Eye opening differentially modulates inhibitory synaptic transmission in the developing visual cortex

Eye opening, a natural and timed event during animal development, influences cortical circuit assembly and maturation; yet, little is known about its precise effect on inhibitory synaptic connections. Here, we show that coinciding with eye opening, the strength of unitary inhibitory postsynaptic currents (uIPSCs) from somatostatin-expressing interneurons (Sst-INs) to nearby excitatory neurons, but not interneurons, sharply decreases in layer 2/3 of the mouse visual cortex. In contrast, the strength of uIPSCs from fast-spiking interneurons (FS-INs) to excitatory neurons significantly increases during eye opening. More importantly, these developmental changes can be prevented by dark rearing or binocular lid suture, and reproduced by the artificial opening of sutured lids. Mechanistically, this differential maturation of synaptic transmission is accompanied by a significant change in the postsynaptic quantal size. Together, our study reveals a differential regulation in GABAergic circuits in the cortex driven by eye opening may be crucial for cortical maturation and function.


Introduction
During neocortical development, sensory experience critically influences neuronal connectivity and synaptic transmission (Katz and Shatz, 1996). In the visual system, maturation of the visual cortex depends on visual afferent activity (Hensch, 2005;Hofer et al., 2009;Pecka et al., 2014). The initial visual inputs experienced through closed eyelids are imprecise and diffuse, and followed by patterned visual inputs after eye opening. In general, experience-dependent maturation of the visual cortex is a gradual process, but it accelerates soon after eye opening when the intensity and frequency of afferent visual activity suddenly and rapidly increase (Lu and Constantine-Paton, 2004). Eye opening in rodents typically occurs over a short period of one to two days (Gandhi et al., 2005). When experimentally synchronized, eye opening drives a rapid series of changes in neuronal activity, protein trafficking, synaptogenesis, synaptic receptor composition, and synaptic transmission and plasticity (Lu and Constantine-Paton, 2004;Yoshii et al., 2003;Zhao et al., 2006). Moreover, eye opening affects neuronal circuit development in the ascending retinothalamo-cortical pathway at every level: the retina (Feller, 2003;Tian and Copenhagen, 2003), superior colliculus (SC) (Zhao et al., 2006), lateral geniculate nucleus (LGN) (Hooks and Chen, 2006;Levitt et al., 2001), and visual cortex (Hoy and Niell, 2015;Ko et al., 2013;Pecka et al., 2014). For example, the probability and strength of excitatory connections between layer 2/3 pyramidal cells (PCs) in the visual cortex significantly increased after eye opening, and these changes were prevented by dark rearing (Ishikawa et al., 2014). Although previous study reported that orientation tuning preferences of fast-spiking interneurons (FS-INs) were dependent on normal visual experience after eye opening (Kuhlman et al., 2011), whether eye opening shapes inhibitory synaptic transmission in the neocortex during development remains unclear.
In the mature neocortex, GABAergic INs can be categorized into three main populations based on the expression of the calcium-binding protein parvalbumin (PV), the neuropeptide somatostatin (Sst), and the ionotropic serotonin receptor (Lee et al., 2010). Sst-expressing INs (Sst-INs), comprising approximately 20~30% of all neocortical INs, are among the most prominent GABAergic IN subtypes in the neocortex (Lee et al., 2010;Pfeffer et al., 2013). Sst expression in the superficial neocortex has typically been associated with Martinotti cells (MCs)-GABAergic INs with ascending axons that arborize in cortical layer one and spread horizontally to neighboring columns (Fino and Yuste, 2011;Ma et al., 2006). A large fraction of Sst-INs preferentially target distal dendrites of PCs (Di Cristo et al., 2004). In addition to providing lateral inhibition to local PC networks, cortical Sst-INs also precisely control the efficacy and plasticity of glutamatergic inputs by regulating postsynaptic spine Ca 2+ signals, synaptic dynamics, dendritic spike bursts, and transformation of dendritic inputs (Chiu et al., 2013;Higley, 2014;Lovett-Barron et al., 2012). Furthermore, cortical Sst-INs not only innervate PCs but also frequently inhibit other cortical INs (disinhibition) (Pfeffer et al., 2013). PV-expressing FS-INs, which constitute~40% of cortical GABAergic neurons, form powerful synapses onto the somatic and perisomatic compartments of PCs (Di Cristo et al., 2004). Although the emergence and maturation of connections from PV-INs to PCs and PV-INs have been reported previously (Lazarus and Huang, 2011); Pangratz-Fuehrer and Hestrin, 2011); Yang et al., 2014), the development of inhibitory synaptic transmission (including amplitude and connectivity) from Sst-INs to PCs and other IN subtypes remains largely unclear. In addition, accumulating evidence indicates that Sst-INs regulate learning-induced and experience-dependent cortical plasticity through feedforward as well as feedback inhibitions, both during early postnatal development and in adulthood (Bloodgood et al., 2013;Chen et al., 2015;Marques-Smith et al., 2016;Oh et al., 2016;Tuncdemir et al., 2016). There is also clear evidence suggesting that PV-INs regulate critical-period experience-dependent plasticity (Hensch, 2005;Krishnan et al., 2015). These observations raise important questions: Does eye opening play a role in the regulation of inhibitory synaptic transmission from Sst-INs and FS-INs to PCs or other types of INs, and if so, what is the synaptic mechanism underlying the regulation?
In this study, we demonstrate that eye opening rapidly weakens the inhibitory synaptic transmission from Sst-INs to PCs, whereas it increases inhibitory synaptic transmission from FS-INs to PCs. Moreover, we show that the maturation of inhibitory synaptic transmission is mediated by differential changes in the postsynaptic quantal size.

Rapid weakening of synaptic transmission from Sst-INs onto PCs coincides with eye opening
To explicitly identify Sst-INs in the neocortex, we crossed the Sst-IRES-Cre mice with the loxPflanked Rosa26reporter-tdTomato mice. Sst-INs in the resulting progeny expressed red fluorescent protein tdTomato in the brain, and this facilitated electrophysiological recordings of Sst-INs (Taniguchi et al., 2011). We focused on layer 2/3 of the primary visual cortex. Consistent with previous reports (Hu et al., 2013;Pfeffer et al., 2013), we observed that~5.2% of tdTomato + neurons expressed PV (5.2 ± 0.3%, nine slices from three mice, Figure 1-figure supplement 1B). Meanwhile,~17.2% of tdTomato + neurons showed the FS properties (36 out of 209, Figure 1-figure supplement 1C) (Hu et al., 2013;Jiang et al., 2015). Furthermore, FS tdTomato + cells exhibited the distinctive basket cell morphology (4 out of 4, Figure 1-figure supplement 1C). These cells were omitted from further analysis. Non-FS tdTomato + cells were further characterized by morphological properties, including ascending axonal arborizations with extensive branching in layer one and horizontal collaterals (19 out of 20, Figure 1-figure supplement 1A), consistent with those of Martinotti cells as previously described (Fino and Yuste, 2011).

Eye opening modulates the Sst-INfiPC and FS-INfiPC synaptic transmission
Although the weakening of synaptic transmission from Sst-INs to PCs was observed at the time of eye opening, it could be induced by intrinsic developmental programs or other mechanisms with coincidental timing rather than by eye opening per se. To determine which factor is responsible for this regulation of synaptic transmission, we first deprived the visual inputs by dark rearing. We darkreared Sst-tdTomato mice from P3 and recorded the synaptic transmission from Sst-INs to PCs in layer 2/3 of the primary visual cortex at P12-15 ( Visual deprivation achieved by dark rearing not only blocks eye opening-induced visual inputs but also eliminates the natural diffuse dark/light stimulation presented through the eyelids before eye opening. Therefore, it is difficult to determine whether the elimination of the weakening of Sst-INfiPC synaptic transmission is induced by eye opening deprivation or by visual deprivation. To address this question, we performed binocular lid suture from P8 to block eye opening and Source data 1. Detailed statistical analysis, detailed data, exact sample numbers, and p values in Figure 4 and For Sst-INfiPC synaptic transmission within the visual cortex (VC), the connection probability was not significantly different among the different groups (c 2 test, p=0.958; Figure 6B). Consistent with dark rearing, there was no significant change in the peak amplitude of uIPSCs in continuously sutured mice between P12-13 and P14-15 ( Figure 6C). Moreover, in continuously sutured mice, the strength of Sst-INfiPC uIPSCs at P17-20 was comparable to that at P12-13 or P14-15 (P12-13, 30.1 ± 5.9 pA; P14-15, 25.2 ± 4.1 pA; P17-20, 17.5 ± 10.6 pA; Figure 6C). These results suggest  that eye opening deprivation prevents the weakening of Sst-INfiPC synaptic transmission. Interestingly, the strength of Sst-INfiPC uIPSCs at P17-20 in mice with eyelids artificially opened was significantly lower than that in continuously sutured mice at P12-13, P14-15, and P17-20 (one-way ANOVA, F (3,68) = 4.231, p=0.008; Figure 6C). These results suggest that artificial eye opening can induce the weakening of Sst-INfiPC synaptic transmission and that this change associated with eye opening is unlikely due to an intrinsic developmental program. Notably, similar findings were observed in the prefrontal Cg1/2 area. In Cg1/2, the probability of Sst-INfiPC connections was not significantly changed among different groups (c 2 test, p=0.987; Figure 6D). The strength of Sst-INfiPC uIPSCs remained unchanged in continuously sutured mice at P12-13, P14-15, and P17-20 ( Figure 6E). Furthermore, the strength of Sst-INfiPC uIPSCs at P17-20 in mice with eyelids artificially opened was significantly smaller than that in continuously sutured mice at P12-13, P14-15, and P17-20 (artificially eye-opened mice: P17-20, 13.6 ± 2.6 pA versus continuously sutured mice: P12-13, 37.3 ± 5.7 pA; P14-15, 54.6 ± 14.7 pA; and P17-20, 39.9 ± 8.3 pA; one-way ANOVA, F (3,39) = 5.607, p=0.003; Figure 6E). In addition, both the connection probability and strength of Sst-INfiFS-IN synaptic transmission were comparable at P12-13 and P14-15 in continuously sutured mice (Figure 6-figure supplement 3).
Together, these results strongly suggest that eye opening differentially modulates the strength of Sst-INfiPC and FS-INfiPC synaptic transmission.
Together, these results suggest the postsynaptic quantal sizes in both of Sst-INfiPC and FS-INfiPC synaptic transmission are altered during eye opening.

Discussion
The plasticity of GABAergic circuitry in the visual critical period in the developing visual cortex has been extensively studied by manipulations that disrupt normal visual experience after eye opening (Griffen and Maffei, 2014;Hensch, 2005;Lefort et al., 2013;Maffei et al., 2004). However, the changes and plasticity of GABAergic circuitry during eye opening are still far from fully understood (Gandhi et al., 2005;Kuhlman et al., 2011;Maffei et al., 2004). In this study, we show the following: (1) eye opening weakens the synaptic transmission from Sst-INs to PCs, but increases the synaptic transmission from FS-INs to PCs; (2) the inhibitory synaptic transmission from Sst-INs to other types of interneurons remains unaltered during eye opening; (3) eye opening-induced alteration of the inhibitory synaptic transmission onto PCs is mediated by changes in postsynaptic quantal size.
We studied the formation of inhibitory synapses onto layer 2/3 PCs and focused on Sst-INs and FS-INs. Although Sst-IRES-Cre line has been widely used to study Sst interneurons in cortical layer 2/ 3 both in vivo and in vitro, the neurons targeted by this line are heterogeneous (Hu et al., 2013;Jiang et al., 2015). Indeed, we observed that~5.2% of Sst-tdTomato neurons in cortical layer 2/3 of visual cortex express PV and~17.2% of Sst-tdTomato neurons show the fast-spiking properties. In spite of this heterogeneity, after removing these fasting-spiking tdTomato + neurons, we observed that the vast majority of tdTomato + cells in cortical layer 2/3 of the neocortex are Martinotti cells (95%, 19 out of 20). We exploited the Sst-tdTomato::Lhx6-EGFP line and identified FS-INs by EGFP + /tdTomatoand fast-spiking properties in this study. Unlike PV-INs that innervate the cell body and proximal dendrites of the layer 2/3 PCs, Sst-INs innervate the distal regions of PCs, including the apical dendrites (Chen et al., 2015;Di Cristo et al., 2004). Notably, the synaptic strength measured in our study represents the strength at the soma of the recorded neuron rather than at the contact sites. These somatic recordings undoubtedly underestimate the distal dendritic tonic currents due to attenuation within the dendrites and limited spatial reach of somatic voltage clamp (Williams and Mitchell, 2008). Since significant differences were observed in neither the total length and complexity of PC dendrites (Figure 1-figure supplement 3) nor the rise time and half-width of uIPSCs between P12-13 and P14-15 mice ( Figure 1F and G), the relative change in the specific Sst-INfiPC connection strength is unlikely due to space-clamp bias. Moreover, the connection probability and strength of Sst-INfiPC uIPSCs exhibit similar developmental properties when we used a cesium-based intracellular solution (improve space clamp) containing a high concentration of Cl -(increase the driving force of uIPSCs) to record the postsynaptic currents.
In addition, we observed that the development of inhibitory synaptic transmission onto PCs displays several distinct features in layer 2/3 of mouse neocortex. First, the strength of Sst-INfiPC synaptic connections is rapidly reduced by~65% from P12-13 to P14-15. Although the developmental synaptic transmission from Sst-INs onto PCs has not been quantified systematically, previous work reported that connection strength from low-threshold spiking interneurons (putative Sst-INs) to excitatory spinal neurons increases from P12-13 to P14-15 in layer 4 of the rat somatosensory cortex (Long et al., 2005). In contrast, a recent study found that both the connection probability and strength of Sst-INfiPC synaptic connections decrease substantially from P14-15 to P20-22 within layer 4 of the mouse visual cortex (Miao et al., 2016). The discrepancies may be due to layer differences (see below). Unlike Sst-INfiPC pairs, our data show that the strength of FS-INfiPC synaptic inputs in layer 2/3 of visual cortex rapidly increases by~140% from P12-13 to P14-15. However, FS-INfiPC unitary conductance was reported to remain unaltered after P8 in layer 5/6 of mouse visual cortex (Pangratz-Fuehrer and Hestrin, 2011). Similarly, Yang et al. observed that the strength of FS-INfiPC uIPSCs is not changed after P9 in layer 5/6 of mouse prefrontal cortex (Yang et al., 2014). These studies imply that FS-INfiPC synaptic transmission does not change in the cortical layer 5/6 during eye opening. In addition, the rapid change of synaptic transmission from Sst-INs and FS-INs onto PCs coincides with the onset of eye opening. However, we cannot determine the exact temporal sequence of the two events (change of synaptic transmission and eye opening) due to the variability in the timing of eye opening (1-2 d) and synaptic responses. Nonetheless, with a controlled eye opening paradigm (Lu and Constantine-Paton, 2004;Yoshii et al., 2003), it will be interesting to further explore the sequence of events within the first 24 hr after eye opening. Lastly, the weakening of Sst-INfiPC synaptic transmission and the strengthening of FS-INfiPC synaptic transmission during eye opening exist not only in layer 2/3 of the visual cortex but also in layer 2/3 of the prefrontal Cg1/2 area. Indeed, eye opening has been shown to affect hippocampal development, and early eye opening accelerated the maturation of synaptic strength (Dumas, 2004). Nevertheless, it still remains unclear whether these changes are induced by light-mediated factors (vision in general) or vision-related factors (e.g. mobility directly or indirectly induced by the siblings or mother).
A major finding from our recordings in layer 2/3 of the visual cortex is that natural eye opening regulates synaptic transmission from Sst-INs and FS-INs onto PCs. This conclusion is based on three lines of experimental evidence. Firstly, visual deprivation (dark rearing) at an early postnatal period can prevent the weakening of Sst-INfiPC synaptic transmission. Secondly, eye opening deprivation (binocular lid suture) can efficiently prevent the changes in both of Sst-INfiPC and FS-INfiPC synaptic transmission. More importantly, we controlled the timing of eye opening by artificially opening the lids from the binocular lid-sutured mice two days after natural eye opening (P16) and compared synaptic responses between siblings with and without eye opening. Our data show that artificially opening the eyes decreases Sst-INfiPC synaptic transmission and increases FS-INfiPC synaptic transmission. These results strongly suggest that eye opening can specifically regulate the Sst-INfiPC and FS-INfiPC synaptic transmission in layer 2/3 of the visual cortex. Of note, a recent study found that although the synaptic strength of Sst-INfiPC connection significantly decreases from P14-15 to P20-22 within layer 4 of the mouse visual cortex, dark rearing does not affect the weakening of Sst-INfiPC synaptic transmission (Miao et al., 2016). Moreover, it has been reported that brief monocular visual deprivation during the time of eye opening selectively reduces the strength of synaptic transmission from PV-INs to PCs in layer 4 of the visual cortex (Maffei et al., 2004). These results suggest that visual deprivation-induced change in GABAergic circuits is layerand cell type-specific. Indeed, accumulating evidence suggests that there are striking differences in morphology, intrinsic electrophysiological properties, and synaptic connectivity between layer 2/3 and layer 4 Sst-INs Xu et al., 2013).
Interestingly, no significant changes were observed in the peak amplitude of uIPSCs from Sst-  (Chittajallu et al., 2013;Lee et al., 2010;Pfeffer et al., 2013). It will be interesting to further investigate the development of synaptic transmission from Sst-INs to various subtypes of Htr3a-INs during eye opening.
Our data suggest that postsynaptic mechanisms may contribute to the developmental change of both Sst-INfiPC and FS-INfiPC synaptic strength. It is well known that GABA A receptors (including different subunits) mediate the majority of inhibitory synaptic transmission in the mammalian cortex (Bowery and Smart, 2006). In addition, different subunits of GABA A receptor are selectively inserted at specific GABAergic synapses (Ali and Thomson, 2008;Nusser et al., 1996). Moreover, previous studies indicated that the expression of GABA A receptor subunits in neocortex changes significantly during early postnatal development (Bowery and Smart, 2006;Fritschy et al., 1994;Heinen et al., 2004). These findings suggest that the changes in subunit expression and/or composition of the GABA A receptor may induce the differential developmental alternations of Sst-INfiPC and FS-INfiPC synaptic strength during eye opening.
The physiological roles for the differential alterations of inhibitory synaptic transmission onto PCs during eye opening remain unclear. Growing evidence suggests that Sst-INs that densely innervate nearby PC dendrites in mouse cortical layer 2/3 are responsible for controlling the efficacy and plasticity of synaptic inputs (Chen et al., 2015;Chiu et al., 2013). Given that in early postnatal life, GABAergic transmission is excitatory to immature postsynaptic neurons (Ben-Ari, 2002;Owens et al., 1996), early emergence of Sst-INfiPC synaptic transmission may enhance the excitability of PCs, thereby promoting their maturation and synaptogenesis (Oh et al., 2016). However, around the time of eye opening, Sst-INs would inhibit PCs. Therefore, decreased Sst-INfiPC inhibition after eye opening could enhance the effect of visual input onto excitatory neurons in the visual cortex by facilitating dendritic events in distal regions (Figure 8). Contrary to Sst-INs, FS-INs control the spike output of PCs by inhibiting their perisomatic sites. Increased FS-INfiPC inhibition after eye opening is a homeostatic response to the reduction of Sst-IN inhibition and the resulting increase in the excitability of PCs (Bloodgood et al., 2013;Chen et al., 2015). We speculate that such enhanced visual input to the visual cortex gated by Sst-INs and the homeostatic rebalancing of inhibition regulated by FS-INs might be important for visual integration, an essential step in visual perception.

Dark rearing and eyelid suture
For dark rearing, pups were raised in dark cages after P3 until sacrificed for in vitro recordings. For eyelid suture, P8 mice were first carefully anesthetized with isoflurane and disinfected with ethanol. The binocular eyelids were sutured with small sterile ophthalmic needles. For artificial eye opening, sutured mice were anesthetized with isoflurane, and the eyelids were carefully opened at P16. After eyelid suture or artificial eye opening, the eyelids were covered with tetracycline ointment, and the pups were kept on warm blankets until fully recovered. If the eyelids of sutured mice were unexpectedly open before recording, the pups were discarded and not included in the recording experiment. (Hu et al., 2013;Pangratz-Fuehrer and Hestrin, 2011). For immature neurons (P5-11), fast-spiking properties were characterized by subthreshold oscillations (Pangratz-Fuehrer and Hestrin, 2011). The variance-mean analysis was performed as previously reported (Mitra et al., 2011;Scheuss and Neher, 2001). Recordings were first carried out in ACSF containing 2 mM Ca 2+ /2 mM Mg 2+ , and then the chamber solution was changed to ACSF containing 3.7 mM Ca 2+ /0.3 mM Mg 2+ and 1 mM Ca 2+ /3 mM Mg 2+ . Trains of 3 action potentials at 20 Hz were elicited, and 30-40 repeated sweeps were recorded, with 10-20 s sweep-to-sweep interval. Recordings with stable baseline were used for analysis. The mean (M) and variance (V) of uIPSC amplitude were calculated for each pulse. The relationship between M and V was fitted to the parabola V = QM À M 2 /N (Q, quantal size; N, number of release sites). Quadratic regression was performed with GraphPad Prism five software (GraphPad Software). Only recordings with R 2 >0.45 (R, regression index) were included for analysis.

Quantification and statistical analysis
Data were analyzed with SPSS 22 software (IBM) and GraphPad Prism five software (GraphPad Software). Statistical significance between groups was tested by two-tailed one-sample t-test, two-tailed unpaired t-test, paired t-test, Mann-Whitney U test, Fisher's exact test, c 2 test, one-way ANOVA and two-way ANOVA. All the detailed test methods, the number of experiments and p values are listed in the source data. All data are presented as mean ±SEM, and the difference was recognized as significant when p<0.05. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.