ABSTRACT
The Structural Maintenance of Chromosomes (SMC) complex plays an important role in chromosome organization and segregation in most living organisms. In Caulobacter crescentus, SMC is required to align the left and the right arms of the chromosome that run in parallel down the long axis of the cell. However, the mechanism of SMC-mediated alignment of chromosomal arms remains elusive. Here, using a genome-wide chromosome conformation capture assay (Hi-C), chromatin immunoprecipitation with deep sequencing (ChIP-seq) and microscopy of single cells, we show that Caulobacter SMC is recruited to the centromeric parS site and that SMC-mediated arm alignment depends on the chromosome partitioning protein ParB. We provide evidence that SMC likely tethers the parS-proximal regions of the chromosomal arms together, promoting arm alignment. Strikingly, the co-orientation of DNA replication and the transcription of highly-expressed genes is crucial for chromosome-wide alignment. Highly-transcribed genes near parS that are oriented against DNA replication disrupt arm alignment suggesting that head-on transcription interferes with SMC translocation and arm alignment. Our results demonstrate a tight interdependence of bacterial chromosome organization and global patterns of transcription.
INTRODUCTION
The chromosomes of all organisms must be compacted nearly three orders of magnitude to fit within the limited volume of a cell. However, DNA cannot be haphazardly packed, and instead it must be organized in a way that is compatible with numerous cellular processes that share the same DNA template, including transcription, DNA replication, and chromosome segregation. This is particularly challenging in bacteria since these chromosome-based transactions happen concomitantly rather than being separated temporally, as in eukaryotes. Application of microscopy-based analyses of fluorescently-labeled DNA loci along with genome-wide chromosome conformation capture assays (Hi-C) have revealed a well-defined in vivo three-dimensional organization of bacterial chromosomes (Badrinarayanan et al., 2015a; Le et al., 2013; Umbarger et al., 2011; Viollier et al., 2004; Wang et al., 2013). Hi-C provides quantitative information about the spatial proximity of DNA loci in vivo by measuring the frequencies of crosslinking between different regions of the chromosome (Lieberman-Aiden et al., 2009). The first application of Hi-C to bacteria examined the Caulobacter crescentus chromosome (Le et al., 2013). Hi-C analysis of Caulobacter confirmed the global pattern of chromosome organization: in cells with a single chromosome, the origin of replication (ori) is at one cell pole, the terminus (ter) is near the opposite pole, and the two chromosomal arms are well-aligned, running in parallel down the long axis of the cell (Le et al., 2013; Viollier et al., 2004) (Fig. 1A). We discovered that a Structural Maintenance of Chromosomes protein (SMC) is crucial for the alignment of the left and right arm of the chromosome in Caulobacter (Le et al., 2013), but how SMC achieves this alignment remains poorly understood.
SMC proteins are widely conserved from bacteria to humans. In eukaryotes, SMC1 and SMC3, together with accessory proteins, form a cohesin complex that holds sister chromatids together until after they achieve a bipolar attachment to the mitotic spindle. The related condensin complex, comprised of SMC2, SMC4, and interacting partners, promotes the compaction of individual chromosomes during mitosis. In most bacteria, there is a single SMC composed of an ATPase “head” domain, a dimerisation “hinge” domain, and an extended antiparallel coiled-coil region in the middle (Nolivos and Sherratt, 2014) (Fig. S1A). Two SMC monomers dimerise, and together with the bacteria-specific proteins ScpA and ScpB, form a tripartite ring that can bring distal DNA segments close together to help organize bacterial chromosomes (Bürmann et al., 2013; Hirano et al., 2001; Mascarenhas et al., 2002; Soppa et al., 2002) (Fig. S1A). The topological entrapment of DNA by a ring-shaped SMC complex has been shown for cohesin and condensin, and for a Gram-positive Bacillus subtillis SMC (Cuylen et al., 2011; Ivanov and Nasmyth, 2005; Murayama and Uhlmann, 2014; Wilhelm et al., 2015). It is likely that topological entrapment is a general feature of all SMC complexes.
How SMC gets loaded onto DNA and topologically entraps DNA is generally well-studied in eukaryotes but not yet completely understood (reviewed in Uhlmann, 2016). In bacteria, SMC loading, translocation and DNA entrapment is less well known, but likely involves the ParA-ParB-parS system (Gruber and Errington, 2009; Lin and Grossman, 1998; Livny et al., 2007; Minnen et al., 2011; Sullivan et al., 2009). ParB is a site-specific DNA-binding protein that nucleates on a centromere-like parS sequence (Mohl and Gober, 1997; Mohl et al., 2001; Toro et al., 2008) and then spreads non-specifically along the DNA, likely forming a large nucleoprotein complex (Breier and Grossman, 2007; Graham et al., 2014; Taylor et al., 2015). ParA, a Walker-box ATPase, interacts with ParB and is required for the segregation of the ParB-DNA complex, and ultimately for the partitioning of replicated chromosomes to daughter cells (Figge et al., 2003; Lim et al., 2014). In B. subtilis, ParB loads SMC onto the chromosome mainly at the ori-proximal parS sites (Gruber and Errington, 2009; Marbouty et al., 2015; Minnen et al., 2016; Wang et al., 2015). Loaded SMC then translocates from parS to distal parts of the chromosome in an ATP hydrolysis-dependent manner (Minnen et al., 2016; Wang et al., 2017). This action is thought to individualize the origins of replicated chromosomes, thereby helping to segregate replicated chromosomes to opposite daughter cells. It has been proposed that B. subtilis SMC loaded at parS translocates the full length of the chromosome to the terminus to promote chromosome arm alignment (Gruber and Errington, 2009; Minnen et al., 2016; Wang et al., 2017). However, ChIP-seq studies indicate that B. subtilis SMC is most enriched near ori, so whether SMC directly promotes arm alignment uniformly across the genome is unclear (Minnen et al., 2016; Wang et al., 2017).
Caulobacter harbours both a canonical SMC-ScpA-ScpB complex as well as a ParA-ParB-parS system. The ParA-ParB-parS complex is essential and cells lacking ParB are not viable (Mohl and Gober, 1997). In contrast, Caulobacter SMC is not required for survival in laboratory conditions (Le et al., 2013). Caulobacter cells lacking SMC grow slightly slower but are not temperature sensitive nor prone to accumulating suppressor mutations as originally suggested (Jensen and Shapiro, 1999). Nevertheless, an ATPase-defective smc mutant shows a severe defect in sister chromosome segregation in this bacterium, suggesting that Caulobacter SMC is also playing a role in chromosome segregation (Schwartz and Shapiro, 2011).
How SMC influences other cellular processes like transcription and is, in turn, influenced by these processes, is not well understood. Abundant, multi-subunit RNA polymerases can translocate through some DNA-bound proteins (Epshtein et al., 2003). In yeast, cohesin is pushed along the chromosome in the same direction as transcription, without dissociating (Lengronne et al., 2004; Ocampo-Hafalla et al., 2016). Similarly, in budding yeast, RNA polymerase can drive the short-range relocation of condensin (D’Ambrosio et al., 2008; Johzuka and Horiuchi, 2007). Whether bacterial RNA polymerase affects SMC has not been systematically investigated. Notably, in almost all bacteria, most genes, especially highly-expressed genes, are transcribed in the same direction as replication, i.e. from ori to ter (Guy and Roten, 2004; Rocha, 2008; Rocha and Danchin, 2003). This co-orientation of genes could help push SMC from ori to ter; alternatively, or in addition, genes transcribed in a head-on orientation could antagonize the translocation of SMC.
Here, we use Hi-C and chromatin immunoprecipitation with sequencing (ChIP-seq), together with microscopy-based analysis of single cells to elucidate the role and the mechanism of SMC in the global organization of the Caulobacter chromosome. We provide evidence that (i) SMC is required for the progressive alignment of the two chromosomal arms, proceeding in the ori-ter direction; (ii) Caulobacter SMC is loaded onto the chromosome at the centromeric parS site and ParB is essential for the SMC-mediated arm alignment; (iii) Caulobacter SMC most likely functions as a tether to actively cohese the chromosomal arms together; and (iv) head-on transcription can profoundly disrupt the alignment of chromosomal arms, likely by interfering with SMC translocation from the centromeric parS site. Altogether, our results demonstrate a tight interdependence of bacterial chromosome organization and global patterns of transcription.
RESULTS
The SMC-ScpA-ScpB complex is required for the alignment of chromosomal arms
Caulobacter cells lacking SMC show a dramatic reduction in inter-arm DNA-DNA interactions (Le et al., 2013) (Fig. 1B, Fig. S1B). To test whether the Caulobacter ScpA and ScpB homologs are also required for inter-arm interactions we generated Hi-C contact maps for ΔscpA and ΔscpB cells (Fig. 1B-D, Fig. S1B). Each strain was grown to exponential phase and then synchronized to obtain a homogeneous population of G1-phase cells before fixing with 1% formadehyde and performing Hi-C. To generate Hi-C contact maps from the raw sequencing data, we divided the Caulobacter genome into 10-kb bins and assigned to corresponding bins the interaction frequencies of informative ligation products (see Methods). Interaction frequencies were visualized as a matrix with each matrix element, mij, indicating the logarithm of the relative interaction frequency of DNA loci in bin i with those in bin j. To emphasize the ori-proximal region, we oriented the Hi-C contact maps such that the ori (0 kb or 4043 kb) is at the center of the x- and y-axis, and the left and the right chromosomal arm are on either side (Fig. 1B).
On the contact map of wild-type Caulobacter, the primary and high-interaction diagonal represents interactions between DNA on the same arm of the chromosome, i.e. intra-arm contacts, while the less prominent secondary diagonal represents DNA-DNA interactions between opposite arms, i.e. inter-arm contacts (Fig. 1B). Note that, in contrast to Hi-C maps reported previously (Le et al., 2013), here we use the logarithm of interaction frequencies to facilitate visualization of these weaker inter-arm interactions. The diagonal pattern of inter-arm contacts on a Hi-C map indicates that each locus on one chromosomal arm interacts with a locus roughly equidistant from the ori on the opposite arm, reflecting a global alignment of the chromosomal arms. Consistent with our previous studies (Le et al., 2013), the inter-arm interactions are significantly reduced in a Δsmc strain. The Hi-C map for ΔscpB revealed a similar decrease in inter-arm interactions, with no strong or obvious change in intra-arm interactions (Fig. 1C-D, Fig. S1B). The Hi-C map for ΔscpA also exhibited a decrease in interarm interactions, though not nearly as significant as for Δsmc and ΔscpB (Fig. 1C-D, Fig. S1B). These data are consistent with ScpA and ScpB forming a complex with SMC that promotes the co-linearity of chromosomal arms in Caulobacter.
ParB induces a progressive alignment of chromosomal arms from ori to ter
In Gram-positive bacteria such as B. subtilis and Streptococcus pneumoniae, SMC is loaded onto the chromosome by ParB at ori-proximal parS sites (Gruber and Errington, 2009; Minnen et al., 2011; Sullivan et al., 2009). To test whether this mechanism is conserved in Caulobacter, a distantly related and Gram-negative bacterium, we used a strain where ParB, which is essential for viability, can be depleted (Mohl and Gober, 1997; Thanbichler and Shapiro, 2006). The promoter of parB at its native chromosomal locus was replaced with a xylose-inducible promoter, Pxyl. Cells grown to exponential phase in the presence of xylose were then washed free of xylose and incubated for five hours in a rich medium supplemented with glucose to inhibit Pxyl activity. Immunoblot analysis with an α-ParB antibody indicated ~2.5 fold decrease in ParB concentration after the five hours in glucose (T = 0 min, Fig. S2A). These cells were then fixed with 1% formadehyde and examined by Hi-C. The contact map of ParB-depleted cells exhibited a clear reduction in inter-arm contacts, similar to Δsmc cells, indicating a role of ParB in maintaining chromosomal arm alignment, possibly by loading SMC onto DNA (Fig. 2A-B, Fig. S2B).
As with B. subtilis SMC (Gruber and Errington, 2009; Wang et al., 2015), the Caulobacter SMC complex may be recruited and loaded onto DNA via a direct interaction with ParB. If so, we reasoned that overexpressing a strong ParB-interacting protein might prevent interactions with SMC and, in turn, disrupt alignment of the chromosomal arms. To test this hypothesis, we performed Hi-C on cells overexpressing a YFP-tagged variant of ParA harboring a K20R substitution. ParA(K20R) is defective in ATP binding and was suggested to bind ParB more tightly than wild-type ParA (Shebelut et al., 2010). Consistent with that conclusion, we found, using a yeast two-hybrid assay, that ParA(K20R) and ParB produced five-fold more activity with a β-galactosidase reporter than did ParA and ParB (Fig. 2C). The Hi-C contact map for a strain overexpressing ParA(K20R) showed a modest, but significant decrease in chromosomal arm alignment compared to a control strain overexpressing wild-type ParA (Fig. 2D-E, Fig. S2C). This result reinforces our conclusion that ParB is required for the SMC-mediated alignment of chromosomal arms in Caulobacter.
We hypothesized that ParB helps load SMC onto the chromosome at parS sites, leading to the SMC-mediated alignment of chromosomal arms. This model predicts that chromosome arm alignment by SMC starts at, or near, parS and then progresses toward the terminus. To investigate the directionality and dynamics of chromosome arm alignment, we again depleted ParB by growing cells in rich medium with glucose for 5 hours, and then added back xylose to induce ParB de novo. Samples were taken 0, 5, 10, 15, 25 and 30 minutes after xylose addition and immediately fixed for Hi-C analysis. Immunoblot analysis with α-ParB antibody showed gradual accumulation of new ParB after adding back xylose (Fig. S2A). At the 0 and 5 minute time points, we observed very little inter-arm interaction, as above (Fig. 2F, Fig. S2D-E). However, over time, the inter-arm interactions increased, beginning close to ori and parS and then extending toward ter (Fig. 2F-G, Fig. S2D-E). The two arms aligned directionally at a rate of ~19 kb per minute (~320 bp per second) and ~16 kb per minute (~267 bp per second) for the left and right arm, respectively (Fig. 2F-G). These data are consistent with a model in which SMC is loaded by ParB, likely at parS sites, and then translocates down the arms toward ter, driving their alignment.
The parS site found adjacent to the parAB operon is ~8 kb from ori in Caulobacter crescentus (Toro et al., 2008). To directly test the model that SMC is loaded onto DNA at parS sites, we inserted a 260-bp DNA fragment containing a parS site ~1800 kb from ori, near the chromosomal terminus (Fig. 3A). We verified that this ectopic parS site was sufficient to recruit ParB, using ChIP-seq analysis of ParB, which binds to the native parS site (Fig. 3B). ChIP-seq of CFP-tagged ParB indicated that most ParB remains associated with the native ori-proximal parS site, but with a significant fraction of ParB also recruited to the ectopic parS at +1800 kb (Fig. 3B). We then performed Hi-C on the strain harboring the additional parS site and observed a new secondary diagonal, emanating from the approximate position of the ectopic parS site (Fig. 3C). The extent of alignment from this ectopic site was less than that associated with the original parS, an issue examined in depth below. We also inserted a second parS at +2240 kb, and this was again sufficient to recruit ParB and induce an alignment of the flanking DNA (Fig. 3). Taken together, our results support a model in which ParB loads SMC at parS sites, leading to the subsequent progressive alignment of flanking DNA.
SMC promotes DNA alignment most effectively for parS-proximal genomic regions
After being loaded at parS sites, SMC likely translocates down the chromosomal arms to drive their alignment. Strikingly however, the extent of inter-arm interactions was not uniform across the entire chromosome indicating that each loaded SMC complex may not travel the entire length of the chromosome. In fact, we noted that in wild-type cells, inter-arm interactions reduced gradually away from parS, leveling out after ~600 kb in either direction (Fig. 1D). Similarly, in the strains in which an ectopic parS site was inserted 1800 kb or 2200 kb from the origin (Fig. 3), the extent to which flanking DNA was "zipped up" and aligned was limited to only a few hundred kilobases.
These observations suggested that SMC is not uniformly distributed across the chromosome and instead may be enriched in the regions showing highest inter-arm interactions. To test this idea, we first used ChIP-seq to map the genome-wide distribution of epitope tagged-SMC. We fused the SMC-encoding gene to a FLAG tag at the N-terminus and placed this allele of smc downstream of a xylose-inducible promoter (Pxyl) on a medium-copy-number plasmid. We then performed Hi-C on Δsmc cells that produced this FLAG-tagged SMC via leaky expression from Pxyl(Fig. S3A). Chromosomal arm alignment was comparable to the wild-type level (Fig. S3B-C), indicating that FLAG-SMC is functional. Overproducing FLAG-SMC in the Δsmc background by adding xylose did not extend arm alignment beyond the wild-type level (Fig. S3B-C). For ChIP, the Δsmc Pxyl-flag-smc strain was grown to exponential phase in a rich medium at 30oC, washed with phosphate-buffered saline and then fixed with 1% formadehyde and crosslinker Gold for immunoprecipitation. SMC-bound DNA was isolated using α-FLAG antibody coupled to sepharose beads. The immunoprecipitated DNA was deep sequenced and mapped back to the Caulobacter genome to reveal enriched genomic sites. As a negative control, we performed α-FLAG ChIP-seq on wild-type Caulobacter, i.e. cells with untagged SMC (Fig. S3D). Individual sequence reads were allocated to regular 1 kb bins along the Caulobacter chromosome. ChIP-seq signals were reported as number of reads within each 1 kb bin in the ChIP fraction of FLAG-tagged SMC minus that of untagged SMC to normalize for non-specific immunoprecipitation (Fig. 4A). Although modest, we observed a clear enrichment of SMC-bound DNA near the parS site (~8 kb from the ori, on the left arm) (Fig. 4A), consistent with SMC loading at parS. SMC decreases in enrichment away from parS but is slightly enhanced close to the ter area (Fig. 4A and the Discussion). We also noted the enrichment of SMC at highly transcribed genes (Fig. 4A-B), most likely as an artefact of non-specific immunoprecipitation (Minnen et al., 2016; Nolivos et al., 2016; Teytelman et al., 2013)
Most notably, FLAG-tagged SMC was enriched above background in a region overlapping parS that extended from approximately +3680 kb to +345 kb, the same approximate region that shows the most extensive inter-arm alignment by Hi-C (Fig. 1D, Fig. 4A, Fig. S4A). We could not detect significant enrichment of FLAG-tagged SMC beyond this parS-proximal region, indicating that SMC is either not appreciably bound to parS-distal regions of the chromosome or its association drops below our limit of detection with ChIP-seq. In either case, we conclude that (i) SMC is not uniformly distributed across the genome and (ii) the enrichment of SMC is strongly correlated with the extent of chromosome arm alignment at the ori-proximal region.
To further test the relationship of SMC enrichment and the alignment of flanking DNA, we generated a strain bearing a relatively large inversion (involving genomic DNA normally between +3611 and +4038 kb) such that parS is relocated ~427 kb away from ori; this strain is referred to as the Flip 1-5 strain (Fig. 4B). ChIP-seq of FLAG-tagged SMC in the Flip 1-5 inversion strain showed an enrichment of DNA surrounding the relocated parS site at a peak level comparable to the ori-proximal parS (Fig. 4A-B), further supporting the conclusion that SMC is loaded by ParB at parS, and that SMC enrichment near parS is independent of ori.
The Hi-C contact map of the Flip 1-5 inversion strain showed a secondary diagonal akin to that in wild-type cells (Fig. 4C). However, the starting point of the arm/flanking DNA alignment was shifted and coincided with the genomic position of the relocated parS (Fig. 4C). This ectopic arm alignment was reduced dramatically in the absence of smc (Fig. 4C-D), indicating that arm alignment in the Flip 1-5 strain still depends on SMC. For the Flip 1-5 strain, the interarm alignment was again strongest in a limited region around parS, extending approximately 330 kb in each direction (Fig. 4D, Fig. S4B). The extent of arm alignment in the Flip 1-5 strain was approximately half that of wild-type cells (~600 kb), consistent with the reduced region showing SMC enrichment by ChIP-seq (Fig. 4B, Fig. S4B).
Collectively, the data presented thus far suggest that SMC loaded at parS may only translocate a limited way down the chromosome, or that most SMC molecules loaded translocate a limited distance. If so, we predicted that if the chromosome expanded, as occurs in elongating, G1-arrested cells, then parS-proximal regions would (i) remain better aligned than other chromosomal regions and (ii) exhibit a stronger dependence on smc for alignment (Fig. S5A). To test this model, we fluorescently labeled pairs of DNA loci, at equivalent distances from ori, but on opposite arms of the chromosome, using an orthogonal ParB/parS system (Badrinarayanan et al., 2015b). A pair of DNA loci were engineered at +200 kb and +3842 kb, i.e. within the parS-proximal domain showing strongest SMC enrichment and highest inter-arm Hi-C values. Another pair of DNA loci were labeled in the middle of each arm (+1000 kb and +3042 kb). And finally, to investigate chromosomal arm alignment near ter, a pair of DNA loci at +1800 kb and +2242 kb were fluorescently labelled. We then measured, for each pair of loci, inter-focus distances in a population of otherwise wild-type and Δsmc cells (Fig. 4E). To allow cells to elongate and expand their chromosomes, we measured inter-focus distances in cells where the only copy of dnaA was driven by an IPTG-regulated Plac promoter. Washing cells free of IPTG produced a population of cells that each contained just one copy of the chromosome and continued to grow but were unable to divide, leading to an elongated cell where the chromosome fills the entire available cytoplasmic space (Kahng and Shapiro, 2003; Le and Laub, 2016).
As cell length increased with time, we observed that the mean inter-focus distances for each pair of DNA loci also increased, consistent with an overall expansion of the chromosome (Fig. 4E). However, the rate of expansion was different, depending on genomic locations of labelled DNA loci. The inter-focus distance for the parS-proximal, 200 kb-3842 kb pair increased only modestly, ~0.5 μm as the cell length tripled (Fig. 4E). In contrast, the inter-focus distance for loci in the middle of the chromosome arms increased ~1.5 μm on average as the cell length tripled (Fig. 4E). Importantly, the inter-focus distance of the ori-proximal loci, but not the other locus pairs, increased in the Δsmc background (Fig. 4E). These microscopy-based analyses support our hypothesis that SMC most effectively aligns the chromosome arms nearest the ori and parS site where it is loaded.
In parallel, we analyzed Hi-C data on the elongated cells resulting from DnaA depletion (Fig. S5B-C) (Le and Laub, 2016). We observed that ~300 kb surrounding parS remains well-aligned even in elongated cells 3 hrs after IPTG withdrawal (Fig. S5B-C). In contrast, chromosomal arm alignment elsewhere on the chromosome was rapidly lost to the same extent as in Δsmc cells after withdrawing IPTG for 2 hrs (Fig. S5D-E). These results agree well with the single-cell microscopy results and together suggest that Caulobacter SMC functions most effectively to align chromosomal arms in close proximity to ori/parS.
Conflicts with transcription influence the SMC-mediated alignment of chromosomal arms
Collectively, our results indicate that SMC is predominantly associated with the parS-proximal region of the chromosome, leading to strongest arm-arm interactions within this region. Additionally, we noted two other features of the arm-arm interactions assessed by Hi-C and ChIP-seq. First, the frequencies of inter-arm interactions extended slightly faster along the left arm compared to the right arm (Fig. 2G). Second, the region showing the strongest inter-arm interactions did not extend as far from parS in the Flip 1-5 strain or in the strains harboring the ectopic parS sites near ter (Fig. 3C, Fig. 4C-D). These observations suggested that the genomic context surrounding parS may influence the translocation of SMC along each arm of the chromosome and, consequently, the patterns of inter-arm interaction.
We hypothesized that highly-expressed genes, particularly those oriented toward parS, may limit the translocation of SMC. To test this idea, we engineered a strain, hereafter called Flip 2-5, in which a ~419 kb DNA segment between +3611 kb and +4030 kb on the left arm was inverted (Fig. 5A, Fig. 6A). This inversion leaves parS at its original location, ~8 kb from ori, but dramatically changes the genomic context of the flanking DNA on one side of parS while leaving the other side unperturbed (Fig. 5A, Fig. S6A-B). Based on RNA-seq data collected from Caulobacter cells growing exponentially in rich medium (Le and Laub, 2016), we chose this segment for inversion as it contains several highly-expressed genes, including a ribosomal rRNA gene cluster and genes encoding ATP synthase, the glycine cleavage system, and cytochrome c oxidase, that all normally read in the ori-ter direction, i.e. co-directionally with replication and SMC translocation (Fig. 5A, Fig. S6A-B) (Le and Laub, 2016).
To assess the density of RNA polymerases directly, we performed ChIP-seq on exponentially-growing cells producing RpoC-FLAG as the only version of RNA polymerase β’ subunit (Fig. 5A, Fig. S6A-B). Separating sequencing reads based on the direction of transcription clearly demonstrated an enrichment of highly-expressed genes transcribed in the ori-ter direction in wild-type cells between +3611 and +4030 kb (Fig. 5A, Fig. S6A). We also confirmed, using ChIP-seq on RpoC-FLAG, that these same genes remain highly expressed in the Flip 2-5 background, but now on the opposite strand such that they read in a ter-ori direction, i.e. head-on with replication and SMC translocation from parS (Fig. 5A, Fig. S6B).
The Hi-C contact map of G1-phase Flip 2-5 cells showed, in sharp contrast to wild-type cells, a pronounced asymmetrical pattern of inter-arm interactions (Fig. 5B, Fig. S7A-B). The interarm interactions in the Flip 2-5 strain manifested as a nearly vertical streak in the Hi-C map, still emanating from a parS-proximal position (Fig. 5B). This vertical streak indicates that the strongest inter-arm interactions now occur between an ~50-80 kb region of DNA on the left side of parS with an ~400 kb segment on the right arm of the chromosome, a pronounced asymmetry compared to the pattern in wild-type cells (Fig. 5B). We confirmed that this asymmetric pattern of inter-arm interactions in the Flip 2-5 strain is still dependent on SMC as this vertical streak disappeared from the contact map of Flip 2-5 Δsmc cells (Fig. 5B). These results strongly suggest that the genomic context of DNA flanking the parS site, and likely the orientation of highly expressed genes, can dramatically influence the pattern of inter-arm contacts.
To assess whether the Flip 2-5 strain also led to a change in the genomic distribution of SMC, we performed ChIP-seq on FLAG-SMC in the Flip 2-5 background (Fig. 6, Fig. S4C). The enrichment of SMC dropped off slightly faster in the first ~150 kb away from parS, in the ori-ter direction, in the Flip 2-5 strain compared to wild type (Fig. 6A, Fig. S4C, Fig. S7C). However, most strikingly, the Flip 2-5 strain exhibited a series of new SMC ChIP-seq peaks, in addition to the one at parS, particularly in the region that was inverted (Fig. 6). Comparing the SMC and RpoC ChIP-seq profiles indicated that these new SMC peaks coincided with the highly expressed gene clusters that had been reoriented in the inversion to read toward ori and parS (Fig. 6C). These new peaks are not artefacts that are normally associated with highly-transcribed genes since (i) they are unique in Flip 2-5 cells but not in the wild type or Flip 1-5 control, and (ii) the shape and the magnitude of the unique peaks in Flip 2-5 are distinct from that of wild type and the Flip 1-5 cells (Fig. 6C). These data strongly suggested that transcription in a head-on orientation with respect to the direction of replication can either impede SMC translocation away from parS or drive the dissociation of SMC from DNA, limiting the extent of inter-arm contacts and, in some cases, produce an asymmetrical pattern of interarm contacts by Hi-C.
To test whether the asymmetrical inter-arm interactions in the Flip 2-5 strain arise because of the reoriented, highly expressed genes within the inverted genomic region, we first tested whether inhibiting transcription eliminated the asymmetrical pattern of inter-arm interactions. To this end, we treated Flip 2-5 cells with rifampicin (25 μg/ml) for 30 min before fixing cells for Hi-C (Fig. 5B). Rifampicin inhibits transcription elongation in bacteria, thereby eliminating actively translocating RNA polymerases from the chromosome (Campbell et al., 2001). As we reported previously for wild-type cells, the inhibition of transcription reduced short-range intraarm contacts (Le et al., 2013). In addition, for the Flip 2-5 strain, the vertical streak was eliminated and the inter-arm interactions reverted to a symmetric pattern on the diagonal (Fig. 5B, Fig. S7B), demonstrating that transcription is required for an asymmetrical inter-arm interaction pattern.
To further test the relationship between the orientation of highly expressed genes and SMC-dependent inter-arm interactions, we constructed three additional strains with different chromosomal inversions (Fig. 7A). We wondered if reversing the transcription orientation of a single highly-transcribing gene cluster, i.e. the rRNA gene cluster, would be sufficient to induce an asymmetrical inter-arm pattern. To test this possibility, we created the Flip 3-4 strain (Fig. 7A). The Hi-C map of G1-phase cells of the Flip 3-4 strain showed negligible changes to the inter-arm alignment compared to non-flipped cells (Fig. 7B-C). However, the inverted region in this strain is ~240 kb away from parS and, as noted above, SMC and SMC-dependent inter-arm interactions are strongest within a limited range around parS. Thus, we reasoned that the effect of an inverted rRNA cluster would be stronger if placed closer to parS. We did so by constructing the Flip 2-4 strain that has DNA between +3788 kb and +4030 kb inverted (Fig. 7A). Note that we could not easily relocate or insert a second ribosomal rRNA cluster next to parS. Hi-C on G1-phase cells of the Flip 2-4 strain showed a pronounced asymmetrical interarm interaction pattern (~20o deviation from a diagonal, starting after 80 kb from ori), though less dramatic than that of the Flip 2-5 cells in which several highly expressed genes in addition to the rRNA locus were inverted (Fig. 7B-C, Fig. 5B).
We further investigated the effect of transcription orientation bias on chromosomal arm alignment by inverting a DNA segment between +3611 kb and +3788 kb, the Flip 4-5 strain. Although this section does not contain an rRNA gene cluster, it includes four highly-transcribed operons that normally transcribe in the ori-ter direction in wild-type cells (Fig. 7A, denoted with square brackets in Fig. 5A). Preceding this segment, DNA between +3788 kb and +4030 kb is largely free of highly-expressed genes oriented toward ori/parS (Fig. 5A). Hi-C on G1-phase cells of the Flip 4-5 strain showed two distinct phases of inter-arm contacts (Fig. 7B-C). The first phase (~290 kb) is a typical set of symmetrical inter-arm contacts as seen in wild-type cells. The second phase coincides with the inverted DNA segment and has a pronounced asymmetrical inter-arm pattern (Fig. 7B-C). Lastly, inhibition of transcription by treating Flip 2-4 or Flip 4-5 cells with rifampicin eliminated the asymmetrical pattern of inter-arm interactions (Fig. S7D). Collectively, our results emphasize that highly-transcribed genes, depending on their transcriptional direction, can dramatically influence the action of SMC and the global organization of a chromosome.
DISCUSSION
Conflicts between SMC and RNA polymerase, and the consequences for bacterial chromosome organization
Chromosomes in all organisms are typically laden with DNA-bound proteins that likely influence the dynamics and movement of SMC (Davidson et al., 2016; Kanke et al., 2016; Ocampo-Hafalla et al., 2016; Stigler et al., 2016; Wang et al., 2017). Time-resolved ChIP-seq in Saccharomyces cerevisiae showed pre-existing cohesin being pushed along the chromosome by the transcription machinery at ~0.5 kb per min (Ocampo-Hafalla et al., 2016). Similarly, short-range relocation of budding yeast condensin (SMC2/4) and the SMC5/6 complex by RNA polymerase has been observed (D’Ambrosio et al., 2008; Jeppsson et al., 2014; Johzuka and Horiuchi, 2007). Our results indicate that the distribution and translocation of Caulobacter SMC, and the consequent alignment of chromosomal arms, is also strongly influenced by RNA polymerase and actively transcribed genes, particularly those oriented toward ori.
In bacteria, the transcription of high-expression genes is notably biased to the ori-ter direction (Fig. S6) (Guy and Roten, 2004; Rocha, 2008; Rocha and Danchin, 2003). The co-orientation of DNA replication and transcription reduces the chance of head-on collisions between RNA and DNA polymerases, which can disrupt transcription, obstruct replication, and elevate the rate of mutagenesis (Hamperl and Cimprich, 2016; Merrikh et al., 2012). Here, we showed using Hi-C that reversing this ori-ter directional bias in a segment of the chromosome as small as ~180 kb induced a major change in the chromosome organization in Caulobacter (Fig. 7), preventing the symmetric alignment of chromosome arms.
The chromosome organization defect that results from highly expressed genes oriented toward ori likely stems from a conflict between SMC and transcription, without involving the replisome. This conclusion is based on two key observations. First, the experiments involving strains with inverted chromosomal regions were performed on synchronous G1-phase Caulobacter cells, i.e. non-replicating cells. Second, the Flip 1-5 strain, where the SMC loading site (parS) was relocated to a mid-arm position, has highly-expressed genes (such as the ribosomal RNA gene cluster) transcribing in the opposite direction to that of the replisome in actively replicating cells. However, this strain did not exhibit a dramatic off-diagonal pattern of DNA interactions (Fig. 4C). Thus, we suggest that head-on conflicts between translocating SMC and RNA polymerase can directly shape bacterial chromosome organization. Additionally, our results on G1 cells demonstrate that DNA replication is not required for SMC translocation or its maintenance on the chromosome.
Given the widespread directional bias of transcription in bacteria (Rocha, 2008), the high conservation of SMC (Nolivos and Sherratt, 2014), and the fact that parS is typically positioned close to ori in most bacteria (Livny et al., 2007), we suspect that a relationship between transcription and SMC similar to that documented here is relevant in other bacteria. Indeed, the rate of SMC-induced alignment of parS-flanking DNA in B. subtillis also appears to be reduced by convergent transcription (Wang et al., 2017). An enrichment of highly-transcribed genes on the leading strand of replication has also been observed in certain regions of the human genome (Huvet et al., 2007) suggesting that a coordination of transcription orientation and SMC translocation may also be relevant in higher organisms.
The mechanism(s) that drive translocation of bacterial SMC from parS remains unknown. It is tempting to speculate that the ori-ter transcription bias might indicate that transcription helps push SMC away from its ori-proximal loading site. However, chromosome arm alignment in wild-type Caulobacter cells treated with rifampicin, which globally inhibits transcription, is slightly extended, not reduced (Fig. S8A-B). E. coli MukBEF, a non-canonical SMC, has been proposed to translocate via a “rope climber” mechanism (Badrinarayanan et al., 2012). In this model, a concerted opening and closing of just one MukBEF dimer in a handcuffed dimer-dimer allows an opened dimer to grab the next DNA segment before releasing the previously closed MukBEF, thereby “swinging” the dimer-dimer complex down DNA (Badrinarayanan et al., 2012). In B. subtilis, recent structural studies suggested that ATP binding, hydrolysis, and release can switch SMC between a rod- and a ring-like conformational states, with this motorlike cycling somehow mediating SMC translocation (Bürmann et al., 2013; Minnen et al., 2016). However, none of these current models can explain the directional movement of SMC from its loading site.
Whether smaller DNA-binding proteins can affect SMC translocation is not clear yet. The structural transition of B. subtilis SMC between a rod-shaped conformation (small lumen) and a ring-shaped conformation (large lumen) suggests that smaller protein-DNA complexes might also influence the dynamics of SMC on the chromosome. Recent in vitro single-molecule experiments in fission yeast showed cohesin’s mobility on flow-stretched DNA was impaired by nucleosomes and other DNA-bound proteins of various size (Stigler et al., 2016). In bacteria, there is a small but highly abundant histone-like protein called HU (Badrinarayanan et al., 2015a). We have seen that chromosome arm alignment in Caulobacter cells lacking HU extends further than that of wild-type cells (Fig. S8C-D), implying that SMC translocation might normally be restricted by HU.
SMC loading at the bacterial centromere parS site: a coupling between chromosome organization and segregation
Our results support the notion that Caulobacter SMC is, like in B. subtilis, recruited at or near parS, loaded in a ParB-dependent manner, and then redistributed towards ter. The Hi-C profile of cells depleted of ParB showed a similar loss in inter-arm interactions as observed with SMC, ScpA, and ScpB mutants (Fig. 1B-D, Fig. 2). Also, the Caulobacter SMC ChIP-seq profile indicates, as with B. subtilis SMC (Gruber and Errington, 2009; Marbouty et al., 2015; Minnen et al., 2016; Wang et al., 2017), highest enrichment near parS, with a progressive decrease toward ter. This pattern is consistent with SMC being recruited at or near parS and then redistributing toward ter.
Whether Caulobacter ParB directly loads SMC at parS is not clear yet. ParB does not show sequence or structural similarity to eukaryotic SMC loader proteins such as NIPBL/Scc2 (Uhlmann, 2016). But there is evidence in B. subtilis and S. pneumoniae that ParB interacts with and directly loads SMC (Gruber and Errington, 2009; Marbouty et al., 2015; Minnen et al., 2016; Sullivan et al., 2009; Wang et al., 2015). A previous report found that Caulobacter SMC and ParB co-localized sparsely, but ParB did not co-immunoprecipitate with SMC (Schwartz and Shapiro, 2011). However, this negative result may reflect a transient and weak interaction between SMC and ParB. A transient interaction would be advantageous in enabling the SMC complex to redistribute away from the initial point of loading.
ParB is a bacterial-specific protein, but likely works closely with and coevolves with SMC to ensure chromosome organization and chromosome segregation in bacteria. Interestingly, S. pneumoniae lacks a ParA homolog, yet retains ParB-parS to recruit SMC (Minnen et al., 2011), underscoring the tight connection of ParB and SMC. It is worth noting that, unlike B. subtilis, deleting the SMC-encoding gene in a wide range of bacteria, including Caulobacter as well as Streptomyces coelicolor, Staphylococcus aureus and S. pneumoniae, does not cause sensitivity to high temperature or fast-growing conditions (Le et al., 2013; Nolivos and Sherratt, 2014). Conversely however, ParA-ParB-parS is essential in Caulobacter but not in B. subtillis (Mohl and Gober, 1997; Mohl et al., 2001; Murray and Errington, 2008). The ParA-ParB-parS system and the SMC complex likely collaborate to ensure proper chromosome segregation and organization, but with slightly different contributions or relative weights in different organisms. Finally, we note that in γ-proteobacteria such as E. coli or V. cholerae, there is a distant SMC homolog, called the MukBEF system (Nolivos and Sherratt, 2014). Notably, ParA-ParB-parS systems do not exist in these bacteria, leaving open the question of how MukBEF complex loads on the chromosome, assuming it requires a specific loader (Nolivos and Sherratt, 2014). E. coli MukBEF was found, by ChIP-seq, to enrich at the replication terminus. We also observed a slight enrichment of Caulobacter SMC near the ter region (Fig. 4A-B). The enrichment of Caulobacter SMC near ter might occur via a parS-independent E. coli-like mechanism, however it does not result in DNA alignment in this area (Fig. 1B). The mechanism of SMC enrichment at ter is currently unknown in Caulobacter.
Evidence that bacterial SMC tethers chromosomal arms together
The Caulobacter SMC complex promotes interactions between loci at approximately equivalent positions on opposite arms of the chromosome up to at least 600 kb from parS. It could be that SMC physically tethers the arms together (an “active alignment” model). Alternatively, SMC could promote alignment passively by compacting each arm separately, reducing the cytoplasmic mobility of each arm and thereby increasing inter-arm interactions (a “passive” alignment model). The Hi-C patterns documented here for the wild-type and various inversion strains are most easily explained by the active alignment model in which SMC physically links DNA from both arms together (Fig. 8). Such a model is also appealing given the notion that SMC can topologically entrap DNA. Moreover, contact probability curves derived from the Hi-C data, which reflect global chromosome compaction, were generally very similar for Δsmc and wild-type cells (Le et al., 2013), suggesting that SMC plays only a minor role in intra-arm compaction in Caulobacter.
In the “active alignment” model, also suggested by studies of B. subtilis SMC (Wang et al., 2017), the inter-arm interactions documented by Hi-C may reflect loop generation by SMC (Fig. 8). In wild-type cells, DNA from each chromosomal arm may be effectively threaded through SMC at approximately similar rates as SMC moves towards ter and away from parS (Fig. 8A-B). In the inversion strains, DNA on the left arm may be threaded through SMC less efficiently than the right arm due to conflicts with convergent transcription (Fig. 8C). As suggested for eukaryotic SMC, loop extrusion or loop enlargement may be a general mechanism for folding chromosomes or bringing distant loci together (Alipour and Marko, 2012; Fudenberg et al., 2016; Goloborodko et al., 2016; Nasmyth, 2001). Processive loop extrusion/enlargement by SMC is also an efficient way to potentially resolve and linearly compact sister chromosomes in a manner that facilitates chromosome segregation (Nasmyth, 2001).
The active arm alignment and loop extrusion model raises the question of how SMC molecules distinguish intra-arm DNA from inter-arm DNA. Hypothetically, if SMC topologically entraps both DNA duplexes right at the parS loading site, instead of just a single duplex, SMC would be able to translocate downwards to ter while tethering both arms of the chromosome (Fig. 8A). Structural studies of SMC in B. subtilis showed that the lumen of this ring-shape complex is large enough to accommodate two DNA duplexes (Bürmann et al., 2013). Further, ParB is known to spread non-specifically on DNA from a parS site to bridge flanking DNA together (Breier and Grossman, 2007; Graham et al., 2014) such that SMC loaded at parS may be effectively trap the DNA duplexes of each arm (Fig. 8A). In B. subtilis, the spreading/bridging property of ParB is essential for an efficient SMC recruitment and inter-arm interactions (Graham et al., 2014; Wang et al., 2015).
Although we favor a model in which SMC is loaded at parS sites and physically tethers the chromosomal arms together, our results suggest that SMC is not uniformly distributed across the genome and that SMC may, in fact, not translocate the full distance from ori to ter, as suggested for B. subtilis. SMC-dependent, arm-arm interactions in Caulobacter are substantially more prominent in the first ~600 kb of the two chromosomal arms. Additionally, our ChIP-seq profile for SMC indicated clear enrichment mainly in this same ~600 kb region. And finally, our microscopy studies (Fig. 4E) demonstrated that different regions of the chromosome responded quite differently to a global extension of the chromosome, as occurs in elongated cells. The parS-proximal regions where SMC is most enriched remained well-aligned in elongated cells, whereas parS-distal regions became significantly less well-aligned, indicating that the two arms were not as tightly cohesed together in these regions.
In Bacillus, DNA-bound SMC, as judged by ChIP analysis, also gradually decreased away from parS, but the arm-arm alignment was maintained more consistently from ori to ter (Gruber and Errington, 2009; Wang et al., 2015). Why does the extent of arm alignment differ in the two organisms and what might cause SMC dissociation from the chromosome? The ATPase activity of the SMC “head” domain is thought to regulate loading as well as ring opening and closing (Minnen et al., 2016). Potentially, if the SMC ring opens, DNA can escape and SMC can dissociate from the chromosome. The intrinsic ATPase rate of Caulobacter SMC is reportedly different from that of Bacillus SMC (Schwartz and Shapiro, 2011). This difference might lead to a different rate of ring opening/closing, resulting in different extents of arm alignment between the two organisms. Alternatively, or in addition, SMC may suffer more frequent conflict with convergently transcribed genes, or be more sensitive to dissociation following such conflicts, leading to less extensive arm-arm interaction.
The extent to which the two arms "zip up" may not matter. If the primary role of SMC-mediated arm-arm interactions is to help enforce the individualization of sister chromosomes immediately after DNA replication, it may only be necessary to ensure that SMC can cohese parS-proximal regions of each chromosome. Indeed, in Caulobacter, ParA-ParB-parS are only required for the segregation of ori-proximal DNA, but not of the distal DNA loci (Badrinarayanan et al., 2015b). Once the ori-proximal DNA is properly segregated, by SMC and ParA-ParB-parS, distal DNA regions follow suit, driven by separate molecular machinery, or more likely without the need of a dedicated system (Badrinarayanan et al., 2015b). In such a case, it may be sufficient to have SMC tether together only a limited region of DNA flanking the parS sites. This model would imply that conflicts between SMC and highly expressed genes oriented toward ori are most detrimental, with respect to chromosome segregation, if they occur in close proximity to parS. Testing this model and further understanding the relationship of SMC and gene expression and its influence on chromosome organization is an important challenge for the future.
ACKNOWLEDGMENTS
We thank Anjana Badrinarayanan, Hugo Brandao, Matt Bush, and Monica Guo for discussion and comments on the manuscript. We thank Lucy Shapiro, Martin Thanbichler, Marie Elliot and Susan Schlimpert for materials. This study was supported by a Royal Society University Research Fellowship (UF140053), a Royal Society Research Grant (RG150448) to T.B.K.L. and a BBSRC grant-in-add (BBS/E/J/000C0683) to the John Innes Centre. M.T.L. is an Investigator of the Howard Hughes Medical Institute.