Undisturbed climbing fiber pruning in the cerebellar cortex of CX3CR1‐deficient mice

Pruning, the elimination of excess synapses is a phenomenon of fundamental importance for correct wiring of the central nervous system. The establishment of the cerebellar climbing fiber (CF)‐to‐Purkinje cell (PC) synapse provides a suitable model to study pruning and pruning‐relevant processes during early postnatal development. Until now, the role of microglia in pruning remains under intense investigation. Here, we analyzed migration of microglia into the cerebellar cortex during early postnatal development and their possible contribution to the elimination of CF‐to‐PC synapses. Microglia enrich in the PC layer at pruning‐relevant time points giving rise to the possibility that microglia are actively involved in synaptic pruning. We investigated the contribution of microglial fractalkine (CX3CR1) signaling during postnatal development using genetic ablation of the CX3CR1 receptor and an in‐depth histological analysis of the cerebellar cortex. We found an aberrant migration of microglia into the granule and the molecular layer. By electrophysiological analysis, we show that defective fractalkine signaling and the associated migration deficits neither affect the pruning of excess CFs nor the development of functional parallel fiber and inhibitory synapses with PCs. These findings indicate that CX3CR1 signaling is not mandatory for correct cerebellar circuit formation.


| INTRODUCTION
Microglia are the resident phagocytes of the central nervous system and act as the main immune effectors against pathogens. They derive from primitive yolk-sac macrophages and establish during early prenatal development in the brain (Ginhoux et al., 2010;Kierdorf et al., 2013), where they subsist throughout the entire life span of the organism by slow turnover (Goldmann et al., 2016).
Besides their function as early sensors for virtually all neuropathological events (Ransohoff & Perry, 2009) additional roles for microglia in the healthy brain during synaptic plasticity and refinement have been described recently (Paolicelli et al., 2011;Schafer et al., 2012;Tremblay, Lowery, & Majewska, 2010). The exact molecular mechanisms regarding microglial-mediated pruning remain unclear.
One proven mechanism depends on classical complement cascade elements acting in an activity-dependent manner, which has been described in the retinogeniculate system (Schafer et al., 2012; Stephan, Barres, & Stevens, 2012;Stevens et al., 2007). Complement components C1q and C3 localize to all synapses but microglia preferentially engulf inputs from "weak" synapses for phagocytosis.
Another possible key player of synaptic remodeling, the fractalkine receptor (CX 3 CR1) signaling, moved into focus of interest.
In the immature cerebellum, a given Purkinje cell (PC) is innervated by multiple CFs (Crepel et al.,1976). Within the first three postnatal weeks, CFs undergo a substantial pruning of redundant connections resulting in the typical CF mono-innervation of PCs in the mature cerebellum (Crepel et al.,1976;Mason & Gregory,1984).
In this study, we tested whether the microglial fractalkine receptor CX 3 CR1, as in the hippocampus (Paolicelli et al., 2011), is a key player for functional pruning of the CF-to-PC synapse. We analyzed, whether recruitment of microglia to the site of CF elimination, particularly the PC and molecular layer, as well as the morphological characteristics of microglia are altered in CX 3 CR1-deficient mice. Further, using electrophysiology, we tested whether CX 3 CR1-deficient mice show a normal postnatal maturation, that is, the normal elimination of redundant CFs within the first 3 weeks after birth. Additionally, we analyzed whether the formation of PF-to-PC synapses as and GABAergic inhibitory synapses on PCs is normal in mice lacking the CX 3 CR1 receptor.

| Genotyping
The genotype of CX 3 CR1 +/+ and CX 3 CR1 −/− mice was confirmed by PCR from tail biopsies using CX 3 CR forward (TCAGTGTTTTCTCCCGCTTGC) plus CX 3 CR-WT reverse (CAGTGATGCTCTTGGGCTTCC) primers or CX 3 CR forward plus CX 3 CR-knock in (eGFP) reverse (GTAGTGGTTGTCGGGCAGCAG) primers, respectively. PCR was performed as follows: one cycle at 95 C of 3 min, 40 cycles comprising 30 s at 95 C, 30 s at 60 C and 1 min at 72 C and one cycle at 72 C of 5 min.

| Image acquisition and quantification of microglia in the cerebellum
For quantification of microglia, IHC sections were imaged with an inverted confocal laser-scanning microscope equipped with an 40x/1.14 NA oil immersion objective (FW 300, Olympus, Hamburg, Germany) and an upright laser-scanning microscope equipped with an 20×/0.8 NA objective (LSM 800 with Airyscan, Zeiss, Jena, Germany).
Stitching of the images acquired with the LSM 800 was accomplished with the Zen 2 (blue edition) software (Zeiss). Combining and pseudocoloring was done in FIJI; linear brightness and contrast adjustments were done homogenously in each combined image. For the assignment of the morphological index, soma size and arborization area of the microglia (Figures 5 and S1) was quantified by using FIJI freehand and polygon selection tool. The morphological index was calculated as described before (Basilico et al., 2019;Tremblay, Zettel, Ison, Allen, & Majewska, 2012).

| Electrophysiological recordings
Acute slice preparation from P9 to P17 CX 3 CR1 −/− , CX 3 CR1 +/+ mice, or C57BL/6 were performed by decapitating and rapidly transferring the brain into ice-cold artificial cerebrospinal fluid (ACSF) saturated with carbogen (95% 0 2 /5% CO 2 ). The ACSF included 20 mM Glucose, 125 mM NaCl, 1.25 mM NaH 2 PO 4 , 26 mM NaHCO 3 , 2 mM CaCl 2 , and 1 mM MgCl 2 at pH 7.4. Slicing procedure and whole-cell patchclamp recordings were performed as described previously (Pätz et al., 2018). For extracellular stimulation of CFs, a stimulation electrode was placed in the GCL in 20-100 μm distance to a patched PC. CF activation was assured by an all-or-none response of the first excitatory postsynaptic current (EPSC) to a paired stimulus resulting in a step-wise stimulus-response curve (SRC) and by paired-pulse depression (Konnerth, Llano, & Armstrong, 1990). The number of CFs innervating the patched PC was estimated by increasing the stimulation strength, which, in the case of multiple CF innervation, results in a stepwise recruitment of CFs (Bosman, Takechi, Hartmann, Eilers, & Konnerth, 2008). In some recordings, the number of CFs per PC was estimated using two to three independent stimulation electrodes.
Here, independent stimulation of individual CFs was ascertained by sequential stimulation of the different CF inputs and absence of paired-pulse depression in crossed stimulations (Bosman et al., 2008;Pätz et al., 2018). To allow proper voltage clamping of CF responses, the normal ACSF solution was supplemented with a submaximal concentration (Pätz et al. 2018; 1 mM) of the rapid glutamate receptor antagonist kynurenic acid. CF and PF responses were acquired in the presence of 10 μM gabazine to block spontaneous GABAergic currents.
For recording of PF paired-pulse behavior (50 ms stimulus interval) and SRCs (repeated every 0.02 s; Figure 6), a stimulation pipette was placed in the ML and PF activation was assured by a graded response of the first EPSC and paired-pulse facilitation (Konnerth et al., 1990).  , and P21 (c). PCs were stained with calbindin (CB, red), microglia with IBA1 (green), and nuclei with DAPI (blue). Note the enrichment of IBA1 + microglia in the PCL around P6 and P8. Scale bars 50 μm. (d-i) Density of IBA1 + microglia in the cerebellar white, and the gray matter as well as the granule cell, the Purkinje cell, molecular, and the external granule cell layer during postnatal development (P4-P21, wild type). Note that the EGCL was detectable only until P15 and that the ML was detectable from P8 onward. Data are presented as median + interquartile range (IQR) (n = 5). Asterisks denote statistical differences between the ages (Kruskal-Wallis test per age group, *p ≤ .05, **p ≤ .01, ***p ≤ .001, ****p ≤ .0001) specific heterogeneity in the cerebellar cortex (Ashwell, 1990;Vela, Dalmau, González, & Castellano, 1995) with features of phagocytotic activity such as phagocytotic cup formation (Perez-Pouchoulen, Van-Ryzin, & McCarthy, 2015). We addressed the question whether microglia become specifically enriched nearby CF collaterals, that is, in the PCL, during the period in which excess CF synapses undergo pruning. To this end we quantified the number of microglia cells, stained by IBA1, in P4-P21 wild-type (WT) mice in a standardized areaof-interest in lobulus IV/V (Figure 1).
In the WM, microglia were present at a relatively stable density of around 600 cells per mm 2 with no significant differences ("n.

| Delayed population of the granule and ML of CX 3 CR1 knockout mice during early postnatal development
Earlier studies have shown that lack of the fractalkine receptor CX 3 CR1 results in reduced microglia density and delayed synaptic pruning in the hippocampus (Pagani et al., 2015;Paolicelli et al., 2011). Moreover, microglial physiology has been suggested to be affected by knockout of CX 3 CR1 as a deteriorated migration and dynamic motility of processes was observed in retinal microglia lacking CX 3 CR1 (Liang et al., 2009;Paolicelli et al., 2014). We hypothesized that CX 3 CR1 signaling may be critically involved in synaptic pruning of CFs and, correspondingly, that deletion of CX 3 CR1 might impair microglial migration and microglia allocation toward the PCL.

| Microglial morphology is not affected by deletion of CX 3 CR1
CX 3 CR1 has been shown to affect the morphology of microglia in the hippocampus (Pagani et al., 2015). Besides investigating the overall distribution of microglia within the cerebellar cortex, we also analyzed the morphological appearance of microglia during early postnatal development. We calculated the morphological index as the ratio of soma size and arborization area (Basilico et al., 2019) for microglia of WT and CX 3 CR1-deficient mice and binned the cells into three age groups: P5-P8, P9-P13, and P15-P21 ( Figure 5).
There was a strong tendency of a declining morphological index during postnatal development of both strains pointing toward the F I G U R E 4 Proximity of microglia processes, PCs, and climbing fibers (CFs). (a) Immunostaining of a sagittal cerebellar slice from a P8 Igsf9-eGFP mouse (full volume z-stack/88 slices). Microglia were stained with IBA1 (green), PCs with calbindin (CB, blue) and CFs with an antibody against GFP (red). The dashed box is shown in higher magnification in (b and b 0 ). (b and b 0 ) Single plane confocal image of microglial processes (green, marked by arrowheads) in close proximity to the CFs (red) in the vicinity of PCs. Scale bars 50 μm ramification of microglia. The morphological index in WT mice thereby decreased from 0.15 at P5-P8 to 0.05 at P9-P13 and 0.1 at P15-P21. Though the difference between the age groups was highly significant in WT and CX 3 CR1 −/− mice (*** between P5-P8 and P9-P13, P15-P21, respectively), significant differences between both strains could be sporadically found in deeper analysis between age and genotype ( Figure S1c,l). Thus, these results suggest that the morphological maturation of cerebellar microglia is mostly undisturbed in CX 3 CR1 −/− mice.

| Formation of PF-to-PC synapses is normal in CX 3 CR1 knockout mice
Previous studies demonstrated that normal formation of PF-to-PC synapses is prerequisite for an undisturbed CF pruning during the late phase of elimination (Kano et al., 1998;).
Here, we found that around P12, the time at which "winner" CFs begin to stabilize their synapses on PC dendrites, microglia colonization of the ML is delayed in CX 3 CR1-deficient mice (Figure 4f). To elucidate, whether PF-to-PC synapse formation is affected by this transient reduced number of microglia cells, we tested whether the establishment of functional PF-to-PC synapses is undisturbed CX 3 CR1-deficient mice.
To this end, we examined the paired-pulse behavior (50 ms interstimulus interval) of PF-EPSCs in P9-P13 CX 3 CR1 +/+ and CX 3 CR1 −/− mice ( Figure 6). No significant difference in the paired-pulse ratios was found between WT and KO mice (p = .098, Figure 6a) suggesting that short-term synaptic plasticity and the presynaptic function of PFs seems to be undisturbed in mice lacking CX 3 CR1. To compare the density of functional PF-to-PC synapses, the SRC of PF-EPSCs was analyzed. Figure 6b

| Pruning of CFs is not affected by CX 3 CR1 deletion
Previous studies implicated a role for CX 3 CR1 in the functional maturation and elimination of synapses in the somatosensory cortex and Dots show mean ± SEM of n = 15 PF-PC connections from n = 5 mice per genotype. No difference was observed between WT and KO mice (Wilcoxon rank test) Scale bar = 100 pA and 10 ms the hippocampus (Hoshiko et al., 2012;Paolicelli et al., 2011). We, therefore, hypothesized that fractalkine signaling via CX 3 CR1 may be critically involved in the pruning process of CFs. To this end, we used electrophysiological recordings to analyze CF elimination in WT and in CX 3 CR1-deficient mice.
We quantified the probability of finding more than one CF per PC in P9-P17 WT and CX 3 CR1 −/− mice. For WT mice, as described previously (Bosman et al., 2008), we found that the switch from multipleto mono-innervation occurs mostly within 17 days after birth (Figure 7c, left) showing a relatively strong scatter in each age group.
A similar pattern was observed in CX 3 CR1 −/− mice (Figure 7c, right), with no significant differences compared to WT mice. Thus, CF elimination proceeds normally in CX 3 CR1-deficient mice, arguing against CX 3 CR1 signaling playing a dominant role in CF pruning.

| Inhibitory transmission is not affected by CX 3 CR1 deletion
Recently, microglia have been shown to be crucial for the development of functional inhibitory synapses on PCs (Nakayama et al., 2018) and that the formation and activation of GABAergic inhibitory synapses during the second postnatal week is critically involved in the refinement of CF-to-PC synapses (Nakayama et al., 2012;Nakayama et al., 2018). To test, whether fractalkine signaling via CX 3 CR1 is critically involved the formation of GABAergic inhibitory synapses on PCs, we examined mIPSCs in P9-P13 WT and CX 3 CR1 −/− mice. We found that the amplitude as well as the frequency of mIPSCs in CX 3 CR1 −/− mice do not show significant differences compared to WT mice (p = .24 (Figure 8b), p = .59 [ Figure 8c]). These results indicate that inhibitory synaptic transmission is normal in CX 3 CR1 −/− mice and, hence, that CX 3 CR1 is not crucial for the development of functional GABAergic inhibitory synapses on PCs.

| DISCUSSION
The contribution of microglia-mediated fractalkine signaling via the CX 3 CR1 receptor in synaptic pruning appears to be heterogeneous throughout different brain regions (Lowery, Tremblay, Hopkins, & Majewska, 2017;Paolicelli et al., 2011;Reshef et al., 2017). Here, we report that CX 3 CR1, while relevant for normal microglial population of the cerebellar cortex, is not required for functional pruning of CF-to-PC synapses. Specifically we found that: (a) knockout of CX 3 CR1 leads to a slightly delayed microglial population of the granule and the ML,
In this study, we focused on the impact of CX 3 CR1 deletion on microglial population of the cerebellum and investigated if microglia do contribute to cerebellar circuit development. Glia cells have been associated with pruning-related incorporation and removal of cell debris, comprising CFs collaterals (Song et al., 2008) and postsynaptic elements of PCs (Morara et al., 2001). Here we show, consistent with previous findings (Nakayama et al., 2018;Perez-Pouchoulen et al., 2015), that after birth, microglia translocate from the WM toward the cerebellar cortex ( Figure 2). We further show that population of the PCL and the ML increases during early postnatal development peaking around P8-P13 (Nakayama et al., 2018), which coincides with the period of CF pruning in mouse cerebellum (Hashimoto, Ichikawa, et al., 2009;Scelfo & Strata, 2005). At the same time, microglia specifically extend their processes toward PC somata (Figure 4), raising the possibility that microglia indeed participate in the pruning process of CF-to-PC synapses.
Interestingly, microglia seem to rarely direct engulf CFs during postnatal development in the cerebellum raising the possibility that the majority of CFs might not be pruned by direct engulfment but rather indirectly by synaptic mechanisms, in which microglia may be involved (Nakayama et al., 2018). Besides, hippocampal microglia have been shown to modulate presynaptic structures by selective partial phagocytosis and to be able to induce postsynaptic spine head filopodia by selective engulfment of dendritic spine heads (Weinhard et al., 2018). Both functions support the hypothesis that microglia rather not directly phagocytose material in the cerebellum unlike in other brain regions like the postnatal retinogeniculate system (Schafer et al., 2012).
Our findings suggest, that microglia specifically populate the PCL at the time window of CF elimination giving the possibility of their contribution in the developmental elimination of CFs. Whether microglia directly remove presynaptic and/or postsynaptic structures or whether glia-dependent mechanisms stimulate CF elimination indirectly (Nakayama et al., 2018) requires further investigations.

| CX 3 CR1 signaling and synaptic pruning
In the healthy brain, expression of the chemokine receptor CX 3 CR1 is restricted to microglia and macrophages, while expression of fractalkine, CX 3 CL1, its only known ligand, is confined to selected neurons indicating their contribution in neuron-microglia interaction (Harrison et al., 1998;Hughes, Botham, Frentzel, Mir, & Perry, 2002;Jung et al., 2000;Nishiyori et al., 1998). Here, CX 3 CR1 −/− mice were used to analyze the possible contribution of fractalkine signaling in CF pruning in the developing cerebellum.
CF elimination starts around P7 resulting in CF-to-PC monoinnervation around P14 (Hashimoto, Ichikawa, et al., 2009;Scelfo & Strata, 2005). Regression of CFs has been described by the probability of finding multiple CFs per PC, which is around 90% at P7 and drastically declines to about 30% during the second postnatal week (Bosman et al., 2008) and to <2% in mature animals (Eccles, Llinás, & Sasaki, 1966). Our data show a prolonged probability of finding and frequency (c) of mIPSCs. Note that there is no significantly difference between WT and CX 3 CR1 −/− mice (Wilcoxon rank test). Data were obtained from 28 cells from n = 14 WT mice and 25 cells from n = 7 CX 3 CR1 −/− mice. Data are presented as median + SEM of median (bars) as well as individual data points (black and gray for WT and CX 3 CR1 −/− mice, respectively) multiple innervation in WT mice and CF-to-PC mono-innervation at P17 (Figure 6), which is somewhat later than previously proposed (Hashimoto, Ichikawa, et al., 2009;Scelfo & Strata, 2005). The latter studies were performed on PCs located in rostral lobules II-VI (Hashimoto, Yoshida, et al., 2009) or unspecified areas (Bosman et al., 2008). Thus, use of different lobules may affect the probability, although the number of CFs in lobules of the caudal and rostral cerebellum are assumed being equal after P7 (Hashimoto et al., 2001).
Here, we show that CX 3 CR1 −/− mice did neither show a disturbed pruning of CFs ( Figure 6) nor a disturbed migration of microglia toward the PCL at the time of CF elimination (Figure 3). Thus, in contrast to synaptic refinement in the cortex, hippocampus and olfactory bulb (Hoshiko et al., 2012;Paolicelli et al., 2011;Reshef et al., 2017), elimination of excess CFs does not rely on fractalkine signaling. In view of the weak expression of fractalkine in the cerebellum compared to the other brain areas (Harrison et al., 1998;Nishiyori et al., 1998), CX 3 CR1 may be a region-specific mediator of synaptic refinement.

| CX 3 CR1 signaling and formation of functional synapses
Microglia have been reported to induce synapse formation in, for example, the hippocampus (Lim et al. 2013) and the somatosensory cortex (Miyamoto et al. 2016). In the mouse barrel cortex, microglial CX 3 Cr1 signaling is critical for long-term remodeling of synapses (Gunner et al., 2019). In the cerebellum, microglia have recently been shown to be crucial for normal formation of functional GABAergic inhibitory synapses on PCs during postnatal development (Nakayama et al., 2018). A disturbed inhibitory synaptic transmission, caused by severe reductions of microglia cells, has been associated with an impairment of CF synapse elimination (Nakayama et al., 2018). Furthermore, the CX 3 CR1 signaling pathway has been reported to affect functional synapse maturation (Hoshiko et al., 2012;Zhan et al., 2014), plasticity (Maggi et al., 2009;Paolicelli et al., 2011;Rogers et al., 2011), and activity (Heinisch & Kirby, 2009). Here, we report that the number of microglia cells in the granule and ML of CX 3 CR1-decifient mice is slightly reduced during the second postnatal week (Figure 3). We show that, at the same time, the formation of functional PF and GABAergic inhibitory synapses on PCs, both prerequisites for proper CF maturation (Kano et al., 1997;Nakayama et al., 2012), is normal in CX 3 CR1-KO mice (Figures 6 and 8). Thus, absence of CX 3 CR1 and the accompanied transient reduction of cerebellar microglia neither influences the development of PF-to-PC synapses nor the inhibitory transmission on PCs. In line with our results, an undisturbed formation of PF-to-PC synapses was also reported when microglia where almost absent during postnatal development (Nakayama et al., 2018) suggesting that microglia are not crucial for proper formation of PF synapses. The observed normal establishment of cerebellar GABAergic inhibitory synapses in CX 3 CR1-deficient mice, in line with previously reported data (Nakayama et al., 2018), implies that fractalkine signaling is not involved in microglia-mediated formation of functional inhibitory synapses in the developing cerebellum (Nakayama et al., 2018). Likewise, there is evidence that the development of GABAergic synapses in the hippocampus is not affected by knockout of CX 3 CR1 (Bertot, Groc, & Avignone, 2019). Taken together, CX 3 CR1 seems not to be substantial for normal maturation of GABAergic synapses but may have a heterogeneous role in the functional development of glutamatergic synapses in different brain regions (Bertot et al., 2019;Zhan et al., 2014).

| CX 3 CR1 signaling and microglial mobility and morphology
Fractalkine has been shown to induce migration of cortical (Maciejewski-Lenoir, Chen, Feng, Maki, & Bacon, 1999) and retinal microglia (Zhang, Xu, Liu, Ni, & Zhou, 2012). Earlier studies found impaired microglia dynamics in various brain regions of CX 3 CR1-deficient mice suggesting transient defects such as slowing of movements of microglial processes and reduced migration and infiltration capacity (Gunner et al., 2019;Liang et al., 2009;Lowery et al., 2017;Paolicelli et al., 2011). Our data on microglial population of the developing cerebellum of CX 3 CR1 −/− mice (Figure 3) show that fractalkine signaling modulates the migratory behavior of microglia also in the cerebellum, indicating that fractalkine signaling is universally required for normal microglia migration across all brain regions. Since the observed effects indicate a transient delayed, but otherwise normal population of the different layers of the cerebellar cortex, CX 3 CR1 seems not to play a major role in populating the cerebellum cortex.
We found that CX 3 CR1 deficiency does not affect microglia morphology ( Figure 5). These results are similar to findings in the retina (Liang et al., 2009). In hippocampal slices and the somatosensory cortex of CX 3 CR1 −/− mice, on the other hand, a reduced mobility of microglia has been reported (Hoshiko et al., 2012;Pagani et al., 2015).
However, this discrepancy may be explained by the short postnatal time period in which microglia may be affected. Altogether, this study implies that fractalkine signaling plays a region-specific role in microglia migration and morphology and in synaptic refinement. For pruning of CF-to-PC synapses, CX 3 CR1 is not an essential player.

| CONCLUSION
In summary, we report that CX 3 CR1 deficiency leads to transient defects in microglial population of the cerebellar cortex, but does not impair pruning of excess CF synapses. This suggests that CX 3 CR1 exhibits a rather heterogeneous role in synaptic remodeling within specific brain regions. Whether microglia contribute to activity-dependent plasticity of CF-to-PC signaling and how fractalkine signaling is involved in these processes remains subject of further investigations.

ACKNOWLEDGMENTS
The authors thank Gudrun Bethge, Susann Hähnel, and Judith Leyh for technical assistance and scientific advices and Hartmut Schmidt for advices on statistical analyses. This work was supported by funds of the German Research Foundation to J. E. and I. B. (EI 342/5) and to I. B. (BE 2272/1-3).

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.