Evidence supporting oxidative stress in a moderately affected area of the brain in Alzheimer’s disease

The pathogenesis of Alzheimer’s disease (AD) remains to be elucidated. Oxidative damage and excessive beta-amyloid oligomers are components of disease progression but it is unclear how these factors are temporally related. At post mortem, the superior temporal gyrus (STG) of AD cases contains plaques, but displays few tangles and only moderate neuronal loss. The STG at post mortem may represent a brain region that is in the early stages of AD or alternately a region resistant to AD pathogenesis. We evaluated expression profiles and activity of endogenous anti-oxidants, oxidative damage and caspase activity in the STG of apolipoprotein ε4-matched human AD cases and controls. Total superoxide dismutase (SOD) activity was increased, whereas total glutathione peroxidase (GPX), catalase (CAT) and peroxiredoxin (Prx) activities, were decreased in the AD-STG, suggesting that hydrogen peroxide accumulates in this brain region. Transcripts of the transcription factor NFE2L2 and inducible HMOX1, were also increased in the AD-STG, and this corresponded to increased Nuclear factor erythroid 2-related factor (NRF-2) and total heme-oxygenase (HO) activity. The protein oxidation marker 4-hydroxynonenal (4-HNE), remained unchanged in the AD-STG. Similarly, caspase activity was unaltered, suggesting that subtle redox imbalances in early to moderate stages of AD do not impact STG viability.


Results
Detection of advanced Aβ but early tau pathology in AD-STG. As previously described, the STG region of the AD cases (AD-STG) in this cohort showed typical AD pathology with significantly more plaques (both diffuse and cored) and AT8-positive neurons (Fig. 1A,B and Supplementary Table 1) 19 . Similarly, immunoblotting showed markedly increased AT8 tau-positive bands (between 45-79 kDa) in the AD-STG (Fig. 1B, refer to Supplementary Fig. S1). NFTs were relatively sparse, and neuronal counts were lower (16%) but not significantly different to controls (Supplementary Table 1). Taken together, the paucity of NFTs and minimal loss of neurons suggest that the disease, tau pathology in particular, is at an early stage in the AD-STG compared with the medial temporal lobe structures (Braak stage V/VI) from these same patients. AD cases also showed a significantly lower mean RNA integrity number (RIN) and brain pH and these two parameters were correlated with one another (r 2 = 0.2, p = 0.002) (Supplementary Table 1). Although no differences in the mean fixation (storage) time (p = 0.06) between AD cases and controls were noted, fixation time was noted to range from 6 to 155 months. Additionally, a difference in post-mortem interval (PMI) was observed between the two groups (Supplementary Table 1; p = 0.04).
Assessing anti-oxidant gene expression in AD and control STG. The disparity in mean RNA integrity number (RIN) identified between the cases and controls had the potential to skew gene expression analyses. In particular housekeeping genes need to reflect RIN differences between cases and controls rather than no difference as per normal 19 . The housekeeping gene Succinate Dehydrogenase Complex Flavoprotein Subunit A (SDHA) satisfied this modified criterion while displaying no difference between cases and controls in a RIN-matched subcohort (n = 21, data not shown). With SDHA as house-keeping gene, significant increases in both NFE2L2 (encoding the transcription factor NRF-2) (p = 0.008) and HMOX1 (encoding HO-1, the inducible form of HO) (p = 0.009) mRNA were determined in the AD-STG (Table 1). In contrast, there were no differences in HMOX2 (encoding the constitutively expressed form of HO,  or SOD1 mRNA with the latter varying widely in the AD-STG. Anti-oxidant protein levels and activity. There was no difference in SOD1, SOD2 and GPX1 protein levels between AD and control groups as judged by direct ELISA (Fig. 2A). However, total SOD activity increased significantly (p = 0.04) (Fig. 2B), while total GPX activity was decreased in the AD-STG compared to controls (p = 0.02) (Fig. 2C). This data was consistent with an imbalance in H 2 O 2 producing vs consuming enzymatic activity and warranted further investigation to assess other endogenous H 2 O 2 -consuming enzymes.
Surprisingly, the capacity for the AD-STG to degrade exogenously added peroxides was slightly greater than control-STG (p = 0.002), as judged by an ex vivo assays measuring rates of H 2 O 2 consumption (Fig. 3A). To understand the specific contribution of CAT to this peroxide consuming activity in control and AD brains, H 2 O 2 decomposition by added STG-homogenate was reassessed in the presence of aminotriazole (AT), a CAT-inhibitor. Under these conditions, H 2 O 2 consumption diminished more in the control group than in the AD cohort, suggesting a higher level of catalase specific activity in control STG (i.e., compared the difference in activity between the white bars (control) and then the black bars (AD) in the absence and presence of AT, Fig. 3A). Overall, we determined a 1.5-fold lower CAT activity in the AD-STG than the corresponding enzymatic activity in the control-STG (p < 0.001) (Fig. 3B). This decreased CAT activity exacerbates the documented loss of GPX activity in the same STG tissues (Fig. 2C) and together imply that the AD-STG shows a limited ability to enzymatically decompose accumulating H 2 O 2 .
Overall, total consumption of H 2 O 2 is likely to be mostly ascribed to a combination of CAT and heme-peroxidase activity in the clarified brain homogenates as sodium azide (SA; a highly active heme-peroxidase inhibitor) 20 , completely inhibited peroxide consumption in both the control and AD cohorts (Fig. 3A). Note, although azide is a potent inhibitor of catalase activity, azide-insensitive catalase activity has been reported in bacteria 21 , and this may impact on the interpretation of these inhibition studies.
The peroxiredoxin (Prx) family of anti-oxidant enzymes, are also a pathway for eliminating H 2 O 2 , and have also been implicated in the pathogenesis of AD 22 . Therefore, we measured total peroxiredoxin activity by assessing the selective consumption of cumene-hydroperoxide. Similar to catalase and total GPX activity, peroxiredoxin activity decreased slightly (and significantly) (p = 0.004) in the AD-STG compared to the controls (Fig. 3C). Together these data indicate that the AD-STG is less capable of detoxifying accumulating (low molecular weight) peroxides via multiple antioxidant enzymic pathways compared with the corresponding control tissue.
Additionally, total and cytosolic NRF-2 protein expression was assessed using a combination of immunohistochemistry and western blotting respectively. Consistent with the gene response, a statistically significant increase in total NFE2L2 expression was observed in the AD cohort (p = 0.05) ( Fig. 4A-C). To assess potential differences in NRF-2 subcellular localisation, NRF-2 protein levels were also assessed in cytosolic homogenate fractions. A recent consensus in the literature indicates that the molecular weight for NRF-2 occurs between 95-100 kDa 23 therefore, bands detected at mass 95 kDa were used quantify a significant decrease in cytosolic NRF-2 levels in the AD cohort (p = 0.004) ( Fig. 4D and E, refer to Supplementary Fig. S2), suggesting that NRF-2 levels are inherently lower in the AD-STG and/or that NRF-2 may be sequestered to the nucleus in this cohort.
Total HO activity, a downstream transcriptional target of NRF-2, was also assessed in the homogenized brain tissues. Overall, we determined a significant increase in total HO activity, as indicated by increased bilirubin production in the ex vivo assay (p = 0.04) ( . This result taken together with the corresponding gene studies that showed a selective increase in the inducible HMOX1 mRNA but not HMOX2 mRNA, suggest that active HO-1 protein is expressed in the AD-STG and this likely accounts for the parallel increase in total HO activity in these same tissues.  Table 1. Summary of STG gene expression in AD and control samples a . a Levels of mRNA in STG brain homogenates were assessed by RT-qPCR as described in the methods section Note, for HMOX 1 and HMOX 2, two controls and one case was removed due to non-compliant melt curve. For NFE2L2 and SOD 1, two samples failed to generate PCR products (due to low quality RNA) and one other sample was removed due to a failure in obtaining a PCR product for SHDA (hence normalizing to the house keeping gene was not possible in this case). Assessing oxidative damage in the AD-STG. Oxidative tissue damage was determined by immunostaining for 4-hydroxynonenal (4-HNE) + . There were no difference between cases and controls ( Fig. 6A,B) suggesting that despite the documented decrease in antioxidant enzymes responsible for detoxifying H 2 O 2 , oxidative protein damage remained unchanged. Semi-quantitation of the 4-HNE + staining confirmed no difference in levels of this biomarker of oxidation (Fig. 6C). While this specific form of oxidized damage was unaffected, the altered patterns of anti-oxidant response elements documented in the AD-STG are consistent with these tissues experiencing a subtle heightened oxidative stress.
Caspase-dependent apoptosis. To investigate whether changes in H 2 O 2 in the AD-STG promoted pro-apoptotic cell death pathways, caspase 3/7 activity was determined in control-and AD-STG homogenates. This biomarker of apoptosis was marginally higher in the AD-STG homogenates relative to the controls, although this did not reach statistical significance (p = 0.08) (Fig. 7). Therefore, our data are consistent with unaltered capsase-mediated cell death (i.e., intrinsic apoptosis) in the AD-STG.

Correlations between biochemical markers and AD pathology.
Overall there was a paucity of correlations identified and some inconsistency in the relationships using different biomarkers of AD. From those that showed linear relationships total HO activity was significantly and positively correlated with STG plaque areal fraction (p = 0.04, r 2 = 0.21) (Fig. 8A) as well as total NFTs in the AD-STG (r 2 = 0.24, p = 0.03) (Fig. 8B). In addition, total SOD activity was positively correlated with total NFTs (r 2 = 0.23, p = 0.03) (Fig. 8C), while total CAT (Fig. 8D) and Prx (p = 0.02, r 2 = 0.26) (Fig. 8E) activities were negatively correlated with AT8-tau positive cells. Interestingly no relationships between the areal fraction of plaques and any other antioxidant protein activities were identified in the brain tissues analysed here. Nor were plaques related to cytosolic or total NRF-2 expression (see Supplemental Fig. S4). Also, we found no significant correlations between AT8-tau and total NFT

Discussion
Despite its inherent retrospective nature post-mortem brain tissue remains the primary vehicle for mechanistic investigations in humans. We have previously suggested that investigating the less affected areas of the AD brain may represent a surrogate for more severely affected areas at an earlier stage in the pathogenesis of AD 24 . Albeit it is also possible that the STG is resistant to AD rather than belatedly affected. In an effort to understand early molecular events associated AD we investigated the STG, a region that is moderately affected in the AD brain with characteristic Aβ pathology, including neuritic plaques, but only mild tau-related changes and minor neuronal loss.
The popular ACH states that a build-up of soluble Aβ oligomers in the brain parenchyma triggers a sequence of neurotoxic events, including oxidative stress that eventually leads to NFT formation and neuronal death 25 . An alternative hypothesis has been proposed that places Aβ accumulation secondary to oxidative stress 26 . Here we specifically hypothesized that examining oxidative stress in a belatedly affected area of the AD brain (as evidenced primarily by Aβ accumulation) would allow us to help clarify the temporal relationship of oxidative stress in AD pathogenesis. The majority of our findings appear to support an increase in oxidative stress with plaques being associated with subtle changes in redox balance insufficient to result oxidative lipid/protein damage.
Attenuation of the antioxidant actions of peroxide-consuming enzymes CAT, GPX and peroxiredoxins with a parallel increase in SOD activity (responsible for converting the superoxide radical anion to H 2 O 2 ) likely creates an intracellular milieu containing excess peroxides in the AD-STG. Although we were able to show decreases in specific peroxidase and catalase activity, the total H 2 O 2 consumption measured in STG clarified homegenates was Additionally, H 2 O 2 consuming activity was assessed following pre-treatment with the catalase inhibitor, aminotriazole (AT) and sodium azide in both subject groups. (B) H 2 O 2 consumption attributed to catalase (difference between vehicle and AT consumption rates) displayed as a fold change relative to the control group (n = 5 controls and n = 5 AD subjects) (C) A modified FOX assay displaying the peroxiredoxin specific consumption of cumen H 2 O 2 (n = 19 control and n = 19 individual AD cases). Data represent mean ± SD; *p < 0.05, **p < 0.01 & ***p < 0.0001. initially higher in AD-STG. We employed a strategy using specific inhibitors or peroxide substrates to evaluate contributions of GPX, CAT and peroxiredxin activities, which were all determined to be lower in the AD-STG. Thus, enhanced dismutation of superoxide radical anion leads to elevated H 2 O 2 that is ineffectively degraded by virtue of decreased capacity for the STG to consume this pro-oxidant. That the total H 2 O 2 consumption was initially higher in the AD-STG than in control-STG requires further discussion. This apparent paradox may reflect methodological limitations or that H 2 O 2 is consumed by reactions with unbound metal ions released through tissue processing. Others have demonstrated that transition metals increase around plaques and NFTs in the AD a b c d e . Total HO activity in AD and control STG. Microsomal fractions of STG homogenates exposed to hemin and biliverdin reductase allowed for the HO dependent conversion of heme-to-bilirubin. Data reflect bilirubin levels, as determined by HPLC analysis. Data represent mean ± SD, n = 19 for control and n = 21 individual AD cases. *p < 0.05.  15 . However, our data indicating that sodium azide treatment inhibited the majority of ex vivo H 2 O 2 consumption, strongly suggesting that free metals play only a minor role in this process, with a majority of transition metals likely to remain bound to proteins and unable to participate in unregulated redox reactions. Furthermore, the increased consumption of exogenous peroxide is unlikely to reflect a methodological limitation as the outcome from the studies with the heme-poison sodium azide points to other pathways of ex vivo H 2 O 2 consumption linked to other peroxidases that are present in brain tissues and not identified here. A study by Gsell and colleagues provides further evidence for an accumulation of H 2 O 2 in the AD brain, as judged by assessments of the ratio of SOD/CAT activity. This study reported that SOD activity remained unchanged in the AD brain, however similar to outcomes we document here, a decreased CAT activity yielded an overall increase in oxidative stress through an accumulation of H 2 O 2 . These results were consistent in multiple post-mortem brain regions, ranging from mild to severely affected regions of AD 29 and we have built on this outcome by demonstrating decreased activity for other enzymes responsible for H 2 O 2 consumption in the AD brain.
Another study reports decreased mitochondrial SOD 2 levels and unchanged SOD1 levels in patient serum, relative to age and gender matched controls 30 . While any deviation from age and gender matched controls may still reflect altered redox homeostasis in AD, differences in SOD levels reported in our present study may reflect difference in redox states between the peripheral and central nervous system of early stage AD patients. Notably, small sample size, has been suggested to contribute to the inconsistent SOD serum levels that have been reported in the literature (reviewed by Chang et al. 31 ). Thus, Chang and colleagues identified that the three studies with over 80 cases all report significantly increased SOD levels. Moreover, differences in SOD levels may reflect different disease stages. Consistent with our biochemical outcomes measured in the STG, which reflects the early changes in AD etiology, comparisons of leukocyte SOD levels in patients with severe and mild AD vs corresponding control subjects suggest that SOD levels are elevated in the early stages of AD, however appear to be depleted in the later stages of disease progression 32 . Although our data suggests a likely accumulation of H 2 O 2 in the AD-STG, we found no evidence of increased ROS-mediated damage at least in the form of 4-HNE + immuno-staining. As previous reports indicate that differences in tissue-storage time influence immuno-staining intensities 33 , and so may impact on the interpretation of this immune-staining approaches, 4-HNE + immuno-staining was repeated with a sub-cohort that were more closely matched for storage time; no difference in result was obtained. Despite no change in 4-HNE + , we chose to assess additional oxidative markers to corroborate this outcome and demonstrate oxidative stress in the AD-STG. The transcription factor NRF-2 is now considered a master redox switch that initiates a suite of cyto-protective genes that are central to cell survival 34 . Both gene and possibly nuclear localisation of NRF-2 were increased in the AD-STG, suggestive of stress-induced NRF-2 activation early in AD pathogenesis. By contrast, studies using highly affected areas of the AD brain suggest that nuclear levels, and thus activation, of NRF-2 decrease in the AD brain 35,36 . This suggests that NRF-2 protein levels may decrease with increasing severity of AD pathology. Overall, our data together with outcomes from others 37,38 provide further evidence to suggest that therapies targeting an enhancement of NRF-2 dependent pathways may serve as a promising treatment for pre-symptomatic AD subjects.
Here we found an increase in expression of inducible HO-1 protein; a downstream target of transcriptional NRF-2. As levels of HMOX-2 mRNA remained unchanged the simplest explanation for the observed increase in total HO activity, measured in an ex vivo assay employing homogenates from AD-STG, is likely attributed solely to alteration in the inducible HO-1 protein. Although previous studies using AD hippocampal tissues also reported increased HO-1 expression, this has not translated to increased activity or neuroprotection 39 . While our data suggest that total HO activity may be enhanced and protective early in the development of AD pathology, it must be recognized that our assessment of HO activity was determined using an ex vivo assay that added liver-derived biliverdin reductase to convert biliverdin into bilirubin, with the latter biomarker of HO activity detected by liquid chromatography. To add biological context to our result, recent studies have indicated decreases in biliverdin reductase activity with increasing severity of AD pathology 40,41 thereby potentially limiting the conversion of biliverdin to the more potent antioxidant bilirubin. Limitations in the conversion of biliverdin to bilirubin may therefore be an important feature of whether HO-1 confers neuro-protection in different individuals, and in turn, potential resistance to developing advanced AD pathology.
Stocker and colleagues 42,43 have demonstrated soluble bilirubin to be a potent inhibitor of lipid peroxidation. Also, inhibition of HO-1 activity in stress induced endothelial cells has demonstrated increased sensitivity to ROS mediated damage, with restored effects of cell viability attributed to the addition of bilirubin 44 . A gene deletion of biliverdin reductase has further been associated with increased endogenous oxidative stress 45 , suggestive of a direct role for bilirubin in conferring protection by HO-1, albeit that carbon monoxide, another by-product of the catabolic heme pathway can also confer neuroprotection 46 .
Our data suggests a requirement for a functional HO/BV/BR system and that augmenting this pathway in AD may be a target for future therapy. However, while the NRF-2/HO-1 pathway may seem an attractive therapeutic target for AD 47 a comprehensive understanding of mechanisms regulating HO-1 expression in brain cells, whether other enzymes such as biliverdin reductase are active, and how these mechanisms link to neuropathological changes is essential before this pathway can be exploited for therapeutic development. Consistent with this conclusion, studies with a transgenic AD mouse model demonstrate 48,49 increased levels of oxidative stress, AD pathology and enhanced cognitive dysfunction when double crossed with a knockout of antioxidant defences, including deficiencies in SOD 50,51 , and GPX 52 , while overexpression of SOD appears to reduce ROS levels, memory deficits and plaque load in respective AD mice 48,49 . Similarly, inducing NRF-2 activity in transgenic AD mice yields enhances neuroprotection against oxidative stress attributed to increased HO-1 expression 53,54 .
On balance the majority of data obtained here suggest that oxidative stress is subtly elevated in the AD-STG, and potentially in the early stages of AD pathogenesis, as judged by altered activity of a number of antioxidant response elements albeit in the absence of detectable oxidative protein damage. The limited extent of pathology and largely preserved complement of neurons observed in the STG suggests that a degree of resistance to oxidative stress is operative. Our findings are consistent with a mechanism whereby the STG, relative to more severely affected areas in the medial temporal lobe, retains the ability to dismutate the superoxide radical anion into H 2 O 2 , stimulating NRF-2 to increase total HO-1 and perhaps BVR activity to ultimately resist neuronal loss that is characteristic of AD.

Methods
Case selection. As previously described 19 frozen and fixed specimens of superior temporal gyrus (STG) from 21 age, gender-and APOE ε4-matched AD cases and 19 neurologically normal controls were obtained from New South Wales Brain Banks (NSWBB) following approval by the University of Sydney's Human Research Ethics Committee (#HREC 11245/USYD 12734) and the NSW Scientific Advisory Committee. Pathological AD was diagnosed by a neuropathologist using standard diagnostic criteria 55 and cases were either Braak stage IV (n = 1), V (n = 4) or VI (n = 12); four cases did not fit staging criteria (see Supplementary Table S1). Brains were from subjects that were to the best of our knowledge unaffected by any infectious disease, or other known neurological disorders. Any brain sample obtained with post-mortem intervals (PMI) > 2.6 days or significant agonal events were excluded from this study. Control brain tissue employed here was derived from a cohort of coronial cases, with the primary mode of death recorded as myocardial infarction.
Tissue preparation. Samples of frozen brain tissue (30 mg) were suspended in 600 μL lysis buffer containing 6 μL beta-mercaptoethanol as provided by the Bioline Isolate II RNA Kit (Bioline, Australia; BIO-5207) then homogenized with a rotating piston arrangement (Wheaton Specialty Glass, NJ, USA) at 500 r.p.m. 56 , and centrifuged (5 min, 13,000 × g) prior to RNA extraction. Total RNA was extracted according to the manufacturer's instructions then quantified using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, USA); RNA quality was determined using a BioAnalyzer 2100 (Agilent Technologies, Santa Clara, USA) and individual RIN values shown in Supplementary Table S1. All samples were stored at −80 °C prior to use.
Additional samples of frozen human brain STG tissue (~100 mg) were cut into small pieces, suspended in 2 mL Buffer (50 mM PBS, pH 7.4) containing: 1 mM EDTA, 10 μM butylated hydroxytoluene and a Protease Inhibitor Cocktail tablet (Roche Diagnostics, Bern Switzerland; 1 tablet/50 mL buffer). The suspension was homogenized as described above and the homogenates divided into 0.2 mL aliquots, immediately snap frozen in liquid nitrogen and stored at −80 °C. As required, homogenates were thawed at 20 °C then centrifuged (5 min, 13,000 × g) to clarify the supernatant for use in all biochemical and immune-assays (see below). Protein concentrations for all samples were determined using the Bicinchoninic Acid (BCA) assay (Sigma-Aldrich, St Louis, USA) and average values used to normalize all subsequent biochemical parameters measured. Gene expression. Complementary DNA (cDNA) was synthesized using the Tetro cDNA Synthesis Kit (Bioline, Australia; BIO-65043) and gene expression was determined using RT-qPCR performed with a Roche LightCycler 480 system (Roche Diagnostics Corporation, Indianapolis, Indiana, USA) according to MIQE recommendations 57 as previously described 19 . Primer efficiencies and sequence pairs used for all RT-qPCR analyses are shown in Table 2. Melt curves were generated using the LightCycler 480SW v1.5.1 software to demonstrate a single PCR product (see Supplementary Fig. S7). RT-qPCR analysis was carried out using the GenEx qPCR data analysis software (v5.3.2) (MultiD Analyses, Goteborg, Sweden). Importantly, correlations were performed to assess the impact of RNA integrity number (RIN) and tissue pH on expression levels of the genes of interest studied here as described in detail elsewhere 19 . This approach afforded the selection of a suitable reference gene that normalized RIN/pH differences between case and control tissue samples.
In addition, FFPE-STG sections were immunostained for 4-hydroxy-2-nonenal (4-HNE + ) and NRF-2. Briefly, sections were rehydrated and antigen retrieved using a decloaking chamber (Biocare Medical, CA, USA). Samples were heat retrieved at 95 °C for 30 min in a pH 9 buffer (Agilent Technologies, Santa Clara, CA, USA, #S2367), cooled to 23 °C and treated with 3% v/v H 2 O 2 in dH 2 O to block for endogenous peroxidase activity. Sections were then incubated with a serum-free protein block (Agilent Technologies, #X0909) for The oxidation biomarker 4-HNE was selected as a measure for tissue oxidation status as critically reviewed by Esterbauer and colleagues 58 . The 4-HNE immuno-sections were imaged simultaneously using the same parameters with an Axio Scan.Z1 slide-scanning microscope (Carl Zeiss, Jena, Germany). Entire sections were scanned using a 10x objective and imported into MetaMorph software program (v7.6 Universal Imaging Corp., Downington, Pa, USA). The lasso tool was used to outline the grey matter (GM) and white matter (WM) for each section and the threshold tool used to calculate the percentage of area stained in the GM and WM of each case.
Where required NRF-2 immunoreactivity was imaged using an Axio Lab.A1 microscope (Carl Zeiss, Jena, Germany). Sections were imaged using a 40x objective, and converted to JPEG files for analysis via the ZEN software (Carl Zeiss, Jena, Germany). Pixel counts were then determined using ImageJ software. The average percentage of area stained in the GM of each section was used for statistical analysis.
Western blot analysis. Western blots were performed on tissue homogenates as described previously 59 .
Clarified supernatants were, diluted to matching protein concentration (23 μg total) and heated with Laemmli loading buffer (BioRad, Hercules, CA; 95 °C, 5 min) and separated on a 12% TGX gel (Bio-Rad, #161-0180) by SDS-PAGE. Prior to protein transfer to the blotting membrane all protein bands were visualized using a ChemiDoc TM MP system (BioRad) and total band densities were obtained via Image Lab v5.2. Proteins were then transferred onto a PVDF membrane, blocked in 5% w/v skim milk powder (in 0.1% v/v TBS-T; 1 h, at 23 °C), then incubated with a mouse anti-Phospho-Tau (AT8; Thermo Scientific, Waltham, MA, #MN1020) overnight at 4 °C. Membranes were blocked at 4 °C and incubated with a rabbit anti-NRF2 antibody (Thermo Fisher Scientific, #710574) for 3 h, at 23 °C and then incubated with either an anti-mouse or anti-rabbit HRP-conjugated secondary antibody (dilution 1:5000 v/v; 1 h at 23 °C). Protein bands were visualized using Luminata Forte HRP substrate (Millipore, Billercia, USA) and imaged on a ChemiDoc TM MP system (BioRad). Antioxidant activity assays. Total Superoxide dismutase (SOD) activity. The combined SOD1/2 (or total SOD) activity 60 was assessed in samples of clarified supernatants (20 µL volume loaded) by measuring the inhibition of pyrogallol auto-oxidation monitored at 405 nm at 5 min intervals over 1 h of monitoring to establish a linear rate of superoxide radical anion evolution. Total SOD activity was expressed as a relative rate per min and normalized to the total protein level in the corresponding samples.
Total Glutathione peroxidase (GPX) activity. Assessment of total GPX activity was based on the GPX-dependent consumption of NADPH monitored at 340 nm 61  decomposition was assessed by the gradient of the linear data analysis (coefficient R 2 ranging 0.9-0.95) and normalized to total protein in the corresponding samples with CAT activity ascribed specifically to an AT-inhibited fraction of peroxide consumption.
Total Peroxiredoxin (Prx) activity. A modified ferrous oxidation-xylenol orange (FOX) assay was used to assess total Prx activity as described previously 64 . Cumene hydroperoxide was selected to enhance specificity in this enzyme activity assay as Prx activity shows enhanced catalytic efficacies and specificities toward tertiary lipid hydroperoxides 65 . Samples of clarified supernatants (3.5 µL) were incubated with 100 µM cumene hydroperoxide and 100 µM dithiothreitol (final volume of 10 µL) for 5 min at 20 °C. An aliquot of FOX solution (250 µM ammonium sulfate, 100 mM D-sorbitol and 125 µM xylenol orange) was then added to yield a total volume of 100 µL and the mixture was incubated at 20 °C. After 30 min, absorbance was read at 560 nm using a FLUOstar Omega reader and data was normalized to total protein.
Total hemeoxygenase (HO) activity. Clarified supernatants from brain homogenates (500 μL) were subject to ultracentrifigation (100 000 × g; 1 h at 4 °C). The resulting microsomal pellets were resuspended in 100 µL PBS and used to determine total heme-oxygenase (HO 1/2) protein activity. For total HO activity microsomes were mixed at a final ratio f 2:1 (brain homogenate-microsomes/isolated at liver-microsomes) in 100 µL buffer (250 mM sucrose, 20 M Tris, pH 7.4). Next, the microsomal mixtures were treated with 1 mM NADPH, 2 mM D-glucose-6-phosphate (G-6-P), 1U glucose-6-phosphate dehydrogenase and 1 µL of 2.5 mmol/L hemin and incubated at 37 °C in the dark. After 1 h, the reaction was stopped by addition of 100 µL ethanol/DMSO (95:5 v/v). The mixture was then centrifuged at 13000 × g for 5 min, and the supernatant was assayed for bilirubin with liquid chromatography 20 . Finally, total HO activity was determined as the protein normalized level of bilirubin.
Caspase activation. Caspase-3 and -7 activities were assessed using a commercial kit (Caspase-Glo TM 3/7, Promega, Fitchburg, USA) as per the manufacturer's instructions. Briefly clarified supernatants (50 μL) were diluted according to individual protein concentrations to yield final protein concentrations of 0.5 mg protein/ mL. These protein-adjusted samples were then added to Caspase-Glo TM 3/7 reagent (1:1 v/v, 20 °C for 60 min) and luminescence measured using a FLUOstar Omega reader (BMG LABTECH, Offenburg, Germany). Total caspase 3/7 activity was represented as mean luminescence after 60 min of monitoring.
Statistical analysis. All data are presented as the mean ± SD. Statistical tests including two-tailed t-tests (for paired data sets), and ANOVA were performed with GraphPad Prism (v6); p ≤ 0.05 was regarded as significant.
Data availability statement. The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).