Zinc-induced heterodimer formation between metal-binding domains of intact and naturally modified amyloid-beta species: implication to amyloid seeding in Alzheimer’s disease?

Zinc ions and modified amyloid-beta peptides (Aβ) play a critical role in the pathological aggregation of endogenous Aβ in Alzheimer’s disease (AD). Zinc-induced Aβ oligomerization is mediated by the metal-binding domain (MBD) which includes N-terminal residues 1–16 (Aβ1–16). Earlier, it has been shown that Aβ1–16 as well as some of its naturally occurring variants undergoes zinc-induced homodimerization via the interface in which zinc ion is coordinated by Glu11 and His14 of the interacting subunits. In this study using surface plasmon resonance technique, we have found that in the presence of zinc ions Aβ1–16 forms heterodimers with MBDs of two Aβ species linked to AD: Aβ containing isoAsp7 (isoAβ) and Aβ containing phosphorylated Ser8 (pS8-Aβ). The heterodimers appear to possess the same interface as the homodimers. Simulation of 200 ns molecular dynamic trajectories in two constructed models of dimers ([Aβ1–16/Zn/Aβ1–16] and [isoAβ1–16/Zn/Aβ1–16]), has shown that conformational flexibility of the N-terminal fragments of the dimer subunits is controlled by the structure of corresponding sites 6–8. The data suggest that isoAβ and pS8-Aβ can be involved in the AD pathogenesis by means of their zinc-dependent interactions with endogenous Aβ resulting in the formation of heterodimeric seeds for amyloid aggregation.


Introduction
The key event in pathogenesis of Alzheimer's disease (AD) is associated with amyloid-beta (Aβ) polypeptide, which undergoes transition from a soluble (physiologically normal) monomeric form into neurotoxic dimers and oligomers; this finally results in the accumulation of Aβ aggregates with characteristic supramolecular structure (amyloid plaques) in specific brain regions (Masters & Selkoe, 2012). Zinc ions play a critical role in the pathological aggregation of Aβ in vivo (Bush, 2013). Although intact Aβ spontaneously aggregates in vitro in the presence of zinc ions (Alies, Hureau, & Faller, 2013), the resultant Aβ aggregates are non-toxic in vivo (Meyer-Luehmann et al., 2006). In contrast to intact Aβ, its chemically and/or structurally modified variants found in the amyloid plaques of AD patients may act as seeds for the formation of pathological aggregates of Aβ (Meyer-Luehmann et al., 2006). However, the molecular mechanism responsible for conversion of the native conformation of endogenous Aβ into the pathological one (induced by the modified Aβ variants) remains unknown (Cummings, 2004).
Recently, it has been demonstrated that Aβ phosphorylated at Ser8 (pS8-Aβ) promotes toxic aggregates formation during the pathogenesis of AD (Kumar et al., 2011(Kumar et al., , 2012. We have found that in vivo administration of Aβ containing isoAsp7 (isoAβ), the most common modified Aβ variant detected in the amyloid plaques (Roher et al., 1993), accelerates progression of cerebral amyloidosis in transgenic mice . IsoAβ was also neurotoxic for neuronal cell cultures (Mitkevich et al., 2013). In this context, it is important that both the pathogenically relevant modifications of Aβ (Ser8 phosphorylation and Asp7 isomerization) as well as so-called the 'English' familial mutation (H6R) associated with the inherited form of AD (Hori et al., 2007;Janssen et al., 2003;Ono, Condron, & Teplow, 2010) are localized in the metal-binding domain (MBD) of Aβ. This domain comprising the N-terminal residues 1-16 (Kozin, Zirah, Rebuffat, Hoa, & Debey, 2001;Tsvetkov et al., 2010;Zirah et al., 2006) is a good model for the analysis of interactions between zinc ions and Aβ: in contrast to the full-length Aβ, MBD does not aggregate in aqueous solutions at physiological pH Kozin et al., 2001;Zirah et al., 2006).
In this study using a surface plasmon resonance (SPR) technique we have demonstrated for the first time that in the presence of zinc ions intact Aβ MBD (Aβ 1-16 ) forms stable heterodimers with Aβ MBD containing isoAsp7 (isoAβ 1-16 ) and also with Aβ MBD containing phosphorylated Ser8 (pS8-Aβ 1-16 ). Using an available structure of the Aβ MBD homodimer carrying the English mutation H6R [H6R-Aβ 1-16 /Zn/H6R-Aβ 1-16 ] (PDB code 2MGT) as a prototype we have generated spatial models of the homodimer [Aβ 1-16 /Zn/Aβ 1-16 ] and the heterodimer [isoAβ 1-16 /Zn/Aβ 1-16 ]. Simulation of 200 ns molecular dynamic trajectories in two constructed models of the dimers have shown that conformational flexibility of the N-terminal fragments of the dimer subunits is controlled by the structure of corresponding sites 6-8.

Materials
Synthetic peptides (purity > 98% by RP-HPLC) corresponding to human Aβ MBD, its isoforms, and rat Aβ MBD (ratAβ 1-16 ) were purchased from 'Biopeptide Co., LLC' (USA) ( Table 1). All synthesized peptides used as ligands for immobilization contained the C-terminal tetraglycylcysteine tag. Various paired combinations of these peptides have been used in the SPR-based study in which one peptide was immobilized on the optical chip as a ligand, while the other was added as a soluble analyte interacting with the immobilized ligand. Standard optical chips CM5 and special reagents were purchased from GE Healthcare (USA). These included: HBS buffer (150-mM NaCl, 3-mM EDTA, .005% surfactant P20, 10-mM HEPES, pH 7.4); 10-mM acetate buffer (pH 4.5); a thiol coupling kit. Solution containing 10-mM HEPES (pH 6.8) and 100-μM ZnCl 2 was used as a working buffer for SPR analysis of zinc-dependent dimerization of MBDs of various Aβ isoforms.
of analyte solutions (Figure 1(A)) through the working and control channels at a flow rate of 5 μL/min for 5 min. During each experiment, at least five solutions with different analyte concentrations were used. All SPR measurements were repeated three times. After each measurement, the optical chip surface was regenerated by injecting HBS buffer for 30 s. Analyte samples in the concentration range from 2 to 20 μM were prepared in the working buffer containing 100-μM ZnCl 2 . In accordance with results of previous studies (Kozin et al., 2001(Kozin et al., , 2011, a series of pilot experiments without zinc ions in the working buffer or zinc ion replacement for other divalent ions (Ca 2+ , Mg 2+ , Cu 2+ and Fe 2+ ) confirmed a specific role of zinc ions for Aβ MBD oligomerization.
The obtained sensograms were analyzed by means of the standard software 'BIAevaluation v.4.1' employing different mathematical kinetic models based on global fitting of theoretical curves to the experimental sensograms.

Molecular dynamic simulation of the dimers
Molecular dynamic simulation has been performed using the 'GROMACS 4.6.5' software package (Van Der Spoel et al., 2005) and the 'Amber ff99SB-ILDN' force field (Lindorff-Larsen et al., 2010). This force field has an improved parameterization for the side-chain torsion angle potential of the aspartate residue, which is particularly important for our problem (Lindorff-Larsen et al., 2010). Implementation of the non-standard residue isoaspartate did not introduce new atom types for 'Amber ff99SB-ILDN'. All parameters for the bonded and nonbonded interactions needed for describing non-standard residue, except partial charges, have been already defined in the force field and used without modifications (Spitaleri et al., 2008). Partial charges for the new residue type have been calculated in Chimera as described above (Supplementary Table 1).
The starting model of the dimer has been placed in the cubic cell with a minimum distance between the protein and the box of .8 nm. TIP3P water molecules have been added to the system and Na + ions have been used to neutralize total charge (Jorgensen, Chandrasekhar, Madura, Impey, & Klein, 1983). The system was minimized using a steepest descent algorithm. Positions for the peptide atoms were restrained and the system was equilibrated with 100 ps of constant volume (NVT) molecular dynamic followed by 100 ps of constant pressure (NPT) molecular dynamic. Molecular dynamic trajectory of 200 ns was calculated using the NPT ensemble. Positions of the atoms for the Aβ residues 11 and 14 together with the coordinated zinc ion were restrained at all steps of simulation using corresponding harmonic potential implemented in 'GROMACS 4.6.5' and a force constant of 9999 kJ/(mol nm 2 ) for each of three dimensions. Calculations have been performed at temperature 300 K, pressure 1 bar, with 2 fs integration step using the Berendsen barostat and the velocity rescale method for the thermostat. The particle-mesh Ewald method (Darden, York, & Pedersen, 1993) has been implemented to treat long-range electrostatic interactions and the LINCS algorithm controlled lengths of covalent bonds (Hess, Bekker, Berendsen, & Fraaije, 1997).
Clustering analysis of the similarity of the conformations from the trajectory has been accomplished using g_cluster tool of 'GROMACS 4.6.5' (Van Der Spoel et al., 2005). Clustering has been performed separately for each peptide chain of the dimer using single linkage algorithm with the RMSD cut-off 1.5 Å. Conformations from molecular dynamic trajectories have been sampled with 100 ps step. RMSD have been calculated over all atoms of the peptide chain.

Biosensor analysis of the interactions between
MBDs of different variants of Aβ in the presence of zinc ions: SPR-based evidence for heterodimer formation Zinc-dependent interactions have been studied between the immobilized ligands (isoAβ 1-16 or pS8-Aβ 1-16 ), and the soluble analytes (isoAβ 1-16 , pS8-Aβ 1-16 and Aβ 1-16 ). Analysis of resultant sets of sensograms performed by means of the BIAevaluation software revealed the best correspondence with the Langmuir theoretical model of the 1:1 interaction (Figure 1(B), Table 2). The closeness of the fit is confirmed by the statistical value χ 2 , which represents the deviation of experimental sensogram from the fitted curve (Biacore Life Sciences, 2004). The lower the χ 2 value the higher the correspondence to the theoretical curve is. Among five theoretical models used for description of the most evident types of molecular interactions between immobilized ligand and added (free) analyte, the lowest χ 2 value was obtained only in the case of the Langmuir theoretical model of the 1:1 (Table 2).
Although the SPR technology can detect the fact of intermolecular interactions, stochiometry of ligandanalyte interactions can be evaluated not only by the best fitting to the theoretical model, but also by the R max /R im ratio. R im is the total amount of the immobilized ligand and R max is a maximum analyte binding capacity of the chip with the immobilized ligand. The R im value (in RU) is obtained from immobilization sensogram and the R max value (in RU) is calculated by the BIA evaluation software from global fitting and extrapolation of steady-state biosensor signals to total saturation of the chip. Table 3 shows that in all investigated pairs of ligand-analyte interactions the R max /R im ratio did not exceed 1. Lower values observed for some examined ligand-analyte pairs may be attributed to steric limitations for analyte binding.

Identification of the key MBD segment involved in formation of the zinc-induced heterodimers
Importance of various MBD sites for zinc-induced dimerization has been evaluated during analysis of interaction of immobilized pS8-Aβ 1-16 with added pS8-Aβ 1-16 or Aβ 1-16 or their fragments (Figure 2). Figure 2 shows that the immobilized pS8-Aβ 1-16 formed complexes only with peptides containing the fragment Aβ 11-14 (EVHH). This fragment was previously identified as the critical one for the formation of soluble zincbound pS8-Aβ 1-16 homodimers (Kulikova et al., 2014). For independent validation of the key role of the Aβ 11-14 site, we have also analyzed zinc-induced interaction between the immobilized Aβ 1-16 and the synthetic peptide corresponding to the rat Aβ MBD (ratAβ 1-16 ). It is well documented that rat Aβ MBD and human Aβ MBD differ by three amino acid residues (Istrate et al., 2012), one of which is located just in the segment 11-14 (Table 1). Figure 3 shows that ratAβ 1-16 did not interact with the immobilized Aβ 1-16 . This obviously means that replacement of only one amino acid residue in the 11-14 segment dramatically impairs (or even abolishes) heterodimer formation.
Analysis of the molecular dynamic trajectories revealed that fragments 1-10 are remarkably flexible in both intact Aβ 1-16 and isoAβ 1-16 peptide chains, however,  the normal Aβ 1-16 chain is more mobile (Figure 4,  Supplementary Figure 1). Dimer structures relax during the first nanosecond of simulation, as reflected by the initial rapid increase of the RMSD (Figure 4). During the rest part of the trajectories RMSD fluctuates near the average values (average RMSD over the 2-200 ns part of the trajectory for the homodimer: .35 nm for chain A and .43 nm for chain B; for the heterodimer: .51 nm for the isoAβ 1-16 chain and .41 nm for the Aβ 1-16 chain). These fluctuations correspond to the conformational transition of the 1-10 fragment of the peptide chains ( Figure 5). Fluctuations have higher amplitude and are more frequent for the intact chains of the dimers indicating that the normal Aβ 1-16 chain is more flexible than the isoAβ 1-16 chain. This finding is supported by the cluster analysis. In the case of the isoAβ 1-16 chain 54.5% of all conformations sampled from the trajectory with 100 ps step are grouped in three large clusters, while for the intact chain of homodimer overall cluster heterogeneity is higher (Supplementary Figure 2).
Structured bended fragments stable along the trajectory have been found in the region of residues 6-8 of the intact peptide chains Aβ 1-16 (Figure 6(A)). Structure of the bend is stabilized by the hydrogen bond between HO-group of S8 side-chain and carboxyl group of D7 side-chain ( Figure 6(A), Supplementary Figure 3).  Elongation of the isoAβ 1-16 backbone with the extra CH 2 group disrupted hydrogen bond between side chains of the residues 7 and 8. The corresponding fragment 6-8 of the isoAβ 1-16 demonstrates a strong propensity toward extended conformations (Figure 6(B)) and is more flexible (Table 5).
Earlier, it was demonstrated that the fragment including residues 11-14 is a primary site for recognition and binding of zinc ion by the intact Aβ (Tsvetkov et al., 2010;Zirah et al., 2006). This site is critically important for the formation of the zinc-bound interface in the homodimers formed by Aβ 1-16 (Kozin et al., 2011), pS8-Aβ 1-16 (Kulikova et al., 2014), and H6R-Aβ 1-16 (Kozin et al., 2015). The same fragment constitutes the zincinduced interface in the homodimers formed by the peptides corresponding to intact Aβ residues 11-18 . Taking into consideration all these data, it is reasonable to conclude that the fragment 11-14 plays a key role not only in the molecular mechanism of zincdependent homodimerization but also in heterodimerization of different human Aβ variants.
Simulation of 200 ns molecular dynamic trajectories for the homodimer [Aβ 1-16 /Zn/Aβ 1-16 ] and the heterodimer [isoAβ 1-16 /Zn/Aβ 1-16 ] demonstrated some differences in conformational behavior of the fragments 1-10. The region including residues 6-8 of the intact peptide adopted stable (along the trajectory) conformation of the bend, while corresponding fragment of isoAβ 1-16 existed in the extended state ( Figure 6(A), (B)). Our analysis of molecular dynamic trajectories suggests that the structure of the conformational motif 6-8 modulates motions of the peptides. Bended conformation of the 6-8 site in Aβ 1-16 yields higher flexibility for the intact peptide chain (Figure 4(A)), while extended conformation of the same site in isoAβ 1-16 makes peptide chain more rigid (Figure 4(B)). These variations in the protein chain dynamic properties can significantly influence the zincmediated oligomerization process for different Aβ forms: more rigid structures are more susceptible for oligomerization (Parker et al., 1999).
Driving forces responsible for the formation of neurotoxic Aβ dimers and oligomers in AD still remain unknown (Jucker & Walker, 2013;Prusiner, 2012). On the one hand, Aβ oligomers isolated from both soluble and insoluble fractions of AD postmortem brain tissue (Larson & Lesné, 2012) are detergent resistant; on the other hand, synthetic analogs of intact Aβ are not amyloidogenic in vivo (Meyer-Luehmann et al., 2006). Assuming that the behavior of isolated MBDs may be extrapolated to the full-length Aβ, it is reasonable to propose the following scenario in AD: occasionally appeared pathologically altered Aβ molecule forms zincinduced heterodimer with a native Aβ molecule, which adopts altered conformation. Acting as pathogenic seeds, these heterodimers cause further propagation of the pathological processes resulting in Aβ-induced neurotoxocity and/or initiation of AD. Chemically modified Aβ variants (Meyer-Luehmann et al., 2006), particularly, isoAβ , may well serve as putative candidates responsible for such amyloid seeding in this disease.

AD
Alzheimer's disease Aβ amyloid-beta MBD metal-binding domain pS8-Aβ amyloid beta with the phosphorylated Ser8 residue isoAβ amyloid-beta with the isomerized Asp7 residue SPR surface plasmon resonance

Supplementary material
The supplementary material for this paper is available online at http://dx.doi.org/10.1080/07391102.2015.1113890. .61 Notes: RMSD has been calculated over the family of 200 conformations from the molecular dynamic trajectory taken with 1 ns step. Structures of each family have been superimposed over N, Cα and C atoms of the same residues 6, 7 and 8.