Title: Sequence Determinants of Enhanced Amyloidogenicity of Alzheimer Aβ42 Peptide Relative to Aβ40
Abstract: Aggregation of proteins into insoluble deposits is associated with a variety of human diseases. In Alzheimer disease, the aggregation of amyloid β (Aβ) peptides is believed to play a key role in pathogenesis. Although the 40-mer (Aβ40) is produced in vivo at higher levels than the 42-mer (Aβ42), senile plaque in diseased brains is composed primarily of Aβ42. Likewise, in vitro, Aβ42 forms fibrils more rapidly than Aβ40. The enhanced amyloidogenicity of Aβ42 could be due simply to its greater length. Alternatively, specific properties of residues Ile41 and Ala42 might favor aggregation. To distinguish between these two possibilities, we constructed a library of sequences in which residues 41 and 42 were randomized. The aggregation behavior of the resulting sequences was assessed using a high throughput screen, based on the finding that fusions of Aβ42 to green fluorescence protein (GFP) prevent the folding and fluorescence of GFP, whereas mutations in Aβ42 that disrupt aggregation produce green fluorescent fusions. Correlations between the sequences of Aβ42 mutants and the fluorescence of Aβ42-GFP fusions in vivo were confirmed in vitro through biophysical studies of synthetic 42-residue peptides. The data reveal a strong correlation between aggregation propensity and the hydrophobicity and β-sheet propensities of residues at positions 41 and 42. Moreover, several mutants containing hydrophilic residues and/or β-sheet breakers at positions 41 and/or 42 were less prone to aggregate than Aβ40 wherein these two residues are deleted entirely. Thus, properties of the side chains at positions 41 and 42, rather than length per se, cause Aβ42 to aggregate more readily than Aβ40. Aggregation of proteins into insoluble deposits is associated with a variety of human diseases. In Alzheimer disease, the aggregation of amyloid β (Aβ) peptides is believed to play a key role in pathogenesis. Although the 40-mer (Aβ40) is produced in vivo at higher levels than the 42-mer (Aβ42), senile plaque in diseased brains is composed primarily of Aβ42. Likewise, in vitro, Aβ42 forms fibrils more rapidly than Aβ40. The enhanced amyloidogenicity of Aβ42 could be due simply to its greater length. Alternatively, specific properties of residues Ile41 and Ala42 might favor aggregation. To distinguish between these two possibilities, we constructed a library of sequences in which residues 41 and 42 were randomized. The aggregation behavior of the resulting sequences was assessed using a high throughput screen, based on the finding that fusions of Aβ42 to green fluorescence protein (GFP) prevent the folding and fluorescence of GFP, whereas mutations in Aβ42 that disrupt aggregation produce green fluorescent fusions. Correlations between the sequences of Aβ42 mutants and the fluorescence of Aβ42-GFP fusions in vivo were confirmed in vitro through biophysical studies of synthetic 42-residue peptides. The data reveal a strong correlation between aggregation propensity and the hydrophobicity and β-sheet propensities of residues at positions 41 and 42. Moreover, several mutants containing hydrophilic residues and/or β-sheet breakers at positions 41 and/or 42 were less prone to aggregate than Aβ40 wherein these two residues are deleted entirely. Thus, properties of the side chains at positions 41 and 42, rather than length per se, cause Aβ42 to aggregate more readily than Aβ40. The deposition of insoluble proteins into amyloid plaque is associated with various diseases including Alzheimer disease, prion encephalopathies, Parkinson disease, and Huntington disease (1Koo E.H. Lansbury Jr., P.T. Kelly J.W. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 9989-9990Crossref PubMed Scopus (602) Google Scholar, 2Carrell R.W. Lomas D.A. Lancet. 1997; 350: 134-138Abstract Full Text Full Text PDF PubMed Scopus (820) Google Scholar, 3Thomas P.J. Qu B. Pedersen P.L. Trends Biochem. Sci. 1995; 20: 456-459Abstract Full Text PDF PubMed Scopus (484) Google Scholar, 4Slipe J.D. Annu. Rev. Biochem. 1992; 61: 947-975Crossref PubMed Scopus (409) Google Scholar). In Alzheimer disease, amyloid β (Aβ) 2The abbreviations used are: Aβ, amyloid β; GFP, green fluorescent protein; HFIP, hexafluoroisopropanol; HPLC, high pressure liquid chromatography; ThT, thioflavin T. peptides aggregate into fibrils, which accumulate as insoluble neuritic plaques. A variety of genetic, neuropathological, and biochemical studies suggests that either the fibrils themselves or precursors on the pathway toward these fibrils play a causative role in Alzheimer disease (1Koo E.H. Lansbury Jr., P.T. Kelly J.W. Proc. Natl. Acad. Sci. U. S. A. 1999; 96: 9989-9990Crossref PubMed Scopus (602) Google Scholar, 5Selkoe D. Physiol. Rev. 2001; 81: 741-766Crossref PubMed Scopus (5196) Google Scholar, 6Scheuner D. Eckman C. Jensen M. Song X. Citron M. Suzuki N. Bird T.D. Hardy J. Hutton M. Kukull W. Larson E. Levy-Lahad E. Viitanen M. Peskind E. Poorkaj P. Schellenberg G. Tanzi R. Wasco W. Lannfelt L. Selkoe D. Younkin S. Nat. Med. 1996; 2: 864-870Crossref PubMed Scopus (2282) Google Scholar, 7Weggen S. Eriksen J.L. Das P. Sagi S.A. Wang R. Pietrzik C.U. Findlay K.A. Smith T.E. Murphy M.P. Bulter T. Kang D.E. Marquez-Sterling N. Golde T.E. Koo E.H. Nature. 2001; 414: 212-216Crossref PubMed Scopus (1336) Google Scholar, 8Geula C. Wu C. Saroff D. Lorenzo A. Yuan M. Yankner B. Nat. Med. 1998; 4: 827-831Crossref PubMed Scopus (502) Google Scholar, 9Schenk D. Barbour R. Dunn W. Gordon G. Grajeda H. Guido T. Hu K. Huang J. Johnson-Wood K. Khan K. Kholodenko D. Lee M. Liao Z. Lieberburg I. Motter R. Mutter L. Soriano F. Shopp G. Vasquez N. Vandevert C. Walker S. Wogulis M. Yednock T. Games D. Seubert P. Nature. 1999; 400: 173-177Crossref PubMed Scopus (2965) Google Scholar, 10Bucciantini M. Giannoni E. Chiti F. Baroni F. Formigli L. Zurdo J. Taddei N. Ramponi G. Dobson C.M. Stefani M. Nature. 2002; 416: 507-511Crossref PubMed Scopus (2164) Google Scholar, 11Walsh D.M. Klyubin I. Fadeeva J.V. Cullen W.K. Anwyl R. Wolfe M.S. Rowan M.J. Selkoe D.J. Nature. 2002; 416: 535-539Crossref PubMed Scopus (3721) Google Scholar). Aβ peptides are produced in vivo by proteolytic cleavage of the amyloid precursor protein by β and γ secretases (5Selkoe D. Physiol. Rev. 2001; 81: 741-766Crossref PubMed Scopus (5196) Google Scholar). Because of inhomogeneous cleavage by γ secretase, Aβ peptides range in length from 39 to 43 residues. Among these peptides, the 40-mer (Aβ40) and the 42-mer (Aβ42) are abundant in diseased brains (12Kuo Y. Emmerling M.R. Vigo-Pelfrey C. Kasunic T.C. Kirkpatrick J.B. Murdoch G.H. Ball M.J. Roher A.E. J. Biol. Chem. 1996; 271: 4077-4081Abstract Full Text Full Text PDF PubMed Scopus (556) Google Scholar, 13Franz G. Beer R. Kampfl A. Engelhardt K. Schmutzhard E. Ulmer H. Deisenhammer F. Neurology. 2003; 60: 1457-1461Crossref PubMed Scopus (191) Google Scholar, 14Roher A.E. Lowenson J.D. Clarke S. Woods A.S. Cotter R.J. Gowing E. Ball M.J. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 10836-10840Crossref PubMed Scopus (623) Google Scholar), and these two peptides are the main components of the neuritic plaques in the parenchyma of diseased brains (5Selkoe D. Physiol. Rev. 2001; 81: 741-766Crossref PubMed Scopus (5196) Google Scholar, 14Roher A.E. Lowenson J.D. Clarke S. Woods A.S. Cotter R.J. Gowing E. Ball M.J. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 10836-10840Crossref PubMed Scopus (623) Google Scholar). With the exception of the C-terminal amino acids, Ile41 and Ala42, the sequences of Aβ40 and Aβ42 are identical. Despite their 95% sequence identity, Aβ40 and Aβ42 display dramatically different behaviors both in vivo and in vitro. Biochemical and immunocytochemical studies show that although Aβ40 is major component in cerebrospinal fluid and plasma, senile amyloid plaques formed in vivo are composed primarily of Aβ42 (5Selkoe D. Physiol. Rev. 2001; 81: 741-766Crossref PubMed Scopus (5196) Google Scholar, 14Roher A.E. Lowenson J.D. Clarke S. Woods A.S. Cotter R.J. Gowing E. Ball M.J. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 10836-10840Crossref PubMed Scopus (623) Google Scholar, 15Selkoe D.J. Neuron. 1991; 6: 487-498Abstract Full Text PDF PubMed Scopus (2209) Google Scholar, 16Gravina S.A. Ho L. Eckman C.B. Long K.E. Otvos Jr., L. Younkin L.H. Suzuki N. Younkin S.G. J. Biol. Chem. 1995; 270: 7013-7016Abstract Full Text Full Text PDF PubMed Scopus (593) Google Scholar). Moreover, although both peptides aggregate into fibrils in vitro, Aβ42 does so more rapidly than Aβ40. Concentrated solutions of Aβ40 are stable for days, whereas comparable solutions of Aβ42 aggregate almost immediately (17Jarrett J.T. Lansbury Jr., P.T. Cell. 1993; 73: 1055-1058Abstract Full Text PDF PubMed Scopus (1933) Google Scholar, 18Harper J.D. Lansbury Jr., P.T. Annu. Rev. Biochem. 1997; 66: 385-407Crossref PubMed Scopus (1420) Google Scholar). Not only is Aβ42 more prone to aggregate than Aβ40, but the pathways toward aggregation are also different. Recently, Teplow and coworkers (19Bitan G. Kirkitadze M. Lomakin A. Vollers S. Benedek G. Teplow D. Proc. Natl. Acad. Sci. U. S. A. 2003; 100: 330-335Crossref PubMed Scopus (1106) Google Scholar, 20Bitan G. Vollers S. Teplow D. J. Biol. Chem. 2003; 278: 34882-34889Abstract Full Text Full Text PDF PubMed Scopus (265) Google Scholar) found that carefully prepared aggregate-free Aβ40 occurs as monomers, dimers, trimers, and tetramers in rapid equilibrium. In contrast, Aβ42 preferentially forms pentamer/hexamer units (paranuclei) that assembled further into beaded superstructures similar to early protofibrils. The different aggregation behaviors of Aβ40 and Aβ42 led us to question the role of the two C-terminal residues, Ile41 and Ala42. Is it simply the increased length of Aβ42 that causes its increased amyloidogenicity? Or are particular features of the Ile41 and Ala42 side chains important for amyloid formation? To address these questions, we performed random mutagenesis on positions 41 and 42. The change in aggregation behavior resulting from the mutations was monitored by using fusions of the Aβ42 variants to GFP. Fluorescence of these fusions is inversely correlated with the propensity of the fused Aβ mutant sequence to aggregate (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar); GFP fused to wild type Aβ42 does not fluorescence, whereas fusions to less aggregating mutants display increased fluorescence. By correlating the observed levels of fluorescence with the identities of the side chains at positions 41 and 42, we determined the side chain properties at these positions that are responsible for the enhanced aggregation of Aβ42 relative to Aβ40. Construction of Libraries of Aβ42-GFP Fusions—Oligonucleotides were purchased from Integrated DNA Technologies (Coralville, IA), and restriction enzymes were purchased from New England Biolabs. Mutations were incorporated into a synthetic Aβ42 gene (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar) by PCR using DeepVent polymerase (New England Biolabs) and an Ericomp Easycycler™ Twinblock thermocycler. After PCR, the mutagenized Aβ42 gene inserts and the pET 28 vector containing the GFP gene (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar) were doubly digested with BamHI and NdeI. The digested insert and vector were then ligated together using T4 ligase. Plasmids were transformed into the XL1-Blue strain of Escherichia coli (Stratagene) and plated for overnight growth on plates containing 50 μg/ml kanamycin. The primer for random mutagenesis at position 41 and 42 had the following sequence: 5′-TCTTCTGGATCCNNNNNNCACCACGCCGCCCACCAT-3′. The primer for mutagenesis encoding a combinatorial mix of polar residues used the degenerate codon NAN. For a combinatorial mix of hydrophobic residues we used the degenerate codon NTN (where N denotes a mixture of A, G, C, and T). Screening of Green/White Phenotype—DNA libraries were isolated from E. coli strain XL1-Blue, transformed into BL21(DE3) (23Studier F.W. Rosenberg A.H. Dunn J.J. Dubendorff J.W. Methods Enzymol. 1990; 185: 60-89Crossref PubMed Scopus (6006) Google Scholar), and plated onto nitrocellulose paper (Millipore NC-HATF 83 mm) on LB plates containing 50 μg/ml kanamycin. Following overnight growth at 37 °C, the nitrocellulose papers were transferred to LB plates containing 50 μg/ml kanamycin and 1 mm isopropyl 1-thio-β-d-galactopyranoside and incubated at 30 °C for 4 h to induce the expression of the Aβ-GFP fusion protein. Colonies were counted and the green/white phenotype was recorded (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar). Fluorescence Measurements—To enable measurement of the fluorescence of the Aβ42-GFP fusions in vivo, colonies were picked, and cultures were grown in LB liquid media containing 50 μg/ml kanamycin. When cultures reached an absorbance at 600 nm of 0.8, expression was induced by addition of isopropyl 1-thio-β-d-galactopyranoside to a concentration of 1 mm, and growth was continued for an additional 5 h 30 min at 30 °C. After induction, cultures were diluted in Tris-buffered saline to an A600 nm of 0.15. Fluorescence was measured using a 50B spectrofluorimeter (PerkinElmer Life Sciences) with excitation at 490 nm and emission at 510 nm. Expression of Aβ42-GFP fusions was assessed by removing 200 μl of cell culture and analyzing the whole cell content by SDS-PAGE. Correlation of Fluorescence with Biophysical Properties—The fluorescence of the Aβ42-GFP fusions in vivo was plotted against the sum of hydrophobicities or against the sum of the β-sheet propensities at positions 41 and 42. Hydrophobicities were based on the scale of Kyte and Doolittle (24Kyte K. Doolittle R. J. Mol. Biol. 1982; 157: 105-132Crossref PubMed Scopus (17296) Google Scholar), and β-sheet propensities were based on the scale of Minor and Kim (25Minor Jr., D. Kim P. Nature. 1994; 367: 660-663Crossref PubMed Scopus (554) Google Scholar). In a further analysis, fluorescence was plotted against the predicted aggregation rates using Equation 1 as developed by Chiti et al. (26Chiti F. Stefani M. Taddei N. Ramponi G. Dobson C.M. Nature. 2003; 424: 805-808Crossref PubMed Scopus (931) Google Scholar). ln(νmut/νwt)=AΔHydr+B(ΔΔGcoil−α+ΔΔGβ−coil)+CΔCharge(Eq.1) In Equation 1, νmut and νwt are the aggregation rates of the mutant and wild type sequences, respectively; ΔHydr is the difference in hydrophobicity, and (ΔΔGcoil-α +ΔΔGβ-coil) is the difference in the propensity to convert from α-helix to β-sheet. (Note, the two ΔΔs are added rather than subtracted. This is because subtraction is already accomplished by the definition of the terms. In the first term α is subtracted from coil, and in the second term coil is subtracted from β.) ΔCharge is the difference in charge, and A, B, and C are empirically determined constants. Synthetic 42-Residue Peptides—Crude peptides were purchased from the Keck Institute, Yale University, and were then purified using reverse phase HPLC. Their identities were confirmed by mass spectrometry, and purity was assessed by analytical reverse phase HPLC. The purities of the peptides were greater than 92%. After purification, peptides were lyophilized and dissolved in hexafluoroisopropanol (HFIP). HFIP was removed by blowing argon over the sample. Samples were then dissolved in 100 μl of Me2SO/500 μl of 4 m NaOH. Concentrations were determined using the extinction coefficient of tyrosine. Aggregation of Mutant Peptides in Vitro—Peptides were dissolved at a concentration of 10 μm in 50 mm NaH2PO4, 100 mm NaCl, 0.02% NaN3 (pH 7.3-7.4). Following incubation at 30 °C for 1 day, samples were centrifuged for 30 min at 60,000 × g. Half of the supernatant was removed and loaded onto reverse phase HPLC to quantify the monomer remaining in solution. To correlate peak size with concentration, a series of standards was run using a range of concentrations of the Asp41-Gln42 mutant. The relationship between peak size and peptide concentration was confirmed using same concentrations of wild type Aβ42, the hydrophobic mutant Leu41-Leu42, and the green control mutant Ser19-Pro34. Thioflavin T Assay—Because the kinetics of fibril formation can be affected by small quantities of seeds, preexisting seeds were removed by the following treatment (27Jao S. Ma K. Talafous J. Orlando R. Zagorski M.G. Amyloid: Int. J. Exp. Clin. Investig. 1997; 4: 240-252Crossref Scopus (104) Google Scholar). Lyophilized peptides were dissolved in trifluoroacetic acid at ∼1 mg/ml and sonicated for 15 min. Trifluoroacetic acid was then removed by blowing argon over the sample. The dry sample was resuspended in 2 ml of HFIP. An aliquot corresponding to 0.5 mg of peptide was dried under argon and then dissolved in 300 μl of Me2SO and mixed with 5 ml of 8 mm NaOH. The solution was then centrifuged at 40,000 × g for 30 min to remove any insoluble material. The supernatant was removed, and phosphate-buffered saline was added to a final concentration of 50 mm NaH2PO4, 100 mm NaCl, 0.02% NaN3. The pH was adjusted to 7.4-7.5 with 200 mm formic acid. 500-μl aliquots were removed every 30 min and mixed with 2.4 ml of a ThT solution (7 μm thioflavin T, 50 mm glycine-NaOH (pH 8.5)). Fluorescence was measured with excitation at 450 nm, and emission at 490 nm. A High Throughput Screen for Aβ Aggregation—Random mutagenesis of position 41 and 42 can produce 400 possible sequences (including wild type). Synthesis, purification, and characterization of the aggregation behavior of all 400 peptides would be laborious and extremely expensive, particularly because the Aβ42 peptide is notoriously difficult to synthesize and purify (28Murakami K. Irie K. Morimoto A. Ohigashi H. Shindo M. Nagao M. Shimizu T. Shirasawa T. Biochem. Biophys. Res. Commun. 2002; 294: 5-10Crossref PubMed Scopus (123) Google Scholar, 29Burdick D. Soreghan B. Kwon M. Kosmoski J. Knauer M. Henschen A. Yates J. Cotman C. Glabe C. J. Biol. Chem. 1992; 267: 546-554Abstract Full Text PDF PubMed Google Scholar). Therefore, we developed a high throughput screen using GFP as a reporter tag (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar). The folding of GFP and the formation of its active chromaphore occur relatively slowly (30Cubitt A. Heim R. Adams S. Boyd A. Gross L. Tsien R. Trends Biochem. Sci. 1995; 20: 448-455Abstract Full Text PDF PubMed Scopus (1207) Google Scholar). Therefore, peptides or proteins fused to the N terminus of GFP can have a dramatic impact on fluorescence. Sequences that aggregate rapidly prevent formation of a correctly folded fluorescent GFP. In contrast, sequences that are soluble or aggregate slowly allow GFP to fold into its native fluorescent structure. In a systematic study using 20 different test proteins, Waldo et al. (31Waldo G.S. Standish B.M. Berendzen J. Terwilliger T.C. Nat. Biotechnol. 1999; 17: 691-695Crossref PubMed Scopus (725) Google Scholar) demonstrated that the fluorescence of E. coli cells expressing fusions to the N terminus of GFP correlated with the solubility of the test protein expressed alone. We have shown previously that the fluorescence of Aβ42-GFP fusions can be used as an unbiased screen for the sequence determinants of Aβ amyloidogenesis (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar). Colonies of E. coli expressing fusions to wild type Aβ42 or to mutants of Aβ42 that favor rapid aggregation are white. In contrast, colonies expressing fusions to soluble (or slowly aggregating) mutants of Aβ42 are green. Here we used Aβ42-GFP fusions to monitor the phenotypic variation resulting from amino acid substitutions at positions 41 and 42 of Aβ42. This enabled a high throughput screen in which more than 2500 colonies could be screened per plate. To ensure that the screen had sufficient dynamic range to distinguish a range of phenotypes, we first performed pilot experiments using fusions of GFP to the wild type sequences of either Aβ42 or Aβ40. (The latter construct can be considered as the null mutant in which residues 41 and 42 are not merely substituted but deleted entirely.) At 37 °C both fusions produced white colonies. Thus, at this temperature, the screen does not have sufficient dynamic range. In contrast, at 30 °C, colonies expressing fusions to Aβ42 were white, whereas those expressing fusions to Aβ40 were green. Thus, at this lower temperature, where expression and subsequent aggregation are slower, the difference in aggregation propensity between Aβ40 and Aβ42 is easily distinguished. Further pilot experiments showed that at 37 °C, virtually all mutations at positions 41 and 42 produced white colonies, whereas at 30 °C libraries of mutants at 41 and 42 produced both green and white colonies. Because expression at 30 °C provided excellent dynamic range, all further experiments were performed at this temperature. A Random Library of Mutants at Positions 41 and 42—A library of random mutants at positions 41 and 42 of Aβ42 was constructed using synthetic oligonucleotides to incorporate random bases (NNN) at the final 2 codons of a synthetic gene encoding Aβ42 (see “Materials and Methods”). The library of mutant genes of Aβ42 was then fused to the 5′ end of a gene encoding GFP. In this construct, the Aβ42 sequence is separated from GFP by a 12-residue linker encoding the sequence Gly-Ser-Ala-Gly-Ser-Ala-Ala-Gly-Ser-Gly-Glu-Phe (31Waldo G.S. Standish B.M. Berendzen J. Terwilliger T.C. Nat. Biotechnol. 1999; 17: 691-695Crossref PubMed Scopus (725) Google Scholar). The library of fusions was then transformed into XL1 Blue cells. Transformation yielded more than 5000 colonies. Because there are only 400 possible combinations of amino acids positions 41 and 42, this library is adequate to sample all (or nearly all) of the possible sequences (TABLE ONE). The library was then extracted from XL1 Blue cells and transformed into BL21(DE3) cells for protein expression and high throughput screening.TABLE ONEDiversity of the libraries Random at both 41, 42 Hydrophilic at both 41,42 Hydrophobic at both 41,42 Pro41-Random42 Gly41-Random42 Random41-Pro42 Random41-Gly42 Random41-Ala42 Ile41-Random42 No. colonies analyzed >5000 >5000 >5000 >2000 >2000 >2000 >1500 >1000 >1200 No. possible amino acid combinations 400 49 25 20 20 20 20 20 20 Open table in a new tab To assess the effect of amino acid substitutions at residues 41 and 42 on the aggregation of Aβ42, the library of Aβ42-GFP fusions was expressed at 30 °C, and the resulting colonies were characterized. In a typical experiment, 1520 colonies were analyzed. Of these, 140 colonies (9%) were white, similar to colonies expressing GFP fusions to wild type Aβ42. The remaining 1380 colonies (91%) showed some level of green fluorescence (TABLE TWO). 34 clones from this library were chosen arbitrarily for further characterization by sequence analysis and fluorescence measurements. The correlation between sequence and fluorescence is shown in Fig. 1.TABLE TWOGreen/white phenotypic screening of colonies expressing Aβ(42)-GFP fusions Random at both 41, 42 Hydrophobic at both 41,42 Hydrophilic at both 41,42 Pro41-Random42 Gly41-Random42 Random41-Pro42 Random41-Gly42 Random41-Ala42 Ile41-Random42 White colonies 140 5000 640 17 18 25 30 37 44 Green colonies 1380 0 4160 210 160 220 290 153 141 Open table in a new tab The Effect of Hydrophobicity at Residues 41 and 42—As shown in Fig. 1A, GFP fusions to mutants with hydrophobic residues at positions 41 and 42 display low fluorescence, similar to fusions to wild type (Ile41-Ala42)Aβ42. In contrast, fusions to mutants with hydrophilic residues at positions 41 and 42 show high fluorescence, similar to fusions to Aβ40. These results suggest that hydrophobic side chains at positions 41 and 42 increase the propensity of Aβ42 to aggregate. To confirm this correlation, we produced two additional libraries of mutants. In one library, residues 41 and 42 were mutated to a combinatorial mixture of hydrophobic residues. In the other library, these residues were mutated to a combinatorial mixture of hydrophilic residues. The hydrophobic library was constructed using the degenerate DNA codon NTN to encode a mixture of nonpolar residues including Ile, Leu, Val, Met, and Phe (32Hecht M. Das A. Go A. Bradley L. Wei Y. Protein Sci. 2004; 13: 1711-1723Crossref PubMed Scopus (177) Google Scholar). The hydrophilic library was constructed using the NAN codon to encode a mixture of polar residues including Glu, Gln, Asp, Asn, Lys, His, and Tyr (where N denotes a mixture of the four DNA bases). Both libraries were plated, and protein was expressed at 30 °C. For the 41/42-hydrophobic library, >5000 colonies were analyzed and none were green. Thus, randomly chosen hydrophobic residues at positions 41 and 42 support Aβ42 aggregation. In contrast, for the 41/42-hydrophilic library, 4800 colonies were analyzed, and 87% of them (4160 colonies) were green. This result indicates that randomly chosen polar residues at positions 41 and 42 disrupt the aggregation of Aβ42. From the hydrophilic library, only 640 colonies (13%) were white. We tested 64 of these colonies for expression of the Aβ42-GFP fusions, and we found that all but one of them did not express the fusion protein. (Because we used the NAN degenerate codon at positions 41 and 42, it was expected that the library would contain stop codons, which prevent protein expression.) The other white colony expressed normal levels of the fusion protein. Sequence analysis of this clone revealed that it had arginine at both positions 41 and 42. Thus, although most randomly chosen polar residues at positions 41 and 42 interfere with Aβ aggregation, arginine is somehow special and allows aggregation (see below). Effect of the β-Sheet Propensity of Residues 41 and 42—Amyloid structures are dominated by β-sheet secondary structure (33Serpell L.C. Biochim. Biophys. Acta. 2000; 1502: 16-30Crossref PubMed Scopus (833) Google Scholar, 34Makin O.S. Atkins E. Sikorski P. Johansson J. Serpell L.C. Proc. Natl. Acad. Sci. U. S. A. 2005; 102: 315-320Crossref PubMed Scopus (538) Google Scholar), and previous studies suggested a strong correlation between the tendency of a sequence to form amyloid and the β-sheet propensities of its component amino acids (26Chiti F. Stefani M. Taddei N. Ramponi G. Dobson C.M. Nature. 2003; 424: 805-808Crossref PubMed Scopus (931) Google Scholar, 35Johansson J. Weaver T.E. Tjernberg L.O. Cell. Mol. Life Sci. 2004; 61: 326-335Crossref PubMed Scopus (27) Google Scholar, 36Päiviö A. Nordling E. Kallberg Y. Thyberg J. Johansson J. Protein Sci. 2004; 13: 1251-1259Crossref PubMed Scopus (51) Google Scholar, 37Kallberg Y. Gustafsson M. Persson B. Thyberg J. Johansson J. J. Biol. Chem. 2001; 276: 12945-12950Abstract Full Text Full Text PDF PubMed Scopus (272) Google Scholar, 38Pawar A.P. Dubay K.F. Zurdo J. Chiti F. Vendruscolo M. Dobson C.M. J. Mol. Biol. 2005; 350: 379-392Crossref PubMed Scopus (511) Google Scholar). Of the 34 sequences analyzed above, 10 sequences contained proline or glycine at positions 41 or 42. These residues have very low intrinsic propensities for β-sheet structure (39Creighton T. Proteins,. W. H. Freeman & Co., New York1992: 171-200Google Scholar). All 10 of these mutants displayed high levels of fluorescence (Fig. 1B). To study further the role of β-sheet propensity at residues 41 and 42, we constructed additional libraries in which β-sheet breakers were incorporated intentionally at position 41 or 42. Four libraries were constructed as follows: Pro41-Random42, Gly41-Random42, Random41-Pro42, and Random41-Gly42. For each library, 20 different sequences are possible. To ensure extensive sampling, 178-320 colonies were analyzed for each library (TABLE TWO). For all four libraries, ∼90% of colonies were green. The non-green colonies were assayed for protein expression, and none expressed the fusions. (As described above, non-expressers arise from stop codons encoded by the NNN degenerate codon.) These results indicate that the ability of residues 41 and 42 to form β-sheet structure is important for Aβ aggregation. Effect of the Site of Mutation—To assess which of the two positions affects aggregation more significantly, we constructed two additional libraries. In each of these libraries one of the two C-terminal residues was held constant, whereas the other was randomized. These libraries were called Ile41-Random42 and Random41-Ala42. (The wild type sequence is Ile41-Ala42.) For each library, nearly 200 colonies were analyzed. The Random41-Ala42 library yielded 37 white colonies (19% white), and 153 green colonies. The Ile41-Random42 library yielded 44 white colonies (24% white), and 141 green colonies (TABLE TWO). Because the fraction of white colonies in the two libraries is not significantly different, we conclude that the wild type residues at positions 41 and 42 contribute to aggregation to a similar extent. Aggregation of Mutant Peptides in Vitro—Waldo et al. (31Waldo G.S. Standish B.M. Berendzen J. Terwilliger T.C. Nat. Biotechnol. 1999; 17: 691-695Crossref PubMed Scopus (725) Google Scholar) and Wurth et al. (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar) demonstrated the validity of GFP fusions as a screen for the aggregation behavior of proteins and peptides. In particular, Wurth et al. (21Wurth C. Guimard N.K. Hecht M.H. J. Mol. Biol. 2002; 319: 1279-1290Crossref PubMed Scopus (196) Google Scholar) showed that the fluorescence in vivo of a mutan