Compaction and Segregation of DNA in Escherichia coli

1. Introduction

In contrast to the chromatin in eukaryotic cells, bacterial DNA occurs free in the cell, not surrounded by a membrane. Nevertheless, the DNA can be observed as a distinct, central region, called the nucleoid. What macro-molecular interactions or activities induce this phase separation, compact the DNA and prevent it from dispersing throughout the cytoplasm?

Already in 1956, Mason and Powelson [1] showed by phase-contrast microscopy of cells grown in rich medium and supplemented with gelatin, that a bright, central structure expands and divides in step with the growing cell. This light appearance of the nucleoid structure suggests that it has a low refractive index and thus a low density, indicative of a lower protein concentration. At that time, it was still assumed that bacteria contained “nuclei” or nuclear structures and that they possessed “a mitotic apparatus consisting of a centriole and spindle” (see for a discussion of the changing views of the nucleoid Robinow and Kellenberger [2]).

Why is the DNA not dispersed throughout the whole cell, but does occur in distinct regions as observed by phase-contrast microscopy? This question has been studied by many groups applying molecular dynamics simulations [3,4] and theoretical computations based on equilibrium statistical mechanics [5,6]. These involve formulations of the free energy of the system according to models for interactions between the supercoiled DNA and macromolecular crowders that cause the compaction of the nucleoid. Whereas Odijk’s so-called depletion theory [5], merely considered the interaction of supercoiled DNA with soluble proteins (see Appendix A), the approach of the Männik-group [6,7] also considered the crowding interaction of DNA with larger particles like polyribosomes. Both theories predict the phase separation between cytoplasm and nucleoid, but the question remains whether the low refractive index as observed by phase contrast microscopy can be confirmed. Here, this question is discussed by comparing the properties of cells described in both approaches (see Appendix B).

Nucleoid compaction can also be explained if the solvent quality of the cytoplasm for DNA is considered [8] or through transcriptional activities that induce folding of transcribed and supercoiled regions [9,10]. Also for these studies is the question whether the low refractive index of the nucleoid is ensured.

The mechanism(s) of DNA compaction can be expected to influence DNA segregation. Because replication and segregation go hand-in-hand, the movement of replicated DNA strands through the parental, un-replicated DNA network already starts at initiation within the so-called initial replication bubble or orisome [11]. From here, the newly synthesized chromosome arms, containing the leading and lagging strands, have to segregate to separate halves of the daughter nucleoids [12,13]. There are two views that try to explain the disentanglement of daughter strands and the formation of daughter nucleoids: (i) an active segregation mechanism like the tripartite ParAB-parS system [14], occurring in the majority of bacterial species [15,16,17]. The system may be helped by SMC (structural maintenance of chromosome) proteins and loop extrusion mechanisms [18,19], or (ii) a passive segregation process based on de novo DNA synthesis. Here, it is proposed that the newly synthesized chromosome arms or replichores do not become mixed or entangled because of their physical differences (explained in Section 4.1, Figure 8). From initiation onwards, the arms are maintained as separate entities, excluding each other and enlarging by de novo DNA synthesis in the meshwork of parental DNA. This is proposed in the “four-excluding-arms model” that explains not only how the two daughter strands segregate into the prospective daughter cells, but also how each chromosome arm ends up in an opposite halve of the two newly-formed nucleoids.

2. The Nucleoid as a Low-Density Structure

2.1. Immersive Refractometry

The observations (Figure 1A) of Mason and Powelson [1] were confirmed by studies using PVP [20] or bovine serum albumin (BSA; Figure 1B; [21]) to increase the refractive index of the external medium surrounding the cells. As discussed by Barer et al. [22] such an increase in the density of the immersion medium reduces the light scattering of the dense bacterium and abolishes the disturbing halo around the cell. At the same time, the phase contrast of internal structures within the cell, like the nucleoid, is increased. In addition, this immersion technique enables the estimation of the refractive index and thus of the density of the internal structure, when its light intensity is equal to that of an external medium with known concentration. Mason and Powelson [1] argued that the observed low-refractive index structures did not arise from an effect of the gelatin on the cell, because the same structures could be seen after fixation and specific staining of DNA with the Feulgen reaction.

Mason and Powelson [1] also showed how the bright, central structure in the cells expands and divides in step with the growing cell. Likewise, the movie made by Yamaichi and Niki [23] from rapidly growing E. coli cells immersed in gelatin showed how the light nucleoid areas enlarge and divide (Figure 1C), confirming that DNA replication and nucleoid segregation go hand in hand [24].

Recently, a reduction in density of the nucleoid as compared to the cytoplasm was confirmed by Spatial Light Interference Microscopy (SLIM). Using this quantitative phase-contrast microscope, Oldewurtel et al. [25] observed in E. coli cells growing in media without gelatin or BSA, a decrease in refractive index of about 10-30% at the site of the nucleoid. It is reassuring that the low-refractive index structures could also be directly visualized in growing bacteria by optical diffraction tomography (ODT; also called digital holographic microscopy, DHM; [26]), without immersion of cells in a dense medium (Figure 2). In this study the nucleoid shows (Figure 2) a lower refractive index (1.35) compared to that of the cytoplasm (1,37); these values are somewhat lower than those obtained by Valkenburg and Woldringh [27] discussed below.

Applying immersive refractometry (Figure 3), Valkenburg and Woldringh [27] determined the refractive indices (RI) of cytoplasm and nucleoid in slow growing E. coli B/r cells. They also calculated theoretical RI values, using both the data of Churchward and Bremer [28] for the macromolecular composition of E. coli and the measurements of cellular and nucleoid volumes of their cells. For these volume measurements an early confocal scanning light microscope (CSLM) was used, developed by Brakenhoff et al. [29]. (see for a review of its rediscovery Nanninga, [30]). This microscope showed an improved optical resolution and visualized the unstained nucleoid by mere absorption contrast (see Figure 1 in [27]). Assuming that the nucleoid only contained DNA and that all proteins and RNA were located in the cytoplasm, the theoretical value for the RI of nucleoid was lower than the experimental value. To match both RI values, it was proposed that the nucleoid must contain, besides the DNA, about 8.6% of proteins, whereas the cytoplasm would contain 21% protein and RNA. These values resulted in a reduction of the macromolecular density for the nucleoid of about 30% as compared with the density of the cytoplasm [27]. In the next sections, the question is discussed whether these observations can be explained on the basis of polymer-physical considerations.

2.2. Polymer-Physical Explanation of a Low-Density Nucleoid by Odijk’s Depletion Theory

Can the phase separation and the reduced density of macromolecules in the nucleoid as compared to that of the cytoplasm described above (Figure 1, Figure 2 and Figure 3), be explained by polymer physical interactions between DNA and proteins? Although it is generally accepted that macromolecular crowding, both in vitro [32,33] and in the cell [7] can lead to DNA compaction, the specific role of different crowders like soluble proteins or polyribosomes has remained unclear. Here, the free-energy approach of Theo Odijk, his so-called depletion theory [5], based on supercoiled DNA self-interactions and DNA-protein cross-interactions [34], is shortly summarized from the viewpoint of a biologist (see also the supplementary information in ref. [35]).

Based on concepts of equilibrium statistical mechanics [36], polymer physicists regard linear, double-stranded DNA as a semi-flexible, polyelectrolyte polymer consisting of stiff, freely-jointed segments in which chemical details like the sequence of base pairs are not considered. The segments can move relative to each other and because of the thermal motion that causes the surrounding solvent molecules to bounce continuously against the polymer (with an energy of ~1 k_BT), the elastic DNA rod undulates and takes the shape of a worm-like chain (Figure 4a). However, the chain is also resistant to bending, which is reflected by straight segments with a so-called persistence length, P. The chain can thus be viewed as a random walk with a step length longer than P, also called the Kuhn length, A, which is a statistical entity with the length of about 2P. Fluctuations of the long, thin chain due to Brownian motion, make the Kuhn segments to collide with each other causing the so-called excluded volume effect. Because the chains have a finite volume that is excluded to the rest of the chain and because a chain cannot pass through itself, the rod-like segments exclude a volume with the size of A²d_eff, representing an expansion or swelling of the long chain (see also Figure 3 in [37]).

These considerations for linear DNA also hold for plectonemic, supercoiled DNA. Under the bombardment of solute molecules, the superhelical DNA will also behave like an elastic structure causing a strong excluded volume effect through interactions between supercoiled Kuhn segments. We now consider a “superwormlike chain” to have a persistence of Ps and an effective Kuhn-length of A_s = 2 P_s (Figure 4b).

Equations for the free energy of the principal excluded volume interactions between the segments of the superhelical DNA itself and the steric, repulsive cross-interactions between the DNA double helix and soluble proteins are discussed in the Appendix A. Starting-point for the calculations in Odijk’s depletion theory [5] is obtained by formulating the excluded volumes for both interactions (B_self in Figure 4b and B_cross in Figure 4c). These expressions, scaled by thermal energy (k_BT and by the total volume of the system (V_cell), give us the free energy of the supercoiled DNA $(F_{self})$ and of its interaction with soluble proteins (F_cross). For the latter expression, only the high number of small, soluble proteins (~10⁶) is considered, while the possible contribution of large polyribosomes was omitted; because they occur in a much smaller number (~8000), their influence on the energy balance as calculated in Appendix A was assumed to be negligible [5].

The total free energy of the nucleoid F_nuc, when occurring dispersed throughout the cell, can be expressed as the sum of three free-energies, F_cross, F_self and a so-called mixing term, F_mix, expressing the electrostatic repulsion between the soluble proteins:

F_nuc = F_cross + F_self + F_mix (cf. Equation (15) in [5]).

In view of the phase contrast microscope observations (Figure 1), it can be assumed that a phase separation exists between nucleoid and cytoplasm. By taking the derivative of the above equation with respect to cell volume we obtain the force for compaction or osmotic pressure, π. By taking the derivative with respect to the number of proteins, we obtain the force for mixing or chemical potential, μ. A minimal free energy situation or thermodynamic equilibrium is imposed by equalizing these two forces throughout the two phases (Figure 4d). This results in two so-called co-existence equations (see Equations (16) and (17) in [5] and Appendix A). From these two equations and a third equation on the protein volume fraction of total proteins, the protein volume fractions in cytoplasm (v_c) and nucleoid (v_n), as well as the volume of the nucleoid (V_n) can be calculated and compared with the experimental values. It should be noted that the full- thermodynamic equilibrium here assumed for the calculations, does not hold for the living cell which is in a stationary dynamic (or steady-state) equilibrium.

As shown in Appendix A, interaction between the superhelical DNA and the soluble proteins (Figure 4c) “overwhelms” the self-energy of the DNA (Figure 4b), leading to the observed phase separation in a minimal energy situation. The results [5,34] indicated that the protein volume fraction in the cytoplasm of v_c = 0.166 is much larger than the protein volume fraction in the nucleoid (v_n = 0.06), as calculated for a nucleoid volume of about V_n = 0.1, explaining the observed refractive index difference between nucleoid and cytoplasm in the phase contrast microscope (Figure 1). However, as discussed in the next section, new estimates have indicated that the volume of the nucleoid may be much larger (about 3-fold; see Table 1 below).

(Here, it should be mentioned that Theo Odijk has re-evaluated the formalism of the equations developed for his 1998-depletion theory, with the tentative conclusion that multiple solutions could be possible. Regarding the effect of polyribosomes (see section 3.1 below), these could cause a perturbation not much larger than 10%; he will come back to this problem in future work (personal communication).)

2.3. Fluorescence Microscopy: New Estimates of Nucleoid Volume

Because in slow-growing E. coli cells, the nucleoid has a simple, rod-like shape, its volume has been estimated by several groups in such cells using fluorescence microscopy. The DNA was either tagged with fluorescent proteins or stained with DAPI. In Table 1 their results are compared with the early measurements of unstained E. coli B/r cells [27] obtained with a CSLM. The rare pictures that have been made with the CSLM show an exceptionally small nucleoid with a volume of 0.07 µm³. A slightly larger volume of 0.12 µm³ was estimated from an archived, original print marked with the original instrumental CSLM-magnification.

The measurements obtained from fluorescent microscope images of E. coli K-12 cells show a much larger nucleoid volume (0.23-0.27 µm³, column 4 in Table 1). Although with the CSLM, we look at the nucleoid visualized by light absorption [29] and in the fluorescence microscope at the light emitted from fluorochrome excitation, it is unlikely that this can explain the different volume measurements. We therefore decided to re-measure images obtained from the same E. coli B/r strain grown under similar conditions (unpublished growth experiment by P.G. Huls, 2001). As shown in Table 1 (last row), three nucleoid volumes were obtained, depending on the threshold used (Figure 5; see Table A2 in Appendix B).

An exceptionally large nucleoid volume (0.7 μm³) has been measured by the group of Jacobs-Wagner, using the Oufti open-source software package [8]. Other estimates by the groups of Kleckner [38] and Männik [7], arrive at values of 0.27 and 0.23 µm³ for nucleoid volume in newborn or early cell-cycle cells, comparable to the 0.24 µm³ value obtained for the DAPI-stained cells of Huls (see Table 1) at a threshold of 0.5 (Figure 5b). In the next section, these larger nucleoid volumes will be discussed and used for comparing the different polymer-physical approaches of Odijk [5] and the group of Männik [6].

3. Different Views on DNA Compaction

3.1. Compaction through Polyribosome Exclusion

In his analysis, Odijk [5] only considered the high number of soluble proteins (~10⁶ per cell), because it was assumed that larger crowder particles, being present in a relatively small number like ribosomes (8000 in slowly growing cells), would exert a negligible influence on the energy balance. However, later studies proposed that crowders like polyribosomes may well have an effect on nucleoid compaction because of their larger size. The groups of Bae-Yeun Ha [3] and Jaan Männik [7] have implemented Odijk’s free-energy approach, while taking into account not only the numbers, but also the size of the crowders. It should be noted that the nucleoid compaction here discussed, is the effect of steric interactions between crowders (proteins, (poly-)ribosomes) and DNA. A different view, discussed in Section 3.2 below, is that nucleoid compaction is not caused by steric interactions between DNA and crowders, but by the poor solvent quality for DNA in the cytoplasm.

In the study of the group of Männik [7], compaction of nucleoid volume was measured by decreasing the volume of the cell through an increase of the osmolality of the suspension medium (a hyperosmotic shock with NaCl) or by mechanical squeezing of cells in a microfluidic chip in which pressure could be applied. Their observations show that a 30% increase in crowder concentration can cause a three-fold decrease in nucleoid volume in living cells. For measuring nucleoid volume small cells in the early stage of the cell cycle were used, because they contain nucleoids with a simple ellipsoidal shape that can be measured relatively easy (see Table 1, row 4, column 3 and 4).

In an elaborate, theoretical follow-up of the experimental work of Yang et al. [7] the group of Männik has further investigated the different contributions of crowder species like soluble proteins as small crowders (diameter 5 nm) and polyribosomes as large crowders (radius of gyration 35 nm), interacting with supercoiled DNA (see Figure 1 in [6]). Their work is here summarized in the nucleoid compression curve of Figure 6. The curve shows that a rapid exclusion of polyribosomes from the nucleoid (green, dashed line) already generates a phase separation between cytoplasm and nucleoid (red, solid line), in a transition from cell (a) to (b) in Figure 6. The solid, red line shows that under the normal physiological condition about 95% of the soluble proteins are still present in the nucleoid (blue, dashed line). Only upon further increase of the concentration of small crowders, a decrease in nucleoid volume is obtained accompanied by depletion of soluble proteins from the nucleoid (Figure 6; transition from cell (b) to (c)).

The proposal of Männik and coworkers that polyribosomes are the dominant factor in establishing a phase separation between cytoplasm and nucleoid is based on considering hypothetical cells where only polyribosomes are present. The compaction curve of these simulated cells shows initially the same phase separation as when polyribosomes are present together with proteins. However, with only polyribosomes as crowder, relative nucleoid volume (V_nuc/V_cell) decreases not further than ~0.4 (Figure 2b in [6]). In addition, the simulation shows that when only proteins are present, a mixed state of the nucleoid is maintained (compare cell (a) in Figure 6). Phase separation (cell (b) in Figure 6) now only occurs at a relative high protein number when n_c/n_c0 > 1.4 (cf. Figure 2c in [6]). The very large effect of polyribosomes is contrasted by the observations of Cunha et al. [33] that small polyethylene glycol molecules (radius of gyration ~1.5 nm) are effective in compacting nucleoids isolated from E. coli.

In the theoretical approach of the Männik-group, polymer physical assumptions are made and input values are used that differ from the work of Odijk [5]. This complicates a comparison between the two approaches. For instance, minimalization of total free energy of the cell, is obtained by Chang et al. [6] using best-fit values for the parameters g and a of their Equation (2). By applying this equation as formulated by Cunha et al. [33], the free energy of DNA self-interactions, F_self, includes the crosslinking of the DNA in the nucleoid. As a result, the F_self has a somewhat lower value (0.58 x 10⁴ k_BT) than the value (1.9 x 10⁴ k_BT) obtained by Odijk [5] who considered the mere excluded volume interaction between the supercoiled Kuhn segments.

In the approach of Odijk [5], energy minimalization is obtained by equalizing the chemical potentials and osmotic pressures in nucleoid and cytoplasm (Figure 4d), as explained in Appendix A. Furthermore, Chang et al. [6] used different input values regarding supercoil persistence length (50 nm versus 80 nm in Odijk’s model), the supercoil contour length (493 µm versus 630 µm, considered by Odijk) and the number of supercoiled (Kuhn) segments (N_s = 6700 versus 4000 estimated by Odijk, [5]). In addition, different input values have been used for ribosomes (6000 versus 8000 considered by Odijk) and for the total number of soluble (non-ribosomal) proteins in cytoplasm and nucleoid (0.23 x 10⁶ versus 1.6 x 10⁶ considered by Odijk; see Table A2 in Appendix B); this 7-fold difference could have a great impact on the different conclusions. In the work of the Männik-group [6], the number for soluble proteins was again obtained by fitting their model to the experimental data (see Figure 1c in [6]), whereas Odijk’s value was obtained from the physiological measurements of Bremer and Dennis [43].

Männik and co-workers [6] also emphasized that “the experimental uncertainties for several parameters entering the expression of the total free energy are considerable.” This certainly holds for the volume of the nucleoid (see Table 1) and for the number of soluble proteins assumed to be present in slow growing E. coli cells (see Figure 7 below). These are the parameters that determine the density and thus the refractive index (RI) of nucleoid and cytoplasm, as discussed in Section 2.1. If phase separation is only the effect of the exclusion of polyribosomes and if soluble proteins would diffuse throughout the cell, what protein-concentration difference and thus RI difference between cytoplasm and nucleoid can be expected? To evaluate this question, a comparison is made in Figure 7 of cytoplasmic and nucleoid volumes and of numbers of ribosomal and soluble proteins between an average E. coli K-12 cell as described in [6] and of the re-measured E. coli B/r cells of Huls (2001), both presented in Table 1. A detailed description of the relevant assumptions and data is given in Table A2 of Appendix B. The comparison shows that for the cells in [6], a reduction of the RI-value in the nucleoid relative to the cytoplasm of 33% is obtained (Figure 7a). This is in agreement with the reduction observed in [27] as described in Section 2.1. For the cells of Huls (2001), the RI-reduction is only 22% (Figure 7b), which can be ascribed to the higher protein concentration assumed to be present in these cells. It suggests that given the assumptions made for this cell, additional depletion of soluble proteins from the nucleoid to a concentration of ~13 mg/ml and an increase of protein concentration (soluble and ribosomal) in the cytoplasm to~ 20 mg/ml is needed to obtain a RI-reduction in the nucleoid of about 34% (see Table A2 in Appendix B).

The total volume fraction of soluble proteins in the cells of [6] calculated from Figure 7a is 0.017; the protein volume fraction calculated for the cells of Huls (Figure 7b) is 0.101. Future estimates and measurements will have to show whether the 7-fold lower input value for soluble proteins as used in [6] is realistic.

3.2. Compaction through Transcriptional Activities

A different approach to DNA compaction was taken by the group of Jacobs-Wagner [8]. Their study is not based on free-energy minimization of steric self- and cross-interactions, but on the solvent quality of chromosomal DNA in the cytoplasm. As a starting point, they are considering the DNA to be a “random-coil polymer” (see Figure 1A of [8]), in which the compacted DNA chains will contact each other and form a network with cross-points depending on the DNA concentration and the quality of the cytoplasmic solvent, consisting of water, proteins, ribosomal subunits and polyribosomes. The distance between the cross-points, the so-called correlation length, ξ, determines the mesh size of the network. They calculated the solvent quality theoretically, using their Equation (1) in [8]. Important parameters for this equation are a Kuhn length for the double helix of 60 nm (rather than 90 nm as given in Figure 4a), a DNA concentration in the nucleoid of ~7 mg/ml, which is about 3x lower than the value obtained in other studies (see Table 1) and an average mesh size of 50 nm. This last value was determined experimentally using probes (GFP-µNS particles) varying in size from 50 to 150 nm. The observations showed that the apparent average mesh size in the nucleoid must be around 50 nm (see Figure 2C,F in [8]). With the above input values, the so-called Flory-exponent was calculated. This exponent is indicative for the solvent quality and gave a value of ν = 0.36. Such a low value (< 0.5) implies that the cytoplasm behaves as a poor solvent for DNA. Here, it should be emphasized that in contrast to the branched DNA superhelix models used by Odijk ([5]; see also Figure 5 in Wegner et al., [44]) and Männik and coworkers ([6]; see also Figure 3B in [34] and [45]), the group of Jacobs-Wagner only considered circular, non-supercoiled DNA for their calculations and poor-solvent simulations (see Figure 3 in [8]).

The results of their experiments on the spatial distribution of (poly)ribosomes and on the effect of transcription inhibition by rifampicin, causing expansion of the nucleoid (see below), made the authors suggest that RNAs in general cause an effective poor solvent quality for DNA in the cytoplasm. These results [8] agree with recent super-resolution and single-molecule fluorescence-microscope studies (review: [46]), suggesting that large proteins [47] and free ribosomal subunits (~20 nm diameter) are able to diffuse through the nucleoid [48,49,50]. These microscopic observations suggest that the nucleoid does not deplete soluble proteins and larger particles (up to 20 nm diameter) and thus contains a heterogeneous mixture of DNA and cytoplasmic components (as suggested by the poor-solvent simulations in Figure 3 of [8]).

In a recent study, a different mechanism of DNA compaction has been described by Bignaud et al. ([10]; see for reviews [18,51]). Based on high resolution chromosome-conformation-capture analyses (Hi-C), Bignaud and coworkers propose that folding of the chromosome is obtained by transcription-induced supercoiled regions in association with SMC proteins. So-called transcription insulating domains (TIDs), that tend to contact each other and cluster together, are suggested to fold the chromosome and to function as its primary building blocks (see Figure 4l in [10]). Their contact maps reveal the occurrence of a succession of short, transcription-induced, compact domains alternating with unstructured, highly-expressed gene regions, that inhibit supercoil diffusion. Previously, Lioy et al. [52] already identified some ~30 chromosome self-interacting domains (or CIDs), with an average size of 150 kb. As to be expected, the Hi-C patterns induced by transcriptional activity disappeared upon inhibition with rifampicin (Figure 1b in [10]).

It is evident that the inhibition of transcription initiation by rifampicin, causing the dissociation of polyribosomes and of 70S ribosomes and the degradation of mRNA, must have a profound effect on the transcriptional units as described by Bignaud et al. in their Figure 4l [10]. If these transcriptional units play a role in chromosome folding, rifampicin is expected to annihilate compaction of the nucleoid. Indeed, rifampicin has been found to change the appearance of the nucleoid in all microscopic studies compilated in Table 2. Because growth rate [6] and the fixation procedure [51] have been suggested to influence nucleoid appearance, they are mentioned in Table 2. In 7 of the 18 studies, a full dispersal of the nucleoid has been noted, all in fast-growing cells; in 2 of these studies cells had been fixed (see column 2 and 3 in Table 2). In the other studies, the changed appearance was described as a radial contraction combined with a longitudinal expansion, both in fast and slow growing cells. An explanation for this behavior was proposed by Mondal et al. [53] from computer simulation studies: polyribosomes are preferentially localized at the endcaps of the cell, where the nucleoid becomes compressed axially. When rifampicin causes their dissociation, the nucleoid will expand in the long axis, while the ribosomal 30S and 50S subunits now form a thicker layer along the cylindrical wall, compressing the nucleoid radially. Some studies [48,54,55] seem to confirm this explanation, which predicts that there is still depletion of ribosomes and proteins from the nucleoid as described in Section 2.2 by the depletion theory of Odijk [5].

Where authors suggest that the nucleoid becomes fully dispersed (indicated in the last column of Table 2 with “yes”), it seems plausible that dissociated ribosomal subunits and soluble proteins fully penetrate the nucleoid, annihilating any phase separation between cytoplasm and nucleoid ([6]; see Figure 7a). This would imply that phase contrast microscopy of rifampicin-treated cells immersed in a high refractive index medium (see Figure 1A–C), would show no region in the cell with a lower refractive index. Future microscopic studies will have to solve the disagreements as evident from the studies compiled in Table 2.

The heterogeneous nucleoid structure proposed by Xiang et al. [8] on the basis of poor solvent simulations and by Bignaud et al. [10], will have to comply with the density difference visualized by phase-contrast microscopy (Figure 1) and must also be understood in terms of the polymer-physical and thermodynamic rules of equal osmotic pressure and chemical potential in the two phases as mentioned above in Figure 4d. The same holds for the proposal of Bakshi et al. [61] that the dissociated 30S and 50S ribosomal subunits can mix with DNA causing nucleoid expansion.

4. Segregation and Movement of Chromosome Arms (Replichores)

4.1. Replichore Movement to Opposite Halves of the Nucleoid

As depicted in Figure 8a, the circular chromosome of E. coli can be divided in two chromosome arms, also called left (L) and right (R) replichores, that run from origin to terminus. Bi-directional replication from the origin results in a so-called replication bubble at initiation, in which two nascent pairs of replichores (red and blue), connected by an origin, represent the two daughter chromosome arms. These have to segregate to the two prospective daughter cells. When studying the necessary movement of daughter arms and their left and right replichores, the physical properties of nucleoid compaction described in the previous sections has to be taken into account. This implies properties like the permeability of the nucleoid to macromolecules involved in transcription and translation. As described by Kohiyama et al. [64] the replication bubble starts with a DnaA-based hyperstructure that integrates the many proteins involved in metabolic and regulatory pathways necessary for the initiation of replication. Here, however, only the structural aspects of the nascent chromosome arms and their behavior during segregation are considered. Does the segregation process take place in a heterogeneous region where small and larger protein complexes (polyribosomes) are mixed with the nucleoid as suggested in some fluorescence microscope studies [48,49,50]? Or does segregation occur in a relatively homogeneous nucleoid where small, soluble proteins are depleted to some extent from the nucleoid as proposed by Odijk ([5]; see also [34])?

Studying the paths of fluorescent loci in slow-growing E. coli cells by time-lapse experiments, the groups of Sherratt [12] and Hansen [13,65] have found that the E. coli chromosome is arranged with its left and right replichore lying separated in opposite halves of the nucleoid (Figure 8b). Their important observations and views (reviews [66,67,68] form the basis for the discussion below and as well for the speculative proposal of a passive segregation model in Section 4.2. Using the constructs of Flemming Hansen in which the origin and two loci on the left and right replichore were tagged with three different colors, the observations of Wang et al. [12] and of Nielsen et al. [13] could be confirmed by measuring the simultaneous movement of the three loci [63]

Figure 8. Schematic representation of the circular chromosome and of the localization of replichores in the nucleoid. (a) Left and right chromosome arms (replichores) with replicated origins (green circles) in the replication bubble. During the replication or C-period [69], the arms are replicated by the two replisomes (red and blue triangles), each synthesizing a leading strand and a lagging strand with Okazaki-fragments (short triangles). Several fluorescently labeled loci (red and blue stars) are indicated, that will show the position of the arms in the growing cells of Flemming Hansen (see below). The two daughter replichores (dashed arrows) will each end up in the prospective daughter cells with their replichores in two opposite halves of the nucleoid. The origin localizes at mid-cell between the left and right replichores that are connected by the terminus (black cross). (b) Position in a slow-growing E. coli cell (cf. Figure 5), with the two chromosome arms in the so-called transversal arrangement. Initiation of DNA replication at mid-cell has formed a replication bubble, which results in a replichore pattern indicated as Left-ori-ori-Right (L-O-O-R). The terminus occurs in a stretched connection between both replichores and migrates after duplication and cell division from the new pole to the cell center. (c) For comparison, a longitudinal arrangement of chromosome arms (cf. Caulobacter [16] or Agrobacterium [17]). Replication bubble is positioned at the stalked cell pole; the terminus at the other tip of the nucleoid [16].

Figure 8. Schematic representation of the circular chromosome and of the localization of replichores in the nucleoid. (a) Left and right chromosome arms (replichores) with replicated origins (green circles) in the replication bubble. During the replication or C-period [69], the arms are replicated by the two replisomes (red and blue triangles), each synthesizing a leading strand and a lagging strand with Okazaki-fragments (short triangles). Several fluorescently labeled loci (red and blue stars) are indicated, that will show the position of the arms in the growing cells of Flemming Hansen (see below). The two daughter replichores (dashed arrows) will each end up in the prospective daughter cells with their replichores in two opposite halves of the nucleoid. The origin localizes at mid-cell between the left and right replichores that are connected by the terminus (black cross). (b) Position in a slow-growing E. coli cell (cf. Figure 5), with the two chromosome arms in the so-called transversal arrangement. Initiation of DNA replication at mid-cell has formed a replication bubble, which results in a replichore pattern indicated as Left-ori-ori-Right (L-O-O-R). The terminus occurs in a stretched connection between both replichores and migrates after duplication and cell division from the new pole to the cell center. (c) For comparison, a longitudinal arrangement of chromosome arms (cf. Caulobacter [16] or Agrobacterium [17]). Replication bubble is positioned at the stalked cell pole; the terminus at the other tip of the nucleoid [16].

The measurements showed that in newborn cells that had not yet initiated replication, the three loci occurred in the typical pattern of L-O-R in 80% of the cells (Figure 9a). This high percentage can be interpreted to indicate that the two replichores had not mixed, but had ended up separated in the two halves of the nucleoid (see Figure 8b and Figure 9a). If the replichores would have intermingled the three loci would occur in three patterns (LOR; LRO; RLO) and the percentage of the L-O-R pattern would only be 33%. After initiation and origin duplication, the cells contained 4 spots showing the patterns L-O-O-R (44%) and O-L-O-R (46%). In the majority of 4-spot cells with increased length (L>2.3 µm), the origins have moved apart and passed either one (36%; O-L-O-R / O-R-O-L; Figure 9b) or both of the replichore loci (26%; O-L-R-O). This indicates that soon after duplication, one of the origins has passed an un-replicated locus on one of the replichores (see also Table A1 in [63]). Figure 9c shows a pattern in cells with 6 spots indicative for an almost fully replicated chromosome. As indicated by the schematic chromosome image in Figure 9c, the symmetric L-O-R-L-O-R arrangement is obtained when the leading strands move faster (through transcriptional activity?) than the lagging strands. This behavior is described by Mäkelä et al. (see Figure 5A in ref. [70]), who attribute it to the MukBEF-dependent binding of DnaN (β2-clamps) primarily to the lagging strands.

To obtain the various patterns in the transversal arrangement, replicating left and right arms have to pass each other once or twice. As will be explained below (Section 4.2, Figure 10 and Figure 11), the formation of these patterns and thus the mechanism of movement of loci or arms results from the development, already in the initial replication bubble, of replichore domains that do not become mixed. These separate domains within the nucleoid are assumed to enlarge and displace each other through continued DNA replication.

It should be emphasized here that, except for the origin (Figure 9b), the occurrence of three adjacent spots (L-L-O-O-R-R) was never observed. This can be interpreted to indicate that replichore loci separate almost immediately after their replication (see also discussion on cohesion in Section 4.3). This contrasts unpublished measurements of E. coli FH4035 cells (see construct of Flemming Hansen in ref. [63]), grown in LB-medium at 30°C (doubling time 66 min) and treated with 30 µg/ml nalidixic acid for 2-3 generations: the short filaments showed duplicated, colocalized spots, mostly separated in the short axis, suggesting that de novo DNA synthesis is required for segregation in the long axis of the cell.

Because it has been suggested that the active process of transertion [71,72] may play a role in the movement of segregating daughter strands, experiments were performed with cells treated with 300 µg/ml rifampicin that inhibits transcription and growth, but allows run-off DNA synthesis and residual divisions [63]. If replicated spots would not segregate in rifampicin-treated cells we would expect to observe a high percentage of cells with adjacent spots (L-L-O-O-R-R). However, as can be seen from Table 3, the replichore patterns found in non-growing cells are similar to those in the control cells. In view of (i) the significantly decreased percentage of cells with 4 or 5 spots, resulting from continued run-off replication (Table 3, column 4), (ii) the absence of cells with adjacent LL-OO-RR spots, and (iii) the decrease in percentage of cells showing the origin lying outside of the other loci (Figure 9b; see Table 2 in [63]) in both growing and non-growing cells, we conclude that chromosome movements in non-growing cells occur in a similar way as in growing cells: in both cases the two arms migrate to different halves of the nucleoid with the origin in between. This result contradicts a previous proposal that the process of transertion drives DNA segregation [71,72] and supports the hypothesis that segregation is passively driven by the process of de novo DNA synthesis. Ideas of a replication-driven segregation have been proposed previously by the groups of Grossman [73], Hansen [13,65], Sherratt [12], Austin [65,74], and Wiggins [75]. A direct link between DNA replication and chromosome organization has been demonstrated by the group of Sherratt [66] and will be discussed in the next section.

Wang and Sherratt [76] already presented evidence that inhibition of transcription with rifampicin did not affect segregation of the origin in E. coli, in agreement with the results presented in Table 3. The conclusion of these experiments [63] is that segregation continues during run-off DNA replication, but with a more random ordering of replichores (see columns 7-9 in Table 3). With respect to the distances measured between loci pairs (LL, OO, RR) in (unfixed) rifampicin-inhibited cells, it can be noted that these were about 0.3 µm smaller than in the growing control cells (Table 3 in [63]). This result can be ascribed to the smaller average length of rifampicin-treated cells (column 6 in Table 3), partly due to inhibited elongation and to residual divisions (see Appendix C, Figure A2).

Regarding the active process of transertion, it should be noted that the separation of daughter nucleoids after termination of replication, is dependent on cell growth (Figure A1; see also Figure 3 in [63]). Active processes like transertion and also cell constriction may thus be involved in stimulating this second step of the segregation process.

4.2. Four-Excluding-Arms Model for Segregation ⁽¹⁾

⁽¹⁾ Presented at the Lorentz workshop, “Escherichia coli DNA-segregation patterns in the presence and absence of growth.” Leiden, 18 June 2012.

It is tempting to assume that movement of replicated chromosome arms (Figure 8a) to opposite halves of the daughter nucleoids (Figure 8b), is only possible if the four newly synthesized DNA arms do not become mixed or entangled but are maintained as separate entities right from the start of the replication-segregation process. This demixed state could be achieved with the help of an active enzymatic mechanism like energy consuming (motor) proteins and topoisomerases. These proteins include, ATPases like topoisomerase IV and SMC complexes, like SMC-ScpAB in Bacillus subtilis and C. crescentus, the MukBEF complex in E. coli, and the MksBEF complex in a wide range of other bacterial species [77]. It is generally believed that E. coli DNA is organized by an axial core formed by MukBEF complexes, which promote individualization of chromosome arms and linear compaction of the chromosome through loop extrusion [19,78]. Sherratt and coworkers have also proposed that a linear MukBEF axial core could direct the symmetric (L-R-L-R) segregation of replichores (see Figure 9c) at the replication fork through differential binding of the β2-clamps to the lagging strands (see Figure 5A in [70]). Previously Jun and Mulder [79] already proposed that asymmetric constraints of leading and lagging strands may cause the formation of different patterns during their segregation.

A simpler, mechanistic explanation for the different segregation patterns of replichores is given by the “four-excluding-arms model”, as proposed here and illustrated in Figure 10. It is hypothesized that the chromosome arms containing the nascent strands synthesized by the four Pol III-replicases (indicated in Figure 10a), become separated through physical exclusion, because of topological and physiological differences between the leading and lagging double strands. As proposed in Figure 10b, the arms with the nicked lagging strands will have a random-coil structure because the ~20 Okazaki-fragments have first to become ligated (see Appendix C), whereas the leading arms can directly become supercoiled and transcribed. It can be assumed that, once the four nascent arms in the initial replication bubble have formed separate blobs, their physical entanglement becomes unlikely, because the mixing of such blobs is energetically unfavorable. Their demixed state in the parental DNA network could represent a minimal energy situation, just as the phase separation between DNA and cytoplasm as described in Section 2.2 represents a minimal energy situation. Future considerations by polymer physicists should reveal whether such a de-mixed state between newly synthesized and parental DNA is feasible. (see discussion in [80] and supplementary information in [35]).

It is proposed that the four nascent replichore arms continue to remain separated and to form separate micro-domains (Figure 10c), that enlarge and displace each other. This exclusion of chromosome arms may be helped by the low DNA diffusion coëfficient as determined both in liberated nucleoids [33] and in living cells [81]. Because mixing or entanglement of the newly synthesized replichores is prevented right from the beginning, the process of entropic de-mixing [79] at a later stage may not be necessary.

An immediate separation of the origins in the replication bubble is to be expected if the replicated DNA in the origin region increases in mass and is free to move, while the two replisomes remain tethered to the un-replicated parental DNA which they are reeling-in. Such tethering of the replisomes would force the duplicated origins to move apart as indicated by the double black arrow in Figure 10b. It can be envisaged that once the initial blobs have been established, they develop into larger, separate domains that are fed by de novo DNA synthesis (Figure 10c). This causes the disappearance of the initial replication bubble, as visualized in Figure 10b, which could represent a structure with replisomes that differ from the ongoing ones (Figure 10c). Intriguingly, such a mechanistic distinction may have been referred to by Khodursky et al. [82] in their discussion of the effects of inhibition of initiating replication forks on thymineless death (TLD).

While the 4 micro-domains enlarge through de novo DNA synthesis within the meshwork of un-replicated, parental DNA, they are envisaged to first form a tetrahedron in the replication bubble (Figure 10c and Figure 11a). Further enlargement of the domains by the two replisomes will force the domains to displace each other and to rearrange in the long axis of the narrow nucleoid (see Appendix C). Such rearrangements will cause different patterns in cells with 6 spots as depicted in Figure 9c. The domain rearrangements could be related to large-scale structural changes and rapid movement of certain loci (“snaps”) as described by Joshi et al. [84]. As previously proposed by Sherratt and coworkers [66], transient pausing of one replisome, causing a lower velocity of DNA synthesis of one replisome and thus of domain expansion, may cause loci in the smaller domains to be pushed to mid-cell by the faster expanding domains synthesized by the other replisome. This results in the patterns like R-O-L L-O-R or L-O-R R-O-L, as indicated in Figure 11(b2).

In Figure 11c the transition to a faster growth rate (nutritional shift-up) is considered. Because the cells become wider [85], there is no need for the enlarging microdomains to adopt the different side-by-side patterns. Instead, the tetrahedron conformation of microdomains is maintained as previously described for spherical, thymine-limited cells by Zaritsky et al. [86]. In addition, because in the wider cells no re-arrangements of domains are necessary, there is no need for the origins to pass the replicating chromosome arms (see Figure 9b). As a result, the origins keep moving apart at the tip of the developing nucleoid towards the cell poles, even upon re-initiation, resulting in a longitudinal arrangement of the chromosome arms as described in [74].

Could this proposal of the formation of 4 excluding and expanding domains starting in the initial replication bubble (Figure 10a) be confirmed by the chromosome-conformation-capture technique or Hi-C [9], for instance, by increased interactions within the 4 domains? This is not to be expected as long as the contacts occur as intra-arm interactions on the same replichore, already visible as the primary diagonal on the Hi-C contact map (see for E. coli Figure 1A in Lioy et al. [52]). Only if the domains of nascent right and left replichore arms were to intermingle, inter-arm interactions would be visualized as a secondary diagonal on the Hi-C contact map (see for C. crescentus Figure 3Ad in [77]). Absence of such interactions in published E. coli Hi-C maps, is in agreement with the “four-excluding-arms model” here proposed.

The organization of the chromosome as proposed by the four-excluding-arms model is compatible with the higher-order level of organization of the chromosome into 4 macrodomains (Ter, Ori, Right and Left) and 2 non-structured regions as described by [87]; see also [52]). As soon as the left and right replichores have become replicated and the four excluding micro-domains (Figure 10b), have developed into four macro-domains these will develop in the duplicated structured and unstructured macro-domains as described in [87] (see also [9]).

For the transversal arrangement of the chromosome as depicted in Figure 11a, the asymmetric deposition of newly synthesized DNA requires stretched regions of replicated DNA that feed the newly developing nucleoids (see Figure 4B panel 3 and 4 in [63]. These “feeding strands” can be expected to induce superdiffusive motions of loci, which have occasionally been observed [81,88].

Is the four-excluding-arms model able to explain the distances between replicated spots as documented for the slow growing E. coli K-12 cells in [63]? In Appendix C the volumes of the developing domains have been calculated for the slow growing E. coli cells depicted in Figure 11 a and b. Although the calculations apply to a different strain (E. coli B/r), the results show that the distances obtained by mere de novo DNA synthesis (1.8 µm in Figure A1, cell b-4), are similar to those measured in the E. coli K-12 strain in [63] (1.4 - 1.6 µm in Figure C2 a, b).

4.3. Comparison between Bacteria and Eukaryotic Cells and the Phenomenon of Cohesion

A fundamental difference between bacteria and eukaryotes lies in the pairing and alignment of sister chromatids throughout G2-phase, until the chromatids become condensed in the metaphase of mitosis [92]. During this cohesion-period the eukaryotic sister chromatids are held together by ring-shaped cohesin protein complexes, thus overcoming the gap between the time when the genome is replicated and when the chromatids are physically separated. However, already during S-phase, sister chromatids become segregated through the loop-extrusion activities of condensin II complexes that prepare the chromosomes for their further condensation by condensin I [93]. Subsequently, the chromatids become separated by proteolysis of cohesins and transported by microtubules of the mitotic spindle, processes not occurring in bacteria. Hirano ([94]) emphasized that “sister chromatids in eukaryotes are already well resolved by metaphase before they are subjected to poleward movement in anaphase…” and that it is therefore reasonable to hypothesize that the resolution process in eukaryotes is mechanistically equivalent to the segregation process in bacteria (see also [80]).

Contrary to the proposition of Bates and Kleckner [39], it has been argued that there is no trustful experimental evidence for a eukaryotic cohesion phenomenon in bacteria [12,95] and that the replicated DNA strands can immediately be separated if no precatenanes have been formed behind the forks that cause interstrand entanglements [96]. Estimation of the cohesion time is generally based on the independent measurements of, for instance, the copy number of the gln locus in comparison with its appearance as fluorescent foci in the cells [84]. See for discussions of “cohesion-experiments” in E. coli, [97,98,99].

While there seems to be no functional necessity for initial linking of bacterial daughter strands by cohesion, analogously to cohesion in eukaryotes, the passive four-excluding-arms model presented here suggests an immediate separation of nascent DNA strands in the replication bubble. However, active mechanisms for bringing strands transiently together on some distance from the replisome for processes like recombination and mismatch repair can be envisaged. Such an organization, involving SeqA-complexes on a distance of ~250 nm from the replisome (cf. distances in Figure A1 and Table A3), has been described by the group of Skarstad [100].

5. Conclusion

Various microscopic studies of bacterial DNA have demonstrated that the nucleoid represents a low refractive index region that is well separated from the cytoplasm (Figure 1, Figure 2 and Figure 3). Polymer-physical and biochemical studies, including chromosome-conformation-capture techniques (Hi-C), have tried to understand this phase separation, either considering the free energy of DNA-crowder interactions [5,6] or chromosome folding by transcriptional activities [10]. An important tool in both approaches is the effect on nucleoid compaction after inhibiting transcription with rifampicin (cf. Table 2). Better microscopy studies of living cells are required to evaluate and understand the volume expansion of the nucleoid in rifampicin-inhibited cells. Also, it is not always clear whether the different proposals for DNA compaction conform to the observation of a low refractive-index nucleoid. A final conclusion has to await better agreement on the biological input values of the number of soluble proteins in an E. coli cell and better estimates of the volume of the nucleoid (see Figure 5). Regarding the polymer-physical approaches to minimalize the total free energy of the cells, it would be helpful if they would be accompanied by a more educational explanation for biologists (see [34,35]), to understand the sources of input values and the complicated computations.

Studies on bacterial DNA segregation indicate two different views: Either, resolution and movement of replicated daughter strands is performed by a dedicated, active process based on DNA loop extrusions through structural maintenance of chromosome (SMC) complexes, or by the passive process of de novo DNA synthesis, as described here. If, in the initial replication bubble (Figure 10b), intermingling of the newly synthesized DNA strands would occur, it is to be expected that the entanglements could only be resolved with an elaborate mechanism of topoisomerases and SMC proteins [19]. However, the different physical properties of nascent leading and lagging chromosome arms (Figure 10a), together with different gene expression activities between the two replichores, could prevent the mixing of the 4 daughter strands right from the beginning. In that case, the secret of segregation lies in the build-up of the replication bubble: if no initial mixing occurs due to their different physical properties, the 4 chromosome arms will exclude each other and will become confined in 4 individual domains without the necessity of de-mixing (Figure 10b).

So far, microscopic observations have not given any indication for the existence of the 4 excluding domains. Further developments of techniques for pulse labelling of nascent DNA, like performed by Spahn et al. [48] seem promising if they could be applied to slow-growing E. coli cells. As already described in Small Things Considered (https://schaechter.asmblog.org/schaechter/2019/11/forbearance-with-the-escherichia-coli-nucleoid.html), the Dekker group studied the dynamics of a ring-shaped nucleoid, “opened up” and replicating in artificially widened cells. The studies show how in living cells duplicated origin-spots are often positioned in or near low-density, blob-like DNA domains [101]. In another study by this group [102], it is shown how, upon initiation of DNA replication, the duplicated origins first move apart in random directions, but re-orient towards the long axis after some time when the amount of replicated DNA has locally increased. These observations also reflect what may happen during the hypothetical transition from slow to fast growth as depicted in Figure 11c,d.

Further developments in spatial light interference microscopy [103], or digital holographic microscopy combined with optical diffraction tomography [26] and improved labeling techniques for nascent DNA strands [48], will be necessary to evaluate the above hypothesis of the four-excluding-arms created in the initial replication bubble. When a more detailed quantification of the number of proteins involved in the replication bubble will become available, calculations of the free energy state of the proposed 4 nascent arms, as performed by Odijk for the whole nucleoid (Figure 5 in [44]), for the construction of the DNA replication cycle during shift-up could become possible. Such calculations might support the above proposal of passive DNA segregation in the form of the hypothetical 4 excluding domains (Figure 10b) that gradually replace the parental nucleoid.