Fibrillar amyloid deposits are defining pathological lesions in Alzheimer’s disease brain and are thought to mediate neuronal death. Amyloid is composed primarily of a 39–42 amino acid protein fragment of the amyloid precursor protein (APP), called amyloid β-protein (Aβ). Because deposition of fibrillar amyloid in vitrohas been shown to be highly dependent on Aβ concentration, reducing the proteolytic release of Aβ is an attractive, potentially therapeutic target. Here, the turnover rate of brain Aβ has been determined to define treatment intervals over which a change in steady-state concentration of Aβ could be measured. Mice producing elevated levels of human Aβ were used to determine approximate turnover rates for Aβ and two of its precursors, C99 and APP. Thet ½ for brain Aβ was between 1.0 and 2.5 hr, whereas for C99, immature, and fully glycosylated forms of APP695 the approximate t ½ values were 3, 3, and 7 hr, respectively. Given the rapid Aβ turnover rate, acute studies were designed using phorbol 12-myristate 13-acetate (PMA), which had been demonstrated previously to reduce Aβ secretion from cells in vitro via induction of protein kinase C (PKC) activity. Six hours after intracortical injection of PMA, Aβ levels were significantly reduced, as measured by both Aβ40- and Aβ42-selective ELISAs, returning to normal by 12 hr. An inactive structural analog of PMA, 4α-PMA, had no effect on brain Aβ levels. Among the secreted N-terminal APP fragments, APPβ levels were significantly reduced by PMA treatment, whereas APPα levels were unchanged, in contrast to most cell culture studies. These results indicate that Aβ is rapidly turned over under normal conditions and support the therapeutic potential of elevating PKC activity for reduction of brain Aβ.
- protein turnover
- amyloid-β protein
- amyloid precursor protein
- Alzheimer’s disease
- phorbol ester
- protein kinase C
The behavioral deficits associated with Alzheimer’s disease (AD) result from the loss of neurons and synapses, primarily in the cortex and hippocampus. Amyloid deposits in these regions are an invariant pathological feature of AD and result from the aggregation of a 39–42 amino acid long protein known as amyloid β-protein (Aβ) (Selkoe, 1996). This protein has been shown to be neurotoxic in vitro when present in an aggregated form (Pike et al., 1993). Evidence obtained from the study of familial forms of AD in which mutations exist in either the amyloid precursor or presenilin genes (Hardy, 1997) implicates elevated secretion of the 42 amino acid form as a likely etiological event in disease development. This form is most fibrillogenic in vitro (Jarrett et al., 1993) and is more abundant in amyloid deposits in the AD brain than shorter, less fibrillogenic forms (Iwatsubo et al., 1994; Savage et al., 1995; Yamaguchi et al., 1995). Under normal conditions, however, soluble Aβ ending at residue 40 is more abundant than the 42 residue form (Haass et al., 1992; Seubert et al., 1992; Shoji et al., 1992;Vigo-Pelfrey et al., 1993).
The amyloid precursor protein (APP) is processed to form Aβ and other derivatives by at least three proteases, identified by their cleavage specificities (Selkoe, 1996). α-Secretase generates a secreted N-terminal APP fragment (APPα) and also destroys the Aβ domain. The remaining C-terminal fragment (9 kDa) can be processed further by γ-secretase, leading to the secretion of a 3 kDa protein (P3) and formation of a 6 kDa C-terminal fragment retained by the cells. β-Secretase cleaves APP at the N terminus of Aβ, leading to the secretion of an N-terminal fragment (APPβ) that is 16 residues shorter than APPα and a cell-associated, C-terminal fragment (C99) that can be processed further by γ-secretase to generate the C terminus of the Aβ protein.
To develop an experimental system in which to measure pharmacological effects on APP processing in vivo, we have used a gene-targeted mouse harboring the Swedish familial Alzheimer’s disease (FAD) mutation and a humanized Aβ domain (Reaume et al., 1996). These mice express readily detectable levels of human Aβ and, unlike conventional APP transgenic animals, express endogenous levels of APP under normal developmental and tissue-specific control. In this model, we previously demonstrated increased cleavage at the β-secretase site, a finding well documented in vitro in FAD model systems (Citron et al., 1992; Cai et al., 1993; Citron et al., 1994).
Knowledge of Aβ turnover rate in vivo is necessary to determine treatment intervals over which a change in Aβ levels could be measured. Here, we demonstrate that mechanisms are present in mouse brain that eliminate Aβ, C99, and APP within several hours of their generation. This knowledge was exploited to initiate studies of agents that could lower Aβ levels in the brain. Phorbol 12-myristate 13-acetate (PMA) has been used extensively in vitro to modulate APP processing and, as a result, lower secreted Aβ levels (Buxbaum et al., 1992; Hung et al., 1993). This compound is highly selective for the activation of protein kinase C (PKC) (Newton, 1995) and has been used previously to modulate brain processes in vivo (Cope et al., 1984; Routtenberg et al., 1986; Baranyi et al., 1987; Deitrich et al., 1989). Here, intracortical injections of PMA resulted in short-term reductions in Aβ levels in mouse brain, thus validating in vivo the modulatory effect of PKC activation on Aβ secretion that has been described in cultured cell systems. We have extended these observations by demonstrating specific effects on levels of both Aβ40 and Aβ42. Interestingly, we did not observe increased APPα concentrations in brain coincident with the reduction in levels of Aβ, contrary to effects of PMA seen in cell culture. This highlights the importance of measuring Aβ levels directly, as well as other APP derivatives, when evaluating modulators of APP metabolism in vivo.
MATERIALS AND METHODS
Antibodies. Rabbit polyclonal antibody 1153 was generated against the first 28 amino acids of human Aβ (Savage et al., 1994). Monoclonal antibody 6E10 and biotinylated 6E10 were both purchased from Senetek (Maryland Heights, MO) and recognize an epitope within Aβ1–17. Rabbit polyclonal antibody 97 was directed against the 30 amino acids at the C terminus of APP (Reaume et al., 1996). Aβ40- and Aβ42-selective polyclonal antibodies (affinity-purified; Aβ40, lots 4434804, 4434805, 4434806, and 4434807; Aβ42 lot 4434417) were obtained from Quality Controlled Biochemicals (Hopkinton, MA). Polyclonal antibody 54 was generated against the peptide sequence SEVNL and is specific for Swedish FAD mutant APPβ (Siman et al., 1995). A monoclonal antibody against actin (clone C4) was purchased from Boehringer Mannheim (Indianapolis, IN). Both goat anti-rabbit and goat anti-mouse IgG1 were purchased from Southern Biotechnology Associates (Birmingham, AL).
Infusion of [35S]methionine into gene-targeted mice. Mice at 6 months of age were anesthetized with a mixture of 120 mg/kg ketamine and 12 mg/kg xylazine. [35S]Methionine (New England Nuclear, Boston, MA) at 500 μCi/100 μl (with 0.9% saline diluent) was infused into the femoral vein of each mouse over 30 min. Rate and volume of infusion were chosen to approach steady-state plasma levels of isotope (Garlick and Marshall, 1972). At various time points after the midpoint of the infusion, mice were anesthetized with avertin (1.25% 2,2,2-tribromoethanol and 2.5% 2-methyl-2-butanol), and blood was withdrawn via intraventricular puncture. The mice were then perfused with 15 ml of Ringer’s solution at room temperature over 5 min, and brains were removed, rapidly frozen, and stored at −70°C.
Detection of radiolabeled and steady-state Aβ, C99, and APP. Brain supernatants were processed for immunoprecipitation as described previously (Reaume et al., 1996). Briefly, brains were homogenized in 6 m guanidine and 50 mm Tris, pH 7.5, and centrifuged at 100,000 × g, and supernatants were dialyzed against PBS with protease inhibitors. From three-fourths of the dialysate, C99 and Aβ were immunoprecipitated using the polyclonal antiserum 1153 and Pansorbin (Calbiochem, San Diego, CA). From the remaining one-fourth of the dialysate, APP was immunoprecipitated using polyclonal antiserum 97. Proteins immunoprecipitated with antiserum 1153 were resolved by electrophoresis on 10–20% Tris-tricine polyacrylamide gels (Owl Scientific, Woburn, MA) and transferred to a polyvinylidene difluoride membrane. Proteins immunoprecipitated using antiserum 97 were resolved using Laemmli gels (4–20%, Owl Scientific) and transferred to nitrocellulose. Dried membranes were exposed to phosphorimage screens (Molecular Dynamics, Sunnyvale, CA). Relative intensities of protein bands were determined using ImageQuant software (Molecular Dynamics).
After exposure of the phosphorimage screen to detect radiolabeled proteins corresponding to Aβ, C99, and APP, steady-state levels of these proteins were detected by immunoblotting. Membranes were wetted with transfer buffer and blocked with 5% nonfat dry milk in Tris-buffered saline (TBS). Monoclonal antibody 6E10 and enhanced chemiluminescence were used to detect these APP forms as described previously (Reaume et al., 1996).
Treatment of mice with phorbol ester. Gene-targeted mice from 3–6 months of age were anesthetized with ether. PMA (40 nmol) or 4αPMA (40 nmol) (phorbol esters from Alexis Biochemicals, San Diego, CA) or a corresponding volume of vehicle (2.5 μl, 30% DMSO and 0.9% saline) was injected unilaterally into the parietal cortex, 2.5 mm down from the surface of the head. Six or 12 hr later, animals were anesthetized with avertin, brains were removed, and parietal cortex samples from both hemispheres were Dounce-homogenized together in 2 ml of 0.2% diethylamine (DEA) and 50 mm NaCl. Brain homogenates were centrifuged at 100,000 × g, and recovered supernatants were neutralized to pH 8.0 with 2 mTris-HCl. Extracts were diluted 1:1 with 5% fetal clonal serum (HyClone, Logan, UT) and 1% nonfat dry milk in TBS and analyzed for Aβ concentration using the ELISAs described below.
Aβ40- and 42-specific ELISAs. For the 40-specific ELISA, Fluoronunc plates (Nunc, Naperville, IL) were coated with goat anti-rabbit IgG at 1:300 in 0.1 m sodium bicarbonate. This was followed by an Aβ40-selective polyclonal antibody at 1:300 in 5% fetal clonal serum and 1% nonfat dry milk in TBS. The wells were blocked further in this same solution, without antibody. Tween 20 (0.1%) in TBS was used as the wash solution between each indicated step. Brain DEA extracts containing Aβ were diluted 1:1 in blocking solution and applied to the plates overnight at 4°C. Aβ was detected using biotinylated 6E10 at 1:5000 and avidin–alkaline phosphatase (Cappel/ICN, Costa Mesa, CA) at 1:500. Bound phosphatase was detected using 4-methyl umbelliferyl phosphate (4-MeUP; Sigma, St. Louis, MO) and read at 360/460 nm. Specificity of this ELISA for Aβ1–40 was tested using recombinant C100 (comprising the last 100 amino acid residues of APP; Savage et al., 1994), Aβ1–42, or Aβ1–43 (Bachem, King of Prussia, PA). These proteins were added to standard curves comprising Aβ1–40 (Bachem), and their ability to alter the curve was determined. ELISA signals are reported as femtomoles of Aβ per milligram of total extracted protein based on Aβ standard curves generated in each experiment.
The design of the 42-selective ELISA was the same, except the capture antibody was 6E10 and the detecting antibody was selective for Aβ42. Fluoronunc plates were coated with goat anti-mouse IgG1 as described above. Monoclonal antibody 6E10 was used at 1:1000 in 5% fetal clonal serum and 1% nonfat dry milk, followed by an additional block in 5% nonfat dry milk in TBS. Intermediate washes were as above. Brain DEA extracts were diluted in blocking solution and applied for overnight capture at 4°C. Aβ42 was detected using the 42-specific polyclonal antibody at 1:200, followed by a goat anti-rabbit IgG–alkaline phosphatase conjugate (Southern Biotechnology Associates, Birmingham, AL) at 1:5000. 4-MeUP substrate was used as above. Specificity of this ELISA for Aβ was tested by comparing signals generated with Aβ1–40, Aβ1–43, and recombinant C100 with signals generated using Aβ1–42.
Because C99 and APP could also bind to the 6E10 capture antibody used in the 42-specific ELISA and are present at much higher concentrations than Aβ42 in the mouse brain, we examined DEA extracts for the presence of C99 and APP. Neutralized DEA extracts were adjusted to 1× radioimmunoprecipitation assay buffer (50 mm Tris base, pH 8.0, 150 mm NaCl, 1% Triton X-100, 0.25% sodium deoxycholate, 0.1% sodium dodecyl sulfate, 2 mm EDTA, and 1 mm benzamidine) and immunoprecipitated with Ab 97 and Pansorbin overnight at 4°C. Some samples were spiked with 5 ng of recombinant C100. Precipitated proteins were electrophoresed on 10–20% Tris-tricine gels or 4–20% Laemmli gels (both from Owl Scientific) to visualize C99 or APP, respectively. Proteins were immunoblotted and detected using 6E10 and ECL as above.
Detection of APPα and APPβ. From the DEA extracts described above, 250 μg aliquots of protein from each sample were methanol-precipitated for 1–2 hr at −20°C. Precipitates were air-dried and resuspended in equal volumes of Laemmli sample buffer. Aliquots of each sample (50 μg) were electrophoresed in triplicate using 4–20% Laemmli gels and transferred to nitrocellulose. Membranes were blocked using 5% nonfat milk powder in TBS and incubated with either 6E10 at 1:2000 or 54 at 1:500 for detection of APPα or APPβ, respectively. Relative band intensity (visualized using ECL as described above) was determined using a Docugel V Scanalytic system (CSP, Inc., Billerica, MA). As a control for protein loading and transfer variabilities between samples, immunoblots were stripped using 100 mm β-mercaptoethanol and 2% SDS at 37°C for 1 hr and reblotted using a monoclonal anti-actin antibody at 1:2000. Band densities representing secreted APP fragments were then normalized to the actin signal densities contained within each respective sample.
Radiolabeling in vivo
We have studied the metabolism of C99, Aβ, and APP in vivo using infusion of [35S]methionine into the femoral vein of the gene-targeted mouse. The turnover of all examined species occurs within 1 d. Identity of these three proteins (Figs. 1,2) is confirmed by molecular weight and reactivity with two Aβ/APP-specific antibodies (1153 and 6E10), one of which is specific for human APP and derivatives (6E10). Radiolabeled proteins immunoprecipitated with 1153 migrated between 3.5 and 6 kDa (Fig. 1 A), at 14 kDa (Fig.1 A), or between 97 and 200 kDa (Fig.2 A) and aligned precisely with proteins immunodetected with 6E10 (Figs. 1 D,E,2 D). These proteins are, therefore, identified as Aβ, C99, and APP, respectively. All three immunoblotted proteins also co-electrophoresed with APP-related standards (APP, Fig.2 E; Aβ and C99, data not shown).
C99 achieved its maximal observed specific activity between 1 and 2 hr after infusion (Fig. 1 A). Thet ½ estimate based on decay of C99 signal over time was ∼3 hr (Fig. 1 B). Peak specific activity of the Aβ peptide was achieved by 4–7 hr after infusion with the signal returning to baseline levels by 16 hr (Fig.1 A,C). Because there were no data points on the downward slope of the Aβ decay curve, an approximatet ½ for Aβ is best estimated from its rate of formation, because, at steady-state, the rate of protein synthesis equals the rate of turnover. The specific activity of the Aβ protein increased from background at 2 hr after infusion to peak incorporation between 4 and 7 hr (i.e., 2–5 hr later). Therefore, we estimate the t ½ to be between 1.0 and 2.5 hr. The coincident decay of C99 and formation of Aβ in vivo supports the precursor–product relationship of these two APP derivatives, respectively (Golde et al., 1992; Cai et al., 1993; Perez et al., 1996). To determine absolute levels of these fragments during the course of the experiment, and also to confirm their identities, immunoblotting was performed using 6E10 (Fig. 1 D,E). Absolute levels of Aβ and C99 were relatively constant at each time point, whereas their specific activities were changing. This confirms that the turnover measurements were performed under steady-state conditions. A radiolabeled protein that migrated slightly slower than C99 (Fig. 1 A) was not labeled with 6E10 (Fig.1 E). This protein, therefore, was not derived from APP and was nonspecifically precipitated.
Full-length APP was immunoprecipitated using the Cterminal-specific Ab 97. Two major proteins were evident on both the phosphorimage (Fig. 2 A) and immunoblot with 6E10 (Fig. 2 D). On both images, the lowerM r band was sharply focused, and the upper band was more diffuse, as is typically seen with mature, fully glycosylated APPs (e.g., Oltersdorf et al., 1990). Most of the APP made in the brains of our gene-targeted mice (Fig. 2 E) is APP695, as expected for rodent brain (Rockenstein et al., 1995). The major proteins immunoprecipitated with Ab 97 and immunodetected with 6E10 electrophoresed precisely with both of the major APP forms extracted from the brain of a transgenic rat overexpressing human APP695. In addition, the lowest M r band detected in the gene-targeted brains co-electrophoresed with the lowest band isolated from human cortex, which is immature APP695. APP-like proteins (APLPs) are also present in mouse brain (Wasco et al., 1992) and could contribute to the phosphorimage signals. This is unlikely, because proteins seen on the Western blot using 6E10 (which does not recognize APLPs) precisely co-electrophorese with the radiolabeled proteins immunoprecipitated with Ab97.
Immature APP695 attained its peak specific activity by 1 hr (Fig.2 B), and the t ½ of this material appeared to be ∼3 hr. The specific activity of mature APP695 peaked between 1 and 2 hr and fell to background by 16 hr (Fig.2 C). Therefore, the t ½was ∼7 hr.
ELISAs detect Aβ40 and Aβ42 in brains of gene-targeted mice
Using sandwich ELISAs selective for Aβ peptides with C termini ending at residue 40 or 42, Aβ was detected in brain extracts. The 40-selective assay (Fig. 3 A) was linear to 240 fmol/ml and sensitive to 12 fmol/ml. This assay was >1000-, 5000-, or 10,000-fold more selective for Aβ1–40 compared with Aβ1–43, C100, or Aβ1–42 standards, respectively. Brain Aβ40 levels measured in the homozygous, gene-targeted mouse brain with this ELISA are 120 fmol/mg protein.
The 42-selective assay (Fig. 3 B) was linear to 480 fmol/ml and sensitive to 7.4 fmol/ml. This assay was 400-, 10,000-, or 16,000-fold more selective for Aβ1–42 compared with Aβ1–43, Aβ1–40, or C100 standards, respectively. Selectivity in the Aβ42 ELISA resided in the detecting antibody, unlike the 40-specific ELISA. Because the DEA extracts contain little membrane-associated C99 (Fig.3 D) or APP (data not shown), the nonselective 6E10 capture reagent was not saturated by C99 or full-length APP and was therefore free to capture DEA-extracted Aβ species. Whole-brain Aβ42 levels measured using this ELISA are 16 fmol/mg protein. Both ELISAs also responded appropriately to twofold differences in brain Aβ driven by gene dosage. Extracts from homozygous mice revealed a twofold higher signal in the ELISA compared with heterozygous mice with only one copy of the targeted allele (Fig. 3 C), confirming a previous observation made using Western blotting (Reaume et al., 1996).
Phorbol ester reduces cortical Aβ and APPβ levels acutely
Because the Aβ protein is cleared within a few hours of its synthesis in the mouse brain, we examined the acute effect of the phorbol ester PMA on the level of Aβ and other APP derivatives. Highly significant 30–35% reductions in levels of Aβ40 (from 121.4 to 84.1 fmol/mg) and Aβ42 (from 20 to 12.9 fmol/mg) were seen in parietal cortex 6 hr after intracortical injection of 40 nmol of PMA (Fig. 4; p < 0.0001). By 12 hr, the effect on Aβ40 was lost (Fig.5 A). Aβ42 levels were not examined at 12 hr. Intracortical injection of 40 nmol of 4αPMA, an analog of PMA unable to activate PKC (Castagna et al., 1982; Kikkawa et al., 1983; Nichols et al., 1987), failed to reduce Aβ40 levels at 40 nmol 6 hr after injection (Fig. 5 B). The hippocampus was also examined and was inconsistently affected by PMA injection (data not shown). It is possible that this highly lipophilic compound did not travel a significant distance from the injection site.
APPα and APPβ levels were also examined 6 hr after PMA injection from the same extracts used to measure Aβ. In contrast to expectations from cell culture experiments, APPα levels were unchanged by this compound (Fig.6 A). APPβ levels, on the other hand, were significantly reduced by 32% in the PMA-treated group (p < 0.02) (Fig. 6 B). The degree of difference in levels of the two forms of secreted APP compared with controls was essentially identical whether the values were normalized to the actin band subsequently visualized within each blotted sample.
We examined the synthesis and turnover of human Aβ present within the gene-targeted mouse brain and used this information to design studies aimed at the reduction of brain Aβ. Under the control of the mouse APP promoter, this model provided appropriate cellular production of human Aβ and allowed the study of Aβ and APP metabolism under physiologically relevant APP concentrations. Aβ, C99, and APP undergo virtually complete clearance within 1 d in this model. The estimated t ½ for Aβ is between 1.0 and 2.5 hr and was estimated from the rate of synthesis, because, at steady-state levels of protein, the rate of synthesis equals the rate of turnover. In addition to metabolic turnover of Aβ, rapid transport of this protein out of the brain into either blood or CSF could also contribute to its clearance. Although Aβ transport into brain across the blood–brain barrier has been reported (Zlokovic et al., 1993), transport of Aβ out of brain parenchyma has not yet been demonstrated.
Western blot data show steady-state levels of all APP-related proteins over the course of the experiment. Thet ½ values predicted here for immature APP (3 hr), fully glycosylated APP (7 hr), and C99 (3 hr) were based on turnover of those molecules that acquired [S35]methionine during the original pulse and via reuse of label released from rapidly metabolized proteins. Rates of synthesis for C99 and APP could not be used to obtain more precise estimates of t ½, because the specific activity of both forms of APP and C99 had already peaked at the earliest time examined. A more precise definition of the half-lives of these proteins based on the rate of synthesis in vivo would require a shorter pulse duration and the measurement of points between time 0 and 1 hr.
We have shown that peripheral infusion of [35S]methionine over 30 min is sufficient to visualize Aβ and its precursors in mouse brain, despite the presence of only one methionine per Aβ molecule, and a background of physiological levels of unlabeled methionine. In studies that examined turnover of total protein in brain using TCA precipitation, there was concern that a stable pool of radiolabeled precursor be maintained to act as a saturating pulse (Garlick and Marshall, 1972; Dunlop et al., 1975; Reith et al., 1978). In these experiments, very long infusion periods or large bolus injections of radiolabel were delivered to saturate the precursor pools. Continuous infusion of [35S]methionine and other radiolabeled amino acids directly into small subregions of the brain has also been used to study the synthesis and transport of substance P (Sperk and Singer, 1982;Torrens et al., 1982; Krause et al., 1984) and other neuropeptides (Kochman et al., 1982; Krause et al., 1982). This report demonstrates whole-brain synthesis and degradation of specific proteins using peripheral infusion of radiolabeled amino acid over a relatively short infusion interval.
The synthesis and turnover of APP and APP fragments have been studied extensively in vitro (Weidemann et al., 1989; Oltersdorf et al., 1990; Haass et al., 1991; Golde et al., 1992; Busciglio et al., 1993; Siman et al., 1993; Perez et al., 1996). Pulse–chase experiments using [35S]methionine showed processing of APP to fully glycosylated, proteolytically cleaved forms within 1 hr in non-neuronal cells (Oltersdorf et al., 1990; Haass et al., 1991; Golde et al., 1992;Busciglio et al., 1993; Siman et al., 1993). In contrast, the half-life of immature APP was 3 hr in the neuronal NT2N cell line (Wertkin et al., 1993), identical to the half-life we estimate in vivo. The gene-targeted mice have the Swedish FAD mutation within the context of mouse APP, unlike the wild-type, human APP present in NT2N cells. When turnover rates of wild-type versus Swedish mutant APP have been compared in vitro, they were found to be identical (Perez et al., 1996). Therefore, the turnover rate estimate of the APP present in the gene-targeted mice likely reflects the in vivo turnover of wild-type APP as well.
The turnover of C-terminal derivatives measured in various cell culture systems has ranged from 1–2 hr to >8 hr (Oltersdorf et al., 1990;Haass et al., 1991; Estus et al., 1992; Busciglio et al., 1993; Siman et al., 1993; Martin et al., 1995). Our estimate of thet ½ of C99 in vivo from expression of the Swedish mutant transgene falls within that range. The cellular compartment in which β-secretase activity generates the N terminus of Aβ differs for APPs containing the Swedish FAD mutation compared with wild-type forms (Haass et al., 1995; Thinakaran et al., 1996). A 13.5 kDa fragment (analogous to C99 here) appeared more rapidly in mouse N2a cells transfected with Swedish mutant APP than wild-type APP (Thinakaran et al., 1996). Therefore, the rate of formation of C99 may also be more rapid in our gene-targeted mouse system expressing Swedish mutant APP compared with wild-type APP.
Aβ turnover in vitro has been studied recently (Naidu et al., 1995; Qiu et al., 1996, 1997). The peak specific activity of radiolabeled Aβ secreted from Chinese hamster ovary (CHO) cells transfected to overexpress APP 695 occurred at 6 hr (Naidu et al., 1995), similar to the result in our gene-targeted system. Estimates of Aβ t ½ values are also comparable between 1.5 and 3 hr. Similar rates of Aβ secretion have been found using CHO cells expressing either wild-type or Swedish mutant forms of APP, whereas the amount of Aβ secreted was elevated in the presence of the mutation (Perez et al., 1996). We predict that turnover rates of Aβ determined in this study will estimate the rates for Aβ generated from wild-type APP if the compartments in which Aβ is metabolized are the same, and Aβ turnover mechanisms are not saturated.
Specific proteases that metabolize Aβ are beginning to be identified. A serine protease–α2-macroglobulin complex was identified in preparations of pancreatic trypsin or fetal bovine serum that degraded exogenous Aβ, as well as Aβ secreted by cells (Qiu et al., 1996). A nonmatrix metalloprotease and a serine protease secreted by both CHO cells and BV-2 microglial cells have also been reported to catabolize Aβ (Qiu et al., 1997). Microglial cells in brain may, therefore, secrete a protease important for Aβ turnover in vivo. Turnover of Aβ could also occur in an intracellular compartment, because Aβ42 uptake from extracellular medium has been described (Knauer et al., 1992); also, the de novo synthesis of intracellular Aβ has been reported in primary neurons (Tienari et al., 1997), NT2N cells (Wertkin et al., 1993), and 293 or CHO cells transfected with Swedish mutant APP (Martin et al., 1995, Perez et al., 1996). Rates of Aβ synthesis and clearance determine its steady-state level and dysfunction of either process could lead to the elevation of Aβ to concentrations critical for fibril formation.
Using this information concerning the turnover rate of Aβ in mouse brain, we investigated the effects of intracortically injected phorbol esters on brain Aβ metabolism. The development of these gene-targeted mice was critical to our study because Aβ is undetectable in wild-type mouse brain with either our ELISA or immunoblotting methods. A highly significant 30–35% reduction was demonstrated in levels of both Aβ40 and Aβ42 6 hr after administration of PMA. Six hours was chosen as the earliest postdose interval after PMA, allowing 2–6 half-lives for Aβ clearance. This is the first report of the selective effects of PMA on Aβ40 and Aβ42 as measured by ELISA. For the elevation of PKC activity to be considered a therapeutically relevent approach for the treatment of AD, it was necessary to demonstrate reduction in levels of brain Aβ42 with a compound known to impact this pathway. The selective effect of phorbol ester stimulation on Aβ42 was examined previously using an in vitro system, whereas Aβ40 was measured as part of the “total” Aβ that remained (Citron et al., 1996). In our study, cortical levels of Aβ40 returned to baseline by 12 hr, possibly reflecting the half-life of PMA in mouse brain, which was determined to be 9.6 hr (Dietrich et al., 1989). We expect the Aβ42 followed a similar time course of recovery, although this fragment was not measured at 12 hr.
The activity of PMA on APP processing in mouse brain is presumably through its selective PKC-stimulatory function (Newton, 1995;Nishizuka, 1995), because an analog that is unable to activate PKC, 4α-PMA, had no effect on Aβ levels. We were unable to measure changes in PKC translocation after PMA delivery in mouse brain, because the DEA extraction buffer led to the release of membrane-associated PKC (data not shown). Therefore, it was not possible to measure concomitant effects of PMA on Aβ levels and PKC activity.
The role of PKC in modulating APP processing was determined by the discovery that stimulation of PKC either directly (Caporaso et al., 1992) or by activation of muscarinic receptor subtypes linked to this second messenger pathway (Buxbaum et al., 1992; Nitsch et al., 1992;Farber et al., 1995), led to an increase in the secretion of APPα and concomitant reduction in both Aβ (Buxbaum et al., 1993; Hung et al., 1993) and APPβ (Felsenstein et al., 1994; Jacobsen et al., 1994). These results suggested either increased processing of APP via an α-secretase-mediated pathway (by increased secretase activity or a rerouting of full-length APP to compartments involved in α-secretase processing) or decreased activity via a β-secretase-mediated pathway. We chose to study the effect of a phorbol ester in vivo, because PKC itself is ubiquitously expressed, whereas select cell surface receptors linked to PKC have a more focal distribution.
In contrast to Aβ, APPα levels were unchanged 6 hr after PMA delivery, whereas APPβ levels decreased significantly. This correlation of Aβ reduction with APPβ reduction suggests that there is reduced cleavage of APP by β-secretase in the presence of phorbol esters, as suggested previously (Buxbaum et al., 1993; Hung et al., 1993; Felsenstein et al., 1994; Jacobsen et al., 1994). The lack of significant elevation in levels of APPα after phorbol ester treatmentin vivo could be attributed to more efficient turnover of this protein compared with in vitro systems or, alternatively, to a dissociation of systems regulating Aβ/APPβ secretion and APPα secretion in the mouse brain. These processes have been reported to be dissociated in certain in vitro systems (Gabuzda et al., 1993; Dyrks et al., 1994; Fuller et al., 1995). Recently, however, increased APPα secretion was reported in a rat model in which PKC activity was constitutively upregulated after treatment with menthylazoxymethanol in utero (Caputi et al., 1997), supporting the influence of this pathway on APP processingin vivo. Together, these studies emphasize the importance of measuring all APP fragments of interest when testing pharmacological modulators of APP processing in brain. In drug discovery efforts, for example, the degree to which a compound is effective in lowering Aβ levels should not be based exclusively on the measurement of a surrogate marker, such as APPα. In addition, test systems must effectively model both synthesis and clearance of brain Aβ and other APP-processing fragments; cultured cells may only partially represent these processes.
Collectively, our data show that the gene-targeted mouse is a useful model for the study of agents that modulate Aβ levels in brain. Turnover of brain Aβ protein under physiological conditions occurs within several hours. This makes possible the design of further studies to investigate agents that modulate Aβ levels in vivowhether by modulation of PKC activity or by alternative pathways independent of PKC.
We thank Drs. Matthew Miller, Bruce Jones, and Dorothy Flood of Cephalon and Dr. Jim Krause of Washington University for discussions that helped initiate this work, Thomas Emmons for help with preparation of figures, Renee Simmons and Edwin McCabe for their excellent care of the gene-targeted animals, and Drs. Frank Baldino and Jeffry Vaught for their continuing support of this work.
Correspondence should be addressed to Dr. Mary Savage, Cephalon, Inc., 145 Brandywine Parkway, West Chester, PA 19380.