TAK1 is an essential regulator of BMP signaling in ... - BioMedSearch

The EMBO Journal Review Process File - EMBO-2009-70648

Manuscript EMBO-2009-70648

TAK1 is an essential regulator of BMP signaling in cartilage Jae-Hyuck Shim, Matthew Greenblatt, Min Xie, Michael Schneider, Weiguo Zou, Bo Zhai, Steven Gygi Corresponding author: Laurie H. Glimcher, Harvard School of Public Health

Review timeline:

Submission date: Editorial Decision: Revision received: Accepted:

16 February 2009 11 March 2009 12 May 2009 18 May 2009

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

11 March 2009

Thank you very much for submitting your research manuscript for consideration to The EMBO Journal editorial office. As you will see from the attached comments, all three referees indicate clear interest in your study that establish not only TAK1 function in chondrocytes but also reveal a rather specific role of TAK1 downstream of BMP but not TGFbeta. However, two of these scientists demand further modifications that will have to involve further experimental work. Most importantly, ref#3 requests a better and more detailed phenotypic characterization of the skeletal phenotype, with the major issue being that additional time points should enlighten phenotypic changes at later stages of embryonic development. Ref#2 is mostly concerned with the TAK1/Smad interactions and does not only request better images but ideally also confirmation that the catalytic-inactive mutant does not simply compromise the binding surface. Conditioned on such modifications, we are glad to invite submission of an appropriately revised version of your research manuscript. I do also have to remind you that the decision on acceptance or rejection still depends on the content of the final version of your manuscript and that it is EMBO Journal policy to usually allow a single round of revisions only. Thank you for the opportunity to consider your work for publication. I look forward to your revision. Yours sincerely, Editor The EMBO Journal _____ REFEREE REPORTS:

© European Molecular Biology Organization

1


Referee #1 (Remarks to the Author): Shim et al in this paper analyze the function of TAK1 during chondrogenesis, they provide genetic and molecular evidence that TAK1 acts downstream of BMP signaling, furthermore they show that TAK1 mediates phosphorylation of Smad1; this allowed them to propose a novel model for BMP signaling in chondrocytes. I think this is an outstanding paper that should be published as such. Referee #2 (Remarks to the Author): The paper by Shim et al. describes the analysis of mice carrying a targeted cartilage-specific deletion of the Tak1 gene (Tak1col2) and addresses the molecular mechanisms uderlying the role of Tak1 in BMP signalling during chondrogenesis. The authors show that Tak1col2 mice display a dramatic postnatal phenotype, involving severe postnatal growth retardation, shortening of the long bones of the limbs, and joint abnormalities. This phenotype shows striking phenotypic similarities with mice carrying deletions of BMP receptor subunits in chondrocytes. The authors therefore conclude that Tak1 plays a role in BMP signalling pathways in chondrocytes. Indeed, they show that Tak1-deficient chondrocytes display a reduced activation of the downstream Smad and MAP kinase pathways following stimulation with BMP2/7, whereas the Smad pathway activation following TGF induction remains unaffected. The authors moreover show that Tak1 is capable of phoshorylating both Smad1 as well as Smad 2 , and that Tak1 can directly interact with Smad proteins in an inducible manner. Overall the results presented by Shim and colleagues are of good quality and support the conclusions drawn by the authors. A few issues, however, would need to be addressed before publication. Authors show that ID1 expression in hypertrophic chondrocytes of Tak1col2 mice is considerably reduced (fig. 3C) however, in fig. 4C they show by RT-PCR that ID1 relative mRNA levels in non stimulated Tak1fl/fl and in Tak1-deficient chondrocytes are identical. What is the reason for this discrepancy? In Fig. 5 , the authors show that Smad proteins directly interact with Tak1. In Fig. 5A , lower panel, authors should show the amount of glutathione-resin-bound GST and GST-Smad1 proteins as controls. The inference that Tak1 interacts with BMP-responsive Smads in an inducible manner because the kinetic Tak1-Smad1 association (fig. 5b) is similar to the kinetic of Smad1/5/8 phopshorylation is a little pushed, it would be better to show this experimentally, the presented IPs moreover are of poor quality (especially the panel showing P-Smad1/5/8 IP) , I realise the difficulty of performing these experiments, nevertheless, due to the importance of the conclusions drawn, it would be advisable to show a better image of the IPs and to include control IPs showing the lack of crossreaction of the anti-Tak antibodies. Using a catalytically-inactive form of Tak1 the authors conclude that the kinase activity of Tak1 is required for its interaction with Smad proteins. Formally , however, it cannot be ruled out that the mutation that renders the Tak1 protein catalytically inactive is somehow affecting the Smadinteraction surface of Tak1. A mapping of the Smad-interaction region within Tak1 would clarify this point. Minor points: on page 7 (lines 18, 19 and 21 , and probably elesewhere in the text) the single proximal long bone of the fore limb should be spelt humerus, not "humerous" besides, a few typos are present in the text e.g. on page 4 line 2 , "intraembryoinc" Referee #3 (Remarks to the Author): Using a conditional knockout (CKO) approach, Shim and collaborators report important new findings concerning the role of TGFbeta activated kinase 1 (TAK1) in cartilage. Mice deficient for


2


TAK1 in chondrocytes present chondrodysplasia and joint abnormalities that are reminiscent of skeletal defects observed in mice in which bone morphogenetic (BMP) signaling is impaired. TAK1 was initially characterized as a key regulator of the MAPK pathway activated downstream of TGFbeta and BMP. Interestingly, the authors show that TAK1 responds essentially to BMP-induced signals (by activating both MAPKs and Smads) but not TGFbeta-induced signals (although TAK1 affects TGFbeta-mediated activation of p38) in chondrocytes in vitro. Consistent with this finding, they show that the expression of genes normally induced by BMP is decreased in absence of TAK1 in cartilage in vivo. Lastly, the authors revealed that TAK1 phosporylates Smad1 in vitro upon activation of BMP signaling. Although it is well known that BMP signaling plays a critical role in skeletal development, the large number of ligands and receptors, as well as the vital role of many component of this network during early development has precluded a straightforward use of KO mice to characterize more precisely the functions of BMP signaling in this process. The work of Shim and collaborator present a new exiting mouse model that brings new insights into the signaling cascade that regulates skeletal development downstream of BMP. Overall, this work is well presented and the data of good quality. However, the manuscript could be significantly improved by addressing the following points: Main criticism: An important criticism of this manuscript concerns the very incomplete characterization of the skeletal phenotype of Col2-cre-mediated TAK1 CKO mice. The authors mention that E18.5 CKO embryos are indistinguishable from control animals, although newborn CKO mice shown in supplementary figure 2 are much smaller than control mice. It is unlikely that this difference appears so suddenly. It is more likely that this results from defects in endochondral bone development occurring during fetal development. It would be informative to show in one single figure (not as supplementary figure) the phenotype of the growth plate chondrocytes of control and CKO mice at various stages of embryonic development (E15.5 and/or E16.5) in addition to the postnatal stages (newborn and P20) already shown. It would be important also to perform in situ hybridizations at these various stages using Ihh, colX and osteopontin probes to evaluate carefully how TAK1 affects hypertrophic maturation of chondrocytes. In addition, the authors should evaluate whether chondrocyte proliferation and apoptosis is affected in absence of TAK1. Lastly, the joint phenotype is not presented well enough. Are all the joints affected? The fusion of the tarsals in CKO mice is difficult to visualize in figure 2B. It would be nice to present alizarin red/alcian blue staining of the elbow of CKO mice instead of (or in addition to) the radiographs, in order to present a "threedimensional view" of the phenotype. The joints are formed during fetal development. Thus, it is also important to look at joint formation in CKO animals during embryonic development in order to identify when the defects occur. Is there a defect of joint specification and/or joint cavitation? Is the expression of GDF5 and other key players in joint development affected in absence of TAK1? Minor points: 1- the rationale of the work done by Shim and collaborators is sometimes not clearly presented. For instance, the authors do not present why they were initially interested in studying the role of TAK1 in cartilage. They do not present any scientific rationale for using chondrocytes immortalized with SV40 large T antigen, to demonstrate the role of TAK1 downstream of BMP. In fact, in vitro experiments performed with these cells have little biological relevance since TAK1 is expressed in quiescent hypertrophic chondrocytes, not in proliferating chondrocytes. Similarly, it is not clear why the authors used human kidney embryonic cells (HEK293) to demonstrate that TAK1 interacts with and phosphorylates Smad1. Is this also happening in chondrocytes? Moreover, they do not explain clearly enough why evaluating whether Smad proteins are phosphorylated and where they are phosphorylated by TAK1 is important. 2- If doable, it would be nice to show that the expression of genes specifically induced by Smad2 are not downregulated in vivo in growth plate chondrocytes. 3- the authors wrote in page 9 that Ihh expression is downregulated in hypertrophic chondrocytes in absence of TAK1, although it is not clear whether figure 3C shows a true downregulation of Ihh expression or simply a reduction of the number of pre-hypertrophic chondrocytes, which express Ihh. The authors could replace the term "hypertrophic" by "pre-hypertrophic" since Ihh expression is restricted to pre-hypertrophic chondrocytes, but is not expressed in hypertrophic chondrocytes. 4- the organization of the many figures presenting the in vitro data (between that in the main manuscript and the supplementary figures) is somewhat confusing. There is perhaps a way to simplify this presentation. The supplementary figure S4F mentioned in page 11 does not exist. 5- the authors could specify what the Tlx2-Luc and 3TP-Luc reporter constructs they used in figures


3


4B and 4C are. Is the 3TP-Luc construct activated only by Smad2 or also by MAPKs? 6- the authors could discuss the potential role of TAK1 in TGFbeta-induced activation of p38, since p38 phosphorylation is decreased in absence of TAK1 in chondrocytes in vitro. 7- in the experiment described in figure 7, what would be the status of phospho-ERK1/2 and phospho-JNK1/2?

1st Revision - authors' response

12 May 2009

We are submitting our revised manuscript "TAK1 is an essential regulator of BMP signaling in cartilage" by Jae-Hyuck Shim, Matthew Greenblatt, Min Xie, Weiguo Zou, Bo Zhai, Steven Gygi, Michael Schneider and myself for your reconsideration for publication in The EMBO Journal. We have performed additional experiments to address the suggestions and concerns raised by the reviewers. The major issue of concern by Reviewer 3, was the need to provide a better and more detailed phenotypic characterization of the skeletal phenotype. To address this, we have analyzed the embryonic phenotype of Tak1col2 mice at E14.5, E16.5 and E18.5. We have performed in situ hybridization analysis of the embryos at E16.5 and E18.5 with Collagen X to determine the status of the growth plate phenotype at these time points (Revised Figure 2C). This confirmed our observation that overall growth and development of long bones is normal during the embryonic period and that the onset of the phenotype is sometime between E18.5 and the neonatal period, and does not reflect an overall defect in endochondral ossification. Analysis of the joints of E14.5 embryos further defined the elbow phenotype, demonstrating a delay in the cavitation of the elbow joint (Revised Figure 2F, lower panels). We thank the reviewer for suggesting these experiments as it they uncovered, as the reviewer suspected, a prenatal phenotype. Additionally, we have analyzed the proliferation and apoptosis of chondrocytes in the growth plate by performing PCNA and TUNEL staining and demonstrated that the reduction in the size of the growth plate of Tak1col2 mice is due to both a reduction in chondrocyte proliferation and an increase in the apoptosis of terminal hypertrophic chondrocytes (Revised Figure 2D). The major issues of concern by Reviewer 2, were twofold: 1) a better image of the TAK1/Smad interactions and 2) a confirmation that the catalytic-inactive mutant does not simply compromise the binding surface and 3) identification of the Tak1 domains that interact with Smad1. We repeated the experiment as suggested by the reviewer and replaced it with a figure of better quality (Revised Figure 5B). Next, we confirmed that the K63W catalytically inactive mutation does not affect the Smad-interaction surface of Tak1, by demonstrating that this mutation does not affect the interaction between TAK1 and TAB1 (Revised Supplementary Figure 4D), and by demonstrating specificity of the effect on TAK1 interactions. These data demonstrate that the K63W mutation does not cause gross misfolding of TAK1. Additionally, as suggested by the reviewer, we have used a series of TAK1 deletion mutants to map the Smad-interaction region within TAK1 (Revised Supplementary Figure 4B). TAK1 binds to Smad1 through amino acids 200-300, thus the K63W mutation is outside the interaction domain between Smad1 interaction domain of TAK1. Additional specific comments have been addressed below as follows: Referee #1 (Remarks to the Author): Shim et al in this paper analyze the function of TAK1 during chondrogenesis, they provide genetic and molecular evidence that TAK1 acts downstream of BMP signaling, furthermore they show that TAK1 mediates phosphorylation of Smad1; this allowed them to propose a novel model for BMP signaling in chondrocytes. I think this is an outstanding paper that should be published as such. We thank this reviewer for his/her enthusiasm for our study.


4


Referee #2 (Remarks to the Author): The paper by Shim et al. describes the analysis of mice carrying a targeted cartilage-specific deletion of the Tak1 gene (Tak1col2) and addresses the molecular mechanisms uderlying the role of Tak1 in BMP signalling during chondrogenesis. The authors show that Tak1col2 mice display a dramatic postnatal phenotype, involving severe postnatal growth retardation, shortening of the long bones of the limbs, and joint abnormalities. This phenotype shows striking phenotypic similarities with mice carrying deletions of BMP receptor subunits in chondrocytes. The authors therefore conclude that Tak1 plays a role in BMP signalling pathways in chondrocytes. Indeed, they show that Tak1-deficient chondrocytes display a reduced activation of the downstream Smad and MAP kinase pathways following stimulation with BMP2/7, whereas the Smad pathway activation following TGFβ induction remains unaffected. The authors moreover show that Tak1 is capable of phoshorylating both Smad1 as well as Smad 2, and that Tak1 can directly interact with Smad proteins in an inducible manner. Overall the results presented by Shim and colleagues are of good quality and support the conclusions drawn by the authors. A few issues, however, would need to be addressed before publication. Authors show that ID1 expression in hypertrophic chondrocytes of Tak1col2 mice is considerably reduced (fig. 3C) however, in fig. 4C they show by RT-PCR that ID1 relative mRNA levels in non stimulated Tak1fl/fl and in Tak1-deficient chondrocytes are identical. What is the reason for this discrepancy? Substantial evidence indicates that chondrocytes in the terminal regions of the growth plate are stimulated by BMP ligands in vivo (Chung et al, 2001; Kronenberg, 2003; Retting et al, 2009; Yoon & Lyons, 2004; Yoon et al, 2006). As shown in Revised Figure 3A and C, the regions of phosphoSmad1/5/8 staining and ID1 expression overlap in the chondroepiphysis in vivo. Thus, the relevant comparison to the in vivo expression data is the BMP-stimulated in vitro cultures, not the unstimulated values. The unstimulated culture values are included as a control to demonstrate ID1 induction upon stimulation with BMPs (Revised Figure 4A). In Fig 5, the authors show that Smad proteins directly interact with Tak1. In Fig. 5A, lower panel, authors should show the amount of glutathione-resin-bound GST and GST-Smad1 proteins as controls. This was added as per the reviewer’s suggestion (Revised Figure 5A, lower panels). The inference that Tak1 interacts with BMP-responsive Smads in an inducible manner because the kinetic Tak1-Smad1 association (fig. 5b) is similar to the kinetic of Smad1/5/8 phosphorylation is a little pushed, it would be better to show this experimentally, the presented IPs moreover are of poor quality (especially the panel showing P-Smad1/5/8 IP), I realize the difficulty of performing these experiments, nevertheless, due to the importance of the conclusions drawn, it would be advisable to show a better image of the IPs and to include control IPs showing the lack of crossreaction of the anti-Tak1 antibodies. We repeated the experiment as suggested and replaced it with a figure of better quality (Revised Figure 5B). Using a catalytically-inactive form of Tak1 the authors conclude that the kinase activity of Tak1 is required for its interaction with Smad proteins. Formally, however, it cannot be ruled out that the mutation that renders the Tak1 protein catalytically inactive is somehow affecting the Smadinteraction surface of Tak1. A mapping of the Smad-interaction region within Tak1 would clarify this point. We have addressed this concern by demonstrating that this mutation does not affect the interaction between TAK1 and TAB1 (Revised Supplementary Figure 4D), demonstrating that the effect of the K63W mutation on TAK1 interactions is specific for its interactions with Smads. Hence the K63W mutation does not cause gross misfolding of TAK1. Additionally, we have used a series of TAK1 deletion mutants to map the Smad-interaction region within TAK1 as suggested (Revised Supplementary Figure 4B). TAK1 binds to Smad1 through amino acids 200-300, thus the K63W


5


mutation is outside the interaction domain between Smad1 interaction domain of TAK1. The domain containing the K63W mutation has been crystallized, allowing for a structural explanation of the effects of the K63W mutation on protein folding (Brown et al, 2005). The K63 side chain faces towards the solvent-exposed side of the ATP binding pocket of TAK1. Thus, while it prevents TAK1 catalytic activity by destabilizing ATP binding, it is not involved in the formation of internal salt bridges or other interactions that would significantly stabilize the overall folding of TAK1. Thus, it is unlikely that this mutation in a solvent-facing side chain would destabilize a separate, distal domain of the protein (amino acids 200-300). In addition, this mutant has been characterized in other systems (Thiefes et al, 2005) and has been found to function as a specific dominant-negative of TAK1. Considering the convergence of our own interaction data between TAK1 and TAB1, X-ray crystallographic structural data of TAK1, and functional data in our own and other cellular systems, the K63W mutation is very well characterized, we believe supports our conclusions. Minor points: on page 7 (lines 18, 19 and 21 , and probably elesewhere in the text) the single proximal long bone of the fore limb should be spelt humerus, not "humerous" Corrected. besides, a few typos are present in the text e.g. on page 4 line 2 ,"intraembryoinc" Corrected. Referee #3 (Remarks to the Author): Using a conditional knockout (CKO) approach, Shim and collaborators report important new findings concerning the role of TGFbeta activated kinase 1 (TAK1) in cartilage. Mice deficient for TAK1 in chondrocytes present chondrodysplasia and joint abnormalities that are reminiscent of skeletal defects observed in mice in which bone morphogenetic (BMP) signaling is impaired. TAK1 was initially characterized as a keyregulator of the MAPK pathway activated downstream of TGFbeta and BMP. Interestingly, the authors show that TAK1 responds essentially to BMP-induced signals (by activating both MAPKs and Smads) but not TGFbeta-induced signals (although TAK1 affects TGFbeta-mediated activation of p38) in chondrocytes in vitro. Consistent with this finding, they show that the expression of genes normally induced by BMP is decreased in absence of TAK1 in cartilage in vivo. Lastly, the authors revealed that TAK1 phosporylates Smad1 in vitro upon activation of BMP signaling. Although it is well known that BMP signaling plays a critical role in skeletal development, the large number of ligands and receptors, as well as the vital role of many component of this network during early development has precluded a straightforward use of KO mice to characterize more precisely the functions of BMP signaling in this process. The work of Shim and collaborator present a new exiting mouse model that brings new insights into the signaling cascade that regulates skeletal development downstream of BMP. Overall, this work is well presented and the data of good quality. However, the manuscript could be significantly improved by addressing the following points: Main criticism: An important criticism of this manuscript concerns the very incomplete characterization of the skeletal phenotype of Col2-cre-mediated TAK1 CKO mice. The authors mention that E18.5 CKO embryos are indistinguishable from control animals, although newborn CKO mice shown in supplementary figure 2 are much smaller than control mice. It is unlikely that this difference appears so suddenly. It is more likely that this results from defects in endochondral bone development occurring during fetal development. We have directly addressed the onset of the phenotype by analyzing Tak1col2 mice during embryonic time points E14.5, E16.5 and E18.5. We discuss these findings in greater detail below, but while the joint phenotype at the elbow is present at E14.5, the alteration in chondrocyte


6


maturation in the growth plate doesnít manifest until sometime between E18.5 and the neonatal period. This suggests that BMP ligands involved in joint cavitation utilize MAPK signaling differently than BMP ligands involved in regulating the growth plate. Alternatively, it may simply reflect the additional presence of other MAPKKK family members that can compensate for the loss of TAK1 during the embryonic period in the growth plate but not during joint cavitation. It will be of interest to further explore the biochemical basis for this difference in future studies. It would be informative to show in one single figure (not as supplementary figure) the phenotype of the growth plate chondrocytes of control and CKO mice at various stages of embryonic development (E15.5 and/or E16.5) in addition to the postnatal stages (newborn and P20) already shown. It would be important also to perform in situ hybridizations at these various stages using Ihh, colX and osteopontin probes to evaluate carefully how TAK1 affects hypertrophic maturation of chondrocytes. As per the reviewer’s suggestion, we have added a figure showing ColX in situ staining of the chondroepiphysis at E16.5 and E18.5 (Revised Figure 2C). We did not find a noticeable difference in chondrocyte maturation or the overall structure of the chondroepiphysis at these time points. Additionally, we performed in situ hybridization for osteopontin on the femurs of E16.5 Tak1col2 mice (Revised Supplementary Figure 2C). The Tak1col2 and littermate control show a similar extent of ossification, confirming that the early embryonic stages of endochondral ossification are not altered in Tak1col2 mice. These results confirm that the growth plate phenotype observed first manifests sometime between E18.5 and p0 and only becomes dramatically evident by around 1 week of life. This suggests that TAK1 might be redundant with another MAPKKK family member in the growth plate during the embryonic period. In addition, the authors should evaluate whether chondrocyte proliferation and apoptosis is affected in absence of TAK1. As the reviewer suggested, we have extended our results demonstrating a growth defect in Tak1col2 mice to include an analysis of chondrocyte proliferation and apoptosis. This analysis was performed using Tunel and PCNA staining of the chondroepiphysis (Revised Figure 3D). Tak1col2 chondrocytes show both a reduction in PCNA-positive prehypertrophic chondrocytes and an increase in Tunel-positive terminal hypertrophic chondrocytes. This suggests that the runting observed in Tak1col2 mice reflects both decreased chondrocyte proliferation and increased chondrocyte apoptosis. Notably, inducible postnatal deletion of Indian hedgehog (IHH) results in a reduction in PCNA positive proliferating chondrocytes, though it was not found to affect chondrocyte apoptosis (Maeda et al, 2007). Thus, at least some of the effects of TAK1 to promote chondrocyte survival are not likely to be secondary to decreased IHH secretion. Lastly, the joint phenotype is not presented well enough. Are all the joints affected? This point has been clarified in the manuscript as per the reviewerís suggestion. As in the brachypodism mice with a mutation in the BMP ligand GDF5, the postnatal Tak1col2 mice only display fusion of the carpals and tarsals (Revised Figure 2G). Also as in the brachypodism mice, architecture of the elbow is altered (Revised Figure 2E,F). Other joints are not affected. Additionally, we have extended our findings regarding the joint phenotype by analyzing E14.5 embryos and found a delay in the formation of the joint between the radius and the humerus (Revised Figure 2F, area indicated by arrows). This is further discussed below. The fusion of the tarsals in CKO mice is difficult to visualize in figure 2B. It would be nice to present alizarin red/alcian blue staining of the elbow of CKO mice instead of (or in addition to) the radiographs, in order to present a "three-dimensional view" of the phenotype. As suggested by the reviewer, pictures of the alizarin red/alcain blue-stained elbows were added to the radiographs previously presented (Revised Figure 2E, lower panels). They show ectopic ossification present in the joint space of CKO mice, a feature we have previously noticed in other models of spontaneous elbow dislocation in mice (unpublished data). The images of the alizarin red/alcain blue stained tarsals demonstrating tarsal fusion in CKO mice was enlarged to make the fusion more apparent (Revised Figure 2G, lower panels).


7


The joints are formed during fetal development. Thus, it is also important to look at joint formation in CKO animals during embryonic development in order to identify when the defects occur. Is there a defect of joint specification and/or joint cavitation? Is the expression of GDF5 and other key players in joint development affected in absence of TAK1? We appreciate these helpful suggestions. As mentioned above, we have extended our findings regarding the joint phenotype in Tak1col2 mice to include an analysis of E14.5 embryos. Notably, there is a failure to fully form the joint between the radius and the humerus, and at E14.5 the radius and the humerus are present as a continuous skeletal element (Revised Figure 2F, lower panels). In our analysis of postnatal Tak1col2 mice, this joint is present, though with distorted and abnormal architecture (Revised Figure 2E,F). Taking these two results together, this suggests a delay in the cavitation of the elbow joint, as would be expected in the case of defective signaling by BMP family ligands, especially GDF5. In future studies, we hope to further explore the role of MAPK and Smad signaling in GDF5-mediated joint specification, and formally demonstrate that TAK1 is downstream of GDF5 signaling in vivo. In order to directly differentiate between defective joint specification versus defective joint cavitation downstream of GDF5, we performed in situ hybridization analysis for GDF5 on the E14.5 embryo joints. Despite repeated efforts, we were unable to get robust, reproducible staining of GDF5 on the E14.5 embryo joints due to technical problems. Minor points: The rationale of the work done by Shim and collaborators is sometimes not clearly presented. For instance, the authors do not present why they were initially interested in studying the role of TAK1 in cartilage. The introduction has been changed to emphasize the rationale behind our study. Briefly, there are little in vivo data describing the functional role of MAPK signaling cascades in chondrocytes in vivo. TAK1 was originally described as a MAPKKK activated downstream of TGF /BMP ligands. Given that both TGF and BMP ligands are known to function in the maintenance and development of bone, we considered conditional ablation of TAK1 as a reasonable starting point to determine both which MAPKKK genes function in the skeleton and what the overall contribution of MAPK activation was to cartilage in vivo (Yamaguchi et al, 1995). Previously, efforts to address these points were frustrated by both extensive redundancy at multiple levels within MAPK signaling cascades and by embryonic lethality of these pathway mutants. In characterizing the Tak1col2 phenotype, we found evidence that TAK1 mediates an unexpected and surprising crosstalk between MAPK signaling cascades and Smad1/5/8 activation, thus we further pursued this mechanistic aspect of the phenotype. They do not present any scientific rationale for using chondrocytes immortalized with SV40 large T antigen, to demonstrate the role of TAK1 downstream of BMP. In fact, in vitro experiments performed with these cells have little biological relevance since TAK1 is expressed in quiescent hypertrophic chondrocytes, not in proliferating chondrocytes. We appreciate the reviewerís concern that immortalized chondrocytes are in some respects a nonphysiologic system, and we have taken steps to both validate the use of this model and compliment the weaknesses of this approach. Foremost, all of the key experiments performed in the immortalized chondrocyte lines were repeated or analogous experiments performed in primary chondrocytes. Many observations, such as reduced Smad1/5/8 phosphorylation were further confirmed in vivo. However, some approaches to biochemical analysis of TAK1 functions in BMP signaling are not tractable using solely primary chondrocytes (i.e., luciferase reporter assays). Thus, use of the immortalized chondrocyte lines compliments and strengthens our findings in vivo and in primary chondrocyte cultures. SV40 large T antigen-induced immortalization is widely used as a platform for biochemical analysis in mouse embryonic fibroblasts (MEFs), and this approach has been used in chondrocytes as well (Kobayashi et al, 2005; Mallein-Gerin & Olsen, 1993; Shim et al, 2005). We have carefully characterized our immortalized chondrocytes lines to address the major concern


8


of the reviewer, that TAK1 is not expressed in our cultures due to a lack of chondrocytes in a prehypertrophic or hypertrophic state of differentiation. First, we have demonstrated robust expression of TAK1 in our immortalized chondrocyte lines and robust deletion of TAK1 in Tak1col2 -derived cultures (Revised Supplementary Figure 3A). Furthermore, we have characterized the differentiation status of the immortalized cultures and demonstrated that, like primary cultures, they contain cells in a mixture of differentiation states, including hypertrophic chondrocytes, as demonstrated by quantitative real time PCR for ColX (Revised Supplementary Figure 3B). As mentioned above, we have also functionally validated this system by confirming that multiple key observations are repeatable in chondrocyte lines, primary chondrocytes, and in cartilage in vivo. For example, our findings of reduced Id1 induction and reduced Smad1/5/8 phosphorylation, were confirmed in chondrocyte cell lines, primary chondrocytes and in vivo (unpublished data and Revised Figure 3 and 4). Thus, we feel that using the immortalized chondrocyte lines in parallel with primary chondrocytes, strengthens our conclusions by the use of multiple confirmatory approaches Similarly, it is not clear why the authors used human kidney embryonic cells (HEK293) to demonstrate that TAK1 interacts with and phosphorylates Smad1. Is this also happening in chondrocytes? HEK293 cells are commonly used for overexpression studies relevant to a variety of biochemical analyses due to their high transfection efficiency. Please note that the data demonstrating the interaction between the endogenous TAK1 and Smad1 in wild type chondrocytes were present in the original manuscript (Revised Figure 5B). Moreover, they do not explain clearly enough why evaluating whether Smad proteins are phosphorylated and where they are phosphorylated by TAK1 is important. It has been well established that Smad proteins are signal transducers and transcriptional comodulators of BMP signaling pathways and phosphorylation of R-Smad1/5/8 (receptor-regulated Smad) at the conserved carboxyl terminal SVS sequence (S435/S438) activates BMP signaling. The phosphorylation level of R-Smads determines the degree of their transcriptional activity (Chacko et al, 2001; Kretzschmar et al, 1997; Macias-Silva et al, 1998; Qin et al, 2001). We demonstrate that TAK1 mediates Smad1 phosphorylation at C-terminal serine residues (S435/S438) (Revised Figure 6), although it would be necessary to determine additional Smad1 phosphorylation sites by TAK1 in future study. More generally, determining the location of TAK1mediated phosphorylation is a necessary precursor to further studies to determine the functional importance of this phosphorylation event. If doable, it would be nice to show that the expression of genes specifically induced by Smad2 are not downregulated in vivo in growth plate chondrocytes. In vivo evidence suggests that TGF family members play very modest roles in regulating growth plate chondrocytes. Three main approaches have been published to determine the role of TGF signaling in chondrocyte biology in vivo: TGF RII floxed-allele mice bred to Col2-cre, TGF RII floxed allele mice bred to Prx1-cre, and TGF RII dominant-negative transgenics driven by the Col2 promoter. TGF RIIPrx1 mice display a growth defect, but TGF RII col2 and col2-TGF RIIdn transgenic mice display only an osteoarthritis phenotype without major growth defects. Thus, these in vivo studies indicate that while TGF signaling functions to prevent the development of osteoarthritis and in a non-chondrocyte mesenchymal cell to regulate long bone elongation, it has only very slight cell-intrinsic effect on chondrocytes of the growth plate (Baffi et al, 2004; Seo & Serra, 2007; Serra et al, 1997). Notably, the col2-TGF RII dominant negative mice were characterized for their expression of IHH and were found to display modestly increased IHH expression (Serra et al. 1997). This is in direct contrast to our results, which indicate a decrease in IHH expression. This offers further support for our hypothesis that TAK1-deficiency selectively impairs BMP and not TGF signaling in chondrocytes in vivo. Additionally, TGF RII-deficient mice display characteristic defects in the morphology of vertebrae, sternum, and the base of the skull (Baffi et al., 2004). In contrast, all such


9


defects are not present in the Tak1col2 mice, offering functional evidence against a role for TAK1 in TGF signaling. Unfortunately, due to the lack of a clear role for TGF signaling in chondrocytes of the growth plate in vivo, little work has been done to characterize TGF target genes in growth plate chondrocytes and it is unclear what genes specifically require Smad2 for their expression in the growth plate in vivo. The authors wrote in page 9 that Ihh expression is downregulated in hypertrophic chondrocytes in absence of TAK1, although it is not clear whether figure 3C shows a true downregulation of Ihh expression or simply a reduction of the number of pre hypertrophic chondrocytes, which express Ihh. The authors could replace the term "hypertrophic" by "pre-hypertrophic" since Ihh expression is restricted to pre-hypertrophic chondrocytes, but is not expressed in hypertrophic chondrocytes. We appreciate the reviewerís suggestion to distinguish between an overall reduction in levels of secreted IHH versus simply noting a reduction in the number of prehypertrophic chondrocytes. We have extended the results demonstrating a reduction in IHH levels by confirming a reduction in the expression of a canonical IHH target gene, patched, in vivo (Revised Figure 3C). Patched transcript levels reflect a functional measure of secreted IHH levels in vivo ihh (Maeda et al, 2007). We performed in situ hybridization analysis for patched on Tak1fl/fl and Tak1col2 chondroepiphyses and found a reduction in patched expression not only in the chondroepiphysis, but also in the bone collar. Since col2-cre mediated deletion of TAK1 is restricted to chondrocytes, the reduction in patched expression in the bone collar further rules out the possibility of a direct, cell intrinsic, effect of TAK1 to regulate patched expression by a means other than altered IHH secretion. The in situ data for IHH indicates that the reduction in IHH secretion is probably due to both a reduction in the number of IHH-expressing pre-hypertrophic chondrocytes and a reduction in the per-cell expression of IHH. We agree with the reviewerís concern that the predominant expression of IHH is in prehypertrophic chondrocytes and have changed the manuscript accordingly. However, examination of IHH expression data from ourselves and others (i.e. Chung et al, 2001) indicates that IHH is also expressed in hypertrophic chondrocytes. The organization of the many figures presenting the in vitro data (between that in the main manuscript and the supplementary figures) is somewhat confusing. There is perhaps a way to simplify this presentation. The supplementary figure S4F mentioned in page 11 does not exist. Corrected. This was reorganized as per the reviewer’s suggestion. The authors could specify what the Tlx2-Luc and 3TP-Luc reporter constructs they used in figures 4B and 4C are. The 3TP-Luc reporter gene that contains Smad binding elements (SBEs), AP-1 sites, and the plasminogen activator inhibitor-1 (PAI-1) promoter, is widely used and highly TGF -responsive (Wrana et al, 1992). As a homeobox gene, Tlx2 expression is directly regulated by the heterocomplex of Smad1 and Smad4 downstream of BMP signaling. The Tlx2-Luc reporter gene contains Tlx2 promoter element that responds to BMP ligands through its BMP responsive elements (BREs) (Tang et al, 1998; Xiao et al, 2003). Is the 3TP-Luc construct activated only by Smad2 or also by MAPKs? The 3TP-Luc reporter gene can be activated by both Smad2 and MAPKs because it contains Smad binding elements (SBEs), AP-1 sites, and the plasminogen activator inhibitor-1 (PAI-1) promoter (Frey & Mulder, 1997; Wrana et al, 1992). However, it has been suggested that TGF -induced MAPK activation is more closely associated with the ability of TGF to inhibit DNA synthesis than with the ability of TGF to regulate AP-1 dependent gene expression important for both TGF 1 production and control of the extracellular matrix (Frey & Mulder, 1997). The authors could discuss the potential role of TAK1 in TGFbeta-induced activation of p38, since


10


p38 phosphorylation is decreased in absence of TAK1 in chondrocytes in vitro. This was added to discussion as per the reviewer’s suggestion. In the experiment described in figure 7, what would be the status of phospho-ERK1/2 and phosphoJNK1/2? Phosphorylation levels of ERK1/2 and JNK1/2 MAP kinases by BMP ligands were too weak to be detected in lentiviral infected Tak1col2 chondrocytes by immunoblotting analysis. Even in the wt chondrocytes, BMP-induced ERK1/2 and JNK1/2 MAPK phosphorylation was barely detected by immunoblotting analysis (Revised Figure 4E). References Cited Baffi MO, Slattery E, Sohn P, Moses HL, Chytil A, Serra R (2004) Conditional deletion of the TGFbeta type II receptor in Col2a expressing cells results in defects in the axial skeleton without alterations in chondrocyte differentiation or embryonic development of long bones. Dev Biol 276(1): 124-142 Brown K, Vial SC, Dedi N, Long JM, Dunster NJ, Cheetham GM (2005) Structural basis for the interaction of TAK1 kinase with its activating protein TAB1. J Mol Biol 354(5): 1013-1020 Chacko BM, Qin B, Correia JJ, Lam SS, de Caestecker MP, Lin K (2001) The L3 loop and Cterminal phosphorylation jointly define Smad protein trimerization. Nat Struct Biol 8(3): 248-253 Chung UI, Schipani E, McMahon AP, Kronenberg HM (2001) Indian hedgehog couples chondrogenesis to osteogenesis in endochondral bone development. J Clin Invest 107(3): 295-304 Frey RS, Mulder KM (1997) TGFbeta regulation of mitogen-activated protein kinases in human breast cancer cells. Cancer Lett 117(1): 41-50 Kobayashi H, Tanaka N, Asao H, Miura S, Kyuuma M, Semura K, Ishii N, Sugamura K (2005) Hrs, a mammalian master molecule in vesicular transport and protein sorting, suppresses the degradation of ESCRT proteins signal transducing adaptor molecule 1 and 2. Journal of Biological Chemistry 280(11): 10468-10477 Kretzschmar M, Liu F, Hata A, Doody J, Massague J (1997) The TGF-beta family mediator Smad1 is phosphorylated directly and activated functionally by the BMP receptor kinase. Genes Dev 11(8): 984-995 Kronenberg HM (2003) Developmental regulation of the growth plate. Nature 423(6937): 332-336 Macias-Silva M, Hoodless PA, Tang SJ, Buchwald M, Wrana JL (1998) Specific activation of Smad1 signaling pathways by the BMP7 type I receptor, ALK2. J Biol Chem 273(40): 25628-25636 Maeda Y, Nakamura E, Nguyen MT, Suva LJ, Swain FL, Razzaque MS, Mackem S, Lanske B (2007) Indian Hedgehog produced by postnatal chondrocytes is essential for maintaining a growth plate and trabecular bone. Proc Natl Acad Sci U S A 104(15): 6382-6387 Mallein-Gerin F, Olsen BR (1993) Expression of simian virus 40 large T (tumor) oncogene in mouse chondrocytes induces cell proliferation without loss of the differentiated phenotype. Proc Natl Acad Sci U S A 90(8): 3289-3293 Qin BY, Chacko BM, Lam SS, de Caestecker MP, Correia JJ, Lin K (2001) Structural basis of Smad1 activation by receptor kinase phosphorylation. Mol Cell 8(6): 1303-1312 Retting KN, Song B, Yoon BS, Lyons KM (2009) BMP canonical Smad signaling through Smad1 and Smad5 is required for endochondral bone formation. Development 136(7): 1093-1104 Seo HS, Serra R (2007) Deletion of Tgfbr2 in Prx1-cre expressing mesenchyme results in defects in


11


development of the long bones and joints. Dev Biol 310(2): 304-316 Serra R, Johnson M, Filvaroff EH, LaBorde J, Sheehan DM, Derynck R, Moses HL (1997) Expression of a truncated, kinase-defective TGF-beta type II receptor in mouse skeletal tissue promotes terminal chondrocyte differentiation and osteoarthritis. J Cell Biol 139(2): 541-552 Shim JH, Xiao C, Paschal AE, Bailey ST, Rao P, Hayden MS, Lee KY, Bussey C, Steckel M, Tanaka N, Yamada G, Akira S, Matsumoto K, Ghosh S (2005) TAK1, but not TAB1 or TAB2, plays an essential role in multiple signaling pathways in vivo. Genes & Development 19(22): 2668-2681 Tang SJ, Hoodless PA, Lu Z, Breitman ML, McInnes RR, Wrana JL, Buchwald M (1998) The Tlx-2 homeobox gene is a downstream target of BMP signalling and is required for mouse mesoderm development. Development 125(10): 1877-1887 Thiefes A, Wolter S, Mushinski JF, Hoffmann E, Dittrich-Breiholz O, Graue N, Dorrie A, Schneider H, Wirth D, Luckow B, Resch K, Kracht M (2005) Simultaneous blockade of NFkappaB, JNK, and p38 MAPK by a kinase-inactive mutant of the protein kinase TAK1 sensitizes cells to apoptosis and affects a distinct spectrum of tumor necrosis target genes. Journal of Biological Chemistry 280(30): 27728-27741 Wrana JL, Attisano L, Carcamo J, Zentella A, Doody J, Laiho M, Wang XF, Massague J (1992) TGF beta signals through a heteromeric protein kinase receptor complex. Cell 71(6): 1003-1014 Xiao C, Shim JH, Kluppel M, Zhang SS, Dong C, Flavell RA, Fu XY, Wrana JL, Hogan BL, Ghosh S (2003) Ecsit is required for Bmp signaling and mesoderm formation during mouse embryogenesis.[see comment]. Genes & Development 17(23): 2933-2949 Yamaguchi K, Shirakabe K, Shibuya H, Irie K, Oishi I, Ueno N, Taniguchi T, Nishida E, Matsumoto K (1995) Identification of a member of the MAPKKK family as a potential mediator of TGF-beta signal transduction. Science 270(5244): 2008-2011 Yoon BS, Lyons KM (2004) Multiple functions of BMPs in chondrogenesis. J Cell Biochem 93(1): 93-103 Yoon BS, Pogue R, Ovchinnikov DA, Yoshii I, Mishina Y, Behringer RR, Lyons KM (2006) BMPs regulate multiple aspects of growth-plate chondrogenesis through opposing actions on FGF pathways. Development 133(23): 4667-4678

The revised manuscript has been seen and approved by all of the listed authors. Thank you very much for the opportunity to revise our manuscript. We believe the additional experiments have strengthened the story considerably


12