New level of regulation of MITF activity by direct interaction with RAF proteins

Charlène Estrada Institut Curie PSL Research University Liliana Mirabal-Ortega Institut Curie PSL Research University Florent Dingli Institut Curie PSL Research University https://orcid.org/0000-0002-7715-2446 Laetitia Besse Institut Curie PSL Research University Cedric Messaoudi Institut Curie PSL Research University https://orcid.org/0000-0003-3535-7723 Damarys Loew Institut Curie PSL Research University https://orcid.org/0000-0002-9111-8842 Celio Pouponnot Institut Curie PSL Research University Corine Bertolotto Centre Méditerranéen de Médecine Moléculaire Alain Eychene Institut Curie Université Paris Sud https://orcid.org/0000-0002-6818-7225 Sabine Druillennec (  sabine.druillennec-rodiere@curie.fr ) Institut Curie INSERM U1021 CNRS UMR3347 https://orcid.org/0000-0003-0237-1465


INTRODUCTION
Cutaneous melanoma is an aggressive tumour arising from malignant transformation of melanocytes 1 . The RAS/RAF/MEK/ERK is a key signaling pathway frequently mutated in cutaneous melanoma since activating mutations in either NRAS or BRAF genes occur in 15-20% and 40-50% of cases, respectively, the two main mutations being NRAS Q61K and BRAF V600E 2,3 . RAS is a GTPase activated via membrane-bound receptors upon stimulation by growth factors. In its GTP-bound form, RAS recruits effectors at the membrane and stimulates a number of downstream intracellular signaling pathways, including the MAPK/ERK pathway 4 . The three RAF serine-threonine kinases, ARAF, BRAF and CRAF conserved in vertebrates are among the main RAS effectors. RAF activation enables subsequent activation by phosphorylation of MEK1 and MEK2, which in turn activate ERK1 and ERK2 5 . Once translocated in the nucleus, ERK regulates a wide variety of transcription programs leading to modulation of key biological processes such as cell proliferation, survival, migration or differentiation 6 .
Using conditional knockout of BRAF and/or CRAF in a mouse melanoma model induced by NRAS Q61K , we showed that while BRAF is required downstream of activated NRAS for tumour initiation, both BRAF and CRAF play compensatory functions during late phases of melanomagenesis, thus highlighting the addiction of melanoma to the RAF/ERK pathway 7 . Interestingly, we demonstrated that in the absence of BRAF and CRAF, ARAF alone can sustain both ERK activation and proliferation in NRAS-mutated melanoma cells. In this context, we also observed that ARAF homodimers are sufficient to induce ERK paradoxical activation by vemurafenib, a widely used BRAF V600E inhibitor, in the absence of both BRAF and CRAF. Our results suggested a dependency toward ARAF kinase, as well as a possible role of ARAF in resistance mechanisms in cutaneous melanoma. The potential role of ARAF in NRAS-induced melanoma was further strengthened by an in silico search in public databases that allowed to identify patients with metastatic melanomas harbouring an ARAF mutation associated with activating NRAS mutations 7 . Nevertheless, ARAF remains the less studied member of the RAF family because: i) ARAF displays the lowest kinase activity towards MEK compared to other 4 RAF proteins 8 , ii) in most cellular models, the role of ARAF is hidden by the predominant roles of BRAF and CRAF.
Microphtalmia-associated transcription factor (MITF) is a master regulator of the melanocytic lineage since it is essential for the differentiation, survival and proliferation of melanocytes 9,10 . MITF belongs to the MiT family, gathering bHLH-LZ domain transcription factors (TFEB, TFEC and TFE3), that can homo-or hetero-dimerize to regulate gene expression 11 . Expressed in about 80% of human melanoma 12,13 , MITF displays a central regulatory role in melanoma cell phenotypic plasticity. A proposed rheostat model suggests that the global level of MITF activity correlates with the phenotype of melanoma cells: at high levels of activity, MITF sustains the proliferative state of melanoma cells while at lower levels, MITF is associated with an invasive and stem-like phenotype [14][15][16][17] . This model has been recently refined with six different states some being resistant to therapy but still on the basis of MITF activity 18,19 . In line with its central role, MITF is finely regulated to ensure the homeostasis of melanocytes or melanoma cells 10 . Among its numerous post-translational regulators, MITF is regulated by ERK2, that phosphorylates the S73 residue inducing both proteasome-mediated degradation and increased activity via the recruitment of p300/CBP transcription cofactor [20][21][22][23][24] . Altogether, the posttranslational regulation of MITF by ERK pathway has opposite consequences regarding MITF activity, depending on cellular context 13 .
To better characterize the role of ARAF in NRAS-driven melanoma, we searched for new ARAF interactors by mass spectrometry. Our results showed that ARAF directly binds to the transcription factor MITF. ARAF/MITF complexes were found in the cytosol of NRAS-mutated melanoma cells. Not only ARAF, but also BRAF and CRAF interacted with MITF and the kinase active status was required to allow complex formation. At the functional level, RAF/MITF interaction modulates MITF nuclear localization, thus regulating its transcriptional activity. Taken together, these results highlight a new level of regulation of MITF by RAF, two key players of melanoma biology.

Identification of new ARAF partners by large-scale analysis
Although our knowledge of ARAF kinase has enlarged over the last decade 25 , ARAF remains understudied compared to the other members of RAF family. Owing to the redundant roles and the high homology of RAF kinases as well as the weak kinase activity of ARAF compared to BRAF and CRAF, it is challenging to study ARAF function in most cellular models where BRAF and CRAF are also expressed. In addition, attribution of a specific function to each RAF kinase is further hampered by their propensity to heterodimerize, especially when looking for binding partners. In the present study, we took advantage of a genetically engineered NRAS-driven melanoma mouse model allowing concomitant ablation of BRAF and CRAF to investigate the role of ARAF. Tumor cells derived from these mice constitute a well-adapted model to study ARAF function in the melanoma context in absence of BRAF and CRAF expression (ARAF-only cells) 7 . The ARAF interactome was established by immunoprecipitation of the endogenous ARAF protein from ARAF-only or control cells followed by analysis of the immune complexes by mass spectrometry in label-free conditions (Fig. 1). ARAF-only cells, which emerged after braf and craf genes ablation in melanoma cultures established from primary NRAS-induced tumours, are highly dependent on ARAF expression for their growth and survival 7 .
Control cells display normal levels of BRAF and CRAF, but express low level of ARAF shRNAmediated knockdown, thus allowing relative quantification of the data. The distribution of the 2700 ARAF-interacting proteins in ARAF-only or control cells is illustrated by the volcano plot (Fig. 1a, Table 1). Well-known ARAF partners were found enriched in ARAF-only cells, such as its upstream activator, the GTPase NRAS, or its substrate MEK1, indicating the reliability of the experimental approach ( Table 1). In order to identify ARAF relevant partners, which functionally impact melanoma cell proliferation, we developed a siRNA-based functional screen on 99 selected targets (Fig. 1a). These 99 interactors were selected as followed: 69 were chosen among 359 proteins enriched in ARAF-only cells with the following parameters: number of peptides>9, ratio>2 and adjusted p-value<0.001 (Fig.   1a). 15 were from the 72 proteins exclusively identified in ARAF-only cells. We also included 15 6 proteins that were found both in our current dataset and in the ARAF interactome published by Zhang et al. 26 . ARAF-only cells growth was followed upon knock-down of the selected partners by siRNA pools transfection ( Supplementary Fig. 1). Among the 99 partners tested, 16 impacted the growth of ARAF only melanoma cells. It appeared that 11 ARAF partners had an anti-proliferative effect while 5 proteins were pro-proliferative (labelled in green and red, respectively in Fig. 1b). Among the 132 interactors identified in the ARAF proteome by Zhang et al. 26,27 , 107 were commonly found in our dataset, showing the robustness of the approach. Twenty-four of the common identified partners were included in our screen: 9 were selected by the previously described parameters and 15 additional were Among the 16 partners that impact melanoma proliferation, we decided to focus on MITF since it represents a key transcription factor for melanoma progression that can be involved in therapy-resistance mechanism. It is also well known to be regulated by the MAPK/ERK pathway 10,13 . Of note, the ARAF interactome by Zhang et al. could not identify MITF as an ARAF partner since it was performed on a heterologous model overexpressing tagged ARAF in HEK293 cells that do not express MITF 26,27 . We next confirmed the pro-proliferative effect of MITF in ARAF-only cells by using two distinct siRNA against MITF in comparison to control siRNA. Since we previously demonstrated that ARAF-only cells rely on ARAF for their proliferation, we included a siRNA targeting ARAF as a positive control (Fig.   1c). Both siRNA against MITF decreased the growth of ARAF-only melanoma cells. Moreover, we observed a good correlation between the effect on cell proliferation and the level of extinction of MITF expression induced by the different siRNA (Fig. 1d), demonstrating that MITF is required for ARAFonly cells growth.

ARAF directly interacts with MITF
The interaction between ARAF and MITF was confirmed by direct coimmunoprecipitation experiments of endogenous proteins in ARAF-only cells (Fig. 2a). As shown in Fig. 2b, ARAF/MITF complexes were also detected by Proximity Ligation Assay (PLA) in ARAF-only cells, further revealing that this interaction occurred in the cytoplasm of melanoma cells. Importantly, this interaction appeared to be direct since complex formation was observed between ARAF and MITF human purified recombinant proteins, in an in vitro coimmunoprecipitation assay (Fig. 2c).

Characterization of the RAF/MITF interaction
While the connection between the ERK/MAPK pathway and MITF is well established in melanoma 13 , a direct interaction between RAF kinases and MITF has never been previously demonstrated. To further substantiate this observation, we tested whether this interaction was specific of ARAF or shared by all the RAF kinases. HEK293T cells were cotransfected with MITF and each of the three different HAtagged RAF proteins. Anti-HA1 immune complexes were then probed with an anti-MITF antibody.
Interestingly, we observed that MITF could interact not only with ARAF but also with BRAF and CRAF, the two others members of the RAF family ( Fig. 3a-b-c). We confirmed the existence of cytoplasmic BRAF/MITF and CRAF/MITF endogenous complexes by PLA experiments in NRASmutated murine melanoma cells as observed for ARAF ( Supplementary Fig. 2). This is the first evidence of a direct interaction between a RAF kinase and MITF, two key players in melanoma cell biology. Although an MITF interactome has been previously reported, RAF kinases were not identified in this study since the authors focused specifically on nuclear interactors by performing nuclear purification 28 . Since the MAPK/ERK pathway is dysregulated by NRAS, but also BRAF mutations in melanoma, we investigated the ability of MITF to interact with the constitutively active BRAF V600E mutant. This is the most frequent BRAF mutation in human cancers, which is highly prevalent in melanoma and which markedly increases BRAF kinase activity 29 (Fig. 3d). We observed that MITF strongly interacts with BRAF V600E with an increased affinity compared to wild-type BRAF. To evaluate the requirement of the RAF kinase activity, we also tested the interaction with the BRAF K499M kinasedead mutant (BRAF KD ) containing a Lys-to-Met substitution in its kinase domain (Fig. 3d). In contrast 8 to BRAF V600E , the capacity of BRAF KD to bind to MITF was decreased as compared to wild-type BRAF.
Therefore, the strength of the binding directly correlates with the activation state of the RAF proteins since MITF strongly interacts with the activated mutant of BRAF, and much less with the BRAF kinasedead mutant. These results suggest that not only an active form of the RAF kinase is required to allow the interaction with MITF, but also that the MITF/BRAF complex formation can occur in a BRAFmutated context. We next investigated the role of the different domains of BRAF in the interaction with MITF, by using truncated forms of the protein (Fig. 3e). HEK293T cells were cotransfected with plasmids encoding MITF and either the C-terminus or N-terminus part of BRAF, or both. Of note, it was previously demonstrated that the N-terminus regulatory domain of RAF proteins binds to their C-terminus kinase domain in order to regulate their activity 30 . Accordingly, N-and C-terminus parts co-precipitated when co-expressed (Fig. 3e). Following C-terminus immunoprecipitation in the absence of the N-terminus, a strong interaction with MITF was observed indicating that the N-terminal part of BRAF is not required for MITF binding. Moreover, in these conditions, the presence of the N-terminus did not strengthen the interaction between MITF and the C-terminus (Fig. 3e, right panel). On the opposite, a weak interaction with MITF was seen when the N-terminal domain was immunoprecipitated in the absence of the Cterminus (Fig. 3e, left panel). However, complex formation between the N-terminus and MITF was strongly increased in the presence of the C-terminal part suggesting that, in this condition, the Nterminus does not interact directly with MITF but through the C-terminal domain. Altogether, the results indicate that complex formation with MITF involves the C-terminus region of RAF proteins that contains the kinase domain. These observations also suggest the requirement of a functional kinase domain to stabilize the interaction between RAF and MITF.

Functional role of RAF/MITF interaction
We next investigated how the MITF/RAF complex formation could affect the respective subcellular localization of each partner, knowing that RAF kinases are cytosolic, whereas MITF can shuttle between the cytosol and the nucleus. HEK293T cells were transfected with epitope-tagged MITF and ARAF, BRAF or CRAF and subcellular localization was analysed by immunofluorescence (Fig. 4). When expressed alone, MITF was mainly nuclear while RAF proteins displayed a clear cytoplasmic localization. However, when co-expressed, a relocalization of MITF from nucleus to cytoplasm was observed, indicating that complexes between MITF and ARAF, BRAF or CRAF, are cytoplasmic in agreement with previous observations in PLA experiments (Fig. 2b, Supplementary Fig. 2). This suggests that RAF proteins may retain MITF in the cytoplasm.
In order to better understand the functional consequence of this cytoplasmic interaction, we next investigated how RAF proteins could affect MITF transcriptional activity. HEK293T cells were transfected by constructs encoding a luciferase reporter gene under the control of the MITF-regulated tyrosinase promoter, together with increasing amounts of plasmids encoding the RAF proteins. We found that RAF kinases overexpression led to a decrease in MITF transcriptional activity, in a dosedependent manner. Both BRAF and CRAF overexpression strongly suppressed MITF transcriptional activity while ARAF, which possesses a weaker kinase activity, reduces MITF activity to a lesser extent (Fig. 5). Thus, the inhibition of MITF transcriptional activity by RAF proteins appears to be correlated with their kinase activity. Accordingly, the luciferase activity was strongly decreased by BRAF V600E mutant as compared to wild-type BRAF while the BRAF kinase-dead mutant had no effect on MITF transcriptional activity. These results showed that binding to RAF kinases negatively regulates MITF

DATA AVAILABILITY
The datasets produced in this study are available in the PRIDE database 43       Complexes were visualized as red dots by using a fluorescent microscope in wild type cells, namely a NRAS-mutated murine cell line that expresses ARAF, BRAF and CRAF. Cell nuclei were stained with DAPI. ARAF-only cells in which BRAF and CRAF have been deleted were used as a control. Scatter plots represent the average number of dots per nucleus of at least five replicates (n = 5). Means with standard deviations are shown. *** p-value=0.0002, ** p-value<0.0014 compared by unpaired t-test. UNIPROT