Our modeling and experimental studies were driven by a desire to understand the role of GABAergic inhibition in regulating synaptic plasticity, especially ocular dominance (OD) plasticity in the visual cortex in response to contralateral eye deprivation. We first examined whether synapses in the mouse primary visual cortex are modifiable by STDP.
Layer 2/3 synapses in V1 are modified by STDP prior to and at the peak of the critical period of OD plasticity
Precocious OD plasticity can be triggered by enhancing GABAergic transmission within V1 
, suggesting that the machinery for OD plasticity is operational before its natural onset, but lies dormant until local GABAergic inhibitory circuits mature. To examine whether the basic mechanisms for synaptic plasticity are present at glutamatergic connections prior to the onset of OD plasticity, we compared the ability to induce STDP at layer 2/3 synapses in acute slices of V1 prior to the onset (postnatal day 16–18) and at the peak (P26–30) of OD plasticity. Long-term synaptic depression (LTD) or potentiation (LTP) was induced using a STDP protocol. Postsynaptic action potentials were evoked by current injection from the recording electrode, bypassing the need for inputs to summate to bring the cell to spike threshold. Whole-cell current-clamp recordings were made from layer 2/3 pyramids (), in the presence of 10 µM picrotoxin. EPSPs were continuously evoked at a frequency of 0.2 Hz throughout the experiment from a field electrode placed in layer 4. EPSPs were monitored for a baseline period of 5 minutes, then paired 100 times with an action potential (AP), and further monitored for 20–45 minutes. To induce LTD, the AP was timed to precede the EPSP by 9+/−2 ms. To induce LTP, the AP was timed to follow the EPSP by 9+/−2 ms.
STDP is inducible in mouse primary visual cortex before and after the onset of the critical period of OD plasticity.
An example of LTD induction from a P17 slice is shown in . The average initial slope of the EPSP during the baseline period was 0.44 mV/ms. Following the AP-EPSP pairing protocol, the average initial slope of the EPSP decreased to 0.27 mV/ms. We calculated the EPSP slope ratio (EPSP slope post-pairing/EPSP slope pre-pairing) to compare plasticity across slices and ages. In both young and mature slices there was a significant reduction of the mean EPSP slope ratio following the LTD protocol. The mean EPSP slope ratio was 0.72+/−0.09 in young slices (p<0.05, paired t-test, n
12); and 0.72+/−0.13 in mature slices (p<0.05, paired t-test, n
12). There was no significant difference between the two ages (), determined using either a t-test (p
0.97) or a Kolmogorov-Smirnov (KS) test (p
0.99), which is sensitive to differences in data distribution as well as the mean.
An example of LTP is shown in . The baseline EPSP slope was 0.28 mV/ms. Following the AP-EPSP pairing protocol, the EPSP slope increased to 0.42 mV/ms. Similar to the LTD protocol, the LTP protocol induced significant plasticity at both ages. The mean EPSP slope ratio was 1.32+/−0.10 in young slices (p<0.05, paired t-test, n
11), and was 1.36+/−0.09 in mature slices (p<0.05, paired t-test, n
11). There was no significant difference between the two ages (t-test, p
0.76; KS, p
We also compared the ability to induce STDP at local recurrent connections within layer 2/3 in the two age groups (). To stimulate local recurrent connections, the field electrode was placed laterally within 50 microns of the recorded cell. Similar to layer4→layer2/3 connections, we found that the activated synapses were modifiable by STDP in both young and mature slices. In response to the LTD protocol the mean EPSP slope ratio was 0.79+/−0.09 in young slices (p<0.05, Wilcoxon signed rank, n
10), and was 0.78+/−0.09 in mature slices (p<0.05, Wilcoxon signed rank, n
9). There was no significant difference between the two ages (KS, p
0.25). In response to the LTP protocol the mean EPSP slope ratio was 1.47+/−0.16 in young slices (p<0.05, Wilcoxon signed rank, n
8), and 1.29+/−0.10 in mature slices (p<0.05, Wilcoxon signed rank, n
9). There was no significant difference between the two ages (KS, p
Therefore, by bypassing the requirement for input summation, we demonstrated that STDP was similarly induced at layer 4→ layer 2/3 connections as well as local recurrent connections in mouse primary visual cortex both prior to the onset and at the peak of the critical period of OD plasticity. Our results, along with others 
, raise the possibility that the ability to induce plasticity at glutamatergic synapses may not be a primary factor in determining the onset of OD plasticity.
Maturation of GABAergic synaptic and intrinsic properties in primary visual cortex
To examine the changes of GABAergic inhibition onto V1 pyramidal neurons during the critical period, we assayed the maximal inhibitory input onto layer 2/3 pyramids at two developmental ages, just prior to the onset (young) and during (mature) the critical period. Postsynaptic responses in layer 2/3 pyramidal neurons were recorded in response to stimulation of layer 4, which evoked a mixed excitatory-inhibitory response (). Similar to previous reports 
, we found that inhibitory drive increased with age relative to excitatory drive (), and that the maximal inhibitory charge significantly increased with age, while the maximal excitatory charge was stable (). Parvalbumin-containing (Pv+) basket cells make up ~50% of GABAergic interneurons in rodent V1, and it has been shown that there is a ~2-fold increase in the number of Pv+ basket presynaptic terminals surrounding pyramidal somata during the critical period 
. To determine if there was a corresponding increase in synaptic function, we recorded from synaptically connected Pv+ interneuron to pyramidal neuron pairs in layer 2/3. Pv+ interneurons were recorded using either BAC transgenic mice in which the Pv promoter drives GFP 
or Pv-cre mice 
injected with a recombinant adeno-associated virus that expresses GFP specifically in Pv+ basket cells 
. We found that peak inhibitory synaptic conductance increased by 1.8-fold during the critical period compared to prior to the onset of critical period, while there was a 25% decrease in the synaptic decay time-constant (, ). In contrast to inhibitory connections, paired recording of pyramidal neurons revealed that the peak excitatory synaptic conductance was similar between young (n
10) and mature (n
10) layer 2/3 connections (0.17+/−0.09 and 0.23+/−0.14 nS, respectively).
Maturation of GABAergic synaptic and intrinsic properties in primary visual cortex.
Synaptic properties of layer 2/3 neurons of V1.
In addition to synaptic properties, we characterized the intrinsic properties of Pv+ interneurons () and found that the current input/spike out curve shifted during the critical period: for the same stimulus input, the number of output spikes was greater during the critical period compared to that prior to the onset of the critical period. Thus, the gain of Pv+ interneurons increased during the critical period. In addition, there was a corresponding decrease in spike half-width ().
Intrinsic properties of layer 2/3 neurons of V1.
These electrophysiological results are summarized in and . Relative to excitatory connections, synaptic inhibition significantly matured with age. These results do not exclude the possibility that there are subtle developmental changes in synaptic excitation. In contrast to synaptic properties, the intrinsic properties of pyramidal cells, including input resistant, changed with age, as previously reported 
A simple point conductance model driven by two convergent input pathways modifiable by STDP
Modeling studies have demonstrated that in response to a change in the temporal pattern of presynaptic spike times, STDP implemented in its most basic form, re-organizes the population of synapses converging onto a postsynaptic neuron such that the most coherent inputs are strengthened, while the remaining synapses are weakened 
. The outcome of STDP driven re-organization is that the net excitatory drive across a population of synapses converging onto a single postsynaptic neuron is stabilized and a subset of presynaptic inputs controls postsynaptic spike timing. Here we extended the Song-Miller-Abbott (2000) model to include two distinct convergent input pathways, and tested the effects of altering presynaptic spike times within a pathway on the ability of the pathway to control postsynaptic spike timing, across a range of inhibitory levels. It has been observed that strong synapses can in some situations undergo less potentiation than weak synapses 
, therefore we also ran the simulation in a weight-dependent mode in which the amount of potentiation was inversely related to synaptic size.
We used an integrate-and-fire model neuron driven by 2 presynaptic pathways that each contained 40 synaptic inputs. As in Song et. al. (2000), a function F
) determined the amount of excitatory synaptic modification arising from a single pair of pre- and postsynaptic spikes separated by a time Δt
). Excitatory synaptic conductance was not allowed to exceed a maximum value gmaxex
. If the modification function pushed the synaptic weight past the gmaxex
value, the weight was reset to the appropriate limiting value. The maximum amount of modification for a single pre- and postsynaptic spike pair corresponded to a 0.5% change of gmaxex
. This function provides a reasonable approximation of the dependence of synaptic modification on spike timing observed experimentally, and makes no assumptions regarding the mechanism(s) of STDP. The model neuron also received inhibitory conductance: each excitatory conductance was followed by an inhibitory conductance of fixed amplitude with a delay randomly varying between 4 to 10 ms. For the simulations in , the ratio of the amplitude of inhibitory conductance over gmaxex
) was 0.264, and the average initial synaptic strength (ISS), was the same for both pathways, set to 25% of gmaxex
. This amount of inhibition was sufficient to maintain the postsynaptic neuron in a balanced mode, defined by an excitatory-inhibitory ratio of 1.1–1.2 at the threshold for action potential generation 
A simple IAF model neuron driven by two convergent input pathways subject to the STDP rule.
A minimal number of parameters were used to generate presynaptic spike trains (see Methods
). We then altered two of these parameters to modify the temporal relationships among inputs within and between the two pathways. First, we altered the temporal correlation
between pathway 1 (P1) and pathway 2 (P2), defined as whether or not the two pathways share coincident presynaptic spike times. Second, we altered the temporal coherence
(1/σ) among inputs within a single pathway, which refers to the degree of temporal clustering of spike times; the value σ represents the temporal jitter (ms) of presynaptic spike times ().
In our baseline condition, Input Regime I (), presynaptic spike times were correlated between P1 and P2, and presynaptic spike times within the two pathways had the same degree of high temporal coherence (σP1
3). This input regime represents features of a normal binocular neuron, which receives converging and correlated inputs from the two eye pathways, and activity within each pathway displays high temporal coherence driven by the same visual stimulus. Ten independent trials of the simulation were run. Cross-correlation analysis of individual presynaptic spike trains versus the postsynaptic spike train demonstrated that each pathway was capable of driving postsynaptic events during the initial 50 seconds of simulated time, and also in the final phase of the simulation (). Cross-correlation results are schematized in . As expected for this input regime, the mean synaptic weight for each pathway was unchanged during the course of the simulation (), both pathways maintained the ability to control postsynaptic spike timing throughout the simulation, as indicated by the dashed lines in .
In Input Regime II (), presynaptic spike times between the two pathways were de-correlated, while the high degree of temporal coherence within each pathway (σP1
3) was preserved. This input regime represents features of a binocular neuron during strabismus, in which inputs from the two eyes are de-correlated and spike times within each eye-specific pathway display high temporal coherence. Ten independent trials of the simulation were run. The mean synaptic weight of one pathway strengthened, at the expense of the opposing pathway, such that the total excitatory driving force was maintained and the firing rate of the postsynaptic neuron was stable. The outcome of which pathway, P1 or P2, dominated occurred at chance level.
For Input Regime III (), in addition to temporal de-correlation between the pathways, temporal coherence was reduced in Pathway2 (σP1
6, a σ ratio of 1
2). This input regime represents features of a binocular neuron during monocular deprivation, in which the two eye-specific pathways are uncorrelated and the temporal structure of activity differs between the two pathways. The open eye views high-contrast patterns, therefore the open-eye pathway likely has a relatively higher degree of temporal coherence compared to the closed-eye pathway 
. The temporal structure of the presynaptic spike trains used in the simulation is shown in . Ten independent trials were run. Pathway1, with higher temporal coherence, always attained the higher mean synaptic weight and emerged to drive spike output. In addition, the spike times of the postsynaptic neuron became controlled by P1 in all trials. Our model thus demonstrates that the stabilizing and competitive properties of STDP first described by Song et. al. (2000), also apply to the condition of two independent convergent pathways, and that the pathway with relatively higher temporal coherence will dominate in driving postsynaptic spike times when the initial synaptic strength is equal between pathways.
Initial synaptic strength confers a competitive advantage
In addition to the temporal structure of presynaptic inputs, initial synaptic strength plays a major role in driving postsynaptic spiking and is also likely to contribute to the outcome of STDP-driven re-organization of inputs. Inputs with higher synaptic strength have an advantage because fewer active synapses are required to evoke a postsynaptic action potential. Indeed, at the retinotectal projection in tadpoles, it has been demonstrated that the extent to which a given pathway potentiates in response to asynchronous stimulation of convergent inputs is dependent on initial synaptic strength 
Using Input Regime III (σ ratio of 1
2), we challenged the ability of P1, the temporally more coherent pathway, to control postsynaptic spike timing by increasing the initial synaptic strength (ISS) of P2. The ISS ratio (P2/P1) was varied from 1.0 to 2.0. Using the same level of inhibition as in (gI/gmaxex
0.264), we found that the fraction of simulation trials in which P1 controlled postsynaptic spike timing decreased with increasing strength of P2 (, dashed line). When P2 was initially 50% stronger (ISSP2/P1
1.5), it dominated in driving postsynaptic spike activity in only half of the trials. Thus when the ISSP2/P1
ratio was ≥1.5, P1, the pathway with higher temporal coherence, failed to direct postsynaptic spike timing above chance level. This result, however, formally contradicts with the results of OD plasticity in V1.
The level of inhibition determines whether temporal coherence or synaptic drive controls postsynaptic spike times in the IAF model.
The majority of binocular neurons in V1 are not equally driven by the two eye-specific pathways. In rodents, approximately 70% of binocular neurons 
are characterized as class 2/3 cells, preferentially driven by the contralateral eye inputs. Despite this contralateral bias, contralateral eye closure during the critical period shifts the response properties of class 2/3 cells such that they become dominated by the initially weak, but temporally more coherent ipsilateral eye inputs. Our result in , in which the pathway with higher temporal coherence fails to direct postsynaptic spikes, is thus inconsistent with the OD shift of class2/3 neurons induced by monocular deprivation. In the following section we tested the hypothesis that synaptic inhibition can constrain STDP and promote the selective strengthening of temporally coherent inputs at convergent pathways, even when challenged with inputs of higher initial synaptic strength.
Inhibition biases STDP to favor temporal coherence over initial synaptic strength
The stabilizing influence of STDP on net excitatory drive is highly dependent on the non-linearity of the spike generation process 
. Given that synaptic inhibition has been shown to potently influence input summation by restricting the temporal window over which inputs are able to effectively cooperate 
, we tested if increasing the amplitude of synaptic inhibition in Input Regime III could bias the weaker but temporally more coherent pathway (P1) to control postsynaptic spike timing. The simulation was run at 6–12 different levels of inhibition and the strength of P2 was increased 2 to 3-fold relative to P1. The simulation was run for 30–50 independent trials for each parameter pair, and the same presynaptic spike train was used for each level of inhibition for a given ISSP2/P1
ratio (). We found that the range of ISSP2/P1
ratios in which P1 out-competed P2 was extended when inhibition was high, and that for a given ISSP2/P1
ratio, higher levels of inhibition increasingly biased the outcome of STDP-driven competition to favor P1 over P2 (). For example, in the case that P2 was set to be 50% stronger than P1 (ISSP2/P1
1.5) and the amplitude of inhibition was set to ≥0.792 gI/gmaxex
, P1 dominated in 100% of the trials, while at lower levels of inhibition (gI/gmaxex
0 to 0.264), P2 out-competed P1 in roughly 50% of the trials. We examined a range of relative temporal coherence values and found that increasing inhibition had a similar effect as above in cases that the ISS ratio was sufficiently high to give an advantage to P2 ().
As previously reported 
, we found that the input resistance of layer 2/3 pyramidal cells decreased with age (). A change in input resistance (Rin
) could potentially influence summation and therefore impact the effect of inhibition in our simulation. However, there was a parallel change in the membrane time-constant (τmem
). Given the relationship, τmem
* whole-cell capacitance, whole-cell capacitance remained stable. In acute slice experiments, values of Rin
are typically about 30% higher than in vivo studies 
. We examined a physiologically realistic range of whole-cell capacitance values in our simulation (from 0.125 nF to 0.375 nF) and found that the effect of increasing inhibition was independent of whole-cell capacitance (Figure S3
In the above simulations we implemented contralateral bias as an increase in initial synaptic strength. However, contralateral bias in vivo could be the result of an increased number of inputs rather than increased synaptic strength (or a combination of the two). Therefore we examined the effect of increasing inhibition across an increasing number of P2 inputs and found that as with synaptic strength, increased inhibition helped to ensure that the temporally coherent pathway out-competed the pathway with an initially stronger synaptic drive ().
A likely mechanism by which stronger inhibition biased STDP to favor P1 is that inhibition narrowed the window of input cooperation, thereby preferentially restricting the less coherent inputs in P2 from contributing to postsynaptic spike generation. If this is the case, then the relative ability of P1 synapses to drive a postsynaptic spike event (synaptic efficacy) should increase with inhibition, independent of the STDP learning rule. To examine this possibility, we compared the efficacy of P1 and P2 synapses in a simulation implemented as in , except without applying the STDP learning rule. As expected, increasing the level of inhibition (from gI/gmaxex
0.264 to 0.729) decreased the relative efficacy of P2 synapses. At a high level of inhibition, P1 synapses had a slight advantage in driving postsynaptic events 2–8 ms preceding the postsynaptic spike (, compare A&B), a temporal window for which STDP-mediated potentiation is the strongest. This slight advantage of P1 was robustly magnified when the STDP learning rule was applied in the simulation. Within the first 100 spikes of the simulation, the relative efficacy of P1 synapses increased compared to the no-STDP condition (, compare B&D). P1 synapses completely dominated by the end of the simulation (, compare D&F). Analysis of synaptic efficacy was also performed on the results from the initial simulations shown in (see Figure S4
). Similarly, we found that the pathway having a slight advantage after the first 100 spikes of the STDP simulation would ultimately dominate.
The relative contribution of input pathways with different temporal coherence and initial synaptic strength to postsynaptic spike events is dependent on inhibition.
We previously showed that at a low level of inhibition (gI/gmaxex
0.264), P2 out-competed P1 in 50% of the STDP simulation trials (, ISSP2/P1
1.5). It was surprising therefore to find that in the absence of STDP, P2 dominated in driving postsynaptic spikes for the full 20 ms time window (, upper left). Given this advantage, P2 would be expected to dominate in 100% of the STDP simulation trials rather than only 50%. The reason that P2 did not dominate in 100% of the trials was that the number of presynaptic spikes occurring after
each postsynaptic spike was greater for P2 than P1 (), thus leading to more LTD in P2 than in P1 when the STDP learning rule was applied. Importantly, the difference in the number of presynaptic spikes following
postsynaptic spike events between the two pathways was similar for both low and high levels of inhibition (). Therefore, the preferential restriction of P2 inputs ( A,B) was due to a specific decrease in their synaptic efficacy rather than a increase of their LTD at higher levels of inhibition.
It has been shown that under some conditions, the amount of synaptic potentiation is dependent on synaptic strength 
, and that this can impact the outcome of STDP 
. Therefore, we ran the same simulation shown in with an additional weight-dependent rule in which the amount of LTP was inversely related to synaptic size (). The ability of stronger inhibition to confer a competitive advantage to P1 was maintained.
Influence of inhibitory neuron gain on STDP
Our whole-cell recordings from pyramidal neurons in cortical slices revealed that while the amplitude of unitary IPSCs increased at the onset of the critical period, there was a concomitant decrease in the synaptic decay time-constant, thus the developmental change in unitary synaptic charge does not scale equally with amplitude. When the 25% decrease in synaptic decay time-constant that we experimentally observed was implemented in the simulation, we found that the effectiveness of increased inhibition in ensuring that P1 out-competed P2 was diminished (). We hypothesized that the developmental increase in gain of spike output in GABAergic interneurons that we experimentally observed could compensate for the decrease in the synaptic decay time-constant. We tested this hypothesis by first demonstrating that an increase in gain of spike output increased the probability of P1 out-competing P2 across a range of initial synaptic strength values (). In this case the synaptic inhibitory decay time-constant was set to the original value of 5.75 ms, and increased gain was implemented as the probability of a given inhibitory input spiking twice. For example, a gain of 1.0 corresponded to a given presynaptic input spiking only once, a gain of 1.5 corresponded to a 50% probability that a given presynaptic would generate a second spike, and a gain of 2 signified that a presynaptic input would spike twice. Pv+ interneuron-to-pyramidal synapses display short-term depression in which the amplitude of the second IPSC is 10–30% reduced compared to the first IPSC (Figure S5
). To account for short-term depression in our simulation, the amplitude of the conductance on the second spike was reduced by 20%. We found that increasing spike output gain increased the probability of P1 out-competing P2. Next, the gain was increased at the same time the synaptic decay time-constant was reduced to 75% of the original value. We found that an increase in gain could indeed compensate for the decreased synaptic decay time-constant.
An increase in gain can compensate for the decrease in the decay time-constant of synaptic inhibition.
In summary, our modeling results show that both temporal coherence and initial synaptic strength of synaptic inputs can confer a competitive advantage at convergent pathways modifiable by STDP, but synaptic inhibition can constrain STDP to favor temporally coherent inputs at the expense of inputs with stronger initial synaptic strength. These results have implications for ocular dominance plasticity, particularly for class 2/3 neurons. Given that OD plasticity involves correlation-based Hebbian mechanisms 
, a prerequisite for a binocular neuron to shift its ocular dominance towards the open eye is that its spike times must correlate with the open eye pathway. For a class 2/3 neuron during contralateral eye closure, this means that the weaker but more coherent inputs of the open eye pathway must increase the correlation of their spike times with the spike times of the postsynaptic neuron. Studies using single unit recordings generally classify cells as 2/3 if the contralateral drive is 1.5 fold or greater than the ipsilateral drive 
, thus the ISSP2/P1
ratios we used are well within the range of experimental observations, and the effects that we observe on STDP may also apply to cells that are borderline class 4 cells. Our modeling results demonstrate that the maturation of GABAergic inhibition can constrain STDP so that the spike output of class 2/3 neurons, which are dominated by the contralateral eye at the time of its closure, become increasingly correlated with the ipsilateral, open-eye input. Our simulation further shows that inhibition mediates such an effect by reducing the synaptic efficacy of the less coherent, even though stronger, inputs. Therefore, a prediction from our model is that maturation of GABAergic inhibition must be sufficiently strong to more potently decrease synaptic efficacy at the peak versus prior to the onset of the critical period of OD plasticity.
Maturation of visual cortical synaptic inhibition reduces synaptic efficacy
We used a visual cortical slice preparation to examine whether the maturation of GABAergic inhibition reduces synaptic efficacy, and whether this effect correlates with the onset of OD plasticity. Postsynaptic responses in layer 2/3 pyramidal neurons were recorded following stimulation of layer 4, which evoked a mixed excitatory-inhibitory response (). The time course of evoked IPSCs outlasted the EPSCs by roughly 7-fold. This was due to the slower kinetics of the GABAA receptors compared to that of the AMPA receptors and to the presence of polysynaptic IPSCs.
Previous studies have shown that stimulus-evoked inhibition can reduce the synaptic efficacy of asynchronous EPSPs for up to 30–50 ms 
, a time course that matches the duration of the evoked GABAA
current measured here. We used a two-pathway stimulation paradigm to examine the effect of synaptic inhibition on synaptic efficacy at layer 4 to layer 2/3 connections (). The stimulation intensity of both pathways was normalized to spike threshold in layer 2/3 pyramidal neurons to facilitate comparison across slices and animals. A test pathway (Ptest
) was stimulated at threshold intensity such that an action potential in a layer 2/3 pyramidal neuron was generated with a probability of approximately 0.5 (see methods
). The ability of Ptest
to trigger a postsynatpic action potential was then challenged by stimulating a leading pathway (Plead
) 40 ms earlier. The stimulation intensity of the Plead
was set such that an action potential in the postsynaptic pyramidal neuron was triggered with >0.9 probability. Ptest
were verified to be independent pathways to avoid short-term plasticity such as synaptic depression. The use of the 40 ms interval between the two stimuli ensured that postsynaptic spikes evoked by Plead
stimulation did not over lap with the postsynaptic responses evoked by Ptest
stimulation. Trials in which only Ptest
(Test Only) was stimulated were interleaved with those in which Plead
were sequentially stimulated (Lead-Test).
The potency of one input pathway to suppress the probability of a convergent pathway from triggering a postsynaptic spike in layer2/3 pyramidal neurons in V1 is developmentally regulated.
The first experiment was done in cell-attached mode, with an intact intracellular chloride gradient. We found that at the peak of the critical period (P26–30), Plead
stimulation reduced the spike probability (ρ) of layer 2/3 pyramidal neurons in response to Ptest
stimulation by 37+/−0.09% (lead-test: ρ
0.32+/−0.08, test alone: ρ
9, Wilcoxon signed rank, p<0.02, ). Prior to the onset of the critical period (P16–18), however, there was little if any effect of Plead
stimulation on spike probability of layer 2/3 pyramidal neurons triggered by Ptest
stimulation (lead-test: ρ
0.66+/−0.09, test alone: ρ
0.62+/−0.06, mean difference: Δ+0.04+/−0.08, n
9, Wilcoxon signed rank, p
We then repeated the above stimulation protocol using whole-cell recordings of layer 2/3 pyramidal neurons. Similar to cell-attached recording, spike probability in layer 2/3 pyramidal neurons in response to Ptest
stimulation was significantly reduced by Plead
stimulation at the peak of the critical period (lead-test: ρ
0.25+/−0.07, test alone: ρ
0.58+/−0.07 mean difference: Δ−0.034+/−0.08, n
12, Wilcoxon signed rank, p<0.003, ), while no such effect of Plead
stimulation was found prior to the onset of the critical period (lead-test: ρ
0.56+/−0.08, test alone: ρ
0.55+/−0.06, mean difference: Δ+0.01+/−0.11, n
12, Wilcoxon signed rank, p
To examine whether the reduction in synaptic efficacy of Ptest
was mediated by synaptic inhibition, we repeated the whole-cell experiment in a bathing solution containing 3 mM divalent cations and 3 µM bicuculline methiodine (BMI). This resulted in an 80% block of synaptically evoked GABAA
current (data not shown) without inducing epileptic activity in cortical slice. In the presence of raised cation concentration but in the absence of BMI, spike probability in layer 2/3 pyramidal neurons in response to Ptest
stimulation was significantly reduced by Plead
stimulation at the peak of the critical period (lead-test: ρ
0.23±0.10, test only: ρ
0.56±0.07, mean difference: Δ−33+/−0.10, n
12, ), similar to the above results of . The effect of Plead
stimulation was blocked in the presence of BMI (lead-test: ρ
0.75±0.05, test only: ρ
11, mean difference: Δ+0.18±0.08, ). In contrast, prior to the onset of the critical period, BMI had little impact on the ability of Plead
to reduce spike probability in response to Ptest
stimulation (no BMI, lead-test: ρ
0.45±0.07, test alone: ρ
0.53±0.02, mean difference: Δ−0.08±0.07, n
12, ; BMI, lead-test: ρ
0.60±0.50, test alone: ρ
0.49±0.02, mean difference: Δ+0.11±0.06, n
10, ). A 2-way ANOVA was performed to determine whether the age-dependent effect of GABAA
blockade was significant. The change in spike probability (lead-test – test only) for layer 2/3 pyramidal cells was compared across treatment groups (). The interaction between age and BMI treatment was significant (p<0.05), and the effect of BMI on spike probability was significant (p<0.001). Subsequent pairwise comparisons using the Holm-Sidak method revealed a significant difference between BMI treated and control cells in the mature age group, but not in the young age group. We conclude that there is a developmental increase in the ability of synaptic inhibition to decrease synaptic efficacy during the critical period in mouse visual cortex.
Reduction of synaptic efficacy of Ptest by Plead stimulation is mediated by synaptic inhibition.
We noted that reduced GABAA
conductance revealed the presence of a slight summation among inputs. This effect was also seen by Mittman et. al. (2005); given the rapid kinetics of AMPA receptors, this effect was unlikely due to synaptic conductance evoked by stimulation of Plead
. The effect could not be explained by a change in input resistance (young: control, 262.38±3.89, BMI, 261.62±4.96; mature: control, 132.50±3.38, BMI, 135.57±2.34 MΩ), suggesting that the slight summation observed in the condition of 80% GABAA
block may be due to a voltage-dependent persistent sodium conductance induced by stimulation of Plead 
Further evidence for the role of chloride conductance in mediating the reduction in spike probability at the peak of the critical period was obtained by using DIDS-fluoride in the recording pipette to block anionic conductances, including GABAA 
. Because the drug was in the pipette, the blockade was specific to the recorded cell. Spike probability in response to Ptest
stimulation was significantly reduced by Plead
stimulation (lead-test: ρ
0.61±0.07, test alone: ρ
0.32±0.06, mean difference: Δ+0.28+/−0.01, n
8, Wilcoxon signed rank, p<0.03, ). In this case, there was a 2.8 fold change in input resistance that likely contributed to summation (break in: 154±37, stable: 348±50 MΩ). The change in spike probability due to stimulation of Plead
for all treatments is summarized in .
In summary, synaptic inhibition evoked by the Plead was effective in reducing synaptic efficacy of Ptest at the peak of but not prior to the onset of the critical period of OD plasticity.