Conservation of dynamic localization among MinD and MinE orthologues: oscillation of Neisseria gonorrhoeae proteins in Escherichia coli

Authors


Summary

Min proteins are involved in the correct placement of division septa in many bacterial species. In Escherichia coli (Ec) cells, these proteins oscillate from pole to pole, ostensibly to prevent unwanted polar septation. Here, we show that Min proteins from the coccus Neisseria gonorrhoeae (Ng) also oscillate in E. coli. Green fluorescent protein (GFP) fusions to gonococcal MinD and MinE localized dynamically in different E. coli backgrounds. GFP–MinDNg moved from pole to pole in rod-shaped E. coli cells with a 70 ± 25 s localization cycle when MinENg was expressed in cis. The oscillation time of GFP–MinDNg was reduced when wild-type MinENg was replaced with MinENg carrying a R30D mutation, but lengthened by 15 s when activated by MinEEc. Several mutations in the N-terminal domain of MinDNg, including K16Q and 4- and 19-amino acid truncations, prevented oscillation; these MinDNg mutants showed decreased or lost interaction with themselves and MinENg. Like MinEEc–GFP, MinENg–GFP formed MinE rings and oscillated in E. coli cells when MinDEc was expressed in cis. Finally, in round E. coli cells, GFP–MinDNg appeared to move in a plane parallel to completed septa. This pattern of movement is predicted to be similar in gonococcal cells, which also divide in alternating perpendicular planes.

Introduction

Cell division in bacteria involves the assembly of FtsZ and other key cell division proteins at the septum (Chen and Beckwith, 2001). This process is regulated by two different mechanisms: one involving nucleoid occlusion, the other involving the action of the Min proteins (Woldringh et al., 1990; Margolin, 2001a; Rothfield et al., 2001). The latter proteins, MinC, MinD and MinE, are present in many Gram-negative microorganisms including the bacillus Escherichia coli (de Boer et al., 1988) and the coccus Neisseria gonorrhoeae (Ramirez-Arcos et al., 2001). In the more extensively investigated E. coli (Ec) system, it is generally believed that MinC inhibits septum formation in dividing cells and is recruited to the membrane by the ATPase MinD (Rothfield et al., 2001). MinEEc is a bifunctional protein with two domains, the N-terminus that inhibits the action of MinCDEc at mid-cell, and the C-terminus that confers topological specificity to the division septum (Zhao et al., 1995). Although inactivation of MinC or MinD in E. coli produces a minicell phenotype indicative of cell division at polar regions, the overexpression of these proteins induces filamentous cells as a result of cell division inhibition at all potential division sites (de Boer et al., 1989). Using green fluorescent protein (GFP) fusions, MinCEc and MinDEc were shown to co-oscillate from one pole to the other in E. coli cells (Hu and Lutkenhaus, 1999; Raskin and de Boer, 1999a,b). MinCEc oscillation is dependent on MinDEc, and MinCDEc oscillation is also dependent on MinEEc (Hu and Lutkenhaus, 1999; Raskin and de Boer, 1999a,b). MinEEc–GFP forms a mobile ring that also oscillates along with MinCDEc but lags behind in a zone closer to mid-cell (Fu et al., 2001; Hale et al., 2001; Margolin, 2001b). The migration of MinEEc towards a polar zone is believed to stimulate MinDEc oscillation by inducing its ATPase activity (Hu and Lutkenhaus, 2001). ATP hydrolysis may result in dissociation of MinCDEc from the membrane at one pole and reassembly of the division inhibitors at the opposite polar zone (Hu and Lutkenhaus, 2001). The self-association of MinD has been reported (Szeto et al., 2001), further supporting the hypothesis that the ATPase activity of MinD, as well as its ability to oscillate, requires MinD dimerization (Hu and Lutkenhaus, 2001). Recently, MinDEc has been reported to polymerize upon binding to ATP and phospholipids. Interestingly, the addition of MinEEc caused the disassembly of the MinDEc array by stimulating its ATPase activity. It is thus proposed that MinEEc is the key factor for initiating MinDEc oscillation (Hu et al., 2002).

Initial studies with N. gonorrhoeae (Ng) have shown that MinC and MinD are also involved in cell division inhibition in this coccus. Disrupting the expression of either protein results in aberrant cell division and cell morphology, in some cases leading to cell lysis (Ramirez-Arcos et al., 2001; Szeto et al., 2001). Interestingly, heterologous expression of MinCNg and MinDNg in wild-type E. coli also induces filamentation, and E. coli minC and minD mutants were complemented by their gonococcal orthologues (Ramirez-Arcos et al., 2001; Szeto et al., 2001). These observations suggest conservation of MinC and MinD function across bacterial species with different morphologies, and prompted us to investigate whether the N. gonorrhoeae Min proteins might oscillate in both rod-shaped E. coli and round cells. Because of the limitations in localizing gonococcal Min proteins in N. gonorrhoeae cells as a result of their small size, we used round E. coli KJB24 to model what might be observed in gonococci. This E. coli strain divides in alternating perpendicular planes, as does N. gonorrhoeae (Westling-Häggström et al., 1977; Begg and Donachie, 1998). It has been shown that round E. coli cells can also divide in parallel planes if grown in Methocel medium (Cooper, 1997), or in planes alternating in three dimensions if grown under thymine limitation and mecillinam treatment (Zaritsky et al., 1999). Recent studies showed that GFP–MinDEc and GFP–MinEEc oscillate in E. coli KJB24 (Corbin et al., 2002). The fusion proteins almost always oscillated parallel to the long axis of the round cells (Corbin et al., 2002). Here, we demonstrate that GFP–MinDNg oscillates in both rod-shaped and round E. coli cells, although the oscillation is slower than that of GFP–MinDEc. The GFP–MinDNg oscillation is activated by MinE from either E. coli or N. gonorrhoeae. Various MinDNg and MinENg mutations and truncations were produced, and their contribution to the oscillation pattern of GFP–MinDNg is examined. Furthermore, MinENg–GFP also oscillates in E. coli cells when expressed with MinDEc and exhibits a movement that is similar to the oscillation of MinEEc–GFP. This is the first study to investigate the oscillation of Min proteins in heterologous backgrounds, specifically Min proteins from coccal species in cylindrical or spherical E. coli cells.

Results

GFP–MinD Ng oscillates in E. coli

We have shown recently that gonococcal MinD can both complement an E. coli minD mutant and arrest normal E. coli cell division when it is overexpressed (Szeto et al., 2001). However, to assess fully the heterologous function of MinDNg in E. coli backgrounds, we investigated whether a GFP–MinDNg fusion would oscillate in E. coli in a manner similar to E. coli MinD. Microscopic observations of E. coli WM1032 (Δmin, Table 1) cells expressing GFP–MinDNg along with untagged MinENg from pSR15 (Table 1) re-vealed a dynamic behaviour showing that fluorescence shifted from one polar zone to the other, and back to the initial polar zone, every 70 ± 25 s, depending on the length of the cells (Fig. 1, Table 2). We were able to record up to 10 oscillation events before the signal became too weak to detect. The oscillation of the gonococcal protein was slower than that of GFP–MinDEc which exhibited a movement of 37 ± 12 s from one polar zone to the other and back to the initial one (Table 2). GFP–MinDNg also oscillates in min + E. coli cells with similar oscillation periods to minE. coli cells (data not shown), indicating that the presence of MinDEc does not significantly inhibit the dynamic localization of the orthologous MinD.

Table 1. . Bacterial strains and plasmids used in this study.
Strains/plasmidsRelevant genotypeSource/reference
  • a . P trc was mutated at the −10 and −35 regions.

  • b . P ADH1 , yeast promoter, which regulates expression of the gal4 gene.

  • c

    . Encodes for the binding domain of GAL4

  • d

    . Encodes for the activation domain of GAL4

Strains
 E. coli DH5α[supE44ΔlacU169 (80 lacZ ΔM15) hsdR17 recA1 endA1 gyrA96 thi-1 relA1]Gibco
 E. coli WM1032MG1655 lacU169 ΔminCDE::kan Sun and Margolin (2001 )
 E. coli KJB24W3110 [rodA (Am)] Begg and Donachie (1998)
 E. coli C41(DE3)[FompT hsdSB(rB- mB-) gal dcm Δ(srl-recA) 306::Tn10 (tetR) (DE3)] Miroux and Walker (1996)
Plasmids
 pDSW209 AmpR Ptrca::gfp Weiss et al. (1999 )
 pWM1255 AmpR Ptrc::gfp–minDEc, minEEc Corbin et al. (2002 )
 pWM1255NgEAmpR Ptrc::gfp–minDEc, minENgThis study
 pSR15AmpR Ptrc::gfp–minDNg, minENgThis study
 pSR15ΔEAmpR Ptrc::gfp–minDNgThis study
 pSR15ER30DAmpR Ptrc::gfp–minDNg, minENg (R30D)This study
 pSR15EcEAmpR Ptrc::gfp–minDNg, minEEcThis study
 pJDE1AmpR Ptrc::gfp–minDNg (K16Q), minENgThis study
 pJDE2AmpR Ptrc::gfp–minDNg (N-4aa), minENgThis study
 pJDE3AmpR Ptrc::gfp–minDNg (N-19aa), minENgThis study
 pWM1079CmR Para::minC′Ec (last 80 bp), minDEc, minEEc-gfp Sun and Margolin (2001 )
 pWM1079-HCmR Para::minC′Ec, minDEc, minEEc–gfp (HindIII religate)This study
 pWM1079NgECmR Para::minC′Ec, minDEc, minENg–gfpThis study
 pSRE-GFPCmR Para::minC′Ng (last 100 bp), minDNg, minENg–gfpThis study
 pSREcE-GFPCmR Para::minC′Ng, minDNg, minEEc–gfpThis study
 pGBT9AmpR PADH1b::gal4(BD)cClontech
 pGAD424AmpR PADH1::gal4(AD)dClontech
 pGBT9minDAmpR PADH1::gal4(BD)-minDNg Szeto et al. (2001 )
 pGADminDAmpR PADH1::gal4(AD)-minDNg Szeto et al. (2001 )
 pGBT9minEAmpR PADH1::gal4(BD)-minENgThis study
 pGADminEAmpR PADH1::gal4(AD)-minENgThis study
 pJminD17AmpR PADH1::gal4(BD)-minDNg (K16Q)This study
 pJminD8AmpR PADH1::gal4(BD)-minDNg (N-4aa)This study
 pJminD9AmpR PADH1::gal4(BD)-minDNg (N-19aa)This study
 pJminD16 AmpR PADH1::gal4(AD)-minDNg (K16Q)This study
 pJminD12AmpR PADH1::gal4(AD)-minDNg (N-4aa)This study
 pJminD13AmpR PADH1::gal4(AD)-minDNg (N-19aa)This study
 pET30aKanR PT7::6×HisNovagen
 pEC1KanR PT7::minENg-6×HisThis study
Figure 1.

GFP–MinD Ng localization in rod-shaped E. coli . Time-lapse fluorescence images of rods of E. coli WM0132 (Δ min ) expressing GFP–MinD Ng from pSR15 (P trc :: gfp-minD Ng , minE Ng ). Times are indicated in seconds. Arrows show the shift of GFP–MinD Ng from pole to pole for three different cells: cells 1, 2 and 3 with complete oscillation cycles of 30, 45 and 60 s respectively. DIC, differential interference contrast imaging.

Table 2. . Oscillation patterns of GFP–MinD Ng and MinE Ng –GFP.
PlasmidGenotype    N (%) n Average oscillation cycle (range) (s)a
  • a

    . A complete cycle of fusion protein oscillation involves movement from one polar zone to the opposite one, and back to the initial polar zone.

  • b . Unpaired t -test analysis of the difference between the oscillation periods of cells expressing GFP–MinD Ng from pSR15 and pSR15ER30D yielded P  = 0.001.

  • N , total number of cells observed in two microscopy fields; %, percentage of the total number of cells that oscillated; n , number of cells analysed to obtain the average of oscillation periods; NA, not applicable; NE, the oscillation period was ‘not examined’ because of under- or overexpression of the fusion protein.

pSR15Ptrc::gfp–minDNg, minENg117 (97.4)1470 ± 25 (30120)b
pSR15ΔEPtrc::gfp–minDNgNANANo oscillation
pSR15EcEPtrc::gfp–minDNg, minEEc23 (78.2) 985 ± 27 (60120)
pSR15ER30DPtrc::gfp–minDNg, minENg(R30D)135 (96.3)1440 ± 15 (3060)b
pWM1255Ptrc::gfp–minDEc, minEEc22 (99.9)1037 ± 12 (2463)
pWM1255NgEPtrc::gfp–minDEc, minENgNANANE
pJDE1Ptrc::gfp–minDNg (K16Q), minENgNANANo oscillation
pJDE2Ptrc::gfp–minDNg (4aa N-terminal truncation), minENgNANANo oscillation
pJDE3Ptrc::gfp–minDNg (19aa N-terminal truncation), minENgNANANo oscillation
pSRE-GFPPara::minDNg, minENg–gfpNANANE
pSREcE-GFPPara::minDNg, minEEc–gfpNANANE
pWM1079-HPara::minDEc, minEEc–gfp69 (74.0) 7124 ± 14 (105150)
pWM1079NgEPara::minDEc, minENg–gfp54 (66.7) 7174 ± 18 (135190)

Either MinENg or MinEEc promote GFP–MinDNg oscillation

To determine the contribution of MinE to this oscillation pattern, gonococcal minE was deleted from pSR15, and the new plasmid, pSR15ΔE (Table 1), was transformed into E. coli WM1032. In these transformants, fluorescence was distributed throughout the cells, and GFP–MinDNg oscillation was not detected (Table 2). Although Western blots showed that GFP–MinDNg protein levels in these cells were similar to the protein levels in cells transformed with pSR15, the fusion protein had a slightly higher molecular weight (Fig. 2A). MinENg was not detected in WM1032 cells expressing pSR15ΔE (Fig. 2B). To investigate whether E. coli MinE was able to stimulate the GFP–MinDNg oscillation, N. gonorrhoeae minE was replaced with minEEc in pSR15, and the new plasmid, pSR15EcE (Table 1), was expressed in E. coli WM1032. Interestingly, the oscillation of GFP–MinDNg was activated by E. coli MinE, although the time for pole-to-pole movements was extended by 15 s (Table 2). MinD and MinE protein levels from these cells were comparable with those from cells expressing the control fusion protein encoded by pSR15, as shown by Western blotting (Fig. 2).

Figure 2.

Expression of GFP–MinD Ng fusions in E. coli WM1032.

A. Western blot using anti-MinD Ng antiserum.

B. Western blot using anti-MinE Ng antiserum. E. coli cells expressing GFP–MinD Ng from lanes: (1) pSR15 [P trc :: gfp-minD Ng , minE Ng ]; (2) pSR15ER30D [P trc :: gfp-minD Ng , minE Ng (R30D)]; (3) pSR15EcE [P trc :: gfp-minD Ng , minE Ec ]; (4) pSR15ΔE [P trc :: gfp-minD Ng ]; (5) pJDE1 [ gfp-minD Ng (K16Q), minE Ng ]; (6) pJDE2 [P trc :: gfp-minD Ng (4aa N-terminal truncation), minE Ng ] and (7) pJDE3 [P trc :: gfp-minD Ng (19aa N-terminal truncation), minE Ng ].

Replacement of E. coli minE with gonococcal minE in pWM1255 (GFP–MinDEc, Table 1) generated pWM1255NgE (Table 1), and cells transformed with this plasmid showed very faint fluorescence, making it impossible to detect any oscillatory movements of GFP–MinDEc (Table 2). We also created an R30D mutation in minENg, which is found at the junction between the anti-MinCD and the topological specificity domains of MinE proteins (King et al., 2000). The wild-type minENg was replaced with the R30D mutation in pSR15 to generate pSR15ER30D (Table 1). E. coli cells expressing this new plasmid exhibited more rapid GFP–MinDNg oscillations than cells with pSR15, registering shifts from one polar zone to the other, and back, every 40 ± 15 s (Table 2). The difference in oscillation periods between GFP–MinDNg expressed from pSR15 and pSR15ER30D was considered significant (P < 0.01, unpaired t-test; Table 2). Protein levels of MinDNg and MinENg expressed from pSR15ER30D were similar to those expressed from the control plasmid pSR15 (Fig. 2). Western blotting with anti-GFP antisera further verified that cellular levels of GFP–MinDNg variants were similar to those of the control GFP–MinDNg (data not shown).

Mutations in the N-terminal domain of MinD Ng affect the oscillatory movement of GFP–MinD Ng

The region(s) of MinD involved in its oscillation have not been determined. In this study, we selected various N-terminal MinDNg mutants to determine the role of the N-terminus of MinDNg in the oscillation of GFP–MinDNg in E. coli cells. Sequence alignment showed a highly conserved lysine residue at position 16 of MinDNg in the ATP-binding domain of this protein (Szeto et al., 2001). Therefore, a K16Q mutation was inserted into MinDNg, which was fused to GFP generating pJDE1 (Table 1). In addition, four-amino-acid (N-4aa) and 19-amino-acid (N-19aa) truncations of the N-terminal domain were constructed to either leave or remove the ATP-binding domain of MinDNg (amino acids 10–17) respectively. These truncated MinDNg variants were fused to GFP creating pJDE2 and pDE3 respectively (Table 1). In each of these constructs, MinENg was expressed in cis. E. coli cells expressing the GFP fusions to the modified MinD proteins did not show any oscillatory movements, and fluorescence was distributed throughout the cells (Table 2). Western blots were performed to determine whether MinDNg and MinENg were expressed in these cells at similar levels to those from the control pSR15 encoding wild-type GFP–MinDNg. Levels of mutant MinDNg proteins from pJDE1 and pJDE2 were comparable with those from pSR15 (Fig. 2A). Slightly less expression was observed from cells containing the N-19aa truncation from pJDE3 (Fig. 2A). Similar results from Western blots were obtained when the membranes were probed with anti-GFP antisera (data not shown). MinENg protein levels from cells expressing any of the MinDNg mutations were comparable with those of cells expressing wild-type MinDNg from pSR15 (Fig. 2B).

We used the yeast two-hybrid system to determine whether the failure of the proteins to oscillate may have been caused by a lack of interaction between mutant MinDNg and MinENg or between the mutant MinDNg proteins with themselves. As shown in Table 3, interaction between MinDNg and MinENg was lost with the K16Q MinDNg mutant and with both N-4aa and N-19aa truncated MinDNg proteins. Similarly, there were no self-interactions between the truncated MinDNg mutants (Table 3). However, the K16Q mutant interacted weakly with itself (Table 3). Mutant MinDNg proteins were tested for their interactions with wild-type MinDNg as a control to ensure that the mutants were actually synthesized in the yeast cells. The only interaction that was lost was between wild-type MinDNg and the N-19aa truncated MinDNg (data not shown).

Table 3. . Effects of MinD mutations on Min protein interactions.
Fusion to GAL4
DNA-BD
Fusion to GAL4
DNA–AD
Interaction detected
(colony lift)
β-galactosidase activity
(Miller units)
  1. a . Mutations in MinD Ng .

MinDNgMinDNgYes 27.49 ± 5.28
MinENgMinDNgYes 12.94 ± 3.25
MinDNg(K16Q)aMinDNg(K16Q)aYes  1.68 ± 0.09
MinENgMinDNg(K16Q)aNo  0.08 ± 0.06
MinDNg(N-4 aa)aMinDNg(N-4 aa)aNo 0.033 ± 0.002
MinENgMinDNg(N-4 aa)aNo  0.00
MinDNg(N-19 aa)aMinDNg(N-19 aa)aNo0.0325 ± 0.002
MinENgMinDNg(N-19 aa)aNo 0.011 ± 0.015

Gonococcal MinE–GFP localizes dynamically in E. coli cells

The oscillation of MinDNg in the presence of MinENg indicated that gonococcal MinE is functional in E. coli as well. Therefore, to observe whether MinENg oscillates in E. coli, we made several constructs containing MinE from either E. coli or N. gonorrhoeae in fusion with GFP. Initially, E. coli minE was replaced with N. gonorrhoeae minE in pWM1079-H (MinEEc–GFP, Table 1), leaving MinDEcin cis (pWM1079NgE, Table 1). E. coli WM1032 cells transformed with pWM1079NgE showed gonococcal MinE–GFP oscillation. The oscillation cycle, involving formation of the gonococcal MinE ring, movement to one polar zone, dissolution, reformation, movement to the opposite polar zone and then back to the initial polar zone, took 174 ± 18 s (Fig. 3, Table 2). This cycle is longer than that observed in E. coli cells expressing E. coli MinE–GFP from pWM1079-H, which took between 124 ± 14 s to complete a similar oscillation event (Table 2). The slow oscillation of MinENg–GFP relative to GFP–MinDNg probably results from partial inactivation of MinENg by the GFP tag (Hale et al., 2001), which might slow the enhancement of ATP hydrolysis by MinD (Hu and Lutkenhaus, 2001).

Figure 3.

MinE Ng –GFP localization in E. coli . Time-lapse fluorescence images of E. coli WM1032 (Δ min ) expressing MinE Ng –GFP from pWM1079NgE (P ara :: minD Ec , minE Ng - gfp ). Times are indicated in seconds.

Gonococcal MinE oscillation in cells expressing pWM1079NgE depended on the co-expression of untagged E. coli MinD. To co-express MinENg–GFP and MinDNg from the same plasmid, we constructed pSRE-GFP (Table 1). E. coli cells transformed with this plasmid showed very bright fluorescence throughout the cells, even without induction, making it impossible to monitor any oscillation that may have occurred (Table 2). Cells in which gonococcal minE was replaced with E. coli minE in pSRE-GFP generating pSREcE-GFP (Table 1) showed very faint fluorescence, which also prevented any observation of oscillatory movements (Table 2). Western blotting using anti-MinENg and anti-GFP antisera confirmed very high levels of expression in cells containing pSRE-GFP and low expression levels in cells containing pSREcE-GFP (data not shown).

GFP–MinDNg oscillates in round E. coli cells

We analysed the effect of expressing GFP–MinDNg in round E. coli cells in an effort to model what might be observed in gonococcal cells. E. coli KJB24 (rodA, Table 1), was selected as it divides in alternating per-pendicular planes, as does N. gonorrhoeae (Westling-Häggström et al., 1977; Begg and Donachie, 1998). When these round cells expressed GFP–MinDNg from pSR15 or pSR15ER30D, two cell populations were observed. Approximately 20% were normal-sized cells, probably representing younger cells, whereas the majority were enlarged cells probably caused by overexpression of GFP–MinDNg. Similar observations were obtained when MinDNg was overexpressed in gonococcal cells (Szeto et al., 2001). In the larger cells, the axis of GFP–MinDNg oscillation appeared to be oriented randomly, and the fluorescence moved from one point to another in the cells without any obvious pattern (Fig. 4A). This is similar to the behaviour of GFP–MinDEc in symmetrical cells of KJB24 (Corbin et al., 2002). However, in normal-sized cells, the oscillation of GFP–MinDNg in diplococci with apparently completed septa was characterized by a back and forth motion parallel to the septum, with a 15 s time lapse from one end of the cell to the other (Fig. 4B). This oscillatory movement is represented in our proposed model of gonococcal MinD localization (double-headed arrows in Fig. 5C). A similar oscillatory behaviour of GFP–MinDEc was observed in KJB24 cells that had just completed septation (Corbin et al., 2002).

Figure 4.

GFP–MinD Ng localization in round E. coli . Time-lapse fluorescence images of round E. coli KJB24 ( rodA ) expressing GFP–MinD Ng from pSR15 (P trc :: gfp-minD Ng , minE Ng ).

A. Enlarged cells showing apparently unco-ordinated movements (arrows show shifts in one cell).

B. Normal-sized cells showing a defined oscillation pattern parallel to the septum (arrows show shifts in one cell).

Times are indicated in seconds. DIC, differential interference contrast imaging.

Figure 5.

Model for Min action in N. gonorrhoeae .

A. MinD Ng -ATP (circles) localizes to one end of the coccus and recruits MinC Ng (squares) to inhibit division there. Stimulation by MinE Ng (diamond) results in continued end-to-end oscillation of MinCD Ng (double-headed arrow). Division is also inhibited where the chromosomal DNA replication machinery (replisome, triangle) is located.

B. The replicating chromosome and oscillating MinCD Ng inhibit cell division at all sites except at a location opposite to the replisome (arrow), resulting in an initial invagination of the cells.

C. Completion of septation after chromosome segregation results in a MinCD Ng oscillation pattern perpendicular to the first one.

D. The new oscillation pattern and the attachment of the replisome to a new intracellular site allow division to proceed in only one location.

E. Tetrad formation as a result of two perpendicular division planes.

Discussion

We show that the dynamic behaviour of the Min system, so elegantly demonstrated in E. coli, probably exists in N. gonorrhoeae. We reported recently that gonococcal MinD was functional in E. coli, because it could complement an E. coli minD mutant, and its overexpression caused cell filamentation (Szeto et al., 2001). We also demonstrated that MinD from either N. gonorrhoeae or E. coli could self-interact (Szeto et al., 2001), providing evidence for the prediction that E. coli MinD oscillation requires its dimerization (Hu and Lutkenhaus, 2001). In the present study, we extended these findings by showing that GFP–MinDNg oscillates in E. coli Δmin cells in a similar fashion to E. coli MinD; however, the complete oscillation cycle is longer.

An outstanding question is whether the behaviour of GFP–MinDNg in coccal cells is also dynamic. Because of the difficulties in performing localization experiments in N. gonorrhoeae cells due to their small size, we used round E. coli KJB24, which divide in a manner similar to N. gonorrhoeae (Westling-Häggström et al., 1977; Begg and Donachie, 1998), to gain insight into this question. When GFP–MinDNg was expressed in round E. coli cells, the fusion protein also displayed dynamic localization patterns. In these cells, expression of the GFP–MinDNg fusion was obtained even without IPTG induction, and the majority of the cells showed increased size and an apparently unco-ordinated oscillatory movement of MinDNg. This phenomenon may be caused by a dosage effect, because strain KJB24 contains native MinD, and we have shown previously that overexpression of MinD in round cells caused them to enlarge (Szeto et al., 2001). A similar behaviour was observed in KJB24 cells, with no obvious long axis, expressing GFP–MinDEc (Corbin et al., 2002). However, in round cells that retained their original size, the localization of GFP–MinDNg followed a complete oscillation cycle (from one end of the cell to the other and then back to the initial end) of 30 s. This period is two or three times shorter than the oscillation period in rod-shaped E. coli, suggesting that some aspects of cell geometry influence oscillation times. GFP–MinDNg moved parallel to the septum in round cells undergoing septation, suggesting that the next plane of division, the medial zone of the perpendicular plane, might be left free for septum formation. Similar oscillatory movements were observed upon expression of GFP–MinDEc in KJB24 diplococci that were in the last stages of septation (Corbin et al., 2002). Although these results suggest that the ability of MinD and MinE to move is independent of cell shape and species-specific protein accessory factors, cell size might be a determinant for co-ordinated or apparently unco-ordinated oscillation patterns. Moreover, the results suggest that Min protein movement may adapt to different division patterns of species with different morphologies. Therefore, we predict that, if MinD from any species is expressed along with MinE, and the accessory factors required for oscillation, such as membrane phospholipids, are present, it should display dynamic localization. For example, it would be interesting to see whether Bacillus subtilis MinD, which does not oscillate in its host (Marston et al., 1998), might be able to display intracellular motility when expressed with MinE from any other species.

Round E. coli cells can act as a heterologous host for the gonococcal Min system. Based on the oscillatory movements of GFP–MinDNg in regular-sized E. coli KJB24 cells, as well as on previous evidence that there is a synergistic relationship between MinCNg and MinDNg (Szeto et al., 2001), we propose a preliminary model for Min protein function in N. gonorrhoeae (Fig. 5). This model represents gonococcal cells containing only one nucleoid; however, further studies need to be done in order to determine whether these cells are single or multinucleate. In Fig. 5A, MinD is shown oscillating from one end, or hemisphere, of a coccus to the other. MinC is recruited by MinD, resulting in division inhibition at one cell half. Stimulation of the ATPase activity of MinD by MinE at either cell half would lead to the continued oscillation of MinCD. The nucleoid itself would probably play a significant role in determining a division site (Yu and Margolin, 1999; Corbin et al., 2002). Recent observations in E. coli and B. subtilis have shown that the bacterial chromosomal DNA replication machinery (replisome) is localized at mid-cell, suggesting that replicating chromosomes are threaded through a stationary replisome (Lemon and Grossman, 1998; Koppes et al., 1999). In cocci such as N. gonorrhoeae, the presence of the replicating DNA surrounding the replisome would probably inhibit cell division until the nucleoids have been properly segregated (Fig. 5B; note bottom of cell). Together, the replicating chromosome and oscillating MinCD would inhibit cell division at all sites, except at a location opposite the replisome (Fig. 5B; white arrow), resulting in an initial invagination of the cell as septation proteins are assembled. This ‘one-sided’ invagination is highly characteristic of gonococcal cells (Fitz-James, 1964) and is frequently observed in our own electron microscope micrographs (data not shown). Completion of septation after chromosome segregation may then result in a MinCD oscillation pattern (Fig. 5C) perpendicular to the first (Fig. 5A), yet parallel to the newly formed cell wall that separates daughter cells. We observed such a pattern in the present study. This new oscillation pattern, and the attachment of the replisome to a new intracellular site, should again allow cell division to proceed in only one location, in this case at the interface between two daughter cells (Fig. 5D). Electron microscopic studies have shown that the second perpendicular division event in diplococcal N. gonorrhoeae is characterized by the co-invagination of cell membranes between each cell (Fitz-James, 1964). Ultimately, the result is a tetrad of daughter cells (Fig. 5E).

We found that the N-terminal region of MinD is essential for its oscillatory movement. A truncation of only four amino acids at this region resulted in loss of protein movement. This may result from the lost interaction of the truncated protein with itself and with MinENg, as shown in the yeast two-hybrid experiments. These results suggest that the first four amino acids of MinD are essential to obtain a proper protein configuration that allows Min protein interactions, and therefore oscillation, to occur. It is expected that similar results would be observed if MinD from other bacteria are also truncated at these residues as they are highly conserved among different species (Szeto et al., 2001). A mutation in the ATP-binding domain of MinDNg, K16Q, also resulted in loss of oscillation by the fusion protein. Similarly, the analogous K16A muta-tion of B. subtilis MinD resulted in loss of proper protein localization at the polar zones in B. subtilis cells (Karoui and Errington, 2001). The significant decrease in self-interaction with the MinDNg K16Q mutant suggests that the localization pattern of MinDNg may be affected by the alteration of its ATP binding or hydrolysis activities. Interestingly, the K16Q mutant lost its interaction with MinENg as well, demonstrating again that alteration of the N-terminal domain of MinDNg is directly involved in its interaction with MinENg. Because MinD proteins from different species display high sequence conservation, it would be interesting to test the contribution of middle or C-terminal domains to the intracellular motility of this protein.

Further evidence for the essential role of MinE in the localization of MinDNg comes from the expression of GFP–MinDNg from pSR15ΔE, in which minENg was deleted. These cells showed evenly distributed fluorescence and no oscillatory movement of the fusion protein. Although we could not ascertain whether GFP–MinDEc would oscillate in the presence of MinENg, the replacement of gonococcal minE with E. coli minE restored the movement of GFP–MinDNg in the E. coli cells, even though the cycle was slowed down. These results suggest that the interaction between MinD and MinE from different bacteria is probably not as efficient as interactions between proteins from the same species. This could conceivably lead to less efficient stimulation of MinD ATPase activity, which would result in a slower oscillation cycle or a loss of MinD movement altogether.

Several mutations at the N-terminus of MinEEc affect the oscillation of GFP–MinDEc by either inhibiting or slowing down fusion protein movement (Hu and Lutkenhaus, 2001). We decided to generate a new mutation in MinENg, R30D, and test its effects on the oscillation pattern of GFP–MinDNg. This mutation is found at the junction between the anti-MinCD and the topological specificity domains of MinE proteins from different species, including Neisseria and E. coli (King et al., 2000). Interestingly, when the mutant MinENg was expressed in cis with a GFP–MinDNg fusion, the oscillation time of the fusion protein decreased significantly. We speculate that the R30D mutation stabilizes either gonococcal MinE or its inter-action with MinDNg, resulting in increased stimulation of MinDNg movement.

A MinENg–GFP fusion with E. coli MinD expressed in cis exhibited MinE rings and an oscillatory pattern similar to those reported for E. coli MinE. These results suggest that the heterologous dynamic behaviour of the Min proteins is not restricted to MinD but can be extended to MinE, and probably to MinC as well.

We were able to extend the proposed models of Min functionality and showed that Min proteins from the Gram-negative coccus N. gonorrhoeae display distinct patterns of movement when expressed in E. coli. There are various recently reported theoretical models that explain why Min oscillation may occur (Howard et al., 2001; Meinhardt and de Boer, 2001; Kruse, 2002). Meinhardt and de Boer (2001) proposed that the pole-to-pole oscillation of the Min proteins arises from a self-organizing system that is independent of environment and prelocalized determinants. Our interspecies data demonstrate that N. gonorrhoeae Min proteins oscillate in E. coli independently of their native gonococcal environment.

The results presented in this paper demonstrate that Min proteins from different bacterial genera that vary in morphology and division patterns can have similar dynamic behaviours. This work extends our previous studies (Ramirez-Arcos et al., 2001; Szeto et al., 2001), further supporting E. coli as a model for examining the function of heterologous Min proteins. Because MinD and MinE from different species move, it appears that amino acid differences observed between E. coli and N. gonorrhoeae orthologues are not crucial for dynamic localization. Therefore, evolutionary changes in these proteins have been shown to have little effect on intracellular movement or MinD–MinE interactions. Our work also emphasizes the essential role of MinE in GFP–MinDNg oscillation. We also determined that the N-terminal domain of MinDNg is directly involved in the interaction of this protein with MinENg and with itself, and therefore in its oscillatory movement. We showed that E. coli MinE is able to activate gonococcal GFP–MinD and that MinENg–GFP is also functional in the E. coli system. Finally, based on our studies in round E. coli cells, we propose that MinDNg might oscillate in gonococcal cells in a manner that accommodates the typical division pattern of alternating perpendicular planes.

Experimental procedures

Strains and growth conditions

Bacterial and yeast strains used in this study are listed in Table 1. E. coli DH5α was used as a host to clone minDNg in fusion with GFP. E. coli strains WM1032 (ΔminCDE) and KJB24 (rodA) were used for the GFP–MinDNg, GFP–MinDEc, MinEEc–GFP and MinENg–GFP localization studies. E. coli C41(DE3) was used as an expression strain for the purification of MinENg−6×His. E. coli WM1032 cells expressing GFP–MinDNg or GFP–MinDEc fusions were grown at 32°C for 6–8 h at ≈ 250 r.p.m., whereas cells expressing MinEEc–GFP or MinENg–GFP fusions were grown at 28°C for 6–8 h at ≈250 r.p.m. For expression of GFP–MinDNg in E. coli KJB24, cells were grown overnight at room temperature without any agitation. All cylindrical E. coli cells were grown in Luria–Bertani (LB) medium (Difco) supplemented with 100 µg ml−1 ampicillin (Amp) and 40 µM IPTG when expressing GFP–MinD fusions, or with 20 µg ml−1 chloramphenicol (Cm) and 0.005% arabinose when expressing MinE–GFP fusions. Round KJB24 cells expressing GFP–MinDNg were supplemented with 50 µg ml−1 thymine and not induced with IPTG. E. coli cells were grown in LB supplemented with 50 µg ml−1 kanamycin (Kan) when expressing MinENg−6×His. Saccharomyces cerevisiae SFY526 was used in yeast two-hybrid assays to study Min protein interactions. Yeast were grown at 30°C on YPAD media or on the appropriate synthetic drop-out (SD) media, as described in the Clontech Yeast Two-Hybrid Manual.

Oligonucleotide primers, polymerase chain reaction (PCR) and inverse PCR (IPCR)

The oligonucleotide primers used for PCR and IPCR amplifications were designed using primer designer (Scientific and Education Software) and were synthesized by the University of Ottawa Biotechnology Research Institute. The sequences of these primers are available on request. PCRs were carried out in a Perkin-Elmer Gene Amp PCR system 9600 thermocycler using the following protocol: 3 min at 94°C; 30 cycles of denaturation for 15 s at 94°C, annealing for 15 s at temperatures varying from 48°C to 51°C (depending on the primer pair used) and extension at 72°C for 0.5–9 min (depending on the expected product size); and a final 5 min extension at 72°C. Reactions were carried out in a final volume of 100 µl containing the following reagents: 1× PCR buffer containing 1.5 mM MgCl2, 0.2 mM dNTPs, 0.2 µg of each primer and 2.5 U of Taq DNA polymerase for PCR amplifications (Roche) or Vent DNA polymerase (New England Biolabs) for IPCR amplifications. Cell suspensions of N. gonorrhoeae and E. coli were prepared by diluting cells from overnight cultures in double-distilled H2O. Cell concentrations were adjusted using the McFarland equivalence turbidity standard 0.5 (Remel) to provide chromosomal DNA templates for PCR. Plasmid DNA templates were adjusted to 0.01 µg ml−1 for IPCRs.

Construction of gonococcal and E. coli GFP–MinD fusions

To obtain a GFP fusion with the N-terminus of N. gonorrhoeae MinD and to provide MinENgin cis, minDNg and minENg were co-amplified using primers incorporating EcoRI and BamHI restriction sites at their 5′ and 3′ ends, respectively, and ligated into pDSW209 (Table 1) to generate pSR15 (Table 1).

To investigate the contribution of MinE to the localization patterns of GFP–MinD from either N. gonorrhoeae or E. coli, derivatives of pSR15 (GFP–MinDNg, Table 1) and pWM1255 (GFP–MinDEc, Table 1) were constructed. To delete minENg from pSR15, a 5703 bp fragment was IPCR amplified from pSR15 template using primers that annealed to regions flanking minENg. Each primer incorporated a BamHI site, allowing the IPCR amplicon to be religated to generate pSR15ΔE (Table 1). To replace minENg with minEEc in pSR15, a primer pair flanking minENg was used for IPCR to generate a linearized pSR15 derivative with minENg removed. A BamHI site was incorporated at one end of the amplicon, while the other end remained blunt ended (minDENg intergenic region). E. coli minE was PCR amplified with Vent polymerase using primers that would incorporate BamHI at the 3′ end of the gene, while the 5′ end remained blunt ended. The IPCR amplicon from pSR15 and E. coli minE were ligated at their compatible BamHI and blunt end sites to generate pSR15EcE (Table 1). Using a similar strategy, wild-type minENg in pSR15 was replaced with a variant gene containing a R30D mutation created by site-directed mutagenesis, generating pSR15ER30D (Table 1). Likewise, minEEc was replaced with minENg in pWM1255 (Corbin et al., 2002), resulting in pWM1255NgE (Table 1).

To assess the role of the N-terminus of MinDNg in the dynamic behaviour of this protein, three different constructs were made. Fragments of 1082 bp, 1070 bp and 1025 bp, each containing modified minDNg and wild-type minENg, were PCR amplified using primers incorporating EcoRI and BamHI restriction sites at the 5′ and 3′ ends of the amplicons respectively. The minDNg modifications included a K16Q mutation in the ATP binding site, and 4- and 19-amino-acid (aa) N-terminal truncations. Amplicons were ligated into pDSW209 producing plasmids pJDE1, pJDE2 and pJDE3 respectively (Table 1).

Construction of gonococcal and E. coli MinE–GFP fusions

A GFP fusion to the C-terminus of MinENg was constructed as follows. Plasmid pWM1079, which contains the last 80 bp of minCEc (minC′Ec) and all of minDEc and minEEc in fusion with GFP (Sun and Margolin, 2001), was digested with HindIII and religated, leaving a unique XhoI restriction site. The new plasmid, pWM1079-H (Table 1), was digested with SacI and XhoI to remove E. coli minC′, minD and minE. A 1200 bp fragment containing the last 100 bp of minCNg and all of minDNg and minENg was PCR amplified with a primer pair containing SacI and XhoI at the 5′ and 3′ ends respectively. This amplicon was ligated with SacI–XhoI-digested pWM1079-H to generate pSRE-GFP (Table 1). To replace minEEc with minENg in pWM1079-H, a primer pair that flanked minEEc and would be extended in opposite directions by IPCR, was used to generate a linearized pWM1079-H amplicon with minEEc removed. This amplicon contained a XhoI site in one end while remained blunt ended at the other. A fragment containing N. gonorrhoeae minE was PCR amplified with Vent polymerase using primers that would incorporate XhoI at the 3′ end of the gene while the 5′ end remained blunt ended (minDEEc intergenic region). The IPCR amplicon from pWM1079-H and N. gonorrhoeae minE were subsequently ligated at their compatible XhoI sites and blunt ends to generate pWM1079NgE (Table 1). Similarly, minENg was replaced with minEEc in pSRE-GFP, and pSREcE-GFP was created (Table 1).

Yeast two-hybrid assays

Modified minDNg containing a K16Q mutation in the ATP binding site, and 4aa and 19aa N-terminal truncations, was PCR amplified individually using primers incorporating EcoRI and BamHI restriction sites at the 5′ and 3′ ends respectively. Each gene was ligated in frame with the GAL4 DNA-binding domain (BD) and the GAL4 activation domain (AD) of pGBT9 and pGAD424 (Clontech) respectively. The plasmids obtained were pJminD17 (BD-MinDNg K16Q), pJminD8 (BD-MinDNg 4aa N-terminal truncation), pJminD9 (BD-MinDNg 19aa N-terminal truncation), pJminD16 (AD-MinDNg K16Q), pJminD12 (AD-MinDNg 4aa N-terminal truncation) and pJminD13 (AD-MinDNg 19aa N-terminal truncation) (Table 1). The minENg gene was also PCR amplified with primers incorporating EcoRI and BamHI restriction sites at its 5′ and 3′ ends respectively. This gene was fused to both domains of GAL4 generating plasmids pGBT9minE (BD-MinENg) and pGADminE (AD-MinENg) (Table 1). Fusions were confirmed to be in frame by DNA sequencing. The plasmids were transformed singly or in combination into S. cerevisiae SFY526, and yeast colonies were screened for β-galactosidase activity using the colony-lift and liquid assays as described before (Szeto et al., 2001).

Purification of MinENg-6xHis

To purify MinENg, the coding region of minENg was PCR amplified with primers incorporating NdeI and XhoI restriction sites at its 5′ and 3′ ends respectively. This gene was fused in frame to the C-terminal 6×His tag of pET30a (Novagen) generating pEC1 (Table 1). The fusion was confirmed to be in frame by DNA sequencing. A 250 ml log-phase culture of E. coli C41(DE3) carrying pEC1 was induced with 0.4 mM IPTG for 4 h at 37°C with shaking at 250 r.p.m. Protein purification was performed as described previously (Ramirez-Arcos et al., 2001; Szeto et al., 2001). The protein was eluted using elution buffer described by Novagen: 1 M imidazole, 0.5 M NaCl and 20 mM Tris-HCl (pH 7.9).

Western blots

Polyclonal anti-MinENg antiserum was produced and affinity purified according to similar protocols to those described for the production of anti-MinCNg and anti-MinDNg antisera (Ramirez-Arcos et al., 2001; Szeto et al., 2001). Cell extracts were prepared by resuspending cells in 100 µl of PBS, pH 7.4. Protein fractions were separated by SDS-PAGE, and protein concentrations were standardized before membrane transfer by densitometry as described previously (Ramirez-Arcos et al., 2001; Szeto et al., 2001). SDS-PAGE-resolved proteins were transferred to Immobilon-P membranes (Millipore), and Western blotting was conducted with a few modifications to previously described methods (Ramirez-Arcos et al., 2001; Szeto et al., 2001). For GFP–MinDNg detection, the membrane was incubated with 1:800 anti-MinDNg antiserum for 1 h at room temperature. For MinENg–GFP detection, the membrane was incubated with 1:100 anti-MinENg antiserum overnight at 4°C. For MinE detection in cells expressing GFP–MinDNg with untagged MinENg (or MinEEc), we used biotinylated anti-rabbit IgG (1:5000; Sigma) as a secondary antibody, followed by incubation with ExtrAvidin–alkaline phosphatase conjugate (1:150 000; Zymed). Protein samples from cells expressing GFP–MinDNg and MinENg–GFP fusions were also blotted and probed with anti-GFP antiserum (Clontech) diluted 1:100, and incubated overnight at 4°C, according to the same conditions as described above.

Microscopy

For fluorescence and differential interference contrast (DIC) imaging, ≈ 3 µl of log-phase-grown cells were mixed with 3 µl of 2% low-melting-point agarose in LB and immediately applied to a microscope slide. The subsequently immobilized cells were viewed with an Olympus BX60 fluorescence microscope and a 100× oil immersion objective, and images were captured with a Photometric CoolSnap fx CCD camera and qed software. Time-lapse fluorescence images were taken every 10–15 s, with exposure times ranging from 0.5 to 1.2 s. Images were saved and compiled with adobe photoshop software. Two microscopy fields were analysed for each cell population that showed oscillatory movements. The total number of cells was determined in these fields, and the percentage of cells showing oscillation was calculated. From these cells, a subset of cells was chosen to determine their average oscillation periods. We investigated whether the difference in oscillation periods between WM1032 cells expressing pSR15 and pSR15ER30D was significant. Statistical analysis was performed using the unpaired t-test (Young and Veldman, 1977). The difference was considered significant for P-values of 0.01. Standard errors were determined on the averages of the oscillation periods calculated for the cells analysed.

Acknowledgements

This project was funded by grants from the Canadian Institutes for Health Research (CIHR) No. 78020 to J. R. Dillon and grants from the National Institutes of Health (NIH) R01-GM61704-01 and the Texas Advanced Technology Program to W. Margolin. J. Szeto was supported by a NSERC (Natural Sciences and Engineering Research Council, Canada) scholarship. We thank Dr D. S. Weiss for providing us with the vector pDSW209. We are also grateful to Dr K. Begg and Dr J. E. Walker for supplying E. coli strains KJB24 and C41(DE3) respectively. We are grateful to D. Tessier (University of Ottawa) for his assistance in the production of anti-MinENg antisera.

Ancillary