|Home | About | Journals | Submit | Contact Us | Français|
SorLA/LR11 (250 kDa) is the largest and most composite member of the Vps10p-domain receptors, a family of type 1 proteins preferentially expressed in neuronal tissue. SorLA binds several ligands, including neurotensin, platelet-derived growth factor-bb, and lipoprotein lipase, and via complex-formation with the amyloid precursor protein it downregulates generation of Alzheimer's disease-associated Aβ-peptide. The receptor is mainly located in vesicles, suggesting a function in protein sorting and transport. Here we examined SorLA's trafficking using full-length and chimeric receptors and find that its cytoplasmic tail mediates efficient Golgi body-endosome transport, as well as AP-2 complex-dependent endocytosis. Functional sorting sites were mapped to an acidic cluster-dileucine-like motif and to a GGA binding site in the C terminus. Experiments in permanently or transiently AP-1 μ1-chain-deficient cells established that the AP-1 adaptor complex is essential to SorLA's transport between Golgi membranes and endosomes. Our results further implicate the GGA proteins in SorLA trafficking and provide evidence that SNX1 and Vps35, as parts of the retromer complex or possibly in a separate context, are engaged in retraction of the receptor from endosomes.
SorLA is a 250-kDa type 1 receptor and a member of the Vps10p-domain (Vps10p-D) receptor family (16). It is a mosaic protein and has also been designated LR11, due to distinct structural similarities to the classical members of the low-density lipoprotein (LDL) receptor family (38). SorLA is mainly expressed in various parts of the nervous system, notably in cortical neurons, hippocampus, the cerebellum, and the spinal cord, but appreciable amounts are also found in alternative tissues such as testis, ovary, lymph nodes, distal kidney tubules, and vascular smooth muscle cells (12, 16, 18, 27, 38). Similar to the other Vps10p-D receptors Sortilin and SorCS1-3, SorLA carries a multi-ligand binding N-terminal Vps10p-D (15). The domain contains a propeptide that is removed by cleavage in the trans-Golgi network (TGN) (15). In contrast to Sortilin, SorLA does not depend on the propeptide for normal processing and transport in the biosynthetic pathway, but in both receptors cleavage is a prerequisite for binding of ligands to the Vps10p-D (15, 21, 37). The domain is targeted by peptides and growth factors and is responsible for the high-affinity binding of neurotensin and glia cell line-derived neuronal factor to SorLA (15, 21, 37). Whereas the Vps10p-D constitutes the entire, or most of the luminal part of Sortilin and SorCS1-3, SorLA is far more composite. Apart from the Vps10p-D it comprises a β-propeller domain with an associated epidermal growth factor class B-like motif, a cluster of 11 LA (LDL-receptor class A) repeats, and a juxtamembrane cluster of fibronectin type III repeats (16, 38). Similar β-propeller domains and clusters of LA-repeats constitute characteristic structural elements in members of the LDL receptor family and account for the binding of a broad variety of ligands to classical family members such as LDL-related protein (LRP)/LR1 and megalin (14, 23). Accordingly, many of these ligands, including receptor-associated protein (RAP), platelet-derived growth factor-bb (PDGF-bb), lipoprotein lipase, apolipoprotein E, and elements of the plasminogen activator system, also bind to SorLA (10, 15, 34).
SorLA is subject to various degrees of tumor necrosis factor α-converting enzyme-mediated cleavage, and ligands may bind to soluble receptors released to the extracellular fluid by cleavage, as well as to full-length receptors on the plasma membrane (13). The function of soluble receptors, for instance as carrier proteins, is virtually undescribed but, like LRP, the full-length receptors are capable of uptake and endocytosis of bound ligands and facilitate signal transduction in response to ligand binding (10, 39). Thus, overexpression of SorLA enhances PDGF-bb induced migratory and invasive activity of cultured smooth muscle cells and the receptor promotes internalization of cell-bound urokinase-type plasminogen activator-type 1 plasminogen activator inhibitor (uPA-PAI) complexes. The latter findings may implicate SorLA in the development of atherosclerosis and demonstrate that, compared to endocytosis by LRP, the uptake of uPA-PAI-1 complex via SorLA is slow and does not contribute to the clearance and recycling of urokinase receptors.
Highly interesting but completely different aspects of SorLA function have been introduced by the observation that SorLA interacts with the amyloid precursor protein (APP) (1, 2). APP is a transmembrane protein closely associated with the development of Alzheimer's disease (AD), a neurodegenerative disorder characterized by neurofibrillary tangles and amyloid plaques. The main components of the plaques are aggregates of amyloid-β (Αβ) peptides originating from sequential proteolysis of APP by β-secretase (the β site of APP-cleaving enzyme [BACE]) and γ-secretase (presenilin 1) activity (33). Recent studies show that coexpression of APP and SorLA results not only in complex formation between the two but also in SorLA-dependent translocation of APP and a concomitant drastic decrease in the generation of Aβ peptides (1, 24, 33). The findings suggest that APP and SorLA may interact early in the biosynthetic pathway and that SorLA protects APP from BACE activity by steric hindrance and/or diversion of APP trafficking. The fact that Aβ peptide generation is significantly increased in SorLA knockout mice (1) and the observation that SorLA appears to be significantly downregulated in the brains of AD patients further emphasize the notion that SorLA antagonizes Aβ peptide generation and AD in vivo (28).
The physiological role(s) of SorLA is far from clarified, but in view of the findings described above and the many different types of ligands targeting the receptor, there can be little doubt that SorLA is multifunctional. Most likely, the receptor can take on different functions depending on the environmental setting. This includes both mechanisms relating to events at the cell surface and, as exemplified by the recent findings in APP, intracellular transport (1, 33). Considering that SorLA predominates in vesicular and perinuclear compartments, the latter context may be particularly important. Like the other Vps10p-D receptors SorLA has a short (54 residues) cytoplasmic domain (cd). Its C terminus constitutes a site for binding of GGA1 to -3 (Golgi membranes localizing γ-adaptin ear homologous ADP-ribosylating factor binding cytosolic adaptor proteins), which are believed to contribute to the sorting of receptors, e.g., Sortilin and the mannose 6-phosphate receptors (MPRs), from Golgi membranes to endosomes and the regulated secretory pathway (22, 25, 35, 40). The cd further comprises a series of potential motifs for transport by alternative adaptors. The motifs include the sequence F12ANSHY17, akin to functional sites found in the mannose receptor and in classical members of the LDL receptor family, as well as an acidic cluster (AC) and a couple of dileucine-like motifs.
We have examined here SorLA trafficking and show that its cellular localization differs from that of the two related receptors Sortilin and LRP. Using wild-type (wt)- and chimeric receptors, we probed SorLA's participation in Golgi membrane-endosome transport and established functional cd sequence motifs for endocytosis and sorting by mutational analysis. Finally, we have applied specific knockdown (RNA interference [RNAi]) of cytosolic adaptors and disruption of sorting sites in the SorLA-cd to clarify the putative involvement of GGAs, AP-1 and -2, PACS-1, and elements of the retromer complex in SorLA trafficking.
SorLA and Sortilin were expressed in eukaryotic cell lines by using the expression vectors pcDNA3.1(−)zeo and pcDNA6/Myc-His (Invitrogen Corp., United Kingdom) (15, 21). SorLA-cd was inserted in pGEX4T-1, expressed in BL21(DE3) cells, and purified by using glutathione-Sepharose beads according to the manufacturer's recommendations (Amersham Pharmacia, Buckinghamshire, United Kingdom). Chimeric receptors containing the luminal and transmembrane domain of the cation-independent MPR (MPR300) or the interleukin-2 receptor (IL-2R/Tac) fused to the intracellular domain of SorLA were constructed as described previously (22). In brief, cDNA encoding the wt cd of SorLA was amplified by using a standard PCR technique and appropriate primers to generate a 5′ HindIII site and a 3′ XhoI site for cloning in the pcDNA3.1/zeo(−)-IL-2R/Tac vector or a 5′ NheI site and a 3′ XhoI site for cloning in the pMPSVHE-MPR300 vector.
CHO-K1 cells were cultured in HyQ-CCM5 CHO medium (HyClone, Logan, UT). HEK293 cells, HepG2 cells, mpr−/− mouse embryonic fibroblast (MEF) cells (19) and μA1−/− MEF cells (20) were all cultured in Dulbecco modified Eagle medium (BioWhittaker, Veriers, Belgium) supplemented with 10% fetal calf serum (Invitrogen). Cells were transfected with FuGENE 6 transfection reagent (Roche, Germany), and stably transfected clones were selected by using zeocin at 500 μg/ml (Invitrogen), hygromycin at 500 μg/ml (Invitrogen), or blasticidin at 10 μg/ml (InvivoGen).
Transfection with specific and nontargeting small interfering RNA (siRNA; 100 nM) was performed with DharmaFECT transfection reagents (Dharmacon RNA Technologies) according to the manufacturer's protocol. The protein expression was analyzed by immunoblotting 72 h after transfection with specific antibodies against PACS-1 (a gift from Gary Thomas), Vps35 (goat polyclonal antibody; Abcam), AP-2 μ-chain (chicken polyclonal antibody; Abcam), AP-1 μ-chain (rabbit polyclonal antibody), and SNX1 (rabbit polyclonal antibody; Abcam and Bethyl Laboratories, Inc., Montgomery, TX). Blots were exposed in a Fuji film LAS1000 machine and horseradish peroxidase-labeled bands were quantified by using Fuji film Multi-Gauge software.
For inhibition of lysosomal enzymes, 0.1 μg of leupeptin and 0.05 μg of pepstatin (both from Sigma Aldrich)/ml were added to cell cultures 36 h after RNAi. Protein expression was determined in cell lysates after an additional 24 h of incubation.
CHO cells expressing IL-2R/SorLA chimeric receptors were incubated at 4°C in medium containing anti-Tac (Roche) labeled with 125I (ca. 3 × 104 cpm/ml). Unbound tracer was removed by washing, and the cells were reincubated in fresh medium at 37°C. At given times, incubation was stopped, and the cells were exposed to pH 2.5 in a Tris-HCl buffer (4°C). Radioactivity released from the cells by this procedure was defined as surface associated (not internalized). To visualize internalization of unlabeled anti-Tac, cells were washed, fixed in 3.7% formaldehyde containing 0.5% saponin, and finally stained with an Alexa Fluor 488 goat anti-mouse immunoglobulin antibody (Invitrogen).
Immunocytochemistry involving wt receptors was performed in stable transfectants using primary rabbit antibodies raised against SorLA-Vps10p-D and a monoclonal anti-Sortilin-Vps10p-D antibody. Rabbit anti-human TGN46 (Serotech, Norway) and rat anti-mouse Lamp1 (GL2A7; Hybridoma Bank, Iowa) antibodies were used for staining, and analysis was carried out in a LSM510-META laser scanning confocal unit using a ×63 water objective lens (N/A 1.2; Carl Zeiss, Germany). Alexa Fluor 488 and 568 antibodies were used for secondary labeling of all fixated cells (Invitrogen). Alternatively, internalized SorLA-Alexa-567-conjugated anti-SorLA complexes were examined by live imaging in 293 cells stably cotransfected with SorLA and GFP-EEA1. Acidic vesicles (lysosomes) were visualized in cells preincubated with LysoTracker (Invitrogen). Using 293 cells, SorLA and the coated pit were visualized in a digital Philips CM10 electron microscope according to a pre-embedding system, wherein 0.8-nm immunogold goat anti-rabbit immunoglobulin G (Aurion Immunogold) is used with silver enhancement (Aurion R-Gent SE-EM; EMS) on 10-mm serial frozen sections and TAAB812 epoxy resin from TAAB, Ltd.
The expression level of chimeric receptors in transfected mpr−/− MEFs was determined as previously described (9). In brief, mpr−/− transfectants cultured in 35-mm dishes were permeabilized and incubated with 5 μg of monoclonal antibody (2C2) recognizing a luminal epitope of MPR300. After removal of unbound antibody, cell-bound 2C2 was measured by using a sandwich enzyme-linked immunosorbent assay. The chimeric receptor expression was calculated as the nanograms of bound 2C2 antibody per milligram of total protein and compared to the level of endogenous MPR300 in normal MEFs.
The enzymatic activity of lysosomal β-hexosaminidase, β-glucoronidase, and β-galactosidase was determined as previously described (19). A Victor-3 microplate reader (Perkin-Elmer) was used for fluorometric measurements.
For biolabeling, cells were washed and preincubated in cysteine- and methionine-free medium (modified Eagle medium; Sigma, St. Louis, MO) prior to reincubation in similar medium, with or without brefeldin A, containing ~200 μCi of l-[35S]cysteine and l-[35S]methionine (Pro-Mix; Amersham Pharmacia)/ml. For labeling of HepG2 and mpr−/− cells the medium was supplemented with 2 to 5% dialyzed fetal calf serum. The subsequent chase was done in full medium.
Immunoprecipitations were performed on medium or cell lysates containing 1% Triton X-100 supplemented with protease inhibitors (CompleteMini; Roche), using GammaBind G-Sepharose beads (Amersham Pharmacia) coated with the relevant protein-specific antibodies (21). Precipitated proteins were subjected to reducing polyacrylamide gel electrophoresis, and gels were exposed on Fuji film imaging plates in a FLA-3000. Bands were quantified by using Fuji film MultiGauge software.
Yeast two-hybrid analysis was conducted by using the Matchmaker LexA system (Clontech) as described previously (22). The SorLA cytosolic tail was inserted into the pBAD42 vector and cotransformed with a pLexA-vps35 construct (a gift from J. Bonifacino) and plated on plasmid selection plates. Resulting colonies were replica plated on X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) induction plates, together with controls for autoreactivity and using the VHS domain from GGA2 as a positive control.
For pull-down experiments, fresh mouse brain was homogenized in 3 ml of lysis buffer (20 mM HEPES-KOH, 125 mM KCl, 5 mM MgCl2, 320 mM sucrose, and 0.1 mM EDTA [pH 7.5] containing protease inhibitor). The homogenate was centrifuged (10 min, 5,000 × g); Triton X-100 was added to a final volume of 1.25%, followed by incubation for 10 min on ice. Cytosolic fractions were generated by centrifugation for 15 min at 20,000 × g and subsequently 60 min at 100,000 × g at 4°C. Supernatants were incubated with 100 μl of glutathione-Sepharose prebound with 200 μg of GST, GST-SorLA-cd, or GST-SorCS1c-cd for 2 h at 4°C. Protein complexes were subsequently subjected to AP-1 γ-chain or AP-2 α-chain immunoblotting (mouse AP-1 γ and AP-2 α antibodies were from BD Transduction Laboratories).
For preparation of postnuclear supernatant (PNS), cells were washed and harvested in Tris-buffered saline with 0.4 mM phenylmethylsulfonyl fluoride at 583 × g for 10 min at 4°C. Cells were resuspended in 10 ml of homogenization buffer (0.25 M sucrose, 10 mM HEPES-KOH [pH 7.2], 1 mM EDTA [pH 7.5], 1 mM magnesium acetate, 0.4 mM phenylmethylsulfonyl fluoride) and centrifuged at 583 × g for 7 min at 4°C. Cells were resuspended in 800 μl of homogenization buffer and disrupted by five passages through a 21-gauge needle, followed by eight passages through metal cell cracker with a 9-μm gap. The PNS was obtained after removal of nuclei and unbroken cells by centrifugation at 1,843 × g for 7 min at 4°C.
The cell components in the PNS were separated by velocity gradient centrifugation (SW41Ti rotor; 100,000 × g, 18 min) over a continuous 0.3 to 1.2 M sucrose gradient prepared by mixing at 45°C for 10 min and 80°C for 1 min using a BioComp gradient master. Fractions (1 ml) were collected by using a BioComp piston gradient fractionator. Selected fractions were pooled and subjected to equilibrium gradient centrifugation (SW41Ti rotor; 100,000 × g, overnight) over a 0.8 to 1.2 M sucrose step gradient. Fractions were collected and analyzed by Western blotting. Antibodies to the molecular markers (Golgi membranes, TGN, and endosomes) vti1a and vti1b were from BD Sciences.
To reduce potential type 1 errors related to multiple comparisons, sorting by the MPR300/SorLA constructs (in the presence of M6P) was statistically evaluated by a one-way analysis of variance. If this test was significant at a P value of <0.05 and passed the normality and equal variance test, individual comparisons between MPR300/Sorla wt and the other constructs were performed by a Bonferroni t test. Only probabilities associated with preplanned comparisons were tested.
SorLA shows structural similarities to Sortilin, as well as to LRP. The three receptors share a number of ligands and are all capable of endocytosis, but in terms of subcellular localization they clearly differ. LRP has a much higher expression at the plasma membrane than the two Vps10p-D receptors, and although SorLA and Sortilin both predominate intracellularly, colocalization is only partial. In double transfectants expressing Sortilin and SorLA, SorLA displays a more dispersed vesicular expression pattern than Sortilin, which concentrates mainly in perinuclear compartments (Fig. (Fig.1A).1A). In accordance with this observation, subcellular fractionation shows a clear difference in the overall localization of the two wt receptors in HEK293 double transfectants (Fig. (Fig.1B).1B). As is apparent from Fig. Fig.1B1B (lower panels), this difference becomes particularly distinct when TGN46 containing fractions obtained in velocity gradients are subjected to additional separation on an equilibrium gradient. A similar discrepancy was observed in relation to endogenous LRP (Fig. (Fig.1B1B).
For further analysis of SorLA trafficking, we constructed two types of chimeric receptors (Fig. (Fig.2)2) combining the SorLA-cd with the luminal and transmembrane domains of either the IL-2R or of the MPR300. Both chimeras (IL-2R/SorLA and MPR300/SorLA) were generated as a series of receptors containing either the wt SorLA-cd, various truncated cd's, or mutant cd's comprising amino acid substitutions in candidate sorting motifs (Fig. (Fig.2B2B).
Initially, CHO-K1 cells, which have no endogenous expression of IL-2R, were stably transfected with the IL-2R/SorLA chimeras. For identification of cd motifs involved in SorLA endocytosis, the transfectants were incubated (4°C, 2 h) with 125I-labeled monoclonal anti-Tac. Unbound antibodies were subsequently removed by washing, and the cells were reincubated in warm medium. At the times indicated (Fig. (Fig.3A),3A), incubations were stopped, and the amount of internalized receptor was measured as cell-associated radioactivity not released upon treatment with acid (pH 2.5). As determined by this procedure, endocytosis mediated by the SorLA wt-cd was both rapid and quantitatively efficient, and at 15 min <25% of the receptors were left at the cell surface. On the other hand, truncated chimeras containing only the first 10 or 29 residues of the SorLA-cd, exhibited next to no internalization (Fig. (Fig.3A,3A, left panel), demonstrating that the N-terminal half of the cd does not harbor independent functional sites for endocytosis. In agreement with this, selective disruption of the NPXY-like motif (F12N14Y17) did not change the rate of internalization (Fig. (Fig.3A,3A, right panel). Similarly, a small C-terminal truncation (ΔM51VIA54), disrupting the established GGA binding and a potential PDZ-domain type 2 binding site, did not hamper internalization, whereas removal of the last 14 residues caused a significant reduction (~35% at 15 min; Fig. Fig.3A,3A, left panel). It follows that the segment D30-P50, which contains a large AC and the dileucine-like motif MI, is decisive in SorLA endocytosis. While further analysis using alanine substitutions ascribed little or no function to the M41I42, even modest changes in the AC demonstrated a significant functional impact and substitution of D30D31,E34DDED38 with alanines drastically slowed down internalization (Fig. (Fig.3A,3A, right panel). Also shown are similar results obtained with chimeras (Fig. (Fig.3B),3B), as well as full-length SorLA constructs (Fig. (Fig.3C)3C) by confocal microscopy and detection of receptor-antibody complex after 0 to 30 min of incubation at 37°C. We conclude that the AC constitutes the single most important endocytic motif in SorLA.
The cytosolic adaptor PACS-1 targets cd's comprising acidic clusters and is essential to endocytosis and the retrograde transport of Furin (36). A GST-PACS-1(117-266) fusion protein provided efficient pull-down of wt SorLA from cell lysates, suggesting that PACS-1 might have a similar role in SorLA trafficking (see Fig. S1A in the supplemental material). However, RNAi and transient knockdown (>90%) of PACS-1 did not affect internalization of IL-2R/SorLA-125I-labeled anti-Tac complexes in 293 transfectants (see Fig. S1B in the supplemental material).
SorLA also interacts with the AP-2 adaptor complex, and AP-2 α-chain of brain extracts was efficiently precipitated by a SorLA-cd GST fusion protein but not by a truncated SorLA-cd not containing the AC [Δ(30-54)], by GST alone, or by the cd of SorCS1c, another member of the Vps10p-D receptor family (Fig. (Fig.4A,4A, upper panel). To determine the functional involvement of AP-2, we therefore examined SorLA endocytosis before and after disruption of the AP-2 complex. As outlined (Fig. (Fig.4A,4A, lower panel), knockdown by >75% of the AP-2 μ-chain, caused SorLA to remain on the plasma membrane, whereas receptors in cells transfected with a nontargeting siRNA exhibited normal internalization and a vesicular localization after 30 min at 37°C. As determined by measuring the receptor-mediated uptake of 125I-labeled anti-Tac, endocytosis (at 30 min) amounted to <10% upon μ-chain knockdown compared to ~70% in the absence of RNAi (Fig. (Fig.4A).4A). We conclude that AP-2 is essential to SorLA endocytosis.
After internalization, IL-2R/SorLA in complex with anti-Tac rapidly accumulates in the perinuclear compartments of CHO cells. Prolonged incubations (>1 h) suggested redistribution in intracellular vesicles but little or no recycling to the cell surface (not shown). A closer examination of the routing of wt SorLA was performed in 293 cells. It appears that SorLA, similar to Sortilin, is internalized via coated pits (Fig.4Ba). Within the first 10 to 20 min, practically all receptor-antibody complexes initially found on the cell surface were translocated to EEA1-associated early endosomes, and after ~30 min at 37°C colocalization with TGN46 was seen (Fig. 4b and c). Occasional colocalization with Lamp-1 was also observed in the same cells, but only after about 30 min of incubation (Fig.4Bd). Finally, tracking (live imaging) of internalized receptor-antibody complexes for up to 2 h, confirmed that, once internalized, the receptors remain in intracellular compartments avoiding translocation to strongly acidic (pH < 5.5, lysotracker positive) vesicles (see Fig. S2 in the supplemental material).
We next decided to examine whether SorLA, like its relative Sortilin, engages in Golgi body-endosome trafficking. To that end, the MPR300/SorLA chimeras were expressed in mpr−/− mouse fibroblasts. The mpr−/− cells lack the MPRs and therefore secrete the hydrolases that are normally transported to the late endosomes by these two receptors. Hydrolase-sorting and functional lysosomes can be restored by transfecting the cells with wt MPR300 or any receptor with an ectodomain that can bind the hydrolases and a cd capable of mediating Golgi body-endosome transport. The MPR300/SorLA chimeras can bind the hydrolases, and the sorting capacity of their respective cd's can consequently be evaluated by monitoring the secretion and/or lysosomal homing of relevant hydrolases in mpr−/− transfectants. As is apparent from Fig. Fig.5,5, stable expression of a chimera carrying the SorLA wt-cd reestablished lysosomal homing of β-hexosaminidase, β-glucuronidase, and β-galactosidase and normalized lysosomal morphology. Between 70 and 80% of the hydrolases produced in mpr−/− cells, versus <15% of those synthesized in transfectants, were found in the culture medium. In the presence of mannose 6-phosphate (M6P; 5 mM), which prevents reuptake of secreted hydrolases via receptors on the surface membrane, the amount of enzymes in the medium was only increased by ca. 15%. It follows that upon expression of MPR300/SorLA the major fraction of hydrolases is subject to direct TGN-endosome sorting and never reaches the medium.
Similar results were obtained by biolabeling and pulse-chase of cathepsin D (Fig. (Fig.6A).6A). The expression of chimeras caused a substantial reduction in the secretion of newly synthesized cathepsin D and, as demonstrated by subsequent Western blotting, this was accompanied by a significant increase in the conversion of procathepsin D to the mature enzyme signifying sorting to, and processing in, the lysosomes (Fig. (Fig.6B6B).
Taken together, our data establish that at similar levels of expression, the cd of SorLA is just as efficient for Golgi body-endosome sorting as the cd's of MPR300 and Sortilin (22).
Evidence of sorting by wt SorLA was obtained in CHO cells devoid of LRP but expressing SorLA and a secretable construct (without the C-terminal endoplasmic reticulum retention sequence HNEL) of the SorLA and LRP binding ligand RAP. After biolabeling, the cells were reincubated in fresh medium, and at given times RAP was immunoprecipitated from medium and cell lysates. The results showed that on coexpression with SorLA, secretion of RAP to the medium was reduced by ca. 30%. Thus, SorLA can target soluble ligands in the biosynthetic pathway for transport (see Fig. S3 in the supplemental material).
Results of β-hexosaminidase sorting in mpr−/− cells transfected with MPR300/SorLA mutants are depicted in Table Table1.1. Obviously, truncated receptors missing the C-terminal half (amino acids 30 to 54) of the SorLA-cd were incapable of sorting. A more detailed examination of this segment revealed that sorting was hampered by alanine substitutions in the AC and the adjacent dileucine-like M41I42 motif. Moreover, the presence of M6P had only a marginal effect on the release of hydrolase in cultures transfected with truncated chimeras without the AC, a finding in good agreement with the observations presented above regarding internalization. A modest but statistically significant increase (~15%) in the secretion of β-hexosaminidase was also seen upon the exchange of M51 and after the exchange of D47 or D48 for alanine. Inasmuch as both D48 and M51 are essential to the interaction between SorLA and GGAs (17), the findings strongly indicate that two types of motifs, i.e., the GGA-binding site and the AC-dileucine motif, govern the Golgi body-endosome trafficking conveyed by the SorLA-cd.
The involvement of GGAs was further supported by subcellular fractionation, which demonstrated a redistribution of SorLA after disruption of its C-terminal GGA-binding site (Fig. (Fig.7E).7E). Additional experiments were performed to identify alternative cytosolic adaptors with a function in SorLA sorting. Since the AC-dileucine motif is considered a classical target for interaction with the AP-1 adaptor complex, we initially examined the possible involvement of this complex by expressing full-length SorLA in AP-1 μ-chain deficient fibroblasts (20). Lack of the μ-chain abrogates AP-1 functions and, as determined by immunofluorescence (Fig. (Fig.7A),7A), staining for full-length SorLA exhibited a broader vesicular distribution in deficient cells than in wt fibroblasts. This difference was confirmed by subcellular fractionation, which demonstrated a significant redistribution of SorLA in AP-1 μ-chain-deficient cells (Fig. (Fig.7B).7B). In agreement with these findings, the wt SorLA-cd, but not the truncated construct Δ(30-54), was found to mediate the pull-down of AP-1 from brain lysates (Fig. (Fig.7C).7C). To determine whether this interaction implicated Golgi body-endosome transport, we next examined the influence of μ-chain knockdown on the sorting of β-hexosaminidase in mpr−/− cells transfected with MPR300/SorLA. The results depicted in Fig. Fig.7D,7D, clearly demonstrate that knockdown (>90%) was accompanied by a substantial increase in cellular secretion of β-glucuronidase. Similar results were obtained in three separate experiments showing that direct endosomal sorting from Golgi bodies is seriously impaired in the absence of a fully functional AP-1 complex, as demonstrated for the lysosomal sorting of major histocompatibility complex class I complexed with viral protein gp48 (26).
Additional analysis was performed to elucidate the possible involvement of the mammalian retromer complex, which has been reported to partake in endosome-to-Golgi body retrieval of MPRs and Sortilin (3, 30). This complex comprises five subunits, including sorting nexin1 (SNX1) and Vps35, and a recent finding in HepG2 cells has shown that deficiency in SNX1 is accompanied by a downregulation of Sortilin expression, most likely due to mis-sorting and subsequent degradation of the receptor in lysosomes (M. Mari, unpublished results). We were able to confirm this result. Moreover, in five separate experiments in which RNAi reduced the level of SNX1 in HepG2 by >90%, we found a significant decrease (40 to 80%) in the expression of both Sortilin and SorLA (Fig. (Fig.8A).8A). As determined by immunoprecipitation of receptors in biolabeled HepG2-cells, synthesis was not affected by RNAi (not shown). On the other hand, treatment with inhibitors of lysosomal enzymes was found to increase receptor expression by ca. 30% in wt Hep-G2 cells and, notably, by >60% in cells subjected to SNX1 knockdown (Fig. (Fig.8A).8A). Thus, SNX1 deficiency seems to downregulate SorLA expression by enhancing its redistribution to and degradation in the lysosomes.
Corresponding experiments in HepG2 cells were carried out to evaluate the impact of Vps35 deficiency. As can be seen (Fig. (Fig.8B),8B), Vps35 knockdown had little effect on Sortilin expression (<20% decrease) but did reduce SorLA by 40 to 50% (in each of five experiments). The collected results evidently suggest that the retromer complex is involved in SorLA sorting. It is therefore interesting that the SorLA-cd, in a yeast two-hybrid analysis (see Fig. S4 in the supplemental material), reacted strongly with the GGA2 adaptor (positive control) but not with Vps35, which is considered the receptor-binding element of the mammalian retromer (3, 5).
Application of SNX1 RNAi did not affect β-hexosaminidase sorting in mpr−/− cells transfected with MPR300/SorLA. However, this result cannot be considered fully informative since we never managed to reduce SNX1 by >60% in this cell type. In contrast, knockdown of PACS-1 in mpr−/− was highly effective (>90%). PACS-1 has recently been implicated in Golgi body-endosome trafficking (29), but in our hands PACS-1 deficiency did not affect β-hexosaminidase sorting by MPR300/SorLA (see Fig. S1C in the supplemental material). Notably, uptake of hydrolases from the medium (difference between enzyme activity found in the medium of unsupplemented cultures and of cultures supplemented with M6P) was unchanged by PACS-1 deficiency. This finding agrees with our findings regarding receptor internalization and underscores the lack of any detectable influence of PACS-1 on the endocytosis of SorLA.
Only a small group of receptors, which includes the Vps10p-D receptor Sortilin and the two MPRs, are known to run a regular service between the TGN and the early and late endosomes. The present report adds SorLA to the chosen few. Our findings demonstrate that although SorLA is expressed on the plasma membrane, the vast majority of receptors are intracellular and engage in Golgi body-endosome transport. Thus, the expression of MPR300/SorLA prevents the secretion of hydrolases in mpr−/− cultures, restores lysosomal morphology and promotes the conversion of hydrolases from their proform to active mature enzymes. The finding that secretion of RAP is downregulated upon coexpression with SorLA further demonstrates that the wt receptor has the capacity to target newly synthesized soluble ligands for sorting in the biosynthetic pathway. In agreement, we find that SorLA sorting depends on cytosolic interactors that have previously been shown to facilitate Golgi body-endosome shuttling. Still, one must keep in mind that our results are based mainly on chimeric receptors carrying the SorLA-cd. It can therefore not be excluded that trafficking of the full-length wt receptor is influenced by additional factors not reflected by the present study.
Like Sortilin and the MPRs, SorLA predominates intracellularly, and only ca. 5 to 10% of the receptors are present on the plasma membrane at any given time (15). Accordingly, SorLA is endocytic and has been shown to mediate internalization of several ligands (10, 15, 34). Our present findings show that endocytosis relies on the C-terminal half of the cytoplasmic tail and identifies the AC (D30-D38) as the major functional motif. Alanine substitutions in the AC impaired endocytosis, and exchange of the entire cluster seriously slowed the process. A truncation immediately after the AC also affected internalization, but selective mutations in this segment did not. It is therefore likely that the latter truncation inhibits endocytosis by hampering the AC and not because it harbors additional functional motifs. Thus, an AC but none of the classical motifs (YXXΦ, dileucines, or NPXY) for endocytosis via coated pits seem active in the SorLA-cd. Acidic clusters are targeted by PACS-1, which partakes in the endocytosis and retrograde sorting of furin (8). PACS-1 may in fact bind SorLA, but receptor internalization was virtually unchanged at <10% of normal PACS-1 levels. In contrast, endocytosis was clearly affected by changes in the AP-2 complex. SorLA binds to the AP-2 complex, and inactivation of AP-2 complex by μ-chain knockdown almost completely blocks internalization. It follows that AP-2 is essential to endocytosis of the receptor, whereas SorLA, similar to Sortilin, does not depend on PACS-1 for this function.
After internalization, SorLA accumulates in early endosomes and, within 30 min, relocation to TGN46 positive and, to a minor degree, Lamp-1-positive vesicles is visible. The internalized receptors do not return to the plasma membrane but appear to incur cycling between intracellular compartments. In agreement, our findings in mpr−/− cells show that SorLA is engaged in Golgi body-endosome transport. The MPR300/SorLA retained hydrolase precursors in the cells and promoted their processing to mature enzymes by delivering them to the lysosomes. We have previously reported similar results for Sortilin (22); however, the overall localization of Sortilin and SorLA differs, suggesting alternative routes of trafficking. Like Sortilin, SorLA depends on at least two sites in its cd for TGN-endosome sorting. One of the two is situated at the extreme C terminus of the cd and implicates the GGA-binding motif D48XXM51. Exchange of D or M for alanine or removal of M by truncation (ΔMVIA) disrupts the motif and prevents binding of all three GGAs to SorLA (17). Here we find that sorting of β-glucuronidase is impaired in mpr−/− cells if the MPR300/SorLA carries a dysfunctional GGA-binding motif. This and the fact that the M51VIA54 truncation alters the subcellular localization of the wt receptor strongly indicate the involvement of GGAs in the TGN-endosome trafficking of SorLA.
The other involved site maps to the AC-dileucine-like motif. This type of motif is a well-known candidate for interaction with the tetrameric adaptor protein complexes, notably AP-1 which, along with the GGAs, is implicated in sorting of MPRs at the TGN (6). The present findings clearly demonstrate that AP-1 is similarly important to SorLA sorting. Fluorescence microscopy, as well as subcellular fractionation, presented a significant difference between localization of wt SorLA in μ-chain-deficient mouse fibroblasts and their wt counterparts, and μ-chain knockdown seriously impaired hydrolase transport in MPR300/SorLA mpr−/− cells.
Recent findings provide evidence that GGA3 and PACS-1 in complex with casein kinase may act in concert to promote transport of the MPR300 from TGN and their return from endosomes (29). As shown here, GGAs engage in SorLA sorting. We therefore challenged PACS-1 function by RNAi, but found no detectable effect on hydrolase-sorting in mpr−/− transfectants. Evidently, PACS-1 is either not involved or its function is subject to redundancy.
We finally probed for the participation of SNX1 in SorLA sorting. SNX1 is part of the retromer, a multisubunit that mediates retrograde transport of MPRs between endosomes and the TGN (3). The mammalian retromer comprises the subunits SNX1, SNX2, Vps26, Vps29, and Vps35 (11). The functional organization of the complex is not entirely clear, but it appears that SNX1 and SNX2 mediate binding to phosphoinositides in highly curved endosomal membranes (7), whereas Vps35 may account for capture of cargo receptors (3, 31). Current evidence suggests that Sortilin is a target for retromer sorting, and a recent finding has established that SNX1 knockdown in HepG2 cells is accompanied by a significant decrease in cellular Sortilin resulting from a reduction in receptor half-life (M. Mari, unpublished results). We confirm this finding and show that SorLA is subject to a similar downregulation and reduction of half-life in response to deficiency of SNX1. In all probability the findings reflect that SNX1 is needed (at least in HepG2 cells) for the retraction of Sortilin and SorLA from the endosomes and enables them to escape degradation in the lysosomes. It is less clear whether this also implicates the retromer complex. The fact that deficiency in Vps35 was accompanied by a significant reduction in SorLA does suggest that sorting of this receptor involves the retromer. In contrast, Sortilin expression was (surprisingly) unchanged under the same conditions. This underscores in the first place a difference in the trafficking of the two receptors and, second, provides evidence that SNX1 function is not limited to the context of the retromer complex. The apparent lack of interaction between the SorLA-cd and Vps35 furthermore indicates that the retromer may target receptors via alternative subunits.
The results obtained with RAP show that wt SorLA may target soluble ligands in the TGN for transport, e.g., lipoprotein lipase, and thereby regulate their release. However, SorLA also binds and sorts the transmembrane protein APP (1), which produces the Alzheimer's disease-associated Aβ-peptide when processed by β-secretase activity. After complex formation, SorLA protects APP from β-secretase activity and inhibits the generation of Aβ-peptide in cells (1, 24, 33). Accordingly, SorLA seems to be downregulated in brains of individuals with AD. In the light of our present findings it is therefore highly interesting that a similar decrease in Vps35 and other elements of the retromer complex has also been associated with AD and the generation of Aβ-peptide (32). As demonstrated, Vps35 and (in particular) SNX1 deficiency may downregulate SorLA in cells. Thus, a functional connection between low levels of SorLA and the two proteins in AD seems conceivable, and it could be speculated that defects or deficiency in SNX1 and/or Vps35 (as parts of the retromer or in a different context) might form a basis for accelerated Αβ-peptide production.
The MIND-center is sponsored by The Lundbeck Foundation.
Published ahead of print on 23 July 2007.
†Supplemental material for this article may be found at http://mcb.asm.org/.