The Centromere: Epigenetic Control of Chromosome Segregation during Mitosis

Abstract

A fundamental challenge for the survival of all organisms is maintaining the integrity of the genome in all cells. Cells must therefore segregate their replicated genome equally during each cell division. Eukaryotic organisms package their genome into a number of physically distinct chromosomes, which replicate during S phase and condense during prophase of mitosis to form paired sister chromatids. During mitosis, cells form a physical connection between each sister chromatid and microtubules of the mitotic spindle, which segregate one copy of each chromatid to each new daughter cell. The centromere is the DNA locus on each chromosome that creates the site of this connection. In this review, we present a brief history of centromere research and discuss our current knowledge of centromere establishment, maintenance, composition, structure, and function in mitosis.

The centromere is the assembly site for the mitotic kinetochore. It is epigenetically defined by the presence of the histone H3 variant CENP-A.

Centromeres were first described by Walther Flemming (1882) in his pioneering characterization of cell division. Flemming noticed that the distinct threads of chromatin (later called chromosomes) each had one (and only one) site of primary constriction, where the total width of the chromosome appeared smaller than the rest of that chromosome. We now recognize that Flemming was reporting the distinct heterochromatin found specifically at centromeres. The term “chromatin” was originally proposed by Flemming as the chromosomes he was observing stain darkly when cells are treated with aniline dyes. Since then, the definition of chromatin has evolved, and now describes the histone proteins and associated proteins that package DNA.

Although more than 130 years later we lack a detailed understanding of numerous aspects of centromere biology, one universally accepted fact is that centromeres have an essential role in the segregation of chromosomes during mitosis. The centromere is a large chromatin-containing protein complex that forms the assembly site for the mitotic kinetochore, itself a megadalton protein complex that binds spindle microtubules to segregate the chromatids during anaphase (Fig. 1). In addition to microtubule attachment, kinetochores are also the site of mitotic checkpoint activation, which prevents anaphase onset in the presence of unattached kinetochores. Without the centromere, no kinetochore would form and cells could not segregate their chromosomes. Thus, the centromere is of crucial importance for chromosome segregation and mitotic control.

Figure 1.

Introduction to centromere function and organization during mitosis. (A) Before and during the early stages of mitosis, centromeres (green circle) recruit kinetochore proteins (yellow discs). During prometaphase, the kinetochore forms the attachment sites for spindle microtubules (red rods). Once both kinetochores of all sister chromatid pairs are stably and correctly attached to microtubules, pulling forces exerted by microtubules (dashed arrows) cause migration of linked sister chromatids to the metaphase plate. At anaphase, sister chromatid cohesion is dissolved and the centromere and kinetochore harness microtubule-dependent forces that pull each sister chromatid to opposite ends of the dividing cell. (B) Basic architecture of the centromere in mitosis. Centromeric chromatin consists of specialized nucleosomes containing the histone H3 variant centromere protein (CENP)-A. CENP-A recruits a network of centromere proteins (green) that are collectively known as the constitutive centromere associated network (CCAN). Kinetochore proteins (yellow), specifically recruited by the CCAN for mitosis, attach to spindle microtubules.

The mitotic centromere:kinetochore can be thought of as a single multiprotein complex. However, kinetochore proteins localize to centromeres before and during mitosis to enable chromosome segregation, whereas centromere proteins (CENPs) persist throughout the cell cycle. CENPs (listed in Fig. 3) are collectively called the constitutive centromere-associated network (CCAN) (Cheeseman and Desai 2008). This nomenclature system has provided some clarity in the field, but it should be noted that many CCAN proteins dynamically associate and dissociate from centromeres, and some function in microtubule regulation during mitosis. Eukaryotic centromeres vary widely in their complexity and structure, ranging from point centromeres of budding yeast that generate a single microtubule-binding site, to holocentric centromeres of nematodes that decorate the entire chromosome, to regional centromeres of vertebrates that provide a distinct attachment site for multiple microtubules. Despite this variation, the core function of the centromere, to form the kinetochore to bind microtubules so that chromosomes can be equally segregated in mitosis, is conserved. Thus, understanding the molecular underpinnings of the centromere is fundamental to establishing how cells faithfully maintain their genomes.

Figure 3.

Overview of centromere proteins and centromere architecture. Centromere proteins are grouped based on individual complexes, often based on the phenotypic consequence of protein depletion in cells and, more recently, biochemical characterization. CENP-A nucleosomes directly recruit CENP-C and possibly the CENP-N/L heterodimer. CENP-C recruits the CENP-H/I/K/M complex that in turn is required for CENP-T centromere localization. How this occurs remains unclear as specific CENP-C:CENP-H/I/K/M and CENP-H/I/K/M:CENP-T/W/S/X interactions have not been identified. The possibility that CENP-T/W/S/X proteins wrap centromeric DNA in nucleosome-like structures, and whether they associate with H3 nucleosomes at centromeres, is currently a topic of intense research. The functions and subcomplexes that comprise the remaining CENP proteins O/P/Q/R and U remain unclear. Note that both Saccharomyces cerevisiae and Saccharomyces pombe have centromere proteins not listed here (as they lack a known human homolog).

CENP-A: THE CENTROMERE-SPECIFIC HISTONE

Eukaryotic DNA is assembled around histone proteins into protein:DNA complexes known as nucleosomes. Nucleosomes throughout most of a chromosome contain two copies of each histone protein H2A, H2B, H3, and H4, together forming an octameric complex with DNA (Fig. 2). Centromeres are unique from the rest of the chromosome in that they feature nucleosomes containing a histone H3 variant, called CENP-A, in place of histone H3 (Fig. 2). CENP-A was initially discovered as a human autoantigen in CREST syndrome patients, and was subsequently shown to copurify with histones, be present in purified nucleosomes, and be essential across eukaryotic model organisms, as mutation or disruption of CENP-A causes a complete failure in centromere and kinetochore formation (Earnshaw and Rothfield 1985; Earnshaw et al. 1986; Palmer et al. 1987, 1991; Sullivan et al. 1994). In humans, functional centromeres always contain CENP-A, including “neocentromeres” that form on noncentromeric chromosomal loci (Saffery et al. 2000; Lo et al. 2001; Warburton, 2004). Experimental overexpression of Drosophila CENP-A^CID causes the ectopic localization of CENP-A^CID to noncentromere regions, which in turn causes ectopic kinetochore formation and numerous mitotic errors (Heun et al. 2006). When CENP-A^CID is targeted to a noncentromeric region by expressing CENP-A^CID-LacI fusion protein in cells containing chromosomally integrated Lac operator arrays, the localized CENP-A^CID causes ectopic kinetochore formation in Drosophila (Mendiburo et al. 2011), and partial kinetochore formation in humans (Gascoigne et al. 2011). Thus, the presence of CENP-A in chromatin is the defining mark of centromeres.

Figure 2.

CENP-A nucleosome structure and possible variants. A cartoon of a conventional octameric H3 nucleosome is shown, together with the most prominent models of CENP-A nucleosome composition, homotypic CENP-A octamers, heterotypic CENP-A/H3 octamers, and tetrameric CENP-A hemisomes. See main text for specific details.

The Epigenetic Nature of Centromeres

Understanding how CENP-A is assembled and maintained only at centromeres is central to understanding centromere and kinetochore function. A simple model for how CENP-A is targeted to the centromere is that specific centromeric DNA sequences dictate CENP-A nucleosome position. This model appears true in the “point” centromeres of some budding yeasts, including Saccharomyces cerevisiae, which has a defined 125-bp centromeric DNA sequence consisting of three centromere-DNA elements (CDEs) that are necessary and sufficient for centromere specification and kinetochore assembly (Clarke and Carbon 1980; Fitzgerald-Hayes et al. 1982; Cottarel et al. 1989). A single CENP-A^Cse4 nucleosome is positioned at CDEII in a centromeric DNA sequence-dependent manner (Spencer and Hieter 1992; Stoler et al. 1995; Meluh et al. 1998; Furuyama and Biggins 2007), although recent evidence suggests multiple accessory CENP-A^Cse4 nucleosomes, proposed to not contribute to kinetochore formation, may exist at centromeres (Coffman et al. 2011; Lawrimore et al. 2011; Haase et al. 2013). In contrast to budding yeast, fission yeast and higher eukaryotes, including humans, have “regional” centromeres that span kilo- to megabases of highly repetitive DNA. Importantly, in many eukaryotes, including humans, the underlying DNA sequence appears dispensable for centromere specification and function. Instead, CENP-A nucleosomes appear to epigenetically define the centromere.

The evidence for epigenetic specification of higher eukaryotic centromeres is compelling. Sequence-driven centromere specification would presumably impart significant selective pressure toward sequence conservation. However, higher eukaryotic centromeric DNA is the most rapidly evolving region of the genome (Csink and Henikoff 1998; Schueler et al. 2001; Padmanabhan et al. 2008), and centromeres remain the most poorly characterized loci in the human genome. The most prominent evidence for epigenetic centromere specification is the discovery of naturally occurring neocentromeres that form on a DNA locus unrelated to the endogenous chromosomal centromere. In some cases neocentromeres maintain the equal segregation of distinct chromosomal fragments where the original centromere continues to function on the original chromosome (Warburton 2004). In other cases, in which a stable neocentromere forms on a chromosome that also has its original centromere, the original centromere can be silenced and the neocentromere replaces the function of the original centromere. In either case, neocentromeres facilitate kinetochore formation, chromosome segregation, become stably inherited over multiple generations, and support development, all despite the lack of the underlying centromere-DNA sequence (Warburton et al. 1997; Saffery et al. 2000). Importantly, CENP-A is present at the site of all functional neocentromeres, but is lost from the original centromeric locus concurrent with endogenous centromere silencing (Warburton et al. 1997; Lo et al. 2001). In Schizosaccharomyces pombe, deletion of a centromere is rescued by spontaneous neocentromere formation typically close to telomeric regions, and is dependent on formation of heterochromatin (Ishii et al. 2008; Kagansky et al. 2009). Neocentromere formation has since been engineered in chicken DT40 cells, by selection of surviving clones after deletion of the endogenous Z chromosome centromere, confirming that neocentromeres maintain many features of the original centromere, without the underlying DNA (Shang et al. 2013). Taken together, these data show that the centromeres of many higher eukaryotes can be epigenetically maintained.

The Structure of CENP-A-Containing Nucleosomes

A current topic of intense study and debate is the structure and composition of CENP-A nucleosomes. Numerous CENP-A-containing nucleosome structures have been proposed, but two general models remain the most discussed (Fig. 2). One model suggests that CENP-A nucleosomes are octameric, wrap ∼120 bp of DNA, and resemble conventional H3 nucleosomes, whereas another model suggests that CENP-A wraps DNA as a hemisome, a nucleosome that contains one copy of CENP-A, H4, H2A, and H2B. Evidence supporting the presence of hemisomes comes from isolation of CENP-A^CID/H4/H2A/H2B tetramers from cross-linked chromatin immunoprecipitations (ChIPs), a smaller nucleosome height than octameric H3 nucleosomes when measured by atomic force microscopy (AFM) (Dalal et al. 2007; Dimitriadis et al. 2010; Bui et al. 2012), positive supercoiling of DNA in yeast centromere plasmids (Furuyama and Henikoff 2009), and cleavage mapping of H4:DNA interactions within budding yeast nucleosomes (Henikoff et al. 2014). The reliability of AFM measurements is questionable, as reconstituted octameric CENP-A nucleosomes can appear shorter when measured by AFM (Miell et al. 2013), a result that itself has been challenged (Codomo et al. 2014; Miell et al. 2014; Walkiewicz et al. 2014). The notion that CENP-A can function as a hemisome will not be corroborated until a hemisome is successfully isolated and characterized biochemically.

In contrast to a CENP-A hemisome, the crystal structure of an octameric CENP-A nucleosome has been resolved, revealing only subtle differences to canonical octameric H3 nucleosomes (Tachiwana et al. 2011). Octameric CENP-A nucleosomes wrap slightly less DNA than H3 nucleosomes (∼120 bp vs. 147 bp), resulting in different DNA entry/exit angles from the nucleosome, and several exposed residues exist in CENP-A’s extended loop 1 region that are important for CENP-A nucleosome stability (Tachiwana et al. 2011). In support of the octameric CENP-A nucleosome model, ChIP-Seq data revealed that CENP-A mononucleosomes wrap more DNA in cells than hemisomes could wrap (Hasson et al. 2013), immunoprecipitation of Drosophila CENP-C^CID mononucleosomes revealed the presence of CENP-A^CID dimers (Zhang et al. 2012), and CAL1-assembled CENP-A^CID nucleosomes in Drosophila are octameric (Chen et al. 2014). Finally, arrays of octameric CENP-A nucleosomes support functional centromere and kinetochore formation in vitro (Guse et al. 2011). Taken together, the most widely accepted and experimentally supported model is that CENP-A nucleosomes exist as octamers, which—other than the presence of CENP-A rather than histone H3—resemble canonical nucleosomes in their composition.

One emerging alternative model, which may reconcile the interpretation of octameric and hemisomal CENP-A nucleosomes, is that CENP-A nucleosome composition may change during progression through the cell cycle. Indeed, two recent studies in two different organisms proposed cell-cycle-dependent transitions between two CENP-A molecules per nucleosome (i.e., an octamer) to one CENP-A molecule per nucleosome (Bui et al. 2012; Shivaraju et al. 2012). These studies interpreted the changes as an octamer to hemisome transition. It is also possible a population of CENP-A nucleosomes exist as heterotypic octamers containing one molecule of CENP-A in addition to one molecule of histone H3 during specific stages of the cell cycle (Fig. 2) (Foltz et al. 2006). Overexpression of CENP-A has recently been shown to cause CENP-A/H3.3 heterotypic octameric nucleosomes assembly on chromosomal arms (Lacoste et al. 2014). However, as the predominant CENP-A nucleosome species at centromeres appears to be homotypic octamers, the functional relevance of heterotypic octamers remains unclear (Foltz et al. 2006; Kingston et al. 2011). Determining whether CENP-A nucleosomes change composition during the cell cycle and how those transitions occur is an important future goal.

DISTINGUISHING FEATURES OF CENTROMERIC CHROMATIN

How CENP-A Differs from Histone H3

The human CENP-A gene encodes a 140-residue protein (histone H3 is 136 amino acids) that is 48% identical to histone H3 overall and 62% identical across the carboxy-terminal 93 amino acids that contain the histone fold domain. Several key differences between CENP-A and H3 impart the centromere-specific functions of CENP-A. The amino-terminal 40 amino acids of CENP-A, the most divergent domain from histone H3, undergoes specific posttranslational modifications that are thought to influence CENP-A nucleosome structure and function, although their significance remains largely unclear (Zeitlin et al. 2001; Kunitoku et al. 2003; Bailey et al. 2013; Boeckmann et al. 2013). The most amino-terminal α helix in CENP-A (residues 48–56) is three residues shorter than that observed in histone H3 (residues 45–56), yet this small difference results in a loss of a DNA interaction that causes different entry/exit angles of DNA from the CENP-A nucleosome, protects less DNA in classic nucleosome micrococcal nuclease digestion assays, and most likely has a key influence of overall centromeric chromatin structure (Conde e Silva et al. 2007; Panchenko et al. 2011; Tachiwana et al. 2011). In the histone fold domain of CENP-A, a two-residue insertion (R80 and G81) in a loop domain is exposed in the CENP-A nucleosome and influences CENP-A nucleosome stability (Sekulic et al. 2010; Tachiwana et al. 2011). Importantly, a central region (residues 75–114 in humans) constitutes the CENP-A targeting domain (CATD) (Black et al. 2004) that is essential for numerous aspects of CENP-A function and structure, including binding of the CENP-A chaperone Holliday junction recognition protein (HJURP) and the centromere protein CENP-N (discussed below). Finally, CENP-A has an extended hydrophobic carboxy-terminal tail of six amino acids that binds the essential centromere protein CENP-C (Carroll et al. 2010; Guse et al. 2011; Fachinetti et al. 2013). The primary sequences of CENP-A homologs are quite divergent, suggesting the mode of CENP-A-mediated specification of centromere formation may be distinct between organisms. For example, a CENP-N homolog, which binds the CATD of CENP-A, has not been found in Drosophila or Caenorhabditis elegans.

Histone Modifications

To describe centromeric chromatin as CENP-A-containing nucleosomes is an oversimplification. Human centromeres range from 0.3 to 5 Mb in length of highly repetitive α satellite DNA. CENP-A-containing nucleosomes are thought to constitute only a subpopulation of centromeric chromatin. When centromeric chromatin is extended into long individual fibers CENP-A nucleosomes are found interspersed with H3 nucleosomes at centromeres (Zinkowski et al. 1991; Blower et al. 2002). These H3 nucleosomes are enriched in specific posttranslational modifications including histone H3 lysine 4, lysine 9, and lysine 36 dimethylation (Sullivan and Karpen 2004; Bergmann et al. 2011). In particular, H3K4me2 appears to be a centromere-specific modification that has been shown to promote CENP-A assembly into chromatin (Bergmann et al. 2011). In S. pombe, artificial tethering of the histone methyltransferase Clr4 to euchromatin promotes neocentromere formation (Ishii et al. 2008). Targeting of transcriptional activators and repressors to human artificial chromosome (HAC) centromeres influences kinetochore formation and CENP-A assembly, although the precise mechanisms remain largely unclear (Nakano et al. 2008; Cardinale et al. 2009; Ohzeki et al. 2012). In addition, tethering either the histone acetyltransferase domains of p300 and PCAF or the histone methylase Suv39H1 to HACs upsets the balance between histone H3K9 methylation and acetylation, promoting centromere formation when acetylated or acting as a barrier to CENP-A nucleosome proliferation when trimethylated (Ohzeki et al. 2012). Histone modification of non-CENP-A nucleosomes is therefore likely to have a role in centromere maintenance and function. In addition, CENP-A nucleosomes themselves are likely to be subject to modification; CENP-A ubiquitination by CUL3/RDX ubiquitin ligase has been shown to stabilize soluble CENP-A in Drosophila (Bade et al. 2014).

Other Centromere-Specific DNA-Binding Proteins

Several DNA-binding proteins, in addition to CENP-A nucleosomes, associate with centromeres and influence centromere function. The histone-fold-containing proteins CENP-T, -W, -S, and -X assemble into dimers of CENP-T/W, tetramers of CENP-S/X, and potentially a heterotetramer of CENP-T/W/S and X (Fig. 3). In vitro, a heterotetramer of CENP-T, -W, -S, and -X was proposed to protect ∼100 bp of DNA from nuclease digestion, in a similar manner to H3 and CENP-A nucleosomes (Nishino et al. 2012). More recently, a CENP-T/W/S/X octamer has been shown to bind ∼100 bp of DNA and potentially induce positive supercoils (as opposed to negative supercoils induced by nucleosomes) (Takeuchi et al. 2014). However, whether a CENP-T/W/S/X nucleosome-like species exists at centromeres remains unclear, as the path of DNA wrapping around these proteins is unknown and the nuclease protection pattern is distinct from that of a nucleosome. Histone fold proteins are not exclusively nucleosomal proteins that wrap DNA, but also include several different classes of DNA-binding factors. CENP-T/W and CENP-S/X may instead function as two independent complexes. CENP-S/X plays noncentromeric roles in DNA repair independently of CENP-T/W (Singh et al. 2010; Tao et al. 2012; Zhao et al. 2014), CENP-S/X depletion does not affect CENP-T centromere recruitment (Amano et al. 2009), and CENP-S is dispensable in engineered neocentromeres that are positive for CENP-T in chicken DT-40 cells (Shang et al. 2013). One possibility is that binding of CENP-S/X to CENP-T/W confers some centromere-specific function, but the ability of CENP-T to function independently from CENP-S suggests CENP-T/W/S/X does not form an obligate nucleosome-like particle (Hori et al. 2008a; Amano et al. 2009). Indeed, before identification of the CENP-T/W/S/X tetramer, CENP-T/W was suggested to associate primarily with histone H3 rather than CENP-A (Hori et al. 2008a; Ribeiro et al. 2010). Consistent with this, although CENP-T is lost from centromeres completely lacking CENP-A, its centromere levels are largely unaffected by a 90% CENP-A reduction (Fachinetti et al. 2013). Therefore, CENP-T/W/S/X, either as a complex or separately, may interact specifically with non-CENP-A centromeric DNA in a non-nucleosome-like manner. Recent data suggests that CENP-T and CENP-W are targeted for degradation unless they are in a heterodimeric complex with each other, suggesting that either a CENP-T/W/S/X nucleosome-like complex or separate complexes of CENP-T/W and CENP-S/X will be the functionally relevant species at centromeres (Chun et al. 2013). Ascertaining whether a CENP-T/W/S/X nucleosome exists within centromeric chromatin, and assessing its relevance to centromere function, are immediate goals for the field.

Possible Roles for Centromeric DNA Sequence?

What is the role, if any, of centromeric DNA sequence? The regional centromeres of many eukaryotes are characterized by repetitive DNA sequences. Human centromeres consist of A/T-rich 171-bp α-satellite repeats. Human centromeric DNA contains the 17-bp “CENP-B box,” which directly recruits the centromere protein CENP-B (Masumoto et al. 1989). CENP-B knockout mice are for the most part phenotypically normal (Hudson et al. 1998; Kapoor et al. 1998; Perez-Castro et al. 1998) and functional neocentromeres lack CENP-B (Saffery et al. 2000). Moreover, CENP-B is absent from the human Y chromosome (Earnshaw et al. 1987) and no CENP-B functional homologs have been identified in Xenopus, zebrafish, C. elegans, or Drosophila to date. Thus, once a centromere is formed, CENP-B appears to be dispensable for centromere function.

On the other hand, CENP-B has been suggested to promote the formation of new centromeres. Studies using HACs as a substrate for new centromere assembly showed that satellite repeats containing CENP-B box mutations were much less efficient at forming a functional centromere (Ohzeki et al. 2002). Consistent with this, fibroblasts from CENP-B knockout mice fail to form new centromeres on artificial chromosomes (Okada et al. 2007). CENP-B may also play a second role in initiating heterochromatin formation, as CENP-B box containing DNA, integrated into a chromosome arm, is specifically trimethylated on H3K9 in CENP-B containing mouse embryo fibroblasts, but not in fibroblasts from a CENP-B knockout mouse (Okada et al. 2007). The CENP-B DNA-binding domain can position CENP-A nucleosomes on DNA in vitro (Tanaka et al. 2005), and CENP-B boxes appear to phase CENP-A nucleosome position in vivo (Hasson et al. 2013).

Partial reduction in CENP-A levels has no effect on CENP-B (Fachinetti et al. 2013). However, when CENP-A is completely abolished, CENP-B centromere levels drop by around 50%, revealing a previously uncharacterized dependence of CENP-B on CENP-A nucleosomes (Fachinetti et al. 2013). Mutation of centromeric CENP-A revealed CENP-B centromere recruitment requires the CENP-A amino-terminal tail through an unknown mechanism, and CENP-B:CENP-A amino-terminal cooperation may contribute to sustaining localization of centromere proteins (Fachinetti et al. 2013). Thus, CENP-B may represent a redundant pathway for CENP-A nucleosome stabilization, positioning, and/or recruitment of other centromere-specific components. It is interesting to place this in the context of studies that find that human centromeres can remain functional even after extensive, but not complete, depletion of CENP-A (Liu et al. 2006; Fachinetti et al. 2013). CENP-B may specify the minimal sites of CENP-A nucleosomes sufficient for functional centromere formation, in this way acting as a backup to guard against CENP-A loss. Thus, in the case of severely limited CENP-A, centromere specification may no longer be epigenetically maintained, as the CENP-B box may have a role. In the same manner, an intriguing possibility is that CENP-B has a partially redundant role in maintaining CENP-A nucleosome position through DNA replication (see below). Depletion of CENP-A to a level that still maintains centromere formation, coupled with CENP-B depletion, may reveal some interesting centromere phenomena.

CENTROMERE PROTEIN RECOGNITION OF CENP-A CHROMATIN

CENP-A nucleosomes direct the assembly of functional centromeres by recruiting centromere proteins, both directly through binding and indirectly through the assembly of higher-order protein complexes. Two proteins, CENP-N and CENP-C, interact directly with CENP-A nucleosomes to promote centromere assembly (Carroll et al. 2009, 2010). Both CENP-N and CENP-C only bind nucleosomal CENP-A, and do so through recognition of different elements of the nucleosome. CENP-N binds CENP-A nucleosomes by recognizing CENP-A’s CATD, whereas CENP-C binds CENP-A’s divergent carboxy-terminal tail and the acidic patch on histones H2A and H2B (Carroll et al. 2009, 2010; Kato et al. 2013).

The functions of CENP-N are not well established. CENP-N depletion causes a loss of multiple other centromere components (Foltz et al. 2006; McClelland et al. 2007; Carroll et al. 2009), so CENP-N appears to be a key building block of the centromere. Indeed, S. cerevisiae CENP-N^Chl4 is required for de novo centromere formation (Mythreye and Bloom 2003). However, CENP-N^Chl4 is not required to sustain previously established centromeres, suggesting it may function redundantly at existing centromeres. In addition, centromeric CENP-N protein levels decrease before mitosis, suggesting that CENP-N may not be required to recruit kinetochore proteins (McClelland et al. 2007; Hellwig et al. 2011). Although CENP-N binds the CATD, this may not be necessary for CENP-N to localize to centromeres; centromeres generated by LacI-CENP-C and LacI-CENP-T fusions, which are proposed to be negative for CENP-A, are still positive for CENP-N (Gascoigne et al. 2011), and CENP-N can be recruited to H3/CENP-A chimeras that lack the CATD in vitro (Guse et al. 2011). In S. cerevisiae, CENP-N^Chl4 forms a heterodimer with CENP-L^Iml3 that interacts with CENP-C^Mif2, suggesting that CENP-C could mediate recruitment of CENP-N/L (Guo et al. 2013; Hinshaw and Harrison 2013).

In contrast to CENP-N, CENP-C has a clear role in both kinetochore formation and centromeric chromatin maintenance (discussed below). CENP-C is specifically recruited to centromeres through hydrophobic interactions between the CENP-A carboxyl terminus, and conserved domains in the CENP-C protein provide specificity for CENP-C recruitment to centromeres (Carroll et al. 2010; Kato et al. 2013). The nucleosome-binding domain of CENP-C contacts both the CENP-A carboxyl terminus and the acidic patch of H2A/H2B and with DNA (Kato et al. 2013). Histone H3 and CENP-A chimeras containing only the carboxy-terminal six amino acids of CENP-A are sufficient for CENP-C recruitment in Xenopus Guse et al. 2011). In human cells CENP-C centromere recruitment is reduced after replacement of CENP-A with H3/CENP-A chimeras lacking the carboxy-terminal CENP-A region (Fachinetti et al. 2013). S. cerevisiae CENP-C^Mif2 dimerizes through a region near its carboxyl terminus (Cohen et al. 2008), presenting the possibility of a CENP-C dimer interacting with a CENP-A octameric nucleosome, or higher-order structures involving two CENP-A nucleosomes, consistent with suggestions that CENP-C is important for higher-order centromere structure (Ribeiro et al. 2010).

Regulation of CENP-C centromere localization at endogenous centromeres is more complicated than direct recruitment by CENP-A. In cells in which CENP-A is progressively removed from centromeres, a significant population of CENP-C maintains its localization despite the absence of CENP-A (Fachinetti et al. 2013). This suggests that CENP-C may be recruited to centromeres through multiple redundant mechanisms, and also raises the possibility that CENP-C may contribute to its own stability at centromeres, which has implications for our understanding of how centromeres could be maintained through CENP-A assembly and DNA replication. As CENP-C is key for establishing the kinetochore, its resistance to decreasing CENP-A nucleosome numbers provides a potential explanation for why all but the complete depletion of CENP-A fails to result in a strong centromere defect.

KINETOCHORE AND CENTROMERE COMPOSITION

Broadly speaking, centromeric proteins have two temporally distinct roles. First, before and during mitosis, centromere proteins build the kinetochore. Second, during mitotic exit and in interphase (in humans), centromere proteins recruit the factors that replenish CENP-A nucleosomes and maintain centromeric chromatin. In this section, we briefly focus on how centromeres promote mitotic kinetochore formation, and then discuss the role of the centromere in CENP-A reassembly. For an in-depth overview of kinetochore function in mitosis, see an accompanying review of the kinetochore (Cheeseman 2014).

The hub of the mitotic kinetochore, and the machinery responsible for robust microtubule attachment, is the KMN protein network: KNL1, the Mis12 complex (Mis12, Nsl1, Nnf1, and Dsn1) and the Ndc80 complex (Ndc80/Hec1, Nuf2, Spc24, and Spc25). In all organisms, an essential function of the constitutive centromere proteins is to recruit the KMN network before and during mitosis. The amino terminus of CENP-C interacts with Nnf1 and is required for the G₂ centromere targeting of the Mis12 complex in both Drosophila and human cells (Milks et al. 2009; Przewloka et al. 2011; Screpanti et al. 2011). A second essential step in KMN protein assembly at centromeres is binding of the Spc24/25 components of the Ndc80 complex by the phosphorylated amino-terminal tail of CENP-T. This conserved interaction forms stable, load-bearing attachments to microtubules via the Ndc80 complex (Bock et al. 2012; Schleiffer et al. 2012; Malvezzi et al. 2013; Nishino et al. 2013). A nonphosphorylatable CENP-T amino-terminal tail disrupts kinetochore function as it cannot fulfill these functions (Gascoigne et al. 2011). Interestingly, the Mis12 complex also binds Spc24/25 in the same manner as CENP-T, and CENP-T and Mis12 compete for Ndc80 binding, which may influence the stability of kinetochore microtubule attachments (Bock et al. 2012; Schleiffer et al. 2012). Furthermore, phosphoregulation of Mis12 and/or CENP-T could control switching between different microtubule-binding modes within kinetochores. These findings may have important implications for how kinetochores may positively select for correct microtubule attachments and destabilize erroneous attachments.

To date, 17 constitutively associated CENP proteins have been identified (Fig. 3). Broadly speaking, during mitosis the CENP proteins appear to regulate the CENP-C/T:KMN network and the KMN network:microtubule interactions. As a rule, depletion of any CENP protein impairs centromere function, causing kinetochore:microtubule attachment and chromosome congression defects (Fukagawa et al. 2001; Foltz et al. 2006; McAinsh et al. 2006; Okada et al. 2006; McClelland et al. 2007; Hori et al. 2008b; Carroll et al. 2009; Toso et al. 2009; Amaro et al. 2010). Although the function of CENP-C and CENP-T has been described in the most detail, biochemical analyses of centromere subcomplexes are also contributing to our understanding of centromere function. In addition to CENP-T/W/S/X mentioned above, CENP-L^Iml3 forms a heterodimer with CENP-N^Chl4 in S. cerevisiae, and a homologous interaction between CENP-L and the CENP-N amino terminus has been identified in human cells (Carroll et al. 2009; Hinshaw and Harrison 2013). A centromere subcomplex of CENP-H/I/K/M has also recently been reconstituted biochemically, and cell-based assays have revealed that CENP-H/I/K/M is required for CENP-C-mediated CENP-T centromere localization (A Musacchio, pers. comm.). CENP-H, CENP-I, and CENP-K were previously shown to bind KNL-1 directly (Cheeseman et al. 2008), suggesting CENP-H/I/K/M may also contribute to kinetochore formation.

The function of the remaining centromere proteins CENP-O, -P, -Q, -R, and -U remains largely unclear. In vitro, CENP-U and a CENP-Q octamer bind microtubules (Amaro et al. 2010), and CENP-U is also proposed to bind Hec1 (Hua et al. 2011). Aurora B–mediated phosphorylation of CENP-U within a CENP-O/P/Q/U complex is required for efficient recovery from spindle-damaging agents in cells (Hori et al. 2008b). Recently, fission yeast CENP-Q and CENP-O have been shown to associate with members of the Mis18 complex that regulates CENP-A assembly (see below) (Subramanian et al. 2014).

In conclusion, although centromeres are perhaps best considered as one large protein complex, the recent careful biochemical analyses of centromere subcomplexes described above have greatly aided our understanding of centromere function. Before these achievements, cell-based work had created a confusing landscape; knowing whether an observed effect is direct or indirect or affected by other consequences of the manipulation has been extremely challenging. We anticipate that further reconstitution of the centromere will continue to provide a key framework for our understanding of centromere function in cells.

CENTROMERE MAINTENANCE AND CENP-A ASSEMBLY

The amount of DNA within a cell doubles every cell cycle as a result of DNA replication during S phase. Conventional histones are assembled concurrently with DNA replication, with a network of histone chaperones and modifiers in place to regenerate the prereplication chromatin environment after passage of the replication fork (reviewed in Probst et al. 2009). In contrast to canonical chromatin, CENP-A nucleosome replenishment in higher eukaryotes occurs independently of DNA replication. In humans, CENP-A assembly starts after anaphase, and finishes before the end of G₁ (Fig. 4) (Jansen et al. 2007). DNA replication-independent assembly is specific to CENP-A nucleosomes, and is conferred by the CATD, as pulse-labeled H3^CATD chimeric proteins mimic the assembly characteristics of CENP-A, at least at endogenous centromeres that also contain CENP-A (Black et al. 2007; Bodor et al. 2013). The timing of CENP-A assembly is, in part, governed by cell-cycle kinases, including Cdk1:cyclin B. When Cdk1 is inhibited, CENP-A assembly occurs prematurely (Silva et al. 2012), and when cyclinA:Cdk1 levels or APC activity are perturbed in Drosophila, CENP-A^CID assembly fails (Erhardt et al. 2008). Although CENP-A assembly remains a poorly understood process, we discuss here some of the relatively well-characterized proteins involved in CENP-A assembly in turn, before summarizing other factors that are less well understood.

Figure 4.

Functions of centromeric chromatin through the human cell cycle. Centromeric chromatin, defined by CENP-A-containing nucleosomes, is specifically assembled during mitotic exit and G₁. Assembly causes the CENP-A copy number at centromeres to double (represented by transition from pink to red chromatin). During DNA replication/S phase, replication of centromere-DNA results in the distribution of CENP-A onto the two nascent DNA strands. This causes a twofold reduction in CENP-A at each centromere-DNA sequence (red to pink chromatin) (see Fig. 5 for more details). Note that it is this “diluted” centromeric chromatin (pink) that is responsible for building functional kinetochores before and during mitosis and, thus, for segregating chromosomes. Broadly speaking, centromere proteins retain localization at CENP-A chromatin through all stages of the cell cycle. How centromere proteins respond to changing CENP-A protein numbers within chromatin is not clear.

HJURP: The CENP-A Chaperone

Canonical histones require chaperones for DNA-replication-dependent assembly (Ransom et al. 2010). CENP-A, as a specialized histone, has a specialized chaperone known as HJURP in humans or Scm3 in yeast. S. cerevisae Scm3 was identified in immunopurifications of CENP-A^Cse4 and is required to target CENP-A^Cse4 to centromeres (Camahort et al. 2007; Mizuguchi et al. 2007; Stoler et al. 2007). Scm3 can assemble CENP-A^Cse4 into nucleosomes in vitro (Shivaraju et al. 2011). HJURP (Holliday junction recognition protein) was originally named because it was suggested to bind Holliday junctions (Kato et al. 2007) and was subsequently identified in purifications of soluble CENP-A complexes that represent a prechromatin assembly intermediate (Dunleavy et al. 2009; Foltz et al. 2009). HJURP and Scm3 bind directly to a dimer of CENP-A and histone H4 through recognition of the CATD domain of CENP-A by HJURP/Scm3’s conserved amino terminus (Foltz et al. 2009; Shuaib et al. 2010; Cho and Harrison 2011; Hu et al. 2011; Zhou et al. 2011). In addition, HJURP forms contacts with CENP-A outside of the CATD that enhance the stability of the HJURP:CENP-A:H4 complex (Bassett et al. 2012). Recognition of the CATD provides at least a partial explanation for why the CATD of CENP-A is essential for CENP-A maintenance and forms the minimal domain in CENP-A required for CENP-A assembly. A histone H3^CATD chimera is recognized by HJURP and stably incorporated into centromeric chromatin (Black et al. 2004; Bodor et al. 2013). The importance of HJURP in centromere maintenance was confirmed by tethering LacI-HJURP fusion protein to Lac operators at ectopic chromosomal loci, which consequently assemble CENP-A and form largely functional kinetochores (Barnhart et al. 2011; Hori et al. 2013). In Drosophila, which lack an HJURP homolog, CAL1 assembles CENP-A^CID nucleosomes in an extremely similar manner to HJURP in vertebrates (Chen et al. 2014).

The levels of soluble CENP-A:HJURP complex increase as cells enter mitosis, but this complex does not associate with centromeres until cells proceed through anaphase, concurrent with the timing of CENP-A assembly (Dunleavy et al. 2009; Foltz et al. 2009). Structural studies indicate that the HJURP:CENP-A:H4 and Scm3:Cse4:H4 trimers preclude CENP-A-H4 tetramer formation (Hu et al. 2011; Zhou et al. 2011). Thus, an extremely interesting problem is how the HJURP chaperone complex releases CENP-A to facilitate transfer of CENP-A/H4 into chromatin. Because a dimer of CENP-A:H4 is carried by each HJURP molecule, either two rounds of dimer delivery must occur to assemble an octameric nucleosome or some other intermediate containing only a CENP-A/H4 tetramer must exist. Of note, the carboxy-terminal region of HJURP contains an HJURP dimerization domain, suggesting that (HJURP:CENP-A:H4)₂ exists as a soluble complex away from chromatin, and that this complex may deliver two CENP-A:H4 dimers simultaneously to sites of new CENP-A assembly (Fig. 5) (Zasadzinska et al. 2013). Similar observations have been made for Scm3 (Dechassa et al. 2014). Interestingly, HJURP recruitment to centromeres can occur without CENP-A binding or HJURP dimerization, but HJURP dimerization is required for CENP-A assembly, confirmed by elegant experiments that rescued deletion of the HJURP dimerization motif with insertion of an unrelated dimerization domain (Zasadzinska et al. 2013). Yeast Scm3 localizes to centromeres independently of CENP-A^Cse4/Cnp1, and also self-associates (Mizuguchi et al. 2007; Pidoux et al. 2009; Williams et al. 2009), suggesting that the mechanism of HJURP/Scm3-mediated CENP-A assembly is conserved. The major limiting factor remaining in our understanding of HJURP function is identifying how HJURP is recruited to centromeres; links between HJURP and other components of the CENP-A assembly machinery remain unclear, although Mis18β (see below) has recently been suggested to interact with HJURP (Wang et al. 2014).

Figure 5.

CENP-A assembly in humans. (1) “Parent” CENP-A nucleosomes (from the previous cell cycle), directly or indirectly specify the sites of new CENP-A assembly. (2) The centromere proteins shown in the figure; CENP-C, -N (which directly bind CENP-A nucleosomes), -H, -I, -K, and -M have been experimentally implicated in CENP-A assembly. (3) In addition to centromere proteins, the Mis18 complex, consisting of Mis18α, Mis18β, and M18BP1, is also required for CENP-A assembly, and most likely modifies (M) the chromatin and/or the recruitment of specialized loading factors. (4) Chromatin modifiers and chaperones, such as RbAp46 and 48, also have a role in modifying chromatin during CENP-A assembly. (5) Centromere proteins and the Mis18 complex somehow recruit HJURP, the CENP-A-specific chaperone. HJURP binding to CENP-A precludes CENP-A:H4 tetramer formation, and HJURP dimerization is required for CENP-A assembly, so one possible model is that nascent CENP-A is delivered as two dimers in a (HJURP:CENP-A:H4)₂ complex, as shown. (6) The proteins or chromatin features that mark the site of new CENP-A assembly, the “placeholders,” remain unclear. (7) Finally, once CENP-A is assembled into chromatin, factors such as MgcRacGAP, which interacts with the Mis18 complex, stabilize incorporated CENP-A.

The Mis18 Complex

Mis18 was discovered in S. pombe as a protein that, when mutated, causes errors in chromosome segregation (Hayashi et al. 2004). Humans possess two homologs of Mis18, Mis18α and Mis18β, and another associating protein called Mis18-binding protein 1 (M18BP1) (Fig. 5). All three proteins interact but it is not known whether Mis18α, Mis18β, and M18BP1 function in CENP-A assembly exclusively as a complex. A homologous M18BP1 protein, Kinetochore Null 2 (KNL2), was discovered in a screen for kinetochore defects in C. elegans and M18BP1 has been identified in humans, Xenopus, and Arabidopsis (Fujita et al. 2007; Maddox et al. 2007; Moree et al. 2011; Lermontova et al. 2013). The function of the Mis18 complex remains unclear, and more proteins are likely to be involved. Indeed, two recent studies in fission yeast have independently identified Eic1/Mis19 and Eic2/Mis20 as novel Mis18 complex components (Hayashi et al. 2004; Subramanian et al. 2014). Eic1 may be the S. pombe homolog of Mis18BP1. Depletion of any one Mis18 complex component (apart from Eic2/Mis20) prevents new CENP-A assembly (Fujita et al. 2007; Maddox et al. 2007; Hayashi et al. 2014; Subramanian et al. 2014). However, in contrast to HJURP, overall cellular CENP-A levels are not diminished after Mis18 depletion, suggesting that the soluble HJURP:CENP-A:H4 remains stable without Mis18 activity and the Mis18 complex functions at chromatin rather than on soluble CENP-A.

Mis18α, Mis18β, and M18BP1 are recruited to centromeres at an early stage in CENP-A assembly, during anaphase/telophase of mitosis and in early G₁ (Fujita et al. 2007; Maddox et al. 2007; Dambacher et al. 2012). The Mis18 complex is required for the recruitment of HJURP to centromeres in a number of model organisms (Pidoux et al. 2009; Williams et al. 2009; Barnhart et al. 2011; Moree et al. 2011), and in human cells, depletion of HJURP has no effect on Mis18α. Together with the early localization of the Mis18 complex to centromeres, these data have led to suggestions that the Mis18 complex is a key player in CENP-A assembly process. Artificially tethering HJURP to chromatin can bypass the requirement for M18BP1; thus, a simple model is that M18BP1 association precedes HJURP binding and is required for proper targeting of the HJURP chaperone complex to centromeres. Whether the Mis18 complex also plays an active role in the transfer of CENP-A into chromatin is unknown.

The specific activities of the Mis18 complex may involve histone and/or DNA modification. Mis18α is required for the recruitment of DNA methyltransferases to centromeres, which are in turn required for normal CENP-A centromere levels (Kim et al. 2012). Mis18 mutation in fission yeast causes an increase in centromeric H3 and H4 acetylation (Hayashi et al. 2004), and treatment with histone deacetylation inhibitors can rescue Mis18α depletion (Fujita et al. 2007). Human M18BP1 and C. elegans KNL2 have SANT/Myb DNA-binding domains, placing M18BP1/KNL2 at the necessary location to modify histones or DNA, either directly or indirectly. SANT/Myb domains are also found in histone modifiers such as Ada2 histone acetyl transferase (HAT) (Boyer et al. 2002), and the SANT/Myb domain of the telomere-binding protein TRF2 alters chromatin structure in vitro (Baker et al. 2009). Of note, the S. pombe transcriptional regulator Teb1, which also contains SANT/Myb domains, has also been shown to be required for Cnp1^CENP-A assembly (Valente et al. 2013).

Constitutive Centromere Proteins and CENP-A Assembly

The fact that constitutive centromere proteins are present at centromeres during interphase (as opposed to kinetochore proteins) perhaps reflects their functions in maintaining centromeres through the cell cycle (Fig. 5). CENP-C depletion prevents CENP-A assembly onto sperm chromatin in Xenopus egg extracts, and CENP-C is required to target Xenopus M18BP1 isoform 1 to metaphase centromeres (Moree et al. 2011). An analogous system may function in Drosophila in which the CENP-A assembly factor Cal1 requires CENP-C to be recruited to centromeres in metaphase (Erhardt et al. 2008; Mellone et al. 2011). Like Mis18α, CENP-C interacts with DNA methyltransferases that may facilitate chromatin remodeling to promote CENP-A assembly (Gopalakrishnan et al. 2009). Fusing CENP-C and CENP-I proteins with LacI, when targeted to lacO sites integrated into the chromosome, support formation of centromeres that are resistant to subsequent IPTG treatment (which removes the fusion protein) (Hori et al. 2013). This indicates that functional CENP-A nucleosomes are assembled into chromatin as a consequence of CENP-C or CENP-I ectopic localization (Hori et al. 2013). Indeed, the individual CENPs that comprise the recently identified CENP-H/I/K/M complex have been implicated in CENP-A assembly in chicken and human cells (Takahashi et al. 2000; Okada et al. 2006).

Additional Factors Involved in CENP-A Assembly

Several other factors are known to function in CENP-A assembly but their activities are not exclusively devoted to centromeres. Depletion of the RbAp46/48 class of histone chaperones, Mis16 in S. pombe or RbAp46 and 48 in humans (by temperature-sensitive mutation or RNAi) causes loss of CENP-A from centromeres (Hayashi et al. 2004). In human cells, RbAp46/48 purify with the HJURP chaperone complex (Dunleavy et al. 2009) and depletion of RbAp46/48 causes a reduction in CENP-A protein levels, indicating that they play a chaperone role for CENP-A (Hayashi et al. 2004). Unlike HJURP, RbAp46/48 have important roles in regulation of conventional chromatin. RbAp46 and 48 are both subunits of the CAF-1 (chromatin assembly factor-1) complex that is required for the replication-coupled assembly of H3:H4 (Verreault et al. 1996; Tagami et al. 2004). Thus, it appears likely that CENP-A assembly shares some chromatin modifiers with conventional histone H3 assembly.

In S. cerevisiae, the E3 ubiquitin ligase Psh1 has been shown to target CENP-A^Cse4 for degradation (Hewawasam et al. 2010; Ranjitkar et al. 2010). Psh1 localizes to centromeres, but is thought to act on noncentromeric CENP-A^Cse4. Psh1 specifically acts on CENP-A^Cse4 as it recognizes the CATD. Because of this, Scm3 binding to the CATD protects CENP-A^Cse4 from Psh1-mediated degradation. Deletion of Psh1 causes assembly of CENP-A^Cse4 into noncentromeric chromosomal loci. These data suggest that mechanisms exist to detect and degrade Scm3-free CENP-A^Cse4 to prevent incorrect assembly. The role of CENP-A ubiquitination in regulating CENP-A assembly may be conserved, as ubiquitination of Drosophila CENP-A^CID by the E3 ubiquitin ligases Ppa and CUL3 can destabilize or stabilize CENP-A^CID, respectively (Moreno-Moreno et al. 2011; Bade et al. 2014).

At this stage, it is unclear if Psh1, or a related pathway, can act on CENP-A^Cse4 already within chromatin. Of note, the recent characterization of the GTP exchange factor MgcRacGAP suggests that, in addition to removal of free CENP-A, correctly assembled CENP-A nucleosomes are stabilized. Pulse labeling of SNAP-CENP-A, to selectively label “old” CENP-A nucleosomes carried over from the previous cell cycle, suggests that MgcRacGAP depletion specifically destabilizes nascent CENP-A nucleosomes (Lagana et al. 2010). However, other than the discovery that MgcRacGAP is degraded during mitotic exit (perhaps explaining its delayed centromere localization) (Nishimura et al. 2013), little is known about the role of GTP hydrolysis and MgcRacGAP in CENP-A assembly or stabilization. MgcRacGAP is also known to have important roles in cell proliferation through regulation of cytokinesis, as a member of the centralspindlin complex (White and Glotzer 2012). In addition to MgcRacGAP, subunits of the RSF complex (remodeling and spacing factor), which regulates canonical histone modification, have also been implicated in CENP-A chromatin stabilization (Obuse et al. 2004; Izuta et al. 2006; Perpelescu et al. 2009).

In summary, great progress has been made in the understanding of CENP-A assembly in the last few years, most noticeably in the identification and characterization of novel CENP-A assembly factors such as HJURP and M18BP1. We anticipate that the next few years will include consolidation of what remains a somewhat fragmented picture. As a final consideration, a comprehensive understanding of CENP-A assembly requires identification of the mechanism for targeting new CENP-A to the correct positions (Fig. 5). Several models have been proposed for a “placeholder,” a marking mechanism that identifies where to assemble new CENP-A nucleosomes, including naked DNA, histone H3.3 nucleosomes (Dunleavy et al. 2011), hybrid CENP-A/H3.3 nucleosomes, a CENP-T/W/S/X complex (Nishino et al. 2012), hemisomes (Bui et al. 2012), or an unidentified component. The chromatin-associated factors that dictate the positions of new CENP-A assembly will most likely be targets for some of the assembly factors discussed above.

CENTROMERE LONGEVITY AND CENP-A MAINTENANCE DURING DNA REPLICATION

Pulse-chase labeling of CENP-A, to specifically track incorporated CENP-A across multiple cell divisions, shows a 50% reduction in labeled CENP-A per chromosome with each replication cycle (Dunleavy et al. 2011; Bodor et al. 2013). In addition, gene deletion of CENP-A causes CENP-A centromere levels to reduce by 50% each cell cycle (Fachinetti et al. 2013). Thus, in the absence of CENP-A assembly, CENP-A is progressively depleted from centromeres after multiple cell generations (as a result of DNA replication; Fig. 6). These data highlight the remarkable persistence of existing CENP-A nucleosomes. To date, a mechanism that recycles correctly positioned, chromatinized CENP-A has not been identified. Thus, CENP-A protein deposited during early development provides the epigenetic mark of centromeres indefinitely. Moreover, histone H4 shows the same stable characteristics as CENP-A specifically at centromeres, suggesting that H4 is also stabilized as part of the CENP-A nucleosome (Bodor et al. 2013). This is not an intrinsic property of centromeric chromatin, as other nucleosomal species show no distinct centromere-specific stabilization (Bodor et al. 2013). What provides CENP-A nucleosomes with this longevity remains unclear, but histone H3^CATD chimeric nucleosomes possess many of the same characteristics, suggesting that the specific nucleosome structure conferred by the CATD may have a role (Bodor et al. 2013).

Figure 6.

Models of CENP-A nucleosome distribution during DNA replication. (A) CENP-A (red) nucleosomes remain as homotypic octamers after passage of the replication fork. In the model shown, octameric histone H3.1 (cyan) nucleosomes occupy the gaps left by the twofold reduction in CENP-A nucleosomes on each DNA strand. Other possibilities include histone H3.3, naked DNA, or other species of specialized centromeric chromatin (such as CENP-T/W/S/X). (B) Each CENP-A nucleosome is split and segregated to both DNA strands. CENP-A:H4 dimers form heterotetramers with H3/H4 dimers, resulting in heterooctamer formation. (C) Octameric CENP-A nucleosomes are split and not replenished, resulting in the formation of tetrameric hemisomes. As hemisomes are incapable of wrapping as much DNA as octameric nucleosomes, this model results in an excess of free DNA, the fate of which is unclear (?). All these models assume octameric CENP-A nucleosomes are the prereplication conformation. Note that the composition of centromeric chromatin after DNA replication becomes the template for CENP-A assembly in the next cell cycle (see Figs. 3 and 4).

Although CENP-A nucleosomes display remarkable longevity, CENP-A nucleosome disruption and reassembly on daughter DNA strands is a necessary part of centromere maintenance during DNA replication. As DNA replication proceeds, CENP-A histones are distributed equally to each new DNA strand (Figs. 4 and 6). Establishing how CENP-A nucleosomes are segregated, whether the composition of CENP-A nucleosomes changes as a result of redistribution, and how the network of centromere proteins differs between G₂ and G₁, are all major unanswered questions that have profound implications for understanding centromere maintenance (Fig. 6). Kinetochores are assembled on centromeres that have been reorganized as a result of DNA replication, and CENP-A assembly during mitotic exit and G₁ uses a replicated centromere as a template.

The challenge of maintaining epigenetic marks on DNA as the DNA is replicated is applicable to the entire genome. Presumably, given the specialized nature of centromeric chromatin, specialized systems to maintain CENP-A likely exist. However, other than the observation that CENP-A histones are equally distributed between replicated DNA strands (Dunleavy et al. 2009), very little is known about potential changes in CENP-A nucleosome structure and/or composition. Important insights will come with a more complete understanding of the role of the centromere proteins in maintaining CENP-A nucleosomes across the DNA replication fork, as recent observations show that many centromeric proteins show distinct stability during S phase. CENP-T and CENP-W completely turn over during S phase, but increase in abundance in S/G₂ relative to G₁ (Prendergast et al. 2011). CENP-S and CENP-X also assemble at centromeres during S/G₂ (Dornblut et al. 2014). CENP-N is stabilized specifically at the end of S phase through an unknown mechanism (McClelland et al. 2007; Hellwig et al. 2011). CENP-C and CENP-H are stabilized at centromeres during DNA replication (Hemmerich et al. 2008). As CENP-C and CENP-N, the two centromere components that directly bind CENP-A, stabilize during S phase, and CENP-C can still localize to centromeres that have completely lost CENP-A (Fachinetti et al. 2013), a speculative possibility is that the network of centromere proteins maintain centromere position during underlying passage of the replication fork and disruption of CENP-A nucleosomes. In this manner, centromere proteins may play a key role in promoting CENP-A nucleosome reformation on replicated centromeric DNA. Finally, CENP-B binding to the CENP-B box may play a role in repositioning of CENP-A nucleosomes during replication (Tanaka et al. 2005).

CONCLUDING REMARKS

Remarkable progress has been made in the understanding of centromeres since the identification of the first centromere proteins in the mid-1980s (Earnshaw and Rothfield 1985; Palmer et al. 1987). Given that the rate at which new centromere proteins are being identified appears to be slowing, we may be nearing a complete parts list of centromere proteins. How these components form a functional centromere/kinetochore is also becoming clearer. The identification of biochemically discrete complexes of centromere and kinetochore proteins, that share phenotypes when perturbed, suggests modularity in centromere assembly and functions. For example, the role of a CENP-T-containing complex in recruiting the Ndc80 complex and the function of CENP-C in recruiting of the Mis12 complex, suggests different functions of the kinetochore are brought to centromeres through different activities of core centromere proteins. However, despite this modularity, there is significant cross talk between centromere components to stably form a functional centromere and kinetochore; Mis12 and CENP-T compete for Ndc80 binding (Bock et al. 2012; Schleiffer et al. 2012; Nishino et al. 2013) and a CENP-N/L heterodimer binds CENP-C (Carroll et al. 2009; Hinshaw and Harrison 2013), presumably as part of a CENP-A nucleosome-containing complex. Pinpointing how and when different centromere modules interact remains an exciting challenge.

Significant advances in our understanding of how centromeric chromatin is specified and assembled have also been made in the past decade. Important cell and biochemical studies have led to the identification of Scm3/HJURP and the Mis18 complex as central regulators of new CENP-A assembly and have shown the distinctive cell-cycle regulation of CENP-A assembly. Establishing how centromeres direct CENP-A assembly to the correct location, to the correct level, and at the right time, are all key future goals. At this time, we lack an understanding of the changes within chromatin that occur during transfer of soluble CENP-A into nucleosomal CENP-A. This is a challenging problem, as manipulating distinct populations of CENP-A (i.e., parent nucleosomes or soluble CENP-A) in cells is not currently possible. Overcoming this obstacle will be important to establish how CENP-A nucleosomes direct CENP-A assembly.

Complementary to our understanding of centromere and kinetochore assembly, studies of the centromere have yielded new insights into epigenetic regulatory mechanisms in eukaryotes. Centromeric nucleosomes and the CENP-A histone may represent one of the best examples of true epigenetic inheritance. As we continue to make progress in our understanding of how CENP-A marks the centromere and how cells interpret that mark to maintain centromeric chromatin, we are likely to gain important insights into other epigenetic inheritance mechanisms. Furthermore, the specific recognition of centromeric chromatin to uniquely specify the functions of the centromere is an exemplar of how chromosomal loci are specialized for distinct functions.