Research ArticleAlzheimer’s Disease

Lysosomal Sorting of Amyloid-β by the SORLA Receptor Is Impaired by a Familial Alzheimer’s Disease Mutation

See allHide authors and affiliations

Science Translational Medicine  12 Feb 2014:
Vol. 6, Issue 223, pp. 223ra20
DOI: 10.1126/scitranslmed.3007747


SORLA/SORL1 is a unique neuronal sorting receptor for the amyloid precursor protein that has been causally implicated in both sporadic and autosomal dominant familial forms of Alzheimer’s disease (AD). Brain concentrations of SORLA are inversely correlated with amyloid-β (Aβ) in mouse models and AD patients, suggesting that increasing expression of this receptor could be a therapeutic option for decreasing the amount of amyloidogenic products in affected individuals. We characterize a new mouse model in which SORLA is overexpressed, and show a decrease in Aβ concentrations in mouse brain. We trace the underlying molecular mechanism to the ability of this receptor to direct lysosomal targeting of nascent Aβ peptides. Aβ binds to the amino-terminal VPS10P domain of SORLA, and this binding is impaired by a familial AD mutation in SORL1. Thus, loss of SORLA’s Aβ sorting function is a potential cause of AD in patients, and SORLA may be a new therapeutic target for AD drug development.


Sortilin-related receptor with A-type repeats (SORLA; also known as SORL1 or LR11) is a type 1 membrane protein widely expressed in neurons of cortex and hippocampus (1, 2). SORLA is best recognized for its role as a neuronal sorting receptor for the amyloid precursor protein (APP) [reviewed in (3)]. APP is a major etiologic agent in Alzheimer’s disease (AD) that is proteolytically processed to neurotoxic peptides 40 to 42 amino acids in length, termed Aβ40 and Aβ42, respectively. SORLA interacts with newly synthesized APP molecules in the Golgi and prevents trafficking of this precursor into cellular compartments where secretases reside. Consequently, overexpression of SORLA in cultured cells reduces amyloidogenic processing, whereas loss of the receptor in genetically engineered mice enhances amyloid-β (Aβ) peptide production and senile plaque deposition (46). The relevance of SORLA for the neurodegenerative process in AD patients is supported by studies demonstrating reduced expression of the receptor in the brains of some individuals with sporadic AD (7). Also, genetic analysis strongly supports association of SORL1 variants with the risk of developing sporadic AD (810). Remarkably, recent findings suggest that missense and nonsense mutations in SORL1 may underlie autosomal dominant early-onset forms of AD not linked to mutations in APP, PSEN1, or PSEN2 (11).

On the basis of studies in cell lines, mouse models, and patients, SORLA mutations are considered a major risk factor for AD. Because concentrations of this receptor are inversely correlated with amyloidogenic processing rates, ways to up-regulate receptor expression may represent a new therapeutic approach in AD. Previous studies in mouse models have solely addressed the pathophysiological consequences of reduced receptor concentrations for acceleration of Aβ production and deposition, a situation shared by patients with SORL1 risk alleles (12, 13). However, experiments addressing the presumed beneficial effects of elevated SORLA in vivo have not yet been reported. Here, we generated a new mouse strain that overexpresses SORLA in the brain, and we explored the relevance of increased SORLA for amyloidogenic processing.


Transgenic mice exhibit neuronal overexpression of SORLA

We used homologous recombination in embryonic stem cells to introduce a human SORLA complementary DNA (cDNA) into the murine Rosa26 gene locus. Expression of the cDNA is driven by the cytomegalovirus early enhancer/chicken β-actin promoter (CAG) element after Cre recombinase–mediated excision of a preceding transcription stop element (fig. S1). Mice carrying the SORLA cDNA were crossed with the Cre deleter strain of mice to generate animals carrying one expressible copy of the SORLA transgene in addition to the endogenous Sorl1 gene locus (Rosa26Tg/+). As shown by Western blotting (Fig. 1A) and by enzyme-linked immunosorbent assay (ELISA) (Fig. 1B), SORLA concentrations in the cortex and hippocampus of adult Rosa26Tg/+ animals were elevated about fourfold compared to control mice that expressed SORLA from the endogenous Sorl1 locus only (Rosa26+/+) (P < 0.001).

Fig. 1. Characteristics of transgenic mice overexpressing SORLA.

(A) Western blot analysis showing overexpression of SORLA in the cortex and hippocampus of 8-week-old mice carrying one copy of the human SORLA transgene (Rosa26Tg/+, lanes 5 and 6) compared to wild-type (WT) animals (Sorl1+/+, lanes 3 and 4). Tissues from Sorl1−/− mice (lanes 1 and 2) were used as a negative control for SORLA immunoreactivity. Detection of Na/K ATPase served as a loading control. The arrowhead highlights the protein band corresponding to SORLA in hippocampal tissues. (B) Quantification of SORLA by ELISA in cortex and hippocampus of 8-week-old Rosa26+/+ and Rosa26Tg/+ mice. Mean values of duplicate measurements of individual mice are shown. ***P < 0.001, Student’s t test. (C) Immunodetection of SORLA (red), astrocyte marker glial fibrillary acidic protein (GFAP) (green), and neuronal marker NeuN (blue) on free-floating sections of cortex and hippocampus of 8- to 12-week-old mice either WT for the murine Sorl1 locus (Sorl1+/+) or homozygous mutant for Sorl1 but carrying one copy of the human SORLA transgene (Rosa26Tg/+, Sorl1−/−). The insets in the merged micrographs indicate localization of SORLA to intracellular vesicles in NeuN-positive cells of the respective tissues. No SORLA immunoreactivity was seen in Sorl1−/− mice used as a negative control. To better appreciate the distinct localization of SORLA to the neuronal cell layer, the hippocampal section is shown both with single SORLA staining and as a merged image. Scale bar, 50 μm.

To achieve even higher concentrations of SORLA in the brain, we bred Rosa26Tg/+ mice to generate mice carrying two copies of the transgene (Rosa26Tg/Tg). Although we are able to derive adult Rosa26Tg/Tg mice, a reduced Mendelian ratio of 10% compared to the expected 25% suggested impaired viability of this genotype (fig. S2A). Further analysis at embryonic day 13.5 and at the newborn stage documented normal embryonic development but perinatal lethality of about half of the Rosa26Tg/Tg newborns (fig. S2A). The reason for the perinatal lethality of some animals remained unclear as they appeared grossly normal upon external inspection or histological examination. Also, the incidence of death was not related to individual SORLA expression as the concentrations of the receptor in brain extracts were similar in Rosa26Tg/Tg newborns that died or that survived (fig. S2B). To circumvent any problems with impaired viability of Rosa26Tg/Tg, we focused our subsequent analysis of this genotype on newborn mice. For analysis of adults, we used Rosa26Tg/+ and compared them to Rosa26+/+ animals. Because SORLA concentrations were only modestly increased in Rosa26Tg/Tg mice compared to Rosa26Tg/+ mice (fig. S2B), Rosa26Tg/+ mice were deemed sufficient to test the consequences of high receptor activity for amyloidogenic processes in the adult brain.

Neuronal overexpression of SORLA reduces Aβ in the brain

Initially, we confirmed proper expression and subcellular localization of human SORLA in the brains of transgenic mice. To do so, we crossed Rosa26Tg/+ animals with Sorl1−/− mice to eliminate any background signal from the endogenous murine receptor. Human SORLA localized to neurons in cortex and hippocampus of adult (Rosa26Tg/+; Sorl1−/−) mice with a vesicular pattern that was typical of the murine receptor in wild-type Sorl1+/+ animals (Fig. 1C). No obvious SORLA immunoreactivity was seen in glia. Neuronal overexpression of SORLA in Rosa26Tg/+ and Rosa26Tg/Tg newborns significantly reduced the concentrations of murine Aβ40 and Aβ42 peptides compared to animals expressing the endogenous receptor only (Rosa26+/+) (Fig. 2, A and B; P < 0.001). The reduction in Aβ peptide concentrations was comparable in Rosa26Tg/+ and Rosa26Tg/Tg newborns, substantiating the notion that carrying two SORLA transgenes does not increase receptor expression significantly compared to carrying one transgene. Overall, SORLA concentrations were inversely correlated with Aβ concentrations in these animals as shown by linear regression analysis (Fig. 2C). Overexpression of SORLA did not affect the concentration of murine soluble sAPPα compared to wild-type animals (Fig. 2D).

Fig. 2. SORLA overexpression decreases murine Aβ concentrations in the mouse brain.

(A and B) Murine Aβ40 (A) and Aβ42 (B) were determined by ELISA in brain extracts from newborn (P1) mice either Rosa26+/+ (murine SORLA), Rosa26Tg/+ (murine SORLA and one human SORLA transgene), or Rosa26Tg/Tg (murine SORLA and two human SORLA transgenes). (C) Linear regression analysis documented inverse correlation of SORLA with Aβ concentration in P1 mouse brains. (D) Concentrations of soluble sAPPα were identical in P1 brain extracts of Rosa26+/+, Rosa26Tg/+, and Rosa26Tg/Tg mice as determined by ELISA. sAPPα and SORLA were measured in duplicate. Aβ40 and Aβ42 were single measurements from individual mice. ***P < 0.001, Student’s t test.

Next, we confirmed the effect of SORLA overexpression on APP metabolism in an established mouse model of AD (the PDAPP mouse strain) (14). As documented for murine APP, overexpression of SORLA specifically reduced human Aβ40 and Aβ42 (Fig. 3, C to E; P < 0.001) but not human sAPPα and sAPPβ (Fig. 3, A and B) in newborn PDAPP mice. The total concentration of human APP was not affected by SORLA concentrations (Fig. 3F). Normal concentrations of total APP and soluble APP products suggested unimpaired precursor processing rates in neonates overexpressing SORLA compared to wild-type animals with normal SORLA expression. A specific decrease in Aβ40 and Aβ42 in cortex (Aβ40: P < 0.001; Aβ42: P < 0.05) and hippocampus (Aβ40: P < 0.01; Aβ42: P < 0.05) was also seen in adult Rosa26Tg/+ mice compared to controls (Fig. 4, A and B). Similar to newborn mice, soluble APP concentrations did not change in adult Rosa26Tg/+ animals, although there was a tendency for lower sAPPα and sAPPβ concentrations in the cortex (Fig. 4, C and D). Finally, unimpaired activity of α- and β-secretases in adult Rosa26Tg/+ mice was also shown by quantification of C-terminal fragments (CTFs) C83 and C99, respectively (fig. S3). Together, these data indicated a distinct impact of receptor activity on Aβ catabolism in a mouse model that overexpresses SORLA.

Fig. 3. SORLA overexpression decreased human Aβ in newborn PDAPP mice.

(A to D) Concentrations of human sAPPα (A), sAPPβ (B), Aβ40 (C), and Aβ42 (D) were determined by ELISA in brain extracts from newborn (P1) PDAPP mice carrying the human APPV717F variant. The animals were Rosa26+/+ (murine SORLA), Rosa26Tg/+ (murine SORLA and one human SORLA transgene), or Rosa26Tg/Tg (murine SORLA and two human SORLA transgenes). Reduced concentrations of Aβ40 and Aβ42, but not sAPPα and sAPPβ, were seen in mice carrying one or two copies of the human SORLA transgene (Rosa26Tg/+ and Rosa26Tg/Tg) compared to WT mice (Rosa26+/+) with normal concentrations of mouse SORLA. (E) Linear regression analysis documented inverse correlation of SORLA with human Aβ concentration in P1 mouse brains. (F) Human APP was unchanged in P1 brain extracts of Rosa26Tg/+ and Rosa26Tg/Tg mice compared to Rosa26+/+ animals as determined by ELISA. Mean values of duplicate measurements of individual mice are shown. ***P < 0.001, Student’s t test.

Fig. 4. SORLA overexpression decreases human Aβ concentrations in adult PDAPP mice.

(A to D) Human Aβ40 (A), Aβ42 (B), sAPPα (C), and sAPPβ (D) were determined by ELISA in cortical and hippocampal extracts from 20-week-old PDAPP mice either Rosa26+/+ or Rosa26Tg/+. Reduced Aβ40 and Aβ42 concentrations, but not sAPPα and sAPPβ concentrations, were seen in mice carrying one copy of the human SORLA transgene (Rosa26tg/+) compared to WT controls (Rosa26+/+). *P < 0.05; **P < 0.01; ***P < 0.001. Student’s t test was used to perform statistical analysis. Grubbs’ test was applied to detect significant outliers.

Aβ degradation pathways are not altered in mice overexpressing SORLA

A number of pathways have been implicated in catabolism of Aβ in the brain, including enzymatic degradation in the extracellular space (15, 16), clearance by endocytic receptors (17), or transport across the blood-brain barrier (18, 19). Levels of mRNA or protein of the Aβ-degrading enzyme neprilysin and insulin-degrading enzyme (IDE) were not changed in newborn Rosa26Tg/+ or Rosa26Tg/Tg animals compared to wild-type control mice (Rosa26+/+) (fig. S4, A to C). Similarly, mRNA and protein concentrations of apolipoprotein E (APOE) (fig. S4, G to I) or of the neuronal receptors sortilin and LRP1 (fig. S4, D to F) were the same in all three genotypes. We did see a modest reduction in neprilysin concentrations in Rosa26Tg/Tg animals (fig. S4, A and B). However, reduced neprilysin was unlikely to be responsible for the decreased Aβ concentration in Rosa26Tg/Tg animals because lower concentrations of neprilysin would be expected to increase Aβ. Plasma concentrations of Aβ40 were identical in adult PDAPP mice crossed with either Rosa26+/+ or Rosa26Tg/+ mice (fig. S5), arguing against enhanced transport of the peptide across the blood-brain barrier as the underlying cause of reduced Aβ in adult Rosa26Tg/+ animals.

Last, we explored the possibility that SORLA activity, directly or indirectly, may affect clearance of Aβ peptides from the brain interstitial fluid (ISF). Several receptors have been implicated in endocytic clearance of Aβ peptides in free form or when bound to carriers such as APOE (17). To test this concept, we applied in vivo microdialysis experiments to dynamically assess the baseline concentrations of soluble ISF Aβ as well as the ISF Aβ half-life. Using established protocols (20), we first sampled hippocampal ISF for 6 hours in freely moving animals to determine a stable baseline concentration of the peptide. Subsequently, a potent γ-secretase inhibitor (compound E) was administered intraperitoneally, and the elimination kinetics of ISF Aβ was determined for 15 hours. Unexpectedly, we found that basal concentrations of ISF Aβ40 were unchanged in the brain parenchyma of adult PDAPPRosa26Tg/+ mice compared to PDAPPRosa26+/+ control animals (Fig. 5A). In addition, the rate of clearance of soluble ISF Aβ40 was identical in both genotypes (Fig. 5, B and C). Although in vivo microdialysis experiments are not able to distinguish between the relative contributions of cellular clearance and blood-brain barrier export to Aβ concentrations in ISF, our data exclude a role for SORLA overexpression in either pathway.

Fig. 5. SORLA overexpression does not alter the catabolism of brain ISF Aβ.

(A) Basal interstitial (ISF) Aβ40 concentrations in 7-month-old PDAPPRosa26+/+ mice (n = 7) and PDAPPRosa26Tg/+ mice (n = 8). (B) Clearance of ISF Aβ40 in PDAPPRosa26+/+ (n = 7) and PDAPPRosa26Tg/+ (n = 7) mice. A stable 6-hour baseline period was obtained, followed by intraperitoneal injection of the γ-secretase inhibitor compound E to halt Aβ production. (C) Aβ40 concentrations during the elimination phase [as in (B)] were transformed with the common logarithm. Log-transformed values were fit with linear regression.

SORLA binds to soluble Aβ through its VPS10P domain

Unchanged concentrations of ISF Aβ40 in mice overexpressing SORLA (Fig. 5) contrasted with the decreased concentrations of the peptide seen when measuring Aβ concentrations in cortical and hippocampal extracts of newborn (Figs. 2 and 3) and adult mice (Fig. 4). This observation suggested that the differences seen in brain extracts in these mouse models mainly reflected cell-associated rather than soluble peptide molecules.

On the basis of the ability of related VPS10P domain receptors, such as sortilin, to direct intracellular target proteins for lysosomal degradation (2124), we hypothesized that a function for SORLA could be intracellular catabolism of newly produced Aβ peptides. Accordingly, we tested the ability of SORLA to directly bind to soluble Aβ peptides using a fluorescence polarization assay that detects fast-dissociating protein-peptide interactions. To this end, several peptides including Aβ40 were labeled with Alexa Fluor 488 via a cysteine residue added at the N terminus (Fig. 6A) and mixed with varying concentrations of the receptor protein. Because the VPS10P domain had been recognized as a binding site for peptides in sortilin previously (25), we focused our analysis on potential Aβ binding to this extracellular domain in SORLA. As shown in Fig. 6B, fluorescently labeled Aβ40 showed higher fluorescence polarization values when mixed with purified SORLA VPS10P domain, indicating a physical interaction between the two proteins (Fig. 6B, open circles). Curve fitting of the concentration-dependent binding data revealed that the KD (dissociation constant) value for the binding between SORLA VPS10P and Aβ40 was 106 ± 5.6 nM (n = 3). Using a competition assay with a series of shorter overlapping peptides, we mapped the major SORLA-binding segment to an internal 15–amino acid sequence in Aβ40 spanning amino acid residues 6 to 20 (Fig. 6A and fig. S6). A peptide containing this sequence (termed Aβ6–20) bound to the SORLA VPS10P domain with a concentration dependency indistinguishable from that of Aβ40 (Fig. 6B, diamond symbols). The binding behavior of SORLA VPS10P toward Aβ40 and Aβ6–20 was similar to that for the internal propeptide of SORLA and neurotensin, two established peptide ligands of this receptor (Fig. 6B). In contrast to the situation with SORLA, the VPS10P domain of the related receptor sortilin bound to neurotensin but failed to bind to Aβ40 or Aβ6–20 (Fig. 6C). We also measured the peptide binding ability of the SORLA VPS10P under various pH conditions. A pH dependency could not be tested for Aβ40 directly because this peptide aggregates under low pH conditions. However, as shown in Fig. 6, D and E, binding of both Aβ6–20 and the SORLA propeptide occurred only at a pH above 5.0, documenting a pH sensitivity for these interactions. Our results underscored the ability of SORLA to act as an Aβ binding receptor and the effect of low pH (as in endosomal/lysosomal compartments) on the disruption of this ligand-receptor interaction.

Fig. 6. Direct and pH-dependent binding of Aβ to the VPS10P domain of SORLA.

(A) Peptides used in the binding assay. Each peptide carries an extra Cys residue at the N terminus used for labeling with Alexa Fluor 488–maleimide. (B and C) Binding of the peptides to the VPS10P domain of SORLA (B) and sortilin (C) assessed by fluorescence polarization assay. Fixed concentrations of labeled peptides (100 nM) were incubated with increasing concentrations of purified VPS10P proteins. From each point, fluorescence polarization values obtained in the absence of the VPS10P domain were subtracted, and the binding was expressed as the net fluorescence polarization increase. (D and E) pH dependency of peptide binding by the SORLA VPS10P domain. Fluorescence polarization values of the fluorescent peptide (100 nM) and SORLA VPS10P domain (1 μM) were measured in buffer with varying pH, and the binding was expressed as net fluorescence polarization increase after the blank subtraction. Fluorescent properties of Alexa Fluor 488 did not change significantly within this pH range.

SORLA activity increases lysosomal targeting and degradation of nascent Aβ peptides

To test the ability of SORLA to affect the catabolism of intracellular Aβ peptides, we generated neuroblastoma SH-SY5Y cell lines stably expressing a human wild-type APP695 transgene in the absence (SY5Y-A) or presence (SY5Y-A/S) of a human SORLA transgene. Replicate layers of SY5Y-A and SY5Y-A/S cells were treated with the γ-secretase inhibitor DAPT to block de novo production of Aβ peptides. Subsequently, turnover of intracellular Aβ peptides was determined by ELISA in cell extracts. As seen in Fig. 7A, intracellular concentrations of Aβ decreased significantly faster in SY5Y-A/S cells compared to SY5Y-A cells over a 4-hour time period (P < 0.01). In addition, subcellular fractionation studies to enrich for lysosomes documented twofold higher levels of Aβ associated with this organelle in SY5Y-A/S compared to SY5Y-A cells (Fig. 7C; P < 0.05). This effect was specific for Aβ because the lysosomal concentration of acid phosphatase, a marker for this compartment, was unaltered when comparing the two cell lines (Fig. 7B). Combined, these data supported the notion of enhanced lysosomal targeting and degradation of Aβ peptides in neuroblastoma cells overexpressing SORLA.

Fig. 7. SORLA increases lysosomal targeting and catabolism of nascent Aβ.

(A) SH-SY5Y cells overexpressing human APP695 with (SY5Y-A/S) or without (SY5Y-A) SORLA were treated with 5 μM DAPT to block γ-secretase activity. At the indicated time points, cells were lysed and the concentration of Aβ in cells was determined by ELISA. **P < 0.01; ***P < 0.001. Data are means ± SEM of 8 to 12 replicates expressed as percent of time point 0. Differences between the clones were determined for the individual time points using Student’s t test. The inset documents expression of SORLA, APP, and tubulin (loading control) in the cell extracts by Western blot analysis. (B and C). Lysosomes were isolated from SY5Y-A and SY5Y-A/S cells by Percoll gradients, and the percent distribution of activity of the lysosomal enzyme marker acid phosphatase (B) and of Aβ40 (C) in the individual fractions was determined. Acid phosphatase activity in each fraction is given as a percent of total activity in all fractions combined (set to 100%). Aβ concentrations are given as percent recovery from the total amount of Aβ applied for fractionation. Identical amounts of acid phosphatase activity were seen in peak lysosomal fraction 9 of both cell clones, but Aβ40 concentrations were twofold higher in SY5Y-A/S compared with SY5Y-A cells (21.11 ± 0.31% versus 11.31 ± 1.72%; *P < 0.05). The data are means of three replicates ± SEM (Student’s t test).

The ability of SORLA to bind to Aβ is disrupted by a familial AD mutation

Recently, several missense mutations have been identified in patients with early-onset AD of unclear etiology. Of particular note, the one mutation showing segregation in a pedigree localized within the VPS10P domain (G511R), suggesting impaired binding of Aβ to SORLA as an underlying molecular defect in this family (11). To explore this possibility, we developed a simple assay to evaluate Aβ binding to wild-type SORLA and to a mutant form of the receptor containing the VPS10P domain mutation using an alkaline phosphatase (AP)–Aβ fusion protein. Wild-type and mutant variants of the SORLA VPS10P domain were expressed at similar levels in human embryonic kidney (HEK) 293T cells (Fig. 8A). However, in contrast to the wild-type VPS10P domain, the G511R variant failed to bind to Aβ (Fig. 8B; P < 0.001). To ultimately test the impact of this familial AD mutation on the intracellular catabolism of Aβ peptides, we assessed the half-life of intracellular Aβ in SY5Y clones expressing full-length versions of the wild-type (SY5Y-A/S) or the mutant receptor variants (SY5Y-A/G511R) (Fig. 8C). In line with an inability to bind to Aβ, the turnover of the intracellular peptide in cells expressing G511R was much lower than in cells expressing the wild-type receptor (SY5Y-A/S) (P < 0.001), and was not significantly different from cells lacking this protein (SY5Y-A).

Fig. 8. Familial AD mutation in SORLA abolishes binding to Aβ and affects intracellular Aβ catabolism.

(A) Supernatants from HEK293T cells transfected with constructs encoding the VPS10P domain (carrying a TARGET tag) of the WT or the G511R variant of SORLA were immunoprecipitated and separated by SDS–polyacrylamide gel electrophoresis, followed by staining with the Oriole fluorescent gel stain. (B) An AP–Aβ1–32 fusion protein was transiently expressed in HEK293T cells and incubated with the WT or G511R SORLA VPS10P domain captured on antibody-coated beads. Bound AP–Aβ1–32 was quantified by determining the AP activity associated with the beads (n = 3; mean ± SE; ***P < 0.001, Student’s t test). (C) Replicate cultures of SH-SY5Y cells overexpressing human APP695 without SORLA (A) or with the full-length versions of the WT (A/S) or G511R variant (A/G) receptor were treated with 5 μM DAPT to block γ-secretase activity. At the indicated time points, cells were lysed and the concentration of Aβ in cells was determined by ELISA. Data are means ± SEM of eight to nine replicates expressed as percent of time point 0. Difference between the clones was determined for the different time points using one-way analysis of variance (ANOVA) followed by Bonferroni’s multiple comparison test. The clearance of Aβ in SY5Y-G511R cells (blue line) was much lower than in SY5Y-A/S cells (***P < 0.001) and comparable to that in cells lacking SORLA (SY5Y-A). The inset documents expression of SORLA, APP, and tubulin (loading control) in cell extracts by Western blot analysis.


SORLA is a negative regulator of Aβ production, and low concentrations of this receptor may contribute to the onset and progression of neurodegeneration in individuals carrying SORL1 gene risk variants (7). In support of this hypothesis, sequence variations in SORL1 have been shown to affect the rate of gene transcription or translation, resulting in lower receptor concentrations in the brains of carriers of the risk allele (26). Conceptually, increasing SORLA activity could represent a therapeutic option to decrease the amount of amyloidogenic products in the brain. Here, we have generated a transgenic mouse model overexpressing SORLA to provide proof of concept for such an approach. Overexpression of SORLA in the brain of our mouse model was restricted to neurons that also expressed this receptor endogenously. Despite the use of the ubiquitous CAG promoter, no obvious SORLA immunoreactivity was seen in glia that do not express the murine receptor in vivo (Fig. 1C). Possibly, some cell-intrinsic mechanisms may be required to sustain stable expression of this receptor in distinct cell types (such as neurons) in vivo.

Mice heterozygous for the SORLA transgene were viable and fertile and did not present any obvious pathological phenotypes despite four- to sevenfold higher receptor concentrations. We observed an impaired viability in mice homozygous for the transgene; the reasons for this still require clarification. Increasing receptor concentrations above a certain threshold may have resulted in adverse effects. Although we failed to see a correlation between brain SORLA concentrations and perinatal death in Rosa26Tg/Tg newborn mice, overexpression of the receptor in other tissues, not considered here, may have had an impact. Still, pharmacological interventions are unlikely to achieve levels of stable overexpression as high as in Rosa26Tg/Tg animals (ninefold).

In line with a protective function for SORLA in amyloidogenic processes, increasing neuronal receptor concentrations reduced total brain Aβ concentrations by threefold. This effect was seen for the murine and the human peptides in cortex and hippocampus of newborn (Figs. 2 and 3) and adult (Fig. 4) mice. Whereas a beneficial effect of SORLA overexpression on reducing total Aβ concentrations may have been anticipated on the basis of studies in cultured cells (4, 27), the molecular mechanism of receptor overactivity in our mouse model was surprising. A large body of work in cell lines and receptor-deficient mice has firmly established a role for SORLA as a sorting receptor that mediates intracellular transport and processing of APP [reviewed in (3)]. The pathophysiological relevance of this sorting function has received independent support from studies on the cytosolic adaptor protein complex retromer. This adaptor complex guides trafficking of SORLA through the intracellular compartments of neurons, and the low activity of the retromer in AD patients is believed to cause SORLA malfunction and, consequently, enhanced APP processing (2830).

Given the above, it was surprising to see that overexpression of SORLA in vivo did not significantly affect cellular APP processing but did affect total brain Aβ concentrations compared to wild-type mice. Given that soluble peptide concentrations in the adult ISF (Fig. 5) were unchanged, SORLA seemed to mainly affect the pool of cell-associated peptides in this model. The likely mechanism was traced to the ability of SORLA to bind to Aβ (Fig. 6) and to promote lysosomal targeting (Fig. 7C) and catabolism (Fig. 7A) of this peptide in neuronal cells. Because Aβ concentrations decreased considerably faster in cells expressing SORLA compared to parental SY5Y cells (Figs. 7A and 8C), the pool of peptides measured in cell extracts was likely to be derived from endogenous production rather than cellular uptake.

It is well appreciated that Aβ accumulates in vulnerable neurons in the AD brain and that this phenomenon precedes deposition of senile plaques in mouse models and in AD patients [reviewed in (31)]. Some studies suggest that intracellular Aβ oligomers may be a major cause of synaptic dysfunction seen in the diseased brain (3234). However, it is unclear whether intraneuronal Aβ mainly derives from endogenous production or from endocytic uptake, or both (31). Together, our data from transgenic mice and neuronal cell lines suggest a model whereby a fraction of Aβ produced in neurons is directly targeted to lysosomes for catabolism, and that this catabolic pathway depends on the activity of the Aβ receptor SORLA. In support of this hypothesis, SORLA has been shown to interact with a number of cytosolic adaptors including PACS1 and the retromer complex that sort cargo between Golgi and late endosomes/lysosomes (27, 28, 35, 36). Also, SORLA has been shown to bind to the lipolytic enzyme lipoprotein lipase in the biosynthetic pathway of neurons and to direct it for lysosomal degradation (24).

Previous studies have unambiguously documented an inverse correlation between SORLA concentrations and APP processing rates comparing wild-type mice with animals lacking receptor expression (12, 13). Although we observed a tendency for lower sAPP concentrations in the brains of adult PDAPPRosa26Tg/+ mice (Fig. 4C), a clear inhibitory effect of raising SORLA activity above wild-type on APP processing was not seen in our model. Likely, wild-type mice already express sufficient amounts of SORLA to guide proper trafficking of APP, and further increasing receptor concentrations does not affect APP routing. Rather, excess SORLA molecules as in neurons of Rosa26Tg/+ mice may accentuate the function of this receptor in the lysosomal sorting of Aβ.

The molar ratio of APP/SORLA in PDAPPRosa26Tg/+ mice and PDAPPRosa26Tg/Tg mice is 1:1.5 and 1:1.8, respectively. A similar ratio is seen in the human brain (1:1.2), arguing that the lysosomal sorting function of SORLA may be important in vivo. The pathophysiological relevance of this receptor activity receives support from the recent identification of mutations in SORL1 in families with autosomal dominant early-onset AD (11). A total of two nonsense and five missense mutations have been identified throughout the receptor coding sequence. So far, no functional analysis has been provided as to whether any of these sequence alterations destroy receptor activities. Our data document that the mutation G511R, which targets the VPS10P domain, disrupts the ability of SORLA to bind to Aβ (Fig. 8B) and to promote intracellular catabolism of this peptide (Fig. 8C). These observations suggest impaired lysosomal sorting of the peptide as a molecular mechanism underlying this familial SORL1 mutation. Because APP binds to the cluster of complement-type repeats in SORLA, intracellular sorting of APP is probably not affected in the G511R receptor variant. However, another missense mutation (Asn1358Ser) has been identified in the APP binding domain of SORLA, suggesting that different functions of the receptor may be lost in the various familial mutations (11). Characterization of these additional familial mutations in cell and mouse models as shown for the G511R SORLA variant in this study will aid in dissecting the various functions of this receptor in AD-related pathologies.


Study design

This study aimed to investigate the consequences of SORLA overexpression on AD pathological processes. To address this question, we generated a new transgenic mouse strain that overexpresses human SORLA in the mouse brain. Using ELISA, Western blot analyses, and in vivo microdialysis experiments, we evaluated the impact of higher SORLA brain concentrations on APP processing and Aβ concentrations within different body compartments (plasma, ISF, and soluble brain extracts) in adult mice and, in some cases, also in newborns. The genotype of the mice was kept blinded until after statistical analyses of the respective samples had been performed. Sample sizes were predetermined on the basis of previous experience. In addition, to determine the mechanism by which higher SORLA concentrations affected Aβ catabolism, we treated parental neuroblastoma SH-SY5Y cells and SH-SY5Y cells stably overexpressing SORLA with DAPT and analyzed by ELISA changes in Aβ intracellular concentrations over time. These experiments were replicated at least three times.

Reagents, mouse models, and cell lines

Polyclonal antisera directed against LRP1, SORLA, APP, APOE, and CTFs were produced in-house. Antibodies directed against sortilin (BD Transduction Laboratories), Na/K ATPase (adenosine triphosphatase), α-tubulin, NeuN (Millipore), GFAP (DAKO), neprilysin (Santa Cruz Biotechnology), IDE (Abcam), α-tubulin (Covance), and β-actin (Sigma) were obtained from commercial suppliers. β-Amyloid peptide (1–40, HiLyte Fluor 488–labeled) was purchased from AnaSpec. PDAPP (14), Cre Deleter (Taconic), and Sorl1−/− (4) strains of mice have been described before. Mice carrying an inducible human SORLA cDNA in the Rosa26 locus were generated by standard homologous recombination in murine embryonic stem cells. The pCAG-DEST targeting vector contains the cytomegalovirus early enhancer/chicken β-actin promoter (CAG) element and the neomycin phosphotransferase gene driven by the mouse phosphoglycerate kinase 1 promoter and flanked by loxP sites. This construct was provided by R. Mort (University of Edinburgh).

Determination of protein concentrations and APP processing in tissues and in cell lines

Intensities of immunoreactive bands of Western blot analyses were quantified by optical densitometry with AIDA Image Analyzer (Raytest) software. Commercial ELISA kits were used to measure concentrations of APP (Life Technologies), human sAPPα/β and Aβ (Meso Scale Discovery), as well as murine sAPPα and Aβ (IBL) in brain extracts, mouse plasma, and cell lysates and supernatants. Concentrations of SORLA protein were determined by custom-made ELISA as previously described (26). For determination of Aβ catabolism in SH-SY5Y cells, replicate cell layers in 24-well format were treated with 5 μM DAPT, and cell lysates were collected at designated time points for ELISA.

Subcellular fractionation studies for lysosomes

Isolation of lysosomes from cells was performed according to Graham (37). In brief, SY5Y-A and SY5Y-A/S cells were harvested and homogenized in HM buffer (0.25 M sucrose, 10 mM Hepes, pH 7.0) with a Wheaton glass homogenizer. After centrifugation (800g for 10 min), the protein concentration was determined, and equal amount of protein was mixed with bovine serum albumin (BSA) and Percoll stock solution (20% final Percoll concentration). After ultracentrifugation (36,000g for 30 min, Beckman JA 25.15 rotor), 0.3-ml fractions were collected and supplemented with NP-40 to a 0.5% final concentration. Each fraction was centrifuged at 100,000g for 1 hour to pellet Percoll. To identify fractions enriched for lysosomes, the activity of lysosomal acid phosphate was determined by mixing 50 μl of each fraction with 0.2 ml of substrate solution (3 mg of p-nitrophenyl phosphate hexahydrate in 90 mM sodium acetate buffer, pH 5.0) and incubation for 30 min at 37°C. Thereafter, 0.2 ml of NaOH (0.25 M) was added to stop the reaction. Acid phosphatase activity was evaluated by measuring the absorbance of each fraction at 410 nm.

In vivo microdialysis experiments

We performed in vivo microdialysis to assess brain ISF Aβ40 in awake mice as previously described (38). Briefly, unilateral guide cannula and 2-mm microdialysis probes (BR-2, Bioanalytical Systems) were implanted into the hippocampus. Microdialysis perfusion buffer was artificial cerebrospinal fluid containing 4% BSA with a constant flow rate of 1.0 μl/min. ISF samples were assessed for Aβ40 with a sandwich ELISA. Anti–Aβ35–40 HJ2 was used as a capture antibody, and anti–Aβ13–18 HJ5.1-biotin was used as detecting antibody (39). Basal levels of ISF Aβ40 were defined as the mean concentration of Aβ40 over 6 hours. To assess ISF Aβ clearance, compound E was administered intraperitoneally at 20 mg/kg to halt Aβ production.

Quantitative reverse transcription polymerase chain reaction

Real-time reverse transcription polymerase chain reaction (PCR) was performed on cDNA reversely transcribed from total RNA with a 7900HT Real-Time PCR system (Applied Biosystems) (40). Transcript levels were determined with commercial Assay-on-Demand primer/probe sets (Applied Biosystems). Each template was run in triplicate, and relative expression levels were calculated by normalization to 18S ribosomal RNA as internal standard.

Fluorescence polarization assay

VPS10P ligand peptides containing additional cysteine at the N terminus were synthesized and labeled with Alexa Fluor 488–C5 maleimide (Molecular Probes) according to the manufacturer’s recommendation. The SORLA VPS10P domain fragment (amino acid residues 1 to 753) with C-terminal His tag was produced and purified as described previously (41). The human sortilin VPS10P domain (amino acid residues 1 to 758) was produced by a similar strategy. For fluorescence polarization assay measurements, 90-μl solution of VPS10P proteins in 50 mM tris (pH 7.0) and 150 mM NaCl containing 3 μM BSA were introduced into a quartz microcuvette and mixed with a 10-μl portion of peptide solution at a final concentration of 100 nM. After incubating at room temperature for 30 min, Alexa Fluor 488 fluorescence was measured with a Hitachi F-7000 fluorescence spectrophotometer, with excitation and emission wavelengths at 490 and 520 nm, respectively. Fluorescence polarization values were defined as FP = (I0I90)/(I0 + I90), where I0 and I90 are the fluorescence intensities parallel and perpendicular to the plane of incident light, respectively.

AP fusion reporter binding assay

Aβ peptide was fused C-terminally to AP with pAPtag-5 vector (42), followed by transient expression in HEK293T cells. Only the first 32-residue portion was used to make the fusion protein (AP–Aβ1–32) because inclusion of the last 8 residues prevented secretion into the medium. The wild-type or G511R mutant SORLA VPS10P domain fragments carrying a C-terminal TARGET tag (41) were captured onto Sepharose beads via immobilized P20.1 antibody against TARGET tag and incubated with culture medium containing either AP alone or AP–Aβ1–32 for 1 hour at 4°C. After washing three times with 20 mM tris (pH 7.5) and 150 mM NaCl [tris-buffered saline (TBS)], the immunocomplexes were eluted with C8 peptide (0.2 mg/ml) in TBS, and the AP activity was measured as previously described (42). For each experiment, the amount of the AP activity obtained with AP alone was subtracted from that with AP–Aβ1–32 as a blank.

Statistical analysis of data

Statistical analyses were performed with Prism 5.0 software (GraphPad Software). Results were analyzed with two-tailed Student’s t test or a one-way ANOVA where appropriate. Post hoc differences in Fig. 7C were compared with Bonferroni’s multiple comparison test. In Fig. 4, a Grubbs’ test was used to detect significant outliers.


Fig. S1. Targeting strategy to generate mice with inducible SORLA overexpression.

Fig. S2. Viability and SORLA levels in transgenic mice.

Fig. S3. SORLA overexpression does not affect production of C-terminal APP fragments.

Fig. S4. Analysis of pathways in brain Aβ catabolism.

Fig. S5. SORLA activity does not affect concentration of Aβ in plasma.

Fig. S6. Mapping of the SORLA-binding region within the Aβ sequence.


  1. Acknowledgments: We are indebted to R. Mort (University of Edinburgh) for providing the CAG construct and to T. Pantzlaff, C. Kruse, and M. Schmeisser for the expert technical assistance. Funding: Studies were funded by grants from the Helmholtz-Association (to T.E.W.), the Alzheimer Forschung Initiative (to T.E.W.), and the Thyssen Foundation (to T.E.W.), by Grant-in-Aid for Scientific Research (A) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) (to J.T.), and by the “Platform for Drug Discovery, Informatics, and Structural Life Science” grant from MEXT (to J.T.). Author contributions: Experiments were performed by S.C., A.-S.C., and T.B. (Figs. 1 to 4 and figs. S2 to S5), by V.S. (Figs. 7 and 8C), by F.L. (Fig. 5), and by S.T.-N. and Y.K. (Figs. 6 and 8, A and B, and fig. S6). E.-M.F. and A.F. generated the transgenic mouse models by blastocyst injection of targeted embryonic stem cell clones. D.M.H., J.T., and T.E.W. designed the experiments and evaluated all data. T.E.W. wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article