Extended sister-chromosome catenation leads to massive reorganization of the E. coli genome

Abstract In bacteria, chromosome segregation occurs progressively from the origin to terminus within minutes of replication of each locus. Between replication and segregation, sister loci are held in an apparent cohesive state by topological links. The decatenation activity of topoisomerase IV (Topo IV) is required for segregation of replicated loci, yet little is known about the structuring of the chromosome maintained in a cohesive state. In this work, we investigated chromosome folding in cells with altered decatenation activities. Within minutes after Topo IV inactivation, massive chromosome reorganization occurs, associated with increased in contacts between nearby loci, likely trans-contacts between sister chromatids, and in long-range contacts between the terminus and distant loci. We deciphered the respective roles of Topo III, MatP and MukB when TopoIV activity becomes limiting. Topo III reduces short-range inter-sister contacts suggesting its activity near replication forks. MatP, the terminus macrodomain organizing system, and MukB, the Escherichia coli SMC, promote long-range contacts with the terminus. We propose that the large-scale conformational changes observed under these conditions reveal defective decatenation attempts involving the terminus area. Our results support a model of spatial and temporal partitioning of the tasks required for sister chromosome segregation.


INTRODUCTION
Prokaryotic and eukaryotic chromosomes are not randomly folded, but consist of well-defined structural entities with a complex hierarchical organization. The regulation of this network and its functional interplay with gene expression or other chromosomal metabolic processes such as DNA repair, replication and segregation have been actively investigated in a number of species (1). Improvement in imaging techniques of living cells, as well as the development of genomic approaches such as chromosome conformation capture (3C/Hi-C3) techniques have highlighted a mosaic of intertwined structural features including loops, domains, and compartments (2).
In bacteria, an important part of genome folding relies on the presence of free DNA supercoils. Independent supercoiling domains were first observed in the 1970s through a combination of molecular biology and electron microscopy approaches (3,4). These textbook pictures consist of a succession of large (∼200 kb) DNA plectonemes, or microdomains, that are delimited by topological insulators. Genomic and recombination assays confirmed the existence of these microdomains, but found that they are likely smaller in size (10-50 kb), delimited by stochastic barriers (5,6). These results suggest that microdomains are not static, and that their size and position can be modulated by DNA transactions such as transcription (7) or the binding of proteins to DNA (8). The democratization of 3C/Hi-C methodologies provided the opportunity to further assess the impact of DNA supercoiling on chromosome folding. To date, 3C/Hi-C data revealed the organization of large chromosomal features, such as the alignment of replication arms (9)(10)(11), macrodomains (12) or inter-chromosomal contacts (13). At smaller scale, Hi-C contact maps of every bacterial genome studied to date display self-interacting domains (CIDs, 20-200 kb), whose significance remain unclear (9)(10)(11)(12)(13)(14). The boundaries in-between CIDs correlate with the presence of long, highly expressed genes or genes coding membrane proteins (9,12,15).
Experiments and simulations suggest that supercoils are able to organize bacterial genome because they would condense DNA and promote the disentanglement of topological domains (16) (. DNA supercoiling is under tight homeostatic control by topoisomerases. In Escherichia coli, Topoisomerase I (Topo I), Topoisomerase III (Topo III), DNA Gyrase and Topoisomerase IV (Topo IV) have all well characterized enzymatic activities (17,18), but understanding the roles played by topoisomerases in chromosomal compaction, folding and organization is still a work in progress. DNA Gyrase, that promotes the formation of free supercoils, is the best candidate for regulation of chromosome organization through supercoiling. Hi-C analysis following gyrase inhibition was studied in Caulobacter crescentus, revealing modest changes in chromosome conformation with a slight decrease of 20-to 200-kb contacts that reduced the sharpness and positions of CID boundaries (9). This observation agrees with recombination data showing that topoisomerase alterations reduce supercoiling domain's size (19). The contribution of other topoisomerases in establishing, maintaining, and regulating genome-wide DNA contacts has, so far, not been investigated.
The E. coli SMC complex MukBEF was shown to be linked to DNA supercoiling, with MukBEF defects being suppressed by Topo I mutation (20). The suppression correlates with an excess of negative supercoiling promoted by DNA gyrase. In addition, Hi-C analysis of a MukB mutant showed an important loss of long-range contacts (12), suggesting that MukB promotes contacts between distant pairs of loci. One hypothesis is that MukB could promote DNA loops in vivo (21)(22)(23)(24). For instance, it is possible that MukB, organised in an axial core, extrudes plectonemic microdomains (25). Interestingly, MukB interacts with ParC, the catalytic subunit of Topo IV. Their interaction changes the catalytic properties of Topo IV (26,27) and modulates its localization (28,29). In vitro assays suggest that a MukB -Topo IV interaction promotes DNA compaction by forcing the intramolecular knotting activity of Topo IV (30). Single molecule imaging reveals that a small portion of ParC molecules are associated with MukB in small clusters, (29). The absence of MukB leads to a 2-fold reduction of Topo IV clusters, suggesting that MukB influences the localization of Topo IV in the cell Topo IV is the main bacterial decatenase. It catalyzes the elimination of precatenanes and catenanes formed during replication (31), and is essential for proper chromosome segregation (32)(33)(34)(35). The requirement of PriA, the main replication fork restart protein, for the survival of Topo IV thermosensitive mutants (parE ts and parC ts ), suggests that Topo IV inactivation increases replication fork stalling (36). Nevertheless, chromosome replication in parE ts mutant cells occurs at a similar rate to that seen in wt cells (35). Topo IV activity is highest at the dif site positioned in the middle of the Terminus of replication macrodomain, presumably to solve catenanes (37)(38)(39). However, it also presents hundreds of putative activity sites dispersed on the genome, presumably used for the removal of precatenanes (37).
Here, we investigate, using Hi-C, the effect of different topoisomerases on E. coli chromosome conformation. Inactivation of Topo IV generates the most significant changes in chromosome organization, with a pair of long-range contact patterns, hereafter referred to as butterfly wings, expanding from the terminus macrodomain flanking regions. Imaging experiments confirmed the chromosome reorganization when Topo IV activity is reduced, suggesting that under these conditions distant sister loci segregate when they contact the terminus. Because MatP and MukB also influence the butterfly motif, we propose that it reveals a particular genome folding dedicated to decatenation. We also showed that when Topo IV is deficient, Hi-C reveals intermolecular contacts between sister chromosomes. Topo III limits these contacts suggesting that it removes a significant portion of precatenanes. Together, our results highlight the ability of Hi-C to unveil chromosomal topological features, including inter-chromosomal contacts, to improve our understanding of the mechanisms governing chromosome segregation.

Strains:
The strains used for this study can be found in Supplementary Table 1. All strains are derived from MG1655. All strains were grown in minimal media A (0.26 M KH 2 PO 4 , 0.06 M K 2 HPO 4 , 0.01 M tri sodium citrate, 2 mM MgSO 4 , 0.04 M (NH4) 2 SO 4 ) supplemented with 0.2% of casamino acids and 0.5% of glucose. The strains containing the thermosensitive allele of parC (parC ts ) were grown at 25 • C and shifted to 30 • C for 60 min, which is semipermissive temperature; the strains containing the thermosensitive allele of parE (parE ts ) were grown at 30 • C and shifted at 42 • C for 60 min at most. Time course of parE ts shift was performed by growing the cells at 30 • C before performing a shift for 10, 20, 30, 40, 50 and 60 min. Strains containing the expression plasmid pBAD were cultivated with 0.2% arabinose to induce the expression of the gene under the control of the PBAD promoter the arabinose was present for the entire duration of the culture. For each imaging or Hi-C experiment a wt MG1655 strain was grown in the exact conditions as the mutant and used as reference for subsequent comparisons.
Drugs and antibiotics: Inhibition of Topo I was done with a 5 min treatment with 50 ng/l of Topetecan (40). DL-Serine Hydroxymate (SHX, Sigma CAS number 55779-32-3), an inhibitor of seryl-tRNA synthetase which triggers the stringent response and prevents new rounds of replication, was used at a 10mg/ml working concentration for 90 min. The efficiency of the drug is checked by FACS. Rifampicin was used for 10 min at a 100 ng/l working concentration to inhibit transcription.
Flow cytometry: The amount of DNA was monitored using either a BD fortessa cytometer with a 488 nm argon laser and 515-545 nm emission filter ( Figure 3) or a Beckman cytoflex with a 561 nm laser and a 610/20 filter (Supplementary Figure 1) at a maximum of 5000 event per second. Calibration was done with the samples of the stationary phase Hi-C analysis of the wild type (wt) and parE ts strains. For each panels, symmetric halves of the normalized contact map binned at 5kb with the wt on the bottom and the parE ts after 60min of shift to non-permissive temperature (42 • C) on the top, and the corresponding ratio matrix. (B) Hi-C analysis of the parC ts strain after 60min of shift to non-permissive temperature (30 • C). Genome coordinates are indicated by the x and y axes. Macrodomains are represented by light green (ori), dark green (right), red (ter), blue (left), gray (NR/NL). matS sites are represented as grey bars. For the normalized contact maps, the color scale of the frequency of contacts between two regions of the genome is indicated below (arbitrary units), from white (rare contacts) to dark red (frequent contacts). For the ratio matrices, a decrease or increase in contacts in the mutant cells compared with the control is represented with a blue or red color, respectively. White indicates no differences between the two conditions. (C) Schematic representation of the loci tagged with parS/ParB fluorescent reporters in the right replichore and used for the experiments presented in (D) to (F).  of the WT, parCts at 25 • C and parEts at 30 • C. For all samples, ∼10 8 cells were fixed in 70% EtOH, washed, marked with propidium iodide (2-20 g/ml), washed again, and resuspended in sterile 1× PBS pH 7.2 prior to analysis. FCSalyzer software (https://sourceforge.net/projects/fcsalyzer/) was used for data analysis.
Hi-C libraries: Hi-C libraries were generated as recently described in (41). 30 ml of culture was grown in Minimal Medium A supplemented with casaminoacids and glucose until OD 600 nm ∼0.2. Protein-DNA interactions were crosslinked by the addition of 37% formaldehyde (3% final concentration) for 30 min at room temperature with gentle agitation. Crosslinking was quenched with 2.5 M glycine (0.4 M final concentration) for 20 min at room temperature with gentle agitation. Fixed cells were then collected by centrifugation (4000 × g, 10 min 4 • C), washed once in 1× PBS and snap frozen on dry ice and stored at -80 • C until use. To proceed to the digestion of the cells, pellets were thawed on ice and resuspended in 1.2 ml of 1× TE+ complete protease inhibitor (EDTA-free, Roche). Cells were transferred to a VK05 tubes containing glass spreads, and then lysed with precellys (V750; 5 × 30 s). Uncross linked proteins were then solubilized by incubating the lysed cells with 10% SDS (0.5% final) for 10 min at room temperature. 1 ml of lysed cells was then added to the digestion mix (3 ml of dH 2 O, 0.5 ml of 10× NEBuffer1, 0.5 ml of 10% Triton-X 100 and 1000U HpaII) and incubated 3 h at 37 • C. Digested cells were pelleted by centrifugation and resuspended in 398 l of water before being added to the biotinylation mix (10× ligation buffer without ATP; dAGTtp 3.3 mM; biotine-14-dCtp 0.4 mM; 50U klenow (NEB5U/l)) and incubated 45 min at 37 • C with agitation. Ligation mix (10× ligation buffer without ATP; BSA 10 mg/ml; ATP 100 mM; 250 U ThermoFisher T4 DNA ligase) was then added to the biotinylated DNA and the mix was incubated 3 h at room temperature with gentle agitation. Protein-DNA complexes were then reverse crosslinked by adding the 20 l of EDTA 0.5 M, 80 l of SDS 10% and and 100 l of proteinase K 20 mg/ml and incubating at 65 • C overnight. DNA extraction is made by phenol-chlorophorm and precipitation with ethanol. DNA is washed with 70% ethanol, resuspended in 130 l of TE and then treated with 20 mg/ml RNAse for 30 min at 37 • C. Samples were sonicated using a Covaris S220 instrument to obtain fragments between 300 and 500 bp, then purified by AMPure XP beads. The Illumina process was performed according to manufacturer recommendation, with are 12 cycles of amplification. The size of the DNA fragment in the libraries are checked on TAE 1% agarose gel and subjected to paired-end sequencing on an Illumina sequencer (NextSeq500-550-75 cycles). Generation of contact maps: Generation of contact maps was done with E. coli analysis pipeline developed by Axel Cournac (https://github.com/axelcournac/EColi analysis). Briefly, reads recovered from the sequencing were aligned with bowtie2 on the reference E. coli genome (NC 000913), the two ends were merged and the reads were filtered to have a mapping quality strictly >30. Then each read was assigned to a HpaII restriction fragment (fragment attribution), >80% of the reads on average are conserved as informative reads. The contact matrices files are then generated, binned at 2, 5 or 10kb and normalized through the sequential component normalization procedure (SCN (42)).
Ratio matrices and ratio plots: In order to visualize the differences between two contact maps, their ratio is calculated as described in (12). Briefly, contacts made in the mutant map are divided by the contacts made in the control map for each loci. Increase or decrease of contact in the mutant compared to the control will appear red or blue respectively and no changes will appear white. To visualize the contact signal at a smaller scale we use the ratio plot representation, as described by (12). Briefly, the ratio-plots summarize the differences of contacts between two conditions made by each bin with its neighboring bins, regardless of their orientation. For each condition, the average contacts (AC) made by a bin along the genome with bins upstream and downstream at increasing distances (from 5 kb to 1000 kb or to 2000 kb, in 5 kb increments), was computed The log 2 (ACmutant/ACwt) was then displayed as a heatmap.
Viability assay: For each strains, cells were serially diluted in LB (10 -1 to 10 -8 ) and 1l of each dilution is deposed on LB agar plates. Plates were incubated overnight at 30 • C.
DNA density analysis: Cells were grown using the same conditions as Hi-C. 1 ml of culture of a control sample at 30 • C and a sample after a 60 min shift at 42 • C for each strain were fixed every 20 min (t0-t60), using an equal volume of 1× PFA (37% paraformaldehyde + glutaraldehyde + 1× PBS). DAPI (0.5 g/ml) is added and the mixture is incubated 10 min in the dark. Cells were then washed and deposited on microscopy slides containing a freshly made agarose pas (1× PBS + 1% agarose). Samples were observed with a Spinning disk (YokogaWa) W1 system on a Zeiss inverted confocal microscope with a 63×, phase objective. To avoid bleaching, each field of view was only observed in the phase contrast channel before acquisition. Using ImageJ, the density of DAPI fluorescence in each nucleoid was calculated as a proxy for nucleoid density.
Loci positioning analysis: Cells were grown akin to Hi-C samples (above). For the experiments with the parE ts allele, 1ml of culture of a control at 30 • C and a sample after shift at 42 • C for each strain, were fixed every 20 min (t0-t60), using an equal volume of fixation medium (3.7% formaldehyde, 0.006% glutaraldehyde, 1× PBS). Cells were pelleted and resuspended in 100 l of fresh Minimal medium A. Three drops of 2l are deposited on glass slides containing a freshly made agarose pad (1× PBS + 1% agarose). Samples were observed with a Spinning disk (Yokogawa) W1 system mounted on a Zeiss inverted confocal microscope at 630-fold magnification. The position of foci in the cell in each condition was analyzed with the MicrobeJ (https: //www.microbej.com/, (43)) and ObjectJ plugins of ImageJ https://sils.fnwi.uva.nl/bcb/objectj/. Two color localization was performed in wt and parCts strains with 7 couples of tags forms with combination of gusC parS pMT1 (ter), aidB parS P1 locus (ori), ygeB parS P1 (left), yfgE parS pMT1 (left 2), yoaC parS pMT1, gusC parS P1, crl parS P1 and cynX parS pMT1 tags (44). Cells were grown at 25 • C in Minimal medium A supplemented with casaminoacids and glucose until culture reached OD = 0.1 and then cells were shifted to 30 • C for 1 h. Cells were observed live on agarose pad on a thermo-controlled stage with a Zeiss inverted epifluorescence microscope equipped with led illumination. Inter focal distance distributions correspond to the average of 3 independant experiments. An average of 600 cells were analyzed per strain.

Topo IV inactivation induces large-scale chromosome conformation changes in enterobacteria
To investigate the respective contribution of each topoisomerase to genome-wide bacterial chromosome organization, we applied the capture of chromosome conformation (Hi-C) methods (41,45) (Material and Methods) to exponentially growing E. coli cells mutated or inactivated for each of the four topoisomerases. The ratio between wt and mutant/depleted cells contact maps were plotted (Materials and Methods) (12).
The inactivation of Topo IV using thermosensitive alleles of either the parC (parC ts ) or parE (parE ts ) subunits (33) resulted in significant changes in the global chromosomal architecture ( Figure 1A and B). Chromosome folding changes were observed upon a 1h shift at 30 or 42 • C respectively for the parC ts and the parE ts mutants. At these temperatures viability and DNA content were modestly affected (Supplementary Figure 1). We observed for both mutants changes in mid-to long-range DNA contacts in three regions of the genome: First, an increase in shortto medium-range contacts (50-300 kb) from oriC to positions located at ∼1 Mb and ∼2 Mb on the right and left replichores, respectively. Second, a strong reduction of midrange contacts in the terminus region (coordinates 1315-1830 kb, Figure 1A, B, Supplementary Figure 2) accompanied by an increase of very short-range contacts (<50 kb, 10 bins). This particular terminus pattern also exhibits a strong (CID-like) border around the dif site. Finally, a peculiar butterfly-like signal with two wings of long-range contacts emerges from the flanking regions of the Ter, extending towards oriC. These changes in chromosome folding are temperature dependent, reproducible and rescued by a plasmid expressing ParE into the parE ts mutant cells (Supplementary Figure 3A-C). To broaden these observations, we applied Hi-C to another gamma-proteobacterium, Salmonella typhimurium for which thermosensitive Topo IV mutations are available (46,47). Similar conformational changes were observed (Supplementary Figure 3D and E).
To assess the specificity of these signals, we disrupted the function of the three other topoisomerases of E. coli. To reduce Topo I activity, we used the topA31 allele (48), a Nucleic Acids Research, 2022, Vol. 50, No. 5 2641 topB deletion mutant was used to study Topo III (49), and DNA gyrase was studied using a thermosensitive mutant of gyrB (50). In contrast to Topo IV alterations, inactivation of Topo I and gyrase activity resulted in a decrease of shortrange contacts. These observations are consistent with the involvement of supercoiling in the definition of CIDs (9). Alteration in Topo III resulted in a slight gain in short-range contacts, with no other significant changes in the chromosome conformation of these cells detected (Supplementary Figure 3F-H). Altogether, our observations suggested that the genome folding changes observed in the parE ts and parC ts strains were genuine consequences of the alteration of Topo IV activities.

Hi-C features observed in Topo IVts mutant cells are concomitant with nucleoid reorganization
We complemented our Hi-C approaches with fluorescence imaging of individual genomic loci. We selected parS/ParB systems that have been extensively validated in E. coli. Unfortunately, the very poor fluorescence yield of the lowly expressed ParB P1-CFP tag at 42 • C did not allow us to perform experiments with the parE ts strain. We focused therefore on experiments with the parC ts strain 1 h upon shift to 30 • C. Imaging further supported that nucleoid organization and compaction undergoes drastic changes following parC ts inactivation. The number and positioning of different parS/ParB chromosomal tags were observed ( Figure  1C-J and Supplementary Figure 4). In the WT cells, upon shift to 30 • C, the two copies of loci close to oriC preferentially localized around the 1 4 and 3 4 positions and loci located on the side of the chromosome arms located at the 1 4 and 60% positions following the canonical L-R-L-R pattern. Loci close to the terminus localized to the new pole in newborn cells and at mid-cell in older cells. In the parC ts strain, the number of cells with two copies of each fluorescent tag decreased and the cell age at which this second copy appears was delayed ( Figure 1E and I). Since replication is only modestly perturbed by Topo IV alteration in the parC ts strain at 30 • C (Supplementary Figure 1B), this might reflect segregation defects. Localization of the tags was also perturbed, with most oriC, left and right foci localizing closer to mid-cell in between the 1 4 and 3 4 positions. Simultaneously ter foci frequently migrated away from midcell. These observations suggest that global chromosome conformation was affected and / or cell cycle course was perturbed.
Simultaneous visualization of two fluorescent tags allowed us to measure interfocal distances in between distant genomic loci, a proxy for genome compaction ( Figure 1F and J). First, we analyzed the distances separating close loci (102 and 197 kb). In the parC ts strain at 30 • C, we measured a significant shift of the distribution of the distances between the crl and cynX tags ( Figure 1F) and the yoaC and gusC tags ( Figure 1J) toward short distances. Since the distances of both close and far couples were reduced we postulate that it reflects both inter and intra sister chromatid compaction. These observations agree with the increased frequency of short-to medium-range contacts observed with Hi-C experiments.

Kinetic analysis of genome folding changes imposed by crippling of Topo IV
We performed a kinetic analysis of chromosome reorganization using Hi-C and imaging, of chromosome folding after Topo IV inactivation. Hi-C revealed progressive increase in short-to mid-range contacts, which became prevalent along chromosome arms after 40 minutes (Figure 2A). Concomitantly the butterfly wings expanded from positions 1 and 2 Mb. The terminus region behaved differently, showing a persistent lack of short-range interactions over the entire kinetics (Figure 2A). DAPI staining revealed that the nucleoid reorganized with a comparable kinetics. In the first 20 min, we observed a global expansion of the DAPI signal leading to the apparent of fusion of the different DNA masses present in a single cell ( Figure 2B). At later timepoint the DNA density increased in the nucleoid and it relocalized to mid-cell leaving empty large polar regions. In a 60 min period at 42 • C the average DAPI staining density increased by 1.5-fold compared to wt ( Figure 2C). Altogether, these time course experiments show that while nucleoid progressively reorganize upon inactivation of Topo IV, specific contacts accumulate at discrete positions surrounding the ter domain. Therefore that Topo IV plays a role in suppressing the formation of short-to medium-range and longrange contacts within or between sister chromosomes.

The absence of Topo IV activity results in the co-localization of distant loci with the terminus of the chromosome
We characterized further the two butterfly-wing stripes emanating at a 45 • angle from the edges of the terminus region in the ratio contact maps between Topo IV and wt ( Figure  3A  These signals correspond to an enrichment in medium and long-range contacts by ∼200 kb regions that in one direction span all the way up to oriC, while on the other direction, are blocked by dif, which represents a barrier preventing contacts across each halves of the terminus ( Figure 3A and B). Measuring the contact frequency of individual bins with the rest of the genome, we investigated how each 5kb bin interacts with the terminus region and therefore contribute to the butterfly signal ( Figure 3C and D). As illustrated for two representative regions (position 3010 and 4440 kb), a small (≈1.5-fold), but significant, contact enrichment with the dif zone was detected several megabases away from dif. This enrichment involved several (4 to10) bins symmetrically localized around dif. The contacts between the dif zone and the 3010 kb or 4440 kb regions present a comparable frequency, suggesting that they do not obey to genomic distance law. We observed, on zoomed panels, that the bin containing dif (green dots on Figure 3D) was not involved in the butterfly wing contacts ( Figure 3D). To test this exclusion of dif, we measured the sum of contacts made by a 100 kb window 2 Mb away from dif ( sign of Figure 3A) with the 1.2 Mb region containing dif (Figure 3E). This window displayed enriched contacts with most bins from the terminus region, but not with dif itself. Their distance independent frequency and the lack of asymmetry at dif strongly suggests that butterfly wings correspond to long-range 3D contacts rather than slithering events. To confirm the propagation of these very long-range contacts at the single cell level, the localization of fluorescently labelled loci positioned in the left replichore (ygeB locus) and terminus region (gusC locus) were observed in cells carrying the parC ts allele ( Figure 3F, G), that displayed butterfly wings at low temperature (30 • C) ( Figure 1C). In these conditions, the left and the ter loci colocalized (interfocal distance < 250 nm, arrows on Figure 3G) more frequently than for the wt. About 18% of the parCts cells show a colocalization of the left and ter foci compared to 9% in the wt. Similar changes in the distribution of distances between ori and ter or right and ter foci were observed (Supplementary Figure  4). Interestingly, the distances between left 2 and right loci did not change; it suggests that it does not correspond to a global compaction but rather a particular genome folding involving the terminus (Supplementary Figure 4). Colocalized ter and left foci were observed at various positions in the cell but are slightly more frequent near mid-cell ( Figure  3I and J). Time-lapse experiments showed that left and ter foci colocalization was transient and involved a rapid migration of the left focus toward an immobile ter focus (Figure 3K). As illustrated on the representative example the left-ter contact did not induce a duplication of the ter or left foci ( Figure 3K). The rapid photobleaching of the CFP tag did not permit to investigate quantitatively these movements and to attribute them to a particular step of the cell cycle of the parCts strain grown at 30 • C. Imaging confirms the existence of long-range contacts between the terminus and distant regions of the chromosome in absence of Topo IV activity. The detection of ori, left and right occasional colocalization events with the terminus in wt cells ( Figure  3I and Supplementary Figure 4) suggests that the peculiar genome folding events detected in the absence of Topo IV activity might also exist in some wt cells. To evaluate a putative role of such contact in chromosome segregation, we measured the distance between sister left and sister ori foci as a function to their distance with the ter focus. In wt cells the shortest distances between sister left and sister ori foci were observed at mid-cell when these foci were the closest from the ter focus (R 2 = 0.18 and 0.16, respectively for left and ori) (Supplementary Figure 5). Although more left and ori foci colocalized with the terminus in the parC ts strain, the localization pattern of newly duplicated sister ori and sister left foci was significantly perturbed (R 2 = 0.08 and 0.006). These observations suggest that duplication of genomic loci might be synchronized or triggered by a 'contact' with the terminus near mid-cell; the butterfly wings pattern ( Figure  3A-D) and the associated extra colocalization with the terminus ( Figure 3G and H) may therefore correspond to an accumulation of unsuccessful duplication attempts.

parE ts inactivation increases inter-chromosomal contacts
The ability of conventional Hi-C to capture intrachromosomal (cis) contacts is well established. However, its ability to reveal trans contacts between sister-chromatids remains limited given the difficulty to distinguish both homologous molecules without incorporation of modified bases (51). Since Topo IV is mainly known for its involvement in the removal of inter-chromosomal links (or catenanes), presumably between allelic (or near-allelic) loci, we sought to test whether the emergence of prominent Hi-C signals spanning throughout the parE ts contact map could result from trans contacts. We therefore sought to distinguish the contribution of intra-versus inter-molecular contacts to these patterns by determining the impact of DNA replication in absence of Topo IV activity. To do this, we blocked replication initiation by treating wt and parE ts cells with the amino acid analog dl-serine hydroxamate (52). Upon SHX addition the distribution of DNA in wt and parE ts cells shifted from a single normal distribution centered around 3N genomes to two peaks respectively centered at 2N and 4N, this suggests that SHX blocked replication initiation but allowed ongoing replication rounds to finish ( Figure 4A). Following SHX treatment, non-replicating cells were then shifted to 42 • C for 1 h. Hi-C ratio maps of replicating versus non-replicating wt cells revealed a significant decrease in mid-range contacts along the entire chromosome in the absence of replication ( Figure  4B). Non-replicating parE ts cells did not show increasing mid-range contacts along the diagonal and did not show the butterfly wings pattern observed for replicating cells ( Figure 4C and Supplementary Figure 6A and B).
Since the stringent response mediated by SHX also reduces transcription of many genes (53), we tested whether absence of the characteristic pattern in SHX-treated cells could result from a poor transcription. Asynchronously growing parE ts and wt cells were therefore treated with the antibiotic rifampicin to inhibit transcription. The partition of the chromosome into three regions in parE ts cells remained visible upon inactivation of Topo IV at 42 • C (i.e. the butterfly wings, a higher mid-range contacts in the oriC proximal region, and lower mid-range contacts in the ter). However, their intensity was reduced. At this stage, it is not possible to distinguish between specific effects of transcription on the pattern mediated by Topo IV alteration or a more trivial alteration of Hi-C matrices' quality by rifampicin (9), (Supplementary Figure 6C, D). Taken together, these results suggest that Topo IV inactivation induces characteristic Hi-C patterns in replicating and transcribing cells, presumably because they correspond to interchromosome contacts.

Butterfly wings signals are linked to chromosome topology
We then wondered whether the circular nature of the E. coli chromosome could lead to topological constraints on the terminus region that would translate into specific contact patterns in the absence of Topo IV. To test this, we performed Hi-C experiments in strains carrying a genome linearized at a tos site, inserted near dif, thanks to the bacteriophage N15 telomerase (54). In this strain, Topo IV is still required for growth (54). However, it is no longer active at dif, and the filamentation phenotype of the matP mutant is now rescued (37). These observations hinted that topological tension resulting from Topo IV inactivation freely diffuses when the N15 telomerase linearizes the duplicated chromosome ( Figure 4D). In the wt strain, linearization of the chromosome near dif resulted in a notable redistribution of long-range contacts towards short and mid-range contacts ( Figure 4E and Supplementary Figure 6E). Upon The replication starts at the oriC site (green box) and progress in a bidirectional manner toward the dif site (red box). We hypothesis that precatenanes are formed during the replication of the linear chromosome in the same manner as in the circular one. Sister chromatids are represented by double helices in red (sister A) and yellow (sister B). Before chromosome segregation, the protelomerase TelN linearize the sister chromosomes at the tos site which could possibly allow for diffusion and resolution of the majority of the precatenanes. Ratio of normalized contacts map binned at 5 kb of (E) linear wt vs circular WT, (F) linear parE ts vs linear wt and (G) linear parE ts linear versus parE ts circular. Macrodomains and interesting positions of the genome are indicated above the plot akin to Figure 1. The y axis indicates the genomic coordinates. A decrease or increase in contacts in the mutant cells compared with the control is represented with a blue or red colour, respectively. White indicates no differences between the two conditions. inactivation of parE ts , the Hi-C contact maps of the linear chromosome displayed a significant loss of short-to mid-range contacts, no butterfly wings and no terminus pattern compared to circular wt chromosomes ( Figure 4F and Supplementary Figure 6F). The comparison between parE ts strains carrying either a linear or a circular genome further revealed a decrease in short/medium contacts along linear chromosomes, in sharp contrast to wt conditions, and compatible with a reduction in the number or density of precatenation links ( Figure 4G). This suggests that the linearization of the strain suppresses the accumulation of interminglements between the sister chromatids. Altogether, these observations strongly support the hypothesis that the enhanced mid-range contacts along chromosome arms, as well as the butterfly wings signals, form in response to a failure of Topo IV to remove inter-chromosomal links.

Topo III activity partially rescues Topo IV deficiencies on chromosome conformation
Topo III (topB) is a type I topoisomerase whose role during the normal cell cycle remains unclear. Both the modest chromosome conformation changes ( Supplementary Figure 3G) and normal viability observed in the mutant (Figure 5A) confirm that in laboratory growth conditions its role is limited. However, part of its activity is revealed when Topo IV is impaired, as Topo III overexpression rescues parE ts growth defects (Lee et al., 2019;Nurse et al., 2003; Figure 5A), whereas topB deletion reduces the viability of parE ts strains at intermediary temperatures ( Figure 5A). When parE ts was inactivated in a topB deletion strain, the Hi-C pattern observed in parE ts mutant cells at the nonpermissive temperature was strengthened ( Figure 5B), with mid-range contacts increase, and stronger butterfly and terminus patterns ( Figure 5B). In the topB parE ts strain, long- range contacts were observed between distant regions and a large central region of the ter macrodomain ( Figure 5C). Their frequency was higher ( Figure 5C) compared to the single parE ts strain ( Figure 3C). As observed for the parE ts strain the dif bin appeared to make less long-range contacts than its neighbors ( Figure 5C). Interestingly, parE ts features were visible in the contact maps at 30 • C in the absence of Topo III (Supplementary Figure 7A), suggesting that at this temperature Topo IV is already partially inactivated, but that the organization defects are suppressed by Topo III. This is in good agreement with the CFU data ( Figure  5A). Ratio plot and 4C-like analysis also revealed important changes in the distribution of short and mid-range contacts outside of the terminus region when both Topo IV and Topo III are inactivated. When plotting the ratio map between the parE ts topB matrix and topB or parE ts matrices we characterized the contribution of Topo III and Topo IV. Butterfly wings, and mid-range contacts appeared in the parE ts topB compared to topB matrices (Supplementary Figure 7B-D). On the other hand, comparing the parE ts topB matrix with parE ts highlighted mostly an increase in short-range contacts ( Figure 5D). The overexpression of Topo III rescued the viability of the parE ts strain at non-permissive temperatures (Nurse et al., 2003, Figure 5A). The Hi-C contact map of a parE ts strain overexpressing Topo III at non-permissive temperature displayed a reduction in short-range contacts along the diagonal, but little changes regarding butterfly wings and mid-range contacts ( Figure 5E and F and Supplementary Figure 7C and 7D). These observations demonstrate that short-range contacts are mediated by a topological structure that can be removed by Topo III, most likely precatenanes with single strand regions near the replication fork. Since butterfly wings, mid-range contact and terminus insulation are not suppressed by Topo III overexpression, they appear as a direct consequences of Topo IV inactivation and might therefore reflect the catenation of fully replicated molecules, a substrate impossible to unlink for Topo III ( Figure 5G).

The position of butterfly wings is determined by MatP/matS, the Ter macrodomain organizer.
The peculiar positioning of the butterfly wings at the edges of the Ter macrodomain ( Figure 3A) prompted us to assess the influence of MatP on these structures. In the absence of MatP and at parE ts non-permissive temperature, the butterfly wings were replaced by a large region displaying increased long-range contacts (up to Mb distances), covering the entire terminus macrodomain ( Figure 6A). The ratio plot of matP parE ts and matP contact maps further show an enrichment in mid-range contacts within the terminus macrodomain in absence of MatP ( Supplementary Figure 8A-D). This observation confirms that the features revealed by Topo IV inactivation are structurally connected, with MatP being responsible for the positioning of the discrete boundaries limiting the entry of putative precatenation links into the terminus macrodomain. These borders define the characteristic butterfly structure and favors the emergence of the terminus pattern.
Based on these observations, we hypothesized that displacing the terminus macrodomain should reorganize the Topo IV-dependent structural features. We used bacteriophage site specific recombination to generate an inversion that moved the region associated with the base of the right butterfly wing (coordinate: 1. 342 702 Mb) to a region now 540 kb upstream the right replichore (coordinate 0. 806 549 Mb) (55) (Figure 6B, Supplementary Figure 8E and F). This inversion resulted in the repositioning of five matS sites (matS 1-4) to a region upstream within the right replichore, which in turn resulted in the generation of a large chromosomal region (coordinates 0.8-1.3 Mb) devoid of matS sites and flanked by two matS regions. We observed a shift in the position of the right butterfly wing that coincided with the new position of the inverted matS sites. The butterfly wing was less precise, and long-range contacts appeared to extend from matS5 to dif. The positioning and strength of the second butterfly signal was not affected by the inversion, suggesting that the two butterfly wings are not functionally interlinked ( Figure 6B, Supplementary Figure 8E and F). Imaging experiments performed in the parC ts matP strain showed a loss of colocalization between the left and ter fluorescent tags ( Figure 6C and D) compared to the single parC ts ( Figure 3G and H). Although the left and ter localization remained affected by Topo IV alteration in the parC ts matP strain (Supplementary Figure 9), the frequency of leftter colocalization was comparable to that of the wt strain (compare Figures 3H and 6C). These observations suggest that the butterfly wings Hi-C signal corresponds to genome reorganizations involving the terminus and orchestrated by MatP binding to matS.

MukB controls the shape of the butterfly wings
The presence of MatP in the terminus macrodomain inhibits MukB activity in wt cells . In the absence of MatP, MukB is able to access the terminus region and has been shown to both accelerate the segregation of terminus loci (56) and change terminus conformation (12). MukB also interacts with Topo IV and modulate its activity (26,27). We therefore tested whether MukB affects the patterns observed when Topo IV is inactive. A mukB deletion in itself did not create butterfly pattern (12), suggesting that disrupting the ParC-MukB interaction does not alter Topo IV activity in the same way as parE or parC inactivation does. When combined with parE ts , the mukB mutation did not abolish the formation of butterfly wings ( Figure 6E). However, their characteristics in the double mutant differ from those in the single parE ts mutant. Comparison of the parE ts mukB mutant with the single parE ts mutant revealed that the basal portion of the butterfly wings were reinforced in the absence of mukB (i.e. higher frequency of contacts) and their length is reduced (Supplementary Figure 8G). In addition, we observed an increase of contacts between all regions within the butterfly, so that they now resemble a self-interacting domain (Supplementary Figure 8G), rather than a stripe ( Figure 6E). Scalograms, i.e. the aggregation of contacts over various scales from each bin (12), clearly show the variations in wing shapes and sizes in the mutant strains (parE ts , parE ts inverted, parE ts mukB and parE ts matP) (Figure 6F). In the parE ts strain, butterfly wings extend from the terminus to the origin of the chromosome. In the absence of MatP or in the inverted strain, they nearly disappear. In ad-  ), parE ts matP versus wt, parE ts mukB vs wt. The y axis indicates the contact distances in kB. A decrease or increase in contacts in the mutant cells compared with the control is represented with a blue or red colour, respectively. White indicates no differences between the two conditions. dition, MatP prevents the formation of mid-range contacts in the terminus region ( Figure 6F). Finally, in the absence of MukB the butterfly wings abruptly stop ∼1 Mb from dif on both replichores. The absence of MukB resulted in the exclusion of the oriC region from the butterfly structure. Results obtained in the parE ts mukB mutant, however, should be taken with precautions since the DNA content and replication cycle of this strain were significantly perturbed (Supplementary Figure 8H). Moreover, the poor viability of the parCts mukB mutant did not allow us to complement Hi-C observations with fluorescent microscopy.
The increase in short range contacts detected along chromosome arms when Topo IV is inactivated (presumably corresponding to the accumulation of precatenanes) did not change in the absence of matP or mukB ( Figure 6F). This suggests that Topo IV partners do not significantly impact precatenanes dynamics. Therefore, MatP is involved in the positioning of the butterfly pattern, while MukB might determine the wing length and contact density ( Figure 6F). Moreover, MatP itself and/or the resulting MukB relocalization might prevents the progression of precatenanes into the terminus ( Figure 6F).

Alteration of each Topoisomerase induces specific chromosome conformations
Although DNA topology plays an important role in bacterial chromosome folding and compaction, the consequence of topoisomerase alteration on these processes are not yet fully understood. We took the advantage of Hi-C protocol improvements in bacteria to analyze chromosome conformation upon alteration of each of the four Topoisomerase of E. coli. Inhibition of Gyrase activity results in a strong reduction of short-range contacts over the entire genome. This agrees with previous report (9) and confirms that supercoiling homeostasis is an important driver of CID organization in bacterial genomes. Surprisingly, because they have opposite catalytic activities, Topo I inhibition also reduces short-range contacts. Topo III and Topo IV inhibitions have the opposite effects, as they both increase shortrange contacts: lightly and homogeneously for Topo III, but much more significantly and with a robust patterning for Topo IV. We investigated in details the nature and determinants of this patterning.

New chromosome contacts emerging from Topo IV inhibition are intermolecular contacts
Topo IV inhibition induces an increase of chromosome contacts in the 50-200 kb range chromosome arms with the exception of a large terminus region that behaves differently. We observed that these new contacts are only established in replicating cells carrying circular chromosomes. Moreover, the activity of Topo III, which appears at very short range (0-50 kb), limits their amount. Altogether, these results demonstrate that short -mid range contacts detected when Topo IV is inactivated are produced by precatenanes ( Figure 7A). Their amplitude, 200-300 kb, can only be explained if catenated sister chromatids slides or fold in 3D on top of each other on a 200-300 kb window. By contrast, Topo III only modulated short scale contacts, which agrees with the idea that soon after replication sister loci are well aligned (no more than 50 kb offset). In the absence of Topo IV some of these precatenane links persist away from the replication fork, out of reach of Topo III action even when overexpressed. Therefore, a few tens of second after their replication, catenane links cannot reach the Topo III activity zone, and might be entrapped by topological barriers ( Figure 7A). The comparison of wild-type replicating and non-replicating cells Hi-C matrices ( Figure 4B) revealed an increase of mid-range contacts when cells replicates. This suggests that during a regular cell cycle Topo IV and Topo III sporadically let few catenation links all over the genome that will contribute to sister chromatid cohesion (32,34). Future Hi-C and sisterC (51) experiments performed on cells with synchronized cell cycle should give important insights on the nature of these inter-sister contacts and their molecular determinants.

Very long-range contacts emerged as a butterfly pattern at the terminus
Butterfly wings observed in the absence of Topo IV involve contact between the dif area (200 kb surrounding dif) and the rest of the chromosome. In wt cells the dif site, in the terminus region, is the major Topo IV activity site, this suggests that a large number of catenation links can be removed at this locus (37,38); perhaps because they accumulate in the replication fork trap or because Topo IV is targeted there by partners such as XerCD, FtsK or MukB (57)(58)(59). It is tempting to speculate that butterfly wings' signal may reflect a malfunctioning of a decatenation machinery active around the dif site in wt cells. The butterfly pattern is an atypical Hi-C feature; these contacts do not follow genomic distance law and are not impacted by cis barriers (Figure 3). The density of contacts inside butterfly wings is dependent on the topological status of the cell (Topo III effect) and on MukB and their position determined by MatP/matS (Figure 6). The butterfly wings signal correlates with an extra colocalization between the terminus region and distant loci in the ori, left or right regions; 10-23% in parC ts and 4-14% in wt cells (Figure 3, supplementary Figure 4). The colocalization events are transient ( Figure 3K) and dependent on the presence of MatP ( Figure 6). Even if ori or left foci colocalization events with the terminus were rare in wt cells, we observed that closest sister ori and left foci are localized at mid-cell and in the proximity of a terminus focus, suggesting that in some cell cycle duplication might be correlated with a contact with the terminus (Supplementary Figure 5). In this model, the terminus region may provide Topo IV to remove persistent catenation links associated to distant regions of the chromosome. One possible explanation for the butterfly wings' signal could be that because Topo IV is inefficient, contacts with the terminus fail to provide decatenation and are therefore prolonged ( Figure 7B). Supporting this model, we observed that butterfly wings were more robust when Topo III was inhibited (i.e. when more catenanes link sister chromatids) and less robust when Topo III was overexpressed.
Several mechanisms might drive butterfly wings' contacts. One possibility is that a protein marking catenation Figure 7. Topo IV alteration reveals that sister chromatid management involves long range genome reorganisations. (A) Graphic representation of putative sister chromatid precatenanes organization, dynamics and homeostasis based on Hi-C data of Topo IV and Topo III mutants. Two pairs of sister loci were represented in pink and yellow respectively. RF stands for replication forks. (B) Graphic representation of the terminus region and the chromosome folding events that take place around it when decatenation is impaired. This region may function as a hub where unresolved entanglements of the two sisters formed during chromosome replication enter by a yet unknown mechanism to be separated by Topo IV. The low decatenation capacity in the parE ts and parC ts strains increase the persistence/recurrence of these events and their detection with Hi-C and imaging. MukB defines the maximum distance of loci able to contact the decatenation hub.
links is involved in protein-protein contacts between the terminus and other regions of the chromosome. A DNA binding protein marking positive supercoils, GapR, has been identified in C. crescentus (60). A yet unknown equivalent of GapR for catenation links (catenation crossing are topologically similar to positive supercoils) might exist in E. coli. Future experiments will be require to test if YejK which interacts with Topo IV and binds DNA (61) or another catenane binding factor, could play such a role. Another possible explanation is that regions of the chromosome presenting excess of catenation links are excluded from the bulk of the genome. Since the absence of MukB changes butterfly wing density, one hypothesis is that catenane links cannot be extruded as plectonemic loops. As MukB does not bind to the terminus neither (25), both catenation links and terminus may be spontaneously excluded from the chromosome bulk covered by MukB and merge in a particular territory of the nucleoid.

Chromosome Conformation Capture proposes a new read out of topological constraints
Most of our knowledge on the control of topological homeostasis by topoisomerases comes from plasmid DNA topology analyses by 1D or 2D gel electrophoresis or electron or AFM microscopy (62). Recently, molecular genomics methods were implemented to map topoisomerase activities or binding on chromosome (63). Hi-C is a popular method to study chromosome conformation, and a derivative to track inter sister contacts was recently designed in eukaryotes (51). The use of loxP recombination to reveal proximity of sisters (34) or their catenation (64) is also a powerful tool to decipher local chromosome topology, with a genome-wide loxP recombination based assay recently adapted to survey sister proximity in V. cholera (65). Our results suggest that Hi-C can also offer a read-out of topological constraints. Coupling all these approaches together, along with imaging, will refine our understanding of chromosome conformation dynamics during the bacteria cell cycle and particularly the contribution of inter-sister contacts.

DATA AVAILABILITY
Sample description and raw sequences are accessible on SRA database through the following accession number: PRJNA730396.