|Home | About | Journals | Submit | Contact Us | Français|
Five point mutations within the amyloid β-protein (Aβ) sequence of the APP gene are associated with hereditary diseases which are similar or identical to Alzheimer’s disease and encode: the A21G (Flemish), E22G (Arctic), E22K (Italian), E22Q (Dutch) and the D23N (Iowa) amino acid substitutions. Although a substantial body of data exists on the effects of these mutations on Aβ production, whether or not intra-Aβ mutations alter degradation and how this relates to their aggregation state remain unclear. Here we report that the E22G, E22Q and the D23N substitutions significantly increase fibril nucleation and extension, whereas the E22K substitution exhibits only an increased rate of extension and the A21G substitution actually causes a decrease in the extension rate.
These substantial differences in aggregation together with our observation that aggregated wild type Aβ(1–40) was much less well degraded than monomeric wild type Aβ(1–40), prompted us to assess whether or not disease-associated intra-Aβ mutations alter proteolysis independent of their effects on aggregation. Neprilysin (NEP), insulin degrading enzyme (IDE) and plasmin play a major role in Aβ catabolism, therefore we compared the ability of these enzymes to degrade wild type and mutant monomeric Aβ peptides. Experiments investigating proteolysis revealed that all monomeric peptides are degraded similarly by IDE and plasmin, but that the Flemish peptide was degraded significantly more slowly by NEP than wild type Aβ or any of the other mutant peptides. This finding suggests that resistance to NEP-mediated proteolysis may underlie the pathogenicity associated with the A21G mutation.
Convergent evidence suggests that accumulation of the amyloid β-protein (Aβ) and its subsequent aggregation initiates a complex cascade that culminates in Alzheimer’s disease (AD) (LaFerla and Oddo, 2005; Walsh and Selkoe, 2004). Five point mutations within the Aβ sequence that are associated with hereditary diseases similar or identical to AD are clustered around the central hydrophobic core of Aβ and include: the A21G Flemish mutation, E22K Italian mutation, E22G Arctic mutation, E22Q Dutch mutation and the D23N Iowa mutation (Fig. 1). These mutations have the potential to impact upon all factors known to regulate Aβ monomer levels, namely production, degradation and aggregation. To date, no simple correlation between Aβ production or aggregation and disease phenotype has emerged, but it, was suggested that the pathogenic effect of the Dutch, Flemish, Italian and Arctic mutations arise as a consequence of their resistance to proteolysis (Tsubuki et al., 2003).
A large number of proteases have been implicated in the catabolism of Aβ, but of these neprilysin (NEP, EC 220.127.116.11) and insulin-degrading enzyme (IDE, EC 18.104.22.168) are the most studied (Eckman and Eckman, 2005; Selkoe, 2001; Turner et al., 2004). NEP is a membrane-bound zinc-metallopeptidase that exists as an ectoenzyme preferentially hydrolysing extracellular oligopeptides on the amino side of hydrophobic residues (Carson and Turner, 2002). NEP has been shown by numerous investigators to be capable of degrading Aβ both in vivo (Iwata et al., 2004) and in vitro (Howell et al., 1995; Kanemitsu et al., 2003; Liu et al., 2007; Shirotani et al., 2001), and its physiologic role has been demonstrated by the finding that genetic ablation of NEP causes elevation of endogenous Aβ (Iwata et al., 2001), whereas transgenic or viral expression of NEP causes a lowering of cerebral Aβ (El-Amouri et al., 2007; Hemming et al., 2007; Iwata et al., 2004; Leissring et al., 2003; Marr et al., 2003). IDE is a zinc-metalloprotease which shows no obvious primary amino acid sequence specificity but has been proposed to recognize a conformation that is prone to conversion to β-sheet structure (Kurochkin, 1998). As with NEP, support for the physiological importance of IDE in regulating Aβ levels comes from the findings that genetic deletion of IDE in mice (Farris et al., 2003; Miller et al., 2003) leads to elevated levels of cerebral Aβ, whereas transgenic over-expression of IDE causes a decrease in brain Aβ levels (Leissring et al., 2003). Unlike NEP and IDE, plasmin (EC 22.214.171.124) does not appear to contribute significantly to the normal catabolism of Aβ (Tucker et al., 2004), but rather seems to play a role in the diseased brain. Plasmin is a serine protease that is produced from its inactive precursor, plasminogen, by the action of tissue-type plasminogen activator (tPA), and mice lacking either tPA or plasminogen clear injected Aβ much less well than wild type mice (Melchor et al., 2003). Moreover, recent studies indicate that high levels of Aβ cause a decrease in tPA activity suggesting that excessive accumulation of Aβ results, in part, due to a loss of plasmin (Cacquevel et al., 2007).
In order to investigate the possibility that intra-Aβ mutations mediate their effect through a common pathogenic mechanism, namely resistance to proteolysis, we examined the ability of NEP, IDE and plasmin to degrade both wild type (wt) Aβ(1–40) and Aβ(1–40) bearing disease-associated single amino acid substitutions (Fig. 1). Here we show that the ability of these proteases to degrade Aβ is strongly retarded by aggregation but that when Aβ peptides are presented in their unaggregated, monomeric state, they are efficiently degraded by NEP, IDE and plasmin, with one exception. The Aβ(1–40)A21G peptide is degraded more slowly by NEP than wt Aβ(1–40) or any of the other mutant peptides studied. This resistance to NEP-mediated proteolysis may represent one mechanism by which the A21G mutation causes increased cerebral accumulation of Aβ. On the other hand, our data suggest that resistance to proteolysis by the major known Aβ degrading enzymes cannot explain the effects of the other four disease-associated intra-Aβ mutations.
Unless otherwise stated all chemicals were purchased from Sigma-Aldrich, St. Louis, MO, USA and were of the highest purity available.
The six peptide sequences shown (Fig. 1) were synthesized and purified at the Bioploymer Laboratory in UCLA. Peptide mass and quantity were determined by a combination of matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Table 1), and quantitative amino acid analysis (Walsh et al., 1997). In all cases peptide purity as determined by analytical HPLC was >95%.
Aggregation was monitored using two distinct assays: (1) sedimentation of aggregates at 100,000g and visualization of the remaining soluble peptides by silver staining of samples separated on sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) and (2) incubation of samples with Thioflavin T to allow for continuous assessment of dye binding. For the sedimentation assay, peptides were dissolved at a concentration of 60 µM in 0.1% NH4OH, diluted with an equal volume of 100 mM Tris-HCl, pH 7.4, and pre-existing aggregates removed by ultracentrifugation at 100,000g and 4°C for 1h. Peptide concentrations were determined using the Micro BCA protein assay (Pierce, Rockford, IL), adjusted to 10 µM, and 60 µl aliquots were incubated at 37°C. At appropriate intervals, peptide solutions were ultracentrifuged at 100,000g and 4°C for 1h and 30 µl of supernatant removed, mixed with 2× sample buffer, boiled for 5 minutes electrophoresed on 10–20% tricine gels (Invitrogen Life Technologies Ltd, Carlsbad, CA, USA) and the Aβ visualized with silver stain according to the method of Shevchenko et al (Shevchenko et al., 1996). A decrease in Aβ concentration following centrifugation is indicative of sedimentation of Aβ aggregates.
For continuous assessment of aggregation, Aβ was incubated in the presence of Thioflavin T and fibril formation estimated by increased fluorescence. Briefly, 75 µl of a 20 µM monomeric Aβ solution (prepared as described for the sedimentation assay) was added to a well of a 96 well micro-titer plate together with 75 µl 20% DMSO and 148.5 µl 20 mM sodium phosphate, pH 7.4 and the first reading taken immediately after addition of 1.5 µl 2 mM Thioflavin T. Samples were incubated at 25°C and shaken at 650 rpm using a VorTemp 56™ Incubator/Shaker with an orbit of 3 mm (Labnet International, Berkshire, UK). Measurements were made at regular intervals using a Molecular Devices SpectraMax M2 microplate reader (Sunnyvale, CA, USA) with excitation and emission at 435 nm (slit width 5 nm) and 485 nm (slit width 10 nm), respectively.
Negative contrast EM was performed as described previously (Walsh et al., 1997). Briefly, sample was applied to a carbon-coated Formvar grid, fixed with glutaraldehyde, stained with uranyl acetate and then viewed using a JEOL 1200 EX or JEOL 2000 transmission electron microscope. All reagents for EM were supplied by Electron Microscopy Sciences (Hatfield, PA, USA).
To prepare aggregate-free, monomeric Aβ solutions, peptides were dissolved in ice-cold HPLC-grade water, diluted with an equal volume of 100 mM Tris-HCl, pH 7.4, to a concentration of 40 µM and then passed through a Microcon® 10 filtration device (Millipore, Bedford, MA, USA). Peptide concentrations were determined using the Micro BCA protein assay and then adjusted to 20 µM, all peptides when normalized in this way produced peaks on HPLC of similar heights, but the degree of silver staining varied somewhat between the different peptides. De-seeding of Aβ solutions using Microcon 10 filters has been shown to effectively produce homogenous monomeric solutions completely free of large Aβ assemblies (Fezoui et al., 2000). Fibril controls were prepared without de-seeding: Aβ(1–40) was dissolved at 1mg/ml in MilliQ water and incubated at 37°C for 1 week. To generate homogenize suspensions of fibrils the sample was sonicated at 5 watts for 30 sec using a Microson ultrasonic cell disruptor (Misonix, Farmingdale, NY). The presence of amyloid fibrils was confirmed by Congo red binding and electron microscopy.
Irrespective of the assembly form used, peptide solutions (250 µl) were incubated at 37°C with one tenth volume of protease to give a final concentration of 70 nM IDE, 80 µM NEP or 156 nM plasmin. Recombinant human IDE was prepared as described previously (Walsh et al., 2002), recombinant human soluble NEP was a gift from Drs. C. Adessi and F. Grueninger (F. Hoffman-La Roche Ltd, Basel, Switzerland) and plasmin was obtained from Calbiochem (San Diego, CA). Protease inhibitors (2 mM 1,10 Phenanthroline, 1 mM PEFA, 5 mM EDTA, 200 µM leupeptin, 1.5 µM aprotinin, 15 µM peptstatin, final concentrations) were added at the times indicated, and the extent of degradation was assessed by reverse-phase HPLC and/or by densitometric analysis of silver-stained samples separated by SDS-PAGE. Briefly, samples (100 µl) were chromatographed on a Discovery® C18 column (4.6 mm × 25 cm) (Supelco, Bellefonte, PA, USA) using a 0–70% acetonitrile gradient generated by a Waters 600 HPLC Controller (Waters Ltd, Milford, MA, USA), and peptides detected by absorbance at 214 nm using a Waters 996 Photodiode Array Detector. The degree of degradation was determined by the change in the peak area of intact peptide. For SDS-PAGE, samples (10 µl) were mixed with 4× sample buffer, boiled for 5 minutes and electrophoresed on 10–20% tris-tricine gels. Gels were silver-stained, and the degree of degradation estimated by densitometric analysis using the Scion Image for Windows Beta 3 program which is available at http://www.scioncorp.com.
The presence of fibrillar aggregates in the 7-day incubated Aβ solutions was assessed essentially as described previously (Walsh et al., 1999). Sample (25 µl) was mixed with 20 µM Congo red in 20 mM potassium phosphate, pH 7.4, containing 0.15 M sodium chloride (225 µl) and incubated for 30 min at room temperature. The absorbance of the resulting solution was then measured at 480 and 540 nm using a Molecular Devices SpectraMax M2 microplate reader (Sunnyvale, CA, USA). All samples were assessed in triplicate and the amount of Congo red bound was calculated using the formula:
Aβ peptides were analyzed by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS). Samples were prepared using the thin-layer method of Beavis and Chait (Beavis and Chait, 1996) using α-cyano-4-hydroxycinnamic acid as matrix. Aβ solutions were first diluted with 50% acetonitrile containing 0.1% TFA in 1:1 ratio and then further diluted 5 times with matrix solution containing saturated matrix in formic acid, water and isopropanol (1:4:4, v/v/v). An aliquot (1 µL) of peptide-matrix solution was spotted onto a sample plate with a preformed ultra thin layer of matrix. Sample spots were washed briefly with ice-cold aqueous TFA (0.1%) and then loaded into the mass spectrometer. Spectra were collected on a Voyager DE-STR time-of-flight mass spectrometer (Applied Biosystems, Foster City, CA) in linear mode with 500 laser shots. Mass calibration was done using bovine insulin as either internal or external mass calibrant.
For comparison of the onset of aggregation, the Tukey-Kramer Multiple Comparisons Test was used to assess the differences between the initial Thioflavin T fluorescence and the fluorescence at each subsequent time point thereafter (GraphPad Prism™ software, GraphPad Inc., San Diego, CA).
To our knowledge, this is the first report to directly compare the aggregation kinetics of wt Aβ(1–40) and Aβ(1–40) peptides bearing each of the five point mutations associated with familial disorders linked to AD. For this study, we used 2 independent methods to monitor the aggregation process and electron microscopy to confirm the presence of amyloid fibrils. Initially, aggregation was monitored by incubating peptides at 37°C and measuring the loss of peptide from the supernate following 100,000 g centrifugation. We have used this protocol extensively, and it allows for detection of sedimentable fibrillar assemblies by monitoring their removal from solution (Clements et al., 1996; Clements et al., 1993; Walsh et al., 1997). When incubated at 10 µM, neither wt Aβ(1–40) nor Aβ(1–40)A21G showed any sign of fibril formation over a four day period (Fig. 2A). Indeed, even at 20 µM, these peptides showed no measurable aggregation, and it was only when they were incubated at 100 µM that appreciable loss of sedimentable aggregates was detected (not shown); at 100 µM, both peptides took more than 3 days to form sedimentable aggregates. In contrast, the rate of aggregation for the other mutant peptides was greatly accelerated compared to that of the wt peptide (Fig. 2A). At 10 µM, peptides bearing the G22 and Q22 mutations readily aggregated and thus disappeared from the supernate within 12 h, whereas Aβ(1–40)D23N essentially disappeared from solution by 24 h. Aβ(1–40)E22K remained in solution for at least 1 day but was completely aggregated into sedimentable assemblies by 48 h. The experiment shown in Fig. 2A is representative of 3 separate experiments, and in each study the order of peptide aggregation was closely similar. However, due to the stochastic nature of protein aggregation of unagitated samples we did observe some variations both between and within experiments. For instance, the duplicates for the 6 h time points for Aβ(1–40)E22G and Aβ(1–40)E22Q are different, with one duplicate showing a substantial decrease at 6 h and the other showing no decrease. Nonetheless, this method clearly shows that Aβ(1–40)E22G and Aβ(1–40)E22Q form sedimentable aggregates most readily, with D23N next fastest, then Aβ(1–40)E22K and finally wt and Aβ(1–40)A21G the slowest. As expected, all peptides studied formed typical amyloid fibrils which were readily detected by EM in both unspun samples and in pellets produced by centrifugation (not shown).
To confirm the veracity of the results obtained using the sedimentation assay and to obtain a more quantitative comparison of aggregation kinetics we employed a Thioflavin T (ThT) binding assay in which samples could be continuously monitored. This approach is particularly useful as it limits the amount of peptide used and provides multiple data points around the time intervals of most interest. Using traditional static ThT binding assays we have previously observed wide fluctuations in the ThT binding of aggregates (Walsh et al., 2001), probably because of variations in the size of the aggregates. Therefore in this study we were careful to vigorously agitate samples throughout the course of the reaction. Such agitation results in faster aggregate formation, but more importantly, the resultant ThT binding data is more consistent among replicate samples. Using this assay, the Aβ(1–40)E22G showed the shortest lag phase, taking only 15 min before the ThT signal increased significantly above the t=0 control (P<0.001, Fig. 2B). In this sort of experiment, the lag phase corresponds to the time taken to form nuclei that can then elongate by addition of monomer (Naiki and Nakakuki, 1996). The lag phase for Aβ(1–40)D23N was somewhat irregular but showed a significant increase at 30 min and all time points thereafter (P<0.01, Fig. 2B). The Aβ(1–40)E22Q had a longer lag phase, being elevated from 60 min (P<0.01, Fig. 2B). The lag phases for wt and Aβ(1–40)A21G were highly similar, with both peptides producing the first sign of appreciable ThT binding at 70 and 80 min (P<0.05 and P<0.0001, Fig. 2B), respectively. Surprisingly the lag phase of Aβ(1–40)E22K was the longest, with this peptide only producing a significant increase in ThT binding at 90 min. The discrepancies between the initiation of aggregation as measured by the lag time for aggregation to be detected by ThT (Fig. 2B) and differences measured by our sedimentation assay (Fig. 2A) probably reflect the ThT assay being much more sensitive, in that it can detect soluble aggregates that are not readily sedimented (Walsh et al., 1999). Both wt and Aβ(1–40)A21G have a shorter lag until aggregation is detected by ThT binding than does Aβ(1–40)E22K, but the time required for Aβ(1–40)E22K to achieve 50% of its aggregation maximum is much shorter (~70 min) than that of either wt (~110 min) or Aβ(1–40)A21G (~140 min, Table 2). It is reasonable to assume that the intensity of the ThT signal is related to both the size and number of aggregates formed and that the G22, Q22 and N23 mutations each decrease the lag time required for initial aggregation while also forming larger sedimentable aggregates at a much faster rate than wt. The K22 mutation slightly increases the lag time required to initiate aggregation, but this peptide can form larger, more readily sedimentable aggregates considerably faster than wt. The lag time for wt and Aβ(1–40)A21G are highly similar, but the time required to achieve maximal aggregation is longer for Aβ(1–40)A21G than wt (Fig. 2B). These results are consistent with prior reports in which aggregation of Aβ(1–40)A21G was found to be similar or slightly slower to that of wt peptide (Clements et al., 1996; Walsh et al., 2001; Walsh et al., 1997) whereas, Aβ(1–40)E22G, Aβ(1–40)E22Q, Aβ(1–40)E22K and Aβ(1–40)D23N have all been shown to aggregate faster than wt Aβ(1–40) (Clements et al., 1996; Melchor et al., 2000; Nilsberth et al., 2001; Sian et al., 2000; Van Nostrand et al., 2001; Walsh et al., 1997) (Fig. 2A and B).
Some studies have reported that Aβ peptides bearing disease-associated mutations are resistant to degradation by NEP and IDE (Morelli et al., 2003; Tsubuki et al., 2003). But since these mutations also alter the aggregation properties of Aβ (Fig. 2) and it has been suggested that aggregated Aβ is less susceptible to degradation (Tucker et al., 2000), we sought to determine if the reported differences in proteolysis resulted from the changes in primary structure per se or simply reflected differences in the peptides’ propensity to aggregate. To this end, we compared the ability of IDE, NEP and plasmin to degrade both aggregated and monomeric Aβ(1–40). Monomeric Aβ was prepared as described previously (Fezoui et al., 2000) and assessed for aggregates by negative contrast electron microscopy, Congo red binding (Crb) and sedimentation at 100,000 g (see Experimental Procedures). All three tests confirmed that the monomeric preparation was free of detectable aggregates (Supplementary Fig. 1).
Using these well-defined preparations, we examined the ability of IDE, NEP and plasmin to degrade both monomeric and aggregated Aβ. In agreement with prior studies (Howell et al., 1995; Iwata et al., 2000; McDermott and Gibson, 1997; Qiu et al., 1998; Tucker et al., 2000) all three enzymes degraded wt Aβ monomer in a time-dependent manner (Fig. 3, upper panels), whereas wt aggregated Aβ was resistant to degradation over the time course studied (Fig. 3, lower panels). Reduced rates of degradation of oligomeric or aggregate forms of Aβ have been observed for certain proteases. For instance, Morelli et al. have shown that the E22Q peptide is degraded by IDE at a significantly reduced rate compared to wt and they hypothesised that the high degree of beta structure in the E22Q peptide was responsible for the reduced degradation (Morelli et al., 2003). Similarly, Tucker et al., have reported that the degradation kinetics of aggregated Aβ by plasmin were much slower than degradation of monomeric Aβ. Using 40 µM Aβ(1–40) and plasmin concentrations more than four-fold greater than those used in our experiments, Tucker et al. observed only a ~30% degradation of aggregated Aβ in a 1 h period (Tucker et al., 2000).
Using the filtration procedure employed above, we prepared aggregate-free monomer solutions of Aβ peptides bearing each of the 5 known disease-associated mutations and asked the simple question can IDE, NEP and plasmin degrade these peptides? All Aβ peptides studied were susceptible to degradation by IDE, NEP and plasmin (Fig. 4A–F). Indeed, the extent of degradation, whether measured by SDS-PAGE/silver stain or HPLC was similar for all peptides, with one notable exception, the A21G peptide. This peptide consistently showed decreased degradation by NEP compared to either wt peptide or the other mutant peptides (compare the G21 lane in Fig. 4B versus the other lanes). We have already demonstrated that the aggregation state of Aβ can influence its degradation (Fig. 3), the aggregation kinetics of A21G are very similar to that of wt Aβ (Fig. 2B) and are much slower than those of the other mutant peptides so the decreased catabolism of A21G by NEP (Fig. 4B) is not due to increased aggregation. Moreover, the fact that the same A21G showed no resistance to proteolysis by IDE or plasmin supports the conclusion that the A21G did not aggregate during incubation with proteases.
To preclude any anomalies resulting from a single means of measuring degradation, we carried out a separate set of experiments in which we used HPLC to monitor the degradation of the wt and mutant peptides. In addition to monitoring the loss of the intact Aβ substrate, HPLC simultaneously detects the appearance of its proteolytic products (Fig. 5D). As with our SDS-PAGE/silver detection procedure (Fig. 4A–C), HPLC (Fig. 4D–F) revealed that all peptides examined were degraded by each of the proteases used, and it also confirmed that the A21G peptide was less degraded by NEP (Fig. 4E). These results are in conflict with the findings of Tsubuki et al, who reported that the A21G, E22Q, E22K and E22G peptides were completely resistant to degradation by NEP (Tsubuki et al., 2003). This apparent contradiction likely results from differences in the preparation of peptides prior to, or aggregation during, the incubation with NEP. We were very careful to use monomeric peptide preparations completely devoid of any detectable Aβ aggregates (SFig. 1). In addition to characterising the peptides before use, we also employed the sedimentation assay to investigate the aggregation state of peptides after incubation at 37°C for two hours, confirming that no significant aggregation occurred during the incubation with proteases (data not shown). Furthermore, to exclude any differences arising due to chemical impurities, we used at least two different batches of each peptide and observed similar degradation profiles with each batch. We conclude that in their monomeric state all 6 peptides are readily degraded by the 3 most well-studied Aβ degrading proteases. However, in all experiments, whether assessed using the SDS-PAGE/silver stain or HPLC paradigms, the A21G peptide was less efficiently degraded by NEP than wt Aβ(1–40) (Fig. 4). Consequently, a more detailed study comparing the degradation of Aβ(1–40)A21G and wt peptide was undertaken.
The observed resistance of the A21G (Flemish) peptide to degradation by NEP suggests that impaired degradation could underlie the enhanced parenchymal and vascular accumulation of Aβ that characterizes the brains of individuals carrying this mutation. In order to investigate this further, kinetic studies were undertaken. NEP was incubated with wt Aβ(1–40) or Aβ(1–40)A21G for the indicated times, samples were electrophoresed and silver stained, and densitometry was used to estimate the amount of intact peptide remaining at each time point. Under the conditions used, wt Aβ(1–40) peptide had a half-life (t½) of ~20 minutes, whereas the t½ for the Aβ(1–40)A21G peptide was almost 3 times as long, taking ~60 min for 50% of the peptide to be degraded (Fig. 5B). In a separate set of experiments, HPLC was used to measure NEP-mediated degradation of wt Aβ(1–40) and Aβ(1–40)A21G. This analysis produced results qualitatively similar to those obtained using the SDS-PAGE/silver stain assay: the half lives for wt Aβ(1–40) and Aβ(1–40)A21G were ~20 min and ~70 min, respectively (Fig. 5C). HPLC also revealed that the NEP degradation profile of Aβ(1–40)A21G differed from that of the other peptides studied. Specifically, two product peaks, present in the NEP-digested wt Aβ(1–40), as well as the Aβ(1–40)E22G, Aβ(1–40)E22K, Aβ(1–40)E22Q and Aβ(1–40)D23N samples, were not found in the NEP-digested Aβ(1–40)A21G reactions (Fig. 5D and E). This latter observation suggests that substitution of alanine by glycine at position 21 changes the accessibility of certain peptide bonds such that NEP does not cleave Aβ(1–40)A21G at all the positions at which it cleaves wt Aβ(1–40).
To address this question, we analyzed the proteolytic fragments of wt Aβ(1–40) and Aβ(1–40)A21G generated by NEP treatment using MALDI-TOF-MS. This revealed that NEP degradation of wt Aβ(1–40) produced a number of proteolytic fragments, including 4–16, 4–17, 1–16, 1–17, 1–19, 1–30 and 1–33 (Fig. 5F). A similar profile was also obtained for Aβ(1–40)A21G (Fig. 5G), but the 1–19 fragment was not detected in 4 separate experiments. However, as in our HPLC analysis (Fig. 5D and E), the levels of proteolytic fragments were much reduced in the Aβ(1–40)A21G samples. Indeed even in samples that were incubated with NEP such that approximately 50% of the parent peptide was degraded the intensity of the A21G products was substantially lower than that observed for the wt peptide (Fig. 5F and G). Thus this makes it difficult to discern whether 1–19 was not produced by NEP proteolysis of Aβ(1–40)A21G or simply that it was produced at levels too low to detect.
This is the first study to simultaneously compare the relative aggregation rates and the susceptibility to degradation of Aβ peptides containing amino acid substitutions corresponding to all the known intra-Aβ mutations linked to AD-like disorders. Here we report that Aβ(1–40)E22G shows the highest propensity for aggregation (i.e., both rapid nucleation and fibril extension) and that Aβ(1–40)E22Q and Aβ(1–40)D23N show slightly slower, but similar aggregation kinetics. Aggregation of Aβ(1–40)E22K is intermediate between that of Aβ(1–40)E22Q and wt Aβ, with a longer lag phase than wt, but a more rapid extension phase. Aβ(1–40)A21G is similar to wt, but with a less rapid extension phase. Increased aggregation has long been assumed to underlie the diseases linked to the E22G, E22K, E22Q, and D23N mutations, however, the relative rates of in vitro aggregation do not show a simple correlation with either the age of disease onset or the type of pathology observed. For instance, the E22G mutation causes a disease indistinguishable from idiopathic AD with an average age of onset of 57+/− 2.7 years (Nilsberth et al., 2001) whereas, the Iowa disease is characterized by severe congophilic amyloid angiopathy (CAA) and a disease onset of 68.9 +/− 4.7 years (Grabowski et al., 2001). Indeed even within the same kindred the disease phenotype can vary dramatically. For instance the A21G (Flemish) mutation can manifest as presenile dementia or cerebral haemorrhage or a mixture of the two, with mean dementia onset at 46.2 +/− 6.1 years (Brooks et al., 2004; Roks et al., 2000). The diseases associated with the E22K and E22Q mutations are characterized by severe CAA but display a highly heterogeneous age of onset (Wattendorff et al., 1995). Thus, while an increased propensity for aggregation is likely to contribute to the aetiology of these diseases; other factors must also influence the disease outcome. Indeed mouse models expressing human APP bearing the A22G mutation form amyloid plaques much more rapidly than mice expressing comparable levels of wild type APP (Cheng et al., 2004), whereas mice expressing E22Q APP form few parenchymal plaques, but exhibit prominent CAA (Herzig et al., 2004).
Because intra-Aβ mutations have the potential to impact upon Aβ degradation as well as aggregation, we asked if peptides bearing intra-Aβ mutations were degraded differently by 3 proteases known to be important for cerebral Aβ catabolism in vivo. Using wt Aβ, we found that proteolysis was highly dependent upon aggregation state and that IDE, NEP and plasmin could not readily degrade fibrillar Aβ. Thus, to distinguish between altered proteolysis resulting from a change in the primary amino acid sequence per se as opposed to a change in aggregation propensity, we were careful to use peptides devoid of aggregates and which exhibited no detectable secondary structure. Owing to the dynamic nature and difficulty obtaining soluble oligomers we did not investigate if oligomers produced from mutant Aβ peptides are catabolized differently, rather we focused on studying degradation of monomeric Aβ peptides. Employing these monomeric, aggregate-free preparations, we found that peptides bearing disease-associated mutations were in general readily degraded by IDE, NEP and plasmin.
Our findings are in conflict with reports that peptides bearing E22Q, A21G, E22K and E22G mutations are almost totally resistant to proteolysis by NEP (Tsubuki et al., 2003) and that the E22G and E22Q peptides are degraded poorly by IDE (Morelli et al., 2003). The most likely explanation for this divergence in results lies in differences in the aggregation state of the peptides used. In this regard, it is particularly noteworthy that in the case of the study by Morelli et al. the authors reported that their Aβ(1–40)E22Q (and to a lesser extent Aβ(1–40)E22G) contained substantial β-structure (Morelli et al., 2003). Both Aβ fibrils and protofibrils have a high β-sheet content, whereas Aβ monomer is natively unfolded (Walsh et al., 1999). Since we found that IDE degrades fibrillar Aβ much less readily than Aβ monomer (Fig. 3), it seems likely that the reported resistance of Aβ(1–40)E22G and Aβ(1–40)E22Q to degradation by IDE (Morelli et al., 2003) was due to the presence of Aβ fibrils and/or protofibrils and not a simple consequence of the amino substitutions rendering these peptides less good substrates. Similarly, the report that NEP did not readily degrade Aβ(1–40)A21G, Aβ(1–40)E22G, Aβ(1–40)E22K and Aβ(1–40)E22Q may also have resulted from aggregation of the peptides used. Although Tsubuki et al. reported that the peptide solutions had not formed fibrils or protofibrils, the extraordinarily high concentration (40 g/L i.e. ~9.2 mM) of peptide used would suggest otherwise (Tsubuki et al., 2003). Another possible explanation for the divergence between our results and those of Tsubuki et al. (Tsubuki et al., 2003) may be the species and form of the NEP used. In the current study, soluble human NEP was used, whereas Tsubuki et al. used mouse NEP of an undefined nature (Tsubuki et al., 2003). It is conceivable that mouse NEP, although highly homologous with human NEP (Li et al., 1995), may act on mutant Aβ differently, or that if Tsubuki et al. used a membrane-bound form of NEP that the mutant peptides readily associated with the membrane and were less accessible for proteolysis.
Whatever the reason for these discordant findings, in numerous experiments using different synthetic peptide batches and different methods to assess degradation we have consistently found that all disease-associated intra-Aβ mutant peptides are degraded similarly by IDE and plasmin and that all except the A21G (Flemish) peptide are degraded equally well by NEP. Given that Aβ(1–40)A21G was readily degraded by IDE and plasmin and that this peptide showed similar aggregation kinetics to wt Aβ(1–40), the observed resistance to NEP is unlikely to be due to altered aggregation but rather is a direct result of the change in primary sequence. NEP preferentially hydrolyses oligopeptides on the amide side of hydrophobic amino acids (Turner et al., 2001) and is thought to cleave Aβ at multiple sites including at phenylalanine 20 which results in liberation of the 1–19 fragment. Glycine is highly disruptive of secondary structure, and it is possible that the A21G substitution distorts the Aβ monomer rendering it a less good substrate for NEP. This notion is supported by a recent structural study that indicates Aβ21–30 containing the A21G substitution is more resistant to digestion by trypsin than wild type Aβ21–30 or Aβ21–30 containing other disease relevant substitutions (Grant et al., 2007). In turn this resistance to NEP-mediated proteolysis would facilitate a build up of Aβ monomer and may explain why the A21G mutation causes increased intracerebral accumulation of Aβ
We are grateful to Drs. Zuhair Nasrallah (UCD) Francesca Paradisi (UCD) and Kevin O’Connor (UCD) for their assistance using the HPLC system and Ann Molloy (UCD) for use of the micro-balance. We also thank Dr. Mark Findeis (Satori Pharmaceuticals Inc, Boston, MA, USA) for advice on the Thioflavin T binding assay and Drs. Celine Adessi and Fiona Grueninger (F. Hoffman-La Roche Ltd, Basel, Switzerland) for the gift of soluble NEP.
This work was supported by Wellcome Trust grant 067660 (DMW), NIH grant AG027443 (DMW & DJS) and the Alzheimer’s Association (RW). The mass spectrometry study was in part supported by an NIH/NCRR Shared Instrumentation Grant.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.