In the first series of experiments we recorded from neurons of the local microcircuit in layer 5A of the rat somatosensory neocortex formed by presynaptic pyramidal cells and postsynaptic FS interneurons. Previous reports indicate that these synapses undergo a transition from PPD to PPF during the first 4–5 postnatal weeks (Angulo et al., 1999). In order to investigate the underlying mechanisms and to compare pyramidal-to-FS synapses in layer 2/3 and layer 5, we performed paired recordings at these synapses.

Cells were identified based on location and shape of the cell body (observed using infrared microscopy) and their firing pattern. Postsynaptic FS interneurons typically had round somata and showed high-frequency non-accommodating and non-attenuating action potential firing upon depolarizing current injection. The presynaptic neurons were slender tufted (layer 5A) pyramidal cells. These cells were located in upper layer 5 (less than 100 μm from the layer 4) and had a smaller cell body compared to the thick tufted cells (layer 5 pyramidal cells). Upon depolarizing current injection, slender tufted pyramidal cells exhibited regular low-frequency spiking with pronounced accommodation.

First we evaluated the stability of recordings and synaptic efficacy in the pairs of slender tufted pyramidal to layer 5 FS cell connections at P14 and P28, respectively. We observed a gradual increase of EPSP amplitudes in FS cells during prolonged whole-cell recordings. After the first 20–30 min EPSP amplitudes reached a steady state which was on average almost two times higher than that at the beginning of an experiment. Figure 1A shows a representative trace and summarizes the average time course of the EPSP amplitude for P14 (n = 9) and P28 (n = 9) animals. Changes in synaptic efficacy might indicate washout of either pre- or postsynaptic factors limiting synaptic transmission. To untangle the pre- or postsynaptic mechanisms underlying the time-dependent change in transmission, we made sequential triple recordings. In the first set of experiments, we continuously recorded from one postsynaptic FS cell while sequentially patching two different presynaptic pyramidal cells. By the time the second pair was obtained, the postsynaptic interneurons had been dialyzed for at least 45 min, ensuring stable internal milieu. However, within the first 15 min of recording with the second pyramidal cell we still observed a gradual increase of EPSPs amplitudes in the postsynaptic FS interneurons (n = 5; data not shown). In inverse experiments the presynaptic cell was kept constant while the postsynaptic interneurons were sequentially changed. Here, run-up of responses was observed only in the first pair (n = 4; data not shown).

Next we analyzed changes in release probability and paired pulse ratios (PPR; EPSP2/EPSP1). We found that in P14 animals, the averaged PPR values decreased from 0.64 ± 0.2 at the beginning of the experiments to 0.46 ± 0.1 at the end of the experiments (n = 20; p < 0.01; Wilcoxon signed rank test). A similar reduction of PPRs was observed in P28 animals where values decreased from 1.1 ± 0.2 to 0.94 ± 0.1 (Figure 1D; n = 21; p < 0.01; Wilcoxon signed rank test). In P28 animals, after prolonged dialysis, facilitation at slender pyramidal to FS cells synapses was washed out almost in all cases. Note that average PPRs in P28 animals were always higher than at P14, both at the beginning or at the end of the experiments (in both cases p < 0.01; Mann-Whitney rank sum test).

Analysis of failures revealed highly diverse data between different pairs. As an example, in P14 animals failure rates ranged from 0 to 22%. However, in all pairs with low initial release probability the failure rate decreased significantly after prolonged dialysis of the cell. At P14, the median initial failure rate was 5% and decreased to 2% during the recording (n = 20; p < 0.01; Wilcoxon signed rank test; Figure 1B). At P28, failure rate medians were 6% (initially) and 2% after 40 min of whole cell dialysis (n = 21; p < 0.01 Wilcoxon signed rank test; Figure 1B1). Taken together, these data suggest that prolonged dialysis of presynaptic slender tufted cells leads to an increase of release probability. The increase of release probability together with the above-mentioned alterations in synaptic efficacy and PPRs are strongly indicative of a presynaptic origin, most likely the washout of factors controlling calcium levels at the release site.

In the interest of stability, all further experiments were carried out after presynaptic pyramidal cells had been dialyzed for 30 min and EPSP amplitudes in FS interneurons had reached a steady state level. Under these conditions median unitary EPSP amplitudes were 1.89 mV (n = 42) and 1.5 mV (n = 43; p = 0.4; Mann-Whitney rank sum test) in P14 and P28 rats, respectively (Figure 1C).

In these experiments, we recorded from neurons of the local microcircuit in layer 2/3 of rat somatosensory neocortex (Reyes et al., 1998). The presynaptic neurons were pyramidal cells, identified by the shape of the soma and the pattern of frequency accommodation of APs upon depolarizing somatic current injection. The target neurons were non-pyramidal, multipolar interneurons, as viewed with infrared video-microscopy and characterized by non-accommodating FS firing patterns (Reyes et al., 1998; Rozov et al., 2001).

Since cortical lamination is formed in an inside-out fashion, development of synaptic properties at connections formed by layer 2/3 pyramidal cells might be delayed relative to those in layer 5. To test this notion we compared release probability, synaptic efficacy and PPRs at layer 2/3 pyramidal to FS cell synapses in P14, P28 and P42 animals. In contrast to the results described above for layer 5 pairs, in most experiments, unitary EPSP amplitudes and PPRs remained constant during the recording time (Figure 2). Although amplitude distributions for all ages were broad, median amplitudes were very similar, being 2.43 mV (n = 69), 2.43 mV (n = 71) and 1.98 mV (n = 27) for P14, P28 and P42 respectively (p = 0.42; Kruskal-Wallis one way analysis of variance on ranks; Figures 2A–C).

Next we analyzed reliability of release by measuring the percentage of failures of the presynaptic AP to evoke an EPSP. Figures 2A–C shows the distribution of failures in pyramidal to FS neurons connections at three developmental stages. Again no significant changes in failure rates were found. Median values were; 0% at p14 (n = 69), 1% at P28 (n = 71) and 0% at P42 (n = 27; p = 0.63; Kruskal-Wallis one way analysis of variance on ranks). In all three ages, APs rarely failed to elicit an EPSP in multipolar neurons, indicating a very reliable synaptic connection.

Previous evidence has indicated marked pronounced short-term depression at synapses from layer 2/3 pyramidal to FS cells (Reyes et al., 1998; Rozov et al., 2001). In line with this observation, the average PPR was 0.58 ± 0.12 (n = 69) for P14, 0.64 ± 0.11 (n = 71) for P28 and 0.71 ± 12 (n = 27) for P42 rats (p = 0.31; Kruskal-Wallis one way analysis of variance on ranks). To rule out the possibility that gradual deterioration of the cells was responsible for the synaptic depression, we analyzed PPRs at the beginning (first 10 min.) and the end (last 10 min.) of each experiment. In most of the experiments PPRs remained constant (Figure 2D). Thus in layer 2/3, the basic properties of synaptic release and type of short term plasticity at pyramidal to FS cells connections do not change between the second and sixth postnatal weeks. Therefore all subsequent experiments were carried out on P14 and P28 animals.

It has been shown that differential dependance of release on extracellular calcium concentration [Ca²⁺]_o could indicate structural differences in release zone organization in facilitating and depressing terminals. It has also been demonstrated that different degrees of saturation of the presynaptic calcium sensor occur during single APs (Rozov et al., 2001). Therefore we tested whether the observed developmental change in PPR goes along with changes in the relation between release probability and [Ca²⁺]_o. Extracellular calcium concentration was varied between 1, 2, 3 and 4 mM, respectively. In order to compare across different cell pairs, averaged EPSP amplitudes and PPRs at each [Ca²⁺]_o were normalized to the values obtained with the control calcium concentration (2 mM).

As illustrated in Figure 3A, dependances of release probability from layer 2/3 pyramidal-to-FS cells terminals on [Ca²⁺]_o in P14 and P28 animals were very similar. EPSP amplitude was already very close to saturation at the normal [Ca²⁺]_o of 2 mM. The averaged data points could be fitted satisfactorily by a Hill equation with an exponent of four and half-effective concentrations (K_1/2) of [Ca²⁺]_o amounting to 1.09 and 1.1 mM, for P14 and P28 respectively.

In both P14 and P28 rats, reduction of [Ca²⁺]_o led to an increase of PPR, and synaptic depression was most pronounced at higher [Ca²⁺]_o (data not shown). However, no statistical difference in these parameters was observed between younger and older animals (at all tested [Ca²⁺]_o concentrations p > 0.05; Mann-Whitney rank sum test).

Next we examine whether a similar dependance between [Ca²⁺]_o and release probability is present at layer 5 pyramidal to FS neurons. Using the same experimental protocol as in layer 2/3 we found no significant differences in calcium-dependance of release between young (P14) and mature (P28) animals at slender pyramidal-to-FS cell terminals (at all tested [Ca²⁺]_o concentrations p > 0.05; Mann-Whitney rank sum test; Figure 3B). Averaged data points were fitted using a Hill equation, yielding an exponent of four and half-effective concentrations (K_1/2) of 2.44 and 2.29 mM [Ca²⁺]_o for P14 and P28, respectively.

Similar to the layer 2/3 connection, reduction of [Ca²⁺]_o led to an increase in PPR and depression was most pronounced at high [Ca²⁺]_o in both P14 and P28 rats. However, we did not find any statistical difference between slices from younger and older animals. Since sensitivity of release to [Ca²⁺]_o at layer 5 pyramidal to FS cell remains the same during the second two postnatal weeks, facilitation at P28 cannot be explained by the change of the calcium sensor sensitivity.

Sensitivity of transmission to exogenously loaded buffers can be used to probe the presynaptic determinants of release probability and short-term plasticity. For example, synapses with a long diffusional distance between calcium channels and calcium sensors are sensitive to both fast and slow exogenous buffers. On the other hand, if calcium channels are located closely to calcium sensors, the effect of calcium buffers is much less pronounced (Neher, 1998). Buffer sensitivity has been successfully used to investigate differences between facilitating and depressing synapses (Rozov et al., 2001). Here, we used the same approach to check whether the spatial arrangement between the calcium source and calcium sensor in layer 2/3 pyramidal cell terminals onto FS cells undergoes developmental modifications (for details, see “Materials and Methods” Section).

Release from pyramidal cell terminals to FS cells was suppressed by EGTA in a concentration dependent manner. This reduction was already significant at a concentration of 0.2 mM. Figure 4A compares the dependance of release on EGTA concentration for P28 layer 2/3 pyramidal to FS cell with data published for the same type of synapses in P14 animals (Rozov et al., 2001). Mean EPSP amplitudes were normalized to the control averaged EPSP amplitude before buffer loading and plotted as a function of buffer concentration. The half effective concentration of EGTA for both P14 and P28 animals was very similar, with values around 7 mM. At both ages, EGTA did not have any statistically significant effects on PPD (p > 0.05; Wilcoxon signed rank test, at all tested EGTA concentrations). PPRs obtained upon buffer loading were normalized to control PPR values, to minimize the effect of scatter in the initial synaptic depression level between different cell pairs (Figure 4A, right plot).

Loading of the fast exogenous buffer BAPTA had a much stronger effect on synaptic release at pyramidal to FS cells compared to EGTA. At P28, the half-effective concentration of BAPTA was about 0.5 mM, comparable to that measured previously in P14 animals (Rozov et al., 2001). Dose-response curves for the effect of BAPTA on synaptic release are shown in Figure 4B. At both ages BAPTA could efficiently suppress synaptic depression at concentrations higher than 0.5 mM (Figure 4B, right plot). Together, these data largely exclude major spatial rearrangements of calcium sources and calcium sensors at the vesicle release site.

For comparison, we tested the effect of calcium buffers on synaptic transmission between slender tufted pyramidal cells and FS neurons in layer 5. Figure 5A shows the effect of the slow calcium buffer EGTA on synaptic release at both ages, P14 and P28. EGTA had relatively small effects on synaptic release. When 10 mM EGTA was loaded into the presynaptic cell, unitary EPSP amplitudes were reduced to 62 ± 11% of the control value in P14. In P28 animals the effect of 10 mM EGTA appeared to be even less pronounced (78 ± 14% of control; Figure 5A). The apparent decrease of EGTA-sensitivity with increasing age was, however, not significant (p > 0.05; Mann-Whitney rank sum test).

Loading of the fast exogenous buffer BAPTA had a much stronger effect on synaptic release at layer 5A pyramidal to FS cell connections than EGTA. A comparison of dose response curves for P14 and P28 animals is shown in Figure 5B. In contrast to EGTA, the effects of BAPTA were significantly less pronounced at later developmental stages. The half effective concentration for BAPTA in P14 animals was around 1 mM which suppressed EPSPs to 0.49 ± 0.12% of control (Figure 5B). In older animals, 2 mM BAPTA was required to suppress release by half (52 ± 10% of control EPSP amplitude). Thus, during maturation of slender tufted pyramidal to FS cell terminals, sensitivity of synaptic release to exogenously loaded buffers is reduced. This finding indicates a shortening of the diffusional distance between calcium channels and calcium sensors. However, this assumption is in apparent contradiction with the reduction of release probability and increased PPRs in P28. One possible explanation might be that reduction in diffusional distance is accompanied by a change in the type of calcium channels.

Presynaptic calcium channel subtypes can change during development, leading to significant changes in the efficacy and dynamics of synaptic transmission. Specifically, the expression of N-type calcium channels keeps growing during the first postnatal weeks and reaches maximum levels at ~P28 in the rat hippocampus and cortex (Jones et al., 1997). In the rat striatum, N-type calcium channel expression is significantly down regulated during the third postnatal week (Martella et al., 2008). Therefore, we compared the contribution of different subtypes of calcium channels to synaptic release at layer 2/3 pyramidal to FS cell synapses in P14 and P28 rats using specific calcium channel blockers.

Bath application of 100 nM of the P/Q-type calcium channel blocker agatoxin IVA (Aga) reduced the amplitude of unitary EPSPs relative to control values to 38 ± 3% (n = 7; p < 0.01; Wilcoxon signed rank test) and to 41 ± 4% (n = 6; p < 0.01; Wilcoxon signed rank test), at P14 and P28 respectively (Figure 6). Of note, this toxin has much higher affinity to P-type channels with a K_D of about 1 nM than to Q-type channels with a K_D of 100 nM. Application of the P/Q- and N-type channel blocker ω-conotoxin (MVIIC) almost entirely blocked synaptic transmission. In the presence of 1 μM of MVIIC, unitary EPSP amplitudes were reduced to 2.1 ± 0.4 (P14; n = 6; p < 0.01; Wilcoxon signed rank test) and to 2.3 ± 1.5% (P28; n = 6; p < 0.01; Wilcoxon signed rank test) of the control value. This toxin blocks Q-type channels more efficiently (K_D of about 5 nM) than P-type channels (K_D of about 500 nM), suggesting a significant contribution of Q-type channels to release (Zhang et al., 1993). The EPSP that remained after adding saturating toxin concentrations could be blocked by 50 μM Cd²⁺ (at both ages p < 0.01; Wilcoxon signed rank test). The remaining release is most probably due to Ca²⁺ influx via R-type calcium channels, since it was completely blocked by 100 μM Ni²⁺ (at both ages p < 0.01; Wilcoxon signed rank test). Application of the specific N-type channel blocker ω-conotoxin (GVIA) at 1 μM suppressed release to 25 ± 4% at P14 (n = 6; p < 0.01; Wilcoxon signed rank test) and to 32 ± 7% at P28 (n = 5; p < 0.01; Wilcoxon signed rank test) of control values, again indicating that N-type channels strongly contribute to evoked release in pyramidal to FS cell connections at both ages (Figure 6A). Thus, at P14 and at P28 chemical transmission between pyramidal and FS neurons is mostly mediated by P/N/Q-type calcium channels. Taken together our data shows that in layer 2/3 pyramidal to FS cell synapses a number of important parameters stay constant between P14 and P28. These include basic release probability, [Ca²⁺]_o dependance, sensitivity to exogenously loaded buffers and the relative contribution of presynaptic P/N/Q-type calcium channels.

In P14 animals, EPSPs in FS cells evoked by stimulation of layer 5A pyramidal cells were blocked to 52 ± 12% (n = 5; p < 0.01; Wilcoxon signed rank test) of the control value by bath application of 100 nM Agatoxin, a P/Q-type calcium channel blocker. Surprisingly 1 μM of GVIA, a specific N-type channel blocker, had only mild effects on release (62 ± 10.2% of control value; n = 7; p < 0.05; Wilcoxon signed rank test) indicating a limited contribution of N-type channels to evoked release at this connection. Moreover, in the presence of the P/Q- and N-type channel blocker MVIIC (1 μM) 11 ± 2% (n = 6; p < 0.001; Wilcoxon signed rank test) of the initial EPSP amplitude remained. The remaining release is most probably due to P-type calcium channels, since a mixture of MVIIC (1 μM) and Agatoxin (100 nM) almost entirely blocked synaptic transmission to 2 ± 1% of control value (n = 5; p < 0.001; Wilcoxon signed rank test; Figure 6B). Thus, in P14 animals synaptic transmission between slender tufted pyramidal and FS neurons is mediated predominately by P/Q-calcium channels, with a small contribution of N type channels.

In P28 animals the effect of GVIA on synaptic release was small (73 ± 18% of control value; n = 7; p < 0.05; Wilcoxon signed rank test) and similar to that in young animals. In contrast to P14, application of Agatoxin did not have any significant effect on synaptic release at P28 (92 ± 4% of control EPSP remained; n = 8; p > 0.05; Wilcoxon signed rank test). This indicates a strongly reduced contribution of P-type calcium channels to synaptic release at layer 5A pyramidal to FS cell synapses. On the other hand, application of MVIIC almost entirely blocked EPSPs down to 1.2 ± 1% (n = 6; p < 0.001; Wilcoxon signed rank test) of the control value, suggesting an increased contribution of Q type calcium channels (Figure 6). In both P14 and P28 animals, the residual release remaining after application of MVIIC and Agatoxin or of MCVII alone was mediated by R type calcium channels, since it could be completely blocked by 50 μM Cd²⁺ (n = 5; p < 0.01; Wilcoxon signed rank test). Thus, based on the differential effects of Agatoxin and MVIIC on evoked EPSP amplitudes and taking into account the differences in K_D, we conclude that at P28 in layer 5A pyramidal to FS cells terminals, release is triggered almost solely by calcium entry through Q type calcium channels.

We investigated local layer 2/3 circuits composed of presynaptic pyramidal cells and postsynaptic parvalbumin positive FS interneurons (Reyes et al., 1998; Rozov et al., 2001). It has been reported that at P28 in connections between pyramidal cells in layer 2/3, as well as in connections between layer 2/3 pyramidal cells and layer 5 pyramidal cells, slight facilitation substitutes the pronounced depression observed in P14 rats (Reyes and Sakmann, 1999). Therefore, one might have expected similar changes in other circuits where layer 2/3 pyramidal cells act as the presynaptic partner. However, our analysis of layer 2/3 pyramidal to FS cell connections at two different ages (P14 and P28) did not reveal any statistical differences in a number of parameters including: initial release probability, unitary EPSP amplitude, PPRs, presynaptic calcium channel composition, and sensitivity to [Ca²⁺]_o and exogenously loaded buffers. These results suggest that developmental modulations of local microcircuits are connection-specific.

Angulo et al. (1999) reported a developmental switch from short-term depression to short-term facilitation in the layer 5 pyramidal to FS cell connections. Although, the presynaptic neurons described in that study belonged to a different subtype (thick tufted pyramidal cells), here we found similar changes take place in another layer 5 circuit formed by slender tufted pyramidal cells and the same type of interneurons. Note, that regardless of the origin of the excitatory input, FS interneurons receive a similar pattern of excitation: more depressing in young than old animals. Interestingly, the time point of the developmental switch in the mode of short-term synaptic plasticity is also similar in both types of connections. One possible explanation for the coincidence of the developmental alterations in two distinct types of terminals targeting the interneuronal subclass might be a common retrograde signal coming from the postsynaptic cell to the two classes of presynaptic cells. In this study, we made an attempt to find out a structural basis for the developmental changes in layer 5 pyramidal cells in the terminals. The major difference between P14 and P28 animals was distinct sensitivity to calcium channel blockers. In this study, we took advantage of the differences in the K_D of P- and Q-type calcium channels to Agatoxin and ω-conotoxin MVIIC. Q-type channels are 100 fold more sensitive to MVIIC (K_D = 5 nM) than P-type (K_D = 500 nM). P-type calcium channels are blocked at much lower concentrations of Agatoxin (K_D = 1 nM) than Q-type (K_D = 100 nM). Thus, at the concentration used in our experiments MVIIC mainly blocked Q-type channels, while Agatoxin acted on the P-type channels. Our results showed that in younger animals, release was mediated by both P- and Q- type channels. A significant contribution of P-type channels is indicated by the fact that only a mixture of MVIIC and Agatoxin resulted in an almost complete blockade of release. In P28 animals Agatoxin did not have any effect on synaptic transmission, suggesting that the role of the P-type calcium channels in transmitter release is negligible.

Surprisingly we found that the developmental increase in PPRs in layer 5A pyramidal to FS cell synapses is accompanied by reduced sensitivity to calcium chelators. The latter would imply a more compact organization of the release sites, with calcium channels more tightly coupled to the calcium sensor. However, if calcium entry remained the same this should lead to an increased release probability, and consequently increased synaptic depression. The most rational way to explain this contradiction is to assume that channels are indeed closer to the sensor, but they provide less calcium at the release site. For example, they might be fewer in numbers.

Another remarkable feature of layer 5A pyramidal cells is the existence of presynaptic washable factors that can reduce synaptic release, since prolonged dialysis of presynaptic slender tufted cells leads to increased release probability and a reduction in PPRs. One possible explanation might be washout of mobile endogenous calcium buffers from presynaptic terminals (Blatow et al., 2003). Indeed, recently it has been shown that layer 5A pyramidal cells express calretinin with the peak of immunoreactivity at P8. However, calretinin expression in the pyramidal neurons decreases at P14 and disappears by P30 (Liu et al., 2014). Therefore calretinin washout can only partly underlay the increase in EPSPs.

Although both of the microcircuits studied here are formed by similar neuronal cell types: presynaptic pyramidal cells and postsynaptic FS interneurons, they differ in a number of their presynaptic release properties. Layer 2/3 pyramidal cell terminals are more sensitive to any manipulation of the intraterminal calcium concentration. Lowering of extracellular calcium leads to stronger reduction of EPSP amplitudes at layer 2/3 pyramidal to FS cells synapses (EPSP norm. 0.31, 1 mM [Ca²⁺]_o, P28) than at those in layer 5 (EPSP norm. 0.63, 1 mM [Ca²⁺]_o, P28). The layer 2/3 pyramidal cell terminals also show much higher sensitivity to exogenously loaded buffers compared to layer 5 pyramidal cells. These observations are in apparent contradiction with the fact that layer 2/3 pyramidal cell inputs to FS cells are characterized by significantly higher unitary EPSP amplitudes (medians 2.43 vs. 1.5; L2 vs. L5; P28; p = 0.007; Kruskal-Wallis one way analysis of variance on ranks) and a slightly lower failure rate (1% vs. 0%; L2 vs. L5; P28) compared to layer 5 pyramidal to FS cell connections. A possible explanation for this might be the different calcium micro-domain organization in layer 2/3 and 5 pyramidal cell terminals. We can speculate that on the one hand the distance for diffusion between calcium channels and the calcium sensor is longer in layer 2/3 terminals than in layer 5 terminals. This would explain the higher sensitivity of release at layer 2/3 synapses to changes in intracellular calcium dynamics. On the other hand, a higher release probability implies higher calcium concentration at the release site, which might be due to a higher density of calcium channels or lower endogenous buffer capacity in layer 2/3 pyramidal cell terminals compared to terminals of layer 5A pyramids. The first assumption is indirectly supported by our finding that terminals of layer 2/3 and layer 5 pyramids express different subtypes of calcium channels, that may lead to a distinct spatial arrangement of calcium sensor and calcium channels at the release sites. A gradual increase in EPSP amplitudes at layer 5 synapses during prolonged recordings is indicative of washout of compounds, possibly calcium chelating molecules that affect calcium signaling.

Finally, the two synapses studied differ in their developmental profiles. While layer 2/3 pyramidal to FS cell terminals do not show any significant developmental changes, excitatory inputs from layer 5A pyramids to FS interneurons undergo drastic modification, changing them from strongly depressing to slightly facilitating synapses. All together our data shows that developmental mechanisms are different in these different synapses and therefore, detailed studies of the maturation of a given type of connection are critical for our understanding of neuronal network functional connectivity.

FS basket interneurons have been shown to play a key role in the generation of gamma rhythms in the neocortex and hippocampus (Fuchs et al., 2007; Cardin et al., 2009; Sohal et al., 2009). Two models are proposed as possible mechanisms for generating gamma oscillations. According to the first, a network consisting of only mutually connected inhibitory interneurons with distinct intrinsic firing properties and GABA_A receptor kinetics, can upon sufficient excitatory drive generate sustained oscillations. The second model is based on reciprocal connections between pools of excitatory pyramidal and inhibitory neurons. In such a network fast excitation and delayed feedback inhibition alternate leading to persistent rhythmic activity (for review, see Buzsaki and Wang, 2012). Indeed, the present findings suggest that excitatory drive of FS cells by local excitatory neurons is an important requirement for the generation of gamma oscillations (Middleton et al., 2008; Tiesinga and Sejnowski, 2009). Furthermore, in animal models of human psychiatric illness, weakening of excitatory input to FS cells, leads to impaired gamma oscillations (Cunningham et al., 2006; Sauer et al., 2015).

Our data suggest that local excitatory synapses between pyramidal cell and FS interneuron in layer 2/3 have significantly higher unitary EPSP amplitudes compared to layer 5. In addition, layer 2/3 connections are characterized by strong synaptic depression. This implies that during rhythmic or bursting activity the probability of recruiting downstream FS neurons is highest at the beginning of the burst. A similar scenario of crosstalk between pyramidal and FS cell is expected in young animals due pronounced PPD. However, in older rat when these synapses switch the mode of short term plasticity to facilitation, shifting efficiency of synaptic transmission towards latter AP in the burst. All together this might result in higher synchrony in upper layer 2/3 compared with the deep layers of the cortex. Indeed, it has been shown with in vivo experiments carried on rhesus monkeys that gamma in superficial layers is characterized by significantly higher synchrony compared to deep layers (Buffalo et al., 2011). It was also found that spike-field coherence in the gamma frequency range was largely confined to the superficial layers, while the deep layers showed maximal coherence at low frequencies. The layer-specific differences in efficacy and short term synaptic plasticity at local pyramidal to FS interneuron synapses might contribute to laminar specificity of network activities.