Centromere proteins CENP-S and CENP-X are members of the constitutive centromere-associated network, which is a conserved group of proteins that are needed for the assembly and function of kinetochores at centromeres. Intriguingly CENP-S and CENP-X have alter egos going by the names of MHF1 (FANCM-associated histone-fold protein 1) and MHF2 respectively. In this guise they function with a DNA translocase called FANCM (Fanconi’s anemia complementation group M) to promote DNA repair and homologous recombination. In the present review we discuss current knowledge of the biological roles of CENP-S and CENP-X and how their dual existence may be a common feature of CCAN (constitutive centromere-associated network) proteins.
A brief overview of the kinetochore and centromere-associated proteins
The faithful segregation of chromosomes during mitosis and meiosis depends on an elaborate assemblage of protein machinery, which ensures that sister chromatids or homologous chromosomes are paired (depending on the type of division) and then pulled to opposite spindle pole bodies/centrosomes by microtubules that have been correctly attached to them. Attachment of microtubules to chromosomes is mediated by kinetochores, which are large multiprotein complexes that are fabricated at centromeric DNA . At the core of this structure is the highly conserved KMN network composed of KNL1, the Mis12 complex (Mis12, Dsn1, Nnf1 and Nsl1) and the Ndc80 complex (Ndc80, Nuf2, Spc24 and Spc25), with the latter being responsible for directly interacting with microtubules. The KMN network is anchored to the underlying centromeric DNA via two distinct mutually exclusive and in some cases redundant interactions with the Spc24-25 domain of the Ndc80 complex [2–4] (Figure 1A). The first of these is routed through the Mis12 complex, which connects to the centromere protein CENP-C that in turn connects to histone H3 or the histone H3 variant CENP-A embedded within nucleosomes at the centromere . The latter is mediated via CENP-T, which forms a complex with CENP-W, CENP-S and CENP-X, and, as discussed below, may also interact directly with centromeric DNA [3,4,6]. In addition there are a further 11 proteins (CENP-H, CENP-I, CENP-K, CENP-L, CENP-M, CENP-N, CENP-O, CENP-P, CENP-Q, CENP-U and CENP-R) that were found to associate with CENP-A chromatin in human cells and contribute to proper centromere and kinetochore function . Together with CENP-A, CENP-C, CENP-S, CENP-T, CENP-W and CENP-X, these proteins form the so-called CCAN (constitutive centromere-associated network). Importantly, the majority of the components that make up the CCAN appear to be conserved in yeast and humans .
The cellular roles of CENP-S and CENP-X
CENP-T, CENP-W, CENP-S and CENP-X
CENP-T, CENP-W, CENP-S and CENP-X each contain histone-fold domains, and recent structural studies have revealed how they mediate dimerization between CENP-T and CENP-W and CENP-S and CENP-X respectively [6,8]. Moreover, in chicken and humans, these two different heterodimers associate via regions within CENP-T and CENP-S to form a heterotetramer that is similar in overall structure to the heterotetramer formed by histone H3 and H4 . The histone H3–H4 heterotetramer can form a tetrasome with 80 bp of DNA wrapped around it, and as CENP-T-W-S-X can introduce supercoils into DNA and protect a 100 bp region from micrococcal nuclease digestion in vitro, it would seem that it too can form a tetrasome . This ability, if replicated in vivo, would enable CENP-T-W-S-X to provide a second point of anchorage for the outer kinetochore additional to CENP-A-containing nucleosomes (Figure 1A). Attachment is mediated by the N-terminal domain of CENP-T, which extends as a long tail from its histone-fold and contacts the Spc24-25 domain of the Ndc80 complex via a conserved peptide motif [2–4]. This interaction is regulated by cyclin-dependent-kinase-mediated phosphorylation of CENP-T, which enhances it in late G2-phase, ensuring that the Ndc80 complex is recruited to kinetochores in time for mitosis [2,3,9]. The important role that CENP-T plays in forging a functional kinetochore is highlighted by its ability when fused to the Tet/Lac repressor to single-handedly recruit Ndc80 to tetO/lacO arrays and, in so doing, promote chromosome segregation [4,10,11].
Under normal conditions CENP-T depends on its heterodimerization with CENP-W to properly associate with kinetochores; however, surprisingly, at least in chicken and budding yeast, CENP-S is not required [4,6,7]. In contrast, the localization of CENP-S and CENP-X to centromeres depends wholly on the interaction between CENP-T and CENP-S . Although CENP-T–CENP-W may be able to localize to centromeres without CENP-S–CENP-X, the importance of the heterotetramer is evident from studies in human and chicken cells showing that CENP-S deficiency causes reduced localization of Ndc80 to the outer kinetochore and an increase in its distance from CENP-T, as well as defects in mitotic progression and chromosome segregation . Moreover, heterotetramerization of these proteins appears to be widely conserved as their putative homologues in fission yeast co-purify from an Escherichia coli expression system and gel filter as a single complex with an estimated molecular mass consistent with a heterotetrameric conformation (C. Bryer, R. Barton and M. Whitby, unpublished work).
CENP-S and CENP-X, also known as MHF1 and MHF2
CENP-S and CENP-X, although originally identified as centromeric proteins, were also found to co-purify from human cells with the FANCM (Fanconi’s anaemia complementation group M) DNA translocase resulting in them being assigned the alternative names of MHF1 (FANCM-associated histone-fold protein 1) and MHF2 respectively [13,14]. However, for simplicity, we will continue to refer to them as CENP-S and CENP-X. FANCM is a component of a DNA-repair network in which defects can result in the rare genetic disease FA (Fanconi’s anaemia) that is characterized by progressive bone marrow failure, various developmental abnormalities, increased cancer incidence and cellular hypersensitivity to DNA interstrand cross-linking agents such as cisplatin and mitomycin-C. A key role for FANCM in this network is to recruit the so-called FA core complex (composed of FANCA, FANCB, FANCC, FANCE, FANCF, FANCG, FANCL, FAAP24 and FAAP100) to stalled replication forks, which facilitates the monoubiquitination of FANCD2 and FANCI that then co-ordinate downstream repair processes [15–19].
FANCM also has roles that are independent of the FA core complex and an appreciation of what these are (or might be) has been gained in part from studies of its yeast orthologues (Fml1 in fission yeast and Mph1 in budding yeast), which operate in an environment from where most other FA proteins are absent . In particular two prominent roles in HR (homologous recombination) have been proposed on the basis of the combined findings of biochemical and genetic experiments: (i) promoting recombination-dependent replication restart (Figure 1B); and (ii) promoting CO (crossover) avoidance during DNA DSB (double-strand break) repair (Figure 1C). The first of these roles is thought to be achieved by an ability to catalyse the reversal of stalled replication forks, which generates a substrate from which HR can be initiated  (Figure 1B). The key data that support this model include: (i) the ability of FANCM/Fml1/Mph1 to catalyse the ATPase-dependent reversal of model replication forks in vitro [22–25]; (ii) the reduction in HR induced by replication fork blockage at the RTS1 barrier in fission yeast when fml1 is deleted or mutated to confer ATPase deficiency [23,24]; and (iii) the requirement for FANCM to promote both the stabilization and restart of stalled replication forks in mammalian cells .
FANCM's second role in HR has mainly been inferred from studies of its yeast and plant orthologues, which reveal a prominent role in CO avoidance during DSB repair in mitotic and/or meiotic cells [23,24,27–29]. This role is probably conserved in vertebrates as both mammalian and chicken DT40 cells deficient in FANCM exhibit increased spontaneous and interstrand cross-link-induced sister chromatid exchange [30–32]. COs are a class of recombinant DNA that classically arise as a consequence of the endonucleolytic resolution of HJ (Holliday junction) intermediates resulting in the reciprocal exchange of DNA between the recombining molecules . During meiosis COs promote the formation of chiasmata that are needed to guide correct chromosome segregation during the first meiotic division ; however, in mitotic cells, they can result in deleterious genome rearrangements if recombination occurs between ectopic homologous sequences . Both Fml1 and Mph1 have been shown to catalyse the dissociation of D (displacement)-loops in vitro [24,28,29]. In vivo, D-loops are formed by strand invasion catalysed by the central recombinase Rad51 and are a precursor of HJs. Through their ability to dissociate D-loops, Fml1 and Mph1 offer an alternative to HJ resolution (and therefore CO formation) by driving repair down a pathway known as SDSA (synthesis-dependent strand annealing), which generates only NCO (non-crossover) recombinants (Figure 1C) [24,28,29].
The interaction between FANCM and CENP-S–CENP-X is important for both the promotion of FANCD2 monoubiquitination and the suppression of spontaneous sister chromatid exchange [13,14]. Similarly, in fission yeast, CENP-S and CENP-X work together with Fml1 to promote HR at blocked replication forks and NCO formation during DSB repair in both mitotic and meiotic cells [14,28]. Biochemical and structural studies have provided some insight into the mechanism by which CENP-S–CENP-X contributes to FANCM/Fml1’s various roles. First, in the absence of CENP-T–CENP-W, two heterodimers of CENP-S–CENP-X associate as a tetramer, which can then interact with a region just on the C-terminal side of FANCM's helicase domain [14,36]. An X-ray crystal structure of this C-terminal domain binding to the CENP-S–CENP-X tetramer shows that the interaction, although not causing an allosteric change, does alter the electrostatic surface potential of the CENP-S–CENP-X tetramer enlarging a region of electropositive charge that probably interacts with DNA . Indeed, CENP-S–CENP-X's ability to bind double-stranded DNA is synergistically enhanced upon interaction with FANCM's C-terminal domain [14,36]. This property may explain how CENP-S–CENP-X is able to stimulate the fork reversal activity of FANCM in vitro [13,14], and presumably also aids substrate targeting in vivo. Certainly mutations in Fml1 that disrupt its interaction with CENP-S–CENP-X result in a reduction in CENP-S recruitment to non-centromeric chromatin, as well as deficient DNA repair, replication fork block-induced recombination and CO avoidance .
Is FANCM/Fml1 needed at centromeres beyond its normal roles in DNA repair?
It is clear that CENP-S–CENP-X is crucial for FANCM/Fml1’s ability to promote DNA repair and recombination, but is the reverse true, i.e. does FANCM/Fml1 support CENP-S–CENP-X in promoting centromeric function? A recent hypothesis postulates the intriguing idea that recombinases may play a role in establishing an active centromeric structure by catalysing HR between the repeated DNA elements that are a feature of so-called regional centromeres found in many eukaryotes including fission yeast and humans . Inter-repeat recombination has the potential of forming a covalently closed circle of DNA maintained via a HJ in its ‘stem–loop’. Such circles could help to direct the assembly of CENP-A-containing nucleosomes, and in this regard it is interesting to note that CENP-A's chaperone in mammals, HJURP, is an HJ-binding protein [38,39]. It was also speculated that tRNA genes, embedded within fission yeast centromeres, could be the initiating sites for HR in this organism based on their known ability to stall replication forks . Fml1 is a prime candidate for driving this recombination through its ability to catalyse fork reversal [23,24]. However, so far, we have found little evidence that loss of Fml1 results in centromere dysfunction . In contrast, deletion of the genes encoding CENP-S and CENP-X in fission yeast results in a marked reduction in cell viability, which is associated with a high frequency of defective chromosome segregation .
The importance of CO control at centromeres
Although HR may play a role in helping to establish a fully functional centromere, there is also evidence indicating that its use at these sites is carefully governed and if possible avoided. This is especially true in regional centromeres due to the presence of repeated DNA elements and the dangers associated with recombining them (see above). For example, inter-chromatid recombination between inverted repeat sequences found at fission yeast centromeres can result in isochromosome formation [40,41], and intra-chromatid recombination within the large arrays of directly orientated α-satellite repeats that make up human centromeres can result in deletions . Such deletions could accrue over time and may account for the age-related centromere ‘loss’ observed in women . Moreover, it has been suggested that centromere shortening might compromise kinetochore assembly/function resulting in higher rates of aneuploidy and/or a failure of cellular proliferation due to persistent activation of the mitotic spindle checkpoint .
The inherent risk of using HR at regional centromeres possibly explains why there are mechanisms in place to avoid the need for its deployment. Included in this is machinery that establishes heterochromatin within centromeric DNA, which can block DSB formation during meiosis and aid replication fork progression by displacing RNA polymerase II from DNA [43–45], as well as factors that protect stalled replication forks from collapsing and therefore requiring HR to promote their restart [46,47]. However, once HR is engaged, as a consequence of replication fork collapse or DSB formation, it appears that, at least in maize, crossing over is suppressed . The ability of Fml1, and presumably FANCM, to dissociate D-loops makes them prime candidates for directing NCO recombination at centromeres. Indeed, it is conceivable that the association between FANCM/Fml1 and CENP-S–CENP-X evolved from a need to concentrate its activity at these chromosomal sites. Consistent with this view it has been shown that FANCM localizes to centromeres in a CENP-S-dependent manner in human cells .
In the present review we have summarized data showing that CENP-S and CENP-X have a dual life in the cell, working both at the centromere to aid kinetochore function and more widely across the genome to promote DNA repair and recombination. Intriguingly, they are not the only CCAN proteins that moonlight in another aspect of chromosome biology. In Xenopus, mouse and human cells, CENP-A, together with at least three other CCAN proteins (CENP-N, CENP-T and CENP-U), is recruited to sites of non-centromeric DNA damage including DSBs, where they are found alongside key DNA-damage-response proteins, including RAD51, 53BP1 (p53-binding protein 1), NBS1 (Nijmegen breakage syndrome 1), ATM (ataxia telangiectasia mutated) and CHK2 (checkpoint kinase 2) . Moreover, the loss of CENP-A from these sites was found to coincide with the completion of DNA repair. Interestingly, the recruitment of CENP-A to DSBs depends on its centromere-targeting domain, suggesting that it assembles at these sites in the same way as it does at centromeres . In this regard it is also interesting to note that its chaperone HJURP is another protein that appears to localize to DSBs and plays a role in HR . Exactly what CENP-A and its fellow CCAN proteins are doing at sites of DNA damage and whether there is any involvement with FANCM and CENP-S–CENP-X is unknown. Answering these questions is an important goal for future research.
The 7th International Fission Yeast Meeting: Pombe 2013: An Independent Meeting/EMBO Conference held at University College London, London, U.K., 24–29 June 2013. Organized and Edited by Jürg Bähler (University College London, U.K.) and Jacqueline Hayles (Cancer Research UK London Research Institute, U.K.).
constitutive centromere-associated network
Fanconi’s anaemia complementation group M
FANCM-associated histone-fold protein
Work in our laboratory is funded by the Wellcome Trust [grant number 090767/Z/09/Z].