Tau Protein Preferentially Associates With Synaptic Mitochondria in a Mouse Model of Tauopathy

A Accumulation of intracellular inclusions of abnormal brillar forms and hyperphosphorylated forms of microtubule-associated protein tau are hallmarks of AD and other tauopathies. Although tau pathology is associated with neuronal dysfunction the mechanism responsible remains obscure. In vitro, pathologically elevated expression of tau alters mitochondrial distribution by impairing cellular tracking and thus may represent an important mediator of mitochondrial abnormalities contributing to neuronal dysfunction. We used the transgenic htau mouse model of tauopathy to investigate in vivo alterations in brain mitochondria in the presence of pathological forms of human tau.

bioenergetics assays, proteomic analysis, and biochemical fractionation experiments in htau mice. These mice express human non-mutant tau (all six isoforms) in the absence of murine tau and suffer progressive tau aggregation and cognitive de cits in conjunction with accumulation of NFTs [38,42].
Here, in htau mice, elevated levels of tau protein were observed to preferentially associate with synaptic mitochondria compared to non-synaptic. Furthermore, there was a signi cant increase in synaptic mitochondrial associated tau protein at 8-months of age compared to 5-month old animals in htau mice only. This correlated with both age-dependent levels of tau phosphorylation and alterations in the bioenergetics of synaptic mitochondria. Additionally, carbonate extraction and protease digestion revealed mitochondrially associated tau is associated with, but not inserted into the mitrochondrial surface. We have identi ed the presence of multiple pathologic forms of tau associated with synaptic mitochondria, indicating a potentially direct role for tau in the impairment of synaptic homeostasis.

Experimental design
The aim of this study was to assess the molecular and functional characteristics of brain mitochondria of a mouse model of AD that carries the human tau gene. Brain mitochondria were assessed by bioenergetic assays, quantitative proteomics, and biochemical assessment. A total of 24 samples types were used (2 mouse genotypes * 2 mitochondrial populations * 2 age-points * 3 biological replicates).

Animals
Htau mice (B6.Cg-Mapt tm1(EGFP)Klt Tg(MAPT)8cPdav/J, Stock# 005491; Jackson Laboratory (JAX), (Bar Harbor, ME) express all six isoforms (including both 3R and 4R forms) of human tau in the absence of murine tau. The generation and characterization of the htau mice has been described previously [38]. All mice used in this study were male and aged to 5 or 8 months and for each experimental protocol, transgenic mice were compared to age-matched controls (C57BL/6J mice, Stock# 000664; JAX). Animals were fed standard mouse chow and water ad libitum and housed in a controlled environment under a 12hour light/dark schedule. Experiments were performed in accordance with National Institutes of Health guidelines under a protocol approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee.

Reagents
All chemicals and reagents were purchased from Sigma Aldrich (St. Louis, MO) unless noted below.

Antibodies
We used the following antibodies in this study: Tau46

Isolation of brain mitochondria
The mice were sacri ced by cervical dislocation and the brain was immediately removed, then rinsed with cold mitochondrial isolation buffer (MSHE): 70 mM sucrose, 210 mM mannitol, 5 mM HEPES, 1 mM EGTA and 0.5% (w/v) fatty acid free BSA (pH 7.2). The protocol was carried out on ice and all subsequent spins were done at 4°C. The brain was homogenized for 10 strokes with a Dounce homogenizer in MSHE containing cOmplete Mini, EDTA-free protease inhibitor cocktail (Roche Diagnostics, Indianapolis, IN). The homogenate was then spun at 1300 x g for 3 min and supernatant was collected. Following a second wash and spin at 1300 x g for 3 min, the supernatants were pooled and spun at 21,000 x g for 10 min. Synaptic and non-synaptic mitochondria were isolated using Percoll density gradient centrifugation as described previously [43]. Following isolation, pelleted mitochondria were resuspended in 1x mitochondrial assay solution (1x MAS; 70 mM sucrose, 220 mM mannitol, 10 mM KH 2 PO 4 , 5 mM MgCl 2 , 2 mM HEPES, 1 mM EGTA, pH 7.2). The isolated mitochondrial protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scienti c, Waltham, MA) and used immediately for bioenergetic analysis or lysed in 4% SDS/DTT and stored at -80°C. Protein quanti cation prior to protein digestion for mass spectrometry was performed using a Pierce 660 nm Protein Assay with BSA standards with the addition of the ionic detergent compatibility reagent (IDCR, Thermo Fisher Scienti c).

Bioenergetic analysis of isolated brain mitochondria
The isolated synaptic and non-synaptic mitochondria were assessed functionally using the Seahorse XF24 analyzer based on the protocol of Rogers [44] with minor alterations as described previously [45]. For the Seahorse experiments, synaptic and non-synaptic mitochondria isolated from three mice per strain (WT and htau) at each of the two ages (5 and 8 months) were utilized for the coupling assay. Each biological replicate (n = 3) had three to four technical replicate wells for the experiment.
Mass spectrometry-based proteomics

Protein digestion
Mouse synaptic and non-synaptic mitochondrial protein sample aliquots (35 µg) used for dataindependent acquisition mass spectrometry were digested with trypsin using lter aided sample preparation [46]. The resultant peptides were desalted using Oasis mixed-mode weak cation exchanges cartridges (Waters, Milford, VA), dehydrated with a Savant ISS 110 SpeedVac concentrator (Thermo Fisher Scienti c) and resuspended in 10 µL of 0.1% formic acid prior to quanti cation using a NanoDrop 2000 UV-vis spectrophotometer (Thermo Fisher Scienti c) in conjunction with the Scopes method for peptide quanti cation by absorbance at 205 nm [47].

SWATH-MS analysis
The samples of peptides (2 µg) from WT and htau mouse synaptic and non-synaptic mitochondrial lysates were analyzed in triplicate (three biological replicates per strain (WT and htau, n = 3) by nano-LC-MS/MS in SWATH-MS mode on the 5600 TripleTOF instrument (SCIEX) and targeted data extraction was performed as previously described [43,45]. All fragment ion chromatograms were extracted and automatically integrated with PeakView software (Version 2.1, SCIEX, Framingham, MA). For peptide identi cation, our published reference spectral library was used [43,45]. This library was generated in ProteinPilot (Version 4.5, SCIEX) using the Paragon algorithm and the default settings. All searches were performed against the UniProt Mus Musculus Proteome UP000000589 containing 17,048 reviewed proteins (Swiss-Prot). Combined results yielded a library of 4,227 proteins identi ed with high con dence (greater than 99%) that passes the global FDR from t analysis using a critical FDR of 1%. In accordance with previously published work [43,45], we selected ve peptides and ve transitions option for quantitative analysis and performed targeted data extraction for each peptide. For each peptide, the fragment ion chromatograms were extracted using the SWATH isolation window set to a width of 10 min and 50 ppm accuracy [43,45]. To calibrate retention times, synthetic peptides (Biognosys AG, Schlieren, Switzerland) were spiked-in to the samples in accordance with the manufacturer's protocol. Data were normalized to the median peak ratios of common proteins in MarkerView software (Version 1.2.1, SCIEX).

Bioinformatics
The Cyber-T Web server (http://cybert.ics.uci.edu/) [48], which implements a t-test using a Bayesian regularization method was used to assess statistical signi cance [49]. Proteins deemed as differentially expressed were p-values < 0.05 from the pairwise post-hoc test (TukeyHSD) that also exhibited Benjamini and Hochberg (BH) q-values < 0.05 to correct for multiple testing. The Search Tool for the Retrieval of Interacting Genes/Proteins (STRING [50]; https://string-db.org) was used to gain further insight into the role of the differentially expressed proteins via network clustering and functional annotation. Log 2 -fold data obtained from the SWATH-MS analysis was clustered heirarchically between genotypes and withing genotypes between ages, using a Multiple Experiment Viewer (MEV, http://mev.tm4.org/, [51]) using the HCL method with a Eucleidian distance metric and complete linkage. Heatmaps were generated in GraphPad 9.

Western blot analysis
Biochemical isolation of synaptic mitochondrial isolates For mitochondrial membrane protein analysis 60 µg of percoll isolated mitochondria were resuspended in 1 mL of 0.1 M Na 2 CO 3 , pH 11.5, and incubated on ice for 30 min. Following the incubation period samples were subjected to ultra-centrifugation at 100,000 x g for 20 min. The insoluble membrane fraction (pellet) solubilized by sonication and boiling at 95°C for 5 min in 30 µL solubilization buffer (SB; 100 mM Tris-HCl pH 7.4, 100mM DTT, 4% w/v SDS). The soluble protein fraction (supernatant) was concentrated by centrifugation with an Omega NanoSep 30K NWCO spin-lter (Pall, Port Washington, NY) at 14,000 x g until near dryness. Proteins were collected from spin-lters by pipette after the addition of 30 µL SB. The soluble protein fraction was then boiled at 95°C 5 min and sonicated brie y. Soluble and insoluble fractions were quanti ed using the Pierce 660 nm Protein Assay (Pierce) with the IDCR (Thermo Fisher Scienti c). Equal masses of soluble and insoluble were analyzed by Western Blot.

Protease treatment of isolated mitochondria
Synaptic and non-synaptic mitochondria from 8 month old male htau mice were isolated by Percoll gradient as described above and nal pellets were resuspended in 200 µL of 1x MAS and protein content was determined by Pierce 660 nm protein assay. 20 µg of total mitochondria were plated in each of 7 wells in a round bottom 96 well plate on ice in a total volume of 15 µL. Protease digests were started by the addition 15 µL of Trypsin diluted in 1x MAS to yield the nal working concentration indicated in Fig. 5.
Plates were incubated on ice for 30 min and mixed by gentle pipetting every 10 min. Once 30 min had passed protease action was halted by the addition of 2 µL of 200 mM phenylmethylsulfonyl uoride, followed by gently pipette mixing and an additional incubation of 10 min on ice. SDS-Page samples were prepared by the addition of 10 µL of 4x sample buffer (Licor, Lincoln, NE), transfer to PCR strips tubes and boiling for 5 min st 95°C. Equal volumes of each sample were then resolved by SDS-Page, transferred to nitrocellulose and immunoblotted as described below with the indicated primary antibodies. Quanti cation was carried out by densitometric measurements made in Light Studio (Licor).

Immunoblot analysis
Isolated mitochondria were lysed in 4% SDS and protein quanti cation was performed using a Pierce 660 nm Protein Assay with BSA standards with the addition of the IDCR (Thermo Fisher Scienti c). 10 µg of lysate was resolved on Nu-PAGE Bolt 4-12% gradient polyacrylamide gels gels using the MES/SDS buffer system (Life Technologies, Carlsbad, CA), transferred to nitrocellulose using an iBlot2 instrument (Invitrogen, Carlsbad, CA). Membranes were blocked with TBS/SuperBlock (Thermo Fisher Scienti c) for 30 min at room temperature, and then probed with the indicated antibodies at appropriate dilutions (see Antibodies section) in Tris buffered saline with 0.1% Tween-20 (TBS-T)/SuperBlock and incubated overnight at 4°C. Blots were washed 3 x 10 min with 1x TBS-T and then incubated with appropriate secondary antibodies (Licor) for 1 hour at room temp. Membranes were again washed 3 x 10 min with 1x TBS-T and imaged using an Odyssey imager (Licor) using appropriate channels. Quanti cation of immunoreactivity was achieved using Image Studio software (Licor).

Statistical analysis
All experiments were conducted with a minimum of 3 biological replicates with or without 3-4 technical replicates (Seahorse only), and statistical analysis was performed using built-in analysis functions in Prism 9 (GraphPad, San Diego, CA). Statistical analysis of the SWATH analysis was determined by Cyber-T. The statistical tests performed are indicated in the gure legends.

Results
Synaptic mitochondria isolated from mice expressing non-mutant human tau isoforms exhibit ageassociated bioenergetic alterations To explore whether alterations in mitochondrial bioenergetics are involved in tauopathy pathogenesis, non-synaptic and synaptic mitochondria were isolated from htau transgenic and WT mice at 5-and 8months of age and oxygen consumption rates (OCR) driven by complex II were measured utilizing the coupling assay [44]. Non-synaptic mitochondria from htau mice did not show alterations in respiration compared to WT mice at either age examined (Fig. 1A). Synaptic mitochondria isolated from htau mice at 8-(but not 5-) months of age exhibit a signi cant increase in the rate of complex II driven state 2 (basal), state 3 (ADP-stimulated), and state 3u (maximum uncoupled) respiration compared to those from agematched WT mice (Fig. 1B). Tau pathology has been described in the brain of htau mice at 8-months [38], thus, mitochondria from the synapse are particularly sensitive to pathologic tau accumulation, and exhibit functional changes reported during aging [45].

Quantitative mitochondrial proteomics reveals accumulation of tau in association with synaptic mitochondria
To explore the molecular mechanisms underlying altered synaptic mitochondrial respiration in the absence of respiratory changes in non-synaptic mitochondria, we investigated the in uence of human tau expression in mice on brain mitochondria protein expression pro les using the quantitative mass spectrometry-based technique SWATH-MS [52]. Synaptic and non-synaptic mitochondria were isolated from 5-and 8-month-old htau transgenic and WT mice, and the proteome was analyzed. In total, 1,578 proteins were identi ed and the complete list of these proteins with quantitative values is provided in Additional File 1: Supplementary Table S1. SWATH sample replicates exhibited a high degree of correlation, suggesting there was minimal sample-to-sample variablity within groups (Additional File 2: Supplemental Figs. 1-4). To uncover which proteins were differentially expressed (DE; FDR q < 0.05) in mitochondria isolated from htau as compared to age-matched WT mice, we used a Bayesian regularized t-test analysis and multiple testing corrections, which revealed 20 (5-months) and 12 (8-months) DE proteins in non-synaptic mitochondria samples, and 54 (5-months) and 31 (8-months) DE proteins in synaptic mitochondria samples (Additional File 1: Supplementary Table S2A-D). As shown in Fig. 2A, only 1 protein (Dephospho-CoA kinase domain containing protein (Dcakd)) was found to be DE in both nonsynaptic and synaptic mitochondria isolated from 5-month old htau mice (as compared to age-matched WT mice), whereas 2 proteins (Mapt (tau) and tubulin alpha-4 chain (Tuba4a)) were found to be DE in synaptic mitochondria isolated from 5-and 8-month old htau mice (as compared to age-matched WT mice). Dcakd was found to be elevated in isolated non-synaptic and synaptic mitochondria from htau mice as compared to WT mice at 5-months (Fig. 2B). However, at 8-months, the levels of Dcakd are similar in isolated mitochondria between htau and WT mice. While Mapt (tau) levels are increased in isolated synaptic mitochondria from htau mice as compared to WT mice at both ages (Fig. 2C), Tuba4a is decreased at 5-months and increased at 8-months (Fig. 2D). Of note, Mapt (tau) levels are increased in isolated mitochondria from 8-as compared to 5-month-old htau mice, highlighting an age-dependent increase.
To uncover protein-protein interaction networks and perform enrichment analysis to gain insight into functional associations of the mitochondrial protein changes, the lists of DE proteins in htau versus agematched WT mice for each mitochondrial population (non-synaptic and synaptic mitochondria) were uploaded to the STRING database. Only interactions which were of high con dence (minimum required interaction score of 0.07) were used to generate the interaction networks using the MCL clustering method (network clusters are denoted by node color). This analysis identi ed 8 interacting proteins organized in 4 networks for the 5-month non-synaptic mitochondria DE proteins (Fig. 3A), 2 proteins organized in 1 network for the 8-month non-synaptic mitochondria DE proteins (Fig. 3B), 29 interacting proteins organized in 7 networks with 9 distinct network clusters for the 5-month synaptic mitochondria DE proteins (Fig. 3C), and 13 interacting proteins organized in 3 networks for the 8-month synaptic mitochondria DE proteins (Fig. 3D). While the networks for the non-synaptic mitochondria DE proteins were composed of single interactions between 2 proteins, the networks for the synaptic mitochondria DE proteins were more complex, evidenced by the increased numbers of nodes and edges. Functional enrichments in the STRING networks revealed gene ontology (GO) biological process (BP) terms that were altered according to our proteomics results (Additional File 1: Supplementary Table S3A-C). While no BP terms were identi ed for the 5-month non-synaptic mitochondria, one term was found for the DE proteins in non-synaptic mitochondria from 8-month htau compared to WT mice were "positive regulation of hydrolase activity" (FDR = 0.0078). In contrast, several BP terms (131 at 5-months and 9 at 8-months) were signi cantly enriched based on the DE proteins in synaptic mitochondria from htau as compared to WT mice, with the top two terms "synaptic vesicle docking" and "glycerol-3-phosphate metabolic process" at 5-months, and "microtubule-based process" and "cytoskeleton organization" at 8-months. Of note, Mapt (tau) was a node in the STRING networks for synaptic mitochondria DE proteins at both ages studied.

Human tau expression alters expression of electron transport chain components
To determine whether proteomic changes affecting the electron transport chain (ETC) and oxidative phosphorylation (OxPhos) correlate with the observed respiratory alterations, we assessed the expression of the ETC component proteins as well as regulators of complex assembly and function (Fig. 4). We used the web served based MEV to cluster log 2 -fold expression data comparing synaptic and non-synaptic data between htau and WT mice at 5-and 8-months of age (Fig. 4A). Clustering indicated that expression changes in ETC components were most similar between synaptic and non-synaptic mitochondria a 5months. Furthermore, changes 8-month synaptic mitochdonria were the most divergent between genotypes, a nding that correlates well with the changes we observed in bioeneregetics of 8-month synaptic mitochondria. Interestingly, the age-dependent alterations of mitochondrial respiratory complex subunit expression in synaptic mitochondria from WT mice were similar to changes observed in htau transgenic mice consistent with the observation of similar synaptic mitochondrial functional alterations with age in htau transgenic and WT mice. However, despite the predominant conservation of age-dependent changes in the synaptic mitochondrial ETC subunits and regulator proteins in htau mice, differences exist which may contribute to the functional differences between htau and WT mice at 8months of age. Therefore, we focused on the synaptic mitochondrial proteomic changes between htau and WT mice at 8-months of age to gain mechanistic insight into the contribution of tau pathology to alterations in synaptic mitochondrial function. Upon examination of the differences in the expression of the ETC subunits between 8-month-old htau and WT mice it becomes apparent that several of the mitochondrial DNA (mtDNA) encoded subunits (Mtnd4 (log2 = -1.43), Mtnd5 (log2 = 1.10), and Mtco1 (log2 = -0.176)) exhibit altered expression (Additional File 1: Supplementary Table 1, Fig. 4A (red row  names)). Although not reaching signi cance in our proteomic analysis due to the stringency of our statistical analysis, lower Mtco1 levels were con rmed via immunoblotting (Fig. 4B). Similar to Mtco1, our proteomic analysis revealed a slight reduction in Uqcrc2 expression in synaptic mitochondria from 8month-old htau compared to WT mice (log2 = -0.256), which was con rmed by immunobot quanit cation ( Supplementary Table 1; Fig. 4). A similar observation was made for Atp5a1 (log2 = -0.44). Since htau mice exhibit hyperphosphorylation and aggregation of tau within the span from 5-to 8-months of age [38], we next sought to characterize tau associated with synaptic mitochondria as our functional and proteomic results suggested this popoulation of mitochondria was the most divergent.

Accumulation of pathologic tau in association with synaptic mitochondria
Phosphorylation of tau at certain serine and threonine residues serves as a biochemical marker for pathologic alterations in AD and fronto-temporal dementia (FTD) [2,10,26,36,38]. We therefore sought to determine whether alterations in synaptic mitochondrial bioenergetics were associated with this hallmark of disease progression. We used immunoblotting to determine if tau was present in puri ed synaptic mitochondrial isolates (Fig. 5). In agreeance with our SWATH analysis (Fig. 2) we observed a signi cant increase in the amount of total tau protein present in htau mice at both 5-and 8-months of age compared with WT mice, which was more pronounced at 8 compared to 5 months of age (Fig. 5A). We also detected CP13 [8,9,53] and PHF1 [54,55] positive immunoreactivity in isolated synaptic mitochondria, which speci cally detect abnormal phosphorylated residues S202 (Fig. 5B) and S396/404 (Fig. 5C), respectively. In synaptic mitochondria we observed low but comparable levels of CP13 immunoreactivity between 5-month-old WT and htau mice, however, at 8-months we detected a signi cant increase of CP13 signal in synaptic mitochondria from htau mice (Fig. 5B). In contrast to CP13, we detected a signi cant increase in PHF1 signal at both 5-and 8-months in synaptic mitochondria from htau mice as compared to WT, and this exaggerated at 8-months (Fig. 5C). In addition to pathologic hyperphosphorylation, tau also changes its conformation during disease progression [56], which nally culminates in aggregation and formation of tangles. These early conformational changes can be monitored by probing with the antibody MC1, which is speci c for a pathologic conformation of tau as it recognizes neuro brillary tangles (NFTs) [56]. In synaptic mitochondria from 5-month-old htau mice there was a trending albeit insigni cant increase in MC1 signal that achieved signi cance at 8-months (Fig. 5D). Interestingly, a signi cant increase in the detection of PFH1 tau was observed at both 5-and 8months timepoints, suggesting increased phosphorylation at S396/404 may be an earlier event in the progression of AD and related tauopathies (Fig. 5C).

Tau localizes to the mitchondrial outer membrane
Other studies have reported the association of tau protein with mitochondria in a number of systems [11,13,28,36,[57][58][59]. This association was determined in some cases to be the result of tau incorporation into the outer mitochondrial membrane, yet in others the consequence of protein-protein interactions with outer mitochondrial membrane proteins, one such protein is voltage dependent anion channel 1 (VDAC1) [60][61][62][63]. With this in mind we sought to determine the mechanism by which tau was co-purifying with synaptic mitochondria in our isolations. To address this, we used sodium carbonate extraction, an established method to differentiate integral mitochondrial membrane proteins from membrane associated mitochondrial proteins. In this assay integral membrane proteins fractionate into an insoluble (pellet; "P") fraction, while peripheral membrane (i.e. membrane associated) proteins segregate into the soluble ("S") fraction. Synaptic mitochondria isolated from 8-month-old htau mice were extracted with sodium carbonate and the resulting fractions were probed by western blot (Fig. 6). As control we probed for the OxPhos panel (Fig. 6A, red) as well as heat-shock protein 60 (HSP60) (Fig. 6A, green). With the exception of ATP synthase F1 subunit alpha (Atp5a1), we primarily detected the proteins of the OxPhos panel in the pellet as expected since most are integral membrane proteins, also as expected we observed the majority of HSP60 in the soluble fraction. Finally, we observed tau protein segregating almost entirely into the soluble fraction (Fig. 6B). Our results suggest that tau associated with synaptic mitochondria is not inserted into the mitochondrial membrane, but likely associated via protein-protein interaction.
To further support our sodium carbonate extraction results, we employed a protease digest assay to determine the rate at which tau was degraded compared to two mitochondrial proteins. We reasoned that if tau was associated with mitochondria via protein-protein interactions it would be degraded at lower concetrations of protease than either integral membrane or mitochondrial matrix proteins. Isolated synaptic and non-synpatic mitochondria from approximately 8-month-old htau mice were digested by trypsin and immunoblotted for tau, VDAC, and succinate dehydrogenase complex subunit A (SDHA) (Fig. 6C). We observed a reduced percentage of remaining full length tau protein at lower trypsin concentration than either SDHA or VDAC, suggesting mitochondrially associated tau is more accessible to proteolytic action. Taken together with our sodium carbonate extraction results, this suggests that tau is present on the mitochondrial outer membrane and this interaction is most likely mediated by proteinprotein interactions.

Tau impairs synaptic bioenergetics and alters electron transport chain subunit expression
We compared the bioenergetic pro les of synaptic and non-synaptic mitochondria and observed an ageand pathological tau-dependent shift in the bioenergetics of synaptic mitochondria only. Speci cally, we determined a signi cant elevation in both basal and maximal respiration in synaptic mitochondria from aged htau mice that was not observed in the young or aged WT controls, nor young htau cohorts. Interestingly, despite elevated basal and maximal respiratory capacity the spare respiratory capacity (SRC) in these mitochondria was impaired. These changes are characteristic of an aged synaptic mitochondrial bioenergetic phenotype, which we previously reported in WT mice [45]. Further, we observed similar age-dependent complex II changes in the Pink1 KO rat model of Parkinson's disease (PD) [64], which in striatal synaptic mitochondria was accompanied by decreased complex I respiration. Studies in other systems have reported similar ndings. For example, Flynn et al [65], observed an impaired SRC in synaptosomes isolated from super-oxide dismutase 2 (SOD2) null mice, which have been reported to mimic tauopathies by accumulation of amyloid and tau deposits [66][67][68][69][70]. Another study reported that genetic ablation of tau expression improved hippocampal mitochondrial function despite reducing expression of proteins of the ETC [71]. Furthermore, studies investigating pathological mutant forms of tau have similarly reported alterations is bioenergetic. One such study by David et al, demonstrated expression if P301L mutant tau in mice resulted in impaired respiratory control ratio and ATP production from cerebral mitochondrial isolates [72]. Notably, the authors also reported increased reactive oxygen species (ROS) production and SOD activity supporting the notion that increased oxidative states promote tauopathy [72]. Several studies have reported increased ROS production as a consequence of complex II driven respiration [73][74][75]. One such study of isolated mitochondria from rat skeletal muscle reported high rates of H 2 O 2 production driven by complex II when complex I and III were inhibited [73].
Furthermore, inhibition of complex II by either malonate or malate completely abrogated H 2 O 2 production [73]. A similar observation was reported in rat cardiac mitochondria [74]. In our assessment of bioenergetics, we observed elevated complex II driven respiration in the synaptic mitochondria from 8month htau mice compared to WT, suggesting that an increase in ROS may similarly contribute to pathology in our system. Indeed elevated ROS has also been reported in models of AD [68,[76][77][78], as well as samples from AD patients [79,80].
In line with our observations that the bioenergetics of synapses (i.e. synaptic mitochondria) are altered by the accumulation of pathological forms of tau protein we observed changes in the expression of components of the ETC including Mtco1. Interestingly, other reports have suggested that reduction of Mtco1 expression is positively selected for as cells trend towards mitochondrial homoplasmy [81].
Notably, decreased expression of Mtco1 correlates with reduced generation of ROS and improved cell survival via impaired apoptosis [81]. Taken together this may suggest a potential mechanism by which early tau accumulation serves a neuroprotective role by reduction of Mtco1 expression. In addition to alterations in Mtco1 expression we observed a decrease in Mtnd4 expression. Mtnd4 is principle component of the minimal assembly core for mitochondrial complex I, and is critical for proper ETC function [82]. Importantly, impaired complex I function has been been reported in several neurodegenerative disorders including PD [83,84] and altered expression of complex I subunit proteins has been reported in AD [85,86]. Recently, a study investigating global proteomic changes reported altered expression of a number of mitochondrial proteins in AD patient brain regions in early and late stages of disease [86]. Notably, expression of several mitochondrial proteins was observed to decrease, speci cally in the entorhinal and parahippocampal cortices, two regions known to be affected by AD [86][87][88]. Furthermore, the study reported a decrease in the expression of Mtnd4 and Mtnd1 in late stage AD [86]. Although, we did not observe a signi cant difference in the expression of Mtnd1, we did observe a signi cant decrease in expression of Mtnd4. Possible explanations for this difference could be the larger group size in the study by Mendonça et al., or intrinsic differences between humans and mice neurophysiology. Another likely contributor to the differences between our study and that of Mendonça is our use of rather strict statistical analysis, which results in increased con dence of our results, could also result in Type II error (not achieving statistical signi cance despite being differentially expressed). Of note our analysis did reveal a trending, albeit not signi cant, decrease in the expression of Mtnd1 in isolates of synaptic mitochondria at 8-months between WT and htau mice. Nevertheless, the congurencies in our analysis and that of Mendonça lends support to the validity of both our study as well as use of the htau model in studying the pathogenesis of AD.
Pathological forms of tau accumulate with synaptic mitochondria We observed signi cant increases in the accumulation of total tau protein as well as pathological forms of the protein (including hyperphosphorylated and NFTs). A number of studies have reported the accumulation of these species in the brains of AD patients and models of AD and related tauopathies [2,8,27,38,54,89]. One study from 2008, reported the accumulation of tau protein in synaptosomes isolated from brains perimortem, where it colocalized with amyloid [90]. Furthermore, using immuno uorescent imaging hyper-phosphorylated tau was observed to primarily in neurites of intact neurons [90]. In another study of postmortem dementia patient samples, synaptic accumulation of both total and hyper-phosphorylated tau was reported in conjunction with ubiquitin-proteosome dysfunction [91]. Indeed, other studies have demonstrated the accumulation of tau impaired mitophagy by protecting the mitochondrial membrane potential to prevent activation of the Pink1/Parkin pathway [57] or by sequestration of cytosolic Parkin preventing its recruitment to damaged mitochondria [92]. Although they did not investigate synaptic accumulation of tau speci cally, Cook et al injected neonatal mice with AAV-Tau P301L on day P0 and observed signi cant widespread accumulation of tau protein and NFTs in these mice by 6 months of age [93]. Tau accumulation in these animals correlated with increased neuroin ammation and cognitive de cits [93]. Conversely, Sahara et al, reported no preferential synaptic accumulation of tau P301L in transgenic mice [94]. They did however note, that tau protein contained within the synaptosomal fraction appeared to be membrane associated [94]. Furthermore, in contrast to our ndings, the author's observed a decrease in the relative amount of tau phosphorylation in synaptic membrane fractions [94]. A study in Drosophila reported tau localization and accumulation in presynaptic terminals and similarly to Sahara et al., it was demonstrated to associate with the pre-synaptic membrane [95]. Our results suggest there may exist an alternate mechanism by which tau accumulates in the pre-synaptic terminal by preferentially associating with mitochondrial membranes. Future studies that focus on determining mechanistically how tau preferentially associates with synaptic mitochondria will be of signi cant interest.

Conclusions
In the current study we have reported that pathological forms of tau protein accumulate and associate with synaptic mitochondria and impair synaptic bioenergetics. Our study highlights the importance of techniques used to separate unique populations of cells or organelles. While our results indicate tau accumulation is more detrimental to the function of synaptic mitochondria, the precise molecular mechanisms that govern preferential associations between tau and synaptic mitochondria remain elusive, and thus should be the subject of detailed future investigations.

Availability of data and materials
The datasets generated and analyzed during the current study are available from the corresponding author upon request.