The Lipoprotein Receptor-related Protein-1 (LRP) Adapter Protein GULP Mediates Trafficking of the LRP Ligand Prosaposin, Leading to Sphingolipid and Free Cholesterol Accumulation in Late Endosomes and Impaired Efflux*

One of the conserved functional pathways linked to engulfment of apoptotic corpses involves two membrane proteins low density lipoprotein receptor-related protein-1 (LRP) and ABCA1 and the LRP adapter protein GULP. Because LRP and ABCA1 play roles in cellular lipid trafficking and efflux, here we addressed whether the third member, the LRP adapter protein GULP, also affects cellular lipid transport. Several lines of evidence show that overexpression of GULP causes glycosphingolipid and free cholesterol accumulation in the late endosome/lysosome compartment that is accompanied by down-regulation of ABCA1 and decreased efflux. Conversely, knockdown of endogenous GULP expression promoted cholesterol flux through the late endosomes and up-regulation of ABCA1, even in the context of a disease state such as Niemann-Pick Type C disease. Mechanistically, we were able to show that trafficking of the LRP ligands α2-macroglobulin and prosaposin, a protein cofactor necessary for glycosphingolipid degradation, are impaired in cells expressing full-length GULP protein, resulting in glycosphingolipid and free cholesterol accumulation in the late endosome/lysosome compartment. On the other hand, knockdown of endogenous GULP results in enhanced targeting of prosaposin and enhanced clearance of glycosphingolipids and cholesterol from the late endosomes. Taken together, these data reveal that GULP/LRP/ABCA1 represents a triad of molecules involved in engulfment and cellular lipid homeostasis.

One of the conserved functional pathways linked to engulfment of apoptotic corpses involves two membrane proteins low density lipoprotein receptor-related protein-1 (LRP) and ABCA1 and the LRP adapter protein GULP. Because LRP and ABCA1 play roles in cellular lipid trafficking and efflux, here we addressed whether the third member, the LRP adapter protein GULP, also affects cellular lipid transport. Several lines of evidence show that overexpression of GULP causes glycosphingolipid and free cholesterol accumulation in the late endosome/lysosome compartment that is accompanied by down-regulation of ABCA1 and decreased efflux. Conversely, knockdown of endogenous GULP expression promoted cholesterol flux through the late endosomes and up-regulation of ABCA1, even in the context of a disease state such as Niemann-Pick Type C disease. Mechanistically, we were able to show that trafficking of the LRP ligands ␣ 2 -macroglobulin and prosaposin, a protein cofactor necessary for glycosphingolipid degradation, are impaired in cells expressing full-length GULP protein, resulting in glycosphingolipid and free cholesterol accumulation in the late endosome/lysosome compartment. On the other hand, knockdown of endogenous GULP results in enhanced targeting of prosaposin and enhanced clearance of glycosphingolipids and cholesterol from the late endosomes. Taken together, these data reveal that GULP/LRP/ ABCA1 represents a triad of molecules involved in engulfment and cellular lipid homeostasis.
Engulfment of apoptotic cells is an important process that plays a critical role in development, organogenesis, cellular homeostasis, wound healing, and autoimmunity. Recently, studies from worms to mammals have identified the existence of a dedicated machinery to clear apoptotic cells and have identified several players involved in this process (reviewed in Refs. 1 and 2). Among the eight genes identified to date in Caenorhabditis elegans, the transmembrane proteins CED-1 and CED-7 together with the cytoplasmic adapter protein CED-6 comprise a signaling pathway leading to recognition and engulfment of apoptotic corpses. The mammalian functional orthologue of CED-1 has been identified as LRP1 (the low density lipoprotein-receptor related protein, also known as CD91) (3). LRP1 is a multiligand receptor and has been shown to play a key role in signaling, lipoprotein uptake, and atherosclerosis (4). The CED-7 homologue ABCA1 has been demonstrated to be involved in phospholipid translocation to the outer leaflet of the cell membrane (5) and in engulfment (6). Hamon et al. (5) have shown that overexpression of ABCA1 in cell lines can promote engulfment and that loss of ABCA1 in mice can cause a partial defect in engulfment, supporting it as a member of the triad. Genetic mutations in ABCA1 have been shown to cause Tangier disease (7)(8)(9)(10), which manifests in very low concentrations of plasma high density lipoprotein and accumulation of cholesteryl ester in macrophagic cells in various tissues (reviewed in Ref. 11). The mammalian functional homologue of CED-6 is GULP and has been shown to be a physiological adapter protein of LRP (3,12).
GULP is a 304-amino acid intracellular adapter protein expressed widely in many cell types (13)(14)(15), and it is composed of an N-terminal phosphotyrosine binding (PTB) 4 domain with an adjacent leucine zipper (LZ) domain and a proline-rich C-terminal region. The PTB domain of GULP has been shown to bind an FXNPXY motif on LRP in vitro (3), and serine/threonine phosphorylation of LRP by protein kinase C␣ regulates the interaction with adapter proteins, including GULP, thereby regulating endocytosis (12). The LZ domain mediates homodimerization of GULP (14) and the dimerization is necessary for its function. 5 The proline motifs within the C terminus of GULP may interact with Src homology 3 domains of other signaling proteins. The current models suggest that GULP functions as an adapter protein linking LRP to downstream effectors that mediate LRP uptake and trafficking (1,3). Because the role of LRP and ABCA1 in lipid transport in mammals has been well established (4,16), we postulated that GULP, the third member of the triad, should also function in the regulation of cellular lipid transport.
Here we will show that GULP expression affects cholesterol and sphingolipid traffic through the late endosomes, creating phenotypes similar to lysosomal storage diseases. It has been clearly established that trafficking and storage of cholesterol are linked to glycosphingolipid metabolism (17)(18)(19). Increasing glycosphingolipids by exogenous addition (20) or by inhibition of breakdown (21) results in concomitant free cholesterol accumulation. Additionally, excessive cholesterol loading will also lead to free cholesterol and sphingolipid accumulation (22). The downstream consequences of cholesterol and sphingolipid accumulation would be impairment of clearance of cholesterol from the lysosome compartment and impaired ABCA1-dependent cholesterol efflux, as in NPC disease (23). We therefore examined pathways whereby GULP could affect sphingolipids, particularly its effect on cellular traffic of various LRP ligands, and we show that it regulates the transport of prosaposin, a previously identified LRP ligand that is delivered to the lysosome (24). Therefore, GULP mediates trafficking of prosaposin, thereby affecting the metabolism of glycosphingolipids and the intracellular cholesterol transport as well as ABCA1-dependent cholesterol efflux.

EXPERIMENTAL PROCEDURES
Materials-Plastic cell culture dishes were obtained from Falcon and ␣ culture media from Invitrogen. Fetal bovine serum, bovine serum albumin, paraformaldehyde, filipin, lactoferrin, and methylamine hydrochloride were purchased from Sigma. LPDS was obtained from the fraction of density greater than 1.213 g/ml. Methyl-␤-cyclodextrin (m␤CD) was from Cerestar (Cargill, Inc., Minneapolis, MI was bacterially expressed as a glutathione S-transferase fusion protein, and recombinant RAP was cleaved from the fusion protein and purified (25). Native ␣ 2 -macroglobulin (␣ 2 M) was purified from human plasma by zinc chelate chromatography (26) and activated by incubation with 1.5 M methylamine hydrochloride at room temperature overnight, and the unreacted methylamine was removed by dialysis. RAP and ␣ 2 M were tested for receptor binding activity as described previously (27). Murine embryonic fibroblasts, MEF-2 (28), and 13-5-1 CHO cells (29) are deficient in LRP. Cy3 labeling dye (Amersham Biosciences) was used to label transferrin (Sigma) and ␣ 2 M according to the manufacturer's instructions.
Fluorescence Microscopy-Most images were taken with an Olympus FV1000 confocal microscope with ϫ100 objective (NA 1.4) with a 440 diode laser and 515 argon laser. Other images were taken with an Olympus 100X U Plan Apochromat, NA 1.35-0.50 objective on an Olympus IX70 inverted microscope equipped with a 12-bit IMAGO SVGA CCD camera and the Till Polychrome IV monochrometer (TILL Photonics GmbH) controlled by TillvisION version 4.0 software. LR73 or CHO cells were grown on glass coverslips or coverslip culture dishes (MatTek Corp., Ashland, MA) to ϳ70% confluence. For live cell imaging, cells were maintained at 37°C in a heated chamber and incubated with Hepes-buffered complete ␣ media, pH 7.4. Cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) constructs were excited at 440 and 514 nm, respectively.
Cells-The LR73 CHO cell line has been described previously (14). FL-GULP, FL-GULP with a mutation in the LZ dimerization domain (L176P/L183P), PTB-GULP, LZ-GULP, the C-terminal domain alone, PTB ϩ LZ, or the LZ ϩ C-terminal domain were all cloned into eukaryotic expression plasmids with either an N-terminal GST or hemagglutinin tag, a C-terminal FLAG tag, or a green fluorescent protein (GFP) tag and stably transfected into LR73 together with a puromycin plasmid.
The positive LR73 clones expressing the protein of interest were identified by Western blotting, and multiple independent clones were routinely analyzed. A GULP polyclonal antibody was generated against the PTB domain of GULP and used for Western blotting at a dilution of 1:5000. The coding sequence for human prosaposin was cloned into a mammalian expression vector with CFP at the C terminus to avoid disrupting targeting information for prosaposin.
Free Cholesterol Staining, Visualization, and Localization-Cells, grown in ␣ media with either 10% FBS or 10% LPDS under a variety of conditions, were fixed by incubation with 3.3% paraformaldehyde for 1 h. Cells were washed extensively with phosphate-buffered saline; the fixative was quenched by a 30-min incubation with 50 mM ammonium chloride. The cells were washed extensively, and 0.05 mg/ml filipin in phosphate-buffered saline was then added and incubated for 1-2 h. Fluorescence microscopy was performed at the excitation wavelength of 390 nm. BODIPY FL C 5 -lactosylceramide (Molecular Probes, Eugene, OR) was equilibrated with FBS and added to cells for 16 h in the presence of 10% FBS, and its accumulation was determined at 505 nm excitation wavelength. The FL-GULP cells were transfected with cDNA constructs for Rab4, Rab5, Rab7, or Rab11 each fused to YFP (excitation wavelength 514 nm; kind gifts of Dr. Marino Zerial). Cells were then fixed and stained for filipin as described above for determination of colocalization of cholesterol and an intracellular compartment. Alternatively, cells were incubated with lysotracker red (10 nM; excitation wavelength 560 nm) in ␣ medium for 15 min prior to washing and visualization. Cells were incubated with Cy3-labeled transferrin or Cy3labeled ␣ 2 M for various times at 37°C before washing the cells and performing fluorescence microscopy in a heated chamber. Cells were fixed and immunostained with a monoclonal antibody to EEA1 (BD Biosciences) or a polyclonal antibody to prosaposin (kind gift of Dr. Carlos Morales), at a 1:100 ratio, and then with an Alexa 594 secondary antibody (Molecular Probes, excitation wavelength 590 nm) for visualization. Prosaposin-CFP was transiently transfected into the parental, FL-GULP-, PTB-GULP-, and AS-GULP-expressing cells and localized with lysotracker staining (or alternatively cotransfected with Rab7). Prosaposin-CFP completely colocalized with prosaposin immunofluorescence.
Clearance of Cholesterol from the Late Endosome/Lysosome-Parental, FL-GULP, and PTB-GULP cells were maintained in 10% LPDS and then grown in 10% FBS for 1, 4, or 16 h. Cells were then fixed and stained as above. As a second experiment, FL-GULP, PTB-GULP, or parental cells were maintained in 10% LPDS and then grown in 10% FBS for 4 h, washed extensively, and then switched back to 10% LPDS for 0, 3, or 12 h. Cells were then fixed and stained as above. For the third experiment, parental, FL-GULP, and PTB-GULP cells were labeled with [ 3 H]cholesterol (equilibrated with 10% FBS) for 36 h. The plasma membrane cholesterol was extracted at 4°C with 10 mM m␤CD for 10 min. The cells were then incubated at 37°C for various times, and the radioactive cholesterol was transferred to the PM and assessed by extracting again at 4°C with 10 mM m␤CD for 10 min. Radioactivity was determined by scintillation counting.
Cholesterol Esterification in GULP-expressing Cells-Parental, FL-GULP, and PTB-GULP cells were labeled by incubation with [  Knockdown of GULP Expression in LR73 Cells-Full-length mouse GULP cDNA was PCR-amplified and digested with NotI and SalI. The purified fragment was then subcloned into pCMV⅐SPORT6 expression vector between NotI and SalI to give the reverse orientation of GULP cDNA. The antisense plasmid was cotransfected into LR73 cells together with a neomycin resistance plasmid (ratio of 10:1) using Lipofectamine 2000 reagent. Clones were selected in 4 mg/ml G418 for 2 weeks. The resultant clones were screened by Western blotting for decrease or loss of GULP protein expression and maintained in 2 mg/ml G418 in ␣-minimal essential media with 10% FBS.
Knockdown of Prosaposin Expression in OV3121 Cells-OV3121 cells are a murine ovarian granulosa tumor cell line (30) similar to CHO cells. Three siRNA primers specific for prosaposin (Ambion) were used in combination for transfection with Lipofectamine 2000, and a scrambled primer was used as a control. Knockdown of prosaposin was confirmed by Western blotting using a polyclonal antibody (a kind gift from Dr. Carlos Morales). Glycosphingolipid synthesis and accumulation were measured following a 30-h incubation of cells with [ 3 H]palmitate. Lactosylceramide and glucosylceramide were extracted and separated by TLC (31).
Cholesterol Efflux Assays-Efflux assays were performed as described (32) except that 25 g/ml apoA-I was used. ABCA1 Western blots were performed as described (32) using a polyclonal antibody from Novus Biologicals.

Accumulation of Free Cholesterol in Cells because of GULP Overexpression-
To address the potential role of GULP in regulating cellular cholesterol metabolism, we either generated stable clones of LR73 CHO cells expressing wild-type or mutant versions of GULP (see supplemental Fig. 1, schematic) or knocked down endogenous expression of GULP. To compare the free cholesterol localization, the cells were grown in fetal bovine serum (FBS)-containing medium, fixed, and stained with filipin, a fluorescent dye specific for free cholesterol. Full-length GULP (FL-GULP)-expressing cells (Fig. 1A) showed multiple, bright, and distinct cholesterol-rich endosomes, whereas the normal cholesterol distribution of parental LR73 cells, cells expressing the PTB domain alone (PTB-GULP), or LR73 cells with markedly reduced levels of endogenous GULP (AS-GULP) presented a normal cholesterol distribution (Fig.  1A). Cells expressing other GULP deletion mutants, the leucine zipper domain alone (LZ-GULP), the C-terminal domain alone, the PTB and LZ domains, the LZ and C-terminal domains, or the full-length GULP with two mutations in the LZ domain (FL-GULP-mLZ) all showed normal cholesterol localization with staining of the cell membrane, diffuse staining of the cytoplasm, and occasionally brighter staining of the perinuclear region (supplemental Fig. 2). To rule out the possibility that the effect seen was because of the GST tag added to the N terminus of GULP, we generated LR73 cells stably expressing FL-GULP with N-terminal hemagglutinin tag, C-terminal GFP tag, or a C-terminal FLAG tag. The phenotypes of these cells were identical to that of the GST-FL-GULP cells (data not shown) ruling out the effects of the tags or their positioning. When we analyzed multiple, independent FL-GULP clones, there was a general correlation between expression level of GULP (supplemental Fig. 3) and the occurrence and amount of the cholesterol-rich endosomes (supplemental Fig. 2). None of the mutant forms of GULP elicited a cholesterol accumulation similar to intact FL-GULP (supplemental Fig. 2). Expression levels of the transfected proteins were about 0.6 -3.8-fold of endogenous GULP expression levels (supplemental Fig. 3). Endogenous GULP expression levels were significantly decreased in AS-GULP cells (supplemental Fig. 4).
GULP Affects Cellular Cholesterol Homeostasis-Interestingly, the accumulation of free cholesterol was reminiscent of the cholesterol trafficking defects seen in cells with NPC disease (33), and this led us to examine cholesterol homeostasis in FL-GULP cells by measuring how cholesterol (either exogenously derived or endogenously synthesized) is converted to cholesteryl ester. Parental, FL-GULP, PTB-GULP, or AS-GULP cells were grown in the presence of either [ 3 H]acetate (to label endogenously synthesized cholesterol and cholesteryl ester) or [ 3 H]cholesterol pre-equilibrated with FBS (to label cholesterol derived from exogenous sources). Although the radioactivity levels of total cholesterol synthesized by endogenous pathways or acquired from exogenous sources (free cholesterol plus cholesteryl ester) were similar between parental and FL-GULP cells, FL-GULP cells had significantly lower cholesteryl ester formation under both types of labeling ( Fig. 1, B and C) as compared with the other cell types. On the other hand, there was a small but significant decrease in the free cholesterol in PTB-GULP and AS-GULP cells compared with parental cells, with exogenously added cholesterol. This result also demonstrates that the PTB-GULP cells display a phenotype indistinguishable from AS-GULP cells, suggesting that the mutant(s) likely represent dominant negative mutations that impair endogenous GULP function. The total cellular cholesterol mass in FBS-treated cells is only slightly elevated in FL-GULP cells, compared with parental LR73 cells, PTB-GULP cells, and AS-GULP cells (Fig. 1D), confirming the results with radioactive cholesterol, but different from the reported cholesterol accumulation in NPC disease. Acyl-CoA:cholesterol acyltransferase activity was measured in a cholesterol esterification assay and showed that in FL-GULP cells there was a significant decrease in the esterification rate compared with parental cells, PTB-GULP cells, and AS-GULP cells (Fig. 1E). This indicated that the FL-GULP cells do not accumulate significantly more cholesterol but sequester it as free cholesterol, instead of storing it as cholesteryl ester (as seen with the parental, PTB-GULP, and AS-GULP cells).
Accumulation of Glycosphingolipid in Cells because of GULP Overexpression-In NPC disease, the affected cells display glycosphingolipid as well as free cholesterol accumulation in the late endosome/lysosome compartment (19,33,34). Thus, we assessed the cellular localization of internalized BODIPY-lactosylceramide, a fluorescent glycosphingolipid. There was plasma membrane and some minor intracellular accumulation in the parental, the PTB-GULP, and the AS-GULP cells ( Fig.  2A). In contrast there was significant accumulation of BODIPY-lactosylceramide in the FL-GULP cells ( Fig. 2A). Moreover, BODIPY-lactosylceramide and filipin staining in FL-GULP cells overlapped (data not shown). As well, newly synthesized glycosphingolipids (lactosylceramide and glucosylceramide) were significantly increased in FL-GULP cells compared with the other cell types (Fig. 2B). Thus, based on this characterization, overexpression of FL-GULP induces localized cholesterol and glycosphingolipid accumulations.
FL-GULP Causes Free Cholesterol Accumulation in Late Endosomes/ Lysosomes-We then confirmed the subcellular location of the choles terol accumulation. In the FL-GULP cells, the filipin staining significantly colocalized with lysotracker red and Rab7, a small GTPase that localizes to the late endosome/lysosome (Fig. 3). However, the filipin staining did not colocalize with Rab5, which labels the early endosomal population, Rab4, which labels the sorting endosomes (data not shown), or Rab11, which labels the recycling endosomes (data not shown) (35). Taken together, these data suggest that the free cholesterol accumulation because of overexpression of FL-GULP occurs in a late endosomal/ lysosomal compartment.
In contrast to the parental cells that show normal late endosomes/ lysosomes, the FL-GULP cells showed an accumulation of enlarged electrolucent late endosomes structures by electron microscopy (Fig. 4) reminiscent of NPC disease cells (36). In Fig. 4, the mitochondria are indicated by arrowheads and late endosomes/lysosomes are indicated by asterisks. When cells were examined close up, we found the appearance of some maturing late endosomes (maturing multivesicular bodies, electrodense) and lysosomes (multilamellar structures) in parental cells. In FL-GULP cells, we see the immature late endosomes structures (multivesicular bodies) with very little electrodense material contained within the late endosomes.
GULP Affects Cholesterol Clearance from Late Endosomes-To determine the effect of FL-GULP in transport and clearance of exogenous lipoprotein cholesterol, parental and FL-GULP cells were grown in the   presence of 10% LPDS for 2 days. This resulted in a significant reduction in cholesterol accumulation in the FL-GULP cells, suggesting that the accumulation of free cholesterol depended on the presence of extracellular lipoprotein (data not shown). The cells were switched to media containing 10% FBS for 1, 4, or 16 h and assessed for cholesterol accumulation. FL-GULP cells accumulated equivalent amounts of cholesterol at 1-and 4-h time points as parental cells (Fig. 5A). After 16 h, the parental cells had cleared the accumulated cholesterol from the late endosomes, whereas the FL-GULP cells retained significant free cholesterol accumulation. The PTB-GULP and AS-GULP cells displayed a pattern similar to the parental cells (data not shown). To examine further the impaired clearance of the internalized cholesterol because of GULP overexpression, we used two approaches. First, we performed a cholesterol pulse-chase experiment. The parental and FL-GULP cells were maintained in media with 10% LPDS and then "pulsed" for 4 h with media containing 10% FBS, after which the cells were washed and switched back to 10% LPDS containing medium; at various time points, free cholesterol accumulation in the cells was monitored. Parental cells quickly cleared the accumulated cholesterol from the late endosome compartment (after 3 h) and had relatively enhanced staining of PM; however, the FL-GULP cells had accumulated late endosomal cholesterol even after 12 h (Fig. 5B). The PTB-GULP and AS-GULP cells displayed a pattern similar to parental cells (data not shown). As a second approach (more quantitative) to measure clearance of internalized cholesterol, we directly tested the rate of return of cholesterol from intracellular compartments (including the late endosome) to the PM. Parental, FL-GULP, PTB-GULP, and AS-GULP cells were labeled with [ 3 H]cholesterol, and the plasma membrane cholesterol was extracted at 4°C with m␤CD followed by a 15-min incubation at 37°C, and then a final extraction at 4°C. FL-GULP cells had a significantly reduced rate of cholesterol transport to the PM compared with parental cells (Fig. 5C), in complete agreement with the pulse-chase experiments above (Fig.  5B). The PTB-GULP and AS-GULP cells had an enhanced rate of transfer of cholesterol to the PM, compared with the parental cells (Fig. 5C). Therefore, the primary consequence of FL-GULP expression on cholesterol transport is a repression of its clearance from the late endosome/ lysosome compartment, whereas knockdown of endogenous GULP results in enhanced clearance.
Reduction of GULP Expression Stimulates Cholesterol Efflux-Because cholesterol efflux to apoA-I is dependent on ABCA1 protein and because the homologues of GULP and ABCA1 are both in the same genetic pathway in C. elegans, we examined the effect of reduced GULP expression on ABCA1-dependent cholesterol efflux to apoA-I. Surprisingly, compared with control cells, the efflux of cholesterol to apoA-I was enhanced almost 200% in the cells with reduced GULP expression (AS-GULP) (Fig. 6A). Furthermore, LR73 cells overexpressing FL-GULP showed a significant decrease in cholesterol efflux (Fig. 6B), essentially opposite of the phenotype seen with reduced GULP expression above. Cholesterol efflux correlated with levels of FL-GULP expression (data not shown). Multiple stable lines expressing GULP protein with mutations or deletions (supplemental Figs. 1 and 2) had greater cholesterol efflux compared with parental LR73 cells, suggesting a possible dominant negative effect of these mutants ( Fig. 6B and data not shown). Efflux of newly synthesized cholesterol (derived from [ 3 H]mevalonate) was significantly decreased in FL-GULP cells compared with parental cells, but significantly increased in PTB-GULP cells and AS-GULP (Fig. 6C). Phospholipid efflux, measured under identical conditions to cholesterol efflux, was also decreased in FL-GULP-overexpressing cells compared with control cells and significantly increased in the PTB-GULP mutant and AS-GULP cells (Fig. 6D). On the other hand, cholesterol efflux to high density lipoprotein showed no significant difference between the different cell types (data not shown). These results suggest that the effect of GULP expression may not only affect cholesterol availability for apoA-I-mediated efflux but also regulate the efflux activity of ABCA1.
Determination of ABCA1 levels by Western blotting showed that ABCA1 expression was enhanced in parental cells under conditions of cholesterol loading, but it was unchanged in FL-GULP-expressing cells (Fig. 6E). However, interestingly, cells expressing PTB-GULP and AS-GULP had significantly enhanced ABCA1 expression levels compared with parental cells in both conditions of cholesterol deprivation and cholesterol loading. This demonstrated that GULP affects ABCA1 protein expression, which correlates with cholesterol and phospholipid efflux. Consistent with this notion, binding of 125 I-apoA-I in the parental, FL-GULP, PTB-GULP, and AS-GULP cells correlated with ABCA1 expression levels (data not shown).
The consequence of impaired GULP function can also be tested in the context of a disease state such as NPC disease. CHO cells with an NPC1 functional mutation (37) were transiently transfected with PTB-GULP-GFP and then fixed and stained with filipin (Fig. 7). NPC cells accumulated large amounts of free cholesterol (Fig. 7, green), but the cells expressing PTB-GULP-GFP (Fig. 7, red) show little accumulation of free cholesterol and normal localization of free cholesterol. This result demonstrates that inactivation of endogenous GULP function results in enhanced trafficking through the late endosomal compartment, which, associated with the increased ABCA1 and efflux shown above, can substantially correct a NPC disease phenotype.

FL-GULP Is a Cytosolic Adapter Protein That Does Not Colocalize with the Accumulated Free Cholesterol in the Late Endosomes/Lysosomes-The
effect of GULP expression is the accumulation of cholesterol in the late endosomes, whereas the effect of inactivation of GULP is the stimulation of trafficking of cholesterol through the late endosomes. To understand the mechanism of this action, we examined whether GULP was recruited to the free cholesterol-containing late endosomes. FL-GULP-GFP was found to be predominantly in the cytoplasm and variably on the plasma membrane; however, in 10% of transfected cells GULP was recruited to distinct punctate structures (Fig. 8A). These structures did not colocalize with filipin or lysotracker (Fig. 8B and data not shown) but instead partially associated with the Rab5 effector protein EEA1   ). B, FL-GULP-GFP did not colocalize with filipin but did partially associate with EE-A1, transferrin, and ␣ 2 M-positive endosomes (indicated by arrows). Scale bar represents 1 m for all panels. (Fig. 8B). Consistent with recruitment to an early endosome compartment, the FL-GULP structures associated with Cy3-labeled transferrinpositive endosomes (Fig. 8B). Interestingly, the acquisition of time lapse images demonstrated that FL-GULP-GFP was not evenly recruited to the transferrin-positive structures, rather it appeared to be enriched in subdomains on these endocytic organelles (see movie attached as supplemental Fig. 5). Thus, the FL-GULP and the transferrin-containing vesicles do not overlap but rather comigrate. Evidence of similar interactions was also obtained between Cy3-␣ 2 M (an LRP ligand) and FL-GULP-GFP (Fig. 8B). Taken together, these data suggest that FL-GULP itself is not enriched in compartments where free cholesterol accumulates but is mainly cytosolic with some association to both the plasma membrane and mobile transferrin-and ␣ 2 M-positive endosomes. These data also suggested that GULP may modulate trafficking of cholesterol by regulating step(s) upstream of the late endosome compartment where the free cholesterol accumulates.
GULP Functions via LRP but GULP Does Not Affect LRP-mediated Uptake of Lipoproteins-We and others have shown previously that GULP and LRP coimmunoprecipitate, suggesting a functional interaction (3,12). We tested the hypothesis that the accumulation of cholesterol might be due to hyper-internalization of cholesterol-enriched lipoproteins. GULP expression did not stimulate uptake of apoE-enriched ␤-very low density lipoprotein (a lipoprotein with a higher affinity for LRP (38)) compared with control cells (Fig. 9A), nor did GULP expression significantly affect uptake of LDL immunodepleted of apoE, a lipoprotein with a higher affinity for the LDL receptor (Fig. 9B). These results indicated that the effect of FL-GULP on LRP does not result in hyper-internalization of cholesterol-containing lipoproteins.
We then assessed the cell surface expression of LRP. Parental, FL-GULP, PTB-GULP, and AS-GULP cells were incubated with 125 Ilabeled ␣ 2 M at 4°C. Binding was measured with increasing concentrations of ligand and also in the presence of RAP in all cell lines (to control for nonspecific binding; but only one line is shown for clarity) (Fig. 9C). FL-GULP cells had very low levels of cell surface LRP, barely above the RAP-inhibitable binding. In contrast, the parental LR73 cells had a mod-erate level of binding, and the PTB-GULP and AS-GULP cells had the highest cell surface expression of LRP. Taken together these data suggested that GULP influences the cell surface presentation of LRP.
Next, we analyzed the effect of GULP in the absence of LRP on lipoprotein uptake. In LRP-deficient CHO cells (13-5-1) (29), the uptake of LDL by the LDL receptor was not affected by expression of GULP (supplemental Fig. 6A). However, compared with control CHO cells, there was a clear accumulation of free cholesterol in endosomes and perinuclear sites of the LRP-deficient 13-5-1 cells but not in the LRPreconstituted 13-5-1 cells (100-10 cells; see supplemental Fig. 6B). In addition, stable overexpression of FL-GULP in LRP-deficient 13-5-1 cells did not elicit accumulation of cholesterol (supplemental Fig. 6B). These results demonstrate that focal accumulation of cholesterol can occur to some degree in LRP deficiency itself but that FL-GULP-induced accumulation is dependent on LRP. These data also suggest that endosomal cholesterol accumulation is not induced by internalization of lipoprotein-derived cholesterol, per se, but that it is rather the mistargeting or mis-trafficking of other LRP ligands that indirectly leads to impairment of cholesterol transport, either during intracellular recycling or in receptor-mediated uptake of lipoproteins.
GULP Causes Mis-targeting of the LRP Ligand ␣ 2 M-␣ 2 M, a high affinity ligand for LRP, is taken up via LRP-mediated endocytosis and trafficked to the late endosomes for intracellular degradation. We followed the trafficking of this LRP ligand in parental, FL-GULP-, PTB-GULP-, and AS-GULP-expressing cells (Fig. 10). In parental, PTB-GULP-, and AS-GULP-expressing cells, ␣ 2 M significantly colocalized with lysotracker in late endosomes. However, in FL-GULP-expressing cells, ␣ 2 M was present in multiple endosomes compartments, mostly at the opposite poles of the cells, with very little colocalization with lysotracker dye. This result demonstrated the mis-targeting of an LRP ligand in FL-GULP cells away from the late endosomes and suggested the possible mechanism of action of GULP. However, this result fails to adequately explain cholesterol accumulation because of FL-GULP overexpression, leading us to look for other LRP ligands. GULP Causes Mis-targeting of the LRP Ligand Prosaposin-Saposins, small heat-stable glycoproteins derived from a common precursor prosaposin, are sphingolipid-binding proteins that act as cofactors for the lysosomal degradation of glycosphingolipids. Prosaposin is also an LRP ligand, and LRP mediates uptake and delivery of prosaposin to the lysosome (24). A transiently expressed epitope-tagged prosaposin (prosaposin-CFP) is synthesized and secreted, re-internalized via LRP, and delivered to the late endosomes, so that by 1 h of post-cycloheximide treatment prosaposin-CFP does accumulate in late endosomes in parental cells. In FL-GULP-expressing cells, prosaposin was present in multiple compartments with only a minor extent of colocalization with the lysotracker dye (Fig. 11) and was completely excluded from large cholesterol-enriched late endosomes. However, in parental cells, prosaposin significantly colocalized with or was adjacent to lysotracker in late endosomes (Fig. 11). Although nonquantitative, colocalization of prosaposin and lysotracker dye was clearly enhanced in PTB-GULP and AS-GULP cells. Essentially the same staining was observed when colocalization with Rab7 was examined (data not shown). These results demonstrate mis-targeting of prosaposin to the late endosome in FL-GULP cells and enhanced targeting of prosaposin in PTB-GULP and AS-GULP cells.
As a control, prosaposin expression was knocked down by siRNA in a murine ovarian cell line (OV3121 (30)) similar to CHO cells, and the cells were assessed for the accumulation of glycosphingolipid and free cholesterol. Cells with decreased endogenous prosaposin expression (confirmed by Western blot, see Fig. 12A) showed increased free cholesterol accumulation and BODIPY-lactosylceramide staining (82 and 88% of cells presented with free cholesterol and glycosphingolipid accumulation, respectively; representative pictures are shown in Fig. 12B). In contrast, only 7 and 12% of control cells showed any accumulation of cholesterol or glycosphingolipid, respectively. The prosaposin knock-down cells had a phenotype very similar to FL-GULP-expressing cells. These results demonstrate that mis-targeting or decreased expression of the LRP ligand prosaposin leads to the accumulation of glycosphingolipid and cholesterol.

DISCUSSION
Here we have made a series of novel observations linking the LRP adapter protein GULP to cellular lipid transport and homeostasis within cells. Expression of FL-GULP results in accumulation of glycosphingolipid and free cholesterol in the late endosomes/lysosomes, impaired cholesterol transport to the ER, and inhibition of free cholesterol clearance from the late endosome/lysosome compartment, thereby affecting its export by ABCA1-dependent efflux. Conversely, knockdown of endogenous GULP expression or expression of dominant negative mutants of GULP results in enhanced clearance of free cholesterol and enhanced ABCA1-mediated efflux. Expression of a dominant negative   mutant of GULP even partially corrected the storage defect in NPC CHO cells. Taken together, these data suggest a role for GULP in modulating glycosphingolipid and cholesterol transport in cells.
GULP Effect on Cholesterol Transport and Clearance from the Late Endosomes-In NPC disease cells, there is accumulation of cholesterol in the late endosomes and prevention of clearance of cholesterol from the late endosomes. No cholesterol is trafficked to the ER, and no regulatory signal occurs in the ER to modulate cholesterol homeostasis. Because of this, the LDL receptor is not down-regulated, and there is continual uptake of LDL-derived cholesterol. This is not the case in FL-GULP-expressing cells, as these cells do not have an increase in cholesterol uptake, synthesis, or total cholesterol mass. There is no significant difference in the uptake and accumulation of lipoprotein-derived cholesterol between the cell types. This represents a difference between NPC disease and the condition induced by FL-GULP expression, suggesting that NPC disease is a more severe phenotype. Clearance of cholesterol from the late endosomes was described in detail by the pulse-chase experiments, in which clearance was shown to be impaired in FL-GULP-expressing cells. Interestingly, although no difference was observed by qualitative filipin staining, trafficking to the plasma membrane (m␤-CD experiment; see Fig. 5C) was significantly stimulated by expression of the PTB-GULP mutant or by knockdown of endogenous GULP.
GULP Effect on Cholesterol Esterification and Efflux-Our experiments show that GULP overexpression affects cholesterol transport out of late endosomes to the ER and the PM, which efflux specifically to apoA-I, and reduces esterification of endogenously synthesized or exogenously acquired cholesterol. This suggested that normal endogenous GULP function imparts a restrictive pressure on cholesterol (normally targeted for transport to the plasma membrane or toward efflux) in the late endosome/lysosome, and that knockdown of endogenous GULP or expression of a mutant GULP releases this pool of cholesterol for efflux and indirectly increases ABCA1 expression level.
The late endosomes and lysosomes play a major role in the sorting and transport of cholesterol and are exquisitely sensitive to modification of their components as exemplified by the complex and not fully understood NPC phenotypes. The NPC phenotype can be induced by increased delivery or decreased catabolism of either sphingolipid or cholesterol (17)(18)(19)22), and accumulation of cholesterol or glycosphingolipid prevents cholesterol mobilization and efflux (20,23,34). Salvioli et al. (39,40) showed that cholesterol levels modulate the subcellular localization of glucosylceramidase, lyso-bisphosphatidic acid, and saposin C, albeit in the context of a NPC cell model, suggesting that the lipids themselves (glycosphingolipids and cholesterol) can disrupt the functioning of the late endosomes. GULP is an adapter protein for LRP and is not localized to the late endosomes, thus we would not predict any interaction between GULP and other proteins of the lysosomal machinery (e.g. NPC1). NPC2, a soluble cholesterol-binding protein, can be secreted from liver cells into the plasma or bile with other lysosomal hydrolases such as cathepsin-D or ␤-galactosidase (41), and NPC2 is reinternalized via mannose 6-phosphate receptors (42), but we cannot exclude the involvement of LRP or GULP at the moment. Interestingly, we found that transient transfection of FL-GULP into different NPC1overexpressing cells (43) reduces cholesterol efflux independently of NPC1 expression levels (data not shown), suggesting that GULP acts independently of NPC1. As well, transfection of PTB-GULP into NPC disease cells enhanced the trafficking through the late endosomal compartment and substantially corrected the NPC disease phenotype (Fig.  7). Similar results were obtained with expression of Rab7 or Rab9 or treatment with oxysterols, which were able to bypass the defect in the NPC disease cells (44,45). Importantly, total cellular cholesterol does not increase in FL-GULP cells compared with control cells. Therefore, this phenotype is unlike NPC disease cells, where cholesterol is increased, and therefore suggests an independent cause of glycosphingolipid and free cholesterol accumulation, not involving NPC1 or NPC2.
Role of GULP in Transport of LRP Ligands-LRP gene expression, unlike the LDL receptor, is not sensitive to cellular cholesterol levels (46). However, LRP can be recruited from an intracellular pool to the cell surface by treatment with insulin, and recruitment can be inhibited by wortmannin, similar to the transferrin receptor (47). Because GULP is a physiological adapter protein of LRP, we propose that overexpressed FL-GULP impairs the proper cellular transport of LRP, its cargo, or associated adapters and scaffold proteins (48) through its interactions in the early/recycling endosomes. However, it is not known whether FL-GULP expression promotes trapping of LRP intracellularly, promotes degradation of LRP, or inhibits recycling of LRP to the plasma membrane and what happens to the associated cargo and adapter proteins, but we are currently addressing these issues. Accumulating evidence suggests that the transition from an early to late or recycling endosome is because of the precisely controlled exchange of cytosolic proteins and adapters (49 -51). The recruitment of these proteins is regulated by a combinatorial network of interactions with receptor proteins, Rab GTPases, and cholesterol-based microdomains (35,51). In principle, if overexpressed GULP were to bind LRP longer than required, the transition of receptor and cargo into the next endocytic compartment would be compromised. In this case, the physiological role for endogenous GULP may be to restrict LRP in the early endocytic structures long enough to properly sort its cargo, whereas overexpressed GULP may dominantly block this maturation step. The absence of GULP, on the other hand, may alleviate any restriction on cargo sorting. This is indeed what we observe in that GULP expression modulates LRP cell surface expression (Fig. 9C) and the residence of LRP cargoes in the early and late endosomes, in the case of ␣ 2 M (Fig. 10) and prosaposin (Fig. 11).
We presume that GULP plays a general role in LRP ligand-mediated internalization and trafficking. We see the most striking phenotype of GULP overexpression on cellular cholesterol distribution without any significant change in total cellular cholesterol levels, but our results are also consistent with a general effect of GULP on endosomal trafficking of LRP ligands. LRP ligands, such as prosaposin, destined for the late endosome, are essential for the degradation of glycosphingolipids in the late endosome/lysosome (24). It is likely that a reduction by GULP in the delivery of prosaposin and possibly other similar cargoes to the late endosome compartments could result in accumulation of cholesterol and glycosphingolipids in the late endosomes of GULP-expressing cells. It has been shown that other soluble lysosomal enzymes like cathepsin D, ␤-glucosylceramidase, and ␣-glucosidase are dependent on LRP for their proper routing to the late endosome (52-54), and thus the effect of GULP may be further implicated in the routing of these proteins. Similarly, reduction of prosaposin levels exacerbates a defect in ␤-glucosylceramidase (as in Gaucher disease) (55), suggesting a functional link between these two proteins. The proposed mechanism of secretion/recapture for prosaposin may be necessary because of the proposed role of saposins as neurotrophic factors (56,57) and as carrier proteins for sphingolipids and gangliosides (58). Saposins have also been shown recently to facilitate lipid presentation to CD1d on T cells (59,60). Alternate pathways in specific cells may also be present, including the mannose receptor and the mannose 6-phosphate receptor for uptake of secreted prosaposin (61), and a fraction of prosaposin is transported intracellularly, mediated by sortillin (62).
The focal accumulation of cholesterol in endosomes and perinuclear sites of the LRP-deficient 13-5-1 cells but not in the LRP-reconstituted 13-5-1 cells is also compatible with a role of LRP in prosaposin transport. The lack of effect of GULP expression in that context (supplemental Fig. 6B) supports a model of LRP-mediated transport of prosaposin regulated by GULP.
Prosaposin Deficiency and Cholesterol Accumulation-Prosaposin knock-out mice (63) or tissues from individuals with a natural prosaposin deficiency (31,64,65) are characterized by sphingolipid accumulation. A recent report (31) clearly identified prosaposin deficiency as a lipid storage disease with significant accumulation of glycosphingolipids in the lysosomes of fibroblasts, macrophages, and cells of the adrenal cortex. Is it also a cholesterol storage disease? Previously, the accumulation of total cellular cholesterol at the cellular or tissue level has not been observed in any of the reported cases of prosaposin deficiency (mouse or human), and we have also not observed this, but only a minor increase in total cellular cholesterol mass (Fig. 1D). On the other hand, there have been many reports that document accumulation of lactosylceramide and glucosylceramide, and we have also observed this effect (Fig. 2, A and B). It is well documented that the affinity of cholesterol for sphingolipids causes their colocalization, and their coaccumulation in lipid storage diseases has been reported (20,33). It has been proposed that accumulation of sphingolipids in late endosomes of sphingolipid storage disease cells creates a molecular trap for cholesterol (20), and a similar mechanism has been shown to operate in sphingomyelinasedeficient macrophages, where cholesterol and sphingomyelin accumulate in late endosomes (66). Several investigators suggest that NPC is primarily a glycosphingolipid storage disease (17,67), and the accumulation of cholesterol is a secondary effect. The effects of GULP overexpression on prosaposin transport and glycosphingolipid metabolism are certainly compatible with the view that defects in NPC primarily affect the regulation of sphingolipid metabolism and secondarily cholesterol transport through the late endosomes. We believe this is the first report of the intracellular accumulation of cholesterol and glycosphingolipid in the late endosomes/lysosomes upon prosaposin mistargeting or deficiency. This phenotype is unlike NPC disease cells (where total cellular cholesterol is increased) but consistent with the phenotype of prosaposin knock-out mice (63) or tissues from individuals with a natural prosaposin deficiency (31,64,65), where total tissue or total cell cholesterol levels are unaltered.
It is intriguing that mammalian homologues of all three members of a functional genetic pathway involved in engulfment of apoptotic cells (conserved from C. elegans to mammals) are linked to intracellular lipid transport and secondarily to cholesterol metabolism in mammalian cells. Thus, interesting future avenues of investigation will be to determine how engulfment, lipid transport, and cholesterol metabolism are linked, and which specific molecular and functional features of LRP, ABCA1, and GULP make these proteins part of an evolutionarily conserved signaling module involved in both engulfment and sphingolipid and cholesterol homeostasis. Our observations further emphasize the importance of understanding the trafficking of cholesterol and the role of adapter proteins such as GULP in the regulation of cellular cholesterol transport and homeostasis. In this context, impairment of GULP function (by knockdown of endogenous GULP or by expression of PTB-GULP or other mutants) may serve to release cholesterol from a storage pool and allow cholesterol export and regulation of its endogenous synthesis. Stimulating an efflux pathway would serve a very important role in foam cells in atherosclerotic lesions and in cholesterol and sphingolipid storage diseases, as GULP inactivation may alleviate those disease state conditions.