FtsQ, FtsL and FtsI require FtsK, but not FtsN, for co-localization with FtsZ during Escherichia coli cell division

Authors


Abstract

During cell division in Gram-negative bacteria, the cell envelope invaginates and constricts at the septum, eventually severing the cell into two compartments, and separating the replicated genetic materials. In Escherichia coli, at least nine essential gene products participate directly in septum formation: FtsA, FtsI, FtsL, FtsK, FtsN, FtsQ, FtsW, FtsZ and ZipA. All nine proteins have been localized to the septal ring, an equatorial ring structure at the division site. We used translational fusions to green fluorescent protein (GFP) to demonstrate that FtsQ, FtsL and FtsI localize to potential division sites in filamentous cells depleted of FtsN, but not in those depleted of FtsK. We also constructed translational fusions of FtsZ, FtsA, FtsQ, FtsL and FtsI to enhanced cyan or yellow fluorescent protein (ECFP or EYFP respectively), GFP variants with different fluorescence spectra. Examination of cells expressing different combinations of the fusions indicated that FtsA, FtsQ, FtsL and FtsI co-localize with FtsZ in filaments depleted of FtsN. These localization results support the model that E. coli cell division proteins assemble sequentially as a multimeric complex at the division site: first FtsZ, then FtsA and ZipA independently of each other, followed successively by FtsK, FtsQ, FtsL, FtsW, FtsI and FtsN.

Introduction

Cytokinesis in bacteria involves the coordinated constriction of cell envelope layers by the septal ring, a cytoskeleton-like organelle that assembles at the division site. This equatorial ring structure remains associated with the invaginating cell membrane until the two daughter cells separate (reviewed in Rothfield et al., 1999; Margolin 2000). Also known as the divisome, the septal ring consists of at least nine essential cell division gene products in Escherichia coli: FtsA, FtsI, FtsK, FtsL, FtsN, FtsQ, FtsW, FtsZ and ZipA (Rothfield et al., 1999; Margolin 2000). All nine proteins localize to the septum, a site in the mid-cell region apparently selected by the minCDE system (Hu and Lutkenhaus, 1999; Raskin and de Boer, 1999a; 1999b; Rowland et al., 2000; Fu et al., 2001; Hale et al., 2001; reviewed in RayChaudhuri et al., 2001). When any one of the cell division proteins is non-functional or absent, cells grow without dividing, leading to the formation of filaments.

Studies of cell division proteins have provided some insight to their individual activities, but how they co-operatively implement cytokinesis remains ambiguous. FtsZ, the best characterized of the nine, is a homologue of the eukaryotic cytoskeletal component tubulin (de Pereda et al., 1996; Lowe and Amos, 1998; Nogales et al., 1998). Like tubulin, it has GTPase and polymerization activities and may provide the force necessary to constrict the cell (de Boer et al., 1992; RayChaudhuri and Park, 1992; Mukherjee et al., 1993; Mukherjee and Lutkenhaus, 1994;1998; Erickson et al., 1996; Yu and Margolin, 1997; Lu et al., 2000). FtsA and ZipA both interact directly with FtsZ (Hale and de Boer, 1997; Wang et al., 1997; Din et al., 1998; Liu et al., 1999; Ma and Margolin, 1999; Hale et al., 2000; Mosyak et al., 2000; Haney et al., 2001), and ZipA stabilizes polymerized FtsZ (RayChaudhuri, 1999; Hale et al., 2000). Interestingly, FtsA is homologous to actin, another eukaryotic cytoskeletal component (Bork et al., 1992; Sanchez et al., 1994; van den Ent and Lowe, 2000). FtsI, also known as penicillin-binding protein 3, appears to be a transpeptidase required specifically for peptidoglycan synthesis at the septum (Spratt and Cromie, 1988; reviewed in Nguyen-Disteche et al., 1998). The largest of the group, the 1330-amino-acid FtsK, appears to be a bifunctional protein: its C-terminal domain facilitates resolution of chromosome dimers during DNA segregation, whereas the N-terminal domain carries out a necessary, but undefined, function in the developing septum; (Draper et al., 1998; Liu et al., 1998; Wang and Lutkenhaus, 1998; Yu et al., 1998a; 1998b; Recchia et al., 1999; Steiner et al., 1999; Barre et al., 2000). The functions of the remaining proteins (FtsL, -N, -Q and -W) are unknown. Furthermore, there are probably other factors involved in cell division that have not yet been identified.

To understand how these cell division proteins work together to form the septal ring, we can examine their dependence on one another for localization. The ability of a particular cell division protein to assemble into a ring structure, at putative division sites in a filament, indicates its independence from the missing or nonfunctional protein for recruitment into the septal ring. Compilation of various localization studies using this approach have yielded a mostly linear dependency pathway, suggestive of a sequential order of assembly (Rothfield et al., 1999; Margolin 2000; see below). The localization of FtsZ at the division site to form the Z ring appears to be the first step in divisome assembly: FtsZ does not depend on any of the other eight proteins for localization, whereas all others depend on its activity. FtsZ is followed closely by FtsA and ZipA; they do not require each other, or any other cell division protein, except for FtsZ. Subsequently, FtsK, -Q, -L, -I, and -N appear to be recruited in that order. The position of FtsW in the pathway is not clear from the published literature.

We had determined previously that there is a strictly linear dependency among FtsQ, FtsL and FtsI for septal recruitment, and that the recruitment of all three depends on both FtsZ and FtsA (Chen et al., 1999; Ghigo et al., 1999; Weiss et al., 1999). However, a number of gaps remained in the dependency analysis of protein assembly at the septum. Furthermore, it seemed possible that some pairs of proteins might depend on each other for assembly, thus not exhibiting the linear sequence that characterized most of the pathway. For instance, FtsK has been assigned its position upstream of FtsQ because it can localize to potential division sites in filaments without functional FtsQ or FtsI, but not in ftsZ or ftsA filaments (Wang and Lutkenhaus, 1998; Wang et al., 1998; Yu et al., 1998a). Yet, the septal localization of FtsQ and other downstream proteins in FtsK-deprived filaments has not been examined. Also, while FtsN appears to be a late recruit to the septal ring, as it needs functional FtsZ, -A, -Q, and -I for localization, and Z rings can form in filaments depleted of FtsN (Addinall et al., 1997), there still might exist a co-dependency with one of the other proteins.

Here, we report that GFP fusions to FtsQ, FtsL and FtsI fail to localize in filaments depleted of FtsK and confirm that FtsZ, FtsA and ZipA do localize to potential division sites in such filaments. Localization studies using GFP fusion proteins also showed that FtsZ, -A, -I, -L and -Q and ZipA can all be recruited into septal rings in the absence of FtsN. These localization results strengthen considerably the sequential model of protein assembly into the septal ring. To give further support to the idea that the divisome is a complex of multiple proteins, we used ECFP and EYFP fusions to demonstrate that FtsQ, FtsL and FtsI co-localize with FtsZ in filaments depleted of FtsN. A summary of all these findings provides a framework for understanding how the septal ring is formed.

Results

Construction of FtsK depletion strains

To assess whether FtsQ, FtsL and FtsI depend on FtsK to localize to the division site, we constructed and compared four different strains (JOE563, JOE600, JOE669 and JOE671), in which FtsK expression could be repressed (see Experimental procedures and Table 4). The first FtsK depletion strain (JOE563) was constructed in the MC4100 background. In this strain, the ftsK::cat-Δ5 null allele (Draper et al., 1998) on the chromosome is complemented by a wild-type ftsK, under the control of an arabinose-dependent promoter (PBAD) on a low-copy plasmid which has an origin of replication from pSC101. The strain formed healthy colonies on media containing arabinose, which induced ftsK expression, but failed to re-streak on media containing glucose, which repressed ftsK expression. Microscopic examination showed that liquid cultures grown with arabinose contained mostly cells with normal morphology (Fig. 1E, I and M), although longer cells and filaments were observed after extended growth in log phase (similar to those in Fig. 1I). In contrast, cultures grown for extended periods in log phase with glucose contained almost all filaments (Figs 1F, J and N), most of which had constrictions suggestive of incomplete division sites (white arrowheads in Figs 1G, K and O).

Table 4. Strains and plasmids.
Strain or plasmidRelevant genetic marker(s) or featuresConstruction, source or referencea
  • a

    . P1 indicates P1 transduction. For example, JOE296 was constructed by infecting MC4100 with P1 lysate made from EC294.

Strains without gfp, ecfp or eyfp
MG1655FprototrophLaboratory collection
BW25141 lacI q rrnB T14 ?lacZWJ16 ?phoBR580 hsdR514 ?araBADAH33 Datsenko and Wanner 2000
?rhaBADLD78galU95 endABT333uidA(?MluI)::pir+recA1 
C600Fthr-1 leuB6 tonA21 lacY1 supE44 thi-1 rbfD1Roberto Kolter, Harvard Medical School
MC4100FaraD139 ?lacU169 relA1 rpsL150 thi mot flb5301Laboratory collection
deoC7 ptsF25 rbsR  
CDK5 aroA::Tn10 ftsK::cat-?5 pcnB::kan Draper et al. 1998
JKD41/pKD123 ftsN::kan Dai et al. 1993
DHB6521SM551 λInCh1 (Kanr) Boyd et al. 2000
EC294MG1655 leu::Tn10David S. Weiss, University of Iowa
JOE296MC4100 araD+leu::Tn10P1 (EC294) × MC4100, select Tetr,
screen Ara+
JOE309MC4100 araD+P1 (BW25141) × JOE296, select Leu+,
screen Ara+
JOE310MC4100 ?araBADAH33P1 (BW25141) × JOE296, select Leu+,
screen Ara
JOE559JOE309/pJC83Transform with pJC83
JOE561JOE309/pJC85Transform with pJC85
JOE593JOE309 ?(λattL-lom)::bla lacIqPBAD-ftsKIntegrate pBADK using λInCh1
JOE662MG1655/pJC85Transform with pJC85
JOE663C600/pJC85Transform with pJC85
Strains with gfp, ecfp or eyfp fusions
gfp fusions:
 EC436MC4100 ?(λattL–lom)::bla lacIqP207gfp–ftsI Weiss et al. (1999)
 EC438MC4100 ?(λattL–lom)::bla lacIqP207gfp–ftsL Ghigo et al. (1999)
 EC442MC4100 ?(λattL–lom)::bla lacIqP207gfp–ftsQ Chen et al. (1999)
 EC447MC4100 ?(λattL–lom)::bla lacIqP210ftsA–gfp Weiss et al. (1999)
 EC448MC4100 ?(λattL–lom)::bla lacIqP208ftsZ–gfp Weiss et al. (1999)
 EC450MC4100 ?(λattL–lom)::bla lacIqP208zipA–gfp Weiss et al. (1999)
 EC452MC4100 ?(λattL–lom)::bla lacIqP207gfp Weiss et al. (1999)
 JOE649JOE309 attφ80::pJC113(P210ftsA–gfp)Integrate pJC113 using pAH123
 JOE650JOE309 attφ80::pJC114(P208ftsZ–gfp)Integrate pJC114 using pAH123
 JOE651JOE309 attφ80::pJC115(P207gfp–ftsI)Integrate pJC115 using pAH123
 JOE652JOE309 attφ80::pJC116(P207gfp–ftsL)Integrate pJC116 using pAH123
 JOE654JOE309 attφ80::pJC118(P207gfp–ftsQ)Integrate pJC118 using pAH123
ecfp or eyfp fusions:
 JOE514JOE309 ?(λattL–lom)::bla lacIqP207ecfp–ftsIIntegrate pJC59 using λInCh1
 JOE515JOE309 ?(λattL–lom)::bla lacIqP207ecfp–ftsLIntegrate pJC60 using λInCh1
 JOE516JOE309 ?(λattL–lom)::bla lacIqP207ecfp–ftsQIntegrate pJC61 using λInCh1
 JOE517JOE309 ?(λattL–lom)::bla lacIqP207eyfp–ftsIIntegrate pJC62 using λInCh1
 JOE518JOE309 ?(λattL–lom)::bla lacIqP207eyfp–ftsLIntegrate pJC63 using λInCh1
 JOE519JOE309 ?(λattL–lom)::bla lacIqP207eyfp–ftsQIntegrate pJC64 using λInCh1
 JOE520JOE309 ?(λattL–lom)::bla lacIqP208ftsA–ecfpIntegrate pJC65 using λInCh1
 JOE521JOE309 ?(λattL–lom)::bla lacIqP208ftsZ–ecfpIntegrate pJC66 using λInCh1
 JOE522JOE309 ?(λattL–lom)::bla lacIqP208ftsA–eyfpIntegrate pJC67 using λInCh1
 JOE523JOE309 ?(λattL–lom)::bla lacIqP208ftsZ–eyfpIntegrate pJC68 using λInCh1
 JOE532JOE310 attφ80::pJC71(P208eyfp)Integrate pJC71 using pAH123
 JOE533JOE310 attφ80::pJC72(P208ftsA–ecfp)Integrate pJC72 using pAH123
 JOE534JOE310 attφ80::pJC73(P208ftsZ–ecfp)Integrate pJC73 using pAH123
 JOE535JOE310 attφ80::pJC74(P208ftsA–eyfp)Integrate pJC74 using pAH123
 JOE536JOE310 attφ80::pJC75(P208ftsZ–eyfp)Integrate pJC75 using pAH123
 JOE608JOE309 attφ80::pJC101(P204myc–eyfp–ftsI)Integrate pJC101 using pAH123
 JOE609JOE309 attφ80::pJC102(P204myc–eyfp–ftsL)Integrate pJC102 using pAH123
 JOE611JOE309 attφ80::pJC104(P204myc–eyfp–ftsQ)Integrate pJC104 using pAH123
 JOE612JOE309 attφ80::pJC105(P204myc–ecfp–ftsI)Integrate pJC105 using pAH123
 JOE613JOE309 attφ80::pJC106(P204myc–ecfp–ftsL)Integrate pJC106 using pAH123
 JOE615JOE309 attφ80::pJC108(P204myc–ecfp–ftsQ)Integrate pJC108 using pAH123
FtsK depletion strains
MC4100-derived:
 JOE563JOE309 ftsK::cat-?5/pJC85P1 (CDK5/pBADK) × JOE561, select Cmr
 JOE698JOE563 ?(λattL–lom)::bla lacIqP210ftsA–gfpP1 (EC447) × JOE563, select low Ampr
 JOE699JOE563 ?(λattL–lom)::bla lacIqP208ftsZ–gfpP1 (EC448) × JOE563, select low Ampr
 JOE700JOE563 ?(λattL–lom)::bla lacIqP207gfp–ftsIP1 (EC436) × JOE563, select low Ampr
 JOE701JOE563 ?(λattL–lom)::bla lacIqP207gfp–ftsLP1 (EC438) × JOE563, select low Ampr
 JOE703JOE563 ?(λattL–lom)::bla lacIqP207gfp–ftsQP1 (EC442) × JOE563, select low Ampr
 JOE704JOE563 ?(λattL–lom)::bla lacIqP208zipA–gfpP1 (EC450) × JOE563, select low Ampr
 JOE705JOE563 ?(λattL–lom)::bla lacIqP207gfpP1 (EC452) × JOE563, select low Ampr
 JOE600JOE309 ftsK::cat–?5 ?(λattL–lom)::bla lacIqPBADftsKP1 (CDK5/pBADK) × JOE593, select Cmr
 JOE655JOE600 attφ80::pJC113(P210ftsA–gfp)P1 (JOE649) × JOE600, select Spcr
 JOE656JOE600 attφ80::pJC114(P208ftsZ–gfp)P1 (JOE650) × JOE600, select Spcr
 JOE657JOE600 attφ80::pJC115(P207gfp–ftsI)P1 (JOE651) × JOE600, select Spcr
 JOE658JOE600 attφ80::pJC116(P207gfp–ftsL)P1 (JOE652) × JOE600, select Spcr
 JOE660JOE600 attφ80::pJC118(P207gfp–ftsQ)P1 (JOE654) × JOE600, select Spcr
MG1655 or C600-derived:
 JOE669MG1655 ftsK::cat-?5/pJC85P1 (JOE600) × JOE662, select Cmr
 JOE671C600 ftsK::cat-?5/pJC85P1 (JOE600) × JOE663, select Cmr
 JOE673JOE669 ?(λattL–lom)::bla lacIqP210ftsA–gfpP1 (EC447) × JOE669, select low Ampr
 JOE674JOE669 ?(λattL–lom)::bla lacIqP208ftsZ–gfpP1 (EC448) × JOE669, select low Ampr
 JOE675JOE669 ?(λattL–lom)::bla lacIqP207gfp–ftsIP1 (EC436) × JOE669, select low Ampr
 JOE676JOE669 ?(λattL–lom)::bla lacIqP207gfp–ftsLP1 (EC438) × JOE669, select low Ampr
 JOE678JOE669 ?(λattL–lom)::bla lacIqP207gfp–ftsQP1 (EC442) × JOE669, select low Ampr
 JOE679JOE669 ?(λattL–lom)::bla lacIqP208zipA–gfpP1 (EC450) × JOE669, select low Ampr
 JOE680JOE669 ?(λattL–lom)::bla lacIqP207gfpP1 (EC452) × JOE669, select low Ampr
FtsN depletion strains
 JOE565
 
JOE309 ftsN::kan/pJC83
Kanr
P1 (JKD41/pKD123) × JOE559, select
 JOE585JOE565 ?(λattL–lom)::bla lacIqP207gfp–ftsQP1 (EC442) × JOE565, select low Ampr
 JOE586JOE565 ?(λattL–lom)::bla lacIqP208ftsZ–gfpP1 (EC448) × JOE565, select low Ampr
 JOE587JOE565 ?(λattL–lom)::bla lacIqP210ftsA–gfpP1 (EC447) × JOE565, select low Ampr
 JOE594JOE565 ?(λattL–lom)::bla lacIqP207gfp–ftsIP1 (EC436) × JOE565, select low Ampr
 JOE602JOE565 ?(λattL–lom)::bla lacIqP207gfp–ftsLP1 (EC438) × JOE565, select low Ampr
 JOE664JOE565 ?(λattL–lom)::bla lacIqP208zipA–gfpP1 (EC450) × JOE565, select low Ampr
 JOE628JOE565 ?(λattL–lom)::bla lacIqP208ftsZ–ecfpP1 (JOE521) × JOE565, select low Ampr
 JOE642JOE628 attφ80::pJC71(P208eyfp)P1 (JOE532) × JOE628, select Spcr
 JOE643JOE628 attφ80::pJC74(P208ftsA–eyfp)P1 (JOE535) × JOE628, select Spcr
 JOE644JOE628 attφ80::pJC75(P208ftsZ–eyfp)P1 (JOE536) × JOE628, select Spcr
 JOE645JOE628 attφ80::pJC101(P204myc–eyfp–ftsI)P1 (JOE608) × JOE628, select Spcr
 JOE646JOE628 attφ80::pJC102(P204myc–eyfp–ftsL)P1 (JOE609) × JOE628, select Spcr
 JOE648JOE628 attφ80::pJC104(P204myc–eyfp–ftsQ)P1 (JOE611) × JOE628, select Spcr
Plasmids
pAH123λcI857, rep101ts origin, Pr-intφ80, AmprBarry L. Wanner, Purdue University
pAH144 oriR R6Kγ, attPHK022, SpcrBarry L. Wanner, Purdue University
pAH162 oriR R6Kγ, attPφ80, TetrBarry L. Wanner, Purdue University
pBAD18arabinose regulation, Ampr Guzman et al. (1995)
pBAD33arabinose regulation, Cmr Guzman et al. (1995)
pBAD42
pBADK
arabinose regulation, Spcr, pSC101 origin
pBAD18-ftsK
Daniel Ritz, Harvard Medical School
Draper et al. (1998)
pDSW204IPTG-regulated promoter, Ampr Weiss et al. (1999)
pDSW207pDSW204–gfp–MCS (fusion vector) Weiss et al. (1999)
pDSW208pDSW204–MCS–gfp (fusion vector) Weiss et al. (1999)
pDSW230pDSW208-ftsZ Weiss et al. (1999)
pDSW233pDSW210-ftsA Weiss et al. (1999)
pDSW234pDSW207-ftsI Weiss et al. (1999)
pDSW236pDSW207-ftsL Ghigo et al. (1999)
pDSW240pDSW207-ftsQ Chen et al. (1999)
pECFP ecfp, AmprClontech
pEYFP eyfp, AmprClontech
pJC2pBAD18–ftsNSee Experimental procedures
pJC50pDSW207(ECFP) (ecfp–MCS fusion vector)See Experimental procedures
pJC51pDSW207(EYFP) (eyfp–MCS fusion vector)See Experimental procedures
pJC52pDSW208(ECFP) (MCS–ecfp fusion vector)See Experimental procedures
pJC58pDSW208(EYFP) (MCS–eyfp fusion vector) BseRI–BsrGI fragment from pEYFP used
to replace same fragment in pJC52
pJC59P207-ecfp-ftsI EcoRI–PvuI fragment containing ftsI from
pDSW234 inserted into same sites of
pJC50
pJC60
P207-ecf-ftsL EcoRI–PvuI fragment containing ftsL
from pDSW236 inserted into same sites
of pJC50
pJC61
P207-ecfp-ftsQ EcoRI–PvuI fragment containing ftsQ
from pDSW240 inserted into same sites
of pJC50
pJC62P207-eyfp-ftsI EcoRI–PvuI fragment containing ftsI from
pDSW234 inserted into same sites of
pJC51
pJC63P207-eyfp-ftsL EcoRI–PvuI fragment containing ftsL
from pDSW236 inserted into same sites
of pJC51
pJC64P207-eyfp-ftsQ EcoRI–PvuI fragment containing ftsQ
from pDSW240 inserted into same sites
of pJC51
pJC65P208-ftsA-ecfp NcoI–PstI fragment containing ftsA from
pDSW233 inserted into same sites of
pJC52
pJC66P208-ftsZ-ecfp NcoI–PstI fragment containing ftsZ from
pDSW230 inserted into same sites
of pJC52
pJC67P208-ftsA-eyfp BseRI–BsrGI fragment from pEYFP used
to replace same fragment in pJC65
pJC68P208-ftsZ-eyfp BseRI–BsrGI fragment from pEYFP used
to replace same fragment in pJC66
pJC69oriRR6Kγ, attPΦ80, Spcr NotI–SphI fragment containing aadA
from pAH144 replaced same fragment of
pAH162
pJC70pJC69-P208-ecfp SphI–ScaI fragment containing
ecfp frmpJC52 inserted into pJC69 cut
with SphI and HincII
pJC71pJC69-P208-eyfp SphI–ScaI fragment containing eyfp
frmpJC58 inserted into pJC69 cut with
SphI and HincII
pJC72pJC69-P208-ftsA-ecfp SphI–ScaI fragment containing ftsA–ecfp
frmpJC65 inserted into pJC69 cut with
SphI and HincII
pJC73pJC69-P208-ftsZ-ecfp SphI–ScaI fragment containing ftsZ–ecfp
frmpJC66 inserted into pJC69 cut with
SphI and HincII
pJC74pJC69-P208-ftsA-eyfp SphI–ScaI fragment containing ftsA–eyfp
frmpJC67 inserted into pJC69 cut with
SphI and HincII
pJC75pJC69-P208-ftsZ-eyfp SphI–ScaI fragment containing ftsZ–eyfp
frmpJC68 inserted into pJC69 cut with
SphI and HincII
pJC76pDSW204-myc-eyfp-MCS (fusion vector)See Experimental procedures
pJC78P204-myc-eyfp-ftsI EcoRI–PvuI fragment containing ftsI from
pDSW234 inserted into same sites of
pJC76
pJC79P204-myc–eyfp–ftsL EcoRI–PvuI fragment containing ftsL
from pDSW236 inserted into same sites
of pJC76
pJC81P204myc–eyfp–ftsQ EcoRI–PvuI fragment containing ftsQ
from pDSW240 inserted into same sites
of pJC76
pJC82
pDSW204–myc–ecfp–MCS (fusion vector) BseRI–BsrGI fragment from pECFP used
o replace same fragment in pJC76
pJC83pBAD33–ftsN SacI–XbaI fragment containing ftsN from
pJC2 inserted into same sites of pBAD33
pJC85pBAD42–ftsK SacI–XbaI fragment containing ftsK from
pBADK inserted into same sites of
pBAD42
pJC87P204myc–ecfp–ftsI EcoRI–PvuI fragment containing ftsI from
pDSW234 inserted into same sites of
pJC82
pJC88
P204myc–ecfp–ftsL EcoRI–PvuI fragment containing ftsL
from pDSW236 inserted into same sites
of pJC82
pJC90
P204myc–ecfp–ftsQ EcoRI–PvuI fragment containing ftsQ
from pDSW240 inserted into same sites
of pJC82
pJC101pJC69–P204myc–eyfp–ftsI SphI–ScaI fragment from pJC78 inserted
into pJC69 cut with SphI and HincII
pJC102pJC69–P204myc–eyfp–ftsL SphI–ScaI fragment from pJC79 inserted
into pJC69 cut with SphI and HincII
pJC104pJC69–P204myc–eyfp–ftsQ SphI–ScaI fragment from pJC81 inserted
into pJC69 cut with SphI and HincII
pJC105pJC69–P204myc–ecfp–ftsI SphI–ScaI fragment from pJC87 inserted
into pJC69 cut with SphI and HincII
pJC106pJC69–P204myc–ecfp–ftsL SphI–ScaI fragment from pJC88 inserted
into pJC69 cut with SphI and HincII
pJC018pJC69–P204myc–ecfp–ftsQ SphI–ScaI fragment from pJC90 inserted
into pJC69 cut with SphI and HincII
pJC113pJC69–P210ftsA–gfp SphI–ScaI fragment from pDSW233
inserted into pJC69 cut with SphI and
HincII
pJC114pJC69–P208ftsZ–gfp SphI–ScaI fragment from pDSW230
inserted into pJC69 cut with SphI and
HincII
pJC115pJC69–P207gfp–ftsI SphI–ScaI fragment from pDSW234
inserted into pJC69 cut with SphI and
HincII
pJC116pJC69–P207gfp–ftsL SphI–ScaI fragment from pDSW236
inserted into pJC69 cut with SphI and
HincII
pJC118pJC69–P207gfp–ftsQ SphI–ScaI fragment from pDSW240
inserted into pJC69 cut with SphI and
HincII
Figure 1.

Localization of GFP fusions to FtsZ, FtsA, FtsQ, FtsL, and FtsI in FtsK depletion strains. See Experimental procedures for growth conditions and microscopy techniques.

A–D. Localization of FtsZ–GFP in MG1655-derived FtsK depletion strain (JOE674).

E–H. Localization of FtsA–GFP in MC4100-derived FtsK depletion strain (JOE698).

I–L. Localization of GFP–FtsQ in MC4100-derived FtsK depletion strain (JOE703).

M–P. Localization of GFP–FtsL in MC4100-derived FtsK depletion strain (JOE701).

Q–T. Localization of GFP–FtsI in MG1655-derived FtsK depletion strain (JOE675).

A, E, I, M and Q. GFP-fluorescence images of cells grown with arabinose to express FtsK.

B, F, J, N and R. GFP-fluorescence images of cells grown with glucose to deplete FtsK.

C, G, K, O and S. Phase-contrast images of FtsK-depleted cells in the same row.

D, H, L, P and T. DAPI-staining of FtsK-depleted cells in the same row.

Double arrowheads in F and H point to a cell with aberrant DNA staining and no FtsA-GFP localization. White arrowheads in G, K and O point to indentations in the filaments of MC4100-derived depletion strains. Black arrowhead in N points to a faint localization signal of GFP–FtsL. Phase-contrast and DAPI images are at half scale compared with the GFP-fluorescence images: black bar (for A, B, E, F, I, J, M, N, Q, R), 10 µm; white bar (for C, D, G, H, K, L, O, P, S, T), 10 µm.

As the partial constrictions may have been a result of residual amounts of FtsK production, we lowered the level of ftsK expression. This was accomplished by integrating a single copy of the PBAD-ftsK construct into the chromosome at the lambda attachment site (see Experimental procedures). The resultant MC4100-derived FtsK depletion strain (JOE600) grew poorly on plates containing arabinose; long cells and filaments became abundant in arabinose liquid cultures after extended growth in log phase, indicative of expression levels below those needed for normal growth (data not shown). The strain failed to grow on plates containing glucose and formed long filaments in glucose liquid cultures, but the filaments still exhibited incomplete constriction sites (data not shown).

The constriction sites that we observed in the FtsK-depleted filaments (of JOE563 and JOE600) were not apparent in FtsK depletion strains constructed by other labs (Draper et al., 1998; Wang and Lutkenhaus, 1998). Differences in strain background and construction might account for these differences in morphology. To test this, we introduced the same ftsK null allele and complementing plasmid into the MG1655 and C600 backgrounds (to generate JOE669 and JOE671 respectively). These strains formed healthy colonies on arabinose plates but not on glucose plates. When grown in liquid media containing arabinose, the cultures consisted of cells of varying length, from short cells to long filaments (Fig. 1A and Q; data not shown for JOE671), possibly as a result of insufficient FtsK expression. When grown in liquid media containing glucose, these strains produced mostly smooth filaments, with no partial indentations (Fig. 1C and S; data not shown for JOE671). Hence, the presence of incomplete constrictions in the FtsK-depleted filaments depends on strain background. Whereas the four FtsK depletion strains exhibited different cell morphologies, nucleoids in the FtsK-depleted filaments were evenly distributed in all four strains (Fig. 1D, H, L, P and T; data not shown).

FtsQ, Fts, and FtsI require FtsK for localization

To determine whether FtsQ, FtsL and FtsI depend on FtsK for septal localization, we introduced translational fusions of gfp to ftsQ, ftsL or ftsI into the chromosomes of three FtsK depletion strains (JOE563, JOE600 and JOE669), either at the lambda or the φ80 attachment site (see Experimental procedures). As positive controls, we also constructed FtsK depletion strains carrying gfp fusions to ftsZ, ftsA or zipA. Expression of the various gfp fusions was driven by IPTG (isopropyl-β-d-thiogalactopyranoside)-inducible promoters. All these fusion constructs had been characterized previously (Chen et al., 1999; Ghigo et al., 1999; Weiss et al., 1999).

Parallel cultures of these strains were grown in rich media containing glucose or arabinose, then chemically fixed for examination by fluorescence microscopy. Similar results were obtained with all three FtsK depletion strains; quantitation of cell length and septal rings was done only in JOE563-derived strains. As expected, when cells were grown with arabinose to induce FtsK expression, all the GFP fusions localized to the division site (Fig. 1A, E, I, M and Q; ZipA localization not shown). The majority of cells exhibited a ring in the mid-cell region (Table 1, % cells with ring). We noticed that cells expressing GFP–FtsQ (Fig. 1I) or higher levels of FtsZ–GFP (at 50 µM IPTG induction; data not shown) appeared longer and more filamentous; comparison of average cell lengths confirmed this observation (Table 1). We did not see this effect in other genetic backgrounds (for instance, see measurements for FtsN depletion strains below). The FtsK depletion strain seems to be more sensitive than strains with other cell division defects to changes in the levels of cell division proteins.

Table 1. Localization frequencies of GFP fusions in FtsK depletion strains a.
GFP
fusion
Strain[IPTG]FtsKb% Cells
w/ring(s)
Total no.
of cells
Total no.
of rings
Total
length
Average length of cells (±SD) Average
length/ringc
Spacing
of ringsd
No ringw/ring(s)Total
  • a

    . All length measurements are in µm. Measurements were made as described in Experimental procedures. SD, standard deviation; NA, not applicable.

  • b

    .+, cells grown with arabinose to induce FtsK expression; –, cells grown with glucose to repress FtsK expression.

  • c

    . The ratio of cell length to number of rings was determined for each filament. The average (±SD) of the ratios was then calculated for each cohort of cells. SD reflects variation in the number of rings per filament.

  • d

    . Total length (column 8) divided by total number of rings (column 7). The spacing is inversely proportional to the frequency of rings.

FtsZJOE69910 µM+
87
100
340
55
299
208
1693
1826
4.4 ± 1.65.1 ± 1.55.0 ± 1.5
33 ± 15
9.2 ± 3.05.7
8.8
FtsZJOE69950 µM+
92
100
183
28
198
197
1243
1387
5.7 ± 2.06.9 ± 3.46.8 ± 3.3
50 ± 19
7.4 ± 1.36.3
7.0
FtsAJOE69880 µM+
93
100
320
54
296
217
1536
1692
4.2 ± 1.34.9 ± 1.44.8 ± 1.4
31 ± 19
8.1 ± 2.35.2
7.8
ZipAJOE70480 µM+
81
100
305
48
249
170
1491
1721
4.1 ± 1.15.1 ± 1.94.9 ± 1.8
36 ± 15
10.6 ± 3.66.0
10
FtsQJOE70310 µM+
83
9
291
46
247
4
1836
1280
4.8 ± 1.86.6 ± 2.56.3 ± 2.5
28 ± 14
NA7.4
320
FtsLJOE70110 µM+
75
38
297
60
225
29
1469
1765
4.0 ± 1.15.3 ± 1.65.0 ± 1.6
29 ± 13
NA6.5
61
FtsIJOE70010 µM+
68
13
311
55
213
7
1661
1692
4.4 ± 1.45.8 ± 1.95.3 ± 1.9
31 ± 13
NA7.8
240

When the FtsK depletion strains were grown with glucose, we saw localization of FtsZ–(Fig. 1B), FtsA–(Fig. 1F) and ZipA–GFP (data not shown) to potential division sites in the filaments, in agreement with previous studies. For all three proteins, the calculated spacings between ring structures (total cell length divided by total number of rings; last column of Table 1) were slightly higher in the filaments compared with dividing cells. This reduction in calculated ring frequency has been observed in other localization studies (Hale and de Boer, 1999; Weiss et al., 1999). However, we could increase the frequency of Z rings in the filaments when we increased the level of FtsZ–GFP expression (Table 1; 10 vs. 50 µM IPTG induction).

In contrast with the results with FtsZ, FtsA and ZipA, we found that filaments expressing GFP fusions to FtsQ (Fig. 1J), FtsL (Fig. 1N) or FtsI (Fig. 1R) rarely exhibited fluorescence at potential division sites; instead, they showed diffuse fluorescence. Occasionally, we observed very faint fluorescent rings in some of these FtsK-depleted cells (Table 1), particularly in those expressing GFP–FtsL, but generally each filament contained only one or two rings (black arrowhead in Fig. 1N). Calculation of the spacings between rings (last column of Table 1) indicated that localization of FtsQ, FtsL and FtsI was greatly reduced in the FtsK-depleted filaments. Thus, FtsQ, FtsL and FtsI depend on the presence of FtsK for septal localization. Taking into account previous localization results, we concluded that FtsK, -Q, -L and -I localize to the division site in that sequence.

FtsQ, FtsL and FtsI localize independently of FtsN

To determine the relationship between FtsN and FtsQ, -L and -I, we examined localization of the corresponding GFP fusions in cells depleted of FtsN. We constructed an MC4100-derived FtsN depletion strain (JOE565) by complementing the ftsN::kan null allele (Dai et al., 1993) with a plasmid carrying the wild-type ftsN allele under PBAD control (see Table 4). As expected, this strain grew well on plates containing arabinose but failed to grow on plates containing glucose. Cells grown in liquid media containing arabinose appeared normal in morphology (data not shown), whereas those grown in glucose media became mostly long, smooth filaments (Fig. 2). Indentations were occasionally observed in these FtsN-depleted filaments (white arrowheads in Fig. 2G). Nucleoids were evenly distributed along the filaments (Fig. 2E and F).

Figure 2.

ZipA, FtsQ, FtsL and FtsI localize to potential division sites in FtsN-depleted filaments. See Experimental procedures for growth conditions and microscopy techniques.

A. Localization of ZipA-GFP in FtsN depletion strain JOE664.

B. Localization of GFP-FtsL in FtsN depletion strain JOE602.

C. Localization of GFP-FtsQ in FtsN depletion strain JOE585.

D. Localization of GFP-FtsI in FtsN depletion strain JOE594.

E and G. DAPI-staining and phase-contrast images of cells in C.

F and H. DAPI-staining and phase-contrast images of cells in D.

White arrowheads in frame G point to partial constrictions occasionally found in FtsN–depleted filaments. Bar, 10 µm.

GFP fusions to FtsZ, FtsA, ZipA, FtsQ, FtsL and FtsI were expressed chromosomally in the FtsN depletion strain, and their localization properties were assessed in the presence of arabinose or glucose. In cultures grown with arabinose to induce FtsN expression, all the GFP fusions localized to the division site in the majority of cells examined (Table 2, % cells with ring). Cells that did not exhibit a mid-cell fluorescence signal were, on average, shorter in length, suggesting that they had finished the previous round of division more recently and were at an earlier stage of the division cycle.

Table 2. Localization frequencies of GFP fusions in FtsN depletion strains a ·
GFP
fusion
Strain[IPTG]FtsNb% Cells
w/ring(s)
Total no.
of cells
Total no.
of rings
Total
length
Average length of cells (±SD) Average
length/ringc
Spacing
of ringsd
No ringw/ring(s)Total
  • a.

    All length measurements are in µm. Measurements were made as described in Experimental procedures. SD, standard deviation; NA, not applicable.

  • b.

    +, cells grown with arabinose to induce FtsN expression; –, cells grown with glucose to repress FtsN expression.

  • c.

    The ratio of cell length to number of rings was determined for each filament. The average (±SD) of the ratios was then calculated for each cohort of cells. SD reflects variation in the number of rings per filament.

  • d.

    Total length (column 8) divided by total number of rings (column 7). The spacing is inversely proportional to the frequency of rings.

FtsZJOE58610 µM+
73
100
643
72
469
249
2395
2620
3.3 ± 1.03.9 ± 0.93.7 ± 0.9
36 ± 14
11.7 ± 4.55.1
11
FtsZJOE58650 µM+
93
100
278
41
259
206
1088
1613
3.0 ± 0.74.0 ± 1.03.9 ± 1.0
39 ± 12
8.1 ± 1.44.2
7.8
FtsAJOE58780 µM+
88
100
688
82
603
368
2441
3124
2.8 ± 0.73.6 ± 0.93.5 ± 0.9
38 ± 16
8.9 ± 2.04.0
8.5
ZipAJOE66480 µM+
78
100
635
89
493
359
2300
3680
3.0 ± 0.73.8 ± 0.83.6 ± 0.9
41 ± 16
11.3 ± 4.74.7
10
FtsQJOE58510 µM+
80
100
629
69
505
231
72568
2572
3.2 ± 0.64.3 ± 1.04.1 ± 1.0
37 ± 18
11.8 ± 4.85.1
11
FtsLJOE60210 µM+
75
100
600
67
450
204
2115
2644
2.8 ± 0.63.8 ± 0.83.2 ± 0.8
39 ± 15
14.5 ± 5.94.7
13
FtsIJOE59410 µM+
74
100
563
67
417
185
2136
2235
3.1 ± 0.64.0 ± 0.93.8 ± 1.0
33 ± 11
13.3 ± 5.35.1
12

When FtsN expression was repressed in the presence of glucose, all GFP fusions also localized to potential division sites in the filaments (Fig. 2A–D; Table 2). Some filaments exhibited more frequent ring structures than others, but all filaments examined had at least one ring; most showed multiple rings. Calculation of the spacings between rings (last column of Table 2) revealed that the frequencies of localization signals decreased about two to threefold in the filaments compared with the normally dividing cells, regardless of the GFP fusion. Z rings were observed more frequently in both arabinose and glucose cultures when the level of FtsZ–GFP was increased by higher induction (50 µM IPTG), but the relative difference in ring frequencies between the two cultures remained similar, about twofold. Thus, the presence of FtsN may have a stabilizing effect on the formation of ring structures for all cell division proteins examined (see Discussion). Nevertheless, the results indicate that FtsZ, FtsA, ZipA, FtsQ, FtsL and FtsI do not require FtsN for localization to the division site.

Co-localization with FtsZ

Knowing that FtsQ, FtsL and FtsI can localize to potential division sites in FtsN-depleted filaments, we wanted to determine whether the sites in which they are seen are defined by the presence of Z rings. Two scenarios are possible in filaments without FtsN: first, FtsQ, -L and -I always co-localize with Z rings; or second, they can localize to sites without Z rings. The second scenario could arise if, for instance, the absence of FtsN leads to instability and disassembly of the Z ring, while FtsQ, -L, and -I remain at the site by attaching to another septal substrate. Co-localization with FtsZ would lend support to the model that cell division proteins are sequentially assembled into a multimeric complex.

To assess co-localization of the cell division proteins, we constructed fusions to ECFP and EYFP, mutant GFPs with different absorption and emission spectra. We fused ECFP or EYFP to the carboxyl termini of FtsZ and FtsA. We also fused Myc-ECFP or Myc-EYFP to the amino termini of FtsQ, FtsL and FtsI; the Myc epitope tag was added at the amino termini of ECFP and EYFP to improve translational efficiency (see Experimental procedures). The gene fusions were placed under the control of an IPTG-regulated promoter and integrated in single copy into the chromosome at the lambda or the φ80 attachment site. In merodiploid cells expressing both the wild-type protein and one of these new fusions, we observed localization of the fusion proteins to the division site (data not shown). Furthermore, we saw localization in only the appropriate fluorescence channel. For instance, cells expressing FtsZ–EYFP or Myc–EYFP–FtsQ did not exhibit septal localization through the ECFP filter set (data not shown).

To test this co-localization system, we first looked at FtsN-depleted filaments expressing both FtsZ–ECFP and FtsZ–EYFP. The fluorescence signals were very dim due to the narrow bandwidths of the excitation and barrier filters for ECFP and EYFP. However, long exposure times allowed us to observe Z rings in the filaments: FtsZ–ECFP and FtsZ–EYFP generally localized to the same potential division sites (Fig. 3A and B). We noted that the localization signals did not always correspond in both fluorescence channels. A Z ring might be fainter in one channel compared with the other. Occasionally, a Z ring was observable in only one channel. A partial explanation for the occasional discrepancy is that the different wavelengths of ECFP and EYFP result in different focal planes for the two channels. Thus, a filament might be in focus through the EYFP channel but slightly out of focus in the ECFP channel. In addition, stochastic distribution of the fusion proteins may exacerbate the situation, causing one ring to have more molecules of the EYFP fusion than the ECFP fusion. Nevertheless, we observed co-localization of FtsZ–ECFP and FtsZ–EYFP in most cases, indicating that this method is feasible.

Figure 3.

FtsQ and FtsI co-localize with FtsZ in FtsN-depleted filaments. Left panels show fluorescence in the ECFP channel, whereas right panels show corresponding fluorescence in the EYFP channel. See Experimental procedures for growth conditions and microscopy techniques. Arrowheads point to localization signals that appear in both channels, whereas double arrowheads point to positions where signals appear in the ECFP but not the EYFP channel.

A and B. Cells expressing both FtsZ–ECFP and FtsZ–EYFP (JOE644).

C and D. Cells expressing both FtsZ–ECFP and Myc–EYFP–FtsQ (JOE648).

E and F. Cells expressing both FtsZ–ECFP and Myc–EYFP–FtsI (JOE645).

G and H. Cells expressing both FtsZ–ECFP and EYFP (JOE642).

To test the co-localization method further, we examined filaments expressing both FtsZ–ECFP and FtsA–EYFP, as various lines of evidence indicate that FtsZ and FtsA interact. Again, we observed co-localization in most cases in which we saw a signal at a potential division site (data not shown). As a negative control, cells expressing EYFP alone and FtsZ–ECFP were examined. Localization signals in the filaments were observed through the ECFP filter set (Fig. 3G), but only diffuse fluorescence was observed through the EYFP filter set (Fig. 3H). This result confirmed that signals in one channel did not bleed through to the other channel.

To determine whether the cell division proteins co-localize in a filament, we then looked at FtsN-depleted filaments expressing both FtsZ–ECFP and Myc–EYFP–FtsQ, –FtsL, or –FtsI. Fusions to FtsQ, FtsL, and FtsI all appeared to co-localize with the Z ring: wherever we observed a band in the EYFP channel (Fig. 3D and F; data not shown for FtsL), we found a corresponding signal in the ECFP channel (Fig. 3C and E; data not shown for FtsL). However, localization signals for FtsQ, FtsL and FtsI were weaker and less frequent than the signals for FtsZ, and we did not always find a band in the EYFP channel that corresponded to the presence of a Z ring in the ECFP channel. The discrepancies between the two channels were particularly dramatic for FtsL (data not shown) and FtsI (double arrowheads in Fig. 3E and F). These results were not surprising, though, because localization studies using GFP fusions had shown that Z rings are more easily detected than ring structures of FtsQ, FtsL or FtsI. Differences in localization frequencies are likely to be a consequence of differences in protein abundance. Despite the low magnitude of the fluorescence signals, our results suggest that FtsQ, FtsL and FtsI molecules aggregate with the Z ring and provide support for the existence of a divisome complex.

Discussion

In this study, we used GFP fusions to assess the septal localization of FtsZ, FtsA, ZipA, FtsQ, FtsL and FtsI in E. coli filaments lacking FtsK or FtsN. We found that all six proteins can localize to potential division sites in FtsN-depleted filaments, but only FtsZ, FtsA and ZipA can do so in FtsK-depleted filaments. Thus, FtsA and ZipA continue to be the only pair that can be recruited independently into the septal ring. There is no evidence of co-dependence for localization among the nine known E. coli cell division proteins. In addition, we demonstrated that FtsA, -Q, -L and -I co-localize with Z rings in FtsN-depleted filaments.

FtsK-depleted filaments: partial constrictions and nucleoid distribution

During the course of our manipulation to deplete cells of FtsK, we noticed that the appearance of ftsK filaments varied, depending on the genetic backgrounds of the strains. In the MC4100 background, indentations along the filaments were observed frequently; some filaments even looked like chains of cells. In contrast, these incomplete constriction sites were infrequent, almost absent, in MG1655- or C600-derived filaments. Both cellular morphologies have been observed in other studies of ftsK mutants (Diez et al., 1997; Draper et al., 1998; Wang and Lutkenhaus, 1998; Yu et al., 1998a). One possible explanation for the variation is that MC4100-derived cells are less sensitive than MG1655- and C600-derived strains to reduced levels of FtsK, and the indentations result from residual FtsK activity. In support of this argument, Wang and Lutkenhaus (1998) were able to eliminate occasional constrictions in FtsK depletion filaments by increasing repression of the complementing ftsK allele. We were unable to eliminate the indentations by reducing the copy number of the complementing ftsK gene, but our PBAD-ftsK construct may be leaky, even under repressing conditions. Another possible explanation for the appearance of indentations is that the reduced level of FtsK leads to inefficient resolution of chromosome dimers, and the septum cannot close completely due to the presence of DNA at the site. This possibility is less likely, as the presence of a chromosome at the constriction site has not hindered cytokinesis in other instances. For example, cells expressing only the N-terminal domain of FtsK continue to divide despite inappropriate chromosome segregation (Liu et al., 1998). In such instances, the chromosome is guillotined (Boyle et al., 2000; Hendricks et al., 2000).

Regardless of the strain background, we saw that nucleoids were distributed normally throughout the FtsK-depleted filaments, as in filaments depleted of other cell division proteins. At first glance, this observation seems contradictory to the chromosome resolution function of FtsK. If chromosomes cannot segregate properly, the nucleoids should not be evenly spaced along the filaments. However, the chromosome segregation defect becomes apparent only when truncated FtsK is produced, resulting in cytokinesis that proceeds without the resolving activity of FtsK's C-terminal domain (Liu et al., 1998; Recchia et al., 1999; Yu et al., 1998b). In our case, cells are unable to divide because ftsK::cat-Δ5 is a complete null allele, with the N-terminal domain of FtsK missing as well (Draper et al., 1998). Recchia et al. (1999) were able to visualize a chromosome segregation defect in ftsZ filaments, presumably because FtsK mislocalizes and fails to faciliate dimer resolution in such filaments, but they used chloramphenicol to condense the chromosome and accentuate the defect. We did occasionally observe aberrant nucleoid staining in the FtsK depletion cultures (for example, see Fig. 1H). In filaments of other cell division mutants, we have attributed this type of staining to cell lysis (Ghigo et al., 1999; Weiss et al., 1999). However, in ftsK mutants, the possibility exists that some of the aberrant staining may reflect improper chromosome segregation.

Reduced ring frequencies in filaments

Although the septal ring can be partially assembled in the absence of cell division, the frequency of its formation appears to be reduced in such filaments. For example, we calculated that, regardless of the protein being localized, the spacing between rings is increased about twofold in ftsN filaments compared with normally dividing cells. The frequency of rings also varies greatly from filament to filament in the same strain, as evidenced by the width of the variance when ring spacing in each filament is averaged for the entire cohort (see standard deviations of average unit length per ring in Tables 1 and 2). This phenomenon has been observed in other localization studies (Chen et al., 1999; Hale and de Boer, 1999). We consider various explanations for this reduction in ring frequency, some of which have been discussed before (Pogliano et al., 1997; Hale and de Boer, 1999). One obvious explanation is that the filaments have aberrant cell envelope physiology, and the integrity of their rings is more easily disrupted during chemical fixation compared with that of normally dividing cells. Fixation can be avoided by examining GFP fusions in live cells, but quantitation of ring formation in live filaments is technically challenging. There is no simple way to adhere large numbers of filaments to the same focal plane and keep them all healthy for the extended period of time needed to take the images.

A second explanation is that, as a filament elongates, the concentration of free FtsZ becomes insufficient to support multiple ring formations. Whereas the total concentration of FtsZ remains the same, most of it is sequestered by the existing, nonconstricting Z rings. There may not be a mechanism for limiting the amount of FtsZ that can be incorporated into a Z ring, and, as sites of nucleation, rings formed at earlier time points may continue to accrue newly synthesized FtsZ protein. As all other cell division proteins depend on FtsZ for localization, they will also only aggregate at the limited number of previously established sites. Another explanation for the reduced frequencies of rings is that GFP fusion proteins, used to visualize the septal rings, accumulate to different extents in the different rings. These experiments were done in merodiploid strains still expressing the wild-type protein. Some rings may contain less fusion proteins than others as a result of random distribution, making these rings fluoresce below the threshold of detection.

The availability of proteins appears to account partly for the reduction in ring frequency, as elevated expression of FtsZ–GFP increased the absolute frequency of Z rings observed in ftsK and ftsN filaments. However, even with higher induction, the spacing between Z rings was still larger in the filaments compared with dividing cells. Hence, protein availability is not a fully satisfactory explanation.

Another possibility, as postulated by Pogliano et al. (1997), is that the activity of a septal ring at one division site may affect the formation or stability of rings at other sites in the cell via some form of checkpoint control; the presence of an unconstricted ring may prevent the formation of rings nearby. They also pointed to MinCD as a known regulator of Z ring assembly. As the oscillation of MinC, MinD and MinE appears to determine the site of septal ring formation, we suggest that the movement of these proteins may dictate the minimum distance between rings in the same cell. In a normally dividing cell, the range of motion for division inhibitors MinCD is limited by the length of the cell, and their inhibitory activity is countered by MinE movement, especially in the mid-cell region (Fu et al., 2001; Hale et al., 2001). However, in a filament, MinCD may inhibit septal ring formation over an extended range, whereas MinE can only counteract their effect in the same amount of space as in normal cells. The increased inhibitory range of MinCD would then account for the increased distance between septal rings. MinC and MinD do appear to have broader ranges of oscillation in FtsZ-inactivated filaments (Raskin and de Boer, 1999a; 1999b), but a more quantitative analysis is needed to determine the ranges of motion of MinC, -D and -E in filamentous cells. Regardless, the Min system clearly plays an important role in determining the frequency of septal ring formation: in min filaments, the spacing of FtsZ rings is significantly decreased, often with two or more rings between segregated nucleoids (Yu and Margolin, 1999).

Finally, the absence of a component in the septal ring may not prevent the assembly of upstream factors, but it may decrease the stability of the complex, resulting in reduced ring frequency. For instance, ZipA appears to stabilize FtsZ polymerization, both in vivo and in vitro (RayChaudhuri, 1999; Hale et al., 2000). FtsN may have a stabilizing effect as well, as overexpression of ftsN can suppress mutations in ftsA, ftsK, ftsQ and ftsI (Dai et al., 1993; Draper et al., 1998). Residual activity of the protein being inactivated or depleted would help explain why some localization signals are stronger than others. Those rings with stronger fluorescence may be more stable because they contain higher levels of the residual protein. This scenario and others described above are not mutually exclusive, and their combinatory effects may cause the reduced ring frequency seen in filaments.

Linear dependency pathway and sequential assembly

Table 3 summarizes results obtained in this and other labs regarding localization dependency among E. coli cell division proteins. Strikingly, a clear line cuts through the table, separating dependent (No's) from independent (Yes's) relationships. The independence of FtsA and ZipA localization stands out as an exception. These results suggest that a linear dependency pathway dictates the localization of FtsZ, FtsA and ZipA, FtsK, FtsQ, FtsL, FtsW, FtsI, and FtsN, in that order. The question remains whether ZipA is a branch point in the pathway: whether FtsK and downstream proteins require it for recruitment into the septal ring. Intriguingly, localization dependency among Bacillus subtilis cell division proteins is quite different from the mostly linear relationship in E. coli. In B. subtilis, interdependence for septal localization appears prevalent (Daniel et al., 1998; 2000; Daniel and Errington 2000; Katis et al., 2000). However, in E. coli, no case has been observed in which two proteins require each other for localization.

Table 3. Summary of localization dependency among E. coli cell division proteins.
Localization of:FtsN
w/oFtsZFtsAZipAFtsKFtsQFtsLFtsWFtsI
  1. (1) Addinall et al., 1996); (2) Addinall et al. (1997); (3) Addinall and Lutkenhaus, 1996); (4) Boyle et al. 1997); (5) Chen et al. (1999); (6) Ghigo et al. (1999); (7) Hale and de Boer (1999); (8) Khattar et al. (1997); (9) Liu et al. (1999); (10) Ma and Margolin (1999); (11) Pogliano et al. (1997); (12) Wang et al. (1998); (13) Wang and Lutkenhaus (1998); (14)Weiss et al. (1999); (15) Yu et al. (1998a); (16) Chen and Beckwith (this study); (17) Mercer and Weiss (personal communication)

FtsZNo (3,7,10)No (7,9,10)No (13,15)No (5)No (6)No (17)No (12,14)No (2)
FtsAYes (1,7)Yes (7,9)No (13,15)No (5)No (6)No (17)No (12,14)No (2)
ZipAYes (7,9)Yes (7,9)      
FtsKYes (13,15,16)Yes (13,15,16)Yes (13,16)No (16)No (16) No (16) 
FtsQYes (1)Yes (3)Yes (5)Yes (13)No (6)No (17)No (14)No (2)
FtsLYes (6)Yes (6)Yes (6) Yes (5)No (17)No (14) 
FtsWYes (4,8,17)Yes (17)Yes (17) Yes (17)Yes (17)No (17) 
FtsI
No (2,12)
Yes (1,11,12,14)Yes (3,12,14)Yes (9,14)Yes (12,13,15)Yes (5)Yes (6)Yes (17) 
FtsNYes (2,16)Yes (16)Yes (16) Yes (16)Yes (16) Yes (16)

The linear dependency pathway in E. coli may reflect the sequence in which the divisome is assembled. Our co-localization results suggest that FtsQ, FtsL and FtsI participate in this putative cell division complex; they aggregate at a site established by FtsZ. Furthermore, the observation that FtsQ co-localizes with FtsZ more frequently than FtsL or FtsI may mirror the temporal sequence of the divisome assembly. It is more probable that potential division sites contain FtsQ than FtsL or FtsI, if FtsQ joins the septal ring at an earlier stage. However, methods other than localization will be required to assess the validity of this model. To date, only ZipA and FtsA have been shown to interact directly with FtsZ (Hale and de Boer, 1997; Wang et al., 1997; Din et al., 1998; Liu et al., 1999; Ma and Margolin, 1999; Hale et al., 2000; Mosyak et al., 2000; Haney et al., 2001). Additional evidence for direct interaction between known components of the septal ring is needed to validate the existence of the divisome complex. Moreover, components of the septal ring probably exist that have not yet been identified. The identities of these unknown components may be essential to establishing the links between known components of the divisome.

Whereas sequential assembly of a multimeric complex is the simplest model for the linear dependency pathway, other scenarios can be invoked to explain the existing data. For example, FtsQ may alter the septal peptidoglycan, providing a substrate that allows localization of FtsI, but the two proteins may not be in the same complex. In addition, the sequence of localization may differ from the order in which the proteins perform their respective functions. For instance, FtsZ localizes first, but it cannot constrict until FtsI starts synthesizing septal peptidoglycan. Nevertheless, the localization results provide a starting point for understanding how the cell division proteins contribute to cytokinesis, and for elucidating the molecular details of this central cellular process.

Experimental procedures

Bacterial strains, plasmids and media

Bacterial strains and plasmids used in this study are listed in Table 4. Rich media used were NZY liquid broth and agar, which differ from Luria–Bertani (LB) in that tryptone (Miller, 1972) is replaced by NZ amine A (Quest International) and the salt concentration is 8 g of NaCl l−1. M63 (Miller, 1972) was used for minimal media, with thiamine (1 µg ml−1), desired amino acids (50 µg ml−1), and the appropriate carbon source (0.2%). Concentrations of antibiotics used in rich media are as follows: ampicillin, 200 (for bla on plasmid), 100 (for pAH123), or 25 (for bla on chromosome) µg ml−1; chloramphenicol, 10 µg ml−1; kanamycin, 40 µg ml−1; spectinomycin, 50 (in liquid) or 100 (in plate) µg ml−1; and tetracycline, 15 µg ml−1. d-Glucose or l-arabinose was added to repress or induce, respectively, the expression of genes under the control of the PBAD promoter (Guzman et al., 1995).

Genetic and molecular biology procedures

Standard techniques were used for cloning and analysis of DNA, polymerase chain reaction (PCR), electroporation, transformation and transduction with P1 phage (Sambrook et al., 1989; Miller, 1992). Stable introduction of genes into the E. coli chromosome at the lambda attachment site via λInCh was performed as described (Boyd et al., 2000). Enzymes used to manipulate DNA were from New England BioLabs or Gibco Life Technologies. Oligonucleotides were obtained from Genosys or Gibco Life Technologies and listed in Table 5. DNA sequencing was performed by the Micro Core Facility at the Department of Microbiology and Molecular Genetics, Harvard Medical School, or by the Tufts University Core Facility.

Table 5. List of primers used.
Primer nameSequenceSite underlined
ftsN5′5′-CATGGAGCTCTGACGAACGAATAAATACAG-3′ SacI
ftsN3′5′-CAATTCTAGATGGGGGGGATTTTGAGGGTT-3′ XbaI
ECFP5′-NcoI5′-CGGGTACCGGTCGCCACCAT-3′(NcoI site is downstream)
ECFP3′-EcoRI5′-GCCGAATTCCTTGTACAGCTCGTCCATGC-3′ EcoRI
ECFP5′-PstI5′-ATACCTGCAGATGGTGAGCAAGGGCGAGGAG-3′ PstI
ECFP3′-HindIII5′-TGCGAAGCTTTACTTGTACAGCTCGTCCAT-3′ HindIII
ECFP-seq25′-TCGATGTTGTGGCGGATCTT-3′ 
Myc-ECFP5′5′-CAGACCATGGAACAGAAACTGATTTCTGAAGA-
AGATCTGCTGATGGTGAGCAAGGGCGAGGA-3′
NcoI

Chromosomal integration at the φ80 att site

Conditionally replicative plasmids carrying the γ replication origin of R6K (denoted oriRR6Kγ) were maintained in BW25141 (Datsenko and Wanner, 2000), which expresses the π protein from a chromosomal copy of the pir gene and allows replication of such plasmids (Kolter et al., 1978). Plasmids containing attP and oriRR6Kγ were integrated into the chromosome of pir cells expressing the appropriate integrase, and the resulting strains were verified by PCR to have a single integrated plasmid with procedures similar to those described elsewhere (Haldimann et al., 1996; 1997; Lu et al., 1998). Specifically, pJC69-derived plasmids were integrated at the φ80 att site by transformation into a pir strain carrying pAH123. Transformed cells were incubated at 42°C for 30 min and 37°C for 1 h to induce expression of intφ80 and to inhibit replication of pAH123. Integrants were selected on plates containing spectinomycin at 37°C. Single integration event was confirmed by PCR.

FtsK and FtsN depletion strains

FtsK and FtsN depletion strains were obtained by transduction with P1 phage. The ftsK::cat-Δ5 or ftsN::kan null allele was transduced into a strain carrying pJC85 or pJC2 respectively. Transductants were selected on plates containing the appropriate antibiotics and l-arabinose to induce expression of the complementing allele from the plasmid. Transduction was confirmed by PCR. Construction of pJC85 is described in Table 4. In JOE593 and JOE600, pJC85 is integrated into the chromosome via λInCh1 (see above). To construct pJC2, we amplified ftsN from the E. coli chromosome with primers ftsN5′ and ftsN3′. The PCR product was digested with SacI and XbaI and ligated into the same sites of pBAD18.

Construction of gene fusions to ecfp or eyfp

Three of the plasmids for making gene fusions to ecfp or eyfp, pJC50, pJC51 and pJC52, were constructed with procedures similar to those used to obtain pDSW207 and pDSW208 (Weiss et al., 1999). To construct pJC50 and pJC51, plasmids for making fusions to the amino terminus of the target protein, we amplified ecfp and eyfp, respectively, from pECFP and pEYFP, with primers ECFP5′-NcoI and ECFP3′-EcoRI; the products were digested with NcoI and EcoRI and ligated into the same sites of pDSW204. Plasmid pJC52, for making ECFP fusion to the carboxy terminus of the target protein, was constructed using primers ECFP5′-PstI and ECFP3′-HindIII to amplify ecfp; the product was digested with PstI and HindIII and ligated into the same sites of pDSW204. Translation of the multiple cloning site (MCS) of pJC52 adds 16 amino acids to the N-terminus of ECFP. Construction of pJC58, for making EYFP fusion to the carboxy terminus of the target protein, is described in Table 4.

Translational fusions of cell division genes to ecfp or eyfp were first constructed to be exactly the same as the gfp fusions, except that the gfp open reading frame (ORF) was replaced by those of ecfp and eyfp(Table 4). The fusions were placed into the chromosome of wild-type cells at the lambda or φ80 attachment site to generate merodiploid cells, which were examined by fluorescence microscopy. We were able to see septal localization of ECFP and EYFP fusions to FtsZ or FtsA, but not of fusions to FtsQ, FtsL or FtsI (data not shown). Immunoblot analysis with anti-FtsQ revealed that neither ECFP–FtsQ nor EYFP–FtsQ were expressed at detectable levels under conditions used to examine green fluorescent protein (GFP)–FtsQ localization (data not shown). The situation was puzzling until we noticed that the lack of localization signal occurred with fusions to the carboxy terminus of ECFP or EYFP. Poor expression of the fusions may have resulted from inefficient translation, as a consequence of the humanized codons of ECFP and EYFP.

To improve translational efficiency, we changed the DNA sequence at the beginning of the ORF by adding a myc tag to the 5′-end of ecfp and eyfp. To obtain pJC76, which is used for making gene fusions to myc-eyfp, we amplified the 5′-end of eyfp from pEYFP with primers ECFPseq2 and Myc-ECFP5′, which contains sequence coding for the Myc epitope tag. The PCR product was digested with NcoI and BseRI and ligated into the same sites of pJC51. Construction of pJC82, which is used for making gene fusions to myc-ecfp, is described in Table 4. The N-termini of Myc–ECFP and Myc–EYFP have the sequence MEQKLISEEDLLMVSK, where MVSK are the first four residues of ECFP and EYFP. Chromosomal integration of constructs that expressed FtsI, FtsL or FtsQ fused to the carboxy terminus of Myc-ECFP or Myc-EYFP were done as before. We were able to observe septal localization of these fusion proteins.

Growth conditions

To deplete cells of FtsK, cells were first grown overnight in rich media containing appropriate antibiotics plus arabinose. Overnight cultures were diluted 1:25 in NZY media, inoculated 1:50–1:100 into the same media containing IPTG (isopropyl-β-d-thiogalactopyranoside) and either glucose or arabinose (plus spectinomycin to select for maintenance of pJC85, if present), and grown until OD600 reached 0.3, when cells in the glucose cultures became filamentous. Cells were grown at either 30°C or 37°C.

To deplete cells of FtsN, cells were grown overnight in rich media containing appropriate antibiotics plus arabinose. Overnight cultures were inoculated 1:250 into rich media containing chloramphenicol, and either glucose or arabinose, and grown until OD600 reached 0.2–0.3. The starter cultures were then inoculated 1:50–1:100 into the same media containing IPTG and grown until OD600 reached 0.2–0.3 again, when cells in the glucose cultures became filamentous. All cultures were grown at 30°C. For co-localization experiments, IPTG was added at a concentration of 10–25 µM.

Fluorescence microscopy

Cells were harvested, chemically fixed, and prepared for microscopy as described previously (Chen et al., 1999). Cells expressing ECFP fusions were not stained with DAPI (4′,6-diamidino-2-phenylindole) because the fluorescence signal of DAPI interferes with that of ECFP. To examine cells expressing GFP, we used an Axioskop 2 microscope (Carl Zeiss) equipped with a 63X plan-Apochromat oil immersion objective (numerical aperture, 1.4) and a 100 W mercury lamp. Filter sets to visualize GFP (HQ: FITC/Bodipy/Fluo3) and DAPI (UV) were from Chroma Technology Corp. Images were captured with an Orca, 12-bit, cooled CCD camera (C4742-95, Hamamatsu Photonics) and Carl Zeiss axiovision software. To examine cells expressing ECFP and EYFP, we used an Axioplan II microscope (Carl Zeiss) or a DeltaVision microscope system (Applied Precision Instruments) that were described elsewhere (Rolls et al., 1999). Images in Fig. 3 were captured with the DeltaVision microscope; exposure times were 25 and 30 s for EYFP and ECFP respectively. Images were exported as TIF files.

Cell length measurements and localization scoring were done as described previously (Chen et al., 1999). Cells with aberrant DAPI staining were not included in the quantitation (see Fig. 1 for an example). Images were processed in adobe photoshop for presentation.

Western blotting

Polyclonal antibodies against the periplasmic domain of FtsQ were used, and immunoblotting was done as described previously (Chen et al., 1999).

Acknowledgements

We thank Joe Lutkenhaus, William Donachie and Barry Wanner for generous provision of strains, plasmids and reagents. We also thank Melissa Rolls, Will Prinz, Pascal Stein and Marc Damelin for help with fluorescence microscopy, and Deborah Hogan for reviewing the manuscript. We are grateful to members of the Beckwith Lab for their assistance and encouragement. David Weiss kindly communicated results before publication. This work was supported by a National Science Foundation pre-doctoral fellowship and a grant (GM38922) from the National Institute of General Medical Sciences. JB was supported by an American Cancer Society Research Professorship.

Ancillary

Advertisement