Chelator PBT2 Forms a Ternary Cu2+ Complex with β-Amyloid That Has High Stability but Low Specificity

The metal chelator PBT2 (5,7-dichloro-2-[(dimethylamino)methyl]-8-hydroxyquinoline) acts as a terdentate ligand capable of forming binary and ternary Cu2+ complexes. It was clinically trialed as an Alzheimer’s disease (AD) therapy but failed to progress beyond phase II. The β-amyloid (Aβ) peptide associated with AD was recently concluded to form a unique Cu(Aβ) complex that is inaccessible to PBT2. Herein, it is shown that the species ascribed to this binary Cu(Aβ) complex in fact corresponds to ternary Cu(PBT2)NImAβ complexes formed by the anchoring of Cu(PBT2) on imine nitrogen (NIm) donors of His side chains. The primary site of ternary complex formation is His6, with a conditional stepwise formation constant at pH 7.4 (Kc [M−1]) of logKc = 6.4 ± 0.1, and a second site is supplied by His13 or His14 (logKc = 4.4 ± 0.1). The stability of Cu(PBT2)NImH13/14 is comparable with that of the simplest Cu(PBT2)NIm complexes involving the NIm coordination of free imidazole (logKc = 4.22 ± 0.09) and histamine (logKc = 4.00 ± 0.05). The 100-fold larger formation constant for Cu(PBT2)NImH6 indicates that outer-sphere ligand–peptide interactions strongly stabilize its structure. Despite the relatively high stability of Cu(PBT2)NImH6, PBT2 is a promiscuous chelator capable of forming a ternary Cu(PBT2)NIm complex with any ligand containing an NIm donor. These ligands include histamine, L-His, and ubiquitous His side chains of peptides and proteins in the extracellular milieu, whose combined effect should outweigh that of a single Cu(PBT2)NImH6 complex regardless of its stability. We therefore conclude that PBT2 is capable of accessing Cu(Aβ) complexes with high stability but low specificity. The results have implications for future AD therapeutic strategies and understanding the role of PBT2 in the bulk transport of transition metal ions. Given the repurposing of PBT2 as a drug for breaking antibiotic resistance, ternary Cu(PBT2)NIm and analogous Zn(PBT2)NIm complexes may be relevant to its antimicrobial properties.


Introduction
The compound 5,7-dichloro-2-[(dimethylamino)methyl]-8-hydroxyquinoline (PBT2) is a terdentate Cu 2+ and Zn 2+ chelator that was previously trialed as a therapeutic to treat Alzheimer's disease (AD). Its anticipated mechanism of action was based on the controversial "metals hypothesis", which proposed that AD results from aberrant interactions of the β-amyloid (Aβ) peptide with endogenous transition metal ions, notably Cu 2+ , causing Aβ to misfold and generate reactive oxygen species (ROS) [1,2]. PBT2 was proposed to prevent these interactions, sequester transition metal ions thought to be trapped within extracellular β-amyloid aggregates, and then enable cellular uptake of these ions by ionophore action [3]. However, repeated phase II clinical trials ultimately provided no evidence for cognitive efficacy of PBT2 in patients with prodromal or mild AD [4,5].
Using electron paramagnetic resonance (EPR) spectroscopy, the Cu 2+ coordination of this class of ligand (L) was first characterized using the non-chlorinated homologue of PBT2 (Figure 1), which was shown to form a terdentate CuL complex and a five-coordinate CuL 2 complex [6]. The proposed structure of CuL 2 has been replicated in the crystal structure of PBT2 [7], and the UV-vis and EPR spectroscopic properties of Cu(PBT2) and Cu(PBT2) 2 have been shown to mirror those of the non-chlorinated homologue [6][7][8][9] (Table  S1). Early EPR studies also showed that both ligands form ternary CuLN Im X complexes ( Figure 1) in which the labile Cl − ligand of CuL is replaced with an imine (N Im ) donor ligand provided by X = imidazole, histamine, L-His, or proteins such as α-synuclein and Aβ (see, in particular, Figure S33 of ref. [6]). However, despite recently identifying a similar EPR spectroscopic signature using a Cu/PBT2/Aβ 1-42 mixture, George and co-workers ascribed the signal to a unique PBT2-inaccessible Cu(Aβ) species and concluded that there was no evidence to support formation of a ternary Cu 2+ complex [10]. Since this could be interpreted by some as a reason for the failure of PBT2 in AD clinical trials, it is important to resolve this contradiction. coordinate CuL2 complex [6]. The proposed structure of CuL2 has been r crystal structure of PBT2 [7], and the UV-vis and EPR spectroscopic Cu(PBT2) and Cu(PBT2)2 have been shown to mirror those of the non-chl logue [6][7][8][9] (Table S1). Early EPR studies also showed that both ligand CuLNIm X complexes ( Figure 1) in which the labile Cl − ligand of CuL is re imine (NIm) donor ligand provided by X = imidazole, histamine, L-His, or p α-synuclein and Aβ (see, in particular, Figure S33 of ref. [6]). However, d identifying a similar EPR spectroscopic signature using a Cu/PBT2/Aβ1-42 m and co-workers ascribed the signal to a unique PBT2-inaccessible Cu(Aβ) s cluded that there was no evidence to support formation of a ternary Cu 2 Since this could be interpreted by some as a reason for the failure of PBT trials, it is important to resolve this contradiction.
To address the above issue, we used EPR spectroscopy to analyze th butions in Cu/PBT2/Aβ mixtures in unprecedented detail and identify th stabilities of ternary Cu(PBT2)NIm X complexes formed with X = Aβ1-40.
To quantify the number and stabilities of ternary Cu(PBT2)NIm Aβ in important to consider how the His residues participate in binary Cu(Aβ pH 7.4, Cu(Aβ) is dominated by {NH2 D1 , C=O D1 , NIm H6 , NIm H13 } and {NH2 D NIm H14 } coordination spheres with indistinguishable EPR spectra [12][13][14]. T solutely required to form Cu(Aβ), whereas only one of His13 or His14 is anchoring of Cu(PBT2) on His6 to form Cu(PBT2)NIm H6 will occur at Cu(Aβ), whereas anchoring on one of His13 or His14 to form Cu(PBT2)N Sequential binding of a second Cu 2+ ion by Cu(Aβ) generates Cu2(Aβ). Th in Cu2(Aβ) remains poorly defined, with one suggestion that binding of t To address the above issue, we used EPR spectroscopy to analyze the species distributions in Cu/PBT2/Aβ mixtures in unprecedented detail and identify the number and stabilities of ternary Cu(PBT2)N Im X complexes formed with X = Aβ 1-40 .

Results
To characterize the Cu 2+ complexes formed by PBT2 in the presence of Aβ, we titrated an equimolar mixture of PBT2 and Aβ 1-40 in PBS pH 7.4 with Cu 2+ and acquired the corresponding EPR spectra ( Figure 2a). To determine the species distribution, each spectrum was decomposed into a linear superposition of basis spectra (Figure 2b) corresponding to Cu(PBT2), Cu(PBT2) 2 , Cu(Aβ 1-40 ), Cu 2 (Aβ 1-40 ), and the putative Cu(PBT2)N Im Aβ complex. The spectrum of Cu(PBT2) was obtained at equimolar Cu/PBT2 stoichiometry and that of Cu(PBT2) 2 at sub-stoichiometric ratios ( Figure S1). The spectra of Cu(Aβ) and Cu 2 (Aβ) were acquired at Cu/Aβ ratios of 1:1 and 2.5:1 ( Figure S2). Attempts to use linear combinations of normalized Cu(PBT2), Cu(PBT2) 2 , Cu(Aβ), and Cu 2 (Aβ) spectra failed to reproduce the EPR spectra of Cu/PBT2/Aβ 1-40 n:1:1 (n = 0-2.67), indicating that additional species must be formed. Indeed, the dominant spectral features did not resemble any of those of the above four species (Figure 2b and Figure S3). Rather, they corresponded closely to those of previously characterized CuLN Im complexes (Table 1) [6,8,11], indicating that Cu(PBT2) can anchor on the imine nitrogen (N Im ) of the His side chains of Aβ 1-40 .  Table 1. The Cu2(Aβ1-40) simulation is a weighted summation of spectra obtained using "first site" parameters (50%) and "second site" parameters (50%). Vertical scales in panels a and b are different. (c) Experimental species distributions (points) resulting from normalization and decomposition of the spectra in panel a ( Figure S4), and theoretical distributions (solid lines) calculated using the relevant formation constants in Table 2 (see also Figure S5  The conditional constants ( ) for formation of Cu(PBT2), Cu(PBT2)2, Cu(Aβ), and Cu2(Aβ) at pH 7.4 have all been previously determined (Table S2), which greatly simplified the task of determining the species distribution of the Cu/PBT2/Aβ1-40 n:1:1 system. Using these values and the EPR basis spectra (Figure 2b), we fitted the series of EPR spectra in Figure 2a Table 1. The Cu 2 (Aβ 1-40 ) simulation is a weighted summation of spectra obtained using "first site" parameters (50%) and "second site" parameters (50%). Vertical scales in panels a and b are different. (c) Experimental species distributions (points) resulting from normalization and decomposition of the spectra in panel a (Figure S4), and theoretical distributions (solid lines) calculated using the relevant formation constants in Table 2 (see also Figure S5 for a comparison of relative and absolute Cu 2+ speciation    Figure S9c). A detailed comparison of Cu 2+ formation constants, including those for the non-chlorinated homologue of PBT2 and various N Im donors, is provided in Table S2. Cu(Aβ), whereas anchoring on one of His13 or His14 to form Cu(PBT2)N Im H13/14 will not. Sequential binding of a second Cu 2+ ion by Cu(Aβ) generates Cu 2 (Aβ). The coordination in Cu 2 (Aβ) remains poorly defined, with one suggestion that binding of the second Cu 2+ ion changes the coordination of the first [15]. However, it will be shown below that a satisfactory explanation of the species distributions requires that His6 remains Cu 2+ -bound and His13 or His14 non-coordinated in Cu 2 (Aβ).

Complex
The conditional constants ( c K) for formation of Cu(PBT2), Cu(PBT2) 2 , Cu(Aβ), and Cu 2 (Aβ) at pH 7.4 have all been previously determined (Table S2), which greatly simplified the task of determining the species distribution of the Cu/PBT2/Aβ 1-40 n:1:1 system. Using these values and the EPR basis spectra (Figure 2b), we fitted the series of EPR spectra in Figure 2a (Table S2) worsened the fit. Although we did not refine the value of c K CuL CuL 2 , we found that large changes from its published value (Table 2) also worsened the fit unless c K Cu Cu(Aβ) was set to a value beyond its accepted range.
The general appearance of the species distributions for Cu/PBT2/Aβ 1-40 n:1:1 (Figure 2c and Figure S5) can be understood as follows: For small n, the 1000-fold greater stability of Cu(PBT2) compared with Cu(Aβ) (Table 2) ensures that Cu 2+ will first bind to PBT2, with the identity of the fourth in-plane ligand then being determined by the relative magnitudes of the stepwise constants c K CuL Im . Thus, for n < 1, Cu 2+ is predominantly bound in a Cu(PBT2)N Im H6 complex with a minor quantity of Cu(PBT2) 2 . For n = 1, Cu(PBT2) 2 is almost entirely replaced by Cu(PBT2)N Im H6 and Cu(PBT2)N Im H13/14 , which only require one PBT2 molecule per Cu 2+ ion. For n > 1, with no free PBT2 molecules available, the additional Cu 2+ is coordinated in the next most stable binary complex, which is Cu(Aβ) (log c K Cu Cu(Aβ) = 10.0 ± 0.1). However, because Cu(Aβ) coordination requires His6, some Cu(PBT2) detaches from His6 and anchors instead on a His13 or His14 side chain, albeit with lower stability, to form Cu(PBT2)N Im H13/14 . As the available sites at His6 are gradually filled by Cu(Aβ), stepwise addition of a second Cu 2+ ion to the peptide occurs (log c K Cu(Aβ) Cu 2 (Aβ) = 8.0 ± 0.1) to form Cu 2 (Aβ). Maximum occupancy of the Cu/PBT2/Aβ 1-40 n:1:1 system is reached at n = 3, with two Cu 2+ ions bound to the "first" and "second" sites of Aβ, and a third Cu 2+ ion bound either to Cu(PBT2) that is free (minor) or anchored to His13/14 of Cu 2 (Aβ) in a Cu(PBT2)N Im H13/14 complex (major). More than three Cu 2+ ions cannot be accommodated by an equimolar PBT2/Aβ mixture, with the excess Cu 2+ ions existing as aqueous copper that will precipitate as [Cu(OH) 2 ] n at pH 7.4. Inclusion of Cu(PBT2)N Im H13/14 , whose formation does not depend on the Cu 2+ loading state of Aβ, was essential to fit the experimental data. Alternative explanations of the physical origin of the lower-affinity ternary complex, such as a change in coordination of the first-bound Cu 2+ ion in Cu 2 (Aβ), can be ruled out because the concentration of Cu 2 (Aβ) relative to that of the ternary complex is too low at n < 2 ( Figure S5).
The fact that c K CuL CuLN H6 Im is 100-fold larger than c K CuL CuLN H13/14 Im indicates that either Cu(PBT2)N Im H6 is stabilized by favorable outer-sphere ligand-peptide interactions and/or that Cu(PBT2)N Im H13/14 is destabilized by such interactions. To distinguish between these possibilities, we repeated the EPR analyses using Cu/PBT2/X 1:1:n systems with relatively unstructured N Im donor ligands from X = imidazole (Figures S6-S8) and histamine ( Figures S9-S11). The EPR spectra of the isolated ternary Cu(PBT2)N Im X complexes were characterized by the same parameters as Cu(PBT2)N Im Aβ spectra (Table 1), confirming that each of these ternary Cu 2+ complexes involves monodentate N Im coordination of the co-ligand (Figure 1). The difference between c K CuL CuLN imidazole Im and c K CuL CuLN H13/14 Im was within experimental error, whereas c K CuL CuLN histamine Im was slightly lower than these constants (Table 2).
Thus, we may conclude that the stability of Cu(PBT2)N Im H13/14 is not strongly influenced by outer-sphere ligand-peptide interactions, whereas such interactions greatly enhance the stability of Cu(PBT2)N Im H6 . To independently verify the EPR method for deriving the conditional formation constants, we also determined the stability of the ternary Cu 2+ complex of the non-chlorinated homologue of PBT2 (L ) with imidazole ( Figures S12-S14) and compared the value with that previously determined using potentiometric titrations [11]. After pH correction of the absolute stability constants (Table S3), c K CuL CuL N imidazole Im was not significantly different from that determined here using EPR (Table S2) and, similar to PBT2, slightly higher than c K CuL CuL N histamine Im (Figures S15-S17).

EPR spectroscopy isolated a common Cu(PBT2)N Im
Aβ spectrum ( Figure S3) for both Cu(PBT2)N Im H6 and Cu(PBT2)N Im H13/14 because they have very similar first coordination spheres. Nevertheless, as has been shown for other terdentate Cu 2+ chelators [16,17], ternary complexes with different N Im donor ligands can be distinguished based on their distinct formation constants, which are determined by outer-sphere interactions to which continuous-wave EPR is typically insensitive. Importantly, the spectroscopic signature of Cu(PBT2)N Im Aβ isolated here for Cu/PBT2/Aβ 1-40 n:1:1 closely matches that reported for the species isolated in Cu/PBT2/Aβ 1-42 1:2:1 [10] (Table 1). In the latter study, the authors ascribed this to a unique PBT2-inaccessible Cu(Aβ) complex and concluded that there was no evidence for ternary complex formation. However, it is clear that CuLN Im X complexes with this spectroscopic signature are formed by PBT2 with a number of N Im donor ligands X (Table 1 and Table S1). Moreover, we demonstrated the requirement for two such complexes-Cu(PBT2)N Im H6 and Cu(PBT2)N Im H13/14 -with distinct stabilities ( Table 2) to explain the species distributions of Cu/PBT2/Aβ 1-40 mixtures. The relatively high stability of Cu(PBT2)N Im H6 compared with complexes formed with other N Im donors might result from stabilizing pi-pi stacking of the aromatic rings of PBT2 and Phe4 or Tyr10, although a combination of electrostatic, steric, and hydrogen-bonding effects may contribute.
Despite the large ternary formation constant for Cu(PBT2)N Im H6 , PBT2 remains a promiscuous Cu 2+ chelator because it is capable of forming a ternary Cu(PBT2)N Im complex with all N Im donor ligands, including ubiquitous His side chains of peptides and proteins in the biological milieu, whose combined effect should outweigh that of a single Cu(PBT2)N Im Aβ complex regardless of its stability. We therefore conclude that PBT2 is capable of accessing Cu(Aβ) complexes with high stability but low specificity. Potential functional implications of Cu(PBT2)N Im complexes have been discussed in our previous studies of the non-chlorinated PBT2 homologue. First, as an alternative to acting as a mobile ion carrier (ionophore) in a lipid membrane, endocytosis of extracellular proteins on which Cu(PBT2) is anchored, followed by release of Cu 2+ in low-pH and/or reducing intracellular compartments, may contribute to the bulk transport of these ions [6]. Second, their production of ROS in the presence of ascorbate [11], in addition to the modulation of cellular ROS signaling following exposure to this class of ligand [17], contrasts with the originally intended ROS-silencing function of PBT2 [18].
Recently, PBT2 has found renewed interest as an antimicrobial compound. Notably, a number of gram-positive bacteria become re-sensitized to previously resistant classes of antibiotics when these antibiotics are supplemented with PBT2 and Zn 2+ in mouse models of wound healing [19] and pneumonia [20]. These results have been attributed to multiple bactericidal mechanisms associated with intracellular Zn 2+ accumulation, including im-pairment of Mn-dependent superoxide dismutase and production of ROS [21]. Although ligands generally have a greater affinity for Cu 2+ compared with Zn 2+ [22], the above effects were observed in response to co-administration of PBT2 (~1 µM) with an excess of Zn 2+ (~100 µM). Therefore, it remains unclear whether ternary Cu(PBT2)N Im complexes may be formed under these conditions. However, given the ability of PBT2 to also form terdentate Zn 2+ chelates [7], a contribution from analogous Zn(PBT2)N Im complexes to the antimicrobial activity of Zn/PBT2 may be speculated.

Sample Preparation
Aβ 1-40 (purity > 95%) was synthesized in the Peptide Technology Facility of the Bio21 Molecular Science and Biotechnology Institute, The University of Melbourne. The lyophilized peptide was dissolved at a nominal concentration of 1 mg/mL in 1,1,1,3,3,3hexafluoroisopropanol and portioned, then the solvent was allowed to evaporate. The resulting peptide film was resuspended at 4 • C in 20% v/v 20 mM NaOH, 70% v/v ultrapure water (MilliQ; MERCK KGAA, Darmstadt, Germany), and 10% v/v 10 × phosphate-buffered saline (PBS; 10 mM phosphate buffer, 2.7 mM KCl, 137 mM NaCl; Sigma). After centrifugation at 14,000× g for 15 min at 4 • C, the supernatant was retained and the peptide concentration was immediately determined using ε 214 = 74,925 M −1 cm −1 [23]. A concentrated stock of 65 CuCl 2 was prepared by stirring 65 CuO ( 65 Cu, >99%; Cambridge Isotope Laboratories, Tewksbury, MA, USA) in concentrated HCl, evaporating excess HCl under heat, then adding ultrapure water. PBT2 was synthesized as previously described [24], and a 1 mM stock solution was prepared by solubilizing the hydrochloride salt directly in PBS.

EPR Spectroscopy
X-band continuous-wave EPR spectra were acquired using a Bruker ESP380E spectrometer fitted with a rectangular TE 102 microwave cavity and a quartz cold finger insert. Microwave frequencies were measured using an EIP Microwave 548A frequency counter and g factors calibrated against the F + line in CaO (g = 2.0001 ± 0.0002) [25]. Experimental conditions are indicated in the figure captions. Background correction was performed by subtraction of the buffer-only spectrum.
The "pepper" function in EasySpin v.5.2.33 [26,27] was used to simulate basis spectra using the static Hamiltonian H = β e B T · g · S + S T · A · I − g n β n B T · I (1) where S and I are the electron and 65 Cu nuclear vector spin operators, g and A are the 3 × 3 electron Zeeman and 65 Cu electron-nuclear hyperfine coupling matrices, β e is the Bohr magneton, β n is the nuclear magneton, and B is the applied magnetic field vector. Rhombic symmetry or higher was assumed for g and A, with principal values of g x , g y , and g z and A x , A y , and A z , respectively. The principal values of g and A and the lineshape parameters (g-A strain model) were varied iteratively using the "esfit" module in EasySpin to minimize the difference between the experimental and simulated spectra.