SorLA/LR11 Regulates Processing of Amyloid Precursor Protein via Interaction with Adaptors GGA and PACS-1*

SorLA has been recognized as a novel sorting receptor that regulates trafficking and processing of the amyloid precursor protein (APP) and that represents a significant risk factor for sporadic Alzheimer disease. Here, we investigated the cellular mechanisms that control intracellular trafficking of sorLA and their relevance for APP processing. We demonstrate that sorLA acts as a retention factor for APP in trans-Golgi compartments/trans-Golgi network, preventing release of the precursor into regular processing pathways. Proper localization and activity of sorLA are dependent on functional interaction with GGA and PACS-1, adaptor proteins involved in protein transport to and from the trans-Golgi network. Aberrant targeting of sorLA to the recycling compartment or the plasma membrane causes faulty APP trafficking and imbalance in non-amyloidogenic and amyloidogenic processing fates. Thus, our findings identified altered routing of sorLA as a major cellular mechanism contributing to abnormal APP processing and enhanced amyloid β-peptide formation.

Golgi network (TGN) and early endosomes (4). Although the exact trafficking route of sorLA has not been established yet, related members of the gene family were shown to move to the cell surface, internalize once, and thereafter shuttle between TGN and endosomal compartments (5).
A hypothesis concerning the relevance of sorLA-mediated transport processes in neurons has been advanced based on studies that identified this protein as intracellular receptor for the amyloid precursor protein (APP), the etiologic agent in Alzheimer disease (AD) (6). APP follows a complex trafficking pathway through secretory and endocytic compartments that determines processing into amyloidogenic and non-amyloidogenic products. It is generally believed that in route to the cell surface, most newly synthesized APP molecules are cleaved by ␣-secretase in a post-Golgi compartment or at the plasma membrane to produce soluble (s)APP␣. Alternatively, some precursor molecules are re-internalized from the plasma membrane and delivered to late endocytic compartments for ␤-secretase (and subsequent ␥-secretase) processing into sAPP␤ and amyloid ␤-peptide (A␤) (7,8).
Because alternative targeting of APP to Golgi compartments or to endosomes is a crucial determinant of proteolytic processing and A␤ production (9), much attention has been focused on elucidating the underlying regulatory mechanisms. A decisive role for sorLA in this process has been suggested because sorLA directly interacts with APP and colocalizes with the precursor in the perinuclear region of neuronal cell types (6,10). Interaction involves binding sites in the luminal as well as in the intracellular domains of both proteins (11,12). Binding to sorLA results in sequestration of APP in intracellular compartments and in reduction of APP processing into both sAPP␣ and A␤ products (6). The significance of sorLA as negative regulator of APP processing in vivo is supported by studies in sorLA-deficient mice that exhibit significantly higher levels of A␤ in the brain compared with control animals (6), similar to the situation seen in patients with sporadic AD that lack receptor expression (13,14). Genetic evidence for a causal role of the receptor in neurodegenerative disease was obtained in epidemiological studies that uncovered association of inherited Sorla/ Sorl1 gene variants with late-onset AD (15,16). Here, we used wild type and trafficking defective mutants of sorLA to decipher the molecular mechanisms governing the intracellular transport route of the receptor, and their relevance for APP transport and amyloidogenic peptide production.

EXPERIMENTAL PROCEDURES
Materials-Antibodies against sorLA were provided by Claus Munck Petersen (University of Aarhus); antisera against APP were produced in-house (1227) or obtained from commercial suppliers (6E10, Signet; JP18957, IBL, Hamburg). Constructs for expression of paGFP-and mRFP-tagged APP and sorLA variants were provided by Robert Spoelgen (Harvard Medical School); HA-PACS-1 and GFP-GGA-1 were from Gary Thomas (Oregon Health Sciences University) and Juan S. Bonifacino (National Institutes of Health, Bethesda, MD), respectively. A plasmid encoding fluorescence-tagged endosomal marker Rab11A-GFP was obtained by cloning the respective cDNA from the UMR cDNA Resource Center into the pEGFPC2 vector from Clontech. The marker pDsRedm-GalT was purchased from Clontech; Alexa Fluor633-conjugated transferrin from Invitrogen.
Cell Culture Experiments-Cellular localization of proteins was detected by sequential scanning confocal immunofluorescence microscopy using Alexa 488-, Alexa 555-, or Alexa 568conjugated secondary antibodies and a ϫ100 Plan-Apochromat oil objective. Subcellular fractionation of cells by discontinuous iodixanol density gradient was performed according to published protocols (6). For pulse-chase experiments, cells were incubated in cysteine-and methionine-free medium (Sigma) for 60 min prior to biolabeling using 200 Ci of L-[ 35 S]cysteine and L-[ 35 S]methionine/ml (Pro-mix, Amersham Biosciences) for 7 min. Thereafter, cells were washed in phosphate-buffered saline, chased for various time points, and lysed. Lysates were precipitated with protein G-Sepharose beads (Amersham Biosciences) pre-coated with antiserum directed against the carboxyl-terminal 15 amino acid residues of human APP (IgG 1227). After incubation for 16 h at 4°C, beads were washed in phosphate-buffered saline and prepared for standard SDS-PAGE and fluorography. For deglycosylation studies, protein extracts were incubated with neuraminidase (Roche) and O-glycosidase (Roche) in 60 mM Na-acetate, pH 5.5, overnight at 16°C, followed by SDS-PAGE and Western blot analysis. For FACS analysis, cells were harvested and incubated in FACS buffer (0.1% fetal calf serum, 1% formaldehyde in phosphate-buffered saline) without (surface staining) or with 0.1% saponin (total staining), followed by incubation with pri-mary anti-sorLA (1:40), and Alexa Fluor 488-conjugated secondary IgG (1:100). Cells were subjected to FACS analysis (10,000 events/gate) using program BD Biosciences CellQuest Pro Version 5.2.1.
Live Cell Imaging-CHO cells on coverslips were transferred in medium buffered with 20 mM HEPES to a live cell chamber heated to 37°C on the microscope stage. A LSM510 inverted confocal microscope (Zeiss) equipped with a pulsed, tunable TiSa laser (Chameleon, Coherent) and a ϫ40 or ϫ63 Plan-Apochromat oil objective was used. paGFP was photoactivated by a 820-nm laser pulse with 5% transmission at zoom 30 for 0.5 s. Stacks of z-sections of GFP (excitation (ex): 488 nm, 3% transmission, emission (em): 500 -550 nm BP) and mRFP/mDsRed (ex: 543 nm, 10 -15% transmission, em: 560 nm LP) were sequentially acquired before and after photoactivation. Z-stacks were aligned for drift in x,y,z direction. Projections of confocal z-stacks are displayed. Single aligned z-sections were used for quantification of mean fluorescence intensity of a selected area using LSM software Image Examiner. For quantification of sorLA in recycling endosomes, representative micrographs from CHO-A/S wt and CHO-A/S gga cells were converted into binary images, and the number of peripheral vesicles either stained for sorLA only or for sorLA and Rab11 were counted manually.
APP Processing Products-The amount of sAPP, carboxylterminal fragments (CTF), and A␤ products were determined by Western blotting (antisera WO2, 1227, and JP18957) or ELISA (KHB3482 for A␤ 40 and KHB3442 for A␤ 42 ; BIOSOURCE). Statistical significance of the data were determined using Student's t test.

RESULTS
Previously, we reported that co-expression with sorLA results in accumulation of APP in the perinuclear region and in reduced exposure on the surface of cultured cells (6). To dissect which step in the intracellular transport route of APP may be influenced by sorLA, we studied its trafficking through the secretory and endocytic compartments in CHO and in human neuronal SH-SY5Y cells in the presence or absence of sorLA.
Initially, we characterized export of nascent APP molecules from the endoplasmic reticulum (ER). To do so, we studied acquisition of carbohydrate modifications on APP, indicative of subsequent ER and Golgi maturation steps. In CHO cell lysates, APP immunoreactivity was seen in three distinct bands (Fig.  1A). Treatment of cell extracts with neuraminidase and O-glycosidase uncovered that the high molecular weight band (sensitive to neuraminidase) represented the mature protein, carrying terminally modified N,O-linked carbohydrates. The intermediate sized band corresponded to APP carrying complex O-linked carbohydrates as it was sensitive to O-glycosi-dase. The lowest molecular weight band represented the immature precursor in the ER as it was resistant to neuraminidase and O-glycosidase treatment. No qualitative difference in this pattern was seen when comparing CHO cells stably expressing human APP 695 (CHO-A) and APP 695 with sorLA (CHO-A/S wt ) (Fig. 1A). However, CHO-A/S wt cells showed an increase in O-glycosylated APP products (middle and upper protein bands) compared with CHO-A cells (Fig. 1A, bracket), suggest-ing accumulation of the mature protein in the presence of sorLA (ratio mature/immature APP: 0.67 Ϯ 0.06 in CHO-A/S wt versus 0.46 Ϯ 0.06 in CHO-A, p ϭ 0.04). This effect was also seen in neuronal SH-SY5Y cells expressing the wild type receptor (supplemental Fig. S1). Accumulation of the mature APP species in the presence of sorLA wt was confirmed by studying the time course of glycosylation in pulse-chase experiments in CHO cells (Fig. 1B). The emergence of partially and fully glycosylated forms of APP (mature APP) followed similar kinetics in cells with and without sorLA (20 min chase), but their protein half-life was significantly extended in the presence of the sorting receptor as shown by determining the ratio of signal intensities for mature versus immature APP species (120 min chase; p Ͻ 0.01) (Fig. 1C).
As trafficking of APP from ER to Golgi compartments was unaffected by sorLA, we investigated its effect on APP transport through the Golgi, the major site of APP concentration in the cell (17). Significant co-localization of sorLA and APP in this compartment has been reported before (6,11). To test the consequence of sorLA activity on APP transit through this organelle, we used live cell imaging to visualize exit of the protein from the perinuclear region. APP immunoreactivity in this region mainly co-localized with ␤-1,4-galactosyltransferase (GalT), a marker for the trans-Golgi/TGN compartments (data not shown). A photoactivable APP-green fluorescent fusion protein (APP-paGFP) was activated in the perinuclear region (identified by co-expression with constitutively fluorescent APP-red fluorescent protein; APP-RFP) by a short laser pulse, and the trafficking out of this area recorded as a decrease in green fluorescence signal (schematic in Fig. 2D). In CHO cells expressing APP only ( Fig. 2A), the precursor exhibited rapid export from the perinuclear region as shown by the immediate appearance of vesicles containing APP-paGFP outside the area of photoactivation. Overall, the decrease in fluorescence signal in the photoactivated region was more than 60% within 12 min after laser pulse (Fig. 2C). Intriguingly, exit of APP-paGFP from the perinuclear region was impaired in the presence of sorLA-RFP as shown by the delayed appearance of APP-paGFP-positive vesicles in distal cell compartments (12 min after pulse; Fig.  2B, p Ͻ 0.005). In fact, the decrease in green fluorescence in this region was not significantly different from that caused by unspecific bleaching of the cells during image acquisition as shown for fixed control cells (Fig. 2C).
As well as regulating export of APP from the trans-Golgi/ TGN, sorLA may also affect retrograde trafficking of precursor molecules internalized from the cell surface to this compartment. To test this hypothesis, we immunolabeled APP molecules on the surface of CHO-A and CHO-A/S wt cells kept at 4°C to block endocytosis. When the cells were shifted to 37°C, endocytic uptake and subcellular targeting of APP was evaluated using co-localization with Alexa 633-transferrin (endosomes) and GalT (trans-Golgi/TGN) (supplemental Fig. S2). No discernable differences were seen for the appearance of APP in endosomes (after 30 min) or in trans-Golgi/TGN (after 120 min) when comparing cells with (CHO-A/S wt ) and without sorLA (CHO-A), indicating that endocytic uptake and retrograde targeting of APP were not influenced by the receptor. So far, our studies had uncovered a critical function for sorLA as trans-Golgi/TGN-localized receptor controlling APP exit from these compartments. To conceptually prove the ability of sorLA to act as retention factor for APP, we generated a receptor variant that carried four extra amino acid residues (KKLN) at the carboxyl terminus (sorLA kkln ). This di-lysine peptide sequence acts as an ER retention motif in some transmembrane proteins. Accordingly, sorLA kkln mainly resided in the ER and was blocked from exit to the Golgi in stably transfected CHO cells as indicated by subcellular fraction studies (Fig. 3A). Aberrant sequestration of sorLA kkln in the ER caused redistribution of APP to the same compartment as deduced from fractionation studies that uncovered APP in ER and Golgi fractions in cells expressing sorLA wt but only in the ER in sorLA kkln expressing cells (Fig. 3A). Sequestration of sorLA kkln in the ER was also demonstrated by faster electrophoretic mobility in SDS-PAGE (Fig. 3B, upper panel) compared with sorLA wt , indicating the absence of terminal glycosylation. In line with a sorLA kkln -dependent blockade of ER exit of APP, accumulation of immature and decrease of mature forms of the protein was seen in cells expressing sorLA kkln (CHO-A/S kkln ) compared with cells without sorLA (CHO-A) (Fig. 3B, middle panel) (ratio mature/immature APP: 0.43 Ϯ 0.06 versus 0.57 Ϯ 0.03, p ϭ 0.02). Finally, the inability of APP to enter distal secretory compartments harboring secretase activities was confirmed by a 66 Ϯ 4.8% decrease in processing to sAPP␣ (Fig. 3B, lower panel; p Ͻ 0.001) and 95% reduction in conversion to A␤ 40 ( Fig. 3C; p Ͻ 0.0009) that was not seen in CHO-A cells. Similar findings were obtained in SH-SY5Y cells expressing sorLA kkln (supplemental Fig. S3).
A number of cellular mechanisms are known that target proteins to and from the TGN including interaction with sorting adaptors GGA and PACS-1 (18,19). Binding of GGA-1 and -2 to a tetrapeptide motif DVPM in the tail of sorLA has been demonstrated before (4). An acidic cluster that may serve as binding site for PACS-1 is also present in the cytoplasmic receptor domain. To dissect regulatory elements in sorLA that convey TGN targeting, we constructed receptor mutants with point mutations in the tail sequence presumed to abolish adaptor bindings (Fig. 4A). Thus, we generated mutants lacking the acidic cluster (sorLA acidic ) or the GGA binding site (sor-LA gga ), or the entire cytoplasmic domain (sorLA ⌬cd ). Activity of these mutants was compared with sorLA wt .
All mutants were stably expressed in parental CHO-A cells harboring the human APP 695 transgene (Fig. 4B). Remarkably, co-expression with sorLA wt but with no other sorLA variant caused accumulation of mature APP, indicating loss of protective function of sorLA by any of the modifications (Fig. 4B). The inability of mutant receptors to protect APP from processing was not due to impaired interaction with the precursor protein  NOVEMBER 9, 2007 • VOLUME 282 • NUMBER 45 as all sorLA variants were able to co-immunoprecipitate APP (Fig. 4C). However, sorLA gga and sorLA acidic failed to co-immunoprecipitate GGA-1 (Fig. 4D) and PACS-1 (Fig. 4E), respectively, consistent with disruption of the individual adaptor binding sites in the mutants. Inability to interact with both GGA-1 and PACS-1 was also seen for sorLA ⌬cd that lacks the entire receptor tail (Fig. 4, D and E).

SorLA Regulates Processing of APP
Next, we investigated the consequences of the sorLA tail mutations on cellular trafficking of the receptor polypeptide and its target APP. Although all sorLA variants exhibited partial colocalization with APP in the perinuclear region as shown by immunofluorescence microscopy of permeabilized cells (Fig. 5), a clear distinction in two receptor categories was shown by surface staining of cells not permeabilized prior to antibody incubation (Fig. 5, insets). Whereas cell lines expressing sorLA ⌬cd and sorLA acidic exhibited pronounced signals for sorLA and APP on the plasma membrane, little cell surface staining was observed in cells expressing sorLA wt and sorLA gga . Quantification of surface exposure by FACS analysis revealed statistical significant differences (p Ͻ 0.004) for sorLA ⌬cd (51.2 Ϯ 6.8% receptor on cell surface) and sorLA acidic (42.9 Ϯ 3.9%) compared with sorLA wt (20.1 Ϯ 5.4%) (see "Experimental Procedures" for experimental details).
Aberrant surface localization of sorLA ⌬cd and sorLA acidic was also confirmed by subcellular fractionation of CHO cell lines using Western blot analysis (Fig. 6A). As reference, the presence of marker proteins for specific cell compartments was evaluated in parallel (Fig. 6B). In all cell lines, distinct peaks for sorLA were seen in fractions containing markers of ER (glucose-related protein 78, Grp78; fractions 3-4) and Golgi (Golgi SNARE 28, GS28; fractions 4 -11). Whereas only 3% of sorLA wt protein was detected in the peak fractions of the plasma membrane (␤ 1 -integrin; fractions 15-16), accumulation of variants sorLA ⌬cd and sorLA acidic in this compartment was evident (23-28% of total protein). This finding was in line with predominant surface localization of these two mutants seen in immunofluorescence (Fig. 5) and FACS sorting (see above) experiments. Surprisingly, in addition to ER and Golgi peak fractions, sorLA gga was also enriched in fractions 16 and 17 that contained endosomes, but partially also overlapped with the plasma membrane (peak fractions 15 and 16). The ratio of sorLA gga protein mass in endosomal versus plasma membrane fractions was increased to ϳ1:1 compared with other receptor variants (ϳ1:3). Because no obvious difference in subcellular localization between sorLA gga and sorLA wt was seen from the immunofluorescence analysis in Fig. 5, we mapped the localization of both proteins more precisely using markers of various endocytic compartments. A subtle but consistent difference was seen with Rab11, a marker of recycling endosomes, that partially colocalized with sorLA gga but to a lesser extent with the wild type protein (supplemental Fig. S4). Quantification of a total of 700 peripheral vesicles for each cell line that stained positive for sorLA identified 22.0 Ϯ 1.6% co-localization with Rab11 in CHO-A/S wt but 43.6 Ϯ 3.6% in CHO-A/S gga (p ϭ 0.0004). This feature suggested faulty recycling to the plasma membrane (rather than retrograde TGN trafficking) for sorLA gga . Cell surface recycling was supported independently by the appearance of some shedded extracellular domains of sorLA gga in the cell supernatant (supplemental Fig. S5). Typically, the ectodomain is released by proteolysis of sorLA by metalloproteases (20), an effect seen most pronounced for cell surface variants sorLA ⌬cd and sorLA acidic , but not for the wild type form of the receptor under these conditions (supplemental Fig. 5).
Given the distinct distribution of various sorLA mutants in either trans-Golgi/TGN, the recycling compartment, or at the plasma membrane, we investigated the effects of these variants on APP processing. When the subcellular fractions of CHO cells in Fig. 6A were probed for the presence of APP, accumulation of the mature protein species was seen exclusively in Golgi fractions of CHO-A/S wt cells but not in this compartment or any other compartment of CHO-A/S ⌬cd (p Ͻ 0.01; Fig.  6C) or other mutant cell lines (data not shown). This finding confirmed that sorLA wt has the unique property to accumulate mature APP molecules in the Golgi, most likely by extending the transit time of the precursor through this compartment. SorLA mutants that lack the ability to interact with GGA and PACS-1 do not parallel such activity.
We also analyzed how altered localization of sorLA (and APP) may affect the generation of APP processing products, initially focusing on the two intracellular variants, sorLA wt and sorLA gga . Consistent with earlier observations, co-expression of APP with sorLA wt resulted in significantly reduced secretion of sAPP␣ (Fig. 7, A and B, p Ͻ 0.001) and A␤ 40 peptide (Fig. 7C,  p ϭ 0.01), the products of non-amyloidogenic and amyloidogenic processing, respectively, compared with cells expressing APP only (CHO-A). In contrast, sorLA gga caused a reduction in A␤ 40 (p ϭ 0.01) but a distinct increase in sAPP␣ production (p Ͻ 0.05) (Fig. 7, A-C). This observation was independently confirmed by measuring C83 and C99, the CTF of APP that arise as co-products of ␣-secretase and ␤-secretase processing, respectively. The ability of sorLA gga to promote ␣-secretase processing was demonstrated by an increase in C83 as compared with parental CHO-A cells. In contrast, sorLA wt impaired C83 production as shown by a decrease in the respective protein band (Fig. 7D). The amount of C99 was reduced by both sorLA variants, demonstrating that the receptors exert their inhibitory effect on amyloidogenic processing via blockade of ␤-secretase activity (Fig. 7E).
As well as studying the effect of intracellular localization of sorLA on APP processing, we also investigated the consequences of wrongful targeting of the receptor to the cell surface with mutants sorLA ⌬cd and sorLA acidic . Whereas sAPP␣ levels were unchanged in sorLA ⌬cd and sorLA acidic compared with CHO-A (Fig. 8, A and B), both sorLA variants caused a massive increase in A␤ 40 levels as shown by ELISA ( Fig.  8C; sorLA acidic p Ͻ 0.02; sorLA ⌬cd p ϭ 0.03). The stimulatory effect on amyloidogenic processing was also seen by determination of A␤ 42 levels that were below detection limit in parental CHO-A and CHO-A/S wt cells, in agreement with A␤ 42 representing less than 10% of the total amyloid peptide production in CHO cells. In contrast to CHO-A and CHO-A/S wt cells, A␤ 42 was easily detectable in CHO-A/S ⌬cd and CHO-A/S acidic (Fig. 8D). An increase in A␤ production was independently confirmed by Western blot analysis that demonstrated substantial amounts of the peptide in supernatants from CHO-A/S ⌬cd and CHO-A/S acidic , but not CHO-A or CHO-A/S wt cell lines (supplemental Fig. 6). Finally, enhanced amyloidogenic activity with sorLA cell surface variants was also documented by detection of significant amounts of sAPP␤ (the co-product of ␤-secretase cleavage) in cell media of the latter cell lines compared with CHO-A. In contrast, levels of sAPP␤ in CHO-A/S wt cells were reduced compared with CHO-A in accordance with the protective function of the wild type receptor (supplemental Fig. 6).
In conclusion, the subcellular localization of sorLA variants and their distinct effects on APP processing are summarized in Table 1. These data have identified molecular mechanisms regulating intracellular trafficking of the receptor; and they have substantiated the physiological relevance of this sorting pathway in determination of APP processing fates in the cell types analyzed here.

DISCUSSION
Although a role for sorLA as suppressor of APP proteolysis has been reported before (6, 10), the molecular mechanism of  1-5). For control, immunoprecipitations were also performed in cells lacking recombinant sorLA (CHO-A; lane 6) or in CHO-A/S wt cells without specific IgG (non-IgG control; lane 7). A weak background signal for APP immunoprecipitation in CHO-A was due to low levels of endogenous sorLA expression in parental CHO cells. D, cells expressing sorLA wt , sorLA gga , and sorLA ⌬cd were transfected with an expression construct for GFP-tagged GGA-1 and immunoprecipitated using anti-sorLA IgG. Panel Input represents Western blots for sorLA and GFP-GGA-1 in cell extracts prior to immunoprecipitation. Panel IP-sorLA is a Western blot analysis for sorLA and GFP-GGA-1 in fractions immunoprecipitated with anti-sorLA IgG (lanes 1-3). Co-precipitation of GFP-GGA-1 is only seen with sorLA wt (lane 1). As negative control, immunoprecipitations were performed in CHO-A/S wt cells without specific IgG (lane 4). E, cells expressing sorLA wt , sorLA acidic , and sorLA ⌬cd were transfected with an expression construct for HA-tagged PACS-1 and immunoprecipitations performed using anti-HA IgG. Panel Input represents Western blots for sorLA and HA-PACS-1 in cell extracts prior to precipitation. Panel IP-PACS1 represents Western blot analysis for HA-PACS-1 and sorLA in fractions immunoprecipitated with anti-HA antibody (lanes 1-3). Co-precipitation with HA-PACS-1 is only seen for sorLA wt (lane 1). Immunoprecipitations were also performed in CHO-A/S wt cells without specific IgG (lane 4). action remained unclear. Now, our studies characterized possible trafficking routes of the sorting receptor, and their dependence on interaction with adaptors GGA and PACS-1. They identified the functional relevance of this trafficking pathway for regulation of APP export from the trans-Golgi/TGN. Most importantly, these studies indicate aberrant trafficking of sorLA as a cell biological mechanism leading to accelerated A␤ production, suggesting possible scenarios that may underlie senile plaque formation in patients with inherited Sorla gene variants (15,16).
Using immunolocalization and live cell imaging we demonstrated that sorLA predominantly resides in trans-Golgi/TGN where it controls release of mature APP to the cell surface, a step considered crucial for conversion into both amyloidogenic and non-amyloidogenic products (21,22). Consequently, sor-LA wt -mediated accumulation of APP in the trans-Golgi/TGN results in impairment of ␣and ␤-secretase cleavage, and a concomitant decrease in A␤ levels. Confinement to the trans-Golgi/TGN is mainly achieved by impairing transition of nascent APP molecules en route through the biosynthetic pathway to the cell surface (Fig. 2). A role for sorLA in re-routing internalized precursor molecules from early endosomes back to the TGN seems less likely as retrograde trafficking of APP is not affected by the presence of the receptor (supplemental Fig. S2).
The unique ability of sorLA to regulate APP routing was confirmed by targeting the receptor to the plasma membrane or to the ER, causing accumulation of APP in the very same compartments (Figs. 3 and 5). No such effect on subcellular localization of unrelated transmembrane proteins such as low density lipoprotein receptor-related protein 1 (LRP1) and interleukin-2 receptor is seen in CHO and SH-SY5Y cells (6), further supporting the specificity of the sorLA and APP interactions.  The central role of the Golgi in APP metabolism is well appreciated as processing of APP requires efficient transition of the precursor through this organelle (21,22) and experimentally disrupting Golgi transition blocks processing (23,24). The relevance of sorLA-mediated Golgi retention of APP was confirmed by the distinct effects that various sorLA trafficking mutants had on amyloid peptide production. Overall, the ability of some mutants to increase but others to decrease processing indicated that impairment of APP processing by sorLA wt was no artifact of protein overexpression, a concern commonly raised with transfection studies. Thus, the inability to retain APP in the Golgi and its enhanced shunt to the plasma membrane seen with cell surface variants of sorLA caused a massive increase in A␤ and sAPP␤, supporting the concept that release to the plasma membrane and entry into endocytic pathways is a major determinant of amyloid peptide formation (23,24). Because sorLA wt does not directly promote endocytosis of APP from the cell surface as shown before (11), our findings suggest that reducing the transition time of APP in trans-Golgi/TGN compartments is sufficient to profoundly enhance amyloidogenic processing. The increase in A␤ and sAPP␤ levels in sorLA ⌬cd and sorLA acidic is impressive and confirmed by independent measurements of the products by ELISA (Fig. 8) and Western blot (supplemental Fig. S6). When interpreting these results one has to bare in mind that amyloidogenic processing in the parental CHO-13-5-1 line is low as it lacks expression of LRP1, a promoter of amyloidogenic processing (25,26). Thus, manipulations that alter processing efficiency may result in massive relative increases in A␤ as observed here (10 -20-fold). As a reference, the average level of A␤ 40 production in sor-LA acidic (ϳ1 ng/mg of cell protein) is similar to the one reported for this line before (0.3 ng/mg of cell protein) (27).
Localization of sorLA to the trans-Golgi/TGN is critically dependent on two motifs in its cytoplasmic domain, a tetrapeptide DVPM and an acidic cluster. Binding of GGA-1/-2 to the tetrapeptide sequence element has been shown in yeast twohybrid screens before, but its relevance for trafficking of sorLA was not examined (4). Binding of PACS-1 to the acidic cluster in sorLA has first been uncovered in this report (Fig. 4E). Functional interaction with PACS-1 seems to be a major determinant of sorLA targeting as mutants lacking the respective binding site behave similar to receptors lacking the entire cytoplasmic domain. The role of PACS-1 in proper localization of sorLA to the late-Golgi/TGN is in line with a similar function for this adaptor in trafficking of other TGN-resident proteins such as furin (28). As well as PACS-1 binding, interaction with GGA-1/-2 is also required for functional expression of sorLA. Based on the trafficking route of other proteins targeted by GGAs (e.g. mannose 6-phosphate receptors, sortilin) (29 -31), one may speculate that lack of GGA-1/-2 interaction infringes the ability of sorLA gga to relocate from early endosomes back to the TGN. This hypothesis is supported by the abnormal appearance of the mutant receptor in recycling endosomes (supplemental Fig. S4) and at the cell surface (supplemental Fig. S5). In cells expressing sorLA gga , APP  may also be retained in recycling compartments and at the plasma membrane, where it is preferentially processed by ␣-secretase as suggested by the increase in sAPP␣ products in CHO-A/S gga (Fig. 7, A and B). Remarkably, previous studies have implicated GGA-1 in APP processing inasmuch as reduced expression of the adaptor enhanced, whereas overexpression of the adaptor decreased A␤ production (32). These effects were independent of a direct interaction between APP and GGA, suggesting the presence of another factor (such as sorLA) that may link both components (33). Although the influence of sorLA variants on APP processing has mainly been tested in CHO cells in this study, identical effects of wild type and mutant forms of the receptor on trafficking of APP in neurons (supplemental Figs. 1 and 3) strongly suggest that similar pathways are operable in neuronal and non-neuronal cell types.
Previously, the crucial role of APP trafficking for processing fates has been established by site-directed mutagenesis of the precursor artificially targeting APP to various organelles, thereby altering the extent of A␤ production (23,34). Because no naturally occurring mutations in APP affecting protein localization have been associated with the risk of AD so far, the relevance of aberrant APP trafficking for onset and progression of sporadic AD in patients remained uncertain. Our studies provide a mechanistic framework for this hypothesis by uncovering a neuronal sorting receptor sorLA that regulates the normal transport of APP. Changes to this pathway that target the sorting receptor to a different cellular compartment dramatically alter the balance between amyloidogenic and non-amyloidogenic processing as shown for various receptor mutants. One can envision a situation where sequence variations that affect adaptor binding or overall expression levels of sorLA in individuals may have dramatic consequences for APP processing kinetics and onset of sporadic AD.