Genome-Wide siRNA-Based Functional Genomics of Pigmentation Identifies Novel Genes and Pathways That Impact Melanogenesis in Human Cells

1Significant effort has been focused on identifying the molecular etiology for pigment variation in skin. 127 mouse coat color genes have been identified, 68 of these genes have human homologues, and 29 of these homologues impact pigmentation in humans. Genetic mapping studies have identified a limited set of genes responsible for skin and eye color variability. Pigment production involves the concerted actions of transcriptional, translational, and intracellular trafficking machinery. MITF, the master regulator of melanogenesis in the mouse hair follicle, activates the transcription of tyrosinase, the rate limiting step in melanogenesis. Tyrosinase is translated in the endoplasmic reticulum and is glycosylated in the Golgi apparatus. Tyrosinase activity is restricted to the melanosome, a melanin specific organelle of poorly defined origin,. While the subtle variation in human skin color is thought to be the result of the complex interaction of multiple genes, the majority of mouse mutants described have segmental or complete absence of pigment. Recent studies have identified partial loss of function mutations that impact the shade of melanin in zebrafish and human skin, but the spectrum of gene targets that regulate pigment shade is unknown. Melanin is expressed in different end organs conferring different functions. Melanin protects the skin, eyes, and brain from toxic insults. Melanin in the inner ear impacts sound conduction. Loss of melanin is thought to play a role in the etiology of age related macular degeneration and Parkinson's disease. Additionally, melanin is aberrantly regulated in human skin disorders such as vitiligo and melasma. Harnessing the molecular mechanisms that regulate melanogenesis to selectively modulate melanin production in the skin, eye, or brain could lead to novel treatments for multiple human pathologies. Pharmacologic modulation of melanin production has primarily focused on identifying inhibitors of tyrosinase, the rate limiting step in pigment production. Currently utilized tyrosinase inhibitors are clinically effective, but are carcinogenic in animal studies. Pharmacologic agonists that stimulate pigmentation in human tissues remain to be identified. A better understanding of the molecular network governing pigment production in the human epidermis is indicated to aid design of agents that inhibit or stimulate pigmentation in human skin.

2Studies to determine the key molecular regulators of melanogenesis in human melanocytes have been hampered by the innate fragility of these cells and the fact that they produce scant amounts of pigment in culture,. To identify novel regulators of melanogenesis in human cells, we utilized MNT-1 melanoma cells to screen a genome-wide synthetic siRNA library for single-gene loci that support melanocyte pigmentation. MNT-1 cells produce substantial amounts of melanin in culture, have a gene expression profile that is most similar to normal melanocytes, and have been used by others to identify pigment regulatory mechanisms that govern normal melanogenesis –. We employed a previously described Dharmacon siRNA library of 84,508 siRNAs corresponding to four unique siRNA duplexes, targeting each of the 21,127 unique human genes arrayed in a one-gene/one-well format on 96 well microtiter plates. A spectrophotometric melanin quantitation assay was coupled with an ATP-dependent luminescence cell viability assay (CellTiter-Glo) to eliminate siRNAs that decrease melanin production as a consequence of impacts on either cell proliferation or cell survival. Using tyrosinase depletion as a positive control, we determined that a 5-day post-transfection incubation period was optimal for quantitative detection of impaired melanin production (Figure S1). Other studies demonstrated that the cell titer glo assay did not interfere with the spectrophotometric quantitation of melanin (data not shown). In order to identify genes that impact both pheomelanin and eumelanin production, we measured melanin content at 405 nm, a wavelength at which both pheomelanin and eumelanin absorb light.

3Raw A405nm absorbance values were normalized to internal reference samples on each plate to permit plate-to-plate comparisons. This analysis was followed by normalization to the experimental mean for each well location calculated from the full data set in order to control for variations in pigment due to plate position effects. Similarly adjusted luminescence values from the multiplexed viability assay were used to generate “normalized absorbance ratios” for each well (Table S1). The distribution of the means of these values from duplicate analyses is shown in Figure S1B. Previous studies have identified 68 genes that regulate pigment production in human cells. Initial examination of our dataset determined that siRNAs directed towards 13 of these 68 genes impaired melanin accumulation without impacting melanocyte survival or proliferation when depleted in these assays, with tyrosinase itself scoring with one of the lowest ratios (2.5 standard deviations below the mean; Table S2). Our current siRNA screening protocol relies on siRNA design algorithms to identify effective siRNA sequences and utilizes a single endpoint assay to identify siRNAs that impact pigment production. Genes whose function is not inhibited by selected siRNAs either secondary to the long half-life of the corresponding protein or secondary to poor siRNA sequence selection would not be identified in our screening approach. Identification of several known regulators of melanogenesis by our screening protocol does give confidence that our approach is sufficiently robust to identify novel regulators of pigment production.

4To facilitate the identification of novel genes that significantly impact melanogenesis, a cutoff of 2 standard deviations below the mean was used to select a candidate hit list. 98 genes were identified as regulators of melanogenesis by our screening approach. Of these 98 genes identified in the primary screen, only 6/98 genes (marked in red, Table 1) exhibited aberrant expression in MNT-1 cells as compared to normal melanocytes, indicating that the screen identified a large number of genes that likely impacted melanogenesis in both primary melanocytes and MNT-1 cells. Two of the genes identified in our screening approach were more recently eliminated from the Refseq database, and were not subject to detailed further evaluation.

5Individually synthesized, pooled siRNAs directed against 35 of the 96 remaining genes selected from the primary screen, as described above, were retested to determine the false-positive rate (Table S4). These genes were randomly selected from the putative target list. To more precisely control for the efficacy of siRNA transfection and to correct for the background absorbance of MNT-1 cells, the ability of each target siRNA to inhibit pigment production was compared to the ability of tyrosinase siRNA to inhibit pigment production using a normalized percent inhibition calculation, and relative pigmentation was assessed visually prior to cell lysis (Figure 1A,B). A Keratin 7 siRNA pool that did not impact pigment production was utilized as a negative control. Four siRNA pools failed to significantly impact pigment production upon retesting and were eliminated from further analysis (Figure 1A, Table S4), giving an estimated false discovery rate of 12.1%. To validate that our candidate siRNAs inhibit the expression of the gene of interest, we utilized quantitative RT PCR to examine if a random selection of candidate siRNA pools inhibited the expression of the appropriate target gene (Figure S2). These results validate that the siRNAs selectively impact the expression of the cognate target gene, although this may not conceivably hold true for all of the siRNAs used in our screen. To eliminate siRNA pools with off-target effects on melanogenesis, the four siRNAs comprising each siRNA pool were retested individually. We found that at least two independent siRNAs against each target gene significantly inhibited pigment production (Figure 1C, Table S4), suggesting that pigmentation phenotypes are not a common consequence of siRNA off-target phenomena. Together, these studies demonstrate that the genome wide siRNA screening platform accurately identified gene targets that specifically impact pigment production.

6Initial examination of existing GO annotation data for our pigment regulators exposed a wide variety of cellular processes represented by the validated and candidate hits (Table 1). Therefore, we employed a focused unbiased approach to identify regulators of tyrosinase, the rate limiting enzyme specifying melanogenesis among novel validated genes supporting MNT-1 pigmentation. Relative accumulation of tyrosinase, the melanogenesis transcription factor MITF, and the melanosomal marker protein Melan-A were examined 96 hours post siRNA transfection. Remarkably, over half of the validated pigment genes appear to be required for tyrosinase protein accumulation (Figure 2A, Figure S3). This defect did not appear to be a gross inhibition of cell fate specification, as Melan-A expression was mostly unaffected. In addition, the sub cellular morphology of PMEL17, a melanosome structural protein, was normal at the level of immunofluorescence detection (Figure S4). Of those pigment genes impacting tyrosinase accumulation, approximately half appear to act at the level of tyrosinase mRNA accumulation (Table 2), and most of these also impaired MITF mRNA accumulation. Given that tyrosinase is an MITF target gene, the pigmentation genes in this later class may represent action at the level of MITF mRNA. A caveat to this interpretation is our observation that siRNA-mediated turnover of tyrosinase mRNA can also lead to inhibition of MITF gene expression (Figure 2A) through a relationship that remains to be defined. Preliminary studies indicated that this phenotype was not a consequence of siRNA off-target phenomenon (Figure S3).

7While pigmentation in humans is a complex multigenic trait, the degree of genetic variation that contributes to melanocyte autonomous pigment production is unknown. To examine the phenotypic penetrance of novel pigmentation genes, identified in MNT-1 cells, in diverse genetic backgrounds, we employed primary human melanocyte cultures isolated from two different individuals. Remarkably, the majority of targets that regulated tyrosinase expression in MNT-1 cells also impacted tyrosinase expression when depleted from darkly pigmented primary melanocytes (Figure 2C, Figure S5A). Approximately half of these targets also inhibited tyrosinase expression when depleted from moderately pigmented melanocytes (Figure 2C, Figure S5B). These results indicate that the primary screen identified a number of genes that impact pigment production in several different genetic backgrounds. Selective activity of some of these targets in different genetic backgrounds suggests that some of these novel regulators of melanogenesis may play a role in human phenotypic variation. Future large scale studies are required to determine if these genes are differentially expressed in different pigment backgrounds.

8For further analyses, we focused on those novel pigmentation genes that impacted tyrosinase expression in all three genetic backgrounds. Among these were two isoforms of aldehyde dehydrogenase, ALDH1A1 and ALDH9A1, well characterized enzymes that regulate ethanol detoxification. A number of chemical inhibitors of these enzymes have been identified, and several of these agents are clinically utilized to induce alcohol intolerance during detoxification interventions; presenting an opportunity for pharmacological validation of the contribution of Aldh activity to melanocyte pigmentation. Disulfiram is an Aldh inhibitor that is toxic to melanoma cells via a mechanism that is independent of Aldh inhibition. However, two non-toxic Aldh inhibitors, cyanamide and Angeli's salt, inhibited pigmentation and tyrosinase protein accumulation in MNT-1 cells at doses that are equivalent to those required for inhibition of Aldh activity in culture (Figure 2D). For quantifying the impact of compound treatment on pigment accumulation, we used our spectrophotometric-based melanin quantitation assay that couples a CellTiter-Glo assay with a melanin quantitation assay to effectively eliminate compounds that impact cell survival or proliferation. Cyanamide did not appear to impact the viability of MNT-1 cells or primary melanocytes in culture. In addition, these compounds impaired UV-induced tyrosinase expression when tested in primary melanocytes (Figure 2D,E).

9Melanosomes are distinct lysosome-related organelles dependent upon appropriate post-golgi sorting events for delivery of functionalizing ‘cargo’ including tyrosinase. Therefore, impaired accumulation of tyrosinase can be a consequence of misrouting to lysosomes and subsequent hydrolysis in that organelle. To define target genes that may participate in this sorting event, lysosome acidification was inhibited by bafilomycin A1 exposure subsequent to target gene depletion. As shown in Figure 3A, a 24 hour inhibition of lysosome acidification rescued tyrosinase accumulation upon depletion of the small G-protein RAB4A, and the small G-protein palmitoyltransferase ZDHHC9. By contrast, bafilomycin did not restore tyrosinase accumulation upon depletion of MSRA, a protein that can protect against oxidative damage through reduction of methionine sulfoxide. These studies offered preliminary evidence that our screening approach did identify novel genes that impact melanosome trafficking/sorting of melanosome protein cargo.

10Gene annotation data was utilized to identify other genes that may regulate melanosome intracellular trafficking/sorting of melanosome protein cargo. Among the panel of validated pigment regulatory genes with phenotypic penetrance in multiple genetic backgrounds was WIPI1 (Figure 2). Wipi1 has been implicated as a human homolog of the yeast autophagy protein ATG18, and is localized to starvation-induced autophagosomes in human cell culture. Two additional autophagy-related proteins, LC3-C and GPSM1/AGS3 were isolated in the primary screen (Table 1). Autophagy, or cellular self-degradation, is a highly conserved cellular pathway that has been associated with cancer formation, neurodegeneration, and aging –. This pathway functions to transport vesicle cargo (autophagosomes) to the lysosome for degradation. Scant evidence currently exists linking autophagy to melanogenesis. Previous studies have documented an abundance of autophagosomes in cells obtained from patients with a disorder of pigmentation (HPS-1) but have hypothesized that their presence is a consequence of the degradation of immature melanosomes within these cells. Other studies have determined that autophagosome components are present in the stage II melanosome, suggesting that parts of the melanosome originate from the autophagosome.

11Our genome wide siRNA screen directly identified autophagy components as novel regulators of melanogenesis. Validation of these targets by siRNA pool deconvolution supported a functional relationship between autophagosome and melanosome biogenesis (Figure 3B). Furthermore, we found that depletion of two additional components required to trigger autophagosome formation, BECN1 or LC3-A, severely impaired pigment accumulation (Figure 3B). Failure to recover these genes in the primary screen is indicative of the false negative rate inevitably associated with high throughput investigations and illustrates the point that our approach is unlikely to identify all known regulators of melanogenesis. Nonetheless, the validation that both Beclin1 and Lc3 impact pigment accumulation is supporting evidence that autophagy impacts melanogenesis. Consistent with this relationship, heterozygous deletion of the autophagy protein Beclin 1 results in a dramatic coat color defect in mice (Figure 3C, Figure S6). Homozygous null mutations are embryonic lethal, however haploinsufficient animals show an interesting chimeric phenotype with normal and hypopigmented hair follicles. The hypopigmented follicles in these mice contain less pigment in the hair follicle bulb as observed on horizontal sections of the hair follicle. Previous studies have determined that only melanoblasts within the hair follicle unit express S100b protein. To determine if the phenotype observed in beclin1 haploinsuffiicient mice is secondary to an impact of beclin1 depletion on melanoblast survival, we attempted to identify S100+ melanoblasts within the hair follicle in horizontally sectioned skin specimens of wild type and Beclin1 haploinsufficient mice. Consistent with published studies, it was difficult to identify S100 positive cells within the hair follicle of wild type mice secondary to interfering melanin. However, in Beclin1 haploinsufficient mice, we determined that S100+ cells were present in the hair follicle (Figure 3D). Data from our siRNA screen indicated that beclin1 depletion does not impact melanocyte survival. Taken together, our siRNA data and histologic analysis suggests that the phenotype observed in the Beclin1 haploinsuficient mice is not a consequence of impacts of Beclin1 on melanocyte survival but is more likely secondary to the impact of beclin1 on melanosome number or melanin content within the hair follicle. As melanosomes are thought to be lysosome related organelles, autophagic machinery may be required for the functional sorting of melanin synthetic machinery. At the cell autonomous level, we found co-localization of the autophagy proteins LC3 and APG5 and the melanosome markers PMEL17 in mature melanosomes (Figure 3E). Thus molecular components required for autophagosome formation are directly implicated in the biogenesis of melanin, either at the level of melanosome formation or melanosome maturation. When coupled with previous data demonstrating that autophagosomes accumulate in cells defective in melanosome maturation, these results indicate that the autophagy pathway is intimately involved in the process of melanosome maturation.

12We have utilized an unbiased, high-throughput functional genomics screening platform to identify critical single gene loci that regulate the notoriously complex, highly regulated process of melanogenesis in human cells. Using this approach, we have identified 92 novel genes that impact pigment production in human cells. The convergence of several of these loci directly on the critical rate-limiting enzyme in melanogenesis, tyrosinase, underscores the power of this approach to identify unrecognized genes that are components of even well characterized enzymatic pathways. The complexity of the network controlling tyrosinase expression uniquely parallels the variation in skin color seen in human skin, underscored by the fact that these mechanisms are differentially active in moderately and darkly pigmented melanocytes. The direct identification of novel pigment modulatory agents highlights the utility of genome wide siRNA screening as a translational approach for deriving novel molecular based treatment strategies.

13MNT-1 cells were a gift of M. Marks (University of Pennylvania). These cells were cultured in DMEM (Invitrogen) with 15% fetal bovine serum (Hyclone), 10% AIM-V medium (Invitrogen), 1xMEM (Invitrogen) and 1× antibiotic/antimycotic (Invitrogen). Darkly pigmented and moderately pigmented melanocytes were purchased from Cascade Biologics. These cells were cultured in Medium 254 with the melanocyte specific HMGS supplement (Cascade Biologics). Beclin 1 heterozygous mice were obtained from Beth Levine. Angeli's salt was a gift from Pat Farmer. Cyanamide was purchased from Sigma. Bafilomycin A1 was purchased from Tocris Biosciences. The genome wide siRNA library used in these studies was previously described. RPMI 1640 (Invitrogen) was media used for creating lipid oligonucleotide mixtures. All transfections utilized Dharmafect-2 transfection reagent (Dharmacon). For western blotting we utilized the following antibodies: tyrosinase (Santa Cruz Biotechnology, cat # sc-7833), MITF (Santa Cruz Biotechnology, sc-56725), Erk1 (Santa Cruz Biotechnology, sc-94), and Melan-A (Santa Cruz Biotechnology, sc28871). Primary antibody dilutions used in these studies ranged form 1∶200 to 1∶1000 and anti mouse or anti rabbit HRP antibodies (Santa Cruz Biotechnology) were used in the immunoblot analysis. Antibodies used for immunofluorescence are described below. S100B antibody was purchased from DakoCytomation.

14High throughput transfection was performed essentially as described with slight modifications. 0.28 pmoles of each siRNA pool in a volume of 30 ul of RPMI was delivered to each of 6 assay plates/master plate using a Biomek FX robotic liquid handler (Beckman Coulter). 0.1 ul of Dharmafect 2 (Dharmacon) in 9.9 ul of RPMI was then delivered to each well using a TiterTek Multidrop. Following a 20–30 minute incubation, 1×104 MNT-1 cells from a trypsin-mediated single-cell suspension were delivered to the siRNA/liposome complexes in a total volume of 200 ul. Plates were incubated for 120 hours at 37°C/5% CO2 after which a Hydra 96 (Robbins-Scientific) was used to removed 100 ul of the medium. 15 ul of CellTiter-Glo Reagent (CTG) (Promega) was delivered to each well and incubated according to manufacturer protocol. Luminescence and absorbance values for each well was recorded using an Envision Plate Reader (Perkin Elmer). Each transfection was performed in duplicate.

15Raw luminescence values collected from the high throughput screen were normalized to internal reference control samples (cells with no siRNA in wells A1–A8) on each plate to allow for plate-to-plate comparisons. These values were used to normalize absorbance values for each well in the plate, effectively controlling for the impact of each siRNA on cell viability. To normalize for positional variation in the plates secondary to prolonged culture times in the humidified incubator, each well in the plate was normalized to the mean value from all wells in the same location. Mean and standard deviation for each data point and the mean and standard deviation of the entire distribution was calculated. siRNAs that produced absorbance/CellTiter-Glo ratios two standard deviations below the mean were subjected to further analysis.

164×103 MNT-1 cells were transfected in 96 well plates with 50 nM candidate siRNA using 0.2 ul dharmafect 2 reagent. 48 hours after transfection, cDNA was prepared from transfected cells utilizing a Cells to Ct kit (Ambion) per the manufacterer's protocol. Primers targeting each candidate gene, tyrosinase, actin and MITF were purchased from Applied Biosystems. An aliquot of each cDNA reaction was then added to each Taqman master mix reaction along with the appropriate primer per the manufacturer's protocol (Applied Biosystems). A 7900HT Fast Real-Time PCR System (Applied Biosystems) was utilized to determine Ct values. Values were normalized using actin and analyzed using the relative quantification mathematical model (Pfaffl).

171×104 MNT-1 cells were plated in a 96 well microtiter plate. 24 hours after plating, cells were incubated with vehicle, hydroquinone Angeli's salt, or cyanamide. 48 hours after drug treatment, cell lysates were prepared and subjected to immunoblotting with a tyrosinase and ERK antibody. Similar protocols were utilized in primary melanocytes. Melanocytes were plated in 96 well microtiter plates in the presence of drug or vehicle. 24 hours after drug treatment, melanocytes were treated with UV. Cell lysates were prepared 24 hours after UV treatment and subjected to immunoblotting. In order to measure an impact of cyanamide on pigment production in melanocytes, primary melanocytes were incubated in the presence of vehicle, phenylthiourea, or increasing concentrations of cyanamide. Cells were incubated for an additional 7 days, with one media change on day 4, prior to collection of absorbance and viability values. For bafilomycin experiments, MNT-1 cells were transfected with 75 nM siRNA in 12 well plates. 80 hours after transfection, 25nM bafilomycin was added. 96 hours after transfection cell lysates were prepared and subjected to immunoblotting.

18For immunoflourescence detection of melanosome and autophagy markers, cells were fixed in 2% paraformaldehyde for 1 hour. Coverslips were washed in PBS, cells were permeabilized with 0.1% Triton-X-100 (MNT-1 cells) or 0.4% saponin (primary melanocytes), and blocked in 1% BSA with 0.1% Tween 20. Cells were incubated with the following primary antibodies: Pmel17 (HMB50, Lab Vision Corporation, 1∶100 dilution), Apg5 (Santa Cruz Biotechnology, sc-33210, 1∶50 dilution), and LC3b (Santa Cruz Biotechnology, sc-28266, 1∶50 dilution), for 1 hour. The secondary antibodies used in this study were Alexa Fluor 514 and 594 purchased from Invitrogen. Coverslips were incubated in a 1∶1000 dilution of the corresponding secondary antibody for 1 hour. Confocal images were acquired using a LSM-510 meta confocal multiphoton microscope.

19Beclin1 haploinsufficient mice were generated as described. All studies involving beclin1 +/+ and beclin1 +/− mice utilized animals that had been backcrossed with C57BL/6J for a minimum of six generations. Representative pictures and figures contained in the manuscript were generated from littermates.

20Mouse skin sections from four wild type and four beclin 1 haploinsufficient mice were fixed in formalin and paraffin embedded. Haematoxylin and eosin staining were performed following standard protocols. Melanin was stained using the Masson-Fontana technique with a neutral red counterstain. S100 staining was performed as described using an eosin counterstain.