Background
The cullin/RING family of ubiquitin ligases comprises a
multifunctional set of enzymes controlling the stability
and activity of cell cycle regulators, transcription
factors, and signaling proteins (reviewed in ref. [ 1]).
Among the best studied examples of cullin/RING enzymes are
the so-called SCF complexes. SCF complexes share homologues
of the core components cullin 1 (CUL1), SKP1, and the RING
domain protein HRT1/RBX1/ROC1, which associate with
different F-box proteins and ubiquitin-conjugating enzymes
[ 1]. F-box proteins specifically bind substrates,
following their phosphorylation in response to activation
of various signaling pathways. Although substrate
phosphorylation is one major determinant of SCF-mediated
ubiquitylation [ 1], covalent modification of the CUL1
subunit through attachment of the ubiquitin-related peptide
NEDD8 also regulates SCF activity. NEDD8 modification
stimulates SCF-dependent substrate ubiquitylation
in vitro [ 2, 3, 4], and mutant CUL1
resistant to NEDD8 modification is defective
in vivo [ 5]. In addition, fission
yeast mutants deficient in the enzymes that attach Ned8p,
or in the
ned8 gene itself, are inviable [
5].
In addition to CUL1, the human genome contains at least
five other cullins, CUL 2, 3, 4A, 4B, and 5. This diversity
is partially recapitulated in
S. pombe , which encodes
pcu3 and
pcu4 , homologues of human CUL3 and
CUL4, for which there are no direct correlates in budding
yeast [ 6]. All human cullins interact with HRT1/RBX1/ROC1
[ 7], are modified by NEDD8 [ 8], and have ubiquitin ligase
activity
in vitro [ 9]. Similarly, the fission
yeast cullins Pcu1p and Pcu4p are neddylated, and
neddylation is critical for their function
in vivo . [ 5].
While it is clear that neddylation affects the activity
of cullin/RING complexes, mechanisms controlling this
modification are just emerging. A recent study by Lyapina
et al. uncovered a novel connection
between the COP9/signalosome (CSN) and cullin neddylation [
10]. CSN is a highly conserved multiprotein complex
consisting of eight subunits, which have been implicated in
a wide variety of regulatory processes, including, cell
cycle control [ 11, 12], signal transduction [ 13],
transcriptional activation [ 14], and plant
photomorphogenesis [ 15, 16]. To date, no distinct
biochemical activity has been identified that can reconcile
this diversity in CSN functions. The eight CSN subunits
show a one-to-one relationship to components of the 19S
proteasome lid complex [ 17, 18, 19] and, like
pcu3 and
pcu4 , at least seven of the eight
human subunits are conserved in fission yeast, but not in
budding yeast. [ 16, 20].
Lyapina
et al. recently showed that
disruption of
S. pombe CSN subunit 1 (
caa1/csn1 ) results in accumulation
of Pcu1p in the neddylated state [ 10]. However, it
remained unclear whether this reflects global control of
multiple cullins by Csn1p as part of the entire CSN
complex. We have examined this question by determining the
effect of disruption of several CSN subunits on Pcu3p
neddylation, localization, and ubiquitin ligase activity.
Our results indicate that the entire CSN complex is
required for proper control of the neddylation state of
multiple cullins.
Results
Pcu3p is modified by Ned8p
Immunoblotting of protein lysates prepared from
S. pombe cells, whose
pcu3 chromosomal locus was modified
to encode Pcu3p with thirteen copies of the Myc epitope
attached to its C-terminus, revealed two distinct species
of Pcu3p, which migrated on SDS gels as two closely
spaced species of roughly equal abundance (Fig. 1A).
Several findings indicate that the slower migrating form
of Pcu3p arose from covalent modification with the
ubiquitin-related protein Ned8p: First, overexpression of
a fusion protein of Ned8p and glutathione transferase
(GST-Ned8p) converted the faster, unmodified form of
Pcu3p into a high molecular weight species that reacted
with antibodies to GST (Fig. 1A). Second, replacing
lysine 729, which is homologous to the conserved
neddylation sites in other cullins, with arginine,
completely abolished the slower migrating, modified form
of Pcu3p (Fig. 1B). In contrast, mutation of a
neighboring conserved lysine (residue 760) did not alter
the mobility of Pcu3p. Thus, as with other
S. pombe cullins [ 5], Pcu3p
undergoes covalent modification by ubiquitin-like Ned8p
at lysine 729.
Multiple csnmutants accumulate neddylated
Pcu3p
Interestingly, fission yeast
caa1/csn1 deletion mutants show a
set of defects, which partially overlaps that found in
pcu3 disruptants: slow growth, cell
elongation, and sensitivity to UV-irradiation, but not
hydroxyurea [ 6, 21]. With respect to the recent
implication of CSN in cullin regulation [ 10], we
considered the possibility that the overlapping
phenotypes of
csn1 and
pcu3 mutants could arise from
interference with Pcu3p function.
Initial support for this hypothesis was provided by
the observation that disruption of
csn1 led to accumulation of Pcu3p
in the slower migrating form (Fig. 2A). This shift in
migration is due to Pcu3p neddylation as shown by
overexpression of GST-Ned8p in
csn1 mutants. This caused an
additional supershift, converting a substantial portion
of the slower migrating form into a GST-Ned8p modified
species (Fig. 2B). Importantly, accumulation of
neddylated Pcu3p was also apparent in strains deleted for
csn3 ,
csn4 , and
csn5 (Fig. 2A), three genes
encoding other putative subunits of the
S. pombe CSN, which co-fractionate
with Csn1p upon size exclusion chromatography (see below,
Fig. 4). However, unlike
csn1 disruptants, which have a
defect in S phase control resulting in slow growth and
cell elongation [ 21], deletion of
csn3 ,
csn4 , and
csn5 genes did not cause cell
elongation, S phase delay, or any other obvious growth
defects (Fig. 2C, 2D, data not shown). This indicates
that the described phenotype of
csn1 mutants [ 21] is unlikely to
be a consequence of interference with Pcu3p activity.
Pcu1p, another fission yeast cullin also accumulated in
the neddylated form in all
csn mutants [ 10] (Fig. 2A),
suggesting that CSN has a general role in the control of
cullin neddylation in fission yeast. The accumulation of
neddylated Pcu3p was prevented in all
csn mutants by providing the
missing CSN subunit on a plasmid (Fig. 3).
The conserved "cysteine box" in Csn5p is not
required for Pcu3p deneddylation
Lyapina
et al. recently demonstrated that
partially purified CSN promotes the cleavage of
Pcu1p-Ned8p conjugates
in vitro [ 10], but it remained
unclear whether the enzymatic activity is contained
within one of the CSN subunits or a tightly associated
peptidase. All peptidases known to cleave ubiquitin- or
SUMO-protein conjugates are cysteine-based proteases
(reviewed in ref. [ 22]). Csn5p contains a motif closely
resembling the conserved cysteine box previously
recognized in the catalytic center of deubiquitylating
enzymes and in the budding yeast proteasome subunit
Rpn11p [ 23]. The complementation assay shown in Fig.
3allowed us to determine the potential role of this motif
in the control of Pcu3p neddylation by replacing the
critical cysteine residue with alanine. This mutant Csn5p
retained full activity for preventing Pcu1p and Pcu3p
hyperneddylation (Fig. 3Alanes 12-15, and data not
shown).
Pcu3p interacts with CSN
We next determined whether the effect of CSN
deficiency on Pcu3p neddylation reflects a physical
interaction of these proteins. Size exclusion
chromatography of lysates prepared from cells harboring
epitope-tagged Pcu3p or CSN subunits at the endogenous
genomic loci, revealed precise co-elution of Pcu3p with
CSN subunits 1, 3, 4, and 5 in high molecular weight
fractions corresponding to approximately 550 kDa (Fig.
4A). Interestingly, Csn5p, unlike the other CSN subunits,
showed a second peak in gel filtration eluting with an
approximate molecular weight of 200 kDa, suggesting that
Csn5p may also be involved in protein complexes separate
from the entire CSN (Fig. 4B).
Based on the co-elution of Pcu3p with CSN subunits in
gel filtration, we determined whether Pcu3p and Csn1p
interact
in vivo at endogenous expression
levels. Pcu3-Myc immunoprecipitated from lysates of cells
expressing HA-tagged Csn1p together with Pcu3p-Myc from
the endogenous promoters revealed co-precipitation of
Csn1p-HA. This interaction was specific as it was not
observed with an irrelevant antibody (Fig. 4Cleft panel)
or with lysate from a strain containing untagged Pcu3p
(Fig. 4C, right panel). In contrast, immunoprecipitation
with HA antibodies failed to precipitate both Csn1p-3HA
and Pcu3p-Myc, potentially due to inaccessibility of the
C-terminal 3HA tag of Csn1p within the CSN complex (data
not shown).
We therefore generated Pcu3p-Myc containing strains
overexpressing N-terminally 6xhistidine-Myc-tagged Csn1p,
Csn3p, Csn4p, or Csn5p. Binding of each of the four
different CSN subunits to nickel beads resulted in
co-purification of Pcu3p, indicative of specific
in vivo interactions (Fig. 4C).
Notably, both modified and unmodified Pcu3p interacted
with CSN subunits (Fig. 4C). We conclude that, at least a
fraction of Pcu3p forms a stable complex with CSN
in vivo .
Pcu3p localization in csnmutants
In
A. thaliana , CSN was shown to be
required for the efficient nuclear accumulation of the
putative RING-type ubiquitin ligase COP1 [ 24], which
mediates ubiquitin-dependent degradation of the
photomorphogenic transcription factor HY5 [ 25]. We
therefore considered the possibility that accumulation of
neddylated Pcu3p in
csn mutants is a consequence of
defective subcellular targeting. Indirect
immunofluorescence staining indicated that endogenous
Myc-tagged Pcu3p is present in the cytoplasm, but
enriched in the nucleus of wild-type cells (Fig. 5). This
localization pattern was fully preserved in all four
csn mutants, indicating that CSN
does not regulate the subcellular localization of
Pcu3p.
Mutation of CSN increases Pcu3p-dependent
polyubiquitylation activity
All human cullins have been shown to assemble into
complexes possessing polyubiquitylation activity
in vitro [ 9, 26]. To determine
whether Pcu3p is associated with such an activity and how
it is affected by CSN, we developed a
substrate-independent
in vitro polyubiquitylation assay,
similar to that originally reported by Lyapina et al. [
27]. As Pcu3p-associated ubiquitin ligase activity was
not described before, we first needed to determine which
of the fourteen different ubiquitin-conjugating enzymes
(UBCs) present in the fission yeast genome associate with
the RING domain protein Pip1p, an essential cofactor of
cullin ubiquitin ligases [ 7, 28, 29, 30]. Binding assays
showed that overexpressed HA-tagged Pip1p specifically
associated with co-overexpressed 6xHis-Myc-tagged Ubc1p,
Ubc7p, Ubc11p, Ubc12p, Ubc13p, Ubc14p, and Mms2p. In the
next step, we confirmed, by immunoprecipitation, that
both forms of Pcu3p-Myc bound endogenous Pip1p in a
CSN-independent manner (Fig. 6B).
Pcu3p-Myc complexes prepared by immunoprecipitation
with Myc antibodies were then incubated with human E1,
ATP, ubiquitin, and the recombinant UBCs that interacted
with Pip1p. Polyubiquitylated species were visualized by
anti-ubiquitin blot. Pcu3p-Myc complexes isolated from
pcu3-13myc cells did not show any
recognizable polyubiquitylation activity above background
with any of the seven UBCs tested (Fig. 6C, lane 2, and
data not shown). In contrast, the ubiquitylation assay
with Pcu3p-Myc complexes purified from
csn5 mutants led to efficient
formation of high molecular weight species reactive with
ubiquitin antibodies (Fig. 6C). This activity was
specific for Ubc7p (Fig. 6C, compare lanes 3 - 7).
Appearance of multiubiquitin chains was inhibited by
mutant ubiquitin lacking all lysines, indicating that
they were specifically generated in the reaction rather
than co-precipitated with Pcu3p-Myc (Fig. 6D). Thus, CSN
deficiency appears to stimulate Pcu3p-associated
ubiquitin ligase activity
in vitro . A similar stimulation
was obtained for substrate-specific polyubiquitylation of
phosphorylated Rum1p by SCF Poppurified from
csn5 mutants (unpublished
observation).
Discussion
Our data confirm and extend the recently discovered
connection between CSN and cullin regulation [ 10] by
showing that deletion of four putative CSN subunits causes
accumulation of Pcu3p in the neddylated state (Fig. 2A).
Co-elution of these four proteins in a 550 kDa complex
(Fig. 4B) strongly suggests that they represent part of the
S. pombe equivalent of the eight
subunit CSN complex found in higher eukaryotes [ 18, 19,
20]. Thus, cullin regulation through modulation of the
balance of Ned8p modification appears to require the entire
CSN complex.
Our study also provides some insight into the role of
deneddylation in cullin function, as we were able to create
a situation in which Pcu1p and Pcu3p are fully neddylated
in vivo (ref [ 10] and data presented
here). Deletion of
pcu1 and
pcu3 results either in lethality or
in severe growth retardation, respectively [ 6]. The normal
growth behavior of
csn3 ,
4 , and
5 mutants (Fig. 2C, 2D), where both
cullins seem constitutively neddylated, argues against a
critical role for dynamic cycles of neddylation and
deneddylation in cullin function in fission yeast.
Consistent with this notion, the subcellular localization
of Pcu3p and Pcu1p were unaffected by mutations in CSN
(Fig. 5) [ 10]. Instead, neddylated Pcu3p and Pcu1p
purified from
csn mutant cell extract exhibit
increased ubiquitin ligase activity
in vitro (Fig. 6B) [ 10]. Thus, CSN
may exert negative control on Pcu1p and Pcu3p by promoting
their deneddylation. However, while mutation of
csn5 caused a modest two-fold
increase in the steady state fraction of neddylated Pcu3p
(Fig. 2A), Pcu3p-associated ubiquitin ligase activity
recovered from this mutant showed a much more dramatic
increase (Fig. 6C). This suggests the possibility that CSN
represses Pcu3p activity by other mechanisms, in addition
to its control of neddylation. Further experiments are
required to determine whether recombinant Csn5p or purified
CSN complex can inhibit Pcu3p-associated ubiquitin ligase
activity.
Lyapina
et al. recently provided evidence for
a cullin deneddylating activity of CSN by demonstrating
that a partially purified preparation of the pig CSN can
deneddylate the fission yeast cullin Pcu1p
in vitro [ 10]. The results presented
here eliminate the attractive hypothesis that the cysteine
box in Csn5p, which is conserved in numerous
deubiquitylating enzymes [ 23], harbors a critical activity
for deneddylation (Fig. 3A). In addition, recombinant Csn5p
failed to deneddylate Pcu1p and Pcu3p
in vitro (unpublished observation).
Our finding that hyperneddylated Pcu3p and Pcu1p accumulate
in all
csn mutants tested (Fig. 2A, and data
not shown), also suggests that control of cullin
neddylation results from cooperation of multiple CSN
components rather than from the activity of a single
catalytic subunit. The CSN complex may, for example, serve
as a platform for recruitment of a separate deneddylating
enzyme. A human NEDD8-processing isopeptidase has been
described previously [ 31]. The potential role of CSN as an
organizing center for certain enzymatic activities may not
be restricted to the control of cullin neddylation, as CSN
was also shown to mediate phosphorylation of c-JUN, IkBa,
and p53 by a tightly associated kinase(s) [ 12, 19].
So far, we were unable to obtain deneddylation of
Pcu3p-13Myc
in vitro by partially purified CSN
preparations active in deneddylation of Pcu1p-13Myc (data
not shown). We can therefore not exclude that CSN's role in
the control of Pcu3p neddylation differs from its role
established for Pcu1p [ 10]. For example, binding by CSN
could protect Pcu3p from the Ned8p-conjugating enzyme
Ubc12p, thus maintaining a fraction of Pcu3p in a
deneddylated state. As this shield is experimentally
removed through deletion of CSN subunits, Pcu3p may
accumulate in the neddylated state. In such a scenario, no
enzymatic activity of purified CSN toward neddylated Pcu3p
would be expected
in vitro .
Finally, our data suggest CSN functions in addition to
cullin regulation that are not shared by all individual
subunits. Out of the four mutants tested, only
csn1 showed the previously described
defect in S phase progression [ 21] (Fig. 2C, 2D), thereby
underscoring the significance of our finding that all CSN
subunits tested here are involved in cullin regulation. It
is likely that some subunits have very specialized
functions within the complex, while others are required for
its broader activities such as control of cullin
neddylation. On the other hand, we found that a sizable
fraction of Csn5p is dissociated from the putative CSN
holocomplex, but enriched in a 200 kDa assembly (Fig. 4B).
It thus appears that part of the diversification of CSN
function may arise from formation of smaller
subcomplexes.
Conclusions
Our study reveals a general role of the
S. pombe CSN complex in the control
of cullin modification by Ned8p. Regardless of the exact
mechanism, CSN may regulate the activity of multiple cullin
ubiquitin ligases through counteracting their stimulatory
Ned8p modification. Future studies will reveal whether some
of the other described functions of CSN also involve
post-translational modification of regulatory proteins by
ubiquitin-like modifiers.
Materials and methods
Plasmids and yeast strains
S. pombe genes for UBCs and
csn3 ,
4 , and
5 were identified in the Sanger
Centre
S. pombe sequence database based on
their homology to the respective human and budding yeast
proteins. Complementary DNAs were amplified by PCR and
subcloned into pRep81.6xHis-Myc, pRep3.6xHis-Myc, or
pRep4.HA. Deletion strains and epitope-tagged strains
were constructed by one-step gene replacement using
PCR-generated fragments containing kanamycin or
ura4 cassettes [ 32]. Growth media,
flow cytometry, and all other relevant
S. pombe techniques were described
previously [ 33].
Immunological techniques
Protein lysates were prepared by bead lysis in a
Fastprep device (Bio 101) in the presence of proteinase
inhibitors, followed by boiling in SDS sample buffer.
Pip1p and Pcu1p were detected with affinity-purified
rabbit antisera. Epitope-tagged proteins were detected by
the monoclonal antibodies 9E10 and 12CA5 prepared from
hybridoma supernatants.
Cell lysates for immunoprecipitation were prepared by
disrupting cells in immunoprecipitation buffer (20 mM
Tris/HCl, pH 7.4; 150 mM NaCl; 0.5 % Triton X-100, 10
μg/ml leupeptin, 10 μg/ml pepstatin, 5 μg/ml aprotinin, 1
mM PMSF). Lysates were cleared and precipitated with the
respective antisera. Immunocomplexes were collected by
binding to protein A or G beads, washed and analyzed by
immunoblotting as described [ 33].
For indirect immunofluorescence staining, cells were
fixed in freshly prepared 3.7% para-formaldehyde for 1h
at 30°C. Cells were washed in PEM (100 mM Pipes pH6.9, 1
mM EGTA, 1 mM MgSO
4 ), and digested with Novozyme (0.5
mg/ml, Sigma) and Zymolyase (0.5 mg/ml, US Biological) in
PEM, 1M sorbitol. Unspecific epitopes were blocked by
incubation in PEM-BAL (PEM plus 1% BSA, 0.1% Na-azide,
0.1 M L-lysine HCl) for 1h. Incubation with monoclonal
Myc antibodies diluted in PEM-BAL was done overnight,
followed by three washes in PEM-BAL. An overnight
incubation with TRITC-labeled anti mouse antibodies was
followed by washing and mounting on polylysine coated
coverslips. Cell nuclei were counterstained with DAPI.
Photomicrographs were obtained with a CCD camera mounted
on a Nikon E600 epifluorescence microscope. Brightness
and contrast were adjusted in Adobe Photoshop 5.0.
In vitroubiquitylation assay
For ubiquitylation reactions, Pcu3p-Myc complexes were
immunoprecipitated from 100 - 200 ug total cell lysates
prepared as described above. Precipiates were washed four
times in 20 mM Tris/HCl, pH 7.4; 150 mM NaCl; 0.5 %
Triton X-100, 10 μg/ml leupeptin, 10 μg/ml pepstatin, 5
μg/ml aprotinin, 1 mM PMS, and equilibrated in 20 mM
HEPES, pH 7.4, 100 mM potassium acetate, 1 mM DTT. The
reaction was started by the addition of a cocktail of 8
μM ubiquitin, ATP regenerating system (2 mM HEPES at pH
7.4, 1 mM ATP, 30 mM creatine phosphate, 1 mM magnesium
acetate, 0.15 mg/ml creatine kinase), reaction buffer (4
mM magnesium acetate, 1 mM DTT), 500 nM bacterially
expressed 6xHis-UBCs, 100 nM 6xHis-tagged human E1, and
0.5 μM ubiquitin aldehyde in a volume of 15 μl. After 90
min at 30°C, the reaction was terminated by the addition
5 × SDS sample buffer. Samples were separated on 11 %
SDS-polyacrylamide gels and analyzed by immunoblotting
with ubiquitin antibodies (Zymed).
Abbreviations
CSN: COP9/signalosome
GST: Glutathione S-transferase
HA: hemagglutinin
SCF: SKP1/Cullin/F-box protein complex
UBC: Ubiquitin-conjugating enzyme
