articles
Unconventional tethering of Ulp1 to
the transport channel of the nuclear
pore complex by karyopherins
Vikram Govind Panse*, Bernhard Küster†, Thomas Gerstberger* and Ed Hurt*‡
*BZH, Im Neuenheimer Feld 328, D-69120 Heidelberg, Germany
†Cellzome AG, Meyerhofstrasse 1, 69117 Heidelberg, Germany
‡e-mail: [email protected]
Published online: 9 December 2002; DOI: 10.1038/ncb893
The ubiquitin-like protein SUMO-1 (small ubiquitin-related modifier 1) is covalently attached to substrate proteins by
ligases and cleaved by isopeptidases. Yeast has two SUMO-1-deconjugating enzymes, Ulp1 and Ulp2, which are
located at nuclear pores and in the nucleoplasm, respectively. Here we show that the catalytic C-domain of Ulp1
must be excluded from the nucleoplasm for cell viability. This is achieved by the noncatalytic N-domain, which tethers Ulp1 to the nuclear pores. The bulk of cellular Ulp1 is not associated with nucleoporins but instead associates
with three karyopherins (Pse1, Kap95 and Kap60), in a complex that is not dissociated by RanGTP in vitro. The Ulp1
N-domain has two distinct binding sites for Pse1 and Kap95/Kap60, both of which are required for anchoring to the
nuclear pore complex. We propose that Ulp1 is tethered to the nuclear pores by a Ran-insensitive interaction with
karyopherins associated with nucleoporins. This location could allow Ulp1 to remove SUMO-1 from sumoylated
cargo proteins during their passage through the nuclear pore channel.
biquitin and ubiquitin-like proteins modulate protein function in the cell through reversible post-translational modification. This post-translational, conjugated state of cellular
proteins has been linked to several cellular regulatory pathways
including cell-cycle progression, apoptosis, differentiation, intracellular targeting and responses to stress1. One such ubiquitin-like
protein is SUMO-1. Similar to ubiquitin, the mechanism of attachment of SUMO-1 to substrate proteins at internal lysine residues is
carried out by specific E1 and E2 enzymes2,3. However, unlike ubiquitination, which targets proteins for proteosome-dependant
degradation, sumoylation modulates the intracellular location,
protein–protein interactions and stability of target proteins4,5.
Characterized substrates for SUMO-1 (called Smt3 in yeast) are
septins, which are present at the bud neck before cytokinesis6. It is
expected, however, that many SUMO-1-carrying substrates exist in
yeast and metazoans. The first identified SUMO-1 substrate is
mammalian RanGAP, the GTPase-activating protein of the nuclear
Ran GTPase, which regulates nucleocytoplasmic transport7,8. Only
the modified form of RanGAP is stably associated with RanBP2, a
protein located at the cytoplasmic face of the nuclear pore complex
(NPC)8. RanBP2 has been shown to promote sumoylation of target
proteins at the nuclear envelope and this has E3-like ligase activity9.
Not much is known about SUMO-1 targets in yeast.
Saccharomyces cerevisiae has two SUMO-1-deconjugating enzymes,
Ulp1, which has a nuclear pore association, and Ulp2, which is located in the nucleus10,11. The significance of having a distinct spatial distribution of Ulp1 and Ulp2 in the cell is not understood, but these
two enzymes may trigger the compartment-specific desumoylation
of target proteins. Here we have investigated how Ulp1 is targeted to
the nuclear pores and the consequences of such an NPC association.
U
Results
Nuclear location of the Ulp1 C-domain is dominant lethal. We first
sought to define the NPC-targeting domain in Ulp1. This protein
consists of a nonconserved N-domain (residues 1–403) and a conserved catalytic C-domain (residues 404–621; ref. 12 and Fig. 1a).
Notably, the expression of only the catalytic C-domain of Ulp1
(Ulp1C) causes a dominant-negative phenotype in yeast12. When
the catalytic domain is inactivated by site-specific mutation, however, Ulp1C is no longer toxic12. This therefore suggested that the
deconjugating catalytic activity of the C-domain is regulated by the
adjacent N-domain.
To analyse the mechanism regulating the intracellular targeting
of Ulp1, we fused the green fluorescent protein (GFP) to Ulp1 and
its various domains. Ulp1–GFP was functional and complemented
the nonviable ulp1 deletion strain (Fig. 1b, middle). Similar to
authentic Ulp1 (ref. 11), Ulp1–GFP showed a nuclear pore association (Fig. 1b, left) and co-clustered with NPCs in the nup133∆
mutant (Fig. 1b, right). When GFP-tagged Ulp1C was expressed
constitutively, it showed a dominant-negative growth phenotype
(data not shown). We therefore placed Ulp1C–GFP under control
of the GAL1 promoter. In glucose medium, GAL1::Ulp1C–GFP
expression was repressed and cells were viable (Fig. 1c). When
GAL1::Ulp1C–GFP expression was induced in galactose medium,
however, the cells stopped growing, often with an increased cell size
(see also ref. 10), and Ulp1C–GFP was concentrated in the nucleus
with a weaker cytoplasmic staining (Fig. 1c). We conclude that a
nuclear or a cytoplasmic location of Ulp1C, or both, may be
responsible for the lethal phenotype.
To distinguish between these possibilities, we attached a nuclear
localization sequence (NLS; from Rlp25)13, a nuclear export
sequence (NES; from PKI)13 or a cytoplasmic retention domain
(from Arc1)14 to Ulp1C–GFP. We found that Ulp1C–GFP–NLS had
an exclusively nuclear location, but it still exerted a dominant-negative lethal phenotype (Fig. 1d). By contrast, Ulp1C–GFP–Arc1 and
Ulp1C–GFP–NES, which were located in the cytoplasm, were not
toxic and could complement the lethal phenotype of a ulp1 knockout strain (Fig. 1e, f). Thus, a nuclear location of Ulp1C is toxic,
whereas a cytoplasmic distribution of this SUMO-1-deconjugating
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
1
articles
a
b
N-domain
Ulp1–GFP Nomarski Confocal
C-domain
Ulp1–GFP Nomarski
Ulp1
Pse1
1
Kap60/Kap95 Coiled coil
150
NPC targeting
340
Ulp1
SUMO deconjugation
403
Ulp1–GFP
621
ulp1∆
c
pGAL::Ulp1C–GFP
pUN100
d
Glucose
pGAL::Ulp1C–GFP–NLS
Galactose
Galactose
pGAL1
pGAL1::Ulp1C–GFP
pGAL1::Ulp1C–GFP–NLS
f
Glucose
Glucose
pGAL1
e
pGAL::Ulp1C–GFP–Arc1
nup133∆
5-FOA
Galactose
Ulp1C–GFP–NES
pGAL1
Ulp1–GFP
pGAL1::Ulp1C–GFP–Arc1
Ulp1C–GFP–NES
pUN100
cc-Ulp1C–GFP
cc-Ulp1C–GFP
Ulp1C–GFP–Arc1
5-FOA
Figure 1 Nuclear location of Ulp1 is dominant lethal. a, Domain organization of
Ulp1. The Ulp1N contains a Pse1-binding sequence (residues 1–150), a Kap60- or
Kap95-binding sequence (150–340) and a coiled-coil domain with NES activity
(340–403). The NPC-targeting domain between residues 1 and 340 is also indicated. b, In vivo location of Ulp1–GFP in ulp1∆ (left) and nup133∆ (right) strains,
analysed by conventional and confocal fluorescence microscopy. For the complementation assays, Ulp1–GFP in pUN100 or empty pUN100 was transformed into
the Ulp1 shuffle strain and growth was measured by dot-spot analysis on 5-FOA
plates after 4 d at 30 °C (middle). c–e, Expression of Ulp1 domains in yeast and
effect on subcellular location and growth. The indicated constructs were analysed
for subcellular location by fluorescence microscopy (left) and for growth on galactose, glucose and 5-FOA-containing plates (right). pGAL1 denotes the GAL1 promoter. Inductions in galactose were carried out for 1 h in wild-type yeast cells and
growth was followed at 30 °C for 6 d. f, Location of Ulp1C–GFP–NES and ccUlp1C–GFP in the ulp1∆ strain. Plasmid containing Ulp1–GFP, Ulp1C–GFP–NES, ccUlp1C–GFP or Ulp1C–GFP–Arc1, or empty pUN100 was transformed into the Ulp1
shuffle strain and 10−1 dilutions of cells were spotted onto 5-FOA plates. Growth
was analysed after 4 d at 30 °C.
enzyme is functional and not toxic to the cells. Apparently, Ulp1
can also carry out its essential role in the cytoplasm, suggesting that
a strict nuclear pore location is not required for the Ulp1 function
(see below).
Two NPC-targeting domains in the N-domain mediate tethering.
It was possible that Ulp1 displaced from the nuclear pores into the
nucleoplasm might trigger (premature) de-sumoylation of nuclear
substrate proteins and thus could be the cause for the toxicity.
Consistent with this notion, functionally active GFP–Smt3 (yeast
SUMO-1) was concentrated in the nucleus but could also be
detected in the cytoplasm (Fig. 2a, top). Only during mitosis did
GFP–Smt3 transiently label the septins (Fig. 2a, bottom), which are
present in the bud neck before cytokinesis6. Thus, SUMO-1 and
sumoylated proteins are predominantly located in the nucleus of
yeast cells.
We next sought to test how the N-domain of Ulp1 (Ulp1N)
tethers UlpC to the nuclear pores and thereby keeps it out of the
nucleoplasm. Ulp1N–GFP (see Fig. 1a) was located in the nuclear
pores, very similarly to full-length Ulp1 (Fig. 2b). Further analysis
of truncation mutants showed that Ulp1N contains two distinct
nuclear pore targeting sequences, one in the first (residues 1–150)
and one in the second (residues 150–403) half of the domain
(Fig. 2b). The separated individual NPC-targeting domains were
less efficient at targeting, however, as increased cytoplasmic stain-
ing was observed with these constructs. Shorter Ulp1∆N constructs
(for example, residues 1–100) were not active in NPC targeting
(Fig. 2b).
We also analysed the subcellular location of N-terminal deletion
mutants of full-length Ulp1. A truncation mutant lacking the first
150 residues (Ulp1∆150−621–GFP) showed a nuclear envelope location, but was partially mislocalized to the cytoplasm (Fig. 2b). A
similar distribution was observed for a Ulp1 mutant lacking the
second half of Ulp1N (Ulp1∆150–340–GFP; Fig. 2b). Taken together,
these data show that two distinct NPC-targeting domains in Ulp1N
are required to mediate efficient NPC tethering.
Ulp1N can be replaced by a nuclear pore targeting sequence. We
considered that if the only function of Ulp1N is to exclude Ulp1C
from the nucleoplasm and to anchor it at the nuclear pores, it
should be possible to replace Ulp1N by a heterologous nuclear pore
targeting sequence and retain function. We therefore fused to
Ulp1C–GFP different nucleoporins (Nsp1C, Nup42 and Nup60)
that are known to be targeted to distinct nuclear pore structures.
Whereas Nsp1 is found on both faces of the NPC, Nup42 is located
exclusively at the cytoplasmic side, and Nup60 is located at the
nuclear side of the NPC15.
Notably,
all
three
constructs,
Ulp1–GFP–Nsp1C,
Ulp1–GFP–Nup42 and Ulp1–GFP–Nup60, showed a nuclear pore
location (Fig. 3a) and did not cause a dominant-negative growth
2
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
articles
a
a
GFP–Smt3
Nomarski
5-FOA
Ulp1C–GFP–Nsp1C
Confocal
Ulp1C–GFP–Nup42
Confocal
Ulp1C–GFP–Nup60
Confocal
Smt3
pUN100
GFP–Smt3
b
Ulp11–403–GFP
Confocal
Ulp1150–403–GFP
Ulp1150–621–GFP
Ulp11–150–GFP
Ulp1∆150–340–GFP
Confocal
b
Ulp11–100–GFP
pGAL1
c
Ulp1
pUN100
pGAL1::Ulp1C–GFP
Ulp1C–GFP–Nup42
Figure 2 Ulp1N has nuclear pore targeting activity. a, In vivo location (left) and
complementation on 5-FOA plates of Smt3–GFP in smt3∆ cells (right). b, In vivo
location of Ulp11–403–GFP, Ulp1150–403–GFP, Ulp11–150–GFP and Ulp11–100–GFP,
expressed in wild-type yeast cells. Cells were analysed by conventional and, in
some cases, confocal fluorescence microscopy. Location of Ulp1150–621–GFP and
Ulp1∆150–340–GFP was analysed in the ulp1∆ strain.
phenotype (Fig. 3b; shown for Ulp1C–GFP–Nsp1C). In addition,
all three Ulp1C constructs fused to nucleoporins could complement the otherwise nonviable ulp1 disruption strain (Fig. 3c).
Thus, redirecting the catalytically active Ulp1C from the nucleus to
the nuclear pores through a functional nuclear pore targeting
domain is sufficient to restore Ulp1 function. We conclude that
Ulp1 can be localized at both faces of nuclear pores without significantly affecting the SUMO-1-deconjugation of substrate proteins.
Ulp1 associates with NPCs through karyopherins. Although these
data showed that Ulp1 was bound to the nuclear pores under steady
state, the mechanism of NPC association remained open. A yeast
two-hybrid assay has indicated that Nup42 and Gle1 bind to Ulp1
(ref. 16). Because the NPC association of Gle1 requires Nup42 (ref.
17), we tested whether the pore location of Ulp1–GFP location is
abolished in a Nup42 deletion strain. We found, however, that Ulp1
remains located exclusively at the nuclear pores with no trace of a
cytoplasmic mislocation in nup42∆ cells (Fig. 4). This suggests that
Ulp1 associates with NPCs by interacting with other components
of the NPC.
To identify the proteins that are associated with Ulp1 in vivo, we
used affinity chromatography to purify from yeast cell lysates Ulp1
tagged with Protein A (Ulp1–ProtA), which was functional and
could replace endogenous Ulp1 efficiently (data not shown).
pGAL1::Ulp1C–GFP–Nsp1C
Ulp1C–GFP–Nup60
Ulp1C–GFP–Nsp1C
Glucose
Galactose
5-FOA
Figure 3 Nucleoporins target Ulp1C to nuclear pores and restore Ulp1
function. a, Nuclear pore location of Ulp1C–GFP–Nsp1C, Ulp1C–GFP–Nup42 and
Ulp1C–GFP–Nup60 in yeast cells, analysed by conventional and confocal fluorescence microscopy. b, Growth of wild-type cells carrying empty pGAL1 (plasmid with
GAL1 promoter), GAL1::Ulp1C–GFP or GAL1::Ulp1C–GFP–Nsp1C constructs in glucose- or galactose-containing medium. Undiluted and 10−1 dilutions of cell suspensions were spotted on the indicated plates and grown at 30 °C for 6 d. c, Growth
of ulp1∆ cells carrying pURA3–ULP1 and transformed with a pUN100 plasmid
encoding Ulp1, Ulp1C–GFP–Nup42, Ulp1C–GFP–Nup60, Ulp1C–GFP–Nsp1C or no
insert (pUN100 plasmid). The growth of transformants on 5-FOA plates at 30 °C for
4 d was analysed after spotting 10−1 dilutions of cells.
Notably, purified Ulp1–ProtA was present in a complex with several
nuclear import receptors: Pse1 (Kap121), karyopherin-β (Kap95)
and karyopherin-α (Kap60; Fig. 5a, lane 1). Proteins that interact
with affinity-purified ProtA bait proteins can be released by treatment with increasing concentrations of MgCl2 (ref. 18). Accordingly,
Pse1, Kap95 and Kap60 were removed efficiently from Ulp1–ProtA
by treatment with 1 M MgCl2 (Fig. 5a, lane 2). In a following step,
Ulp1–ProtA was released from IgG–Sepharose by harsh elution with
acetic acid (Fig. 5a, lane 3).
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
3
articles
Ulp1–GFP
Nomarski
Ulp1–GFP
a
Nomarski
b
Pse1
120 -
Ulp1–ProtA
Ulp1–ProtA
Kap95
Kap60
Kap60
nup42∆
1
Figure 4 Ulp1 remains located at the nuclear envelope in nup42∆ cells.
Nuclear pore location of Ulp1–GFP in nup42∆ cells and its isogenic wild-type strain
NUP42 was visualized by fluorescence microscopy.
2
1 2 3 4 5 6
Western
3
Ulp1–ProtA
1
Mr(K)
60 Ulp1–ProtA
Pse1
120 -
We next determined how much of Ulp1 in the cell is in complex
with these three karyopherins (Fig. 5b). Equivalent amounts of
whole-cell lysate, soluble supernatant, insoluble pellet, flow
through from the column, and a onefold and tenfold equivalent of
the Ulp1–ProtA eluate from the IgG–Sepharose beads were
analysed by SDS polyacrylamide gel electrophoresis (SDS–PAGE)
and Coomassie staining (Fig. 5b, top) or western blotting to detect
Ulp1 (Fig. 5b, bottom). This indicated that the predominant pool
of Ulp1 present in the cell can be purified by IgG–Sepharose chromatography. Because purified Ulp1–ProtA contains the three
karyopherins in almost stoichiometrical amounts (Fig. 5a and b),
the bulk of cellular Ulp1 must be associated with these shuttling
nuclear transport receptors.
A hallmark of importin cargo complexes is that they can be dissociated on incubation with RanGTP19. But in vitro RanGTP could
not release Ulp1 from the three karyopherins (Fig. 5c, left). To verify that RanGTP used in this assay was active, we confirmed that
RanGTP binds to Yrb1 (ref. 20), a yeast RanGTP-binding protein
(Fig. 5c, right). The unexpected finding that RanGTP does not
release Ulp1 from karyopherins may point to an unconventional
interaction between Ulp1 and shuttling nuclear import receptors.
Because Ulp1N mediates the NPC association (see Fig. 2), we
next analysed whether this domain binds directly to karyopherins.
Notably, all three karyopherins, Pse1, Kap95 and Kap60, were firmly associated with affinity-purified Ulp1N1–403–ProtA or
Ulp1N1–250–ProtA (Fig. 5d, lanes 1 and 2). Because Ulp1N contains
two different domains for NPC association (see Fig. 2b), we
analysed whether Pse1, Kap95 and Kap60 bind to these distinct
regions. Notably, Ulp1∆1–150–ProtA bound to Kap60 and Kap95, but
not to Pse1 (Fig. 5d, lane 4). Vice versa, Ulp1 lacking the second
half of Ulp1N (Ulp1∆150–340) bound to Pse1 but not to Kap60 or
Kap95 (Fig. 5d, lane 3). Thus, two distinct binding sequences for
two different karyopherins are present in Ulp1N, both of which
constitute an efficient NPC-tethering domain in vivo.
Ulp1 localization is disrupted by mutants affecting karyopherins.
To obtain further evidence that in vivo the association of Ulp1 with
karyopherin transport receptors mediates the tethering of Ulp1 to
the NPCs, we used a dominant-negative Yrb4∆N (Kap123∆N)
mutant that, similar to other importin-β mutants lacking the Nterminal RanGTP-binding domain, can cause the dissociation of
karyopherins from the nuclear pores21. Notably, on expression of
Yrb4∆N, Ulp1 became significantly mislocalized to the cytoplasm,
but the association of nucleoporins (for example, Nic96) with
nuclear pores was not affected (Fig. 6a). In addition, an increased
association of Ulp1–GFP with the plasma membrane was observed
in Yrb4∆N-expressing cells, although the reasons for this are not
understood. Thus, the pore location of intact Ulp1 can be altered by
a dominant-negative importin-β-like transport receptor, which is
known to compete with other mobile transport receptors for binding to nucleoporins containing FG repeats.
In addition, we tested the subcellular distribution of Ulp1 in
the temperature-sensitive pse1-1 mutant. Notably, Ulp1–GFP
2 3 4 5 6
c
Mr(K)
4
Pse1
120 80 60 -
Kap95
60 -
NUP42
Coomassie
Mr(K)
Mr(K)
60 -
GST–Yrb1
Kap95
Kap60
Rna1
30 -
30 -
Ran
Ran
1
2
3
4
5
6
1
2
d
Mr(K)
120 -
Pse1
80 60 -
Kap95
Kap60
1 2 3 4
Figure 5 Ulp1N forms a RanGTP-insensitive complex with karyopherins.
a, Purified Ulp1–ProtA is associated with Pse1, Kap95 and Kap60. Whole-cell
lysate containing Ulp1–ProtA was passed over IgG–Sepharose. Shown are the
acetic acid eluate (lane 1), the 1 M MgCl2 eluate (lane 2), and the acetic acid eluate
after elution with MgCl2 (lane 3). Proteins were analysed by SDS–PAGE and bands
were identified by mass spectrometry. b, Equivalent amounts of whole-cell lysate
containing Ulp1–ProtA (lane 1), lysate supernatant (lane 2), insoluble pellet (lane 3),
flow through (lane 4) and Ulp1–ProtA acetic acid eluate (lane 5; lane 6 is a 10-fold
concentration) were analysed by SDS–PAGE (top) and western blotting (bottom).
c, Left, Kap95, Kap60 and Pse1 are not released from Ulp1–ProtA on treatment
with RanGTP. Purified Ulp1–ProtA was immobilized on IgG–Sepharose beads and
incubated with buffer (lanes 1 and 2), RanGTP (lanes 3 and 4) or RanGDP (lanes 5
and 6). Eluates were collected (lanes 1, 3 and 5) and then bound proteins (lanes
2, 4 and 6) were released by acetic acid and analysed by SDS–PAGE. Right,
RanGTP binds to Yrb1 (ref. 20). RanGTP (lane 1) or RanGDP (lane 2; generated by
incubation of RanGTP with Rna1) was incubated with glutathione S-transferase
(GST)-tagged Yrb1, which was immobilized on glutathione beads. Shown are the
bound fractions. d, Identification of Pse1- and Kap95/Kap60-binding sites in
Ulp1N. Whole-cell lysates containing Ulp1N1–403–ProtA (lane 1), Ulp1N1–250–ProtA
(lane 2), Ulp1N1–150–ProtA (lane 3) or Ulp1150–621–ProtA (lane 4) were purified over
IgG–Sepharose and the eluates were analysed by SDS–PAGE. The ProtA fusion
proteins are marked with an asterisk and the positions of Pse1, Kap95 and Kap60
are indicated.
significantly mislocalized to the cytoplasm when shifted to the
restrictive temperature (Fig. 6b). Western analysis showed that the
observed cytoplasmic mislocalization of Ulp1–GFP, either on
expression of dominant-negative Yrb4∆N or in the pse1-1 mutant
at restrictive temperature, was not due to enhanced quantities of
Ulp1–GFP (Supplementary Information, Fig. S1).
Ulp1 can replace the nuclear Ulp2-deconjugating function. So
far, these data showed that a nuclear location of the catalytic
Ulp1C domain is deleterious to the cell and that tethering of Ulp1
to the NPCs through karyopherins is the mechanism by which its
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
articles
a
Ulp1–GFP
Nomarski
Nic96–GFP
Nomarski
pGAL1
pGAL1::Yrb4∆N
23 °C
b
Ulp1–GFP
37 °C
c
Ulp1–GFP
pse1-1
Ulp2–GFP
pse1-1
d
23 ° C
37 °C
ulp2∆
+ pRS424
ulp2∆
+ pRS424–Ulp1N
Figure 6 Dissociation of Ulp1 from the nuclear pores. a, In vivo location of
Ulp1–GFP and Nic96–GFP in yeast expressing Yrb4∆N. Strains expressing
Ulp1–GFP and Nic96–GFP were transformed with pGAL::YRB4∆N or empty pGAL
and grown in raffinose-containing medium to an optical density of 0.5 at 600 nm
before Yrb4∆N expression was induced by 2% galactose. After 4 h, the location of
Ulp1–GFP and Nic96–GFP was analysed by fluorescence microscopy. b, In vivo
location of Ulp1–GFP in the pse1-1 temperature-sensitive mutant grown at 23 °C
and shifted to 37 °C (the restrictive temperature) for 3 h. c, Nuclear location of
Ulp2–GFP in ulp2∆ cells, visualized by fluorescence microscopy. d, ulp2∆ cells
were transformed with the high-copy plasmid pRS424 or pRS424Ulp1N–GFP.
Transformants were spotted onto selective SDC plates and incubated at 23 °C or
37 °C for 6 d.
deconjugating function is regulated. Notably, the second SUMO-1deconjugating enzyme in yeast, Ulp2, which has a catalytic domain
that is highly similar to Ulp1C, fulfils its role inside the nucleus11.
Accordingly, we found that GFP-tagged Ulp2 was located exclusively in the nucleus in living cells, and showed no nuclear envelope
staining (Fig. 6c).
It has been shown that a ulp2 knockout strain is viable, but shows
a temperature-sensitive phenotype at 37 °C that can be rescued by
the overproduction of Ulp1 (ref. 11). In addition, overproduction of
a catalytically inactive Ulp1 enzyme (see above) complements the
ulp2∆ temperature-sensitive phenotype11. From our data presented
here, it seemed possible that Ulp1N might be able to displace some
endogenous Ulp1 (an amount too low to be toxic) from the nuclear
pores into the nucleoplasm, thereby substituting for Ulp2 function.
If this assumption were true, then Ulp1N alone might be able to
rescue the ulp2∆ temperature-sensitive phenotype. This was indeed
observed (Fig. 6d). We therefore suggest that Ulp1 can replace the
nuclear Ulp2-deconjugating function when a small pool of Ulp1 is
displaced from the nuclear pores into the nucleoplasm.
Ulp1N also contains an NES signal. If Ulp1-mediated deconjugation
of SUMO-1 requires the strict elimination of this enzyme from the
nucleus, cells may want to keep the intranuclear pool of Ulp1 low.
This could be achieved, for example, by an auxiliary NES in Ulp1.
Notably, adjacent to its NPC-targeting domain Ulp1N contains a
leucine-rich region that is predicted to be a coiled coil (residues
341–403; see Fig. 1a). In addition, NES sequences that are recognized by the general NES receptor exportin-1 (also known as Xpo1)
are leucine-rich with a regular spacing of hydrophobic residues19.
To test whether the predicted coiled-coil sequence in Ulp1 has
NES activity, we left this part (referred to as ‘cc’) attached to Ulp1C
(cc-Ulp1C; residues 341–403) and tested for its intracellular location and complementation. These studies showed that expression
of cc-Ulp1C–GFP was not toxic and could rescue the otherwise
lethal ulp1∆ strain (Fig. 1f). In addition, cc-Ulp1C–GFP showed a
predominantly cytoplasmic location (Fig. 1f), differing from the
predominantly nuclear distribution of Ulp1C–GFP (see Fig. 1c).
To show directly that the coiled-coil part of Ulp1 has NES activity, we fused residues 340–403 (see Fig. 1a) to a mutant of Nmd3
lacking its essential NES domain (nmd3∆NES). Nmd3 is an adaptor protein for NES-mediated export of ribosomal 60S subunits
that requires the exportin-1 receptor13,22. When Nmd3 lacks its
NES, it is mislocalized to the nucleoplasm and cells are not viable
(refs 13, 22, and Fig. 7a). Notably, the coiled-coil region of Ulp1N
fused to nmd3∆NES (nmd3∆NES–GFP–cc) not only mediated
export to the cytoplasm, but also complemented the lethality of the
nmd3 knockout strain (Fig. 7a). In addition, nmd3∆NES–GFP–cc
expressed in the leptomycin (LMB)-sensitive xpo1-T539C mutant23
showed a distinct nuclear accumulation after the addition of LMB
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
5
articles
nmd3∆NES–GFP
Nmd3–GFP
nmd3∆NES–GFP–cc
Nmd3–GFP
nmd3∆NES–GFP
nmd3∆NES–GFP–cc
nmd3∆NES–GFP–Ulp1N
5-FOA
b
XPO1 + LMB
xpo1-T539C + LMB
Nmd3–GFP
nmd3∆NES–GFP–cc
Figure 7 The predicted coiled-coil domain in Ulp1N has NES activity. a, In
vivo location of nmd3∆NES–GFP, Nmd3–GFP and nmd3∆NES–GFP–cc in the
nmd3∆ strain analysed by fluorescence microscopy (top). Plasmids containing the
above constructs were transformed into the Nmd3 shuffle strain and transformants
were grown on 5-FOA plates at 30 °C for 4 d (bottom). b, In vivo location of
Nmd3–GFP and nmd3∆NES–GFP–cc in the LMB-sensitive xpo1-T539C mutant and
its isogenic wild-type (XPO1) strain, 20 min after the addition of LMB to the culture
medium. Signals from GFP fusion proteins were observed by fluorescence
microscopy.
(Fig. 7b). By contrast, the whole of Ulp1N (including the coiledcoil part) fused to nmd3∆NES did not restore Nmd3 function (Fig.
7a, 5-FOA).
To show that the NPC-anchoring activity in Ulp1N is dominant
over the NES activity in the coiled-coil domain, we analysed fulllength Ulp1–GFP in the LMB-sensitive xpo1 mutant strain.
Ulp1–GFP did not accumulate in the nucleus in presence of LMB
in this strain (data not shown). Thus, the nuclear pore targeting
activity of the N-terminal domain is dominant over the NES activity present in the coiled-coil domain. But the coiled-coil domain
could also have other functions (such as dimerization) in addition
to acting as a safety device to keep the catalytic domain out of the
nucleus in situations where some Ulp1 molecules escape the NPCtethering mechanism.
Discussion
We have shown that the location of the SUMO-1-deconjugating
enzyme Ulp1 at the nuclear pore is functionally important. If Ulp1
becomes mislocalized to the nucleoplasm, then cells are not viable.
By contrast, targeting Ulp1 to the cytoplasm does not impair its
essential function; therefore, a nuclear pore location is not strictly
required for Ulp1 function. We do not know the reason for the toxicity that arises when Ulp1 is present in the nucleoplasm, but the
enzyme may deconjugate SUMO-1 from intranuclear substrate proteins, which would be harmful in this compartment. To keep Ulp1
out of the nucleoplasm, Ulp1N has an efficient nuclear pore targeting activity and also an NES located in the predicted coiled-coil
6
domain. Indeed, this coiled-coil domain is sufficient to abolish the
dominant-negative phenotype of Ulp1C. We cannot say, however,
whether this NES is active in the full-length protein. Previously, this
region was reported to be a regulatory element that modulates the
activity of Ulp1C12.
The reason why Ulp2 is not toxic despite its nucleoplasmic location is not known. Ulp2 is not essential and therefore may deconjugate only a few sumoylated substrate proteins. By contrast, Ulp1
is essential and thus seems to be a more promiscuous enzyme
whose activity must be regulated strictly. It is possible that Ulp1
desumoylates distinct nuclear substrate proteins that are imported
into the nucleus or exported from the nucleus, which is consistent
with its nuclear pore location. Thus, regulated SUMO-1 deconjugation of a set of nuclear substrate proteins (which is an essential
process) could be optimally achieved by tethering Ulp1 to the
nuclear pores.
Our other key finding is that Ulp1 becomes localized to the
nuclear pores by a mechanism involving unconventional NPC
anchoring. We have provided much evidence that Ulp1 associates
with the NPC through shuttling transport receptors of the karyopherin family. Unexpectedly, Ulp1 is bound to Kap95, Kap60 and
Pse1 in a tight complex that RanGTP cannot dissociate in vitro.
Although it is possible that Ulp1 interacts with certain nucleoporins18,
the strong binding of almost all cellular Ulp1 to karyopherin transport receptors is intriguing. The most likely explanation is that in vivo
the bulk of Ulp1 is associated with Pse1, Kap95 and Kap60, which
similar to Ulp1 have a predominantly NPC location under steady
state owing to their transient interaction with nucleoporins containing FG repeats. Thus, Ulp1 could be tethered to the active transport
channel of the NPC by piggybacking permanently to shuttling transport receptors and escaping release by RanGTP. Consistent with this
notion, deleting one of the two karyopherin-binding sites in Ulp1N
causes a partial mislocation of Ulp1 to the cytoplasm. Thus, the loss
of physical interaction of Ulp1 with either Pse1 or Kap95 and Kap60
is correlated with a diminished NPC association.
Why RanGTP does not release Ulp1 from the karyopherins is
not known. This insensitivity towards RanGTP suggests that Ulp1
bound to Kap95, Kap60 and Pse1 may not represent a classical
nuclear import intermediate. Our data have shown that Ulp1N
contains at least two different types of signal, which bind to two different karyopherins. Among the many potential schemes, one possibility is that on Ulp1N binding to two different import receptors
(Pse1 and Kap95), the Ran-binding domain of Pse1 and Kap95
could be shielded, causing a decreased sensitivity to RanGTP.
Alternatively, Ulp1 may bind to the import receptors in an
unknown mode that is not dissociated by RanGTP-induced conformational changes in importin-β family members. Further work
is required to distinguish between these possibilities.
Our work has shown that NPC anchoring is not a prerequisite
for the essential function of Ulp1 and that the removal of SUMO1 from, for example, export cargo, could occur in the cytoplasm.
But the strategic location of a deconjugating enzyme in the nuclear
pore channel, through which active transport occurs, would facilitate efficient removal of SUMO-1 from sumoylated export cargoes
during transit. Pse1, Kap60 and Kap95 are known to be concentrated at the nuclear pores in the process of active nucleocytoplasmic
transport. In that position, statistically, they could encounter the
bulk of cellular Ulp1.
The precise location of Ulp1 in the active transport channel is
not known. Ulp1 deficient in either of the karyopherin-binding
domains results in an enhanced cytoplasmic pool, suggesting that
there is a cooperative thermodynamic effect in the interaction of
the tandemly arranged karyopherin-binding sequences, showing
pore-tethering activity, with the transport receptors. This enhanced
cytoplasmic pool may point towards a certain cytoplasmic bias of
Ulp1 at the pore channel.
The process of both SUMO-1 conjugation and deconjugation is
conserved in evolution from yeast to humans. The Saccharomyces
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
articles
pombe and human homologues of Ulp1 have been reported to be
localized at the NPC, suggesting a conserved mechanism for regulated and restricted enzymatic activity at the NPC24,25. In the case of
higher eukaryotes, an association between Senp2, a mammalian Ulp1
homologue, with the FG-repeat domain of Nup153 has been reported25,26. Neither of these two studies showed, however, that Senp2
binds directly to Nup153, because the binding assays were based on
western analysis of Nup153 and were done in lysates containing
other proteins, including importin-β-like transport receptors.
Nup153 forms a tight complex with importin-β in vivo27, and it
remains to be shown whether mammalian Senp2 also binds directly
through its N-domain to receptors of the importin-β family.
The exact functional significance of excluding the SUMO-1deconjugating enzyme from the nucleus remains elusive owing to
the lack of substrates, possibly export cargoes, for this enzyme.
Further genetic and proteomic approaches should elucidate the
substrate specificity of the enzyme, which might shed light on the
reason why the Ulp1 enzyme is tethered to the NPC. Notably, the
SUMO-1 conjugation machinery has been linked to the NPC9,28:
the nucleoporin RanBP2 (Nup358) has SUMO-1 E3-like activity
and is therefore involved in SUMO-1 transfer from Ubc9 to
SUMO-1 targets. Thus, both SUMO-1 conjugation and deconjugation are linked to nucleocytoplasmic transport.
Methods
Yeast strains and molecular biology
The yeast strains and plasmids used in this study are given in the Supplementary Information. DNA
recombinant work, microbiological techniques, plasmid transformation, yeast growth in 5-FOA (fluoro-orotic acid)-, galactose- or glucose-containing medium, and yeast genetic work were done essentially as described29. In addition to standard yeast vectors, we used the following plasmids: pRS315NOP1::GFP and pRS315-NOP1::ProtA30.
Complementation assay
Where possible, gene constructs were tested for their functionality by complementation of the otherwise lethal or sick phenotype in the respective yeast disruption strains. The in vivo complementation
assay for NES-mediated export using the nmd3∆NES reporter construct was similar to described
assays13,22; however, we used a modified nmd3∆NES–GFP construct containing a new polylinkercloning site after the GFP gene insert, which will be published elsewhere (T.G. and E.H., manuscript in
preparation). For LMB treatment of the LMB-sensitive (xpo1-T539C) and LMB-resistant (XPO1) S.
cerevisiae strains, 100 ng ml−1 LMB was added to the culture medium23. Expression of dominant-negative GAL::YRB4∆N in yeast strains was done as described21.
Protein expression, purification and binding
Affinity purification of Ulp1–ProtA, Ulp1N–ProtA and the shorter Ulp1∆N–ProtA truncation constructs by IgG–Sepharose (Amersham) was done essentially as described31. Purified complexes were
dissociated in SDS sample buffer, resolved by SDS–PAGE and stained with Coomassie blue. We carried
out mass spectrometry as described29.
Incubation of Ulp1–ProtA with bound Kap60, Kap95 and Pse1 (immobilized on IgG–Sepharose
beads) with RanGTP or RanGDP was done as described20. RanGTP was tested and found to be functionally active as described20. We incubated affinity-purified Ulp1–ProtA bound to IgG–Sepharose
beads with increasing concentrations of MgCl2, and eluted it in acetic acid as described18.
Microscopy
For fluorescence microscopy, yeast strains were grown in liquid medium to an optical density at 600
nm of 0.5 before the cells were mounted on a microscopic slide and inspected with a conventional
(Zeiss Axioscope; Zeiss, Jena, Germany) or confocal fluorescence (Leica TCS MP, Bensheim, Germany)
microscope.
RECEIVED 25 JUNE 2002, REVISED 13 SEPTEMBER 2002, ACCEPTED 28 OCTOBER 2002,
PUBLISHED 9 DECEMBER 2002.
1. Laney, J. D. & Hochstrasser, M. Substrate targeting in the ubiquitin system. Cell 97, 427–430 (1999).
2. Johnson, E. S., Schwienhorst, I., Dohmen, R. J. & Blobel, G. The ubiquitin-like protein Smt3p is
activated for conjugation to other proteins by an Aos1p/Uba2p heterodimer. EMBO J. 16,
5509–5519 (1997).
3. Johnson, E. S. & Blobel, G. Ubc9p is the conjugating enzyme for the ubiquitin-like protein Smt3p.
J. Biol. Chem. 272, 26799–26802 (1997).
4. Hochstrasser, M. All in the ubiquitin family. Science 289, 563–564 (2000).
5. Melchior, F. SUMO-1 — nonclassical ubiquitin. Annu. Rev. Cell. Dev. Biol. 16, 591–626 (2000).
6. Johnson, E. S. & Blobel, G. Cell cycle-regulated attachment of the ubiquitin-related protein SUMO
to the yeast septins. J. Cell Biol. 147, 981–994 (1999).
7. Mahajan, R., Delphin, C., Guan, T., Gerace, L. & Melchior, F. A small ubiquitin-related polypeptide
involved in targeting RanGAP1 to nuclear pore complex protein RanBP2. Cell 88, 97–107 (1997).
8. Saitoh, H., Pu, R., Cavenagh, M. & Dasso, M. RanBP2 associates with Ubc9p and a modified form
of RanGAP1. Proc. Natl Acad. Sci. USA 94, 3736–3741 (1997).
9. Pichler, A., Gast, A., Seeler, J. S., Dejean, A. & Melchior, F. The nucleoporin RanBP2 has SUMO1 E3
ligase activity. Cell 108, 109–120 (2002).
10. Li, S. J. & Hochstrasser, M. A new protease required for cell-cycle progression in yeast. Nature 398,
246–251 (1999).
11. Li, S. J. & Hochstrasser, M. The yeast ULP2 (SMT4) gene encodes a novel protease specific for the
ubiquitin-like Smt3 protein. Mol. Cell. Biol. 20, 2367–2377 (2000).
12. Mossessova, E. & Lima, C. D. Ulp1-SUMO crystal structure and genetic analysis reveal conserved
interactions and a regulatory element essential for cell growth in yeast. Mol. Cell 5, 865–876 (2000).
13. Gadal, O. et al. Nuclear export of 60S ribosomal subunits depends on Xpo1p and requires a nuclear
export sequence-containing factor, Nmd3p, that associates with the large subunit protein Rpl10p.
Mol. Cell. Biol. 21, 3405–3415 (2001).
14. Galani, K., Grosshans, H., Deinert, K., Hurt, E. C. & Simos, G. The intracellular location of two
aminoacyl-tRNA synthetases depends on complex formation with Arc1p. EMBO J. 23, 6889–6898
(2001).
15. Rout, M. P. et al. The yeast nuclear pore complex: composition, architecture and transport mechanism. J. Cell Biol. 148, 635–651 (2000).
16. Takahashi, Y., Mizoi, J., Toh-E, A. & Kikuchi, Y. Yeast Ulp1, an Smt3-specific protease, associates
with nucleoporins. J. Biochem. (Tokyo) 128, 723–725 (2000).
17. Strahm, Y. et al. The RNA export factor Gle1p is located on the cytoplasmic fibrils of the NPC and
physically interacts with the FG-nucleoporin Rip1p, the DEAD-box protein Rat8p/Dbp5p and a
new protein Ymr255p. EMBO J. 18, 5761–5777 (1999).
18. Simos, G. et al. The yeast protein Arc1p binds to tRNA and functions as a cofactor for the
methionyl- and glutamyl-tRNA synthetases. EMBO J. 15, 5437–5448 (1996).
19. Görlich, D. & Kutay, U. Transport between the cell nucleus and the cytoplasm. Annu. Rev. Cell. Dev.
Biol. 15, 607–660 (1999).
20. Künzler, M., Gerstberger, T., Stutz, F., Bischoff, F. R. & Hurt. E. Yeast Ran-binding protein 1 (Yrb1)
shuttles between the nucleus and cytoplasm and is exported from the nucleus via a CRM1 (XPO1)dependent pathway. Mol. Cell. Biol. 20, 4295–4308 (2000).
21. Schlenstedt, G. et al. Yrb4p, a yeast Ran-GTP-binding protein involved in import of ribosomal protein L25 into the nucleus. EMBO J. 16, 6237–6249 (1997).
22. Ho, J. H., Kallstrom, G. & Johnson, A. W. Nmd3p is a Crm1p-dependent adapter protein for
nuclear export of the large ribosomal subunit. J. Cell Biol. 151, 1057–1066 (2000).
23. Neville, M. & Rosbash, M. The NES-Crm1p export pathway is not a major mRNA export route in
Saccharomyces cerevisiae. EMBO J. 18, 3746–3756 (1999).
24. Taylor, D. L., Ho, J. C., Oliver, A. & Watts, F. Z. Cell-cycle-dependent localisation of Ulp1, a
Schizosaccharomyces pombe Pmt3 (SUMO)-specific protease. J. Cell Sci. 115, 1113–1122 (2002).
25. Hang, J. & Dasso, M. Association of the human SUMO-1 protease SENP2 with the nuclear pore. J.
Biol. Chem. 277, 19961–19966 (2002).
26. Zhang, H., Saitoh, H. & Matunis, M. J. Enzymes of the SUMO modification pathway localize to filaments of the nuclear pore complex. Mol. Cell. Biol. 22, 6498–6508 (2002).
27. Shah, S., Tugendreich, S. & Forbes, D. Major binding sites for the nuclear import receptor are the internal nucleoporin Nup153 and the adjacent nuclear filament protein Tpr. J. Cell Biol. 141, 31–49 (1998).
28. Kirsh, O. et al. The SUMO E3 ligase RanBP2 promotes modification of the HDAC4 deacetylase.
EMBO J. 21, 2682–2691 (2002).
29. Bassler, J. et al. Identification of a 60S preribosomal particle that is closely linked to nuclear export.
Mol. Cell 8, 517–529 (2001).
30. Hellmuth, K. et al. Yeast Los1p has properties of an exportin-like nucleocytoplasmic transport factor for tRNA. Mol. Cell. Biol. 18, 6374–6386 (1998).
31. Siniossoglou, S. et al. Structure and assembly of the Nup84p complex. J. Cell Biol. 149, 41–54
(2000).
32. Seedorf, M. & Silver, P. A. Importin/karyopherin protein family members required for mRNA
export from the nucleus. Proc. Natl Acad. Sci. USA 94, 8590–8595 (1997).
ACKNOWLEDGEMENTS
E.H. was supported by grants from the Deutsche Forschungsgemeinschaft (Leibniz-Programm) and
Fonds der Chemischen Industrie. V.G.P. is the recipient of a long-term fellowship from the Human
Frontier Science Program.
Correspondence and requests for materials should be addressed to E.H.
Supplementary Information accompanies the paper on www.nature.com/naturecellbiology.
COMPETING FINANCIAL INTERESTS
The authors declare that they have no competing financial interests.
NATURE CELL BIOLOGY ADVANCE ONLINE PUBLICATION www.nature.com/naturecellbiology
© 2002 Nature Publishing Group
7