Competitive Mirror Image Phage Display Derived Peptide Modulates Amyloid Beta Aggregation and Toxicity

Alzheimer´s disease is the most prominent type of dementia and currently no causative treatment is available. According to recent studies, oligomeric species of the amyloid beta (Aβ) peptide appear to be the most toxic Aβ assemblies. Aβ monomers, however, may be not toxic per se and may even have a neuroprotective role. Here we describe a competitive mirror image phage display procedure that allowed us to identify preferentially Aβ1–42 monomer binding and thereby stabilizing peptides, which destabilize and thereby eliminate toxic oligomer species. One of the peptides, called Mosd1 (monomer specific d-peptide 1), was characterized in more detail. Mosd1 abolished oligomers from a mixture of Aβ1–42 species, reduced Aβ1–42 toxicity in cell culture, and restored the physiological phenotype in neuronal cells stably transfected with the gene coding for human amyloid precursor protein.


Introduction
Dementia counts for more than 30 million patients worldwide. Alzheimer´s disease (AD) is the most frequent type of dementia and therefore contributes largely to this number [1,2]. Because age is the most important risk factor for AD, the number of AD patients is expected to increase further.
AD is a progressive type of dementia with symptoms like memory loss, apathy, depression, anxiety and behavioral changes as well as neuropathological hallmarks including inflammation, neuronal cell death and loss of brain mass [2][3][4][5][6]. To date, AD can only be treated palliatively and symptomatically, although with only limited success. Throughout our life time, Aβ is constantly produced from the amyloid precursor protein (APP) by βand γ-secretases [7,8].
Currently, soluble Aβ oligomers are assumed to be the major toxic species in AD [9][10][11]. Aβ oligomers were shown to account already at nanomolar concentrations for neuronal atrophy, induction of synaptic dysfunction, formation of pores and thereby disturbance of cellular ion homeostasis and inhibition of long-term potentiation, altogether leading to impaired neuronal function, reduced cognition and a decline in memory [12][13][14][15][16]. The conversion between the different Aβ assembly states is dynamic and reversible [13,14,[17][18][19][20][21][22]. Recent research gave rise to the fact that AD might be a prion-like disease with misfolded Aβ molecules acting as nucleation seeds and initiating aggregate formation by recruiting additional unfolded or oligomeric Aβ peptides and thereby accelerating amyloid growth [23,24].
From the current point of view, elimination of Aβ oligomers may be the most promising therapeutic intervention to AD. Our approach was to identify therapeutically interesting compounds by selection of ligands, which modulate the dynamic equilibrium of Aβ 1-42 species towards decomposition of toxic oligomers by binding of the ligand to monomeric Aβ and stabilizing it within the equilibrium and thereby shifting the equilibrium away from Aβ oligomers.
Phage display selection is suitable to identify small peptide ligand that bind to any target, for example, Aβ. The variation of phage display using the exact mirror image of the original target is called mirror image phage display. This method allows the identification of peptides that consist solely of D-enantiomeric amino acid residues [25,26]. Such D-peptides are more protease-resistant than peptides consisting of L-enantiomeric amino acid residues and have a prolonged half-life in the body [27][28][29][30]. Additionally, D-peptides are assumed to show lower immunogenicity even though immunogenicity levels may depend on the dose and frequency of administration [27-29, 31, 32].
Previously, we have shown that the D-peptide D3, which was derived from a mirror image phage display selection, is able to reduce the amyloid plaque load in transgenic mice, improve cognition and reduce inflammation after oral application [33,34]. This has proven the suitability of mirror image phage display to identify highly active compounds that are stable enough for oral administration.
Here, we set out to repeat this successful strategy, but with an important modification to yield small, D-enantiomeric peptides that show increased specificity for Aβ monomers in order to stabilize them. Based on a previously described combination of selection and counterselection [35] we used D-enantiomeric Aβ 1-42 monomers as bait and added Aβ 1-42 oligomers and fibrils as counterselective agents to achieve our goal.

Separation of Aβ 1-42 species via size exclusion chromatography (SEC)
In order to obtain monomeric and oligomeric Aβ 1-42 fractions, size exclusion gel filtration chromatography was executed following an optimized protocol of Johansson et al. [36]. Seedless Aβ 1-42 in HFIP was dried in a vacuum centrifuge and solved in elution buffer (50 mM sodium phosphate, 150 mM NaCl, pH 7.4) to a concentration of 250 μM. After one minute vortexing, the sample was sonicated for one minute and centrifuged for 60 seconds at 16,000 x g to precipitate eventually undissolved remains. The supernatant was transferred in a syringe to a Superdex 75 10/300 GL column connected to an ÄKTA purifier FPLC chromatography system (both GE Healthcare Europe GmbH, Freiburg, Germany). Size exclusion was executed at a flow rate of 0.6 ml per minute and detection took place at 214 nm. The desired fractions were collected and pooled. Aβ  oligomers eluted at about 8 ml and Aβ 1-42 monomers at approximately 14 ml.

Preparation of Aβ 1-42 samples with a broad range of different sized species
In order to gain a solution of Aβ 1-42 which resembles several toxic and nontoxic Aβ species over a broad size range, 80 μM seedless Aβ 1-42 was incubated in 10 mM sodium phosphate buffer (pH 7.4) for 4.5 hours at RT and shaking at 600 rpm.

Preparation of Aβ 1-42 fibrils
In order to gain Aβ 1-42 fibrils, seedless Aβ 1-42 was incubated at 80 μM in 10 mM sodium phosphate buffer (pH 7.4) at RT for at least 24 hours with shaking at 600 rpm. To separate the fibrils from other species, the sample was loaded onto an iodixanol density gradient as described below. Fractions containing fibrils and high molecular weight (HMW) aggregates (fractions [12][13][14] were used for further experiments.

Density gradient centrifugation (DGC)
In order to separate differently sized Aβ 1-42 assemblies, 100 μl of the sample were added on the top of a discontinuous gradient of iodixanol (OptiPrep, AXIS-SHIELD, Oslo, Norway). The gradient was prepared by layering 260 μl of 50% iodixanol at the bottom of a 11 x 34 mm polyallomer centrifuge tube, overlaid by 260 μl of 40%, 260 μl of 30%, 780 μl of 20%, 260 μl of 10% and 100 μl of 5% iodixanol. The samples were spun at 260,000 x g for three hours at 4°C in an Optima MAX-XP ultracentrifuge with a TLS-55 rotor (both Beckman Instruments, Brea, USA). After centrifugation, 14 fractions of 140 μl each were harvested with a pipette from top to bottom.

Competitive mirror image phage display
In order to obtain specifically Aβ 1-42 monomer binding species, a mirror image phage display with six panning rounds was performed. To reduce binding of phages to larger Aβ 1-42 species, Aβ 1-42 oligomers and fibrils were added, starting from round two, as counter-selective agents. Phages binding to Aβ 1-42 oligomers and fibrils were subsequently removed from the solution leading to an enrichment of phages, which exclusively bind monomeric Aβ 1-42 .
The target peptide, SEC-derived monomeric, D-enantiomeric, N-terminally biotinylated Aβ  , was immobilized to plates with alternating surface properties. In order to avoid selection of plastic binding phages, different plastic surfaces were used in each subsequent round of panning (Nunc 96-well Immobilizer Streptavidin microwell plate, polystyrene (Thermo Fisher Scientific Inc., Waltham, MA, USA), polypropylene and polycarbonate microwell plates 96-well (BioTeZ Berlin Buch GmbH, Berlin, Germany)). According to the manufacturer, the polystyrene plates were pretreated by washing three times with 300 μl of 1x TBS containing 0.05% (v/v) Tween-20 per well. Additionally, in every second round, the surface was blocked with 150 μl of 1x TBS / 0.1% (v/v) Tween-20 / 1% BSA (w/v) per well with gentle shaking for one hour at RT prior to the addition of the target peptide to reduce unspecific binding. Thus, no combination of plate surface and blocking was used twice during the six panning rounds performed.
One-hundred microliters of a 63 nM solution of target peptide diluted in 1x TBS per well was immobilized to the well for five minutes at RT, followed by three washing steps with 150 μl of 1x TBS.
In the first panning round 90 μl of 1x TBS were mixed with 10 μl of the Ph.D.-12 Phage Display Peptide Library (New England Biolabs GmbH, Frankfurt/M, Germany) and added to the well for five minutes of incubation at RT. The solution was removed and 100 μl of 10 μM biotin in 1x TBS / 0.1% (v/v) Tween-20 (TBS-T) were added for five minutes incubation at RT to reduce streptavidin-binding phages. Subsequently, the well was washed four times with TBS-T.
Elution of bound phages was conducted by adding 100 μl of 0.2 M glycine-HCl (pH 2.2) for ten minutes at RT. The solution was removed and subsequently added to a fresh reaction tube with 25 μl of 1 M Tris-HCl (pH 9.1) in order to neutralize the solution. Twenty microliters of the solution were then used for phage titer determination after elution (output titer). The remaining volume was used for amplification of the eluted phages.
Determination of the output titer and phage amplification was conducted according to the distributor´s manual. Shortly, 100 μl of phage dilutions ranging from 10 −2 to 10 −7 were mixed with 100 μl E. coli K12 ER2738 cells at an optical density of 0.6. The mixture was plated together with 800 μl top agar per dilution on LB/Tet/IPTG/XGal Petri dishes (40 x 10 mm) and incubated overnight at 37°C. The next day, plaques were counted and the titer was determined.
For phage amplification, 20 ml E. coli K12 ER2738 cell solution were grown to an optical density of 0.1 and incubated with the remaining volume of eluted phages (105 μl) for 4.5 hours at 37°C and 160 rpm. After incubation, the culture was centrifuged for 20 minutes at 2,300 x g at 4°C and 1 ml of the supernatant was removed and stored at 4°C. The remaining volume was precipitated overnight at 4°C with 7 ml PEG-8000 / 2.5 M NaCl. Subsequently, the solution was centrifuged for 60 minutes at 3,000 x g and 4°C whereupon the supernatant was discarded. After dissolving the pellet in 1 ml of 1x TBS, the sample was centrifuged for five minutes at 9,300 x g at 4°C. The supernatant was mixed with 200 μl PEG-8000/2.5 M NaCl, followed by 60 minutes incubation on ice and a final centrifugation step for 20 minutes at 16,000 x g and 4°C. The supernatant was removed and the pellet was resuspended in 100 μl of 1x TBS. The input titer was measured analogically to the output titration with dilutions ranging from 10 −8 to 10 −13 .
While the concentration of the target peptide remained stable, D-enantiomeric Aβ 1-42 oligomers and high molecular weight (HMW) aggregates/fibrils without an N-terminal biotin tag were added, starting from panning round two in increasing concentrations (round 2: 1 nMround 3: 5 nM-round 4: 10 nM-round 5: 50 nM-round 6: 500 nM) as will be explained below. Aβ 1-42 oligomers and HMW aggregates and fibrils were obtained from SEC and DGC, respectively, as described above. In order to obtain Aβ 1-42 HMW aggregates and fibrils fractions 12, 13 and 14 from the DGC were combined. Due to the addition of the competition step with Aβ 1-42 oligomers and HMW aggregates/, the protocol was adjusted starting from the second panning round as follows. The amplified phages from the previous round were diluted to 1x10 11 phages in 60 μl of 1x TBS and added to the well previously coated with SEC-derived, Denantiomeric, N-terminally biotinylated Aβ 1-42 monomers. Twenty microliters of SEC-derived Aβ 1-42 oligomers and 20 μl of DGC-separated Aβ 1-42 HMW / fibrils, each diluted in 1x TBS to the above mentioned concentration (e.g. for the second panning round 1 nM and 5 nM for panning round 3), were additionally added to the sample. After five minutes of incubation at RT, the solution was removed and the procedure went on with the aforementioned biotin competition step. The amount of washing steps increased with every panning round (4-6-8-10-12-15).

Single phage amplification
Single plaque forming units, each representing phages grown from one single clone, were picked from the output titer plates from round three to six and added to 5 ml of a E.coli K12 ER2738 culture at an optical density of 0.1 grown in LB/tetracycline medium and were then amplified for 4.5 h hours at 37°C at 160 rpm. Cultures were centrifuged for 20 minutes at 3,000 x g and 4°C and the supernatant was removed. Two milliliters were used for DNA extraction and kept at 4°C until usage, one milliliter was aliquoted for single phage ELISA and 0.5 ml were mixed with 0.5 ml 80% sterile glycerin as backup and stored at -80°C.

Single phage ELISA
The specificity and affinity of single phage clones towards the target peptide and the counterselective agents Aβ 1-42 oligomers and HMW aggregates / fibrils was analyzed by ELISA. 320 nM of SEC-derived N-terminally biotinylated D-Aβ 1-42 monomers or 320 nM SECderived N-terminally biotinylated D-Aβ 1-42 oligomers mixed with DGC-derived N-terminally biotinylated Aβ 1-42 HMW aggregates/fibrils were immobilized in duplicates for each single phage clone. Therefore, the total amount of Aβ 1-42 in each well was 150 ng. Amplified single phage clones from panning rounds three to six were analyzed. Additionally, the different species were immobilized to two wells each in order to check the immobilization efficiency by the Aβ-specific antibody 6E10 (Beta Amyloid 1-16 (6E10) monoclonal antibody, BioLegend, Dedham, MA, USA). Non-coated wells were used as additional controls for each phage. Additionally, buffer instead of phages was applied as control for cross reactivity of the anti M13:HRP conjugated antibody to the well surface and the immobilized target in duplicate.
The microtiter plates (Nunc 96-well Immobilizer Streptavidin microwell plate, Thermo Fisher Scientific Inc., Waltham, MA, USA) were pretreated as recommended by the manufacturer. The aforementioned Aβ 1-42 species were diluted in 1x TBS to 320 nM each and 100 μl per well were incubated for 15 minutes at RT with gentle shaking. After washing the wells twice with 150 μl of 1x TBS and blocking for one hour at RT with 1% (w/v) BSA in TBS-T, the plates were washed again three times with 150 μl of TBS-T. During the blocking step, the amplified phages and the buffer control (LB medium) were mixed with 1% (w/v) BSA in TBS-T in a ratio of 1:1 and incubated on a shaker at RT for 20 minutes. Afterwards 100 μl of the phage suspensions, the buffer control and the first antibody solution (6E10 1:1,000 in TBS-T) were given to the wells for one hour with gentle shaking at RT. After washing the plates for five times with 150 μl of TBS-T, 200 μl of TBS-T were added to each well and the plates were incubated at RT for another hour.
The supernatant was removed and 100 μl of the antibody dilutions were added to the adequate wells for one hour at RT (mouse anti bacteriophage M13 major coat protein (p8) HRP conjugated, GE Healthcare Europe GmbH, Freiburg, Germany and goat anti mouse IgG (H+L) HRP conjugated, Thermo Fisher Scientific Inc., Rockford, IL, USA, respectively). The anti-M13 antibody was diluted 1:5,000 in TBS-T and the goat anti mouse IgG antibody was diluted 1:1,000 in the same buffer. Subsequently, the plates were washed ten times with 150 μl of TBS-T and detection was conducted by measuring the conversion of the substrate 3,3',5,5'-tetramethylbenzidine (TMB) by HRP accompanied with a color change. Therefore, 50 μl of a TMB solution (one pill TMB was dissolved in 1 ml DMSO and diluted with 9 ml sterile filtered 0.05 M phosphate citrate buffer) was added to each well and color change was stopped by adding 50 μl of 2 M H 2 SO 4 when the solution in the wells started getting turquoise. The absorption at 450 nm was measured in a microplate reader. After subtraction of the buffer control, the absorption at 450 nm for the wells coated with Aβ 1-42 oligomers and fibrils were normalized to the Aβ 1-42 content of the Aβ 1-42 monomer-coated wells as determined by 6E10.

DNA extraction for sequencing
The single stranded phage DNA was purified as described in the NEB phage display manual (Instruction manual, version 2.7) with the exception that the iodide buffer was exchanged by a 10:1 mixture of 3M sodium acetate (pH 5.2) and TE buffer. Sequencing was conducted by GATC Biotech AG (Konstanz, Germany).

Transmission electron microscopy (TEM)
In order to analyze which Aβ 1-42 species have formed after incubation with the peptide Mosd1, TEM images were taken. Ten micromolar seedless Aβ 1-42 was incubated with or without 10 μM Mosd1 for 24 hours at RT. Twenty microliters of each sample were spotted on a formvar/carbon coated copper grid (Plano GmbH, Wetzlar, Germany) for three minutes. Afterwards the solution was detached with filter paper and the grids were washed three times with 20 μl of ddH 2 O and once with 5 μl 1% aqueous uranyl acetate. Then, 5 μl of the 1% uranyl acetate solution was applied to the grid for one minute for negative staining. The solution was removed and grids were dried overnight. The samples were analyzed with a Libra 120 transmission electron microscope (Carl Zeiss AG, Oberkochen, Germany) operating at 120 kV.

Quantitative determination of interference with Aβ 1-42 aggregate size distribution
The density gradient centrifugation method allows matrix-free separation and fractionation of different Aβ 1-42 species according to their sedimentation coefficients which depend on size and shape.
One-hundred microliters were given on the top of a discontinuous iodixanol gradient and spun in an ultracentrifuge as described above. Fourteen fractions of 140 μl each were harvested from top to bottom after centrifugation and the residual pellet (60 μl) was mixed with 60 μl of 6 M guanidine hydrochloride and boiled for ten minutes. This sample represents the 15 th fraction. The samples were analyzed by RP-HPLC and Tris-tricine SDS-PAGE followed by silver staining.
Peptides were quantified via isocratic reversed-phase high performance liquid chromatography (RP-HPLC). The column used was a Zorbax SB-300-C8 on a 1260 Infinity system (both Agilent Technologies Deutschland GmbH, Böblingen, Germany). Twenty microliters of each sample were injected and run with 1 ml per minute in an aqueous 30% (v/v) acetonitrile / 0.1% (v/v) trifluoric acid buffer as mobile phase and 80°C column temperature to denature Aβ species and separate them from other components, especially iodixanol. The signal was detected at an absorbance of 214 nm. The data were recorded and peaks were integrated using the Chem-Station software (Agilent Technologies). In order to calibrate the column, Aβ 1-42 solutions of known concentration were used to plot peak area versus Aβ 1-42 concentration. The plot equation then allowed calculation of Aβ 1-42 concentration within the samples.
Seedless Aβ 1-42 (200 μM in 10 mM sodium phosphate buffer) were incubated at 37°C and 600 rpm for three days in order to gain seeding-competent Aβ 1-42 fibrils. The sample was centrifuged at 14,000 x g for 45 minutes at 4°C in order to sediment fibrillary Aβ 1-42 content. The supernatant was removed and the remaining Aβ 1-42 fibrils were diluted in 100 μl 10 mM sodium phosphate buffer. After sonication for two minutes, concentration was determined by RP-HPLC. Aβ 1-42 seeds were incubated with or without a fivefold molar excess of Mosd1 for additional 24 hours. The mixture of Aβ 1-42 seeds and Mosd1 was again sonicated for two minutes and added to freshly prepared seedless Aβ 1-42 (15 μM) and 20 μM ThT. The final concentrations were 1.88 μM Aβ 1-42 seeds and 9.4 μM Mosd1. Each sample was added to a 96-well microtiter plate in triplicate and progression of fluorescence values was monitored over time at 24°C. The experiment was conducted independently for three times.
For each sample an asymmetric five parameter fit was used in order to determine the amplitude of relative fluorescence units and the half-life t 1/2 . After checking for Gaussian distribution by D´Agostino-Pearson omnibus normal test, significance was determined by one-way ANOVA.

Cell viability assay (MTT reduction)
PC-12 cells (DSMZ-Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany) were cultured in DMEM medium supplemented with 10% fetal calf serum, 1% antibiotics (Penicillin / Streptomycin) (all Sigma-Aldrich Corp., St.Louis, MO, USA) and 5% horse serum (PAA Laboratories GmbH, Pasching, Germany) on collagen Acoated (Biochrom GmbH, Berlin, Germany) tissue culture flasks (SPL Life Sciences Co., Korea) in a humidified incubator with 5% CO 2 at 37°C and grown for a maximum of 15 passages. Medium was changed every two days and cells were passaged, according to their confluence, every three to five days.
PC-12 cells were seeded in clear, collagen-coated 96-well flat bottom microwell plates (Life Technologies Inc., Carlsbad, CA, USA) at a density of 1x10 4 cells in a volume of 100 μl per well and incubated for 24 hours. Aβ 1-42 / Mosd1 mixtures were incubated as already described in the above mentioned assay for quantitative determination of interference with Aβ 1-42 aggregate size distribution. Mosd1 concentrations of 0/40/80 μM were used. From each well that contained cells 1.25 μl medium were removed and replaced with 1.25 μl of the preparations in order to gain a final concentration of 1 μM Aβ 1-42 and 0/1/0.5 μM Mosd1 in every well, respectively. Every approach was tested three times at least in quintuplicates. The plates were incubated for 24 hours in the incubator with 5% CO 2 at 37°C. Subsequently, 10 μl of the supernatant were taken out of every well and 10 μl of MTT reagent (Cell proliferation kit 1, Roche Diagnostics GmbH, Mannheim, Germany) were added for four hours of incubation in the incubator. Afterwards, 100 μl of solubilization reagent (Roche Diagnostics GmbH, Mannheim, Germany) were added to all wells and the plate was incubated overnight. The plate was shaken on a table shaker for five minutes to distribute the purple colored solution equally within each well and absorbance was measured at 660 nm and 570 nm with a microplate reader (POLARstar OPTIMA, BMG Labtech GmbH, Ortenberg, Germany). Reference values at 660 nm were subtracted from values for formazan absorbance (570 nm). The mean value of cell-free wells was subtracted as background. The arithmetic mean of all measurements per approach was calculated. Results are represented as the percentage of MTT reduction, assuming that the absorbance of control cells was 100%. Cells treated with 0.1% TritonX-100 in 10 mM sodium phosphate buffer (pH 7.4) served as control for dead cells. After analysis for Gaussian distribution by Shapiro-Wilk test, the values were tested for significance using the Mann-Whitney U test.

Western blot to investigate γ-secretase inhibition
Neuro-2a cells (DSMZ, Braunschweig, Germany) and Neuro-2a cells stably transfected with human APP695 were cultured in DMEM with 10% fetal calf serum and 1x non-essential amino acids (both Sigma-Aldrich Corp., St.Louis, MO, USA) in tissue culture flasks in a humidified incubator with 5% CO 2 at 37°C. Fifty thousand cells per well were seeded in 500 μl in a 24 well microwell plate for 24 hours at 37°C and 5% CO 2 to assure attachment to the surface. A stock solution of 1 mM Mosd1 was prepared in sterile water and the γ-secretase inhibitor DAPT was diluted to 5 mM in DMSO, respectively. The dilutions were spun at 1,000 x g for one minute prior to use. Then 0/10/100 μM Mosd1 and 1 μM DAPT were added to the cell supernatant and DMSO as well as water were tested as control. The cells were incubated at 37°C and 5% CO 2 for 24 hours. The cells were analyzed with a laser scanning microscope LSM 710 (Carl Zeiss AG, Oberkochen, Germany). Afterwards, the cells were washed three times with 1x PBS and lysed with 60 μl of NP-40 lysis buffer for 15 minutes. The lysates were resuspended, transferred to reaction tubes and centrifuged for five minutes at 12,000 x g. Afterwards, the supernatants were transferred to new reaction tubes and protein concentration was determined via microBCA assay.
After gel electrophoresis, the proteins were transferred via semi-dry blotting to a PVDF membrane. Protein transfer took place at 1 A and 25 V for 30 minutes. The membrane was blocked with 5% non-fat dry milk powder in 1x TBS-T at 4°C overnight with gentle shaking. Subsequently, the membrane was washed three times with TBS-T at RT and gentle shaking. The primary antibody for the detection of C-terminal fragment of APP (anti-amyloid precursor protein, C-terminal (751-770) rabbit pAb, Merck KGaA, Darmstadt, Germany) was diluted 1:5,000 in TBS-T. Simultaneously, the anti-β-actin antibody (β-actin (8H10D10) mouse mAb, Cell Signaling Technology, Inc., Danvers, MA, USA) was diluted 1:1,000 in TBS-T and both antibodies were given to the membrane, which was incubated overnight at 4°C on a roller mixer, subsequently. The membrane was washed three times with TBS-T and incubated with the secondary antibodies goat anti-rabbit IgG:HRP diluted 1:5,000 in TBS-T and goat antimouse IgG:HRP diluted 1:10,000 in TBS-T (both Santa Cruz Biotechnology, Inc., Dallas, Texas, USA) for three hours at RT on a roller mixer. After three washing steps with TBS-T, the membrane was incubated with the substrate for enhanced chemoluminescence (SuperSignal West Dura Chemoluminescent Substrate, Thermo Fisher Scientific, Inc., Rockford, IL, USA) for five minutes at RT in the dark. The signal was detected with the ChemiDoc MP gel documentation system (Bio-Rad Laboratories, Inc., Hercules, CA, USA).

Results
N-terminally biotinylated, SEC-derived Aβ 1-42 monomers were immobilized to streptavidincoated microwell plates. To avoid selection of plastic binding ligands, different plastic surfaces were used in each round. In particular, polystyrene, polypropylene and polycarbonate served as surface, alternating every round. Additionally, in every second round, the surface was blocked with BSA. Due to this procedure, six different combinations of blocking and surface type were used during the panning rounds and not a single combination occurred twice. This step was important, because only very low concentrations of Aβ 1-42 were immobilized in order to assure its monomeric state. Thus, there was a substantial risk that uncoated surface may have a strong bias during selection. In the first panning round, only Aβ 1-42 monomers were offered to the naïve library of phages. For the next round, amplified phages from the previous round were added together with increasing concentrations of non-biotinylated counterselective Aβ -42 species (SEC-derived oligomers and DGC-derived HMW aggregates and fibrils). After panning, a biotin competition step was conducted to remove phages that compete with biotin. Afterwards the well was washed every round with increasing rigor and still bound phages were eluted and amplified.
We verified the state of all used Aβ conformers by DGC followed by Tris-tricine-SDS-PAGE and silver staining. SEC-derived Aβ 1-42 monomers, SEC-derived oligomers, and fibrillary Aβ 1-42 were analyzed. As shown in Fig 1, prepared monomers are indeed free of any aggregates, as only the most upper fractions 1 and 2 of the DGC gradient contained Aβ. Oligomers can be found predominantly in fractions 4 to 6, and in smaller amounts also in neighboring fractions. Fibrils can be found in the extreme high molecular weight fractions, but obviously, during the DGC run, a substantial fraction of monomers dissociated from them and can be found in fractions 1 and 2.
Single phage clones from round three to six were amplified and sequenced. Sixteen sequences occurred more than once and were therefore tested for binding to monomeric Aβ 1-42 , Aβ 1-42 oligomers and HMW aggregates / fibrils as well as non-coated wells in a single phage ELISA. Additionally, several clones with uniquely occurring sequences and four clones from a previously conducted mirror image phage display were tested. This former mirror image phage display was conducted with almost identical conditions. The experimental setup differed only in two details. Firstly, the counterselective Aβ 1-42 HMW aggregates and fibrils were derived from an iodixanol gradient with different iodixanol concentrations (30-24-18-12-6 and 3% iodixanol instead of 50-40-30-20-10 and 5%). Secondly, the neutralization of the eluate took place in the same well as the elution step. Fig 2 shows the immobilization efficiency of different Aβ 1-42 species detected by Aβ 1-42 specific antibody 6E10. On all plates used for single phage ELISA, Aβ 1-42 monomers and oligomers/fibrils were immobilized in an amount which allowed normalization of the Aβ 1-42 oligomer/fibril content to the content of Aβ 1-42 in Aβ 1-42 monomer-coated wells as determined by 6E10. As shown in Fig 3, the competitive mirror image phage display selection indeed yielded phages that display peptides that specifically bind to Aβ 1-42 monomers as demonstrated by the strong binding to Aβ 1-42 monomer coated wells and notably less binding to wells coated with Aβ 1-42 oligomers and HMW aggregates or fibrils. We picked the clone with the highest signal intensity for monomeric Aβ 1-42 , but also the highest specificity for monomeric Aβ 1-42 as deduced from the high ratio of binding to Aβ 1-42 monomers and binding to oligomeric and fibrillary Aβ 1-42 . The amino acid sequence of the chosen clone 5.60 was YSYLT-SYHMVWR. The corresponding D-enantiomeric peptide, named Mosd1 (monomer specific d-peptide 1), was synthesized from purely D-amino acids with its C-terminus amidated, and was further characterized.
In order to obtain information on how Mosd1 changes Aβ 1-42 aggregation, TEM pictures were taken after 24 hours incubation of Aβ 1-42 with and without an equimolar ratio of Mosd1. The Aβ 1-42 sample without Mosd1 contained a mesh of fibrils (Fig 4A). Additionally, Aβ oligomers and large aggregates (dot-like structures) were present, displaying a broad range of aggregated Aβ 1-42 species. In the sample with Aβ 1-42 and Mosd1 neither fibrils nor oligomers were observed. Instead, amorphous, unstructured aggregates were the major species found on the grid (Fig 4B).
After incubation of 80 μM Aβ 1-42 followed by coincubation with different concentrations of Mosd1 (0/10/20/40/80 μM), samples were loaded on a discontinuous iodixanol gradient and  centrifuged. Fifteen fractions were harvested from top to bottom, applied to Tris-tricine-SDS-PAGE and silver stained. With this experiment we were able to show that Mosd1 abolishes different Aβ 1-42 species from the sample, which contains Aβ 1-42 species of different size, ranging from monomers to fibrils and HMW aggregates.
The distribution of 80 μM Aβ 1-42 , incubated for 4.5 hours without addition of Mosd1, showed a broad range of species distributed from fraction 1 to 12, representing monomers, small and large oligomers, protofibrils, fibrils and HMW aggregates (Fig 5A) [37]. The amount of Aβ 1-42 as estimated from silver staining decreased with the fraction number, indicating substantial amounts of Aβ 1-42 in form of monomers and small oligomers. The most toxic oligomers can be expected in fractions four to six [38]. The addition of Mosd1 shifted the distribution of Aβ 1-42 towards large aggregates (Fig 5B-5E). This aggregates are suspected to be nontoxic [39]. The Aβ 1-42 contents in fractions one to ten were decreased with increasing Mosd1 concentrations. We therefore conclude that Mosd1 decreased the amount of toxic Aβ 1-42 oligomers and shifted the distribution equilibrium towards non-toxic aggregates in a concentration dependent manner.
Silver staining does not allow a quantitative determination of the detected species. In order to gain quantitative information about Aβ 1-42 concentrations in every fraction, samples of each fraction were applied to RP-HPLC chromatography and analyzed for their Aβ 1-42 contents, as described before [40]. The QIAD (quantitative determination of interference with Aβ 1-42 aggregate size distribution) assay fully confirmed the above described behavior of Mosd1 and Single phage ELISA. The relative binding affinity of single phage clones from mirror image phage display (clone numbers 6.xx) and a previously conducted mirror image phage display (clone numbers 5.xx) to SEC-derived biotinylated Aβ 1-42 monomers, oligomers and fibrils as well as the non-coated wells was analyzed. The M13 phage-specific antibody was used for detection. Transformation of substrate by the antibody-conjugated HRP was measured at 450 nm. Amplified single phage clones were added to wells coated with 150 ng Aβ 1-42 monomers (red) or 150 ng Aβ 1-42 oligomers and fibrils (1:1; green) or to wells only coated with streptavidin (blue). Cross reactivity of the M13 phage-specific antibody was tested in an approach without addition of phages. After background subtraction of the anti M13 antibody values, the values for phage to Aβ 1-42 oligomer/fibril binding were normalized to the values of phage to Aβ 1-42 monomer binding according to the outcome of coating efficiency controls with the Aβ specific antibody 6E10. demonstrates that Aβ 1-42 oligomer elimination by Mosd1 is concentration dependent (Fig 6). Most importantly, Mosd1 is able to almost completely eliminate Aβ 1-42 oligomers.
A characteristic property of fibrillar Aβ to is its ability to act as a seed for conversion of monomeric Aβ into fibrillar Aβ assemblies. We therefore analyzed the impact of Mosd1 on  fibril seeded fibrillation of Aβ 1-42 using the Thioflavin T (ThT) assay (Fig 7). Fibrillar Aβ 1-42 seeds were grown and afterwards incubated with or without Mosd1. ThT was added and the mixture was added to freshly prepared seedless Aβ 1-42 . The progression of ThT fluorescence was monitored over time. Compared to freshly prepared seedless Aβ 1-42 (t 1/2 = 19.69 h), addition of fibrillary Aβ 1-42 seeds significantly reduced the time needed to increase fibrillar content (t 1/2 = 0.57 h). When fibrillary Aβ 1-42 seeds were coincubated with Mosd1 prior to addition of freshly prepared seedless Aβ 1-42 , the time for the increase of fibrillary content was significantly reduced as well, yet the time for the increase was significantly longer (t 1/2 = 3.36 h) compared to the sample without Mosd1 (Fig 7A). Furthermore, the amount of fibrillar Aβ 1-42 content in the samples, displayed by the amplitude of relative fluorescence units (Fig 7B), did differ significantly between seedless Aβ 1-42 (set to 100% RFU) and seedless Aβ 1-42 incubated with fibrillary Aβ 1-42 seeds (84%) as well as between seedless Aβ 1-42 and fibrillary Aβ 1-42 seeds coincubated with Mosd1 (74%).
In order to confirm that Mosd1 yields Aβ 1-42 species that are not toxic to cells and Mosd1 can indeed abolish the cytotoxic effect of Aβ 1-42 and rescue cell viability, an MTT assay was performed (Fig 8). PC-12 cells were incubated with Aβ 1-42 with or without different concentrations of Mosd1, derived from the same incubation method as mentioned above. Mosd1 without Aβ 1-42 was tested to ensure that Mosd1 itself is not toxic to the cells. TritonX-100 was used as   in each DGC fraction were determined quantitatively via RP-HPLC. Samples were loaded to a Zorbax 300SB-C8 column connected to a 1260 Infinity HPLC system. Separation of the samples was achieved by elevated column temperature (80°C) and an isocratic mobile phase of 30% acetonitrile/0.1% TFA in water. The averaged concentration of Aβ 1a control for cell death. Untreated cells were set 100% viable and all other values were normalized to this value. Incubation with 0.1% TritonX-100 resulted in 1.6% viable cells. The Aβ 1-42 composition led to a significantly decreased cell viability of 45% and was therefore proven to be toxic. Mosd1 in a concentration of 1 μM did not alter MTT reduction in this assay and showed no toxic effect in PC-12 cells. When coincubated with Mosd1, the toxic effect of Aβ 1-42 was significantly decreased with increasing concentrations of Mosd1. In an equimolar ratio, 86% of the cells stayed viable and with half the concentration of Mosd1 still 66% of the cells were viable. Mosd1 is therefore able to rescue PC-12 cells from Aβ 1-42 induced toxicity.
Neuro-2a cells, stably transfected with human APP695, develop a pathologic phenotype when compared with wild type Neuro-2a cells. Wild type Neuro-2a cells accumulate and grow protrusions between each other and appear polygonal, whereas Neuro-2a cells, which are stably transfected with human APP, appear isolated, spindle shaped and show less protrusions and cell contacts (Figs 9 and 10). Applied to wild type Neuro-2a cells, Mosd1 did not alter their physiological phenotype. The cells clotted and developed protrusions and appeared polygonal (Fig 9). Added to hAPP695-transfected Neuro-2a cells, Mosd1 compensated the pathological phenotype (Fig 10). The cells developed connections and protrusions, grew denser and did partially clot, resembling the phenotype of wild type Neuro-2a cells. This effect is dose dependent, Because Mosd1 binds Aβ 1-42 monomers, it might also bind to the respective part of APP and influence its cleavage by γ-secretase. In order to investigate this potential side activity, we analyzed whether Mosd1 influences γ-secretase activity with APP as a substrate. Neuro-2a cells, stably transfected with the gene coding for human APP695 were incubated with 10 and 100 μM Mosd1, respectively. The cells were lysed, lysates were separated by SDS-PAGE, blotted and the APP C-terminal fragment β (APP CTFβ) was detected by enhanced chemoluminescence. Both concentrations of Mosd1 showed no signs of γ-secretase activity inhibition, i.e. the γ-secretase substrate APP CTFβ was not accumulated as it is the case for the treatment with the γ-secretase inhibitor DAPT (Fig 11) [41,42].

Discussion
The competitive mirror image phage display yielded Aβ 1-42 monomer specific ligands Our competitive mirror image phage display led to single phage clones that preferentially bind to monomeric Aβ  . Sequencing of single phage clones revealed several consensus sequences which occurred multiply. Additionally, ELISA results showed an overall specificity Competitive Mirror Image Phage Display Derived Peptide of most single phage clones for monomeric Aβ 1-42 compared to Aβ 1-42 oligomers and fibrils and non-coated wells. Several single phage clones exhibited a very strong difference in binding affinities between monomeric and aggregated Aβ 1-42 and also the non-coated surface. We picked the most promising clone and investigated the respective D-peptide Mosd1 for further analysis.

Mosd1 modulates Aβ 1-42 aggregation
Via TEM and quantitative determination of Aβ 1-42 aggregate distribution, we were able to show that coincubation of Aβ 1-42 with Mosd1 leads to an altered aggregation pathway of Aβ  . While incubation of Aβ 1-42 alone for 24 hours led to oligomers and fibrillar structures, coincubation with Mosd1 led to large amorphous aggregates that were shown to be non-toxic [30, 39,40]. By combining these results with the experiment regarding interference of Mosd1 with Aβ 1-42 aggregate size distribution mentioned above, we suggest the following mode of action. Mosd1 stabilizes Aβ 1-42 monomers and thereby modulates the dynamic equilibrium of Aβ 1-42 species. Toxic Aβ 1-42 oligomers are decomposed and the resulting Aβ 1-42 species being precipitated and converted into amorphous non-toxic aggregates instead of fibrils. According to Ladiwala et al., Mosd1 can be sorted into class I molecules, which convert soluble Aβ oligomers into large, non-toxic conformers [43]. In addition, Mosd1 with 33% residues being aromatic shares the overall aromatic features of compounds, which belong to this class. Indeed, also polyphenols like resveratrol and derivatives or polyphenolic flavones like kaempferol-3-Orhamnoside are able to cause the development of large, non-toxic, off-pathway Aβ aggregates Competitive Mirror Image Phage Display Derived Peptide as shown in TEM, AFM and SDS-PAGE studies by several research groups [44,45]. Additionally, Mosd1 is able to decelerate Aβ 1-42 aggregation. Therefore, Mosd1 is able to reduce the impact of already formed Aβ 1-42 seeds on monomeric Aβ 1-42 and their conversion into toxic aggregates as shown by Seeding ThT Assay.
According to cell culture experiments, aggregates generated by coincubation of Aβ 1-42 with Mosd1 are not toxic to PC-12 cells, whereas, the mixture of Aβ 1-42 species not treated with Mosd1 showed significant cellular toxicity. Other compounds, which consist of aromates or include aromatic side chains, were also reported to reduce Aβ induced cell toxicity [43,45,46]. This might be contributed to the binding to Aβ 1-42 monomers and prevention of their aggregation as well as to the conversion of toxic oligomeric Aβ 1-42 species into large non-toxic species as shown in the latter experiments.
Additionally, we were able to show that human APP695 transfected Neuro-2a cells, treated with Mosd1, appear healthier compared to untreated cells, which develop a pathological phenotype. Mosd1 is able to reverse the effects of human APP695 expression and its cleavage products like reduced cell contacts and a lack of cellular protrusions.
Despite the fact that Mosd1 reduces Aβ 1-42 toxicity by modulating the equilibrium of Aβ 1-42 species, it needs to be assured, that the compound is safe and does not interfere with related but physiological relevant pathways. For example, γ-secretases are involved in the cleavage of APP, leading subsequently to Aβ. Moreover, γ-secretases are also involved in cleavage of transmembrane Notch and therefore interfere with Notch signaling pathways. Alterations within the Notch pathway can lead to severe side effects [47,48]. According to our results Mosd1 has no effect on γ-secretase activity. Given the usefulness of mirror image phage display to obtain stable D-peptides for synthetically available targets, we wonder why there have been only a limited number of reports describing such approaches [49][50][51][52] and those described in reviews [26,34,53].

Conclusion
Taken together, we have established a novel competitive mirror image phage display for the selection of D-enantiomeric peptides, which bind specifically to monomeric Aβ 1-42 . Using non-biotinylated SEC-derived oligomers and DGC-derived HMW aggregates and fibrils of Aβ 1-42 as counterselective agents, we were able to enrich phages, which bind specifically to the immobilized biotinylated monomeric Aβ 1-42 .
One of our selected D-enantiomeric peptides, Mosd1, shows promising characteristics in several in vitro experiments. This leads to the assumption that Mosd1 is able to stabilize Aβ 1-42 monomers, interferes with seeded growth and modulates Aβ 1-42 aggregation towards nontoxic, amorphous aggregates and therefore rescues cells from Aβ 1-42 derived toxicity and reverses pathological phenotypes from hAPP-transfected neuronal cells.
Mosd1 also does not affect γ-secretase function which makes it safer and more precise than γ-secretase modulating compounds. Its small size should facilitate blood-brain-barrier transfer and the D-enantiomeric conformation enables high proteolytic stability. Further investigations including affinity studies and epitope mapping should provide more information about the mechanism how Mosd1 modulates Aβ  .
We were able to show that a competitive mirror image phage display is a straightforward method to select compounds, which are Aβ 1-42 monomer specific and able to modulate Aβ 1-42 aggregation towards non-toxic species and therefore exhibit high therapeutic potential.