Research paper
Nucleus 3:4, 370-383; July/August 2012; © 2012 Landes Bioscience
Chromosome positioning and the clustering of functionally related loci in yeast is driven by chromosomal interactions 1 Institute of Natural Sciences; Massey University; Auckland, New Zealand; 2Deutsches Krebsforschungszentrum; Biophysics of Macromolecules; Heidelberg, Germany; 3Bio Systems Analysis Group; Friedrich-Schiller-Universität; Jena Centre for Bioinformatics; Jena, Germany; 4Institute of Information and Mathematical Sciences, Massey University; Auckland, New Zealand; 5Département de Biologie Cellulaire; Université de Genève; Genève, Switzerland
These authors contributed equally to this work.
†
Keywords: yeast, genome architecture, proximity-based ligation, modeling
In recent years there has been considerable and growing interest in the 3-dimensional organization of genomes. In this manuscript we present an integrated computational-molecular study that produces an ensemble of high-resolution 3-dimensional conformations of the budding yeast genome. The compaction, folding and spatial organization of the chromosomes was based on empirical data determined using proximity-based ligation. Our models incorporate external constraints that allow the separation of gross organizational effects from those due to local interactions. Our models show that yeast chromosomes have preferred yet non-exclusive positions. They also identify interaction dependent clustering of tRNAs, early firing origins of replication, and Gal4 protein binding sites, yet the cluster composition is dynamic. Our results support a link between structure and transcription that occurs within the context of a flexible genome organization.
Introduction Structure is obvious within the objects and networks we interact with on a daily basis. Yet how structure contributes to the workings of systems we cannot see by eye is by necessity less obvious. Microscopic and molecular analyses have determined that structural features contribute to cellular functions including enzymatic activity and directed movement (e.g., ref. 1). Similarly, the spatial organization of genomic loci is increasingly found to be implicated in gene regulation and thus contributes to the genotype-phenotype transition. However, due to its complexity the spatial organization of genomes has remained recalcitrant to high resolution investigation leaving a gap that prevents a holistic understanding of the genome and its central role within cellular processes. Proximity-based ligation methodologies have been used to globally identify the contact points within and between chromosomes.2-6 The strength of the proximity-based ligation methodologies is their ability to identify loci that are directly or indirectly (i.e., mediated by a defined or unknown complex) held close enough to be cross-linked, within a subset of the test population, at a moment in time. However, these techniques do not reveal the spatial organization of loci outside of cross-linking distance nor do they identify where a given locus is positioned within the nuclear or cellular space.7 Despite this, it is plausible
that the spatial arrangement of the genome inside the nucleus can be reconstructed by combining the entirety of chromosomal interactions measured by proximity-based ligation (e.g., genome conformation capture; GCC2). Single representations of the 3-dimensional organization of the Saccharomyces cerevisiae and Schizosaccharomyces pombe genomes have been generated using empirically obtained chromosomal interaction frequencies that were converted into linear distance approximations.6,8,9 These studies resulted in the proposal of unique genome conformations that do not represent the dynamic nature of nuclear organization. However, it is important to incorporate dynamism because genome organization shows a lot of variability10-12 indicating that it must change from cell to cell and over time. Critically, the chromosomal interactions identified by proximity-based ligation are averaged across the test cell population and the subset of interactions that is present in a single cell cannot be determined. Therefore, it remained unclear: (1) how much the genome organization depends upon the specific subset of interactions that are currently present within it; and (2) what contribution the nuclear membrane and gross localization effects (e.g., centromeres, telomeres and the nucleolus) make to the overall organization. Here we use a coarse grained molecular model of the S. cerevisiae genome to produce an ensemble of genome conformations
*Correspondence to: J. M. O’Sullivan; Email:
[email protected] Submitted: 03/24/12; Revised: 05/22/12; Accepted: 06/01/12 http://dx.doi.org/10.4161/nucl.20971 370
Nucleus
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
Lutz R. Gehlen,1,† Gerd Gruenert,1-3,† M. Beatrix Jones,4 Chris D. Rodley,1,5 Jörg Langowski2 and J.M. O’Sullivan1,*
Figure 1. Heat map of genomic organization in respiro-fermenting yeast cells. (A) The interaction network for exponentially growing yeast cells was determined by GCC (see Methods, sequences were banked in GEO 33018). The genome was divided into 55,000 bp segments. Interactions for each restriction fragment were assigned to the 55,000 bp segment that contained the restriction fragment. Where fragments overlapped two or more segments, proportional assignments were made according to the percentage of overlap between the fragment and segments. (B) Cartoon depicting modeling constraints used in this analysis. Centromeres (C) were attracted to a zone stretching 50 nm from the spindle pole body. Telomeres (T) were pushed from the nucleus center (0,0,0) to a peripheral zone 800 nm from the center and fragments that interact with the rDNA (R) were attracted to within 100 nm of the nucleolus. Information pertaining to compaction levels and interaction patterns for segments was added to the model (see Methods). (C) Snapshot of one possible solution for the yeast genome organization obtained by molecular modeling.
based upon randomly chosen subsets of empirically defined interactions. We observe distinct chromosomal positioning and functional clustering in our conformations. Models that incorporated reduced sets of constraints enabled us to isolate the effects of the nuclear membrane and gross localization and confirmed that the observed organization is largely driven by the chromosomal interactions. Results A coarse grained polymer model of the yeast genome. To derive the structure of the yeast genome beyond cross-linkable proximity from our experimentally obtained chromosome interaction data sets (Fig. 1A), we developed a coarse grained computational model of the yeast genome. In our model, each chromosome is represented by a polymer chain, either a freely-jointed (open chromatin) or worm-like chain (compact chromatin), with an overall compaction level of 70% (Supplemental Methods) and the following external constraints (Fig. 1B): (1) centromeres are attached elastically to a point on the periphery representing the spindle pole body (SPB);13 (2) telomere movement is restricted to the outermost shell (less than 200 nm from the periphery) of nuclear space14 and (3) the rDNA is located opposite the SPB.10 250 simulations were run for each of several conditions, and for each simulation run, inter- and intra-chromosomal interactions were randomly selected from the empirically determined chromosome interaction data sets for incorporation. See the Methods section for further details on simulation parameters. Yeast nuclei exhibit rotational symmetry with respect to the SPB-nucleolus axis.10,12 This rotational symmetry is intrinsic to the external constraints and is therefore present in our models. Thus, all constraints (i.e., external constraints and DNA-DNA interactions) to the system are equally well fulfilled if the whole genome conformation is rotated around the SPB-nucleolus axis
by an arbitrary angle. Therefore, the final conformations of different simulation runs were aligned to each other by minimizing the root mean squared deviation (RMSD) between the models before comparison (see Methods). Yeast chromosomes assume non-random positions. Mammalian chromosomes reside in distinct regions of the nucleus termed chromosome territories.15-19 These territories are not disjunct and a degree of chromosome mixing occurs.18 So far, the existence of chromosome territories within the budding yeast nucleus is debated20,21 although there is growing evidence of their existence.2,6,10,12,21 Therefore, we used our model to investigate if single chromosomes assume preferred positions within the yeast nucleus. The nuclear regions occupied by the individual chromosomes were reconstructed by pooling the positions of the polymer segments representing the respective chromosomes from all 250 aligned conformations. Density contour plots for the segment positions of all chromosomes are shown in Figures 2 and S2. These data reveal that yeast chromosomes assume distinct positions and gross structures, albeit with considerable overlap. This is consistent with observations for higher eukaryotic chromosomes.18 As one would expect, the short chromosomes, which are most restricted by the positions of their centromere and telomeres, tend to stay close to the spindle pole body (Figs. 2A and 3A) while larger chromosomes extend toward the nucleolus (Figs. 2B and 3B). Consistent with the fact that chromosome XII contains the rDNA, this chromosome is most affected by nucleolus positioning (Fig. 2C). Extensive interactions with nucleolar and nucleolus-associated sequences, as observed for chromosome VI and XIV, appear to modify the general chromosome organization such that split domains form (Fig. 2D and E). These data tend to reinforce the role of the nucleolus in chromosome organization.22-26
www.landesbioscience.com Nucleus
371
©2012 Landes Bioscience. Do not distribute
research paper
Research paper
Chromosomal interactions play a key role in the distinct positioning of chromosomes. The distinct positioning and the partially length-dependent separation of chromosomes we observed within our models raise the question: what are the key factors that determine chromosome positioning? As previously mentioned, it is plausible that the positioning of small chromosomes close to the spindle pole body is a consequence of the centromere-SPB interaction (Fig. 3A). However, it was unclear if the other inter- and intrachromosomal interactions that were incorporated into our models play a major role in the global positioning of chromosomes or if they only serve to “fine-tune” the genome structure. Therefore, we determined the effects of intragenomic interactions (both inter- and intrachromosomal) by calculating sets of simulations using different levels of external and internal constraints (Fig. 4A). In the first scenario—the confined model (Fig. 4Ai)—the chromosomes were enclosed into the nucleus without any further constraints. In the second scenario—the constrained model (Fig. 4Aii)—external constraints (i.e., centromere attachment to the SPB, telomere attachment to the periphery, and the spatial separation into nucleolus and the rest of the genome) were added to the confined model (see Methods). In the third scenario—the interaction model (Fig. 4Aiii)—interand intra-chromosomal interactions were incorporated into the constrained model. Change from the confined to constrained model is accompanied by an alteration in chromosome position, consistent with the addition of external constraints. However, it appears that the major change in condensation results from the addition of chromosomal interactions (Fig. 4A; Fig. S2).
372
We quantified the effects of the different constraints by calculating the root mean squared deviation (RMSD) for all possible pairwise comparisons within each set of models generated in scenarios 1–3 above, following rotational alignment of the centers of mass (see Methods). The RMSD measures the deviation between two conformations and as expected it was reduced in the constrained, as compared with the confined model (Student’s t-test p < 10 -15 ; Fig. 4B). However, the largest decrease occurred upon the addition of the chromosomal interactions (i.e., from 5,937 to 4,261; Fig. 4B). This means that the variation between the interaction model genome conformations was much smaller than the variation between the genome conformations in the other models. In other words, the interaction model genome conformations were generally much more similar to each other. This explains our qualitative observations that the nuclear region occupied by a single chromosome is most condensed in the interaction model (Fig. 4A). We conclude that the chromosomal interactions, as determined by proximity-based ligation, play a key role in the distinct positioning of yeast chromosomes within our models. Functional elements form distinct clusters. It is plausible to think that genomic elements that are functionally related are clustered within the nucleus in order to promote their coregulation. However, this kind of regional juxtaposition is not necessarily close and stable enough to lead to cross-linking of the elements by formaldehyde. Therefore, we investigated if groups of similar functional genomic elements show evidence of spatial clustering within our models. We chose four functional elements: tRNAs (tRNA); early firing replication origins; 6,27
Nucleus
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
Figure 2. Reconstruction of individual chromosome positions derived from the population of conformations with interactions. Individual positions were obtained for all chromosome segments from all conformations. Segment densities are shown as contour plots. Density quantiles are as illustrated. (A) Chromosome III, (B) Chromosome VII, (C) Chromosome XII, (D) Chromosome VI, (E) Chromosome XIV.
potential G-quadruplex forming sequences (Gquad); and Gal4 protein binding sites (UASGal4 ; Table S1). tRNA clustering is well documented and has been shown to be dependent upon codensin.28-30 Duan et al. identified two tRNA clusters within the yeast nucleus and also reported clustering of the early firing origins of replication.6 Hence these two element sets were included to enable comparisons of our models with independently generated data sets. Gquads are a form of higher order DNA structure that is composed of stacks of four Hoogsteen bonded guanines. Gquads have been functionally implicated in telomere maintenance, transcriptional regulation and ribosome biogenesis.31,32 UASGal4 sequences form part of the regulatory control system that responds to galactose33 but they are also located at genomic positions outside of the galactose genes. Therefore, it is feasible to hypothesize that these elements are clustering in space as part of their regulatory mechanism. In order to measure clustering of elements within our models, we used the density distribution function (DDF;34 Methods). The DDF provides a measure for the degree of spatial clustering within a point set without making assumptions about the size, shape or number of clusters (see Methods). Point sets that are randomly distributed have a constant distribution function
of value 1, while values greater than 1 for small distances indicate element clustering. In contrast, values less than 1 indicate an exclusion volume surrounding the elements in question. The DDFs for each set of elements were calculated individually for each conformation and the averaged functions are shown in Figures 5 and 6. With the exception of the origins of replication, all other elements tested show a large peak in the DDF for distances less than 40 nm, superimposed with a slowly decaying component for distances greater than 80 nm (Fig. 5; highlighted in yellow). The peak at small values must be partly attributed to the fact that the elements are physically linked as part of chromosomes (illustrated in Fig. S3E). To control for this, we chose equal numbers of randomly selected genomic loci for each element under investigation. The DDFs of these randomly chosen samples show a value of approximately 3 for distances less than 40 nm in all three models (Fig. S3). This suggests that the physical linking of the elements within chromosomes contributes to the increase in DDF at distances less than 40 nm. Furthermore, if elements cluster linearly along a chromosome, this will by necessity contribute to the increase in DDF at short distances (Fig. S3E). Evidence for this linear clustering should also be present within the confined
www.landesbioscience.com Nucleus
373
©2012 Landes Bioscience. Do not distribute
Figure 3. Color reconstruction of preferred chromosome positions in exponentially growing respiro-fermenting yeast cells. Reconstructions were made as in Figure 2 and are viewed at 120° angles rotated about the spindle pole body—nucleolus axis. The spindle pole body is located at the top of each image. (A), preferred positions of the five smallest chromosomes: VIII, red; IX, orange; III, green; VI, light blue; I, dark blue. (B), preferred positions of the five largest chromosomes: IV, red; XV, orange; VII, green; XII, light blue; XVI, dark blue. Chromosomes are shown at 75% density quantile.
model. Comparisons between the interactions and confined models confirm that the peak at small distances is accounted for by physical linkage and linear clustering for the tRNAs, Gquads and UASGal4 elements (Fig. 5B–D). By contrast, the DDFs for the origins of replication show variable levels of exclusion below 80 nm dependent upon origin type and the presence or absence of the interactions (Fig. 6). Together these findings are consistent
374
with earlier observations that early origins of replication27 and other functional elements35 are not randomly located along chromosomes. While the initial peak of the tRNAs, Gquad and UASGal4 density distribution functions is largely explained by linkage and linear arrangement, the slowly decaying component (> 80 nm) is not (Fig. 5B–D). Rather, at distances greater than
Nucleus
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
Figure 4. Inter- and intra-chromosomal interactions make the major contribution to the genome organization within the population of conformations. (A) Chromosome I position and compaction is significantly influenced by the addition of the interaction data. Chromosome I contour map in the: i, confined model; ii, constrained model; and iii, interaction model. (B) The addition of inter- and intra-chromosomal interactions caused a major shift in the root mean squared distances (RSMD) distribution. Addition of the external constraints reduced the mean RMSD from 6,790 to 5,937 (p < 10 -15 in a t-test on 125 independent conformation pairs chosen from each model; Supplemental Methods). The subsequent addition of the chromosomal interactions led to a further reduction of the mean RMSD to 4261 (p < 10 -15, t-test as above). Populations of conformations were rotated to minimize the RMSD between the chromosomes centers of mass. Histograms were calculated for RSMD between pairs of conformations (see text).
www.landesbioscience.com Nucleus
375
©2012 Landes Bioscience. Do not distribute
clustering is accounted for by this general compaction effect (Fig. 5C). However, general genome compaction accounts for a smaller percentage of the tRNA (47%), early (35%) or late origin (55%) or UASGal4 (63%) clustering as the element specific DDFs remain substantially higher than the random fragment DDFs (Figs. 5B, D and 6). In contrast to the situation for the Gquad and UASGal4 DDFs, the constrained DDF for tRNAs and early firing origins is higher than that for random fragments within the interaction model (Figs. 5B and 6B). This suggests that the external constraints play an important role in the clustering of these elements. These results are consistent with observations of a centromeric and nucleolar clustering of tRNAs that have previously been reported in yeast.6,28 However, they do not preclude the possibility that the regulation of transcription also plays a role in tRNA clustering.28-30 tRNAs and early firing origins of replication form stable clusters. The density distribution functions of tRNAs, early origins and UASGal4 sequences indicate that the chromosomal interactions cause specific clustering of these elements. This raises the question if the same elements cluster consistently with the same partners despite the fact that individual conformations incorporate different sets of chromosomal interactions. To answer this question we used a clustering algorithm based Figure 5. Genomic elements cluster within populations of conformations. The inclusion of on point density to detect clusters in each interactions causes significant inter-element clustering that is not explained by linear chrogenome conformation (see Methods). mosome structure. (A) cartoon depicting the three models that were analyzed. (B–D) DDFs The early firing origins show consistent for tRNAs, g-quadruplexes and Gal4p binding sites (UASGal4), respectively. Blue lines, confined clustering, which is initially defined by the model; red lines, constrained model; black lines, interaction model; grey lines, randomly introduction of constraints into the models chosen fragments from the interaction model. The number of random fragments was equal to the number of element fragments under consideration (Supplemental Material). Yellow (Fig. 7). The subsequent introduction of interhighlights the DDF for distances > 80 nm. The broken black line depicts a density ratio of 1. actions into these models stabilizes the nonCDR (non-CLB5p dependent regions) and 80 nm, the sequential addition of the constraints and the inter- unchecked origin (unchecked by RAD53p 27) cluster formation actions results in increases in the DDF values (Figs. 5 and 6). such that clusters (Table S2) are reproduced in up to 50% of the Comparisons of the constrained vs. confined and interaction population (Fig. 7). Our cluster of 11 origins is a subset of the vs. constrained model DDFs identified these increases as being early firing origins identified by Duan et al. The identification highly significant for all elements (t-test, all p values < 10 -15 ; of our early firing cluster as a subset of Duan et al. is consistent see Supplemental Methods). The shift in the interaction model with our identifying stable associations and not origins which are can partly be explained by an increase in general genome com- capable of interacting in an unknown proportion of the populapaction because random fragments also exhibit a shift in their tion. Consistent with their low levels of clustering in the DDF, DDF values upon the addition of the interactions (Fig. S3). the late firing origins do not form any detectable stable clusters The incorporation of constraints into the model makes a large using these parameters (data not shown). Similar to the early firing origins, the tRNAs form into at least contribution (58% of the area between the curve and a density ratio of 1 above a clustering distance of 80 nm) to early origin two clusters which reproducibly contain the same partners in clustering. However, it contributes much less (9% of the area) ≥ 15% of the conformations within the population (Fig. 8C and E, to the clustering that is observed for the late origins (Fig. 6). Tables 1 and 2). This clustering is emphasized by the inclusion Comparisons between the interaction model DDFs for ran- of interactions (Fig. 8B–D) although the core structure is already dom fragments and Gquad elements show that 78% of Gquad visible in the constrained model (Fig. 8B).
376
Nucleus
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
The tRNAs within cluster one and two are a subset of the tRNAs that Duan et al. identified as seeming to cluster with centromeres.6 This is important as it confirms that the models we generated, using independently obtained proximity data, reproduce previous observations. Critically, our data identify a spatial distinction between clusters one and two which was not present in Duan et al. This difference can be attributed to our use of a stricter non-hierarchical clustering methodology and likely represents a finer distinction of this cluster. The tRNA genes that form cluster one are centromere-proximal (distance less or equal than 23,066 ± 12,266 (mean ± SD), range 3,102–5,1870 bp). However, we clearly demonstrate that not all centromere proximal tRNAs form part of cluster one. In fact, only 60% of the tRNA genes that fall within 35,332 bp of centromeres (mean + 1 SD) reproducibly (i.e., > 15% of conformations) form part of this cluster. These results indicate that, for cluster one, tRNA clustering is affected by the centromere proximity. However, the intensification of tRNA clustering upon the addition of the interactions indicates that clustering is not solely determined by the centromere proximity (compare Fig. 8B–D) rather chromosomal interactions contribute to the positioning and clustering of these Figure 6. Early and late firing origins of replication have different clustering characteristics. Similar genomic elements. to our observations for the tRNAs, g-quadruplexes and UASGal4 elements (Fig. 5), the inclusion of Duan et al. also identified a nucleolar interactions causes significant clustering of origins of replication. However, origins of replication cluster6 which is not reconstituted by our have exclusion zones, within which interactions do not occur (highlighted in yellow) indicating that models. This is not surprising because they are spaced throughout the nucleus. (A–D) DDFs for late firing origins of replication, early firing origins, non-CDR origins and unchecked early origins, respectively.6,27 Models are as depicted in Figthe repetitive nature of the rDNA preure 5A; blue lines, confined model; red lines, constrained model; black lines, interaction model; grey vents the accurate positioning of interlines, randomly chosen fragments from the interaction model. The number of random fragments actions involving the rDNA along the was equal to the number of element fragments under consideration (Supplemental Material). Yelnuclear-nucleolar boundary within our low highlights the DDF for distances < 80 nm. The broken black line depicts a density ratio of 1. models. Similar to the situation observed for the tRNAs, the UASGal4 and Gquad elements show a propensity grained molecular models of the S. cerevisiae genome isolated the to be close to other elements of the same type (Fig. 5). However, effects of the external constraints and clearly demonstrated that the composition of the UASGal4 and Gquad clusters is dynamic distinct chromosomal positioning and functional clustering is (Figs. S4 and S5). largely driven by the chromosomal interactions. Single 3-dimensional representations of the S. cerevisiae 6 Discussion and S. pombe9 genomes have been produced from empirically obtained proximity-based ligation data that have been converted We set out to determine how much the genome organization into linear distance approximations. These single structures depends upon the specific subset of interactions that are cur- have been used to make inferences about the organization of the rently present within it; and what contribution the nuclear mem- respective genomes.6,9 brane and gross localization effects (e.g., centromeres, telomeres Microscopic studies show considerable variation in the disand the nucleolus) make to the overall organization. Our coarse tances between pairs of loci within populations of cells10,11,13 and
driven transcription is inhibited, raises the possibility that these interactions repress transcription. Additionally, the stochastic nature of gene expression40 may be reflected in the fact that the clustering occurs in a small portion of the population of conformations (Fig. 8; Figs. S4 and S5). Both of the tRNA clusters we observed were centromere proximal. The fact that we did not detect previously observed nucleolar clusters28,30 in our conformations may result from the simplified nucleolar constraints we implemented. Due to the repetitive nature of the rDNA it is impossible to use proximitybased ligation methods to determine the internal nucleolar structure or the specific rDNA repeat other loci interact with. Yet, there are clearly nucleolar associated sequences.2,26,41 Therefore, excluding rDNA interactions from the model6 causes the nucleolus to become artificially disjoint from the rest of the genome. To address these conflicting issues, we incorporated only two rDNA repeats in our models and implemented interactions between the rDNA and other loci as attractive forces toward the nucleolusnucleoplasm boundary. While this conserves interactions with the rDNA and precludes artificial clustering to one of two rDNA repeats, it may not fully resolve the clustering of elements along the nucleolar boundary. McCune et al. observed linear clustering of large blocks of origins with similar times of activation and argued that this led to a genomic organization of temporal blocks of early and late S-phase replication.27 Duan et al. observed clustering of early origins, but not late origins in their interaction library.6 Both non-CDR and unchecked origins27 showed stable clustering in our analyses. The non-CDR and unchecked origins are the earliest firing origins in S phase and are generally more efficient.27 We propose that the stable clustering we observed can explain origin competence and thus contributes to temporal firing patterns. That is, the clustered origins are biochemically competent to fire as clustering promotes high concentrations of the origin binding proteins and maintains high occupancy rates, consistent with ideas surrounding looping42 and replication factories.43 Our results are also consistent with observations that relocating origins to ectopic sites alters their firing time,44 particularly since the addition of constraints into the model initiates the clustering pattern. This study has demonstrated the feasibility of coarse grained molecular modeling for the reconstruction of yeast genome organization. Our approach has generated an ensemble of conformations that reflect the dynamic nature of genome organization. This has enabled the isolation of the external and internal factors that govern this organization. Future development of these models through the inclusion of cell cycle, immunoprecipitation and transcriptional data will deepen our understanding of the structure function relationship through investigations into the formation and existence of sub-nuclear domains and chromosome segregation. Methods Chromosome interaction data. Chromosome interaction data were derived for Saccharomyces cerevisiae BY4741 [Mata his3Δ1
www.landesbioscience.com Nucleus
377
©2012 Landes Bioscience. Do not distribute
therefore it is generally assumed that global genome organization changes from cell to cell. The dynamic properties of the genome cannot be represented in a single conformation. Therefore, we generated an ensemble of genome conformations in order to represent the dynamic nature of genome organization. However, proximity-based ligation identifies interactions from a population of cells. As such, it is not technically possible to determine the total set of interactions that co-exist within a single cell. We addressed this limitation by randomly selecting interacting loci from the data pool for each simulation run. While this does not solve the problem entirely, as we do not reproduce the exact sets of interactions present in the population, this approach allows us to identify the most robust features of genome organization. The fact that there was significant organizational overlap within the model population supports our approach. The yeast genome is clearly restricted by external constraints including the nuclear membrane and the gross localization of the centromeres, telomeres and nucleolus. However, the extent of these restrictions was unknown when compared with the effect of chromosomal interactions. To address this question we restricted the positioning of the centromeres, the rDNA and telomeres in our models. Although, there is empirical evidence that rDNA 36 and telomeres37-39 can move out of their preferred locations it is unclear what determines their positioning. Therefore, to reduce the complexity of the model, we restricted centromere, rDNA and telomere movement to specified nuclear zones. Our results clearly showed that while the constraints have an effect on global genome structure the addition of the chromosomal interactions makes a non-trivial contribution to genome organization. The chromosomal interactions were clearly responsible for the largest change in the global positioning of chromosomes, caused the largest reduction in inter-conformation variation, and had the greatest effect upon functional element clustering. This indicates that the external constraints alone are insufficient to uniquely determine genome structure. Thus, it is reasonable to conclude that the genome maintains flexibility to rapidly adapt its structure to different conditions without disrupting the external constraints. Independent of the question of whether structure determines function, the fact that the interactions determine large parts of the structure means that a correlation between structure and function is possible. Such a link is supported by the fact that tRNA genes are co-regulated and have been shown to cluster spatially in our models and in microscopic studies.28,30 We observed two tRNA clusters, one of which contained both copies of the CCU decoding proline tRNA [i.e., tP(AGG)N and tP(AGG)C]. Critically, the CCU codon is the second most commonly used proline codon (www.kazusa.or.jp/codon/cgi-bin/showcodon. cgi?species=4932) and therefore the membership of these two tRNAs within this cluster suggests that it is transcriptionally active. This is consistent with microscopic studies that identify clustering of transcriptionally active tRNAs.28,30 The observation of interaction dependent clustering of the UASGal4 and Gquad transcription regulatory elements provides further support for the structural transcriptional linkage in yeast. Specifically, the clustering of the UASGal4 loci in glucose grown cells, when galactose
©2012 Landes Bioscience. Do not distribute Figure 7. Early origin cluster compositions are highly reproduced within the population. Early replicating origin clustering was determined for all individual conformations within each model. Inter-origin clustering is shown as a matrix in which the rows and columns represent all early origins, early non-CDR or early unchecked origins.6,27 Thus, the entry in the ith row and jth column indicates clustering between origins i and j. Black means that the origins were never in the same cluster and red indicates the origins were clustered in more than 50% of the conformations. While interaction patterns are observed upon addition of the constraints (B and E), the addition of interactions further stabilized the clustering and introduced a new overlapping cluster. All early origin clustering observed in: (A), the confined model; (B), the constrained model; (C), the interaction model. Early non-CDR origin clustering observed in: (D), the confined model; (E), the constrained model; (F), the interaction model. Unchecked origin clustering observed in: (G), the confined model; (H), the constrained model; (I), the interaction model. Clustering was determined using the clustering algorithm (see text). Clusters with a minimum of 100 nm regular tetrahedron edge length and a maximum edge length of 150 nm were identified (see Methods).
378
Nucleus
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute Figure 8. For figure legend, see page 380.
www.landesbioscience.com Nucleus
379
leu2Δ0 met15Δ0 ura3Δ0] cells, growing exponentially in glucose, using genome conformation capture (GCC) according to Rodley et al.2 The chromosome interaction network was assembled from a total of 23,735,888 paired end sequences for a total of 1,449,269 unique interactions (interaction files are available upon request; Sequences are banked with GEO 33018). Polymer chain simulations. Polymer modeling rationale. Chromosomes were modeled as coarse grained polymer chains using the classical molecular dynamics simulator LAMMPS.45 Different sets of constraints were imposed on chromosome movement including positioning of centromeres, telomeres and the rDNA (rDNA; Supplemental Methods). Experimentally detected inter- and intrachromosomal interactions were incorporated as harmonic forces. A different subset of the interactions was used in each simulation run in order to reflect the variability in a population of cells. Chromatin density. Chromatin exists at different densities (euchromatin and heterochromatin) at different locations within the genome of an individual cell. We employed two different models to represent compact (30 nm diameter) and open (10 nm diameter) chromatin fibers. Since we are not simulating chromatin using all-atom molecular dynamics, but a coarse grained approach, the linear mass density of the chromatin fiber is an important parameter for our simulations. Compact chromatin was modeled as 30 nm fibers with a linear mass density of 130 bp/nm (or 43x compaction compared with naked DNA) and a persistence length of 200 nm.13 The 30 nm polymer building blocks are designated within the models as spherical segments of 30 nm diameter each of which represent DNA-strand partitions of 3,900 bp (Fig. S1). Since stiffness was considered explicitly this model resembles the worm-like chain (WLC) model for polymer fibers.46 Open chromatin was modeled with a linear mass density of ca 22 bp/nm and a persistence length of 30 nm.13 Angular elasticity was not used for this fiber. Instead, it was represented as a freelyjointed chain with a segment length of 60 nm (the Kuhn length that corresponds to a persistence length of 30 nm).46 Because chromosomes can switch from a condensed to open chromatin form, both types of chromatin fiber (open and compact) were used alternately as part of the modeling of a single polymer chain. Regions of compacted chromatin were identified based on short range intrachromosomal interactions (Supplemental Methods). In our simulations, 70% of the genome was compacted. Model types. We ran simulations with three different sets of constraints. The simplest model is the ‘confined model’ in which the only constraint is that the chromosomes are enclosed into a sphere of 1,000 nm radius.
380
In the “constrained model” we added the restrictions that the centromeres are elastically attached to a fixed point at the nuclear periphery (representing the spindle pole body) 13 and telomere movement is restricted to the outermost shell of the nucleus (less than 200 nm from the periphery).14 The nucleolus(rDNA) is located in a region of the nucleus opposite of the spindle pole body.10 For simplicity, this separation was modeled by restricting the rDNA to the nucleolus by a plane located 700 nm from the center. The “interaction model” extends the “constrained model” by adding harmonic forces between certain segment pairs. These forces represent the experimentally identified chromosomal interactions. The highly repetitive rDNA is present as a tandem repeat that comprises 18.2 kb across 14 uncondensed simulation fragments. Hence, applying the aforementioned method for introducing interactions between rDNA and other DNA segments would pose a problem as all the interactions could converge on a single point. It is plausible that the bulk of the chromosomal interactions between the rDNA and non-rDNA loci occur at the surface of the nucleolus, since non-rDNA sequences are excluded from the nucleolus. Therefore, interactions between rDNA and non-rDNA segments are represented by forces that drive the nonrDNA segment toward the plane that separates the nucleolus from the rest of the genome. Initial simulations for chromosome positioning. If the interaction forces are applied to randomly positioned chromosomes the simulations are likely to get stuck in local optima. Hence in the interaction model we incorporated an initial simulation, in which each of the 16 yeast chromosomes is condensed to a single point “pseudochromosome” and positioned relative to all other chromosomes in an interaction informed manner. A harmonic bond force connects each of the 16 yeast pseudochromosomes. The relaxed length of this bond is dependent upon the number of interactions between each pair of chromosomes. After this initial positioning, each pseudochromosome is expanded into the polymer chain representing the chromosome. Conformation analysis. Similarity of conformations. Several of our analyses require a measure for the degree of similarity of different genome conformations resulting from our simulations. The most straight-forward of these measures is the root mean square difference (RMSD) between two conformations. If the segment positions of one conformation are given by the vectors ai and those of the other conformation by bi then the root mean square difference d between the two conformations is given by
Nucleus
d = (Σ(ai - bi)2)(1/2).
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
Figure 8 (See previous page). tRNA cluster compositions vary within individual conformations in the population. tRNA clustering was determined for all individual conformations within each model. Inter-tRNA clustering is shown as a matrix in which the rows and columns represent 268 individual tRNAs (Table S1). Thus, the entry in the ith row and jth column indicates clustering between tRNAs i and j. Black means that the tRNAs were never in the same cluster and red indicates the tRNAs were clustered in more than 50% of the conformations. (A), tRNA clustering observed in the confined model; (B), tRNA clustering observed in the constrained model; (C), tRNA clustering observed in the interaction model; (D), absolute difference matrix highlighting the changes in cluster stability between the constrained and interaction models; (E), tRNA cluster one (black) and two (red). Clustering was determined using the clustering algorithm (see text). Clusters with a minimum of 100 nm regular tetrahedron edge length and a maximum edge length of 150 nm were identified (see Methods).
Table 1. tRNA genes that are present within the centromeric zone and form part of cluster one tRNA gene identity In
a
# tRNA genes
Out
b
tA(AGC)H tD(GUC)I1
# tRNAs within CZ
Total
Within CZ
In cluster
Out of cluster
Codon frequency (per 1,000 codons)e
11
1
1
0
21.2
c
d
tA(UGC)A
5
1
0
1
16.2
tD(GUC)G1
16
3
2
1
20.2
tE(UUC)C
14
4
3
1
45.6
10
2
2
0
18.4
16
5
4
1
9.8
tD(GUC)I2 tE(UUC)E1 tE(UUC)I tF(GAA)F
©2012 Landes Bioscience. Do not distribute
tE(UUC)M tF(GAA)P1 tG(GCC)C
tG(GCC)F2
tG(GCC)F1 tG(GCC)O2 tG(GCC)P1 tK(CUU)C tK(CUU)P
tI(AAU)N2
13
1
0
1
30.1
tK(CUU)E1
14
4
2
2
30.8
10
3
0
3
27.2
tK(CUU)J tL(CAA)A tL(CAA)C tL(CAA)K tL(UAA)J
7
1
0
1
26.2
tM(CAU)C
tM(CAU)J2
10
2
1
1
20.9
tN(GUU)C
tN(GUU)F
10
3
2
1
24.8
2
2
2
0
13.5
tN(GUU)N2 tP(AGG)C tP(AGG)N tP(UGG)O1
tP(UGG)A
10
2
1
1
18.3
tQ(UUG)H
tQ(UUG)E2
9
2
1
1
27.3
6
1
1
0
6.4
tR(UCU)E
11
1
0
1
21.3
11
6
3
3
23.5
tR(ACG)O tS(AGA)H
tS(AGA)A
tS(AGA)M
tS(AGA)B
tS(AGA)D1
tS(AGA)L
tS(GCU)O tT(AGU)D
tT(AGU)B
3
1
1
0
9.8
11
5
4
1
20.3
14
1
1
0
22.1
tT(AGU)H tT(AGU)I2 tT(AGU)O2 tV(AAC)H tW(CCA)J tY(GUA)F1 tY(GUA)O
6
1
0
1
10.4
8
2
2
0
14.4
tRNA genes that are linearly located within 35,332 bp of yeast centromeres and are present within the cluster. btRNA genes that are linearly located within 35,332 bp of yeast centromeres yet are absent from the cluster. cThe total number of identical tRNA genes that are present in the yeast genome. d CZ, centromere zone (i.e., < 35,332 bp from centromere). edata obtained from www.kazusa.or.jp/codon/cgi-bin/showcodon.cgi?species=4932. a
www.landesbioscience.com Nucleus
381
References 1. Gruber S, Errington J. Recruitment of condensin to replication origin regions by ParB/SpoOJ promotes chromosome segregation in B. subtilis. Cell 2009; 137:685-96; PMID:19450516; http://dx.doi. org/10.1016/j.cell.2009.02.035. 2. Rodley CD, Bertels F, Jones B, O’Sullivan JM. Global identification of yeast chromosome interactions using Genome conformation capture. Fungal Genet Biol 2009; 46:879-86; PMID:19628047; http://dx.doi. org/10.1016/j.fgb.2009.07.006. 3. Lieberman-Aiden E, van Berkum NL, Williams L, Imakaev M, Ragoczy T, Telling A, et al. Comprehensive mapping of long-range interactions reveals folding principles of the human genome. Science 2009; 326:28993; PMID:19815776; http://dx.doi.org/10.1126/science.1181369. 4. Fullwood MJ, Liu MH, Pan YF, Liu J, Xu H, Mohamed YB, et al. An oestrogen-receptor-alpha-bound human chromatin interactome. Nature 2009; 462:5864; PMID:19890323; http://dx.doi.org/10.1038/ nature08497. 5. Iwasaki O, Tanaka A, Tanizawa H, Grewal SIS, Noma KI. Centromeric localization of dispersed Pol III genes in fission yeast. Mol Biol Cell 2010; 21:25465; PMID:19910488; http://dx.doi.org/10.1091/mbc. E09-09-0790.
382
Table 2. tRNA genes that are present within cluster two tRNA gene identity
Chromosome
Start…Stop
tG(GCC)F2
VI
181044…180974
tS(GCU)F
VI
191613…191513
tA(AGC)F
VI
204996…204924
tD(GUC)G1
VII
531610…531681
tE(UUC)G3
VII
541850…541921
tD(GUC)G2
VII
544648…544577
the Supplemental Material for a detailed description of the DDF calculation. Cluster composition. Cluster composition was analyzed using a clustering algorithm based on point density. The underlying premise of this analysis is that clusters are regions of high point density. Density was estimated using the Delaunay triangulation of the point set. Clusters were identified as points connected by small tetrahedra in the triangulation. See Supplemental Material for details. Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed. Acknowledgments
J.O.S. acknowledges support from the Auckland Medical Research Foundation, Massey University Research Fund and Marsden Fund. L.R.G. was supported by a Swiss National Science Foundation fellowship (PBBSP3-130910). C.R. was funded by an HRC PhD scholarship (08/554). J.L. acknowledges support by a collaboration grant from the German Federal Ministry of Education and Research. The authors thank Susan Gasser for her advice and support. Supplemental Material
Supplemental materials may be found here: www.landesbioscience.com/journals/nucleus/article/20971
6.
Duan Z, Andronescu M, Schutz K, McIlwain S, Kim YJ, Lee C, et al. A three-dimensional model of the yeast genome. Nature 2010; 465:363-7; PMID:20436457; http://dx.doi.org/10.1038/nature08973. 7. Grand RS, Gehlen LR, O’Sullivan JM. Methods for the investigation of chromosome organization. In: Urbano KV, Ed. Advances in Genetics Research. New York: Nova Publishers 2011. 8. Dekker J, Rippe K, Dekker M, Kleckner N. Capturing chromosome conformation. Science 2002; 295:130611; PMID:11847345; http://dx.doi.org/10.1126/science.1067799. 9. Tanizawa H, Iwasaki O, Tanaka A, Capizzi JR, Wickramasinghe P, Lee M, et al. Mapping of longrange associations throughout the fission yeast genome reveals global genome organization linked to transcriptional regulation. Nucleic Acids Res 2010; 38:816477; PMID:21030438; http://dx.doi.org/10.1093/nar/ gkq955. 10. Bystricky K, Laroche T, van Houwe G, Blaszczyk M, Gasser SM. Chromosome looping in yeast: telomere pairing and coordinated movement reflect anchoring efficiency and territorial organization. J Cell Biol 2005; 168:375-87; PMID:15684028; http://dx.doi. org/10.1083/jcb.200409091.
Nucleus
11. Schober H, Kalck V, Vega-Palas MA, Van Houwe G, Sage D, Unser M, et al. Controlled exchange of chromosomal arms reveals principles driving telomere interactions in yeast. Genome Res 2008; 18:26171; PMID:18096749; http://dx.doi.org/10.1101/ gr.6687808. 12. Berger AB, Cabal GG, Fabre E, Duong T, Buc H, Nehrbass U, et al. High-resolution statistical mapping reveals gene territories in live yeast. Nat Methods 2008; 5:1031-7; PMID:18978785; http://dx.doi. org/10.1038/nmeth.1266. 13. Bystricky K, Heun P, Gehlen L, Langowski J, Gasser SM. Long-range compaction and flexibility of interphase chromatin in budding yeast analyzed by highresolution imaging techniques. Proc Natl Acad Sci USA 2004; 101:16495-500; PMID:15545610; http:// dx.doi.org/10.1073/pnas.0402766101. 14. Gotta M, Laroche T, Formenton A, Maillet L, Scherthan H, Gasser SM. The clustering of telomeres and colocalization with Rap1, Sir3 and Sir4 proteins in wild-type Saccharomyces cerevisiae. J Cell Biol 1996; 134:1349-63; PMID:8830766; http://dx.doi. org/10.1083/jcb.134.6.1349.
Volume 3 Issue 4
©2012 Landes Bioscience. Do not distribute
Alignment of conformations. The external constraints (i.e., centromeres, telomeres and nucleolus positioning) that are imposed on the chromosomes are rotationally symmetric with respect to the axis defined by the spindle pole body and the center of the nucleolus. Therefore, a conformation fulfills these constraints equally well if it is rotated around this axis by an arbitrary angle. Moreover, the distances between pairs of DNA segments are not changed if the entire genome is rotated. As such, analyses that involve the combination or comparison of the results of different simulation runs are only meaningful if the genome conformations have been previously aligned to each other. Rotational alignment was attained by finding the orthogonal transformation (i.e., rotation or combined rotation and reflection) that minimizes the RMSD between the conformations. However, only a small part of the genome is directly influenced by the presence of constraints and interactions within the models and the flexibility of the unconstrained regions contributes substantially to the RMSD. Therefore, to reduce this noise and focus on the alignment of the global genome arrangement, we calculated the center of mass for each chromosome within each conformation. Optimal alignment matrices were then calculated using an analytical algorithm,47,48 to find the orthogonal transformation that minimizes the RMSD between these simplified point sets. Density distribution functions. The density distribution function (DDF) of a point set is a measure for the “randomness” of the spatial distribution of a set of points.34 The DDF is calculated by considering the average density of points around each point of the set at varying distances. For a random point set, the DDF has a constant value of 1 (apart from noise) because the local densities around the points are all equal to the average point density. An increased value for small distances indicates clustering of points while a decreased value signals an excluded volume around each point. The calculation needs to correct for the smaller volume around points near the border of the nucleus. See
26. O’Sullivan JM, Sontam DM, Grierson R, Jones B. Repeated elements coordinate the spatial organization of the yeast genome. Yeast 2009; 26:12538; PMID:19235779; http://dx.doi.org/10.1002/ yea.1657. 27. McCune HJ, Danielson LS, Alvino GM, Collingwood D, Delrow JJ, Fangman WL, et al. The temporal program of chromosome replication: genomewide replication in clb5Delta Saccharomyces cerevisiae. Genetics 2008; 180:1833-47; PMID:18832352; http://dx.doi. org/10.1534/genetics.108.094359. 28. Thompson M, Haeusler RA, Good PD, Engelke DR. Nucleolar clustering of dispersed tRNA genes. Science 2003; 302:1399-401; PMID:14631041; http://dx.doi. org/10.1126/science.1089814. 29. Haeusler RA, Engelke DR. Spatial organization of transcription by RNA polymerase III. Nucleic Acids Res 2006; 34:4826-36; PMID:16971453; http://dx.doi. org/10.1093/nar/gkl656. 30. Haeusler RA, Pratt-Hyatt M, Good PD, Gipson TA, Engelke DR. Clustering of yeast tRNA genes is mediated by specific association of condensin with tRNA gene transcription complexes. Genes Dev 2008; 22:220414; PMID:18708579; http://dx.doi.org/10.1101/ gad.1675908. 31. Johnson JE, Smith JS, Kozak ML, Johnson FB. In vivo veritas: using yeast to probe the biological functions of G-quadruplexes. Biochimie 2008; 90:125063; PMID:18331848; http://dx.doi.org/10.1016/j. biochi.2008.02.013. 32. Burge S, Parkinson GN, Hazel P, Todd AK, Neidle S. Quadruplex DNA: sequence, topology and structure. Nucleic Acids Res 2006; 34:5402-15; PMID:17012276; http://dx.doi.org/10.1093/nar/ gkl655. 33. Traven A, Jelicic B, Sopta M. Yeast Gal4: a transcriptional paradigm revisited. EMBO Rep 2006; 7:496-9; PMID:16670683; http://dx.doi.org/10.1038/ sj.embor.7400679. 34. Rodieck RW. The density recovery profile: a method for the analysis of points in the plane applicable to retinal studies. Vis Neurosci 1991; 6:95-111; PMID:2049333; http://dx.doi.org/10.1017/S095252380001049X. 35. Képès F. Periodic epi-organization of the yeast genome revealed by the distribution of promoter sites. J Mol Biol 2003; 329:859-65; PMID:12798677; http:// dx.doi.org/10.1016/S0022-2836(03)00535-7. 36. Torres-Rosell J, Sunjevaric I, De Piccoli G, Sacher M, Eckert-Boulet N, Reid R, et al. The Smc5-Smc6 complex and SUMO modification of Rad52 regulates recombinational repair at the ribosomal gene locus. Nat Cell Biol 2007; 9:923-31; PMID:17643116; http:// dx.doi.org/10.1038/ncb1619.
37. Tham WH, Wyithe JS, Ko Ferrigno P, Silver PA, Zakian VA. Localization of yeast telomeres to the nuclear periphery is separable from transcriptional repression and telomere stability functions. Mol Cell 2001; 8:189-99; PMID:11511372; http://dx.doi. org/10.1016/S1097-2765(01)00287-8. 38. Hediger F, Neumann FR, Van Houwe G, Dubrana K, Gasser SM. Live imaging of telomeres: yKu and Sir proteins define redundant telomere-anchoring pathways in yeast. Curr Biol 2002; 12:2076-89; PMID:12498682; http://dx.doi.org/10.1016/S0960-9822(02)01338-6. 39. Bupp JM, Martin AE, Stensrud ES, Jaspersen SL. Telomere anchoring at the nuclear periphery requires the budding yeast Sad1-UNC-84 domain protein Mps3. J Cell Biol 2007; 179:845-54; PMID:18039933; http://dx.doi.org/10.1083/jcb.200706040. 40. Raser JM, O’Shea EK. Control of stochasticity in eukaryotic gene expression. Science 2004; 304:18114; PMID:15166317; http://dx.doi.org/10.1126/science.1098641. 41. Németh A, Längst G. Genome organization in and around the nucleolus. Trends Genet 2011; 27:14956; PMID:21295884; http://dx.doi.org/10.1016/j. tig.2011.01.002. 42. Vilar JM, Saiz L. DNA looping in gene regulation: from the assembly of macromolecular complexes to the control of transcriptional noise. Curr Opin Genet Dev 2005; 15:136-44; PMID:15797196; http://dx.doi. org/10.1016/j.gde.2005.02.005. 43. Cook PR. The organization of replication and transcription. Science 1999; 284:1790-5; PMID:10364545; http://dx.doi.org/10.1126/science.284.5421.1790. 44. Ferguson BM, Fangman WL. A position effect on the time of replication origin activation in yeast. Cell 1992; 68:333-9; PMID:1733502; http://dx.doi. org/10.1016/0092-8674(92)90474-Q. 45. Plimpton S. Fast Parallel Algorithms for Short-Range Molecular-Dynamics. J Comput Phys 1995; 117:1-19; http://dx.doi.org/10.1006/jcph.1995.1039. 46. Cabrera JE, Jin DJ. The distribution of RNA polymerase in Escherichia coli is dynamic and sensitive to environmental cues. Mol Microbiol 2003; 50:1493505; PMID:14651633; http://dx.doi.org/10.1046/ j.1365-2958.2003.03805.x. 47. Kabsch W. Solution for Best Rotation to Relate 2 Sets of Vectors. Acta Crystallogr A 1976; 32:922-3; http:// dx.doi.org/10.1107/S0567739476001873. 48. Kabsch W. Discussion of Solution for Best Rotation to Relate 2 Sets of Vectors. Acta Crystallogr A 1978; 34:827-8; http://dx.doi.org/10.1107/ S0567739478001680.
www.landesbioscience.com Nucleus
383
©2012 Landes Bioscience. Do not distribute
15. Cremer M, Grasser F, Lanctôt C, Müller S, Neusser M, Zinner R, et al. Multicolor 3D fluorescence in situ hybridization for imaging interphase chromosomes. Methods Mol Biol 2008; 463:205-39; PMID:18951171; http://dx.doi.org/10.1007/978-159745-406-3_15. 16. Cremer T, Cremer C. Chromosome territories, nuclear architecture and gene regulation in mammalian cells. Nat Rev Genet 2001; 2:292-301; PMID:11283701; http://dx.doi.org/10.1038/35066075. 17. Branco MR, Branco T, Ramirez F, Pombo A. Changes in chromosome organization during PHA-activation of resting human lymphocytes measured by cryo-FISH. Chromosome Res 2008; 16:413-26; PMID:18461481; http://dx.doi.org/10.1007/s10577-008-1230-x. 18. Branco MR, Pombo A. Intermingling of chromosome territories in interphase suggests role in translocations and transcription-dependent associations. PLoS Biol 2006; 4:138; PMID:16623600; http://dx.doi. org/10.1371/journal.pbio.0040138. 19. Parada LA, McQueen PG, Misteli T. Tissue-specific spatial organization of genomes. Genome Biol 2004; 5:44; PMID:15239829; http://dx.doi.org/10.1186/gb2004-5-7-r44. 20. Haber JE, Leung WY. Lack of chromosome territoriality in yeast: promiscuous rejoining of broken chromosome ends. Proc Natl Acad Sci USA 1996; 93:1394954; PMID:8943041; http://dx.doi.org/10.1073/ pnas.93.24.13949. 21. Zimmer C, Fabre E. Principles of chromosomal organization: lessons from yeast. J Cell Biol 2011; 192:72333; PMID:21383075; http://dx.doi.org/10.1083/ jcb.201010058. 22. Vogel F, Schroeder TM. The internal order of the interphase nucleus. Humangenetik 1974; 25:26597; PMID:4464236; http://dx.doi.org/10.1007/ BF00336904. 23. Stahl A, Hartung M, Vagner-Capodano AM, Fouet C. Chromosomal constitution of nucleolus-associated chromatin in man. Hum Genet 1976; 35:2734; PMID:1002162; http://dx.doi.org/10.1007/ BF00295616. 24. Léger I, Guillaud M, Krief B, Brugal G. Interactive computer-assisted analysis of chromosome 1 colocalization with nucleoli. Cytometry 1994; 16:31323; PMID:7988293; http://dx.doi.org/10.1002/ cyto.990160405. 25. Chubb JR, Boyle S, Perry P, Bickmore WA. Chromatin motion is constrained by association with nuclear compartments in human cells. Curr Biol 2002; 12:439-45; PMID:11909528; http://dx.doi.org/10.1016/S09609822(02)00695-4.
