Early GABAergic OPC synapse disruption impairs tone fear extinction
To investigate the potential impact of early postnatal interneuron-OPC interactions on cognitive functioning in adulthood, we used the NG2cre ;Gcamp3 ;γ2 mice (γ2 mice) in which transient γ2 subunit-mediated GABAergic synapses of OPCs are inactivated by 4-hydroxytamoxifen (4-OHT) injections administered from postnatal day 3 (P3) to P5 (Supplementary Fig. 1a; see Online Methods). In agreement with our previous reports which demonstrated recombination at the γ2 locus and a partial reduction in OPC evoked GABAergic postsynaptic currents (I) in the somatosensory cortex, our current results reveal a near-complete abolishment of I in layer 5 OPCs within the IL region of acute mPFC slices of γ2 mice at P9-P13 (Supplementary Fig. 1b, c). This finding indicates a more pronounced impairment of interneuron-OPC synaptic signaling in the mPFC compared to the somatosensory cortex.
To assess whether γ2-containing GABA-A receptor-mediated currents persist in OPCs into adulthood, we recorded I in layer 5 OPCs of 2-month-old control and γ2 mice. In contrast to the pronounced γ2-dependent currents observed in young animals, evoked I remained detectable at similar levels in both groups and was insensitive to zolpidem (1 μM), a positive allosteric modulator selective for GABA-A receptors containing the γ2 subunit (Supplementary Fig. 1a, d, e). This pharmacological profile as well as the presence of a GABAergic current in the γ2 mice is consistent with prior reports indicating that, while adult OPCs in the cortex and hippocampus express functional GABA-A receptors, these cells lack the γ2 subunit. Together, these data confirm that γ2-mediated GABAergic synapses are developmentally restricted and absent in adult OPCs in the mPFC. To further validate the efficiency of OPC targeting in this region, we quantified recombination by immunolabeling Olig2 cells and enhancing GCaMP3 detection with an anti-GFP antibody in 4-month-old mice. We found 65.5% of recombinant OL lineage cells in control mice and 63.4% in γ2 mice in the mPFC, indicating efficient and comparable recombination for both groups (Supplementary Fig. 2a).
After confirming the early functional inactivation of OPC GABAergic synapses in this cortical region in γ2 mice, we next investigated whether this developmental synaptic alteration would affect mPFC-dependent cognitive functions in adulthood. To assess this, we first evaluated four-month-old γ2 mice using a fear conditioning task, a widely used paradigm to investigate mPFC-dependent processes such as learning, memory, emotional regulation, and perception.
To evaluate learning and memory processes associating environmental cues with aversive experiences, adult mice subjected to 4-OHT injections at P3-P5, underwent a behavioral test where a 30 s conditioned tone stimulus was paired with a foot shock delivered during the last 2 s of the tone in a specific context (Fig. 1a). In the conditioning session, adult control and γ2 mice exhibited similar fear memory acquisition, albeit slightly faster in γ2 mice (Fig. 1b). No significant differences were found between the two groups during the context extinction session (Fig. 1c). However, in the first tone extinction session, although the baseline level (BL, first trial block) of fear expression were comparable in control and γ2 mice (Fig. 1d), control mice showed faster and more extensive fear extinction learning by the end of the protocol (Fig. 1d). To test whether this deficit in fear extinction learning was persistent, we performed a second tone extinction session on the following day (extinction retrieval). We found that, in this session, control mice exhibited significantly lower baseline values and reduced initial freezing levels compared to the first tone extinction session, indicating successful extinction retrieval (Fig. 1e). The tone extinction learned in the first session was therefore retained and successfully retrieved over time in control mice. In contrast, γ2 mice showed persistently high baseline values and a weak extinction, with freezing levels remaining similar to those observed in the first tone extinction session (Fig. 1e), indicating that these mice failed to improve or consolidate fear extinction learning between sessions. Of note, no significant sex differences were observed between male and female mice in both control and γ2 groups (Supplementary Fig. 3a-h). To confirm that the observed behavioral deficits specifically result from early postnatal recombination rather than to the presence of the floxed γ2 alleles, we reproduced the same fear conditioning paradigm in γ2 mice that were not injected with 4-OHT. These animals retain normal developmental expression of γ2 subunits in OPCs during development, while carrying the same Cre and γ2 alleles as the injected littermates. Non-injected γ2 mice showed no differences in fear conditioning behavior compared to controls (Supplementary Fig. 5a-e; Fig. 1b-f), indicating that the presence of the floxed alleles does not affect extinction behavior. We then assessed the effect of inducing recombination specifically in adulthood (Supplementary Fig. 5a). In this condition, no significant differences in tone extinction learning or retrieval were observed compared to non-injected γ2 mice (Supplementary Fig. 5b-e). These findings indicate that the behavioral deficits depend on early postnatal disruption of γ2-mediated GABAergic signaling in OPCs, and that recombination induction in adulthood, when the γ2 subunit is not expressed in OPCs, has no effect.
Given that the deficit was specific to tone extinction and not observed in context extinction, we considered whether this could be due to auditory impairments. To address this, we measured startle responses to noise stimuli of varying intensities, including the one used during fear conditioning (Supplementary Fig. 4a, b). Although this test does not assess potential hidden hearing loss, we found that both groups exhibited comparable startle responses to random noise of varying dB levels. This suggests that major auditory deficits are unlikely to account for the observed differences in fear conditioning (Supplementary Fig. 4c, d). Building on these findings, we further analyzed the extinction learning deficit in γ2 mice throughout different stages of the retrieval session (Supplementary Fig. 3i). The extinction learning deficit in γ2 mice was predominant at early and intermediate stages, but it was also evident at the late stage (Fig. 1f), showing their impaired ability to acquire and consolidate an extinction memory.
Our results indicate that γ2 mice exhibit reduced fear learning and memory capabilities, particularly in tone fear extinction learning, while no major auditory impairments were detected. These results suggest that the disruption of OPC GABAergic synaptic signaling during the early postnatal period leads to long-term cognitive impairments in adulthood, specifically affecting processes related to learning and memory, without evidence of substantial auditory deficits.
The long-lasting cognitive impairments observed in γ2 mice, particularly during tone fear extinction, suggest that these deficits may be caused by underlying neurophysiological alterations. Therefore, we specifically tested the local excitation-inhibition (E/I) balance in the mPFC, a structure involved in regulating cognitive processes such as learning, memory, and emotional regulation. This hypothesis is further supported by the known role of mPFC PV interneurons in regulating the E/I balance by modulating excitatory inputs, which are necessary for higher-order cognitive functions. Since fear conditioning experiments revealed specific deficits in tone fear extinction in γ2 mice, we focused on the IL region of the mPFC, a critical region for fear extinction memory processes.
To test the E/I balance in the IL region of adult γ2 mice, we performed patch-clamp recordings of layer 2/3 pyramidal cells in acute slices, while stimulating in layer V with an extracellular electrode. The stimulation elicited both excitatory (eEPSCs) and inhibitory (eIPSCs) postsynaptic currents in mature IL circuits (Fig. 2a). Our results revealed that, 3 months after inactivating the γ2 subunit-mediated GABAergic synapses in OPCs, γ2 mice exhibited a significant reduction in the mean amplitudes of eIPSCs, but not eEPSCs, in pyramidal neurons (Fig. 2a, b). Consequently, the E/I ratio was higher in γ2 mice, causing a strong E/I imbalance (Fig. 2b).
To assess whether the E/I imbalance observed in the mPFC reflects a broader effect or is more regionally restricted, we extended our analysis to two additional brain regions involved in fear expression and extinction: the amygdala and hippocampus. First, we recorded from principal neurons in the basolateral amygdala (BLA), while stimulating the lateral amygdala (Fig. 2c, d). Similarly, we recorded from CA1 pyramidal neurons in the hippocampus, while stimulating Schaffer collateral inputs from CA3 (Fig. 2e, f). Unlike the IL region, we found no significant differences in eIPSC and eEPSC amplitudes, nor in the E/I ratio between control and γ2 mice in either the BLA or CA1 regions (Fig. 2d, f).
To further evaluate the functional specificity of these alterations, we complemented these recordings with region-specific behavioral tests targeting the amygdala and the hippocampus: the elevated plus maze and the novel object recognition, respectively. In both tasks, γ2 mice performed similarly to controls, showing no significant differences in anxiety levels and recognition memory (Supplementary Fig. 5f, g). These results are consistent with the lack of synaptic alterations in these regions. Moreover, they align with our fear conditioning results, in which γ2 mice exhibited normal fear acquisition, a process largely mediated by the amygdala, and normal contextual extinction, which strongly depends on hippocampal function (Figs. 1b, c).
Since the proper regulation of E/I balance is essential for normal cognitive processing, the observed alterations in the mPFC functioning suggest a functional impairment in this region due to early interneuron-OPC synapse disruption. These findings indicate that the synaptic E/I imbalance is more prominent in the mPFC than the amygdala and hippocampus, suggesting that the mPFC is particularly sensitive to early developmental perturbations.
The observed cognitive and E/I ratio deficits prompted us to investigate whether early disruption of interneuron-OPC synapses results in persistent myelination deficits, particularly in PV interneurons known for their highly plastic myelination. Previous findings in the somatosensory cortex demonstrated that γ2 mice exhibit myelination defects in PV interneurons during juvenile stages. Here, we extended this analysis to the mPFC in adulthood to determine whether these abnormalities persist and specifically affect the myelination of PV interneurons, the primary source of transient GABAergic synaptic input onto OPCs. We found that γ2 mice displayed normal overall myelin coverage and intensity of MBP staining in IL cortical layers (Fig. 3a-c). We then analyzed the length of individual myelin segments, or internodes, along PV axons in the same region. This was achieved by reconstructing high resolution images of axonal segments expressing MBP, PV and the axonal marker SMI-312 (Fig. 3d). Measurements revealed a statistically significant increase in the mean internode length of PV, but not PV, axons of γ2 mice compared to controls (Fig. 3e, f). The myelination defect observed in PV axons was not accompanied by alterations in either the cell density or distribution of PV interneurons across mPFC layers, nor by changes in the total interneuron population identified by GAD67 expression (Supplementary Fig. 6a-e). In addition, quantification of PV signal within SMI-312 axonal compartments — used as a proxy for PV axon density — did not reveal significant differences between control and γ2 mice (Supplementary Fig. 6f, g). Furthermore, we found no differences in the cell density or proportion of Olig2/CC1 OPCs and Olig2/CC1 OLs between control and γ2 mice (Supplementary Fig. 2b-d). These results suggest that the early disruption of γ2-mediated synapses in OPCs impacts predominantly PV interneuron myelination, without affecting interneuron survival or the maturation of OL lineage cells in the mPFC.
Although OPCs form GABAergic synapses only transiently during postnatal development and these synapses are absent in adulthood, myelination defects in PV interneurons persists into adulthood of γ2 mice. Notably, these alterations are specific to PV interneurons as we found no significant changes in the overall mPFC myelination pattern, length of internodes of PV axons, OL lineage cell densities, PV interneuron survival, the distribution of interneurons within cortical layers or of the estimated PV axon density.
The persistent PV interneuron dysmyelination as well as the electrophysiological and cognitive impairments observed in γ2 mice raised the question of whether these deficits could be corrected through targeted interventions. To this end, we explored two distinct strategies aimed at addressing the underlying structural and behavioral abnormalities: enhancing PV interneuron activity and promoting myelination.
Previous research has demonstrated that using a modified form of the human M3 muscarinic (hM3Dq) receptor, activated by injections of the inert clozapine metabolite clozapine-N-oxide (CNO) for 14 days, can specifically promote activity-dependent myelination of PV interneurons in the adult mPFC. In addition, acutely increasing PV interneuron activity with a chemogenetic approach should also rebalance the observed reduced cortical inhibition, improving E/I impairments (Fig. 2a, b). Building on this finding, we explored whether selectively targeting PV interneurons with a chemogenetic approach could effectively address dysmyelination and fear learning deficits (Fig. 4a). We performed bilateral intracranial injections into the IL region of the mPFC in γ2 mice using either pAAV-S5E2-dTom-nlsdTom (dTom) or pAAV-S5E2-hM3D(Gq)-P2A-dTomato (hM3Dq-dTom) viral vectors (Fig. 4a, b). Approximately 80% of cells infected by these viral constructs were identified as PV interneurons in this region (Fig. 4b, c). To confirm the functional efficacy of the hM3Dq-mediated activation of PV interneurons, we first performed whole-cell patch-clamp recordings in acute mPFC slices from γ2 mice injected with pAAV-S5E2-hM3D(Gq)-P2A-dTomato. The dTom neurons, that express hM3Dq, exhibited an increased excitability in the presence of CNO while nearby dTom pyramidal neurons did not, indicating that CNO effectively enhanced PV interneuron activity (Supplementary Fig. 7a-d). This supports the specificity and efficacy of hM3Dq-mediated chemogenetic manipulation. We then performed twice-daily in vivo injections of CNO over two weeks in dTom-injected and hM3Dq-dTom-injected γ2 mice (Fig. 4a). Our results demonstrate that CNO administration did not induce any significant changes in either the coverage or mean intensity of the MBP fluorescence signal compared to those in dTom-injected γ2 mice (Fig. 4d-f). Furthermore, we did not detect any reduction of PV interneuron internode length in hM3Dq-dTom-injected γ2 mice (Fig. 4g), suggesting that increased PV interneuron activity in adulthood does not restore aberrant internode length. To examine whether enhancing PV interneuron activity in vivo, independently of the effect on interneuron dysmyelination, ameliorates cognitive deficits, we conducted the fear conditioning task in both CNO-treated controls and γ2 mice. As expected, no differences were observed in the acquisition or context sessions (Supplementary Fig. 8a, b). Consistent with the lack of effect on PV interneuron internode length, we observed no significant differences during the tone extinction and extinction retrieval sessions between the two viral injected groups (Fig. 4h, i), nor when compared to naive γ2 mice (Fig. 1d, e).
Our second strategy consisted in enhancing myelination in young adult mice, a developmental period when myelination is still actively progressing in the mPFC and cognitive deficits associated with neurodevelopmental disorders are often recognized due to the emergence of visible symptoms. We reasoned that targeting this critical period might offer greater potential for modifying myelination and improving cognitive deficits because: (1) myelination plasticity persists well beyond early postnatal stages in cortical regions, and (2) therapeutic interventions during active myelination phases might rescue structural and functional deficits caused by earlier disruptions. We focused on fractalkine (FKN), a well-studied chemokine known for its role in microglial function, but also shown to promote OPC differentiation into mature OLs during cortical development through its release from GABAergic interneurons. Recent studies demonstrate that FKN can promote myelination under both healthy and demyelinating conditions when administered exogenously. We thus exploited the pro-myelinating potential of FKN to examine whether its exogenous application in mice could compensate for PV interneuron dysmyelination in the mPFC and mitigate the associated learning and memory deficits observed during tone fear extinction. To this end, we performed in vivo FKN infusions into the lateral ventricles of P42-46 mice over a 25-28-day period (Fig. 5a). Our results revealed that in vivo FKN infusions in γ2 mice increased both the coverage and mean intensity of MBP fluorescent signal in the mPFC of infused mice compared to vehicle-treated mice, indicating an enhanced myelination (Fig. 5b-d). However, measurements of internode length of PV/SMI-312 axons in the same animals showed no difference between the two groups, suggesting that this intervention is insufficient to repair PV interneuron-specific myelin abnormalities in γ2 mice (Fig. 5e). To assess whether FKN-induced myelination enhancement could still result in cognitive improvements, we conducted the fear conditioning task in γ2 mice treated with either vehicle or FKN for comparison. As expected, no differences were observed in the acquisition or context sessions (Supplementary Fig. 8c, d). However, no improvements were observed in the tone extinction or extinction retrieval sessions either (Fig. 5f, g), indicating that enhancing ongoing mPFC myelination fails to rectify PV interneuron dysmyelination or improve the associated learning and memory deficits.
Our findings show the irreversible cognitive impairments resulting from early disruption of OPC GABAergic synapses. Despite targeted interventions aimed at increasing PV interneuron activity or enhancing myelination, we observed no improvements in PV interneuron internode length and fear extinction learning and memory. Once early interneuron-OPC interactions are compromised during a critical developmental window, subsequent efforts to reverse structural and functional deficits become particularly challenging. The early postnatal period, particularly the first two postnatal weeks, is therefore essential for ensuring proper cognitive development and function.
The E/I balance in the mPFC, which is disrupted in γ2 mice, plays a fundamental role in generating gamma oscillations (30-90 Hz), high-frequency neural rhythms which support prefrontal-dependent cognitive processing. These oscillations arise from the rapid synchronization of excitatory activity, with PV interneurons playing a central role by providing perisomatic inhibition onto pyramidal cell ensembles, which fine-tunes network discharges. Deficits in E/I balance and PV interneuron myelination could potentially disrupt gamma oscillations, impacting cognitive functions. To investigate this possibility, we recorded local field potentials (LFPs) from the IL region of the mPFC in freely moving mice using an implanted tetrode (Fig. 6a), of which the placement was confirmed postmortem through histological analysis (Fig. 6b). After 10 days of recovery post-implantation, mice were first habituated to wearing the recording cable in their home cage for 3 days. Then, they were placed in an open field environment and later subjected to the fear conditioning task, during which LFP signals were recorded (Fig. 6a, c). We analyzed these signals using fast Fourier transformation (FFT) to generate spectrograms and power spectra allowing us to extract different oscillatory activities during the behavioral tasks (Fig. 6c-f, i, j).
In vivo LFP recordings during the open field test detected both theta (4-12 Hz) and gamma (30-90 Hz) oscillations in the mPFC of control and γ2 mice (Fig. 6c). Contrary to our hypothesis that disrupted myelination would impair gamma oscillations, there were no significant differences in theta or gamma oscillatory activity between the two groups, regardless of whether the mice were in the center or periphery of the arena (Fig. 6e-h). These results are consistent with the absence of significant differences in exploratory behavior between control and γ2 mice, including distance traveled, time spent, immobility time in the center and periphery of the arena, as well as the number of crossings or average speed (Supplementary Fig. 9a-f). These findings indicate that deficits in PV interneuron myelination did not result in alterations in either oscillatory activity of IL region or exploratory behavior under the conditions of the open field task.
The consequences of PV interneuron dysmyelination in the mPFC may be task-specific, potentially impacting complex processes rather than general locomotor or exploratory behaviors. These same implanted mice were therefore subjected to the fear conditioning protocol which is designed to assess more cognitively demanding processes (Fig. 6a). In vivo LFP recordings were performed during the tone extinction retrieval session, when more pronounced behavioral differences between control and γ2 mice were observed (Fig. 1e). During this session, we optimized recordings by using a Faraday cage and minimizing external noise. Analysis of the oscillatory activity revealed no significant differences in the relative power of theta oscillations between the two groups during tone presentation and inter-trial intervals (ITI) (Fig. 6i-k). However, γ2 mice exhibited a significant reduction in gamma oscillations throughout both periods, suggesting a specific disruption of neural dynamics at this frequency range (Fig. 6i, j, l). These results indicate that early postnatal disruption of interneuron-OPC signaling leads to persistent PV interneuron dysmyelination and E/I imbalance in the mPFC, which in turn drive alterations in gamma oscillations and the subsequent impairments in fear extinction learning observed in γ2 mice.
The absence of significant gamma oscillation changes in the open field, in contrast to the pronounced reduction observed during fear extinction retrieval, suggests that PV interneuron dysmyelination predominantly affects higher cognitive processes, without necessarily disrupting basal neural activity or basic behaviors like exploration or locomotion.
Gamma oscillations can be subdivided into low and high frequency ranges, each contributing distinctly to cognitive processes. High-gamma frequencies have been associated with encoding processes, whereas low-gamma frequencies support retrieval in human episodic memory. To investigate whether the differences in gamma oscillations observed during fear extinction retrieval between control and γ2 mice were frequency-specific, we conducted a detailed analysis of both low-gamma (30-45 Hz) and high-gamma (55-90 Hz) power. Our results revealed a significant reduction in the relative power of low-gamma oscillations in γ2 mice during both tone presentation and ITI of fear extinction retrieval, while high-gamma oscillations showed no significant differences (Fig. 7a, b). These results suggest that deficits in neuronal processing in mPFC of γ2 mice are primarily associated with alterations in low-gamma oscillations.
Gamma oscillations are known to couple strongly with theta oscillations during cognitive processes, playing a key role in coordinating network-wide communication. Our analyses reveal that the strength of the theta phase-low-gamma amplitude and theta phase-high-gamma amplitude coupling increased during tone presentation compared to the ITI in both control and γ2 mice (Supplementary Fig. 10b, f). This increase was observed across the entire session as well as at the different stages within the session (Supplementary Fig. 10d, e, h, i), indicating a robust network activation and an ability in both groups to discriminate the tone stimuli. This strong theta-gamma coupling serves as a hallmark of effective cognitive processing, especially since no significant changes in coupling were observed between the center and periphery of the arena in the open field task (Supplementary Fig. 9g, h). However, despite the clear increase in coupling during tone presentation compared to ITI, there were no significant differences between control and γ2 mice in the tone/ITI coupling ratio, either for low-gamma or high-gamma oscillations (Supplementary Fig. 10c, g). The lack of coupling defects shows that the coordination of broad network interactions remains intact, despite a decrease in low-gamma power.
To further dissect the temporal dynamics of gamma power alterations, we analyzed the relative power of low- and high-gamma oscillations at different stages of the retrieval session: early, intermediate, and late (Fig. 7c, d). We found that low-gamma power was significantly reduced during the early and intermediate stages in γ2 mice compared to controls (Fig. 7c) during tone presentations and ITI, while it remained unaffected during the late stage. These early and intermediate reductions in low-gamma power correspond with the higher freezing levels observed in γ2 mice at these stages (Fig. 1f), suggesting a temporal relationship between low-gamma disruption and behavioral deficits. In contrast, high-gamma power exhibited a trend toward reduction that did not reach statistical significance and remained comparable between control and γ2 mice throughout all three stages of the retrieval session (Fig. 7d), further supporting the conclusion that the observed alterations are more pronounced in the low-gamma frequency range.
In summary, our findings suggest that low-gamma oscillations in the mPFC play a more critical role than high gamma oscillations in supporting fear extinction retrieval. The reduction in low-gamma power, particularly during the early and intermediate stages of the retrieval session in γ2 mice, points to its essential role in enabling neuronal network activity required for the successful fear extinction learning and memory. This reduction, alongside the lack of changes in theta-gamma coupling, is consistent with an intracortical mPFC network dysfunction, resulting from PV interneuron impairments which are key for intracortical projections.
