Niemann-Pick Type C (NP-C) disease, caused by mutations in either human NPC1 (hNPC1) or human NPC2 (hNPC2), is characterized by the accumulation of unesterified cholesterol in late endosomes. Although it is known that the NP-C proteins are targeted to late endosomal/lysosomal compartments, their delivery mechanisms have not been fully elucidated. To identify mechanisms regulating NP-C protein localization, we used Saccharomyces cerevisiae, which expresses functional homologs of both NP-C proteins – scNcr1p and scNpc2p. Targeting of scNcr1p to the vacuole was perturbed in AP-3-deficient yeast cells, whereas the delivery of scNpc2p was affected by deficiencies in either AP-3 or GGA. We focused on the role of the AP-3 pathway in the targeting of the mammalian NP-C proteins. We found that, although mouse NPC1 (mNPC1) and hNPC2 co-localize with AP-3 to a similar extent in fibroblasts, hNPC2 preferentially co-localizes with AP-1. Importantly, the targeting of both mammalian NPC1 and NPC2 is dependent on AP-3. Moreover, and consistent with the NP-C proteins playing a role in cholesterol metabolism, AP-3-deficient cells have reduced levels of cholesterol. These results provide information about how the NP-C proteins are targeted to their sites of action and illustrate the possibility that defective sorting of the NP-C proteins along the endocytic route can alter cellular cholesterol.
Niemann-Pick disease Type C (NP-C) is a fatal autosomal recessive disorder characterized by the lysosomal accumulation of lipids. Specifically, exogenously derived cholesterol is aberrantly accrued in late endosomes and lysosomes (Liscum et al., 1989; Pentchev et al., 1985), and it is thought that this defect is due to an impairment in the retrograde movement of lipids from these organelles (Davies et al., 2000; Neufeld et al., 1999) to the plasma membrane or endoplasmic reticulum (ER) (Cruz et al., 2000). Although the two mutated genes responsible for the disorder have been identified as human NPC1 (hNPC1) and human NPC2 (hNPC2) (Carstea et al., 1997; Loftus et al., 1997; Naureckiene et al., 2000), the precise functions of the gene products remain unknown. hNPC1 is a highly conserved integral membrane protein of 1278 amino acids composed of 13 transmembrane-spanning domains, a conserved sterol-sensing domain and a C-terminal di-leucine motif, which has been proposed to be dispensable for proper function (Carstea et al., 1997; Davies and Ioannou, 2000; Scott et al., 2004). Examination of the subcellular distribution of hNPC1 reveals that the protein localizes to vesicles containing LAMPI (Garver et al., 2000; Higgins et al., 1999), LAMPII (Neufeld et al., 1999) and RAB9 (Higgins et al., 1999), indicating that the protein resides in late endosomes. The proper intracellular localization of hNPC1 seems to be important for its function, because a mutant form of the protein that accumulates in the ER is causative of NP-C disease (Blom et al., 2003).
hNPC2 is the second protein associated with NP-C disease (Naureckiene et al., 2000). hNPC2 is a secreted 151 amino acid protein containing an N-terminal ER signal peptide and an MD-2-related lipid-recognition domain (Inohara and Nunez, 2002). The protein is bound by the mannose-6-phosphate receptor (Naureckiene et al., 2000) and co-localizes with cathepsin D (Zhang et al., 2003), indicating that hNPC2 is located, in part, in the lysosome.
The mechanisms that target the NP-C proteins to the late endosome/lysosome have not been fully investigated; however, intrinsic signals and modifications found within the NP-C proteins suggest how these proteins could be targeted to their respective organelles. hNPC1 contains a C-terminal di-leucine motif (Carstea et al., 1997; Loftus et al., 1997; Scott et al., 2004), which – in other membrane proteins – is necessary and sufficient to confer endocytic targeting (Bonifacino and Traub, 2003; Johnson and Kornfeld, 1992a; Johnson and Kornfeld, 1992b; Letourneur and Klausner, 1992). However, in the case of NPC1, its C-terminal di-leucine motif is not sufficient for its late endosomal localization (Scott et al., 2004). As mentioned above, hNPC2 is modified by the addition of a mannose-6-phosphate group (Naureckiene et al., 2000), which is a signal for targeting soluble molecules to the lysosome when recognized by the mannose-6-phosphate receptor (Kornfeld and Mellman, 1989; Pohlmann et al., 1995). Consistent with this modification, the targeting of NPC2 is regulated by mannose-6-phosphate receptors (Willenborg et al., 2005).
For intracellular trafficking, different adaptor proteins (APs) recognize intrinsic targeting signals in proteins and sort these cargo molecules into vesicles for transport between organelles (Bresnahan et al., 1998; Dietrich et al., 1997; Fujita et al., 1999; Heilker et al., 1996; Hofmann et al., 1999; Honing et al., 1998; Peden et al., 2001). There are several types of adaptors, including the heterotetrameric AP complexes AP-1, AP-2, AP-3 and AP-4 as well as the monomeric Golgi-localized, γ ear-containing, ARF-binding (GGA) proteins (reviewed in Boehm and Bonifacino, 2002). The AP-1, AP-2, AP-3 and GGA proteins are, for the most part, functionally and structurally conserved in all eukaryotes, including in Saccharomyces cerevisiae (reviewed in Boehm and Bonifacino, 2002). In contrast to what has been observed in multicellular organisms, none of the components of these complexes are essential for cell viability in yeast (reviewed in Boehm and Bonifacino, 2002). The finding that both of the yeast homologs of the NP-C proteins are able to functionally replace the loss of their respective mammalian homologs (Berger et al., 2005b; Malathi et al., 2004), in conjunction with the conservation of adaptor function, makes yeast an ideal model system in which to examine the effects of the AP complexes on NP-C protein sorting. We can then use the studies from the simple yeast model system to guide our studies of mammalian NPC1 and NPC2 trafficking.
In this study we examine the mechanisms that control the proper sorting of mouse NPC1 (mNPC1) and hNPC2 to their target organelles. We first exploit the yeast model system to analyze the requirements for trafficking of the yeast NP-C homologs S. cerevisiae Ncr1p (scNcr1p) (Berger et al., 2005a; Malathi et al., 2004; Zhang et al., 2004) and scNpc2p (Berger et al., 2005b). Results of the survey of APs in yeast implicate the AP-3 complex in trafficking both scNcr1p and scNpc2p to the yeast vacuole. These findings in yeast were then used as a guide to examine the role that the AP-3 complex plays in the transport of the mNPC1 and hNPC2 proteins in mammalian cells. We describe the role that the AP-3 complex plays in the localization and function of the NP-C proteins and identify an endocytic compartment in which the NP-C proteins and regulatory sorting machinery reside together. This study indicates that a single sorting mechanism, mediated by AP-3, can regulate the targeting of proteins that contain topologically distinct sorting information by both direct and indirect mechanisms.
Adaptor protein complexes direct the proper targeting of yeast Ncr1p and Npc2p to the vacuole
As a first step towards understanding the mechanisms that govern mammalian NP-C protein localization, we took advantage of the genetic tractability of S. cerevisiae to analyze the machinery required to traffic the yeast NP-C proteins, scNcr1p and scNpc2p, to the vacuole (Berger et al., 2005b; Malathi et al., 2004). To investigate the transport pathways for the yeast NP-C homologs, a genetic analysis was performed by exploiting deletion mutants of the conserved, but non-essential, AP complexes GGA1 and GGA2, AP-1, AP-2 and AP-3 (Boehm and Bonifacino, 2002). We analyzed the effect that deletion of the individual members of each of the AP complexes had on the localization of both scNcr1p, which is normally localized to the vacuolar membrane (Berger et al., 2005a; Malathi et al., 2004; Zhang et al., 2004), and scNpc2p, which is normally localized to the vacuolar lumen (Berger et al., 2005b) (Fig. 1). As a control, we also examined the localization of alkaline phosphatase (ALP), a known AP-3 cargo molecule (Cowles et al., 1997; Stepp et al., 1997) that localizes to the yeast vacuolar membrane in wild-type cells (Fig. 1).
Individual deletion mutants of the yeast AP-1 components APL4, APL2, APM1, APM2 and APS1, homologous to γ, β1, μ1A, μ1B and σ1A of mammalian AP-1, respectively (Boehm and Bonifacino, 2002), were transformed with individual multicopy plasmids encoding the GFP fusion proteins scNcr1p-GFP or scNpc2p-GFP, or, as a control, GFP-ALP. The subcellular localization of each protein was assessed by direct fluorescence microscopy. Consistent with previous observations that individual deletions of the AP-1 complex members yield no overt phenotypes (Nakai et al., 1993; Phan et al., 1994; Rad et al., 1995; Stepp et al., 1995; Yeung et al., 1999), we found that, in the majority of mutants tested, the localization of scNcr1p, scNpc2p or ALP was not affected (supplementary material Fig. S1). AP-1-dependent protein mis-sorting has been identified when individual AP-1 mutants are combined with a clathrin heavy chain mutant; however, there is variability in the penetrance of these phenotypes depending on which subunit is studied (Valdivia et al., 2002; Yeung et al., 1999). We therefore examined the role of the GGA proteins GGA1 and GGA2 in the trafficking of the yeast NP-C proteins. GGA proteins mediate protein sorting from the Golgi in a way similar to AP-1 complex sorting from the Golgi (Boehm and Bonifacino, 2002). We analyzed the localization of scNcr1p-GFP, scNpc2p-GFP or control GFP-ALP by direct fluorescence microscopy in cells deficient for both GGA1 and GGA2 (gga1Δgga2Δ). Much like AP-1/clathrin heavy chain double-mutant cells, gga1Δgga2Δ cells exhibit defects in carboxypeptidase Y sorting (Hirst et al., 2000; Zhdankina et al., 2001). Our results indicate that neither scNcr1p nor our control ALP was mislocalized (Fig. 2). However, scNpc2p was localized to the lumen of the vacuole and also accumulated in small punctate structures that were absent in wild-type cells. These structures were observed in close proximity to the plasma membrane as well as surrounding the vacuole (Fig. 2). These results suggest that scNpc2p is trafficked through prevacuolar compartments prior to reaching the vacuole, similar to what is observed for other proteins dependent upon AP-1 or the GGA proteins in yeast (Boehm and Bonifacino, 2002). Furthermore, these results highlight the conservation between the NPC2 orthologs, because both are targeted by AP-1/GGA-controlled pathways in both yeast (this work) and mammals (Willenborg et al., 2005).
The role of the AP-2 complex was assessed by examining the localization of scNcr1p-GFP, scNpc2p-GFP or GFP-ALP in each of the following strains: apl3Δ, apl1Δ, apm4Δ or aps2Δ. These yeast mutants contain a deletion of the gene encoding the homolog of α, β2, μ2 or σ2 of mammalian AP-2, respectively (Boehm and Bonifacino, 2002). These cells were examined by direct fluorescence microscopy for localization of each GFP fusion protein. Results from this analysis indicate that single deletions of the AP-2 complex members have no effect on the localization of scNcr1p or ALP to the vacuolar membrane, nor does it affect the targeting of scNpc2p to the vacuolar lumen (Fig. 3 and supplementary material Fig. S2). These results suggest that the AP-2 complex in yeast is not required for localization of the NP-C proteins.
AP-3 complex involvement in NP-C protein trafficking was assessed by examining the localization of scNcr1p-GFP, scNpc2p-GFP or GFP-ALP in apl5Δ, apl6Δ, apm3Δ or aps3Δ cells, deletion mutants of genes that encode homologs of δ, β3A, μ3A and σ3A of mammalian AP-3, respectively (Boehm and Bonifacino, 2002). The subcellular localization of the GFP fusion proteins was assessed by direct fluorescence microscopy. The AP-3 complex traffics ALP in yeast (Cowles et al., 1997; Piper et al., 1997; Stepp et al., 1997), as evidenced by GFP-ALP accumulation in punctate structures throughout the cell in addition to the vacuolar membrane in all AP-3 pathway mutants (Fig. 4). When the localization of scNcr1p and scNpc2p was assessed, these proteins were also mislocalized in the majority of the AP-3 mutant cells (Fig. 4). Notably, scNpc2p, a soluble protein, accumulated in both punctate and tubular structures surrounding the vacuole in all AP-3 mutants tested (Fig. 4). scNcr1p accumulated in punctate structures, similar to what was observed for ALP, in all AP-3 mutants except for apm3Δ. The finding that scNcr1p localizes correctly in the apm3Δ mutant background is consistent with previously published work in which apm3Δ was the only AP-3 mutant examined (Zhang et al., 2004).
Based on the vacuolar localization of scNcr1p in an apm3Δ mutant background, Zhang et al. concluded that scNcr1p trafficked through the vacuolar protein sorting (VPS) pathway and not the ALP pathway (Zhang et al., 2004). However, because scNcr1p and scNpc2p mislocalize in all other AP-3 mutants (Fig. 4), we investigated whether scNcr1p and scNpc2p can traffic through either pathway to the vacuole. The VPS pathway was assessed by examining the localization of both scNcr1p and scNpc2p in a vps4Δ mutant background (Zahn et al., 2001). Results from this analysis showed no discernible mislocalization of scNcr1p from the vacuolar membrane or accumulation in enlarged multilamellar pre-vacuolar compartments typical of this class (E) of VPS mutants (Raymond et al., 1992) (supplementary material Fig. S3). By contrast, scNpc2p accumulated in these compartments while still remaining luminal, suggesting that scNpc2p can traffic through the VPS pathway as well as the AP-3 pathway (supplementary material Fig. S3).
Taken together, our analysis of NP-C protein trafficking in yeast supports a model in which scNcr1p traffics to the vacuole in an AP-3-dependent manner, whereas proper scNpc2p localization depends on both the AP-1/GGA/VPS and AP-3 pathways.
Mammalian NPC1 is targeted by the AP-3 complex
Based upon the finding that yeast Ncr1p, which can functionally replace mammalian NPC1 (Malathi et al., 2004), is trafficked to the vacuole via an AP-3-mediated mechanism, we hypothesized that the mammalian NPC1 protein is also trafficked in an AP-3-dependent manner. To test this hypothesis, we first examined whether mNPC1 and the AP-3 complex co-localize. Because we lack an antibody capable of detecting endogenous mNPC1 by indirect immunofluorescence, we assessed the localization of mNPC1-GFP expressed from an adenovirus construct capable of rescuing Npc1–/– phenotypes (Paul et al., 2005). Furthermore, and as shown below (see Fig. 5 and Fig. 6A), mNPC1-GFP expressed from an adenoviral vector recapitulated the known localization of NPC1 to LAMPI-positive late endosomal compartments (Paul et al., 2005). Adenovirally delivered mNPC1-GFP was co-localized with AP-3 by infecting immortalized AP-3 δ–/– mocha mouse fibroblasts transduced with a retrovirus encoding the δ subunit of AP-3 (AP-3+/+ cells) or an empty retrovirus (AP-3–/– cells) as a control. The localization of both mNPC1 and the δ subunit of AP-3 was assessed in the AP-3+/+ cells by indirect immunofluorescence confocal microscopy. Results from this analysis show that ∼20% of all mNPC1-GFP-positive puncta co-localize with AP-3 and indicate that this co-localization is specific because mNPC1-GFP co-localization with the adaptor complex AP-1 is 3.8-fold lower than with AP-3 (Fig. 5 and Fig. 6A). Furthermore, the absence of AP-3 did not increase the extent of co-localization between AP-1 and mNPC1 (Fig. 5 and Fig. 6A). As a control, we detected no AP-3 staining in the AP-3–/– cells (Fig. 5). Collectively, these findings indicate that mNPC1 and AP-3 reside in a subset of the same organelles.
To analyze the effect that loss of the AP-3 complex has on protein localization, mNPC1 was co-localized with markers of various organelles in the endocytic pathway. Early/recycling endosomes were revealed with antibodies directed against the transferrin receptor (TfR) (supplementary material Fig. S4) (Dunn et al., 1989); the transition between early and late endosomes was identified using antibodies against the SNAREs Vti1b and syntaxin 8 (Fig. 5 and supplementary material Fig. S4, Syn8) (Atlashkin et al., 2003; Hirst et al., 2004; Kreykenbohm et al., 2002; Prekeris et al., 1999; Pryor et al., 2004); and late endosomes/lysosomes were labeled using antibodies against LAMPI (Fig. 5) (Lewis et al., 1985), which traffics in an AP-3-dependent manner (Dell'Angelica et al., 1999; Yang et al., 2000). Immunolocalization of mNPC1-GFP with the AP-3 cargo molecule LAMPI in AP-3+/+ and AP-3–/– cells indicates that, consistent with previous studies (Garver et al., 2000; Higgins et al., 1999; Neufeld et al., 1999), there is almost total (∼80%) co-localization between these proteins, irrespective of the presence or absence of AP-3 (Fig. 5 and Fig. 6A). mNPC1 localization to LAMPI-positive late endosomal/lysosomal compartments is selective, because only a tenth of mNPC1-GFP-positive puncta co-localized with TfR-positive early/recycling endosomes (Fig. 6A and supplementary material Fig. S4).
AP-3 budding profiles containing LAMPI are present in domains of early endosomes (Peden et al., 2004). Therefore, we hypothesized that in the absence of AP-3, mNPC1 could accumulate in endocytic compartments upstream of LAMPI-positive late endosomes. To explore this hypothesis, we analyzed the co-localization of mNPC1-GFP with TfR and the endosomal SNAREs Vti1b and syntaxin 8 in AP-3–/– cells, and, as a control, AP-3+/+ cells (Fig. 5 and Fig. 6A,B and supplementary material Fig. S4). Results demonstrated that the co-localization between mNPC1 and TfR was not affected by the absence of AP-3. However, the co-localization of mNPC1 with the early/late endocytic SNAREs Vti1b and syntaxin 8 increased two- and three-fold, respectively, in AP-3–/– cells (Fig. 5 and Fig. 6A,B and supplementary material Fig. S4). Similar findings were obtained in freshly isolated primary skin fibroblasts from two different AP-3-deficient mice, grizzled and mocha (Peden et al., 2002), respectively (Fig. 6B). Importantly, the same phenotype was also observed when the AP-3 cargo molecule LAMPI was co-localized with Vti1b and syntaxin 8 (Fig. 6C and supplementary material Fig. S5). These observations indicate that, at steady state, mNPC1 preferentially localized with the adaptor complex AP-3 and its cargoes, such as LAMPI. Furthermore, our findings suggest that, much like LAMPI, the subcellular distribution of mNPC1 in Vti1b–syntaxin-8-positive endosomes is affected by the lack of AP-3.
Previous studies have shown that AP-3 deficiency leads to increased targeting of AP-3 cargoes to the plasma membrane and this enrichment has been widely used as a reliable assay for AP-3 function (Dell'Angelica et al., 1999; Di Pietro et al., 2006; Janvier and Bonifacino, 2005; Peden et al., 2004; Salazar et al., 2006; Styers et al., 2004). To determine whether mNPC1 also has a similar AP-3-dependent increase in plasma membrane distribution, surface levels of endogenous mNPC1 were examined by biotinylation of AP-3–/– and AP-3+/+ cells followed by streptavidin-precipitation of biotin-conjugated proteins. Precipitates were resolved by SDS-PAGE and analyzed by immunoblotting using an antibody directed against NPC1 to detect endogenous mNPC1. As controls, we also examined the surface levels of endogenous LAMPI, an AP-3-dependent cargo molecule, as well as that of the TfR and caveolin 1 (Cav1), neither of which is transported in an AP-3-dependent manner (Dell'Angelica et al., 1999). Furthermore, Cav1 and β actin were used to examine the behavior of proteins that, although not containing extracellular epitopes, could co-precipitate with biotinylated transmembrane proteins. Results indicate that, as expected, there is an increase in LAMPI distribution to the plasma membrane in AP-3–/– cells as compared with AP-3+/+ cells, with no difference detectable for TfR or Cav1 (Fig. 7A). When the distribution of mNPC1 was examined, as was discovered with LAMPI, there was an increase in the surface level of mNPC1 (Fig. 7A). Quantification of Cav1 normalized surface content indicates an even greater increase for mNPC1 (sevenfold) than that observed for LAMPI (fivefold; Fig. 7B). This finding indicates that mNPC1 possesses a similar targeting behavior to that described for known AP-3 cargoes, such as LAMPI, which suggests that mNPC1 is trafficked to the late endosome by the AP-3 complex, as was suggested by results with yeast Ncr1p.
AP-3 deficiency leads to decreased detection of unesterified cholesterol
To determine whether cholesterol levels can be regulated in an AP-3-dependent manner, we examined the cellular distribution and content of cholesterol in AP-3+/+ and AP-3–/– cells using filipin, a fluorescent marker for unesterified cholesterol (Demel and Van, 1965), and Amplex Red, a cholesterol oxidase-based assay to quantitatively measure total cholesterol levels (Amundson and Zhou, 1999). Cholesterol distribution was qualitatively analyzed by epifluorescence and deconvolution microscopy. A comparison of the localization pattern of cholesterol between AP-3–/– and AP-3+/+ cells indicates that there are no appreciable qualitative differences in the subcellular compartmentalization of unesterified cholesterol (Fig. 8A). By contrast, quantitative assessment of total unesterified cholesterol by filipin flow cytometry demonstrated that filipin staining of AP-3–/– cells was diminished by ∼50% as compared with AP-3+/+ cells (Fig. 8B,C). Similarly, total cellular cholesterol was reduced by 30% (n=11, P<0.002; Fig. 8D) as determined using Amplex Red. Taken together with the increase in plasma membrane mNPC1 in AP-3–/– cells, this decrease in cellular cholesterol suggests that the AP-3 complex regulates the function of cholesterol transport proteins, such as NPC1, by affecting their spatial distribution within the cell.
AP-3 deficiency affects the fate of mammalian NPC2
Our analysis of the yeast Npc2p protein reveals that its localization to the vacuolar lumen is dependent on AP-3 adaptor complex and AP-1/GGA sorting machinery (see Figs 2 and 4). To test the hypothesis that the AP-3 complex facilitates the trafficking of the non-membrane-associated hNPC2 protein in mammalian cells, we examined whether there was co-localization between the AP-3 complex and hNPC2. We assessed the co-localization of the functional fusion protein hNPC2-GFP, expressed from an adenovirus construct (Berger et al., 2005b), with AP-3 in AP-3+/+ cells or, as a control, AP-3–/– cells. As an additional control, we co-localized hNPC2 with the adaptor complex AP-1, an adaptor that binds to the cytosolic domains of the mannose-6-phosphate receptors that preferentially deliver NPC2 from the Golgi complex to lysosomal organelles (Willenborg et al., 2005). Protein distribution was analyzed by indirect immunofluorescence confocal microscopy. Results from this analysis show that half of all hNPC2 puncta were positive for AP-1 (Fig. 6D) and that these AP-1-positive puncta were concentrated around the perinuclear area, consistent with a Golgi-endosome distribution (Fig. 9). The distribution and extent of co-localization between AP-1 and hNPC2 was not affected by the absence of AP-3 (Fig. 6D and Fig. 9). As predicted from our yeast data, hNPC2 also co-localized with AP-3-positive organelles (Fig. 6D and Fig. 9), although to a lesser extent than that observed for mNPC1 (Fig. 6A,D and see Fig. 5). No AP-3 fluorescence was detected in the control AP-3–/– cells (Fig. 9).
We, and others, have previously co-localized hNPC2 with markers for lysosomal and Golgi structures (Berger et al., 2005b; Blom et al., 2003). To test whether loss of the AP-3 complex alters the steady-state localization of hNPC2, we co-localized hNPC2-GFP with AP-1 γ, an adaptor found in Golgi and endosomes, and the endosomal markers TfR, Vti1b, syntaxin 8 and LAMPI in either AP-3+/+ or AP-3–/– cells. Consistent with our previous reports (Berger et al., 2005b), co-localization of hNPC2 with LAMPI and TfR was restricted to a few discrete puncta that represent ∼10% and ∼20% of all hNPC2-positive structures, respectively (Fig. 6D and Fig. 9; supplementary material Fig. S6). In addition, immunolocalization of hNPC2 indicated that there was no detectable effect of the loss of AP-3 on the steady-state distribution of hNPC2 in AP-3–/– cells compared with AP-3+/+ cells (Fig. 6D). hNPC2 co-localized predominantly (∼50%) with the endosome markers Vti1b and syntaxin 8 (Fig. 6D and Fig. 9; supplementary material Fig. S6). However, and in contrast to mNPC1, the subcellular distribution of hNPC2 in these Vti1b- and syntaxin 8-positive compartments was minimally affected by the absence of AP-3 (Fig. 6D and Fig. 9; supplementary material Fig. S6), consistent with the preferential localization of hNPC2 with AP-1.
Because hNPC2 is a secreted protein, we hypothesized that, if the AP-3 complex was involved in transporting this protein, there might be an alteration in the amount of hNPC2 detected in the media similar to the changes in mNPC1 surface levels observed in AP-3–/– cells. To test this hypothesis, we infected equal numbers of AP-3+/+ and AP-3–/– cells with adenovirus encoding hNPC2-GFP or GFP alone. The infected cells were grown for 72 hours and the media was collected in order to assess the amount of hNPC2-GFP present. Cell lysates were prepared as controls for cellular hNPC2 protein levels from each sample. Immunoblot analysis revealed the expected 47 kDa hNPC2-GFP detectable in the culture media from AP-3+/+ cells with no detectable changes in hNPC2-GFP levels in cell lysates from wild-type and AP-3-deficient cells (Fig. 10). However, the amount of hNPC2 detected in the media from AP-3–/– cells is greatly diminished compared with media from AP-3+/+ cells (Fig. 10). As a control, an equal amount of the residual GFP was detected in the media from either cell type. These results support the hypothesis that, in addition to the AP-1/GGA sorting machinery (see Fig. 2) (Willenborg et al., 2005), the subcellular fate of mammalian NPC2 is also regulated by the AP-3 complex, probably by an indirect mechanism.
Mammalian NPC1 and NPC2 are hypothesized to function in retrograde lipid transport from late endosomes and lysosomes (Liscum, 2000), suggesting that the localization of NPC1 and NPC2 to these organelles is of primary importance to their function. Unraveling the mechanisms that regulate NP-C protein transport is therefore crucially important to understanding the pathology of the disease. Furthermore, these trafficking pathways might offer new therapeutic targets.
We have exploited yeast cell biology to assess the routes by which the two NP-C homologs, scNcr1p and scNpc2p, traffic to the vacuole. Our analysis indicates that there is a requirement for the AP-1/GGA pathway in trafficking scNpc2p to the vacuolar lumen. Interestingly, there is mis-targeting of scNpc2p to punctate vesicles throughout the cell as well as to tubular structures that surround the vacuole in gga1Δgga2Δ mutants. This localization to structures outside of the vacuole suggests that scNpc2p might normally traffic through intermediary compartments, such as the prevacuolar compartment, which become easily detectable when the GGA proteins are deficient. The probability that scNpc2p traffics through intermediary compartments is strengthened by our finding that scNpc2p accumulates in the lumen of class E compartments, which accumulate in a vps4Δ background (Zahn et al., 2001). These results are consistent with the finding that hNPC2 binds to and is dependent upon the mannose-6-phosphate receptor for lysosomal targeting (Naureckiene et al., 2000; Willenborg et al., 2005) and with our observation that hNPC2 extensively co-localizes with AP-1.
In contrast to the GGA proteins, which seem to only affect scNpc2p localization, our findings indicate that the AP-3 pathway is required for the proper transport of both NP-C proteins. We found that, as seen for ALP (a protein trafficked specifically by the AP-3 complex in yeast) (Cowles et al., 1997; Piper et al., 1997; Stepp et al., 1997), scNcr1p and scNpc2p are also dependent on the AP-3 complex for proper trafficking. When each individual deletion of the AP-3 complex was evaluated for its effects upon scNcr1p localization, except for apm3Δ, there was a clear mislocalization of the protein to cytoplasmic punctate structures in addition to the normal vacuolar membrane localization in the majority of mutants, similar to what was seen for ALP (see Fig. 4).
scNcr1p is found throughout the cytoplasm in punctate structures in all AP-3 mutants tested, except for apm3Δ (see Fig. 4), consistent with previous work in which scNcr1p localization was examined only in this single AP-3 mutant background (Zhang et al., 2004). The finding that this single deletion mutant does not alter the scNcr1p trafficking pattern suggests that the μ subunit of the yeast AP-3 complex might be dispensable for the transport of specific cargo molecules, because the localization of both ALP and scNpc2p is altered in the apm3Δ mutant (see Fig. 4). The observation that a deficiency in the yeast Apm3p subunit does not manifest evident scNcr1p phenotypes is not unprecedented. Deficiencies in the Caenorhabditis elegans AP-2 μ2 (Boehm and Bonifacino, 2002; Grant and Hirsh, 1999) and AP-3 μ3 (Shim and Lee, 2005) subunits do not fully reproduce the extent and quality of the phenotypes observed with deficiencies in any of the other three AP-2 and AP-3 adaptor subunits. Although the mechanism for the phenotypic differences among subunit deficiencies remains unknown, a possible explanation is the presence of partial adaptor complexes constituted by just two subunits, which has been described for mouse models deficient in the AP-3 β3 chain (Peden et al., 2002; Yang et al., 2000). In this regard, recent evidence indicates that, in some cases, yeast AP-3 complexes assembled with mutant Apm3p subunits are unable to support the targeting of membrane proteins containing tyrosine-based motifs to the same extent as observed for apm3Δ cells, in which the Apm3p subunit is absent altogether (Wen et al., 2006). However, these same mutant complexes show no defect in the targeting of an AP-3-trafficking-dependent protein containing a di-leucine-based motif to the vacuolar membrane (Wen et al., 2006), suggesting that deficiencies in the Apm3p subunit might lead to targeting defects in only a subset of AP-3 cargoes. Alternatively, the differential effect on scNcr1p localization might be due to differences in the recognition of the protein, because the endosomal sorting signal for scNcr1p has not been defined; although, the mammalian NPC1 protein contains a di-leucine motif at its C-terminus (Watari et al., 1999) and separate sorting information in the sterol sensing domain (Scott et al., 2004). Further analysis will be necessary to determine why the loss of APM3 does not alter the trafficking pattern of scNcr1p.
Zhang et al. reported that scNcr1p did not traffic through the AP-3 pathway, based on their analysis of apm3Δ cells, but rather through intermediary compartments that are a part of the VPS pathway (Zhang et al., 2004). When we examined whether scNcr1p could transit through the VPS pathway, we found that, unlike scNpc2p localization, there was no detectable defect in the localization of scNcr1p. In contrast to Zhang et al., our analysis of the VPS route was restricted to studies just in a vps4Δ background (supplementary material Fig. S3). Therefore, a distinctive possibility is that mutations in the VPS pathway as well as in the AP-3 route could both affect the localization of scNcr1p but to different degrees depending on the specific nature of the mutations, the genetic loci affected or the assays used in our studies and those by Zhang et al. Thus, much like scNpc2p, the targeting of scNcr1p could be modulated by other targeting mechanisms in addition to the AP-3 route.
The finding that the AP-3 complex facilitates the localization of the NP-C proteins in yeast directed us to examine whether the trafficking of the mammalian NP-C proteins was also AP-3-dependent in mammalian cells. Previous work has demonstrated that a deficiency in the AP-3 complex leads to an accumulation of AP-3-dependent cargo molecules at the cell surface without affecting their endosomal/lysosomal steady-state localization (Dell'Angelica et al., 1999). Indeed, our results indicate that, as seen for the AP-3-transported LAMPI protein (see Fig. 7) (Dell'Angelica et al., 1999), mNPC1 is mis-targeted to the plasma membrane in AP-3-deficient mouse fibroblasts to an even greater extent than LAMPI. Concomitant with mNPC1 redistribution to the plasma membrane, mNPC1 co-localization with the early/late endosome markers Vti1b and syntaxin 8 increases, suggesting that mNPC1 might be delayed in these compartments. Similar results are seen for LAMPI. Importantly, both Vti1b and syntaxin 8 levels are not affected by a deficiency in AP-3 (Salazar et al., 2006), with Vti1b transport from endosomes being dependent on EpsinR-AP-1 mechanisms (Hirst et al., 2004). These results are consistent with the notion that mNPC1, like scNcr1p, is an AP-3 cargo molecule. This finding is very interesting in light of our evidence showing that AP-3–/– cells have less detectable unesterified and total cholesterol than AP-3+/+ fibroblasts, with no obvious effect on the steady-state cholesterol distribution (see Fig. 8). The precise mechanism leading to decreased cholesterol content in AP-3-deficient cells remains unclear, but it is reasonable to hypothesize that it is related to the redistribution of NPC1, NPC2 and/or other late endosomal proteins that are mis-targeted in the absence of AP-3. Of particular interest is the observation that genetic ablation of mouse LAMPI and LAMPII, which are both AP-3 cargoes (Bonifacino and Traub, 2003), alters the cellular content of cholesterol detected by filipin staining (Eskelinen et al., 2004). Alternatively, other proteins, such as MLN64, a membrane protein targeted to late endosomes that is predicted to bind cholesterol, based upon structural studies (Alpy et al., 2001; Murcia et al., 2006), could contribute to the phenotypes observed in AP-3–/– cells. Irrespective of the precise mechanism that leads to reduced cellular cholesterol levels, targeted perturbation of AP-3 complex function might help ameliorate cholesterol accumulation defects seen in a number of disorders, including in NP-C.
We also examined AP-3-dependent trafficking of hNPC2. Based upon our results showing increased trafficking of mNPC1 to the plasma membrane, we hypothesized that, if the transport of hNPC2 was AP-3-dependent, we would detect an increase in hNPC2 in media. Surprisingly, although our analysis indicated that hNPC2 trafficking is affected by AP-3, we found that our ability to detect hNPC2 in the media of AP-3-deficient cells was greatly diminished as compared to media from AP-3+/+ cells. How can the AP-3 pathway affect a mannose-6-phosphate-modified luminal protein? It is speculative at this point, but it is attractive to consider the possibility that AP-3 plays an indirect role in the transport of hNPC2, with a membrane-bound protein, such as NPC1, or another AP-3-interacting membrane protein, binding the small soluble protein. Alternatively, AP-3 might regulate the targeting of a factor that does not bind hNPC2, yet is required for hNPC2 targeting via an AP-3-independent route; such a factor could be, for example, a vesicular SNARE protein that directly binds AP-3 (Martinez-Arca et al., 2003), such as VAMP7 (also known as SYBL1). The trafficking and expression of this vesicular SNARE has recently been shown to be affected by AP-3 deficiencies (Salazar et al., 2006). This type of mechanism has been postulated to explain the phenotypes observed with carboxypeptidase Y, a mannose-6-phosphate-modified luminal protein, in AP-3-deficient yeast cells (Bonangelino et al., 2002).
Recent studies show that the lysosomal targeting of NPC2 is dependent upon the presence of the mannose-6-phosphate receptors in fibroblasts (Willenborg et al., 2005). This observation is consistent with our findings in yeast and mammalian cells. Yeast Npc2p is trafficked in a GGA-dependent as well as an AP-3-dependent manner (see Figs 2 and 4). Similarly, hNPC2 predominantly co-localizes with the adaptor complex AP-1 and to a lesser extent with AP-3 in mouse fibroblasts. These observations support a model in which AP-3 might play an indirect role in the targeting of the soluble hNPC2 protein, possibly by increasing the surface display of a hNPC2-binding partner that allows for greater uptake of the free protein from the media. However, we cannot rule out the possibility that the loss of hNPC2 detection could occur by a change in the secretion of the protein. Whether the effect of AP-3 deficiency on hNPC2 detection in the media is due to enhanced uptake into endosomes or the inability of hNPC2 to be transported into an appropriate compartment for efflux from the cell remains to be explored. Irrespective of the mechanism, the factors that can most readily influence the function and targeting of NPC2 are those that reside with it in a common compartment. It is of interest to point out that hNPC2 is present in Vti1b–syntaxin-8-positive endosomes, a localization that, in the case of mNPC1, becomes evident in cells that lack functional AP-3 complexes. Further analyses will be necessary to delineate the exact role that AP-3 plays in the transport of hNPC2.
Our studies indicate that the AP-3 complex facilitates the trafficking of the NP-C proteins. These findings indicate that the targeting of the NP-C proteins is controlled by multiple sorting machineries (AP-1/GGA/VPS and AP-3). In addition, the finding that hNPC2 transport is AP-3-dependent demonstrates that the sorting of a non-membrane-bound protein might be controlled indirectly by the AP-3 complex. This complex and the proteins that are transported by it (Dell'Angelica et al., 1998; Dell'Angelica et al., 1999; Faundez and Kelly, 2000; Honing et al., 1998; Robinson, 2004; Salazar et al., 2005; Salazar et al., 2004a) could provide new insight into the function of NPC1 and NPC2, and could provide novel therapeutic targets for the treatment of NP-C disease.
Materials and Methods
Strains, plasmids and cell culture
All DNA manipulations were performed according to standard protocols (Sambrook and Russell, 2001) and all yeast media was prepared by standard methods (Adams et al., 1997). All yeast strains and plasmids are described in Table 1. Chemicals were obtained from Fisher Scientific (Pittsburgh, PA), Sigma Chemical (St Louis, MO), Stratagene (La Jolla, CA) or USBiological (Swampscott, MA) unless otherwise noted.
HEK293 cells were grown in DMEM (4.5 g/l glucose) (Cellgro, Herndon, VA) supplemented with 10% FBS (Hyclone, Logan, UT), 100 units/ml penicillin and 100 μg/ml streptomycin (Cellgro). Immortalized AP-3 δ–/– mocha mouse fibroblasts transduced with a retrovirus encoding the δ subunit of AP-3 (AP-3+/+ cells) or with an empty retrovirus (AP-3–/– cells) were grown in DMEM supplemented with 10% FBS, 100 units/ml penicillin, 100 μg/ml streptomycin and 200 μg/ml hygromycin B (Peden et al., 2004).
Primary antibodies used in this study were: anti-mouse LAMPI (1D4B) used at 1:1000 and monoclonal anti-AP-3 δ (SA4) ascites used at 1:500 (Developmental Studies Hybridoma Bank, University of Iowa); mouse monoclonal anti-TfR used at 1:1000 and anti-NPC1 used at 1:500 (Zymed Laboratories, San Francisco, CA); anti-α-tubulin used at 1:1000 (DM1A) and anti-β actin (actin) used at 1:5000 (Sigma); rabbit anti-GFP used at 1:5000 (Synaptic Systems, Goettingen, Germany); anti-mouse Caveolin1 (Cav1) used at 1:2000 and mouse monoclonal anti-AP1 γ used at 1:500 (BD Biosciences Transduction Labs); mouse monoclonal anti-syntaxin 8 used at 1:500 and mouse monoclonal anti-Vti1b used at 1:500 (BD Bioscience Pharmigen). Secondary antibodies used were: Alexa-Fluor-568-conjugated goat anti-mouse used at a 1:1000 dilution (Molecular Probes, Invitrogen Detection Technologies, Carlsbad, CA), and goat anti-rabbit HRP and goat anti-mouse HRP used at 1:7000 (Zymed Laboratories, Invitrogen Immunodetection, Carlsbad, CA).
An adenovirus encoding hNPC2-GFP was created as previously described (Berger et al., 2005b). Briefly, hNPC2 was amplified by PCR from human peripheral blood mononuclear cell cDNA (kind gift of Silvija I. Staprans, Emory University, Atlanta, GA). hNPC2 was first cloned into the pEGFP vector (Clontech, Palo Alto, CA) to create an in-frame C-terminal GFP fusion. PCR was used to amplify the entire cDNA fused to GFP followed by cloning into the AdEasy Adenoviral Vector system (Stratagene). In short, hNPC2-GFP was cloned NotI-HindIII into the pShuttle-CMV vector followed by homologous recombination into the pAdEasy-1 adenoviral backbone. The resulting recombinant DNA was transfected into HEK293 cells using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) for plaque formation. Virus was amplified in HEK293 cells and purified using the BD Adeno-X Virus Purification Kit (Clontech) as directed.
Unesterified cholesterol levels for mocha (AP-3–/–) and rescued (AP-3+/+) cells (Peden et al., 2004) were quantified by flow cytometry. Briefly, cells were placed in suspension using PBS with 5 mM EDTA followed by fixation in 4% paraformaldehyde in PBS at 4°C for 20 minutes. After fixation, the paraformaldehyde was quenched by washing the cells with PBS supplemented with 25 mM glycine. Cells were incubated with 50 μg/ml of filipin for 2 hours at 37°C to label unesterified cholesterol. Excess filipin was removed by washing the cells with PBS three times for 5 minutes each. Unesterified cholesterol levels were quantified using a MoFlo High Performance Cell Sorter (DakoCytomation, Fort Collins, CO). Results were analyzed using FlowJo software version 4.4.4 (Tree Star, Ashland, OR).
Total cellular cholesterol was determined using an Amplex Red Cholesterol Assay Kit (Molecular Probes, A12216) and following manufacturer's instructions. Briefly, cells that had been seeded in six-well plates were incubated with or without methyl-β-cyclodextrin (MβCD) at 10 mg/ml for 1 hour at 37°C in DMEM media to partially deplete cholesterol. Incubating cells for longer periods of time with MβCD compromised cell viability. Cells were washed twice in ice-cold PBS and lysed in 100 μl of buffer A (150 mM NaCl, 10 mM HEPES, 1 mM EGTA and 0.1 mM MgCl2, pH 7.4) containing 0.5% Triton X-100 plus Complete anti-protease mixture. Both free cholesterol and cholesterol esters were analyzed in 5 μl of whole-cell extract. Fluorescence was measured with a Synergy HT (Biotek, Winooski, VT) microplate reader using an excitation of 540 nm and a fluorescence detection of 575 nm. Results were expressed as μg cholesterol/μg protein and then calculated as the percentage of control. All determinations were performed at least in triplicate in three independent experiments (n=11).
Direct fluorescence microscopy
scNcr1p and scNpc2p were localized in living yeast cells as the C-terminal GFP fusion proteins, scNcr1p-GFP (pAC1184) and scNpc2p-GFP (pAC1185), as previously described (Berger et al., 2005a; Berger et al., 2005b). Alkaline phosphatase (ALP) was localized in living yeast cells as the N-terminal GFP fusion protein, GFP-ALP (Cowles et al., 1997). Wild-type yeast cells (ACY402), gga1Δgga2Δ yeast cells (ACY969), deletions of AP-1 complex members (apl4Δ, apl2Δ, apm1Δ, apm2Δ and aps1Δ), deletions of AP-2 complex members (apl3Δ, apl1Δ, apm4Δ and aps2Δ) or deletions of AP-3 complex members (apl5Δ, apl6Δ, apm3Δ and aps3Δ) were transformed with scNCR1-GFP, scNPC2-GFP or GFP-ALP plasmids and viewed by direct fluorescence microscopy using an Olympus BX60 epifluorescence microscope equipped with an Olympus UPlan Apo 100×/1.35 oil iris DIC objective, a GFP optimized barrier filter and a Photometrics Quantix digital camera. All images were collected using IPlab Spectrum software.
Immunofluorescence of cultured cells was performed as previously described (Faundez et al., 1997; Wei et al., 1998). Briefly, cells were placed on ice and fixed in 4% paraformaldehyde in PBS for 20 minutes. Following fixation, cells were quenched by washing twice with PBS containing 25 mM glycine and then once in PBS. Cells were permeabilized by incubating with block containing 15% horse serum, 0.02% saponin in PBS, 2% BSA and 1% fish skin gelatin for 1 hour at room temperature. Primary antibody incubation in block was performed for 1 hour at 37°C. Cells were washed three times with block and incubated with secondary antibodies diluted in block for 1 hour at room temperature. Cells were washed three times with block, once with PBS and were then mounted in gelvatol. Cells were visualized by confocal microscopy using a Zeiss Axiovert 100 M microscope coupled to HeNe1 and argon ion lasers. All images were viewed and acquired using a Plan Apochromat 63×/1.4 oil DIC objective and Zeiss LSM 510 sp1 software. Co-localization analyses were performed using Metamorph software (Universal Imaging, Downingtown, PA) and obtained from at least ten images collected from two coverslips per condition and per independent experiment. All determinations were performed in three independent experiments.
Immunoflorescence was performed as described (Salazar et al., 2004b). Images were acquired with a scientific-grade cooled charge-coupled device (Cool-Snap HQ with ORCA-ER chip) on a multi-wavelength, wide-field, three-dimensional microscopy system (Intelligent Imaging Innovations, Denver, CO), based on a 200M inverted microscope using a 63× numerical aperture 1.4 lens (Carl Zeiss, Thornwood, NY). Immunofluorescent samples were imaged at room temperature using a Sedat filter set (Chroma Technology, Rockingham, UT), in successive 0.25 μm focal planes. Out-of-focus light was removed with a constrained iterative deconvolution algorithm (Swedlow et al., 1997).
AP-3–/– and AP-3+/+ cells were grown to confluence and biotinylated as a monolayer (Salazar and Gonzalez, 2002). Cells were lifted off plates at 4°C with PBS-10 mM EDTA and sedimented at 800 g for 5 minutes. Cell pellets were resuspended in 5% SDS, 0.15 M Tris-HCl pH 6.7, 30% glycerol diluted 1:3 (vol:vol) with RIPA buffer (25 mM Tris-HCl pH 8.2, 50 mM NaCl, 0.5% NP-40, 0.5% DOC, 0.1% SDS, 0.1% azide) plus Complete anti-protease mixture. After a brief sonication (2 times 5 seconds each) samples were diluted 1:10 with PBS containing 0.5% NP-40. Cell debris was removed by sedimentation at 15,000 g for 20 minutes. A sample of 500 μg of each supernatant was precipitated with streptavidin-conjugated Sepharose beads over 5 hours at 4°C. After extensive washes with PBS containing 0.5% NP-40, bead-bound material was eluted with SDS-sample buffer, and proteins were resolved by SDS-PAGE and analyzed by immunoblot. Immunoblots were probed with an anti-NPC1 antibody, an anti-LAMPI antibody, an anti-TfR antibody and an anti-Cav1 antibody, or, as a control, an anti-β actin antibody. Immunoreactive bands were visualized by chemiluminescence.
Media detection assays
Detection of hNPC2 in media was assessed by infecting 2×105 fibroblasts with adenovirus encoding either human NPC2-GFP or GFP alone. After 7 hours, the cells were washed with DMEM and then grown in 800 μl of fresh media. After 72 hours, the media was collected and cell lysates were prepared by washing cells twice with PBS followed by lysis. 65 μl of media and 30 μg of lysate (5% of total) were used for immunoblot analysis. Proteins were resolved by SDS-PAGE and transferred to a PVDF membrane. Immunoblotting was performed by standard methods (Towbin et al., 1979). Blots were probed with an anti-GFP rabbit polyclonal antibody, an anti-AP-3 δ antibody or an anti-tubulin antibody. Immunoreactive bands were visualized by chemiluminescence.
Experimental conditions were compared with the non-parametric Wilcoxon-Mann-Whitney rank sum test using Synergy KaleidaGraph v3.6.2 (Reading, PA).
We thank Annette L. Boman, Todd R. Graham, Andrew A. Peden and Silvija I. Staprans for supplying reagents, and members of the Faundez and Corbett labs for helpful discussions. This work was supported by a pre-doctoral NRSA fellowship from the National Institute of Neurological Disorders and Stroke, National Institutes of Health, to A.C.B. (NS044743), a research training grant in Dermatology from the National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, to M.L.S. (2T32AR007587-11), grants from the National Niemann-Pick Disease Foundation to A.H.C. and to R.A.M., and grants from the National Institutes of Health to R.A.M. (NS40564) and to V.F. (NS42599 and GM 077569). An award from the American Heart Association supported E.W.
Supplementary material available online at http://jcs.biologists.org/cgi/content/full/120/20/3640/DC1
↵* Present address: NIH, NCI Experimental Immunology Branch, Bethesda, MD 20892, USA
- Accepted August 13, 2007.
- © The Company of Biologists Limited 2007