Impaired astrocytic Ca2+ signalling in awake Alzheimer’s disease transgenic mice

Increased astrocytic Ca2+ signaling related to amyloid plaques has been shown in Alzheimer’s disease mouse models, but to date no reports have characterized behaviorally induced astrocytic Ca2+ signalling in such mice without the confounding effects of anesthesia. Here, we employ an event-based algorithm to assess astrocytic Ca2+ signals in the neocortex of awake-behaving tg-ArcSwe mice and non-transgenic wildtype littermates while monitoring pupil responses and behavior. We demonstrate an attenuated astrocytic Ca2+ response to locomotion and an uncoupling of pupil responses and astrocytic Ca2+ signalling in 15-months old plaque-bearing mice. This points to a potential decoupling of neuromodulatory activation and astrocytic Ca2+ activity, which may account for some of the cognitive dysfunctions observed in Alzheimer’s disease.


Introduction 41
Since astrocytic Ca 2+ signals were first discovered in the early 1990's they have been the object of 42 numerous studies exploring their roles in brain physiology and pathophysiology. Importantly, such 43 signals have been shown to occur in response to a wide array of neurotransmitters and to trigger the 44 release of substances that affect neuronal signalling and the vasculature. A growing body of evidence 45 suggests that astrocytic Ca 2+ signals play important roles in higher brain functions such as memory 46 formation and cortical processing, mediated in part through the neuromodulatory systems of the brain 47 The field of astrocytic Ca 2+ signalling is undergoing a revolution as developments in optical imaging 58 and genetically encoded fluorescent sensors now allow us to monitor these signals in awake-behaving 59 mice, without the confounding effects of anesthesia (Srinivasan et al. 2015; Bojarskaite et al. 2020). red. Scale bars: 50 μm. (E) 3D visualization of the imaging plane relative to amyloid plaques, with lines 98 representing shortest distance from plaque to ROI. We found a low correlation between distance to nearest 99 plaque and gross level of astrocytic Ca 2+ signalling.

102
Two-photon imaging of awake-behaving tg-ArcSwe mice 103 To characterize astrocytic Ca 2+ signalling in awake tg-ArcSwe mice and nontransgenic littermates we 104 employed two-photon microscopy of cortical layer 1-3 astrocytes in the somatosensory cortex 105 expressing GCaMP6f. The glial fibrillary acidic protein (GFAP) promoter was used to target 106 astrocytes ( Fig 1A). Amyloid plaques were visualized in vivo by methoxy-X04 delivered by 107 intraperitoneal injection (Fig. 1A). Methoxy-X04 enters the brain and specifically stains parenchymal 108 Aβ plaques and cerebrovascular deposits (Klunk et al. 2002), and has been used for in vivo imaging in The tg-ArcSwe mice were imaged at ~15 months of age. At this age they present with amyloid-β 121 plaques throughout the cortical mantle, and score poorly on behavioral tasks (Codita et al. 2010;122 Lillehaug et al. 2014; Lord et al. 2006) (Fig. 1B). Aβ-plaques were characterized by loss of cells and 123 severely perturbed tissue morphology, including autophagic vacuoles (Fig 1B). Even so, relatively 124 normal cellular morphology was present at short distances away from amyloid plaques, and astrocytes 125 faithfully expressed the GCaMP6f Ca 2+ sensor 3 weeks after viral transduction (Fig. 1A,B). 126 Astrocytic Ca 2+ signals in quiet wakefulness are preserved in tg-ArcSwe mice spreading to nearby astrocytes. Such Ca 2+ waves were found in 10-15% of recordings from tg-131 ArcSwe mice (Fig. 1D). The number of such clear pathological events were few compared to the 132 overall astrocytic Ca 2+ signalling we found without the highly confounding effects of anesthesia. 133 Consequently, the gross level of astrocytic Ca 2+ signalling was similar in mutant mice and their 134 littermates as measured by ROA frequency, ROA density (the active fraction of a compartment) as 135 well as event size and duration in the full field-of-view (FOV) and across the different astrocytic 136 subcompartments ( Fig 1C and Supplementary Fig. 2). We were not able to detect a clear correlation in 137 astrocytic Ca 2+ signalling measured by ROA density and the distance from nearest amyloid plaque (in 138 3D) (slope = -0.00020, Fig. 1E the tg-ArcSwe mice, they were allowed to move freely on a custom built disc-shaped treadmill 147 (Bojarskaite et al. 2020). All mice exhibited both running and behavioral quiessence, and the level of 148 running between the two genotypes were comparable ( Supplementary Fig. 1). Running was 149 accompanied by an increase in pupil size and a brisk increase in astrocytic Ca 2+ signalling typically 150 involving most of the astrocytes in the field-of-view (FOV) in both genotypes ( Fig. 2A). When astrocytic Ca 2+ signals were analysed using a linear mixed effects regression model, a lower ROA 152 density rise rate was found in tg-ArcSwe mice when assessing the full FOV (0.31 in WT vs. 0.20 in 153 tg-ArcSwe, p = 0.032), and astrocytic processes (0.32 in WT vs. 0.20 in tg-ArcSwe, p = 0.032), 154 whereas the Ca 2+ responses were not significantly different in astrocytic somata and endfeet (0.46 in 155 WT vs. 0.39 in tg-ArcSwe, p = 0.23, and 0.35 in WT vs. 0.30 in tg-ArcSwe, p = 0.11, 156 respectively) (Fig. 2B). Max ROA density values in WT vs. tg-ArcSwe were significantly different 157 when assessing the full FOV (0.63 vs. 0.42, p = 0.033), near significantly different when assessing 158 astrocytic processes and endfeet (0.72 vs. 0.53, p = 0.053 for processes and 0.69 vs. 0.51, p = 0.068 159 for endfeet), and not significantly different for astrocytic somata (0.81 vs. 0.71, p = 0.25) (Fig. 2B). immediately before the air puff were excluded from the analyses. We found no signs of habituation to 203 the stimulus in terms of behavioral response ( Supplementary Fig. 4). Interestingly, tg-ArcSwe mice 204 were more prone to react with running behavior during startle responses than WT littermates (  and WT littermates (Fig. 4). 241

disease. 274
Methodological advances now allow astrocytic Ca 2+ signals to be studied in awake animals. 275 Benefitting from this opportunity we show that tg-ArcSwe mice sustain a pattern of behaviorally 276 induced astrocytic Ca 2+ signalling similar to that found in littermate controls. However, the signals are pharmacologically or by pharmacogenetics have been demonstrated to be beneficial (Holland,307 Robbins, and Rowe 2021). In our study we find a largely retained level of pupil response in the AD 308 mice during startle responses, which under physiological circumstances would suggest that the  In our awake-behaving mice, we found no clear correlation between Ca 2+ activity and 3D distance 325 reconstruction of plaque positions relative to the imaging plane (Fig. 1E). The apparent discrepancies 326 in the litterature and the present study may be due to different AD mouse models, different age groups 327 investigated, different Ca 2+ indicators employed or lastly the effects of removing anesthesia allowing for a much richer repertoire of astrocytic Ca 2+ signalling to emerge, effectively masking a potential 329 weak correlation. 330 331 Astrocytes have a highly complex and specialized morphology, and this morphology is known to 332 change in reactive astrogliosis (Escartin et al. 2021). The majority of astrocytic processes are much 333 smaller than what can be clearly delineated by non-super resolution optical microscopy, but to what 334 extent these small processes are altered in reactive astrogliosis is unknown, although the astrocytic 335 territories are known to be preserved (Wilhelmsson et al. 2006). We cannot rule out that gliosis- The present study underscores the importance of studying astrocytic Ca 2+ signals in unanesthetized 344 mice, and to carefully consider animal behavior when interpreting astrocytic Ca 2+ dynamics. By lifting 345 the confounding effects of anesthesia we found that the astrocytic hyperactivity previously reported in 346 Alzheimer's disease mouse models was only a part of the total picture. At first glance, the 347 physiological Ca 2+ responses were remarkably well preserved, and not characterized by a general 348 increase in astrocytic Ca 2+ signalling. However, behavior like quiet wakefulness and locomotion are 349 not static entities, and the degree of activation of all relevant parameters needs to be taken into 350 account with statistical modelling to be able to conclude if there are relevant differences between the 351 genotypes. We were able to demonstrate attenuated Ca 2+ dynamics and an uncoupling between 352 astrocytic Ca 2+ signalling and arousal in the tg-ArcSwe mice, and given the growing spectrum of roles 353 ascribed to astrocytic Ca 2+ signalling in higher brain functions, the present findings may highlight one 354 cause for the cognitive decline of AD patients. Mice were anaesthetized with isoflurane (3% for initiation, then 1-1.5% for maintenance) in room air 389 enriched with 20% pure oxygen, and given buprenorphine 0.1 mg/kg s.c. preemptively for analgesia. 390 Bupivacain was administered subcutaneously over the skull, and left for 10 minutes before a boat 391 shaped skin flap was removed. After removing the skin, a 2.5 mm diameter craniotomy was drilled 392 over the somatosensory cortex with center coordinates 3.5 mm lateral and -1.5 mm posterior to 393 Bregma. Virus was injected (70 nL at 35 nL/min at 200 µm depth from the brain's surface) at three 394 evenly spaced locations positioned to stay clear of large blood vessels, and a glass plug consisting of 395 two coverslips glued together were placed in the craniotomy, slightly pressing the dura to prevent dural overgrowth (Bojarskaite et al. 2020). The surrounding area of the skull was sealed with 397 cyanoacrylate glue and a layer of dental cement. Post-operatively the mice were given meloxicam 2 398 mg/kg for two days. Only animals with normal post-operative recovery were included in the study. 399 Mice were left to recover for a minimum of two weeks before habituation to head-fixation and 400 imaging. than 20 degrees/s the last 10 seconds prior to start of running. Spontaneous running was not defined 426 within 30 seconds following air-puff. Quiet wakefulness was defined as continuous segment of at 427 least 10 s duration with less than 2 degrees/second locomotion, as well as no locomotor activity faster 428 than 2 degrees/s for 15 seconds before segment start. Quiet wakefulness episodes were not defined 429 within 30 seconds following air-puff. Quiet wakefulness periods were reviewed manually through the 430 IR-surveillance video recordings to ensure animals were awake when sitting still. 431

Pupillometry 432
Pupil size was recorded with a Basler Dart USB camera (daA1600-60um) with a 25mm fixed focal 433 length lens and 2X fixed focal length lens extender (Edmund optics, items #59-871 and #54-356). 434 The pupil was back-illuminated with the spillover light from the two-photon microscope laser coming 435 littermates were sacrificed at 12 months of age and used for qPCR analysis. These animals were 474 anesthetized as described above and decapitated. The brains were extracted, and the left hemisphere 475 was dissected into the frontal cortex, hippocampus, cerebellum and the rest of the brain. This tissue 476 was frozen and stored in -80ᴼC pending analysis. 477  Table 1. Astrocytic Ca 2+ signals were studied by means of the ROA densitya number between 0 and 1 550 indicating the fraction of the compartment with activity. Here, the area may be the entire field-of-view 551 (FOV), or we may limit ourselves to the area identified as belonging to cellular subcompartments; the astrocytic processes, somata or endfeet. In each trial we had time series of ROA density lasting 553 around 300 s (approximately 9000 frames). Within each trial, the startle period was defined as starting 554 from the air puff at 150 s and lasting 600 frames (~20 seconds). In addition, one or more time periods 555 within the trial could be identified as spontaneous runs, or as periods of quiet wakefulness. A first 556 important question concerns the choice of summary statistics adequately describing the Ca 2+ response 557 in such periods of interest (runs and startle), which are dynamic behavioral states that entail both 558 acceleration, steady locomotion and deceleration. We have studied the mean and max ROA density 559 and the ROA density rise rate which is defined as the maximal increase in ROA density over a 560 maximum of 50 frames. Initial explorations indicated that the main results are fairly robust to the 561 choice of window length, and 50 frames appeared to be a sensible choice compared to the kinetics of 562 astrocytic Ca 2+ signals. The ROA density rise rate is meant to capture some of the dynamics in 563 astrocytic Ca 2+ signaling, and can be understood as the maximum acceleration inside the time period 564 of interest. The rise rate has a high correlation with the maximum, and if the length of the window is 565 increased sufficiently these two statistics tend to become almost identical (since most traces are close 566 to zero at some point in the trial). See the Supplementary Fig. 5A which displays both the max ROA 567 density and the rise rate in an example. 568 569 Pupil size measurements had a coarser time resolution than for the astrocytic Ca 2+ signals; see 570 example in Supplementary Fig. 5B, showing the pupil sizes around a startle response. The pattern in 571 the figurea sharp increase in pupil size after air puff, before a gradual declinewas quite typical. 572 Therefore, we chose to only consider the measurements in a time-window of 6.67 seconds on each 573 side of the airpuff (or start of running). We defined the pupil dilation as the relative increase in the 574 ratio of pupil diameter to eye diameter, inside this window. In other words, we calculate the average 575 ratio before the air puff (or start of running) and after the airpuff, compute the difference and divide 576 by the average before the air puff. 577

RNA isolation and real time PCR
Interpreting hierarchical plots 578 We have chosen to present some of our data in the form of hierarchical plots (see Figure 2B and C, 579 Figure 3B and C, and Supplementary Fig. 1, 2 and 3). These plots allow the reader to assess the 580 degree of separation between the genotypes and at the same time get an impression of the variation at 581 different levels of the analysisin our case, the variation between different mice of the same 582 genotype and the variation between repeated measurements on the same mouse. At the lowest level -583 the trial levelwe have points representing the observations themselves. For the spontaneous runs, 584 there are sometimes more than one run per trial and in that case the points are the median ROA 585 density rise rate in these runs. The maximal number of spontaneous runs in a single trial was 5 586 (average: 1.6 runs per trial). At the middle level we have the median ROA density rise rate for each mouse, and at the top level the median ROA density rise rate for each genotype. The lines between the 588 levels indicate which observations belong to each mouse, and to each genotype respectively. We have 589 made similar plots for the max ROA density also. The hierarchical plots are a useful tool for 590 exploratory analysis. They are also meant to promote transparency in scientific reporting, and to 591 highlight the importance of intra-group variation. Still, it is important to realize that the impression 592 conveyed by the plots might not be identical to the results from statistical modelling. In the plots, we 593 do not include the influence of various technical and biological covariates which one typically would 594 include in a statistical model. Some of the variation between observations belonging to the same 595 mouse and between different mice of the same genotype may be explained by such covariates, as we 596 will see in the next section. 597

598
Statistical analyses were conducted in R (version 4.1.1). The ROA density rise rate was modelled by 599 linear mixed effect regression models which were fitted using the glmmTMB package (Brooks,600 Kristensen, and Van Benthem 2017). We conducted two sets of analyses: (i) to investigate potential 601 differences in ROA density rise rate between the two genotypes; (ii) to investigate the relationship 602 between ROA density rise rate and the pupil dilation, including potential differences between the two 603 genotypes with respect to this relationship. For (i) the coefficient of primary interest is the one 604 belonging to the genotype variable, while for (ii) we are interested in the effect of pupil dilation on the 605 ROA density rise rate, as well as the interaction between this pupil effect and the genotype. In both of 606 these sets of analyses we adjusted for the following fixed effect covariates: the level of optical zoom 607 (2 levels), the depth of the measurements (in μm) and the maximal speed in the relevant time window. 608 We included random intercepts for each mouse (5 WT and 6 tg-ArcSwe). We analyzed the ROA 609 density max and mean values with similar models as the ones described here for the rise rate. For (i), 610 the number of observations ranged between 77 to 117 trials with startle data (depending on the 611 subcompartment) and between 72 and 109 episodes of spontaneous running, while for (ii) the number 612 of observations ranged between 60 and 86 for the startle data and between 35 and 44 for the 613 spontaneous runs. There were less observations for the (ii) analyses because some episodes/trials had 614 missing or incomplete pupil measurements. 615

616
The sensitivity of our result to these modelling choices were assessed by various robustness checks, 617 see next section. The adequacy of model assumptions was investigated by residual plots (Hartig,618 n.d.). In the cases where the residual plots indicated deviations from the assumption of constant 619 residual variance, we extended the model by allowing the residual variance to vary as a function of 620 genotype. The reported p-values are based on the t-distribution, with degrees of freedom as provided 621 from the glmmTMB package. No corrections for multiple comparisons were applied. 622 signalling ( Figure 1E) we considered the mean ROA density in each ROI (n=10988) in the quiet 625 wakefulness episodes. If present, any effect of plaque proximity on Ca 2+ signalling should be 626 discernible among ROIs observed in the same episode. The dashed lines in Figure 1E show the effect 627 of distance on Ca 2+ signalling within each episode, and they form an uncertain picture: in some 628 episodes there is a weak positive relationship, with seemingly higher mean ROA density further away 629 from plaques, while in many episodes there is a negative relationship, with somewhat higher mean 630 ROA density close to plaques. The overall line is found by fitting a linear mixed effect model with 631 mean ROA density as the response, with a fixed effect of distance and with each episode having its 632 own random intercept and slope for the distance effect. 633

634
For the results presented in Figure 4: Unless otherwise stated, the data are presented as mean ± 635 standard error of the mean (SEM). A p-value equal to or below 0.05 was considered statistically 636 significant. Mann-Whitney U-test was used to analyse the number of processes, total length of 637 processes in µm, process thickness in µm 3 and area fraction of positive GFAP staining. Two-way 638 ANOVA followed by Sidak post hoc comparison was used to analyse the number of intersections.