|Home | About | Journals | Submit | Contact Us | Français|
Aggregation of the amyloid β peptide (Aβ) plays a key role in the molecular etiology of Alzheimer’s disease. Despite the importance of this process, the relationship between the sequence of Aβ and the propensity of the peptide to aggregate has not been fully elucidated. The sequence determinants of aggregation can be revealed by probing the ability of amino acid substitutions (mutations) to increase or decrease aggregation. Numerous mutations that decrease aggregation have been isolated by laboratory-based studies. In contrast, very few mutations that increase aggregation have been reported, and most of these were isolated from rare individuals with early onset Familial Alzheimer’s disease. To augment the limited data set of clinically derived mutations, we developed an artificial genetic screen to isolate novel mutations that increase aggregation propensity. The screen relies on expression in E. coli of a fusion of Aβ to green fluorescent protein (GFP). In this fusion, the ability of the GFP reporter to fold and fluoresce is inversely correlated with the aggregation propensity of the Aβ sequence. Implementation of this screen enabled the isolation of 20 mutant versions of Aβ with amino acid substitutions at 17 positions in the 42-residue sequence of Aβ. Biophysical studies of synthetic peptides corresponding to sequences isolated by the screen confirm the increased aggregation propensity and amyloidogenic behavior of the mutants. The mutations were isolated using an unbiased screen that makes no assumptions about the sequence determinants of aggregation. Nonetheless, all 16 of the most aggregating mutants contain substitutions that reduce charge and/or increase hydrophobicity. These findings provide compelling evidence supporting the hypothesis that sequence hydrophobicity is a major determinant of Aβ aggregation.
Postmortem studies of the brains of Alzheimer’s disease (AD) patients reveal significant quantities of senile plaque. Biochemical analyses of the amyloid fibrils in these plaques indicate that Aβ peptides are the primary components of the fibrils1,2. These Aβ peptides are produced by proteolytic cleavage of the Amyloid Precursor Protein (APP). Because cleavage of APP can occur at several sites, Aβ peptides occur in several different lengths, with the 40-residue Aβ40 and the 42-residue Aβ42 being the most abundant. Although Aβ40 is produced in larger amounts, Aβ42 aggregates more readily, and increased ratios of Aβ42/Aβ40 have been observed in the brains of AD patients 3,4.
The molecular details of Aβ aggregation, and the mechanism through which this aggregation causes AD are not fully understood. Nonetheless, a large number of studies support the “amyloid cascade” hypothesis5, which posits that accumulation of aggregated Aβ initiates a multi-step cascade that ultimately leads to Alzheimer’s disease. Several lines of evidence support this hypothesis: First, genetic studies show that several forms of Familial Alzheimer’s Disease (FAD) are caused by mutations either in APP or in the enzymes that process APP. Both classes of mutations increase the production and/or aggregation of Aβ42, and lead to the early onset of AD6–8. Second, early onset AD is also observed in Down syndrome, wherein trisomy of chromosome 21, which encodes APP, leads to increased production of Aβ429–12. Third, construction of transgenic animals, including nematodes, fruit flies, and mice, has demonstrated that introduction of APP and/or Aβ produces cognitive and behavioral impairment13–15. Finally, studies of enzymes that metabolize Aβ confirm the relationship between Aβ accumulation and AD. For example, decreased expression of insulin degrading enzyme (IDE) or neprilysin – both of which are known to degrade Aβ– leads to increased accumulation of Aβ, and ultimately to AD. In contrast, overexpression of these enzymes reduces Aβ levels and attenuates Aβ-related memory deficit 16–20. Together, these studies provide a compelling case for the role of Aβ aggregation in the pathogenesis of Alzheimer’s disease.
Although Aβ accumulation and aggregation clearly play a role in AD, recent studies indicate that the insoluble fibrils themselves may not be the toxic species. Instead, it now appears that oligomers or intermediates in the aggregation process are the major toxic species in AD. For example, Lesne et al. demonstrated that extracellular accumulation of a 56 kDa soluble oligomer of Aβ42 (presumably a dodecamer) causes memory deficits in transgenic mice21. Similarly, Walsh et al. demonstrated that small oligomers of Aβ inhibit long term potentiation of neurons, resulting in memory deficits, whereas monomers or fibrils of Aβ showed no effect22,23.
To enhance understanding of the molecular etiology of AD, we and others have probed the amino acid sequence determinants of Aβ aggregation 24–29. Previously, our lab developed an artificial genetic system to screen for mutations in the sequence of Aβ42 that prevent aggregation24. By using this system to screen randomly generated libraries of mutations, we demonstrated that replacement of nonpolar residues with polar residues inhibited aggregation and caused dramatic increases in the solubility of Aβ42. More recently, we also showed that at many positions in the Aβ42 sequence, random mutations of nonpolar residues to other nonpolar residues had little or no effect, thereby demonstrating that “generic” hydrophobic residues – rather than particular nonpolar side chains – are sufficient to promote the aggregation of Aβ42.
Complementary studies by both Wetzel’s group and Shirasawa’s group used proline scanning mutagenesis to demonstrate that disruption of the β-sheet regions of Aβ decreases aggregation propensity 27,28. Thus, mutagenesis experiments have shown that both sequence hydrophobicity and β-sheet propensity are key determinants of aggregation. Experimental and bioinformatics approaches by Chiti et al. support these findings – both for Aβ42 and for other amyloidogenic proteins29.
In addition to the laboratory-generated mutations described above, naturally occurring mutants in the human population provide insights into the sequence determinants of Aβ aggregation. Several examples of familial early onset AD are caused by mutations in Aβ that increase its aggregation propensity. For example, the Dutch mutant, Glu22→ Gln, increases Aβ aggregation and leads to early onset AD30.
Laboratory-based studies of the sequence determinants of Aβ aggregation have focused primarily on mutations that decrease aggregation. In contrast, genetic studies of early onset FAD in the human population have discovered mutations that increase aggregation propensity. In this later class, however, only a few mutants are known – presumably because those mutations that cause the most dramatic increase in aggregation are lethal and do not survive in the population. To augment the clinically isolated collection of aggregation prone mutants in Aβ, we have developed an unbiased screen for mutations that increase aggregation. Here we describe the implementation of this screen to isolate a collection of mutations that increase aggregation propensity beyond that of wild-type Aβ.
Previously, our lab described a high throughput screen for mutations in Aβ42 that inhibit aggregation. Our screen relied on fusions of Aβ42 to green fluorescent protein (GFP). In such fusions, the correct folding and fluorescence of GFP depends on the solubility of Aβ4224,31. Consequently, fusions of wild type Aβ42 to GFP yield colorless samples. However, mutations in Aβ42 that inhibit aggregation allow GFP to fold, and yield fluorescent samples. This fusion system was adapted to high throughput screening by expressing the fusion protein in E. coli and screening for green colonies on Petri dishes. We have used this system previously (i) to develop an unbiased screen for random mutations that diminish Aβ42 aggregation24; (ii) to probe the importance of side chains at positions 41 and 42 in causing Aβ42 to aggregate more readily than Aβ4026; (iii) to demonstrate that generic hydrophobic residues (as compared to specific side chains) at many positions in the Aβ42 sequence are sufficient to promote aggregation and fibrilogenesis25; and (iv) to develop a high throughput screen for small molecule inhibitors of Aβ42 aggregation32.
In the current report, we demonstrate that Aβ-GFP fusions can be used not only to isolate inhibitors of aggregation (either mutations in the sequence or exogenous small molecules), but also to find amino acid substitutions that actually enhance aggregation propensity.
Isolating mutants with enhanced aggregation propensity required that we use the Aβ-GFP screen in an inverse form: Instead of searching for rare green colonies (indicating Aβ solubility and GFP folding) amidst collections of white colonies, we sought to isolate rare white colonies (indicating Aβ aggregation and GFP misfolding) amidst collections of green samples. This inverse screen was made possible by modifying our original screen in two significant ways: First, instead of using fusions to Aβ42, we used fusions to the less aggregating Aβ40. Second, instead of running the screen at 37°C, we ran it at 30°C, where aggregation occurs more slowly. With these modifications, wild-type Aβ40-GFP fusions produce colonies that are slightly fluorescent26. Thus, we could isolate mutants that enhance aggregation by searching for rare white colonies amidst collections of slightly green colonies.
We constructed a library of random mutations in Aβ40 using error prone PCR. GFP fusions to this library were expressed at 30°C and screened for white colonies, indicating enhanced aggregation relative to the wild-type Aβ40-GFP fusion. In principle, however, white colonies could result either from amino acid substitutions in Aβ40 that increase its aggregation propensity or from spurious mutations including (i) deletion of all or part of the GFP construct; (ii) frameshifts and/or stop codons; (iii) diminished protein expression. To ensure that the white phenotype was not due to deletions, frameshifts, stop codons, or reduced expression, all white colonies were assayed for protein expression by SDS-PAGE (data not shown), and only those clones that expressed at levels similar to the wild-type Aβ40-GFP fusion were pursued for further studies. The sequences of these mutants are shown in Figure 1.
The aggregation propensities of the mutant sequences of Aβ40 were compared to wild type Aβ40 (and Aβ42) by measuring the fluorescence of cultures expressing the corresponding GFP fusions. As shown in previous work, there is a direct correlation between the fluorescence of such cultures and the solubility of the Aβ-GFP fusion: Fusions yielding lower fluorescence are less soluble24–26. Figure 2 shows that mutant versions of Aβ40-GFP fusions display a range of fluorescence. Some are similar to wild type Aβ40-GFP fusions, while others are considerably lower. Strikingly, some of the mutations in Aβ40 produce signals even lower than wild type Aβ42, indicating that these amino acid replacements cause Aβ40 to aggregate even more readily than wild type Aβ42. For example, GFP fusions to mutants WM1-WM5 (WM = White Mutant) show less fluorescence than the wild type Aβ42-GFP fusion. Considering that WM1-WM5 are mutants of Aβ40, and wild type Aβ40 is much less prone to aggregate than wild type Aβ42 3,4,33, these results indicate that the amino acid substitutions present in WM1-WM5 exert a substantial effect on the aggregation propensity of Aβ.
To confirm that mutations isolated by screening GFP fusions in E. coli actually increase the aggregation propensity of the Aβ peptide in isolation, we compared the rate of aggregation of mutant and wild-type forms of synthetic peptides (without GFP). Five peptides were synthesized and characterized: WM1 (Gln15→ Leu) and WM5 (Asp23→ Tyr) were chosen because they are the most aggregating single mutants. These were compared with a peptide corresponding to the clinically isolated Dutch mutant (Glu22→ Gln). All of these mutants were synthesized in the context of Aβ40. In addition, two versions of the wild-type sequence, Aβ40 and Aβ42, were also synthesized.
Each peptide was dissolved at 20 μM and incubated at 37°C under quiescent conditions. At various time points, aliquots were removed and mixed with Thioflavin T (ThT), which is widely used to assay for amyloid aggregates34. Binding of ThT was quantified by measuring fluorescence at 490 nm. As anticipated from the results with the GFP fusions (Fig. 2), mutants WM1 and WM5 have a much higher propensity to aggregate than the corresponding wild-type peptide, Aβ40 (Figure 3A). Moreover, at early time points (< 5 hours), WM1 and WM5 aggregate even faster than wild-type Aβ42. This is striking because the mutants were studied in the context of the shorter and less aggregation-prone 40-residue peptide.
After long incubations (1 to 2 weeks), Aβ42 produced slightly more aggregation than the WM1 and WM5 mutants in Aβ40 (Figure 3B). Both Dutch (in the context of Aβ40) and wild type Aβ40 do not produce significant quantities of amyloid until several days had elapsed. The relatively slow aggregation of the Dutch mutant was unexpected, since Dutch had been reported previously as a mutation that increases aggregation propensity 35. (Under agitated condition the Dutch mutant did indeed aggregate more rapidly than wild type Aβ40, with ThT fluorescence appearing after 30 min for the Dutch mutant and only after 150 min for wild type Aβ40 – data not shown).
The effects of the mutations on peptide fibrilogenesis were assessed by electron microscopy (EM). Peptides were dissolved at 20 μM in phosphate buffer and incubated under quiescent conditions. EM images were recorded after 1, 3, 7, 14 and 28 days. After 1 day of incubation, fibrils were observed for wild-type Aβ42 and for mutants WM1 and WM5 (Fig. 4A). Again, it is noteworthy that although the WM1 and WM5 mutants were studied in the context of the shorter Aβ40 sequence, they formed fibrils in the same time period as wild type Aβ42. As shown in figure 4A, wild type Aβ40 and the Dutch mutant of Aβ40 did not show fibrils after one day; these peptides formed fibrils more slowly, with fibrils only observed in EM images after a week of incubation (Fig. 4B). The EM results are consistent with the ThT assays: WM1 and WM5 (in the context of Aβ40) and wild type Aβ42 form fibrils rapidly, whereas the Dutch mutant and wild type Aβ40 are much slower.
The ThT and EM experiments described above were designed to detect the later steps in Aβ aggregation, namely the formation of amyloid fibrils. To compare the behavior of mutant and wild-type sequences in the earlier steps of the aggregation pathway, we monitored the disappearance of soluble peptide. Samples at a concentration of 10 uM were incubated at 37°C under quiescent conditions. After 1 or 3 days of incubation, the samples were centrifuged at 100,000 g for 30 min to pellet insoluble material. The amount of soluble peptide remaining in the supernatant was then quantified by reverse phase HPLC. In each case, the entire chromatogram was analyzed for the presence of soluble peptide corresponding to mutant sequences and/or multimeric species. As shown in figure 5, following one day of incubation, almost no soluble peptide could be detected for WM5(Aβ40) or wild type Aβ42. Thus these peptides had rapidly aggregated into insoluble material. In contrast, the Dutch mutant and wild type Aβ40 contained significant soluble peptide even after 3 days of incubation. These results are consistent with the fluorescence data for the GFP fusions (Fig. 2), the ThT assays (Fig. 3), and the EM analysis (Fig. 4).
The solubility studies of the WM1 mutant produced an unexpected result. WM1 had the lowest fluorescence in the GFP fusion assay, and showed the fastest kinetics of fibril formation in the ThT and EM assays; yet a significant amount of WM1 peptide remained soluble after one day of incubation. Analysis after 3 days of incubation still revealed soluble peptide (WM1 shows significant amount of soluble peptide even at 20 μM, the concentration at which ThT and EM assays were performed; data not shown). These observations are discussed below.
We described the construction of an artificial genetic system to screen for amino acid substitutions that increase the aggregation propensity of the Alzheimer’s peptide. The screen enabled the isolation and characterization of 20 different mutant sequences of Aβ40. The first 16 of these (WM1-WM16) show unambiguous phenotypes indicating enhanced aggregation relative to wild type Aβ40 (Figs. 1 & 2). [WM17-WM20 displayed borderline phenotypes similar to wild type, and will not be discussed further.]
Analysis of the 16 sequences with enhanced aggregation propensity (WM1 to WM16) shows that all of them contain amino acid substitutions that increase hydrophobicity. For example, WM1 has a Gln→ Leu mutation, and WM5 has an Asp→ Tyr mutation. In another set of examples, WM2 and WM13 both contain a Lys→ Thr substitution at the 16th residue. However, WM2 has the additional mutation Ala→ Val at position 21, while WM13 has the additional mutation Val→ Ala at position 24. Apparently, the increased hydrophobicity of the Ala→ Val substitution relative to the Val→ Ala substitution accounts for the enhanced aggregation propensity of WM2 relative to WM13 (Fig. 2).
WM5 (Asp23→ Tyr), WM9 (Glu22→ Ala) and WM17 (Glu22→ Asp) have single mutations in the region that forms a turn in structural models of Aβ fibrils 36–39. This is also true for the clinically isolated mutants, Dutch (Glu22→ Gln), Arctic (Glu22→ Gly) and Iowa (Asp23→ Asn).
As shown in figures 1 and and2,2, WM5 and WM9 lost a charge and show enhanced aggregation propensity. In contrast, WM17 has a mutation that conserves charge and has an aggregation propensity similar to wild type Aβ40. These data support the hypothesis that increased hydrophobicity and diminished charge enhance aggregation propensity (Ref 29 and below).
The phenotype of WM1 (Gln15→ Leu) in the GFP fusion system showed this sequence has the highest aggregation propensity of the mutants in the collection (Figs. 1 & 2). The enhanced propensity of WM1 to aggregate was confirmed by biochemical studies with synthetic peptide: Both the ThT assay (Fig. 3) and EM studies (Fig. 4) demonstrate that WM1 forms fibrils much more rapidly than wild type Aβ40. However, quantification of soluble peptide by HPLC showed that following a full day of incubation, a significant quantity of WM1 peptide remained in solution (Fig. 5). Taken together, these results show that (i) WM1 forms amyloid fibrils rapidly, but (ii) nonetheless a pool of material remains in solution.
To better understand these results, we analyzed the structural model of Aβ fibrils, built from solid state NMR constraints by Tycko and coworkers36. As shown in figure 6, in this model Gln15 is packed against Val36, Gly37 and Gly38. The mutation of Gln15 to leucine may interfere with this packing and thereby enable a significant fraction of the sample to avoid aggregation and remain in solution. At the same time, the increased hydrophobicity associated with the Gln→ Leu substitution might enhance the rate of aggregation for that fraction of the sample that successfully nucleates a structure with altered packing around the Leu side chain. The altered packing around the mutant leucine side chain may account for the observation that the initial fibrils formed by WM1 are shorter than those formed by other peptides (Fig. 4A). Dynamic equilibrium between these short fibrils and monomeric peptide would sustain significant levels of soluble material (Fig. 5).
Which features of an amino acid sequence are responsible for peptide aggregation? Dubay et al. formulated a theory to account for aggregation propensity as a function of the following properties: hydrophobicity, β-sheet propensity, α–helix propensity, binary patterning40 of polar and nonpolar residues, and charge.41 According to their theory, the aggregation propensity of a sequence (Pagg) can be calculated with the following equation,
where Ihyd, Iα, Iβ, Ipat and Ich are terms corresponding to the hydrophobicity, α-helix propensity, β-sheet propensity, binary patterning, and charge of a sequence, respectively. The coefficients αhyd, αα, αβ, αpat, αch assign the appropriate weighting for each of these parameters.
The fluorescence of mutant forms of Aβ40 fused to GFP can be compared with the aggregation propensity predicted by this equation. However, it must be noted that aggregation can include the formation oligomers or fibrils or both, and recent studies by Luheshi et al. indicate that the sequence dependence of the oligomerization rate and the fibrilization rate are similar but not identical42. Because we do not know how our mutants affect the relative rates of oligomerization versus fibrilization, we have compared the fluorescence of our mutations to the cumulative aggregation rate indicated by the equation above. As shown in figure 7, there is a strong inverse correlation between GFP fluorescence (which reports on solubility) and the aggregation propensity predicted by this equation.
The occurrence of early onset Familial Alzheimer’s Disease (FAD) has led to the clinical isolation of several mutations that alter the aggregation behavior of Aβ. Three clinical mutants that have been studied extensively are Dutch (E22Q), Iowa (D23N), and Arctic (E22G). In all three cases, the mutant peptide aggregates into protofibrils and/or fibrils more rapidly than wild-type Aβ 30,43–45. All three of these mutants contain amino acid substitutions that reduce charge and increase hydrophobicity. However, it is difficult to draw general conclusions about aggregation propensity from a collection comprising only three mutations at two positions in the sequence.
To supplement this naturally occurring collection, we have used an artificial genetic selection to isolate novel mutations that enhance Aβ aggregation. This new collection augments the naturally-occurring collection in the following ways:
Mutagenesis of Aβ40 was performed using nucleotide analogs as described in Ref. 46. Nucleotide analogs, dPTP and 8-oxo-dGTP (Trilink biotech, San Diego CA) were used for error prone PCR using Taq polymerase (Promega, Madison Wisconsin). After the first round of PCR, products were purified and used as templates for a 2nd round of PCR to replace the nucleotide analogs with A, G, T, and C. Purified PCR products were double digested using NdeI and BamHI (NEB, Ipswich MA) then cloned into a pET28 vector containing the GFP gene.
Plasmid pET28 vectors containing libraries of mutated Aβ40-GFP fusion genes were transformed into XL1-Blue competent cells (Stratagene, La Jolla CA) and plated for overnight growth as described previously24. Libraries of plasmids from these plates were recovered then transformed into BL21(DE3) (Stratagene, La Jolla CA) for screening. After transformation, cells were plated onto nitrocellulose filters placed on top of LB plates containing Kanamycin (35 μg/ml). After overnight growth at 37°C, the nitrocellulose filters were transferred onto a plate containing 1mM IPTG to induce protein expression. To produce the appropriate dynamic range of phenotype, plates were incubated at 30°C. At this temperature, colonies expressing wild type Aβ40-GFP fusion appear light green26. To select mutants with enhanced aggregation propensity, white colonies were picked. Since a white phenotype could also result from the failure of protein expression, all white colonies were also checked for protein expression. Those that expressed at levels similar to the wild-type fusion were sequenced. The GFP fluorescence of each mutant was quantified as describe in Ref. 24.
Crude peptides were purchased from the Keck Institute at Yale University, and purified using a C4 reverse phase (RP) column. Solvent gradients were run at 65°C using Solvent A: 95% water, 5% Acetonitrile, 0.1% TFA; and Solvent B: 50 % Acetonitrile, 50% water, 0.1% TFA. Molecular weights of the purified peptides were confirmed using mass spectrometry, and purity was checked using an analytical RP-HPLC C4 column (Vydac, CA). Purified peptides were treated with TFA to remove pre-existing aggregates47.
Peptides were dissolved in 300 μl DMSO and diluted with 6 ml of 8 mM NaOH (20 μM final peptide concentration). Concentrated PBS buffer was added to the solution to adjust the pH (final concentration: 50 mM NaH2PO4, 100 mM NaCl, 0.02%NaN3, pH 7.3–7.4). Samples were incubated at 37°C under quiescent conditions. At various time points, 500 μl of sample were mixed with 2.4 ml of a solution of ThT (7 μM ThT, 50 mM Glycerol-NaOH, pH 8.5) and fluorescence was measured at 490 nm (excitation at 450 nm).
Solutions were prepared as described above at peptide concentrations of 20 μM. Samples were incubated at 37°C under quiescent condition for 1, 3, 7, 14, 21, and 28 days. Following the incubation, Formvar carbon-coated grids were floated on a drop of each sample for 2 min, washed twice with distilled water, and then stained for 2 min with 1% uranyl acetate. Samples were imaged using a Zeiss 912ab Electron Microscope.
Peptides were dissolved to give a final concentration of 10 μM in 50 mM NaH2PO4, 100 mM NaCl, 0.02%NaN3 (pH 7.3~7.4). Each sample was incubated at 37°C under gently agitated or quiescent conditions. After incubation, samples were centrifuged at 100,000 X g for 30 min to remove insoluble materials, and soluble peptides in the supernatant were quantified by analytical RP-HPLC. Since different mutant peptides and different oligomeric states would be expected to elute at different points in the HPLC gradient, the entire chromatogram was scanned for the presence of peaks corresponding to the elution of any soluble species of peptide.
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.