Genetic Interactions between Neurofibromin and Endothelin Receptor B in Mice

When mutations in two different genes produce the same mutant phenotype, it suggests that the encoded proteins either interact with each other, or act in parallel to fulfill a similar purpose. Haploinsufficiency of Neurofibromin and over-expression of Endothelin 3 both cause increased numbers of melanocytes to populate the dermis during mouse development, and thus we are interested in how these two signaling pathways might intersect. Neurofibromin is mutated in the human genetic disease, neurofibromatosis type 1, which is characterized by the development of Schwann cell based tumors and skin hyper-pigmentation. Neurofibromin is a GTPase activating protein, while the Endothelin 3 ligand activates Endothelin receptor B, a G protein coupled receptor. In order to study the genetic interactions between endothelin and neurofibromin, we defined the deletion breakpoints of the classical Ednrb piebald lethal allele (Ednrbs-l) and crossed these mice to mice with a loss-of-function mutation in neurofibromin, Dark skin 9 (Dsk9). We found that Neurofibromin haploinsufficiency requires Endothelin receptor B to darken the tail dermis. In contrast, Neurofibromin haploinsufficiency increases the area of the coat that is pigmented in Endothelin receptor B null mice. We also found an oncogenic mutation in the G protein alpha subunit, GNAQ, which couples to Endothelin receptor B, in a uveal melanoma from a patient with neurofibromatosis type 1. Thus, this data suggests that there is a complex relationship between Neurofibromin and Endothelin receptor B.


Introduction
Pigment cells arise during mammalian development through one of two known lineages. One lineage arises directly from multipotent neural crest cells, while the other lineage initiates within bipotential melanoblast-Schwann cell precursors [1,2]. Pigment cells initially migrate through the dermis, and then later enter the epidermis and hair follicles, which are epidermal appendages [2,3,4,5]. Mature pigment cells, called melanocytes, produce melanin. Melanocytes in the hair follicles pigment the hair, while melanocytes in the dermis and the inter-follicular epidermis determine skin color.
The Endothelin receptor B (Ednrb) is a seven transmembrane G protein coupled receptor expressed in melanocytes [6]. Ednrb is activated by the Endothelin 3 (Edn3) ligand [7]. Endothelin signaling transmits survival/proliferation signals into melanocytes by way of heterotrimeric G proteins, including the alpha subunits, Gnaq and Gna11 [8,9]. Mice lacking either Edn3 or Ednrb have a very similar phenotype and are hypo-pigmented due to a lack of melanocytes during development [10,11,12]. The coats of mice completely lacking Ednrb (the Ednrb s-l allele) range from being completely white to being white with small pigmented patches on the head and/or rump [12]. In addition, these mice exhibit a completely unpigmented glabrous skin (tails, ears, and feet) [13].
Several lines of evidence suggest that endothelin signaling plays an important role in melanocyte survival in the dermis, but has little, if any, effect on melanocytes located in the inter-follicular epidermis. First, transgenic over-expression of Edn3 in keratinocytes causes hyper-pigmentation of the dermis, but not of the interfollicular epidermis [14]. Second, there are no white coat patches when Ednrb knockout occurs after E12.5, when melanoblasts, immature melanocytes, first migrate into the epidermis [15]. And finally, activating mutations in Gnaq and Gna11 cause dermal, but not epidermal, hyper-pigmentation [13].The dermal hyperpigmentation of Edn3 over-expressing mice dissipates when the transgene is shut off, indicating that endothelin signaling continues to regulate the density of melanocytes in the adult dermis [14].
Neurofibromin (Nf1) is a GTPase activating protein that also regulates pigmentation [16,17,18]. In humans, inherited heterozygous mutations in NF1 cause neurofibromatosis type 1, which is characterized by skin hyper-pigmentation and Schwann cell based tumors. These phenotypes are caused by a combination of haploinsufficiency and somatic loss of heterozygosity [19,20]. Nf1 mutant mice exhibit a darker dermis and epidermis [21]. Conditional haploinsufficiency of Nf1 induced in Schwann cellmelanoblast precursors around E11.5 recapitulates the dark dermis phenotype of Nf1+/2 mice, while conditional Nf1 haploinsufficiency induced in already committed melanocytes does not [21]. This indicates that Nf1 haploinsufficiency has its effect on dermal pigmentation during a limited period of time. A developmental connection between Schwann cells and melanocytes might underlie the susceptibility of these two types of cells to Nf1 loss. However, the cell type(s) required to recapitulate the darker epidermis phenotype have yet to be identified.
Because haploinsufficiency of neurofibromin and over-expression of Endothelin 3 both cause increased numbers of melanocytes to populate the dermis during mouse development, we are interested in how these two signaling pathways might intersect. To address this, we have examined genetic interactions between Nf1 and Ednrb in mice.

Determination of the Ednrb s-l Deletion Breakpoints
According to southern blot analysis, the entire coding region of the Ednrb gene is deleted in the Ednrb s-l (piebald lethal) allele, which is commercially available [12,22]. To be able to use this allele in our crosses, we needed to determine the location of the Ednrb s-l deletion breakpoints, in order to verify that no other genes are included in the deletion and to permit genotyping by PCR (polymerase chain reaction).
Our strategy was to determine where PCR either did, or did not, produce a product in DNA from Ednrb s-l /Ednrb s-l mice, using DNA from wildtype mice as a positive control. We designed our PCR primer pairs using sequence from the Ensembl mouse genome database (m37). Beginning 20 Mb away from Ednrb on either side, we sequentially designed primer pairs at denser intervals, targeting the region between the last known primer pair that produced a product and first known primer pair that failed. After several rounds, we were able to narrow this deletion flanking region to ,1 kb on either side of the deletion. Next, we designed several primer pairs in which the left primer was positioned in one deletion flanking region and the right primer was positioned in the other. Some of these primer pairs produced a product in Ednrb s-l / Ednrb s-l DNA, but not wildtype DNA. We sequenced these PCR products, and found that they contained the breakpoints.
The Ednrb s-l deletion encompasses 97,637 base pairs of chromosome 14 ( Figure 1a). According to the Ensembl genome database (m37), no other genes are located in the region that is deleted. Subsequently, for each animal to be genotyped, we used two different PCR reactions. One contains primers that span the deletion and is positive only if the Ednrb s-l allele is present, while the other contains primers within the deleted region and is positive only if the wildtype allele is present ( Figure 1b).

Interactions between Ednrb s-l and Nf1 Dsk9
To study genetic interactions between Nf1 and Ednrb, we made use of a loss-of-function Nf1 allele, Nf1 Dsk9 . The Dsk9 mutation is a missense in the GTPase accelerating protein related domain (GRD) of neurofibromin [21]. We crossed Nf1 Dsk9 /+ mice to Ednrb s-l /+ and then crossed the resulting Ednrb s-l /+; Nf1 Dsk9 /+ mice to Ednrb s-l /+ mice. The coats of Ednrb s-l /Ednrb s-l mice range from being completely white to being white with small pigmented patches on the head and/or rump [12,22].
In the progeny from this cross, we measured the mean pixel intensity of group photographed pieces of tail dermis, split from the epidermis using sodium bromide. The mean pixel intensity correlates with the skin color of the sample (Figure 2a versus b) [13,21] and permits statistical analysis. The mean pixel intensity of +/+; Nf1 Dsk9 /+ dermis is greater than +/+; +/+ dermis and Ednrb sl /+; +/+ dermis (Figure 2a-b, p = 0.0328 and p = 0.00748, respectively, student's ttest), while the mean pixel intensity of Ednrb s-l /+; Nf1 Dsk9 /+ dermis is not significantly different than +/+; +/+. This suggests an additive effect of Ednrb s-l /+ with Nf1 Dsk9 /+. As expected, the tail dermis of Ednrb s-l /Ednrb s-l mice is completely unpigmented (Figure 2a) [13]. The tail dermis of Ednrb s-l /Ednrb s-l ; Nf1 Dsk9 /+ mice also lacks visible pigment (Figure 2a). Therefore, Nf1 Dsk9 /+ is not able to compensate for a complete lack of Ednrb in the tail dermis.

GNAQ Mutation in NF1 Uveal Melanoma
There is an interesting association between dermal hyperpigmentation and uveal melanoma, which is a melanoma of the uveal tract of the eye. For example, the nevus of Ota, a hyperplasia of the dermis, is associated with an increased risk of uveal melanoma in people of Western European descent [23]. In addition, constitutively active, oncogenic mutations in either GNAQ or GNA11 are found in ,75% of common nevi of the dermis and in ,83% of uveal melanomas [24]. Furthermore, constitutively active mutations in GNAQ/GNA11 [24,25] and loss of function mutations in NF1 [26] both lead to hyper-active MAP kinase signaling. To examine whether oncogenic GNAQ or GNA11 mutations are required in uveal melanomas bearing NF1 mutations, we examined a uveal melanoma tumor from a patient with neurofibromatosis type 1, which is known to be caused by inherited heterozygous mutations in NF1 [19,20]. There are multiple reports of uveal melanoma in patients with neurofibromatosis type 1 in the literature, despite the rarity of uveal melanoma. For examples, see [27,28,29].
The uveal melanoma from a patient with neurofibromatosis type 1 that we studied was previously described [30] and was available in the pathology archives at the Vancouver General Hospital. We sequenced GNAQ and GNA11 in DNA extracted from the formalin-fixed, paraffin-embedded sample. We found that this tumor bears a somatic glutamine 209 to proline substitution in GNAQ (GNAQ Q209P ), which is known to cause constitutive activity [25,31,32] (Figure 3a-b). This finding suggests that even in the context of NF1 haploinsufficiency, constitutive activity of GNAQ plays an important oncogenic role in uveal melanoma.

Discussion
We have studied the genetic interactions between Ednrb and Nf1 in pigmentation, examining the result of bringing loss of function mutations in Ednrb and Nf1 together in mice. When mutant alleles of two different genes are brought together in an animal, one possible outcome is that the phenotype of one allele is expressed while the other is not. This is referred to as epistasis. In our experiments, we found that Nf1 Dsk9 haploinsufficiency requires Ednrb to darken the tail dermis of mice (Figure 2a,b). Thus, Ednrb is epistatic to Nf1 with respect to dermis skin color. Since the dermis of Ednrb s-l /Ednrb s-l mice is completely albino, this suggests that endothelin signaling is required prior to or in parallel with Nf1 to support melanoblast formation/survival/proliferation in the dermis during development. However, we also observed that Nf1 haploinsufficiency increases the amount of pigmented coat spots in mice that lack Ednrb (Figure 2c). Since Ednrb is not absolutely required for coat pigmentation, Nf1 Dsk9 appears to be able to increase melanocyte occupancy of hair follicles during development, in the absence of Ednrb. Thus, our data suggests that these two pathways do not act in concert all the time. This is supported by the finding of a constitutively active GNAQ mutation in an NF1+/2 uveal melanoma, which otherwise might be redundant, if EDNRB/GNAQ signaling completely overlapped with NF1 [8,9].
How might Ednrb be essential for dermal pigmentation in the tail, but not hair follicle occupancy? It is likely caused by the reduced requirement for Ednrb in the head and rump regions, as compared with the rest of the body. During embryogenesis, Ednrb sl /Ednrb s-l embryos exhibit greatly reduced numbers of melanoblasts, at or before embryonic day (E) 10.5 [10]. This suggests that there is an early requirement for Ednrb in melanocyte development. However, melanoblasts expressing the melanocyte marker, dopachrome tautomerase (Dct), persist in the head and rump skin of Ednrb s-l /Ednrb s-l embryos [10]. This varied regional requirement for Ednrb could be due to environmental differences, and/or to differences in the origin of the melanoblasts themselves. The trunk dermis is derived from the somatic dermamyotome, while the head dermis is derived from the cranial neural crest. At least some of the melanoblasts in the head arise from Schwann cell precursors associated with cranial nerves, while others appear to arise directly from neural crest cells, such as those at the midbrain-hindbrain border [33]. Newly formed melanoblasts located close to cranial nerves exhibit a greater reduction in number in Ednrb tm1Nrd /Ednrb tm1Nrd embryos compared to the other melanoblasts in the head, suggesting that the non-Schwann cell derived melanocytes have a reduced requirement for Ednrb [33].
The colonization of hair follicles by melanoblasts occurs only when the hair follicles are first forming [34]. The percent of the coat that is pigmented reflects the number of melanoblasts that were available in the epidermis at that stage. Each self-contained hair follicle retains its occupied or unoccupied status into adulthood. Endothelin signaling appears to be dispensable for melanocytes once they are located in the epidermis [14,15]. Thus, melanoblasts in the head and rump regions that manage to develop/survive in the absence of endothelin signaling should be able to expand freely to their maximum potential once they enter the epidermis [2,3,4,5].
We have found that Nf1 haploinsufficiency increases the area of the head and rump coat that is pigmented in Ednrb null mice. This indicates that melanoblasts with a reduced requirement for Ednrb can be stimulated by Nf1 haploinsufficiency to increase in number in the epidermis. Given the findings of others [33], described above, this suggests there might be a role for Nf1 outside Schwann cell derived melanoblasts. Because Schwann cell precursors arise from neural crest cells, there is currently no cre line that would specifically target melanoblasts arising directly from the neural crest, which could be used to test this hypothesis.
In summary, our data suggests that there is a complex relationship between Nf1 and endothelin signaling, and that there are multiple roles for Nf1 in pigmentation. If one copy of Nf1 is lost, more melanoblasts are produced from Schwann cell precursors, which darkens the dermis [21]. However, these melanoblasts require Ednrb for their production and/or their survival/proliferation. If both copies of Nf1 are lost, on-going melanocyte proliferation/survival is enhanced in both the dermis and the epidermis of the tail, via a melanocyte cell autonomous mechanism [21]. Finally, on a background with reduced melanoblast numbers, Nf1 haploinsufficiency can increase the number of melanoblasts in the head and rump epidermis, via an Ednrb-independent mechanism.

Mouse Husbandry
This study was carried out in strict accordance with the recommendations of the Canadian Animal Care Committee. All  In all experiments, we collected samples between 2 and 3 weeks of age, before the Ednrb sl /Ednrb s-l mice develop megacolon. Ednrb s-l mice were obtained from the Jackson Laboratories [22]. Nf1 Dsk9 mice were recovered during an ENU mutagenesis screen at the Institute for Experimental Genetics in Germany [35]. Both strains were made predominantly C57Bl/6J.

Genotyping
Genomic DNA from ear skin biopsies or embryonic membranes was extracted using Qiagen DNeasy Blood and Tissue Kit (Invitrogen). 50 ng of DNA was used for each PCR reaction, including 1X PCR buffer containing 1. Dermal-epidermal Separation and Pixel Intensity of Tail Skin 1 cm piece of skin from the middle of the tail was removed from the tail bones of 2 week old mice, incubated in 2 M sodium bromide for 2 hours at 37uC, and separated using fine forceps. Skin samples to be compared to one another were photographed within a single picture. Dermal skin pigmentation was quantified using ImageJ in terms of average pixel intensity.

Uveal Melanoma
An archival, paraffin embedded, enucleated uveal melanoma from a patient with neurofibromatosis type 1 was obtained under the approval of the institutional review board at University of 2-3 week old mice is shown. The mean pixel intensity of +/+; Nf1 Dsk9 /+ dermis is greater than +/+; +/+ dermis (p = 0.0328*, student's ttest) and Ednrb sl /+; +/+ dermis (p = 0.00748**, student's ttest), while the mean pixel intensity of Ednrb s-l /+; Nf1 Dsk9 /+ dermis is not significantly different than +/+; +/+. The pixel intensity is not zero in Ednrb s-l /Ednrb s-l tail dermis, with or without Nf1 Dsk9 , because the photograph is not completely white in terms of pixels, even while there is no melanin. c) The percent of the coat that is pigmented (mean area +/2 S.E.M.) in Ednrb s-l /Ednrb s-l ; +/+ and Ednrb s-l /Ednrb sl ; Nf1 Dsk9 /+ mice is shown. On average, spots of coat pigmentation account for 9.8% of the area of the coat in Ednrb s-l /Ednrb s-l ; Nf1 Dsk9 /+ mice, compared to 1.1% in Ednrb s-l /Ednrb s-l ; +/+ mice (p = 0.003925***, student's ttest). No spotting was observed in any other genotype (data not shown). In b and c, individual dots in graphs represent different animals, the total number of which is given by n for each genotype. doi:10.1371/journal.pone.0059931.g002 Figure 3. Analysis of GNAQ and GNA11 in a NF1 uveal melanoma. a) DNA was extracted from a uveal melanoma from a patient with neurofibromatosis type 1. The DNA was used to sequence exons 4 and 5 of GNAQ and GNA11. The tumor exhibited a CAA to CCA (glutamine to proline) substitution in GNAQ exon 5 at residue 209, a known oncogenic hotspot. No other mutations in GNAQ or GNA11 were found. b) DNA extracted from surrounding normal tissue of the patient exhibited the expected sequence at GNAQ 209 (CAA), indicating that the Q209P mutation in the tumor is somatic. doi:10.1371/journal.pone.0059931.g003 British Columbia and the Vancouver General Hospital (previously described in [30].) The data was analyzed anonymously. Tumor tissue from five 8 mM thick sections was micro dissected, digested with proteinase K, and heat inactivated. 1 uL (,5 ng of DNA) was used to amplify GNAQ and GNA11 exons 4 and 5 as previously described [24,25]. Sequencing 5 mL of diluted PCR product (35 ng) was mixed with 2 mL of Exo-SAP-IT PCR product clean-up mix (Affymetrix). 3 mL of Exo-SAPed PCR product, 1 pM of primer, and 3 mL of Big dye mix (v3.1, Applied Biosystems) were used for each sequencing reaction, and run on an Applied Biosystems 3730 DNA Analyzer.

Statistics
Jump was used to determine that all data was distributed normally using the Shapiro-Wilk W test. p values of statistical significance were calculated using either student's ttest.