Early life stress impairs postnatal oligodendrogenesis and adult behavior through activity-dependent mechanisms

Exposure to stress during early life (infancy/childhood) has long-term effects on the structure and function of the prefrontal cortex (PFC) and increases the risk for depression and anxiety disorders. However, little is known about the molecular and cellular mechanisms of these effects. Here we focused on changes in a mouse model of early stress induced by chronic maternal separation (MS). Unbiased mRNA expression profiles in the medial PFC (mPFC) of MS pups (P15) identified an increased expression of myelin-related genes and a decreased expression of immediate early genes. We further analyzed oligodendrocyte lineage markers and performed birthdating experiments to show a precocious oligodendrocyte differentiation in the mPFC of MS pups that led to a depletion of the oligodendrocyte progenitor pool in adults. Pharmacogenetic inhibition of neuronal activity in the developing mPFC of control animals induced a similar premature differentiation of the oligodendrocyte lineage, while the transient increase of mPFC activity in MS pups prevented this change. Moreover, the developmental diminution of mPFC activity mimicked the emotional and cognitive phenotypes observed in MS, while activating the mPFC during MS prevented the emergence of depression and cognitive disorders in adulthood. Altogether, our results identify an activity-dependent maturation process in the developing mPFC that is influenced by early life stress and has critical functions on postnatal oligodendrogenesis as well as adult emotional and cognitive behaviors.


INTRODUCTION
The environment a child encounters during early life plays a critical role in normal brain development. It has been repeatedly documented that early life stress (ELS), which includes various forms of child abuse and neglect, increases the risk to develop psychiatric disorders [1][2][3][4] . Accordingly, brain imaging studies in patients exposed to childhood adversity show altered brain activity and myelin content, notably affecting the medial prefrontal cortex (mPFC) which constitutes a critical hub for emotional control and cognition [5][6][7][8] . However, the developmental mechanisms underlying these long-term effects of early stress remain unclear.
To identify the pathophysiological mechanisms leading to ELS phenotypes, in particular as regards to mood and anxiety disorders, several preclinical models have been developed 9,10 . The most commonly used rodent model of ELS involves daily maternal separation (MS) during the early postnatal period, a time when pups normally spend most of their time huddled in the mother's nest 9,11 . Rodents exposed to MS during the first postnatal weeks display, as adults, a permanent impairment of emotional and cognitive behaviors, together with an altered physiology, cytoarchitecture, and myelination of the mPFC [12][13][14][15] . Strikingly, these behavioral and structural defects closely reproduce the phenotypes of humans exposed to childhood trauma 10,16 . Several of these alterations are already apparent at juvenile or adolescent stages 14,[17][18][19][20] but the causal mechanisms remain poorly understood.
A critical role of neuronal activity in cortical maturation has been demonstrated in the primary sensory cortical maps (visual, somato-sensory, auditory) 21,22 in which the activity of the incoming axons can be easily manipulated. In these classical model systems, neuronal activity has been shown to play critical roles, not only in the maturation of excitatory and inhibitory synaptic connections, but also in the onset of myelination that marks the stabilization of brain circuits. Activity thus appears as the mechanistic link between changes in the environment and alterations in the brain's developmental trajectory. We hypothesize that similar influence of neuronal activity likely exists in associative cortices such as the mPFC. Supporting this hypothesis, recent electrophysiological recordings and molecular analysis showed that maternal care modulates neuronal activity in mPFC 14,18,23,24 . However, the role of neuronal activity during critical periods of PFC development has not been tested yet.
In the present study, we focused on cellular and molecular changes occurring in the mPFC at the outset of ELS. We started with an unbiased transcriptome screen in the mPFC, comparing MS with standard facility raised (SFR) pups. Unexpectedly, we identified major changes in gene pathways associated with myelination and neuronal activity. Turning to cellular analyses, we documented a precocious differentiation of oligodendrocyte progenitor cells (OPC) in MS P15 pups, which coincided with an increase in myelin transcripts. Conversely, the OPC pool was depleted in adults. To determine the role of activity in these phenotypes, we manipulated locally mPFC neuronal activity in pups using designed receptors exclusively activated by designed drugs (DREADDs) 25 . We found that local and transient decrease in neuronal activity is sufficient to induce premature OPC differentiation, while the transient increase in neuronal activity prevents it. Additionally, we demonstrate that the emotional and cognitive disorders induced by ELS in adulthood are controlled by developmental changes in mPFC neuronal activity. Taken together, our results demonstrate a fundamental role of mPFC postnatal activity in the etiology of ELS behavioral phenotypes and identify oligodendrogenesis as a potential underlying mechanism in this process.

Early life stress (ELS) alters the expression of immediate early genes and genes involved in myelin formation in the developing medial prefrontal cortex (mPFC).
As a preclinical model of ELS, we used the common model of chronic maternal separation (MS), where pups are daily separated for 3h from their dam starting at postnatal day 2 (P2) and until P14 ( Figure 1A) 11,15 . To identify early molecular candidates potentially involved in the permanent alterations induced by ELS, we focused on the chronic effect of MS by analyzing P15 mPFC, sampled 24h after the last maternal separation, and we performed an unbiased genetic screen using high throughput RNA sequencing. To minimize inter-individual variations and dissection biases, mPFC of 3 males from different litters were pooled and 4 independent samples per condition were tested. After normalization by the RUVSeq R package v3.1 26 and using edgeR, we found a total of 86 genes differentially expressed between MS and SFR animals (adjusted p value <0.05 and change in transcript level >30%, Supplementary Table 1). Among the 56 up-regulated genes, Gene Ontology (GO) analyses revealed a major enrichment in biological processes related to myelin ensheathment and oligodendrocytes differentiation (red dots in Figure 1B and Supplementary Figure 1). Similar analyses of the 30 down-regulated genes displayed enrichment in cellular response to calcium (Supplementary Figure 1) likely due to the strong representation of several immediate early genes (IEG) (blue dots in Figure 1B). To validate these changes we performed quantitative PCRs on mPFC mRNAs obtained from a new cohort of MS and SFR animals analyzed at P15. This second cohort confirmed the decrease in IEG transcripts (Fos, Fosb and Arc, Figure 1C) and the increase in myelin-related transcripts (Mag, Mog and Plp1, Figure 1D). Altogether, this unbiased approach identified for the first time an early concomitant alteration of transcripts related to IEG and myelin ensheathment in the mPFC of postnatal animals exposed to chronic maternal separation.

Premature OPC differentiation in ELS pups.
To determine possible causes of the increase in myelin related transcripts, we investigated oligodendrogenesis in MS animals. The postnatal period corresponds to an intense proliferation of oligodendrocyte progenitors cells (OPC) and, P15 marks the initiation of differentiation into myelinating oligodendrocytesin the mPFC 20,27 . We performed triple immunostaining of Olig2, PDGFRα and CC1 to distinguish OPC (Olig2 + PDGFRα + , Figure 2A) from differentiated oligodendrocytes (OL, Olig2 + CC1 + ) 28 . At P15, we observed a significant increase in the density of OL in the mPFC of MS animals compared to SFR animals (F(1,7)=21.84 p=0.0023), while the density of OPC was unaltered (F(1,7)=0.032 p=0.863, Figure 2B). At adult stages, we observed a significant decrease in OPC density (F(1,8)=6.28 p=0.036), while no difference in OL was detected (F(1,8)=0.022 p=0.8865, Figure 2C). To test whether early differentiation contributes to these observations, sequential staining of the dividing progenitors was performed using two thymidine analogs at P10 (iodo-deoxyuridine, IdU) and P15 (bromo-deoxyuridine, BrdU, Figure  2D). This interval was chosen because OPC differentiate into OL in approximately 5 days 27 . In the mPFC of P15 MS animals, the fraction of OPC that kept dividing between P10 and P15 (triple labeled with Olig2 + IdU + BrdU + , blue bars in Figure 2E, F(1,5)=12.62 p=0.0163) was significantly reduced, while the differentiating fraction (Olig2 + IdU + BrdU -, purple bars, F(1,5)=7.234 p=0.0433) was increased. Thus, our results demonstrate that MS induces a precocious differentiation of the OPC pool during development, as a potential cause for depletion of the OPC pool in adults. To assess how oligodendrogenesis defects could be related with the alterations in IEG we previously observed, we performed immunostaining on P9 MS mPFC (Supplementary Figure 2). We noticed that the decrease in IEG expression preceded the defects in oligodendrogenesis thus suggesting a chronic down-regulation of neuronal activity in the MS pups. Considering the important role of neuronal activity in adult myelination 29 , we further tested how manipulation of neuronal activity over the early postnatal period could modify oligodendrogenesis.

Pharmacologic manipulation of mPFC postnatal activity bi-directionally controls developmental oligodendrogenesis.
We used viral delivery of AAVs expressing DREADDs to transiently and bi-directionally manipulate neuronal activity in the mPFC 25,30 . AAV8-hSyn-hM4Di-mCherry or AAV8-hSyn-hM3Dq-mCherry viral constructs were bilaterally injected into the mPFC of all P1 pups ( Figure  3A) and they further received daily injections of saline (SAL) or the inert ligand Clozapine-N-Oxyde (CNO) from P2 to P14. Pups injected with the inhibitory construct were raised in the SFR condition (referred to as SFR+hM4Di, Figure 3A). Conversely, pups injected with the activatory construct were submitted to the MS protocol (referred to as MS+hM3Dq). We first verified the timing and efficiency of expression. mCherry expression in cortical neurons was visible 36h after injection (at P2.5, data not shown) and at P15, 38%±5.2 of mPFC NeuN + cells were mCherry + upon hM4Di injections and 51.8%±4.4 upon hM3Dq injections. In addition, 95,98%±0.06 of the mCherry + cells were NeuN + , 72,59%±3.75 were Tbr1 + or Ctip2 + and only 6.17%±1.24 were GABA + ( Figure 3B-C). No co-expression of mCherry with Olig2 or PDGFRα or the oligodendrocyte-specific antibody CC1 could be detected (data not shown) supporting that viral infections essentially targeted neurons, with a bias for deep-layer glutamatergic neurons (Tbr1 + or Ctip2 +) .
To evaluate the efficiency of the DREADD constructs, we used ex vivo patch clamp electrophysiological recordings. We observed a robust increase in neuronal excitability upon exposure to CNO (30µM) or to its potential metabolite clozapine (1µM), in hM3Dq-mCherry + layer 5 pyramidal neurons from P10 mPFC (Supplementary Figure 3). In hM4Di-mCherry + neurons, exposure to CNO or clozapine did not modify neuronal excitability (Supplementary Figure 3), which may be due to the preferential pre-synaptic targeting of hM4Di in cortical neurons 30,31 and to the low expression of G protein-coupled inwardly-rectifying potassium (GIRK) channels in the cortex at this age 32 . Of note, we observed a significant difference in excitability between hM4Di-mCherry + and hM3Dq-mCherry + cells in the absence of any ligand, suggesting that hM3Dq DREADDs are partially activated in basal conditions (p<0.0001 F(6,144)=85.479, see controls in Supplementary Figure 3B and 3D). Nevertheless, the inhibitory action of hM4Di was demonstrated by a significant decrease in mPFC Egr1 expression among hM4Di-mCherry + neurons, in P8 and P15 animals sacrificed 2 hours after CNO injection (Supplementary Figure 3). Additionally, a sustained decrease in cFos labeling was observed in P15 mPFC of SFR+hM4Di mice sacrificed 24h after chronic CNO exposure (P2-P14), while a sustain increase was observed in P15 MS+hM3Dq mice (Supplementary Figure 3). Altogether, these results support the efficacy of our approach for mimicking or preventing the chronic decrease in IEG expression in the mPFC induced by MS during the postnatal period.
To test whether changes in early neuronal activity play a causal role in the MS-induced oligodendrogenesis defects, P15 and adults brains were analyzed after SFR+hM4Di or MS+hM3Dq treatments, as described above. We found an increased density of OL at P15 (F(1,7)=15.991 p=0.005, Figure 3D), and a decrease density of OPC in adults (F(1,9)=7.809 p=0.0209, Figure 3E) in SFR+hM4Di animals after P2-P14 CNO exposure. Conversely, we found an increase in the density of OPC at P15 in MS+hM3Dq animals with P2-P14 CNO exposure (F(1,7)=6.52 p=0.0379, Figure 3F), that was followed by a late increase in OL density (F(1,8)=25.795 p=0.001, Figure 3G). Therefore, transient inhibition of mPFC neurons reproduced the oligodendrogenesis defects of mice exposed to MS, suggesting early exit of the cell cycle. On the other hand, transient activation of the mPFC was sufficient to prevent the MS-induced defects in oligodendrogenesis. These results indicated a causal role of altered mPFC neuronal activity on the precocious oligodendrogenesis observed in MS animals.

Postnatal local decrease in mPFC activity recapitulates MS behavioral phenotype in adults.
Adult animals exposed to MS have been shown to display behavioral phenotypes, such as increased depression and cognitive deficits, suggestive of alterations in mPFC development. We tested the causal role of developmental pharmacologic manipulations in the PFC on these phenotypes. Because variability in MS effects in different mouse strains has been reported 11 , we began with a characterization of the behavioral phenotype of adult BALBcJ mice exposed to chronic MS ( Figure 4A). In our protocol, MS males showed decreased exploration (Mann-Whitney p=0.0002) and time spent in the center in the open field (OF, F(1,25)=4.317 p=0.0482, Figure 4B and C) as well as decreased time spent in the open arm of the elevated plus maze (EPM, F(1,22)=5.438 p=0.0293, Figure 4D). They further showed an increased burying behavior in the marble test (F(1,32)=5.402 p=0.0266, Figure 4E), when compared with SFR animals. In addition, MS males displayed an increased time floating in the forced swim test (FST, F(1,26)=6.048 p=0.0209, Figure 4F) and an impaired working memory in the sequential novel object recognition test (SNOR, F(1,23)=4.75 p=0.0396,) which measures the preference in exploring an object seen less recently than another 18,33 . These results confirmed an increase in anxiety and depression-related behaviors in MS animals and impaired in a working memory task highly linked to mPFC functioning 33 .
Strikingly, adults SFR+hM4Di animals exposed to P2-P14 CNO exhibited behavioral alterations very similar to those of MS animals, when compared with SAL treated animals. These included reduced exploration of the OF (distance F(1,24)=6.433 p=0.0181, Figure  As an additional control, to exclude any unwarranted effects of the effects of chronic exposure to CNO, a P2-P14 CNO/SAL cohort of animals was tested for adult behavior and showed no behavioral change (Supplementary Figure 5). Altogether, our data show that decreased mPFC activity during development is sufficient to recapitulate salient features of the adult's MS phenotype.

The postnatal local increase in mPFC activity prevents the emergence of depression-related MS phenotype.
To test for the requirement of altered mPFC activity in MS behavioral phenotype, we assessed the behavior of animals exposed to the MS protocol with or without chemogenetic activation (hM3Dq) of the mPFC ( Figure 5A). When compared to SAL-treated MS animals, CNO treated MS mice did not show a difference in the behavioral parameters related to anxiety, including: the However, CNO exposed MS+hM3Dq animals floated less in the FST ( Figure 5F, F(1,29)=6.508 p=0.0163) and display an improved performance in the working memory-test ( Figure 5G, Mann-Whitney p=0.0463). To strengthen our analysis of the depression-related phenotype we used the sucrose splash test, which measures self-care and hedonic behaviors 34 . We observed an increased grooming behavior in CNO exposed MS+hM3Dq animals as compared to SAL ( Figure 5H, F(1, 25)=6.532 p=0.171). Conversely, animals solely exposed to MS display a decrease in grooming when compared with SFR animals ( Figure 5I, F(1,25)=5.618 p=0.0258). Hence, transient mPFC activation prevented the emergence of the depression and cognitive aspects of MS phenotypes, but not of the anxiety-related aspect.

DISCUSSION:
Using an unbiased approach, we identified predominant alterations in plasticity and myelinrelated genes within the maturing mPFC of an early life stress model. Our analysis of the oligodendrocyte lineage showed a precocious differentiation upon MS and a subsequent depletion of the OPC pool in adult brains. We took advantage of pharmacogenetic tools to further establish that a transient and local alteration of mPFC neuronal activity during postnatal development was sufficient to recapitulate or prevent the defects in oligodendrogenesis and in mood behaviors induced by MS. Our results thus demonstrate a direct role of postnatal mPFC activity in the etiology of emotional and cognitive neurodevelopmental disorders and raise the possibility that postnatal oligodendrogenesis is involved in this process.

The pre-weaning period is critical for mPFC myelination
We found that MS (P2-P14) induces a significant increase in the expression of myelin-related genes at P15, correlating with a precious differentiation of the oligodendrocyte lineage and permanent alterations of the adult OPC pool. These results identify P2-P14 as a critical period for oligodendrogenesis in the mPFC, which coincides with the critical period for emotional behaviors, including mood and anxiety disorders 35,36 . Previously, another critical period for oligodendrogenesis has been identified at juvenile stages (P21-P35) when transient social isolation induces permanent deficit in adult social interactions associated with decreased myelin content and OPC proliferation 37 . Of note, adult exposure to various chronic stress has only transient consequences on behavior and mPFC myelination [37][38][39] . Therefore, our results, by showing the importance of the P2-P14 period, suggest that successive and/or expanded critical period(s) occur in the mPFC to establish anxiety, mood and social circuits. These two postnatal periods correspond to different stages of oligodendrogenesis that could be differentially affected by stressors although both result in decreased myelin content in mPFC at adult stages 12,20 .
The developmental mechanisms involved in adult hypomyelination after MS are still disputed 7,12 . Our data favor a precocious differentiation of the oligodendrocyte lineage resulting in a depletion of the OPC pool in adults. This interpretation is supported by a recent clinical paper indicating that childhood adversity in human biases oligodendrocyte lineage toward a more differentiated phenotype in the vmPFC, with more pronounced effects in samples from young 7 individuals that died closer to ELS experience 7 . In addition, several other studies of sensory deprivation during critical periods or exposure to early life stress similarly reported, in other brain structures than mPFC, a transient and immediate increase in myelin content followed by a permanent decrease in adulthood [40][41][42][43][44] . However, another study by Yang et al. 20 reported a deficit in OPC differentiation when studying effects of MS in the mPFC of Sprague Dawley rats. The discrepancy with the present results could be due to a number of technical differences such as species, and timing of analysis and proliferation labeling. Identifying the precise mechanisms by which ELS affects oligodendrogenesis should ultimately clarify this debate, while overall these data support critical effects of MS on oligodendrocyte maturation processes.

Neuronal activity controls early stages of oligodendrogenesis
Recent advances have established an important role for experience and neuronal activity in controlling adult oligodendrogenesis and myelination 29 . Additionally, our unbiased approach identified an important decrease of several IEG in the P15 mPFC after MS, as a proxy for a deficit in neuronal plasticity 45 . Converging reductions of Fos and FosB further suggest that there is a chronic dampening of neuronal network activity in the mPFC. These results are in agreement with previous reports in the rat MS model showing a local decrease in specific IEG expressions and a reduced neuronal synchrony in the mPFC, as revealed by in vivo recordings 14,18 .
We further demonstrate that stimulation of neuronal activity in the developing mPFC favors OPC proliferation at the expense of differentiation and prevents the emergence of oligodendrogenesis defects induced by MS. Conversely, chronically decreasing activity induces a bias toward OPC differentiation, thus mimicking the MS phenotype. To the best of our knowledge, this is the first demonstration of a causal role of neuronal activity on mPFC oligodendrogenesis and on postnatal oligodendrogenesis. Nevertheless, in adults, increased neuronal activity stimulates proliferation and differentiation of OPC and increases myelination in the long run 29,46,47 . This is therefore opposite to the effect that we observe during early postnatal life, since we show that increased neuronal activity stimulates proliferation at the expense of differentiation. Our results suggest that early stressors induce a pathological precocious differentiation which thereafter alters myelin content, or the capacity of myelin plasticity in adult life. Therefore, different activity-dependent regulations of oligodendrogenesis could be at play during development. This is supported by changes in the molecular properties of OPCs during this postnatal period 48,49 , including the establishment of transient synaptic connectivity between OPCs and interneurons 29,50 .

Early neuronal activity as a common regulator for psychiatric disorders?
Another important finding of the present study is the direct role of activity during critical periods of early life, for the etiology of emotional and cognitive disorders. The fundamental role of the excitatory/inhibitory balance in the establishment of critical periods has been identified in all primary sensory cortices 21,51,52 . Our study provides the first clear demonstration that locally modifying mPFC neuronal activity during a critical period durably affects adult emotional behavior. The mPFC neuronal network is still being assembled during the postnatal period, in particular with the maturation of outgoing and cortico-cortical glutamatergic and GABAergic circuits. Since our pharmacogenetic approach targeted both types of neurons, it remains to be determined how this affects circuit refinement. However, on a translational perspective, our 8 results demonstrate that early mPFC activity tightly controls two major MS-related endophenotypes, namely oligodendrogenesis and depression-related behavior, thus establishing a common cause for these typical traits of emotional disorders [53][54][55] . A direct role of oligodendrogenesis activity in the etiology of emotional disorders has previously been suggested by recent studies in adults reporting emotional deficits after genetic alterations of oligodendrogenesis 37,[56][57][58][59][60] . Another line of evidence showed that the social avoidance induced by chronic stress correlate with myelin alterations in the mPFC and that restoring myelination is sufficient to rescue it, hence reinforcing the correlations between myeline defects and behavioral defects 39,61 . Present results establish a critical role of mPFC postnatal neuronal activity in the etiology of ELS behavioral alterations, and further support an implication of early oligodendrogenesis in adult behavior.

Animals
All experiments performed in mice were in compliance with the standard ethical guidelines (European Community Guidelines and French Agriculture and Forestry Ministry Guidelines for Handling Animals-decree 87849). All dams were first-time pregnant BALB/cJ Rj mothers shipped from Janvier Labs ~ 7 days before delivery. At arrival, pregnant females were grouphoused by 2 and litters sizes were homogenized from 8 to 14 animals with both genders represented. After weaning at P21, all the mice were group-housed (4-5 per cage), and kept under standard laboratory conditions (22 ± 1°C, 60% relative humidity, 12-12 hrs light-dark cycle, food and water ad libitum) in ventilated racks. Males and females were similarly exposed to maternal separation protocol (MS) and/or viral injections, but only males were used for results while females were sacrificed at weaning except for: (i) the first cohort of MS animals where both genders were behaviorally tested, (ii) females exposed to P1 viral injection were sacrificed at P8 or P14 to test for CNO efficacy using Egr1 staining, and (iii) a cohort of SFR females exposed to chronic postnatal CNO.

Maternal Separation Protocol:
Maternal separations (MS) of pups from their dams were conducted daily between 1:00 and 4:00 P.M., starting at postnatal day 2 (P2) and terminating at P14. Briefly, dams were first removed from the home cage and placed into a clean cage, then pups, were collected and placed together into another clean cage. The two cages were distanced to avoid vocalizations. After the 180 min, pups and then dams were returned to their home cage. Control animals were standard facilityreared (SFR) and did not receive extra handling. Animals were sexed at P1 to allow homogeneous distribution of males and first cage change occurred at P2, then every week.

RNA-sequencing
For the transcriptomic profiling, bilateral mPFCs from SFR (n=12) and MS (n=12) animals were dissected manually at P15 (Figure 1), fast frost on dry ice and collected in 100 µL of RNALater® (Ambion, Life Technology). 3 unilateral mPFCs from different litters were pooled to generate 4 samples by treatments that were shipped to Penn State Hershey Genome Sciences Facility. Total RNA was extracted using mirVana kit (Thermo Fisher Scientific) and a bead mill homogenizer (Bullet Blender, Next Advance) together with an equivalent mass of stainless steel beads (Next Advance) were used to homogenize the tissue. The cDNA libraries were prepared using SureSelect Strand Specific RNA Library Preparation Kit (Agilent Technologies) and RNA and cDNA qualities were assessed using BioAnalyzer Kits (Agilent Technologies, RNA 6000 Nano and High Sensitivity DNA). The 12 libraries were loaded onto TruSeq SR v3 flow cells on an Illumina HiSeq 2500 and run for 50 cycles using a single-read recipe (TruSeq SBS Kit v3). The Illumina CASAVA pipeline Version 1.8 was used to extract de-multiplexed sequencing reads and FastQC (version 0.11.2) was used to validate the quality of the raw sequence data. Additional quality filtering was performed using FASTX-Toolkit with a quality score cutoff of 20. Next, alignment of the filtered reads to the mouse reference genome (mm10mm10) was done using Tophat (version 2.0.9) allowing 2 mismatches. Raw read counts were calculated using HTSeq-count 62 . The "remove unwanted variation" (RUVg) strategy ("RUVseq" R package, v3.1) as described by Risso et al. 26 was used along with edgeR 63 to identify differentially expressed genes (DEG) between MS and SFR animals. First we normalized the raw read counts by selecting a set of "in-silico empirical" negative controls, i.e., 5000 least significantly DEG genes based on a first-pass DEG analysis performed prior to normalization. Normalized read counts were applied using the negative binomial GLM approach implemented in edgeR using the Likelihood Ratio Test method. Significantly DEG were defined to be those with q-value < 0.05 calculated by the Storey et al. method 64 .

RT-qPCR
mPFC were dissected and homogenized in 1ml of TriReagent (Sigma) using a Polytron®PT13000D (Kinematica) and RNA were extracted according to standard protocols. Genomic DNA was removed by digestion with Amplification Grade DNase I (Sigma-Aldrich). First-strand cDNA was synthesized by reverse transcription of 5μg of total RNA with Superscript-II reverse transcriptase (Invitrogen) according to standard protocols. Reverse transcriptase was omitted in some samples as negative control. Relative expression levels of Gapdh, Arc, Fos, Fosb, Mag, Mog and Plp1 mRNAs were determined by real time quantitative PCR (RT-qPCR) using Absolute SYBR Green Mix (ABgene) and sets of specific primers (Supplementary Table 2) on a 96 wells plate (Thermo Scientific AB-0600). Gene expression was normalized to mouse Gapdh mRNA expression. Data were analyzed with the 2-ΔCt method on MxPro qPCR software (Agilent Technologies), and values are expressed as the mean of duplicates.

AAV injections in the PFC
AAV8-hSyn-hM3D(Gq)-mCherry or AAV8-hSyn-hM4D(Gi)-mCherry (Addgene #50474 and #44362) were used as specified. The titers were within the range: 10 12 -10 13 particles/mL. Injections were done on P1 mice pups anesthetized on ice (1minute or until the absence of pinch reaction) and kept on a frozen pad during all the procedure (max 8 min). The cages were put on a heat pad 30min before the onset of the experiment and litters were separated from the dams and put on another heated cage with some bedding right before the onset. Pups were put in prone position to maintain the skull flat and the skin covering the skull was gently opened with a single incision using a razor blade. The mPFC was targeted using as a reference the inferior cerebral vein (~2-3 mm posteriorly) and the superior sagittal sinus (~2mm laterally). Bilateral injections (200 nL each) were performed at 1.1 mm deep from the skull surface using a pulled glass capillary (30-50 µm tip diameter PCR micropipette, Drummond Scientific Company) mounted on a hydraulic micromanipulator MO-10 (Narishige, Japan). After injection, the opened skin on the skull was sealed using tissue adhesive glue (Vetbond #1469, 3M, France) and pups were gently warmed up. All pups from the same litters were injected to avoid competition for feeding and they were returned altogether to homecage at the end of the procedure. Careful supervision was provided on the hours following the procedure to favor mother acceptance and re-glue the scare if necessary. Afterward, cages were returned to the normal facility and daily supervision was provided.

Tissue processing
For cortical ontogenetic analysis, mice were sacrificed at postnatal (P8-P15) and adult stages. All mice were anesthetized (Pentobarbital 0.5 mg/g) and fixed by intracardiac perfusion of 4% paraformaldehyde in 0.1 M phosphate-buffered (PB pH=7.2) using a peristaltic pump (Thermoscientific Masterflex). Brains were quickly removed, post-fixed overnight in the same fixative solution and cryoprotected 2 days in 30% sucrose containing 0.01% sodium azide. 35 µm-thick coronal sections were prepared using a cryostat (Leica) or a cryotome (Microm Microtech). Serial sections from the whole brain were collected as series of 6. Tissue sections were used immediately for immunohistochemistry, if not, they were stored at -20°C in cryoprotective solution (30% ethylene glycol and 30% sucrose in 0.1 M phosphate buffer, pH=7.4).

Images acquisition
Fluorescent images were acquired on a Leica SP5 confocal system equipped with an Argon laser (for the 488nm excitation), a Diode 561nm and HeNe 633nm. The Z-series stacks of confocal images were acquired at 1024 x 1024 pixels resolution, with a pinhole set to one Airy unit and optimal settings for gain and offset.

Behavioral Studies
Female and male mice were tested separately for emotional behaviors following the series of behavioral tests starting at P80. Several cohorts were conducted for MS and SFR+hM4Di phenotype and none has undergone all the tests. For the single MS+hM3D cohort, the tests were performed in the following order: open-field, elevated plus maze, burying marble test, sequential objects recognition test, splash test and the two-days forced swim test, with 3 to 7 days between each test according to the stress of the test. All behavioral testing was done during the light cycle between 9 am and 6 pm and animals were transferred to the testing room in their home cage at least 30 min before the onset of the test for habituation. To eliminate odor cues, each apparatus was thoroughly cleaned with alcohol after each animal and males and females were conducted separately. Testing took place under ambient light conditions (50-80 lux).
Open Field test: Exploration and reactivity to a novel environment were assessed using an open field (OF) test. Mice were placed in the center of rectangular plastic grey boxes (H24,5xL35xL55cm) and activity was recorded for 20 min. Total distance and time spent in the center of the OF were automatically recorded using AnyMaze tracking software (Stoelting Co).

Elevated Plus Maze test:
The maze is a grey plus-crossed-shaped apparatus, with two open arms (30 cm) and two closed arms (30 m) linked by a central platform and located 40cm above the floor. Mice were individually put in the center of the maze facing an open arm and were allowed to explore the maze for 10 min. The time spent and the entries into the open arms were automatically measured using AnyMaze tracking software and used as anxiety indexes.
Burying Marble test: Housing cages (20x36 cm) were filled with 6 cm of wood bedding (instead of housing white chip cellulose bedding) and 3x4 marbles were evenly distributed as a grid. Mice were individually put in a corner of the cage and the number of marbles buried on more than 2/3 of their surface was manually recorded every minute for 20 min.
Splash Test: Group housed animals were placed in a new cage with bedding for at least 30 min in a room with dim light (40 lux). Mice were then individually sprayed on the back twice (~2x0,6 mL) with 20% sucrose solution in water and placed in a corner of their home cage. Grooming behaviors, including licking, stroking and scratching were manually recorded using AnyMaze software keys.

Sequential Novel Object Recognition test:
This test is an adapted version of the novel object recognition test more specific to mPFC functioning and was performed as previously described 33 . Animals were first habituated to the arena (same grey boxes as OF), without stimuli for 10 min daily for 2 d before the commencement of the behavioral testing. This task comprised two sample phases and one test trial. In each sample phase, the subjects were allowed to explore two copies of an identical object for a total of 4 min. Different objects were used for sample phases 1 and 2 (miniature car then plastic hook, or glass tube then cork of Falcon tube), with a delay between the sample phases of 1h. The test trial (3 min) was given 3 h after sample phase 2. During the test trial, one copy of the objects from sample phase 1 and one copy of the objects from sample phase 2 were used. The positions of the two objects were counterbalanced between the animals. If temporal order memory is intact, the subjects will spend more time exploring the object from sample 1 (i.e., the object presented less recently) compared with the object from sample 2 (i.e., the novel object).
Porsolt Forced Swim test: Testing of behavioral despair was carried out in two consecutive days using a glass cylinder (40 cm X 20 cm diameter) filled with water (23-25°C) up to 3/4. Mice were tested once a day for 6 min. All the swim sessions were videotaped and manually tracked using AnyMaze software keys.

Statistics:
The Gaussian distributions of the values were tested on Prism software using d'Agostino and Pearson or Shapiro-Wilk for small n. T-test or ANOVA test were then performed when possible and otherwise Mann-Whitney test was applied and specified in the text. All results are expressed in mean ± sem. * p<0.05, ** p<0.01 and *** p<0.005.  A Scheme represents the specificity of the oligodendrocyte lineage markers using triple immunostaining. Oligodendrocyte progenitor cells (OPC) express Olig2 and high levels of PDGFRα while mature oligodendrocytes (OL) express Olig2 and high levels of the CC1 antigen. Scale bar represents 20µm. B and C Graphs show the density of OPC (dark orange) and OL (light orange) in the P15 (B) and adult (C) mPFCs of SFR and MS animals. D Scheme represent the protocol of sequential labeling of OPC using IdU injections at P10 and BrdU injections at P15 to differentiate the fraction of P10 OPC that are still cycling at P15 (blue, IdU + BrdU + ) from those which start differentiating by then (purple, IdU + BrdU -). E Graph shows an increased fraction of differentiated OPC in P15 mPFC of MS animals as compared to SFR animals. Figure 3: Neuronal activity bi-directionally controls oligodendrocyte differentiation in the developing mPFC. A Scheme represents the experimental paradigms of the chronic treatment with CNO or SAL solutions from P12 to P14 after local and bilateral injections of pharmacogenetic virus. All animals were either exposed to inhibitory hM4D(Gi) virus and SFR protocol (blue, SFR+hM4Di) or exposed to excitatory hM3D(Gq) virus and MS protocol (red, MS+hM3Dq). B and C Images show mCherry labeling in P15 mPFC (B) and the specificity of viral expression to neuronal lineage by NeuN and mCherry co-labeling (C). Scale bars represent 500 µm (B) and 20 µm (C). D-G Graphs show the density of OPC (dark orange) and OL (light orange) in the P15 (D, F) and adult (E, G) mPFCs of SFR+hM4Di (D, E) and MS+hM3Dq (F, G) animals. SFR+hM4Di (blue) animals exposed to behavioral tests in adulthood. B and B' Graphs show the distance travelled in an open field (OF) by SFR or MS animals (B) and by SFR+hM4Di animals exposed to chronic SAL or CNO (B'). C and C' Graphs show the time spent in the center of the OF during a 20min test. D and D' Graphs show the time spent in the open arm of an Elevated Plus Maze (EPM) during a 10min test. E and E' Graphs show the number of buried marbles at each minutes of a 15min test. F and F' Graphs show the time floating per min of a 6 min forced swim test (FST). G and G' Scheme represents the experimental paradigm of the sequential object recognition test with 3min of two identical Object1 exploration, followed 1h after by 3min of two identical Object 2 exploration, and 3h after by 4min of one Object1 and one Object2 exploration test. Graphs show the preference ratio for Object1. G Graph shows the preference ratio for Object1. H and I Graphs show the cumulative time grooming after a splash with 20% sucrose solution by SFR+hM3Dq animals exposed to chronic SAL or CNO (H) and by SFR or MS animals (I).
A P0 P2 P15 P14 [gene]*100/[GAPDH] D Figure 1: Analysis of differentially expressed genes in the mPFC of P15 animals exposed to maternal separation highlights myelin ensheathment and immediate early genes.