Ribosomal DNA and the nucleolus in the context of genome organization

In recent years, we have come to recognize many levels of chromosome organization in the nucleus. On the macro scale, transcriptionally active and inactive chromatin segregates into global compartments. Locally, the genome is organized into topologically associated domains or TADs (Nuebler et al. 2018). The regions between TADs are referred to as TAD boundaries, and these boundaries tend to be gene dense and highly transcribed. Components of nuclear envelope such as lamins and nuclear pore complexes also contribute to the genome organization by association with heterochromatic and euchromatic regions of the genome, respectively (Fraser et al. 2015). The nucleolus is a prominent nuclear body that contains ribosomal DNA regions from different chromosomes and is a site of ribosomal biogenesis. In this article, we review features of ribosomal DNA genes, turn our attention to proteins that associate with the ribosomal DNA, and then discuss how the rDNA may be organized in three dimensions. Finally, we summarize recent developments pertaining to how the nucleolar compartment may be separated from the nucleoplasm and how it reacts to stress. The latter two sections highlight some of the unanswered questions in nucleolar biology.

The nucleolus is the largest compartment of the interphase nucleus. The major activity that occurs in the nucleolus is the transcription and processing of ribosomal RNA. The long precursor ribosomal RNA (pre-rRNA) molecule is transcribed by RNA polymerase I (RNAPI), and then further processed via cleavage and base modification to produce 18S, 5.8S, and 28S structural ribosomal RNA. These structural RNAs are assembled together with 5S rRNA and ribosomal proteins to generate small and large ribosomal subunits (Russell and Zomerdijk 2005). In all organisms, ribosomal genes are present in multiple repeat units to satisfy the high demand for ribosomes. Processing of the pre-rRNA transcript into the 18S rRNA (for the small ribosomal subunit), the 5.8S, and the 28S rRNA (for the large ribosomal subunit) occurs in the nucleolus. These RNAs are assembled with approximately 80 different ribosomal proteins to form ribosomes (de la Cruz et al. 2015). In plants and animals, 5S rRNA genes, also encoding an essential structural RNA of ribosomes and also present in tandem repeats, have their own cluster in a separate genomic location and are transcribed by RNA polymerase III (RNAPIII), outside of the nucleolus. The biological reason why 5S genes are localized in a separate genomic compartment and are transcribed by a different polymerase remains unknown.

In plants and animals, rDNA genes are organized in tandem repeats that are spread among multiple chromosomes. In human and mouse genomes, rDNA genes are present in several hundred nearly identical copies, and their actual number varies among individuals (Gibbons et al. 2015; Xu et al. 2017). In the human genome, clusters of rDNA genes are partitioned among the short arms (p-arms) of five acrocentric chromosome pairs: 13, 14, 15, 21, and 22 (Fig. 1a) (Henderson et al. 1972). All 5S repeat units in the human genome are positioned on the distal end of chromosome 1. The distribution of rDNA genes is somewhat similar in the mouse genome, where rDNA repeats are on chromosomes 12, 15, 16, 18, and 19 (Fig. 1b), and 5S genes are on the distal end of chromosome 8 (Kurihara et al. 1994; Matsuda et al. 1994). Human 45S genes are positioned closely above the centromere, where they occupy most of the p-arms and are flanked by heterochromatic proximal and distal junctions (PJ and DJ) (McStay 2016). In the mouse genome, where all chromosomes are telocentric, rDNA gene clusters are positioned closely downstream of the centromeres and are surrounded by heterochromatin.

Open image in new window

Nucleoli form around transcriptionally active repeats of ribosomal DNA genes that are also referred to nucleolar organizer regions or NORs. NORs are genetic loci that contain repeated rDNA genes in tandem array. Mammalian rDNA transcription units are 43–45 kb in length (Gonzalez and Sylvester 1995; Grozdanov et al. 2003). Each rDNA repeat unit consists of the coding region producing a precursor transcript (pre-rRNA), often referred to as 45S (or 47S) that is processed into three mature rRNA molecules: 18S, 5.8S, and 28S. Boundaries of the coding region contain external transcribed spacers (5′ETS and 3′ETS), and coding parts of the 45S sequence are separated by internal transcribed spacers (ITS1 and ITS2). Sequences encoding pre-rRNAs are separated by long intergenic spacers (IGSs) approximately 30 kb in length. Mammalian rDNA transcription units are flanked at their ends by one or more terminator sequences that stop the transcript elongation by RNAPI. The rRNA coding sequences are highly conserved in animals, plants, and fungi. However, the IGS sequence is less conserved and contains regulatory elements including promoters and repetitive enhancer elements. There are two promoters within the vertebrate rDNA repeat unit: the gene promoter and the spacer promoter. Transcription of the 45S pre-rRNA is controlled by the gene promoter that consists of two parts—a core promoter and an upstream control element (Haltiner et al. 1986; Learned et al. 1986). The spacer promoter, located within the IGS, can give rise to the non-coding RNA that is thought to participate in gene silencing (Mayer et al. 2008, 2006; Santoro et al. 2010). In the case of active genes, RNAPI elongation from the spacer promoter seems to be immediately arrested by TTF1 at the adjacent spacer termination site (Herdman et al. 2017). The region distal to the spacer promoter contains enhancer elements. The rest of the IGS appears to be devoid of known regulatory elements, containing simple repeated sequences and transposable elements (Fig. 2).

Open image in new window

The first direct observations of the physical structure of transcribed rDNA repeats were made decades ago in chromatin spreads from amphibian oocytes. In amphibian oocytes, the number of rDNA genes is greatly amplified by a mechanism that is still not understood and could reach hundreds to thousands of copies per oocyte. The biological purpose of this rDNA amplification presumably is to meet the high demand for ribosomes during early embryogenesis. These amplified rDNA genes that also somehow become extra-chromosomal form numerous nucleoli (Brown and Dawid 1968). Nucleoli from amphibian oocytes provided ample material to study rDNA by electron microscopy. After treatment with low ionic strength buffer that causes nucleoli to de-condense and the DNA to unfold, rDNA genes with associated transcription machinery were spread onto a slide for subsequent electron microscopy analysis. This method of visualizing transcribed genes was termed the “Miller spread” by the name of its inventor, Oscar L. Miller, Jr. (Miller and Beatty 1969a, b) (Fig. 4).

Open image in new window

Visualization of rDNA repeats by the Miller spread technique allowed precise correlation of morphology with the known rDNA sequence (Bakken et al. 1982; Sollner-Webb and McKnight 1982). Overall, rDNA repeats visualized by this method resemble Christmas trees in a row, separated from each other by a naked stem. Each “Christmas tree” represents a transcribed gene with nascent transcripts attached to the rDNA template. The transcribed portion of the gene is covered by particles containing rDNA transcription factors and RNAPI (Scheer 1987). “Branches” extending from the DNA molecule are nascent pre-rRNA transcripts with terminal knobs that consist of RNA processing machinery (Mougey et al. 1993). The length of these transcripts gradually increases in the direction of transcription: short transcripts are in the beginning stage, while the longest transcripts are near completion. The chromatin axis of transcribed rDNA does not contain nucleosomes (Foe 1978; Scheer 1978). Transcribed regions of rDNA genes are interspersed by a non-transcribed spacer that appears as a region of naked DNA with occasional nucleosomes. Sometimes, short transcripts originating from spacer-promoter sequences can be visualized (Scheer 1987; Williams et al. 1981). To this day, our fundamental understanding of the physical chromatin structure of rDNA is derived in large part from Miller spreads. Much less is understood about how this “Christmas tree” chromatin with all its associated structural and transcriptional machinery is organized and folded in the nucleolus in three dimensions.

The genome of eukaryotic cells is compartmentalized. Chromosomes within the nucleus occupy distinct compartments called chromosome territories (Bolzer et al. 2005; Cremer et al. 1993; Meaburn and Misteli 2007). Chromosomal interaction maps obtained by sequencing techniques such as high-resolution chromosome conformation capture (Hi-C) suggest that chromatin is organized in topologically associated domains (TADs), defined as associated peaks of contact frequency. Formation of TADs has been explained by a “DNA loop extrusion” model, where loops of DNA are extruded through the lumen of SMC complex protein rings (cohesins and/or condensins) and constrained by the insulator protein CTCF (Alipour and Marko 2012; Dixon et al. 2016; Fudenberg et al. 2016; Rao et al. 2014). In this model, the DNA looping process brings together distant genomic elements and effectively compacts chromatin into chromosomes. Topoisomerase binding sites tend to overlap with cohesin and CTCF, indicating that the decatenation activity of topoisomerases may be necessary to relieve topological stress in the process of loop extrusion (Canela et al. 2017). Therefore, SMC proteins, CTCF, and topoisomerases are all part of the basic machinery that organizes chromosomes. Hi-C analysis and mathematical modeling also demonstrated compartmentalization of the genome on a higher level, where large stretches of chromatin tend to associate with each other into active (euchromatic or “A”) and inactive (heterochromatic or “B”) global compartments, simply due to biphasic separation of open (transcriptionally active) and closed (transcriptionally inactive) chromatins (Lieberman-Aiden et al. 2009; Nuebler et al. 2018).

rDNA essentially organizes its own compartment—the nucleolus—that is spatially separated from other chromatin domains and is distinct from A or B compartments. In fact, it is insulated by a “shell” of heterochromatin, discussed further below. rDNA is the most highly transcribed region in the genome. Within the nucleolus, strands of rDNA repeats from multiple chromosomes must be spatially arranged in a manner sustainable for this extensive transcription. Inside the nucleolus, rDNA is loosely packed, except under conditions of nucleolar stress, when the gene clusters form compact “caps” at the nucleolar periphery (van Sluis and McStay 2017). Given its unique sequence and compartmentalization, could the loop extrusion model of genomic organization be applicable to the rDNA?

As previously discussed, basic architectural proteins that organize chromatin into TADs (SMC proteins, CTCF, and topoisomerases) bind rDNA repeats at defined locations upstream of each promoter. All parts of this architectural complex are essential to the transcriptional activity and the stability of the ribosomal DNA in the nucleolus. Yet, certain features of the transcriptionally active rDNA chromatin such as the binding of UBF, lack of nucleosomes, immense number of growing transcripts (“Christmas trees”), and propensity to form R loops may challenge the extrusion of rDNA through the rings of SMC complexes. The repetitive nature of the rDNA sequence prevents analysis by the current Hi-C methods which cannot discriminate different repeats. Therefore, there is no direct evidence for TADs in the rDNA. However, some evidence of topological associations within the rDNA were obtained from a different chromosome conformation capturing method termed 3C that can detect spatial associations between different parts of the same sequence. One of these studies detected interactions between the promoter and the terminator regions of the gene, which can be interpreted as gene looping between the sites of transcription initiation and termination (Nemeth et al. 2008). Another study showed that under conditions of boosting rDNA transcription with c-Myc, upstream and downstream parts of the gene become juxtaposed, which can also be interpreted as formation of DNA loops between transcription initiation and termination sites (Shiue et al. 2009) (Fig. 5a). Another interpretation of these results may be that the inactive portion of the gene becomes looped out, effectively bringing together the termination site of one gene with the initiation site of the other. The non-coding part of the gene would have fewer obstacles for being extruded through SMC complexes (Fig. 5b). This configuration is more in line with the observation that boundaries between TADs tend to be at the highly transcribed regions. Hypothetically, juxtaposing initiation and termination sites would increase the transcription efficiency by promoting re-initiation of the RNAPI transcription within one coding region, or between adjacent coding regions of the rDNA. It remains an open question whether topological associations occur within a single rDNA gene, between neighboring genes in an array of repeats, or perhaps between genes from different arrays co-localized in the same nucleolar compartment.

Open image in new window

The possibility of topological associations between different arrays of rDNA genes implies that rDNA genes from different chromosomes may be physically linked. Indeed, “satellite associations,” between short arms of human acrocentric chromosomes have been observed in early cytogenetics studies, but the physical nature of these associations and their functional consequence has remained unexplored (Ardito et al. 1978; Ferguson-Smith and Handmaker 1961; Jacobs et al. 1976; Zhdanova 1972). Interestingly, the short arms of acrocentric chromosomes are the sites of chromosomal fusion in Robertsonian translocations, one of the most frequent structural chromosomal re-arrangement in humans (Gardner et al. 2011). It is tempting to speculate that satellite associations and Robertsonian translocations are related. For example, inter-chromosomal rDNA interactions could provide physical proximity that specifically facilitates acrocentric fusions.

The nucleolus is surrounded by a “shell” of chromatin, a dense, mostly heterochromatic layer of DNA encircling the nucleolar periphery (Ferreira et al. 1997; Nemeth and Langst 2011; Sadoni et al. 1999). Investigations of chromatin surrounding the nucleolus showed that associations of genomic regions with the nucleolus may be non-random. For instance, earlier studies showed that centromeres tend to be positioned spatially close to the nucleolus in various model systems (Carvalho et al. 2001; Haaf and Schmid 1989; Ochs and Press 1992). Naturally, centromeric heterochromatin of human acrocentric chromosomes flanking the rDNA repeats is associated with nucleoli, because of its physical proximity to the rDNA. However, the centromeres and pericentromeric regions of other non-rDNA chromosomes can also be found at the nucleolar periphery. For instance, consistent association of nucleoli with centromeres of chromosomes 1, 9, and the whole chromosome Y has been documented (Leger et al. 1994; Stahl et al. 1976). An imaging study in budding yeast where certain genomic loci were tagged with fluorescent reporters suggested that the nucleolus was an important landmark for gene positioning and showed that certain genes required for ribosome biogenesis tend to be positioned close to the nucleolus (Berger et al. 2008). Integrating a fluorescent reporter at various sites in mammalian cells revealed that loci in close proximity to the nucleolus are more restricted in their movements than more nucleoplasmic genomic regions, suggesting that these sites may be physically attached to the nucleolar compartment (Chubb et al. 2002).

Genomic studies of perinucleolar chromatin revealed a set of conserved genomic regions associated with the nucleolus that was termed nucleolar-associated domains (NADs) (Nemeth and Langst 2011). Genome-wide mapping of nucleolus-associated chromatin showed that most human chromosomes contain consistent NADs. NAD chromatin tends to contain satellite, centromeric, and pericentromeric DNA, has low gene density, and is enriched in transcriptionally repressed genes (Nemeth et al. 2010; van Koningsbruggen et al. 2010). Besides simple physical proximity, certain proteins were found to have tethering roles in the formation of NADs. For instance, centromere-bound proteins CENPC1 and INCENP were shown to be required for tethering centromeres to the nucleolus, possibly through association with the centromeric alpha-satellite RNA (Wong et al. 2007). Also, the DNA-binding protein CTCF was demonstrated to promote chromatin association with the nucleolus through its interaction with nucleolar protein nucleophosmin, thus recruiting its bound regions to the nucleolar periphery (Yusufzai et al. 2004). Similarly, the Drosophila homolog of nucleophosmin, the nucleoplasmin-like protein (NLP), together with CTCF, is required for anchoring centromeres to the nucleolus (Padeken et al. 2013). Taken together, findings from genomic and imaging studies highlight the role of the nucleolus in maintaining the three-dimensional chromatin organization of the entire nucleus.

The nucleolus self-assembles in interphase around transcriptionally active rDNA genes. Electron microscopy of the animal nucleolus allows visualization of three distinct ultrastructural regions: the fibrillar center (FC), surrounded by the dense fibrillar component (DFC), which is embedded in the peripheral granular component (GC) (Pederson 2011). In addition to DNA and RNA, these layers consist of hundreds of proteins (Boisvert et al. 2007). rDNA transcription occurs within the FC. The processing of nascent transcripts (cleavage of pre-rRNA, rRNA folding, and modification of certain bases) takes place within the DFC, where large and small ribosomal subunits also begin to assemble. Incorporation of the mature 18S rRNA (for the small ribosomal subunit) and the 5.8S and 28S rRNA (for the large subunit) continues within the DFC and GC, where ribosomal proteins and 5S rRNA enter the nucleolus from the outside. When the pre-ribosomal components reach the GC, the assembly of pre-ribosomal subunits is nearly done, and assembled subunits are exported. In this tripartite anatomy, the FC component is populated by the basal rDNA transcriptional machinery including UBF and RNAPI. The DFC is occupied by the early RNA processing factor fibrillarin and other RNA processing and modification factors. The GC is filled with proteins involved in late rRNA processing and assembly (e.g., nucleolin and nucleophosmin) and pre-ribosomal particles. The assembly of functional ribosomes is completed in the cytoplasm, where large and small subunits associate, and a few additional proteins are incorporated in the functional ribosomes (Fromont-Racine et al. 2003; Thomson et al. 2013).

This tripartite anatomy reflects a functional hierarchy within the nucleolus. Importantly, the nucleolus exhibits physical cohesiveness and distinct boundaries without having a membrane. The biophysical properties that allow a nucleolus to exist without a membrane have not been clear, but a new paradigm has emerged for thinking about the nucleolus as a phase-separated body inside the nucleus (Brangwynne et al. 2011). Phase separation refers to the idea that the proteins, RNAs, and other local constituents confer a local viscosity that prevents mixing with the surrounding nucleoplasm. Many nuclear bodies have been proposed to be phase separated, such as the Cajal bodies, P-bodies, and stress granules. In the case of the nucleolus, purified constituent putative RNA-binding proteins such as fibrillarin, an rRNA methyltransferase, and nucleolin, a histone H1 binding protein, were shown to form liquid droplets in vitro. Furthermore, these two proteins, which are constituents of the inner dense FC and the outer GC, respectively, could recapitulate at least one aspect of layered nucleolar anatomy, with nucleolin droplets forming around fibrillarin droplets (Feric et al. 2016).

RNA is proposed to be a key component in many instances of phase-separated nuclear bodies, and naturally, RNA is present at extremely high concentrations in an active nucleolus, making it plausible that ongoing transcription and high RNA levels participate in the separation of the nucleolar compartment from the nucleoplasm. In fact, nucleation by rRNA dictates the precision of nucleolus assembly in Drosophila melanogaster embryos, altering the assembly process from a stochastic nucleation-limited process to a growth-limited, high-precision event (Falahati et al. 2016). The state of transcription and the amount of rRNA influence the kinetics of nucleoli formation (Berry et al. 2015). Furthermore, protein-mediated chromosomal crosslinks may help drive separation of the rDNA from the rest of the chromatin (Hult et al. 2017). Interestingly, when RNAPI transcription is halted, the nucleolar morphology changes dramatically, displaying characteristics of nucleolar stress. Future work aimed at elucidating how nucleolar stress impacts the rDNA structure and the biophysical properties of nucleoli will improve our understanding of the organization of this organelle as a whole. Phase separations are postulated to affect local enzyme reaction rates, nuclear organization, and sequestration of factors (Shin and Brangwynne 2017), all of which are relevant for nucleolar function.

In addition to a high abundance of RNA, major factors for phase separation are proteins with intrinsically disordered domains (Lin et al. 2017). These types of domains have the capacity to interact with many different proteins and adopt different structures. Perhaps most importantly for phase separation, they can promote multimerization. While fibrillarin and nucleolin have been shown to behave as liquid droplets in vitro, it will be interesting to further examine the nucleolar proteome, which consists of hundreds of proteins (Ahmad et al. 2009; Leung et al. 2006), for proteins that can form concentration-dependent aggregates (Khan et al. 2018). In drosophila embryos, some nucleolar proteins appear to be recruited actively to the nucleoli, while others display properties of phase-separated components that act independently of ribosomal DNA (Falahati and Wieschaus 2017). A systematic approach will reveal components of the nucleolar proteome that potentially contribute to the phase separation of the nucleolus from the nucleoplasm and will further highlight the nucleolar processes that contribute to this separation. A more complete understanding will require both biochemical and cell biological approaches, and much remains to be done. For a review on how concepts of phase separation apply to nucleoli, see Mangan et al. (2017).

Cellular stressors that cause the rDNA damage and disrupt rRNA transcription (ionizing radiation, genotoxic agents, and transcription inhibitors) are accompanied by extensive reorganization of the nucleolar structure (Fig. 6). For instance, actinomycin D, a DNA intercalating agent that inhibits transcriptional elongation by RNAPI, causes translocation of the rDNA and associated proteins to the nucleolar periphery where they form compact granules termed “caps” (Boulon et al. 2010). Ionizing radiation and other DNA-damaging agents tend to exert similar effects. Nucleolar “caps” contain condensed rDNA with FC and some DFC proteins (Floutsakou et al. 2013; Shav-Tal et al. 2005). Nucleolar “caps” appear to maintain layers of non-overlapping inner and outer subdomains that contain FC and DFC proteins, respectively (van Sluis and McStay 2017).

Open image in new window

It is not clear which specific changes in rDNA structure manifest as formation of the stress “caps,” and whether the formation of these “caps” is mediated primarily by the rDNA compaction or aggregation of associated proteins. It was proposed that the formation of nucleolar “caps” in response to double-stranded breaks in rDNA sequesters damaged rDNA genes to the nucleolar periphery to facilitate their repair by the homologous recombination machinery (Harding et al. 2015; van Sluis and McStay 2015; Warmerdam et al. 2016). However, changes in rDNA organization and nucleolar biophysical properties upon stress need to be better characterized.

Of note, perturbations in nucleolar function activate the p53 tumor suppressor pathway that causes cell cycle arrest and potentially apoptosis. Activation of p53 in this case involves certain ribosomal proteins, such as RPL5 and RPL11. When rDNA transcription is slowed down or stalled, or if ribosome biogenesis is disrupted for other reasons, free ribosomal subunits remain in excess. These un-incorporated ribosomal components can bind and inhibit the p53 ubiquitin ligase MDM2, causing accumulation of the p53 protein (Holmberg Olausson et al. 2012; Olson 2004; Warner and McIntosh 2009; Zhang and Lu 2009). Un-incorporated 5S rRNA contributes to p53 accumulation as well (Donati et al. 2013; Onofrillo et al. 2017; Russo and Russo 2017). This underscores the important role of the nucleolar structure in sensing cellular stress and maintaining homeostasis.

Nucleolar morphology is not only altered in response to cellular stress, but also appears to change with age. In two recent studies of aging cells, nucleoli tended to get larger with age, and their ribosome production output was increased (Buchwalter and Hetzer 2017; Duncan et al. 2017). The two models for aging examined were Hutchison-Gilford progeria, a rare disease associated with accelerated aging, and oocytes from aged mice. A complimentary study found that small nucleoli and reduced ribosome biogenesis were associated with extended lifespan in worms and mice (Tiku et al. 2017). An integrated explanation for these findings may be that protein homeostasis decreases with age. These studies suggest nucleolar morphology may provide a window into cellular physiology. Large, fused nucleoli have long been used to predict poor prognosis in cancer (Derenzini et al. 2009). In the future, it will be important to understand what drives these morphological changes and their functional significance.

The nucleolus serves as a major organizer for the nucleus, via the rDNA repeats themselves as well as nucleolar-associated domains on other chromosomes, many of which are heterochromatic. Many proteins are associated with rDNA, some of which are specific, such as RNAP1, and some of which are more general, such as topoisomerase, cohesin, and CTCF. In future studies, it will be important to understand how the ribosomal DNA repeats are organized in the nucleolar compartment, the biophysical properties of the compartment, and how these features change under stress and disease conditions. Since nucleolar morphology is an important diagnostic and prognostic factor in clinical pathology (Derenzini et al. 2009), understanding its structure-function relationship would increase its utility in evaluating the cellular state in health and disease.