All-D-enantiomers of beta-amyloid exhibit similar biological properties to all-L-beta-amyloids.

The amyloidogenic peptide beta-amyloid has previously been shown to bind to neurons in the form of fibrillar clusters on the cell surface, which induces neurodegeneration and activates a program of cell death characteristic of apoptosis. To further investigate the mechanism of Abeta neurotoxicity, we synthesized the all-D- and all-L-stereoisomers of the neurotoxic truncated form of Abeta (Abeta25-35) and the full-length peptide (Abeta1-42) and compared their physical and biological properties. We report that the purified peptides exhibit nearly identical structural and assembly characteristics as assessed by high performance liquid chromatography, electron microscopy, circular dichroism, and sedimentation analysis. In addition, both enantiomers induce similar levels of toxicity in cultured hippocampal neurons. These data suggest that the neurotoxic actions of Abeta result not from stereoisomer-specific ligand-receptor interactions but rather from Abeta cellular interactions in which fibril features of the amyloidogenic peptide are a critical feature. The promiscuous nature of these beta-sheet-containing fibrils suggests that the accumulation of amyloidogenic peptides in vivo as extracellular deposits represents a site of bioactive peptides with the ability to provide inappropriate signals to cells leading to cellular degeneration and disease.

The amyloidogenic peptide ␤-amyloid has previously been shown to bind to neurons in the form of fibrillar clusters on the cell surface, which induces neurodegeneration and activates a program of cell death characteristic of apoptosis. To further investigate the mechanism of A␤ neurotoxicity, we synthesized the all-D-and all-Lstereoisomers of the neurotoxic truncated form of A␤ (A␤ [25][26][27][28][29][30][31][32][33][34][35] ) and the full-length peptide (A␤ 1-42 ) and compared their physical and biological properties. We report that the purified peptides exhibit nearly identical structural and assembly characteristics as assessed by high performance liquid chromatography, electron microscopy, circular dichroism, and sedimentation analysis. In addition, both enantiomers induce similar levels of toxicity in cultured hippocampal neurons. These data suggest that the neurotoxic actions of A␤ result not from stereoisomer-specific ligand-receptor interactions but rather from A␤ cellular interactions in which fibril features of the amyloidogenic peptide are a critical feature. The promiscuous nature of these ␤-sheet-containing fibrils suggests that the accumulation of amyloidogenic peptides in vivo as extracellular deposits represents a site of bioactive peptides with the ability to provide inappropriate signals to cells leading to cellular degeneration and disease.
Alzheimer's disease (AD), 1 vascular dementia, and hereditary cerebral hemorrhage with the Dutch type are diseases that share an invariant pathological feature, the accumulation of an amyloidogenic peptide into insoluble fibrillar extracellular deposits. In all three cases, the major component of the extracellular debris is the ␤-amyloid peptide (A␤) that is derived from the proteolytic processing of the large membraneanchored amyloid precursor protein (APP) encoded by a single gene located on chromosome 21 (1). However, the biological significance of these amyloid deposits has been extensively debated as to whether they are a causative factor of each disease or merely a metabolically inert end product lacking in biological activity. Evidence in support of a causative role for A␤ in neuropathology comes from genetic analysis of the APP gene where several autosomal dominant mutations have been linked with AD and hereditary cerebral hemorrhage with the Dutch type (2,3). In a recent in vitro study, the ␤-APP 717 mutation consistently caused a significant increase in the percentage of the longer and more amyloidogenic A␤ 1-42 over the shorter A␤   (4). Incorporation of this same mutation into a transgenic mouse model yields A␤ deposition and neuropathology that closely parallels that observed in AD (5). Additional in vivo evidence suggesting that the A␤ peptide itself may be biologically active comes from a transgenic model overexpressing A␤ 1-42 , in which A␤ transgene expression was detected in a variety of peripheral tissues but histopathological changes were restricted to the brain. Moreover, the neurodegeneration was largely limited to the cerebral cortex, hippocampus and amygdala, all areas affected in AD, and was essentially undetectable in the cerebellum, which is typically not affected in AD (6). Finally, the amount of ␤-amyloid that accumulates in the brain appears to correlate well with the decline of brain function (7).
Insights into the inherent biological activity associated with A␤ have come from in vitro studies that show synthetic A␤ can spontaneously assemble into ␤-sheet-containing fibrils (8,9), that this fibrillar-A␤ can induce neuritic dystrophy in neuronal cultures similar to that seen in the AD brain, and that the mechanism of A␤-triggered degeneration is via programmed cell death (PCD) (10). These observations have led to the general hypothesis that the biological activity of A␤ is dependent on its transformation into a highly stable protease-resistant antiparallel ␤-sheet conformation and higher order quaternary assemblies (11) similar to those found in senile plaques. This is of fundamental importance because it suggests that the biological activity of A␤ is dependent on protein conformation and the transition into this conformation. The consequences of such a relationship between biological activity and protein conformation are critical to understanding the role of A␤ and other ␤-pleated sheet protein assemblies such as prion protein in disease.
In order to understand the degenerative processes induced by A␤, it is essential to define the characteristics of A␤ salient to its function as a neurotoxic stimulus. In a previous study, we used a series of synthetic A␤ peptides with progressively truncated C-termini to demonstrate that the length of this hydrophobic region is a crucial determinant of peptide ability to both aggregate and induce neurotoxicity in vitro (12). We have also synthesized a series of truncated A␤ peptides to examine the effects of N-terminal heterogeneity, which occurs in vivo on the assembly and biological activity of A␤. The N-terminal truncated isoforms produced enhanced aggregation into neurotoxic ␤-sheet fibrils, which suggests that these truncated peptides may initiate the pathological neurodegeneration in AD by acting as a nucleation site for A␤ deposition (13). Thus far, we have observed that assembled, bioactive A␤ peptides exhibit ␤-sheet structure and that amino acid substitutions that dis-rupt A␤ assembly also prevent ␤-sheet structure and abolish toxicity (14). Further analysis of peptides will be useful in elucidating the specific requirements for both the assembly and bioactivity of A␤.
Since numerous ligand-target interactions are stereospecific, one means to both examine the nature of the A␤-cellular interactions and assess the validity of several proposed mechanisms of A␤-induced cell death is to determine whether A␤ bioactivity exhibits stereospecificity. Similar issues of ligand stereospecificity have been investigated in several recent studies by comparing the binding and or activities of D-and L-enantiomers of small peptide ligands (15,16). In the current study, we have utilized a comparable paradigm, synthesizing the all-Dand all-L-amino acid stereoisomers of the truncated biologically active A␤ [25][26][27][28][29][30][31][32][33][34][35] and the full-length A␤ 1-42 peptide, and compared their physical and biological properties to determine whether the interaction of A␤ with biologically relevant cells is stereospecific.
Circular Dichroism-The mean residue ellipticity of A␤ 25-35 peptides (25 M in 5 mM potassium phosphate, pH 7.3) was determined using a Jasco J-720 spectropolarimeter equipped with a computerized data processor, as described previously (14). Samples were loaded into a 1.0-cm path length quartz cell and measured over a 190 -250-nm wavelength range at 0.5 nm increments. Data from eight scans were averaged and subtracted from base-line values but otherwise are unsmoothed. The instrument was calibrated with a 0.06% (w/v) solution of d-camphorsulfonate.
Electron Microscopy-For ultrastructural analysis, 25 M samples of A␤ peptides (20 mM MOPS buffer, pH 7.4) were adsorbed onto 200 mesh formvar grids and stained with 2% uranyl acetate prior to viewing with a Zeiss 10CR transmission electron microscope at 80 kV transmission (14).
Tissue Culture-Cultures of hippocampal neurons from gestational day 18 rat pups were prepared as described previously (12). Cultures were plated at 2.5 ϫ 10 4 cells/cm 2 on poly-L-lysine-treated multiwell plates and maintained in serum-free DMEM supplemented with N-2 components. After two days in vitro, cultures were exposed to the various A␤ peptides for 24 h, after which cell viability was determined on the basis of trypan blue exclusion (12,18). Raw data were statistically compared by analysis of variance followed by Scheffé f-test.

RESULTS
The initial experiments to investigate whether an all-D-enantiomer of an amyloidogenic peptide retains biological activity were performed with all-D-A␤ [25][26][27][28][29][30][31][32][33][34][35] . This is the smallest commonly studied fragment of A␤ that retains both the ability to form ␤-sheet-containing fibrils and neurotoxicity (19). An additional advantage of A␤ [25][26][27][28][29][30][31][32][33][34][35] is that it can be modeled relatively easily, and information on the alignment of the antipa-rallel strands as well as the importance of various side chain interactions in formation of the A␤ fibrils and in defining the surface topography necessary for neurotoxicity can be investigated. In the antiparallel ␤-sheet conformation, the surface topography is determined by the amino acid sequence and the alignment of adjacent strands in the ␤-sheet. In a computergenerated model of A␤ [25][26][27][28][29][30][31][32][33][34][35] in which the two peptides are maximally overlapped, a cluster of positively charged lysine residues is observed on one face of the ␤-sheet while the other face of the ␤-sheet contains primarily hydrophobic residues. Previous studies with A␤ 25-35 containing single amino acid substitutions have confirmed the importance of the sequence for retaining the properties of A␤ (14).
The biological properties of the two enantiomers were then tested by applying the enantiomers to primary cultures of rat hippocampal and cortical neurons that have previously been used to assay the neurotoxic activity of A␤ (12,18). The all-D-A␤ [25][26][27][28][29][30][31][32][33][34][35] produced visible aggregates in the tissue culture wells and appeared to bind to the surface of neurons equally as well as the all-L-A␤ [25][26][27][28][29][30][31][32][33][34][35] (Fig. 2, A-D). Noticeable neuronal degeneration was apparent at 12 h, and extensive cell death was observed at 24 h for both enantiomers. In order to compare the levels of neurotoxicity between the two enantiomers, a dose response curve was generated. As can be seen in Fig. 2E, the all-D-A␤ [25][26][27][28][29][30][31][32][33][34][35] produced similar toxicity to the all-L-A␤ 25-35 over the entire range of concentrations tested. The specificity of the neurotoxicity was determined by analyzing peptides with scrambled sequences of both the all-L and the all-D enantiomers. Neither scrambled sequence produced detectable neurotoxicity over the entire range of concentrations utilized for the dose response curve (data not shown).
During the comparison of computer-generated models of A␤ [25][26][27][28][29][30][31][32][33][34][35] in antiparallel ␤-sheet conformation, we discovered, that with perfect alignment of the individual strands of peptide, that a pseudo-axis of symmetry was generated do to the planar nature of the ␤-sheet such that the distribution of the surface groups produced topochemically similar enantiomers. Other cases of topochemically similar peptides that bind to stereoselective receptors and posses similar activities have been reported (20,21). Based on the high level of bioactivity associated with the all-D-A␤ 25-35 , we next determined whether the all-D-enantiomer of the full-length A␤ 1-42 peptide would also bind to cells and produce similar neurotoxicity to the all-L-A␤ 1-42 . Although the ␤-sheet-containing fibrils of the A␤ 1-42 are predicted to form planar sheets, the longer length of the A␤ 1-42 peptide would reduce the probability of formation of topochemically similar enantiomers since the surface topography of A␤ 1-42 would be far more complex than with A␤ [25][26][27][28][29][30][31][32][33][34][35] .
Highly purified all-D-A␤ 1-42 was subjected to CD, and the spectra were compared with the all-L-A␤ 1-42 (Fig. 3A). The all-D-enantiomer produced the expected mirror image spectra, indicating similar secondary structure for the enantiomers. The peptides were then examined by electron microscopy, and, while the filamentous structures were different from those observed with A␤ 25-35 , the A␤ 1-42 enantiomers produced fibrils that were indistinguishable from each other (Fig. 3, B and C). The ability to bind certain dyes, such as Congo red and thioflavine T, is a characteristic property of amyloidogenic peptides (22)(23)(24) and can be used to measure the amount of peptide in ␤-sheet-containing fibrils (25). Analysis of assembled peptides of both enantiomers indicates that the all-D-A␤ 1-42 binds thio-flavine with intensity equal to that of the all-L-A␤ 1-42 (data not shown).
The biological response to the full-length all-D-A␤ 1-42 by neurons was assayed as described above for A␤ [25][26][27][28][29][30][31][32][33][34][35] . The all-D-enantiomer clearly binds to the neurons since clusters of the fibrillar A␤ 1-42 can be seen over much of the cell surface within 6 h of adding the peptide to the cultures (Fig. 4, C and D). Perhaps more importantly, the fibrillar clusters induce extensive neurodegeneration over a 24-h time course. It should be noted that not all neurons respond equally to fibrillar forms of A␤, as previous studies have shown that neurons that are immunopositive for GABA are resistant to the neurotoxic activity of A␤ (26). A comparison of the relative neurotoxicity of

DISCUSSION
This study was designed to probe the stereospecificity of the interaction between A␤ and the plasma membrane of cultured neurons that in vitro leads to programmed cell death (10,27,28). According to classic receptor pharmacology, a D-stereoisomer of an amino acid or peptide would not be predicted to exhibit bioactivity comparable with the native L-peptide. For example, glutamate receptors readily discriminate L-versus D-antagonistic agents (29). We have analyzed both the physical and biological properties of the all-D-enantiomers of A␤ [25][26][27][28][29][30][31][32][33][34][35] and A␤  and have compared them with their corresponding all-L-enantiomers. With the exception of the CD spectropolarimetry study, which produced mirror image spectrums, both all-D-enantiomers exhibited essentially identical physical and biological properties to their all-L-enantiomers.
A number of studies have been done utilizing D-enantiomers of various ligands to investigate the stereospecific requirements for binding to their respective receptor proteins. In the cases of three peptide hormones, bradykinin (30), oxytocin (31), and angiotensin (32), that must interact with chiral receptors on the plasma membrane, the all-D-forms of the ligands were inactive. However, different results were obtained in the case of a synthetic ␤-endorphin analog that contained 18 D-amino acid residues in the C-terminal portion of the peptide but 5 Lresidues within the actual binding site. The all-D-containing region was designed to form a left-handed amphiphilic helical segment that was topochemically similar to the native righthanded amphiphilic helix. The D/L chimeric peptide retained equal ability to bind and to activate the opiate receptor (21). In some cases, the all-D-peptide enantiomers can still resemble the parent compound, both in the overall spatial arrangements and with respect to the electronic nature of the functional groups. In the case of the antibiotic enniatin B, the topochemically similar enantio-enniatin B possessed similar antimicrobial activity (20). In two recent studies for example, the all-Dpeptide analogs were found to bind with similar affinities to their respective receptors. In the first one, two all-D-amphiphilic helical peptides were shown to interact with calmodulin in a sterically malleable fashion (16,33), and the second example reported that the laminin segment containing the IKVAV amino acid sequence, which is responsible for cell attachment and tumor-promoting activities, was retained in the all-D-peptide. Peptide analogs with either alternating D-L-substitutions or randomized IKVAV sequence were inactive, indicating that the sequence and conformational status of the domain contribute to the biological activity but that no stereospecific requirement exists (15).
Although bilayer lipids and membranes are also chiral and contain numerous asymmetric centers, the partitioning of chiral channel-forming antibiotic peptides into membranes does not require a specific chirality. The all-D-analogs of cecropin, magainin II amide, and melittin were equally effective when tested on achiral synthetic planar bilayers and as antibiotics against bacteria containing chiral membranes (34). The reports that A␤ 1-40 forms giant multivalent cation channels when incorporated into synthetic bilayers (35,36) and our findings that the neurotoxic activity associated with both A␤ [25][26][27][28][29][30][31][32][33][34][35] and A␤  are not chirally dependent are consistent with the results obtained with the channel-forming antibiotic peptides. Unfortunately, after many years of extensive study on A␤, there is no definitive evidence that A␤ forms channels in neurons. Another more likely possibility is that A␤ is active as a membrane perturbant, which may alter the microenvironment between the bilayer and membrane-bound enzymes or receptors.
We currently favor a mechanism dependent on the interaction of A␤ with membrane receptor proteins on the surface of neurons, and other cell types such as astrocytes, because assembled fibrillar forms of the amyloidogenic peptides are required for activity (12,14,37). It is possible, for example, that A␤ acts as a ligand to cross-link receptors at the cell surface and activates cell death pathways via activation-induced cell death similar to Fas (38). Consistent with a mechanism involving membrane receptors, we have shown that the lectin ConA, which forms clusters of membrane glycoproteins on the cell surface, also causes neurodegeneration and apoptotic death in cultured neurons similar to that observed with A␤ while succinyl ConA, which binds but does not cross-link, is inactive (39). Recently, Burdick, et al. (40) have shown that a substantial portion of the A␤ that binds to cells can be removed by treatment with trypsin and several receptors that appear to bind A␤ peptides have been identifed (41)(42)(43)(44)(45)(46). In the case of the receptor for advanced glycation end products (RAGE), some evidence has been presented that it may be directly involved in A␤induced neurotoxicity (45). Experiments with all-D-A␤ and these putative A␤ receptors are in progress and should provide information on the specificity of these receptors for A␤. Thus, we suggest that extracellular macromolecular assemblies such as A␤ can serve as stimuli or agonists that trigger a particular sequence of cellular reactions in neurons that initiate an apoptotic program of cell death. These PCD agonists are characterized in part by ␤-sheet fibrillar structure but, in addition, have the common ability to access critical signal transduction and downstream mechanisms that drive PCD.
Other amyloidogenic proteins, which do not share sequence homology with A␤ (e.g.. prion and amylin), do form structurally similar extracellular deposits and have been found to have similar neurotoxic activity in vitro (47,48). One possible mechanism that could explain the common biological activity seen with different amyloidogenic peptides is ␤-sheet augmentation, whereby a peptide forms a "peptide-surface association" (49) either by inserting itself into a ␤-sheet-containing domain (50), which has been proposed for other diseases involving protein conformational changes (51)(52)(53)(54), or by adding to the edge of an anti-parallel ␤-strand, as has been implicated in regulating protein associations governing signal transduction pathways and assembly interactions in certain viral capsids (49). The ␤-sheet augmentation mechanism provides much greater flexibility than classical domain-domain association because the peptide is not constrained by a rigidly folded domain. The specificity in this model is dependent on the ability of the peptide to augment an appropriate ␤-strand on a protein. Consistent with this model, A␤ has shown a pronounced ability to bind to other proteins, such as ␣-1-antichymotrypsin, and transthyretin which are rich in ␤-sheet. Finally, the cell surface contains numerous proteins with Ig superfamily homology with extensive ␤-sheet content, which include receptors (55) and cell adhesion molecules (i.e. NCAM and N-cadherin) (56), and in several reports, the cell surface appears to be able to actually nucleate A␤ assembly (40,57,58), which is also consistent with the model (49).
The fact that the all-D analogs of A␤ retain bioactivity may present new avenues for therapeutic intervention by allowing the D-enantiomers of inhibitory peptides to be utilized. An approach similar to this has recently been used to identify an all-D-amino acid opioid peptide with analgesic activity capable of crossing the blood brain barrier using a synthetic combinatorial library made up of D-amino acid hexapeptides (59). In addition, the identification of D-peptide ligands through mirror image phage display using genetically encoded libraries (60) offers the promise of rapidly screening for D-peptide ligands that can block assembly and or neurotoxicity of the A␤ peptide. All-D-ligands are generally resistant to proteolysis and D-amino acid proteins are reported to have low immunogenicity, thus making them useful for pharmacological applications (61).
The results obtained in this study suggest that the neurotoxic activity of A␤ is independent of the classical stereoisomerspecific ligand-receptor interaction. Rather, A␤-induced neurotoxicity is dependent on the primary sequence of the peptide that regulates both the ability of the peptides to assemble into active conformations and bind to cellular surfaces. While mechanisms dependent on the perturbation of cellular membranes or the formation of calcium ion channels cannot be excluded, the requirement for higher order protein assemblies by amyloidogenic peptides for biological activity does not readily support these mechanisms. Further investigation of the mechanism of A␤-induced toxicity will likely benefit attempts to understand the neurodegeneration that occurs in AD and perhaps other amyloid-related disorders.