In Vivo Role of Alternative Splicing and Serine Phosphorylation of the Microphthalmia-Associated Transcription Factor
Julien Debbache, M. Raza Zaidi, Sean Davis, Theresa Guo, Keren Bismuth, Xin Wang, Susan Skuntz, Dragan Maric, James Pickel, Paul Meltzer, Glenn Merlino, Heinz Arnheiter


The microphthalmia-associated transcription factor (MITF) is a basic helix-loop-helix leucine zipper protein that plays major roles in the development and physiology of vertebrate melanocytes and melanoma cells. It is regulated by post-translational modifications, including phosphorylation at serine 73, which based on in vitro experiments imparts on MITF an increased transcriptional activity paired with a decreased stability. Serine 73 is encoded by the alternatively spliced exon 2B, which is preferentially skipped in mice carrying a targeted serine-73-to-alanine mutation. Here, we measured the relative abundance of exon 2B+ and exon 2B RNAs in freshly isolated and FACS-sorted wild-type melanoblasts and melanocytes and generated a series of knock-in mice allowing forced incorporation of either alanine, aspartate, or wild-type serine at position 73. None of these knock-in alleles, however, creates a striking pigmentation phenotype on its own, but differences between them can be revealed either by a general reduction of Mitf transcript levels or in heteroallelic combinations with extant Mitf mutations. In fact, compared with straight serine-73 knock-in mice with their relative reduction of 2B+ Mitf, forced incorporation of alanine 73 leads to greater increases in MITF protein levels, melanoblast and melanocyte numbers, and extent of pigmentation in particular allelic combinations. These results underscore, in vivo, the importance of the link between alternative splicing and post-translational modifications and may bear on the recent observation that exon 2B skipping can be found in metastatic melanoma.

ALTERNATIVE splicing and post-translational modifications are among the major mechanisms by which individual genes generate multiple protein products. In fact, modern sequencing technologies and proteomic analyses have shown that most if not all multiexon pre-mRNAs give rise to alternatively spliced mature mRNAs (Nilsen and Graveley 2010) and that post-translational modifications modulate the activities of most eukaryotic proteins (Mann and Jensen 2003). In a simple sense, then, the two mechanisms are linked as post-translational modifications depend on the presence of the particular exons encoding the modifiable residues or proteolytic cleavage sites. Here, we present a mouse model that allows us to probe this link in vivo by separately targeting a splice site and a codon for a biologically relevant phosphoacceptor site in the gene encoding the pigment cell transcription factor, microphthalmia-associated transcription factor (MITF). In fact, pigmentation is particularly suitable for such studies as it provides for an easily visible and highly sensitive readout of gene function.

Mitf plays a crucial role in the development and function of melanin-bearing pigment cells in skin, eye, and inner organs (Hodgkinson et al. 1993; Arnheiter et al. 2006; Arnheiter 2010). Its protein product is a basic helix-loop-helix leucine zipper transcription factor that regulates target gene promoters by binding specific E box sequences as homodimers or heterodimers with the related proteins TFEB, TFE3, and TFEC (Strub et al. 2011). Mitf mutations have been found from zebrafish to man and their analyses have established that Mitf controls specification, proliferation, death, differentiation, and stress responses of normal pigment cells and regulates migration, metastasis, and death of melanoma cells (Arnheiter et al. 2006; Hoek and Goding 2010).

In vitro experiments have shown that phosphorylation of MITF serine 73 increases the protein’s transcriptional activity about threefold (Hemesath et al. 1998) and that serine-73 phosphorylation alone or double serine-73/serine-409 phosphorylation leads to polyubiquitination and degradation of MITF (Wu et al. 2000; Xu et al. 2000). To address the question of whether serine-73 phosphorylation plays a role in vivo, we have previously generated targeted mice in which codon 73 was changed to one encoding alanine 73, which cannot be phosphorylated (Bismuth et al. 2008). These mice had a normally pigmented coat and normal eyes, a finding that was later confirmed by using a transgenic bacterial artificial chromosome (BAC) rescue strategy with BACs engineered to encode alanine 73 alone or alanine 73 and alanine 409 together (Bauer et al. 2009). A complication of these experiments was the fact that the Mitf exon, which encodes serine 73, exon 2, contains a 5′ alternative splice site that divides it into an exon 2A of 20 codons and an exon 2B of 56 codons, the latter including codon 73. This arrangement normally leads to a major transcript (at least 90% of total Mitf RNA in skin or heart) that contains the entire exon 2, and a minor one in which exon 2B, and hence codon 73, is missing, although the open reading frame is maintained (Bismuth et al. 2008; Bauer et al. 2009). Interestingly, the codon-73 mutation in both targeted mice and BAC transgenics is associated with a substantial decrease of the transcripts containing exon 2B (representing only ∼10% of total Mitf mRNA in the targeted mice and ∼45% in the BAC rescue mice), and a concomitant increase of the transcripts lacking exon 2B (Bismuth et al. 2008; Bauer et al. 2009). This splice change, which is likely due to the fact that codon 73 is embedded in a splice enhancer sequence whose affinity for the serine/arginine-rich splice regulator SRSF5 is reduced upon mutation (Wang et al. 2009), precluded meaningful conclusions on the specific role of serine 73. To separate the potential effects of mutations at residue 73 from the splice-dependent absence of this residue, we retargeted Mitf in a way that renders the alternative 5′ splice site nonfunctional and incorporates either a wild-type serine, a phosphomimetic aspartate, or an alanine at position 73. We find that although the new mutations cause no visible coat phenotypes on their own, they sensitize melanocytes to the phenotypic effects of reductions in total Mitf levels and to the effects of extant Mitf alleles in compound heterozygotes. Intriguingly, when tested under such conditions, the targeted allele containing the wild-type serine behaves as a hypomorph, while that containing the alanine behaves as a hypermorph. The results show that alternative splicing of exon 2B is indeed relevant in vivo, although in an unexpected manner and in a way that would be difficult to detect solely on the basis of in vitro experiments.

Material and Methods

Melanoblast/melanocyte purification and RNA-Seq data

Melanocytes were purified by FACS sorting from doxycycline-treated bitransgenic mice (iDct-GFP mice, strain FVB/N) that express H2B–GFP fusion proteins specifically in Dct-positive melanocytes, using previously published protocols (Zaidi et al. 2011a,b,c). To purify melanoblasts from E15.5 and E17.5 C57BL/6 iDct-GFP embryos and P1 pups, the dams were fed doxycycline-fortified chow for the entire duration of gestation until harvest. For P3 and P7 pups, doxycycline was injected intraperitoneally at 80 μg/g body weight 24 hr before harvest. Total RNA was prepared from FACS-sorted melanoblasts/melanocytes according to standard Illumina RNA-Seq paired-end protocol and sequenced on the Illumina GAIIx to 80 bp per read. For analysis of alternative exon 2B splicing, a set of possible splice sites was determined and assigned to the aligner, indicating that a given splice variant might exist. The sequences were aligned using the genomic short-read nucleotide alignment program (Wu and Nacu 2010) to mouse genome assembly mm9. A custom Python script was then used to count introns that had one boundary at position 97,944,374 on chromosome 6 (5′ end of exon 3). The analysis revealed that only three splice sites were used, one at position 97,943,348, yielding 2B+ Mitf, one at position 97,943,180, yielding 2B Mitf, and a single read for one at position 97,879,927, yielding Mitf with a junction between exon H and exon 3.

Targeting constructs, minigenes, and expression plasmids

The 17.9-kb targeting construct described by Bismuth et al. (2008) was used as template to PCR amplify a 1.2-kb EcoR1/HindIII-flanked fragment encompassing part of intron 1, exon 2A/2B, and part of intron 2/Neo-loxP cassette. The PCR-amplified fragment was cloned into pcDNA 3.1 and used for PCR mutagenesis. In a first round, the exon 2A/2B junction was altered by changing four bases without changing the coding sequence. In a second round, the previously introduced ApaLI site was changed back to the wild-type sequence and codon A73 was either left intact or was changed into one encoding either aspartate or wild-type serine. Sequence-confirmed clones for each of the three different codons at residue 73 were digested with AflII and AgeI and used to replace the corresponding AflII/AgeI fragment in a 7.5-kb BamHI fragment representing the 3′ arm of the original construct. The resulting constructs were digested with BamHI and ligated to a BamHI-linearized construct containing the thymidine kinase gene and the 5′ arm of the original construct, yielding three final reengineered targeting constructs. The correct orientation of the 7.5-kb BamHI fragment was verified by PCR with two primer pairs described in Supporting Information, Table S1. To test for the effect of the 5′ splice junction mutation, an Mitf minigene comprising exon 1M, intron 1, exon 2A/B, intron 2, and exon 3 was subjected to PCR mutagenesis to alter the 5′ splice junctions and codon 73 as in the targeting contructs, using the Quickchange Site-Directed Mutagenesis kit (Stratagene). For expression analysis, N-terminal estrogen-receptor (ER)-coupled Mitf 6a+ pBABE vectors were obtained from Colin Goding (Carreira et al. 2005) and subjected to PCR mutagenesis to obtain MitfS73A (ER S73A) and MitfΔ2B (ER 2B) expression plasmids, using a primer set described in Table S1. All constructs were sequence verified.


Mice carrying the allele Mitftm1.2Arnh, containing a S73A mutation with or without a floxed Neo cassette in intron 2, have been described (Bismuth et al. 2008). For the purpose of this study, Mitfmi-S73A mice from a separate targeting experiment were used and are designated Mitftm7Arnh (containing the neomycin resistance cassette) and Mitftm7.1Arnh (lacking the neomycin resistance cassette) (MGI:5050706). For targeting the exon 2A/2B 5′ splice site in conjunction with mutations at serine 73, LC3 embryonic stem (ES) cells [genotype (C57BL/6Nx129S6)F1] were electroporated with 20 µg of the respective NotI linearized targeting constructs and grown under standard G418/FIAU selection protocols. After screening, targeted ES cells were injected into C57BL/6N blastocysts and chimeric mice were bred with C57BL/6J to test for germ line transmission and to establish Neo-cassette–containing targeted lines. To remove the floxed Neo cassette, mice were bred with C57BL/6•129S4-Prm1-Cre deleter mice. After removal of the Neo casette, mice were either intercrossed to obtain homozygous targeted mice or crossed with mice carrying other mutant Mitf alleles. The official designation of these targeted mice are: Mitftm4Arnh (MGI:5050700) and Mitftm4.1Arnh (MGI:5050701) for the splice junction/S73 allele; Mitftm5Arnh (MGI:5050702) and Mitftm5.1Arnh (MGI:5050703) for the splice junction/S73A allele; and Mitftm6Arnh (MGI:5050704) and Mitftm6.1Arnh (MGI:5050705) for the splice junction/S73D allele, each with or without the Neo cassette, respectively. C57BL/6•129S2 Kittm1Alf (Bernex et al. 1996) heterozygous males were bred with Mitfmi-S73AΔneo and Mitfmi-S-S73AΔneo homozygous females. Double heterozygous offspring were then crossed with the corresponding Mitf homozygotes to establish E12.5 and E15.5 embryos and P1 pups homozygous for Mitf and heterozygous for Kittm1Alf. X-gal–labeling was done as described (Wen et al. 2010). For generation of melanocyte lines, mice were crossed with B6•Cdkn2atm1Rdp mice to obtain double homozygous Mitf/Cdkn2a mutants. The bitransgenic iDct-GFP mice (Zaidi et al. 2011a) used for RNA-Seq were kept on a FVB/N background.

Reverse transcription, RT–PCR, and real-time (q)RT–PCR, Western blots

Random primed cDNAs were made from 500 ng of total RNA extracted from heart or 50 ng extracted from melanocyte lines using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems) and RNAse inhibitor cocktail (Invitrogen). PCR and qPCR were performed as described (Bharti et al. 2010), using primers described in Table S1. For Western blots, 600 µg of heart tissue or 120,000 cultured cells were separated by SDS–PAGE and blotted onto nitrocellulose membranes. The blots were exposed to appropriate antibodies and bands were revealed using SuperSignal West Pico chemiluminiscent substrate. Quantitation was done using a Biorad ChemiDoc XRS+ system.

Genotyping and Southern analysis

DNA extracted by standard procedures from tissue samples was analyzed by PCR, using primers indicated in Table S1. For Southern blots, liver DNA was prepared, digested as indicated, and probed using PCR-amplified probes as previously published (Bismuth et al. 2008).


Photographs of knock-in mice and their intercrosses with other Mitf alleles were taken between 3 weeks and 2 months of age.

Cell lines stably expressing Mitf proteins

pBABE-ER vectors allowing expression of HA-tagged exon 6A+ MITF fused with the estrogen-responsive portion of the estrogen receptor were obtained from Colin Goding (Carreira et al. 2005). Virus was produced by cotransfection of pBABE vectors and pCL Ampho vectors in 293T cells, and ARPE-19 cells (spontaneously immortalized human retinal pigment epithelium cells) were infected and selected with puromycin (1 μg/ml) without subsequent cloning. Stable cell lines were prepared allowing expression of only ER (ER Ø) or MITF-ER with wild-type S73 (ER WT), S73A (ER S73A), or 2B MITF (ER 2B). For induction, cells were treated with 300 nM 4-hydroxy-tamoxifen (TM) for 48 hr. Total extracts were probed with 6A5 anti-MITF antibodies (Bharti et al. 2008).

Flow cytometry

For assaying MITF expression, ARPE19 cells were washed and fixed with 4% paraformaldehyde, permeabilized, and incubated with mouse monoclonal antibody 6A5 and R-phycoerythrin-goat antimouse IgG1 as second antibody. Flow cytometry was performed using a FACSVantage SE flow cytometer (Becton Dickinson) and the data were processed using CellQuest software. The gating was determined empirically and applied in the same way to all experiments.


Alternative splicing of exon 2B of Mitf in melanocytes

Previous results have shown that heart and skin of wild-type mice express Mitf splice variants that specifically lack exon 2B (Hallsson et al. 2000; Bismuth et al. 2008; Bauer et al. 2009). To confirm these findings in freshly isolated embryonic and postnatal melanocytes, we purified the respective cells by FACS, using the recently described bitransgenic iDct-GFP mice that allow for tetracyclin-regulated expression of a histone H2B–GFP fusion protein in melanoblasts and melanocytes (Zaidi et al. 2011a,b,c). For these experiments, we used wild-type embryos at E15.5 and E17.5 and postnatal mice at P1, P3, and P7. RNA-Seq of sorted cells clearly showed the presence of Mitf RNA. As shown in Figure 1A, detailed analyses revealed that although in all samples the majority of reads across the 5′ alternative splice junctions of exon 2 corresponded to the exon 2B-exon 3 sequence, a small subset corresponded to the exon 2A-exon 3 sequence. To further confirm these results, we subjected RNA from P1 and P3 FACS-sorted cells to real-time RT–PCR as described (Bharti et al. 2008). The results showed a 2B/(2B + 2B+) ratio of between 1.85 and 4.36%. These values were similar to those in passage 10 melanocytes obtained from P2 mice that were wild-type at Mitf and homozygous for a Cdkn2a targeted null allele used to prevent cell senescence (Sviderskaya et al. 2002) (Figure 1B). In contrast, in corresponding melanocytes from P2 skin of Mitf-S73A mutated, Cdkn2a−/− mice (labeled S73A in Figure 1B), the ratio was 90.75%. These results indicate that freshly isolated embryonic or postnatal melanocytes or cultured melanocytes express a small amount of Mitf mRNAs lacking exon 2B and that a serine-73 mutation substantially increases this alternatively spliced transcript.

Figure 1 

Detection of Mitf isoforms in melanocytes in vivo. (A) GFP+ melanoblasts from bitransgenic (iDct-GFP) embryos and postnatal pups of the indicated ages were isolated and subjected to sorting by FACS. Corresponding RNA-seq data were analyzed as described in Materials and Methods. The ratios of the reads across exon 2A/exon 3 and exon 2B/exon 3 [2B/(2B + 2B+)] as well as the total number of reads analyzed are shown. (B) Relative abundance of Mitf 2B+ and Mitf 2B RNA prepared from in vivo isolated P1 and P3 melanocytes, expressing GFP and FACS-sorted as for A, was determined by qRT–PCR using the appropriate primer pairs (Table S1). In addition, RNA from melanocyte cell lines derived from either Cdkn2a−/−; Mitf+/Mitf+ and Cdkn2a−/−; Mitfmi-S73A/Mitfmi-S73A P1 dorsal skins and cultured for 10 passages served as controls. The calculated percentages correspond to the ratio of 2B/(2B + 2B+).

A targeted mutation in the exon 2A/2B junction of Mitf prevents exclusion of exon 2B in vitro and in vivo

As shown in Figure 2A and Figure S1, the RNA sequence of the alternative 5′ splice site, CACCAG′GUAAAG, is close to the consensus splice sequence, C/AAG′GUA/GAGU. To ensure efficient inclusion of exon 2B despite codon-73 mutations, we deliberately mutated this splice site in genomic DNA to yield RNAs with the sequence CAUCAA′GUUAAA. The mutation was expected to render the 5′ splice junction nonfunctional without, however, changing the encoded amino acid sequence. We first tested whether the mutation would function in the intended way in a minigene construct that comprised exon 1M-exon 3 of Mitf. Upon transfection into heterologous cells, this construct normally recapitulates the alternative exon 2B splicing and its dependence on the sequence of codon 73. When both the 5′ splice junction and codon 73 were mutated, however, only exon 2B+ RNA and no aberrant splice products were produced in vitro (not shown). We then prepared three separate targeting constructs that carried the same splice mutation along with a codon for either wild-type serine, aspartate, or alanine (Figure 2A; for details, see Materials and Methods). After electroporation into embryonic stem cells, three lines of targeted mice designated Mitfmi-S-S73S, Mitfmi-S-S73A, and Mitfmi-S-S73D (S-S for splice site and serine 73) were obtained and characterized by Southern blot analysis (Figure 2, B and C) and by sequencing (not shown). Together with Mitfmi-S73A mice, which do not carry the 5′ splice site mutation, we established eight separate lines, four of them containing a floxed neomycin resistance cassette in intron 2 (see Figure 2A), and four corresponding ones in which this cassette was removed by breeding with Prm1-Cre deleter mice.

Figure 2 

Generation of four knock-in Mitf alleles. (A) Schematic representation of the Mitf genomic structure and the mutations generated in Mitfmi-S73Aneo [previously described (Bismuth et al. 2008) and retargeted for the purpose of this study] and in the three novel splice alleles. Sequence modifications at the exon 2A–2B junction as well as at the S73 codon are highlighted in red, and the position of the neomycin resistance cassette in intron 2 is shown. (B and C) Differential restriction profile of the Mitf locus before and after targeting and Southern blot analysis of wild-type and one line each of homozygous targeted mice. DNA was digested as indicated and probed with the respective 5′ and 3′ probes shown in B.

For RNA analyses, we first used heart tissue because heart expresses relatively high levels of Mitf RNA and, unlike melanocytes, is not altered in a major way by Mitf mutations. RT–PCR analyses showed the presence of both exon 2B+ and exon 2B RNA in wild-type mice; the predominant lack of exon 2B in mice homozygous for the Mitfmi-S73AΔneo allele; and the exclusive presence of exon 2B in hearts of mice homozygous for either of the three 5′ splice-mutant alleles [regular RT–PCR in Figure S2A, qRT–PCR (Bharti et al. 2010) in Figure S2B]. Total levels of Mitf RNAs, averaged from quantitative RT–PCR analyses using primers for exons 6-7 and 8-9 (Table S1), were similar to wild type in the mutants in which the Neo cassette has been removed (Figure S2C). To determine RNA levels in mice in which the Neo cassette was left intact, we used primers specifically covering the mutated exon 2 region to avoid measuring aberrant RNAs potentially starting from the Neo cassette. The results showed that when the Neo cassette was present, Mitf RNAs containing part or all of exon 2 were reduced by ∼50–75% (Figure S2D). Immunoprecipitation/immunoblotting assays of hearts of the different mutants indicated that Mitfmi-S73AΔneo mice produced MITF protein with a faster electrophoretic mobility, consistent with the predominant expression of MITF lacking exon 2B in such mice (Figure S2E). In contrast, the different S-S73 mutants all produced predominantly full-length MITF protein, but their relative expression levels differed. Compared to Mitfmi-S-S73SΔneo mice, the levels were ∼1.3 fold higher in Mitfmi-S-S73DΔneo and ∼1.6 fold higher in Mitfmi-S-S73AΔneo mice, suggesting that the particular residue at position 73 influences protein stability (Figure S2E).

To determine RNA and protein expression specifically in melanocytes, we then tested above described Mitf+/Mitf+; Cdkn2a−/− and Mitfmi-S73AΔneo/Mitfmi-S73AΔneo; Cdkn2a−/− melanocyte lines as well as a corresponding line derived from P2 skin of Mitfmi-S-S73AΔneo/Mitfmi-S-S73AΔneo; Cdkn2a−/− mice. RT–PCR assays showed a predominent band representing the Mitf 2B+ transcript in wild type, a double band representing the Mitf 2B+ and Mitf 2B transcripts in Mitfmi-S73AΔneo, and a single band representing the Mitf 2B+ transcript in Mitfmi-S-S73AΔneo (Figure 3A). As schematically depicted in Figure 3B, MITF protein can undergo numerous post-translational modifications in melanocytes and other cell types, including phosphorylations, sumoylations, and caspase cleavage. These modifications may lead to proteins with distinct electrophoretic mobilities on Western blots. In fact, as shown in Figure 3C, wild-type melanocytes showed the characteristic double band of MITF protein representing S73-phosphorylated and S73-nonphosphorylated exon 2B+ protein. Mitfmi-S73AΔneo lines, however, showed a band with an electrophoretic mobility corresponding to nonphosphorylated MITF 2B+ and one corresponding to MITF 2B. In contrast to the relative amounts of the respective transcripts, however, the MITF 2B+ protein band was more prominent than the MITF 2B protein band. This observation suggests that compared to the wild-type full-length protein, either nonphosphorylated full-length MITF is more stable or MITF lacking exon 2B less stable, provided exon 2B+ and exon 2B mRNA share equal translational efficiency. In addition, Mitfmi-S73AΔneo lines showed one band of lower and one of higher electrophoretic mobility (marked with arrows in Figure 3C) that may represent differently modified or cleaved products. Mitfmi-S-S73AΔneo lines showed a single RT–PCR product corresponding to the Mitf 2B+ transcript and a simpler Western blot signal with a prominent band corresponding to nonphosphorylated MITF 2B+ protein and one band each of higher and lower electrophoretic mobility (arrows), again potentially representing differently modified or cleaved products. Quantitation of the Western blot bands revealed that Mitfmi-S73AΔneo and Mitfmi-S-S73AΔneo cells had higher MITF levels than the corresponding wild-type cells (Figure 3C), consistent with the notion that lack of phosphorylation at residue 73 may increase accumulation of full-length MITF because of increased protein stability.

Figure 3 

Mitf exon 2B alternative splicing profile at the RNA and protein level in melanocyte lines. (A) RNA extracts were prepared from passage 10 Cdkn2a−/− melanocyte lines from wild-type B6, Mitfmi-S73AΔNeo (S73A), and Mitfmi-S-S73AΔNeo (S-S73A) mice and subjected to RT–PCR. (B) Schematic diagram of select post-translational modifications of MITF seen in melanocytes and other cell types. Indicated are phosphorylation, sumoylation, and caspase cleavage sites, while sites for acetylation, ubiquitination, and other modifications are not shown. (C) Western blots of protein extracts of the cell lines used in A, probed with 6A5 anti-MITF antibodies. The double band seen in the wild-type sample corresponds to phosphorylated and nonphosphorylated exon 2B+ MITF. Mitfmi-S73A cells show both a minor band corresponding to 2B MITF and a major band corresponding to 2B+ MITF, in addition to a band of higher molecular weight and one of lower molecular weight, likely representing another post-translationally modified form and a degradation product (arrows). Mitfmi-S-S73A cells show a predominant 2B+ MITF band and no 2B MITF band. Also, there is a band of higher molecular weight and one of lower molecular weight (arrows). MITF quantification relative to the endogenous β-tubulin is indicated below the gel.

Genetic dissection of pigmentation phenotypes associated with the targeted mutations

To determine the phenotypic consequences of the above mutations, we performed extensive breeding tests. First, we generated mice containing the neomycin resistance cassette that were either homozygous for the four targeted Mitf alleles or were compound heterozygotes with any of three extant Mitf alleles: Mitfmi-vga9, which is a null allele due to insertion of a transgene array in the M-Mitf promoter (Hodgkinson et al. 1993); MitfMi-wh, a semidominant allele characterized by an Ile212-to-Asn mutation in the DNA-binding basic domain (Steingrímsson et al. 1994); and Mitfmi, a semidominant allele characterized by the lack of an Arg in a row of four Arg in the basic domain (Steingrímsson et al. 1994). The targeted homozygotes showed normal eyes and normal pigmentation in coat, ears, and feet except for a white belly spot, which on average was largest in Mitfmi-S-S73SNeo but almost always absent in Mitfmi-S-S73ANeo mice (Figure 4, A–D). Similar phenotypic differences between the targeted alleles were seen in compound heterozygotes with Mitfmi-vga9. While Mitfmi-vga9/Mitf+ mice are normally pigmented (Figure 4E) and Mitfmi-vga9/Mitfmi-vga9 mice totally white and microphthalmic (Figure 4H), the compound heterozygotes showed various extents of white spotting whereby Mitfmi-S-S73ANeo/Mitfmi-vga9 mice had the smallest areas of white coats; Mitfmi-S-S73SNeo/Mitfmi-vga9 mice were almost completely white; and Mitfmi-S73ANeo/Mitfmi-vga9 and Mitfmi-S-S73DNeo/Mitfmi-vga9 mice were intermediate (Figure 4, F, G, I, and J). In contrast, no such differences between the targeted alleles were seen in combination with MitfMi-wh as all mice had a fawn coat with white spots of similar sizes (Figure S3A). Only rarely did we observe small black spots in these mice, which may reflect somatic loss-of-function mutations in the dominant-negative MitfMi-wh allele (an example shown in Figure S3A for Mitfmi-S-S73ANeo/MitfMi-wh). We also saw no major codon-73–dependent differences in combinations with Mitfmi as all mice were totally white except for small black spots that appeared in highest frequency and sizes in Mitfmi-S-S73ANeo/Mitfmi mice but were always absent in Mitfmi-S-S73SNeo/Mitfmi mice (Figure S3B). These results indicate that in alleles containing the neomycin resistance cassette, the presence or absence of exon 2B and the phosphorylatability or permanent negative charge of residue 73 are associated with phenotypic differences that are, however, only seen in particular allelic combinations.

Figure 4 

Phenotypes associated with the four knock-in alleles alone and in combination with extant Mitf alleles. (A–J) Controls and Neo-cassette–containing lines. (A–D) White belly spots found in homozygotes of the indicated genotypes. Note that Mitfmi-S-S73A homozygotes show no belly spots, in contrast to the other lines. (E) Mitfmi-vga9 heterozygotes show no pigmentary phenotype while Mitfmi-vga9 homozygotes are completely white and microphthalmic (H). (F, G, I, and J) Heteroallelic combinations with Mitfmi-vga9. Note different degrees of white spotting with the different targeted alleles. (K–R) Lines lacking the Neo cassette (Δneo). (K–N) Regardless of the genotype, all homozygotes are normally pigmented and indistinguishable by visual inspection from wild-type B6 mice, except that Mitfmi-S73AΔneo/Mitfmi-S73AΔneo mice have a tail with generally lighter pigmentation compared to the other genotypes. (O–R) Heteroallelic combinations with MitfMi-wh. Note the darker pigmentation seen in combinations with either Mitfmi-S73AΔneo or Mitfmi-S-S73AΔneo compared to combinations with Mitfmi-S-S73DΔneo, which yield mice with a coat that was only slightly darker, and combinations with Mitfmi-S-S73SΔneo, which yield mice that are indistinguishable from MitfMi-wh/Mitf+ heterozygotes. Each photograph in O–R represents littermates.

In a second round of breedings, we used mice lacking the neomycin cassette to generate targeted homozygotes or compound heterozygotes as above. As shown in Figure 4, K–N, homozygotes were all normally pigmented regardless of the splice and serine-73 mutations. In contrast, phenotypic differences between the different targeted Mitf alleles were seen in combination with MitfMi-wh (Figure 4, O–R). Compared with MitfMi-wh/Mitf+ littermates, both Mitfmi-S73AΔneo/MitfMi-wh and Mitfmi-S-S73AΔneo/MitfMi-wh mice had a considerably darker coat, yet they still carried white belly spots typical of the presence of the MitfMi-wh allele. However, the coat was only slightly darker in Mitfmi-S73DΔneo/MitfMi-wh mice, and Mitfmi-S73SΔneo/MitfMi-wh mice were phenotypically indistinguishable from MitfMi-wh/Mitf+ mice (Figure 4, O–R). Because it was theoretically possible that the presence of MITF protein that cannot be phosphorylated at residue 73 might influence the phosphorylation status of the MITFMi-wh protein and so influence the stability and dominant-negative activity of the latter, we also performed immunoprecipitation/immunoblot assays of heart tissues of the different allelic combinations. As shown in Figure S2F, however, there was no clear indication for a change in phospho-MITFMi-wh as the bands in Mitfmi-S73AΔneo/MitfMi-wh hearts (lane 5) appeared as the sum of those seen in Mitfmi-S73AΔneo/Mitfmi-S73AΔneo (lane 3) and MitfMi-wh/ MitfMi-wh (lane 8) hearts. Furthermore, none of the targeted alleles showed any differences in combinations with Mitfmi-vga9 or Mitfmi (Figure S3, C and D). These results again indicate that phenotypic differences between the different targeted alleles can be revealed only in particular allelic combinations.

Melanoblast and melanocyte accumulation is differently affected by Mitfmi-S73A and Mitfmi-S-S73A

The darker coat associated with the presence of the serine-73-to-alanine mutated MITF protein in MitfMi-wh compound heterozygotes could be explained by the accumulation of a greater number of melanoblasts and melanocytes in the corresponding mice. The availability of Kittm1Alf, a Kit null allele characterized by the insertion of a bacterial lacZ gene in exon 1 of Kit, allowed us to specifically and reliably label melanoblasts during development. In fact, previous results have shown that the nuclear β-GAL produced by Kittm1Alf is not affected by Mitf mutations, at least not in melanoblasts (Hou et al. 2000). It needs to be kept in mind, however, that Kittm1Alf/Kit+ mice have white feet, tail tips, and belly spots as only two fully functional Kit alleles allow for completely normal melanoblast development. Hence, the use of Kittm1Alf gave us the added opportunity to compare the phenotypic effects of the different targeted Mitf alleles in conjunction with a separate mutation affecting melanogenesis. Intercrosses showed, however, that homozygosity for Mitfmi-S73A and Mitfmi-S-S73A did not grossly change the Kittm1Alf/Kit+ phenotype except that on average, the belly spots seemed slightly larger with both Mitf alleles. We then harvested corresponding E12.5 and E15.5 embryos as well as P1 skin and subjected the specimens to X-gal labeling (Figure 5). At E12.5, in a representative area around the developing eye, Kittm1Alf/Kit+; Mitfmi-S-S73A/Mitfmi-S-S73A embryos showed similar numbers of X-gal–labeled cells as Kittm1Alf/Kit+; Mitfmi-S73A/Mitfmi-S73A embryos. At E15.5 and P1, however, representative areas showed slightly, but significantly higher numbers of labeled cells in Kittm1Alf/Kit+; Mitfmi-S-S73A/Mitfmi-S-S73A compared to Kittm1Alf/Kit+; Mitfmi-S73A/Mitfmi-S73A mice. This was also reflected in the number of labeled cells in pigmented hair follicles in a representative dorsal area between the forelimbs. The results suggest that the absence of the entire exon 2B, and the selective absence of a phosphorylatable residue at position 73, affect the accumulation of melanoblasts and melanocytes differentially. The findings are consistent with the facts that the presence of the S-S73A allele in some allelic combinations leads to decreased sizes of white spots (or increased numbers of pigmented spots) compared to mice carrying the S73A allele (see Figure 4, A, B, F, and G; Figure S3B).

Figure 5 

Melanocyte numbers in developing embryos. Embryos homozygous for either Mitfmi-S73AΔNeo or Mitfmi-S-S73AΔNeo and heterozygous for Kittm1Alf were X-gal–labeled and blue cells counted in selected regions at E12.5 and E15.5 and in P1 skin and P1 pigmented hair follicles. Numbers from 7 to 10 fields from at least three embryos are shown relative to those observed in Kittm1Alf/Kit+ control samples. Statistical significance (two-tailed unpaired t test) is indicated (*P < 0.1; **P < 0.01; ***P < 0.001).

Because Mitfmi-S73A is a complex allele, giving rise to both exon 2B as well as exon 2B+ protein (see Figure 3C), we finally assessed what the separate effects of each of these two isoforms might be in cells in culture. For this, we used ARPE19 cells, which are spontaneously immortalized human retinal pigment epithelial cells that are low in endogenous MITF and yet are derived from a pigmented cell type, and infected them with retroviral vectors expressing either (wild type) MITF 2B+ (corresponding to MITF derived from the Mitfmi-S-S73S allele), MITFS73A, or MITFΔ2B proteins, each fused with the estrogen-responsive portion of the ER (termed ER WT, ER S73A, and ER 2B in Figure 6). In such cells, MITF proteins normally are found in the cytoplasm at relatively low levels but accumulate at high levels in nuclei after tamoxifen induction. In fact, Western blots revealed the expected MITF bands and electrophoretic mobilities after tamoxifen induction (Figure 6A). The cells were then subjected to flow cytometry, using two-color fluorescence labeling for MITF protein expression and total DNA content (DAPI) for cell cycle analysis. As shown in Figure 6B, 48 hr of tamoxifen induction increased the percentage of gated cells showing above-threshold levels of MITF protein. Intriguingly, wild-type MITF-expressing cells showed distinct MITFlow and MITFhigh populations, whereas MITFS73A- and MITFΔ2B-expressing cells had a distinctly lower MITFlow population. This suggested that the MITFlow population was generated by reduction of MITF by degradation of the S73-phosphorylated form of MITF. As shown in Figure 6C, tamoxifen induction of the different MITF isoforms had differential effects on DNA content of cells. While there was little change when the tag alone and no ectopic MITF was present, wild-type MITFhigh cells showed a decrease in the percentage of cells in S phase and an increase in the percentage of cells in G0/G1 48 hr after tamoxifen induction. In contrast, the percentage of MITFS73A-high and MITFΔ2B-high cells in S phase was increased, rather than decreased, after tamoxifen induction, though less prominently with MITFΔ2B compared to MITFS73A. This in vitro analysis with heterologous cells and overexpressed, cDNA-derived inducible proteins suggested that wild-type MITF has antiproliferative activities but that the lack of exon 2B, or the absence of phosphorylation at residue 73, yield MITF proteins lacking antiproliferative activity. The results confirm earlier findings with an S73A mutated protein analyzed in different cells (Bismuth et al. 2005) and are consistent with the in vivo genetic observations that the relative increased accumulation of nonphosphorylatable MITF protein leads to increased melanoblast/melanocyte numbers, increases in the size and extent of pigmentation in certain heteroallelic combinations, and increased ability to compensate a strong dominant-negative allele, MitfMi-wh.

Figure 6 

Effects of the S73A mutation and exon 2B deletion on protein levels and DNA content in stable cell lines inducibly expressing MITF. ARPE-19 cells expressing either pBABE-ER (ER Ø), MITF wild-type S73 (ER WT), MITF S73A (ER S73A), or MITF 2B (ER 2B) proteins were treated with 4-OH-tamoxifen (TM) for 48 hr before collection, and total extracts were probed with 6A5 anti-MITF antibodies. (A) TM-treated, wild-type MITF expressing cells show the characteristic double band of S73-phosphorylated and nonphosphorylated MITF protein, S73A-MITF expressing cells only one band corresponding to nonphosphorylated MITF, and 2B MITF-expressing cells a single band corresponding to the 2B isoform. (B and C) ER MITF and ER only (ER Ø) expressing ARPE-19 cell lines were incubated with or without TM, harvested after 48 hr, and labeled for MITF and DNA content, using DAPI. (B) Flow cytometric analysis displays the percentage of cells expressing relatively low (L) or high (H) MITF intensity and (C) their relative cell cycle stages based on DNA content. Gating was determined empirically and applied equally for all samples.


Numerous in vitro assays have shown that MITF is regulated by post-translational modifications, which in turn depend on the presence of the exons carrying the modifiable residues. Nevertheless, among the many spontaneous and induced Mitf mutations from zebrafish to man, few have been found so far to affect splicing or a specific post-translationally modifiable residue. This is not surprising given the fact that, for instance, targeted mutations in the serine-73 phosphoacceptor site as described here, along with those used in BAC transgenic rescue strategies (Bauer et al. 2009), produce few if any phenotypes on their own. It is only in conjunction with additional perturbations, such as the reduction in overall Mitf RNA levels or the combination with other Mitf alleles, that we start to see the effects of mutations in serine 73, or, in addition, the subtle differences between a serine-73 mutation and the lack of the entire exon carrying serine 73.

Our analysis of the phenotypes produced by the targeted alleles benefited from the wide dynamic range of pigmentation and from the fact that the targeted mutations were produced in the endogenous gene rather than in BACs. Targeting the endogenous gene allows for a direct comparison between the different lines as their transcript levels are all similar to wild type. In contrast, Mitf BAC transgenic lines show variable RNA expression levels (Bauer et al. 2009) likely due to the fact that the available BACs do not contain the entire 214-kb Mitf gene and so may lack the boundary elements necessary for sufficient insulation from neighboring sequences encountered at random transgene insertion sites. Comparable RNA expression levels were all the more important because the task was to evaluate the effects of mutations that may affect the stability of the mutated proteins. It has been shown in vitro that serine-73–mutated MITF has a similar activity on a test target gene promoter as nonphosphorylated wild-type MITF, but a lower activity when compared to phosphorylated wild-type MITF (Hemesath et al. 1998). In subsequent in vitro experiments, it was reported that S73A-singly mutated MITF or S73A/S409-doubly mutated MITF have an increased stability. Hence, it became important to determine whether there might be similar increases in the stability of mutated MITF protein in vivo and whether a possible reduction in transcriptional activity was compensated for by increased protein stability.

The simplest explanation for all genetic results presented in this article is indeed an increased stability of the mutated proteins. First, we find that both in an unrelated tissue, the heart, and in melanocytes, S73A-mutated MITF accumulates to higher levels than corresponding wild-type MITF. Interestingly, in heart, S73D-mutated MITF accumulates to levels intermediate between those in S73A- and S73S-mutated MITF, reflecting the intermediate phenotype that the Mitfmi-S-S73D allele produces in some allelic combinations. This suggests that a permanent negative charge at residue 73 neither reflects fully the properties of serine-73–phosphorylated MITF nor those of residue-73–nonphosphoryated MITF. Second, homozygotes for the mutations with wild-type transcript levels are all fully pigmented. Third, the two mutations resulting in MITF that cannot be phosphorylated at residue 73 (Mitfmi-S-S73AΔNeo and Mitfmi-S73AΔNeo) are capable of partially compensating the effects of the strongly dominant-negative MitfMi-wh allele. Fourth, when the Mitf RNA expression levels of the Neo-cassette–containing alleles were further reduced in compound heterozygotes with the null mutant Mitfmi-vga-9, MITFS73A-expressing mice, with the lowest RNA levels, produced the most extensive pigmentation, and MITFS73S-expressing mice, with higher RNA levels, the least amount of pigmentation. Fifth, in vitro, both MITFS73A and MITFΔ2B have milder antiproliferative effects than corresponding wild-type MITF, or they may even stimulate cell proliferation given the nonlimiting amounts of growth factors to which cultured cells are normally exposed. Nevertheless, despite this in vitro finding, the number of melanocytes found postnatally in vivo is not substantially different in mutants compared to wild type. This is likely due to the fact that only excess amounts of growth factors such as KIT ligand or endothelin 3 would allow for prolonged survival of supernumerary melanocytes in vivo.

The above observations imply that increased accumulation of the mutated MITF proteins may not only compensate for the reduction of their activities observed in vitro, but may even overcompensate such activity reductions. This underscores the in vivo importance of post-translational modifications that would seem to allow for precise regulation of the activities of MITF. The results presented in this article also imply that MITFS73A and MITFΔ2B have slightly different effects in vivo. Although Mitfmi-S73A can produce both full-length and exon 2B-deleted MITF, and although the full-length MITFS73A protein may be relatively more stable than the internally truncated one, Mitfmi-S73A and Mitfmi-S-S73A mice still differ slightly in melanoblast/melanocyte numbers when they are counted during development and postnatally, or in extent of pigmentation in the adult if RNA levels are reduced below a threshold amount by presence of a Neo cassette in combination with the Mitf null allele Mitfmi-vga9. Although subtle differences in the total levels of MITF may explain these variations, it is equally plausible that exon 2B has additional functions besides simply providing the substrate for serine-73 phosphorylation. In fact, it is conceivable that wild-type MITF, MITFS73A, and MITFΔ2B differentially stimulate their target genes. This may be important when considering allele-specific genetic interactions of Mitfmi-bws, an Mitf allele with an exon 2B splice bias but low RNA levels and a wild-type serine 73 (Wen et al. 2010). It may also be important to explain the recent finding that human MITFΔ2B, when used to rescue pigmentation in zebrafish with a temperature-sensitive mutation in mitfa, leads to increased divisions of differentiated, pigmented cells (Taylor et al. 2011). Moreover, it may become important to explain the recent finding that a melanoma metastasis carries a somatic mutation at the downstream exon 2B splice junction that leads to elimination of exon 2B (Cronin et al. 2009). Although alterations in MITF may be only one among many pathways whose changes allow for differentiated cells or melanoma cells to escape normal control of cell proliferation, it may well be that MITF splice alterations, genomically fixed or dynamically regulated, provide for both efficient and precise modulation of melanocyte or melanoma physiology.


We thank Colin Goding for microphthalmia-associated transcription factor–estrogen receptor constructs, and L. Baweke, the National Institute of Neurological Disorders and Stroke (NINDS) Animal Health and Care Section, and the NINDS sequencing facility for excellent support. This work was supported by the intramural research program of the National Institutes of Health, NINDS, the National Cancer Institute, and the National Institute of Mental Health.


  • Received October 20, 2011.
  • Accepted February 9, 2012.

Literature Cited

View Abstract