Plasticity in Single Axon Glutamatergic Connection to GABAergic Interneurons Regulates Complex Events in the Human Neocortex

Abstract

In the human neocortex, single excitatory pyramidal cells can elicit very large glutamatergic EPSPs (VLEs) in inhibitory GABAergic interneurons capable of triggering their firing with short (3-5 ms) delay. Similar strong excitatory connections between two individual neurons have not been found in nonhuman cortices, suggesting that these synapses are specific to human interneurons. The VLEs are crucial for generating neocortical complex events, observed as single pyramidal cell spike-evoked discharge of cell assemblies in the frontal and temporal cortices. However, long-term plasticity of the VLE connections and how the plasticity modulates neocortical complex events has not been studied. Using triple and dual whole-cell recordings from synaptically connected human neocortical layers 2-3 neurons, we show that VLEs in fast-spiking GABAergic interneurons exhibit robust activity-induced long-term depression (LTD). The LTD by single pyramidal cell 40 Hz spike bursts is specific to connections with VLEs, requires group I metabotropic glutamate receptors, and has a presynaptic mechanism. The LTD of VLE connections alters suprathreshold activation of interneurons in the complex events suppressing the discharge of fast-spiking GABAergic cells. The VLEs triggering the complex events may contribute to cognitive processes in the human neocortex, and their long-term plasticity can alter the discharging cortical cell assemblies by learning.

Fig 1. A subset of human neocortical PCs innervate FSINs with VLEs.

(A) Triple whole-cell recording demonstrating the rich glutamatergic connectivity in the human neocortex layers 2–3 (L2–3). Two PCs (PC1 red and PC2 green) in L2–3 synaptically excite the same FSIN (blue). (A1) Partial reconstruction of the cells with color-coding presented in the schematic inset. Scale 25 μm. Confocal images illustrate positive immunoreactions of the neurobiotin (nb, Cy3)-filled interneuron axon boutons for vgat (Cy5, arrows in merged image) and pv (Alexa488, arrow). Scales 5 μm. (A2–3) Sample traces show presynaptic spikes (superimposed) and postsynaptic currents in the synaptic connections (cells voltage clamped at −60 mV). PC1 generates large monosynaptic EPSC in the interneuron (A2), whereas PC2 evokes small EPSC in the same cell (A3). In addition, PC2 is synaptically connected to PC1. The EPSCs from the PCs show fast kinetics in the interneuron, whereas EPSC in the PC–PC connection is slow. (A4) The two glutamatergic inputs to the FSIN show very different amplitude EPSPs and distinct paired-pulse ratios in current clamp (at Em –69 mV). (B) Histograms show the distribution of average EPSP amplitude (1 mV bin, failures excluded) in 16 identified L2–3 PC–PC pairs (B1) and in 22 PC–FSIN pairs (B2). (B3) The very large and the small amplitude EPSCs from PCs to FSINs show similarly fast time-to-peak time. Values are average EPSCs (of at least five) from individual pairs. The underlying data are shown in S1 Data.

Fig 2. Single fiber connections to FSINs with large EPSP show group I mGluR-dependent LTD.

(A) Paired recordings from synaptically connected layer 2–3 PCs and FSINs with large amplitude EPSCs/EPSPs show LTD. The LTD is generated by the presynaptic PC firing 40 Hz bursts (5 pulses, 40 times), while the postsynaptic cell is at Em. (A1) Partial reconstruction of one recorded PC (red, axon orange)–FSIN (blue, axon light blue) pair with large EPSCs/EPSPs. Scale 50 μm. L2–3 indicates layer 2–3. Schematic shows experimental design and color-coding for the cells and traces. Confocal micrographs illustrate vgat+ and pv+ axon bouton of the FSIN filled with neurobiotin (nb, scale 2 μm). A PC spike and averaged EPSC (5 at −60 mV) in the cell pair below. Scales 1 nA and 100 pA/5 ms. (A2) Single AP-evoked EPSP amplitude (interval 10 s) in the same experiment at baseline and following the PC 40 Hz burst firing (arrow at 0 time point). The afferent single fiber burst firing induced LTD (at 20−25 min p < 0.001, paired t-test). The EPSPs (blue, average of 5 at Em –62 mV) and presynaptic cell spikes (red) at different time points and one 40 Hz burst illustrated on top. The FSIN is at Em during the recording. (A3) Mean ± standard error of the mean (s.e.m.) in five PC–FSIN pairs with large EPSP (5.85 ± 0.59 mV at baseline, showing no failures) show prominent LTD (30 s bin, baseline-normalized, n = 5 pairs, p < 0.01, Wilcoxon test). (A4) The LTD requires group I mGluRs. A PC–FSIN pair with large EPSP in the presence of group I mGluR blockers LY367385 (100 μM) and 2-Methyl-6-(phenylethynyl)pyridine hydrochloride (MPEP, 25 μM) (applied 5 min before the bursts indicated by arrow). The EPSPs (blue, at Em −67 mV) and presynaptic cell spikes (red) shown on top. The FSIN is at Em during recording. (A5) Mean ± s.e.m. of similar PC–FSIN pairs with large EPSPs (8.42 ± 2.83 mV in baseline, showing no failures) in four experiments (30 s bin, baseline-normalized). The underlying data are shown in S2 Data. (B) EPSP analyses indicate presynaptic LTD. (B1) LTD (n = 5) is associated with an increased ratio of the EPSP amplitude SD/mean illustrated here as decreased baseline-normalized CV⁻² (mean ± s.e.m. black asterisk, p < 0.05, Mann-Whitney test). Red asterisk compared with the non-LTD experiments (n = 4) (p < 0.05, Mann-Whitney test). (B2) Likewise, the EPSP amplitude PPR (1^st versus 2^nd EPSP) is reduced in the LTD experiments (black asterisk, p < 0.05, Mann-Whitney test), but not in the presence of group I mGluR blockers. Red asterisk indicates significance between the groups (p < 0.05, Mann-Whitney test). Baseline-normalized time window is 20–25 min after afferent bursts. The data are available in S3 Data. (B3) Sample traces from one experiment above showing the PC firing (paired-pulse 50 ms)-evoked EPSPs in the FSIN during baseline and in LTD.

Fig 3. LTD fails in weak single-fiber PC-FSIN connections, but is generated by coactivity of multiple glutamatergic fibers.

(A) LTD fails in PC–FSIN pairs with small EPSP. (A1) Schematic shows experimental design. A presynaptic PC spike (red) with postsynaptic EPSCs (blue, average of 5 at −60 mV) in one recording and confocal micrographs of the FSIN axon (nb, neurobiotin) with pv+ boutons (scale 5 μm, arrows point colabeling in merged image). (A2) One PC–FSIN pair with the EPSPs in baseline and after the afferent cell 40 Hz bursts (arrow). Averaged EPSPs (5) at Em −72 mV on top at different time points and a 40 Hz burst. Postsynaptic cell is at Em (current clamp) during the recording and the bursts. (A3) Mean ± s.e.m. (30 s bin, baseline-normalized) of similar experiment in five PC–FSIN pairs with small amplitude EPSPs (1.89 ± 0.43 mV in baseline with failures). (A4) Failure of the LTD in weak PC–FSIN connections is not due to insufficient postsynaptic depolarization. Plot shows EPSP in one PC–FSIN pair before and following the presynaptic bursts, now paired with FSIN depolarization beyond the firing threshold (see Methods). Averages of EPSPs (5 at Em −66 mV) and a 40 Hz burst with simultaneous depolarization (30 mV, 250 ms) in voltage clamp shown on top. (A5) Mean ± s.e.m. of five similar experiments with small EPSP (1.44 ± 0.22 mV in baseline with failures) PC–FSIN pairs (baseline-normalized, 30 s bin). The underlying data are shown in S4 Data. (B) Connections between PCs exhibit small amplitude EPSPs with no long-term plasticity when PC1 bursts, while PC2 is at resting membrane potential. (B1) EPSP amplitude in one experiment before and after the 40 Hz presynaptic cell bursts (arrow, postsynaptic cell at Em −78 mV). Averaged EPSPs (five at Em) shown on top with a 40 Hz burst, and a schematic showing the experimental design. (B2) Mean ± s.e.m. of baseline-normalized EPSPs (1.40 ± 0.30 mV in baseline with failures) in four PC–PC pairs as in B1 (30 s bin) (S4 Data). (C) Activation of multiple afferent pathways to FSINs using extracellular stimulation reveals group I mGluR-dependent LTD in weak PC–FSIN synapses (S5 Data). (C1) One experiment with monosynaptic EPSC in FSIN (voltage clamped at −60 mV) at baseline and following the 40 Hz bursts applied to the stimulation pathway (arrow at 0 time point). Inset traces (averages of 5) show evoked EPSCs in baseline and in LTD. The monosynaptic component is indicated by dotted vertical line. Schematic shows experimental design. (C2) Mean ± s.e.m. of seven baseline-normalized experiments as in C1 showing the LTD in control conditions (open symbols, p < 0.001, paired t-test) and blockade of the LTD in experiments with LY367385 (100 μM) and MPEP (25 μM) (solid symbols, n = 7, paired t-test). (C3) Generation of group I mGluR-dependent LTD by 40 Hz stimulation is conserved in mammalian neocortex occurring also in rat FSINs. Multiple fiber extracellular stimulation with LTD in rat L2–3 somatosensory cortex FSINs. Open symbols show experiments in control conditions (n = 5, p < 0.01) and solid symbols in the presence of LY367385 (100 μM) and MPEP (25 μM) (n = 5) (Wilcoxon test). Blockers for glutamate N-methyl-D-aspartatereceptors (NMDARs) (DL-2-Amino-5-phosphonopentanoic acid; DL-APV, 100 μM) and GABA_ARs (PiTX, 100 μM) were present in C1–C3. (C4-C5) Likewise, LTD of the EPSCs in both species is associated with an increased amplitude SD versus the mean. Data shows decreased CV⁻² (baseline-normalized at 20 min after 40 Hz) in LTD in control conditions, but not when LTD is blocked in the presence of group I mGluR blockers (p < 0.05 between groups, Mann-Whitney test).

Fig 4. The LTD depresses discharge of FSINs in complex events.

(A) Large EPSPs and APs in FSIN (blue) with short 3 to 5 ms delay elicited by single PC spike (red, peak indicated by vertical line). The figure shows four consecutive cycles with 10 s interval in one vgat+ and pv+ FSIN at Em (−69 mV) (note short AP duration, the positive peaks are indicated in the abscissa). Schematic shows experimental setting. (B) Single PC spike triggers disynaptic GABAergic currents in the layers 2–3. Dual whole-cell PC recordings (voltage-clamp) show that single PC1 APs trigger dIPSCs in PC2 with high probability and short delay. Schematic shows experimental design. (B1) Traces show consecutive events (4) in one experiment. The dIPSC onsets are marked in abscissa. (B2) Histograms (ordinates normalized and show from 0 to 1) illustrate delay distribution of the first dIPSC onset in four experiments (each 6–9 min, indicated as pairs 1–4) in control conditions. Most evoked dIPSCs are phase-locked to the presynaptic PC spike with <10 ms delay with obvious moderate variability of the mode of delay between experiments (S6 Data). (C) The dIPSCs are generated by glutamatergic excitation. Sample traces show five consecutive dIPSCs between PC1 and PC2 in baseline conditions and the blockade with AMPAR blocker GYKI53655 (25 μM) (see S4 Fig). (D–E) The single AP-evoked dIPSCs between PCs are stable over a long period (S6 Data). (D) Raster plot illustrates timing of dIPSC onset (in PC2) evoked by an AP in the presynaptic PC (PC1) in one 30 min experiment. Consecutive (6) presynaptic spikes and dIPSCs at different time points shown on top. (E) Three experiments (pairs 1, 2, and 3) as in D (pair 1), illustrated with histograms showing the dIPSC onset delay in different time windows (0–5 min, 10–15 min, and 20–25 min). The dIPSC probability and delay are stable for at least 30 min (Chi-square test). (F–G) dIPSCs show LTD after presynaptic PC 40 Hz burst firing (S6 Data). (F) 40 Hz spike bursts in the presynaptic PC (similar to the LTD experiments in Fig 2) induce LTD of the dIPSCs. Raster plot shows dIPSCs in one paired PC recording. After baseline, the 40 Hz spike bursts in PC1 (at 0 time point, dotted horizontal line) induce permanent depression of the dIPSC occurrence. Traces illustrate the presynaptic cell spike and the dIPSCs (6) at baseline and the absence of dIPSCs after 20 min. (G) Three similar experiments (pairs 1, 2, and 3) as in F (pair 1), showing the LTD of dIPSCs after the 40 Hz presynaptic bursts (BL −5 to 0 min) (p < 0.01, Chi-square test). (H–I) The LTD of dIPSCs is blocked with group I mGluR antagonists (S6 Data). (H) Similar experiment as in F, but in the presence of LY367385 (100 μM) and MPEP (25 μM). Traces on top show the pre- and disynaptic currents at baseline and 20 min after the 40 Hz bursts. (I) Three experiments as in H (pair 1) illustrated with dIPSC onset delay histograms at the baseline (−5 to 0 min) and at two time windows after the 40 Hz presynaptic bursts.