The Anti-Amyloidogenic Action of Doxycycline: A Molecular Dynamics Study on the Interaction with Aβ42

Abstract

The pathological aggregation of amyloidogenic proteins is a hallmark of many neurological diseases, including Alzheimer's disease and prion diseases. We have shown both in vitro and in vivo that doxycycline can inhibit the aggregation of Aβ42 amyloid fibrils and disassemble mature amyloid fibrils. However, the molecular mechanisms of the drug's anti-amyloidogenic property are not understood. In this study, a series of molecular dynamics simulations were performed to explain the molecular mechanism of the destabilization of Aβ42 fibrils by doxycycline and to compare the action of doxycycline with those of iododoxorubicin (a toxic structural homolog of tetracyclines), curcumin (known to have anti-amyloidogenic activity) and gentamicin (an antibiotic with no experimental evidence of anti-amyloidogenic properties). We found that doxycycline tightly binds the exposed hydrophobic amino acids of the Aβ42 amyloid fibrils, partly leading to destabilization of the fibrillar structure. Clarifying the molecular determinants of doxycycline binding to Aβ42 may help devise further strategies for structure-based drug design for Alzheimer's disease.

Keywords: Alzheimer’s disease; amyloid-beta protein; curcumin; doxycycline; iododoxorubicin; molecular dynamics.

Figure 1

Chemical structures of (a) 4′-iodo-4′-deoxydoxorubicin (IDOX) and (b) doxycycline.

Figure 2

Molecular models of the amyloid Aβ42 fibrils and the small molecules tested. For the simulations, we considered two different polymorphs of Aβ42 fibril structures, with five peptides of the 2MXU PDB entry (a) and five peptides of the 5OQV PDB entry (b). Structures of candidate anti-amyloidogenic molecules: doxycycline (c), IDOX (d), curcumin (e) and gentamicin (negative control, f).

Figure 3

Molecular dynamics (MD) simulations in the absence and presence of the small molecules. Root-mean-square deviation (RMSD) calculated using the C_α of each residue (only one trajectory per systems is shown) (a,c). The root-mean-square fluctuations (RMSFs) (b,d) were measured using the C_α of each residue of the central peptide, excluding the first 200 ns of the MD simulations, and were averaged over triplicate simulations.

Figure 4

Average contact time between anti-amyloidogenic molecules and the two different polymorphs of Aβ42 fibrils, 2MXU (a) and 5OQV (b) during the MD simulation. The contact time is the amount of time (as a percentage) during which the molecule is in contact with the fibril. The 5 ns window is meant to exclude sporadic contacts due to the high simulated concentration.

Figure 5

Doxycycline binding sites on Aβ42 fibrils. Superposition of doxycycline coordinates (center of mass) during the MD simulations (excluding the first 200 ns) in the presence of the two polymorphs of Aβ42 fibrils (2MXU, (a); 5OQV, (f).) Panels (b) and (g) show the main interaction sites. When simulated with the 2MXU structure, doxycycline stably bound to the side chain of M35 (c), within the groove formed by I32 and L34 (d) and within the groove formed by F19 and L17 (e). In the simulations with the 5OQV structure, doxycycline stably interacted with the hydrophobic groove formed by V39 and I41 (h), as well as with the groove formed by F20 and V18 (i).

Figure 6

Iododoxorubicin binding to (a–d) 2MXU and (e,f) 5OQV fibril structures. Superposition of iododoxorubicin coordinates (center of mass) during the 2MXU MD simulations (a,b) and snapshots of the main binding regions (c,d). Superposition of IDOX coordinates during the 5OQV MD simulation (d) and a snapshot of the main binding area (f).

Figure 7

Curcumin binding to (a,b) 2MXU and (c–g) 5OQV Aβ fibrils. Superposition of curcumin coordinates (center of mass) during the 2MXU MD simulations (a) and a snapshot of the main binding area (b). Superposition of curcumin coordinates during the 5OQV MD simulation (c) and snapshots of the main binding area (d–g).