Deregulation of mitochondrial F1FO-ATP synthase via OSCP in Alzheimer’s disease

F1FO-ATP synthase is critical for mitochondrial functions. The deregulation of this enzyme results in dampened mitochondrial oxidative phosphorylation (OXPHOS) and activated mitochondrial permeability transition (mPT), defects which accompany Alzheimer’s disease (AD). However, the molecular mechanisms that connect F1FO-ATP synthase dysfunction and AD remain unclear. Here, we observe selective loss of the oligomycin sensitivity conferring protein (OSCP) subunit of the F1FO-ATP synthase and the physical interaction of OSCP with amyloid beta (Aβ) in the brains of AD individuals and in an AD mouse model. Changes in OSCP levels are more pronounced in neuronal mitochondria. OSCP loss and its interplay with Aβ disrupt F1FO-ATP synthase, leading to reduced ATP production, elevated oxidative stress and activated mPT. The restoration of OSCP ameliorates Aβ-mediated mouse and human neuronal mitochondrial impairments and the resultant synaptic injury. Therefore, mitochondrial F1FO-ATP synthase dysfunction associated with AD progression could potentially be prevented by OSCP stabilization.

C linical observations have suggested that mitochondrial dysfunction is among the earliest manifestations of Alzheimer's disease (AD) and constitutes a hallmark pathological feature of this neurological disorder 1,2 . Previous studies have also identified impaired mitochondrial oxidative phosphorylation (OXPHOS) as a feature of mitochondrial defects in AD individuals 2,3 and AD animal models [4][5][6] . Compromised mitochondrial OXPHOS efficiency results in lowered mitochondrial bioenergetics and exaggerated production of free radicals 7 . Indeed, ATP deficiency and oxidative damage are characteristics of brains from AD patients 2 . Impaired mitochondrial OXPHOS efficiency is closely associated with dysfunction of mitochondrial respiratory enzymes, including mitochondrial complex I to IV, as well as the defect of F1FO-ATP synthase. Previous studies have mostly focused on the dysfunction of mitochondrial complex IV in AD [8][9][10] . However, this concept has been recently challenged [11][12][13] . In fact, in addition to mitochondrial respiratory enzyme defects, increasing evidence implicates the dysfunction of mitochondrial F1FO-ATP synthase in AD [14][15][16][17] .
The mitochondrial F1FO-ATP synthase, which includes three components (F1, FO and the peripheral stalk), is a critical mitochondrial OXPHOS enzyme involved in the regulation of mitochondrial ATP production and in the maintenance of the mitochondrial membrane potential 18,19 . The F1FO-ATP synthase can both synthesize ATP and degrade ATP when operating in reverse to generate proton backflow, increasing mitochondrial membrane potential when it is critically low [19][20][21] . In addition to its vital function in mitochondrial OXPHOS, recent studies have shown that this enzyme contributes to the formation of the mitochondrial permeability transition pore (mPTP) 22,23 through the interaction of its oligomycin sensitivity conferring protein (OSCP) subunit with cyclophilin D (CypD), the key regulator of mPTP [22][23][24] . Extensive formation of mPTP is a severe mitochondrial pathological event that leads to collapsed mitochondrial membrane potential (mDC), decreased mitochondrial OXPHOS capacity, elevated reactive oxygen species (ROS) generation and, eventually, cell death 25 . Indeed, mPTP activation is thought to be a key mechanism of mitochondrial stress in AD and has been proposed to underlie its characteristic synaptic dysfunction and cognitive decline 5,26 . Given its role in mitochondrial OXPHOS 18 and mPTP formation [21][22][23] , the deregulation of mitochondrial F1FO-ATP synthase may predispose to compromised OXPHOS efficiency and sensitized mPTP formation, which are two hallmark mitochondrial defects in AD. However, to date information on the dysfunction of F1FO-ATP synthase in AD has remained limited. Accordingly, the underlying molecular mechanisms causing the defect of this enzyme in AD remain unresolved.
In this study, we compare the levels of major F1FO-ATP synthase subunits in the brains from AD individuals, mild cognitive impairment (MCI) patients and non-AD control subjects, and find a selective decrease in the levels of OSCP during the progression of AD. We also find that, in a mouse model of AD that overexpresses the human form of amyloid beta (Ab), the loss of OSCP is more prominent in synaptic mitochondria. In addition to OSCP loss, we also detect a direct physical interaction between OSCP and Ab in the brains from AD cases, as well as in AD mice. Such OSCP aberrations disrupt F1FO-ATP synthase stability, leading to severe mitochondrial dysfunction and synaptic injury. Further in vivo studies show the deleterious impact of F1FO-ATP synthase dysfunction on the development of mitochondrial defects in AD mice. Importantly, the restoration of OSCP ameliorates the Ab-mediated mitochondrial dysfunction and synaptic injury in mouse or human neurons, further supporting the role of OSCP deregulation in mitochondrial dysfunction under Ab-rich conditions. Therefore, mitochondrial F1FO-ATP synthase dysfunction that results from OSCP aberrations may constitute a primary AD event that can be prevented by OSCP protection, suggesting OSCP as a potential new therapeutic target for AD.

Results
Loss of OSCP in AD subjects and 5xFAD mice. To examine changes of F1FO-ATP synthase in AD brains, we used immunoblotting to compare the expression levels of the major subunits of the mitochondrial F1FO-ATP synthase in protein extracts from the temporal lobe of non-AD, MCI, and AD subjects (Supplementary Table 1). Among the F1FO-ATP synthase subunits tested here, OSCP was slightly reduced in MCI brains, while this decrease was exaggerated in AD patients (Fig. 1a). Notably, the expression levels of the other major F1FO-ATP synthase subunits, including a, b, c, a, and b were not significantly changed in either MCI or AD brains ( Supplementary Fig. 1a,b), implying that the OSCP loss is not likely a result of F1FO-ATP synthase reduction in AD. Next, we performed histological studies on brain sections from the temporal lobes of AD and control subjects. Our results showed a dramatic reduction of OSCP expression in neurons from the temporal lobe of AD individuals (Fig. 1b). A similar reduction of OSCP also occurred in an AD mouse model (5xFAD mice). To determine whether synaptic mitochondria are more vulnerable to OSCP loss under conditions of elevated Ab, we separated synaptic mitochondria and nonsynaptic mitochondria from 5xFAD mice at 4 and 9 months old that mimic the amyloidopathy and behavioral changes typical of the early-, and middle to late-stages of AD, respectively 27 . The purity of isolated mitochondria was verified by means of the detection of high abundance of the mitochondrial protein voltage-dependent anion channel (VDAC) without contamination of b-actin or synaptic vesicles ( Supplementary Fig. 2). Synaptic mitochondria demonstrated a significant OSCP decrease in young 5xFAD mice, and this OSCP loss was exacerbated with age ( Fig. 1c). However, a significant OSCP reduction in nonsynaptic mitochondria was detected only in old 5xFAD mice (Fig. 1d), suggesting that in Ab-rich environments synaptic mitochondria are more susceptible to OSCP loss. In sharp contrast, the expression levels of a, b and c, as well as a and b subunits remained unaltered in either synaptic or nonsynaptic mitochondria from 5xFAD mice at any tested age (Supplementary Fig. 3a-d), indicating that OSCP expression is selectively suppressed in parallel with the manifestation of AD-like symptoms in 5xFAD mice.
OSCP loss impacts mitochondrial and synaptic function. To investigate the impact of OSCP loss on neuronal mitochondrial function we genetically downregulated OSCP expression in primary cultured mouse neurons by using specific OSCP shRNA. Nontarget shRNA was used as control. Quantitative analysis showed substantially decreased OSCP levels in OSCP knockdown neurons (OSCP KD, Fig. 2a). The key functions of mitochondrial F1FO-ATP synthase are to maintain mDC and to produce ATP. To determine the impact of OSCP loss on F1FO-ATP synthase function, we first measured mDC in OSCP knockdown neurons via staining with tetramethylrhodamine methyl ester (TMRM), an indicator of mDC (ref. 5). In comparison with control neurons, OSCP knockdown neurons exhibited significantly decreased intensities of TMRM (Fig. 2b), suggesting a collapsed mDC in response to OSCP loss. Furthermore, in agreement with reduced mDC, ATP generation was markedly reduced in OSCP knockdown neurons (Fig. 2c), suggesting deregulation of mitochondrial OXPHOS efficiency. In view of the close correlation between compromised OXPHOS efficiency and mitochondrial ROS production, we next compared mitochondrial ROS levels by staining with MitoSox Red, a specific fluorescent indicator of mitochondrial superoxide abundance 5,6 . Our results showed significantly elevated MitoSox Red intensity in mitochondria from OSCP knockdown neurons (Fig. 2d). In addition, in comparison to the control neurons OSCP downregulated neurons also displayed significantly reduced mitochondrial population in neurites (Fig. 2e).
OSCP deregulation has also been previously linked to mPTP formation [22][23][24] . To determine whether the loss of OSCP sensitizes mPTP opening, we subjected neurons to calcein AM-cobalt chloride quenching assay 28 . When we exposed the neurons to ionophore, which increases mitochondrial calcium overloading to trigger mPTP formation 29 , OSCP knockdown neurons exhibited larger reductions in mitochondrial calcein intensity than control neurons (Fig. 2f). CypD is a known regulator of mPTP 25 . Inhibiting CypD pharmacologically with cyclosporin A (CsA) substantially suppressed mPTP formation in control neurons; however, mPTP activation associated with OSCP downregulation was not blocked by the addition of CsA (Fig. 2f). These results seem to implicate that mPTP activation induced by loss of OSCP does not rely on the functional status of CypD. This was further supported by unaltered CypD expression levels in OSCP knockdown neurons detected by immunoblotting in parallel with calcein AM-cobalt chloride quenching assay ( Supplementary Fig. 4a). Furthermore, by downregulating OSCP in CypD-deficient neurons, we found that mPTP activation resulting from OSCP deficiency was not protected by depletion of CypD ( Supplementary Fig. 4b,c), which is in agreement with the abolished effect of CsA treatment (Fig. 2f). Taken together, these data suggest that the loss of OSCP sensitizes mPTP opening regardless of the functional status of CypD.
Because the major function of OSCP is to stabilize the F1FO complex by interacting with F1, we tried to further probe the association between OSCP loss-of-function and mPTP formation in neurons by disrupting the interaction of OSCP with F1. To this end, we downregulated the b-subunit, which is the major F1 component that interacts with OSCP in the integral F1FO-ATP synthase 30 (Supplementary Fig. 5a). Neurons in which the b-subunit was knockeddown demonstrated significantly suppressed mDC and reduced ATP generation ( Supplementary  Fig. 5b,c) in patterns similar to those that we observed in OSCP-deficient neurons (Fig. 2b,c). Importantly, b-subunit knockdown neurons displayed substantial mPTP activation ( Supplementary Fig. 5d). These results indicated that increased mPTP formation is closely associated with F1FO complex disruption, further highlighting a key role of OSCP loss-offunction in triggering mPTP opening. These findings also support previous reports that increased F1 dissociation from the F1FO complex contributes to mPTP formation 23,31,32 .
Mitochondria play a critical role in maintaining synaptic and neuronal function 26,33 . To determine whether loss of OSCP induces synaptic dysfunction, we measured synaptic density by . OSCP shRNA-treated neurons showed a significant reduction in synaptic density (Fig. 3a), suggesting a deleterious influence of OSCP loss-associated mitochondrial dysfunction on synaptic function. To directly determine the influence of OSCP loss on synaptic transmission we performed whole-cell voltage-clamp recordings to examine changes in action potentialindependent miniature excitatory postsynaptic currents (mEPSCs). The downregulation of OSCP significantly reduced the average mEPSC amplitude in OSCP knockdown neurons in comparison to their nontarget shRNA-treated controls (Fig. 3b,d,e), without significantly altering mEPSC frequency (Fig. 3b,c). Furthermore, direct stimulation of postsynaptic glutamate receptors by 'puffing' glutamate onto cultured neurons resulted in currents of significantly smaller amplitude in OSCP knockdown neuron (Fig. 3f). Taken together, these data suggest that OSCP loss impairs synaptic function, which parallels the well-documented synaptic dysfunction, and particularly a loss of postsynaptic function, in the pathology of AD 34 . Energy deprivation, oxidative stress and mPTP formation are hallmarks of mitochondrial defects in AD brains 2,5,35,36 . Synaptic loss and impaired synaptic transmission further characterize AD-related pathological changes 34,[37][38][39] . Therefore, our results suggest that reduced OSCP expression is associated with AD-like mitochondrial dysfunction and the resultant synaptic failure.

Interaction of OSCP with Ab results in OSCP loss-of-function.
Ab is a key mediator of AD and previous studies have shown that Ab deposits in AD mitochondria, targeting several mitochondrial proteins 2,5,35,40 . We therefore explored whether OSCP is a binding partner of Ab in AD brains by co-immunoprecipitating temporal lobe protein extracts using anti-OSCP antibody. We detected OSCP/Ab interaction in AD brains (Fig. 4a), as well as in MCI brains that showed brain Ab deposition ( Supplementary  Fig. 6). In contrast, non-AD brains did not exhibit the OSCP/Ab complex (Fig. 4a). The OSCP/Ab complex was not detected in AD brain tissue when the OSCP antibody was replaced by nonimmune IgG (Fig. 4a), validating the specificity of OSCP/Ab binding. Similarly, the OSCP/Ab complex was also found in 5xFAD mice by co-immunoprecipitation (Fig. 4b). Further confocal microscopy studies showed extensive colocalization of OSCP and Ab in neocortex and hippocampus from 5xFAD mice (Fig. 4c), suggesting an interaction between OSCP and Ab in vivo.
To determine whether OSCP directly interacts with Ab we next performed an in vitro pull-down assay using glutathione S-transferase (GST)-tagged OSCP as the bait protein.  Physical binding of Ab could result in functional changes of its target proteins 2 . Therefore, it is important to determine whether the Ab interaction influences OSCP function by tethering F1 via binding with aand b-subunits 41 . To address this question, GST-tagged OSCP was pre-incubated with or without Ab1-42. After washing off unbound Ab1-42, we conducted in vitro pull-down assays and found that the binding of OSCP to aor b-subunits was significantly reduced by Ab1-42, suggesting that the interplay between OSCP and Ab1-42 disrupts the ability of OSCP to bind to the F1 entity (Fig. 4f). To further evaluate the association between Ab and impaired OSCP function, we generated deleted forms of OSCP based on the predicted Ab binding sites on OSCP ( Supplementary Fig. 7a-c). We found that the ability of OSCPD107-121 to interact with Ab was significantly reduced (Fig. 4e). Furthermore, OSCPD107-121 showed extensive binding with aor b-subunits regardless of the presence or absence of Ab1-42 (Fig. 4f). These results further indicate a direct and specific influence of Ab on OSCP, leading to reduced mitochondrial bioenergetics and activated mPTP formation.
To examine the impact of Ab on OSCP binding to the F1FO-ATP synthase within mitochondrial membranes we exposed brain mitochondria to Ab1-42. As predicted, co-immunoprecipitation showed OSCP/Ab complexes in Ab-exposed mitochondria, indicating that Ab1-42 can enter mitochondria and bind to OSCP within the F1FO-ATP synthase ( Supplementary Fig. 8a). Next, we examined F1FO-ATP synthase stability by conducting blue-native PAGE (BN-PAGE) 42,43 . Mitochondria were exposed to vehicle-, Ab1-42-or Ca 2 þtreatment. Ca 2 þ is thought to disrupt the F1FO complex through OSCP 32 . Ab induced a marked increase in F1 dissociation from the F1FO complex, similar to Ca 2 þ -treatment, while little OSCP was detected in the free F1 ( Supplementary Fig. 8b,c). The identification of F1FO dimer, monomer and F1 by b subunit blots was further validated by using an antibody against F1 ( Supplementary Fig. 8d). Together these data suggest that the Ab-induced F1FO complex instability is at least partly due to a reduction in OSCP's ability to hold F1 and FO together. As expected, Ab1-42-treated mitochondria displayed mitochondrial OXPHOS inhibition as evident from the significantly suppressed ATP synthesis ( Supplementary Fig. 8e). To dissect the influence of Ab on mitochondrial OXPHOS enzymes, we measured the enzymatic activities of mitochondrial complexes I through IV and OSCP aberrations and 5xFAD mouse mitochondrial dysfunction. We hypothesized that if the above effects of OSCP loss and OSCP/Ab interplay on F1FO-ATP synthase function could be extrapolated to an in vivo setting, 5xFAD mice would display F1FO-ATP synthase dysfunction, particularly in synaptic mitochondria, which demonstrated early and extensive OSCP reduction, as well as Ab deposition ( Supplementary Fig. 9a,b). To address this, we first measured mitochondrial OXPHOS function in 5xFAD mitochondria. Although nonsynaptic mitochondria from 5xFAD mice also demonstrated an age-and genotypespecific effect ( Supplementary Fig. 10a-c), synaptic mitochondria from 5xFAD mice exhibited an early and marked decrease in the mitochondrial respiratory control ratio, ATP synthesis and F1FO-ATP synthase catalytic activity (Fig. 5a-c). In contrast, mitochondrial complex IV, whose deactivation is thought to be the major OXPHOS defect in AD 8-10 , exhibited only a relatively mild decrease ( Supplementary Fig. 10g,h), again suggesting that F1FO-ATP synthase deregulation contributes to the extensive OXPHOS suppression in 5xFAD mice. Next, we examined F1FO complex proton-flow coupling, which reflects F1FO complex integrity 44,45 . Synaptic mitochondria from 5xFAD mice demonstrated a significant increase in oligomycin-insensitive respiration (that is, uncoupled electron transport, Fig. 5d). This was further supported by a F1FO-ATP synthase coupling assay which showed that 5xFAD synaptic mitochondria had markedly blunted sensitivity to oligomycin-A inhibition (Fig. 5e,f). Such changes were also seen in nonsynaptic mitochondria, but the effect size was considerably smaller (Supplementary Fig. 10d-f). These results imply that F1FO complex destabilization occurs in 5xFAD brain mitochondria. Direct evidence of F1FO complex instability in 5xFAD mitochondria was collected using BN-PAGE. Free F1 was identified by immunoblotting using specific antibodies recognizing the b subunit as previously described 42 . We found an age-dependent increase in F1 dissociation (Fig. 5g,h), confirming the destabilization of the F1FO complex in 5xFAD mitochondria. Lastly, because mPTP formation is thought to be a consequence of F1FO complex uncoupling via OSCP deregulation, we measured mPTP Results shown are representatives from three non-AD and three AD patients. Ab1-42 peptide and AD brain extracts were used as positive controls for Ab immunoreactive bands. (b) Co-immunoprecipitation of OSCP and Ab in 5xFAD mouse neocortex. Results shown are representatives from three mice in each group. Ab1-42 peptide was used as a positive control for Ab immunoreactive bands. (c) Colocalization of OSCP (green) and Ab (red) in neocortex and hippocampus from 5xFAD mice. Neurons were identified by staining for NISSL (blue). Scale bar, 10 mm. (d) OSCP and Ab interaction determined by an in vitro pull-down assay. (e) Amino-acid residue numbers are given for mature OSCP protein (with known mitochondrial signalling peptide removed). SP is the mitochondrial signalling peptide. Wild-type OSCP and OSCPD107-121 were used for pull-down assay. OSCPD107-121 displayed lowered capacity to interact with Ab compared to wild-type OSCP. (f) In vitro pull-down assay showed that Ab1-42 suppresses the ability of OSCP, but not OSCPD107-121, to bind aand b-subunits as demonstrated by decreased aand b-immunoreactive bands. n ¼ 6-10 samples per group. Error bars represent s.e.m.
formation susceptibility and found that 5xFAD synaptic mitochondria exhibited a significantly increased response to Ca 2 þ -induced mitochondrial swelling ( Supplementary Fig. 10i), which is in agreement with our previous findings 33 . Therefore, these data show a strong correlation between OSCP deregulation and mitochondrial dysfunction in AD-relevant pathophysiological settings.
Protection of OSCP restoration on Ab-exposed mouse neurons. To further address the role of OSCP deregulation for the induction of mitochondrial dysfunction in an Ab-rich environment we examined whether the Ab-mediated mitochondrial dysfunction can be attenuated by OSCP restoration. Therefore, we overexpressed OSCP in mouse neurons, aiming to restore OSCP levels and to dilute the OSCP/Ab interaction which would be expected to protect OSCP function against Ab toxicity. The control and OSCP overexpressing (OSCP OE) neurons were exposed to a treatment with oligomeric Ab1-42. By immunoblotting and immunostaining (Fig. 6a,b), we found that Ab-mediated OSCP reduction in control neurons was significantly ameliorated by OSCP upregulation. Further mitochondrial functional assays showed that OSCP overexpression substantially attenuated the decreased mDC, lowered ATP production and decreased neuritic mitochondrial population as well as sensitized mPTP formation in Ab-treated neurons ( Fig. 6c-f). Notably, OSCP overexpression by itself did not significantly alter the expression level of CypD ( Supplementary  Fig. 11). To investigate the specific protection of OSCP restoration against Ab toxicity on neuronal mitochondria we overexpressed the F1FO-ATP synthase b-subunit, which forms the catalytic core of the F1FO-ATP synthase. We then treated the control and b-subunit-overexpressing (b-subunit OE) neurons with Ab. Ab treatment did not induce detectable changes in the expression levels of b-subunit in either control or b-subunit OE neurons when compared with their vehicle-treated counterparts ( Supplementary Fig. 12a), which supports our findings of the unaltered b-subunit levels in AD subjects ( Supplementary  Fig. 1a,b) as well as in 5xFAD mice ( Supplementary Fig. 3a,c). However, further experiments showed that the overexpression of b-subunit did not demonstrate significant protection against Abinduced mDC collapse, ATP reduction or decrease in neuritic mitochondrial population, as well as mPTP activation ( Supplementary Fig. 12b-e), further supporting a specific protective effect of OSCP restoration in AD-related conditions. We used exogenous Ab treatment to mimic the high level of Ab over-production in AD subjects; however, it is not clear whether OSCP restoration has similar protective effects in neurons that generate endogenous Ab. To address this question we used primary neuron cultures from 5xFAD mice. In comparison to nonTg neurons, cultured 5xFAD neurons demonstrated a significant loss of OSCP expression, which was prevented by OSCP overexpression (Supplementary Fig. 13a). Importantly, OSCP overexpression in 5xFAD neurons exhibited protective effects on mDC, ATP production and neuritic mitochondrial population, as well as mPTP formation Aβ b e f ( Supplementary Fig. 13b-e) against endogenous Ab toxicity in similar patterns as we observed in neurons treated with exogenous Ab, confirming the protective effects of OSCP restoration against Ab toxicity. Given the crucial role of mitochondria in sustaining synaptic transmission and plasticity 46 , and the attenuation of mitochondrial defects by OSCP overexpression, we next examined whether OSCP restoration also protects synaptic function against Ab toxicity. Control and OSCP OE neurons were exposed to vehicle-or oligomeric Ab1-42-treatment before the synaptic density was analysed using immunofluorescent staining of PSD95 and vGlut1. Oligomeric Ab1-42-treated control neurons displayed a significant reduction in synaptic density in comparison to control neurons receiving vehicle treatment (Fig. 7a). In sharp contrast, the Ab-induced synaptic loss was significantly ameliorated by OSCP overexpression (Fig. 7a). OSCP overexpression by itself did not affect baseline levels of synaptic density (Fig. 7a). In addition, OSCP overexpression protected against the reduction in mEPSC amplitude induced by Ab toxicity (Fig. 7b,d,e). The relatively preserved mEPSC frequency of Ab-treated neurons (Fig. 7b,c) may reflect distinct time-dependent changes in pre-and post-synaptic function during acute Ab treatment 47 , and/or may result from presynaptic calcium accumulation due to mPTP activation 48 , which could affect asynchronous neurotransmitter release and thus mEPSC frequency 49 . Moreover, in OSCP overexpressing neurons the postsynaptic response to glutamate stimulation was indistinguishable from that in control neurons even in the presence of exogenous Ab (Fig. 7f). Taken together, our results suggest that the restoration of OSCP protects neuronal mitochondrial and synaptic function from Ab toxicity.
Protection of OSCP restoration on Ab-exposed human neurons. Because AD is a human disease, it is of considerable interest to know whether OSCP restoration confers similar protection to mitochondria in Ab-exposed human neurons. Therefore we derived human neurons from human neural stem cells (Supplementary Fig. 14) and treated them with oligomeric Ab1-42. Immunobloting for OSCP levels and co-immunoprecipitation of OSCP and Ab revealed that the OSCP loss and the formation of the OSCP/Ab complex (Fig. 8a,b) observed in AD individuals were mirrored in Ab-treated human neurons. Next, OSCP was overexpressed in human neurons (Fig. 8c). Control and OSCP OE human neurons were then exposed to oligomeric Ab1-42 at 500 nM and then processed for our assays of mitochondrial function. Our results showed that Ab-induced mitochondrial dysfunctions in ATP production, mDC and neuritic mitochondrial population, as well as mPTP formation (Fig. 8d-g) were substantially reduced by OSCP overexpression in human neurons. Moreover, the overexpression of OSCP in human neurons did not have a detectable influence on baseline levels of the assayed parameters in comparison to control neurons ( Fig. 8d-g), which further supports our observations in OSCP OE mouse neurons.

Discussion
In this study, we find that OSCP loss and OSCP/Ab interaction constitute the major OSCP alterations in the brains from AD patients and 5xFAD mice. Among the major subunits of F1FO-ATP synthase, OSCP is selectively decreased in the brains of AD subjects and 5xFAD mice. Moreover, the early decrease of OSCP expression in synaptic mitochondria from 5xFAD mice indicates that synaptic mitochondria are more vulnerable to Ab-induced mitochondrial alterations. This reveals a novel form of synaptic mitochondrial stress in AD and supports a causative role of synaptic mitochondrial dysfunction in the development of early synaptic dysfunction in the disease 33 . Another critical finding of this study is the interaction of OSCP and Ab. In recent years, the accumulation of Ab in mitochondria, and particularly in synaptic mitochondria, has received considerable attention 33 . Ab is probably transported via a translocase in the outer mitochondrial membrane 50 and impacts mitochondrial function via multiple pathways, including the interaction of Ab with several mitochondrial proteins such as CypD 5 , Amyloid betabinding alcohol dehydrogenase 35 and dynamin-like protein 1 (ref. 40). Our finding of significant interplay between OSCP and Ab furthers our understanding of the intracellular influence of Ab toxicity in AD. Indeed, given the extensive Ab accumulation in mitochondria, we cannot exclude the possibility that mitochondrial Ab may bind to other proteins embedded in the inner mitochondrial membrane or to other subunits of mitochondrial F1FO-ATP synthase, which could also be involved in compromised F1FO-ATP synthase function in AD. In view of  the critical role that the b-subunit plays for the catalytic functions of F1FO-ATP synthase, we examined whether the b-subunit is a binding partner of Ab. Our data (not shown) suggest that the b-subunit is not a likely binding partner of Ab. Future investigation will explore whether Ab may affect other mitochondrial proteins that contribute to mitochondrial F1FO-ATP synthase dysfunction, as well as mPTP activation in AD. Our findings of reduced OSCP levels and the interaction of Ab with OSCP, which impair mitochondrial function in AD-related pathological settings, implicate that Ab toxicity induces OSCP aberrations, leading to mitochondrial F1FO-ATP synthase dysfunction in AD-sensitive brain areas. However, increasing evidence suggests that AD is a systemic disease. In addition to AD-affected brain areas, Ab is detected in circulating blood 51 and platelets 52,53 , as well as in other tissues such as intestine and kidney 54 . The systemic distribution of Ab raises the possibility of Ab interaction with OSCP in cells outside the nervous system, which may potentially cause mitochondrial dysfunction. Moreover, in view of the complexity of mitochondrial abnormalities in AD our findings of OSCP aberrations may only provide one mechanism for systemic mitochondrial dysfunction. Therefore, it will be important for future studies to evaluate the impact of OSCP alterations in AD on a systems level.
Suppressed mitochondrial OXPHOS is a hallmark of mitochondrial defects in AD 2,3 . However, the detailed mechanisms of this mitochondrial deficit in AD are not fully understood. It is well accepted that AD is accompanied by pronounced changes in the function of mitochondrial complex IV (refs 8-10). Indeed, we have found a significant decrease of complex IV activity in 5xFAD neuronal mitochondria when comparing it to the corresponding complex IV activity in nonTg mice. However, the contribution of complex IV dysfunction to ATP deficiency in AD is still controversial [11][12][13] . Given the redundant capacity of complex IV in mitochondria, it has been argued that the slight decrease in complex IV activity does not explain severe ATP deprivation in AD 11 , suggesting that other mitochondrial OXPHOS enzymes also become dysfunctional. F1FO-ATP synthase is a critical OXPHOS enzyme synthesizing ATP 18,19 . To date, our knowledge on the functional state of mitochondrial F1FO-ATP synthase in AD is extremely limited. Earlier studies suggested that F1FO-ATP synthase is not involved in AD because the F1FO-ATP synthase enzyme activity in the brains or platelets of AD patients 55 appeared to be unchanged, and increased enzymatic activity was observed in platelets from probable AD subjects 56 . However, this concept has been challenged in recent years by evidence that showed alterations of this enzyme in AD-sensitive brain regions, and particularly in neurons [14][15][16][17] . Indeed, as we show in this study, synaptic mitochondria are more vulnerable to Ab-induced F1FO-ATP synthase dysfunction in 5xFAD mice, whereas nonsynaptic mitochondria are relatively preserved. Further, we show that F1FO complex uncoupling is a prominent defect in 5xFAD mouse brain mitochondria. The coupling state of this enzyme in AD, however, was largely overlooked in previous studies, which failed to fully evaluate F1FO-ATP synthase deregulation in AD. Therefore, our results showing that F1FO-ATP synthase deregulation is associated with suppressed neuronal mitochondrial OXPHOS efficacy provide a novel mitochondrial mechanism of neuronal stress in AD and impact our current understanding of mitochondrial OXPHOS deficits in this disease.
The molecular identity of mPTP has long been a critical scientific issue 23 . Recently, uncoupled mitochondrial F1FO-ATP synthase was identified as the molecular basis of mPTP 23 . Specifically, the prevalent model proposes that OSCP is the binding target of CypD, which is a non-pore forming regulator of mPTP opening. The interaction of CypD with OSCP triggers the instability of the F1FO complex, thus resulting in the dissociation of F1 and eventually the formation of a nonselective leak channel within c rings 22,24 . These findings have suggested a key role of OSCP deregulation in the induction of mPTP. However, whether this model applies to neuronal mitochondria has not been comprehensively investigated. By downregulating OSCP in neuronal mitochondria, we found substantially sensitized mPTP formation. Importantly, this OSCP loss-associated mPTP activation is indispensable of the function of CypD, thus serving as strong evidence of the role of OSCP deregulation for inducing mPTP formation in neurons. Our in vivo results further link mPTP over-activation to OSCP aberrations in AD-relevant pathological settings. Sensitized mPTP formation is well documented as a key mitochondrial dysfunction in AD. Our previous studies have shown that mPTP blockade by CypD depletion protects mitochondrial and neuronal stress in a mouse model of AD, suggesting that mPTP regulation could be a therapeutic strategy for AD treatment 5,26 . The physiological function of mPTP has been noticed in recent years [57][58][59] . Molkentin and colleagues 57 found increased susceptibility of heart failure in CypD-deficient mice outlining the risks associated with inhibiting the physiological functions of mPTP and CypD. Therefore, it would be preferable to develop a strategy that can reduce mPTP over-activation under pathological states like AD while preserving mPTP function under normal physiological conditions. In this study, we found that OSCP restoration reduces Ab-induced mPTP over-activation without having a detectable influence on baseline mPTP formation. Therefore, protecting OSCP seems to be a promising avenue for the regulation of mPTP for AD treatment.
Lastly, mitochondrial dysfunction has been identified as a causative factor of synaptic failure in AD. Therefore, supporting mitochondrial function seems to be a viable strategy to protect synaptic strength and plasticity to delay the cognitive decline in  Figure 9 | Schematic summary. With the progress of AD, brain mitochondria gradually undergo OSCP loss and Ab accumulation in AD-relevant conditions. OSCP loss and OSCP/Ab interaction impair OSCP function to keep F1FO complex integrity. This leads to severe mitochondrial dysfunction, including decreased ATP production, collapsed mitochondrial membrane potential, and increased ROS production and release, as well as the activation of mitochondrial permeability transition pore formation. Such mitochondrial deregulation results in dampened neuronal function and eventually neuronal death.
AD patients 1,2 . We found that loss of OSCP induces synaptic dysfunction, which was paralleled by reduced mitochondrial function. Notably, the restoration of OSCP mitigated Ab-mediated mitochondrial dysfunction and further preserved synaptic function as evidenced by our measures of synaptic density and synaptic transmission. Therefore, our results confirm the deleterious impact of mitochondrial injury on synaptic function and further highlight the therapeutic role of OSCP restoration for the protection of synaptic strength to ameliorate cognitive impairments in AD.
In summary, we have uncovered a novel mechanism of mitochondrial dysfunction mediated via OSCP disruption in an AD-relevant pathological setting. Moreover, our results indicate a role of mitochondrial F1FO-ATP synthase deregulation in the development of AD mitochondrial defects that has long been overlooked. However, other factors may also contribute to mitochondrial F1FO-ATP synthase dysfunction in AD. For example, a previous study has shown that the a-subunit interacts with extracellular domain of amyloid beta precursor protein (APP) and Ab on neuronal surface ATPase 60 . In addition, oxidative stress disrupts ATP synthase activity 19 , and ATP synthase a-subunit O-GlcNAcylation is decreased in AD-related conditions 61 . These changes may also potentially affect mitochondrial F1FO-ATP synthase. Thus future studies need to fully explore mitochondrial F1FO-ATP synthase dysfunction in AD. Nevertheless, the most parsimonious interpretation of our findings is that F1FO-ATP synthase deregulation via OSCP links mitochondrial defects to AD (Fig. 9) and constitutes a novel target for AD therapy.

Methods
Mice. Animal studies were approved by the University of Texas at Dallas Institutional Animal Care and Use Committee (IACUC) and were performed in accordance with the National Institutes of Health guidelines for animal care. 5xFAD mice overexpress a human form of mAPP-bearing mutations (SwFlLon) and PSEN1 mutations (M146L and L286V) linked to familial AD. CypD-deficient mice (B6;129-Ppif tm1Jmol /J, mixed gender) and 5xFAD mice (B6SJL-Tg(APPSwFlLon, PSEN1*M146L*L286V) 6799Vas/Mmjax, mixed gender) were obtained from Jackson Laboratory. Mice were allocated randomly to experimental groups for the various experimental measurements based on genotyping. Four and nine months old nonTg and 5xFAD mice of mixed genders were used in the experiments. Day 0-1 nonTg, 5xFAD and CypD-deficient pups were used for primary neuron culture. The investigators performing the experiments did not select the mice allocation. The genotype of all the transgenic animals were double checked by performing PCR and/or dot blots before the experiments. Mice with wrong genotyping were excluded. The number of mice was determined by our previous data and power calculation to ensure the minimal numbers of mice as needed were used in the experiments.
Human samples. Frozen brain samples and paraffin-embedded brain slices were requested from UT Southwestern Medical Center ADC Neuropathology Core, supported by ADC grant (AG12300) under a protocol approved by The UT Southwestern Medical Center with informed consent from all subjects and the study adhered to the Declaration of Helsinki principles.
Mitochondria preparation. Synaptic and nonsynaptic mitochondria were isolated from tissue as previously described, brain tissues were homogenized in ice cold isolation buffer (225 mM mannitol, 75 mM sucrose, 2 mM K2PO4, 0.1% BSA, 5 mM Hepes, 1 mM EGTA (pH 7.2)) with a Dounce homogenizer (Wheaton). The resultant homogenate was centrifuged at 1,300g for 3 min, and the supernatant was layered on a 3 Â 2-ml discontinuous gradient of 15, 23 and 40% (vol/vol) Percoll and centrifuged at 34,000g for 8 min (flying time) on Beckman Coulter ultracentrifuge (Optima XPN-90 Ultracentrifuge). After centrifugation, the interface between 15 and 23% (Band containing synaptosomes) was collected. Additionally, the interface between 23 and 40% (containing nonsynaptic mitochondria) was removed and collected. The fractions were then resuspended in isolation buffer containing 0.02% digitonin and incubated on ice for 5 min. The suspensions were then centrifuged at 16,500g for 15 min. The resulting loose pellets were washed for a second time by a centrifugation at 8,000g for 10 min. Pellets were collected and resuspended in isolation buffer. Percoll density gradient centrifugation was performed as described above for a second time. The interface between 23 and 40% (mitochondria released from synaptosomes) was collected and resuspended in isolation buffer to centrifuge at 16,500g for 15 min. The resultant pellet was resuspended in isolation buffer followed by a centrifugation at 8,000g for 10 min. The final synaptic mitochondrial pellet was resuspended in isolation buffer and stored on ice during experiments. Protein concentration was determined using the Bio-Rad DC protein assay (Bio-Rad Laboratories).
ATP synthase catalytic activity assay. ATP synthase activity was measured spectrophotometrically using NADH-linked, ATP-regenerating system 62 . Briefly, mitochondria of appropriate amount were placed in ATP synthase Assay buffer (100 mM Tris-HCl (pH 7.4), 2 mM MgCl 2 , 50 mM KCl, 0.2 mM EDTA, 0.23 mM NADH and 1 mM phosphoenolpyruvate). The reaction was triggered by the addition of 0.4 M ATP-Mg and recorded on a spectrophotometer (Ultrospect 2100, Amersham Biosciences) at OD340 nm for a total of 600 s at 10-s intervals.
ATP synthase coupling assay. Mitochondrial fractions (15 mg) were placed in ATP synthase assay buffer (100 mM Tris-HCl (pH 7.4), 2 mM MgCl 2 , 50 mM KCl, 0.2 mM EDTA, 0.23 mM NADH and 1 mM phosphoenolpyruvate) in the presence of oligomycin-A at 0, 0.4 and 1 mg oligomycin-A per mg mitochondrial protein for 15 min at room temperature. After the incubation, ATP synthase activity was measured spectrophotometrically using NADH-linked, ATP-regenerating system. ATP synthesis assay. Aliquots of mitochondria were analysed by using the ATP Luminescent assay kit (Abcam). Briefly, mitochondria were placed in isolation buffer (225 mM mannitol, 75 mM sucrose, 2 mM K 2 PO 4 , 0.1% BSA, 5 mM HEPES, 1 mM EGTA (pH 7.2)). Mitochondria were energized with 5 mM glutamate/malate and ATP synthesis was induced with the injection of 500 mM ADP. ATP determination was accomplished following the manufacturer's instructions.
Mitochondrial respiration assays. Purified mitochondria were energized by glutamate (5 mM) and malate (5 mM) and subjected to respiration assays on a Clark electrode. Oxygen consumption was triggered by the addition of ADP. The mitochondrial respiratory control ratio was defined as the ratio of State III respiration/State IV respiration. To measure oligomycin-insensitive respiration, mitochondria were energized by glutamate (5 mM) and malate (5 mM) and oxygen consumption was triggered by the addition of ADP as described above. ADP-triggered respiration was collected as total respiration. Oligomycin (5 mM) was then added and the after-oligomycin respiration was collected. The oligomycin-insensitive respiration was calculated as the percentage of after-oligomycin respiration in total respiration.
Mitochondrial swelling assay. Mitochondria were suspended in 0.5 ml of swelling assay buffer (50 mg of mitochondrial protein, 150 mM KCl, 5 mM HEPES, 2 mM K 2 HPO 4 , 5 mM glutamate (pH 7.3)). Mitochondrial swelling was triggered by the addition of calcium (500 nmol mg À 1 of protein). Swelling was observed by immediately and continuously recording changes at OD540 nm by using a spectrophotometer (Ultrospect 2100, Amersham Biosciences) for a total of 600 s at 10-s intervals.
Co-immunoprecipitation. Mitochondria from cerebral cortices of transgenic mice or human subjects were suspended in buffer (500 mg ml À 1 , 50 mM Tris, 150 mM NaCl, 1 mM EDTA, protease inhibitors (Calbiochem, set V, EDTA free), 0.1% NP-40, pH 7.5) and then subjected to five freeze-thaw cycles, followed by centrifugation at 12,000g for 10 min at 4°C. We immunoprecipitated the resulting supernatant with mouse antibody to OSCP (Santa Cruz, #sc-365162, 0.5 mg IgG per 100 mg protein) at 4°C overnight, followed by an incubation with protein A/G agarose (Pierce) for 2 h at room temperature. Nonimmune IgG (0.5 mg IgG per 100 mg protein) was used as negative control. We subjected the resultant immunoprecipitant to immunoblotting with antibody to Ab (CST, #8243, 1:4,000).
GST-OSCP pulldowns. GST pulldown was performed according to manufacturer's protocol (Pierce). Briefly, the human OSCP cDNA (Gene Name: ATP5O. NCBI Gene ID: 539) or deleted form of OSCP cDNA was transformed into BL21 (DE3) pLysS Escherichia coli (Promega) using the pGEX-4t-1 plasmid (GE Healthcare). After transformation and selection a single colony was chosen for PCR to verify positive transformation. After overnight growth and induction by IPTG (Sigma-Aldrich), bacteria were pelleted and then were lysed by sonication in 1 Â PBS containing 0.2 mM PMSF and 100 mg ml À 1 lysozyme. After sonication bacterial debris was removed by centrifugation at 12,000g for 15 min at 4°C. Supernatant was collected and incubated with glutathione agarose high-capacity, high-performance resin (Pierce) for 2 h. Glutathione beads were then washed and incubated overnight at 4°C with mitochondria lysates or Ab peptide. After washing, protein was eluted from the beads and separated by SDS-PAGE. Immunoblotting or coomassie staining was performed to visualize results.
Protein-protein predictions. HADDOCK (high ambiguity-driven proteinprotein docking) 63 was used to predict protein-protein interaction between ATP5O (PDB# 2WSS; s chain) and amyloid beta (PDB# 1Z0Q). The modelling was performed in ambiguous interaction restraints with CPORT'S active and passive constraints. Models with z-scores below À 1.4 were selected and amino acids on OSCP within 4 Å of amyloid beta were determined using PyMOL 64 .
Deletion constructs. Regions of interest were determined by both HADDOCK modelling and based on literature analysis 65,66 . Four OSCP deletions constructs were made; OSCPD1-18, OSCPD168-190, OSCPD1-18; 168-190 and OSCPD107-121. Theses constructs were generated via PCR and inserted into a GST-bearing vector (pGEX-4t-1). The vector was subsequently transformed into BL21(DE3) pLysS E. coli (Promega). After transformation and selection of a single colony was chosen for PCR to verify positive transformation. Protein products were obtained via IPTG induction and purification with glutathione agarose high-capacity, high-performance resin (Pierce).
Oligomeric Ab preparation. Ab1-42 peptide (GenicBio) was diluted in 1,1,1,3,3,3,-hexafluoro-2-propanol to 1 mM using a glass gas-tight Hamilton syrings with a Teflon plunger. The clear solution was then aliquoted in microcentrifuge tubes, and it was dried by vaporation in the fume hood. Peptide film was diluted in DMSO to 5 mM and sonicated for 10 min in bath sonicator. The peptide solution was resuspended in cold HAM'S F-12 to 100 mM and immediately vortexed for 30 s. The solution was then incubated at 4°C for 24 h.
ELISA assay for mitochondrial Ab measurement. Ab level in mitochondrial fractions was measured by using human Ab1-40 and Ab1-42 ELISA kits (Life Technologies) following the manufacturer's instructions.
Mouse neuron culture. Cortices or hippocampi were dissected from day 0-1 pups in cold Hank's buffer (without Ca2 þ and Mg2 þ ), dissociated with 0.05% trysin at 37°C for 15 min followed by 10 times trituration in ice cold neurobasal A medium. Cells were then passed through 40 mm mesh cell strainer (Corning) and centrifuged for 2 min at 200g. The pellet was gently resuspended in neuron culture medium (neurobasal A with 2% B27 supplement, 0.5 mM L-glutamine, 50 U ml À 1 penicillin, and 50 mg ml À 1 streptomycin) and plated on poly-D-lysine-(Sigma-Aldrich) coated culture plates (Corning) or Lab-Tek chamber slides (Nunc, 177445) with an appropriate density.
Human neural stem cell culture and neural cell differentiation. Human neural stem cells (StemPro Neural Stem Cells, #A10509-01, Life Technologies) were cultured and differentiated into neurons as manufacturer's instruction. Differentiated neurons were determined by the morphology as well as the staining of neuronal-specific marker bIII tubulin using an antibody against bIII tubulin (CST, #D71G9, 1:300) followed by a secondary antibody of Alexa Fluor@ 594 (Life Technology, #A-11037, 1:500).
OSCP knockdown in mouse primary neurons. Lentivirus-expressing shRNA targeted to mouse OSCP was packaged with lentivirus shRNA construct (clone TRCN0000076166: 5 0 -CCGGGCTTCCTGAGTCCAAACCAAACTCGAGTTTGG TTTGGACTCAGGAAGCTTTTTG-3 0 , Sigma-Aldrich), packaging vector psPAX2 (Addgene) and envelope vector pMD2.G (Addgene). Lentivirus-expressing nontarget shRNA control (SHC002, Sigma-Aldrich) was used as a control. Mouse primary neurons were cultured for 3 days before infection with lentivirus at a multiplicity of infection (m.o.i.). of 5. The virus containing medium was removed after 2 h and fresh culture medium was replaced to continue culturing. Neurons were treated and collected for experiments after a further 7 days in culture.
b Subunit knockdown in mouse primary neurons. Lentivirus-expressing shRNA targeted to mouse b subunit was packaged with lentivirus shRNA construct (clone TRCN0000076228: 5 0 -CCGGCTGCAACTGATCTCTCCATATCTCGAGATATG GAGAGATCAGTTGCAGTTTTTG-3 0 , Sigma-Aldrich), packaging vector psPAX2 (Addgene) and envelope vector pMD2.G (Addgene). Lentivirus-expressing nontarget shRNA control (SHC002, Sigma-Aldrich) was used as a control. Mouse primary neurons were cultured for 3 days before infection with lentivirus at an m.o.i. of 5. The virus containing medium was removed after 2 h and fresh culture medium was replaced to continue culturing. Neurons were treated and collected for experiments after a further 7 days in culture.
OSCP overexpression in neurons. Human OSCP cDNA (Gene Name: ATP5O. NCBI Gene ID: 539) were inserted in to lentivirus vector with human polyubiquitin promoter-C (Addgene). Lentiviruses were packaged and applied on primary neurons similar as shRNA lentiviral vector. Oligomeric Ab (500 nM) was used on neurons for 24 h before cell collection.
b Subunit overexpression in neurons. Human b-subunit cDNA (Gene Name: ATP5B. NCBI Gene ID: 506) were inserted in to lentivirus vector with human polyubiquitin promoter-C (Addgene). Lentiviruses were packaged and applied on primary neurons similar as shRNA lentiviral vector. Oligomeric Ab (500 nM) was used on neurons for 24 h before cell collection.
Preparation and treatment of OSCP OE human neurons. Neural stem cells were plated on poly-ornithine (Sigma-Aldrich) and laminin (Life Technologies) coated culture plates in complete StemPro neural stem cell SFM (Life Technologies) at 2.5 Â 10 4 cells per cm 2 . After 24 h, the medium was replaced with neural differentiation medium (neurobasal A medium containing 2% B27 and 0.5 mM L-glutamine; Life Technologies) for neuron differentiation. 3 days after differentiation, 10 mM fluorodeoxyuridine (Sigma-Aldrich) and 10 mM uridine (Sigma-Aldrich) were added to remove mitotic cells. After 2 days differentiation cells were infected with expressing lentivirus at an m.o.i. of 5, then were treated with 500 nM Ab for 4 days after 7 days differentiation.
Mitochondrial membrane potential assay. Neurons were incubated with 200 nM TMRM (Sigma-Aldrich) which is cell-permeable red colour fluorescent dye and a specific indicator of mitochondrial membrane potential. After the incubation of TMRM for 30 min in an incubator, the dye was washed by using pre-warmed neurobasal A medium and the TMRM staining on neuronal mitochondria was imaged on an inverted fluorescent microscope with on-stage incubator (Nikon). The mitochondrial membrane potential in neurites was analysed by using Nikon NIS Advanced Research software.
Mitochondrial superoxide assay. Mitochondrial superoxide was determined by using Mitosox Red (Life Technologies) 5,26,48 . Neurons were incubated with 2 mM MitoSox Red for 30 min in an incubator followed by washing using pre-warmed Neurobasal A medium. The images of Mitosox Red staining were collected on a Nikon inverted confocal microscope with on-stage incubator. The intensity was subsequently analysed by using Nikon NIS Advanced Research software.
Calcein AM-cobalt chloride quenching assay. Neurons were subjected to the labelling of 1 mM calcein (Life Technologies) and then incubated with 1 mM cobalt chloride (Sigma-Aldrich) to remove cytosolic calcein staining. The changes of mitochondrial calcein were recorded in the absence or presence of ionophore, A23187 (Sigma-Aldrich) which induces mitochondrial calcium overloading on a Nikon inverted fluorescent microscope with on-stage incubator (37°C, 5% CO 2 ) for 30 min. The results were subsequently analysed by using Nikon NIS Advanced Research software. (Sigma-Aldrich, #M4403, 1:400) followed by goat anti-mouse IgG conjugated with Alexa 488 (Invitrogen, #A11029, 1:500). Images were collected under a Nikon confocal microscope. Neuritic segments 20 mm away from soma were used for the analysis. Neuritic mitochondrial population was calculated as the area of a neurite occupied with mitochondria.
BN-PAGE. BN-PAGE was performed to separate mitochondrial F1FO-ATP synthase. Purified mitochondria were pelleted and resuspended in Solubilization Buffer (NativePAGE Sample Buffer (Life Technologies), Protease Inhibitor (Calbiochem, set V, EDTA free), and 1 mM PMSF with 3.33% Digitonin) and incubated on ice for 30 min. Samples were centrifuged at 12,000g for 10 min at 4°C, the supernatants was recovered and mixed with G-250 Solution ( 1 4 of detergent percentage). Samples were loaded on 3-12% NativePAGE Novex gel (Life Technologies) to separate the proteins. The intensity of coomassie staining was measured to determine the equal loading amount of samples. BN-PAGE was then transferred onto PVDF membrane with Mini-PROTEAN Tetra electrophoresis system (Bio-Rad). Proteins were subsequently fixed with 5% acetic acid and Native Mark ladder (Life Technologies) was visualized by Ponceau S staining. Blocking was performed for 1 h at RT with 5% nonfat milk. Immunoblot was performed against ATP synthase b subunit (Santa Cruz, #sc-33618, 1:5,000) or F1 (Abcam, #ab109867, 1:2,000) and OSCP (Santa Cruz, #sc-365162, 1:6,000). Images were collected on a Bio-Rad Chemidoc Imaging System. Image J software (National Institutes of Health) was used to analyse the scanned blots and to quantify protein signal intensity. For Ab treatment experiment, purified brain mitochondria were exposed to vehicle or 5 mM oligomeric Ab in for 30 min on ice and subsequently exposed to the incubation in the presence or absence of Ca 2 þ for 10 min. Mitochondria were then washed with mitochondria isolation buffer via centrifugation at 8,000g for 10 min. The pelleted mitochondria were used for BN-PAGE as described above.
Dot blot. Dot blot was performed to confirm 5xFAD genotype. Mitochondrial or brain cortical extracts were loaded onto nitrocellulose membrane and allowed to dry. Membrane was blocked for 1 h at RT with 5% nonfat milk. Immunoblot was performed with anti-amyloid beta IgG (CST, #8243, 1:5,000). Images were taken with ChemiDoc MP System (Bio-Rad).
Electrophysiology. Whole-cell voltage-clamp recordings were obtained from cultured cells at room temperature using recording artificial cerebrospinal fluid (ACSF) containing (in mM): 126 NaCl, 2.5 KCl, 1.2 Na 2 HPO 4 , 25 Na 2 HCO 3 , 10 glucose, 2 CaCl 2 , and 1 MgCl 2 , bubbled with 95% O 2 -5% CO 2 (refs 26,70). Miniature excitatory postsynaptic currents (mEPSCs) were pharmacologically isolated by adding 75 mM picrotoxin and 1 mM tetrodoxin to the recording ACSF. Electrodes (2-3 MO open tip resistance) were filled with (in mM): 130 CsCl, 20 TEA, 10 HEPES, 2 MgCl 2 , 0.5 EGTA, 4 Na-ATP, 0.3 Na-GTP, 14 phospocreatine, and 2 QX-314, pH 7.2 (CsOH). Recordings were performed using an Axon Multiclamp 700B amplifier (Molecular Devices, Union City, CA, USA) and the data were acquired using Axograph X (Axograph Scientific, New South Wales, Australia). Cells were recorded at a holding potential of À 65 mV, and access resistance of the recorded cells was monitored throughout the experiment and a o20% change was deemed acceptable. The frequency and amplitude of miniature postsynaptic currents were measured from 3 min continuous recording using MiniAnalysis (Synaptosoft, Decatur, GA, USA), with a threshold set at two times the RMS baseline noise. To directly measure changes in postsynaptic glutamate receptor function, we 'puffed' glutamate onto cultured hippocampal neurons during voltage-clamp recordings. Glass pipettes (1-2 mm tip) were filled with ACSF containing 1 mM L-glutamic acid and 0.5 mM Alexa Fluor 594 hydrazide. Pipettes were placed within 30 mm of the soma of the recorded neuron and glutamate was pressure ejected through a pneumatic drug ejection system (NPI Instruments, Tamm, F.R.G.), triggered by the recording software. Air pressure was set to 0.15-0.2 MPa and puff duration was between 15 and 20 ms. Fluorescence video imaging of Alexa 594 was used to ensure that glutamate puff size was comparable and that no leakage occurred between puffs. Peak amplitude was calculated from the average of at least 10 sweeps.
Statistical analysis. Two-way ANOVA followed by Bonferroni post hoc analysis or Student t-tests wherever appropriate were used for repeated measure analysis on SPSS software (IBM software). The distribution and variance were normal and similar in all groups. Po0.05 was considered significant. All data were expressed as the mean ± s.e.m.