|Home | About | Journals | Submit | Contact Us | Français|
Reticulon 3 (RTN3) was initially identified as a negative modulator of BACE1, an enzyme that cleaves amyloid precursor protein (APP) to release β-amyloid peptide. Interestingly, RTN3 can also form aggregates upon accumulation, and increased RTN3 aggregation correlates with the formation of RTN3 immunoreactive dystrophic neurites (RIDNs) in brains of Alzheimer’s cases. Transgenic mice expressing RTN3 alone develop RIDNs in their hippocampus but not in their cortex. To determine the in vivo effects of RTN3 and preformed RIDNs on amyloid deposition, we crossed bi-transgenic mice expressing APP and presenilin 1 (PS1) mutations with mice over-expressing RTN3. We found that amyloid deposition in the cortex, the hippocampal CA3 region and the dentate gyrus was significantly reduced in triple transgenic mice compared to bi-transgenic controls. However, reduction of amyloid deposition in the hippocampal CA1 region, where RIDNs predominantly formed prior to amyloid deposition, was less significant. Hence, preformed RTN3 aggregates in RIDNs clearly offset the negative modulation of BACE1 activity by RTN3. Furthermore our study indicates that the increased expression of RTN3 could result in an alteration of BACE1 intracellular trafficking by retaining more BACE1 in the ER compartment where cleavage of APP by BACE1 is less favored. Our results suggest that inhibition of RTN3 aggregation is likely to be beneficial by reducing both amyloid deposition and the formation RIDNs.
Neuritic plaques, a hallmark of Alzheimer’s disease (AD) pathology, refer to β-amyloid peptide (Aβ) deposits surrounded by activated microglia, reactive astrocytes and dystrophic neurites (Tanzi and Bertram, 2005). Dystrophic neurites with the characteristic morphology of swollen dendrites and/or axons are recognizable by antibodies specific to various proteins including ubiquitin (Kowall and Kosik, 1987;Onorato et al., 1989;Perry et al., 1987), neurofilament (Dickson et al., 1999) and GAP-43 (Masliah et al., 1992). We have recently demonstrated that a distinct population of dystrophic neurites, marked by an antibody specific to reticulon 3 (RTN3), represents more abundant dystrophic neurites (Hu et al., 2007). At the current stage, the spatial and temporal relationship among dystrophic neurites that are immunoreactive to different antibodies remains unclear.
The formation of dystrophic neurites is often regarded as an event downstream to amyloid deposition because transgenic mice expressing familial mutant amyloid precursor protein (APP) produce dystrophic neurites that surround amyloid plaques (Games et al., 1995;Holcomb et al., 1998). However, dystrophic neurites are not commonly found to surround Aβ deposits in a diffused form (Joachim et al., 1989), suggesting that Aβ alone may not be sufficient to trigger the formation of various dystrophic neurites in human. Whether dystrophic neurites will form prior to amyloid deposition and whether the occurrence of dystrophic neurites will affect the formation of amyloid deposition are important questions to be answered. In this study, we have addressed these questions by using an animal model over-expressing RTN3.
RTN3 is a member of the reticulon (RTN) family of proteins that have demonstrated neurological functions (Oertle and Schwab, 2003;Yan et al., 2006). We and others have shown that RTN proteins, particularly neuronal RTN3, interact with BACE1 and that this interaction negatively modulates BACE1 cleavage of APP (He et al., 2004;He et al., 2006;Murayama et al., 2006;Wojcik et al., 2007). While examining the in vivo role of RTN3 in AD pathogenesis, we have also found that RTN3 is enriched in RTN3 immunoreactive dystrophic neurites (RIDNs) in AD brains. More importantly, transgenic mice overexpressing RTN3 (Tg-RTN3) develop RIDNs predominantly in their hippocampi (Hu et al., 2007), and this correlates with the formation of RTN3 aggregates in susceptible brain regions. Because of this unique feature present in this animal model, we asked two important questions in this study: 1) whether increased expression of RTN3 would reduce amyloid deposition via negative modulation of BACE1 activity; 2) whether preformed RIDNs in this model would affect amyloid deposition due to the presence of RTN3 aggregation. To address these questions, Tg-RTN3 mice were bred with transgenic mice expressing Swedish mutant APP and mutant presenilin 1 (PS1). The brain samples from the triple-genic mice were compared with the parental bi-genic mice. Although RTN3 overexpression reduced amyloid deposition in most brain areas, the aggregated RTN3 in preformed RIDNs reduced the negative modulation of BACE1 activity by RTN3. The knowledge from this study has advanced our understanding of the in vivo role of RTN3 in AD pathogenesis and also of the impact of dystrophic neurites on amyloid deposition.
Tg-RTN3 mice were generated in the lab as described previously (Hu et al., 2007). Briefly, Tg-APPsw/PSEN1DE9 mice (Tg-PA) were purchased from Jackson Laboratory (stock #004462). Tg-RTN3;APPsw/PSEN1DE9 mice (Tg-R3PA) were generated by crossing of Tg-RTN3 mice with APPsw/PSEN1DE9 mice. All mice in the study were maintained and used according to the protocols approved by the Institutional Animal Care and Use Committee at the Cleveland Clinic. HR3M cells were the HEK-293-derived cells that stably expressed myc-tagged RTN3 (He et al, 2004). HM cells were the HEK-293-derived cells that stably expressed HA-tagged BACE1 (Qahwash et al., 2003). Antibodies against β-amyloid peptide (6E10), calnexin (C4731), myc tag, β-actin (A5441) and APP C-terminus (A8717) were purchased from Sigma- Aldrich. Antibodies R454 and R458 recognized the N-terminus and C-terminus of RTN3, respectively, and were generated by the lab (He et al, 2004). The antibody against BACE1, B279 was made by the lab and the 3d5 antibody was a generous gift from Dr. Robert Vassar (Northwestern University Feinberg School of Medicine, Chicago, IL). Alexa Fluor 488-and 568-labeled secondary antibodies were purchased from Invitrogen. DAB (3,3′ Diaminobenzidine Tetrahydrochloride, D5905) and thioflavine S were purchased from Sigma-Aldrich. Avidin biotin complex elite kit was purchased from Vector Laboratories (Burlingame, CA). Complete protease inhibitor tablet was purchased from Roche Biosciences (Palo Alto, CA). 4–12% and 12% Bis–Tris NuPage gels and 16% Tris-glycine gels were purchased from Invitrogen (Carlsbad, CA). Super Signal West Pico kit and cell surface protein biotinylation kits were purchased from Pierce (Rockford, IL).
Immunohistochemical and confocal experiments were performed according to standard methods as described previously (He et al., 2007). Briefly, the right half of brains of Tg-R3PA and Tg-PA mice (180 days old) were sagittally sectioned at 16-μm thickness using a cryostat after 4% paraformaldehyde fixation and O.C.T. embedding. In immunohistochemical staining, the fixed brain sagittal sections were initially reacted with anti-Aβ monoclonal antibody 6E10, and subsequently detected by the complex of HRP-conjugated secondary antibody and DAB. Thioflavine S staining was performed as follows. After an initial wash in distilled water, prefixed frozen brain sections were incubated in 0.3% TritonX-100 in PBS (1X) for 30 min. Following an additional wash in water, the sections were stained with 1% Thioflavine S in water. Then, the stained sections were incubated in 70% ethanol for 5 min, washed again, mounted with Vecta Shield mounting media, and examined using Leica fluorescence microscopy and software Magna FIRE. For the Aβ plaque examination, sixteen brain sagittal sections (160 μm apart) per mouse were immunolabeled. Aβ deposition was detected with primary antibody 6E10 (1:1000 dilution) and secondary antibody anti-mouse IgG conjugated with Alexa Fluor 568 (1:2000 dilution). Both cellular RTN3 and RIDNs were detected with primary antibody R458 (1:1000 dilution) and secondary antibody anti-rabbit IgG conjugated with Alexa Fluor 488 (1:1000 dilution). The images were examined and captured with a Leica confocal microscope. Laser 581nm was used for Alexa Fluor 568 and the fluorescence signals were displayed in the red channel, whereas laser 488nm was used for Alexa Fluor 488, and results were visualized in green channel.
Quantification of β-amyloid plaques was conducted on the images captured from brain sagittal sections stained with antibody 6E10 by either immunohistochemical or immunoconfocal staining (Borchelt et al., 1997). Sections of each mouse brain were imaged and the areas and densities of the plaques were measured by the Image J software (NIH). The total counts of neuritic Aβ plaques in sixteen sections per mouse brain were determined in an unbiased fashion (Dong et al., 2004). Group means of plaque numbers were calculated based on a sum of the counts from individual animals in each group. The β-amyloid plaque area (plaque load) was determined and the images captured with a Leica confocal microscope (20x objective) and quantified using the image J software (NIH) as a percentage of either total cerebral cortical area or hippocampal area or sub-hippocampal area from 8 sections in each animal. An intensity threshold level was established for discrimination between plaque immunoreactivity and background labeling. The threshold for detection was held constant throughout the image quantification. Vascular signals were excluded in the analysis.
Statistical analysis was performed using GraphPad Prism. One tailed Student’s t-test was used to calculate statistical significance between two groups of data. One-way ANOVA was performed when statistical analysis involves three or more groups. F-test is automatically conducted by the software.
The insoluble Aβ1–40 and Aβ1–42 were extracted from the aggregated Aβ1–40 and Aβ1–42 in the brain tissues by the guanidine hydrochloride method (Wang et al., 2006). The levels of Aβ1–40 and Aβ1–42 in the left cerebral cortex and hippocampus were separately analyzed using sandwich ELISA according to lab standards as previously described (Yan et al., 1999).
Frozen dissected tissues of mouse brains were used for Western blot analysis. Brain tissues were homogenized in 1% CHAPS dissolved in (1x) PBS solution, Roche complete protease inhibitors (1x) and 0.1 mM phosphatase inhibitor Na3VO4 on ice. Total proteins were extracted from the homogenates by rotating on a rotator for 30 min at 4°C. After centrifugation at 15,000 × g for 120 min, supernatants were collected. Total protein concentrations of the supernatants were determined with the BCA protein assay reagent (Pierce). Equal amounts of the supernatant (40 μg total protein) were resolved on 4–12% bis–tris NuPage gels followed by standard Western blotting with indicated antibodies. Chemiluminescent signals for proteins on Western blots were scanned and their integrated density values were calculated with a chemiluminescent imaging system (Alpha Innotech, San Leandro, CA).
The HR3M and HEK293 cell lines were first grown in DMEM media for 24hrs in 10mm plates to about 70% confluence and then cells were transfected with BACE1-expressing plasmid DNA for 24h using the lipofectamine transfection reagent according to the manufacturer’s guidelines. Transfected cells were subjected to subcellular fractionation on a sucrose gradient according to procedures described previously (Yan et al., 2001). The gradients were centrifuged at 22,000 rpm for 18h at 4°C in a Beckman SW41Ti rotor. 12 × 1ml fractions were collected from the top of the gradient and the distributions of BACE1, RTN3 and APP along the gradient were analyzed by western blot. Antibodies B279, APP C-terminal and R458 were used for detection of BACE1, APP and RTN3, respectively. Antibodies against calnexin, β-COP, TGN38 and EEA1 were used to indicate the compartments of ER, Golgi and early endosomes, respectively. Alternatively, HM cells were transfected with either empty vector or RTN3 and subcellular compartments of transfected cells were fractionated on a iodixanol gradient according to published procedures (Xia et al., 1998).
Detection of RTN3 on the cell surface was performed in two different experimental conditions. First, BACE1 expression plasmid vector was transiently transfected in both HEK-293 cells and HR3M cells to enhance detection of BACE1 because of the existing low detection sensitivity of BACE1 antibody. After 24-h transfection, cell surface proteins were biotinylated by incubation of the live transfected cells with 0.25 mg/ml EZ-Link Sulfo-NHS-SS-Biotin at 22°C as described (Yan et al., 2001). Briefly, after quenching and washing, biotinylated cells were lysed in lysis buffer (20mM Hepes at pH 7.9, 10 mM KCl, 1 mM EDTA, 1 mM EGTA, 1% Nonidet NP-40, 10% Glycerol and protease inhibitors) at 4°C for 30 min on rotator and centrifuged for 5 min at 15,000 × g. Supernatants were collected as total proteins. Portions of the total proteins were used in the neutravidin pull-down experiment to isolate biotinylated surface proteins. The total proteins and the surface proteins were then resolved on 12% Bis–Tris NuPage gels and analyzed by Western blotting. In the second experiment, BACE1-expressing HM cells were transfected with 10 μg of either empty vector DNA or RTN3-expressing vector DNA for 48 h. Biotinylation and pull-down experiments were performed according to the same above procedures. Western blot analysis was performed as outlined above.
We have previously shown that overexpression of RTN3 increases its interaction with BACE1, thereby sequestering BACE1 to cleave its APP substrate (He et al., 2004). This finding led us to hypothesize that RTN3 plays a role in AD via suppressing Aβ production and reducing amyloid deposition in vivo. To test our hypothesis, we first generated transgenic mice over-expressing RTN3 driven by a mouse prion gene promoter. As described recently (Hu et al., 2007), among three lines of generated Tg-RTN3 mice, Line 3 (L3) mice express the highest levels of the RTN3 transgene. Western blot analysis of protein lysates from this line of mice showed that the levels of RTN3 in Tg-RTN3 mice were about 4-fold higher than that in wild-type mice (Figure 1A). BACE1 levels were not noticeably affected by the increased expression of RTN3 in this line of mice (Figure 1A). Similarly, the levels of high molecular weight APP species were also comparable in both genotypes of mice (Figure 1A). However, the levels of CTF99, a C-terminal fragment of APP cleaved by BACE1, were slightly reduced (Figure 1A). When the ratio of CTF99 over APP was calculated, there was a clear reduction of CTF99 levels in Tg-RTN3 mice (0.18±0.02 vs 0.11±0.01; n=3, P<0.001). APP was also cleaved by BACE1 at the β′ site to produce CTF89 (Tomasselli et al., 2003;Huse et al., 2002). However, under some electrophoresis conditions, CTF89 was often inseparable from CTF83, a C-terminal membrane bound product of APP following cleavage by α-secretase within the Aβ domain (comparing Figure 1 to Supplemental Figure 1). When the ratio of CTF99 over CTF83, which contained CTF89, was calculated, this reduction of CTF99 was also obvious (0.14±0.02 vs 0.07±0.01; P<0.001, n=3). Together, these data indicate that higher in vivo RTN3 levels cause reduced cleavage of APP by BACE1. Interestingly, the reduction of CTF99 levels was only repeatedly seen in lysates prepared from the cerebral cortex of Tg-RTN3 mice, and was less obvious in lysates from the hippocampus (Figure 1B; CTF99/APP: 0.12±0.02 in Tg-R3PA vs 0.09±0.01 in Tg-PA; P<0.05, n=3). This phenomenon of regional disparity will be further discussed below.
Since amyloid deposition has never been formed in mice expressing only endogenous APP, we bred Tg-RTN3 L3 mice with the Tg-APPswe/PSEN1ΔE9 bi-transgenic mice (abbreviated as Tg-PA mice) to generate Tg-RTN3/APPswe/PSEN1ΔE9 triple transgenic mice (abbreviated as Tg-R3PA) in order to determine whether increased expression of RTN3 would affect amyloid deposition. Tg-PA mice express both the familial mutant PS1 with a deletion of exon 9 and the Swedish mutant APP under the control of a prion promoter, and this line of mice is widely used as an AD model to replicate amyloid deposition in animals (Borchelt et al., 1997). Examination of triple transgenic mice (Tg-R3PA) allowed us to detect the effect of RTN3 on Aβ production and amyloid deposition in comparison to the Tg-PA mice, which normally develop amyloid plaques at six months of age. Upon reaching six months of age (exactly at 180 days), Tg-PA and Tg-R3PA mice were killed and their brains were collected for both biochemical and immunohistochemical examinations that included measuring Aβ levels by ELISA and quantification of Aβ plaque density. To minimize the potential gender effect on Aβ production and amyloid deposition, only female mice were recruited in our present biochemical and immunohistochemical studies. To obtain consistency, the left half of brains of both Tg-PA mice and Tg-R3PA mice were used for biochemical assays and the right half was used for immunohistochemical analysis.
In immunohistochemical staining of fixed brain sagittal sections, Aβ deposits in the cerebral cortex of Tg-R3PA mice, detected by monoclonal antibody 6E10 that specifically reacts with human Aβ N-terminal region, were visibly fewer than those in the comparable cortical region of Tg-PA mice (Figure 2A–B). However, the reduction in the number of condensed Aβ deposits in the entire hippocampus of Tg-R3PA mice was less significant (Figure 2A–B).
Quantification of Aβ plaque numbers in the cerebral cortex and hippocampus of Tg-R3PA and Tg-PA mice revealed that the average number of cortical neuritic Aβ plaques was 661.2 ± 148.2 in 16 brain sagittal sections of Tg-R3PA mice (140 μm apart between two sections; n=4), which is significantly lower than the corresponding plaque number of 1605 ± 177.4 in the comparable sections of Tg-PA mice (n=5) (reduced by 55.5%; P<0.01, unpaired t-test).. In addition, we also determined the area occupied by the compact Aβ plaques using Image-J software to quantify Aβ plaque density or load. Our result showed that 0.342 ± 0.0743% of the examined frontal cerebral cortex of PA mice was occupied by condensed Aβ deposition, whereas only 0.124 ± 0.0567% of the examined area of R3PA mice was occupied by condensed Aβ deposition (Figure 2C; decreased by 63.8%; P<0.05, unpaired t test).
To further confirm the above observation, we stained Aβ plaques by another commonly used method, thioflavin-S staining, and the results also showed fewer plaques in the cortical region of Tg-R3PA mice than in the corresponding region of Tg-PA mice (Figure 2D–G). The difference in the hippocampal regions was smaller (Figure 2F–G), consistent with the above immunohistochemical staining results.
Clearly, a differential effect of RTN3 on amyloid deposition exists, consistent with the altered CTF99 levels in Tg-RTN3 mice. To our knowledge, this regional differential effect on Aβ deposition appears unique considering the fact that expression of the RTN3 transgene in these two regions is similar. This differential effect suggests the presence of additional factors in either cerebrum or hippocampus that led to this disparity. Based on our observation that RIDNs were spontaneously produced mainly in the hippocampus but not in the cerebrum of Tg-RTN3 mice (Hu et al., 2007), we speculated that the pre-formation of RIDNs might have caused this regional disparity.
To add further support to our above hypothesis, we performed confocal examination by double-labeling using monoclonal antibody 6E10 to detect amyloid plaques and RTN3-C-terminus-specific antibody R458 to detect RIDNs. We found that large amounts of dispersed RIDNs were found only in the Tg-R3PA hippocampal CA1 area, spreading to CA3 with a few also identified in the dentate gyrus (Figure 3D). RIDNs in these areas were first formed at the age of three months (Shi et al., 2009). Dispersed RIDNs were scarce in the cortex at this age (Figure 3A–B), consistent with our previous report (Hu et al., 2007). Noticeably, increased expression of RTN3 actually promoted formation of RIDNs in surrounding amyloid deposits (arrowhead in Figure 3B). Overall, the size of amyloid plaques was larger in Tg-PA mice than in Tg-R3PA mice while RIDNs were more obvious in Tg-R3PA mice than in Tg-PA mice (Figure 3).
As shown in Figure 2C, the Aβ plaque load in the entire hippocampus was not significantly different between Tg-R3PA and Tg-PA mice. Since RIDNs are formed mainly in the CA1 region and less in the CA3 and dentate gyrus regions, further quantification of these sub-regions was performed. We found that only the number of condensed Aβ plaques in the CA1 region of Tg-R3PA mice (enriched in RIDNs) as not significantly reduced in comparison to that of Tg-PA mice (34.0 ± 6.26 per 16 sections vs 47.3 ± 4.67 per 16 sections; P>0.05, unpaired t-test), while the plaque number was significantly reduced in the Tg-R3PA mouse CA3 (19.2 ± 5.54 vs 38.0 ± 3.46 per 16 sections; reduced by 49.5%, P<0.05) and dentate gyrus regions (50.4 ± 14.2 vs 90.0 ± 5.03 per 16 sections; reduced by 44.0%; P<0.05) (Figure 3E). An alternative quantification that measures plaque number per square mm of examined samples also demonstrated the consistency of these results (Figure 3F).
Collectively, we have shown here that the increased level of RTN3 largely and significantly reduced the formation of Aβ plaques in R3PA cerebrum and most of the hippocampus. The offset inhibitory effect in the hippocampal CA1 region by increased expression of RTN3 is likely related to the pre-formed RTN3 aggregates in RIDNs. We have previously demonstrated that the occurrence of RIDNs in the Tg-RTN3 mouse hippocampus correlates with the formation and levels of RTN3 aggregates (Hu et al., 2007), and that BACE1 does not interact with RTN3 aggregates (He et al., 2006). Although BACE1 is also suggested to mark dystrophic neurites (Zhao et al., 2007), we found that RIDNs did not overlap with the dystrophic neurites marked by BACE1 antibody (data not shown), indicating that BACE1 and aggregated RTN3 are partially segregated into two different populations of neuritic regions. Since RIDNs formed earlier than amyloid plaques in our animal models, preformed RIDNs appeared to adversely affect the process of amyloid deposition in the CA1 region.
The levels of Aβ, especially Aβ1–42, govern the process of amyloid deposition in AD (Sisodia and George-Hyslop, 2002;Tomita and Iwatsubo, 2004;Golde and Younkin, 2001). As previously reported in AD mouse studies, Aβ1–42 is much easier to aggregate than Aβ1–40, and the aggregated Aβ1–42 and Aβ1–40 are dissolvable in guanidine hydrochloride (Jankowsky et al., 2004). To measure Aβ1–40 and Aβ1–42 levels by sandwich ELISA, total Aβ was extracted from the specified mouse brain samples by the guanidine hydrochloride method as described previously (Wang et al., 2006). The Aβ1–40 (29.9 ± 1.97 vs 39.6 ± 2.17 pmol/g) and Aβ1–42 levels (377.4 ± 50.72 vs 596.9 ± 14.68 pmol/g) were significantly lower in the cerebral cortex of Tg-R3PA mice than those in Tg-PA mice (Figure 4A, *P<0.05, **P<0.01, n=5). Noticeably, the Aβ1–40/Aβ1–42 ratio was not altered (data not shown), consistent with our previous observation of the inhibitory effects of RTN3 on BACE1 in vitro (He et al., 2004). The hippocampal levels of insoluble Aβ1–40 (39.4 ± 4.58 vs 45.9 ± 6.44 pmol/g; P=0.229) and Aβ1–42 (309.9 ± 42.60 vs 364.3 ± 62.22 pmol/g; P=0.255) in Tg-R3PA mice were also lower than that in Tg-PA mice (Figure 4B). Although there appears to be a smaller reduction in Aβ levels than in amyloid deposition, previous studies also showed similar phenomena in other animal models including BACE1 heterozygous mice (McConlogue et al., 2007). Overall, our ELISA results correlate with the immunohistochemical results that show reduced production and deposition of Aβ in the cerebral cortex of Tg-R3PA mice.
We also performed Western blot analysis on total protein extracts from cerebral cortical homogenates of Tg-PA and Tg-R3PA mice (6 months old) to examine the levels of the BACE1-cleaved C-terminal product CTF99, which was assessed by an antibody recognizing either Aβ N-terminus (6E10) or APP C-terminus (A8717). It should be noted that the brain full-length APP and secreted N-terminal fragments were not separable on the Western blot. The total levels of these proteins appeared to not be significantly altered, reflecting that expression of APP in Tg-R3PA mice is comparable to that in Tg-PA mice. While the level of RTN3 was validated to be about 4-fold higher in Tg-R3PA cerebrum, the level of APP CTF99 was noticeably reduced, especially when 6E10 was used to detect human APP CTF99 (Figure 5A). This slight reduction was clearly manifested by comparing the ratios of CTF99 over either CTF83 (1.97±0.07 vs 1.64±0.12; P<0.05, n=4) or total APP (0.96±0.04 in Tg-R3PA vs 0.73±0.07 in Tg-PA; P<0.05, n=4). On the other hand, the total hippocampal CTF99 levels were not significantly reduced, consistent with the morphological and Aβ results (Figure 5B; CTF99/APP: 1.31±0.07 in Tg-R3PA vs 1.26±0.03 in Tg-PA; P<0.05, n=4). Thus, reduced amyloid deposition in R3PA mice is attributable to reduced BACE1 activity in R3PA mice and preformed RIDNs affecting negative modulation of RTN3 on APP processing.
Although both our in vitro and in vivo studies have demonstrated that increased expression of RTN3 will bind to and negatively affect BACE1 activity by preventing BACE1 access to its APP substrate, the physiological function of RTN3 upon its interaction with BACE1 is still unclear. To further address this mechanistic question, we tested the potential role of RTN3 in cellular trafficking as discussed previously (Wakana et al., 2005). We reasoned that the cellular distribution of BACE1 would be affected if increased expression of RTN3 would alter BACE1 trafficking. To investigate the potential effect of RTN3 on the intracellular localization of BACE1, we first performed sucrose gradient centrifugation to fractionate the cellular compartments of the HEK-293 and HR3M cells. HR3M cells were established by stably expressing RTN3 in HEK-293 cells (He et al., 2007). Our result showed that in HR3M cells, the majority of BACE1 was localized in the ER fractions, leaving a small amount of BACE1 in the Golgi fractions (Figure 6A). In contrast, a significant portion of BACE1 was localized in the Golgi of the HEK-293 cells (fractions 3 and 4) and a much smaller portion was localized in the early endosomes (fraction 2), consistent with previous reports (Yan et al., 2001;Creemers et al., 2001;Pastorino et al., 2002). We also found that endogenous levels of APP detected with antibody A8717 were mostly in the fractions containing ER and plasma membrane proteins in both cases, but appeared more obvious in HR3M cells. It should be noted that full length APP levels were higher in HR3M cells than in HEK-293 cells, consistent with our previous reports that increased expression of RTN3 elevates the levels of full length APP in cultured cells (He et al., 2004) and in sciatic nerves (Shi et al., 2009).
To further confirm this observation, we transfected either empty vector or RTN3 expression construct into HM cells that stably express BACE1. After transfection for 48 hrs, these cells were subjected to subcellular fractionation using the iodixanol gradient method according to procedures described previously (Xia et al., 1998). We found that BACE1 displayed bi-modal distribution in the ER and Golgi fractions in the control-transfected cells but significantly more BACE1 was enriched in the ER compartments when RTN3 was significantly overexpressed (Figure 6B).
It has been previously shown that a small percentage of BACE1 is detectable on the cell surface (Huse et al., 2000;Yan et al., 2001;Zou et al., 2007;Kinoshita et al., 2003;Hu et al., 2006). Potentially, the increased retention of BACE1 in the ER compartment would reduce surface expression of BACE1. To test this possibility, we compared the cell surface localization of BACE1 in HEK-293 cells or HR3M cells by employing biotin-labeling of surface proteins. To confirm this, we transiently transfected these two cell lines with equal amounts of HA-tagged human BACE1 expression construct and the transfection was allowed to proceed for 24 hrs. Cell surface proteins were then biotinylated as described previously (Yan et al., 2001), and biotinylated proteins were specifically pulled-down with neutravidin-beads for Western blot analysis. Although the expression of HA-tagged BACE1 in HR3M cells was similar to that in HEK-293 cells, the ratio of surface BACE1 to total cellular BACE1 was significantly lower in HR3M cells compared to HEK-293 cells (Figure 7A). Based on quantification from three independent sets of experiments, we found that 7.67 ± 0.649% of total cellular BACE1 in HEK-293 cells was detectable on the cell surface, but only 3.40 ± 0.486% of total BACE1 in HR3M cells was detected on the cell surface, reflecting an approximately 55% reduction of surface BACE1 upon increased expression of RTN3 (Figure 7B, P<0.01, n=3). In an alternative experiment, BACE1-stably-expressed HM cells were transiently transfected with either RTN3 plasmid DNA or empty vector for 48 hrs. The biotinylation and pull-down experiments were conducted in the same manner as above. Consistent with the data outlined above, we observed a significant reduction of surface BACE1 in cells upon overexpression of RTN3 (Figure 7D, E). The percentage of surface BACE1 in the total BACE1 was 5.49 ± 0.679% in the control, whereas it was decreased to 1.36 ± 0.0694% when RTN3 was overexpressed.
Higher levels of RTN3 are expected to reduce BACE1 processing of APP. Consistent with this, total full length APP was indeed significantly higher in HR3M cells than in HEK-293 cells (P<0.01, n=3, t test; Figure 7A, D). We also verified that mRNA levels of APP in these two cell lines were similar (data not shown), thereby confirming that this elevated level of full length of APP is due to decreased processing of BACE1 as discussed previously (He et al., 2004;Murayama et al., 2006). Interestingly, the cellular trafficking of APP was also altered as evidenced by altered subcellular fractionation (Figure 6A) and the percentage of APP on the cell surface (Figure 7C and F, P<0.01, n=3). This reduced surface level of APP in cells was in line with the observation of fractionation on the sucrose gradient..
Altogether, we demonstrated that increased expression of RTN3 indeed altered cellular trafficking of BACE1 and APP. An increased retention of BACE1 in the ER compartment is also not favorable for optimal processing of APP by BACE1, which requires an acidic pH environment. Unlike BACE1, membrane trafficking of APP might have been indirectly affected by increased expression of RTN3 because direct interaction between RTN3 and APP was not detected. Whether the altered interaction between BACE1 and APP upon overexpression of RTN3 contributes to the altered cellular trafficking of APP remains to be investigated.
RTN proteins are a group of integral membrane proteins predominantly associated with the ER (Yan et al., 2006;Oertle et al., 2003). While their biological functions remain poorly understood, diverse biochemical and functional studies suggest various potential roles of these proteins, including the shaping of the tubular ER, regulated trafficking of vesicle proteins, neurite growth and exocytosis (GrandPre et al., 2000;Kiseleva et al., 2007;Steiner et al., 2004;Chen et al., 2000;Voeltz et al., 2006;Hu et al., 2007). Altered expression levels of RTNs are found in a variety of disease states, including neurological disorders (Yan et al., 2006;Fergani et al., 2005;Bandtlow et al., 2004). We and others have demonstrated that RTNs interact with BACE1 and negatively modulate BACE1 protease activity (He et al., 2004;He et al., 2006;Murayama et al., 2006;Wojcik et al., 2007;Kume et al., 2009). In the present study, we further demonstrated that increased expression of RTN3 in our mouse models significantly reduced Aβ production, amyloid deposition and plaque load. Specifically, the over-expression of RTN3 in Tg-R3PA mice significantly decreased the number of neuritic Aβ plaques by about 55% in the cerebral cortex, 50% in the CA3 region, and 44% in the dentate gyrus region compared to the same regions in Tg-PA mice. On the other hand, our results are the first to suggest that pre-formed RIDNs largely in the Tg-R3PA CA1 region adversely affect the negative modulation of BACE by RTNs, thereby affecting the process of amyloid deposition. There are many potential mechanisms that could cause the formation of dystrophic neurites in AD patients, and the formation of RIDNs correlates with the formation of RTN3 aggregates. Our results imply that inhibition of RTN3 aggregation will reduce the formation of RIDNs and amyloid deposition in AD.
RIDNs in Tg-RTN3 mice occur as early as 3 months of age (Shi et al., 2009), which is earlier than the onset of the Aβ plaque formation in Tg-PA mice (about 5–6 months of age). The hippocampal CA1 region is the predominant region of this formation. In our study, the expression levels of RTN3 in the hippocampus and cerebral cortex were comparable (Figure 1) even though most of the RIDNs developed in the hippocampal CA1 region and only a few RIDNs were scattered in the cerebral cortex. The observed biochemical difference between the hippocampus and the cerebral cortex was perhaps due to the existence of a large number of RTN3 aggregates in the hippocampus but not in the cerebral cortex of Tg-R3PA mice (Hu et al., 2007). It remains to be determined why the hippocampus is more susceptible to the formation of RTN3 aggregates. However, because of this difference, the overall beneficial effect of RTN3 over-expression was significantly diminished in the hippocampus of Tg-R3PA mice as the insoluble Aβ1–40 and Aβ1–42 levels as well as the amounts of dystrophic Aβ plaques were lowered to a lesser extent. We have recently demonstrated that aggregated RTN3 does not interact with BACE1, whereas the free-form of RTN3 binds to BACE1 and inhibits it (He et al., 2006), indicating that monomeric status is critical for RTN3 inhibitory function of BACE1. RTN3 adopts a ω-shape structure with two ends facing the cytosolic side (He et al., 2007), and its C-terminal QID motif is required for the interaction with the BACE1 C-terminal membrane-proximal region (He et al., 2006). The interaction of RTN3 with BACE1 is essential for the inhibitory role of RTN3 in the attenuation of BACE1-mediated APP processing. As the major component of RIDNs is aggregated RTN3, the formation of RIDNs on a large scale reduces available RTN3 monomer intracellularly, which in turn liberates BACE1 from the RTN3-mediated inhibition.
If pre-formation of RIDNs in the mouse hippocampus is avoidable, inducing expression of RTN3 could be an alternative approach to achieve reduction of amyloid deposition. We found that over-expression of RTN3 in regions other than CA1 sufficiently caused a reduction in amyloid plaque formation, although it did not completely block Aβ production. Partial reduction of BACE1 activity has been shown to reduce amyloid deposition (Singer et al., 2005;McConlogue et al., 2007;Laird et al., 2005), and our results are in line with these observations. It should also be noted that the APP model used in the present study differs from the PDAPP or Tg-2576 mice used in the above publications, but inhibition of BACE1 is commonly believed to reduce amyloid deposition.
The mechanism by which increased expression of RTN3 reduces BACE1 processing of APP was investigated by testing the potential role of RTN3 in trafficking of membrane proteins. Genetic studies suggest that the RTN family of proteins may shape tubular structures (Tolley et al., 2008;Voeltz et al., 2006), but other studies show that RTN proteins can affect vesicle trafficking through interactions with vesicle transporter proteins (Wakana et al., 2005;Liu et al., 2007;Iwahashi and Hamada, 2003;Steiner et al., 2004). Our present study has revealed that increased expression of RTN3 causes retention of more BACE1 in the ER compartments and diminishes the trafficking of BACE1 to the cell surface. The potential explanation for this altered trafficking is that RTN3 is not a fast movable ER trafficking molecule, and interaction between RTN3 and BACE1 causes BACE1 to transit slowly from the ER to the late secretory compartments. In line with this observation, we found that the surface BACE1 was reduced significantly when RTN3 was over-expressed. Interestingly, cellular trafficking of APP was also affected upon over-expression of RTN3 (Figures 6 and and7).7). Although RTN3 does not interact with APP directly based on our co-IP experiments (He et al., 2004), the increased retention of APP in the ER may be related to elevated levels of BACE1 in the ER in this circumstance, and an interaction between APP and BACE1 may indirectly retain more APP in the ER. Increased retention of APP in the ER would explain the slightly reduced levels of CTF83, a processing product of APP by α-secretase as shown in Figure 1.
Altered intracellular trafficking and localization of BACE1 are expected to affect APP metabolism and Aβ production. Normally, the majority of BACE1 is present in the trans-Golgi network (TGN) and endosomes while the rest exists on the cell surface and in the ER (Yan et al., 2001;Walter et al., 2001;Creemers et al., 2001;Lin et al., 2000;Vassar et al., 1999). Cleavage of APP at the β-site by BACE1 occurs in both early secretory compartments and endocytic pathways (Yan et al., 2001;Kinoshita et al., 2003;Shi et al., 2001;Huse et al., 2002;He et al., 2005). Consequently, more BACE1 is accumulated in the ER environment in which processing of APP by BACE1 is in a less favorable pH condition as suggested previously (Yan et al., 1999;Vassar et al., 1999;Sinha et al., 1999;Hussain et al., 1999;Lin et al., 2000). Increased retention of BACE1 in the ER is expected to enhance degradation of BACE1 as suggested recently (Tesco et al., 2007). Perhaps tight interaction of RTN3 with BACE1 prevents not only access of BACE1 to its APP substrate for proper processing but also its degradation. While the elevated levels of RTN3 can retain BACE1 in the ER compartment, it is unclear whether the exit of BACE1 out of the ER compartment requires RTN3. The complete deficiency of RTN3 and its family members will be tested in order to address this question.
In summary, our findings have established the dual roles of RTN3 in vivo, and our results suggest that altered cellular trafficking of BACE1 in vitro and in vivo can also account for reduced production of Aβ and its deposition in the brain. Importantly, our results also indicate that the preformed RIDNs will adversely favor the production of Aβ, and higher levels of RIDNs than amyloid plaques have already been found in the hippocampal region of AD postmortem brains (Shi et al., 2009). Hence, the prevention of the formation of RIDNs will be a necessary step to enhance the effect of RTN3 on amyloid deposition.
We thank Dr. Xiangdong Zhou for much help during the studies and Chris Nelson for critic reading of this manuscript. Dr. Robert Vassar (Northwestern University) provided us with the BACE1-specific monoclonal antibody that was used for confocal staining of BACE1 in brain samples. This work is supported by NIH grants to RY (AG025493), and awards from Ralph Wilson Foundation and Alzheimer’s Association to RY. QS is partially supported by a young investigator award from NARSAD and MP is supported by AHAF postdoctoral fellowship.