Specific Interaction of Gαi3 with the Oa1 G-Protein Coupled Receptor Controls the Size and Density of Melanosomes in Retinal Pigment Epithelium

Background Ocular albinism type 1, an X-linked disease characterized by the presence of enlarged melanosomes in the retinal pigment epithelium (RPE) and abnormal crossing of axons at the optic chiasm, is caused by mutations in the OA1 gene. The protein product of this gene is a G-protein-coupled receptor (GPCR) localized in RPE melanosomes. The Oa1-/- mouse model of ocular albinism reproduces the human disease. Oa1 has been shown to immunoprecipitate with the Gαi subunit of heterotrimeric G proteins from human skin melanocytes. However, the Gαi subfamily has three highly homologous members, Gαi1, Gαi2 and Gαi3 and it is possible that one or more of them partners with Oa1. We had previously shown by in-vivo studies that Gαi3-/- and Oa1-/- mice have similar RPE phenotype and decussation patterns. In this paper we analyze the specificity of the Oa1-Gαi interaction. Methodology By using the genetic mouse models Gαi1-/-, Gαi2-/-, Gαi3-/- and the double knockout Gαi1-/-, Gαi3-/- that lack functional Gαi1, Gαi2, Gαi3, or both Gαi1 and Gαi3 proteins, respectively, we show that Gαi3 is critical for the maintenance of a normal melanosomal phenotype and that its absence is associated with changes in melanosomal size and density. GST-pull-down and immunoprecipitation assays conclusively demonstrate that Gαi3 is the only Gαi that binds to Oa1. Western blots show that Gαi3 expression is barely detectable in the Oa1-/- RPE, strongly supporting a previously unsuspected role for Gαi3 in melanosomal biogenesis. Conclusion Our results identify the Oa1 transducer Gαi3 as the first downstream component in the Oa1 signaling pathway.


Introduction
Hypopigmentation mutations affecting melanin synthesis or melanosomal biogenesis in the retinal pigment epithelium (RPE) of mammals are known to have profound effects on the developing retina and visual pathways, including abnormal crossing of the optic axons, nystagmus, strabismus, foveal hypoplasia, and reduced visual acuity [1]. Two forms of albinism are commonly recognized: oculocutaneous albinism (OCA), in which neither the eye nor the skin or hair are pigmented and ocular albinism (OA), which affects primarily the eye pigmentation.
Ocular albinism type 1 (OA1, also called Nettleship-Falls type), is the most common form of ocular albinism. It has an estimated prevalence of 1/50,000 in the general population of the United States [2]. Although the cutaneous manifestations of OA1 are very mild, affected patients present abnormal macromelanosomes in both the RPE and skin [3]. In contrast to other forms of albinism, melanin is not dramatically reduced in OA1; in fact, this disease is characterized by the unusual coexistence of the typical albino visual defects with a substantial amount of melanin in the eyes [4]. Different types of mutations in the OA1 gene have been associated with ocular albinism type 1 (http://albinismdb.med.umn.edu/ oa1mut.html#mutations). OA1, the protein product of the OA1 gene, is a G protein-coupled receptor localized to RPE melanosomal membranes and the initiator of the observed abnormal visual phenotype in ocular albinism. The position of OA1 within these membranes, with its N-terminal towards the lumen of the melanosome and C-terminal towards the cytoplasm, suggests that it may function as a novel intracellular GPCR activated by the binding of a melanosomal ligand. This ligand could thus regulate melanosomal biogenesis through activation of specific G-proteins found in the RPE cytoplasm [5]. In addition, this localization of OA1 is consistent with its proposed role as a stop signal for melanosome overgrowth during melanogenesis [6,7] and could explain the changes in RPE phenotype observed in ocular albinism: mutations or deletions in OA1 producing a non functional OA1 protein would allow a continuous vesicular traffic of membrane proteins to melanosomes resulting in the formation of macromelanosomes [3].
It has been shown that the endogenous OA1 from human melanocyte extracts co-immunoprecipitates mainly with the alpha subunits of the Gi subfamily of heterotrimeric G-proteins [8]. The Gai subfamily comprises three closely related members, Gai1, Gai2, and Gai3, which share 85 to 95% amino acid sequence identity and partial overlapping expression patterns. In addition to their established function in signal transduction across the plasma membrane, these heterotrimeric Gi proteins localize to intracellular membranes and have been implicated in the regulation of membrane trafficking and fusion events along the secretory and endocytic pathways, such as vesicle formation by the endoplasmic reticulum, the Golgi/secretory pathway, and vesicle trafficking and fusion [9,10,11]. We have recently shown that mouse Oa1 and Gai3 play an important role in the determination of melanosomal size and density and that both signal in the same pathway to regulate axonal guidance at the optic chiasm [12]. In the present study, we used Gai knockout mouse models to investigate whether the other two members of the Gai family of proteins, Gai1 and Gai2, are involved in the regulation of size and density of the RPE melanosomes. We also studied the effects of the loss of both Gai1 and Gai3 using the corresponding double knockout mice (heretofore called DKO). In addition, we analyzed the specific interaction of each Gai protein with Oa1 in in-vitro immunoprecipitation and GST-pull-down experiments. For the latter, two polypeptides corresponding to the Oa1 third intracellular loop (i3) and carboxy-terminal tail (CT), the two regions that have been shown to play a critical role as selectivity determinants in receptor-G-protein interactions [13,14], were tested. Furthermore, we investigated the expression and cellular response of Gai proteins in the RPE of Oa1-/mice using Western blots and ADPribosylation. Altogether, our in-vivo and in-vitro data demonstrate a previously unknown role of Gai3 as the specific protein downstream of Oa1 that regulates the size and density of RPE melanosomes.

Size, density and morphology of RPE melanosomes
To investigate the involvement of Gai proteins in the regulation of size and density of RPE melanosomes, we analyzed electronmicrographs of RPE melanosomes from each Gai1-/-, Gai2-/-, Gai3-/and DKO mice, using 129 Sv wild-type mice as controls, as discussed in Materials and Methods. We had previously shown that Gai3-/mice, like Oa1-/mice, have larger RPE melanosomes than those of control mice, and that they each have reduced density of melanosomes [12] (Figures 1A, B, C and D). Notably, the two background strains for these knock-out mice (129 Sv and C57Bl/6NCrl, hereafter B6/NCrl) differ substantially in melanosome size and density, making clear the need for appropriate controls when considering these features of the RPE. Oa1-/macromelanosomes are conspicuously larger than Gai3-/macromelanosomes, but the latter are still significantly larger than those found in the 129 Sv control mice [12]. Our current electron microscopy results compare the relative roles of Gai1 and Gai2 to that of Gai3 upon melanosomal size, density and shape, and assess the effects of combining the loss of two of them in a double knockout mouse.
We first examined the size of melanosomes in the RPE of these different Gai knockout mice. The RPE of Gai1-/mice shows melanosomes that appear slightly enlarged ( Figure 1E), while melanosomes in Gai2-/mice appear no different ( Figure 1F), when compared with the melanosomes of 129 Sv mice ( Figure 1C). By contrast, the RPE in DKO mice appears to contain enlarged melanosomes relative to those in 129 Sv mice, being comparable in size to those in the RPE of Gai3-/mice ( Figure 1D). In order to quantify these differences, we measured melanosomal size, and determined the relative frequency of larger, macro-melanosomes (.5000 nm 2 ) within the RPE of the different Gai-/mice. A total of 2501 129 Sv, 1285 Gai1-/-, 1609 Gai2-/-, 1115 Gai3-/-and 1997 DKO melanosomes were sampled. Comparison of the area of melanosomes from each mouse line ( Figure 2A) shows that the frequency of larger melanosomes significantly differs between the groups, (ANOVA, p,0.05). Gai3-/-RPEs have the highest percentage of melanosomes larger than 5000 nm 2 (6.58%61.47), and post-hoc Tukey tests confirmed that this group is significantly larger than the 129 Sv control retinas (p,0.05). Thus, loss of only Gai3 yielded a significant increase in the presence of larger melanosomes. Combining the loss of Gai1 with Gai3 proteins in the DKO mice certainly did not worsen the abnormal morphology of the Gai3-/-RPE melanosomes. Together, these results suggest that only Gai3 has a function in the determination of the size of these organelles.
Second, we evaluated melanosome density (number of melanosomes/RPE mm 2 ) in the RPE of all the Gai-/and their 129 Sv control mice ( Figure 2B). While melanosomal density need not be inversely related to melanosome size, we previously found that it was in both Gai3-/and in Oa1-/mice [12], and so extended these analyses to the present knock-out mice. Gai1-/and Gai2-/mice showed comparable densities to those of 129 Sv mice, while Gai3-/mice had a large reduction (37.3%) in melanosomal density when compared to 129 Sv mice. DKO mice also showed a comparable effect (a 34.6% reduction). Statistical analysis confirmed an effect of group (ANOVA, p,0.05), and post-hoc Tukey tests confirmed that Gai3-/and DKO are significantly different from both 129 Sv and Gai2-/mice (p,0.05). These results and the fact that the loss of Gai1 alone had no significant effect on either melanosome size or density leads us to conclude that of the three heterotrimeric Gai proteins, Gai3 may be the endogenous downstream protein in the Oa1 signaling cascade that controls melanosomal size and density.
Third, we compared the morphology of the melanosomes in the RPE of the different Gai-/mice with that of their control mice, 129 Sv, to determine the percentage of melanosomes with a round shape. We classified all melanosomes according to their sphericity level, as described in Materials and Methods. Comparisons of 129 Sv with each Gai-/mouse studied showed that all animals except for Gai2-/have a similar percentage of RPE melanosomes that are round: 129 Sv 8.75%60.25, Gai1-/-9.37%60.77, Gai2-/-5.39%60.37, Gai3-/-9.12%61.85, and DKO 7.65%60.31. One-way analysis of variance, however, failed to reveal any effect of group despite the appearance of fewer round melanosomes in Gai2-/-( Figure 2C). Analysis of retinal function in all Gai-/-mice by electroretinography ERG responses obtained from129 Sv, Gai1-/-, Gai3-/and DKO mice are summarized in Figure 3A and from 129 Sv and Gai2-/mice in Figure 3B. In each panel, mean response amplitudes (61sd) are plotted against stimulus intensity (left to right). The retinal function of Gai1-/-, Gai3-/and DKO mice, as judged by the electroretinograms, appears to be not significantly different and within normal limits from control retinas ( Figure 3A). Curiously, the ERGs of the Gai2-/mice show a reduced b-wave amplitude ( Figure 3B), suggesting that the lack of Gai2 had some other impact on rod-mediated, but not on cone-mediated, retinal function, but the reason for this difference is not yet understood.
The Gai3 protein specifically interacts with the melanosomal GPCR Oa1 Seven transmembrane GPCRs are characterized by their ability to couple with and activate heterotrimeric G proteins in response to a ligand binding mainly through two well-known functional regions: the third intracellular loop (i3) and the carboxy-terminal tail (CT) [8,14]. Given that Oa1 shares all typical hallmarks of GPCRs, to characterize the potential interactions of Oa1 and Gai3 we carried out in-vitro binding studies in which recombinant fusion polypeptides of glutathione S-transferase with i3 and CT regions of Oa1, immobilized on glutathione-agarose beads, were incubated with in-vitro-synthesized 35 S-labeled Ga proteins. Figure 4A shows a diagram indicating how the GST fusion proteins were obtained using the PGEXT4-2 vector, as detailed in Material and Methods. Figure 4B corroborates that the apparent molecular masses of the fusion proteins, after SDS-PAGE electrophoresis, correspond to the expected 29 kDa for the GST::Oa1-i3 fusion protein and 37 kDa for the GST::Oa1-CT fusion protein; and Figure 4C shows the apparent molecular masses of the in-vitro synthesized 35 S-labeled Ga proteins.
SDS-PAGE of the Oa1-Gai interacting complexes eluted from the agarose beads and Coomassie blue staining of the gel showed that each lane contained the same amount of GST-fusion protein: lanes 2-7 and 8-13 correspond to GST::Oa1-CT and GST::Oa1-i3 reactions, respectively ( Figure 4D). Interestingly, the autoradiograph of the same gel demonstrated that Gai3 is the only one of all Ga proteins tested that specifically binds to Oa1 ( Figure 4E, lanes 4 and 10). No binding was detected when GST::Oa1-i3 and GST::Oa1-CT were incubated with 35 S-labeled Gai1, Gai2, Gas, or Gaq or when GST alone, used as control, was incubated with one of the 35 S-labeled Ga proteins (we used Gai3). Also, incubation of beads having only GST-fusion proteins (negative control) did not show any non-specific Oa1 binding with TNT rabbit retinoculocyte lysate or with the beads ( Figure 4E, lanes 7 and 13).
Mass spectrometry also identifies Gai3 as the specific OA1-interacting Gai protein To identify RPE proteins that interact with human OA1 and gain a better understanding of this GPCR's function, we used a human antibody against OA1 in immunoprecipitation (IP) reactions followed by SDS-PAGE. For these experiments, we dissected the RPE from donor, adult human eyes, which allowed us to obtain sufficient amount of protein. Half of the gel was stained with Coomassie blue to visualize the protein bands. Bands in the appropriate molecular mass of Gai proteins (39-41 kDa) and OA1 (45-48 kDa) were excised, destained, trypsinized and sent for mass spectrometry analysis to the Pasarow Mass Spectrometry core facility at UCLA. The other half of the gel was transferred to nitrocellulose membranes to confirm by Western blot analysis, using antibodies to the mass spectrometryidentified interacting protein, the specificity of the OA1 partner.
As we anticipated, mass spectrometry of the ,38-41 kDa excised band identified Gai3 as one of the proteins immunoprecipitated together with OA1 (Table 1). We corroborated the specific interaction of these proteins by incubating the blot with anti-Gai3 antibody and visualizing the Gai3 band using the ECL detection system ( Figure 5A). Given that the commercially available antibodies against Gai3 recognize both, Gai3 and Gb, Gb was another protein immunoprecipitated together with OA1 and identified by mass spectrometry (Table 1). RPE lysates not immunoprecipitated with antibodies against OA1 but incubated with pre-immune serum did not show any G-protein on the corresponding lane of the blot ( Figure 5A). Further confirmation of the OA1 and Gai3 interaction was obtained by performing a reciprocal IP experiment using lysates containing proteins of adult, human RPE and the anti-Gai3 antibody. After separating the immunoprecipitated proteins by SDS-PAGE and transferring them to nitrocellulose membranes, the blot was incubated with the anti-OA1 antibody. Our results show the presence of two forms of OA1, previously described with apparent molecular weights of 45 and 48 kDa [15], in the lane corresponding to the RPE proteins immunoprecipitated with the anti-Gai3 antibody and the absence of the OA1 protein in the lane corresponding to the RPE lysates incubated with pre-immune serum ( Figure 5B).
Gai3 compensates the lack of Gai1 in Gai1-/-mouse RPE Compensatory increases elevating a Gai subunit level in a tissue when another Gai subunit is missing from it have been previously observed [16].To test this possibility, we used Western blots incubated with Gai common antibody to analyze the expression of  Proteins that cover a wide variety of functions in signal transduction, pre-mRNA processing and cytoskeleton assembly [32].  experiments. After incubation of the crude membrane preparations from RPE with PTX and 32 P-nicotinamide adenine dinucleotide ( 32 P-NAD), an autoradiography of the 32 P-labeled proteins separated on 6 M Urea-SDS-PAGE was obtained. Two bands at 40 and 41 kDa corresponding to Gai2 and Gai1 plus Gai3, respectively, were observed on the lane of the gel containing the RPE from control mice ( Figure 7A, lane 1). In contrast, the RPE from Oa1-/mice only showed the band corresponding to the 32 P-ADP ribosylated Gai2. No sign of Gai3 or Gai1 was present on the autoradiograph ( Figure 7A, lane 2). To corroborate these results, a Western blot was prepared with the same samples, using the Ga common antibody. Figure 7B shows the Gai2 and Gai1 plus Gai3 bands in the control RPE (lane 1), but in Oa1-/-RPE, comparable levels of Gai2 but a very minimal amount of either or both Gai1 and Gai3 are seen (lane 2).

Discussion
We have previously demonstrated by in-vivo studies on mice that the heterotrimeric G protein Gai3 signals in the same transduction pathway controlled by Oa1 to regulate melanosomal biogenesis and axonal growth through the optic chiasm [12]. However, the interaction between Oa1 and the two other members of the Gai family of proteins, Gai1 and Gai2, remained to be explored. In this paper, taking advantage of the availability of knockout mice for each of the Gai genes, and of DKO (Gai1-/-, Gai3-/-) mice, we tested the hypothesis that Oa1 transmits its signal through one specific Gai protein and demonstrated with in-vitro and in-vivo experiments that this protein is indeed Gai3.
To investigate the involvement of all Gai proteins in the regulation of size, density and shape of RPE melanosomes, electron micrographs of the RPE from 3-month-old Gai1-/-, Gai2-/-, Gai3-/-, DKO and 129 Sv were analyzed. With regard to size, our results indicate that only the loss of function of Gai3 significantly increases the number of macromelanosomes in the RPE. Even with loss of function of both Gai1 and Gai3, no added effect in the size of DKO melanosomes was observed when compared to melanosomes of Gai3-/-RPE, suggesting that Gai1 is not involved in establishing the size of RPE melanosomes. We then looked at the number of melanosomes per unit area of RPE, and found that melanosomal density in Gai1-/and Gai2-/were not significantly reduced from that in 129 Sv wild-type RPE. Conversely, melanosomal density was reduced significantly and similarly in Gai3-/and in DKO RPEs when compared to wildtype RPE. This again supports the notion that Gai1-/does not contribute to the RPE melanosomal density decrease observed in DKO mice. We also established that despite the appearance of fewer round melanosomes in Gai2-/-RPEs, there is not a major change in the percent frequency of round melanosomes in all Gai-/-RPEs when compared to those in the wild-type 129 Sv. In addition, as demonstrated by electroretinography, the retinas of Gai3-/as well as those of DKO mice are functional despite the phenotype of their RPEs ( Figure 3A) similar to what was previously observed in the ERG of Oa1-/-mice [12]. Thus, these results together lead us to conclude that of the three heterotrimeric Gai proteins, Gai3 is the Oa1-associated protein involved in the regulation of RPE melanosomal size and density.
Potential binding sites for heterotrimeric G-proteins on GPCRs have been localized to the cytoplasmic loop 3 (i3) and carboxyterminal tail (CT) of the seven-transmembrane receptors [14]. These intracellular segments are the ones that determine the interaction with specific G proteins, and as a result, which of several possible signaling pathways are activated [17,18,19,20]. Therefore, we generated GST-fusion proteins with the Oa1-i3 and Oa1-CT segments and used them in GST pull-down assays to determine which Gai protein bound in-vitro to Oa1. Results of these experiments showed that of the three heterotrimeric Gai proteins, Gai3 is the only one that binds specifically to Oa1. Most important, immunoprecipitation of RPE proteins using Oa1 antibodies, followed by SDS-PAGE and mass spectrometry analysis of the proteins, brought down Gai3 together with Oa1, conclusively identifying Gai3 as an Oa1-specific interacting protein. These results were further confirmed by immunoprecipitation experiments of RPE proteins using Gai3 antibodies, which also demonstrated that Oa1 co-immunoprecipitated with Gai3. Collectively, these data indicate that the GPCR Oa1 initiates in the RPE the signal transduction cascade that controls melanosomal size and density through activation of Gai3.
Compensatory increases elevating a Gai subunit level in a tissue when another Gai subunit is missing from it have been previously observed [16]. This raises the possibility that in the RPE of Gai knockout mice, at the protein level, compensatory expression of the other Gai subunits may reduce the effect that the loss of a particular Gai protein has on the control of melanosomal size and density. To test this, we measured the levels of the three Gai proteins in the RPEs of Gai1-/-, Gai2-/-, Gai3-/and DKO mice using Western blotting with a Gai common antibody. Our results suggest that at the protein level, only the loss of Gai1 from the RPE may be compensated for by Gai3. However, with respect to functional compensation of the melanosomal phenotype, we cannot conclude that Gai3 functionally compensates for the loss of Gai1 since our results show that Gai1-/melanosomes are not significantly different from those observed in control mice, even though they are a bit larger. Similarly, the fact that the DKO phenotype is no worse than that of Gai3-/alone with respect to size, density or shape of RPE melanosomes, suggests that Gai1 is not critical for the regulation of these parameters. Furthermore, compensatory increases, if any, in Gai1 protein in the Gai3-/retina seem to play no role in melanosome biogenesis. Although Gai2 levels are considerably increased in the Gai1-/animals and to a quite lesser extent in Gai3 knockouts, we have shown in the in vivo experiments that Gai2 is not involved in the determination of melanosome size. Given the high levels of this protein in the RPE, it must play an important, though different, role in this tissue than Gai3, a role having quite discriminable effects, evidenced in the ERG itself ( Figure 3B).
ADP-ribosylation of Ga subunits is a covalent modification catalyzed by PTX in which an ADP-ribose moiety is attached to the C-terminus of the protein. This prevents the Gi proteins to interact with their receptors and, therefore, it causes functional inactivation of all signal transduction pathways. ADP-ribosylation with NAD 32 P-labeled on its ADP-ribose moiety allowed us to tag Gai subunits in membranes and to learn about their involvement in cellular responses. We carried out this reaction using RPE membranes of Oa1-/and their congenic B6/NCrl mice. Interestingly, our results showed that the three Gai subunits are present in the RPE of control mice, but that only Gai2 is in the Oa1-/-RPE. Similarly, when we used samples from the same RPE membranes to perform Western blotting with the Ga common antibody we found that Gia2 levels in -/-Oa1 RPE are comparable to those in the control RPE, but that there is very minimal, if any, amount of either or both Gai1 and Gai3 in the Oa1-/-RPE. These interesting results will lead to a whole series of studies in the future.
It is well established that in addition to their important roles in many pathways of transmembrane signaling, where they participate in processing and sorting of incoming signals as well as in adjusting the sensitivity of the signaling system, heterotrimeric G-proteins are also localized to the Golgi complex [21], where they are involved in the formation of secretory vesicles from the trans-Golgi network (TGN) [9]. Gai3, in particular, acts as an inhibitor of intra-Golgi and post-Golgi trafficking [22], and has been found, among other places, in the membranes of secretory vesicles in pancreatic acinar cells, from where it facilitates the fusion between zymogen granules and/or the expulsion of vesicular contents [23].
The specific function of Gai3 in the RPE has not been identified. Our working hypothesis is that after the bulk of Oa1, together with tyrosinase, TRP1 and other compounds involved in melanin synthesis, has reached the stage II melanosomes, Oa1 starts to activate Gai3, which in turn inhibits the traffic of vesicles carrying the membrane proteins required for melanization from the TGN. However, melanization occurs with no problem in stage III and IV melanosomes, as they use the proteins already present. This would explain why the levels of Oa1 decrease from those in stage II melanosomes to those in stages III and IV [24]. Oa1 is not being renewed in the melanosomes after its activation of Gai3 because there are no longer vesicles bringing this protein to the melanosomes. Thus, the function of Gai3 in the RPE would be to control the size of melanosomes through the inhibition of vesicle trafficking from the TGN to the melanosome, a function previously thought to be carried by Oa1 [25].
On the basis of our results and the accumulated information in the literature about GPCRs, Gai3 functions and melanogenesis, we are proposing a hypothetical model for the beginning of the Oa1 signaling cascade ( Figure 8A).
In this model, at the end of stage II melanogenesis, when melanosomes have acquired their melanogenic proteins and the bulk of Oa1, a signal must turn on an endogenous lumenal ligand to activate Oa1 at the melanosomal surface membrane. Like in other GPCR cascades, activated Oa1 will cause the exchange of GDP for GTP on the Gai3 subunit of the heterotrimeric G protein localized to the surface of the same membrane, activating Gai3 (Gai3*) -which then separates from the bc subunits-and leading to the release of Gai3* and Gbc to the RPE cytoplasm. Gai3* will at this stage inhibit the vesicular traffic of membrane proteins from the TGN to the melanosome. Prenylation of the Gc subunit will target the Gbc complex to the endoplasmic reticulum, where it will be processed fully before it is delivered to the Golgi. There, Gbc will bind Gai3 and after post-translational modification of Gai3, the G protein heterotrimer may leave the TGN and get transported to the melanosome surface membrane using the classic secretory pathway.
This model could also explain the presence of macromelanosomes in the RPE of ocular albinism patients or Oa1 and Gai3-/mice ( Figure 8B). Mutations in the OA1 gene in humans could render the OA1 protein incapable of activating the heterotrimeric Gai3 on the surface membrane of the melanosome. The same effect would be observed in the absence of Oa1 or Gai3 in knockout mice. Without the active form of Gai3*, inhibition of the vesicular traffic of melanin-related proteins to stage II melanosomes cannot occur, and the continuous supply of this material to the melanosome would result in the formation of abnormally large organelles, the macromelanosomes.
In summary, while the precise function of Gai3 in RPE melanosome biogenesis remains to be delineated, it is clear that this protein plays an important role in the control of the size of RPE melanosomes. As a consequence, Gai3 is also controlling RPE pigmentation, which seems to be necessary during embryonic stages for the proper decussating pattern of the optic axons. Because Gai3-/mice, like Tyr-/and Oa1-/mice, all show abnormalities in the decussation of their optic axons [12], Gai3 appears to be the common effector by which these three distinct RPE phenotypes affect the retinal ganglion cells as their axons navigate the optic chiasm.

Ethics Statement
All experiments involving mice were carried out using protocols approved by the UCLA Animal Research Committee, and in accordance with the ARVO Statement for Use of Animals in Ophthalmic and Vision Research.
Animal and human tissues C57BL/6NCrl (B6/NCrl) mice and congenic Oa1 knock-out mice (Oa1-/-) were obtained from The Charles River Labs, USA and Italy, respectively, and bred at UCLA. Gai1-/-, Gai2-/-, Gai3-/-and the DKO mice were previously generated on the 129 Sv background. The genotype of these mice was determined by Southern blot analysis of mouse tail genomic DNA as described by Jiang et al. [26]. Mice were housed and bred in conventional cages and environmental conditions at the animal facilities of UCLA.  Healthy human donor eyes were obtained from the National Disease Research Interchange (Philadelphia, PA) and immediately frozen in liquid nitrogen. The donor eyes were handled in compliance with the Declaration of Helsinki.

Electron Microscopy
3 month-old 129 Sv, Gai1-/-, Gai2-/-, Gai3-/and DKO mice were deeply anesthetized by an intraperitoneal injection of 120 mg/kg sodium pentobarbital, and perfused intracardially with 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.4. Eyes were enucleated, rinsed in 0.1 M phosphate buffer, post-fixed with 1% buffered osmium tetroxide, dehydrated in graded ethanol, and embedded in araldite 502. Sections for electron microscopy (60-70 nm) were cut on a Leica Ultracut UCT and collected on 200 mesh uncoated copper grids. For the ultrastructural analysis, stained sections (5% uranyl acetate and 0.4% lead citrate) were observed with a 910 Zeiss electron microscope. The RPE fields analyzed were photographed using a Keenview TM digital camera. For quantification of melanosomes and determination of their area in the RPE, we analyzed 50 micrographs for each 129 Sv and DKO mice and 30 micrographs for each Gai1-/-, Gai2-/and Gai3-/mice at 16,0006 magnification using the analySIS TM software for LEO 900 TEM, version 3.2. (Soft Imaging System, Lakewood, CO). Images were cropped with Adobe Photoshop (Adobe Systems Inc., San Jose, CA). A total of 2501 129 Sv, 1285 Gai1-/-, 1609 Gai2-/-,1115 Gai3-/and 1997 DKO melanosomes were studied, sampling 10 eyes for each 129 Sv and DKO mice and 6 eyes for each Gai1-/-, Gai2-/and Gai3-/mice. The data for the Gai3-/mice were previously reported, but are included herein for direct comparison.
Melanosomal size and density. We used the magic wand feature of the Soft Imaging System analySIS software to select and measure the area of every individual melanosome and of the total RPE area containing the melanosomes in each micrograph, as we have done previously in the analyses of Gai3-/and the Oa1-/mice and their corresponding controls [12]. We analyzed the difference in the melanosomal size, the percent frequency of melanosomes in the .5000 nm 2 size range, and the number of melanosomes per RPE area among five mouse groups: 129 Sv, Gai1-/-, Gai2-/and Gai3-/and DKO. Comparisons among all groups were done using a one way analysis of variance (ANOVA) followed by post-hocTukey tests to identify significant differences between individual pairs of groups, conservatively using animal averages for each measure and thus an n of 5 for the 129 Sv and DKO groups and an n of 3 for the Gai1-/-, Gai2-/and Gai3-/groups. A p-value of less than 0.05 was considered to be statistically significant.
Melanosomal Morphology. To determine melanosomal shape we used the particle detection analysis of the Soft Imaging System analySIS software for LEO 900 TEM, version 3.2. The classification of the shape of the organelles is given by how round the organelle is. A melanosome with a shape factor of 1 is round, while elliptical melanosomes have a shape factor less than 1. We compared the percentage of round melanosomes for each particular mouse line. The differences among all mouse groups [129 Sv, Gai1-/-, Gai2-/-, Gai3-/and DKO] were compared using one-way ANOVA and post-hoc Tukey tests, all as above.

Electroretinography
Mice were anesthetized with an intraperitoneal injection of xylazine (0.5 mg/ml) and ketamine (1 mg/ml) in normal saline. In adult mice, a dose of 0.1 ml was administered. Body temperature was maintained at 38uC with a heating pad. Pupils were dilated with Atropine (1%). A gold-wire electrode was placed on the corneal surface of the right eye and referenced to a gold wire in the mouth. A needle electrode in the tail served as the ground. Responses were amplified (Tecktronix AM 502 Differential Amplifier, 610,000) band pass filtered (0.1-300 Hz), digitized using an I/O board (PCI-6221, National Instruments, Austin, TX) in a personal computer, and averaged. A signal rejection window was used to eliminate electrical artifacts. All stimuli were presented in a large integrating sphere painted with a highly reflective white matte paint (#6080, Eastman Kodak Corporation, Rochester, NY). Rod mediated responses were obtained with blue flashes (Wratten 47A; I max = 470 nm) varied over an intensity range of 3.5 log units. Cone-mediated responses were obtained with white flashes on a rod saturating background (32 cd/m 2 ).
Production of GST::OA1-i3 and GST::OA1-CT recombinant proteins Mouse Oa1 cDNA was subcloned into the mammalian expression vector pCDNA3.1/V5-His-Topo using EcoRI and PstI sites in the polylinker. Using this construct, a 108 bp DNA fragment corresponding to the third cytosolic loop (i3) of OA1 (residues 213-248), with flanking fragments corresponding to the BamHI and SalI restriction enzymes, was amplified by PCR using the following primers: Forward, 59-TGGA TCCTTTCACAA-GACAGTGACTTCA-39; and Reverse, 59-AGTCGACTCAT-TTG AAAAAACGGGTCTTGAT-39. Similarly, a 275 bp DNA fragment corresponding to the C terminal (CT) of OA1 (residues 314-405), with flanking fragments corresponding to the BamHI and SalI restriction enzymes, was amplified by PCR using the following primers: Forward, 59-TGGATCCACAGGATG-CAGCCTGGATGTC-39; and Reverse, 59-AGTCGACTCA-GAGTTCCCCCTGGGCTTGGGA-39. The 25 ml reaction contained 2.5 mg of each 59 primer and 39 primer, 100 ng/ml pCDNA3.1/V5-His-Topo-OA1, 0.6 Units of Taq DNA polymerase (Invitrogen Corporation, Carlsbad, CA), 16 PCR buffer, 2.5 mM MgCl 2 and 250 mM dNTP mix. After 30 cycles (melting for 3 min at 94uC; annealing for 1.5 min at 53uC; extension for 1 min at 72uC), the PCR reactions were subjected to an additional 5 min incubation at 72uC. Each PCR reaction yielded a single product that was subsequently isolated and gel purified, cut with BamHI and SalI, and subcloned into the pGEX-4T-2 vector (Amersham Bioscience, now GE Healthcare, Piscataway, NJ) using T4 ligase. Ligation reactions were used for transformation of BL21 (DE3) E. coli, and were selected on LB-ampicillin agar plates. Positive clones were screened by restriction analysis and DNA sequencing.

Preparation of GST-Fusion Proteins
E. coli cultures carrying the pGEX-4T-2 constructs were grown until they reached mid-logarithmic growth phase (0.4-0.5 A 600 ). Isopropyl-b-D-thiogalactopyranoside (IPTG) was then added to a final concentration of 0.1 mM to induce GST protein expression, and the cultures were further incubated for 4 h at 30uC. GST was Figure 8. Proposed hypothetical model of the Gai3 regulation of RPE melanosomal size. A) An unknown luminal ligand (L?) turns on wildtype Oa1, which activates Gai3. The active Gai3 in turn inhibits vesicular traffic of proteins from the TGN to the melanosome, controlling in this way its size. B) Mutated OA1 is unable to activate Gai3 and, thus, the continuous supply of melanin-related proteins to the melanosome results in the formation of very big organelles, the macromelanosomes. doi:10.1371/journal.pone.0024376.g008 purified from cultures incubated at 37uC for 3 h. The GST::Oa1-i3 fusion protein was obtained as previously described [27] with few modifications. Briefly, bacteria were harvested by centrifugation at 4,000 g for 15 min and the pellets were resuspended in 2 ml of STE Buffer (10 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, 25 mg/ml PMSF and protease inhibitors cocktail) containing 50 mg/ml of lysozyme (added immediately prior to resuspension), and incubated on ice for 15 min. Dithiothreitol (DTT) was then added to a final concentration of 2.5 mM. Bacteria were lysed by the addition of 1.0% N-lauroylsarcosine (Sarkosyl), disrupted on ice for 1 min using a MISONIX Sonicator 3000 (power level 3, 50% cycle) and centrifuged at 13,000 g for 10 min at 4uC to remove insoluble debris. To obtain the GST::Oa1-CT fusion protein, a similar procedure was followed, but no lysozyme or Sarkosyl was used.

Purification of GST-tagged proteins
GST-tagged proteins were purified from bacterial lysates by affinity chromatography using immobilized Glutathione Sepharose TM High Performance (GE Healthcare). 1 ml GSTrap HP column was equilibrated with 5 ml of binding buffer (PBS, pH 7.3) and 1 ml of bacterial lysate was loaded onto it. The column was washed with 5 ml of binding buffer and the GST-tagged protein was eluted with 3 ml of elution buffer (50 mM Tris-HCL, 10 mM reduced glutathione, pH 8.0).

GST activity assay
The GST specific activity was measured spectrophotometrically using the GST assay kit (Sigma-Aldrich Corporation, St. Louis, MO) according to the manufacturers' instructions, using 1 mM 1chloro-2,4-dinitrobenzene (CDNB) and 2 mM reduced-glutathione (GSH) as substrates at 340 nm and 25uC. Fractions with good GST activity (,600-800 mg/ml) were used for incubations with invitro transcribed/translated 35 S-methionine-labeled Ga proteins (see below).

In-vitro synthesis and 35 S-methionine-labeling of Ga Proteins
Gai1, Gai2, Gai3, Gas and Gaq were in-vitro transcribed, translated and labeled with 35 S-methionine using the pAGA-2 vector containing the cDNA of the corresponding Ga protein and the TNT Coupled Reticulocyte Lysate system (Promega, Madison, WI) with T7 RNA polymerase. The resulting 35 S-labeled proteins (1.5 mg of each) were boiled in 26 SDS-PAGE sample buffer containing 3% b-mercaptanol, loaded onto a 10% Tris-Glycine gel and electrophoresed at 45 V until the dye had run out of the gel. After washing with water for 10 min (3 times), and immersing in 5% glycerol for 15 min, the gel was dried and visualized by autoradiography. The relative expression of 35 S-labeled Ga proteins was analyzed by densitometric measurement using Quantity One 1-D analysis software (version 4.4, Bio-Rad, Hercules, and CA). The mean density for each band, expressed as number of pixels, was used to calculate the ratio of each protein to the 35 S-labeled Ga protein that had the lowest number of pixels (ratio = 1). This number was used to determine the appropriate volume of each 35 S-labeled Ga protein that was used for the GSTpull down assay.

GST-pull-down assays
For in-vitro protein-protein interaction assays, 10 ml of each 35 S-Ga-protein expressed from the pAGA-2 vector containing the same amount of protein were incubated with 50 ml of Immobilized Glutathione agarose resin (Thermo Fisher Scientific Inc., Wal-tham, MA), ,1 mg of GST protein or the GST-fusion protein (GST::Oa1-i3 or GST::Oa1-CT), and 100 ml of Binding buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 0.5% Triton 1006) for 2 h at room temperature (RT). The beads were recovered by centrifugation at 3,000 g, 4uC, for 1 min, washed for 5 min (5 times) with 500 ml of ice-cold Binding buffer, and centrifuged again for 1 min at 8,000 g. The protein-protein interacting complex was eluted from the beads with 30 ml of 26 SDS-PAGE sample buffer containing 3% b-mercaptanol and its components were separated by SDS-PAGE as above, on a 10% Tris-glycine gel. The gel was fixed and Coomassie blue-stained to confirm that the fusion proteins were present in comparable amounts and had the right molecular weight. After destaining, the gel was dried and exposed to X-ray film overnight (280uC) to visualize the interacting 35 S-Ga-proteins.

Mass Spectrometry Analysis
Several steps were followed to prepare samples for mass spectrometry: Immunoprecipitation and Western Blotting. Normal human RPE was dissected from frozen donor human eyes. Tissue was homogenized on ice with a Dounce homogenizer in 500 ml RIPA lysis buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% Na-deoxycholate, 0.1% SDS, 1 mM PMSF, 16 complete protease inhibitor cocktail (Roche Molecular Biochemicals, Indianapolis, Indiana), and 5 mM N-ethylmaleimide (Sigma-Aldrich Corporation). Homogenates were pre-cleared with 100 ml of protein G-Sepharose (Thermo Fisher Scientific, Inc.) for 1 h at 4uC. Lysates (500 ml) were first incubated with 17 ml of anti-OA1 (ABCAM, Cambridge, MA) or with 17 ml anti-Gai3 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) rabbit polyclonal primary antibodies, at 4uC overnight, and then with 50 ml of protein G-Sepharose for 2 h at RT. The complex-bound resin was washed 5 times with IP buffer (25 mM Tris-HCl, 150 mM NaCl; pH 7.2) and 3 times with water. Immunoprecipitated complexes were eluted with 26 SDS-PAGE sample buffer containing 3% b-mercaptanol and each of duplicate aliquots were loaded onto a different half of a 12% Tris-glycine gel. SDS-PAGE was carried out overnight at 44 V. Half of the gel was stained with Coomassie blue to visualize protein bands for mass spectrophotometry analysis and the other half was transferred to a nitrocellulose membrane (Hybond ECL, Amersham Biosciences) overnight at 33 V. The blot was then incubated with 1:10,000 anti-Gai3 or anti-OA1 rabbit primary antibodies, and with 1:10,000 goat anti-rabbit secondary antibodies. The Enhanced Chemiluminescence (ECL) detection reagent (Amersham Biosciences) was used to visualize the bands.
Tryptic Digestion. Protein bands of interest on the Coomassie blue-stained gel were excised and placed in microtubes. In-gel digestion with trypsin was performed according to standard procedures routinely used in the Pasarow Mass Spectrometry core facility at UCLA. Briefly, the gel pieces were distained with 10 ml acetonitrile (ACN) for 30 min at RT and treated with 10 mM DTT/100 mM NH 4 HCO 3 (200 ml) for 1 hour at 37uC. Samples were then alkylated with 55 mM iodoacetamide (Sigma-Aldrich Corporation). The gel pieces were washed with 100 mM NH 4 HCO 3 for 15 min at RT, dehydrated with ACN and digested with 1.25 mg trypsin (Promega) in 100 ml of 50 mM NH 4 HCO 3 , on ice, for 45 min. They were then incubated overnight in 10 ml of 50 mM NH 4 HCO 3 , without trypsin. The digests were extracted twice with 100 ml of 50% ACN/0.5% formic acid at RT for 60 min with constant mixing, the extracts were pooled and dried and each sample was then reconstituted with 8 ml of 2% formic acid and sent for nano-liquid chromatography tandem mass spectrometry (nLC-MSMS) analysis at the Pasarow core facility.
Database Analysis. The mass spectra were searched against a human trypsin indexed database, with variable modifications of carboxyamidomethylation, methionine oxidation, and deamination of asparagine residues using the Bioworks software (Thermo Fisher) based on the SEQUEST algorithm implemented in Discoverer software (Thermo Fisher). Spectra were also searched using Mascot software (Matrix Science, UK) and results with p,0.05 (95% confidence interval) were considered significant and indicating identity.

Preparation of RPE Membranes
The RPEs from Gai1-/-, Gai2-/-, Gai3-/-, DKO, Oa1-/and their corresponding 129 Sv and B6/NCrl control mice (12 eyes of each), were dissected, collected and frozen in liquid nitrogen. Each sample was homogenized in 250 ml of 27% (wt/wt) sucrose/1 mM EDTA/10 mM Tris?HCl, pH 7.5, in Dounce homogenizers. Homogenates were centrifuged for 5 min at 1,000 g and the supernatants were again centrifuged at 12,000 g for 20 min to obtain the RPE membranes. Quantification of protein from melanin-containing RPE membranes was carried out by the method of Sedmak and Grossberg, which depends on the conversion of Coomassie brilliant blue G-250, in diluted acid, from a brownish-orange to an intense blue color [28]. With this very sensitive method the melanin interference is negligible. The G-250 dye was prepared as a 0.06% solution in 19% perchloric acid (w/v) and was filtered through Whatman No. 1 filter paper to remove any undissolved material. The assay consisted of adding 0.5 ml of the G-250 dye to 5 ml of homogenized RPE membranes, mixing immediately, and determining absorbance at 620 nm.

Western blot analysis
The Gai1, Gai2, and Gai3 protein levels in RPE membranes from 129 Sv, Gai1-/-, Gai2-/-, Gai3-/and DKO mice, were measured using an anti-Ga common antibody (Cell Signaling technology, Danvers, MA) that recognizes all the Gai and Gao proteins. Proteins (15 mg per lane) were separated by SDS-PAGE as above on 9% acrylamide/bis-acrylamide gels containing 6 M urea and blotted onto nitrocellulose. Blots were incubated with 1:2,000 anti-Ga common polyclonal anti-rabbit primary antibody, overnight, at 4uC and with 1:5,000 goat anti-rabbit secondary antibody for 3 hours at RT. The ECL detection reagents were used to visualize the bands.