Alzheimer’s drug PBT2 interacts with the amyloid beta 1-42 peptide differently than other 8-hydroxyquinoline chelating drugs

Abstract

Although Alzheimer’s disease (AD) was first described over a century ago, it remains the leading cause of age-related dementia. Innumerable changes have been linked to the pathology of AD; however, there remains much discord regarding which might be the initial cause of the disease. The “amyloid cascade hypothesis” proposes that the amyloid beta (Aβ) peptide is central to disease pathology, which is supported by elevated Aβ levels in the brain before the development of symptoms and correlations of amyloid burden with cognitive impairment. The “metals hypothesis” proposes a role for metal ions such as iron, copper, and zinc in the pathology of AD, which is supported by accumulation of these metals within amyloid plaques in the brain. Metals have been shown to induce aggregation of Aβ and metal ion chelators have been shown to reverse this reaction in vitro. 8-Hydroxyqinoline-based chelators showed early promise as anti-Alzheimer’s drugs. Both 5-chloro-7-iodo-8-hydroxyquinoline (CQ) and 5,7-dichloro-2[(dimethylamino)methyl]-8-hydroxyquinoline (PBT2) underwent unsuccessful clinical trials for the treatment of AD. To gain insight into the mechanism of action of 8HQs, we have investigated the potential interaction of CQ, PBT2, and 5,7-dibromo-8-hydroxyquinoline (B2Q) with Cu(II)-bound Aβ(1-42) using X-ray absorption spectroscopy (XAS), high energy resolution fluorescence detected (HERFD) XAS, and electron paramagnetic resonance (EPR). By XAS, we found CQ and B2Q sequestered ~83% of the Cu(II) from Aβ(1-42), whereas PBT2 sequestered only ~59% of the Cu(II) from Aβ(1-42), suggesting that CQ and B2Q have a higher relative Cu(II) affinity than PBT2. From our EPR, it became clear that PBT2 sequestered Cu(II) from a heterogeneous mixture of Cu(II)Aβ(1-42) species in solution leaving a single Cu(II)Aβ(1-42) species. It follows that the Cu(II) site in this Cu(II)Aβ(1-42) species is inaccessible to PBT2 and may be less solvent-exposed than in other Cu(II)Aβ(1-42) species. We found no evidence to suggest that these 8HQs form ternary complexes with Cu(II)Aβ(1-42).

Graphical Abstract

1. Introduction

Alzheimer’s disease (AD) is the leading cause of age-related dementia, accounting for approximately two thirds of all dementia cases. Alois Alzheimer first described the disease in 1907;^1–2 however, AD was not recognized as a common cause of dementia or a major cause of death for almost 70 years. In 2021, AD was the 6^th leading cause of death in the United States with more than 6 million people living with the disease.³ One in 3 seniors dies of AD or another dementia.³ Without significant developments in preventing or curing the disease, the prevalence of AD is projected to more than double by 2050, affecting an estimated 12.7 million Americans.³ In 2021, dementias cost $355 billion in the United States alone, but in 2050 this number is expected to reach $1.1 trillion.³

Brain changes associated with AD may begin more than 20 years prior to the onset of symptoms.^4–6 In a recent study, patients carrying genetic mutations that cause AD were found to have significantly elevated levels of amyloid beta (Aβ) in the brain up to 22 years before symptoms were expected to develop (based on the age their parents developed AD symptoms).⁵ Additionally, glucose metabolism began to decrease approximately 19 years – and brain atrophy began approximately 13 years – before the expected onset of symptoms.⁵ These results suggest that several changes may compound over time to result in AD pathology.

Post-mortem AD brain tissues typically show significant inflammation on both a macroscopic scale, with generalized brain atrophy and shrinkage from cell loss, as well as on a microscopic scale, with debris from dead and dying neurons and astrocyte and microglial activation.^7–9 Other microscopic brain changes associated with AD pathology include cleavage of the amyloid precursor protein to form soluble oligomeric Aβ and extracellular amyloid plaques, along with accumulation of hyperphosphorylated tau proteins to form intracellular neurofibrillary tangles. Because amyloid plaques are sometimes observed in aged individuals without clinical symptoms of AD,^10–16 the specific contribution of amyloid plaques to AD pathology has been intensely debated.¹⁷ However, various studies have shown that amyloid plaque burden correlates with reduced cognitive function and these have been reviewed previously.^18–19

According to the “amyloid cascade hypothesis”, the Aβ peptide plays a central role in the pathogenesis of AD. The amyloid precursor protein (APP) is cleaved through two potential pathways by metalloproteases: the nonamyloidogenic or the amyloidogenic pathway. Normally, the extracellular APP domain is cleaved by α-secretase, followed by intramembranous domain cleavage by γ-secretase; however, in the amyloidogenic pathway, the initial cleavage of the extracellular APP domain is mediated by BACE 1, resulting in the formation of 39–43 amino acid Aβ peptides.^20–24 In AD brain plaques, several Aβ peptide forms with up to 43 amino acids, and various N- and C-terminal truncations, have been identified.^25–27 Herein, the notation Aβ(x-y), where x is the first, N-terminal amino acid and y is the last, C-terminal amino acid, is used to describe the forms of Aβ studied here and elsewhere. The most prevalent alloforms of Aβ found in the AD brain are Aβ(1–40) and Aβ(1-42),^28–30 with the slightly longer Aβ(1-42) form proposed to be the more toxic form because of its propensity for oligomerization.³¹

Soluble Aβ peptides are found in cerebrospinal fluid and blood plasma of all humans, including the young and those without any dementia-like symptoms; what triggers the conversion from the apparently non-toxic, soluble form to the toxic, amyloidogenic forms of Aβ observed in AD is largely unknown. However, dyshomeostasis of metals such as iron,³² copper,³³ and zinc^{32, 34} has been implicated in the aggregation of Aβ and these studies have been extensively reviewed.^35–41 Several other possible triggers for nucleation and oligomerization of amyloid peptides have been proposed, including lipids, gangliosides, cholesterol, and sugars, and these mechanisms have been recently characterized⁴² or recently reviewed.^43–46 Despite recent contention surrounding the role of metals in the “amyloid cascade hypothesis”,⁴⁷ there is indisputable evidence to support altered metal ion homeostasis – especially of copper – in AD.^48–49 Several studies, including meta-analyses of large data sets combined over time and laboratories, have shown that overall copper levels are ~10–40% lower in AD brains,^50–53 depending on the brain region, and ~12% higher in AD serum^54–58 compared to healthy controls.

The “metal hypothesis” of AD suggests that the interaction of Aβ with iron, copper, and/or zinc at glutamatergic synapses promotes aggregation and precipitation of Aβ in the form of insoluble amyloid plaques, which then results in neuronal network disruption and furthers AD pathogenesis.⁵⁹ This hypothesis is supported by several studies that have found metal cations such as iron, copper, and zinc to be approximately 23 to 26% higher in amyloid plaques compared with surrounding neuropil in the human AD brain.^60–64 In addition, redox-active metals such as iron and copper have been connected to the possible formation of harmful and reactive low molecular weight species such as hydroxyl radicals within amyloid plaques in the brain,^{37–38, 65–66} which could potentially contribute to the observed oxidative damage.

Pharmaceuticals have been developed to target metal ions in amyloid plaques in attempts to ameliorate aberrant metal-Aβ interactions (Figure 1). Early studies showed that metals such as iron, copper, and zinc induced aggregation of Aβ in vitro, that this aggregation was reversible through metal ion chelation by common metal ion chelators such as ethylenediaminetetraacetic acid (EDTA) and diethylenetriamine pentaacetic acid (DTPA), and that chelation could reduce hydrogen peroxide production.^{32–33, 67–70} Metal chelators such as ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), N,N,N′,N′-tetrakis(2-pyridinylmethyl)-1,2ethanediamine (TPEN) and bathocuproine were also shown to aid in solubilization of Aβ from post-mortem AD brain tissue.^71–72 Subsequently, several bifunctional compounds that interact with both metals and Aβ were designed by Lim et al. and also reported to disrupt aggregation of the Aβ peptide and lessen neurotoxicity.^73–76

Figure 1. Structures of various metal ion chelators explored in the treatment of Alzheimer’s disease.

Full chemical names are explained in the text.

Cherny et al. reported a 49% decrease in brain amyloid in a transgenic AD mouse model treated orally with clioquinol (5-chloro-7-iodo-8-hydroxyquinoline; PBT1; CQ).⁷⁷ These studies precipitated much of the research that has investigated the use of 8-hydroxyquinoline-based compounds in the treatment of neurodegenerative disorders such as AD. 8-Hydroxyquinolines (8HQs) have since been shown to suppress Aβ oligomer formation in vitro, both in the presence and in the absence of metal ions,^78–80 and to suppress amyloid deposition in vivo.^{77, 81} Two 8HQs, namely CQ and PBT2 (5,7-dichloro-2-[(dimethylamino)methyl]-8-hydroxyquinoline), showed significant promise early on in mouse models of AD and underwent clinical trials to assess their efficacy and safety in human clinical trials for neurodegenerative diseases.^82–88 However, despite some promising results from early, small clinical trials, including high safety and tolerability outcomes, as well as improved cognition and lowered levels of Aβ in serum, it is evident that there is much that is not yet understood about the disease pathology and about the mechanisms of action of these compounds. In subsequent larger, extended clinical trials investigating the anti-Alzheimer’s properties of PBT2, treatment reportedly did not show significant cognitive improvements over the placebo group.^89–90 Patients administered PBT2 had a lower amyloid burden after the 1-year treatment compared to their initial values; however, the small (n=15) placebo group also showed an inexplicably reduced amyloid burden as measured by positron emission tomography (PET) scans using Pittsburgh compound B (PiB; a radioactive analog of thioflavin T).^89–90 It is likely that the lack of significant results from monitored outcomes in these trials were influenced by several factors, possibly including individual variance in disease biomarkers^91–92 and the relatively small number of patients in the study. For example, amyloid burden has been more recently found to change with age in complex ways that depend on several factors, such as the presence of the ε4 variant of the apolipoprotein E gene.⁹³ Additionally, the majority of studies evaluating the applicability of clinical PiB-PET imaging in the diagnosis and evaluation of AD have used the outcomes from a single PiB-PET test,^93–95 not repeat tests over time, so it is possible that the sensitivity to PiB might change with repeated measurement. Overall, the results from clinical trials investigating PBT2 appear to highlight the ongoing uncertainty concerning whether metal-Aβ biochemistry is a cause or an effect of AD.

Although the outcomes of clinical trials investigating CQ and PBT2 appear largely unsuccessful, many more metal-protein attenuating compounds (MPACs) have been designed, synthesized, and tested for effectiveness against a host of different peptides linked to neurodegenerative diseases. Because many of these subsequent MPACs have chemically similar structures and metal-coordinating ligands, further understanding of 8HQ-based compounds – the first MPACs – and their mechanism(s) of action is still pertinent to current endeavours in the field. For example, 1-methyl-1H-imidazole-2-carboxaldehyde isonicotinoyl hydrazone (X1INH) was recently reported to attenuate abnormal Cu-α-synuclein interactions and to lessen protein aggregation.⁹⁶ Several other examples have been recently reviewed.^97–101

In this study, we investigated the potential interaction of 8HQs, shown at the bottom of Figure 1, with Cu(II)-bound Aβ(1-42) using X-ray absorption spectroscopy (XAS), high resolution fluorescence detected (HERFD) XAS, and electron paramagnetic resonance (EPR). We found that CQ and B2Q (5,7-dibromo-8-hydroxyquinoline) sequestered approximately 83% of the Cu(II) from Aβ(1-42), whereas PBT2 sequestered only approximately 59% of the Cu(II) from Aβ(1-42). From EPR it appears that the Cu(II) that remains bound to Aβ(1-42) is bound in a specific conformation, which may suggest that one conformation of Cu(II)Aβ(1-42) has a higher Cu(II) affinity than PBT2 or the Cu(II) is inaccessible to this larger, more flexible Cu(II) chelator. We found no evidence of a ternary PBT2-Cu(II)-Aβ(1-42) complex, contrary to some previous studies¹⁰² and in line with others.⁸⁰

2. Results and Discussion

2.1. Cu K-edge X-Ray Absorption Spectroscopy and High Energy Resolution Fluorescence Detected XAS

We have previously published detailed analyses of XAS and HERFD-XAS of both Cu(II)Aβ(1-42)¹⁰³ and Cu(II)-bis-8HQs^104–105 in aqueous surfactant solutions; these spectra are used only for comparison herein (Figure 2). The HERFD-XAS technique¹⁰⁶ results in the collection of generated X-ray fluorescence with better resolution than the natural line width. In conventional K-edge XAS, the short lifetime of the 1s core hole causes spectral broadening, and consequent loss of detail and chemical sensitivity. As described previously,^103–105 Cu K-edge HERFD-XAS uses the Cu Kα₁ fluorescence line and observes a small subset of the 2p_3/2→1s transitions that produce the Cu Kα₁ fluorescence, eliminating most of the lifetime broadening from the 1s core hole and improving the spectral resolution.

Figure 2. Cu K-edge near-edge HERFD-XAS (solid) and XAS (dashed) spectra of Cu(II)Aβ(1-42) with 2 molar equivalents of 8-hydroxyquinolines CQ, B2Q, or PBT2.

Spectra of Cu(II) complexes of CQ, B2Q, and PBT2, prepared as 2 8HQs: 1 Cu(II), are shown for comparison. Spectra are offset vertically by 0.5 for clarity. The vertical dotted line indicates the 1s→4p + LMCT shake-down transition.

X-ray absorption near-edge spectra, from conventional XAS and HERFD-XAS, of Cu(II)-bound Aβ(1-42) in aqueous surfactant solutions after the addition of 8HQs are shown in Figure 2. In both conventional XAS and HERFD-XAS near-edge spectra, the addition of two molar equivalents of either CQ or B2Q to a solution of Cu(II)Aβ(1-42) produces spectra that more closely resemble Cu(II)-bis-CQ or Cu(II)-bis-B2Q spectra, respectively (Figure 2). These near-edge spectra suggest that CQ and B2Q almost completely sequester Cu(II) ions from Cu(II)-bound Aβ(1-42) such that the resulting spectra are highly similar to those of Cu(II)-bis-CQ or Cu(II)-bis-B2Q in solution. In contrast, when two molar equivalents of PBT2 are added to a solution of Cu(II)Aβ(1-42), the resultant spectra contains features of both Cu(II)Aβ(1-42) and Cu(II)-bis-PBT2 spectra (Figure 2).

2.2. CQ and B2Q Sequester Most of the Cu(II) from Aβ(1-42)

Fitting the linear combination of Cu(II)Aβ(1-42) and Cu(II)-bis-CQ to XAS edge spectra of Cu(II)Aβ(1-42) with 2 molar equivalents of CQ suggests that 83 ± 2% of the Cu(II) is bound to CQ as a bis complex (Figure 3). Likewise, fitting the linear combination of Cu(II)Aβ(1-42) and Cu(II)-bis-B2Q to XAS edge spectra of Cu(II)Aβ(1-42) with 2 molar equivalents of B2Q also suggests that 83 ± 1% of the Cu(II) is bound to B2Q as a bis complex. A linear combination of 41 ± 2% Cu(II)Aβ(1-42) and 59 ± 2% Cu(II)-bis-PBT2 results in a good fit of the near-edge spectrum of Cu(II)Aβ(1-42) with 2 molar equivalents of PBT2 (Figure 3), suggesting that PBT2 removes a smaller fraction of the Cu(II) from Aβ(1-42) compared with CQ and B2Q.

Figure 3. Cu K-edge near-edge XAS showing linear combination fits of Cu(II)Aβ spectra with 2 molar equivalents of (A) CQ, (B) B2Q, and (C) PBT2.

In each panel A-C the data is shown as a black solid line, the fit is shown as a colored dashed line, the Cu(II)-bis-8HQ complex is shown as a colored solid line, and Cu(II)Aβ(1-42) is shown as a dashed black line. The intensity of both the spectra of Cu(II)-bis-8HQ complexes and the spectra of Cu(II)Aβ(1-42) has been adjusted to reflect their best fit proportion. The e.s.d. was ± 2% in (A) and (C) and ± 1% in (B).

Findings that CQ and B2Q sequester most of the Cu(II) from Aβ(1-42) agree with previously published formation constants for Cu(II) complexes with 8HQs.^107–110 However, the finding that PBT2 appears to sequester a smaller fraction of Cu(II) from Aβ(1-42) compared to CQ and B2Q appears to contrast with the previously determined formation constant for Cu(II)PBT2.¹¹¹ The formation constants, log K₁ and log K₂, for the Cu(II)-bis-CQ complex¹⁰⁷ and the Cu(II)-bis-B2Q complex,¹⁰⁹ which are similar to those published for the Cu(II)-bis-8HQ complex,^{108, 110} are listed in Table 1. The Cu(II) affinity K of the Aβ peptide has been proposed previously to be anywhere from ~32 μM to < 1 pM (Table 1). More recently, Alies et al. re-evaluated the Cu(II) affinities proposed in previous studies and more carefully accounted for various factors including the technique and buffering system used. Alies et al. found the re-evaluated Cu(II) affinity in most previous works agreed on a log K of ~ 10 for Aβ (K ≈ 100 pM). The Cu(II) affinities of most 8HQs appear to be higher than those of Aβ, assuming log K values of the Cu(II) binding sites in Aβ peptides are approximately 10 as was proposed by Alies et al. (Table 1).¹¹²

Table 1.

Cu(II) affinities of 8HQs and Aβ from the literature.^a

8HQ	Log K1	Log K2	Ref.
8HQ	13.49	12.73	Johnston and Freiser,108 and Stevenson and Freiser110
CQ	12.50	10.90	Budimir et al.107
B2Q	12.01	11.23	Gupta et al.109
PBT2	13.61	5.95	Sgarlata et al.111
AP	Log K1	Log K2	Ref.
1–40	5.4		Atwood et al. 199833
1–42	6.5		Atwood et al. 199833
1–40	5.8		Garzon-Rodriguez et al.113
1–42	5.7		Garzon-Rodriguez et al.113
1–40b	10.3	7.9	Atwood et al. 2000114
1–42b	17.2	8.3	Atwood et al. 2000114
1–28	7.0		Syme et al. 2004115
1–40	5.8		Hou et al.116
1–16	7.0		Ma et al.117
1–16	7.0	5.0	Guilloreau et al.118
1–42	6.3		Danielsson et al.119
1–28	6.4		Danielsson et al.119
1–40	4.5		Tõugu et al.120
1–16	9.2		Hong et al.121
1–40c	9.0		Hatcher et al.122
1–40b	9.4		Hatcher et al.122
1–42	10.2–11.2		Sarell et al.123
1–40	7.2		Rózga et al.124
1–16, 1–40	10	7.7–8.0	Alies et al.112
1–16	9.2		Conte-Daban et al.125
1–28	9.3		Conte-Daban et al.125
1–40	9.4		Conte-Daban et al.125

^aValues listed are those reported from various techniques in approximately chronological order. Values listed were reported for physiological pH values (in the range pH 7 – pH 7.5). Reported pH values are indicated where more than one affinity constant was provided for the pH range of interest;

^bValues reported for pH 7.4;

^cValues determined at pH 7.2

8HQs with 2-position substituents were found to have more varied formation constants for the Cu(II) complex, which was proposed as an indication of whether the 2-position substituent might be involved in Cu(II) coordination.¹¹⁰ 8HQs with 2-position substituents that likely sterically inhibit Cu(II) complex formation were found to have lower log K₁ and log K₂ values compared with unsubstituted 8HQ.¹¹⁰ 8HQs with 2-position substituents that likely induce tridentate Cu(II) chelate formation were found to have higher log K₁ values, but lower log K₂ values than 2-CH₃ substituted 8HQ.¹¹⁰ Stevenson and Freiser proposed that the striking difference between log K₁ and K₂ values could be explained by the initial formation of a tridentate 8HQ complex with the 2-position amine acting as a ligand, along with the phenolate O and pyridine N. However, one of the ligands – either the amine N or the phenolate O – would need to be released to bind a second 8HQ moiety and form a Cu(II)-bis-8HQ complex.¹¹⁰ PBT2 was recently reported to have a Cu(II) affinity log K₁ value of 13.61 and a log K₂ value of 5.95 (Table 1),¹¹¹ which is in agreement with formation constants from other 2-position substituted 8HQs that initially act as tridentate chelates and subsequently form bidentate, bis chelate complexes. However, the log K₁ value proposed for the formation of the 1 Cu(II): 1 PBT2 complex seems to contrast with the findings here because the log K₁ value of PBT2 was proposed to be higher than those of CQ and B2Q. Based on the published Cu(II) binding affinities, PBT2 would be proposed to sequester more Cu(II) from Aβ than CQ or B2Q. Instead, it appears that Cu(II) sequestration from Aβ(1-42) by PBT2 may be more complex than can be accounted for by a simple binding constant model; there may be other contributing factors.

The formation of ternary complexes between Cu(II), 8HQs, and other small molecules (e.g. amino acids and neurotransmitters) and peptides has been proposed previously.^{102, 111, 126} Sgarlata et al. found evidence for ternary complexes with Cu(II), either 8HQ, CQ, or PBT2, and amino acids such as glycine, glutamate, and histidine.¹¹¹ Kenche et al. proposed that 2-[(dimethylamino)methyl]-8-hydroxyquinoline formed a ternary complex with Cu(II) and Aβ(116).¹⁰² Similarly, Mital et al. found 2-[(dimethylamino)methyl]-8-hydroxyquinoline to form a metastable intermediate complex with Cu(II) and Aβ(4–16).¹²⁶ Surprisingly, 8HQs have also been found to interact with Aβ(1-42) in the absence of metals, disrupting aggregation of Aβ(1-42) and stabilizing formation of dimeric 8HQ-Aβ(1-42) species.⁸⁰ Our data do not show strong evidence supporting formation of a ternary complex between Cu(II), Aβ(1-42), and CQ, B2Q, or PBT2, although a ternary complex could constitute a minor species not fit in our model (Figure 3). Instead, our results suggest that PBT2 is a weaker Cu(II) chelator compared with CQ and B2Q and sequesters approximately 59% of the Cu(II) from Aβ(1-42) compared with the approximately 83% sequestered by CQ or B2Q.

2.3. PBT2 Has a Lower Relative Cu(II) Affinity Than CQ and B2Q, and Does Not Appear to Form Ternary Complexes with Cu(II)Aβ(1-42)

As described above, the near-edge of Cu(II)Aβ(1-42) with 2 molar equivalents of PBT2 is best fit to a linear combination of Cu(II)-bis-PBT2 and Cu(II)Aβ(1-42) (Figure 3 C). The EPR spectrum of Cu(II)Aβ(1-42) is magnetically non-dilute (Figure 4), consistent with a non-homogeneous mixture of at least 2 species or spin-spin coupling of distant Cu(II) centers. Solutions of Cu(II)Aβ(1-42) at physiological pH have previously been shown to be a mixture of at least 2 species.^{103, 115, 118, 127–130} The two species that coexist at physiological pH are commonly termed ‘component I’ and ‘component II’. Component I is the dominant species below pH ~6.5 and is more recently thought to represent a Cu(II) coordination environment in which the 3 histidine residues rapidly exchange such that only 2 histidine nitrogens are bound simultaneously.^{129, 131–133} The N-terminal amino nitrogen, and possibly the carboxylate O, of Asp1 are thought to complete the Cu(II) coordination sphere in component I.^{123, 129, 132, 134} Component II is the dominant species above pH ~9 and the Cu(II) coordination environment in this species is less clear. Recent results suggest that all 3 histidines may simultaneously coordinate in component II,^{123, 129, 131} although other studies have suggested coordination of a single histidine in component II.^{130, 135} The other 1–3 ligands required to complete the coordination sphere are unknown, with various ligands proposed in more recent studies.^{123, 129–131, 133}

Figure 4. EPR spectra of Cu(II)Aβ(1-42) with PBT2.

Spectra include (a) Cu(II)-bis-PBT2, (b) Cu(II)Aβ(1-42) with 2 molar equivalents of PBT2, and (c) Cu(II)Aβ(1-42). The resultant spectrum after subtraction of the Cu(II)-bis-PBT2 spectrum (a) from that of Cu(II)Aβ(1-42) with 2 molar equivalents of PBT2 (b) is shown in (d) (solid) with its simulation (dashed line). (e) The second derivative of spectrum (d) (solid) with its simulation (dashed line). The second derivative was smoothed using 3-point smoothing. Simulation parameters are listed in Table 2.

Although most of the literature EPR spectra of Cu(II)Aβ(1-x) solutions appear to be characteristic of magnetically isolated S=½ Cu(II) species and thus can be considered magnetically dilute, these spectra are primarily of C-terminally truncated Cu(II)Aβ(1–16). There are only two literature examples of EPR spectra of the full Cu(II)Aβ(1-42), and although the spectra appear magnetically dilute, the concentrations used in those studies are much lower than that used herein (e.g. 50 μM and 100 μM vs. 0.5 mM).^{123, 136} This raises the question of whether the effect might be concentration dependent where spin-spin coupling between Cu(II) atoms ~7 to 10 Å apart in more concentrated solutions could contribute to the EPR spectrum appearing magnetically non-dilute.

Subtraction of the Cu(II)-bis-PBT2 spectrum from the Cu(II)Aβ(1-42) + 2 PBT2 spectrum results in a spectrum that has similarities to that of Cu(II)Aβ(1-42), but appears to be magnetically dilute and homogeneous (Figure 4 d, e). One explanation is that one Cu(II) conformation in the known mixture of components I and II at pH 7.4 in Cu(II)Aβ(1-42) is resistant to Cu(II) sequestration by PBT2. In support of this explanation, the g_∥ and A_∥ values of the difference spectrum (Figure 4 d) appear to be more similar to those previously reported for component II compared with component I.^{123, 129–130, 132–133} Another possibility is that the formation of some Cu(II)-bis-PBT2 interferes with the spin-spin coupling of distant Cu(II) atoms that made the original Cu(II)Aβ(1-42) spectrum appear non-dilute magnetically. A combination of these effects is also possible – PBT2 may sequester Cu(II) from one of the two described Aβ components at pH 7.4 and Cu(II)-bis-PBT2 may disrupt stacking of Cu(II)Aβ(1-42) chains in the remaining component. Alternatively, there may be some aggregation of Cu(II)Aβ(1-42) in the original solution as the long 1–42 form is highly prone to aggregation. With regards to the last alternative scenario, however, the presence of surfactant has previously been shown to prevent aggregation of Aβ,^137–138 and the short timescale between solubilization of lyophilized Aβ(1-42), the addition of Cu(II), and subsequent flash freezing of the samples in this study disfavours aggregate formation. Fluctuations in pH are also an unlikely explanation. It is improbable that the pH of the Aβ(1-42) solution changes significantly with the addition of 8HQ because of the excess buffer (i.e., 100 mM MOPS) present – especially with PBT2 because the stock solution was also prepared in buffered surfactant solution. Additionally, it is possible that the ratio of component I to component II is impacted by the buffered surfactant solution compared to other studies; however, the concentration of surfactant is consistent across all samples examined herein. Therefore, effects from the surfactant also cannot explain the differences observed with the addition of different 8HQs. We would also like to note that the presence of 10% DMSO in solutions with CQ or B2Q, but not those with PBT2, could influence the Aβ peptide structure between these solutions. However, because of the hydrophobic nature of the CQ and B2Q ligands, DMSO was required for solubility. Based on early studies, DMSO is likely to inhibit or reduce the formation of β-sheet secondary structure in Aβ.¹³⁹ Surfactants such as DTAB, used to help ensure monomeric peptide,^137–138 also favor the formation of α-helices over β-sheet secondary structure.^140–141 Therefore, in the presence of the surfactant, it seems unlikely that 10% DMSO would have a significant effect on the peptide secondary structure,¹⁴² other than to also help ensure that the peptide remains monomeric through the prevention of β-sheet formation.

2.4. Curve-Fitting of the EXAFS

EXAFS curve-fitting results are shown in Figure 5 and described in Table S1. As has been reported previously by our group, both CQ and B2Q form planar bis complexes with Cu(II) in which both the CQ or B2Q ligands and the Cu(II) are all co-planar.¹⁰⁴ EXAFS spectra of both Cu(II)Aβ(1-42) with 2 equivalents of CQ or B2Q, were fit using a model of Cu(II)-bis-CQ or B2Q, respectively (Figure 5; Table S1), and found to give reasonable agreement between the data and the model. These fitting results provide support for sequestration of most of the Cu(II) from Aβ(1-42) by CQ and B2Q.

Figure 5. Cu K-edge EXAFS spectra and Fourier transforms of Cu(II) complexes with Aβ(1-42) and with CQ, B2Q, or PBT2.

Data is shown in black and fits are shown in color. Fit details are listed in Table S.1. EXAFS spectra are offset by 12 and Fourier transforms are offset by 1 vertically for clarity. The dashed box highlights the ~6 Å peak further discussed in the text.

We previously found the EXAFS of a solution of excess PBT2: Cu(II) to be best fit to a distorted 4-coordinate Cu(II)-bis-PBT2 coordination environment, with 2 pyridine nitrogen ligands and 2 phenol oxygen ligands.¹⁰⁵ Additionally, we previously found the EXAFS of a solution of equimolar Aβ(1-42) and Cu(II) at pH 7.4 to be best fit to an average Cu(II) coordination environment consisting of two histidine nitrogens and two additional oxygen or nitrogen ligands.¹⁰³ EXAFS spectra of Cu(II)Aβ(1-42) with 2 molar equivalents of PBT2 were reasonably well fit to 59% Cu(II)-bis-PBT2 and 41% Cu(II)Aβ(1-42) (Figure 5; Table S1), using the proportions obtained from near-edge fitting analyses above (Figure 3). Our previous parameters obtained for both Cu(II)-bis-PBT2¹⁰⁵ and Cu(II)Aβ(1-42)¹⁰³ were used herein to obtain the fit shown in Figure 5 and detailed in Table S1. The multiplicity of each path, N, was reduced to the proportion of its corresponding species (Table S1) based on the results from near-edge fitting detailed above (Section 2.2). The distances between the absorbing copper and the backscattering atoms were refined as a set of linked paths. The fitting method used is further detailed in Methods Section 4.7.

A small peak at ~6.3 Å in the Fourier transform is observed in all three spectra of Aβ + 8HQs and is not accounted for in our models of Cu(II)-bis-8HQs or Cu(II)Aβ(1-42). We Fourier filtered this peak and found it to be best fit to a Cu^…Cu interaction with a distance of 6.30 Å. We propose that this peak arises from stacking of Cu(II)-bis-8HQs, possibly between Aβ peptide strands. This finding may be a result of the higher relative concentrations of both Aβ(1-42) and Cu(II) used in these XAS studies. Higher relative copper concentrations are required for sufficient signal-to-noise in Cu(II) XAS samples, where the sample must be moved in the X-ray beam to avoid photoreduction to Cu(I), as we have previously discussed.^{103–104, 143}

2.5. Discussions of in vivo Applicability

Our findings detailed herein are not intended to directly represent in vivo conditions. The conditions in the Alzheimer’s disease brain are complex and disease mechanisms are likely multi-faceted. For example, in vivo there are many more Aβ species present, compared with the purified Aβ(1-42) solutions used herein, although the presence of 400 mM detergent (DTAB) and 100 mM MOPS may aid in approximating in vivo conditions through pH buffering at pH 7.4 and molecular crowding. In addition, it is likely that Aβ exists in vivo in several forms, including the monomeric, soluble form, the oligomeric, fibrillar form, as well as in amyloid plaques, and that not all Aβ peptides might coordinate metal ions. However, we have examined a simplified model as a step to furthering our understanding of what might happen in vivo, while building upon previous studies that have primarily examined the truncated Aβ(1–16) that does not exist in the AD brain.

We have shown in vitro evidence to support weaker Cu(II) sequestration from Aβ(1-42) by PBT2, compared with CQ and B2Q. Based on these findings one might hypothesize that PBT2 would have poorer anti-Alzheimer’s effects; however, studies of AD animal models have shown PBT2 to have superior effects compared with CQ.⁸¹ Previous in vivo studies have described both CQ and PBT2 as ionophores or MPACs that can redistribute metals from areas of significant accumulation (e.g., amyloid plaques) to metal-deficient areas (e.g., neurons) rather than simply chelating aberrant metals.^{81, 144–145} Although it is very difficult to predict what may happen in vivo, our reported results also provide support for a more complex mechanism of action rather than simple Cu(II) sequestration. In the presence of other truncated forms of Aβ in vivo, namely N-terminally truncated Aβ(4-x) species containing amino terminal Cu(II)- and Ni(II)-binding (ATCUN) motifs, which are expected to have Cu(II) affinities more than 3 orders of magnitude greater than Aβ(1-x) species,¹⁴⁶ Cu(II) may not be as readily sequestered by 8HQ-based compounds. Additionally, amyloid plaques have been shown to be mixtures of many different aggregated peptides, as well as lipids, and metals.^{60–64, 147–153} Sequestration of metal ions trapped in plaque cores by 8HQs might be lowered due to the presence of these other materials. However, lipids on the plaque periphery may aid in permeation of the lipophilic 8HQs into plaques to reach the aggregated protein cores. Additionally, Cu(II) is found bound to soluble Aβ species that may be more readily accessible to sequestration by 8HQs. Therefore, it is possible that 8HQs act as ionophores, sequestering Cu(II) from accessible binding sites in soluble Aβ species and mobilizing this Cu(II) to deficient neurons, similar to previous proposals.^{81, 144–145}

To further address questions surrounding the potential interactions between Aβ peptides, metals, and 8HQs, future studies might build upon the in vitro experiments detailed herein. For example, examination of 8HQ interactions with oligomeric Aβ species, although challenging because of the potential for non-specific metal binding (i.e., both within and between Aβ peptides), is the logical next step in understanding these interactions. Now that we have established the Cu(II) distribution between Aβ(1-42) and 8HQs in a relatively simple model, future studies might build on these results to explore other Aβ peptides, other metal ions, and other metal-binding compounds.

3. Conclusions

XAS and HERFD-XAS near-edge data presented herein suggest that CQ, B2Q, and PBT2 all sequester Cu(II) from Aβ(1-42). However, CQ and B2Q appear to have higher relative Cu(II) affinities than PBT2, sequestering most (~83%) of the Cu(II) from Aβ(1-42). PBT2 has a lower relative Cu(II) affinity compared with CQ and B2Q, sequestering only approximately 59% of the Cu(II) from Aβ(1-42). Interestingly, the fraction of Cu(II) that remains bound to Aβ(1-42) appears to belong to a single Cu(II)Aβ(1-42) species. This result is interesting because Cu(II)Aβ(1-42) is known to bind Cu(II) in at least two different Cu(II) coordination environments at pH 7.4 (components I and II). It is possible that PBT2 may sequester Cu(II) from only one of these components, which may have a lower relative Cu(II) affinity than the other component. Alternatively, lower solvent accessibility of the Cu(II) in at least one of the components may also play a role in reducing Cu(II) sequestration. The bulkier 2-position substituent of PBT2 may impact the ability of this 8HQ to access a less solvent accessible Cu(II) site in Aβ(1-42). We do not find strong evidence to support the formation of 8HQ-Cu(II)-Aβ(142) in contrast to some previous studies.

4. Materials and Methods

4.1. Preparation of Aβ(1-42)

Aβ(1-42) was prepared and purified as described previously.¹⁰³ Briefly, BL21 cells expressing a plasmid encoding Aβ(1-42) were grown in M9 minimal media and induced at OD = 0.8 with isopropyl β-D-1-thiogalactopyranoside (IPTG). Cells were pelleted and lysed, and inclusion bodies were collected, washed, and treated with 8 M guanidine hydrochloride to extract the peptide. The peptide extract was treated with hexafluoroisopropanol (HFIP) to monomerize then purified using size exclusion chromatography (250mm × 4.6mm, C18, 300 μm pore size) using a one-hour gradient from 0.1% aqueous TFA to 50% acetonitrile with 0.1% TFA. Purified Aβ(1-42) eluted around 30 minutes. The Aβ(1-42) peptide concentration was determined using the UV−visible absorption of tyrosine at 280 nm. Aβ(1-42) was obtained in high purity (Figure S4), again treated with HFIP, and lyophilized. Aβ(1-42) was weighed and stored as aliquots of ~0.4 mg of dry powder at −80 °C.

4.2. Synthesis of PBT2

PBT2 (5,7-dichloro-2-[(dimethylamino)methyl]-8-hydroxyquinoline) was synthesized as published previously.^{105, 154} Briefly, triethylamine was slowly added to a solution of 5,7-dichloro8-hydroxyquinoline-2-carboxaldehyde, followed by sodium triacetoxyborohydride, and the mixture was stirred at room temperature overnight. Dichloromethane was added, the mixture was washed with sodium bicarbonate, dried, and concentrated. The resultant residue was extracted with diethyl ether and concentrated. HCl was added and the mixture was concentrated in vacuo. After washing with dichloromethane, PBT2•HCl was obtained as a pale straw-coloured solid (~15–30%; Figure S3).

4.3. XAS Sample Preparation

A stock solution of dodecyl trimethylammonium bromide (DTAB) was prepared in diH₂O to a final concentration of 400 mM in 100 mM MOPS, pH 7.4. The solution was then heated to ~60°C, sonicated for ~10 min, and allowed to cool. Lyophilized Aβ(1-42) was dissolved in this aqueous buffer to a final concentration of 0.54 mM; aqueous Cu(II) from a stock solution of 100 mM copper chloride in diH₂O was added to the peptide solution to a final concentration of 0.50 mM (1 Aβ: 1 Cu(II)).

Stock solutions of 10 mM CQ and B2Q were prepared in neat DMSO because of the high hydrophobicity of the compounds.¹⁰⁴ A stock solution of 10 mM PBT2 was prepared in 400 mM DTAB in 100 mM MOPS buffer at pH 7.4 because of the relatively low hydrophobicity of the hydrochloric salt of PBT2. Two molar equivalents of each 8HQ were added to an aliquot of Cu(II)Aβ(1-42) in 400 mM DTAB in 100 mM MOPS buffer at pH 7.4. The final concentration of DMSO was 10% in solutions with CQ and B2Q; no DMSO was present in solutions with PBT2. Glycerol was not added as a glassing agent due to its propensity to accelerate photoreduction in the X-ray beam.^{103, 143, 155} Instead, samples were flash-frozen to minimize the formation and growth of ice crystals. Aqueous samples were loaded into 3 mm acrylic sample cuvettes sealed with metal-free tape, flash frozen in a slurry of isopentane cooled in liquid nitrogen, and stored in liquid nitrogen prior to experimentation. Mixing, loading, and flash freezing of the samples was accomplished in < 60 s for all samples to minimize the potential for Aβ oligomerization. We expect the mixture of Cu(II), Aβ(1-42), and 8HQ to have reached equilibrium in a significantly shorter time.

4.4. Synchrotron X-Ray Absorption Spectroscopy

Cu K-edge data were collected on the biological XAS beamline (BL) 7–3 at the Stanford Synchrotron Radiation Lightsource (SSRL) at SLAC National Laboratory in Menlo Park, California, USA using the data acquisition program, XAS-Collect.¹⁵⁶ The SPEAR storage ring contained 500 mA at 3.0 GeV. Beamline 7–3 utilizes a Si(220) double crystal monochromator and a rhodium-coated vertically collimating mirror upstream of the monochromator, which achieves harmonic rejection by adjusting the mirror cut-off angle (e.g. Cu K-edge spectra are collected at the 12 keV cut-off). Harmonic rejection was alternatively achieved by detuning the monochromator crystal by 60%. Samples were maintained at 45° to the incident beam at ~10 K using a liquid helium flow cryostat (Oxford instruments, Abingdon, UK) during data collection. Incident and transmitted X-rays were measured using nitrogen-filled gas ionization chambers. XAS was collected in fluorescence mode by monitoring the Kα fluorescence using a 30-element germanium detector at 90° to the incident beam.

The monochromator energy was calibrated through reference to a standard foil measured simultaneously with the sample. Specifically, the first energy inflection of the Cu foil was used to calibrate the spectrum energy to 8980.3 eV. Each data set is a collection of a minimum of 5 scans and extends to a minimum k range of 15 Å⁻¹. To decrease the risk of photoreduction of Cu(II) compounds, samples were scanned at multiple positions on the same sample (i.e., using a 0.5 × 12 mm beam and 0.5 mm vertical movements) and the near-edges of successive scans were compared for loss of the 1s→3d transition indicative of photoreduction.

XAS data were analyzed using the EXAFSPAK set of computer programs (https://www-ssrl.slac.stanford.edu/exafspak.html), as previously described.¹⁵⁷ Quantitative analyses utilized FEFF825^158–159 to calculate the ab-initio theoretical phase and amplitude for use by the curve-fitting program, OPT (a component of EXAFSPAK).

4.5. Synchrotron High Energy Resolution Fluorescence Detected X-Ray Absorption Spectroscopy

Near-edge HERFD XAS measurements were carried out on BL 6–2 at SSRL with the SPEAR3 storage ring containing 500 mA at 3.0 GeV as previously described.^{103–105, 143} Briefly, a Si(311) double crystal monochromator was used with a 6-element array of Si(444) crystal analyzers to record the Cu Kα₁ emission line.¹⁶⁰ Harmonic rejection was achieved by setting the cut-off energy of the upstream Rh-coated mirror to 18 keV. Incident and transmitted X-rays were monitored using helium- and nitrogen-filled gas ionization chambers, respectively. Aluminum filters upstream of the incident ion chamber were used to minimize X-ray exposure and thus minimize the resultant photodamage. Samples were maintained at 10 K using a liquid helium flow cryostat (Oxford instruments, Abingdon, UK) and were inclined at 45° to the incident X-ray beam. Energy calibration of the incident beam monochromator was determined relative to the lowest-energy inflection of a copper foil, which was taken to be 8980.3 eV. Data reduction and analyses were carried out as previously described using the EXAFSPAK suite of computer programs.¹⁶¹

4.6. Near-Edge Fitting

A near-edge spectrum of a mixture of species is the sum of individual component spectra, where the height of the normalized edge of each component is proportional to the fraction of the element in that form. Least squares fitting was used to quantitatively analyze the proportion of each spectrum in mixtures of Aβ(1-42) with 8HQs. In least-squares fitting, the function F is minimized as:

$$F=\frac{1}{N}\sum _{j=1}^N{(y_{j,obs}-y_{j,\mathit{\text{calc}}})}^2$$ (1)

where the calculated intensity is given by:

$$y_{j,\mathit{\text{calc}}}=\sum _{i=1}^m{x}_i{I}_{i,j}$$ (2)

The parameters are defined as follows: N is the total number of energy points, j is an energy point number, y_i,obs is the observed intensity of a normalized mixed spectrum at energy j, y_i,calc is the intensity of the calculated spectrum at energy j, m is the number of components, i is the component number, x_i is the proportion of the element of interest in component i and I_i,j, is the normalized intensity of the spectrum of component i at energy j. Fractions of Aβ(1-42) and 8HQs are reported as a percentage with the estimated standard deviation (e.s.d.) (Section 2.2; Figure 3).

4.7. EXAFS Curve-Fitting

Geometry optimized structures of copper CQ and B2Q complexes were generated as previously described.¹⁰⁴ The calculated paths were then used in the curve-fitting program, OPT, in EXAFSPAK to fit the data. The multiplicity for each multiple scattering path (N), was set to an integer value predicted by the model. Fitting parameters included the following: distance between absorbing atom (i.e., Cu) and the backscattering atom (R), mean square deviation in R (Debye−Waller factor; σ²), and the energy offset to the threshold energy (ΔE₀). For the Cu K-edge, E₀ is assumed to be 9000 eV. R parameters that differed by <0.1 Å were fit as a group of related, linked paths, where the percent change in each parameter was held to be the same. The σ² values for each distance, R, were also refined as a group of related paths, where the percent change in each parameter was held to be the same, for the entire structural model. ΔE₀ was estimated through comparisons to Cu(II) standards and was permitted to float only in final refinements of the fit.

Histidine parameters were calculated as previously described, using the EXAFSPAK program, HIS-FLAP.¹⁰³ A geometry optimized structure for the 2 N 2 O coordination environment in Cu(II)-bis-PBT2 was generated as previously described.¹⁰⁵ For fitting the CuAβ(1-42) + PBT2 spectrum, the N for each scattering path was reduced to the appropriate fraction based on the proportion of each complex (i.e. 59% Cu(II)-bis-PBT2 and 41% Cu(II)Aβ) indicated by the near-edge fit. The R values and ΔE₀ were refined as a set of linked parameters. The σ² values were maintained from previous fitting of Cu(II)A_β¹⁰³ and Cu(II)-bis-PBT2¹⁰⁵ and were not further refined.

The fit error function F is defined as

$$F={\left\{\sum k^6{({\chi }_{\mathit{\text{calc}}}-{\chi }_{\mathit{\text{expt}}})}^2/\sum {\chi }_{\mathit{\text{expt}}}^2\right\}}^{1/2}$$ (3)

Noise contributions with frequencies higher than the longest bond-length were estimated as previously described.¹⁶² Fitting parameters for each best fit model are detailed in Table S1.

4.8. Electron Paramagnetic Resonance

X-band EPR spectra were collected on samples of 0.56 mM Aβ(1-42) and 0.49 mM ⁶³CuCl₂ in 400 mM DTAB buffered with 100 mM MOPS to pH 7.4 and with 25% glycerol. Aβ(1-42) was from the same batch as was used for the XAS measurements, with the addition of glycerol as a glassing agent. Two molar equivalents of PBT2 were added to the solution of Cu(II)Aβ(1-42) and spectra were collected. X-band EPR spectra of Cu(II)PBT2 were also collected under the same solution and conditions for comparison. Isotopically enriched ⁶³Cu was used to avoid broadening of resonance lines because of the different magnetic moments of naturally abundant isotopes (69% ⁶³Cu, g_n = 1.4840; 31% ⁶⁵Cu, g_n = 1.5900).

X-band EPR (~9.43 GHz) used a Bruker EMX spectrometer with a Bruker ER 4122SHQE resonator. Typical spectrometer conditions were 0.1 mT modulation amplitude, 5 mW applied microwave power and a temperature of 121 K. The spectrum was an average of eight 83.89 second sweeps, with 2048 points per scan. EPR powder lineshape simulation used a modified version of the program QPOW.^163–164

4.9. Density Functional Theory Calculations

Density functional theory (DFT) calculations utilized Dmol³ Materials Studio 7.0, as previously described.¹⁰⁴ Briefly, Dmol³ geometry optimization calculations employed the generalized gradient approximation,¹⁶⁵ and Becke Exchange,¹⁶⁶ Perdew,¹⁶⁷ and Lee, Yang, and Parr functionals.¹⁶⁸ Optimization calculations used all-electron core treatment and were calculated in vacuo (i.e. without a solvent reaction field). Calculations used symmetry parameters where appropriate to define the point group of the complexes.

Table 2.

EPR simulation parameters for Cu(II)-bis-PBT2 and the spectrum resulting from the subtraction of Cu(II)-bis-PBT2 from Cu(II)Aβ(1-42) + PBT2.^a

Cu(II) Complex	Solution Conditions		g ⊥
		g ∥	g x	g y	A∥ (mHz)	A⊥ (mHz)	Ref.
Cu(II)-bis-PBT2	400 mM DTAB, 100 mM MOPS, pH 7.4 400 mM	2.272	2.062	2.052	440	10	Summers et al.105
Cu(II)Aβ(1-42) *	DTAB, 100 mM MOPS, pH 7.4	2.242	2.072	2.015	425	10	This work

^aValues determined in this study are shown in bold; errors are estimated to be ±0.005 for g_∥ and ±0.002 for g_⊥; estimated errors for A_∥ and A_⊥ are ±9 and ±2, respectively;

^*From simulations of spectra d and e in Figure 4. Hyperfine A strain broadening in A_x and A_y of 80 mHz was also required to best simulate the data.

Synopsis.

8-Hydroxyquinoline-based chelators interact with and chelate Cu(II), as well as other metals and compounds. When 8-hydroxyquinoline-based chelators, CQ and PBT2 were added to aqueous solutions of the amyloid beta peptide, CQ and B2Q were found to sequester ~83% of the Cu(II) from Cu(II)Aβ(1-42), whereas PBT2 sequestered only ~59% of the Cu(II). Based on the results of X-ray absorption spectroscopy and EPR, it appears that the Cu(II) in a single Cu(II)Aβ(1-42) species is inaccessible to PBT2.