Mild deficiency of mitochondrial Complex III in a mouse model of Alzheimer’s disease decreases amyloid beta plaque formation

Background: For decades, mitochondrial dysfunctions and the generation of reactive oxygen species have been proposed to promote the development and progression of the amyloid pathology in Alzheimer’s disease, but this association is still debated. In particular, it is still unclear if mitochondrial dysfunctions are a trigger or rather a consequence of the formation of amyloid aggregates, and in particular, the role of the different mitochondrial oxidative phosphorylation complexes in Alzheimer’s patients’ brain remains poorly understood. Methods: To study how mitochondrial Complex III defects affect amyloid beta pathology in vivo, we partially knocked out mitochondrial Complex III (CIIIKO) in mature forebrain neurons of an Alzheimer’s mouse model that develops plaque pathology (APP/PS1). Results: We found that Complex III dysfunction in adult neurons induced mild oxidative stress which did not correlate with increased amyloid beta accumulation. In fact, CIIIKO-AD mice showed decreased plaque number, decreased Aβ42 toxic fragment and altered amyloid precursor protein cleavage pathway. Conclusions: Our results support a model in which mitochondrial dysfunction is not the cause of amyloid oligomer accumulation but rather a consequence of amyloid beta toxicity.

Scarcer are data on the other complexes and in particular of Complex III. Decreased Complex III core protein 1 was observed in the temporal cortex of patients with AD (20), mitochondrial preparations of frontal cortex of AD patients show deficiencies of Complex II, III and IV but no differences in respiratory supercomplexes (21). Other reports show how Complex III activity and Complex III/ Citrate Synthase (CIII/CS) ratio are higher in AD patients' platelets (22).
Despite years of study, the role of the different mitochondrial OXPHOS complexes in the brain of AD patients remains poorly understood. In particular, it is still unclear if mitochondrial dysfunctions are a trigger or rather a consequence of AD pathology. ROS can affect amyloid pathology by enhancing beta-secretase and gammasecretase activity and exacerbating amyloid beta fragment (Aβ) production and aggregation (23)(24)(25)(26), but amyloid pathology, on the other side, can also negatively affect mitochondrial function. Aβ directly associates with mitochondria, inhibits OXPHOS (15,(27)(28)(29)(30)(31)(32)(33), and can alter fission and fusion processes (34).
We previously reported that severe isolated Complex IV deficiency led to a reduction in plaque formation and Aβ steady-state levels in the APP/PS1 mouse model of AD (35) and that depletion of mtDNA also led to a decrease in plaques in the same model (36).
These findings were surprising and raised the question that severe OXPHOS defects may not reflect age related OXPHOS decay, which are mild. To directly analyze the contribution of a mild Complex III deficiency to the amyloid pathology, we crossed APP/PS1 AD transgenic mice with a RISP neuron specific conditional KO in which the assembly and function of mitochondrial Complex III is impaired in a subset of post-mitotic neurons, particularly CamKIIα+ neurons in the cerebral cortex and the hippocampus.

Mouse Procedures
The AD transgenic mice carry a mutant APP Swe and a mutant PSEN1 dE9 allele (The Jackson Laboratory, Stock #34832) (37), CaMKIIα-CreERT2 transgenic mice express a tamoxifen-inducible Cre recombinase under the control of the mouse calcium/calmodulin-dependent protein kinase II alpha (CaMKIIα) promoter region (38) (The Jackson Laboratory, Stock #012362). Mice were crossed with RISP FF mice where exon 2 of the RISP gene is flanked by two loxP sites (39). Tamoxifen was prepared by first dissolving the powder in ethanol (20 mg per 100 µl) and mixing this solution with 900 µl corn oil for a final concentration of 20 mg/ml. 5-month-old AD-RISP KO mice (RISP FF -AD +/--CaMKIIαCreERT2 +/-) and control mice (RISP FF -AD +/-) were injected IP with 130 mg per kilogram body weight of tamoxifen once a day for 5 consecutive days. All experiments and animal husbandry were performed according to a protocol approved by the University of Miami Institutional Animal Care and Use Committee. Mice were housed in a virus-antigen-free facility of the University of Miami, Division of Veterinary Resources in a 12-h light/dark cycle at room temperature and fed ad libitum.

Quantitative PCR
Total DNA was extracted from the cortex and hippocampus using standard phenol/chloroform extraction.
Quantitative PCR reactions done using SYBR chemistry (SsoAdvanced Universal Master Mix SYBR Green, Bio-Rad) were performed on a Bio-Rad CFX96/C1000 qPCR machine. The comparative ddCt method was used to determine the relative levels of undeleted and deleted UQCRFS1 to a control genomic region (40). To estimate the levels of "undeleted UQCRFS1" we designed primers to amplify a region in Exon 2 of levels UQCRFS1 (F:AACCAAGATGAGTACAGACA, R:AGAACCAAGAAGGAGATTGA); to estimate the levels of "deleted UQCRFS1" we designed primers in the intron sequences flanking Exon 2 of UQCRFS1 (F:TCATCCGAGACCCAGCAA, R:AGCACATAGCAGAGATACAA). Primers for beta-Actin were (F:GCGCAAGTACTCTGTGTGGA, R:CATCGTACTCCTGCTTGCTG).

Western Blots
Mice were perfused with phosphate-buffered saline, brains were isolated and different regions were dissected and homogenized in PBS supplemented with Complete protease inhibitor mixture (Roche) [12].

Enzymatic Activity Assays
Complex IV and citrate synthase activities were measured in tissue homogenates by spectrophotometric methods (41). Briefly, homogenates from cortex and hippocampus were prepared using a hand-held rotor (VWR) to homogenize tissue in PBS plus protease inhibitor mixture (Roche) on ice. Cytochrome c (2 mM) reduced with sodium dithionite was added to homogenates in 10 mM KH2PO4, 1 mg/ml BSA, and 120 mM lauryl maltoside.
Samples were read at 550 -580 nm with the slope reading taken for 2' at 37°C. Potassium cyanide (240 mM) was used to inhibit the reaction to ensure slope was specific to COX activity. Readings were normalized by protein concentration. Homogenates for citrate synthase were added to 50 mM Tris-HCl, pH 7.5, 20 mM acetyl CoA, 10 mM 5,5'-dithiobis (2-nitrobenzoic acid), and 0.2% Triton X-100 and performed at 412 nm with 50 mM oxaloacetate to start the reaction. Readings were obtained for 5' at 30°C. Normalization was again to protein concentration.

Beta-Secretase Activity Assay
Beta-secretase activity was measured in cortical and hippocampal lysates using a FRET-based substrate, H-RE(EDANS)EVNLDAEFK(DABCYL)R-OH (Calbiochem), which contains a beta-secretase cleavage sequence with Swedish-type mutations. Homogenates lysed in PBS-1% Triton-X and protease inhibitor cocktail (Roche) were centrifuged at 10,000 x g for 5' at 4ºC, and the supernatant was used for the analysis. The activity assay was carried out on 4 μg of protein in a 100-μl volume reaction volume that contained 12.5μM substrate and 20 mM sodium acetate (pH 4.4). The fluorescence from the released EDANS fluorophore was measured at 37°C with excitation and emission wavelengths of 355 and 460 nm, respectively using a Synergy H1 hybrid reader (Biotek). Background fluorescence from the substrate alone was subtracted from the readings. Fluorescence (arbitrary units) was recorded 30' after the initiation of reaction.

Quantification of Aβ42 fragment
Cortices and hippocampi were homogenized in cold guanidine buffer [ were then mounted with SurgiPath Sub-X mounting media (Leica). Images were captured with an Pathscan Enabler 5. As previously described (35) the total number of amyloid plaques per tissue area (cerebral cortex and hippocampus) were counted from ten non-consecutive stained sections per animal (n=3-4 per group) scanning the whole area. Fiji program was used for stereological counting.

Nissl staining:
PFA-fixed frozen sections were dried for 3 hours, then incubated in 1:1 ethanol/chloroform overnight for defattening. Slides were then rehydrated through 100% ethanol to distilled water. Staining was performed in warm (37ºC) 0.1% cresyl violet solution for 5-10 minutes (Abcam 246817). Slides were then rinsed in running distilled water, serially dehydrated to 100% ethanol 2x5 minutes, cleared in xylene 2x 5 minutes, and mounted with a permanent mounting medium. Images were captured with an Pathscan Enabler 5.

Immunostaining of Oxidized Nucleic Acids
Histological sections were obtained as previously described above. After deparaffinization and hydration, sections were microwave-heated twice for 4' in 10 mM sodium citrate buffer (pH 6.0). Sections were then permeabilized in 0.2% Triton X-100 for 30' and endogenous peroxidase activity was quenched by immersing section in 100% methanol containing 0.3% H2O2 for 30'. After blocking sections with 10% NGS for 1 h at RT, sections were incubated with monoclonal antibody against 8-hydroxy-2'-deoxyguanosine (oh 8 dG) and 8-hydroxyguanosine (oh 8 G) (1:1,000; QED Bioscience) for 16 h at 4°C in 5% NGS. The bound primary antibody was detected as stated above using streptavidin peroxidase/DAB staining. For a negative control, AD sections were pre-incubated with 10 µg/µl RNase (Qiagen) and 0.04 units/µl DNase (Promega) in 50 µl of 1´ RQ buffer (Promega) for 3 hours at 37ºC prior to the primary antibody incubation. For a positive control, AD sections were pretreated with 30% H2O2 for 10 minutes before blocking with NGS.

OxyBlot
Cortices and hippocampi were homogenized in cold PBS +protein and phosphatase inhibitors. OxyBlot measures carbonyl groups introduced into proteins by oxidative reactions with ozone or oxides of nitrogen or by metal catalyzed oxidation. Carbonyl groups were quantified on 15mg of protein homogenates using S7150 Sigma-Aldrich OxyBlot Protein Oxidation Detection Kit.

Statistical Analysis
Two-tailed, unpaired Student t test was used to determine the statistical significance between the different groups.
If more than two groups were analyzed, significance of the differences was evaluated by one-way ANOVA followed by Bonferroni post-test. Double asterisks indicate p< 0.01, a single asterisk indicates p< 0.05. Error bars represent standard error of the mean (SEM).

Generation of CIII KO -AD mice
To analyze the effects of Complex III depletion on AD pathology, we developed and characterized an AD transgenic mouse carrying mutant APP and mutant presenilin 1 (APP/PS1) (37) with adult-onset Rieske Iron Sulfur Protein (RISP) defects. RISP is encoded by the UQCRFS1 gene and is a catalytic subunit of mitochondrial OXPHOS Complex III. Because APP/PS1 mice develop amyloid deposits in cortex and hippocampus starting at 6-7 months of age (42), we wanted to induce a Complex III defect in adult mice. Therefore, we used a CaMKIIα-CreERT2 transgenic mouse in which the expression of the Cre recombinase is tamoxifen-inducible (38) (Fig.1B).
To verify that cre recombinase excised exon 2 of the UQCRFS1 gene specifically in brain of CIII KO -AD mice after tamoxifen induction, we analyzed DNA extracted from cortex and spleen of CIII KO -AD and AD mice injected with tamoxifen ( Fig.1B). We could detect the gene deletion in cortex of CIII KO -AD mice, but not in spleen of CIII KO -AD mice and in cortex of control mice (Fig1C), indicating the specificity of the system. We estimated the levels of exon 2 excision in cortex and hippocampus by qPCR and found that both areas had 50% of the UQCRFS1 gene deleted (Fig.1D). Considering that the samples used to extract DNA contained not only CaMKIIα + neurons but also glia and CaMKIIαneurons, the excision of the UQCRFS1 gene in the target neurons was actually higher than the 50% detected by this method. No significant differences were detected in the excision levels between cortex and hippocampus (Fig.1E).

Effects of ablation of RISP on OXPHOS complexes
To verify the levels of RISP protein in CaMKIIα + neurons, we performed a western blot assay on cortical and hippocampal homogenates of AD and CIII KO -AD mice (Fig.2). When normalized to total protein loading, RISP protein was not reduced in cortex ( Fig. 2A-2D) or in hippocampus ( Fig. 2A-2G). When normalized to a mitochondrial membrane protein, VDAC, the level of RISP was reduced to 62% in cortex (Fig.2B)  belonging to the other complexes were significantly affected. We performed this analysis also in males (Suppl. Fig.1) which also showed an increase of SDHA in the hippocampal homogenates. Interestingly, in males, COXI (Complex IV) was also significantly increased in the CIII KO -AD mice compared to control. To determine if the increased levels of subunits of Complex II in females and of Complex II and IV in males were due to increased mitochondrial mass, we measured the steady-state levels of two other non-OXPHOS mitochondrial membrane proteins, VDAC1 and TIM23. VDAC1 was increased in cortex from both females and males but unchanged in hippocampus ( Fig. 2H-I, Suppl. Fig.1H-I). TIM23 was increased only in hippocampal homogenates from females ( Fig. 2H-I).
To determine if the induced decrease of RISP was enough to cause a defect in Complex III activity, we measured OXPHOS complex enzymatic activities in cortical and hippocampal homogenates. Complex III activity was reduced in cortex to 65% (± 7) compared to control (Fig.3A), in hippocampus it was reduced to 61% (± 3.5) (Fig.3C). These homogenates also contain glial and endothelial cells, in which the UQCRFS1 gene has not been knocked out. Because of this, we have to consider an underestimation of the actual decrease in Complex III activity in affected neurons. Complex IV and citrate synthase (CS) activity were not changed in cortical and hippocampal homogenates (Fig.3B, 3D, Suppl. Fig.2).
In order to analyze if the loss of RISP caused a reorganization of the respiratory complexes, mitochondrial proteins were extracted with digitonin and separated by blue native gel electrophoresis followed by western blot. Complex I, III and IV were analyzed using anti-NDUFA9, RISP and COX1 antibodies, respectively. TIM23, VDAC1 and SDHA were used as mitochondrial loading control. Mitochondrial supercomplexes are indicated in the figure as HMW (High Molecular Weight) and CI+CIII2. Complex V and II were analyzed using ATP5a and SDHA antibodies, respectively. We detected a decrease of CI+CIII2 architectures in cortex of CIII KO -AD female mice, and an increase of free Complex I (Fig.3E-F). We found similar results in hippocampus, but because of a higher variability, the analysis did not reach statistical significance (Fig.3E-G). Assembled Complexes IV, V and II did not show changes.

Neurodegeneration and neuroinflammation in CIII KO -AD mice
The brains of 8-month-old CIII KO -AD mice did not show significant differences in size or weight compared to AD controls. To detect if the mild defect in Complex III affected the neuroanatomy of the brain, we first analyzed the gross morphology of different brain regions by hematoxylin & eosin staining. We did not detect any gross nor massive anatomical alteration (Suppl. Fig. 3A). We then performed Nissl staining (which is commonly use to stain nucleic acids in the nervous tissue) and immunohistochemistry staining for NeuN (marker of neuronal nuclei) on PFA-fixed frozen sections. We did not detect gross changes in the neuronal population in different areas analyzed (Suppl. Fig. 3B, Fig.4A). We then performed western blots analysis to measure the levels of neuronal marker TUJ1 and synaptic marker Synaptophysin on cortical and hippocampal homogenates of CIII KO -AD and AD controls. We detected a decrease in TUJ1 in cortex from CIII KO -AD females compared to AD controls (Fig.4B), while we did not detect significant changes in males (Suppl. Fig.3D).
Neurodegeneration and metabolic defects are often accompanied by neuroinflammation, therefore we examined the extent of gliosis in these animals. We performed immunohistochemistry and western blot analysis with antibodies against GFAP (marker of glial cells) and Iba1 (marker of microglia) on both females and males. We did not detect significant changes, neither in the steady-state levels of both markers (Fig. 4D, Suppl. Fig.3F) nor in the morphology of GFAP and Iba1 positive cells (Fig.4C, Suppl. Fig.3E).

CIII KO -AD mice show fewer amyloid plaques and fewer Aβ42 fragment
Starting at 6 months of age, AD mice accumulate numerous amyloid plaques, with most of them present in the cerebral cortex and the hippocampus (36). By 8-9 months, the plaque number reaches 1.5-2 plaques per mm 2 (35,36) . To analyze if the partial defect of Complex III affected the formation of the amyloid plaques, we performed a stereological count of amyloid plaques on 8-month-old AD and CIII KO -AD mice. We visualized the plaques by immunohistochemistry on serial coronal sections with an anti-human Aβ antibody (6E10) that recognizes the amyloidogenic portion of APP. Brains from CIII KO -AD females showed significant reductions in the number of plaques compared to AD brains, in cortex and hippocampus ( Fig. 5A-B).
The plaque number is correlated with Aβ42 content and to secretase activity (44,45). To determine whether the diminished number of plaques in CIII KO -AD mice correlated with the amount of Aβ42 peptides, we performed an ELISA on protein homogenates from cortex and hippocampus of AD and CIII KO -AD mice. Lysates from CIII KO -AD females' brains showed a reduction of Aβ42 of ~66% in cortex and of ~50% in hippocampus (Fig.5C). We consequently analyzed APP processing by performing western blot analyses of the different fragments derived from the APP sequential cleavage. We used an antibody against the carboxy terminal portion of APP (Ctf, amino acid residues 676-695) that recognizes both the un-cleaved APP and the C-terminal fragment, formed after the first secretase cleavage (either by alpha-or beta-secretase). From this analysis we detected a decrease in the ratio between the carboxy terminal fragment (Ctf) and the full-length APP in homogenates from the cortex, suggesting a downregulation of this processing (Fig. 5D). In hippocampus we did not detect significant changes.
One of the crucial events triggering the formation of amyloid fragments from APP is the cleavage by betasecretase. We measured beta-secretase activity with a FRET-based assay in cortical lysates from AD and CIII KO -AD mice brains, but we did not detect significant differences between the 2 groups (Fig. 5E), suggesting that a decrease in beta-secretase activity was not responsible for the reduced levels of Ctf/APP. The steady state levels of the major gamma-secretases involved in the formation of Aβ42 (Nicastrin and Presenilin 1) in cortical and hippocampal homogenates, were also not altered (Fig.5F). To investigate the possibility that the decrease in Aβ42 fragments occurred because of an increase in its degradation, we analyzed the levels of two proteins involved in general cellular pathways for protein degradation: p62 and LC3B (autophagy pathway), and 20S proteasome (ubiquitin proteasome system, UPS). We did not detect significant changes in the levels of p62 nor in LC3B lipidation (Suppl. Fig.4A). We did detect a decrease of the 20S proteasome levels in cortical but not in hippocampal homogenates (Suppl. Fig.4B). Brains from CIII KO -AD mice males, did not show significant changes in plaque numbers (Suppl. Fig.5A-B), Aβ42 content (Suppl. Fig.5C), or APP processing (Suppl. Fig.5D).

CIII KO -AD mice show increased oxidative stress
Plaque formation has been correlated with oxidative stress. AD patient brains show increased levels of lipid peroxidation, DNA strand breaks, and oxidized DNA bases (46). To investigate for signs of oxidative stress, we measured the steady-state level of SOD2, a protein that is typically increased as a consequence of oxidative stress. We detected increased levels of SOD2 in cortex of CIII KO -AD females compared to AD, while in hippocampus we did not detect changes (Fig.6A). Oxidative stress also causes oxidative modification to proteins with consequent addition of carbonyl groups to protein side chains. To provide a more direct indication of the oxidative state of the different brain regions, we performed an OxyBlot (Sigma), that identifies these carbonyl groups. Cortical homogenates showed proteins with more carbonylated groups, indicating a higher status of oxidative stress (Fig. 6B). Hippocampus did not show any significant changes, similar to the analysis performed with SOD2.
Previously we showed that the cRISP KO animals displayed oxidative stress in specific regions of the brain, in particular in the piriform cortex (47). To analyze if in our model the oxidative stress was also localized in specific regions of the brain, we performed an 8-hydroxy-deoxy-guanosine (8-OHdG)/8-hydroxy-guanosine (8-OHG) staining. 8-OHdG and 8-OHG are a product of DNA and RNA oxidative damage respectively, therefore an indirect result of the presence of ROS. We did not detect changes between AD and CIII KO -AD mice (Fig. 6C), showing that the mild CIII defect did not exacerbate oxidative damage to nucleic acids.

Discussion
Although the primary cause of Alzheimer's disease (AD) is still unknown, mitochondrial OXPHOS dysfunction has been implicated in the development and progression of AD. Several studies reported OXPHOS defects in postmortem AD brains (14,16,17,48) and mutations in mitochondrial DNA (mtDNA), which encodes proteins for several OXPHOS complexes (10)(11)(12). Moreover, it has been shown recently that mitochondrial Complex I abnormalities are associated with tau load in AD patients (49). By better understanding the metabolic consequences of OXPHOS defects in the CNS and in AD pathology, novel therapeutic approaches potentially applicable to AD could be developed.
Different OXPHOS deficiencies influence development and brain pathology in distinctive ways. We have previously compared the phenotypes of mice in which Complex I (NDUFS3) (50), Complex III (RISP) or Complex IV (COX10) (47) have been knocked out in neuronal cells. These models showed some interesting differences.
The Complex I neuronal KO (Ndufs3 nKO) mice died around 5 months with a fatal encephalopathy, ataxia, neuronal cell death and massive gliosis (50). The Complex III deficient mice (Uqcrfs1 nKO) were all dead by 5 months and showed degeneration of the piriform cortex with high levels of oxidative damage (47). The Complex IV deficient mice (Cox10 nKO) survived much longer with a progressive CNS defect and showed a neuronal loss localized to the hippocampus (47). These findings are also different from our observations with a model of neuronal mtDNA damage, which showed prominent degeneration of the striatum prior to the degeneration of the cortex and hippocampus (51).
Therefore, different OXPHOS deficiencies could influence the development of AD pathology in distinctive ways.
Focusing on the amyloid cascade, it is surprising that in most of the studies on mouse models of AD, inhibition of different mitochondrial complexes resulted in a decrease of plaque burden and Aβ accumulation. Zhang et al. (52) reported how a mild inhibition of Complex I reduced levels of Aβ and phospho-Tau in three AD animal models.
In these animals, there was no sign of oxidative damage or inflammation and mitochondrial bioenergetics was increased. AMP-activated protein kinase was upregulated, GSK3β activity was reduced, and restoration of axonal trafficking resulted in elevated levels of neurotrophic factors and synaptic proteins (52). Our lab previously generated a neuron-specific Complex IV conditional KO mouse (knocking out Cox10, subunit of CIV) crossed with the APP/PS1 mouse (COXd-AD) (35). The plaque burden in cortex and hippocampus was significantly decreased in COXd-AD mice compared to AD controls. This reduction was accompanied by a reduction of total Aβ42 without a significant alteration in the level of APP. In that model, beta-secretase activity was lower and oxidative stress was reduced. The Aβ pathology in this model, however, was studied at 4 months, because the ablation of COX10 caused a massive neurodegeneration.
In a different model (AD-mito-PstI), we induced mtDNA damage in cortex and hippocampus of APP/PS1 mouse for 2 months (6 to 8 months of age) (36). Also, in this model, the OXPHOS deficiency caused by mtDNA depletion had a negative effect on Aβ42 content and, consequently, on plaque burden. This effect was associated with an alteration in APP processing, independent from beta-secretase activity and from oxidative stress. On the other hand, in the mutator mouse, that show accumulation of mtDNA mutations and premature aging, Aβ42 levels and plaque density were increased. The increased amyloid pathology was not caused by an elevated Aβ production but by a decreased clearance by the insulin degrading enzyme (IDE) (53).
A frequent criticism, leveled by proponents of OXPHOS defects causing plaque and Aβ accumulation, is that the OXPHOS defect in these different models was too severe and already present from birth, not mimicking mild defects associated with normal aging. To address this criticism, we analyzed the consequences of a mild and adult-onset Complex III deficiency on the amyloid pathology. We chose Complex III, because defects in this complex have been associated with increased ROS (47). The lack of Complex III activity in neurons (analyzed in previous RISP nKO mice) caused premature death by 5 months of age. To avoid such a strong phenotype, we used an inducible model in which the expression of cre-recombinase is activated by tamoxifen injections at 5months of age. This caused a deletion of encoded protein Rieske iron-sulfur protein (RISP), catalytic subunit of Complex III, with a consequent mild Complex III deficiency in CamKIIα positive neurons. This mild CIII deficiency allowed us to avoid a massive neurodegeneration and to analyze the amyloid pathology 3 months later, at 8 months of age.
There was a difference in phenotypes between females and males. Sex differences in APP/PS1 mice are well known, with females accumulating amyloid at an earlier age than males and building up more amyloid deposits (54). These differences are not limited to the AD pathology (55)(56)(57)(58). In our model, the decrease in plaque burden and Aβ fragment was evident only in cortex and hippocampus of female mice, while amyloid pathology in males was not affected by the defect of Complex III. Moreover, we detected a mild decrease of TUJ1 in cortex of females but no change in the same regions in males.
Still, this model also showed a decrease of plaque burden caused by a lower amount of Aβ42 fragment. In cortex, this was accompanied by an alteration in the APP processing (lower level of Ctf/APP), but not by a difference in secretases activity or steady-state levels. Oxidative stress might affect the plaque formation by enhancing the expression and activity of BACE-1 (59,60) and by enhanced secretion of soluble APPβ. In vitro and in vivo studies showed how Complex I and III-derived ROS lead to elevated levels of Aβ (24). Moreover, recent studies have also reported that the treatment of different AD mouse models with mitochondrial-targeted antioxidants have protective effects that includes decrease in Aβ production and plaque accumulation (61,62). Therefore, the working hypothesis in the previous models was that a lower level of oxidative stress might have been the cause of a decrease in plaque number (35).
The fact that in our model we detected increased oxidative stress, led us to the conclusion that the mechanism by which interfering with OXPHOS complexes activity leads to a decrease of plaques is independent from ROS formation. The consideration that in the models in which mtDNA and Complex I were impaired, the decreased number of plaques was accompanied by no change in ROS, also support our conclusion. Although ROS can exacerbate AD pathology, it is not clear if it has a primary role in the development of the disease.
What may be more compatible with the existing data, is that mitochondria dysfunction is a consequence of Aβ toxicity. The Aβ fragment's typical localization is extracellular, but different studies showed that it can be found in other organelles, including mitochondria, in both human AD brain and mouse models of AD where it causes impairment of electron transport chain (ETC) complexes and increased ROS production (32,63). Aβ can actively be imported into mitochondria through cellular trafficking systems and colocalize with the ETC inhibiting Complex IV (29,64). Aβ may also contribute to mitochondrial dysfunction through disruption of mitochondrial fusion and fission processes, leading to mitochondrial fragmentation (34).
Our results, together with previous reports, support a model in which mitochondrial dysfunction in AD mouse models is not the cause of Aβ oligomer accumulation. Rather, our data are more compatible with a model in which OXPHOS function is decreased because of Aβ toxicity.

Conclusions
We reported how mitochondrial Complex III dysfunction in adult neurons of a mouse model of amyloid pathology induced mild oxidative stress which did not correlate with increased amyloid beta accumulation. CIII KO -AD mice showed decreased plaque number, decreased Aβ42 toxic fragment and altered amyloid precursor protein cleavage pathway. Our results support a model in which mitochondrial dysfunction is not the cause of amyloid pathology but rather a consequence of Aβ toxicity.

List of abbreviations:
Aβ: Amyloid beta

Ethical Approval and Consent to participate
Not applicable

Consent for publication
All authors consented to the publication. Francisca Diaz passed away prior to the submission of the research paper.

Availability of supporting data
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare no competing interests.

Funding
This work was supported primarily by the Florida Biomedical Foundation Ed and Ethel Moore Alzheimer's Disease Research Program grant 5AZ06 (CTM), the National Institute of Health Grants K01AG057815 (MP), and 1R01NS079965 (CTM).

Authors' contributions
MP designed the research, performed the experiments, analyzed, and interpreted data, and wrote the manuscript.
FD performed BN-PAGE analysis, enzymatic activity, and contributed intellectually to the research. NN performed qPCR analysis CSG and PI assisted in plaque counting, RB contributed intellectually to the research, CTM planned the project together with MP and contributed to the writing of the manuscript. All authors read and approved the final manuscript.