The mitochondria-targeted antioxidant MitoQ inhibits memory loss, neuropathology, and extends lifespan in aged 3xTg-AD mice

Abstract
Oxidative stress, likely stemming from dysfunctional mitochondria, occurs before major cognitive deficits and neuropathologies become apparent in Alzheimer’s disease (AD) patients and in mouse models of the disease. We previously reported that treating 2- to 7-month-old 3xTg-AD mice with the mitochondria-targeted antioxidant MitoQ (mitoquinone mesylate: [10- (4,5-Dimethoxy-2-methyl-3,6-dioxo-1,4-cyclohexadien-1-yl)decyl](triphenyl)phosphonium methanesulfonate), a period when AD-like pathologies first manifest in them, prevents AD-like symptoms from developing. To elucidate further a role for mitochondria-derived oxidative stress in AD progression, we examined the ability of MitoQ to inhibit AD-like pathologies in these mice at an age in which cognitive and neuropathological symptoms have fully developed. 3xTg- AD female mice received MitoQ in their drinking water for five months beginning at twelve months after birth. Untreated 18-month-old 3xTg-AD mice exhibited significant learning deficits and extensive AD-like neuropathologies. MitoQ-treated mice showed improved memory retention compared to untreated 3xTg-AD mice as well as reduced brain oxidative stress, synapse loss, astrogliosis, microglial cell proliferation, Aβ accumulation, caspase activation, and tau hyperphosphorylation. Additionally, MitoQ treatment significantly increased the abbreviated lifespan of the 3xTg-AD mice. These findings support a role for the involvement of mitochondria-derived oxidative stress in the etiology of AD and suggest that mitochondria- targeted antioxidants may lessen symptoms in AD patients.

1.Introduction
Alzheimer’s disease is a neurodegenerative disorder characterized by increasing cognitive impairment and dementia coincident with progressive damage to brain neuronal tissue.Neuropathological hallmarks of the disease include the development of extracellular amyloid plaques in the brain and the formation of neurofibrillary tangles (NFT) within neurons. The plaques are primarily composed of amyloid β (Aβ) peptides cleaved from the amyloid precursor protein (APP; O’Brien and Wong, 2011) and the tangles of hyperphosphorylated tau, a cytoskeleton-associated protein (Weingarten et al., 1975; Delacourte et al., 1999; Frost et al., 2015). Oxidative stress occurs early in the development of AD, preceding development of most plaques and tangles and continuing throughout disease progression. Considerable evidence suggests that this stress may be a key mediator of the disease (Gutteridge, 1994; Praticò et al., 2002; Pope et al., 2008; Praticò, 2008; Moreira et al., 2010; Federico et al., 2012; Castellani et al., 2016).A primary source of oxidative stress are reactive oxygen species (ROS) derived from the mitochondria electron transport chain (ETC; Nicholls and Ferguson, 2013; Halliwell and Gutteridge, 2015). Production of ROS by mitochondria occurs when electrons leak from the ETC to reduce dioxygen to the free radical superoxide (O2.-). The O2.- converts to additional ROS and other reactive species (RS). MitoQ is a mitochondria-targeted antioxidant composed of ubiquinone, a component of the ETC, covalently bound via a ten carbon chain to triphenylphosphonium (TPP+), a lipophilic cation that targets the ubiquinone moiety to the inner mitochondrial membrane driven by the high electrochemical potential across that membrane (Kelso et al., 2001; Smith et al., 2003). TPP+ remains on the matrix side of the membrane, allowing the ubiquinone to penetrate the membrane where respiratory complex II reduces it toubiquinol. MitoQ acts as an antioxidant when oxidized back to ubiquinone by various RS (James et al., 2005; Smith and Murphy, 2010).

Complex II reduces the ubiquinone to repeat the cycle.Because MitoQ is a poor substrate for complex I and III, it cannot substitute for endogenous ubiquinone and thus does not take part in mitochondrial respiration (James et al., 2005). It instead acts as a renewable antioxidant. MitoQ can cross the blood-brain barrier and, following a Nernstian distribution driven by the high potential across the inner mitochondrial membrane, concentrates hundreds of fold in mitochondria as compared to serum concentrations (Murphy and Smith, 2007). Ad libitum exposure to MitoQ in drinking water results in establishment of a steady-state concentration across this membrane within a few days (Smith et al., 2003). MitoQ is excreted unchanged or with sulfation and glucuronidation in the urine and bile (Li et al., 2007). Long-term ad libitum water administration of MitoQ to mice causes no major changes in physical activity, oxygen consumption, food consumption, body weight, or a number of other physical parameters (Rodriguez-Cuenca et al., 2009).3xTg-AD mice express three mutant human transgenes, two that cause early-onset AD (APPswe and PS1M146V) and one that causes frontotemporal dementia (tauP301L; Oddo et al., 2003). While all mouse models of AD have limitations (Li et al., 2016), 3xTg-AD mice are perhaps the best for investigating development of AD as AD-like pathologies appear in them in an age- dependent sequence comparable to that seen in humans. Cognitive deficits occur as early as four months of age (Billings et al., 2005). Mitochondrial dysfunction and oxidative stress precede these deficits, comprising some of the earliest pathological signs noted in these mice (Resende et al., 2008; Yao et al., 2009).

By six months of age, the presence of synaptic dysfunction and extracellular plaques are detectable. Tau pathology, present in this model as NFT, appears at about 12 months of age (Oddo et al., 2003). We previously evaluated the effect of MitoQtreatment on cognitive performance and neuropathology in young 3xTg-AD mice (McManus et al., 2011). In that study, we exposed 3xTg-AD mice to MitoQ in their drinking water for five months beginning at two months of age. This treatment prevented cognitive decline and associated AD-like pathologies.To further assess the role that oxidative damage has in disease progression, we treated female 3xTg-AD mice with MitoQ after AD-like pathology had developed. Beginning at 12 months of age and continuing for five months, mice received MitoQ in all drinking water. This treatment significantly improved spatial memory retention and inhibited brain oxidative stress, astrogliosis, microglia cell proliferation, amyloid plaque formation, Aβ accumulation, tau hyperphosphorylation, and formation of NFT. MitoQ-treatment also increased 3xTg-AD lifespan, suggesting that the neuropathology ameliorated by MitoQ treatment is responsible for the accelerated death rate of these mice.

2.Material and methods
Mitoquinone mesylate ([10-(4,5-Dimethoxy-2-methyl-3,6-dioxo-1,4-cyclohexadien-1- yl)decyl](triphenyl)phosphonium methanesulfonate) complexed to β-cyclodextrin was a gift from Michael P. Murphy via GlycoSyn Technologies. We purchased all other reagents from Sigma-Aldrich unless otherwise noted.The 3xTg-AD mouse model used in this study expresses three mutant human transgenes: amyloid precursor protein (APPswe), presenilin-1 (PS1M146V), and four-repeat tau (tauP301L; Oddo et al., 2003). Both of the mutations in amyloid precursor protein and presenilin-1 are associatedwith early-onset familial forms of AD and the tau mutation with frontotemporal dementia. AD- like symptoms of the disease appear in these mice as early as three-four months of age and continue to progress with time. We obtained founding 3xTg-AD breeders and control mice with the same 129/C57BL/6 hybrid genetic background but lacking the transgenes from The Jackson Laboratory. The study used only female 3xTg-AD mice and control mice because female 3xTg- AD mice develop greater Aβ burden than male mice (Carroll et al., 2010). The Jackson Laboratory also indicates that the female AD phenotype is more consistent than the male one.Beginning at 12 months of age and continuing for five months, female mice 3xTg-AD mice were administered 100 μM MitoQ complexed to β-cyclodextrin (1:4 ratio) in all drinking water. β-cyclodextrin, a cyclic oligosaccharide, was included to aid in solubilizing MitoQ, a lipophilic compound. Due to rapid degradation in the intestine and the ability of intestinal sugar transporters to transport only monosaccharides, β-cyclodextrin is not absorbed into the blood stream (Shimpi et al, 2005).

Control animals did not receive β-cyclodextrin in their water as the quantity used to solubilize the MitoQ led to daily mouse consumption that was well below that known to have detectable effects on mouse plasma lipids levels (Wagner et al., 2008).Littermate controls and nonTg controls with the same 129/C57BL6 genetic background had access to drinking water that did not contain MitoQ. All mice were group housed in our animal facilities, given access to the same rodent chow, Bed-o’Cobs bedding (Andersons Lab), and were maintained on a 12 h light/dark cycle. All animal procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.Following the five-month treatment period, mice were assessed for spatial memory retention using the Morris Water Maze (MWM; Morris, 1984). For eight consecutive days, mice underwent acquisition trials in which they were trained to find and escape onto a hidden platform within the water maze. The water maze consisted of a circular aluminum tank (4ft diameter) filled with opaque water and one slightly submerged plexiglass platform,14 cm in diameter.Water, made opaque with nontoxic white tempura paint, was maintained at 24 ± 1°C. Behavioral assessments were conducted as previously described with minor changes (McManus et al., 2011). Mice were placed on the hidden platform before the first acquisition trial for 10 s to reduce stress and establish existence of an escape platform. Acquisition trials followed in which mice were placed in the water maze at one of four predetermined starting points and allowed a 60 second free swim to escape onto the platform. Mice unable to find the platform were manually guided there and allowed 30 s on the platform to become familiar with distinct spatial cues in the test area. Each mouse underwent four trials each day with a 30 s rest on a warm towel between trials.

Trials continued until mice met escape latency criterion defined as reaching the escape platform within 20 s or less. Spatial bias was determined in probe trials 1.5 and 24 h after the last acquisition trial. The platform was then removed and mice allowed a 60 s free swim. The time spent in the quadrant where the platform had been previously located was determined. To account for possible sensorimotor deficits, mice were subjected to a cued acquisition trial following the last probe trial. In the cued trials, the platform was replaced in the pool and visibly marked with a flag. Mice were placed in the maze at a novel position and allowed to find the newly placed platform. Times to reach the platform and swim speed were determined. Each mouse was handled and assessed for general health prior to cognitive assessment. All trials wererecorded and analyzed using Ethovision Tracking software (Noldus Inc.) and SigmaPlot 11.1 (Systat Software).Mice were sacrificed in accordance with our animal use protocol with carbon dioxide followed by cervical dislocation. Brains were rapidly removed and split sagittally. Each cerebral hemisphere was either fixed in 4% paraformaldehyde for immunohistochemistry or used fresh or snap-frozen and stored at -80°C for immunoblotting or biochemical assays.Harvested brain tissues were homogenized in radio immunoprecipitation assay (RIPA) buffer (50 mM Tris, 0.5% Sodium deoxycholate, 1% Triton X-100, 150 mM NaCl) supplemented with a protease inhibitor cocktail (Sigma-Aldrich Cat# P2714). Samples were centrifuged at 13,000 rcf at 4°C for 15 m using an Eppendorf 5417R centrifuge. Equal amounts of protein were separated via SDS-PAGE, and transferred to PVDF (Millipore) membranes for 1 hour in cold transfer buffer. The membranes were blocked with 5% non-fat milk in TBS-T (10 mM Tris-HCL, 100 mM NaCl, and 0.1% Tween-20) for 30 minutes at room temperature. Afterward, membranes were incubated with primary antibodies against glial fibrillary acid protein (GFAP) at 1:1000 (Thermo Fisher Cat# PA3-16727, RRID: AB_2109795), anti-synaptophysin at 1:1000 (Millipore Cat# MAB525820UG, RRID: AB_11214133), anti-nitrotyrosine at 1:1000 (Invitrogen Cat# ab61392, RRID: AB_942087), or anti-tau 5 at 1:500 (Santa Cruz Cat# sc-58860, RRID: AB_785931) at 4°C overnight.

Membranes were then washed for at least 20 m in TBS-T andincubated at room temperature in an anti-mouse or anti-rabbit HRP-linked secondary antibody at 1:1000 (Cell Signaling Cat# 7076 and 7074, RRID: AB_330924 AB_2099233) followed by another 20 m wash. Anti-β-actin at 1:300 (GenScript Cat# A00730-40, RRID: AB_914100), and anti-β-tubulin at 1:1000 (Thermo Cat# PA1-41331, RRID: AB_2210397) were used for loading controls. Proteins were detected using Chemiluminescent ECL Western Blotting Substrate (Pierce).Individual cerebral hemispheres were fixed for 48 h in 4% paraformaldehyde, embedded in paraffin, cut into 12 um sections, and mounted on glass slides. Sections were deparaffinized and rehydrated through a series of incubations in xylene and ethanol. Following rehydration, antigen retrieval was achieved with heated 10 mM sodium citrate buffer pH 6.0 at 95°C for 10 m in a humidity chamber. Antigen-retrieved sections were then incubated for 30 m in 0.3% H2O2 in MeOH and blocked with Vectastain Universal blocking serum (Vector Laboratories) at room temperature for 30 m. Following blocking, sections were incubated with anti-AT8 (Thermo Cat# MN1020, RRID:AB_223647), anti- tau5 (Santa Cruz Cat# sc-58860, AB_785931), and anti- AB42 (Bioss Inc. Cat# bs-0107R, RRID:AB_10858046) overnight at 4°C. Sections were visualized using an ABC immunoperoxidase kit from Vector Laboratories with diaminobenzidine substrate.Soluble and insoluble fractions of Aβ(1-42) were detected in both whole brain tissue and hippocampal tissue with a BetaMark Colorimetric ELISA kit (Covance). Briefly, tissue washomogenized in ice cold 0.6% SDS lysis buffer (50 mM Tris, 2 mM EDTA, 150 mM NaCl) supplemented with a protease inhibitor cocktail (Sigma-Aldrich Cat# P2714).

Samples were centrifuged at 25,000 rcf at 4°C for 1 h. Supernatant containing soluble amyloid or amyloid peptide standards were added to ELISA plates in duplicate and incubated overnight at 4°C. The following day the 96-well plate was thoroughly washed, incubated with tetramethylbenzidine substrate for 50 m at room temperature while protected from light, and the colorimetric product was determined by absorbance at 620 nm with a SpectraMax M2 microplate. The Bradford assay (Pierce) was utilized to ensure equal loading of protein.Caspase 3/7 activity was measured using a Caspase-Glo 3/7 kit (Promega) following the manufacturer’s instructions. Briefly, brain tissue from each treatment group was homogenized in ice-cold hypotonic extraction buffer (25 mM HEPES, pH 7.5, 5 mM MgCl2, 1 mM EGTA) and centrifuged at 13,000 rpm for 15 minutes at 4°C. Protein concentrations were determined with the Bradford Assay (Pierce) and then diluted with PBS to achieve equal protein loading. Samples were incubated in a white-walled, clear-bottom 96-well plate for 1 hour with equal volume Caspase-Glo reagent. Luminescence (relative light units) was measured by a SpectraMax M2 microplate reader.Graphical representations were made and statistical significance measures determined with SigmaPlot 11.1 (Systat Software). Appropriate statistical analyses were conducted for experiments based on data distribution. Statistical comparisons were made via one-way ANOVAon Ranks with Dunn’s or Holm-Sidak post-hoc tests or by repeated measures ANOVA. Error bars represent ± SEM.

3.Results
3xTg-AD mice were supplied with 100 μM MitoQ in their drinking water beginning at 12 months after birth and continuing until the18th month. At the end of the treatment period, spatial learning and memory retention were assessed with the MWM (Morris, 1984). All mice were able to achieve a baseline criterion defined as finding and escaping to the MWM platform in 20 s or less during the training trials (Fig 1A). Eighteen-month-old, MitoQ-treated 3xTg-AD mice reached criterion two days before age-matched littermate 3xTg-AD controls. Thus, MitoQ treatment significantly improved performance during acquisition trials.To determine the effect of MitoQ treatment on memory retention, the escape platform was removed and mice were allowed a 60 s free swim. Spatial bias was measured by the amount of time the animals spent looking for the platform in the quadrant where the platform was previously positioned. Memory retention was determined at 1.5 and 24 h after the last training trial. The nonTg mice had significantly better memory retention at both time points as compared to 3xTg-AD mice. MitoQ increased retention in the 3xTg-AD mice to that of the nonTg mice at both time points (Fig. 1B). All mice were evaluated prior to the start of the training trials for general health. Swim speed was measured in the first trial. There were no discernable differences between the treatment groups, indicating that the mice had equal ability to reach the platform(Fig. 1C). Following the last acquisition trial, mice were subjected to a cued trial where the platform was placed back into the MWM at a visible novel location marked with a flag. All mice escaped the MWM to the cued platform within the same period. No significant differences were observed among groups, indicating that genotype or MitoQ treatment had no effect on sensorimotor capabilities (Fig. 1D; p = 0.107).

Synaptic dysfunction is a likely contributing factor causing the cognitive deficits in 3xTg-AD mice. There is a marked increase in synaptic dysfunction with increasing age in these mice coincident with a significant decline of the presynaptic vesicle glycoprotein synaptophysin in their brains (Oddo et al., 2003; McManus et al., 2011). MitoQ treatment of young 3xTg-AD mice inhibits synaptic loss concurrent with MitoQ-mediated inhibition of spatial memory retention deficits (McManus et al., 2011). Our behavioral data with the aged MitoQ-treated animals suggested that this treatment may have a protective effect on synaptic function in aged 3xTg-AD mice, as well. We used immunoblotting to quantify synaptophysin levels in cortical tissues of 18- month-old female nonTg, 3xTg-AD, and 3xTg-AD mice that had received MitoQ treatments for the preceding five months. The 3xTg-AD mice had much lower levels of synaptophysin than nonTg animals (Fig. 2). MitoQ treatment significantly inhibited synaptophysin loss in these animals, suggesting that preservation of synapses could be behind the improved cognitive performance of the MitoQ-treated mice.Increased oxidative stress, a consequence of the overproduction and/or reduced clearance of RS, precedes most other neuropathologies associated with AD (Nunomura et al., 2001, 2006; Praticò et al., 2002; Castellani et al., 2016). MitoQ’s antioxidant moiety, ubiquinone, is reduced bycomplex II into ubiquinol, returning ROS-forming electrons to the ETC and regenerating the antioxidant ubiquinol moiety (Frei et al., 1990; James et al., 2005; Smith and Murphy, 2010).Oxidative stress can cause damage to DNA and proteins. Superoxide reacts with nitric oxide (NO) to produce peroxynitrite (ONOO-), a RS that can cause oxidative modifications to tyrosine residues in proteins.

Nitrated modifications to proteins are elevated in humans with AD and in mouse models of AD (Smith et al., 1997; McManus et al., 2011). Our previous studies revealed that MitoQ attenuated ONOO–associated reactive species in cortical neurons in vitro and decreased nitrated modification of proteins in the brains of young female 3xTg-AD mice (McManus et al., 2011). To determine MitoQ’s effect on nitrated modifications in late stage AD- like pathology, we measured nitrotyrosine levels in 18-month-old female mouse brain tissue with immunoblots (Fig. 3). Nitrotyrosine levels were greatly elevated in the 3xTg-AD brains as compared to that in the brains of nonTg animals. While MitoQ did not eliminate the presence of nitrated products, it did significantly reduce the presence of nitrotryosine compared to non- treated 3xTg-AD mice.Increased numbers of reactive astrocytes (astrogliosis) and microglia cells, indicative of injury and inflammation, are found in postmortem brain tissue of human AD patients and in the brains of mouse models of AD (Oddo et al., 2003; Janelsins et al., 2005; McManus et al., 2011; Hansen et al., 2018). In our previous work, we showed that MitoQ treatment inhibited astrogliosis in young 3xTg-AD mice (McManus et al., 2011). To evaluate the effect MitoQ had on astrogliosis in late stage AD-like pathology, we used the astrocyte marker GFAP (Yang and Wang, 2015).MitoQ reduced GFAP band density in immunoblots of 3xTg-AD mouse brains by ~4-fold(Fig.4A) indicating it decreased astrocyte proliferation. Immunoblots of ionized calcium binding adaptor molecule 1 (Iba1), a microglial/macrophage marker (Imai et al., 1996), were elevated in the 3xTg-AD brains compared to nonTg controls. Figure 4B shows that Iba1 levels in brains from MitoQ-treated 3xTg-AD animals were significantly lower than Iba1 levels in the brains of both untreated nonTg animals and the brains of MitoQ-treated 3xTg-AD mice indicating significant inhibition of microglia cell proliferation by MitoQ.Considerable evidence supports the existence of a positive feedback loop between Aβ deposition and oxidative stress. Introduction of pro-oxidants into neuronal cultures promotes the production of Aβ and stimulates signaling pathways that contribute to increased APP cleavage (Misonou et al., 2000; Tamagno et al., 2002, 2005, 2008; Quiroz-Baez et al., 2009).

Specific interactions between Aβ and mitochondria can further propagate mitochondrial dysfunction and contribute to accelerated release of RS (Caspersen et al., 2005; Manczak et al., 2006; Mao et al., 2012). A direct correlation between oxidative stress and increased Aβ production in vivo has been demonstrated by introducing oxidizing agents directly into the hippocampus of wild-type mice, and observing a localized increase in Aβ deposition and pathological conformational changes associated APP cleavage enzymes (Arimon et al., 2015).Intraneuronal Aβ staining is detectable in the brains of 3xTg-AD mice by four months after birth and appears extracellularly by six months (Oddo et al., 2003). By 12 months after birth, these pathologies are well established. Consistent with a role for RS in increasing brain Aβ levels, our previous work demonstrated that treating young 3xTg-AD mice for five months with MitoQ greatly inhibited the deposition of Aβ within their brains. To determine whether treatmentof older animals with MitoQ would have a similar effect, we used immunohistochemistry and ELISA to evaluate Aβ(1-42) burden in the brains of untreated 3xTg-AD mice, 3xTg-AD mice treated with MitoQ for 5 months, and nonTg mice. Photomicrographs of cortical Aβ(1-42)-stained sections revealed intracellular and extracellular Aβ(1-42) deposition (Fig. 5A). Sections from nonTg mice showed little if any Aβ(1-42) staining, while those from MitoQ-treated 3xTg-AD animals had reduced staining compared to the untreated 3xTg-AD mice. Soluble Aβ(1-42) in the brains of 3xTg-AD mice was about 10-fold higher than in nonTg mice (Fig 5B). Five months of MitoQ treatment reduced this level by more than 2-fold to about 4-fold higher than that found in the brains of nonTg mice. The reduction in Aβ(1-42) deposition in the brains of 3xTg-AD mice by MitoQ treatment provides further evidence that increased mitochondria-derived RS modulates APP cleavage and increases Aβ deposition.MitoQ treatment decreased neurofibrillary tangles, total tau levels, phosphorylated tau levels, and caspase activity in the brains of 3xTg-AD miceTau pathology manifests in the 3xTg-AD mouse model in the form of NFT and is detectable in their brains by 12 months of age (Oddo et al., 2003).

We used immunohistochemistry to investigate formation of NFT in the brains of 18-month-old female nonTg, 3xTg-AD, and 3xTg- AD mice that had received MitoQ treatment for the preceding five months. Figure 6A shows that MitoQ-treated 3xTg-AD mice had fewer NFT in their cortex than did untreated 3xTg-AD mice. No tangles were apparent in nonTg mice. We quantified total tau protein expression and tau phosphorylation levels by immunoblot. Total tau levels were significantly elevated in the brains of 3xTg-AD mice compared to nonTg ones. MitoQ-treated mice had total tau levels similar to those of nonTg mice, indicating that MitoQ decreased the build-up of tau in the brain (Fig 6B). Tau phosphorylation levels, occurring at AD-associated phosphorylation residues Ser202/205,were about 10-fold higher in immunoblot band densities in the 3xTg-AD brains than in the nonTg ones. MitoQ treatment reduced phosphorylated tau to a level that was indistinguishable from that of nonTg brains (Fig.6C).Caspase-mediated cleavage of tau is involved in the development of NFT (Rissman et al., 2004; Rohn et al., 2008; de Calignon et al., 2010). Given that stabilization of oxidative stress in young 3xTg-AD mice inhibits elevated caspase activity (McManus et al., 2011), we evaluated MitoQ’s effect on caspase activity in 18-month-old female 3xTg-AD mice that had received five months of MitoQ treatment. Caspase-3/7 activity was measured in brain tissue using a luminescent caspase activity assay. Capase 3/7 activity was elevated about 3.5-fold in the 3xTg- AD brains as compared to the nonTg ones. Relative luminescent values demonstrated that MitoQ treatment blocked caspase activation by about 20% relative to 3xTg-AD mice (Fig. 7).

The median lifespan of 3xTg-AD mice is ~22 months with females living a bit longer than males (Rae and Brown, 2015). The median lifespan of nonTg mice with the same genetic background (129/C57BL/6) is ~34 months, also with females living a bit longer than males. Death of 3xTg- AD mice occurs as early as 12 months after birth. The common time frame of NFT development and the occurrence of mortality in the 3xTg-AD mouse model suggest that tau pathology may have a role in shortening the lifespan of these mice. As previously reported, we saw significant mortality in untreated 3xTg-AD mice by 12 months of age. MitoQ-treated mice had a lifespan that was indistinguishable from the nonTg mice within the 18-month period of the study (Fig. 8). These data show that, in addition to improved cognition, reduction of brain oxidative stress, and neuropathology, MitoQ extended the lifespan of this mouse model of AD.

4.Discussion
Evidence of oxidative stress is found in the brains of individuals who have died with AD and in the cerebrospinal fluid of patients with mild cognitive impairment, a condition that can be prodromal to AD (Nunomura et al., 2001, 2006; Praticò et al., 2002; Castellani et al., 2016). This stress precedes most of the neuropathological hallmarks of AD, suggesting that it may lie upstream from them. Mitochondrial dysfunction also occurs early in AD progression and is the likely source of the RS causing oxidative stress (Gutteridge, 1994; Pope et al., 2008; Moreira et al., 2010; Federico et al., 2012; Castellani et al., 2016). Oxidative stress and mitochondrial dysfunction are found not only in human AD but also in AD-like pathology in mouse models of the disease (Velliquette et al., 2005; Anantharaman et al., 2006; Butterfield et al., 2001, 2006, 2007; Resende et al., 2008; Yao et al., 2009; McManus et al., 2011). The appearance of oxidative stress in the development of both human and mouse models of AD and the success of antioxidants in treating AD-like symptoms in preclinical animal trials has led to a number of human antioxidant clinical trials for treating AD (Persson et al., 2014). These trials have largely been unsuccessful (Castellani et al., 2016). Possible reasons for these failures include poor ability of the antioxidants to cross the blood-brain barrier, inability to reach and/or enter mitochondria in sufficient concentrations, treatment so late in the disease that irreparable damage has occurred, or that oxidative stress is merely associated with AD and is not critical for its etiology in humans as it appears to be in animal models. Several AD trials have used N-acetyl-L-cysteine (L-NAC), a compound that does not target mitochondria but acts as an antioxidant in part by increasing levels of glutathione, a major cellular antioxidant. A concentration of about 10 mM L-NAC is needed to block neuronal degeneration caused by exposing cultures of cortical neurons to Aβ (Hsiao et al., 2008).

Similar concentrations of L- NAC are needed to block the apoptotic death of NGF-deprived sympathetic neurons in culture, a form apoptosis that appears to require increased mitochondrial production of ROS (Kirkland et al, 2001). MitoQ blocks this death at nM concentrations. A one nM concentration of MitoQ will block Aβ-induced death of cortical neurons in culture (McManus et al, 2011). This is a concentration seven orders of magnitude lower than the L-NAC concentrations required to do so. It is uncertain how much L-NAC can cross the human blood brain barrier and whether such concentrations can be reached in the brains of treated patients (Shahripour et al., 2014). Another antioxidant, vitamin E, has also had little success in several AD clinical trials (Persson et al., 2014). Like L-NAC, vitamin E does not target mitochondrial ROS production acting instead to target lipid peroxidation (Halliwell and Gutteridge, 2015) a form of oxidative damage lying downstream of mitochondrial ROS and that is largely independent of other ROS-induced damage such as DNA and protein oxidation. Problems with drug pharmacodynamics and pharmacokinetics are found with other antioxidants tested to date (Castellani et al., 2016). The development of mitochondria-targeted antioxidants, such as MitoQ, that cross the blood-brain barrier and concentrate in mitochondria, the principal source of most ROS, provide a better tool for determining whether mitochondria-associated oxidative stress is important for the disease and, perhaps, may lead to new and novel therapies for treating it (Zhao et al., 2004; James et al., 2005; Murphy and Smith, 2007; McManus et al., 2011).
We previously demonstrated the ability of MitoQ to inhibit cognitive decline and AD-like neuropathologies in young 3xTg-AD mice. That study focused on prevention of disease development (McManus et al., 2011). MitoQ treatment began two months after birth and continued for five months, a period during which the first AD-like pathologies become manifest.

The current study focused not on disease prevention but on therapy for disease that had already become patent. We evaluated the effects of MitoQ treatment on cognitive decline and neuropathologies in 3xTg-AD mice starting at 12 months after birth and continuing until 18 months of age. During this period, all of the known AD-like pathologies are present and are progressing in severity (Oddo et al., 2003). As in the younger 3xTg-AD mice, we found that MitoQ treatment of older mice was effective in improving spatial memory retention. Levels of the synaptic protein synaptophysin were significantly higher in the brains of MitoQ-treated mice than in the brains of untreated ones indicating that MitoQ treatment inhibited synapse loss and suggesting that this inhibition might underlie the observed cognitive improvement. Quantification of protein nitrotyrosine levels indicated that, at least by this measure, the brains of 18-month-old 3xTg-AD animals were under considerably greater oxidative stress than nonTg brains. MitoQ treatment greatly reduced nitration but not to the level found in nonTg animals. Overexpressing mitochondria superoxide dismutase in transgenic mouse models of AD decreases Aβ production and inhibits cognitive decline (Li et al., 2004; Esposito et al., 2006; Massad et al., 2009). Similarly, treating young 3xTg-AD mice with MitoQ prevents increased levels of Aβ in their brains (McManus et al., 2011). Aβ peptides are produced by sequential cleavage of APP by β- and γ-secretases. Oxidative stress can increase Aβ production by increasing β- and γ-secretase expression and activity via stress-activated kinases (Tong et al., 2005; Tamagno et al., 2002, 2005, 2008; Karupagounder et al., 2009; Zhang et al, 2011). Oxidative stress can also lead to modification of caspase activity (Circu and Aw, 2010), and activated caspases can cleave APP (Banwait et al., 2008; Bredesen et al., 2010; Mukherjee and Williams, 2017). Aβ has toxic effects on mitochondria (Lustbader et al., 2004; Sorrentino et al., 2018). Aβ can enter mitochondria via the TOM import machinery and induce mitochondrial dysfunction, at least in part, by inhibiting preprotein maturation (Petersen et al., 2008; Mossman et al., 2014). This inhibition imbalances the organelle proteome causing decreased respiration and increased ROS production. Therefore, increased Aβ in AD may involve a positive feedback cycle in which mitochondria-derived oxidative stress increases Aβ production, and the Aβ then increases ROS via its effects on mitochondria. Consistent with such a cycle, MitoQ decreased both oxidative stress and Aβ levels in the 18-month-old 3xTg-AD brains.

Increased numbers of astrocytes and microglial cells are found in both human AD brains and the brains of 3xTg-AD mice (Oddo et al., 2003; Janelsins et al., 2005; McManus et al., 2011; Hansen et al., 2018). It is uncertain as to whether these inflammatory processes contribute to AD or are downstream of the neuropathological processes. There is evidence that microglial cells are involved in the etiology of the disease, as well as evidence that they are protective against it (Hansen et al., 2018). The ability of MitoQ treatment to reduce Aβ levels, reactive astrogliosis, and microglial cell proliferation is consistent with the proliferation of both astrocytes and microglial cells lying downstream of the Aβ pathology. The amyloid hypothesis postulates that Aβ deposition lies upstream of NFT formation (Hardy and Allsop, 1991). Aβ deposition occurs earlier in human AD and the 3xTg-AD mouse model of AD than do NFT (Oddo et al., 2003; Selkoe and Hardy, 2016). Many studies have confirmed the interdependence of Aβ and tau pathology (Sato et al., 2018). Mutations in human APP that increase amyloid deposition also lead to increased levels of tau in neurons (Bateman et al., 2012; Selkoe and Hardy, 2016). Introduction of Aβ into neuronal cultures can cause hyperphosphorylation of tau, necessary for formation of NFT (Jin et al., 2011). Experimental treatments that reduce Aβ burden also reduce NFT in 3xTg-AD mice (Oddo et al., 2004; Rasool et al., 2013; Dai et al., 2017). In the present study, MitoQ treatment reduced Aβ accumulation, NFT formation, and both total amounts of tau and hyperphosphorylated tau in the 3xTg-AD brains. Cleavage of tau by caspases can lead to tau aggregation and phosphorylation (Rissman et al., 2004; Rohn et al., 2008). Thus, MitoQ decreased two events thought to lie upstream of tau pathology, Aβ accumulation, and caspase 3 activity.

The increased lifespan in the MitoQ-treated 3xTg-AD mice was intriguing. The reason that these animals die earlier than nonTg ones has not been reported and we did not do necropsies of dead animals. However, the most likely explanation for the extended survival was the reduction in neuropathologies associated with the disease. A similar extension of lifespan by MitoQ treatment of a transgenic C. elegans strain that expresses high levels of Aβ indicates that excess Aβ levels can have deleterious effects cross- phyla and that targeting mitochondria can help alleviate these effects (Ng et al., 2014). Further investigation of the means of death of these mice and the effect of MitoQ treatment on mortality will be needed to determine the mechanism. Comparing some of the parameters measured in our previous study of the effects of MitoQ on young animals to the effects of MitoQ on these parameters in the aged animals of this study revealed that MitoQ was highly effective in blocking progression of almost all AD-like symptoms in the young mice but merely inhibited their progression in the older animals (McManus et al., 2011). Synaptophysin (immunoblot band densities) in young 3xTg brains averaged about 60% of those from the nonTg brains. MitoQ treatment increased this to 100% of the nonTg level. The synaptophysin levels in the brains of the aged 3xTg-AD mice was lower than that in the brains of the younger animals with only about 30% of the synaptophysin level in the nonTg brains. MitoQ treatment increased this to about 60% of the nonTg brains. The 3xTg- AD nitrotyrosine levels in the brains of the young animals were about 3.8-fold higher than in the nonTg mouse brains as compared to about 2.8-fold higher in the brains of the older animals.

MitoQ treatment decreased the nitrotyrosine band density to nonTg levels in the earlier study while, in the present study, nitrotyrosine levels decreased by about 64% but remained about 1- fold higher than in the nonTg brains. The 3xTg-AD GFAP in the original study was about 2.5- fold higher than that of the nonTg brains. MitoQ treatment reduced this to the nonTg level. In the current study, GFAP was about 6-fold higher than in the nonTg brains. MitoQ treatment reduced this to about 2-fold higher than nonTg levels. As have others (Belfiore et al., 2018), we found that the young 3xTg-AD mice in the early study had relatively low brain Aβ levels. The Aβ concentration increased by several orders of magnitude by 18 months of age. In the earlier study, MitoQ decreased Aβ burden from about 6.5 pg/mg total protein to about 1.8 pg/mg (72% decrease) as compared to the nonTg level of about 0.5 pg/mg. Nanogram levels of Aβ were found in the brains of the 18-month-old animals. We found about 5 ng/mg protein in the brains of the untreated 3xTg mice as compared to about 0.4 ng/mg protein in the nonTg animals. MitoQ treatment reduced the concentration to about 2 ng/mg protein (60% decrease). MitoQ treatment was far more effective in reducing caspase activity in the brains of the young mice (to control levels of nonTg animals) than in the aged ones where activity was reduced only by about 31%. Thus, the trend of the effect of MitoQ treatment on the investigated parameters was similar in the old as compared to the young mice but the degree of the effect was blunted. However, even though not as effective at ameliorating pathology in the old animals as the young ones, the data demonstrate considerable therapeutic efficacy in them suggesting the possibility that MitoQ might be useful in treating AD after it has become patent.

In summary, our findings show that the mitochondria-targeted antioxidant MitoQ reduced cognitive and AD-like neuropathological symptoms in old 3xTg-AD mice. In humans, amyloid plaques can occur in the absence of cognitive deficits, and there is a poor correlation between the plaques and neuropathology (Maarouf et al., 2011; Wirth et al., 2013; Altman et al., 2015). Those findings, combined with the plethora of failed clinical trials targeting Aβ (Hardy and De Strooper, 2017), lends credence to the idea that these pathologies may contribute to AD but are perhaps not the singular cause of the disease. Evidence that oxidative stress and mitochondrial dysfunction precedes most of the major AD pathologies and likely contributes to them suggests that targeting them may be more effective in treating the disease than targeting Aβ or tau. This idea is further supported by recent work showing that enhancing mitochondria proteostasis in an AD mouse model reduces Aβ-associated proteotoxic stress (Sorrentino et al., 2018). We, and others, have shown that antioxidants targeted directly to mitochondria, the source of most RS, are effective in reducing AD pathology in animal models of AD (Szeto, 2006; McManus et al., 2011; Mao et al., 2012). It should also be noted that while 3xTg-AD mice exhibit many of the symptoms seen in AD patients, they do not at any age show loss of neurons in the brain, a prominent feature of human AD (Virgili et al., 2018). Therefore, the positive effects on 3xTg- AD cognition by MitoQ that we report here were likely due solely to synapse preservation. At least some of the failure of AD drug trials may result from the fact that significant, irreversible brain damage with neuron loss has already Mitoquinone occurred by the time the treatment started. Early detection and treatment before this has occurred is imperative. It would be of interest to determine whether MitoQ can prevent neuronal loss in an AD mouse model that does lose neurons, such as the 5xFAD model (Oakley et al., 2006).