Subcellular localization and stability of MITF are modulated by the bHLH-Zip domain

Summary

Microphthalmia-associated transcription factor (MITF) is a member of the basic helix-loop-helix leucine zipper (bHLH-Zip) family and functions as the master regulator of the melanocytic lineage. MITF-M is the predominant isoform expressed in melanocytes and melanoma cells and, unlike other MITF isoforms, it is constitutively nuclear. Mutational analysis revealed three karyophilic signals in the bHLH-Zip domain of MITF-M, spanning residues 197-206, 214-217 and 255-265. Structural characterization of the MITF protein showed that basic residues within these signals are exposed for interactions in the absence of DNA. Moreover, our data indicate that neither DNA binding nor dimerization of MITF-M are required for its nuclear localization. Finally, dimerization-deficient MITF-M mutants exhibited a significantly reduced stability in melanoma cells when compared to the wild type protein. Taken together, we have shown that, in addition to its well-established role in DNA binding and dimer formation, the bHLH-Zip domain of MITF modulates the transcription factor’s subcellular localization and stability.

Introduction

Microphthalmia-associated transcription factor (MITF) plays a key role in the establishment of various cell types including melanocytes, the pigment-producing cells of the vertebrate body (Steingrimsson et al., 2004). In fact, MITF controls the expression of genes important for melanocyte proliferation, migration, survival and differentiation and is thus often referred to as the master regulator of this lineage (Cheli et al., 2011). Mice carrying mutations at the microphthalmia (mi) locus exhibit severe pigmentary defects of the eyes and coat and some alleles can lead to eye abnormalities, hearing loss, osteopetrosis and mast cell deficiency (Hodgkinson et al., 1993; Steingrimsson et al., 2004). In humans, MITF mutations have been associated with Waardenburg syndrome type 2A (WS2A) and Tietz syndrome (TS), two autosomal-dominant hypopigmentation and deafness disorders (Pingault et al., 2010; Tassabehji et al., 1994). Most of the reported WS2A and TS mutations interfere with the DNA binding ability or the transcription activation potential of MITF (Grill et al., 2013). Furthermore, somatic mutations in MITF and amplification of the gene have been observed in metastatic melanomas (Cronin et al., 2009; Garraway et al., 2005), and a recurrent germline mutation in MITF predisposes to melanoma development (Bertolotto et al., 2011; Yokoyama et al., 2011).

MITF is a member of the MYC supergene family of basic helix-loop-helix leucine zipper (bHLH-Zip) transcription factors (Hodgkinson et al., 1993; Hughes et al., 1993). Proteins of this family all share a basic region that is required for DNA binding and a HLH-Zip domain important for dimerization. DNA binding of MITF can only occur upon homodimerization or heterodimerization with its close relatives, transcription factor EB (TFEB), TFEC and TFE3 (Hemesath et al., 1994). As a consequence of differential promoter usage, MITF is transcribed into several isoforms that differ in their N-termini and show tissue-specific expression patterns (Bharti et al., 2009). MITF-M is the predominant isoform expressed in melanocytes and, unlike other MITF isoforms, it is constitutively nuclear (Takebayashi et al., 1996). In contrast to TFEB and TFE3, two master regulators of autophagy, MITF-M lacks the N-terminal domain important for mTORC1-mediated cytoplasmic retention (Settembre et al., 2012).

Nuclear transport of macromolecules through the nuclear pore complex is an energy-dependent process mediated by members of the karyopherin family. These soluble transport receptors recognize their cargo via small peptide motifs, termed nuclear localization signals, which are characterized by a cluster of basic lysine and arginine residues (Stewart, 2007). MITF has been reported to interact with several members of the nuclear transport machinery, including importins and nucleoporins (Laurette et al., 2015; Perera et al., 2015). A nuclear localization signal has been identified in the basic region of MITF, spanning residues 213 to 218 (Takebayashi et al., 1996; Zhang et al., 2012). However, several lines of evidence point to the presence of multiple regions in MITF that modulate its nuclear localization. For example, the mutant MITF-M protein encoded by the mouse Mitf^ew allele lacks amino acid residues 187-212 and has been reported to localize to the cytoplasmic compartment of HEK293T cells, despite the presence of the above mentioned nuclear localization signal (Takebayashi et al., 1996). Furthermore, mutation of K307 in MITF-A (which corresponds to K206 in MITF-M) has been shown to interfere with the transcription factor’s nuclear localization (George et al., 2016). Finally, C-terminal truncations at residues R214 and R259 of MITF-M, two mutations associated with WS2A, resulted in a cytoplasmic enrichment of the mutant proteins (Grill et al., 2013). In this study, we have used EGFP fusions and in vitro mutagenesis to thoroughly investigate the domains in MITF-M that mediate its nuclear localization and determine its stability.

Materials and Methods

Cloning and expression constructs

The p3XFLAG-CMV^™-14 construct expressing the plus isoform of mouse Mitf-M (MITF-M-FLAG) was kindly provided by Colin Goding. Human TFE3 was cloned into a piggybac vector (pBac TFE3-FLAG-HA) downstream of a tetracycline response element (TRE) with Gibson Assembly (New England Biolabs, Ipswich, MA) using the following primers: 5’-CTCAAACAGTGTATATCATTG-3’, 5’-tgatatacactgtttgagATGTCTCATGCGGCCGAAC-3’, 5’-tcatggtctttgtagtcGGACTCCTCTTCCATGCTG-3’, 5’-GACTACAAAGACCATGACGG-3’. The melanocyte-specific plus isoform of mouse Mitf-M cDNA was cloned as a fusion into the multiple cloning site of the pEGFP-C1 expression vector (Clontech, Mountain View, CA) by using HindIII and ApaI restriction sites (EGFP-MITF). Mutations were introduced into Mitf using the Q5 Site-directed Mutagenesis Kit (New England Biolabs) according to the manufacturer’s instructions and confirmed by sequencing (Genewiz, Essex, UK). Mutagenesis primers are listed in Table S1.

Cell culture and transient transfections

HEK293T, 501MEL, SKMEL28 and LU1205 cells were maintained in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum (both from Gibco, Thermo Fisher Scientific, Waltham, MA) at 37°C and 5% CO₂. The day before transfection, cells were seeded onto 8-well glass slides or 24-well plates at a density of 2x10⁴ or 5x10⁴ cells per well, respectively. Transient transfections were performed using FuGENE HD reagent (Promega, Madison, WI) according to the manufacturer’s instructions. Following incubation for 24 hours, cells were subjected to immunofluorescence stainings or Western blotting.

Immunofluorescence stainings and confocal microscopy

For imaging of EGFP-tagged MITF-M constructs, cells were fixed with methanol-free 4% formaldehyde (Thermo Scientific, Rockford, IL) diluted in PBS, followed by DAPI (Thermo Fisher Scientific) staining for 15 minutes at room temperature. For antibody stainings, formaldehyde-fixed cells were incubated with blocking buffer (5% normal goat serum/0.3% Triton X-100 in PBS) for 1 hour and subsequently stained with mouse anti-MITF (clone C5, Thermo Scientific MS771PABX, 1:200) or mouse anti-FLAG (clone M2, Sigma-Aldrich F3165, 1:5000) antibodies diluted in PBS containing 1% BSA and 0.3% Triton X-100 overnight at 4°C. Finally, cells were incubated with an Alexa Fluor 546 anti-mouse IgG secondary antibody (Life Technologies, Carlsbad, CA) for 1 hour at room temperature, counterstained with DAPI and mounted in Fluoroshield medium (Sigma-Aldrich, St. Louis, MO). Imaging was performed using a FluoView FV1200 laser scanning confocal microscope (Olympus) equipped with a PlanApo N 60X/1.40 ∞/0.17 Oil Microscope Objective. The environmental temperature was set to 17°C. Images were digitally zoomed (1.5X) using FluoView 4.2 acquisition software.

Subcellular fractionations

Cellular compartments were separated using a detergent-based fractionation kit (Cell Signaling Technology, Danvers, MA) according to the manufacturer’s instructions. Approximately 2x10⁶ HEK293T cells overexpressing the respective EGFP-tagged MITF-M constructs were used. Whole cell lysates, representing approximately 20% of sample, and isolated fractions (cytoplasm, membranes/organelles and nuclei) were subjected to SDS-PAGE and Western blotting. The efficiency of separation was determined using compartment-specific markers.

Co-immunoprecipitations

For co-immunoprecipitation studies, EGFP-tagged MITF-M constructs were co-expressed with MITF-FLAG or pBac TFE3-FLAG-HA and a plasmid containing a tetracycline-controlled transactivator (tTA) in a 10:10:1 ratio, respectively. Cells were taken up in lysis buffer (50mM Tris-HCl, 150mM NaCl, 1mM EDTA, 1% Triton X-100, pH7.4) supplemented with protease inhibitors (Sigma-Aldrich) 24 hours post-transfection. The soluble fraction was incubated with ANTI-FLAG M2-Affinity Gel (Sigma-Aldrich) on a roller shaker for 4 hours at 4°C. Proteins were eluted using 3XFLAG peptide at a final concentration of 150 ng/μl (Sigma-Aldrich). After three washing steps, samples were further processed for Western blotting.

Western blotting

Cells were lysed in SDS sample buffer (2% SDS, 5% 2-mercaptoethanol, 10% glycerol, 63 mM Tris-HCl, 0.0025% bromophenol blue, pH6.8) and boiled for 5 minutes at 95°C. Proteins were separated by SDS-PAGE and blotted onto methanol-activated polyvinylidene difluoride membranes (Thermo Scientific). After blocking with 5% BSA (Sigma-Aldrich) in Tris-buffered saline, membranes were probed with the following primary antibodies overnight at 4°C: mouse anti-MITF (clone C5, Thermo Scientific MS771PABX, 1:2000), mouse anti-FLAG (clone M2, Sigma-Aldrich F3165, 1:5000), rabbit anti-GFP (Abcam ab290, 1:4000), mouse anti-β-Actin (clone C4, Millipore MAB1501R, 1:10000), rabbit anti-β-Actin (clone 13E5, Cell Signaling Technology 4970, 1:4000), mouse anti-MEK (clone L38C12, Cell Signaling Technology 4694, 1:1000), rabbit anti-AIF (clone D39D2, Cell Signaling Technology 5318, 1:1000) and rabbit anti-Vimentin (clone D21H3, Cell Signaling Technology 5741, 1:1000). Following three washing steps in TBS containing 0.1% Tween-20, membranes were incubated with DyLight 800 anti-mouse or DyLight 580 anti-rabbit IgG secondary antibodies (Cell Signaling Technology) for 1 hour at room temperature. The blots were scanned with Odyssey imaging system (LI-COR Biosciences, Lincoln, NE) and Image Studio version 2.0. Quantification of band intensities was performed using ImageJ software (Schneider et al., 2012).

Protein degradation assay

HEK293T, 501MEL, SKMEL28 and LU1205 cells overexpressing EGFP- or FLAG-tagged MITF-M mutant constructs were treated with 70 μM cycloheximide (CHX, Sigma-Aldrich) in the presence or absence of 2 μM Vemurafenib (PLX4032, Selleck Chemicals, Houston, TX) for 3 hours, before being lysed in SDS sample buffer and subjected to Western blotting.

Statistical analysis

All data are expressed as mean ± SEM. Statistical analyses were conducted using GraphPad Prism 6 (GraphPad Software). An unpaired t-test was used for the comparison of two means. Analysis of more than two groups was performed using a one-way ANOVA followed by Bonferroni’s post-hoc correction. For subcellular fractionations, a two-way ANOVA followed by Bonferroni’s post-hoc correction was used to determine statistical significance compared to wild type MITF-M. P-values lower than 0.05 were considered statistically significant.

Results

EGFP-MITF mirrors the localization pattern of endogenous MITF

MITF is endogenously expressed in human 501MEL cells and, apart from a minor cytoplasmic fraction, shows a predominantly nuclear localization as visualized by MITF-specific antibody staining (Figure 1a). Importantly, staining of untagged MITF-M overexpressed in HEK293T cells revealed a localization pattern similar to the native MITF protein (Figure 1a). With the aim of generating a tool for identifying regions in MITF that define its subcellular localization and stability, we cloned the melanocyte-specific M-isoform of MITF into an EGFP expression vector (EGFP-MITF-M). Confocal imaging revealed a strong nuclear signal of the fusion protein upon overexpression in 501MEL and HEK293T cells (Figure 1b). Similar to endogenous MITF, a small portion of EGFP-MITF-M was detectable in the cytoplasm of both cell lines. Overexpression of an empty vector (EGFP-EV) resulted in both a nuclear and cytoplasmic signal for GFP, which, due to its small size, can freely diffuse through the nuclear pore (Figure S1a). To rule out EGFP-mediated effects on the subcellular localization of EGFP-MITF-M, we also analyzed the staining pattern of MITF-M carrying a FLAG tag at its C-terminus (MITF-M-FLAG). Indeed, the MITF-M-FLAG protein was predominantly present in the nucleus of 501MEL and HEK293T cells (Figure S1b). Considering the high abundance of endogenous MITF in 501MEL cells, which might interfere with the subcellular localization of our fusion constructs by means of dimerization, we used HEK293T cells for further analyses. To validate our imaging data, we performed subcellular fractionations of HEK293T cells overexpressing untagged MITF-M or EGFP-MITF-M. Both proteins showed a similar distribution among the cytoplasmic, membrane/organelle and nuclear fractions (Figure 1c). The purity of the respective compartments was confirmed by mitogen-activated protein kinase kinase (MEK), apoptosis-inducing factor (AIF) and Vimentin (VIM) staining, respectively. Quantification of band intensities revealed that 33% of untagged MITF-M localized to the MEK-positive cytoplasmic compartment, 6% of MITF-M localized to AIF-positive membranes and organelles, and 61% of the protein was found in the VIM-positive nuclear fraction (Figure 1d). EGFP-MITF-M showed a similar distribution, with 19% of the protein present in the cytoplasmic compartment, 6% in the membrane and organelle fraction and 75% in the nucleus. Taken together, these data validate our EGFP-MITF-M fusion construct as a suitable tool for monitoring effects on the transcription factor’s subcellular localization and stability upon genetic manipulation.

Figure 1. MITF-M overexpression constructs accurately reflect the subcellular localization of endogenous MITF.

(a) Representative confocal images of endogenous MITF in 501MEL cells (left panel) and untagged MITF-M overexpressed in HEK293T cells (right panel). MITF was visualized using an MITF (C5) antibody. (b) Representative confocal images of EGFP-MITF-M overexpressed in 501MEL (left panel) and HEK293T (right panel) cells. Lower panels show merged images with DAPI. Scale bars: 10 μm. (c) Western blot analysis of subcellular fractions isolated from HEK293T cells. Untagged MITF-M and EGFP-MITF-M constructs were visualized using MITF (C5) or GFP antibodies, respectively. The purity of cytoplasmic (CYT), membrane/organelle (MO) and nuclear (NUC) fractions was confirmed by mitogen-activated protein kinase kinase (MEK), apoptosis-inducing factor (AIF) and vimentin (VIM) staining. Whole cell lysates (WCL) were used to represent total protein. (d) Band intensities in CYT, MO and NUC were quantified with ImageJ software and are depicted as percentage of the total amount of protein present in the three fractions. Error bars represent standard error of the mean (SEM) of three independent experiments.

N-terminal or C-terminal protein truncations affect the dimerization potential and nucleocytoplasmic distribution of MITF-M

In a next step, we generated various N- and C-terminally truncated versions of MITF-M to map out regions that affect the subcellular localization and stability of the transcription factor. To this end, we deleted the first 70 to 220 amino acids from the N-terminus (Ndel70, Ndel120, Ndel170 and Ndel220) or inserted stop codons at defined residues (Q258X, N278X, S298X and K316X) of EGFP-MITF-M (Figure 2a). We then overexpressed the truncated proteins in HEK293T cells and analyzed their size and band patterns by Western blotting. Besides the expected size differences of the truncated proteins ranging from approximately 50 to 70 kDa, we noticed drastic changes in the band pattern of various truncation products when compared to the full-length protein (Figure 2b). While EGFP-MITF-M, hereafter referred to as wild type (WT), typically resolves as a double band on a standard Western blot, the upper band was not detected for the N-terminal deletion constructs. This is most likely caused by a loss of S73 phosphorylation in these truncated MITF-M proteins (Hemesath et al., 1998). Interestingly, the upper band of the C-terminally truncated Q258X and N278X mutants appeared less intense, whereas S298X and K316X again showed a double band, comparable with wild type MITF-M. Staining with an MITF-specific antibody (C5) revealed a similar band pattern for the fusion proteins as observed with the GFP antibody except for Ndel170 and Ndel220, which gave no positive signal (Figure 2b). These data suggest that the epitope of the MITF (C5) antibody is located between residues 120 and 170 of MITF-M.

Figure 2. Truncated MITF-M proteins show changes in band patterns and dimerization ability.

(a) Schematic of N-terminal and C-terminal MITF-M truncation constructs. For N-terminal truncations, the first 70, 120, 170 or 220 amino acids were deleted from the MITF-M sequence. C-term truncations were generated by introducing stop codons at position 258, 278, 298 or 316 of MITF-M using site-directed mutagenesis. (b) Western blot analysis of EGFP-tagged MITF-M mutants overexpressed in HEK293T cells. MITF-M constructs were visualized using GFP or MITF (C5) antibodies, Actin was used as a loading control. (c-d) Co-immunoprecipitation of C-terminally truncated EGFP-MITF-M constructs with wild type MITF-M-FLAG (c) or TFE3-FLAG-HA (d) followed by Western blotting. Lysates of HEK293T cells were precipitated with a FLAG-specific antibody and analyzed for the presence of C-terminally truncated MITF-M constructs using a GFP antibody.

Considering the complete or partial lack of the HLH-Zip domain in the Q258X and N278X mutants, respectively, we assessed the ability of C-terminally truncated MITF-M proteins to dimerize with their wild type counterpart. To this end, we co-expressed EGFP-tagged MITF-M constructs with wild type MITF-M-FLAG in HEK293T cells and performed FLAG-immunoprecipitations. Our results indicate that wild type MITF-M, as well as S298X and K316X were able to interact with MITF-M-FLAG. In contrast, Q258X and N278X mutants exhibited a strongly reduced homodimerization ability, with the residual dimerization most likely mediated by the intact HLH domain (Figure 2c). Furthermore, Q258X and N278X failed to co-immunoprecipitate with FLAG-tagged TFE3 when co-expressed in HEK293T cells (Figure 2c). These findings indicate that both mutants lack the ability to homo- and heterodimerize with MITF and TFE3, respectively.

We then analyzed the subcellular localization of N- and C-terminally truncated MITF-M proteins in HEK293T cells by confocal imaging. The N-terminal deletion construct Ndel70 revealed a minor cytoplasmic fraction similar to wild type MITF-M, whereas Ndel120 and Ndel170 were mostly absent from the cytoplasmic compartment (Figure 3a). Interestingly, deletion of the first 220 residues resulted in a slightly increased cytoplasmic localization of the mutant protein when compared to the wild type. Despite a predominantly nuclear localization, the C-terminal truncation product Q258X lacking the leucine zipper, was also highly abundant in the cytoplasmic compartment (Figure 3b). In contrast, the subcellular localization of N278X was comparable to wild type MITF-M, whereas both S298X and K316X were primarily nuclear. Overexpression of these truncated MITF-M proteins in 501MEL cells resulted in similar localization patterns, indicating that this is not a cell-type specific phenotype (Figure S2). In addition, we performed subcellular fractionations of HEK293T cells and in accordance with our imaging data, Ndel120 and Ndel170 as well as S298X and K316X were mainly present within the nuclear fraction, whereas Ndel220 and Q258X showed a strong cytoplasmic signal (Figure 3c-d). As expected, the subcellular localization of Ndel70 and N278X was similar to that of wild type MITF-M. Quantifications of band intensities confirmed that the amount of protein located in the cytoplasm was significantly increased for Ndel220 and Q258X when compared to the wild type (Figure 3e). Of note, the fractionation results obtained for Ndel220 did not reflect the nuclear localization pattern of this mutant seen in the confocal images. Such discrepancy between the two methods has been reported previously (Petrulis et al., 2000) and might be explained by the leakage of nuclear components into the cytosol during detergent-based fractionation. Due to the fact that Ndel220 lacks the basic region and thus fails to associate with DNA, it seems likely that the mutant protein leaks out of the nucleus in the process of extract isolation. Taken together, our data suggest two regions in MITF-M, namely 170-220 and 258-278, to play a role in mediating its nuclear localization. Given the nuclear presence of Q258X and N278X, two dimerization-deficient MITF-M mutants, we further conclude that dimerization is not a prerequisite for the nuclear localization of the transcription factor.

Figure 3. N-terminal and C-terminal protein truncations affect the nucleocytoplasmic distribution of MITF-M.

(a-b) Representative confocal images of N-terminally (a) or C-terminally (b) truncated EGFP-tagged MITF-M constructs overexpressed in HEK293T cells. Lower panels show merged images with DAPI. Scale bars: 10 μm. (c-d) Western blot analysis of subcellular fractions isolated from HEK293T cells. MITF-M constructs were visualized using a GFP antibody. Whole cell lysates (WCL) were used to represent total protein. (e) Band intensities in CYT, MO and NUC were quantified with ImageJ software and are depicted as percentage of the total amount of protein present in the three fractions. Error bars represent standard error of the mean (SEM) of three independent experiments. Two-way Anova with Bonferroni’s multiple comparisons test was performed to determine statistical significance. *, p < 0.05.

Arginines 214-217 in the basic region of MITF-M determine its nuclear localization

Deletion of R217 of MITF-M has been described both in mice (the Mitf^mi allele) and WS2A patients. The mutant protein encoded by Mitf^mi acts in a dominant-negative manner by interfering with the nuclear transport and DNA binding ability of MiT/TFE family members (George et al., 2016; Grill et al., 2013; Takebayashi et al., 1996). Confocal imaging confirmed that deletion of R217 in our EGFP-MITF-M construct (R217del) led to an increased cytoplasmic localization of the mutant protein in HEK293T cells when compared to the wild type (Figure 4a). However, a significant portion of R217del was still present within the nucleus, suggesting that DNA binding is not required for the nuclear retention of MITF.

Figure 4. The basic region of MITF-M comprises a nuclear localization signal.

(a) Representative confocal images of EGFP-tagged R217del, R214/217A, R214-217A and R214-217del MITF-M constructs overexpressed in HEK293T cells. Lower panels show merged images with DAPI. Scale bars: 10 μm. (b) Western blot analysis of subcellular fractions isolated from HEK293T cells. MITF-M constructs were visualized using a GFP antibody. Whole cell lysates (WCL) were used to represent total protein. (c) Band intensities in CYT, MO and NUC were quantified with ImageJ software and are depicted as percentage of the total amount of protein present in the three fractions. Error bars represent standard error of the mean (SEM) of three independent experiments. Two-way Anova with Bonferroni’s multiple comparisons test was performed to determine statistical significance. *, p < 0.05.

The basic region of MITF-M contains a stretch of four basic residues (R214-217) that has been previously described to act as a nuclear localization signal (Takebayashi et al., 1996; Zhang et al., 2012). In order to assess the involvement of these residues in the nuclear localization of MITF-M, we generated various EGFP-tagged MITF-M constructs with these arginines mutated to alanines in different pairs (R214/215A, R215/216A, R216/217A, R214/217A) or all together (R214-217A). As expected, replacing all four arginines with alanines resulted in a predominantly cytoplasmic localization of the R214-217A mutant protein in HEK293T cells (Figure 4a). The same result was obtained for R214-217del and R214-217E mutants (Figure 4a and S3a). MITF-M fusion proteins with two of the four arginines mutated to alanines were equally distributed among the nuclear and cytoplasmic compartments (Figure 4a and S3a). Again, these mutants exhibited comparable localization patterns in 501MEL cells (Figure S3b). Subcellular fractionations followed by Western blotting revealed a prominent band for R217del, R214/217A, R214-217A, and R214-217del in the cytoplasm of HEK293T cells, whereas the signal detected in the nuclear compartment was strongly reduced when compared to the wild type (Figure 4b). Quantification of band intensities showed a significant enrichment of R217del, R214/217A, R214-217A and R214-217del mutants in the cytoplasmic fraction (Figure 4c). Notably, the R217del and R214-217del proteins were highly abundant in the membrane and organelle fraction. Taken together, these findings confirm that arginines 214-217 in MITF-M represent a nuclear localization signal and that the four residues are all important for this function.

Residues 197-206 and 255-265 in the bHLH-Zip domain of MITF-M show karyophilic properties

Takebayashi et al. have reported that the mutant Mitf^ew protein, which lacks amino acid residues 187-212, showed an increased cytoplasmic staining pattern when compared to its wild type counterpart (Takebayashi et al., 1996). These findings point to the presence of an additional nuclear localization signal in MITF. Indeed, a cluster of basic amino acid residues is located upstream of the DNA binding domain, ranging from residues R197 to K206. Mutation of lysine and arginine residues within this region to alanines (K/R197-206A) resulted in a predominantly cytoplasmic localization of the mutant protein (Figure 5a). While substitution of R197 with serine (R197S), a mutation found in melanoma tumors (https://portal.gdc.cancer.gov/), only had minor effects, replacement of residues 203-206 with alanines (K/Q/R203-206A) again showed drastic changes in the nucleocytoplasmic distribution of MITF-M in HEK293T cells (Figure 5a). Of note, we observed similar localization patterns of these constructs in 501MEL cells (Figure S4a). Subcellular fractionations and quantification of band intensities further confirmed that the majority of both K/R197-206A and K/Q/R203-206A proteins was present in the cytoplasmic fraction of HEK293T cells (Figure 5b-c). The R197S mutant also showed a significant cytoplasmic enrichment when compared to the wild type (Figure 5c). These data strongly suggest that residues 197-206 are involved in directing MITF-M into the nucleus.

Figure 5. Two additional karyophilic clusters in the bHLH-Zip domain of MITF-M modulate its nuclear localization.

(a-b) Representative confocal images of EGFP-tagged K/R197-206A, R197S and K/Q/R203-206A (a) or K/R255-265A, K/R255/256A and K/R259/263/265A (b) MITF-M constructs overexpressed in HEK293T cells. Lower panels show merged images with DAPI. Scale bars: 10 μm. (c) Western blot analysis of subcellular fractions isolated from HEK293T cells. MITF-M constructs were visualized using a GFP antibody. Whole cell lysates (WCL) were used to represent total protein. (d) Band intensities in CYT, MO and NUC were quantified with ImageJ software and are depicted as percentage of the total amount of protein present in the three fractions. Error bars represent standard error of the mean (SEM) of three independent experiments. Two-way Anova with Bonferroni’s multiple comparisons test was performed to determine statistical significance. *, p < 0.05.

In an attempt to explain the increased cytoplasmic presence of Q258X (Figure 3b) and R259X mutant MITF (Grill et al., 2013), we searched the MITF sequence for additional clusters of basic amino acids located downstream of residue R259. Indeed, we found two arginine- and lysine-rich regions in the HLH-Zip domain of MITF, spanning amino acids 255-265 and 270-280. While mutating the basic amino acids between residues 270 and 280 (H/K/R270-280A) did not interfere with the subcellular localization of the transcription factor (data not shown), K/R255-265A mutant MITF-M revealed an increased cytoplasmic signal in HEK293T cells when compared to its wild type counterpart (Figure 5a). Interestingly, we observed a strong cytoplasmic abundance of both K/R255/256A and K/R259/263/265A mutants, suggesting that all basic residues located between residues 255 and 265 are important for the nuclear localization of MITF-M (Figure 5a). Overexpression of these constructs in 501MEL cells confirmed the results obtained for HEK293T cells (Figure S4b). According to our subcellular fractionation data, the K/R255-265A mutant was significantly enriched in the cytoplasm of HEK293T cells and band intensities of K/R255/256A and K/R259/236/265A were similar to those of K/R255-265A in the respective compartments (Figure 5b-c). These findings indicate that residues 255-265 located in the HLH-Zip domain represent another karyophilic sequence in MITF-M. Importantly, characterization of the MITF protein structure (Pogenberg et al., 2012) showed that all three potential nuclear localization signals are exposed for interactions in the absence of DNA (Figure 6a). Finally, multiple sequence alignment revealed that the basic residues within these signals are highly conserved among the MiT/TFE family of transcription factors (Figure 6b).

Figure 6. Potential nuclear localization signals in MITF-M are exposed for interactions and conserved among the MiT/TFE family of transcription factors.

(a) Crystal structure of MITF in the absence of DNA. An MITF dimer is shown in cyan and green, the three potential nuclear localization signals are highlighted in purple. Basic residues within these signals are labelled on one MITF protomer. Image was drawn using PyMOL software (PyMOL Molecular Graphics System, Schrödinger, Inc.). (b) Multiple sequence alignment of human MITF, TFEB, TFEC, TFE3 and MYC was performed using ClustalWS and Jalview 2.10.3 software. Conserved lysine (red) and arginine (magenta) residues are highlighted.

Dimerization-deficient MITF-M mutants exhibit a decreased stability in melanoma cells

We next sought to investigate whether the subcellular localization and/or dimerization ability of MITF-M affect its stability. To address this question, we first overexpressed our N-terminally truncated MITF-M mutants in HEK293T and 501MEL cells and assessed their stability in comparison to the wild type protein after inhibiting translation with cycloheximide (CHX). Western blot analyses revealed that after 3 hours of CHX treatment, levels of wild type MITF-M decreased by around 25% or 35% in HEK293T or 501MEL cells, respectively (Figure 7a-b). A similar decrease in protein levels was observed for all N-terminal deletion constructs, suggesting that the N-terminus of MITF does not affect protein stability (Figure 7a-b). In this context, it is interesting to note that the mitogen-associated protein kinase (MAPK) pathway has been reported to trigger MITF degradation through extracellular signal-regulated kinase (ERK)-mediated phosphorylation of S73 (Wu et al., 2000), a residue lacking in Ndel120, Ndel170 and Ndel220. To test whether BRAF/ERK signaling plays a role in regulating the stability of MITF in BRAF(V600E) mutant melanoma, we treated 501MEL cells overexpressing EGFP-MITF-M with CHX and the selective BRAF(V600E) kinase inhibitor Vemurafenib (PLX4032) for 3 hours. However, protein levels of the wild type protein remained unchanged in the presence of Vemurafenib when compared to CHX treatment alone (Figure 7c). We also assessed the stability of FLAG-tagged S73A mutant MITF-M in 501MEL cells but did not observe any effects on protein levels when compared to the wild type (Figure 7d). These findings suggest that neither hyperactive BRAF signaling nor S73 phosphorylation are involved in modulating the stability of MITF.

Figure 7. The N-terminus of MITF is not involved in regulating protein stability.

(a-d) Western blot analysis of HEK293T (a) or 501MEL cells (b-d), treated with cycloheximide (CHX, 70 μM), Vemurafenib (PLX4032, 2 μM) or vehicle control (DMSO) for 3 hours. MITF-M constructs were visualized using GFP (a-c) or FLAG antibodies (d). Levels of phosphorylated extracellular signal-regulated kinase (p-ERK) were analyzed to confirm BRAF inhibition, Actin was used as a loading control. Quantification of band intensities was performed using ImageJ software. Total MITF levels were normalized to Actin, CHX groups were standardized to the corresponding DMSO group (set to 1). Error bars represent standard error of the mean (SEM) of three independent experiments. One-way Anova with Bonferroni’s multiple comparisons test (a-b) or unpaired student’s t-test (c-d) was performed to determine statistical significance.

Finally, we also determined the stability of our C-terminally truncated MITF-M mutants in the presence of CHX. When overexpressed in HEK293T cells, levels of the mutant proteins dropped by around 30-35% in response to CHX, which was comparable to the wild type (Figure 8a). Surprisingly, the two dimerization-deficient MITF-M mutants Q258X and N278X exhibited a significantly reduced stability in 501MEL cells upon CHX treatment, whereas S298X and K316X showed no effects (Figure 8b). To further assess the possibility of dimerization affecting MITF stability, we introduced several mutations into the leucine zipper region with the aim of disrupting the dimerization potential of the full-length protein. Co-expression of these EGFP-tagged zipper mutants with MITF-M-FLAG followed by FLAG-immunoprecipitations revealed that some of the mutants retained their ability to dimerize, whereas L257/267/274/281/288R (Zip mut 3), L267/274/281R (Zip mut 5) and L267/274/281W (Zip mut 6) clearly failed to interact with their wild type counterpart (Figure S5a). Of note, Zip mut 3 also exhibited an impaired heterodimerization potential with TFE3-FLAG (Figure S5b). Interestingly, all dimerization-deficient MITF-M proteins localized to the nucleus of HEK293T cells (Figure S5c), further suggesting that the nuclear localization of MITF is independent of its ability to dimerize. Similar to what has been observed for Q258X and N278X in HEK293T cells, Zip mut 3 showed no obvious changes in protein levels upon CHX treatment when compared to wild type MITF-M in this cell line (Figure 8c). However, protein levels of Zip mut 3 were significantly diminished in 501MEL cells, again pointing towards cell type-specific effects on MITF stability (Figure 8c). Importantly, when overexpressed in BRAF(V600E) mutant SKMEL28 or LU1205 melanoma cells, Zip mut 3 also exhibited a significantly reduced stability upon CHX treatment when compared to the wild type (Figure 8d). Taken together, our data indicate that dimerization, rather than subcellular localization, is important for the stability of MITF-M in melanoma cells.

Figure 8. Dimerization-deficient MITF-M mutants are less stable in melanoma cells.

(a-c) Western blot analysis of HEK293T (a, c left panel) or 501MEL cells (b, c right panel), treated with cycloheximide (CHX, 70 μM) or vehicle control (DMSO) for 3 hours. MITF-M constructs were visualized using a GFP antibody, Actin was used as a loading control. (d) Western blot analysis of SKMEL28 and LU1205 cells treated with CHX (70 μM) or DMSO for 3 hours. MITF-M constructs were visualized using a GFP antibody, Actin was used as a loading control. Quantification of band intensities was performed using ImageJ software. Total MITF levels were normalized to Actin, CHX groups were standardized to the corresponding DMSO group (set to 1). Error bars represent standard error of the mean (SEM) of three independent experiments. One-way Anova with Bonferroni’s multiple comparisons test (a-b) or unpaired student’s t-test (c-d) was performed to determine statistical significance. *, p < 0.05.

Discussion

A common way to regulate the activity of transcription factors involves their sequestration in the cytoplasm until adequate signals trigger their nuclear import. Such a mechanism has been described for TFEB and TFE3, which are retained in the cytoplasm under basal conditions through mTORC1-mediated phosphorylation and 14-3-3 binding (Roczniak-Ferguson et al., 2012). In response to cellular stress such as nutrient deprivation, mTORC1 activity is switched off, TFEB and TFE3 are dephosphorylated by calcineurin and subsequently translocate into the nucleus where they regulate the expression of their target genes (Medina et al., 2015; Puertollano et al., 2018; Roczniak-Ferguson et al., 2012). Owing to differential usage of alternative promoters, the MITF protein is generated in various isoforms (Bharti et al., 2009). With the exception of the short MITF-M isoform, all other isoforms share the domain encoded by exon 1B1b, which has been reported to play a role in the shuttling of MITF-A between the nuclear and cytoplasmic compartments (Lu et al., 2010). In TFEB and TFE3, this domain contains the region involved in interactions with Rag GTPases, which is required for their recruitment to lysosomes and subsequent phosphorylation by mTORC1 (Settembre et al., 2012). Indeed, MITF-M is primarily present within the nucleus, suggesting that its nuclear localization is mostly dependent on nuclear localization signals and the nuclear transport machinery.

Using in vitro mutagenesis, we have mapped three clusters of basic amino acid residues in MITF-M that modulate its nuclear localization. Two of the signals are located in the DNA binding region, spanning residues 197-206 and 214-217, and a third signal is present in the HLH-Zip domain, ranging from residue 255 to 265. Sequence alignment indicates that the three potential nuclear localization signals are highly conserved among the MiT/TFE family of transcription factors. Therefore it seems likely that the respective amino acid sequences also have the potential to direct MITF-A as well as members of the TFE family into the nucleus. For c-MYC, another bHLH-Zip transcription factor, a nuclear localization signal has been identified upstream of its basic region (Dang and Lee, 1988). However, a c-MYC mutant lacking this region was shown to remain partially nuclear, also suggesting the presence of additional nuclear localization signals in the protein. Indeed, residues that correspond to R214-217 in MITF-M have been reported to play a role in the nuclear transport of c-MYC as well (Dang and Lee, 1988). Altogether, these findings suggest that proteins of the bHLH-Zip family comprise several nuclear localization signals, ensuring that they are shuttled into the nucleus when needed.

Structural characterization of MITF has revealed that the three clusters of karyophilic amino acid residues identified in this study are exposed for interactions in the absence of DNA. This strongly implies that proteins of the nuclear import machinery and other regulatory factors directly interact with these residues. Indeed, importin beta 1 (KPNB1), nucleoporin 153 (NPM153) and transportin 1 (TNPO1) have been identified as interaction partners of MITF in 501MEL cells by using a mass spectrometry approach (Laurette et al., 2015). Furthermore, importins 7 and 8 were shown to be involved in the nuclear transport of MITF-H in pancreatic cancer cells (Perera et al., 2015). We have used siRNAs against all of the above-mentioned proteins to determine whether any of these factors are involved in the nuclear import of endogenous MITF in 501MEL cells. Although the mRNAs for the individual import proteins were effectively knocked down, we did not observe any effects on the nuclear localization of MITF (data not shown). This might be explained by a redundancy in nuclear import pathways, with several importins, multiple nuclear localization signals and post-translational modification events being at play. Alternatively, the nuclear transport proteins may be too stable for short-term knockdown experiments to reveal noticeable differences. In addition to various importins, exportin 1 has been reported to interact with MITF (Laurette et al., 2015). However, the regions in the protein important for its nuclear export remain to be identified. Despite the presence of a leucine-rich nuclear export signal consensus sequence (LxxxLxxLxL) between residues 280 and 290 (la Cour et al., 2003), mutation of these amino acids revealed no effect on the subcellular localization of MITF-M (data not shown). By using truncated MITF-M mutants, we observed that deletion of the first 120 residues of MITF-M resulted in an almost exclusive nuclear localization, whereas deleting the first 70 residues did not affect nucleocytoplasmic shuttling of the mutant protein. It is thus tempting to speculate that the region between amino acid residues 70 and 120 plays a role in the nuclear export of MITF-M. The nuclear transport mechanisms of MITF-M certainly deserve further attention.

Several mutations in MITF have been reported to affect the transcription factor’s subcellular localization. For example, the R217del mutation is found in WS2A and TS patients as well as in mice carrying the original Mitf^mi mutation (Leger et al., 2012; Steingrimsson et al., 2004; Tassabehji et al., 1994). Previous work has suggested that this single amino acid deletion severely disrupts the nuclear localization of MITF (Takebayashi et al., 1996). In osteoclasts and other cell types, however, this mutant is predominantly nuclear (Bronisz et al., 2006; Grill et al., 2013; Hershey and Fisher, 2004), suggesting that the subcellular localization of R217del depends on the cellular context. Our results indicate that deletion of R217 leads to an enrichment of MITF-M in the cytoplasm of both HEK293T and 501MEL cells. While confocal images indicate that the cytoplasmic concentration of R217del is intermediate between wild type and R214-217A mutant MITF-M, subcellular fractionations suggest that there is little difference between the R217del and R214-217A mutants. Of note, similar inconsistencies were observed for the Ndel220 construct. In this context it is worth mentioning that detergent-based fractionation can lead to significant leakage of nuclear components into the cytosolic fraction (Petrulis et al., 2000). Regarding the impaired DNA binding potential of both R217del and Ndel220 mutants, nuclear leakage of the two proteins during extract isolation might explain the increased cytoplasmic enrichment seen on Western blots. Taken together, R217 forms part of a nuclear localization signal in MITF-M and deletion of this residue partially interferes with the transcription factor’s import into the nucleus. The Mitf^mi-ew mouse mutation, which affects splicing of Mitf such that the protein lacks residues 187-212, is another example of a mutation that has been shown to affect the nuclear import of MITF. Takebayashi et al. reported that around 40% of the mutant MITF^ew protein localized to the cytoplasmic compartment of HEK293T cells (Takebayashi et al., 1996). Moreover, mutation of K206 has recently been described in a patient with COMMAD syndrome, a complex disorder exhibited upon compound heterozygosity of two semi-dominant WS2A or TS mutations, and was shown to negatively affect the nuclear localization of MITF (George et al., 2016). Both observations are consistent with the presence of a karyophilic amino acid sequence upstream of the basic domain, spanning residues 197-206. Finally, the R259X truncating mutation, which is associated with WS2A, has been reported to interfere with the nuclear localization of MITF (Grill et al., 2013). This is in agreement with our findings demonstrating that residues 255-265 form yet another potential nuclear localization signal.

It is interesting to note that several lysine residues located within the two karyophilic regions involving residues 197-206 and 255-265, namely K205, K206 and K265, are predicted acetylation sites (Hou et al., 2014). Acetylation has been shown to modulate the nuclear localization and/or retention of various transcription factors, including HNF4 and NRF2 (Kawai et al., 2011; Soutoglou et al., 2000). It is thus tempting to speculate that acetylation might also play a central role in defining the subcellular localization of MITF. Another important point to address is whether nuclear retention of MITF requires DNA binding. The four consecutive arginines 214-217 directly interact with DNA (Pogenberg et al., 2012), suggesting that if DNA binding is impaired, the protein may not be able to stay in the nucleus. However, R217del MITF-M fails to bind DNA (Grill et al., 2013) and still partly localizes to the nucleus of HEK293T and 501MEL cells, thus arguing against this notion. In agreement with this finding, Ndel220 mutant MITF-M, which lacks the whole DNA binding region, shows a prominent nuclear signal on confocal images. The potential nuclear localization signal we have mapped between residues 255 and 265 is not directly involved in DNA binding although it may play a role in dimerization. Of note, various dimerization-deficient MITF-M mutants, including Q258X, N278X and Zip mut 3, are highly abundant within the nucleus. Therefore, dimerization does not seem to be required for the nuclear transport of the transcription factor. Interestingly, all dimerization-deficient MITF-M mutants used in this study exhibit a significantly reduced stability in 501MEL, SKMEL28 and LU1205 melanoma cell lines, whereas no such effect was observed in HEK293T cells. Importantly, all three melanoma cell lines carry the BRAF(V600E) mutation, strongly suggesting that altered signaling mediates the effects on MITF turnover. In this context, KIT ligand-induced phosphorylation of MITF at S73 has been reported to modulate the transcription factor’s half-life in the melanocytic lineage (Wu et al., 2000). However, abolishing phosphorylation of S73 showed no effects on the stability of MITF-M in 501MEL cells. In line with this observation, inhibition of BRAF signaling did not influence the levels of MITF in this particular cell line. It is interesting to note that the upper band intensities of dimerization-deficient Q258X, N278X and zipper mutants 3, 5 and 6 are strongly reduced when compared to the wild type protein. This might indicate that MITF is preferentially phosphorylated upon dimerization, which in turn might affect its stability. The pathways that regulate MITF stability require further investigation.

Given the importance of MITF in melanoma as well as Waardenburg and Tietz syndromes, proper characterization of the protein, including its various isoforms, is required for advancing our understanding of the nature of these diseases. In this study we have established that in addition to its well-known role in DNA binding and dimer formation, the bHLH-Zip domain of MITF modulates the transcription factor’s subcellular localization and stability.

Significance.

Given the central role of MITF in melanoma and pigmentation disorders, proper characterization of the protein is required for advancing our understanding of the nature of these diseases. Here, we report three conserved karyophilic signals in MITF, which orchestrate the transcription factor’s nuclear localization. Importantly, variants in a number of the residues involved have been associated with Waardenburg syndrome type 2A and Tietz syndrome. Furthermore, we show that dimerization is crucial for MITF stability in cells of the melanocytic lineage. Together, these findings provide novel insights into possible means of targeting MITF localization and protein stability for therapeutic intervention.