Chapter 104

Segregation of Cell Structures

STEPHEN COOPER

ANALYSIS OF SEGREGATION IN BACTERIA

A newborn cell on its trajectory toward division must double all of its components. How much of the new material (or old material) present in a dividing cell goes to one daughter cell and how much goes to the other? Between the two extremes of completely dispersive segregation (old and new material go equally to both daughter cells) and a completely conservative segregation (all new material goes to one cell and all old material goes to the other), other possibilities arise. If the cell were simply an amorphous blob of cytoplasm that divided randomly at division, a simple equipartition of all cell components at division would be expected. But the cell is not that simple, and each of the three major cell components—the DNA, the cell surface, and the cell cytoplasm—has a different segregation pattern.

If each molecule of the cell had a different segregation pattern, or if there were a large number of patterns for many different molecules, the description and understanding of the segregation rules would pose an insuperable problem. Fortunately, one can simplify the problem by apportioning the material of the cell into three categories. These three categories—the cytoplasm, the genome, and the cell surface—have been used before to describe and simplify the growth pattern of the cell (22). A particular molecule is considered a member of a category or group by its location and not necessarily by its biochemical properties.

The history of segregation analysis began when Van Tubergen and Setlow (82) studied the behavior of a number of macromolecules as cells grew and divided. Cells were labeled with radioactive precursors of cell components, and the labeled cells were "chased" or grown for a number of generations in unlabeled medium. At selected times, the cells were fixed and autoradiographed, and the number of grains per cell was determined. If there were an equipartition of material, daughter cells would be expected to receive equal shares of the original labeled material, and a Poisson distribution of grains would be expected. A non-Poissonian distribution would suggest a segregation process other than equipartition. By determining the number of generations prior to achieving a non-Poissonian distribution, the number of independent segregating subunits in the original cell could be estimated. Van Tubergen and Setlow (82) concluded from the observation of a non-Poissonian distribution of grains in thymidine-labeled cells that only a limited number of chromosomes or DNA strands were present in the bacterial cell, and these units were conserved after synthesis. Later studies confirmed and extended this work (15, 59). These DNA units segregated as stable units and did not subdivide with further growth; this result is consistent with the semiconservative replication of the bacterial DNA. In contrast, when protein or cell wall was specifically labeled, a dispersive or nonconservative segregation pattern was observed.

Forro and Wertheimer (34, 35) expanded on these results by autoradiographing microcolonies formed by pulse-labeled or long-term-labeled cells. After long-term labeling, microcolonies had two heavily labeled cells and two lightly labeled cells. These results were consistent with the cells having one or two replicating chromosomes, equivalent to cells growing with a 40-min interdivision time (22). Because the cells were not arranged in any order when the microcolonies were formed, the spatial pattern of DNA segregation could not be determined. This point has been the object of considerable effort and will be analyzed below.

These basic observations will now be elaborated to present further evidence regarding the details of segregation of different cell components. Each analysis will be presented in the following form. First a simple description of our current understanding of the basic segregation pattern will be presented, omitting various details for the sake of clarity. This will be followed by a discussion of the experimental basis for the proposed segregation pattern. Finally, an analysis of alternative views will be presented.

SEGREGATION OF CYTOPLASM

Cytoplasm is generally believed to be dispersed randomly and nonconservatively to each of the daughter cells (Fig. 1). The evidence for this pattern of segregation comes from two sources. First, Van Tubergen and Setlow (82) demonstrated that amino acid- and uridine-labeled material segregated randomly and dispersively. However, even diaminopimelic acid (DAP)-labeled material gave a dispersive segregation pattern, even though cell wall does not have a completely dispersive segregation pattern (see below). More persuasive evidence for dispersive, random segregation of cytoplasm is that cells labeled and placed on a membrane elution apparatus ("baby machine") divide their material equally, such that the radioactivity per cell is halved at each generation (19). No departure from this pattern is seen over a period of at least 10 generations. This is strong evidence that protein segregates in a random manner (22; unpublished results).

Fig. 1Segregation of peptidoglycan. For a cell with the surface totally lableled (black lines), there is a random segregation of the side wall material (indicated by the decreasing shading at each generation). The old pole material is conserved (indicated by the white material). Thus, three distinct areas at the cell are postulated: the older poles where the density of label does not change, new poles which are unlabled, and side walls where the density of label halves at each generation.

Source:

Random cytoplasmic segregation is also supported by the absence of other proposals as to how any nonrandom patterns could arise. There is no morphological evidence for compartments or structures that could produce a nonrandom or conservative segregation pattern.

SEGREGATION OF THE BACTERIAL CELL SURFACE

Arrangement of Peptidoglycan on the Escherichia coli Surface

The peptidoglycan of the bacterial cell is a meshlike structure made of glycan chains cross-linked by chains of amino acids. As reviewed elsewhere (22 [specifically chapter 6], 23; this volume, chapter 6), the glycan chains appear to go circumferentially around the outside of the rod-shaped bacterial cell, much like hoops encircling a barrel. Electron microscopic analysis of partially digested cell walls (85) and the results of controlled sonication studies (83) suggested that the strands of the cylindrical side wall are indeed arranged in a hooplike pattern. When cell wall fragments were observed, the strands went preferentially in one direction, perpendicular to the long axis of the cell. Growth in cell length, according to this structure, occurs by the insertion of new hoops between the existing hoops. Other arrangements such as helical glycan strands are difficult to visualize. If the glycan strands were parallel to the long axis, then strand insertion would lead to cells growing in width rather than length, and a helical arrangement would lead to changing angles of helix as new strands were inserted within the helices. The arrangement of strands at the pole is unknown at this time.

Although there may not be a perfect alignment, as has been argued by Koch (56), a hoop arrangement is the best way to allow lengthwise extension of the growing cell. New strands are inserted between preexisting glycan chains, and the cell grows in a lengthwise direction between divisions. If the glycan chains were placed in the axial direction—i.e., in the lengthwise direction of the cell—then insertion of new strands between preexisting glycan chains would lead to an increase in cell circumference. It is clear from even the earliest observations of living cells in the light microscope (75) that rod-shaped cells grow lengthwise, which is strong a priori support of the hooplike strand arrangement model.

Biochemical analysis indicates that glycan strands are relatively short and cannot extend around the entire circumference of the cell (41). Therefore, the peptidoglycan layer is made up of short overlapping strands that collectively go completely around the cell. The absence of long-range order in this arrangement does not change the fundamental conclusion that the insertion of new glycan strands between resident glycan strands leads to the growth of the cell in the axial or lengthwise direction.

Cell wall segregation brings up many questions. Where does the old wall go at division, where is the new wall synthesized and inserted, and what is the inheritance pattern of peptidoglycan over a number of division cycles? These questions suggest that the segregation of preexisting material is determined by the topological insertion pattern of new material. Segregation is the mirror image of the location and insertion mode of new wall material. An example of the relationship of insertion to segregation is that a thoroughly conservative mode of surface segregation would lead to an insertion pattern whereby only half of the growing cell would have surface label following a pulse-label.

Idealized Pattern of Surface Synthesis

Cell growth leads to both conservative and dispersive modes of segregation because the poles and the side wall behave differently in their segregation patterns. This can be shown with cells fully labeled in the peptidoglycan by a specific label such as DAP. The poles act as conserved segregation units, with both labeled poles going to separate daughter cells. The side wall segregates in a dispersive manner: as new strands are introduced between old strands, the labeled side wall material is diluted by new material. Thus, after one generation, the side wall is uniformly labeled at half the specific activity of the side wall in the initial cell. (An important exception to uniform labeling and insertion in the side wall is noted below.) This idealized pattern of cell surface segregation (Fig. 2) is supported by a large amount of experimental evidence.

Fig. 2Segregation of peptidoglycan. For a cell with the surface totally lableled (black lines), there is a random segregation of the side wall material (indicated by the decreasing shading at each generation). The old pole material is conserved (indicated by the white material). Thus, three distinct areas at the cell are postulated: the older poles where the density of label does not change, new poles which are unlabled, and side walls where the density of label halves at each generation.

Source:

The segregation pattern is a reflection of the synthetic pattern of the cell wall. The surface of E. coli and other rod-shaped bacteria is considered to be a cylinder capped by two hemispheres. These polar hemispheres are stable, with no growth occurring in them once they have formed (57, 58). Side wall synthesis, on the other hand, is spread throughout the side wall and does not appear to occur in any zonal pattern. If we know that new material is inserted, at least in the side wall area, between any preexisting pair of hoops, then by definition we must have a dispersive pattern of side wall segregation. The most compelling evidence indicates that the overall pattern is uniform incorporation over the side wall (89, 90).

From this model, a number of questions arise. For example, is there a sharp distinction between the side wall and the pole, or is this merely an anthropomorphic construct that does not actually exist for the cell? Given that there are poles, we can ask whether the poles are completely stable or whether there is some turnover or metabolism of polar peptidoglycan. With regard to the side wall, is insertion completely random over the length of the cell, or are there some preferred sites of strand insertion or some connections between strands that are resistant to the intrusion of a new one?

Location of Insertion of Peptidoglycan on the Cell Surface

The earliest experiments on cell wall growth used fluorescent antibodies to label the cell surface. These studies indicated that in E. coli, the side wall grows diffusely with no conserved areas (4, 17, 66). Additional autoradiographic evidence showed that new cell wall material could be inserted over the entire surface of the cell (11, 12). Similar results were obtained for the matrix protein attached to the peptidoglycan (5).

There have been numerous reports that peptidoglycan is not stable and is actually released to the medium (24, 37, 39). This is a complication for segregation studies that is difficult to evaluate at this time, since the amount of release reported varies between 7 and 50% per generation.

The clearest demonstration of this random pattern of insertion is the autoradiographic evidence showing that there are no apparent zones of preferred synthesis prior to invagination (90). Biochemical support came from studies of the peptidoglycan acceptor-donor radioactivity ratio (ADRR), a technique that measures the pattern of strand insertion into the cell wall. These studies also indicated that there were no conserved areas of the cell wall and that new material could be inserted between any two strands (10, 26). (See discussion of ADRR below.)

The question of whether cell wall growth is symmetrical— i.e., whether it is equal in both of the prospective two new daughter cells or occurs primarily or entirely in only one—is an important problem. The symmetrical nature of cell growth was conclusively demonstrated (at least for peptidoglycan) by the sophisticated statistical analysis of Verwer and Nanninga (84). They analyzed the distribution of radioactive DAP on each of the two cell halves of dividing cells. They determined whether differences in the number of grains in one half of the cell were due solely to statistical variation or whether there was a significant difference that would fit an asymmetrical pattern of peptidoglycan synthesis. Their results showed that synthesis of cell wall in the two cell halves was the same, indicating that there is no conservation of material, with one daughter cell receiving a majority of the new peptidoglycan material.

Additional evidence for random insertion over the side wall comes from membrane elution experiments demonstrating that there is no sudden drop in the elution of peptidoglycan label from the membrane as would be expected if there were a conservative mode of synthesis (M. Ma and S. Cooper, unpublished results). Since elution of label can continue for up to 10 generations, it can be concluded that insertion, and hence peptidoglycan segregation, is random with no large conserved subunits. (To be rigorous about this analysis, it should be stated that any regular and precise pattern of zonal growth and conservation of zones is excluded by this experiment; zonal growth that is completely random and not associated with any particular cellular location during the division cycle may not be excluded by this experiment.)

At some time in the life of the cell, new poles are formed in the center of the cell; these poles are made up of completely new material. Thus, the newborn cell is composed of three zones of cell surface: an older pole made in some previous generation, a new pole just made, and a cylindrical surface in between, made up of a mixture of old and new strands.

Evidence for Zonal Growth

Although a nonzonal pattern of synthesis is indicated by the data cited above, the first two decades of the study of bacterial cell wall growth were dominated by the idea of growth zones, i.e., localized areas of cell surface growth. There are three sources for this idea. One is the early recognition that in Streptococcus spp., the cell wall is a rigid structure that grows with one zone of wall growth; old cell wall is not metabolized, and new material is not inserted within it. This zonal pattern of growth was readily extended to gram-negative bacteria. The second source of the idea of zonal growth is the analogy to DNA synthesis. DNA replication is regulated by the insertion of new replication points at the origins of DNA. DNA segregation is conservative, as befits our understanding of the mode of DNA replication. This model was applied to cell wall synthesis, suggesting that there may be growth zones and that at some time or times during the cycle, new growth zones are inserted. Such a zonal mode of growth would lead to a conservative or semiconservative mode of segregation. The third element that supported the idea of zonal growth was the proposal of the replicon model to explain DNA segregation. In the absence of a visible mitotic apparatus, it was proposed that the regular segregation of DNA at division could be explained by the binding of DNA strands to the cell surface, with cell wall growth taking place between the bound DNA strands (50, 51). The wall growth between the surface-bound DNA strands could lead to separation and sequestration of DNA in the two new daughter cells. It was believed that zonal growth, particularly in the center of the cell, was an important requirement for DNA segregation.

The experimental support for zonal growth came mainly from autoradiographic experiments. When cells were labeled for a short time with DAP and analyzed for the location of grains by autoradiography using the electron microscope, it was found that for cells of all sizes there was a preferential location of zones of incorporation in a relatively narrow band at the center of the cell (70). This result implied a preferential zone of growth in the center of the cell. New zones would appear in the new daughter cells at some later time. Immunofluorescence analysis also supported the insertion of discrete zones of synthesis in the side wall of rod-shaped cells (16). When pulse-chase experiments were performed, the evidence for zones was ambiguous (77), and it was concluded that there was a randomization of the material in the initial zone. At the same time that autoradiography suggested a zonal growth mechanism, kinetic measurements of cell wall synthesis during the division cycle indicated that in the middle of the cycle there was a sudden doubling in the rate of peptidoglycan synthesis (48, 67). This result was supportive of the idea of zonal growth. In retrospect, the early indications of a central zone of growth were probably due to elevated incorporation at the new septum.

Donachie and Begg proposed the unit cell model as a specific type of zonal growth (31). This model proposes that the cells grow only from one pole, producing one daughter cell with completely new and one with old peptidoglycan in the side wall. Support for the unit cell model came from microscope observations of cells growing in only one direction. One difficulty with these experiments is that one cannot eliminate a preferential attachment of one cell pole to the substrate and free movement of the other pole, which could produce the appearance of growth in one direction. However, this model got additional support from the finding (6, 7) that phage attachment sites are inserted asymmetrically on the cell surface. Verwer and Nanninga (84) disproved the unit cell model, at least for peptidoglycan, by analyzing the distribution of radioactive DAP on each of the two halves of dividing cells. The unit cell model is also incompatible with the pattern of elution of DAP from cells bound to the membrane in a membrane elution experiment (18).

Is the Pole Metabolized?

A sophisticated analysis of earlier DAP incorporation experiments suggested that there could be a small amount of incorporation of label into old poles (57). A subsequent reanalysis of the data from high-resolution, electron microscopic, autoradiographic studies (using a computer program that corrected for effects of normalization of cell lengths) indicated that the poles, once formed, were extremely stable (58).

An interesting result with regard to the question of pole stability is the report that the major outer membrane protein, OmpA, is concentrated at the poles (8). Cephalexin prevented the insertion of this protein at the poles. Removal of the antibiotic allowed the randomly dispersed protein to migrate to the pole and possibly to the newly forming septum. This result suggests some metabolism of the pole after its formation.

Growth of the pole in a narrow band, termed the "leading edge" model, has been proposed on the basis of autoradiographs indicating that at all stages of pole growth, the width of labeled material incorporated into the pole is invariant (87).

Peptidoglycan Segregation at the Strand Level

This discussion has dealt primarily with the segregation and insertion of peptidoglycan at the level of visible areas of growth. At a deeper level, we may consider whether strand insertion occurs between every strand or whether there are some connections between strands that cannot be broken. The ADRR method has been introduced to answer this question (10, 21, 26). The ADRR is a measure of how much labeled DAP is in the donor penta- or tetrapeptide, compared to the acceptor position, during a short period of labeling. A zero ADRR indicates that all of the radioactivity is in the donor position and that new strand insertion into peptidoglycan is occurring by single-strand insertion. Some ADRR measurements are consistent with a two-strand insertion mechanism (10), while others indicate a single-strand insertion mechanism (26). There has been disagreement regarding the meaning of the ADRR values (21, 24, 30, 33). Anyone interested in using ADRR analysis should become familiar with these papers.

Höltje (49) has proposed a model in which three strands are inserted at the same time that one strand is removed from the peptidoglycan. His model explains the maintenance of the specific shape of the bacterial cell. By having the newly inserted strands follow along a previously existing strand, cell circumference is kept constant. A processive degradation of a resident strand, coupled with the simultaneous insertion of new strands, coordinates growth while maintaining cell shape. Höltje’s proposal also accounts for the turnover of peptidoglycan. No biochemical results are supportive of this model (chapter 6, this volume). Furthermore, this model is inconsistent with ADRR measurements indicating single-strand insertion (26).

It has been observed that the degree of cross-linking increases after insertion (9). This is a surprising result, since it is hard to imagine how new cross-links can be formed once a strand is inserted. The removal of the final d-alanine from the peptidoglycan pentapeptide by a carboxypeptidase precludes cross-linking, as the energy for transpeptidation is no longer available. Furthermore, as the peptidoglycan appears to be under stress due to turgor pressure (55), it is probable (but as yet unproven) that cross-links cannot be introduced between strands that are physically too far apart. One explanation for the observed increase in cross-linking is that there may be a replacement of low-cross-linked strands by more highly cross-linked strands. With a uniform stress over the surface, the less densely cross-linked strands will be stressed more at each cross-link. This is because the stress, which is constant over the surface of the cell, is spread among fewer cross-links in a low-density cross-linking region than in a more densely cross-linked region. Thus, the low-density cross-linked regions will be preferentially replaced by new strands of peptidoglycan material. Since, by definition, the less densely cross-linked strands have a cross-linking value below the average of the inserting strands, there would be an increase in cross-linking of the resident peptidoglycan as it ages (22). A numerical example can illustrate this process. Consider a peptidoglycan with an average of 25% cross-linking made up of equal amounts of 20, 25, and 30% regions. A newly inserted strand has an average 25% cross-linking. If this new strand preferentially replaces a 20% region, the average degree of cross-linking would increase. This "evolutionary" mechanism allows the cell to continue to strengthen its peptidoglycan as the peptidoglycan drifts toward higher cross-linking values.

Zonal versus Dispersive Growth of the Side Wall Peptidoglycan: Conclusion

The conclusion of this review is that growth of the side wall peptidoglycan of gram-negative, rod-shaped bacteria is dispersive. Zonal growth has been reported, but my conclusion is that the side wall is subdivisible down to the molecular level, with no evidence for major zones of growth or conservation of material.

Segregation of Membrane Lipids

The measured pattern of membrane synthesis during the division cycle, using glycerol and palmitic acid as membrane labels, is similar to the pattern of peptidoglycan synthesis during the division cycle (36). A simplified view of membrane synthesis is that it is made in response to the increase of surface as determined by the growth of the peptidoglycan layer. One idea that arises from this observation is that the segregation of membrane will be similar to the segregation of peptidoglycan. But because peptidoglycan is a more rigid structure, and membrane may be more fluid, there could be differences in segregation patterns between membrane and peptidoglycan.

Early support for the dispersive segregation of membranes comes from density shift experiments (81, 88). In experiments formally analogous to the Meselson-Stahl experiments with DNA, it was observed that the membrane grows in a nonconservative manner. These experiments were sensitive enough to find a small amount of conservation of membrane during growth. In more measurements over many more generations, Green and Schaechter (40) labeled cells with glycerol and used autoradiography to demonstrate that membrane segregation was dispersive for eight generations. Thereafter, the presence of conserved units of membrane (e.g., floating islands of original membrane from the labeled cell) was indicated by analysis of the label over the individual cells. Calculations indicated that there are approximately 256 independently segregating subunits in the cells. (In this experiment, Arthur Koch in a footnote extended the Poisson distribution to cells of different original size and with the age distribution of a growing culture. This equation includes the important ideas that cells of all ages are not represented equally in a culture and that the cells in the culture are not all of the same size.)

Segregation of Membrane Proteins

Studies of the segregation of membrane proteins have produced findings that support different and apparently irreconcilable models. Some experiments have suggested a zonal growth model, while others have proposed a dispersive and nonconservative mode of protein insertion.

A zonal pattern of growth has been suggested for both the inner (29) and outer (2, 3, 51, 53, 54, 81) membranes. Kepes and Autissier looked for the unequal segregation of β-galactoside permease as a membrane marker among the progeny after cessation of enzyme synthesis (2, 3, 53, 54). Studies on cytochromes supported a nondispersive segregation pattern (78). In contrast, dispersive segregation was found in studies of the membrane-bound anaerobic nitrate reductase in E. coli (13). Additional experiments on β-galactoside permease (1) and phage receptors (60) support a conservative insertion segregation mode.

Segregation of membrane components has also been studied by pulse labeling and autoradiography. Finding that the labeled material is inserted diffusely over the surface is supportive evidence for random segregation. Electron microscopic studies of the insertion of outer membrane proteins indicated that there was no clear zonal growth (45, 78, 86). The protein studied, the LamB protein, was inserted diffusely over the entire surface. Analysis of the potential for diffusion of proteins within the outer membrane indicates that this does not account for the observed homogeneous distribution of pulse-inserted outer membrane proteins (86). This work supports the random segregation model of surface segregation.

On the other hand, earlier work on other outer membrane proteins (OmpF and OmpA) using autoradiography supported the zonal insertion model (6, 8). In addition, other studies of membrane proteins using a fluorescent antibody supported zonal insertion (29). Studies of outer membrane receptor proteins for phage lambda (71) or T6 (6, 7) supported a zonal model of surface growth as well. At this time it is difficult to reconcile these experiments with others suggesting that there is a symmetrical synthesis of the bacterial cell surface (85).

There is no reliable information on the segregation of flagella and fimbriae. It is generally assumed, without any strong evidence either way, that the fimbriae and flagella are made at random points on the cell surface and are segregated along with their nearby cell surface peptidoglycan or membrane components.

Summary View of Membrane Segregation

How can this apparently contradictory and confusing set of results be summarized to give a unified picture of membrane growth during the division cycle? My personal conclusion is that there is no significant zonal growth in the side wall of gram-negative rod-shaped bacteria for the membrane or peptidoglycan.

As a final thought, the cell may not benefit from a zonal mode of growth. A diffuse model of surface growth avoids the accumulation of patches of material that are old and which may have various errors of aging. With the continuous insertion of newer material next to older material, the cell surface can be constantly strengthened and errors can be repaired in a manner analogous to proofreading for informational macromolecules. Whether errors in a zone could be repaired by noninsertional means has yet to be determined.

SEGREGATION OF DNA

During unhindered, exponential, balanced bacterial growth, essentially every cell in a bacterial culture is viable. This implies that every cell in a culture contains a genome, since termination of replication occurs sometime prior to division, and therefore at least two complete genomes will be present in the dividing E. coli (chapter 102, this volume). The problem of nucleoid segregation is to define how the two genomes are apportioned so that each daughter cell always gets one genome. Nucleoid separation and partition of completed nucleoids to two daughter cells have been discussed in detail elsewhere (46, 61, 63, 72, 74; chapter 105, this volume).

I will now turn to a discussion of the inheritance and segregation of the individual strands of DNA to daughter cells. It is not always appreciated that every cell can receive a genome with perfect fidelity, while at the same time individual strands may segregate nonrandomly. As will be seen, this analysis of strand segregation gives an unexpected insight into potential mechanisms for the movement of DNA into newly formed daughter cells.

Experimental Analysis of Strand Segregation

Lin et al. (62) analyzed strand segregation by using the chain-forming methylcellulose (Methocel) technique. This method allows the growth of E. coli in chains so that the cells retain their respective order in the chain. A random segregation pattern was observed (62). We now understand that this observation was due to the fact that the cells being analyzed in these experiments contained up to eight or more labeled strands, and this complexity precluded any observation of nonrandom segregation. Subsequently, in a technical tour de force, Pierucci and Zuchowski (69) demonstrated that segregation of strands is nonrandom. These authors studied cells with known chromosome configurations (i.e., cells with either two or four strands) and determined the locations of the labeled strands in chains of cells formed in Methocel. They compared their data with a number of models and proposed that in each cell, one strand segregates randomly and one segregates nonrandomly. Nonrandom segregation was also found in membrane elution experiments (68).

A simplified approach utilizing presegregation of labeled DNA prior to chain formation confirmed the nonrandom pattern but led to a different conclusion. Tritiated thymidine-labeled cells were allowed to grow in unlabeled medium for a few generations prior to chain formation; these cells had only one labeled strand, and consequently only one cell in a chain was labeled (28). It was observed that in a chain of four cells, the outermost cell was preferentially labeled. Nonrandom segregation was greatest at slow growth rates, with a more random pattern appearing at faster growth rates. The nonrandom pattern fit a rule stating that strands go preferentially to the same pole that they went to in the previous generation. Because the probability that a strand will continue in the same direction it went previously is greater than 0.5, the model was called the strand-inertia model. There was no permanent association of any DNA strand with either pole (27). Thus, nonrandom strand segregation is probabilistic rather than deterministic in nature (Fig. 3).

Fig. 3Nonrandom DNA strand segregation. Consider an exponentially growing culture with DNA labeled at generation 0. Before labeling, there is one replicating but unlabeled chromsome. At the instant of labeling (generation 0), a cell has two labeled strands as the cell has one replicating chromosome), as indicated by the striped lines. The corresponding unlabeled strands are indicated by the solid lines. The cells will grow for a number of generations (N) and then be allowed to form chains in Methocel. At N+2 generations (a four-cell chain), one can determine the proportion of chains labeled in the inner and outer position. Here only the labeled strands are shown, as these are the strands observed by autoradiography. More cell chains are labeled in the outer position (p) than in the inner position (q). Although this result varies slightly with growth rate, the basic mechanism and underlying quantitative pattern are constant at all growth rates. Equivalent results are seen in cells forming chains immediately after pulse-labeling. The analysis is somewhat different, as there are two labeled cells in a four-cell chain. More inner cells than outer cells are labeled, since the unlabeled complementary DNA strands (not seen in autoradiography) are acting in the same manner as the strands observed in the N+2 generation. Because of equipartition, the labeled strands, which segregate away from the unlabeled strands, now go to the inner position of the four-cell chains. This is because the older, unlabeled strands go preferentially to the outer position of four-cell chains as shown in the N+2 generation.

Source:

A more accurate way to measure the nonrandom segregation of DNA has been introduced with a long-term, automated-ratio method using the baby machine (Cooper and Ma, unpublished results). Cells are labeled with tritiated thymidine and a ¹⁴C-amino acid, placed on a membrane, and eluted for up to 10 generations. The amino acid label per cell is halved each generation. After the third generation, the thymidine label per cell is reduced by less than half each generation. What is happening is that a constant fraction of the extant thymidine is eluted each generation, but this amount is less than half of the amount of bound label. Thus, the label in the bound cells is not decreasing by half each generation. This pattern of radioactivity release is due to the nonrandom segregation pattern. Because of the differences in segregation of the protein (random) and the DNA (nonrandom), the ratio of tritiated thymidine to ¹⁴C label increases. The slope of this ratio, plotted on a semilogarithmic graph, is a direct and sensitive measure of the degree of nonrandom segregation. Segregation values from 55/45 up to 60/40 have been obtained by this method (Cooper and Ma, unpublished results). Strand segregation is nonrandom and probabilistic in nature and, as we shall see below, serves as an important indicator that the cell surface is an agent of strand segregation.

The question can be raised as to whether the apparently slight deviation from 50/50 is important, since an all-or-none segregation pattern may have been expected. As will be seen below, it is just this slight deviation from randomness, obtained through rather precise measurements, that makes this finding important. This value fits with a model that relates nonrandom DNA segregation to the asymmetry of the cell in terms of cell wall insertion and segregation. It is to this model that we now turn.

Helmstetter-Leonard Model for DNA Strand Segregation

An elegant mechanistic explanation for nonrandom DNA segregation has been proposed by Helmstetter and Leonard (42, 43). When cells were pulse-labeled and allowed to form chains immediately with no presegregation of the labeled DNA, the center cells of four-cell chains were preferentially labeled. This is exactly what is expected from the presegregation finding that the labeled strands preferentially go to the outer cells of a four-cell chain. Helmstetter and Leonard noted that segregation of these new strands preferentially to the center cells in the four-cell chains mimicked the synthesis and segregation of cell wall material. If one distinguishes between old and new cell wall material, new and old material is randomly placed over the side wall, new material appears preferentially in the newly forming poles, and the outer poles of the cell contain all of the older cell wall material (Fig. 2). At division, the side wall has been intercalated with new cell wall material such that the amounts of old and new material in the side wall are the same in the two daughter cells. In the center of the dividing cell, however, the new poles of the daughter cells have been synthesized from new material. At the outer ends of the dividing cell, the poles are made up solely of old material, with no new material inserted in the poles during the previous generation of growth. Thus, in every E. coli cell one pole is brand new, having been formed at the previous division; the other pole is older. Helmstetter and Leonard proposed that this asymmetry is the basis for the asymmetric segregation of DNA. Their model states that a newly synthesized DNA strand tends to bind preferentially to, and to segregate with, new cell wall. Segregation will then be nonrandom. The degree of nonrandomness in this model is determined by the fraction of the total surface that is pole material. The greater the fraction of pole material, the more nonrandom the segregation pattern. Thus, a short, thick cell will give a more nonrandom pattern than a similarly sized cell that is thin and long. An alternative formulation is that older strands preferentially bind to older cell wall material. The ultimate result is the same, but the biochemical tests might be different. Helmstetter et al. have presented other formulations whereby the nonrandom segregation pattern is discussed in terms of excluded volumes and excluded areas (43, 44). The binding of the new DNA strand to the new cell wall material occurs only once, and for one generation. Afterwards, the DNA strand acts in reaction to the segregation pattern of the complementary, newer daughter strand.

This model is consistent with seemingly contradictory experimental results. The degree of nonrandom segregation decreased with increasing growth rate when measured by the presegregation Methocel method (28). In contrast, an invariant degree of nonrandom segregation was observed with the membrane elution method (68). The explanation of this apparent paradox has been presented in detail elsewhere (22). Briefly, in the cell population as a whole, segregation appears more random with increasing growth rate because the proportion of symmetrical segregation to asymmetrical segregation increases. In the Methocel method, there is an increase in the proportion of cells with random segregation at higher growth rates. This is because poles present in these cells were both made during the period just before strand attachment and thus, from the strand’s point of view, are equivalent. With the membrane elution method, cells are continuously lost from the membrane, and at one particular division there is a clear indication of nonrandom segregation without the interference of these symmetrical cells.

The important point is to relate DNA strand segregation to the cell surface. In the slowly growing cell, the daughter cell DNA "chooses" between an old pole, present during DNA synthesis, and a new pole, made after DNA synthesis ended. In the more rapidly growing culture, the four cells produced during the time of DNA replication and nucleoid partition are divided between two kinds of cells: (i) cells with an old (original) pole and a new pole and (ii) cells with two "new" poles in the sense that they were made during the period of replication and segregation. Neither of the poles in these cells was made before the initiation that is related to the segregation event. Thus, some of the DNA would segregate asymmetrically (the cell has both a very old pole and a new pole), and some would segregate symmetrically (there are two new poles).

The explanatory power of the Helmstetter-Leonard model should not be overlooked. It takes two contradictory results, accepts both of them, and explains them with a unified model that fits the data (22). As the Helmstetter-Leonard model is based on the relative shape of cells, the constancy of segregation values (obtained by the membrane elution method) implies that cell shape at different growth rates is invariant. A constant cell shape has been invoked to explain a large variety of other data (22). If the segregation model of Helmstetter and Leonard is proven to be correct, then the constant segregation data (68) support a constant cell shape (22).

Segregation is more nonrandom at lower temperatures (25). One implication of this result, in view of the Helmstetter-Leonard model, is that at lower temperatures the cells should be squatter than at higher temperatures. This change in shape would increase the area of the cell devoted to pole, thus making DNA segregation more nonrandom. When the length-to-width ratio was determined in cells growing at different temperatures, there was indeed a change in shape; cells decreased their length more than their width (80). This result supports the shape determination of nonrandom segregation (44).

Minichromosome Segregation

The bacterial chromosome exhibits nonrandom strand segregation as well as equipartition of chromosomes. Equipartition means that even though strand segregation is nonrandom with regard to direction, each cell always has one chromosome. In contrast, it has been proposed that the minichromosome (an autonomously replicating circle of DNA controlled by the normal bacterial origin) exhibits nonrandom segregation but does not exhibit equipartition (42). This leads to the appearance of minichromosomeless cells and explains how, in the face of cell cycle-specific replication of the minichromosome, there is a drift toward higher minichromosome content in cultures under continuous antibiotic selection (22, 42, 52, 76).

A different model explaining the nonrandom segregation of DNA has been considered (14). It was proposed that the nonrandom segregation of the DNA, while not perfectly deterministic, could be described as "nonrandom, with a certain degree of randomness." A critical analysis of this model has been presented (22).

Relationship of DNA Strand Segregation to Cell Envelope Segregation

The first experiments on DNA segregation defined the number of units to be considered during a segregation experiment. At the time the early experiments were performed, it was just becoming known that there were but a few units of segregating genetic material in the cell; i.e., the cell did not contain a large number of randomly assorting chromosomes. This result is subtly related to the proposal by Jacob et al. (50) of the replicon model. From observations that there was an equipartition of DNA to daughter cells, they postulated an attachment of DNA to the cell surface; they proposed that the cell surface was analogous to the mitotic apparatus of higher cells (72, 74; M. Schaechter and U. von Freiesleben, personal communication). One of the predictions of the replicon model is that the DNA may be permanently attached to one of the poles of the bacterial cell. If the attachment of DNA to the surface were permanent, and if the attachment site was a fixed position in the cell relative to the poles of the cell, then a particular strand would always segregate toward one pole at division. This permanent attachment has not been observed. There does appear, however, to be some relationship of cell surface growth to the segregation of DNA. There is very likely a "bacterial equivalent of mitosis" (74; Schaechter and von Freiesleben, personal communication), but it does not appear to have the same mechanical relationship to DNA separation as the complex mitotic apparatus of eukaryotic cells. Because side wall growth of the rod-shaped E. coli is dispersive, the DNA attachment to both side wall and polar material leads to more probabilistic predictions.

Do Nucleoids Jump at Termination?

It has been proposed that the positioning of newly formed nucleoids, from the center of the cell at 0.5 to new positions at 0.25 and 0.75, occurs with a sudden movement or "jump" that cannot be accounted for by cell growth (32, 47). Løbner-Olesen and Kuempel (63) have proposed an alternative view of the apparently rapid movement of nucleoids away from the center of the cell whereby origins of chromosomes are separated continuously after initiation. It is the final formation of the separated nucleoid that makes them appear to jump to new locations. The original experiments of Schaechter et al. (73) produced data indicating that nucleoid formation was a continuous process (i.e., new daughter nucleoids are being formed prior to termination) and that the nucleoids observed were connected until termination occurred (25).

CELL POLARITY

A recent report has suggested that even E. coli, which is believed to be the quintessentially symmetrical cell, may be functionally asymmetric with regard to its cell surface (65). The observation leading to this proposal was the finding, by immunoelectron microscopy, of chemoreceptor complexes clustered predominantly at the older cell poles. Thus, the cell is asymmetrical with respect to the location of particular molecules.

This finding of polarity in E. coli has led to the proposal that polarity is a general and important phenomenon (64). In contrast, one may view the observed polarity as merely the inevitable consequence of the fact that all rod-shaped cells have poles of different ages. More generally, all bacteria have one old and one new pole. It may be the age of the pole that determines its chemical constitution; i.e., there may be no inherent function, meaning, or purpose to the observed polarity. When the younger pole eventually matures and becomes an older pole, it will gain the same chemical characteristics that were found in the original older pole. In this view, even the complex life cycle of Caulobacter crescentus is a simple elaboration of the basic cell cycle pattern of E. coli (20, 22). The linear sequence of events—from initiation of DNA replication, to replication, to partition and segregation, and finally to pole formation—is not complete at the separation of cells but continues as the pole matures after cell division. This maturation is not as evident in E. coli as in C. crescentus, with its sequence from a bald pole, to a pole with a flagellum, to the formation of a stalk. This maturation has been given the term "E period" since it logically follows the B, C, and D periods of DNA synthesis and segregation (20, 22). The point of the E-period proposal is that the poles forming in the center of the cell are younger than the poles at the ends of the cell. Such an unavoidable age difference may lead to the observation of polar differentiation. This holds true for E. coli as well as for C. crescentus.

METHODOLOGICAL CONSIDERATIONS

It is generally believed that topological models require optical (e.g., light or electron microscopic) observations as a source of experimental support. This is not the case, as can be seen by the successful study of the segregation of each of the cell components using membrane elution methods rather than optical methods. It is possible to infer segregation patterns from membrane elution studies by reconstructing the segregation pattern from the radioactivity eluted at each generation.

PROBLEMS REMAINING FOR SEGREGATION OF CELL STRUCTURES

There are a number of results in the literature that do not conform to the idealized models presented for the segregation of each of the parts of the cell. These experiments give results that disagree with the general model (such as those proposing zonal growth or unit cell growth). How should we handle these early data? Should we continue to spend a great deal of time with these older results when newer, improved methodologies have led to different results? In particular, how should older results be viewed when newer results are explained in satisfying molecular terms?

I propose that the data to date, after choices are made between different and perhaps opposing results, fit the general models presented here and summarized in Fig. 1, 2, 3. Segregation of the cytoplasm is dispersive, segregation of the cell wall is understood in terms of conserved poles and dispersive, nonzonal side wall segregation, and strand segregation of DNA is nonrandom and follows, in a formal way, the pattern of wall synthesis and segregation.

Our understanding of cytoplasm and cell surface synthesis enables us to understand the segregation pattern of these cell components. The mechanisms determining the nonrandom DNA segregation pattern, however, remain a mystery. The primary remaining problem with regard to understanding the apportionment of material to daughter cells at division relates to DNA segregation.

ACKNOWLEDGMENTS

Comments and suggestions regarding this review were given by Arthur Koch, William Donachie, Moselio Schaechter, Conrad Woldringh, and Jochen Höltje. The thoughtful, creative, and collaborative editing by Alexandra Cooper has made this article a joy to write. I thank her for this and look forward to many more collaborative efforts.