The links between chromatin spatial organization and biological function

Before the advent of high-throughput molecular biology methods, microscopy was the main method to study the arrangement of chromosomes within the cell nucleus. Molecular biology methods such as ChIP (chromatin immunoprecipitation) are the main approach of studying the interaction between proteins and specific genomic sites. Although microscopy techniques and ChIP are still widely used, the newly developed technique of 3C (chromosome conformation capture) has dramatically increased the observational resolution with respect to genome properties and the nuclear organization of chromosomes [1].

Several different microscopy approaches, including light, electron and fluorescence microscopy, are available today. In this section, we focus on fluorescence-based techniques with respect to the study of nuclear architecture. The nuclear organization is determined through the establishment of reference point positions within the nucleus such as the NM (nuclear membrane), the nucleolus or a certain chromatin region. In studies assessing movement of chromatin from one nuclear compartment to another, nuclear structures might not only be relevant as a reference point, but also indicate a biological function [2,3].

For live-cell imaging purposes, generating a fusion between the protein of interest and a fluorescent tag is an important approach to study chromatin organization. GFP is the most widely used fluorochrome, but other alternatives with different emission wavelengths have emerged in recent years. The resulting hybrid proteins can be visualized directly under the microscope at the same time as stimuli and/or growth conditions are altered, allowing for real-time analysis of relocalization events within the nucleus. Live-cell imaging can also be used to follow the position of a specific locus in the genome. This is done by the integration of lacO repeats into a specific genomic locus together with the integration at another site with a LacR–GFP fusion protein that is able to bind to the lacO repeats, thereby creating a detectable GFP signal from the locus of interest [4,5].

FISH (fluorescence in situ hybridization) is based on fluorescently labelled oligonucleotide probes that bind complementarily to either DNA or RNA. This method cannot be used for live-cell imaging purposes, since cells must be fixed before hybridization with the probe. It has nevertheless revolutionized our understanding of nuclear architecture. With refinements of the technique such as usage of multiple fluorescently labelled probes in combination with advanced image analysis, 3D structures can be generated. In one study, all 24 human chromosomes could be labelled in the nucleus using a technique of combinatorial labelling on the basis of the usage of seven fluorochrome sets [6].

ChIP is the most commonly used technique to study the association between well-defined proteins and genomic regions. The basic approach is the cross-linking of proteins with DNA, for example using formaldehyde, followed by shearing of the chromatin into soluble particles. Subsequently, the protein–DNA complex is pulled down using an antibody with affinity to the protein of interest. When the DNA-binding sequences are known and a comparison is sought between different conditions, the pulled-down DNA can be analysed by PCR. For whole-genome analysis purposes, however, the ChIP technique is now combined with high-throughput techniques such as microarrays (ChIP-chip) or sequencing (ChIP-seq) by newly developed MPS (massive parallel sequencing) techniques. This high-throughput approach has allowed for a wide coverage of protein–DNA interactive components of the human epigenome (http://www.roadmapepigenomics.org).

The 3C methods are primarily focused on structure and interactions at the intra- and inter-organizational level of chromosomes and chromosome regions. 3C was first reported in 2002, and variants of the method have been developed since such as 4C (circularized chromosome conformation capture), 5C (carbon copy chromosome conformation capture) and HiC (genome-wide chromosome conformation capture) [1]. These methods aim to chart long-range chromatin interactions, and the main goal is to generate a 3D representation of chromosomes on the basis of an estimation of the frequency at which different genomic loci are in close proximity to each other. The initial steps include fixation that cross-links the chromatin, followed by, for most purposes, restriction enzyme digestion. The fragmentation, in turn, generates shorter complexes of protein and DNA, and the DNA is then ligated in order to combine sequences that are in close spatial proximity to each other. In a following step, a 3C library is generated with short DNA sequences containing the restriction site in the middle. In the initial 3C approach, the generation of a matrix of interaction frequencies was carried out with either semi-quantitative or quantitative PCR using a range of primer pairs. The aim of this approach is the interactive relationship between two genomic loci. In the 4C approach, one locus of interest is screened against multiple other positions over the whole genome. Analysis is performed with either microarray technology or MPS. In the 5C approach, multiple genomic loci are screened against each other in order to generate shorter-interaction, usually below 1 Mb, frequency matrices. From these data, chromosome conformation can be predicted and visualized in a 3D fashion. The HiC methodology is similar to that of 5C, but the comparison is genome-wide [7].

The DamID method is used to determine association of chromatin to any protein, making use of the enzyme Dam (DNA adenine methyltansferase) from Escherichia coli that creates a DNA modification, namely DNA adenine methylation, normally not present in eukaryotic cells [8]. Briefly, a fusion protein is generated between Dam and a protein of interest, resulting in the methylation of all the genomic loci positioned in the vicinity of the fusion protein. In the next step, the methylated regions are amplified using PCR and analysed by high-throughput techniques, i.e. microarrays or sequencing.

The DNA in the nucleus is highly compacted and organized, in order to enable it to fit within the boundary of the cell nucleus. A necessity considering that the DNA molecules from a single cell, should they be stretched out, would together reach a length of 2 m. The basic units of organization are the nucleosomes, which are spaced in an even array throughout the genome, like beads on a string.

The nucleosome consists of a stretch of 146 bp of DNA wrapped 1.65 turns around a histone octamer [9]. In addition to the canonical histones H2A, H2B, H3 and H4 building up the octamer, there is a wide range of histone variants present in nucleosomes at specific genomic loci whose presence affects the biological function in various ways. The majority of the histone variants reported have been found to replace H3 and H2A, for example CENP-A (centromere protein A). This well-studied histone variant is conserved from yeast to humans and substitutes for histone H3. It is involved in the kinetochore assembly at the central core centromere and thus plays an important role in chromosome segregation during mitosis and meiosis [10]. Another histone variant, H2A.Z, reported to be involved in many biological processes, replaces the canonical histone H2A at certain sites [11]. Among the many functions of H2A.Z, it has been suggested that the exchange of this histone from nucleosomes facilitates DNA double-strand break repair, but it has also been proposed to be involved in the differentiation of embryonic stem cells through interactions with PcG (polycomb group) proteins. In Schizosaccharomyces pombe, this histone variant has been implicated to play a role in chromosome segregation and centromeric silencing [12].

The N-terminal tails of the histones protrude out from the histone octamer and are subjected to post-translational modifications that dictate the accessibility to the DNA by, e.g., transcription factors. Some post-translational modifications can also be inherited through multiple cell divisions in a manner that allows distinct chromosomal loci to retain their specific signature of epigenetic marks [13]. The most commonly targeted amino acids are lysine and arginine, and the most studied modifications are methylation, acetylation, ubiquitylation, SUMOylation and phosphorylation, but many others have discovered, such as glycosylation and crotonylation [14,15].

Euchromatin is characterized by high gene density and is commonly more accessible to transcriptional activation compared with heterochromatin. Histones within euchromatic areas are characterized by acetylation and by H3K4me2/3 (di/tri-methylated histone H3 Lys⁴). Heterochromatin, on the other hand, is characterized by low transcriptional activity, low abundance of genes and enrichment in repetitive DNA sequences. Furthermore, at the chromatin level, heterochromatin is hypoacetylated and has high amounts of H3K9me2/3 (di/tri-methylated histone H3 Lys⁹). This histone modification is in turn important for the recruitment of HP1 (heterochromatin protein 1), called Swi6 in S. pombe, and binds to H3K9me2/3 with its chromodomain. The binding to chromatin induces a conformational switch in Swi6 that drives oligomerization and subsequent spreading of heterochromatin [16]. In addition, in fission yeast, it is clear that RNAi is involved in heterochromatin formation, a process that is dependent on the proteins Ago1, Dcr1 and Rdp1 [17].

Heterochromatin can be divided further into constitutive and facultative heterochromatin. The latter is highly compacted and is defined by a more or less constant epigenetic signature and is found at chromosomal structures such as telomeres and centromeres where it assembles on repetitive DNA sequences. Facultative heterochromatin, on the other hand, is more abundant in terms of gene content and changes in chromatin state, i.e. transcriptionally active compared with state, can occur.

The linear arrangement of nucleosomes comprises the most basic order of chromatin called the 11 nm fibre. This is then probably followed by the formation of a 30 nm fibre that will subsequently organize itself within the limits of each chromosome in the cell nucleus. Very little is known about the driving forces behind chromatin organization, but molecular crowding and interaction with the nuclear periphery seem to be contributing factors [8,18]. The chromatin is surrounded by the NM that is composed of an INM (inner nuclear membrane) and an ONM (outer nuclear membrane). In higher eukaryotes, the INM has a nuclear lamina composed of intermediate filaments, lamins and associated proteins, in contact with chromatin [8] (Figure 1A). Lamins are lacking in fission yeast, but there are the INM proteins Lem2 and Man1, which contain the conserved LEM (Lap2/emerin/Man1) domain [19]. Moreover, these two proteins have overlapping functions together with another IMN protein named Ima1, and all three proteins contacts chromatin [19–21].

In addition, genes encoding tRNA are also involved in organizing the chromatin, since, in both mammalian cells and S. pombe, tRNA genes have been shown to co-localize [22,23]. Moreover, there is growing evidence for TFIIIC (transcription factor IIIC), which is necessary for the transcription of tRNA by RNA polymerase III, to be involved in chromatin organization in mammalian as well as in yeast cells. There is a substantial portion of TFIIIC-bound loci devoid of polymerase III both in human and S. pombe cells [24,25]. In S. pombe, these TFIIIC-bound loci cluster at the nuclear periphery [24].

Finally, repetitive DNA sequences in the genomes contribute to the organization within the cell nucleus. Heterochromatin is found at repetitive sequences, and these tend to cluster together [26]. In S. pombe, the 13 full-length Tf2 retrotransposons, found at different positions along the DNA co-localize at one to three nuclear spots in the cell nucleus via a special type of chromatin formed by conserved CENP-B proteins and the methyltransferase Set1 [27].

The nuclear organization of eukaryotic organisms is highly complex and dynamic. Besides the various structural components in the nucleus, the intra- and inter-chromosomal interactions, as well as contacts between chromatin and nuclear components is not static, but is steadily changing in response to environmental cues such as shifting availability of nutrients and developmental processes [2,3,28].

The chromosomal arrangement in the nucleus is organized into an intricate 3D structure. The interactions between different chromosomal regions play an important role in gene regulation. Mammalian chromosomes occupy discrete areas in the cell nucleus, called CTs (chromosome territories) (Figure 1A). They can be visualized by FISH techniques where unique probes for each chromosome are designed to individually paint the chromosomes [6]. In fission yeast, visualizing chromosome I and II also revealed distinct localization of the chromosomes [29] (Figure 1B). In addition, mapping the DNA interactions by 3C technologies strengthened further the existence of CTs in fission yeast [30]. Despite the defined organization of chromosomes into CTs, looping out of genomic loci does occur. This can happen when transcriptionally active loci translocate to spots in the cell nucleus with high concentrations of polymerase II, so-called transcription factories, at least in erythroid progenitor cells [31]. In S. pombe, there is evidence that genes regulated by the same transcription factor tend to co-localize, indicating the existence of transcription factories in this yeast [30].

Investigating the 3D organization of the DNA in the mammalian cell nucleus by 3C technologies has demonstrated that the chromatin folds into discrete globules of a few hundreds of kilobases [32,33]. Moreover, these domains have a distinct ToR (time of replication) [34,35]. These findings fit very well with the long-term notion that mammalian DNA replication occurs in replication factories that changes positions throughout the S-phase in a distinct spatiotemporal pattern [36]. Although the functional consequences of this organizational aspects is not clear at this point, there is some experimental evidence that some domains change ToR throughout development, which is accompanied by a change in subnuclear localization [37] (see below). Interestingly, the distinct spatiotemporal pattern of DNA replication is also evident in fission yeast [38].

The positioning of the chromosomes in relation to the NM and other structures is of great importance when it comes to gene expression. With the help of DamID, the LADs (lamina-associated domains) in the human genome were found to be enriched in transcriptionally repressed genes [8] (Figure 1A). Furthermore, these regions tend to be flanked by gene promoter regions that point away from the LADs, CpG islands and the protein CTCF (CCCTC-binding factor). When DamID was used to study LADs in single cells, it was observed that the contact of LADs to the nuclear lamina is dependent on the activity of the methyl H3K9 methyltransferase G9a and levels of H3K9me2 [39]. Furthermore, in mammalian cells, the part of the genome associated with the nucleolus, NADs (nucleoli-associated domains) are enriched in poorly expressed genes. Not surprisingly, an overlap between the LADs and the NADs has been reported indicating a similar subnuclear environment of transcriptional repression at the nuclear periphery as well as in the vicinity of the nucleolus [40]. Interestingly, there was a clear correlation between LADs and late ToR. In addition, if the replication time changed during development, there was also a change in the localization with respect to the nuclear lamina [37]. Moreover, in a recent study of chromatin using a large collection of different human cell types, several distinct chromatin profiles were identified: one transcriptionally active and several transcriptionally inactive clusters, one of them high in H3K9me3 [28]. Furthermore, the inactive chromatin states were enriched for contacts with the nuclear lamina [8]. Interestingly, the study by Zhu et al. [28] indicated that in vitro culturing in itself could promote the formation of heterochromatic regions with H3K9me3, and thus NM association, most probably due to a response to serum.

In S. pombe, transcriptionally repressed regions enriched in H3K9me2/3, such as centromeres, telomeres and mating type region, are localized to the nuclear periphery [41] (Figure 1B). In the case of centromeres, they are attached to the NM via Sad1, a SUN (Sad1 and UNC-84) domain protein inserted into the INM that in turn connects to Ksm1, containing a conserved KASH (Klarsicht, ANC-1 and Syne/Nesprin homology) domain, in the ONM. The Sad1–Ksm1 partners have been shown to have overlapping functions with a protein without any conserved domains called Csi1 in S. pombe [42,43]. Telomeres are attached to the NM by the connection of several proteins. Bqt3 is inserted into the NM and contacts Bqt4, which in turn binds Rap1 that finally binds Taz1 bound to the telomeric repeat region [44]. Moreover, both LEM domain-containing proteins, i.e. Man1 and Lem2, are necessary for the proper attachment of telomeres to the nuclear periphery, and Man1 was shown using DamID to be in contact with the subtelomeric regions [19,21]. Interestingly, the attachment of telomeres to the NM is not necessary for telomere integrity or function, since in a bqt4Δ mutant background, telomeres are no longer bound to the NM without any obvious consequences for telomere length or silencing. However, the detachment of telomeres during mitosis by the phosphorylation of Rap1 by Cdc2 and other kinases is critical for proper chromosome segregation [45].

Finally, in a recent study of Atf1-bound stress-response genes, these loci were shown to associate with the nuclear pores together with Dcr1 and other components of the RNAi pathway in S. pombe, thereby allowing the degradation of transcripts under non-induced conditions [46]. Another study also revealed the localization of gene clusters induced by nitrogen starvation to be localized to the NM under non-expressed conditions and to translocate to a more interior nuclear position upon activation [2]. Moreover, during induction and translocation, a drastic fall in nucleosome density over the gene bodies was detected [47].

The correlation between subnuclear localization and transcription has been established in a wide range of organisms. In response to environmental cues, there is a physical relocalization of chromosomal regions between the nuclear periphery and nuclear interior. What remains to be elucidated is whether the movement is a cause or a consequence of transcriptional activation. This question could be addressed by utilizing live-cell imaging combined by tethering chromosomal loci to the NM, preventing the movement, and at the same time quantifying gene expression in the absence or presence of an environmental cue. Another unresolved issue is what drives the movement of chromosomes; is the movement stochastic or motor-driven?

The nuclear organization of eukaryotes exhibits an ordered structure and evolutionary conservation. In the present article, we have touched upon different aspects of how the nuclear organization of chromatin is associated with transcription and replication, with a focus on studies conducted in fission yeast and humans. Studies in these two evolutionarily distant organisms show that the distribution of chromatin and chromosomes are non-random. There is a distinction in spatial proximity of different chromosomal regions with respect to the NM. This distinction is reflected in differences of transcription levels and control of replication timing.

3C

chromosome conformation capture

4C

circularized chromosome conformation capture

5C

carbon copy chromosome conformation capture

CENP

centromere protein

ChIP

chromatin immunoprecipitation

CT

chromosome territory

Dam

DNA adenine methyltansferase

DamID

DNA adenine methyltransferase identification

FISH

fluorescence in situ hybridization

HiC

genome-wide chromosome conformation capture

INM

inner nuclear membrane

LAD

lamina-associated domain

LEM

Lap2/emerin/Man1

MPS

massive parallel sequencing

NAD

nucleoli-associated domain

NM

nuclear membrane

ONM

outer nuclear membrane

TFIIIC

transcription factor IIIC

ToR

time of replication

Members of the Bjerling group and Birger Scholz are thanked for reading and commenting on the paper before submission.

The work was supported by grants from Goran Gustafsson Foundation, the Swedish Cancer Society [grant number 2011/360] and the Swedish Research Council [grant numbers 521-2011-2437 and 621-2011-468].