Procedures of animal care and slice preparation approved by the “Universidad Autónoma de Madrid” and “Consejo Superior de Investigaciones Científicas” follow the guidelines laid down by the European Council on the ethical use of animals (Directive 2010/63/EU) and every effort was made to minimize animal suffering and number. The procedures have been described in detail elsewhere (Nuñez et al., 2012).

Recordings were obtained from slender tufted L5A PNs (N = 244) and confirmed by intracellular biocytin staining (Nuñez et al., 2012). In control ACSF, basal stimulation induced an EPSP that above a threshold depolarization level triggered a single AP (Figure 1C). Increasing stimulation intensity could increase the number of APs (1–3 APs, with frequent failures) without further modifying the response. In contrast, under P_ITX (50 μM) blockade of GABA_A inhibition, the EPSP could trigger at short delays of 5 ± 3 ms 1 AP or a brief high frequency burst of 2 and occasionally 3 APs with failures, that lasted 20 ± 8 ms (N = 14). The AP burst rode on a slow high amplitude depolarization wave (peak amplitude 53 ± 6 mV; duration 375 ± 60 ms; same cells; Figures 1C–F, 2A–C). The slow depolarization wave was not modified when stimulation intensity was further increased, displaying an all-or-none behavior.

The amplitude and duration of the slow depolarization wave was markedly reduced when either nifedipine (20 μM; reaching values of 47 ± 7% of the controls; P < 0.001; N = 10) or D-AP5 were superfused (50 μM; reaching values of 28 ± 5% of the controls; P < 0.001; N = 6) (Schiller and Schiller, 2001; Nuñez et al., 2012). When both D-AP5 and nifedipine were superfused the slow depolarization wave was totally suppressed (P < 0.001; N = 9; Figures 1F,G). In addition, a robust cytosolic Ca²⁺ signal that could be recorded in the soma, where the ΔF/F₀ reached values that were 52 ± 9% of the controls (P < 0.001; N = 6), while in basal and apical dendrites, the ΔF/F₀ attained values that were 19 ± 9% of the controls (P < 0.001; same cells; Figures 1B–E). Much smaller Ca²⁺ signals (8 ± 1% of controls, P < 0.01; same cells) were induced when a single AP was triggered in the absence of Ca²⁺ spikes (Figures 1B,C). The all-or-none behavior of the slow depolarization wave following EPSPs and the robust dendritic Ca²⁺ signals suggests that the slow depolarizations were Ca²⁺ spikes.

We first tested if EPSP-Ca²⁺ spikes triggered by low-frequency basal stimulation could induce LTP. Under blockade of GABA_ARs with P_ITX (50 μM), basal stimulation induced inward EPSCs at −65 mV with mean peak amplitudes of −160 ± 9 pA (P < 0.001; N = 6). After a 5–10 min control recording of EPSCs under voltage-clamp with stimulation at 0.3 Hz, the recording was switched to current-clamp and stimulation intensity was increased until EPSP-Ca²⁺ spikes were evoked without failure. In these conditions the delay between the EPSP and the AP burst-Ca²⁺ spike was fixed for a given experiment although it could fluctuate between ≈ 2 and ≈ 20 ms in different experiments. EPSP-Ca²⁺ spikes were applied 60 times at 0.2 Hz, the recording was switched back to voltage-clamp, and stimulation intensity and frequency restored to the initial control conditions. The repeated EPSP-Ca²⁺ spikes induced a robust LTP typified by an increase in the mean peak EPSC amplitude that in ≈ 30 min reached values that were 210 ± 45% of the control (P < 0.001; N = 23; Figures 2D,E).

We also tested if the same basal stimulation protocol could induce plasticity when synaptic inhibition was active in control ACSF. In control ACSF Ca²⁺ spikes were never evoked by repeated basal stimulation and EPSC amplitudes were essentially identical to the controls, reaching values that were 97 ± 2% of the control (P > 0.05; N = 8; Figure 2E) in ≈ 30 min. The above results suggest that this LTP was intimately dependent on active dendritic mechanisms under close control by GABA_A inhibition (Wigström and Gustafsson, 1983; Kampa et al., 2007; Marlin and Carter, 2014).

To verify the pre- or postsynaptic origin of this LTP we first tested for possible increases in release probability at excitatory synapses. There were no modifications in PPR or the 1/CV² ratio (Figures 3A–C), suggesting that this LTP was not associated with changes in the probability of glutamate (Glu) release, and that there was no presynaptic contribution to it. In addition, both EPSCs and currents evoked by Glu puffs (that bypass the presynaptic components of Glu transmission) were potentiated to similar values (166.5 ± 22.3%; P < 0.01; N = 6 and 169.2 ± 23.7, P < 0.05; same cells, for the EPSCs and Glu currents, respectively) following the induction of LTP (Figures 3D,E).

To determine the contribution of the Ca²⁺ spike to the induction of this LTP under P_ITX (50 μM), we inhibited Ca²⁺ spikes through a blockade of NMDARs with D-AP5 (50 μM), and of L-type VGCCs with nifedipine (20 μM). Basal stimulation was increased well above the intensity (x2) at which the EPSP triggered APs to compensate for the EPSP amplitude reduction caused by the blockade of the NMDA component. In these conditions the Ca²⁺ spike was blocked (Figures 1F,G) and basal stimulation (60 times at 0.2 Hz) was unsuccessful in inducing LTP (EPSCs reached values that were 98 ± 9% of the controls, P > 0.05; N = 6; Figure 4B).

In L5 PNs STDP requires pairing an EPSP with an AP burst induced by depolarizing current injection to rescue AP back-propagation through the generation of a Ca²⁺ spike (Larkum et al., 1999b; Kampa et al., 2004). Accordingly, we tested the effects of avoiding the AP burst by antagonizing voltage-gated Na⁺ channels with intracellular QX-314 under PITX (50 μM). Under QX-314-loading (5 mM in the pipette solution) APs were inhibited and the EPSP and Ca²⁺ spike remained (see below). In addition, under QX-314 the peak amplitude of Ca²⁺ spikes (58 ± 5 mV, N = 6) were even larger than those linked with the induction of LTP (53 ± 6 mV, see above), likely indicating that what was required for LTP was not just the Ca²⁺ rise, but that Na⁺-mediated back-propagating APs played a key role. In these conditions, repeated basal stimulation (60 times at 0.2 Hz) induced a robust LTD instead of LTP, and EPSCs reached values that were 48 ± 6% of the controls (P < 0.01; N = 6) ≈ 30 min after the onset of stimulation (Figure 4B). These results suggest that the AP burst and Ca²⁺ spike were essential to the induction of this LTP.

To determinate the contribution of the cytosolic Ca²⁺ rise in the LTP induction, we tested the effects of BAPTA-loading, which chelates Ca²⁺, preventing a rise of Ca²⁺ in the cytosol. BAPTA-loading (40 mM in the pipette solution) did not modify the EPSP-Ca²⁺ spike (Figure 4C), and repeated basal stimulation (60 times at 0.2 Hz) was unable to induce LTP while EPSCs reached values that were 103 ± 6% of the controls (P > 0.05; N = 6) ≈ 30 min after the induction process (Figures 4B,C).

Because Ca²⁺ release from IP3-sensitive stores and Ca²⁺ induced-Ca²⁺ release (CICR) through ryanodine receptors can contribute to the cytosolic Ca²⁺ rise we tested the effects of blocking the ryanodine receptors with intracellular ruthenium red (Ru-Red 400 μM in the pipette solution). Ru-Red blocked the LTP without modifying the EPSP-Ca²⁺ spike, while EPSCs reached values that were 99 ± 11% of the controls (P > 0.05; N = 6; Figure 4C) ≈ 30 min after the induction process. We next examined the effects of blocking IP₃Rs with intracellular heparin (5 mg/ml in the pipette solution). In these conditions the LTP was inhibited without modification of the EPSP-Ca²⁺ spike and the EPSCs reached amplitudes that were 86 ± 8% of the controls (P > 0.05 N = 6) ≈ 30 min after 60 basal stimulations at 0.2 Hz (Figure 4D). Therefore, a rise in cytosolic Ca²⁺ was required for the induction of this LTP and was produced by influx through NMDARs, L-type VGCC and release from intracellular stores.

Synaptic plasticity is controlled by intracellular cascades in which G-protein coupled receptors (GPCRs) play key roles (Mukherjee and Manahan-Vaughan, 2013). Therefore, we first tested the effects of blocking G-proteins by loading the PN with GDPβS. With intracellular GDPβs (1 mM in the pipette solution) EPSP-Ca²⁺ spikes continued but failed to induce LTP. Thirty min after the induction protocol, EPSCs reached values that were 101 ± 6% of the controls (P > 0.05; N = 6; Figures 5A,D). Since metabotropic receptors are coupled to G-proteins we checked whether metabotropic Glu receptors (mGluRs) were involved in LTP induction. Superfusion with MCPG (1.0 mM), a group I/II mGluR antagonist, prevented LTP and, ≈ 30 min after induction, EPSCs reached values that were 92 ± 25% of the controls (P > 0.05; N = 4, Figure 5D). Although there was a small but not significant increase in EPSC amplitude, there was no LTP with the selective mGluR1 antagonist LY367385 (50 μM) and EPSC amplitudes reached values that were 109 ± 25% of the controls (P > 0.05; N = 5; Figures 5C,D). In contrast, a robust LTP was induced ≈30 min after the EPSP-Ca²⁺ spike in the presence of the mGluR5 selective antagonist MPEP (5.0 μM), with EPSC amplitudes that reached 196 ± 32% of the controls (P < 0.01; N = 6; Figures 5C,D). Therefore, mGLuR1 activation was required to induce the LTP.

Muscarinic AChRs (mAChRs) play a key role in certain forms of long-term enhancement of excitatory synaptic transmission (Fernández de Sevilla et al., 2008; Buchanan et al., 2010; Fernández de Sevilla and Buño, 2010; Nuñez et al., 2012; Dennis et al., 2016). Accordingly, we tested the effects of the non-selective mAChR antagonist atropine (0.3 μM), which prevented LTP induction. Under atropine EPSCs reached values that were 101 ± 2% of the controls (P > 0.05; N = 5) ≈ 30 min after the induction process (Figure 5D). DAU5884 (1 μM), a selective subtype 3 mAChR antagonist, blocked the LTP and EPSCs reached values of 103 ± 4% of the controls (P > 0.05; N = 5) ≈ 30 min after induction (Figures 5B,D). Pirenzepine (75 nM), a selective M1 mAChR antagonist, did not prevent the LTP and EPSCs reached values of 168 ± 5% of the controls (P < 0.01; N = 5) ≈30 min after induction (Figure 5D). Nicotinic AChRs (nAChRs) have also been involved in the regulation of transmitter release and synaptic plasticity. Therefore, we tested the effects of blockade of α4β2-containing and α7-containing nAChRs using MMA (10 μM) plus MLA (50 μM). This prevented LTP induction and EPSC reached values that were 102 ± 4% of the controls (P > 0.05; N = 5; Figure 5D). Taken together these results suggest that this LTP required the participation of GPCRs and activation of mGluR R1 and subtype M3 mAChRs, as well as nAChRs. Note that none of these treatments modified the EPSP-Ca²⁺ spikes, suggesting that the inhibition of LTP occurred downstream of the cytosolic Ca²⁺ rise.

Cytosolic Ca²⁺-mediated activation of intracellular kinases can induce LTP through an increase in the number of functional AMPARs in dendritic spines (Fernández de Sevilla et al., 2008). Kinases can also enhance NMDAR-mediated responses by changing the biophysical properties of NMDARs (Fernández de Sevilla and Buño, 2010). We therefore tested whether activation of the CaMKII was required to induce the LTP. Blockade of CaMKII with the peptide inhibitor AIP (5 μM in the pipette solution) suppressed LTP without preventing the EPSP-Ca²⁺ spike and ≈30 min after the induction process EPSCs reached values that were 93 ± 10% of the controls (P > 0.05; N = 4; Figures 6A,C).

We have shown that Ca²⁺ release from endoplasmic reticulum stores by activation of IP₃Rs plays a key role in the cholinergic LTP in CA1 PNs (Fernández de Sevilla and Buño, 2010). The Ca²⁺ released plays a key role in long-term enhancement excitatory synaptic transmission (see above). Accordingly, we tested the effects of inhibiting the production of IP3 by blocking the PLC translocation with intracellular U73122. U73122 (2 mM in the pipette solution) inhibited the LTP without preventing the EPSP-Ca²⁺ spike and EPSCs reached values that were 97 ± 18% of the controls (P > 0.05; N = 6; Figures 6A,C). We next investigated the effects of inhibiting PKC with intracellular chelerytrine (5 μM); there was no effect on LTP and EPSCs reached values that were 191.47 ± 30% (P < 0.001; N = 5) of the controls ≈30 min after induction. In contrast, PKA blockade with H89 dihydrochloridre (10 μM in the pipette solution) had no effect on the EPSP-Ca²⁺ spike but did inhibit the LTP and ≈30 min after induction, EPSCs reached values that were 92.4 ± 4% of the controls (P > 0.05; N = 6; Figures 6B,C). The above results suggest that this LTP requires CaMKII, PLC, and PKA activation.

Protocols that produce strong membrane depolarization induce LTP while those that cause modest or local depolarization generate LTD (Holthoff et al., 2004). This bidirectional behavior is thought to be caused by the level of Ca²⁺ influx during dendritic depolarization (Artola et al., 1990; Bliss and Collingridge, 1993; Neveu and Zucker, 1996; Dan and Poo, 2004; Kampa et al., 2007). Accordingly, we analyzed the effects of membrane hyperpolarization during the induction process. Hyperpolarization to −100 mV decreased the amplitude and duration of averaged EPSP-Ca²⁺ spike to 27 ± 2 mV and 50 ± 8 ms (P < 0.01; N = 5). In these conditions repeated basal stimulation (60 times at 0.2 Hz) induced an LTD that rapidly reached values of 63 ± 1% of the controls (P < 0.001; N = 5; Figure 7A). A NMDAR blockade with D-AP5 (50 μM) reduced the average amplitude and duration of Ca²⁺ spikes to 25 ± 4 mV and 55 ± 13 ms (P < 0.01; N = 5). In these conditions, basal stimulation (as above) produced a potent LTD that reached values that were 56 ± 3% of the controls (P < 0.01; N = 6) (Figure 7B). Nifedipine (20 μM) blockade of L-type VGCC also reduced the amplitude and duration of the Ca²⁺ spikes to 21 ± 1 mV and 100 ± 9 ms (P < 0.01; N = 5). Thirty min after induction, repeated basal stimulation (as above) induced a slowly declining LTD that reached values of 77 ± 5% of the controls (P < 0.05; N = 5; Figure 7C). A single or a pair of AP and Ca²⁺ spikes was followed by hyperpolarization in Figures 7A–C.

Moreover, a robust LTD that reached values of 48 ± 6% of the controls (P < 0.01; N = 6; Figures 7D, 4B) was induced when APs were inhibited by QX-314. Under QX-314 Ca²⁺ spikes were larger than those linked with the LTP (see above). These results suggest that what was required for LTP was not just the Ca²⁺ rise, but that Na⁺-mediated back-propagating APs played a role. In contrast, with the LTP that increased gradually in amplitude following induction, the LTD in these circumstances was fully developed at the end of the induction process.

The Ca²⁺ spike is strongly linked with Ca²⁺ influx, consequently an analysis of the relationship between the Ca²⁺ spike and the cytosolic Ca²⁺ signal could provide a direct estimate of the conditions that induce the bidirectional synaptic plasticity. Therefore, we recorded the somatic Ca²⁺ signals associated with the Ca²⁺ spike in control ACSF and under PiTX, PiTX + D-AP5, and PiTX + nifedipine. In control ACSF small Ca²⁺ signals with ΔF/F₀ values of 8 ± 1% of the controls (P < 0.01; N = 6) were induced when APs were triggered in the absence of Ca²⁺ spikes. Under PiTX (50 μM) the ΔF/F₀ reached much higher values that were 52 ± 9% of the controls (P < 0.001; N = 6), whereas lower values of 25 ± 5% (P < 0.00; same cells) were attained when D-AP5 (50 μM) was added to block NMDARs. Under nifedipine (50 mM) added to inhibit L-type VGCC, ΔF/F₀ achieved values of 27 ± 5% from the controls (P < 0.001; N = 5) (Figure 7E). Taken together the above results suggest that the direction of the induced change in synaptic plasticity could be regulated by the degree of membrane depolarization and the cytosolic Ca²⁺ level attained during the EPSP-Ca²⁺ spike.

To verify the pre- or postsynaptic origin of this LTD we first tested for possible decreases in release probability at excitatory synapses. We plotted the noise-free coefficient of variation (CV_NF) vs. time. Under all LTD-inducing manipulations, the CV_NF values remained below 1.0 (Figure 7F), suggesting the absence of significant changes in Glu release probability during an LTD and meaning that the synaptic depression originated postsynaptically.

We first tested if repetitive low-frequency deflections delivered at the principal whisker (Figure 8A) could induce long-term response changes resembling those that occur in vitro. The 37 L5 neurons recorded were either silent or displayed a low spontaneous firing rate (0.5–2 APs/s), and responded to contralateral displacements of the principal whisker. Control whisker-evoked responses had on average 3.5 ± 0.5 APs/stimulus (measured from 0 to 100 ms after the stimulus onset). The low spontaneous AP firing rate and the activation by deflection of the principal whisker provide strong support to the notion that recordings were obtained from L5A PNs, as has been described previously (Manns et al., 2004; de Kock and Sakmann, 2009).

Following a 60 s control stimulation at 0.5 Hz a train of 40 stimuli at 1.0 Hz induced a long-lasting enhancement of the response (or LTP) from 3.9 ± 0.4 APs/stimulus in control to 5.3 ± 0.6 APs/stimulus, measured 30 min after induction in 16 neurons out of 23 (or 66%; P < 0.001; Figures 8B,D,E), while 5 neurons (22%) reduced their response (or LTD; from 3.5 ± 0.5 APs/stimulus in the control to 2.4 ± 0.1 APs/stimulus; P < 0.01; Figures 8C–E) and 2 neurons (9%) were not affected (from 3.2 ± 0.6 APs/stimulus in the control to 3.4 ± 0.4 APs/stimulus after induction; P > 0.05). Figure 8E shows the time course of the AP response plasticity.

Interestingly, the whisker-evoked response was enhanced during the 1 Hz induction from 3.9 ± 0.41 APs/stimulus in the control to 5.5 ± 0.58 APs/stimulus (P < 0.01; N = 5) in neurons that showed LTP (Figure 8B). In contrast, in neurons that showed LTD whisker-evoked responses were not altered during induction (from 3.5 ± 0.48 APs/stimulus in control to 2.9 ± 0.37 APs/stimulus; P > 0.05; N = 5; Figure 8C).

We next tested if blockade of NMDARs by injection of D-AP5 (50 μM; 0.1 μl) in L5 (see Section Materials and Methods) prevented plasticity. To exclude possible artifacts caused by this manipulation we checked that the D-AP5 injection did not modify AP amplitudes. Under the effects of D-AP5 the whisker-evoked response measured ≈30 min after the 1 Hz induction stimulation train was essentially identically to the control response and plasticity was absent (from 2.6 ± 0.38 in control to 2.5 ± 0.36 APs/stimulus after induction; P > 0.05; N = 14; Figure 8D). Therefore, the above results suggest that the activation of NMDA receptors in L5 play a key role in the genesis of the bidirectional synaptic plasticity.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.