Tackling amyloidogenesis in Alzheimer’s disease with A2V variants of Amyloid-β

We developed a novel therapeutic strategy for Alzheimer’s disease (AD) exploiting the properties of a natural variant of Amyloid-β (Aβ) carrying the A2V substitution, which protects heterozygous carriers from AD by its ability to interact with wild-type Aβ, hindering conformational changes and assembly thereof. As prototypic compound we designed a six-mer mutated peptide (Aβ1-6A2V), linked to the HIV-related TAT protein, which is widely used for brain delivery and cell membrane penetration of drugs. The resulting molecule [Aβ1-6A2VTAT(D)] revealed strong anti-amyloidogenic effects in vitro and protected human neuroblastoma cells from Aβ toxicity. Preclinical studies in AD mouse models showed that short-term treatment with Aβ1-6A2VTAT(D) inhibits Aβ aggregation and cerebral amyloid deposition, but a long treatment schedule unexpectedly increases amyloid burden, although preventing cognitive deterioration. Our data support the view that the AβA2V-based strategy can be successfully used for the development of treatments for AD, as suggested by the natural protection against the disease in human A2V heterozygous carriers. The undesirable outcome of the prolonged treatment with Aβ1-6A2VTAT(D) was likely due to the TAT intrinsic attitude to increase Aβ production, avidly bind amyloid and boost its seeding activity, warning against the use of the TAT carrier in the design of AD therapeutics.

. Analysis of 1.5 μs explicit solvent MD simulations of the Aβ1-6 WT and Aβ1-6 A2V peptides. (A) Apolar character of the peptide SAS represented as the ratio between SAS apolar and the total SAS. (B) Gyration radius distribution. (C) Analysis of secondary structure propensity. "Structure" indicates residues possessing a defined secondary structure, in this case structure indicates residues in a "turn" configuration. "Coil" indicates residues that do not display a defined secondary structure. Analysis of the secondary structure was carried out with DSSP. (D) Typical compact "turn" and elongated "coil" configurations reported for the Aβ 1-6 A2V and Aβ 1-6 WT , respectively. (E) Analysis of the most populated structural clusters. Representative structures of the six most probable clusters were reported. The coil configuration has been highlighted in green, the turn in red and a partly folded turn in orange.   WT . In polarizedlight and EM studies, both peptides were used at 0.125 mM, molar ratio = 1:1 or 1:4 respectively, with 20 days incubation. From 5-20 days, 1:1 co-incubation of the two peptides (B,D) displayed a lower amyloid fibril content respect to Aβ 1-42 WT alone (A,C), showing protofibrils, short fibrils and disaggregated granular and C. elegans models indicated that this mutation shifts APP processing towards the amyloidogenic pathway with increased production of amyloidogenic peptides. Furthermore, the A2V substitution in the Aβ sequence (Aβ A2V ) increases the propensity of the full-length peptides (i.e., Aβ 1-40 and Aβ 1-42) to adopt a β -sheet structure, boosts the formation of oligomers both in vitro and in vivo and enhances their neurotoxicity [8][9][10] . Following the observation that humans carrying the mutation in the heterozygous state do not develop AD, we carried out in vitro studies with synthetic peptides that revealed the extraordinary ability of Aβ A2V to interact with wild-type Aβ (Aβ WT ), interfering with its nucleation or nucleation-dependent polymerization 7 . This provides grounds for developing a disease-modifying therapy for AD based on modified Aβ A2V peptides retaining the key functional properties of parental full-length Aβ A2V .
These results prompted us to develop a prototypic compound by linking Aβ 1-6 A2V (D) to an all-D form of TAT sequence [TAT(D)], a peptide derived from HIV that powerfully increases virus transmission to neighbour cells 11 , and is widely used for brain delivery of drugs [12][13][14] . Here we report that this compound [Aβ 1-6 A2V TAT(D)] has strong anti-amyloidogenic effects in vitro, leading to inhibition of oligomer, amyloid fibril formation and of Aβ -dependent neurotoxicity. Preclinical studies showed that a short-term treatment with this peptide in an AD mouse model prevents Aβ aggregation and amyloid deposition in the brain but longer treatment unexpectedly increases amyloid burden, most likely due to the TAT intrinsic attitude to enhance Aβ production and to avidly bind amyloid and boost its seeding activity, warning against the use of this carrier in therapeutic approaches for AD.

Results
In silico molecular modeling of Aβ A2V peptide variants. To predict the structural basis of the anti-amyloidogenic effect of Aβ 1-6 A2V (D), a comparative conformation analysis of WT and mutated Aβ 1-6 was carried out with all-atom classical MD simulations in explicit solvent. Both Aβ 1-6 WT and Aβ 1-6 A2V are intrinsically disordered peptides characterized by high flexibility. Nevertheless, the substitution of Ala2 with a Val residue induces significant changes in the appearance of the peptide in solution, resulting in an increase of the apolar character of the solvent accessible surface (SAS) (Fig. 1A) and in a modification of the gyration radius distribution in the Aβ 1-6 A2V . Figure 1B shows that the probability distribution of the gyration radius is characterized by a global shift to smaller values and by the appearance of a shoulder in the distribution corresponding to gyration radii of 0.5 nm.
An analysis of the secondary structure content displayed by the peptides (Fig. 1C) shows that, while both Aβ 1-6 A2V and Aβ 1-6 WT display a predominant coil configuration, Aβ 1-6 A2V shows a slightly higher propensity to form secondary structure motifs involving two to three residues. Aβ 1-6 A2V in fact displays a propensity to form a turn involving the Glu3, Phe4 and Arg5 residues (Fig. 1D). The most populated structural clusters 15 (Fig. 1E), in Aβ 1-6 WT are characterized by an elongated coil structure accounting for 52.6% of the configurations, while the compact "turn" state is only the third most probable cluster, with a population of around 9%. Conversely, in the Aβ 1-6 A2V , while the most populated structure is still an elongated coil (32%), the "turn" configuration is the second most populated structural cluster (31%).
Both Aβ 1-6 WT and Aβ 1-6 A2V under physiological conditions are characterized by intramolecular salt bridges such as those between Asp1 and Arg5 or Glu3-Arg5. In the extended coil configuration (Fig. 1E), salt bridges can be dynamically formed and dissociated without requiring a specific rearrangement of the peptide backbone. However, in the turn configuration salt bridges are typically dissociated; the interaction of the apolar Val2 sidechain with the Arg5 sidechain stabilizes such a dissociated state. The additional sterical hindrance to the rearrangement induced by the Val2 sidechain also contributes to the stabilization of the turn configuration of the A2V peptide.
The propensity of the A2V mutant to adopt a Glu3-Arg5 turn configuration characterized by a significant lifetime can be interpreted as the probable source of the heterotypic interaction of the Aβ 1-6 A2V with full-length Aβ , which results in hindering its assembly. Aβ1-6 A2V retains the in vitro anti-amyloidogenic features of the parental full-length peptide. We previously showed that Aβ 1-6 A2V (D) destabilizes the secondary structure of Aβ 1-42 WT  Treatment of SH-SY5Y cells with Aβ 1-6 WT (D) or Aβ 1-6 A2V (D) showed that neither is toxic for living cells even at high concentrations (20 μ M) ( Fig. 2A,B) and that both peptides are able to reduce the toxicity induced by Aβ 1-42 WT (Fig. 2C,D). However, Aβ 1-6 A2V (D) showed a stronger effect in counteracting the reduction of cell viability caused by Aβ 1-42 WT (Fig. 2D), suggesting that the A-to-V substitution actually amplifies the protective effects of the six-mer peptide.
material. E,F: Representative Tapping mode of AFM images as determined by amplitude error data of Aβ 1-42 WT oligomers. Aβ 1-42 WT peptide 100 μ M in phosphate buffer 50 mM, pH 7.4 was incubated at 4 °C for 24 h alone (E) (Z range: -10/ + 10 mV) or in presence of Aβ 1-6 A2V TAT(D) (F) (Z range: -10/ + 25 mV). The molar ratio of Aβ 1-42 WT to Aβ 1-6 A2V TAT(D) was 1:4. Scale bar: 1 μ m, inset: 200 nm. (G,H): height plot profiles obtained along different lines traced on the topographic AFM images. Overall, these effects were already evident in the 1:1 mixture of the two peptides (data not shown), suggesting that the inhibition of Aβ 1-42 WT aggregation by Aβ 1-6 A2V TAT(D) is a dose-dependent effect. Aβ1-6 A2V TAT(D) maintains the in vitro anti-amyloidogenic properties of Aβ1-6 A2V (D). Aβ 1-6 A2V (D) alone does not efficiently cross either the blood brain barrier (BBB) or cell membranes (data not shown). This is an important feature that would deeply limit its use as an in vivo anti-amyloidogenic drug. So, we linked this peptide to the all-D TAT sequence to improve the translocation of Aβ 1-6 A2V (D) across the BBB and cell membranes, minimize the degradation of the peptide and reduce the immune response elicited by the molecule. The resulting compound [Aβ 1-6 A2V TAT(D)] destabilizes the secondary structure of Aβ 1-42 WT . Indeed, CD spectroscopy studies showed that Aβ 1-6 A2V TAT(D) inhibits the acquisition of β -sheet conformation by Aβ 1-42 WT (data not shown), thus affecting the folding of the full-length peptide.
We tested the ability of Aβ 1-6 A2V TAT(D) to inhibit the fibrillogenic properties of the full-length Aβ in vitro and found that the compound hindered Aβ 1-42 WT aggregation (Fig. 3). Polarized light and electron microscopy studies on aggregates of Aβ 1-42 WT formed after 20 days incubation with or without Aβ 1-6 A2V TAT(D) revealed that the mutated peptide hinders the formation of amyloid structures (Fig. 3B) and reduces the amount of fibrils generated by the full-length peptide (Fig. 3D). Moreover, AFM analysis (Fig. 3E,H) showed that Aβ 1-6 A2V TAT(D) actually interferes with the oligomerization process of Aβ 1-42 WT . Indeed, monomeric Aβ 1-42 WT , incubated alone at a final concentration of 100 μ M, formed a family of small oligomers of different size within a range of 6-20 nm in diameter (~ 70%) (Fig. 3E,G). Conversely, the co-incubation with Aβ 1-6 A2V TAT(D) resulted in the formation of very small globular structures with a range of 5-8 nm in diameter and height of 200-400 pm (~ 70%), large and thin structures, apparently very rich in water (width: 500-700 nm; height: 200-500 pm). Notably, only rare oligomeric structures were detected (Fig. 3F,H).
These effects were observed by incubating Aβ 1-42 WT and Aβ 1-6 A2V TAT(D) at a 1:4 molar ratio, but they were also evident at equimolar concentrations of the two peptides.
All these findings indicated that the designed Aβ 1-6 A2V TAT(D) peptide is particularly efficient at inhibiting Aβ polymerization and toxicity in vitro, and identified it as our lead compound for the subsequent in vivo studies.

Aβ1-6 A2V TAT(D) is able to cross the BBB after its intraperitoneal administration. The stabil-
ity of Aβ 1-6 A2V TAT(D) and Aβ 1-6 A2V TAT(L) peptides to serum proteases was determined using MALDI-TOF. The D-peptide remained stable for 48 h, while the L-peptide was rapidly degraded in about 1 h (Supplementary Fig. 3 S). To check the efficacy of TAT peptide in delivering Aβ 1-6 A2V to the brain, we measured the levels of Aβ 1-6 A2V TAT(D) in brain tissue of mice treated with the compound i.p. once a week at the dose of 10 mg/kg for five months, 1 or 24 h after the last treatment. The analysis showed that Aβ 1-6 A2V TAT(D) brain levels were 145.03 ± 38.26 pg/mg tissue (mean ± SEM) and 62.02 ± 28.68 after 1 or 24 h, respectively (Supplementary Table  3S  The treatment started at the age of 4 months, when the deposition of amyloid in the brain of these animals usually begins, becoming consistent by 6 months. The schedule of treatment was established based on the results of a pilot study with the same mouse model, showing that the best anti-amyloidogenic effects are obtained by treating animals i.p. once a week with 10 mg/kg peptide (data not shown).
The short-term treatment (2.5 months) of APPswe/PS1dE9 mice with Aβ 1-6 A2V TAT(D) resulted in the exciting reduction of Aβ production and aggregation (Fig. 5B), and prevention of amyloid deposition in the brain (Fig. 5C,D) of the group treated once a week with Aβ 1-6 A2V TAT(D). This group showed an increase of Aβ 1-42 in the soluble fraction and a concomitant decrease in the insoluble fraction, suggesting that the treatment induces a transfer of Aβ 1-42 from the insoluble to soluble compartment of brain tissue, which may reflect a disaggregation of amyloid deposits. This view is corroborated by the fact that aggregated Aβ was reduced in both compartments. Moreover, the decrease of aggregated Aβ in the soluble fraction is associated with an increase of monomeric Aβ 1-42 levels, suggesting that A2V-based therapy changes the dynamic equilibrium between Aβ aggregates and Aβ monomers, resulting in a reduction of the most toxic Aβ species (i.e., soluble Aβ assemblies).
The reduction of amyloid deposition involved all brain areas, including the hippocampus, and was more prominent in the frontal and entorhinal cortices and olfactory bulbs (Fig. 5D).
These findings were paralleled by a slight effect on cognitive function (Fig. 5A), as suggested by performances of mice treated with Aβ 1-6 A2V TAT(D) in the Novel Object Recognition test (NOR). The results of the study showed a trend towards cognitive impairment in the control group and an opposite tendency of mice treated with Aβ 1-6 A2V TAT(D), whose cognitive performances were no worse than at the beginning of the treatment with Aβ 1-6 A2V TAT(D). In fact, the NOR test scores showed a small, albeit not statistically significant, improvement. However, the final outcome of treatment (after 5 months) was an unexpected substantial increase in amyloid burden (Fig. 6C) and attenuation of the effects on Aβ production (Fig. 6B), while the results of the behavioral assessment by NOR suggested preserved cognitive function in the group treated with Aβ 1-6 A2V TAT(D) compared with the saline-treated mice, as indicated by significant differences in the scores for recognition and discrimination indexes obtained for the two experimental group of mice at the end of treatment (Fig. 6A).
Aβ1-6 A2V TAT(D) induces a shift in APP processing toward the amyloidogenic pathway in a chronic treatment. The unexpected outcome of the prolonged treatment with Aβ 1-6 A2V TAT(D) spurred us to search for the causes underlying the removal of the anti-amyloidogenic effects previously observed after 2.5 months of treatment with the test compound.
To this end, we investigated the effects of Aβ 1-6 A2V TAT(D) on APP processing and found that treatment with Aβ 1-6 A2V TAT(D) resulted in a shift of APP processing towards the amyloidogenic pathway, leading to an increased C99:C83 ratio in the brain of APPswe/PS1dE9 mice (Fig. 7A-D), which presumably paves the way for overproduction of Aβ . Interestingly, the effects on APP processing were detected only after prolonged treatment with Aβ 1-6 A2V TAT(D) and were not observed after short-term treatment schedules, i.e., after 3 weeks in the huA-PP Swe /moAPP 0/0 or after 2.5 months in the APPswe/PS1dE9 mice.
TAT peptide alone binds amyloid vigorously. Concomitant studies performed in our labs showed that Aβ 1-6 A2V TAT(D) has a special propensity to target β -amyloid deposits (Fig. 7E-H) and that this is a specific attribute of the TAT(D) sequence.
These data suggest that the anti-amyloidogenic effects of Aβ 1-6 A2V (D) are overcome in the chronic treatment by the TAT intrinsic attitude to promote amyloidogenic APP processing, bind to amyloid and boost its seeding activity, leading to an increase of amyloid burden in vivo. Aβ1-6 A2V TAT(D) induces an immune response after peripheral administration in mice. Finally, we tested the Aβ 1-6 A2V TAT(D) immunogenic properties and found that the chronic treatment with this compound results in the production of not negligible levels of serum IgG against Aβ 1-6 A2V TAT(D) as well as against full-length Aβ (Fig. 8A,B). Interestingly, the treatment with Aβ 1-6 A2V TAT(D) did not induce production of IgG against Aβ 1-6 A2V peptide (Fig. 8C), while 3 out of 9 mice displayed slightly increased levels of IgG against TAT(D) (Fig. 8D). These data suggest that both Aβ 1-6 A2V and TAT(D) are not per se strongly immunogenic, while the combination of the two peptides might confer strong immunogenicity to Aβ 1-6 A2V TAT(D) compound that, however, did not result in any evident side effect on brain, as revealed by neuropathologic studies (data not shown).

Discussion
During the last few decades, huge efforts have been made to develop disease-modifying therapies for Alzheimer, but the results of these attempts have been frustrating. The anticipated increase of AD patients in the next few decades makes the development of efficient treatments an urgent issue 16 . In order to prevent the disease and radically change its irreversible course, a long series of experimental strategies against the main molecular actors of the disease (Aβ and tau) 17 or novel therapeutic targets 18 have been designed based on purely theoretical grounds 19 as well as on evidence mainly deriving from preclinical observations in AD animal models 20 . However, few strategies proved suitable for application in human clinical trials, and none proved to be really effective 21 .Our approach differs from previous strategies -mainly those involving modified Aβ peptides that have been found to inhibit amyloidogenesis 19,22 -since it is based on a natural genetic variant of amyloid-β (Aβ A2V ) that occurs in humans and prevents the development of the disease when present in the heterozygous state 7 .
In this context, we carried out in vitro and in vivo studies that revealed the extraordinary ability of Aβ A2V to interact with Aβ WT , interfering with its aggregation 8 . These findings were a proof of concept of the validity of therapeutic strategies based on the use of Aβ A2V variant, and prompted us to develop a new disease-modifying treatment for AD by designing a six-mer mutated D-isomer peptide [Aβ 1-6 A2V (D)] linked to the short amino acid sequence derived from the HIV TAT peptide, widely used for brain delivery, to make the translocation of Aβ 1-6 A2V (D) across the BBB feasible.
The use of TAT as a carrier for brain delivery of drugs has been employed in several experimental approaches for the treatment of AD-like pathology in mouse models 12,13 . Recently, intraperitoneal administration of a TAT-BDNF peptide complex for 1 month was shown to improve the cognitive functions in AD rodent models 23 .
with Aβ 1-6 A2V TAT(D) vs mice treated with saline solution (control group). Immunohistochemistry with 4G8 antibody, original magnification 10× (C). Quantification of amyloid burden by 'plaque count' and 'area fraction' in the two experimental groups: saline treated animals (blue columns) and Aβ 1-6 A2V TAT(D)-treated mice (red columns) (D). Evidence of reduction of amyloid burden in the group treated with Aβ 1-6 A2V TAT(D). The significance of the results was calculated via a Mann-Whitney U test. Statistical differences were considered significant if p < 0.05.
Scientific RepoRts | 6:20949 | DOI: 10.1038/srep20949 A previous study showed that, following its peripheral injection, a fluorescein-labelled version of TAT is able to cross the BBB, bind amyloid plaques and activate microglia in the cerebral cortex of APPswe/PS1DE9 transgenic mice 24 . TAT was then conjugated with a peptide inhibitor (RI-OR2, Ac-rGffvlkGr-NH2) consisting of a retro-inverted version of Aβ 16-20 sequence 25 that was found to block the formation of Aβ aggregates in vitro and to inhibit the toxicity of Aβ on cultured cells 25 . Daily i.p. injection of RI-OR2-TAT for 21 days into 10-month-old APPswe/PS1DE9 mice resulted in a reduction in Aβ oligomer levels and amyloid-β burden in cerebral cortex 24 .
We followed a similar strategy and initially demonstrated that Aβ 1-6 A2V (D), with or without the TAT sequence, retains in vitro the anti-amyloidogenic properties of the parental full-length mutated Aβ , since it is effective at hindering in vitro the production of oligomers and fibrils, the formation of amyloid and the toxicity induced by Aβ 1-42 WT peptide on SYSH-5Y cells.
Based on these results, we then decided to test in vivo the anti-amyloidogenic ability of Aβ 1-6 A2V TAT(D). The compound proved stable in serum after i.p. administration in mice, able to cross the BBB and associated with an immune response that was not found to cause any brain damage.
Short-term treatment with Aβ 1-6 A2V TAT(D) in the APPswe/PS1DE9 mouse model prevented cognitive deterioration, Aβ aggregation and amyloid deposition in brain. Unexpectedly, a longer treatment schedule, while retaining the results for cognitive impairment, attenuated the effects on Aβ production and increased amyloid burden, most likely due to the intrinsic amyloidogenic properties of TAT.
Indeed, we found that TAT(D), unlike Aβ 1-6 A2V (D), has a strong ability to bind amyloid deposits. This avidity for amyloid could boost the intrinsic seeding activity of amyloid plaques via a continuous and self-sustained recruitment of Aβ aggregates, leading to an exacerbation of the amyloidogenesis.
A similar effect of TAT was described in a study 26 reporting that HIV TAT promotes AD-like pathology in an AD mouse model co-expressing human APP bearing the Swedish mutation and TAT peptide (PSAPP/TAT mice). These mice indeed showed more Aβ deposition, neurodegeneration, neuronal apoptotic signalling, and phospho-tau production than PSAPP mice.
Moreover, TAT was found to increase Aβ levels by inhibiting neprilysin 27 or enhancing β -secretase cleavage of APP, resulting in increased levels of the C99 APP fragment and 5.5-fold higher levels of Aβ 42 28 . The same study reported that stereotaxic injection of a lentiviral TAT expression construct into the hippocampus of APP/ presenilin-1 (PS1) transgenic mice resulted in increased TAT-mediated production of Aβ in vivo as well as an increase in the number and size of Aβ plaques. This is consistent with our findings, indicating a shift in APP processing towards the amyloidogenic processing in vivo at the end of the 5-month treatment with Aβ 1-6 A2V TAT(D) that was not observed in shorter treatment schedules with the same compound.
Therefore, these data suggest that the final outcome of our in vivo studies with Aβ 1-6 A2V TAT(D) is the result of side effects of the TAT carrier, whose amyloidogenic intrinsic activity neutralized the anti-amyloidogenic properties of the Aβ A2V variant. Nevertheless, we believe that the approach based on the use of Aβ A2V variant can be successfully used in treating AD, because of its potential ability to tackle the main pathogenic events involved in the disease, as suggested by the natural protection against the disease which occurs in human heterozygous A673V carriers.
Interestingly, recent studies produced evidence in favour of a natural protection against AD in human carriers of the A2T Aβ mutation 29 , another human Aβ variant characterized by an alanine-to-threonine substitution at the same APP codon of the A2V-Aβ mutation (APP-A673T or Aβ A2T variant). This mutation segregates almost exclusively with individuals who never showed any sign of dementia, suggesting a possible protective role against AD in the Icelandic and Finnish populations 30 . Additional studies tried to clarify the molecular basis of the A2T-induced protection for AD, suggesting a likely composite mechanism including effects on APP processing (with consequent decrease of Aβ production), and on Aβ structure, aggregation and neurotoxicity [31][32][33] .
More recent papers reported that the A673T variant is extremely rare in other cohorts from the US and Asia and does not play a substantial role in risk for AD outside of Icelandic and Scandinavian geographic areas 34,35 . However, the discovery of protective genetic variants like Aβ A2V and Aβ A2T , although rare, should prompt a novel vision of genetic studies, which until now has been limited to the identification of pathogenic variants, expanding the genetic research into the detection of "protective" DNA variations as a useful foundation for the design of efficient disease-modifying treatments in medicine.
Finally, we would emphasize that, regardless its optimal BBB delivery abilities and cell penetrating activity 36 , some intrinsic properties of TAT sequence can promote amyloidogenesis in long treatment schedules and should be carefully taken into account whenever this peptide is employed in the design of therapeutic strategies for neurological diseases [37][38][39] , particularly AD 40 .

Materials and Methods
In silico studies. The impact of the A2V mutation on the structural features of the Aβ 1-6 peptide was investigated using classical molecular dynamics (MD) simulations with an all-atom approach using the AMBER ff99SB-ILDN force field 41 . The solvent environment was explicitly simulated using the Tip3p water model and applying three-dimensional periodic boundary conditions. A 1.0 nm cutoff for the direct calculation of non-bonded interactions was applied, and long-range electrostatics were treated with the particle-mesh Ewald approach. The systems were prepared by dispersing the Aβ 1-6 WT and Aβ 1-6 A2V peptides in a cubic box of side 5.0 nm, filled with water molecules, and were initially relaxed through energy minimization in order to avoid in the group treated with Aβ 1-6 A2V TAT(D) in comparison with mice treated with saline solution (control group). The significance of the results was calculated using a Mann-Whitney U test. Statistical differences were considered significant if p < 0.05. nonphysical contact between the solute and solvent molecules, then equilibrated with a 5 ns MD simulation at constant ambient pressure and temperature (300 K). The production runs were then extended for 1.5 μ s each. Covalent bonds involving hydrogen were constrained by applying the LINCS algorithm, allowing a simulation timestep of 2 fs 42 . Temperature was controlled by applying the Bussi-Donadio-Parrinello thermostat, while pressure was controlled with an isotropic Parrinello-Rahman barostat 43 . All simulations were carried out using GROMACS 4.6.5 42 . In vitro studies. Peptide synthesis. All synthetic peptides were prepared as previously described 7 via solid-phase Fmoc chemistry on an Applied Biosystems 433 A peptide synthesizer, characterized by matrix-assisted laser desorption/ionization mass spectrometry (MALDI). Their purity was always above 95% (See Supplementary  Information for details).

. Effects of TAT(D) on APP processing (A-D) and binding of peptides to amyloid deposits (E-H).
Densitometric quantification of APP C-terminal fragments revealed an increase in the C99:C83 ratio in mice treated with Aβ 1-6 A2V TAT(D) every 7 days for 5 months (B,D), while no differences in C99:C83 ratio were observed at the beginning of the treatment (A,C). Statistical analysis was performed using a Student's t-test and differences were considered significant if p < 0.05. Immunohistochemistry with biotinylated peptides, original magnification 20×. Amyloid plaque immunohistochemistry in brain sections of CRND8 transgenic mice. E and F show transgenic mouse brain stained with Biot-Aβ 1-6 WT TAT(D) and Biot-Aβ 1-6 A2V TAT(D) respectively. (G): transgenic mouse brain stained with Biot-TAT(D). (H): transgenic mouse brain stained with Biot-Aβ 1-6 A2V (D). Arrow indicates a plaque that is not stained by Biot-Aβ 1-6 A2V . No immunoreactivity was detected in control non-transgenic mice slides co-incubated with Biot-D peptides (data not shown). temperature for 2 minutes. Samples were washed and dried under gentle nitrogen flow. AFM analyses were carried out on a Multimode AFM with a Nanoscope V system (Veeco/Digital Instruments). AFM images were analysed using the Scanning Probe Image Processor data analysis package. All the topographic patterns and SPIP characterization were verified via additional measurements on a minimum of five different, well separated areas.
Binding of Biot-Aβ1-6 WT/A2V TAT(D) to amyloid. It was tested on brain sections of transgenic and control mice of CRND8 strain 44 . To unequivocally identify the fragment(s) of the peptide responsible for binding to plaques we used three separate segments: Biot-Aβ 1-6 WT (D), Biot-Aβ 1-6 A2V (D) and Biot-TAT(D). Seven-micrometre thick brain sections fixed in Carnoy and embedded in paraffin were used for staining. The immunoreaction was revealed by incubation with the avidin-biotin complex (Vectastatin ABC Kit, Vector Laboratories) and diaminobenzidine (DAB) as the substrate for horseradish peroxidase.
Cell models. Toxicity studies with Aβ 1-6 WT (D), Aβ 1-6 A2V (D) or Aβ 1-6 A2V TAT(D) and the short peptides' ability to inhibit the toxicity induced by Aβ 1-42 WT in SH-SY5Y cell cultures were analyzed following previously described protocols 7 . Each experiment was performed in triplicate; the data provided were obtained as the mean of three independent assays.
In vivo studies. Animals. Animal care and experimental procedures were conducted in accordance with European Union (2010/63/EU) and Italian (D. Lgs. 26/2014) legislations and followed the applicable rules and guidelines of the Animal care surveillance Committee of the INCB. All the experiments involving animals were approved by the Animal care surveillance Committee of the INCB and by the Italian Ministry of Health. All animals were sacrificed by cervical dislocation under deep anesthesia (telazol 20 mg/kg and medetomidine 1,5 mg/ kg i.p.).
In vivo assessment of translocation across BBB. APP23 mice (Harlan, Correzzana, Italy) were treated by intraperitoneal injection with Aβ 1-6 A2V TAT(D) (10 mg/kg) or vehicle (saline solution) and sacrificed 1 h or 24 h after the last treatment. Aβ 1-6 A2V TAT(D) was quantified in mouse brain using HPLC-MS/MS 45 . The stability of Aβ 1-6 A2V TAT(D) to protease digestion was also investigated by MALDI TOF MS (See Supplementary  Information for details). In vivo anti-amyloidogenic effects of Aβ1-6 A2V TAT(D) in the huAPP Swe /moAPP 0/0 mouse model. Transgenic APP23 mice expressing the Swedish double mutation in the human APP gene 46 and knock-out for endogenous APP (huAPP Swe /moAPP 0/0 ) were generated in our lab by crossing APP23 mice with APP 0/0 animals. Mice (n = 6) were treated intraperitoneally (i.p.) every day for 21 days with Aβ 1-6 A2V TAT(D) (10 mg/kg) or saline solution and culled 24 h after the last treatment.
In vivo anti-amyloidogenic effects of Aβ1-6 A2V TAT(D) in the APPswe/PS1dE9 mouse model. Heterozygous four-month-old female APPswe/PS1dE9 transgenic mice were used in this study. These mice express human APP carrying the Swedish mutation together with human presenilin 1 (PS1) carrying the pathogenic dE9 mutation 47 . Animals were divided in 2 experimental groups (n = 10 each), treated i.p. once a week with saline solution or Aβ 1-6 A2V TAT(D) and culled 2.5 (intermediate time) and 5 months (final time) after treatment. In both mouse transgenic lines, the brain was removed and dissected into 2 hemibrains by midsagittal dissection. One half was immediately stored at 80 °C for biochemical assays, the other immediately immersed in formalin (10%) overnight for immunohistochemical studies.
Behavioral tests. The Novel Object Recognition test (NOR) was used to examine the cognitive performance of animal models involved in preclinical studies, following well consolidated protocols 48 . The results of the NOR were expressed by using: -Discrimination index (time exploring novel object -time exploring familiar object)/(time exploring novel object + time exploring familiar object).
Biochemical studies. The left hemisphere of each brain was homogenized in 7 volumes of 10 mM Tris-HCl, pH 7.4, added with cOmplete Mini Protease Inhibitors cocktail (Roche), sonicated for 1 min using an ultrasonic homogenizer (SONOPULS) and centrifuged at 100,000 xg for 1 h at 4 °C. The supernatant was saved as the soluble fraction; the pellet was extracted in 70% formic acid and neutralized with 20 volumes of 1 M Tris (insoluble fraction).
Aβ 40, Aβ 42 and aggregated Aβ were measured in both soluble and insoluble fractions by ELISA (Invitrogen). Each experiment was performed in triplicate. The analysis of APP processing followed the protocol described above. After centrifugation, the supernatant was collected as the SDS-soluble fraction, which was analyzed by western blot with the A8717 antibody (Sigma). Signal intensity of the bands was measured by Quantity One (BioRad).
Neuropathological studies. Coronal slices of the right hemibrain were embedded in paraffin and cut (7μ m); sections were de-waxed in xylene and hydrated through serial alcohols to water. After formic acid (80%) pre-treatment, sections were incubated overnight with anti-Aβ antibody (4G8, 1:4000; Covance). The primary antibody signal was detected with a biotinylated secondary antibody followed by horseradish streptavidin peroxidase and visualized with DAB. Amyloid deposition was quantified in mouse brain using Aβ immunostaining (4G8) and the staining intensity was evaluated semiquantitatively using a scale range from 0 to 5 by light microscopy. The assessment was conducted in two adjacent sections of the same brain area 49 . A parallel quantification of the Aβ plaque load was performed using image analysis software (NIS-elements-Nikon) [50][51][52] .
Measurement of serum antibodies against Aβ peptides. Sera were collected from mice before (T0) and at the end of the treatment (5 months), and tested using ELISA for the presence of IgG against Aβ 1-6 A2V TAT(D), the full-length Aβ 1-42, Aβ 1-6 A2V (D) and TAT(D) following previously described protocols for antigen-specific mouse IgG detection 53 , except blocking buffer that was PBS 3% milk. Anti-Aβ 1-42 IgG antibodies from 6E10 IgG clone (Covance), and sera from mice immunized with Aβ 1-6 A2V TAT(D) or with Aβ 1-6 A2V emulsified in Complete Freund's adjuvant were used as positive controls.

Statistical analysis.
Mann Withney U-test was used to compare data obtained by NOR test (Discrimination Index, Preference Index, Recognition Index). Comparison of cell viability after administration of Aβ 1-42, alone or together with Aβ 1-6 A2V TAT(D) or Aβ 1-6 WT TAT(D), was performed by Student t-test. Student t-test was also used to compare (i) amyloid burden in immunohistochemical studies, (ii) Aβ 40, Aβ 42 and aggregated Aβ levels obtained by ELISA tests, and (iii) relative amounts of APP C-terminal fragments after Western Blot quantification. Two tailed P value equal or less than 0.05 was considered statistically significant. All calculations were performed using GraphPad Prism 5.