USP10 inhibits genotoxic NF-κB activation by MCPIP1- facilitated deubiquitination of NEMO

Transcription

1 Manuscript EMBO USP10 inhibits genotoxic NF-κB activation by MCPIP1- facilitated deubiquitination of NEMO Jixiao Niu, Yuling Shi, Jingyan Xue, Ruidong Miao, Shengping Huang, Tianyi Wang, Jiong Wu, Mingui Fu, Zhao-Hui Wu Corresponding author: Zhao-Hui Wu, University of Tennessee Health Science Center Review timeline: Submission date: 23 January 2013 Editorial Decision: 19 February 2013 Re-submission: 21 June 2013 Editorial Decision: 22 July 2013 Revision received: 20 August 2013 Editorial Decision: 02 October 2013 Additional correspondence (author): 03 October 2013 Additional correspondence (editor): 10 October 2013 Additional correspondence (author): 17 October 2013 Editorial Decision: 21 October 2013 Revision received: 24 October 2013 Accepted: 31 October 2013 Editor: Hartmut Vodermaier Transaction Report: (Note: With the exception of the correction of typographical or spelling errors that could be a source of ambiguity, letters and reports are not edited. The original formatting of letters and referee reports may not be reflected in this compilation.) Editor: 1st Editorial Decision 19 February 2013 Thank you for submitting your manuscript on USP10/MCPIP1-mediated NEMO deubiquitination for consideration by The EMBO Journal. We have now received the comments from three expert referees, which you will find copied below. I am afraid that these reports do not offer sufficient support for publication of this study in The EMBO Journal. Although the finding that MCPIP1- dependent DUB activity is mediated by USP10 and targeted specifically to linear ubiquitin chains in the context of genotoxic NF-kB activation is acknowledged as potentially interesting and important, all referees raise a large number of well-taken concerns that in our view have the potential to undermine the key conclusions and interpretations of the manuscript. I will not repeat all these points of criticism in detail here; however, it is apparent that all referees are concerned about technical issues, significance and conclusiveness of many of the presented results. In view of this long list of overlapping criticisms, I am afraid we feel that it would be premature to consider this manuscript a strong enough candidate for an EMBO Journal paper to invite a revision at this present stage. I would in this case not exclude the possibility of looking once more at a new version of this work, should future efforts guided by the comments of our reviewers allow to decisively substantiate the results and conclusions of this study. However it is only fair to say that such a new submission European Molecular Biology Organization 1

2 would only be sent back to our referees if we felt the key concerns have been exhaustively and conclusively addressed, and if the novelty of the findings would still be uncompromised at the time of resubmission. I am sorry we cannot be more positive at this stage, but would nevertheless like to thank you for opportunity to consider this work for The EMBO Journal, and hope that you will find our referees' detailed comments and suggestions helpful. Referee #1: Manuscript by Niu et al explores the roles of zinc finger-containing protein MCPIP1 and deubiquitinase USP10 in genotoxic NF-kB activation. The authors claim that USP10, which interacts with MCPIP1 and via MCPIP1 with NEMO, removes linear polyubiquitin chains from NEMO. As a result, DNA damage induced NF-kB activation is decreased. - The main deficiency of this study is the absence of solid evidence that removal of linear polyubiquitination from NEMO by USP10 is critical for the observed effect on NF-kB activation. USP10 can cleave different ubiquitin linkage chains, not just linear polyubiquitin, and NEMO is just one of the substrates of USP10. The authors claim from their reconstitution/overexpression experiments that K63-linked ubiquitination is not critical for NF-kB activation. However, they never showed that NEMO is not ubiquitinated with K63-linked or other chains. Other ubiquitin chains, such as K6, K11 or K48 could also play a role in DNA damage induced NF-kB activation and the authors should examine if these chains are involved in their system and if USP10/MCPIP1 can cleave these other ubiquitin chains following DNA damage stimuli. This should be done with several time points to establish the kinetics of DUB reaction, and in the presence or absence of DUB, not just at one time point as shown now in 6C. They should also reconstitute U2OS cells with K11R, K48R, N-methyl-ubiquitin and K7R ubiquitin constructs. K7R and N-methyl-ubiquitin would be especially informative because K7R ubiquitin would allow the formation of only linear linkage and N-methylated ubiquitin would specifically prevent the formation of linear polyubiquitination. Finally, the authors should downregulate components of LUBAC to independently confirm the importance of linear polyubiquitination for DNA damage induced NF-kB activation. - The effect of USP10 knockdown is always assessed in the context of MCPIP1 overexpression. The authors should knockdown USP10, treat cells with DNA damaging agents without MCPIP1 overexpression and then assess NEMO polyubiquitination, gene expression and apoptosis (figures 6E, 7C, 7D, S5A, S5B, S6B). - The authors should purify recombinant USP10 and NEMO and investigate deubiquitination of NEMO in reconstituted DUB assay without possible contamination from co-purified proteins from overexpression in mammalian cells. - From figure 5C it seems that NEMO binds MCPIP1 and USP10 even without DNA damaging stimulus. Although interaction is enhanced after 2h with dox, the authors should perform more careful analysis with several time points to investigate how fast is this complex formed after DNA damaging stimuli. They could also do these experiment in ubiquitin mutants reconstituted cells to examine in NEMO ubiquitination is required for binding to MCPIP1 and USP10. - The authors state on page 12 that they identified USP10 as a MCPIP1-interacting protein by mass spectrometry. This statement needs to be supported by more data - for example, which other proteins were identified in this pull-down, did they also find NEMO (as they claim that MCPIP1 binds NEMO), they should show peptide traces or peptide region of USP10 that was identified. In addition, what was used as a positive and negative control in these experiments? Specific points: - In some experiments that authors observe MCPIP1 upregulation (1A-B) but in some not (2D), Why? - The activation of apoptosis is barely visible in figure 7A and there are almost no differences for cleaved caspase-3 in 7G. The same is true for supplementary figure S7B where there is basically no difference in survival of MCPIP1 +/+ or -/- cells after irradiation. In addition, the authors do not provide any information on how long was this assay. These data should be shown on a linear graph with clear indication of how many cells were examined and how many times. European Molecular Biology Organization 2

3 - It s not clear why the authors reference Komander et al 2009 study when that paper shows differential association of ciap1 UBA domain for linear and K63 chains in pull-down assay? They go on and state that the UBA domain of MCPIP1 may also show promiscuous binding to both linear and K63 ubiquitin chains. The authors should determine exact binding affinity of MCPIP1 UBA domain for different ubiquitin chains using purified UBA to experimentally verify this important point. Referee #2: This manuscript describes an interesting negative feedback mechanism that limits the activation of NFkB in response to genotoxic stress. MCPIP1, the focus of this study, has variously been described as an RNAse or a DUB. The present study provides strong evidence that MCPIP1 is induced in response to genotoxic stress in an IKK dependent and IkB-sensitive manner and suggests that one mechanism by which MCPIP1 suppresses NFkB signalling may be through recruitment of a DUB, USP10, to NEMO. This, the authors propose, results in removal of linear chains from NEMO and consequential inactivation of the pathway. The evidence regarding a key role for MCPIP1 in the attenuation of NFkB signalling in response to genotoxic stress is very convincing. The identification of NEMO as the key substrate for MCPIP1-associated deubquitination amongst a multitude of potential ubiquitinated targets appears somewhat arbitrary. Overall this is a very comprehensive treatment of the potential mode of action of MCPIP1 through recruitment of a DUB to NEMO, although both the identification of the DUB and its substrate seem a bit arbitrary in this story. I only have a few specific comments listed below: 1. Figure 3A: I could not see any validation of this linear ubiquitin chain specific antibody here. This is mandatory information and a validation of the specificity (e.g. on different ubiquitin chains) should be included. 2. Page 10: ubiquitin switching system. The authors state in the text that endogenous ubiqitin was significantly depleted by shub - a western blot showing free ubiquitin levels is required to assess this point. Figure 4A - what is HA-tagged in this figure? Molecular weight markers need to be shown for all ubiquitin blots, to allow the reader to assess where the Ub bands are in respect to the unmodified protein. 3. Figure 4E - is the UBA domain required for a) association of MCPIP1 with NEMO, and b) deubiquitylation of NEMO? 4. Page 12: The identification of USP10 as a MCPIP1 interacting partner is said to be based on a proteomic study - are these data not shown? 5. The authors' interpretation may well be the simplest one, given the direct association between MCPIP1 and USP10 as well as NEMO, however it cannot be excluded that USP10 activity is required elsewhere in the cascade. A loss of ubiquitin chain linkages on NEMO may be a mere consequence of inhibition of the cascade at an alternative upstream point, or crosstalk between p53- dependent regulatory mechanisms, which have previously implicated USP10 function, and NFkB signalling cascade. Alterations in signalling and NEMO ubiquitylation may for example also be observed if USP10 was required to stabilise any negative negative regulator of the pathway. This needs to be discussed. Referee #3: The manuscript by Niu and colleagues reports on a novel role of MCPIP1 and USP10 in the negative regulation of genotoxic stress induced gene expression. More specifically, they European Molecular Biology Organization 3

4 demonstrate that MCPIP1 is induced by genotoxic stress and mediates the removal of linear polyubiquitin chains from the IKK adaptor protein NEMO, lowering NF-kB activation. Whereas MCPIP1 has previously been proposed to function as a deubiquitinase, the authors contradict this by using recombinant MCPIP1. In contrast, they show that deubiquitination of NEMO by MCPIP1 is mediated by its ability to bind linear ubiquitin chains (on NEMO) and to recruit USP10, which is shown to be responsible for NEMO deubiquitination. The resulting decrease in NF- B activation lowers the expression of NF-kB dependent cell survival genes, which is also reflected by a higher apoptotic response in the presence of MCPIP1. Finally, the authors also demonstrate the in vivo role by comparing the effect of irradiation of wild type and MCPIP1 knockout mice on NF-kB activity and apoptosis in the small intestine. Linear ubiquitination has only recently been found to play an important regulatory role in NF-kB signaling and the authors previously reported a role for this type of modification in genotoxic stress induced NF-kB activation. Until now, however, the mechanisms that negatively regulate this pathway in the context of genotoxic stress were unknown. The present finding is therefore of significant interest. Moreover, it expands the already diverse mechanisms by which MCPIP1 has been shown to modulate proinflammatory gene expression in response to specific immune receptors. The paper is well written and the conclusions are based on several well designed and complementary approaches with cell lines such as overexpression or knockdown of MCPIP1 and USP10, as well as various mutants, use of MCPIP1 knockout cells, and several biochemical assays. However, a number of issues still need to be addressed. 1. The authors conclude that increased MCPIP1 expression upon genotoxic stress is due to the NFkB dependent upregulation of MCPIP1 expression. However, it should be mentioned that the increase in MCPIP1 mrna expression is mostly mild (2-3 fold) and although the authors show binding of p65 to specific regions of the MCPIP1 gene, it is still possible that the effects of NF-kB inhibitors represent indirect effects. Moreover, since MCPIP1 is known to be regulated at the level of mrna stability, regulation at this level may also contribute to the observed changes. The authors should therefore mutate the identified NF-kB binding site and test its effect in an NF-kB dependent luciferase reporter assay. Also pulse-chase experiments may reveal a possible contribution of mrna degradation. 2. NF-kB activation is in all experiments evaluated by EMSA, showing increased binding to a specific DNA probe. However, no information on the identity of the specific NF-kB family member is indicated. Is this p65? (this may require supershift experiments to answer). Also, how specific is the signal? (competition with non-labeled probe?, other bands visible in whole gel?). Importantly this type of assay only tells that NF-kB has moved to the nucleus, not that IKKs are activated (which one would expect if NEMO is the target). The authors should therefore also analyze the effect of MCPIP1 overexpression and deficiency on IKK activity (e.g. measuring IKK phosphorylation or IKK activity in an in vitro kinase assay). This would exclude that MCPIP1 and USP1 simply affect the nuclear translocation of NF-kB (and maybe also NEMO). In addition, some of these figures could also benefit from supporting data showing the effect of MCPIP1 on the induction of NF- B target genes. 3. In fig. 3B, the authors show that MCPIP1 inhibits LUBAC induced linear ubiquitination of NEMO. In this experiment they use HOIP and HOIL overexpression (although this should still be mentioned in the legend). Why is overexpression of HOIP/HOIL as such (without Etop) not inducing linear ubiquitination of NEMO as it is known that overexpression of LUBAC is sufficient to activate NF-kB (as also shown in Fig. 4A)? 4. The authors propose that MCPIP1 binds to linear ubiquitinated NEMO. Why then does one not see the binding of MCPIP1 to higher running NEMO in fig. 3D (last lane)? Also, why can binding of MCPIP1 to NEMO be detected without stimulation of the cells if this is dependent on linear ubiquitination (Fig3D)? 5. On page 10 the authors conclude 'Taken together, these results further support that MCPIP1- dependent inhibition of NEMO linear ubiquitylation may diminish sequential activation of TAK1 and IKK, resulting in decreased NF- B activation upon DNA damage.' However, this does not make sense as the effect of MCPIP1 on linear ubiquitination of NEMO is seen after 2h stimulation with Etop or CPT and the kinase activity is already investigated after 90 min. 6. In fig. 4A the authors show results obtained in cells depleted for ubiquitin and reconstituted with wt and Ub(K63R) to show that LUBAC can induce NF- B activation independent of K63 polyubiquitination. I do have some problems with this figure and the conclusions. First, what does the HA antibody detect? NEMO? HA-ubiquitin? This is not clear from the figure or legend. Second, the authors should also use anti-linear ubiquitin to show that LUBAC is indeed only inducing linear ubiquitination. In this context and as a control, the authors should also detect this linear European Molecular Biology Organization 4

5 ubiquitination in cells reconstituted with N-terminally tagged ubiquitin that cannot be used for linear ubiquitination. 7. The authors show that MCPIP1 binds linear polyubiquitin via its UBA domain and suggest that this allows MCPIP1 to bind linear ubiquitinated NEMO. However, MCPIP1 has also been reported to disassemble K63 chains in an UBA dependent manner, suggesting that this domain also binds K63 chains. To define UBA specificity and its identity as a real linear ubiquitin binding domain, the authors should carry out affinity binding experiments of MCPIP1 with an equimolar mixture of K63 and linear chains (as recently reported by Kensche et al., J Biol Chem Jul 6;287). 8. Fig4E may also lead to wrong conclusions as immunoprecipitating MCPIP1 from transfected cells might pull down endogenous ubiquitin chains or other proteins that bind to linear ubiquitin (as is clear from the controversial findings with MCPIP1 purified from HEK293T cells versus other sources). A formal demonstration of the affinity of MCPIP1 for a specific type of ubiquitin chain would require studies using recombinant MCPIP1 purified from non-mammalian cells. Along the same lines, Fig 4B/C should not be included in the paper, as it is mentioned in the text that recombinant MCPIP1 did not show DUB activity (it may still be shown in the supplementary data). I would also suggest to rephrase the paragraph as it is very misleading. 9. If binding of MCPIP1 to linear ub via its UBA domain is important for the NF-kB inhibitory effect of MCPIP1, then one should also expect that UBA mutation affects the NF-kB inhibitory potential of MCPIP1. This should be tested. 10. From fig. 5B, the authors conclude that MCPIP1-USP10 binding is etoposide inducible. However, this is a bit misleading as the effect is minor and most likely only reflects the higher levels of MCPIP1 after etoposide. The authors should rephrase this in the text. 11. the authors show that the inhibitory effect of MCPIP1 can be counteracted by USP10 silencing (Fig. 6E and Fig. S6B). However, this experiment needs a control showing the effect of USP10 silencing on non-transfected cells. If real, this alone should increase etoposide induced linear ubiquitination of NEMO and gene expression. 12. the authors suggest that USP10 is able to deubiquitinate linear ubiquitin chains. However, evidence is based on USP10 purified from transfected HEK293T cells. As it is clear from the MCPIP1 data that using proteins purified HEK293T cells may be misleading because of the copurification of active components that may be responsible for the observed effects, the authors should use recombinant USP10 from other sources to make their point. 13. The description of 'small intestine' or 'intestinal tissues' for the experiment shown in Fig. 7 is very vague. It is known that different regions may behave differently. Is this proximal, distal,... How are the extracts prepared? 14. Fig. 7G: why are the NF-kB bands running at different height? This seems to reflect a change in the composition of the bound NF-kB complex. 15. For different experiments, the authors use different stimuli to induce genotoxic stress. For example, Etop is used for HEK293T and HT1080 cells (Fig1A and 1B), Dox is used for MDA-MB- 231 and MEF cells (Fig1F and S1B), IR for HEK293T cells (Fig2B). This is sometimes very confusing and the reason for this is not clear. Also, conclusions are sometimes made from 2 experiments using different stimuli. I would suggest trying to combine data from 1 stimulus and have data with other stimuli as supplementary figures. 16. Similarly, the authors use two different IKK inhibitors. Bay is used to test the involvement of NF-kB in MCPIP1 protein and mrna upregulation (Fig1C and 1D), whereas TPCA-1 is used for p65 binding (Fig1E). They then combine everything to make their conclusion out of it. 17. MCPIP1 has previously been shown to also regulate JNK activation in response to LPS and IL- 1b. Do MCPIP1 and USP10 also affect this pathway in response to genotoxic stress? This would give valuable information on the specificity of the effect. 18. Fig. S7B: the effect of MCPIP1 -/- on % survival is very minor. Therefore this experiment would benefit from additional data using Etop and Dox as stimuli. Is this similar? Minor comments: 1. The writing suffers from some grammatical and spelling errors; especially 'articles' are missing in many cases. 2. Legends should indicate how many times the experiments have been performed. Also if a mean value is given, the number of samples to calculate this mean should be provided. In addition, p- values should be indicated. 3. In general, legends should be more explanatory. For some figures there is also no indication of time of stimulation (e.g. FigS1D and S1E). Which tags are on which proteins?... European Molecular Biology Organization 5

6 4. Fig2C: also use the full name of CPT as it is also not mentioned in the text. 5. a schematic representation of the domain structure of MCPIP1 and indicating the location of several mutants used in this study would be very helpful. 6. From Fig2C the authors can't conclude that the absence of MCPIP1 leads to an extended duration of NF- B binding (instead of activation), as only one time point is shown. In fact, from Fig2D it is clear that at the time used in Fig2C, the binding is not extended but rather enhanced. 7. Fig 3F-H. How is phosphorylation detected? Radioactive? Phospho-specific antibodies? 8. Fig. 6A: the NF-kB signal should be quantified as done for the other figures. Re-submission 21 June 2013 I respectfully enclose the revised version to previous manuscript (EMBOJ ) entitled USP10 inhibits genotoxic NF-κB activation by disassembling linear polyubiquitin anchored on NEMO to be considered for publication in the EMBO Journal. We have revised the manuscript in order to best address the comments of the reviewers. I believe that the input of the reviewers has improved the final version of our manuscript. I provide a point-by-point answer to each of the reviewer s comments. Referee #1: Manuscript by Niu et al explores the roles of zinc finger-containing protein MCPIP1 and deubiquitinase USP10 in genotoxic NF-kB activation. The authors claim that USP10, which interacts with MCPIP1 and via MCPIP1 with NEMO, removes linear polyubiquitin chains from NEMO. As a result, DNA damage induced NF-kB activation is decreased. - The main deficiency of this study is the absence of solid evidence that removal of linear polyubiquitination from NEMO by USP10 is critical for the observed effect on NF-kB activation. USP10 can cleave different ubiquitin linkage chains, not just linear polyubiquitin, and NEMO is just one of the substrates of USP10. The authors claim from their reconstitution/overexpression experiments that K63-linked ubiquitination is not critical for NF-kB activation. However, they never showed that NEMO is not ubiquitinated with K63-linked or other chains. Other ubiquitin chains, such as K6, K11 or K48 could also play a role in DNA damage induced NF-kB activation and the authors should examine if these chains are involved in their system and if USP10/MCPIP1 can cleave these other ubiquitin chains following DNA damage stimuli. This should be done with several time points to establish the kinetics of DUB reaction, and in the presence or absence of DUB, not just at one time point as shown now in 6C. They should also reconstitute U2OS cells with K11R, K48R, N-methyl-ubiquitin and K7R ubiquitin constructs. K7R and N-methyl-ubiquitin would be especially informative because K7R ubiquitin would allow the formation of only linear linkage and N-methylated ubiquitin would specifically prevent the formation of linear polyubiquitination. Finally, the authors should downregulate components of LUBAC to independently confirm the importance of linear polyubiquitination for DNA damage induced NF-kB activation. Following the reviewer s suggestion, we further examined whether NEMO can be modified by other forms of polyubiquitin chain with distinct linkages. Using antibodies specifically recognizing K63- (CST#5621) or K48- (CST#8081) polyubiquitin linkage, we did not detect high-molecular-weight modification of NEMO in etoposide-treated 293T cells, while the polyubiquitinated NEMO were readily detected by antibodies recognizing ubiquitin or linear-linked polyubiquitin at later time points after etoposide treatment (Suppl-Fig. 3A). In contrast, we were able to detect polyubiquitinated forms of ELKS and IκBα with K63-linkage or K48-likage specific ubiquitin antibodies, respectively, in 293T cells after stimulation (Suppl-Fig. 3B). To our knowledge, antibody specifically recognizing K6-chain is not available. Meanwhile we were not able to obtain K11-chain specific antibody from commercial source (The vendor of this antibody, Millipore, has this product on hold due to technical issue). We thereby took an alternative approach to interrogate whether NEMO can be modified with K6- or K11-linked polyubiquitin upon DNA damage. We cotransfected 293T cells with Myc-NEMO and HA-Ub-K6 only, -K11 only, -K48 only or K63 only construct. Upon Etoposide treatment, we were able to detect NEMO polyubiquitination with anti-linear ubiquitin antibody in cells transfected with individual single lysine-only ubiquitin European Molecular Biology Organization 6

7 mutants, but not with anti-ha antibody (Suppl-Fig. 3C). All these data suggested that linear-linkage is the major form of NEMO polyubiquitination upon DNA damage. It will be ideal if we were able to generate U2OS cell lines in which endogenous ubiquitin were replaced with K48R, K11R, 7KR or N-Me-Ub to examine the importance of these specific linkage in genotoxic NF-κB signaling pathway. However, generating these ubiquitin-replacing stable cell systems will probably take more than a year s effort which will significantly delay the process of revision. We will continue this line of experiments in my follow-up studies, and hope the reviewer find our data from experiments using linkage-specific antibodies and single lysine only ubiquitin mutants as alternative approaches convincing. As the reviewer suggested, we examined genotoxic drug-induced NF-κB activation in HOIP-deleted HEK293T cells and MEFs with HOIL-1 or Sharpin deficiency. Consistent with our previous results (Niu et al, EMBO J, 2011), Genotoxic drugs (CPT and Dox)-induced NF-κB activation measured by EMSA was significantly decreased in HOIL-1 -/- and Sharpin -/- MEFs, compared with that in their wild type counterparts (see data below). Interestingly, although residual NF-κB activation in HOIPdepleted cells was detected at a much reduced level upon genotoxic treatment, this low level NF-κB activation was no longer sensitive to USP10 overexpression, suggesting that USP10-mediated inhibition of genotoxic NF-κB signaling is primarily dependent on its activity to attenuate linear polyubiquitination (Suppl-Fig. 6D). Consistently, we were able to detect Etoposide-induced NF-κB activation in the Ub-K63R replaced U2OS cells, which was still sensitive to overexpression of USP10-WT, but not -C424A mutant (Suppl-Fig. 6E). Meanwhile, we found NEMO linear ubiquitination was induced by etoposide treatment in Ub-K63R U2OS cells (Please see additional data in our response to Reviewer 3-Major point 6). These data suggest that non-k63 polyubiquitin chains, including linear chains, may be sufficient to mediate genotoxic NF-κB activation. As our data indicate that NEMO is mainly ubiquitinated with linear chain, and is the only protein identified thusfar to be linear ubiquitinated upon DNA damage, the USP10-dependent inhibition of genotoxic NF-κB activation in Ub-K63R cells further supported that USP10 may attenuate genotoxic NF-κB signaling by disassembling linear polyubiquitin chains anchored on NEMO. Meanwhile, we also discussed the potential additional mechanisms by which USP10 may inhibit genotoxic NF-κB signaling in revised text as suggested by the reviewer 2 (please see our response to comment 5). - The effect of USP10 knockdown is always assessed in the context of MCPIP1 overexpression. The authors should knockdown USP10, treat cells with DNA damaging agents without MCPIP1 overexpression and then assess NEMO polyubiquitination, gene expression and apoptosis (figures 6E, 7C, 7D, S5A, S5B, S6B). As the reviewer suggested, we depleted USP10 with specific sirna and examined its impact on genotoxic drug-induced NF-κB activation. USP10 knockdown alone markedly enhanced etoposideinduced NF-κB activation (suppl-fig. 5D). Consistently, genotoxic drug-induced Caspase 3 activation was substantially decreased in cells where USP10 was depleted by sirna (suppl-fig. 8E). This observation was further supported by our data showing that the induction of NF-κBdependent anti-apoptotic genes, such as c-iap1 and c-iap2, upon genotoxic stress was significantly augmented by USP10 knockdown (suppl-fig. 8F). Moreover, depletion of USP10 remarkably enhanced NEMO linear ubiquitination in cells treated with genotoxic drug (Fig. 6C). All this body of evidence supports a critical role of USP10 in negatively regulating genotoxic NF-κB activation, resulting in enhanced cell apoptosis upon DNA damage. - The authors should purify recombinant USP10 and NEMO and investigate deubiquitination of NEMO in reconstituted DUB assay without possible contamination from co-purified proteins from overexpression in mammalian cells. We cloned the DUB domain of USP10 into bacteria expression vector and purified recombinant histagged USP10 wild type and C424A mutant from E. coli (suppl-fig. 6A). Consistent with data of Flag-purified USP10 from mammalian cells, recombinant USP10 was able to disassemble linear- European Molecular Biology Organization 7

8 linked tetra-ubiquitin in vitro, while catalytic inactive mutation C424A abolished this activity (Fig. 6D). Moreover, recombinant USP10 wt, but not C424A mutant, was able to dismantle linear ubiquitin chain attached on NEMO in vitro (Fig. 6E). These data further supported the notion that USP10 can function as a DUB to directly disassemble NEMO-anchored linear ubiquitin chain. - From figure 5C it seems that NEMO binds MCPIP1 and USP10 even without DNA damaging stimulus. Although interaction is enhanced after 2h with dox, the authors should perform more careful analysis with several time points to investigate how fast is this complex formed after DNA damaging stimuli. They could also do these experiment in ubiquitin mutants reconstituted cells to examine in NEMO ubiquitination is required for binding to MCPIP1 and USP10. Following the reviewer s suggestion, we further analyzed the interaction between NEMO/MCPIP1/USP10 following genotoxic stimulation (Fig. 5D). We detected a weak interaction between USP10 and MCPIP1, with NEMO to a lesser extent, in resting cells. The interaction of these three proteins was substantially and continuously increased until 2 hours after etoposide treatment. Interestingly, we noticed that the association of NEMO with MCPIP1/USP10 complex was decreased while an increase of USP10-MCPIP1 interaction was detected at 3 hour after etoposide treatment. We reasoned the decrease of NEMO association with USP10/MCPIP1 may correlate to the USP10-depedent inhibition of NEMO linear ubiquitination, suggesting a critical role of linear ubiquitination in enhancing NEMO association with MCPIP1/USP10 complex. In contrast, USP10-MCPIP1 interaction is unlikely dependent on NENO linear ubiquitination, but may be affected by the increased MCPIP1 level after DNA damage as suggested by the reviewer 3. - The authors state on page 12 that they identified USP10 as a MCPIP1-interacting protein by mass spectrometry. This statement needs to be supported by more data - for example, which other proteins were identified in this pull-down, did they also find NEMO (as they claim that MCPIP1 binds NEMO), they should show peptide traces or peptide region of USP10 that was identified. In addition, what was used as a positive and negative control in these experiments? We now included the Sypro Ruby staining of immuno-purified MCPIP1 protein complex from Flag- MCPIP1 transfected HEK293T cells which was subjected to mass spectrometry analysis (suppl-fig. 5A). In parallel, cells transfected with empty vector was used as control. Selected unique peptide sequences identified were listed along with excised gel slice region. We did not capture NEMO in this proteomic analysis, possibly due to relatively low amount of NEMO associated with MCPIP1 in resting cells. Accordingly, neither of the other two IKK complex components, IKKβ and IKKα, was identified in this analysis, supporting a critical role of stimulation in enhancing MCPIP1-IKK complex interaction as shown in a recently study (Iwasaki et al. Nat Immunol. 2011, 12: ). Interestingly, we detected another MCPIP1-interacting protein identified by the same study (Iwasaki, 2011), β-trcp, in our proteomic analysis, suggesting a basal interaction between these two proteins. Specific points: - In some experiments that authors observe MCPIP1 upregulation (1A-B) but in some not (2D), Why? We repeated the experiment in Fig. 2D and the new data indicated that MCPIP1 level was indeed increased in MEF cells treated with genotoxic agents. - The activation of apoptosis is barely visible in figure 7A and there are almost no differences for cleaved caspase-3 in 7G. The same is true for supplementary figure S7B where there is basically no difference in survival of MCPIP1 +/+ or -/- cells after irradiation. In addition, the authors do not provide any information on how long was this assay. These data should be shown on a linear graph with clear indication of how many cells were examined and how many times. We now provide new data from repeated experiments in Fig 7G and longer exposure of the blot in Fig 7A, which both showed clear pro-apoptotic function of MCPIP1 in response to genotoxic treatment. Also, we repeated the cell survival experiments upon IR (shown in new suppl-fig. 8B) or Doxorubicin treatment (suppl-fig. 8C) and presented the survival curves in linear graphs as the reviewer suggested. The additional information of experimental procedure has been added in revised European Molecular Biology Organization 8

9 method section and in figure legend. Our data consistently support that MCPIP1 deficiency confers a survival benefit to cells upon genotoxic stress. - It s not clear why the authors reference Komander et al 2009 study when that paper shows differential association of ciap1 UBA domain for linear and K63 chains in pull-down assay? They go on and state that the UBA domain of MCPIP1 may also show promiscuous binding to both linear and K63 ubiquitin chains. The authors should determine exact binding affinity of MCPIP1 UBA domain for different ubiquitin chains using purified UBA to experimentally verify this important point. In the paper of Komander et al, ciap-c (UBA) was listed as a domain with similar specificity to both K63-linked and linear tertra-ubiqutin (Fig. 4E, EMBO Rep, 2009). To determine the specificity of MCPIP1 UBA to different linkages, we purified recombinant GST-MCPIP1 UBA domain, which was subjected to a similar in vitro pull down assay used by Komander et al. As shown in Fig. 4E, GST-MCPIP1-UBA was able to pull down both K63-Ub 4 and linear Ub 4, when they were presented alone or in mixture (1:1 ratio). As the difference of specific ubiquitin chain linkage may affect the ability of ubiquitin antibody to recognize the polyubiquitin peptides (see Lanes 1 and 2 of input in Fig. 4E, which were loaded with equal amount of polypeptides), it is difficult to directly compare the quantity of bait-bound ubiquitin chains with different linkages based on the signal from western blots in our pull down assay. Therefore, our data suggest that MCPIP1 UBA domain can bind to both K63-linked or linear polyubiquitin, potentially with different affinity. Referee #2 : This manuscript describes an interesting negative feedback mechanism that limits the activation of NFkB in response to genotoxic stress. MCPIP1, the focus of this study, has variously been described as an RNAse or a DUB. The present study provides strong evidence that MCPIP1 is induced in response to genotoxic stress in an IKK dependent and IkB-sensitive manner and suggests that one mechanism by which MCPIP1 suppresses NFkB signalling may be through recruitment of a DUB, USP10, to NEMO. This, the authors propose, results in removal of linear chains from NEMO and consequential inactivation of the pathway. The evidence regarding a key role for MCPIP1 in the attenuation of NFkB signalling in response to genotoxic stress is very convincing. The identification of NEMO as the key substrate for MCPIP1-associated deubquitination amongst a multitude of potential ubiquitinated targets appears somewhat arbitrary. Overall this is a very comprehensive treatment of the potential mode of action of MCPIP1 through recruitment of a DUB to NEMO, although both the identification of the DUB and its substrate seem a bit arbitrary in this story. I only have a few specific comments listed below: 1. Figure 3A: I could not see any validation of this linear ubiquitin chain specific antibody here. This is mandatory information and a validation of the specificity (e.g. on different ubiquitin chains) should be included. We used the anti-linear ubiquitin antibody developed by Dr. Iwai s lab, which has been shown to specifically recognize linear linkage in previous reports (Tokunaga et al. Nat Cell Biol, 11:123, 2009). To further characterize the NEMO polyubiquitination upon DNA damage, we used antibodies specifically recognizing K63- or K48-linkage as well as ubiquitin mutants harboring only K6 or K11 to examine whether NEMO can be conjugated with K63-, K48-, K6 or K11-linked polyubiquitin chain in cells exposed to genotoxic treatment. Our failure of detecting NEMO ubiquitination upon DNA damage by K63- or K48-linkage specific antibody indirectly supported the specificity of this linear ubiquitin antibody. Our data suggested that linear chain is the predominant form of NEMO ubiquitination in response to DNA damage (suppl-fig. 3A-C). Please also see our response to the reviewer 1 (Major points 1). 2. Page 10: ubiquitin switching system. The authors state in the text that endogenous ubiqitin was significantly depleted by shub - a western blot showing free ubiquitin levels is required to assess this point. Figure 4A - what is HA-tagged in this figure? Molecular weight markers need to be shown for all ubiquitin blots, to allow the reader to assess where the Ub bands are in respect to the European Molecular Biology Organization 9

10 unmodified protein. We used an inducible ubiquitin-replacement system developed by Dr. James Chen lab (Xu et al, Mol Cell, 2009). In shub-stable cells, we were able to observe significant decrease of endogenous ubiquitin in response to tetratcycline treatment as shown in suppl-fig. 4A (Lane 4 compared to Lane 1). As described in the original report (Xu et al,2009), the exogenous ubiquitin (WT or KR mutant) was introduced by a construct harboring tandem ubiquitin linked with IRES. The second copy the exogenous ubiquitin is tagged with HA while the first copy is untagged. We used anti-ha blot to show that exogenous ubiquitin was successfully induced by Tet treatment. We have included annotation in the figure legend. Following the reviewer s suggestion, protein weight markers are now included in all blots showing polyubiquitination. 3. Figure 4E - is the UBA domain required for a) association of MCPIP1 with NEMO, and b) deubiquitylation of NEMO? To address the reviewer s question, we carried out the co-ip experiment as shown in suppl-fig 4E. Deletion of UBA domain at N-terminus of MCPIP1 decreased, but not abolished, the interaction between NEMO and MCPIP1, suggesting additional domain of MCPIP1 also mediated NEMO- MCPIP1 association. Our data also indicated that the CCCH-Zn Finger domain may be essential for interaction between NEMO and MCPIP1, as C306R mutation completely disrupted their association. Furthermore, we found deletion of UBA domain substantially decreased the ability of MCPIP1 to inhibit NEMO linear ubiquitylation (suppl-fig 4F, lane 6), suggesting an important role of MCPIP1-UBA association with linear Ub chain in mediating NEMO deubiquitylation. 4. Page 12: The identification of USP10 as a MCPIP1 interacting partner is said to be based on a proteomic study - are these data not shown? We have included the proteomic analysis data in suppl-fig 5. Please also see our detailed response to the reviewer 1 (Major point 5). 5. The authors' interpretation may well be the simplest one, given the direct association between MCPIP1 and USP10 as well as NEMO, however it cannot be excluded that USP10 activity is required elsewhere in the cascade. A loss of ubiquitin chain linkages on NEMO may be a mere consequence of inhibition of the cascade at an alternative upstream point, or crosstalk between p53- dependent regulatory mechanisms, which have previously implicated USP10 function, and NFkB signalling cascade. Alterations in signalling and NEMO ubiquitylation may for example also be observed if USP10 was required to stabilise any negative negative regulator of the pathway.this needs to be discussed. We agree with the reviewer that USP10 may also negatively regulate NEMO linear ubiquitylation indirectly by inhibiting ubiquitylation of other proteins, such as ELKS, whose ubiquitylation is required for optimal NEMO linear ubiquitylation. Although our data indicated that USP10 can remove linear polyubiquitin attached on NEMO in vitro (Fig. 6E) and inhibit genotoxic NF-κB activation in Ub-K63R U2OS cells (suppl-fig. 6E), we cannot rule out the possibility that USP10 may also attenuate genotoxic NF-κB signaling by inhibiting other ubiquitylation events in normal cells. The role of USP10 in regulating p53 stability also has been well documented and the crosstalk between p53 and NF-κB signaling may play a critical role in regulating apoptosis in cells exposed to DNA damage. We have included the discussion on these potential alternative mechanisms in our revised text (Pg.20). Referee #3: The manuscript by Niu and colleagues reports on a novel role of MCPIP1 and USP10 in the negative regulation of genotoxic stress induced gene expression. More specifically, they demonstrate that MCPIP1 is induced by genotoxic stress and mediates the removal of linear polyubiquitin chains from the IKK adaptor protein NEMO, lowering NF-kB activation. Whereas MCPIP1 has previously been proposed to function as a deubiquitinase, the authors contradict this by using recombinant MCPIP1. In contrast, they show that deubiquitination of NEMO by MCPIP1 is mediated by its ability to bind linear ubiquitin chains (on NEMO) and to recruit USP10, which is European Molecular Biology Organization 10

11 shown to be responsible for NEMO deubiquitination. The resulting decrease in NF-κB activation lowers the expression of NF-kB dependent cell survival genes, which is also reflected by a higher apoptotic response in the presence of MCPIP1. Finally, the authors also demonstrate the in vivo role by comparing the effect of irradiation of wild type and MCPIP1 knockout mice on NF-kB activity and apoptosis in the small intestine. Linear ubiquitination has only recently been found to play an important regulatory role in NF-kB signaling and the authors previously reported a role for this type of modification in genotoxic stress induced NF-kB activation. Until now, however, the mechanisms that negatively regulate this pathway in the context of genotoxic stress were unknown. The present finding is therefore of significant interest. Moreover, it expands the already diverse mechanisms by which MCPIP1 has been shown to modulate proinflammatory gene expression in response to specific immune receptors. The paper is well written and the conclusions are based on several well designed and complementary approaches with cell lines such as overexpression or knockdown of MCPIP1 and USP10, as well as various mutants, use of MCPIP1 knockout cells, and several biochemical assays. However, a number of issues still need to be addressed. 1. The authors conclude that increased MCPIP1 expression upon genotoxic stress is due to the NFkB dependent upregulation of MCPIP1 expression. However, it should be mentioned that the increase in MCPIP1 mrna expression is mostly mild (2-3 fold) and although the authors show binding of p65 to specific regions of the MCPIP1 gene, it is still possible that the effects of NF-kB inhibitors represent indirect effects. Moreover, since MCPIP1 is known to be regulated at the level of mrna stability, regulation at this level may also contribute to the observed changes. The authors should therefore mutate the identified NF-kB binding site and test its effect in an NF-kB dependent luciferase reporter assay. Also pulse-chase experiments may reveal a possible contribution of mrna degradation. We agree the fold induction of MCPIP1 mrna level upon DNA damage was mild, which is consistent with our previous reports in which genotoxic stress-induced NF-κB-dependent gene induction was generally mild compared with those induced by strong NF-κB activators such as TNFα and LPS. It may also reflect a negative feedback response that MCPIP1 destabilizes the mrna of itself (Iwasaki et al. Nat Immunol. 2011, 12: ). Nevertheless, we detected consistent increase of MCPIP1 protein level in cells treated by genotoxic drugs. To examine whether the increase of MCPIP1 transcription is a direct effect of NF-κB-dependent transactivation, we generated MCPIP1 promoter and enhancer-driven luciferase reporter constructs harboring both NFκB-binding regions following the reviewer s suggestion. As shown in Fig 1F, deletion of either NFκB-binding region in MCPIP1 promoter or enhancer abrogated the genotoxic drug-induced luciferase activity, which is consistent with our data showing p65 enrichment at respective regions upon genotoxic treatment by ChIP assay (Fig 1E). All these data indicated that NF-κB may directly upregulate MCPIP1 transcription in response to genotoxic stimulation. 2. NF-kB activation is in all experiments evaluated by EMSA, showing increased binding to a specific DNA probe. However, no information on the identity of the specific NF-kB family member is indicated. Is this p65? (this may require supershift experiments to answer). Also, how specific is the signal? (competition with non-labeled probe?, other bands visible in whole gel?). Importantly this type of assay only tells that NF-kB has moved to the nucleus, not that IKKs are activated (which one would expect if NEMO is the target). The authors should therefore also analyze the effect of MCPIP1 overexpression and deficiency on IKK activity (e.g. measuring IKK phosphorylation or IKK activity in an in vitro kinase assay). This would exclude that MCPIP1 and USP1 simply affect the nuclear translocation of NF-kB (and maybe also NEMO). In addition, some of these figures could also benefit from supporting data showing the effect of MCPIP1 on the induction of NF-κB target genes. To verify the specificity of our gelshift assay, we performed supershift assay along with cold-probe competition using whole cell extracts from etoposide-treated HEK293T cells. As shown below, the radioactive band signal represents majorly p65/p50 herterodimer of NF-κB, which can be attenuated by excessive cold probe and supershifted by antibodies against p65 or p50. We also used IKK and TAK1 kinase assays to monitor the NF-κB signaling kinase cascade in cells exposed to genotoxic stimuli and the impact of MCPIP1 on kinase activation by DNA damage (Fig 3F, 3G, 3H, suppl-fig 3E, 3F, 3G). Moreover, the induction of NF-κB-target genes, such as c-iap1/2, Bcl-xL, IL-6, IL-8, Cox-2 and TNFα, and the influence of MCPIP1/USP10 overexpression/depletion on these genes European Molecular Biology Organization 11

12 transcription upon genotoxic treatment were also measured by qrt-pcr (Fig. 7B, 7D, suppl-fig 7A, 7B) 3. In fig. 3B, the authors show that MCPIP1 inhibits LUBAC induced linear ubiquitination of NEMO. In this experiment they use HOIP and HOIL overexpression (although this should still be mentioned in the legend). Why is overexpression of HOIP/HOIL as such (without Etop) not inducing linear ubiquitination of NEMO as it is known that overexpression of LUBAC is sufficient to activate NF-kB (as also shown in Fig. 4A)? Overexpression of LUBAC Complex does induce basal level of NEMO linear ubiquitination, which could be significantly enhanced upon genotoxic treatment. In the experiment shown in Fig. 3B, HOIL-1 and HOIP were not overexpressed and the blots represented the endogenous protein level of HOIP and HOIL-1. We are sorry for the confusion caused by figure legend and we have added further detail in revised figure legends. 4. The authors propose that MCPIP1 binds to linear ubiquitinated NEMO. Why then does one not see the binding of MCPIP1 to higher running NEMO in fig. 3D (last lane)? Also, why can binding of MCPIP1 to NEMO be detected without stimulation of the cells if this is dependent on linear ubiquitination (Fig3D)? The amount of linear-ubiquitylated form of NEMO is rather small which is very difficult to be visualized by antibody detecting NEMO. Moreover, our data indicates that MCPIP1 may bind to NEMO with multiple domains, as deletion of UBA domain did not completely abolish association between MCPIP1 and NEMO (suppl-fig 4E). The abrogation of NEMO-MCPIP1 interaction by C306R mutation suggests that the MCPIP1 Zn finger is essential for NEMO-MCPIP1 association, which can be further stabilized by MCPIP1 UBA domain binding with linear ubiquitin chain. 5. On page 10 the authors conclude 'Taken together, these results further support that MCPIP1- dependent inhibition of NEMO linear ubiquitylation may diminish sequential activation of TAK1 and IKK, resulting in decreased NF-κB activation upon DNA damage.' However, this does not make sense as the effect of MCPIP1 on linear ubiquitination of NEMO is seen after 2h stimulation with Etop or CPT and the kinase activity is already investigated after 90 min. Our data indicated that endogenous MCPIP1 induction served as a negative feedback response to attenuate genotoxic NF-κB activation. Mechanistically, MCPIP1 could inhibit TAK1/IKK activation via mediating USP10-dependent NEMO deubiquitination in cells exposed to DNA damage, which served as a mechanism attenuating genotoxic NF-κB signaling. The TAK1 /IKK kinase assays and NEMO linear ubiquitination experiments we performed here were in cells overexpressing MCPIP1 (which bypassed the MCPIP1 induction) or MEFs with MCPIP1 genetic deletion, which may result in earlier and stronger phenotype observed here. Also, it is plausible that NEMO linear ubiquitination takes place at earlier times than 2 hours after genotoxic treatment, which may be not abundant enough for detection by immunoblotting, but sufficient for regulating IKK/TAK1 activation by DNA damage. Possibly, endogenous basal level MCPIP1 may also play a role in controlling the magnitude of TAK1/IKK activation at these earlier time points. 6. In fig. 4A the authors show results obtained in cells depleted for ubiquitin and reconstituted with wt and Ub(K63R) to show that LUBAC can induce NF-κB activation independent of K63 polyubiquitination. I do have some problems with this figure and the conclusions. First, what does European Molecular Biology Organization 12

13 the HA antibody detect? NEMO? HA-ubiquitin? This is not clear from the figure or legend. Second, the authors should also use anti-linear ubiquitin to show that LUBAC is indeed only inducing linear ubiquitination. In this context and as a control, the authors should also detect this linear ubiquitination in cells reconstituted with N-terminally tagged ubiquitin that cannot be used for linear ubiquitination. We are sorry for the confusion caused by lack of detail in our figure legend and methods section. As we addressed to the reviewer 2 (point 2), the HA-blot was used to detect exogenous HA-Ub induced by tertracycline treatment. As we shown below, in Ub-K63R-replaced U2OS cells, genotoxic treatment-induced polyubiquitylation of NEMO can be detected by antibody recognizing linear linkage, but not by anti-ha, confirming an intact N-terminus of ubiquitin is required for linear linkage forming. 7. The authors show that MCPIP1 binds linear polyubiquitin via its UBA domain and suggest that this allows MCPIP1 to bind linear ubiquitinated NEMO. However, MCPIP1 has also been reported to disassemble K63 chains in an UBA dependent manner, suggesting that this domain also binds K63 chains. To define UBA specificity and its identity as a real linear ubiquitin binding domain, the authors should carry out affinity binding experiments of MCPIP1 with an equimolar mixture of K63 and linear chains (as recently reported by Kensche et al., J Biol Chem Jul 6;287). Following the reviewers suggestion, we purified recombinant MCPIP1-UBA domain fused to GST and carried out an in vitro pull down assay. As shown in Fig 4E, MCPIP1 UBA domain was able to pull down both K63-linked and linear tetraubiquitin peptides when they were presented alone or in mixture (1:1 ratio). This result further supports the notion that UBA domain of MCPIP1 binds to both K63- and linear ubiquitin linkage, potentially with differential affinity. Please also see our response to reviewer 1 (Specific point 3). 8. Fig4E may also lead to wrong conclusions as immunoprecipitating MCPIP1 from transfected cells might pull down endogenous ubiquitin chains or other proteins that bind to linear ubiquitin (as is clear from the controversial findings with MCPIP1 purified from HEK293T cells versus other sources). A formal demonstration of the affinity of MCPIP1 for a specific type of ubiquitin chain would require studies using recombinant MCPIP1 purified from non-mammalian cells. Along the same lines, Fig 4B/C should not be included in the paper, as it is mentioned in the text that recombinant MCPIP1 did not show DUB activity (it may still be shown in the supplementary data). I would also suggest to rephrase the paragraph as it is very misleading. As we now shown in new Fig 4E, recombinant GST-MCPIP1 UBA domain was able to bind to K63- and linear tetraubiquitin, supporting the specificity of MCPIP1 for binding respective European Molecular Biology Organization 13

14 polyubiquitin linkage. Following the reviewer s suggestion, we also have moved the original Fig 4B/C to supplementary data and rephrased the corresponding text to improve the clarity. 9. If binding of MCPIP1 to linear ub via its UBA domain is important for the NF-kB inhibitory effect of MCPIP1, then one should also expect that UBA mutation affects the NF-kB inhibitory potential of MCPIP1. This should be tested. As the reviewer suggested, we examined the impact of UBA domain deletion on the ability of MCPIP1 to inhibit genotoxic NF-κB activation. As shown in Fig 4C, deletion of UBA substantially undermined the MCPIP1-dependent inhibition of NF-κB activation in cells treated with genotoxic drugs, suggesting a critical role of UBA domain in MCPIP1-mediated inhibition of genotoxic NFκB signaling. 10. From fig. 5B, the authors conclude that MCPIP1-USP10 binding is etoposide inducible. However, this is a bit misleading as the effect is minor and most likely only reflects the higher levels of MCPIP1 after etoposide. The authors should rephrase this in the text. We agree that the increase of MCPIP1 level after genotoxic stimulation may contribute to the increased association between USP10-MCPIP1, which may be indicated by our further analysis of this interaction with additional time points after genotoxic stimulation (Fig. 5D). We have revised the text accordingly to include this potential mechanism. 11. the authors show that the inhibitory effect of MCPIP1 can be counteracted by USP10 silencing (Fig. 6E and Fig. S6B). However, this experiment needs a control showing the effect of USP10 silencing on non-transfected cells. If real, this alone should increase etoposide induced linear ubiquitination of NEMO and gene expression. As shown in new Fig 5G (lane 5/6) and suppl-fig 5D, depletion of USP10 indeed enhanced NF-κB activation by genotoxic agents. We also observed increase NEMO linear ubiquitination in USP10- depleted cells (Fig 6C). Moreover, NF-κB-target gene c-iap1/2 induction upon DNA damage was further enhanced by USP10 depletion (suppl-fig 8F) 12. the authors suggest that USP10 is able to deubiquitinate linear ubiquitin chains. However, evidence is based on USP10 purified from transfected HEK293T cells. As it is clear from the MCPIP1 data that using proteins purified HEK293T cells may be misleading because of the copurification of active components that may be responsible for the observed effects, the authors should use recombinant USP10 from other sources to make their point. As also suggested by the reviewer 1, we purified recombinant USP10 protein (DUB domain) from E. coli, and used the recombinant USP10 to carry out in vitro DUB assay using Ub 4 peptide with different linkages as well as linear ubiquitinated NEMO as substrates. As shown in Fig 6D/E, recombinant USP10 was able to cleave linear tertaubiquitin into single ubiquitin moiety and reduce linear ubiquitination of NEMO in vitro, while catalytic inactive mutation C424A abolished this activity. These data support an important role of USP10 in disassembling linear ubiquitin chain attached on NEMO upon DNA damage. Please also see our response to the reviewer1 (Major point 3). 13. The description of 'small intestine' or 'intestinal tissues' for the experiment shown in Fig. 7 is very vague. It is known that different regions may behave differently. Is this proximal, distal,... How are the extracts prepared? We used the proximal section of mouse small intestine (Jejunum) tissues for the experiments and detailed methods for preparing protein extracts and mrna form mouse tissue have been included in revised methods section. 14. Fig. 7G: why are the NF-kB bands running at different height? This seems to reflect a change in the composition of the bound NF-kB complex. We repeated the experiments with freshly prepared samples and now show in new Fig 7G, NF-κB complexes visualized by our EMSA in both WT and MCPIP-KO samples were about the same size. European Molecular Biology Organization 14

15 To explore the potential composition of the major NF-κB complexes activated by IR, we also carried out supershift assay as the reviewer suggested. As shown below, the complex observed in our EMSA assay is likely composed of p50 and p For different experiments, the authors use different stimuli to induce genotoxic stress. For example, Etop is used for HEK293T and HT1080 cells (Fig1A and 1B), Dox is used for MDA-MB- 231 and MEF cells (Fig1F and S1B), IR for HEK293T cells (Fig2B). This is sometimes very confusing and the reason for this is not clear. Also, conclusions are sometimes made from 2 experiments using different stimuli. I would suggest trying to combine data from 1 stimulus and have data with other stimuli as supplementary figures. Both etoposide and doxorubicin are Topoisomerase II inhibitors. The various genotoxic drugs and IR we used in our study all generate double strand breaks (DSB) on genomic/mitochondrial DNA in treated cells. The reason we used multiple DNA-damaging drugs and IR for treatment is to confirm the observation we made are conserved cellular responses to DNA damage, especially to lethal DSB. 16. Similarly, the authors use two different IKK inhibitors. Bay is used to test the involvement of NF-kB in MCPIP1 protein and mrna upregulation (Fig1C and 1D), whereas TPCA- 1 is used for p65 binding (Fig1E). They then combine everything to make their conclusion out of it. Bay11 is a classical IKK inhibitor which has been used in a variety studies on NF-κB signaling. TPCA-1 is a relatively new IKK inhibitor whose activity is more potent and more specific in inhibiting IKKβ. As IKKβ is the essential catalytic component of IKK complex in genotoxic NF-κB signaling, we reasoned to use two different IKK inhibitors will provide more convincing results demonstrating the critical role of genotoxic NF-κB kinase signaling cascade in regulating MCPIP1 induction and gene transcription. 17. MCPIP1 has previously been shown to also regulate JNK activation in response to LPS and IL- 1b. Do MCPIP1 and USP10 also affect this pathway in response to genotoxic stress? This would give valuable information on the specificity of the effect. Following the reviewer s suggestion, we examined the JNK activation upon genotoxic treatments (suppl-fig 7F and 7G). Consistent with previous reports, we observed significant increase of JNK activation in response to genotoxic stress in MCPIP1 deficient MEF cells and in USP10-depleted HEK293Tcells, supporting a critical role of MCPIP1/USP10 in regulating JNK activation by genotoxic stimulation, potentially via modulating upstream TAK1 activation. 18. Fig. S7B: the effect of MCPIP1 -/- on % survival is very minor. Therefore this experiment would benefit from additional data using Etop and Dox as stimuli. Is this similar? We repeated the MCPIP1 MEF cell survival assay after IR and showed the new results in linear graph of survival percentage as suggested by the reviewer1 (suppl-fig 8B), this new data clearly supported the advantage of MCPIP1-deficiency in MEF cell survival after IR treatment. Moreover, we observed similar survival advantage of MCPIP1-/- MEFs in response to Dox treatment (suppl- Fig 8C). Minor comments: 1. The writing suffers from some grammatical and spelling errors; especially 'articles' are missing in many cases. We thank the reviewer s comment and have sought help form professional editorial service to improve the clarity and accuracy of our writing. We hope the reviewer find our revised manuscript improved. European Molecular Biology Organization 15

16 2. Legends should indicate how many times the experiments have been performed. Also if a mean value is given, the number of samples to calculate this mean should be provided. In addition, p- values should be indicated. Following the reviewer s suggestion, we have revised the figure legends accordingly. 3. In general, legends should be more explanatory. For some figures there is also no indication of time of stimulation (e.g. FigS1D and S1E). Which tags are on which proteins?... As we responded to the previous comment, we have revised the legends to increase the clarity. 4. Fig2C: also use the full name of CPT as it is also not mentioned in the text. We have included the full name of CPT, camptothecin, in the text and figure legends. 5. a schematic representation of the domain structure of MCPIP1 and indicating the location of several mutants used in this study would be very helpful. As the reviewer suggested, a diagram of MCPIP1 domain and annotation the critical residues has been included in Fig 4D. 6. From Fig2C the authors can't conclude that the absence of MCPIP1 leads to an extended duration of NF-κB binding (instead of activation), as only one time point is shown. In fact, from Fig2D it is clear that at the time used in Fig2C, the binding is not extended but rather enhanced. We agree that the time point used in Fig 2C does not reflect the impact of MCPIP1 on duration of NF-κB activation in response to DNA damage. As we shown in new Fig 2D, NF-κB activation, measured by EMSA, has almost returned to basal level at 16 h after Dox treatment while we still detected substantial NF-κB activation at the same time point in MCPIP1-deficient cells, supporting a role of MCPIP1 in affecting the duration of genotoxic NF-κB activation. We have revised our text accordingly. 7. Fig 3F-H. How is phosphorylation detected? Radioactive? Phospho-specific antibodies? The radioactive signal of kinase substrate phosphorylation was exposed to phosphorscreen and detected by cyclone phosphoimager. We have included this information in figure legends and respective methods section. 8. Fig. 6A: the NF-kB signal should be quantified as done for the other figures. Quantification data has been provided in the revised figure. 2nd Editorial Decision 22 July 2013 Your revised manuscript on USP10-MCPIP1 has now been evaluated once more by the original referees 1 and 3, as well as - subsequently - by an additional arbitrating referee with particular DUB expertise; you will find their comments copied below for your information. The reason for this procedure was that the original referees, while acknowledging significant overall improvement, still retained major reservations regarding the analyses of USP10 DUB activities and ubiquitin chain specificities. I therefore asked a trusted arbitrating referee to comment on these issues, and I am happy to inform you that they confirmed the overall significance of the presented findings, in particular the identification of MCPIP1 as a potentially new type of DUB adaptor. At the same time, referee 4 however too remained unconvinced by the current evidence for USP10 being a linear ubiquitin chain-specific DUB, and considers alternative interpretations - such as USP10 removing ubiquitin chains en bloc from NEMO - more likely. In addition, they raise a number of other issues regarding the analysis of DNA damage responses and apoptosis, as well as minor points regarding controls, presentation and explanations. European Molecular Biology Organization 16

17 Given the importance of the topic and of the already clearly supported data, I would be inclined to offer you an exceptional additional round of revision of this work, in order to substantiate the issues regarding the genotoxic stress/apoptosis involvement of USP10/MCPIP1, and especially to clarify the biochemical role of USP10 in NEMO deubiquitination. In this respect, the following points will be essential for eventual acceptance: - addressing referee 4's point major point 1 by conducting the requested time course experiment, and by reconsidering the conclusions on USP10 chain specificity. In case of reinterpretation along the lines suggested by referee 4, further attempts to biochemically verify USP10 chain specificity may not be necessary. - I agree with the referee that it may be best to leave out the mirna regulation data, which are currently anyway only part of a supplementary figure. - regarding referee 4's points 3-5, please try to improve these issues especially by utilizing more comparable conditions between different experiments as suggested. - please address all the various specific/minor points raised by all three referees, by altered discussion/presentation or by showing additional (control) data where appropriate. On the other hand, I don't feel it would be necessary to follow referee 1's suggestion to look at USP10 effects on p53, provided that the biochemical role of USP10 will be addressed as discussed earlier. - please consider moving a revised/corrected version of the summary model presented in Figure S8G into the main figures, either as a sub-panel or as an additional stand-alone figure. Please also include more of the Material and Methods in the main manuscript, as this section is currently predominantly hidden in the Supplement. - finally, please provide source data files for the various electrophoretic gels and blots, in order to make the primary data more accessible and to allow readers to judge the full gels from which the minimized panels were derived. We would ask for a single PDF/JPG/GIF file per figure comprising the original, uncropped and unprocessed scans of all gel/blot panels used in the respective main and supplementary figures. These should be labelled with the appropriate figure/panel number, and should have molecular weight markers; further annotation would clearly be useful but is not essential. A ZIP archive containing these individual files can be uploaded upon resubmission (selecting "Figure Source Data" as object type) and would be published online with the article as a supplementary "Source Data" file. Once you are finishing revising the manuscript, please follow the instructions and URL below to resubmit a detailed response letter and separate text, figure and SI files as detailed below. Of course, please do not hesitate to contact me should you have any questions or comments regarding the referee reports and this decision - especially since I have to emphasize that this will have to be the final round of revision for this manuscript at our Journal Referee #1 : In the latest version of their manuscript the authors have attempted to address many concerns raised by three reviewers. They have performed additional experiments and provided much needed clarifications for the text and figures. However, they have not addressed the catalytic selectivity of deubiquitinase (DUB) USP10. Reviewers 1 and 3 asked them to check whether USP10 can cleave polyubiquitin chains composed of linkages other than linear (which is expected and extremely common among USP family of deubiquitinases). However, this has not been addressed. Even more frustrating is their claim in response to point 12 from reviewer 3 that they have done it (but they have not) - only to again go into the statements about the cleavage of linear tetraubiquitin by USP10. This purposeful avoidance of an important issue raised by two reviewers suggests that USP10 may not cleave only linear ubiquitin chains. It is not clear why the authors are forcing the issues of linear polyubiquitin chain disassembly when it is well known that other chains, especially K63 are involved in DNA damage response and genotoxic stress. So far only 2 DUBs have been reported to European Molecular Biology Organization 17

18 cleave linear polyubiquitin chains - CYLD (nonselectively as it also cleaves K63 and other chains) and recently reported OTULIN/Gumby (which possesses striking specificity for linear chains). Thus, if USP10 is a linear ubiquitin chains specific DUB, this would be very important finding. However, since USP10 deubiquitinates p53 (Cell Sep 30;147(1):223-34), a protein that can be modified by a variety of linkages including K48, it is very unlikely that the observed activity of USP10 stems exclusively from its putative cleavage of linear polyubiquitin chains. Instead, it seems that potential cleavage of linear ubiquitin chains assembled on Nemo is just one of USP10 substrates and not exclusive or essential component of its function. Thus, they should check how USP10 knockdown affects p53 stability and ubiquitination following DNA damage stimuli. Specific points: USP10 knockdown did not really "substantially decrease" caspase-3 activity but merely delayed it - figure S8E. Similarly, USP10 depletion had very small effect on NF-kB regulated gene induction: the authors show examples of c-iap1/2 whose fold induction goes from 1.6x to 2.1x (S8F). How can that be significant? Maybe the authors should look for alternative genes whose expression is more severely affected during genotoxic stress response. Referee #3: The authors have extensively revised a previously submitted version of their manuscript, taking into account multiple referee comments and suggestions. Several additional experiments were included in the revised paper or the response to referees. Overall the manuscript has significantly improved and major conclusions are now well supported by the data that are reported. In my opinion, the paper contains many novel findings on a topic that is of high interest to the field, the more because MCPIP1 seems to be a key regulator of inflammatory responses. I only have some very minor comments: - in the text the authors mention the use of genotoxic drugs in Figure 3D, although no drugs are used in this experiment. Please correct. - in the legend of Fig3F-G it is not mentioned how long the cells were stimulated with TNF or CPT - in the legend of supplementary Fig3B the presence of PS-341 was mentioned. The reason for its use is not mentioned in the text, which may be confusing for some of the readers. Referee #4 : The manuscript by Niu et al reports that MCPIP1 acts an adaptor for USP10 to target NEMO, explaining the role of MCPIP1 in inhibiting NF B signaling. This work is important, as it clears an artifact from the literature, namely that MCPIP1 is a DUB itself. The following claims are well supported by the presented data: - MCPIP1 is induced after genotoxic stress - MCPIP1 negatively regulates NF B activation in an RNAse independent manner - MCPIP1 binds NEMO and affects NEMO ubiquitination - MCPIP1 binds USP10, and USP10 is the DUB for NEMO - MCPIP1 and USP10 are involved in apoptosis induced by genotoxic stress. These results are significant and worth publishing in EMBO J. To assess what type of Ub chains USP10 disassembles in NEMO, the authors make use of state-ofthe-art tools, namely the Ub replacement strategy, and a linear-specific antibody developed by the Iwai laboratory. This antibody is the only one that shows (long) linear Ub chains on NEMO, which many labs have not been able to observe. Unfortunately the antibody is not widely distributed, and it is still unclear what it actually recognizes. Main points: 1) The results seem to support that NEMO may be ubiquitinated with linear chains, but this does not make USP10 a Met1-DUB, but rather a NEMO DUB. Indeed, the authors do not provide convincing evidence that USP10 cleaves linear chains: the assay shown in Fig 6D suggests hardly any activity European Molecular Biology Organization 18

19 against linear chains, which is consistent with work by Sixma and colleagues that report low activity of USP DUBs against this chain type (Faesen et al, Chem & Biol, 2012). If the authors want to keep their claim that USP10 is specific for linear chains on NEMO, they would have to show specificity of the DUB, which is missing from this manuscript. I would be surprised if USP10 was not much more active against isopeptides. This had been requested by Reviewer 1, and was not addressed, and I agree with that Reviewer that this is unacceptable. Moreover, there is additional evidence that USP10 does not necessarily target chains. All assays in Fig 6 show complete removal of Ub chains by USP10. Although the quality of the linear Ub chains blots is quite poor, a time course of the in vitro cleavage reaction (Fig. 6E) could resolve whether over time the signal decreases gradually (ie 'from-the-top') or whether the entire smear disappears. Instead, the authors should consider the possibility that USP10/MCPIP is a NEMO-DUB, which cleaves the first linkage between NEMO and Ub rather than the chain itself. All of the data supports this, as NEMO does not seem to be modified any longer after USP10 is present. This is a simple issue of wording in the results and abstract, which currently implies USP10 to be linear specific. 2) The data on MCPIP in mirna stability is a massive distraction from the main story and should be removed. It does not add anything to the main findings, and decreases readability of the text. It is also not tested whether USP10 interaction affects MCPIPs RNase activity. This should be expanded and published elsewhere. 3) DNA damage part: there is no assessment of the extent of the DNA damage after Etoposide treatment, and variations of the extent of the damage could be significant. RNAse activity of MCPIP1 could modulate the DNA damage. What is missing are blots against gamma-h2ax, or any other marker of damage, to allow assessment of how severe the damage is. 4) The data for the apoptosis assays are not consistent and not well presented in the Figure and text. Why is etopoxide used with MCPIP overexpression and then compared to MCPIP -/- cells treated with IR or doxyrubicine (Figure S8a,b). It is also not shown that NEMO linear ubiquitination (or USP10-dependent deubiquitination) affects NFkB activity and apoptosis. The apoptosis part only focuses on MCPIP, which might be due to other roles of the protein, but should more focus on USP10. 5) Figure 3E: why is there such a big difference in the interaction between MCPIP and NEMO depending on the drug that was used (CPT and etoposide should have similar effects)? Doxorubicin should be included to make sure this is not an artifact, especially since this is a single experiment. In addition to these main points, there are numerous smaller issues with the text and presentation: 6) The in vitro DUB assay of MCPIP against all chain types should be shown. 7) The repression caused by MCPIP overexpression appears quite variable (see fig. 2 A-C and fig 4C, where repression in the latter is much less, and effect upon doxorubicin is greater than etoposide - just the opposite to what is observed in fig. 2). Please report statistics for these experiments. 8) Figure 5H: overexpression of USP10 should decrease NFkB activation, and it doesn't seem the case if comparing lanes 3 and 6, since it looks like they loaded more in lane 3 than 6, therefore the effect in both conditions should be more or less the same. 9) As a general comment, the abbreviations used in the Figures are inconsistent, and most are not described in the legends, or are described differently. It is impossible to understand the figures without reading all legends and the corresponding text! The Figures have to be labeled clearer, and single letter abbreviations should be removed. 10) In most Figures, loading controls are heavily overexposed and do not allow a judgment of equal European Molecular Biology Organization 19

20 loading. 11) Also, all blots that only show a single band size, the size marker is missing. 12) In some blots, the background is too light. In Figure S4A a darker exposure should be shown for instead of white boxes. 13) For all linear Blots after NEMO IP, they should put asterisks to the massive band below 56 kd, which may be Heavy Chain, because the IP Ab is rabbit as well as the Iwai linear chain Ab. However, this is also the size where NEMO runs, so it could be misunderstood in that this linear specific Ab would be able to recognize NEMO (see comment re antibody above). 14) Figure 2E would benefit if they would add in the Figure that they used Etopoxide there (for the different time points). 15) Figure 4E needs similar input amounts for Linear and K63 and maybe silver staining to avoid Ab specificity preferences towards one of the chain linkages. 16) In Figure 6c it seems as if USP10 is also upregulated after Etopoxide treatment, but this was never mentioned in the text nor explored further. 17) The pictures they chose to show in Figure 7F are not representative considering the quantification is correct. They would better show pictures with a more obvious difference. 18) Figure S8D is either missing a Caspase blot or the authors were referring to S8F instead (but then Figure S8D doesn't make any sense). 19) the English can be improved in many instances. 1st Revision - authors' response 20 August 2013 Referee #1: In the latest version of their manuscript the authors have attempted to address many concerns raised by three reviewers. They have performed additional experiments and provided much needed clarifications for the text and figures. However, they have not addressed the catalytic selectivity of deubiquitinase (DUB) USP10. Reviewers 1 and 3 asked them to check whether USP10 can cleave polyubiquitin chains composed of linkages other than linear (which is expected and extremely common among USP family of deubiquitinases). However, this has not been addressed. Even more frustrating is their claim in response to point 12 from reviewer 3 that they have done it (but they have not) - only to again go into the statements about the cleavage of linear tetraubiquitin by USP10. This purposeful avoidance of an important issue raised by two reviewers suggests that USP10 may not cleave only linear ubiquitin chains. It is not clear why the authors are forcing the issues of linear polyubiquitin chain disassembly when it is well known that other chains, especially K63 are involved in DNA damage response and genotoxic stress. So far only 2 DUBs have been reported to cleave linear polyubiquitin chains - CYLD (nonselectively as it also cleaves K63 and other chains) and recently reported OTULIN/Gumby (which possesses striking specificity for linear chains). Thus, if USP10 is a linear ubiquitin chains specific DUB, this would be very important finding. However, since USP10 deubiquitinates p53 (Cell Sep 30;147(1):223-34), a protein that can be modified by a variety of linkages including K48, it is very unlikely that the observed activity of USP10 stems exclusively from its putative cleavage of linear polyubiquitin chains. Instead, it seems that potential cleavage of linear ubiquitin chains assembled on Nemo is just one of USP10 substrates and not exclusive or essential component of its function. Thus, they should check how USP10 knockdown affects p53 stability and ubiquitination following DNA damage stimuli. We regret our negligence of leaving the data of USP10 in vitro DUB assay for K63 chain out of our previous revision. Now we show in revised Fig S6C, USP10 cleaves K63-linked tetraubiquitin in European Molecular Biology Organization 20

21 vitro with higher activity than that of USP10-dependent cleavage of linear tertraubiqutin, confirming that USP10 harbors promiscuous DUB activity to cleave ubiquitin chains with different linkages. Specific points: USP10 knockdown did not really "substantially decrease" caspase-3 activity but merely delayed it - figure S8E. Similarly, USP10 depletion had very small effect on NF-kB regulated gene induction: the authors show examples of c-iap1/2 whose fold induction goes from 1.6x to 2.1x (S8F). How can that be significant? Maybe the authors should look for alternative genes whose expression is more severely affected during genotoxic stress response. We repeated the gene expression quantification experiments in USP10-depleted cells. We found c- IAP1/2 induction was consistently further increased in USP10-depleted cells, compared with WT cells, in response to genotoxic treatment, which is statistically significant. Following the reviewer s suggestion, we also examined the induction of Bcl-xL by genotoxic treatment, which also showed consistent enhancement in USP10-depleted cells. These data are presented in revised Fig S8F. Referee #3: I only have some very minor comments: - in the text the authors mention the use of genotoxic drugs in Figure 3D, although no drugs are used in this experiment. Please correct. We have revised the text accordingly. - in the legend of Fig3F-G it is not mentioned how long the cells were stimulated with TNF or CPT The time of treatments has been included in revised legend. - in the legend of supplementary Fig3B the presence of PS-341 was mentioned. The reason for its use is not mentioned in the text, which may be confusing for some of the readers. We have revised the legend of Fig S3B as the reviewer suggested to increase the clarity. Referee #4: Main points: 1) The results seem to support that NEMO may be ubiquitinated with linear chains, but this does not make USP10 a Met1-DUB, but rather a NEMO DUB. Indeed, the authors do not provide convincing evidence that USP10 cleaves linear chains: the assay shown in Fig 6D suggests hardly any activity against linear chains, which is consistent with work by Sixma and colleagues that report low activity of USP DUBs against this chain type (Faesen et al, Chem & Biol, 2012). If the authors want to keep their claim that USP10 is specific for linear chains on NEMO, they would have to show specificity of the DUB, which is missing from this manuscript. I would be surprised if USP10 was not much more active against isopeptides. This had been requested by Reviewer 1, and was not addressed, and I agree with that Reviewer that this is unacceptable. Moreover, there is additional evidence that USP10 does not necessarily target chains. All assays in Fig 6 show complete removal of Ub chains by USP10. Although the quality of the linear Ub chains blots is quite poor, a time course of the in vitro cleavage reaction (Fig. 6E) could resolve whether over time the signal decreases gradually (ie 'from-the-top') or whether the entire smear disappears. Instead, the authors should consider the possibility that USP10/MCPIP is a NEMO-DUB, which cleaves the first linkage between NEMO and Ub rather than the chain itself. All of the data supports this, as NEMO does not seem to be modified any longer after USP10 is present. This is a simple issue of wording in the results and abstract, which currently implies USP10 to be linear specific. European Molecular Biology Organization 21

22 Following the reviewer s suggestion, we performed additional in vitro NEMO deubiquitination assay using recombinant USP10. As now shown in revised Fig. 6E, USP10 incubation efficiently decreased NEMO-attached polyubiquitin en bloc, instead of shortening the polyubiquitin chain on NEMO. Also we found USP10 cleaved isopeptide linkage, such as K63-linkage, more efficiently than cleaving linear peptide bond (Fig S6C). These data support the notion as the reviewer suggested that USP10 may function as a DUB primarily cleaving the entire polyubiquitin chain off NEMO, thereby inhibiting genotoxic NF-κB signaling. We have revised the manuscript accordingly. 2) The data on MCPIP in mirna stability is a massive distraction from the main story and should be removed. It does not add anything to the main findings, and decreases readability of the text. It is also not tested whether USP10 interaction affects MCPIPs RNase activity. This should be expanded and published elsewhere. We have removed this portion of data from the manuscript. 3) DNA damage part: there is no assessment of the extent of the DNA damage after Etoposide treatment, and variations of the extent of the damage could be significant. RNAse activity of MCPIP1 could modulate the DNA damage. What is missing are blots against gamma-h2ax, or any other marker of damage, to allow assessment of how severe the damage is. We preformed immunoblotting experiments to detect H2AX phosphorylation in HEK293 cells exposed to Etoposide. Surprisingly, we observed significant increase of γ-h2ax in MCPIP1- overexpressed HEK293 cells (shown below), which suggested MCPIP1 may play a role in regulating DNA repair. We will further explore the potential mechanisms and report the data in future publications. 4) The data for the apoptosis assays are not consistent and not well presented in the Figure and text. Why is etopoxide used with MCPIP overexpression and then compared to MCPIP -/- cells treated with IR or doxyrubicine (Figure S8a,b). It is also not shown that NEMO linear ubiquitination (or USP10-dependent deubiquitination) affects NFkB activity and apoptosis. The apoptosis part only focuses on MCPIP, which might be due to other roles of the protein, but should more focus on USP10. In our previous studies, we found, unlike in HEK293 cells, etoposide generally do not induce robust NF-κB activation in MEF cells. Therefore, we chose treatment with Dox or IR, which effectively induce NF-κB activation in MEFs, to study the impact of interfering genotoxic NF-κB activation on MEF cell apoptosis in response to DNA damage. Our previously publication (Niu et al. EMBO J, 2011) and additional data (Fig. S6F) support a critical role of LUBAC-mediated NEMO linear ubiquitination in regulating genotoxic NF-κB activation. Our data also indicate that depleting USP10 substantially increased NF-κB-regulated anti-apoptotic gene induction (Fig. 7D and S8F) and decreased caspases activation in cells treated with genotoxic agents (Fig. S8E). We hope the reviewer found these data supportive and consistent with our conclusion. 5) Figure 3E: why is there such a big difference in the interaction between MCPIP and NEMO depending on the drug that was used (CPT and etoposide should have similar effects)? Doxorubicin should be included to make sure this is not an artifact, especially since this is a single experiment. European Molecular Biology Organization 22