Microglia undergo rapid phenotypic transformation in acute brain slices but remain essential for neuronal synchrony

Acute brain slices represent a “workhorse” model for studying the central nervous system (CNS) from nanoscale events to complex circuits. While slice preparation inherently involves tissue injury, it is unclear how microglia, the main immune cells and damage sensors of the CNS shape tissue integrity ex vivo. To this end, we have studied the mechanisms of microglial phenotype changes and contribution to neuronal network organisation and functioning in acute brain slices. Using a novel ATP- reporter mouse line and microglia reporter mice, we show that acute slice preparation induces rapid, P2Y12 receptor (P2Y12R) dependent dislocation and migration of microglia, paralleled with marked morphological transformations driven by early ATP surges and subsequent ATP flashes. Gradual depolarization of microglia is associated with the downregulation of purinergic P2Y12R and time-dependent changes of microglia-neuron interactions, paralleled by altered numbers of excitatory and inhibitory synapses. Importantly, functional microglia not only modulate synapse sprouting, but the absence of microglia or microglial P2Y12R markedly diminishes the incidence, amplitude, and frequency of sharp wave-ripple activity in hippocampal slices. Collectively, our data suggest that microglia are inherent modulators of complex neuronal networks, and their specific actions are indispensable to maintain neuronal network integrity and activity ex vivo. These findings could facilitate new lines of research resulting in improved ex vivo methodologies and a better understanding of microglia-neuron interactions both in physiological and pathological conditions.


Introduction 20
Since its first applications several decades ago (Andersen, 1981;Yamamoto & McIlwain, 21 1966), the acute brain slice preparation technique has become a key tool in the field of neuroscience 22 and has extensively contributed to our understanding of cellular physiology. As the methodology of In line with this, the spatial distribution of microglial processes within the slice showed similar, 91 but even more pronounced alterations during the incubation process ( Fig 1G). Microglial process 92 density in the top layer increased significantly by 21% (p<0.05) as early as 2 hours of incubation and 93 by 29% (p<0.001) after 5 hours of incubation (Fig 1H, blue). Contrary to microglial cell bodies, process 94 density at the bottom layer showed a significant drop by 15% (p<0.05) already after 2 hours of 95 incubation (Fig 1H, green). Here, we also observed 23% (p<0.05) lower process density at the bottom 96 region of slices compared to top after 5 hours of incubation (Fig 1H; black statistical indication, 97 independent t-test). Importantly, the total percentage of area covered by microglial processes dropped 98 to half between the 0-minute (acute slices fixed immediately after cutting) and 5-hour timepoints 99 (p<0.0001, independent t-test, Fig 1I). 100 To capture these changes in real time, slice preparations were transferred into a recording 101 chamber for confocal imaging (Fig 1J). The native signal of microglia (CX3CR1 +/GFP ) was continuously 102 imaged for at least 6 hours after slice preparation (Fig 1K, Supplementary Video 1-2). We found that 103 individual cell behaviours correlated with the quantitative measurements above (Fig 1E-F Stence et al., 2001). Decrease in the area covered by microglial processes (Fig 1I) suggested robust 143 changes in microglia morphology taking place in acute slice preparations. This behaviour was also 144 evident in the data gathered by our imaging experiments, where we saw rapid morphological 145 transformation of individual microglial cells (Fig 2A, Supplementary Video 3). To define the spatio-146 temporal characteristics of these pronounced morphological changes, preparations were immersion-147 fixed at different timepoints during incubation. In parallel, a group of mice were transcardially perfused 148 to obtain slice preparations from the same brain area as controls ( Fig 2B). Confocal images were  shape and retraction of processes translate to higher sphericity values) increased by ~50% on average 155 already after 20 minutes of incubation (independent t-test, p<0.0001), and peaked at ~100% increase 156 (p<0.0001) between 1 and 2 hours of incubation both in regions of the cortex and in the hippocampus 157 ( Fig 2F). At the same time, total number of process endings dropped by ~50% after 20 minutes of 158 incubation (p<0.0001), and this reduction also peaked between 1-and 2 hours of incubation with ~80% 159 decrease in average (Fig 2G, p<0.0001). We also conducted this experiment in an older group of animals 160 (p.n.: ~95) and observed the same extensive changes concerning both cortical and hippocampal regions 161 ( Supplementary Fig 1). Based on these observations, we concluded that microglia showed rapid 162 morphological changes in acute slice preparations. N-methyl-D-glucamine (NMDG)-based cutting solution, and slices were incubated in a low Na + , 205 sucrose-based ACSF while floating in a Styrofoam "boat" with a netting at the bottom on the surface 206 of an incubation chamber; Lab #3 used an ice-cold, sucrose-based ACSF solution, and slices were 207 incubated in a storage chamber filled with carbonated standard ACSF (for details see: Methods). Of 208 note, we also tested slice preparation using a room-temperature cutting solution, which did not alter the 209 course of microglial changes (data not shown). In all laboratories, slices were immersion-fixed in 4%   To compare different laboratories and acute slice preparation  218 techniques, two independent laboratories received a protocol to synchronize acute slice fixation timepoints 219 and fixation methods. 220 B. Independent laboratories were using their own native acute slice preparation method to harvest and fixate 221 slices at specific timepoints according to the protocol. 222 C. Acute slice preparations were sent to our laboratory after fixation and preparation for transport (see :  223 Methods), where they were treated together during immunostaining, imaging, and morphological analysis. Rapid downregulation of P2Y12R is accompanied by gradual 235 microglial depolarization during the incubation process 236 To further investigate the extent of microglia transformation during the incubation process, we 237 next examined changes in microglial P2Y12 receptor (P2Y12R) levels, a core microglial receptor especially in response to injury when exposure to high ATP/ADP levels. Therefore, rapid P2Y12R 259 downregulation is expected to be accompanied with membrane depolarization. To test this, we 260 performed electrophysiological recordings from microglia in acute slices. As in previous experiments, 261 acute slice preparations were placed into an interface type incubation chamber for recovery and then 262 transferred to a recording chamber at different timepoints during incubation, to measure microglia in 263 whole cell patch-clamp configuration. Targeting of microglial cells was guided by the intrinsic GFP 264 signal across the hippocampal CA1-CA3 stratum lacunosum-moleculare and stratum radiatum regions, 265 and below ~40 µm measured from the slice surface ( Fig 4F). Our results showed that microglial cells 266 became gradually more depolarized throughout the total 5 hours of the incubation process (Fig 4G, left; 267 N=158 cells measured in slices from a total of 10 animals), while we did not observe significant changes 268 in input resistance (Fig 4G, right). Taken together, microglia undergo rapid phenotype changes in acute 269 slice preparations, as characterized by cell body and process translocation (Fig 1, 3), quick 270 morphological shift towards a reactive shape (Fig 2, 3), as well as early P2Y12R downregulation and 271 gradual depolarization of resting membrane potential (Fig 4).  Since injury-induced ATP markedly affects microglial phenotypes and recruitment via P2Y12R A. Outline of the two-photon imaging experiment performed on acute brain slices. Slices were prepared from 335 mice expressing the ATP sensor (GRABATP on the cell surface of VGlut1-positive neurons to measure changes 336 in extracellular ATP levels. 10 min-long imaging sessions were followed by Z-stack acquisition after slice 337 cutting, for up to 5 hours. Altogether, 3*5 hippocampal (CA3) and 3*5 cortical (layer 2-3) data sets were 338 collected from n=3 mice 1 hour apart. 339 B. Representative images of the video sets taken at the earliest manageable timepoint after slice cutting. Whitney, ns: not significant, *: p<0.05, **: p<0.01. 366

Supplementary Figure 2. Characteristics of extracellular ATP signal changes in acute brain slices 367
A. MFI data collected from hippocampal or cortical slices imaged 1-2h (red), 2-3h (blue), 3-4h (orange) or 4-5 368 (green) hours after slice preparation show significantly slower ATP reduction, than those from the earliest 369 timepoints (0-1h, black). Note, that data were collected from regions of interests where no apparent 370 flashing ATP activity was observed. Two-way ANOVA, ****P<0.0001 for all (red, blue, orange, green) 371 compared to 0-1h (black). 372 B. MFI data obtained from Z stacks. ATP signal intensity and subsequent reduction over time is most apparent 373 next to the slice surface (0-25, 25-50um) compared to deeper layers (75-100um). Two-way repeated 374 measures ANOVA, Dunnett's multiple comparison test. 375 C. Gradient in ATP signal intensity from the slice surface to the ~100 µm depth of the slice is maintained for at 376 least 5 hours after slice preparation. Two-way repeated measures ANOVA, Dunnett's multiple comparison 377 test. 378 of GABAergic synapses (Fig 6D, right), we did not observe significant increase immediately after slice 401 preparation, however we measured a prominent decrease in contacts during the incubation process, with 402 a 36% drop after 2 hours of incubation (p<0.01). Based on these observations, we concluded that 403 microglia-neuron interaction sites underwent rapid and progressive changes, as somatic coverage and 404 contact prevalence on glutamatergic synapses significantly increased immediately after slice 405 preparation, followed by a gradual decrease over time. 406 Given the observed changes in microglial contact prevalence on synapses, we next examined 407 how the density of synaptic elements changed during the incubation process. We used the quantitative 408 post-embedding labelling method to precisely determine both glutamatergic and GABAergic synaptic 409 density changes (Fig 6E) in acute slices. We found significant increases in both glutamatergic (20% 410 increase compared to 0 min, p<0.01) and GABAergic (9% increase compared to 0 min, p<0.05) synaptic 411 density after 1 hour of incubation (Fig 6F), similarly to previous reports ( (CTRL) and depleted (DEPL) condition, we used the previously described post-embedding labelling 437 method ( Fig 6I). We found that time-dependent synaptic density changes after slice preparation were 438 markedly influenced by microglial actions (Fig 6J). To our surprise, the absence of microglia abolished 439 the gradual increase of glutamatergic synaptic density observed under control conditions and resulted 440 in significantly lower synapse densities from 20 minutes after slice preparation and onward, reaching 441 27% lower synaptic density values in DEPL compared to CTRL after 5 hours, (p<0.001, Fig 6D, left).

442
Even more prominent differences were observed regarding GABAergic synaptic densities (Fig 6J,   443 right). The absence of microglia led to a marked initial increase of GABAergic synaptic density, peaking 444 at 1 hour, followed by significant drop at 2 hours. This course fundamentally differs from that observed 445 under control conditions, which peaks at 2 hours (19% higher in CTRL, p<0.001, Fig 6J, right). These   The absence of microglia or microglial P2Y12 receptors 489 dysregulates neuronal network activity in acute slice preparations 490 It is well known that slices need a recovery period after cutting, to enable reliable and stable 491 electrophysiological measurements. In line with this, we observed significant increase in SWR 492 parameters after 2 hours in control slices (Supp . Fig 4), which paralleled marked, time-dependent 493 changes in GABAergic and glutamatergic synapse densities.

494
Because the absence of microglia resulted in markedly lower glutamatergic and GABAergic 495 synaptic densities after 2 hours of incubation (Fig 6J), we set out to examine whether synchronous 496 events are also positively affected by the presence of microglia and compared features of spontaneously 497 occurring SWR activity measured from control (CTRL), or microglia depleted (DEPL) acute slice 498 preparations (Fig 7A-B). Simultaneous recordings from both conditions allowed precise quantification 499 and comparison of the differences concerning the occurrence of SWR activity between the two 500 conditions. We observed that 18 out of 36 slices (50%) presented detectable SWR activity in CTRL, 501 whereas only 4 out of 36 slices (11%) from DEPL (Fig 7C). This striking difference confirms that 502 microglial actions are necessary for the emergence of physiological-like network activity in ex vivo 503 slices. Furthermore, our results showed significant differences between spontaneously occurring SWR 504 activity recorded from CTRL or DEPL acute slice preparations, as we observed ~3.5-fold decrease in 505 SWR amplitude, ~1.9-fold decrease in SWR rate and ~2.1-fold decrease in ripple amplitude registered 506 from DEPL when compared to CTRL slices (Fig 7D). These results indicate that microglia can 507 effectively support the neuronal network in slice preparations during the incubation process, and can 508 positively influence the occurrence, amplitude, and frequency of spontaneously occurring SWR 509 activity.

510
To examine whether P2Y12 mediated actions are contributing to the observed effects, we 511 repeated the same experiment by comparing slices from control and P2Y12 KO mice (Fig 7E-F).

512
Similarly to microglia depleted mice, SWR occurrence was lower in the absence of P2Y12Rs (Fig 7G,   513 33%, 10 slices from 30 in total) when compared to control slices (46.6%, 14 slices out of 30). In this 514 case, we also observed significant differences in SWR parameters, as SWR amplitude in average 515 showed a ~1.7-fold decrease (Fig 7H, left), while SWR ripple amplitude showed ~1.3-fold decrease in 516 P2Y12 KO animals (Fig 7H, right). SWR rate also decreased ~1.3-fold compared to control slices, 517 however due to a higher variance, this difference was not significant (Fig 7H, middle). Taken together, 518 these results suggest that the positive influence of microglia on SWR activity observed in acute slice 519 preparations is in part mediated by microglial P2Y12R action. A. Schematic representation of the experiment. CX3CR1 +/GFP littermates (N=8; p.n.: ~65 days) were used to 560 create acute hippocampal slice preparations and placed into an interface-type incubation chamber for at 561 least 1 hour of recovery time. Subsequently, slices were transferred at specific timepoints (hourly after 1-5 562 hours) into a recording chamber to measure sharp-wave ripple (SWR) activity via local field potential 563 recordings (LFP) registered from the CA3 pyramidal layer of the hippocampus. 564 B. Representative LFP recordings measured after 1 (yellow) or 2 hours (purple) of incubation. Grey lines 565 represent detected SWR events (bars 10 s, 50 µV In this study, we provide a comprehensive assessment of microglial function in acute slice 574 preparations in an experimentally relevant timeframe, using sample sets from multiple expert 575 laboratories. We show that while microglia are undergoing marked time-dependent phenotype changes, 576 they are instrumental to maintain ex vivo spontaneous activity of the neuronal network in a P2Y12R-577 dependent manner. We believe that these fundamental mechanisms of microglia-neuron interactions 578 should be considered for studies on microglial and neuronal functions ex vivo, while also have important 579 implications for the acute slice methodology and could facilitate improvements in modelling. 580 We first studied how the slice preparation procedure affected microglial cells in acute slices.   704 In order to minimize bacterial contamination, all the tools and containers used for slice 705 preparation and incubation were routinely cleaned before and after experiments with 70% ethanol and 706 were rinsed extensively with distilled water. For acute slice preparation, mice were decapitated under 707 deep isoflurane anesthesia. The brain was removed and placed into an ice-cold cutting solution, which 708 had been bubbled with 95% O2-5% CO2 (carbogen gas) for at least 30 min before use. The cutting

714
After acute slice preparation, slices were placed into an interface-type holding chamber for 715 recovery. In an interface-type chamber, slices are laid onto a mesh just slightly submerged into the 716 artificial cerebrospinal fluid (ACSF), therefore the oxygenation of the tissue is mainly realized by the 717 direct exposure to humidified oxygen-rich air above the slices. This chamber contained standard ACSF 718 at 35°C that gradually cooled down to room temperature. The ACSF solution contained the following 719 (in mM): 126 NaCl, 2.5 KCl, 26 NaHCO3, 2 CaCl2, 2 MgCl2, 1.25 NaH2PO4, 10 glucose, saturated 720 with 95% O2-5% CO2. Immediately after slice preparation/given timeframes of incubation/after 721 recordings, slices were immersion-fixed for 1 hour with 4% PFA solution. In the case of PSB treated 722 acute slices, both cutting and standard ACSF solutions used to prepare the slices contained 10 µM PSB-    768 300 µm thick acute slices were immersion fixed immediately after slicing (0 minute) or after 769 20 minutes, 1 hour, 2 hours or 5 hours spent in an interface-type incubation chamber (Fig 1A). and these masks were subtracted from the original tiff files to get images containing microglial 780 processes only. As a validation for volume-related quantifications in acute slices, the average thickness 781 for each preparation across the incubation procedure was measured (mean±SD: 262±7 µm), which 782 showed no substantial differences between groups (Kruskal-Wallis test, p>0.05). For quantification, a 783 measuring grid was placed onto the entire thickness of cross-sections (Fig 1 C), which divided the 784 thickness into 7 equal zones. Cell body numbers were counted within these grids, and microglial process   797 Acute brain slices (300µm thick) were prepared from 80-day old CX3CR1 +/GFP mice as described above. or after 20 minutes, 1 hour, 2 hours or 5 hours spent in an interface-type incubation chamber (Fig 1A). 806 Fixed slices were washed in 0.1M PB, flat embedded in 2% agarose blocks, and re-sectioned on a 807 vibratome (VT1200S, Leica, Germany) at 100 μm thickness (Fig 1B). Sections selected from the middle 808 region of incubated slices were immunostained with antibodies, and DAPI (for primary and secondary 809 antibodies used in this study, please see Table 1.). Preparations were kept in free-floating state until 810 imaging to minimize deformation of tissue due to the mounting process. Imaging was carried out in   822 Before the immunofluorescent labelling, the 50 µm thick sections were washed in PB and Tris-823 buffered saline (TBS). Thorough washing was followed by blocking for 1 hour in 1% human serum

848
For the analysis of synaptic contact prevalence, confocal stacks with triple immunofluorescent 849 labeling (pre-and postsynaptic markers and microglia) were analyzed using an unbiased, semi-850 automatic method. First, the two channels representing the pre-and postsynaptic markers were exported 851 from a single image plane. The threshold for channels were set automatically in FIJI, the "fill in holes" 852 and "erode" binary processes were applied. After automatic particle tracking, synapses were identified 853 where presynaptic puncta touched postsynaptic ones. From these identified points we selected a subset 854 in a systematic random manner. After this, the corresponding synapses were found again in the original 855 Z-stacks. A synapse was considered to be contacted by microglia, when a microglial process was closer 856 than 200 nm (4 pixels on the images).

857
Post-embedding immunofluorescent labelling and quantitative analysis 858 The technique described by Holderith et al. (Holderith et al., 2021.) was used with slight 859 modifications. 300 µm thick acute slices were cut from CX3CR1 +/GFP mouse line and then immersion 860 fixed immediately after slicing (0 minute) or after 20 minutes, 1 hour, 2 hours or 5 hours spent in an 861 interface-type incubation chamber (Fig 1 B) The excitation wavelength was set to 920 nm for detecting the signal and 550/88 nm emission filter and 916 a GaAsP NDD PMTs detector was used. Time lapse data were analyzed using the NIS-Elements   Acute slice preparations were gathered at each recording day (6 slices/animal, 450 µm thick) 962 in a pairwise manner from control and microglia depleted animals while using the same solutions and 963 equipment. The slice preparation sequence was alternated throughout the recording days between the 964 two groups, as well as the chambers that were used for the incubation process, to minimize artefacts 965 that might have been introduced by variance in slice preparation or incubation quality. After at least 1 966 hour of incubation, slices from both conditions were transferred together in a pairwise manner to a dual peaks. Based on this, we identified the ripple cycle closest to the SWR peak and used its negative peak 989 as triggering event for averages to preserve ripple phase. Taking the absolute value of the ripple 990 bandpassed signal and low pass filtering it, we calculated the ripple power peak (Ripple amplitude).

991
After detection, ~100 consecutive events were selected for quantification, where highest values were 992 measured along the whole recording.

993
Quantification and statistical analysis 994 All quantitative assessments were performed in a blind manner.