Partial Rescue of Ocular Pigment Cells and Structure by Inducible Ectopic Expression of Mitf-M in MITF-Deficient Mice

Purpose Complete deficiency of microphthalmia transcription factor (MITF) in Mitfmi-vga9/mi-vga9 mice is associated with microphthalmia, retinal dysplasia, and albinism. We investigated the ability of dopachrome tautomerase (DCT) promoter-mediated inducible ectopic expression of Mitf-M to rescue these phenotypic abnormalities. Methods A new mouse line was created with doxycycline-inducible ectopic Mitf-M expression on an Mitf-deficient Mitfmi-vga9 background (DMV mouse). Adult DMV mice were phenotypically characterized and tissues were collected for histology, immunohistochemistry, and evaluation of Mitf, pigmentary genes, and retinal pigment epithelium (RPE) gene expression. Results Ectopic Mitf-M expression was specifically induced in the eyes, but was not detected in the skin of DMV mice. Inducible expression of Mitf-M partially rescued the microphthalmia, RPE structure, and pigmentation as well as a subset of the choroidal and iris melanocytes but not cutaneous melanocytes. RPE function and vision were not restored in the DMV mice. Conclusions Ectopic expression of Mitf-M during development of Mitf-deficient mice is capable of partially rescuing ocular and retinal structures and uveal melanocytes. These findings provide novel information about the roles of Mitf isoforms in the development of mouse eyes.

A lthough all melanin-bearing pigment cells of vertebrates come from the neuroectoderm, they can be divided into two principally distinct classes. The retinal pigment epithelium (RPE) cells are derived from the neuroepithelium of the ventral forebrain, and the melanocytes in skin and its appendages and various extracutaneous locations are derived from the neural crest. Nevertheless, the development of both neuroepitheliumand neural crest-derived melanocytes depends on the same gene, microphthalmia-associated transcription factor (Mitf), which encodes a set of distinct isoforms of a helix-loop-helixleucine zipper transcription factor collectively called MITF proteins. 1 During mouse development, Mitf is first expressed around embryonic day 9.5 in precursors of the pigment cells, soon followed by the expression of a gene whose protein product is later involved in melanogenesis, dopachrome tautomerase (Dct). 2 In the RPE, Mitf expression reaches its peak during the following embryonic days but then is progressively reduced. 3,4 Among the neural crest-derived cells, Mitf expression marks the melanocyte precursors called melanoblasts, and Mitf continues to be expressed in their melanocytic derivatives along with Dct, particularly in skin and feather and hair follicles. 2 MITF has many functions in melanocyte development and maintenance, chief of which are the regulation of cell specification, proliferation, and differentiation. In the total absence of functional Mitf, as seen in mice homozygous for the Mitf mi-vga9 allele, mice exhibit microphthalmia and have neither pigmented RPE cells nor pigmented neural crest-derived cells in skin, iris, choroid, inner ear, or heart. 3 Nevertheless, there is a fundamental difference in the way these two types of pigment cells respond to Mitf loss-of-function mutations. Neural crestderived, Dct-expressing melanoblasts can be seen for only a short time before they disappear. 3 In contrast, in the presumptive RPE, the mutant cells hyperproliferate and continue to express Dct. 3 These findings prompted us to test whether inducible ectopic expression of Mitf, if achieved early enough during development and at sufficient levels, might rescue the pigment cells in question, and whether later removal of Mitf might affect the cells once formed. To achieve ectopic expression, we prepared a line of transgenic mice modeled after our previously described transgenic mice in which a Dct promoter-driven reverse tetracycline-controlled transactivator (Dct-rtTA) activates GFP expression under the control of a tetracycline-responsive element (TRE). 5 This system allows for targeting GFP expression to both neural crest-derived melanoblasts/melanocytes and RPE cells during all stages of development and in adulthood. 5 We anticipated that a similar targeted expression might be achieved for Mitf.

Derivation of DMV Mice
We produced a transgene construct, TRE-Mitf-M-V5 (Mi-V5), in which a cDNA of mouse Mitf-M (419 residue; referred as [þ] isoform) with a V5 tag sequence at its carboxyl end (Mi-V5) was placed under the control of a TRE ( Supplementary  Fig. S1A). The Mi-V5 transgene construct was injected into zygotes heterozygous for Mitf mi-vga9 and Dct-rTA obtained by appropriate crosses and eventually generated a bi-transgenic line of mice on an Mitf-deficient background ( Supplementary  Fig. S1B). This line of mice is hereafter called DMV, and the littermate without the Mitf transgene DV. All mouse experiments were conducted following National Institutes of Health guidelines for the care and maintenance of mice and ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Experiments were approved by the Animal Care and Use Committee at National Cancer Institute Frederick. All experiments were performed on mice older than 1 month.

Mouse Genotyping
Genomic DNA (gDNA) was prepared for genotyping from mouse tail clips by the HotSHOT protocol. 8 In brief, mouse tail clips, 2-to 3-mm long, were immersed in 75 lL of alkaline lysis buffer (25 mM of NaOH and 0.2 mM of EDTA, pH ¼ 12) and heated at 958C for 30 to 40 minutes. After cooling on ice for 1 minute, 75 lL of neutralization buffer (40 mM of Tris-HCl, pH ¼ 5) were added. The PCR reaction contained 13 GoTaq Green Mix (No. M712; Promega), primers (0.5 lM each), gDNA from HotShot method (2 lL), and water for a final volume of 25 lL (Supplementary Table S1).
The genotyping of the Mitf mi-vga9 (vga9) allele requires two PCR reactions, Mi-UP and LacZ. Homozygous vga9 yields LacZpositive, no-Mi-UP results. Heterozygous vga9 yields both LacZ and Mi-UP positive results. Wild-type Mitf yields no-LacZ, but Mi-UP-positive results. The genotyping PCRs for Dct-rtTA and TRE-Mi-V5 use primers of rtTA and TRE-Mi, respectively.

Reverse Transcription-Polymerase Chain Reaction
The tissues harvested from mice were flash frozen in liquid nitrogen and stored at À808C. To prepare RNA, the frozen tissue was pulverized using Cryoprep System (#CP-01; Covaris, city, state, country), and the powder was lysed for RNA preparation following the instruction of RNeasy Mini Kit (#74106; Qiagen, Germantown, MD, USA). cDNA was synthesized using Invitrogen SuperScript III First-Strand Synthesis System (#18080051; ThermoFisher Scientific). The PCR reaction mixture was prepared as mentioned in the Mouse genotyping section. The primers and thermocycles are described in Supplementary Table S2.

Scoring of Eye Size
Eye size was scored as ''micro,'' ''medium,'' or ''large'' based on photographs of adult mice taken by CS and CGC. Mice were then classified by genotype and dox treatment. Eyes that could not clearly be seen in the photographs were excluded from evaluation. Classification was performed for 136 eyes from DV mice, 43 eyes from DMV Àdox mice, 83 eyes from DMV þdox mice treated with 2 g/kg dox and 35 eyes from DMV þdox mice treated with 0.2 g/kg dox. A X 2 test was performed to analyze the distribution using Prism version 7 (Graphpad Software, La Jolla, CA, USA).

Histology and Immunohistochemistry
Tissues were fixed in 10% neutral buffered formalin for 24 hours and then moved to 70% ethanol until being sent to Histoserv, Inc. (Gaithersburg, MD, USA) for paraffin embedding, sectioning, and hematoxylin and eosin (H&E) staining. The immunohistochemistry protocol was performed as published 9 with the addition of hydrogen peroxide melanin bleaching following antigen retrieval. Briefly, slides were deparaffinized in xylene and rehydrated through graded alcohols. Antigen retrieval was performed using Target Retrieval Buffer pH 6 (#S1699; Dako, Santa Clara, CA, USA) with steam (Farberware Programmable Pressure Cooker; Farberware, Vallejo, CA, USA) for 10 minutes followed by further incubation for 10 minutes on the benchtop. For melanin bleaching, 10% hydrogen peroxide was heated to 608C, slides were incubated at 608C in warm H 2 O 2 for 10 minutes and then washed in TBS for 5 minutes. After addition of a nonspecific protein blocker (X0909; Dako), slides were stained with a Dct antibody (Pep8h, gift from Vincent Hearing, National Cancer Institute) by incubation at a 1:5000 dilution overnight at 48C. The staining was revealed using ImmPRESS anti-rabbit alkaline phosphatase polymer (Vector Laboratories, Burlingame, CA, USA), developed with Vector Red chromogen (Vector Laboratories) followed by counterstaining with Mayer's hematoxylin (#MHS32; Sigma Aldrich, St. Louis, MO, USA) before coverslipping.

Ectopic Expression of Mitf-M in MITF-Deficient Mice During Embryogenesis Rescued Ocular Size, But Not Coat Pigmentation
One of the most abundant isoforms of MITF in neural crestderived melanocytes is MITF-M while in the RPE other isoforms (MITF-A, MITF-H, MITF-D) are more abundant, 10-12 although low levels of MITF-M can also be found, at least in adult human and bovine RPE. 13 As available evidence suggests that the different isoforms do not regulate vastly different sets of target genes, we reasoned that dox-inducible ectopic expression of MITF-M in Mitf mi-vga9 homozygotes might be able to rescue not only neural crest-derived melanocytes but RPE cells as well (Fig. 1A). The function and inducible expression of the Mi-V5 construct was confirmed using an Mitf-responsive reporter assay and Western blotting in cell lines expressing tetracyclinecontrolled transactivators (Figs. 1B, 1C). The Mi-V5 construct was then used to derive the DV and DMV mice on an Mitfdeficient Mitf mi-vga9 background (Supplementary Figs. S1A, S1B). As anticipated, when mothers carrying DMV embryos were treated with dox, transgenic Mitf-M could be induced in the offspring ( Fig. 2A). During embryogenesis, pigment was grossly visible in the optic cup of DMV þdox embryos by 13.  Fig. S3D, inset). When the dox-treated DMV pups were allowed to grow to adulthood, their coat remained as white as that of untreated DMV controls or transgene-negative Mitf mi-vga9 homozygotes. Their eyes, however, were considerably larger and visibly pigmented compared with those of untreated controls (Fig. 2B). Among the organs tested, Mitf-M induction was dependent on dox and efficient in the eye. Nevertheless, while dox-dependent in the eye, transgene expression was leaky in the intestine and the olfactory bulb. Interestingly, we never detected induction in the skin.
Mitf-M and total Mitf expression were similar in control C57BL/6 and DMV þdox eyes, as was the expression of Dct, although there were individual variations in expression levels (Fig. 3A). Variation in eye size in DMV þdox mice was often related to the level of transgene expression (Figs. 3A, 3B). Large eyes were in general significantly (P < 0.0001) more common in DMV þdox mice (28%) than in DMV Àdox (5%) or DV mice (4%) (Figs. 3C, 3D). This variability was reduced after treatment with 2 g of dox/kg as opposed to 0.2 g of dox/kg ( Supplementary Fig. S4). Rarely, large eyes were also seen in the DV population, perhaps due to genetic variability introduced from an initial C3H outcross arranged to increase fertility when establishing the transgenic line ( Supplementary  Fig. S5). Pigmented hairs, however, were never found on any of the DMV pups or adults, perhaps because the ectopic Mitf-M expression started too late in embryogenesis, remained at too low levels, or lasted for too short a time period to rescue melanoblasts. From these results, we concluded that Mi-V5 was inducible and functional in DMV mice but only in their eyes.

Mitf-M Expression Partially Restored Retinal Structure but Not Function
The above findings were intriguing and prompted us to ask whether pigment cell rescue in the eye was restricted to the RPE or involved other eye structures as well, in particular also neural crest-derived iris and choroidal cells. In addition to microphthalmia, Mitf mi-vga9 mice have structural abnormalities in the eyes, including RPE hyperplasia and loss of normal retinal architecture. 3,[14][15][16] The histologic analysis of adult eyes (Fig. 4), indicated not only an increase in size but also an overall improvement of the anatomy of the eye, showing a more easily identifiable and less dysplastic lens, and pigmented choroidal and iris melanocytes, either isolated or clustered in small groups. These latter cells were also positive for DCT, consistent with the suggestion that they represent true melanocytes (Fig. 4C). Nevertheless, rescue of these cells is incomplete and estimated not to exceed 5% compared with control. Dox treatment also led to a more organized RPE, with a number of its cells pigmented, in contrast to untreated controls, whose RPE remained unpigmented and hyperplastic (Figs. 4A middle row, 4B, 4Ca, 4Cb). Among the various improvements in eye anatomy we also saw an improvement of the layered structure of the retina. While in untreated controls, the outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), and inner plexiform layer (IPL) are all thinned out (Fig. 4A, middle row), these layers are by comparison thicker in the treated DMV mice (Fig. 4A, middle row, and 4B). Nevertheless, photoreceptor outer and inner segments (OS and IS) remain absent regardless of treatment (Fig. 4B). Improvement in the RPE was additionally supported by expression of Best1, an RPE marker, in rescued RPE but not controls (Fig. 5A). However, expression of the retinol dehydrogenase Rdh5, which participates in the regeneration of 1-cis-retinal required for photoreceptor function, was not rescued (Fig. 5A). 16 We also tested whether the rescued eyes would respond to visual stimuli by measuring their visualevoked response (VER) in electroretinograms. Not surprisingly, they did not (Fig. 5B). This finding is consistent with previous work showing that the RPE is integral for normal photoreceptor layer development. 17 Lastly, we asked whether later dox removal would have any effect on the pigmented cells. Hence, we discontinued treatment in two mice at 33 days and one mouse at 72 days after birth but saw no significant gross or microscopic changes in the eyes for up to 90 days thereafter, suggesting that MITF-M expression is not required for maintenance of ocular elements.

Expression of Mitf-M under the control of a Dct promoter in
Mitf-null mutant mice was able to partially rescue the structure of the eye, though not its visual function, and along with it a fraction of the RPE cells and choroidal and iris melanocytes, although the neural crest origin of these latter rescued cells still remains to be confirmed. No rescue, however, was achieved for cutaneous melanocytes. This was similar to a previous study showing a partial rescue of ocular pigmentation but not hair pigmentation using a tyrosinase-rtTA transgene in albino mice. 18 We posit that the failure to rescue skin melanocytes and variable rescue of ocular pigment cells could  be due to either toxicity of the rtTA system, missing a critical window of MITF expression before DCT expression arises, and/or insufficient MITF expression driven by the transgenic Dct promoter. Although toxicity of rtTAs have been reported, 19 Dct-rtTA LacZ animals show rtTA expression in both eyes and cutaneous melanocytes 20 and expression of the transgenes throughout development is routine in our laboratory for visualizing cutaneous melanocytes. It is possible, however, that the period in normal development between the first onset of MITF-M expression and subsequent DCT expression is critical for survival of pigment cell precursors in the skin. Because the use of Dct-rtTA would miss such a window, this would account for failure to rescue cutaneous melanocytes. Additionally, the dose of MITF-M may have been insufficient to rescue skin melanocytes and fully rescue the RPE. Although our Dct-rtTA TRE-H2BGFP system successfully results in expression of GFP in melanocytes, various causes, including integration site position of the transgene, strain background, dose responses, and epigenetic silencing, have been suggested for variegated or insufficient transgene expression in mice. 18,[21][22][23][24] Although DMV mice have near normal MITF expression in the eyes in adult mice, we unfortunately cannot assess expression in cutaneous melanocytes because failure of expression (for any reason) necessarily results in an absence of skin melanocytes. Previous work has shown that expression of inducible GFP with this Dct-rtTA is higher when both transgenes are homozygous as opposed to heterozygous. 5 Combined with the better rescue observed at the higher dox dose, this suggests that homozygous Dct-rtTA Mi-V5 mice may show a more complete rescue of ocular structures and pigment cells.
With respect to eye development, our findings confirm the critical role of RPE in retina development and raise interesting questions concerning the functional role of distinct MITF isoforms. Previous results have shown that selective nonconditional knockouts of Mitf-D or conditional suppression of Mitf-D by Dct-Cre-mediated knockout of Pax6 led to shifts in the expression of other MITF isoforms but no significant changes in total Mitf expression and no visible perturbations in eye development. 14,25 Moreover, the lack of MITF-M in mice homozygous for the extant allele Mitf mi-bw , although leading to the absence of choroidal melanocytes and melanocytes in the anterior layer of the iris, nevertheless leaves the RPE (and pigmentation of the posterior layer of the iris) as well as the overall structure of the eye intact. 26 That MITF-M, normally absent in the developing RPE in mice, partially rescued Mitfnull mutant RPE is, however, consistent with the earlier observation that a close relative of MITF, TFEC, rescued the RPE of Mitf mi-rw homozygotes. 14 Our findings suggest that Mitf-M þ6aa isoform expression during development can rescue RPE hyperplasia in MITF-null mice, in contrast to recent reports that only the -6aa isoform suppressed proliferation in a human RPE cell line. 27 The rescue of iris and choroidal as opposed to cutaneous melanocytes might indicate a different timing or level of expression of the rescue transgene in the different cell types, perhaps hinting at the earlier notion that distinct subpopulations of neural crest-derived melanocytes are not all created equal. In any event, our model for inducible expression of MITF may lend itself to future studies of the functional role of Mitf isoforms, Mitf relatives and other genes for the development of the various subpopulations of melanin-bearing pigment cells in mice.