The nickel-chelator dimethylglyoxime inhibits human amyloid beta peptide in vitro aggregation

One of the hallmarks of the most common neurodegenerative disease, Alzheimer’s disease (AD), is the extracellular deposition and aggregation of Amyloid Beta (Aβ)-peptides in the brain. Previous studies have shown that select metal ions, most specifically copper (Cu) and zinc (Zn) ions, have a synergistic effect on the aggregation of Aβ-peptides. In the present study, inductively coupled plasma mass spectrometry (ICP-MS) was used to determine the metal content of a commercial recombinant human Aβ40 peptide. Cu and Zn were among the metals detected; unexpectedly, nickel (Ni) was one of the most abundant elements. Using a fluorescence-based assay, we found that Aβ40 peptide in vitro aggregation was enhanced by addition of Zn2+ and Ni2+, and Ni2+-induced aggregation was facilitated by acidic conditions. Nickel binding to Aβ40 peptide was confirmed by isothermal titration calorimetry. Addition of the Ni-specific chelator dimethylglyoxime (DMG) inhibited Aβ40 aggregation in absence of added metal, as well as in presence of Cu2+ and Ni2+, but not in presence of Zn2+. Finally, mass spectrometry analysis revealed that DMG can coordinate Cu or Ni, but not Fe, Se or Zn. Taken together, our results indicate that Ni2+ ions enhance, whereas nickel chelation inhibits, Aβ peptide in vitro aggregation. Hence, DMG-mediated Ni-chelation constitutes a promising approach towards inhibiting or slowing down Aβ40 aggregation.

www.nature.com/scientificreports/ to the intrinsic presence of metallic ions, including Cu 2+ , Ni 2+ and Zn 2+ , as revealed by the ICP-MS metal analysis conducted in the present study (see above). Addition of 10 µM NiSO 4 to the mixture increased the average aggregation rate by 2.5-fold (Table 1), while addition of 100 µM NiSO 4 resulted in a dramatic 5.7-fold increase compared to the no supplemental metal control, suggesting the divalent cation Ni 2+ can bind to the Aβ 40 peptide and enhance its aggregation ( Fig. 1 and Table 2). A similar effect was observed when NiCl 2 was used (instead of NiSO 4 ) as source of Ni 2+ (data not shown); hence, the nature of the counterion does not appear to play a role in (or interfere with) the observed aggregation. Upon addition of 10 and 100 µM Zn(II), a fivefold and 14-fold increase in Aβ 40 peptide aggregation rate was observed compared to the control, respectively (Table 2), in agreement with previously published studies [27][28][29] ; in contrast, addition of 10 or 100 µM CuSO 4 had no significant effect on the aggregation rate (Table 2). This result (lack of aggregation) could be due to the pH used in our study    (Table 2). Furthermore, addition of 500 µM or 1000 µM DMG led to partial or full inhibition of the aggregation; in the latter case, we measured flat or even decreasing average fluorescence rates (reported as < 1% of control, Table 2). Hence this dose-dependent inhibitory effect suggests that (i) DMG is able to pull metals away from the Aβ 40 peptide and (ii) the Aβ 40 peptide aggregation observed in absence of supplemented metals is likely due to the intrinsic presence of metallic ions (including Ni 2+ ) within the recombinant peptide preparation, since the addition of the chelator leads to inhibition. When increasing amounts of DMG were added to the reaction mixture in presence of 100 µM Ni 2+ , Cu 2+ or Zn 2+ , results with mixed outcomes were obtained. Complete inhibition was observed in presence of Ni ( Fig. 1 and Table 2) and only partial inhibition was seen in presence of Cu or Zn, however Zn was still able to induce Aβ 40 peptide aggregation ( Table 2). The respective efficacy (or lack thereof) of DMG in presence of Cu, Ni and Zn correlates with the chelator's respective affinity for each metal, as revealed by mass spectrometry analysis of metal-DMG complexes (see below).

Effect of pH on Aβ 40 peptide aggregation in presence of metals or DMG.
To study the effect of pH on Aβ 40 peptide aggregation (in presence of metals or DMG), additional ThT-based aggregation experiments were conducted at pH 6.5, 7.5 or 8.5, with 25 µM Aβ 40 , in absence or presence of NiSO 4 (25 µM), CuSO 4 (25 µM), ZnSO 4 (10 µM), or DMG (100 µM) ( Fig. 2 and data not shown). Overall, aggregation rates at alkaline pH 8.5 (black symbols) were lower compared to pH 7.5 (grey symbols) or pH 6.5 (white symbols). As previously observed, addition of Zn(II) led to the fastest and sharpest increase in fluorescence (triangles) under all pHs tested. Interestingly, Ni(II)-induced aggregation (squares) was faster at pH 6.5, compared to pH 7.5, while it was absent at pH 8.5 (black squares). Aβ 40 peptide aggregation in presence of Cu(II) or DMG was negligible under all 3 pH conditions tested (data not shown).
Nickel binding to human recombinant Aβ 40 peptide is confirmed by isothermal titration calorimetry. Isothermal titration calorimetry (ITC) has been already used to analyze copper or zinc binding to various Aβ peptides, including Aβ 40 [51][52][53] . In the current study, we used ITC to determine whether nickel can bind to the Aβ 40 peptide. The peptide (same used in ThT-based aggregation assays) was present in the sample cell at a concentration of 20 μM. Twenty injections of NiSO 4 (1 mM solution, 5 μM increments in sample cell) were performed every 5 min under constant stirring (350 rpm) at 25 °C, and the heat release was measured (Fig. 3). The heat release profile indicates Ni binding to the peptide (Fig. 3, top Panel). The best fit of Ni titration (Fig. 3, bottom Panel) suggests an apparent stoichiometry of less than 1 mol Ni(II) per mole of Aβ 40 peptide (~ 0.7), in range with previously reported stoichiometry ratios of 1:1 for Cu(II) or Zn(II), and Aβ 40 54 . The apparent K d value for Ni is approximately 4.2 μM, similar to that previously reported of 7 ± 3 μM for Zn 53 . Furthermore, the  www.nature.com/scientificreports/ ΔH (enthalpy) and ΔS (entropy) were found to be − 5 kJ/mol and 86 mol/J/K, suggesting the Ni-Aβ 40 binding event can be considered both exothermic and spontaneous. Injection of DMG (instead of nickel) into the sample cell containing Aβ 40 peptide did not induce any significant change, indicating that DMG cannot bind to the peptide (data not shown). Hence, this result suggests the inhibitory effect of DMG on Aβ 40 peptide aggregation, as observed with ThT-based assays, is due to nickel chelation, rather than direct DMG-Aβ 40 peptide inhibitory interaction.  Fig. S1).

Analysis of DMG in brain samples using FTICR-MS and NMR. The fact that DMG inhibits Aβ 40
peptide aggregation in vitro suggests it might be able to do the same in vivo, however DMG would first need to cross the blood-brain barrier (BBB). To determine whether DMG can localize to the mouse brain, we used FTICR-MS (see above) and Nuclear Magnetic Resonance (NMR). NMR was successfully used to detect DMG in the livers of mice subjected to daily oral doses (6. 1 mg) of aqueous DMG for 3 days 55 . In the present study, the same treatment was administered (e.g. one daily oral delivery for 3 days), brain samples were processed and analyzed by NMR and FTICR-MS and compared to brain samples from (no DMG) control mice. Unfortunately, both methods failed to identify DMG (whether by itself or metal-chelated) in brain samples.

Discussion
To study Aβ peptide in vitro aggregation, one can choose to use either chemically synthetic peptides, or recombinant peptides, expressed in, and purified from, organisms such as E. coli. Synthetic Aβ preparations have been associated with various problems, such as presence of impurities in the preparation, incorporation of the L-form of amino-acids (e.g. D-His, D-Met, D-Arg) instead of the L-form during synthesis, or reproducibility issues in terms of quality and yield, to a point that even batch-to-batch variations have been reported [56][57][58] . On the other  www.nature.com/scientificreports/ hand, the expression and purification of recombinant Aβ peptides also bring their own limitations and issues, including low yield, reduced solubility and presence of oxidized amino-acids (e.g. Met 35 -sulfoxide) 58 . One major difference between synthetic and recombinant Aβ peptides though, often overlooked in the literature, is the absence and presence of metals associated with each preparation, respectively. Indeed, any protein or peptide showing natural affinity for one (or several) metal(s), as it is the case with Aβ peptides for copper or zinc 24 , will likely encounter (and bind to) the metals within the host (E. coli or other hosts). Hence recombinant Aβ peptides are likely to be already associated with metals upon purification, in contrast to synthetic peptides. Given that both Cu and Zn enhance Aβ peptide aggregation, one can expect that Cu or/and Zn-containing recombinant Aβ peptide will be "naturally" more prone to aggregation than their synthetic counterparts. This could account for differences reported in a study by Finder and coworkers, who found that recombinant Aβ 42 peptides (likely metal-bound) aggregated faster and were more neurotoxic than synthetic Aβ 42 peptides (likely metal-depleted) 58 40 aggregation, whereas DMG-mediated Ni-chelation inhibits it . Moreover, Ni(II) was found to be more efficient than Cu(II), and less efficient than Zn(II), respectively, at promoting Aβ 40 aggregation, under the conditions tested in our study. Since various parameters (such as pH and temperature) have been previously shown to have an effect on metal-induced aggregation 30,59 , we tested the effect of pH on Ni-dependent aggregation. Three buffers with similar salt content (192 mM NaCl) but various pHs (6.5, 7.5, or 8.5) were used. Interestingly, acidic pH (6.5) conditions increased Ni-induced aggregation compared to the control pH (7.5), www.nature.com/scientificreports/ whereas Ni-induced aggregation was abolished at pH 8.5. The increased Ni-induced aggregation at acidic pH, as observed in the present study, is in agreement with previous published data from Atwood et al., who reported an increase of Aβ 1-40 aggregation in presence of 1 μM Ni at pH 6.6, compared to pH 7.4 30 . Likewise, the same study correlated acidic pH (6.6) with enhanced aggregation, in presence of either Cu or Zn (both at 20 μM). Herein, Zn-induced aggregation was slightly higher at pH 7.5 compared to 6.5, and significantly faster compared to pH 8.5. The effect of pH on Cu-induced aggregation was negligible, but it is worth noting that the effect of Cu was very limited throughout our ThT-based assays, for a reason yet to be determined. The effect of temperature on Ni-dependent Aβ 40 aggregation was not tested with the fluorescence-based method, as all assays were carried out at 37 °C. However, Ni-Aβ 40 binding was also observed at 25 °C, as shown by ITC (see below). Although results obtained with both methods cannot be directly compared nevertheless we can report that Ni binding (to Aβ40) happens both at 25 °C and 37 °C. Since aggregation in presence of a particular metal (e.g., nickel) suggests initial metal-peptide binding, we further investigated the likelihood of Ni binding to Aβ 40 , by using ITC. The calorimetry-based method has been successfully used in the past to study Zn binding to Aβ 40 , both at low (10 μM) and high (70 μM) concentrations 53 . In the current study, we only looked at the effect of Ni on low Aβ 40 concentration, with a starting concentration of Aβ 40 in the sample cell at 20 μM. After Ni was injected via 20 consecutive injections, every 5 min (5 μM increments), a heat profile characteristic of independent metal-binding was observed. Although the apparent K d (4.2 μM) is similar to that reported for Zn 53 , the apparent stoichiometry (0.7 mol of Ni per mole of Aβ 40 ) is significantly lower than that previously reported by Drochioiu and colleagues, who found that synthetic Aβ 40 peptide displays high affinity toward nickel ions with up to three Ni 2+ ions bound per Aβ 40 peptide 60 .However, the discrepancy between our results and theirs could be due to the nature of Aβ 40 peptide used, and the type of analytical method used to analyze Ni-Aβ 40 . In our study, we used a purified recombinant Aβ 40 peptide, and ITC, whereas Drochioiu et al. used synthetic Aβ 40 peptide, electrospray ion trap mass spectrometry (ESI-MS) and circular dichroism (CD). Nevertheless, results from both groups indicate that Aβ 40 can bind nickel with high affinity. Furthermore, our results confirm that DMG inhibits Aβ 40 aggregation through Ni chelation (not direct contact with the peptide), since titration of the peptide with DMG did not induce any peptide conformational change, as observed with ITC.
Given the presence of Cu 2+ , Ni 2+ , Zn 2+ in recombinant Aβ 40 peptide, combined to their respective effect on Aβ 40 peptide aggregation, metal chelation therapy towards AD constitutes a valid approach. However, the risk of chelation therapy is that removal of essential metal ions will lead to serious adverse effects (for instance, irondeficiency anemia) as pointed out by other researchers 35 . Hence it is preferable to use chelators with select affinity towards non-essential metals: the Ni-specific chelator DMG is therefore a good candidate. Indeed, DMG has been used for many years to detect, quantitate or decrease Ni levels in various environments; it can also be used to inhibit the growth of bacteria, including multidrug resistant Enterobacteriaceae, as recently demonstrated by our group 55 . Mammalian hosts do not contain known Ni-dependent enzymes, which makes Ni-chelation therapy an attractive approach 47 . On the other hand, most bacteria, including pathogenic ones, require nickel as cofactor for one or several enzymes, such as [Ni-Fe] hydrogenase(s) 61 or urease 47 . Thus, DMG-mediated inhibition of these enzymes, as demonstrated with Salmonella Typhimurium hydrogenases or Klebsiella pneumoniae urease, leads to bacterial growth inhibition, both in the mouse and in the wax moth animal models 55 . In the present study, we showed that DMG can drastically reduce, and even abolish Aβ 40 peptide aggregation. The inhibitory effect was observed in absence of supplemental metal, as well as in presence of copper, nickel or even zinc (albeit with lower efficacy). Although thioflavin is a popular reporter of amyloid aggregation, it mostly binds to β-sheet rich fibrils 62,63 . Therefore, our conclusions on the effect of Ni and Ni-chelation (DMG) on Aβ peptide-aggregation must be limited at this time to the β-sheet content. Additional experiments will be needed to determine whether Ni and its chelator have a broader effect on other Aβ peptide conformations. Likewise, further experiments will be conducted to test whether DMG can inhibit the aggregation of other physiologically relevant Aβ peptides, for instance Aβ 42 .
The current study is not the first one to report inhibitory effect of a nickel chelator on Aβ peptide aggregation. Indeed, Reinhardt and colleagues reported beneficial effects of the nickel chelator disulfiram on AD hallmarks, including inhibitory effects on Aβ 42 peptide aggregation 64 . The study however was not aimed at establishing any link between Ni and AD. The authors found that disulfiram increased synthesis of the metalloproteinase α-secretase, resulting in secretion of the neuroprotective APP cleavage product sAPPα and thus preventing formation of the amyloidogenic βA peptides 64 . The concentration of disulfiram shown to have inhibitory effects on peptide aggregation was significantly lower compared to DMG concentrations used in our study, however the disulfiram drug is highly toxic, even at low doses, with concentrations higher than 5 μM inducing cytotoxicity 64 . This finding correlates with previous studies linking disulfiram with negative outcomes, such as elevated nickel levels in rat brains 65 , elevated nickel levels in body fluids of patients with chronic alcoholism 65,66 , as well as hepatotoxicity in humans 67,68 .
If DMG were to be used in a clinical trial against AD, it might not only inhibit Aβ plaque formation, but also Ni-requiring microorganisms. This, however, would not necessarily be a negative outcome, in light of the link between pathogens and AD, known as the "infection hypothesis of AD" [69][70][71] . The list of pathogens potentially linked to AD includes viral, fungal and bacterial species. Among bacteria directly or indirectly associated with AD, one can find Helicobacter pylori 72,73 , E. coli 74 and Salmonella Typhimurium 75 , all of which require Ni as cofactor for one or several enzymes (for a review, see Maier and Benoit 47 ). In the case of H. pylori, another protein is relevant to the pathogen/AD link. The gastric pathogen produces abundant amounts (2% of total protein) of a small histidine-rich protein (Hpn) that has been shown to develop amyloid-like fibrils in vitro 76 . The continuous production of Hpn by the bacterium during decades of chronic gastric infection could result in leakage of the protein, first into the bloodstream and eventually into the brain, potentially triggering AD, as hypothesized by Ge and Sun 77 . More generally, an antimicrobial role for Aβ peptides (as part of the brain's ancient immune system)  75 .
In summary, Ni could affect the onset and the progression of AD through two different mechanisms, as depicted in our proposed model (Fig. 4). The first mechanism involves the binding of Ni 2+ to Aβ (Aβ 40 , possibly Aβ 42 ), eventually leading to aggregation, plaque formation and AD; this would comply with the metal hypothesis of AD. The second mechanism involves the use of Ni 2+ as a required cofactor for various enzymes (e.g. Niglyoxalase, Ni-superoxide dismutase, Ni-acireductone dioxygenase, [NiFe] hydrogenases and urease, see 47 ) of pathogens previously shown to play a role in Aβ peptide aggregation; alternatively some of these pathogens might contribute to AD independently of Aβ plaque formation. Both scenarios would fit the "infectious hypothesis of AD". Whether one Ni-dependent mechanism is preferred over the other, or both actively contribute to the onset and/or the progression of AD, nevertheless a DMG-mediated Ni-chelation strategy is at the intersection of both (the metal and the infectious) hypotheses. Thus, it is likely to interfere and disrupt both mechanisms, eventually slowing down or stopping the progression of AD. More than a century after Alois Alzheimer fist described the disease, and without any cure on the horizon, new therapeutic strategies are urgently needed to combat this neurodegenerative disease; we believe that nickel chelation (via DMG treatment) is a promising AD-combatting strategy that warrants further research. Amyloid beta metal analysis. Metal levels for 20 elements (Li, Be, Al, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Rb, Sr, Cd, Cs, Ba, Pb, U) were determined by inductively coupled plasma mass spectrometry (ICP-MS). Briefly, 0.5 mg of lyophilized human recombinant Aβ 40 peptide (expressed in E. coli, purified and manufactured by rPeptides, Watkinsville, GA) was resuspended in ultrapure water to a final concentration of 5 mg/mL, digested overnight with concentrated trace metal grade nitric acid, heated for 2 h at 95 °C and subjected to ICP-MS using a Thermo X-Series II ICP-MS (Center for Applied Isotope Studies, University of Georgia, Athens, GA). The same treatment and analysis were performed on three other kit components: TBS 10X (reaction buffer), Thioflavin stock (400 μM) and NaOH 10 mM (used to resuspend the Aβ 40 peptide).

Chemicals
Amyloid beta aggregation. The effect of metals, DMG, and/or pH on human recombinant Aβ 40 aggregation was monitored using a thioflavin T (ThT)-based kit, following the manufacturer's recommendation (kit# A-1180-1, rPeptides, Watkinsville, GA, USA). This kit contains human recombinant Aβ 40 peptide (> 97% pure, as determined by manufacturer's HPLC) with the following sequence DAEFRHDSGYEVHHQKLVFFAEDVG-SNKGAIIGLMVGGVV, as provided by the manufacturer. Briefly, standard assays were conducted in triplicate in black polystyrene 96-well plates, in presence of Tris Buffer Saline (TBS) pH 7.4, or TBS pH 8.  Ni can bind to Aβ peptides, leading to aggregation and plaque formation (left side, metal hypothesis of AD). In addition, Ni is required as cofactor for enzymes (such as hydrogenase and urease) of pathogens previously shown to play a role in Aβ peptide aggregation (right side, infection hypothesis of AD). The Ni-chelator DMG could inhibit Aβ peptide aggregation and the progression of AD (red crosses), either directly (left side) or indirectly (through pathogen inhibition, right side).  4 OH 0.2%, TBS 1X, pH 7.4 ("NTBS")). A volume of 500 μL was loaded onto the ITC sample cell, and the injection syringe was filled with 50 μL of either NTBS buffer (control), 1 mM NiSO 4 , or 1 mM DMG. All samples were degassed for 15 min at 25 °C before use. Titration was initiated using a program for 20 injections (2.38 μL each, every 5 min) with continuous stirring (350 rpm) at constant temperature (25 °C). ITC data were analyzed using NanoAnalyze 1.2 software (TA Instruments). Data obtained with the control experiment (Aβ 40 peptide in sample cell, buffer in syringe) were subtracted from each experiment to account for any injection-related heat change. The Ni-Aβ 40 experiment was done in triplicate, with a representative data set shown in figures. Detection of DMG and DMG-metal complexes in mouse brains. All procedures were performed in accordance with the relevant guidelines and regulations and approved by the University of Georgia IACU committee, and the study was carried out in compliance with the ARRIVE guidelines (http:// www. nc3rs. org. uk/ page. asp? id= 1357). A group of 6 (C57/BL) mice was used for this experiment: 3 mice were given 0.2 mL of 100 mM DMG (~ 6.1 mg) every day for three days and 3 mice were used as (no DMG) controls. Mice were euthanized by CO 2 asphyxiation and cervical dislocation. Brains were quickly removed and frozen at − 80 °C. Upon thawing, brains were cut into pieces and homogenized in 2 mL sterile deionized water, incubated for 1 h at 90 °C, sonicated for 20 s and spun down (16,800 × g for 6 min). Supernatants were passaged through a 0.45 μm filter unit and analyzed by FTICR-MS (see above) and Nuclear Magnetic Resonance (NMR), as previously described 55 .