Morphogenesis of Escherichia coli

Animal cells.The aim of this section is to introduce the structural embedding of the cytoplasmic membrane in a cellular framework. To make this point clear, I will describe the membrane of an animal cell. In particular, the development of our knowledge of the erythrocyte membrane is illuminating. Originally, the erythrocyte membrane, like all other membranes, was visualized in the electron microscope as a triple-layered structure distinctly separated from other cellular structures. This image arose because the fixative used (KMnO₄) is quite destructive to all other cellular components. With the onset of better fixatives (glutaraldehyde) and especially because of the impressive emergence of fluorescent light microscopy, a completely different picture of cellular organization arose: the cytoskeleton came to the fore.

Nowadays, the plasma membrane of animal cells is considered an intermediate between the cytoskeleton and the extracellular matrix. The cytoskeleton is an all-pervading network of polymerized proteins, and the extracellular matrix is likewise constituted of organized polymers. Transmembrane protein dimers (integrins) serve to integrate the intracellular and extracellular systems. The remarkable observation that the two compartments show a comparable orientation of actin-containing microfilaments (inside) and of fibronectin (outside), as revealed by fluorescence light microscopy, underlines their structural continuity via the plasma membrane (focal adhesion sites) (4).

Bacteria.In the foregoing eukaryotic example, the plasma membrane functions as a structural intermediate between the extracellular matrix/cell wall and parts of the cytoskeleton. Can one extrapolate from this example to bacteria? The small size of bacteria makes them difficult objects for (fluorescence) microscopy. To further develop this point, it is essential to refer to the basics of prokaryotic cytology. As will be outlined below, the absence in prokaryotic cells of a nuclear membrane and of the endoplasmic reticulum (ER)-Golgi machinery, makes a direct structural link between cell envelope and its chromosome possible. Actually, transcription and translation, which over the course of evolution of eukaryotes became spatially separated, are not separated in bacteria. Protein transport in gram-negative bacteria implies protein export from the cytoplasm to the periplasm. Therefore, the bacterial periplasm is topologically equivalent to the ER lumen whereas the bacterial cell wall is equivalent to the eukaryotic extracellular matrix. Thus, the bacterial cytoplasmic membrane combines properties of the ER membrane and the eukaryotic plasma membrane.

Cytoplasmic membrane and cotranslational protein synthesis.How does one arrive at a more integrated view of the bacterial cytoplasmic membrane? During cellular growth, proteins with various functions have to be inserted into the cellular membrane, either permanently or transiently. For at least some proteins, evidence is available for cotranslational protein transport through the cytoplasmic membrane (160, 168). While in vitro studies have concentrated on posttranslational translocation of envelope proteins, it is likely that, in vivo, both co- and posttranslational translocation exists for many proteins (for reviews, see references76, 114, and 158). Since bacterial transcription and translation are coupled (152), some functionally active parts of the nucleoid (encoding membrane or secreted proteins) are linked to the cytoplasmic membrane (Fig.6) (52, 94, 115, 182).

• Open in new tab
• Download powerpoint

Fig. 6.

Structural continuity between nucleoid and envelope through cotranscriptional biosynthesis of membrane proteins. Transmembrane proteins (white), involved in the synthesis of peptidoglycan, may restrict lateral diffusion of proteins being translocated (grey), and vice versa. im, inner or cytoplasmic membrane; om, outer membrane; pg, peptidoglycan. Modified from reference182 with permission of the publisher.

Cytoplasmic membrane and cell wall.How is the cytoplasmic membrane associated with cell wall components, notably the murein layer? As mentioned above, murein assembly is carried out by the HMW PBPs. Typically, the HMW PBPs possess a short amino-terminal cytoplasmic tail, one transmembrane segment, and a large periplasmic domain. This topology makes it likely that nascent murein is functionally associated with integral membrane proteins (Fig. 5). Also, the end products of the cytoplasmic steps leading to the mono- and disaccharide pentapeptide are covalently linked to undecaprenyl phosphate in the cytoplasmic membrane (lipids I and II, respectively [Fig. 3]). Lipid II also serves as an intermediate for lipopolysaccharide synthesis (for a review, see reference139); in addition, peptidoglycan synthesis and phospholipid synthesis appear tightly coupled (39, 132). This coupling most probably occurs at the level of the undecaprenyl phosphate-linked precursor (39).

The above presentation therefore shows that the cytoplasmic machinery of protein export results in cotranscriptional linkage of DNA to the cytoplasmic membrane (94), whereas at the same time the periplasmic machinery of murein synthesis is also linked to the cytoplasmic membrane (Fig. 6) (129).

I believe that there is sufficient evidence to make the point that the envelope and cytoplasm constitute a dynamically integrated structural entity. In other words, when bacterial shape is considered (admittedly most prominently reflected in the shape of its sacculus), the structural organization of the whole cell must be taken into account.

Elongation factor Tu.Many years ago, Minkoff and Damadian (105) advocated the presence of an all-pervading actin-like cytoplasmic network (Fig. 7). This network was thought to be responsible for cellular contraction and swelling as related to potassium uptake. Other researchers at that time also collected evidence for the occurrence of actin (and myosin) inE. coli (115). Very recently, proteins that bound to rat brain anti-actin antibody, to phalloidin, and to DNase I were detected in E. coli (51). However, their genes have not yet been identified. Returning to the protein of Minkoff and Damadian (105), it later appeared to be EF-Tu, which bears only a limited similarity to actin (140). What remains interesting, however, is that EF-Tu can polymerize in vitro into threadlike filaments and elongated sheets (6) and that it is present in excess over ribosomes in the cell. About one-quarter of the roughly 70,000 EF-Tu molecules are thought to be active in protein synthesis (74), while the remainder might play a structural role in the cell (6, 74, 120). Bundles of EF-Tu polymers are not likely to occur in vivo, because they would have been observed in electron microscopic thin sections. On the other hand, if EF-Tu filaments consisting of linear arrays of monomers were present in the cytoplasm, we would not expect to visualize them by current electron microscopic techniques, while in the case of fluorescent labeling, the resolution might be too limited. Therefore, one has to conclude that the occurrence of EF-Tu filaments in E. coli remains a possibility.

• Open in new tab
• Download powerpoint

Fig. 7.

Contractile proteins and the regulation of cell hydration in E. coli. Ion selectivity (Na⁺ or K⁺) depends on the cytoplasmic space available. Reprinted from reference 105 with permission from the American Society for Microbiology.

Interestingly, in this context, the eukaryotic counterpart of EF-Tu, EF1a, also occurs in large excess in the cell. More strikingly, EF1a can bind to actin filaments and microtubules in vitro as well as in vivo (for a review, see reference 19). Actin filaments can be cross-linked by EF1a (131), whereas microtubules become destabilized (148). These observations indicate that EF1a affects cytoskeletal organization, presumably reflecting “compartmentalization of translation in eukaryotic cells” (19). In fact, the close association of polyribosomes with heart myofilaments had already been observed many years ago (16). It is difficult to accept that such compartmentalization is superfluous in a bacterium.

FtsZ.A stronger potential cytoskeletal candidate is the essential cell division protein FtsZ, which also occurs in large quantities (about 20,000 copies per cell) (11). FtsZ accumulates at the cell center, presumably in a polymerized form, to direct the division process. However, even during division, a substantial amount of FtsZ remains in the cytoplasm (166,168). Given the polymerization potential of FtsZ (Fig.8) (13, 40, 109), one might surmise that FtsZ filaments could occur in the cytoplasm to function as a kind of supportive framework. Recently, the green fluorescent protein technique has been used to localize FtsZ in the cell. After overproduction, FtsZ-green fluorescent protein fusions became visible as spirals in the cytoplasm (95). Whether these structures are aggregates or polymers is not clear, nor whether FtsZ at wild-type levels can form spirals. However, it should be emphasized that if such elements do occur, they will be present most probably in the form of thin linear arrays of monomers (Fig. 8).

• Open in new tab
• Download powerpoint

Fig. 8.

In vitro assembly of FtsZ polymers and subsequent absorption to a polycationic lipid monolayer. Protofilaments (linear sequences of FtsZ monomers) occur singly or grouped into slender sheets. Rings have also formed. Reprinted from reference40 with permission of the publisher.

Even so, the picture which emerges from this discussion of FtsZ reminds one of the redistribution of microtubular and microfilament arrays during the eukaryotic cell cycle. The cellular distribution of FtsZ is not the same in dividing and nondividing cells; therefore, the principle of spatial redistribution of cytoskeletal elements seems to hold for E. coli too.

Cytoplasmic ultrastructure.How should one imagine the cytoplasmic structure beyond the ribosome? A helpful visualization is presented in Fig. 9. It depicts a 100-nm window of E. coli cytoplasm (50; see also Fig. 2 in reference 184). Ribosomes, mRNA, tRNAs, proteins, and double-stranded DNA have been drawn to scale and according to their estimated respective concentrations. The “empty” space between these macromolecules has been enlarged 10 times in the lower figure (Fig. 9b). Water molecules have been depicted as the predominant species, and some larger molecules or ions reside in between. No distinction has been made between bulk water and water of hydration (44). The postulated cytoplasm-supporting structure (see above) should then be conceived as filaments built from proteins of approximate sizes, as depicted in Fig. 9a. Somehow, structured multienzyme complexes have to be fitted between the ribosomes and the putative filamentous elements (129, 151). Clearly, the degree of structural organization in the cell, ranging from ribosomes to water, is not known at present, mainly because adequate in vivo techniques are lacking. However, such a cytoplasmic “structure” involves the basic organization of living matter and therefore represents a major challenge for science to resolve.

• Open in new tab
• Download powerpoint

Fig. 9.

(a) A 100-nm window of the E. coli cytoplasm. Ribosomes and other components have been drawn to scale. (b) A close-up of part of the window as indicated. Depicted are water molecules, some larger molecules, and part of a protein. Reprinted from reference50 with permission of the publisher. See also reference 184.

Cell division proteins.So far, we have concentrated on the periplasmic aspects of cell division, namely, the ingrowth of the peptidoglycan layer. The important demonstration by electron microscopic immunogold labeling that FtsZ assembles underneath the cytoplasmic membrane in the cell center (11) emphasizes the relevance of cytoplasmic events. FtsZ can bind GTP and possesses GTPase activity (11, 30, 140). In the presence of guanine nucleotides, FtsZ is able to polymerize in vitro, and various structures have been observed: sheets, protofilaments (a linear sequence of monomers), and rings (Fig. 8) (41). The putative GTP-binding site of FtsZ bears limited resemblance to eukaryotic tubulins (for reviews, see references 91 and100). These properties, in conjunction with the characteristic localization of FtsZ during division, have led to the idea that polymerized FtsZ might function as a contractile cytoskeletal apparatus (for a review, see reference 93). Interestingly, the overall sequence of prokaryotic FtsZ resembles eukaryotic γ-tubulin more than it resembles α- or β-tubulin (100). γ-Tubulin occurs in spindle pole microtubules, which are specifically involved in the positioning of microfilaments in the division plane of animal cells (reference42 and references therein).

Until now, I have mentioned two proteins that are specifically involved in cell division: PBP3 and FtsZ. Many others are known, and they are mostly designated Fts proteins (for reviews, see references93 and 164). Little is known about the roles of most of them, although most are localized to the site of constriction. On the other hand, their cellular location clearly shows how they are distributed over three cellular compartments: cytoplasm, cytoplasmic membrane, and periplasm. Again, one is reminded of the structure of focal adhesions. The main bulk of many Fts proteins lies in the periplasm (near to the murein), while they are anchored in the cytoplasmic membrane: FtsI (PBP3), FtsL (53), FtsN (25), and FtsQ (15, 37). Others span the cytoplasmic membrane several times: FtsK (7) and FtsW (71). In addition to FtsZ, FtsA (136) and ZipA (54) are cytoplasmic division proteins.

Divisome and subassemblies.It might be expected that division proteins assemble into one multimeric structure, which carries out the division process. This structure has been called septalsome (64, 65), divisome (119), or septator (5). I prefer the term divisome, because in my view division proceeds through constriction and not through septation in E. coli (80, 123). How the various proteins spatially and temporally interact in the divisome is largely unknown. Very recently, FtsN (3) and PBP3 (174) have been localized at midcell. Genetic (159) and biochemical (121) evidence has been presented for the interaction of FtsA and PBP3. Indications have also been found for the interaction of FtsA and FtsZ (2, 95, 165), of PBP3 and FtsW (102), of PBP3 and PBP1b (67), and, finally, of FtsZ and ZipA (Z-interacting protein [54]). Furthermore, peptidoglycan hydrolases appear to bind to the HMW PBPs (67). In contrast to FtsZ, most division proteins occur in a limited number of copies per cell. For FtsA or FtsQ, for instance, there are not enough molecules to cover the circumference of the cell. Also, FtsA and FtsZ function at a certain ratio (24, 34), implying that the FtsZ ring is not saturated with cell division proteins. Therefore, as a tentative model, one can envision that the divisome is composed of subassemblies which are connected by FtsZ polymers (Fig.10). Regarding the putative composition of a divisome subassembly (Fig. 11), it is important to realize that it has to carry out localized peptidoglycan synthesis. Therefore, it appears reasonable that MraY (producing lipid I) and MurG (producing lipid II) (Fig. 3), which provide disaccharide pentapeptides, should be considered essential components. Also included are hydrolytic enzymes (66) (Fig.5). DNA loops (Fig. 6) are absent. Clearly, the structural connection between the cytoplasm and periplasm during division is quite different from the one during cell elongation.

• Open in new tab
• Download powerpoint

Fig. 10.

Divisome composed of the FtsZ ring and divisome subassemblies.

• Open in new tab
• Download powerpoint

Fig. 11.

Model of a divisome subassembly. Note that the structure unites the cytoplasm and periplasm. Although the association between individual pairs of some proteins has been demonstrated (see the text), the presumed multiple interactions have not yet been characterized. The gene products FtsA, FtsI (PBP3), FtsK, FtsL, FtsN, FtsQ, FtsW, and FtsZ have been denoted A, I, K, L, N, Q, W, and Z, respectively. Lyt. indicates that lytic enzymes must be present (see also Fig. 5); cm, cytoplasmic membrane.

Three distinct events.Microscopy of the filamentous phenotypes at the nonpermissive temperature has shown thatftsZ mutants possess nonindented lateral walls. In contrast, other fts mutants can initiate division, after which division becomes aborted as revealed by partial, blunt constrictions (8, 155). This has been interpreted to mean that FtsZ is the first division protein to become active and that the participation of other Fts proteins follows. In support of this conclusion, it has been shown recently, that the FtsZ ring can form at the nonpermissive temperature in ftsA, ftsI (pbpB), andftsQ mutants (1). It has also been shown that FtsA can become localized to the cell center in the presence of nonfunctional PBP3 and FtsQ (2). Earlier data obtained by electron microscopic autoradiography (176) revealed that impairment of PBP3 did not inhibit the initiation of division, implying, in agreement with genetic evidence (8, 154), that this protein is involved in a later stage of division. For this reason, it has been argued that a PIPS step precedes the activation of PBP3 (119, 176). Although PIPS has not been characterized as yet, penicillin-insensitive enzymes acting on peptidoglycan have been described (119).

This brief survey suggests the following sequence of events. (i) FtsZ recognizes a specific target on the cytoplasmic membrane at the cell center and polymerizes there to form the FtsZ ring. Very recently, the possibility has been raised that the target is ZipA (54), although it now appears that ZipA ring formation is dependent on FtsZ (26a). (ii) Penicillin-insensitive peptidoglycan synthesis is activated. (iii) PBP3, FtsA, and FtsN associate with the ring (FtsK, FtsL, FtsQ, and FtsW might also become associated at this stage). (iv) Division progresses while peptidoglycan synthesis is carried out by PBP3 and other members (Fig. 5) of the multienzyme complex (the enzymatic functions of FtsK, FtsL, FtsN, FtsQ, and FtsW are not yet known). According to the scheme in Fig. 10, FtsZ would play a triple role during division: first, to establish the ring, second, to switch on PIPS, and third, to connect the divisome subassemblies. It is during the last stage that PBP3 and its companions are presumed to become active.

Do FtsA and FtsQ play a role in peptidoglycan synthesis?Thus far, I have emphasized the interaction between cell division gene products and the peptidoglycan-synthesizing system. In the divisome, FtsZ occupies a dominant position. However, archaebacteria such as the halobacteria do not contain peptidoglycan, yet they employ FtsZ (100, 170). These species have S-layer glycoproteins (containing glycan strands) superimposed on their cytoplasmic membrane and, not unexpectedly, lack β-lactamases (101). Like mycoplasmas, these organisms might also lack ftsA andftsQ genes (43, 170) (remember also thatftsQ, ftsA, and ftsZ form one transcriptional unit in E. coli [Fig. 4]). Probably, therefore, FtsA and FtsQ play a role in peptidoglycan synthesis (170). Some indications support this idea. Overproduction of FtsA leads to bulging cells, presumably because of local increase in peptidoglycan synthesis (169). It might also be speculated that membrane-bound FtsA (143) interacts with MurG and or MraY (Fig. 3 and 11). Regarding E. coli FtsQ, the SXXK and SXN sequences characteristic of some PBPs and β-lactamases (45) are present in this protein (14), also suggesting a role in peptidoglycan assembly. However, what argues against this suggestion is that the above sequences could not be found in Haemophilus influenzae, an organism which closely resembles E. coli.

Motor proteins?As has been shown by electron microscopic autoradiography, [³H]DAP-incorporation remained constant in three classes of dividing cells, grouped according to the degree of constriction. This has been interpreted to mean that “enzymes actively involved in murein synthesis become more and more closely packed” as division ensues (176). Consequently, during constriction, the divisome subassemblies are approaching each other, perhaps with the aid of motor proteins acting on FtsZ polymers (12). FtsA might be involved in movement of the divisome compartments, because it belongs to the actin family (143). Do motor proteins exist in E. coli? A 177-kDa motor-like protein (MukB [62]) has been implicated in nucleoid segregation. The overall organization of MukB resembles that of the so-called SMC proteins (133). These are thought to play a role in chromosome condensation, as required for mitosis. So far, there is no evidence for a role of motor proteins in the E. colidivision process.

Drosophila.Polarity in Drosophilacenters around the establishment of its anteroposterior and dorsoventral axes. In the Drosophila egg, asymmetric distribution of some constituents is present from the outset. These are the follicle and nurse cells surrounding the egg, which determine its internal three-dimensional chemical topography. This can be visualized by the asymmetric distribution of mRNAs coding for anterior and posterior proteins, respectively. For instance, bicoid mRNA (produced by nurse cells) becomes concentrated at the anterior end of the egg, establishing a morphogen gradient (for a review, see reference4). There does not seem to be an intracellular organizing agent that establishes the polarized distribution of the constituents of the embryo. Obviously, the early embryo contains topological memory due to its descent. Does this also apply to the yeast cell?

Saccharomyces cerevisiae.Polarity in S. cerevisiae research has focused on the selection of a new budding site (for a review, see reference 18). During this process, gene products have to find their targets (other gene products), which are located at the cellular membrane. How is a new bud site chosen?

Haploid cells choose a new bud site next to remnants of the previous division (axial bud site selection [reference 18and references therein]). One of the several proteins involved in site selection is Bud10p, a transmembrane protein. Interestingly it bears some resemblance to integrins of animal cells (55) (integrins connect the cytoskeleton with the extracellular matrix).

In diploids, a bud may arise opposite the previous division site. Such a specific location implies in itself an established polarity. In fact, division remnants might play a role here too. It is conceivable that part of the division remnant has been carried to the pole of the newly formed bud to provide recognition points for proteins involved in budding. Thus, both in haploids and in diploids, remnants of the previous division are used to establish a new bud site.

Bacteria.How does a prokaryotic cell compare to the above scheme? Well-known research themes are sporulation inBacillus spp. and the development of specialized cells inCaulobacter.

In Bacillus, spores tend to develop near the oldest pole (reference 63 and references therein), although the spore does not seem to have a preference for the old or the new chromosome (86). In fact, immunofluorescent staining of FtsZ has revealed that upon entry into sporulation, midcell localization of FtsZ switches to a bipolar pattern (90). Eventually, septation occurs at one pole only. How does asymmetry arise, and what are its consequences? At present, it is not known by what mechanism the spore-related septum is chosen. The consequences of its asymmetric placement are clearer; this leads to differences in local activities of transcription factors. As a result, a spore develops and the mother cell dissolves (for a review, see reference 41).

In Caulobacter, cellular polarity is marked by the placement of polar appendages. A predivisional cell carries a flagellum at one end and a stalk at the other. The two cellular compartments also differ in gene expression, which has been shown to be related to preexisting cellular asymmetry (for a review, see reference146). In these examples, polarity or asymmetry is quite apparent.

However, what about detecting polarity if easily discernible structures like spores and stalks do not occur? At the envelope level, at least three distinct cellular areas can be distinguished (23): the old pole, which is virtually inert with respect to [³H]DAP incorporation (85), and the lateral wall and the newly forming polar cap, which are involved in [³H]DAP incorporation (for reviews, see references23 and 122). As outlined above, the cell wall is connected to cytoplasmic structures via the cytoplasmic membrane (Fig. 6). At the newest pole, these connections are clearly less tight, because it is there that plasmolysis takes place most readily (112). Thus, each new cell is born with preestablished polarity. At the DNA level, it has been found that semiconservatively replicated DNA strands do not distribute randomly over the daughter cells (for a review, see reference23). The newest strand shows a tendency to be located in the cellular compartment near the newest pole (60). These observations further add to the notion of preexisting polarity in E. coli. Can further evidence of polarity (in the absence of cytological markers) be demonstrated?

Three recent E. coli examples (Fig.13) arose through immunogold labeling of chemoreceptor proteins (99) and the cell division protein FtsZ (11) and fluorescence microscopy of ZipA (54). The first is localized at the poles, and the division proteins are found midway between the poles. To adopt these locations, target sites have to be there first, suggesting the occurrence of specific domains in the cytoplasmic membrane. It has been shown that inE. coli, complexes of chemoreceptor proteins and two cytoplasmic proteins (CheA and CheW) are localized predominantly to a cell pole (Fig. 13a) (99). Polar location appeared to depend on the combined presence of chemoreceptor and CheW. Together, they seem to recognize a target at the cytoplasmic membrane. How did the target arise? It has been speculated (146) that it represents a remnant from a previous division. If so, polarization in yeast and polarization in E. coli have the same underlying principle: polarization arises from polarization.

• Open in new tab
• Download powerpoint

Fig. 13.

(a) Polar localization (arrow) of an E. colichemoreceptor (Tsr) as revealed by immunogold electron microscopy. Reprinted from reference 99 with permission of the publisher. (b) FtsZ localization at the cell center underneath the cytoplasmic membrane (arrows) at a nucleoid-free area, shown by immunogold electron microscopy. Reprinted from reference11 with permission of the publisher. (c) ZipA ring as revealed by fluorescence of a protein fusion between ZipA and green fluorescent protein. Bar, 2 μm. Reprinted from reference54 with permission of the publisher. (d) Immunogold labeling of FtsZ with monoclonal antibody MAb12 (165).

Nucleoid partitioning and cell division.It seems almost trivial to state that division serves to distribute duplicated chromosomes to new daughter cells. It is therefore not a big step to presume that nucleoid partitioning and cell division are also temporally and structurally related in E. coli. Alternatively, is bacterial division autonomous?

Some time ago, it was stated that the nucleoid exerts a “veto power” (61) over cell division, meaning that the nonpartitioned nucleoid represents a physical barrier for the division process (reference 183 and references therein). An experimental result underscoring this notion is shown in Fig.14. It represents an autoradiographic experiment where the incorporation of [³H]DAP into peptidoglycan was monitored with an electron microscope (111). Because the resolution of this technique is limited to about 100 nm (80), the experiment was executed with adnaX(Ts) mutant, in which DNA replication stops immediately at the nonpermissive temperature. Under these conditions, cells make filaments, and comparatively large DNA-less and DNA-containing areas can be distinguished. In other words, this makes it possible to compare [³H]DAP incorporation into the envelope bordering DNA with envelope that does not. A striking result is that in the vicinity of the nucleoid, there is less [³H]DAP incorporation than elsewhere (nucleoid occlusion [183]). Extrapolating this outcome to the much shorter wild-type cell, this suggests that after nucleoid separation, a circular envelope domain arises in the cell envelope, where [³H]DAP incorporation is less strongly inhibited, allowing peptidoglycan synthesis to increase locally as needed for the onset of division. Clearly, this can happen only when the FtsZ ring has also been formed (Fig. 13b). How should one envision the reduction of peptidoglycan synthesis in the vicinity of the nucleoid? In an attempt to give an explanation, I will return to a previous section of this review.

• Open in new tab
• Download powerpoint

Fig. 14.

Autoradiography of [³H]DAP incorporation (a and b) and of [³H]leucine incorporation (c). (a) Silver grain distribution over whole-mount ftsZ filaments. (b) Silver grain distribution over whole-mount dnaXfilaments. (c) Silver grain distribution over dnaX filaments pulse-labeled with [³H]leucine. The shaded areas in panels b and c represent the position of the nucleoid. Reprinted from reference 111 with permission of the publisher.

Cytoplasmic membrane domains.It seems reasonable to suppose that cotranscriptional synthesis of polypeptides (94) occurs most prominently in the vicinity of the nucleoid. Actually, in the experiment documented in Fig. 14, [³H]leucine incorporation was also measured (111). (It should be realized that in this experimental setup, no distinction could be made between synthesis of cytoplasm- and membrane-bound protein.) It was found that the incorporation of leucine was most intensive near the nucleoid. Therefore, I suggest that two extra envelope-associated domains should be added to the ones mentioned before. These putative domains (Fig.15) are distinguished by the relative enrichment of cotranscriptional DNA/membrane structures on the one hand and the relative enrichment of membrane components of the peptidoglycan-synthesizing machinery on the other. It would seem that proteins in membranes are confined to discrete areas.

• Open in new tab
• Download powerpoint

Fig. 15.

Proposed envelope domains in E. coli: old pole, new pole, nucleoid-associated domain, and domain between pole and nucleoid-associated domain. n, nucleoid.

By using “optical tweezers,” an impression could be gained about the trajectories that proteins can follow in an animal membrane (single-particle trapping). This has led to the concept of a membrane skeleton composed of domains (88). These domains have a size range of 0.1 to 1 μm, suggesting that a limited number would be present in E. coli (Fig. 1 in reference86). However, it is not clear whether the factors that limit domain size in animal cells are the same as those inE. coli.

Nevertheless, I have tried to develop the notion that cytoplasmic membrane domains are topologically determined by the spatial position of the functionally active nucleoid.

Cytoplasmic domains?In the above, the focus has been on domains in the cytoplasmic membrane. What about the cytoplasm in relation to cell division? It has been assumed that DNA replication and peptidoglycan synthesis withdraw pyrimidine nucleotides from the same pool (153). If so, one can imagine that termination of DNA replication might result in a local surplus of deoxyribonucleotides at the cell center (183). Part of this could then be used to produce extra UDP-GlcNAc, the starting material for peptidoglycan synthesis (Fig. 3). However, elongating and dividing cells have the same levels of peptidoglycan precursors (87). Therefore, the idea can work only if the cell can provide for a local (compartmentalized) increase in murein precursors at the expense of murein precursors elsewhere (119).

Division site selection without invoking the nucleoid.Is bacterial cell division regulated at the envelope level? There are two views that favor this idea. The first is based on the observation that when plasmolysis is induced in growing cells, spaces or bays arise, in particular at a cell pole. However, a fraction of such bays tend to occur in the cell center in nondividing cells or at one-quarter and three-quarter of the cell length in dividing cells, i.e., at future division sites. Although morphologically distinct, the bays have been linked genealogically to so-called periseptal annuli (20). These are circumferential zones of adhesion between the cytoplasmic membrane and peptidoglycan layer that run completely around the cell cylinder, and the region between them defines the periseptal domain where constriction starts (20, 21). Such annuli become visible in the cell center upon plasmolysis of dividing cells, and they have been most clearly observed in a Salmonella typhimurium lkyD mutant, where cell separation is impaired (96). Does the occurrence of a plasmolysis bay at a future division site reflect a property related to cell division? As mentioned above, the first visible sign of incipient division in nonplasmolyzed cells is the positioning of FtsZ at and of ZipA in the cytoplasmic membrane in the cell center. This implies that the property referred to above has no connection with FtsZ, because FtsZ and ZipA are not yet located to the future division site; i.e., only one ring is found in normally dividing cells (11). The genealogical relationship between plasmolysis bays and periseptal annuli has been questioned by Woldringh (180), and the controversy has not yet been settled (cf. MicroCorrespondence [1995] in Mol. Microbiol. 17:597–599 and reference 130).

The second model (the central stress model) for the triggering of cell division on the cell envelope level (irrespective of the nucleoid) has been presented by Koch and Höltje (84). They assume differences in stress between the cytoplasmic membrane and the peptidoglycan layer. These differences are supposed to be related to the respective elastic properties of the two layers. Division occurs when “the threshold is locally exceeded” (80). This would result in conformational changes of proteins, thus providing for a signal that induces local wall growth in the middle of the cell. Although plausible, the model is difficult to test by present techniques.

The central-stress model (84) is not necessarily in conflict with the nucleoid occlusion model (183) as expanded here. The cytoplasmic domains depicted in Fig. 15 might well give rise to differential stresses along the lateral wall. What all models have in common is the differentiation of the cell envelope into domains.

Protein-protein interaction.Much is known about protein-protein interactions for proteins which are diffusely inserted in the bacterial envelope and which are targeted to the cytoplasmic membrane, the periplasm, or the outer membrane. However, when considering division, we are dealing with a sharply localized target: the cell center. As outlined above, the composition of the cytoplasmic membrane might be different at the site of division, not only with respect to proteins but perhaps also with respect to phospholipid composition (80, 127, 128). A component X has been postulated as a connecting element between cytoplasm (FtsZ) and periplasm (peptidoglycan) (119). This protein, together with other components (lipids?), might create a specific membrane domain favorable for the binding of FtsZ and ZipA. Of course, the question of how ZipA arrives at the cell center is still open. Although PBP3 and FtsA can also associate with FtsZ, it is likely that this occurs at a later stage of the division process (see the previous section). It has also been suggested that cytoplasmic FtsZ has to change its conformation (exposing its hydrophobic segments) to bind to its cytoplasmic membrane target (165, 166). FtsA, which can bind ATP (143), could be instrumental in causing this change. However, the fact that the FtsZ-ring can form in the presence of a defective FtsA molecule (1) argues against such a role for FtsA. FtsA might reach FtsZ (while forming a divisome subassembly [Fig. 10]) by diffusion through the cytoplasm (however, see below). Transmembrane proteins like PBP3 and FtsQ might diffuse through the cytoplasmic membrane and become stuck when they meet ZipA and FtsZ. A similar mechanism has been suggested for the assembly of the chemoreceptor complex at the cellular pole (98). It should be added that PBP3 can occur in a dimeric form (187), like PBP1b (188). Dimerization of PBPs has been incorporated in the three-for-one model (Fig. 3).

Spatial distribution of mRNAs.In the foregoing, it has been assumed implicitly that divisome components are randomly positioned in the cytoplasm or in the cytoplasmic membrane before being assembled into a localized divisome. As an alternative, again taking Drosophila as the example (Fig. 12), mRNAs might be present at defined cellular locations. For example, in the case of cytoplasmic FtsZ, preferential cellular location of its gene or of its messengers would not occur. In contrast, at division, spatial rearrangement of the chromosome would direct the ftsZ gene to the cell center to provide for mRNAs at the spot. Is such a scenario likely? At a first glance, it is not. The cell size difference betweenDrosophila and E. coli is so huge that it is difficult to believe that diffusion will not distribute proteins inE. coli extremely rapidly. On the other hand, the importance of the structural organization should not be underestimated (Fig. 9). In fact, recent developments in microscopy such as the green-fluorescence technique (17) might allow these ideas to be tested. Such an approach should give us an idea about the degree of bacterial macromolecular organization as related to polarized processes.

Introduction.First, I will summarize some notions which have been developed in the foregoing sections. (i) A cylindrical soap bubble is an analog for the shape of E. coli(79, 80). (ii) With respect to cell elongation and cell division at the peptidoglycan level, specificity resides in the respective enzymes being active and not in the chemical makeup of the sacculus. This argues in favor of physical forces in shaping the cellular form. (iii) The sacculus is not an independent entity. It is structurally connected to the cytoplasmic membrane via the transmembrane HMW PBPs. Many membrane proteins as such become inserted through a mechanism of cotranslational protein synthesis. Translation and transcription in turn go hand in hand, thus connecting (mobile?) DNA loops to the cytoplasmic membrane. (iv) E. colidoes not have an eukaryote-like cytoskeleton. However, arguments have been presented to suggest that there is a cytoplasmic supportive structure composed of EF-Tu and FtsZ. (v) The structural connection between periplasm and cytoplasm at sites of cell elongation is quite different from the one at the site of cell division. This difference results from rearrangement of the macromolecular fabric, which is believed to be due, at least in part, to the partitioning status of the nucleoid. (vi) Various envelope domains have been proposed to exist: old poles, new poles, lateral wall next to the nucleoid, and lateral wall between nucleoid and pole. For each domain, the interaction between the cytoplasm and periplasm is supposedly different. (vii) As a consequence, each newborn cell is polarized from the outset. Polarization arises from polarization. (viii) Eubacteria as well as archaebacteria (without peptidoglycan) use the cell division protein FtsZ to carry out division. Although some prokaryotes can do without peptidoglycan, E. coli cannot. It should be borne in mind that the essential cell division gene ftsI encodes a penicillin-binding protein (PBP3).

Cell shape and E. coli. Depending on the growth conditions, E. coli varies in size; fast-growing cells are bigger than slowly growing ones (for a review, see reference123). At all growth rates, its shape can be roughly approximated by a cylinder with hemispherical ends. However, cells differ in their aspect ratio: the mean cell length (L) divided by twice times the radius (2R) (185). Obviously, cells with a small aspect ratio are short while a relatively large part of their surface is occupied by polar caps. In contrast, long cells have quite a different surface proportion of side wall and polar cap material. In an earlier section, it was mentioned that PBP2 is involved in cell elongation whereas PBP3 participates in polar cap formation. Specific antibiotics against these PBPs (mecillinam and cephalexin, respectively) result in inhibition of peptidoglycan synthesis as related to cell shape (Fig.16). That is, the higher the aspect ratio, the smaller the PBP3/PBP2-inhibition ratio (122,177). Clearly, the two PBPs play a role in maintaining cell shape.

• Open in new tab
• Download powerpoint

Fig. 16.

Relation between cell shape of MC4100 lysAand the PBP3/PBP2 inhibition ratio. The average cell dimensions have been measured, and the length/width ratio (L/2R) has been calculated. Briefly, the different shapes were obtained by varying the osmolality of the medium and the temperature and by usingpbpA (encoding PBP2) and ftsZ derivatives of MC4100 lysA. Further details can be found in the legend to Fig. 6 in reference 177. Each culture was probed for inhibition of [³H]DAP incorporation by PBP3-specific cephalexin (10 μg/ml) and PBP2-specific mecillinam (2 μg/ml). The ratios of these inhibition percentages were taken as the PBP3/PBP2 ratio. Reprinted from reference 177 with permission of the publisher.

Recently, it was suggested (67) that the sacculus as such would act as a template to be copied by a multienzyme complex (Fig. 5). However, this does not seem likely, because it is difficult to imagine how the aforementioned multienzyme complex could do more than produce glycan strands. It is still a big step from 30-nm glycan strands (Fig.2) to an intact sacculus.

This brings us back to a question posed at the beginning of this review: where does gene expression end, and where does physics come in? It was also mentioned that there seems to be no clear-cut relation between chemical composition and cellular shape. So how should one understand the shape of E. coli?

Surface stress theory.A model which can account for the rod shape of E. coli by analogy to the “morphological” behavior of soap bubbles has been developed by Koch in recent years (82, 83; for reviews, see references79 and 80). In E. coli, mass increase and hydrostatic pressure (22) take the place of the bubble blower. As outlined above, a cylindrical soap bubble can form between two fixed rings (Fig. 1), which are equivalent to the more or less rigid polar caps. The cylindrical shape becomes possible because the membrane of the soap bubble is fluid. According to Koch (80), four criteria must be fulfilled for stable cylindrical elongation. (i) The lateral wall has to enlarge according to a dispersive mode of insertion of building blocks. This has been demonstrated by electron microscopic autoradiography for the incorporation of [³H]DAP into the peptidoglycan layer and by immunogold labeling for the insertion of lipoprotein and LamB in the outer membrane (for a review, see reference 122). (ii) The polar caps should be relatively rigid. Reanalysis of the autoradiographic data of Woldringh et al. (181) concerning [³H]DAP incorporation has revealed that this is the case (83). The same result was obtained when cells that had incorporated d-Cys were chased into ad-Cys-free medium. Isolation of sacculi and labeling ofd-Cys with immunogold after biotinylation revealed diluted label in the lateral wall and nondiluted label in the cell poles (33). (iii) The criterion of sufficient turgor pressure also seems to be met (78). (iv) The growing sacculus “must be one in which the mathematics for fluid membranes apply and not the mathematics for stresses in a static solid” (80). In other words, are the lateral sides of a monolayered sacculus also fluid? This is very likely. In situ during growth, fluidity is achieved by lytic and synthetic enzymes, which are active all over the lateral surface. “Sacculus fluidity” would thus depend on the proper balance between synthetic and lytic activities in relation to its growth. This suggests that PBP2 plays a central role in maintaining this balance. Therefore, it would seem that the “morphogene” pbpA codes for a gene product (PBP2) that contributes to the maintenance of sacculus fluidity. It does not seem to determine cell shape but to contribute to the production of a suitable structure for physical factors to act upon.

Cell division in E. coli could be explained if the surface tension (T) were lowered in the center of the cell. It has been calculated that a twofold change in T would be sufficient to cause invagination of the cell envelope (Fig.17) (variable T model [80, 81]). In Fig. 17, it can also be seen that the extent of the decrease in T affects the shape of the constriction. As mentioned previously (118), impairment of PBP3 by mutation or by the PBP3-specific antibiotic furazlocillin leads to a constriction, similar in shape to that simulated by a 1.33- or 1.55-fold reduction in T. Interestingly, pointed polar caps have been observed in another temperature-sensitive PBP3 mutant (pbpB ^r1) when grown at the nonpermissive temperature (155). Pointed polar caps might be expected if the reduction in T is 1.75 (Fig. 17). These striking correlations indicate but do not yet prove that the activity of PBP3 is influenced by surface tension. In the case of long and smoothftsZ filaments, T obviously is not reduced at potential division sites. Still, it is difficult to understand what “soap bubble property” can account for their morphological stability. To test the variable T model, one would like to measure T at the cell center, which unfortunately is not possible at present.

• Open in new tab
• Download powerpoint

Fig. 17.

Extent of constriction as a function of the local fold decrease in surface tension, T. The square in the upper figure refers to the lower graph. If the fold decrease of Tis greater than 2, division can take place (lower left figure). In the case of, for instance, a 1.33- or 1.5-fold decrease in T, division cannot proceed and cells with so-called blunt constrictions arise (lower right figure). This situation applies after impairment of PBP3. Modified from reference 81 with permission of the publisher.

Hydrostatic pressure has also been invoked in relation to tip growth in fungi, although in some organisms no hydrostatic pressure seemed to be present (58). This is a puzzling situation, which could reflect the technical difficulty in measuring hydrostatic pressure or our limited understanding of a morphogenetic process (see also reference 175).

Cell shape and the interaction of cytoplasm and periplasm.Does the macromolecular fabric of the envelope and underlying cytoplasm play a role in maintaining the rod shape? It has been suggested in relation to the three-for-one model that existing template strands determine the length of the glycan chains coming in, thus also determining the cell diameter (66). However, as pointed out by Koch (80), the stress-bearing existing fabric will have a quite different spatial organization from the incoming unstressed polymers, making the idea of a glycan chain yardstick perhaps unlikely. What about the E. coli cytoplasm? Clearly, the ease with which spheroplasts are formed after the sacculus is degraded by lysozyme precludes a dominant role for a cytoskeleton-like structure in maintaining the rod shape. It might be speculated that the hypothetical cytoplasmic supportive structure helps to subdivide the cytoplasm into discrete functional domains. This would allow the safe transcription of (for instance) the pbpA gene and proper insertion and activation of its gene product (PBP2) in its membraneous environment, suitable for peptidoglycan synthesis. Perhaps this cytostructural context forms the material link between genotype and phenotype (at least in E. coli [57]). As has already been discussed extensively above, the relevance of cytoplasmic organization (polymerization of FtsZ and divisome assembly) is much more apparent in the case of cell division.