Antithetical NFATc1-Sox2 and p53-mir200 signaling networks govern pancreatic cancer cell plasticity

Transcription

1 The EMBO Journal Peer Review Process File - EMBO Manuscript EMBO Antithetical NFATc1-Sox2 and p53-mir200 signaling networks govern pancreatic cancer cell plasticity Shiv K. Singh, Nai-Ming Chen, Elisabeth Hessmann, Jens Siveke, Marlen Lahmann, Garima Singh, Nadine Voelker, Sophia Vogt, Irene Esposito, Ansgar Schmidt, Cornelia Brendel, Thorsten Stiewe, Jochen Gaedcke, Marco Merenberger, Howard C. Crawford, William R. Bamlet, Jin-San Zhang, Xiao-Kun Li, Thomas C. Smyrk, Daniel D. Billadeau, Matthias Hebrok, Albrecht Neesse, Alexander Koenig, and Volker Ellenrieder Corresponding author: Volker Ellenrieder, University Medical Center Goettingen Review timeline: Submission date: 21 July 2014 Editorial Decision: 20 August 2014 Revision received: 16 November 2014 Editorial Decision: 25 November 2014 Revision received: 09 December 2014 Accepted: 10 December 2014 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: Thomas Schwarz-Romond 1st Editorial Decision 20 August 2014 Thank you very much for your interest in The EMBO Journal and submitting your study on the role of regulatory circuits driving (pancreatic) cancer cell plasticity. Two expert referees have submitted rather consistent comments on your results. These reveal certain interest but also disclose further demands that would amount to major experimental revisions. These should center around the establishment of truly causal relationships as well as rule-in/out the directness of Sox2 regulation by and as putative key target of mir200. This would also clarify/differentiate possible ZEB protein mediated regulation. Along the same, mostly substantiating lines of experiments, gain- and loss of function approaches should be employed and complemented with further reaching expression analyses from primary patient samples as requested by ref#2 and to strengthen the proposed mechanistic network. As both scientific reports are explicit and self-explanatory, I see no further need to repeat the EMBO 1

2 The EMBO Journal Peer Review Process File - EMBO individual requests at any length. Conditioned on addressing these demands, I am very happy to invite your study for a single round of major revisions. Please understand that an amended version will have to undergo a second round of peer- assessment before a final decision on publication in The EMBO Journal would be reached. Therefore, please kindly estimate the demands and focus your efforts to avoid disappointments later in the process. If you see necessary, please do not hesitate to get in touch (due to time constrains preferably via e- mail), also to discuss timeline/feasibility/necessary extensions in light of the transmitted comments. Lastly, I have to formally remind you that The MBO Journal allows only one round of major revisions and the final decision will depend on strength and completeness of the ultimate version of your study. REFEREE REPORTS: Referee #1: This manuscript entitled "Antithetical NFATc1-Sox2 and p53-mir200 signaling networks govern pancreatic cancer cell plasticity" by Singh S. et al. present very interesting data on a critical role of NFATc1-SOX2 in pancreatic cancer progression via activation of the EMT program. Comprehensive analyses in mouse tumor models coupled with molecular analyses in cell lines and human tumors provided quite convincing evidence that induction of NFATc1 and SOX2 are critical for EMT in pancreatic cancer, which is correlated with PDAC progression. What is less convincing in the current version of the manuscript is the role of the p53-mir200 axis in the proposed model due to the following reasons. 1. Fig. 7C shows that overexpression of mirna200c can suppressed EMT gene expression, but it is unclear how many fold of overexpression this is and how it is compared to induction of endogenous mir200 in the KNC mouse tumor model. Because Fig. 7A shows that P53 reepxression can induce mir200c by only 3-fold. The critical experiment is to repress mir200 in the Kras/NFATc1 cells to test whether it results in the similar EMT phenotypes and SOX2 induction, as shown in Fig. 2E with depletion of p53. This experiment is essential to determine whether the role of p53 here is through mir Another key question not answered is whether SOX2 is the key target gene downstream of mir200. No data is provided to show that SOX2 contains mir200-binding sites. Instead, the mir200/zeb double-feedback loop has been well characterized in many cell systems. Therefore, it is equally possible that mir200's key role here is to directly repress ZEB proteins. 3. It will further strengthen the argument to show that repression of mir200s are correlated with increased NFATc1 expression in the mouse and human pancreatic cancer models presented in Fig.1 Additional points: 1. In both the abstract and the model presented in Fig. 8, inflammation is proposed to be the key inducer of NFATc1. Although NFATc1 is an inflammation-induced gene, no evidence is provided showing that the induction of NFATc1 in this setting is due to inflammation. Therefore, this should be revised. 2. Snail1 was shown as a direct target of NFATc1 and Sox2. It is unclear whether Snail1 is the only EMT transcription factor to be directly bound by NFATc1/SOX2 or Zeb1 and Twist1 are also direct targets. 3. In several figures, please explain the experiments in more detail about the figure legends or EMBO 2

3 The EMBO Journal Peer Review Process File - EMBO experimental procedure. For example, Fig.5L, it is not clear whether HA-NFATc1 was transfected into the cell line or the original NFATc1 transgene was immunoprecipitated. If this is an overexpression system, then the interaction between these two proteins are quite weak. 4. In the title, "cancer cell plasticity" is very vague and "Epithelial-Mesenchymal Transition in pancreatic cancer cells" might be more precise. Referee #2: Singh et al. show data about the role of the transcription factor NFATc1 in regulating the plasticity of pancreatic cancer cells in a mouse model and in samples from PDAC patients. They propose a model in which NFATc1 controls EMT and cancer stem cell-like phenotype through the activation of SOX2 transcription, which is counteracted by the (previously reported) EMT-suppressive ability of p53 and mir-200c. The model is interesting but requires further validation in cell lines, an assessment of the translational relevance in patient samples, and the paper requires a substantial statistical revision. Major: - In order to better support the model proposed in Fig8 the experiment in Fig4 should be repeated overexpressing NFATc1 in Panc cells with the simultaneous knockdown of SOX2 and/or with the overexpression of p53, to show that the NFATc1-SOX2 complex is crucial for EMT. - To support the notion that the pathway is active in human PDAC, the NFATc1-SOX2 coexpression should be also shown in patient samples by immunohistochemistry. This would complement the results as shown in FigE3. Additionally, the results of Fig1A, showing a higher proportion of poorly differentiated tumors in NFATc1 positive tumors, should be tested also in patients showing whether NFATc1 is a general driver of EMT and differentiation irrespectively of the p53 status. - Statistical analysis is missing in many experiments throughout the paper; a statistical revision is required. Minor - The authors should explain the reason behind the inconsistent number of patients used in the IHC in the Fig1C and Table E1. - What is the fold change of EMT genes in the array result? - How many sirnas were tested per gene silencing experiment? Were combination of sirnas (like smart pool) used? If not some data need to be reproduced with other sirnas. - The authors state that the cells in the Fig2 had a dedifferentiated phenotype. Pictures of the cells showing morphological alteration could be provided. - Many qpcr result plots have inconsistent scaling in the Y-axis. All graphs should be normalized as in the Fig2A. - Is the morphology of the NFATc1-siRNA cells somehow altered? - On page 11 there is a mention of Fig3H and 3I, which are not present in this submission. 1st Revision - authors' response 16 November 2014 EMBO 3

4 Point-to-point response We thank the Reviewers for their critical analyses and helpful suggestions which we addressed in full to improve the impact of our data and the quality of the manuscript. As described in detail below, we have performed a number of additional experiments and have extensively revised the manuscript in response to the Reviewers comments. We believe that the new results obtained in this revision strengthen our key findings of newly discovered antithetical NFATc1:Sox2 and p53/mir-200c signaling networks that promote EMT and stemness in pancreatic cancer. Below is a point to point response to the individual comments. Additional data have been displayed in the point to point reply to facilitate rapid assessment for the Reviewer, while figures quoted in the revisions had already been shown in the initial version of the manuscript. Reviewer #1 We would like to thank the Reviewer for his general appreciation of our work and his/her insightful comments. This manuscript entitled "Antithetical NFATc1-Sox2 and p53-mir200 signaling networks govern pancreatic cancer cell plasticity" by Singh S. et al. present very interesting data on a critical role of NFATc1-SOX2 in pancreatic cancer progression via activation of the EMT program. Comprehensive analyses in mouse tumor models coupled with molecular analyses in cell lines and human tumors provided quite convincing evidence that induction of NFATc1 and SOX2 are critical for EMT in pancreatic cancer, which is correlated with PDAC progression. What is less convincing in the current version of the manuscript is the role of the p53- mir200 axis in the proposed model due to the following reasons. Fig. 7C shows that overexpression of mirna200c can suppress EMT gene expression, but it is unclear how many fold of overexpression this is and how it is compared to induction of endogenous mir200 in the KNC mouse tumor model. Because Fig. 7A shows that P53 reepxression can induce mir200c by only 3-fold. 1

5 The critical experiment is to repress mir200 in the Kras/NFATc1 cells to test whether it results in the similar EMT phenotypes and SOX2 induction, as shown in Fig. 2E with depletion of p53. This experiment is essential to determine whether the role of p53 here is through mir200. We thank Reviewer #1 for these comments and apologize for not displaying the mirna expression data upon overexpression. We have included these data showing 150fold and 70fold induction of mir-34a and mir-200c, respectively, following overexpression in KNPC cells in Expanded View E5D of the revised manuscript. MiRNA-34a and mirna-200c expression in KNPC cells following overexpression. Additionally, Expanded view E5B shows the endogenous mir-34a and mir-200c levels in KPNC and KNC cells (not shown here, please see actual figure). Here we observed a 30fold down-regulation of mir-34a expression and a 3-4fold decrease of mir-200c expression in KNPC cells compared to KNC cells. As requested, we also performed expression analyses in KNC cells following depletion of mir-200c. Successful depletion of mir-200c led to an up-regulation of the EMT markers Snail and Vimentin as well as of Sox2 as displayed in Expanded View E5E. Additionally, we extended our studies and conducted Western Blot analysis, which showed strong up-regulation of ZEB1, Snail, Twist and Vimentin after suppression of mir-200c (Figure 7E of the revised manuscript). 2

6 QRT-PCR reveals successful mirna-200c depletion in KNC cells (upper left panel) and displays upregulation of the indicated markers (upper right panel, both graphs are depicted in Expanded View E5E of the revised manuscript). Western Blot analysis following mir-200c loss in KNC cells (lower panel, included in Fig. 7E of the revised manuscript). In the current manuscript we demonstrate that EMT and stemness characteristics are controlled by the p53 status in pancreatic cancer. Furthermore, recent reports have established the p53/mir-200c loop as a central player in the suppression of cancer cell progression (Chang et al., 2011). Here we show that depletion of mir-200c leads to a strong increase of EMT and stemness markers on RNA and protein level in cells with a wt p53 status and overexpressing NFATc1 (KNC). We believe that these new results strongly support our hypothesis that the p53/mir-200c loop can counteract the NFATc1/Sox2 mediated induction of EMT and stemness in pancreatic cancer. 3

7 Another key question not answered is whether SOX2 is the key target gene downstream of mir200. No data is provided to show that SOX2 contains mir200- binding sites. Instead, the mir200/zeb double-feedback loop has been well characterized in many cell systems. Therefore, it is equally possible that mir200's key role here is to directly repress ZEB proteins. We fully agree with Reviewer #1 s assertion that mir200 might directly affect expression of ZEB1. In fact, there is growing evidence for the notion that mir200 regulates both Sox2 and ZEB1. ZEB1 has been described as a crucial inducer of tumorigenicity, invasion and metastases (Wellner et al., 2009). ZEB1 expression is tightly controlled by mir-200c (Brabletz and Brabletz, 2010). Moreover and as discussed in the new version of our manuscript, Brabletz and colleagues performed an outstanding study in which they describe a ZEB1/miR-200 feedback loop operational in breast and pancreatic cancer (Brabletz et al., 2011). Activation of ZEB1 in these models results in enhanced Notch activation via inhibition of mir-200 s repressive effects on Notch pathway components. Since ZEB1 inhibits activation of mir-200c, it indirectly induces stemness features and drug resistance in tumor cells. Additionally to ZEB1 and BMI1, Sox2 represents a direct target of mir-200c and the negative regulation of Sox2 by mir-200c depends on direct binding to a mir-200c binding site on the 3 UTR of Sox2 (Lu et al., 2014). In accordance to the literature, mir-200c over-expression in pancreatic cancer cells results in a significant down-regulation of EMT markers, including ZEB1 (Fig. 7C, D), while silencing of the micro-rna leads to its up-regulation (Fig. 7E, E5E), clearly confirming a negative effect of mir-200c on EMT marker expression. Additionally, the same experimental settings displays a significant attenuation of Sox2 expression following overexpression of mir-200c (Fig. 7C,D), while loss of the microrna results in Sox2-upregulation (E5E). Silencing of Sox2 reduces NFATc1 binding to and enhancer activation and promoter activity of Snail1 (Fig. 5M), a process associated with a reduction of EMT marker expression on mrna level (Fig. 5N). Therefore, mir-200c dependent reduction of stemness and EMT features in pancreatic cancer entails two inter-connected pathways: mir-200c directly represses 4

8 ZEB1 and further EMT transcription factors. In addition, mir-200c reduces Sox2 expression and thus targets EMT via an indirect pathway as well. It will further strengthen the argument to show that repression of mir200s are correlated with increased NFATc1 expression in the mouse and human pancreatic cancer models presented in Fig.1 As requested by Reviewer #1, we compared mir-200c and NFATc1 expression levels in human and murine pancreatic cancer models. The three human pancreatic cancer cell lines L3.6, Panc1 and Imim-PC1 differ in their NFATc1-expression levels. Panc1 is the cell line with the strongest expression. L3.6 expresses moderate levels of NFATc1, while Imim-PC1 cells are almost negative for NFATc1 expression. As shown below, the mir-200c expression inversely correlates with NFATc1 expression, as the strongest mir-200c signal could be detected in Imim-PC1 cells (low for NFATc1 expression), while detection of mir- 200c failed in Panc1 cells (high for NFATc1 expression). QRT-PCR shows mir-200c and NFATc1 expression in human pancreatic cancer cells lines. 5

9 Moreover, three out of four tumor bearing KPC mice showed an inverse correlation of NFATc1 and mir-200c expression levels (compare right panel of the graph below). To further extend our observations from mice and cell lines, we performed analyses in human pancreatic cancer samples. As the stromal component dominates in pancreatic cancer and tumor cells represent only a small percentage of the tumor burden, we performed micro-dissection of pancreatic cancer samples to isolate tumor-cell rich areas of pancreatic cancer. After micro-dissection, RNA was isolated to investigate the NFATc1 and mir-200c expression levels in these samples. As depicted here, NFATc1 and mir-200c expression inversely correlate in 6 out of 8 human pancreatic cancer specimens. Left panel: NFATc1 and mir-200c expression levels in NFATc1-positive and negative human pancreatic cancer samples following tissue-micro-dissection of tumor cell-rich areas. Right panel: mir-200c expression in KPC tumors with differential NFATc1 expression. 6

10 All together, these data indicate an inverse correlation of NFATc1 and mir-200c expression in pancreatic cancer and thus strongly support our notion of antithetical NFATc1:Sox2 and p53:mir-200c pathways in pancreatic cancer. Additional points: In both the abstract and the model presented in Fig. 8, inflammation is proposed to be the key inducer of NFATc1. Although NFATc1 is an inflammation-induced gene, no evidence is provided showing that the induction of NFATc1 in this setting is due to inflammation. Therefore, this should be revised. We thank the Reviewer for this important comment. Indeed, NFATc1 activation is tightly linked to acute or chronic inflammation, but our studies do not allow any conclusion regarding the impact of inflammation on induction of EMT and stemness in pancreatic cancer. Therefore, we changed the corresponding statements in the abstract and adapted the model in Figure 8 as requested. Proposed model of antithetical NFATc1-Sox2 and p53-mir-200 networks in pancreatic cancer plasticity 7

11 Snail1 was shown as a direct target of NFATc1 and Sox2. It is unclear whether Snail1 is the only EMT transcription factor to be directly bound by NFATc1/SOX2 or ZEB1 and Twist1 are also direct targets. To address this important question, we performed ChIP analysis in KNPC cells and pulled NFATc1 and Sox2 to study their binding affinities to ZEB1 and Twist1. As shown here, neither NFATc1 nor Sox2 bind to the ZEB1 enhancer or promoter, respectively. In the case of Twist1, we detected a weak binding of NFATc1 to the Twist1 enhancer and promoter. We observed no significant enrichment of Sox2 binding on the Twist enhancer, while binding was detectable on the Twist1 promoter. In contrast to that and highly reproducible, NFATc1 binding was observed on the Snail enhancer, while Sox2 enrichment was detected on the Snail promoter (Figure 5M, 7F, 7G). Therefore we conclude, that cooperative NFATc1:Sox2-mediated induction of EMT in pancreatic cancer is the consequence of Snail1 activation by direct enhancer and promoter binding. Nevertheless, our expression analyses following depletion of Sox2 (Figure 5N) or pharmacological or genetic inactivation of NFATc1 (Figure 4B, C) display that down-regulation of EMT markers is not limited to Snail1, but also includes Twist1 and ZEB1. Therefore we consider that NFATc1 and Sox2 indirectly target ZEB1 and Twist1 or bind to additional promoter and enhancer regions of both EMT genes. ChIP-analysis in KNPC cells showing NFATc1 and Sox2 occupancy on Twist and Zeb1 promoter and enhancer. 8

12 In several figures, please explain the experiments in more detail about the figure legends or experimental procedure. For example, Fig.5L, it is not clear whether HA- NFATc1 was transfected into the cell line or the original NFATc1 transgene was immunoprecipitated. If this is an overexpression system, then the interaction between these two proteins are quite weak. We apologize for the confusion and have expanded the explanations of the experimental designs in out figure legends and in the material and methods section of the revised manuscript. With regard to the specific question raised, the immunoprecipitation presented in Figure 5L shows an interaction of endogenous HA- NFATc1 and Sox2 in KNPC cells. In the title, "cancer cell plasticity" is very vague and "Epithelial-Mesenchymal Transition in pancreatic cancer cells" might be more precise. We agree with Reviewer #1 that "cancer cell plasticity" is a somewhat vague term. We did have extensive discussions among the authors on this topic, but feel that considering that our studies are not limited to EMT regulation in pancreatic cancer, but also investigate stemness features, we prefer to keep the current title, as "cancer cell plasticity" summarizes EMT and stemness programs. 9

13 Reviewer #2 We thank the Reviewer for his/her supportive and constructive comments that we have addressed in full as described in detail below. General Comments Singh et al. show data about the role of the transcription factor NFATc1 in regulating the plasticity of pancreatic cancer cells in a mouse model and in samples from PDAC patients. They propose a model in which NFATc1 controls EMT and cancer stem celllike phenotype through the activation of SOX2 transcription, which is counteracted by the (previously reported) EMT-suppressive ability of p53 and mir-200c. The model is interesting but requires further validation in cell lines, an assessment of the translational relevance in patient samples, and the paper requires a substantial statistical revision. Major: In order to better support the model proposed in Fig8 the experiment in Fig4 should be repeated overexpressing NFATc1 in Panc cells with the simultaneous knockdown of SOX2 and/or with the overexpression of p53, to show that the NFATc1-SOX2 complex is crucial for EMT. We thank Reviewer #2 for this important comment. As requested, we analyzed the impact of Sox2-depletion on NFATc1-mediated up-regulation of EMT transcription factors in human pancreatic cancer cell lines. For this approach we stably transfected Sox2 shrna in L3.6 cells. Following over-expression of NFATc1 in control shrna clones, we observed a strong up-regulation of the EMT-markers ZEB1 and Snail1. In contrast to wild-type cells, Sox2 depletion diminished the effect of NFATc1 overexpression in this cell line. Of note, Sox2 shrna treatment does not affect the expression levels of NFATc1. The data are included in Fig. 5O and Expanded View E3F, E3G of the revised manuscript. These results support our finding that Sox2 expression is critical for NFATc1 mediated induction of key genes that promote EMT in pancreatic cancer. 10

14 QRT-PCR shows successful Sox2 depletion and NFATc1 overexpression in L3.6 cells (upper panels) as well as abolished NFATc1-mediated upregulation of EMT markers in the absence of Sox2 (lower panel). The figures are included in Fig. 5O and Expanded View E3F, E3G in the revised manuscript. To support the notion that the pathway is active in human PDAC, the NFATc1-SOX2 co-expression should be also shown in patient samples by immunohistochemistry. This would complement the results as shown in FigE3. Additionally, the results of Fig1A, showing a higher proportion of poorly differentiated tumors in NFATc1 positive tumors, should be tested also in patients showing whether NFATc1 is a general driver of EMT and differentiation irrespectively of the p53 status. We thank Reviewer #2 for this crucial comment and have performed comprehensive additional immunohistochemical analyses to address this issue. Sox2 and NFATc1 staining in a TMA of 42 pancreatic cancer patients is displayed in Expanded View 11

15 E3E of the revised manuscript. In congruency with our results obtained from analysis in primary pancreatic cancer cell lines, we observed co-expression of Sox2 and NFATc1 in 79% of human pancreatic cancer samples. Thus our new results are in agreement with our previous findings and support the relevance of an oncogenic NFATc1-Sox2 axis in pancreatic cancer. Pie chart shows the percentage of NFATc1-positive human pancreatic cancer samples among Sox2 positive tumors (the figure is included in Expanded View E3E of the revised manuscript) To investigate, whether NFATc1 activation correlates with dedifferentiation irrespectively of the p53 status we sequenced a set of human pancreatic cancer samples to allow distinction between p53 wild-type and p53 mutant situations. Expanded View E1A shows the percentage of well (left panel: G1/G2) and poorly (right panel: G3) differentiated human pancreatic cancer samples with respect to the p53 status among NFATc1 positive samples. 88.9% of the well-differentiated NFATc1 positive tumors displayed a p53 wild-type situation (light blue), while only 36.4% of the G3 tumors bared a wild-type p53 status. In contrast to that, p53 mutations were detected in only 11.1% of the well differentiated tumors, while 63.6% of the G3 tumors had a p53 mutation. Therefore, only in one third of the tested human pancreatic cancer samples, NFATc1 drives dedifferentiation despite functional p53- signaling, whereas p53 tumor suppressor functions were disrupted in the wide majority of NFATc1-positive and poorly differentiated human pancreatic cancer samples. Hence, NFATc1 driven dedifferentiation can occur irrespectively of the p53 status, but the high correlation between dedifferentiation and p53 inactivation in NFATc1 positive tumors argues that p53 deficiency significantly accelerates NFATc1- driven dedifferentiation in pancreatic cancer. 12

16 Statistical analysis is missing in many experiments throughout the paper; a statistical revision is required. We thank the Reviewer for this note and have performed a careful statistical revision of the paper. For example, we calculated p-values, accentuated significant (p<0.05) differences and included a description of our statistical analysis in the materials and methods section of the manuscript. Minor The authors should explain the reason behind the inconsistent number of patients used in the IHC in the Fig1C and Table E1. We apologize for the confusion caused by inconsistent numbers of patients in Figure 1C and Table 1. NFATc1 and p53 stainings were performed in a total of 161 patient samples. This high number of samples was achieved by combining two different TMA datasets (Mayo Clinic TMA, TMA derived from Irene Esposito, Munich). Clinical data, including the survival of the patients, was only available from the Mayo Clinic TMA data base. Therefore the correlation of the p53 status, age of the patients at diagnosis, survival and grading could only be conducted in a total of 129 patient samples. This explanation was added to the Material and Methods section. What is the fold change of EMT genes in the array result? Our revised manuscript comprises an Exel file of the array which contains the fold changes of gene expression. 13

17 How many sirnas were tested per gene silencing experiment? Were combination of sirnas (like smart pool) used? If not some data need to be reproduced with other sirnas. We apologize for the omission of this information in the original version of the manuscript. To guard against off-target effects we used a set of 2-3 sirnas for all of our experiments. This information is now included in the Material and Methods section of the revised manuscript. The authors state that the cells in the Fig2 had a dedifferentiated phenotype. Pictures of the cells showing morphological alteration could be provided. We thank Reviewer #2 for this suggestion and admit that our statement Loss of p53 caused robust transcriptional activation of gene signatures implicated in EMT, stemness, and metastasis ( ) and these expression changes were accompanied by acquisition of a fast migrating, dedifferentiated phenotype, evidenced by wound healing experiments and time-lapse microscopy (Figure 2C) (pp.8-9 of the revised manuscript) was somewhat misleading. KNC cells already display a very spindleshaped phenotype and we did not observe additional morphological changes of the cells following depletion of p53 (please compare light microscopy images below). Nevertheless, we do see a significant enhancement of EMT marker expression and an accelerated wound-closure rate upon p53 silencing. To clarify this issue we deleted the term dedifferentiated in the quoted passage in the revised manuscript. 14

18 Light microscopy reveals the cell morphology of KNC cells following depletion of NFATc1 and p53, respectively. Western Blot analysis was used for knockdown-control. Many qpcr result plots have inconsistent scaling in the Y-axis. All graphs should be normalized as in the Fig2A. As requested, we critically revised our graphs, re-labeled the Y-axis and displayed our mrna results as x-fold expression wherever it seemed suitable. In the case of figures 4A, 5A and E5B, we believe that displaying the data as relative mrna expression levels is preferable, as it demonstrates expression of the indicated markers in different GEMM tissues in which we cannot normalize to a basal expression level. Is the morphology of the NFATc1-siRNA cells somehow altered? 15

19 Indeed, the morphology of KNC cells is altered following NFATc1 depletion. The cells demonstrate a less spindle-shaped phenotype, possibly due to reduced activation of mesenchymal traits (please see light microscopy images and Western Blot analysis for knockdown-control on p.15 of the point-to-point response). On page 11 there is a mention of Fig3H and 3I, which are not present in this submission. We thank the Reviewer for this important statement and apologize for the incorrect citing of our data. The conservation of NFAT binding sites within the Sox2 gene among vertebrates and NFATc1 binding in Sox2 enhancers are not displayed in Figure 3H and 3I, but in Figure 5H and 5I. The revised manuscript contains the correct references to the appropriate figures. 16

20 The EMBO Journal Peer Review Process File - EMBO nd Editorial Decision 25 November 2014 Thank you very much for the revised study that was highly appreciated by one of the original referees. REFEREE REPORT: Referee #2: Excellent revision work! EMBO 4