Hermansky-Pudlak Syndrome Protein Complexes Associate with Phosphatidylinositol 4-Kinase Type II α in Neuronal and Non-neuronal Cells*

The Hermansky-Pudlak syndrome is a disorder affecting endosome sorting. Disease is triggered by defects in any of 15 mouse gene products, which are part of five distinct cytosolic molecular complexes: AP-3, homotypic fusion and vacuole protein sorting, and BLOC-1, -2, and -3. To identify molecular associations of these complexes, we used in vivo cross-linking followed by purification of cross-linked AP-3 complexes and mass spectrometric identification of associated proteins. AP-3 was co-isolated with BLOC-1, BLOC-2, and homotypic fusion and vacuole protein sorting complex subunits; clathrin; and phosphatidylinositol-4-kinase type II α (PI4KIIα). We previously reported that this membrane-anchored enzyme is a regulator of AP-3 recruitment to membranes and a cargo of AP-3 ( Craige, B., Salazar, G., and Faundez, V. (2008) Mol. Biol. Cell 19, 1415-1426 ). Using cells deficient in different Hermansky-Pudlak syndrome complexes, we identified that BLOC-1, but not BLOC-2 or BLOC-3, deficiencies affect PI4KIIα inclusion into AP-3 complexes. BLOC-1, PI4KIIα, and AP-3 belong to a tripartite complex, and down-regulation of either PI4KIIα, BLOC-1, or AP-3 complexes led to similar LAMP1 phenotypes. Our analysis indicates that BLOC-1 complex modulates the association of PI4KIIα with AP-3. These results suggest that AP-3 and BLOC-1 act, either in concert or sequentially, to specify sorting of PI4KIIα along the endocytic route.

Membranous organelles along the exocytic and endocytic pathways are each defined by unique lipid and protein composition. Vesicle carriers communicate and maintain the composition of these organelles (2). Consequently defining the machineries that specify vesicle formation, composition, and delivery are central to understanding membrane protein traffic. Generally vesicle biogenesis uses multiprotein cytosolic machineries to select membrane components for inclusion in nascent vesicles (2,3). Heterotetrameric adaptor complexes (AP-1 to AP-4) are critical to generate vesicles of specific composition from the different organelles constituting the exocytic and endocytic routes (2)(3)(4).
The best understood vesicle formation machinery in mammalian cells is the one organized around the adaptor complex AP-2 (5). This complex generates vesicles from the plasma membrane using clathrin. Our present detailed understanding of AP-2 vesicle biogenesis mechanisms and interactions emerged from a combination of organellar and in vitro binding proteomics analyses together with the study of binary interactions in cell-free systems (5)(6)(7)(8)(9). In contrast, the vesicle biogenesis pathways controlled by AP-3 are far less understood. AP-3 functions to produce vesicles that traffic selected membrane proteins from endosomes to lysosomes, lysosome-related organelles, or synaptic vesicles (10 -13). AP-3 is one of the protein complexes affected in the Hermansky-Pudlak syndrome (HPS; 3 Online Mendelian Inheritance in Man (OMIM) 203300). So far, mutations in any of 15 mouse or eight human genes trigger a common syndrome. This syndrome encompasses defects that include pigment dilution, platelet dysfunction, pulmonary fibrosis, and occasionally neurological phenotypes (14,15). All forms of HPS show defective vesicular biogenesis or trafficking that affects lysosomes, lysosome-related organelles (for example melanosomes and platelet dense granules), and, in some of them, synaptic vesicles (11)(12)(13). Most of the 15 HPS loci encode polypeptides that assemble into five distinct molecular complexes: the adaptor complex AP-3, HOPS, and the BLOC complexes 1, 2, and 3 (14). Recently binary interactions between AP-3 and BLOC-1 or BLOC-1 and BLOC-2 suggested that arrangements of these complexes could regulate membrane protein targeting (16). Despite the abundance of genetic deficiencies leading to HPS and genetic evidence that HPS complexes may act on the same pathway in defined cell types (17), we have only a partial picture of protein interactions organizing these complexes and how they might control membrane protein targeting.
In this study, we took advantage of cell-permeant and reversible cross-linking of HPS complexes followed by their immunoaffinity purification to identify novel molecular interactions. Cross-linked AP-3 co-purified with BLOC-1, BLOC-2, HOPS, clathrin, and the membrane protein PI4KII␣. We previously identified PI4KII␣ as a cargo and regulator of AP-3 recruitment to endosomes (1,18). Using mutant cells deficient in either individual HPS complexes or a combination of them, we found that BLOC-1 facilitates the interaction of AP-3 and PI4KII␣. Our studies demonstrate that subunits of four of the five HPS complexes co-isolate with AP-3. Moreover BLOC-1, PI4KII␣, and AP-3 form a tripartite complex as demonstrated by sequential co-immunoprecipitations as well as by similar LAMP1 distribution phenotypes induced by down-regulation of components of this tripartite complex. Our findings indicate that BLOC-1 complex modulates the recognition of PI4KII␣ by AP-3. These data suggest that AP-3, either in concert or sequentially with BLOC-1, participates in the sorting of common membrane proteins along the endocytic route.

EXPERIMENTAL PROCEDURES
Antibodies-Antibodies used in these studies are listed in Table 1 except for the vps33b antibodies. An antiserum was prepared against rat vps33b peptide DTLTAVENKVSKLVTD-KAAGKITDAFSSL (amino acids 450 -478 of the NCBI record NP_071622.1). This peptide is conserved in human and mouse vps33b. Peptide synthesis and rabbit immunization were performed by Alpha Diagnostic International (San Antonio, TX).
Cross-linking and Immunoprecipitation-Detailed procedures have been described previously (1). Briefly cells were washed twice in cold PBS plus CaCl 2 and MgCl 2 (PBS-CM) and incubated with 1 mM dithiobis(succinimidyl propionate) (DSP) (Pierce) or DMSO alone in PBS-CM for 2 h on ice. The reaction was stopped by adding 25 mM Tris, pH 7.4, for 15 min on ice and washing twice in PBS. Cells were lysed for 30 min on ice in buffer A (150 mM NaCl, 10 mM HEPES, 1 mM EGTA, and 0.1 mM MgCl 2 , pH 7.4), 0.5% Triton X-100, and Complete TM antiprotease mixture (Roche Applied Science). Homogenates were clarified by sedimentation at 16,100 ϫ g for 10 min. 500 -800 g of supernatant were immunoprecipitated using Dynal magnetic beads M450 (Invitrogen) covered with monoclonal antibodies against AP-3 ␦ subunit, ␥-adaptin (AP-1), or transferrin receptor for 3 h at 4°C. Beads were washed six times in buffer A plus 0.1% Triton X-100 for 5 min each and incubated with SDS-PAGE sample buffer for 5 min at 75°C. Samples were loaded on 4 -20% PAGE-SDS Criterion precast gels (Bio-Rad) and analyzed by immunoblot.
Sucrose Sedimentation and Mass Spectrometry-Clarified Triton-soluble supernatants from PC12 cells either treated with DSP or DMSO alone were sedimented in a 5-20% sucrose gradient prepared in buffer A plus 0.5% Triton X-100 during 13 h at 187,000 ϫ g in an SW55 rotor (1.5 mg per gradient) (20). 20 samples were collected from the bottom (250 l each) and  A total of ϳ58 mg of clarified Triton-soluble supernatant from DSP-treated PC12 cells was fractionated by sucrose gradient sedimentation (ϳ20 gradients per preparation). AP-3 migration in the gradient was determined by immunoblot. Two preparations were performed. In one preparation sucrose fractions 1-8 (from the bottom) were pooled. In a second preparation, fractions 3-5 were pooled and further analyzed. Pooled fractions containing cross-linked AP-3 complexes were diluted in buffer A, and samples were subjected to immunoaffinity chromatography with AP-3 ␦ antibodies bound to Dynal M450 magnetic beads. Binding was performed for 3 h at 4°C. After six washes in buffer A plus Triton X-100, AP-3 cross-linked complexes were eluted with a 50 M concentration of a peptide corresponding to the epitope recognized by the anti-␦ SA4 monoclonal antibody (AQQVDIVTEEMPENALPSDEDDKD-PNDPYRA) corresponding to the amino acids 680 -710 of human ␦-adaptin (NCBI:AAD03777; gi:1923266). Elution was performed for 2 h at 0°C. Peptide eluted material was precipitated on ice with 10% trichloroacetic acid for 30 min, and protein pellets were washed twice with 1:1 ethanol/ether (Ϫ20°C), dried, and resuspended in 0.1 N NaOH. Solubilized protein pellets were incubated in Laemmli sample buffer (Bio-Rad) at 75°C for 5 min, and proteins were resolved in a single lane of a 4 -20% PAGE-SDS Criterion gel (Bio-Rad). Proteins were stained with SYPRO Ruby (Bio-Rad), and the lane was divided into 18 fractions. Each fraction was subjected to in-gel digestion with sequencing grade trypsin (Promega, Madison, WI) at 37°C overnight, and the peptides were extracted as described previously (21). Protein identification was performed with nanohigh performance liquid chromatography (HPLC)-tandem mass spectrometry (MS/MS) analysis (for details, see Ref. 22); a QSTAR XL (Applied Biosystems, Foster City, CA) hybrid quadrupole tandem mass spectrometer interfaced with an Ultimate nano-HPLC system (LC Packings, Sunnyvale, CA) was used. The mass spectrometer was operated in positive ion mode using information-dependent acquisition to acquire a single MS scan (m/z 400 -1900 scan range) followed by up to two MS/MS scans (m/z 50 -1900 scan range). A rolling collision energy was used for the MS/MS scans. The data were analyzed using the ProID (Applied Biosystems) and Mascot (Matrix Science) search algorithms. We defined the following criteria to consider probable positive protein identification. Proteins were included for analysis 1) if represented by at least one peptide with a Mascot score Ͼ45 unless the proteins was already known to interact with AP-3 and/or 2) if represented in at least two of three spectrometry analyses. Peptides identified in a single MS run were also included if the corresponding proteins were part of a complex in which other subunits were identified in a distinct MS run.
Membrane Preparation-HEK293T cells, either treated or not with DSP, were washed in PBS and homogenized by 18 passages in a cell cracker in buffer A containing Complete antiprotease mixture. Homogenate was sedimented at 27,000 ϫ g for 40 min to generate a P1 membrane fraction and an S1 super-natant. Cytosol free of membranes was obtained by centrifugation of S1 at 210,000 ϫ g for 30 min in a Beckman Coulter TLA120.2 rotor. Membrane fractions or cytosol was solubilized in buffer A plus 0.5% Triton X-100 for 30 min on ice. Triton extracts were clarified by centrifugation at 16,100 ϫ g for 10 min, and supernatants were immunoprecipitated followed by Western blot analysis.
Clathrin-coated Vesicle (CCV) Isolation-Clathrin-coated vesicles were prepared according to Girard et al. (23). Briefly PC12 cells were homogenized in buffer B (100 mM MES, pH 6.8, 0.5 mM EGTA, and 1 mM MgCl 2 ) plus Complete antiprotease mixture. Homogenates were centrifuged at 17,000 ϫ g for 20 min in a Sorvall SS-34 fixed angle rotor to obtain a supernatant (S1) and a pellet (P1). S1 was centrifuged at 56,000 ϫ g for 1 h in a Beckman Type 40 ultracentrifuge fixed angle rotor to generate P2 and S2 fractions. P2 pellets were resuspended in buffer B plus antiprotease mixture and laid over a layer of 8% (w/v) sucrose in buffer B prepared in 50% (v/v) deuterium oxide. After 2 h of centrifugation at 116,000 ϫ g in a Beckman SW55 rotor, the pellet of CCVs was resuspended in buffer B and kept at Ϫ80°C.
HEK293T cells were transfected twice on alternate days with siRNA-CONTROL 1 or 2 oligos or with siRNA oligos to silence the human PI4KII␣, the BLOC1 subunit pallidin, or the AP-3 ␦ subunit. Twenty-four hours after the second transfection cells were reseeded in 12-well plates for Western blots or on coverslips for immunofluorescence analysis. Samples were processed the next day.

Identification of Proteins Co-purifying with Cross-linked AP-3
Complexes-We used chemical cross-linking to stabilize labile protein interactions followed by purification of cross-linked protein complexes containing HPS gene products. We chose DSP, a homobifunctional cell-permeable cross-linker with a 12-Å spacer arm. This reagent contains a disulfide bond that allows cleavage of cross-linked products (25)(26)(27). Previously we demonstrated that the interaction of AP-3 with its cargo/regulator PI4KII␣ on membranes could be structurally and functionally explored in vivo using DSP whole-cell cross-linking (1). Here we focused on the adaptor complex AP-3, which is affected in HPS2 syndrome (28), because it shows the most severe systemic and neurological phenotype of all HPS affected loci (29).
We first sought non-saturating in vivo DSP incubation conditions that could stabilize a small fraction of known AP-3interacting proteins into high molecular weight complexes, "in vivo controlled cross-linking." Triton-soluble extracts of control and DSP-treated PC12 cells were fractionated by sucrose sedimentation. We determined the sedimentation of AP-3 and its membrane protein cargoes, ZnT3 and PI4KII␣ (1,18,20). After DSP treatment, a fraction of AP-3 exhibited a slightly increased sedimentation coefficient above 9.4 S (Fig. 1A). Similarly a discrete pool of ZnT3 and PI4KII␣ was incorporated into complexes with a sedimentation coefficient between 9.4 and 16.5 S (Fig. 1A, lanes 1-7). The sedimentation of a non-AP-3 cargo synaptic vesicle protein, synaptophysin (20), remained unaffected by DSP treatment. Moreover the total protein sedimentation pattern between control and DSPtreated samples remained identical (Fig. 1B). These results show that a limited set of protein-protein interactions are stabilized in vivo by DSP treatment. We independently confirmed these findings by immunoprecipitation of AP-3 complexes from [S 35 ]methionine-labeled whole-cell Triton extracts from control and DSP-treated PC12 cells. Because the majority of AP-3 remains uncross-linked in whole cell extracts of DSPtreated cells, we predicted that major radiolabeled polypeptides precipitated with AP-3 antibodies should be similar between DSP-treated and untreated cells. In fact, radiolabeled bands immunoprecipitated with AP-3 ␦ antibodies were identical irrespective of whether cells were treated or not treated with DSP (Fig. 1C, compare lanes 5 and 6). Moreover these bands differed from those immunoprecipitated with transferrin receptor (TrfR) antibodies (Fig. 1C, lanes 3 and 4). These experiments show that AP-3 complexes are selectively immunoisolated from cross-linked cells without background due to excessive cross-linking or nonspecific binding.
pare lanes 2 and 6). Importantly PI4KII␣ interaction with AP-1 was negligible in wild type cells, yet it increased beyond background levels in mocha cells. These results are consistent with our findings that AP-1 and PI4KII␣ colocalize only in AP-3-null cells (18).
To assess whether proteins identified by mass spectrometry associate into a complex, we asked whether clathrin, BLOC-1 (detected by its subunit pallidin), and PI4KII␣ sedimented together with AP-3 by sucrose sedimentation analysis (Fig. 3A). Triton-soluble extracts from control and DSP-treated PC12

Proteins co-purifying with cross-linked AP-3 complexes
The table depicts proteins identified by AP-3 immunoaffinity chromatography. ⌺ peptides and ⌺ Mascot scores corresponds to the sum of all peptides identified and their individual Mascot scores in three MS/MS analyses from two independent purifications. Clathrin binding motifs were identified using gi-defined primary sequences fed into the Eukaryotic Linear Motif (ELM) search engine. Relevant previous proteomes correspond to those of 1) AP-3 microvesicles (18), 2) brain clathrin-coated vesicles (7), 3) HeLa cell clathrin-coated vesicles (8)  cells were fractionated by sucrose sedimentation. Sucrose fractions were immunoisolated with AP-3 ␦ antibodies followed by immunoblot analysis of immunocomplexes (Fig. 3A). Clathrin, BLOC-1, and PI4KII␣ co-precipitated with AP-3 as a heterogeneous complex whose sedimentation ranged from 9.4 to 16.5 S. Such a complex was not observed in uncross-linked PC12 Triton-soluble extracts (Fig. 3A, lanes 1-8). AP-3, BLOC-1, and PI4KII␣ could form a tripartite complex. We tested this hypothesis performing sequential immunoprecipitations with antibodies against AP-3 ␦ and an HA epitope engineered into either PI4KII␣ or PI4KII␣L61A/L62A. PI4KII␣L61A/L62A is a mutant that lacks a dileucine sorting motif necessary to bind AP-3, thus providing a stringent control for specificity (Fig. 3B) (1). HEK293 cells transfected with HA-tagged forms of wild type and PI4KII␣L61A/L62A were treated with DSP, and crosslinked complexes were precipitated with AP-3 ␦ antibodies. Cross-linked complexes bound to AP-3 beads were efficiently eluted with the ␦ epitope peptide (Fig. 3B, compare lanes 3 and  4 with lanes 5 and 6). Eluates underwent a second round of immunoprecipitation with anti-HA antibodies (Fig. 3B, lanes 7  and 8). PI4KII␣-HA co-precipitated AP-3 and BLOC-1 subunits detected by immunoblot, yet this tripartite complex failed to be isolated from cells expressing PI4KII␣L61A/L62A-HA.
These data indicate that BLOC-1, PI4KII␣, and AP-3 associate into a high molecular weight complex. Collectively our results demonstrate that in vivo controlled cross-linking is a viable approach to identify proteins associated into labile macromolecular complexes. In the case of AP-3, the macromolecular complex possesses as a minimum BLOC-1 and PI4KII␣. AP-3, BLOC-1, and PI4KII␣ Concentrate in Clathrin-coated Vesicles-Adaptors and clathrin cycle between cytosolic and membrane-bound states. Once on membranes, adaptors and clathrin are concentrated along with membrane proteins into CCVs (3,4). We analyzed the subcellular distribution of AP-3, PI4KII␣, BLOC-1, and clathrin by fractionation of DSP-crosslinked PC12 cells into soluble and total membrane fractions (Fig. 4A, lanes 1-6 and 7-12, respectively). Cytosol and membranes were solubilized with detergent, and cross-linked AP-3 complexes were isolated with AP-3 ␦-adaptin antibodies (Fig.  4A, lanes 5 and 6 and lanes 11 and 12). Cytosolic fractions were free of membranes as evidenced by the negligible content of PI4KII␣ (Fig. 4A, compare lanes 13 and 14 with lanes 15 and  16). AP-3 complexes isolated from cross-linked cytosolic fractions contained AP-3 and the BLOC-1 subunit pallidin but were devoid of other cytosolic proteins found in AP-3 cross-linked complexes obtained from whole-cell extracts such as clathrin (Fig. 4A, lane 6). In contrast, AP-3 complexes isolated from cross-linked membrane fractions were enriched in AP-3 and its membrane protein cargo/regulator PI4KII␣, clathrin, and the BLOC-1 subunit pallidin (Fig. 4A, compare lanes 6 and 12). Binding of these proteins to AP-3 was not due to cytosolic contaminants in the membrane fraction as indicated by the absence of tubulin (Fig. 4A, lane 12). Furthermore background binding of all these proteins was negligible as determined by precipitations using beads lacking antibodies or decorated with transferrin receptor antibodies (Fig. 4A, lanes 1-4 and 7-10). We confirmed the association of AP-3, clathrin, and PI4KII␣ in membranes immunoisolating cross-linked clathrin complexes from DSP-treated PC12 cells (Fig. 4B). Detergent-soluble extracts from vehicle-and DSP-treated cells were immunoisolated with the clathrin monoclonal antibody X22. AP-3 and PI4KII␣ co-isolated with clathrin yet only in the presence of DSP (Fig. 4B, compare lanes 9 and 10). Similarly immunoisolation of cross-linked AP-3 complexes yielded clathrin and PI4KII␣ in DSP-treated extracts (Fig. 4B, compare lanes 7 and  8). In contrast; AP-3, clathrin, and PI4KII␣ levels were negligible in control immunoisolations performed with empty beads (Fig. 3B, lanes 5 and 6) or transferrin receptor antibody-containing beads (Fig. 4B, lanes 3 and 4). A, clarified detergent-soluble extracts from DSP-treated PC12 cells were resolved by sucrose sedimentation as described in Fig. 1D. Fractions 1-13 were immunoprecipitated (IP) with AP-3 ␦ antibodies, and immunocomplexes were analyzed by immunoblot (IB) with the indicted antibodies. Note that the bands in pallidin blots seen in the absence and presence of DSP correspond to IgG light chains. Both immunoprecipitating and immunoblotting antibodies were raised in mice. B, HEK293T cells transfected with PI4KII␣-HA wild type or the mutant PI4KII␣L61A/L62A-HA were incubated with DSP and lysed, and supernatants were immunoprecipitated with AP-3 ␦ antibodies. AP-3 immunocomplexes were eluted with 50 M ␦ antigenic peptide during 3 h at 4°C (Eluate; lanes 3 and 4). After peptide elution, beads were heated at 75°C for 10 min to release remaining AP-3 complexes (Beads; lanes 5 and 6). ␦ antigenic peptide eluate (lanes 3 and 4) or heat eluted material remaining in beads (lanes 5 and 6) were immunoprecipitated again with rabbit antibodies against HA epitope to isolate complexes containing HA-tagged PI4KII␣ (lanes 5-8). Immunocomplexes were resolved by SDS-PAGE and analyzed by Western blot. AP-3 isolates a ternary complex containing PI4KII␣-HA and the BLOC1 complex (lane 8). PI4KII␣L61A/L62A-HA mutant severely reduced BLOC-1 and AP-3 associated to the kinase (lane 7). ⌽ denotes that the procedure was not performed. Lanes 1 and 2 represent the initial input in the experiment (10%). The arrow indicates IgG chains.
AP-3, BLOC-1, and PI4KII␣ Co-precipitate with the HPS Complexes BLOC-2 and HOPS-The co-isolation of two HPS complexes, AP-3 and BLOC-1, with PI4KII␣ prompted us to examine whether other HPS complexes not detected by massspectrometry in PC12 cells could co-precipitate with crosslinked AP-3 complexes. Our rationale was founded in the detection of the HOPS subunit vps33b in PI4KII␣-containing microvesicles (18) and the binary interactions between BLOC-1 and BLOC-2 described by Di Pietro et al. (16). To test for the presence of BLOC-2 and HOPS complex subunits in crosslinked AP-3 complexes, we used mouse skin fibroblasts and human HEK293 cells. In contrast with PC12 cells, endogenous BLOC-2 and HOPS complex subunits were readily detectable in these cell types (Figs. 5 and 6). Cross-linked AP-3 complexes were immunoisolated from detergentsoluble extracts from wild type mouse fibroblasts incubated in the absence and presence of DSP (Fig.  5A, lanes 1-6). Similar to our findings in PC12 cells, AP-3 crosslinked complexes were co-isolated with clathrin, BLOC-1, and PI4KII␣ from DSP-treated wild type fibroblasts (Fig. 5A, lane 6). In addition, the BLOC-2 subunit Hps6 ruby was also detected in these AP-3 cross-linked complexes containing PI4KII␣ (Fig. 5A, lane 6). The specificity of a BLOC-2 interaction with AP-3 cross-linked complexes was confirmed by the absence of AP-3, PI4KII␣, BLOC-1, and the BLOC-2 subunit Hps6 ruby from transferrin receptor immunoisolations (Fig. 5A,  lane 4). We further tested the association of BLOC-2 subunits with AP-3 cross-linked complexes using mouse fibroblasts carrying genetic deficiencies in HPS genes. First AP-3 ␦ antibody immunoisolations performed with ␦-null mocha fibroblasts abrogated the detection of the BLOC-2 subunit Hsp6 ruby (Fig. 5A, mh/mh, lane 12). Finally deficiencies in the BLOC-1 subunit pallidin decreased the content of Hps6 ruby in AP-3 cross-linked complexes (Fig. 6C, Pldn pa/pa , compare lanes 1Ј and 2Ј).
We focused on HEK293 cells to test whether HOPS complex subunits could be co-isolated with AP-3 cross-linked complexes. These cells endogenously express high levels of the HOPS subunit vps33b (Fig. 5B, lanes 7 and 8). Much like PC12 cells and mouse skin fibroblasts, AP-3 complexes isolated from DSP-treated HEK293 cell extracts co-precipitated with PI4KII␣ and the BLOC-1 subunit pallidin. However, in addition to these proteins, the HOPS complex subunit vps33b was readily detectable in cross-linked AP-3 complexes (Fig. 5B, lane  6). The presence of vps33b in AP-3 complexes was selective because Vps33b was absent from control immunoisolations with transferrin receptor antibody-coated beads (Fig. 5B, lane  4) and from AP-3 ␦ antibody immunoisolations out-competed with a ␦ antigenic peptide (Fig. 5C, compare lanes 12 and 18). The interaction of HOPS complex subunits with AP-3 crosslinked complexes was not limited just to endogenous vps33b. In fact, exogenous expressed vps39-GFP or vps41-Myc also associated selectively with cross-linked AP-3 complexes isolated from HEK293 cells (Fig. 5C, lanes 14 and 16, respectively). Similar to vps33b, PI4KII␣, and BLOC-1 these proteins were eliminated by competing AP-3 ␦ antibody with a ␦ antigenic peptide. These results indicate that subunits of three HPS complexes, BLOC-1, BLOC-2, and HOPS, co-isolate with AP-3.  5 and 6), AP-3 ␦ antibodies (lanes 7 and 8), and clathrin X22 antibodies (lanes 9 and 10). Immunocomplexes were resolved by SDS-PAGE and analyzed by immunoblot. C, CCV-enriched fractions (lane 6) prepared as described by Girard et al. (23) were resolved by SDS-PAGE and analyzed by immunoblot. Compare CCV with total homogenate (H) and fractions P1, S1, P2, and S2 (see "Experimental Procedures"). All lanes contain 1 g of protein.
These results indicate that AP-3 forms a complex(es) with clathrin, PI4KII␣, BLOC-1, BLOC-2, and HOPS. A tripartite complex between AP-3, BLOC-1, and PI4KII␣ is at the core of these interactions where BLOC-1 selectively regulates the abundance of PI4KII␣ that associates with AP-3.
AP-3 and BLOC-1 Down-regulation Phenocopies PI4KII␣ Knockdown-Association between AP-3, BLOC-1, and PI4KII␣ suggests that deficiencies in components of this complex should alter traffic of membrane proteins to lysosomes in a similar fashion. We tested this prediction analyzing the appearance of LAMP1 enlarged endosomes after down-regulation of PI4KII␣, AP-3, or BLOC-1. LAMP1 is a lysosomal membrane protein that appears in enlarged endosomes induced by acute down-regulation of PI4KII␣ (1). Rescue of this phenotype requires the presence of an AP-3 dileucine sorting motif as well as the kinase activity in PI4KII␣ (1). siRNA-mediated downregulation of PI4KII␣ (Fig. 7A, lane 2) reduced the content of  1, 2, 7, and 8), with beads coated with TrfR (lanes 3, 4, 9, and 10), or with ␦ antibodies (lanes 5, 6, 11, and 12). Immunocomplexes were resolved by SDS-PAGE and analyzed by immunoblot with antibodies against AP-3 and subunits of BLOC-1 (pallidin) and BLOC-2 (Hsp6 ru). B, HEK293 Triton-soluble cell supernatants treated with (even lanes) and without DSP (odd lanes) were immunoprecipitated with empty beads (lanes 1 and 2) and beads coated with monoclonal antibodies against TrfR (lanes 3 and 4) or the ␦ subunit of AP-3 (lanes 5 and 6). Immunocomplexes were resolved by SDS-PAGE and analyzed by immunoblot with the antibodies specified to the left. C, Triton-soluble supernatants of untransfected HEK293 cells or cells transfected with Myc-tagged vps41 or GFP-tagged vps39 treated with (even lanes) and without DSP (odd lanes) were immunoprecipitated with beads decorated with ␦ AP-3 antibodies (lanes 7 and 18). Controls were performed by competition with 10 M ␦ peptide (lanes 7-12). Protein complexes were resolved by SDS-PAGE and analyzed by immunoblot (IB) with the indicated antibodies. All inputs correspond to 5%. this enzyme without affecting the levels of either components of the tripartite complex (AP-3 or BLOC-1) or proteins that co-precipitate with AP-3 and PI4KII␣, such as clathrin and the HOPS subunit vps33b. Similarly down-regulation of pallidin or AP-3 ␦ selectively decreased the levels of subunits found in each one of these individual protein complexes (Fig. 7A, lanes 4 and  5). Reduction in the content of multiple AP-3 or BLOC-1 subunits by siRNA knockdown of only one subunit recapitulates phenotypes observed in single gene deficiencies affecting AP-3 and BLOC-1 in mice (32,33). Acute siRNA-mediated downregulation of PI4KII␣ induced the appearance of LAMP1-positive enlarged endosomes in 25% of the cells analyzed. In contrast ϳ5% in control siRNA-treated cells displayed this phenotype (Fig. 7, B and C, gray bars). Similarly acute downregulation of AP-3 or BLOC-1 subunits induced LAMP1 enlarged organelles with the same penetrance as PI4KII␣ siRNA (Fig. 7C, gray bars). In addition to LAMP1 enlarged endosomes, either BLOC-1 pallidin knockdown (Fig. 7B), PI4KII␣, or AP-3 ␦ siRNA-treated cells (data not shown) possessed "doughnut"-shaped LAMP1-positive organelles clustered in the periphery of cells (Fig. 7B). Control cells exhibited this phenotype in less than 2% of cells in contrast with 30 -50% of either BLOC-1-, AP-3-, or PI4KII␣-down-regulated cells (Fig. 7C, black bars). These data provide functional evidence supporting a sorting role of a BLOC-1-PI4KII␣-AP-3 tripartite complex.

DISCUSSION
Here we used an in vivo chemical cross-linking approach coupled with immunoaffinity chromatography to identify and explore the role of molecules co-purifying with AP-3 ( Fig. 1). This strategy yielded 51 molecules identified by mass spectrometry in PC12 cells (Table 3 and Fig. 8). Prominently represented among these proteins were the membrane-anchored proteins PI4KII␣ and ZnT3 and cytosolic factors such as BLOC-1 subunits and clathrin chains. Moreover a targeted search for subunits from other HPS complexes identified BLOC-2 and HOPS in non-neuronal cells (Figs. 5 and 6). These results provide novel insight into the molecular organization of HPS protein complexes indicating that AP-3 forms complexes with clathrin and cargoes, such as PI4KII␣, and three other HPS complexes, BLOC-1, -2, and HOPS. The relevance of these associations is illustrated by the effect that genetic deficiencies in HPS complexes exert on the composition of AP-3 crosslinked complexes. For example, an AP-3 hypomorph allele (Ap3b1 pe/pe ) predictably decreased the association of PI4KII␣, BLOC-1, and clathrin to cross-linked AP-3 complexes (Fig. 6). However, a BLOC-1-null background (Plnd pa/pa ) significantly decreased the association of PI4KII␣ and BLOC-2 with AP-3 cross-linked complexes without affecting clathrin (Fig. 6). These data are consistent with a model where BLOC-1 modulates the association of PI4KII␣ to AP-3. Furthermore our present findings in BLOC-1-deficient cells provide a molecular mechanism to understand our previous observation that PI4KII␣ and AP-3 colocalization is reduced in BLOC-1-deficient Pldn pa/pa cells (34). In vivo chemical cross-linking uniquely allows the isolation of multiprotein complexes containing coats, adaptors, enzymatic regulators of vesicle biogenesis, and specific membrane protein cargoes recognized by adaptors. In the present case, PI4KII␣ and ZnT3, two well established AP-3 cargoes, associate with AP-3 but not transferrin receptor. Our data provide biochemical, genetic, and functional evidence for complexes containing a minimum of AP-3, BLOC-1, and PI4KII␣ (Figs. 3B, 6, and 7, respectively).
interact with AP-3 or lay in proximity to the adaptor. To minimize those molecules that could spuriously co-purify with cross-linked AP-3, we used diverse criteria to filter protein identification by mass spectrometry (22), immunoprecipitation specificity with AP-3 but neither transferrin receptor nor AP-1 antibodies, and the absence of AP-3-immunoprecipitated proteins in AP-3-null Ap3d mh/mh cells (Fig. 2). Using these criteria 50% of those components identified in our studies were also represented in other relevant analyses such as those performed in clathrin-coated vesicles by us and others (7,8), synaptic vesicles (31), AP-3 microvesicles (18), and AP-3 complexes assembled into liposomes (30). Our analysis does not represent an exhaustive identification of all components of an AP-3 interaction "network." In vivo controlled cross-linking possesses inherent limitations. First, the nature of the molecules retained in cross-linked complexes is strictly dependent on the crosslinker chemistry. Second, antibodies used to isolate crosslinked components select for complexes where the epitope is exposed. This is evident with antibodies against ␦-adaptin that recognize an AP-3 epitope in the proximity of where VAMP7-TI binds AP-3 (35). This could explain our inability to detect this v-(R)-SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) in cross-linked AP-3 complexes. Third, a component attached to detergent-insoluble membranes or the cytoskeleton (36) would be poorly represented. Finally we cannot assess the in vivo stoichiometry of components present in immunoprecipitated cross-linked complexes. This idea is based on the subsaturating DSP conditions used to stabilize supramolecular complexes yet still maintain low background of spurious interactions. This is an important consideration because either HPS complexes could function   (16,40). Annx II, annexin II; HC, heavy chain; LC, light chain.