Dr. Lakshmi Goyal
Cell Editorial Office
1100 Massachusetts Avenue
Cambridge, MA 02138
Dieter A. Wolf, M.D.
December 18, 2002
Re.: Manuscript C0209044
Dear Dr. Goyal,
Enclosed please find a revised version of our manuscript “Fission yeast COP9/signalosome
suppresses cullin activity through recruitment of the deubiquitylating enzyme Ubp12p”.
In the attached reply letter, we have discussed the reviewers’ criticism point-by-point. In
addition, we have included thoroughly controlled new data in Fig. 6, which we believe
substantially strengthen our model for the role of CSN/Ubp12p in cullin ubiquitin ligase
assembly. We hope that the revised manuscript will be acceptable for publication.
Regarding the decision whether this manuscript should be published in Cell or Molecular
Cell, we would like to ask you to consider the conceptual novelty of our data not only from the
point of view of CSN, but also with regards to the function of deubiquitinating enzymes. In our
view, there are very few cases, in which the substrate targeting mechanism of a member of the
confusingly large DUB family is as clearly established as in this report. For most of the 14
different DUBs in yeast, and in particular Ubp12p, we know little more than that they are
dispensible for growth. How these proteases, which have very similar enzymatic activities in
vitro, are targeted to specific substrates in vivo, is in most cases unknown. Our data clearly show
that recruitment of Ubp12p to CSN targets it to its cullin substrates.
Secondly, Antony Carr’s lab will shortly be resubmitting to Cell a revised version of a
manuscript dealing with the role of fission yeast CSN in checkpoint control and protein
degradation. Bodo Stern is handling this manuscript. Both Dr. Carr and myself feel that a back-
to-back publication of both papers in Cell would ideally present these major advances in
understanding CSN that might even merit a Minireview written by Dr. Deshaies, Dr. Finley, or
one of the reviewers.
Dieter A. Wolf
This reviewer raises two general concerns: (1.) imprecisions in English language use, and (2.)
“extensive model-building” that is “not very convincing”. In response to the first criticism, we
have removed all ambiguous language, in order to improve clarity, as suggested by the reviewer.
Regarding the second concern, we refer to our reply to reviewer 2, where we discuss in detail our
newly added data (revised Fig. 6) that strengthen our model.
A) The reviewer asks for clarification regarding the inhibitory effect of o-phenanthroline (o-PT)
on Ubp12p and other DUBs. In particular, the reviewer’s concern appears to pertain to the
concentration of o-PT used in our study (5 mM), although a misunderstanding may underlie this
concern. The reviewer states that “Any effect requiring (>100 mM) is probably non-specific, not
to mention 5 mM”. As 5 mM is far below 100 mM, but in the middle of the range of effective
concentrations of o-PT as stated by the vendor (1 – 10 mM, Sigma-Aldrich), we are unclear what
the concern is. We show that 5 mM o-PT inhibits the CSN-associated DUB activity and purified
Ubp12p and Yuh1p in vitro. In addition, it can restore Pcu3p-associated ubiquitin ligase activity
suggesting that it does not indiscriminately interfere with all enzyme functions. But it is correct
that we do not know the exact mechanism of DUB inhibition by o-PT. Using p-PT, as suggested
by the reviewer, would not answer this question. If we find that o-PT is effective while p-PT is
not, this would point to the potential involvement of divalent metals in DUB activity, although as
the reviewer states correctly, it would remain “quite unclear why Ubp12p is inhibited by a
chelating agent if it's a sulfhydryl protease”. If we find that both reagents are effective, the
mechanism of inhibition would seem even less clear. We wish to point out that none of our
conclusions depend on the exact mechanism of DUB inhibition by PT. We merely used PT,
because it was recently shown to inhibit other proteasome- and CSN-associated proteases (Cope
et al., 2002). In addition, we show that ubiquitin aldehyde and NEM inhibit Ubp12p, arguing that
it is a cysteine protease as generally believed.
B) The reviewer questions our proposition in the discussion that proximity of substrate adapters
to E2s presumably confers instability. This proposition rests on the previous demonstration that
F-box protein mutants, which no longer associate with cullins are resistant to the well-established
autocatalytic destruction mechanism (Galan and Peter, 1999; Wirbelauer et al., 2000; Zhou and
Howley, 1998). We do not see why this “unavoidable property” would be irreconcilable with
being an “important feature of cell regulation”. In fact, we believe it is an important mechanism
that enables a rapid change in substrate-specificity of cullin ubiquitin ligases by rapid turn-over
and exchange of adapters. But we also argue that mechanisms that counteract this rapid
autocatalytic degradation are required for an efficient assembly of cullin ubiquitin ligases. We
provide evidence for this mechanism in the revised manuscript (Fig. 6).
C) We are now providing additional evidence supporting the model in Fig. 6 and therefore
decided to retain it. Contrary to the reviewer's opinion, we have found in discussions with
colleagues that this model does facilitate the interpretation of our results.
D) The statement that F-box proteins bind phosphorylated substrates was omitted.
E) Ambiguities in the language have been resolved throughout the manuscript. We agree with
the reviewer that this improves the clarity of the manuscript.
The reviewer suggests multiple experiments to test predictions resulting from our model for a
potential role of Ubp12p and CSN in cullin ubiquitin ligase assembly. Some of these suggestions
may result from minor misunderstandings that we will address first.
A) 1. “...one might expect to see accumulation of ubiquitinated cullin, Skp1, or Rbx1 in Ubp12
mutants”. As we show in our model (former Fig. 6), we do not believe that cullins, Hrt1, or Skp1
are subject to autoubiquitylation. At least for the cullins, we have directly tested this by blotting
Pcu1p and Pcu3p immunocomplexes with anti-ubiquitin antibodies (Figs. 1A; 2B,C; 4C,D,
6A,B). In addition, as the reviewer states, there is no reported evidence for Hrt1 or Skp1
The confusion may have arisen from a revised statement in the discussion: “In order to
prevent autoubiquitylation of cullins and associated adapters, mechanisms must exist that either
prevent their interactions with UBCs or counteract the activity of the UBCs they associate with”.
For completeness, we have included cullins as theoretical targets of autoubiquitylation along
with adapters due their common proximity to UBCs, although we agree there is no evidence for
this and our own data disprove this possibility. Cullins were consequently omitted form this line
of argument in the revised manuscript. Clearly, Ubp12p can not counteract the ubiquitylation of
proteins that are not ubiquitylated.
2. “The model in Fig. 6 shows the isolated F-box protein associated with COP9, but is there
evidence for this?”. For clarification, we wish to note that the model shows isolated cullins in
association with CSN, but not adapters. In fact, we speculate that adapter binding may release
the cullin from CSN, a possibility we are currently pursuing. This is one of the predictions of our
model as the reviewer notes, but testing this prediction will require substantially more work with
purified components, which we feel is beyond the scope of the present manuscript.
3. The reviewer speculates that “the F-box protein remains bound to cullin during the
deneddylation cycle”, a possibility which would be consistent with “the finding that hCSN can
deneddylate and inactivate preformed SCFSkp2 complexes in vitro”. This is a possibility, but in
light of published suggestions that F-box proteins are degraded together with their substrates
(Zhou and Howley, 1998), we favor the possibility that cullins are deneddylated after releasing
their adapters. We are not aware of a study showing that preformed SCFSkp2 complexes can be
deneddylated by hCSN in vitro, although this finding would not preclude the possibility that
adapters are released before deneddylation occurs in vivo. To our knowledge, Yang et al. have
shown that purified CSN can inhibit p27 ubiquitylation in vitro, but the effect on CUL1
neddylation was not examined (Yang et al., 2002). Considering that the CSN preparation used in
these experiments likely contained Ubp12, CSN-mediated inhibition of p27 ubiquitylation in
vitro could also have been caused by the counteracting activity of Ubp12.
B) Despite these potential minor misunderstandings, the reviewer suggests several experiments
concerning F-box protein stability that are strongly supported by our model. Specifically, our
model predicts that due to failed inhibition of autocatalytic destruction, cullin adapter proteins
may be less stable in csn and ubp12 mutants. In the revised manuscript, we have included
experiments confirming this prediction (Fig. 6) and thus provided concrete evidence supporting
As the putative adapters associating with Pcu3p are currently unknown, we focused our
attention on F-box proteins. This required first confirming that Pcu1p-associated ubiquitin ligase
activity is subject to similar negative control by CSN/Ubp12p as Pcu3p. This was confirmed by
testing the effect of csn deletion on Pcu1p activity. The results show that Pcu1p activity, like
Pcu3p activity, is derepressed in csn mutants (Fig. 6A of the revised manuscript). Purified hCSN
as well as recombinant Ubp12p in return, inhibit Pcu1p activity retrieved from csn mutants, as
seen with Pcu3p (Fig. 6B). Thus, the inhibitory effect of CSN/Ubp12p is a general phenomenon
affecting multiple cullins.
In the next step we asked, as suggested by the reviewer, whether the Pcu1p-associated F-box
protein Pop1p is less stable in csn and ubp12 mutants as predicted. As shown in Fig. 6C of the
revised manuscript, this is indeed the case. The effect, which was independently verified in three
experiments, was subtle but reproducible (half-life of Pop1p is 40.7 ± 3.0 min in wild-type cells,
27 ± 1.7 min in csn5 and 30.9 ± 2.1 ubp12 mutants). A t-test revealed that the difference in half-
life between wild-type and mutants was significant (p = 0.007 and 0.011, respectively), whereas
the difference between csn5 and ubp12 was insignificant (p = 0.1).
We were not surprised to see that the destabilizing effect is subtle, as a stronger effect would
have been expected to cause accumulation of the CDK inhibitor Rum1p, which should lead to
the strong cellular phenotype of pop1 mutants. A previous study already indicated that Rum1p
stability is not strongly affected in csn1 mutants, potentially due to the added layer of control
through Rum1p phosphorylation (Lyapina et al., 2001).
We believe that the new data lends significant credibility to our model suggesting that
CSN/Ubp12p function contributes to efficient cullin ubiquitin ligase assembly by stabilizing
adapter proteins. Clearly more experiments will be required to elucidate the entirety of this
mechanism, including the question of whether adapter binding releases cullins from
CSN/Ubp12p-mediated inhibition. However, we feel that the data presented in this report,
identifying Ubp12p as a new CSN-associated enzymatic activity regulating cullin function by
counteracting adapter protein instability, are significant enough to warrant publication at this
point in time.
C) A further point raised by this reviewer and reviewer 4 concern our immunostaining data
suggesting that CSN is required for efficient nuclear accumulation of Ubp12p. Although the
concerns of reviewer 4 may be partly caused by the poor quality of the paper copies received by
this reviewer, it is true that the cytoplasmic redistribution of Ubp12p in csn5 mutants is not
complete. However, we have included in Fig. 5C of the revised manuscript, new micrographs of
Ubp12p localization in csn4 mutants. Again, there was a striking, though not absolute
redistribution of Ubp12p to the cytoplasm.
In evaluating these data, we believe it is helpful to keep in mind that the nuclear compartment
of fission yeast cells is substantially smaller than the cytoplasmic compartment. Based on a
nuclear diameter of 2.5 micrometers and an average cell size of 4 by 4 by 10 micrometers, Dr. R.
McIntosh, perhaps the most experienced S. pombe electron microscopists who we consulted in
this matter, estimates a cytoplasmic/nuclear ratio of ~16. In other words, uniform cell staining as
observed for Ubp12p in most csn mutant cells (revised Fig. 5C) would correspond to greater than
90% of Ubp12p residing in the cytoplasm. The more evident nuclear staining in wild-type cells
may therefore indicate a roughly equal distribution of Ubp12p in the cytoplasm and the nucleus,
although this is difficult to quantify exactly.
In describing our immunostaining results in the previous version of the manuscript, we had, in
fact, underappreciated the large estimated cytoplasmic/nuclear ratio in fission yeast. We have
therefore rephrased the description of Ubp12p staining in wild-type cells from “largely nuclear”
to “roughly equal cytoplasmic and nuclear distribution”.
Our working hypothesis for the future is that the two peaks of Ubp12p observed in gel
filtration (Fig. 5B) correspond to the nuclear, CSN-bound (high molecular weight), and
cytoplasmic (low molecular weight) fractions, respectively. If this is true, the gel filtration data
would suggest that Ubp12p approximately equally localizes to the nucleus and the cytoplasm in
wild-type cells (Fig. 5B), whereas it would be largely (maybe 90%) cytoplasmic in csn mutants.
In agreement with this interpretation, our coimmunoprecipitation data in Fig. 5A show that
Pcu3p binding to Ubp12p is greatly reduced, but not completely abolished in csn mutants. We
can further test these possibilities, once we have established reliable protocols for biochemical
separation of soluble nuclear and cytoplasmic proteins, which is not trivial in S. pombe.
D) Minor Points
1. The statement that F-box proteins associate with phosphorylated substrates has been qualified
to reflect newer evidence that F-box proteins also bind unphosphorylated substrates.
2. We greatly appreciate the reviewer's clarifying comments regarding the species specificity of
GST-UBC3 activity that escaped our attention. The relevant section in the manuscript was
revised to reflect these insights.
1)“In Fig. 1A, the authors suggest that the difference in Pcu3 ubiquitination is only 2-fold
different in csn mutants. It doesn't look that small a difference to me...”. We show that Pcu3p
neddylation, not ubiquitination, is twofold different. This is apparent from the lower panel of
Fig. 1A, showing that in wild-type cells approximately 50% of Pcu3p is neddylated, whereas
in csn mutants all of it is neddylated. Therefore, the neddylated forms increase two-fold.
However, Pcu3p-associated ubiquitylation activity increases from virtually undetectable
levels in wild-type to a robust activity in all csn mutants examined (Fig. 1A, top panel). We
argue “This stimulation far exceeds the approximate twofold increase in Pcu3p-13Myc
neddylation observed in csn mutants over wild-type.” We assume that this explanation
clarifies the issue.
2)“...does the human signalosome actually fail to deneddylate Pcu3 or is it just slower?”. During
the past year, we have spent a lot of effort on determining whether CSN can deneddylate
Pcu3p. Whereas our hCSN preparation can efficiently deneddylate Pcu1p (Fig. 1B) and
budding yeast Cdc53p in vitro (Wee et al., 2002), we have never detected deneddylation of
Pcu3p-13Myc even when the amount of hCSN and the incubation times were substantially
increased. Since in vivo, there is clearly an effect of CSN deletion on Pcu3p-13Myc
neddylation (Fig. 1A and Zhou et al., 2001), the large C-terminal tag may interfere with
hCSN-mediated Pcu3p-13Myc deneddylation in vitro. Support for this notion is provided by
our new data showing that hCSN also fails to deneddylate C-terminally tagged Pcu1p-13Myc
in vitro (Fig. 6B). On the other hand, tagged cullins seem to be good substrates for CSN in
vivo, as tagging alone does not lead to accumulation exclusively in the neddylated state (Fig.
1A and 6A). While we have been unable to provide a satisfactory explanation for this
discrepancy, the failure of hCSN to deneddylate C-terminally tagged cullins in vitro has
allowed us to examine its role in the control of cullin activity independent of its role in
deneddylation. This provided the basis for the identification of Ubp12p.
3)The reviewer questions the specificity of our CSN/Ubp12p co-immunoprecipitation
experiments and asks whether we have “other non-specific binding controls” and whether we
“can do the coIP in the opposite direction”. We believe that the CSN/Ubp12 interaction is
specific, as it is maintained during multiple steps of chromatographic purification of hCSN
Moreover, in fission yeast, we show that Ubp12p-proA bound to IgG beads co-purifies
four different CSN subunits. We were unable to perform a reciprocal immunoprecipitation
where Myc-tagged CSN subunits are precipitated, as Ubp12p-proA would also bind to the
Myc antibodies via its protein A moiety. Regarding a non-specific binding control, it is
difficult to know which protein to choose, unless it was previously established that it does not
interact with Ubp12p or its other associated proteins (CSN, cullins etc.). We have therefore
tested whether tubulin is co-precipitated with Ubp12p-proA, which was not the case (data not
shown), however, we are unceratin whether this is really a more relevant control than the
“bead only” control we include with Fig. 3B.
4)Regarding our immunostaining data, we refer to our discussion of this point in response to the
comments of reviewer 2. In response to the reviewer’s suggestion, we have zoomed in on
cells in a smaller field of the micrographs and increased their contrast (revised Fig. 5C).
5)“...one suspects that there are multiple ubps functioning redundantly to support this important
role”. The finding that all budding yeast ubp single mutants, and even some quadruple
mutants are viable, points to a significant degree of functional redundancy. On the other hand,
distinct biochemical defects were detected in some single mutants that are not complemented
by redundant genes. The same appears to be true for fission yeast ubp12 mutants, which by
themselves have a clear biochemical defect in inhibiting cullin-associated ubiquitin ligase
activity (Fig. 4C) and F-box protein instability (Fig. 6C), which appear not to be
complemented by any other UBP. We attribute this to the substrate targeting function of CSN,
which creates specificity by recruiting Ubp12p, but not other DUBs, to cullins. In our view,
this is one of the most important findings of our study.
The biochemical defects in ubp12 mutants do not lead to an obvious cellular phenotype,
however. We suspect that there may be circumstances in the life-cycle of fission yeast, in
which Ubp12p function in ubiquitin ligase assembly is advantageous without impacting
directly on cell viability under normal growth conditions. This could be as subtle as
maintaining “economy” in the synthesis of ubiquitin ligase components or as a fail-safe
mechanism when controls at the level of substrate modification cave in. It could be argued
that eukaryotes evolved so many different and specific UBPs, in order to ensure viability
under various conditions rather than “putting all eggs into one basket”.
Cope, G., Suh, G. S. B., Aravind, L., Schwarz, S. E., Zipursky, S. L., Koonin, E. V., and
Deshaies, R. J. (2002). Role of Predicted Metalloprotease Motif of Jab1/Csn5 in Cleavage of
NEDD8 from CUL1. Science 298, 608-611.
Galan, J. M., and Peter, M. (1999). Ubiquitin-dependent degradation of multiple F-box proteins
by an autocatalytic mechanism. Proc Natl Acad Sci U S A 96, 9124-9129.
Lyapina, S., Cope, G., Shevchenko, A., Serino, G., Tsuge, T., Zhou, C., Wolf, D. A., Wei, N.,
Shevchenko, A., and Deshaies, R. J. (2001). Promotion of NEDD8-CUL1 conjugate cleavage by
COP9 signalosome. Science 292, 1382-1385.
Wee, S., Hetfeld, B., Dubiel, W., and Wolf, D. A. (2002). Conservation of the
COP9/signalosome in budding yeast. BMC Genetics 3:15.
Wirbelauer, C., Sutterluty, H., Blondel, M., Gstaiger, M., Peter, M., Reymond, F., and Krek, W.
(2000). The F-box protein Skp2 is a ubiquitylation target of a Cul1-based core ubiquitin ligase
complex: evidence for a role of Cul1 in the suppression of Skp2 expression in quiescent
fibroblasts. Embo J 19, 5362-5375.
Yang, X., Menon, S., Lykke-Andersen, K., Tsuge, T., Di, X., Wang, X., Rodriguez-Suarez, R. J.,
Zhang, H., and Wei, N. (2002). The COP9 Signalosome Inhibits p27(kip1) Degradation and
Impedes G1-S Phase Progression via Deneddylation of SCF Cul1. Curr Biol 12, 667-672.
Zhou, C., Seibert, V., Geyer, R., Rhee, E., Lyapina, S., Cope, G., Deshaies, R. J., and Wolf, D.
A. (2001). The fission yeast COP9/signalosome is involved in cullin modification by ubiquitin-
related Ned8p. BMC Biochemistry 2, 7.
Zhou, P., and Howley, P. M. (1998). Ubiquitination and degradation of the substrate recognition
subunits of SCF ubiquitin-protein ligases. Mol Cell 2, 571-580.