BRN2 is a non-canonical melanoma tumor-suppressor

While the major drivers of melanoma initiation, including activation of NRAS/BRAF and loss of PTEN or CDKN2A, have been identified, the role of key transcription factors that impose altered transcriptional states in response to deregulated signaling is not well understood. The POU domain transcription factor BRN2 is a key regulator of melanoma invasion, yet its role in melanoma initiation remains unknown. Here, in a BrafV600E PtenF/+ context, we show that BRN2 haplo-insufficiency promotes melanoma initiation and metastasis. However, metastatic colonization is less efficient in the absence of Brn2. Mechanistically, BRN2 directly induces PTEN expression and in consequence represses PI3K signaling. Moreover, MITF, a BRN2 target, represses PTEN transcription. Collectively, our results suggest that on a PTEN heterozygous background somatic deletion of one BRN2 allele and temporal regulation of the other allele elicits melanoma initiation and progression.

3) The data to support the claim of two novel mechanisms linking BRN2 loss to PTEN expression are preliminary. a. The microarray results provided are under-powered. In the methods the authors state that 13 tumors were profiled. In the presented results, only 7 total were used, inclusive of 5 experimental and 2 controls. What was the rationale for omitting the other 6 tumors? Were the samples representative of independent mice? Why were only 2 controls used? That the study was underpowered is clear from Table S3 which shows that no genes were significantly altered in expression when considering the adjusted p value. b. The relationships between either BRN2 or MITF and PTEN are both novel and of interest to the community. These studies could be strengthened by performing the experiments in more than one human cell line, by using a luciferase reporter to determine wither transcription factor binding increases or decreases transcription, and by performing both the BRN2 and MITF experiments in the same sets of cells. As is, it is unclear how the MITF data presented in figure S8 are connected with the rest of the manuscript.
4) The title, abstract, introduction and discussion contain misrepresentations of the previous literature and exaggerations of the results. a. Although the authors cite the majority of relevant work, this reviewer disagrees with the representations of the data within those studies and suggests a more careful reading and comprehensive assessment of the cited papers. Specifically, the contribution of BRN2 to melanoma metastasis has been looked at in vivo by multiple groups as has the consequence of BRN2 inhibition in normal melanocytes, and the relationship between MITF and BRN2 is more complex then suggested. b. Many of the conclusions drawn are exaggerated and need to be toned down. In particular, conclusions based upon the TCGA analyses and microarray data should be omitted unless the issues outlined above are addressed.

Reviewer #2 (Remarks to the Author):
In this study the authors argue that BRN2 is a haplo-insufficient tumor suppressor whose loss promotes melanoma initiation and progression. The mechanistic basis for this is proposed to be BRN2 regulation of PTEN transcription, either directly or via repression of MITF. Whereas previous studies have indicated that BRN2 promotes melanoma invasion (e.g. Goodall et al., 2008;Fane et al., 2017), the role of BRN2 in melanoma initiation and progression remains to be fully investigated. This report provides important clues and contributes to a better understanding of the role of this transcription factor in melanomagenesis and possibly other BRN2-dependent tumors The BRN2 locus shows frequent copy number loss in cutaneous melanoma metastases, although these deletions also affect many other genes implicated in melanomagenesis (Fig. 1A-B). Additional data supporting that BRN2 is the key contributor to melanoma initiation and progression would significantly strengthen the manuscript.

Major issues
Mono-allelic loss of BRN2 is associated with worse overall survival. Considering that central thesis is that BRN2 is a tumor suppressor whose monoallelic loss promotes melanoma initiation and progression, frequency of BRN2 loss (in addition to mRNA levels) at early/pre-metastatic stages of melanoma should be included to further support the weak association BRN2 low vs. high mRNA levels with survival ( Fig. 1G). Is this an early event in melanomagenesis?
Mouse model indicates that mono-allelic loss of BRN2 in neonatal mice promotes melanomagenesis. Recommend further experiments to support BRN2 tumor suppressive function. Does re-expression of BRN2 in cells that have lost BRN2 suppress (i) colony formation in vitro (ii) tumor formation in mice?
Would be useful to show survival data for the different cohorts of mice as well as histology of the primary tumors including staining with melanoma and proliferation markers.

Do tumors progress to metastatic melanoma beyond LN infiltration?
What effects does bi-allelic loss of BRN2 in melanoma cells have on tumorigenicity? What is the impact of silencing the other genes in Chr6 (e.g. Arid1A, ROS)?
Evidence from clinical samples of melanoma in situ/invasive melanoma vs. adjacent precursor nevi suggests that BRN2 levels increase during transition to invasive melanoma (Zeng et al., 2018). Please discuss this and address whether mono-allelic loss of BRN2 occurs early in melanoma development.
All main figure in vitro analyses rely on comparisons of one transformed cell line and one nontransformed melanocyte (e.g. following BRN2 knockdown; Fig. 6D). For greater generalizability please include additional cell lines such as those in Fig. S8 and include cells with loss of PTEN.
Authors suggest that BRN2 may induce PTEN transcription partially by repressing MITF. While references for the link between BRN2 and MITF are cited, no evidence is presented to show that BRN2 inhibits MITF expression in these models or that BRN2 knockdown/loss enhances MITF expression. PTEN levels are instead monitored following direct knockdown of MITF. Include data showing regulation of MITF by BRN2 and corresponding PTEN levels.
To demonstrate that BRN2 loss can also repress PTEN indirectly via the induction of MITF, concomitant silencing of BRN2 and MITF should be performed. Likewise, MITF levels could be assessed in BRAF/PTEN/BRN2 tumors.
MITF promotes proliferation, whereas low MITF is associated with an invasive, slow-cycling phenotype. Could MITF also be contributing to the effects of Brn2 loss on proliferation and LN infiltration?
Since the overall conclusion of the study is that "BRN2 loss reduces PTEN transcription in vitro and in vivo, thus ramping up PI3K signaling and inducing both the initiation of melanoma and the formation of metastases" it would be interesting to determine the effects of BRAFi+PI3Ki inhibitors on BRAF/PTEN/BRN2 tumor initiation and LN metastasis.
Minor issues BRN2 mRNA levels are lower in invasive melanomas (T3+T4) than localized. Please incorporate Fig. S2B into Fig. 1 as this shows the most clinical relevant staging (Clark's level superseded by AJCC staging).
What is PTEN mutation status in cell lines used?
Based on the phenotype elicited by loss of Brn2, the authors state that "Brn2 acts as a tumor suppressor in vivo, and its loss induces melanoma initiation and increases tumor growth rate"; reword to more precisely indicate that loss of Brn2 in the context of monoallelic PTEN… "Cell lines express Pten and Brn2 mRNA…(data not shown)" -include the data in supplemental figures.
"Cell number was monitored" after BRN2 knockdown (Fig. 4D). Cell cycle analysis to assess relative proportion of cells proliferating would further support the conclusions.
Reduction in PTEN positivity upon BRN2 loss (mouse tumor IHC - Fig. 6A) is not reflected in Western blot levels (Fig. 6B). PTEN protein levels look comparable between WT and BRN2 when normalized to actin intensity. Provide densitometry values normalized to actin in Fig. 6B. Also probing for MITF is suggested.

5
In the manuscript "BRN2 is a non-canonical melanoma tumor-suppressor", the authors use a combination 6 of TCGA data analysis, a novel mouse model, and in vitro cell line experiments to investigate how loss of 7 BRN2 influences human and mouse melanoma progression. The main conclusions the authors draw are 8 that haploinsufficiency of BRN2 is sufficient for melanoma initiation, that haplo-insufficiency of BRN2 is 9 sufficient for melanoma metastasis, that BRN2 loss increases PTEN expression, and that PTEN expression 10 is increased through two novel molecular mechanisms. Although the manuscript does contain some well-11 conducted and interesting experiments, the majority of the authors' conclusions are unsubstantiated.

13
Strengths of the study 14 15 1) The authors have convincingly demonstrated an interesting, previously unknown, relationship between 16 BRN2 loss and PTEN activation. Specifically, the authors demonstrate:

22
2) The mouse model the authors have generated is elegant and likely will be useful for the field.

24
3) Overall, this reviewer believes these observations are novel and would be of interest to the fields of 25 melanoma genetics, melanocyte and neural crest development, and cancer cell signaling.

26
This reviewer encourages the authors to consider re-framing this study to focus on these strengths.

28
We thank the reviewer for these positive comments

30
Weaknesses of the study 31 32

33
The analyses conducted with the TCGA data are flawed in logic, approach and statistical analyses. Over-34 all, this reviewer does not agree that the authors can conclude that "The BRN2 locus is frequently deleted 35 in human melanoma and reduced BRN2 mRNA levels correlate with reduced overall survival and 36 worsening of prognostic factors" as they claim.

39
We understand the reviewer's point of view. We built this study on the previous literature showing the

51
BRN2 is located on Chr. 6q, which is among the most frequently lost chromosomal arms in melanoma 52 (Guan, et al (1998) Cancer Genet Cytogenet, PMID: 9844599). Any gene on this arm will demonstrate the 53 same pattern the authors present here. To make their argument, the authors must convincingly show that 54 2 BRN2 is lost by focal deletions at a frequency significantly greater than all other genes on Chr. 6q.

57
One of the goals of this study, as the reviewer points out, was to generate a relevant mouse model for 58 human melanoma to evaluate the causal role of the loss of BRN2 during melanomagenesis, and especially 59 during initiation since it is the best way to address functionally this question. Indeed, and we fully agree 60 with the reviewer; Brn2 is located on chromosome 6q and this chromosomal arm is frequently lost.

62
As suggested by the referee we have added to our manuscript the 1,2 references since Bastian et al 63 analyzed 32 primary melanoma and Guan et al 21 melanoma cell lines. From the TCGA, our study 64 validates and solidifies the previous findings using a total of 367 melanomas.

66
As presented in Figure 1B some deletions are limited to the BRN2 locus, but indeed the vast majority 67 cover other loci including CCNC, ROS1 and ARID1B, which are more apical than the BRN2 locus. We re-68 analyzed the SKCM TCGA study data using clinical annotations last updated August 8, 2019 to evaluate 69 overall survival comparing the mono allelic loss and the diploid state for BRN2, CCNC, ROS1 and ARID1B.

70
These data are shown in Figure 1C and Figure S1. It appears that at this stage of the follow up the clinical 71 study, statistical significance is not reached for BRN2 (p = 0.067), but for each of the other genes the 72 statistical significance is less significant than of BRN2 (CCNC, p = 0.086; ROS1, p = 0.27; ARID1B, p = 0.66)

74
There is therefore a trend for BRN2, and potentially for CCNC, to demonstrate a lower overall survival 75 when mono-allelic compared with the diploid state. These trends were not observed for ROS1 and 76 ARID1B. In conclusion, after this more precise analysis the referees statement that 'Any gene on this arm 77 will demonstrate the same pattern the authors present here' is not correct.

79
Moreover, as the referee will appreciate, this study does not focus on CCNC, ROS1 and/or ARID1B, and 80 our goal was not to show that these genes were or were not of importance during melanomagenesis, 81 though in the future, it may be of importance to analyze in vivo the role of CCNC, ROS1 and ARID1B during 82 melanoma initiation.

84
The first goal of this analysis was not to show that BRN2 was specifically lost in human melanoma, as a 85 result of a focal deletion targeting BRN2, but to evaluate the number of BRN2 alleles that were lost in 86 human melanoma in a descriptive way. The second goal of this analysis was to evaluate whether this loss 87 was independent of BRAF or NRAS mutation status to better design our mouse melanoma model (see 88 Figure S2).

90
Importantly we demonstrate genetically in mice that the loss of Brn2 on a Braf/Pten background induces 91 more efficiently the number of melanomas per mouse. The mouse models revealed that the loss of Brn2 92 did not affect the level of expression of Ccnc, Ros1 and/or Arid1b in melanoma.

94
This study highlights the potential importance of Brn2 in melanomagenesis in vivo but it is not designed to 95 address the function(s) of the other genes on Chr.6q.

98
b. The authors have selectively omitted data that refute their hypothesis. When this reviewer conducted 99 the same analysis with the same TCGA data, the graphed frequencies were replicated (36% diploid, 53% 100 Het, 1.1% deletion), however there were also 7.1% gain and 0.6% amplification. As the gain percentage is 101 greater than the deletion percentage, the figure caption "BRN2 locus is frequently deleted in human 102 melanoma" is false.

105
We understand the point of the referee. However, we did not omit this information, it was mentioned in 106 the original version (lines 13-14 p6) the other categories (gain and amplification). We have replaced the caption of Figure 1A "The Brn2 locus is frequently deleted in human melanoma …" by "One Brn2 allele is 3 frequently lost in human melanoma…". We also added in the figure legend "Gain and/or amplification of 109 the BRN2 locus (7.9%) are not displayed".

112
c. Similarly, the authors omit Clark's Level 1 in Figure 1E, and provide no rationale for the grouping of 113 T1+T2 and T3+T4 in figure S2B. Further, the data distributions do not appear to be appropriate for 114 application of the Mann-Whitney test. The authors should provide the raw data or statistical tests 115 demonstrating that the shape of all distributions is the same. Overall, the data relating BRN2 mRNA 116 expression to prognostic factors are unconvincing and should be substantiated by orthogonal studies.

119
We did not want to omit Clark's Level 1 in the previous Figure 1E. Indeed Clark's levels are only relevant 120 for primary melanoma. In order to simplify our message, we decided to remove the panels associated 121 with the Clark's level and Breslow index (previous Figure S2B). It is important to note that modification of

128
A third category of "6q loss" should also be included, and compared to focal deletions of BRN2 and each 129 other gene on this chromosomal arm.

132
As the referee mentioned, loss of 6q and survival has been very well established. Therefore we did not 133 want to display these data. However, we evaluated the loss of 6q (n = 61) with the mono-allelic loss of 134 Brn2 (n = 132) (see Figure S1P). Indeed, it appears that the 6q loss is associated with a worse prognostic 135 than the BRN2 mono-allelic loss. This result indicates that other gene(s) located on 6q are of importance 136 in melanomagenesis. As suggested by the referee we have added to our manuscript the 3 reference where 137 they analyzed 53 cases.

140
2) Although the authors convincingly show with their mouse model that partial loss of BRN2 increases 141 melanoma proliferation, this finding is expected from many previous publications, as the authors already 142 discuss. The authors go on to claim that haplo-insufficiency is "sufficient" for both initiation and 143 metastasis. However, these conclusions are not substantiated by the presented data:

145
1A6. 146 1) We do not claim that haplo-insufficiency of Brn2 is "sufficient" for both initiation and metastasis. We 147 claim that the haplo-insufficiency of Brn2 promotes more efficiently both initiation and metastasis 148 formation on a (Tyr::CreERt2/°; Braf oncflox/+ ; Pten f/+) genetic background that was tamoxifen induced.

149
It was very clear in the legend and the text. Perhaps, it was confusing in the abstract. We changed this 150 sentence as follows "Here, in a BrafV600E Pten+/-context, we show that BRN2 haplo-insufficiency 151 promotes melanoma initiation and metastasis."

153
2) The referee focuses only on proliferation ( Figure 2C). There is another very important issue to consider; 154 the number of independent melanomas is higher in Braf-Pten-Brn2-het/hom mice than in Braf-Pten-Brn2-

155
WT mice showing that independent initiation is promoted when the Brn2 gene is defloxed ( Figure 2B). We

171
Brn2-WT mice were sacrificed at 4 weeks with an average of 8 tumors even they did not reach a total 172 volume of 2cm^3 except one mouse that was sacrificed earlier (three weeks).

173
As mentioned in the Materials and Methods, mice were killed either four weeks after the appearance of 174 the tumors or when the total volume of the tumors reached 2 cm^3. We added in the materials and 175 methods that the mice were checked two-three times a week to estimate the volume of the tumors. The

177
These results collectively support the conclusion that initiation (proliferation and senescence

178
[bypass/escape]) is more effective when the amount of Brn2 is lower than normal (het or hom). We 179 added this information in the figure legends.

182
b. Similarly, the authors claim an observed increase in metastasis. However, the authors note that the 183 number of metastatic events were actually the same. This is despite an increase in primary tumor number 184 and size. The observed differences were the size of the lymph node and the number of melanocytes in the 185 node -both of which could be explained by more rapidly dividing cells. A more reasonable conclusion is 186 that the number of metastatic events is the same upon BRN2 loss (or even less relative to total primaries), 187 but that the BRN2 het melanocytes proliferate faster. The authors should consider quantifying distal 188 metastases, which the activated BRAF, PTEN deleted mice are known to induce. If partial loss of BRN2 can 189 substitute for the loss of the second PTEN allele, the expectation is that activated BRAF, partial PTEN loss, 190 partial BRN2 loss mice will also present with distal metastases.

193
Our goal was not to say that the "number of metastatic events were actually the same" as the referee

206
We have added a sentence in the discussion on this issue.

208
In our conditions, we did not observe distal metastases during the follow-up period that is constrained by

218
c. The authors only observe a difference on the specific background of activated BRAF and PTEN partial 219 loss, but neither PTEN diploid nor PTEN full loss. This is a niche circumstance, the relevance of which to 220 human melanoma is unclear, as PTEN loss generally occurs quite late during human melanoma 221 progression. In either case, the conclusions of BRN2 loss being "sufficient" for any of these phenotypes 222 are misleading. The authors also chose not to present the PTEN full loss data, nor comment on BRN2 full 223 loss data, both of which are critial controls.

226
We agree with the reviewer that the BRAF activated and a PTEN partial loss represents a specific genetic 227 background. These two genetic events (BRAF and PTEN) are the most frequent alterations of the MAPK 228 (50-60%) and PI3K (30-50%) signaling pathways, respectively. The activation of BRAF and the loss of PTEN 229 are two independent events occurring in about 15-30% of the cases. In this sense, we do not believe that 230 it is a niche circumstance for these patients. We developed this point for humans in Figure S2.

232
We now included the Pten diploid/full loss and Brn2 full loss:     Figure S3). This information is now included in the text and the full data are now presented.

252
3) The data to support the claim of two novel mechanisms linking BRN2 loss to PTEN expression are 253 preliminary.

256
Regarding this point, concerning the regulation of PTEN by BRN2, we show that:

257
(i) In vivo the level of PTEN is reduced when Brn2 is reduced.

258
(ii) In cellulo we show that the reduction of Brn2 induces a reduction of Pten mRNA in three independen 259 human melanoma cell lines (DAUV-1, Gerlach and SKmel28).

260
(iii) ChIP experiments shows that Brn2 binds to the promoter and to exon 1 of Pten.

261
(iv) Initially, we contacted several times a group at the MD Anderson to get the human PTEN promoter, 262 but were unfortunately unable to obtain it from the authors of BBRC, 292, 422-426. "Promoter analysis of 263 tumor suppressor gene PTEN: identification of minimum promoter region". We therefore cloned the PTEN 264 promoter upstream of a luciferase reporter and showed that increases expression of BRN2 activates the 265 PTEN promoter and decreasing the BRN2 level decreases PTEN promoter activity. It was indeed a real tour 266 de force to get this construct since the amount of GC is very high in this promoter. In order to show that 267 this regulation is conserved during evolution, we also cloned the mouse PTEN promoter. We showed that 268 BRN2 regulates PTEN promoters in both species. These constructs will of course be available for the 269 community.

279
(iii) Using the same Cut and Run seq approach, four distal enhancers exhibited greater (at least 2-fold)

283
We believe that at this point, these data are sufficient to show that BRN2 and MITF regulate the

289
What was the rationale for omiting the other 6 tumors? Were the samples representative of independent 290 mice? Why were only 2 controls used? That the study was under-powered is clear from Table S3 which   291 shows that no genes were significantly altered in expression when considering the adjusted p value.

294
The reviewer is correct. We increased the number of samples to generate similar information.

295
Indeed we initially wanted to profile 13 tumors but we had a major problem with the extraction of RNA 296 from WT for unknown reasons. In this respect, we generated more mice and tumors from new crosses.

297
We produced five more Braf-Pten-Brn2-het tumors and three more Braf-Pten-Brn2-WT tumors. We also 298 generated tumors that were homozygous for Brn2 (n=10). RNAs were extracted prior to repeating the 299 transcriptomic analysis with all samples (10 from Braf-Pten-Brn2-het, and 5 from Braf-Pten-Brn2-WT 300 tumors, and 10 from Braf-Pten-Brn2-homo). A new Table S3 is now provided. Moreover, we indicated in 301 the Materials and Methods that all tumors were derived from independent mice. We analyzed these 302 transcriptomic analyses accordingly. The results are given in Figure 4 and Figure

317
We partly answered this question in 1A10.    * We initially evaluated the connection between BRN2 and PTEN on two cell lines. One was human (Dauv-326 1) and one is mouse (melan-a). It shows that this regulation is conserved during evolution; this is an 327 important point to stress. We tested our hypothesis on three novel human melanoma cell lines (Gerlach 328 SKmel28, and 501Mel) as requested by the reviewer. As requested, we performed the knock down of 329 BRN2 in these three cell lines (Gerlach, SK28, and 501Mel), and we confirmed that the reduction of BRN2

340
In the discussion, we come back on the high complexity of regulation of MITF by BRN2 associated with 341 AXL.

343
To answer the referee concerning the luciferase reporter experiment, we cloned, sequenced and 344 generated the mouse and human PTEN promoter constructs driving luciferase. We transfected the human

366
We would like to answer this point, but it would have helped if the referee had been more precise. We 367 have added several citations that we omitted in the previous version.

368
The referee considers that the contribution of BRN2 to melanoma metastasis has been looked at in vivo. If   We fully agree with the reviewer that the relationship between MITF and BRN2 is highly complex. Here 380 the goal of this article is not to study the relationship between MITF and BRN2, but it is to bring to the 381 light for the first time that PTEN plays a role in this complexity.

389
We have now moderated our conclusions when needed. All TCGA and microarray data studies are 390 correlative. In this respect, we cannot conclude until appropriate functional studies are performed.

396
In this study the authors argue that BRN2 is a haplo-insufficient tumor suppressor whose loss promotes 397 melanoma initiation and progression. The mechanistic basis for this is proposed to be BRN2 regulation of The BRN2 locus shows frequent copy number loss in cutaneous melanoma metastases, although these 406 deletions also affect many other genes implicated in melanomagenesis (Fig. 1A-B). Additional data 407 supporting that BRN2 is the key contributor to melanoma initiation and progression would significantly  Mono-allelic loss of BRN2 is associated with worse overall survival. Considering that central thesis is that 415 BRN2 is a tumor suppressor whose monoallelic loss promotes melanoma initiation and progression, 416 frequency of BRN2 loss (in addition to mRNA levels) at early/pre-metastatic stages of melanoma should 417 be included to further support the weak association BRN2 low vs. high mRNA levels with survival (Fig. 1G).

426
We performed the analysis of the publically available data published by Shain et al (2018) to evaluate the 427 number of BRN2 alleles in nevi and melanoma that arose from these nevi. Nineteen pairs 428 (nevus/melanoma) could be analyzed as shown in Figure S1D. It appears that 25% (5 out 19) of the 429 melanoma presented a mono-allelic loss of BRN2 compared to nevi. In this respect, we may conclude that 430 this BRN2 mono-allelic loss occurred during the early steps of melanomagenesis.

433
Mouse model indicates that mono-allelic loss of BRN2 in neonatal mice promotes melanomagenesis.

434
Recommend further experiments to support BRN2 tumor suppressive function.

437
Being on a pure C57BL/6 background (crucial), this mouse model demonstrates, and not only indicates,

449
Of course, to decipher the various cellular mechanisms occurring during melanomagenesis, it is indeed 450 crucial to perform in cellulo experiments with the associated limitations. Our study demonstrates that 451 Brn2 acts as a tumor-suppressor, this was not shown previously, and as pointed out above is not only 452 relevant to the genetics of melanoma, but also to conditions when BRN2 might be down-regulated by 453 conditions in that microenvironment that would decrease its promoter activity.

456
Does re-expression of BRN2 in cells that have lost BRN2 suppress (i) colony formation in vitro?

460
* Colony formation represents the ability of the cells to grow and to form a colony in vitro which is an 461 assay to evaluate some parts of the metastasis process. Colony formation is associated with melanoma 462 progression but not with melanoma initiation.

464
* In two different studies it was shown that the reduction of Brn2 levels has no effect on colony formation

486
The presence or absence of Brn2 did not decrease the ability of these melanoma cell lines to grow in 487 syngeneic mice (data not shown). In other words, it appears that the absence of Brn2 in these melanoma

503
2) The Braf-Pten melanoma models are widely used in the field and they are accepted to be one of the

515
The growth of the tumors is so rapid that we had to sacrifice the mice for ethical issues. Consequently a 516 full analysis of metastasis cannot be performed with these transgenic mice according to ethical rules.

519
What effects does bi-allelic loss of BRN2 in melanoma cells have on tumorigenicity?

522
We now included the data corresponding to melanoma initiation in Brn2 homozygous mice ( Figure 2). On

532
What is the impact of silencing the other genes in Chr6 (e.g. Arid1A, ROS)?

535
While we focused our study on Brn2, we regard this as an important question.

537
Similar in vivo experiments could be performed with Ros and/or Arid1a mice. Knowing that Arid1b may 538 interact with Brg1 (Smarca4), we may expect to have a positive output according to our results obtained 539 previously 9 . Arid1b mice homozygous for a null allele die perinatally. Arid1b Floxed mice (exon 5) were 540 generated 10 . The full knockout of Ros1 was performed by Carmen Birchmeier 11 . In the absence of Ros1, 541 mice are viable but males are not fertile, it is therefore possible to address this question with such mice.

542
To generate such combination (CreERt2/Braf/Pten/Arid1A or Ros) on a pure genetic background, produce 543 enough mice, follow the tumors, it would take over than 3 yrs. These experiments cannot reasonably be 544 expected from us in a reasonable time. Moreover, this article provides a large amount of novel 545 information on the function of BRN2.

547
Since the levels of Arid1b and Ros in mouse melanoma lacking or not Brn2 were not affected, the 548 phenotype arising as a consequence of loss of Brn2 and is not obviously link to Arid1B/ROS levels in the 549 mouse. This information was in the discussion (p17 of the previous ms). In addition, we showed that the 550 level of expression of Arid1b or Ros are not correlated with the OS in patients (see Figure S1)

553
Evidence from clinical samples of melanoma in situ/invasive melanoma vs. adjacent precursor nevi 554 suggests that BRN2 levels increase during transition to invasive melanoma (Zeng et al., 2018). Please 12 discuss this and address whether mono-allelic loss of BRN2 occurs early in melanoma development.

558
In the Zeng model, they observe that the loss of CDKN2A locus leads to the upregulation of CDK4/6 that 559 induces E2F1, which is able to induce BRN2 and promoting in their assay motility and invasion. This

585
* We initially evaluated the connection between BRN2 and PTEN in two cell lines. One was human (Dauv-586 1) and one is mouse (melan-a). It shows that this regulation is conserved during evolution; this is an 587 important point to stress. We tested our hypothesis on three novel human melanoma cell lines (Gerlach 588 SK28, and 501Mel) as requested by the reviewer. As requested, we also performed the knock down of 589 BRN2 in these three cell lines (Gerlach, SK28, and 501Mel), and confirmed that the reduction of BRN2

600
In the results and discussion, we are coming back on the high complexity of regulation of MITF by BRN2 601 that includes their mutual regulation. Moreover, we bring to the light AXL and its regulation. Indeed, AXL 602 is associated with melanoma metastasis. We have shown that Braf-Pten-Brn2-het cells produce more Axl 603 than Brn2-WT/hom (see Figure 4H). Moreover, when the levels of both Brn2 and Mitf are decreased in 604 human melanoma cell lines, the level of AXL is induced (see Figure S6D,H,L,P,T).

606
* Figure S8 on MITF is now a main figure (Figure 6)

617
The reviewer is right, we do not present any evidence that BRN2 inhibits MITF expression in this mouse 618 model. As previously showed in Figure 1

647
We showed that Brn2 can bind directly to the Pten promoter, and that Brn2 induces the Pten promoter 648 using a classical luciferase assay. We also show that independently Mitf represses Pten mRNA levels. See

657
MITF promotes proliferation, whereas low MITF is associated with an invasive, slow-cycling phenotype.

658
Could MITF also be contributing to the effects of Brn2 loss on proliferation and LN infiltration?

661
The information given by the referee "MITF promotes proliferation, whereas low MITF is associated with 662 an invasive, slow-cycling phenotype" is coming from some melanoma cells in culture and are accepted as

674
Since the overall conclusion of the study is that "BRN2 loss reduces PTEN transcription in vitro and in vivo,

689
Though undoubtedly a good experiment, our research ethics committee did not approve the use of 690 BRAFi+PI3Ki inhibitors in mice before the tumors appear. Therefore we could not assess their effects on

698
What is PTEN mutation status in cell lines used?

701
T1 and T2 are already invasive melanoma by definition (AJCC classification). We are unsure of the 702 meaning of "localized"; when it is limited to epidermis it is referred as in situ (Tis). However, there is a 703 conflict with the request from reviewer #1. In order to avoid any misrepresentation we removed the 704 former Figure S2B that is not essential for the main message of the article.

706
We included a table (Table S3) giving the genetic status for BRAF and NRAS, and the relative protein level 707 of BRN2 and PTEN of the melanoma cell lines used in this study.

710
Based on the phenotype elicited by loss of Brn2, the authors state that "Brn2 acts as a tumor suppressor 711 in vivo, and its loss induces melanoma initiation and increases tumor growth rate"; reword to more 712 precisely indicate that loss of Brn2 in the context of monoallelic PTEN…

715
We modified the sentence. Indeed, we tested in mice the role of Brn2 in a context Braf(V600E) and Pten 716 heterozygous and showed that Brn2 acts as a tumor suppressor.

719
"Cell lines express Pten and Brn2 mRNA…(data not shown)" -include the data in supplemental figures.

722
It is now included as Figure S6 and table S3.

725
"Cell number was monitored" aper BRN2 knockdown (Fig. 4D). Cell cycle analysis to assess relative 726 proportion of cells proliferating would further support the conclusions.

729
The number of BrdU+ cells is determined from the solid tumors ( Figure 3C,D), as Ki-67 ( Figure 3A,B). We

738
Only one tumor per mouse was considered corresponding to the biggest one.

739
Tumors had the same size since we harvested tumors for transcriptomic analyses when they reached a 740 size of 1 cm^3. This information is now included in the materials and methods.

743
Reduction in PTEN positivity upon BRN2 loss (mouse tumor IHC - Fig. 6A) is not reflected in Western blot 744 levels (Fig. 6B). PTEN protein levels look comparable between WT and BRN2 when normalized to actin

753
Also probing for MITF is suggested.

756
Mitf is expressed at a very low level in advanced Braf/Pten tumors (see Laurette et al, 2020 - Figure S3).

757
Unfortunately, we could not detect MITF neither by western blotting nor by IHC in these mouse tumors.

759 760
The authors have made significant changes to their original manuscript in the form of additional clarity of interpretations, transparency of data, and substantial additional experiments and analyses. In doing so, they have substantially strengthened the manuscript and convincingly support each of their main conclusions. I congratulate the authors on the excellent study which I believe both provides both novel and interesting insight to the field as well as a useful new model system and datasets. A few minor edits are requested below.
Minor requests. 1. I still fundamentally disagree with the author's continued choice to omit the gain/amplification category in Fig 1A. It is unethical to selectively omit this category visually from the figure, presumably because it does not cleanly fit with the hypothesis. It is good the authors mention the category in the text, but still, there is no reason to not also add the column visually into the figure for those readers who might internalize the main figures, but miss the sentence in text, and therefore inadvertently misinterpret the data. As the authors allude to in their introduction and discussion, studies of the role of BRN2 in melanoma are wrought with often conflicting observations and confusion, in large part to the practice of "hiding" conflicting data in the supplements, text mention only, or (data not shown). The present study is thorough, excellent and adds substantial clarity to the field due to the careful and nuanced approaches -this reviewer urges the authors to not soil such excellent work through data hiding. Being transparent here will not detract from the overall impact of the study and will, instead, help the field to move past clearly oversimplified binary / linear models and appreciate the nuances and complexities of tumor initiation. Fig S1D does not appear to support the conclusion the authors draw from it as stated in text lines 105-106. The authors present 5 cases where BRN2 was lost in melanoma compared to nevi and 4 cases where it was gained. Thus the statement "It appears that ~30% (5 out 18) of melanomas presented a mono-allelic loss of BRN2 compared to nevi….we may conclude that this BRN2 mono-allelic loss may occur during the early steps of melanomagenesis" seems misleading when 4 out of those 18 cases had a gain. The analysis is useful and should be kept in the manuscript, but if the 5/18 is highlighted, so too should be the 4/18 for transparency, and the authors should discuss and/or tone down the conclusion.

Similarly, the analysis presented in
3. Text line 307 seems to be incomplete: "these results corroborate the immunohistological studies of these tumors. Indeed, ..... (In process)." Overall, an excellent and important study.

Reviewer #2 (Remarks to the Author):
This is a revised manuscript; the study offers some new insight into the role of Brn2 for melanoma initiation and progression. It would be helpful to reorganize and edited the Manuscript to more accurately describe the data and conclusions.
For example, Fig S1N shows that survival is associated BRN2 loss in the context of monoallelic loss of PTEN without providing a rationale for the reader. Perhaps it would be better to present the data in S1N following the paragraph describing 'Co-occurrence of BRN2 loss with mono-allelic loss of PTEN (line 133)' Figs. 1C, E: TCGA data retrieved August 2019 should be updated.
It would be interesting to follow-up on several of the gene ontology enrichment process identified by validating them in the mouse models e.g. inflammation, angiogenesis. The data showing BRN2 is able to directly regulate PTEN transcription are reasonable and novel. The authors also propose that BRN2 is regulating PTEN indirectly via MITF, however they still do not present evidence for BRN2 modulating MITF. Instead, they show that MITF can regulate PTEN (Fig. 6), which is not central to the main conclusions, unless a link between BRN2 and MITF is shown in their models and perform epistatic experiments, for example by evaluating if MITF overexpression could rescue PTEN expression in the context of Brn2 loss.
Other/minor comments -Clear conclusions summarizing results for each set of experiments will be helpful. -1C, E: TCGA data retrieved August 2019 should be updated.
-Analysis of data from precursor nevi/melanoma pairs indicates that mono-allelic loss of BRN2 can occur early in melanomagenesis (5 of 18 samples), as discussed in the results. However, gain of BRN2 occurred in 4 of 18 samples. Please also specify this in the main text.
-If the authors wish to conclude that 'BRN2 promotes the bypass of senescence (by reducing Pten levels)…', this could be experimentally tested by evaluating the % of senescent cells in PTEN+/-+ BRN2 WT, BRN2+/, BRN2-/-. -Several conclusions are stated based on data not shown. If conclusions are important for the overall findings of the study and data is available, it should be included. For example, "The presence or absence of Brn2 did not decrease the ability of these melanoma cell lines to grow in syngeneic mice (data not shown). Based on the data not shown, the authors conclude that "the absence of Brn2 in these melanoma cells does not affect the implantation of the cells on the body wall, the proliferation after their transformation or the induction of angiogenesis in an immunocompetent environment." -Line 133: first the authors state that: "There was no significant correlation between BRAF or NRAS mutation and BRN2 loss (mono-or bi-allelic), neither in human melanoma samples nor the human cell-line panel ( Figure S2A,B) ". Then a few lines down (143) they state "found the most frequent genetic constellation that co-occurs with BRN2 loss in melanoma to be BRAFV600X mutation and mono-allelic PTEN loss ( Figure S2E)". Please review and edit as appropriate -Hyperactivation of PI3K/AKT was identified via GSEA; the authors then argue that this was validated based on substantial increased pAKT-S473 and pS6 -S235/236. However, this does not seem to be the cased in S4K, M and O where levels of basal pAkt are fairly similar among all genotypes. Please explain.
-Induction of PI3K/AKT should be validated in tumor samples from the different mouse models -MEK/BRAF inhibitors seem to downregulate BRN2. Brn2 hom loss seems to sensitize cells to PLX treatment, please discuss. Also, It is odd that MAPK + PI3K do not cooperate in the context of Brn2 loss. -The finding that Brn2 het exhibit Low MITF/High Axl seems to contradict the findings that those cells are more sensitive to MAPKi compared to WT, as the MITFlow/AXl high generally is associated with resistance to BRAF/MEK. -While the AXL mRNA levels are statistically significantly increased, the magnitude of increase is rather small, except in 501Mel. Additionally, mRNA levels may not necessarily reflect RTK activation. Authors should more accurately state their findings - Fig This reviewer was asked to specifically comment on CUT&RUN part of the manuscript.

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): The authors have made significant changes to their original manuscript in the form of additional clarity of interpretations, transparency of data, and substantial additional experiments and analyses. In doing so, they have substantially strengthened the manuscript and convincingly support each of their main conclusions. I congratulate the authors on the excellent study which I believe both provides both novel and interesting insight to the field as well as a useful new model system and datasets.
We thank the reviewer for his/her very positive comments and for the constructive criticism that indeed led to a much-improved manuscript.
A few minor edits are requested below.
Minor requests. 1. I still fundamentally disagree with the author's continued choice to omit the gain/amplification category in Fig 1A. It is unethical to selectively omit this category visually from the figure, presumably because it does not cleanly fit with the hypothesis. It is good the authors mention the category in the text, but still, there is no reason to not also add the column visually into the figure for those readers who might internalize the main figures, but miss the sentence in text, and therefore inadvertently misinterpret the data. As the authors allude to in their introduction and discussion, studies of the role of BRN2 in melanoma are wrought with often conflicting observations and confusion, in large part to the practice of "hiding" conflicting data in the supplements, text mention only, or (data not shown). The present study is thorough, excellent and adds substantial clarity to the field due to the careful and nuanced approaches -this reviewer urges the authors to not soil such excellent work through data hiding. Being transparent here will not detract from the overall impact of the study and will, instead, help the field to move past clearly oversimplified binary / linear models and appreciate the nuances and complexities of tumor initiation.
Reviewer #2 (Remarks to the Author): This is a revised manuscript; the study offers some new insight into the role of Brn2 for melanoma initiation and progression. Answer 1. We thank the Reviewer for recognizing the new insights offered by this manuscript. Indeed, this study reveals the role of BRN2 during melanoma initiation in vivo, which had not been demonstrated previously, and it adds some new insight into the role of BRN2 for melanoma progression.
It would be helpful to reorganize and edited the Manuscript to more accurately describe the data and conclusions. For example, Fig S1N shows that survival is associated BRN2 loss in the context of monoallelic loss of PTEN without providing a rationale for the reader. Perhaps it would be better to present the data in S1N following the paragraph describing 'Co-occurrence of BRN2 loss with monoallelic loss of PTEN (line 133). Answer 3. In the first submission, we presented the recovered TCGA data in May 2017. In the second submission, we submitted recovered TCGA data in August 2019. Differences between May 2017 and August 2019 were negligible. The differences between August 2019 and February 2021 will undoubtedly be less significant than in the previous update. Indeed, the number of updates performed recently is very small because the vast majority of patients have already died. Therefore, this modification does not seem to be essential, unless the editor wishes us to do so.
It would be interesting to follow-up on several of the gene ontology enrichment process identified by validating them in the mouse models e.g. inflammation, angiogenesis. Answer 4. This study presented in a supplementary figure (Fig S5) is of course of interest, but this information would not dramatically reinforce the main conclusions of this article, and we feel it would be best left for a follow up manuscript. The data showing BRN2 is able to directly regulate PTEN transcription are reasonable and novel. The authors also propose that BRN2 is regulating PTEN indirectly via MITF, however they still do not present evidence for BRN2 modulating MITF. Instead, they show that MITF can regulate PTEN (Fig.  6), which is not central to the main conclusions, unless a link between BRN2 and MITF is shown in their models and perform epistatic experiments, for example by evaluating if MITF overexpression could rescue PTEN expression in the context of Brn2 loss.
Answer 5. It is already well established that BRN2 can either positively or negatively regulate transcription of MITF (see for example: 1,2 ), a key protein of the melanocyte lineage (see introduction). As already mentioned, the role of BRN2 in melanomagenesis has been controversial. Our aim was to show that both BRN2 and MITF can regulate PTEN. Does the activation of PTEN by BRN2 occur at the same time as MITF repressing PTEN? At this point, it is difficult to answer this question, since BRN2 may activate or repress MITF. A better understanding of the regulation of MITF by BRN2 is needed, and it will be necessary to continue to study this specific question, which has been unresolved since 2008, when Brn2 was found to activate or repress the transcription of MITF; likely both scenarios can occur depending on specific conditions or the presence of unidentified co-factors. Here, the indirect regulation of Pten by Brn2 via Mitf is included as a possibility in the Discussion. To address this question, one option would be to use a genetic approach to delete specifically the binding sites of BRN2 or/and MITF on the PTEN locus, or alternatively to modify the activity/amount of BRN2 and/or MITF and evaluate the resulting levels of PTEN mRNA. This cannot be included in this article, as these experiments will require the generation of novel mouse models in keeping with the in vivo approach used throughout the manuscript. However, we recognize the Reviewer's point and consequently toned down the possible indirect mRNA regulation of PTEN by BRN2 via MITF. Please see line 468.
Other/minor comments -Clear conclusions summarizing results for each set of experiments will be helpful.
-Analysis of data from precursor nevi/melanoma pairs indicates that mono-allelic loss of BRN2 can occur early in melanomagenesis (5 of 18 samples), as discussed in the results. However, gain of BRN2 occurred in 4 of 18 samples. Please also specify this in the main text.
Answer 7. We now have included a mention of gains of BRN2 in the text, as suggested by the Reviewer (see also reply to Reviewer 1). Please see lines 105.
-If the authors wish to conclude that 'BRN2 promotes the bypass of senescence (by reducing Pten levels)…', this could be experimentally tested by evaluating the % of senescent cells in PTEN+/-+ BRN2 WT, BRN2+/, BRN2-/-. Answer 8. It is known that melanoma initiation is the result of two main cellular processes: initial proliferation and bypass/escape of senescence as mentioned in the introduction. Here, we show that melanoma initiation is induced since the number of individual melanomas is higher when BRN2 is absent (Brn2 F/F) or decreased (Brn2 F/+). In other words, low/null BRN2 induces proliferation, but it is not sufficient to promote melanoma initiation since the production of Braf V600E induces a wellcharacterized OIS. Moreover, we had mentioned in the text "…BRN2 may act in vivo as an MITF repressor, we have observed that in a non-tumoral context, the specific knock-out of Brn2 in vivo in melanocytes increases the level of Mitf (publication in preparation). In addition, it has been shown that the decrease/loss of PTEN promotes senescence bypass (see for instance Conde-Perez et al., 2015). The lack of BRN2 would contribute to the bypass of senescence by decreasing the levels of PTEN. These physiological and molecular arguments support the hypothesis that the loss/decrease of BRN2 favours the bypass of senescence. We toned down this issue in different sentences. Please see lines 198, 496.
-Several conclusions are stated based on data not shown. If conclusions are important for the overall findings of the study and data is available, it should be included. For example, "The presence or absence of Brn2 did not decrease the ability of these melanoma cell lines to grow in syngeneic mice (data not shown). Based on the data not shown, the authors conclude that "the absence of Brn2 in these melanoma cells does not affect the implantation of the cells on the body wall, the proliferation after their transformation or the induction of angiogenesis in an immunocompetent environment." Answer 9. Following the Reviewer's recommendation, we now present the results associated with the growth of mouse melanoma cells, mutant or not for Brn2, in syngeneic mice. We included these results as a supplementary table (Table S4) in the text, as follows: "The presence or absence of Brn2 did not decrease the ability of these melanoma cell lines to grow in syngeneic mice (Table S4)." Please see line 235. For this experiment, one hundred thousand cells of each line were subcutaneously injected in C57BL/6 mice. All mice presented tumors from the six tested cell lines (two cell lines per genotype -Brn2 +/+, Brn2 F/+, and Brn2 F/F). This information is included in the Material and methods section of the supplemental information and in the table note.
As to the sentence: "the absence of Brn2 in these melanoma cells does not affect the implantation of the cells on the body wall, the proliferation after their transformation or the induction of angiogenesis in an immunocompetent environment", we included it to simply remind the reader of the significance of cell growth after subcutaneous injection in mice. We modified the text accordingly. -Line 133: first the authors state that: "There was no significant correlation between BRAF or NRAS mutation and BRN2 loss (mono-or bi-allelic), neither in human melanoma samples nor the human cell-line panel ( Figure S2A,B) ". Then a few lines down (143) they state "found the most frequent genetic constellation that co-occurs with BRN2 loss in melanoma to be BRAFV600X mutation and mono-allelic PTEN loss ( Figure S2E)". Please review and edit as appropriate Answer 10. The reviewer refers to the supplementary Figure S2. I am not sure I understand the referee's comment. In the first sentence, we do not include the status of PTEN ( Figure S2A,B) and in the second sentence we include the status of PTEN ( Figure S2E). More precisely, with the full paragraph in italic "There was no significant correlation between BRAF or NRAS mutation and BRN2 loss (mono-or bi-allelic), neither in human melanoma samples nor the human cell-line panel ( Figure S2A,B)." This corresponds to the first cited sentence. Then, the reviewer omits to refer to "We then searched for co-occurring CNAs of other known melanoma-associated genes and found that mono-allelic loss of BRN2 co-occurred with mono-allelic loss of PTEN in approximately 40% of the human melanoma samples in TCGA and in the human cell-line panel ( Figure S2C,D). We next evaluated the concomitant BRN2 locus alterations, BRAF/NRAS mutations, and CDKN2A/PTEN alterations and ….". then he refers to "found the most frequent genetic constellation that co-occurs with BRN2 loss in melanoma to be BRAF V600X mutation and mono-allelic PTEN loss ( Figure S2E)." Taking into account the full context of these sentences, we believe that there is no formal opposition between these statements. To avoid any confusion we modified the text "BRAF V600X mutation and mono-allelic PTEN loss" to BRAF V600X mutation together with mono-allelic PTEN loss. Please see line 141.
-Hyperactivation of PI3K/AKT was identified via GSEA; the authors then argue that this was validated based on substantial increased pAKT-S473 and pS6 -S235/236. However, this does not seem to be the cased in S4K, M and O where levels of basal pAkt are fairly similar among all genotypes. Please explain.
Answer 11. The referee is right, it is not the case in Figure S4K,M,O: these panels refer to cells grown in 2D-culture in the presence of 10% FCS that are selected for efficient proliferation in vitro, a situation quite different from in vivo settings. We added a comment in the corresponding legend in the supplemental information.
-Induction of PI3K/AKT should be validated in tumor samples from the different mouse models Answer 12. We already showed that the level of Pten was reduced in Braf-Pten-Brn2-het tumors compared with Braf-Pten-Brn2-WT tumors ( Figure S5D). To address the reviewer comments, we presented two sets of data. In Figure S5D, we now present the western blots of WT, het and homozygous tumours for Akt, pAkt, S6, and pS6. Moreover, we evaluated the percentage of cells expressing Pten in Braf-Pten-Brn2-WT compared to Braf-Pten-Brn2-het tumors ( Figure 5A). We now included the percentage of cells expressing Pten in Braf-Pten-Brn2-hom tumors. It appears that the percentage of cells producing Pten is higher in Braf-Pten-Brn2-WT melanoma than in Braf-Pten-Brn2het or hom melanoma. From these results, we may conclude that the number of cells in which the PI3K/AKT signalling pathway is activated is increased in Braf-Pten-Brn2-het or hom melanoma compared to Braf-Pten-Brn2-WT melanoma. We understand that we indirectly answer the reviewer but are convinced that it is sufficient at this point. Please see line 335.
-MEK/BRAF inhibitors seem to downregulate BRN2. Brn2 hom loss seems to sensitize cells to PLX treatment, please discuss. Also, It is odd that MAPK + PI3K do not cooperate in the context of Brn2 loss.
Answer 13. It has been shown that BRAF and PI3K induce BRN2 3,4 . In consequence, it is expected that the level of BRN2 is decreased in the presence of such inhibitors. In addition, in the absence of BRN2, melanoma cells are indeed more sensitive to PLX treatment (Herbert et al, 2019). In addition, it has been recently reported (Pierce et al., 2020) that human "Melanoma cells expressing higher levels of BRN2 generally showed higher IC50 values compared with those lines with low or no BRN2 expression." This lower sensitivity with higher BRN2 levels is coherent with the function of BRN2 in DNA repair (Herbert et al., 2019). We added this information in the Discussion. Please see line 563.
We were indeed expecting that MAPK and PI3K would cooperate, but it is not the case in our model system. Deciphering this question is of importance but, not being the main focus of this article, it is best left to a follow-up study.
-The finding that Brn2 het exhibit Low MITF/High Axl seems to contradict the findings that those cells are more sensitive to MAPKi compared to WT, as the MITFlow/AXl high generally is associated with resistance to BRAF/MEK. While the AXL mRNA levels are statistically significantly increased, the magnitude of increase is rather small, except in 501Mel. Additionally, mRNA levels may not necessarily reflect RTK activation. Authors should more accurately state their findings For these articles, the experimental settings are very different from ours. They generate in vivo or in cellulo resistant cells to BRAF/MEK inhibitors and evaluate the level of RNA/proteins including MITF and AXL in sensitive and resistant cells. Indeed, it appears that MITF-low/AXL-high cells are more resistant than normal cells. A large set of gene expression modifications occurred during the selection process, including, e.g., changes in the expression of EGFR or PDGFRb that may affect sensitivity to MAPKi. However, as far as we know, the status of BRN2 was not examined in these studies.
In our case the experimental situation is completely different. Besides the mutations of the Braf and Pten loci, a single mutation was differentially generated at the level of the BRN2 locus, and the cells were not randomly selected for drug resistance, they were naïve. Moreover, the level of Mitf is always low in all Braf-Pten melanoma cells. These key differences make any comparison difficult/impossible.
That said, the referee is right, there is a slight increase of Axl mRNA, but we do not evaluate the activity/amount of Axl protein in these mouse tumors or cell lines since it was not the main focus of this article. We now make the point in the text that the level of AXL is clearly induced in melanoma cell lines in which the level of Axl was originally low (501Mel and SK28). Moreover, we included in the discussion the following sentence "In the future, we will have to evaluate the kinase activity of Axl in this context, and the consequences of its inhibition in cellulo and in vivo to understand the Mitf/Brn2/Axl ménage à trois." It is now included on line 505.
In general, those are good points, but are not the main focus of the article.
Answer 15. There were no cases in which CCNC gene-expression was below the 1 TPM threshold we had defined to represent absence of expression. As such this panel is not informative, but we included it for the sake of completion. We made a point of this issue in the figure legend -see supplemental information.
-It would be better to use consistent nomenclature for all genes; e.g. BRN2 in Answer 17. The Brn2 deletion is indeed inducible. Tamoxifen has to be added to induce the translocation of CreERt2 to the nucleus to generate the mutation.