Báo cáo Y học: Beyond the ABCs of CKC and SCC Do centromeres orchestrate sister chromatid cohesion or vice versa? potx

Centromere-proximal regions usually correspond to sites of avid and persistent sister chromatid cohesion mediated by the conserved cohesin complex.. Keywords: centromere; kinetochore; CE

Beyond the ABCs of CKC and SCC

Do centromeres orchestrate sister chromatid cohesion or vice versa?

Pamela B Meluh1and Alexander V Strunnikov2

1

Memorial Sloan-Kettering Cancer Center, Laboratory of Mechanism and Regulation of Mitosis, New York, USA;2Unit of Chromosome Structure and Function, NIH, NICHD, Laboratory of Gene Regulation and Development, Bethesda, MD, USA

The centromere–kinetochore complex is a highly specialized

chromatin domain that both mediates and monitors

chromosome–spindle interactions responsible for accurate

partitioning ofsister chromatids to daughter cells

Cen-tromeres are distinguished from adjacent chromatin by

specific patterns ofhistone modification and the presence of

a centromere-specific histone H3 variant (e.g CENP-A)

Centromere-proximal regions usually correspond to sites of

avid and persistent sister chromatid cohesion mediated by

the conserved cohesin complex In budding yeast, there is a

substantial body ofevidence indicating centromeres direct

formation and/or stabilization of centromere-proximal

cohesion In other organisms, the dependency ofcohesion on

centromere function is not as clear Indeed, it appears that pericentromeric heterochromatin recruits cohesion proteins independent ofcentromere function Nonetheless, aspects of centromere function are impaired in the absence of sister chromatid cohesion, suggesting the two are interdependent Here we review the nature ofcentromeric chromatin, the dynamics and regulation ofsister chromatid cohesion, and the relationship between the two

Keywords: centromere; kinetochore; CENP-A; histone; methylation; heterochromatin; sister chromatid cohesion; cohesin; chromatin immunoprecipitation

I N T R O D U C T I O N

Chromosomes are duplicated during S phase in a process

that entails not only DNA replication, but also replication of

the chromatin itself Thus, the distribution and modification

state ofnucleosomes, as well as other DNA-associated

proteins that organize the genome and specify patterns of

gene expression must be maintained The process ofhigh

fidelity DNA replication is well understood at this point [1]

Much less is known about the propagation ofchromatin

structure and organization Presumably, this is

accom-plished to a large degree by chromatin assembly factors that

deposit nucleosomes concomitant with DNA replication, as

well as by ÔinstructionsÕ encoded in the DNA sequence (e.g

sequence-specific protein binding sites, intrinsic bends, etc.)

However, there are many epigenetic phenomena that cannot

be explained in this way Thus, various mechanisms for the

self-propagation of pre-existing chromatin states have

been proposed [2,3] We imagine that as for the DNA, replication ofchromatin must also be a faithful process One specialized chromatin domain whose faithful duplication is paramount to accurate chromosome segre-gation is the centromere–kinetochore complex (hereafter, centromere or CKC) The CKC plays both a mechanical and a regulatory role during mitosis Centromeres of paired sister chromatids capture dynamic microtubules (MTs) within the mitotic spindle and exert force upon them As mitosis proceeds, sister centromeres ultimately interact most stably with MTs from opposite spindle poles [4] Such bipolar attachment ensures that each daughter cell will receive a precise complement ofchromosomes The fidelity ofchromosome transmission is enhanced not only by the geometry ofstable centromere–MT interac-tions, but also by the action ofa centromere-based regulatory system called the mitotic checkpoint that monitors CKC–MT interactions and delays the onset of anaphase until stable bipolar attachment is achieved (reviewed in [5])

Genomic integrity is further enhanced by the cell cycle dependent deposition ofprotein complexes that mediate association, precise alignment, and efficient packaging of sister chromatids following replication Chief among these factors are the evolutionarily conserved cohesin and condensin complexes (Table 1) [6] These complexes con-tribute to the structural maintenance ofchromosomes and accurate chromosome transmission during meiosis and mitosis Not surprisingly, both complexes are essential and contain SMC protein pairs (structural maintenance ofchromosomes [7]) in addition to unique components that presumably confer functional specificity

Numerous observations suggest an intimate relationship exists between the formation and function of the CKC and

Correspondence to P B Meluh, Memorial Sloan-Kettering Cancer

Center, Program in Molecular Biology, Laboratory ofMechanism

and Regulation ofMitosis, 1275 York Ave.,

New York, NY 10021, USA.

Fax: + 1 646 422 2062, Tel.: + 1 212 639 7679,

E-mail: p-meluh@ski.mskcc.org

Abbreviations: CKC, centromere–kinetochore complex; SCC, sister

chromatid cohesion; SMC, structural maintenance ofchromosomes;

CAR, cohesin-associated region; MT, microtubule; ChIP, chromatin

immunoprecipitation.

Dedication: This Minireview Series is dedicated to Dr Alan Wolffe,

deceased 26 May 2001.

(Received 28 January 2002, revised 11 March 2002,

accepted 18 March 2002)

Table 1 Quick guide to key cohesion, centromeric and cell cycle proteins described in the text.

Protein in:

S cerevisiae; S pombe;

Condensin complex

Smc2; Cut14; HCAP-E ATPase, coiled-coil domain DNA binding activity

Smc4; Cut3; HCAP-C ATPase, coiled-coil domain DNA binding activity

Ycs4; Cnd1; HCAP-D2 HEAT repeats (BIR repeats)

Ycg1/Ycs5; Cnd3; HCAP-G HEAT repeats

Cohesin complex

Scc1/Mcd1; Rad21;

hRAD21/hSCC1

Cleaved by Esp1 at anaphase onset; C-terminal fragment normally degraded by Ub-dependent proteolysis via N-end rule

selectively retained at paired sister centromeres until Meiosis II anaphase, when cleaved, presumably by Esp1; proper localization requires Spo13

Other factors that promote SCC

on cohesin; Pds5 S.p interacts with cohesin complex

Ctf7/Eco1; Eso1; ?? C 2 H 2 Zn finger-like

domain, Gcn5-related N-acetyl-transferase super family

Interacts with PCNA; maturation ofcohesion; Ctf7/Eco1 possesses in vitro lysine acetyltransferase activity toward itself, Scc1/Mcd1, Scc3, and Pds5

domain; Eso1 also has a polymerase domain (Pol eta) Trf4 & Trf5; (Cid12?);

(POLS?)

DNA polymerase sigma Cohesion establishment during S phase; Trf4 associates

physically with both Smc1 and Smc2

Mad2; Mad2; MAD2 Mitotic checkpoint protein Accumulates on unattached CKC’s during prophase and

prometaphase (as do other mitotic checkpoint proteins); Mad2 (or Bub1 and BubR1) can physically associate with APC Cdc20 to inhibit its activity

CAR (Cohesin

Associated Region)

Often intergenic A+T-rich region DNA sequence bound by cohesin (in some cases, can promote

cohesion when introduced at an ectopic site); centromeric DNA has portable CAR activity; may or may not correspond

to a site of cohesion Factors that can disrupt SCC

Esp1; Cut1; Separase CD-clan caspase-like protease Cleaves Scc1/Mcd1 to promote sister chromatid separation at

the metaphase-to-anaphase transition; regulated by Securin Pds1; Cut2; Securin/PTTG D-box (APC substrate) Chaperone and inhibitor ofEsp1/Separase; degraded via

APC-directed Ub-dependent proteolysis during mitosis

required for clearance of Scc1/Mcd1 cleavage products APC/C or ‘‘Anaphase

Promoting Complex’’

RING-H2 domain subunit, cullin subunit

Multi-subunit ubiquitin-ligase (E3) for mitotic progression and cyclin B destruction; Cdc20 is the specificity factor for Pds1;

Cdc5; Plo1; Polo Polo Kinase (S/T kinase) Among other things, may phosphorylate Scc1/Mcd1 to enhance

its Esp1-dependent cleavage

enhance its separase-dependent cleavage; high Cdc2 kinase activity leads to inhibitory phosphorylation ofseparase

the establishment and maintenance ofsister chromatid

cohesion (SCC) Cytologically, centromeres ofmitotic and

meiotic chromosomes appear as sites ofavid cohesion, and

may, in fact, direct the formation and maintenance of a

centromere-proximal domain ofcohesin and other cohesion

promoting factors [8–12] Consistent with this, in Drosophila

and vertebrate cells, cohesin subunits at centromeres are

specifically retained until anaphase, whereas the vast

majority ofcohesin dissociates from chromosome arms

during prophase [13–16] Similarly, during meiosis I,

meiosis-specific cohesin complexes persist at centromeres,

even while arm cohesion is dissolved to facilitate dissolution

ofthe chiasmata [17,18] Retention ofSCC at centromeres

during meiosis I is, ofcourse, critical for the reductional

pattern ofmeiosis I chromosome segregation However,

SCC is also essential for accurate equational chromosome

segregation during mitosis (and meiosis II), and when compromised, leads to chromosome nondisjunction and loss [17,19–24]

SCC at or near the centromere clearly opposes the MT-dependent pulling forces exerted by the spindle prior to the onset ofanaphase, and therefore helps to prevent premature dissociation ofsister chromatids [25] Perhaps more importantly, centromere-proximal cohesion (and in the case oflarger centromeres, perhaps condensin-mediated packaging) apparently serves to sterically constrain sister CKCs in a Ôback-to-backÕ orientation, thereby precluding merotelic or monopolar attachments in favor of bipolar attachment ofthe paired sisters to the spindle [25–27] Finally, the metaphase-to-anaphase transition is triggered

by the timely degradation ofa particular cohesin subunit known as Scc1/Mcd1 in S cerevisiae, and Rad21 in other

Table 1 continued.

Protein in:

S cerevisiae; S pombe;

Heterochromatin proteins

Lysine 9; necessary for proper chromosome segregation; required in S pombe for centromeric (but not arm) cohesion and recruitment ofcohesin to silent chromatin (including centromeres)

??; Clr4; SUV39H1 Chromo domain, SET domain Heterochromatin protein; histone H3 Lysine 9 methyltransferase Centromere proteins

Cse4; Cnp1; CENP-A Histone H3 fold domain Essential CKC determinant; may replace histone H3 in

centromere specific nucleosomes; uniquely marks centromeric chromatin; required for proper localization of CENP-C and INCENP to CKC; might directly or indirectly recruit cohesin

localization depends on CENP-A

localization to CKC in mid-mitosis is dependent on cohesin Ipl1; Ark1; Aurora B/AIM-1 Aurora kinase Chromosomal passenger protein; interacts with INCENP;

localization to CKC in mid-mitosis is dependent on cohesin; substrates include histone H3, CENP-A, and Ndc10 (S cerevisiae centromere protein)

Bir1; Bir1/Cut17; Bir1/Survivin IAP-related Chromosomal passenger protein; interacts with INCENP and

Aurora B kinase; localization to CKC in mid-mitosis is dependent on cohesin

Ndc80/Hec1; Ndc10; HEC All subunits have coiled-coil

domains

Component ofconserved CKC protein complex (includes Ndc80/Hec1, Nuf2, Spc25, Spc24); in S cerevisiae, Ndc80/Hec1 physically and genetically interacts with cohesin subunit Smc1; Nuf2 S.c and Spc25 S.c also interact with Smc1

for cleavage and localization to the spindle midzone in anaphase; null mutants show high frequency of premature sister chromatid separation in meiosis I

sister chromatid separation) in meiosis I

meiosis I; mutants fail to localize Rec8 properly and show equational division (i.e premature sister chromatid separation) in meiosis I

??; ??; MEI-S322 Coiled-coil domain Drosophila protein present at CKCs ofboth meiotic and mitotic

chromosomes; dissociates at anaphase; required for main tenance ofSCC at the CKC in mitosis and especially in meiosis

species In so far as the CKC controls this transition

through activation ofthe mitotic checkpoint [5], it does so

by indirectly inhibiting the degradation ofthe Scc1/Mcd1

cohesin subunit (reviewed in [28–30]) Below we review in

further detail the nature of centromeric chromatin and SCC

and the relationship between them

T H E N A T U R E O F C E N T R O M E R I C

C H R O M A T I N

Centromeres are defined genetically by phenotypic or

molecular markers [31] that always segregate away from

one another during meiosis I Centromeres can be

visua-lized cytologically as the primary constriction ofmetaphase

chromosomes in vertebrate cells, and correspond to the site

ofkinetochore formation Centromeric chromatin

struc-ture and composition have been studied primarily

in human and rodent cells, Drosophila, Caenorhabditis elegans, fission yeast, and budding yeast, using a combi-nation ofgenetics, cell biology and more recently, bioin-formatics With the apparent exception of budding yeast and possibly C elegans, natural centromeres occur within the context ofconstitutive heterochromatin As such, these centromeric regions are largely nontranscribed, exert position effects on gene expression, and show reduced recombination They typically contain extended arrays of repetitive, often A+T-rich, DNA sequence elements (e.g alpha satellite in human cells) [31–34], and by indirect immunofluorescence and/or chromatin immunoprecipita-tion, are organized by nucleosomes that are hypo-acetyl-ated and hyper-methylhypo-acetyl-ated (i.e on lysine 9 ofhistone H3) (Fig 1) [35–38] The latter features account for the

Fig 1 Cross talk between centromere components and cohesion proteins Diagram summarizes potential relationships compiled from findings from several species The centromeric chromatin is characterized hypo-acetylated, hyper-methylated (ƠMeÕ) histone H3-containing nucleosomes as well as the CENP-A-containing variety (ƠÃ) The degree ofnucleosome interspersion is unresolved and may be species-specific Histone H3 methylation by ƠSETÕ domain-containing methyltransferases [e.g Clr4 S.p , Su(var)3–9 D.m ; SUV39H1] is critical for recruitment of chromo domain proteins (e.g HP1 or Swi6 S.p .) and cohesin to heterochromatic regions Recruitment ofother centromere proteins (e.g CENP-C (ƠCÕ); INCENP; and possibly the Ndc80/Hec1 complex) is dependent on CENP-A (ƠÃ) Under certain conditions CENP-A can be shown to recruit cohesin, and Ndc80/Hec1 genetically and physically interacts with cohesin Mitosis-specific centromere proteins such as the mitotic checkpoint protein Mad2 indirectly promote SCC by inhibiting the activity ofAPC towards Pds1/Securin Conversely, cohesin is important for the recruitment ofsome centromeric proteins, namely the chromosomal passenger proteins aurora B/Ipl1, INCENP, and Bir1 Tension exerted across the CKC leads to disruption of SCC (at least in budding yeast) and release ofpassenger proteins Aurora B/Ipl1 kinase, which can phosphorylate (ƠPÕ) CENP-A, as well as histone H3, has been implicated in detecting tension at the CKC and may therefore play an active role in the dissolution of SCC, perhaps by acting on CENP-A The mechanism(s) whereby CENP-A distribution and HMTase activity are directed to centromere-proximal regions is currently unknown but clearly ofgreat interest.

presence ofchromo domain-containing heterochromatin

proteins (which bind lysine 9-methylated histone H3) at

centromeres in Schizosaccharomyces pombe (Swi6 and

Clr4), Drosophila and mammalian cells [HP1 and Su

(var)3–9] Recently, these chromo domain proteins have

been implicated in the recruitment and/or stabilization of

centromere-proximal cohesin [25,39]

The pattern ofcore histone tail modification within

centromeric chromatin is a critical determinant ofCKC

structure and function Treatment of S pombe or human

cells with trichostatin A, a histone deacetylase inhibitor,

leads to increased histone acetylation throughout the

genome Within centromeric chromatin, increased

acetyla-tion correlates with profound effects on centromere

struc-ture and function, including decreased centromeric gene

silencing, loss ofassociated HP1 proteins, and pronounced

chromosome segregation defects [36,40] Similarly,

muta-tions in S pombe and Drosophila that affect centromeric

gene silencing or position effect also affect centromere

structure and function [25,41–43] In several cases, the genes

defined by these mutations encode histone-modifying

enzymes such as deacetylases (Clr3) or methylases (Clr4),

strongly supporting the idea that centromeric chromatin

must ÔsportÕ a particular histone code

Centromeric chromatin is distinct from surrounding

chromatin, not only with respect to its histone modification

pattern as discussed above, but also at an even more

fundamental level of chromosome organization: namely,

that ofthe nucleosome itself While the sequence and size

ofunderlying centromeric DNA varies greatly among

organisms [2,34], centromeric chromatin in all eukaryotes

studied to date, including budding yeast and C elegans,

contains a unique and essential histone H3 variant

(reviewed in [44]) The founding member of this group,

CENP-A, was originally identified as a prominent

auto-antigen in human CREST serum [45] CENP-A is a

constitutive centromere component, and localizes to the

inner kinetochore plate ofmitotic chromosomes [46] In

mammalian cells, CENP-A is incorporated into

nucleo-some-like particles along with histones H2A, H2B, and H4

[47,48] and in vitro, CENP-A can replace histone H3 within

reconstituted nucleosomes [49] Importantly, alpha satellite

DNA, the major DNA repetitive element present at human

centromeres, copurifies with CENP-A-containing

nucleo-somes [50] Given that CENP-A is found only at active

centromeres [46,51], it has been proposed that centromeric

chromatin is uniquely marked by centromere-specific

nucleosomes in which CENP-A replaces histone H3

Whether such specialized nucleosomes are present at

centromeres in other organisms remains to be determined,

but is likely to be the case For example, in

Saccharo-myces cerevisiaeor S pombe alterations in histone dosage

can affect centromere chromatin structure and impair

chromosome segregation [52–55], as do certain point

mutations in histones H2A and H4 [56,57] While such

phenotypes could reflect the indirect effect of altered gene

expression, allele-specific genetic interactions between

his-tone H4 and the yeast CENP-A homolog, Cse4, suggest

these two proteins physically associate, possibly in the

context ofa nucleosome-like particle [57,58] Regardless, it

is clear from genetic studies that CENP-A is required for

the assembly ofa functional kinetochore, and in its

absence other essential centromere components, such as

CENP-C, are no longer recruited to the CKC [59–64] That said, CENP-A is not sufficient for centromere specification When a functional Cse4–Gal4 DNA binding domain fusion is directed to an ectopic locus, neocentro-mere activity is not detected [65] (P Meluh, unpublished results) Similarly, when overexpressed, CENP-A is misdi-rected to noncentromeric chromatin, where it recruits and/

or stabilizes some (e.g CENP-C and hSmc1), but not all centromere factors [147]

S I S T E R C H R O M A T I D C O H E S I O N

Sister chromatid cohesion (SCC) refers to the physical, and

as we now know, protein-mediated linkage that exists between replicated sister chromatids from the onset of DNA replication until anaphase Thus, SCC exists throughout a significant portion ofboth the mitotic and meiotic cell cycles Moreover, SCC persists, at least at or near centromeres, even when M phase is prolonged by drug treatment or checkpoint activation Establishment, main-tenance and timely resolution ofSCC are essential steps in ensuring proper chromosome segregation and, accordingly, the genetic stability ofa eukaryotic cell Mutations that impair any ofthese steps would be expected to cause gross chromosome missegregation, cell growth arrest and/or inviability Uncovering such relevant phenotypes as prema-ture sister chromatid separation in prometaphase or failure

to separate sisters in anaphase has led to the genetic identification ofa number ofSCC structural and regulatory components (Table 1) [19–22,24,66–68] These genetic find-ings have been strikingly consistent with biochemical approaches aimed at defining the molecular nature of SCC (reviewed in [6,29,30])

Thus, in a few short years, we have gone from regarding SCC as a Ôcytological formalismÕ to a clear appreciation ofSCC as a complex cellular process that involves specific cis-acting chromosomal loci (e.g the centromere), dozens of protein factors, and that is intimately associated with the cell cycle machinery The precise molecular mechanism whereby sister chromatids are held together remains elusive, as biochemical activities and dependency relationships for assembly have been assigned to only a few SCC proteins

On the other hand, our knowledge about SCC regulation and the chromosomal localization ofSCC activity has accumulated at a fast-pace Chromosomal ÔaddressesÕ for SCC activity can be divided into two major classes: centromeric cohesion and chromosomal arm cohesion As mentioned above, location is not the only distinction between these two classes Centromeric and arm cohesion can be differentially regulated, such that centromere-prox-imal cohesion is more stable and thus, presumably, the more relevant target ofcell cycle regulation [69] The molecular basis for the distinction is the subject of great interest

P R O T E I N S I N V O L V E D I N

E S T A B L I S H M E N T O F S I S T E R

C H R O M A T I D C O H E S I O N

What mediates SCC? The landmark study by Holloway

et al [70] established that at least one noncyclin protein must be degraded via ubiquitin-dependent proteolysis to allow for sister chromatid separation during mitosis This observation, combined with studies that ruled out DNA

topological constraints as the sole mediator ofSCC [20,71],

led to the proposal that SCC must be protein mediated This

view has since been validated by genetic and biochemical

studies that have identified chromosome-associated proteins

involved in SCC (reviewed in [6,29,30])

Perhaps the best characterized ofthese SCC factors is

the evolutionarily conserved cohesin complex, components

ofwhich have been genetically identified in several model

organisms Cohesin was first biochemically identified as a

multicomponent complex in Xenopus embryonic extracts

[13], and similar complexes have since been purified from

yeast [72,73] and other vertebrate species [14,74,75] In all

cases, the cohesin complex consists offour types of

subunit, some ofwhich have tissue- or developmental

stage-specific paralogs (Table 1 [17,18,74,76,77]) The

cohesin core consists ofa heterodimeric pair ofSMC

(structural maintenance ofchromosomes) proteins, Smc1

and Smc3 Like the related Smc2–Smc4 heterodimer found

in the condensin complex [78], the Smc1–Smc3

heterod-imer forms an extended coiled-coil with two catalytic

domains possessing DNA-binding and ATPase activities

[6,79] Smc1 and Smc3 are likely to interact through their

hinge regions [80,81] to form a clamp around the chromatin fiber [82] In budding yeast, Smc1 and Smc3 are constitutively bound to chromatin throughout the cell cycle and presumably serve as platforms for assembly of mature cohesin early in S phase At that time, two additional subunits, Scc3 (i.e SA1/STAG in mammals) and Scc1/Mcd1 (called Rad21 in other organisms), are recruited to chromatin in an Smc1- and Smc3-dependent fashion [10,20] It is possible that assembly of tetrameric cohesin also occurs in a soluble phase during S phase to compensate for chromatin replication Interestingly, the

S pombeScc3 homolog, Psc3, does not behave as a stable component ofthe cohesin complex, but this could simply reflect the method ofcell lysate preparation [73] Regard-less, it is believed that within replicated chromosomes only the mature tetrameric form of cohesin is competent to bridge sister chromatids This view is supported by the DNA binding properties ofcohesin in vitro [79], and also

by the fact that cleavage of Scc1/Mcd1 at the metaphase-to-anaphase transition and the accompanying loss ofScc3 from chromatin coincides with, and indeed is a prerequis-ite for, dissolution of SCC [67] (see below)

Fig 2 Regulation of sister chromatid cohesion establishment and release during the budding yeast cell cycle Schematic diagram summarizing factors cited in the text that govern SCC Left panel Several key regulatory steps in SCC formation and resolution are shown in grey boxes Green arrows indicate poleward pulling forces exerted on the centromere–kinetochore complex (CKC) by the mitotic spindle Blue arrows indicate a chromatin recoil force that may allow for transient re-establishment of SCC at the CKC in budding yeast Right panel Consequences of impaired SCC Metaphase arrest, chromosome loss (ÔCutÕ phenotype), and/or nondisjunction can result from SCC misregulation Ac, acetylation; P, phos-phorylation; APC, anaphase promoting complex.

Other factors essential for the establishment and

main-tenance ofSCC have been identified, largely through genetic

screening (Table 1; Fig 2) Although these proteins

colo-calize with cohesin in chromatin and/or genetically interact

with cohesin subunits, they are not stably associated with

soluble cohesin complexes One such factor is the conserved

Pds5/Spo76/BimD protein [22,75,83–86] Mutations in Pds5

homologs confer SCC and chromosome segregation defects

similar to those seen in cohesin mutants, yet Pds5 is not an

integral part ofthe cohesin complex per se For example, in

S cerevisiae, Pds5 is significantly less abundant than

cohesin itself[85], and although Pds5 and cohesin have

been shown to physically interact in S pombe and human

cells, only a subfraction of total cohesin is associated with

Pds5 [75,86] Nonetheless, Pds5 homologs undergo

cell-cycle regulated localization to mitotic and meiotic

chromo-somes in their respective organisms, and at least in budding

and fission yeast, chromatin localization is dependent upon

cohesin function [22,85,86] However, cohesin complex

localization to chromosomes is independent ofPds5

func-tion Thus, Pds5 either acts downstream ofcohesin in an

SCC assembly pathway, or it provides an SCC fidelity or

optimization function In this regard, S pombe strains

lacking Pds5 are viable and establish SCC in S phase

normally However, pds5D mutants are unable to maintain

SCC during prolonged G2-arrest [86] In contrast, in

budding yeast, Sordaria, and A nidulans, Pds5 is essential

for both establishment and maintenance of SCC

[22,84,85,87] These observations suggest that Pds5

pro-motes maturation or stabilization ofthe cohesin-mediated

linkage between sisters, and that different organisms require

Pds5 activity to different extents

Although cohesin subunits can associate with chromatin

thought out the cell cycle, execution point studies indicate

that productive and faithful SCC strictly requires that

mature cohesin assembly and deposition occur in S phase

[21,72,88–90] Several proteins promote this timely

incor-poration ofcohesin into chromatin, and hence, SCC For

example, mutations in the highly conserved Scc2 (Mis4 in

S pombe) and Scc4 proteins, phenocopy cohesin mutants

in that loading ofScc1/Mcd1 and Scc3 onto chromatin in

S phase, and consequently SCC, fails [91,92] However,

unlike cohesin mutants, loss ofScc2 or Scc4 function does

not preclude assembly ofsoluble mature cohesin complexes

[91] Thus, Scc2 and Scc4 might function in early S phase to

chaperone Scc1/Mcd1 and Scc3 to pre-existing

chromatin-bound Smc1–Smc3 heterodimers or to make newly

repli-cated chromatin accessible for cohesin assembly

The sequence and/or the genetic interactions ofother

proteins required for cohesion suggest that productive

cohesin deposition is intimately associated with the act of

replication Budding yeast Ctf7/Eco1 and the related

S pombe protein Eso1 [72,89,90] are required specifically

for establishment of SCC in S phase Cohesin binding to

chromatin per se is independent ofCtf7/Eco1/Eso1, but

SCC linkages are either not formed or break prematurely in

ctf7/eco1/eso1 mutants Ctf7/Eco1 was recently shown to

exhibit lysine acetyltransferase activity in vitro, both towards

itselfand toward cohesin subunits, suggesting that Ctf7/

Eco1 promotes establishment ofSCC via post-translational

modification ofcohesin [93] Ctf7/Eco1 and Eso1 interact

with PCNA both physically and genetically [86,89] and

Eso1 itselfhas a DNA-polymerase-like domain that is

functionally separable from the domain involved in SCC These data suggest a possible mechanistic link between replication fork movement and cohesin deposition onto chromatin The budding yeast Trf4 protein reinforces such a connection Trf4 is itself a DNA polymerase and when inactivated, leads to a profound defect in cohesion estab-lishment [94,95] To date, it has not been possible to genetically separate the polymerase activity ofTrf4 from its role in cohesion While the DNA polymerase domains in proteins required for SCC might be a red herring, Ctf18, an RFC-like protein has also been recently implicated in SCC [89,96] Thus, an attractive hypothesis is that a polymerase switch similar to that which occurs during Okazaki fragment synthesis might take place at cohesion sites to facilitate duplication and assembly of cohesion complexes when sister chromatids are in close proximity [95,96] This model remains to be tested

Obviously, what Ôduplication and assemblyÕ entails must await structural and biochemical characterization ofthe critical proteins However, a detailed understanding ofthe molecular basis ofSCC will also require the development of

a comprehensive in vitro system This will be an ambitious undertaking given that SCC assembly seems tightly coupled

to DNA and chromatin replication Moreover, as described below, it is unclear what would constitute an adequate DNA template on which to reconstitute SCC as our understanding ofthe nature ofcohesion sites in vivo is meager

C I S - A C T I N G S I T E S R E Q U I R E D

F O R S I S T E R C H R O M A T I D C O H E S I O N

Given that cohesin complex deposition occurs coincident with replication, or shortly thereafter, why invoke the existence ofspecific heritable, DNA- or chromatin-encoded cohesin binding sites? This concept may have derived in part from the observation that centromeric regions often appear

as avid and persistent sites ofcohesion In addition, models ofhigher order mitotic chromosome structure envision two closely associated sister chromatid cores with sister DNA loops extending in opposite directions [97,98] Cohesin-mediated Ôspot-weldingÕ ofhomologous sister chromatid domains would be a prerequisite for such models

To date, the most comprehensive analyses ofchromo-somal sites potentially required for SCC have been carried out in budding yeast Two fundamentally different approa-ches have been used One approach has exploited chromatin immunoprecipitation (ChIP) [99] and makes the assumption that sites ofcohesin binding (e.g Scc1/Mcd1 or Smc1) correspond to sites ofcohesion A second approach has been

to functionally map the sequence(s) required for SCC using centromere-based plasmids Like authentic chromosomes, these minichromosomes are replicated in S phase and persist

as paired sister minichromatids until anaphase [8,71] The ChIP studies have revealed that cohesin complexes

do not bind uniformly along yeast chromosomes Rather, cohesin associated regions (CARs) are 300–1000 bp in length and sparsely distributed on the chromosome, occur-ring only every 8–13 kbp on average [9,10,100,101] It is noteworthy that such CARs correspond to only a few nucleosomes’ worth ofDNA There is, however, one striking exception to this rule, namely, an enormous con-centration ofcohesin binding sites occurs over a 10–20 kb domain encompassing each centromere [9,10,73,100] It is

important to note that functional centromeres, which in

yeast are specified by a 125 bp DNA sequence, are both

necessary and sufficient for formation of much larger

cohesin domains [8,10] Moreover, using a recombinase

strategy, Megee et al have shown that the centromere is a

major determinant ofSCC on yeast minichromosomes [8]

Taken together these data suggest that the centromere

somehow directs cohesin recruitment (or stabilization) over

a long distance and that the resulting assemblage is directly

responsible for centromere-dependent SCC The

mechan-ism ofsuch localized recruitment (e.g active spreading,

polymerization, post-translational modifications), ifit

occurs at all, is unknown

There are several observations that seem at odds with this

Ôcentromere-directed SCCÕ model for budding yeast SCC

First, by ChIP, cohesion proteins are not enriched to the

same extent around the centromeres ofminichromosomes

as they are around centromeres within the chromosome [8]

Moreover, localization ofcohesion proteins within cells or

chromosome spreads as determined by microscopic

tech-niques does not agree with the ChIP data By indirect

immunofluorescence or GFP-tagging, cohesion proteins

are broadly distributed on chromosomes, and do not appear

to be enriched at centromeres in yeast [20,73,102,103]

(A V Strunnikov, unpublished results) or, for that matter,

in vertebrate cells prior to prometaphase [16,74,75,104]

Finally, one ofthe most paradoxical observations is that

soon after their replication, pericentric regions of sister

chromatids in yeast appear to Ôsplit apartÕ owing to

MT-dependent spindle forces [10,105–107] In other words,

despite their apparent load ofcohesion proteins,

centro-meric regions may not be closely paired

These discrepancies could in part be explained by the

in vivo and in vitro binding preference of the cohesin

complex for A+T-rich DNA [9,100,101,108] However,

they might also reflect the limitations ofChIP, an assay that

relies on the geometry ofreactive amino groups to promote

formaldehyde-dependent protein–DNA cross-linking [99]

Conceivably, the microscopic localization data accurately

reflect a broad distribution ofcohesion proteins on

chro-mosomes In this case, ChIP must be revealing a functional

or structural difference in cohesion proteins or chromatin as

a whole around the centromere, such that cohesion proteins

are more accessible to cross-linking reagents and/or

anti-bodies This altered state could be related to the weakened

(or absence of) cohesion at the centromeres in mitosis (see

above) or to the stretching ofcentromere chromatin [26,109]

that could alter DNA conformation, making it more

accessible to cross-linking

Another possibility is that we have been misled by the

pervasive concept ofSCC as Ômolecular glueÕ, which

connotes SCC as inert and unchanging Perhaps cohesion

protein binding at the centromere, or for that matter, at all

CARs, is dynamic In this case, the apparent binding

differences revealed by ChIP would reflect localized

differ-ences in cohesin on-off rates (as influenced by DNA

sequence, chromatin structure, Smc1–Smc3 enzymatic

cycle, local kinase activities, etc.) Dynamic cohesin binding

is consistent with several observations: (a) cohesin proteins

rapidly dissociate from the chromosome when a CAR (i.e

the centromere) is excised [8]; (b) over-expression ofcohesin

subunits can expand CAR size as determined by ChIP

(P Megee, Department ofBiochemistry and Molecular

Genetics, University ofColorado Health Sciences Center,

DN, USA, personal communication); (c) CARs are Ôport-ableÕ, but only in a qualitative sense, the range over which

a given CAR promotes cohesin binding varies with its chromosomal (or minichromosomal) location; (d) cryptic cohesion sites on a minichromosome are revealed when its centromere is excised [9,10]; and (e) the transient but persistent separation ofpaired sister centromeres following replication (i.e ÔbreathingÕ) [26,105–107] and that re-associ-ation ofsister centromeres early in mitotic prophase requires cohesin [26] Taking these considerations into account, it is not presently possible to pinpoint a specific sequence within the centromeric region that could rightly be called a Ônucleation siteÕ for cohesin recruitment

In higher eukaryotes, the existence ofnatural cis-acting sites that mediate SCC is largely inferred, based on current models ofchromosome architecture Assuming there are specific cohesion sites, those on chromosome arms and at the centromere are differentially regulated in vertebrate cells Thus, cohesin largely disappears from chromosome arms during prophase, but persists at centromeres until anaphase [13–16] (see below) Another difference appears to be that once established in S phase, SCC on chromosome arms is less static than that at centromeres, as sister sequences can display dynamic association and dissociation behavior [110] The nature ofthis phenomenon remains to be elucidated Although the nature and organization ofSCC sites in higher eukaryotes remains elusive, several repetitive DNA sequences have been shown to promote (albeit misregulated) SCC when integrated into an ectopic chromosomal location [111,112] Most notable among these, apropos this review, is the normally centromeric human alpha-satellite repeat That human centromeric DNA can promote SCC in mammalian cells is consistent with findings from budding yeast with one important distinction Namely, in this case, SCC establishment does not correlate with centromere function because integrated alpha-satellite DNA rarely directs formation of a functional centromere Similarly, the repressed centromeres ofstable dicentric chromosomes often remain heterochromatic and show avid SCC [113] This suggests that it is possible to genetically separate centromeric cohesion and segregation functions and lends support to studies in S pombe indicating cohesin is attracted to heterochromatin and possibly plays an inde-pendent structural role in interphase chromatin [25,39]

R E G U L A T I O N O F S I S T E R C H R O M A T I D

C O H E S I O N

In budding yeast, positive regulation ofSCC establishment

is mediated by strong transcriptional induction of SCC1/ MCD1 and SCC3 expression, accompanied by smaller increases in the mRNA levels for other genes involved in cohesion establishment Indeed, cluster analysis ofgenome-wide expression data from budding yeast revealed that several cohesion protein genes, including SCC1/MCD1, are coregulated by the MBF transcription factor along with many DNA biosynthetic genes [114]

Once established in S phase, how is cohesion maintained?

Is it stable or dynamic? Besides an obvious requirement for continued integrity ofthe cohesion proteins (and perhaps the underlying CAR, as in the case ofcentromere-proximal cohesion), there is little information about the mechanism of

SCC maintenance As noted above, chromatin association

ofat least some cohesion proteins may be dynamic Indeed,

one ChIP study suggested that Scc1/Mcd1 is redistributed

from arm sites to centromeric regions during the transition

from the S phase to mitosis [100] This has not been

rigorously tested, but iftrue, would be reminiscent of

cohesin dynamics in higher eukaryotes, where the bulk of

cohesin and Pds5 dissociates from chromosome arms as

cells enter mitotic prophase [15,16,75] While cohesin may

be concentrated (or selectively retained) at centromeres

when yeast cells enter mitosis, as noted above,

centromere-proximal cohesion per se appears to be lost or weakened

prior to metaphase (Fig 2 [26,105–107]) Iftrue, then the

crucial regulatory step governing loss ofSCC at the

metaphase-to-anaphase transition in yeast must involve

arm cohesion sites, or at least the most centromere-proximal

CARs that do not ÔbreatheÕ ( 30 kb from the centromeric

DNA) [26,106,115]

Work from several labs has revealed that the mechanism

whereby such hypothetical ÔstrongÕ cohesion sites are

dissociated or disassembled entails proteolytic destruction

ofthe cohesin subunit Scc1/Mcd1 (reviewed in [29,30,

116,117]) First, it was shown that at the nonpermissive

temperature, esp1–1S.c..mutants are defective for sister

chro-matid separation and subsequent anaphase [118] Second, it

was noted that a subset ofScc1/Mcd1p undergoes specific

proteolytic cleavage at the onset ofanaphase that is

Esp1-dependent [67] Importantly, a noncleavable form of Scc1/

Mcd1 behaves in a dominant fashion to retard sister

chromatid separation, thus mimicking the esp1–1 phenotype

[67,68] Similar observations were made for S pombe

mutants in the Esp1-related protein Cut1 [73,119] Esp1

was subsequently shown to be a conserved caspase-like

CD-clan cysteine endopeptidase (i.e ÔseparaseÕ) capable of

cleaving Scc1/Mcd1 in vitro [120] In vivo, this cleavage event

is followed by the destruction of Scc1/Mcd1 fragments in a

Ubr1-dependent fashion (i.e by the N-end rule degradation

pathway) [121] Destruction ofScc1/Mcd1 is accompanied

by dissociation ofScc3 (and Pds5) from chromosomes,

which together presumably inactivate the cohesin complex

[85] In this way, the critical connections that hold sisters

together from S phase through early mitosis are literally cut

by Ômolecular scissorsÕ in the form of Esp1/separase at the

onset ofanaphase

The elucidation ofseparase function provided great

insight into the destruction ofSCC, but in itself, neither

explained the timeliness ofthat destruction, nor the

requirement for ubiquitin-mediated proteolysis in sister

chromatid separation (i.e as mediated by anaphase

pro-moting complex/cyclosome or APC/C) [70] Indeed,

bud-ding yeast Esp1 and fission yeast Cut1 are present

throughout the cell cycle and may have additional functions

in spindle morphogenesis and exit from mitosis [104,119,

122–127] Not surprisingly, the activity ofEsp1 towards

Scc1/Mcd1 is coordinated with the cell cycle by multiple

levels ofregulation (Fig 2)

Esp1 exists in a complex with a ÔsecurinÕ protein known as

Pds1 in budding yeast [118] Unrelated proteins that are

functionally equivalent to Pds1 have been identified in many

systems and appear to regulate Esp1 in two ways (i.e

vSecurin in Xenopus, PTTG in human cells, Cut2 in

S pombe [72,128–130]) In each case, sequestration by its

cognate securin inhibits separase’s sister chromatid

separ-ation activity In addition, Pds1 (and perhaps its analogs) serves as a chaperone/activator for Esp1, in part by promo-ting its efficient nuclear targepromo-ting [127] Pds1 contains a B-type cyclin destruction box and accordingly is degraded

in mitosis in an APCCdc20-dependent fashion [131,132] The regulated degradation ofPds1 in mitosis frees Esp1

to act on Scc1/Mcd1, and explains the requirement for ubiquitin-dependent proteolysis in the metaphase-to-ana-phase transition However, recent studies suggest that similar to authentic caspases, Esp1 itselfmight undergo proteolytic cleavage to become catalytically active [16] In addition, Xenopus separase is subject to inhibitory phos-phorylation by Cdc2 [133]

Scc1/Mcd1 cleavage is also regulated at the level of substrate in that phosphorylation ofScc1/Mcd1 enhances its cleavage by Esp1-like proteins both in vitro and in vivo [73,120,134] Phosphorylation has been attributed to the Cdc2 kinase in human cells [74] and to the polo-like protein kinase Cdc5 in yeast [134] Cdc5 is proposed to facilitate cleavage ofScc1/Mcd1 via phosphorylation ofthe endo-peptidase recognition site; however, whether Cdc5 directly phosphorylates Scc1/Mcd1 in vivo is unknown [134] Nonetheless, when Cdc5 is inactivated, separation ofsister chromatid arms is delayed compared to that ofcentromeric regions [134] This observation is consistent with an earlier study showing that separation ofsister centromeric regions

is unaffected by mutations that prevent Esp1-mediated cleavage ofScc1/Mcd1 [26], and provides additional evidence that during an unperturbed mitosis, arm cohesion sites in yeast are ÔstrongerÕ (i.e more important in resisting mitotic spindle forces) than centromeric ones

As mentioned previously, the apparent dichotomy of cohesin sites at arms and centromeres found in yeast is paralleled in Drosophila and mammalian cells In these metazoans, the bulk ofcohesin dissociates from chromo-some arms during the late stages ofmitotic chromochromo-some condensation, hinting that only a minor, virtually invisible, fraction of cohesin remains on chromosomes to maintain SCC through metaphase [14,15,75] Experiments both in vivo and in vitro showed that such dissociation ofcohesin in prometaphase is not accompanied by proteolysis ofhRad21 [16,74] The residual pool ofchromatin-associated cohesin, revealed by overexpressing a tagged form of hRad21, localizes to centromere-proximal regions [16] As in yeast, this fraction of hRad21 is apparently removed from chromatin in an Esp1-dependent fashion, because cleavage oftagged hRad21 was observed [16] Indeed, in subsequent experiments, overexpression ofa noncleavable mutant of hRad21 was shown to block anaphase and cytokinesis [68] Thus, during mitosis in higher eukaryotes, cohesin is removed from sister chromatids by at least two distinct mechanisms such that centromeres, which selectively retain cohesin, appear to be the ÔstrongerÕ sites ofSCC

While yeast and metazoans both regulate centromeric

vs arm SCC differentially, they seem to do so in opposite ways (Fig 2) One wonders ifthis is really the case, given that cohesion proteins and cell cycle machinery are widely conserved The apparent differences might simply reflect the timing with which different organisms assemble their mitotic spindles and establish bipolar chromosome attach-ment We suggest that in all organisms, centromere-proximal cohesion is absolutely critical for establishing stable bipolar attachment, whereas either

centromere-proximal SCC or arm SCC can resist outward pulling

spindle forces once bipolar attachment is achieved In

yeast, the spindle forms in S phase and centromeres (which

have measurable MT binding activity at this point [135])

can, in principle, establish stable bipolar attachment soon

after their replication In contrast, plant and animal

spindles form in M-phase, following nuclear envelope

breakdown, thus maintenance ofcentromere-proximal

SCC well into mitosis is essential for proper mitotic

chromosome segregation

Due to SCC, stable bipolar spindle attachment places

centromeric chromatin under tension Conceivably this

tension is transduced (mechanically or chemically) to

centromeric cohesin, such that interchromatid cohesion is

weakened without dissociation or cleavage ofcohesin This

would explain why in budding yeast, sister centromeres can

ÔbreatheÕ soon after replication, whereas a similar change in

centromeric cohesion would not occur in higher eukaryotes

until bipolar spindle attachment is established during

prometaphase Notably, intersister kinetochore distance

does in fact increase in a MT-dependent fashion at this stage

in vertebrate cells [136] The notion that centromeric

cohesion is by design tension-sensitive seems inconsistent

with observations that cohesin association is favored and

perhaps enhanced at or near centromeres Indeed, as

discussed above, mitotic centromeres in yeast and

metazo-ans are preferential sites of cohesin binding However, in

budding yeast, cohesin binding at centromeres has been

most often examined in nocodazole-arrested cells, but to our

knowledge, such results have never been directly compared

to cells synchronously traversing an unperturbed cell cycle

Thus it is possible that cohesin binding at centromeres is

enhanced as a consequence ofmitotic checkpoint activation

Alternatively, the persistence ofcohesin at centromeres

despite loss ofcohesion might reflect a second

noncohesion-related role for cohesin at centromeres (see below)

C E N T R O M E R E S A S S P E C I A L I Z E D

S I T E S O F C O H E S I O N

As highlighted in the preceding sections, SCC at or near

centromeres somehow differs from that on sister chromatid

arms One simple explanation for this is that all cohesin

complexes are not created equally Perhaps distinct modes

ofregulation reflect differences in the subunit composition

and/or post-translation modification ofcohesin complexes

Indeed, genome sequencing projects have revealed potential

variant cohesin subunits within a single organism, and

distinct cohesin complexes have been identified

biochemi-cally [74,75]

Meiosis is one natural situation where differences in

centromere-proximal vs centromere-distal SCC are both

functionally significant and might reflect cohesin complexes

ofdistinct composition To ensure reductional division in

meiosis I, it is imperative that sister chromatids remain

paired at their centromeres Meiotic cells express a variant

ofthe essential cohesin subunit Scc1/Mcd1, called Rec8,

that is thought to replace Scc1/Mcd1 in a meiosis-specific

cohesin complex [17,18,137–139] Despite significant

func-tional redundancy between these two complexes, they are

likely to play distinct roles in meiosis The most fascinating

property ofRec8 is its persistence at sister centromeres until

anaphase ofmeiosis II, whereas Scc1 and the bulk ofRec8

are removed prior to the onset ofanaphase I to facilitate dissolution ofchiasmata This pericentromeric fraction ofRec8 is critical for kinetochore attachment and proper chromosome alignment on both the meiosis I and meiosis II spindles Presumably, pericentromeric Rec8 is spared from cleavage and degradation at anaphase of meiosis I via interaction with an unknown protective factor(s) [140,141] Conceivably, this protective factor simultaneously stabilizes centromeric Rec8 and suppresses

or alters centromere segregation functions Candidates for such factors might be defective in mutants that exhibit precocious sister chromatid separation in meiosis I (e.g yeast bub1 mutants [142] or Drosophila mei-S332 mutants [143,144]) or that seem to bypass meiosis I altogether (e.g spo12, spo13, and slk19 mutants [140,145])

C E N T R O M E R E S R E C R U I T C O H E S I O N

P R O T E I N S , B U T D O C O H E S I O N

P R O T E I N S R E T U R N T H E F A V O R ?

As discussed above, functional centromeres in budding yeast are necessary and sufficient to promote cohesin binding over an extended chromosomal domain and in metazoans, cohesin is preferentially retained at centromeres until anaphase It follows that centromeric proteins would

be implicated in cohesin recruitment [10,25,39,146,147] However, it is too soon to know which, ifany, centromere protein directly mediates cohesin targeting because the dependency relationships for centromere assembly have not been fully elucidated Nonetheless, examples of potential cross-talk between cohesin and centromeres have been described (summarized in Fig 1) When experimentally mistargeted to noncentromeric chromatin in HeLa cells, CENP-A, but not CENP-C, causes corecruitment (or stabilization) hSMC1 [147] This suggests that CENP-A might interact with cohesin The conserved and essential Ndc80/Hec1 complex (comprised ofNdc80/Hec1, Nuf2, Spc24, and Spc25 [148,149]); might also serve to recruit cohesion proteins to centromeric regions Components of the Ndc80/Hec1 complex colocalize with centromeres and impact chromosome segregation in budding and fission yeast, as well as in human cells [146,148] Intriguingly, Ndc80/Hec1 interacts both physically and genetically with the cohesin subunit Smc1 It remains to be tested whether cohesin deposition (or selective retention) at centromeres requires Ndc80/Hec1, or vice versa

It is possible that cohesin assembly at centromeres is not directed by a specific centromere protein, but rather is an indirect consequence ofthe overall state ofhistone modifi-cation Two properties ofheterochromatin, namely, histone hypoacetylation [35,36,40] and histone H3 lysine 9 methyla-tion [38,150], are required for faithful chromosome segrega-tion As noted earlier, centromeres in many organisms are heterochromatic Moreover, several heterochromatin pro-teins that localize to centromeres via their chromo domains (which bind lysine-9-methylated histone H3) have also been implicated in centromere function (i.e SUV39H1, HP-1, Clr4, Swi6 [40,55,151–154]) Recent studies in fission yeast indicate that cohesin may be ÔattractedÕ to (or stabilized at) centromeric heterochromatin via interaction with such chromo domain proteins In the absence ofSwi6, neither Rad21 nor Psc3 (SCC3S.p.) are present at centromeres, and centromeric cohesion, but notably not arm cohesion, is