Abstract
Our knowledge of bacterial nucleoids originates mostly from studies of rod- or crescent-shaped bacteria. Here, we reveal that Deinococcus radiodurans, a relatively large, spherical bacterium, possessing a multipartite genome, and well-known for its radioresistance, constitutes a valuable system for the study of nucleoids in cocci. Using advanced microscopy, we show that as D. radiodurans progresses through its cell cycle, it undergoes coordinated morphological changes at both the cellular and nucleoid level. D. radiodurans nucleoids were found to be highly condensed, but also surprisingly dynamic, adopting multiple distinct configurations and presenting a novel chromosomal arrangement in which oriC loci are radially distributed around clustered ter sites maintained at the centre of cells. Single-molecule and ensemble studies of the abundant histone-like HU protein suggest that its loose binding to DNA may contribute to this remarkable plasticity. These findings clearly demonstrate that nucleoid organization is complex and tightly coupled to cell cycle progression.
Introduction
In all organisms, genomic DNA is compacted several orders of magnitude and yet must remain accessible for essential DNA-related processes including DNA replication, repair and transcription. In bacteria, packaging of genomic DNA is achieved by several mechanisms, including DNA supercoiling 1 and DNA compaction by nucleoid-associated proteins (NAPs), such as HU 2–4. However, recent studies now indicate that additional factors, such as molecular crowding and depletion forces, also play important roles in determining the volume of the cell occupied by the nucleoid 5–7, and suggest that cell shape and size may critically influence nucleoid organization 7,8.
So far, a vast majority of the studies of bacterial nucleoids have focused on three model bacteria, E. coli, B. subtilis and C. crescentus, all of which are either rod- or crescent-shaped bacteria. Recent developments in single-molecule and genome-wide analytical techniques have started to shed light on how these model bacteria organize, compact and segregate their chromosomes, three processes that are intimately connected. Genetic, microscopy and chromosome conformation capture studies have revealed that nucleoids are organized into micro- and macrodomains 9, the origins of which are still unclear (reviewed in 10 and have been shown to adopt an overall helical arrangement 11–13, which is distributed along the longitudinal axis filling a large fraction of the cell volume (~70% in E. coli). Mapping of specific chromosome loci such as the origin (oriC) or the termination sites (ter) have also revealed that chromosomes in model bacteria can be seen to adopt two stereotypical configurations. Under certain conditions, they may arrange longitudinally, along an ori-ter axis with the oriC and ter regions located at opposite poles of the cell, or transversally in which both the oriC and ter are localized at mid-cell and the left and right arms are located on either sides 2,13–15.
Although in all bacteria irrespective of their shapes, chromosome organization and segregation are tightly coupled to the division process, in cocci, the directionality of chromosome segregation has additionally been proposed to play a key role in positioning the division site 16. The two most studied cocci are the ovoid Streptococcus pneumoniae and the spherical Staphylococcus aureus, two major human pathogens. So far, however, the small size of these cocci has largely restricted the study of nucleoid rearrangements and concomitant morphological changes occurring during their cell cycle. In contrast, Deinococcus radiodurans is a non-pathogenic, relatively large (~2 µm in diameter) coccus, displaying an outstanding resistance to DNA-damaging agents including ionizing radiation, UV light and desiccation 17–19. Several factors have been proposed to contribute to this outstanding phenotype: (i) a highly efficient and redundant DNA repair machinery, (ii) the presence of numerous antioxidant metabolites 18,20–22 that contributes to an increased protection of the proteome, and (iii) the unusual properties of its genome. D. radiodurans has indeed been shown to possess a complex, multipartite genome composed of four replicons each of which are present in multiple copies ranging from 4 to 10 copies, depending on its growth phase 23–25. Moreover, D. radiodurans’ genome has been reported to be more condensed 26 than that of radio-sensitive bacteria, such as E. coli, and to adopt an unusual ring-like structure 27,28. These unusual features have been suggested to facilitate genome maintenance and repair 29,30.
Using both spinning-disk time-lapse microscopy and super-resolution imaging, we have performed a detailed analysis of the morphological changes that occur at the cellular and nucleoid level as D. radiodurans grows and divides in alternating perpendicular planes. These data reveal that D. radiodurans nucleoids are indeed highly condensed, while remaining surprisingly dynamic, adopting multiple distinct configurations as the bacterium progresses through its cell cycle. Studies of the dynamics of the highly abundant histone-like HU protein, which is the major NAP that coats the genomic DNA in D. radiodurans 31 reveal that it only associates loosely with DNA, thereby facilitating the structural rearrangements of the nucleoid. Finally, we followed the choreography of the oriC and ter loci of chromosome 1 during the various stages of the cell cycle and show that they exhibit very different distributions within the cell with the oriC loci being radially distributed around the centrally located ter sites. Taken together, these findings demonstrate that the properties of D. radiodurans make it particularly well suited for the study of nucleoid organization in cocci, and provide new, compelling evidence indicating that bacterial nucleoids are complex and dynamic entities that are tightly coupled to cell shape, cell cycle progression and septal growth.
Results
Morphological changes occurring during the growth of D. radiodurans
To follow the morphological changes occurring during the growth of D. radiodurans, cells were labelled with the membrane dye Nile Red and deposited on an agarose pad for either time-lapse imaging on a spinning-disk confocal microscope or super-resolution imaging using Point Accumulation for Imaging in Nanoscale Topography (PAINT) microscopy (see Methods and Figure S1 for a comparison of the two imaging systems). All cell measurements presented below were extracted from super-resolution images (Figure 1A and Figure S2), while the order and duration of the various stages of the cell cycle were derived from spinning-disk microscopy images and videos.
Exponentially growing D. radiodurans cells cultured in rich medium form diads 32 that transiently form tetrads before the start of a new cell cycle (Figure 1B). D. radiodurans cells divide sequentially in two perpendicular planes 33,34. Starting from an elliptical and largely symmetric diad (Phase 1), the cells initiate their cell cycle by a phase of cell growth (Phase 2) leading to a slight invagination at the junctions between the central septum and the cell periphery (Figure 1B and Figure S2 and S3). This phase ends with the appearance of bright foci corresponding to the start of the growth of the new septa (Phase 3). These foci appear in the centre of the central septum and in the middle of the opposite peripheral cell wall. We observed that one foci may appear before the other in a given phase 3 cell and septal growth in the two cells forming the diad were frequently seen to be asynchronous (eg. cells with asterisk in Figure 1A). Phase 4 corresponds to cells in which the growth of the new septa is partial and in Phase 5, the newly formed septum is about to close. Finally, in Phase 6 septal growth is complete and tetrads are formed (Figure 1B). As the cells progress through Phases 3 to 6, the invagination at the diad junction becomes more pronounced and the length of the central septum progressively decreases (Figure S3) until disappearing completely as the tetrads fall apart to form two diads (Figure 1B and C). In stationary phase cells, the tetrads no longer separate into diads and the number of Phase 6 cells increases markedly (Figure S4). Moreover, a substantial fraction of these tetrads engages in a new division cycle, but cell cycle progression is very slow and only a small number of octads are seen (Figure S4).
Various cell parameters were extracted from our images (Figure S2 and S3) and were used to determine the average shape, size and volume of cells at each phase of the cell cycle (Figure 1C and D). D. radiodurans cells were found to grow linearly throughout their cell cycle (Figure S5), as seen by the progressive increase in cell volume starting from a mean volume of 1.87 ± 0.18 µm3 in Phase 1 until a volume of 3.60 ± 0.34 µm3 in Phase 5 just before cytokinesis (Figure 1D). The change in cell volume was also accompanied by a change in ellipticity of the cells as they progress through the cell cycle (Figure S3). Starting from hemispheres in Phase 1, individual cells become quasi ellipsoids by the end of Phase 5. In contrast, in stationary phase, cell growth arrests and the mean volume of cells in tetrads decreases to 1.65 ± 0.34 µm3 (Figure S4).
To assess the duration of each growth phase in exponential cells, two approaches were used. Durations were either deduced from the distribution of phases in the population of exponentially growing cells when deposited on an agarose pad and directly observed under the microscope (the fraction of cells in each phase reflects the relative duration of these phases, Figure 1E interior annulus), or by time-lapse imaging of cells (Figure 1E exterior annulus and Supplementary Movie S1). Very similar results were obtained for the two approaches and the total cell cycle duration measured in our time-lapse experiments (134 ± 11 min) was close to the doubling time of 130.1 min derived from our growth curves performed on D. radiodurans cells grown in liquid medium in a shaking incubator (Figure S5). Cells spend 91% of their time as diads and 9% as tetrads, and septal growth (Phases 3, 4 and 5) occurs during more than half of the cell cycle. Phases 1, 5 and 6 are the shortest, lasting between 11 and 15 min, while Phases 2, 3 and 4 are longer and last 25 to 35 min (Figure 1E).
Cell growth occurs in both peripheral and septal cell walls
In bacteria, cell growth is associated with expansion of the cell wall and peptidoglycan (PG) synthesis and remodelling. We thus investigated whether growth of the cell wall occurs (i) in the septal regions only, (ii) throughout the septal and peripheral cell walls, or (iii) only in the peripheral cell walls of D. radiodurans (Figure 2A). For this, cells were pulse labelled with two dyes: BODIPY-FL 3-amino-D-alanine (BADA), a green dye that is stably incorporated into the pentapeptide chain during PG synthesis in the peripheral cell walls only and does not reorganize during the cell cycle, and Nile Red that diffuses into lipid membranes and thus labels the whole cell wall, including newly synthesized septa during the cell cycle. The samples were then washed to remove any unbound dye prior to imaging. The use of two lasers prevented D. radiodurans from growing under the microscope, so cells were observed shortly after the labelling procedure (T=0) and then again after a complete cell cycle (T=130 min). The mode of cell wall growth could be deduced from such images by comparing the lengths of green- and red only-labelled cell walls per cell (Figure 2A) at T=0 with that observed after a complete cell cycle (T=130 min). If growth is restricted to the septal region (model i), which is not labelled with the BADA dye at T=0, the length of the green-labelled cell wall should remain constant after one cell cycle; if growth occurs throughout all cell walls (model ii), the lengths of both the green- and red-labelled cell walls should increase, and if growth is restricted to the outer periphery of the cells (model iii), then the red only-labelled cell wall should remain constant (Figure 2A). For ease of analysis, we performed our measurements on Phase 1 cells that displayed complete BADA (green) labelling of their peripheral cell wall, but only Nile Red staining of their central septum at T=0 (Figure 2B) and could thus be compared to our proposed models presented in Figure 2A. After a complete cell cycle, Phase 1 cells clearly displayed an altered labelling pattern (Figure 2B and 2C).
Measurements revealed that the lengths of both green- and red-labelled regions had significantly increased (Figure 2D), thus indicating that cell wall growth occurs throughout the entire cell surface of D. radiodurans cells including septal regions (Figure 2A), as reported for S. aureus (Monteiro et al, 2015).
Next, we evaluated the position and angle of the newly growing septa relative to the central septum in exponentially growing cells. In all cells, septal growth was found to occur precisely in the middle (50% ± 1.5%) at a 90° ± 7° angle to the central septum (Figure S6). Early electron micrographs of D. radiodurans cells suggested that septal closure occurs through a closing door mechanism and not as a diaphragm 33. Three-dimensional imaging of Nile Red stained cells confirmed this (Figure S6). The closing septum grows from both sides of the cells leaving a gap that stretches across the whole cell.
Nucleoid organization during the cell cycle
To follow the three-dimensional morphology of D. radiodurans nucleoids as a function of their cell cycle in exponentially growing and stationary phase bacteria, we used spinning-disk confocal microscopy on live cells. The nucleoids were visualized either by using a genetically modified strain of D. radiodurans in which the gene encoding the highly abundant HU protein was endogenously fused to mCherry (Table 1), or by staining the DNA with the Syto9 dye (Figure 3). The high extent of colocalization observed when double staining D. radiodurans nucleoids with both HU-mCherry and Syto9 (Figure S7) confirmed that HU largely coats the genomic DNA. Importantly, the same nucleoid structures (described below), including the previously described ring-like or toroid structure, were observed irrespective of the mode of labelling, and all four replicons of D. radiodurans assembled into a single nucleoid entity. When using Syto9, cell membranes could additionally be stained with Nile Red in order to distinguish the different phases of the cell cycle (Figure 3).
Nucleoid conformations were assessed in Syto9 and Nile Red stained wild-type D. radiodurans cells and representative shapes of nucleoids were retrieved for each phase of the cell cycle (Figure 3). In addition, time-lapse three-dimensional imaging of either Syto9 stained nucleoids (Movie S2) or HU-mCherry expressing D. radiodurans (Movie S3) was performed during a complete cell cycle to determine the chronological order in which these different conformations appeared. These videos clearly highlight the dynamic nature of the nucleoids. In Phases 1 and 2, a majority of nucleoids (respectively 52% and 42%) adopted toroidal shapes with an average diameter of 0.95 ± 0.10 µm (Figure S2). Additionally, in Phase 1, a substantial fraction (28%) of nucleoids were found to form condensed, undefined structures, which were seen to be short-lived (minutes) in our time-lapses and rapidly became toroids, which appeared to be more stable and long-lasting. Due to the reduced resolution of the spinning-disk images in the Z axis, we cannot rule out that these undefined structures may correspond to top views of toroids with an interior ring size below the resolution limit of the images. In Phase 2, 16% of nucleoids presented a square configuration with bright vertices, and a significant fraction (21%) also adopted a more open configuration (open ring or crescent). In Phase 3, these two nucleoid configurations were found to be the most prevalent (respectively 25 and 30%) and the crescent-shaped nucleoids were positioned with their convex sides facing the central septum of the diad (Figure 3A). The fraction of toroidal-shaped nucleoids decreased significantly in Phase 3 and instead they were seen to open up to form crescents. In some cases, an intermediary step was observed in which the toroids became squares with 3 or 4 of their vertices exhibiting strong fluorescent foci before opening up to adopt a crescent shape. In Phase 4 cells, most of these configurations disappeared and instead a large proportion of nucleoids were now elongated rods (54%) with an average length of 1.48 ± 0.25 µm and an apparent full width at half maximum of 0.52 ± 0.08 µm (Figure S2). More complex branched structures (35%) were also observed in Phase 4. Remarkably, these structures were all positioned with their long axis perpendicular to the future division axis (Figure 3A). Time-lapse imaging revealed that these elongated structures resulted from the opening and stretching of the crescent-shaped nucleoids seen in Phase 3. In Phase 5 cells, in which the formation of the newly dividing septa was nearly complete, nucleoids appeared mostly as branched, elongated structures (35%), as seen in Phase 4, or as double rings connected by a thin thread (42%), in which each ring was positioned in one of the two future daughter cells. A clear constriction was observed around the center of these nucleoids resulting from the two closing septal segments (Figure 3A). Time-lapse imaging revealed that the rod-like structures predominantly found in Phase 4 progressively formed more complex, branched structures that eventually formed double crescents or rings connected by short stretches of DNA. Finally, in Phase 6 cells, chromosome segregation was complete and the newly formed nucleoids presented similar configurations as those seen in Phase 1 cells, i.e. condensed, undefined structures (45%) or toroids (45%). These findings are summarized in a simulated movie that illustrates the principal morphological changes of the nucleoids observed in exponentially growing D. radiodurans cells as they progress through their cell cycle (Figure 3B and Movie S4). Interestingly, as observed previously when characterizing the different phases of the cell cycle, the two cells composing a diad were not necessarily synchronized in terms of nucleoid morphology. However, it was clear that the morphological changes at the nucleoid level were highly coordinated with cell cycle progression and septal growth (Figure 3B).
Nucleoid compaction
Three-dimensional images of Syto9-and Nile Red-stained D. radiodurans were also used to determine the volume occupied by the nucleoid in D. radiodurans as a function of its cell cycle (Figure 4A; Methods). For comparison, the same procedure was used for analysis of E. coli cells (Table 1 and Figure S7). Unlike the cell volume (Figure 1C), the nucleoid volume did not increase linearly throughout the cell cycle (Figure 4A). Instead it was found to remain constant (~0.7 µm3) during Phases 1 to 3 and then to increase in Phases 4 and 5 to reach a maximum volume of ~1.2 µm3 just before cytokinesis. The newly formed nucleoids in Phase 6 exhibited the smallest volumes (~0.6 µm3). As a result, the fraction of the cell occupied by the nucleoid varied as a function of the cell cycle from a minimal value of 0.30 ± 0.06 observed in Phase 4 to a maximal value of 0.36 ± 0.06 detected in Phase 1 cells (Figure 4B). In comparison, the mean cell and nucleoid volumes of E. coli cells were found to be 1.33 µm3 and 0.83 µm3 respectively, corresponding to a cell fraction occupied by the nucleoid of 0.65 (Figure S7), in agreement with previously reported values 11. These quantitative measurements thus confirm that D. radiodurans possesses a highly condensed nucleoid occupying approximately one third of the cell volume, but also reveal that its level of compaction varies during its cell cycle. We also determined the mean volume of Syto9-stained nucleoids in D. radiodurans expressing HU-mCherry, and found that these were slightly higher than that of wild-type cells (Figure S7), perhaps as a consequence of steric hindrance caused by the mCherry fusion protein. Exponentially growing D. radiodurans cells have been reported to possess four to ten copies of their 3.2 Mbp genome 25, so the DNA density in an average D. radiodurans nucleoid (~0.7 µm3) should be in the range of 18 to 46 Mbp/µm3. In E. coli, which possesses one to two copies of a 4.6 Mbp genome, the DNA density in its nucleoid should be between 5.5 and 11 Mbp/µm3, corresponding to a density at least 3 times lower than that observed in D. radiodurans cells.
Finally, Syto9-and Nile Red-stained stationary phase D. radiodurans tetrads were also analyzed to determine the level of nucleoid compaction in such cells (Figure S4). The average fraction of the cell occupied by the nucleoid was found to be 0.19, which is significantly lower than the highest level of DNA compaction measured in exponentially growing Phase 4 cells (Figure 4B).
Nucleoid and HU dynamics
To further characterize D. radiodurans nucleoids in exponentially growing bacteria, we used super-resolution photoactivated localization microscopy (PALM) to image a D. radiodurans strain in which the most abundant NAP, HU, was fused to PAmCherry (Table 1). PALM experiments were performed exclusively on live bacteria because fixation protocols strongly affected the nucleoid organization (loss of diverse shapes and decondensation of the DNA). Several of the nucleoid morphologies described above (eg. crescent- and rod-shaped nucleoids) could be seen in the reconstructed PALM images, but most of the distinguishing features and details were lost (Figure 5A). Although, PALM provides higher spatial resolution, acquiring a complete dataset in PALM microscopy takes several minutes on our set-up. The loss of details in our PALM images thus most likely reflects the dynamic nature of the nucleoids in the second to minute timescale, as confirmed by a spinning-disk microscopy time-lapse experiment in which images of HU-mCherry were acquired in live D. radiodurans with short interval times (Movie S5). This time-lapse video reveals that even though most of the nucleoid conformations last for several minutes (Figure 3 and Movies S2 and S3), the nucleoids are mobile and can be seen to rotate and reorganize in the second timescale.
HU has been shown to play a major role in the compaction and organization of the nucleoid in D. radiodurans 25,31 and to be essential for viability 35. We therefore wondered whether HU may contribute to the remarkable plasticity of D. radiodurans nucleoids. Both single-molecule and ensemble measurements were carried out to probe the dynamics of HU proteins in live, exponentially growing D. radiodurans. Single-particle tracking PALM (sptPALM) experiments were performed on the HU-PAmCherry expressing D. radiodurans strain in which the trajectories of individual HU-PAmCherry molecules were reconstituted (Figure 5B). Cumulative probability distribution (CPD) analysis 36 of 1000 individual tracks extracted from several datasets revealed that only a single population of molecules could be distinguished, which displayed a confined diffusion (with a confinement radius of ~400 nm; Figure S8) with a mean apparent diffusion coefficient of ~0.32 µm².s−1. This value, which is significantly lower than that obtained for freely diffusing fluorescent proteins in bacteria (6-9 µm².s−1; 37), but also higher than that of immobile, DNA-binding proteins (D*<0.2 µm².s−1 38,39), clearly indicates that HU is not tightly associated with the genomic DNA. Ensemble measurements of HU mobility probed by fluorescence recovery after photobleaching (FRAP) on HU-mCherry expressing D. radiodurans cells confirmed these findings and revealed that 95% of HU molecules were mobile exhibiting a half-life of 0.75 s (Figure 5C). These data are consistent with a FRAP study of the E. coli NAP, H-NS 37 and suggest that even though HU coats the DNA, it most likely only associates transiently with DNA, thereby allowing the nucleoid to rapidly reorganize as the cells progress through their cell cycle.
Distribution and choreography of oriC and ter sites during the cell cycle
To follow the choreography of the oriC and ter loci during the cell cycle of exponentially growing D. radiodurans, we used engineered D. radiodurans strains, GY15787 and GY15800, in which a heterologous ParABS system is used to fluorescently label respectively the oriC and ter loci of chromosome 1 (Table 1), allowing them to be detected as bright foci within the cells 25. Three colour 3D spinning-disk images were recorded on such cells in order to visualize GFP-labelled oriC or ter loci, hydroxycoumarin-amino-D-alanine (HADA)-stained cell walls and HU-mCherry labelled nucleoids (Figures 6A and D). When analyzing these images, we chose to restrict our analysis to the bright ParB-labelled foci that could unambiguously be discriminated from the background fluorescence. This most likely explains why a reduced number of foci were counted per cell compared to the original study 25.
Throughout the cell cycle, the number of ter foci in cells was systematically lower than that of oriC foci (Figure 6A), as reported previously 25. The number of oriC foci was found to double as the cells progress from Phase 1 to Phase 5, going from a mean number of 2.25 foci/cell to 4.44 foci/cell in Phase 5 (Figure 6B). After septal closure, in Phase 6, the number of oriC foci was almost divided by 2, reaching 2.38 foci/cell. This suggests that DNA replication occurs all along the cell cycle until all copies of oriC from chromosome 1 have been duplicated. In contrast, the number of ter foci remains much lower around 1.15 foci/cell and constant for most of the cell cycle until Phase 5 where it increases to reach 1.5 foci/cell.
We also investigated the localization and distribution of the two loci within the cells (Figure 6C). The ter foci were found to be specifically located to the central region of cells during most of the cell cycle. It was only in Phases 1 and 2 that the ter foci were found to be slightly off-centre. As the cells progress from Phase 1 to 3, the ter loci progressively move away from the central septum to relocate in the centre of the growing cells. In all phases, however, a majority of ter foci were located along the new division axis and 47% of them were still localized in this region at late stages of septal closure in Phase 5, just before cytokinesis, indicating a very late segregation of ter. In contrast, the oriC foci displayed a very different cellular distribution and could be seen throughout the cell cytoplasm with only a minor fraction localized along the new division axis and in the centre of the cells. Multiple oriC foci were thus seen to be radially distributed around a single centrally located ter foci for most of the cell cycle. This was particularly visible in Phases 1 and 6. Interestingly, both ter and oriC seemed to be excluded from the peripheral region of the cell bordering the cell wall.
No clear correlation could be established between foci localization and nucleoid morphology. However, particular configurations could be seen (Figure 6D). ter foci could for example be observed occasionally in the centre of the toroid- and square-shaped nucleoids, i.e. in regions with low DNA density, devoid of HU-mCherry labelling. In the late stage of nucleoid segregation (Phase 5), ter foci were also seen to localize on the thin thread connecting the partially segregated nucleoids (double rings), close to the new division axis and the closing septum. oriC, in contrast, was often seen at the periphery of the nucleoids, and occasionally at the outer tips of the branched nucleoids observed in Phase 4 cells.
Discussion
In this work, we have characterized the cell cycles of exponentially and stationary growing D. radiodurans, enabling us to define 8 distinct phases, and have followed the changes in the cell and nucleoid morphology occurring during these processes.
We determined that exponentially growing D. radiodurans diads grow throughout their cell cycle via remodelling of their entire cell surface (peripheral and septal cell walls) and divide in alternate orthogonal planes. As proposed in an early study of D. radiodurans 33, our 3D images confirmed that growth of the dividing septum occurs through a closing door mechanism and not as a diaphragm as has been reported for ovococci 40,41. One side of the septum starts growing precisely from the middle of the central septum, while the other grows from the middle of the opposite peripheral cell wall, both at a 90° angle to the central septum. This observation is surprising and will be the subject of future studies. Septal growth lasts for two-thirds of the cell cycle (Phases 3-6). Asynchronization of the division process was frequently observed: (i) between two cells that form a diad, resulting in cells of the same diad being in different phases of the cell cycle, or (ii) within a given cell, in which one side of the closing septum started growing earlier or faster than the other side, although overall, no difference in growth speed was observed between the two sides of the closing septum (Figure S3). In exponentially growing D. radiodurans, once the dividing septum closes (Phase 6), the newly formed tetrads are relatively short-lived, lasting for a dozen of minutes before splitting into two diads to initiate a new cell cycle (Phase 1). Cytokinesis results from the progressive increase in the curvature of the central septum (and reduction in its length) until the two diads eventually separate, creating a small gap between the two adjacent diads (see separating tetrad, bottom left in Figure 1A). It was recently shown in S. aureus, that the splitting of the diads into single cells occurs on the millisecond timescale driven by the turgor pressure 42,43. In D. radiodurans, visualizing this splitting process was more difficult, since the cells do not undergo any substantial morphological changes upon cytokinesis. Instead the separation appears to be slower and progressive, most likely catalyzed by the enzymatic processing of the cell wall in this region of the cell.
In every organism, nucleoid organization and cell division are intimately linked. The present study presents the first morphological characterization of the nucleoids of a coccus bacterium. The nucleoids of D. radiodurans were found to display a remarkable plasticity, adopting multiple, distinct conformations as the cells progress through their cell cycle. Our single-molecule and ensemble measurements suggest that the dynamic nature of these nucleoids may be facilitated by the loose binding of the major D. radiodurans NAP, the histone-like protein HU, to the genomic DNA. To our knowledge, this is the first report of such a wide diversity of nucleoid structures in bacteria. In addition to the previously reported toroid-shape, the nucleoid was also found to adopt an elongated branched structure that aligned parallel to the central septum. This arrangement perpendicular to the newly growing septum provides clear evidence that nucleoid organization and chromosome segregation are tightly coupled to the cell division process. In many model bacteria including S. aureus, E. coli and B. subtilis, nucleoid occlusion (NO) systems, typically composed of sequence specific DNA-binding proteins, play a central role as spatial and temporal regulators of cell division by preventing the assembly of the divisome over the nucleoid 44. Interestingly, in D. radiodurans Phase 3 cells in which septal growth is starting, three major conformations of nucleoids can be observed: toroids, squares and crescent-shaped nucleoids formed by the opening of the ring-like structures prior to their elongation and alignment with the central septum in Phase 4. Changes in the cell morphology and in particular start of septal growth, thus seem to precede the major rearrangements of the nucleoid, which suggests that the nucleoid is not the major determinant of the division site in D. radiodurans. Although no NO system has so far been identified in D. radiodurans, we cannot exclude that such a system might also be at play in the regulation of cell division in this organism.
The Min system that prevents the formation of the Z-ring in aberrant locations 45, might be the main factor ensuring that cytokinesis occurs at midcell in D. radiodurans. Many spatial features have been proposed as key determinants of pole-to-pole oscillations of Min in E. coli, such as the highly negative membrane curvature at the poles 46, the longest axis of confinement 47, the longest possible distance for the diffusion of MinD 48 or the symmetry axes and scale of the cell shape 49. D. radiodurans diads do not possess any poles per se. In Phase 1, cell ellipticity is close to 1 and cells possess an infinite number of symmetry planes (i.e. every plane orthogonal to the central septum). It is only when the cells start growing in Phase 2 (with a major change in the length of the major axis and no change in minor axis length leading to a decrease in the ellipticity of the cells), that the number of symmetry planes is reduced to two, both orthogonal to the central septum, one having its normal vector parallel to the long axis of the cell and the other with its normal vector parallel to the second short axis. A possible oscillation of MinCDE along the long axis of the ellipsoidal shaped cell, parallel to the central septum, could be the basis for the precise placement and angle of growth of the dividing septum, orthogonal to the central septum and along the short axis of the cell. In such a scenario, the Min system would sense the geometry of the cell and start to oscillate in Phase 2.
As bacterial nucleoids fill a large fraction of the total cell volume, a common hypothesis suggests that the cell size and shape may play a major role in chromosome positioning and sizing. It was recently proposed that cell shape and size were ‘sensed’ indirectly by chromosomes via the pressure exerted by depletion forces induced by cytosolic crowders 8,50, which would promote the compact shape and central localization of nucleoids in cells. In the case of D. radiodurans, several studies have described its nucleoid as being highly condensed 26–30, but this visual impression is in part due to the relatively large size of D. radiodurans cells that have a mean volume of ~2.6 µm3. We therefore assessed the volume occupied by the nucleoids of D. radiodurans and compared them to that of E. coli, which had previously been reported to occupy ~70% of the cell volume 11. Our measurements revealed that the mean volume of D. radiodurans nucleoids is very similar to that of E. coli (0.7 vs 0.8 µm3), but because of the larger cell volume, the fraction of the cell occupied by the nucleoid is significantly lower in the case of D. radiodurans and was found to vary from 0.30 to 0.35 depending on the phase of the cell cycle. This fraction even decreased below 0.2 in stationary phase cells, which are smaller and present highly condensed nucleoids. Such small cell fractions occupied by the nucleoid suggest either that depletion forces do not play a major role in nucleoid positioning and compaction in D. radiodurans or that D. radiodurans cells possess an unusually high concentration of crowders.
Because D. radiodurans possesses multiple copies of its genome, compared to a single copy in E. coli, the average DNA density in D. radiodurans nucleoids is thus significantly higher than in E. coli, thereby confirming the impression that D. radiodurans nucleoids are particularly condensed. Interestingly, the nucleoids were found to be most compact in Phases 4 and 5, during the final stages of chromosome segregation. The specific positioning of the nucleoids in the center of cells and parallel to the central septum, together with their increased levels of compaction during these late stages of the cell cycle are reminiscent of metaphase chromosomes in eukaryotic cells that align at midcell prior to segregation of each one of the two chromosomes into the two daughter cells.
Finally, we compared the number and distribution of the oriC and ter loci of chromosome 1 as a function of the cell cycle. As previously reported 25, there were significantly more oriC foci in cells than ter sites irrespective of the stage of the cell cycle. This has been proposed to result from differences in the duration of periods of transient cohesion of the oriC and ter loci from the various copies of chromosome 1 after replication - this period being longer for ter - and/or a very late replication of ter compared to oriC 25. Our cartography of the localizations of oriC and ter loci in D. radiodurans at each phase of the cell cycle also revealed very distinct distribution patterns for the two loci and allow us to propose a first model for nucleoid organization and choreography in a coccus bacterium possessing a complex, multicopy genome (Figure 7). Chromosome arrangement in D. radiodurans is neither longitudinal nor transversal, but instead radial with oriC sites distributed all around centrally located ter sites. The ter loci were clearly retained and clustered by a yet unknown mechanism in the central region of the cells until just before cytokinesis (Figure 7B and C). ter foci could indeed still be seen in the small gap formed by the closing septal doors in Phase 5 cells. Segregation of replicated ter loci occurred at a very late stage of the cell cycle (Phase 5; Figure 7C) and the duplicated ter loci then progressively migrated to the centre of the two new sister cells. In contrast, oriC loci localized throughout the cell, with the exclusion from membrane proximal regions, ruling out any direct anchoring of oriC loci to the cell walls as has been observed in some bacteria 51–53. Further work will be needed to understand the detailed choreography of this complex genome and how it relates to the various nucleoid structures observed in this study, but our findings suggest that cell shape may critically influence both nucleoid organization and chromosome arrangement.
Methods
Bacterial cultures
All Deinococcus radiodurans strains (wild-type, HU-mCherry, HU-PAmCherry GY17031, and oriC/ter labelled strains GY15787 and GY15800, Table 1) were grown at 30°C with shaking in TGY2X supplemented with the appropriate antibiotics. Escherichia coli strain BL21 (DE3) (Table 1) was grown at 37°C with shaking in LB medium. For microscopy experiments, D. radiodurans cells were pre-grown the day before and then diluted for an overnight growth until reaching exponential (OD650 ~0.3-0.5) or stationary phase (OD650 >5). To establish the growth curve of wild-type D. radiodurans cells, cultures inoculated at different cell densities were measured every 2 h and then combined to obtain a complete growth curve. The final growth curve corresponding to the mean of three independent experiments was fitted to an exponential growth curve (Y=Y0*exp(K*X), with Y0 corresponding to the Y value at T=0 and K the growth rate) using GraphPad Prism 6, and the doubling time was calculated as ln2/K.
Sample preparation for microscopy
General sample preparation
Glass coverslides used for PALM/PAINT imaging of D. radiodurans were treated in an ozone oven for at least 10 min. 1.75% low-melting agarose (Biorad) pads dissolved in minimal medium 54 (for PALM) or TGY2X (for spinning-disk microscopy) were poured on a cover slide inside a frame made with double-faced tape. A glass coverslip was placed on top of the agarose pad in order to flatten the surface of the agarose while hardening. This coverslip was removed once the agarose had solidified. For imaging, bacterial cultures were centrifuged 3 min at 3000 × g and were resuspended in 10 µl of medium or washed in high purity PBS (GIBCO). 1 µl of this cell suspension was deposited on the pads. The drop was spread over the surface of the pad by rotating the coverslip. A second coverslip was placed over the agarose pad containing the sample, thereby immobilizing the bacteria on the pad.
Nucleoid and cell membrane staining
To determine the growth phase and cell volumes, cells were incubated with Nile Red for 15 min at 30°C, with agitation and then placed on an TGY agarose pad. For single timepoint images acquired with the spinning disk microscope, the Nile Red was used at a concentration of 30 µM. For correlative spinning-disk microscopy and PAINT imaging the concentration was 300 nM. For PAINT imaging alone, the concentration was 15 nM. To determine nucleoid volumes, cells in exponential phase were incubated with Nile Red (30 µM) and Syto9 (150 nM) for 15 min at 30°C, with agitation and then placed on a TGY2X agarose pad. In all cases, the cells were rinsed to remove excess dye prior to imaging.
Peptidoglycan labeling
Cells in exponential growth phase were stained with the fluorescent BADA dye 55 at a concentration of 31 µM, incubated at 30°C with agitation for 15 min. Then Nile Red was added (15 µM) for another 15 min at 30°C with agitation. The cells were then rinsed in TGY2X at 30°C in order to remove unbound dyes. Cells were then deposited on TGY2X agarose pads for imaging (T=0) or returned to the incubator until T=130 min.
Time-lapse imaging
For imaging of the cell growth and division process, cells in exponential growth phase were incubated with Nile Red (15 µM) for 15 min at 30°C, with agitation. Cells were then rinsed in TGY2X before imaging. For time-lapse imaging of the nucleoids, exponentially growing HU-mCherry expressing cells or Syto9-stained (150 nM for 5 min) wild-type cells were used. In all cases, cells were deposited on an TGY2X agarose pad designed with air holes to oxygenate the cells 56. The samples were maintained at 30°C during image acquisition.
Foci acquisition
A saturated pre-culture of oriC or ter labelled HU-mCherry expressing D. radiodurans cells (Table 1) was diluted 60X in TGY2X supplemented with 3.4 µg/mL chloramphenicol and grown for a further 5 h until reaching an OD650 between 0.3 and 0.5. Cells were then stained with HADA (30 µM) for 30 min and deposited on a TGY2X agarose pad.
Image acquisition
PALM/PAINT imaging
Wide-field illumination was achieved by focusing the laser beams to the back focal plane of a 100x 1.49-numerical-aperture (NA) oil immersion apochromatic objective lens (Olympus). The intensity and time sequence of laser illumination at the sample were tuned by an acousto-optical tunable filter (AOTF; Quanta Tech). Near circular polarization of the laser beams was ensured by inserting a polychromatic quarter-wave plate downstream the AOTF. 8 to 20 cells were imaged per field of view. Typically, for each sample 2 or 3 fields of view on the same agarose pad were imaged and at least three independent experiments were performed on different days. Sample drift was corrected in ImageJ using gold nanobeads (Sigma) deposited on the agarose pads nearby the bacteria.
Single particle tracking
Data were acquired with continuous 561 nm light illumination, at low power (130 W/cm²) using a frametime of 30 ms. The Trackmate plugin for ImageJ was used to localize the particles in each frame and to connect the coordinates into trajectories 57. Simple LAP (Linear Assignment Problem) Tracker generated the tracks, with a maximal allowed linking distance set to 200 nm and a maximal frame interval between two spots to be bridged set to 2. Only tracks that contained more than four points were exported into MATLAB for further processing. The TrackArt MATLAB software 36 was used to analyze trajectories and determine cumulative probability distributions and apparent diffusion coefficients. Trajectories shorter than 10 frames or with a minimal individual mean square displacement (MSD) fit R² values below 0.9 were filtered out.
Spinning disk imaging
Spinning-disk confocal microscopy was performed on the M4D cell imaging platform at IBS using an Olympus IX81 microscope and Andor iXon Ultra EMCCD Camera with the laser beams focused to the back focal plane of a 100X 1.49-numerical-aperture (NA) oil immersion apochromatic objective lens (Olympus). The intensity and time sequence of laser illumination at the sample were tuned by an acousto-optical tunable filter. For cell and nucleoid volume analyses, series of Z-planes were acquired every 100 nm using a piezo stage, whereas 200 nm was used for the multiplex imaging of oriC and ter labelled D. radiodurans strains. Fluorescence emission was collected through a quad-band Semrock™ Di01-T405/488/568/647 dichroic mirror and the corresponding single band emission filters. For Nile Red time-lapse imaging (Movie S1), single plane image sets (2D) were acquired every 10 min, for a total period of 4 h, using very low 561 nm laser power (few micro-Watts) and 100 ms exposure times. Nucleoid time-lapse imaging (Movies S2 and S3) was achieved by acquiring 3D images (Z-planes were acquired every 100 nm) every 10 min for a period of 3 h. An additional 3D time-lapse video of HU-mcherry was also performed with short interval times (20 sec) over a period of 10 min (Movie S5).
FRAP measurements
A time-lapse acquisition of the fluorescence channel (mCherry) was performed for 60 timepoints at the fastest acquisition speed (10 timepoints before and 50 timepoints after photobleaching). Photobleaching (Andor FRAPPA module) was set at the wavelength used for imaging, targeting a single point (crosshair) of the analyzed nucleoid. Bleaching was performed using a low intensity setting (2% of the nominal maximum laser power), for 160 µs dwell time and 12 repeats to achieve a sufficiently small bleached region and with a loss of at least 50% of the original signal intensity. FRAP analysis was performed using EasyFRAP after measuring intensities using 58, as described in the original publication 59.
Data analysis
Cell measurements
To calculate the volume of each cell, an ellipse was fitted to the cellular membrane of Nile Red-labelled cells in 2D images (Figure S2A). Measurements of the shorter (Lshort) and longer (Llong) axes of these ellipses were extracted from such fits. The distance between the central septum and the extremity of the exterior of the ellipse along the short axis was measured (P), to retrieve the distance between the septum and the center of the ellipse along the short axis (w). With all these values, the total volume of the cell was computed using the equations presented in Figure S2A with the cells approximated as prolate ellipsoids. The length and width of the elongated and toroid shaped nucleoids were measured as described in Figure S2C and D.
Cell wall growth analysis
To determine the extent of BADA (green) versus Nile Red (red) staining in Phase 1 cells, the perimeter of each cell was measured by fitting an ellipse to the Nile Red-labelled cell membrane. The fraction of cell wall labelled with BADA was then determined by measuring the two angles as shown in Figure S2B. The perimeter of Nile Red only stained membranes (without any BADA labeling) for an angle α, is retrieved from the equation (1). The BADA labelled perimeter was then determined by computing the lengths of the two sections of cell wall devoid of BADA labeling and subtracting them from the whole perimeter of the ellipse.
Nucleoid volume measurements
Z-stacks of the Syto9 fluorescence signal corresponding to the nucleoids were deconvoluted in Volocity using 7 iterative cycles and a calculated PSF (488 nm/520 nm). Cropped images in which the fluorescence signal was homogeneous throughout the full field of view were then exported to the Imaris (Bitplane) software, where 3D objects corresponding to the nucleoids were created based on the automatic thresholding of the fluorescence signal. The volumes of these objects were then extracted for further analysis. The same procedure and settings in Imaris were used to extract nucleoid volumes from all our samples, including E. coli nucleoids.
Foci distribution
Using the Volocity software, Z-stacks of the GFP signal were iteratively deconvoluted using a calculated PSF (488 nm/520 nm). In the case of ter, Z-projections for each channel were concatenated into multi-channel stacks. In the case of oriC, the same multi-channel stacks were assembled, but we also verified that fluorescent foci were not lost by the Z projection. Fluorescent foci were manually marked in ImageJ to extract their xy coordinates as well as the dimensions and phase of the cells in which they were found. The number of foci and their distribution were then automatically determined with respect to the average cell dimension of D. radiodurans cells at each phase of the cell cycle derived from PAINT imaging.
Statistical analysis
The non-parametric Mann-Whitney tests were performed with GraphPad Prism 6 in order to assess the difference in cell dimensions and nucleoid volumes during the cell cycle. P values below 0.01 were considered as significant and were indicated with asterisks: **P ≤ 0.01, ***P≤0.001, **** P≤0.0001.
Contributions
KF and JT designed the research. KF, FL and JPK performed the microscopy experiments. PS provided the genetically modified strains of D. radiodurans. YSW synthesized the HADA dye. KF, FL, JPK, DB and JT analyzed the data. KF, DB and JT wrote the manuscript and all authors discussed the results and approved the manuscript.
Declarations of interest
The authors declare no competing interests.
Data availability statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Acknowledgements
We thank Prof. Yves Brun and Prof. Michael Van Nieuwenhze for providing us with the BADA dye. This work was supported by CEA Radiobiology grant (A-IRBIO-01-19) and the CEA Irtélis program (PhD grant of KF). This work used the M4D imaging platform of the Grenoble Instruct-ERIC Center (ISBG : UMS 3518 CNRS-CEA-UGA-EMBL) with support from FRISBI (ANR-10-INBS-05-02) and GRAL (ANR-10-LABX-49-01) within the Grenoble Partnership for Structural Biology (PSB).