MLN64 Mediates Mobilization of Lysosomal Cholesterol to Steroidogenic Mitochondria* 210

This study demonstrates that the steroidogenic acute regulatory protein-related lipid transfer (START) domain-containing protein, MLN64, participates in intracellular cholesterol trafficking. Analysis of the intracellular itinerary of MLN64 and MLN64 mutants tagged with green fluorescent protein showed that the N-terminal transmembrane domains mediate endocytosis of MLN64 from the plasma membrane to late endocytic compartments. MLN64 constitutively traffics via dynamic NPC1-containing late endosomal tubules in normal cells; this dynamic movement was inhibited in cholesterol-loaded cells, and MLN64 is trapped at the periphery of cholesterol-laden lysosomes. The MLN64 START domain stimulated free cholesterol transfer from donor to acceptor mitochondrial membranes and enhanced steroidogenesis by placental mitochondria. Expression of a truncated form of MLN64 (ΔSTART-MLN64), which contains N-terminal transmembrane domains but lacks the START domain, caused free cholesterol accumulation in lysosomes and inhibited late endocytic dynamics. The ΔSTART-MLN64 dominant negative protein was located at the surface of the cholesterol-laden lysosomes. This dominant negative mutant suppressed steroidogenesis in COS cells expressing the mitochondrial cholesterol side chain cleavage system. We conclude that MLN64 participates in mobilization and utilization of lysosomal cholesterol by virtue of the START domain's role in cholesterol transport.

thesis in the endoplasmic reticulum (1). Endocytosed cholesterol moves to the late endocytic compartment and subsequently modulates homeostatic responses via SREBP-SCAP located in the endoplasmic reticulum and Golgi (2). Mutations in NPC1 and NPC2 cause Niemann-Pick type C (NPC) disease, characterized by a phenotype of excessive cholesterol accumulation in lysosomes (3,4). Although NPC1 and NPC2 are required for egress of lysosomal cholesterol, the mechanism by which cholesterol is removed from the late endocytic pathway, a dynamic tubular network that communicates with lysosomes, is not known.
The mysteries of lysosomal cholesterol transport began to be unraveled when the gene responsible for the majority of cases of NPC disease, NPC1, was cloned (3). NPC1, a protein with a sterol-sensing domain, spans the surface membrane of the late endosomal compartment, and cellular uptake of low density lipoproteins promotes enrichment of NPC1 in late endosomes (5)(6)(7). The sterol-sensing domain was shown to be critical for cholesterol transfer, confirmed by the findings based on sitedirected mutagenesis and protein trafficking analysis of NPC1 protein (5)(6)(7). NPC1 was recently suggested to be a permease that may be involved in material transfer across the endosomal/lysosomal limiting membrane (8). The critical role of another protein, HE1, in clearing cholesterol out of lysosomes was established when it was discovered that the HE1 gene is mutated in the second and less frequent form of Niemann-Pick type C disease and is, therefore, the NPC2 gene. HE1 encodes a soluble cholesterol-binding protein (4,9).
Recently, MLN64, a protein that contains a C-terminal StAR-related lipid transfer (START) domain was found to reside in late endosomes and lysosomes (10 -12). The N terminus of MLN64 consists of a 6-amino acid leader sequence and four anchoring transmembrane domains (10). The START domain binds cholesterol and can stimulate sterol transfer between membranes (10,13,14). The topology of MLN64 in late endosomal membranes predicts that the START domain projects into the cytoplasm, where it is positioned to effect cholesterol transfer to opposing membranes (12). The location of a START domain protein in late endocytic pathway with NPC1 and HE1 suggests that MLN64 is a component of the machinery that facilitates cholesterol egress from lysosomes. To investigate this possibility, we elucidated the trafficking, in living cells, of wild type and mutant MLN64 proteins fused with green fluorescent protein (GFP). We found that a truncated MLN64 without the START domain (⌬START-MLN64) is a dominant negative mutant. The dominant negative action of this mutant caused extensive cholesterol accumulation in CHO cells and COS-7 cells accompanied by inhibition of late endosomal tubular trafficking, similar to the NPC phenotype that is caused by loss of function of NPC1 and NPC2 (3,4,7). This dominant negative mutant also inhibited steroidogenesis in cells. An in vitro assay provided evidence that the START domain mediates cholesterol transfer between membranes and stimulates mitochondrial metabolism of cholesterol into steroid hormones. These findings strongly suggest that MLN64 plays a role in the maintenance of intracellular cholesterol homeostasis.

Construction of Expression Plasmids for MLN64-GFP, MLN64-CFP,
⌬START-MLN64-GFP, START-MLN64-GFP, and NPC1-YFP-The human MLN64 cDNA has been previously described (10). MLN64 cDNA encoding either full-length or specific domains of the MLN64 protein were fused to the green fluorescent protein (pEGFP-N2; CLONTECH Laboratories Inc., Palo Alto, CA). All of the cDNAs used in making these fluorescent fusion proteins were generated by PCR using either Pfu or TaqDNA polymerases. The PCR primer pair used for the MLN64-GFP (forward, ATTAGAATTCATGAGCAAGCTGCCCAGGG-AG; reverse, TAATGTCGACTGATGCGCTGTCGCAGGTGAAAG) spanned the cDNA encoding for amino acids 1-436 of the human MLN64 protein. The MLN64-CFP was constructed by replacing the GFP of MLN64-GFP with CFP from pECFP-N2 (CLONTECH). The START domain for MLN64 (amino acids 216 -445) was also PCRamplified and fused to the GFP vector. A shared forward PCR primer (ATTAGAATTCCGCCACCATGGGGTCTGACAATGAATCAGATG) for generation of the GFP construct was used with the START-MLN64-G-FP-specific reverse primer (TAATGTCGACTGATGCGCTGTCGCAGG-TGAAAG). The ⌬START-MLN64-GFP fusion protein was generated by restriction digestion with XmnI and SmaI and gel isolation of the vector DNA plus cDNA encoding amino acids 1-225 followed by ligation of the blunt ends to form the ⌬START-MLN64-GFP construct. All constructs were sequenced and compared with the human MLN64 cDNA sequence (GenBank TM accession numbers NM 006438 and NM 006445, respectively). The human StAR-RFP fusion protein was constructed with the RFP tag pDsRed1 N1 vector (CLONTECH) on the C terminus of the full-length coding sequence of StAR. The human NPC1-YFP fusion protein was constructed by inserting NPC1 together with the six-histidine linker from vector pNES6 (7) into pEYFP-N1 (CLONTECH) in frame with the N terminus of YFP. The function of NPC1-YFP was confirmed by analyzing cholesterol clearance in CT60 cells using filipin staining (data not shown). The F2 expression plasmid, which encodes a fusion protein consisting of the P450 side chain cleavage enzyme and its electron donor (adrenodoxin and adrenodoxin reductase) partners (15), was generously provided by Dr. Walter L. Miller (University of California, San Francisco). An expression plasmid for murine type II 3␤-hydroxysteroid dehydrogenase was kindly provided by Dr. Anita Payne (Stanford University). Plasmids for transfection were prepared using the Qiagen Maxiprep system.
Cell Culture and Transfection-For assays of steroidogenic activity (16) of the GFP fusion constructs, COS-1 were cultured in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum and 50 g of gentamycin/ml in 5% CO 2 , 95% air at 37°C for cell propagation and plating. These cells were obtained from the American Type Culture Collection (Manassas, VA). COS-1 cells were plated at 30,000 cells/well in 12-well plates on day 0. On day 1, cells were transfected with FuGENE6 (Roche Molecular Biochemicals) plus plasmid DNA in the presence of serum as described in the manufacturer's protocol. Cells were co-transfected overnight with 250 ng of each of the 3␤-hydroxysteroid dehydrogenase and F2 expression plasmids and 500 ng of the GFP fusion vectors. On day 2, medium was changed to Dulbecco's minimal essential medium containing 10% fetal calf serum. Some cultures were incubated with (22R)-hydroxycholesterol, a soluble steroidogenic substrate, to assess maximal hormone production (10). After 24 h, medium was collected and frozen at Ϫ20°C until steroid analysis (see "Progesterone Assays"), and cells were harvested for protein determinations.
To examine utilization of LDL-derived cholesterol, human LDL was reconstituted with [ 3 H]cholesterol linoleate by the method of Krieger et al. (17). COS-1 cells transfected with the various MLN64 constructs and the steroidogenic enzymes as described above were cultured in medium supplemented with 10% lipoprotein-deficient serum and 50 g of protein/ml (70 g of cholesterol/ml)-reconstituted LDL (0.85 ϫ 10 5 dpm/g of cholesterol). The culture medium was collected 40 h later, and radiolabeled progesterone was isolated as described by Soto et al. (18).
Wild type CHO cells, CT60 CHO cells, and COS-7 cells were used for microscopic analysis. Cells were maintained in Ham's F-12 medium (CHO cell lines) or McCoy's medium (COS-7 cells and fibroblasts) containing 10% fetal bovine serum (HyClone, Logan, UT). Before transfection, cells were seeded in LabTek chambered coverslips and grown for 20 h to achieve ϳ60% confluence before transfection. All of the cell lines were transfected using FuGENE6 according to the manufacturer's instructions with some modification. For each chamber of the two-chamber coverslips, cells were incubated with mixtures containing 1 g of DNA and 4 l of FuGENE6 reagent in 400 l of Ham's F-12 medium for 5 h followed by culture in 10% fetal bovine serum-containing Ham's F-12 medium. Transfected cells were either imaged as live cells at different times of expression of the transfected cDNAs or as fixed cells (in 3% paraformaldehyde for 30 min) stained with 50 mg/ml filipin (Polysciences, Warrington, PA) to examine lysosomal cholesterol content (7). For live cell imaging, cells were grown on chambered coverslips (LabTek), and temperature was maintained using a combination of a heating stage and an Objective Heater System (Bioptechs, Butler, PA).
Progesterone Assays-Medium samples from transfected cells were assayed using progesterone Coat-A-Count tubes and reagents (Diagonistic Products Corp., Los Angeles, CA) as described by the manufacturer. Statistical tests were performed using the JMP 3.1.5 computer program (SAS Institute Inc., Cary, NC). One-way analysis of variance was used to analyze the effect of the expression plasmids on steroidogenic activity. Tukey-Kramer mean separation tests were performed for comparison between the means with p Ͻ 0.05 taken as the level of significance.
Immunoblotting-Western blotting was used to confirm expression of the GFP fusion proteins and the coupling of GFP to the MLN64 sequences. The rabbit MLN64 antiserum was generated against recombinant MLN64 START domain as a His 6 fusion protein (see below). An antiserum to GFP was obtained from Roche Molecular Biochemicals. Western blotting was carried out as described by Watari et al. (10).
Production of Recombinant MLN64 START Domain and Assay of Biological Activity-The human MLN64 START domain (amino acid residues 216 -445) was expressed in Escherichia coli as a His 6 tag protein with the tag on the START domain C terminus using a pET vector, and the protein was purified using previously described methods (13,14). The recombinant protein was used to generate rabbit anti-MLN64 antibodies. It was also tested for sterol transfer activity as described by Kallen et al. (14) in an assay that measures transfer of [ 14 C]cholesterol from unilamellar egg phosphatidylcholine to heattreated liver mitochondria acceptor membranes. Preliminary studies demonstrated that steroidogenic mitochondria from bovine corpus luteum or human placenta also served as equivalently effective acceptor membranes for START domain-stimulated sterol transfer. Transfer of [ 3 H]triolein incorporated into the liposomes was measured to correct for liposome fusion or trapping. The sterol transfer activity of recombinant MLN64 START domain was compared with that of sterol carrier protein 2.
Preparations enriched in syncytiotrophoblast mitochondria were prepared from human term placenta as described by Martinez et al. (19). The mitochondria were incubated in 120 mM KCl, 10 mM MOPS, 0.5 EGTA, 10 mM isocitrate, 0.1% bovine serum albumin, 0.4 g of aprotinin/0.1 ml, 1 M leupeptin, pH 7.4, with or without recombinant proteins (19). Reactions were terminated by the addition of methanol, and progesterone was assayed as described above. Progesterone present at zero time was subtracted from values obtained at various time points of incubation.
Pharmacological Perturbations-Wild type CHO cells were treated with progesterone (10 g/ml) plus LDL (100 g/ml) to induce cholesterol accumulation in lysosomes (7,20). To assess the effect of microtubule disruption on MLN64-GFP trafficking, cells were cultured in 10% fetal bovine serum-containing medium with 10 M nocodazole (Sigma) for up to 2 h at 37°C.
Labeling Endocytic Pathway in Living Cells-Late endocytic compartments (late endosomes and lysosomes) were labeled by incubating cells with fluorescent dextran (5, 7). Cells were cultured at 37°C, in medium containing 1 mg/ml rhodamine-dextran (dextran, tetramethylrhodamine, M r 70,000, lysine-fixable; Molecular Probes, Inc., Eugene, OR) for 1 h of uptake plus 1 h of chase (2 h of uptake/chase time), or 17 h of uptake plus 1 h of chase (18 h of uptake/chase time) before live cell imaging.
Confocal Microscopy and Time Lapse Imaging-Confocal images were collected using an LSM410 confocal microscope system equipped with a krypton-argon Omnichrome laser with excitation wavelengths of 488 and 568 nm for GFP and rhodamine (and Alexa Fluor 568), respectively (Carl Zeiss Inc., Thornwood, NY), and an argon ion laser (Coherent, Santa Clara, CA) with an excitation wavelength of 351 nm for filipin. Emissions were collected with a C-apochromat 63 ϫ 1.2 NA water immersion objective (Zeiss) using a 514 -540-nm band pass filter for GFP, a 590-nm long pass filter for rhodamine and DsRed RFP, and a 400 -435-nm band pass filter for filipin.
Time lapse images were captured using a cooled CCD camera (Cool-SNAP HQ Monochrome; Photometrics, Tucson, AZ), with a Sony Interline high quantum efficiency chip and custom macros provided by BioVision Technologies. The camera was attached to an Axiovert 35 inverted microscope (Zeiss) with a plan Apochromat 100 ϫ 1.4 NA oil immersion objective (Zeiss) or a C-apochromat 63 ϫ 1.2 NA water immersion objective (Zeiss). The excitation was from a 100-watt Hg-Arc lamp (Atto Instruments Inc., Rockville, MD) with standard fluorescein and rhodamine optics for GFP and RFP, respectively. For CFP and YFP, exciters 395-425 and 460 -480 nm, dichroic 505DCXR, and emitter 505-555 nm were used. For dual color time lapse imaging of CFP/ YFP or GFP/RFP in co-transfected cells, cells were imaged at 20°C to slow late endosomal tubular movement. Either CFP and YFP or GFP and RFP were excited and collected sequentially with a time interval of 0 s to minimize time lag during switching channels. IPLab software system (Scanalytics, Fairfax, VA) and custom scripts provided by Bio-Vision Technologies were used for image capture, processing, animation, and export to QuickTime movies. Adobe Photoshop 5.0 was also used for image processing (Adobe Systems Inc., San Jose, CA). Although MLN64 has been localized to late endosomes (12), the path that it takes to reach this destination has not been identified. To study the biosynthetic pathway of MLN64, we followed the expression of MLN64-GFP in living CT60 CHO cells that have mutations in the NPC1 gene and thus display a cholesterol trafficking defect (Fig. 2). 12-17 h after transfection, we found that MLN64-GFP was mainly associated with the plasma membrane ( Fig. 2A). Very little MLN64-GFP fluo-rescence was present in cholesterol-laden lysosomes that were labeled by rhodamine dextran uptaken by the cells for 17 h before transfection (data not shown) (7). 72 h after transfection, MLN64-GFP was mainly present as ring-shaped structures (Fig. 2B) recognized by immunostaining with anti-C-terminal MLN64 antibody (Fig. 2C). Filipin staining confirmed that MLN64-GFP was at the periphery of cholesterol-loaded lysosomes (data not shown). These observations confirm that MLN64-GFP travels first to the plasma membrane before being targeted to the late endocytic compartment. In wild type CHO cells, MLN64-GFP followed an identical biosynthetic pathway to the plasma membrane (data not shown) and then to the late endocytic pathway (see below), as in CT60 cells.

Characterization of the Expression of Wild Type and Mutant
The ⌬START-MLN64-GFP followed a biosynthetic trafficking pathway similar to that of wild-type MLN64, first to the plasma membrane (Fig. 2D) and then to the cholesterol-laden lysosomes (Fig. 2E), showing that ⌬START-MLN64-GFP contains the essential signals for plasma membrane localization and lysosomal targeting similar to newly synthesized MLN64. ⌬START-MLN64-GFP is not recognized by the anti-C-terminal MLN64 antibody (Fig. 2F). START-MLN64-GFP, the mutant lacking the N-terminal domains, was present only in a cytoplasmic and nuclear pattern 16 -72 h after transfection, further substantiating the role of the N-terminal domain in directing MLN64 to the plasma membrane and then to the late endocytic compartment (see below). ⌬START-MLN64-GFP (Fig. 2G) and MLN64 Constitutively Traffics within the Cholesterol-regulated, Dynamic Late Endosomal Tubulovesicular Network-We used WT CHO cells transiently transfected with MLN64-GFP or stably expressing MLN64-GFP. MLN64-GFP was found in late endosomal tubulovesicular compartments that constitutively traffic in a microtubule-dependent manner (supplemental movies 1 and 2). The dynamic late endosomal tubular movement is inhibited by nocodazole (10 M), which disrupts the microtubular network. Similar to previous findings on NPC1 trafficking (7), within 8 min after nocodazole treatment, the MLN64-GFP-containing endosomal compartments ceased to bud tubules and gathered in the perinuclear region, showing little movement (data not shown).
Analysis of living cells cotransfected with MLN64-CFP (Fig.  3, A and C) and NPC1-YFP (Fig. 3, B and D) confirmed that MLN64 colocalizes and traffics in the same dynamic late endosomal tubules as NPC1 (supplemental movie 3). As we predicted, MLN64 containing late endosomal trafficking is sensitive to its cholesterol content. In WT CHO cells transfected with MLN64-GFP and exposed to LDL and progesterone, MLN64-GFP (Fig. 3E) was localized at the periphery of lysosomes (Fig. 3F), similar to the location of NPC1 (7,21). Late endosomal tubules were rarely observed in the progesterone plus LDL-treated cells. Late endosomal tubular trafficking was also not observed in cholesterol-laden NPC2 and CT60 cells transfected with MLN64-GFP (data not shown). These data show that both MLN64 and NPC1 traffic in late endosomal tubules and the tubular mobility is inhibited when cells accumulate endocytosed cholesterol.
We examined the potential for the endosomal tubules to come into close proximity with mitochondria by imaging cells co-transfected with MLN64-GFP and StAR-RFP, which labels the mitochondrial matrix (supplemental movie 4). We noted parallel alignment of late endosomal tubules with mitochondria in living cells. The late endosomal tubules extending from vesicles increase the surface area of the late endocytic compartment and provide more sites for potential close association with mitochondria, where cholesterol transfer could be effected. Because MLN64-GFP was not anticipated to enter into the mitochondrial matrix, close approximation, but not overlap of the and J, and Fig. 6, A and B). Time lapse movies of ⌬START-MLN64-GFP-expressing living cells show no late endosomal tubular trafficking and immobile ⌬START-MLN64-GFP containing ring-shaped compartments in the perinuclear region of the cell (supplemental movie 5). The dynamic contact and parallel alignment of mitochondria with late endosomal tubules, present in MLN64-GFP-transfected cells, were not present in cells co-transfected with ⌬START-MLN64-GFP and StAR-RFP. The StAR-RFP enters mitochondria and labels the matrix (supplemental movie 6).
Endocytosis assays of WT CHO cells transfected with ⌬START-MLN64-GFP revealed an alteration in the lysosomal filling of the endocytosed fluid phase marker, dextran (Fig. 5). NPC1-GFP-containing late endosomes were shown previously to load with fluorescent dextran in 2 h of uptake/chase time (5,7). In WT CHO cells MLN64 (Fig. 5A) containing late endosomes loaded with dextran (Fig. 5B) in 2 h of uptake/chase time (Fig. 5C, merged image of Fig. 5, A and B). Co-localization of MLN64-GFP (Fig. 5D) and endocytosed dextran (Fig. 5E) is shown in 18 h of uptake/chase time (Fig. 5F, merged image of Fig. 5, D and E). In comparison, in WT CHO cells transfected with ⌬START-MLN64-GFP and incubated with fluorescent dextran in 2 h of uptake/chase time, ⌬START-MLN64-GFP and dextran were in separate compartments (Fig. 5I, merged image of Fig. 5, G and H). Dextran (Fig. 5H) was not present in the ⌬START-MLN64-GFP-containing ring-shaped structures (Fig.  5G). After 18 h of uptake/chase time, dextran (Fig. 5K) was present only associated with ⌬START-MLN64-GFP-containing rings (Fig. 5F, merged image of Fig. 5, J and K). This delayed dextran loading of the late endocytic compartment indicates that 1) ⌬START-MLN64-GFP acts at the latest stage of the endocytic pathway; 2) expression of ⌬START-MLN64-GFP shifts the late endocytic pathway from a dynamic tubular reticulum to separate and immobile compartments; and 3) the alteration of late endocytic compartments impairs the fusion capability of late endocytic compartments with earlier endo-cytic compartments, a phenomenon similar to a novel NPC phenotype revealed by the previous cell fusion study (7).
Because of the altered late endocytic pathway, as described above in both WT CHO cells and COS-7 cells transfected with ⌬START-MLN64-GFP, we examined cholesterol accumulation in these cells that exhibit a phenocopy of lysosomal storage diseases. Interestingly but not surprisingly, filipin staining showed that ⌬START-MLN64-GFP expression resulted in extensive cholesterol accumulation in the abnormally enlarged lysosomes. ⌬START-MLN64-GFP-containing ring-shaped structures (Fig. 6, A and inset) are at the periphery of cholesterol-laden lysosomes (Fig. 6, B and inset), similar to the location of ⌬START-MLN64 and MLN64 in transfected CT60 cells. Neither wild-type MLN64-GFP (Fig. 6C), START-MLN64-GFP (Fig. 6E), nor GFP expression (Fig. 6G) induced cholesterol accumulation in lysosomes (Fig. 6, D, F, and H). In wild-type CHO cells (as well as COS-7 cells; see below), wild-type MLN64-GFP was located in late endocytic compartments throughout the cytoplasm (Fig. 6C), whereas the START-MLN64-GFP (Fig. 6E) and GFP (Fig. 6G) were present in a cytoplasmic and nuclear pattern 12 h to 3 days after transfection.
As shown in Fig. 7, the dominant negative effect of ⌬START-MLN64-GFP was also found in COS cells. COS-7 cells expressing MLN64-GFP (Fig. 7, A, C, and D, and supplemental movie 7) were similar to wild-type CHO cells expressing MLN64-GFP, and COS-7 cells expressing ⌬START-MLN64-GFP (Fig. 7, B and E, and supplemental movie 8) were similar to wild-type CHO cells expressing ⌬START-MLN64-GFP. The ⌬START-MLN64-GFP mutant caused cholesterol accumulation in lysosomes (Fig. 7, F and inset, and supplemental movie 8) and inhibited trafficking of late endocytic tubules in COS-7 cells, whereas the START-MLN64-GFP mutant (Fig. 7G) was located to the nucleus and cytoplasm and did not induce cellular cholesterol accumulation in COS-7 cells.
To explore the biochemical consequences of expressing ⌬START-MLN64-GFP, we examined the effects of co-expression of the GFP constructs on progesterone production in COS cells transfected with plasmids encoding the steroidogenic machinery. To correct for efficiency of expression of the co-transfected steroidogenic enzymes, we normalized the progesterone production by dividing it with the progesterone production of companion cells incubated in the presence of 22-OH-cholesterol, a substrate that readily enters into mitochondria. This hydroxycholesterol substrate supports maximal steroidogenic activity, limited only by the expression of the cholesterol side chain cleavage system, and thus provides an index of the transfection efficiency of the steroidogenic enzymes. We have used this normalization method routinely in our structure-function studies of the StAR protein (10,16). Any adverse effects of the co-transfected GFP constructs on expression of the steroidogenic enzymes would be reflected in reduced conversion of 22-OH-cholesterol into progesterone. Cells transfected with the empty GFP vector produced equivalent amounts of progesterone as cells transfected with ⌬START-MLN64-GFP (5.68 versus 5.99 ng/dish, respectively) (Table I). However, the COS-1 cells transfected with the empty GFP vector produced ϳ40% less progesterone than cells transfected with the other GFP constructs when incubated with 22-OH-cholesterol, demonstrating reduced total steroidogenic potential. The normalized progesterone production data as well as the relative steroidogenic activity (with the empty GFP vector results arbitrarily set to a value of 1.0) revealed a significant inhibitory effect of the ⌬START-MLN64-GFP on steroidogenesis. The MLN64-GFP construct caused a modest but not significant increase in normalized progesterone production (1.3-fold above the empty GFP vector), whereas the START-MLN64-GFP construct increased normalized progesterone production more than 3.7fold, a finding consistent with our previous observations (10). The ability of the START-MLN64-GFP to increase steroid production indicates that fusion of GFP to the C terminus does not abrogate the sterol transfer activity of the START domain. To examine the effects of ⌬START-MLN64-GFP on the utilization of LDL-derived cholesterol, we incubated transfected COS-1 cells with LDL reconstituted with [ 3 H]cholesteryl linoleate and assessed the formation of radiolabeled progesterone. The formation of radiolabeled progesterone by cells transfected with ⌬START-MLN64-GFP (1944 Ϯ 422 dpm/culture dish, mean Ϯ S.D., n ϭ 3) was significantly less (p Ͻ 0.01) than cells transfected with either the GFP empty vector (6111 Ϯ 888 dpm/ culture) or MLN64-GFP (13323 Ϯ 2821 dpm/culture). The significantly reduced steroidogenic activity indicates that expression of ⌬START-MLN64-GFP impairs the utilization of the lysosomal pool of cholesterol for steroidogenesis.

DISCUSSION
The present study demonstrates that 1) MLN64 resides in the NPC1-containing dynamic late endosomal reticulum; 2) overexpression of the transmembrane domain of MLN64 (⌬START-MLN64) induces cholesterol accumulation in lysosomes and inhibits utilization of lysosomal cholesterol by mitochondria for steroidogenesis; and 3) the START domain of MLN64 transfers cholesterol between membranes and stimulates mitochondrial steroidogenesis in vitro.
MLN64 Is a Component of the Dynamic Late Endosomal Tubular Reticulum, and the Dominant Negative MLN64 Mutant Causes Niemann-Pick C Phenocopy-Studies on living cells show that MLN64 is a component of a dynamic late endocytic reticulum that is composed of rapidly trafficking late endosomal tubules (5, 7). Co-localization of MLN64 and NPC1 in the dextran-labeled late endocytic tubulovesicular compartments indicates that these two membrane-spanning proteins could function in similar membrane microdomains. A previous somatic cell fusion study revealed that these late endocytic compartments normally undergo rapid content exchange, including membrane and fluid phase contents (7). MLN64 molecules may be rapidly equilibrated in the late endocytic pathway through these dynamic tubular movements that facilitate interactions between MLN64 and other components, such as NPC1 and HE1, in the late endosomal reticulum.
Visualizing the dynamics of the late endocytic pathway led to finding the dominant negative MLN64 mutant. The ⌬START-MLN64-GFP contains all of the predicted transmembrane domains of MLN64 but lacks the START domain at the C-terminal half of the MLN64 protein. This N-terminal mutant is sufficient to mediate the biosynthetic trafficking of MLN64 from its site of synthesis, via the plasma membrane, to its final destination in the late endocytic compartment. ⌬START-MLN64 not only follows the identical biosynthetic pathway as wild type MLN64 but also co-localizes with wild type MLN64 at the periphery of cholesterol-laden lysosomes. We had shown previously that cholesterol accumulation retards late endocytic tubular formation and mobility (7). While analyzing trafficking of MLN64 mutants, we noted that the morphology, distribution, and movement of endocytic compartments were dramatically altered in wild type CHO cells and COS-7 cells expressing ⌬START-MLN64-GFP. In these cells, ⌬START-MLN64-GFP containing ring-shaped structures aggregated in the perinuclear region and showed limited saltatory movement. Fluorescent filipin cytochemistry showed accumulation of cholesterol in lysosomes in ⌬START-MLN64-GFP-transfected cells, and the mutant protein is present associated with periphery of The finding that ⌬START-MLN64 colocalizes with wild type MLN64 and that overexpression of ⌬START-MLN64 results in cholesterol accumulation in lysosomes suggests that endogenous MLN64 facilitates egress of cholesterol out of these lysosomes. The ability of truncation mutants consisting of the transmembrane domains to prevent the function of endogenous wild-type proteins involved in cholesterol metabolism has also been reported in the case of sterol response element-binding protein cleavage-activating protein (23). The dominant negative effect caused by ⌬START-MLN64 also suggests that another unknown protein may promote the function of MLN64 and that the overexpressed ⌬START-MLN64 interacts with this component, thus blocking the endogenous MLN64 func- tion. The identification of the factor that interacts with MLN64 may lead to the further understanding of protein-protein interactions in the late endocytic compartments as well as cholesterol trafficking. Among the potential MLN64-interacting proteins are NPC1 and HE1, both of which are involved in cholesterol trafficking and homeostasis in the late endocytic compartment. Alternatively, MLN64 may function as a homodimer, and the truncation mutant may prevent dimerization of endogenous MLN64.

The START Domain of MLN64 Mediates Utilization of LDLderived Lysosomal Cholesterol by Steroidogenic Mitochondria-
The START domain of MLN64 shares high (35%) sequence identity within the START domain in the StAR protein, which is abundant in steroidogenic cells (24). The START domains of both MLN64 and StAR are similar in their abilities to bind cholesterol and stimulate steroidogenesis (10,13). StAR is known to mobilize cholesterol from the cholesterol-rich outer mitochondrial membrane to the inner mitochondrial chamber, where the first step in steroidogenesis occurs, conversion of cholesterol to pregnenolone (25). In this study, biochemical analysis demonstrates that steroidogenesis by mitochondria is inhibited when blockage of cholesterol trafficking out of lysosomes by ⌬START-MLN64 occurs. Our data also show that the START domain in MLN64 can transfer cholesterol from donor liposomes to acceptor mitochondrial membranes and stimulate progestin synthesis by placental mitochondria, suggesting that MLN64 effects cholesterol egress out of lysosomes and mediates utilization of lysosomal cholesterol by mitochondria. The ϳ40% inhibition of progesterone production in COS cells transfected with ⌬START-MLN64-GFP compared with that observed with MLN64-GFP is consistent with the 50% reduction in progesterone secretion by granulosa treated with drugs that produce the Niemann-Pick type C phenotype of lysosomal cholesterol storage (25). The inhibition of utilization of LDL-derived cholesterol for steroidogenesis was, however, more profound (68% inhibition compared with empty vector-GFP and 85% inhibition compared with MLN64-GFP), indicating a key role for MLN64 in the trafficking of LDL-derived cholesterol. Collectively, these findings indicate that a portion of the steroidogenic pool of cholesterol moves through the late endosomal system prior to reaching mitochondria, where the first committed step in steroid hormone biosynthesis takes place.
The MLN64-containing compartment may deliver cholesterol substrate to mitochondria in steroidogenic cells, perhaps through transient interactions of the MLN64 START domain associated with the dynamic late endosomal compartment with the outer mitochondrial membranes. It is noticeable that both late endosomal tubular trafficking and mitochondria movement are microtubule dependent (7,26). Using dual color time lapse imaging techniques, we found that the MLN64-GFPcontaining late endosomal tubules align parallel to StAR-RFPlabeled mitochondria and transiently contact mitochondria, providing evidence for spatial interaction between two sterolrelated intracellular organelles (supplemental movie 4). Because we have observed fragments of MLN64 containing the START domain in placental cell extracts (10) and also in placental mitochondrial preparations, 2 it is possible that the cleaved START domain transports cholesterol from the late endosomes/lysosomes to the mitochondria. The START domain may also promote movement of cholesterol from the outer to the inner mitochondrial membrane, where the cholesterol side chain cleavage enzyme resides. This is believed to be the process stimulated by StAR (25). However, the human placenta does not express StAR (16), and we have previously postulated that MLN64 subserves this role of StAR in the human placenta (10). The present observations are fully consistent with this notion. The fact that exogenous MLN64 START domain augments mitochondrial steroidogenesis indicates that whatever endogenous MLN64 is present in the mitochondrial preparations is not sufficient to drive maximal hormone production. This raises the possibility that regulation of MLN64 levels contributes to the control of placental steroidogenesis. Based on our observations on the effects of the ⌬START-MLN64-GFP construct on accumulation of free cholesterol and impairment of steroidogenesis, we conclude that MLN64 is directly involved in lysosomal cholesterol mobilization. MLN64 may have an additional role in steroidogenic cells, where it mediates delivery of substrate to the mitochondrial cholesterol side chain cleavage system.

TABLE I
The ⌬START-MLN64-GFP construct inhibits steroidogenesis COS-1 cells were transfected with plasmids expressing the human cholesterol side chain cleavage and 3␤-hydroxysteroid dehydrogenase enzymes and the indicated plasmids. Cells were cultured in the absence (Ϫ) or presence (ϩ) of (22R)-hydroxycholesterol (5 g/ml). Progesterone production was determined by radioimmunoassay. Values present are means Ϯ S.E. from eight separate experiments in which each treatment group contained triplicate cultures. Relative steroidogenic activities of the constructs are calculated normalizing progesterone production to total cholesterol side chain cleavage and 3␤-hydroxysteroid dehydrogenase activity as determined in the presence of (22R)-hydroxycholesterol. The normalized data were expressed relative to the empty GFP vector control to determine relative activity. Normalized data were analyzed for statistical significance by Mann-Whitney U test. Values with a different superscript letter (a-c) are significantly different from all others in the column (p Ͻ 0.02).