Chemistry of Alzheimer's Disease

Early Ideas

Early ideas about the chemistry


Membranes, Radicals, and Beta Sheets

A basic polyunsaturated membrane

The amyloid beta peptide can damage neuronal membranes by lipid peroxidation and by generating non-ion-specific pores.  The most toxix species is not the beta amyloid monimer but rather a soluble oligomeric form, dimer, trimer, etc. Lipid peroxifation requires an unsaturated membrane and an initialing radical that can penetrate hydrocarbon interior.  The brain is rich in polyunsaturated lipids.  The penetrating radical should be generated outside the membrane in an aqueous environment and be long-lived enough and hydrophobic enough to enter the bilayer.  Recognizing that removal of the alpha-C hydrogen atom from any amino acid residue would generate a captodatively stabilized C-centred radical, we examined the possibilities of generating such a radical in a bet asheet, and which residues would be the most susceptible to alpha-C-H oxidation.  The answer is a glycine residue in a beta sheet.

Model of Cu(II) binding to Abeta

What about the Copper?

What is the affinity of Abeta for Cu(II) and Cu(I)?

Computations agree with experiment: for Cu(II)/Abeta, DG = -36 kJ/mol

The is some disagreement re binding of Cu(I): we calculate DG = -73 kJ/mol

How is the copper bound?

The graphic shows a model of the major binding motif.  Cu(II) is bound to all three His residues, the carbonyl between H13 and H14, and the carbonyl of Ala2, in a trigonal bipyramidal arrangement.  As many as 3 other pH-dependent structures have been proposed.

Cu(I) is tightly bound to His13 and His14 through the delta N atoms of the imidazole groups, in a linear dicoordinated configuration.

What is its reduction potential of Cu(II)/Abeta?

Computations agree with experiment: Eo(Cu(II)/Abeta, Cu(I)/Abeta) = 0.31 eV

The reduction is slow, requiring two higher energy intermediate structures.

TS for H abraction by Met(+) from Gly

Is Met35 really involved?

Early literature says "YES"

Can it be oxidized by Abeta-bound Cu(II)?

Computations say "not directly".  Either a Cu(I)-generated hydroxyl radical is involved, or a second Met.  The latter suggests that at least a dimer must be involved.

Can the oxidizrd Met35 abstract a hydrogen atom from a Glycine backbone?

Computations say "not easily", and they must be close to each other in a beta sheet.

The graphic shows a model transition structure for H atom abstraction by Met(+) from a Gly in a beta sheet.  The early calculations indicated a gas phase activation energy of 49 kJ.mol.

If glycine, which glycine?

Abeta ultimately oligomerizes as a beta sheet.  We showed that in a beta sheet, only a glycine residue is susceptible to oxidation by H atom abstraction, and that an S-oxidized Met can effect the transformation.  Of the six Gly residues, computations pointed to Gly29 and Gly33 as the most likely targets for oxidation on energetic grounds.  Experimentally, the G33V mutant of Ab exhibits considerably reduced neurotoxicity (Kanski, J.; Varadarajan, S.; Aksenova, M. Butterfield, D. A. Biochim. Biophys. Acta, 2002, 1586, 190-198).

The Radical Model of Alzheimer's

Radical Model

The Radical Model of Alzheimer's

The Radical Model proposes that beta amyloid first aggregates into a soluble oligomer with antiparallel beta sheet structure and bearing a bound Cu(II).  The amyloid-bound copper(ii) species generates a methionine sulfidyl radical either by direct electron transfer or indirectly by a hydroxyl radical.  The methionyl radical in turn generates a captodatively stabilized alpha-C-centred radical at a glycyl residue, which can be long-lived enough to infiltrate into the interior of a polyunsaturated and oxygen-rich neuronal membrane.  The membrane-insinuated alpha-C-centred glycyl radical then abstracts a H-arom to generate a more stable pentadienyl radical, which will react with oxygen to initiate lipid peroxidation.  If the glycyl radical first intercepts an oxygen molecule before or after membrane insertion, the resultant peroxy species can also initiate lipid peroxidation by the same mechanism.

Those Neurotoxic Oligomers

Consensus is that the most neurotoxic species of beta amyloid are soluble aggregates (oligomers), rather than the monomeric form, or fibrular forms that make up the senile plaques that are one of the hallmarks of AD.  While it is clear that amyloid fibrils possess a cross, in sync, antiparallel structure, oligomers appear to have an antiparallel beta sheet configuration.  Toxicity has been shown to increase with extent of aggregation.  Beyond that, little is known with certainty about the structures of oligomers, or their mechanism of toxicity.

The beta amyloid dimer (Ab(1-42)2)

Tjernberg showed that the site of Ab most likely to initiate aggregation is the central hydrophobic core, L17VFFA21.  Various fragments of Ab containing  this core aggregate as antparallel beta sheets and many ultimately form fibrils. We carried out long-duration (about 25 microsec in total), molecular dynamics simulations and found that the most stable conformation of the Ab(1-42) dimer is that shown at the right.

The most stable Ab(1-42) dimer

Ab(1-42) dimer from MD simulations

The most stable structure of the beta amyloid dimer derived from MD simulations. The blue and teal correspond to one strand, the red and pink to the other. The central region of the pink and teal is the LVFFA core in antiparallel configuration


Basic requirements

Design strategy
Example of SMD-US methodology

Example of SMD-US methodology Steered Molecular Dynamics is used to define the reaction coordinate. Then at selected points, a harmonic potential is applied and the system is equilibrated for 20 - 40 ns. Weighted histogram analysis (WHAM) is used to tease out the more accurate reaction coordinate and associated free energy changes.

Example of results from SMD-US methodology

Example of results from SMD-US methodology. Besides the black PMF curve with error bars, we also monitor H-bond count (blue curve), and separations of salt bridges (red and green curves).

Beta Sheet Inhibitors


If oligomeric Ab is neurotoxic, and it aggregates as beta sheets, could one design molecules that can bind specifically to Ab and inhibit the beta sheet formation?  That was the question that Samir Roy sought to answer in his PhD thesis research.  At the time this work was initiated, Ab(1-42) was far too large and flexible to be studied by ab initio computational means, or even empirical MD simulations.  As a model for full length Ab, we selected Ab(13-23), which, for short, we called Rec (for recognition site).  This fragment encompasses the central hydrophobic core and includes three charged residues, K16, E22, and D23 which could be used to enhance binding affinity of the inhibitor.  It also include two His residues, H13 and H14, which are the primary binding sites for Cu(II) and Cu(I).  A strategic choice was to adopt a peptidic strand for the inhibitor as well, in order to maximize its selectivity for Rec (i.e., Ab) over the rest of the mish-mash of proteins that comprise a real biological system.

The design strategy

Having adopted a peptide platform for the inhibitor, the following requirements must be met:

  1. The length is 8 residues. Any larger and the immune system would destroy it.
  2. Use D-amino acids, and/or pseudopeptides to render it resistant to enzymatic degradation.
  3. Select residues of complimentary polarity: oppositely charged residues adjacent to each other and matching non-polar residues for the hydrophobic core.  The oppositely charged residues may be used to direct the complexation to parallel or antiparallel beta sheets.
  4. Add methyl groups to the backbone to inhibit propagation of the beta sheet across the “inhibitor”

Four classes of inhibitor were designed by docking to Rec with the MOE semi-flexible docking software:

  1. SGA: L-residues which bind as an antiparallel beta sheet
  2. SGB: D-residues which bind as an antiparallel beta sheet
  3. SGC: L-residues which bind as a parallel beta sheet
  4. SGD: D-residues which bind as a parallel beta sheet

The methodology

The few best candidates from the MOE docking results in each class were subjected to MD simulations, using the GROMACS software, the Gromos96 53a5 forcefield, SPC water, in a 6 x 6 x 6 nm3 cube with periodic boundary conditions, at 1 atm pressure and 310 K.  After first equilibrating the complex between Rec and the Inhibitor, steered molecular dynamics were applied to establish a reaction coordinate for dissociation by slowly pulling the two fragments apart to complete separation.  Then 30 “windows” equally spaced along the path were selected and subjected to Umbrella Sampling.  The difference between the high and low points along the potential of mean force curve were taken as the free energy difference between bound and unbound structures, i.e., the binding affinity, DGR-S.  The procedure applied to the Rec dimer yielded a binding affinity, DGR-R = 53 kJ/mol.  The inhibitor, S should bind maximally to R, i.e., DGR-S > DGR-R, and minimally to itself (DGS-S as small as possible).

A measure of effectiveness of inhibition is given by DGeff =  2 DGR-S – DGR-R – DGS-S

The results

Selected pseudopeptides (S)  in each class are listed in the table below.  A positive value for DGeff indicates that the peptide should be an effective beta sheet blocker (and prevent oligomerization of Ab).


Conclusions of the inhibitor study

The free energy changes in the table below correspond to stoichiometric amounts.  One sees that SGB1 and SGC1 should make the best beta sheet inhibitors, although all classes may be effective if one can adjust the concentrations.

The ultimate challenge

Our last project was to modify SGC1 by adding a chelator with a high affinity for Cu(I).  The modifies inhibitor we called TGC1.  TGC1 has the additional property that the oxidation potential of its Cu(I) complex is too low to permit the generation of ROS. TGC1 should not only inhibit beta sheet formation but also remove any bound Cu from Abeta and prevent the formation of ROS. In the next section we describe this research.

Peptide Structure







SGA3 N-Acetyl-Daba1-Orn2-MeLeu3-Phe4-MePhe5-Leu6-Ala7-Glu8-NH2    56  46 3
SGB1 N-Acetyl-daba1-orn2-leu3-mephe4-phe5-mephe6-leu7-glu8-NH2 62 45 26
SGC1 N-Acetyl-Glu1-Ala2-MePhe3-Phe4-MePhe5-Leu6-Orn7-Daba8-NH2   53 26 27
SGD1 N-Acetyl-glu1-leu2-mephe3-phe4-mephe5-leu6-orn7-daba8-NH2 50 32 15

where the lower case designates the D-amino acids and Daba = diaminobutyric acid, Orn = ornithine,

MeLeu = N-methylleucine, MePhe = N-methylphenylalanine. All these pseudo-peptides mimic the

R section of the Ab peptide; having complimentary charged residues between the hydrophobic

core. The all L-amino acid SGA and SGC bind to R in antiparallel and parallel mode, respectively.

The all-D-amino acid SGB and SGD bind to R in an antiparallel and parallel mode, respectively.

Data from: Banafsheh Mehrazma, Stanley Opare, Anahit Petoyan and Arvi Rauk, D-Amino Acid Pseudopeptides as Potential Amyloid-Beta Aggregation Inhibitors, Molecules 2018, 23, 2387-2410; doi:10.3390/molecules23092387


Search for a Cu(I) chelator

Ab initio methods using the CAMB3LYP hybrid functional were applied to search for possible Cu(I)/Cu(II) chelators.  The criterion was free energy of binding.  Geometry and ZPVE and thermal corrections (change of enthalpy to 298K and entropy at 298 K, were determined with the 6-31+G(D) basis set with final zero point enthalpy calculated at the 6-311+G(2df,2p) level.  Solvation was added by computation with the 6-31+G(d) basis set and SCRF = IPCM. 

A variety of potential ligands, based on amines, thioethers, and imidazoles, were explored.  The best ligand assembly, shown at the right involved a pyridine scaffold and two imidazoles forming a three-point chelation framework.  For Cu(II), the fourth ligand was a water molecule.  At the para position of the pyridine was placed a glutamic acid residue (the N-terminal residue of SGC1).

TGC1 - Rec Complexes, with and without Cu(I)

It remained to be proved that modification of SGC1 to TGC1, or attachment of Cu to Rec or to TGC1, did not disrupt or weaken its binding to Rec (and by extension, to Abeta).  Indeed, equilibration by MD showed that all three combinations, TGC1-Rec, Cu(I)/Rec-TGC1, and Cu(I)/TGC1-Rec formed parallel beta sheets as expected.  The Cu(I)/TGC1-Rec complex is shown at the right. The red backbone is that of Rec.

SMD_US PMF Analysis for TGC1-Rec-Cu(I)

The SMD_US procedures described above showed that the attachment of Cu(I) to either Rec or TGC1 enhanced the binding energy, which was greatest for Cu(I)/TGC1-Rec.  The PMF curve is shown at the right,



Optimum ligand assembly for Cu(II)/Cu(I) chelation, and properties

Cu chelator

The final choice for Cu(I)/Cu(II) chelator. The binding affinities are sufficiently high for this complex to extract copper from Abeta. The reduction potential is low enough to prevent the reduction of oxygen to superoxide.

Complex between Cu/TGC1 and Rec

The complex between Cu-bound TGC1 and Rec

The complex between Cu-bound TGC1 and Rec

Dissociation PMF for Cu/TGC1-Rec

Binding of Cu/TGC1 to Rec

Binding of Cu/TGC1 to Rec