Carbonic anhydrase inhibition selectively prevents amyloid β neurovascular mitochondrial toxicity

Summary Mounting evidence suggests that mitochondrial dysfunction plays a causal role in the etiology and progression of Alzheimer's disease (AD). We recently showed that the carbonic anhydrase inhibitor (CAI) methazolamide (MTZ) prevents amyloid β (Aβ)‐mediated onset of apoptosis in the mouse brain. In this study, we used MTZ and, for the first time, the analog CAI acetazolamide (ATZ) in neuronal and cerebral vascular cells challenged with Aβ, to clarify their protective effects and mitochondrial molecular mechanism of action. The CAIs selectively inhibited mitochondrial dysfunction pathways induced by Aβ, without affecting metabolic function. ATZ was effective at concentrations 10 times lower than MTZ. Both MTZ and ATZ prevented mitochondrial membrane depolarization and H2O2 generation, with no effects on intracellular pH or ATP production. Importantly, the drugs did not primarily affect calcium homeostasis. This work suggests a new role for carbonic anhydrases (CAs) in the Aβ‐induced mitochondrial toxicity associated with AD and cerebral amyloid angiopathy (CAA), and paves the way to AD clinical trials for CAIs, FDA‐approved drugs with a well‐known profile of brain delivery.

Carbonic anhydrases (CA) are enzymes involved in the reversible conversion of carbon dioxide and water into bicarbonate and protons. They are present in all the vertebrates, showing different intracellular locations and regulating pH and ion transport. CA-VA and CA-VB have a mitochondrial localization (Ghandour, Parkkila, Parkkila, Waheed & Sly, 2000). CA-II, known as cytoplasmic, was also recently shown by proteomic profiling to be increased in brain mitochondria in aging and neurodegeneration (Pollard, Shephard, Freed, Liddell & Chakrabarti, 2016). CA inhibitors (CAIs) are used to treat a variety of disorders including glaucoma, epilepsy, neuropsychiatric disorders, and acute mountain sickness (Aggarwal, Kondeti & McKenna, 2013;Fossati et al., 2016;Huang et al., 2010).
In this study, we examine multiple mitochondrial pathways of amyloid toxicity in neuronal and cerebral endothelial cells (ECs), and evaluate CAIs as active regulators of these processes. We analyze changes in mitochondrial membrane potential, production of ATP, emission of ROS (reactive oxygen species), mitochondrial and cytoplasmic calcium influx, as well as activation of caspase 9 and cell death. While unveiling mechanistic insights into the deleterious mitochondrial actions of amyloid, we propose and test a novel therapeutic approach for preventing these deleterious events. We analyze the role of the CAI methazolamide (MTZ) and, for the first time, its analog acetazolamide (ATZ), on specific Ab-mediated pathways of mitochondrial dysfunction and apoptotic cell death, in both neuronal cell lines and microvascular ECs, challenged, respectively, with Ab42 and the vasculotropic Ab40-Q22 (Fossati, Ghiso & Rostagno, 2012a). Importantly, we include the analysis of mitochondrial toxicity in cerebral endothelial cells. The deposition of amyloid (predominantly Ab40) around cerebral vessels and microvessels, known as cerebral amyloid angiopathy (CAA), is today recognized as an integral part of the disease. In addition to the well-known neurodegenerative pathology caused by the parenchymal deposition of Ab (mainly in its 42 amino acids form), CAA is known to cause vascular damage, microand macro-hemorrhage, apoptosis, and dysfunction of the entire neurovascular unit. These neurovascular effects further exacerbate the pathology and progression of the disease (Revesz et al., 2009;Zlokovic, 2008). Mutations in the Ab peptide generate variants such as the Ab40-Q22 mutant, which are associated with CAA, hemorrhagic stroke, and early-onset dementia in AD familiar forms, induce aggressive endothelial cell damage, and can represent useful tools to study amyloid-mediated vascular pathology (Fossati et al., 2010).
MTZ was first selected from a drug library for its ability to inhibit cytochrome C (CytC) release from isolated mitochondria, showing beneficial effects in models of Huntington's disease (Wang et al., 2008) and ischemia-reperfusion injury (Wang et al., 2009). Albeit pointing to a mitochondrial effect of MTZ, the mechanisms of action were not fully clarified. MTZ prevented cell death and CytC release in cellular and mouse models of Ab-induced neurodegeneration (Fossati et al., 2016). This is the first study expanding the analysis to other members of the CAIs family and deeply analyzing the mitochondrial mechanism of action of these drugs. Here, we thoroughly examined the effects of two different FDA-approved and clinically used members of the CAI family (MTZ and ATZ) on Ab-mediated mitochondrial damage, and we tested for the first time if the protective effect induced by MTZ on CytC release, the resulting caspase-9 activation, and apoptosis, were also exerted by an analog CAI, ATZ.
The FDA has approved MTZ and ATZ for use in glaucoma decades ago. CAIs are also currently approved for the prevention of acute mountain sickness and related cerebral edema and as diuretics. Furthermore, ATZ is used in the treatment for idiopathic intracranial hypertension and normal pressure hydrocephalus (Alperin et al., 2014). Their use in these neurological disorders as well as in epilepsy (Aggarwal et al., 2013) confirms the ability of these drugs to reach the brain at effective concentrations. Due to the long-term use of MTZ and ATZ in chronic conditions, the efficacy and the safety of their systemic administration have been widely assessed (Wright, Brearey & Imray, 2008), making clinical trials for CAIs in AD a concrete possibility. Our novel findings on the mitochondrial effects of MTZ and ATZ against neuronal and vascular amyloid toxicity justify the selection of these drugs as a therapeutic strategy for AD and CAA.
2 | RESULTS 2.1 | Ab treatment elicits mitochondrial membrane depolarization and increases mitochondrial H 2 O 2 production. CAIs counteract both effects First, to determine the concentrations of MTZ and ATZ effective to decrease the apoptotic effect of the Ab peptides in both cell lines, we conducted a dose-response experiment measuring DNA fragmentation ( Figure 1), which showed that ATZ is about 10-fold more effective than MTZ at inhibiting apoptosis in both cell types. Afterward, to clarify the molecular mechanisms responsible for the mitochondrial effects of Ab and to determine whether CAIs exert a protective effect on these processes, we analyzed the main pathways responsible for maintaining mitochondrial function. Preservation of mitochondrial membrane potential (DΨ) is an essential element for cell physiology, survival, and energetic function. Indeed, mitochondrial membrane depolarization is known to precede and facilitate apoptotic cell death. We studied the effects of Ab42 and Ab40-Q22 on mitochondrial membrane potential in neuronal cells (SH-SY5Y) and ECs, respectively. Membrane potential was measured using TMRM fluorescent probe. Our data clearly showed that challenge with aggregated (oligomeric) forms of the Ab peptides induced a depolarizing effect on the mitochondrial membrane of neuronal cells and ECs, after only 45 min of treatment ( Figure 2a). This effect was especially dramatic in the case of microvascular ECs, where DΨ was reduced more than 60% by Ab. It is interesting that the magnitude of the membrane depolarization induced by Ab fibrils and monomers was significantly lower, compared to the depolarization induced by the oligomeric forms. The scrambled Ab42 peptide, used as a negative control, did not exert any effect on mitochondrial DΨ ( Figure 2a). On the other hand, addition of 10 lM of FCCP to completely depolarize mitochondria resulted in loss of fluorescent signal in both cell lines (Supporting Information, Figure S1), confirming that in our experimental conditions, TMRM fluorescence decrease reflects the degree of mitochondrial depolarization. Both MTZ and ATZ were able to rescue DΨ to values similar to those observed in control. 100 lM was used as starting point for MTZ, due to our data indicating an effect of this or higher doses on preventing cell death [ (Fossati, Todd, Sotolongo, Ghiso & Rostagno, 2013;Fossati et al., 2016) and Figure 1]. ATZ, used for the first time in this study, was able to prevent the loss of DΨ and to maintain the potential at the level of control cells at a significantly lower concentration (10 lM). To demonstrate that this effect was specifically due to the inhibition of CAs by MTZ and ATZ, we used N-methyl acetazolamide (100 lM), a structural analog of ATZ unable to inhibit CAs. Treatment of SH-SY5Y cells with the analog exerted no effect on DΨ, either under control conditions or in the presence of Ab.
Increased production of mitochondrial H 2 O 2 is a classical signal of mitochondrial dysfunction and an essential mediator of cell death (Singh, Sharma & Singh, 2007). H 2 O 2 is a membrane permeable second messenger, as well as a potent precursor of other ROS generation (Turrens, 2003). H 2 O 2 production is also tightly regulated by DΨ. We measured the levels of H 2 O 2 produced by isolated mitochondria purified after neuronal and ECs treatment with Ab, in the presence or absence of the CAIs. Ab induced a significant increase in the amount of H 2 O 2 generated by isolated mitochondria (three-  Fluo-4/cytoplasmic Ca 2+ ), we measured the levels of mitochondrial and cytoplasmic-free Ca 2+ (Figure 4). Due to their charge, rhodamine-based calcium probes are known to be preferentially localized in mitochondria (Smithen et al., 2013). This fact was confirmed under our experimental conditions, as shown in Supporting Information, Figure S2a, insert.
Surprisingly, we found that an acute treatment (45 min) with pre-aggregated oligomeric Ab peptides decreased mitochondrial-free calcium levels in both cell types (Figure 4a F I G U R E 2 Mitochondrial membrane depolarization induced by Ab oligomers is prevented by CAIs. Graphs showing mitochondrial membrane potential (DΨ) measured by TMRM in SH-SY5Y neuronal cells (left) and microvascular ECs (right). (a) The monomeric and fibrillar forms of the peptides induced a much lower mitochondrial membrane depolarization. It is interesting that the scrambled form of the peptide did not exert any effect on DΨ in any of the cell lines. Ab42, on its different aggregation states, was always added at 10 lM, while the final concentration of Q22 was 50 lM. (b) DΨ in the neuronal cells (left panel) is reduced to about 65% of control cells. The reduction is completely prevented by MTZ and ATZ. In ECs cells (central panel), DΨ is reduced to 35% of the control levels by Ab, and reverted to above 80%, after treatment with the peptide in the presence of CAIs. N-methyl acetazolamide (100 lM), a structural analog of ATZ unable to inhibit CAs, showed no effect on DΨ either under control conditions or in the presence of Ab, in SH-SY5Y cells (right panel). Data in histograms are mean AE SEM of, at least, three independent experiments and cytoplasmic Ca 2+ in neuronal cells, while in ECs, the highest doses of MTZ and ATZ were able to counteract the effect exerted by the peptide (Figure 4b, d).
To exclude experimental artifacts, we subjected the cells to different loading and washing times, which consistently resulted in decreased levels of free calcium in both mitochondrial and in the intensity of the fluorescence. In addition, uneven distribution of the dye within mitochondria could also occur when using this probe. However, rhodamine-based fluorescence probes have been extensively used in the literature as a method to assay mitochondrial calcium Babcock, Herrington, Goodwin, Park & Hille, 1997;Boitier, Rea & Duchen, 1999), and we conducted all the experiments accurately and using the appropriate controls.
2.3 | Carbonic anhydrase inhibition does not affect intracellular pH or ATP generation CA catalyzes the interconversion of CO 2 and H 2 O to HCO À 3 and protons, through a reversible reaction (Meldrum & Roughton, 1933). To examine a possible effect of CAIs on proton flux across the inner mitochondrial membrane and energy production, we measured cellular ATP levels in permeabilized cells, using a luciferin-luciferase assay. Dissimilar results were obtained in SH-SY5Y and ECs. Treatment of SH-SY5Y cells with Ab induced a modest decrease in ATP levels (p = 0.05, Figure 5c). These levels remained unchanged upon treatment with MTZ or ATZ in combination with the peptide. As a control, treatment with ATP synthase inhibitor oligomycin induced a sharp decrease (p < 0.0001). In contrast, treatment of ECs with the Q22 peptide induced a 21.6% increase in steady-state ATP levels (p = 0.001, Figure 5c). This increase was unaffected by cotreatment with either MTZ or ATZ. It is interesting that oligomycin, similar to Ab, also increased ATP luminescence in the ECs (p = 0.0002).

| Ab-induced apoptosis and caspase activation are prevented by CAIs
Apoptotic cell death is a well-known contributor to neurovascular degeneration in AD. CytC release from dysfunctional mitochondria and the resulting caspase-9 activation are known to play key roles in the apoptotic process. We have recently reported a protective effect of MTZ against apoptotic cell death in models of Ab-induced toxicity (Fossati et al., 2016). Here, we tested for the first time if the protective effect induced by MTZ on caspase activation and CytC release was also exerted by the analog CAI ATZ.
We analyzed the effect of MTZ on amyloid-mediated mitochondrial cell death pathways, showing that Ab-induced CytC release ( Figure 6a and b), caspase-9 activation (Figure 6c and d), and DNA fragmentation (Figure 1a-d) were inhibited by MTZ starting at 100 lM concentration, which confirmed our recent work (Fossati et al., 2016). Importantly, ATZ was effective in preventing CytC release and caspase-9 activation, as well as apoptosis, at concentrations 10 times lower than MTZ, starting at concentrations ≤10 lM (Figures 1a-d and 6). ATZ completely reverted Ab-induced caspase-9 activation ( Figure 6).

| DISCUSSION
Mitochondrial dysfunction is an early and causal step in AD pathology, tightly linked to neurodegeneration and promoting cognitive impairment (Hirai et al., 2001;Swerdlow, Burns & Khan, 2014;Swerdlow & Khan, 2009;Swerdlow et al., 2010). The apoptotic outcome has been attributed to the pathological effects of Ab intermediate aggregation species (particularly oligomers and protofibrils) on mitochondrial pathways (Fossati et al., 2010(Fossati et al., , 2012b. However, the specific biochemical pathways leading to mitochondrial dysfunction in the presence of amyloid, as well as the resulting activation of cell death pathways in neurovascular cells, are still unclear.
Here, we revealed that CAs might be previously unrecognized key targets in these processes. Our results clearly showed induction of mitochondrial membrane depolarization and increased mitochondrial H 2 O 2 production, in response to Ab-challenge, in both neuronal and cerebral microvascular ECs. MTZ and ATZ, two different members of the CAI family, were effective at inhibiting the mitochondrial dysfunction pathways induced by Ab. Intriguingly, other mitochondrial parameters, such as ATP production and pH, were not equally affected. Moreover, mitochondrial and cytoplasmic calcium flux did not seem to be essential for the mechanism of action of the CAIs.
In the presence of Ab, mitochondrial membranes were strongly depolarized and mitochondrial production of H 2 O 2 was increased.
These data are concordant with previous reports, showing that the Ab peptide affects the production of different types of ROS, including H 2 O 2 (Kaminsky & Kosenko, 2008). The essential role played by mitochondrial H 2 O 2 production in the activation of the apoptotic pathway in our study is concordant with previous work showing that increments in the generation of this molecule appear early after Abchallenge (Milton, 2004;Tabner et al., 2005). In our model, H 2 O 2 production and mitochondrial membrane depolarization, which is also a key process in AD pathogenesis (Moreira et al., 2010), appear as primary inductors of Ab-mediated apoptotic cell death and as the main targets of CAIs. Both parameters were clearly reverted when CA was inhibited by MTZ and ATZ.
It is interesting that despite the increase in H 2 O 2 produced by mitochondria isolated after cell treatment with Ab (Figure 3a), other intracellular ROS measured in whole cells were not increased in response to Ab in our model (Figure 3b  where likely H 2 O 2 is not continually present, due to its ability to cross membranes. The fact that H 2 O 2 is highly unstable and that it reacts with lipids and proteins, inducing peroxidation, is also in line with the proposed hypothesis.
To determine whether CAIs affect energy production or proton availability after Ab challenge, we measured ATP levels in the presence or absence of CAIs. It is interesting that despite cell type differences likely due to different coupling properties, MTZ or ATZ treatment did not affect the steady state of ATP, either in control conditions or in cells treated with Ab. These results suggest that energy production and proton flux are not involved in the CAIs' mechanism of action, which seems primarily driven by the prevention of mitochondrial DΨ changes and by the reduction in mitochondrial H 2 O 2 release. Thus, the protective effects of CAIs against amyloid peptides are not due to an increase in mitochondrial respiration. It is interesting that the amyloid peptides alone exerted differential effects on SH-SY5Y and ECs, inducing a slight decrease in ATP levels in neuronal cells and a substantial increase in cerebral

(c)
F I G U R E 5 Intracellular pH and ATP production are not affected by CAIs. Measurement of the intracellular pH after 3 hr of treatment with the pre-aggregated peptide is shown in (a) for SH-SY5Y and for ECs. The same pH measurement after 16 hr of treatment with Ab without preaggregation is represented in (b) (Ab42 10 lM or Q22 50 lM). Cellular pH is not significantly affected by Ab and CAIs (MTZ 100 or 300 lM and ATZ 10 or 100 lM). (c) Bar histograms showing ATP production in response to Ab peptides or to peptides in the presence of CAIs for SH-SY5Y and ECs cell cultures. ATP production is measured by a luminometric assay (CellGlo, Promega) and is represented as A.U. Data in histograms are mean AE SEM of, at least, three independent experiments ECs. A possible explanation is that these cell types handle mitochondrial dysfunction by differentially resorting to aerobic glycolysis (Newington et al., 2011). In fact, we can presume fundamentally different glycolytic metabolism for neuronal cells (such as SHSY-5Y) and ECs. This may explain the different effects induced by the ATP synthase inhibitor oligomycin in our experiments and strengthens our conclusion that the protective mechanisms of CAIs are independent of ATP production.
Maintaining the proper mitochondrial calcium concentration is imperative for cell survival. In fact, mitochondria, jointly with the ER, are the two main organelles in charge of keeping calcium cell homeostasis. However, while in the ER the levels of free calcium are kept within the physiological range by a group of proteins called calsequestrins (MacLennan & Wong, 1971), mitochondria lack any specific protein to exert this action, and the mechanism governing this process is still unclear.
Previous work showed that Ca 2+ homeostasis is dysregulated in cellular models of AD, as well as in human AD brains (Berridge, Bootman & Lipp, 1998;Celsi et al., 2009;Garwood et al., 2013). Surprisingly, our data showed decreased levels of free calcium, both  (Granatiero et al., 2015). Discordant results have been reported regarding the effects of Ab on cellular and mitochondrial calcium, with studies showing data both consistent and contradictory with our findings (Abramov, Canevari & Duchen, 2003, 2004. The observed effect on mitochondrial and cytoplasmic calcium concentration was partially rescued by the highest concentrations of MTZ and ATZ in microvascular ECs, reaching values similar to those found under control conditions, while no modulation by CAIs was observed in neuronal cells. This is key evidence that mitochondrial CAIs effects on Ab-induced toxicity are independent of calcium uptake in both mitochondria and cytoplasm, as even in conditions in which CAIs do not affect mitochondrial and cytoplasmic Ca 2+ flux (neuronal cells), CAIs are able to inhibit loss of DΨ and production of H 2 O 2 , as well as caspase activation and cell death. In the absence of an obvious effect of CAIs on mitochondrial Ca 2+ levels, we concluded that the observed effects are independent of Ca 2+ homeostasis. Thus, rather than examining pathways in which fluctuating levels of mitochondrial Ca 2+ has been shown, as the opening of the mitochondrial permeability transition pore or the ER-mitochondria interactions, we opted to focus our studies on other mitochondrial parameters affected by CAIs. One of the main consequences of the over dosage of CAIs in humans is the imbalance in the serum electrolyte levels and the resultant change in blood pH (Crandall, Bidani & Forster, 1977). Small increases in the pH of the solutions where MTZ and ATZ are dissolved are also linked to increases in their solubility (Jiang et al., 2013) and may induce changes in the bioavailable concentration of the CAIs, introducing more complexity in our study. Therefore, we monitored possible changes in the intracellular pH upon cell treatment with the CAIs, which could affect Ab-induced toxicity. No significant changes in the pH were found (Figure 5b and c). This allowed us to exclude that the protective effects of CAIs were secondary to pH modulation.
As expected, CytC release and caspase-9 activation were induced by Ab challenge in both neuronal and microvascular ECs.
DNA fragmentation, indicating apoptotic cell death, was also increased by Ab in both cell types. This is the first study showing that two different CAIs were able to counteract the detrimental effects of Ab on mitochondrial dysfunction and apoptotic cell death and to analyze their mitochondrial mechanisms of action. While previous findings in neurodegenerative diseases have proposed a role of MTZ in the prevention of mitochondrial dysfunction and CytC release (Fossati et al., 2013(Fossati et al., , 2016Wang et al., 2008Wang et al., , 2009), these studies did not explore the potential protective effects of other CAIs, and did not show that the effects were specifically due to CA inhibition. We hypothesized that the prevention of Ab-mediated mitochondrial dysfunction may be due to a direct effect on mitochondrial and/or cellular CAs, as shown by the lack of effect of an inactive ATZ analog. Albeit none of the CAIs available today are fully selective for one of the enzymes, both MTZ and ATZ have high activity on mitochondrial CA (CA-VA and -VB) (Supuran, 2008).
In our models, CAIs prevent Ab-induced apoptosis by inhibiting loss of DΨ and production of H 2 O 2 , as well as CytC release from the mitochondria. A possible mechanism responsible for the prevention of mitochondrial depolarization and H 2 O 2 production is that pharmacological inhibition of mitochondrial CAs slows down the production of HCO À 3 , limiting Krebs cycle and electron transport chain, and thus reducing the production of H 2 O 2 and subsequent oxidative stress.
This mechanism is also proposed by Shah's group, who showed that inhibition of CAs rescued high-glucose induced mitochondrial dysfunction, ROS production, and pericyte loss in diabetic mice (Price, Eranki, Banks, Ercal & Shah, 2012;Shah, Morofuji, Banks & Price, 2013). The known effects of CAIs on specific ion channels, aquaporins (Kamegawa, Hiroaki, Tani & Fujiyoshi, 2016), or other receptors which interact with Ab on mitochondrial or cell membranes may also be mechanisms responsible for the amelioration of Ab-induced mitochondrial dysfunction (Aggarwal et al., 2013). More studies will be needed to further clarify these molecular mechanisms.
Our results suggest a new and critical role for CA inhibition in the regulation of Ab-induced neuronal and microvascular toxicity, both essential underlying processes of AD etiopathology, through an effect of the CAIs on specific pathways of mitochondrial dysfunction. Importantly, the mitochondrial effects are exerted not only by MTZ, but also by another member of the CAIs family, ATZ, which is effective at even lower concentrations. These concentrations are in the range of those achieved clinically in the brain. Although clinical trials will be required to ultimately demonstrate effects in AD or other dementias, the translatability of our findings will be increased if the protective effects of CAIs are demonstrated in transgenic animal models of amyloidosis. We have previously shown reduced caspase activation and neuronal death after Ab intracerebral injection in mice treated with a concentration of MTZ of 10 mg/kg, that after allometric scaling is significantly under the maximum recommended dose for human adults (Fossati et al., 2016). Promising studies in transgenic mouse models of amyloidosis are currently ongoing in our laboratory. The physiological relevance of this approach is further highlighted by recent proteomic studies showing increased CAII in the mitochondria in aging and neurodegeneration (Pollard et al., 2016), as well as by our group's recent findings demonstrating the presence of multiple CA enzymes in amyloid plaques within the AD human brain (Drummond et al., 2017).
This study, clarifying for the first time the mitochondrial molecular mechanisms of MTZ and ATZ protection against Ab toxicity, paves the way for future clinical trials aimed to repurpose these and other FDA-approved CAIs against AD and dementia.

| Cell cultures
The neuroblastoma SH-SY5Y cell line was purchased from the Amer-

| Ab peptides
Synthetic Ab42 was synthesized using N-tert-butyloxycarbonyl chemistry by James I. Elliott at Yale University and purified by reverse-phase high-performance liquid chromatography on a Vydac  (Fossati et al., 2010), and lyophilized.
Peptides were subsequently dissolved in DMSO to a 10 mM concentration, followed by the addition of deionized water to 1 mM concentration and by further dilution into the medium in which the experiments were run, to a final concentration of 10 lM in the case of Ab42 and to 50 lM in the case of the Ab40-Q22. Peptide treatments were performed in EBM-2 supplemented with FBS 1% and in DMEM:F12 with no FBS or 1% FBS, for ECs and SH-SY5Y cells, respectively.

| Peptide preparation
In brief, HFIP-pretreated and lyophilized peptides were resuspended in DMSO to a 10 mM concentration, immediately prior to use. After that, peptides were directly used in the case of the monomeric preparations and added to the cells at the specified concentrations.
Peptide pre-aggregation to obtain oligomeric and fibrillar preparations was performed following the protocol published by Dahlgren et al. (2002). Aggregates were then added to the cells at the desired concentrations, alone or in cotreatment with the different CAIs.
Monomeric, oligomeric, or fibrillary state of the preparations was tested by EM as we previously published (Fossati et al., 2010).

| Amplex Red
Cells were plated on 6-well plates and treated with the peptides and/or CAIs for the experimental times. Plates were centrifuged for 10 min at 200 RCF and washed twice with PBS 19. Afterward, to isolate the mitochondria, 500 ll of homogenization buffer (75 mM sucrose, 225 mM mannitol, 5 mM Tris-HCl pH = 7.4, 1 mM PMSF, and protease inhibitor cocktail) was added. Cells were scraped in this buffer, collected in glass tubes, and grinded exactly 80 times with a pellet pestle, keeping everything always on ice. Cells were then centrifuged at 800 RCF for 5 min at 4°C. The supernatant was collected and centrifuged again at 800 RCF for 5 min at 4°C. The supernatant was again collected and centrifuged at 11700 RCF for another 5 min at 4°C. After this, the supernatant was discarded, and the pellet was resuspended in 100 ll of homogenization buffer.
The samples were centrifuged again at 11700 RCF for 10 min at

| Mitochondrial and cytoplasmic calcium assay
This assay was performed as previously published in F. In brief, cells were plated on 25-mm optical borosilicate poly-L-lysine-coated sterile glass covers, (Thermo Fisher, Waltham, Massachusetts, USA) at a 70% confluence. 24 hr later, cells were loaded with 5 lM Rhod-2 AM or 2.5 lM Fluo-4 AM on HBSS for 45 min. Then, cells were washed twice with HBSS and incubated for an additional 15 min on fresh HBSS, without Rhod-2 AM or Fluo-4 AM. After that, HBSS was replaced again by fresh HBSS, glasses were mounted on microscopy chambers, and experiments were conducted, after adding the pre-aggregated peptides and/or the CAIs. Cells were mounted on Sykes-Moore chambers (BellCo, Vineland, New Jersey, USA) and imaged every 15 s for 45 min, at a 209 magnification, using a Nikon fluorescent microscope (Chiyoda, Tokyo, Japan). The calcium ionophore ionomycin (5 lM) was added at the end of the experiments to test whether the cells were still functional and able to maintain calcium gradient between cytoplasm and extracellular media after 45 min of experiment and to estimate the degree of maximal fluorescence (Supporting Information, Figure S2). Images were analyzed using NIS-Elements and ImageJ software. Specifically, at least 10 ROIs per field were marked in mitochondrial (Rhod-2) and cytoplasmic areas (Fuor-4) in all the images, and a time measurement of the intensity in both fluorophores was conducted. The fluorescence for each time point was normalized to the fluorescence value measured at time = 0 s. and expressed as a percentage of this initial fluorescence. Results of typical experiment are shown in Supporting Information, Figure S2. Please note that under these experimental conditions, a spontaneous moderate increase on Fluo-2 and Rhod-2 fluorescence was observed, even under control conditions (Supporting Information, Figure S2a-b). Due to the time frame of these experiments, the peptide was pre-aggregated as described above to obtain oligomers before being added to the cells.

| pH measurement
Cells were plated on 96-well fluorescence microtiter microplates (Thermo Fisher Scientific). The day after, cells were treated with the different CAIs and/or peptides. After treatment, intracellular pH measurement was performed using the pHrodo red AM kit (Thermo Fisher Scientific), following the indications provided by the manufacturer.

| ATP measurement
Steady-state levels of ATP were estimated using the Cell Titer-Glo, according to manufacturer's instructions. In brief, 5 9 10 3 cells per well of both ECs and SH-SY5Y cells were plated on 96-well plates.
The day after, cells were treated with either 10 lM Q22 or 50 lM Ab42 peptides, respectively. Peptides were added alone or in combination with increasing concentrations of MTZ (100 and 300 lM) or ATZ (10 and 100 lM). Controls were treated with either DMSO or 5 lM oligomycin. After 3 hr of treatment, the plates containing the cells were equilibrated to room temperature for 30 min prior to addition of the luciferin/luciferase/cell lysis mixture. Absolute luminescence from quadruplicate experiments was recorded using a SpectraMax plate reader (Molecular Devices, Sunnyvale California, USA).

| Statistical analysis
Statistical significance of differences between groups was determined by Student's t test or two-tailed Student's t test. Moreover, in experiments containing more than two groups, the statistical significance was determined by ANOVA with turkey post-hoc test. For the statistical analysis and the graphical representation, we used Origins Lab (Northampton, Massachusetts, USA) software and GraphPad Prism (GraphPad, La Jolla, CA). Values of p ≤ 0.05 were considered significant. (* p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001).

ACKNOWLEDGMENTS
We sincerely thank Drs. Jorge Ghiso and Agueda Rostagno for sharing experimental and space resources during the first period of this study, and Dr. Erik R. Swanson for providing us with N-methyl acetazolamide, the structural analog of ATZ. This work was supported by grants from the American Heart Association 13SDG16860017, the Alzheimer's Association NIRG-12-240372, the Leon Levy Fellowship in Neuroscience, the Blas Frangione Foundation to SF, NIH grants AG008051 and NS073502 to TW and NIH grants AG13616, AG022374, AG12101, AG057570 awarded to MdL. MES was partly a fellow from CIEN-Reina Sofia Foundation (Carlos III Health Institute, Spanish Ministry of Economy).