|Home | About | Journals | Submit | Contact Us | Français|
We have determined the three-dimensional (3D) architecture of the Caulobacter crescentus genome by combining genome-wide chromatin interaction detection, live-cell imaging, and computational modeling. Using chromosome conformation capture carbon copy (5C) technology, we derive ~13 Kb resolution 3D models of the Caulobacter genome. These models illustrate that the genome is ellipsoidal with periodically arranged arms. The parS sites, a pair of short contiguous sequence elements involved in chromosome segregation, are positioned at one pole of this structure, where they nucleate a compact chromatin conformation. Both 5C and imaging experiments demonstrate that placing these sequence elements at new genomic positions yields large-scale rotations of the genome within the cell. Utilizing automated fluorescent imaging, we orient the genome within the cell and illustrate that within the resolution of our data the parS proximal region is the only portion of the genome stably attached to the cell envelope. Our approach provides an experimental paradigm for deriving insight into the cis-determinants of 3D genome architecture.
The three-dimensional (3D) architecture of the genome both reflects and regulates its functional state (Dekker, 2008; Fraser and Bickmore, 2007; Misteli, 2007; Thanbichler and Shapiro, 2006a). For example, chromatin loops that place promoters and distant enhancers within close spatial proximity play important roles in eukaryotic transcriptional regulation (Kagey et al., 2010; Miele and Dekker, 2008; Tolhuis et al., 2002; Vernimmen et al., 2007). In addition to illustrating the functional ramifications of genome folding, such examples of long-distance interactions also illustrate that knowledge of the primary sequence of the genome is not necessarily sufficient to infer where individual portions of the genome lie within space. Hence, to characterize the folding patterns of genomes and to derive insight into the functional relationships between loci it is necessary to comprehensively interrogate the spatial positioning of many loci. However, such comprehensive and high-resolution investigations have been technically challenging.
The recent development of several high-throughput technologies, including automated fluorescent imaging (Viollier et al., 2004) and chromosome conformation capture (3C)-based approaches (Dekker et al., 2002; Dostie et al., 2006; Duan et al., 2010; Fullwood et al., 2009; Lieberman-Aiden et al., 2009; Simonis et al., 2006; Zhao et al., 2006) has begun to enable studies of genome-wide chromosome folding. Fluorescent microscopy-based approaches allow the accurate determination of the sub-cellular positions of increasing numbers of defined chromosomal loci, while high-throughput 3C-based approaches enable quantification of inter-loci interaction frequencies that can subsequently be used to infer the average 3D distances between these loci. Studies utilizing one or both of these approaches have highlighted the potential of genome-wide studies of chromosome structure and have begun to reveal specific features of chromosome folding including the transcription-based compartmentalization of the human nucleus (Baù et al., 2010; Lieberman-Aiden et al., 2009; Osborne et al., 2004; Shopland et al., 2006; Simonis et al., 2006) and the correlation between a locus’ genomic and sub-cellular positioning in bacteria (Nielsen et al., 2006; Teleman et al., 1998; Viollier et al., 2004; Wang et al., 2006b). However, the detailed structures of genomes are only beginning to be revealed and many details, including the identities of the sequence elements that define such structures, await further elucidation.
We sought to determine the high-resolution 3D structure of an entire genome and to utilize the resulting structure to identify the sequence elements that define its architecture. Towards this goal, we studied the synchronizable bacterium, Caulobacter crescentus (Caulobacter, hereafter) whose circular chromosome is organized such that the origin and terminus of replication reside near opposite poles of the cell and other loci lie along the long axis in an order correlating with their genomic distance from the origin (Viollier et al., 2004). To derive higher resolution insights into the folding of this genome we employed a multi-pronged approach which utilized chromosome conformation capture carbon copy (5C) (Dostie et al., 2006), 3D modeling (Baù et al., 2010), and live-cell imaging. Using 5C, we measured 28,730 contact frequencies between loci spanning the entire Caulobacter genome and used these frequencies to derive 3D models of the genome in wild-type and genetically modified strains. Coupling these data with highly automated fluorescent microscopy, we oriented our models with respect to specific cellular landmarks. Our models demonstrate that the chromosomal arms are periodically arranged. Moreover, both our 5C and microscopy data indicate that changing the position of a short (10 Kb) region including the parS sites affects the orientation of the entire 4 Mb chromosome within the cell.
Spatial distances between genomic loci can be inferred from the frequencies at which these segments physically interact within a population of cells. Regions of the genome positioned close in space contact frequently, while regions located far apart rarely contact. To determine the frequency of such physical interactions, we used chromosome conformation capture (3C) technology (Dekker et al., 2002), which utilizes cross-linking with formaldehyde in conjunction with spatially constrained ligation to assess the average spatial proximity of genomic loci (Figure 1A). Each 3C experiment yields a genome-wide library of ligation products whose relative frequencies reflect the 3D structure of the genome.
We developed a 3C protocol optimized for bacteria and employed it to generate genome-wide ligation product libraries from Caulobacter swarmer cells, one of Caulobacter’s two cell types. To ascertain whether these libraries accurately reflected the spatial arrangement of the genome, we determined whether the frequencies of particular ligation products within these libraries (contact frequencies) were consistent with known features of the Caulobacter chromosome spatial organization. Caulobacter swarmer cells contain a single ~4 Mb circular chromosome that is divided into a left and right arm by a diametrically opposing origin and terminus of replication which reside near opposite poles of the cell (Viollier et al., 2004). Thus, we expected that the origin and terminus would rarely contact. We used PCR to determine the frequency of ligation products (contact frequencies) between the origin and terminus and found that, as expected, the contact frequencies between loci positioned near the origin and terminus were significantly lower (< 256 times) than contact frequencies between loci positioned near each other within the origin and terminus regions (Figure 1B). Thus, we conclude that 3C libraries accurately reproduce previously established features of Caulobacter genome spatial organization.
We next used chromosome conformation capture carbon copy (5C) to comprehensively determine the composition of our swarmer 3C libraries. 5C, described in detail before (Dostie et al., 2006; Dostie et al., 2007), uses highly multiplexed ligation-mediated amplification of pairs of plus and minus strand probes to measure the relative frequencies of tens of thousands of 3C ligation products in parallel. In our modified 5C approach, bar-coded oligonucleotide probes were used to specifically detect and amplify ligation junctions between restriction fragments, and the frequency of such junctions was quantified by high-throughput polony-sequencing of the ligated probe pairs (Shendure et al., 2005) (Figure 1A and S1A).
To interrogate the Caulobacter genome at high density, we designed 339 5C probes with 169 complementary to the plus and 170 complementary to the minus strand of restriction fragments (Figure 1C). These probes cover the Caulobacter genome in an alternating fashion with adjacent fragments queried by opposite strand probes, enabling the measurement of 28,730 contact frequencies and the uniform sampling of the Caulobacter genome at an average resolution of ~13 Kb (std deviation of 13 Kb).
We generated genome-wide 5C libraries for synchronous Caulobacter swarmer cells and found that the resulting contact frequencies were highly correlated across replicates (r > 0.94 - Table S1), indicating a high degree of reproducibility of individual contact frequencies. We collectively represent these frequencies as two-dimensional contact maps in which the plus and minus strand probed restriction fragments are arranged along orthogonal axes. The contact frequency between any given pair of fragments is thus located at the intersection of the corresponding row and column (Figure 1D).
The swarmer cell 5C contact map contains two prominent diagonals that intersect near the center (corresponding to the region containing the terminus − ~2 Mb) and corners (corresponding to the region containing the origin - 0 and 4 Mb) of the map (Figure 1D). To understand the structural implications of these diagonals it is helpful to consider what such a map would look like for an unconstrained circular chromosome. In such a structure, interactions would exclusively occur between loci proximal in the genome, yielding a contact map with a single diagonal stretching from the upper-left to lower-right corners of the contact map (Figure S1B,C). In addition to this diagonal, the swarmer cell contact map contains a second diagonal which stretches from the lower left to the upper right corner, indicating the presence of a prominent set of interactions between fragments separated by large genomic distances. Correspondingly, the contact frequency profiles of most fragments contain a peak of short-range interactions centered about the genome coordinates of these fragments as well as a second peak of long-range interactions. Interestingly, these long-range interactions are centered about loci on the opposite arm that are roughly equidistant from the origin (Figure 1E). Collectively, these short and long-range peaks indicate that the swarmer chromosome structure is ellipsoidal with loci roughly symmetric about the origin folded close in space and a long axis connecting regions near the origin and terminus (Figure 1E).
The swarmer 5C contact map allows us to identify the maximally polar regions of the ellipsoidal swarmer chromosome. As these regions are approached along either arm of the chromosome, the genomic distance to the contacting region on the opposite arm becomes smaller until the proximal short-range and the long-range (inter-arm) peaks in restriction fragment contact frequency profiles merge. Thus, the maximally polar regions of the genome correspond to the positions in the contact map at which the two diagonals intersect (Figure S1D). We find that one polar locus corresponds to a region near the origin and the second to a region near the terminus. Thus, our data are consistent with previous microscopy data that indicated that the origin and terminus are positioned near opposite poles of the cell (Viollier et al., 2004). Importantly, the resolution of our 5C data allows us to more precisely define the genomic locations of the regions located at the extreme poles. We find that the origin proximal maximally polar region is located 25 +/− 17 Kb to the left of the origin and contains the parS sequence elements (Livny et al., 2007; Mohl and Gober, 1997) (Figure S1E,F) and that the terminus-proximal maximally polar region lies 42 +/− 17 Kb to the left of the dif site (Figure S1E,F) (Jensen, 2006). Of note, the parS sites are known to act as centromere-like elements in Caulobacter (Toro et al., 2008), and the dif sites play a role in the resolution of chromosome dimers formed during replication and segregation (Jensen, 2006). Thus, our contact frequency data suggest that the chromosome is folded about sequence elements that are critical to chromosome segregation.
Next, we used the swarmer 5C data to generate models of the 3D conformation of the Caulobacter swarmer cell genome. Our 5C data represent a comprehensive set of contact frequencies between loci spanning the entire Caulobacter genome. It has previously been demonstrated that contact frequencies are reliable proxies for the average spatial distances between loci (Lieberman-Aiden et al., 2009; Miele et al., 2009) and we observed a similar relationship in our data (Figure S2A,B). Therefore, we can convert 5C contact frequencies into average spatial distances which can subsequently be used to derive 3D models.
To generate models we utilized the Integrative Modeling Platform (IMP, http://www.integrativemodeling.org) (Alber et al., 2007) to search for the 3D conformations of the Caulobacter genome that satisfy the collective distances inferred from our 5C contact maps (Figure 2A). This approach yields population-average conformations, which represent the dominant folding patterns across the cell population, but does not reveal the variations in detailed conformations that undoubtedly exist within this population. Specifically, we utilized fluorescent microscopy derived distances between hundreds of chromosomal loci located on the same arm (Figure S2B) in conjunction with 5C data to derive a calibration curve mapping between contact frequencies and spatial distances (See Supplemental Experimental Procedures) and we then used that relationship to transform 5C contact frequencies into spatial distances (Figure 2A-I and Supplemental Experimental Procedures). Next, treating each restriction fragment as a point in space, we connected fragments with springs with equilibrium (relaxed) distances equal to the 5C-derived distances (Figure 2A-II). The 3D coordinates of all fragments were then stochastically initialized (Figure 2A-III) and in an iterative process which utilizes simulated annealing, the spatial locations of fragments were randomly modified until a structure that satisfied as many of the 5C-derived distances as possible was attained (Figure 2A-IVa). We repeated this optimization 56,000 times to generate an ensemble of structures that was maximally consistent with our 5C data. Finally, we structurally superimposed and clustered the 10,000 best models (i.e. those with the fewest violations of the 5C distance constraints) into groups of highly similar structures (Figure 2A-IVb).
We find that the models of the Caulobacter swarmer chromosome group into only four structurally similar clusters (Table S2), suggesting that while the chromosome may be flexible and folded somewhat differently in each cell, there are a limited number of dominant conformations. To represent the swarmer model clusters, we present 3D density maps in which the volume occupied by each fragment indicates the variation in this fragment’s positioning across the models within the cluster (Figure 2B). Such volume elements may represent dynamic variability in a fragment's spatial positioning (Elmore et al., 2005; Fiebig et al., 2006; Reyes-Lamothe et al., 2008) and also likely encompass local structures such as topological domains which are below the resolution of our data (Postow et al., 2004; Stein et al., 2005).
In 3D coordinate space we find that the differences between models are largely the result of mirroring of portions of the structures. The presence of mirrors is expected as perfect mirror images are indistinguishable by 5C and are equivalent in distance space (where optimization is performed by IMP). Indeed, clusters 1 and 2 as well as clusters 3 and 4 are pairs of nearly perfect mirror images (Figure S2D). Furthermore, clusters 1 and 2 are partial mirrors of clusters 3 and 4 (Figure S2E). Thus, the four model clusters represent small variations on a single conformation.
The models within all clusters display structural features apparent from the raw contact maps: their overall geometry is ellipsoidal with loci that are roughly equidistant from the origin folded close in space. The origin proximal maximally polar fragment (along the long axis) is located ~7 Kb to the left of the parS elements (Figure 2C and Table S2), which is within the resolution of 5C data (~13 Kb). Based upon fragment intra-cluster variability we determined that the effective resolution of the models is between 175 and 225 nm (Figure S2C). We expect that the actual resolution of individual models may be much higher as models may be locally very similar along their length but diverge significantly at only a few positions.
A prominent novel feature emerges from all four clusters: the arms are wound sinusoidally through space with roughly 1.5 period repeats per arm. The partial mirroring between clusters 1 and 2 and clusters 3 and 4 has the effect of causing the arms to be either intertwined (clusters 3 and 4) or separated (clusters 1 and 2). We favor the intertwined conformation as the corresponding model clusters have lower variability (Figure S2C) and lower IMP objective function scores (Table S2). However, it is possible that both conformations exist within a population of swarmer cells.
Our models suggest that the parS sites play a direct role in organizing the swarmer chromosome. Such a finding is consistent with recent analyses that have suggested that these sequence elements are specifically anchored to the Caulobacter old cell pole through interactions with the ParB and PopZ proteins (Bowman et al., 2008; Ebersbach et al., 2008; Toro et al., 2008). Based upon these independent lines of evidence, we hypothesized that the orientation of the entire Caulobacter chromosome is defined by parS-based anchoring. To test this hypothesis we studied an inversion strain (ET166) in which the parS sites were moved ~400 Kb away from the origin (Figure 3A). We performed 5C on this strain and discovered that the resulting contact map displays a marked displacement of the diagonal representing contacts between the two arms towards the upper-left corner of the map (Figure 3B), indicating a change in the pattern of interactions between opposite arm loci. 3D models generated using these data show a striking genome-wide clockwise rotation of loci within the ellipsoidal structure, thereby positioning the parS sites at a structural pole (Figure 3C, Figure S3A and Table S3). Thus, changing the position of the parS sites in the genome resulted in a large-scale re-organization of its 3D architecture. This dramatic re-orientation of the chromosome was not observed in a control strain, ET163, which carried a similar, but slightly smaller (by 10 Kb), inversion that did not change the positions of the parS sites (Figure 3A,D–E, and Figure S3C). Therefore, we conclude that the rotation observed in ET166 was a direct result of moving the 10 Kb region containing parS sites (parS region, hereafter), to a new genomic location.
5C-derived contact frequencies reflect the relative positions of loci but do not indicate the sub-cellular positions at which contacts occur. Thus, we next turned to in vivo microscopy to place the observed structures within the context of the cell. Starting with the inversion strains described above, we generated a set of derivative strains which had lacO arrays inserted at different chromosomal locations, thereby allowing us to utilize LacI-CFP fusion proteins to visualize the sub-cellular positions of these loci in living cells (Figure 4A).
Using custom-built microscope control software (Christen et al., 2010) and automated image analysis (Toro et al., 2008), we first measured the sub-cellular positions of over 60 different loci in at least 500 wild-type cells. The resulting images illustrate that loci on the same arm are arranged in a roughly linear fashion along the long axis of the cell (Figure 4B) and hence are consistent with both our models as well as previous microscopy data (Viollier et al., 2004). We next analyzed the inversion strains and confirmed that in strain ET166, in which the parS sites were located ~400 Kb away from the origin, the entire chromosome did in fact rotate in a clockwise fashion within the cell in response to the repositioning of the parS region (Figure 4B). Such a rotation was not observed in the control strain, ET163, which displayed only a subtle rotation of loci in the opposite direction (Figure 4C). Importantly, the terminus region was amongst the portions of the genome whose sub-cellular position was changed in ET166 relative to wild-type, indicating that as our 3D models predicted, the terminus is not anchored to a pole of the cell.
To confirm the importance of the parS region in shaping the organization of the swarmer genome we created an additional inversion strain, ET322, in which this region was moved to the right of the origin (as opposed to ET166, where it was moved to the left of the origin) (Figure S4). We expected that such a positioning of the parS region would result in a wholesale rotation of the chromosome similar to that seen in ET166, but in the opposite direction. Indeed, we find that in this inversion strain, the parS sites remained at the cell pole, while the rest of the chromosome, including the terminus, rotated as expected (Figure S4).
Thus, both our 3D models and live-cell imaging demonstrate that the only locus whose position is fixed in the cell is the 10 Kb region containing the parS sites. The gross orientation of the entire chromosome is determined, to a large extent, by the relative position of this region. At our current resolution we are unable to determine whether or not additional (i.e. non-parS) anchor points exist within the 10 Kb region we have manipulated. In fact, since ParB, the DNA-binding protein which binds to the parS sites, itself interacts with a number of other polarly localized proteins including, MipZ (Thanbichler and Shapiro, 2006b), and PopZ (Bowman et al., 2008; Ebersbach et al., 2008), it seems likely that the parS anchoring region represents a large nucleoprotein complex that could encompass tens of kilobases of sequence.
To assess whether there exist additional constraints on chromosome folding outside of the parS sites, we wished to determine whether the arms of the chromosome run in register from pole to pole. If the arms were to run in parallel along the long axis without any additional structural constraints, we would expect that opposite arm loci positioned at similar genomic distances from the parS sites would be located at similar positions along the long axis of the structure/cell. Thus, for each fragment in each model we identified the closest opposite arm fragment with respect to positioning along the long axis of the models and then compared the genomic distances of these loci to the parS sites (the maximally polar loci). For the majority of the wild-type genome, we find that the closest opposite arm fragments are indeed roughly equidistant from the parS sites, with a small deviation from symmetry near the opposite pole (Figure 5A). Thus, our models indicate that the arms of the wild-type swarmer chromosome are well aligned.
We performed a similar analysis for strain ET166, the inversion strain in which the parS sites were moved ~400 Kb to the left of the origin. Interestingly, we find that despite the fact that much of the genome rotates in register, placing the parS sites at a structural pole, there are a large number of loci located outside of the inverted region which despite residing at different genomic distances from the parS sites are aligned along the long axis of the cell (Figure 5B). A similar, and more extreme lack of symmetry was also observed in the models for strain ET163, the inversion strain in which the parS sites remained near the origin (Figure S5A). Such findings suggest that there are additional factors (See discussion), possibly located within the inverted regions, which control the positioning of the chromosomal arms.
We next used our models to assess if the swarmer cell genome is uniformly compacted along its length, or if instead there are regions of unusual compaction, as has been observed in other systems (Gilbert et al., 2004; Mercier et al., 2008). The average distance amongst sets of neighboring loci in compact genomic regions should be smaller than the corresponding distances in less tightly packed regions. Therefore, we can utilize the mean distance between neighboring loci in a model as a measure of compaction. To ensure that that such a measure accurately reflects compaction, we first corrected the observed inter-fragment distances for differences in genomic site-separation. Specifically, for each model we utilized the mean relationship between intra-arm site-separation and spatial distance to define expected inter-loci distances. We then calculated the log ratios of these expected distances to the observed distances and determined the average ratios for sets of neighboring fragments, thereby defining compaction scores. This process was repeated for each model in our clusters and the cluster-wide average compaction scores were plotted against genome position.
We found that in wild-type cells, fragments located near both poles showed above average levels of compaction (Figure 6A,B). However, the region to the left of the parS sites was arranged in a particularly compact fashion (Figure 6A). Intriguingly, we found that in the inversion strains (ET163 and ET166), this region of high compaction traveled along with the parS sites (Figures 6A,C and S5B). Specifically, chromosomal regions of average compaction became highly compact when the parS sites were moved nearby, and regions that were compact became less condensed when separated from the parS elements (Figures 6C and S5B). While such changes in compaction may partially be the result of the movement of regions closer to/away from a structural pole, two lines of evidence indicate that polar localization alone does not entirely explain these changes. First, we observe that in ET166 the compaction level of the region moved closer to the parS sites increases significantly more than the region moved closer to the opposite pole (Figure 6A). Secondly, while in both wild-type and ET166 cells there are regions near the poles opposite the parS sites that are arranged somewhat compactly, these regions are asymmetrically positioned with respect to these poles (Figure 6B). Such asymmetry would not be expected if polar localization alone rendered a region of the genome compact. Thus, we conclude that the parS sites nucleate a compact chromosome conformation that spreads 100–200 Kb into flanking genomic regions.
The results described above reveal that there are strong constraints on locus positioning along the long axis of the cell. We next set out to explore whether similar constraints are present along the short axis by assessing if the Caulobacter chromosome is free to rotate about the long axis of the cell. We constructed three double-label strains in which the markers were located on opposite arms at similar genomic distances from the parS elements (Figures 7A). We then determined the number of cells in which each arm was closest to the ventral side of the cell and found that the queried loci did not preferentially localize to a given side of the swarmer cell (Figure 7A and S6).
To confirm that loci do not have defined positions along the long axis of the cell, we also analyzed a large set of strains in which a single locus was labeled. We expected that if the arms of the chromosome do in fact reside on a particular side of the cell, the distributions of the positioning of such loci along the short axis of the cell would reveal such a preferential localization pattern. However, the distributions of 38 markers at unique chromosomal sites in a total of ~200,000 cells illustrate that the arms were equivalently distributed about the short axis of the cell (Figure 7B). Thus, our results demonstrate that while the Caulobacter chromosome is oriented along the long axis of the cell through the action of the parS elements, there is no such organization along the short axis, and instead the chromosome is free to rotate about the long axis of the cell.
Genomes must be compacted by at least three orders of magnitude to fit within their sub-cellular volumes. While such compaction could in theory yield highly disordered structures, recent evaluations of genome-wide chromosome folding have illustrated that genomes reproducibly adopt specific spatial arrangements (Dekker et al., 2002; Duan et al., 2010; Lieberman-Aiden et al., 2009; Viollier et al., 2004). Such investigations have also begun both to enable deeper analyses of the complexities of 3D genome architecture and to reveal the forces and cis-elements that define such organization. Here, we have investigated the high-resolution folding patterns of the Caulobacter swarmer cell genome using high-throughput chromatin interaction detection, 3D modeling, and fluorescent microscopy, and have used the resulting structures to investigate the cis-elements that define the local and global arrangement of this bacterial genome.
Our contact-based 3D models and fluorescent microscopy illustrate that the parS sites define the global structure of the Caulobacter genome. Our data indicate that these sites reside at the pole of the wild-type swarmer chromosome/cell and that moving a 10 Kb region containing them elsewhere in the genome yields large-scale rotations of the chromosome that reposition these elements at the cell/structural pole. The parS sites are conserved sequence elements present within 69% of sequenced bacterial genomes (Livny et al., 2007) and are the first sequences to segregate regardless of their positions in the genome (Toro et al., 2008). In addition, evidence suggests they anchor the chromosome to the pole via interactions with the ParB and PopZ proteins (Bowman et al., 2008; Ebersbach et al., 2008). We therefore propose that our findings are a consequence of the parS sites being the first genomic elements to segregate to the opposite pole where they become anchored and thereby establish the global orientation of the genome. Such a model is supported by the fact that in both wild type and ET166 cells loci segregate in an order roughly corresponding to their genomic distances from the parS sites (Toro et al., 2008; Viollier et al., 2004). Thus, our findings indicate that the global orientation of the chromosome may be defined by the order of segregation and that the rotated global arrangements of the genome observed in inversion strains ET166 and ET322 are the result of a perturbation of the order of segregation of loci that results from the movement of the parS sites.
In addition to shaping the global orientation of the Caulobacter genome, we show that the parS sites also nucleate a compact chromatin conformation over ~100 Kb, as moving these elements elsewhere in the genome condenses the regions placed near these sites and de-condenses the regions moved away. We hypothesize that these sites yield such a compact conformation due to interactions with ParB which itself is thought to physically interact with Structural Maintenance of Chromosomes (SMC), a protein known to play a role in chromosome compaction and segregation across a number bacterial species including Caulobacter (Britton et al., 1998; Jensen and Shapiro, 1999; Wang et al., 2006a). Consistent with such an explanation, it has been demonstrated that in Bacillus subtilis ParB (SpoOj) spreads from the parS sites into adjacent regions of the genome (Breier and Grossman, 2007) and recruits the SMC complex to these regions (Gruber and Errington, 2009; Sullivan et al., 2009). Thus, the data presented here, which demonstrate for the first time that the centromeric region of a bacterial chromosome is particularly compact in vivo, connect SMC’s previously noted effects upon chromosome segregation and compaction.
We observe that the contact frequencies between loci on the same arm decline as the genomic distances between these loci increases and that there are no obvious demarcated “blocks” of interactions in the swarmer contact maps which would denote regions of the genome that are physically isolated from each other. Thus, consistent with previous microscopy data (Viollier et al., 2004), our contact data indicate that the Caulobacter swarmer cell genome does not contain large physically and genetically isolated domains similar to the macrodomains observed in Escherichia coli (Espeli et al., 2008; Mercier et al., 2008; Valens et al., 2004).
Our models also elucidate the detailed arrangement of the arms of the chromosome and demonstrate that the chromosomal arms are arranged in a periodic fashion. Interestingly, a helical arrangement of newly replicated DNA has been observed in B. subtilis (Berlatzky et al., 2008). While the mechanism behind such a periodic arrangement in Caulobacter and/or B. subtilis is yet to be unraveled, such arrangements could represent an energetic minimum (Maritan et al., 2000). Alternatively, these highly regular folding patterns could be the consequence of interactions between the genome and helically arranged cytoskeletal proteins (Alyahya et al., 2009; Dye et al., 2005; Gitai and Shapiro, 2003).
We find that opposite arm loci equidistant from the parS elements are aligned at similar positions along the long axis of the wild-type swarmer chromosome structure. However, movement of the parS sites elsewhere in the genome leads to the emergence of regions of the structure in which opposite arm loci are no longer well aligned. These mis-alignments suggest that there are additional constraints on the positioning of loci along the long axis of the structure/cell. Consistent with this conclusion, we find that in ET163, the inversion strain in which the parS elements remain in their wild-type locations, the cellular positions of many loci throughout the chromosome change (with the exception of the parS sites). In keeping with the segregation-based model posed above, the inversions in strains ET163 and ET166 could affect the timing of segregation of loci and thereby influence the alignment and positioning of the arms of the chromosome. Additionally, it is conceivable that genes located on opposite arms need to be maintained close in space and hence may be recalcitrant to the global rotation brought forth by the movement of the parS elements in strain ET166.
By utilizing highly automated fluorescent microscopy we have also determined the orientation of the swarmer cell chromosome within the cell. These studies indicate that loci have no preferential locations about the short axis of the cell and therefore that the chromosome has no preferential orientation about this axis. This finding illustrates that the parS sites represent the only sequence elements that stably anchor the chromosome to the cell. However, it remains possible that events such as transertion (Woldringh, 2002), the simultaneous transcription, translation, and insertion of membrane proteins into the cellular envelope, may transiently couple the genome to the membrane.
The work presented here illustrates how a comprehensive study of genome 3D architecture can provide insight into the roles of sequence elements in defining this structure. The experimental paradigm we introduce is general and could be used in conjunction with genetic perturbations to elucidate the roles of nucleoid associated proteins, cis-elements, and DNA-templated processes such as transcription and replication in shaping the folding of the genome. With additional advances, including decreases in DNA sequencing costs, such a paradigm could also be applied to larger eukaryotic genomes to further elucidate the complex relationships between genome sequence, structure, and function.
More detailed experimental procedures including descriptions of our inversion strain construction, polony sequencing, model generation, 5C data/model analysis, and live-cell imaging procedures can be found in the Supplemental Experimental Procedures.
Synchronous swarmer cells were isolated as previously described (Alley et al., 1993) and were subsequently cross-linked with 1.0% formaldehyde. Cells were pelleted, resuspended in 1X TE at a final concentration of 1 × 107 cells/µL, and lysed with Ready-Lyse (Epicentre) lysozyme (20 U/µL). Cells were then solubilized with 0.5% SDS, and chromatin was digested with BglII (0.94 U/µL) in 1X buffer 3 containing 1.0% Triton X-100. Digested chromatin was subsequently solubilized again using SDS (0.9%) and was ligated at a final concentration of 0.25 ng/µL in 1X T4 DNA ligase buffer containing 1% Triton X-100. 3C libraries were purified via phenol-chloroform extraction and ethanol precipitation.
Melting temperature normalized probes were designed such that their sequence was complementary to the sequence upstream of BglII restriction sites with adjacent probes containing sequence corresponding to opposite strands (plus vs. minus). Plus strand probes were flanked by the forward emulsion PCR primer (see Supplemental Experimental Procedures) and minus strand probes were flanked by the reverse complement of the reverse emulsion PCR primer. 3C libraries (1.8 ng/µL) were denatured at 95°C for 10 minutes in 1X AMP Ligase buffer, probes were subsequently annealed for 14 hours at 65°C at a concentration of 0.2 nM (per probe), and ligation was subsequently performed at 65°C for 1 hr using AMP ligase (0.0167 U/µL). Ligated probes were amplified via PCR (see Supplement Experimental Procedures).
We acknowledge Tony Tsai for preliminary microscopy data and Marian Walhout, Zhiping Weng, Andrew Tolonen, Jae Kim, Kun Zhang, and Nikos Reppas for helpful suggestions/reading of the manuscript. We thank the IMP community (http://www.integrativemodeling.org/) especially Daniel Russell, Ben Webb and Andrej Sali as well as the Chimera developers (http://www.cgl.ucsf.edu/chimera/) including Thomas Goddard and Tom Ferrin. Work conducted by MAU, MAW, GJP, and GMC was supported by a Department of Energy GTL center grant (to GMC). ET was partially supported by the Smith Stanford Graduate Fellowship and LS was supported by National Institutes of Health Grants R01 GM51426 and R24 GM073011-04. MAM-R and DB were funded by the Spanish Ministerio de Ciencia e Innovación (BIO2007/66670 and BFU2010/19310) and JD was supported by a grant from the National Institutes of Health (HG003143) and a W.M. Keck Foundation Distinguished Young Scholar Award.
Author ContributionsMAU, MAW, and JD conceived of using 5C to analyze Caulobacter chromosome structure. MAU conducted all of the 5C experiments with GP providing sequencing support. Modeling was conceived of by MAM-R, JD, and DB, and was performed by MAU, MAW, and DB with support from MAM-R. Subsequent analyses were conceived of and performed by MAU, MAW, and JD with aide from GMC and MAM-R. ET and LS conceived of using chromosome rearrangements to study genome organization. ET constructed the inversion strains and performed the microscopy-based studies. S-HH and MJF wrote the image analysis and microscope control software. MAU, ET, MAW, JD, MAM-R, HHM, LS, and GMC wrote the paper.