Address correspondence and reprint requests to Dr. G. W. Rebeck at Alzheimer Research Unit, Massachusetts General Hospital, 149 13th Street, Charlestown, MA 02129, U.S.A.
Abstract : Apolipoprotein E and α2-macroglobulin (α2M) are genetic risk factors for late-onset Alzheimer's disease, and both bind a cell surface receptor, the lowdensity lipoprotein receptor-related protein (LRP). To investigate the role of LRP on preventing the accumulation of β-amyloid peptide (Aβ), we examined the effects of α2M on the clearance of endogenous Aβ. Studies were performed in primary Tg2576 transgenic mouse cortical neuronal cultures expressing human mutant amyloid precursor protein (APP) 695. This system allowed us to follow endogenous Aβ using immunoblots to detect monomeric forms of the peptide. Aβ and APP levels were measured in conditioned media. We found that activated α2M (α2M*) substantially decreased soluble Aβ levels and had no effect on secreted or full-length APP levels. Native α2M, which is not a ligand for LRP, did not affect Aβ levels. The receptor-associated protein, which inhibits interaction of all ligands with LRP in vitro, prevented α2M*-induced decreases of soluble Aβ levels. These data suggest that α2M* affects soluble Aβ clearance rather than Aβ production. Further studies showed that similar Aβ clearance via an LRP-mediated pathway was observed after treatment with another LRP ligand, lactoferrin. Taken together, these data demonstrate that α2M* enhances the clearance of soluble Aβ via LRP in cortical neurons.
The central component of senile amyloid plaques in Alzheimer's disease (AD) is the β-amyloid peptide (Aβ), derived from the proteolytic processing of the parent molecule, named the amyloid precursor protein (APP). Aβ is produced continuously as a soluble peptide, suggesting the existence of clearance mechanisms that normally prevent its accumulation and aggregation. So far, several cell surface receptors have been implicated in Aβ clearance, including the scavenger receptor A (Paresce et al., 1996), the receptor for advanced glycation end products (Yan et al., 1996), and the low-density lipoprotein receptor-related protein (LRP) (Narita et al., 1997).
LRP, a 600-kDa multifunctional cell surface receptor, is associated in late-onset AD from several perspectives. Genetic studies have implicated two ligands of LRP as risk factors in AD : apolipoprotein E (Strittmatter and Roses, 1996) and, more recently, α2-macroglobulin (α2M) (Blacker et al., 1998 ; Liao et al., 1998), Both proteins bind Aβ in vitro (Strittmatter et al., 1993 ; Narita et al., 1997 ; Hughes et al., 1998) and immunostain Aβ deposits in vivo (Namba et al., 1991 ; Strauss et al., 1992). Our current study focuses on the possibility that α2M is involved in Aβ clearance via an LRP-mediated endocytic process. α2M is a large tetrameric protein that has established roles as a pan-proteinase inhibitor and in binding and clearance of various small molecules, including cytokines and growth factors. α2M has a unique mechanism of proteinase inhibition : Proteinases cleave a peptide bond within the “bait” region, leading to a conformation change [from α2M to activated α2M (α2M*)] in the tetramer, “trapping” the protease. This conformational change also makes α2M* a competent ligand for binding and clearance by LRP.
To assess the potential effects of α2M* on Aβ clearance, we examined the chronic effects of α2M* exposure on APP and Aβ levels. We isolated neurons from mice transgenic for human APP695 containing the K670N, M671L double mutation (Hsiao et al., 1996). These neuronal cultures have robust expression of secreted APP and readily detectable levels of Aβ. Primary neurons also express high levels of α2M* receptor/LRP, consistent with other reports (Holtzman et al., 1995 ; Ishiguro et al., 1995 ; Tooyama et al., 1995). The advantages for using this model are, first, that the Aβ level detected in the media was endogenously produced by neurons and is comparable to levels in human CSF (Ida et al., 1996). Second, Aβ levels could be measured by immunoblotting, with which the Aβ band at 4 kDa can clearly be distinguished from the other immunoreactive material, for example, Aβ dimers and multimers.
Previous studies have shown that α2M* can bind Aβ in vitro (Zhang et al., 1996 ; Du et al., 1997 ; Narita et al., 1997 ; Hughes et al., 1998) and that complexes of Aβ with α2M* can be cleared by LRP on fibroblasts (Narita et al., 1997). Here we examined the effects of α2M* on Aβ clearance in cultured cortical neurons. We find that α2M* treatment specifically and significantly increased clearance of soluble Aβ via LRP, without changing the secreted and full-length APP levels. Our data suggest that LRP-mediated clearance of Aβ is potentially an important mechanism for preventing Aβ accumulation in vivo.
Human recombinant receptor-associated protein (RAP) was prepared from a glutathione S-transferase fusion protein as described (Williams et al., 1992). α2M* was isolated from human plasma and activated by treatment with methylamine, as described (Ashcom et al., 1990). Lactoferrin was purchased from Sigma Chemical Co. (St. Louis, MO, U.S.A.). WO-2, a mouse monoclonal antibody against Aβ residues 5-8, was kindly provided by Dr. Konrad Beyreuther (University of Heidelberg) (Ida et al., 1996).
Cortical neurons were isolated from embryonic day 16 Swiss-Webster mice, which had been bred with heterozygous transgenic mice overexpressing human APP695 with the K670N, M671L double mutation (Hsiao et al., 1996), as described (Rebeck et al., 1998). DNA from each embryo was analyzed using the PCR with APP-specific primers to determine its transgenic status. Individual cortices were dissociated in calcium-free saline and plated on poly-D-lysine (Sigma)-coated tissue culture dishes at the density of 1.5 × 106 cells/ml. The neurons were grown in Neurobasal media (GibcoBRL) plus 10% fetal bovine serum. One hour after plating, medium with serum was replaced with medium containing B-27 supplement (GibcoBRL). Treatment with 5-fluoro-2′-deoxyuridine (20 μg/ml) on days 2 and 5 after plating minimized nonneuronal cell proliferation.
Treatment with α2M*
Primary neuronal cultures were treated with 500 nMα2M*. α2M* was added at 4-5 days in vitro, and fresh medium containing α2M* was added at 8 days in vitro. A similar treatment protocol was applied to other LRP ligands or drugs, such as lactoferrin (500 nM) and RAP (1 μM). Conditioned medium was collected at 9-10 days in vitro and stored at -20°C for western blot analysis. Control cultures (sister cultures) were not treated with α2M* or other LRP ligands. Treatment comparisons were only made between neurons isolated from the same embryo.
Aβ and APP analyses
For experiments with primary neurons, equal volumes of conditioned media (50-80 μl per experiment) were analyzed for total Aβ by immunoblot analysis, as described (Ida et al., 1996). In brief, proteins were denatured, reduced, and separated by 10-20% Tris-tricine sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Novex). Proteins were transferred to nitrocellulose at 380 mA for 45 min, and the blotted membrane was boiled in 0.1 M phosphate-buffered saline for 5 min to enhance the signal. Nitrocellulose was blocked with 0.25% bovine serum albumin in phosphate-buffered saline containing 0.05% Tween 20 for 30 min at room temperature. The blots were incubated with 1 μg/ml antibody WO-2 in phosphate-buffered saline containing 0.05% Tween 20 and 0.25% bovine serum albumin overnight at 4°C. Immunoreactivity was detected using horseradish peroxidase-linked anti-mouse IgG, and the blots were developed with chemiluminescent reagent and exposed to film. Analyses with a Bio-Rad GS-700 imaging densitometer were recorded as percentages of the sister cultures (untreated cultures).
For analysis of the full-length APP, cytoplasmic lysates were prepared from primary mouse cortical neuronal cultures treated with α2M*. The neuronal cultures were lysed in 50 mM Tris-HCl (pH 8.0) containing 0.5 M NaCl, 4 μM leupeptin, 2 μM pepstatin, 1.5 μM aprotinin, 400 μM phenylmethylsulfonyl fluoride, and 0.1% Triton X-100. The total cell lysates were centrifuged at 2,000 rpm for 30 s, and the supernatant was analyzed by immunoblotting as mentioned as above.
Results from several cultures were pooled for statistical analyses. Data are expressed as mean ± SEM values. Statistical significance was determined by one-way ANOVA followed by the Fisher post hoc test for multiple comparisons. p < 0.05 was considered indicative of a statistically significant difference.
Characterization of Aβ from primary neurons of Tg2576 transgenic mice
Primary neurons from Tg2576 embryos express high levels of soluble Aβ, which can be detected by immunoblot using the sensitive Aβ antibody WO-2 (Ida et al., 1996). With WO-2, which recognizes Aβ residues 5-8, we were able to detect human Aβ standards at levels from 10 pg to 10 ng (Fig. 1A). This result indicated that the sensitivity of this assay is similar to that reported earlier with WO-2 antibody (Ida et al., 1996). In media from primary Tg2576 transgenic neurons, WO-2 specifically recognizes the 4-kDa Aβ band and the 100-kDa secreted APP band (Fig. 1B). Of importance is that we do not detect high-molecular-weight aggregates or large amounts of oligomers in the conditioned media, suggesting that the majority of soluble Aβ is monomeric.
To determine the effects of α2M* on the levels of soluble Aβ, primary neuronal cultures producing Aβ were treated with α2M*, and the amount of Aβ in the media was measured by western blot. In both control and α2M*-treated neurons Aβ was detectable at 18 h, and levels increased markedly in media conditioned for 5 days (Fig. 2). α2M* treatment diminished the amount of Aβ in a time-dependent manner. Aβ levels were greatly inhibited after incubation for 3 days with α2M*, and this inhibition was enhanced on day 5. For subsequent experiments, we treated cells for 4—5 days, because at this time point the α2M*-induced inhibition is substantial and maintained. In 10 individual embryos, α2M* substantially decreased soluble Aβ levels in conditional media by an average of 65% (Fig. 3A and B). In addition, α2M* does not affect Aβ aggregation as measured in the presence of sodium dodecyl sulfate and reducing agent (Fig. 3A).
To control for nonspecific effects of α2M*, we examined the influence of native α2M (which is not an LRP ligand) on Aβ levels. Native α2M did not influence the levels of Aβ from the cortical neurons (Fig. 3C and D). This result indicates that only α2M* can specifically decrease soluble Aβ levels.
RAP inhibits the effects of α2M* on soluble Aβ
RAP is a chaperone protein that facilitates the proper folding and subsequent trafficking of LRP within the early secretory pathway (Bu and Rennke, 1996). RAP is not present on the neuronal surface or the extracellular matrix, but binding sites for RAP are present on the surface and throughout the cytoplasm of neurons, indicative of the distribution of members of the low-density lipoprotein receptor family, including LRP (Zheng et al., 1994). RAP binds to multiple sites on LRP and antagonizes the binding of all known LRP ligands to this receptor in vitro. To test if the tremendous decrease in Aβ levels induced by α2M* was due to enhanced LRP-mediated clearance, we coincubated cultures with RAP at the concentration of 1 μM, which is known to block the interactions between LRP and all its ligands. The α2M*-induced decreases of soluble Aβ level were abolished by co-incubation with RAP (Fig. 4A and B). RAP alone did not alter the levels of soluble Aβ (Fig. 4C and D) and secreted APP (Fig. 4C). These data suggest that α2M* affects soluble Aβ levels by promoting clearance of Aβ via LRP and not by promoting degradation or aggregation of Aβ.
Effects of α2M* on secreted APP and full-length APP
Next, we examined whether treatment with α2M* had an effect on secreted APP or total cellular APP, to evaluate whether the decline in Aβ content was due to a more generalized effect on APP. The effects of α2M* on secreted APP695 are summarized in Fig. 5, showing that α2M* treatment had no effect on the level of secreted APP (Fig. 5A is representative immunoblot, and Fig. 5B demonstrates mean values of embryos studied). Similarly, no changes were seen in expression of full-length APP (Fig. 5C and D). Thus, the α2M*-induced reduction in levels of soluble Aβ is independent of changes in APP levels.
LRP is a large endocytic receptor that binds multiple ligands. Therefore, it was of interest to determine whether soluble Aβ levels were decreased by another LRP ligand. Among the diverse ligands from LRP involved in AD, lactoferrin is of particular interest due to its robust association by immunostaining in the AD brain (Osmand and Switzer, 1991 ; Rebeck et al., 1995). We therefore tested lactoferrin as a possible carrier for LRP-mediated Aβ clearance. Lactoferin treatment decreased soluble Aβ levels in five individual cultures by 51% (Fig. 6A and B), similar to the α2M*-induced reduction in soluble Aβ levels (Fig. 3). To test if the reduction in Aβ levels by lactoferrin was due to enhanced LRP-mediated clearance, we coincubated cultures with 1 μM RAP. RAP treatment prevented lactoferrin's effects on reducing Aβ levels (Fig. 6C and D). This result indicates that lactoferrin affects soluble Aβ levels by promoting clearance of Aβ via LRP, suggesting that lactoferrin and α2M* share a common pathway for clearance of soluble Aβ.
LRP is a multifunctional receptor with four distinct ligand binding domains and at least 14 identified ligands (Strickland et al., 1995), including apolipoprotein E, α2M, and lactoferrin. It is involved in receptor-mediated endocytosis, directing ligands to degradation via the late endosome and lysosome (Bu et al., 1994). There is a striking relationship between LRP and the Aβ deposits in AD. Not only LRP itself (Rebeck et al., 1993 ; Tooyama et al., 1993) but also most LRP ligands studied so far are components of plaques (Rebeck et al., 1995). We hypothesized that the accumulation of LRP and its ligands in plaques may be a result of impaired LRP-mediated Aβ clearance. In the current study, we found that one of these LRP ligands, α2M*, had a marked effect on Aβ clearance via LRP.
Our cell culture data support the hypothesis that α2M* promotes Aβ clearance rather than either altering APP processing or affecting extracellular Aβ degradation. First, we observed no changes in the levels of secreted and full-length APP from cells expressing APP695, an isoform that is not observed to interact with LRP (Kounnas et al., 1995). Second, our experimental design mitigates against a direct effect of α2M* or RAP on Aβ synthesis because the K670N, M671L double mutant transgene is believed to cause an increase in Aβ levels by altered processing of APP before it reaches the cell surface (Haass et al., 1992, 1995 ; Essalmani et al., 1996). Third, the methylamine-activated form of α2M, which binds with LRP, decreased soluble Aβ levels, but the native form of α2M, which does not bind to LRP, had no effects on soluble Aβ levels. Fourth, reductions in Aβ levels were not observed with RAP treatments alone, suggesting that disruption of LRP interactions did not lower Aβ levels. Finally, α2M* did not alter Aβ levels when LRP clearance mechanisms were blocked with RAP. These data demonstrate that the effects of α2M* on Aβ are via LRP-mediated uptake rather than simply by reducing Aβ production or promoting degradation or aggregation. Our findings of clearance of endogenous Aβ by primary neurons support earlier work showing clearance of preformed Aβ-α2M* complexes by fibroblasts (Narita et al., 1997).
Several groups have shown that nonactivated α2M and α2M* forms bind Aβ (Zhang et al., 1996 ; Du et al., 1997 ; Hughes et al., 1998), and two studies have demonstrated that α2M* binding reduces Aβ aggregation (Du et al., 1998 ; Hughes et al., 1998). Qiu et al. (1996) showed that α2M can promote degradation of Aβ by binding both Aβ and an Aβ-protease. Our data strongly support a role of α2M* in Aβ clearance rather than through the mechanism of increasing Aβ degradation, because (a) native α2M had no effects on Aβ levels, (b) the decrease in Aβ levels after treatment with α2M* could be completely eliminated by RAP, and (c) α2M* had no effects on soluble Aβ levels in the absence of neurons (data not shown). Furthermore, in our system, α2M* does not affect Aβ aggregation as we observed no formation of Aβ aggregates by western blot in either α2M*-treated or untreated cultures.
In addition, we have demonstrated that another LRP ligand, lactoferrin, can enhance LRP-mediated clearance of Aβ. Previous studies have shown that other LRP ligands, including apolipoprotein E, can promote Aβ clearance in smooth muscle cells (Urmoneit et al., 1997) and neurons (Jordan et al., 1998). The simplest model to explain the effect of LRP ligands on Aβ levels is that each can enhance the normal clearance of Aβ by primary neurons, leading to reduced levels of Aβ in the conditioned media. Although no evidence is available for the effect of lactoferrin binding capacity on Aβ, it is possible that lactoferrin binds Aβ, based on the immunostaining of lactoferrin on amyloid in AD brain. Our result that RAP can abolish the lactoferrin-induced decrease in Aβ level suggests that lactoferrin is sharing the same pathway as α2M* for the clearance of soluble Aβ mediated by LRP. In vivo, α2M* (Strauss et al., 1992), lactoferrin (Fillebeen et al., 1998), and apolipoprotein E (Poirier et al., 1991) are up-regulated in response to CNS damages, perhaps in a coordinated way to clear debris. Their presence on Aβ deposits in the AD brain (Rebeck et al., 1995) may represent a common failed mechanism of Aβ clearance.
These data provide a context to interpret the genetic associations of apolipoprotein E, α2M*, and LRP with AD. A deficiency of LRP-mediated clearance in the CNS may predispose toward AD by causing a relative increase in Aβ level. Further evidence of how each of these genetic factors alters Aβ clearance may elucidate a common mechanism underlying Aβ deposition in AD. Thus, enhancing LRP clearance may be a pharmacological approach to decreasing Aβ and its associated pathological changes in vivo.
We thank Dr. Karen Hsiao for providing the Tg2576 transgenic APP mice. We thank Dr. Konrad Beyreuther for the WO-2 antibody. This work was funded by a grant from the American Health Assistance Foundation and grants AG14473 (to G.W.R.) and AG12406 (to B.T.H.) from the National Institutes of Health.