Mitochondrial ROS control neuronal excitability and cell fate in frontotemporal dementia

The second most common form of early‐onset dementia—frontotemporal dementia (FTD)—is often characterized by the aggregation of the microtubule‐associated protein tau. Here we studied the mechanism of tau‐induced neuronal dysfunction in neurons with the FTD‐related 10+16 MAPT mutation.


RESEARCH IN CONTEXT
1. Systematic review: Tau is involved in a number of neurodegenerative disorders such as frontotemporal dementia. Different mechanisms have been described to understand its role in neurodegeneration, including synaptic dysfunction, mitochondrial alterations, oxidative stress, and calcium deregulation. Although neurodegeneration is a hallmark of dementia, memory decline appears prior to neuronal loss in a number of neurodegenerative disorders, including FTD. Synaptic plasticity is thought to represent a key mechanism underpinning learning and memory. Indeed, synaptic dysfunction is the best correlate with the progression of cognitive decline in tauopathies such as Alzheimer's disease (AD). 9 Glutamate is the main excitatory neurotransmitter in the central nervous system acting through either ionotropic glutamate receptors (plasmalemmal ion channels such an N-methyl-D-aspartate [NMDA] and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid [AMPA] receptors that mediate Ca 2+ and Na + influx) or metabotropic glutamate receptors (mGlu, protein G-coupled receptors that mediate Ca 2+ release from intracellular stores, among other signaling pathways). However, excessive activation of glutamate receptors prompts hyperexcitability in neurons, resulting in intracellular calcium overload that triggers a pathological process known as excitotoxicity, which ultimately leads to neuronal death. [10][11][12] Impaired glutamatergic transmission has been previously demostrated in tauopathies, pointing to tau-induced dysfunction of NMDA and AMPA receptors. 13,14
• Tau-induced mitochondrial ROS alter the trafficking of specific glutamate receptors.
• Mitochondrial antioxidants prevent calcium overload and excitotoxicity in FTD.
• Extracellular 4R tau impairs glutamatergic signaling by the same mechanism.
Reactive oxygen species (ROS) are able to modulate physiological and pathological signal transduction in the brain. NMDA glutamate receptor activation can trigger superoxide production by nicotinamide adenine dinucleotide phosphate (NADPH) oxidase, contributing to cell signaling, but also to neuronal damage. 15,16 Conversely, the function of NMDA receptors can also be modified by ROS, impairing synaptic function in AD. 17,18 Specifically, ROS produced by the mitochondria regulate diverse physiological processes in brain cells, including signal transduction, [19][20][21] but overproduction of mitochondrial ROS can trigger cellular dysfunction and neuronal death. 22 Indeed, our previous studies in induced pluripotent stem cell (iPSC)-derived neurons with the FTD-related 10+16 MAPT mutation show that mitochondrial ROS overproduction is a key pathological event of tau-induced pathology. 8 Mitochondrial dysfunction in these cells leads to oxidative stress and neuronal death that can be prevented with mitochondrial antioxidants. In addition, 10+16 neurons are more vulnerable to calcium overload 5,23 and exhibit severe functional impairments including a depolarized resting membrane potential and changed neuronal excitability due to reduced Na v 1.6 expression. 24 Here, we have tried to understand how this system of impairments interact in the course of neurodegeneration, by exploring a possible link between them and the dysregulation of the glutamatergic signaling. We have found that overproduction of ROS in the mitochondria of the 10+16 neurons leads to an increase in the surface levels of specific subunits of NMDA and AMPA receptors via protein oxidation.
This impairs the glutamate-induced signal transduction leading to calcium overload. Supplementation of the cells with mitochondrial antioxidants completely recovers the glutamate-induced calcium response in patients' neurons and protects against excitotoxicity.
Similar results were obtained in isogenic-engineered 10+16 MAPT iPSC-derived neurons and in primary neurons treated with extracellular 4R tau. Our results highlight a direct link among mitochondrial dysfunction, oxidative stress, and calcium deregulation in the mechanism of 4R tau-induced neuronal death, which is not restricted to FTD but can be extrapolated to other forms of dementia. These findings demonstrate a key role for mitochondria in pathophysiological signaling and the possibility to modulate its effects with mitochondrial antioxidants.

Materials
Unless otherwise specified, all the materials were obtained from Thermo Fisher Scientific (Life Technologies). NMDA, AMPA, MitoTEMPO, Trolox, and MK801 were obtained from Sigma-Aldrich.
Murphy (Medical Research Council Mitochondrial Biology Unit).

Mitochondrial membrane potential
Mitochondrial membrane potential (ΔΨm) was analyzed as previously described. 31 Cells were loaded for 40 minutes with 25 nM tetramethylrhodamine methyl ester (TMRM, Sigma). Z-stacks were acquired using a Zeiss 710 VIS CLMS confocal microscope equipped with a META detection system and an x40 oil immersion objective (Zeiss). The dye was excited at 561 nm, and the emitted fluorescence was detected above 580 nm. Z-stacks were analyzed and average intensity was calculated using Volocity 3D Image Analysis Software (PerkinElmer).

Cell death
Cells were loaded for 20 minutes at room temperature with 20 μM pro-

Cross-linking assay
To estimate the trafficking of the different AMPA receptor (AMPAR) and NMDA receptor (NMDAR) subunits between the cytosol and the cellular membrane we performed a crosslinking assay with the membrane impermeable crosslinker bis(sulfosuccinimidyl)suberate BS3 (ThermoFisher) as described by Boudreau et al. 32 Briefly, BS3 forms a covalent bond between the cell surface proteins in close proximity leaving intracellular proteins unaffected. Therefore, the apparent molecular weight of the receptors located in the cell membrane increases compared to the non-crosslinked (intracellular) ones, and it is possible to distinguish and quantify each fraction using western blot. A sample not treated with BS3 was included to confirm the position of the intracellular (non cross-linked) band. The absence of higher molecular weight bands in the actin blot serves as a technical control indicating that intracellular proteins were not crosslinked.
When indicated, 100 nM MitoTEMPO was added to the cell media 2 hours before the experiment. Neurons were then washed, the crosslinking reaction was performed incubating them in 0.5 mM BS3 in HBSS for 15 minutes at 37 • C and terminated by adding 100 mM glycine (10 minutes at 4 • C). Solution was washed and protein extracts collected in ice-cold RIPA lysis buffer supplemented with protease and phosphatase inhibitors (Thermo Fisher). Samples were snap frozen, sonicated, and centrifuged at 14000 rpm; and protein content of the extracts was determined by the Pierce BCA protein assay (Thermo Fisher).

Western blot
Fifteen to twenty μg of protein extracts were then fractionated on a sodium dodecyl sulfate polyacrylamide gel (4%-12%; Thermo Fisher), transferred to a polyvinylidene difluoride membrane (Bio-Rad) and blocked with 5% non-fat milk. Membranes were incubated overnight with the corresponding primary antibodies diluted in 5% bovine albu-

Redox proteomics
For redox proteomics, protein extracts were prepared from cell cul- The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE 34 partner repository with the dataset identifier PXD025083. Neurons were monitored for spontaneous firing activity in either cellattached configuration (after formation of gigaseal) or whole-cell (at -60 mV). Once after membrane breakthrough (whole-cell), cells were monitored for the intrinsic passive membrane properties, including the resting membrane potential (V rest ), capacitance (C m ), and input resistance (R in ), as described by Kopach et al. 24 Synaptic activity (excitatory transmission) was examined by recording spontaneous excitatory postsynaptic currents (sEPSCs) at -70 mV. The glutamate-evoked currents were recorded at different membrane potential by applying exogenous glutamate using the rapid-application system. An application pipette was positioned close to a recorded neuron, and glutamate (100 μM in aCSF) was applied in a series of 500-ms pulses (or as indicated), in varied inter-pulse intervals, with a pressure supplied by a two-channel PDES-02DX pneumatic micro ejector system (npi electronic GmbH).

Electrophysiology
Glutamate-evoked currents were analyzed for the peak amplitude calculated as mean amplitude for train of repetitively evoked responses (at least 3 to 5 trials), in every tested cell. Antagonists (10 μM CNQX, 10 μM NBQX, 10 μM MK801) were applied in bath.

Clampfit 10.3 software (Molecular Devices) and OriginPro (Origin
Lab) were used for analysis. Mini Analysis Program (Synaptosoft) was used for off-line analysis of sEPSCs, as described. 35 Briefly, excitatory events were distinguished from baseline noise by setting the appropriate parameters for each individual cell and by manually eliminating false events. The AMPAR-mediated postsynaptic currents were analyzed for the frequency of their occurrence and the amplitude.
For bursting activity, polysynaptic events occurring in the bursts were counted.

Dysfunction of human iPSC-derived neurons with the MAPT 10+16 mutation links to upregulation of AMPAR and NMDAR
We have previously shown that human iPSC-derived cortical neurons with the 10+16 MAPT mutation linked to FTD exhibit pathophysiological excitability at late stages of neurogenesis (≈150 DIV), revealed as a depolarized resting membrane potential associated with impaired firing due to reduced expression of Nav1.6 channels. 24 We therefore used electrophysiology to further explore dysfunctions on these neurons.
In cell-attached mode, iPSC-derived cortical neurons (healthy cohort) demonstrated a sustained spontaneous activity, a constant and regular firing ( Figure 1A), representing genuine human cell activity (i.e., not compromised by membrane breakthrough and potential washing out/dilution of intracellular regulators). On the contrary, iPSC-derived neurons with the 10+16 MAPT mutation displayed bursting, irregular discharge consisting of high-frequency bursts (≈2.6 Hz to 9.3 Hz), fol-lowed by periods of prolonged "silence" ( Figure 1B) associated with membrane depolarization (Figure S1A in supporting information). In whole-cell mode, iPSC-derived neurons with the 10+16 MAPT mutation (held at -60 mV) revealed bursts of spontaneous discharge, again, opposing regular firing in the age-matched control neurons ( Figure 1C).
We observed two characteristic patterns of spontaneous activity in FTD: (1) high-frequency firing associated with sustained depolarization (case 1) and (2) trains of short bursts consisted of few spikes followed by periods of "silence" (case 2; Figure 1C).
To evaluate synaptic dysfunction, we next recorded synaptic events (sEPSC), which revealed about two-fold increase in the frequency of events occurred in neurons with the mutation ( Figure 2B). These currents were AMPAR-mediated because they were recorded at -70 mV and eliminated by the competitive AMPAR antagonist NBQX (Figure 2A). The increased frequency indicates boosted synaptic drive in FTD. Interestingly, the mutation did not change the sEPSC amplitude (P = 0.10; Figure 2B). Consistently, in 10+16 neurons, about ≈41% of all synaptic events occurred within high-frequency bursts ( Figure 2B), further confirming irregular, bursting activity in FTD.
To understand how neuronal dysfunction in FTD relates to changes in glutamate receptor functioning, we next recorded glutamate-evoked currents. The currents were evoked by locally applied glutamate (see Methods) and recorded in human iPSC-derived neurons at different membrane potentials ( Figure 2C) in response to train of repeated glutamate puffs (500-ms to 5-s pulses; Figure 2D). In 10+16 iPSC-derived neurons, the amplitude of glutamate-evoked currents recorded at -70 mV was increased ≈7.6 times compared to that in control (Figure 2D). Because it was recorded at -70 mV and fully eliminated by a selective AMPA/kainate receptor antagonist CNQX, this indicates upregulated AMPAR in FTD neurons. The glutamate-evoked current amplitude was also increased when recorded at +40 mV (above ≈4.4 times, Figure 2E). Such an increase in the current amplitude was reduced by CNQX (by ≈50%, P < 0.001 paired comparison), reflecting the AMPAR-mediated component, while the subsequent application of an activity-dependent NMDAR antagonist, MK-801, fully eliminated the remaining current ( Figure 2E). This indicates that in FTD human neurons with the MAPT mutation, the glutamate-evoked current increase is due to upregulation of both AMPAR and NMDAR.

Increased AMPAR-and NMDAR-mediated Ca 2+ influx in iPSC-derived neurons with the MAPT 10+16 mutation
Impairment of glutamatergic signaling by AMPAR and NMDAR upregulation might disturb intracellular Ca 2+ homeostasis, which is a known mechanism leading to neuronal death in tau-induced FTD. 4 [Ca 2+ ] c with most of them depicting a second (delayed) peak (Figure 3A). AUC was ≈two times larger; while recovery of basal [Ca 2+ ] c significantly reduced (≈to half) in FTD neurons ( Figures 3C, 3D), and ultimately, was only achieved by washing out the agonist ( Figure 3A).
Basal [Ca 2+ ] c before stimulation was however similar between cohorts ( Figure S2A in supporting information). Nevertheless, trains of spontaneous [Ca 2+ ] c rise were observed in some neurons from patient 2 ( Figure 3A), in agreement with previous results. 5 We also explored the calcium response to other stimuli that further helped us to distinguish neurons and glial cells ( Figure 3A). 36 To further confirm the increase in AMPAR/NMDAR-mediated Ca 2+ permeability, we stimulated the cells with the selective agonists. Bath application of NMDA (20 μM, in Mg 2+ -free medium) or AMPA (20 μM) to neuronal cultures evoked a [Ca 2+ ] c rise, which was, again, of a higher amplitude, with a larger AUC, in neurons with the 10+16 MAPT mutation compared to healthy controls ( Figures S3A, S3B, S3D, S3E). Interestingly, application of kainate induced a higher rise in the patients, but [Ca 2+ ] c at the end of the experiment was not significantly different between all the controls and patients (Figures S2C, S2F). These data point to the tau-induced upregulation of Ca 2+ -permeable AMPAR and NMDAR in FTD neurons. . Box plots represent the median, 25, and 75 percentiles. Non-parametric Kruskal-Wallis H test was used to determine whether there were statistically significant differences between controls and patients, or between the treatments in each group (ns: non-significant, ***P < 0.0001) frame) demonstrated a robust transient [Ca 2+ ] c rise in wt-tau neurons in response to repetitive localized applications of glutamate (10 μM), whose kinetics well resembled a spatial-temporal profile of agonist application/diffusion (a few seconds time-course; Figure 4C).

Introduction of the 10+16 MAPT mutation impairs calcium signaling and mitochondrial function in genetically engineered neurons
Glutamate-evoked [Ca 2+ ] c transients were eliminated by CNQX (Figure 4D). In genetically engineered 10+16 MAPT neurons, the [Ca 2+ ] c rise had a dramatically slower decay and a higher peak amplitude compared to the isogenic wt tau cells (Figures 4C, 4E). Electrophysiological recordings also demonstrated increased glutamate-induced currents in 10+16 MAPT neurons, which were eliminated by CNQX at -70 mV ( Figure S4A in supporting information) and CNQX with APV at 40 mV ( Figure S4B). Importantly, genetically engineered 10+16 neurons also mimicked the mitochondrial dysfunction we previously described in 10+16 FTD neurons. 8 Engineered 10+16 neurons exhibited a hyperpolarized mitochondria and an increased rate of cytosolic and mitochondrial ROS production ( Figures 4F-4G, Figures S4C, S4D) compared to their isogenic wt tau neurons. Depolarization of the mitochondria with rotenone reduced ROS production in 10+16 neurons ( Figures   S4C, S4D), indicating mitochondrial hyperpolarization was the underlying cause of the elevated ROS, as previously observed in patients' neurons. 8 Thus, both calcium-deregulation and mitochondrial dysfunction appear to be specifically mediated by this tau mutation.

Mitochondrially located antioxidants prevent the glutamate-induced calcium deregulation in MAPT 10+16 iPSC-neurons
Indeed, we previously showed that 10+16 FTD-neurons are more vulnerable to calcium-induced cell death 5 and that mitochondrial ROS overproduction is a key pathological event, leading to neuronal death that could be prevented with mitochondrially targeted antioxidants. 8, 23 We therefore reasoned whether the neuroprotec-tive effect of mitochondrial (and other) antioxidants could be exerted by targeting AMPAR and NMDAR-mediated calcium overload. We  Figures 5A-5C, Figure S5A). In contrast, treatment with D-PUFAs did not significantly changed peak amplitude or AUC, nor facilitated recovery of basal [Ca 2+ ] c in the patients (Figures 5A-5C, Figure S5A

Mitochondrial antioxidants reduce the surface levels of specific AMPAR and NMDAR subunits elevated in patients' neurons
Excessive ROS production might exert a pathogenic role by mediating the oxidation of numerous proteins and altering their structure, hence function. 37,39 Some amino acid oxidations can be reversed (i.e., methionine or cysteine oxidation), while amino acid carbonylation remains irreversible. Accumulation of reversibly oxidized amino acids may act as a regulatory switch. 40,41 To further understand the role of mitochondrial ROS overproduction in FTD pathogenesis we next explored the pattern of protein oxidation in neurons using redox proteomics.
As expected, we observed an increase in the amount of oxidized peptides in patients compared to control ( Figure 6A). Pre-incubation of neurons with MitoQ effectively reduced the number of oxidized peptides (by > 20%) in FTD samples ( Figure 6A) and shed some light on potential protein candidates to explain how mitochondrial ROS mediates tau pathology and glutamatergic dysfunction ( Figure 6B). Notably, MitoQ reversed the oxidation of proteins such as MAP1B, clathrin, or Hsc70, known to regulate AMPAR and NMDAR trafficking, so we next explored this possibility by analyzing the distribution of these receptors between cytosol and plasma membrane of neurons using a crosslinking assay as described in Methods.

AMPARs are tetramers composed of different subunits (GluA1-4),
with GluA1 and GluA2 predominant, and the lack of editing of GluA2 conferring the receptor calcium permeability. There was no significant difference in the total levels of all four AMPAR subunits (GluA1-4) measured with a pan AMPAR antibody ( Figure 6C), or specifically GluA1 ( Figure 6F) between control and FTD neurons. However, their surface expression was significantly higher in patients ( Figures 6D, 6E, 6G, 6H).  (Figures 6D, 6E, 6G, 6H), with no significant effect on total subunit content ( Figures 6C, 6F). This indicates increased mitochondrial ROS contributes to the trafficking of AMPARs, and specifically the calciumpermeable subunit GluA1, to the cellular membrane in patients.
NMDARs are also tetramers composed of two obligatory NR1, plus NR2(A-D) (or more rarely NR3) subunits, which confer the receptor-specific signaling properties. NR1 total and surface levels were similar between control and FTD neurons, and were unaffected by mitoTEMPO (Figures 6I-6K). However, NR2B, known to predominately locate in extrasynaptic membranes and contribute to excitotoxicity, 42 was highly expressed in patients ( Figure 6L), and MitoTEMPO significantly reduced its presence in the cell membrane ( Figures 6M, 6N). These results indicate that although membrane expression of NMDARs is similar as indicated by NR1, specific NR2Bcontaining receptors involved in excitotoxicity are upregulated in the patients and can be modulated by mitochondrial antioxidants, providing a mechanistic basis of the neuroprotective action of these compounds.

Extracellular 4R tau impairs the glutamate-induced calcium response of control neurons by increasing mitochondrial ROS production
In the recent years, increasing evidence supports that tau spreads through the brain in a "prion-like" manner, in a mechanism involving extracellular tau release and uptake by cells. 43,44 We found that iPSCderived neurons were able to secrete tau and interestingly, tau secretion in patient 2 was significantly higher than in control ( Figure S6B in supporting information). Notably, this patient also showed spontaneous calcium oscillations in the absence of any stimulation (Figure 3A), which is consistent with other authors showing that increased neuronal activity stimulates the release of tau. 45 We then hypothesized if secreted extracellular tau was also able to alter the glutamate- Box plots represent the median, 25-and 75 percentiles. Non-parametric Kruskal-Wallis H test was used to determine whether there were statistically significant differences between controls and patients, or between the treatments in each group (ns: non-significant, *P < 0.05, ** P < 0.01, *** P < 0.0001). D-F, Patch-clamp recordings of the glutamate-evoked currents in iPSC-derived neurons. D, Representative whole-cell recordings in FTD-iPSC-derived neurons with the MAPT 10+16 mutation without treatment (red) and post-treatment with MitoTEMPO (2 hours pre-incubation) at -70 mV (lower traces) and 40 mV (upper traces) in response to local glutamate (Glu) puffs (as indicated). Statistical comparison of the glutamate-evoked current amplitudes at -70 mV (E) and 40 mV (F) between control (gray) and FTD groups without treatment (red) and treated with mitochondrial antioxidant (blue). Data are mean ± standard error of the mean (SEM). *** P < 0.001; **P < 0.01 (unpaired t-test) (1)(2)(3)(4), detected with a pan-AMPA antibody, (H) AMPAR subunit GluA1, (K) NMDAR subunit NR1, (N) NMDAR subunit NR2B. Neuron-specific βIII tubulin was used as a loading control. Absence of higher molecular weight bands in the actin band (E) confirms that intracellular proteins were not crosslinked. Histograms show the quantification of the number of samples indicated in brackets, lines represent the median. For the quantification of the total levels, intracellular and surface levels were added. In the total levels group, some additional experiments using non-crosslinked samples were also included. In all cases, data was normalized to control for each experiment. Statistical significance between control and patients was analyzed with the non-parametric Mann-Whitney test in all cases except (G), which followed a normal distribution and two-sample t-test was used, number of samples analyzed is indicated in brackets. Specific  26 This suggests a specific role for 4R tau isoform, whose production is enhanced by the 10+16 mutation, in the mechanism of action. To confirm this point, we used K18 tau, which is a construct from wt 4R tau comprising the four-repeat region of the protein, to treat primary rat cortical neurons (300 nM K18 tau, 24 h).
Similarly to human iPSC-derived neurons, tau-treated primary neurons also showed upregulated calcium responses to a physiological concentration of glutamate (5 μM), which were restored with the mitochondrial antioxidant MitoQ (Figures 7C-7G). Notably, mitochondrial ROS production was also increased by K18 tau (Figures 7H, 7I).
Calcium deregulation in tau-treated primary neurons became even clearer under exposure of these cells to toxic concentrations of glutamate. Thus, 50 μM glutamate induced the initial peak and delayed calcium deregulation typical for this concentration, which was higher in tau-treated cells (Figures S6G, S6H, S6K). And again, the mitochondrial antioxidant MitoQ significantly decreased the calcium signal in response to 50 μM glutamate in both control and tau-treated neurons ( Figures S6I-S6K). The specific role of 4R tau was further confirmed by treating primary neurons with extracellular 3R tau. Interestingly, 3R tau also induced an altered calcium response in the neurons but by a different mechanism than 4R tau, because it was not modulated by mitochondrial ROS (Figure S6L), which indeed were only upregulated by 4R and not 3R tau ( Figure S6M). Taken together, these results confirm that extracellular or secreted 4R tau recapitulate the pathological phenotype of 10+16 mutation.

DISCUSSION
Here we report a direct link among mitochondrial ROS, calcium signaling, and glutamatergic transmission deregulation, which might lead to early dysfunction preceding neuronal loss in tauopathies, and, according to our results, is also involved in the mechanism of neurodegeneration.
Our data demonstrate that in 10+16   contribute to the altered glucose metabolism occurring in neurodegenerative disorders, which might also affect the synaptic function, due to its high energy requirements. We previously showed glucose metabolism was altered in 10+16 neurons, 8 which were able to maintain ATP levels at the age they were studied by different compensatory mechanisms. This chronic impairment in the energy supply might be the trigger for neuronal dysfunction and neurodegeneration. Indeed, we demonstrate that mitochondrial bioenergetics dysfunction, shown by the increased mitochondrial membrane potential, is a key event in the pathology, as it drives the increased mitochondrial ROS production.
Proteomics data suggest that the protective effect of MitoQ is related to the regulation of AMPAR and NMDAR trafficking. Oxidation of several cytoskeletal-related proteins closely related to tau was also recovered by MitoQ, such as the beta 3 and 4a tubulins, which are components of the microtubules, together with the cytoplasmic dynein, myosin 10, or the microtubule-associated protein 1B (MAP1B), which might affect axonal transport. Importantly, MAP1B has been shown to modulate synaptic transmission by regulating AMPAR [61][62][63] and NMDAR trafficking. 64 Clathrin heavy chain oxidative modifications were also reduced by MitoQ in the patients' cells.
This protein has a major role in the formation of coated vesicles essential for clathrin-mediated endocytosis, which is implicated in the AMPA [65][66][67] and NMDA 68,69 receptors' internalization. Another component of this machinery, the heat shock cognate 71, Hsc70, a constitutively expressed heat shock protein important for the uncoating of the clathrin-coated vesicles, 70 was also recovered by MitoQ. Alterations in Hsc70 might lead to the impairment of clathrin-mediated endocytosis, 71 and therefore, AMPAR and NMDAR internalization.
Interestingly, oxidative modifications of this protein were also found in AD brains by redox proteomics. 59 In addition, the tyrosine-protein phosphatase Shp-2 also has a role in the regulation of NMDA functionality, by regulating its phosphorylation. 72 Taken together, redox proteomics data show that MitoQ was able to reverse the oxidative modifications found in several proteins implicated in AMPAR and NMDAR trafficking in patients' cells, suggesting this could be the mechanism for the altered glutamatergic signaling induced by mitochondrial ROS.
Here, we show that tau pathology alters glutamatergic signaling, and, conversely, it has also been described that overactivation of AMPARs and NMDARs are able to trigger tau hyperphosphorilation and pathology. 73,74 Increased neuronal activity also leads to enhanced tau secretion and propagation. 45 We show that iPSC-derived neurons are able to secrete tau, and importantly, extracellular tau (specifically the 4R isoform, either patient-secreted or the recombinant K18 fragment) reproduces the alterations in the glutamatergic signaling observed in the mutated iPSC neurons. Rather than a direct effect of extracellular tau in the receptors at the membrane site, the mechanism appears to be also mediated by tau-induced mitochondrial ROS, as the altered calcium signaling is restored in the presence of mitochondrially targeted (intracellular) antioxidants. Thus, this could be one of the mechanisms for tau pathology propagation and amplification in FTD and other tauopathies.

CONFLICTS OF INTEREST
The authors declare no competing interests.