Blockade of the pro‐fibrotic reaction mediated by the miR‐143/‐145 cluster enhances the responses to targeted therapy in melanoma

  1. Serena Diazzi
  2. Alberto Baeri
  3. Julien Fassy
  4. Margaux Lecacheur
  5. Oskar Marin‐Bejar
  6. Christophe A Girard
  7. Lauren Lefevre
  8. Caroline Lacoux
  9. Marie Irondelle
  10. Carine Mounier
  11. Marin Truchi
  12. Marie Couralet
  13. Mickael Ohanna
  14. Alexandrine Carminati
  15. Ilona Berestjuk
  16. Frederic Larbret
  17. David Gilot
  18. Georges Vassaux
  19. Jean‐Christophe Marine
  20. Marcel Deckert
  21. Bernard Mari
  22. Sophie Tartare‐Deckert

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

No abstract available

Article activity feed

  1. Version published to 10.15252/emmm.202115295
  2. Review Commons

    Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer #1 (Evidence, reproducibility and clarity):

    The manuscript is interesting and well presented. The authors propose the use of an antifibrotic drug to attenuate resistance to RTK inhibitors.

    **Specific comments***

    • It is not entirely clear how Nintedanib decreases tumour growth. It may be due to its effect on resistant melanoma cells as proposed, but it could also be due to the effect on CAFs. This should be at least discussed. *

    The reviewer asks about a potential effect of Nintedanib on CAFs in our mouse model. While we show that Nintedanib has a direct action on melanoma cells in vitro, the in vivo situation can indeed be more complex. We agree that we cannot rule out the possibility that its therapeutic efficacy could be attributed in part to inhibition of CAFs, knowing that BRAF inhibitors has been shown to activate CAFs in melanoma, generating a host-tumor niche that can mediate therapeutic escape. However, addressing the contribution of CAF in vivo is challenging and would represent an entire new study. As requested by the reviewer, we have discussed this important issue and added 3 new references (see discussion section lines 377-381).

    • A potential caveat is that drug used is non-specific as it also blocks PDGFR signalling. Hyperactivation of RTKs is a mechanism of BRAFi resistance and for example in Figure 1J, they see that BIF1120/Nintedanib has a significant effect on BRAFi-resistant cells, which may indicate that the growth inhibition seen in allografts could be a combination of an "anti-fibrotic" role and its own activity inhibiting the survival of resistant cells. This needs to be considered.*

    We thank the reviewer for this interesting issue. Nintedanib was chosen due to its inhibitory action on extracellular matrix deposition and as an example of a rapidly available drug to be exploited therapeutically to increase the effect of targeted therapy and delay the emergence of therapy-resistant cells. We recognize that a possible disadvantage of Nintedanib could be due to its multi-targeted nature (e.g. PDGFR (α and β), FGFR-1, -2, -3, -4 and VEGFR-1, -2, -3 as well as Src, Lck or Lyn) but it is one of the only approved molecules for the treatment of fibroproliferative diseases. Upregulation of PDGFRβ/AKT signaling was previously shown to contribute to acquired resistance in M238R (Shi et al. Cancer Res. 2011;71:5067-74 ; Nazarian et al. Nature. 2010;468:973-7). Our in vitro results indicate that Nintedanib inhibits survival of these resistant cells along with a decrease in their myofibroblast-like dedifferentiated phenotype (Fig. 1 I-J).

    To meet the reviewer’s comment, we have now addressed the contribution of PDGFRβ inhibition in Nintedanib’s effects on resistant cells. We have performed experiments on M238R using the selective PDGFR inhibitor CP673451 in comparison with Nintedanib (please see results section lines 120-127 and new Supplementary Fig. S1F-H). The data show that selective inhibition of the PDGFR pathway attenuates the myofibroblast-like signature typical of resistant cells to a similar degree as Nintedanib and affects melanoma cell viability (new Supplementary Fig. S1G-H). However, administration of CP673451 showed less efficiency than Nintedanib in inducing a phenotype switch toward a more differentiated phenotype (new Supplementary Fig. S1G). To further confirm the implication of RTK pathway in the phenotype observed, we analyzed the tyrosine phosphorylation status of EGFR, PDGFR and FGFR (another RTK inhibited by Nintedanib) and activation of AKT in M238R melanoma cells upon treatment with Nintedanib or CP673451 (new Supplementary Fig. S1F and additional results for the reviewers). Nintedanib had no effect on FGFR tyrosine phosphorylation and slightly decreased pEGFR levels. However, we found that the two inhibitors showed similar efficiency in decreasing phospho-PDGFRβ and phospho-AKT levels (Supplementary Fig. S1F). The results section has been modified according to these new results (lines 126-127).

    Altogether these data suggest that inhibition of PDGFR signaling likely plays a prominent role in the efficacy of Nintedanib in vitro on M238R survival. Thus, as proposed by the reviewer, we can predict that the growth inhibition induced by Nintedanib seen in vivo could be a combination of its "anti-fibrotic" action and PDGFR inhibitory activity inhibiting the survival of resistant cells. It is important to note that, compared to Nintedanib, inhibition of PDGFR/AKT signaling by the CP673451 compound is not sufficient to direct melanoma cells to a more differentiated state. This is now discussed in the manuscript (Discussion section lines 404-405).

    • Does the viability decrease in BRAFi-sensitive cells? For instance, in the parental cells?*

    This information was already addressed in the manuscript. As shown in Supplemental Fig. S1D, Nintedanib had no effect on BRAFi-sensitive M238P viability. We have also confirmed this result using a crystal violet viability assay on M238P and UACC62 cells treated with different doses of BIBF1120.

    • Figure 1 b-e, in vivo and in vivo experiments. *How many animals were used? Collagen decrease is not quantified (statistics missing).

    We apologize for this omission and have now added the number of animals in the legend of Fig.1 (n = 6). We have also performed statistics for collagen quantification and included this analysis in Fig.1F (see lines 720/723). We also provide to the referee the detailed statistical analysis of mature collagen fibers between the different treatment groups.

    • The title is not accurate. "prevent" resistance in melanoma is an overestimation because the cells do become resistant, albeit later.*

    We agree with the reviewer and we have modified the title accordingly. The new title is now: “Blockade of pro-fibrotic response mediated by the miR-143/-145 cluster prevents targeted therapy-induced phenotypic plasticity and delays resistance in melanoma”.

    Reviewer #1 (Significance):

    As the authors discussed, they and others have previously studied the contribution of ECM and stromal remodelling to resistance to targeted therapies in melanoma. Previous data from E. Sahai´s lab show that BRAFi activate CAFs and increase the production and remodelling of the extracellular matrix, but in this work, they look at a cell-autonomous mechanism mediated by miRs that promotes fibrosis and propose the use of an antifibrotic drug to attenuate resistance to RTK inhibitors.

    __Reviewer #2 (Evidence, reproducibility and clarity (Required)): __ In this very interesting study, Diazzi and colleagues show that during adaptation to MAPK-targeted therapy (MAPKi), melanoma cells upregulate a miRNA profibrotic cluster (miR-143, -145), which drives a phenotypic switch towards a drug resistant undifferentiated mesenchymal-like state. From the miRNA targets, authors identify FSCN1 as a gene that needs to be downregulated during adaptation to MAPKi by the miRNAs, since FSCN1 ablation promotes the drug resistant phenotype. Importantly, authors show in a preclinical mouse melanoma model that the anti-fibrotic drug nintedanib (BIBF) improves response to MAPKi and delays onset of resistance.

    The study conclusions are convincing and the data are adequately replicated and presented, authors should be commended for having the manuscript in such good shape. However, there are a few issues that authors should clarify/expand.*

    We sincerely thank the reviewer for his/her careful review and constructive comments.

    The study starts with the in vivo YUMM1.7 model and combination BRAFi+MEKi, and then authors use this combination in many in vitro experiments. However, when studying resistant lines, only BRAFi-resistant and -sensitive pairs were used. I would suggest including more validation of the upregulation of the miRNA and the fibrotic genes on BRAFi+MEKi-resistant lines, and this could be easily gathered from published transcriptomes of several BRAFi+MEKi-resistant melanoma lines from Roger Lo's lab (Song et al 2017 Cancer Discov, including M238, M229, M249 used by the authors). To complement this approach, miRNA expression could be evaluated in large collections of melanoma cell lines classified as more or less undifferentiated (correlating with more or less resistance) as in Tsoi 2018 Cancer Cell and Verfaille 2015 Nat Commun.

    We thank the reviewer for these interesting suggestions. We have performed several analyses, summarized below:

    • First, we have analyzed the expression of the miRNA-143/-145 cluster and pro-fibrotic signature by qPCR in A375 parental and BRAFi/MEKi double resistant melanoma cell lines described in Shen et al. Nat Commun. 2019;10:5713. We observed the upregulation of both mature miRNAs along with a pro-fibrotic signature in several A375 DR clones compared to parental cells. This new result is described in the results section (lines 147-150) and shown in new Supplementary Fig. S2B. In addition, we have included in the results section the important information that the undifferentiated/mesenchymal-like BRAFi-resistant M229R and M238R cells used in our work also displayed cross-resistance to MEKi (results section, line 112 and 1 new reference).

    • Second, as recommended, we have also fully (re)analyzed the mentioned studies and associated datasets. We provide a summary of the different studies including samples number, design of the study, platform used and accession.

    A general observation is that unfortunately, none of these published studies provided an available small RNA-seq dataset, which thus does not allow quantifying the expression levels of mature miRNAs. However, some interesting observations have been uncovered from these datasets, confirming at least in part some of our data:

    i) The dataset from Song et al. 2017 compared 18 isogenic parental versus resistant cell lines. Two subsets of resistant cells were identified, with MAPK addiction (Ra) or Resistance with MAPK redundancy (Rr). The expression of the pri-miR-143/145 precursor, named MIR143HG, was detected in these cells and was found significantly upregulated in Rr cell lines compared to parental cells. Of note, MIR143HG was also part of the Rr specific signature associated with a mesenchymal phenotype. This interesting observation is now discussed in the manuscript (Discussion section, lines 392-394).

    ii) The dataset from Tsoi et al. 2018 focused on transcriptome analysis of 53 human melanoma cell lines including paired acquired resistance sublines established from patient biopsies. Unfortunately, MIR143HG expression is not detected in this dataset, probably due to a limited sequencing depth. Interestingly, we found that FSCN1 expression was decreased in most mesenchymal-like resistant cell lines compared to their parental counterpart. These data cannot be added in the manuscript since we cannot correlate the expression of the miRNAs with their target.

    iii) The dataset from Verfaillie et al. 2015 revealed transcriptomic analyses on 11 short-term cultures derived from patient biopsies before therapy and gave access to RNA-seq data of tumors with a proliferative or an invasive phenotype. MIR143HG is not detected and FSCN1 expression does not appear to be associated with a specific phenotype. We have performed qPCR-based expression of miR-143-3p and miR-145-5p in some of these short-term cultures, confirming that miR-143/-145 expression is not associated with a specific phenotype in therapy naïve melanoma cells (results for referees, see below). Expression of miR-143-3p and miR-145-5p in each short-term culture was compared to the average expression of the analyzed miRNA in the proliferative short-term cultures. These results are consistent with the findings of our study describing that expression of the miR-143/145 cluster is triggered by the inhibition of the BRAF oncogenic pathway.

    Related to this, the clinical relevance would increase if findings were validated using patient samples, for example, from published transcriptomes (Hugo 2015 Cell, Song 2017 Cancer Discov, Wagle 2014 Cancer Discov...) or even from TCGA, which could be used to identify if patients with high miRNA have worse prognosis.

    We agree with the reviewer about the importance of providing clinical data supporting our observations. We have carefully analyzed all these profiling studies and provide below a summary.

    Overall, these studies have several limitations: i) as underlined above, expression of the miRNA cluster is specifically induced in response to therapy and is not present (or barely) in tumors at diagnosis; ii) no small RNA-seq datasets are available yet; iii) melanoma tumors are highly heterogeneous and invaded with stroma, especially CAFs and vessels that also express these miRNAs. We have looked at the expression of the MIR143HG precursor in these datasets and it was not present, probably due to low to medium sequencing depths in these clinical studies.

    We have also carefully explored TCGA datasets to look at possible association between prognosis and mature / precursor miRNA as well as miRNA target (FSCN1) expression in skin cutaneous melanoma (SKCM) using the tools developed by Anaya et al. 2016, PeerJ Computer Science 2:e67. Cox regression models and Kaplan-Meier analysis (using different percentiles) did not show any association of our candidates with survival on a cohort of 459 SKCM patients (median survival of 2.4 years).

    Finally, during the revision process, we could have access to 9 relapsed melanoma for research purposes from the Dermatology Department of Nice University Hospital (CHU) following treatment with targeted therapies, immunotherapies or a combination of them. We have analyzed in these biopsies the expression of fibrotic/mesenchymal genes, FSCN1 and the miR-143/145 cluster compared to the mean expression of the same genes/miRNAs in therapy naïve patient-derived xenografts (MEL003, MEL006, MEL015, MEL047). Our first results indicate that relapsed tumors acquire a strong fibrotic signature which is associated to increased expression of the miR-143/-145 cluster and decreased expression of FSCN1 (8 out of 9 patients).

    These results are encouraging and represent a good indicator for further clinical validation but are not solid enough to be incorporated in the manuscript. Overall, validation of our hypotheses in patient samples would require an entire new and highly complex clinical study comparing tumors at diagnosis with relapsed tumors after targeted therapies and ideally processed using single-cell RNA-seq and/or RNA FISH to take into account the stromal compartment.

    • While blocking the miRNA improves BRAFi response (Fig.3H), it is not clear that this combination would overcome resistance (using resistant lines), although authors show that BIBF does overcome resistance (Fig.1J). *This also applies to line 277 "… mirroring the effect of miR143/145 ASOs, forced expression of FSCN1 in M238R cells decreased viability in the presence of BRAFi (Fig.5H)." However, the miRNA ASOs were used in parental cells (Fig.3H).

    To meet the reviewer’s comment, we have conducted new experiments in resistant melanoma cells using different approaches to silence simultaneously the 2 mature miRNAs: i) an ASO-directed RNAse H degradation of the miR-143/145 precursor, as described by Plaisance et al., JACC Basic Transl Sci. 2016, 1:472-493 to knock-down the pri-miRNA in cardiomyocytes, and ii) a combination of the 2 anti-miRs ASOs. Unfortunately, the first approach failed to efficiently inhibit the expression of mature miR-143-3p and miR-145-5, suggesting that the miR-143/145 cluster has a different precursor gene in melanoma than the one described in cardiomyocytes.

    Concerning the second approach, as expected, the 2 anti-miRs ASOs as well as the combination of the 2 ASOs efficiently targeted the mature miRNAs (new Supplementary Fig.S6C). Inhibition of miR-145-5p alone and combined inhibition of the two miRNAs significantly affected the viability of BRAFi resistant melanoma cells (M238R) in the absence of BRAFi (new Supplementary Fig.S6D) in a similar way as Nintedanib/BIBF (Fig. 1J).

    • Analysis of cytoskeletal changes. Text (lines 284-287) is missing references, regarding "…morphological changes with cells assuming flattened spindle-like shape" and "..function of FSCN1 in F-actin microfilaments reorganization...".*

    We apologize for these omissions and have added the relevant references in the text (lines 305/306).

    Besides, authors say that transient overexpression of miRNAs reproduced these morphological changes as shown by F-actin staining. These would have benefited from including also side-by-side comparison of BRAFi treatment on these cell lines. To my knowledge, these melanoma lines (M238, M229, etc) have not been characterized in that regard (F-actin, focal adhesions). In Nazarian et al 2010, only brightfield pictures are shown in a supplementary figure.

    The same applies to YAP and especially MRTF activation upon miRNA overexpression, and whether this mirrors what BRAFi does to YAP and MRTF. In Misek et al 2020 and Kim et al 2015 YAP and MRTF were shown to be more enriched in the nucleus in resistant than in parental cells. Kim et al also show in time course experiments that there is significantly higher nuclear YAP after 7-14 days of BRAFi treatment. In the present manuscript, authors seemed to have assessed nuclear YAP/MRTF after 72h miRNA overexpression. Does it mirror MAPKi?

    As suggested by the reviewer, we have compared side-by-side the effect of oncogenic MAPK pathway inhibition to the effect of miR-143 or miR-145 overexpression on cytoskeleton and focal adhesion dynamics as well as YAP and MRTFA nuclear translocation in M238P, M229P and UACC62P melanoma cells. These analyses clearly show that transient overexpression of miR-143-3p or miR-145-5p mirrors the effects of BRAF or BRAF/MEK inhibition after 3 days on mechanopathways and acto-myosin remodeling. We thank the referee for this comment, which is helpful for the interpretation of the data. The new additional panels have been included in new Fig. 6B-D, new Fig. 7B-D, new Supplementary Fig. S10B-D and new Supplementary Fig. S11C-D.

    Regarding the decreased proliferation/survival after miRNA overexpression, is it truly slow cycling and not combined with some cell death? Table S1 has a "cell death of tumor cell lines" theme after miRNA overexpression.

    Following the reviewer suggestion, Annexin V/DAPI staining has been performed in M238P cells upon transient overexpression of miR-143 or miR-145. No significant cell death was observed (new Supplementary Fig. S4D). Detailed statistical analysis and quantification of the experiment is provided. Staurosporine (Stauro) treatment was used as a positive control of cell death induction.

    Related to this, in Supp. Fig.4C the effect on the cell cycle effect is very small, is this significant? It is unclear when the cell cycle was assessed after miRNA overexpression (72h?), it could be a matter of timing. According to Fig.3E, there is a reduction in growth from 60-72h onwards.

    We performed, as suggested by the reviewer, cell cycle analysis at longer timing after transfection (96 hours) (new Supplementary Fig. S4C). We observed a significant accumulation of melanoma cells in G0/G1 phase upon miR-143 or miR-145 overexpression and a significant decrease of the percentage of cells in S phase. Detailed statistical analysis of the described experiment is provided.

    Statistics. While multiple comparison tests were used, most graphs have asterisks on top of some bars, and it is unclear what is being compared with what. For example, Fig.2B have asterisks on top of BRAFi+MEKi group, does it mean it is significant vs vehicle group? In this and other similar cases (1J, 2C, S1B and others), a comparison against the combination group (BRAFiMEKi+BIBF) is also relevant. This should be revised throughout manuscript.

    As recommended by the reviewer, statistical analysis have been modified in the mentioned figures: Fig. 1J (lines 732/733), Fig. 2B (lines 745/746), Fig. 2C (lines 749/750) and Fig. S1B (see new figures and lines 251/252 of Supplementary materials).

    ***Minor:**

    -For all the studies using stable cell lines, authors should state how long after transduction and selection experiments were performed. *

    As recommended, we have now added this information (see lines 8-12 of Supplementary materials).

    - Authors only show single miRNA overexpression or inhibition. However, both miRNA are upregulated upon MAPKi. Did authors try the double overexpression or blockade?

    As suggested by the reviewer, we experimented the double blockade in M238P and 1205Lu cells treated with MAPK inhibitors. Results are presented in new Fig. 3B, 3D, 3H and Supplementary Fig. S6A-B. Overall, combined inhibition of the two miRNAs had an effect comparable or more significant than the single miRNA inhibition depending on the cellular parameter analyzed.

    Concerning the double overexpression, we already experimented lentivirus-mediated stable overexpression of the two miRNAs in two melanoma cell lines. Results are presented in Supplementary Fig. S5A-F and confirmed the functional effects observed by the single miRNA overexpression.

    - For the 1205Lu xenograft experiment, authors should also show the tumour growth curves, and explain how long treatment was and when miRNA expression was analysed (endpoint?). In addition, why in 5A there are only 3 dots (mice?) per group, while in 5B there are more (6-7 in control, 4-5 in BRAFi)?

    We apologize for this omission. We have added line 270 of the manuscript the reference to the previous study in which the experiment is described. miRNA expression was analyzed in tumors at the endpoint of the experiment i.e. 2 weeks after Vemurafenib treatment start. Moreover, we performed again the analysis of FSCN1 and miR-143/145 expression with the same number of mice (n = 6), please see new Fig. 5A.

    - In a few graphs, the axis legend should give more information. For example, Fig.2 says Fold change, and it should be Fold change expression, or similar; Fig.4G fold change FSCN mRNA expression; Fig. S2 log2 expression (resistant/par), S5A...

    We have corrected this and modified y-axis legends in the corresponding figures.

    *- Fig.1E-G and S1B. *Is this at endpoint for each group?

    Yes, it is as stated in the materials and methods section.

    - Fig.3H and S6B. how long were these experiments?

    Experiments shown in Fig. 3H and Fig. S6B were carried out during 72 h. This information has been included in the legend of the corresponding figures.

    - Fig.7B and D. Why the MRTFA signal in miR-neg and siCTRL is so different? Same for UACC in S11A vs s11D.

    We apologize for this inaccuracy. We have revised the figures to show more representative pictures (new Figs. 7B, 7D and S11A, S11D and new Fig. 6C).

    • Fig.5C and 5E. FSCN1 knockdown in 5C is very efficient, while not so much in 5E. However, effects on MITF, AXL etc in 5C are quite impressive. are these knockdowns representative?

    We again apologize for this inaccuracy. We performed a new experiment and we are now showing a more representative FSCN1 knockdown in new Fig. 5E.

    - Fig.6-7 legend. When mentioning scale bar, it reads uM, should it be um?

    We have corrected this mistake.

    • Fig.7A. In the graph, the "YAP nuclear enrichment", do the numbers represent the nuclear/cytoplasm ratio?

    Yes, numbers represent the nuclear/cytoplasm ratio. This information was added in the legend of the corresponding figures.

    - When showing migration and a picture (Fig.3F, 5D, S4D, S5E...), the blue over dark background is difficult to see, using greyscale or a brighter pseudocolour would help

    We thank the reviewer for this useful suggestion. We have done this and used the gray scale to improve the quality of the pictures.

    Reviewer #2 (Significance):

    These findings have important preclinical implications, since the study proposes a biomarker of resistance (profibrotic signature) and importantly, a potential new therapy to delay MAPKi resistance in melanoma (BIBF). It could also apply to other BRAFmutant cancers and diseases cursing with fibrosis.

    Field of expertise: melanoma, drug resistance, cytoskeleton*

    Reviewer #3:

    Major comments:

    The manuscript is well written, data are convincing, well presented and supportive of the conclusions.

    We thank the reviewer for his/her interest about our study and supportive comments.

    **Minor points that may be improved:**

    - The expression of miR-143/145 increases in melanoma cell lines treated with BRAFi and/or MEKi for 72h (Fig. 2B, Supp. Fig. 2B-F), and also after the development of resistance to MAPK-targeted therapies (Fig. 2A, Supp. Fig. 2A). The transient overexpression of miRs in therapy-naive cells leads to cells de-differentiation toward a mesenchymal/MAPK resistant state. On the other hand, these cells become more sensitive to BRAFi treatment when combined with LNA-mediated inhibition of miRs activity. It would be important to determine if the same occurs also in resistant cells, or whether MAPKi-resistance is established, cells are no longer sensitive to miRs blockade.

    The answer to this point is common to the point 2 raised by the reviewer #2.

    According to reviewers suggestion, we have conducted new experiments in resistant melanoma cells using different approaches to silence simultaneously the 2 mature miRNAs: i) an ASO-directed RNAse H degradation of the miR-143/145 precursor, as described by Plaisance et al., JACC Basic Transl Sci. 2016, 1:472-493 to knock-down the pri-miRNA in cardiomyocytes, and ii) a combination of the 2 anti-miRs ASOs. Unfortunately, the first approach failed to efficiently inhibit the expression of mature miR-143-3p and miR-145-5, suggesting that the miR-143/145 cluster has a different precursor gene in melanoma than the one described in cardiomyocytes.

    Concerning the second approach, as expected, the 2 anti-miRs ASOs as well as the combination of the 2 ASOs efficiently targeted the mature miRNAs (Supplementary Fig.S6C). Inhibition of miR-145-5p alone and combined inhibition of the two miRNAs significantly affected the viability of BRAFi resistant melanoma cells (M238R) in the absence of BRAFi (new Supplementary Fig.S6D) in a similar way as BIBF (Fig. 1J).

    - In 2 out of 4 melanoma PDX samples naïve/resistant to combo BRAFi/MEKi therapy, the expression level of miR-143/145 cluster correlates with the de-differentiated transcriptomic profile of resistant tumor. How is Fascin1 expression in these samples?

    The reviewer legitimately asks about the expression level of the miR-143/-145 target FSCN1 in the PDX samples used in the study. Expression of FSCN1 in PDX resistant vs naïve samples has been assessed by RT-qPCR. Results are provided. We observed decreased expression of FSCN1 in only 1 out of the 2 samples showing increased miR-143/145 expression. This can be due to the heterogeneity of the subpopulations composing the tumor sample. It would have been interesting and probably more informative to test FSCN1 expression also at protein level since often miRNA molecular targets are inhibited at translation level but unfortunately we did not have the access to protein extracts corresponding to these samples.

    - The clinical relevance of the data could be strongly improved by assessing the expression of the miRs cluster and of its target Fascin1 in resistant subsets of patients, comparing their expression to patients before treatment, making use of available datasets.

    We agree with the reviewer about the importance of providing clinical data supporting our observations. We have carefully analyzed all available profiling studies and datasets and provide below a summary.

    Overall, these studies have several limitations: i) as demonstrated in our study, expression of the miRNA cluster is specifically induced in response to therapy and is not present (or barely) in tumors at diagnosis; ii) no small RNA-seq datasets are available yet; iii) melanoma tumors are highly heterogeneous and invaded with stroma, especially CAFs and vessels that also express these miRNAs. We have looked at the expression of the MIR143HG precursor in these datasets and it was not present, probably due to low to medium sequencing depths in these clinical studies.

    We have also carefully explored TCGA datasets to look at possible association between prognosis and mature / precursor miRNA as well as miRNA target (FSCN1) expression in skin cutaneous melanoma (SKCM) using the tools developed by Anaya et al. 2016 PeerJ Computer Science 2:e67. Cox regression models and Kaplan-Meier analysis (using different percentiles) did not show any association of our candidates with survival on a cohort of 459 SKCM patients (median survival of 2.4 years, see Kaplan plots below).

    Finally, during the revision process, we could have access to 9 relapsed melanoma for research purposes from the Dermatology Department of Nice University Hospital (CHU) following treatment with targeted therapies, immunotherapies or a combination of them. We analyzed in these samples the expression of fibrotic/mesenchymal genes, FSCN1 and the miR-143/145 cluster compared to the mean expression of the same genes/miRNAs in therapy naïve patient-derived xenografts (MEL003, MEL006, MEL015, MEL047). Our results indicate that relapsed tumors acquire a strong fibrotic signature which is associated to increased expression of the miR-143/145 cluster and decreased expression of FSCN1 (8 out of 9 patients).

    This represents a good indicator for further clinical validation but is not solid enough to be incorporated in the manuscript. Overall, validation of our hypotheses in patient samples would require an entire new and highly complex clinical study comparing tumors at diagnosis with relapsed tumors after targeted therapies and ideally processed using single-cell RNA-seq and/or RNA FISH to take into account the stromal compartment.

    Minor comments:

    - Fig. 4C, lower legend: M238P not M238S.

    We apologize for this mistake and corrected it.

    Reviewer #3 (Significance):

    **Nature and significance of the advances:**

    The findings not only suggest the combination therapy with the anti-fibrotic drug Nintedanib to be effective in enhancing MAPKi treatment in melanoma, reducing the development of resistance, but identify the molecular mechanism via the induction o the miR-143/145 cluster and the effects on the target Fascin1.

    **Compare to existing knowledge**

    These two miRNAs have been shown to have both oncogenic and oncosuppressor activities and have already been involved in EMT induction. The findings add yet one more piece to the puzzle.

    **Audience**

    This manuscript is not only of interest for oncology researchers but also of general interest or the understanding of fundamental biological processes and their effects on cancer therapy.

    **Your expertise**

    Molecular biologist and cancer research, transcriptional control of tumor transfromatin and progression including EMT, microRNAs -143/145

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary:

    In the present work Diazzi and co-authors describe the mechanism through which the anti-fibrotic drug Nintedanib potentiates MAPK-targeted therapy efficacy in melanoma cells. Nintedanib prevents the MAPK-induced pro-fibrotic response and is associated with loss of miR-143/-145 cluster expression. These miRs promote melanoma cells de-differentiation towards a pro-fibrotic mesenchymal-like state that correlates with resistance to MAPK inhibitors. Looking for miR-143/-145 targets responsible for this phenotype switch, the authors identified Fascin1 as a crucial regulator of cytoskeleton dynamics and mechanopathways.

    Major comments:

    The manuscript is well written, data are convincing, well presented and supportive of the conclusions.

    Minor points that may be improved:

    • The expression of miR-143/145 increases in melanoma cell lines treated with BRAFi and/or MEKi for 72h (Fig. 2B, Supp. Fig. 2B-F), and also after the development of resistance to MAPK-targeted therapies (Fig. 2A, Supp. Fig. 2A). The transient overexpression of miRs in therapy-naive cells leads to cells de-differentiation toward a mesenchymal/MAPK resistant state. On the other hand, these cells become more sensitive to BRAFi treatment when combined with LNA-mediated inhibition of miRs activity. It would be important to determine if the same occurs also in resistant cells, or whether MAPKi-resistance is established, cells are no longer sensitive to miRs blockade.
    • In 2 out of 4 melanoma PDX samples naïve/resistant to combo BRAFi/MEKi therapy, the expression level of miR-143/145 cluster correlates with the de-differentiated transcriptomic profile of resistant tumor. How is Fascin1 expression in these samples?
    • The clinical relevance of the data could be strongly improved by assessing the expression of the miRs cluster and of its target Fascin1 in resistant subsets of patients, comparing their expression to patients before treatment, making use of available datasets.

    Minor comments:

    • Fig. 4C, lower legend: M238P not M238S

    Significance

    Nature and significance of the advances:

    The findings not only suggest the combination therapy with the anti-fibrotic drug Nintedanib to be effective in enhancing MAPKi treatment in melanoma, reducing the development of resistance, but identify the molecular mechanism via the induction o the miR-143/145 cluster and the effects on the target Fascin1.

    Compare to existing knowledge

    These two miRNAs have been shown to have both oncogenic and oncosuppressor activities and have already been involved in EMT induction. The findings add yet one more piece to the puzzle.

    Audience

    This manuscript is not only of interest for oncology researchers but also of general interest or the understanding of fundamental biological processes and their effects on cancer therapy.

    Your expertise

    Molecular biologist and cancer research, transcriptional control of tumor transfromatin and progression including EMT, microRNAs -143/145

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In this very interesting study, Diazzi and colleagues show that during adaptation to MAPK-targeted therapy (MAPKi), melanoma cells upregulate a miRNA profibrotic cluster (miR-143, -145), which drives a phenotypic switch towards a drug resistant undifferentiated mesenchymal-like state. From the miRNA targets, authors identify FSCN1 as a gene that needs to be downregulated during adaptation to MAPKi by the miRNAs, since FSCN1 ablation promotes the drug resistant phenotype. Importantly, authors show in a preclinical mouse melanoma model that the anti-fibrotic drug nintedanib (BIBF) improves response to MAPKi and delays onset of resistance.

    The study conclusions are convincing and the data are adequately replicated and presented, authors should be commended for having the manuscript in such good shape. However, there are a few issues that authors should clarify/expand.

    1. The study starts with the in vivo YUMM1.7 model and combination BRAFi+MEKi, and then authors use this combination in many in vitro experiments. However, when studying resistant lines, only BRAFi-resistant and -sensitive pairs were used. I would suggest including more validation of the upregulation of the miRNA and the fibrotic genes on BRAFi+MEKi-resistant lines, and this could be easily gathered from published transcriptomes of several BRAFi+MEKi-resistant melanoma lines from Roger Lo's lab (Song et al 2017 Cancer Discov, including M238, M229, M249 used by the authors). To complement this approach, miRNA expression could be evaluated in large collections of melanoma cell lines classified as more or less undifferentiated (correlating with more or less resistance) as in Tsoi 2018 Cancer Cell and Verfaille 2015 Nat Commun.

    Related to this, the clinical relevance would increase if findings were validated using patient samples, for example, from published transcriptomes (Hugo 2015 Cell, Song 2017 Cancer Discov, Wagle 2014 Cancer Discov...) or even from TCGA, which could be used to identify if patients with high miRNA have worse prognosis.

    1. While blocking the miRNA improves BRAFi response (Fig.3H), it is not clear that this combination would overcome resistance (using resistant lines), although authors show that BIBF does overcome resistance (Fig.1J). This also applies to line 277 ".. mirroring the effect of miR143/145 ASOs, forced expression of FSCN1 in M238R cells decreased viability in the presence of BRAFi (Fig.5H)." However, the miRNA ASOs were used in parental cells (Fig.3H).
    2. Analysis of cytoskeletal changes. Text (lines 284-287) is missing references, regarding "..morphological changes with cells assuming flattened spindle-like shape" and "..function of FSCN1 in F-actin microfilaments reorganization..". Besides, authors say that transient overexpression of miRNAs reproduced these morphological changes as shown by F-actin staining. These would have benefited from including also side-by-side comparison of BRAFi treatment on these cell lines. To my knowledge, these melanoma lines (M238, M229, etc) have not been characterized in that regard (F-actin, focal adhesions). In Nazarian et al 2010, only brightfield pictures are shown in a supplementary figure. The same applies to YAP and especially MRTF activation upon miRNA overexpression, and whether this mirrors what BRAFi does to YAP and MRTF. In Misek et al 2020 and Kim et al 2015 YAP and MRTF were shown to be more enriched in the nucleus in resistant than in parental cells. Kim et al also show in time course experiments that there is significantly higher nuclear YAP after 7-14 days of BRAFi treatment. In the present manuscript, authors seemed to have assessed nuclear YAP/MRTF after 72h miRNA overexpression. Does it mirror MAPKi?
    3. Regarding the decreased proliferation/survival after miRNA overexpression, is it truly slow cycling and not combined with some cell death? Table S1 has a "cell death of tumor cell lines" theme after miRNA overexpression.

    Related to this, in Supp. Fig.4C the effect on the cell cycle effect is very small, is this significant? It is unclear when the cell cycle was assessed after miRNA overexpression (72h?), it could be a matter of timing. According to Fig.3E, there is a reduction in growth from 60-72h onwards.

    1. Statistics. While multiple comparison tests were used, most graphs have asterisks on top of some bars, and it is unclear what is being compared with what. For example, Fig.2B have asterisks on top of BRAFi+MEKi group, does it mean it is significant vs vehicle group? In this and other similar cases (1J, 2C, S1B and others), a comparison against the combination group (BRAFiMEKi+BIBF) is also relevant. This should be revised throughout manuscript.

    Minor:

    -For all the studies using stable cell lines, authors should state how long after transduction and selection experiments were performed.

    -Authors only show single miRNA overexpression or inhibition. However, both miRNA are upregulated upon MAPKi. Did authors try the double overexpression or blockade?

    -For the 1205Lu xenograft experiment, authors should also show the tumour growth curves, and explain how long treatment was and when miRNA expression was analysed (endpoint?). In addition, why in 5A there are only 3 dots (mice?) per group, while in 5B there are more (6-7 in control, 4-5 in BRAFi)?

    -In a few graphs, the axis legend should give more information. For example, Fig.2 says Fold change, and it should be Fold change expression, or similar; Fig.4G fold change FSCN mRNA expression; Fig. S2 log2 expression (resistant/par), S5A...

    -Fig.1E-G and S1B. Is this at endpoint for each group?

    -Fig.3H and S6B. how long were these experiments? Fig.7B and D. Why the MRTFA signal in miR-neg and siCTRL is so different? Same for UACC in S11A vs s11D.

    -Fig.5C and 5E. FSCN1 knockdown in 5C is very efficient, while not so much in 5E. However, effects on MITF, AXL etc in 5C are quite impressive. are these knockdowns representative?

    -Fig.6-7 legend. When mentioning scale bar, it reads uM, should it be um?

    -Fig.7A. In the graph, the "YAP nuclear enrichment", do the numbers represent the nuclear/cytoplasm ratio?

    -When showing migration and a picture (Fig.3F, 5D, S4D, S5E...), the blue over dark background is difficult to see, using greyscale or a brighter pseudocolour would help.

    Significance

    These findings have important preclinical implications, since the study proposes a biomarker of resistance (profibrotic signature) and importantly, a potential new therapy to delay MAPKi resistance in melanoma (BIBF). It could also apply to other BRAFmutant cancers and diseases cursing with fibrosis.

    Field of expertise: melanoma, drug resistance, cytoskeleton

  5. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    The manuscript is interesting and well presented. The authors propose the use of an antifibrotic drug to attenuate resistance to RTK inhibitors.

    Specific comments

    1. It is not entirely clear how Nintedanib decreases tumour growth. It may be due to its effect on resistant melanoma cells as proposed, but it could also be due to the effect on CAFs. This should be at least discussed
    2. A potential caveat is that drug used is non-specific as it also blocks PDGFR signalling. Hyperactivation of RTKs is a mechanism of BRAFi resistance and for example in Figure 1J, they see that BIF1120/Nintedanib has a significant effect on BRAFi-resistant cells, which may indicate that the growth inhibition seen in allografts could be a combination of an "anti-fibrotic" role and its own activity inhibiting the survival of resistant cells. This needs to be considered.
    3. Does the viability decrease in BRAFi-sensitive cells? For instance, in the parental cells.
    4. Figure 1 b-e, in vivo and in vivo experiments. How many animals we used? Collagen decrease is not quantified (statistics missing).
    5. The title is not accurate. "prevent" resistance in melanoma is an overestimation because the cells do become resistant, albeit later.

    Significance

    As the authors discussed, they and others have previously studied the contribution of ECM and stromal remodelling to resistance to targeted therapies in melanoma. Previous data from E. Sahai´s lab show that BRAFi activate CAFs and increase the production and remodelling of the extracellular matrix, but in this work, they look at a cell-autonomous mechanism mediated by miRs that promotes fibrosis and propose the use of an antifibrotic drug to attenuate resistance to RTK inhibitors.

  6. Version published to 10.1101/2021.07.02.450838v1 on bioRxiv