The BLOC-3 subunit HPS4 is required for activation of Rab32/38 GTPases in melanogenesis, but its Rab9 activity is dispensable for melanogenesis

HPS4 biogenesis of lysosome-related organelles complex 3 subunit 2 (HPS4) is one of the genes whose mutations have been associated with Hermansky–Pudlak syndrome (HPS), characterized by ocular albinism and susceptibility to bleeding because of defects in the biogenesis of lysosome-related organelles such as melanosomes. HPS4 protein forms a BLOC-3 complex with HPS1, another HPS gene product, and the complex has been proposed to function as a guanine nucleotide exchange factor (GEF) for RAB32, a member of the Rab small GTPase family (Rab32), and Rab38 (Rab32/38-GEF) and also as a Rab9 effector. Although both Rab32/38 and Rab9 have been shown previously to be involved in melanogenesis in mammalian epidermal melanocytes, the functional relationships of these small GTPases with BLOC-3 remain unknown. In this study, we used site-directed mutagenesis to generate HPS4 mutants that specifically lack either Rab32/38-GEF activity or Rab9-binding activity and investigated their involvement in melanogenesis of melan-le cells (an HPS4-deficient melanocyte cell line derived from light ear mice). Melan-le cells exhibit a clear hypopigmentation phenotype, i.e. reduced expression and abnormal distribution of tyrosinase and reduced melanin content. Although re-expression of WT HPS4 completely rescued this phenotype, the Rab32/38-GEF activity–deficient HPS4 mutant failed to restore melanin content and tyrosinase trafficking in these cells. Unexpectedly, as WT HPS4, the Rab9 binding–deficient HPS4 mutant completely rescued the phenotype. These results indicate that activation of Rab32/38 by HPS4 (or BLOC-3) is essential for melanogenesis of cultured melanocytes and that Rab9 likely regulates melanogenesis independently of HPS4.

BLOC-3, a heterodimeric complex of HPS1 and HPS4, has recently been reported to be a specific GEF for Rab32 and Rab38 (Rab32/38) (16). Interestingly, BLOC-3 has also been reported to be a Rab9 effector (17), and involvement of Rab9A in melanogenesis has been reported by two independent groups (18,19). Based on these findings, Rab9A and Rab32/38 have been proposed to constitute a Rab cascade through BLOC-3 in melanogenesis (11,16,18,20), the same as Rab5 and Rab7, which constitute a Rab cascade through Mon1-Ccz1 in endosome maturation (21). However, whether the BLOC-3mediated Rab cascade actually functions in melanocytes has never been experimentally investigated.
To clarify the functional relationship between HPS4 and these Rabs, in this study we performed rescue experiments by using genetically HPS4-deficient melanocytes (i.e. melan-le cells, which have reduced melanin content and a defect in tyrosinase trafficking) and HPS4 mutants specifically lacking Rab32/38 activation ability or Rab9A-binding ability. The results showed that a Rab9 binding-deficient HPS4 mutant completely restored melanogenesis in melan-le cells but that a Rab32/38-GEF activity-deficient mutant failed to restore either melanin content or tyrosinase trafficking. Our findings indicate that the Rab32/38-GEF function of HPS4 is independent of its Rab9A binding, and an alternate role of Rab9A in melanogenesis is discussed.

HPS4, not HPS1, is an active Rab9-specific binding protein
HPS4, a component of the BLOC-3 complex (simply called BLOC-3 hereafter), has been shown to directly interact with Rab9A and Rab9B but not with Rab5A, Rab7, or Rab27A (17). However, because mammals contain ϳ60 different Rabs, HPS4 may be capable of interacting with other Rabs with higher affinity than with Rab9. To determine the Rab binding specificity of HPS1 and HPS4, we first performed yeast two-hybrid assays by using constitutively active (CA) and constitutively negative (CN) mutants of 62 different mammalian Rabs as bait and HPS1 and HPS4 as prey, as described previously (22)(23)(24). The results of the comprehensive screening showed that HPS4 strongly interacted with an active form of Rab9A/B (Fig. 1, boxed) consistent with a previous report (17). A negative form of Rab40B/C appeared to be positive for HPS4, but because of their relatively weak interactions, we did not pursue them in the subsequent analysis. By contrast, none of the CA/CN Rabs interacted with HPS1 ( Fig. 1). We therefore concluded that HPS4, not HPS1, is an active Rab9-specific binding protein.

Mapping of the Rab9-binding domain in HPS4 by truncation analysis
We next attempted to identify the minimum region of HPS4 that is required for Rab9 binding by performing a systematic truncation analysis ( Fig. 2A). HPS4 contains an N-terminal longin domain (16), but no known protein motifs are present in the large C-terminal portion. We therefore prepared five N-terminal truncated mutants (⌬N1-⌬N5) and five C-terminal truncated mutants (⌬C1-⌬C5) and performed yeast two-hybrid assays ( Fig. 2A). Interestingly, deletion of either the N-terminal 200 amino acids or the C-terminal 200 amino acids of HPS4 completely abrogated its Rab9 binding activity (Fig. 2B). Further truncation analysis indicated that amino acids 150 -570 of HPS4 ( Fig. 2A, dashed box) are necessary and sufficient for its Rab9 binding activity (Fig. 2C).

Identification of the amino acid residues in HPS4 for Rab9 binding by site-directed mutagenesis
The minimum Rab9-binding domain identified above contains ϳ400 amino acids, and it is clearly larger than the usual Rab-binding domains (ϳ100 amino acids) of the Rab effectors that function in melanocytes, e.g. the Rab27-binding domain of Slac2-a/melanophilin, the Rab32/38-binding domain of Varp, and the Rab36binding domain of Rab interacting lysosomal protein (RILP) (23,25,26). Because these known Rab-binding domains are well conserved in different species of vertebrates, we compared seven HPS4 sequences from different vertebrate species: human, bovine, Figure 1. Rab binding specificity of HPS1 and HPS4. The Rab binding specificity of mouse HPS1 and HPS4 was determined by yeast two-hybrid assays. Yeast cells containing both pGBD-C1 plasmids, which express a CA form (which mimics the GTP-bound form) or a CN form (which mimics the GDP-bound form) of mouse or human Rab lacking the C-terminal Cys residues (positions are indicated in the left panel), and pAct2 plasmids, which express mouse HPS1 or HPS4, were streaked on SC-AHLW (lacking adenine, histidine, leucine, and tryptophan) selection medium and incubated at 30°C. Note that HPS4 specifically interacted with the active form of Rab9A and Rab9B (boxed), whereas neither Rab interacted with HPS1.

G55W/G59M mutations in the longin domain of HPS4 impaired Rab32/38-GEF activity
It has been shown previously that the N-terminal longin domain of HPS4, which is not required for Rab9 binding (Fig. 2), is involved in GEF activity toward Rab32/38 (16). To investigate whether the longin domain and Rab9-binding domain of HPS4 are functionally related, we also attempted to produce its Rab32/38-GEFactivity-deficientmutantbysite-directedmutagenesis. Although nothing is known about the residues of HPS4 that are critical for its Rab32/38-GEF activity, we utilized recent structural data in regard to the thermophilic fungus Chaetomium thermophilum Ccz1 (27) because Ccz1 contains a similar longin domain, and Ccz1, together with Mon1, functions as a heterodimeric GEF for Rab7/Ypt7 (21,28). Comparison of the  (16). The Rab9(CA/CN) binding activity of each HPS4 truncation mutant is summarized at the right of each construct. B and C, Rab9 binding activity of HPS4 truncation mutants. Yeast cells containing pGBD-C1-Rab9A/B(CA/CN)⌬Cys plasmids and pAct2 plasmids expressing HPS4 truncation mutants were streaked on SC-LW (lacking leucine and tryptophan) or SC-AHLW medium (selection medium) and incubated at 30°C. Note that amino acids 150 -570 of HPS4 are the minimal Rab9-binding site.

Structure-function analysis of HPS4 in melanogenesis
longin domains of HPS4 and Ccz1 revealed that several amino acids were highly conserved, and we especially focused on Gly-55 and Gly-59 of HPS4 (Fig. 5A) because these two Gly residues are critical for the GEF activity of CtCcz1, and the yeast Ccz1(G47W/G51M) mutant is unable to rescue a vacuolar fragmentation phenotype of ccz1⌬ cells (27). We therefore introduced similar mutations into the longin domain of HPS4 (G55W and G59M) by site-directed mutagenesis (Fig. 5A, arrowheads). Although the Sec2 and VPS9 GEF domains have been demonstrated to physically interact with a CN form of Rabs (mimics their substrates) by conventional yeast two-hybrid assays (23,29,30), HPS4 alone was unable to interact with CN-Rab32/38 (Fig. 1). Because heterodimer formation by HPS1 and HPS4 is required for Rab32/38-GEF activity (16), both BLOC-3 components would be necessary for the interaction with CN-Rab32/38. To confirm that they are, we evaluated the interaction between BLOC-3 and CN-Rab32/38 or CA-Rab32/38 by performing yeast tri-hybrid assays as described under "Experimental procedures." As expected, BLOC-3 containing the WT HPS4 specifically interacted with CN-Rab32/38 but was unable to interact with CA-Rab32/38 or CA/CN-Rab7 (Fig. 5B, top row). In contrast, BLOC-3 containing the HPS4(G55W/G59M) mutant failed to interact with CN-Rab32/38, even in the presence of HPS1 (Fig. 5B, center row), suggesting that the HPS4(G55W/G59M) mutant lacks Rab32/38-GEF activity. Because these two mutations are located in the longin domain, which, as described above, is not essential for Rab9 binding (Fig. 2), the HPS4(G55W/G59M) mutant interacted normally with Rab9 (Figs. 4A, lane 3, and 5C) in addition to HPS1 (Fig. 4B, lane 5). In contrast, the Rab9 bindingdeficient HPS4(S541A/N543A) mutant still interacted with CN-Rab32/38 in the presence of HPS1 (Fig. 5B, bottom row), suggesting that this mutant retains Rab32/38-GEF activity. We therefore concluded that the G55W/G59M and S541A/ N543A mutants of HPS4 would be useful tools for analyzing the functional relationships between the Rab32/38-GEF activity and Rab9 binding activity of HPS4 in cultured melanocytes.

Structure-function analysis of HPS4 in melanogenesis Rab32/38-GEF activity, not Rab9 binding activity, of HPS4 is essential for melanogenesis in melanocytes
In the final set of experiments, we investigated whether the HPS4 mutants described above could support melanogenesis in melanocytes. To do so, we focused on an HPS4-deficient immortal melanocyte cell line (melan-le), which was established from a light ear mouse (31), and performed rescue experiments by stably expressing WT and mutant HPS4 in melan-le cells by retrovirus infection. Because of the absence of functional HPS4 protein in melan-le cells, melan-le cells should contain a reduced amount of active Rab32/38. To confirm this, we first observed the intracellular localization of endogenous Rab32 by performing an immunofluorescence analysis (Fig.  6A). We did not test Rab38 because Rab32 and Rab38 have been shown to be functionally redundant in melanocytes (15, 18, 32) and because our anti-Rab38 antibody does not function well in immunofluorescence analyses. As expected, re-expression of HPS4 with an HA tag in melan-le cells restored the perinuclear punctate localization of Rab32 (Fig. 6A, bottom left), whereas hardly any Rab32 signals were observed in control melan-le cells (Fig. 6A, top left), presumably because inactive Rab32 is mostly present in the cytosol. It should be noted that the GEF activity-deficient HPS4(G55W/G59M) mutant failed to restore punctate localization of Rab32 (Fig. 6A, bottom right). Unexpectedly, however, the Rab9 binding activity-deficient HPS4(S541A/N543A) mutant fully restored Rab32 localization (Fig. 6A, top right), suggesting that Rab32 is correctly activated and localized at the proper sites in HPS4(S541A/N543A)-reexpressing cells. To validate the immunofluorescence data, we performed GTP-Rab32 pulldown assays using the GTP-Rab32binding domain of Varp (RBD32), as described previously (18), to check the activation status of Rab32 in WT and mutant HPS4-expressing cells. Although hardly any active Rab32 was trapped by the RBD32 beads in control melan-le cells (Fig. 6B, lane 1), re-expression of HPS4(WT) in melan-le cells clearly increased the active Rab32 level (Fig. 6B, lane 2). Similarly, reexpression of the HPS4(S541A/N543A) mutant in melan-le cells also increased the active Rab32 level (Fig. 6B, lane 3), and no apparent difference in the amount of GTP-Rab32 was observed between HPS4(S541A/N543A)-and HPS4(WT)-expressing cells. In contrast, the GEF activity-deficient HPS4(G55W/G59M) mutant was unable to increase the GTP-Rab32 level (Fig. 6B, lane  4). These results, taken together, indicated that the HPS4(S541A/ N543A) mutant can normally activate Rab32 (and presumably Rab38) in melan-le cells, the same as WT HPS4, and thus it is possible that the activation of Rab32/38 by HPS4 occurs independently of Rab9. To further investigate this possibility, we evaluated the effect of Rab9A knockdown on the active Rab32 level in black mouse-derived melan-a cells (33). Rab9B knockdown was not performed because we showed previously that Rab9B is not expressed in melan-a cells (18). The results showed that neither Rab32 punctate localization nor the GTP-Rab32 level were affected in Rab9A-knockdown cells compared with control cells (Fig. 7, A and B). However, because Rab32 signals were mildly decreased in Rab9A knockdown cells, Rab9A may partly be involved in endosomal localization of Rab32 independent of HPS4 (Fig. 7A).
Because Rab32/38, targets of BLOC-3, are known to regulate the trafficking of melanogenic enzymes to melanosomes (15, 18,23,32,34,35), we investigated the impact of the two HPS4 mutants on tyrosinase trafficking and melanin synthesis in melan-le cells. In control melan-le cells, tyrosinase signals were restricted to the perinuclear region (Fig. 8A, top left panel,  arrow), and hardly any black melanosomes were observed (Fig.  8A, top right panel). Re-expression of HPS4(WT) clearly restored the peripheral distribution of tyrosinase (Fig. 8A, second left panel), and black-pigmented melanosomes were restored (Fig. 8A, second right panel). The increased tyrosinase protein level and melanin content were also confirmed by immunoblotting (Fig. 8, B and C) and by measuring optical density at 490 nm (Fig. 8, D and E), respectively. Re-expression of the Rab9 binding activity-deficient HPS4(S541A/N543A) mutant restored the peripheral distribution of tyrosinase and melanin content to the same level as in HPS4(WT)-expressing

Structure-function analysis of HPS4 in melanogenesis
cells, whereas re-expression of the GEF activity-deficient HPS4(G55W/G59M) did not (Fig. 8, A-E). We therefore concluded that the Rab32/38-GEF activity of HPS4 is necessary and sufficient for trafficking of tyrosinase to melanosomes and mel-anin synthesis in melanocytes and that the function of HPS4 is independent of Rab9, even though Rab9 itself is required for tyrosinase trafficking and melanin synthesis, as reported previously (18,19). , and rat (Rn) HPS4 and yeast Ccz1. Identical residues and conserved residues in HPS4 are shown against a black background and gray background, respectively. Two Gly residues, Gly-47 and Gly-51, in yeast Ccz1 have been shown to be essential for Ypt7-GEF activity (27), and the corresponding Gly residues were also conserved in mammalian HPS4 and were the focus of the site-directed mutagenesis to create Rab32/38-GEF activity-deficient HPS4. B, Rab32/38(CA/CN) binding activity of BLOC-3 (HPS1/4) WT and mutants as revealed by yeast tri-hybrid assays. Yeast cells containing pHPS1 plasmids, pAct2 plasmids expressing HPS4(WT or mutants), and pGBD-C1-Rab7/32/38(CA/CN)⌬Cys plasmids were streaked on SC-LWU (lacking leucine, tryptophan, and uracil) or SC-AHLWU (lacking adenine, histidine, leucine, tryptophan, and uracil) medium (selection medium) and incubated at 30°C. C, the normal Rab9 binding activity of HPS4(G55W/G59M) (GW/GM). Two-hybrid assays were performed as described for Fig. 3B.

Figure 6. Effect of HPS4 mutants on localization and activation of Rab32 in melan-le cells.
A, stable expression of the Rab9 binding-deficient HPS4 mutant (S541A/N543A (SA/NA)), not of the Rab32/38-GEF activity-deficient mutant (G55W/G59M (GW/GM)), restored clear punctate localization of Rab32 in melan-le cells, the same as HPS4(WT) did. HA-tagged HPS4(WT, S541A/N543A, or G55W/G59M) was stably expressed in melan-le cells by retrovirus infection followed by puromycin selection. Immunostained Rab32 was analyzed with a confocal fluorescence microscope. Cells that did not contain clear Rab32 puncta are outlined with a broken line. Expression of HPS4(WT or SA/NA) in melan-le cells significantly increased Rab32 signals compared with control melan-le cells (melan-le, 1.00 Ϯ 0.0643 (means Ϯ S.E., relative intensity); melan-le ϩ HPS4(WT), 2.33 Ϯ 0.110; melan-le ϩ HPS4(SA/NA), 2.01 Ϯ 0.123; p Ͻ 0.01, Tukey's test), whereas expression of HPS4(GW/GM) in melan-le cells had no significant effect (1.06 Ϯ 0.0823). Scale bar ϭ 20 m. B, the level of GTP-Rab32 in melan-le cells stably expressing HPS4(WT and mutants). Active Rab32 pulldown assays were performed as described under "Experimental procedures" (18). In brief, GSH-Sepharose beads coupled with T7-GST-RBD32 were incubated with a lysate of melan-le cells expressing HPS4(WT or mutants). Proteins bound to the beads were analyzed by immunoblotting with anti-Rab32 antibody (top panels) and anti-HA tag antibody (bottom panels) or by Amido Black staining for T7-GST-RBD32. The amount of endogenous Rab32 in the reaction mixture was adjusted before incubation with the beads (i.e. lysates for HPS4(WT) and HPS4(SA/NA) were diluted twice with the binding buffer). The positions of the molecular mass markers (in kilodaltons) are shown on the left. Note that HPS4(SA/NA) activated Rab32 in melan-e cells, the same as HPS4(WT) did, but HPS4(GW/GM) did not.

Discussion
Rab32/38 and Rab9 have been reported to be involved in melanogenesis (15,18,19), but whether these small GTPases regulate the trafficking of melanogenic enzymes synergistically or independently has remained unknown. Recent identification of BLOC-3 as a Rab32/38-GEF (16) and HPS4 as a Rab9 effector (17) prompted many researchers, including us, to hypothesize that HPS4 serves as a functional link between Rab32/38 and Rab9 during melanogenic enzyme trafficking; that is, that Rab9 -BLOC-3-Rab32/38 constitutes a cascade in which Rab9 recruits BLOC-3 to an organelle membrane, where it then activates Rab32/38 (11,16,18,20). In this study, we tested this hypothesis by evaluating the functional relationship between Rab32/38 and Rab9 by preparing Rab32/38-GEF activitydeficient and Rab9 binding activity-deficient HPS4 mutants (Figs. 3, 4, and 5) and then performing rescue experiments using HPS4-deficient melan-le cells and these HPS4 mutants. The results, however, showed that the activation, localization, and function of Rab32/38 in melanogenesis entirely depend on the Rab32/38-GEF activity of BLOC-3 and are mostly independent of Rab9A (Figs. 6, 7, and 8). Thus, Rab32/38 and Rab9 are likely to contribute independently to melanogenesis in melanocytes.
Although the BLOC-3-mediated Rab cascade model described above is a fascinating model, the results of our study clearly demonstrated that Rab9 is dispensable for activation and localization of Rab32/38, at least in melanocytes. How, then, is Rab9 involved in melanin synthesis? Because Rab9 is thought to regulate endosome-to-trans-Golgi network transport (36,37), and various molecules other than melanogenic enzymes, e.g. transporters and the fibril protein premelanosome protein (PMEL), need to be transported to immature melanosomes for maturation (5,6), Rab9 may be involved in the transport or recycling of these molecules during melanogenesis, and that may indirectly affect the levels of melanogenic enzyme expression. Alternatively, Rab9 may bind to the Rab32/ 38-GAP RUTBC1 and regulate Rab32/38 recycling or localization (18). Actually, because the Rab32 signals were mildly affected in Rab9A knockdown cells (Fig. 7), Rab9 binding activity of RUTBC1 may be involved in localization of Rab32/38. Further analysis of a Rab9 binding-deficient RUTBC1 mutant in RUTBC1 knockdown/knockout cells will be necessary to resolve this issue.
Although the BLOC-3-mediated Rab cascade is highly unlikely to function during melanogenesis in melanocytes, the cascade may still play an important role in other types of membrane trafficking in different cell types. Actually, Rab32/38 have been reported to regulate the biogenesis of other lysosomerelated organelles and to prevent pathogenic microbial infections (38). We have especially noted the latter function of Rab32/38 because they localize to bacterial pathogencontaining endosomes, and BLOC-3 has been shown to be required for this process (39 -42). It will be interesting to investigate whether the BLOC-3-mediated Rab cascade is involved in defense against pathogenic microbial infections. A Rab9 binding-defective HPS4 mutant we developed in this study will be a useful tool for evaluating the importance of the BLOC-3mediated Rab cascade during microbial infection.
It is now widely thought that Rab-GEFs are major determinants of Rab membrane localization (43). Our findings support this thinking because Rab32/38 did not show any clear membranous localizations in the absence of BLOC-3 GEF activity (Fig. 6). However, immunofluorescence analysis revealed hardly any colocalization of HA-tagged HPS4 with Rab32/38 in rescued melan-le cells; instead, it was mostly localized in the cytoplasm (data not shown). We therefore speculate that another binding partner of BLOC-3 is required for its transient localization at immature melanosomes or endosomes. Alternatively, Rab32/38 themselves may contain targeting sequences, e.g. a hypervariable C-terminal domain, that target them to specific organelles (44). Further extensive research will be necessary to determine the precise molecular mechanism by which BLOC-3 defines the site of Rab32/38 activation.
In summary, we have identified the critical amino acids responsible for the Rab32/38-GEF activity and Rab9 binding activity of HPS4 and have demonstrated, by means of loss-offunction HPS4 mutants, that BLOC-3-GEF activity toward Rab32/38 is essential for melanogenesis in melanocytes, and that Rab9 binding activity is not. In the future, the HPS4 mutants developed in this study will be useful for analyzing other BLOC-3-mediated membrane trafficking events, such as defense against pathogenic microbial infection.

Figure 7. Effect of Rab9A knockdown on localization and activation of endogenous Rab32 in melan-a cells. A, localization of endogenous Rab32 in
Rab9A knockdown melan-a cells. Melan-a cells were transfected with siRNAs against Rab9A and then immunostained with anti-Rab32 antibody. Although no apparent change in punctate localization of Rab32 was observed in Rab9A-knockdown cells, their Rab32 signals were mildly but significantly reduced compared with control melan-a cells (siControl, 1.00 Ϯ 0.0819 (means Ϯ S.E., relative intensity); siRab9A#1, 0.556 Ϯ 0.0495; and siRab9A#2, 0.679 Ϯ 0.0559; **, p Ͻ 0.01, Dunnet's test). Scale bars ϭ 20 m. B, the same level of active Rab32 between control cells and Rab9 knockdown melan-a cells. Active Rab32 pulldown assays were performed as described under "Experimental procedures" (18), and the description of the immunoblot data is the same as for Fig. 6B. The positions of the molecular mass markers (in kilodaltons) are shown on the left.

Plasmid construction and site-directed mutagenesis
cDNA encoding the ORF of mouse HPS1 was amplified from Marathon-Ready adult mouse brain and testis cDNAs (Clontech/Takara Bio, Shiga, Japan) by PCR using the following specific primers containing a BamHI linker (underlined) or a stop codon (bold): HPS1 Met primer, 5Ј-GGATCCATGAAGTGC-GTGTTGGTGGC-3Ј; HPS1 stop primer, 5Ј-CTAGGGCAG-GGTCACCCGG-3Ј. The truncated mutants of HPS4 shown in Fig. 2 were produced by PCR using specific primers and fulllength HPS4 cDNA (18) as a template. Mutant HPS4 expression

Cell cultures, transfections, and stable expression of HPS4
The black mouse-derived immortal melanocyte cell line melan-a and the light ear mouse-derived immortal melanocyte cell line melan-le were obtained from the Wellcome Trust Functional Genomics Cell Bank and cultured as described previously (31,33,52). COS7 cells and Plat-E cells (a kind gift from Dr. Toshio Kitamura) were maintained at 37°C under 5% CO 2 in DMEM (Fujifilm Wako Pure Chemical, Osaka, Japan) containing 10% fetal bovine serum and antibiotics. Retrovirus production and infection were performed essentially as described previously (53). Stable melan-le cell lines were obtained by puromycin selection (1.5 g/ml for 2-3 days). Cells were transfected with plasmid DNAs and siRNAs using Lipofectamine 2000 and RNAiMAX (Thermo Fisher Scientific), respectively, according to the manufacturer's instructions.

Co-immunoprecipitation assays in COS-7 cells and active-Rab32 pulldown assays in melanocytes
COS-7 cells were co-transfected for 24 h with pEF-T7-HPS4 (WT or mutants) and pEF-FLAG-HPS1 or pEF-FLAG-Rab9A using Lipofectamine 2000. The transfected cells were lysed with lysis buffer (50 mM HEPES-KOH (pH 7.2), 150 mM NaCl, 1 mM MgCl 2 , and 1% Triton X-100 supplemented with complete EDTA-free protease inhibitor mixture (Roche)), and the cell lysates were incubated for 1 h at 4°C with anti-FLAG tag antibody-conjugated agarose beads (Sigma-Aldrich). The beads were washed three times with washing buffer (50 mM HEPES-KOH (pH 7.2), 150 mM NaCl, 1 mM MgCl 2 , and 0.1% Triton X-100), and proteins bound to the beads were analyzed by 10% SDS-PAGE followed by immunoblotting with the appropriate antibodies, as indicated in each figure. Immunoreactive bands were visualized by enhanced chemiluminescence. COS-7 cell extracts that had been transfected with pEF-T7-GST-RBD32 (18) were lysed with lysis buffer and allowed to react with GSH-Sepharose beads for 1 h at 4°C. The beads were washed three times with washing buffer and then incubated for 1 h at 4°C with lysates from melan-le cells stably expressing HPS4 (WT or mutants) or not expressing HPS4 at all, Rab9Aknockdown melan-a cells, or control melan-a cells. The proteins bound to the beads were analyzed by 10% SDS-PAGE followed by immunoblotting with the antibodies indicated in each figure.

Immunocytochemistry
Immunostaining was performed essentially as described previously (18,52). In brief, cultured cells were fixed with 4% paraformaldehyde or 10% TCA and stained with specific antibodies against tyrosinase and Rab32. The stained cells were examined for fluorescence with a confocal fluorescence microscope (Fluoview 1000-D; Olympus, Tokyo, Japan) through an objective lens (ϫ60 magnification, numerical aperture 1.40, Olympus) and with Fluoview software (version 4.1a, Olympus). The fluorescence signals of Rab32 were captured at random (40 and 63 cells each in Figs. 6A and 7A, respectively) with the confocal microscope and quantified with ImageJ software (version 1.52i, National Institutes of Health).

Melanin assays
Melanin assays were performed essentially as described previously (15, 23). In brief, melanocytes were solubilized in lysis buffer, the pigment was pelleted by centrifugation at 17,400 ϫ g for 10 min at 4°C, and the pellet was dissolved in 1 M NaOH/ 10% DMSO for 30 min at 100°C. Melanin content was measured as optical density at 490 nm with a microplate reader (model 680XR, Bio-Rad) and normalized to total protein content.

Sequence analysis
The amino acid sequences of various species of HPS4 were aligned using ClustalW multiple alignment programs (54) pro-

Statistical analysis
Statistical tests were performed using Tukey's test or Dunnett's test, and p Ͻ 0.05 was considered statistically significant.