The MreB and Min cytoskeletal-like systems play independent roles in prokaryotic polar differentiation


  • Yu-Ling Shih,

    Corresponding author
    1. Department of Molecular, Microbial, and Structural Biology, University of Connecticut Health Center, Farmington, CT 06032, USA.
      E-mail; Tel. (+1) 860 679 2203; Fax (+1) 860 679 1239.
    Search for more papers by this author
  • Ikuro Kawagishi,

    1. Division of Biological Science, Graduate School of Science, Nagoya University, Chikusa-Ku, Nagoya 464-8602, Japan.
    Search for more papers by this author
  • Lawrence Rothfield

    1. Department of Molecular, Microbial, and Structural Biology, University of Connecticut Health Center, Farmington, CT 06032, USA.
    Search for more papers by this author

E-mail; Tel. (+1) 860 679 2203; Fax (+1) 860 679 1239.


Establishment of an axis of cell polarity and differentiation of the cell poles are fundamental aspects of cellular development in many organisms. We compared the effects of two bacterial cytoskeletal-like systems, the MreB and MinCDE systems, on these processes in Escherichia coli. We report that the Min proteins are capable of establishing an axis of oscillation that is the initial step in establishment of polarity in spherical cells, in a process that is independent of the MreB cytoskeleton. In contrast, the MreB system is required for establishment of the rod shape of the cell and for polar targeting of other polar constituents, such as the Shigella virulence factor IcsA and the aspartate chemoreceptor Tar, in a process that is independent of the Min system. Thus, the two bacterial cytoskeletal-like systems act independently on different aspects of cell polarization.


Establishment of an axis of cell polarity and the subsequent differentiation of the cell poles play important roles in cellular organization in a wide range of organisms. In rod-shaped and other asymmetric cells the axis of cell polarity is defined by the long axis of the cell. The axis of polarity can also be defined by using the plane of cell division as a frame of reference. In this case the poles are defined as the positions on either side of the division site that are located farthest from the mid-point of the division plane.

Polar differentiation is affected by the preferential localization of specific cell components at one or both poles. In bacterial species, these include proteins involved in locomotion, cell adhesion, division site placement, cell differentiation, chemotaxis and others (reviewed in Janakiraman and Goldberg, 2004a). The mechanisms responsible for the establishment of cell polarity and differentiation of the cell poles in prokaryotic cells are poorly understood.

The actin homologue MreB is required for cell shape maintenance in rod-shaped bacteria, such as in Escherichia coli and Bacillus subtilis. MreB is normally organized into coiled membrane-associated cytoskeletal structures that extend along the length of the cell (Jones et al., 2001; Kruse et al., 2003; Shih et al., 2003; Figge et al., 2004). A role for MreB in polar differentiation events has been suggested by the observations that depletion of MreB leads to defects in localization of several polarized proteins in Caulobacter crescentus (Gitai et al., 2004), and to defects in chromosome segregation in several organisms (Kruse et al., 2003; Soufo and Graumann, 2003; Gitai et al., 2004). Based on these observations it has been proposed that the MreB cytoskeleton plays an essential role in induction of bacterial cell polarity (Gitai et al., 2004).

A second group of proteins, MinCDE, are involved in a striking example of polarization that is required for the establishment of the division site at mid-cell in E. coli and other organisms. During this process membrane-associated MinCDE polar zones oscillate repeatedly between the two ends of the cell during the course of each cell cycle. This keeps the time-averaged MinCDE concentration high near the cell poles but low near the centre of the cell. Because MinC is an inhibitor of septation, this prevents assembly of the cell division machinery at sites nearer the cell extremities and effectively establishes the division site near mid-cell (reviewed in Rothfield et al., 1999; 2001). The Min proteins resemble MreB in being organized into pole-to-pole helical arrays that coil around the body of the cell (Shih et al., 2003). Assembly and disassembly of the MinCDE polar zones appears to reflect the redistribution of Min protein molecules along this cytoskeletal-like helical framework (Shih et al., 2003).

In this study, we compared the roles of the E. coli MreB and Min proteins in establishment of the cellular axis of polarity and in the differentiation of the cell poles. The studies indicated that the MreB and Min systems play independent roles in polarization events in E. coli. The MreB cytoskeleton was required for establishment of the rod shape and for polar targeting of a number of polarized components, as indicated by the loss of polar localization of the Shigella virulence factor IcsA and the aspartate chemoreceptor Tar in ΔmreB cells, in a process that was independent of the Min system. This is in contrast to the Min proteins that were capable of defining an axis of cell polarity in the absence of MreB as shown by establishment of a stable end-to-end oscillation cycle in spherical ΔmreB cells, which preceded formation of the FtsZ ring and division septum. Cellular asymmetry did not seem to be required in establishment of the polarized oscillation of the Min proteins. Thus, the two cytoskeletal-like systems appear to act independently in establishment of an axis of polarity and in the differentiation of the cell poles of E. coli.


The role of MreB in localization of MinD and other polarized cellular components was examined using mreB deletion strains YLS2 (mreB::cat) and YLS3 (ΔmreB). Both strains showed loss of the normal rod-shaped morphology of wild-type cells. Cells were generally spherical (Fig. 1A). The morphologies were similar to those described in studies of other E. coli mreB mutants (Wachi et al., 1987; Kruse et al., 2003). In septating cells, the two daughter cells were often of unequal size (arrows in Fig. 1A), indicating that the division plane was not always placed at the mid-point of the parental cell. The mreB mutants also showed slow growth rates and cold sensitivity, losing viability within a week at 4°C. The doubling times of YLS2 during log-phase growth at 30°C, 37°C and 42°C were 34 min, 29 min and 31 min, respectively, compared with 26 min, 17 min and 16 min for the MC1000 parent, and growth of YLS2 slowed further in late log phase.

Figure 1.

Division pattern and MreB immunofluorescence.
A. Nomarski images of strain YLS3 (ΔmreB). Arrows indicate cell division sites. Scale bar, 1 µm.
B and C. Immunofluorescence of MreB in strains MC1000 (min+) (B) and RC1 (ΔminCDE) (C).
D. FtsZ expression levels by immunoblot analysis of 2.5 µg of crude cell extract of MC1000 (mreB+) (lane 1), YLS2 (MC1000/mreB::cat) (lane 2), YLS3 (MC1000/ΔmreB) (lane 3), AB1157 (mreB+) (lane 5) and YLS26 (AB1157/mreB::cat) (lane 6). Lane 4, FtsZ (0.1 µg).
E. MreB immunoblot analysis of 2.5 µg of crude cell extracts of MC1000 (lane 1), YLS2 (lane 2), YLS3 (lane 3), AB1157 (lane 5) and YLS26 (lane 6). Lane 4, His-MreB (0.05 µg).

To minimize the likelihood of suppressor mutations, the mreB::cat cassette was also transferred by linear DNA transformation from YLS2 into two recipient strains, MC1000 and AB1157. The efficiencies of transfer were similar to that of a control minCDE::cat cassette (Experimental procedures). The resulting mreB::cat mutants in both backgrounds showed the same phenotypic characteristics described above for the parental strain. This suggests that the phenotypes associated with YLS2 are directly linked to the mreB deletion.

It has been recently reported that an mreB::cat mutation was lethal unless complemented by increasing the cellular FtsZ concentration or by the presence of another suppressor mutation (Gerdes et al., 2004). One difference between the present mutations and those previously studied was the presence in the present strains of the N-terminal 18 amino acids of MreB and the untranslated 127 bp 3′ end of the mreB compared with the 12 N-terminal amino acids and 36 bp 3′ end of the mreB mutant described previously (Kruse et al., 2003). It is not known whether this played a role in the survivability of the mutant. Measurement of cellular FtsZ by quantitative immunoblot analysis indicated that the concentration of FtsZ in the mreB::cat strains was approximately twofold higher than in the mreB+ parental strains (illustrated in Fig. 1D). It is conceivable that the strains picked up mutations unlinked to mreB during the original construction and also during transfer into both recipient strains that resulted in the FtsZ elevation. It is also possible that the loss of MreB feeds back positively on ftsZ expression.

Expressing mreB alone in trans from pLE6 (Plac-mreB), by growth in 0–250 µM IPTG, failed to restore the rod shape to mreB::cat cells. Restoration of rod shape to the E. coli mreB::cat mutant required coexpression of mreB, mreC and mreD, as previously reported for other mreB deletion mutants (Kruse et al., 2003). In contrast to the results with mreB::cat cells, expression of mreB alone by growth of strain YLS3/pLE6 (ΔmreB/Plac-mreB) in the presence of 40–60 µM IPTG restored rod shape to ΔmreB cells (approximately 81% rod-shaped cells). The cells often were shorter than wild-type cells and frequently contained slight bends or kinks or showed a branched morphology. These results suggest that replacement of the internal sequence of mreB by the cat cassette also has polar effects on transcription of the downstream genes that are relieved by removal of the cat cassette.

Comparison of MreB and MinD helical structures

We have previously shown that Yfp-MreB forms membrane-associated helical structures that coil around the cell from pole to pole in wild-type E. coli cells (Shih et al., 2003). However, the Yfp moiety causes significant loss of MreB function as shown by the failure of Yfp-MreB to fully complement strain mreB129 (Shih et al., 2003), a spontaneous mreB mutant (Wachi et al., 1987), and the inability of Yfp-MreB to restore rod shape to ΔmreB cells (data not shown). Because of the negative effects of the Yfp moiety on MreB function, we used immunofluorescence to re-examine the localization pattern of MreB. This confirmed that MreB helical structures were present along the length of both wild-type (Fig. 1B) and ΔminCDE cells (Fig. 1C). The helical densities of the MreB structures were 1.6 ± 0.3 and 1.5 ± 0.2 turns per µm in wild-type and ΔminCDE cells respectively. This confirmed that formation of the MreB helical structures is independent of the Min system.

In separate experiments, the helical density of the Yfp-MinD coiled structures in unfixed random images of strain HL1/pFX40 (ΔminDE/Plac-yfp::minD minE) was 2.2 ± 0.4 turns per µm. The difference in helical density between the MreB and MinD coiled arrays is statistically highly significant (P < 10−3). This result supports the view that MreB and MinD are present in different helical structures in E. coli, as previously suggested from studies of Yfp-MinD- and Yfp-MreB-labelled cells (Shih et al., 2003).

Localization and oscillation of Min proteins in the absence of MreB

Min proteins are present in two prominent polarized structures in rod-shaped E. coli cells, the membrane-associated MinCDE polar zone and the MinE ring that caps the polar zone. To determine whether MreB is required for the formation and oscillation of the MinCDE polar zone and MinE ring, we colocalized Yfp-MinD and MinE-Cfp in mreB::cat spherical cells (Fig. 2A–C). This revealed that membrane-associated Yfp-MinD zones were visible in 21% of the cells (124/592) (Fig. 2A1–C1). Bright MinE-Cfp foci were present at the edge of the MinD zones in 28% of cells that contained MinD zones (Fig. 2A2–C2). Three-dimensional reconstruction of optically sectioned cells confirmed that the MinE-Cfp foci were part of a ring structure (Fig. 2D6), equivalent to the MinE ring that caps the MinD polar zone of wild-type cells when it approaches mid-cell (Raskin and de Boer, 1997; Fu et al., 2001; Shih et al., 2002). MinE rings had not been identified in previous studies of spherical cells of ΔrodA mutants (Corbin et al., 2002). In addition to the MinE ring, a less intense fluorescent zone of MinE was also present that coincided with the MinD zone (Fig. 2A2–C2, yellow arrow), equivalent to the MinE component of the polar zone in wild-type cells (Raskin and de Boer, 1997; 1999; Fu et al., 2001). Similar patterns were present in ΔmreB and mreB::cat cells that were singly labelled with Yfp-MinD or MinE-Yfp (data not shown). Gfp-MinC in mreB::cat cells also formed membrane-associated zones that were similar to the MinD zones (Fig. 2E). This is consistent with studies in wild-type cells showing that MinC accompanies MinD within the polar zone (Hu and Lutkenhaus, 1999; Raskin and de Boer, 1999). We conclude that MreB is not required for the formation of membrane-associated MinCDE polar zones and MinE rings that are similar to the polar zones and MinE rings of wild-type cells.

Figure 2.

Distribution of Min proteins and FtsZ in wild-type and ΔmreB cells.
A–D. Yfp-MinD (green) and MinE-Cfp (red) in strain YLS2/pYLS68 (mreB::cat/Plac-yfp::minD minE::cfp) grown in the presence of 10 µM IPTG. (A1) The white arrow indicates a MinD loop. (A2–C2) Polar zones (yellow arrow) and MinE-Cfp rings (white arrow) in three different cells. Yfp-MinD (panel 1), MinE-Cfp (panel 2) and overlaid images (panel 3). (D) MinE-Cfp ring is shown in raw image of one slice from an optically sectioned cell (panel 4), in the same slice after deconvolution (panel 5) and in a three-dimensional reconstruction of the stack of deconvolved images (panel 6).
E. Gfp-MinC in three cells of strain YLS2/pYLS49-2 (mreB::cat/Plac-gfp::minC minD minE). Panels 7–9, fluorescence images showing polar zones of Gfp-MinC; panels 7’−9’, Nomarski images.
F–I. Colocalization of Yfp-MinD (green) and FtsZ-Cfp (red) from strains MC1000 (mreB+)/pFX40 (Plac-yfp::minD minE)/pWM1409 (Para-ftsZ::cfp) (F) and YLS3 (ΔmreB)/pFX40/pWM1409 (G–I). Cells were grown overnight at 30°C in the presence of 10 µM IPTG and 0.0001% arabinose, diluted into fresh medium without inducer for ∼ 3 h before viewing. Panels 10, 11 and 12 show Yfp-MinD, FtsZ-Cfp and overlaid images respectively. Arrows indicate the FtsZ rings.
J and K. FtsZ immunofluorescence of two cells of strain YLS2 (mreB::cat) showing perpendicular FtsZ rings. Arrows indicate the ends of the two perpendicular FtsZ rings.
L. Time-lapse images of strain YLS2 (mreB::cat) at 75 min intervals, showing perpendicular division planes arising during two successive division cycles. Arrows indicate planes of division.

When Yfp-MinD was expressed in mreB::cat cells together with MinE, the membrane-associated MinD zones underwent repetitive oscillations in a significant number of cells (Fig. 3A). This was also observed by Thanedar and Margolin (2004), although the details of the oscillation patterns were not defined. The average period of oscillation was 106 ± 36 s. The MinE and MinC polar zones and MinE rings showed similar oscillation patterns (data not shown). In 84% of cells (Table 1) the MinD zones oscillated exclusively between the two ends of the cell as illustrated in Fig. 3A–C. In the remaining cells, the MinD zones either oscillated predominantly from end to end with an occasional movement elsewhere (7% of cells), or moved irregularly between multiple sites that were not always located at opposite ends of the cell (9% of cells) (Fig. 3D). End-to-end oscillations accounted for 95% of the observed movements of MinD zones in the total population (Table 1).

Figure 3.

Oscillation patterns of Yfp-MinD polar zones in cells of strain YLS2/pFX40 (mreB::cat/Plac-yfp::minD minE). Intervals of the time-lapse photographs are indicated in seconds.
A–C. Pole-to-pole oscillation in symmetric (A), asymmetric (B) and septating (C) cells.
D. Oscillation along multiple axes.
E. Drifting pattern of movement around the circumference of the cell.

Table 1.  Types of oscillation.
 Type of oscillationaType of cells
  • a

    . Cells were categorized as described in Fig. 3. For each category (septating, non-septating, all), rows 1 and 2 show the number of cells showing the indicated oscillation type; numbers in parentheses are the ratio of cells with the indicated oscillation type to the total number of cells in the category. Rows 3 and 4 show the total number of movements of MinD zones from one location to another; numbers in parentheses are the ratio of movements of the indicated type to the total number of movements in all cells of the category. Drifting motion was not included in this analysis.

  • b

    . See Fig. 3C.

  • c

    . All oscillation cycles were end-to-end (Fig. 3A–C).

  • d

    . One or more movements were not end-to-end (Fig. 3D).

Number of cellsEnd-to-end  42 (0.91)c  30 (0.75)c  72 (0.84)c
Other   4 (0.09)d  10 (0.25)d  14 (0.16)d
Number of movementsEnd-to-end292 (0.97)258 (0.93)550 (0.95)
Other 10 (0.03)  18 (0.07)  28 (0.05)

Septal constrictions were visible in 53% of cells, permitting the unambiguous identification of cell poles in these cells (defined as the regions of the cell envelope in the future daughter cells that were farthest away from the nascent division plane). In 91% of these cells, the MinD zones moved exclusively between the poles (Fig. 3C, Table 1). These results show that MinD can establish a polar localization pattern and a polarized axis of oscillation in the absence of MreB.

Interestingly, the presence of significant cellular asymmetry or septal ingrowth was not required for the pole-to-pole oscillation pattern (Fig. 3A, Table 1). In ΔmreB cells that lacked visible septa, the incidence of end-to-end oscillations was similar in symmetric and asymmetric cells (82% versus 88%), with asymmetry defined as length/width ratio ≥ 1.1. This contrasts with the report that Min oscillations in spherical ΔrodA cells were most often random (Corbin et al., 2002). In addition to the oscillating MinD polar zones described above, MinD zones in a small subset of mreB::cat cells showed no visible discontinuous or oscillatory movement pattern but instead showed a drifting pattern of movement. In these cases, the MinD zones moved processively to adjacent sites around the cell circumference (Fig. 3E), usually continuing all or most of the way around the cell, with occasional reversals of direction. Drifting movement was very rarely observed in cells with nascent septa. It has not yet been established whether the drifting mobility behaviour precedes establishment of the polarized oscillation pattern that characterizes most cells.

Relation of MinD zones to division site placement

The results described above indicate that MinD polar zones localize normally in relation to nascent septa in the absence of the MreB protein. We further defined the relation of the MinD zones to the division site by comparing the localization patterns of FtsZ and MinD, using FtsZ as an early marker for development of the division machinery.

In immunofluorescence studies of wild-type cells, the FtsZ ring appears as a band across the cell cylinder or as paired dots on either side of the cylinder near mid-cell (Addinall et al., 1996). Consistent with previous studies of FtsZ-Gfp (Thanedar and Margolin, 2004), FtsZ labelled with the Gfp derivative Cfp formed typical ring structures at mid-cell in most wild-type cells (Fig. 2F11). Discrete FtsZ rings or paired foci were present in approximately 45% of ΔmreB cells (Fig. 2G11–I11), consistent with the observation that spherical ΔmreB cells underwent regular cycles of septation and division. In many cells the FtsZ rings and nascent septa were not located precisely at the mid-point of the cell (Fig. 2H and I), giving rise to the production of daughter cells of unequal sizes (Fig. 1A). The asymmetric placement of the division site might simply result from the larger volume and diameter of the spherical cells as compared with rod-shaped cells. Asymmetric division may also result from abnormal arrangement of the Min filamentous structures in the ΔmreB cells (see below).

Double-label experiments with FtsZ-Cfp and Yfp-MinD showed that the relation of the MinD polar zone to the FtsZ ring in ΔmreB cells (Fig. 2G–I) was similar to the relation of the MinD polar zones to nascent septa (Fig. 3C). The MinD polar zones in the spherical or septating ΔmreB cells were located in one side of the cell and did not extend across the plane defined by the Z-ring. There were many cells that contained MinD polar zones with no visible FtsZ ring. In contrast, there were essentially no cells that contained an FtsZ ring but lacked Yfp-MinD polar zones. This suggests that establishment of the oscillating polar zones preceded FtsZ ring assembly, as it does in wild-type cells.

In random micrographs the membrane-associated MinD zone was not always exactly parallel to the plane of the FtsZ ring (Fig. 2G). This implies that although the MinD zone undergoes polarized oscillation in the absence of MreB, the precise targeting of the MinD zones to the pole (defined by the ends of a line extending at right angles from the mid-point of the division plane to the edge of the cell) may be affected either by the absence of MreB or by the spherical cell shape. The increased volume of the ΔmreB cells could itself perturb the positioning of the Min polar zones if diffusion dynamics play a key role in establishment of the oscillatory pattern (Meinhardt and de Boer, 2001; Huang and Wingreen, 2004).

Time-lapse studies of cells through two division cycles showed that the second division plane was placed at right angles to the primary division plane (Fig. 2L). Consistent with this observation, two FtsZ rings appeared to cross each other at right angles in occasional cells (Fig. 2J and K). These rings presumably mark the planes of the division events during successive division cycles. Thus, the mechanism that leads to alternation of division planes at right angles to each other in spherical cells does not require the MreB cytoskeleton. A similar pattern of alternating perpendicular division planes occurs in spherical bacteria such as Neisseria (Westling-Haggstrom et al., 1977) and in spherical E. coli rodA mutants (Begg and Donachie, 1998) and in spherical cells resulting from mecillinam treatment (Pas et al., 2001). This pattern contrasts with the division pattern in wild-type rod-shaped cells in which successive division planes are always parallel to each other.

MinD structures inΔmreB cells

The cellular organization of Yfp-MinD in ΔmreB cells was examined in further detail to determine whether they contained elements similar to the MinD helical structures that coil around rod-shaped wild-type cells. The spherical ΔmreB cells contained curved or coiled filamentous structures, but did not show continuous helical structures that could be followed around the entire cell. The structures were visible in unprocessed images (Figs 2A1 and 4A–D and H), confirming that they were not processing artefacts. They were also seen in fixed cells that were subjected to optical sectioning and deconvolution to remove out-of-focus information (Fig. 4E and F) and in three-dimensional reconstructions prepared from the stack of deconvolved images (Fig. 4G).

Figure 4.

Yfp-MinD structures in cells of strain YLS2/pFX40 (mreB::cat/Plac-yfp::minD minE).
A–D. Raw images of loops (A, B and D) or curved filaments (C) of Yfp-MinD in unfixed cells. Loops are indicated by arrows in (D).
E and F. Deconvolved images of Yfp-MinD fluorescence from a single focal plane in two individual cells showing extended filamentous structures.
G. Three-dimensional reconstructed image of Yfp-MinD showing extended filamentous and coiled (arrows) structures.
H. Time-lapse images of unfixed cells during an oscillation cycle, showing polar zone fluorescence moving alternately between coils at the two ends of the cell. Units in seconds.

The most common Yfp-MinD structures were single circles or loops (Figs 2A1 and 4A and B), or occasionally multiple loops (Fig. 4D and G), that were visible within a polar zone. These circle or loop structures appeared to be located on the cell surface and were not observed to wind around the circumference. It is not established yet whether these were part of a more extended coiled structure as is the case in wild-type rod-shaped cells. In some cells time-lapse studies showed that the Yfp-MinD in the polar zone oscillated between loops located at opposite ends of the spherical cell (Fig. 4H). This is of interest as it has previously been shown that MinD polar zones in rod-shaped wild-type cells oscillate from pole to pole by redistribution between polar coils of a cytoskeletal-like MinD helical array (Shih et al., 2003).

Extended MinD structures that formed a poorly defined network of curved or looped elements were seen in a number of cells (Fig. 4E–G). In rare cells, several evenly spaced curved filaments ran across the width of the cell (Fig. 4C), consistent with periodic membrane-associated structures. However, the detailed long-range organization of the extended structures described above has not yet been successfully elucidated. It is not known whether there is a relationship between the structures in the ΔmreB cells and the helical MinD structure that extends along the long axis of wild-type cells. The alternation of division planes that occurs in successive division cycles in the spherical ΔmreB cells (Fig. 2L) may complicate interpretation of the long-range organization of the MinD structures. The orientation and long-range organization of the Min filamentous structures in the ΔmreB cells could reflect the difference in geometrical constraints of rod-shaped and spherical-shaped cells or could be affected by the absence of the longitudinal MreB helical arrays.

Localization of other polarized proteins

To further investigate the establishment of polarity in ΔmreB cells, we studied the localization pattern of the IcsA protein of Shigella and the aspartate chemoreceptor protein Tar. ‘Poles’ in ΔmreB cells were identified visually in cells containing nascent septa and were defined as the regions on either side of the division site that were farthest from the mid-point of the division plane.

The IcsA protein of Shigella is located at the cell pole, where it mediates actin tail formation in the host cell cytoplasm when the organism is present in eukaryotic host cells (reviewed in Goldberg, 2001). We studied the IcsA507−620 fragment which contains a polar targeting domain and localizes at cell poles and division sites when expressed in E. coli (Charles et al., 2001). The previous observation that the localization pattern of IcsA507−620-Gfp at poles and division sites was suppressed by coexpression of full-length IcsA confirms the specificity of IcsA507−620-Gfp localization (Janakiraman and Goldberg, 2004b).

As previously reported, the IcsA507−620-Gfp foci were located at one cell pole (Fig. 5A1) or both poles (Fig. 5A2 and 3) or at division sites (Fig. 5A3) in wild-type MC1000 cells. It has been shown that the localization of IcsA507−620-Gfp is independent of the Min system (Janakiraman and Goldberg, 2004b). In contrast, there was no consistent localization pattern of IcsA507−620-Gfp in ΔmreB cells (Fig. 5B). The number of IcsA507−620-Gfp foci per ΔmreB cell varied from zero to greater than five (Fig. 5B4 and E). The foci in the ΔmreB cells appeared to be randomly distributed, with no preference for division sites or cell poles. IcsA507−620-Gfp foci were occasionally observed at one end of the cell with a frequency similar to their presence at other points around the periphery of the cell. In most septating cells, different numbers of foci were present in the two daughter cells, with 42% of septating cells showing a twofold or greater difference in the number of foci between the two daughter cells. This contrasts with the distribution in wild-type cells (Fig. 5A). In a parallel experiment, IcsAΔ507−620-Gfp, in which the polar targeting domain is absent, was diffusely distributed in wild-type (Fig. 5C) and ΔmreB cells (Fig. 5D) under the same induction conditions. We conclude that MreB is required for the polar localization of IcsA507−620.

Figure 5.

localization of IcsA507−620-Gfp and Tar-Gfp in wild-type, Δmin and ΔmreB cells.
A and B. Localization of IcsA507−620-Gfp (red) and membrane stain FM4-64 (green) in individual cells of strains MC1000/pMPR402 (Para-icsA507−620::gfp) (A) and YLS3/pMPR402 (B). Arrows indicate IcsA507−620-Gfp foci. Scale bars, 1 µm.
C and D. Localization of IcsAΔ507−620-Gfp in cells of strains MC1000/pMAC362 (Para-icsAΔ507−620::gfp) (C1) and YLS3/pMAC362 (D1 and 3). Micrographs C2, D2, D4 are the Nomarski images of cells in C1, D1, D3 respectively. Scale bars, 1 µm.
E. Distribution of IcsA507−620-Gfp foci in ΔmreB cells. (a) Non-septating cells: top line shows number of foci. (b) Septating cells: top line shows number of foci in each of the two daughter cells (e.g. 2/3 indicates that one daughter has two foci and the other has three foci). Asterisk (*): cells showed specified number of foci as percentage of total non-septating or septating cells.
F–H. Localization of Tar-Gfp in cells of strain MC1000/pDS1030 (Para-tar::gfp) (F), YLS24 (Δmin)/pDS1030 (G) and YLS3 (ΔmreB)/pDS1030 (H). Arrows indicate invaginations at the division sites. Scale bars, 1 µm.

We also examined the localization pattern of the aspartate chemoreceptor protein Tar, a member of the chemotaxis family that clusters at the cell pole (Homma et al., 2004). Tar-Gfp was expressed in wild-type, Δmin and ΔmreB cells. In wild-type cells, Tar-Gfp showed the previously described (Homma et al., 2004) bipolar localization pattern (Fig. 5F). The polar foci of Tar-Gfp were also seen in Δmin cells (Fig. 5G). In some cells (Fig. 5G), Tar-Gfp was organized into an apparent helical pattern along the length of the cell. This has previously been observed and studied and will be published separately (D. Shiomi and I. Kawagishi, pers. comm.). In contrast, in ΔmreB cells the Tar-Gfp foci appeared to be randomly positioned around the cell periphery, in both septating and non-septating cells (Fig. 5H). We conclude that MreB is also required for the polar localization of the Tar protein. The MreB requirement for polar localization of IcsA and Tar contrasts with the MreB-independent polarization of the MinCDE proteins.


We conclude that the MreB and Min cytoskeletal-like systems can act independently in establishment of an axis of polarity in E. coli. MreB is required both for rod-shaped determination and for targeting of components to the cell poles. The Min system, on the other hand, is capable of establishing an axis of polarized localization and oscillation in the absence of MreB.

Min system

It is clear that the MreB cytoskeleton play no role in the polarization of the Min system as the MinCDE polar zones and MinE rings assembled normally and underwent their usual repetitive pole-to-pole oscillation pattern in spherical mreB::cat cells. On the basis of these results we suggest that the axis of polarity in spherical cells is determined by the Min system, with the oscillating MinCDE system then determining the placement and orientation of the FtsZ ring and division apparatus as it does in rod-shaped wild-type cells. In this view, there are no cell ‘poles’ until the axis of oscillation is selected. Assembly of the Min oscillatory system precedes formation of the FtsZ ring as indicated by the observation that Min polar zones were present in many ΔmreB cells that did not contain FtsZ rings, and by analogy to division site selection in wild-type cells where the Min oscillatory system assembles normally in the absence of FtsZ (Raskin and de Boer, 1999). The observation that the Min polar zones in non-septating cells can undergo a type of ‘drifting’ movement around the cell circumference (Fig. 3E) and that a small fraction of cells show an indeterminate oscillation pattern is consistent with a model in which the polar zone can sample many different axes or sites until a preferred axis for pole-to-pole movement is found as discussed below.

Selection of the division site in the spherical cells is not random, as shown by the fact that the FtsZ ring and division septum were assembled in alternating perpendicular planes in successive division cycles. The orientation of the Min oscillation axis therefore must also alternate orientations during successive division cycles. We have confirmed this in time-lapse studies (data not shown). This implies that selection of the correct Min oscillation axis relies on topological information from the previous division cycle. We speculate that the topological memory may be a function of the residual division site from the preceding division cycle, with the ‘new pole’ becoming an excluded site for Min oscillation. This would restrict the new axis of oscillation to a plane parallel to the preceding division plane (i.e. perpendicular to the preceding axis of oscillation).

The presence of oscillating MinD polar zones has also been described in spherical cells resulting from loss of rodA (Corbin et al., 2002; Ramirez-Arcos et al., 2002). There is no evidence for direct interaction between RodA and MreB, although it has been suggested that they might be part of the same multiprotein complex (Kruse et al., 2005). It has been proposed that the Min oscillation axis in spherical ΔrodA cells is determined by cellular asymmetry that is caused by assembly of an FtsZ arc and asymmetric septal invagination (Corbin et al., 2002). In the present study, we found that the end-to-end oscillation is dominant in spherical ΔmreB cells without apparent asymmetry and without evidence of early septal ingrowth. This implies that detectable asymmetric geometry is not required for establishing the axis of Min oscillation in ΔmreB cells. Based on our observations, it is more likely that both placement of the division site and development of cell asymmetry are the consequences of the polarized Min oscillation (Fig. 6A). It has also been shown that self-perpetuating oscillations can be generated by mathematical modeling in symmetrical spherical cells (Huang and Wingreen, 2004).

Figure 6.

Models of the polarization events in the ΔmreB cells of E. coli.
A. Polarized oscillation of the Min proteins in spherical ΔmreB cells. The end-to-end oscillation of Min proteins leads to the establishment of the cell ‘poles’, defined as the farthest positions on the opposite sides of the division septum.
B. Localization of polarized proteins (open circles) in wild-type and ΔmreB cells. Loss of polar localization in the ΔmreB cells (right side) could be due to absence of MreB or to the difference in cellular geometry of the rod-shaped and spherical cells.

Any attempt to explain the intriguing perpendicular alternation of division planes in the spherical ΔmreB cells must confront the fact that this pattern changes to a pattern of parallel orientation of successive division planes when MreB is expressed and cell shape changes from sphere to rod. Thus, an event associated with the sphere-to-rod transition can supersede the mechanism responsible for the pattern of perpendicularly alternating division planes in the spherical cells.

MreB system

A number of specific polar differentiation events in E. coli were MreB-dependent and Min-independent. These included the polar localization of the aspartate chemoreceptor Tar and the Shigella virulence protein fragment IcsA507−620. [Nilsen et al. (2005) have also observed loss of polarity of IcsA507−620 in ΔmreB cells.] These results are consistent with previous reports that implicate MreB in localization of several polarized proteins in C. crescentus (Gitai et al., 2004). The basis of the local accumulation of the proteins in non-polar locations in ΔmreB cells is not known. It could reflect redistribution of polar nucleation sites, if they exist. Further work will be needed to answer this question. It is likely that the MreB cytoskeleton will also be required for the localization of other proteins that are preferentially found at one or both cell poles (Fig. 6B).

We also confirmed previous observations (Kruse et al., 2003; Soufo and Graumann, 2003; Figge et al., 2004; Gitai et al., 2004) that chromosome segregation to the two ends of the cell is defective in the absence of MreB (data not shown). This led to the presence of approximately 18% anucleate cells in ΔmreB cultures and occasional cells in which the division septum guillotined a segment of the nucleoid (data not shown). At the molecular level, it has been shown that deletion or overexpression of MreB leads to apparent mislocalization of oriC and terC in E. coli (Kruse et al., 2003), and MreB was shown to directly or indirectly interact with the origin-proximal region of the chromosome, presumably to mediate chromosome segregation in C. crescentus (Gitai et al. 2005).

It is not clear why MreB is required for these polar targeting events. The spiral MreB structure that extends from pole to pole consists of one or more helical filamentous structures (Kruse et al., 2003; Shih et al., 2003; Gitai et al., 2004). The structural similarity between MreB and actin (van den Ent et al., 2001) suggests that the MreB helical filaments may act as tracks for delivery of cargo proteins to the poles, presumably in cooperation with motor proteins, or that the helices are composed of shorter filaments that themselves are moved to the poles together with cargo proteins. The ends of the pole-to-pole MreB filaments could also act as nucleation sites for anchoring protein arrays at the poles. Alternatively, MreB may act indirectly in polar targeting. This could, for example, involve changes in polar murein, which differs from murein elsewhere along the cylindrical cell (de Pedro et al., 1997; 2001). There is evidence that this is associated with changes in distribution of surface proteins (de Pedro et al., 2003; Nilsen et al., 2004). As the MreB cytoskeleton appears to regulate the pattern of murein assembly during cell growth (Daniel and Errington, 2003; Figge et al., 2004), it may also influence the organization of murein at the cell poles. A secondary effect on inner membrane organization might then be responsible for the protein targeting events. It will be of interest to determine whether the characteristic lack of turnover of polar murein that characterizes the ends of rod-shaped cells also occurs at the ‘poles’ of spherical ΔmreB cells. The possibility also exists that other components, such as MreC, MreD, PBP2 and RodA, may also be involved in the polar targeting function, or that geometric effects of the rod shape might play a role in these events. Further work will be needed to clarify these important points.

Experimental procedures

Strains and plasmids

The following E. coli strains and plasmids were used (only relevant genotypes are shown): strains MC1000 (ΔlacΔara) (de Boer et al., 1989), AB1157 (Δara) (Bachmann, 1972), RC1 (ΔlacΔaraΔminCDE) (Rowland et al., 2000), HL1 (ΔminDE) (Hale et al., 2001), and plasmids pFX40 (Plac-yfp::minD minE) (Shih et al., 2002), pLE6 (Plac-mreB), pLE7 (Plac-yfp::mreB), pLE9 (Plac-mreB mreC mreD) (Shih et al., 2003) and pYLS68 (Plac-yfp::minD minE::cfp) (Shih et al., 2002). Strain YLS2 (mreB::cat) was generated using the Red recombination system (Datsenko and Wanner, 2000). The mreB coding sequence between 55 bp and 1013 bp was replaced with three stop codons followed by a cat cassette transcribed in the opposite direction to mreB. YLS2 was cured for the cat cassette (Datsenko and Wanner, 2000) to generate YLS3 (ΔmreB), which retains the stop codons after mreB nucleotide 54. Strains YLS1 (min::cat) and YLS24 (Δmin) were created using the same principles. Plasmid pDS1030 (pBAD24-Tar-Gfp) was generously provided by Daisuke Shiomi before publication. Plasmids pMPR402 (Para-icsA507−620::gfp) and pMAC362 (Para-icsAΔ507−620::gfp) (Charles et al., 2001) and pWM1410 (Para-ftsZ::cfp) were generous gifts from Marcia Goldberg and William Margolin respectively. pYLS100 (PT7lac-his::mreB) was constructed by polymerase chain reaction (PCR)-mediated introduction of mreB into pET15b. pYLS120 (Plac-yfp::mreB mreC mreD) was constructed by PCR-mediated introduction of mreB mreC mreD into the large XbaI/HindIII fragment of pYLS68.

To transfer the mreB:cat mutation to other host cells, the chromosomal mreB::cat cassette was PCR amplified from YLS2, yielding a fragment identical to the DNA fragment used to introduce mre::cat during construction of strain YLS2. The linear DNA (100–200 ng) was transferred by electroporation into MC1000 and AB1157 (108−9 competent cells) (Datsenko and Wanner, 2000), selecting for chloramphenicol resistance. This avoids carrying over any extragenic suppressor mutation(s) into the recipient cells. Transformation efficiencies were approximately 500 and 100 transformants per µg of mreB::cat DNA into MC1000 and AB1157 respectively. In parallel experiments, a minCDE::cat mutation was transferred using the same technique into MC1000 and AB1157 with efficiencies of approximately 100 and 40 transformants per µg of DNA into MC1000 and AB1157 respectively. It should be noted that transfer efficiency may be affected by the nature of the flanking sequences. The strain AB1157/mreB::cat was named YLS26.

Other methods

Growth and induction conditions and visualization of Yfp- or Cfp-tagged proteins (Shih et al., 2002) and immunofluorescence methods (Justice et al., 2000) were described previously except that 100 mM phosphate buffer pH 7.4 was used during fixation for FtsZ and pH 6.7 for MreB and Alexa fluor 488- and Alexa fluor 594-conjugated goat anti-rabbit secondary antibodies were used (Molecular Probes). Anti-MreB antiserum was raised against His-MreB and purified by absorption to purified His-MreB bound to PVDF membrane, followed by elution with 0.2 M glycine (pH 2.14) and renaturation with an equal volume 1.5 M Tris-base (pH 8.8).

Fluorescence image acquisition and image processing was performed essentially as previously described (Shih et al., 2003) except that an Olympus PlanApo 60×, 1.4 oil objective was used. Analysis (Table 1) was limited to cells in which at least four clear movements of a MinD zone were observed within a single focal plane. The oscillation rate of MinD was determined as previously described (Shih et al., 2002). In spherical cells, end-to-end oscillation was defined as repetitive movements between opposite sides of the cell. Helical densities, defined as number of coils (identified as transverse bands) per µm, represent the average of measurements from at least 75 cells.

To visualize cell membrane, cells were stained with 1 µg ml−1 FM4-64 [N-(3-triethylammoniumpropyl)-4-(6-(4-(diethylamino)phenyl)hexatrienyl) pyridinium dibromide] for 10 min followed by extensive washing with saline. Chroma filter cube UMWIG was used to visualize FM4-64 fluorescence.

For FtsZ quantitative Western blot analysis, the band intensity of the test sample was compared with band intensities of a range of concentrations of FtsZ (Shih et al., 2002).


We thank Dr Mary J. Osborn and Dr Marcia Goldberg for helpful discussions. We also thank Dr Daisuke Shiomi for providing plasmid pDS1030 before its publication. This work was supported by National Institute of Health Grant GM060632-05.