Free Heme and Amyloid-β: A Fatal Liaison in Alzheimer’s Disease

General

Most experiments were performed in 10 mM phosphate buffered saline (PBS, Sigma-Aldrich, Steinheim, Deutschland), pH 7.4. For measurements at pH 6.5 a citrate-phosphate buffer (CPB) prepared from 10 mM citrate (Merck, Darmstadt, Germany), 20 mM disodium hydrogen phosphate, and 140 mM NaCl (both Sigma-Aldrich, Steinheim, Germany) was used. Hemin (chloride salt of ferric protoporphyrin IX, Sigma Aldrich, Steinheim, Germany) as a model for free heme was freshly prepared on a daily basis as a 10 mM stock solution in 0.1 M NaOH [24, 35].

Aβ with the human-specific sequence was obtained from Rockland, Limerick, PA, USA (Aβ_1 - 16) or AlexoTech, Umeå, Sweden (Aβ_1 - 40). Selected mutants and the rodent variant of Aβ_1 - 16 were synthesized in the Core unit DNA technologies of the IZKF Leipzig (Interdisziplinäres Zentrum für Klinische Forschung, Interdisciplinary Center for Clinical Research), Medical Faculty, University Leipzig, Leipzig, Germany. Stock solutions (500 μM) of Aβ were prepared in PBS (Aβ_1 - 16) or 10 mM borax buffer, pH 9.0 (Aβ_1 - 40) and stored at –20°C until usage. The latter buffer was prepared by supplementing sodium borate (Sigma-Aldrich, Steinheim, Germany) with 140 mM NaCl and adjusting the pH value with 0.1 mM NaOH (Honeywell, Seelze, Germany).

Peroxidase activity

The peroxidase activity of Aβ-heme complexes was quantified by performing 2,2^′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid)- (ABTS-) based absorbance measurements [19, 36] in a plate reader (TECAN Infinite M200, Männedorf, Switzerland): Briefly, the non-aggregating Aβ_1 - 16 [25] (1–160 μM) was pre-incubated for 5 min [37] with hemin (1–10 μM) and ABTS (1 mM, Sigma-Aldrich, Steinheim, Germany) in PBS (pH 7.4) or CPB (pH 6.5) at 37°C by using 96-well plates. After H₂O₂ (Sigma-Aldrich, Steinheim, Germany) addition (usually 500 μM) via an injector device the formation of ABTS radical cations (ABTS^•+) was followed spectroscopically at 734 nm (ɛ₇₃₄ = 1.5×10⁴ M^-1 cm^-1) [36, 38] for up to 90 min. All given concentrations are final ones. From the observed initial slope of the absorbance increase the peroxidase activity was determined and expressed as mU/ml [39].

The H₂O₂ stock solution used for the initiation of the peroxidase activity was freshly prepared from a 30 % stock solution by diluting in bi-distilled H₂O, determined spectroscopically at 230 and 240 nm (ɛ₂₄₀ = 74 M^-1 cm^-1, ɛ₂₃₀ = 43.6 M^-1 cm^-1) and stored at 4°C till injection [40, 41]. In some experiments, instead of H₂O₂ injection a glucose/glucose oxidase-system (G/GO-system) was applied for a constant hydrogen peroxide generation. Thereby 10 mM glucose and 2–200 mU/ml GO were used to realize different H₂O₂ production rates.

Dityrosine formation

The peroxidase activity of Aβ-heme complexes was also measured by following the formation of dityrosine bridges from the human-specific Y10 residue [25]. Briefly, 100 μM of the Aβ_1 - 16 peptide were pre-incubated with hemin (0.1–100 μM) at 37°C. After H₂O₂ injection (usually 500 μM) the formation of dityrosine was followed for up to 90 min via fluorescence measurements (excitation: 320 nm, emission: 410 nm) [19, 42] by using the TECAN plate reader. Again in some experiments, the G/GO-system was used instead of H₂O₂ injection. Moreover, in selected samples 1 μM horseradish peroxidase (HRP, Sigma-Aldrich, Steinheim, Germany) was added instead of hemin [21].

The formation of dityrosine due to the peroxidase activity of Aβ-heme complexes was also verified by using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry. Briefly, 100 μM Aβ_1 - 16 were incubated with up to 2 μM hemin or 1 μM HRP. Then 100 μM H₂O₂ were added and the samples were incubated for 20 h at 37°C. Afterwards 4 μM hemopexin (Hpx) was applied to bind hemin. The samples were subsequently centrifuged through a filter with 30 kDa cut-off to remove both Hpx-hemin complexes and HRP. Then 0.5 M 2,5-dihydroxybenzoic acid (DHB, Sigma-Aldrich, Steinheim, Germany) in methanol supplied with 0.1% trifluoracetic acid (TFA, Sigma-Aldrich, Steinheim, Germany) was added as a matrix and the samples were analyzed in in the positive mode of a Bruker Autoflex (Bruker Daltonics GmbH, Leipzig, Germany) mass spectrometer equipped with a 337 nm nitrogen laser [43].

Amyloid-β fibrillation

The aggregation of Aβ_1 - 40 was followed by using Thioflavin T (ThT, Sigma-Aldrich, Steinheim, Germany) [44]. Briefly, the integration of ThT into the peptide fibrils was followed for about 72 h at 37°C by performing fluorescence measurements (excitation: 440 nm, emission: 485 nm) [32, 44] in the TECAN plate reader. While 50 μM Aβ and 20 μM ThT were used throughout [45], in selected experiments heme (up to 10 μM), HRP (1 μM) and/or H₂O₂ (500 μM) were added. The peptide was pre-incubated with heme, HRP, and/or H₂O₂ for 4 h at 37°C. Afterwards ThT was added and the formation of Aβ fibrils was followed at 37°C for up to 96 h. Thereby the fluorescence intensities of the samples were measured every 30 min after shaking the 96-well plate for 10 s.

For each condition the obtained values were corrected for the fluorescence intensities observed in the absence of Aβ_1 - 40 [45]. By fitting the curves to a sigmoidal function (equation 1) the fibrillation half-time (t_1/2, corresponds to h) and the apparent fibrillation rate constant (k_obs, corresponds to 1/τ) were determined [32, 47]. If no sigmoidal fitting was possible a hyperbolic fitting function (equation 2) was applied to determine t_1/2 and k_obs. y = ( F 1 + m 1 x ) + ( F 2 + m 2 x ) 1 + e - x - h τ(1) y = F 1 - F 2 1 + e x - h τ + F 2 + m 2 x(2)

The formation of Aβ fibrils was verified by using scanning transmission electron microscopy (STEM) [31, 32]. Briefly, 10 d after the ThT measurements 1 μl of selected samples were applied on formvar coated copper grids, dried for 2 h and negatively stained with 1% uranyl acetate [14]. STEM was performed on a Zeiss SIGMA electron microscope (Zeiss, Jena, Germany) equipped with a STEM detector and ATLAS software. From the STEM pictures the morphology was analyzed [11].

In vivo measurements

For the in vivo investigation, Tg2576 mice, which contain human AβPP with the Swedish double mutation (K670N, M671L) under control of a hamster prion protein promoter [48], were used. Age-matched (18 months) non-transgenic littermates served as control mice. All animal experiments were conducted at Leiden University after being approved by the Institutional Animal Care and Use Committee in accordance with the NIH Guide for the care and use of laboratory animals.

Magnetic resonance angiography (MRA) measurements were conducted on a vertical wide-bore 17.6 T Bruker spectrometer (Bruker BioSpin GmbH, Rheinstetten, Germany) with a 1000 mTm^-1 actively shielded imaging gradient insert. A birdcage radio-frequency coil with an inner diameter of 2 cm was used. The system was interfaced to a PC running Topspin 2.0 and Paravision 5.0 software. Time-of-flight angiograms were obtained by using a three-dimensional gradient-echo sequence and generation of maximum-intensity projections [49].

Histological analysis for the identification of Aβ deposits, CAA and iron was performed as described previously [49, 50]: First, mice were anesthetized and transcardially perfused with PBS, pH 7.4, followed by 4% buffered paraformaldehyde (ThermoShandon, Runcorn, UK) through the left cardiac ventricle. Afterwards the brain was dissected out, placed in the same fixative for 48 h, dehydrated, and embedded in paraffin. Subsequently 5 μm coronal sections were cut using a vibratome.

To detect Aβ, brain sections were subjected to immunohistochemistry using anti-Aβ (6E10) or anti-Aβ_1 - 40 (BC40) antibodies. Immuno-labeling was visualized by using the Vectastain ABC kit (Vector laboratories, Bulingame, UK) according to the manufacturer’s instructions. Aβ-associated redox-active iron was detected histochemically as described previously [51, 52]. Briefly, the brain sections were incubated in 7% potassium ferrocyanide in 3% HCl solution for 15 h and subsequently incubated in 3.5 mM 3,3’-diaminobenzidine (DAB) and 4.4 mM H₂O₂ for 10 min. Images of the histological sections were obtained using a Leica DM RE HC microscope, interfaced to a Leica DC500 3CCD digital camera (both Leica Microsystems GmbH, Wetzlar, Germany).

Statistics

All ABTS-based kinetic measurements were performed in triplicate. The peroxidase activity was calculated from the initial slope of the averaged kinetic curve whereby the given standard deviations correspond to the coefficient of determination for the applied linear curve fit. Also the fluorescence measurements for the determination of dityrosine formation were performed in triplicate and an average curve was created. From the stable fluorescence plateau, the amount of dityrosine was determined whereby the given standard deviation corresponds to the oscillations of the fluorescence values within this time frame. The MALDI-TOF measurements for the verification of dityrosine were performed in triplicate and mean and standard deviation of the relative peak intensities were calculated.

For the aggregation measurements with ThT, again at least three independent experiments were performed. The obtained kinetic curves were averaged, corrected, and fitted to a sigmoidal or hyperbolic curve to determine t_1/2 and k_obs. The standard deviation given for these values corresponds to the coefficient of determination obtained during the curve fit. The average fibril diameter was determined by analyzing the microscopic pictures. Thereby in each sample ten randomly chosen fibrils were measured to calculate mean and standard deviation of the fibril diameter.

For all results, significant differences were determined by using a Student’s two tailed t-test after verifying the Gaussian distribution of the data (Shapiro-Wilk test). Thereby weakly significant (*p≤0.05), significant (**p≤0.01), or strongly significant (***p≤0.001) differences were determined.

The presented results from the MRA measurements and the microscopic analysis of the brain sections are representative examples obtained during the analysis of multiple measurements. Five Tg2576 and five control mice were used for the in vivo studies.

Heme peroxidase activity promotion by Aβ

Using ABTS-based kinetic measurements with Aβ_1 - 16 we observed a clear enhancement of the heme-derived peroxidase activity in the presence of the peptide. Furthermore, the peroxidase activity of Aβ-heme complexes showed a characteristic pH- and H₂O₂-dependence. As illustrated by the representative kinetic curves shown in Fig. 1A, while free heme (10 μM, grey dashed line) led only to a minor and slow formation of ABTS^•-, in the presence of increasing Aβ concentrations (10–180 μM, grey - black lines) constantly elevating ABTS oxidation rates were observed. In fact, as shown in Fig. 1B in the presence of 2 μM peptide already a strongly significant elevated peroxidase activity (1.18±0.01 mU/ml) was observed as compared to the heme control (1.03±0.01 mU/ml). With increasing Aβ-heme ratio an almost linear increase in the peroxidase activity was observed with the highest value (11.89±0.18 mU/ml) at 18-fold peptide excess over heme. Yet as illustrated by the data obtained from measurements with 1 μM heme (peroxidase activity: 0.13±0.00 mU/ml) and up to 100 μM peptide (Fig. 1C) at even higher Aβ-heme ratios a saturating effect can be observed. In the absence of heme, the peptide exhibited no peroxidase activity (not shown).

OPEN IN VIEWER

Fig.1

Aβ-derived acceleration of heme-dependent peroxidase activity. 10 μM heme and up to 180 μM Aβ_1 - 16 (A and B) or 1 μM heme and up to 100 μM Aβ_1 - 16 (C) were pre-incubated in PBS in the presence of ABTS. After H₂O₂ addition (500 μM), the formation of ABTS^•- was followed spectroscopically. While in (A) kinetic samples for 10 μM heme alone (grey, dashed) or in the presence of 10 μM, 60 μM, 120 μM, or 180 μM peptide (grey to black lines) are shown, in (B) the peroxidase activity calculated from the initial slopes is plotted against the Aβ-heme ratio. The corresponding data for 1 μM heme and up to 100 μM Aβ are displayed in (C). In (D) the relative Aβ-derived promotion of the heme-dependent peroxidase activity is shown for both conditions.

Most interestingly, by plotting the relative increase in the peroxidase activity (referred to the value obtained in the sole presence of heme) against the Aβ-heme ratio (Fig. 1D) for the experiments with 10 μM heme (black), an approximate 12-fold increase in the peroxidase activity was already observed at an 18-fold peptide excess. In contrast, by using 1 μM heme (grey), a comparable effect was only observed in the presence of a 100-fold peptide excess. At a 10-fold excess of Aβ_1 - 16, a 7.8-fold higher peroxidase activity was found for 10 μM heme while with 1 μM heme, a 10-fold peptide excess led only to an about 2.9-fold higher peroxidase activity.

As shown in Supplementary Figure 1, we also performed measurements at pH 6.5. As illustrated by the displayed selected kinetic curves (Supplementary Figure 1A) and proofed by determining the corresponding activities the peroxidase activity of 10 μM heme (1.96±0.09 mU/ml) was already about 90% higher as compared to the value observed at pH 7.4. Also upon application of increasing peptide concentrations (Supplementary Figure 1B), higher peroxidase activities were observed at pH 6.5 (black) than at pH 7.4 (white). Yet the peptide-dependent promotion of the heme-derived peroxidase activity was not completely equal for both pH values: While at pH 7.4, an 18-fold peptide excess led to an almost 12-fold increase in the peroxidase activity (see also Fig. 1D); at pH 6.5, a Aβ-heme ratio of 18:1 led only to an about 7.8-fold higher ABTS^•- formation rate.

Instead of a H₂O₂ bolus addition, we used the glucose/glucose oxidase-system (G/GO-system) in selected experiments in order to create a constant hydrogen peroxide production during the peroxidase measurements. As shown in Fig. 2A, for all applied GO concentrations (2–200 mU/ml) again a clear Aβ-heme ratio-dependent increase in the peroxidase formation rate was observed. Thereby a minor H₂O₂ dependence can be observed: At 200 mU/ml GO (black) an Aβ-heme ratio of 100/1 instead of 1/1 led to an about 7.0-fold higher peroxidase activity. However, by using 20 mU/ml (grey) or 2 mU/ml (light grey) GO, a comparison of the named Aβ-heme ratios showed only an about 3.7-fold or 3.2-fold increase in the ABTS^•- production rate. As calculated from a calibration curve (Supplementary Figure 2), the used amount of GO yields to a constant production of 2.62 μM, 25.69 μM, or 159.05 μM H₂O₂/min, respectively, which meets the expected H₂O₂ production rates.

OPEN IN VIEWER

Fig.2

Effect of Aβ on the H₂O₂-dependent heme peroxidase activity. In (A) 1 μM heme and up to 100 μM Aβ_1 - 16 were pre-incubated in PBS in the presence of ABTS and 2 mU/ml (light grey), 20 mU/ml (grey) or 200 mU/ml GO (black). After addition of 10 mM glucose the formation of ABTS^•- was followed and the initial peroxidase activity determined. In (B) for 10 μM heme (squares) or 1 μM heme/10 μM Aβ_1 - 16 (circles) the ABTS oxidation was followed after addition of 20–5000 μM H₂O₂. The calculated initial peroxidase activities were related to the particular values observed at 20 μM H₂O₂.

Yet, as illustrated in Fig. 2B, by incubating either 10 μM heme (squares) or 10 μM Aβ and 1 μM heme (circles) with increasing H₂O₂ concentrations the increase in the relative peroxidase activity (referenced to the values obtained at 20 μM H₂O₂) was almost equal: For instance, the application of 500 μM H₂O₂ led to an 11.3-fold higher activity in the case of the Aβ-heme complex while for heme alone a 9.1-fold increase was observed. These results proof again the heme-based enzymatic activity of Aβ-heme complexes. Still, again a minor effect of the peptide on the H₂O₂ concentration-dependent promotion of the heme-derived peroxidase activity may be guessed.

Aβ sequence-dependent heme peroxidase activity promotion

Next we tested the relevance of the human-specific amino acid sequence of Aβ for the promotion of the heme-dependent peroxidase activity by incubating 10 μM heme with a 10-fold excess of different Aβ peptides. Thereby we focused on sequence differences between the human and rodent peptide, but also included selected mutants described in the literature. The results clearly show that the enzymatic activity of Aβ-heme complexes strongly depends on the human-specific sequence of the peptide.

As illustrated by the representative kinetic curves in Fig. 3A, the exchange of one of the three human-specific amino acids of Aβ_1 - 16 (R5, Y10, H13) to the rodent equivalent (G5, F10, R13) already led to considerably lower curves in comparison to the human peptide (black line). Yet the ABTS^•- formation of these single mutants (dark grey lines) seems to be still stronger in comparison to the mouse peptide (light grey line) and certainly stronger than the peroxidase activity of heme alone (light grey dashed line). As shown by the exemplary curves in Fig. 3B, the corresponding double mutants (R5G, H13R; R5G, Y10F; Y10F, H13R) (grey lines) apparently showed an even lower peroxidase activity than the single mutants that was, however, still higher than the ABTS oxidation rate of the mouse peptide (light grey line). As both the human and the rodent Aβ peptide contain a histidine at position 14, an artificial double mutant was also included where both the human-specific H13 and H14 were replaced by glycine. This double mutant seemed to show an even lower peroxidase activity than the mouse peptide, which was, however, still stronger than the value observed for heme alone (light grey dashed line).

OPEN IN VIEWER

Fig.3

Aβ sequence-dependent acceleration of the heme-derived peroxidase activity. 10 μM heme alone or in the presence of 100 μM Aβ were pre-incubated in PBS in the presence of ABTS. After H₂O₂ addition (500 μM) the formation of ABTS^•- was followed spectroscopically. Both in (A) and (B) kinetic sample spectra for heme alone (light grey, dashed) or in the presence of the human (black line) or the rodent (grey line) peptide are shown. Also kinetic examples for peptides with one (A) or two (B) mutations of the human-specific amino acids R5, Y10, and H13 are included. In (C) for all tested peptides the peroxidase activity calculated from the initial ABTS^•--derived absorbance increase are shown. Thereby also values obtained for single mutants of human Aβ_1 - 16 described in the literature are included.

In Fig. 3C, the peroxidase activity for all samples determined from the initial slope is shown. While for heme alone, a peroxidase activity of 1.03±0.01 mU/ml was found, the peroxidase activity in the presence of a 10-fold excess of human Aβ was about 8-fold higher (7.99±0.27 mU/ml). In contrast by using rodent Aβ, only an about 2.5-fold higher value (2.49±0.01 mU/ml) was observed. Upon mutation of one of the human-specific amino acids in Aβ_1 - 16 with the rodent equivalent significant (R5G) or highly significant (Y10F, H13R) lower peroxidase activities were observed as compared to the human peptide. By using single mutants described in the literature (R5N, Y10G, H13A), again a highly significant reduction of the Aβ-mediated promotion of the heme-dependent peroxidase activity was found. In fact, the effect of these single mutants was always stronger as compared to the corresponding rodent mutations (see, for example, Y10F and Y10G). However, all tested single mutants still led to a strongly significant higher peroxidase activity than the rodent Aβ peptide.

The peptides where two of the three human-specific amino acids were replaced by the rodent equivalents (R5G, Y10F; R5G, H13R; Y10F, H13R) had even lower promoting effects on the heme-derived peroxidase activity. Still the obtained values were strongly significant higher than by using the rodent peptide. Since both in the human and the rodent Aβ sequence a histidine residue is present at position 14 of the peptide, we guessed that a mutation of the human-specific histidine 13 might be counterbalanced by H14. In fact, by applying a corresponding double mutant (H13G, H14G), the obtained peroxidase activity (2.06±0.03 mU/ml) was strongly significantly lower than for the rodent peptide. All tested peptides exhibited no peroxidase activity in the absence of heme (not shown).

Y10 as an endogenous peroxidase substrate of Aβ-heme complexes

By measuring the formation of dityrosine bridges, the human-specific Y10 in the human peptide was found as an important endogenous substrate for the peroxidase activity of Aβ-heme complexes. Thereby the efficiency of dityrosine formation depended on the Aβ-heme ratio, the H₂O₂ concentration, and the Aβ sequence. The peroxidase-derived formation of human Aβ dimers due to Y10-dependent dityrosine formation was also verified by MALDI-TOF mass spectrometry.

In Fig. 4A representative kinetic curves illustrate the fluorescence increase at 410 nm (excitation: 320 nm) after the addition of 500 μM H₂O₂ to 100 μM Aβ_1 - 16 in the presence of different heme concentrations. While at an Aβ-heme ratio of 10:1 (10 μM heme, black), almost no fluorescence increase was observed; at 2.25 μM heme (grey, Aβ-heme ratio: 44.4:1), a fast and strong increase in the fluorescence at 410 nm took place, indicating a quick formation of dityrosine bridges. However, at even higher Aβ-heme ratios (e.g., 200:1, light grey), again lower fluorescence values were observed, showing a lower yield of dityrosine. In the sole presence of the peptide (dashed line), no dityrosine formation took place.

OPEN IN VIEWER

Fig.4

Formation of dityrosine bridges due to heme-derived peroxidase activity. The formation of dityrosine was quantified by following the increase in the fluorescence at 410 nm (excitation: 320 nm) after adding up to 500 μM H₂O₂ to 100 μM Aβ_1 - 16 in the presence of different heme concentrations. As illustrated for 500 μM H₂O₂ in (A) and quantified for 500 μM (black) and 100 μM (white) H₂O₂ in (B) at an Aβ-heme ratio of about 30-50:1 an optimal dityrosine formation was observed. Comparable results were obtained by using the G/GO-system (C). Upon application of single and double mutants of human Aβ_1 - 16 (D) a dityrosine formation was never observed in the absence of Y10. In the presence of the named amino acid a fairly comparable dityrosine formation rate was found for all tested peptides.

By using the data obtained from control measurements with HRP (not shown), the displayed maximal fluorescence values at 410 nm were translated to the amount of dityrosine formed after H₂O₂ addition. The corresponding data for the measurements in the presence of 500 μM H₂O₂ (black) and 100 μM H₂O₂ (white) are shown in Fig. 4B. For 500 μM H₂O₂, starting at an Aβ-heme ratio of 12.5:1 (corresponds to 8 μM heme) up to an Aβ-heme ratio of 133.3:1 (0.75 μM heme) significant higher fluorescence values as compared to the negative control (only peptide, dashed line) were observed. Thereby the highest dityrosine formation was found at an Aβ-heme ratio of 44.4:1. The determined 21.08±2.02 μM dityrosine correspond to a relative yield of 42.2%, given the application of 100 μM peptide. In the presence of 100 μM H₂O₂, a comparable ratio dependence was observed. Under these conditions, the highest dityrosine formation was found at an Aβ-heme ratio of 33.3:1 (13.37±1.80 μM, 26.7% relative yield).

As shown in Fig. 4C, by using 10 mM glucose and glucose oxidase (black: 200 mU/ml, grey: 20 mU/ml, white: 2 mU/ml) as a H₂O₂-generating system an optimal dityrosine formation was always observed in the presence of 2.25 μM heme, which corresponds to an Aβ-heme ratio of 44.4:1. Moreover, the results from the measurements with the G/GO-system confirm the observation that in the presence of more H₂O₂ higher dityrosine yields are obtained although the values observed at 200 mU/ml GO (10.30±1.49 μM, 20.6 % relative yield) and 2 mU/ml GO (7.42±1.21 μM, 14.8% relative yield) showed no significant difference.

We also tested the dityrosine formation from rodent Aβ as well as the single and double mutants of human Aβ mentioned before (see above) by incubating the peptides (100 μM) with 3 μM heme in the presence of 500 μM H₂O₂. As illustrated by the quantitative analysis of the obtained data (Fig. 4D), all Y10-containing peptides showed a comparable dityrosine formation. For the R5G mutant, weakly significant higher final fluorescence values were seen, while for H13R and H13A, significant higher final fluorescence values were obtained, suggesting bigger dityrosine yields for theses peptides. In fact, while the obtained fluorescence value for human Aβ_1 - 16 at the given Aβ-heme ratio (33.3:1) was 18.62±1.39 μM dityrosine, for H13R, an about 2-fold higher yield (37.97±6.99 μM dityrosine) was calculated. Most interestingly for the artificial histidine double mutant (H13G, H14G), no significant reduction in the dityrosine yield was observed despite the strongly significant lower peroxidase activity observed in the corresponding ABTS measurements (see Fig. 3C). In striking contrast, in the absence of Y10 (rodent Aβ, Y10-single and -double mutants) no dityrosine formation was observed at all, proving the specificity of the fluorescence measurements.

The formation of dityrosine bridges in Aβ_1 - 16 due to the peroxidase activity of Aβ-heme complexes was reconfirmed using MALDI-TOF mass spectrometry (Supplementary Figure 3). As illustrated by the representative spectra shown in Supplementary Figure 5A-C, in the absence of H₂O₂ (dashed lines) only peaks at m/z 1956.0 (Aβ + H⁺) followed by a typical isotopic pattern were found, corresponding to monomeric Aβ_1 - 16. In the absence of any further additions (Supplementary Figure 3A), the incubation of the peptide with 100 μM H₂O₂ (straight line) did not lead to any new peaks. The same holds for the addition of H₂O₂ to 100 μM peptide in the presence of 1 μM heme (Supplementary Figure 3B). In contrast, upon incubation of Aβ with H₂O₂ in the presence of 2 μM heme (Supplementary Figure 3C), a new peak at m/z 3911.2 (Aβ-Aβ + H⁺) was found, corresponding to dimeric Aβ_1 - 16 with a dityrosine bridge. Again this new peak was followed by a typical isotopic pattern. A much more prominent dityrosine formation was observed after incubation of Aβ_1 - 16 with H₂O₂ and 1 μM HRP (Fig. 5D). An average relative dityrosine formation of 39.6±9.0% was observed in the HRP-containing sample while in the presence of 2 μM heme only about 3.1±0.7% was found, indicating a weak but still notable peroxidase activity in the latter case. Although not visible in the given sample (Supplementary Figure 3D), the statistical analysis showed that in the presence of HRP even in the absence of H₂O₂ (dashed line) a small dityrosine formation (2.6±1.4 %) takes place.

OPEN IN VIEWER

Fig.5

Effect of heme-derived peroxidase activity on the Aβ fibrillation kinetics. Aβ_1 - 40 (50 μM) was pre-incubated with free heme (0.1–10 μM, dark – light grey) or HRP (1 μM, light grey, dashed) in the absence (A) or presence (B) of H₂O₂ (500 μM) before adding ThT (20 μM). Without H₂O₂ (A) for the peptide alone (black, dashed), a t_1/2 value of 49.58±0.99 h and a k_obs value of 0.99±0.01 h^-1 was observed while increasing heme concentrations constantly delayed and decreased the ThT-dependent fluorescence. In the presence of 1 μM HRP, almost no fluorescence intensity increase was observed. Also H₂O₂ alone (B) inhibited the fluorescence increase while the additional presence of free heme increased the effect. In the presence of H₂O₂ and 10 μM heme or 1 μM HRP, a slight (heme) or strong (HRP) hyperbolic increase in the ThT-derived fluorescence was observed instead of a sigmoidal kinetics. In (C) and (D), the t_1/2 and k_obs values are given for the experiments without (white) or with (black) H₂O₂ whereby sometimes a logistic curve fit (dashed) was applied.

Effect of peroxidase activity of Aβ-heme complexes on Aβ fibrillation

The ThT-based fluorescence measurements suggest inhibitory effects of free heme and H₂O₂ on the fibrillation of Aβ_1 - 40. Moreover, upon co-incubation of higher amounts of HRP or higher heme concentrations with H₂O₂, a totally different kinetics of the fluorescence increase was observed, possibly implicating a peroxidase-mediated change in the Aβ fibrillation behavior. However, as verified by STEM in all samples, a formation of Aβ fibrils took place while differences in the morphology were clearly observable.

As shown in Fig. 5A in the absence of further additions the incubation of Aβ_1 - 40 (50 μM, black dashed line) with ThT led to a typical sigmoidal increase in the fluorescence with a fibrillation half-time (t_1/2) of 51.60±0.01 h and an apparent rate constant (k_obs) of 0.99±0.01 h^-1 (see also Supplementary Table 1). Yet the curves observed after incubation with free heme (0.1–10 μM, dark grey – light grey lines) show a continuous shift in the lag phase as well as lower final fluorescence values, indicating a delay and partial inhibition of the Aβ fibrillation. In fact, in the presence of 10 μM heme, t_1/2 and k_obs values of 83.65±2.71 h and 0.14±0.01 h^-1 were observed, suggesting a delayed and impaired formation of Aβ fibrils. Most interestingly in the presence of HRP (1 μM, light grey dashed line), almost no increase in the ThT-derived fluorescence was observed within 96 h, which may indicate an efficient inhibition of the Aβ fibrillation in the presence of the peroxidase. A fibrillation half-time of 71.07±0.66 h and a k_obs value of 0.37±0.05 h^-1 were determined under these conditions.

As shown by the kinetic curves in Fig. 5B, after pre-incubation of Aβ_1 - 40 with 500 μM H₂O₂ (50 μM, black dashed line) the ThT-based fluorescence measurements showed a delayed (t_1/2: 73.96±0.59 h) and slower (k_obs: 0.13±0.00 h^-1) increase in the fluorescence, indicating again a disturbed fibrillation of the peptide. Upon co-incubation of Aβ with 500 μM H₂O₂ and free heme (0.1–10 μM, dark grey – light grey lines), a further delay and decrease in the fluorescence intensity increase was observed. Of note, at 10 μM heme a fitting of the kinetic data to a sigmoidal curve was no longer possible. Instead a hyperbolic equation was applied, which better meets the small initial fluorescence increase observed under the described conditions. As shown in Supplementary Table 1, in this sample the half-maximal fluorescence intensity was achieved after about 1.15±5.33 h. Most interestingly also in the presence of 500 μM H₂O₂ and HRP (1 μM, light grey dashed line) an instant fluorescence increase was observed. The curve fitting revealed a t_1/2 value of 2.97±1.42 h. Both half-times lay within the 4 h ThT-free pre-incubation time of Aβ_1 - 40 with heme, HRP, and/or H₂O₂.

In Fig. 5C, the t_1/2 values of all samples are compared whereby the white bars represent the H₂O₂-free samples while in black the results obtained in the presence of H₂O₂ are given. This representation clearly shows the retarding effect of free heme, HRP, and H₂O₂ on the Aβ_1 - 40 fibrillation as illustrated by higher fibrillation half-times of the ThT-sigmoidal fluorescence increase. In striking contrast, upon incubation of the peptide with 500 μM H₂O₂ and either 10 μM heme or 1 μM HRP a hyperbolic fluorescence increase and considerably lower t_1/2 values were observed (dashed), suggesting different fibrillation kinetics of Aβ under these conditions. Maybe the peroxidase activity of HRP/Aβ-heme complexes is responsible for these differences. In Fig. 5D, the corresponding k_obs values are shown, clearly illustrating the decelerating effect of free heme on the ThT-derived fluorescence increase even in the absence of H₂O₂. In the presence of H₂O₂, even slower values were observed, which, however, increased slightly at higher heme concentrations. In the presence of 500 μM, H₂O₂ and either 10 μM heme or 1 μM HRP a comparable k_obs value was obtained from the applied logistic curve fitting (dashed).

We also created STEM pictures of the described samples to confirm the Aβ fibrillation as well as to study the long-term effect of free heme, HRP, and H₂O₂ on the formation of peptide fibrils. As illustrated by the representative microscopic pictures given in Fig. 6, both in the absence (Fig. 6A-E) and in the presence (Fig. 6F-J) of H₂O₂, an ultimate formation of Aβ fibrils was observed. Thereby an average fibril diameter of 14.4±2.3 nm was observed throughout the samples. However, differences in the fibril morphology and density were sometimes clearly visible. While the presence of 0.1 μM heme (Fig. 6B, G) and/or 500 μM H₂O₂ (Fig. 6F, G) seem to have no effect on the fibrillation of 50 μM Aβ_1 - 40 (A) in the presence of 1 μM (Fig. 6C, H) or 10 μM (Fig. 6D, I) heme, the formation of a more dense Aβ fibril network seems to take place. Thereby, more amorphous fibrils were observable in the absence of H₂O₂ (Fig. 6C, D), while in the presence of H₂O₂ (Fig. 6H-I), straight and dense fibrils were visible. In the presence of HRP (Fig. 6E, J), again amorphous fibrils were observed in the absence of H₂O₂ (Fig. 6E), while in the presence of H₂O₂ (Fig. 6J), straight and dense fibrils were visible. Yet in the latter case, much less fibril formation was observed as compared to the heme-containing samples.

OPEN IN VIEWER

Fig.6

Aβ fibril formation in the presence of peroxidase components. Aβ_1 - 40 (50 μM) was pre-incubated with 0 μM (A, F) –10 μM heme (D, I) or 1 μM HRP (E, J) either in the absence (A-E) or presence (F-J) of 500 μM H₂O₂. After 4 h, ThT (20 μM) was added for the kinetic measurements given in Fig. 5. The shown representative STEM pictures were taken from samples prepared about 10 days after the ThT measurements. In all samples, Aβ fibrils with an average diameter of 14.4±2.3 nm are clearly visible (see inserts) while differences in the overall fibril structure morphology and density (see overview) were observed. Especially in the presence of higher heme concentrations, a denser but more amorphous fibril network was observed in the absence of H₂O₂ (C-D) while in the presence of H₂O₂, the fibril network resembled the negative control (A, neither heme nor H₂O₂). In the sample containing HRP and H₂O₂ (J), a less and shorter fibrils were observed. The scaling bars correspond to 500 nm (overview) and 250 nm (inserts).

Determination of Aβ-derived cerebral angiopathies in vivo

The MRA measurements clearly showed a decreased blood flow in the middle cerebral artery (MCA) in the Tg2576 mice, correlating well with the increased perivascular deposition of Aβ in this AD model. Moreover, a co-localization of the peptide with free iron was observed in the brain tissue of the animals.

As Aβ is responsible both for microhemorrhages [53] and for erythrocyte alterations [54] in AD, we finally performed in vivo measurements to illustrate a possible pathway for the formation of Aβ-heme complexes at AD. As illustrated in Fig. 7A-B, the MRA studies showed a marked disturbance of the blood flow in the branches of the MCA of Tg2576 mice (Fig. 7B), while in the healthy control (Fig. 7A) normal signal intensities were observed throughout the vessels. In both cases, a sagittal view of the head of age-matched (18 months) animals is shown. The identification of cerebral arteries was performed by inspecting the angiograms under various angles and comparing the vessels to literature data [49, 55–57].

OPEN IN VIEWER

Fig.7

Blood flow disturbances and co-localization of Aβ and iron in Tg2576 mice. MRA measurements were performed for the visualization of the blood vessels while CAA, Aβ deposits, and iron accumulation were detected via staining of the brain sections. While the angiogram of an 18-month-old healthy control animal (A) showed no blood flow alterations, in the corresponding picture of a Tg2576 mouse (B) blood flow defects are clearly visible (arrow). As shown in (C), the latter correspond with a marked deposition of Aβ_1 - 40 (arrow) in the pial branches of the MCA. Moreover, in the Tg2576 mice a co-localization of Aβ plaques (D) and iron (E) in the adjacent brain sections was observed (small arrows). The inserts given in (D-E) represent an about 8-fold magnification of the regions marked with the red arrows. The scaling bars correspond to 500 μm (C) and 50 μm (D-E).

As illustrated by the microscopic picture given in Fig. 7C, the blood flow disturbances in the AD model animals coincide with the massive deposition of Aβ_1 - 40 in the territory of the pial branches of the MCA. In the healthy control animals, no accumulation of the peptide in the vessel walls was observable (not shown). Furthermore, the analysis of adjacent brain sections showed a coincidence of amyloid plaques (Fig. 7D) with the deposition of free iron (Fig. 7E) in the Tg2576 mice. These results leave open the possibility that the iron could be the leftover of free heme, which, considering the CAA shown in Fig. 7A-C, may result from microhemorrhages and hemolytic events in the brain tissue of AD patients. In summary, the detected Aβ-derived blood flow alterations and the co-localization of free iron with Aβ plaques may indicate the transient formation of Aβ-heme complexes during the pathology of AD.

The applied system is suitable to test the peroxidase activity of Aβ-heme complexes

In this study, ABTS-based absorbance measurements were used as a well-established method to study the peroxidase activity of Aβ-heme complexes [19]. Thereby we followed the ABTS^•+ formation at 734 nm to avoid interferences from other assay components [36] including free heme [58, 59]. Furthermore, most experiments were conducted at an excess of ABTS to avoid over-oxidation to ABTS^2 + and related side products [60]. The small absorbance decrease observed at longer measuring times may be attributed to ABTS backformation [60]. It was mainly observed in the presence of tyrosine-containing Aβ sequences, suggesting a radical transfer to the peptide. The measurements with free heme in the absence of Aβ yielded activities of 0.13 mU/ml (1 μM heme) and 1.03 mU/ml (10 μM heme), respectively, proving an adequate correlation between the concentration of the peroxidase-active component and the determined initial enzymatic activity.

Aβ promotes the peroxidase activity of free heme in a concentration-dependent way

The ABTS measurements clearly showed that Aβ promotes the peroxidase activity of free heme in a concentration-dependent manner. Thereby at ratios higher than 20:1 a saturating effect was observed, maybe indicating the formation of Aβ-heme complex with the named stoichiometry. Comparable results were also observed in other studies whereby, however, only ratios up to 8:1 were applied [19, 27, 61]. In line with others in the absence of free heme, we observed no peroxidase activity [25], proving a heme-based enzymatic activity of Aβ-heme complexes.

The relative effect of Aβ on the heme-derived peroxidase activity also depended on the absolute heme concentration: At higher heme concentrations, stronger effects of the peptide were observed. These results indicate the formation of Aβ-heme complexes with multiple heme units. Moreover, they suggest a non-linear increase in the peroxidase activity upon cerebral accumulation of free heme, especially in the presence of elevated Aβ levels observed in AD patients.

Another new aspect addressed in this study is the H₂O₂ dependence of the Aβ-heme complex-mediated peroxidase activity. Thereby either H₂O₂ was directly applied or the G/GO-system was used, which allows a continuous H₂O₂ production [62]. A direct comparison of the H₂O₂ concentration-dependent increase in the relative peroxidase activity yielded comparable effects for free heme (10 μM) and for Aβ-heme complexes (10 μM/1 μM), again proving a heme-based peroxidase activity of the complexes. Still at higher constant H₂O₂ production rates, a stronger promotion of the enzymatic activity with increasing Aβ concentrations was observed, which may indicate a coordinating effect of the peptide with regard to the heme activation by H₂O₂. In fact, in chordata heme peroxidases, the amino acids on the distal heme site are known to regulate the access of H₂O₂ toward the heme center [63, 64].

Selected studies at pH 6.5 showed higher peroxidase activities of Aβ-heme complexes as compared to the values obtained at pH 7.4. While the pH-dependent peroxidase activity of Aβ-heme complexes was never tested before, these results are well in line with the results obtained with heme peroxidases [63, 64]. The studies also revealed an influence of the pH value on the Aβ-derived promotion of the peroxidase activity, showing a stronger effect of the peptide at neutral conditions. These results can be explained by the fact that both, the binding of the peptide to heme [27] and the subsequent promotion of the heme-mediated peroxidase activity are influenced by the pH value [25, 35]: While for the binding of the heme a histidine residue (most likely H13) with a neutral net charge is responsible [19, 65], an acidic arginine (R5) promotes the stabilization of H₂O₂ as the sixth heme ligand [20] and its subsequent homolytic cleavage [15, 27]. Y10 is regarded as the second amino acid contributing to the peroxidase activity of Aβ-heme complexes [15, 25].

As up to low micromolar heme concentration were found in the brain tissue of AD patients, the performed peroxidase activity measurements (1/10 μM heme) are somewhat comparable to the physiological conditions [28]. Yet the applied Aβ concentrations (up to 100 μM) are far beyond the physiological range [66, 67]. Still such high peptide concentrations were necessary to explore the Aβ-heme ratio-dependent peroxidase activity. Moreover, the aggregation of Aβ as well as its deposition around blood microvessels [53, 68] may lead to high local concentrations of the peptide, especially at places of hemolysis-derived formation of free heme. Regarding H₂O₂ again unphysiological high concentrations were applied. Still the lowest concentration used in our studies (20 μM, see Fig. 2B) approximately meet (patho-) physiological H₂O₂ concentrations in the blood [69]. In contrast, by using 500 μM H₂O₂, a suitable high peroxidase activity could be observed to investigate, e.g., the Aβ sequence-dependent enzymatic activity of Aβ-heme complexes.

The three amino acids responsible for the binding (H13) of free heme and the subsequent peroxidase activity of Aβ-heme complexes in the presence of H₂O₂ (R5, Y10) resemble the corresponding structures in chordata heme peroxidases [19, 70]. Moreover, they represent the only differences between the human and the mouse sequence of the peptide [19, 20]. While we included only peroxidase activity experiments with Aβ_1 - 16 in the manuscript some preliminary experiments as well as literature data [61] clearly indicate a much higher peroxidase activity of the long peptide in the presence of free heme.

Human Aβ has the strongest impact on the peroxidase activity of free heme

We next tested the impact of the peptide sequence of Aβ_1 - 16 on the promotion of the heme-mediated peroxidase activity. Thereby special attention was paid to the three amino acids mutated in the human peptide (R5, Y10, H13) in comparison to the rodent peptide (G5, F10, R13). Already published mutational studies only cover a direct comparison between human Aβ and the rodent equivalent [27] or the replacement of single human-specific Aβ amino acids by non-rodent equivalents, namely R5N, Y10G, and H13A [19, 25]. Still the reported results are comparable to ours: While others also clearly showed a lower peroxidase of Aβ-heme complexes formed with the rodent peptide as compared to complexes formed with human Aβ, the differences were not quantified and only Aβ-heme ratios up to 8:1 were used [27]. In our studies, the rodent peptide only doubled the peroxidase activity of free heme, while in the presence of the human peptide, an about 8-fold higher peroxidase activity was observed.

Regarding single mutants for Y10G a marked decrease in the peroxidase activity was reported [25], which corresponds to our results. In fact, in our studies Y10G turned out as the single mutant with the strongest effect, suggesting Y10 as an important key player during the peroxidase activity of Aβ-heme complexes. However, for the corresponding rodent mutation (Y10F), only an about 1.4-fold decreased peroxidase activity was observed. As suggested by a recent publication, the named mutants influence the structure of the peptide, which also affects the peroxidase activity of the formed Aβ-heme complexes [61]. Regarding the artificial mutation R5N, again a strong decreasing effect on the peroxidase activity was reported [19], which is only partially reflected by our results. Again, the corresponding rodent single mutation (R5G) had a weaker effect. These differences may be attributed to varying experimental conditions, including the applied Aβ-heme ratio and H₂O₂ concentration. Another possible explanation comes from a recently published review: As stated by Kohdarahmi et al., R5 may have only a minor effect on the peroxidase activity of Aβ-heme complexes, most likely by influencing the structure and/or aggregation behavior of Aβ [61].

For H13A (and H14A), almost no effect on the peroxidase activity was reported, which was attributed to the fact that H13 (present in human Aβ) can be replaced by H14 (present in human and rodent Aβ), and vice versa, in regard to the binding of free heme [19]. Accordingly for the H13G, H14G double mutation, a peroxidase activity at the level of free heme was observed [19]. Regarding the single mutants, the results are only partially comparable to ours where a single mutation of H13 (H13R or H13A) already led to a considerably decreased peroxidase activity. Yet in line with others [19], upon application of the H13G, H14G double mutation, we observed a massive reduction of the peroxidase activity. In fact, the obtained value was even slightly lower than the peroxidase activity obtained in the presence of free heme and rodent Aβ_1 - 16. Yet unlike shown by others, it still exceeded the peroxidase activity observed in the sole presence of free heme [19].

As reported previously [61], for human Aβ (see Fig. 1), but not, for example, for the murine peptide, increasing Aβ-heme ratios lead to a considerable increase in the peroxidase activity. Thus, the differences obtained by using a 10/1 ratio may be small by applying equimolar Aβ-heme complexes.

In summary, regarding the tested single mutants, the replacement of Y10 showed the strongest effect, while lower impacts were found for H13 und R5. Yet considering that H13 may be replaced by H14 [33], the strongest impact was found for histidine (H13/H14), followed by the human-specific residues Y10 and R5. In fact, H13/14 is responsible for the binding of free heme [26, 65] while Y10 and R5 contribute to the subsequent peroxidase activity of Aβ-heme complexes [19, 25]. Most interestingly Y10 is not only regarded as a redox-active center contributing to the peroxidase activity of Aβ-heme complexes [19] but also as an endogenous target for this enzymatic activity, leading to the formation dityrosine [21, 29].

Human-specific Y10 is an endogenous peroxidase substrate of Aβ-heme complexes

Our studies confirmed the results of others [21, 25] that upon peroxidase activity of Aβ-heme complexes Y10 can be oxidized to form 1,3-dityrosine [42, 71]. Yet the impact of the Aβ-heme ratio and the amount of H₂O₂ on the dityrosine yield was never tested before.

Our fluorescence measurements suggest an optimal dityrosine formation at an Aβ-heme ratio of about 40:1, which contradicts the continuous increase of the peroxidase activity at higher peptide excesses. These apparently discrepant results may be explained by the fact that a strong peroxidase activity leads to the over-oxidation of tyrosine to products like trityrosine, isotrityrosine, and pulcherosine, which decrease the amount of dityrosine [71, 72]. Both upon direct addition of H₂O₂ and by applying the G/GO-system a greater dityrosine yield was obtained at higher H₂O₂ concentrations, which suggests that multiple peroxidase cycles of the Aβ-heme complexes are needed per dityrosine molecule to be formed.

Another new aspect addressed in this work is the effect of the Aβ_1 - 16 sequence on the dityrosine formation. In line with others, in the absence Y10 we observed no dityrosine formation at all, proving the specificity of the fluorescence measurements [19, 25]. Yet by testing single and double mutants of human Aβ_1 - 16, we also observed an influence of R5, H13, and H14 on the dityrosine yield: Strangely enough, by replacing the named amino acids by the rodent or an artificial equivalent, we always observed an at least slightly higher dityrosine formation. The most striking differences were observed for the H13R single mutant and the H13G, H14G double mutant were a doubled (H13R) and a not significantly reduced (H13G, H14G) dityrosine yield coincide with a slightly (H13R) or strongly (H13G, H14G) reduced peroxidase activity. As the named histidine residues are responsible for the proximal heme coordination, we guess that their replacement leads to the formation of less stable Aβ-heme complexes [19]. Accordingly, a lower peroxidase activity is observed, which, however, may translate to a higher dityrosine yield as over-oxidations and/or the formation of other tyrosine oxidation products than dityrosine are diminished [71, 72].

To confirm the formation of dityrosine upon Aβ-heme peroxidase activity, we also performed selected MALDI-TOF measurements. While in the presence of H₂O₂ we were able to show a minor dityrosine formation in the presence of free heme and a moderate product formation in the presence of HRP, the obtained product yields are scarcely comparable to the fluorescence measurements. Given the limited quantification of mass spectrometry data, in future studies dityrosine antibodies [19, 25] or thin-layer chromatography [71] should be applied. As a mechanism for the reported dityrosine formation, we propose a direct one-electronic oxidation by the activated Aβ-heme peroxidase [73], followed by a recombination of the formed tyrosyl radicals. In fact, Y10 is located directly at the active site of in Aβ-heme complexes, suggesting a direct oxidation by the activated oxoferryl state [28]. Yet it has to be stated that free ROS, including hydroxyl radicals (^•OH) [74], may also lead to Aβ-derived dityrosine formation under oxidative stress conditions of AD [42].

In summary, the obtained results clearly show the formation of dityrosine from Y10 due to the peroxidase activity of Aβ-heme complexes. As already reported in the literature the Y10-derived dityrosine formation in the presence of H₂O₂ [25] promotes the aggregation of Aβ via stabilization of peptide dimers [21, 29], which may be relevant for the progression of AD.

The formation of Aβ-heme complexes has several effects on Aβ fibril formation

Yet while Aβ-heme complex-mediated dityrosine formation promotes the fibrillation of the peptide, free heme is well known to inhibits Aβ aggregation [33] and even leads to the dismantling of partly formed fibrils [26, 31]. In fact, free heme interacts both with H13/H14, leading to the formation of peroxidase-active Aβ-heme complexes [27], and with F19/F20, which inhibits the fibrillation of the peptide [3, 26, 33]. Yet the mentioned studies on Aβ fibrillation in the presence of free heme have often been performed only at low Aβ-heme ratios [26, 31], lack long-term observation [31] and do not include the application of H₂O₂ [75] as a precondition for the Aβ-heme complex-derived dityrosine formation. By using ThT-based fluorescence measurements [31], in this study the named issues were addressed and clearly showed that even at low ratios (50:0.1) free heme already starts to delay and inhibit the formation of Aβ fibrils. Still it has to be stated that the initial interaction of ThT with the negatively charged peptide is of electrostatic nature [44, 45], which may provide an additional explanation for the observed lower fluorescence values.

Most interestingly also H₂O₂ itself apparently disturbed the formation of peptide fibrils. Yet these results may also be explained by the fact that oxidized Aβ fibrils yield a weaker fluorescence signal than native ones during the ThT measurements [21]. Moreover, we cannot rule out a direct oxidation of the dye. The same holds for the measurements in the presence of H₂O₂ and heme or HRP as well as for the HRP control. One interesting aspect observed during the measurements with ThT is the different kinetics in the fluorescence increase observed after application of 1 μM HRP or 10 μM heme and H₂O₂: In both cases a hyperbolic instead of sigmoidal fluorescence increase was observed. To our knowledge, such results were never obtained before and may indicate a completely different fibrillation behavior of Aβ in the presence of peroxidase activity [47].

The also performed STEM measurements clearly showed that despite the (apparent) inhibitory effect of heme, H₂O₂, and HRP on the Aβ fibrillation within the first 96 h, in all samples a peptide fibrillation finally takes place. Thereby Aβ fibrils with a diameter of about 14 nm were always observed, which is in the expected range [21, 34]. Still, while in the sole or combined presence of H₂O₂ or low amounts of free heme, the fibrils strongly resembled the control (Aβ_1 - 40 alone) at higher heme concentrations or in the presence of HRP different fibril morphologies were found. These results are in line with others [21] who also observed different fibril networks under oxidative conditions [21]. The observed denser fibril networks in the sole presence of higher heme concentrations are somewhat surprising as no peroxidase activity should take place in the absence of H₂O₂. Yet it may be explained by the release of free iron, which is well known to promote the aggregation of Aβ [26].

In summary, ThT measurements, especially at the chosen experimental conditions [44, 45], are a suitable tool to show the strong inhibitory effect of free heme on Aβ fibrillation. Yet they are not completely reliable under conditions were oxidative processes take place. Still the STEM measurements clearly indicate that the Aβ-heme complex-derived peroxidase activity influences the fibril morphology. Most likely the formation of dityrosine disturbs the classical Aβ fibrillation pathway, leading to an enhanced formation of, e.g., protofibrils [34]. This is nicely illustrated by the positive control (HRP + H₂O₂) where thinner and shorter fibrils were observed. Thus, free heme, upon complexation with Aβ, subsequent peroxidase activity and dityrosine formation, actually seems to promote the dimerization/oligomerization of the peptide.

While Alzheimer plaques contain both Aβ_1 - 40 and Aβ_1 - 42, in the current study, the shorter peptide was used as it was already discussed as the more toxic one, maybe due to its slower aggregation as compared to Aβ_1 - 42 [16]. Moreover besides this small difference in the aggregation kinetics both peptides were reported to be well comparable regarding their affinity toward heme [33]. Thus, our results are most likely well transferable to Aβ_1 - 42.

Yet it has to be clearly stated that the peroxidase activity measurements performed with Aβ_1 - 16 are only partly comparable to the aggregation studies performed with the full length peptide: As discussed in a recent publication [61], free heme most likely binds not only to H13/H14 in the hydrophilic part of the peptide but also to F19/20 in the hydrophobic part of Aβ_1 - 40/42 [26, 27]. In fact, the latter second heme binding site was already shown to affect the peroxidase activity of Aβ-heme complexes [37].

Most interestingly as even in mature Aβ fibrils the amino acids responsible for the formation of peroxidase-active Aβ-heme complexes (R5, Y10, H13/H14) are part of the unstructured solvent-facing side chains [11, 77]. Accordingly, the aggregation state of the peptide was shown to have no influence on this enzymatic activity [3, 37]. In fact, a recent review shows a positive correlation between Aβ aggregation and the peroxidase activity of Aβ-heme complexes [61]. In line with our results, this effect is attributed to peroxidase complexes with an Aβ-heme ratio higher than one [28, 61]. The last question addressed in this study refers to possible sources of free heme as a precondition for the formation of Aβ-heme complexes in the brain tissue of AD patients.

Aβ-derived CAA and hemolysis could lead to the formation of Aβ-heme complexes

We also performed MRA measurements and, as reported before [49], detected a decreased blood flow in Tg2576 mice as compared to the healthy control. Thereby in line with others [53, 66], a clear correlation with the vascular deposition of Aβ was observed. Furthermore, in the animals a co-localization of free iron with Aβ plaques was found, which again confirms published results [20, 21].

Originally a disturbed mitochondrial heme metabolism in neurons was suggested as a source for free heme and iron in the brain [23, 78]. Yet, based on the obtained results and literature studies, we suggest that hemoglobin may represent the major for cerebral free heme in AD. As illustrated in Fig. 8, the vascular deposition of Aβ (CAA) [6], resulting cerebral microhemorrhages [53, 68] and the hemolytic effect of the peptide [54, 66] could account as the major mechanisms. This theory is confirmed by the fact that hemolysis-associated acute phase proteins like Hpx [4, 79] and heme oxygenase 1 [80, 81] show a higher expression in AD patients. Moreover, hemoglobin-derived peptides are elevated in the brain of AD patients [82].

OPEN IN VIEWER

Fig.8

Proposed model for Aβ-derived microvessel destruction and hemolysis as a source for cerebral free heme. While monomeric Aβ readily crosses the blood-brain barrier, Aβ fibrils deposit around cerebral blood vessels and lead to microhemorrhages. A subsequent Aβ monomer and/or fibril-derived hemolysis of erythrocytes leads to the accumulation of free heme in the brain tissue. Especially in the presence of elevated cerebral Aβ levels, these processes will lead to the formation of Aβ-heme complexes, which, in the presence of H₂O₂, exhibit a peroxidase activity. The latter causes, e.g., oxidative stress in the brain tissue, neuronal death, and neurotransmitter destruction, thus, may considerably contribute to the pathology of AD.

In blood, Aβ is quickly taken up by erythrocytes and leads to increased cell sizes and decreased deformability of the cells [66, 67]. This effect, which leads to hemolytic events in the tight brain capillaries [83], was already attributed to the induction of oxidative stress in the erythrocytes [54, 84]. Thus, we cannot rule out a contribution of Aβ-heme complex-mediated peroxidase activity during these processes. Most interestingly, Aβ was also shown to inactivate catalase in erythrocytes, leading to an increased H₂O₂ concentration in these cells [85]. Moreover, Aβ-derived phospholipid oxidation in erythrocytes leads to increased endothelial binding of the cells [86, 87] and vascular destructions [86], providing positive feedback mechanisms for the accumulation of erythrocytes and free heme in the brain tissue. Up to micromolar cerebral heme concentrations were found in AD patients [28]. In addition endothelial and smooth muscle cells of cerebral microvessels can also produce AβPP [53], a process, which may be elevated upon Aβ-derived microhemorrhages.

Thus, the presented in vivo results confirm already published pathological features of AD, including CAA as well as the coinciding deposition of Aβ plaques and free iron. Still, considering literature data these observations may represent final remains of Aβ-derived microhemorrhagic and hemolytic events, which could lead to the formation of blood-derived Aβ-heme complexes as a possible hallmark of AD pathology. Accordingly, AD was already suggested as a thrombohemorrhagic disorder [88] whereby Aβ-heme complexes play a key role [20]. In fact, these complexes are most likely also formed in vivo and the resulting peroxidase activity contributes to the pathology of AD [89]. In order to prove the connection between CAA, in future studies we will try to detect free heme in the brain tissue of Tg2576 mice [28]. Furthermore dityrosine formation in the brain tissue of this AD model will also be addressed [90]. In summary, realizing the peroxidase activity of Aβ-heme complexes as an important pathological mechanism of this neurodegenerative disease may open the door for the development of new therapeutic strategies, including the application of antioxidants [22, 24] and/or heme peroxidase inhibitors [34]. Recently also the binding of free heme to the tau protein was shown [91], suggesting an even broader role of free heme at AD. Hyperphosphorylated tau protein and tau-containing neurofibrillary tangles are major hallmarks of AD besides Aβ plaques [20, 92].