Background
Alzheimer Disease (AD) is a multifactorial, age-related neurodegenerative disease of the elderly that leads to progressive memory loss, impairments in behavior, language, visual-spatial skills and ultimately death. The disease is characterized by a progressive neuronal loss and accumulation of extracellular senile plaques composed of amyloid beta (Aß) and intracellular neurofibrillary tangles (NFT) composed of hyperphosphorylated tau in selective brain areas such as hippocampus and cortex [
1]. According to 2015 World Alzheimer’s Report, 47.5 million people had AD-related dementia worldwide, including 5.4 million Americans, and projected the numbers to rise to 75.6 million by 2030 and to 131.5 million by 2050. Over 9.9 million new cases of AD-related dementia are diagnosed every year worldwide [
2]. Dementia has a huge economic impact on our society and the estimated total healthcare cost of dementia worldwide in 2015 was $818 billion. Currently, there are no drugs or agents available to treat or to prevent AD.
Several decades of intensive research have revealed that multiple cellular changes have been implicated during the course of AD including widespread oxidative damage, extensive neuroinflammation, and aberrant cytoskeletal alteration among which mitochondrial dysfunction is an early prominent feature in susceptible neurons in the brain from AD patients and models of AD, and likely plays a critical role in the pathogenesis of AD [
1,
3,
4]. Indeed, a reduced rate of brain metabolism preceding functional impairment is one of the best documented abnormalities in AD which is likely due to deficiency in several key enzymes of oxidative metabolism including cytochrome oxidase (COX) that have been consistently demonstrated in AD [
3,
4]. However, mechanisms underlying mitochondrial dysfunction in AD remains elusive.
Mitochondria are dynamic organelles that continuously fuse with each other to form larger tubular networks and divide into smaller structures, a process regulated by the balance of fission/fusion machinery mainly involving several large GTPases: mitochondrial fission is regulated by cytosolic protein DLP1 which translocates to mitochondria during fission while mitochondrial fusion is regulated by mitofusin 1 and 2 on the outer mitochondrial membrane and OPA1 on the inner mitochondrial membrane [
5]. The delicate balance between mitochondrial fission and fusion is crucial in the maintenance of healthy population of mitochondria and proper mitochondrial distribution, morphology and function, disruption of which causes human diseases, especially neurological diseases [
5]. For example, thus far, genetic mutations were found in Mfn2 and reported to be associated with Charcot-Marie-Tooth (CMT) disease, the most common inherited neurological disorders, accounting for up to 20 to 30% of all axonal CMT type 2 cases, to a lesser degree, also be associated with optic atrophy, clinical signs of first motor neuron involvement, and early onset stroke [
6‐
8]. Increasing evidence suggested that an abnormal mitochondrial dynamics is likely involved in the mitochondrial structural damage and dysfunction in AD: several groups demonstrated that overexpression of familial AD-causing amyloid precursor protein (APP) mutants or exposure to soluble oligomeric Aβ caused changes in the expression of mitochondrial fission and fusion proteins and profound mitochondrial fragmentation in neuronal cells which led to ultrastructural damage to mitochondria and mitochondrial dysfunction along with neuronal deficits such as synaptic abnormalities in vitro [
9‐
14]. Aβ-induced changes in mitochondrial dynamics and distribution are early events in vivo in
Drosophila models [
15,
16]. Abnormal mitochondrial distribution and round, swollen, and damaged mitochondria consistent with enhanced fragmentation are also documented in AD mouse models [
17‐
19]. Fibroblasts from AD patients and AD cybrid cell models also demonstrated abnormal mitochondrial dynamics and dysfunction [
20,
21]. Indeed, ultrastructural deficits and abnormal distribution of mitochondria were evident in pyramidal neurons in AD brain [
9]. Although no specific mutations in fusion genes were associated with AD, studies from several groups consistently demonstrated a significantly reduced expression of large GTPases involved in fusion (i.e., Mfn1, Mfn2 and OPA1) in the brain of AD patients, implicating that mitochondrial fragmentation, largely due to disrupted mitochondrial fusion, likely occurs in the susceptible neurons in AD brain [
12,
22,
23]. However, whether disruption of mitochondrial fusion causes mitochondrial deficits and neurodegeneration in AD-afflicted brain areas has not been determined. To address the causal role of disrupted mitochondrial fusion in neurodegeneration and other AD-related deficits in AD-afflicted brain areas, we disrupted mitochondrial fusion by knocking out Mfn2 in the hippocampus and cortex and found that ablation of Mfn2 caused neuronal degeneration in vivo in a temporal order starting with significant mitochondrial fragmentation and dysfunction and increased oxidative stress, followed by cytoskeletal alterations and inflammatory response that culminates in eventual neuronal death, all features characterizing AD, thus establishing the critical role of mitochondrial fragmentation in mitochondrial dysfunction and neuronal degeneration and the pathogenesis of AD.
Methods
Transgenic mice
Experimental mice were generated by crossing B6.Cg-Tg(Camk2a-cre)T29-1Stl/J mice (Jackson Laboratory) with B6.129(Cg)-Mfn2
tm3Dcc
/J mice (Mfn2loxP). PCR of genomic DNA was used for routine genotyping of offspring to detect the presence of the CAMKCre gene and the Mfn2loxP gene. We chose these mice since the Cre-recombinase was shown to be expressed in the forebrain and hippocampus, highest in the CA1 layer. Mice were housed under standard conditions and all animal studies were performed following an approved protocol through the Case Western Reserve University IACUC board. At various ages, mice were perfused with saline and brain tissue collected and bisected with one half frozen (with brain areas cortex, hippocampus, cerebellum and brainstem area separated) and the other half fixed in 10% buffered formalin. For this constitutive expression study, Mfn2 cKO mice from ages 4, 8, 12, 18, 28, and 78 weeks (at least n = 3 per age group) and age-matched control mice (no Cre+, at least n = 3 per age group) were collected.
PCR
To confirm recombination of the floxed gene, DNA was extracted from frozen samples of cortex, hippocampus, cerebellum and brainstem using the Wizard genomic DNA kit. PCR was performed using primers designed to amplify a 240 bp band only after Cre recombination has occurred.
Immunoblotting
For Western blot analysis frozen dissected hippocampal and cortical tissues were individually homogenized in 10X volume Cell lysis buffer (Cell signaling) with added protease and phosphatase inhibitors (Roche) and centrifuged at 14,000 g for 20 min in a refrigerated microcentrifuge. Protein concentration of the supernatant was determined using BCA protein assay. Proteins, 10 μg per lane (specified in figure legends), were resolved with SDS-PAGE and transferred to Immobilon (Millipore). Blots were blocked in 10% nonfat milk for 1 h and probed with primary antibodies overnight. After washing 5X in TBS-tween, HRP-conjugated secondary antibodies (Cell Signaling) were applied for 1 h, rinsed again and bands detected using ECL (Santa Cruz or Millipore). Loading control was anti-actin (Millipore). Bands were quantified using QuantityOne (Biorad) or ImageJ.
Immunohistochemistry
Formalin fixed brain samples were embedded in paraffin and 6 μm sections were cut using a microtome and placed on coated slides. Immunohistochemistry was performed using the peroxidase anti peroxidase method as previously described [
24]. Briefly, sections were deparaffinized in 2 changes of xylene, dehydrated through descending series of ethanol, and finally into Tris buffered saline (TBS: 50 mM Tris, 150 mM NaCl, pH = 7.6). Endogenous peroxidase was removed with a 30-min incubation in 3% H
2O
2. Antigen retrieval using citrate buffer and pressure cooking (Biocare) was performed for most IHC experiments. After blocking for 30 min in 10% normal goat serum (NGS), primary antibodies were applied and incubated overnight at 4 °C. Species specific secondary antibodies and PAP complexes were applied and after 3 changes of tris buffer, the sections were developed with DAB (Dako) and slides were rinsed in dH2O, dehydrated and coverslipped with Permount. Antibodies used were directed against Cre-recombinase (Millipore), GFAP (Invitrogen), MAP2 (Millipore), NeuN (Millipore), COXI (Molecular Probes), HO-1 (Enzo), iba1 (Wako), AT8 (Thermo Fisher), and mitochondria complex cocktail (Abcam).
To detect protein carbonyls, the dinitrophenol binding -assay was also performed. After tissue sections were rehydrated, 50 ul of DNP solution was applied (20 mM DNP, 0.5% TFA, 92.5% DMSO) and incubated for 15 min at 37 °C. The sections were rinsed with changes of acetic acid until no yellow color could be further removed. After rinsing thoroughly with water and then with TBS, sections were blocked and DNP detected with rabbit antibody to DNP (1/10,000). For western blot analysis, samples were first denatured and then treated with DNP solution following manufacturer’s instructions (Millipore), sample buffer directly added and the entire sample run on the gel.
TUNEL method was used to label cells undergoing apoptosis following manufacturer’s instructions (Roche). After TUNEL method, the sections were stained with DapI and mounted with Fluoromount (SouthernBiotech) and FITC-labelled apoptotic cells were imaged on a Zeiss Axiophot.
Image quantification
Images were acquired on a Axiophot with a axiocam (Zeiss). Quantification was performed on these images using Axiovision software and either the number of cells stained per area, or density of stained structures was determined using Axiocam image analysis software. To measure hippocampal size, H&E stained sections from brain areas collected from Bregma 1.08–1.8, representing the complete hippocampus and similar ventricle presentation were imaged and measured in a blinded fashion. To measure total cortical size, sections from similar levels were imaged at the same magnification, the cortical area cropped using Photoshop and the area was quantified.
Electron microscopy
For electron microscopic analysis, brain samples from control and KO mice were collected and fixed as previously described following brain dissection [
19]. Brain slices of about 1 mm thick were made and small areas of the CA1 region of the hippocampus and cortex were sampled and embedded in Epon. Semithin sections were prepared and stained with toluidine blue to clearly note the CA1 regions. Images of pyramidal neurons from the CA1 region or cortex with initial segment of axon visible were obtained. Mouse genotype was blinded to the electron microscopist. Mitochondria parameters were quantified using Image J and included aspect ratio (length/width) and size (area in μm
2).
Mitochondrial oxygen consumption measurement
The real-time measurement of oxygen consumption rate (OCR) in synaptic mitochondria in synaptosomes was performed using the Seahorse XF24 Analyzer (Seahorse Bioscience, North Billerica, MA), according to the manufacturer’s instructions. If needed, ATP synthase inhibitor oligomycin (1 μM), uncoupler FCCP (4 μM) and complex I inhibitors antimycin A (1 μM) and rotenone (1 μM) were injected sequentially. After measurement, cells and synaptosomes were lysed and OCR data was normalized by total protein as previously described [
24].
Discussion
Studies from multiple groups suggest that mitochondrial fragmentation largely due to disrupted mitochondrial fusion contributes to mitochondrial dysfunction and neuronal deficits in AD. To gain a better understanding of whether and how mitochondrial fragmentation causes AD-related deficits in the hippocampal and cortical neurons, in this study, we developed a transgenic mouse model of disturbed mitochondrial fusion in the pyramidal neurons of hippocampus and cortex by tissue-specific knockout of Mfn2. Interestingly, we found that ablation of Mfn2 caused a series of pathological events including mitochondrial damage and dysfunction, oxidative stress, inflammatory response, and cytoskeletal changes that eventually led to significant neurodegeneration in the hippocampus and cortex in vivo.
In our study, mitochondrial fragmentation became apparent in the CA1 hippocampal neurons since 8 weeks of age as reflected by significantly decreased mitochondrial aspect ratio. This was accompanied by increased ultrastructural damage to mitochondria as reflected by the loss of integrity of the internal structures. In older cKO mice (i.e., 28 weeks of age), swollen mitochondria with significantly enlarged size (i.e., doubled area) and almost complete round shape (aspect ratio close to 1) demonstrated almost complete loss of cristae, suggesting severe damage to mitochondria, a phenotype that was also observed in older DLP1 KO mice [
28] and in the Mfn2 KO Purkinje Cells [
48]. Mitochondrial function is dependent on their intact structure. Indeed, increased oxidative stress and impaired mitochondrial function such as reduced expression of OXPHO proteins and decreased mitochondrial respiratory parameters were also first observed in the Mfn2 cKO mice at 8 week of age when ultrastructural damage was observed. While the causal relationship between mitochondrial dysfunction and damage, and oxidative stress remains to be teased out, it was suggested that excessive mitochondrial fission leads to reduced mtDNA copy number [
29] which could lead to reduced expression of essential OXPHOS proteins as we confirmed by the western blot analysis and cause mitochondrial dysfunction. Prior studies from multiple groups demonstrated that excessive fission or fusion caused deficits in complex assembly [
30‐
32] likely through changes in the cristae shape and curvature or other uncharacterized mechanism that also negatively impacted mitochondrial function. Indeed, dramatic changes in the cristae organization was noted in the 8 weeks old animals. Mitochondrial dysfunction unavoidably causes increased oxidative stress which could damage mitochondrial ultrastructure and further exacerbate mitochondrial dysfunction. The progression of mitochondrial ultrastructural damage from relatively mild changes such as multilamellar appearance and vacuolation with partially broken cristae in Mfn2 cKO at 8 weeks of age to more severe appearance such as the extreme loss of internal cristae structure and significantly swollen mitochondria likely reflected the accumulation of oxidatively damaged mitochondria along the course. It is of importance to note that no mitochondrial defects were noted in the brain of the Mfn2 cKO mice at 4 weeks of age which excludes the potential complication due to developmental abnormalities. These results clearly demonstrated that Mfn2 ablation-induced mitochondrial fragmentation caused mitochondrial structural damage and dysfunction in the hippocampus and cortex in vivo.
The most striking change in the Mfn2 cKO mice is the apparent neurodegeneration in the hippocampal area at 18 weeks of age where near 90% of the CA1 hippocampal neurons were wiped out. While this extreme loss of neurons was not apparent in individual fields of the cortical area, taken together with the progressive shrinkage of the entire cortex with age, significant cortical neuron loss was apparent at 18 weeks of age when around 25% neurons were lost. Indeed, TUNEL assay did reveal that these neurons die by apoptosis in both the hippocampus and cortex at 18 weeks of age. Significant cortical neuronal loss continues to progress and reached around 50% at 28 weeks of age. It was noted that disabling mitochondrial function could produce the same pathological changes. For example, malonate treatment, which inhibits succinic dehydrogenase, resulted in strikingly similar hippocampal neurodegeneration [
33]. Given that mitochondrial dysfunction far precedes neurodegeneration in this Mfn2 cKO model, our results clearly demonstrated that disrupted mitochondrial fusion could lead to neurodegeneration in AD-affected brain areas through mitochondrial dysfunction.
Interestingly, Mfn2 ablation induced neurodegeneration in the hippocampus and cortex is preceded by a series of pathological events in temporal order that may suggest a causal relationship between these events. Increased oxidative stress was observed in the Mfn2 cKO mice at the age of 8 weeks. It must be noted that oxidative stress is also a prominent and early feature of AD [
34]. For example, advanced glycation end-products, lipid peroxidation, protein nitration and carbonyl formation have all been found to be increased in the brain of AD [
34]. Further induction of HO-1 was determined to be a relatively early neuronal response as its appearance co-localized with the Alz50 tau epitope in degenerating neurons in AD [
35]. At 12 weeks of age, cytoskeletal and neurofibrillary changes became apparent in the Mfn2 cKO mice. As temporally later events, cytoskeletal and neurofibrillary changes likely lie downstream of mitochondrial dysfunction and/or increased oxidative stress. In this regard, ample evidence demonstrated that oxidative stress could impact the posttranslational modification and lead to increased proteolysis of microtubule associated proteins, reduce the ability of microtubule to polymerize and cause severing of actin microfilaments and thus impair cytoskeletal structure in both neuronal and non-neuronal cells [
36‐
38]. Soluble Aβ oligomers has been shown to cause proteolysis of microtubule associated proteins including (MAP2) during apoptosis [
39]. Similarly, oxidative stress causes increased tau phosphorylation and promotes the conformational changes and fibrillation/aggregation of tau protein which leads to AD-related neurofibrillary changes [
35,
40‐
42]. An inflammatory response was also observed in the Mfn2 cKO mice at 12 weeks of age which likely is also caused by mitochondrial dysfunction and/or oxidative stress. The inflammatory response, as shown by increased activation of microglia and astrocytes, precedes the observed neurodegeneration at 18 weeks of age. There are many studies that suggest inflammation as an early sign of AD disease progression. An inappropriate immune response may promote AD by increasing the production of Aβ and reducing removal of amyloid plaques by microglia [
43]. Overall, our study demonstrated that mitochondrial fragmentation caused by disruption in mitochondrial fusion could initiate a cascade of abnormal changes that are relevant to the important pathological changes during the course of AD and lead to neurodegeneration in the hippocampus and cortex in vivo.
In addition to mitochondrial fragmentation and ultrastructural damage, we found an abnormal distribution of mitochondria, especially in the neuronal process, in the pyramidal neurons of Mfn2 cKO mice months before neurodegeneration. Loss of Mfn2 leads to decreased number of mitochondria in the axon and dendritic processes. In this regard, it is important to note that such an abnormal distribution of mitochondria was also noted in the brain of AD patients [
9] and AD animal models [
19]. This observation replicated prior findings where Mfn2 deletion also caused a prominent defect in mitochondrial content and transport in the processes of dopaminergic neurons which set DA neurons to degenerate 1 to 2 months later in retrograde manner [
44,
45]. While it is not clear whether such abnormal mitochondrial distribution contribute to neurodegeneration in the hippocampal and cortical neurons in Mfn2 cKO mice since some neuron population could survive almost complete loss of mitochondria from its processes [
46], it could contribute to neuronal dysfunction such as synaptic deficits of importance to AD as we demonstrated before [
9]. Although altered expression of various mitochondrial fission/fusion proteins could differentially impact mitochondrial transport [
12], Mfn2 may impact mitochondrial transport through its direct interaction with Miro and Milton [
47], adaptor proteins crucial for kinesin-mediated transport of mitochondria.
As we demonstrated in this study, CA1 neurons are susceptible to Mfn2 loss. However, these neurons are resistant to DLP1 loss [
25]. It is of interest to note that, in the contrary, dopaminergic neurons and Purkinje cells are susceptible to both the loss of Mfn2 and DLP1 [
45,
46,
48‐
50], respectively. It is believed that different cells, including various neuron populations, have very different mitochondrial morphology according to their specific metabolic needs, and hence possess unique balance on the regulation of mitochondrial dynamics. Therefore, such a difference in the tolerance of loss of Mfn2 and DLP1 suggests that, comparing to other neuron populations, CA1 neurons more critically depend on mitochondrial fusion than fission for their function and survival which makes mitochondrial fusion a better target for restoring mitochondrial dynamic balance in AD. Alternatively, Mfn2 has been implicated in the regulation of mitochondrial properties besides fusion such as mitochondrial transport [
47]. This could point to a difference in mitochondrial distribution to explain the preferential requirement of Mfn2 in the hippocampus for neuronal survival. However, this appears unlikely since studies of DLP1 loss in dopaminergic neurons show a similar abnormality in mitochondrial movement suggesting that DLP1 may also play a role in mitochondrial transport [
46]. In this regard, the loss of DLP1 in the hippocampus increased the distance between mitochondria in the dendritic processes which was well tolerated, although overall content and number were unchanged [
25].