REPORTS

Resolution of Sister Telomere

Association Is Required for
Progression Through Mitosis
Jasmin N. Dynek and Susan Smith*
Cohesins keep sister chromatids associated from the time of their replication
in S phase until the onset of anaphase. In vertebrate cells, two distinct pathways
dissociate cohesins, one acts on chromosome arms and the other on centromeres. Here, we describe a third pathway that acts on telomeres. Knockdown
of tankyrase 1, a telomeric poly(ADP-ribose) polymerase caused mitotic arrest.
Chromosomes aligned normally on the metaphase plate but were unable to
segregate. Sister chromatids separated at centromeres and arms but remained
associated at telomeres, apparently through proteinaceous bridges. Thus, telomeres may require a unique tankyrase 1 dependent mechanism for sister
chromatid resolution before anaphase.
Telomere length is regulated by the
TTAGGG repeat binding protein TRF1 (1)
and its associated factor tankyrase 1 (2), a
member of the poly(ADP-ribose) polymerase (PARP) enzyme family (3). Tankyrase
1 localizes to human telomeres (4 ), as well
other subcellular sites (5, 6 ). Tankyrase 1
contains five discrete TRF1-binding sites
(7 ) and controls TRF1 binding to telomeric
DNA through ADP-ribosylation (4 ). When
tankyrase 1 is overexpressed in human
cells, TRF1 is released from telomeres and
degraded by the proteasome (8). Long-term
overexpression of tankyrase 1 induces

telomere lengthening in a reaction that

depends on its catalytic PARP activity and
on telomerase (2, 8, 9).
To inhibit tankyrase 1 expression, we transfected tankyrase 1 small interfering RNA
(siRNA) duplexes (10) into asynchronously
growing HeLaI.2.11 cells (11). Tankyrase 1
protein was lost at the earliest time point (24
hours) and throughout the time course (Fig.
1A). A similar knockdown was achieved with a
different tankyrase 1 siRNA duplex (fig. S1).
Cells treated with tankyrase 1 siRNA exhibited
a dramatic mitotic arrest, as measured by immunofluorescence (Fig. 1B and fig. S2) and

fluorescence-activated cell sorting (FACS)

analysis, which indicated a threefold increase in
the G2/M population (Fig. 1C). This phenotype
was observed with two different tankyrase 1
siRNA duplexes, and it was observed in cells
independent of their telomere length or telomerase status (table S1).
Closer examination of the cells arrested
in mitosis indicated a predominance of abnormal mitotic figures, which fell into three
distinct classes (Fig. 1D), each peaking in
number at a different point in the tankyrase
1 siRNA time course (fig. S3). The earliest
phenotype, fat, displayed a broad 4,6diamidino-2-phenylindole (DAPI) stain on
the metaphase plate (Fig. 1D, b). The second had misaligned chromosomes (Fig.
1D, c). And the third and terminal phenotype, aberrant displayed a deteriorated
spindle and a condensed and disordered
DAPI stain (Fig. 1D, d). Aberrant cells did
not reenter the cell cycle (fig. S4A) and
were not apoptotic (fig. S4B).
To determine whether tankyrase 1 expression could rescue the abnormal mitotic
phenotype, we cotransfected cells with
tankyrase 1 siRNA and with wild-type
(WT) or PARP-dead (HE/A) (9) tankyrase
1 plasmids containing a single-base mismatch to the siRNA oligonucleotide, which
rendered them resistant to the siRNAmediated knockdown (fig. S5). Analysis of
the population of mitotic cells expressing
tankyrase 1 indicated that wild-type (but
not PARP-dead) tankyrase 1 efficiently res-

Fig. 1. Knockdown of tankyrase 1 expression

results in mitotic arrest. (A) Immunoblot analysis of whole-cell extracts from HeLaI.2.11 cells
transfected without (mock) or with (TNKS1)
tankyrase 1 siRNA. (B) Histogram showing the
percentage of cells in interphase or mitosis
after transfection with tankyrase 1 siRNA.
About 1000 cells were scored by immunouorescence for each time point. (C) FACS analysis
of HeLaI.2.11 cells 48 hours after transfection
with tankyrase 1 siRNA. y axis, cell number. (D)
Immunouorescence analysis of methanolxed HeLaI.2.11 cells stained with DAPI (blue)
and -tubulin antibody (green) after treatment
with tankyrase 1 siRNA indicates a progression
of three classes of abnormal mitotic cells. Scale
bar, 2 m. (E) Tankyrase 1 siRNA cells were
cotransfected with siRNA-resistant tankyrase 1
wild-type (WT) or HE/A plasmids. Histogram
shows the ratio of normal to abnormal
tankyrase 1 expressing mitotic cells scored by
immunouorescence as described in table S2.
(F and G) Immunouorescence of methanolxed HeLaI.2.11 cells transfected with
tankyrase 1 siRNA showing (F) measurements
of the DNA (DAPI) and spindle ( tubulin) or
(G) centromere disposition by staining with
DAPI (blue), -tubulin antibody (red), and antibody against centromeres (ACA) (green).
Scale bar, 2 m.

www.sciencemag.org SCIENCE VOL 304 2 APRIL 2004

97

REPORTS
cued the abnormal mitotic phenotype (Fig.
1E and table S2).
To gain insight into the mechanism of
the arrest induced by tankyrase 1 siRNA
treatment, we focused on the earliest phenotypes. Measurements of the fat and
misaligned metaphases indicated that the
DAPI staining was much broader and the
spindle much shorter than a wild-type
metaphase (Fig. 1F and fig. S6) and more
resembled an early anaphase figure. Consistent with anaphase entry, centromeres
frequently appeared separated (Fig. 1G, b
and c), which suggested that chromosomes
aligned on the metaphase plate and underwent centromere separation, but then were
unable to proceed with anaphase.
Time-lapse analysis of mitosis in a
mock-transfected cell and a cell treated
with tankyrase 1 siRNA showed that, in
both cases, chromosomes congressed normally and by 38 min were aligned on the
metaphase plate. However, although the
mock-transfected cell continued to progress
through mitosis, the tankyrase 1 siRNA cell
remained arrested in metaphase (Fig. 2A;
Movies S1 and S2). The period of metaphase arrest in tankyrase 1 siRNA cells
varied. For example, one cell (Fig. 2B,
arrowhead) remained in metaphase from 24
to 188 min (164 min), whereas two other
flanking cells (Fig. 2B, arrows) remained
arrested in metaphase for the duration of
imaging (260 min) (Movie S3).
Progression through mitosis requires
ubiquitin-mediated proteolysis of a number
of regulatory proteins by the anaphasepromoting complex (APC) (12). The APC
substrates, cyclins A and B, were degraded in
fat mitotic figures (Fig. 2, C and D), which
indicated that the APC was active. Separation
of sister chromatids also depends on the
APC, where the final, irreversible step is
cleavage of the SCC1 cohesin subunit (13).
The SCC1 subunit was cleaved in tankyrase 1
siRNA mitotic cells (Fig. 2E). Thus, the arrest in mitosis was not due to a defect in the
APC. Finally, a negative immunostain for the
histone -H2AX (fig. S7) indicated that the
arrest in mitosis was not due to DNA damage.
Cells transfected with tankyrase 1
siRNA entered mitosis and proceeded to
metaphase normally; chromosomes aligned
on the metaphase plate and sister centromeres separated. However, at that point, the
cells were arrested and were delayed or
unable to proceed through anaphase. We
wondered whether abnormal telomere associations were preventing sister chromatids
Skirball Institute of Biomolecular Medicine, New York
University School of Medicine, 540 First Avenue, New
York, NY 10016, USA.
*To whom correspondence should be addressed. Email: smithsu@saturn.med.nyu.edu

98

from separating and moving to the poles.

Tankyrase 1 is a positive regulator of telomere length (2). Thus, we asked if there
was a gross defect in telomere replication
in tankyrase 1 siRNA cells by measuring
DNA replication using 5-bromo-2deoxyuridine (BrdU) incorporation. Telomeric DNA from cells transfected with
tankyrase 1 siRNA sedimented as heavylight DNA, consistent with at least one
round of DNA replication and similar to
mock-transfected cells (Fig. 3A). Thus,
tankyrase 1 siRNA did not detectably inhibit telomere replication.
An alternative explanation for the mitot-

ic arrest phenotype in tankyrase 1 siRNA

cells could be covalent ligation of chromosome ends. Inhibition of the telomere repeat
binding protein TRF2 results in loss of the
3 G-strand telomere tail, ligation of telomere restriction fragments, and end-to-end
fusions as observed in metaphase spreads
(11). We performed similar analyses to address this question in cells transfected with
tankyrase 1 siRNA, and there was no indication of loss of G-strand tails or telomeric
fusion products (Fig. 3B).
We further assayed for covalent ligation
of telomeres in metaphase spreads using
standard procedures, which include hypo-

Fig. 2. Characterization of the mitotic arrest in tankyrase 1 siRNA cells. (A and B) Time-lapse video
live-cell imaging of a HeLa cell line expressing histone H2B tagged with green uorescent protein
(HeLa-H2B-GFP cells) (24) 36 hours after transfection with tankyrase 1 siRNA. (A) Progression of
mitosis in a mock-transfected cell versus a tankyrase 1 siRNA cell for 108 min. Scale bar, 10 m.
(B) Larger eld which includes the tankyrase 1 siRNA cell shown in (A), (arrowhead) and two
anking cells (arrows) that remain in metaphase for 260 min. Scale bar, 10 m. (C and D)
Immunouorescence analysis of mitotic cyclins in methanol-xed HeLaI.2.11 cells stained with
DAPI (blue) and cyclin A or B (red) 20 hours after transfection with tankyrase 1 siRNA. Scale bar,
5 m. (E) Immunoblot analysis of whole-cell extracts from HeLa-SCC1-myc cells 40 hours after
treatment with tankyrase 1 siRNA. TNKS1 cells were isolated by mitotic shake-off. Arrow indicates
the SCC1myc cleavage product.

2 APRIL 2004 VOL 304 SCIENCE www.sciencemag.org

REPORTS
tonic treatment. Note that this treatment,
which was developed to allow visualization
of separated chromosome arms, may release chromosomal proteins (14 ). Typically, in these preparations, sister chromatids
remain associated at their centromeres because of tenacious protein-protein interactions (Fig. 3C). Because centromeres were
no longer associated in cells transfected
with tankyrase 1 siRNA (Fig. 1G and see
Fig. 4B, below), chromosomes appeared as
separated chromatids, and end-to-end fusions were not observed (Fig. 3C).
Alternatively, sister telomeres could associate through proteinaceous bridges,
which may not survive the hypotonic swelling used in metaphase spread preparations.

To address this question, we turned to a

fluorescent in situ hybridization (FISH)
replication timing assay (15), where cells
were isolated by mitotic shake-off and
fixed directly (without hypotonic treatment) to avoid artifactual separation of sister chromatids. In control mitotic cells, telomeric regions generally appeared as two
doublets, which indicated that telomeres
had replicated and separated (Fig. 4A; table
S3). In contrast, in cells transfected with
tankyrase 1 siRNA, two singlets were frequently observed, which indicated that the
telomeric regions had not separated (Fig.
4A; table S3). Centromere probes appeared
as two doublets in both mock-transfected
cells and cells transfected with tankyrase 1

Fig. 3. Telomeres undergo

DNA replication and are
not covalently fused in
tankyrase 1 siRNA cells.
(A) Telomeric DNA was
isolated from HelaI.2.11
cells after 48 hours of cotreatment with tankyrase
1 siRNA and BrdU, and
fractionated on CsCl density gradients. Telomeric
DNA was measured by
dot-blot analysis using a
32
P-labeled telomeric probe.
The positions of light-light
(LL), heavy-light (HL), and
heavy-heavy (HH) DNA are
indicated. (B) Southern blot
analysis of telomere restriction fragments isolated from
HeLaI.2.11 cells after 48
hours of treatment with
tankyrase 1 siRNA, and fractionated by pulsed-eld gel
electrophoresis. The gel was
hybridized to a 32P endlabeled (CCCTAA)4 oligonucleotide under native or
denatured conditions. (C) Telomeric PNA FISH analysis (FISH using peptide nucleic acid probes) of
metaphase spreads of HeLaI.2.11 cells collected 48 hours after transfection with tankyrase 1 siRNA,
swollen in hypotonic buffer, and xed in methanol-acetic acid. Telomeric repeats were detected by
using a Cy3-(CCCTAA)3 PNA probe (red). DNA was stained with DAPI (blue). Scale bar, 5 m.

Fig. 4. Sister telomeres (unlike arms and centromeres) remain associated

in cells treated with tankyrase 1 siRNA. (A to D) Chromosomespecic FISH analysis of HeLaI.2.11 cells collected by mitotic shakeoff 48 hours after treatment with tankyrase 1 siRNA. Cells were

siRNA, but showed greater separation in

the latter (Fig. 4B).
In contrast to telomere-specific probes,
arm probes showed no difference between
cells treated with tankyrase 1 siRNA and
mock-transfected cells and generally appeared as two doublets (Fig. 4C; table S3).
Thus, unlike telomeres, chromosome arms
had separated. Finally, we performed double
FISH using arm- and telomere-specific
probes from the same chromosome simultaneously. In tankyrase 1 siRNA cells, chromosome arms appeared as two doublets, whereas, in the same cell, telomeres appeared as
two singlets (Fig. 4D). Thus, in tankyrase 1
siRNA cells, chromosome arms and centromeres had replicated and separated, but telomeres [although replicated (Fig. 3A)] remained associated.
Knockdown of tankyrase 1 expression
blocked cells from completing mitotic anaphase. Although sister chromatids separated
at their centromeres and along their arms,
they remained associated at their telomeres.
Thus, the inability of sister chromatids to
segregate to daughter poles appeared to be
due to a persistent telomere association. A
role for telomeres in sister chromatid separation during mitosis has been suggested in
Tetrahymena (16 ), Schizosaccharomyces
pombe (17), and Drosophila (18).
Our results are consistent with the hypothesis that telomeres of sister chromatids normally associate and that resolution of this
association is required for progression
through anaphase. Our finding that telomere
resolution is blocked, whereas centromere
and arm resolution appears to occur normally, indicates a unique mechanism for
telomeres which is distinct from arms and
centromeres. In vertebrates, cohesins are released from chromosomes by two distinct
posttranslational mechanisms (13, 19). The
bulk of cohesins are released during prophase
by a phosphorylation-mediated mechanism
(20, 21), whereas the remaining centromeric

xed directly in methanol-acetic acid without hypotonic swelling and

hybridized to chromosome-specic uorescently labeled probes: telomeres (green), arms (red), and centromeres (red). DNA was stained with
DAPI (blue). Scale bar, 2 m.

www.sciencemag.org SCIENCE VOL 304 2 APRIL 2004

99

REPORTS
cohesins are released at metaphase by proteolytic cleavage (22, 23). Our finding that
tankyrase 1 PARP activity is required to rescue the abnormal mitotic phenotype implicates a third posttranslational mechanism,
poly(ADP-ribosyl)ation, in sister chromatid
resolution. Whether telomeres are held together by cohesins or by telomere-specific
proteins, such as TRF1 and its associated
factors, remains to be determined.

1.
2.
3.
4.

References and Notes

B. van Steensel, T. de Lange, Nature 385, 740 (1997).

S. Smith, T. de Lange, Curr. Biol. 10, 1299 (2000).
S. Smith, Trends Biochem. Sci. 26, 174 (2001).
S. Smith, I. Giriat, A. Schmitt, T. de Lange, Science
282, 1484 (1998).
5. S. Smith, T. de Lange, J. Cell Sci. 112, 3649 (1999).
6. N. W. Chi, H. F. Lodish, J. Biol. Chem. 275, 38437
(2000).
7. H. Seimiya, S. Smith, J. Biol. Chem. 277, 14116
(2002).

8. W. Chang, J. N. Dynek, S. Smith, Genes Dev. 17, 1328

(2003).
9. B. D. Cook, J. N. Dynek, W. Chang, G. Shostak, S.
Smith, Mol. Cell. Biol. 22, 332 (2002).
10. S. M. Elbashir et al., Nature 411, 494 (2001).
11. B. van Steensel, A. Smogorzewska, T. de Lange, Cell
92, 401 (1998).
12. J. M. Peters, Mol. Cell 9, 931 (2002).
13. I. C. Waizenegger, S. Hauf, A. Meinke, J. M. Peters, Cell
103, 399 (2000).
14. Y. Ohnuki, Chromosoma 25, 402 (1968).
15. R. Or et al., Chromosoma 111, 147 (2002).
16. K. E. Kirk, B. P. Harmon, I. K. Reichardt, J. W. Sedat,
E. H. Blackburn, Science 275, 1478 (1997).
17. H. Funabiki, I. Hagan, S. Uzawa, M. Yanagida, J. Cell
Biol. 121, 961 (1993).
18. M. M. Donaldson, A. A. Tavares, H. Ohkura, P. Deak,
D. M. Glover, J. Cell Biol. 153, 663 (2001).
19. A. Losada, M. Hirano, T. Hirano, Genes Dev. 12, 1986
(1998).
20. I. Sumara et al., Mol. Cell 9, 515 (2002).
21. A. Losada, T. Yokochi, R. Kobayashi, T. Hirano, J. Cell
Biol. 150, 405 (2000).
22. S. Hauf, I. C. Waizenegger, J. M. Peters, Science 293,
1320 (2001).

SUMO Modication of
Huntingtin and Huntingtons
Disease Pathology
Joan S. Steffan,1 Namita Agrawal,2 Judit Pallos,2
Erica Rockabrand,1 Lloyd C. Trotman,3 Natalia Slepko,2
Katalin Illes,1 Tamas Lukacsovich,2 Ya-Zhen Zhu,1
Elena Cattaneo,4 Pier Paolo Pandol,3
Leslie Michels Thompson,1,5* J. Lawrence Marsh2*
Huntingtons disease (HD) is characterized by the accumulation of a pathogenic
protein, Huntingtin (Htt), that contains an abnormal polyglutamine expansion.
Here, we report that a pathogenic fragment of Htt (Httex1p) can be modied
either by small ubiquitin-like modier (SUMO)1 or by ubiquitin on identical
lysine residues. In cultured cells, SUMOylation stabilizes Httex1p, reduces its
ability to form aggregates, and promotes its capacity to repress transcription.
In a Drosophila model of HD, SUMOylation of Httex1p exacerbates neurodegeneration, whereas ubiquitination of Httex1p abrogates neurodegeneration.
Lysine mutations that prevent both SUMOylation and ubiquitination of Httex1p
reduce HD pathology, indicating that the contribution of SUMOylation to HD
pathology extends beyond preventing Htt ubiquitination and degradation.
HD is a dominant neurodegenerative disorder
caused by the expansion of a polyglutamine
[poly(Q)] repeat in Htt (1, 2). In HD and
Department of Psychiatry and Human Behavior, Gillespie
2121, University of California, Irvine, CA 92697, USA. 2Department of Developmental and Cell Biology, 4444 McGaugh Hall, University of California, Irvine, CA 92697, USA.
3
Molecular Biology Program and Department of Pathology,
Sloan-Kettering Institute, Memorial Sloan-Kettering Cancer
Center, 1275 York Avenue, New York, NY 10021, USA.
4
Department of Pharmacological Sciences, Center of Excellence on Neurodegenerative Diseases, University of Milano,
Via Balzaretti 9, 20133 Milano, Italy. 5Department of Biological Chemistry, D240 Medical Sciences I, University of
California, Irvine, CA 92697, USA.
1

*These authors contributed equally to this work.

To whom correspondence should be addressed. Email: jlmarsh@uci.edu

100

other poly(Q) diseases, mutant proteins or

pathogenic poly(Q) peptides produced by
proteolytic processing aggregate into nuclear
and/or cytosolic inclusions in neurons and in
neuronal processes. These aggregates also
contain other cellular proteins including
transcription-regulating proteins, chaperones,
proteasome subunits, and ubiquitin (35).
Pathogenic poly(Q) proteins can be
modified in ways that change their cellular
function or fate. Htt is subject to ubiquitination, which normally targets proteins for
degradation (6, 7 ). Mutations in ubiquitin
ligases enhance poly(Q) toxicity in Drosophila, mouse, and cell models (810),
whereas overexpression of Parkin, an E3
ubiquitin ligase, can reduce poly(Q) aggre-

23. F. Uhlmann, F. Lottspeich, K. Nasmyth, Nature 400,

37 (1999).
24. T. Kanda, K. F. Sullivan, G. M. Wahl, Curr. Biol. 8, 377
(1998).
25. We thank G. Wahl for H2B-GFP-HeLa cells, J. M.
Peters for SCC1-mycHeLa cells; M. Pagano for antibody against cyclin A; A. J. North (Bio-Imaging Resource Center, Rockefeller University) for assistance
with live imaging; and D. Roth, T. Meier, and members of the Smith laboratory for comments on the
manuscript. J.N.D. was supported by an NIH Predoctoral Training Program in Cell and Molecular Biology
(GM07238-28). This work was supported by grants to
S.S. from the Edward Mallinckrodt, Jr., Foundation,
the New York City Council Speakers Fund for Biomedical Research, and the NIH (R01 CA95099-01).
Supporting Online Material
www.sciencemag.org/cgi/content/full/304/5667/97/DC
Materials and Methods
Figs. S1 to S7
Tables S1 to S3
References
Movies S1 to S3
16 December 2003; accepted 2 February 2004

gation and suppress cytotoxicity (11).

Thus, ubiquitination appears to reduce
poly(Q) toxicity, presumably by promoting
the degradation of toxic proteins.
SUMOylationthe covalent attachment
of SUMO-1 to lysine residuesis a posttranslational modification process [for reviews, see (12, 13)] that is biochemically
similar to, but functionally distinct from,
ubiquitination. SUMOylation can alter the
function or subcellular localization of proteins, and competition between SUMO-1
and ubiquitin for identical target lysines
can protect some proteins from degradation
(1416 ). The majority of SUMO-modified
proteins are located in the nucleus (17 ), and
SUMOylation can have a direct effect on
nucleocytoplasmic transport. Here, we investigate how SUMOylation of Htt might
affect HD pathogenesis.
Truncated Htt [Httex1p 97QP (Fig. 1A)
(18)] and HISSUMO-1 colocalize when
transfected into immortalized striatal nerve
cells from cell line 12.7 (19) (Fig. 1B), reflecting either direct modification of Htt by
SUMO-1 or colocalization of Htt with other
SUMOylated proteins. To identify possible
modifications of mutant Httex1p (7, 20),
which contains only three lysine residues
[K6, K9, and K15 (Fig. 1A)], HISSUMO-1
or HIS-ubiquitin was coexpressed with
Httex1p either intact or with the lysines mutated to arginine [K63 R6 (K6R), K9R, and
K15R] in both striatal cells and HeLa cells.
We also compared Htt fragments either with
or without (97QP and 103Q, respectively) the
proline-rich domain immediately following
the poly(Q) region (Fig. 1A). Both proteins
can be SUMOylated or ubiquitinated (Fig.
1C, lower panel) and a single primary
SUMOylated species predominates, although
more complex SUMO or ubiquitin modifications can be seen [Fig. 1D and supporting

2 APRIL 2004 VOL 304 SCIENCE www.sciencemag.org