P2Y13 receptors regulate microglial morphology, surveillance, and resting levels of interleukin 1β release

Abstract Microglia sense their environment using an array of membrane receptors. While P2Y12 receptors are known to play a key role in targeting directed motility of microglial processes to sites of damage where ATP/ADP is released, little is known about the role of P2Y13, which transcriptome data suggest is the second most expressed neurotransmitter receptor in microglia. We show that, in patch‐clamp recordings in acute brain slices from mice lacking P2Y13 receptors, the THIK‐1 K+ current density evoked by ADP activating P2Y12 receptors was increased by ~50%. This increase suggested that the P2Y12‐dependent chemotaxis response should be potentiated; however, the time needed for P2Y12‐mediated convergence of microglial processes onto an ADP‐filled pipette or to a laser ablation was longer in the P2Y13 KO. Anatomical analysis showed that the density of microglia was unchanged, but that they were less ramified with a shorter process length in the P2Y13 KO. Thus, chemotactic processes had to grow further and so arrived later at the target, and brain surveillance was reduced by ~30% in the knock‐out. Blocking P2Y12 receptors in brain slices from P2Y13 KO mice did not affect surveillance, demonstrating that tonic activation of these high‐affinity receptors is not needed for surveillance. Strikingly, baseline interleukin‐1β release was increased fivefold while release evoked by LPS and ATP was not affected in the P2Y13 KO, and microglia in intact P2Y13 KO brains were not detectably activated. Thus, P2Y13 receptors play a role different from that of their close relative P2Y12 in regulating microglial morphology and function.

which allow them to sense their environment and respond to pathological stimuli. In the healthy brain, microglia are highly motile ramified cells that constantly extend and retract their processes to survey the CNS tissue (Davalos et al., 2005;Nimmerjahn, Kirchhoff, & Helmchen, 2005). We recently showed that this mode of microglial motility is regulated by the membrane potential generated by tonic activity of the twopore domain potassium channel THIK-1 . Under pathological conditions, such as brain tissue damage, disease or infection, microglia become activated, upregulating surface receptors such as MHCII and A 2A , while their morphology becomes less ramified, with fewer, shorter processes, and larger somata (Kreutzberg, 1996;Orr, Orr, Li, Gross, & Traynelis, 2009). Depending on the nature of the stimuli, microglia can also undergo a complete transformation into different functional and morphological, for example, amoeboid, states and secrete neurotoxic inflammatory, or neuroprotective anti-inflammatory factors (Hanisch & Kettenmann, 2007).
ATP is a major neurotransmitter in the CNS, which is also released by injured cells and serves as a key stimulus for triggering microglial responses (Biber, Neumann, Inoue, & Boddeke, 2007;Burnstock, 2006;Zimmermann, 1994). Microglia detect ATP and its degradation product ADP through the purinergic receptor P2Y 12 and respond by sending out their processes to surround regions of tissue damage: a process called directed motility (Haynes et al., 2006). P2Y 12 is a G iprotein-coupled receptor (GPCRs), which inhibits adenylyl cyclase activity upon activation (Communi et al., 2001;von Kugelgen & Hoffmann, 2016). In addition to P2Y 12 receptors, transcriptome data suggest that microglia express P2Y 13 receptors at levels almost as high as for P2Y 12 (Zhang et al., 2014). P2Y 12 and P2Y 13 receptors share a similar pharmacology and molecular structure (48% identical at the amino acid level; Perez-Sen et al., 2017). However, although the function of P2Y 12 receptors has extensively been studied, mainly because of its clinical relevance to platelet-mediated thrombosis and the role of P2Y 12 in directing microglial processes to damage sites, the significance of P2Y 13 receptors for microglial function remains elusive.
Transcriptome data suggest that in the brain P2Y 13 mRNA is expressed predominantly by microglia, much less by oligodendrocyte precursor cells and myelinating oligodendrocytes, and not by other brain cell types (Zhang et al., 2014). A recent study supports these data showing that microglia in the brain, and not neurons, astrocytes, or neural progenitor cells, express P2Y 13 mRNA under basal conditions, using fluorescence in situ hybridization (Stefani et al., 2018). However, P2Y 13 protein expression may be low (Haynes et al., 2006), possibly due to constitutive ubiquitination and proteasomal degradation (Pons et al., 2014). P2Y 13 mRNA is upregulated in pathological conditions such as demyelination (Bedard, Tremblay, Chernomoretz, & Vallieres, 2007) and neuropathic pain (Kobayashi, Yamanaka, Yanamoto, Okubo, & Noguchi, 2012;Niu et al., 2017), suggesting a role of this receptor in neuroinflammation. However, its function in microglia, especially under physiological conditions, is still unclear, although it may evoke a rise of [Ca 2+ ] i in response to ADP (Zeng et al., 2014) and be involved in the release of proinflammatory cytokines (Liu et al., 2017). Increasing evidence also supports the notion that ADP may act through P2Y 13 receptors to regulate the structural complexity of microglia (Matyash, Zabiegalov, Wendt, Matyash, & Kettenmann, 2017;Stefani et al., 2018) and contribute to the homeostatic control of adult hippocampal neurogenesis in situ (Stefani et al., 2018).
In this study, we expand this knowledge and describe the effect of genetically deleting P2Y 13 receptors on microglial morphology and function in the brain. Knocking out P2Y 13 results in less ramified microglial cells and a reduction in brain surveillance. It also increases the P2Y 12 -mediated THIK-1 current, although the directed motility (but not the speed of individual process movements) towards an ADP source or brain injury is slower in the P2Y 13 KO mice, as the shorter processes need to grow further and thus take a longer time to reach their target. The membrane potential of P2Y 13 KO microglia is similar to that of WT microglia, indicating that the deramified morphology is not due to voltage changes , but has some aspects of a "primed-like" microglial phenotype. This is further supported by higher baseline levels of interleukin 1β release in P2Y 13 KO mice. However, in the intact brain of P2Y 13 KO mice, no changes of classical microglial activation markers were detected. Together, these findings indicate that P2Y 13 -deficient microglia exist neither in a typical nonactivated nor an activated state, but that they have adopted a partly deramified state which lies in between those two boundaries.

| Mice
Generation of the P2Y 13 KO mouse strain and gross characterization were described previously (Fabre et al., 2010). In brief, they were generated through homologous recombination of a construct in which the first noncoding exon and 182 base pairs of the second exon of p2ry13 were replaced with a neomycin resistance cassette to generate loss of P2Y 13 expression. This construct was integrated into embryonic stem cell genomic DNA following electroporation. Blastocysts with this allele were generated and implanted into pseudopregnant females. Deletion of P2Y 13 in this mouse line was confirmed via real-time PCR of liver samples. (Confirmation of the deletion of microglial P2Y 13 using antibody labeling or Western blots was not possible because neither of the two commercially available P2Y 13 antibodies that we tested (Alomone APR017 and Abcam ab108444) nor the P2Y 13 antibody, that was kindly provided by David Julius, labeled microglia (in brain slices or when isolated) or western blots of brain or spleen tissue specifically in the wild-type mice; instead there was labeling of cells, and multiple western blot bands, in both WT and KO tissue ( Figure S1). KO mice are healthy, fertile and show no behavioral abnormalities. For all experiments, mice with a C57BL/6-background (backcrossed at least 12 times) were used, and WT and P2Y 13 KO mice of either sex were age-matched and usually littermates. For microglial imaging in real-time, for some experiments, the P2Y 13 KO mice were bred with transgenic mice expressing eGFP under control of the Iba1 promoter (Hirasawa et al., 2005) in which microglia are labeled by eGFP. Preweaning animals were housed with their mother and sometimes the father as well; weaned animals were housed in groups of 2-5. Housing was in individually ventilated cages. Animal procedures were carried out in accordance with the guidelines of the UK Animals (Scientific Procedures) Act 1986 and subsequent amendments, and a local ethical review board gave approval. Ages are stated in the figure legends for each experiment.
Electrode junction potentials were compensated. I-V relations were from responses to 200 msec voltage steps ranging from −124 mV to +56 mV in 30 mV increments. Voltage-and current-clamp recordings were performed using a MultiClamp 700B amplifier (Molecular Devices). Currents were filtered at 1 kHz, digitized (10 kHz) and analyzed off-line using pClamp10 software.

| Two-photon imaging and evoked directed motility
Microglia in hippocampal slices were imaged at 34-36 C, at a depth of 50-100 μm in the slice (to avoid studying superficial microglia that had started to become activated by the slicing procedure) using a Zeiss LSM 710 or 780 microscope (with a 20X lens, NA 1.0) and a Spectraphysics Mai Tai DeepSee eHP Ti:sapphire infrared laser. For imaging of microglia labeled with isolectin B 4 -Alexa 594, the laser was tuned to a wavelength of 800 nm, while for imaging cells labeled with eGFP a wavelength of 920 nm was used with a pixel dwell time of 1 μs. Generally the laser was adjusted to 1.8% of its maximum power at 800 nm or 6-8% at 920 nm, corresponding to~5 mW and~12 mW, respectively, at the preparation, that is, well within the intensities used by others (Hines, Hines, Mulligan, & Macvicar, 2009;Pfeiffer, Avignone, & Nagerl, 2016;Wake, Moorhouse, Jinno, Kohsaka, & Nabekura, 2009).
Ablation of a small volume of tissue (laser damage) was performed by illuminating a~10 μm radius spot with the laser intensity increased 30-fold and the pixel dwell time increased to 100 μs. For imaging of directed process motility in response to a pipette filled with 1 mM ADP or to a laser ablation, stacks of 21-31 slices imaged at 2 μm depth intervals were acquired every 30 s (the image analysis was performed to maximum intensity of stacks having a total z-dimension of 12 μm as described below). For imaging of microglial surveillance, stacks of 21-31 slices imaged at 2 μm depth intervals were acquired every 60 s. Images were typically 512 by 512 pixels and covered a square field of view 200-250 μm wide.

| Image analysis
Analysis of two-photon images was performed using custom-written ImageJ (NIH) and MATLAB scripts, available at https://github.com/ AttwellLab/Microglia. For analysis and quantification of microglial surveillance, each slice of every stack was filtered with a median filter after subtraction of smooth continuous background with the ImageJ "subtract background" plugin with a ball size of 30 pixels. The 4D stacks were then registered first for lateral drift, then rotated 90 on their side, registered for z-drift and rotated back to their original orientation. We then performed a maximum intensity projection. Cells of interest were individually selected by manually drawing a region of interest (ROI) around an area including all their process extensions over the whole duration of the resulting 2D movie, and erasing data around that ROI. These 2D movies of individual cells were then manually binarized and saved as independent files. The binarization was performed on high resolution and preprocessed images with the experimenter blinded to the genotype. The choice of the threshold value was based on the intensity and morphology of the cells. As a general binarization criterion, the threshold value was set to a value ensuring that all the microglial processes, even the thinner ones, were visible in all the different time frames.
To quantify surveillance, for each movie, starting with the second frame, we subtracted from each binarized frame F t the preceding frame F t − 1 and created two binarized movies, PE consisting of only the pixels containing process extensions (F t − F t − 1 > 0) and PR consisting of only the pixels containing process retractions (F t − F t − 1 < 0). In both PE and PR, all other pixels are set to 0. The surveillance index B is defined as the sum over all nonzero pixels in PE + PR, that is: where Σ pixels denotes a sum taken over all nonzero pixels, and in most graphs is displayed normalized to its average over the first 20 min of the experiment (<Σ pixels [PE + PR] > control where < > control denotes a temporal average taken over the 20-min control period of the experiment), or normalized to the value for WT mice. The surveillance index provides a measure of the brain volume that is surveyed by a microglial cell in a given time. It is affected both by the rate at which processes elongate and shorten, and by the overall number of the microglial processes and their length.
To quantify ramification, the resulting movies were processed using MATLAB. In each movie frame, the MATLAB functions bwarea (uk.mathworks.com/help/images/ref/bwarea.html) and bwperim (uk. mathworks.com/help/images/ref/bwperim.html) with 8-connected neighborhood were used to quantify, respectively, the area and the perimeter of the cell. The ramification index R is defined as the ratio of the perimeter to the area, normalized by that same ratio calculated for a circle of the same area. Specifically: Thus, R = 1 if the cell is a perfect circle. The more ramified the cell, the larger R is.
To quantify the area-normalized motility index, a measure of how well the cell surveys the brain given its size, we normalized the surveillance index of each individual cell to its mean area. The area of each cell was measured in maximum intensity projections for every time frame and then averaged over all time frames to get the mean area (in pixels) per cell. We then averaged the ratios (surveillance index/mean area) of all cells for the WT and all cells for the KO mice and reported these numbers.
For analysis and quantification of microglial directed process motility in response to a laser ablation to simulate brain injury, or to a pipette filled with 1 mM ADP, each slice of every stack was first filtered with a median filter (which replaces each pixel value with the median value of the 3 pixel × 3 pixel array centered on that pixel). We then performed a maximum intensity projection of stacks having a total z-dimension of 12 μm (with the pipette tip or the laser ablation in the middle), registered (Thevenaz, Ruttimann, & Unser, 1998), and binarized the resulting two-dimensional movies (setting the threshold for binarization manually). The resulting movies were processed using a MATLAB script inspired by the algorithm described by Gyoneva et al. (2014). Briefly, after the user manually clicks on the final target of chemotactic processes (the tip of the glass pipette containing ADP), the algorithm divides the surrounding area into concentric circles with radii at 2 μm intervals, and then segments these circles into 32 radial sectors, thus creating 32 patches between every two consecutive concentric circles. Then, for each frame, starting from the center, the algorithm searches in every radial sector for the first patch containing

| Microglia immunostaining and Sholl analysis
Microglial morphology was assessed by Sholl analysis (Sholl, 1953). To compare changes in microglial morphology produced by deletion of P2Y 13 expression, three KO and three control wild-type (WT) mice were killed by sodium-pentobarbital overdose and fixed by transcardial perfusion with 4% paraformaldehyde (PFA). Brains were then removed and further fixed for 24 hr in 4% PFA. Horizontal sections (75 μm, "magic cut"; Bischofberger et al., 2006) were cut on a vibratome (Leica). Tissue sections were blocked and permeabilized in 10% goat serum and 0.5% Triton X-100 in PBS for 4 hr at room temperature. Slices were incubated at 4 C overnight in rabbit anti-Iba1 (Synaptic Systems 234003) antibody prepared in blocking solution.
After washing three times for 20 min in PBS, slices were incubated overnight in secondary antibody (goat anti-rabbit Alexa Fluor 488).
Slices were then washed in PBS, incubated in DAPI (Invitrogen) and mounted on glass slides using DAKO fluorescence mounting medium.
All images were acquired using a 63× oil immersion objective (NA 1.4) on a Zeiss LSM700 confocal microscope. All analysis was carried out with the experimenter blinded to the genotype. Cell reconstructions were carried out using 3D automatic cell tracing in Vaa3D software (http://www.vaa3d.org), which is available online and uses the APP2 (all-path-pruning 2.0) algorithm to re-construct ramified cells in 3D (Xiao & Peng, 2013). In brief, the algorithm uses the gray-weighted image distance transform (https://uk.mathworks.com/help/images/ ref/graydist.html) to assess the area occupied by the cell. This can be done either automatically or manually. It then allows the experimenter to manually perform a background thresholding method and applies the distance transform algorithm to specify the cell body as the main seed node. It subsequently reconstructs a "tree" where the seed node is the cell body, the branch nodes are the microglial branches that connect different processes (segments) and the leaf nodes are the terminals. It then applies a length-based hierarchical pruning method to code the final reconstruction of the cell. Finally, the tracing method produces a file with all the hierarchical information stored and morphological parameters are then extracted using custom code written in MATLAB. The code used is available at https://github.com/ AttwellLab/Microglia.

| Microglia isolation from adult mouse brain and immunocytochemistry
Microglia isolation from adult mice was performed following the protocol described in Lee and Tansey (2013). In brief, three adult WT and three adult KO mice were perfused with ice-cold sterile PBS under deep anesthesia (sodium pentobarbital). Mice were then killed and brains removed and finely minced with a blade on ice, followed by incubation of the tissues in dissociation medium containing DNase I (20 U/ml, Invitrogen), Dispase II (1.2 U/ml, Roche), and Papain (1 mg/ml, Sigma-Aldrich) for 30 min at 37 C. Tissues were subsequently mechanically dissociated (with the use of Pasteur pipettes) and filtered into single-cell suspensions before being centrifuged at 250 g for 4 min. The cell pellets were then re-suspended in 37% isotonic Percoll (4 ml per brain) and loaded into 15 ml conical tubes in between 30 and 70% layers of Percoll. The Percoll gradients were then centrifuged at 300g for 40 min at 18 C. The 70-37% interfaces, that included microglia, were collected, diluted in HBSS (1×, Sigma Aldrich) and centrifuged at 500g for 7 min at 4 C. The pellets were re-suspended in HBSS, washed three times and then seeded on poly-L-lysine precoated coverslips in 24-well plates and used for immunocytochemistry. 2.9 | ELISA measurements of interleukin-1β and TNFα release As previously described (Charolidi, Schilling, & Eder, 2015;, hippocampal slices (300 μm thick) were prepared in ice-cold HEPES-buffered medium (MEM bubbled with O 2 , pH 7.4, 42360-032, Gibco) under sterile conditions. To induce inflammasome activation and IL-1β release, slices were exposed to inflammatory-like stimuli (Bernardino et al., 2008). Each slice was placed on a Millicell cell culture insert (12 μm pore size, PIXP01250, Merck Millipore) and transferred into 24-well plates containing 800 μl and are similar with those that have been described in the literature to induce inflammatory response for mice (Hines, Choi, Hines, Phillips, & MacVicar, 2013). The amounts of IL-1β and TNFα released into the medium were measured by ELISA, using mouse Quantikine IL-1 beta/IL-1F2 (R&D Systems MLB00C) and mouse TNFα Quantikine (MTA00B) kits, respectively. Data were from at least three mice per experiment, from each of which two brain slices were used per experimental condition. Immediately after collecting the media, photographs of slices were taken and the slice surface area was determined using Image J. To compare data between brain slices in different conditions, the amounts of IL-1β and TNFα released into the medium were normalized to the slice surface area. The data were then further normalized to the mean of the control values obtained in WT mice.

| RNA isolation and quantitative RT-PCR
Total RNA was extracted from whole-brain tissues using TRIzol

| Western blot
Twenty micrograms of total protein extract from whole brain and spleen tissues were resolved on NuPAGE 4-12% Novex Bis-Tris Gels (Invitrogen, NP0321) and transferred onto nitrocellulose membranes.
Blots were probed with antibodies to P2Y 12 (kindly provided by Dr. David Julius, UCSF) and to P2Y 13 (Abcam, ab108444) followed by horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody (1:2,000, Thermo Fisher Scientific, 65-6120). Antibody binding was detected using the Luminata Crescendo Western HRP substrate (Sigma, WBLUR0500). To normalize for protein content, membranes were stripped and re-probed with anti-actin antibody (1:5,000, Sigma, A2066). Quantification of the protein expression level was done with ImageJ software using the built-in Gel Analysis protocol.  show a less complex morphology, with fewer and shorter processes compared to their wild-type counterparts.
To investigate how microglial surveillance is affected by knocking out P2Y 13 receptors, we live-imaged microglia genetically labeled with eGFP in hippocampal slices from WT and P2Y 13 KO mice using twophoton microscopy, and quantified changes of surveillance minute-byminute (see section 2). Experiments were carried out less than 4 hr after brain slicing, on microglial cells located 50-100 μm deep in the slice, to avoid microglial activation (Hanisch & Kettenmann, 2007;Kurpius et al., 2006). Briefly, the surveillance index is a measure of the number of image pixels surveyed per unit time and depends on both the number of cell processes and their speed of movement. Knocking out P2Y 13 receptors significantly reduced surveillance of the brain by microglia (Figure 2a To test whether these effects may originate from elevated cAMP levels due to the disruption of the major P2Y 13 -gated G i -protein coupled signaling pathway (Bernier et al., 2019), while imaging WT brain slices we applied forskolin, which activates adenylate cyclase and raises cAMP. This reduced surveillance by 43% (p = 1.2 × 10 −9 , Figure S2a,c,d) and ramification by 22% (p = 3.5 × 10 −11 , Figure S2a,e,f), suggesting that the altered morphology and surveillance of P2Y 13 KO microglia, which resemble what occurs after forskolin treatment in WT slices, may be mediated by a rise in cAMP levels.
These results show that microglia in mice without P2Y 13 receptors are less ramified than microglia in WT mice, and survey less brain

| Microglial surveillance still occurs in the absence of P2Y 12 and P2Y 13 function
We have recently shown that short-term block of P2Y 12 receptors with the selective antagonist PSB-0739 does not alter surveillance of the brain by microglia, implying that the normal level of ADP in the extracellular space of brain slices is too low to activate P2Y 12 receptors . P2Y 13 receptors have a similar affinity for ADP (Abbracchio et al., 2006), and so are also not (25 μM), a commonly used antagonist for P2Y 13 receptors, and measured their surveillance and ramification. We found that MRS 2211 reduced both parameters in wild-type microglia, but surprisingly also reduced them (by a similar amount) in P2Y 13 KO microglia ( Figure S3a , b). This is inconsistent with the suppression of surveillance being mediated by block of P2Y 13 receptors. Indeed, MRS 2211, which is a competitive antagonist of P2Y 13 receptors (pIC 50 = 5.97, in the presence of 100 nM ADP) is also an antagonist, albeit 20-fold less powerful , for P2Y 12 receptors and for P2Y 1 receptors (expressed by neurons and astrocytes; Zhang et al., 2014). However, neither activation nor blockade of P2Y 1 receptors, by applying MRS 2365 (10 μM) and MRS 2179 (25 μM), respectively, altered surveillance (see Figure S4a,c), while MRS 2365 weakly reduced ramification ( Figure S4b,d). These data, in combination with the lack of effect on surveillance of the P2Y 12 receptor blocker PSB-0739 , imply that the MRS 2211-evoked effects do not reflect an action on P2Y 1 or P2Y 12 receptors.
We considered the possibility that P2Y 13 and P2Y 12 might affect each other's expression, or form dimers and act cooperatively perhaps because each receptor affects the trafficking of the other to the cell surface membrane (see next section) or because of a need for ADP to bind to both P2Y 12 and P2Y 13 (assuming a dimeric receptor) for receptor activation to occur. We initially quantified the total P2Y 12 protein level in the P2Y 13 KO brain tissue by western blot and found that it was not significantly different compared to the WT tissue ( Figure S5a, b). Immunohistochemistry with a P2Y 12 antibody also showed that its protein distribution in the P2Y 13 KO microglia in fixed hippocampal slices was broadly similar to that in wild-type microglia ( Figure S5c).
To test the effect on microglial surveillance of blocking both P2Y 12 and P2Y 13 receptor function, we applied the selective P2Y 12 antagonist PSB-0739 to P2Y 13 KO slices and quantified any change in surveillance index that occurred. Blocking P2Y 12 receptors with PSB-0739 (2 μM) did not significantly affect surveillance of microglia in the P2Y 13 KO or WT (Figure 2e,f). This confirms our previous finding that the physiological levels of ATP and ADP in the brain are too low to affect microglial surveillance by activating P2Y 12 receptors  and presumably also P2Y 13 receptors, which have a similar high affinity for ADP (Abbracchio et al., 2006).
In summary, knocking out P2Y 13 receptors reduces microglial ramification and thus surveillance of the brain. These effects were also seen with the supposedly specific P2Y 13 blocker MRS 2211, but MRS 2211 also had this effect in the P2Y 13 KO and so acts by a mechanism not involving P2Y 13 receptors. Surveillance still occurred with P2Y 12 receptors blocked in the P2Y 13 knock-out, implying that no purinergic stimulation of either of these receptors is needed for microglial surveillance to occur.
3.3 | Microglia lacking P2Y 13 receptors show a normal resting potential but a larger ADP-evoked K + current To test whether the effect of knocking out P2Y 13 receptors on ramification and surveillance was mediated by changes of the membrane potential, as found previously when blocking THIK-1 K + channels in the microglial membrane , we per-  (Figure 3a,b). Applying ADP from a puff pipette (containing 100 μM ADP) located above the slice evoked an outward current at 0 mV as reported previously , however in the P2Y 13 KO the current density was~50% larger ( Figure 3c,d). The ADP-evoked current reflects the activation of THIK-1 K + channels by ADP binding to P2Y 12 receptors but THIK-1 channels are also tonically active and thus set the resting potential . Despite the increase of the ADPevoked current in the P2Y 13 KO, the tonic activity of THIK-1 was unaffected, since the resting potential, the tetrapentylammonium ATP/ADP release and activation of P2Y 12 receptors should more effectively attract microglial processes to a damage site (Haynes et al., 2006), but in fact this did not occur as the chemotactic response was attenuated in the KO (see below).
To test whether the larger ADP-evoked THIK-1 currents may reflect a raised cAMP level when P2Y 13 receptors are knocked out (see above), we perfused WT microglial cells for~10 min with cAMP via the patch solution, which enhanced the ADP-evoked current density by >70% (n = 9; p = .017: see Figure S2b). This suggests that the larger ADPevoked THIK-1 currents of P2Y 13 KO microglia may be mediated via elevated cAMP levels, in line with cAMP mimicking the reduced morphology and surveillance seen in P2Y 13 KO microglia ( Figure S2a,c-f).

| P2Y 13 receptors regulate the rate of microglial directed motility
To investigate a possible effect of P2Y 13 receptors on microglial directed motility, we imaged slices with two-photon microscopy in acute hippocampal slices prelabeled with fluorescently tagged isolectin B 4 (see section 2). Inserting a 1 mM ADP filled glass pipette (with a tip diameter of~2-3 μm) carefully into the slices induced directed motility of microglial processes to the tip of the pipette. This resulted in complete process convergence on the tip of the pipette in both WT and P2Y 13 KO slices (Figure 4a,b), however the time needed for the convergence of P2Y 13 KO microglial processes (time to 1/e) was 46% longer than for their WT counterparts (p = .04, Figure 4e,f).
We also evoked directed motility in slices of both genotypes in response to focal brain injury by applying a laser ablation of~10 μm diameter (Figure 4c,d). The two methods initiated similar microglial responses but with different kinetics (Figure 4a-d). The laser ablation induced an overall faster microglial response with the complete convergence of processes occurring~8 min after the ablation, while a chemotaxis induced by locally elevating ADP through a glass pipette took~15 min (Movie S2). Microglial processes took 22% longer to arrive at the ablation site when P2Y 13 was knocked out (p = .05, Figure 4e,f). Analysis of the processing speed revealed no differences between the WT and the KO (Figure 4g).
Overall, our results show that P2Y 13 receptors are necessary to maintain an efficient chemotactic response of microglial processes to CNS injury.
3.5 | P2Y 13 receptors reduce the resting level of interleukin 1β release IL-1β and TNFα are two major pro-inflammatory cytokines generated in response to infection and contribute to tissue injury during disease.
The production of IL-1β from innate immune cells such as macrophages and microglia requires the formation of inflammasome complexes to activate caspase 1, which generates IL-1β from its inactive precursor. Inflammasome assembly is a multi-stage process, involving priming by a Toll-Like Receptor agonist such as the bacterial coat component lipopolysaccharide (LPS), followed by a fall of intracellular [K + ] evoked by an activating signal such as ATP (Munoz-Planillo et al., 2013). We have previously shown that, in combination with priming by LPS, activation of P2Y 12 and P2Y 13 receptors with 2-MeSADP leads to inflammasome-mediated production of interleukin-1β: a process in which P2Y 12 receptors presumably evoke K + loss through the activation of THIK-1 K + channels .
To determine whether P2Y 13 receptors play a role in the generation and release of immune mediators when microglia become activated we used acute hippocampal slices from wild type and P2Y 13 KO mice and quantified (by ELISA) the level of the released IL-1β in the supernatant (see section 2). Applying LPS (50 μg/ml) for 6 hr evoked some IL-1β release, which was greatly enhanced when co-applied with ATP for the last 3 hr of LPS exposure, both in WT and P2Y 13 KO brain slices ( Figure 5a). In the KO, IL-1β release evoked by LPS and ATP was slightly but not significantly higher than in the WT. However, the level of baseline IL-1β release in the absence of added LPS or ATP, was significantly higher in P2Y 13 KO slices than in WT slices ( Figure 5a), suggesting that P2Y 13 receptors suppress the release of IL-1β in baseline conditions. This may also suggest that microglia exhibit a different functional phenotype under resting conditions, however, P2Y 13 KO microglia did not appear to be activated in the intact brain as judged by their cell body area (Davis, Salinas-Navarro, Cordeiro, Moons, & De Groef, 2017;Gyoneva et al., 2014), which was 50.6 ± 3.78 μm 2 in the WT and 47.0 ± 1.7 μm 2 in the KO.
We further tested whether the release of another cytokine, TNFα, is affected when P2Y 13 receptors are deleted using the same procedure as for IL-1β. TNFα is a proinflammatory cytokine, which is released by glial cells in the brain (mainly microglia and astrocytes) in response to pathogen-or damage-associated molecular patterns, in a process that does not require inflammasome assembly (and so presumably does not require K + loss from the cell via THIK-1). LPS evoked a similar release of TNFα in WT and P2Y 13 KO brain slices, which was independent of the presence of ATP (Figure 5b), indicating that P2Y 13 receptors do not regulate the production of TNFα by microglia.
To further investigate the impact of P2Y 13 receptor deletion on the functional state of microglia in the intact brain, we evaluated the mRNA expression of selected microglial target genes in whole brains from P2Y 13 KO mice and aged-matched WT controls by RT-PCR. Target genes were selected according to their expression level in different functional states of microglia and were classified as being F I G U R E 5 P2Y 13 KO increases baseline release of interleukin 1β in brain slices, however, the microglia in the intact P2Y 13 KO brains are not detectably activated. (a) ELISA-measured IL-1β levels released from hippocampal slices from 8 WT (at P56-P93) and 8 P2Y 13 KO mice (at P58-P105) exposed to no drugs, LPS (50 μg/ml), ATP (1 mM), or ATP + LPS, showing an enhanced resting level of IL-1β release in the KO. (b) ELISAmeasured TNFα levels released from hippocampal slices from 3 WT and 3 P2Y 13 KO mice exposed to the same protocol as (a), showing a lack of effect of knocking out P2Y 13 receptors on TNFα release. (c-d) mRNA levels of selected "activation" (il6, il-1β, iNos), "priming" (tlr2, csf1, lgals3), and "homeostasis" (p2ry12) marker genes in whole brain samples from 5 WT (at P64-P115, 4 males, 1 female) and P2Y 13 KO mice (at P62-P115, 4 males, 2 females), relative to Gapdh, as measured by quantitative RT-PCR. (e) Transcript levels of the selected markers shown in c and d in P2Y 13 KO brains expressed as n-fold change relative to their WT aged-matched controls. The number of mice is shown on bars associated with "activation" (il1b, il6, iNos), "priming" (tlr2, lgals3, tlr2), or "homeostasis" (p2ry12), following Holtman et al. (2015). The mRNA expression of all the genes tested by this method was not different in P2Y 13 KO mice (Figure 5c-e). Thus, our results show that P2Y 13 receptors suppress the release of IL-1β in brain slices, however, microglial cells lacking P2Y 13 receptors were not detectably activated.

| DISCUSSION
P2Y 12 and P2Y 13 receptors are both highly expressed in microglia at the mRNA level (Zhang et al., 2014), both use ADP as their preferred physiological agonist with an EC 50 value of~60 nM (see overview in Abbracchio et al., 2006) and share a similar pharmacological profile (Zhang et al., 2002). Both receptors couple predominantly to Gα i proteins and so their activation will inhibit the adenylate cyclase and thus decrease intracellular cAMP levels. P2Y 12 activation has been shown to mediate directed motility of microglial processes to an ATP/ADP source (Haynes et al., 2006), but the function of microglial P2Y 13 receptors is unknown.
We took advantage of the existence of a P2Y 13 receptor knockout mouse (Fabre et al., 2010) to investigate the role of this receptor in microglial function. Although P2Y 13 is highly expressed at the mRNA level in microglia (Stefani et al., 2018;Zhang et al., 2014), its function has been little studied. This partly reflects a lack of specific pharmacological tools to activate or inhibit this receptor. MRS 2211 is often used as a "specific" P2Y 13 receptor blocker, but we found that it affected microglia even when P2Y 13 receptors were knocked out, thus implying that it is not a specific tool to study the role of P2Y 13 receptors in any tissue. Haynes et al. (2006) and Stefani et al. (2018) found that, despite confirming P2Y 13 mRNA expression in microglia (Zhang et al., 2014), its expression could not be detected at the protein level.
The inability to antibody-label P2Y 13 may reflect a high rate of constitutive proteasomal degradation (Pons et al., 2014), or may alternatively suggest that P2Y 13 protein is only expressed at high levels in microglial culture (Quintas, Vale, Goncalves, & Queiroz, 2018) or upon microglial activation in situ .
We found that knock-out of P2Y 13 receptors had four main effects on microglia. (a) It reduced their ramification (Figure 1) which consequently reduced their surveillance of the brain parenchyma. (b) It enhanced the outward THIK-1 K + current evoked by ADP activating P2Y 12 receptors. This is mediated by an increased coupling of P2Y 12 receptors to THIK-1 channels (and not an increased number of THIK-1 channels) since the tonic activity of THIK-1 and thus the resting potential was not changed (Figure 3). (c) Surprisingly, despite the potentiation of P2Y 12 receptor function as assessed by ADP-evoked THIK-1 activation, the time needed for P2Y 12 receptor-mediated convergence of microglial processes to an ADP source or to local injury triggered by a laser ablation was prolonged in the P2Y 13 KO (Figure 4). (d) The release of the pro-inflammatory cytokine IL-1β was increased in the P2Y 13 KO in the absence of stimulation with LPS and ATP ( Figure 5), suggesting the adoption of a different functional state with some resemblance to mild activation, although microglial cells were not detectably activated as judged by the lack of expression of classical microglial activation markers ( Figure 5). Thus, P2Y 13 contributes to maintaining microglial morphology and surveillance and suppresses inappropriate IL-1β release.
The attraction of microglial processes to an ATP source or laser lesion involves activation of P2Y 12 receptors by ADP (Haynes et al., 2006) and is independent of THIK-1 activity . P2Y 12 receptor protein levels were not changed in the P2Y 13 KO (at least at the level of the whole brain: Figure S5) and the P2Y 13 KO microglia show an enhanced THIK-1 K + current (Figure 3) evoked by ADP , which suggests that the sensi- , and a larger P2Y 12 -activated THIK-1 current ( Figure S2b).
Despite this increased sensitivity to ADP, the P2Y 13 KO cells moved their processes at a similar speed to those of the WT towards an ADP source, where they nevertheless arrived later because the KO cells are less ramified and have shorter processes, and thus they have to grow further than the WT cell processes until they reach their target.
Previously we have shown that, in nonactivated microglia, ramification is controlled by the cell's membrane potential, so that microglia become less ramified when the cell is depolarized . The loss of ramification seen upon deleting P2Y 13 cannot be explained in this way because the resting potential of the cells was unchanged (Figure 3e), and may arise from the fact that the microglia have adopted a different functional state (Norden & Godbout, 2013;Perry & Holmes, 2014). RT-PCR analysis of the two genotypes at the whole-brain level did not reveal differences of the activation state of P2Y 13 KO microglia (Figure 5c,e), but this may reflect a lack of sensitivity of the assay. The fact that more IL-1β protein is released in the P2Y 13 KO (Figure 5a), while the mRNA level for il-1β is unchanged (Figure 5c,e) may reflect a poor correlation between changes of protein level and transcriptional, translational, and post-translational processes, which has previously been described across different systems, conditions, and developmental stages (Liu, Beyer, & Aebersold, 2016;Peshkin et al., 2015;Taniguchi et al., 2010).
We suggest above that some of the effects of P2Y 13 deletion may reflect an increase of cAMP concentration in the cell, but other signaling pathways may also be associated with the P2Y 13 receptor. These include G q -coupled elevations of intracellular [Ca 2+ ] mediated by internal store release and activation of store-operated calcium channels (Carrasquero et al., 2009;Zeng et al., 2014), Erk1/2 activation (Ortega, Perez-Sen, Delicado, & Teresa Miras-Portugal, 2011), and regulation of transcription through a rise in nuclear calcium level and activation of nuclear factor erythroid-derived 2-like 2, a transcription factor that regulates the oxidative stress response (Espada et al., 2010;Lyubchenko et al., 2011). P2Y 13 may also inhibit Akt (Chatterjee & Sparks, 2012), which is an important mediator of P2Y 12mediated actin re-organization. In addition, the levels of P2Y 13 may directly influence the levels of other purinergic receptors found at the microglial membrane due to the influence that dimerization between different P1/P2 receptors can have on receptor trafficking and retention at the cell surface (Schicker et al., 2009). However, much of the evidence pertaining to the downstream signaling pathways associated with P2Y 13 should be interpreted with caution, especially in microglia, for the following reasons. First, many studies were conducted in cell types other than microglia, or in cell lines or cultured cells, and there may be differences in the signaling between cell types. Second, almost all of these studies relied highly upon the use of MRS 2211 to block P2Y 13 receptors, but we have now shown this is not a selective P2Y 13 antagonist. Furthermore, the failure, to date, to demonstrate microglial P2Y 13 protein expression in vivo suggests either that the commercially available P2Y 13 antibodies lack specificity or that the constitutive proteasomal degradation of the receptor shown to regulate the surface expression of the receptor in hepatocytes (Pons et al., 2014) also happens in microglia in vivo (which would imply that a low level of surface receptors can still have a significant effect on cell function).
The high mRNA expression level for P2Y 13 in microglia, and the P2Y 13 KO-evoked effects on microglial function that we have characterized, both suggest that P2Y 13 plays a direct role in regulating microglial function. Our finding that knocking out P2Y 13 receptors results in less ramified microglia in the brain under resting conditions is in line with recent studies showing a general reduction in the structural complexity of microglia in P2Y 13 KO mice (Stefani et al., 2018) and in mice with constitutive deletion of the ATP degrading enzymes CD39 and CD73 (which prevents hydrolysis of extracellular ATP and thus the formation of the P2Y 13 receptor agonist ADP; Matyash et al., 2017).
Nevertheless, the P2Y 13 knock-out mouse that has been used is a global KO, and so the effects seen could in principle be mediated by a loss of P2Y 13 elsewhere in the body. Expression of P2Y 13 receptor mRNA occurs not only in the brain, but also in the liver, spleen, lymph nodes, and bone (Communi et al., 2001). P2Y 13 KO mice show a decrease in hepatic high-density lipoprotein (HDL) uptake, hepatic cholesterol content, and biliary cholesterol output, although their plasma HDL and other lipid levels are normal (Fabre et al., 2010).
These metabolic changes result in a substantial decrease in the rate of macrophage-to-feces reverse cholesterol transport (RCT), a process whereby excess peripheral cholesterol, especially that in macrophage foam cells, is taken up to be incorporated into HDL particles and delivered to the liver for excretion in the feces. Thus, although it has not yet been studied, a rise in microglial cholesterol level might occur as a result of these peripheral changes. Since cholesterol levels regulate microglial function (Bohlen et al., 2017) and high cholesterol is reported to lead to microglial activation (Zatta, Zambenedetti, Stella, & Licastro, 2002), it is conceivable that the effects we report are mediated by altered systemic cholesterol trafficking. Construction of a microglial-specific P2Y 13 receptor KO mouse will help further investigations of the role of P2Y 13 receptors in microglia, without interference from the actions of this receptor in peripheral tissues.