|Home | About | Journals | Submit | Contact Us | Français|
Duplication and segregation of chromosomes involves dynamic reorganization of their internal structure by conserved architectural proteins, including the structural maintenance of chromosomes (SMC) complexes cohesin and condensin. Despite active investigation of the roles of these factors, a genome‐wide view of dynamic chromosome architecture at both small and large scale during cell division is still missing. Here, we report the first comprehensive 4D analysis of the higher‐order organization of the Saccharomyces cerevisiae genome throughout the cell cycle and investigate the roles of SMC complexes in controlling structural transitions. During replication, cohesion establishment promotes numerous long‐range intra‐chromosomal contacts and correlates with the individualization of chromosomes, which culminates at metaphase. In anaphase, mitotic chromosomes are abruptly reorganized depending on mechanical forces exerted by the mitotic spindle. Formation of a condensin‐dependent loop bridging the centromere cluster with the rDNA loci suggests that condensin‐mediated forces may also directly facilitate segregation. This work therefore comprehensively recapitulates cell cycle‐dependent chromosome dynamics in a unicellular eukaryote, but also unveils new features of chromosome structural reorganization during highly conserved stages of cell division.
The chromosomes of prokaryotes and eukaryotes display multiple levels of hierarchical organization, whose dynamic changes influence or regulate metabolic processes including gene expression and DNA replication and repair (Taddei & Gasser, 2012; Wang et al, 2013; Dekker & Mirny, 2016). The improper coordination of chromosome condensation and segregation during the cell cycle can lead to important structural abnormalities and result in cell death or diseases such as cancer (Valton & Dekker, 2016). In recent years, major advances in imaging and chromosome conformation capture approaches (Dekker et al, 2002; Lieberman‐Aiden et al, 2009; 3C, Hi‐C) have complemented earlier work by describing at an unprecedented resolution the multiple hierarchical layers of genome organization. A variety of remarkable 3D chromosomal structures have been described in a number of species, including in unicellular organisms such as bacteria and yeasts.
The genome of budding yeast Saccharomyces cerevisiae presents a Rabl organization driven by (i) centromeres clustering at the spindle pole body (SPB, S. cerevisiae microtubule organizing center), (ii) telomeres tethering to the nuclear envelope, (iii) the nucleolus where the rDNA is sequestered opposite to the SPB, and (iv) chromosome arm length (Burgess & Kleckner, 1999; Taddei & Gasser, 2012). Hi‐C experiments have confirmed this Rabl organization, but the existence of sub‐megabase structures within yeast chromosomes similar to mammalian topological associated domains or their bacterial equivalent is still controversial (Duan et al, 2010; Hsieh et al, 2015; Eser et al, 2017). Importantly, genomic analysis of chromosome 3D architectures has usually been done using asynchronous populations, in which cells are found in various stages of the cell cycle. However, the initiation and progression of replication, followed by the segregation of the sister chromatids (SCs) into daughter cells, is expected to modify the genome higher‐order organization. Recent studies have unveiled cell‐cycle stage‐specific genome‐wide topological variations in bacteria, yeast, fly, and mammals (Naumova et al, 2013; Guidi et al, 2015; Marbouty et al, 2015; Hug et al, 2017). As expected, in all species the largest reorganization transition is associated with SC condensation, a fundamental process occurring concomitantly to their individualization, and facilitating their proper segregation.
Pioneer studies on yeasts proved essential to study these processes. Mutations in cell‐division cycle (cdc; Hartwell et al, 1973) genes can block the cell cycle progression, enabling the study of global and/or local chromosome reorganization at specific cycle phases (Hartwell et al, 1973; Guacci et al, 1994; Sullivan et al, 2004; Renshaw et al, 2010; Rock & Amon, 2011). The evolutionary conserved structural maintenance of chromosomes (SMC) proteins bind to chromosomes and modify their structure in spatially and temporarily regulated manner during the cell cycle (Aragon et al, 2013; Uhlmann, 2016). Cohesins, such as Scc1, promote SC cohesion during DNA replication (Blat & Kleckner, 1999; Glynn et al, 2004) and get cleaved at the metaphase‐to‐anaphase transition (Uhlmann et al, 1999). At the same time, condensins such as Smc2 are loaded onto SCs to facilitate their segregation (Renshaw et al, 2010; Stephens et al, 2011; Hirano, 2012). In fission yeast, the binding of SMCs modifies the level of chromosome compaction defined as the ratio between long (> 10 kb)‐ and short‐range (< 10 kb) contacts at specific loci (Mizuguchi et al, 2014; Kim et al, 2016).
While Hi‐C studies on mammalian and drosophila cells have confirmed this compaction change and provided important insights on the organization of mitotic chromosomes’ internal structure (Naumova et al, 2013; Hug et al, 2017), no comprehensive analysis of the 4D dynamics of the chromosomes during an entire eukaryotic cell cycle has been achieved. To explore new chromosomal structural features over the cell cycle progression, we analyzed the internal folding and overall organization of S. cerevisiae genome over 15 synchronized time points and the role of cohesin and condensin using Hi‐C (Dekker et al, 2002; Lieberman‐Aiden et al, 2009). This analysis provides a broad overview and in‐depth insight on SMC‐dependent structural transitions resulting in chromosome individualization and segregation, including a potential role for a condensin‐dependent loop in contributing to the segregation of the rDNA cluster.
Hi‐C libraries were generated from cell cultures synchronized in G1 with elutriation (Marbouty et al, 2014) and/or arrested at different stages of the cell cycle through thermosensitive (ts) cdc mutations (Fig 1A; Hartwell et al, 1973). After sequencing, the corresponding normalized genome‐wide contact maps were computed (bin: 5 kb; Fig 1C and D, left panels; Fig EV1; Materials and Methods; Cournac et al, 2012).
These 2D maps were translated into 3D representations to visualize the main folding features (Lesne et al, 2014; e.g., centromeres and telomeres clustering in G1; Figs 1B and EV1). These 3D structures are average representations of the contact frequencies quantified over a population of cells and therefore do not represent the exact structure found in individual cells. For instance, on these 3D representations all the telomeres loosely cluster together. In a single nucleus, telomeres rather form small groups scattered all around the nuclear membrane (Taddei & Gasser, 2012). Since in different cells the composition of these clusters differs, all telomeres end up being regrouped together in the average 3D structure that reflects the population average of contacts. In addition, they are not polymer models and cannot be interpreted as such. Nevertheless, these representations conveniently highlight important structural features not readily apparent in the 2D maps (Mercy et al, 2017).
The differences between two conditions were determined by computing the log‐ratio between the maps (bin: 5 kb † ; Fig 1C; Materials and Methods). The color scale reflects the variations in contact frequency for each bin between two different contact maps. The ratio of contact maps generated from two independent G1 cell populations (experimental replicates) displays a relatively homogenous white (i.e., null) signal, corresponding to little differences between them (Fig 1C, right panel). These minor variations between the maps result in occasional faint colored areas and reflect experimental noise (Appendix and Materials and Methods). On the other hand, the ratio between exponentially growing G1 and quiescent G0 cells contact maps (Fig 1D, right panel) shows a strong difference in inter‐telomere contact frequencies, reflecting the formation of the telomeres hyper‐cluster characteristic of the G0 metabolic state (Guidi et al, 2015; Fig 1D, black arrowheads).
Multiple maps can also be compared altogether by computing their pairwise distance matrix, showing that the genome organization of cells in anaphase (cdc15) differs the most compared to other time points (Fig 1E; Materials and Methods). The overall similarities/differences between datasets can then be summarized using principal component analysis (PCA; Fig 1F). This 2D representation shows that the experimental duplicates (such as G1, or anaphase cdc15) clustered together, while the distance increases progressively between G1 (obtained with either elutriation or cdc6 ts mutant), metaphase (cdc20), and the distant anaphase (cdc15) datasets.
Altogether, these comparisons highlight major changes in chromosome higher‐order architecture taking place in cells progressing throughout the cell cycle into metaphase and anaphase.
To decipher the chromosome structural changes that take place during replication, synchronized G1 cells were released into S phase and Hi‐C maps generated for six time points sampled from two independent kinetics (Figs 2A and EV2; Materials and Methods). The PCA reveals a progressive structural evolution from G1 to late S/G2 phase (Fig 2B). The dependency of the contact probability P on genomic distance reflects the chromosome compaction state (Lieberman‐Aiden et al, 2009; Naumova et al, 2013; Mizuguchi et al, 2014). The P(s) shows a gradual and consistent enrichment in long‐range intra‐chromosomal contacts (> 20 kb) with respect to short‐range (< 10 kb) during replication (Fig 2C). This compaction change is absent when replication is impaired, for instance, in the absence of the replication‐checkpoint regulator cdc6 (Piatti et al, 1995), even though cells enter mitosis and engage into segregation of non‐replicated chromosomes (Fig 2D, left panel). The progressive increase in long‐range contacts stops with the completion of S phase, when it reaches the level observed in cells arrested at the G2/metaphase transition (G/M) with the microtubule‐depolymerizing drug nocodazole (Jacobs et al, 1988; Fig 2D, middle panel). The crossing of the P(s) slopes from the early to late replication time points occurs around 10–20 kb (Fig 2C, highlighted in gray), a window within the range of the spacing reported between cohesin binding sites (~11 kb on average; Glynn et al, 2004), suggesting that this change in compaction could be due to cohesin activity. In agreement with the key role of cohesin in sister‐chromatid folding during replication, Scc1 depletion using an auxin‐inducible degron scc1‐aid strain prevents the enrichment in long‐range contacts in S/G2 (Fig 2D, right panel). This result supports the hypothesis that distant regions enriched in cohesin are tethered together, resulting in chromatin loops (Guillou et al, 2010).
The Scc1‐dependent compaction occurs concomitantly with a gradual individualization of the SC pairs throughout replication, as shown by the overall increase in the ratio between intra‐ and inter‐chromosomal contacts from 63 ± 10% in G1 (six time points) to 73 ± 4% in S/G2 (four time points) and illustrated by the ratio between G1 and G2 maps (Fig 2E, top right ratio). In sharp contrast to this overall decrease in inter‐chromosomal contacts, the centromeres of different chromosome tend to strongly cluster in G2. In the absence of the cohesin Scc1, intra‐chromosomal contacts in G2 cells decrease to levels similar to or even below G1 (Fig 2E, bottom left ratio), while the major binding sites for cohesin (i.e., centromeres; Glynn et al, 2004) also exhibit a reduced level of contacts (Fig 2F; Appendix Fig S1). These results suggest that cohesins affect the genome organization through the gradual compaction of SC, the clustering of centromeres, and chromosome individualization. Although yeast chromosomes are shorter than mammalian chromosomes, they similarly change their internal conformation and individualize themselves prior to entering metaphase.
In budding yeast, replication initiates at discrete autonomously replicating sequences (ARSs; Brewer & Fangman, 1987). ARSs display partially stochastic activation, with only a subset of origins activated early during S phase. The distribution of early origins is uneven, with an enrichment in pericentromeric regions, and a depletion in subtelomeric regions. The genome‐wide pattern of ARS activation timing defines a population‐average replication timing program (Raghuraman et al, 2001). To investigate the link between genome organization and replication timing, the read coverage of the Hi‐C libraries was used to compute the replication timing profile of the cell population for each of the time point, and follow their progression through S phase. The average profile correlates well with previously published pattern (Raghuraman et al, 2001; McCune et al, 2008; Fig 3A; Materials and Methods). To visualize the progression of replication on the higher‐order architecture of the genome, we colored the 3D structures recovered from three early replication time points according to their replication progression status. The superimposition of the three structures recapitulates intuitive properties of yeast replication program, with a “replication wave” propagating from the centromeric regions enriched in early origins, through chromosomal arms, and toward the late replicating subtelomeric regions (Fig 3B and C; red and blue signal, respectively).
We also asked whether our data support the proposed co‐localization of adjacent early replication origins (Kitamura et al, 2006; Knott et al, 2012; Saner et al, 2013). We found a statistically significant enrichment in contacts between these positions and their surrounding regions, but whether it results from an active co‐localization or from their positioning in the pericentromeric regions co‐localized due to the Rabl organization remains unclear (not shown). More analyses are required to solve this question and integrate the different observations.
After replication, cells progress into mitosis (M phase). During metaphase, microtubules originating from opposite SPBs attach to the kinetochores of the two SCs (London & Biggins, 2014). The anaphase‐promoting complex (APC) co‐activator Cdc20 is essential for the proper activation of separase, resulting in the cleavage of cohesin and SC segregation in anaphase (Uhlmann et al, 1999; Visintin et al, 1997; 20). In the absence of Cdc20, cohesins are not cleaved and cells remain blocked in metaphase. Another key player in mitosis progression is the Cdc15 kinase which promotes mitotic exit at the end of anaphase by activating cytokinesis (Rock & Amon, 2011). In the absence of Cdc15, cells are therefore blocked into late anaphase. The higher‐order changes in the organization of chromosomes that take place during metaphase and anaphase were investigated using populations of cells synchronized with conditional mutants of cdc20 and cdc15, respectively. Contact maps of cdc20‐, cdc15‐, and cdc15‐arrested cells released into permissive conditions were generated to characterize chromosome reorganization throughout M phase (Figs 4A and EV3; Materials and Methods). PCA shows that the major structural change occurs during mitotic exit and that cells released from the cdc15 arrest display after 60 min a G1‐like genome structure, reflecting the fact that the entire cell cycle is now covered by our analysis (Fig 4B). The P(s) reveals a strong increase in short‐range contacts (< 10–20 kb) from G2 to anaphase, exceeding G1 levels which are only restored after anaphase completion (Fig 4C, left panel). This increase in short‐range contacts and the accompanying drop in long‐range contacts suggest the formation of an elongated, stretched structure. Upon spindle destabilization using the microtubule‐depolymerizing drug nocodazole in cdc15‐arrested cells (cdc15 noc), the two segregated chromosomal masses get closer as shown by imaging of DAPI‐stained cells (Fig 4C, inset; Fig EV4; Materials and Methods), in agreement with former reports (Jacobs et al, 1988). In these cells, the stretched chromosomal structure disappears as shown by a P(s) that now overlaps the G2 curve (Fig 4C, right panel). Besides the change in P(s), the global contact pattern of cdc15‐arrested cells remains unaltered following nocodazole treatment (Fig 4D, upper right ratio). Altogether, these results show that microtubule‐dependent segregation forces contribute to the stretching the chromosomes in anaphase, possibly in combination with additional constraints resisting this force such as the cohesion of SC arm extremities (see Discussion).
Nocodazole is commonly used to synchronize cells at the G2/M transition. We took advantage of having contact maps of cdc20‐arrested cells in metaphase to compare them with those obtained from nocodazole‐arrested cells (Fig EV4; Materials and Methods). The ratio map appeared globally similar, although we noticed in the presence of nocodazole a small drop in inter‐chromosomal contacts (Fig 4D, bottom left ratio). Chromosome 12 (chr12) also presents a peculiar signal at the level of the rDNA cluster (Fig 4E, left panel), with an enrichment in contacts between the two flanking regions of the rDNA cluster in G2/M nocodazole‐treated cells compared to cdc20‐arrested cells (Fig 4E, right panel). These results indicate that the G2/M nocodazole arrest is associated with a destabilization of the chr12 structure at the level of the rDNA locus. The intra‐chromosomal contact increase within chr12 is also accompanied by a global decrease in inter‐chromosomal contacts in the presence of nocodazole (Fig 4D, bottom left ratio). Remarkably, chr12 organization was not affected when cdc15‐arrested cells were treated with nocodazole (Fig 4D, top right ratio). Altogether, these observations point to a role for the microtubule array in maintaining the organization of the nucleolus inside the nucleus, before its segregation in anaphase. In summary, while chromosome structures are overall similar in cell synchronized in G2 by nocodazole or in a cdc20 ts mutant, nocodazole‐arrested cells present a slightly different nucleolus structure (and, by extension, chr12). One interpretation could be that the condensation of the rDNA is not yet completed in G2/M nocodazole arrest and that as a result, rDNA flanking regions are freer to contact each other's.
The comparison of cdc15 and cdc20 maps shows an increase in centromere clustering in anaphase, leading to the formation of a prominent polymer brush structure (Daoud & Cotton, 1982; Fig 4F, bottom left ratio, yellow arrowheads). Such increase is in agreement with the role of condensin in forming a “spring” of chromatin at pericentromeric regions at the metaphase‐to‐anaphase transition (Stephens et al, 2011). Surprisingly, a peculiar loop pattern appears on chr12 in cdc15‐arrested cells, bridging the centromere and the centromere–proximal left flanking region of the rDNA cluster (see pink arrowheads in Fig 4F, G and H). Upon release from the cdc15 arrest, the telomere–proximal right flanking region of the rDNA cluster becomes strongly isolated from the rest of the genome (Fig 4F, upper right ratio; cdc15+40 min), while the contacts of the centromere–rDNA loop intensify (Fig 4H; cdc15+40 min). After completion of mitosis and re‐entry in interphase (cdc15+60 min), the loop disappears (Fig 4H). Interestingly, this loop can be seen in asynchronous populations while it is only present in anaphase (Fig 4I). 3D representations illustrate the dramatic reorganization of chr12 and the formation of the loop bridging centromeric region and the rDNA (Fig 4J, pink arrowheads). Microtubules are not required to maintain this loop in anaphase, since it remains present in cdc15‐arrested cells treated with nocodazole (Fig 4D, upper right ratio; Fig 4I and K), suggesting that the left flanking region of the rDNA is physically bound through an unknown mechanism to the centromeric regions. These results complement imaging studies showing that the rDNA exhibits a dense, line‐like shape that extends throughout the nucleus at anaphase (2.1 SD, 0.2 μm; Sullivan et al, 2004).
The proper condensation and segregation of the rDNA cluster requires the nucleolar release of the Cdc14 phosphatase. Cdc14 mediates a shutdown of rDNA transcription, facilitating the loading of the Smc2 condensin and hence the condensation of the cluster (Yoshida et al, 2002; D'Amours et al, 2004; Sullivan et al, 2004, 14; Machín et al, 2006; Clemente‐Blanco et al, 2009). In addition, topoisomerase II (Top2), which decatenates the intertwining structures that appear between SCs during replication, is also required for rDNA segregation to proceed (Sullivan et al, 2004; D'Ambrosio et al, 2008; Baxter et al, 2011; Leonard et al, 2015). We investigated the influence of those factors on the 3D structure of the rDNA locus during anaphase (Figs 5A and EV5; Materials and Methods).
First, Smc2 depletion in smc2‐aid cdc15‐arrested strain affects anaphase genome organization by (i) reducing centromere clustering and (ii) suppressing the formation of the rDNA loop, with a resulting contact map highly similar to the cdc20 map (Fig 5B, bottom left ratio). Therefore, condensins are responsible for the observed increase in inter‐centromere contacts at anaphase compared to metaphase (Fig 4F, bottom left ratio), while they are also required for the formation of the loop bridging the centromere of chromosome 12 with the rDNA cluster (two loci enriched in condensin deposition). The smc2 cdc15 and cdc14 maps are strikingly similar (Fig 5C, bottom left ratio). The 3D representations of smc2 cdc15 and cdc14 cells (Fig 5E) and the rDNA contact plots with the rest of chr12 (Fig 5F) illustrate the loss of the rDNA loop in the absence of Smc2 and/or Cdc14. In addition to this effect, both mutants also display the same decrease in centromere clustering compared to cdc15 cells (Fig 5C, upper right ratio; Fig 5G), pointing at their functional relationship on the same pathway.
The organization of the genome was also compared in cdc15‐arrested cells in the presence or absence of Top2, top2‐aid cdc15‐arrested (Fig 5D, upper right ratio; Fig 5E and F; Materials and Methods). Top2‐depleted cells display a strong decrease in contacts between the telomere–proximal region of chr12R and the rest of the genome (including chr12L). The signal is consistent with the essential role played by Top2 in rDNA segregation, showing that the non‐segregated regions are isolated from the segregated chromosomal sets. The comparison between top2 cdc15 and cdc20 cells reveals an enrichment in contacts at centromeres and the persistence of the centromere–rDNA loop in the Top2 mutant (Fig 5D, bottom left ratio). These results indicate that the formation of these condensin‐dependent structures in anaphase is independent from the decatenation and/or the segregation of the rDNA cluster.
This study consists of an experimental and analysis framework to systematically investigate and compare chromosome folding and organization at different stages of the cell cycle. We applied Hi‐C to populations of cells synchronized at different points of the cycle, generating genome‐wide, 5‐kb‐resolution contact maps which unveil their average 3D genome organization. The global influence of cohesin, condensin, and topoisomerase 2 has been investigated in the corresponding mutants, as well as the effects of the microtubule‐depolymerizing drug nocodazole. Comparative approaches between contact maps provided a global view of the structural transitions between the different stages of the cycle, some expected, such as chromosome compaction during replication, and others that had not been described before, such as topological structures involving the rDNA cluster.
An overview of chromosome structural changes during the cell cycle can be summarized from centromere contacts, intra‐/inter‐chromosomal contact ratio, and short‐/long‐range contact ratio computed for each of the time points (Fig 6A).
Centromere clustering gradually increases during the cell cycle, through the establishment of sister‐chromatid cohesion during replication, and through condensin‐dependent clustering during anaphase (Fig 6A, upper panel). A potential consequence of this increased clustering in anaphase could be the generation of a stronger polymer brush, that is, the mechanical phenomenon that leads to the self‐organization of a polymer tethered to a surface into stretched, non‐intermingling structure (de Gennes, 1987). Interestingly, the strengthening of the polymer brush organization could consequently contribute to chromosome individualization during anaphase. The intra‐/inter‐chromosomal contact variations reflect the successive phases of chromosome individualization and intermingling, with individualization taking place during replication (cohesin‐dependent) and during anaphase (spindle‐dependent; Fig 6A, bottom panel). The intra‐/inter‐chromosomal contact ratio correlates strongly with centromere clustering (c = 0.72, p = 10−4), with both ratios peaking during anaphase exit.
Short‐/long‐range contact ratio recapitulates the three different internal folding (I, II, and III) states of chromosomes (G1, G2, and anaphase; Fig 6B, middle panel). These three states can be determined based on a quantitative analysis of the significance of changes between P(s) curves obtained using several replicates in different phases of the cycle (Fig 7). During replication, cohesins mediate the compaction of chromosomes from state I to II. The chromosomes are then stretched by the mitotic apparatus during anaphase (state III) before returning to state I in G1. The mechanical constraint imposed by the anaphase spindle appears responsible for the state III stretching, as a nocodazole treatment results in relaxation of chromosomes, which switch back to state II. Imaging of the two sets of segregated chromosome during nocodazole treatment supports this spring relaxation effect, with the two masses being brought back together upon the depolymerization of microtubules. The nature of the mechanical constraints remains unknown, but it is tempting, in light of our observation of chr12 behavior (below), to propose a role for condensins in actively promoting this movement. In this scenario, condensins could favor the segregation of sister chromatids by pulling the chromosomes toward the centromere cluster. As a result, the loss of microtubule and tethering to the SPB may lead cohesin to actively pull back the segregated region together. We anticipate that whether condensins play an active role in the segregation of chromosomes in addition to the pulling force imposed by the microtubule spindle will be thoroughly investigated in the years to come.
In addition, we also show that the two main regions of condensin deposition, that is, the centromeres and the rDNA locus, are bridged during anaphase through a condensin‐dependent mechanism resulting in a loop‐like structure on the right arm of chromosome 12. Whether this structure is systematically found in all cells, or only in a subset of the population, remains to be determined through single‐cell imaging approaches such as FISH analysis. Although the precise mechanisms of formation remain unknown as well as its functional importance, we show that the setting up of the loop depends on condensin. Several mechanisms can be envisioned for the generation of this loop. One possibility is that starting from regions with a high condensin density, an active mechanism such as DNA extrusion through the action of condensins would pull the centromere and the rDNA cluster together (Fig 6C). Condensin depletion (leading to disruption of the loop) is associated with segregation defects. Overall, this structure therefore appears to de facto play a role in the segregation of the rDNA cluster, potentially through the application of a force that would drag the rDNA region to the centromere cluster before the completion of anaphase. A consequence of this model, would be that a similar loop extrusion mechanism could facilitate the segregation of other chromosomes as well. In this case, one or more loops could actively facilitate the segregation of large regions of chromosomes toward the tethered centromeres, down the telomeric regions. Chromosome 12, in this scenario, would appear as an exception with the large rDNA cluster generating a physical barrier in the middle of the right arm that is not present in other chromosomes. More experiments are nevertheless needed to investigate this proposed role. Yeast chromosome 12 could therefore prove a convenient model to study the action of loop extrusion mechanism (Alipour & Marko, 2012).
The importance of the rDNA loop remains to be further characterized as well as its similarity with loops found in other eukaryotic species. Overall, our exhaustive dataset opens new avenues for the comprehensive analysis of the 3D chromosome choreography during replication and segregation and brings to light new perspectives regarding these fundamental processes.
All strains were grown in rich medium (YPD: 1% bacto peptone (Difco), 1% bacto yeast extract (Difco), and 2% glucose), except for YKL051 (MET3‐HA‐CDC20) that was grown in synthetic complete medium deprived of methionine (SC: 0.67% yeast nitrogen base without amino acids (Difco), supplemented with a mix of amino acids, uracil and adenine, and 2% glucose). Cells were grown at either 30°C or 23–25°C (the later temperature corresponding to the permissive temperature of the conditional thermosensitive mutations cdc6‐1, cdc14‐3, and cdc15‐2; see below for details). Dataset corresponding to the quiescent state (G0) comes from already published data by Guidi et al (2015) and was obtained by carbon source exhaustion. All strains are described in Table EV1.
To recover G1 daughter cells, the exponentially growing cultures were elutriated—a physical method of synchronization, used to separate cells according to their density and sedimentation velocity (see Appendix Supplementary Methods; Marbouty et al, 2014). The G1 daughter cells recovered through elutriation were suspended in fresh YPD at 30°C for 30 min, so they could recover from the elutriation procedure (i.e., stay in PBS). To minimize the potential variability introduced by the age heterogeneity of the bulk population, G1 daughter cells were used as starting point for all cell cycle synchrony and in combination with genetic and chemical synchronization methods (see below).
G1 elutriated cells were released into S phase to analyze genome conformation during this stage. 2 × 109 G1 cells—originating from the same elutriated fraction to minimize heterogeneity in replication initiation—were inoculated into 150 ml YPD at 25°C (to slow down replication fork progression). Upon release, the synchronized cultures were sampled every 5 min and the cells analyzed through FACS, revealing an approximate lag of 130 min before replication restart. Therefore, aliquots were cross‐linked and processed into Hi‐C libraries at 135, 140, 145, 150, 155, 160, and 165 min. The progression of each fraction throughout the S phase (from G1 to G2) was monitored with flow cytometry.
Synchronizations using thermosensitive (ts) cdc strains (Hartwell et al, 1973) were all performed starting from elutriated G1 daughter cells growing in non‐permissive temperature conditions designed to arrest the progression of the cycle at specific phases. See Appendix Supplementary Methods for details of synchronization procedures of strains YKL052 (cdc14‐3), YKL053 (cdc15‐2), and YKL054 (cdc6‐1).
Chemical synchronization was also performed on elutriated G1 daughter cells.
Synchronization at the G2/M transition was achieved by restarting G1 cells (strain YKL050) in YPD at 30°C for 1 h, followed by the addition of nocodazole (Calbiochem; 15 μg/ml) and incubation for another 2 h at 30°C. Cells arrested in G2/M with nocodazole were either processed into Hi‐C libraries, or washed and inoculated in fresh YPD medium at 30°C. The washing of nocodazole allowed G2/M synchronized cells to proceed into M phase (cells sampled after 20, 45, 60 and 90 min were processed into Hi‐C libraries).
To investigate the constraints imposed by the spindle during anaphase, elutriated YKL053 cells were elutriated and the recovered G1 daughter cells processed and blocked into anaphase using the cdc15‐2 thermosensitive mutation. A sample of the population was then incubated with nocodazole (15 μg/ml) for 20 min. A sample was released at permissive temperature in the presence of nocodazole for 20 min. Finally, a sample was released at the permissive temperature for 20 min before being incubated with nocodazole for 20 min.
For synchronization in metaphase, a system allowing induced depletion of cdc20 was used (MET3‐HA‐CDC20; strain YKL051). Elutriated G1 daughter cells were restarted in YPD complemented with 50 μg/ml methionine for 5 h at 30°C. Cells arrested in metaphase were split into different aliquots. One sample was immediately processed into a Hi‐C library, while two others were washed, suspended in SC medium without methionine, and processed into Hi‐C after 20 and 40 min.
To investigate the influence of SMC on chromosome organization, strains carrying auxin‐inducible degron (aid) versions of Scc1 (strain YKL055) and Smc2 (YKL056) proteins were processed into Hi‐C libraries. The degradation of these proteins is induced when auxin (IAA) is added to the medium at a final concentration of 2 mM. Both asynchronous populations of strains YKL055 and YKL056 were elutriated in the absence of IAA. G1 daughter cells were incubated in YPD supplemented with IAA at 30°C. A sample of the YKL055 population (scc1‐aid) was processed into a Hi‐C library in late S/G2 (see Release into S phase). For the YKL056 population (smc2‐aid), the cells were arrested in late anaphase using the cdc15‐2 mutation also present in the genome of this strain, before being processed into a Hi‐C library.
To study the influence of topoisomerase II‐mediated decatenation on chromosome organization, we used a strain (YKL057) in which TOP2 gene is tagged by aid (top2‐aid) and that also carries the cdc15‐2 mutation. An asynchronous exponentially growing culture of YKL057 cells was split into two fractions incubated for 3 h at the non‐permissive temperature of 37°C in either the presence or absence of IAA (20 mM). The synchrony of each time point was monitored with flow cytometry and microscopy, and the cells were processed by Hi‐C.
About 5 × 106 cells were fixed in ethanol 70% and stored at 4°C overnight. Cells were then pelleted, washed, and incubated in sodium citrate 50 mM (pH7.4) complemented with RNase A (10mg/ml; Roche) for 2 h at 37°C. Next, Sytox green (2 μM in sodium citrate 50 mM; ThermoFisher) was added and cells incubated for 1 h at 4°C. Flow cytometry was performed on a MACSQuant Analyzer (Miltenyi Biotec), and data were analyzed using FlowJo X 10.0.7 software (Tree Star).
Fractions of cells fixed in ethanol 70% and stored at 4°C overnight were pelleted and washed three times for 5 min in 1× PBS. Cells were permeabilized by immersion in 0.2% Triton X‐100 (Biosolve) for 5 min. To remove the Triton, cells were pelleted and washed three times in 1× PBS. The liquid was aspirated and cells were suspended in DAPI labeling solution (2 μg/ml in 1× PBS) for 10 min at room temperature. Before imaging acquisition, the labeling solution was aspirated and the cells were washed three times for 5 min in 1× PBS. Cells were imaged at 350 nm excitation wavelength with Nikon fluorescence microscope (Camera Andor Neo sCMOS, software Andor IQ2 2.7.1, LED Lumencor Spectra X).
Hi‐C libraries were generated using the four‐cutter enzyme DpnII through a protocol adapted from Belton et al (2012). The protocol is detailed in Appendix Supplementary Methods. The resulting libraries were used as template for the Illumina amplification by PE‐PCR primers and paired‐end‐sequenced on the NextSeq 500 or HiSeq 2000 Illumina platform (2 × 75 or 2 × 150 bp kits; see Table EV2 for details).
Raw Hi‐C data were processed as follows. PCR duplicates were removed using the 6 Ns present on each of the custom‐made adapter and the 2 trimmed Ns. Paired‐end reads were mapped independently using Bowtie 2.1.0 (mode: –very‐sensitive –rdg 500,3 –rfg 500,3) against the S. cerevisiae reference genome (S288C). An iterative alignment, with an increasing truncation length of 20 bp, was used to maximize the yield of valid Hi‐C reads (mapping quality > 30). Only uniquely mapped reads were retained. On the basis of their DpnII restriction fragment assignment and orientation, reads were classified as either valid Hi‐C products or unwanted events to be filtered out (i.e., loops and non‐digested fragments; for details, see Cournac et al, 2012, 2016). To generate contact matrices for all time points along the cycle, filtered Hi‐C reads were binned into units of single restriction fragments, and then, successive fragments were assigned to fixed size bins of either 5 or 50 kb. Bins that exhibited a high contact frequency variance (< 1.5 Standard Deviation or 1.5–2 SD. from the mean) were filtered out for all maps to allow pairwise comparison of the data. On average, around 15 million of valid reads were used to build each contact map. To remove potential biases resulting from the uneven distribution of restriction sites and variation in GC content and mappability, the contact maps were normalized using the sequential component normalization (SCN) procedure (Cournac et al, 2012).
To assess the similarity between normalized matrices, these were binned at 50 kb and quantile‐normalized (Hicks & Irizarry, 2015). We then measured their similarity by computing the Euclidean distance between them. In order to visualize similarities between sets of matrices, we did a principal component analysis (PCA) of the pairwise distance matrix between samples.
Polymers display a decrease in contact probability, P(s), as a function of the genomic distance, s. The degree of decay of P(s) was often interpreted as informative of the polymer state. To compute the intra‐chromosomal P(s) plots, pair of reads aligned in intra‐chromosomal positions were partitioned by chromosome arms. Reads oriented toward different directions or separated by < 1.5 kb were discarded to filter for self‐circularizing events. For each chromosome, read pairs were log‐binned in function of their genomic distance s (in kb), according to the following formula:
The P(s) plot is the histogram computed on the sum of read pairs for each bin. This sum is weighted by the bin size 1.1(1+bin) (because of the log‐binning), as well as the difference between the length of the chromosome and the genomic distance s. The difference acts as a proxy for the number of possible events.
To obtain the 4C‐like intra‐ and inter‐chromosomal contact profiles for rDNA and centromeres, adjacent bins were indexed on the respective chromosomes. The resulting indexed and filtered matrices at either 5‐ or 50‐kb bin were normalized using SCN (see Generation and normalization of contact maps). The profiles for the selected bins were plotted and compared using Matlab (no smoothing was applied).
The replication profile was computed from the raw 5‐kb‐binned contact maps. Firstly, G1 replicates were averaged and the sum of contact over each 5‐kb bin was computed. The same computation was repeated for datasets obtained from cells released into S phase. To obtain the replication timing, we computed the ratio of these two signals and smoothed this ratio using a running‐average window of six bins.
The 3D representations of the contact maps were generated using ShRec3D (Lesne et al, 2014) on the normalized contact maps, filtered for low‐signal bins. First, the algorithm computes the distance matrix from the contact map, by assuming that distances are inversely proportional to the normalized contact counts. A shortest path algorithm is then used to insure that the distance matrix satisfies the triangular inequality. Finally, we used Sammon mapping to recover the optimal 3D coordinates from the distance matrix (Morlot et al, 2016). All the 3D structures presented here were rendered using VMD (Humphrey et al, 1996). Besides the cautiousness regarding the interpretation of 3D structure we mention in the main text, we also underline that the 3D structures are not used to compare datasets: All computational analyses are performed using the contact map data.
To compare contacts between centromeric regions, the sum of normalized inter‐chromosomal contacts between 100‐kb regions centered on centromeres was computed and divided by the total number of normalized inter‐chromosomal contacts between all chromosomes. To compare short‐ versus long‐range contacts, a ratio of intra contacts was computed as follows. The number of intra contacts involving fragments positioned < 30 kb apart was divided by the number of intra contacts involving fragments positioned more than 30 kb apart, for all chromosomes. For intra‐ versus inter‐chromosomal contacts, the total number of normalized intra‐chromosomal contacts was divided by the sum of normalized inter‐chromosomal contacts.
To assess for the contribution of experimental variability to the variations in contacts between different conditions, we proceeded as follows. Density histograms displaying the distribution of the log2 contact ratio of all elements of Hi‐C matrices (50‐kb bins) between pairs of biological and experimental replicates (3×G1, 2×G2, 3×M) were computed and compared to similar histograms computed from pairs of Hi‐C matrix obtained in different experimental conditions (see Appendix Fig S2).
An estimation of the replicate variability at the centromeres was obtained by plotting the boxplots representing the distribution of the log2 contact ratios between pairs of biological and experimental replicates only of the bins encompassing the centromeres (50‐kb bins; see mask; Appendix Figs S3 and S4). The same computation was performed on pairs of matrices obtained in different conditions to estimate the statistical significance of the variations. All replicates were taken into account. P‐values were obtained by the Wilcoxon signed‐rank test.
LL‐S and RK designed research. LL‐S performed the experiments, with contributions from GM, AT, and HM. TMG generated the smc2‐aid mutant. VFS and JM analyzed the data, with contributions from LL‐S. LL‐S, JM, and RK interpreted the data and wrote the manuscript.
The authors declare that they have no conflict of interest.
Expanded View Figures PDF
Review Process File
We thank Martial Marbouty and Axel Cournac for contributions to the early stage of this project. Sample description and raw contact maps are accessible on the GEO database through the following accession number: GSE90902. Raw sequences are accessible on SRA database through the following accession number: SRP094582. We are also grateful to Armelle Lengronne, Stephane Marcand, Philippe Pasero, Etienne Schwob, Emmanuelle Fabre, Nancy Kleckner, and Angela Taddei for sharing strains and for discussions. Vittore Scolari and Heloise Muller were partly supported by Pasteur‐Roux‐Cantarini postdoctoral fellowships. This research was supported by funding to R.K. from the European Research Council under the 7th Framework Program (FP7/2007‐2013, ERC grant agreement 260822), from Agence Nationale pour la Recherche (MeioRec ANR‐13‐BSV6‐0012‐02), and from ERASynBio and Agence Nationale pour la Recherche (IESY ANR‐14‐SYNB‐0001‐03).
The EMBO Journal (2017) 36: 2684–2697
See also: C Barrington et al (September 2017)
1Correction added on 15 September 2017 after first online publication: Bin size was corrected from 5 to 50 kb.
Julien Mozziconacci, Email: rf.ueissuj.cmtpl@iccanocizzom.
Romain Koszul, Email: firstname.lastname@example.org.