW233. The EcoRI site changes the N terminus of FtsA from MIK to MEFIK. The linker sequence is EFNNNLQMS, where EF are the last two residues of FtsA and MS are the first two residues of GFP.

(iii) zipA-gfp.

zipA was amplified by PCR with pJC3 as template and GACGAATTCTAGTAGTGGCAAGGTGTTAGAACAACAG and CCATATGCATGTTGTTGTTGGCGTTGGCGTCTTTGACTTC as primers. The amplified DNA fragment was digested with EcoRI and NsiI (sites underlined) and ligated into pDSW208 that had been cut with EcoRI and PstI, which is compatible with NsiI, to create pDSW242. Because the 3′-most base of the first primer hybridizes 9 bp upstream of the start codon for zipA, the translational start used in pDSW242 is the native start site for zipA and no amino acid changes have been introduced into the protein. The linker sequence is NANNNMHMS, where NA are the last two residues of ZipA and MS are the first two residues of GFP.

(iv) gfp-ftsI.

ftsI was amplified by PCR with pLMG173 as template and GC ACCATGgaattcAACAACAACAAAGCAGCGGCGAAAACGCAG  and pBAD3′-alt (28) as primers. The amplified DNA was digested with NcoI (site underlined) and HindIII (site present downstream of ftsI in pLMG173) and ligated into the same sites of pBAD24 to create pDSW172. The FtsI protein encoded by pDSW172 has an altered N terminus owing to the incorporation of restriction sites for NcoI, EcoRI (lower case in primer sequence), and three asn residues (MKA→MEFNNNKA). After verifying that this ftsI allele complemented an ftsI23(Ts) mutant, it was obtained from pDSW172 on an EcoRI-HindIII restriction fragment and ligated into the same sites of pDSW207 to create pDSW234. The fusion protein encoded by pDSW234 has the linker sequence YKEFNNNKA, where YK are the last two residues of GFP and KA are the second and third residues of FtsI (i.e., the initiating methionine of FtsI is absent in the fusion protein). Plasmid pDSW254, a kanamycin-resistant derivative of pDSW207, was made as follows. A PstI fragment carrying the kan gene from pUC-4K (Pharmacia) was made blunt ended with T4 DNA polymerase in the presence of all four dNTPs and then ligated into pDSW207 that had been digested with FspI and ScaI. These restriction enzymes each cut pDSW207 once within the bla gene, so the cloning procedure results in replacing most of bla with kan.

(v) Fusions to FtsI swap proteins.

To make gfp-III, the III gene was amplified by PCR with pLD30 as template and GCACCATGgaattcAACAACAACAAAGCAGCGGCGAAAACGCAG and pBAD3′-alt (28) as primers. The amplified DNA was digested with EcoRI (site in lower case) and HindIII (site present downstream of III in pLD30) and ligated into the same sites of pDSW207 to create pDSW246. The fusion protein encoded by pDSW246 has the same linker sequence as that in GFP-FtsI. gfp-IFI, on plasmid pDSW249, also has this linker sequence and was constructed similarly, except that the template for PCR was pLD118. To make gfp-FFI, the FFI gene was amplified by PCR with pLD43 as template and CGAGAATTCAACAACAACATGGATGTCATTAAAAAGAAACATTGGTGGC and pBAD3′-alt as primers. The amplified DNA was digested with EcoRI (site underlined) and HindIII (site present downstream of FFI in pLD43) and ligated into the same sites of pDSW207 to create pDSW247. The fusion protein encoded by pDSW247 has the linker sequence YKEFNNNMD, where YK are the last two residues of GFP and MD are the first two residues of FFI, whose amino terminus is derived from malF. gfp-FII, on plasmid pDSW248, also has this linker sequence and was made similarly, except that the template for PCR was pLD57.

Complementation. (i) Complementation of Ts mutations.

GFP fusion constructs and control plasmids were transformed into a strain carrying a Ts allele of ftsI, selecting for ampicillin resistance on NZY plates at 30°C. Isolates were tested for complementation by streaking onto NZY-ampicillin with 0 to 50 μM IPTG (to modulate expression of the plasmid-borne gfp fusions being tested), and growth was scored after 18 h at 42°C. Identical control plates were incubated at 30°C.

(ii) Complementation of null mutations.

These tests were done in an FtsI depletion strain that has a transposon insertion in the chromosomal ftsI gene and a wild-type copy of ftsI on a plasmid. The transposon confers kanamycin resistance, while the plasmid confers chloramphenicol resistance. GFP fusion constructs in single copies at the lambda att site were transduced with P1 into the depletion strain, selecting for resistance to 25 μg of ampicillin/ml on NZY-kanamycin-chloramphenicol plates that contained 0.2% arabinose to induce the plasmid borne wild-type ftsI. Transductants were made phage free and were tested for complementation by streaking onto NZY-ampicillin-kanamycin-chloramphenicol with 0.2% glucose (to repress the plasmid-borne ftsI) and 0 to 50 μM IPTG (to induce the chromosomal gfp fusion). Plates were incubated at 30, 37, and 42°C, and growth was scored after 18 to 24 h. Control plates contained 0.2% arabinose rather than glucose.

Generation and affinity purification of polyclonal antibodies against FtsI.

Polyclonal antibodies against the periplasmic domain of FtsI were raised in New Zealand White rabbits (Covance, Denver, Pa.). The protein used as antigen was obtained as follows. The periplasmic domain of ftsI (residues 41 to 577) was amplified by PCR with pLMG173 as template and GAGACCATGGCACGCGTAGCGTGGTTACAAG and TTCCGCTCGAGTGCCACAAATTCATTTTTATC as primers. The amplified DNA was digested with NcoI and XhoI (sites underlined) and ligated into the same sites of pET-26b(+) (Novagen, Madison, Wis.) to create pDSW163. This plasmid directs synthesis of cloned inserts under control of a T7lac promoter. In addition, it provides a pelB leader sequence to promote protein export and a C-terminal hexa-histidine tag to facilitate protein purification. Plasmid pDSW163 was transformed into BL21(DE3). After a 2-h induction with 1 mM IPTG, FtsI(peri)-His₆ was overproduced to ∼3% of total cell protein and was found almost exclusively in cytoplasmic inclusion bodies despite the presence of the leader sequence. The insoluble FtsI(peri)-His₆ was purified under denaturing conditions (6 M urea) by nickel affinity chromatography according to instructions in the pET system manual (Novagen). The final preparation was ∼95% pure, and the yield was ∼5 mg/liter of culture. Protein to be used for raising antibodies was dialyzed against water to remove urea, and the precipitated FtsI(peri)-His₆ was recovered and concentrated by centrifugation.

For affinity purification of anti-FtsI antibodies, FtsI(peri)-His₆ was dialyzed into coupling buffer (100 mM NaPO₄ [pH 7.0], 400 mM NaCl, 3 M guanidine), and 2 mg of protein was coupled to 1 ml of AminoLink resin (Pierce Chemical Co., Rockford, Ill.) according to the manufacturer’s instructions. The coupling efficiency was >90%. The column was then equilibrated with Tris-buffered saline (TBS) (25 mM Tris [pH 7.4], 150 mM NaCl, 3 mM KCl). Anti-FtsI antiserum (5 ml) was diluted with 10 ml of TBS and passed three times over the column to allow binding of anti-FtsI antibodies. The column was washed with 10 ml of TBS containing 500 mM NaCl. Bound antibodies were then eluted with 0.1 M glycine (pH 2.5), and 1-ml fractions were collected into tubes containing 0.1 ml of 1 M Tris-HCl (pH 8.0). Peak fractions were identified by immunoblots against purified FtsI(peri)-His₆ that had been spotted onto nitrocellulose. Most of the anti-FtsI activity eluted in fraction 2. This fraction was then passed over a 1-ml column of immobilized E. coli lysate (Pierce Chemical Co.) to remove antibodies that recognize soluble E. coli proteins. Fractions of 0.5 ml were collected and analyzed for protein by Bradford assay (Bio-Rad, Hercules, Calif.). Fractions 2 and 3 contained 75% of the input protein and were pooled and dialyzed against 10 mM NaPO₄ (pH 7.0)–250 mM NaCl. Anti-FtsI antibodies were then concentrated threefold to a volume of ∼300 μl by ultrafiltration in a Centricon 30 (Amicon, Beverly, Mass.). Purified anti-FtsI was stored at −20°C at a concentration of 0.7 mg of immunoglobulin G (IgG)/ml in 10 mM NaPO₄ (pH 7.0), 250 mM NaCl, 10 mg of bovine serum albumin/ml, and 50% glycerol.

Growth and processing of cells for protein localization experiments.

To prepare cells for protein localization using GFP, an overnight culture grown in NZY containing 25 μg of ampicillin/ml was diluted 1:2,000 into a 250-ml flask containing 20 ml of LB plus IPTG at the concentrations indicated above to induce expression of gfp fusions. When the optical density at 600 nm (OD₆₀₀) reached ∼0.2 (3 to 4 h), cells were fixed directly in growth medium and processed for microscopy as described below. For experiments involving temperature-sensitive mutants, the cultures were grown at 30°C to an OD₆₀₀ of ∼0.2, at which time a sample was fixed (T = 0), and 4 ml was transferred to a flask containing 16 ml of LB (with 0% NaCl) and IPTG at 42°C. Additional samples were fixed after 30, 45, and 60 min of incubation at 42°C. For experiments involving the FtsI-specific β-lactam antibiotics cephalexin, piperacillin, and furazlocillin, the overnight culture was grown in NZY containing 40 μg of kanamycin/ml. This culture was diluted 1:2,000 into LB and IPTG and was grown at 37°C to an OD₆₀₀ of ∼0.2. A sample was fixed (no drug control), and 4 ml was transferred to a flask containing 16 ml of LB plus IPTG and the desired antibiotic. Samples were fixed after 30, 45, 60, and 75 min of incubation. For experiments involving induction of sfiA, an inhibitor of FtsZ ring formation, the overnight culture was grown in NZY containing 40 μg of kanamycin/ml. This culture was diluted 1:4,000 into LB containing kanamycin and IPTG and was grown to an OD₆₀₀ of ∼0.05 (2.5 h) at which time arabinose or glucose was added to a final concentration of 0.2% to induce or repress expression of sfiA, respectively. Samples were fixed 60 min after addition of the sugar. For experiments involving depletion strains, the overnight culture was grown in NZY containing 40 μg of kanamycin/ml, 30 μg of chloramphenicol/ml, 25 μg of ampicillin/ml, and 0.04% arabinose. This culture was diluted 1:50 into 20 ml of NZY containing the same antibiotics, 0.01% arabinose, and IPTG and was grown at 30°C to an OD₆₀₀ of ∼0.2 (3 h). Cells were washed once with NZY and inoculated 1:100 into the same medium with either 0.01% arabinose or 0.2% glucose. Samples were fixed when the OD₆₀₀ reached ∼0.1 (4 h).

Cells were fixed by a procedure based on that previously described for IFM (47). A 500-μl sample of cells in exponential growth was added directly to a microfuge tube containing 20 μl of 1 M NaPO₄ (pH 7.4), 100 μl of 16% paraformaldehyde, and 0.4 μl of 25% glutaraldehyde. Fixation was for 15 min at room temperature followed by 30 min on ice. Fixed cells were washed twice by centrifugation and resuspension in 1 ml of phosphate-buffered saline (PBS) (10 mM NaPO₄ [pH 7.4], 150 mM NaCl, and 15 mM KCl), then resuspended in PBS at an OD₆₀₀ of ∼1.0. Cells (10 μl) were applied to 15-well multitest slides (ICN Biochemicals, Aurora, Ohio) that had been pretreated with poly-l-lysine (Sigma, St. Louis, Mo.). After 10 min to allow cells to absorb to the slide, wells were washed three times with 10 μl of PBS and allowed to air-dry for ∼1 min. Cells were rehydrated with 10 μl of PBS, the PBS was removed, and 10 μl of PBS containing 0.2 μg of 4′,6-diamidino-2-phenylindole (DAPI) (Sigma)/ml was applied. After a 15-min incubation in the dark, cells were washed twice with 10 μl of PBS and mounted in PBS containing 50% glycerol. Slides could be stored in the dark at −20°C for >1 week without loss of GFP signal.

A sample of whole cells was harvested at the end of each experiment to verify expression of the gfp fusions by Western blot analysis: cells from 2 ml of culture, typically at an OD₆₀₀ of ∼0.3, were pelleted by centrifugation and resuspended in 100 μl of sodium dodecyl sulfate-polyacrylamide gel electrophoresis gel-loading buffer. Samples were boiled for 5 min prior to electrophoresis.

Growth was monitored throughout each experiment by OD₆₀₀. We observed a marked reduction in growth, apparently due to lysis, about an hour after cultures were shifted to conditions that induce filamentation (e.g., the presence of FtsI-specific β-lactams or ftsA, ftsZ, ftsI, and ftsQ temperature-sensitive mutants at 42°C).

Cells were processed for IFM as described (57), except that we used a new anti-FtsI antibody (described above) at a dilution of 1:4,000.

Microscopy.

Fluorescence micrographs were recorded on an Olympus BX60 microscope equipped with a 100× UPlanApo objective. Three filter sets were used: a U-MWIBA filter for GFP and fluorescein, a U-MNUA filter for DAPI, and a U-MWG filter for propidium iodide. Typical exposure times were 2 to 4 sec for GFP, 10 sec for fluorescein, and 0.1 sec for DAPI and propidium iodide Images were captured using a cooled charge-couple device camera (Princeton Instruments) and a personal computer with MetaMorph software version 3.0 (Universal Imaging Corp.). Images were processed using Adobe Photoshop 4.0.

Scoring of fluorescence micrographs.

Cells and filaments were measured and scored for the presence or absence of an FtsI ring(s). Rings were identified as fluorescent bands that extended at least halfway across the diameter of the cell. Small spots of fluorescence at the middle of nonconstricting cells were not scored as FtsI rings, although they probably indicate an early stage in ring formation. In highly constricted cells bands that extended across the cell appeared as a spot but were scored as FtsI rings since they extended across the diameter of the rod at that point. The length of cells and filaments was measured using NIH Image software (developed at the U.S. National Institutes of Health and available on the Internet at http://rsb.nih.gov/nih-image/) to compare fluorescence images to a calibration standard. Only cells and filaments that were entirely within the image were scored, although this introduced a slight bias against longer filaments. Finally, only filaments with regularly spaced nucleoids (as judged by DAPI staining) were scored. Severe nucleoid segregation defects were typically observed in ∼10% of the filaments, although sometimes the figure was as high as 60%. None of the Fts proteins localized in such filaments, which we suspected were dead.

Western blotting.

Samples typically contained the equivalent of 200 μl of culture at an OD₆₀₀ of 0.3. Proteins were separated on sodium dodecyl sulfate–10% polyacrylamide gels and transferred to nitrocellulose with a semi-dry blotter (Bio-Rad). Anti-FtsI was used at a dilution of 1:80,000. The secondary antibody was goat anti-rabbit IgG conjugated to horseradish peroxidase and diluted 1:25,000, and detection was performed with SuperSignal chemiluminescent substrate (both from Pierce Chemical Co.) and exposure to X-OMAT film (Kodak, Rochester, N.Y.). Densitometry of film was done with a Fluor-S MultiImager system (Bio-Rad).

RESULTS

A fusion of gfp to ftsI.

To further our studies of FtsI localization, we constructed a gfp-ftsI translational fusion that places a bright variant of GFP (16) at the amino terminus (cytoplasmic domain) of FtsI. We then integrated the gfp-ftsI fusion into the chromosome at the λ attachment site (attB). In addition to gfp-ftsI at attB, a wild-type copy of ftsI was present at the normal chromosomal locus. Expression of gfp-ftsI was driven from an IPTG-inducible promoter.

In cells grown in rich media (NZY or LB) containing 2.5 μM IPTG, our standard growth conditions for studying localization, the steady-state level of GFP-FtsI was similar to that of wild-type FtsI as judged by Western blotting (Fig. 1). When these cells were fixed and examined by fluorescence microscopy, about half of them exhibited a bright band of fluorescence at their midpoints, indicating that GFP-FtsI localizes to the division site (Fig. 2 and Table 2). The frequency of localization of GFP-FtsI to the midcell was mildly temperature dependent, being observed in 59% ± 5% of the cells at 30°C (n = 6 independent experiments in which at least 200 cells were scored), 50% ± 5% at 37°C (n = 9), and 44% ± 4% at 42°C (n = 3). GFP is known to fold better at lower temperatures (31), but we have not investigated whether this is the basis of the temperature sensitivity that we observed. Localization was not observed in cells expressing gfp alone (Table 2) or when GFP was fused to MalF, a membrane protein involved in maltose transport (data not shown).

FIG. 1.

Steady-state levels of FtsI and GFP-FtsI fusion proteins as determined by Western blotting with an anti-FtsI antibody. Molecular mass standards are indicated to the left of the blot, and the positions of FtsI and GFP-FtsI are shown to the right. The GFP fusion protein produced is indicated above each lane. The faint band at the position of GFP-FtsI in the first lane is due to cross-reaction of the primary antibody with a protein of unknown identity. The intensity of this band does not respond to IPTG. In contrast, the proteins identified as GFP-FtsI respond to IPTG and are also detected with anti-GFP antibodies (not shown). The strains used were MC4100, EC436, EC505, EC507, EC509, and EC511.

FIG. 2.

Localization of GFP-FtsI, GFP-III, and GFP-FFI in cells fixed during exponential growth. Arrows indicate examples of FtsI rings at division sites. These are very faint in the case of GFP-III and not detected in GFP-FFI. In the case of GFP-FtsI, 10 of the 13 cells shown were judged to have an FtsI ring at the division site. The strains used were EC436, EC505, and EC507.

TABLE 2.

Localization frequencies of and complementation by GFP-FtsI and GFP-FtsI swap proteins

Construct	Structurea	Localization frequencyb		Complementation
		Cells scored (n)	Band at midcell (%)	Tsc	Nulld
GFP-FtsI		421	61	+++	+++
GFP-III		419	6	++	+
GFP-FFI		461	<1	−	ND
GFP-FII		382	<1	−	ND
GFP-IFI		400	<1	−	ND
GFP alone		421	0	−	−

^aOpen boxes represent domains of FtsI. Hatched boxes represent domains derived from MalF. Stippled boxes denote GFP. E and M indicate borders of the transmembrane domain where restriction sites for EagI and MscI, respectively, have been introduced. The former changes a Trp to Gly and the latter changes an Arg to His. Diagrams of fusion proteins are not to scale. The cytoplasmic and transmembrane domains of FtsI contain 23 and 17 amino acids, respectively, while those from MalF contain 16 and 20 amino acids. The periplasmic domain of FtsI contains 537 amino acids. 

^bStrains used were EC436, EC505, EC507, EC509, EC511, and EC452. Fluorescent bands for GFP-III were faint. Symbols: −, no growth; ++, fair growth; +++, good growth. 

^cStrains used were LMG64 transformed with pDSW234, pDSW246, pDSW247, pDSW248, pDSW249, and pDSW207. 

^dStrains used were EC550, EC552, EC556. ND, not determined. Symbols: −, no growth; +, poor growth; +++, good growth. 

Previously we reported that 10 to 20% of cells have polar FtsI as determined by IFM. With GFP as a reporter, <2% of cells had polar localization when induced with 2.5 μM IPTG, and these polar localization signals were very weak compared to midcell fluorescence. When cells were grown in the presence of 20 μM IPTG, septal localization was still observed in about half of the cells, but 5 to 10% exhibited strong polar localization, and foci of fluorescence were also observed at a variety of other locations (data not shown). The physiological significance of these signals is, however, doubtful since GFP-FtsI is overproduced about 10-fold at 20 μM IPTG. At higher levels of expression, background fluorescence obscured visualization of GFP-FtsI at the division site.

As suggested by the normal appearance of the cells in Fig. 2, gfp-ftsI did not seriously perturb cell division when induced with 2.5 μM IPTG in a merodiploid. We therefore tested whether gfp-ftsI complemented a null mutation in ftsI. To do this, we used an ftsI depletion strain, EC548, in which the chromosomal copy of ftsI has been inactivated by a transposon insertion. EC548 is kept alive by a plasmid-borne wild-type ftsI allele whose expression is under control of an arabinose-dependent araBAD promoter. We transduced gfp-ftsI into EC548 to create EC552. Growth of EC552 was no longer arabinose-dependent, indicating that gfp-ftsI complements the transposon insertion (Table 2). Moreover, growth was not IPTG-dependent either, indicating that the basal level of expression of gfp-ftsI from the P₂₀₇ promoter is sufficient. This level of expression appears to be equivalent to the normal expression level of ftsI as judged from Western blots of merodiploids grown in liquid media (data not shown).

We considered the possibility that the complementation observed was due not to GFP-FtsI per se but to FtsI released from the fusion protein by proteolysis. Western blotting revealed a small amount of normal-length FtsI in the depletion strain, about 5% the level of FtsI in a wild-type strain (not shown). Whether the small amount of proper-length FtsI is sufficient to support division is not known, but depletion experiments indicate that cells filament when they have 20% the normal level of FtsI (57). Moreover, if GFP-FtsI were nonfunctional, we would expect it to interfere with division. We think this because GFP-FtsI localizes to the division site and because Broome-Smith et al. (13) have shown that a catalytically inactive mutant protein (serine 307→cysteine) is a dominant negative. Thus, the simplest interpretation of our complementation experiment is that GFP-FtsI supports division.

Localization of FtsI with altered membrane anchors.

FtsI consists of a small cytoplasmic domain, a single membrane-spanning segment, and a large periplasmic domain that encodes the transpeptidase activity involved in peptidoglycan synthesis (reviewed in reference 44). Replacement of the membrane anchor (i.e., cytoplasmic domain plus membrane-spanning segment) with a cleavable signal sequence from OmpA results in production of the periplasmic domain of FtsI as a soluble protein in the periplasm (22). Although this soluble form of FtsI is catalytically active, it does not complement an ftsI(Ts) mutation, indicating that the membrane anchor is important for cell division (25).

Previously, we investigated whether the cytoplasmic domain and membrane-spanning segment serve only to tether the periplasmic domain to the membrane or whether these small domains have a more sophisticated function (28). To do this, we introduced restriction sites at either end of the membrane-spanning segment and used these sites to replace the cytoplasmic domain and membrane-spanning segment with analogous parts of other membrane proteins. We found that these so-called “swap” proteins inserted into the membrane and retained transpeptidase catalytic activity as determined by a penicillin-binding assay. Nevertheless, they failed to support cell division. We speculated that the swap proteins might be defective in localization to the division site.

To test this hypothesis, we fused GFP to several swap proteins: III, FFI, FII, and IFI. The three-letter names indicate the source of each domain in the swap proteins. For example, III refers to a protein in which all three domains—cytoplasmic, membrane-spanning, and periplasmic—are derived from FtsI. This protein is encoded by the ftsI allele that carries restriction sites flanking the membrane-spanning segment. These restriction sites introduced a single amino acid change at each end of the membrane-spanning segment. Although the III gene is not fully wild type, it complemented an ftsI null mutant (transposon insertion) in our previous study (but see below). FFI, FII, and IFI refer to proteins which carry a cytoplasmic domain and/or membrane-spanning segment from MalF, a maltose transport protein with no role in cell division. None of these swap alleles complemented an ftsI null mutant.

To test for localization of the swap proteins, cells expressing GFP fusions to the swap genes were fixed and examined by fluorescence microscopy. The GFP-III protein localized, albeit poorly, while GFP-FFI, GFP-FII, and GFP-IFI did not appear to localize at all (Fig. 2 and Table 2). Western blotting indicated that the stability and abundance of each swap protein was similar to that of GFP-FtsI (Fig. 1). Because the III protein has a localization defect, we conclude that the sequence of the borders of the membrane spanning segment are important for localization. Because the FII and IFI proteins have a more severe localization defect, the cytoplasmic domain and/or membrane spanning segment may also be important for localization (see Discussion).

The localization defect observed for GFP-III was not expected given our previous report that the III protein supports cell division. In those studies, we scored complementation by the III allele as somewhat weaker than that of the wild type. Both genes were expressed at high levels from pBR-derived plasmids with an estimated copy number of 30 to 50 per cell. In contrast, for localization studies we expressed gfp-III at physiological levels from a single-copy chromosomal gene. Besides expression level, the other obvious difference is the presence of the GFP tag, but this does not appear to disturb wild-type FtsI, making it unlikely that GFP is the cause of the defect in the GFP-III protein.

Consistent with our previous report, we found that plasmid-borne gfp-III complemented an ftsI23(Ts) mutation almost as well as gfp-ftsI, while gfp-FFI, gfp-FII, and gfp-IFI did not complement at all. To assay gfp-III for complementation under conditions similar to those used for localization, we transduced gfp-III into the ftsI depletion strain to create EC556. Growth of EC556 was not dependent upon arabinose, indicating that gfp-III complements the transposon insertion (Table 2). But complementation was poor compared to that by gfp-ftsI, especially at 42°C. Below 50 μM IPTG, the colonies were rough and flat. Cells from these colonies examined under the microscope were filamentous. At 50 μM IPTG, colony morphology was normal, and only a few of the cells were long filaments. In contrast, EC552, which expresses gfp-ftsI, exhibited normal colony and cell morphology throughout the range of 0 to 50 μM IPTG. These observations are consistent with the III protein having a localization defect that can be partly relieved by overproduction.

Localization of GFP-FtsI in the presence of FtsI-specific β-lactams.

Several β-lactam antibiotics, among them cephalexin, furazlocillin, and piperacillin, cause E. coli cells to filament. These antibiotics specifically inactivate FtsI by acylation of serine 307 in the transpeptidase catalytic site (24, 45). To test whether FtsI needs its transpeptidase activity to localize to a division site, we examined the effect of these β-lactams on localization of GFP-FtsI (Fig. 3A and B).

FIG. 3.

Localization of FtsI in filaments formed by treating cells with FtsI-specific β-lactam antibiotics. Upper panels show strain EC522, expressing gfp-ftsI, after 75 min in the presence of cephalexin. (A) GFP. (B) DAPI. Lower panels show FtsI detected by IFM of strain MC4100 after 45 min in the presence of furazlocillin. Arrows indicate FtsI localized to potential division sites. (C) Fluorescein. (D) Propidium iodide.

Addition of any of the three above-mentioned antibiotics prevented septation but did not prevent localization of GFP-FtsI to potential division sites (Fig. 3A and B and data not shown). After 75 min in the presence of these antibiotics, cells were about 18 μm long, and >97% had GFP-FtsI localized to at least one potential division site, while at least 10% had three or more GFP-FtsI bands. When normalized to cell length, the frequency of GFP-FtsI localization in the filamentous population decreased only about twofold to a spacing of about one ring per 10 μm. This spacing is similar to that of FtsZ rings as determined by IFM of filaments obtained by treating cells with cephalexin (46). As shown below, localization of GFP-FtsI requires a Z ring. We therefore conclude that the mild reduction of GFP-FtsI localization observed in filaments is a secondary consequence of delaying Z-ring assembly.

We also used IFM to localize FtsI in furazlocillin-treated cells (Fig. 3C and D). In some filaments, FtsI was localized to potential division sites, but in other filaments it was not detected. In addition, fluorescence was sometimes observed at various locations in the filaments. We have seen similar cell-to-cell variation in cultures that have not been treated with an antibiotic; indeed, this sort of inconsistency is one reason we now rely primarily on fusions to GFP when studying localization of FtsI.

The simplest interpretation of these results is that FtsI localizes to potential division sites in the presence of β-lactams, but our IFM procedures do not detect FtsI very consistently. An alternative which we cannot exclude is that β-lactams diminish localization of FtsI but not of GFP-FtsI. This would be the case if, for example, GFP-FtsI does not react with any of the three β-lactams tested, although this seems unlikely because GFP-FtsI does not confer resistance to cephalexin or piperacillin (not shown).

We verified by Western blotting that none of the antibiotic treatments affected the level of FtsI or GFP-FtsI (reference 46 and data not shown).

Dependency of localization on FtsZ, FtsA, FtsQ, and FtsL.

To determine whether localization of FtsI to the division site depends on the activities of three other essential division proteins—FtsZ, FtsA, and FtsQ—we transduced Ts mutations in each of these proteins into a strain carrying gfp-ftsI at attB and a wild-type copy of ftsI at the normal chromosomal locus. The ftsZ84(Ts) allele encodes a protein that fails to make Z rings at the nonpermissive temperature (46), although rings form within minutes upon return to the permissive temperature (4). The ftsA12(Ts) allele encodes a protein that fails to localize to the division site at the nonpermissive temperature (5). The localization behavior of the protein encoded by ftsQ1(Ts) has not been reported. Our findings with the Ts mutants are presented in Fig. 4 and Table 3.

FIG. 4.

Dependence of localization of GFP-FtsI on FtsZ, FtsA, FtsQ, and FtsL. In each case, one or more filaments is shown together with nonfilamenting cells that do exhibit GFP-FtsI bands and serve as internal controls for microscopy and subsequent image processing. These latter are the short cells in the micrographs. Filaments are as follows: EC458 [ftsZ(Ts)] after 60 min at 42°C; EC436/pDSW259 (sfiA induced) 60 min after addition of arabinose; EC455 [ftsA(Ts)] after 45 min at 42°C; EC538 (FtsQ depletion) in glucose; EC607 (FtsL depletion) in glucose. Controls were EC436 after 60 min at 42°C, EC436/pDSW259 60 min after addition of glucose, EC436 after 45 min at 42°C, EC538 in arabinose, and EC607 in arabinose.

TABLE 3.

Localization of GFP-FtsI rings in fts temperature-sensitive mutants

Mutanta	Growth temperature (duration)	No. of cells scored	Cell length (avg ± SD) (μm)	Total no. of FtsI rings	% of cells with ring(s)	Spacing of FtsI ringsc (μm of cell length/ring)
Wild type	30°C	381	4.1 ± 1.0	221	58	7.1
	42°C (60 min)	338	4.5 ± 1.4	146	43	10
ftsZ(Ts)	30°C	265	5.4 ± 1.8	112	42	13
	42°C (60 min)	127	23.9 ± 6.2	4b	3	760
ftsA(Ts)	30°C	224	4.9 ± 1.2	128	57	8.6
	42°C (60 min)	96	24.4 ± 5.1	2b	2	1,200
ftsQ(Ts)	30°C	116	9.6 ± 5.1	20b	17	56
	42°C (45 min)	110	22.8 ± 7.8	5b	5	500

^aStrains used were EC454 for the wild type, EC458 for ftsZ84(Ts), EC455 for ftsA12(Ts), and EC457 for ftsQ1(Ts). 

^bFaint. 

^cTotal length of cells (or filaments) divided by total number of FtsI rings. 

In the ftsZ84(Ts) background growing at 30°C, cells were only slightly longer than the wild type, and just over 40% had a prominent band of GFP-FtsI at the division site compared to nearly 60% in the wild type. Upon shifting to 42°C, cells became filamentous, and GFP-FtsI localization was rarely observed. When normalized to cell length, the frequency of GFP-FtsI localization was about 100-fold lower after an hour at the nonpermissive temperature. This is probably an underestimate of the dependency of FtsI localization on FtsZ, as the few rings scored at the nonpermissive temperature were quite faint and might reflect leakiness of the temperature sensitive allele.

We also found that GFP-FtsI failed to localize when Z-ring formation was inhibited by overproduction of SfiA (also called SulA). SfiA is ordinarily induced as part of the SOS response to DNA damage and binds directly to FtsZ, blocking the assembly of Z-rings (reviewed in reference 37). To test the effect of SfiA on localization of GFP-FtsI, we cloned sfiA into pBAD18-kan (27), an expression vector with an arabinose-inducible promoter. Cells transformed with pDSW259 (= pBAD18-kan carrying sfiA) filamented upon addition of arabinose to the culture; after an hour they averaged 15.8 ± 4.1 μm in length, and only 2% had a (faint) GFP-FtsI band (Fig. 4). In contrast, uninduced cells were 3.8 ± 0.8 μm long, and >50% exhibited localization of GFP-FtsI to the midcell. As a further control, we showed that arabinose did not affect cell length or GFP-FtsI localization when cells were transformed with empty vector, pBAD18-kan (data not shown). Finally, we verified that induction of sfiA blocked assembly of Z rings (data not shown). For this experiment, we transformed pDSW259 into a strain that carries an ftsZ-gfp fusion.

In the ftsA12(Ts) background, cells were relatively normal in length at 30°C and close to 60% had a prominent band of GFP-FtsI at the division site. At the nonpermissive temperature, cells filamented and only 2 to 3% of the filaments had an FtsI ring. When normalized to cell length, the frequency of localization was about 100-fold lower in the Ts mutant at the nonpermissive temperature. The occasional localization signals scored in the filaments were very faint and likely reflected leakiness of the ftsA12(Ts) allele.

Although the doubling time of our ftsQ1(Ts) strain at 30°C was similar to that of the wild type, the cells were markedly filamentous. Apparently, the ftsQ1(Ts) allele is somewhat defective even at the permissive temperature. Fewer than 20% of the cells fixed at 30°C exhibited localization of GFP-FtsI, and these localization signals were often quite faint, suggesting that localization of FtsI depends upon FtsQ. Moreover, localization decreased in cells grown at 42°C, and was down about 50-fold as compared to the wild type after 45 min at the nonpermissive temperature. It was difficult to obtain data at later times after the temperature shift because growth (increase in OD₆₀₀) was poor and the filaments were mostly unhealthy as judged by defects in nucleoid segregation. We also observed dependency of GFP-FtsI localization on FtsQ in an FtsQ-depletion strain, EC538. This strain has a transposon insertion in the chromosomal ftsQ allele; cell division depends upon a plasmid-borne wild-type ftsQ gene that is under the control of an arabinose-dependent promoter. When grown in the presence of arabinose, cells appeared normal and about 50% had GFP-FtsI localized to the midcell. Upon shift to glucose, cells filamented and localization of GFP-FtsI was rarely observed (Fig. 4; Table 4).

TABLE 4.

Localization of GFP-FtsI in FtsQ and FtsL depletion strains

Straina	Sugarb	No. of cells scored	Cell length (avg ± SD) (μm)	No. of FtsI rings	% of cells with ring(s)	Spacing of FtsI ringsd (μm)
FtsQ depletion	Arabinose	194	4.2 ± 1.0	98	51	8.4
	Glucose	123	14.7 ± 5.6	2c	2	906
FtsL depletion	Arabinose	247	5.4 ± 1.5	156	63	8.5
	Glucose	130	20.7 ± 7.9	4c	3	672

^aStrains used were EC538 for FtsQ depletion and EC607 for FtsL depletion. 

^bArabinose = inducing condition; glucose = depletion condition. 

^cFaint. 

^dTotal length of cells (or filaments) divided by total number of FtsI rings. 

Since we have no ftsL(Ts) allele, we used an FtsL depletion strain, EC607, to investigate whether FtsI localization depends upon the presence of FtsL. When the depletion strain was grown in the presence of the inducer arabinose, cells were relatively normal in length and >50% had GFP-FtsI localized to the midcell. In the presence of glucose, cells were filamentous and localization of GFP-FtsI was essentially not observed (Fig. 4; Table 4). We conclude that FtsI requires FtsL to localize to the division site.

None of the above treatments (temperature shift, induction of sfiA, depletion of FtsQ and FtsL) changed the level or stability of FtsI or GFP-FtsI as judged by Western blotting (data not shown).

Localization of other division proteins in an ftsI(Ts) mutant.

We fused gfp to ftsZ, ftsA, zipA, ftsQ, and ftsL, and placed these fusions on the chromosome using lambda InCh. The fusions to ftsQ and ftsL and their behavior in an ftsI23(Ts) mutant will be described elsewhere (15, 23). The fusions to ftsZ, ftsA, and zipA are similar to those previously published (29, 38). The fusion to ftsZ complements the ftsZ84(Ts) allele, but the fusion to ftsA does not complement the ftsA12(Ts) allele (data not shown). The fusion to zipA and was not tested for complementation.

To determine whether localization of any of these proteins depends upon FtsI, we transduced the ftsI23(Ts) allele into strains that express each of these gfp fusions and compared localization at the permissive and nonpermissive temperatures. Confirming results obtained with IFM, FtsZ (2, 46) and FtsA (5) localized well at 30°C and 42°C, implying that localization of these proteins to the division site is independent of FtsI (Table 5). The ZipA-GFP protein also localized well at both temperatures (Fig. 5; Table 5). We conclude that ZipA does not require FtsI to localize to the site of septation.

TABLE 5.

Localization frequencies of FtsZ-GFP, FtsA-GFP, and ZipA-GFP in wild-type and ftsI23(Ts) mutant backgrounds

GFP fusion	Genetic background and conditionsa	No. of cells scored	Cell length (avg ± SD) (μm)	% of cells with the indicated no. of rings:							Spacing of ringsc (μm)
				0	1	2	3	4	5	>5
FtsZ-GFP	Wild type, 30°C	158	3.4 ± 0.9	10	90b	0	0	0	0	0	3.8
	Wild type, 42°C	133	2.7 ± 0.7	6	94	0	0	0	0	0	2.8
	ftsI23(Ts), 30°C	128	5.7 ± 1.8	2	93	5	1	0	0	0	5.5
	ftsI23(Ts), 42°C	87	26.8 ± 6.7	0	2	17	28	33	14	6	7.5d
FtsA-GFP	Wild type, 30°C	115	3.1 ± 0.9	9	91	0	0	0	0	0	3.4
	Wild type, 42°C	143	2.5 ± 0.6	15	85	0	0	0	0	0	3.0
	ftsI23(Ts), 30°C	129	6.9 ± 3.2	2	88	8	2	1	0	0	6.7
	ftsI23(Ts), 42°C	95	28.9 ± 8.4	1	2	16	16	33	17	16	7.3
ZipA-GFP	Wild type, 30°C	205	3.1 ± 0.8	18	82	0	0	0	0	0	3.8
	Wild type, 42°C	244	2.6 ± 0.7	11	89	0	0	0	0	0	3.0
	ftsI23(Ts), 30°C	234	4.5 ± 1.3	11	89	0	0	0	0	0	5.0
	ftsI23(Ts), 42°C	95	21.3 ± 6.0	0	3	29	27	31	9	0	6.8

^aStrains used were EC484, EC486, EC479, EC481, EC489, and EC491. Cultures were grown at 30°C and then shifted to 42°C for 60 min. 

^bValues exceeded 95% in other experiments in which ftsZ-gfp was more highly expressed. 

^cTotal length of cells (or filaments) divided by total number of FtsZ, FtsA, or ZipA rings. 

^dThe spacing of FtsZ rings does not increase as much at higher levels of induction. 

FIG. 5.

Localization of ZipA-GFP in an ftsI23(Ts) mutant background (EC491). Cells were grown at 30°C prior to fixation (A) and were shifted to 42°C for 60 min prior to fixation (B).

DISCUSSION

Little is known about the mechanism(s) by which bacterial proteins are localized to discrete subcellular sites such as the septum or the pole(s). In a couple of instances of polar localization, proteolysis plays an important role by clearing the target protein from the rest of the cell (6, 20, 49), but what protects the target protein from proteolysis at the pole(s) is not yet known. At least two proteins that localize to the division site, FtsA and ZipA, probably do so by binding directly to another protein, FtsZ, that is already present at that site; how FtsZ finds the division site remains to be determined (18, 29, 39, 54). There are several additional instances, most related to cell division, in which localization of a protein depends upon (prior?) localization of another protein(s) (3, 36, 53, 59). In these cases, it is not yet clear whether the interactions are direct.

Here we have used a fusion of gfp to ftsI to study targeting of FtsI to the division site of E. coli. gfp-ftsI was integrated into the chromosome with the aid of a newly constructed lambda vector, lambda InCh, that greatly simplifies this process. gfp-ftsI was expressed at physiologically appropriate levels under control of an IPTG-regulatable promoter. Although gfp-ftsI complemented a null mutation in ftsI, most of our studies were done in merodiploids that also had a wild-type copy of ftsI in the normal chromosomal location (the 2-min region).

Confirming previous results obtained by IFM (53, 57), GFP-FtsI localized to the division site in about half of the cells growing in rich medium, with a doubling time of 25 to 30 min. Strong polar localization of GFP-FtsI was not observed unless the protein was overproduced 10-fold relative to wild-type FtsI, suggesting that our previous observation of polar FtsI with IFM was an artifact. Polar FtsI was not reported by Wang et al. (53), who used IFM.

FtsI’s membrane anchor is important for septal localization.

How are the membrane proteins involved in septum assembly targeted to the division site? Recent progress argues against the naive hope that these proteins might share a targeting motif. No such motif is apparent from sequence comparisons. More convincingly, septal localization depends on the cytoplasmic domain in the case of ZipA, the periplasmic domain of in the case of FtsN, and the membrane anchor in the case of FtsK (3, 29, 59).

As a first step towards defining sequences in FtsI that are involved in targeting the protein to the division site, we fused GFP to several swap proteins in which the cytoplasmic domain and/or membrane spanning segment of FtsI had been replaced with analogous parts of MalF, a maltose transport protein with no role in cell division (28). Interestingly, none of the swap proteins localized to the division site. We draw three conclusions from this finding.

(i) Failure of the FFI, FII, and IFI proteins to localize plausibly accounts for their failure to complement null mutations in ftsI. Accounting for the complementation defect is an issue because all three swap proteins appear to insert into the membrane with the proper topology and retain penicillin-binding activity (28).

(ii) The membrane anchor (i.e., the cytoplasmic domain plus the membrane spanning segment) has a role in targeting FtsI to the division site.

(iii) The poor localization and complementation exhibited by the GFP-III protein, which differs from GFP-FtsI by just one amino acid at each end of the transmembrane domain, indicates that the borders of the membrane-spanning segment are important for FtsI function.

With respect to the role of the membrane anchor in septal localization, the simplest interpretation of our findings is that both the cytoplasmic domain and the transmembrane segment are required. However, preliminary results from our lab indicate that most of the cytoplasmic domain can be deleted without loss of ftsI function in cell division; only deletions that approach the presumed border between the cytoplasmic domain and membrane-spanning segment fail to complement a Ts mutant when expressed from plasmids. Thus, it is possible that our FII swap affects both the cytoplasmic domain and the membrane-spanning segment.

The membrane-spanning segment of FtsI appears to be unusually short, with only 17 amino acids separating the Arg residues (arbitrarily) defined as the ends of the segment. Lengthening the membrane-spanning segment by just one residue abolishes FtsI function in cell division (28). Moreover, the short length of the transmembrane domain, but not its sequence, is conserved in FtsI proteins from other organisms (data not shown). In contrast, the transmembrane domains of other bitopic cell division proteins in E. coli appear to range in length from 20 (FtsL) to 28 (FtsN) residues, while E. coli PBP2, which appears to be the primary transpeptidase for elongation, is predicted to have a transmembrane domain of 25 amino acids (7, 17, 26). These observations suggest that the length of the transmembrane domain of FtsI is important.

What might be the mechanism by which FtsI’s short membrane-spanning segment directs the protein to the septum? Perhaps the cytoplasmic membrane is thin at the division site. In this case, the short transmembrane domain would function directly as a localization signal, and recruitment of FtsI to the division site might depend upon membrane alterations induced by other division proteins that localize to this site prior to FtsI (Fig. 6) (see below). Interestingly, several bitopic membrane proteins found in the eukaryotic Golgi apparatus have short transmembrane domains (∼17 amino acids) that serve as targeting signals (41). Golgi membranes have little cholesterol and are therefore thinner than cytoplasmic membranes in eukaryotes. An alternative possibility is that accommodating the short membrane-spanning segment in the lipid bilayer forces the periplasmic domain into a precise orientation that is necessary for its assembly into a protein complex. For example, Höltje and coworkers (32) have evidence for a protein complex involving FtsI and several other enzymes involved in peptidoglycan metabolism. If these interactions (or others that remain to be described) are important for septal localization, then the primary localization determinants in FtsI would be sequences in the periplasmic domain. Although these ideas are speculative, they make testable predictions about the behavior of certain mutants.

FIG. 6.

Dependency relationships for septal localization of several E. coli division proteins. The first event is polymerization of FtsZ into the Z ring at the future division sites. The other proteins then localize to that site in the order indicated. The positions of ZipA, FtsK, and FtsW are only partially established. Unpublished work from our lab indicates that ZipA localization requires a Z ring but not FtsA. FtsK requires FtsA but not FtsI. FtsW is known to localize, but what happens in fts mutant backgrounds is not known. The model is based on our own results in this paper, references 15, 23, 46, and 57, and reports from several other labs (2, 3, 5, 38, 53, 59).

The transpeptidase catalytic activity might not be required for septal localization.

We found that GFP-FtsI localizes to potential division sites in the presence of three different β-lactam antibiotics that specifically inhibit the transpeptidase activity of FtsI—cephalexin, piperacillin, and furazlocillin. GFP-FtsI was localized in nearly 100% of the filaments, and the localization signals were quite strong. Our findings conflict with those of Wang et al. (53), who used IFM and reported that furazlocillin prevents localization of FtsI. However, in our hands wild-type FtsI localized in furazlocillin-treated cells (filaments) as determined by IFM, although the data were not as clean or as reproducible as those we obtained with GFP-FtsI. The simplest interpretation of our observations is that FtsI does not require its transpeptidase activity for septal localization. Nevertheless, it remains possible that GFP-FtsI responds aberrantly to β-lactams. FtsI’s transpeptidase activity can also be inactivated genetically by changing serine 307 in the active site to cysteine. This mutant protein blocks cell division when overproduced (13), perhaps because it competes for localization to the division site. Efforts are underway in our lab to fuse this form of FtsI to GFP so that the localization behavior of a catalytically inactive FtsI protein can be studied directly.

FtsI is a late recruit to the septal ring.

Localization of GFP-FtsI to the division site required the Z ring, FtsA, FtsQ, and FtsL. The requirement for FtsZ and FtsA has also been observed using IFM (53). Conversely, we found that many other division proteins localize in an FtsI-independent fashion—this was shown here for FtsZ, FtsA, and ZipA, and will be published elsewhere for FtsQ and FtsL (15, 23). In the cases of FtsZ and FtsA, our findings confirm results obtained by IFM (2, 5, 46). Taken together, these findings indicate that FtsI is a late recruit to the septal ring.

Our results, together with those from other labs (2, 3, 5, 38, 59), suggest that there is a simple, linear pathway for sequential recruitment of proteins to the division site (Fig. 6). Whether the sequence reflects a temporal order and/or assembly of a multiprotein complex is not yet known. Two aspects of the model are particularly striking. First, the division proteins that reside primarily in the cytoplasm (FtsZ, FtsA, and ZipA) all precede those that reside primarily in the periplasm (FtsQ, FtsL, FtsI, and FtsN). Are the signals that govern where to divide generated in the cytoplasm rather than the cell envelope? The second striking feature of the model is its linearity. There are as yet no examples of codependency as would be the case if, for example, two proteins localized to the division site as a heterodimer or cooperative interactions among two or more proteins were needed to stabilize their assembly into a multiprotein complex. Likewise, there are no branches, as would be the case if two proteins bound independently to the same target. Do some of the proteins act sequentially at the division site rather than as part of a single multiprotein complex? So far, the only convincing evidence for direct interactions among the division proteins are for FtsZ with FtsA and ZipA (18, 29, 54). Nevertheless, one potential use of information on the dependency of septal localization is that it suggests which protein pairs are the most promising candidates to test for direct interactions in vitro.