Identification of the primary peptide inhibitor contaminant of fibrillation and toxicity in synthetic Amyloid-β42

Understanding the pathophysiology of Alzheimer disease has relied upon the use of amyloid peptides from a variety of sources, but most predominantly synthetic peptides produced using t-butyloxycarbonyl (Boc) or 9-fluorenylmethoxycarbonyl (Fmoc) chemistry. These synthetic methods can lead to minor impurities which can have profound effects on the biological activity of amyloid peptides. Here we used a combination of cytotoxicity assays, fibrillation assays and high resolution mass spectrometry (MS) to identify impurities in synthetic amyloid preparations that inhibit both cytotoxicity and aggregation. We identify the Aβ42Δ39 species as the major peptide contaminant responsible for limiting both cytotoxicity and fibrillation of the amyloid peptide. In addition, we demonstrate that the presence of this minor impurity inhibits the formation of a stable Aβ42 dimer observable by MS in very pure peptide samples. These results highlight the critical importance of purity and provenance of amyloid peptides in Alzheimer’s research in particular, and biological research in general.

After observing qualitative differences in the morphology of aggregating fibrils of r Aβ42 125 and sAβ42, we sought to determine if differences might also exist in the quantitative 126 dynamics of fibrillation. To this end we performed a standard fluorescence-based assay 127 to track the kinetics of aggregation in real-time. Thioflavin T (ThT) is a benzothiazole 128 dye that binds specifically to fibrillar Aβ and has been used for many years as a marker 129 in histological identification of Aβ plaques in brain tissue. ThT binds an interaction 130 surface unique to Aβ fibrils whereupon its fluorescence emission is greatly enhanced 109 . 131 Thus an increase in ThT signal represents increased binding to fibrils and is therefore a 132 direct readout for aggregation. Using the ThT dye assay, we observed distinct 133 aggregation dynamics between rAβ42 and sAβ42 (Error! Reference source not 134 found.). The kinetics of aggregation is set apart by two features of the fluorescent 135 traces. Firstly, the lag time to reach the half-maximal signal, t 1/2 , is about 10% longer in 136 the sAβ42 reaction indicating alterations to the early aggregation steps. Secondly, the 137 fast-phase aggregation trajectory is less steep for the sAβ42. Together with the EM 138 data, this indicates that the more highly branched fibrils observed in the synthetic 139 sample form with reduced kinetics when compared to the more linear fibrils of 140 recombinant peptide (Error! Reference source not found.2).

Mutations within the Aβ42 Glycine Zipper Motif Alter Fibrillation and Toxicity 146
We have observed that the ability of Aβ42 to aggregate correlates with its toxicity. This 147 correlation has been noted before[9,10,11], however it is not clear by what mechanism 148 the tendency toward fibrillation would drive toxicity. To further explore this connection, 149 we assayed if manipulations which impact aggregation behavior could also alter toxicity. 150 For this experiment we utilized the G37L mutant of the Aβ42 peptide. This mutation 151 disrupts the glycine zipper motif (Fig 4) that is important for the normal homo-152 oligomerization of Aβ42[10,12]. It has been observed that the G37L mutant peptide can 153 act as a dominant negative in an aggregation assay[9,10,13]. Thus, we performed a 1:1 154 mixing experiment between wild-type and G37L Aβ42 and applied this peptide to PC12 155 cells in the same toxicity experiment as above (Error! Reference source not 156 found.). We found that the G37L peptide exhibited a protective effect from Aβ42 157 toxicity (Fig 5). This result supports the claim that features of Aβ42 involved in 158 oligomerization are also important in toxicity. should still be able to produce a measurable effect even diluted 20 fold. Strikingly, we 187 observed that only 5% of the synthetic peptide was sufficient to perturb the normal 188 dynamics of Aβ42 aggregation (3). Both the time to t 1/2 and slope of rapid aggregation 189 phase indicate that the dynamics of aggregation have indeed been altered by the 190 minority species present in the synthetic peptide. In contrast, 5% doping of recombinant 191 Aβ42 into the synthetic peptide was not able to improve aggregation behavior (Error! Reference source not found.), again suggesting the sAβ42 contains an inhibitor of 193 aggregation. We next used this doping strategy to assess toxicity of Aβ42 mixtures in 194 PC12 cells. In agreement with our ThT data, just a small dose of synthetic peptide was 195 capable of reducing the toxicity of the recombinant peptide by a measurable amount 196

IDENTIFICATION OF Aβ INHIBITORS 206
Our data indicate that a potent inhibitor of both the aggregation (Error! Reference 207 source not found.) and toxicity (Fig 6) of Aβ42 exists as a minority product present in 208 the commercially obtained synthetic peptide. This finding has wide implications due to 209 the ubiquitous use of synthetic Aβ42 in AD research. However, due to the recent 210 commercial availability of recombinant Aβ42, it may not be necessary to improve upon 211 commercial synthesis and purification to avoid these inhibitory contaminants. Instead, 212 these unknown inhibitors may prove valuable as a starting point for the rational design 213 of molecules for therapeutic intervention in AD. Furthermore, these inhibitors serve as a unique and novel tool for probing the relationship between Aβ42 aggregation and 215

toxicity. 216
We began our study of the contaminants of synthetic Aβ42 by optimizing a 217 chromatographic method for both purification and analysis. We found conditions under 218 which we could elute peaks of recombinant Aβ42 from a C8 reverse-phase column in a 219 mobile phase of acetonitrile with TFA as a counter ion. We then used this method to 220 analyze the synthetic Aβ42 sample for the presence of contaminants. The same sharp 221 main peak existed, however a number of additional minor peaks were observed. There 222 was a distinct peak that eluted just before the main peak, as well a significant shoulder 223 on both the leading and trailing edges of the main peak (Fig 7). We also analyzed both recombinant and synthetic peptide samples by MALDI-TOF 231 mass spectrometry (Fig 8). This analysis revealed that the synthetic peptide sample 232 contains a diverse collection of contaminants; however due to their relatively low 233 abundance no specific identifications from this MALDI-TOF cocktail were possible. Most Aβ42-HIS). The Aβ42-HIS was doped into recombinant Aβ42 at 5% mole fraction, a 255 much higher concentration than might normally appear in the synthetic peptide, and 256 aggregation was monitored by ThT fluorescence (Fig 9). We found that even this large 257 dose of Aβ42-HIS produced a minor alteration in the recombinant peptide's velocity of 258 aggregation. Therefore, we conclude that although Aβ42-HIS may be capable of inhibiting Aβ42 aggregation, it alone is not sufficient to illicit effects of the magnitude 260 observed. Therefore, we believe that other as yet undiscovered inhibitors in synthetic 261 Aβ42 must exist. Using RP-HPLC (Fig 7) and mass spectrometry (Fig 8), we observed many 268 contaminating minority products in synthetic Aβ42. To determine if any of these 269 possessed inhibitory activity, synthetic Aβ42 samples were fractionated by RP-HPLC 270 and then assayed for inhibition of aggregation using the ThT assay. Specifically, the 271 main peak of synthetic Aβ42 was isolated from the contaminating material that eluted on 272 either side of it, yielding fractions labeled 1, 2 and 3 (Fig 7). As before, fractions were 273 doped into recombinant Aβ42 at 5% wt/wt for the aggregation assay. Significant 274 changes in aggregation dynamics were not observed in recombinant peptide doped with 275 fractions 2 and 3, but we found that fraction 1 contained extremely potent inhibitory 276 activity (Fig 10). It is important to note that this fraction was still a complex mixture; 277 therefore, it remains unclear whether its inhibitory activity is caused by a single or 278 multiple species. Without further analysis by more sensitive methods such as tandem 279 MS-MS, it is impossible to ascertain the identity of the active agents isolated in fraction 280 1, although based on the MALDI-TOF experimental results (Fig 8) we suspect them to 281 be related peptides. With the quantity of relevant contaminants available for study being extremely limiting we were prompted us to adopt a candidate-based approach towards 283 identification of Aβ42 activity inhibitors. 284 There are several sequence features which are known to pose challenges to 288 standard solid phase peptide synthesis methods, and therefore can inform rational 289 prediction of peptide synthesis byproducts for a given target. Highly hydrophobic 290 aggregation-prone sequences such as found in Aβ42 are known to be particularly 291 difficult to generate [14]. Traditional synthesis methods proceed from the carboxy-292 terminus towards the amino-terminus coupling a single amino acid at a time each 293 followed by a round of deprotection. Given that the carboxy-terminal 1 / 3 of Aβ42 is 294 derived from the transmembrane helix of APP (Fig 4), early steps in the synthesis of 295 Aβ42 are particularly susceptible to aggregation of the nascent peptide chain [14]. 296 Peptide aggregation competes with binding of synthesis reagents including deprotection 297 agents and subsequent amino acid residues. Because of the iterative nature of the 298 synthesis process, if either deprotection or coupling fails for a particular molecule in one 299 round, that molecule may well participate in future rounds of synthesis, yielding a final 300 peptide lacking only a single amino acid. In addition to these challenges, valine to valine 301 coupling reactions, which are used in Aβ42 synthesis, are known to be of lower 302 efficiency[15] than other peptide couplings in general. These facts led us to predict that 303 omission of valine 39 (hereinafter Aβ42Δ39) generated a putative byproduct of 304 synthesis that could be responsible for the observed inhibitory activity. This hypothesis 305 is supported by the presence of a peak mass observed by MALDI mass spectrometry 306 approximately 100 Da below the main peak. Additionally, the HPLC fraction containing 307 the inhibitory activity eluted slightly before Aβ42, indicating slightly reduced 308 hydrophobicity that is also consistent with deletion of a valine residue. 309 Absolute mass resolution by MALDI-TOF is inversely correlated with analyte 310 mass[16], and therefore the 3% difference in mass of Aβ42Δ39 from the total peptide 311 would prove difficult to resolve using solely this method. Digestion of the sample with 312 trypsin would generate a C-terminal fragment Aβ42 29-42 (Fig 4) was found to ionize very poorly by MALDI, leading to very weak signal. We did not see 329 evidence of a peak near 1153Da although we considered this uninformative as the 330 Aβ42Δ39 would represent only a very small fraction of the already weak signal from the 331 29-42 fragment. 332

333
To overcome the sensitivity hurdles of MALDI-TOF in the identification of potential 334 contaminants, we turned to ESI-Orbitrap mass spectrometry analysis on both 335 recombinant and synthetic Aβ42 (Fig 11). This data confirmed the presence of the 336 many contaminating species in synthetic Aβ42 observed by MALDI MS and HPLC. The 337 ESI-Orbitrap data allowed us to resolve individual peaks out of the shoulder present on 338 the Aβ42 peak seen by MALDI MS from approximately 4200 Da to 4500 Da. 339 Importantly, a peak of 4417.25 Da was observed, consistent with the presence of 340 Aβ42Δ39. Based on this finding, we proceeded with direct evaluation of Aβ42Δ39 as an 341 inhibitor of Aβ42 aggregation. Also of note was the presence of a peak in the 342 recombinant sample of Aβ42 dimer, M/Z 1290.52. This indicates not only that small 343 oligomers can form under the inhibitory conditions used (1% NH 4 OH), but that these 344 oligomers can remain intact through electrospray ionization, trapping and detection. The 345 absence of this peak in the synthetic sample appears to serve as further evidence of its 346 reduced ability to aggregate. 347