Interaction between amyloid precursor protein and Nogo receptors regulates amyloid deposition

Abstract

Excessive production or accumulation of β-amyloid (Aβ) peptides in human brains leads to increased amyloid deposition and cognitive dysfunction, which are invariable pathological features in patients with Alzheimer's disease (AD). Many cellular factors can regulate the production of Aβ. In this study, we show that a family of proteins named Nogo receptor proteins (NgR1 to NgR3) regulates Aβ production via interaction with amyloid precursor protein (APP). Further mapping of the interacting domain indicates that a small region adjacent to the BACE1 cleavage site of APP mediates interaction of APP with Nogo receptor proteins. Our results also indicate that increased interaction between Nogo receptor and APP reduces surface expression of APP and favors processing of APP by BACE1. When NgR2 was ablated in AD transgenic mice expressing Swedish APP and PS1ΔE9, amyloid deposition was clearly reduced (0.66% of total measured area in APP_swe/PS1ΔE9/NgR2^−/− mice vs. 0.76% of total measured area in APP_swe/PS1ΔE9 mice). Our results demonstrate that down-regulation of NgR expression is a potential approach for inhibiting amyloid deposition in AD patients.—Zhou, X., Hu, X., He, W., Tang, X., Shi, Q., Zhang, Z., Yan, R. Interaction between amyloid precursor protein and Nogo receptors regulates amyloid deposition.

One of the invariable pathological features of Alzheimer's disease (AD) is the presence of amyloid plaques in the brain (1). The major proteineous components in amyloid plaques are 39- to 43-aa β-amyloid (Aβ) peptides, which are released from a much larger amyloid precursor protein (APP) through sequential cleavages by 2 endopeptidases: BACE1 (β-secretase) and γ-secretase (2–4). Production of Aβ is disrupted if APP is processed by α-secretase and BACE2, both of which cleave APP within the Aβ region (5–8). Therefore, the levels of Aβ can be dysregulated if the activity of any of these proteases is perturbed.

To identify proteins that can modulate BACE1 activity, we and others demonstrated that reticulon (RTN/Nogo) proteins interact with BACE1 and that this interaction negatively modulates BACE1 activity (9–13). Although the mammalian RTN family consists of 4 members (RTN1 to RTN4; refs. 14–17), RTN3 has been more extensively studied because of its highly enriched expression in neurons. Transgenic mice overexpressing RTN3 in neurons exhibit reduced amyloid deposition in their cortical regions (18), consistent with the role of RTN3 in the negative modulation of BACE1.

In the initial study of the role of RTN proteins in the regulation of BACE1 activity, Fournier et al. (19) reported that a novel leucine-rich repeat protein interacts with RTN4, alternatively known as Nogo, by expression cloning. This novel protein was named as Nogo receptor (NgR; NgR1) for its putative role in inhibiting axonal outgrowth and neuritic sprouting through interaction with Nogo (20). Further in silico alignments identified 2 additional homologous proteins, NgR2 and NgR3 (21–23); all 3 NgR proteins are largely expressed by neurons (24). Despite the presence of 3 mammalian NgR proteins, most current biochemical and genetic studies have been focused on NgR1. For example, it has been shown that NgR1 interacts with all 3 myelin inhibitory proteins, Nogo-A, MAG, and OMgp, to modulate neurite outgrowth (25–28). NgR1 is also implicated in the regulation of intracellular trafficking of cholesterol via interacting with Niemann-Pick type C2 protein (29). Genetic mutation at the NGR locus and NgR1 deficiency are linked as genetic risks for neuropsychiatric disorders such as schizophrenia (30–32)

Our interest in NgR proteins stems from the interaction between BACE1 and Nogo. We asked whether binding of Nogo with NgR would constrain its interaction with BACE1, thereby altering BACE1 activity. Contrary to our expectation, we found that overexpression of NgR1 had no effect on the physical interaction between BACE1 and Nogo, but instead, NgR1 coimmunoprecipitated with the BACE1 substrate APP. Our initial observation was consistent with another study that showed interaction of NgR1 with APP (33). Although Park et al. (33) suggested that only NgR1 interacts with APP, we found that APP interacts with all 3 NgR proteins. Because the interaction between NgR2 and APP exhibited the strongest effect on elevating Aβ production in cultured cells, our experiments have been focused on NgR2. We mapped the binding domain in APP to the residues adjacent to the BACE1 cleavage site (between 558 and 599 based on the full length of 695 aa). Although NgR2 does not interact with BACE1, increased interaction between NgR2 and APP in cells could reduce surface expression of APP, and this altered cellular trafficking of APP by NgR2 indirectly favors processing of APP by BACE1. Consistent with this in vitro result, genetic deletion of NgR2 in an AD transgenic mouse model, which expresses Swedish APP and mutant PS1 (34), decreases amyloid deposition. Collectively, Nogo and Nogo receptor family proteins clearly have differential effects on Aβ production.

MATERIALS AND METHODS

Cell lines and reagents

Human HEK-293 cells, maintained under standard culture conditions, were transfected with NgR1- or NgR2-containing C-terminal flag tag and stable cell lines named NgR1-HEK and NgR2-HEK were generated by G-418 selection (80 μg/ml). Antibody R461, which recognizes the C terminus of Nogo (RTN4; ref. 9), and antibody B279, which recognizes the residues 295–310 of BACE1 (35), were generated in our laboratory. Antibody 8717 recognizes the APP C terminus, and monoclonal M2 antibody specifically reacts with the flag tag; both antibodies with anti-flag M2 affinity beads were purchased from Sigma (St. Louis, MO, USA). Monoclonal antibody 22C11 was purchased from Chemicon (Temecula, CA, USA), and 6E10 was purchased from Signet (Dedham, MA, USA). BACE2 antibody was purchased from Affinity Bioreagents (Golden, CO, USA). Polyclonal antibody against NgR2 was generated by immunization with a KLH-conjugated NgR2 peptide (EELDLGDNRHLRS). This antibody is only sensitive to overexpressed NgR2 in cultured cells but not the endogenously expressed NgR2 in brains on a Western blot.

Generation of NgR expression constructs

DNA fragments encoding human NgR1 to NgR3 were obtained by PCR amplification using templates from human brain libraries and were inserted into the pflag-CMV vector (Sigma). Mouse NgR1- to NgR3-coding DNA fragments were similarly generated and inserted into pCDNA3-CMV vectors with no additional tag sequence added. APP-BACE2 expression construct was generated by fusion of the entire extracellular domain of APP with the transmembrane and C-terminal region of BACE2. To generate truncation APP mutants, either the XbaI or ECoRI site was first created in the desired location of APP by site-directed mutagenesis according to instructions from the QuikChange site-directed mutagenesis kit (Stratagene, Palo Alto, CA, USA), and the mutated APP was then digested with either XbaI or EcoRI to delete ∼300 DNA bases, as illustrated in the corresponding figures. All mutated expression constructs were validated by double-strand DNA sequencing.

Creation of NgR2-knockout mice

Genomic DNA spanning the NgR2 gene was purchased from Invitrogen (Carlsbad, CA, USA). The 5′ (3.7 kb) and 3′ (4.6 kb) arms of DNA fragments, which flank exon 3 of the mouse NgR2 gene, were obtained by PCR amplification and sequentially subcloned into the targeting Neo/Tk vector (as outlined in Fig. 5A). The cloned DNA fragments were confirmed by both restriction digestions and DNA sequencing. The generated targeting vector was then transfected into 129 SvEvTac embryonic stem (ES) cells by electroporation. Selection of positive ES cells was confirmed by Southern blot analysis, and the targeting vector-integrated ES cells were injected into blastocysts derived from C57BL/6J ES cells. The injected blastocysts were then transferred into pseudopregnant female mice to give birth to chimeric mice, which were further bred with wild-type (WT) C57BL/6 to generate NgR2 heterozygous and homozygous knockout mice. To reduce the possible effects of genetic background on amyloid deposition, as well as other possible phenotypes, the founder mice were backcrossed with C57BL/6 mice for 4 generations to have relatively pure genetic background before breeding with transgenic mice expressing APP/PS1ΔE9 (36) purchased from The Jackson Laboratory (Bar Harbor, ME, USA).

Figure 5.

Validation of genetic deletion of NgR2 in mice. A) As illustrated, mouse NgR2 is encoded by 3 exons, and the targeting construct was constructed to remove exon 3. After homologous recombination, the original 7.5-kb BstXI fragment in WT mice was replaced by the 5.8-kb Neo-containing fragment in NgR2-null mice. DNA size markers are 9.4 and 6.5 kb, respectively. Solid line near the last BstXI site indicates Southern blot probe. B) Northern blot analysis confirmed that the 2.4-kb NgR2 transcript was abolished in NgR2-deficient mice. β-Actin was used as a loading control.

Genotyping and Southern and Northern blot analyses

The following pairs of primers (from 5′ to 3′ end) were used for PCR-based genotyping: CAGCCTGGACATTTCTGTGA (forward primer) and GGGGAACTTCCTGACTAGGG (reverse primer) were used to validate NgR2 deficiency; TAACTCTTCCTCCCCCATCC (forward primer) and TCCAGGCACCCTAAAAAGTG (reverse primer) were used to detect WT NgR2 gene. Southern blotting was performed with a DIG-labeled DNA probe following the protocol from Roche (Indianapolis, IN, USA). DNA probes for Southern blotting were synthesized by PCR amplification using the following primer set (from 5′ to 3′ end): TTAACACGTTTCCCCAGAGC and CGACGTGTGGACCTTCTAGG. Northern blotting was carried out with a DIG Northern starter kit (Roche). Double-strand DNA template was synthesized first by PCR amplification with this pair of primers (AGCCTGCCTACTGCTGACAC and GAGCTGGCAACGATACAGGT). T7 promoter sequence was added to the 5′ end of the reverse primer to synthesize the DIG-labeled antisense RNA probe by an in vitro transcription.

Immunoprecipitation and Western blot assay

Cultured cells were first grown in DMEM for 24 h in 10-mm plates to ∼70% confluence and then transfected with the indicated expression constructs. After incubation for 48 h, cell lysates were prepared according to the procedures for immunoprecipitation (IP) with monoclonal anti-M2 flag affinity beads. The extensively washed immunocomplexes were eluted and resolved on a 4–12% NuPage Bis-Tris gel from Invitrogen for standard Western blot assays with the indicated primary antibody and an appropriate horseradish peroxidase-conjugated secondary antibody. To perform Western blotting with mouse brain samples, brain tissues were homogenized in modified RIPA buffer, as described previously (37), and equal amounts of proteins were loaded on NuPage gels for Western blot analysis.

Quantification of Aβ peptides using ELISA

Soluble and insoluble Aβ_1–40 and Aβ_1–42 were differentially prepared from the frozen brain tissues by the guanidine hydrochloride method (38). The levels of Aβ_1–40 and Aβ_1–42 in the mouse brain samples were quantified by sandwich ELISA according to the procedure previously described (39). Results were obtained from 6 female APP_swe/PS1ΔE9 and 8 female APP_swe/PS1ΔE9/NgR2^−/− mouse brain samples. Mouse Aβ_1–40 measurement was performed according to the protocol provided by the manufacturer (Wako, Osaka, Japan).

Quantification of amyloid plaque load

Quantification of amyloid plaques was conducted with sagittal sections of mouse brain. Amyloid plaques were marked with 6E10 antibody followed by 3,3′-diaminobenzidine (DAB) visualization. Images were captured with a Leica light microscope (Leica Microsystems, Wetzlar, Germany) and plaque load, defined as percentage of total area, was measured from 8 evenly spaced sections chosen from each animal by ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA). An intensity threshold level was established to distinguish authentic plaque signal from the background intensity. The threshold for detection was held constant throughout the image quantification. Five female mice at the age of 8 mo from each group were used for the experiment.

Cell surface protein biotinylation

Cell surface protein biotinylation and detection were performed as described previously (35). Briefly, WT mouse NgR2 expression plasmid or control vector was transiently transfected into HEK-293 cells. After 24 h transfection, cell surface proteins were biotinylated by incubation of the live transfected cells with 1 mg/ml EZ-Link Sulfo-NHS-SS-Biotin at 4°C for 30 min. After being quenched and washed, biotinylated cells were lysed in a low-salt buffer (20 mM HEPES at pH 7.9, 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 1% Nonidet P-40, and 10% glycerol and protease inhibitors) at 4°C for 30 min on a rotator and centrifuged for 5 min at 15,000 g. Supernatants were then collected, and portions of the total supernatants were used in the neutravidin pulldown experiment to isolate biotinylated surface proteins. The total protein and the surface protein were then resolved on 4–12% Bis-Tris NuPage gels and analyzed by Western blotting.

RESULTS

Interaction of NgR proteins with APP

NgRs are synthesized as a family of proteins that are localized onto the cell surface via GPI linkage (40), and Nogo presumably interacts with NgR1 via an intercellular connection on the cell surface. Three NgRs share similar structural organization with ∼40% amino acid sequence identity (21). Because Nogo can negatively modulate BACE1 activity (16) and Nogo binds to NgR1 (20), we set out to test whether NgRs affect APP processing. We cloned all 3 NgR proteins by PCR amplification and generated expression constructs either with or without flag tags at their C termini. Expression of these constructs in 125.3 cells, which stably express Swedish mutant APP (39), showed expression of 3 proteins at the expected sizes (Fig. 1A). APP is a highly glycosylated protein, with both N-linked and O-linked glycosylation. The specific interaction between NgRs and APP was detected by using 2 antibodies: monoclonal 6E10, which recognizes the N terminus of the Aβ region (Fig. 1A), and polyclonal antibody A8717, which recognizes the C terminus of APP (Fig. 1B). Evidently, all 3 NgRs preferentially interacted with the less matured full-length APP, as anti-flag affinity beads clearly pulled down only the fast migrating full-length APP. APP is a type I transmembrane protein present in all secretory compartments, and full glycosylation occurs in the late Golgi and trans-Golgi network (TGN). Most of the C-terminal flag-tag of NgRs is expected to be removed in the late secretory compartments due to a GPI linkage. Our results suggest that the interaction between APP and NgRs begins at the early secretory compartments.

Figure 1.

Interaction of NgR proteins with APP. Expression constructs were transfected in 125.3 cells (A, B) and HM cells (C) for 48 h. Post-transfection, the lysates were prepared for IP with antiflag antibody followed by Western blot analysis with the specific antibodies. A, B) Immunocomplexes were examined with antibody 6E10 recognizing Aβ regions (A) and antibody A8717 that recognizes APP C terminus (B). Longer exposure of the film showed no interaction of NgRs with CTF83. C) Antibody B279 reacts with residues 295–310 of BACE1, and NgRs do not interact with BACE1. Arrows indicate nonspecific IgG bands.

Notably, the membrane-bound APP C-terminal fragments did not interact with NgRs (Fig. 1B), suggesting that this interaction is not due to an overexpression artifact from membranous associations. To further support this, under the same coimmunoprecipitation (co-IP) conditions, flag-tagged NgRs did not pull down overexpressed BACE1, another type I transmembrane protein, as antibody B279 only detected overexpressed BACE1 in the lysates but not in the immunocomplexes (Fig. 1C). Thus, we showed that NgRs specifically interact with APP and that this interaction does not require the APP C-terminal fragment, which includes the transmembrane domain.

To further study this interaction, we generated stable cell lines expressing NgR1 or NgR2 to test their interactions with various APP mutants. We found that expression of the secreted form of APP in NgR1- or NgR2-stably expressing cells failed to detect any interaction (data not shown) suggesting that anchoring of APP on the membrane was perhaps required. To confirm this further, we generated a chimeric protein that fused the entire extracellular domain of APP with the transmembrane domain plus the C-terminal tail of BACE2 (illustrated in Fig. 2A). Like APP, BACE2 is also a type I transmembrane protein and the cellular expression pattern of BACE2 resembles that of APP (7). Expression of this chimeric protein in cells stably expressing either NgR1 or NgR2 showed an interaction between NgR and the chimeric protein APP-BACE2 (Fig. 2A; example of NgR2 with the chimeric protein). Since various APP mutants were previously generated in our laboratory to perturb the cleavages of APP by BACE1 (41, 42), we tested whether NgRs would interact with APP in the disrupted Aβ region. Despite a perturbed Aβ domain, these APP mutants retained their interactions with NgR1 and NgR2 (Fig. 2B). Hence, our results further demonstrate that the Aβ domain is not required for this specific interaction, while anchoring of the protein on the membrane is required.

Figure 2.

Aβ domain of APP is dispensable for the interaction. A) Chimeric protein was generated by fusion of the extracellular domain from APP and the transmembrane domain and C-terminal tail from BACE2. Expression of this chimeric protein in a cell line stably expressing NgR2 was detectable by antibody 22C11, which recognizes the N-terminal domain of APP, and BACE2 antibody. NgR2 clearly interacts with this chimeric protein. B) Various APP mutants, named APP_21A6, APP_IA4KK, APP_EFI, and APP_WA4I, were expressed in a cell line stably expressing NgR2. Co-IP was performed with anti-flag beads; immunocomplexes were examined with antibody A8717. While the full-length APP or any APP mutant was coimmunoprecipitated with NgR2-flag, there was no co-IP of NgR2 with APP-CTFs.

A small region adjacent to the BACE1 cleavage site mediates APP interaction with NgRs

To identify the specific domain that mediates binding of APP to NgRs, we performed sequential deletion of the N-terminal domain of APP, as illustrated in Fig. 3. Expression of these truncated APP mutants in either NgR1 or NgR2 cell lines showed that deletion of residues 500 to 599 (the construct APP_Δ500–599) abolished the interaction between APP and NgRs (Fig. 3, middle left panel). Fusion of this fragment or an even smaller region together with the rest of the C-terminal fragments and the first 113 amino acids of APP produced 2 additional expression constructs: APP_Δ113–517 and APP_Δ113–557. It appears that a small region (residue 558 to 599) was sufficient to mediate the interaction between APP and NgRs in our co-IP experiments (Fig. 3, bottom left panel). We also noted that APP_Δ113–211 exhibited a stronger binding to NgR2, implying that the potential change in tertiary structure may favor this interaction by exposing the binding pocket. These co-IP experiments consistently showed that NgR2 interacted with the less mature APP mutant proteins, as flag beads pulled down mainly the faster-migrating bands (Fig. 3). Collectively, our results showed that NgRs strongly interact with APP via a core region spanning residues 558 to 599. Although this region covers the BACE1-cleavage site (between residues M595-D596), the sequence immediately surrounding this cleavage site is not required, as demonstrated in Fig. 2B (lane APP_IA4KK).

Figure 3.

Mapping of the binding domain of APP. Various APP mutants with the indicated deletion of ∼100 residues as illustrated were generated, and their expressions in a NgR2-expressing cell line were confirmed by Western blot analysis with antibody A8717. NgR2 interacted with all APP truncation mutants except APP_Δ500–599. Co-IP between NgR2 and APP_Δ113–557 indicates that residues within the region between residues 558 to 599 are required for this interaction. Top left panel: protein inputs from a NgR2-expressing cell line overexpressing deletion mutants. Bottom left panels: IP prepared from cell lysates using anti-Flag antibody and blotted with C-terminal APP antibody. Right panel: schematic diagram of sequential deletion mutation of full-length APP.

NgR-APP interaction elevates Aβ production in cultures

To explore whether the interaction between NgRs and APP would affect APP metabolism, we transiently transfected NgR expression constructs in 125.3 cells, which stably express Swedish mutant APP and produce high levels of Aβ (39). The conditioned medium from the treated cells was aliquoted for quantifying Aβ_1–40 and Aβ_1–42 by ELISA. We found that overexpression of human NgR2 significantly elevated both Aβ_1–40 and Aβ_1–42 levels (Fig. 4). The concentration of Aβ_1–40 in hNgR2flag-expressing cells was 53.63 ng/ml compared with 33.12 ng/ml in mock-transfected cells (n=4; P<0.01, Student's t test). The concentration of Aβ_1–42 peptide, which is more fibrillogenic, was also significantly increased in NgR2-transfected cells (9.57 ng/ml in hNgR2flag-transfected cells vs. 5.97 ng/ml in control cells; n=4; P<0.01, Student's t test). Overexpression of WT mouse NgR2 protein, which shares ∼94% homology with its human orthologue, displayed similar effects on elevated Aβ production (Fig. 4; 50.4 ng/ml in mNgR2-transfected cells vs. 33.12 ng/ml in control cells; n=4; P<0.01, Student's t test). The enhancing effects on the production of Aβ_1–40 and Aβ_1–42 from NgR1 (data not shown) or NgR3 overexpression were relatively weaker (Fig. 4). In addition, similarly increased Aβ levels induced by NgR2 overexpression were observed in another N2A-APP cell line, which stably expresses Swedish mutant APP in mouse neuroblastoma Neuro-2a cells (data not shown). All of the above results suggest that NgRs, by binding with the APP molecule, alter Aβ production but do not cause a shift in the ratio of Aβ_1–40 to Aβ_1–42. It should also be noted that overexpression of Nogo or Nogo family member RTN3 causes significant reduction of Aβ_1–40 and Aβ_1–42 under the same experimental conditions (9). Hence, the effect of Nogo and Nogo receptors on Aβ production is divergent.

Figure 4.

Effect of NgR-APP interaction on Aβ production. Indicated expression constructs were expressed in 125.3 cells in quadruplets for 48 h, and the conditioned medium was portioned into aliquots for quantification of Aβ_1–40 (A) and Aβ_1–42 (B) by ELISA. *P < 0.05, **P < 0.01 vs. flag control.

NgR2 deficiency causes reduced amyloid deposition in mice

To investigate the in vivo effect of NgRs on Aβ production, we chose to generate NgR2-knockout mice because of the stronger effect on Aβ production in cultured cells. We generated a targeting construct to remove exon 3, which encodes the C-terminal 250 aa (residue 172 to residue 420; Fig. 5A). Although we failed to generate specific antibodies that can recognize endogenously expressed NgR2 after several attempts, and there were no commercial antibodies available for detecting endogenously expressed NgR2, we managed to validate genetic deletion of NgR2 in the knockout mice by employing Southern and Northern blot analyses. In NgR2-knockout mice, it was clear that the 7.5-kb BstXI fragment was completely replaced by the 5.8-kb fragment that contains a partial Neo sequence (Fig. 5A). The removal of exon 3 rendered the NgR2 mRNA unstable as the 2.4-kb transcript was totally eliminated in the NgR2-null mice (Fig. 5B). NgR2-null mice are viable and fertile. Moreover, their life span and body weight were also comparable to their WT littermates (data not shown). Detection of no obvious discernible abnormalities in NgR2-null mice is not surprising, as NgR1-knockout mice are also similarly viable and healthy (33).

To create NgR2 deficiency in an AD mouse model, we bred APP_swe/PS1ΔE9 double-transgenic mice with NgR2 homozygous mice, and the obtained APP_swe/PS1ΔE9/NgR2^+/− offspring were further bred with NgR2 heterozygotes to knock out NgR2 in this AD animal model. Brain samples were prepared from both 2- and 8-mo-old APP_swe/PS1ΔE9/NgR2^−/− and APP_swe/PS1ΔE9 mice for quantification of Aβ levels by ELISA (18). The total Aβ levels from 2-mo-old APP_swe/PS1ΔE9/NgR2^−/− samples showed a slight reduction, although not reaching statistical significance, when compared with APP_swe/PS1ΔE9 samples (data not shown). However, a reduction of Aβ (both Aβ_1–40 and Aβ_1–42) became more obvious in samples from 8-mo-old APP_swe/PS1ΔE9/NgR2^−/− mice compared with APP_swe/PS1ΔE9 control littermates (Fig. 6A). This reduction was consistent with the reduced Aβ deposition in APP_swe/PS1ΔE9/NgR2^−/− mouse brains (Fig. 6B). Quantification of amyloid plaque load with ImageJ also showed a clear reduction in amyloid plaque density within the cortex and the hippocampal region (Fig. 6C; 0.66% of total measured area in APP_swe/PS1ΔE9/NgR2^−/− mice vs. 0.76% of total measured area in APP_swe/PS1ΔE9 mice; n=5 animals; P<0.05, Student's t test). Hence, NgR2 deficiency can reduce Aβ production and the subsequent formation of amyloid plaques.

Figure 6.

Reduced Aβ and amyloid deposition in NgR2-null mice. A) Levels of soluble and insoluble Aβ_1–40 and Aβ_1–42 extracted from brain tissues were measured by sandwich ELISA. Aβ values are expressed as mean ± se Aβ concentration (ng/g tissue). Difference in the values of both soluble and insoluble Aβ between 2 genotypes of mice was significant (n=6 and 8). *P < 0.05; Student's t test. B) Immunohistochemical analysis of condensed Aβ deposits in brain sagittal sections of 8-mo-old mice. Amyloid plaques were detected by monoclonal antibody 6E10. C) Plaque load was determined by percentage of the area occupied by Aβ-immunoreactive condensed deposits in total examined area. Imaging data for Aβ plaque load were collected from the DAB-stained sections by antibody 6E10, and analyzed with Image J software (n=5). *P < 0.05; unpaired t test.

NgR2 regulates surface expression of APP

To explore the mechanism underlying the reduced amyloid deposition resulting from NgR2 deficiency, we examined levels of APP and its processed products. We found that the overall levels of full-length APP were slightly elevated in 8-mo-old APP_swe/PS1ΔE9/NgR2^−/−mice compared with their APP_swe/PS1ΔE9 littermates (Fig. 7A; P>0.05, Student's t test), suggesting that less APP was perhaps metabolized to produce Aβ, consistent with the reduced levels of Aβ in these mice. Since the ratio of BACE1-cleaved CTF99 to α-secretase-cleaved CTF83 is an indicator of whether the balance between BACE1 cleavage and α-secretase cleavage of APP is shifted, we compared this ratio between APP_swe/PS1ΔE9/NgR2^−/− and APP_swe/PS1ΔE9 samples on the gels. It appears that the elevation of full-length APP was due to the reduced ratio of CTF99 to CTF83 (Fig. 7A; 1.22 in APP_swe/PS1ΔE9/NgR2^−/− vs. 1.63 in APP_swe/PS1ΔE9; n=6 animals; P<0.05, Student's t test). An additional nonparametric Mann-Whitney U test was used to determine whether there was a significant difference between the 2 groups, and the result indicated the same (P<0.05; n=6/group). This reduction was also obvious in 8-mo-old hippocampal lysates from NgR2-null mice compared with their WT littermates (Fig. 7B; 0.021 in NgR2^−/− vs. 0.029 in WT mice; n=5 animals; P < 0.05, Student's t test; P<0.05, Mann-Whitney U test). The reduced ratio of CTF99 to CTF83 clearly indicates that NgR2 deficiency favors α-secretase cleavage of APP. In separate experiments, we also measure mouse Aβ1–40 levels, and a slight reduction of mouse Aβ1–40 was seen in even 4-mo-old NgR2 deficiency mouse cortical samples (225±35 fmol/g tissue in NgR2^−/− vs. 286±20 fmol/g tissue in WT mice; P<0.05, Student's t test).

Figure 7.

Analysis of APP and its processed C-terminal fragments. A) Brain lysates were prepared from 8-mo-old APP_swe/PS1ΔE9/NgR2^−/− mice or their APP_swe/PS1ΔE9 littermates for Western blot analysis with antibody A8717, which recognizes the C terminus of APP. Equal amounts of proteins were loaded onto the gel and verified by the antibody to calnexin. B) Similar Western blot analyses were performed by using brain lysates prepared from NgR2-null mice and their WT littermates. C) Brain cortical samples from 4-mo-old NgR2-deficient mice and their WT littermates were prepared for measuring mouse Aβ levels. There was a small reduction of mouse Aβ1–40 in samples from NgR2-knockout mice compared with WT control (n=6 mice) *P < 0.05.

Since NgRs do not interact with BACE1 directly, it is intriguing to consider why changes in levels of NgRs would shift normal APP processing by BACE1 and α-secretase. To explore this, we examined surface expression of APP in cells transfected with WT NgR2. We found that overexpression of NgR2 in HEK293 cells appeared to reduce the surface expression of APP, as the levels of biotinylated APP were further decreased compared with their levels in the vector-transfected cell lysates (Fig. 8). Although the endogenous expression of BACE1 on the cell surface was not detectable by currently available commercial or our own BACE1 antibodies, the surface level of ADAM10 was not significantly altered by the overexpression of NgR2. Together, we inferred that the reduced surface expression of APP may favor processing by BACE1, as endogenous levels of BACE1 are highly enriched in the later Golgi and TGN (35, 43). Increased processing of APP by BACE1 in the early secretory compartments is known to favor competitive cleavages of APP by BACE1 over α-secretase (4, 44).

Figure 8.

Cell surface protein biotinylation. HEK293 cells were transfected with WT NgR2 expression plasmid DNA or empty vector DNA for 48 h, and surface proteins were labeled by biotinylation. Neutravidin was used to pull down biotinylated surface proteins, followed by Western blotting. Surface biotinylated APP is significantly lowered in NgR2-overexpressing cells, while surface expression of ADAM17 was not significantly affected by NgR2. APP was detected by antibody A8717 and NgR2 by polyclonal NgR2 antibody. Calnexin and actin are shown as loading controls. *P < 0.05; Student's t test and Mann-Whitney U test.

DISCUSSION

NgR was initially identified as a receptor for Nogo, and the Nogo-NgR interaction regulates axonal outgrowth (19). NgR has 2 other family members: NgR2 and NgR3. All 3 NgR proteins (NgRs) share similar structural organization, which contains a stretch of 8 leucine-rich-repeat (LRR) domains. This stretch of LRR domains, separated by the short LRR N- and LRR C-terminal domains, are the most conserved among NgR members (between 52 and 59% identity), while the C-terminal GPI anchoring domain is the least conserved (between 16 and 18% identity). Together, these domains constitute the extracellular portion of the protein that mediates interaction with various proteins. Here, we showed that NgRs interact with the extracellular domain of APP and that this interaction appears to regulate Aβ production and amyloid deposition. Reducing NgR interaction with APP is a potential approach to decrease amyloid deposition in AD patients.

NgRs are a family of GPI-anchored proteins; this GPI anchor is composed of glycosylphosphatidylinositol, which can be attached to the C-terminal residues of NgRs during post-translational modification and anchor NgRs on the cell surface outer leaflet. In this study, the flag tag was fused to the C terminus of NgRs to avoid disruption of the extracellular organization; the flag tag is likely removed after the GPI attachment and before reaching the cell surface. We found that the C-terminal flag tag preferentially pulled down all immature APP variants in our co-IP experiments, indicating that the interaction between NgRs and APP begins as early as in the endoplasmic reticulum (ER) compartment, where full glycosylation has not yet occurred. If using a NgR1-specific antibody for similar co-IP experiments, NgR1 interacts with both mature and immature forms of APP (33). This interaction is specific because NgR-flag proteins did not coimmunoprecipitate with overexpressed BACE1, which is largely enriched in the ER, TGN, and endosomes. More importantly, the interaction between NgR and APP naturally occurs in the mouse brain (33). Our mapping study indicated that residues 558 to 599 in APP could mediate its interaction with NgRs, and this is consistent with the prior report that showed ectodomain APP in mediating its interaction with NgR (33).

While there is consistency in the physical interactions between NgRs and APP, the effect of this interaction on Aβ generation does not appear to be completely congruent. Park et al. (33) showed that increased expression of NgR1 reduced Aβ generation in N2A cells stably expressing Swedish APP, while mice deficient in NgR1 showed increased amyloid deposition. More intriguingly, infusion of a soluble NgR fragment in APP/PS1 transgenic mice reduced amyloid plaque deposits and dystrophic neurites (33). Our results appear contrary with regard to Aβ production. We found that increased expression of NgR1 slightly enhanced production of Aβ in 2 different cultured cell lines (data not shown), and this modest increase was also seen in human neuroblastoma SH-SY5Y cells overexpressing RTN4R (NgR1) (45). Hence, our results were consistent with this study. Moreover, a recent study (46) using a traumatic brain injury model also showed that the same NgR1-null mice failed to cause enhanced amyloid deposition, contrary to the original report by Park et al. (33). Nevertheless, infusion of a soluble NgR fragment in an AD mouse model can reduce amyloid deposition, suggesting that this fragment exerts an inhibitory activity in Aβ production in vivo. Because of these intriguing observations, it is clear that NgR proteins can regulate APP metabolism even though the role of NgR1 in Aβ production requires further investigation.

In this study, we showed that increased expression of NgR2 enhanced production of Aβ in cultured cells and mice deficient in NgR2 had reduced amyloid deposition. We noted that the reduction of Aβ plaque load in an AD model with NgR2 deficiency was not substantial (reduced by ∼15%) and that the presence of other NgR members might compensate for some loss of NgR2 or replicate the NgR2 effect. Since the interaction between APP and NgRs is evident and consistent in different studies, it will be more important to examine an AD model with deficiency in all NgR proteins to avoid possible redundancy.

The reduction of Aβ production in NgR2-null mice appears to stem from the reduced processing of APP by BACE1 because of a reduction in CTF99. Although there was no obvious physical interaction between NgRs with BACE1, NgRs could affect cellular trafficking of APP. It appears that less APP was transported to the cell surface on overexpression of NgR2 in cultured cells. Endogenous BACE1 is mostly enriched in the late Golgi and TGN, while α-secretase is more enriched on the cell surface. If cells are treated with brefeldin A, which can block the transport of membrane proteins beyond the ER compartment, nascent APP is cleaved by BACE1, but there is hardly any cleavage from α-secretase (35), suggesting that a majority of α-secretase is located in the late secretory compartments. Hence, the reduced APP on the cell surface may be favorable for APP processing by BACE1.

In summary, we show that increase NgR expression elevates Aβ generation while mice deficient in NgR2 have reduced Aβ levels and amyloid deposition. Down-regulation of NgRs is a potential therapeutic strategy to reduce amyloid deposition. Alternatively, blocking the interaction between NgRs and APP may also inhibit the formation of amyloid plaques.