Amyloid β Peptides Do Not Form Peptide-derived Free Radicals Spontaneously, but Can Enhance Metal-catalyzed Oxidation of Hydroxylamines to Nitroxides*

Amyloid β (Aβ) peptides play an important role in the pathogenesis of Alzheimer’s disease. Free radical generation by Aβ peptides was suggested to be a key mechanism of their neurotoxicity. Reports that neurotoxic free radicals derived from Aβ-(1–40) and Aβ-(25–35) peptides react with the spin trapN-tert-butyl-α-phenylnitrone (PBN) to form a PBN/⋅Aβ peptide radical adduct with a specific triplet ESR signal assert that the peptide itself was the source of free radicals. We now report that three Aβ peptides, Aβ-(1–40), Aβ-(25–35), and Aβ-(40–1), do not yield radical adducts with PBN from the Oklahoma Medical Research Foundation (OMRF). In contrast to OMRF PBN, incubation of Sigma PBN in phosphate buffer without Aβ peptides produced a three-line ESR spectrum. It was shown that this nitroxide is di-tert-butylnitroxide and is formed in the Sigma PBN solution as a result of transition metal-catalyzed auto-oxidation of the respective hydroxylamine present as an impurity in the Sigma PBN. Under some conditions, incubation of PBN from Sigma with Aβ-(1–40) or Aβ-(25–35) can stimulate the formation of di-tert-butylnitroxide. It was shown that Aβ peptides enhanced oxidation of cyclic hydroxylamine 1-hydroxy-4-oxo-2,2,6,6-tetramethylpiperidine (TEMPONE-H), which was strongly inhibited by the treatment of phosphate buffer with Chelex-100. It was shown that ferric and cupric ions are effective oxidants of TEMPONE-H. The data obtained allow us to conclude that under some conditions toxic Aβ peptides Aβ-(1–40) and Aβ-(25–35) enhance metal-catalyzed oxidation of hydroxylamine derivatives, but do not spontaneously form peptide-derived free radicals.

As the leading cause of dementia, Alzheimer's disease (AD) 1 is characterized by a loss of memory and neurons. These characteristics have been associated with brain lesions known as neurofibrillary tangles composed of Tau protein and amyloid plaques, which consist of amyloid ␤ (A␤) peptide. In a small number of families, mutations in the genes for the amyloid peptide precursor (APP, chromosome 21) to A␤ peptide for presenilin-1 (PSEN-1, chromosome 14), presenilin-2 (PSEN-2, chromosome 1) and for an anonymous gene on chromosome 12 have been associated with AD. In contrast, Saunders and colleagues (1) discovered that a significant percentage of patients inheriting one or more of the epsilon-4 alleles of apolipoprotein-E (APOE4, chromosome 19) were at risk of acquiring AD at an earlier age than their counterparts expressing the more common epsilon-3 allele of APOE. The observation that AD patients with mutant genes for APP, PSEN-1, or PSEN-2 have higher levels of A␤ peptide than non-diseased controls (2,3) argues that A␤ peptide may cause AD. Further supporting this idea is the observation that APOE4 patients with AD display more amyloid plaques than those with APOE3 (4). Based on neuropathological, genetic, and biochemical associations, the presence of amyloid ␤ peptide is a key component of AD.
A crucial link between A␤ peptide and AD was provided by Yankner and colleagues (5), who showed that fibrillar aggregates of A␤ peptide were toxic to neurons. This seminal observation inspired a global quest to define the mechanism by which fibrillar A␤ peptide mediated neurotoxicity. One of the mechanisms proposed suggests that neurons exposed to A␤ peptide suffer severe oxidative stress that may lead to their death. Clear evidence of oxidative stress in Alzheimer's disease has been provided by Smith et al. (6), who found that neurons of AD patients contained nitrotyrosine modifications of proteins, which were not detected in age-matched control brains. Increased levels of lipid peroxides (7), reactive aldehydes such as hydroxynonenal (8), and oxidized DNA (9) have also been reported in AD brains, providing additional evidence of an oxidative stress component in the disease.
The exact nature of the radical species generated in Alzheimer's diseased brains is unknown. In one line of investigation, Behl et al. (10) demonstrated that cells exposed to fibrillar A␤ peptide responded by releasing hydrogen peroxide and dying, a process that could be inhibited by the application of catalase to degrade the released peroxide. Subsequent reports have demonstrated cellular release of superoxide (11) and nitric oxide (12) in response to A␤ peptide treatment. The effect of these radicals may be potentiated since A␤ peptides also appear to inhibit the cellular redox mechanisms that normally protect cells from oxidative stress (13).
In a provocative hypothesis, Hensley et al. (14) suggest that the A␤ peptide itself spontaneously generates free radicals that can damage cells. By mixing the spin trap N-tert-butyl-␣-phenylnitrone (PBN) with neurotoxic forms of the A␤ peptide such as A␤-(1-40) and A␤-(25-35), they report the generation of ESR-detectable radical adducts in cell-free solutions (14 -17).
They suggest that these radicals are generated by methionine oxidation (14,15) or by the fragmentation of A␤ peptide into smaller oligopeptide radicals (14,15) that act as "shrapnel" to damage and eventually kill cells. Recently, these results were described as possibly artifactual due to contaminants in some of the preparations (18).
We have reinvestigated the spin-trapping studies of spontaneous free radical formation by A␤ peptides. In contrast to previous literature (14 -17), we now report that neurotoxic A␤-(1-40) and A␤- (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35) in the presence of the spin trap PBN do not form ESR-detectable radical adducts spontaneously. Amyloid/PBN radical adducts (14 -17) reported earlier were found to be di-tert-butylnitroxide and tert-butylhydronitroxide, which are formed by oxidation of the corresponding hydroxylamines. We have investigated the possibility that toxic A␤ may potentiate metal-catalyzed oxidation of hydroxylamine derivatives. Transition metals like iron and copper frequently contribute to reactions where molecules are oxidized, but the transition metal oxidation of hydroxylamine derivatives has not been previously studied. A body of emerging work demonstrates that copper and iron are significantly associated with amyloid plaques in the brains of patients with AD, but not with the neuropil lacking plaques nor with control neuropil from healthy brains (19). Since A␤ is retained on metal chelate columns charged with copper (20), and copper dramatically increases the rate of A␤ aggregation into amyloid fibrils (21), then complexes of A␤ and copper may contribute to the pathological oxidative stress associated with AD. We now report the effects of neurotoxic A␤-(1-40) and A␤- (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35) and non-toxic A␤-(40 -1) on the oxidation of hydroxylamine derivatives in the presence and absence of metals. The data obtained allows us to conclude that the previously reported formation of amyloid/ PBN radical adducts (14 -17) in the Sigma PBN treated with A␤ peptides can be explained as A␤ peptide-stimulated, metalcatalyzed oxidation of the corresponding hydroxylamine impurities in Sigma PBN.
Preparation of PBN Stock Solution-Stock solutions of PBN (150 mM) in HPLC grade water were used to prepare final 50 mM PBN solutions.
Preparation of TEMPONE-H Stock Solutions-TEMPONE-H was dissolved in oxygen-free HPLC grade water (30-min argon bubbled water) with 1 mM Desferal. Desferal was used to decrease the spontaneous oxidation of hydroxylamines catalyzed by traces of transition metal ions. The concentration of TEMPONE-H in the stock solutions was 1 mM. Prior to the experiments, the stock solution was kept under a flow of argon in a cool, air-tight place.
Preparation of tert-Butylhydronitroxide Solution-tert-Butylhydronitroxide was obtained by auto-oxidation of 1 mM N-tert-butylhydroxylamine in water or by oxidation of 10 M solution of N-tert-butylhydroxylamine with 10 M K 3 [Fe(CN) 6 6 ] and the iron-chelating agent Desferal. Oxidation of hydroxylamine impurities in Sigma PBN was performed using 5 M Computer Simulation-Computer simulations and spin-trap data base searches were performed using a computer simulation program, 2 the details of which have been described elsewhere (23).

Reinvestigation of Spontaneous Free Radical Formation by A␤ Peptides-
The ability of A␤ peptides to spontaneously form free radicals was studied by ESR using the spin trap PBN from the OMRF Spin Trap Source. A␤-(1-40), A␤-(40 -1), and A␤-(25-35) peptides were tested in parallel with a control that did not contain A␤ peptides (Fig. 1). ESR spectra were collected every 40 min over 6 h, and no ESR signals were observed in any mixtures ( Fig. 1, A-D). Therefore, A␤ peptides do not form ESR-detectable radical adducts spontaneously.

, E-H).
This triplet spectrum has a nitrogen hyperfine-coupling constant of 17.14 G. The formation of this triplet spectrum with Sigma PBN, but not OMRF PBN, is consistent with the spectrum arising from an impurity in Sigma PBN.
Formation of Di-tert-butylnitroxide and tert-Butylhydronitroxide from Impurities in Sigma PBN-PBN from Sigma was checked for the presence of impurities that might lead to the formation of ESR signals. After a 6-h incubation of 50 mM aqueous Sigma PBN, a strong ESR spectrum was observed ( Fig. 2A) that consisted of ESR spectra from two nitroxides (Fig. 2, B-D). One nitroxide has a triplet ESR spectrum ( Fig.  2C) with a nitrogen hyperfine coupling constant of 17.16 G, which is very close to the reported nitrogen hyperfine coupling constant for di-tert-butylnitroxide. The second nitroxide has a four-line ESR spectrum (Fig. 2D) that is consistent with a nitrogen hyperfine coupling constant of 14.61 G and a hydrogen hyperfine coupling constant of 13.93 G, which are very close to the reported hyperfine coupling constants for tert-butylhydronitroxide (24,25).
To confirm the chemical structure of nitroxide radicals observed in the solution of Sigma PBN, we obtained experimental ESR spectra of di-tert-butylnitroxide (Fig. 2E) and tert-butyl-hydronitroxide (Fig. 2F). Experimental ESR spectra of di-tertbutylnitroxide and tert-butylhydronitroxide were identical to the ESR spectrum of the PBN solution shown above (Fig. 2). The ESR spectra from the Sigma PBN solution ( Fig. 2A) can be described as a combination of two specific ESR spectra: those of di-tert-butylnitroxide (Fig. 2E) and tert-butylhydronitroxide (Fig. 2F). Thus, during the incubation of Sigma PBN in water, both di-tert-butylnitroxide and tert-butylhydronitroxide were formed.
In order to analyze the mechanism of nitroxide formation in the PBN solution, the effects of the chelating agent Desferal and of ferric iron addition were studied. Fresh solutions of Sigma PBN in Chelex-100-treated phosphate buffer contained trace amounts of di-tert-butylnitroxide as identified by its ESR spectrum (Fig. 3A). After just 2 h of incubation, the ESR signal was increased dramatically (Fig. 3B). A parallel 2-h incubation in the presence of the chelating agent Desferal significantly inhibited formation of di-tert-butylnitroxide (Fig. 3C). Addition of 5 mM K 3 [Fe(CN) 6 ] to the fresh PBN solution caused greater nitroxide formation (Fig. 3D) than that of the 2-h incubation (Fig. 3B).
Linewidth Dependence on the ESR Instrumental Settings-Previously, it was reported that A␤-(1-40) and A␤- (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35) peptides formed radical adducts with identical triplet ESR spectra and nitrogen hyperfine coupling constants of 17.1 G (16), which is actually the same as that reported for di-tertbutylnitroxide (Fig. 2E). Based on the much greater linewidth of the A␤ peptide/PBN reaction product (1.6 G), these authors concluded that this product was not di-tert-butylnitroxide (17). However, the linewidth is not a specific parameter of a nitroxide ESR spectrum (26). Linewidth is very dependent on the ESR instrumental settings, mainly modulation amplitude and microwave power. In order to show that authentic di-tert-butylnitroxide and the nitroxide derived from Sigma PBN can have the same ESR linewidth as has been described for a PBN/ ⅐ A␤ peptide radical adduct, the ESR spectra were obtained using different modulation amplitude settings (Fig. 4). The linewidth of the ESR spectra of di-tert-butylnitroxide was equal to that of the nitroxide from Sigma PBN (Fig. 4, A and D)  when ESR spectra were collected under identical instrument settings. Moreover, both spectra displayed the same linewidth dependence on the modulation amplitude setting. We observed peak-to-peak linewidths of 0.51, 0.96, and 1.56 G when using modulation amplitudes of 0.32, 1.00, and 1.59 G, respectively (Fig. 4).
Previously, it was also reported that non-toxic A␤-(40 -1) peptide formed a radical adduct with a quartet ESR spectrum and equivalent nitrogen and hydrogen hyperfine coupling constants of 14.5 G (16,17), which are very close to those reported for tert-butylhydronitroxide (Fig. 2F). Based on the much greater linewidth of the PBN/ ⅐ A␤ peptide reaction product, it was concluded that this product was not tert-butylhydronitroxide (17). In order to show that tert-butylhydronitroxide and the nitroxide from Sigma PBN solution could have the same ESR spectrum linewidth as the radical described as the PBN/ ⅐ A␤ peptide-(1-40) reaction product, the ESR spectra were obtained using different modulation amplitude settings for the ESR spectrometer (Fig. 5). The linewidth of the ESR spectrum of tert-butylhydronitroxide was the same as for the nitroxide species from the Sigma PBN solution (Fig. 5, A and D). Moreover, the dependence of linewidth on the modulation amplitude was the same. Using modulation amplitudes of 0.32, 1.00, and 1.59 G, the widths of the low-field line were observed to be 0.76, 0.98, and 1.53 G, respectively (Fig. 5).

FIG. 3. Formation of di-tert-butylnitroxide during transition metal-catalyzed auto-oxidation of di-tert-butylhydroxylamine in a solution
It was previously shown that di-tert-butylnitroxide can be formed in a solution of PBN by the metal-catalyzed auto-oxidation of di-tert-butylhydroxylamine. Therefore, in the presence of 1 mM Desferal in phosphate buffer, A␤-(1-40) and A␤-(25-35) stimulate the oxidation of the di-tert-butylhydroxylamine impurity to form di-tert-butylnitroxide.

Effect of Ferric Ions on TEMPONE-H Oxidation-
The potential role of ferric ion in A␤-stimulated hydroxylamine oxidation was studied (Fig. 9). ESR spectra of the fresh solution of TEMPONE-H revealed only a relatively small amount of TEM-PONE (Fig. 9A). Addition of 1 M Fe 3ϩ NH 4 (SO 4 ) 2 to TEM-PONE-H led to a 10-fold increase in the ESR amplitude of the TEMPONE spectra (Fig. 9B) (Fig. 9C). Therefore, ferric ion can oxidize hydroxylamine TEMPONE-H to form nitroxide TEMPONE and, presumably, ferrous ion (Reaction 1), which is air-oxidized (Reaction 2) (27). Hydrogen peroxide did not directly oxidize the hydroxylamine TEMPONE-H in solution containing Desferal (data not shown).

Effect of Cupric Ions on TEMPONE-H Oxidation-
The possible role of cupric ions in A␤-stimulated hydroxylamine oxidation was studied (Fig. 10). Addition of 1 M Cu 2ϩ SO 4 to TEM-PONE-H led to a drastic increase in the ESR amplitude of the TEMPONE spectrum (Fig. 10, A, and D), which was much more pronounced than the effect of 1 M Fe 3ϩ NH 4 (SO 4 ) 2 (Fig. 10B). The chelating agent Desferal (1 mM) completely inhibited oxidation of TEMPONE-H by 1 M Cu 2ϩ SO 4 (Fig. 10C). Therefore, cupric ion can oxidize hydroxylamine TEMPONE-H to form nitroxide TEMPONE and cuprous ion (Reaction 3), which is oxidized by oxygen (Reaction 4) (28).

REACTIONS 3 and 4
Auto-oxidation of Hydroxylamines-In order to clarify the mechanism of auto-oxidation of hydroxylamine derivatives, formation of TEMPONE was determined in the presence of superoxide dismutase, catalase, or Desferal (Fig. 11). TEM-PONE-H was incubated in phosphate buffer for 30 min. ESR spectra of the control sample of TEMPONE-H revealed a small amount of TEMPONE (Fig. 11A). The addition of superoxide dismutase (100 units/ml) to TEMPONE-H led to a 2-fold decrease in the ESR amplitude of the TEMPONE spectra (Fig.  11B). The effect of superoxide dismutase supports the role of Reaction 5 in the auto-oxidation of hydroxylamines (22).
Addition of catalase (10 g) to TEMPONE-H did not significantly change the content of TEMPONE in the sample (Fig.  11C). Addition of the chelating agent Desferal (100 M) to TEMPONE-H greatly inhibited the TEMPONE formation (Fig.  11D). Therefore, transition metals (M nϩ1 ) can oxidize hydroxylamine TEMPONE-H to form TEMPONE (Reaction 6). Reduced transition metals (M nϩ ) are oxidized by oxygen (Reaction 7). Decomposition of hydrogen peroxide by catalase did not affect the oxidation of hydroxylamine TEMPONE-H, excluding a role for a direct oxidation of TEMPONE-H by hydrogen peroxide.
Previously, it was reported that ESR signals formed during the incubation of A␤ peptides with PBN were due to PBN/ ⅐ A␤ peptide radical adducts (14 -17). However, according to the current literature, there are no reported radical adducts of PBN that have similar ESR parameters (a N ϭ 17.1 G). Using Sigma PBN, both the triplet and quartet spectra of di-tertbutyl-and tert-butylhydronitroxides, respectively, were detected. These nitroxides had previously been misinterpreted as novel radical adducts formed from A␤ radicals (14 -17).
Recently, it was suggested that PBN/ ⅐ A␤ peptide radical adducts decomposed to an alkoxyl nitroxide (15,17). This alkoxyl nitroxide would have the same structure as the alkoxyl adduct of 2-methyl-2-nitrosopropane. This assignment is also an error, because the nitrogen hyperfine coupling constant of alkoxyl adducts of 2-methyl-2-nitrosopropane is about 27 G (29,30), which is quite different from the 17.1 G for the radical observed. Moreover, according to the data base (http://epr.niehs.nih.gov/), the hyperfine coupling constant of 17.1 G is very specific to di-tert-butylnitroxide, which is directly supported by the ESR experiments with synthetic di-tert-butylnitroxide presented here (Fig. 2E).
We show in this work that these ESR signals do not appear during the incubation of A␤ peptides with high quality OMRF PBN. Moreover, these ESR signals did appear during the autooxidation of impurities only in the Sigma PBN. These data strongly suggest that the ESR spectra reported as PBN/ ⅐ A␤ peptide radical adducts actually result from the formation of di-tert-butylnitroxide and tert-butylhydronitroxide radicals. These nitroxide radicals are formed during metal-catalyzed auto-oxidation of di-tert-butylhydroxylamine and N-tert-butyl-hydroxylamine (Scheme 1), which are impurities in PBN from Sigma.
The spin trap PBN is usually synthesized by condensation of benzaldehyde with N-tert-butylhydroxylamine (31). Therefore, the trace amount of N-tert-butylhydroxylamine in preparations of PBN is probably due to incomplete purification.
The appearance of di-tert-butylhydroxylamine in PBN can be explained by the presence of di-tert-butylhydroxylamine in the N-tert-butylhydroxylamine commonly used for the synthesis of PBN. Scheme 2 illustrates the mechanism of di-tert-butylhydroxylamine formation from N-tert-butylhydroxylamine. N-tert-butylhydroxylamine is a good reducing agent, which, after two one-electron oxidations, produces 2-methyl-2-nitrosopropane. 2-Methyl-2-nitrosopropane is an unstable compound, which is readily decomposed by heat or light to the tert-butyl radical and nitric oxide (32,33). The tert-butyl radical will react with 2-methyl-2-nitrosopropane (which is actually a well known spin trap) to form di-tert-butylnitroxide (32,33). Di-tertbutylnitroxide can easily be reduced to di-tert-butylhydroxylamine even by excess N-tert-butylhydroxylamine. The presence of di-tert-butylhydroxylamine in the sample of N-tert-butylhydroxylamine was supported by the fact that the ESR spectra of N-tert-butylhydroxylamine oxidized by K 3 [Fe(CN) 6 ] contained small traces of the triplet signal of di-tert-butylnitroxide (Fig.  3F). Therefore, trace amounts of di-tert-butylhydroxylamine are present in many preparations of PBN. Due to the intrinsic sensitivity of ESR, oxidation of trace levels of these impurities can lead to readily detected ESR signals. The assignment of the structure of these nitroxides to specific compounds excludes the possibility that these ESR spectra are derived from amyloid ␤ peptides.
The data obtained allow us to conclude that, under some conditions, the toxic A␤ peptides A␤-(1-40) and A␤- (25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35) enhance transition metal-catalyzed oxidation of hydroxylamine derivatives. Therefore, the previously reported formation of radical adducts with A␤ peptides can be explained as an A␤enhanced, transition metal-catalyzed oxidation of hydroxylamine derivatives found as impurities in commercial PBN. In addition, our data (Fig. 7) demonstrate that A␤ peptides can compete with Desferal for transition metals such as copper, which implies that A␤ peptides may bind redox-active transition metals in vivo where trace metal concentrations are extremely low. Any biological significance of the catalytic role of the toxic A␤ peptides in the transition metal-mediated oxidation of hydroxylamine is yet to be determined. Although these hydroxylamines do not occur naturally, other easily oxidized substances exist in vivo such as ascorbate and GSH.
Moreover, it is known that copper and iron ions catalyze protein damage and may be partly responsible for the alterations of protein damage in vivo (35). Metal-catalyzed protein damage is associated with oxidative modification of amino acids, for example, formation of protein carbonyl groups (35). Therefore, the previously reported inactivation of glutamine synthetase and creatine kinase incubated with A␤-(25-35) (14) could be explained on the basis of metal-catalyzed protein damage.
Previously, it was shown that iron facilitates A␤ toxicity to cultured cells (36). Moreover, it was found that iron metabolism is altered in Alzheimer's disease (37), which, in combination with accumulation of A␤ peptides, could result in peroxidative stress in the brains of Alzheimer's disease patients. It was suggested that lowering the level of available iron could provide a therapeutic approach to Alzheimer's disease (38). Furthermore, it was shown that A␤-(25-35) enhances iron-and copper-catalyzed generation of reactive oxygen species (39). Therefore, our data concerning toxic A␤-stimulated metal-catalyzed oxidation of hydroxylamine derivatives could give insight into the synergetic toxicity of transition metals and A␤ peptides.
The data obtained lead to the conclusion that A␤ peptides do not form ESR-detectable radical adducts spontaneously. Our spin-trapping results and interpretation differ from the previous investigations (14 -17). Our data do not support the conclusion that A␤ peptides spontaneously form free radicals, but demonstrate that hydroxylamine impurities in the spin trap preparations are responsible for the observed ESR spectra. Therefore, the reported spontaneous generation of PBN radical adducts by A␤ peptides has been reinterpreted. Although the spontaneous free radical model of A␤ neurotoxicity has been criticized by Sayre et al. (40), the previous spin-trapping results and interpretations have not been challenged until now.