In vivo imaging of the phagocytic dynamics underlying efficient clearance of adult‐born hippocampal granule cells by ramified microglia

The phagocytosis of dead cells by microglia is essential in brain development and homeostasis. However, the mechanism underlying the efficient removal of cell corpses by ramified microglia remains poorly understood. Here, we investigated the phagocytosis of dead cells by ramified microglia in the hippocampal dentate gyrus, where adult neurogenesis and homeostatic cell clearance occur. Two‐color imaging of microglia and apoptotic newborn neurons revealed two important characteristics. Firstly, frequent environmental surveillance and rapid engulfment reduced the time required for dead cell clearance. The motile microglial processes frequently contacted and enwrapped apoptotic neurons at the protrusion tips and completely digested them within 3–6 h of the initial contact. Secondly, while a single microglial process engaged in phagocytosis, the remaining processes continued environmental surveillance and initiated the removal of other dead cells. The simultaneous removal of multiple dead cells increases the clearance capacity of a single microglial cell. These two characteristics of ramified microglia contributed to their phagocytic speed and capacity, respectively. Consistently, the cell clearance rate was estimated to be 8–20 dead cells/microglia/day, supporting the efficiency of removing apoptotic newborn neurons. We concluded that ramified microglia specialize in utilizing individual motile processes to detect stochastic cell death events and execute parallel phagocytoses.

of their life (Sierra et al., 2010). However, the mechanism underlying the continuous elimination of newborn cells in the DG and its biological role in the homeostasis of the hippocampal neural circuits has not been fully elucidated.
Microglia are resident immune cells that are responsible for a variety of brain functions, both in physiological and pathological conditions (Butovsky & Weiner, 2018;Hammond et al., 2018;Thion et al., 2018;Tremblay et al., 2011). In the resting state, microglia occupy non-overlapping territories, exhibit a ramified morphology, and continuously search the tissue environment using their motile processes (Davalos et al., 2005;Nimmerjahn et al., 2005;Wake et al., 2009). The continuous search for, and frequent contact of microglial processes with nearby neuronal components have been shown to be essential for the proper formation and maintenance of neural circuits via the active pruning of synapses (Paolicelli et al., 2011;Wake et al., 2009;Weinhard et al., 2018;Wilton et al., 2019) or the promotion of synapse development (Huangi et al., 2021;Parkhurst et al., 2013). In contrast, in the presence of tissue injury and inflammation, microglia dramatically change their shape into an amoeboid morphology (Borst et al., 2021;Savage et al., 2019).
Based on their morphological resemblance to peripheral macrophages, it has been assumed that amoeboid microglia are more efficient in eliminating cell corpses and tissue debris (Kettenmann et al., 2011).
However, the simple association of amoeboid morphology and phagocytic activity has been challenged by a number of recent findings (Paolicelli et al., 2022). Nevertheless, previous studies have reported the continuous clearance of newborn GCs by ramified microglia in the adult DG (Abiega et al., 2016;Diaz-Aparicio et al., 2020;Sierra et al., 2010). A special membranous extension at the distal microglial process, termed phagocytic pouch, enwraps the apoptotic newborn neurons in the unperturbed DG. The unique mechanism of elimination of newborn neurons by ramified microglia in the subgranular zone (SGZ) contradicts the traditional view of two distinct classes of microglia, that is, quiescent ramified microglia and amoeboid microglia with active phagocytic activity.
Despite the abundance of apoptotic newborn neurons in the hippocampus, there is limited information about the dynamic interaction between microglial processes and dying neurons, the time course of the engulfment process, and the digestion of phagolysosomal contents. This is partly attributed to the difficulty in visualizing unchallenged microglia in the deep subcortical regions, such as the DG (Kamei et al., 2022;Yang et al., 2010). Previous two-photon imaging studies of adult cortical microglia successfully revealed the dynamic behavior of ramified microglia in physiological conditions (Kondo et al., 2011). However, the surgical procedures used for the implantation of an imaging window above the DG easily induce tissue injury and inflammation, which profoundly alter microglial behaviors (Kamei et al., 2022). Therefore, successful in vivo monitoring of phagocytosis by ramified microglia in the DG requires minimally invasive implantation techniques of imaging windows combined with genetically modified mice expressing distinct fluorescent proteins in newborn neurons and microglia.
Here, we investigated the phagocytic activity of ramified microglia in the SGZ and their interaction with apoptotic newborn neurons in vivo. Dual-color imaging of microglia and apoptotic newborn neurons revealed two important characteristics. First, frequent environmental surveillance and rapid engulfment reduced the time required for dead cell clearance. The digestion of cell corpses within a single microglial process was estimated to be completed within 3-6 h of the initial contact. Second, parallel digestion of multiple dead cells by a single microglial cell increased the clearance capacity. These two characteristics of ramified microglia contributed to their phagocytic speed and capacity, respectively. The estimated cell clearance rate was 8-20 dead cells/microglia/day, thus supporting the efficiency of the removal of apoptotic newborn neurons. From these observations, we propose that ramified microglia are designed to utilize individual motile processes to detect stochastic cell death events and execute parallel phagocytoses effectively.
All animal experiments were approved by, and carried out according to the policies of, the Animal Ethics Committee of the University of Tokyo. Four-to-twelve-week-old mice were used unless otherwise stated.

| Animal treatments
5-Ethynyl-2 0 -deoxyuridine (EdU; Tokyo Chemical Industry #E1057) prepared at 5 mg/mL in sterile phosphate-buffered saline (PBS) was administered intraperitoneally to mice at 50 mg/kg, to label the dividing cells within a specific time window. Tamoxifen (Tam; Sigma-Aldrich #T5648) was dissolved in corn oil at 20 mg/mL and was administered to mice orally using 20G feeding needles or intraperitoneally. The dosage and timing of Tam administration were selected to achieve an appropriate labeling density, as follows: oral administration at 500 mg/kg for either 1, 2, or 4 days, or a single intraperitoneal injection at 200 or 300 mg/kg.

| Immunohistochemistry
Mice were transcardially perfused with 5-10 mL of PBS, followed by 40 mL of 4% paraformaldehyde (PFA) in PBS. The brain was removed from the skull, postfixed overnight in 4% PFA at 4 C, and coronally sliced on a vibratome (VT1000 S; Leica Biosystems) at a thickness of 50 μm. The slices containing the dorsal hippocampus were analyzed in this study. For immunofluorescence labeling, sections were preincubated in PBS with 0.1% Triton X-100 and 2% normal goat serum (NGS) for 30 min, followed by overnight incubation with primary antibodies in 0.1% Triton X-PBS with 2% NGS at room temperature. After thorough washing with 0.1% Triton X-PBS, the sections were incubated with the appropriate secondary antibodies and Hoechst 33342

| Confocal microscopy
Sections were imaged using an A1 confocal laser scanning microscope (LSM) (Nikon) or an FV1000 confocal LSM (Olympus). The following objective lenses were used: 20Â (0.75 Numerical Aperture (NA), CFI Plan Apo VC 20Â; Nikon), 100Â silicon-immersion (1.35 NA, CFI SR HP Plan Apo Lambda S 100Â C Sil; Nikon), 60Â oil-immersion (1.40 NA, CFI Apo Lambda S 60Â Oil; Nikon), or 30Â silicon-immersion (1.05 NA, UPLSAPO30 Â S; Olympus) lenses. The pinhole size was set to 1 Airy unit, and Z-stack images were obtained to cover the sections' entire thickness using a step size of about one-third of the Z-axis resolution.

| Quantification of confocal images
Acquired confocal images were analyzed using the ImageJ (National Institute of Health) or NIS-Elements AR (Nikon) software. Putative pyknotic nuclei were identified based on three criteria: (1) a diameter <6 μm, (2) brighter signal than that of GC nuclei, and (3) lack of intranuclear high-contrast objects. The third criterion eliminated the contaminated microglia, which have small nuclei, comparable to pyknotic ones. The condensed chromosomes of dividing cells could be discriminated from pyknotic nuclei based on their variable thickness.
The size and position of the engulfed structures were measured by determining the border of the microglial cytoplasm surrounding the cell corpse within a single Z-section containing the largest crosssection of the microglial cytoplasm. Although the defined area is larger than the true area of the engulfed object, this measurement is practical as it eliminates possible artifacts induced by the subjective tracing of the phagosome-cytoplasm boundary. The length of microglial processes from the base to the centroid of engulfed structures was measured along the curved paths of microglial processes in 3D. The number of processes of GFP-positive microglia was counted on the 3D-reconstructed images. Processes that projected directly from the microglial cell body were counted as primary processes. Nuclei were judged to be positive for EdU based on the overlap with the Hoechst signal. The organelles were judged to be Lamp-1-or CD68/Lamp-4-positive when the staining was present along their entire perimeters. Cells were counted manually or using the Cellpose software (Stringer et al., 2020).

| Surgical procedure for intravital imaging
An imaging window was surgically implanted over the right dorsal hippocampus for in vivo two-photon imaging, via modification of previous methods (N. B. Danielson et al., 2017;N. B. B. Danielson et al., 2016;Gonçalves et al., 2016;Pilz et al., 2016). The imaging window consisted of a circular glass with a diameter of 3 mm and a thickness of 0.15 mm (3φ No.1; Matsunami Glass) that was adhered to the bottom of a custom-made cylindrical stainless tube (outer/inner diameter, 3 or 3.2 mm/2.8 mm; and height, 2.0 or 2.1 mm; Morishita) using UV-curing adhesive (NOA61 or NOA81; Thorlabs) or epoxy resin (EPO-TEK 353ND; EPOXY TECHNOLOGY). After the intraperitoneal injection of ketamine-xylazine (ketamine, 100 mg/kg; Daiichi Sankyo Company; and xylazine, 10 mg/kg; Bayer), for anesthesia, and mannitol, for edema reduction (1 g/kg; Sigma-Aldrich #M9546), the scalp was removed and an aluminum chamber plate (CP-1; Narishige) was fixed to the skull using dental resin cement (Super-Bond C&B; SUN MEDICAL), for stereotaxic fixation with the head-holding device (MAG-2; Narishige). A circular craniotomy with a diameter of 3.0 or 3.2 mm was performed using a fine-tipped dental drill (ST1 RA (CA) 005; GC Corporation) with the aid of a 3-mm dermal punch (Dermapunch; Maruho). For imaging of the hippocampal DG, the dura was removed and the cortex and the underlying fibers of the external capsule were gently aspirated by a 23 G blunt needle (NIPRO). Next, removal of the CA1 tissue was performed. Specifically, using a 25 or 27 G blunt needle (NIPRO), the alveus of the hippocampus was aspirated, and the CA1 parenchyma was carefully aspirated until the stratum lacunosum-moleculare (LMol) was exposed. The appearance of a loose-fibrous LMol was distinguishable from the above structure under the stereo microscope (S8APO or S9D; Leica Microsystems).
The bleeding was carefully controlled by continuous saline perfusion.
After the exposure of the LMol, the imaging window was gently inserted on the LMol until the bottom of the circular glass was attached to the brain surface. The imaging window was fixed to the surrounding skull with dental resin cement (Super-Bond C&B). The mice survived without major complications for several months after the surgery. In vivo volume imaging data or time-lapse volume imaging data were processed for motion correction using the ImageJ plugins StackReg/ MultiStackReg or Correct 3D Drift, respectively. The time-lapse images were also processed to correct the photobleaching using a plugin called Bleach Correction. Throughout the imaging period, mice were head-fixed onto the head-holding device (MAG-2) and were kept awake or lightly sedated with either 0.5%-1% isoflurane (Pfizer) or a low dose of ketamine-xylazine. The body temperature was maintained in a physiological range using a heating pad (HEATINGPAD-1 and Hot-1; ALA Scientific). During prolonged imaging, water was administered orally to prevent dehydration.

| Quantification of two-photon in vivo images
Two-photon images were mainly analyzed using the ImageJ software.
The temporal changes in size or intensity were analyzed by the MATLAB (MathWorks) software. Microglial morphology was measured based on GFP signals.
Microglial process motility was analyzed by recording the coordinates of the randomly selected process tips in 2-9 consecutive time points. To compensate for the image drift, immobile structures, such as tdTomato-positive neurons, were used as a reference. The motility was calculated from the cumulative distance of the process tip between two frames. The motility of the cell body was calculated from the centroid positions of a randomly selected cell body. The cell body movement was generally smooth and unidirectional.
Therefore, the measurement of the centroid positions at two time points set a few hours apart provided a reliable estimate of the actual translocation. To measure the size of a fluorescent object, such as a newly formed phagocytic cup, a microglial cell body, or a newborn neuron, we manually defined the outline of the structure in the image plane in which the object had the largest profile along the Z-axis. From the segmented area, the centroid position, area size, and the major and minor axes of ellipses fitted to the area were calculated. The length of microglial processes or the number of primary processes in 3D was measured as described in the histological analyses.
The morphology of the engulfed structures/phagosomes was analyzed at each time point using in-house MATLAB programs. The

| Statistics
Continuous variables were presented as the median [first quartile, third quartile]. Statistical analysis was performed using R (Version 4.0.0; R Foundation for Statistical Computing). All data were analyzed using non-parametric tests, as they did not always show normality. Continuous variables were compared using the Mann-Whitney U test for two groups and the Kruskal-Wallis rank-sum test with post-hoc comparisons using the Mann-Whitney U test and Bonferroni corrections for more than two groups. Finally, the correlations of variables were analyzed using Spearman's rank correlation coefficient. All analyses were two-sided, and the significance level was set at P < 0.05.

| The ramified processes of microglia in the SGZ engulf apoptotic newborn cells
First, we analyzed the fixed sections of brains collected from CX3CR1 +/ GFP mice, to evaluate the relationship between apoptotic neural progenitor cells (NPCs) and microglia in the DG of young animals (4-week-old mice). Hoechst 33342 staining revealed pyknotic nuclei, which are a characteristic feature of apoptosis, that were sparsely distributed in both the SGZ and granule cell layer (GCL). GFP-positive microglial processes engulfed these pyknotic nuclei, which were frequently surrounded by a cytoplasm that was immunopositive for cleaved

| In vivo imaging of both microglia and newborn neurons in the DG
To record the entire temporal sequence of phagocytic events, we performed time-lapse in vivo two-photon imaging of microglia and newborn neurons in the mouse DG. To visualize the interactions between microglia and newborn neurons, genetically manipulated mice expressing fluorescent proteins with different colors in the two cell types are necessary.
We generated CX3CR1 +/GFP ; Nes-CreER T2+/À ; Rosa-CAG-LSL-tdTomato +/À mice, which express GFP in microglia and tdTomato in newborn neurons (Imayoshi et al., 2006;Jung et al., 2000;Madisen et al., 2010). We detected a substantial number of tdTomato-positive mature GCs and astrocytes before tamoxifen (Tam) Figure S3D). This window implantation protocol did not induce apparent neuronal death in the DG immediately after surgery and preserved fine neuronal structures, such as dendritic spines, in both the acute and chronic phases ( Figure S4A). By applying the established in vivo imaging technique to CX3CR1 +/GFP ; Nes-CreER T2+/À ; Rosa-CAG-LSL-tdTomato +/À mice, we achieved the simultaneous visualization of GFP-positive microglia and tdTomato-positive newborn neurons in the DG ( Figure S3E). This method was applicable to the chronic in vivo imaging of newly formed GCs after their integration into the hippocampal neural circuits ( Figure S4B, C). These results indicate that this imaging method has considerable potential for the in vivo imaging of physiological events in the DG, with minimal influence on microglial morphology and dynamics.

| In vivo imaging of homeostatic cell clearance by ramified microglia in the SGZ
We performed in vivo two-photon time-lapse imaging of microglia and newborn neurons in the DG of 5-or 13-week-old CX3CR1 +/GFP ; Nes-CreER T2+/À ; Rosa-CAG-LSL-tdTomato +/À mice. Adult neurogenesis in the dentate can be studied from postnatal 5-13 weeks without contaminating early neurogenesis outside the subgranular zone or age-related decline of adult neurogenesis (Kuipers et al., 2015;Nicola et al., 2015). Therefore, in vivo imaging was performed at either the early (postnatal 5 weeks) or late (postnatal 13 weeks) time points to confirm the stability of both neurogenesis and cell clearance within this time window. To induce tdTomato expression, Tam was administered 6-10 days before the implantation of the imaging window.
Imaging sessions started immediately after recovery from the implan-  2b and 1a, b). The phagosomes formed at the process tips underwent intermittent retrograde transport toward the cell bodies, with simultaneous shrinkage of the processes. The completion of phagosome transport, which was associated with a gradual decrease in the phagosomal contents, required 3-6 h. This retrograde translocation explained why newly enwrapped cells identified using in vivo imaging were more distantly located from the cell body than the average position of the phagosomal population detected in the histological sections (in vivo imaging: n = 77 processes, six animals; histology: n = 120 processes, three samples, three animals) (Figure 2d). To eliminate the possibility that the implantation of the imaging window induced artificial microglial behavior, we performed in vivo imaging 4 weeks after window implantation and compared the morphological features and dynamic behavior of the microglial processes that formed phagocytic pouches ( Figure S5). There was no discernible difference between the two sets of imaging data, suggesting that microglial activation in the acute phase of post-surgical recovery was minimal and did not affect phagocytic behavior.
3.4 | Effective surveillance, engulfment, and phagolysosome formation in ramified microglial processes The in vivo visualization of microglial phagocytosis in the SGZ provides a unique opportunity to characterize the cellular machinery implemented in ramified microglia to achieve effective cell clearance.
Our analyses of time-lapse images identified three essential properties of the phagocytosis that was carried out by ramified microglia: (1) effective surveillance of dying cells by motile microglial processes; (2) rapid conversion of the process tips to phagocytic pouches; and Next, we tested the speed of conversion of the process tip to phagocytic pouches. When ramified microglia respond to tissue injury induced by a focused laser beam, reorganization of microglial processes toward the damaged site occurs within 1-3 h (Davalos et al., 2005). This gradual and overall reorganization of microglial morphology is mediated by ATP locally released from the injured site. In contrast, the in vivo imaging of the interaction between apoptotic neurons and microglia in the SGZ revealed the absence of slow global changes in microglial polarity associated with phagocytic events. To In vivo imaging (left) was used to identify and measure newly formed phagosomes (n = 77 processes; six animals). In fixed brain sections (right), all phagosomes, regardless of their previous history, were measured (n = 120 processes; three samples; three animals). Scale bars = 10 μm. ****P < 0.0001.
illustrate this property, we quantified the motility of microglial processes just before engulfment, which revealed high motility indexes, comparable to those of control processes (n = 30 processes; six cells; three animals each for targeting and non-targeting processes) processes; six animals) (Figure 3c, g). We concluded that the formation of phagocytic pouches is a rapid process without slow global changes in the microglial overall morphology and motility.

| Compartmentalized lysosomal digestion and sustained surveillance of microglial processes
Two independent factors may affect the cell-clearance capacity of ramified microglia. The first factor is the speed of detection and digestion of individual apoptotic neurons. As described in the previous section, the simultaneous imaging of microglial processes and target dying neurons indicated that a single microglial process is designed to reduce the time required for the lysosomal digestion of neurons after their entry into the apoptotic pathway. The second factor is the ability of a single microglial cell to perform the digestion of multiple apoptotic cells simultaneously. We speculated that multiple microglial processes could function as independent parallel compartments for cell clearance. To clarify this issue, we performed the following analysis. (d) Size of tdTomato-positive cells before engulfment. Both the area (left) and the length of the major-minor axes after ellipsoid fitting (right) were plotted (n = 25 events; eight animals). (e) Size of newly formed phagocytic structures. Both the area (left) and the length of the major-minor axes after ellipsoid fitting (right) were plotted (n = 81 events; six animals). (f) Similar motility between the processes visualized right before phagocytosis and those not destined for phagocytosis (left). The microglial cell bodies did not exhibit motility, regardless of their engagement in phagocytosis (right) (n = 30 processes; six cells; three animals each for targeting and non-targeting processes). (g) Frequency distribution of the time from the initial microglial contact to the completion of engulfment (n = 54 processes; six animals). Scale bars = 10 μm. n.s.; not statistically significant, *P < 0.05, ****P < 0.0001. objects; three animals) (Figure 5b). We estimated that the reduction in phagosomal volume associated with the retrograde translocation within microglial processes was 94%. This substantial volume reduction indicated that the phagolysosomal degradation of engulfed dying cells mainly occurred within the compartment of a single microglial process. This notion was consistent with the histological finding (e-g) Proportion of CD68/Lamp-4-, Lamp1-, and cathepsin D-positive phagosomes in microglial processes (n = 163 objects, three samples, three animals for CD68/Lamp-4; n = 142 objects, three samples, three animals for Lamp-1; n = 137 objects, three samples, three animals for cathepsin D). (H) Lysosomal density estimated based on the intensity of the anti-CD68/Lamp-4 antibody staining in non-phagocytic processes and cell bodies (n = 13 processes and 10 cell bodies; 5 samples; three animals). Scale bars = 10 μm. **P < 0.01. that most phagocytosed pyknotic nuclei did not exist in cell bodies, but were confined to processes (Figure 1d). The confinement of the lysosomal degradation of engulfed neurons to a single process may be beneficial for the retention of the ability to digest another engulfed cell corpse by the remaining processes. The histological analysis of CD68/Lamp-4-positive lysosomes indicated the moderate enrichment of lysosomal components in the process that was engaged in phagocytosis, together with the substantial retention of lysosomes in cell F I G U R E 5 Legend on next page. bodies and non-phagocytic processes (n = 11-12 processes or cells; six samples; three animals) (Figure 5c, d). The pattern of lysosomal distribution was consistent with the notion that the digestive capacities of the non-phagocytic processes are sufficient for initiating their own engulfment and digestion.
The preserved surveillance activity of the non-targeting processes is another crucial property for the functioning as parallel compartments for cell clearance. In vivo imaging confirmed that the motility of the processes that were not involved in phagocytosis did not change after the phagocytic events (n = 25 processes; five cells; two animals) ( Figure 5e). In addition, phagocytic events in neighboring microglia did not affect the frequency of the microglial process contacting nearby neurons in the SGZ (n = 14 cells; six animals) (Figure 5f). The sus- 3.6 | Phagocytosis of dying neurons by ramified microglia in multiple brain regions The continuous generation of apoptotic neurons is highly restricted to specific brain regions in the adult brain. The phagocytic properties of ramified microglia in the SGZ may be unique to the hippocampal DG. Alternatively, the mechanism may be broadly implemented in ramified microglia resident in other brain regions and at different developmental stages. To clarify this point, we performed a histological examination of multiple brain regions in which apoptotic cells were present (VanRyzin et al., 2019;Wang et al., 2021).
In the visual cortex of P7 mice, where massive apoptosis of chandelier cells was detected (Figure 6a), the ramified microglia exhibited phagocytosis of CC3-positive cells at the tip of their processes, where CD68/ Lamp-4 immunoreactivity was colocalized (Figure 6b). Parallel phagocytosis was also detected in the P7 mouse visual cortex, suggesting a phagocytic mechanism similar to that observed in the adult SGZ (Figure 6c). The immunohistochemical studies of the amygdala in P7 mice and of the sensory cortex in 21-month-old mice further confirmed the presence of phagocytosed pyknotic nuclei at the process tips, where the lysosomal marker CD68/Lamp-4 was colocalized (Figure 6d, e). Thus, we concluded that ramified microglia utilize individual motile processes as a device for effective cell clearance in broad physiological contexts.

| DISCUSSION
In this study, we sought to clarify the mechanism underlying effective cell clearance by ramified microglia in the adult neurogenic niche of the SGZ. We optimized the method of high-resolution in vivo two-photon imaging of the DG, and, in combination with histological analyses, identified the unique phagocytic events of microglia in the adult SGZ. Microglia effectively detected and engulfed the entire apoptotic neurons at the tips of their highly motile processes. The primary phagosomes rapidly matured into phagolysosomes inside the processes, which was accompanied by the progressive digestion of phagosomal contents.
Multiple microglial processes functioned as independent compartments for cell clearance. After initiating the first phagocytic event, other microglial processes sustained environmental surveillance and targeted the F I G U R E 5 Compartmentalized digestion by phagolysosomes in microglial processes. (a) Trajectories of individual phagosomes within microglial processes after the initiation of engulfment. After initial pausing for 50 min, the phagosomes engaged in retrograde movement toward the cell body (n = 6 objects; three animals). Inset images show the same phagosome at t = 109, 154, and 163 min to illustrate the retrograde movement. (b) Correlation between the location of individual phagosomes and their volume. Different symbols represent the traces of identical phagosomes during their transport (n = 6 objects; three animals). (c) Localization of CD68/Lamp-4-positive lysosomal components in individual processes. Both the process engaged in phagocytosis (white arrowhead) and other processes (red arrowhead) contained similar levels of CD68/ Lamp-4. (d) Lysosomal densities in phagocytic and non-phagocytic processes and in cell bodies (n = 11-12 processes or cells; six samples; three animals). (e) Preserved motility of non-phagocytic processes before and after the events of phagocytosis by other processes that extended from the same cell body (n = 25 processes; five cells; two animals). (f) Preserved microglial contact frequency with a preselected newborn neuron before or after the initiation of phagocytosis by a microglial cell in the vicinity (within 40 μm) of the target neuron (n = 14 cells; six animals). (g) Time-lapse in vivo imaging of a GFP-positive microglial cell with multiple phagocytic events (filled and open arrowheads) and continuous surveillance of the tissue environment (arrows). (h) Histological identification of microglia with multiple phagocytic structures that were immunopositive for CD68/Lamp-4 (arrowheads). Scale bars = 10 μm. n.s.; not statistically significant, *P < 0.05, **P < 0.01. second apoptotic neuron. Based on these observations, we postulate that ramified microglia in the SGZ utilize individual motile processes to detect stochastic cell death events and execute parallel phagocytoses effectively. The homeostatic cell clearance afforded by ramified microglia will help us understand how microglia maintain and modulate the neural circuit function via the regulation of adult neurogenesis.

| Methodological considerations
For the proper interpretation of our study results, its technical limitations should be acknowledged. We used CX3CR1 +/GFP ; Nes-CreER T2+/À ; Rosa-CAG-LSL-tdTomato +/À mice (Imayoshi et al., 2006;Jung et al., 2000;Madisen et al., 2010)  The in vivo imaging of ramified microglia in deep brain structures is challenging, and possible microglial activation caused by tissue damage should be carefully evaluated (Xu et al., 2007). In this study, we adopted and optimized the protocol of Danielson et al. (N. B. B. Danielson et al., 2016), which preserves the LMol of the CA1 with a thickness of approximately 100 μm. After surgery, microglia in the uppermost LMol changed their shape, whereas microglia in the intact DG preserved their ramified morphology, similar to those imaged in fixed sections of the intact brain. A stricter evaluation of microglial states using single-cell transcriptomics will be required to rule out possible alterations in phagocytic functions triggered by the window implantation surgery (Hammond et al., 2019). However, the in vivo two-photon imaging of the SGZ microglia performed 4 weeks after window implantation indicated an absence of discernible changes in motile behavior and phagocytic activity between the imaging carried out in the acute postsurgical period and that performed in the chronic phase ( Figure S5). Therefore, we concluded that the in vivo imaging data analyzed in this study faithfully report the dynamic interaction between apoptotic neurons and microglia in the process of homeostatic cell clearance.

| Multiple modes of microglial phagocytosis in vivo
Accumulating evidence from both histological examinations and in vivo live imaging of microglia indicates their active role in the elimination of damaged cells and tissue debris (Butler et al., 2021;Márquez-Ropero et al., 2020). The microglial response to damaged cells has been well documented in diseases and after tissue injury (Das & Chinnathambi, 2020;Nolte et al., 1996;Stence et al., 2001). In these pathological conditions, microglia are activated by inflammatory cytokines and substrates released from damaged cells, change to an amoeboid shape, promote movement toward damaged cells, and initiate phagocytosis. This type of phagocytosis by activated microglia with an amoeboid morphology is similar to the behavior of macrophages in response to peripheral tissue damage (Arandjelovic & Ravichandran, 2015). Traditionally, amoeboid microglia are thought to be highly efficient in eliminating cell corpses and tissue debris. However, recent experimental evidence has questioned the efficacy of phagocytosis by amoeboid microglia (Paolicelli et al., 2022). In addition to this type of phagocytosis performed by amoeboid microglia, the experimental induction of spatially confined tissue damage triggers a different type of microglial response; in this case, microglia undergo polarization of their structure and extend their processes to the injury site, without moving their cell body (Davalos et al., 2005;Haynes et al., 2006). This response was shown to be mediated by locally released ATP and microglial P2Y 12 receptors. The converging microglial processes form a spherical containment around the local injury; however, the initiation of phagocytic activity at the site of conversion has not been investigated. Nevertheless, this microglial response seems to be distinct from the migrating amoeboid microglia that are fully reactivated by tissue damage.
In this study, we provided evidence indicating the existence of a third type of microglial response that initiates effective phagocytosis without movement of the cell body. Previous histological studies identified the presence of a phagocytic pouch at the distal microglial processes in the hippocampal SGZ (Luo et al., 2016;Sierra et al., 2010).
However, the snapshots of microglia engaged in cell clearance were insufficient to provide a comprehensive picture of effective homeostatic cell clearance by ramified microglia. Our systematic approach using in vivo two-photon time-lapse imaging successfully illustrated the successive steps of detection, engulfment, and digestion of dying cells, together with the quantitative analysis of the cell clearance efficiency by a single microglial cell.
The behavior of motile microglial processes just before engulfment indicates the immediate conversion of motile processes to phagocytic pouches. This behavior contrasts with that of microglia in response to the ATP released from the confined tissue damage (Davalos et al., 2005;Haynes et al., 2006). It is reasonable to assume that signals exposed on the surface of damaged cells, such as phosphatidylserine (Darland-Ransom et al., 2008;Fadok et al., 1992;Segawa et al., 2014) or calreticulin (Gardai et al., 2005), trigger phagocytosis via microglial receptors, whereas diffusible factors, such as nucleotides and the fractalkine CX3CL1, may not play a major role (Koizumi et al., 2007;Truman et al., 2008). We expect that, in the future, new imaging technologies will directly identify the cell surface signaling molecules that are involved in the detection of damaged cells by microglia in the SGZ.
Studies of both nervous system development and of the period after seizure-induced enhancement of adult neurogenesis postulated the engulfment and digestion of viable cells by microglia (Cunningham et al., 2013;Luo et al., 2016). The phagocytosis of live cells, which is termed "phagoptosis" may also be present in the unperturbed SGZ (Brown & Neher, 2014). Previous histological studies of the unperturbed adult SGZ reported the microglial engulfment of apoptotic NPCs (Sierra et al., 2010). However, a recent study revealed the phagoptosis of newly generated SGZ neurons after status epilepticus (Luo et al., 2016). Our in vivo two-photon time-lapse imaging elucidated the morphological characteristics of tdTomato-positive newborn neurons right before engulfment. These neurons imaged in vivo exhibited a cell size that was close to that of the intact NPCs, suggesting that they were in the early stages of apoptosis or in the stage that precedes entrance into the cell death cascade. About 25% of the engulfed cells were CC3 positive in our study, whereas the remaining engulfed cells were smaller in size and positive for CD68/Lamp-4.
These two populations may represent the early phase of phagocytosis of CC3-positive apoptotic cells and the later phase of cell digestion, respectively. The scarcity of phagosomes containing CC3-negative cells with a size comparable to that of viable neurons suggests that apoptosis preceded phagocytosis in our study, in contrast to the presence of a substantial fraction of non-apoptotic engulfed neurons in previous studies of phagoptosis (Cunningham et al., 2013;Luo et al., 2016). 4.3 | Phagocytic and tissue searching activities were independently regulated in single microglial processes Ramified microglia in the SGZ continue surveillance of the tissue environment even during the digestion of a cell corpse in one of their processes. The surveillance activity of microglial processes is regulated by the Gi/Rho GTPase pathway and by the TWIK-related Halothaneinhibited K + channel (THIK-1) (Madry et al., 2018;Merlini et al., 2021). In turn, directional process extension is induced by P2Y 12 receptor activation, possibly through the pathway mediated by G i . Furthermore, the difference in cAMP concentration created by norepinephrine G s -coupled receptors or P2Y 12 G i -coupled receptors regulates the balance between thin filopodia and thick processes (Bernier et al., 2019). The ATP released from damaged cells may activate the P2Y 12 receptor and induce cAMP decrease confined to the single process that are engaged in engulfment. The coexistence of motile filopodia-rich processes and stable processes serving for phagocytosis may be explained by this cAMP-dependent regulation.
We consistently observed the intermittent retrograde motility of phagolysosomes in the microglial process after phagocytosis. However, the mechanism underlying this retrograde movement has not been clarified. In the zebrafish optic tectum, apoptotic neurons are removed by microglial phagocytosis, which is initiated by forming phagocytic cups without translocation of the cell body, similar to the phagocytic behavior of ramified microglia observed in the mouse SGZ (Möller et al., 2022). The experimental perturbation of the microtubule system disrupted the pattern of phagocytosis in zebrafish microglia.
Furthermore, the location of the centrosome inside microglial processes determines the location of phagosome formation, possibly through regulated vesicular trafficking. These observations in zebrafish suggest the primary role of microtubule polarity orientation in phagosome formation. We speculate that a similar microtubuledependent mechanism of phagosome formation exists in the SGZ microglia, and that microtubule minus-end directed motors, such as dynein, drive the retrograde phagolysosomal translocation. In summary, ramified microglia utilize individual motile processes to detect stochastic cell death events in the neurogenic niche and execute parallel phagocytoses, which support the homeostasis of adult hippocampal neurogenesis.

CONFLICT OF INTEREST STATEMENT
The authors declare no financial and non-financial competing interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study will be made available upon request to the corresponding author.