Microglia Maintain Homeostatic Conditions in the Developing Rostral Migratory Stream

Visual Abstract

neuroblast migration in the RMS. GFP-labeled microglia in CX3CR-1 GFP/1 mice assemble primarily along the Introduction Microglia are brain-resident immune cells that maintain homeostatic conditions necessary for normal neurodevelopment and orchestrate responses to environmental insults (Pierre et al., 2017;Tay et al., 2017;Hammond et al., 2018;Lenz and Nelson, 2018;Li and Barres, 2018;Ferro et al., 2021). The importance of microglia in normal neurodevelopment was recently underscored by the case of a child born without microglia because of a homozygous mutation in colony stimulating factor 1 receptor (CSF1R), which is necessary to maintain a microglia population (Nandi et al., 2012). This child presented with agenesis of the corpus callosum, ventriculomegaly, and heterotopias (Oosterhof et al., 2019). These mirrored the findings of a Csf1r knock-out mouse model, which also demonstrated ventricular enlargement, disrupted olfactory bulb (OB) architecture and olfactory deficits (Erblich et al., 2011). Microglia functions may be of special relevance to the olfactory system, which is unique in both its vulnerability to environmental insults as well as an extended period of neurogenesis and neuronal migration.
The rostral migratory stream (RMS) is an important neuroblast migratory corridor in the olfactory system. In mice, neuroblasts are continuously generated in the subventricular zone (SVZ) and migrate in chains down the RMS to the OB (Lois and Alvarez-Buylla, 1994;Ponti et al., 2013;Bordiuk et al., 2014). Upon reaching the OB, neuroblasts differentiate into periglomerular and granule cell interneurons and integrate into preexisting circuits (Lledo and Saghatelyan, 2005;. Neuronal migration via the RMS is also active in the human fetus and continues for several months postnatally (Sanai et al., 2011;Wang et al., 2011). The blood vessel scaffold and dense astrocyte tube that wraps the RMS and supports migrating neuroblasts (Bovetti et al., 2007;Snapyan et al., 2009;Grade et al., 2013;Gengatharan et al., 2016;Fujioka et al., 2017;Todd et al., 2017) does not assemble until several weeks after birth (Law et al., 1999;Peretto et al., 2005;Nie et al., 2010;Bozoyan et al., 2012), leaving a functional void that may be filled by microglia.
Sustained neurogenesis and neuroblast migration in the RMS may require "quality assurance" measures via microglia phagocytosis. Microglia phagocytosis is essential during critical periods of development for addressing over-exuberant neurogenesis (Cunningham et al., 2013;Barger et al., 2019) and synaptogenesis (Tremblay et al., 2010;Paolicelli et al., 2011;Gunner et al., 2019;Mallya et al., 2019;Scott-Hewitt et al., 2020). In the RMS of adult animals apoptotic cells accumulated following microglia depletion in the SVZ (Ribeiro  and selective depletion of MERTK in microglia (Fourgeaud et al., 2016), a receptor tyrosine kinase involved in initiating phagocytosis (Rosin et al., 2021), implicating a role for microglia phagocytosis in regulating the neuroblast population. Microglial activities may be especially crucial during the early postnatal period, when there is prolific migration of neuroblasts before the formation of the blood vessel and astrocyte scaffold.
During early postnatal periods a specialized microglia subtype may mediate neurodevelopment. Singlecell sequencing studies identified a microglia subtype with an ameboid morphology and phagocytic markers, including lysosomal-associated membrane protein 1 (Lamp1), CD68, and CLEC7A, in developing white matter tracts during the first postnatal week (Hagemeyer et al., 2017;Hammond et al., 2019;Li et al., 2019). These cells are involved in phagocytosis in developing white matter tracts (Li et al., 2019), but also express the trophic factors insulin-like growth factor 1 (IGF-1) and osteopontin (SPP1; Hagemeyer et al., 2017;Hammond et al., 2019;Li et al., 2019) and are implicated in supporting myelination (Hagemeyer et al., 2017;Wlodarczyk et al., 2017) and axons (Ueno et al., 2013). Intriguingly, postnatal microglia expressing the phagocytic marker CLEC7A are also seen in the RMS (Li et al., 2019), suggesting that a similar microglia subset may be involved in phagocytosis or trophic support in the developing RMS.
We investigated whether microglia regulate the population of migrating neuroblasts in the RMS during the early postnatal period. We examined the distribution and morphology of RMS microglia and employed depletion strategies to gain insights into their function. We found that ameboid microglia distribute along the borders of the early postnatal RMS before the formation of the blood vessel scaffold and astrocyte tube. Microglia depletion did not impact the migratory capacity of neuroblasts but did compromise the homeostasis of the RMS, causing an accumulation of neuroblasts and apoptotic cells that broadened the domain of the early postnatal RMS. Microglia therefore appear critical for maintaining the homeostasis of the RMS neuroblast population and migratory environment.

Animals
Experiments investigating microglia morphology and distribution were conducted with B6.129P2(Cg)-Cx3cr1 tm1Litt /J (The Jackson Laboratory stock #005582) mice, which express EGFP in microglia under control of the Cx3cr1 locus (Jung et al., 2000). The mice were maintained as homozygous colonies and crossed with C57BL/6J mice to produce heterozygous CX3CR1 GFP/1 mice for experiments. To examine the interactions between microglia and neuroblasts, CX3CR-1 GFP/GFP mice were crossed with C57BL/6J-Tg(Dcx-DsRed)14Qlu/J (a gift from Angeliki Louvi), which express the red fluorescent protein variant DsRed under the control of the doublecortin (Dcx) locus (X. Wang et al., 2007). This allowed the generation of CX3CR-1 GFP/1 ; DCX DsRed/1 transgenic mice.
All experiments randomly included both male and female mice, although sex comparisons were not pursued. Mice were housed with a 12 h light/dark cycle with access to standard food and water ad libitum. All animal care and use were performed in accordance with the Institutional Animal Care and Use Committee (IACUC) Yale animal care committee's regulations.

Electroporation
To label individual migrating neuroblasts, stem cells lining the lateral wall of the lateral ventricle were electroporated with a pCAG-tdTomato at P0 and killed at P2, P4, P7, P14. The pCAG-tdTomato plasmid used for electroporation was a gift from Angelique Bordey (Addgene plasmid #83029; http://n2t.net/addgene:83029), which was maintained in Escherichia coli strains stored at À80°C as 10% glycerol:LB stock. P0 mice were anesthetized with hypothermia before injection with tdTomato plasmid. The plasmid solution was injected into one of the brain lateral ventricles using a Picospritzer (General Valve Corporation). Electroporations were made by delivering five pulses of 35 V using a pair of gold tweezers (Genepaddles-542, Harvard Apparatus) connected to an ECM 830 electroporator (BTX Harvard Apparatus).

Liposome injection
Microglia were depleted in perinatal mice by the injection of clodronate-filled liposomes (Clodrosome) at postnatal day (P0). Injection of clodronate-filled liposomes and their subsequent phagocytosis by microglia induces cell death, reflected by a dramatic decrease in the number of microglia 3 d later (Cunningham et al., 2013). A total of 2 ml of Clodrosomes or the control PBS-filled liposomes (Encapsome) were injected into the cerebral lateral ventricles of CX3CR-1 GFP/1 mice bilaterally using a Picospritzer (General Valve Corporation; Table 1). The effect of microglia knock-down was evaluated 3 d later by immunohistochemistry and confocal microscopy. To demonstrate the phagocytic capacity of microglia throughout the RMS, fluorescent liposomes containing the fluorescent dye Dil (Fluoroliposome) were injected into the cerebral lateral ventricles of P1 CX3CR-1 GFP/1 mice and killed at P4 (Table 1). P0 postnatal mice were anesthetized with hypothermia before injection.

BrdU injections
To analyze the density of neuroblasts migrating in the RMS, CX3CR-1 GFP/1 ;DCX DsRed/1 mice were injected at P0 with 25 mg/kg of the thymidine analog BrdU (BD PharMingen) to label newborn neuroblasts and killed at P3. To investigate the number of neuroblasts that reach the glomerular layer following microglia depletion, CX3CR-1 GFP/1 mice were injected with 50 mg/kg of BrdU at P1 and killed at P4.

Tamoxifen injections
To deplete microglia in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 mice, 30 mg/g of tamoxifen (TMX; Sigma T5648) was injected at P0. TMX induces Cre-mediated recombination at loxP sites and removal of the terminal STOP codon sequence, enabling expression of the DTA fragment (Fig. 1). Mice were either killed at P3 or injected with TMX every subsequent 3 d to maintain microglia depletion before P14. Cx3cr1 CreER/1 ; ROSA26 1/1 littermates served as controls. Tissue processing For immunofluorescence, mice were deeply anesthetized with an overdose of Euthasol (Virbac) and perfused transcardially in 0.1 M PBS, pH 7.4 with 1 unit/ml heparin, followed by 4% paraformaldehyde (PFA) (JT Baker) in PBS. Animals were decapitated and the brains carefully removed from the skull. The brains were postfixed in 4% PFA at 4°C overnight. Following PBS washes, brains were transferred to a 30% sucrose and kept at 4°C for 3-4 d for cryoprotection. Brains were then embedded in OCT (optimal cutting temperature) compound (Thermo Fisher Scientific); 25-mm sections were serially collected in both A, Sagittal section showing neuroblasts (red) in a P7 DCX DsRed/1 mouse. There is active migration of neuroblasts from the subventricular zone (SVZ) to the olfactory bulb (OB) via the RMS in the first postnatal week. Scale bar = 250 mm. B, Sagittal section showing microglia (green) in the SVZ, RMS, and OB in a P4 CX3CR-1 GFP/1 mouse. To identify individual migrating neuroblasts (red), the SVZ was electroporated with a tdTomato plasmid at P0 and killed at P4. Microglia line the borders of the RMS and associate with migrating neuroblasts within the core. Scale bar = 50 mm. C, Ultrathin sections showing GFP-immunopositive microglia soma and processes at the electron microscope level in CX3CR-1 GFP/1 mice at P8 and P21. Microglia are pseudo-colored in green. C 1 , C 2 , Ameboid microglia at P4 contact migrating neuroblasts. C 1 , A microglial cell engaged in phagocytosis, as shown by the engulfment of material within the extended phagocytic cup. C 3 , C 4 , Microglia processes contact migrating neuroblasts at P21. C 3 , A microglia between a cluster of migrating neuroblasts. m, microglial cell. n, neuroblast. Scale bar = 2 mm. coronal and sagittal planes using a Reichert Frigocut Cryostat E-2800. Sections were frozen at À20°C until use.
For electron microscopy, the mice were perfused with 4% PFA and 0.75% glutaraldehyde, followed by postfixation in the same fixative for 4 h. Brains were rinsed in PBS overnight and cut on a vibratome (50 mm).

Immunohistochemistry
Sections were thawed at 37°C and treated for antigen retrieval with 0.01 M citrate buffer at pH 6 and 68°C for 30 min. They were incubated in a blocking solution of PBS supplemented with 0.1% Triton X-100 (Sigma-Aldrich) and 0.2% bovine serum albumin (BSA; Sigma-Aldrich) for 1 h at room temperature. Sections were incubated in a mixture of primary antibodies diluted in blocking solution at 4°C overnight, followed by incubation with secondary antibodies diluted in blocking solution at 1:400 for 2 h at room temperature (Table 2). Nuclei were counterstained by incubating the sections with 1 mg/ml DAPI for 10 min (Invitrogen). Sections were washed with PBS after each step. The sections finally mounted with Mowiol 4-88 (Sigma-Aldrich).

Electron microscopy
Tissue was incubated with 2% BSA (Sigma) in PBS for 30 min to block nonspecific binding sites. The tissue was incubated in a chicken polyclonal anti-GFP antibody (Abcam) in blocking buffer for 4 d at room temperature. Tissue was then incubated in biotin-conjugated goat anti-chicken IgY secondary antibodies for 1 h at room temperature. Sections were incubated with the ABC reagent (Vector) for 1 h, followed by a DAB peroxidase reaction. Sections were immediately processed for electron microscopy.
Stained tissue was postfixed with 4% osmium tetroxide for 1 h, dehydrated through graded alcohols, and polymerized in Epon between glass slides and coverslips coated previously with Liquid Release Agent (Electron Microscopy Sciences). Areas of interest were selected based on the DAB labeling and re-embedded in Epon. These re-embedded sections were cut using an ultramicrotome in 70-nm-thick ultrathin sections. These 70-nm sections were examined with a JEOL transmission electron microscope and photographed at primary magnifications of 3000-4000Â.

Fluorescence in situ hybridization
To produce an RNA probe antisense to Igf1 mRNA, forward primer 59-TGAGCTCAGCCAGTTCGTGTGTGGACC GAGGGG-39(containing a BlpI restriction site at its 59 end) and reverse primer 59-TGAAGATCTTAATACGACTCACT ATAGGCGGCTCTTTAGAGGCAGGGACTAGG-39 (containing a BglII restriction site, followed by a T7 promoter sequence, at its 59 end) were used to amplify a 646 nucleotide region of the Igf1 coding sequence (nucleotides 485-1130, spanning the junctions between exons 2 and 3, 3 and 4, and 4 and 5) from cDNA prepared from OB tissue isolated from eight-week-old wild-type (WT) CD-1 mice. Following sequential digestion with BlpI and BglII, the resulting amplicon was cloned into the corresponding restriction sites of plasmid pET-28a such that the Igf1 cassette was flanked by a T7 terminator at its 39 end. Using the resulting construct as template and an Atto 647N RNA Labeling kit (Jena Bioscience), fluorescent RNA probes were generated by in vitro transcription according to the manufacturer's recommendations.

Confocal imaging and processing
All images were taken using a Zeiss LSM800 confocal with either a 10Â or 40Â objective and images processed with Fiji (ImageJ) software. When creating tiled images of entire sagittal sections for illustrative purposes, images were taken with the 10Â objective at 27-mm z-stacks (four Research Article: New Research slices at 6.75-mm step distance). After creating a maximum projection of each channel, a "rolling ball" algorithm was applied to correct for uneven illumination. Cell density analysis was performed on confocal images obtained with the 10Â objective. Z-stacks were 27 mm in depth (four slices at 6.75-mm step distance.) The rostral migratory stream (RMS) was first isolated on a 16bit maximum projection of a 638.9 Â 638.9-mm image of DAPI nuclear staining with a freehand drawing tool. The area of this drawn region of interest (ROI) was measured later to determine the density of stained cells and area ratio of stained processes. The ROI was drawn on the maximum projection of the DAPI channel to prevent bias for incorporating stained cells of interest. A mask of the drawn ROI was then applied to the corresponding channels stained with antibodies against GFP, GFAP, Iba1, BrdU, and CC3 to quantify stained objects within this ROI.
There was increased noise in the GFP and DsRed channels, and thus we applied an additional "despeckle" function, which removes salt-and-pepper noise. Salt-andpepper noise reduction was achieved in the CC3 channel with a "median" filter of pixel radius 1. BrdU staining was punctate, so to blur punctate staining and isolate stained nuclei a "Gaussian blur" function with a pixel s of 2 was applied to the BrdU maximum projection. To segment stained pixels from background we used the Fiji automatic Otsu threshold function, except for segmenting Iba1 and CC3 stained objects, for which we used the Fiji automatic Triangle threshold function. A gray scale attribute filter with an opening function of area minimum of 100 pixels and connectivity of 8 was further applied using the MorphoLibJ integrated library and plugin (Legland et al., 2016) to isolate the cell bodies of GFP-stained and Iba1stained microglia and BrdU-stained neuroblasts. A gray scale attribute filter with an opening function of area minimum of 15 pixels and connectivity of 8 was applied to isolate the cell bodies of DsRed-stained neuroblasts, and of an area minimum of 50 pixels and connectivity of 8 to isolate CC3-stained cells. After making a mask of the images, the density of stained cell bodies was assessed by counting the number of cells with the "Analyze Particles" function and normalizing this number to the area of the drawn ROI for the layer. GFAP staining was quantified by the percent area stained within the drawn ROI.
The density of CC3-stained cells in the RMS descent following 3 and 14 d of microglia depletion was performed on confocal images obtained with the 40Â objective. Zstacks were 18 mm in depth (15 slices at 1.2-mm step distance.) A "rolling ball" algorithm with a radius of 50 pixels was applied to subtract background and salt and pepper noise reduction was achieved with a "median" filter of pixel radius 3. After creating a maximum projection of the CC3 channel, stained pixels were segmented from background using the Fiji automatic Otsu threshold function. To fill in the nucleus we applied an opening filter of a diamond radius 2 using the MorphoLibJ integrated library and plugin (Legland et al., 2016). All segmented CC3stained nuclei were then manually compared with the Zstack of merged DCX, DAPI and CC3 channels to assess whether DCX staining surrounded the isolated CC31 DAPI1 nucleus. Finally, the density of CC31 DCX1 and CC31 DCXÀ stained nuclei were assessed by normalizing the number to the area of the drawn ROI of the RMS from a maximum projection of the DAPI channel. The total number of CC31 stained nuclei was verified using the "Analyze Particles" function.
We employed the ImageJ plugin FracLac to assess the morphology of individual binarized microglial and neuroblast cells (Young and Morrison, 2018). 40Â confocal images were taken of GFP1 microglia and tdTomato1 neuroblasts in the RMS elbow, and maximum projections were made of 24 mm (10 slices at 1.2-mm step distance) to collapse the whole cell and its processes. Maximum projections were obtained for four sections of the RMS elbow per animal, with every cell that had all of its processes contained within the image analyzed. For the GFP channel, a "Despeckle" function was first used to correct for noise, followed by a "remove outliers" function to target bright outliers with a pixel radius of 2 and threshold of 50. The GFP1 cell was isolated from background using the "default" automatic threshold and binarized. The "close" plugin was then used to connect two pixels if they were separated by up to two pixels. To enhance the small features of neuroblasts and remove background in the tdTomato channel, a "difference of Gaussians" was applied to the max projection. The tdTomato1 cell was isolated from background using the "triangle" automatic threshold and binarized before performing fractal analysis.
To perform fractal analysis on an individual cell in the image, a rectangle selection of 100 pixels of both height and width was applied for the ROI of all cells, to ensure that all the cells have the same scale. While fractal shapes are scale-independent, FracLac for ImageJ is dependent on scale, and thus the ROI needs to be consistently sized throughout the data collection (Karperien, 1999;Young and Morrison, 2018). Using the matching maximum projection photomicrograph as a reference, the paintbrush tool was used to isolate the cell of interest by removing adjacent cell processes and connecting fragmented processes (Young and Morrison, 2018). The isolated cell was then converted to an outline using the "outline" function. Fractal analysis data are gathered via box counting, in which a series of grids of decreasing caliber were systematically laid over an image and the number of boxes containing a pixel counted. The fractal dimension (D B ) is the relationship between how a pattern's detail (N) changes with the scale (« ), or resolution, at which the image is considered (Karperien et al., 2013). Fractal analysis was done with the Box Counting in FracLac plugin, and data were collected with four box counting orientations (Num G = 4). Lacunarity (K) is also calculating using the box counting method and is a measure of cell heterogeneity; cells with low lacunarity are homogenous or rotationally invariant (Karperien et al., 2013). Additional cell shape measurements were generated using the "convex hull" (a polygon that encloses all pixels of the binary cell) and "bounding circle" (the minimum circle that can enclose all pixels.) The span ratio is the ratio of the longest length over the longest length of the convex hull (Morrison et al., 2017). Density is the ratio of the area of cell divided by the area of its convex hull. A full list and explanation of FracLac output data are explained in the FracLac for ImageJ manual (Karperien, 1999). The calculations to obtain the fractal dimension (D B ), lacunarity (K), span ratio and density are best explained in the reference guide provided for FracLac for ImageJ (http://rsb.info.nih.gov/ij/plugins/ fraclac/FLHelp/Introduction.htm).
We further used FracLac to evaluate the lacunarity of a max projection of DCX-stained processes in the RMS as a measure of the space occupied by neuroblast processes following 14 d of microglia depletion. Confocal images were obtained with the 40Â objective. Z-stacks were 18 mm in depth (15 slices at 1.2-mm step distance.) After obtaining a maximum projection, DCX-stained processes were segmented from background with a "Otsu" automatic threshold. A mask of the drawn ROI obtained using a freehand drawing tool on the maximum projection of the DAPI channel was applied and DCX-stained processes outside the RMS subtracted from the image. Fractal analysis was done with the Box Counting in FracLac plugin, and data were collected with four box counting orientations (Num G = 4). Lacunarity (K) values are an indicator of the "gappiness" between stained processes, with high lacunarity suggesting greater gaps and variation between stained processes (Karperien et al., 2013).

Statistical analysis
All measures across different sections within an animal from the same RMS region were averaged to yield one sample replicate for statistical analysis. Statistical analysis was performed with GraphPad Prism 9.2 for windows (GraphPad Software Inc). The data were analyzed with a one-way ANOVA followed by Bonferroni's post hoc test or a Student's t test as appropriate. *p ,0.05, **p , 0.01, ***p , 0.001, ****p , 0.0001. Data are shown as the mean 6 SEM (Standard error of the mean). Specific p values for each experiment are shown in the statistical table (Table 3).

Results
Ameboid microglia are present in the developing RMS before the formation of the astrocyte and vascular scaffold There is active migration of neuroblasts from the SVZ to the OB through the RMS during the first postnatal week, as shown by dense staining of DsRed1 neuroblasts in the RMS of DCX DsRed/1 mice (Fig. 1A). During the first postnatal week microglia line the RMS along its lateral borders (Fig. 1B). To examine the interactions between individual microglia and neuroblasts, neuroblasts were electroporated to express tdTomato. Within the RMS, microglia closely appose individual neuroblasts (Fig. 1B, inset). This close interaction between microglia and neuroblasts within the RMS was confirmed with electron microscopy (Fig. 1C 1 -C 4 ), which showed microglia and their processes (green) intimately associated with migrating neuroblasts. We identified neuroblasts based on the following ultrastructural features: multiple nuclei within the nucleus, a dark cytoplasm with many free ribosomes, and the presence of extracellular spaces between neuroblasts (Doetsch et al., 1997;Matsumoto et al., 2019). In the biotin-labeled microglia distinctive features included the accumulation of chromatin on the inner membrane of the nucleus, the presence of lipofuscin granules in the cytoplasm and their comparatively small size relative to neuroblasts within the RMS (Fig.  1C 1 -C 4 ).
Microglia distribute along the RMS before the establishment of the vascular and astrocyte scaffold. While CD31stained vasculature is observed during the first postnatal week, it does not show the characteristic wrapping of the RMS that forms the scaffold during adulthood ( Fig. 2A 1 -A 3 ). Numerous tdTomato-labeled neuroblasts are seen at a distance from the developing vasculature, suggesting that they migrate independently of the vasculature during this period. An increase in CD31-stained blood vessels was observed at P14, with the beginnings of a network of vasculature forming around the RMS at this age ( Fig. 2A 4 ). Microglia closely appose endothelial cells in the RMS at P4 and P21 ( Fig. 2A 5 ,B).
Similarly, an astrocyte tube of thick septate processes that surrounds and infiltrates the RMS is present in the adult mouse (P56) but not during the first postnatal week (P7; Fig. 3C). Instead, radial glia stained with glial fibrillary acidic protein (GFAP) radiate out from the RMS core during the first postnatal week (Fig. 3A). Microglia organized along the periphery of the RMS during the first postnatal week closely intermingled with the GFAP-stained radial glial processes (Fig. 3A). Microglia also contacted the long pial fibers of radial glia with cell bodies in the SVZ (Fig. 3B). The gradual formation and thickening of astrocyte processes over development in the RMS is reflected in the significant increase in GFAP staining across the elbow between P4 and P56 (p , 0.0001; Fig. 3D 2 ). By P56 microglia exhibit a more homogeneous distribution across the RMS (Fig. 3C), reminiscent of the "tiling" observed in the mature cortex (Nimmerjahn et al., 2005). While the microglia distribution changes over the course of postnatal RMS development, there is no significant difference in the density of GFP1 microglia in the RMS between postnatal ages (Fig. 3D 2 ).
The morphology of microglia in the developing RMS may further give insight into their function. Ameboid microglia are round with short processes and appear to be highly motile and phagocytic (Davalos et al., 2020). Conversely, ramified microglia have long and highly branched processes that continuously sample their local environment (Davalos et al., 2005;Nimmerjahn et al., 2005). Microglia exhibit a greater range of morphologies during the early postnatal period in the RMS, including fewer morphologically complex and more "dense" cells (p , 0.0001; Fig. 4). The "dense" microglia observed at P2 and P4 implicate ameboid morphologies in the RMS during the first postnatal week (Fig. 4D). There is a decrease in very "dense" and ameboid cells by P14, which is accompanied by an increase in cell complexity (Fig.  4). Microglia in the RMS show a gradual increase in cell process complexity between P2 and P56, as demonstrated by   Research Article: New Research Figure 3. Microglia are present in the rostral migratory stream (RMS) before the formation of the astrocytic tube and interact with radial glia. A, Sagittal sections of the RMS from CX3CR-1 GFP/1 mice electroporated with a tomato plasmid at P0 and killed at ages P4 and P7. Migrating neuroblasts are depicted with tomato, microglia with GFP, and radial glia and astrocytes with GFAP. Microglia interact with GFAP-stained radial glial processes tangentially oriented at the border of the RMS (orange arrowheads) and with processes radially distributing within its core (orange arrows). Scale bar = 50 mm. B, Sagittal section of radial glia cell pial fibers extending from the SVZ of a CX3CR-1 GFP/1 mouse electroporated with a tomato plasmid at P0 and killed at ages P2. Some microglia interact with pial fibers via a process extension that partially wraps around the fiber in the manner of a "hand-hold" (red arrowheads). Scale bar = 50 mm. C, The RMS elbow from CX3CR-1 GFP/1 mice at P7 and P56. Microglia are represented by GFP (green) and radial glia and astrocytes by GFAP (cyan). Scale bar = 50 mm. Microglia are observed predominately at the periphery of the RMS at P7, where they associate with radial glia processes. While the astrocyte tube is not present at P7, the thick septate processes of the increase in the average fractal dimension (D B ) of binarized microglia (p , 0.01; Fig. 4B 1 ). This increase in the complexity of microglia processes reflects an increased ramification. This change in microglia morphology from ameboid to ramified during RMS development could simply be a manifestation of normal microglial cell development. However, it seems plausible that the functional state of microglia adapt in response to the signals from their local tissue environment (Matcovitch-Natan et al., 2016;Li and Barres, 2018), including cues from the emerging astrocyte and vascular scaffolds.

Neuroblast migratory capacity is independent of microglia in the developing rostral migratory stream
The presence of microglia in the RMS during the first postnatal week prompted the hypothesis that microglia may facilitate neuroblast migration in the absence of the vascular and astrocyte scaffold. We first investigated whether microglia depletion affects the density of neuroblasts migrating down the RMS; neuroblast clumping within the RMS following microglia depletion would implicate a lack of forward migration. Microglia depletion was conducted in CX3CR-1 GFP/1 ;DCX DsRed/1 transgenic mice to allow identification of microglia (GFP) and neuroblasts (DsRed; Fig. 5A 1 ). To further analyze the density of migrating neuroblasts, 25 mg/kg of BrdU was injected at P0 to label actively diving neuroblasts (Fig. 5A 2 ). Microglia depletion was initiated 2 h later with an injection of clodronate-filled liposomes in the cerebral lateral ventricles. PBS-filled liposomes injected in littermates served as a control. Mice were killed at P3 to determine the density of BrdU-labeled and DsRed-labeled neuroblasts in the RMS start and descent following microglia depletion. There was a significant reduction in the density of microglia in both the RMS start and RMS descent in mice injected continued GFAP-stained astrocytes are present in the RMS by P56. D 1 , There is a dramatic increase in GFAP staining between P4 and P56 in the RMS elbow, reflecting the development of the astrocyte tube. D 2 , Microglia are present in the RMS elbow at similar densities at P4, P7, P21, and P56. Figure 4. Microglia in the rostral migratory stream exhibit increased average pattern complexity at later postnatal ages. A 1 , A microglial cell from the RMS of a P4 CX3CR-1 GFP/1 mouse. Scale bar = 25 mm. A 2 , The microglia from A 1 segmented and binarized. A 3 , Fractal analysis data are gathered via box counting, in which a series of grids of decreasing caliber were systematically laid over an image and the number of boxes containing a pixel counted. The fractal dimension (D B ) is the relationship between how a pattern's detail (N) changes with the scale (« ) at which the image is considered. B 1 , There is significantly increased average pattern complexity at later postnatal ages, as shown by increased measures on the fractal dimension. B 2 , Example microglia shapes and their corresponding fractal dimension D B. Scale bar = 25 mm. C, Microglia cell "density" is measured by dividing the area of the cell by the area of its convex hull. The convex hull is in cyan, which is created by connecting a series of straight segments that enclose all the pixels of the binary image. The bounding circle (depicted in magenta) is the smallest circle enclosing all the pixels. D, There is a decrease in the proportion of "dense" and more "ameboid" cells at later postnatal ages. Eight to 20 microglia per animal.
with liposomal clodronate as compared with littermates injected with liposomal PBS (p , 0.001; Fig.  5B 1 ). However, microglia depletion was not accompanied by a difference in the density of either DsRed-labeled or BrdU-labeled neuroblasts in the RMS descent (p = n.s.; Fig. 5B 2 ,B 3 ). While there was a significant increase in the density of DsRed-labeled cells in the RMS start following liposomal clodronate treatment (p , 0.01; Fig. 5B 2 ), there was no significant difference in the density of BrdU-labeled neuroblasts in the RMS start (Fig. 5B 3 ). The absence of difference in BrdU labeling suggests that neurogenesis was not significantly altered following transient microglia depletion. The increase in DsRed staining may reflect a decrease in microglia phagocytosis of neural stem cells in the SVZ, which was not captured by the temporally sparse Figure 5. Microglia depletion after liposomal clodronate injection does not impact neuroblast density in the rostral migratory stream (RMS). A 1 , Coronal sections of the RMS start and descent from P3 CX3CR-1 GFP/1 ;DCX DsRed/1 littermates following injection with liposomal PBS or liposomal clodronate at P0 and subcutaneous 25 mg/kg BrdU injection 2 h later. The RMS region was defined by increased density of DAPI nuclear staining, as shown by the white dashed outline. Scale bar = 100 mm. A 2 , GFP-stained, DsRedstained, and BrdU-stained cells were segmented and binarized with Fiji software. Scale bar = 100 mm. B 1 , Three days after injection of liposomal clodronate, there was a significant decrease in microglia density in both the RMS start and RMS descent. B 2 , Microglia depletion was not accompanied by a corresponding decrease in the density of migrating neuroblasts in the RMS descent, as shown by the density of DsRed-labeled cells in this region. There was a significant increase in the density of DsRed-labeled cells in the RMS start following liposomal clodronate treatment, potentially reflecting a decrease in microglia phagocytosis of neural stem cells in the subventricular zone. B 3 , There was no significant difference in the density of newly generated BrdU-labeled neuroblasts in either the RMS start or descent between liposomal PBS and liposomal clodronate treated littermates. **p , 0.01, ***p , 0.001, ****p , 0.0001. To further examine whether microglia promote neuroblast migration we asked whether microglia depletion impacts the distance migrated by individual neuroblasts in the RMS. To accomplish this we used an alternative model of inducible microglia depletion. Homozygous Cx3cr1 CreER/CreER were crossed with heterozygous ROSA26 eGFP-DTA/1 mice to create Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 and Cx3cr1 CreER/1 ; ROSA26 1/1 littermates. Following tamoxifen injection, the toxic DTA subunit is expressed in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 mice, leading to the ablation of microglia. To label individual neuroblasts and evaluate the distance traveled, we electroporated a tdTomato plasmid into the SVZ at P0. We injected tamoxifen intraperitoneal 2 h later to selectively deplete microglia in those littermates that express the toxic DTA subunit. Littermates that did not express the DTA subunit served as controls. We killed the mice at P3 and assessed the density of tdTomato-labeled neuroblasts in the RMS start, descent and horizontal limb ( Fig. 6A-C) and their morphology (Fig. 6D,E) following microglia depletion.
Tamoxifen injection at P0 significantly depleted microglia 3 d later in both the RMS start and RMS descent (p , 0.0001; Fig. 6C 1 ). However, this was not accompanied by a difference in the density of neuroblasts in the RMS start, descent, or horizontal limb following microglia depletion in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 as compared with Cx3cr1 CreER/1 ; ROSA26 1/1 littermates (p = n. s.; Fig. 6C 2 ). There was also no difference in the ratio of tdTomato-labeled neuroblasts that reached the RMS descent over those in the RMS start, or neuroblasts that reached the horizontal limb compared with those still in the RMS descent for each animal (p = n.s.; Fig. 6C 3 ). The successful migration of neuroblasts to the more distal segments of the RMS following microglia depletion suggests that microglia are not necessary for neuroblast migration in the RMS. We next asked whether microglia depletion influenced neuroblast morphology, potentially hinting at more subtle disorganization of neuroblast migratory patterns (Fig. 6D,E). However, there was no significant difference in neuroblast pattern complexity (p = n.s.; Fig.  6F 1 ), heterogeneity (p = n.s.; Fig. 6F 2 ), density (p = n.s.; Fig.  6F 3 ) or span ratio (p = n.s.; Fig. 6F 4 ) following microglia ablation. The similar morphologies of neuroblasts traversing the RMS further indicates that microglia ablation does not impact the migratory capacity of neuroblasts.

Microglia mediate the homeostasis of the developing rostral migratory stream
The microglia cell distribution and ameboid morphology pattern in the early postnatal RMS suggests that microglia may monitor the RMS environment and apply trophic support or "quality assurance" measures through phagocytosis. Single-cell microglia analysis identified a subset with a distinct gene expression profile in the first postnatal week in the developing corpus callosum that express the phagocytic markers CD68 and CLEC7A and trophic factors IGF-1 and SPP1 (Hagemeyer et al., 2017;Hammond et al., 2019;Li et al., 2019). Microglial expression of receptor tyrosine kinase MERTK, which is important in the adult RMS (Fourgeaud et al., 2016), is also generally upregulated on the surface of embryonic microglia compared with the adult (Rosin et al., 2021).
We capitalized on these recent findings to ask whether microglia in the RMS the phagocytic markers CLEC7A, MERTK, CD68 (Fig. 7A) and trophic factor IGF-1 during early postnatal development. Microglia expressed phagocytic markers while "hugging" neuroblasts in the RMS of P7 CX3CR-1 GFP/1 ;DCX DsRed/1 mice and contained DsRed material within phagocytic cups (Fig. 7B). Microglia also expressed phagocytic markers while "hugging" neuroblasts in the corpus callosum of P7 CX3CR-1 GFP/1 ;DCX DsRed/1 mice (Fig. 9B), consistent with previous studies (Li et al., 2019). Clusters of microglia expressing the phagocytic markers CD68, CLEC7A, and MERTK were also found wrapping DsRed1 neuroblasts in the RMS start and ventral border of the RMS elbow (Fig. 7D-F). Occasional microglia expressing phagocytic markers were also found to wrap migrating neuroblasts in the core of the RMS descent and horizontal limb (Fig. 7C). Microglia in the RMS start of P4 CX3CR-1 GFP/1 mice expressed high transcript levels of Igf1; however, there was heterogeneity of expression, with microglia expressing high transcript levels of Igf1 immediately adjacent to microglia with no appreciable expression (Fig. 7G). continued software. The RMS region was defined by increased density of DAPI nuclear staining, as shown by the white dashed outline. Scale bar = 100 mm. C, There was a significant decrease in the density of Iba1-labeled microglia in the RMS following tamoxifen injection in DTA mice as compared with WT littermates. However, this was not accompanied by a difference in the density of tdTomato-labeled neuroblasts in the RMS start, descent, or horizontal limb. There was also no difference in the ratio of neuroblasts that reached the RMS descent/RMS start, or the ratio of neuroblasts that reached the horizontal limb/the number that were still traveling the descent. D, A segmented and binarized neuroblast. Fractal analysis data are gathered via box counting, in which a series of grids of decreasing caliber were systematically laid over an image and the number of boxes containing a pixel counted. E, A binarized neuroblast showing the convex hull used for morphology analysis. The convex hull is in cyan, which is created by connecting a series of straight segments that enclose all the pixels of the binary image. The bounding circle (depicted in magenta) is the smallest circle enclosing all the pixels. F 1 , There was no significant difference in neuroblast pattern complexity following microglia depletion, as shown by similar measures on the fractal dimension (D B ). F 2 , There was no significant difference in the heterogeneity of binarized neuroblast cells following microglia depletion, as shown by measures of lacunarity (K). F 3 , There was no difference in the "density" of neuroblast morphology following microglia depletion. Density is measured by dividing the area of a cell by the area of its convex hull F 4 , There was no difference in the "span ratio" of neuroblasts in the RMS following microglia ablation. The span ratio is the ratio of the major over minor axes of the convex hull. Nine to 13 neuroblasts per animal. ****p , 0.0001. Figure 7. Specialized microglia subtype in the postnatal rostral migratory stream (RMS). A, Illustration of microglia phagocytic markers. CD68 is a lysosome-associated transmembrane glycoprotein that is upregulated with phagocytosis. CLEC7A, or Dectin-1, is a transmembrane pattern-recognition receptor that triggers phagocytosis and the release of reactive oxygen species (ROS). MERTK is a member of the TAM (Tyro3, Axl, and Mertk) family of receptor tyrosine kinases that recognize soluble bridging proteins that bind externalized phosphatidylserine, including GAS6 and protein S. B, 3D volume rendering of GFP1/CLEC7A1 microglia and DsRed1 neuroblasts in the RMS start of a P7 mouse with Imaris software. The yellow arrow shows an example of a GFP1/CLEC7A1 microglial cell "wrapping" a DsRed1 neuroblast. The yellow arrowhead indicates DsRed-labeled material within a microglia phagocytic cup. Scale bar = 10 mm. C, Sagittal section showing the interactions between microglia (green) and neuroblasts (red) in the RMS (dashed white outline) of a P7 CX3CR-1 GFP/1 ;DCX DsRed/1 mouse. Microglia exhibited the phagocytic marker CD68 while "wrapping" neuroblasts in the RMS. Microglia CD68 expression was marked at the start of the RMS and at the ventral border of the elbow. CD681 microglia also "hugged" neuroblasts radially migrating from the subventricular zone through the corpus callosum and start of the RMS. cc, corpus callosum; st, start; ve, ventral elbow. Scale bar = 250 mm. Scale bar of magnified inset = 25 mm. Microglia expressed the phagocytic markers CLEC7A (D), MERTK (E), and CD68 (F) while wrapping neuroblasts at the start of the RMS (C 2 , D 2 , F 2 ) and at the ventral border of the RMS elbow (C 1 , D 1 , F 1 ). Scale bar = 25 mm. G, Microglia expressing high transcript levels of the trophic factor Igf1 neighbor microglia with no Igf1 expression at the RMS start of a P4 CX3CR-1 GFP/1 mouse. Scale bar = 25 mm.
We next sought to elucidate whether microglia monitor and phagocytose antigens in the RMS. To demonstrate the phagocytic capacity of microglia throughout the RMS, fluorescent liposomes containing the fluorescent dye Dil (Fluoroliposome) were injected into the cerebral lateral ventricles of P1 CX3CR-1 GFP/1 mice and killed at P4. Fluoroliposomes accumulated within microglia to a high degree in the SVZ but were also observed in distal regions of the RMS (Fig. 8). Microglia thus appear to monitor and phagocytose extracellular material within the early postnatal RMS.
To demonstrate a functional role for microglia in regulating the population of migrating neuroblasts in the developing RMS, we depleted microglia for 3 and 14 d postnatally to evaluate whether there was an accumulation of apoptotic neuroblasts in the RMS. Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 and Cx3cr1 CreER/1 ; ROSA26 1/1 control littermates were injected with TMX intraperitoneally starting at P0 and killed at P3 or injected with TMX every 3 d subsequently until killing at P14. To label microglia, neuroblasts and apoptotic cells, immunohistochemistry was performed with primary antibodies against Iba1 (data not shown), doublecortin (DCX), and cleaved caspase-3 (CC3; Figs. 9A, 10A). There was a significant decrease in microglia density in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 animals compared with Cx3cr1 CreER/1 ; ROSA26 1/1 control littermates after 3 d of TMX treatment (RMS start p , 0.0001; RMS descent p , 0.0001). Microglia depletion was associated with a significant increase in the accumulation of CC31/DCX1 apoptotic cells in the RMS start and RMS descent at both P3 (WT vs DTA p , 0.01; Fig. 9B) and P14 (WT vs DTA p , 0.01; Fig. 10B 1 ), suggesting an accumulation of apoptotic neuroblasts. There was no significant difference in the density of CC31/DCXÀ apoptotic cells in the RMS following microglia depletion in either group (Fig. 10B 1 ). An extended period of microglia depletion was also associated with an increase in width of the RMS descent (p , 0.01; Fig. 10B 2 ) and decreased lacunarity of DCX-stained processes within the RMS (p , 0.05; Fig. 10B 3 ). The expanded RMS width and decrease in "gappiness" between stained neuroblast processes may reflect an increase in the number of neuroblasts or neuroblast clumping within the RMS following microglia depletion. Of further interest, we also found patent olfactory ventricles in the core of the RMS in all six mice that expressed the DTA1 subunit following 14 d of microglia depletion (Fig. 10C).

Discussion
We sought to examine whether the ameboid microglia that invade the RMS early in development regulate the population of migrating neuroblasts. We found that microglia line the outer borders of the RMS and intimately associate with migrating neuroblasts in the early postnatal period. Microglia depletion did not impact the migratory capacity of neuroblasts in the postnatal RMS. However, microglia express phagocytic and trophic markers, "wrap" migrating neuroblasts, and contain DsRed-labeled material within phagocytic cups, suggesting possible phagocytosis or trophic support of neuroblasts in the RMS. Furthermore, microglia depletion induced an accumulation of apoptotic and densely packed neuroblasts, an expanded RMS, and patent olfactory ventricles. Together, these results suggest that microglia function to maintain a healthy neuroblast population and environment permissive to neuroblast migration.
Microglia are present in the RMS during the first postnatal week, where they exhibit an ameboid morphology, collect at the borders of the RMS and closely associate with migrating neuroblasts. This microglia distribution occurs before the assembly of the vascular scaffold and astrocyte tube. There was only sparse CD31 blood vessel labeling in the RMS during the first postnatal week, which lengthened at P14 to resemble the scaffold that parallels the path of neuroblast migration in adults (Bozoyan et al., 2012). An astrocyte tube was similarly not observed during the first postnatal week. Instead, GFAP-stained radial glia radiated from the RMS core (Alves et al., 2002). However, neuroblast migration was orthogonal to these processes, suggesting that migration may be independent of radial glia (Alves et al., 2002). A fine network of astrocyte processes coalesced into thick septate processes at P21 that further consolidated into the densely packed astrocytic tube typical of adults at P56 (Peretto et al., 2005). The distribution of microglia flanking the outer borders of the RMS during the first postnatal week suggests that they may provide cues to migrating neuroblasts in the absence of the vascular scaffold and astrocyte tube.
This microglia cell distribution and morphology pattern suggests that microglia monitor the RMS environment and may apply trophic support or phagocytic functions when appropriate. There is a preponderance of ameboid microglia in the RMS during the first postnatal week  before they shift to a more ramified morphology, indicating that microglia functions may differ between these periods. "Dense" and ameboid microglia in the RMS decreased in number by P14, with a concomitant increase in microglia process complexity by P21.
This change in microglia morphology could simply be a manifestation of normal microglial cell development. Microglial cell development follows a step-wise developmental program with distinct temporal stages in gene expression (Matcovitch-Natan et al., 2016;Hammond et al., 2019;Li et al., 2019;Masuda et al., 2019Masuda et al., , 2020. Throughout microglial cell development there is a gradual expression of adult homeostatic genes, such as Tmem119, P2ry12, and Hexb (Bennett et al., 2016;Matcovitch-Natan et al., 2016). Meanwhile, mutations in the transcription factor SALL1 induce an abnormal ameboid microglia morphology and downregulation of the homeostatic genetic signature (Buttgereit et al., 2016;Utz et al., 2020). Alternatively, varying microglia morphologies and shifts in gene expression could reflect different functional Figure 9. Microglia depletion for 3 d postnatally is associated with an accumulation of apoptotic cells in the rostral migratory stream (RMS). A, Coronal sections of the descent of the RMS in Cx3cr1 CreER/1 ; ROSA26 1/1 (left) and in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 (right) littermates following intraperitoneal injection of tamoxifen at postnatal day 0 (P0) and killing at P3. The RMS region was defined by increased density of DAPI nuclear staining, as shown by the white dashed outline. Apoptotic cells were identified with cleaved caspase-3 (CC3) labeling (cyan) and neuroblasts with doublecortin (DCX) staining (red). Scale bar = 50 mm. B, There was a significant increase in the density of CC31/DCX1 cells in the RMS of DTA1 mice following injection of tamoxifen in the RMS descent. There was no significant difference in the density of CC31/DCXÀ cells in the RMS descent following microglia depletion. **p , 0.01. Figure 10. Microglia depletion for 14 d postnatally is associated with an accumulation of apoptotic cells and patent olfactory ventricles in the rostral migratory stream (RMS). A, Coronal sections of the start and descent of the RMS in Cx3cr1 CreER/1 ; ROSA26 1/1 (left) and in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 (right) littermates following intraperitoneal injection of tamoxifen at postnatal day 0 (P0) and every 3 d subsequently until killing at P14. Apoptotic cells were identified with cleaved caspase-3 (CC3) labeling (cyan) and neuroblasts with doublecortin (DCX) staining (red). Scale bar = 50 mm. B 1 , There was a significant increase in the density of CC31/ DCX1 cells in the RMS of DTA mice following 14 d of microglia depletion in the RMS descent, but not in CC31/DCXÀ cells. B 2 , There was a significant increase in the width of the RMS at its widest point at the RMS descent. The RMS region was defined by increased density of DAPI nuclear staining, as shown by the wshite dashed outline in A. B 3 , There was a significant decrease in measures of lacunarity (K) of DCX1 stained processes in the RMS descent following microglia depletion, suggesting decreased gaps in the space between stained processes. C, Coronal sections of the whole olfactory bulb (OB) in Cx3cr1 CreER/1 ; ROSA26 1/1 (left) and in Cx3cr1 CreER/1 ; ROSA26 eGFP-DTA/1 (right) littermates. Pups were injected every 3 d starting at P1 and killed at P14. The RMS region is defined by increased density of DAPI nuclear staining in the OB core, as shown by the white dashed outline. A decrease in the number of microglia (green) is seen in the OB of the DTA1 mouse (right) following TMX injections. Enlarged and patent olfactory ventricles were observed in DTA1 mice (V), as opposed to the closed olfactory ventricles of WT littermates (left). Scale bar = 250 mm. *p ,0.05, **p , 0.01. activities of microglia at different developmental stages. While single-cell RNA-sequencing revealed little microglia transcriptomic heterogeneity in adult mice, microglia isolated from developing mice exhibited heterogeneous gene expression profiles, suggesting diverse functionality (Hammond et al., 2019;Li et al., 2019;Masuda et al., 2019). Postnatal microglia transcript clusters further illustrated varied distribution across brain locations, implicating local specification of microglia function during development (Masuda et al., 2019). Meanwhile, whole genome transcriptional profiling of microglia showed enrichment of biological processes related to neurogenesis and neural development during embryonic life (E14.5) and early postnatal life (P1), but not adulthood (Grassivaro et al., 2020). Thus, specific microglia subsets may occur in a spatially and temporally restricted matter to regulate specific neurodevelopmental events.
We next considered whether the ameboid microglia lining the RMS in the first postnatal week regulate the population of migrating neuroblasts. Microglia do not appear to alter the intrinsic migratory capacity of neuroblasts in the RMS: microglia depletion for 3 d postnatally with two different depletion strategies did not alter the density of neuroblasts within the RMS nor the distance they migrated. These results indicate that microglia are not necessary for neuroblast migration in the early postnatal RMS. This is consistent with the findings of Kyle et al., who found no difference in the density of BrdU-labeled neuroblasts 4 d after BrdU treatment in the RMS of adult animals treated with the oral CSF-1R inhibitor PLX5622 (Kyle et al., 2019). However, microglia depletion for 14 d postnatally did induce an accumulation of apoptotic neuroblasts and decreased lacunarity of DCX-stained neuroblast processes within a broader RMS domain, suggests clumping of apoptotic neuroblasts to a degree that widens the RMS. These findings are consistent with the buildup of apoptotic cells observed after microglia depletion and selective microglial MERTK knock-down in the RMS of adult animals (Ribeiro Xavier et al., 2015;Fourgeaud et al., 2016).
The accumulation of apoptotic neuroblasts following microglia depletion could be secondary to a lack of appropriate microglia phagocytosis of defective or apoptotic neuroblasts. Microglia lining the RMS in the first postnatal week express the phagocytic markers CLEC7A, MERTK, and CD68; in contrast, microglia in the cortical regions outside the RMS do not, suggesting a specific role for microglia phagocytosis in the RMS. Microglia further contain DsRed-labeled neuroblast material within phagocytic cups. This parallels the findings of P7 postnatal microglia that express the phagocytic marker CLEC7A in the corpus callosum (Li et al., 2019). These microglia demonstrated an increase phagocytic capacity, engulfing fluorescent beads in brain slices (Li et al., 2019). These microglia also contained CC3 positive inclusions that were positive for myelin basic protein, suggesting that microglia phagocytose oligodendrocytes. Interestingly, oligodendrocytes that were negative for cleaved caspase three were also frequently observed to be "hugged" by early postnatal microglia, suggesting that they may actively contribute to eliminating oligodendrocytes in developing white matter tracts (Li et al., 2019). Similarly, we found that microglia expressing the phagocytic markers CLEC7A, CD68, and MERTK "hug" neuroblasts in the RMS, potentially suggesting phagocytosis of migrating neuroblasts.
The accumulation of neuroblasts within a wider RMS following microglia depletion suggests that a surplus of neuroblasts may be generated during development, which are subsequently eliminated by microglia. Microglia phagocytose neural precursor cells in both the rat and primate SVZ during all stages of prenatal development (Cunningham et al., 2013) as well as maturing neurons in the OB (Lazarini et al., 2012;Denizet et al., 2017). Microglia can also "multitask," with individual cells simultaneously enveloping and phagocytosing multiple neural precursor cells (Barger et al., 2019). A role of microglia for clearing defective cells via phagocytosis may be reflected in the child born without microglia (Oosterhof et al., 2019); the observed heterotopias may be a result of a lack of appropriate phagocytic clearance of defective neuroblasts, as opposed to a general failure of neuroblast migration in the absence of microglia. However, it remains unknown whether microglia simply phagocytose dead or dying neuroblasts in the RMS, or whether they "kill" neuroblasts in the RMS via phagoptosis. There is evidence that microglia can "kill" cells in neurogenic regions through an extracellular respiratory burst (Marín-Teva et al., 2004;Sierra et al., 2010). Microglia depletion caused an increase in Purkinje cell number in the cerebellum, suggesting that microglia eliminated "living" and not apoptotic Purkinje cells (Marín-Teva et al., 2004). Microglia also phagocytosed mitotically active neural precursor cells independent of apoptotic pathways in the SVZ (Cunningham et al., 2013). Furthermore, despite the accumulation of apoptotic cells in the RMS of Axl and Mertk knock-out animals, there was a ;70% increase in the cellular density of the granule cell and glomerular layers of the OB, suggesting that many neuroblasts phagocytosed by microglia are not apoptotic but instead "eatenalive" (Fourgeaud et al., 2016). Future studies may reveal whether microglia randomly target migrating neuroblasts for phagoptosis or are able to detect aberrantly migrating or otherwise defective neuroblasts.
Alternatively, the accumulation of apoptotic neuroblasts following microglia depletion and microglia "wrapping" of neuroblasts could suggest trophic support of migrating neuroblasts. IGF-1 is also expressed by the CD681 CLEC7A1 microglia subtype in the first postnatal week (Hammond et al., 2019;Li et al., 2019) and supports oligodendrocyte precursor cells (Wlodarczyk et al., 2017) and projecting axons of layer V cortical neurons (Ueno et al., 2013) in the corpus callosum. We also observed IGF-1 expression in RMS microglia during the first postnatal week. Thus, withdrawal of supportive trophic factors, such as IGF-1, following microglia depletion could result in the accumulation of apoptotic neuroblasts that we observed in the RMS. There may also be regional heterogeneity of microglia functions within the RMS, with some microglia being responsible for trophic support and others for neuroblast elimination. Such heterogeneity was demonstrated in ameboid microglia expressing SPP1 in the corpus callosum during the first postnatal week: microglia expressing high transcript levels of Spp1 neighbored microglia with no expression (Hammond et al., 2019). Paralleling this finding, SPP1 was also observed in a subset of CLEC7A1 microglia in the developing corpus callosum of P7 adjacent to CLEC7A1 with no SPP1 expression (Li et al., 2019). Similarly, we observed microglia with high transcript levels of Igf1 neighboring microglia with no expression in the RMS start. There may therefore be local heterogeneity of microglia functions in the early postnatal RMS, with microglia responsible for neuroblast elimination neighboring microglia that support neuroblast survival.
Microglia also phagocytose labeled antigens in distal regions of the RMS, indicating the monitoring and phagocytosis of extracellular material. The enlarged and patent olfactory ventricles we observed following 14 d of postnatal microglia depletion could also be secondary to deficient microglia phagocytosis. It is possible that phagocytic microglia lining the RMS are responsible for clearing apoptotic debris, and thus an accumulation of debris in their absence may obstruct normal CSF outflow, increase CSF pressure and induce ventriculomegaly. This parallels early findings of enlarged and patent olfactory ventricles observed in three-week-old Csf1 null mice that have a significant depletion of microglia (Erblich et al., 2011). An accumulation of apoptotic debris and other extracellular material in the RMS may similarly impede neuroblast migration over time. Thus, another important function of microglia in the developing RMS may be the monitoring of draining antigens and maintenance of RMS homeostasis by responding to these antigens and clearing debris.
In summary, our data suggest that a specialized microglia subtype is involved in maintaining a healthy neuroblast population and environment permissive to neuroblast migration in the RMS during early postnatal development. We found that ameboid microglia closely associate with migrating neuroblasts in the postnatal RMS before the formation of the glial tube and vascular scaffold that facilitate neuroblast migration in adulthood. Microglia in the first postnatal week further express the markers CD68, MERTK, CLEC7A, and IGF-1 and "wrap" migrating neuroblasts within the RMS. While the migratory capacity of neuroblasts was unaffected by microglia depletion, extended microglia depletion induced an accumulation of apoptotic neuroblasts within a broader migratory corridor. Our data suggest that a specialized ameboid microglia subtype expressing the markers CD68, MERTK, CLEC7A, and IGF-1 is critical for maintaining a healthy neuroblast population that enables an environment permissive to neuroblast migration in the RMS. As our understanding of the functional capabilities of microglia continues to expand, these results shed important insight into the role of microglia in regulating the homeostasis of neuroblast migratory corridors during the early postnatal period.