IRAK-M mediates Toll-like receptor/il-1r-induced NFκB activation and cytokine production

Transcription

1 The EMBO Journal Peer Review Process File - EMBO Manuscript EMBO IRAK-M mediates Toll-like receptor/il-1r-induced NFκB activation and cytokine production Hao Zhou, Mingjia Yu, Koichi Fukuda, Jinteak Imm, Peng Yao, Wei Cui, Katarzyna Bulek, Jarod Zepp, Youzhong Wan, Tae Whan Kim, Weiguo Yin, Victoria Ma, James Thomas, Jun Gu, Jian-an Wang, Paul DiCorleto, Paul Fox, Dr. Jun Qin (II) and Xiaoxia Li Corresponding author: Xiaoxia Li, Lerner Research Institute Review timeline: Submission date: 02 July 2012 Editorial Decision: 09 August 2012 Revision received: 14 November 2012 Accepted: 18 December 2012 Editor: Karin Dumstrei 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.) 1st Editorial Decision 09 August 2012 Thank you for submitting your manuscript to the EMBO Journal. Your study has now been seen by three referees and their comments are provided below. While referee #3 is not convinced that the paper is a good fit for publication in the EMBO Journal, both referees # 1 and 2 find that the paper makes an important contribution and are more supportive. Given the comments provided by referee #1 and 2, I would like to invite a revision should you be able to address the concerns raised in full. Please also keep in mind the specific points raised by referee #3. I should add that it is EMBO Journal policy to allow only a single major round of revision and that it is therefore important to resolve the raised issues at this stage. When preparing your letter of response to the referees' comments, please bear in mind that this will form part of the Review Process File, and will therefore be available online to the community. For more details on our Transparent Editorial Process, please visit our website: Thank you for the opportunity to consider your work for publication. I look forward to your revision. European Molecular Biology Organization 1

2 The EMBO Journal Peer Review Process File - EMBO REFEREE REPORTS Referee #1 Understanding the molecular mechanism that regulate TLR-IL-1 responses is of significant interest. The manuscript of Zhou al describes a series of in vitro experiments on the role of IRAK-M in the regulation of TLR (TLR2,7, 9) and IL-1 responses. IRAK-M is known for some time as a negative regulator of TLR/IL-1R induced NF-kB activation by preventing the release of the IRAK1/2 complex from the receptor and its physiological significance has been validated in IRAK-M knockout mice and models of inflammation and infection. Using cells that are deficient for different IRAK family members, the authors now report that IRAK-M can also directly mediate TLR/IL-1R signaling. These experiments are further supported by biochemical data showing that IRAK-M, similar to IRAK1/2, can interact with MyD88/IRAK4. More specifically, they report that IRAK-M triggers TAK-1 independent, but MEKK3-dependent NF-kB activation, which is mainly responsible for a second wave of NF-kB signaling and uncoupled from posttranscriptional control mechanisms. As this pathway seems to regulate some negative regulatory proteins, it is suggested that the inhibitory effect of IRAK-M on TLR signaling may be partially explained by IRAK-M mediated induction of other NF-kB inhibitory proteins. In addition, they show that IRAK-M, by interacting with IRAK-2, also negatively regulates translation of cytokines and chemokines. The experimental design is overall sound. The data are of good quality and support the main conclusions. Overall, these data significantly advance our knowledge on the role of this important mediator (IRAK-M) and the mechanism by which it exerts its function. I only have some minor remarks that should be addressed. 1. The authors start their story using TLR7 stimulation of different IRAK-deficient cells (Fig. 1) and then switch to IL-1R stimulation to study the underlying biochemical mechanism. Also in the rest of the paper they regularly make this switch between TLR7 and IL-1R. This is somewhat confusing. The authors should first do their biochemical studies (Fig. 2) with TLR7 stimulation, and may then confirm this with IL-1R stimulation. 2. page 4, last sentence of 1st section: the authors mention that TLR-2 and TLR-9 induced phosphorylation of IKKa/b is abolished in IRAK1/2/M -TKO- BMDMs, whereas retained in IRAK1/2-DKO cells. This is wrong as suppl. Fig. 1 shows that it is also abolished in IRAK1/2 DKO cells. 3. on page 6, line 5, the authors mention that IRAK-M deficiency decreased TLR7-induced late phase NF-kB activation (after 30 min) in a gel shift assay (Fig. 3A). If 30 min is late phase, what is the timing for the early phase. The figure only shows 0.2 h as an earlier time point, which also shows reduced NF-kB activation. Please clarify. 4. page 7, line 4: it is mentioned that most of the IRAK-M dependent genes were late genes (induced after 1 hour). However, A20 is known to be an early response gene as is also illustrated in Fig. 5. So, is A20 an exception? The authors should at least mention this. 5. page 7, line 18: IRAK-M KO BMDMs had more cytokine and chemokine mrnas in the translation-inactive pool. Should this not read 'translation-active pool'? 6. Fig. 7F: the increased phosphorylation of MNK1 and eif4e is not clear. A better experiment is needed, or at least quantification of the blots should be shown. Referee #2 Previous studies have suggested that IRAK-M inhibits TLR signaling by preventing the dissociation of IRAKs from MyD88 and the formation of IRAK-TRAF6 complex. In this paper, Zhou et al reveals a positive role of IRAK-M in NF-kB activation. They showed that the residual activation of European Molecular Biology Organization 2

3 The EMBO Journal Peer Review Process File - EMBO NF-kB by TLR7 in IRAK1/2 dko mice was abolished by further deletion of IRAK-M. IRAK-M could form a MyDDosome complex with MyD88 and IRAK4, and it could also interact with MEKK3, which leads to IKK activation. The authors suggest that the IRAK-M - MEKK3 axis induces a second wave of NF-kB activation, which turns on the expression of certain genes that are dependent on mrna stabilization by IRAK4. Several of these genes appear to be the negative regulator of NF-kB, explaining in part the hyper-inflammatory phenotypes of IRAK-M ko mice. The authors also show that IRAK-M interacts with IRAK2 and inhibits the translation of several inflammatory genes, providing another explanation for the anti-inflammatory effect of IRAK-M. These results are interesting, and appear to resolve the paradoxical role of IRAK-M in activating NF-kB and suppressing inflammation. Several points need to be addressed: 1) The only evidence that implicates MEKK3 in the 'IRAK-M pathway' is the binding of IRAK-M and MEKK3. More evidence is needed. I can suggest two experiments: a) to test whether the residual activity of NF-kB/p-IkB in IRAK1/2 KO cells could be abolished by depleting MEKK3; b) to test if IRAK-M could still activate NF-kB in 293-I1A cells (Irak1/2-deficient) in cells depleted of MEKK3 or TAK1 (i.e, by RNAi). 2) The authors' results do not exclude a simpler model that could explain their results and the previous phenotypes of IRAK-M ko mice. In this model, IRAK-M is a weaker activator of IKK and it competes with IRAK1/2 for binding to IRAK4. Thus, in the absence of IRAK-M, more IRAK1/2 are recruited to the MyDDosome, leading to stronger IKK activation (e.g, see phosphor-ikka/b in Fig. 3B; see also Kobayashi 2002). In the absence of IRAK1/2, IRAK-M could partially rescue IKK activation, thus appearing as a positive regulator. 3) Based on Figure 5B, the authors suggested that A20 and IkBa levels were lower in IRAK-M ko cells than those in WT cells. However, the western blot in figure 3B showed that IkBa protein level IRAK-M ko cells was comparable if not higher than that in WT cells. 4) Figure 7F, the authors claimed that there was more MNK1 and peif4e phosphorylation in IRAK- M ko cells, but the results are not convincing. I wonder if quantitation of the results would show any significant difference. Also, was MKK3/6 phosphorylated in the absence of stimulation? Minor point: The title is vague; it should state concisely the major new findings of this paper (what are the novel mechanisms?) Referee #3 In this paper Zhou et al argue for a signaling rather than an inhibitory role for IRAKM (aka IRAK3). A general point is that the conclusions are made on the basis of double KOs of IRAK1 and2 and given the hierarchical nature of the Myddosome assembly there is a danger that functions are being unmasked that are not relevant in an IRAK1/2 + background. Many of the conclusions rely on overexpression of transfected genes and semi quantitative methods such as Western blot and EMSA. The first experiments look at the activation of NFkB in double and triple IRAK1,2 M knockout BMDMs using EMSAs. The authors need to show the unbound oligos on the gel so that it is clear that there is a different ratio in the KO cells. Fig 1B shows some predictable patterns for standard TLR?IL1 signalling markers such pikbalpha although they do not have a blot showing what happens to IRAKM. Overall this figures adds little to understanding of the role of IRAKs in signaling. In Fig 2 different cells (MEFS) and a different stimulus (IL1) are used for pull down assays of transfected, overexpressed IRAKM and on the face of it complexes with MyD88, MEKK and TRAF 6 are detected although whether these complexes also contain IRAK1 or 2 is not addressed. Curiously in relation to this they state that MEKK modification (?phosphorylation) is retained in European Molecular Biology Organization 3

4 The EMBO Journal Peer Review Process File - EMBO IRAK1/2 KOs referring to the EMSA in Fig1A. I can see no evidence for this anywhere in Fig 1 or indeed Fig 2. On the basis of the pulldowns they try to draw the conclusion that IRAKM assembles into a Myddosome. They say that IRAKM 'may adopt a similar fold (as IRAK4) ' as if this is some special insight - it is clearly a death domain - and then they do some crude modeling, replacing the IRAK2 subunits in the Myddosome. On the basis of this they home in on some conserved residues around the type 2a interface. Mutagenesis of these residues, unsurprisingly, affects function (there are numerous other interactions that contribute to the Myddosome and thetype 3 interface would be a better one to target). None of this proves that IRAKM forms a signaling Myddosome. They would have to do some assembly experiments and in any case evidence points to IRAK1 and M forming a fourth layer after IRAK2 (unpublished). Moving on to Fig 3 we are back to TLR7 stimulus and BMDMs. In Fig. 3A an EMSA is presented that shows a reduced level of early and late phase NFkB in IRAK M KOs. There are no loading controls again and the differences look ca 2-3 fold so it is impossible to assess the significance of this result. In Fig 3B there do seems to be some differences in late p-ikbalpha and p-ikk but W. blots are not very quantitative and these experiments would need to be repeated several times. A curiosity is that in Fig. 3A, IkBalpha completely disappears by 10 mins and reappears again as a very strong band at 30 mins, while p-ikba, which should be detected just as well by the anti-ikba antibody is strongly induced. This is suspicious as in Fig 3C which should be essentially the same experiment there is no reduction in total IkBalpha at 10mins at all. Maybe photoshop has gone a bit haywire here? And in Fig 3C itself they use a proteasome inhibitor MG132 at a massive 20mM - never mind the off target effects I guess that concentration would kill the cells stone dead. The next three figures present RT-PCR and some cytokine ELISA experiments looking at the level of various cytokines, chemokines and regulatory molecules. The main claim here is that the inhibitory molecules such as A20 are somewhat reduced and the argument is made that IRAKM has some special negative role mediated by the production of these but in the IRAK2,1 KOs they are already significantly reduced so this does not convince for a specific role of IRAKM. In any case this work coheres poorly with the previous results. The final Figure 7 shows a crude fractionation of ribosomes from WT and IRAKM Kos and RT- PCRs of various RNAs. They find ca 2 fold differences in Fig 7B for TNF, IL6 and KC to argue that IRAKM is involved in translational repression. This would appear to be a single experiment (or replicates). It would require several independent repeats to be a credible result. Fig 7E is back to MEFs and IL1 - it shows that IRAKM forms a complex with IRAK2! (see my point above about the myddosome). In conclusion this study seems to be made up of a number of components that do not fit together very well and does not make a coherent argument to support the grand overview presented in Fig. 8. I found this paper very difficult to read - many sentences are difficult to understand as they are constructed of cumbersome and lengthy adjectival clauses. 2nd Revision - authors' response 14 November 2012 European Molecular Biology Organization 4

5 Point-by-point response: Referee #1 Understanding the molecular mechanism that regulate TLR-IL-1 responses is of significant interest. The manuscript of Zhou al describes a series of in vitro experiments on the role of IRAK-M in the regulation of TLR (TLR2,7, 9) and IL-1 responses. IRAK-M is known for some time as a negative regulator of TLR/IL-1R induced NF-kB activation by preventing the release of the IRAK1/2 complex from the receptor and its physiological significance has been validated in IRAK-M knockout mice and models of inflammation and infection. Using cells that are deficient for different IRAK family members, the authors now report that IRAK-M can also directly mediate TLR/IL-1R signaling. These experiments are further supported by biochemical data showing that IRAK-M, similar to IRAK1/2, can interact with MyD88/IRAK4. More specifically, they report that IRAK-M triggers TAK-1 independent, but MEKK3-dependent NF-kB activation, which is mainly responsible for a second wave of NF-kB signaling and uncoupled from posttranscriptional control mechanisms. As this pathway seems to regulate some negative regulatory proteins, it is suggested that the inhibitory effect of IRAK-M on TLR signaling may be partially explained by IRAK-M mediated induction of other NF-kB inhibitory proteins. In addition, they show that IRAK-M, by interacting with IRAK-2, also negatively regulates translation of cytokines and chemokines. The experimental design is overall sound. The data are of good quality and support the main conclusions. Overall, these data significantly advance our knowledge on the role of this important mediator (IRAK-M) and the mechanism by which it exerts its function. I only have some minor remarks that should be addressed. 1. The authors start their story using TLR7 stimulation of different IRAK-deficient cells (Fig. 1) and then switch to IL-1R stimulation to study the underlying biochemical mechanism. Also in the rest of the paper they regularly make this switch between TLR7 and IL-1R. This is somewhat confusing. The authors should first do their biochemical studies (Fig. 2) with TLR7 stimulation, and may then confirm this with IL-1R stimulation. We appreciate the point raised by this reviewer. We initially had technical problems in transducing tagged- IRAKM into BMDMs by retroviral infection. We have now finally succeeded in adenoviral infection on BMDMs. We restored FLAG-IRAKM expression in IRAK1/2/M TKO macrophages and treated them with TLR7 ligand. The data are shown in Fig. 2B, 2H-J (Page 5-6, red). 1

6 2. page 4, last sentence of 1st section: the authors mention that TLR-2 and TLR-9 induced phosphorylation of IKKa/b is abolished in IRAK1/2/M -TKO- BMDMs, whereas retained in IRAK1/2-DKO cells. This is wrong as suppl. Fig. 1 shows that it is also abolished in IRAK1/2 DKO cells. We apologize for the incorrect description in the text. TLR-2/9 induced phosphorylation of IKK α/β was indeed abolished in IRAK1/2 DKO cells. We meant to describe phosphorylation of IkBa. We have corrected the description in the text: Similarly, TLR2- and TLR9-induced phosphorylation of IκBa were still retained in IRAK1/2-DKO-BMDMs, which was completely abolished in IRAK1/2/M-TKO-BMDMs (Supple. Fig. 1), implicating the participation of IRAKM in TLR2- and TLR9-mediated signaling. However, it is important to point out, whereas TLR9-induced IκBa phosphorylation in IRAK1/2-DKO-BMDMs was comparable to that in wild-type cells, TLR2-induced IκBa phosphorylation in IRAK1/2-DKO-BMDMs was substantially reduced compared to that in wild-type cells, suggesting differential utilization of IRAKM by different TLRs (Supple. Fig. 1, Page 4, red). 3. on page 6, line 5, the authors mention that IRAK-M deficiency decreased TLR7-induced late phase NF-kB activation (after 30 min) in a gel shift assay (Fig. 3A). If 30 min is late phase, what is the timing for the early phase. The figure only shows 0.2 h as an earlier time point, which also shows reduced NF-kB activation. Please clarify. We understand the reviewer s concern. We have now repeated the experiments and included more early time points (0.1, 0.2, and 0.5h) in the gel shift assay (Fig. 3A), showing no difference between wild-type and IRAKM KO. 4. page 7, line 4: it is mentioned that most of the IRAK-M dependent genes were late genes (induced after 1 hour). However, A20 is known to be an early response gene as is also illustrated in Fig. 5. So, is A20 an exception? The authors should at least mention this. We understand the reviewer s point. We have now mentioned in the text for the fact that A20 was induced at 30 minutes (Page 7, red). Although A20 was induced at 30 min time point, the peak of induction was at 4h time point (Fig. 5A). The IRAKM-mediated pathway did contribute to its induction as shown in Fig. 5B. 5. page 7, line 18: IRAK-M KO BMDMs had more cytokine and chemokine mrnas in the translation-inactive pool. Should this not read 'translation-active pool'? We apologize about this typo. It should be 'translation-active pool'. We have corrected it in the text (Page 7, red). 6. Fig. 7F: the increased phosphorylation of MNK1 and eif4e is not clear. A better experiment is needed, or at least quantification of the blots should be shown. We have now quantified the blots by ImageJ software for Fig. 7F. Referee #2 Previous studies have suggested that IRAK-M inhibits TLR signaling by preventing the dissociation of IRAKs from MyD88 and the formation of IRAK-TRAF6 complex. In this paper, Zhou et al reveals a positive role of IRAK-M in NF-kB activation. They showed that the residual activation of NF-kB by TLR7 in IRAK1/2 dko mice was abolished by further deletion of IRAK-M. IRAK-M could form a MyDDosome complex with MyD88 and IRAK4, and it could also interact with MEKK3, which leads to IKK activation. The authors suggest that the IRAK-M - MEKK3 axis induces a second wave of NF-kB activation, which turns on the expression of certain genes that are dependent on mrna stabilization by IRAK4. Several of these genes appear to be the negative regulator of NF-kB, explaining in part the hyper-inflammatory phenotypes of IRAK-M ko mice. The authors also show that IRAK-M interacts with IRAK2 and inhibits the translation of several inflammatory genes, providing another explanation for the anti-inflammatory effect of IRAK-M. These results are interesting, and appear to resolve the paradoxical role of IRAK-M in activating NF-kB and suppressing inflammation. Several points need to be addressed: 2

7 1. The only evidence that implicates MEKK3 in the 'IRAK-M pathway' is the binding of IRAK-M and MEKK3. More evidence is needed. I can suggest two experiments: a) to test whether the residual activity of NF-kB/p-IkB in IRAK1/2 KO cells could be abolished by depleting MEKK3; b) to test if IRAK-M could still activate NF-kB in 293-I1A cells (Irak1/2-deficient) in cells depleted of MEKK3 or TAK1 (i.e, by RNAi). This point is well taken. MEKK3 was knocked down in 293-I1A cell by shrna. IRAKM-mediated NFκB activation was almost abolished in MEKK3 knock-down cells compared to that in the control cells (Fig. 3E-F, Page 6-7, red). 2. The authors' results do not exclude a simpler model that could explain their results and the previous phenotypes of IRAK-M ko mice. In this model, IRAK-M is a weaker activator of IKK and it competes with IRAK1/2 for binding to IRAK4. Thus, in the absence of IRAK-M, more IRAK1/2 are recruited to the MyDDosome, leading to stronger IKK activation (e.g, see phosphor-ikka/b in Fig. 3B; see also Kobayashi 2002). In the absence of IRAK1/2, IRAK-M could partially rescue IKK activation, thus appearing as a positive regulator. Reviewer s simpler model could be part of our model. It is correct the for the early time points. TAK1 pathway may compensate the loss of MEKK3 pathway, which leads to hyper phosphorylation of IKK and JNK. However, this model will not explain the loss of NFκB activation and gene expression in IRAKM KO (Fig. 5B). Furthermore, we also showed that the inhibitory effect on the inflammatory cytokines and chemokines is at the posttranscriptional levels (Fig. 6-7). 3. Based on Figure 5B, the authors suggested that A20 and IkBa levels were lower in IRAK-M ko cells than those in WT cells. However, the western blot in figure 3B showed that IkBa protein level IRAK-M ko cells was comparable if not higher than that in WT cells. We understand the reviewer s concern. Our interpretation is the following. IκBα was phosphorylated and degraded in WT, whereas late phase IκBα phosphorylation/degradation was blocked in IRAKM KO. Therefore, although the re-synthesis of IκB was lower in IRAKM KO, the steady-state levels of IκB at late time points were similar to that in WT. 4. Figure 7F, the authors claimed that there was more MNK1 and peif4e phosphorylation in IRAK-M ko cells, but the results are not convincing. I wonder if quantitation of the results would show any significant difference. Also, was MKK3/6 phosphorylated in the absence of stimulation? Minor point: This point is well taken. We have now quantified the blots by ImageJ software for Fig. 7F. The result was significant. The higher band in p-mkk3/6 western blot is the non-specific band. MKK3/6 was not phosphorylated in the absence of stimulation. The title is vague; it should state concisely the major new findings of this paper (what are the novel mechanisms?) How about: IRAKM Myddosome mediates second wave of TLR-IL-1R-induced NFκB activation and modulates cytokine production. We welcome suggestions of titles. Referee #3 In this paper Zhou et al argue for a signaling rather than an inhibitory role for IRAKM (aka IRAK3). A general point is that the conclusions are made on the basis of double KOs of IRAK1 and2 and given the hierarchical nature of the Myddosome assembly there is a danger that functions are being unmasked that are not relevant in an IRAK1/2 + background. Many of the conclusions rely on overexpression of transfected genes and semi quantitative methods such as Western blot and EMSA. The first experiments look at the activation of NFkB in double and triple IRAK1,2 M knockout BMDMs using EMSAs. The 3

8 authors need to show the unbound oligos on the gel so that it is clear that there is a different ratio in the KO cells. Fig 1B shows some predictable patterns for standard TLR? IL1 signalling markers such pikbalpha although they do not have a blot showing what happens to IRAKM. Overall this figures adds little to understanding of the role of IRAKs in signaling. In EMSAs, the labeled probe is always added in excess (which often leads to over-exposure of the free probe band at the bottom of the gel on the films). In EMSA, the DNA-binding protein(s) in question is often the limiting factor, which allows quantitative assessment for the binding ability of the protein complex (NFκB in this case). Usually we let the free probe run out of gel in our experiments. According to the reviewer s suggestion, we re-ran the gel for EMSAs and the whole films are now shown as an attached Figure below. In Fig 2 different cells (MEFS) and a different stimulus (IL1) are used for pull down assays of transfected, overexpressed IRAKM and on the face of it complexes with MyD88, MEKK and TRAF 6 are detected although whether these complexes also contain IRAK1 or 2 is not addressed. We have now done the experiment in IRAK1/2/M-TKO macrophages with restored IRAKM by adenoviral infection following co-immunoprecipitation assays upon TLR7 ligand stimulation (Fig. 2B). Curiously in relation to this they state that MEKK modification (?phosphorylation) is retained in IRAK1/2 KOs referring to the EMSA in Fig1A. I can see no evidence for this anywhere in Fig 1 or indeed Fig 2. The reviewer may have missed the data presented in Western blots shown in Fig. 1B; Fig. 3B; Fig. 3C; Fig. 3D. We never show MEKK3 shift by EMSA in Fig. 1A. The reviewer might be mistaken here for the gel-shift assay?? On the basis of the pulldowns they try to draw the conclusion that IRAKM assembles into a Myddosome. They say that IRAKM 'may adopt a similar fold (as IRAK4) ' as if this is some special insight - it is clearly a death domain - and then they do some crude modeling, replacing the IRAK2 subunits in the Myddosome. We respectively disagree with this comment. The claim of IRAKM-Myddosome is based on abundant genetic and biochemical evidences (Fig. 1B; Fig. 2B; Fig2H-I; Fig. 3A-B and Fig. 5A-B). The modeling was only part of the study to help us design structure-functional analysis. Based on the modeling, we generated very useful point mutants of IRAKM, which were then put back into IRAK1/2/M-TKO macrophages to perform structurefunctional analysis. These studies provide novel insight into the mechanistic role of IRAKM. By the way, the IRAKM DD was modeled based on IRAK2 DD, not IRAK4 DD. It is important to point out, we have previously reported that the DD of IRAK4 is functionally distinct from that in IRAK1 (Fraczek et al, 2008), IRAK2 and IRAKM (unpublished data). On the basis of this they home in on some conserved residues around the type 2a interface. Mutagenesis of these residues, unsurprisingly, affects function (there are numerous other interactions that contribute to the Myddosome and the type 3 interface would be a better one to target). None of this proves that IRAKM forms a signaling Myddosome. They would have to do some assembly experiments and in any case evidence points to IRAK1 and M forming a fourth layer after IRAK2 (unpublished). As stated above, based on the modeling, we generated very useful point mutants of IRAKM, which were then put back into IRAK1/2/M-TKO macrophages to perform structure-functional analysis. These studies are very solid and novel. We observed interaction of IRAKM with IRAK4 and MyD88 in the absence of IRAK1 and IRAK2 (Fig. 2B) and this interaction was disrupted by point mutations predicted by our modeling (Fig. 2H). These studies clearly demonstrate the formation of MYD88-IRAK4-IRAKM complex (IRAKM Myddosome) in the absence of IRAK1 and IRAK2. We do agree with the reviewer that IRAKM can also form a fourth layer with IRAK2, which was to inhibit protein translational control (see our proposed model in Fig. 8 and data in Fig. 6-7). However, this does not exclude the ability of IRAKM to directly interact with IRAK4-MyD88, which is strongly supported by all the genetic and biochem evidences (Fig. 1B; Fig. 2B; Fig2H-J; Fig. 3A-B and Fig. 5A-B). 4

9 We would like to further comment on the specific mutations that we generated. We mutated the residues in interface I and II based on the Nature paper by Wu group since "the type I and type II interactions are more extensive in buried surface area than the type III interaction, Nature 465, page 886, right column, middle). Our mutation experiment proves the point so we did not think mutation of the type III was necessary. However, IRAK4-F25 residue may participate in two interfacial regions out of the "three surface types": type I, II, and III upon Myddosome formation (Nature 465 page 887, right column, line 16 from the top; "Unexpectedly, the F25D mutation not only..."). We mentioned the significance of the IRAK4 F25 in the interface in the text that the IRAK4 F25 backbone oxygen interacts with IRAKM Q78. We now made IRAK4 F25D mutant and retrovirally expressed in IRAK4-deficient fibroblast together with IRAKM. Compared to IRAK4 wild-type, IRAK4 F25D mutant showed reduced interaction with IRAKM upon IL-1 stimulation (Supple. Fig. 2). Moving on to Fig 3 we are back to TLR7 stimulus and BMDMs. In Fig. 3A an EMSA is presented that shows a reduced level of early and late phase NFkB in IRAK M KOs. There are no loading controls again and the differences look ca 2-3 fold so it is impossible to assess the significance of this result. We have repeated this experiment for Fig. 3A as described above. The results are highly reproducible. As far as we understand the assay, it is not common practice to include loading control, since the intensity of the shifted bands are determined by the amounts and binding ability of the protein complex in question (NFκB in this case). The probe is always in access in this assay. The way we control the experiments is the following. We have always measured the protein concentrations for each sample and used the same amounts of protein for all the samples in one gel-shift assay. We have included more detailed description about the gel-shift assay in our material and methods. We have repeated the experiments more than five times. The data are highly reproducible. In Fig 3B there do seems to be some differences in late p-ikbalpha and p-ikk but W. blots are not very quantitative and these experiments would need to be repeated several times. We have repeated the experiments more than five times. The data are highly reproducible. A curiosity is that in Fig. 3B, IkBalpha completely disappears by 10 mins and reappears again as a very strong band at 30 mins, while p-ikba, which should be detected just as well by the anti-ikba antibody is strongly induced. This is suspicious as in Fig 3C which should be essentially the same experiment there is no reduction in total IkBalpha at 10mins at all. Maybe photoshop has gone a bit haywire here? It is difficult to directly compare p-iκb blots with total IκB blots, since they are probed with different antibodies with different affinities. Our interpretation for Fig. 3B is the following. The residual amount of IκB at 10 min in Fig. 3B was probably highly phosphorylated (in terms of percentage of phosphorylated molecules), which was detected by the p-ikba antibody. However, the level of total IkB protein at that time point was too low to be seen with the IkB antibody under the condition that we performed the western blots (developing time and exposure time). This is very common phenomena in our more than 10 years of study with IkB and p-ikb analysis. In Fig. 3C, IκBα blot did not show complete degradation upon TLR7 stimulation as shown in Fig. 3B. It was due to the variations among different experiments. It is also important to point out DMSO was added in the Fig. 3C experiment. IκBα was still degraded in Fig. 3C, although it was slower than that in Fig. 3B. And in Fig 3C itself they use a proteasome inhibitor MG132 at a massive 20mM - never mind the off target effects I guess that concentration would kill the cells stone dead. 5

10 We thank this reviewer for pointing out this terrible mistake!! 20mM is the stock concentration. Working concentration is 20μm. We have corrected it in the text. The next three figures present RT-PCR and some cytokine ELISA experiments looking at the level of various cytokines, chemokines and regulatory molecules. The main claim here is that the inhibitory molecules such as A20 are somewhat reduced and the argument is made that IRAKM has some special negative role mediated by the production of these but in the IRAK2,1 KOs they are already significantly reduced so this does not convince for a specific role of IRAKM. In any case this work coheres poorly with the previous results. We would appreciate if the reviewer would re-review our data about the role of IRAKM shown in IRAKM-KO and IRAK1/2/M-TKO. Our results here also suggest that IRAKM plays an overall negative role in TLR-IL-1R signaling. However, the molecular mechanism is different from what is in the literature. In this manuscript, we provided two novel mechanisms for the regulatory role of IRAKM in TLR signaling (Fig. 8). First, IRAKM is able to function as an intermediate signaling component to transmit signaling upon TLR activation (Fig. 1B; Fig. 2B; Fig2H-J; Fig. 3A-B and Fig. 5A-B). IRAKM interacts with MyD88-IRAK4 to form IRAKM Myddosome to mediate TLR7-induced MEKK3-dependent second wave NFκB activation. However, this IRAKM-dependent pathway only induces expression of genes that are not regulated at the posttranscriptional levels (including inhibitory molecules SOCS-1, SHIP-1, A20 and IκBa), exerting an overall inhibitory effect on inflammatory response. The second novel mechanism is that IRAKM specifically interacts with IRAK2, but not IRAK1, and suppresses TLR7-induced IRAK2-mediated translation of cytokines and chemokines. These findings present a new outlook for how IRAKM modulates TLR signaling and TLR-mediated inflammatory responses. The final Figure 7 shows a crude fractionation of ribosomes from WT and IRAKM Kos and RT-PCRs of various RNAs. They find ca 2 fold differences in Fig 7B for TNF, IL6 and KC to argue that IRAKM is involved in translational repression. This would appear to be a single experiment (or replicates). It would require several independent repeats to be a credible result. Fig 7E is back to MEFs and IL1 - it shows that IRAKM forms a complex with IRAK2! (see my point above about the myddosome). The experiments in Fig. 7 have been repeated for 3 times. IRAKM Myddosome is formed in the absence of IRAK2 (Fig. 1B; Fig. 2B; Fig2H-J; Fig. 3A-B and Fig. 5A-B). In conclusion this study seems to be made up of a number of components that do not fit together very well and does not make a coherent argument to support the grand overview presented in Fig. 8. I found this paper very difficult to read - many sentences are difficult to understand as they are constructed of cumbersome and lengthy adjectival clauses. We respectively disagree with the reviewer about his/her general and negative comments, although we do appreciate his/her expertise in structural biology. I hope that our careful and detailed responses to his/her comments will help him/her to share our excitement about the novel findings on IRAKM. We would also appreciate if the third reviewer would read the comments from reviewer 1 and 2, who share a more positive view about our study. In term of writing style, we will be more than happy to work with the editors to improve according to the standards of EMBO J. Reference Fraczek J, Kim TW, Xiao H, Yao J, Wen Q, Li Y, Casanova JL, Pryjma J, Li X (2008) The kinase activity of IL-1 receptorassociated kinase 4 is required for interleukin-1 receptor/toll-like receptor-induced TAK1-dependent NFkappaB activation. J Biol Chem 283:

11 Figure 1. Whole films of EMSAs 7

12 A. Nuclear extracts prepared from wild-type (WT), IRAK1/2-double deficient (IRAK1/2 DKO), IRAK1/2/M-triple deficient (IRAK1/2/M TKO), IRAK4-deficient (IRAK4 KO) bone marrow-derived macrophages (BMDMs) untreated or treated with TLR7 ligand R848 (1μg/ml) for the indicated times were analyzed by electrophoresis mobility shift assay using an NFκBspecific probe. B. Nuclear extracts prepared from 293-I1A cells transiently co-transfected with wild-type (WT), W74A and E71A/Q78A untreated or treated with IL-1β (1ng/ml) for the indicated times were analyzed by electrophoretic mobility shift assay using an NFκB specific probe. C. Nuclear extracts prepared from wild-type (WT), and IRAKM deficient (MKO) BMDMs untreated or treated with R848 (1μg/ml) for the indicated times were analyzed by electrophoretic mobility shift assay using an NFκB specific probe. 8

13 The EMBO Journal Peer Review Process File - EMBO Accepted 18 December 2012 Thank you for submitting your revised manuscript to the EMBO Journal. Your study has now been seen by referees #1 and 2 and their comments are provide below. As you can see both referees appreciate the introduced changes and support publication here. I am therefore very pleased to proceed with the acceptance of your paper for publication here. Thank you for contributing to the EMBO Journal REFEREE REPORTS Referee #1 The authors have done a good job in addressing all my comments and revising the paper. In my opinion, it is now ready for publication. I also agree with the new title that has been proposed. Referee #2 This revision has addressed my concerns. I think the discovery of a positive role of IRAKM in NFkB activation is unexpected and important and it merits publication in EMBO J. European Molecular Biology Organization 5