The molecular basis for centromere identity and function

Abstract

The centromere is the region of the chromosome that directs its segregation in mitosis and meiosis. Although the functional importance of the centromere has been appreciated for over 130 years, elucidating the molecular features and properties that endow centromeres with the capacity to orchestrate chromosome segregation has remained a central ongoing challenge. The defining feature of most eukaryotic centromeres is the presence of nucleosomes containing the histone H3 variant, CENP-A (also known as CenH3), which specifies this region epigenetically. In this review, we synthesize the research on the central features of centromere identity, the molecular basis for centromere propagation to the chromosomes of daughter cells and gametes, and the mechanisms by which the centromere recruits the kinetochore to establish a connection to spindle microtubules.

The transmission of an intact genome to daughter cells during cell division is a fundamental requirement for the viability of cells and organisms. In eukaryotes, DNA is packaged into chromosomes. Each chromosome must be faithfully replicated and segregated at every cell division. To achieve accurate segregation, chromosomes rely on a specialized region known as the centromere. The centromere recruits the kinetochore, a proteinaceous macromolecular structure that forms attachments to the microtubules of the mitotic and meiotic spindles. Together, centromeres and kinetochores are the central players in chromosome segregation. Defects in centromere or kinetochore function can lead to the loss or disruption of genomic information, resulting in profoundly deleterious developmental defects or disease¹.

The crucial function of the centromere has been appreciated for over 130 years. The centromere was first observed by light microscopy as the chromosomal attachment site for spindle microtubules in dividing cells² (Figure 1a). As the centromere protects and maintains sister chromatid cohesion during mitosis and meiosis^3–6 this region of the chromosome is also visible in many organisms as the primary constriction on condensed mitotic chromosomes (Figure 1b). Geneticists subsequently combined these cytological observations with the analysis of recombinant progeny to translate genetic maps onto physical ones by defining the positions of genes relative to the centromere^{7, 8}.

Figure 1. Visualization of the centromere.

a) Comparison of images of mitotic Salamander cells hand-drawn by Walther Flemming in 1882² (top) with immunofluorescence images of human cells (bottom) stained for microtubules (green), CENP-A (red) and DNA (blue). The images show cells at different phases of a mitotic cell cycle: late prometaphase-metaphase (left), anaphase (middle) and telophase (right). b) Images of the centromere at increasing resolution. Top left: immunofluorescence image of a mitotic chromosome stained for DNA (blue), CENP-A (red) and CENP-B (a marker for the alpha-satellite DNA repeats present at most human centromeres, green). Top right: electron micrograph of centromeric region of a mitotic chromosome showing centromeric chromatin (dark cloud), kinetochore, and microtubules (indicated by arrows). Image courtesy of Conly Rieder. Bottom left: Immunofluorescence image of stretched centromeric chromatin fibers showing patches of CENP-A (red) interspersed with H3, in this case specifically H3 dimethylated on lysine 4 (H3K4me2, green). Image courtesy of Elaine Dunleavy. Bottom right: Crystal structure of the CENP-A nucleosome⁹⁰. PDB ID: 3AN2

Although the centromere has been described extensively by cytological and genetic approaches, defining the molecular features that confer its functions is a central ongoing challenge⁹. When first defining the term centromere in 1936, Cyril Darlington commented that “[the centromere must] be considered in terms of function rather than form, since the function is evident and the form elusive”¹⁰. Elucidating the “form” of centromeres has remained challenging because centromeres require a complex interplay between numerous molecular features that vary across eukaryotes. Despite this complexity and variation, several common themes have emerged regarding the molecular bases of centromere function. In the vast majority of eukaryotes, centromere specification is primarily epigenetic and depends on the presence of specialized nucleosomes containing the histone H3 variant centromere protein A (CENP-A; also known as CenH3). Centromere function requires the combination of CENP-A-containing nucleosomes, features of the underlying DNA sequence, unique combinations of chromatin marks and interactions with kinetochore proteins.

In this Review, we highlight recent work on the molecular basis for centromere function, with a focus on the vertebrate centromere. We describe the current understanding of the genetic and epigenetic features that define centromeres, the mechanisms of centromere propagation, and the recognition of the centromere by the kinetochore. This work is revealing the elusive form underlying the critical functions of the centromere in the propagation of the genome to cells and gametes.

Centromere DNA structure and function

In the majority of eukaryotes analyzed to date, the centromere is specified epigenetically (Box 1), such that specific DNA sequences are neither strictly necessary nor sufficient for centromere function. Instead, the unifying characteristic of most eukaryotic centromeres is the presence of the histone H3 variant, CENP-A. Nonetheless, recent work has highlighted evolutionary and functional preferences for specific DNA structures that strongly indicate that they contribute to centromere function, as we describe in this section.

Box 1: Evidence for the epigenetic nature of the centromere.

The first evidence that the centromere is specified epigenetically came from human patient samples containing dicentric chromosomes in which one centromere was functionally inactivated without changes to its underlying DNA sequence⁷⁰. Subsequent work has observed epigenetic centromere inactivation in dicentric chromosomes in diverse contexts^198–200. Centromere inactivation is also frequently observed in Robertsonian fusions²⁰¹ and isodicentric Y chromosomes generated by sister chromatid recombination of Y chromosome palindromes²⁰². These data indicate that centromere sequences are not sufficient for centromere function.

Compelling evidence that centromere sequences are not necessary for centromere function comes from neocentromeres (reviewed in ⁶⁹). For example, routine karyotyping of a human patient in 1993 revealed a chromosome fragment that had lost its centromeric DNA, but was nonetheless stably maintained in mitosis, assembled a functional kinetochore, and mediated sister chromatid cohesion in the absence of the canonical underlying DNA repeats⁵⁴ (Figure 2a). Subsequent work revealed cases of inherited neocentromeres, demonstrating that these structures are stable in both mitosis and meiosis^{203, 204}. Neocentromeres have also been generated experimentally in diverse organisms by selecting for their ability to rescue acentric chromosomal fragments^{133, 205–208}. Neocentromeres have been observed in otherwise normal karyotypes in which the centromere DNA sequences remain intact, but have lost centromere function²⁰⁴, reinforcing the insufficiency of centromere sequences proposed by observation of dicentric chromosome inactivation.

A common structure for centromeric DNA sequences

Most eukaryotes have monocentric chromosomes, in which a centromere is assembled at a single localized region (Figure 2a). A notable exception are some nematodes (including Caenorhabditis elegans), and some insects and plants, which assemble a diffuse centromere along the entire length of the chromosome, a phenomenon known as holocentricity¹¹ (Figure 2a). Species with monocentric chromosomes can either have point centromeres, containing short DNA sequences, or regional centromeres (Figure 2a)¹², which contain kilobases to megabases of DNA. Point centromeres are found in some budding yeasts¹², including Saccharomyces cerevisiae¹³, and are defined as those centromeres in which the precise centromeric DNA sequence is necessary and sufficient for kinetochore assembly and DNA segregation^14–16. Regional centromeres are typically comprised of repetitive DNA sequences that contribute to, but are not sufficient for, centromere function. However, some organisms contain regional centromeres that are non-repetitive, such as the yeast Candida albicans¹⁷, or have a mixture of repetitive centromeres and non-repetitive centromeres, such as orangutan¹⁸, horse¹⁹, and chicken²⁰. Repetitive centromeres consist of retrotransposons and/or long arrays of simple tandem repeats, referred to as satellite DNA²¹.

Figure 2. Centromere specification.

a) Diagram of the diverse types of centromeres found across eukaryotes. Holocentric chromosomes assemble a diffuse centromere across the whole chromosome. Monocentric chromosomes assemble a centromere at a single localized site on the chromosome, which is visible as a constriction between the chromosomes in mitosis (known as the primary constriction). Monocentric chromosomes can be further divided into those with point centromeres and those with regional centromeres. Point centromeres contain a specific DNA sequence that is sufficient for centromere function (here illustrated with the S. cerevisiae DNA architecture), which assembles a single CENP-A nucleosome. Regional centromeres contain large regions of DNA that is often repetitive (such as alpha-satellite DNA in primates), and assemble numerous CENP-A nucleosomes. b) Model of the DNA sequence of primate centromeres. Primate centromeres are built from alpha-satellite monomers (triangles), which are largely but not completely identical, as indicated by the different colored triangles. Patterns of these monomers arranged head-to-tail are re-iterated over the centromere core (purple) as higher-order repeats. Some monomers within the centromere core contain a sequence termed the CENP-B box, which binds to the centromere-DNA binding protein, CENP-B. The centromere core is flanked by less ordered monomers which comprise the pericentromere (blue). LINEs, SINEs and other satellites (squares) are found interspersed with alpha-satellite monomers in the pericentromere²¹⁴. c) Schematic showing comparison of macaque and human orthologous chromosomes that have undergone centromere repositioning such that the position of the centromere has moved, but the surrounding markers have not, as indicated by the color blocks, which represent syntenic regions. Part c) adapted from³³.

The precise DNA sequences found at centromeres vary dramatically across evolution, and it has been proposed that this rapid evolution is a consequence of meiotic drive²². Despite the dramatic divergence in centromere sequences, regional centromeres possess a modular structure that is shared by many taxa. Regional centromeres typically consist of a central core, where the CENP-A nucleosomes reside, comprised of homogenous ordered repeats, and an outer heterochromatic domain, termed the pericentromere, that typically contains less ordered repeats (Figure 2a, b). For example, centromeres of the fission yeast Schizosaccharomyces pombe contain a centromere core of non-repetitive sequences flanked by perfect inner inverted repeats and less ordered outer repeats²³. Similarly, the Mus musculus centromere core is comprised of minor satellite arrays containing homogenous 120 bp repeats flanked by less-ordered ~234 bp major satellite repeats²⁴. Primate centromeres are built on a single 171 bp monomer termed alpha-satellite^25–28. In humans and other great apes, the alpha-satellite is arranged head-to-tail to form higher order repeats that are themselves re-iterated across the centromere. The human pericentromere contains flanking monomers that lack higher order repeats and share reduced identity between monomers (see ²⁹ for further review of centromeric DNA structure) (Figure 2b). Thus, centromeres frequently arrange their divergent centromere sequences in a common repetitive structure.

Evolutionary preference for repetitive DNA structures

Cytogenetic comparisons between closely related species have revealed that some centromeres adopt new positions over evolutionary time subsequent to a speciation event without transposing the surrounding genetic markers, a phenomenon known as centromere repositioning³⁰ (Figure 2c). These structures are referred to as evolutionary new centromeres (ENCs) and have been observed in primates and other mammals (reviewed in ³¹) and birds³². A striking property of ENCs is that they typically contain the same molecular features as the “old” centromeres within the karyotype, including the species-specific satellite DNAs. For example, all nine ENCs in macaque contain alpha-satellite arrays and large segmental duplications, making them indistinguishable from “old” macaque centromeres³³. Thus, ENCs are postulated to be seeded upon new, non-repetitive DNA sequences in a manner analogous to neocentromeres (Box 1), but subsequently acquire their species-specific satellite DNA over time. The recent ENCs on orangutan chromosome 9 and horse chromosome 11 have not acquired satellite DNA, and may represent intermediates in this maturation process^{18, 19}. Chromosomes harboring ENCs also exhibit a decay of the satellite sequences at the ancestral site³⁴. The acquisition of a modular structure of tandem repeats by ENCs further supports a contribution of such DNA structures to centromere function.

Contributions of DNA sequences to centromere function

Although specific sequences are not necessary or sufficient for centromere function in some contexts, centromere DNA sequences can confer centromere function to exogenous DNA in diverse organisms, indicating that they can have a role in the de novo specification of a centromere. The most striking example of this comes from the budding yeast, S. cerevisiae, which contains a ~125 bp sequence that is sufficient to confer mitotic and meiotic stability to an exogenous minichromosome¹³. Although the discovery of these centromeres was strongly facilitated by the strictly sequence-dependent point centromeres of budding yeast, sequences that confer centromere functions have also be found in organisms with regional centromeres, including Schizosaccharomyces pombe³⁵ and primates³⁶.

Extensive work has sought to use alpha-satellite DNA to build human centromeres de novo and generate human artificial chromosomes (HACs). In pioneering work, cloned alpha-satellite DNA from human chromosomes enabled linear human mini-chromosomes³⁷ and yeast artificial chromosomes³⁸ to be stably inherited in human cells. These systems demonstrated that alpha-satellite DNA was sufficient to initiate centromere formation. The analysis of HAC formation also permitted structure-function studies of the alpha-satellite DNA, revealing a key role for the higher order repeats³⁹. The mechanisms by which alpha-satellite DNA sequences initiate centromere formation are the subject of current investigations. Recently, it was suggested that alpha-satellite arrays adopt chromatin marks that favor the deposition of CENP-A nucleosomes (see below)^{40, 41}. Together, this work is beginning to bridge the gap between the centromere DNA sequences and the epigenetic marks required for centromere function.

DNA sequence-dependent binding proteins at the centromere

The existence of common DNA sequence motifs at the centromeres of most organisms within a species presents the opportunity for recognition by DNA binding proteins that can confer centromere functions. In budding yeast, which contains sequence-specific centromeres, kinetochores are indeed assembled upon a foundation of the sequence-specific DNA binding CBF3 complex, which binds to the CDEIII DNA element at the centromere⁴². In organisms where centromeres can form in the absence of specific centromere sequences, the potential roles for a DNA-sequence-specific binding protein are more challenging to predict, particularly because centromere sequences vary dramatically across species, whereas centromere proteins are largely conserved. The only known centromere sequence element that is conserved between primates and rodents is the CENP-B box^{43, 44}, a 17 bp sequence that binds to the protein CENP-B⁴⁵. The CENP-B box is found in the minor satellite of Mus musculus and some monomers within the HOR of human alpha-satellite repeats. Although Mus musculus and great apes share the CENP-B box, some primates lack CENP-B boxes⁴⁶, and the rodent M. caroli contains a divergent CENP-B box that retains the nine basepairs required for CENP-B binding⁴⁷.

Due in part to its inconsistent conservation, the importance of the CENP-B box and the protein itself remain poorly understood. CENP-B directly interacts with and stabilizes both CENP-A nucleosomes and CENP-C to contribute to centromere function^48–50. However, CENP-B knockout mice are viable^51–53 and neocentromeres are maintained without acquiring CENP-B binding capability⁵⁴. Perhaps most intriguingly, the human Y chromosome centromere lacks CENP-B boxes⁴³ and does not bind detectable CENP-B protein⁵⁵. Similarly, the Y chromosome of Mus musculus lacks the minor satellite sequences that contain the CENP-B box⁵⁶. However, Y chromosome sequences are not sufficient to generate HACs without acquiring other centromeric alpha-satellites from the host cells^{37, 57} and HAC formation requires the CENP-B box^{39, 58}. Together, these data indicate that CENP-B, like the centromere sequences it binds, is not strictly required at the centromere but makes functional contributions to maximize mitotic fidelity that contribute particularly to the generation of centromeres de novo.

Centromere epigenetics and the CENP-A nucleosome

CENP-A is an epigenetic hallmark of centromeres

In most eukaryotes, the defining feature of centromeres is the presence of nucleosomes containing the histone H3 variant CENP-A. CENP-A was first identified as a centromere-specific antigen recognized by antibodies from human patients with the autoimmune disease CREST syndrome⁴⁵. Concurrent and subsequent work found that CENP-A was a component of chromatin with biochemical similarity to histones^59–62, and shared homology with histone H3^{61, 63}. CENP-A homologues have been identified in diverse eukaryotes based on their similarity to histone H3^64–66. As a centromere-specific histone H3 variant, CENP-A provides a compelling candidate for an epigenetic mark of centromere identity^{67, 68}. Consistent with a fundamental requirement for CENP-A in centromere function, CENP-A is found at all identified neocentromeres⁶⁹, as well as the active centromeres of dicentric chromosomes⁷⁰, and is essential for the localization of all known kinetochore components^{48, 71, 72}. Importantly, artificial targeting of CENP-A to an ectopic chromosomal locus is also sufficient to generate structures capable of directing microtubule attachment and chromosome segregation^73–76.

CENP-A nucleosomes possess unique structural properties

The existence of a centromere-specific histone raises intriguing possibilities regarding how CENP-A is specialized to mark the position of the centromere and recruit downstream kinetochore proteins. At the sequence level, CENP-A contains two important regions: a histone fold domain that shares 62% sequence identity with histone H3 in humans, and an N-terminal tail that differs more significantly from H3⁶³ and even between CENP-As from different species⁷⁷ (Figure 3a). Within the histone fold domain, the first loop and second alpha helix (L1-alpha 2) are necessary for targeting CENP-A to the centromere, and are sufficient to confer centromere targeting when introduced into chimeras with histone H3^{78, 79}. Therefore, this region is referred to as the CENP-A targeting domain (CATD) (Figure 3a). Sequences within CENP-A nucleosomes also confer centromere-specific functions through the direct binding of the core kinetochore proteins CENP-N and CENP-C (Figure 3a). In particular, CENP-N binds directly to the CATD of CENP-A ^{76, 80, 81}. CENP-C makes extensive contacts with the CENP-A nucleosome: with the six residues of the CENP-A C-terminal tail ^{80, 82, 83}, with other histones within the CENP-A nucleosome ⁸², and with the CENP-A CATD ^{76, 84}. The CENP-A N-terminal tail has also been implicated in the recruitment of kinetochore proteins in different organisms^{48, 76, 85–87}. Thus, variations between CENP-A and H3 at the sequence level confer centromere specificity and kinetochore assembly properties to CENP-A.

Figure 3. Specialization and propagation of CENP-A.

a) Model of human CENP-A primary and secondary structure showing conservation with histone H3. Each segment corresponds to a single amino acid, and is colored according to its conservation with human H3.1 as indicated. The first N-terminal amino acid, shown detached, represents the cleaved initiator methionine. Barrels represent alpha helices, and rods represent loops. Within the histone fold domain, the helices are designated alpha1 through alpha3, and the loops are designated L1 and L2. L1 and alpha2 comprise the CENP-A targeting domain, which is sufficient to target CENP-A to centromeres due to its interaction with the CENP-A chaperone, HJURP. This region also binds to CENP-N⁸¹ and is important for CENP-C recruitment^{76, 84}. CENP-C also binds to the C terminal residues of CENP-A^{80, 82, 83}. b) Model for the changes to CENP-A chromatin over the cell cycle. The timing of the localization of the CENP-A deposition factors is indicated. At S phase, existing CENP-A is partitioned between the replicated sisters, and gaps filled with histone H3.3. Although centromere localization of M18BP1 precedes recruitment of Mis18alpha and beta¹¹⁶, the precise onset of its localization has not been established. By mitosis, M18BP1 localizes to centromeres, followed by Mis18alpha and Mis18beta at mitotic exit. An HJURP dimer²¹⁵ is recruited in early G1 to direct new CENP-A deposition. New CENP-A is stabilized in late G1 by MgcRacGAP and RSF1. Defining the mechanisms that remove these assembly factors once CENP-A deposition is complete also remains an important open question. c) Model for the two-step regulation of CENP-A deposition. CDK prevents CENP-A deposition outside of G1 phase by inhibiting Mis18 complex localization, Mis18 complex assembly and HJURP recruitment. Plk1 binds to the Mis18 complex to promote CENP-A deposition at centromeres during G1.

CENP-A nucleosomes also have structural distinctions from canonical H3-containing nucleosomes with the potential to make contributions to centromere function (Figure 4). The structural properties of the CATD make the free (CENP-A-H4)₂ tetramer more conformationally rigid than the (H3-H4)₂ tetramer as determined by hydrogen-deuterium exchange, and cause the CENP-A-CENP-A interface to be rotated when compared to the H3-H3 interface in a canonical nucleosome, generating a more compact structure^{88, 89}. However, in the crystal structure of the octameric nucleosome, the CENP-A-CENP-A axis appears similar to the H3-H3 axis from canonical nucleosomes⁹⁰. Recent work indicates that CENP-A nucleosomes in solution sample both forms, and that binding of CENP-C shifts the nucleosome to the state similar to that of canonical nucleosomes⁹¹. In addition, there has been an extensive ongoing debate regarding whether the CENP-A nucleosome forms a hemisome (with one molecule each of CENP-A, H4, H2A and H2B) that wraps DNA in a right-handed manner, or an octamer (reviewed in ⁹²). Finally, CENP-A nucleosomes appear to confer structural alterations to centromeric chromatin. For example, CENP-A arrays are more condensed^{93, 94}, but with a DNA entry and exit site that is loose compared to canonical nucleosomes^{90, 93–96}, a property that is enhanced by CENP-C binding⁹¹. Thus, sequence and structural specializations of CENP-A nucleosomes and CENP-A containing-chromatin generate fundamental distinctions between centromeric chromatin and bulk chromatin.

Figure 4. Centromeric chromatin.

Model of the epigenetic modifications at the core centromere, CENP-A domain, and the pericentromere. In addition to the sequence and structural specializations that differentiate CENP-A chromatin from bulk chromatin, posttranslational modifications of CENP-A nucleosomes contribute to centromere function. Human CENP-A is mono-ubiquitinated at lysine 124 within the histone fold domain by CUL4-RBX1-COPS8²¹⁶ to promote its centromere targeting. Acetylation at this lysine 124 residue has also been reported²¹⁷. Finally, diverse other posttranslational of CENP-A^217–219 and H4 in the CENP-A nucleosome²²⁰ have been described. Defining the functional contributions of these modifications remains a central challenge.

Centromere propagation

Faithful centromere inheritance is critical for the transmission of the genome, as failure to propagate the centromere results in the inability of a chromosome to attach to the mitotic spindle, leading to loss of the chromosome and the information it encodes. On monocentric chromosomes, the spurious formation of a centromere at two distinct loci allows a single chromatid to attach simultaneously to opposing spindle poles, resulting in mis-segregation or fragmentation of the chromosome by spindle forces. The fragmentation of dicentric chromosomes can result in breakage-fusion-bridge cycles that confer cascading chromosomal instability^{97, 98}. Therefore, the centromere must be faithfully inherited at a single site on each chromosome through all mitotic and meiotic divisions (Box 2).

Box 2: Transmission of the CENP-A nucleosome during meiosis.

In addition to its central role in mediating mitotic divisions, the centromere must also be propagated and direct chromosome segregation during meiosis to be transmitted to the progeny. Transmission of Y chromosome neocentromeres between generations²⁰³ demonstrates that the position of the human centromere is heritable through the male germline independently of the underlying DNA sequence. Unlike the majority of canonical histones, CENP-A is not exchanged for protamines during sperm development in mammals⁶⁰, Xenopus²⁰⁹ or Drosophila^{210, 211}, and can therefore provide a template for the centromeres in the progeny. Indeed, in Drosophila, maintenance of CENP-A (CID) in the sperm is required for centromere propagation and the faithful segregation of the paternal chromosomes in the embryo, as sperm chromosomes lacking CENP-A are unable to template a centromere de novo ²¹¹. In contrast, in C. elegans, CENP-A is not continuously maintained throughout meiosis and so does not follow this self-templating pattern, as sperm do not contribute CENP-A following fertilization, and CENP-A is instead provided by the oocyte, which removes CENP-A in pachytene of prophase I and reloads it in diplotene²¹².

In those organisms that maintain their centromeres through meiosis, the molecular mechanisms that replenish CENP-A following meiotic S phase are poorly understood. Several differences from the mechanisms of CENP-A replenishment during the mitotic cell cycle have been proposed. In Drosophila, CENP-A is assembled during prophase I of female meiosis, and during both prophase I and after exit from meiosis II in the male²¹⁰. CENP-A deposition is similarly biphasic during the meiotic divisions to produce male gametes in rye²¹³. The mechanisms that transmit centromere position and features through the germline in vertebrates remain a key unanswered question.

The CENP-A deposition machinery

In most eukaryotes, centromere inheritance requires the transmission of CENP-A nucleosomes to maintain the epigenetic mark on each sister chromatid. Fundamental to this transmission is the striking stability of CENP-A, which does not exchange once it is incorporated at centromeres^{91, 99, 100}, and is conservatively partitioned between the newly replicated sister chromatids during S phase^{91, 99, 100}. Unlike canonical histones, the deposition of new CENP-A is uncoupled from DNA replication, such that the occupancy of CENP-A molecules at the centromere is halved during mitosis when the centromere recruits the complete kinetochore (Figure 3b). The nature of centromeric chromatin during mitosis following this dilution remains an area of active investigation, with current models indicating that the gaps left by this dilution are filled by H3.3¹⁰¹. In human cells, new CENP-A molecules are deposited during the subsequent G1⁹⁹.

The deposition of new CENP-A requires the coordinated activity of several assembly factors (Figure 3b). CENP-A has a dedicated histone chaperone, HJURP^{102, 103}, which recognizes CENP-A as distinct from H3 via specific contacts between the CENP-A targeting domain (CATD) and the N terminal CENP-A-binding domain of HJURP^104–107. The HJURP CENP-A-binding domain is homologous to the yeast CENP-A chaperone Scm3¹⁰⁸, and is sufficient to direct the incorporation of CENP-A at an ectopic locus⁷⁵. HJURP localizes to centromeres only during G1^{102, 103}, when new CENP-A deposition occurs. Consistent with this, HJURP does not participate in the partitioning of CENP-A between sister chromatids during S phase¹⁰⁰.

In addition to HJURP, CENP-A deposition in G1 requires the three-subunit Mis18 complex comprised of Mis18α, Mis18β, and Mis18 binding protein 1 (M18BP1)¹⁰⁹ (also known as KNL2¹¹⁰). Intriguingly, not all components of the Mis18 complex are conserved across eukaryotes, with a single Mis18 homolog in fungi¹¹¹ (without an identified M18BP1), and an M18BP1 homolog (KNL2), but no Mis18α/β homologues in C. elegans¹¹⁰. In D. melanogaster, the Mis18 complex and HJURP functions appear to be combined in a single molecule, CAL1^{112, 113}. M18BP1 has been shown to interact with CENP-C in both human cells and Xenopus laevis^{114, 115}. As CENP-C binds directly to CENP-A nucleosomes as described above, this provides a mechanism to ensure that the Mis18 complex and HJURP are recruited only to sites of pre-existing centromeres to locally direct the incorporation of new CENP-A. The interaction between M18BP1 and CENP-C is crucial for the recruitment of the Mis18 complex during CENP-A assembly in G1 phase in human cells^{114, 116}. However, Xenopus M18BP1 is recruited via CENP-C during mitosis but not interphase, suggesting that additional M18BP1 recruitment mechanisms exist^{84, 115}. CENP-C has also been proposed to contribute to CENP-A deposition beyond Mis18 complex recruitment⁸⁴, including by binding to HJURP directly¹¹⁷. Finally, CENP-C⁹¹, the RSF complex¹¹⁸, and the centralspindlin component MgcRacGAP¹¹⁹ have been implicated in the maintenance of CENP-A once it is incorporated at centromeres. Together, these centromere-specialized assembly factors ensure the specific incorporation of CENP-A at centromeres.

Regulation of CENP-A deposition

Multiple regulatory safeguards have been identified that ensure the faithful deposition of new CENP-A-containing nucleosomes exclusively at centromeres. In metazoa, CENP-A deposition occurs around mitosis or following mitotic exit^{99, 115, 120–122}. This temporal restriction isolates CENP-A deposition from the deposition of canonical H3, which is coupled to DNA replication in S phase. The cell cycle restriction of CENP-A deposition relies heavily on phosphorylation downstream of cyclin-dependent kinase (CDK) ¹²³ (Figure 3c). Ongoing work indicates that CDK negatively regulates CENP-A incorporation at numerous steps. In Drosophila, the degradation of cyclin A plays a key role in deposition of CENP-A^{112, 121}. In human cells, CDK phosphorylates the Mis18 complex subunit M18BP1 to reduce its centromere localization¹²³, and to prevent recruitment of the Mis18α and Mis18β subunits¹¹⁶ outside of G1. CDK phosphorylation of HJURP disrupts its localization to centromeres¹²⁴, whereas CDK phosphorylation of CENP-A itself on serine 68 has been reported to inhibit the CENP-A-HJURP interaction¹²⁵, although the role of serine 68 in CENP-A deposition is controversial^{48, 76, 84, 100, 105, 107}.

In addition to this temporal regulation by CDK, CENP-A deposition requires a licensing step by Polo-like kinase 1 (Plk1)¹¹⁶ (Fig. 3c). Thus, centromere propagation requires a two-step regulatory paradigm analogous to the regulation of DNA replication by CDK and Dbf4-dependent kinase (DDK)¹²⁶. Plk1 binds to and phosphorylates the Mis18 complex to promote Mis18 complex localization and license the centromere for CENP-A deposition¹¹⁶. Bypassing both the CDK regulation of Mis18 complex assembly and Plk1 licensing by constitutively targeting the Mis18α subunit to the centromere results in CENP-A deposition throughout the cell cycle and severe mitotic defects¹¹⁶. This indicates that the temporal isolation of CENP-A deposition is important for centromere function.

Generation of a CENP-A permissive chromatin environment

Although CENP-A is an essential component of centromeres, it is not the sole driver of centromere specification. CENP-A homologs are absent in some organisms, including trypanosomes and some insects with holocentric chromosomes^{127, 128}, raising the possibility that alternate strategies for centromere specification have arisen during evolution. Even in CENP-A-containing organisms, additional molecular features contribute to defining an active centromere, including the properties of the underlying DNA sequence, the composition of the surrounding chromatin, and post-translational modifications of CENP-A itself (Figure 4). Moreover, individual CENP-A molecules are found frequently at non-centromeric sites throughout the chromosomes¹²⁹, indicating that the presence of CENP-A alone is not sufficient for centromere formation.

The core centromere and pericentromere are distinguished not only by organization of their DNA sequence repeats as described above, but also by distinct chromatin signatures that are crucial for their functions. Early studies associated centromeres with heterochromatin ¹³⁰ and subsequent work has found that the pericentromere in particular is heterochromatic, containing hypermethylated H3 lysine 9 (H3K9)^{131, 132}, although non-repetitive centromeres and neocentromeres frequently lack surrounding heterochromatin^{133, 134}. In contrast to the heterochromatic pericentromere, at the core centromere CENP-A-containing nucleosomes are interspersed with canonical H3-containing nucleosomes with transcriptionally permissive marks, particularly dimethylated histone H3 lysine 4 (H3K4me2)^135–137 and H3K36me2⁴⁰ in human and Drosophila cells. Recent analyses of HAC formation and maintenance have revealed that artificially increasing heterochromatin at the alpha-satellite array is detrimental for CENP-A deposition and centromere function^{41, 138}, whereas H3K4me2 and increased H3K9 acetylation promote CENP-A maintenance^{40, 41}. This indicates that both the presence of transcriptionally permissive marks and absence of heterochromatin in the centromere core are important for CENP-A localization to centromeres (Figure 4).

The importance of chromatin marks that are permissive for transcription at the core centromere raises the possibility that transcription of the centromere and pericentromere plays a role in centromere propagation and function. In fission yeast, transcripts from the pericentromeric repeats contribute to the formation of pericentromeric heterochromatin, which in turn is required for de novo CENP-A deposition on mini-chromosomes¹³⁹. In addition, transcripts derived from the centromere core have been reported in diverse organisms^{140, 141} (Figure 4). In human cells, RNA polymerase II and several transcription factors localize to mitotic centromeres¹⁴² and transcripts have been detected from the alpha-satellite sequences of HAC centromeres⁴⁰. Broadly disrupting RNA polymerase I or II results in kinetochore defects^{142, 143} as well as defects in the deposition of new CENP-A nucleosomes¹⁴⁴. However, tethering strong transcriptional activators to the centromere is deleterious to centromere function in many organisms^{138, 145} indicating that the transcriptional requirement for centromere identity and function must be finely tuned.

Chromatin remodelers associated with active transcription have also been implicated in the deposition of new CENP-A (Figure 4), including RSF1, FACT, CHD1, and RbAp46 and 48^{103, 109, 111, 118, 119, 146–148}. These proteins may facilitate new CENP-A deposition through the generation of the necessary transcriptionally permissive centromere core, or may play a direct role in remodeling centromeric chromatin to accommodate its oscillations between maximal and half-maximal CENP-A occupancy throughout the cell cycle (Figure 3b, Figure 4). For example, if H3.3 replaces CENP-A nucleosomes following DNA replication, this H3.3 must be exchanged for new CENP-A during the following G1. The Mis18 complex has also been proposed to contribute to the chromatin remodeling in anticipation of new CENP-A deposition, including by recruiting factors that regulate DNA methylation¹⁴⁹ and histone acetylation^{109, 111}. As a result, tethering a histone acetyltransferase to a HAC centromere can partially complement depletion of the Mis18 complex ⁴¹. However, recent work indicates that the Mis18 complex also functions directly in the CENP-A deposition process by interacting with the HJURP chaperone^{150, 151}.

Restriction of CENP-A deposition

The regulated deposition of CENP-A nucleosomes ensures the epigenetic propagation of the centromere at a persistent location on each chromosome. Many organisms also have strategies to prevent CENP-A deposition at non-centromeric sites, where they could make inappropriate attachments to the mitotic spindle. In S. cerevisiae, mis-targeted CENP-A is removed by the combined action of the FACT chromatin remodeler and the E3 ubiquitin ligase Psh1, which targets ectopic CENP-A for degradation^{152, 153}. In fission yeast, the proteasome subunit Rpt3 interacts with CENP-A and has been implicated in restricting the size of the CENP-A domain¹⁵⁴. However, a similar proofreading mechanism to remove ectopic CENP-A has not yet been identified in vertebrates, consistent with the persistence of CENP-A molecules at non-centromeric sites in the genome in human cells¹²⁹.

CENP-A deposition is also restricted within the centromere. In humans, mouse and chicken, the CENP-A domain occupies only a small portion of the core centromere sequences^{129, 155, 156}. There is significant variation in the size of the CENP-A domain among human chromosomes (between 0.4 Mb and 4.2 Mb for a set of analyzed X and Y chromosomes¹⁵⁷), although an approximately equivalent ratio between the size of the CENP-A domain and the alpha satellite array is maintained¹⁵⁷. The CENP-A domain of neocentromeres is restricted to an even smaller region, with reports between 40 kb and 0.5 Mb^{96, 133, 134, 158}. How the CENP-A domain is restricted in size in vertebrates remains an area of active investigation. Exogenous CENP-A expression in human cells leads to down-regulation of the endogenous CENP-A protein⁹⁹, and CENP-A overexpression far beyond this level results in mis-localization of CENP-A to chromosome arms^{63, 86, 159}. These data indicate that the restriction of the CENP-A domain occurs at least in part at the level of modulating total protein in the cell, as recently proposed in human cells¹²⁹. In chicken and Drosophila, high local concentrations of the CENP-A chaperones HJURP or CAL1, respectively, can also drive centromere expansion^{151, 160}. Intriguingly, these homeostasis mechanisms maintain CENP-A in large excess of the amount required for kinetochore function, as cells depleted of CENP-A to as little as 10 percent or even 1 percent of its initial level recruit kinetochore proteins and at least partially direct chromosome segregation^{48, 71}.

Centromere recognition

The centromere achieves its key function – the segregation of its corresponding chromosome – by recruiting the kinetochore, the macromolecular structure that mediates attachment to the microtubules of the mitotic spindle and acts as a signaling hub to ensure accurate chromosome segregation¹⁶¹. Thus, understanding how centromere form begets its function hinges critically on defining the network that connects the centromere components to the proteins of the kinetochore.

When first proposed, the terms centromere and kinetochore referred to the same structure – the region of the chromosome that attaches to the spindle fiber^{10, 162}. With the increasing resolution of this region by electron microscopy (Fig. 1b), the terms became delineated, with the kinetochore defined as an electron-dense structure in which microtubules embed, and the centromere used to refer to the underlying chromatin¹⁶³. However, the molecular connections between the centromere and kinetochore remain incompletely understood.

Components of the centromere-kinetochore interface

Establishing the architecture of the centromere-kinetochore interface has been accelerated by the discovery of multiple key molecular players over the last ten years^164–167. The proteins of the centromere-kinetochore interface are collectively referred to as the Constitutive Centromere Associated Network (CCAN) (also referred to as the Interphase Centromere Complex (ICEN)) (Figure 5). The CCAN is a group of 16 proteins that localize to the centromere throughout the cell cycle¹⁶¹. These proteins are designated in vertebrates with alphabetical CENP- names (CENP-C, CENP-H, CENP-I, CENP-K, CENP-L, CENP-M, CENP-N, CENP-O, CENP-P, CENP-Q, CENP-U, CENP-R, CENP-T, CENP-W, CENP-S, CENP-X) ^{45, 164–172}, although other CENP- named proteins do not represent constitutive centromere components. Within the CCAN, these proteins can be combined into five groups: CENP-C, the CENP-L-N complex^{81, 173}, the CENP-H-I-K-M complex^{164, 165, 174}, the CENP-O-P-Q-U-R complex^{175, 176}, and the CENP-T-W-S-X complex¹⁷⁷ (Figure 5a). Together, these proteins recognize centromeric chromatin and connect it to the kinetochore.

Figure 5. Contributions of the Constitutive Centromere Associated Network (CCAN) at the centromere-kinetochore interface.

a) Diagram of the proteins of the CCAN. The sixteen proteins of the CCAN, designated by CENP- and a letter, can be grouped into sub-complexes as indicated. The sub-complexes are grouped according to functions that have been reported for at least one of their subunits. KMN: a network of KNL1, Mis12 complex and Ndc80 complex, which together bind to microtubules. b) Comparison of the crystal structures of the tetramer comprised of the histones CENP-A and H4 in the context of the nucleosome (PDB ID: 3AN2)⁹⁰ (H2A, H2B and DNA are excluded for clarity) with the heterotetramer comprised of the histone fold-containing proteins CENP-T, -W, -S, and -X heterotetramer (PDB 3VH5)¹⁷⁷. c) A simplified model of the connectivity from the centromere, to the kinetochore, to the microtubule during mitosis. The contributions of CENP-C and CENP-T to recruiting the microtubule binding-interface of the kinetochore are highlighted, and the other CCAN components are excluded from this model for clarity.

Dissecting the contributions of the CCAN to centromere recognition presents a particular challenge due to their differing functional requirements between organisms. Although the CCAN is largely conserved between yeast and human^{178, 179}, it is dispensable in yeast with the exception of the CENP-U homolog Ame1 and the CENP-Q homolog Okp1^{180, 181}. In mammals, CENP-U is essential for early mouse development¹⁸², but eliminating CENP-U and CENP-Q results in relatively mild phenotypes in tissue culture cells^{176, 182}. In addition, some organisms such as Drosophila and C. elegans have a minimal CCAN, for which the only identified CCAN homolog is CENP-C. In this section, we will review the ongoing work to define the precise molecular roles of the CCAN in kinetochore assembly and faithful chromosome segregation.

Recognition of centromeric chromatin

A central requirement for connecting the centromere and kinetochore is that the CCAN directly recognize centromeric chromatin. Although each of the CCAN proteins can be co-immunoprecipitated with CENP-A nucleosomes, only CENP-C and CENP-N have been reported to bind to nucleosomes directly^{80–82, 91} (Figure 5a). These proteins recognize the key structural distinctions between CENP-A and H3 (see above). Together, CENP-C and CENP-N provide direct connection for kinetochore proteins to CENP-A.

In addition to binding directly to CENP-A, CCAN components make additional contacts with centromeric chromatin to build a robust platform on the centromere. Several CCAN proteins bind directly to DNA, including CENP-C¹⁸³, CENP-Q¹⁷⁵, and the CENP-T-W-S-X complex¹⁷⁷. The CENP-T-W-S-X complex is particularly intriguing, as it is comprised of histone-fold containing proteins^{177, 184} and adopts a structure similar to canonical nucleosomes (Figure 5b). In this structure, CENP-T-W and CENP-S-X form dimer pairs that can be combined into a CENP-T-W-S-X heterotetramer¹⁷⁷ or a (CENP-T-W-S-X)₂ octamer¹⁸⁵. The CENP-T-W-S-X complex wraps DNA, inducing positive supercoils^{177, 185}, and protects a region of ~ 100 bp from micrococcal nuclease digestion¹⁷⁷, indicating that it may integrate directly into centromeric chromatin. The importance of these nucleosome-like properties for CENP-T-W-S-X localization, as well as centromere and kinetochore function, is still being elucidated.

The CCAN directs outer kinetochore assembly

The CCAN assembles the outer kinetochore on a platform of centromeric chromatin. In particular, CENP-C and CENP-T form parallel, but non-redundant pathways that recruit the key microtubule binding proteins of the kinetochore, the KNL1/Mis12/Ndc80 (KMN) network^{159, 186–189} (Figure 5c). Indeed, artificial targeting of fragments of CENP-C or CENP-T to an ectopic chromosomal locus is sufficient to recruit the KMN network and generate a kinetochore-like structure that can direct chromosome segregation^{159, 190}. In budding yeast, CENP-U forms a third pathway to recruit the KMN network¹⁷⁵. In human cells, CENP-I has also been reported to interact with the microtubule binding proteins of the kinetochore¹⁹¹. These protein interactions are regulated in most eukaryotes such that the CCAN only recruits a full kinetochore during mitosis¹⁹². Specifically, phosphorylation by Aurora B kinase promotes interactions between CENP-C and the Mis12 complex during mitosis^{191, 193}. In addition, the Ndc80 complex is sequestered outside of the nucleus throughout interphase, and is thereby spatially separated from the CCAN until mitosis when CDK phosphorylation promotes its direct interaction with CENP-T^{159, 189, 192}.

Although the interactions between the KMN network and the CCAN proteins CENP-C and CENP-T are increasingly well characterized, the contributions of the other CCAN proteins to generating a platform for the kinetochore remain poorly defined. Ongoing work is seeking to define the architecture and associations of the remaining components^{71, 80, 81, 117, 164, 174, 194}. This work indicates that CENP-C is a keystone molecule within the CCAN and is required for the recruitment of several other CCAN components^{80, 117, 175, 194}, in addition to its role in recruiting the KMN network and promoting CENP-A deposition described above. Recent work has also implicated the CENP-H-I-K-M complex in the recruitment of CENP-T-W-S-X¹⁷⁴, potentially providing centromere-specificity to the DNA-binding activity of CENP-T-W-S-X. Further defining the physical and functional relationships between the various CCAN sub-complexes remains a key goal.

Defining the contribution of the CCAN to centromere and kinetochore function

Ultimately, the central challenge that remains at the centromere-kinetochore interface is to define the contributions of the CCAN proteins to centromere and kinetochore function. Beyond building a molecular bridge to the microtubule-binding interface, CCAN proteins may make additional contributions to chromosome segregation (Figure 5a). For example, recent work has suggested that the vertebrate CCAN plays a key role in resisting the forces generated by spindle microtubules¹⁹⁵, as well as controlling metaphase oscillations¹⁹⁶ and chromosome congression through recruiting the motor protein CENP-E¹⁹⁷. In addition, several CCAN proteins, including CENP-C, CENP-N and CENP-I, have been shown to play key roles in the deposition of new CENP-A nucleosomes at centromeres^{66, 80, 81, 114, 115, 146, 190} (Figure 5a), presenting an appealing model for the propagation of the centromere via kinetochore proteins. The ongoing advances in the identification of CCAN components and subcomplexes will provide further insight into the functional contributions of the CCAN.

Conclusions

Research in centromere biology continues to provide important insights into the molecular mechanisms that underlie the specification, propagation, and recognition of this epigenetically defined chromosomal locus. However, many important mysteries remain to be unraveled. For example, continuing to define the contributions of DNA architecture and chromatin marks to CENP-A deposition is crucial for understanding why some sites of spurious CENP-A deposition result in neocentromere formation whereas others are maintained in the genome inertly. Other key goals for future work include establishing the mechanisms by which the centromere is disassembled and re-assembled to allow passage of the DNA replication fork in S phase, understanding the differences in CENP-A transmission during the meiotic cell cycle, and characterizing the connectivity between CENP-A and the full mitotic kinetochore. Through the development of cytological, biochemical, and genetic tools, researchers are defining Cyril Darlington’s form of the centromere in increasing molecular detail. Future work faces the challenge of further dissecting endogenous centromeres and building them de novo to define how the form imparts the function.

Key points.

• Centromeres are defined epigenetically and require the presence of the centromere-specific histone H3 variant CENP-A.

• Although DNA sequences are not strictly required for centromere specification, similarities in the organization of centromere DNA suggest that DNA structures contribute to centromere function.

• CENP-A nucleosomes contain unique sequence and structural features that allow them to stably mark the centromere and be recognized by kinetochore components.

• CENP-A propagation requires specialized deposition factors and tight regulatory control.

• The centromere directs the assembly of the kinetochore via the sixteen subunit Constitutive Centromere Associated Network (CCAN).

Glossary

Centromere

the region of a chromosome that directs its segregation

Evolutionary New Centromere (ENC)

a centromere at a different site from the centromere of the chromosome ancestor, where the movement of the centromere cannot be parsimoniously explained by a simple chromosome rearrangement

Histone chaperone

a protein that binds to histones to facilitate nucleosome assembly

Holocentric

a chromosome that assembles a centromere along the entire length of the chromosome

Human Artificial Chromosome (HAC)

a unit of exogenous DNA that segregates autonomously in human cells

Kinetochore

a macromolecular structure that connects the centromere to the microtubule polymers of the spindle to orchestrate and regulate chromosome segregation

Monocentric

a chromosome that assembles a centromere at a single distinct region along the chromosome

Meiotic drive

preferential transmission of a genetic element during meiosis, such that it is represented in more than 50% of the gametes of a heterozygote

Neocentromere

a region of a chromosome that has the functional characteristics of a centromere but occurs at a site distinct from the site of centromere formation for the chromosome in most organisms of the species, and lacks canonical centromere DNA sequences

Biographies

Iain Cheeseman is a faculty member at the Whitehead Institute for Biomedical Research and the Department of Biology at MIT. He carried out his doctoral work on budding yeast centromeres and kinetochores with David Drubin and Georjana Barnes at the University of California, Berkeley, USA, and then conducted postdoctoral studies on nematode and human kinetochores with Arshad Desai at the Ludwig Institute for Cancer Research, San Diego, USA. His laboratory studies kinetochore function in human tissue culture cells using biochemical and cell biological approaches.

Kara McKinley received her A.B from Princeton University in molecular biology and is currently a graduate student in the Department of Biology at MIT. Kara works in the Cheeseman laboratory at the Whitehead Institute for Biomedical Research on the propagation and recognition of CENP-A-containing chromatin.