A cooperative switch determines the sign of synaptic plasticity in distal dendrites of neocortical pyramidal neurons

Abstract

Pyramidal neurons in the cerebral cortex span multiple cortical layers. How the excitable properties of pyramidal neuron dendrites allow these neurons to both integrate activity and store associations between different layers is not well understood, but is thought to rely in part on dendritic backpropagation of action potentials. Here we demonstrate that the sign of synaptic plasticity in neocortical pyramidal neurons is regulated by the spread of the backpropagating action potential to the synapse. This creates a progressive gradient between LTP and LTD as the distance of the synaptic contacts from the soma increases. At distal synapses, cooperative synaptic input or dendritic depolarization can switch plasticity between LTD and LTP by boosting backpropagation of action potentials. This activity-dependent switch provides a mechanism for associative learning across different neocortical layers that process distinct types of information.

Figure 1. AP-EPSP Pairing Evokes LTP at an L5-to-L5 Connection, but LTD at an L2/3-to-L5 Connection

(A) AP-EPSP pairing at a unitary L5-to-L5 connection resulted in LTP. After the baseline period (blue dotted line and blue trace, top middle inset), pre- and postsynaptic 50 Hz AP trains were paired (pre leading before post by 10 ms; top right inset). The post-pairing responses (red dotted line and red trace, top middle inset) were significantly potentiated (after/before = 195%, p < 0.001; EPSP rise time = 1.9 ms), in agreement with prior studies (Markram et al., 1997b; Sjöström et al., 2001). (Left) Illustration of recording configuration for L5-to-L5 pairs. (B) The same conditions and induction protocol as in (A) in this L2/3-to-L5 paired recording, however, resulted in significant LTD (after/before = 63%, p < 0.001; EPSP rise time 3.5 ms). (Left) Recording configuration.

Figure 2. Identifying Synaptic Locations from EPSP Rise Time and Morphology

(A) Reconstruction of an L2/3-to-L5 connected pair (blue circles, putative synaptic contacts). (B) Using a computer model of the reconstructed L5 neuron in (A), calibrated to the actual EPSP rise time (5.0 ms) of this paired recording by assuming equal contribution of the two putative synaptic contacts, we simulated the activation of synapses in different locations of the dendritic tree. This gave rise to EPSPs with different rise times (color-coded according to synapse location). (C) Based on this simulation, we predicted the relationship between distance and somatic EPSP rise time for many synaptic locations. Dashed line: 3 ms cut-off that defines proximal versus distal inputs (see text). (D) In several paired recordings, putative synaptic contacts were identified (cf. Figure S1), which enabled the verification of the simulation (cf. [B]). (E) The mean location of putative synaptic contacts correlated well (r = 0.89) with the somatic EPSP rise time in L5-to-L5 (black triangles) and L2/3-to-L5 (closed blue circles) paired recordings (n = 13), thus corroborating the model prediction (cf. [C]).

Figure 3. Synaptic Location Determines the Sign of Plasticity

AP-EPSP pairing in L5-to-L5 pairs (closed black triangles, n = 34), L2/3-to-L5 pairs (closed blue circles, n = 19), and extracellular stimulation experiments (open black squares, n = 21) resulted in LTP of proximal inputs (somatic EPSP rise time < 3 ms; 137% ± 4.9%, n = 46), but evoked LTD of distal input (somatic EPSP rise time > 3 ms; 80.2% ± 4.5%, n = 28, p < 0.001). (Inset) In agreement with this, the sign of plasticity was inversely correlated with average distance of putative synaptic contacts (n = 11; also see Figure S1). Dashed red lines are power law fits.

Figure 4. LTP Induction at Distal Synapses Requires Cooperativity among Inputs

(A) Strong extracellular stimulation near the distal dendritic tuft (left) recruited a slow, large EPSP (rise time 5.2 ms) that significantly potentiated when paired with APs at 50 Hz (137%, p < 0.001). (Left) Configuration permitted multiple parallel recordings. (B) A distal L2/3-to-L5 unitary input (rise time 4.2 ms) that was boosted by extracellular stimulation during the induction (compound EPSP 5.7 mV) potentiated significantly after AP-EPSP pairing (137%, p < 0.001). (Left) Stimulation electrode placement is illustrated schematically. (C) Using extracellular stimulation, pairing small distal EPSPs with APs resulted in LTD (closed red squares, 70% ± 5%, n = 15). Recruiting more inputs by increasing extracellular stimulation strength, however, resulted in large EPSPs that potentiated robustly when paired with APs (open red squares, 137% ± 7%, n = 12; p < 0.001), demonstrating a cooperativity requirement for distal LTP. Similarly, although distal unitary connections typically depressed (open blue circles, 86% ± 9%, n = 8; cf. Figure 3), boosted distal unitary connections (cf. [B]) significantly potentiated (126% ± 7%, n = 4, p < 0.05). (D) For distal inputs, LTD (80% ± 5%, n = 25) was observed below a threshold somatic EPSP amplitude (<1.0 mV). Above this threshold, distal plasticity switched to LTP (126% ± 7%, n = 19; p < 0.001). Data is pooled from L5-to-L5 pairs (triangles), L2/3-to-L5 pairs (open blue circles), boosted L2/3-to-L5 pairs (closed blue circles) and extracellular stimulation experiments (open and closed squares). Dashed red line is best fit sigmoid. Only experiments with EPSP rise time > 3 ms (cf. Figure 2) were included in (C) and (D).

Figure 5. Distal LTD is Endocannabinoid-Dependent

(A) Extracellular 50 Hz stimulation of distal inputs (rise time 3.7 ms) in the absence of somatic spiking resulted in depression (41%, p < 0.001). (Left) Recording configuration. (B) Unpaired distal EPSPs evoked LTD (green triangles, 71% ± 6%, n = 10; cf. Golding et al., 2002) that was abolished by the endocannabinoid CB1 receptor blocker AM251 (Gatley et al., 1996) (red squares, 99% ± 3%, n = 6, p < 0.01). Unpaired spike trains (blue circles, 102% ± 6%, n = 10) did not evoke plasticity.

Figure 6. Dendritic Depolarization Switches Distal Plasticity from LTD to LTP

(A) AP-EPSP pairing in conjunction with a brief subthreshold dendritic current injection (0.4 nA, 200 ms, red trace) resulted in LTP of a small, distal input (rise time 4.9 ms). (Left) Recording configuration, also showing sample LTP induction traces (stimulus electrode, green; dendrite, red, “d;” soma, blue, “s”). (B) Pairing of APs and EPSPs in the presence of a brief subthreshold dendritic current injection (247 ± 22 μm from the soma) evoked LTP of small distal inputs (<1 mV; red circles; 163% ± 7%, n = 5). The corresponding amount of subthreshold current injected into the soma, however, resulted in LTD (blue triangles; 88% ± 8%, n = 3). The subthreshold dendritic (0.35 ± 0.03 nA) and somatic (0.31 ± 0.01 nA) current injections were not significantly different (p = 0.36). These results demonstrate that dendritic, but not somatic, depolarization switches distal plasticity from LTD to LTP (p < 0.001).

Figure 7. Conditions that Evoke Distal LTP also Boost bAPs

(A) Amplitude, width, and area of bAPs were increased by a 0.4 nA subthreshold dendritic current injection (red bars, top, and red traces in middle). At the soma, the effect on the AP was moderate (blue bars, top, and blue traces at bottom). Traces with 0.4 nA subthreshold current injection are represented with darker colors than control traces (middle, “APs + depol,” and bottom). (Left) Dendritic recording (“d”) 301 μm from the soma (“s”). (B) The amount of depolarization-induced bAP boosting increased with dendritic recording distance from the soma (Larkum et al., 1999a, 2001; Stuart and Häusser, 2001). The boosting occurred irrespective of the source of the depolarization (dendritic current injection, triangles; synaptic activation, circles; n = 13 cells total). (C) (Left) L5 pyramidal cell filled with Fluo-5F and Alexa Fluor 594. (Right) Line scans at three dendritic locations reveal supralinear Ca²⁺ signals (green) when pairing APs with a subthreshold 200 ms long dendritic current injection (0.4 nA, 260 μm from soma, “d”) in distal dendrites, compared to the arithmetic sum of the APs (red) and the depolarization alone (blue). Note that boosting is widespread in distal dendrites. In this cell, the simultaneous dendritic recording measured a bAP area boosting of 175% ± 0.4%; compare (B). Scale bars, 100 μm, 20% dG/R, 50 ms. (D) APs paired with distal large EPSPs also resulted in supralinear Ca²⁺ signals (green) in distal dendrites, as compared with APs (red) and EPSPs (blue). As with plasticity experiments (Figure 4), EPSPs (3.9 mV; rise time 3.2 ms) were evoked by relatively strong extracellular stimulation near the dendritic tuft (stimulation electrode position 740 μm). Note that boosting is not limited to locations close to the stimulation electrode. Scale bars, 2% dG/R, and see (C). (E) The Ca²⁺ signal boosting due to pairing of APs with depolarization (triangles; 18 positions, four cells) or with large EPSPs (circles; 37 positions, ten cells) increased with distance from the soma, indicating that sufficient dendritic depolarization boosts bAPs throughout the distal apical dendritic tree (Larkum et al., 1999a, 2001; Stuart and Häusser, 2001).

Figure 8. Somatic Hyperpolarization Reduces Basal Dendritic AP-Evoked Ca²⁺ Signals and Abolishes Proximal LTP

(A) (Left) L5 pyramidal cell filled with Fluo-5F and Alexa Fluor 594. (Right) Line scans in basal dendritic locations 1 (21 μm from soma) and 2 (92 μm) reveal reduced Ca²⁺ signals for APs paired with hyperpolarization (red; −0.34 nA, 200 ms, −23 mV max deflection) compared with APs alone (blue). Note the difference in the effect at locations 1 and 2, as well as the absence of an effect in the axon (location 3, 147 μm). Scale bars, 50 μm, 20% dG/R, 200 ms. (B) AP-evoked Ca²⁺ signals in dendrites, but not in axons, are reduced more by hyperpolarization as the distance from the soma increases. This data is consistent with a hyperpolarization-induced reduction in the reliability of AP propagation into the basal (n = 25) as well as the apical (n = 12) and apical oblique (n = 7) dendritic arbors, but not into the axon (n = 6). Data was collected from five cells. Axonal line scans, indicated at origin for clarity, were acquired 129 ± 14 μm from soma. (C) In a connected L5-to-L5 pair, strong postsynaptic hyperpolarization during 50 Hz pairing (Induction 1; −0.41 nA, 200 ms) abolished LTP (Sjöström et al., 2001). The same connection subsequently potentiated due to a strong LTP induction protocol (Induction 2; 0.7 nA, 200 ms; 130%, p < 0.001), demonstrating that its ability to express LTP was not compromised. Scale bar, 100 μm. (D) Somatic hyperpolarization blocked LTP in L5-to-L5 pairs (red triangles; n = 6) as compared with controls without hyperpolarization (green circles; n = 34).

Figure 9. A Dendritic Switch Providing a Learning Rule for Distal Dendritic Long-Term Plasticity

(A) Proximal inputs onto an L5 pyramidal cell are relatively more prone to LTP (green) than distal dendritic inputs, which tend to undergo LTD (red). (B) Sufficient depolarization of the apical dendrite switches distal plasticity from LTD to LTP. Dendritic depolarization even relatively close to the soma can switch the mode of plasticity at the distal apical dendrite.