In Vivo Elevation of Mitochondrial Activity and Amyloid Deposition, but Inversely Correlated, in Early-stage Senescence-Accelerated Mice: A Positron Emission Tomography Study

Background: While marked reductions in neural activity and mitochondrial function have been reported in Alzheimer’s disease (AD), the degree of mitochondrial activity in mild cognitive impairment (MCI) or early-stage AD remains unexplored. Here, we used positron emission tomography (PET) to examine the direct relationship between mitochondrial activity ( 18 F-BCPP-EF) and β-amyloid (Aβ) deposition ( 11 C-PiB), followed by immunohistochemistry for ATPB (an ATP synthase on the mitochondrial inner membrane), in the same brains of senescence-accelerated mouse prone 10 (SAMP10) mice, an Aβ-developing neuroinammatory animal model showing accelerated senescence with deterioration in cognitive functioning similar to that in MCI. Results: The SUVRs of 18 F-BCPP-EF and 11 C-PiB were signicantly higher in the 15-week-old SAMP10 mice than in the control and 5-week-old SAMP10 mice. The two parameters were found to negatively correlate with each other. Consistent with these binding results, the immunohistochemical analysis demonstrated upregulation of ATPB in neurons, astrocytes, and microglia, but not in pericytes, in the 15-week-old SAMP10 mice. Conclusions: The present results provide in vivo evidence of an increase in mitochondrial energy production and amyloid burden at an early stage in SAMP10 mice. The inverse correlation between these two phenomena suggests a concurrent compensatory increase in neuronal energy production and Aβ-induced elevation of neuroinammatory responses in glial cells, as conrmed immunohistochemically. The combination of PET and immunohistochemistry allowed in vivo evaluation of altered mitochondrial activity during neuropathological processes, including Aβ accumulation, in an animal model of AD spectrum disorder.


Introduction
It is well recognized that mitochondrial dysfunction contributes to the neurodegeneration occurring in Alzheimer's disease (AD). Recent ndings suggest that pathological changes that occur in AD brains, such as synaptic and neuronal losses and excessive β-amyloid (Aβ) production, may be induced by mitochondrial dysfunction and increased oxidative stress [1]. In AD patients, mitochondria are reportedly characterized by impaired functioning, including lowered oxidative phosphorylation, decreased adenosine triphosphate production (ATP), increased generation of reactive oxygen species (ROS), and compromised antioxidant defense [2]. Mild cognitive impairment (MCI) is an intermediate condition of impaired cognitive function between normal aging and dementia, and is commonly associated with progression to AD [3,4]. The Aβ deposition rate in patients with MCI who are likely to convert to AD is greater than that in stable patients [5]. However, the in vivo relationship between mitochondrial activity and Aβ level in the state of senescence to MCI remains unclear.
We previously developed an 18 F-BCPP-EF tracer for mitochondrial complex 1, which provides an experimental advantage in that it allows time-course changes in mitochondrial activation to be studied in vivo [6]. Speci cally, this probe can visualize the availability of complex I, the rst component of four electron transport complexes in the inner mitochondrial membrane, and which can be speci cally inhibited by rotenone. Using this probe, we successfully monitored dysfunction in mitochondrial activity in a rat cortical ischemia model [7] and in the parahippocampal region of the early-stage human AD brain [8].
SAMPs (senescence-accelerated mouse prones) are inbred mouse lines showing accelerated aging. There are currently nine independent strains from SAMP1 to SAMP11, which show distinct features of the aging phenomenon. Among these lines, the SAMP10 line shows neuronal loss with Aβ deposition, and impairment of learning and memory due to cortical degeneration in later life [9]. At 8-16 months-of-age, not only are the numbers of neurons reduced, but so are the lengths of dendrites and the spine densities of cortical pyramidal neurons [10]. Preceding this neuronal degeneration, microglia are affected at an earlier stage (~ 3 months-of-age) [11], with the number of segments and tips and the combined lengths of microglial processes being signi cantly decreased. We recently reported that at this stage, during which morphological impairments in microglia occur (i.e., the number of segments and tips and the combined lengths of microglial processes become much reduced), type 2 endocannabinoid receptor (CB2)-positive protective microglia are dominant compared with translocator protein 18 kD (TSPO)-positive in ammatory microglia [12]. Therefore, the merit of using the SAMP10 line at this early stage is that it allows exploration of alterations in molecular events before substantial brain atrophy occurs, which is comparable with the state of the senescence in MCI in humans.
In this study, we used 18 F-BCPP-EF and 11 C-PiB PET tracers and immunohistochemistry to investigate the in vivo relationship between mitochondrial activity and Aβ deposition in the brains of SAMP10 mice during the early stage of neurodegeneration.

Animals
Six 15-week-old senescence-accelerated mouse resistant 1 (SAMR1) mice, which develop normal senescence and are often used as a control line for SAMP mice, and eight 5-week-old and eight 15-weekold senescence-accelerated mouse prone 10 (SAMP10) mice purchased from the SLC Company (Hamamatsu, Japan), were used in this study. The mice were housed with their littermates to a maximum of ve animals in each cage with food and water available ad libitum. All animal protocols and the following experiments were approved by the ethics committees of the Central Research Laboratory at Hamamatsu Photonics and Hamamatsu University School of Medicine. In addition, all applicable institutional and/or national guidelines for the care and use of animals were followed.

Tracer production PET Ligand Syntheses
The HM-18 cyclotron (Sumitomo Heavy Industry, Ltd., Tokyo, Japan) situated at Hamamatsu Photonics PET center was used to produce the positron-emitting radionuclides 11 C and 18 F in 14 N(p,α) 11 C and 18 O(p, n) 18 F nuclear reactions, respectively. The labelled compounds were then synthesized with a modi ed CUPID system (Sumitomo Heavy Industries, Ltd., Tokyo, Japan). 18 F-BCPP-EF was radiolabeled by nucleophilic 18 F-uorination of the corresponding precursor, as described elsewhere [6,7]. The radiochemical purity was more than 99% and the speci c radioactivity was above 50.0 GBq/µmol. 11 C-PiB was synthesized by N-methylation of the nor-compound N-desmethyl-PIB with 11 C-methyl tri ate [13]. The radiochemical purity was more than 96% and the speci c radioactivity was above 35 GBq/µmol.

PET measurements
PET was performed using a high-resolution animal PET scanner (SHR-38000, Hamamatsu Photonics, Hamamatsu, Japan) with an axial eld of view (FOV) of 108 mm, a transaxial FOV of 330 mm, and a transaxial spatial resolution of 2.3 mm in the center. Animals were scanned twice a day, rst with 11 C-PiB, then 2 hours later with 18 F-BCPP-EF; the order of the tracers was not counterbalanced because of radiotracer half-life issues. Iso urane at 1.5-2.0% in oxygen was used to anesthetize the mice for the duration of the entire scans. The animals were placed in the prone position on a xation plate and were then set within the gantry hole of the PET scanner. After a 15-minute transmission scan using an external 68 Ge/ 68 Ga rod source (67 MBq) for attenuation correction, serial emission scans lasting for 70 min and 80 min were performed immediately following injections of 11 C-PiB at a dose of 8 MBq and 18 F-BCPP-EF at a dose of 5 MBq, respectively. The tracers were injected intravenously through a cannula inserted into the tail vein. No arterial sampling was conducted. The PET data were reconstructed using 3D DRAMA (iteration 2, gamma 0.1) with a Gaussian lter of 1.0 mm full width at half maximum (FWHM), yielding a voxel size of 0.65 × 0.65 × 1.0167 mm for the reconstructed images. To obtain anatomical information, Xray CT scans were performed immediately following the PET measurements, using a ClairvivoCT (Shimadzu Corporation, Kyoto, Japan).

Immunohistochemistry
Immunostaining was performed as previously reported [14]. Brie y, mice were anesthetized with chloral hydrate (400 mg/kg) and transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA; pH 7.4). Their brains were removed, post-xed in 4% PFA, and immersed in 30% sucrose in PBS as cryoprotectant until the tissue sank. Tissues were then frozen in dry ice and stored at − 80 °C. Frozen coronal sections of 20-µm thickness were cut using a cryostat. The slides were blocked with 10% donkey serum in PBS containing 0.1% Triton X-100 for 1 hour at room temperature (RT), followed by incubation overnight at 4 °C with primary antibodies. After washing with PBS for three times, the slides were incubated for 1 hour at RT with uorescent conjugated secondary antibodies. Then, they were washed three times with PBS and stained with DAPI to visualize nuclei. Fluorescent images of a single focal plane were obtained by confocal microscopy using a 63 × lens (SP8, Leica, Wetzlar, Germany). The following primary antibodies were used in this study: mouse anti-ATPB (1:200, Abcam, Tokyo, Japan), rabbit anti-CaMKII (1:500, Abcam), rabbit anti-GFAP (1:500, DAKO/Agilent, Santa Clara, US), rabbit anti-Iba1 (1:500, WAKO, Saitama, Japan), and rat anti-PDGFRβ (1:250, kindly provided by Prof. Takakura, Osaka University). The secondary antibodies were as follows: Alexa Fluor 488 anti-rabbit IgG and anti-mouse IgG, Alexa Fluor 568 anti-rabbit IgG, Alexa Fluor 594 anti-mouse IgG, and Alexa Fluor 647 anti-goat IgG (1:500, Thermo Fisher Scienti c, Waltham, USA).

Data and statistical analyses
The PET data were analyzed with PMOD image software (version 3.7; PMOD Technologies Ltd, Zurich, Switzerland). The SUVRs for 11 C-PiB and 18 F-BCPP-EF were estimated by dividing the target SUV by the cerebellar SUV, with the latter being taken to indicate the background level [8,15]. The SUVs were calculated as the measured radioactivity divided by the ratio of the total injected dose to the mouse body weight. As described elsewhere [12,16], ellipsoid volume of interest (nearly 2.9 mm x 0.8 mm x 1.5 mm in diameter) ranging from 14 to 16 mm 3 were placed over the frontal cortex anteriorly under the Bregma by referring to the X-ray CT images [12] (Supplementary Fig. 1). One-way analysis of variance (ANOVA) was performed to compare tracer uptake and mouse age, with the signi cance level set at p < 0.05 with a correction for multiple comparisons. Within each age group, correlation analysis was performed between the two tracer SUVRs ( 11 C-PiB SUVR and 18 F-BCPP-EF SUVR, at either 5 or 15 weeks-of-age) using the false discovery rate (FDR) correction for multiple correlations (p < 0.05), to examine deviations in patterns of the parameters in the living brains of SAMP10 mice in relation to the progression of senescence.

PET ndings
We rst analyzed the mitochondrial activity in the brains of SAMR1 mice at 15 weeks-of-age and SAMP10 mice at 5 and 15 weeks-of-age using 18 F-BCPP-EF. Figure 1 shows the parametric PET images of 18 F-BCPP-EF uptake superimposed on CT images of 15-week-old SAMR1 mice (A) and 5-week-old (B) and 15-week-old (C) SAMP10 mice. The SUVRs of 18 F-BCPP-EF in the brain did not show a signi cant difference between SAMR1 mice and 5-week-old SAMP10 mice (Table 1). By contrast, in 15-week-old SAMP10 mice, the SUVR was higher throughout the brain (Fig. 1C), being signi cantly higher than that of the 5-week-old SAMP10 mice ( Fig. 2A), meaning that mitochondrial oxidative metabolism had increased in the SAMP10 mice around the period of 15 weeks-of-age. Next, we analyzed the Aβ level using 11 C-PiB. The SUVRs of the control SAMR1 mice and 5-week-old SAMP10 mice did not show a signi cant difference (Fig. 1D-E and Fig. 2B). By contrast, the 15-week-old SAMP10 mice (Fig. 1F) showed a signi cantly higher 11 C-PiB SUVR than the 5-week-old mice (Fig. 2B), indicating that Aβ accumulation is detectable by PET at around 15 weeks of age. The lack of a signi cant difference between the 15-weekold SAMP10 and SAMR1 mice was due to a large variation in the 11 C-PiB SUVR in the SAMR1 mice.
Direct comparisons between the 18 F-BCPP-EF and 11 C-PiB SUVRs showed a signi cant negative correlation in 15-week-old SAMP10 mice, but no signi cant correlation in either 5-week-old SAMP10 or 15-week-old SAMR1 mice. This nding indicates that a greater Aβ burden reduces mitochondrial availability in 15-week-old SAMP10 mice, and vice versa.

Immunohistochemical ndings
The PET nding of increased mitochondrial activity at 15 weeks-of-age in SAMP10 mice was followed by a con rmatory examination using immunohistochemical analyses. Double immunostaining against ATPB (a component of ATP synthase on the mitochondrial inner membrane) and CaMKII (a marker for neurons) was performed. In the control mice (15-week-old SAMR1 mice and 5-week-old SAMP10 mice), there were a small number of ATPB positive signals in the soma of neurons in the cerebral cortex ( Fig. 3A-F). By contrast, in the 15-week-old SAMP10 mice, the intensity and numbers of ATPB-positive signals were dramatically elevated ( Fig. 3G-I), which was consistent with the 18 F-BCPP-EF PET nding. The immuno-positive punctuate signals were observed not only in the soma, but also in neurites. This result suggests that neurons are poised to increase ATP production in the early stage of neurodegeneration.
As ATPB immuno-positive signals were also observed in non-neuronal cells, as shown in Fig. 3G, we next analyzed the localization of ATPB in astrocytes. In 15-week-old SAMR1 mice, the GFAP signal was mainly observed in the endfeet of perivascular astrocytes in the cerebral cortex ( Fig. 4A-C), where we also found that ATPB was localized. While the ATPB immuno-positive signal was very low in 5-week-old SAMR1 mice (Fig. 4D-F), both the ATPB signal and the overall GFAP signal were much higher throughout the cerebral cortex in the 15-week-old SAMP10 mice, indicating proliferation of reactive astrocytes consistent with previous reports (Fig. 4G-I, data not shown) [17]. As expected, the ATPB signal was colocalized within the GFAP + astrocytes. Further analysis of the upregulation of ATPB in pericytes was performed, because many ATPB immuno-positive signals were found along a capillary ( Supplementary Fig. 2), and progressive pericyte loss in a mouse model of AD pathogenesis was previously reported [18]. However, most of the ATPB signals along the capillary were located in astrocytes, not in pericytes ( Supplementary  Fig. 2). Figure 5 shows the results of investigations into whether ATPB was localized to microglia. In the control mice (15-week-old SAMR1 and 5-week-old SAMP10 mice), the intensities of the ATPB immuno-positive signals were too weak to detect in Iba1 + microglia. By contrast, a detectable level of ATPB-positive signal was colocalized with Iba1 + microglia in the 15-week-old SAMP10 mice. The microglia in these 15-weekold SAMP10 mice may have been protective microglia, as we recently reported that CB2 + microglia activation was dominant at an early stage of neurodegeneration [12]. Next, we analyzed the expression of TREM2, which reportedly has a protective function to prevent AD progression [19,20]. While the TREM2 immuno-positive signal was very low in the microglia in the control mice ( Fig. 6A-F), intensive TREM2 immuno-positive punctuate signals were observed in Iba1 + microglia in 15-week-old SAMP10 mice ( Fig. 6G-I). This result supports our previous nding of the activation of protective microglia in SAMP10 mice at this early-stage, which is comparable to the state of senescence in early-stage MCI in humans.

Detection of mitochondrial activity and amyloid deposition in SAMP10 mice
In this study, we showed that the SUVRs of 18 F-BCPP-EF and 11 C-PiB in the cerebral cortex were signi cantly higher in 15-week-old SAMP10 mice than in control mice, and that these SUVRs were inversely correlated with each other in 15-week-old SAMP10 mice ( Figs. 1 and 2). To our knowledge, this is the rst study to report a change in mitochondrial activity in SAMP10 mice on in vivo PET imaging, highlighting elevated mitochondrial activity at an early stage in SAMP10 mice, comparable to the early stage of AD spectrum disorder in humans. Although many researchers have investigated frontal lobe atrophy and Aβ-deposition using SAMP10 mice older than 7 months, we here used younger mice to analyze early molecular changes at the beginning of neurodegeneration. Aβ-deposition was not obvious at 5 weeks-of-age, but we did observe an elevation in the SUVR of 11 C-PiB in 15-week-old SAMP10 mice.
As the SUVR of 18 F-BCPP-EF, which re ects oxidative phosphorylation in mitochondria, was also increased at the same stage ( Figs. 1 and 2), it is probable that metabolism had increased to compensate for an energy loss in neurons, which happens with mitochondrial dysfunction in the pathological condition of Aβ-deposition in the AD brain. Although the pathological process is different, this compensatory theory is in line with previous reports of elevated dopamine synthesis in early Parkinson's disease [21,22]. Of course, in a later stage of disease, reduced energy production would occur in association with more serious pathological events, as neurodegeneration progresses [23,24].
A previous metabolomic study showed a difference in the metabolic pathway between MCI and AD patients [25], with the level of pyruvate in cerebrospinal uid being signi cantly higher in MCI patients than in cognitively normal (CN) individuals and AD patients. It is also reported that gene expression of complex I-V subunits in the electron transport chain is elevated in MCI compared with AD and agematched CN individuals [26]. Thus, the increase in mitochondrial activity in 15-week-old SAMP10 mice in the current study may re ect these changes in energy metabolism found in the state of senescence to MCI.

Page 8/21
A Possible Key Player: Glial Cells A possible key event that might be at play in the inverse correlation between Aβ deposition and mitochondrial activity is neuroin ammation, speci cally neuroprotective glial activation. Aβ accumulation affects neurons and neuroin ammatory cells, and upregulation of IL-1β and IFN-γ in 3month-old SAMP10 mice, and IL-6 in later-stage SAMP10 mice, have been observed [27]. In our recent study, neuroprotective microglia were more dominant than neuroin ammatory microglia in the early stage of neurodegeneration in SAMP10 mice [12]. We showed that elevation of mitochondrial energy metabolism, or oxidative phosphorylation, occurred according to the progression of neurodegeneration (Figs. 1 and 2A). Although the high energy requirements of neurons means that they form a major contribution to the oxidative metabolism of the brain, glial cells are also responsible for some of the oxidative metabolism. The oxidative metabolism in microglia changes as symptoms progress. While induction of M1-type in ammatory microglia by lipopolysaccharide leads to a reduction of mitochondrial oxygen consumption and lactate production, these reductions are not caused by IL-4/IL-13, inducers of M2-type protective microglia [28]. This indicates the occurrence of higher metabolism in M2 type microglia. When protective microglia are dominant in the early stage of neurodegeneration, oxidative phosphorylation in mitochondria remains at a high level. However, as the number of protective microglia decreases and neuroin ammatory microglia become prominent with the progression of Aβ deposition in later stages, the oxidative phosphorylation activity in microglia will be reduced by mitochondrial dysfunction, concomitant with that in neurons [28].

Interaction Of Amyloid And Neuronal Mitochondria
One of the mechanisms by which amyloid leads to mitochondrial dysfunction is the transportation of APP into mitochondria [29]. Pre-sequence protein (PreP), a processing enzyme that recognizes mitochondrial-targeting signal peptides and cleaves after protein import, can degrade Aβ in the mitochondria [30]. Interestingly, the proteolytic activity of PreP is decreased in the AD brain [31]. Aβ can be localized to the inner mitochondrial membrane [32], and constituents of the γ-secretase complex, such as nicastrin, APH-1, PEN-2, and presenilin-1, which function in APP processing, are also localized within the mitochondria-associated membrane [33,34]. It has been reported that mitochondrial dysfunction and neurodegeneration occur in model mice with deleted HtrA2, a serine protease that interacts with Aβ, APP, and presenilin-1 within the intermembrane space [35]. These pieces of evidence suggest that Aβ and Aβrelated enzymes are linked with failure of mitochondrial function.

Immunohistochemical Con rmation
Our immunohistochemical analyses presented in Figs. 3 to 5 show that in 15-week-old SAMP10 mice, mitochondrial ATPB, a key enzyme for ATP production, was mainly present in neurons, although some was present in microglia, and slightly elevated levels were also present in reactive/perivascular astrocytes. Because the population of neurons was most abundant in the cerebral cortex and upregulation of ATPB in neuron was prominent compared to astrocytes and microglia, the primary contributor to the higher SUVR of 18 F-BCPP-EF in the 15-week-old SAMP10 mice was considered to be neurons. As the number of GFAP + reactive astrocytes in the cortex had increased (data not shown) [17], and speci c CB2 + protective microglia were activated at this early stage in the SAMP10 mice [12], we speculate that the extent of the polarized neuroin ammatory responses (neurotoxic or neuroprotective) of these glial cells would be of relevance to future neuronal degeneration. The protective cytokines (including neurotrophic factors) that are released from microglia exposed to neuropathic substances such as Aβ might stimulate neurons to supply more glucose and glutamine from perivascular astrocytes. These supplies may enable neurons to survive by temporarily increasing energy production.
The blood-brain barrier breakdown caused by pericyte dysfunction and impairment of platelet-derived growth factor receptor-β (PDGFRβ) signaling have recently been attracting attention as pathological features of AD. Pericytes are involved in the e ux of accumulated Aβ in the brain. Originally, we thought that pericyte activity might be elevated at an early stage of neurodegeneration; however, in 15-week-old SAMP10 mice, there was no increase in the immunostaining level of ATPB in pericytes, while adjacent astrocytes contained a higher level of ATPB ( Supplementary Fig. 2). In this relatively early stage animal model, the contribution of pericytes to accumulation of Aβ in the brain parenchyma may be minimal. Indeed, the number of pericytes starts to decrease after 4 months-of-age in 5xFAD mice [18].
We also observed elevation of TREM2 (an important protein for clearance of Aβ) in microglia in 15-weekold SAMP10 mice (Fig. 6). TREM2 is a key player in the switching of microglia from a homeostatic state to a disease-associated state. Interestingly, soluble TREM2 in cerebrospinal uid is higher in Aβ + Tau + MCI patients than in CN individuals [36]. Furthermore, TREM2 expression in mononuclear cells in the peripheral blood of MCI patients, especially those likely to convert to AD, was signi cantly higher than in CN individuals [37]. TREM2 activates the mTOR pathway that regulates mitochondrial energy production by promoting the synthesis of mitochondrial proteins, including components of MC-1 and MC-5 [38]. Therefore, the correlation between mitochondrial activity and TREM2 expression is reasonable.

Conclusion
In this study, we found that the SUVRs of 18 F-BCPP-EF and 11 C-PiB were signi cantly higher in the brains of 15-week-old SAMP10 mice than in 5-week-old SAMP10 mice and 15-week-old SAMR1 mice, indicating an increase in mitochondrial activity and Aβ deposition occurring in the brain in the early state of senescence towards cognitive impairment and AD spectrum-type neurodegeneration. Interestingly, the SUVR of 18 F-BCPP-EF was negatively correlated with that of 11 C-PiB. Consistent with this PET analysis, we also found that ATPB was strongly upregulated, mainly in the neurons, in the cerebral cortex of 15week-old SAMP10 mice, along with elevated activation of astrocytes and microglia. Hence, Aβ-induced neuroin ammation may shift the net production of ATP from neuronal oxidative metabolism to anaerobic glycolysis, as supported by the negative correlation between mitochondrial activity and Aβ accumulation.
This contention could lead to a therapeutic expectation that a speci c treatment to sustain mitochondrial activity might help ameliorate senescence-related neuroin ammation and degeneration.

Limitations
There are several limitations of note in the current study. First, although we measured changes in the uptake of both 18 F-BCPP-EF and 11 C-PiB in 5-and 15-week-old SAMP10 mice and control SAMR1 mice, the PET data did not reveal the mitochondrial activities of any particular cell types: neurons, microglia, oligodendrocytes and/or astrocytes. As the present study focused on changes in early-stage SAMP10 mice, observations at a later stage might reveal different expression of in ammatory substances and cells; hence, a further study is needed to address this issue in the broader time course of senescencerelated neuronal loss. Second, because the spatial resolution of the PET scanner used was only 2.3 mm, the results may be subject to partial volume effects. To reduce such effects, we tried to set the measurement VOIs to at least twice the size of the FWHM of the scanner. An alternative would be to use autoradiography instead of PET, although measures of parameter change within the same animal cannot be obtained using autoradiography. Third, as we did not make additional PET measurements using a tracer for neuroin ammation in the same animals, we could not evaluate the direct contribution of neuroin ammation to the change in mitochondrial activity. However, taking our recent PET results on neuroin ammation [12] into consideration, we consider an in vivo relationship to be likely, as mentioned above.  The SUVRs of 18F-BCPP-EF (A) and 11C-PiB (B) tracers and the relationships between them (C-E). The 18F-BCPP-EF and 11C-PiB (B) SUVRs were signi cantly higher in the 15-week-old SAMP10 mice than in the controls (A and B, *p < 0.05). A negative correlation between the binding of the two tracers was found in the 15-week-old SAMP10 mice (E). The dotted lines in (E) represent the 95% con dence intervals for the correlation.

Figure 3
Double immunostaining for ATPB (green) and CaMKII (red) in the cerebral cortex of 15-week-old SAMR1 mice (A-C) and 5-week-old (D-F) and 15-week-old (G-I) SAMP10 mice. Note that the ATPB signal had greatly increased and was localized to the neurons at 15-weeks-of-age. Arrowheads indicate ATPB signals. Scale bar: 10 μm.

Figure 4
Immunostaining for ATPB (green) and GFAP (red) in the cerebral cortex of 15-week-old SAMR1 mice (A-C) and 5-week old (D-F) and 15-week-old (G-I) SAMP10 mice. Note that both ATPB and GFAP immunoreactivity were much higher in the 15-week-old SAMP10 mice than in the two control groups. The localization of ATPB was observed in GFAP+ astrocytes. Arrowheads indicate ATPB signals. Scale bar: 10 μm.

Figure 5
Double immunostaining for ATPB (green) and Iba1 (red) in the cerebral cortex of 15-week-old SAMR1 mice (A-C) and 5-week old (D-F) and 15-week-old (G-I) SAMP10 mice. Note that the ATPB signal was much higher and localized to the microglia in the 15-week-old SAMP10 mice. Arrowheads indicate ATPB signals. Scale bar: 10 μm.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.