Search
Close this search box.

Developmental origin of oligodendrocytes determines their function in the adult brain – Nature Neuroscience

Transgenic mouse lines

Sox10-loxP-eGFP-polyA-STOP-loxP-tdTomato (hereafter abbreviated as ‘Sox10-GFP-tdTomorTom’) reporter mice15 express enhanced GFP in all OLCs in the absence of Cre activity. To differentially label vOLCs and dOLCs in the forebrain, Sox10-GFP-tdTom mice were crossed to Emx1-Cre mice (hereafter abbreviated as ‘Emx’)13. To study the functional importance of dOLCs in the adult brain, dorsally derived, Emx1+ OLCs were ablated by crossing the Emx1-Cre line to the Sox10-loxP-GFP-poly(A)-STOP-loxP-DTA (hereafter abbreviated as ‘Sox10-DTA’ or ‘DTA’) mouse line13. To determine the ablation efficiency, all three mouse lines were crossed (hereafter abbreviated as ‘Emx/Tom/DTA’), allowing the visualization of the remaining dOLCs after ablation.

All mice were on a C57BL/6-B6CBA mixed background. Breeding colonies were maintained by mating transgenic mice with wild-type B6CBA-F1 mice. Mice were fed a standard diet and were kept on a 12-h dark/light cycle in individually ventilated cages. Mouse cages were kept at 45–65% humidity and at a temperature range of 20–24 °C. Experiments were performed on either neonatal (P0) or adult (11–14 weeks) animals. In case of the behavioral test, experiments were started at the age of 12 weeks. Mice ages are clearly stated in the text and respective figure legends. For cognitive behavior experiments, only male mice were used. For all other experiments, animals from both sexes were used. All animal experiments conformed to the UK Animals (Scientific Procedures) Act of 1986 and were approved by the Cambridge University local ethics committees before licensing by the UK Home Office.

Behavioral tests

Open field test (general vigilance)

Individual mice were placed in the middle of an opaque, white plastic box (width, 50 cm; length, 50 cm), and their behavior was video-recorded for 5 min. The time spent close to the walls versus the center of the testing chamber was quantified.

Rearing test (general vigilance)

Individual mice were placed in an open-top, clear plastic cylinder (diameter, 15 cm; height, 30 cm), and their behavior was video-recorded for 1 min. The number of rears (both freely and against the wall) was counted.

Hanging bar test (forelimb strength)

For the hanging bar test, a metal bar (diameter, 6 mm) was elevated 30 cm above the table top. Individual mice were positioned on the bar in such a way that only the front paws grabbed onto the bar. If a mouse failed to grab the bar due to bad placement, the experiment was repeated after a brief rest. If a mouse fell within 5 s (not due to poor placement), the experiment was repeated up to three times. Mice were tested on three alternate days (one run per session). The time at which the mice fell off or touched the frame was scored. The following scoring system was used: Score 1 = 1–5 s, Score 2 = 6–10 s, Score 3 = 11–20 s, Score 4 = 21–30 s and Score 5 = >30 s.

Rotarod (locomotor coordination)

Individual mice were placed on a rotarod (Ugo Basile) facing forward. The rotarod was operated in acceleration mode, consistently accelerating from 4 r.p.m. to 40 r.p.m. over the course of 5 min. The speed at which a mouse fell off and the reason for the end of the trial (falling, jumping or passive rotation) were recorded.

Balancing beam (locomotor coordination)

For the balancing beam test, a metal beam (length, 1 m; diameter, 10 mm) was elevated 50 cm above the table top. Mice were trained to cross the beam independently once a day (2–3 test runs per session) for 1 week. Subsequently, mice were tested on three alternate days (two runs per session). All test runs were video-recorded. The time a mouse needed to cross the beam (time a mouse stood still on the beam was not considered) and the number of foot faults were counted.

Horizontal ladder (locomotor coordination)

For the horizontal ladder test, a ladder (length, 1 m; width, 10 cm) with evenly spaced rungs was elevated 50 cm above the table top. Mice were trained to cross the ladder independently once a day (2–3 test runs per session) for 2 weeks. On the test day, three runs of each mouse were video-recorded. The time a mouse needed to cross the ladder (time a mouse stood still on the ladder was not considered), the number of steps taken with one hindlimb and the number of slip-offs or misses of paw placement were counted.

Vertical beam (bradykinesia)

For the vertical beam test, a metal beam (length, 1 m; diameter, 20 mm) was embattled perpendicular to the table top. The vertical beam test does not require any training. The mice were tested on three alternate days (two runs per session). All test runs were video-recorded. The number of steps taken with the hindpaws were counted.

Gait analysis (gait)

The paws of mice were colored with ink to trace their gait on white paper. Mice were trained to walk continuously through a tunnel over white paper once a day (2–3 test runs per session) for 2 weeks. Subsequently, mice were tested on three alternate days (one run per session). Stride length and stride width were measured. Measurements were taken only if the animal walked at a constant speed for at least three body lengths (approximately 30 cm).

Von Frey test (sensation)

One hindpaw of individual mice was touched with continuous force using the electronic Von Frey apparatus (Ugo Basile) until the mouse pulled away the paw. The maximum force needed to provoke a reaction to the stimulus was recorded. From each animal, five measurements were taken.

rCPT (sustained attention)

Testing was performed in a touchscreen operant chamber for mice (Campden Instruments) as previously described46. After habituation, mice were trained to discriminate a visual ‘target’ stimulus from multiple ‘non-target’ stimuli in four stages as previously described24. After successfully completing the training stages, the performance of mice in the rCPT was assessed using the following basic parameters: stimulus set, one ‘target’ and four ‘non-target’ stimuli; SD time, 2 s; inter-trial interval, 5 s. Subsequently, to increase the load on attention, the SD (amount of time a stimulus was shown on the screen) was shortened to 1.5 s, 1.0 s and 0.5 s.

Four main measures of task performance were analyzed: response to the ‘target’ stimulus (hit), no response to ‘target’ stimulus (miss), response to ‘non-target’ stimulus (false alarm) and no response to ‘non-target stimulus (correct rejection). From these measures, the HR and FAR were calculated as follows:

$$begin{array}{l}{rm{HR}}=frac{{rm{Hit}}}{{rm{Hit}}+{rm{Miss}}}qquadqquad{rm{FAR}}=frac{{rm{False}}; {rm{alarm}}}{{{rm{False}}; {rm{alarm}}}+{{rm{Correct}}; {rm{rejection}}}}end{array}$$

Additionally, based on the HR and FAR, the sensitivity (d′), referring to the behavioral discrimination between ‘target’ and ‘non-target’ stimulus, and response bias (c), were calculated as follows:

$$begin{array}{l}{rm{d}}{rm{mbox{‘}}}={rm{z}}({rm{HR}})-{rm{z}}({rm{FAR}})qquadqquad{rm{c}}=-frac{{rm{z}}left({rm{HR}}right)+{rm{z}}({rm{FAR}})}{2}end{array}$$

c reflects the decision criterion used by the animal when deciding whether or not to respond. Thus, a low c indicates a bias of the animal to judge a stimulus as a target; a high c indicates a bias of the animal to judge a stimulus as a non-target.

Reversal learning

Reversal learning was carried out as previously described47. In brief, animals were trained to discriminate between a ‘target’ visual stimulus and a ‘non-target’ visual stimulus that were presented simultaneously side by side in two touchscreen windows. The correct response to the ‘target’ stimulus was always reinforced, whereas the response to the ‘non-target’ stimulus was never reinforced. Stimulus location (left or right) was pseudorandomly determined with the same spatial configuration never presented more than three times in a row. When all animals achieved a high level of performance (>80% correct choice for one session), these contingencies were reversed so that the previously correct ‘target’ stimulus became the incorrect ‘non-target’ stimulus and vice versa. Animals were tested daily on this reversal learning stage until performance reached higher than 80%. Key performance variables were the percentage of correct responses to the ‘target’ stimulus and the number of correction trials.

Extinction learning

Extinction learning was carried out as previously described46. In brief, all animals were initially trained to emit responses to a white stimulus located in the central window of a ‘3-choice’ Perspex mask. All responses were reinforced at this stage. Subsequently, extinction learning was assessed by identical task presentation but with omission of reinforcement. Data are expressed as the percentage of responses to a stimulus per session.

scRNA-seq and snRNA-seq

Tissue dissociation and scRNA-seq

Neocortices of P90 Emx/tdTom (n = 4) and Emx/Tom/DTA (n = 2) were microdissected, and the tissue was dissociated into single-cell suspensions, as previously described29. In brief, mice were transcardially perfused with ice-cold oxygenated artificial cerebrospinal fluid (22 mM NaCl, 0.63 mM KCl, 0.4 mM NaH2PO4 × 2H2O, 6.5 mM NaHCO3, 25 mM saccharose, 5 mM glucose, 0.5 mM CaCl2, 4 mM MgSO4, pH 7.3), and the brains were collected. The tissue was sectioned at the vibratome in ice-cold artificial cerebrospinal fluid, and the CC was microdissected. Tissue dissociation was performed with an Adult Brain Dissociation Kit (Miltenyi Biotec) following the manufacturer’s instructions (the red blood cell removal step was not included). After debris removal, the cells were resuspended in ice-cold 1% BSA in artificial cerebrospinal fluid, filtered with a 30-μm filter (Sysmex Partec) and FACS sorted with a BD Influx system (BD Biosciences) to separate GFP+ and TdTom+ OLCs. An scRNA-seq experiment set with a cohort of control and experimental mice was discarded due to strong batch effects, most likely due to stress during the FACS procedure.

The sorted cells were processed with the Chromium Single Cell A Chip kit v2 and library prep with the Chromium Single Cell 3′ Library & Gel Beads kit v2 (10x Genomics) according to the manufacturer’s instructions. A total of 3,000 cells for each sample were loaded on the Chromium Single Cell A Chip, although a lower number of cells was recovered in singlet and passed the quality control.

Tissue collection and preparation for snRNA-seq

To assess the impact of dOPC ablation immediately when ablation occurs, cortical tissue was analyzed using snRNA-seq of antibody-hashed samples26. Neonatal mice (Emx/tdTom (n = 2) and Emx/Tom/DTA (n = 2)) were euthanized on the day of birth by decapitation, and then the motor and cingulate cortex was dissected in ice-cold oxygenated cutting solution containing 87 mM NaCl, 2.5 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 0.5 mM CaCl2, 4 mM MgSO4, 75 mM sucrose and 20 mM glucose, before samples were snap frozen on dry ice in a sterile microtube. The samples were stored at −80 °C until further processed. To isolate nuclei, tissue fragments from two animals with identical genotype were combined in nuclei isolation buffer (250 mM sucrose, 25 mM KCl, 5 mM MgCl2, 10 mM Tris buffer, pH 8.0, 0.1% NP40, 0.1 mM DTT and 0.4 U μl−1 RNAse inhibitors) and gently homogenized with a plastic pestle on ice. Subsequently, the homogenate was filtered using a 40-μm strainer, and the nuclear pellet was collected upon centrifugation. The pellet was then washed twice with Nuclei Wash and Resuspension (NWR) buffer containing 2% BSA in PBS and 0.4 U μl−1 RNAse inhibitors. Finally, the nuclei were resuspended in ST Staining Buffer (ST-SB: 2% BSA, 0.02% Tween 20, 10 mM Tris, pH 7.5, 146 mM NaCl, 1 mM CaCl2 and 21 mM MgCl2), filtered again with a 40-μm strainer and collected for inspection and counting. Hashtag antibody tagging was done as described previously26 with minor adjustments. Two sets of ablated and two sets of control nuclear samples (1 M nuclei per 100 μl of ST-SB) were first blocked with TruStain FcX (anti-mouse CD16/32, BioLegend, 101319) before the addition of 1 μg per 100 μl of Mab414 nuclei hashing antibodies (TotalSeq A0451-54, BioLegend, 682205, 682207, 682209 and 682211). To remove the unbound antibodies, samples were washed twice with ST Wash Buffer (10 mM Tris, 146 mM NaCl, 1 mM CaCl2, 21 mM MgCl2, 2% BSA, 0.4 U μl−1 of RNAse inhibitor) and again with NWR buffer to avoid the detergent and magnesium carry-over before sequencing. Finally, the nuclei were resuspended in NWR and strained with a 20-μm and a 10-μm strainer before counting, sample pooling and downstream processing with the 10x platform for single-nuclei sequencing.

snRNA-seq bioinformatic analysis

Feature selection and antibody demultiplexing

Cell Ranger software (10x Genomics standard pipeline, version 3.1.0) was run twice on the single-nucleus dataset, first to unravel pre-mRNA expression of genes supported by mm10 genome and second to associate barcoded antibodies to cells, given as a list of features. Intersection of cell barcodes from both modalities was retained. Cells expressing more than 600 genes, with unique molecular identifiers (UMIs) higher than 1,000 and lower than 30,000, were selected. Moreover, cells with more than 1% of RNA counts being associated to mitochondrial environment were removed. After a centered log-ratio transformation, sample of origin assignation was performed for each cell by the HTODemux function from Seurat version 4.0, with a positive quantile value to consider a cell positive set at 0.99. Among the 35,820 valid cells processed by Cell Ranger, 14,294 were targeted by more than one unique antibody barcode; 10,474 cells were classified as negative, not reaching the positive quantile threshold; and 11,052 cells considered as barcoded by a unique antibody were saved for downstream analysis. A total of 5,198 cells were retained for the ablated condition (two antibodies labeling the ablated samples, 2,218 cells and 2,980 cells), and 5,854 cells were selected for the control condition (two antibodies labeling the control samples, 2,639 cells and 3,215 cells).

Clustering and label transfer

To cluster the remaining cells by features expression, dimensionality reduction was performed on log-transformed and scaled UMI counts of the 2,000 most variable features. After selection of the first 30 principal components by the elbow method, clustering was performed by a shared nearest neighbor graph based on k-nearest neighbors (KNN) calculated by Euclidean distances. Modularity optimization was done by the Louvain algorithm at a resolution of 0.6. The 19 clusters highlighted by this method were then analyzed and compared to previous single-cell transcriptomic datasets of developmental mouse brain48,49. Label transfer was performed with canonical correlation analysis using the Seurat functions FindTransferAnchors and TransferData. Final annotation of clusters was done in accordance with the outputted prediction scores from label transfers.

Markers

To unveil highly variable genes in each cell type, the Wilcoxon rank-sum test was used throughout FindAllMarkers function from Seurat, where significant genes (adjusted P value (Padj) < 0.01), expressed in at least 25% of cells in each population and with a log2 fold change higher than 0.25, were retained. Finally, the top 15 genes genes by cell type ranked by log2 fold change were selected to draw the heatmap.

Frequency analysis

For the sake of condition library size and cell type composition balance, the number of cells per condition and cell type was normalized to a percentage of each cell type per condition, divided by the number of conditions, allowing the assessment of cell representation across condition and cell type. We confirmed our findings by applying a Bayesian model for compositional single-cell data, scCODA version 0.1.8 (ref. 27), which highlighted no credible effects of the DTA perturbation on any of the cell population balance. Astrocyte population was selected as the reference cell type, which is assumed to be unchanged in absolute abundance, and parameters inference was calculated via Hamiltonian Monte Carlo (HMC) sampling.

Effect of experimental perturbation

To investigate whether each independent ablated cell presented hallmarks of transcriptional perturbation, we estimated the relative likelihood of observing each cell in each experimental condition. We computed the effect of experimental perturbation using MELD software version 1.0.0 (ref. 28). Following the guidelines presented in ref. 26, we performed L1 normalization on the raw counts for each cell, followed by a non-negative square root transformation. The potential of heat diffusion for affinity-based trajectory embedding (PHATE)50 was then drawn, with a KNN at 10 and an alpha decay parameter at 10, after principal component analysis (PCA) on the transformed counts. After benchmarking the optimal parameters to run the core MELD function, a beta at 67, a KNN at 7 and a mean square error at 0.001162 were used to generate the MELD graph from which estimation of densities of each sample were drawn. The densities were then normalized over sample to calculate the relative likelihood associated to each sample.

scRNA-seq bioinformatic analysis

Feature selection

The count table from the output of the Cell Ranger software was obtained as part of the standard pipeline of the 10x Genomics platform (version 3.1.0). Cells expressing more than 200 genes but fewer than 3,000 were retained. Additionally, cells with more than 15% of RNA counts being associated mitochondrial origin were removed. Counts were normalized using the sctransform function in Seurat version 3.0 (refs. 51,52).

Clustering and label transfer

To cluster the cells, PCA was performed on the 3,000 most variable genes, and components were selected using the elbow method, selecting the first 30 principal components for further clustering. Cells were clustered using the Leiden algorithm in Seurat, with a resolution parameter of 0.6. To label the 11 clusters, we matched the labels to previous literature, where the backSPIN algorithm was used for clustering29. Label transfer was ultimately performed using the FindTransferAnchors and TransferData functions. With the clusters now labeled to their best matching cluster from the literature, our clusters were named accordingly. To integrate the data, the data were individually normalized per sample, and the FindTransferAnchors function was used to integrate the data.

Markers

The Wilcoxon rank-sum test was used to calculate the significant genes in relation to the obtained clusters. Significant genes (P < 0.01) were then selected. For each cluster, an enrichment score (z-score) was calculated using the FindAllMarkers function in Seurat, with a cutoff of 0.25 fold change.

Frequency analysis

The percentages of cells in the respective samples belonging to each cluster were calculated. The frequency graphs were then plotted as a percentage of the total captured cells in each sample, allowing us to make assessments on whether a certain sample is overrepresented or underrepresented compared to the other samples.

Differential gene expression analysis and Reactome pathway enrichment analysis

All comparisons were performed using the Wilcoxon rank-sum test with a Bonferroni-adjusted P value. Significantly differentially expressed genes with P < 0.01 and fold change > 0.1 were selected. Reactome pathways were calculated for each comparison using the R package ReactomePA53.

Effect of experimental perturbation

Similar to the snRNA-seq, we calculated the relative likelihood estimate of observing each cell in each experimental condition using MELD version 1.0.0. The same transformations were used on the raw counts to generate the PHATE projection, and the following parameters were drafted to generate the MELD graph (beta = 30, KNN = 8, mean square error = 0.004893). The densities were again normalized over sample to calculate the relative likelihood associated to each sample. For each cell type, we investigated the likelihood heterogeneity using vertex frequency clustering (VFC). We focused on two heterogeneous clusters, MOL5/6a and MOL5/6d, both of which were broken down in two populations according to their perturbated likelihood.

Immunohistochemistry

Mice were terminally anesthetized and fixed by intracardiac perfusion using 4% (w/v) paraformaldehyde (PFA) in 0.01 M PBS. Tissue was post-fixed in 4% PFA for 2 h at 20–25 °C, cryoprotected with 20% (w/v) sucrose for 24–48 h, embedded and frozen in OCT medium and stored at −80 °C. Tissues were sectioned at 10 μm and collected onto poly-l-lysine-coated glass slides.

Tissue sections were incubated with blocking solution (10% (v/v) normal goat serum and 0.1% (v/v) Triton X-100 in PBS) for 1 h at 20–25 °C. In case antibodies raised in mouse were used, tissue sections were incubated with the MOM kit according to the manufacturer’s protocol (Vector Laboratories). In case the antigen required antigen retrieval for detection, slides were placed in antigen retrieval buffer solution (Dako, S2369, diluted 1:10 in distilled water), which was pre-heated to 95 °C, and slides were incubated for 10 min at 75 °C. Subsequently, tissue sections were incubated with primary antibodies overnight at 4 °C and then secondary antibodies for 1 h at 20–25 °C.

The following primary antibodies were used: CC-1 (anti-adenomatous polyposis coli (APC)) (OP80, mouse, 1:100, Calbiochem), CASPR1 (ab34151, rabbit, 1:1,000, Abcam), CD68 (MCA1957, rat, 1:200, Serotec), GFP (GFP-1020, chicken, 1:1,000, Aves Labs), GFAP (Z0334, rabbit, 1:1,000, Dako), IBA1 (019-19741, rabbit, 1:1,000, Wako), MBP (MCA4095, rat, 1:100, Serotec), NeuN (MAB377, mouse, 1:50, Millipore), tdTomato (AB8787-200, goat, 1:300, Sicgen Antibodies), OLIG2 (AB9610, rabbit, 1:1,000, Merck) and Ankyrin G (N106/36, mouse, 1:200, Antibodies Incorporated). The following secondary antibodies were used: anti-chicken Alexa 488 (donkey, 1:500, Jackson Laboratory), anti-goat Alexa 568 (donkey, 1:500, Thermo Fisher Scientific), anti-rabbit Alexa 647 (donkey, 1:500, Thermo Fisher Scientific), anti-rat Alexa 647 (donkey, 1:500, Thermo Fisher Scientific) and anti-mouse Alexa 647 (donkey, 1:500, Thermo Fisher Scientific). Cell nuclei were visualized by staining with Hoechst 33258 (1:10,000, Biotium). All images were acquired using an SP5 confocal microscope (Leica), except for the CASPR1 staining, which was imaged using an Airyscan microscope (Zeiss) to increase resolution. Cell counts were obtained from 3–6 biological replicates, 3–4 tissue sections per biological replicate. ImageJ (version 2.0.0-rc-68/1.52h and 1.53t), CellProfiler (version 2.2.0) and CellProfiler Analyst (version 2.2.1) were used for image analysis, cell counting and measurement of the size of the selected area.

In situ hybridization

Assessment of OPC numbers

To visualize the distribution of OPCs, in situ hybridization with chromogenic RNAscope was performed. Tissue sections were incubated with RNAscope hydrogen peroxide reagent (ACDBio) for 10 min at 20–25 °C. Subsequently, tissue sections were permeabilized using RNAscope target retrieval reagent (ACDBio) for 5 min at 99–100 °C, followed by protease treatment using RNAscope Protease Plus (ACDBio) for 30 min at 40 °C. Probe hybridization, signal amplification and signal detection were performed using an RNAscope 2.5 HD Detection Kit (BROWN). The procedures were carried out according to the manufacturer’s instructions, except for the ‘AMP 5’ step, in which tissue sections were incubated for 45 min at 20–25 °C. The following probe was used: Pdgfra (mouse, 448431, ACDBio). Cell nuclei were visualized by staining with Hoechst 33258 (1:10,000, Biotium). Cell counts were obtained from five biological replicates, 3–4 tissue sections per biological replicate. ImageJ was used for image analysis, cell counting and measurement of the size of the selected area.

Assessment of neuronal subtype distribution and density at P90

Neuronal populations of the adult (P90) mouse cerebral cortex were characterized by single-molecule fluorescence in situ hybridization (smFISH) and immunohistochemistry for the neuron-specific marker RBFOX3 (NeuN). Three to five biological replicates were analyzed per genotype, with 2–4 whole-brain cryosections (coronal, 16 µm) per animal. As described previously54, smFISH staining was performed using an RNAscope LS Multiplex Fluorescent Assay (ACDBio; ref. 55), which was automated together with RBFOX3/NeuN immunohistochemistry on a BOND RX robotic stainer (Leica Biosystems), followed by spinning disk confocal imaging on an Operetta CLS high-content screening microscope (PerkinElmer) and image analysis using Harmony software (PerkinElmer).

RNAscope catalog probes (https://acdbio.com/catalog-probes) were assigned to four different channels (C1–C4; Supplementary Table 3) and combined in individual BOND runs for 3-plex or 4-plex transcript detection. RNAscope probes against bacterial DapB mRNA as well as against mouse housekeeping genes were used as negative and positive controls, respectively (Supplementary Table 3). The following modifications were applied to the previously published protocol54: for heat and protease treatment, P90 samples were incubated in BOND ER2 buffer, pH 9.0 (Leica Biosystems), at 95 °C for 5 min and in ACD protease reagent at 42 °C for 20 min. The probes in C1, C2 and C3 channels were labeled using Opal 520 (Akoya Biosciences (AB), FP1487001KT), Opal 570 (AB, FP1488001KT) and Opal 650 (AB, FP1496001KT) fluorophores, respectively, at a dilution of 1:1,500. The C4 probe complexes were developed with TSA biotin (AB, NEL700A001KT, 1:500) and streptavidin-conjugated Atto 425 (Sigma-Aldrich, 09260, 1:200). DAPI (Thermo Fisher Scientific, D1306, 0.25 μg ml−1) was used as nuclear counterstain. Slides were coverslipped using ProLong Gold Antifade Mountant (Thermo Fisher Scientific, P36930) and Precision cover glasses thickness No. 1.5H (Paul Marienfeld GmbH & Co. KG, 0107222).

For spinning disk confocal imaging, regions of interest (ROIs) for high-resolution ×40 scans, covering the cerebral cortex across individual hemispheres, were manually selected on low-magnification whole-tissue section ×5 scans. ×5 scans were acquired in wide-field mode with a numerical aperture (NA) 0.16 air objective; ×40 scans were acquired in confocal mode with a ×40 NA 1.1 automated water-dispensing objective. ×40 fields were imaged as a z-stack of 21 planes with a 1-μm step size and with 8% overlap.

To segment single neurons and quantify mRNA spots from high-resolution images, analysis scripts were created on Harmony software (PerkinElmer). Individual neurons were segmented as described previously54 using either Synaptotagmin I (Syt1) or RBFOX3/NeuN to delineate neuronal cytoplasm. The number of Gad2 mRNA spots was calculated per Syt1+ single neuron for the identification of excitatory versus inhibitory neuron populations; the number of Bcl11b (Ctip2), Crym and Fezf2 mRNA spots was calculated per NeuN+ single neuron for the identification of corticospinal motor neuron (CSMN) populations56,57. For neuronal subtype identification, the lowest point of the density curve of the respective spot count signal distribution, plotted on a logarithmic scale, was used for setting the cutoff: 15 mRNA spots for Gad2; five mRNA spots for Bcl11b (Ctip2), Crym and Fezf2. For example, a NeuN+ neuron with more than five Fezf2 spots was classified as Fezf2+ and ≤5 Fezf2 spots as Fezf2.

Cortical layers and areas—that is, motor (MO) area and primary somatosensory (SSp) area—were manually annotated on coronal sections at the plane of the anterior corpus callosum. DAPI staining and layer neuron marker (Cux2, Bcl11b/Ctip2 and Foxp2) expression patterns were used as anatomic landmarks and the Allen Reference Atlas (P56, coronal, image 44)58 as anatomic reference. Anatomic annotation masks were drawn on tissue maps, which were created by stitching ×40 maximum z-projection images in OMERO (University of Dundee & Open Microscopy Environment). Single-neuron objects were annotated with cortical layer and area identity based on their xy position within anatomic annotation masks. For each cortical layer and area, the percentage (%) of each neuronal subtype across all Syt1+/NeuN+ neurons was calculated as well as neuronal subtype density (number of neurons per area (mm2)).

Assessment of neuronal subtype distribution at P0

For the assessment of neuronal subtypes in the P0 mouse cerebral cortex, the same protocol was followed as described above for P90, with the following exceptions: for heat and protease treatment, P0 samples were incubated in BOND ER2 buffer, pH 9.0, at 95 °C for 2 min and in ACD protease reagent at 42 °C for 10 min. Because of increased cell densities in the neonatal cerebral cortex, single cells were segmented from DAPI+ nuclear area plus a small cytoplasmic ring (1-µm perimeter around the DAPI+ nucleus), and RNA spot counts were evaluated in the segmented cell area. A cell with >5 Syt1 spots in the segmented cell area was classified as a neuron; a neuron with >5 Gad2 spots was classified as inhibitory and ≤5 Gad2 spots as excitatory; and a neuron with >5 Ebf1 spots was classified as Ebf1+.

RNA FISH

In situ four-color RNA FISH was performed using primer-padlock 2-oligo hybridization of the RNA targets, as described previously59, except that hybridization was performed in 30% formamide buffer (Merck). In brief, ssDNA oligo probes were designed against the coding regions of target RNA transcripts using Picky2.0 to identify loci without RNA secondary structure or repetitive sequence, followed by blast searching to confirm target specificity. Probe targets and sequences can be found in Supplementary Table 4. Probe hybridization was performed overnight at 40 °C with rocking in a humid chamber. Probes then underwent rolling circle amplification, polyacrylamide gel mounting and proteinase K (Sigma-Aldrich) sample clarification before imaging. Images were acquired on a Zeiss LSM 980 microscope.

Image analysis

Image analysis was performed in arivis Vision 4D, version 4.1.1. In the pipeline, nuclei were first segmented using DAPI staining. A 3-µm region grow was used to define the cell area. PLP+ cells were determined by mean intensity of GFP fluorescence per cell (threshold for positive cell: 400). Finally, individual Tppp dots larger than 0.5 µm were counted per cell. A cell with three or more Tppp dots was counted as positive.

SCoRe and quantification

SCoRe imaging was performed to assess myelination status in motor cortex layer 1 as described previously60,61. In brief, mice were perfused transcardially under terminal anesthesia with 25 ml of warm (30 °C) PBS, followed by ice-cold 4% PFA. The brain was removed from the skull, and each hemisphere was dissected out and flattened by clamming between two glass slides with 1-mm spacer and continuing to fix in 4% PFA at 4 °C for 6 h before changing to PBS. The flat-mounted hemisphere was examined using a Zeiss LSM 880 laser scanning confocal microscope. Then, z-stack images (1,024 × 1,024 pixels) with a 2-mm z-step were captured from the outer surface of the motor cortex with a ×20 water immersion objective, and four tiled images were stitched into a larger view of the region. To quantify, three random ROIs of a volume 300 × 300 × 40 mm were selected and background subtracted, and the myelin signal was segmented and converted to a binary image with the Triangle algorithm in ImageJ. The total volume and length (skeleton) were calculated using Volocity software (Quorum Technologies). Internode length was measured semi-automatically on the image stacks using the ImageJ plugin Simple Neurite Tracer by manually locating adjacent nodes of Ranvier along continuous segments of myelin. A total of 294 internodes in six controls and 248 in five ablated animals were measured.

Electron microscopy

Mice were perfused with 4% (w/v) glutaraldehyde in 0.01 M PBS. Tissue was post-fixed in 4% (w/v) glutaraldehyde at 4 °C overnight. The cortex was dissected, and the tissue was further processed in 2% osmium tetroxide at 4 °C overnight. The tissue was dehydrated in a series of ethanol washes (1× 70% EtOH for 15 min, 1× 95% EtOH for 15 min and 3× 100% EtOH for 10 min (Sigma-Aldrich)), washed twice in propylene oxide for 15 min and incubated in a 1:1 mix of propylene oxide and resin (50% resin, 34% dodecenyl succinic anhydride, 16% methyl nadic anhydride, 2% 2,4,6-Tris(dimethylaminomethyl)phenol (DMP-30), all (v/v), TAAB Laboratories) for at least 3 h at 20–25 °C. The tissue was incubated twice in pure resin for 12 h at 20–25 °C, and samples were embedded in plastic containers in fresh resin for 2 d at 60 °C. Ultra-thin sections of cortical layers 5–6 or CC were cut onto copper grids and stained with uranyl acetate and lead citrate (TAAB Laboratories). The sections were imaged with a H-600 Transmission Electron Microscope (Hitachi High-Tech). Subsequently, G-ratio (ratio of the inner axonal diameter to the total outer diameter) was measured from 100–500 axons per animal, 4–5 animals per age group and per genotype. The G-ratio measurement was carried out from images with magnification at ×3,000 or ×6,000, using the open-source software AxonSeg, with the built-in TEM model and artifact manually corrected62.

Immunogold electron microscopy

Samples were prepared essentially as described previously63. In brief, mice were perfused after anesthesia transcardially with a fixing solution composed of 0.25% glutaraldehyde, 4% formaldehyde and 0.5% NaCl in phosphate buffer pH 7 according to Schultz and Karlsson64. After dissection and preparing horizontal vibratome sections of 300-µm thickness, the ROI (motor cortex layers 5–6) was cut out with a biopsy punch. These pieces of tissue were infiltrated overnight in 2.3 M sucrose in 0.1 M phosphate buffer pH 7, mounted on aluminum pins and frozen in liquid nitrogen. Ultra-thin cryosections were prepared using a diamond knife (cryoimmuno, 35°, Diatome) and a UC6 ultramicrotome equipped for cryosectioning (Leica Microsystems). Ultra-thin sections were placed on hexagonal 100 mesh copper grids (Science Services) and labeled for GFP (polyclonal rabbit, 1:100, Invitrogen, Thermo Fisher Scientific, RRID: AB_221569) or tdTom (polyclonal rabbit anti-RFP, 1:100, Rockland) followed by incubation with protein A-gold (10 nm, Cell Microscopy Core, UMC Utrecht). Images were taken with a LEO912 transmission electron microscope (Zeiss Microscopy GmbH) using an on-axis 2k CCD camera (TRS).

Western blot

Whole-brain samples were lysed in IP lysis buffer (Thermo Fisher Scientific) supplemented with 0.1% Halt protease and phosphate inhibitor (Thermo Fisher Scientific). Tissue lysates were mixed with NuPAGE sample LDS loading buffer (Invitrogen) supplemented with 1× NuPAGE Reducing Agent (10x) (Invitrogen) and denatured for 10 min at 95 °C. Gel run and transfer were performed according to the manufacturer’s instruction for the mini gel tank (Thermo Fisher Scientific). After transfer, the nitrocellulose membranes were blocked in 1× Odyssey blocking solution (TBS, LI-COR) in TBS-T (TBS + 0.1% Tween (Sigma-Aldrich)) for 1 h at 20–25 °C on the shaker. Subsequently, nitrocellulose membranes were incubated with primary antibodies overnight at 4 °C and then secondary antibodies for 1 h at 20–25 °C. The following primary antibodies were used: β-ACTIN (A5441, mouse, 1:5,000, Sigma-Aldrich), GFAP (Z0334, rabbit, 1:1,000, Dako) and GS (ab73593, rabbit, 1:1,000, Abcam). The following secondary antibodies were used: anti-mouse 680 (donkey, 1:15,000, LI-COR) and anti-rabbit 800 (donkey, 1:15,000, LI-COR). Fluorescent signals were detected using the Odyssey apparatus (LI-COR) with an exposure time of 2 min. Band intensities of the developed western blots were determined with Image Studio software (version 4.0, LI-COR).

Isolation of adult oligodendrocytes

Adult male and female mice (2–3 months) were decapitated after lethal injection with phenobarbital. The brains were removed quickly and placed into ice-cold Hibernate A (BrainBits). The meninges and the olfactory bulb were mechanically removed in Hibernate A, and the brain tissue was mechanically minced into 1-mm3 pieces. The tissue pieces were spun down at 100g for 1 min at room temperature, and the tissue was washed in HBSS (no Mg2+ and no Ca2+, Gibco). Each brain was mixed with 10 ml of dissociation solution (34 U ml−1 papain (Worthington) and 20 mg ml−1DNase Type IV (Gibco) in Hibernate A). The brain tissue was dissociated on a shaker (50 r.p.m.) for 40 min at 35 °C. The digestion was stopped by addition of ice-cold HBSS. The tissue was centrifuged (200g, 3 min, room temperature); the supernatant was completely aspirated; and the tissue was resuspended in Hibernate A supplemented with 2% B27 (Gibco) and 2 mM sodium pyruvate (Gibco) (trituration solution). The tissue was allowed to sit in this solution for 5 min. To obtain a single-cell suspension, the tissue suspension was triturated 10 times using first a 5-ml serological pipette and subsequently three fire-polished glass pipettes (opening diameter >0.5 mm). After each trituration step, the tissue suspension was allowed to sediment (approximately 1–2 min), and the supernatant (approximately 2 ml), containing the cells, was transferred into a fresh tube. After each round of trituration, 2 ml of fresh trituration solution was added. To remove accidentally transferred undigested tissue bits, the collected supernatant was filtered through 70-μm cell strainers into tubes that contained 90% isotonic Percoll (GE Healthcare, 17-0891-01, in 10× PBS, pH 7.2 (Life Technologies)). The final volume was topped up with phenol red free DMEM/F12 with HEPES (Gibco) and mixed to yield a homogeneous suspension with a final Percoll concentration of 22.5%. The single-cell suspension was separated from remaining debris particles by gradient density centrifugation (800g, 20 min, room temperature, without break). The myelin debris and all layers without cells were discarded, and the brain cell containing phase (last 2 ml) and the cell pellet were resuspended in HBSS+ and centrifuged (300g, 5 min, room temperature). The cell pellet was resuspended in red blood cell lysis buffer (Thermo Fisher Scientific, 15838518) and incubated for 90 s at room temperature to remove red blood cells. Next, 10 ml of HBSS+ was added to this cell suspension and spun down (300g, 5 min, room temperature). The cell pellets were resuspended in 0.5 ml of modified Miltenyi washing buffer (MWB, 2 mM EDTA (Thermo Fisher Scientific), 2 mM sodium pyruvate, 0.5% BSA in PBS, pH 7.3) supplemented with 10 ng ml−1 human recombinant insulin (Gibco). To this cell suspension, 2.4 μg of Mouse MOG Biotinylated antibody (BAF2439, BioTechne) in 100 μl was added for every 10 million cells. After 25-min incubation, gently shaking at 4 °C, 7 ml of MWB was added. The solution was centrifuged (300g, 5 min, room temperature), and the pellet was resuspended in 80 μl of MWB supplemented with 20 μl of Anti-Biotin MicroBeads (Miltenyi Biotec, 130-090-485) per 10 million cells. The cells were incubated for 15 min, slowly shaking at 4 °C. The secondary antibody was again washed out with 7 ml of MWB, and the sample was centrifuged (300g, 5 min, room temperature). The cell pellet was resuspended in 0.5 ml, and MACS was performed according to the recommendations of the supplier. In brief, an MS column (Miltenyi Biotec, 130-042-201) was inserted into a MiniMACS Separator (Miltenyi Biotec, 130-042-102) and pre-wet with 0.5 ml of MWB. Resuspended cells were put onto one MS column. Subsequently, the column was washed three times using 500 ml of MWB for each wash. Finally, MOG+ cells were flushed out the column with 0.5 ml of MWB. Cells were centrifuged (300g, 5 min, room temperature), and then the pellet was resuspended in 500 μl of TRIzol (Thermo Fisher Scientific, 15596026) and frozen at −80 °C until RNA isolation.

RNA isolation and RT–qPCR

RNA was isolated from freshly purified OLs according to the Direct-zol RNA MicroPrep Kit (Zymo Research, R2061). cDNA was generated using the Maxima First Strand Kit according to the manufacturerʼs instructions (Thermo Fisher Scientific; K1641). For RT–qPCR, the following primers were used: Pdgfra: FWD 5′-GCCTCCATTCTGGAGCTTGT-3′, REV 5′-CTCGATGTTCGGAAGAGGG-3′; Mag: FWD 5′-GGATAATGATTTCAGCTTCTCG-3′, REV 5′-GTAGGGACTATTGAAATACCAG-3′; Tppp: FWD 5′-ACGTGGACATTGTCTTCAGCA-3′, REV 5′-GAGGACACAGCTTTCGTGA-3′. All primers were used at a concentration of 400 mM. The efficiency of each primer was greater than 95% as determined for each primer pair by serial dilutions of OL cDNA. cDNA, primers and the Syber Green Master Mix (Thermo Fisher Scientific, A25741) were mixed as instructed by the manufacturer, and RT–qPCR and melting curve analysis were performed on a QuantStudio 7 Flex Real-Time PCR System (Life Technologies). Fold changes in gene expression were calculated using the Pfaffl method in Microsoft Excel.

Statistics

In all, the graphical data are presented as mean ± s.e.m. Statistical analysis was performed using GraphPad Prism (versions 9 and 10). The data were analyzed for normal distribution using D’Agostino–Pearson omnibus and Shapiro–Wilk normality test. To assess differences between two groups, the two-tailed unpaired Studentʼs t-test or the Mann–Whitney non-parametric test was used, depending on whether the data were distributed normally or not. When comparing more than two groups, a one-way ANOVA test was used followed by Tukey’s post hoc test. If the sample was not normally distributed, a Kruskal–Wallis test combined with Dunn’s post hoc test was used. In case a dataset had two or more variables, a two-way ANOVA was performed. If an interaction of the variables in the dataset was detected in the two-way ANOVA, the variables were analyzed separately by one-way ANOVA combined with Tukey’s post hoc test. If the sample was not normally distributed, a Kruskal–Wallis test (ranks) was carried out. To analyze data obtained by repeated measurements (for example, rCPT, reversal learning and extinction learning), the mixed-effect model was performed using the function ‘lmer’ from the R package ‘lme4’. If an interaction of the variables in the dataset was detected in the mixed-effect model, each group was analyzed separately by simple main effect model.

For behavioral studies, pilot experiments were run on which basis the sample size was determined by power analysis. For all other experiments, no statistical methods were used for sample size determination. All experiments were performed with three or more replicates, unless stated otherwise, which are similar sample sizes to those standard in the field. For cell quantifications, a sufficient number of cells per animal was counted (≥50) to ensure normal distribution around the mean. No animals or data points were excluded from the analyses reported in this paper. There was no randomization when animal and samples were assigned to the various experimental groups, because groups were determined by an animal’s genotype. Experimental conditions were the same for each group.

Data collection was not performed blinded apart from the behavioral experiments, where the animal identity was coded before testing. Data analysis was performed blinded to the conditions of the experiments and automated wherever possible to limit the influence of the experimenter on the outcome. In each figure legend, the number of technical replicates (N), the number of biological replicates (number of animals) (n), the statistical test and the significance levels are indicated. Datasets were considered significantly different at a P value of less than 0.05. Significance levels are presented graphically as follows: *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Reporting summary

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