Optogenetic Targeting of Astrocytes Restores Slow Brain Rhythm Function and Slows Alzheimer’s Disease Pathology

Patients with Alzheimer’s disease (AD) exhibit non-rapid eye movement (NREM) sleep disturbances in addition to memory deficits. Disruption of NREM slow waves occurs early in the disease progression and is recapitulated in transgenic mouse models of beta-amyloidosis. However, the mechanisms underlying slow-wave disruptions remain unknown. Because astrocytes contribute to slow-wave activity, we used multiphoton microscopy and optogenetics to investigate whether they contribute to slow-wave disruptions in APP mice. The power but not the frequency of astrocytic calcium transients was reduced in APP mice compared to nontransgenic controls. Optogenetic activation of astrocytes at the endogenous frequency of slow waves restored slow-wave power, reduced amyloid deposition, prevented neuronal calcium elevations, and improved memory performance. Our findings revealed malfunction of the astrocytic network driving slow-wave disruptions. Thus, targeting astrocytes to restore circuit activity underlying sleep and memory disruptions in AD could ameliorate disease progression.

Additionally, sleep disturbances manifest early in the disease progression. Disruptions in deep non-rapid eye movement (NREM) slow-wave sleep (SWS) are evident in patients with mild AD 15,16 and in individuals with mild cognitive impairment (MCI) 17 . The cortico-thalamic circuit generates slow waves as oscillations alternating between DOWN states, periods of relative quiescence, and UP states, periods of vigorous neuronal activity 18-20 . These slow oscillations, or slow waves of < 1 Hz, are prevalent during NREM sleep 21 and are disrupted in patients with AD 22 as well as in individuals with MCI 17 . Aβ is associated with decreased slow-wave activity (SWA) in asymptomatic adults, and reduced NREM SWA is correlated with increased Aβ deposition in the medial prefrontal cortex 23 . Reduced slow-wave power is also associated with tau pathology 15 . Thus, disturbances in SWA present a potential early biomarker for AD 24,25 .
Slow-wave disruptions observed in AD are recapitulated in mouse models of beta-amyloidosis, in which slow-oscillation power is downregulated by Aβ in young APP mice 26 . Moreover, disruptions in slow oscillations actively contribute to the disease progression 24,26,27 . Slow oscillations play a critical role in consolidation of declarative memories during deep NREM sleep [28][29][30][31][32][33] . Slow oscillations drive the reactivation of short-term hippocampal memories by synchronizing spindle activity in the thalamus, thereby contributing to the long-term plastic changes in neocortical networks and supporting the consolidation of long-term memory in neocortex 34,35 . Boosting slow waves enhances memory performance 30 , while disruptions of slow oscillations are linked to memory impairments 23 . Similarly, memory function is impaired in APP mice [36][37][38] . Since slow oscillations play an important role in declarative-memory consolidation during sleep, therapeutic targeting of slow oscillations could restore sleep and memory de cits as well as slow AD progression.
Little is known about the underlying mechanisms that cause the disruption of slow waves. One of the possible mechanisms is, as Aβ targets synapses, alterations of excitatory and inhibitory neurotransmission contribute to neuronal circuit dysfunctions 26, 39,40 . Both excitatory and inhibitory neurons participate in generation of slow oscillations 18,20,[41][42][43][44][45] . Besides neurons, astrocytes actively contribute to circuit function 46,47 . Astrocytes form the tripartite synapses with pre-and post-synaptic neuronal compartments to regulate synaptic transmission via astrocytic calcium signaling [48][49][50] . Yet, whether astrocytes contribute to the disruption of slow oscillations in APP mice remains unclear.
Astrocytes are electrically silent cells that exhibit calcium transient elevations in response to neuronal stimulation 51 . Importantly, astrocytes are implicated in SWA 52 , and astrocytic calcium transients are synchronized to the neuronal UP states of slow oscillations in healthy anesthetized rats 53 . In addition, more astrocytes than neurons are active during the initiation of SWA, indicating that astrocytes participate in the build-up of slow waves 53 . Electrical stimulation of individual astrocytes increases the activity of the local astrocytic networks and triggers the synchronous UP state transitions in neighboring neurons within somatosensory cortex 54 . Thus, astrocytes actively participate in SWA.
Elevations in resting astrocytic calcium levels were reported in APP mice 55,56 , suggesting that disruption of astrocytic activity may contribute to the cortico-thalamic circuit dysfunction driving slow waves in AD.
Therefore, understanding the role of astrocytes in the disruption of cortical SWA is important. Here, we hypothesized that aberrant astrocytic activity contributes to the disruption of the cortico-thalamic circuit function, such that restoration of astrocytic function will slow AD progression in APP mice. To test this hypothesis, we monitored the calcium transients in individual astrocytes using high-resolution multiphoton microscopy in vivo. We discovered the power of astrocytic calcium transients was decreased in young APP mice. Importantly, optogenetic targeting of astrocytes restored slow oscillations, decreased amyloid deposition, normalized neuronal calcium homeostasis and rescued sleep-dependent memory function in APP mice. Our ndings provided evidence for aberrant astrocytic activity contributing to slow wave disruptions. Thus, astrocytic targeting of SWA emerges as a novel therapeutic target for the prevention or treatment of AD.

Results
Astrocytic calcium transients associated with slow oscillations are disrupted in young APP mice. To determine whether astrocytic calcium signaling was altered in 4-6-month-old APP mice, we monitored calcium transients within individual cortical astrocytes using imaging with multiphoton microscopy. We targeted the genetically encoded calcium indicator Yellow Cameleon 3.6 (YC3.6) to astrocytes in the somatosensory cortex via the GFAP promoter, since cortical slow oscillations traverse somatosensory cortex 57 and slow oscillations are disrupted when imaged in somatosensory cortex 26 . YC3.6 expression was visualized in the somas, processes, and microdomains of astrocytes in non-transgenic control (NTG) and APP mice (Fig. 1A). YC3.6 is a ratiometric (R) probe. Representative ΔR/R 0 traces are shown in Fig. 1B. Here we focused the analyses on the low amplitude, high frequency astrocytic calcium transients that were implicated in slow oscillations 53 (Fig. 1B, blue box), not large amplitude, low frequency astrocytic calcium transients 55 (Fig. 1B, red box). Next, we used Fourier transform analysis to characterize the astrocytic calcium transients. The percentage of astrocytes cycling at the slow-wave frequency (0.2-1 Hz) was high and did not differ signi cantly between APP and NTG mice (Supplemental Fig. 1, 72.11 ± 3.01% of astrocytes in NTG vs 70.15 ± 2.93% of astrocytes in APP mice; n = 300 cells in 7 NTG mice, n = 282 cells in 7 APP mice). The average power spectral density plots of somatic calcium transients for NTG (n = 300 cells in 7 mice) and APP mice (n = 282 cells in 7 mice) are shown in Fig. 1C.
The calcium transient power, which is proportional to the square of the amplitude [A] 2 , at the slow-wave frequency was signi cantly lower in astrocytic somas, processes and microdomains of APP mice compared to that in NTG controls (Fig. 1D Fig. 1F, 0.032 ± 0.00022 for NTG vs 0.020 ± 0.00015 for APP; ****p < 0.0001). The frequency at peak power remained comparable in astrocytic somas, processes, and microdomains across genotypes (Fig. 1D Optogenetic activation of cortical astrocytes restores slow waves in young APP mice. We had previously shown that APP mice exhibit disruptions in SWA compared to NTG controls measured with voltage sensitive dyes 26 . The power of slow waves was signi cantly lower in APP mice compared to NTG mice starting at 3 months of age, while the frequency remained comparable at 0.6 Hz 26 . Since astrocytes actively contribute to slow oscillations and astrocytic calcium transients associated with slow waves are disrupted, we tested whether optogenetic activation of astrocytes would restore neuronal slow oscillations in APP mice. 3-5-month-old APP and NTG mice received viral delivery of ChannelRhodopsin2-mCherry (ChR2) or mCherry lacking ChR2. Viral vectors targeted astrocytes via the GFAP promoter in the left hemisphere ( Fig. 2A).
In addition to ChR2 or mCherry injections in the left hemisphere, animals received viral injections of a voltage sensor Voltron and AAV9-hsyn-Cre into the contralateral somatosensory cortex ( Fig. 2A).
Subsequently, a cannula was implanted over the ChR2/mCherry injection site and a 5mm cranial window was installed over the right somatosensory cortex expressing Voltron (Fig. 2B, C). Cortical slow waves were monitored using wide-eld microscopy, before and during light activation of ChR2/mCherryexpressing astrocytes (Fig. 2D-F). Since slow oscillations propagate as traveling waves, optogenetic stimulation in anterior left cortex was expected to activate the local circuit that would then propagate the wave into the contralateral hemisphere, where slow waves would be monitored with voltage sensors.
To determine whether optogenetic stimulation of astrocytes during slow wave sleep (SWS) impacted the cortical local eld potential (LFP), mice underwent injection of ChR2 in the left hemisphere and implantation of a stereotrode array in the contralateral hemisphere, targeting the region imaged with Voltron. The cortical LFP was recorded as animals slept and underwent optogenetic stimulation, and SWS was identi ed using the theta delta ratio in the setting of prolonged immobility in a sleep posture. Compared to baseline LFP, optogenetic stimulation of ChR2-expressing astrocytes at 0.6 Hz was associated with a small but signi cant increase in LFP power at 0.6 Hz (n = 3 mice, p = 0.030, 1-tailed paired t-test; Supplemental Fig. 2A-C). These results indicated that optogenetic targeting of astrocytes could be used to successfully manipulate neuronal slow oscillations in APP mice.
The rhythmic optogenetic stimulation of astrocytes decreases amyloid beta deposition. Since optogenetic targeting of astrocytes restored slow waves, and SWA actively contributed to the clearance of Aβ 58 , we next determined whether optogenetic restoration of SWA via astrocytes could slow the accumulation of Aβ in APP mice. To that end, ChR2-or mCherry-expressing cortical astrocytes were stimulated with light at 0.6 Hz, 24 hours per day continuously for 2 or 4 weeks in aged APP mice. We decided to stimulate during the day and night, since prior work showed APP mice exhibiting de cits in NREM sleep during the day and night 59,60 . At the end of the light treatment, cranial windows were installed over the right posterior cortex, contralateral to the ChR2/mCherry site. High-resolution multiphoton microscopy allowed monitoring of methoxy-X04-positive amyloid plaques after treatment in vivo (Fig. 3A). Multiphoton images revealed the presence of amyloid plaques in 8-10-month-old APP mice (Fig. 3B). Restoring slow oscillations decreased amyloid deposition in the ChR2 group compared to mCherry controls (Fig. 3B). The frequency distribution of plaque size is shown in Fig. 3C. The size of the plaques was not signi cantly different across conditions (Fig. 3C). However, amyloid plaque burden, which takes into account plaque numbers and size, was signi cantly lower after optogenetic treatment compared to light activation of mCherry ( Fig. 3D; 1.43 ± 0.11% burden for mCherry vs 1.01 ± 0.06% burden for ChR2; mCherry: n = 57 elds of view (FOV) in 8 APP mice; ChR2: n = 82 FOV in 10 APP mice; **P < 0.01). Also, the number of amyloid plaques was signi cantly lower in the ChR2 group compared to that in mCherry controls ( Fig. 3E; 185.9 ± 14.44 plaques/mm 3 for mCherry vs 139.0 ± 5.69 plaques/mm 3 for ChR2; mCherry: n = 57 FOV in 8 APP mice; ChR2: n = 82 FOV in 10 APP mice; *P < 0.05). Therefore, astrocyte-dependent rescue of slow waves reduced the rate of amyloid plaque deposition in APP mice. Random optogenetic stimulation of astrocytes failed to signi cantly alter amyloid plaque number or amyloid plaque burden in APP mice compared to non-treated APP mice (Supplemental Fig. 4A, B).
The rhythmic optogenetic stimulation of astrocytes normalizes neuronal calcium levels in APP mice. Intracellular calcium levels are tightly regulated for healthy neuronal function. APP mice exhibit disruptions in neuronal calcium homeostasis that result in elevated resting calcium levels, or calcium overload 61,62 . Calcium levels were monitored using multiphoton microscopy after 2-4 weeks of optogenetic stimulation. The genetically encoded ratiometric calcium reporter YC3.6 under the CBA promoter was used to target cortical neuronal processes, neurites. YC3.6 allows the determination of absolute calcium concentration in neurons, as well as dynamic changes. We determined whether light activation of ChR2-expressing astrocytes would alter neuronal calcium levels in APP mice. Thus, neuronal calcium was used as a functional readout of treatment e cacy. While the majority of neuronal processes, or neurites, exhibited normal calcium levels in APP mice (Fig. 4A, blue neurons), some neurites exhibited abnormally high calcium (Fig. 4A, red neuron, white arrows). Restoration of slow waves led to a decrease in the percentage of neurons with calcium overload (Fig. 4B). A YC3.6 ratio greater than two standard deviations above the mean value (> 1.73) in NTG neurons was de ned as calcium overload (Fig. 4C red box) 26,27,61,62 . The percentage of neuronal processes exhibiting calcium overload was lower in the ChR2 group compared to the mCherry group (Fig. 4C, D; 12.65 ± 1.8% for mCherry vs 5.79 ± 1.2% for ChR2; mCherry: n = 2182 neurites in 23 FOV of 7 mice; ChR2: n = 2014 neurites in 24 FOV of 6 mice; *p < 0.05).
Thus, astrocyte-dependent rescue of slow waves decreased neuronal calcium overload and restored neuronal calcium homeostasis in APP mice.
The same mice were subjected to a locomotor test to determine whether increases in freezing could be attributed to de cits in locomotor activity. The distance travelled in the open eld was not signi cantly different for the four groups ( Fig. 5D; 2754 ± 285.3cm for NTG mCherry vs 2785 ± 110.2cm for NTG ChR2, 3329 ± 285cm for APP mCherry vs 3125 ± 346.1cm for APP ChR2). The average velocity of the APP mice did not show signi cant group differences ( Fig. 5E; 4.59 ± 0.48cm/s for NTG mCherry vs 4.52 ± 0.20cm/sfor NTG ChR2, 5.61 ± 0.47cm/s for APP mCherry vs 5.21 ± 0.58cm/s for APP ChR2). Our data indicated that the locomotion of the APP mice was not impaired by optogenetic treatment. Thus, slow wave rescue via optogenetic targeting of astrocytes improved sleep-dependent memory consolidation in APP mice.
We subsequently veri ed that optogenetic stimulation of astrocytes did not result in neurodegeneration in APP mice (Supplemental Fig. 5). Thus, blue light did not result in signi cant phototoxicity.

Discussion
Patients with AD exhibit memory impairments and sleep disruptions. Slow oscillations, important for memory consolidation during sleep, are impaired in patients with AD and mouse models of betaamyloidosis. Here we determined whether astrocytes contribute to slow-wave disruption in APP mice.
It is increasingly recognized that astrocytes, in addition to modulating local synaptic plasticity, control the activity of neural networks. Astrocytes actively participate in circuit function by interacting with pre-and post-synaptic neurons via tripartite synapses 48-50 and by communicating with neighboring astrocytes via gap junctions 63-65 . As each individual astrocyte interacts with a multitude of neurons, astrocytes exhibit calcium transients indicative of the activity of the circuit in which that astrocyte is participating. We monitored astrocytic calcium transients of low amplitude and high frequency (0.02-1Hz). We discovered decreases in the power of calcium transients in astrocytes concomitant to aberrant slow oscillations in young APP mice. This nding was in contrast with elevated astrocytic calcium activity of high amplitude and low frequency (~ 0.005Hz) reported in APP mice 55 . Nevertheless, aberrant calcium activity was consistent with the reports of disrupted astrocytic calcium signaling in mouse models of beta-amyloidosis 55,56 . Astrocytic calcium transients of APP mice exhibited maintenance of the slowwave frequency but showed decreases in power starting at 3 months of age, before plaque deposition, yet when signi cant oligomeric Aβ is present in brain parenchyma 26 . This suggests that disruption of astrocytic calcium transients is an early event in the disease progression.
Since astrocytic calcium transients were disrupted, we used optogenetics to stimulate cortical astrocytes at the slow-wave frequency to allow the circuit to reach UP states of high power in APP mice. In doing so, we tested whether optogenetic targeting of astrocytes could restore slow waves. We showed that optogenetics allowed targeting of astrocytes with high delity and spatiotemporal precision within the intact cortico-thalamic circuit driving slow waves in vivo. Slow oscillations were restored by activation of ChR2-expressing astrocytes in APP mice. This suggests that even though slow oscillations are impaired, the structural integrity of the glial-neuronal circuit remains intact. At present, it is unclear whether optogenetic activation of astrocytes leads to restoration of global astrocyte calcium activity that propagates the slow wave signal. Alternatively, optogenetic activation of astrocytes could lead to restoration of local slow wave activity which then propagates via neurons. Finally, both neurons and astrocytes could participate in propagation of slow wave activity. Another limitation of the study was continuous optogenetic stimulation for 14-28 days without respect for natural wake/sleep states of the mouse. Future studies using closed-loop stimulation paradigm where stimulation is limited to bouts of NREM sleep will be necessary to address this limitation. Taken together with our earlier studies, it is clear that optogenetic stimulation of astrocytes as well as neurons leads to restoration of slow wave activity in APP/PS1 mice, suggesting that astrocytes interact with neurons when maintaining slow wave activity 24,26,27 .
APP mice overexpress a human mutant APP transgene and a human mutant PS1 variant, generate Aβ and deposit amyloid plaques starting at 4.5-5 months of age 66 . Accumulation of Aβ results in neuronal dysfunction and circuit impairments 6,67,68 . Here we showed that optogenetic targeting of astrocytes decreased the rate of amyloid beta deposition in APP mice. SWS improves Aβ clearance via the glymphatic system 58 . Since we found that optogenetic targeting of astrocytes restored SWA in APP mice, amyloid clearance via glymphatic pathway is a probable mechanism. Also, previous studies reported a link between Aβ production and SWA impairments in cognitively healthy adults 15,23,69 . Remarkably, increased Aβ deposition within medial prefrontal cortex has been associated with decreased SWA 23 . In addition, the levels of Aβ in cerebrospinal uid increase after SWA suppression using auditory tones 69 . Our nding that restoration of astrocytic network function decreases amyloid plaque deposition serves to con rm and further expand on these previous results. It also strongly suggests that astrocytes regulate slow oscillations and participate in the pathogenesis of AD.
Amyloid beta disrupts neuronal calcium homeostasis which results in calcium elevations. Abnormal neuronal calcium homeostasis has been reported in AD [70][71][72] and transgenic mouse models of betaamyloidosis, such as APP and Tg2576 mice 26,62,73 . The measure of basal neuronal calcium has been adopted as a functional readout of treatment e cacy in animal models of AD [73][74][75][76] . Indeed, we saw neurons with calcium elevations or calcium overload in APP mice similar to earlier reports 26,27,62 . The toxicity of amyloid plaques or soluble oligomers surrounding plaques is responsible for the calcium overload in neurons, since the percentage of neurites exhibiting calcium overload is proportionally related to the proximity to plaques 62 . Furthermore, soluble oligomeric Aβ elicited pronounced calcium elevations in neurons in vivo 61,76 . Our ndings indicate that neuronal calcium elevations in APP mice are prevented by activating astrocytes with optogenetics. Slow-wave restoration in this manner decreased deposition of amyloid plaques and as a consequence restored calcium homeostasis within neurons.
De cits in SWA result in memory impairments 77,78 . Potentiation of SWA during sleep has been linked to improvements in declarative memory consolidation in human subjects 30,[79][80][81] . Thus, SWA is necessary and su cient for memory consolidation during sleep. Patients with Alzheimer exhibit sleep and memory impairments similar to APP mice 36,38 . Even though present study did not assess the effect of optogenetic stimulation on sleep in APP mice and future studies are needed to determine the effect of optogenetic stimulation on sleep-wake cycles of these mice, it did address the effect of optogenetic stimulation on sleep-dependent memory consolidation. The impairments in contextual fear memory start at 6 months of age 37 . Here we saw improvements of fear recall but not fear acquisition after optogenetic treatment. This suggests that aberrant astrocytic function actively contributes to de cits in sleepdependent memory consolidation. Furthermore, these data provide further evidence that disrupted slow oscillations contribute to memory decline in AD. Optogenetic treatment of NTG mice failed to further improve their memory possibly due to a ceiling effect. It is unclear however whether the rescue of memory consolidation in APP mice was due to increased LFP power at 0.6Hz during memory consolidation, due to enhanced amyloid beta clearance or both.
At present, it is unclear whether the bene cial effects of optogenetic stimulation cease after light offset. Future experiments are needed to elucidate the potential of circuit entrainment beyond the stimulation period.
In summary, we demonstrate that astrocytes are critical for maintenance of proper circuit function in APP mice. Restoration of SWA via astrocyte targeting reduces amyloid plaque deposition, restores the calcium homeostasis in neurons, and improves sleep-dependent memory consolidation in APP mice. Our results provide evidence that dysfunctional slow oscillations underlie the memory impairments and pathophysiology of AD, as well as propose that targeting astrocytes to alter SWA might serve as a novel therapeutic strategy against AD.

Experimental animals
The mouse experiments were performed with the approval of Massachusetts General Hospital (MGH) Institutional Animal Care and Use Committees (IACUC, protocol # 2012N000085) for the use of experimental animals. All methods were carried out in accordance with relevant guidelines and regulations. All methods are reported in accordance with ARRIVE guidelines. The transgenic mouse line expressing the Swedish mutation of amyloid precursor protein (APP) and deltaE9 mutation in presenilin 1 (APPswe/PS1dE9; APP mice) 82 and age matched non-transgenic littermate controls (NTG) were used in the study. Age and sex of animals were speci ed in methods and results section for individual experiments. Mice were housed 4 animals per cage in a pathogen-free environment with ad libitum access to food and water on a 12/12 hour day and night cycle.
All stereotaxic injections were performed according to the following procedure. The mice were anesthetized by inhalation of 1.75% iso urane and placed on the stereotaxic instrument with a heating pad to maintain body temperature during the injections. Eye ointment was applied to the animals' eyes prior to viral injections to prevent dryness. Fur was removed from the scalp with a trimmer. Liquid iodine was applied to the exposed skin and wiped clean with 70% isopropyl alcohol swabs 3 times each. An incision was made with sterile surgical scissors to expose the skull. A small burrhole was drilled over each injection site. Injections were performed with a 33-gauge metal needle and a 10 µl glass syringe, which was controlled by a stereotaxic injector. All the stereotaxic injections were performed at the rate of 100 nl/min. Following each intracortical injection, the scalp was sutured and animals were allowed to recover from anesthesia on a heating pad until awake and freely moving. At least four weeks were allowed for viral vector expression prior to imaging or optogenetic manipulation.

Cranial Window Installation And Multiphoton Imaging
Four weeks after AAV2/5-gfa2-YC3.6 injection, a cranial window was implanted over the injection site in the right somatosensory cortex of each APP and NTG mouse to allow monitoring of astrocytic calcium transients with multiphoton microscopy. To achieve this, mice were anesthetized by inhalation of 1.75% iso urane. A round craniotomy was made with a dental drill and a 5mm glass coverslip was placed over the cortex. A mixture of dental cement and Krazy glue was used to secure the coverslip to the surrounding skull. A mode-locked titanium/sapphire laser (Mai Tai; Spectra-physics, Fremont, CA) generated two-photon uorescence with 860 nm excitation, and three photomultiplier tubes (PMTs) (Hamamatsu Photonics, Japan) detectors collected the uorescence emission in the range of 380-480 nm, 500-540 nm, and 560-650 nm. YC3.6-expressing astrocytes were imaged with a 25X water immersion objective (NA = 1.05, Olympus). Time-lapse images were acquired at a resolution of 256 x 256 pixels and sampling frequency at 2 µs/pixel. The power of the laser was kept below 50 mW at the objective to avoid phototoxicity. Astrocytic calcium transients and slow oscillations were monitored in anesthetized mice similar to other reports 53,57,83 .
The cohort of animals that underwent chronic optogenetic stimulation of ChR2-expressing astrocytes was rst subjected to craniotomy with 5mm cranial window implanted over the right somatosensory cortex. Z-stack images of amyloid plaque and AAV2-CBA-YC3.6-expressing neurons (cell bodies and neuronal processes, neurites) were acquired with a 25X water immersion objective at a resolution of 512 x 512 pixels and a sampling frequency at 4 µs/pixel. Methoxy-XO4 (10mg/kg) was injected intraperitoneally a day before the imaging session to detect amyloid plaques.

Optogenetic Stimulation Of Chr2-expressing Astrocytes
Four weeks after ChR2-mCherry or mCherry alone injection in the left hemisphere and the voltage sensor Voltron injection in the right hemisphere, a small uorophore with an appended HaloTag ligand Janelia Flour 525 dye (JF525 dye) (Janelia Research Campus, Ashburn, VA) was injected through the retro-orbital sinus a day before the imaging. JF525 dye was prepared by lyophilizing 100 nanomoles of JF525 dye and mixing in 20 µl of DMSO, 20 µl Pluronic F-127 (20% w/v in DMSO), and 100 µl of PBS. On the day of imaging, NTG and APP mice underwent cranial window installation to allow recording of slow oscillations in the right hemisphere. A 5mm glass coverslip was placed over the right hemisphere. After the cranial window surgery, a ber-optic cannula (length = 2mm, Doric Lenses Inc, Quebec, Canada) was implanted over the ChR2-mCherry or mCherry injection site to activate astrocytes. Cannula tip was positioned above the cortex to prevent penetration and disruption of cortical circuit activity. The cannula and glass coverslip were secured with the dental cement mixture to the surrounding skull. The activity of neurons expressing Voltron labeled with JF525 was recorded with an upright uorescence microscope with 2X objective (NA = 0.14, Olympus A separate cohort of mice underwent chronic light treatment to modulate the astrocytic network function. 400 ms pulses of 473 nm light (5-10 mW) were delivered at 0.6 Hz, 24 hr/day for 14 or 28 days to restore the power of slow oscillations in the APP mice. Mice were housed in microdialysis bowls (Harvard Apparatus, Holliston, MA) with free access to food and water during entire optogenetic treatment.
Following the chronic light treatment, mice were subjected to multiphoton imaging of amyloid plaques and YC3.6-expressing neurons as described above.
Open eld and contextual fear conditioning tests.

Image Analysis
The time-lapse images of cortical slow oscillations that were acquired with the CMOS camera were analyzed with ImageJ. The uorescence intensity from Voltron_JF525 images were measured as the change in the pixel intensity over the baseline (ΔF/F0). Fourier transform analysis was used to quantify the ΔF/F0 traces of Voltron_JF525 images to determine the power of slow oscillations at 0.4-0.6 Hz, and the power of slow oscillations at 1.2 Hz when astrocytes were optogenetically activated at twice of the frequency.
ImageJ was used to process the time-lapse images of YC 3.6-expressing astrocytes and Z-stack images of YC 3.6-expressing neurites. YC 3.6 is a genetically encoded ratiometric calcium sensor. To determine YC3.6 ratio values, ratio images of YFP/CFP were created. Brie y, the images were preprocessed by subtracting the background from the last image of each volume from YFP and CFP channels. A median lter with a radius of 2 was applied to each channel. The regions of interest outlining the cell body of astrocytes or neurites were manually traced and selected using ImageJ. The YC3.6 ratios of the regions of interest (ROIs) were calculated by dividing the emitted uorescence intensity of YFP by CFP. A customwritten Matlab script was used to create the pixel-by-pixel pseudocolored images based on the YFP/CFP ratios. Fourier transform analysis was performed to determine the power and frequency relationships of the calcium transients in astrocytes. We de ned the number of astrocytes cycling at slow wave frequency as the power of the astrocytic calcium transient greater than three standard deviations above the mean.
Amyloid plaque image analysis was performed as follows. First, maximum intensity projections of each Z-stack volume were generated with Matlab. Then, ImageJ was used to manually select the ROIs for each plaque. Lastly, amyloid plaque number and amyloid plaque burden were counted and measured with the custom-written Matlab script. Data were presented as amyloid plaque numbers per cortical volume.
Alternatively, amyloid plaque burden was calculated as percentage of cortical area occupied by amyloid.
The z values of Z-stacks were comparable across conditions.

Statistics
All the statistical analyses were performed using GraphPad Prism (version 6.01). Data was reported as mean ± SEM. Data sets were rst tested with Shapiro-Wilk normality test, D'Agostino-Pearson omnibus normality test or Kolmogorov-Smirnov normality test. Normally distributed data was analyzed using student t test. Mann-Whitney test or Kruskal-Wallis test was used to analyze nonparametric data. *p < 0.05 was considered signi cant for 2 groups.

Data Availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information les). Datasets not included are available from the corresponding author upon reasonable request. Figure 1 Astrocytic calcium transients are impaired at the slow-wave frequency in young APP mice. (A) Multiphoton microscopy image of YC3.6-expressing astrocytes in 4-6 months old NTG and APP mice.

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