Animal models

All experiments were performed in compliance with protocols approved by the Institutional Animal Care and Use Committee at Cold Spring Harbor Laboratory according to protocol 20-3. The following mouse lines were obtained from the Jackson Laboratory: C57Bl/6J (JAX, 000664), NG2-CreERT2 (B6.Cg-Tg(Cspg4-Cre/Esr1*)BAkik/J; JAX, 008538) and Rosa26-CAG-LsL-tdTomato (B6.Cg-Gt(ROSA)26Sortm14(CAG-tdTomato)Hze/J (Ai14); JAX, 007914). NG2-CreERT2 mice were bred with Ai14 mice inhouse to yield NG2-CreERT2tdTomato mice in which OPCs are labeled with tdTomato on tamoxifen (TAM) administration. Except when noted, animals were housed in normal 12-h light:12-h dark cycles at average temperatures of 68–70 °F and 54–58% humidity. Analyses were performed on equal numbers of male and female mice at postnatal days (P)10, P20, P27 and P90. Live imaging was performed on animals between 2 and 6 months of age. No sex differences were observed in this study.

Sensory deprivation and stimulation paradigm

C57BL/6J mice were reared according to a standard 12-h light/12-h dark cycle until P20, at which point they were weaned from their mothers and separated into experimental groups. One cohort continued to be housed in normal light conditions until P27. Two cohorts of mice were placed in a well-ventilated light-proof cabinet (Actimetrics) until P27 (LDR). One of the two cohorts subjected to LDR was collected at P27 in the dark by an investigator using night vision goggles, whereas the other was acutely re-exposed to light for 10 h and then collected. These cohorts are referred to as LDR and LDR + 10, respectively, throughout the paper.

Plexxikon 5622 administration

Mice were fed 1,200-mg kg−1 irradiated freebase Plexxikon 5622-formulated chow (PLX; Research Diets, Inc.: D1110404i), which blocks colony-stimulating factor receptor 1, or control chow produced in parallel (D19101002) from P18 to P27 to pharmacologically deplete microglia from the brain. Mice were fed on the chow ad libitum, and the investigator provided all husbandry for the mice during treatment.

TAM administration

TAM (Sigma, T5648) was dissolved in sunflower oil (Sigma, S5007) overnight at 37 °C with shaking to achieve a working concentration of 20 mg ml−1. Animals were administered a bolus of TAM at 150 mg kg−1 on two consecutive days at a minimum of 2 weeks before imaging and analysis.

Immunofluorescence

Animals were anesthetized with isoflurane and perfused with ice-cold PBS followed by 4% paraformaldehyde (PFA). Brains were removed and incubated in 4% PFA overnight at 4 °C. The next day, the brains were washed by rotating in PBS three times for 10 min each and allowed to sink in 15% and then 30% sucrose at 4 °C. Brains were then embedded in optimal cutting temperature (OCT; VWR, 25608-930) and stored at −80 °C. Coronal or sagittal sections with 25-µm thickness containing V1 and dLGN were collected onto Superfrost Plus microscope slides (Thermo Fisher Scientific, 1255015) using a cryostat and stored at −80 °C. For staining, the sections were washed at room temperature for 10 min in PBS and then dried for 10 min at 60 °C. A hydrophobic barrier was drawn around the samples using an ImmEdge Pen (VWR, 101098-065). A blocking solution (5% fetal bovine serum (FBS) or normal goat serum (NGS) and 0.3% TritonX-100, in PBS) was applied for 1 h at room temperature. Next, the blocking solution was replaced with primary antibodies prepared in a probing solution (5% FBS or NGS and 0.1% Triton in PBS). Primary antibodies were incubated at 4 °C overnight in most cases and for 48 h for staining with rat anti-NG2 (1:250; Thermo Fisher Scientific, MA5-24247). Other primary antibodies used included guinea pig anti-VGluT2 at 1:1,000 (Sigma, AB2251-l); rabbit anti-Iba1 at 1:1,000 (Wako, 019-19741); rat anti-Lamp2 at 1:200 (Abcam, AB13524); rabbit anti-Sox10 at 1:100 (Abcam, AB227680); chicken anti-Homer1b/c at 1:1,000 (Synaptic Systems, 160026); chicken anti-GFAP at 1:500 (Abcam, ab4674); rat anti-MBP at 1:1,000 (Abcam, ab7349); rabbit anti-LRP1 at 1:500 (Abcam, ab92544); rabbit anti EEA1 at 1:1,000 (Abcam, ab2900); and rabbit anti-Rab7 at 1:500 (Abcam, ab137029). After primary incubation, the tissue was washed four times with washing solution (PBS adjusted to 0.1% Triton) for 10 min per wash. The following Alexafluor secondary antibodies (Invitrogen) were diluted to either 1:1,000 or 1:500 in probing solution and incubated on the tissue for 1 h at room temperature: goat anti-guinea pig 647 (A21450); goat anti-guinea pig 555 (A21435); goat anti-guinea pig 488 (A11073); donkey anti-rabbit 405 (A21450); goat anti-rabbit 488 (A11008); goat anti-rabbit 647 (A21428); goat anti-rat 405 (A48261); donkey anti-rat 488 (A21208) and goat anti-rat 647 (A21247). Following secondary incubation, the sections were washed four times with washing solution, mounted with Fluoromount-G (SouthernBiotech, 0100-01) and coverslipped.

Confocal imaging parameters

In most cases, confocal images were acquired on an LSM 710 or LSM 780 (Zeiss) microscope with ×20/0.8 NA (air) and ×63/1.4 NA (oil) objectives. Z-stack images were acquired to capture OPCs and/or microglia in layer 4 of V1.

NG2-CreERT2tdTomato validation

Fixed brain sections from NG2-CreERT2tdTomato mice were stained for Sox10 and NG2 as described above and imaged on a confocal microscope at ×20. To determine the percentage of tdTomato+ cells that were either within the oligodendrocyte-lineage (Sox10+) or putative OPCs (NG2+), maximum projections of the images were generated to be manually counted with the ‘Cell Counter’ plugin in ImageJ. First, markers were placed at all tdTomato+ cells. Next, at each tdTomato+ marker location, the investigator determined and tabulated whether the location was also Sox10+ or NG2+.

PLX validation

To confirm the depletion of microglia by PLX, mice were killed 3 and 10 d after being placed on the PLX or control chow. Sections were immunostained for the microglial marker Iba1, and confocal images of the cortex were acquired with a ×20 objective. Using the ImageJ cell counter plugin, the number of cells in an image was manually counted. To ensure that the population densities of oligodendroglia and OPCs were unaffected, sections at the 10-day timepoint were stained for Sox10 and NG2. Images of ×20 with the equivalent number of z-stacks were taken, and the cell counter plugin was used to count the number of Sox10+ cells (oligodendroglia) and Sox10+NG2+ cells (OPCs). We also stained sections and analyzed mean intensities for the astrocyte marker GFAP and the mature oligodendrocyte–myelin marker MBP as described above to rule out PLX-driven changes unrelated to microglial depletion.

Fluorescence in situ hybridization

Brains were fixed by perfusion in 4% PFA, embedded in OCT and stored at −80 °C until processing. Sections of 25-µm thickness were mounted on Superfrost Plus slides. Multiplexed single-molecule fluorescence in situ hybridization (FISH) was performed using the RNAscope platform V2 kit (Advanced Cell Diagnostics (ACD), 323100) according to the manufacturer’s protocol for fixed-frozen sections. The samples were mounted with Fluoromount-G with DAPI (SouthernBiotech, 0100-20) and coverslipped. Commercial probes obtained from ACD detected the following genes: Pdgfra (480661-C2), Mertk (441241-C3), Calcrl (452281-C3), Arsb (837631) and Ptprj (883051).

FISH with fresh-frozen tissue

Brains were embedded in OCT and stored at −80 °C until processing. A Leica CM3050 cryostat was used to make sections of 20-µm thickness, which were mounted on Superfrost Plus slides. Multiplexed single-molecule FISH was performed using the RNAscope platform V1 kit (ACD, 320850) according to the manufacturer’s protocol for fresh-frozen sections. The samples were mounted with Fluoromount-G with DAPI (SouthernBiotech, 0100-20) and coverslipped. Commercial probes obtained from ACD detected the following genes: Pdgfra (480661-C2), Rorb (444271) and VGluT2 (319171-C2 and 319171).

To quantify VGluT2 transcripts in Pdgfra+ and Rorb+ cells, ×63 confocal images were acquired using an LSM 780 Zeiss microscope. Tissue from three mice was used and a minimum of five images were analyzed per mouse. VGluT2 expression was quantified using an ImageJ macro built inhouse. The macro first thresholded and binarized the DAPI channel before expanding it using the dilate function. This mask was then put through a watershed filter to ensure that cells that were near each other were separated. The resulting DAPI mask was used to create regions of interest (ROIs), where each ROI was considered a single cell. The number of FISH puncta within each ROI was counted using the 3D image counter function within ImageJ. ROIs were identified according to the following criteria: ROIs containing five or more Pdgfra or Rorb molecules were considered to be positive for each of the transcripts.

Reanalysis of single-cell RNA-sequencing data

Data from ref. 23 were downloaded as a raw count matrix from the Gene Expression Omnibus database (GSE102827). The data were processed via the Seurat v3 pipeline using standard parameters. OPC clusters were identified by enriched expression of Pdgfra. For Extended Data Fig. 7, transcripts were plotted across the Uniform Manifold Approximation and Projection (UMAP) using the FeaturePlot function in Seurat.

Engulfment quantification

Preprocessing of ×63 images stained for OPCs and/or microglia was performed in ImageJ. First, the ‘enhance contrast’ command was run so that 0.1% of pixels would be saturated, and then, a mean filter with pixel radius of 1.5 µm was applied. Next, an ROI was drawn around a given cell, cropped into its own file and used in downstream Imaris (BitPlane) processing and analysis. In Imaris, volumetric reconstructions of the fluorescence images were created using the ‘Spots’ and/or ‘Surfaces’ objects. A surface object was used to reconstruct a cell of interest following the guided creation wizard. The investigator then deleted any discontinuous part of the surface that could not be clearly traced back to the soma of the cell, using the fluorescence as a reference. Next, this cell surface was used as the ROI to create a mask of target channels (for example, VGluT2, Lamp2), defined by the signal included within the surface. New surface objects were generated using these masks, which represent the internal contents of the cell. Across conditions within an individual biological replicate, the same creation parameters were used in generating the internalized surfaces. The volumes of the cell and internalized surfaces were collected from the statistics tab in Imaris. The internalized volume was then normalized to a cell’s volume to represent the amount of engulfment by a given cell.

In some cases, masks of the Lamp2 channel within an OPC surface were made to reconstruct lysosomes. To quantify lysosome contents, the same internalization approach was used where the lysosomal surface was treated as the ‘cell’ to mask the target channel.

To quantify the distance of engulfment loci from the center of an OPC, we first defined the center of the cell by manually placing a spots object at the soma of the reconstructed cell using the autodepth-based-on-fluorescence option. Next, a spots object of the masked VGluT2 signal was created using the wizard, with a spots-diameter of 2 µm. This approach captured regions with one or more VGluT2 puncta per spot, which are referred to as ‘loci’ in the main text. The predefined ‘shortest distance to spots’ statistic was used to measure the distances between the engulfment loci and the center of the cell. To quantify the range of synaptic surveillance by a given OPC, the axis-aligned bounding-box statistic of the engulfment locus spots was collected.

Quantification of intact synapse contact and engulfment

NG2-CreERT2tdTomato tissue was stained for the presynaptic marker VGluT2 and the postsynaptic marker Homer1b/c. Images that focused on tdTomato+ OPCs were acquired on a confocal microscope with a ×63 objective. Next, the images were cropped to include single OPCs and transferred into Imaris. Within Imaris, the ‘Coloc’ tab was used to create a new colocalization channel—a channel depicting overlapping signals of VGluT2 and Homer1b/c. The mean intensities of the VGluT2 and Homer1b/c channels were used as thresholds when creating the colocalization channel. A spots object was then created from this new channel, representing intact (containing both pre- and postsynaptic compartments) synapses. An OPC was reconstructed using the Surface object as described above. Finally, the intact synapse spots were filtered for spots less than −0.132 µm from the OPC surface (that is, more than 1 pixel within the surface of the OPC). To estimate the percentage of all VGluT2–Homer1b/c synapses an OPC interacts with, the number of internalized synapses was divided by the total number of synapses within the cropped image.

Structured illumination microscopy

Brains were fixed in 4% PFA and then embedded in OCT and stored at −80 °C. Sections of 25-µm thickness were subjected to immunofluorescence in a free-floating format. Sections were stained with primary and secondary antibodies as described under ‘Immunofluorescence’ above, but in larger volumes of 1 ml. Stained sections were mounted onto thickness no. 1 ½ High-performance Zeiss cover glasses (Thermo Fisher Scientific, 10474379) and then centered onto Superfrost Plus microscope slides with ProLong Gold Antifade Mountant (Thermo Fisher Scientific, P36934). Stained samples were kept at 4 °C until imaging. 3D structured illumination microscopy (3D SIM) images of fixed, stained samples were acquired using an Applied Precision V3 OMX system equipped with a ×100/1.4 NA U-PLANAPO objective (Olympus) and two Cascade II 512 EM-CCD cameras (Photometrics). Stacks of six optical sections (125-nm step) were acquired consecutively in two channels (488 nm and 593 nm) using DeltaVision software (Applied Precision). 3D super-resolution image stacks were reconstructed using SoftWorx 6.5.2 using channel-specific OTFs and Wiener filter settings of 0.001 or 0.002. These image stacks were then imported into Imaris for volumetric reconstructions.

Airyscan imaging

Samples were imaged on a Zeiss LSM900 confocal microscope using the Airyscan mode with a ×63/1.4 NA objective. The image underwent Airyscan processing with autofilter selected within Zeiss Zen. The processed image was then transferred to Imaris for volumetric reconstruction as described above.

STED microscopy

Samples were stained for NG2 and VGluT2 and mounted using ProLong Diamond Antifade (Thermo Fisher Scientific, P36965) and coverslipped with No. 1.5H coverslips (Azer Scientific, ES0107222). Images were acquired on a Leica SPX8 STED microscope with a ×100/1.4 NA objective on Leica Application Suite X software. In Fig. 1j, single slices and maximum projections for two example OPCs are shown. For the maximum projections, some z sections were omitted due to imaging artifacts.

pAAV:hSYN-synaptophysin-mCherry-eGFP (pSynDig)

We purchased the presynaptic ATP Syn-ATP reporter (Addgene, 51819) and the fluorescent marker hSYN–eGFP (Addgene, 50465) plasmids for downstream creation of the probe for synaptic digestion (pSynDig) construct. We amplified an elongated coding region for synaptophysin–mCherry from the Syn-ATP construct (Addgene, 51819; forward primer, 5′-GCGCAGTCGAGAAGGTACCGGCAGCAATGGACGTGGTG-3′, 5′-reverse primer: CCTTGCTCACCATGGTGGCGGGTCCCTTGTACAGCTCG-3-′). The elongated coding region (1,728 bp) was then gel-purified using the Qiagen MinElute Gel Extraction Kit (28604) and used in downstream Gibson assembly with the New England Biolabs HiFi DNA Assembly Cloning Kit (NEB, E5520S) to insert the amplified region into the hSyn–eGFP vector (Addgene, 50465) following its linearization with BamHI (NEB, R0136S). The resulting plasmid was gel-purified and sequenced before packaging into an adeno-associated virus (AAV) to yield AAV9-hSYN-pSynDig at a titer of 1.2 × 1014 genome copies per ml.

pSynDig validation and quantification

We confirmed that pSynDig injections into the dLGN labeled VGluT2+ TC inputs in V1 by imaging mCherry+, eGFP+ and VGluT2+ puncta (following immunostaining with an antibody against VGluT2) in confocal images taken with a ×63 objective at Nyquist settings for downstream deconvolution using Huygens Essentials (SVI). Images were deconvolved based on imaging parameters (for example, pixel resolution, excitation wavelength, number of excitation photons, depth of acquisition, numerical aperture of objective and medium and emission wavelength), and the Huygens express deconvolution wizard was set to conservative deconvolution as a means of increasing image resolution and signal to noise. We measured the intensity of each channel using line intensity quantification in ImageJ. We observed largely overlapping mCherry and eGFP signal colocalized with VGluT2, as expected. We next measured the degree of colocalization between the eGFP and mCherry signals using the mean Pearson’s correlation coefficients across three images per animal. As expected, we observed a small percentage of puncta that were mCherry+ but eGFP−. To verify that these mCherry+ eGFP− puncta represented inputs in the process of lysosomal degradation, we immunostained the tissue for the lysosomal marker Lamp2 and took 3D images (z-stacks) on a confocal microscope using a ×63 objective. In postprocessing, we applied a Gaussian blur of 0.132 µm. Maximum projections of the z-stacks were made, and then, the mean intensities of the mCherry, eGFP and Lamp2 channels were quantified. Images were then processed using Imaris, where lysosomes were reconstructed using a Surface object. In the wizard, the recorded mean intensities from ImageJ were used as the threshold value to define the surfaces. Masks of the eGFP and mCherry channels were made from the lysosome surfaces to analyze the within-lysosome pSynDig signal. Surfaces of the masked eGFP and mCherry signals were created, this time using 0.75 * (recorded mean intensity) as the threshold. To quantify the intensity of mCherry and eGFP outside of lysosomes, the masked channels were subtracted from the original channels, thereby removing any signal that was within a lysosome. mCherry and eGFP surfaces were generated from these subtracted channels, using the same wizard thresholds as for the within-lysosome group. The sum intensity statistic of the eGFP and mCherry signals was collected for the within- and outside-lysosome surfaces, before being normalized to the respective mCherry signal. The example image in Extended Data Fig. 8a was taken with a ×63 objective with ×1.5 digital zoom, before being deconvolved for clarity.

To quantify pSynDig in OPCs, the same approach was used as for the within-lysosome group, where OPC surfaces were reconstructed and used to mask the mCherry and eGFP channels. In Fig. 2f, only mCherry surfaces are shown and they are pseudocolored white or magenta if they overlapped with an eGFP surface or not, respectively.

AAV injections

Mice aged between 2 and 6 months were injected with meloxicam (2.5 mg kg−1, subcutaneous) before being anesthetized using isoflurane (SomnoSuite, Kent Scientific; 3–5% induction, 1–2% maintenance). Once anesthetic depth was achieved, mice were placed onto a stereotaxic apparatus where body temperature was maintained using a heating pad. Mice were then unilaterally injected with either AAV9-hSYN-eGFP (Addgene viral prep 50465-AAV9) or AAV9-hSYN-pSynDig (500 µl with a flow rate of 50 nl min−1) into the right hemisphere dLGN (x = 2.15, y = −2.15 and z = −2.9 mm from bregma). Following surgery, animals were administered Flunixin (10 mg kg−1) and allowed to recover on a heating pad before returning to their home cages.

Chronic window implantation

Mice (n = 10) aged between 2 and 6 months, previously injected with meloxicam (2.5 mg kg−1, subcutaneous) were anesthetized using isoflurane (3–5% induction, 1–2% maintenance), and body temperature was maintained with a heating pad throughout surgery and during initial recovery. After the initial AAV injection (as described above), a craniotomy of >3 mm in diameter was drilled using a dental drill (RWD, 78001) over V1 at approximately +2.5 mm lateral and −2.9 mm posterior from bregma. A 3-mm glass coverslip, sterilized with 70% ethanol, was then placed over the craniotomy and a mixture of surgical glue (Vetbond, 3M) and cyanoacrylate glue was used to secure the coverslip onto the skull. The skull was covered with a thin layer of Vetbond and then sealed with dental cement (Ortho-Jet, Land Dental). Finally, a custom-made head bar was secured onto the skull using luting cement (Metabond, C&B). Mice were then administered flunixin meglumine (10 mg kg−1, intraperitoneal), allowed to recover on a heating pad until ambulatory and then allowed to recover for 1 to 2 weeks before imaging. After recovery, the quality of the windows was checked before imaging, and mice with suboptimal windows were killed and used for downstream immunofluorescence quantification in fixed tissue.

In vivo two-photon imaging

For in vivo imaging experiments, mice were secured into a custom head mount and movement restraint system before being placed onto the two-photon microscope. Mice were imaged using a custom two-photon system (Independent Neuroscience Services) with a ThorLabs tunable Tiberius laser. Laser wavelength was tuned to 980 nm to image both tdTomato and eGFP concurrently. We used a ×16/NA 0.8 water immersion lens (Nikon), and light was captured using two photomultiplier tubes fitted with filters (520–565 and >565 nm) for eGFP and tdTomato, respectively. While imaging at a depth between 150 and 350 μm from the pia, the laser power was kept below 30 mW to avoid photodamage. Imaging volumes were captured at near Nyquist settings (either 512 × 512 or 1,024 × 1,024 for time-lapse and single-time-point recordings, respectively; resulting in voxels ≤264 × 264 × 1,000 nm) and were selected for fields with tdTomato+ cells with distinguishable OPC morphology and TC inputs. For time-lapse recording, volumes were taken once per minute. Raw image files were then processed using Huygens software (SVI) for cross-talk correction, followed by registration along the z and t dimensions and processing with the multiphoton deconvolution wizard (set to conservative deconvolution as previously described for pSynDig experiments). Cross-talk-corrected, registered and deconvolved images were then imported into Imaris and Surface objects of OPCs and TC inputs were created, with surfaces of OPCs being limited to the OPC soma and major processes. TC inputs were then classified by their distance to the OPC’s surface using the ‘shortest distance to surfaces’ filter function in Imaris, with TC inputs that were greater than 0 nm from the OPC surface classified as noncontacting inputs, inputs with a distance to OPCs of 0 nm classified as contacting inputs and inputs with a distance of less than −270 nm classified as ‘engulfed’ inputs. The classified thalamocortical inputs’ volumes were then averaged per video and per time frame and normalized to the average volume of noncontacting inputs over the entire imaging period to avoid bleaching effects where applicable. For data presentation, the OPC surface volume was pseudocolored with green and classified input volumes were pseudocolored using the ‘overlapped volume ratio to surface’ function.

Note on identifying OPCs in vivo

Because we typically imaged NG2-CreERT2tdTomato mice more than 2 weeks after administration of TAM, some of the Cre-expressing OPCs had differentiated to mature oligodendrocytes labeled with tdTomato by the time of imaging. However, OPCs could be easily distinguished from oligodendrocytes based on their oblong, bean-shaped somata compared to oligodendrocytes, the somata of which were more spherical (Extended Data Fig. 1).

Cell isolation and flow cytometry

Mice were first anesthetized with isofluorane and subsequently perfused transcardially with ice-cold PBS. Brains were then removed from the skull and the cortices were dissected and chopped into ~3-mm pieces for overnight enzymatic digestion with 0.5× Accumax (Thermo Fisher Scientific, SCR006). The tissue was further homogenized in a buffer solution (150 mM HEPES, 1× HBSS, 1% BSA, 2 mM EDTA and 5% glucose) by gentle pipetting using a 1-ml cut pipet-tip, and once again with a 1-ml uncut pipet-tip. OPCs and microglia were then enriched using a 40% Percoll in HBSS solution and centrifuged at 600g for 25 min. The cell pellet was washed and incubated with a CD16–CD32 receptor blocking antibody at 1:100 (Thermo Fisher Scientific, 14-0161-82) for 10 min. Next, the glial cells were stained with viability dye live/dead aqua stain at 1:1,000 (Thermo Fisher Scientific, L34957), and antibodies were directed against cell surface proteins for 30 min. The OPC population was identified using A2B5-AF488 at 1:100 (Thermo Fisher Scientific, FAB1416G) in combination with CD140a–PDGFRA-PE-Cy7 at 1:100 (BioLegend, 135912), and microglia were selected based on the expression of CD11b-PerCP-Cy5.5 at 1:100 (BioLegend, 101227) and CD45-Pacific Blue at 1:100 (BioLegend, 157211). To stain the intracellular synaptic material, the cells were first fixed with 1% PFA in HBSS for 10 min at room temperature, permeabilized with FoxP3/TRN (Life Technologies, 00-5523-00) staining buffer set according to the manufacturer’s protocol and stained with the following antibodies: SNAP25-AF 647 at 1:100 (BioLegend, 836311) or SYN1-AF 647 at 1:100 (Cell Signaling, 11127S). OPCs and microglia were analyzed with BD Dual Fortessa, and the results were analyzed using CytoexploreR24. For the gating strategy details, see Extended Data Fig. 6.

Blinding

Experimenters were blinded to conditions for quantitative imaging experiments. One experimenter collected the tissue and assigned it a randomized label before providing the blinded tissue to another experimenter for analysis. After data acquisition and processing, the data were plotted in GraphPad by L.C., A.F. or Y.A. after which the samples were unblinded.

Statistics and reproducibility

For all imaging datasets, we analyzed a minimum of three images over a set of three mice. For data computed based on the individual cells, multiple cells were taken from a minimum of three images per mouse (and a minimum of three mice per condition where applicable) and used as representative data points. Littermate controls were used whenever possible, for example, in the PLX experiments. For in vivo imaging, a minimum of three mice were used for analysis to create representative OPC volumes. Experimental mice were subject to exclusion based on window integrity as well as the signal-to-noise ratio (SNR) of fluorescent protein emissions. Both single volumes and time-lapse volumes were subject to exclusion based on the motion artifact as well as overt changes in SNR over the imaging session. Three independent replicates, with one mouse per condition within a given replicate, were used for flow cytometry datasets. Each independent replicate was analyzed separately and then pooled for data presentation.

Sample sizes and the number of mice/cells used were based on means and standard deviations found within our own lab and by others in similar experiments2. No statistical methods were used to predetermine the sample size.

All data are presented as the n representing individual cells or mouse averages, with exact n information in the figure legends. Parametric tests were used for Gaussian datasets, with nonparametric tests used when parametric tests were not applicable. Before unblinding and finalizing analyses, we removed outliers using the ‘Identify Outliers’ function in GraphPad (using the robust regression and outlier removal (ROUT) method with Q (maximum desired false discovery rate) = 1%). All statistical analyses were performed in GraphPad by L.C., A.F. or Y.A. with details described in the figure legends.

Figure production

Representative images were rendered in either Imaris or ImageJ, where contrast and brightness were altered across the entire image for ease of viewing.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.