Oxytocin Regulates Synaptic Transmission in the Sensory Cortices in a Developmentally Dynamic Manner

The development and stabilization of neuronal circuits are critical to proper brain function. Synapses are the building blocks of neural circuits. Here we examine the effects of the neuropeptide oxytocin on synaptic transmission in L2/3 pyramidal neurons of the barrel field of the primary somatosensory cortex (S1BF). We find that perfusion of oxytocin onto acute brain slices significantly increases the frequency of miniature excitatory postsynaptic currents (mEPSC) of S1BF L2/3 pyramidal neurons at P10 and P14, but reduces it at the later ages of P22 and P28; the transition occurs at around P18. Since oxytocin expression is itself regulated by sensory experience, we also examine whether the effects of oxytocin on excitatory synaptic transmission correlate with that of sensory experience. We find that, indeed, the effects of sensory experience and oxytocin on excitatory synaptic transmission of L2/3 pyramidal neurons both peak at around P14 and plateau around P18, suggesting that they regulate a specific form of synaptic plasticity in L2/3 pyramidal neurons, with a sensitive/critical period ending around P18. Consistently, oxytocin receptor (Oxtr) expression in glutamatergic neurons of the upper layers of the cerebral cortex peaks around P14. By P28, however, Oxtr expression becomes more prominent in GABAergic neurons, especially somatostatin (SST) neurons. At P28, oxytocin perfusion increases inhibitory synaptic transmission and reduces excitatory synaptic transmission, effects that result in a net reduction of neuronal excitation, in contrast to increased excitation at P14. Using oxytocin knockout mice and Oxtr conditional knockout mice, we show that loss-of-function of oxytocin affects baseline excitatory synaptic transmission, while Oxtr is required for oxytocin-induced changes in excitatory synaptic transmission, at both P14 and P28. Together, these results demonstrate that oxytocin has complex and dynamic functions in regulating synaptic transmission in cortical L2/3 pyramidal neurons. These findings add to existing knowledge of the function of oxytocin in regulating neural circuit development and plasticity.

The development and stabilization of neuronal circuits are critical to proper brain function. Synapses are the building blocks of neural circuits. Here we examine the effects of the neuropeptide oxytocin on synaptic transmission in L2/3 pyramidal neurons of the barrel field of the primary somatosensory cortex (S1BF). We find that perfusion of oxytocin onto acute brain slices significantly increases the frequency of miniature excitatory postsynaptic currents (mEPSC) of S1BF L2/3 pyramidal neurons at P10 and P14, but reduces it at the later ages of P22 and P28; the transition occurs at around P18. Since oxytocin expression is itself regulated by sensory experience, we also examine whether the effects of oxytocin on excitatory synaptic transmission correlate with that of sensory experience. We find that, indeed, the effects of sensory experience and oxytocin on excitatory synaptic transmission of L2/3 pyramidal neurons both peak at around P14 and plateau around P18, suggesting that they regulate a specific form of synaptic plasticity in L2/3 pyramidal neurons, with a sensitive/critical period ending around P18. Consistently, oxytocin receptor (Oxtr) expression in glutamatergic neurons of the upper layers of the cerebral cortex peaks around P14. By P28, however, Oxtr expression becomes more prominent in GABAergic neurons, especially somatostatin (SST) neurons. At P28, oxytocin perfusion increases inhibitory synaptic transmission and reduces excitatory synaptic transmission, effects that result in a net reduction of neuronal excitation, in contrast to increased excitation at P14. Using oxytocin knockout mice and Oxtr conditional knockout mice, we show that loss-of-function of oxytocin affects baseline excitatory synaptic transmission, while Oxtr is required for oxytocin-induced changes in excitatory synaptic transmission, at both P14 and P28. Together, these results demonstrate that oxytocin has complex and dynamic functions in

INTRODUCTION
The wiring of neural circuits is an intricate developmental process, regulated by a combination of intrinsic and extrinsic cues (Katz and Shatz, 1996;Crair, 1999;Sur and Rubenstein, 2005;Blankenship and Feller, 2010). In rodents, wiring of the cerebral cortex occurs mostly during the first 4 weeks of postnatal development (Micheva and Beaulieu, 1996). This process is regulated by genetic programming, in combination with environmental factors (Feldman and Brecht, 2005;Fox and Wong, 2005;Nithianantharajah and Hannan, 2006;Sale et al., 2009). The anatomical and functional properties of neurons in the sensory cortices are particularly sensitive to modification by environmental stimuli during a limited developmental window, known as the ''sensitive period'' (Knudsen, 2004;Luby et al., 2020). An extreme form of sensitive period is the ''critical period'', where appropriate experience is essential for the normal development of a pathway or set of connections (Hensch, 2004). The most well-studied example of the critical period is the formation of ocular dominance columns in the visual cortex (Wiesel and Hubel, 1963). Subsequent studies showed that different aspects of visual cortical development have different sensitive/critical periods (Hensch, 2004;Hooks and Chen, 2007). Other cortical regions also have various sensitive/critical periods for different aspects of their development (Neville and Bavelier, 2002;Erzurumlu and Gaspar, 2012;Kral, 2013).
In previous work, we identified a new form of experiencedependent cross-modal plasticity in the sensory cortices, by showing that deprivation of sensory inputs in one modality cross-modally delayed the development of other sensory cortices (Zheng et al., 2014). Specifically, we showed that deprivation of somatosensory inputs through whisker deprivation (WD) reduced excitatory synaptic transmission in L2/3 pyramidal neurons of both the barrel field of the primary somatosensory cortex (S1BF), and the primary visual cortex (V1), at both P7 and P14. We further showed that the neuropeptide oxytocin, mostly synthesized in the paraventricular nuclei of the hypothalamus (PVH) and the supraoptic nuclei (SON), is an important mediator of this form of plasticity. Specifically, at P14, oxytocin knockout mice had reduced excitatory synaptic transmission, similar to the effects of WD, while perfusion of oxytocin onto acute brain slices or in vivo injection of oxytocin enhanced excitatory synaptic transmission (Zheng et al., 2014).
An important remaining question is whether this form of experience-and oxytocin-dependent plasticity has a sensitive/critical period. Here, we address this question by examining the effect of experience and oxytocin on synaptic transmission on L2/3 pyramidal neurons at different developmental time points. Our results show that this experience-dependent plasticity in the sensory cortices has a critical period ending around P18. We further show that the effect of oxytocin on excitatory and inhibitory synaptic transmission, as well as the expression of oxytocin receptors in the cerebral cortex, change during cortical development. Together, these results demonstrate that the developmental effects of oxytocin on excitatory and inhibitory synaptic transmission are dynamic and complex.

Animals
All animal procedures complied with the animal care standards set forth by the US National Institutes of Health and were approved by the Institutional Animal Care and Use Committee at the Institute of Neuroscience, Chinese Academy of Sciences, and of Peking University. Mice on C57BL/6 background were raised in a specific pathogen-free (SPF) environment and grouphoused under a 12h-12 h light-dark cycle with food and water provided ad libitum from the cage lid. Their health status was monitored routinely.
Recordings were made from 2-3 cells per slice, and 2-3 slices per mouse; for drug bath application experiments, one cell per slice was recorded. For mEPSC recordings, cells were held at −70 mV in voltage-clamp, with pipette resistance of 3-4 MΩ in the presence of tetrodotoxin (TTX, 0.5 µM) and picrotoxin (50 µM) to block Na + channels and GABA A R, respectively. For mIPSC recordings, cells were held at −60 mV in the presence of TTX (0.5 µM) and NBQX (10 µM) to block Na + channels and AMPAR respectively. A brief hyperpolarization (10 mV, 100 ms) was given to monitor series and input resistances every 10 s. Cells with changes of input or series resistance greater than 20% were excluded from analyses. All cells analyzed had a series resistance <25 MΩ. Liquid junction potential and series resistance were uncompensated. Data were analyzed in MiniAnalysis (Synaptosoft, Fort Lee, NJ) with detection thresholds of 5 pA and 6 pA, for mEPSC and mIPSC, respectively. Data were analyzed blinded to the experimental condition.

Fluorescent In situ Hybridization, Immunohistochemistry, and Quantitation
In situ hybridization was performed as previously described (Wu et al., 2009;Xiu et al., 2014;Duan et al., 2018). Oxtr probes were cloned into the BamHI and Frontiers in Cellular Neuroscience | www.frontiersin.org EcoRI sites of pBluescript vector using forward primer cgcggatccGTTGGCACGGGTCAGTAGT, and reverse primer ccggaattcAATGCTTTCTGGGATGTCCTAA. RNA probes were labeled using DIG RNA Labeling Mix (Roche, Cat# 11277073910). Anti-Digoxigenin-AP Fab fragments (Roche, Cat# 11093274910, RRID: AB_514497) were used for DIG labeling of Oxtr. For colocalization with various cell type markers, the following primary antibodies were co-incubated with anti-digoxigenin-AP: CaMKIIβ (Abcam, Cat# ab34703, RRID: AB_2275072) and Somatostatin (SST, Santa Cruz, Cat# sc-7819, RRID: AB_2302603). The following secondary antibodies were used: Donkey anti-Rabbit Alexa Fluor 488 (Thermo Fisher Scientific, Cat# A-21206, RRID: AB_2535792), Donkey anti-Goat Alexa Fluor 488 (Thermo Fisher Scientific, Cat# A-11055, RRID: AB_2534102), both diluted 1:1,000. Fast red (Roche, Cat# 11496549001) was used for visualization of in situ hybridization. GAD67-GFP transgenic mice were used to label GABAergic neurons. Sections were mounted with Fluoromount medium (Sigma-Aldrich, Cat# F4680). S1 Layer 2/3 images (1,024 × 1,024) were acquired on a Nikon A1 confocal microscope with S Fluor 40× Oil DIC H N2 Optics (N.A. = 1.3). Image analysis was performed using Image-Pro Plus (Media Cybernetics, Rockville, MD, USA), blinded to the experimental condition. For colocalization analysis, images of each channel were separately thresholded, and colocalization was defined as one or more pixels of overlap between the two conditions. Oxtr in each cell type was measured as a ratio of total Oxtr area in the section. For measuring the percentage of Oxtr positive marker, the area of Oxtr colocalizing with each marker was ratioed over that of total marker area. Image analysis was carried out with no post-acquisition modifications. For example images, brightness/contrast was adjusted with linear ranges using ImageJ (N.I.H., Bethesda, MD, USA). P14 and P28 sections were adjusted with the same parameters.

Real-Time qPCR and Oxytocin Peptide Measurements
Total mRNA was extracted from the cerebral cortex, hippocampus, and hypothalamus, using TRIzol reagent (Invitrogen, Cat# 15596018). First-strand cDNA was generated using the M-MLV reverse transcriptase (Promega, Cat# M1701) according to the manufacturer's protocols. Real-time qPCR was performed using SYBR Green Master Mix (TaKaRa, Cat# RR420A) on a LightCycler 480 (Roche Applied Science). All reactions were carried out in duplicates, and the comparative C T method was used for comparisons between samples. The following primers were used: Oxytocin peptide concentration from S1, V1 and plasma were measured using an ELISA kit (Pheonix Pharmaceutics, EK-051-01), as previously described (Zheng et al., 2014).

Statistical Analysis
Statistical analysis was performed using GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA). Data are presented as mean ± SEM. Unpaired two-tailed Student's t-test (for sample pairs) or one-way ANOVA (for three or more samples) followed by Tukey's multiple comparison tests were used, depending on the number of samples. For oxytocin perfusion experiments, paired two-tailed Student's t-test were used. For electrophysiological experiments, n represents the number of neurons. For in situ hybridization experiments, n represents the number of brain sections. For other experiments, n represents the number of mice. Typically, three or more mice were used for each experimental condition. Data were analyzed blinded to the experimental condition.
Following dark-rearing (DR), another unimodal sensory deprivation paradigm, similar results were obtained. DR litters showed a significant reduction in mEPSC frequency at the earlier time points of P7 and P14 (Zheng et al., 2014); at P18, however, mEPSC frequency was not different between DR mice and those reared under standard lighting conditions (Ctrl; Supplementary Figures 1C,D), in both S1BF (Ctrl: 1.71 ± 0.27 Hz; DR: 1.52 ± 0.23 Hz; P = 0.62) and V1 (Ctrl: 1.83 ± 0.40 Hz; WD: 1.56 ± 0.28 Hz; P = 0.57). Together, these results suggest that experience-dependent synaptic plasticity in the sensory cortices has a sensitive/critical period ending around P18.
Since we have previously shown that sensory experience regulated excitatory synaptic transmission via the neuropeptide oxytocin (Zheng et al., 2014), we asked if oxytocin regulated excitatory synaptic transmission with a similar sensitive/critical period. We thus bath applied oxytocin (1 µM) onto acute S1BF brain slices of P14, P18, and P28 mice, and measured mEPSC frequency and amplitude of S1BF L2/3 pyramidal neurons, before and after oxytocin application. Consistent with our previous report (Zheng et al., 2014), in P14 mice, oxytocin significantly increased mEPSC frequency (Ctrl: 4.10 ± 0.35 Hz, OXT: 5.43 ± 0.52 Hz; P < 0.001; Figures 1A,B). We also observed a small, but significant, reduction in mEPSC amplitude (Ctrl: 13.01 ± 0.63 pA, OXT: 12.17 ± 0.58 pA; P < 0.05; please see ''Discussion'' section for discussion on all mEPSC amplitude changes). mEPSC frequency, reflecting release probability of individual synapses and total synapse number of the cell, and mEPSC amplitude, reflecting the size of individual synapses, both contribute to total synaptic strength. To more directly measure the effect of oxytocin on the total synaptic inputs of L2/3 neurons, we calculated the total charge transfer per second and found it to be significantly higher following oxytocin application (Ctrl: 148.32 ± 11.23 pAms; OXT: 193.10 ± 16.35 pAms; P < 0.01), consistent with oxytocin increasing total excitatory synaptic input of L2/3 pyramidal neurons at this developmental stage.
In P18 mice, oxytocin application had no significant effects on mEPSC frequency (Ctrl: 5.80 ± 0.47 Hz, OXT: 5.65 ± 0.57 Hz; P = 0.60) or amplitude (Ctrl: 9.71 ± 0.21 pA, OXT: 9.39 ± 0.15 pA; P = 0.08; Figures 1C,D). This result suggested that the effect of oxytocin on promoting excitatory synaptic transmission also had a critical period ending around P18, similar to the sensitive/critical period observed following sensory deprivation by WD or DR (Supplementary Figures  1A-D). Consistently, at P18-P21, sensory deprivation by WD or DR did not significantly reduce oxytocin peptide level, in S1 or V1 (Supplementary Figures 1E,F). This contrasts with significantly reduced oxytocin levels in S1 and V1 of P14 mice following WD or DR (Zheng et al., 2014).

Developmental Changes in Oxytocin Receptor Expression in Different Neuronal Types
What biological changes may account for, or at least contribute to, dynamic changes in the effect of oxytocin on synaptic transmission? Oxytocin primarily signals through the oxytocin receptor (OXTR), a G protein-coupled receptor expressed widely in the brain (Gimpl and Fahrenholz, 2001;Jurek and Neumann, 2018). OXTR expression is developmentally dynamic and is regulated by experience (Vaidyanathan and Hammock, 2017). In the mouse cerebral cortex, Oxtr mRNA and OXTR protein expression, as well as radioligand labeling of receptors, all showed peak receptor expression at P14 (Hammock and Levitt, 2013;Zheng et al., 2014;Mitre et al., 2016). However, it is not known if Oxtr expression is mostly in glutamatergic or GABAergic neurons at this age. We thus performed in situ hybridization of Oxtr mRNA, in combination with immunohistochemistry for the beta subunit of Ca 2+ /calmodulin-dependent protein kinase II (CaMKIIβ) or somatostatin (SST), respectively labeling glutamatergic (excitatory) neurons or a subclass of GABAergic (inhibitory) neurons previously shown to express Oxtr (Nakajima et al., 2014). Oxtr in situ hybridization was also carried out using GAD67-GFP mice, in which GABAergic neurons are labeled with the green fluorescent protein (GFP; Tamamaki et al., 2003).
In S1, at both P14 and P28, Oxtr mRNA partially colocalized with all three markers (Figures 3A-C); colocalization was defined as the overlap between the two signals at the pixel level. Between P14 and P28, the distribution of Oxtr mRNA changed, from mostly colocalizing with CaMKIIβ at P14 (CaMKIIβ:   73.56 ± 2.51%; GAD67: 26.0 ± 2.06%; SST: 20.45 ± 1.07%), to relatively even distribution between glutamatergic and GABAergic neurons at P28 (CaMKIIβ: 46.69 ± 3.42%; GAD67: 46.99 ± 3.39%; SST: 26.28 ± 1.98%; Figure 3D). In addition, the ratio of Oxtr-expressing cells decreased from P14 to P28 in both glutamatergic neurons (P14: 79.79 ± 2.04%; P28: 47.98 ± 6.17%) and GABAergic neurons (P14: 64.39 ± 4.43%; P28: 35.74 ± 4.86%; Figure 3E), consistent with previous reports showing peak Oxtr expression in the cerebral cortex at P14. At both time points, a very high proportion of SST neurons expressed Oxtr (P14: 95.39 ± 1.12%; P28: 87.96 ± 2.66%; Figure 3E). These results show that Oxtr expression is dynamic during development, and has distinct expression patterns at P14 and P28. Not having a specific OXTR antibody on hand, we confirmed our in situ results using the ''oxytocin binding'' method. Persistent activation of OXTR, a G protein-coupled receptor, leads to its endocytosis and internalization, together with its ligand oxytocin (Gimpl and Fahrenholz, 2001); thus cells expressing functional OXTR have significant oxytocin binding capacity and can be labeled using an antibody against oxytocin following ligand binding. In N2a cells, application of oxytocin resulted in specific labeling of OXTR-expressing cells with oxytocin antibody, but not neighboring cells not expressing OXTR; this effect was blocked by pre-incubation with oxytocin antibody and did not occur upon incubation with the closely related neuropeptide vasopressin (Supplementary Figure 2). We then treated acute brain slices with oxytocin for 20 min, fixed the brain slices, and immunostained for oxytocin and the pan-neuronal marker NeuN (Figure 4). In S1 of P14 mice, oxytocin immunoreactivity, marking cells with internalized OXTR, colocalized significantly with NeuN and was relatively high in the superficial layers of the cerebral cortical (layers 2/3), as compared to the deeper layer (layer 5; Figure 4A). As a control for specificity, cortical brain slices incubated with the closely related neuropeptide vasopressin were not labeled with oxytocin antibody (Figure 4B). In S1 of adult mice, the oxytocin immune-reactive cells distributed relatively evenly across superficial and deeper layers ( Figure 4C). This dynamic pattern of OXTR expression in different cortical layers during development is consistent with a recent report using an OXTR reporter mouse line (Newmaster et al., 2020). The results of these oxytocin labeling experiments are also consistent with those of our Oxtr in situ hybridization experiments, although the labeling efficiency is lower. Specifically, the proportion of oxytocin immune-reactive cells that are GAD67-positive increased from 15.31 ± 0.30% at P14 to 29.17 ± 4.17% in adult mice. In both age groups, over 90% (P14: 90.30 ± 3.33%; adult: 90.28 ± 1.39%) of oxytocin immune-reactive cells co-labeled with NeuN, consisting with high OXTR expression in neurons (Figures 4D,E). In addition to the cerebral cortex, internalized oxytocin also colocalized with NeuN in the hippocampus, amygdala, and lateral septum of P14 mice (Supplementary Figures 3A-C). Oxt -/-, 1.98 ± 0.12 Hz; n = 20; P < 0.05 vs. Ctrl), while mEPSC amplitude is not significantly affected (Ctrl, 10.95 ± 0.38 pA; Oxt +/− , 10.25 ± 0.22 pA; Oxt -/-, 10.74 ± 0.35 pA). One-way ANOVA followed by Tukey's multiple comparison test for all panels. *P < 0.05, **P < 0.01, unpaired t-test.
Labeling of cortical neurons was also achieved using FITC-oxytocin (see ''Materials and Methods'' section for details; Supplementary Figure 4A). Importantly, FITC-oxytocin did not label cells in the PVH of EIIa-Cre; Oxtr fl/fl (EIIa-Cre; Oxtr cKO) mice, where Oxtr is removed from the very early embryo (Lakso et al., 1996;Lee et al., 2008), thus demonstrating the requirement of OXTR for oxytocin binding in vivo (Supplementary  Figures 4B,C).

Oxytocin and Sensory Experience Have Similar Sensitive/Critical Periods in Regulating Excitatory Synaptic Transmission of L2/3 Neurons
Here, we found that sensory experience and oxytocin regulate excitatory synaptic transmission in L2/3 pyramidal neurons of the sensory cortices with a similar sensitive/critical period, peaking around P14 and ending around P18 (Figure 1 and Supplementary Figures 1A-D; Zheng et al., 2014). Consistently, sensory experience regulates oxytocin expression with a similar time course, elevation at P14, and essentially no changes at P18 (Supplementary Figures 1E,F; Zheng et al., 2014). Curiously, we observed an increase in oxytocin level in S1 of DR mice (Supplementary Figure 1F), possibly due to homeostatic compensation. The effects of sensory experience and oxytocin on excitatory synaptic transmission are both directional, with sensory deprivation and loss-offunction of oxytocin reducing excitatory synaptic transmission, and environmental enrichment and exogenous oxytocin application increasing synaptic transmission (Figures 1, 5 and Supplementary Figure 1; Zheng et al., 2014). The above evidence suggests that oxytocin may function as a mediator of early experience-dependent plasticity in L2/3 pyramidal neurons of the sensory cortices.
As to what biological change ends the sensitive/critical period, we can only speculate. Oxytocin expression in the hypothalamus increases steadily between P7 and P60 (Zheng et al., 2014), thus it is unlikely that a sharp change in oxytocin expression leads to closure of this sensitive/critical period. Oxtr mRNA, OXTR protein, oxytocin binding capacity, and OXTR reporter expression in the sensory cortices has been reported to peak around P14 and drops significantly at P21 and P28 (Hammock and Levitt, 2013;Mitre et al., 2016;Newmaster et al., 2020; Figure 3). Thus, a reduction in Oxtr expression may contribute to the closure of the sensitive/critical period. It is probably one of many factors that contribute. To better understand this form of experience-dependent plasticity in the sensory cortices, more mechanistic studies, as well as a deeper understanding of the physiological function of this sensitive/critical period is needed.

Effects of Oxytocin on Synaptic Transmission at P14
At P14 (and the earlier time point of P10), loss-of-function of oxytocin reduces mEPSC frequency of S1BF L2/3 pyramidal neurons, while its application to acute brain slices increases mEPSC frequency (Figures 1, 5). Furthermore, loss-of-function of Oxtr blocks oxytocin-induced increase in mEPSC frequency (Figure 6). Oxtr expression is high in L2/3 glutamatergic neurons and low in GABAergic neurons at this age (Figures 3, 4). Consistently, oxytocin application does not significantly affect inhibitory synaptic transmission. Increased excitatory synaptic transmission and no change in inhibition add to an increase in total excitatory input of L2/3 pyramidal neurons.
An interesting fine point is the difference between the effect of Oxt -/knockout and Oxtr cKO: Oxt -/mice have reduced excitatory synaptic transmission under baseline conditions, at both P10 and P14 (Figure 5), while in EIIa-Cre; Oxtr cKO or Nex-Cre; Oxtr cKO mice, baseline excitatory synaptic transmission is unaffected, but the effect of oxytocin application is blocked (Figure 6). Given that EIIa-Cre is expressed from the early embryo, before implantation (Lakso et al., 1996), EIIa-Cre; Oxtr cKO mice should have completely or near complete knockout of Oxtr. This opens up the possibility that oxytocin may have developmental effects independent of its receptors.
Another subject of interest is whether oxytocin affects mEPSC amplitude, a parameter that correlates with the size of individual synapses. In Oxt -/mice, and in Oxtr cKO mice, mEPSC amplitude was similar between knockout mice and littermate controls (Figures 5, 6). Thus loss-of-function of Oxt or Oxtr does not affect mEPSC amplitude. In all oxytocin application experiments, however, a small but significant reduction in mEPSC amplitude was often observed (Figures 1, 6). A small reduction in mEPSC amplitude, in addition to being a biological phenomenon, could also be an artifact due to increased serial resistance during the course of whole cell patch-clamp recordings. To minimize this problem, we only analyzed recordings in which series and input resistances changed by less than 20% over the course of the experiment. Also, in P18 mice, neither mEPSC amplitude nor frequency was affected by oxytocin application (Figures 1C,D). Thus, the small reduction in mEPSC amplitude following oxytocin application is likely to be a bona fide biological phenomenon. However, since the magnitude of the increase in mEPSC frequency is much larger than the reduction in mEPSC amplitude, the total charge transfer, representing total excitatory input of the neuron, is increased following oxytocin application at P14.

Effects of Oxytocin on Synaptic Transmission at P28
In P28 mice, oxytocin application reduces mEPSC frequency and increases mIPSC frequency (Figures 1, 2) of S1BF L2/3 pyramidal neurons. The effect of acute oxytocin application on reducing mEPSC frequency was confirmed by in vivo oxytocin injection (Figures 1H-J) and blocked in Nex-Cre; Oxtr cKO mice (Figures 6G,H).
These results, together with those of P14, suggest that oxytocin has distinct effects on synaptic transmission at P14 and P28: increasing total excitatory inputs (increased mEPSC frequency and no changes in inhibition) of L2/3 pyramidal neurons at P14, and reducing it (reduced mEPSC frequency and increased mIPSC frequency) at P28. What physiological changes may underlie these switches? As discussed above, oxytocin expression increases steadily between P7 and P60 (Zheng et al., 2014), and thus is unlikely to account for the above described switch. Our in situ hybridization and oxytocin binding results, together with published data using a variety of approaches to measure OXTR level (Hammock and Levitt, 2013;Mitre et al., 2016;Newmaster et al., 2020), suggest that OXTR level is higher at P14, as compared to P21 and P28. It is relatively straightforward for high receptor expression to be associated with a higher level of signal transduction and increased transmission (e.g., P10 and P14), and lower receptor expression to be associated with no changes (e.g., P18); however, the change from increase at P14 to reduction at P28 presumably requires additional alterations in downstream signal transduction components.
In terms of GABAergic synaptic transmission, our results (Figures 3, 4) suggest increased relative expression of OXTR in GABAergic neurons in P28 and older mice, as compared to P14. An increase in OXTR expression in GABAergic cells would presumably increase OXTR-dependent signaling in these cells. Since we observed an increase in GABAergic input to L2/3 pyramidal neurons at P28, the increased OXTR expression likely enhanced the synaptic output of GABAergic neurons onto pyramidal neurons. A recent study indeed showed that oxytocin can enhance the excitability of SST neurons, thereby reducing the level of spontaneous activity (Maldonado et al., 2021).
Thus, combining the results on glutamatergic and GABAergic transmission, it seems that lower expression of OXTR is associated with no effects on synaptic transmission, while higher OXTR expression is associated with increased transmission. The exception is excitatory synaptic transmission in P28 and older mice, where oxytocin application significantly reduces excitatory synaptic transmission in neurons expressing a low level of OXTR. We hypothesize that change in the level of one or more OXTR downstream signaling component mediates this effect.
We focused in S1 L2/3 pyramidal neurons for this study. Recent work showed that oxytocin affects spontaneous network events differentially in S1 and V1 (Maldonado et al., 2021). Given the complexity of the effects of oxytocin, studies on more cell types, more brain regions, and at more developmental stages, as well as more in-depth investigations of OXTR downstream signaling under these different conditions, are needed for a full understanding of its function.

Implications for Neural Circuit Development and Plasticity
Oxytocin has been shown to affect many aspects of neural circuit development and function, including regulating excitatory or inhibitory synaptic transmission, altering neuronal firing rates and patterns, and modulating the transition of GABA from excitatory to inhibitory (Stoop, 2012;Hammock, 2015;Marlin and Froemke, 2017;Ben-Ari, 2018). Our study adds to existing knowledge by showing that oxytocin regulates the excitatory and inhibitory synaptic transmission of L2/3 pyramidal neurons in a developmentally dynamic manner. An immediate implication of this finding is that giving the same dose of oxytocin to an individual may have different effects, sometimes opposite, depending on the developmental stage of the individual.
Because oxytocin can promote trust, eye contact, and facial memory, it has been proposed as a therapy for the treatment of autism spectrum disorders (ASD), a developmental disorder with deficits in social communication (Green and Hollander, 2010;Insel, 2010;Meyer-Lindenberg et al., 2011;Yamasue et al., 2012;Miller, 2013). A very large proportion of individuals with ASD also are hypo-or hypersensitive to sensory inputs (Marco et al., 2011;Suarez, 2012), consistent with the function of oxytocin in regulating cortical neural circuit wiring. However, clinical trials investigating the effectiveness of oxytocin as a treatment for ASD reported mixed results (Guastella and Hickie, 2016;Ooi et al., 2017;Keech et al., 2018;Huang et al., 2021). The dynamic developmental effects of oxytocin likely add to the difficulty of obtaining consistent results. Given the heterogeneity of ASD, the developmental switch point at which oxytocin function shifts from being overall excitatory in L2/3 pyramidal neurons to overall inhibitory may be different for different individuals. If we consider that oxytocin regulates the function of many types of neurons and that many of these functions may be developmentally dynamic and brain-region specific, the situation becomes exceedingly complex.
In addition to identifying complexity, what potential directions do we see moving forward? First, a deeper understanding of oxytocin-OXTR downstream signaling, as well as of other pathways mediating oxytocin signaling, would be important to understand the diversity of its physiological functions. Second, taking developmental stage/age into account may contribute towards more consistent results in both animal and human studies. The developmental stage/age may need to be defined functionally, and therapies may have to optimize treatment windows. Third, especially for patient studies, attempts to subtype or subclass may reduce the heterogeneity of outcomes. In the end, we hope that deeper mechanistic insights eventually translate to effective therapies for patients.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Institutional Animal Care and Use Committee at the Institute of Neuroscience, Chinese Academy of Sciences and of Peking University.