REPORTS
Distinct Cohesin Complexes
Organize Meiotic Chromosome
Domains
Tomoya S. Kitajima,1* Shihori Yokobayashi,1*
Masayuki Yamamoto,1 Yoshinori Watanabe1,2†
Meiotic cohesin complexes at centromeres behave differently from those along
chromosome arms, but the basis for these differences has remained elusive. The
fission yeast cohesin molecule Rec8 largely replaces its mitotic counterpart,
Rad21/Scc1, along the entire chromosome during meiosis. Here we show that
Rec8 complexes along chromosome arms contain Rec11, whereas those in the
vicinity of centromeres have a different partner subunit, Psc3. The armassociated Rec8-Rec11 complexes are critical for meiotic recombination. The
Rec8-Psc3 complexes comprise two different types of assemblies. First, pericentromeric Rec8-Psc3 complexes depend on histone methylation-directed
heterochromatin for their localization and are required for cohesion during
meiosis II. Second, central core Rec8-Psc3 complexes form independently of
heterochromatin and are presumably required for establishing monopolar attachment at meiosis I. These findings define distinct modes of assembly and
functions for cohesin complexes at different regions along chromosomes.
Eukaryotic sister chromatid cohesion is established during S phase and is maintained
throughout G2 until M phase of the cell cycle.
This cohesion is mediated by cohesin, a multisubunit complex (1–4). The mitotic cohesin
complex is composed of two SMC (structural
maintenance of chromosome) family pro-
teins, Smc1 and Smc3, and two accessory
subunits, Scc1 and Scc3 [Psm1, Psm3,
Rad21, and Psc3 in fission yeast, respectively
(5)], all of which are essential for cohesion
function and cell proliferation. In meiosis,
Rad21/Scc1 is dispensable (6, 7) and is largely replaced by its meiotic counterpart, Rec8
(8–10). During meiotic prophase, Rec8 complexes play a central role in establishing sister
chromatid cohesion and facilitating recombination. In addition, centromeric Rec8 is required for ensuring that each kinetochore
within a sister chromatid pair attaches to the
same spindle pole (monopolar attachment) in
fission yeast (6, 9). At anaphase I, Rec8 is
disrupted along the arms, whereas centromeric Rec8 persists so that recombined homologs
separate but sister chromatids move together
to the same spindle pole (7, 9, 11). At meiosis
II, when sister chromatids separate, centromeric Rec8 is disrupted. Thus, Rec8 complexes at centromeres and along chromosome
arms appear to play different roles and must
be differentially regulated through meiosis.
Fission yeast have two Scc3-like proteins,
Psc3 and Rec11 (12, 13). Psc3 plays an essential role in sister chromatid cohesion during
mitosis by forming a complex with Rad21 (fig.
S5) (5). In contrast, Rec11 is meiosis-specific,
and its mutation reduces recombination, like
Rec8 (14, 15). Thus, circumstantial evidence
suggests that Rec11 may function together with
Department of Biophysics and Biochemistry, Graduate School of Science, University of Tokyo, 2SORST,
Japan Science and Technology Corporation, Hongo,
Tokyo 113-0033, Japan.
1
*These authors contributed equally to this work.
†To whom correspondence should be addressed. Email: [email protected]
Fig. 1. Distinct roles of the cohesin subunits of Rec8,
Rec11, and Psc3 during meiosis. (A) The cut3-GFP dots
were monitored in meiotic cells arrested by mei4⌬ at
late prophase I (19). Examples of wild-type and rec11⌬
cells are shown at bottom. (B) One of the homologs was
marked with cen1-GFP, cen2-GFP, and cen3-GFP and
monitored for segregation pattern during meiosis. The
segregation patterns of cen-GFP during meiosis are illustrated with examples of rec11⌬ and rec8⌬ cells. The
equational pattern of wild-type cells (*) is mostly caused
by recombination (19). (C) The segregation pattern of
cen3-GFP marked on both homologs was monitored
after meiosis I by arresting cells with the mes1 mutation.
Examples of rec11⌬ cells are shown at bottom. (D) The
indicated cells were grown on the yeast extract agar
(YEA) plate. Note that rec11⫹ is not expressed during
mitosis, therefore rad21⌬ rec8⫹o.p. psc3⫹ rec11⫹ cells
sustain viability by the Rec8-Psc3 pair. The same cells
were induced to meiosis and monitored for the segregation pattern of cen2-GFP marked on one chromosome.
Representative cells are shown at bottom.
1152
16 MAY 2003 VOL 300 SCIENCE www.sciencemag.org
REPORTS
Rec8 as a component of the meiosis-specific
cohesin complex. To evaluate this possibility,
we determined whether Rec11 is required for
sister chromatid cohesion in meiosis by monitoring green fluorescent protein (GFP) fluorescence associated with the cut3 locus (cut3GFP), which is located at the middle of the left
arm of chromosome 2. During prophase of
meiosis I, wild-type cells show two cut3-GFP
dots (Fig. 1A), indicating stable associations
between each pair of sister chromatids. However, rec11⌬ cells often contain three or four
cut3-GFP dots (Fig. 1A), representing dissociation in these regions, as observed in rec8⌬
cells. This premature separation of sister cut3
sequences appears in rec11⌬ cells at the end of
premeiotic DNA replication (fig. S1) and is not
an indirect effect of decreased recombination,
as sister cohesion is unimpaired in cells lacking
rec12 (the SPO11 homolog in fission yeast)
and, thus, lacking meiotic recombination entirely (16). Identical results were obtained by monitoring ade3-GFP, which lies in the middle of
the left arm of chromosome 1 (17). Hence,
Rec11, like Rec8, plays a crucial role in sister
chromatid cohesion along chromosome arms
during meiotic prophase I.
To determine whether Rec11 is also required for centromere functions, we marked a
centromere-linked sequence (cen-GFP) on only
one of the two homologs in a zygote and monitored segregation of the GFP dots during meiosis (Fig. 1B). rec8⌬ cells undergo predominantly equational (i.e., ⫹/–, ⫹/–) chromosome
segregation at meiosis I (9). If Rec11 worked
with Rec8 at centromeres, rec11⌬ cells would
show this equational segregation pattern at meiosis I. However, nearly all pairs of sister chro-
Fig. 2. In meiosis, Psc3 and
Rec11 locate at centromeres and
along chromosome arms, respectively. (A) Rec11-HA or
Psc3-HA was counterstained
with Rec8-GFP and DAPI (4⬘,6⬘diamidino-2-phenylindole)
in
meiotic prophase (19). In the
merged images, Rec11-HA or
Psc3-HA is represented by red
and Rec8-GFP by green. Early
and middle prophase nuclei were
selected according to the Rec8GFP pattern. In rec11⌬, these
nuclei cannot be discriminated
because of the decrease of arm
Rec8 throughout prophase. (B) Schematic map of chromosome 1 and the primers used (cnt, dg,
lys1, mes1). ChIP assays (19) with GFP antibodies were used to measure Rec8-GFP, Rec11-GFP, or
Psc3-GFP levels in wild-type cells and Rec8-GFP levels in rec11⌬ cells throughout the indicated
chromosome sites. The ratio of cnt/mes1 in each cell is shown at the bottom. (C) The locations of
Rec8-GFP, Rec11-GFP, Psc3-GFP, and Psm3-GFP were examined during the progression of meiosis.
matids in rec11⌬ cells move together to the
same pole (reductional-like ⫹/⫹, –/– pattern)
(Fig. 1B). Recombination-deficient rec12⌬
cells also show a ⫹/⫹, –/– segregation pattern,
and deletion of rec8⫹ from either rec11⌬ or
rec12⌬ cells shifts their segregation pattern
from reductional-like to equational. Therefore,
monopolar spindle attachment of sister centromeres is independent of crossing over or chiasmata formation and is directly regulated by
centromeric Rec8 in a Rec11-independent
manner.
Despite the fact that sister chromatid movement appears normal, rec11⌬ cells nevertheless
exhibit low spore viability (⬍70% viable) (15,
17). Experiments in which cen3 were marked
with GFP on both homologs revealed that
⬎20% of rec11⌬ cells exhibit homolog nondisjunction, in which both homolog pairs move to
the same pole at meiosis I (Fig. 1C). Nonrandom segregation of homologs in rec11⌬ cells
(different from rec12⌬ cells) can be explained
by residual levels of recombination (15) and,
presumably, residual cohesion as well. rec11⌬
cells undergo faithful disjunction during meiosis II (fig. S2C), indicating that centromeric
cohesion persists through meiosis I. Thus, the
abnormal chromosome segregation and reduced spore viability of rec11⌬ cells stem from
defective arm cohesion and recombination with
ensuing nondisjunction at meiosis I.
Disruption of monopolar attachment in
rec8⌬ cells but not in rec11⌬ cells suggests that
Rec11 is not the sole partner of Rec8. Though
Rec11 is indeed meiosis-specific, Psc3 is expressed during both mitosis and meiosis (fig.
S2A), suggesting some meiotic role for Psc3.
Mitotic cells carrying a temperature-sensitive
allele of psc3 (psc3-2T) (18) displayed extensive separation of cut3-GFP dots (fig. S2B),
whereas arm cohesion of psc3-2T cells is completely intact during meiotic prophase (fig.
S2C). The psc3-2T mutation slightly enhances
the dissociation of cut3 sequences in a rec11⌬
background (fig. S2C), suggesting that Psc3
may assist arm cohesion if Rec11 is absent in
meiosis and that Psc3 may secure the residual
meiosis I disjunction of rec11⌬ cells. In contrast to the dramatic reduction of meiotic recombination in rec11⌬ cells, wild-type levels
of recombination occur in psc3-2T cells (fig.
S2D). Moreover, the overexpression of psc3⫹
leads to partial recovery of the arm cohesion
defect of rec11⌬, but the recombination defect
remains unimproved (fig. S3). Thus, Psc3 may
play a minor role in cohesion along meiotic
chromosome arms, but it has no ability to promote recombination.
Because Rec11 is dispensable for the centromeric functions of meiotic cohesin, Psc3
might instead partner with Rec8 at centromeres.
Indeed, psc3-2T cells show defects in sister
chromatid segregation during meiosis (fig.
S2E). To definitively illuminate the meiotic
roles for Psc3, we exploited the fact that ectopic
www.sciencemag.org SCIENCE VOL 300 16 MAY 2003
1153
REPORTS
expression of the Rec8-Rec11 pair sustains viability of rad21⌬ psc3⌬ cells (Fig. 1D), thus
allowing meiotic induction in the complete absence of Psc3. The control rec8⫹ overexpressing (o.p.) psc3⫹ rec11⫹ cells undergo proper
meiotic chromosome segregation, both reductional (meiosis I) and equational (meiosis II).
However, rec8⫹o.p. psc3⌬ rec11⫹o.p. cells
show defective sister chromatid movement at
both meiotic divisions (Fig. 1D). Thus, Psc3
plays a crucial role in kinetochore regulation at
both meiotic divisions, and this function of Psc3
cannot be replaced by overexpression of Rec11.
In contrast, the Rec8-Rec11 pair sustains mitotic growth better than Rec8-Psc3 (Fig. 1D; fig.
S4A), underscoring the meiosis specificity of
kinetochore regulation by Rec8-Psc3.
The distinct functions of Rec11 and Psc3
should be reflected by differential localization
along meiotic chromosomes. In situ immunofluorescence showed that both Rec11 and Rec8
appear only during meiosis, whereas Psc3 is
detected during both mitosis and meiosis. During meiotic prophase, Rec8 first appears at the
centromeres and later distributes throughout the
chromosome (Fig. 2A) (9). Rec11 does not
coexist with Rec8 at centromeres in early meiotic prophase, but later it colocalizes with the
distal chromosomal signals of Rec8. Conversely, Psc3 almost exclusively colocalizes with
centromeric Rec8 throughout prophase I (Fig.
2A). To more precisely delineate the localizations of these cohesin proteins, we used a chromatin immunoprecipitation (ChIP) assay (19)
using primers that amplify the centromeric central core (cnt) or pericentromeric (dg) regions, a
centromere-proximal arm region (lys1), and a
middle arm region (mes1) (Fig. 2B). The data
reveal that Rec8 associates more with centromere regions than with chromosome arms (20),
whereas Rec11 associates with arm regions
more than with the centromere. Although
Rec11 is not fully excluded from centromeres,
the centromere-enrichment ratio (cnt/mes1) of
Rec11 is less than one-fifth that of Rec8. In
contrast, Psc3 associates exclusively with the
centromere, showing a centromere-enrichment
ratio ⬎4 times higher than Rec8 and ⬎20 times
higher than Rec11 (Fig. 2B). In rec11⌬ cells,
the association of Rec8 is reduced selectively at
arm sites, as shown by both ChIP and immunofluorescence assays (Fig. 2, A and B,
rec11⌬). Moreover, when Rec8-GFP is expressed ectopically during mitosis, it localizes
to centromeres much more efficiently if Psc3 is
coexpressed than if Rec11 is coexpressed (fig.
S4). Furthermore, immunoprecipitation experiments demonstrated that Rec8 interacts with
Psc3, Rec11, and Psm3/Smc3 in vivo (fig. S5).
At anaphase of meiosis I, Rec11 signals become faint in the nucleus, but Psc3 persists,
together with Rec8, at the clustered centromeres (Fig. 2C). The centromeric Rec8-Psc3
dots disappear at meiosis II. The foregoing
results suggest a scheme in which arm-associ-
1154
Fig. 3. Pericentromeric
localization of Rec8Psc3 requires histone
H3 methyl transferase
and heterochromatin
and is required for
persisting
cohesion
throughout meiosis I.
(A) Location of Rec8GFP at the centromeres was examined
by ChIP assays in the
indicated strains arrested at meiotic
prophase. Central core
(cnt, imr) and pericentromeric regions (dg,
dh) were examined.
(B) One of the homologs marked with
cen1-GFP was monitored for segregation
pattern in tetra-nucleated cells. (C) Separation of sister cen2GFP dots after meiosis
I is evident in clr4⌬
cells. (D) The Rec8 signal is nearly lost from
centromeres in clr4⌬
cells arrested after
meiosis I by the mes1
mutation.
ated Rec8 cooperates primarily with Rec11 to
promote recombination and maintain arm cohesion until the end of meiosis I, whereas centromeric Rec8 works together with Psc3 throughout meiosis I until meiosis II.
How do the two types of Rec8 complexes
associate with distinct regions of chromosomes? Recent studies from our laboratory and
others have shown that Rad21-Psc3 complexes
preferentially assemble at pericentromeric heterochromatin regions. This assembly depends
on histone H3 methylation by Clr4/Suv39hRik1 and the consequent binding of the heterochromatin protein Swi6/HP1 (18, 21, 22). Psc3
interacts with the chromodomain of Swi6 in a
two-hybrid assay (18), whereas Rec11 does not.
Therefore, Rec8-Psc3 complexes could target
the centromeric regions via the Psc3-heterochromatin interaction. To test this possibility,
we assessed the effects of swi6 or clr4 deletion
on Rec8 localization to centromeres. The level
of Rec8 association with pericentromeric regions is markedly reduced in these mutants,
whereas Rec8 is still enriched at the central core
(Fig. 3A). These results suggest that preferential
localization of Rec8-Psc3 at centromeres is promoted by two independent mechanisms, a Clr4Swi6 – dependent mechanism at pericentromeric regions and a Clr4-Swi6 –independent
mechanism at the central core. The replacement
of histone H3 by its variant CENP-A at the
central core (23) might provide a basis for the
localization of Rec8-Psc3 there.
The restriction of histone modificationdependent localization of cohesin to the peri-
centromeric region prompted us to address the
role of pericentromeric cohesin in normal meiotic chromosome segregation. In swi6⌬ and
clr4⌬ cells marked on one pair of sister chromatids with cen1-GFP, sister chromatid pairs
move together to the same nucleus during meiosis I, indicating that monopolar attachment is
intact in these mutants (Fig. 3B). Moreover,
homologous chromosomes undergo faithful
disjunction at meiosis I in these mutants (Fig.
3C) (17). At meiosis II, however, sisters fail to
segregate properly, undergoing nondisjunction
in 20 to 40% of cells (Fig. 3B). The defect in
meiosis II is more penetrating in clr4⌬ cells
than in swi6⌬ cells, with a pattern approximating random segregation. This is reconcilable
with the more thorough decrease in pericentromeric Rec8 in clr4⌬ cells (Fig. 3A). Reinforcing the foregoing results, clr4⌬ cells frequently
display precocious separation of cen2-GFP signals (Fig. 3C) and a decreased level of Rec8 at
the centromeres if arrested after meiosis I (Fig.
3D). Thus, Clr4-Swi6 – dependent enrichment
of Rec8-Psc3 at pericentromeric regions is required to preserve centromeric cohesion
through meiosis I, thereby ensuring equational
segregation in meiosis II (Fig. 4). The defects
accompanying loss of pericentromeric heterochromatin in fission yeast resemble those of
Drosophila lacking the MEI-S332 protein,
which is a proposed guardian of centromeric
cohesion in meiosis II (24). Although we do not
know the precise mechanism for how the centromeric cohesin is protected from degradation
until meiosis II, our results suggest that peri-
16 MAY 2003 VOL 300 SCIENCE www.sciencemag.org
REPORTS
homologs tightly together to ensure disjunction
at meiosis I. Thus, the meiosis-specific cohesin
subunit Rec11/STAG3 might have evolved to
strengthen or develop the arm function of cohesin. We have shown that defects in the armspecific Rec11-associated cohesin cause meiosis I nondisjunction. This phenotype differs
markedly from the more disruptive chromosome segregation seen in rec8⌬. Therefore,
such arm-specific cohesion factors may be particularly important to consider with regard to
human reproductive problems like Down syndrome, which arises mainly from nondisjunction in meiosis I.
References and Notes
Fig. 4. A model for the action of meiotic cohesins. The arm-specific complex Rec8-Rec11
required for recombination and for holding homologs is disrupted during meiosis I. The heterochromatin-dependent location of Rec8-Psc3
at pericentromeric regions is required for persisting centromere cohesion until meiosis II.
The central core Rec8-Psc3 is presumably required for establishing monopolar attachment
of sister kinetochores for meiosis I.
centromeric cohesin might be the major target
of protection from degradation at anaphase I.
Given that pericentromeric Rec8 is dispensable
for the monopolar attachment, Rec8 bound to
the centromeric central core could be responsible for this role at meiosis I (Fig. 4), as predicted in our previous model (6, 20).
Previous studies have established that the
cohesin complex in meiosis is different from
that in mitosis. Our findings indicate that the
species of cohesin complex also varies with
location along the chromosome, thereby organizing distinct chromosome domains. Mammals have at least three Scc3-like proteins,
STAG1-3/SA1-3. STAG3 is meiosis-specific,
localizing along chromosome arms during meiotic prophase I but becoming undetectable
thereafter (25), indicating that STAG3 is most
likely an ortholog of Rec11. Drosophila also
has a meiosis-specific Scc3-like protein (26),
suggesting the conservation of this principle.
During mitosis, a substantial release of cohesin
from chromosome arms accompanies the process of chromosome condensation, producing
two recognizable chromatids (27). Nonetheless, important levels of cohesin remain, especially at centromeres, until the onset of anaphase. Cohesin enrichment at centromeres
during mitosis is mediated through an interaction between cohesin and heterochromatin in
fission yeast and also, presumably, in mammals
(18, 28). This system might be insufficient for
the more complex process of meiosis, as cohesion along chromosome arms plays a crucial
role in promoting recombination and holding
1. K. Nasmyth, Annu. Rev. Genet. 35, 673 (2001).
2. J. Y. Lee, T. L. Orr-Weaver, Annu. Rev. Cell. Dev. Biol.
17, 753 (2001).
3. T. Hirano, Genes Dev. 16, 399 (2002).
4. F. Uhlmann, Curr. Biol. 13, 104 (2003).
5. T. Tomonaga et al., Genes Dev. 14, 2757 (2000).
6. S. Yokobayashi, M. Yamamoto, Y. Watanabe, Mol.
Cell. Biol., in press.
7. F. Klein et al., Cell 98, 91 (1999).
8. S. Parisi et al., Mol. Cell. Biol. 19, 3515 (1999).
9. Y. Watanabe, P. Nurse, Nature 400, 461 (1999).
10. P. Pasierbek et al., Genes Dev. 15, 1349 (2001).
11. S. B. Buonomo et al., Cell 103, 387 (2000).
12. Y. F. Li, M. Numata, W. P. Wahls, G. R. Smith, Mol.
Microbiol. 23, 869 (1997).
13. A. Toth et al., Genes Dev. 13, 320 (1999).
14. L. C. DeVeaux, G. R. Smith, Genes Dev. 8, 203 (1994).
15. M. D. Krawchuk, L. C. DeVeaux, W. P. Wahls, Genetics
153, 57 (1999).
16. M. D. Cervantes, J. A. Farah, G. R. Smith, Mol. Cell 5,
883 (2000).
17. Y. Watanabe, unpublished data.
18. N. Nonaka et al., Nature Cell Biol. 4, 89 (2002).
19. Materials and Methods are available as supporting
online material at Science Online.
20. Y. Watanabe, S. Yokobayashi, M. Yamamoto, P.
Nurse, Nature 409, 359 (2001).
21. J. Nakayama, J. C. Rice, B. D. Strahl, C. D. Allis, S. I.
Grewal, Science 292, 110 (2001).
22. P. Bernard et al., Science 294, 2539 (2001).
23. K. Takahashi, E. S. Chen, M. Yanagida, Science 288,
2215 (2000).
24. T. L. Orr-Weaver, Cell 99, 1 (1999).
25. I. Prieto et al., Nature Cell Biol. 3, 761 (2001).
26. B. McKee, personal communication.
27. A. Losada, M. Hirano, T. Hirano, Genes Dev. 16, 3004
(2002).
28. P. Bernard, R. C. Allshire, Trends Cell Biol. 12, 419
(2002).
29. We thank J. P. Cooper for critical reading of the
manuscript and G. R. Smith, S. I. S. Grewal, K. Nabeshima, A. Yamamoto, Y. Hiraoka, and M. Yanagida
for cell strains. We thank all the members of our lab
for their help. This work was supported in part by
grants from the Ministry of Education, Science and
Culture of Japan.
Supporting Online Material
www.sciencemag.org/cgi/content/full/300/5622/1152/
DC1
Materials and Methods
Figs. S1 to S5
References
19 February 2003; accepted 14 April 2003
Tumor Response to Radiotherapy
Regulated by Endothelial Cell
Apoptosis
Monica Garcia-Barros,1 Francois Paris,1 Carlos Cordon-Cardo,2
David Lyden,3 Shahin Rafii,5 Adriana Haimovitz-Friedman,4
Zvi Fuks,4* Richard Kolesnick1*†
About 50% of cancer patients receive radiation therapy. Here we investigated
the hypothesis that tumor response to radiation is determined not only by
tumor cell phenotype but also by microvascular sensitivity. MCA/129 fibrosarcomas and B16F1 melanomas grown in apoptosis-resistant acid sphingomyelinase (asmase)– deficient or Bax-deficient mice displayed markedly reduced baseline microvascular endothelial apoptosis and grew 200 to 400%
faster than tumors on wild-type microvasculature. Thus, endothelial apoptosis
is a homeostatic factor regulating angiogenesis-dependent tumor growth.
Moreover, these tumors exhibited reduced endothelial apoptosis upon irradiation and, unlike tumors in wild-type mice, they were resistant to single-dose
radiation up to 20 grays (Gy). These studies indicate that microvascular damage
regulates tumor cell response to radiation at the clinically relevant dose range.
Ionizing radiation is a widely used therapy
for solid tumors and is thought to act by
directly targeting tumor clonogens, also
known as stem cells (1, 2). Tumor curability
is believed to be determined by the most
resistant clonogen, because one surviving
stem cell appears sufficient for reconstituting
tumor growth (3, 4). This model appears
relevant to several normal tissues, particularly those classified as rapid-turnover systems.
For example, gastrointestinal (GI) damage is
believed to result from direct interaction of
radiation with the clonogenic compartment at
the crypt of Lieberkühn base (5, 6). However,
we recently reported that microvascular endothelial apoptosis is required for clonogenic
cell dysfunction (7). GI damage was prevented when endothelial cell apoptosis was inhibited genetically by asmase–/– depletion or
pharmacologically by intravenous basic fi-
www.sciencemag.org SCIENCE VOL 300 16 MAY 2003
1155