Aß40 mutants protect Drosophila against Aß42 toxicity despite their amyloidogenic properties in the non-transgenic mouse brain


 BackgroundSelf-assembly of the amyloid-β (Aβ) peptide into aggregates, from small oligomers to amyloid fibrils, is fundamentally linked with Alzheimer’s disease (AD). However it is clear that not all forms of Aβ are equally harmful, and that linking a specific aggregate to toxicity also depends on the assays and model systems used [1, 2]. Though a central postulate of the amyloid cascade hypothesis, there remain many gaps in our understanding regarding the links between Aβ deposition and neurodegeneration.MethodsIn this study, we examined familial mutations of Aβ that increase aggregation and oligomerization, E22G and DE22, and induce cerebral amyloid angiopathy, E22Q and D23N. We also investigated synthetic mutations that stabilize dimerization, S26C, and a phospho-mimetic, S8E, and non-phospho-mimetic, S8A. To that end, we utilized BRI2-Aβ fusion technology and rAAV2/1 based somatic brain transgenesis in mice to selectively express individual mutant Aβ species in vivo . In parallel we generated PhiC31-based transgenic Drosophila melanogaster expressing wild type (WT) and Aβ40 and Aβ42 mutants, fused to the Argos signal peptide to assess the extent of Aβ42-induced toxicity as well as to interrogate the combined effect of different Aβ40 and Aβ42 species.ResultsWhen expressed in the mouse brain for 6 months, Aβ42 E22G, Aβ42 E22Q/D23N, and Aβ42WT formed amyloid aggregates consisting of some diffuse material as well as cored plaques, whereas other mutants formed predominantly diffuse amyloid deposits. Moreover, while Aβ40WT showed no distinctive phenotype, Aβ40 E22G and E22Q/D23N formed unique aggregates that accumulated in mouse brains. This is the first evidence that mutant Aβ40 overexpression leads to deposition under certain conditions. Interestingly, we found that mutant Aβ42 E22G, E22Q, and S26C, but not Aβ40, were toxic to the eye of Drosophila . In contrast, flies expressing a copy of Aβ40 (WT or mutants) in addition to Aβ42WT, showed improved phenotypes, suggesting possible protective qualities for Aβ40.ConclusionsThese studies suggest that some Aβ40 mutants form unique amyloid aggregates in mouse brains, despite protecting against Aβ42 toxicity in Drosophila , which highlights the significance of using different systems for a better understanding of AD pathogenicity and more accurate screening for new potential therapies.

early and central process in the development and progression of AD [3,4]. Aβ with differing C-termini is produced through the sequential cleavage of the amyloid-β precursor protein (APP) by β-and γ-secretases [5,6]. Although the major species generated is Aβ40, Aβ42 is much more amyloidogenic and is considered the toxic species despite being generated at lower levels (~5-10% of total A ). It is currently unknown why Aβ, which is a naturally produced protein, begins to misfold and aggregate in the brain. While most instances of AD are sporadic, 10-15% of AD cases are due to familial mutations of A and result in extensive amyloid pathology within the brain and vasculature, called amyloidosis cerebral amyloid angiopathy (CAA) [5]. Strong evidence suggests that the vast majority of these mutations are associated with earlier disease onset, faster amyloid aggregation, and more aggressive toxicity to the cells [7]. It is widely believed that amyloid aggregates in the brain can form diffuse as well as compact plaques. Nevertheless, the exact role of mutant Aβ species in the underlying pathology has not been shown. The pathological phenotypes caused by mutations that alter amino acid residues within the Aβ sequence are variable, and the underlying pathogenetic mechanisms are not fully understood [8]. In vitro studies have revealed that bril formation of Aβ is a complex process in which nucleation of assembly is the rate-limiting step [9,10]. Recent data support the notion that intra-Aβ amino acid substitutions affect peptide self-association [11]. Several familial forms of AD, characterized by single amino acid mutations at residues E22 or D23 of Aβ, located in the turn of the -hairpin, include the Italian (E22K), Arctic (E22G), Dutch (E22Q), and Iowa (D23N) familial mutants. Aβ40 and Aβ42 variants containing these mutations have faster folding nucleation and have been found to be more neurotoxic and to aggregate more readily than the wild-type (WT) peptide in in vitro experiments. A 42 E22Q, D23N, and E22K mutations cause hereditary cerebral hemorrhage with amyloidosis, supporting the evidence that Aβ mutants at positions 22 and 23 show increased neurotoxicity than wild-type Aβ [12,13]. The E22G mutation causes early onset AD that involves enhanced proto bril formation [14]. Another familial mutation reported in recent years, termed Osaka ( E22), causes severe dementia, cerebellar ataxia, and gait disturbances in the absence of senile plaques [15]. It has been shown that deletion of E22 results in enhanced oligomerization of Aβ [16].
To further understand Aβ aggregation dynamics, previous studies have introduced mutations to stabilize Aβ aggregates. It has been proposed, for example, that Aβ dimer, designed by a cross-linked replacement of Ser 26 with cysteine, rapidly forms toxic proto brils [17,18]. Additionally, it was postulated, that phosphorylation of serine residue 8 promotes aggregation by stabilization of β-sheet conformation of Aβ and increased formation of oligomeric Aβ aggregates that represent nuclei for brillization [19].
A number of AD mouse models were developed in recent years, based on overexpression of transgenes containing FAD mutations. Most of these models exhibited age-dependent amyloid deposition in the brain along with thio avin-S-positive plaques, including compact plaques with dense cores that are reminiscent of those seen in human AD [20,21]. Different promoters and genetic backgrounds prevent comparison of these models and the fact that full-length APP is overexpressed can in uence the respective development of behavioral and pathological features. Several years ago new mouse models were developed based on genetic constructs that express fusion proteins between the BRI2 protein, involved in amyloid deposition in familial British dementia and Aβ. These mice express Aβ peptides in absence of APP. BRI2-Aβ40 mice do not exhibit amyloid pathology, whereas BRI2-Aβ42 mice accumulate detergent-insoluble Aβ as they age [22].
Here, we used mouse models to investigate the potential pathogenic role of mutant Aβ peptides in vivo. To do so, we used recombinant adeno-associated virus (rAAV) vectors to express A E22G, E22Q/D23N, E22, S8E phospho-mimetic and S8A non-phospho-mimetic, and S26C dimer mutants using the BRI2 fusion strategy [23][24][25]. This approach allows individual delivery of Aβ mutants to the mouse brain to compare their aggregation patterns [25,26]. Although mouse models are useful at examining amyloid pathology and glial involvement, they typically do not show AD relevant neurodegenerative changes. Therefore, we then expressed A 40/A 42 E22G, E22Q, and A 42 S26C in Drosophila to examine neurotoxicity of the mutant Aβ transgenes under similar expression levels. The questions we are asking are: Do Aβ mutant aggregates in vivo display the characteristics of individual Aβ strains? Does Aβ that aggregates faster cause neuronal toxicity? Does Aβ40 bearing an aggregation-prone mutation, accumulate in vivo and cause toxicity? Does the theory of "templating" the aggregation work when the seed is formed from individual Aβ species and not from brain homogenate?

Methods
Mice and neonatal injections B6C3H-F1 mice were obtained from Envigo. All animal procedures were performed with approval from the University of Florida Institutional Animal Care and Use Committee. All animals were housed three to ve to a cage and maintained on ad libitum food and water with a 12 h light/dark cycle. Intracerebroventricular injections of rAAVs were carried out on day P0 as described previously [27]. Two microliters of rAAV2/1 encoding Aβ42 WT, Aβ42 E22G, Aβ42 E22Q/D23N, Aβ42 ΔE22, Aβ42 S8A, Aβ42 S26C, Aβ40 WT, Aβ40 E22G, or Aβ40 E22Q/D23N were administered bilaterally. At endpoint, mice were euthanized, brains harvested, one hemibrain was xed overnight in 4% paraformaldehyde solution at 4 °C for immunohistochemical staining. Another hemibrain was ash frozen for biochemical fractionation and ELISA.
Immunohistochemical imaging Right hemisphere of all injected mice was xed in formalin, embedded in para n, sectioned, and stained with a biotinylated pan-Aβ antibody Ab5 (1:500, human Aβ1-16 speci c; T.E. Golde). Immunohistochemically stained sections were captured using the Scansope XT image scanner (Aperio; Leica Biosystems) or BX 60 (Olympus) and analyzed using ImageScope program.
Generation of transgenic ies expressing mutant A β species To generate mutant transgenic ies expressing comparable levels of Aβ mutants, all cDNA sequences (Aβ40 WT, E22G, E22Q, and S26C, and Aβ42 WT, E22G, E22Q, and S26C) were cloned under the UAS of the pJFRC-MHU vector, which carries an attB site for site-directed integration. All Aβ peptides were fused to the Argos signal peptide to ensure secretion. The resulting constructs were microinjected into yellow white (yw) embryos at Rainbow Transgenics (Camarillo, CA) and targeted to the same genomic location, the attP2 site on chromosome 3, to achieve similar expression levels in vivo. At least two transgenic lines for each Aβ construct were established. The ies were raised and maintained at 25 °C in regular media prior to experimentation. To express the Aβ constructs, we combined these transgenic lines with the glass multimer reporter (GMR)-Gal4 (all eye cells) and ELAV-Gal4 (pan-neural) drivers. pJFRC-MUH was a gift from G. Rubin (plasmid #26213; Addgene; [29]).
Drosophila Eye imaging To generate the eye images, we crossed GMR-Gal4 or GMR-Gal4;Aβ40/42 stocks with each mutant transgene at 25 °C. Two days after eclosion, we collected females from the progeny, serially dehydrated in ethanol, air-dried in hexamethyldisilazane (Electron Microscope Sciences), and metal-coated for photodocumentation in a Jeol 5000 scanning electron microscope.
Negative geotaxis assay To assess the ability of the different Aβ peptides to disrupt locomotor function, we crossed Aβ ies with the ELAV-Gal4 driver to direct pan-neural expression of the transgenes. All crosses were set at 25 °C. After eclosion, virgin females were collected in groups of 25 ies and transferred to 26 °C to perform climbing assays as described by us [30,31]. Brie y, each day, groups of 25 ies were transferred to at bottom empty vials (9.5 cm high, 1.7 cm in diameter) and analyzed for their ability to climb to the top. At the beginning of each test session, we rmly tapped the vial against the bench surface forcing ies to the bottom. After eight seconds, we recorded the number of ies that climbed up the walls of a vial above a 5-cm mark. Scores recorded were the mean number of ies climbing to the criterion during each daily session.
Statistical analyses For climbing assays in ies, we calculated the climbing index for each Aβ transgene. The climbing index is the percentage of ies climbing above the criterion of 5-cm mark (no. ies above 5cm mark/total no. of ies x 100). The climbing index was averaged from six trial of ve replicates, containing 25-30 ies, for each day and normalized. All replicates were averaged to compare climbing by ANOVA with Duncan's post hoc analysis (GraphPad Software). Final images were created using Photoshop CS5 (Adobe Systems). All values in the text and gures represent means ± standard error of the mean.

Results
Overexpression of Aβ peptides via AAV delivery results in amyloid deposits in the mouse brain.
FAD-related Aβ mutant species are prone to accelerate aggregation and increase toxicity compared to Aβ WT [14,17,[32][33][34][35][36][37]. We selected speci c mutations associated with aggressive formation of aggregates, such as oligomers or brils, both in a test tube and in vivo. Figure 1 illustrates the numerous mutants that were assessed in this study. All BRI2-Aβ40 and Aβ42 mutant constructs were packaged into the rAAV2/1 viral cassette, expressed in HEK cells, and the truncation and proper secretion of the Aβ peptide to the media was con rmed (Fig. S1). Various levels of Aβ, were detected by Western blotting and sandwich Elisa, suggesting differences in half-life and stability of the different mutants. Further, all constructs were packaged into rAAV2/1 and injected into newborn mice as described previously [27,38,39]. rAAV-EGFP was used as a control. Each viral construct was delivered into two litters and brains were extracted at 6 months post injection (4-6 mice per group). One hemibrain was frozen for biochemical analysis and the other was xed and para n embedded for immunohistochemistry. We stained the brain sections with a pan-Aβ antibody.
The data shown in Fig. 2A and Table 1 demonstrates robust amyloid deposition in mice injected with rAAV-BRI2-Aβ42 WT, Aβ42 E22Q, Aβ42 E22Q/D23N, Aβ42 ΔE22, Aβ42 S26C as well as Aβ42 S8A and to a small extent Aβ42 S8E, 6 months after injection. As shown in Fig. 2B, despite extensive variability, it is clear that mutants Aβ42 E22G and Aβ42 E22Q/D23N as well as Aβ42 WT were detected in both SDSsoluble and the SDS-insoluble, FA-soluble fractions, suggesting increased prevalence of compact, "cored" plaques, whereas Aβ42 ΔE22, Aβ42 S8A, Aβ42 S8E, and Aβ42 S26C deposits were more SDS-soluble, corresponding to more diffuse plaques. Interestingly, overexpression of the phosphomimetic Aβ42 S8E resulted in sparse deposits, with very low levels of both SDS-soluble and FA-soluble Aβ42, whereas nonphospho-mimetic Aβ42 S8A exhibited increased deposits, suggesting that phosphorylation does not play a signi cant role in Aβ42 deposition. We have previously shown that overexpression of WT Aβ40 resulted in no amyloid pathology [30]. However, when a series of BRI2-Aβ40 mutants were overexpressed in the neonatal brain, we observed that Aβ40 E22G and E22Q/D23N aggregated and accumulated in the brain (Fig. 3A). Interestingly, amyloid deposits of the Aβ40 E22Q/D23N double mutant were detected in both the SDS fraction and FA fraction, whereas Aβ40 E22G deposits were almost entirely SDS-soluble, suggesting a more diffuse type of amyloid aggregate (Fig. 3B). Aβ40 WT accumulation and aggregation was not detected and neither was ΔE22, S8A, S8E, nor S26C ( Fig. 3A and B). Notably, accumulation of Aβ40 E22G and E22Q/D23N is the rst evidence that Aβ40 overexpression leads to deposition under certain conditions.

Neurotoxic assessment of mutant Aβ peptides in Drosophila eye
The Drosophila eye provides an unparalleled and reliable platform to study the contributions of neurotoxic amyloids in vivo. Its photoreceptor neurons are grouped within 800 ommatidia that form an external symmetrical array of hexagonal structures, which are particularly sensitive to Aβ42 insults [40].
Thus, we used this paradigm to compare the neurotoxic properties of wild-type and mutant Aβ peptides upon speci c expression in the eye with the GMR-Gal4 driver. We found that ies expressing one copy of the Aβ40 WT or Aβ40 mutants (Aβ40 E22G, S26C, and E22Q) displayed highly organized ommatidia with even distribution of bristles, consistent with the normal phenotype observed in control ies expressing LacZ (compare insets in Fig. 4). In contrast, and consistent with our previous observations [30], a single copy of the Aβ42 WT induced a more aggressive phenotype characterized by disorganization, ommatidial fusion and partial lack of bristles (Fig. 4). Importantly, ies expressing Aβ42 mutants (Aβ42 E22G, S26C, and E22Q) exhibited more severe and extensive disorganization with ommatidial perforations and reduction of the eye size (Fig. 4).
Next, to quantify the neurotoxic contribution of the different Aβ transgenes, we assessed locomotor performance in adult ies as an alternative functional assay for Aβ42-mediated toxicity. For this, we panneurally expressed a GFP-attP2 control transgene or the different Aβ constructs under control of the ELAV-Gal4 driver and evaluated the ability of female ies to climb vertically from day 1 of age. After a blinded assessment of all the genotypes, we calculated the mean percentage of ies that climbed above 5 cm per day (climbing index).
An initial exploratory climbing assay comparing Aβ40 WT and GFP control showed no statistical differences in phenotype between both genotypes (Fig. S2A). This data con rms the eye phenotype results observed previously in Fig. 4.
Then, we determined if the Aβ42 E22G, S26C, and E22Q mutations affected the behavioral output of the ies compared to Aβ42 WT. As previously reported [41], the climbing ability of ies expressing Aβ42 WT was signi cantly impaired compared to control ies expressing only GFP (p < 0.0001). For instance, locomotor activity of the Aβ42 WT ies decreased rapidly, reaching 50% climbing ability by day 7, stopping altogether by day 16 (Fig. S2B). Interestingly, Aβ42 S26C mutant line reached 50% climbing at day 4 suggesting a more deleterious phenotype (p < 0.001), however as age increased, climbing pro ciency declined to a similar tread as Aβ42 WT ies reaching a total halt by day 16 (p < 0.05). In contrast, ies carrying the mutations E22G or E22Q showed high toxicity during development, leading to substantial pupal lethality. Luckily, enough escapers for both genotypes were recovered to create 3 replicates although their locomotor function was so greatly impaired that only 5% of ies climbed on the rst day of testing, stopping, altogether, by day 2 (Fig. S2B).
Aβ40 suppresses the progressive dysfunction induced by Aβ42 Next, to examine the potential neuroprotective effect of Aβ40 over Aβ42 toxicity, we generated ies expressing one copy of the transgene encoding each Aβ40 peptide (Aβ40 WT, E22G, S26C, and E22Q) over an Aβ42 WT background. Eye phenotypes in the Aβ42 background co-expressed with Aβ40 showed a slight betterment on phenotype, manifested as a decrease in the number of fused ommatidia and loss of bristles (Fig. 5A). Moreover, we detected mild changes regarding ommatidia disorganization in the anterior part of the eye compared to the control sample (Aβ42; LacZ) (Fig. 5A).
In addition, we investigated the effect of co-expressing Aβ42 WT ies with a second copy of Aβ42 WT or Aβ42 mutant lines. Surprisingly, only Aβ42/Aβ42 E22Q ies showed a markedly exacerbated phenotype characterized by a signi cant increment in the presence of ommatidial perforations and necrotic spots in the eye; while co-expression of Aβ42 WT, S26C, and E22G showed an elliptical shaped eye with ommatidial disorganization, perforation and fusion characteristic of Aβ42/Aβ42 control (Fig. 5A).
Pan-neural (elav-Gal4) co-expression of human Aβ42 peptide with potential modi ers could induce an age-related improvement in climbing pro ciency. We showed before (Fig. 5A) that expression of Aβ40 on an Aβ42 background led to a mild improvement on the structure of the adult eye of the y. Based on this, we decided to analyze the possible physiological bene ts of co-expressing Aβ42 peptides with Aβ40 WT and mutants in motor neurons on geotaxis (Fig. 5B).
Similar to the data shown in Fig. S2B, control females expressing GFP alone reached the 50% climbing index by day 13 and continued to move for 15 more days. In contrast, ies co-expressing Aβ42 and GFP displayed severe locomotor dysfunction, reaching 50% climbing index by day 7 coming to a total halt in climbing by day 22 (Fig. 5B). As initially observed in the eye phenotypes, ies co-expressing Aβ42 and Aβ40 WT performed slightly better than control ies expressing Aβ42/GFP. Unexpectedly, ies coexpressing Aβ42 and Aβ40 E22Q performed as well as GFP control ies reaching 50% climbing by day 13, however the rate of climbing performance quickly deteriorated by day 15 followed by a slow decline and total stop by day 25. Aβ40 co-expression appears to block the deleterious effect of Aβ42 early in the adulthood but is unable to maintain it as age progresses leading to de ciency on climbing.
Then, to better understand the effects of these combinations we assessed the climbing performance of the different Aβ42 with Aβ40 combinations every 5 days (Fig. 5C and Table 2). Strikingly, while no statistical differences in phenotype manifested in younger ies, from day 10 to day 20, ies expressing Aβ40 mutations (S26C, E22G, and E22Q) performed signi cantly better that ies co-expressing Aβ42 with GFP. By day 25, however, only the ies co-expressing Aβ40 S26C and E22Q performed better than the control Aβ42/GFP, while Aβ42/Aβ40 WT and Aβ42/Aβ40 E22G phenotypes were similar to those coexpressing Aβ42/GFP. This pattern was maintained until day 30 when ies stopped climbing. All P values are the result of comparing the climbing index (%) of each Aβ42 WT/Aβ40 WT, Aβ42 WT/Aβ40 E22G, Aβ42 WT/Aβ40 S26C, and Aβ42 WT/Aβ40 E22Q genotype with respect to Aβ42 WT/GFP.
As expected, a similar experiment with ies co-expressing Aβ42 with an extra copy of Aβ42 WT and mutants (S26C, E22Q, and E22G) showed a signi cant increase in toxicity when compared to control ies carrying a single copy of Aβ42 (Aβ42/GFP) (Fig. S3). Interestingly, Aβ42/Aβ42 S26C performed signi cantly better between days 5 and day 10 (p < 0.0001) than other combinations suggesting a milder decrease in toxicity for carriers of mutation S26C. Of note, Aβ42 co-expression with Aβ42 E22G resulted in high pupal lethality and this group had to be removed from the experiment (Fig. S3).

Discussion
Although the majority of AD cases are sporadic, there are several FAD mutations with amino acid substitutions in APP that alter Aβ aggregation rates and result in accelerated disease progression with other pathologies, such as CAA. These mutations have been exploited in mouse models as tools to study Aβ aggregation and to inform therapeutic development. Despite recent reports showing that Aβ can behave like a prion-like strain and FAD mutations inducing unique phenotypes [42], the link between Aβ aggregation and neurodegeneration is unclear. In this study we overexpressed Aβ mutant peptides in the absence of APP in the brains of neonatal mice and showed predisposition to aggregation of Aβ42 WT and mutants. Further, we examined the effect of exclusively expressed Aβ mutant peptides on eye structure and behavior in Drosophila. Our results suggest that while Aβ42 species that lead to robust aggregation in mice, result in higher levels of toxicity in ies, mutations in Aβ40, E22G and E22Q/D23N, formed amyloid deposits despite being protective in the y model. Thus, the choice of model system used to study Aβ aggregation and neurodegeneration is crucial for understanding the link between them.
We employed the BRI2 system to selectively express the Aβ mutations without APP. The BRI2 system utilizes fusion constructs in which the sequence encoding the 23-amino-acid ABri peptide at the carboxyl terminus of the transmembrane protein BRI is replaced with a sequence encoding Aβ [25]. Constitutive processing of the resultant BRI2-Aβ fusion proteins in transfected cells resulted in high-level expression and secretion of the encoded Aβ peptide. AAV2/1 vectors encoding BRI2-Aβ cDNAs, were previously used to achieve high-level hippocampal expression and secretion of the speci c encoded Aβ peptide in the absence of APP overexpression [26].
Differential levels of expression may be due to a variability of transfection e ciency as well as by variability in furin cleavage e ciency. Thus, if a mutation in the Aβ sequence causes aggregation of the fusion protein, it may make the furin cleavage and, as a result, Aβ secretion, less e cient. Also, since SDS-PAGE is performed in reducing conditions, some of the low molecular weight soluble Aβ aggregates may be seen as a monomer on Western blot but might not be detected by ELISA.
Despite these limitations, when overexpressed in the mouse brain, Aβ mutants aggregate and present as unique phenotypes. Thus, Aβ42 WT, E22G, and E22Q/D23N resulted in more profound SDS insoluble, FAsoluble aggregates, corresponding to compact plaques while overexpression of Aβ42 ΔE22, S8A, S8E, and S26C, resulted in mostly SDS-soluble material, corresponding to diffuse plaques. Indeed, the E22G mutation favored fast Aβ brillization and aggregation [43,44]. Our ndings suggest that the E22G and E22Q/D23N mutations affected the aggregation kinetics most profoundly. These mutations accelerated the overall aggregation by the modulation of the nucleation processes, whereas the elongation process is not signi cantly affected [45]. This shift in kinetics resulted in amyloid deposition, even with Aβ40, while Aβ40 WT and other Aβ40 mutants did not result in deposition [42]. It is important to note that both familial AD, and CAA-related mutations, such as E22Q, E22G/D23N, ΔE22, as well as rationally designed mutations S8A, S8E, and S26C, led to amyloid deposition when overexpressed in the mouse brain.
Another interesting aspect of the FAD mutations within the Aβ sequence is that they lead to remarkable phenotypic diversity in the abundance of CAA [46,47]. In our study, we did not detect CAA following unique BRI2-Aβ overexpression, suggesting that vascular deposits require a diverse mix of Aβ species.
It has been extensively reported that healthy patients present greater deposition of Aβ40, while most familial and sporadic AD cases have increased Aβ42 deposition or an augmented Aβ42:Aβ40 ratio. This is believed to be due to Aβ42 rapidly forming more stable aggregates than Aβ40 in unhealthy brains [48][49][50][51]. In addition, pathogenic mutations within the sequence are described to signi cantly increase oligomerization. For instance, both E22Q and E22G aggregate to form proto brils and brils more rapidly than WT Aβ42 [42]. Our Drosophila results support these ndings. As shown in Fig. 4, Aβ40 WT ies presented highly organized ommatidia with even distribution of bristles similar to the healthy control group. In contrast, a single copy of the Aβ42 WT or mutant peptides (Aβ42 E22G, S26C, and E22Q) induced a classic rough eye phenotype characterized by disorganized ommatidial assembly, ommatidial fusions, and loss of interommatidial bristles consistent with a more toxic effect of Aβ42 peptides. These data were then supported by our negative geotaxis assays that showed a signi cant detriment in climbing ability in those ies expressing Aβ42 peptides; while Aβ40-associated phenotypes were indistinguishable from control GFP recapitulating some of the ndings by Jonson et al. in 2015 (Suppl. Figure 2) [52]. Moreover, our Aβ42 E22G and E22Q lines showed signi cant toxicity during development, leading to substantial pupal lethality. This coincides with previous studies where Aβ42 E22G led to signi cantly high rates of lethality in development as well as detriment in climbing capacity compared to Aβ42 WT ies [53]. One explanation resides in the hypothesis that mutations at position 22, including E22Q and E22G, increase neurotoxicity in Aβ42 by stabilizing a C-terminal core that accelerates aggregation [54].
The results of our study illustrate the differences between Aβ40 and Aβ42, supporting other studies that demonstrated the signi cance between the structural differences among both peptides [55,56]. Additionally, it has been suggested that small changes in the Aβ42:Aβ40 ratio affect aggregation kinetics, the morphology of the resulting amyloid brils and synaptic function both in vitro and in vivo [57]. However, our results also demonstrate that Aβ40 could potentially protect against Aβ42 aggregation.
Flies co-expressing Aβ40 on an Aβ42 background, had a moderately improved ommatidial organization, suggesting a protective effect of Aβ40 peptides against Aβ42. However, although we observed a slight recovery on the overall climbing performance of the Aβ40/Aβ42 ies compared to Aβ42/GFP control group, addition of Aβ40 did not reach statistically signi cant difference in climbing ability. Interestingly, and contrary to what we expected, the introduction of the known mutations S26C, E22G, and E22Q in the Aβ40 peptide did not signi cantly further improve the Aβ40 protective effect over Aβ42 observed in the Drosophila eye. However, these mutations signi cantly improved the climbing ability of Aβ42-expressing ies, reaching 50% climbing performance signi cantly later when compared to the Aβ42/GFP control group, a signi cance that was observed through later age points (day 10, 15, and 20). These results were striking. Multiple reports suggest that introduction of mutations in Aβ40 peptides induce an amyloidogenic phenotype similar to Aβ42 [54,[58][59][60]. More recently, Yoo et al showed that the heterozygous E22G pathogenic mutation of Aβ40 enhance misfolding of Aβ via cross-seeding from Aβ42 WT bril that could potentially be responsible for early onset AD phenotypes [61].
While the exact sequence of events that causes AD remain to be identi ed, aggregation of the Aβ peptide is a critical step in this process. We believe that, although most cases involve WT Aβ, signi cant insights on the steps of aggregation can be gained by studying the effects of point mutations implicated in earlyonset AD.

Limitations
First, FAD represents a very small fraction of overall AD cases, thus, studying the formation of amyloid pathology on Aβ mutants does not represent sporadic AD. The lack of direct correlation between amyloid accumulation, Aβ-induced toxicity and neurodegeneration is a limitation to a single model study, emphasizing the importance of using two or more Aβ overexpression models, such as mice and fruit ies.
Second, the somatic brain transgenics AAV delivery technology, utilized in our research, results in substantial variability of amyloid deposition. Future studies focusing on adult injections of AAV-BRI-Aβ under neuronal promoter are advisable. Third, despite the elevated number of individuals used, Drosophila negative geotaxis assay showed elevated variability within the replicates. This variability caused that some of the trends lacked signi cant power. A more sensitive approach that accounts for ner changes in drosophila motor performance could provide more accurate and informative data. Fourth, although we expect levels of expression to remain identical between the various mutant lines, we cannot rule out the possibility of different degrees in protein accumulation and toxicity that could be responsible for some of the differences observed between samples. Recapitulation of the same results in future experiments will strengthen the data.

Conclusions
In summary, by using complimentary approaches of expressing mutations without APP and in both mice and Drosophila, we demonstrated that although some Aβ40 mutants show amyloidogenic properties in the mouse brain, they can be protective against Aβ42-induced toxicity in y eye phenotypes and behavioral assays. Since Drosophila are a better model of amyloid-induced toxicity than mice, these results emphasize the importance of utilizing multiple models for screening therapeutic agents.   Brain expression of Aβ42 WT and mutants results in unique amyloid deposition. Newborn B6C3F1 pups were bilaterally injected ICV with 4 µl rAAV1-BRI-Aβ42 (1013 vg/ml) (WT, E22G, E22Q/D23N, E22, S8A, S8E, or S26C). After 6 months mice were euthanized, and brains were extracted and processed. (A) Representative brain sections were stained with pan-Aβ antibody and counterstained with hematoxylin.

Abbreviations
Scale Bar, 60μm, 200μm, 500μm, n=4-10. (B). Second hemibrain was sequentially extracted in 2% SDS followed by 70% FA and Aβ levels were quanti ed using Aβ sandwich ELISA with C-terminal speci c mAb as capture and pan-Aβ mAb as detection. Each dot represents an individual mouse brain, n=4-10. E22G and E22Q/D23N Aβ40 mutants deposit in the mouse brain. P0 newborn pups were injected with rAAV1-BRI2-Aβ40 WT or following mutants, E22G, E22Q/D23N, E22, S8A, S8E, or S26C. Mice were aged 6 months and brains were extracted and processed. (A) Representative brain sections were stained with anti-pan-Aβ antibody and counterstained with hematoxylin. Scale Bar, 60μm, 200μm, 500μm, n=4-10. ( ) Second hemibrain was sequentially extracted in 2% SDS followed by 70% FA and each fraction was subjected to Aβ sandwich ELISA with C-terminal speci c mAb as capture and pan-Aβ mAb as detection to quantify Aβ42 levels. Each dot represents an individual mouse brain, n=4-10. Phenotypes produced by various Aβ peptides in the Drosophila eye. Panels show SEM images from y eyes with the indicated genotypes. Control ies expressing LacZ alone show highly organized eyes with hexagonal lenses. Expression of extracellular Aβ40 WT, E22G, S26C, and E22Q mutants showed slightly more disorganized ommatidia with no change in size or structure. Expression of extracellular Aβ42 WT results in small eyes with severe ommatidial disorganization and fusion. Aβ42 E22G, S26C, and E22Q mutants showed higher disorganization with presence of fusion in ommatidia and sporadic necrotic points.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.