Peptides for Therapy and Diagnosis of Alzheimer’s Disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder with devastating effects. The greatest risk factor to develop AD is age. Today, only symptomatic therapies are available. Additionally, AD can be diagnosed with certainty only post mortem, whereas the diagnosis “probable AD” can be established earliest when severe clinical symptoms appear. Specific neuropathological changes like neurofibrillary tangles and amyloid plaques define AD. Amyloid plaques are mainly composed of the amyloid-β peptide (Aβ). Several lines of evidence suggest that the progressive concentration and subsequent aggregation and accumulation of Aβ play a fundamental role in the disease progress. Therefore, substances which bind to Aβ and influence aggregation thereof are of great interest. An enormous number of organic substances for therapeutic purposes are described. This review focuses on peptides developed for diagnosis and therapy of AD and discusses the pre- and disadvantages of peptide drugs.


INTRODUCTION
Alzheimer's disease (AD) is a devastating neurodegenerative disorder and the most common cause of dementia. The clinical characteristics are difficulties with memory, apathy and depression, impaired judgment, disorientation, confusion and other. The greatest risk factor for AD is age. In 2007, AD affected 27 million people world-wide with steadily increasing tendency. By 2050, the prevalence is estimated to quadruple, thereby raising significant economic problems, not to mention the suffer of each affected individual [1].
The pathological hallmarks of AD are the presence of neurofibrillary tangles and amyloid deposits in the brain of the patient, as already defined by Alois Alzheimer in 1907 [2]. Neurofibrillary tangles are aggregates of paired helical filament composed of the abnormally phosphorylated and -folded tau protein. Tau is a hydrophilic microtubule binding protein which is expressed in six human isoforms of 352 to 441 amino acid residues [3][4][5].
A is the major component of the amyloid plaques. It consists of 39 to 43 amino acid residues. A , especially A 1-42, is prone to aggregation and undergoes formation from monomers to oligomers, larger intermediate forms like protofibrils, and, insoluble fibrils and plaques [6]. A is derived from the amyloid precursor protein (APP) by sequential activities of the -and -secretases [7][8][9]. As originally suggested by the amyloid cascade hypothesis, it appears likely that A peptides and their aggregated forms initiate cellular events leading to the pathologic effects of AD. According to a previous version of the amyloid cascade hypothesis, fibrillar forms of A , deposited in amyloid plaques, have been thought to be responsible for neuronal dysfunction [7][8][9][10][11]. More recent studies support that diffusable A oligomers including protofibrils, prefibrillar aggregates and so called A -derived diffusible ligands (ADDLs), are the major toxic species during disease development and progression [12][13][14].
Currently, only palliative therapies for AD are available. Acetylcholine inhibitors like Donepezil, Galantamine and the NMDA receptor antagonist Memantine have been approved for clinical use as treatment of cognitive symptoms.
Although it is still controversial if A is the causative agent of AD, inhibition of A production and aggregation are often *Address correspondence to this author at the Forschungszentrum Jülich, ICS-6, D-52425 Jülich, Germany; Tel: +49 2461 612100; Fax: +49 2461 612023; E-mail: d.willbold@fz-juelich.de addressed for therapy development. As a consequence, the majority of AD therapeutic research has been focused on the A peptide. Less effort has been directed towards the development and validation of tau-targeted therapeutic compounds. A number of tau fibril formation inhibitors, derived from multiple chemical classes, have been identified, as reviewed elsewhere [15]. To date, only one tau fibrillization inhibitor, the phenothiazine methylene blue [16], has entered in vivo evaluation. Promising first results have been presented at the ICAD meeting in 2008, but up to now have not been published in a peer-reviewed journal. Currently, only one peptide compound addressing tau pathology is known. Davunetide (DAP) is an eight amino acid peptide derived from the activity-dependent neuroprotective protein ADNP. It decreases tau phosphorylation and A levels in tau transgenic mice and 3 x transgenic (tg)-AD mice. The intranasal formulation AL-108 is currently in clinical development [17][18][19][20][21].
Peptides, which are specified as (linear) molecules consisting of two or more (<100) amino acids residues, are today reasonable alternatives to chemical pharmaceuticals. They are key regulators of biological functions and offer high biological activity associated with high specificity and low toxicity. The peptide market is growing fast due to an increased number of therapeutic targets, improved delivery methodologies, the establishment of large biological and synthetic peptide libraries, and high throughput screening or selection. Today, 67 therapeutic peptides are on the market, 150 in clinical phases and more than 400 in the pre-clinics. In spite of this progress, the development of peptide drugs can be severely hampered by their short half-life in vivo. In general, peptides are rapidly degraded by proteases, and their nature implies problems for administration and delivery, especially to the brain. These problems can at least partially be overcome as peptide chemistry permits a variety of methods for peptide modification or the use of D-enantiomeric amino acid residues [31][32][33].
A variety of small peptides that inhibit aggregation of A and reduce its toxic effects were already described and a fraction of them shown to be effective in AD rodent animal models. Additionally, A binding peptides, developed for a suitable use in in vivo strated to bind A with an affinity in the submicromolar range. Employing surface plasmon resonance, binding to A oligomers and fibrils, but not to monomers could be demonstrated. D1 stained amyloid plaques in the brain tissue sections derived from AD patients, whereas other, non-A amyloidogenic deposits, were not stained [45,46]. D1 was also tested for its in vivo binding characteristics in APP/PS1 transgenic mice. Upon direct injection into the brain, D1 bound very specifically to A 1-42, staining all dense deposits in the brain but not diffuse plaques, which contain mainly of A 1-40 and are not AD specific [47]. This demonstrated that D1 might be suitable for further development into a molecular probe to monitor A 1-42 plaque load in the living brain.
In 2008, Larbanoix et al. selected A 1-42 binding peptides using a random disulfide constrained heptapeptide phage display library. Two clones (see sequences in Table 1) were enriched. The K d -values for the phage clones, on which several peptide copies are displayed, were in the picomolar range. After peptide synthesis including biotinylation, the binding affinities dropped to the micromolar range. Nevertheless, preliminary in vivo studies in transgenic mice, in which the functionalized peptides were used as contrast agents, showed high contrast effects in APP V717I /PS1 A246E transgenic mice, but not in wildtype controls. For the experiments, however, the blood brain barrier (bbb) had to be permeabilized artificially with 25 % mannitol [48]. Very recently, Larbanoix et al. designed another linear hexapeptidic phage display library based on the A 1-42 amino acid sequence and selected against aggregated A 1-42 as a target. Two of 26 selected clones, presenting highest binding affinities to A 1-42, were translated to synthetic peptides with biotine label (Pep1: LIAIMA and Pep2: IFALMG, corresponding A fragment IIGLMV [31][32][33][34][35][36] ) and presented lower K d values (still in the micromolar range) as the peptides described in the first article. Pep1 and Pep2 were highly hydrophobic and are expected to pass the bbb very well. The peptides did not show any sign of toxicity in cell culture. The specific interaction of both peptides with amyloid plaques in human brain tissue was demonstrated by immunohistochemistry [49].

THERAPEUTIC PEPTIDES (SEE TABLES 2A-D)
A -sequence Derived Peptides (see Table 2A) In 1996, Tjernberg and coworkers searched for an A ligand to interfere with A -self interaction and polymerization. The strategy was to identify binding sequences within A and, based on their primary structures, to synthesize A derived peptide ligands. The short A fragment KLVFF (A [16][17][18][19][20] was identified to bind full length A and to prevent the fibrillization thereof. Alanine substitution experiments revealed that amino acids Lys16, Leu17 and Phe20 were critical for A interference [50]. A molecular modeling study suggested that the association of full length A and the KLVFF peptide lead to the formation of atypical antiparallelsheet structures stabilized by Lys16, Leu17 and Phe20 [51]. The effects of KLVFF containing or derived synthetic peptides was further approved in a variety of studies, e.g. by the one of Matsunaga et al. [52], see Table 2A. Additionally, it was proposed that conjugates, bearing several copies of the KLVFF sequence or the retro-inverso version thereof, linked to dendrimers or to branched poly(ethylene glycol) moieties, possess superior affinity and efficiency [53,54]. In 2008, Austen et al. designed KLVFF derived compounds to address very early aggregation intermediates of A , i.e. A oligomers. The idea was to add water soluble amino acids residues to KLVFF, thereby generating the peptides OR1 (RGKLVFFGR) and OR2 (RGKLVFFGR-amid). Both peptides inhibited A fibrillogenesis, whereas OR2 additionally inhibited oligomer formation and A toxicity on SY5Y cells, supporting the idea that particularly oligomers are responsible for the cytotoxic effects of A [55]. Unlike OR2, the retro-inverso D-enantiomeric version (RI-OR2) of the peptide was highly resistant to proteolysis and stable in human plasma and brain extracts [56]. Additionally, it  [49] bbb: blood brain barrier. D/L: describes peptide conformation.       was reported that retro-inversions of OR1 and OR2 increase the inhibitory effects of the peptides [57]. In 2004, Fülop et al. developed an A aggregation inhibitor based on the A 31-34 sequence IIGL, which also plays a fundamental role in A aggregation and cytotoxicity [58][59][60]. As in the study described above, the strategy was to link a solubilizing amino acid residue to the original sequence RIIGL. In contrast to propionyl-IIGL, another derivative of the same sequence, PIIGL did not self-aggregate and was not toxic to cells in culture. RIIGL inhibited the formation of A fibrils and reduced A cytotoxicity [61].
In 2008, Fradinger et al. prepared a series of A C-terminal fragments (A x-42; x = 28-39). The authors of the article tested the hypothesis that C-terminal peptides of A should possess high affinity to full length A and might disrupt oligomer formation, as the C-terminus is supposed to be a key region controlling A aggregation. Cell viability assays identified A 31-42 and 39-42 to be the most effective inhibitors of A induced cell toxicity. The peptides additionally prevented the disturbance of synaptic activity by A oligomers. To investigate the in vitro mechanism of action, dynamic light scattering, photo-induced cross-linking and discrete molecular dynamics were applied and gave clear hints that the peptides inhibit A induced toxicity by stabilizing A in non-toxic oligomers [62].
The inheritance of the apolipoprotein (apo) E4 allele has been identified as a major genetic risk factor for sporadic AD [63]. All Apo isoforms are discovered to act as pathological chaperones and propagate A fibril formation, with apo E4 being the most efficient isoform [64,65]. Sadowski et al. investigated weather blocking of the interaction between apo E4 and A can have therapeutic effects [66]. In earlier studies, Ma et al. demonstrated that the synthetic peptide A 12-28 can be used as inhibitor of apo E4-A interaction, inhibiting A fibril formation in vitro [67]. Sadowski et al. modified the A 12-28 sequence: substitution of valine at position 18 to proline rendered the peptide non-amyloigogenic and untoxic whereas the affinity of the peptide to apo E4 was not affected. The use of D-amino acids and end-protection increased the serum halflife of the peptide substantially. A 12-28P blocked A -apoE4 interaction and reduced A fibrillogenesis and toxicity in vitro. The peptide was bbb-permeable and inhibited A deposition in AD transgenic mice [66].

-sheet Breaking Peptides (see Tables 2B-2E)
The first -sheet breaker peptides were reported by Soto et al. in 1996 [68]. In general, the term -sheet breaker describes compounds, containing an A recognition or binding motif which provides specificity, combined with an A oligomer or fibril disrupting D-4F Ac-DWFKAFYDKVAEKFKEAF-NH2 Apo A-I mimetic peptide D Inhibits A deposition and improver cognitive performance of AD tg mice after oral application [120] D/L: describes peptide conformation. TEG: triethyleneglycol. +: positively charged amino acids. H: hydrophobic amino acid. X: hydrophobic or polar amino acid motif which can consist of charged amino acids, prolines, methylated amino acids, cholyl-groups and others. By far the most A aggregation inhibiting peptides are -sheet breaker, demonstrating the effectiveness of these compounds. In most cases, the A recognition domain is based on the A sequence.
-sheet Breaking Peptides Based on the KLVFF Sequence (see Table 2B) Starting in 1996, Ghanta et al. designed A binding hybrid peptides based on the KLVFF-binding sequence in addition to a disruption domain consisting of a chain of charged amino acids like KKKKKK or RRRRRR. Interestingly, some of the hybrid peptides accelerated A aggregation, but reduced A toxicity. Presumably, the peptides speeded up the association of potentially toxic A oligomers into less toxic A fibrils as demonstrated by a range of biochemical and biophysical methods. The ability of the compounds to increase solvent tension was a very strong predictor on the effect on A aggregation. Chains of charged amino acids without A recognition motif, used in the assays as control, did not exert comparable strong effects [69][70][71][72][73][74][75]. In a similar study, Sun et al. investigated hybrid peptides composed of the critical A binding domain EIVY and solvent disruptive sequences poly E, K or R. The hybrid peptides EIVY-EEEE and EIVY-KKKK enhanced A fibrillization, whereas EIVY-RRRR inhibited A aggregation and altered the morphology of A fibrils to amorphous aggregates. The ability of the hybrid peptides to interfere with A aggregation was also discussed in the context of their abilities to change the surfacetension in solutions [76], in Table 2E.
In 2003, Gordon et al. replaced the amide bonds of A 16-20 with ester bonds in an alternating fashion. The ester peptide A 16-20e was monomeric under solvation conditions, inhibited A aggregation and disassembled existing fibrils. A 16-20e, could, however, build dimers. These data demonstrated that interference with backbone hydrogen bonding is therapeutically attractive [77,78].
Other -sheet breakers based on the KLVFF sequence are AMY-1 and AMY-2. Both contain alpha, alpha-disubstituted amino acids at alternating positions and arrest A fibril growth. Instead, large globular aggregates are formed [79]. Rangachari et al. investigated the , -dehydroalanine ( Ala) containing peptides P1 (KLVF-A-I A) and P2 (KF-A-A-A-F). The design of P2 was based on the experience that peptides containing the FxxxxF motif bind to the groove composed of the GxxxG motif (amino acids G33-G37) in A fibrils [80] , -dehydro-amino acid residues are known to be strong inducers of specific peptide conformations. Additionally, they increased resistance to proteolysis in vivo. Both peptides under investigation inhibited A aggregation [81]. Table 2C)

Proline Based -sheet Breaker (see
One peptide compound which was extensively investigated in vitro, as well as in animal models, was introduced 1996 by Soto et al [68]. The authors developed a peptide partially homologous to the central hydrophobic region of A (amino acids 17-21: LVFFA), containing the amino acid proline to prevent the formation ofsheet structure and to inhibit A amyloid formation. Proline has special characteristics and is a well-known -sheet blocker [82]. To increase the solubility of the peptide compound, charged amino acids were added to the ends. The so called "inhibitor of fibrillogenesis" (iA 1: RDLPFFPVPID) did not aggregate itself, inhibited amyloid formation and disassembled existing fibrils in vitro. The relative dissociation constant of iA 1 was determined by fluorescence spectroscopy to be approximately 80 nM. In order to enhance bbb permeability and to reduce generation of immune responses in vivo, the peptide was shortened, and the derivatives iA 3 (seven amino acids) and iA 5 (five amino acids) were proven to be similar good or even better A aggregation inhibitors in vitro than the basic compound, compared e.g. in ThT assays. The D-enantiomeric version of iA 1 was as effective as the L-version, and more resistant to proteases [68]. In later studies, iA 5 was shown to inhibit A cytotoxicity in cell culture. It also reduced A fibrillogenesis in a rat brain model of amyloidogenesis. Male Fischer-334 rats were injected with freshly solubilized A 1-42 directly into the amygdala and the animals were sacrificed 8 days later. Co-injection of iA 5 in 20 % molar excess lead to significantly reduced plaque deposition [83]. In addition, it was shown that iA 5 induced disassembly of fibrils, reduced A induced histopathological changes like neuronal shrinkage and the extent of interleukin-1 positive microglia cells surrounding A deposits [84]. The chronic intraperitoneal administration of the peptide to the rat model described above lead to a significant improvement of spatial learning acquisition in Morris water maze tests and working memory tests. To perform these studies, iA 5 was end protected by N-terminal acetylation and Cterminal amidylation [85], as the non-protected peptide was unstable in blood and proteolytically degraded very rapidly [86]. In 2002, two different AD transgenic mouse models (double transgenics overexpressing human APP with London mutation V717I and human PS1 with A246E mutation, and single transgenics only overexpressing human APP V717I) were used to demonstrate that intraperitoneal injected, end-protected iA 5 reduced amyloid plaque formation, neuronal cell death and brain inflammatory processes. Pharmacokinetic studies in mice and rats demonstrated good stability of the end-protected peptide as well as high capability of the peptide to cross the bbb. The mechanism of peptide action remained unclear, but administration of large doses of the peptide did not lead to antibody production in the treatment and evaluation period [86].
In 2004, Datki and Coworkers investigated the effects of a LPFFD based peptide, amino acid sequence LPYFD, on the A induced changes on neuroblastoma cells. Neurite degeneration and tau aggregation were significantly decreased. Additionally, A induced cell toxicity was reduced [87]. The pentapeptide LPYFDamid protected neurons against A toxicity in vitro and in vivo. Intraperitoneally administered LPYFD-amid crossed the blood brain barrier in rats at least to a certain extent, and protected against synaptotoxic effects of A up to 3.5 h after i.p. injection [88,89]. Table 2D)

Methylated -sheet Breaking Peptides (see
Several teams have studied the effects of N-methyl amino acid incorporation into peptides. For the resulting compounds, the term "meptides" was established. The first methylated -sheet breaker was described by Hughes et al. After it has been suggested that A 25-25, amino acid sequence GSNKGAIIGLM, resembles a biologically active region of A that forms large -sheet fibrils and is highly cytotoxic [90][91][92], the authors used it as a full length model for A 1-42 and synthesized six N-methylated derivatives to prove that those could prevent A wildtype aggregation and cytotoxicity. As the derivatives were homologous to A , they were expected to bind the wildtype form and to prevent further addition of A monomers. It was assumed that N-methylation could block hydrogen bonding at the outer edge of the assembling amyloid, disrupting peptide-peptide interactions that promote A fibrillization. As expected, N-methylated peptide variants had significant influence on A aggregation, as investigated using a variety of biophysical methods. Notably N-methyl-Gly-33 was shown to be amenable to inhibit A aggregation and cytotoxicity. The localization of the Nmethyl group was very critical as some of the other peptides did not prevent A aggregation, but altered fibril morphology [93,94]. In 2001, Gordon et al. described the synthesis and biochemical characterization of rationally designed "meptides" based on the KLVFF sequence, containing N-methyl amino acids in alternating positions of the sequence. One of the compounds, termed A 16-22m (NH 2 -K(Me-L)V(Me-F)F(Me-A)E-CONH 2 ) was shown to be highly soluble in aqueous media and monomeric in buffer solution. It inhibited A fibrillization and disassembled preformed A fibrils in vitro. Inhibition was sequence specific and dependent of N-methylization. Protease resistance of the methylated peptide was increased in comparison to the unmethylated A 16-22 peptides [78]. The A 16-20m peptide, a truncated version of the peptide described above, was synthesized in order to eliminate the charged Glu residue, providing the inhibitor with a net positive charge. A 16-20m was effective to inhibit A polymerization and to dissassemble preformed fibrils. In addition, it was highly water soluble despite its composition of hydrophobic amino acids. The peptide passed spontaneously model phospholipid bilayers and cell membranes, suggesting promising pharmacological properties [95].
In 2004, Cruz et al. developed the peptide "inL", based on the KLVFF A recognition element. The authors added an additional Lys to the N-terminus, in order to increase solubility, and an Nmethyl-20F in order to block A aggregation [96]. The peptide inhibited A toxicity in cell culture very efficiently. In 2009, the corresponding D-peptide "inD", as well as the retro-inverso peptide "inrD", was investigated. Both D-enantiomeric peptides were more resistant against protease degradation, as expected. The retroinverso peptide was shown to be a more effective inhibitor of A aggregation than the other two peptide versions [97].
The peptidic inhibitor PPI-1019, also known as Apan, is derived from the D-enantiomeric Cholyl-LVFFA-NH2 (see chapter other -sheet breaking peptides). The cholyl-group was replaced by a methyl-group and the C-terminal residue was changed from Dalanine to D-leucine. PPI-1019 completed phase I and II clinical trials and was found to be safe, well tolerated and amenable to cross bbb. After peptide administration, levels of A 1-40 in the CSF increased, which might be discussed as a sign for A clearance out of the brain [98,99].
In 2006, Kokkoni et al. optimized "meptides" based on the KLVFF sequence in a large approach based on five peptide libraries. Peptide length, methylation sites, end-blocking, side chain identity and chirality were varied. The most interesting compound, judged by A fibrillization and cell cytotoxicity inhibition activity, was D-[(chGly)-(Tyr)-(chGly)-(chGly)-(mLeu)]-NH 2 and rules could be stated to predict peptide performance. Ideal inhibitors should be D-peptides, possess a free N-but an amidated C-terminus and residues one to four should be large, branched hydrophobic side chains. Only one methylated amino acid was essential [100]. Table 2E)

Other -sheet Breaking Peptides (see
In 1999, Findeis et al. started the development of new -sheet breaking peptides derived from 15-residue A peptides. The anti-A fibrillization activity of the peptides was enhanced by modification of their amino terminus, where different organic reagents were attached. In subsequent libraries, the size of the inhibitors as well as the amino acid sequences were optimized, and finally, the lead compound cholyl-LVFFA-OH, designated PPI-368, was identified. PPI-368, the acyl-D-amino acid analogue PPI-433 and the amide analogue PPI-457 inhibited A polymerization potently. The latter two were stable in monkey CSF for at least 24 h. Unfortunately, hepatic first -pass elimination was reported for all compounds after biodistribution studies were performed [101,102]. In 2009, Fryman-Marom et al. introduced the D-enantiomeric -sheet breaker NH 2 -D-Trp-Aib-OH, which combined an indole and -aminobutyric acid (Aib) [103]. Aib has been shown to induce helical conformations and to disrupt -sheet structures, the Ramachandran plot indicating that Aib even has the potential to be an better -sheet breaker than proline [104]. NH 2 -D-Trp-Aib-OH specifically interacted with (untoxic) low molecular weight A oligomers and inhibited their growth to larger, celltoxic A forms in vitro. The compound was stable, safe, orally bioavailable, and crossed the bbb in the range of 4 to 8%, depending on the route of administration. It reduced the amount of plaques in the brains of AD transgenic mice and improved their cognitive performance [103].

Peptides Selected in Combinatorial Libraries
Combinatorial peptide libraries offer a powerful technique to select highly specific peptide ligands for different pharmacological interesting targets [44,105].
Already in 1997, Blanchard and colleaques selected so-called "decoy peptides" using a combinatorial library of approximately 43000 individual sequences composed of the D-amino acids Ala, Ile, Val, Ser, Thr and Gly. The 6 residue peptides were chosen by their ability to complex with tagged A 25-35 peptide and hadsheet forming potential, associating with A and blocking aggregation. Some of the selected peptides abolished the calcium influx, caused by aggregated A 25-35 or A 1-42 in cell culture [106,107].
Schwarzman and al. selected inhibitors of amyloid formation by screening of a FliTrx random peptide library. 12 residue peptides were displayed on the surface of E. coli by fusion to a flaggelar protein. Synthetic A 1-42 was used as a target molecule in five rounds of biopanning. Four groups of A binding peptides were isolated (see Table 2F), two of which were enriched by positively charged amino acids. Most clones were shown to bind monomeric A , but exhibited very low binding to fibrillar A 1-42 [108].
Orner et al. identified peptides that bind to A in either monomeric or fibrillar state using phage display approaches with monomeric or fibrillar A as targets, respectively. Two libraries were designed, guided by the group's previous studies on the KLVFFK 6 peptide (see above). The first library displayed sequences with the PoPoPoKLVFFPoPoPo motif, where Po indicates a residue with polar side chain. The second library contained sequences with XXXKLpLpArArPoPoPoPo motif, where X is any amino acid, and Lp and Ar indicate residues with lipophilic and aromatic side chains. Peptides with selectivity for monomeric versus fibrillar A could be identified and most of the selected peptides bound to A 10-35 with higher affinity than the parent peptide Ac-KLVFFKKKKK-OH. Peptides selected for A monomer binding did not affect aggregation, whereas peptides selected to bind fibrillar A increased the aggregation of A dramatically, altering the morphology of the resulting aggregate. This effect was clearly correlated with affinity of the peptides to the N-terminal part of A [109].
Taddei et al. reported on an approach with the aim to inhibit the catalytic production of H 2 O 2 by A , which is dependent on A 's superoxide dismutase (SOD)-like activity. A phage display procedure with 6-and 15mer peptide libraries was applied to select for peptides that target the active site of human A 's SOD-like activity, in order to prevent its interaction with redox-active metal ions. As the SOD-like activity site is not present in rat A , a counterselection step with rat A was employed in the phage display procedure. A 1-40 or A 1-42 were used as targets. 25 peptides which bound to A were identified, and two of the three most enriched peptides, named amyloid neutralizing agents (ANA) 1 to 3, were shown to significantly reduce A 's SOD-like activity in cell culture. A 15mer peptide additionally reduced A toxicity in cell culture and seemed to be comparably potent as the known A metal-mediated redox activity inhibitor Clioquinol [110]. A very sophisticated system was used by Baine et al. to select for peptides that inhibit A aggregation in two combinatorially diverse peptide libraries. The goal was to select peptides which bind the two hydrophobic patches of A and block aggregation by highly charged and polar aspartatic acid residues. A 1-42 was genetically fused to EGFP. When expressed in E. coli, aggregation of A inhibits the correct folding of EGFP and therefore its fluorescent properties. In the selection process, randomized peptides were coexpressed with A -EGFP. Peptides which were resistant to degradation by cellular proteases and inhibited A aggregation permitted EGFP to be folded properly. Colonies with the brightest fluorescence were chosen for further characterization. Three candidate peptides were selected and characterized, being capable to inhibit A aggregation. One of them even disaggregated preformed A fibrils [111].
Recently a highly specific D-enantiomeric ligand for A has been identified using a mirror image phage display approach with a huge randomized 12-mer peptide library (> 1 billion different peptides). The dominant peptide sequence RPRTRLHTHRNR was obtained, referred to as D3. D3 modulated A aggregation and inhibited A toxicity in cell culture. In vitro data clearly demonstrated that D3 is able to precipitate toxic A oligomers into large, high-molecular-weight, nontoxic, ThT negative, nonamyloidogenic amorphous aggregates that fail to act as seeds in A fibril formation assays. D3 did not increase the concentration of monomeric A . Computational simulations of an A nonamer in the presence and absence of D3 proved strong interactions between the arginine-rich D3 and negatively charged groups of A , which were expected to compensate the charge on the A surface and reduce solubility and promote the aggregation of A . Moreover, D3 binding also showed effects on the topology of the A oligomers, which induced a large twist and facilitated the formation of nonfibrillar aggregates [112,113]. Van Groen et al. demonstrated the usage of FITC-labelled D3 for both in vitro and in vivo staining of A -1-42 in the brains of transgenic AD-model mice [47]. Additionally, D3 was proven to have notable bbb permeability in an in vitro bbb cell culture model which further demonstrated the therapeutic potential of D3 [114]. Most recently, oral treatments of A transgenic mice with D3 yielded significant cognitive improvement, reduction of plaque load and plaque-related inflammation [112]. In 2010, Müller-Schiffmann et al. reported on the D3 hybrid compound JM169, which combined the D-enantiomeric peptide with a -sheet breaking compound via a linker substance. The authors demonstrated that the hybrid compound was more efficient in vitro than the sum of its components and had novel properties [115].
In 2009, Paula-Lima et al. used a phage display approach to select peptides binding to the aggregated form of A . One of the identified heptapeptides with the amino acid sequence GNLLTLD (designated GN peptide) was detected to be homologous to the Nterminal domain of mammalian apolipoprotein A-I. Apo A-I, the major protein component of high-density lipoprotein (HDL), has a central role in reverse cholesterol transport [116,117] and antioxidant as well as anti-inflammatory properties [118]. It was shown that purified human apoA-I and A formed complexes. The interaction of apo A-I also rendered the morphology of amyloid aggregates [119]. Likewise in 2009, Handattu et al. evaluated the apo A-I mimetic peptide D-4F, synthesized from D-amino acids and coadministered with pravastatin, as a treatment for AD transgenic mice [120]. D-4F was developed based on the presence of lipidassociating amphipathic -helices in apo A-I and possessed the ability to avidly bind lipids [121,122]. Most studies of D-4F were focused on its potential role in atherosclerosis management [123,124]. Several studies suggested that AD may have an inflammatory component similar to atherosclerosis that is associated with very small vessels such as arterioles [125][126][127]. In the study, groups of male mice were treated with D-4F and pravastatin via the drinking water. In comparison to the controls, the treated group showed significantly increased cognitive behavior, reduced plaque deposition and reduced inflammatory responses [120].
Very recently, Kawasaki et al. constructed a random library to obtain peptide inhibitors specific to inhibit formation of soluble 37/48 kDa A 1-42 oligomers. The random library was based on the LPFFD sequence: XX-P-XXX, where X means any amino acid. Novel peptides containing arginine residues were enriched while panning for soluble A 1-42. Selected ligands with the strongest affinity to A contained three arginine residues and suppressed formation of 37/48 kDa oligomers and kept the monomeric form of A even after 24 hours of incubation [128].

CONCLUSION
During the past years, several peptide inhibitors of A aggregation have been investigated for their applicability as new therapeutic lead compounds. In conclusion, it must be stated that only very few, iA 5 [84][85][86], A 12-28P [66], LPYFDa [88,89], trp-Aib [103], D-4F [120] and D3 [47,112,113] were proven to be effective in rodent mouse models. Only one compound, PPI-1019, is tested in clinical trials. Despite the high diversity of peptides, combined with their simplicity, high specificity, low toxicity and high biological activity [31,33], peptides are susceptible to proteolytic degradation and in general do not circulate for more than a few minutes in blood [129]. Additionally, peptides and do not cross membranes very well [130].
Several strategies were already applied to overcome high protease susceptibility of peptides and to improve bbb permeability. Different chemical modifications, including incorporation of conformationally constrained amino acids, or modifications of the peptide backbone, have been performed. For example, Adessi et al. introduced a methyl-group at the nitrogen atom of one amid bond in iA 5, resulting in a 10-fold higher in vivo life-time in comparison to the unmodified iA 5 [131]. Moreover, end-protection is commonly used to shield peptides against proteolytic degradation as well as to increase bbb permeability [66,89,132]. Another promising strategy to improve the peptide stability is the use of Denantiomeric amino acids which are considered to be rather protease resistant in vivo and in addition often less immunogenic than the respective L-peptides [42,66,133]. Additionally, D-peptides can be taken up systematically after oral administration [134].
It was assumed for a long time that A deposited in extracellular amyloid fibrils and plaques were the major pathogenic species in AD. However, over the past decade, accumulating evidence suggests that A oligomers are the toxic moiety responsible for synaptic dysfunction and neuronal cell loss [12]. At the moment it is unclear whether peptide inhibitors should be targeted to monomeric A , oligomeric A or A fibrils deposited in plaques. It is also unclear if A 1-40 or A 1-42 should be addressed and weather compounds need to cross the bbb in order to be effective. More work is necessary to elucidate the nature of the most synaptotoxic A species during AD development and progression. The ongoing research on peptides which target distinct A species, as well as the investigation of their influence on A aggregation and toxicity, will provide further understanding of the molecular mechanisms involved in AD.