1. Neuroscience
Download icon

Dendritic spikes in hippocampal granule cells are necessary for long-term potentiation at the perforant path synapse

  1. Sooyun Kim  Is a corresponding author
  2. Yoonsub Kim
  3. Suk-Ho Lee
  4. Won-Kyung Ho  Is a corresponding author
  1. Seoul National University College of Medicine, Korea
Research Article
  • Cited 0
  • Views 2,364
  • Annotations
Cite this article as: eLife 2018;7:e35269 doi: 10.7554/eLife.35269

Abstract

Long-term potentiation (LTP) of synaptic responses is essential for hippocampal memory function. Perforant-path (PP) synapses on hippocampal granule cells (GCs) contribute to the formation of associative memories, which are considered the cellular correlates of memory engrams. However, the mechanisms of LTP at these synapses are not well understood. Due to sparse firing activity and the voltage attenuation in their dendrites, it remains unclear how associative LTP at distal synapses occurs. Here, we show that NMDA receptor-dependent LTP can be induced at PP-GC synapses without backpropagating action potentials (bAPs) in acute rat brain slices. Dendritic recordings reveal substantial attenuation of bAPs as well as local dendritic Na+ spike generation during PP-GC input. Inhibition of dendritic Na+ spikes impairs LTP induction at PP-GC synapse. These data suggest that dendritic spikes may constitute a key cellular mechanism for memory formation in the dentate gyrus.

https://doi.org/10.7554/eLife.35269.001

Introduction

The cortico-hippocampal circuit is implicated in the formation, storage, and retrieval of spatial and episodic memories (Lisman, 1999). The dentate gyrus (DG), the first stage in the hippocampal circuitry, receives abundant excitatory projections from the entorhinal cortex via the perforant-path (PP) synapses. Theoretical models of hippocampal function propose that the DG is critically involved in pattern separation and that synaptic transmission and plasticity at PP-granule cell (GC) synapses in the DG is required to remove redundant memory representations (Marr, 1971; McNaughton and Morris, 1987; Treves and Rolls, 1994). In agreement with theoretical predictions, knockout of N-methyl-D-aspartate (NMDA) receptors in GCs impairs long-term potentiation (LTP) and the ability to rapidly form a contextual representation and discriminate it from previous similar memories in a contextual fear conditioning task (McHugh et al., 2007). Moreover, GCs that were activated by contextual fear conditioning, referred to as memory engram cells, present clear signatures of synaptic potentiation such as a larger AMPA-NMDA ratio and a greater density of dendritic spines (Ryan et al., 2015). Thus, knowledge of plasticity at PP-GC synapses is essential for understanding the hippocampal function.

Attempts to understand synaptic plasticity rules at PP-GC synapses date to studies of Bliss and Lomo (1973), who first demonstrated LTP in the hippocampus by high-frequency stimulation of the PP fibers in vivo. More recently, theta-burst high-frequency stimulation (TBS) of the PP has been used to induce LTP at PP-GC synapses (Schmidt-Hieber et al., 2004; McHugh et al., 2007; Ge et al., 2007). However, the underlying mechanisms for the induction of LTP with TBS at these synapses remain unclear. Associative forms of synaptic plasticity depend on a presynaptic activity (e.g. excitatory postsynaptic potential, EPSP) and a postsynaptic signal (e.g. action potential, AP). Classically, a backpropagating AP (bAP) provides the associative postsynaptic signal at the synaptic site for the induction of LTP (Hebb, 1949; Magee and Johnston, 1997; Dan and Poo, 2006; Feldman, 2012). However, axosomatic APs are poorly propagated back into the dendrites of the GCs (Krueppel et al., 2011) and are unlikely to contribute to the induction of LTP. In addition, mature GCs are relatively silent during exploration (Schmidt-Hieber et al., 2014; Diamantaki et al., 2016) and fire with a low number of APs. Thus, spike-timing-dependent plasticity (STDP) that depends on an axosomatic postsynaptic AP will rarely occur under natural conditions (Feldman, 2012). The pronounced attenuation of AP backpropagation, the low occurrence of APs, along with the distinct intrinsic features of mature GCs, such as a hyperpolarized resting membrane potential and reduced excitability (Scharfman and Schwartzkroin, 1990; Mongiat et al., 2009; Pernía-Andrade and Jonas, 2014), cannot explain how distal synaptic PP inputs can be potentiated. Resolving this question has critical implications for understanding both the mechanism and function of memory encoding and retrieval (Ryan et al., 2015).

An alternative postsynaptic signal that may contribute to LTP induction at PP synapses innervating distal dendrites of GCs is a locally generated dendritic spike (Sjöström et al., 2008). Synaptic plasticity at distal synapses may occur in the absence of axosomatic APs via dendritic sodium spikes, which provide an alternative source of postsynaptic depolarization necessary for LTP induction (Golding et al., 2002; Kim et al., 2015). In GCs, linear somatic responses can be seen when glutamate is applied in the dendrites (Krueppel et al., 2011). However, without directly accessing the electrical properties of the distal GC dendrites, it is still not known whether synaptic stimulation can generate local dendritic spikes in GCs.

To address this question, we performed subcellular patch-clamp recordings on the thin dendrites of GCs. We found that inhibition of dendritic Na+ channels prevented LTP induction via TBS at PP-GC synapses. Because dendritic Na+ channels could generate local spikes that were independent of axosomatic AP generation and were not affected by voltage attenuation, we suggest that Na+ spikes in the dendrites provide the postsynaptic signal necessary for the induction of LTP at PP-GC synapses.

Results

We examined GCs with an input resistance (Rin) <200 MΩ (Rin = 108.4 ± 2.6 MΩ; resting membrane potential: –81.3 ± 0.2 mV; see ‘Materials and methods’) which corresponds to the mature GC population in the acute hippocampal slices from rats (Schmidt-Hieber et al., 2004).

TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs

We first induced long-term potentiation (LTP) in GCs by theta-burst stimulation (TBS) of the PP synapses in the outer third of the molecular layer (Figure 1A). To activate the PP synapses, the tip of a stimulation electrode was placed in close proximity to the dendrite (<50 µm) using fluorescence microscopy (‘Materials and methods’). When bursts of postsynaptic APs were evoked during TBS, TBS-induced LTP caused an increase in the amplitude of the excitatory postsynaptic potentials (EPSPs; from 6.72 ± 0.71 mV to 12.69 ± 1.44 mV, n = 6, p<0.05; Figure 1B). However, axosomatic APs in GCs are substantially attenuated with distance from the soma (Krueppel et al., 2011), and therefore are not likely to provide the necessary depolarization at distal dendrites. To test whether axosomatic APs are required for the induction of LTP at these distal synapses, we either locally applied tetrodotoxin (TTX) to the proximal axon, soma, and proximal dendrites in a subset of experiments (6 of 13 experiments) or adjusted the stimulus intensity to prevent axosomatic AP initiation and backpropagation occur during TBS. The absence of axosomatic spikes did not block LTP induction, indicating that bAPs are not critical for this form of LTP (from 6.96 ± 0.40 mV to 9.96 ± 0.86 mV, n = 13, p<0.01; Figure 1C). Intriguingly, activation of the PP synapses during TBS often produced voltage responses with a fast depolarizing phase similar to somatic events observed during dendritic spike generation in other types of neurons (Figure 1C,D; Golding and Spruston, 1998; Golding et al., 2002; Jarsky et al., 2005; Losonczy and Magee, 2006; Kim et al., 2012). Consistent with previous results (Golding and Spruston, 1998; Golding et al., 2002; Losonczy and Magee, 2006; Kim et al., 2015), various shapes of putative dendritic spikes were observed during TBS (Figure 1D). These voltage responses were first distinguished from EPSPs without dendritic spikes by the spikelet waveform (Figure 1D) and then identified as putative dendritic spikes when the peak of the temporal derivative (dV/dt) of somatic voltage responses was larger than 2.5 mV/ms. Typically, weak (2.5 mV/ms ≤ dV/dt < 10 mV/ms) and strong (dV/dt > 10 mV/ms) putative dendritic spikes were identified (weak dendritic spikes: orange, 4.2 ± 0.4 mV/ms, n = 17; strong dendritic spikes: red, 14.5 ± 0.8 mV/ms, n = 11; p<0.0001; Figure 1D). The dV/dt of the EPSPs without any putative dendritic spikes was 1.3 ± 0.01 mV/ms (n = 2498), which was significantly smaller than that of weak putative dendritic spikes (p<0.0001; Figure 1D). These putative dendritic spikes were accompanied by a sustained plateau potential (Figure 1D; see also Figure 2B). Remarkably, the presence of these weak and strong putative dendritic spikes during TBS was correlated with LTP induction (in the presence of putative dendritic spikes, 173.2 ± 12.1%, n = 7; in the absence of putative dendritic spikes 114.1 ± 12.8%, n = 6; p<0.005; Figure 1E,G). Indeed, we found a strong correlation between the number of putative dendritic spikes observed during TBS and the magnitude of LTP (r = 0.77; p<0.005; n = 13; Figure 1F). These results suggest that dendritic spikes but not axosomatic APs contribute to LTP induction at PP-GC synapses.

Putative dendritic spikes during theta-burst stimulation (TBS) induction are required for long-term potentiation (LTP) at the perforant-path (PP) to granule cell (GC) synapse.

(A) Maximum intensity projection of confocal stack fluorescence images of a GC (top) indicating the medial molecular layer (MML) and the outer molecular layer (OML). Synaptic responses of the PP were evoked by electrical stimulation in the OML. Scale bar is 50 µm. (bottom) Theta-burst LTP induction protocol. (B) Representative time course of EPSPs (top) and summary plot (bottom) before and after TBS of the PP synapses. Red line denotes average EPSP baseline value. Increment in EPSP amplitude denotes LTP. Representative traces, which correspond to the numbers in the time course plot, show the average of 30 EPSP traces before and 25–30 min after TBS (for all subsequent figures). Inset: example of the first burst of TBS responses showing initiation of multiple axosomatic APs during TBS induction upon increasing the stimulus intensity. (bottom, right) Summary bar plot of average EPSP amplitude before and after TBS indicating a significant increment in synaptic responses 25–30 min after TBS stimulation (n = 6, *p<0.05). (C) Same as (B) but no axosomatic AP initiation occurs during TBS. Inset: representative example of TBS responses. Arrow indicates a putative dendritic spike. (bottom right) Summary plot showing that a significant potentiation in EPSP amplitude was induced after TBS stimulation even in the absence of axosomatic APs (n = 13, **p<0.01). (D) (left) Representative somatic voltage traces when TBS stimulation evokes putative dendritic spikes (top) or not (bottom). The arrows indicate weak (orange) and strong (red) putative dendritic spikes and accompanying plateau potentials (blue arrow). (right) Somatic voltages (top row) and corresponding dV/dt (bottom row) of weak (orange) and strong (red) putative dendritic spikes in the burst responses (left) on an expaned time scale. Note that dV/dt values of the putative dendritic spikes increase sharply in a non-linear manner. (bottom, right) Summary bar graphs of dV/dt peak values of somatically recorded EPSPs with or without dendritic spikes (grey: EPSPs without putative Dspikes; orange: with weak putative Dspikes; red: with strong putative Dspikes). (E) Average time course of EPSPs when TBS stimulation evokes putative dendritic spikes and in the absence of putative dendritic spikes during TBS (black). LTP is induced only if putative dendritic spikes are present during TBS stimulation. (F) The number of putative dendritic spikes are significantly correlated with the magnitude of LTP. Black lines represent linear regressions (n = 13). (G) Bar summary graph and individual experiments (circles) indicating that experiments showing the occurrence of putative dendritic spikes (with putative Dspikes, red) during TBS induction induced a significant increment of LTP. Note that there are no significant differences in the magnitude of LTP when TBS stimulation evokes axosomatic APs (black) or putative dendritic spikes (red). Bars indicate mean ± SEM; circles represent data from individual cells. Lines connect data points from the same experiment. *0.01 ≤ P < 0.05. **p<0.01. ****p<0.0001. Single-cell data (top panel in B and C) and mean data (bottom panels in B and C, and E; mean ± SEM). Vertical gray bars in B, C, and E indicate the time point of the induction protocol.

https://doi.org/10.7554/eLife.35269.002
Figure 2 with 1 supplement see all
Pairing protocols did not induce LTP at the lateral PP-GC synapses.

(A) Representative time courses of EPSP amplitudes before and after pairing presynaptic stimulation of the PP synapses and postsynaptic action potentials with short time intervals (+10 ms, Pre–post sequence, black, top left inset; –10 ms, Post–pre sequence, green, top right inset). Both pairing protocols induce no significant changes in EPSP, suggesting that action potential (AP) backpropagation is not necessary for LTP induction. Right inset shows that LTP was not induced after pairing of EPSPs and APs in both pre-post (top) and post-pre sequences (bottom). (B) The average EPSP time courses of pre-post (black) and post-pre (green) induction protocols of pairings between synaptic responses and postsynaptic APs, showing that low-frequency pairing protocols failed to induce LTP. (C) Summary data indicating that experiments showing the occurrence of putative dendritic spikes during TBS protocol induced a significant increment of LTP, whereas pairings of synaptic stimulation with postsynaptic APs did not show a statistically significant LTP induction, independent of temporal order. Bars indicate mean ± SEM; circles represent data from individual cells. Lines connect data points from the same experiment. ***p<0.005. Single-cell data (A) and mean data (B); mean ± SEM). Vertical gray bars in (A) and (B) indicate the time point of the induction protocol.

https://doi.org/10.7554/eLife.35269.004

To further test whether axosomatic spikes contribute to the induction of LTP at the PP-GC synapse, we used associative pairing protocols (Figure 2). Presynaptic activity was either followed (pre-postsynaptic sequence) or preceded (post-presynaptic sequence) at a 10 ms interval by single or AP bursts (2 APs at 100 Hz). Both protocols failed to induce significant changes in EPSP amplitude (pre-postsynaptic sequence, control: 4.97 ± 0.38 mV, after induction: 5.83 ± 0.86 mV, n = 15, p=0.23; post-presynaptic sequence, control: 4.78 ± 0.54 mV, after induction: 4.91 ± 0.99 mV, n = 9, p=0.54; Figure 2A,B), suggesting that APs of axonal origin are not important for potentiation. We further performed similar experiments at the medial PP–GC synapses by stimulating axons in the middle third of the molecular layer (Figure 2—figure supplement 1A). Pairing EPSPs and APs in both pre-postsynaptic and post-presynaptic sequences did not induce LTP of EPSPs (pre-postsynaptic sequence, control: 4.86 ± 0.40 mV, after induction: 5.09 ± 0.99 mV, n = 7, p=0.9015; post-presynaptic sequence, control: 3.86 ± 0.26 mV, after induction: 4.27 ± 0.57 mV, n = 6, p=0.56; Figure 2—figure supplement 1B–E), consistent with previous studies that reported no synaptic potentiation after similar pairing protocols (Yang and Dani, 2014; Lopez-Rojas et al., 2016). Together, these results imply that dendritic spikes, rather than axosomatic APs, are the essential signal for LTP induction at distal GC synapses.

LTP by TBS at PP–GC synapses requires NMDARs and Na+ channels

We next examined the receptors involved in LTP expression at the PP-GC synapses. Bath application of the NMDAR antagonist DL-AP5 (50 µM) abolished LTP (control: 8.89 ± 0.95 mV; after induction: 9.88 ± 1.21 mV, n = 9, p=0.50; Figure 3A,E,F). While sustained plateau potentials during LTP induction were inhibited by DL-AP5 in the external solution, fast rising events remained unchanged (Figure 3B). Thus, we next tested whether dendritic voltage-gated Na+ channels are involved in LTP. To block dendritic Na+ channels without affecting synaptic transmission, we included the intracellular sodium channel blocker QX-314 (5 mM) in the whole-cell patch pipette. Stimulus intensity was set to elicit baseline EPSP of similar or slightly larger amplitude than that required to produce putative dendritic spikes during TBS (Figure 3C,D). This manipulation abolished both putative dendritic spikes (0 of 7 cells; Figure 3D) and TBS-induced LTP (control: 9.24 ± 1.21 mV, after induction: 8.77 ± 1.67 mV, n = 7, p=0.69; Figure 3C–E). These results indicate that TBS-induced LTP in GCs depends on the activation of NMDARs and voltage-gated Na+ channels on the postsynaptic dendritic membrane, reinforcing the idea that dendritic Na+ spikes may play a pivotal role in TBS induction at distal synapses.

Induction of LTP at PP–GC synapses requires activation of NMDARs and the involvement of Na+ channels.

(A) An example of the time course of EPSP amplitude when TBS protocols was applied in the presence of the NMDAR antagonist, DL-AP5 (50 µM). Red line indicates the average EPSP amplitude. Insets show that EPSP did not increase in the presence of DL-AP5 (orange) after TBS, but a robust LTP was induced when TBS was applied in standard saline (ACSF, black). (B) (left) Long-duration plateau potentials are mediated by NMDA receptor channels. Somatically recorded voltages in response to high-frequency burst sitmulation (green, 10 shocks, 100 Hz) under control (black) and in the presence of DL-AP5 (50 µM; orange). The inset shows putative dendritic spikes (indicated by the red arrow) before and after the addition of DL-AP5; Note that putative dendritic spikes are resistant to the NMDAR blockers. (right) Summary of the effects of DL-AP5 on plateau potentials. Peak amplitdue of plateau potentials were measured after the stimulus (indicated by dashed lines; Control: 36.6 ± 2.2 mV; DL-AP5: 25.6 ± 3.2 mV; n = 6, *p<0.05). (C) A representative time course of EPSP amplitude before and after TBS when the cells were dialyzed with a sodium-channel blocker, QX-314 (5 mM). Inset shows that the averaged EPSP amplitude did not change when blocking sodium channels with QX-314 (blue) despite TBS induction. In contrast, when cells were dialyzed with the standard intracellular solution, the amplitude of EPSP increased after TBS (i.e. LTP). (D) The effects of intracellular QX-314 on dendritic spike initiation in response to high-frequency PP sitmulation. (left) Representative traces of EPSPs in response to PP stimulation with (blue) and without (black, control) QX-314. Putative dendritic spikes (arrow) were observed only under control condition. (right) Bar graphs indicate the baseline EPSP amplitude (EPSPcontrol: 7.05 ± 0.54 mV, n = 7; EPSPQX-314: 9.24 ± 1.21 mV, n = 7; p=0.10) and the number of putative dendritic spikes during TBS induction (control: 4.14 ± 1.06, n = 7; QX-314: 0, n = 7; ***p<0.005) in two groups. (E) Summary plot of TBS-induced LTP experiments in the presence of DL-AP5 (orange) and with dialysis of QX-314 in the recording pipette (blue). Both treatments prevented the induction of LTP at the PP to GC synapse. (F) Summary bar graph and individual average EPSP amplitudes after TBS in control (standard saline, black), bath application of DL-AP5 (orange) and dialysis of QX-314 (blue). Treatments with DL-AP5 and QX-314 showed a significant difference compared to the standard LTP induction (DL-AP5, *p<0.05; QX-314, ***p<0.005; compared to control in Figure 1I). Representative traces in A and C correspond to the numbers (1 and 2) n the time-course plot. Bars indicate mean ± SEM; circles represent data from individual cells. Lines connect data points from the same experiment. Single-cell data (A,C) and mean data (E; mean ± SEM). Vertical gray bars in (A, C,) and (E) indicate the time point of the induction protocol.

https://doi.org/10.7554/eLife.35269.008

Backpropagation of axosomatic APs in the dendrites of GCs

To directly dissect the dendritic mechanism that determines the induction rules of synaptic plasticity in GCs, we employed subcellular patch-clamp techniques to analyze AP backpropagation and initiation in GC dendrites (Figure 4A). First, we characterized backpropagation in GCs with somatic current injection evoking trains of APs at the soma while simultaneously recording in the dendrite (Figure 4B). The AP always appeared first in the somatic recording and then in the dendrites (Figure 4C). Similar to the previous report (Krueppel et al., 2011), the peak amplitude of the bAPs attenuated as a function of distance from the soma (Figure 4D,E). At dendritic distances beyond 150 µm from the soma, the peak amplitude of bAPs was reduced to ~36% of the amplitude of somatic APs (soma: 93.6 ± 1.7 mV; dendrite: 33.7 ± 3.1 mV; n = 10, p<0.005; Figure 4D). The attenuation per dendritic length is much more pronounced in GCs compared to the other neocortical (Stuart et al., 1997; Nevian et al., 2007) and hippocampal pyramidal neurons (Spruston et al., 1995; Kim et al., 2012). It is interesting to note that the extent of bAP attenuation when normalized to the overall length of GC dendrites (mean dendritic length: 278 ± 7.4 μm; n = 11) is quite similar to layer five pyramidal neuron dendrites (cf. Supplementary Figure 3B in Nevian et al. (2007); see also Krueppel et al. (2011); Figure 4F), suggesting that the failure of LTP induction during pairing protocols at distal GC synapses (Figure 2) might be explained by the voltage attenuation of bAPs, as was reported in layer five pyramidal neurons (Letzkus et al., 2006; Sjöström et al., 2008); Feldman; 2012). We also estimated the conduction velocity of bAPs by analyzing the AP latencies measured at the half-maximal amplitude of the rising phase. APs were initiated axonally and propagated back into the dendrites with a velocity of 226 µm/ms (Figure 4G; n = 56; corresponding to 0.2–0.3 m/s, Senzai and Buzsáki, 2017). This conduction velocity is slower than those measured in the apical and basal dendrites of layer five pyramidal neurons (apical dendrites, 508 µm/ms; basal dendrites, 341 µm/ms; Nevian et al. (2007) and Stuart et al., 1997) and is similar or lower than the velocity estimated in the apical dendrite of other hippocampal principal neurons (Spruston et al., 1995; Kim et al., 2012). In summary, these experiments show that bAPs in GCs propagate into the dendrites with substantial voltage attenuation and moderate conduction velocity.

Properties of backpropagating APs in the dendrites of GCs.

(A) Morphological reconstruction of a GC with representative double somatic and dendritic whole-cell recording configuration used to analyze the AP backpropagation. Scale bar is 50 µm. (B) A train of APs elicited by a 1 s current pulse applied at the soma. Black traces indicate somatic voltage and corresponding current; red traces indicate dendritic voltage and corresponding current. (C) First AP in the train displayed at expanded time scale. Voltage traces (soma in black, dendrite in red) indicate that the AP is initiated first near the soma and propagated back into the dendrites with a lower amplitude. (D) Summary graph to compare somatic (black) and dendritic (red) AP peak amplitude. Bars indicate mean ± SEM; circles represent data from individual cells. Lines connect data points from the same experiment. ***p<0.005. (E) Scatter plot of peak amplitude of the backpropagating AP against the absolute physical distance of the recording site from the soma (56 somatodendritic recordings). The blue curve represents a mono-exponential fit to the data points between 0 and 212 µm. (F) Scatter plot of the bAP amplitude normalized to the corresponding axosomatic AP amplitude plotted against the distance of the recording site scaled to the total dendritic length (278 ± 7.4 μm; n = 11). The blue curve is a mono-exponential fit to the data. (G) Scatter plot of AP latency as a function of the distance from the soma (56 somatodendritic recordings) together with a linear regression (blue line) to compute the average conduction velocity of the AP into the dendrites; dendritic AP propagation velocity was 226 µm/ms. Single-cell data (E–G, red) and mean data (E), black; mean ± SEM).

https://doi.org/10.7554/eLife.35269.010

Ionic mechanisms of AP backpropagation

To determine the ionic mechanisms underlying the strong voltage attenuation and the moderate conduction velocity of bAPs, we assayed the somatic and dendritic distribution of voltage-gated Na+ and K+ currents in outside-out patches isolated at various locations using pipettes of similar open tip resistance (soma: 17.9 ± 0.5 MΩ, n = 24; dendrite: 19.4 ± 0.5 MΩ, n = 36; Figure 5). Depolarizing voltage pulses from –120 mV to 0 mV evoked TTX-sensitive inward Na+ currents in the majority of outside-out patches excised from both soma and dendrites (Figure 5A and Figure 5—figure supplements 1 and 2). Pooled data demonstrated that the current amplitude of Na+ channels is uniformly distributed over the dendritic membrane (Figure 5D). On average, the peak amplitude of Na+ currents was not significantly different between somatic and dendritic patches (soma: –6.81 ± 1.36 pA, n = 19; proximal dendrite (within 100 µm, PD): –5.02 ± 1.55 pA, n = 8; distal dendrite (beyond 100 µm, DD): –7.17 ± 1.36 pA, n = 21; p>0.99 for all cases, Kruskal-Wallis test with Dunn’s multiple comparison; Figure 5D). Surprisingly, very large transient K+ currents that appeared to activate and inactivate rapidly in response to voltage pulses from –120 mV to +70 mV was found in patches obtained from the dendrites (Figure 5B). These inactivating components of K+ channel activity were reduced by the A-type K+ channel blocker 4-aminopyridine (4-AP, Figure 5B and Figure 5—figure supplements 1 and 2). Plotting the current amplitude along the dendrite demonstrated that GC dendrites show significantly larger A-type K+ currents than the soma (soma: 30.4 ± 4.2 pA, n = 21; PD: 61.5 ± 8.7 pA, n = 7; DD: 102.3 ± 15.3 pA, n = 21; soma vs. PD, p=0.07; soma vs. DD, p<0.0001; PD vs. DD, p=0.96, Kruskal-Wallis test with Dunn’s multiple comparisons; Figure 5E). Consistent with this finding, we found that bath application of 5 mM 4-AP in simultaneous somatic and dendritic voltage recordings results in a significant enhancement of the peak amplitude and duration of bAPs in the dendrites, indicating that dendritic A-type K+ current (IA) contributes to voltage attenuation of bAPs in GCs (Figure 5—figure supplement 3). Finally, the delayed rectifier K+ current components, which was largely inhibited by 20 mM extracellular tetraethylammonium (TEA, Figure 5C and Figure 5—figure supplement 1), showed a low and uniform expression over the entire somatodendritic axis (soma: 19.6 ± 2.2 pA, n = 21; PD: 18.3 ± 3.8 pA, n = 7; DD: 21.7 ± 3.8 pA, n = 22; p>0.99 for all cases, Kruskal-Wallis test with Dunn’s multiple comparisons; Figure 5C,F). Taken together, these voltage-clamp data reveal that a markedly high density of K+ channels and a uniform density of Na+ channels are present in the dendrites of GCs, which are likely to account for both the strong dendritic AP attenuation and the moderate AP propagation velocity.

Figure 5 with 3 supplements see all
Differential Na+ and K+ channel densities in the dendrites of GCs.

(A) Averages of Na+ current recorded from outside-out patches from soma (black, averages of 25–27 sweeps) and dendrite (red, 110 μm, averages of 20 sweeps) in response to a test pulse potential to 0 mV (bottom). Na+ currents were recorded in the presence of 4-AP (5 mM) and TEA (20 mM). Left, soma; right, dendrite; Top, control; bottom, currents in the presence of 0.5 μM TTX in the bath. Leak and capacitive currents were subtracted by a ‘P over –4’ correction procedure. Note that the remaining outward current is the resistant K+ current component to 5 mM 4-AP (Figure 5—figure supplements 1 and 2; Hoffman et al., 1997). Blockade of outward K+ currents by extracellular 4-AP (5 mM) only had a negligible effects on the peak amplitude of Na+ currents (Figure 5—figure supplement 2). (B) Averages of A-type K+ current evoked in outside-out patches excised from soma (black, averages of 6–8 sweeps) and dendrite (red, 110 μm averages of 15–19 sweeps) in response to a test pulse potential to +70 mV (top). Transient A-current was measured by subtraction of traces with a −40 mV prepulse from those with a −120 mV prepulse. Left column, soma; Right column, dendrite; Top row, control; Bottom row, currents in the presence of 5 mM 4-AP in the bath. (C) Averages of delayed rectifier K+ current evoked in outside-out patches excised from soma (black, averages of 6–8 sweeps) and dendrite (red, 124 μm, averages of 10–18 sweeps) in response to a test pulse potential to +70 mV (top). Delayed rectifier K+ current was measured by a −40 mV prepulse. Left column, soma; Right column, dendrite; Top row, control; Bottom row, currents in the presence of 20 mM TEA in the bath. See also Figure 5—figure supplement 1. (D, E, F) (top) Plot of amplitude of Na+ channel activity (D), A-type K+ channel activity (E), and delayed rectifier K+ channel activity (F) as a function of distance from the soma, demonstrating that various channels are differentially expressed across the length of GC dendrites. Data from 19, 21, and 21 somatic (black circles) and 29, 28, and 29 dendritic patches (red circles). Blue circles represent the average of somatic recordings. (bottom) Summary bar graph showing the peak amplitude of Na+ (D), A-type K+ (E) and delayed rectifier K+ (F) channel activity in the soma, proximal dendrite (<100 μm) and distal dendrite (≥100 μm). Bars indicate mean ± SEM; circles represent data from individual experiments. n.s., not significant; ****p<0.0001 by Kruskal Wallis test with post hoc multiple comparison using Dunn’s test.

https://doi.org/10.7554/eLife.35269.012

A burst of PP stimuli can evoke dendritically initiated Na+ spikes in GCs

Voltage-gated Na+ channels may contribute to the generation of dendritically initiated local spikes in regions where the dendrites are very small (Holmes, 1989). Indeed, depolarizing dendrites with brief current pulse injection evoked local spikes in distal dendrites of GCs (55 of 63 recordings; Figure 6A–C). In the proximal domain (within 100 µm from the soma), dendritic spikes were not detected and depolarizing dendrites resulted in an axosomatic AP that propagated back to the dendritic recording site (Figure 6D). At dendritic recording locations 70 to 150 µm from the soma, dendritic spikes were associated with axosomatic spikes. For distances larger than 150 µm from the soma (approximately corresponding to the outer molecular layer), current injections robustly initiated isolated dendritic spikes (Figure 6D). Pharmacological analysis revealed that dendritic spikes were resistant to 200 μM CdCl2 and 50 μM NiCl2 but were inhibited by 0.5 μM TTX, indicating that they were mediated by dendritic voltage-gated Na+ channels rather than by Ca2+ channels (TTX: 11.3 ± 3.3%, n = 3; CdCl2: 93.9 ± 7.8%, n = 4; NiCl2: 97.9 ± 10.2%, n = 4; Figure 6—figure supplement 1). Because our results indicate that GC dendrites contain a high density of transient, A-type K+ channels (Figure 5B,E), we further explored how these channels influence dendritic spike initiation. We combined dual somatodendritic recordings and focal application of 4-AP (10 mM) to the dendritic patch (Figure 6—figure supplement 2A). Local block of IA significantly decreased the current threshold for dendritic spike initiation, suggesting that dendritic IA affect the generation of dendritic spikes (Figure 6—figure supplement 2B,C).

Figure 6 with 2 supplements see all
Initiation of dendritic spikes in GCs.

(A) Schematic recording configuration of a simultaneous somatic (black) and dendritic (red) patch-clamp recording on a GC. (B) Local dendritic spikes (arrow) in GCs evoked by dendritic current injection pulses of increasing amplitude (black, voltage in the soma; red, voltage in the dendrite). (C) The relationship between voltage amplitude (black, soma; red, dendrite) and dendritic injection resembles a step-function, suggesting the all-or-none nature of the dendritic spike (arrow). Peak values of dendritic spikes were measured after subtraction of scaled subthreshold dendritic responses. (D) A summary plot showing whether a spike was evoked first in the dendrite (right, red) or in the soma (left, black) for an increasing current pulse injection at the dendrite. Red squares indicate the cells showing a dendritic spike followed by an axosomatic spike. Red triangles show the cells showing isolated dendritic spike. Black circles indicate the cells showing only an axosomatic spike. The green shaded area approximately corresponds to the outer molecular layer. The black circle and red lines in the plot indicate the soma and the dendrite, respectively.

https://doi.org/10.7554/eLife.35269.020

We next examined whether Na+ spikes in GC dendrites can be evoked by synaptic stimulation of PP inputs. Extracellular synaptic stimulation was combined with simultaneous double patch-clamp recordings from the soma and dendrites (Figure 7A). Dendritic spikes could be triggered by stimulating the PP inputs with TBS protocol (Figure 7A,B). Repeated trials of a high-frequency PP stimulation using the same stimulus intensity produced variable dendritic responses with subthreshold EPSP, weak dendritic spikes, or strong dendritic spikes that appeared either in isolation or associated with axosomatic APs (Figure 7B). When dendritic spikes were present, the dV/dt of the corresponding somatic voltages showed variable peak amplitudes that were similar to those of the putative dendritic spikes during TBS induction shown in Figure 1D (EPSPs with a dendritic spike: dV/dt = 7.1 ± 1.1 mV/ms, n = 26; p=0.58 compared to EPSPs with putative dendritic spikes: dV/dt = 8.2 ± 1.0 mV/ms, n = 28 in Figure 1D; Figure 7C,D). Overall, dendritic recordings reveals that a high-frequency burst activation of PP synapses can produce dendritic Na+ spikes that are manifested as rapidly rising voltage responses at the soma (Figure 1D).

Relationship between the dV/dt of somatically and dendritically recorded voltages during dendritic spikes generation.

(A) (top) Schematic recording configuration of a triple pipette consisting of electrical stimulation of the PP synapses in the OML, dendritic patch-clamp recording (120 µm from the soma, red) and somatic whole-cell recording (black). Scale bar is 50 µm. (bottom) Dendritic spikes can be identified as larger spikes in the dendrite (red arrow) with the corresponding small spike at the soma (black). (B) (left) Somatic and dendritic voltages in response to a high-frequency PP stimulation with a constant stimulus intensity are shown superimposed for comparison. The recording site on the dendrite is 147 μm from the soma. (right) Traces of somatic (black) and dendritic (red) voltage responses (top row) on the left (red box) and corresponding dV/dt (bottom). dV/dt peak amplitudes of each traces were summarized in the table (bottom). The red arrows indicate the dendritic spikes. When dendritic spikes were present, the corresponding somatic voltage changes were used for analysis of dV/dt peaks. For comparison, somatic and dendritic membrane voltages and corresponding dV/dt traces during axosomatic AP generation are also shown on the right (dashed box). Encircled numbers indicate correspondence between traces in B) and data points in C). (C) (left) Summary plot of peaks in somatic dV/dt against corresponding peaks in dendritic dV/dt for four simultaneous somatodendritic recordings in response to high-frequency burst stimulation of the PP synapses (dendritic recording sites are from 136 µm to 175 µm from the soma). (right) An enlarged view of the box (orange) in the left panel. Black circles, in the absence of dendritic spikes; red circles, in the presence of the dendritic spikes; gray circles, data points from the first EPSPs (EPSP1). Note that somatic dV/dt peak amplitudes of subthreshold EPSP1s were comparable to those of the somatic traces when dendritic spikes were present. (D) Summary bar graphs of dV/dt peak amplitudes of somatically recorded dendritic spikes in C) (red) and in Figure 1D. Bars indicate mean ± SEM; circles represent data from individual cells.

https://doi.org/10.7554/eLife.35269.025

Dendritically initiated Na+ spikes are required for TBS-induced LTP at PP–GC synapses

Finally, we examined whether dendritic Na+ spikes are necessary for TBS-induced LTP at PP synapses. To this end, we applied a low concentration of TTX (10 nM; Kim et al., 2015) to block dendritic Na+ spikes without interfering synaptic transmission (Figure 8A,B). Inhibition of dendritic Na+ spikes prevented TBS-induced LTP (control: 6.58 ± 0.51 mV; after induction: 7.04 ± 0.88 mV, n = 8; p=0.64; Figure 8C–E), without affecting the baseline EPSP amplitude (control: 5.21 ± 0.78 mV; after TTX: 5.06 ± 0.68 mV, n = 11; p=0.83; Figure 8—figure supplement 1). All together, these results suggest that dendritically generated local Na+ spikes in response to TBS of the PP are necessary for LTP induction in GCs.

Figure 8 with 1 supplement see all
Blockade of Na+ channels prevent dendritic spikes and LTP induction by TBS.

(A) Schematic diagram showing the experimental configuration. In these experiments, the sodium channel blocker TTX was applied by bath perfusion. Scale bar is 50 µm. (B) Bath application of TTX (10 nM) prevents the generation of dendritic spikes evoked by synaptic stimulation (black), as seen on the somatic whole-recording trace (blue). Note that decreasing availability of Na+ channels by 10 nM TTX had a negligible effect on the EPSP1 (see also Figure 8—figure supplement 1). (C) Time course of EPSP amplitudes in the presence of TTX (10 nM). The induction of LTP by TBS can be prevented by bath application of TTX, since the average EPSP amplitude does not change. (D) Average EPSP before and after TBS shows no changes in amplitude in the presence of TTX. (E) Summary bar graph and single experiments (circles) indicating that the TTX application does not produce an increment in the EPSP amplitude and therefore no statistically significant LTP induction. Bars indicate mean ± SEM; circles represent data from individual cells. Data points from the same experiment are connected by lines. Single-cell data (C) and mean data (D); mean ± SEM). Vertical gray bars in C) and D) indicate the time point of the induction protocol.

https://doi.org/10.7554/eLife.35269.027

Discussion

In summary, the present study demonstrates several major findings. First, our results show that conventional STDP protocols (Dan and Poo, 2006; Feldman, 2012) do not trigger synaptic potentiation at PP-GC synapses. Second, a physiologically relevant TBS paradigm can efficiently induce LTP in GCs without axosomatic APs. Finally, direct dendritic recordings revealed that LTP induction requires dendritic spikes. To our knowledge, these studies show the first direct evidence for dendritic spike generation in GCs and the role of these spikes during induction of LTP in the dentate gyrus network. Whether our findings also hold for GCs at early stages of development remains to be determined.

Considerable evidence has shown that the electrical properties of neurons at the cellular and subcellular levels are brain region-specific and cell-type-specific and endow unique rules and characteristics on circuit function (Sjöström et al., 2008; Stuart and Spruston, 2015). Hyperpolarized resting potential, strong dendritic voltage attenuation, and low occurrence of APs are the known electrophysiological characteristics of mature GCs (Scharfman and Schwartzkroin, 1990; Schmidt-Hieber et al., 2004; Mongiat et al., 2009; Krueppel et al., 2011; Pernía-Andrade and Jonas, 2014), which are distinct from other types of hippocampal principal neurons (Spruston et al., 1995; Kim et al., 2012; Stuart and Spruston, 2015). While backpropagated APs are an important associative signal for triggering plasticity at the synaptic site via the voltage-dependent relief of Mg2+ block of NMDARs (Magee and Johnston, 1997; Dan and Poo, 2006; Feldman, 2012; Mishra et al., 2016), the above features of mature GCs are highly unfavorable for LTP induction at distal synaptic contacts. Accordingly, we found that GCs do not show LTP at PP-GC synapses during standard STDP (Figure 2 and Figure 2—figure supplement 1), which is consistent with recent reports (Yang and Dani, 2014; Lopez-Rojas et al., 2016).

However, several studies have demonstrated EPSP-AP pairing protocol-induced synaptic potentiation at the same synapses (Levy and Steward, 1983; Lin et al., 2006). Given that Levy and Steward (1983) and Lin et al. (2006) employed in vivo and in vitro field recording configurations, respectively, the discrepant results could be attributed to GCs at different stages of maturation as immature GCs exhibit a lower threshold for LTP induction (Schmidt-Hieber et al., 2004; Ge et al., 2007) or under different recording circumstances that were exposed to various neuromodulators. For example, Yang and Dani (2014) reported that the pairing protocol that showed no synaptic potentiation could induce reliable LTP at PP-GC synapses after D1-type dopamine receptor activation by which dendritic A-type K+ currents are suppressed. Because IA is known to limit the backpropagation of APs (Hoffman et al., 1997), suppression of IA can boost AP backpropagation, allowing sufficient dendritic depolarization for LTP induction at distal synapses. Our observations of a high density of dendritic IA and their effects on AP backpropagation in GCs directly support the finding of Yang and Dani (2014). Therefore, under physiological conditions, long-lasting dendritic depolarization during theta oscillation (Buzsáki, 2002) or activation of neuromodulatory systems (Hamilton et al., 2010; Yang and Dani, 2014) may cause inactivation of dendritic IA and trigger pairing-induced LTP (Lin et al., 2006; Brunner and Szabadics, 2016).

Although the presence of A-type K+ channels in GC dendrites had been shown in earlier immunocytochemical studies (Birnbaum et al., 2004; Monaghan et al., 2008; Menegola et al., 2008), it has been proposed that dendritic A-type K+ channels have no significant impact on bAP-induced Ca2+ signals (Krueppel et al., 2011). Krueppel et al. (2011) results appear to be inconsistent with our present data, which show the strong expression of functional A-type K+ channels in GC dendrites. In contrast to that study, we directly tested the effect of 4-AP in simultaneous soma-dendrite recordings by using both global and local application methods. Our results show that the effect of local puff application of 4-AP to the dendritic recording site is much smaller than that of global application. As Krueppel et al. (2011) used local application halfway between the soma and the Ca2+ imaging site, the impact of 4-AP on bAPs in their study could be underestimated.

We further found that high-frequency stimulation of PP synapses is efficient in eliciting dendritic spikes required for LTP induction. As reported in a recent in vivo study, mature GCs are exposed to abundant functional glutamatergic inputs from the entorhinal cortex during theta rhythm (Pernía-Andrade and Jonas, 2014; Schmidt-Hieber et al., 2014). Moreover, several lines of evidence demonstrated that mature GCs are highly innervated by PP synaptic connections while receiving a powerful perisomatic inhibition (Dieni et al., 2013; 2016; Temprana et al., 2015). Under these in vivo network conditions, strong dendritic excitation by high-frequency PP inputs and strong perisomatic inhibition may promote amplification of dendritic responses without axosomatic APs. Therefore, dendritic spikes are likely physiologically relevant signals for the induction of LTP. Support for the idea of cooperative LTP at PP-GC synapses has also come from McNaughton et al. (1978), who demonstrated that high-frequency stimulation of PP synapses could induce synaptic enhancement in the absence of GC discharges.

This specific high-frequency stimulation-dependent synaptic potentiation in GCs presumably stems from distinct functional and geometrical features of GC dendrites. A moderate density of dendritic Na+ channels together with higher input impedance of the distal dendrites (Hama et al., 1989; Schmidt-Hieber et al., 2007; Holmes, 1989) suffices for initiating spikes locally in distal dendrites that have small capacitive load but not for supporting active backpropagation of axosomatic APs. Although GCs have relatively short dendrites, a high K+ to Na+ current ratio in these thin-caliber dendrites imposes a strong distance-dependent attenuation of axosomatic APs, leading to a lack of pairing-induced LTP at distal synapses. Thus, Na+ spikes in the dendrites contribute the postsynaptic depolarization necessary for the induction of associative plasticity (Golding et al., 2002; Kim et al., 2015). However, it should be noted that dendritic Na+ spikes were accompanied by NMDAR-mediated plateau potentials and therefore these two dendritic events could act in concert to trigger LTP in GCs (Schiller et al., 2000).

Consequently, our findings suggest that in the absence of axonal firing (Alme et al., 2010; Diamantaki et al., 2016), dendritic spikes would allow a silent GC to participate in the storage of memories via LTP induction. It would further permit a functional separation between a storage phase mediated by Na+ channels in GC dendrites, and a recall phase (O'Neill et al., 2008) that effectively activates the CA3 neurons (Vyleta et al., 2016) via Na+ channels in GC axons.

Materials and methods

Slice preparation and electrophysiology

Request a detailed protocol

Acute hippocampal slices (thickness, 350 μm) were prepared from the brains of 17- to 25-day-old Sprague-Dawley rats of either sex. Rats were anesthetized (isofluorane, Forane; Abbott) and decapitated rapidly. All the experiments were approved by the University Committee Animal Resource in Seoul National University (Approval #: SNU-090115–7). All brains were sliced coronally, and dorsal slices displaying all the subregions of the hippocampal formation were used for the experiments (Coronal sections located between 4.3 mm and 5.7 mm from the posterior end of the right hemisphere). Slices were prepared in an oxygenated ice-cold sucrose-containing physiological saline using a vibratome (VT1200, Leica), incubated at ~36°C for 30 min, and subsequently maintained in the same solution at room temperature until the recordings. Recordings were performed at near-physiological temperature (33–35°C) in an oxygenated artificial cerebral spinal fluid (ACSF).

Patch pipettes were made from borosilicate glass capillaries (outer diameter = 1.5 mm, inner diameter = 1.05 mm) with a horizontal pipette puller (P-97, Sutter Instruments). The open-tip resistance of patch pipettes was 2.5–6.5 MΩ and 11–30 MΩ for somatic and dendritic recordings, respectively. Current-clamp recordings were performed with an EPC-10 USB Double amplifier (HEKA Elektronik). In current-clamp recordings, series resistance was 8–80 MΩ. Pulse protocols were generated, and signals were low-pass filtered at 3 or 10 kHz (Bessel) and digitized (sampling rate: 20–50 kHz) and stored using Patchmaster software running on a PC under Window 10. Resting membrane potential (RMP) was noted immediately after rupture of the patch membrane. Rin was determined by applying Ohm’s law to the steady-state voltage difference resulting from a current step (±50 pA). Pipette capacitance and series resistance compensation (bridge balance) were used throughout current-clamp recordings. Bridge balance was checked continuously and corrected as required. Experiments were discarded if the resting membrane potential depolarized above –70 mV and were stopped if the resting membrane potential or Rin changed by more than 20% during the recording.

All experiments were performed on visually identified mature GCs on the basis of the relatively large and round-shaped somata, and the location of the cell body under DIC optics. GCs located at the superficial side of the GC layer in the suprapyramidal blade were preferentially targeted. These cells had the average RMP of –81.3 ± 0.2 mV and Rin of 108.4 ± 2.6 MΩ (n = 165), that is similar to characteristic intrinsic properties of mature GC population (Schmidt-Hieber et al., 2004). Cells were filled with a fluorescent dye, Alexa Fluor 488 (50 µM, Invitrogen) and imaged with an epifluorescence system mounted on an upright microscope equipped with a 60 x (1.1 N.A.) water immersion objective lens. Focal electrical stimulation (100 µs pulses of 1–35 V) was applied on isolated dendrites in the outer third of the molecular layer (within 100 µm of the hippocampal fissure) by placing a glass microelectrode (0.5–3 MΩ) containing 1 M NaCl or ACSF in the vicinity of the selected dendrite (typically at <50 µm distance), guided by the fluorescent image of the dendrite. All experiments were performed in the presence of the GABA receptor antagonist picrotoxin (PTX, 100 μM) and CGP52432 (1 μM).

To record voltage-gated Na+ or K+ currents, outside-out patches were excised from the soma and the dendrite with pipettes of similar geometry and open-tip resistances (18.8 ± 0.4 MΩ, n = 60, ranging from 12.5 to 24.4 MΩ) for comparison of channel density between soma and dendrite (Kim et al., 2012). Ensemble K+ currents were evoked by a pulse protocol consisting of a 50–200 ms prepulse to –120 mV followed by a 200 ms test pulse to 70 mV. Na+ currents were generated by a pulse sequence comprised of a 100 ms prepulse to –120 mV and a 30 ms test pulse to 0 mV. In all cases, the holding potential was –90 mV before and after the pulse protocol. Voltage protocols were applied to outside-out patches once every 3 and 5 s for Na+ and K+ current recordings, respectively. Leak and capacitive currents were subtracted online using the pipette capacitance compensation circuit of the amplifier and a ‘P over –4’ correction procedure.

Subcellular dendritic patch-clamp recording

Request a detailed protocol

Dendritic recordings from GCs were obtained similarly as described previously (Kim et al., 2012). First, Alexa Fluor 488 (50 µM, Invitrogen) diluted in an internal solution was loaded into cells via a somatic recording pipette. Second, after ~10 min of loading, fluorescently labeled dendrites were traced from the soma into the molecular layer using epifluorescence microscope. Finally, fluorescent and infrared differential interference contrast (IR-DIC) images were compared, and GC dendrites were patched under IR-DIC.

Stimulation protocols for the induction of long-term potentiation (LTP)

Request a detailed protocol

LTP was induced by either theta burst stimulation (TBS; Schmidt-Hieber et al., 2004) or pairing protocols. TBS induction protocol consisted of burst of EPSPs (10 stimuli at 100 Hz) repeated 10 times at 5 Hz. These episodes were repeated four times every 10 s. The pairing protocol comprised 300 repetitions of a presynaptic stimulation and one or two postsynaptic APs at the different time intervals at 1 Hz. A postsynaptic AP was evoked by a brief current injection to the soma (2 ms, 3 nA). For LTP experiments, baseline EPSPs evoked by stimulating presynaptic axon fibers at 0.1 Hz were measured for ~10 min after whole-cell recording. After LTP induction protocol, EPSPs were recorded for 30 min. LTP magnitude was evaluated as the percentage of EPSP baseline (5 min) after 25 to 30 min after the induction protocol. In a subset of recordings, local application of TTX (1 µM) to the perisomatic area was used to prevent axosomatic AP initiation and backpropagation during TBS induction, without affecting evoked EPSPs.

Solutions and chemicals

Request a detailed protocol

The extracellular solution for dissection and storage of brain slices was sucrose-based solution (87 mM NaCl, 25 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 7 mM MgCl2, 0.5 mM CaCl2, 10 or 25 mM glucose, and 75 sucrose). Physiological saline for experiments was standard ACSF (125 mM NaCl, 25 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 1 mM MgCl2, 2 mM CaCl2, and 25 mM glucose). TTX and 4-AP were applied either via bath perfusion (0.5 µM and 10 mM, respectively) or by local application (1 µM and 10 mM, respectively) with a pressure application system (Picospritzer 3, General Valve). Pressure pulses had durations of 0.2 s and amplitudes of ~10 psi. CdCl2 and NiCl2 were applied in the bath at a concentration of 200 µM and 50 µM, respectively.

For whole-cell recording and K+ current recording in outside out patches, we used K+ rich intracellular solution that contained 115 mM K-gluconate, 20 mM KCl, 10 mM HEPES, 0.1 mM EGTA, 2 or 4 mM MgATP, 10 mM Na2-phosphocreatine, and 0.3 mM NaGTP, pH adjusted to 7.2–3 with KOH (~300 mOsm). In a subset of experiments, 50 µM Alexa 488 and 0.1–0.2% biocytin (wt/vol) were added to the internal solution for labeling during the experiment or after fixation, respectively. All drugs were dissolved in physiological saline immediately before the experiment and perfused on slices at a rate 4–5 ml min−1. These included: DL-AP5, PTX, CGP52432, TTX, 4-AP, and TEA. Intracellularly delivered QX-314 (N-(2,6-dimethylphenylcarbamoylmethyl) triethylammonium chloride) were directly added to the pipette solution before the experiment was started. TTX and DL-AP5 (D,L-2-amino-5-phoshonovaleric acid; 50 µM) were purchased from Tocris Bioscience; CGP52432 were from Abcam; All other drugs were from Sigma-Aldrich.

Immunohistochemistry

Request a detailed protocol

Post-hoc morphological analysis of GCs was performed as described previously with slight modifications (Kim et al., 2012). Slices were fixed overnight at 4°C in 4% paraformaldehyde in 100 mM phosphate buffer solution (PBS), pH7.4. After fixation, slices were rinsed several times with PBS and permeabilized with 0.3% Triton X-100 in PBS. Subsequently, slices were treated with 0.3% Triton X-100% and 0.5% BSA in PBS to prevent any unwanted. Next, slices were treated with 0.3% Triton X-100% and 0.2% streptavidin-cy3 in PBS and were again incubated overnight in 4°C. Finally, slices were mounted with DAKO S3023 medium.

Data analysis

Request a detailed protocol

Custom-made routines written in Igor Pro 6.3 (Wavemetrics), Stimfit 0.15 (Guzman et al., 2014) or Prism (Graphpad) were used for data analysis and statistical testing. Spike threshold value was determined as the time point at which the derivative of voltage exceeded 40 V/s at the soma or 20 V/s at the dendrite. To determine the conduction velocity, latency–distance data were fit with a linear regression, and the velocity was calculated from the linear regression slope.

All values indicate mean ± standard error of the mean (SEM), with ‘n’ denoting the number of experiments. To test statistical significance, a two-sided nonparametric Wilcoxon signed-rank test or Wilcoxon rank sum test were used (unless noted otherwise). For comparisons of more than two groups, a Kruskal-Wallis test with Dunn’s multiple comparisons for post-hoc testing was used. Differences with p-value less than 0.05 were indicated with an asterisk and considered significant. Membrane potential values were displayed without correction for liquid junction potentials. Distances were measured from the soma to the dendritic recording site along the dendritic trajectory (Kim, 2014).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
    The Organization of Behavior
    1. DO Hebb
    (1949)
    John Wiley & Sons.
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
    Simple memory: a theory for archicortex
    1. D Marr
    (1971)
    Philosophical Transactions of the Royal Society B: Biological Sciences 262:23–81.
    https://doi.org/10.1098/rstb.1971.0078
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
    Synaptic mechanisms of sparse activity in hippocampal granule cells during mouse navigation
    1. C Schmidt-Hieber
    2. H Wei
    3. M Häusser
    (2014)
    Society for Neuroscience Abstract.
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59

Decision letter

  1. John Huguenard
    Reviewing Editor; Stanford University School of Medicine, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Dendritic spikes in hippocampal granule cells are necessary for long-term potentiation at the perforant path synapse" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Gary Westbrook as the Senior Editor. The following individual involved in review of your submission has agreed to reveal his identity: Maarten Kole (Reviewer #1). The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This report documents Nav-dependent electrogenesis in dendrites of dentate gyrus granule cells and shows how this electrogenesis is critical for perforant path long-term potentiation. The experiments are technically demanding and the overall work is of high quality, and should be published. However, the reviewers have several concerns that need to be addressed. Notably, there was a general concern that the paper is over-hyped, and there is really not a need for this. The results are clean and should be simply presented to document the findings.

Essential revisions:

1) There is a lack of clarification and experimentation why the observed "high" Ia-type current density blocks excitation in the axial direction but still enables local dendritic Na+ spikes. The ion channel distribution profiles alone (Figure 5) are insufficient to explain this since direct experiments addressing their role in spike-timing dependent failures is lacking. It is neither discussed nor addressed experimentally how the specific granule cell cable structure contributes. The interplay between ionic currents and the passive cable properties are the true backbone upon which depolarizations either spreads or attenuates. One prediction is that local block of IA (e.g. with 4-AP) would cause back-propagation facilitation without changing local dendritic spikes. Dual whole-cell voltage recording during synaptic stimulation and local somatic 4-AP application would be one way to address this issue.

2) The pharmacology experiments in Figure 6—figure supplement 1 are flawed (subsection “Dendritically initiated Na+ spikes are required for TBS-induced LTP at PP–GC synapses”, first paragraph). Dendritic spikes are expected to be mediated by nickel-sensitive T-type Ca2+ channels (at 50 – 100 µM, see e.g. Schmidt-Hieber, Jonas and Bischofberger,et al. 2004) and insensitive to cadmium (a high voltage-activated calcium channel blocker). Experiments with nickel (or TTA-P2) would more firmly demonstrate what the precise pharmacology of the observed dendritic spike is; sodium, T-type calcium channels or both. In fact, it seems cadmium did block a late component in the depolarization in the dendritic recording? Please quantify and clarify.

3) The role of dendritic spikes is introduced with a lot of confidence in the Abstract as a 'key mechanism for memory formation'. However, it remains unexplained why the authors see it in some but not all DG cells. The age range of 17-24 days is not 'mature' (subsection “Slice preparation and electrophysiology”, third paragraph). Only in the aforementioned subsection it is mentioned that morphology may be different for the DG cells showing dendritic spikes but quantification and details are lacking. Were dendritic spikes seen in cells with different input resistances? Were there any electrophysiological or morphological correlates? Given the well-known neurogenesis in this region the authors will need to identify the DG cells and show whether dendritic spikes are a true general phenomenon or limited to young immature cells.

4) The authors find that the expression of LTP is correlated with the presence of fast rising events at the soma during induction, which they suggest reflects the generation of dendritic spikes. These somatic events need to be better quantified (see also point 5). What were their kinetics and amplitude? How many of these events were "necessary" for LTP to be observed? Was there a correlation between the number of these events and the magnitude of LTP?

5) The experiment showing that blocking somatic APs using local somatic TTX applications had no effect on LTP induction is a critical. Yet, it is not well documented. It is only mentioned briefly in the subsection “LTP by TBS at PP–GC synapses requires NMDARs and Na+ channels”, with data apparently pooled with other data in Figure 1D.

6) The rationale for Figure 4 is unclear. Firstly, the impact of the time taken for APs to propagate from the soma to the distal dendrites (1 ms) is minimal. Secondly, why were timings of 1 to 3 ms used in these experiments? Classical STDP is induced using +/- 10 ms timing as the +10 ms time coincides (approximately) with when an AP would be evoked by a synaptic input. What is the rational for +1 and +3 ms timing? In its current form this figure is of limited value.

7) Figure 6 shows, using dual somatic and dendritic recording, that isolated dendritic Na spikes can be evoked at distal dendritic locations during dendritic current injection and importantly also during TBS stimulation. These data potentially provide evidence that the fast rising events observed at the soma during LTP induction are indeed dendritic spikes. If true, at the soma these events should have the same properties and characteristics. Was this the case? This analysis is critical to the argument that dendritic spikes are necessary for LTP induction.

8) In regard to the data in Figure 6, while not essential, it would have been good to show using dendritic recording that APV blocks dendritic spikes evoked by TBS stimulation. This would explain why the somatic representation of these events is absence in APV and why LTP was blocked under these conditions.

9) The argument made in the third paragraph of the Discussion that dendritic K channels may limit AP backpropagation is inconsistent with the observations of Krueppel et al. (2011), who found that 5 mM 4-AP had no impact on dendritic calcium transients mediated by bAPs.

10) There is an overly hyped description of the results, and issues with some of the interpretations of the data:

A) LTP is not learning, synaptic or behavioral. LTP is a form of long-term plasticity that certainly plays a role in learning. The authors go way overboard in selling the significance of their results. Some examples of this are:

Abstract: "associative learning at distal synapses"

Introduction, first paragraph: "synaptic learning" "similar memories"

Discussion, first paragraph: "formation of memories"

Discussion, fourth paragraph: "synaptic learning"

B) I suggest that the authors rewrite these sections more in keeping with the role of LTP as one (of many) cellular/synaptic changes that are important for learning and memory.

11) I would quibble with the authors' conclusions that dendritic spikes are necessary and/or required for LTP. For example, if Na channels were blocked, could LTP be induced if there was still sufficient depolarization in the dendrites? I suspect that the answer to this is yes. Furthermore, in each of the examples shown of a dendritic spike (Figure 1D, Figure 1—figure supplement 1, Figure 6D) there appears to be an accompanying plateau potential, presumably due to NMDA receptors. So, which is required, the Na spike or plateau potential, or both. Given data from other cell types, perhaps the Na spike facilitates the triggering of an NMDA plateau, both of which are required or necessary for LTP induction "under physiological conditions".

12) In Figure 3 the authors explore spike back propagation from soma into dendrites. A plot of amplitude vs distance is shown in Figure 3E. These results are similar to those from Krueppel et al. However, I think the authors should also plot the amplitude as a function of relative distance as Krueppel et al. did in their Figure 1E. This is useful for comparing bAP spike amplitude with other cell types that have different dendritic lengths. It looks to me that the attenuation is not that different (as a function of total length) from pyramidal cells, so I would rewrite the sentence “The bAP attenuation was more pronounced (length constant of 182 µm; n = 46) than in the dendrite of other hippocampal principal neurons (Spruston et al., 1995; Golding et al., 2002; Kim et al., 2012)”. I would also suggest an experiment in which they measure bAP amplitude before and after adding 4AP in the bath to test whether the bAP amplitude is sensitive to block of A-type K channels. Their data in Figure 5 suggest that it would be but Krueppel suggest not. Otherwise, if Na channels have uniform density (their Figure 5), what is the mechanism for the declining bAP amplitude with distance.

13) The data from outside-out patches and shown in Figure 5 are very nice (and impressive) given the small size of the dendrites. However, the figure legend states that all the recordings were performed with 5 mM 4AP and 20 mM TEA in the bath. If this is correct, what is the outward current shown in 5A?

14) While the authors suggest a "unique" distribution of Na and K channels (Abstract), one could argue that their results for GCs are qualitatively similar to those from CA1 pyramidal cells, but different from the conclusions of Krueppel et al. The results presented are nonetheless important since so little is known about DGC dendrites.

[Editors’ note: this article was subsequently rejected after discussions between the reviewers, but the authors were invited to resubmit for further consideration.]

Thank you for submitting your work entitled "Dendritic spikes in hippocampal granule cells are necessary for long-term potentiation at the perforant path synapse" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Dan Johnston (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife. Although there is significant novelty in the new findings regarding potential roles of dentate gyrus granule cell dendritic spikes in plasticity, numerous issues remain in this revision. Thus at this point we cannot accept this manuscript for publication, however we would be willing to consider a new submission if you decide to fully address the remaining critiques.

Reviewer #1:

In this revision, Kim and co-authors have addressed most of my major concerns and provided new experimental data in support of the general conclusion that distal dendritic sodium-mediated spikes contribute to LTP at the perforant path (PP) synapses. The manuscript has improved and in my opinion builds a strong case for dendritic spike generation. Although the experiments are in general compelling it still contains errors and statements sometimes lack quantification (see below).

Regarding the request for more detail how the authors calculated outside-out surface area the authors write, "we used Hu and Jonas, 2014". But when reviewing Figure 5 and subsection “Ionic mechanisms of AP backpropagation”, there is an obvious mistake, which was present already in the first version of this manuscript. The y-axis shows negative values for sodium conductance density, which is biophysically implausible. I have the impression that the authors rather plotted current density (unit amperes per square micrometre) but labelled their axes with conductance density (unit siemens). This error seems to propagate through the data presentation for the A-type and delayer rectifier K+ density distributions. I don't think that a correction for driving force was applied based on the numbers and it is neither clear whether and how area was corrected for. The authors may want to consult the article of Schmidt-Hieber and Bischofberger, 2010 (J. Neurosci. 30(30); p. 10233), read also the supplement and thoroughly revisit this part of their study. As requested earlier, it is important they spell out how data were obtained and compare their numbers with previously published data. Both for the sodium and potassium measurements.

Reviewer #2:

Overall the authors have done a good job of addressing the points raised by the reviewers. Nevertheless, I have suggestions for additional changes to the manuscript that do not require new experiments:

1) I was a little surprised to see no clear correlation between the number of putative dendritic spikes observed at the soma during TBS stimulation and the magnitude of LTP (Author response image 3). Also, the number of putative dendritic spikes associated with LTP induction seems very low. As the authors indicate this may be because they have missed detecting dendritic spikes in their somatic recordings, as suggested by the data shown in Author response image 4. The capacity to detect dendritic spikes in their somatic recordings is key to the idea that LTP induction requires dendritic spike generation, as discussed in the subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs” and concluded from the data in Figure 1. As indicated in my original review, a characterisation of the somatic events detected when dendritic spikes are observed directly during dendritic recordings would provide direct evidence that the fast rising events observed at the soma during TBS stimulation are indeed dendritic spikes. It would also be important in my opinion to quantify how reliably dendritic spikes can be detected by somatic recordings.

2) Some of the supplemental data is key to the story in my opinion. In particular, I am referring to the data in Figure 1—figure supplements 1 and 2. I suggest parts (or all) of Figure 1—figure supplement 1 is included in Figure 1, and parts (or all) of Figure 1—figure supplement 2 is included in Figure 2.

3) I still think the rational for Figure 4 is weak. This figure does not add much to the paper. The observed bAP conduction velocities are not that different from that seen in pyramidal cells where standard +10/-10 ms EPSP-AP timing protocols can evoked STDP at proximal synapses.

Reviewer #3:

The authors have nicely addressed my previous concerns, and I have no other major comments.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "Dendritic spikes in hippocampal granule cells are necessary for long-term potentiation at the perforant path synapse" for further consideration at eLife. Your revised article has been favorably evaluated by Gary Westbrook (Senior Editor), a Reviewing Editor, and two reviewers. Both reviewers think that the manuscript has been improved, yet additional clarification is needed, as support for the main conclusions of the paper requires additional analysis and improved presentation.

Specific points:

1) There is a little confusion in the first section of the Results ("TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPS"). The authors first state "To ensure that no axosomatic AP initiation and backpropagation occur during TBS, we locally applied tetrodotoxin (TTX) to the GC axon, soma, and proximal dendrites in a subset of experiments (6 out of 13 experiments)", then later you say "To test the contribution of bAPs in this form of LTP, we applied strong synaptic stimulation without perisomatic TTX application". It would be better to first discuss and show the control case in the absence of TTX, describe as you do the presence of putative dendritic spikes during synaptic stimulation and how these are correlated with the magnitude of LTP, and only then introduce the idea that bAPs are not required showing both that the magnitude of LTP is not related to the presence of APs under control conditions and that LTP persists when APs were blocked.

2) The finding that APV "abolished LTP”, but had little if any impact on putative dendritic spikes (Figure 2B and subsection “LTP by TBS at PP–GC synapses requires NMDARs and Na+ channel”: "fast rising events remained unchanged") questions the causal role of these putative dendritic spikes in LTP induction. The authors address this by using QX-314 to block Na+ channels, finding that this "abolished both plateau potentials and TBS-induced LTP". Do the authors mean "abolished both putative dendritic spikes and TBS-induced LTP"? Assuming this is the case some quantification of the effect of QX-314 on putative dendritic spikes is warranted (E.g. data on the number of putative dendritic spikes in control versus QX-314). Note, also the figure reference in the aforementioned subsection should be Figure 2C, D not Figure 2B-D.

3) The authors state that "Pooled data demonstrated that a moderate density of Na+ channels is distributed over the dendritic membrane", yet then indicate dendritic Na+ current densities of between 136 pS/um^2 (proximal) and 206 pS/um^2 (distal). These are high not moderate densities. One model of AP backpropagation in granule cells Krueppel et al. (2011) used dendritic Na+ channel densities of around 2 mS/cm^2 (or 20 pS/um^2). That is, almost a factor of 10 lower than estimated by the authors. The estimated densities of A-type and delayed rectifier type K+ channels are also very high. Given some uncertainty in the accuracy of estimating the patch membrane area based on pipette capacitance, it might therefore be better to simply state the peak current amplitude rather than current density.

4) Addition of the data showing the association between putative dendritic spikes observed at the soma (based on dV/dt) and the direct observation of dendritic spikes during dendritic recordings is most welcome. In my view this data (Figure 5—figure supplement 3) is critical to the paper and therefore should not be "buried" in the supplemental data, but should be included as part of Figure 5 of the manuscript. I would argue the data in Figure 1—figure supplement 1, showing that APs are not required for LTP, is also critical to the story and therefore should also be presented as a main figure rather than supplemental data.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Dendritic spikes in hippocampal granule cells are necessary for long-term potentiation at the perforant path synapse" for further consideration at eLife. Your revised article has been favorably evaluated by Gary Westbrook (Senior Editor), and the Reviewing Editor, John Huguenard.

The manuscript has been improved but there are a few remaining issues that need to be addressed before acceptance, as outlined below:

1) There is some confusion regarding the figure labeling in the Results text, especially around new Figure 5 (formerly Figure 4, subsection “Ionic mechanisms of AP backpropagation”). Please check carefully to ensure that figures are cited correctly.

2) Although strong dendritic spikes are defined as > 10 mV/ms in the subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs”, the definition of weak dendritic spikes is more ambiguous, especially given that some thresholds must have been set to distinguish between EPSPs and weak dendritic spikes. This discriminant belongs in the text. The text table in Figure 7B suggests it is on the order of 3 mV/ms as the minimum somatic detection of a DS.

https://doi.org/10.7554/eLife.35269.039

Author response

Essential revisions:

1) There is a lack of clarification and experimentation why the observed "high" Ia-type current density blocks excitation in the axial direction but still enables local dendritic Na+ spikes. The ion channel distribution profiles alone (Figure 5) are insufficient to explain this since direct experiments addressing their role in spike-timing dependent failures is lacking. It is neither discussed nor addressed experimentally how the specific granule cell cable structure contributes. The interplay between ionic currents and the passive cable properties are the true backbone upon which depolarizations either spreads or attenuates. One prediction is that local block of IA (e.g. with 4-AP) would cause back-propagation facilitation without changing local dendritic spikes. Dual whole-cell voltage recording during synaptic stimulation and local somatic 4-AP application would be one way to address this issue.

We thank the reviewer for this suggestion. Bath application of 4-AP increased both the peak amplitude (142 ± 2.6%, n = 6) and the duration (378 ± 59%, n = 6) of backpropagating APs in the dendrites. It indicates that A-type K+ channels affect AP backpropagation, as the reviewer suggested (now presented in Figure 5—figure supplement 1 in the revised manuscript). Next, focal 4-AP application at the dendritic site lowered the current threshold for dendritic spike initiation (Control, 0.84 ± 0.15 nA; 4-AP, 0.76 ± 0.14 nA, n = 8, P < 0.005), suggesting that A-type K+ channels may contribute to the regenerative dendritic spikes by, for example, confine them to local dendritic locations (now in Figure 6—figure supplement 1 in the revised manuscript). We now discuss these findings in the Results subsections “Ionic mechanisms of AP backpropagation” and “Dendritically initiated Na+ spikes are required for TBS-induced LTP at PP–GC synapses”.

Finally, we also discuss the effect of the GC dendritic morphology on the specific dendritic function as suggested (Discussion, sixth paragraph).

2) The pharmacology experiments in Figure 6—figure supplement 1 are flawed (subsection “Dendritically initiated Na+ spikes are required for TBS-induced LTP at PP–GC synapses”, first paragraph). Dendritic spikes are expected to be mediated by nickel-sensitive T-type Ca2+ channels (at 50 – 100 µM, see e.g. Schmidt-Hieber, Jonas and Bischofberger, 2004) and insensitive to cadmium (a high voltage-activated calcium channel blocker). Experiments with nickel (or TTA-P2) would more firmly demonstrate what the precise pharmacology of the observed dendritic spike is; sodium, T-type calcium channels or both. In fact, it seems cadmium did block a late component in the depolarization in the dendritic recording? Please quantify and clarify.

As the reviewer cited, T-type calcium spikes are observed in newly generated granule cells (i.e., cells with input resistance > 1.5 GΩ, Schmidt-Hieber et al., 2004, Figure 3). Our dataset includes only mature granule cells (input resistance: 108.0 ± 2.6 MΩ). As expected, bath application of 50 µM NiCl2 did not block dendritic spikes (97.9 ± 10.2% of peak amplitude, 4 experiments).

Bath application of 200 µM CdCl2 did not affect the magnitude of the late-depolarizing response of the dendritic spikes evoked by brief current injection (98.9 ± 14.7% of baseline, 3 experiments), and documented it in the legend of Figure 6—figure supplement 1in the revised manuscript. We think that this response is due to the morphological and passive electrical properties of the GC dendrites (Author response image 1).

We acknowledge the suggestion of the reviewer and add now the nickel experiments in Figure 6—figure supplement 1and the cadmium effects in the legend. In this way, we hope to emphasize the sodium origin of the dendritic spikes.

Author response image 1
Effects of Cd2+ on a late component in the dendritic depolarization.

(A, B) Somatic (black) and dendritic (red) voltage responses to dendritic current injection pulses with increasing amplitude (A: 200 pA; B; 700 pA). Voltage responses in the presence of CdCl2 were recorded 10 minutes after bath application of CdCl2. Note the decrease in the subthreshold depolarizing response of the dendritic trace after bath application of CdCl2 in A. (C) Summary of the effect of CdCl2 on the late depolarizing component in 3 dendritic recordings, indicating that CdCl2 did not affect the magnitude of the late depolarizing phase of the dendritic spikes.

https://doi.org/10.7554/eLife.35269.033

3) The role of dendritic spikes is introduced with a lot of confidence in the Abstract as a 'key mechanism for memory formation'. However, it remains unexplained why the authors see it in some but not all DG cells. The age range of 17-24 days is not 'mature' (subsection “Slice preparation and electrophysiology”, third paragraph). Only in the aforementioned subsection it is mentioned that morphology may be different for the DG cells showing dendritic spikes but quantification and details are lacking. Were dendritic spikes seen in cells with different input resistances? Were there any electrophysiological or morphological correlates? Given the well-known neurogenesis in this region the authors will need to identify the DG cells and show whether dendritic spikes are a true general phenomenon or limited to young immature cells.

We thank the reviewer for pointing out these issues. To address the reviewers points, we have recorded newly generated GCs (>1GΩ, see Schmidt-Hieber et al., 2004, Figure 1E). Unlike mature GCs (<200 MΩ), dendritic spikes were not observed in response to high-frequency synaptic stimulation of PP synapses (Figure 2 for reviewers; 4 cells). Direct dendritic recordings in newly generated GCs is not possible due to its rudimentary dendritic arborization (Schmidt-Hieber et al., 2004, Figure 1A-D). Therefore, we have softened our statements on the developmental effects of the dendritic spike generation (Discussion, first paragraph) and add a more carefully wording on memory formation all over the text. To state clearly that we recorded from mature GC neurons, we detailed the intrinsic electrical properties (such as resting membrane potential and input resistances) of the cells in our dataset in the Results section and revised accordingly the Materials and methods section.

Author response image 2
High-frequency PP stimulation does not elicit dendritic spikes in young GCs (>1GΩ).

(A) DIC-image of young GCs located ate the deep side of the GC layer. Scale bar is 50 µm. (B) Somatic voltage responses to current pulses (–10 and +10 pA pulses). (C) Somatic voltage responses to high-frequency synaptic stimulation of the PP inputs (‘Materials and methods’).

https://doi.org/10.7554/eLife.35269.034

4) The authors find that the expression of LTP is correlated with the presence of fast rising events at the soma during induction, which they suggest reflects the generation of dendritic spikes. These somatic events need to be better quantified (see also point 5). What were their kinetics and amplitude? How many of these events were "necessary" for LTP to be observed? Was there a correlation between the number of these events and the magnitude of LTP?

We provide a definition of the somatic readout of a dendritic spike as the first temporal derivative of the voltage (i.e., dV/dt) which is > 5 mV/ms which is based on a rigorous quantification of our dataset (Figure 1—figure supplement 1C and D of the revised paper). We also analyze the correlation between the number of dendritic spikes and LTP magnitude (Author response image 3) and found no correlation (r = 0.096, P = 0.83). We attribute this lack of correlation to the difficulty to unequivocally count dendritic spikes in the somatic recording (Author response image 4).

We add now a sentence in the text, “By examining the temporal derivative (dV/dt) of somatic voltage responses, we define dendritic spikes as those with dV/dt > 5 mV/ms in all further analysis and experiments (Figure 1—figure supplement 1).”

Author response image 3
Magnitude of LTP is not correlated with the number of fast rising events at the soma during TBS.

Data that were used in this analysis were taken from Figure 1G.

https://doi.org/10.7554/eLife.35269.035
Author response image 4
Dendritic spikes often appeared completely attenuated at the soma.

Dual soma-dendrite recordings reveal that high-frequency synaptic stimulation of the PP inputs sometimes evoked dendritic spikes without a clear corresponding somatic responses, ‘spikelets’.

https://doi.org/10.7554/eLife.35269.036

5) The experiment showing that blocking somatic APs using local somatic TTX applications had no effect on LTP induction is a critical. Yet, it is not well documented. It is only mentioned briefly in the subsection “LTP by TBS at PP–GC synapses requires NMDARs and Na+ 140 channels”, with data apparently pooled with other data in Figure 1D.

We apologize for the confusion in the description of the experimental procedures. In Figure 1B-D, somatic APs were absent in all experiments, however perisomatic TTX application was used in a subset of recordings. We have revised the text to describe the experiments clearly (subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs”).

6) The rationale for Figure 4 is unclear. Firstly, the impact of the time taken for APs to propagate from the soma to the distal dendrites (1 ms) is minimal. Secondly, why were timings of 1 to 3 ms used in these experiments? Classical STDP is induced using +/- 10 ms timing as the +10 ms time coincides (approximately) with when an AP would be evoked by a synaptic input. What is the rational for +1 and +3 ms timing? In its current form this figure is of limited value.

We agree with the reviewer that the findings deserve an appropriate explanation. The conduction velocity on the AP in the dendrites is 222 µm/ms (Figure 3). Thus, the AP originated at the soma will arrive at the distal PP synapses (~200–~278 µm from the soma) about one millisecond later. For that reason, we argued that evoking a somatic AP 10 ms after the synaptic stimulation (i.e. classical STDP protocol, Markram et al., 1997) will not provoke a coincidental AP-mediated depolarization with the presynaptic input in the distal GC dendrites (i.e. AP will arrive too late at the synaptic contact). Furthermore, distal dendritic EPSPs in GCs are remarkably brief (~8.8 ms; Schmidt-Hieber et al., 2007). We accounted for both effects by adjusting the induction protocol with short pairings (1–3 ms) expecting to match the synaptic depolarization at the distal GC synapses. We found that it does not produce LTP, and concluded that the somatically initiated AP is not relevant for plasticity at the distal synapses in GCs. We now document this rationale in the Results section (subsection “Backpropagation of axosomatic APs in the dendrites of GCs”) and in the revised Figure 4.

7) Figure 6 shows, using dual somatic and dendritic recording, that isolated dendritic Na spikes can be evoked at distal dendritic locations during dendritic current injection and importantly also during TBS stimulation. These data potentially provide evidence that the fast rising events observed at the soma during LTP induction are indeed dendritic spikes. If true, at the soma these events should have the same properties and characteristics. Was this the case? This analysis is critical to the argument that dendritic spikes are necessary for LTP induction.

We thank the reviewer for this comment. We have characterized the somatic responses when dendritic spikes were elicited. Thus, dendritic spikes can be identified as the first temporal derivative of the somatic voltage which is larger than 5 mV/ms. (now presented in the subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic 119 bAPs” and Figure 6B of the revised paper). Please see also our answer #4.

8) In regard to the data in Figure 6, while not essential, it would have been good to show using dendritic recording that APV blocks dendritic spikes evoked by TBS stimulation. This would explain why the somatic representation of these events is absence in APV and why LTP was blocked under these conditions.

We have followed the reviewer’s suggestion and shown that bath-application of 50 µM DL-AP5 reduces plateau potentials at the soma, indicating that they have a NMDAR-component (68.6 ± 5.5% of baseline, 6 experiments; now in Figure 1—figure supplement 2 of the revised paper). Plateau potentials were more prominent at the dendritic locations, as shown in Author response image 5 (2 cells).

Author response image 5
NMDAR-dependent sustained plateau potentials in dendrites.

(A) Diagram illustrating the recording configuration of a simultaneous somatic (black amplifier) and dendritic (red amplifier) patch-clamp recording on a GC combined with bath application of the NMDAR blocker, DL-AP5. Scale bar is 50 µm. (B) Blockade of NMDARs by a bath application of DL-AP5 (50 μM) abolished sustained plateau potentials. The decrease in plateau potentials were more prominent in the dendrites.

https://doi.org/10.7554/eLife.35269.037

9) The argument made in the third paragraph of the Discussion that dendritic K channels may limit AP backpropagation is inconsistent with the observations of Krueppel et al. (2011), who found that 5 mM 4-AP had no impact on dendritic calcium transients mediated by bAPs.

We have completely revised the role of K+ channels in the AP backpropagation and generation of dendritic spikes to the view of our new experiments. It is now documented in Figure 5—figure supplement 1 andFigure 6—figure supplement 2,and discussed in Results subsections “Ionic mechanisms of AP backpropagation” and “Dendritically initiated Na+ spikes are required for TBS-induced LTP at PP–GC synapses”, see also our answer #1 for details). To address the concern of the reviewer, we have performed additional experiments using both global and local 4-AP application methods. We found that the effect of focal dendritic application of 4AP on the AP backpropagation (Author response image 6) is virtually smaller than that of global application (see also Figure 5—figure supplement 1). We justify the effects of 4-AP on dendritic spikes due to the focal application procedure of our experiment (Discussion section, fourth paragraph).

Author response image 6
Effect of dendritic A-type K+ channel blockade on AP backpropagation.

(A) Diagram of the experimental setup illustrating the recording configuration of a simultaneous somatic (black amplifier) and dendritic (red amplifier) patch-clamp recording on a GC combined with focal application of 4-AP directly to the dendritic patch. Scale bar is 50 µm. (B) (Left) Sample traces of somatic (black) and dendritic (red) APs evoked by somatic current injection under control (top) and after puff application of 10 mM 4-AP (bottom). (Right) First AP in the train displayed at expanded time scale. Local application of 4-AP to the dendrites selectively affect the duration of APs in the dendrites. (C) Summary of the effects of local dendritic application of 4-AP on peak amplitude and duration of APs. (Left; Soma: 100.8 ± 1.3%; Dendrite: 100.6 ± 6.5% ; n = 6, P = 0.98; Right; Soma: 105.6 ± 2.3%; Dendrite: 130.5 ± 5.2%; n = 6, ***P < 0.005) in 6 somatodendritic recordings. Bars indicate mean ± SEM; circles represent data from individual cells.

https://doi.org/10.7554/eLife.35269.038

10) There is an overly hyped description of the results, and issues with some of the interpretations of the data:

A) LTP is not learning, synaptic or behavioral. LTP is a form of long-term plasticity that certainly plays a role in learning. The authors go way overboard in selling the significance of their results. Some examples of this are:

Abstract: "associative learning at distal synapses"

Introduction, first paragraph: "synaptic learning" "similar memories"

Discussion, first paragraph: "formation of memories"

Discussion, fourth paragraph: "synaptic learning"

B) I suggest that the authors rewrite these sections more in keeping with the role of LTP as one (of many) cellular/synaptic changes that are important for learning and memory.

Precise wording was given through the text to denote that long-term potentiation is the cellular correlate of learning and memory.

11) I would quibble with the authors' conclusions that dendritic spikes are necessary and/or required for LTP. For example, if Na channels were blocked, could LTP be induced if there was still sufficient depolarization in the dendrites? I suspect that the answer to this is yes. Furthermore, in each of the examples shown of a dendritic spike (Figure 1D, Figure 1—figure supplement 1, Figure 6D) there appears to be an accompanying plateau potential, presumably due to NMDA receptors. So, which is required, the Na spike or plateau potential, or both. Given data from other cell types, perhaps the Na spike facilitates the triggering of an NMDA plateau, both of which are required or necessary for LTP induction "under physiological conditions".

As the reviewer suggested, the dendritic Na+ spike may contribute to the NMDAR activation necessary for the induction of LTP (now stated in the Discussion, sixth paragraph). To demonstrate this effect, we dialyzed GC cells with 5 mM QX-314 while stimulating PP synapses at high-frequencies to show that without proper sodium channel activation, it is not possible to induce LTP (92.0 ± 8.1% of EPSP baseline, 8 experiments), as in Mishra et al., 2016. We added a summary graph in Figure 2D.

12) In Figure 3 the authors explore spike back propagation from soma into dendrites. A plot of amplitude vs distance is shown in Figure 3E. These results are similar to those from Krueppel et al. However, I think the authors should also plot the amplitude as a function of relative distance as Krueppel et al. did in their Figure 1E. This is useful for comparing bAP spike amplitude with other cell types that have different dendritic lengths. It looks to me that the attenuation is not that different (as a function of total length) from pyramidal cells, so I would rewrite the sentence “The bAP attenuation was more pronounced (length constant of 182 µm; n = 46) than in the dendrite of other hippocampal principal neurons (Spruston et al., 1995; Golding et al., 2002; Kim et al., 2012)”. I would also suggest an experiment in which they measure bAP amplitude before and after adding 4AP in the bath to test whether the bAP amplitude is sensitive to block of A-type K channels. Their data in Figure 5 suggest that it would be but Krueppel suggest not. Otherwise, if Na channels have uniform density (their Figure 5), what is the mechanism for the declining bAP amplitude with distance.

We thank the reviewer for these suggestions. We have analyzed the dendritic distance in biocytin-filled GCs (278 ± 7.4 μm; n=11) and plotted it versus the normalized amplitude of the backpropagating AP. This plot is now in Figure 3E of the revised manuscript and discussed in the Results subsection “Backpropagation of axosomatic APs in the dendrites of GCs”.

The role of A-type potassium channels on bAP and dendritic spikes is now extensively discussed (Figure 5—figure supplement 1 and Figure 6—figure supplement 2) in the revised manuscript and also addressed in our answer #1 and #9.

13) The data from outside-out patches and shown in Figure 5 are very nice (and impressive) given the small size of the dendrites. However, the figure legend states that all the recordings were performed with 5 mM 4AP and 20 mM TEA in the bath. If this is correct, what is the outward current shown in 5A?

We thank the reviewer for his/her awareness on the extreme difficulty of the experiments. The small outward potassium current (~28 pA) is the resistant component to the 5 mM application of 4-AP (EC50= ~1 mM, see Hoffman et al., 1997, Figure 2D). We had previously seen similar outward currents in the dendrites of CA3 neurons (Kim et al., 2012, Figure 4). However, the small sodium currents present in the GC dendrites makes the outward currents more pronounced. We now mention the presence of the remaining 4-AP resistant K+ component in the Figure 5 legend.

14) While the authors suggest a "unique" distribution of Na and K channels (Abstract), one could argue that their results for GCs are qualitatively similar to those from CA1 pyramidal cells, but different from the conclusions of Krueppel et al. The results presented are nonetheless important since so little is known about DGC dendrites.

We agree with the reviewer and removed the word ‘unique’ from the manuscript.

[Editors' note: the author responses to the re-review follow.]

Reviewer #1:

In this revision, Kim and co-authors have addressed most of my major concerns and provided new experimental data in support of the general conclusion that distal dendritic sodium-mediated spikes contribute to LTP at the perforant path (PP) synapses. The manuscript has improved and in my opinion builds a strong case for dendritic spike generation. Although the experiments are in general compelling it still contains errors and statements sometimes lack quantification (see below).

Regarding the request for more detail how the authors calculated outside-out surface area the authors write, "we used Hu and Jonas, 2014". But when reviewing Figure 5 and subsection “Ionic mechanisms of AP backpropagation”, there is an obvious mistake, which was present already in the first version of this manuscript. The y-axis shows negative values for sodium conductance density, which is biophysically implausible. I have the impression that the authors rather plotted current density (unit amperes per square micrometre) but labelled their axes with conductance density (unit siemens). This error seems to propagate through the data presentation for the A-type and delayer rectifier K+ density distributions. I don't think that a correction for driving force was applied based on the numbers and it is neither clear whether and how area was corrected for. The authors may want to consult the article of Schmidt-Hieber and Bischofberger, 2010 (J. Neurosci. 30(30); p. 10233), read also the supplement and thoroughly revisit this part of their study. As requested earlier, it is important they spell out how data were obtained and compare their numbers with previously published data. Both for the sodium and potassium measurements.

We apologize for this mistake. We have taken the suggestion of the reviewer very seriously, and thoroughly revised this figure. To measure the membrane area of the outside-out patches and re-analyze the conductance density, we closely followed Schmidt-Hieber and Bischofberger 2010 as the reviewer suggested. First, we estimated the membrane patch area with pipettes of resistance and geometry identical to that used for our experiments in Figure 5(now presented in the new Figure 4—figure supplement 1 in the revised manuscript). We determined the linear relationship between patch area and pipette conductance under our experimental condition, resulting in A(gP) = 4.7124 x gP+ 0.21421, where A is membrane patch area (µm2) and gPis pipette conductance (µS). Re-analysis of our voltage-clamp recording data using our parameters revealed that overall values of conductance density are larger than those calculated by the parameters established by Hu and Jonas (2014). Nevertheless, our new results corroborate our previous conclusion that a markedly high density of K+ channels and a moderate and uniform density of Na+ channels are present in the dendrites of GCs. These data have been incorporated and are discussed in the revised manuscript in Figure 4D–F. We also have revised the Materials and methods section accordingly (subsection “Data analysis”).

Reviewer #2:

Overall the authors have done a good job of addressing the points raised by the reviewers. Nevertheless, I have suggestions for additional changes to the manuscript that do not require new experiments:

1) I was a little surprised to see no clear correlation between the number of putative dendritic spikes observed at the soma during TBS stimulation and the magnitude of LTP (Author response image 3). Also, the number of putative dendritic spikes associated with LTP induction seems very low. As the authors indicate this may be because they have missed detecting dendritic spikes in their somatic recordings, as suggested by the data shown in Author response image 4. The capacity to detect dendritic spikes in their somatic recordings is key to the idea that LTP induction requires dendritic spike generation, as discussed in the subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs” and concluded from the data in Figure 1. As indicated in my original review, a characterisation of the somatic events detected when dendritic spikes are observed directly during dendritic recordings would provide direct evidence that the fast rising events observed at the soma during TBS stimulation are indeed dendritic spikes. It would also be important in my opinion to quantify how reliably dendritic spikes can be detected by somatic recordings.

We thank the reviewer for the suggestion. As the reviewer suggested, we have characterized the fast depolarizing somatic events when dendritic spikes were present (now presented in the new Figure 5—figure supplement 3 in the revised manuscript). As expected, dendritic spikes lead to a significant acceleration of the rising phase of the somatic EPSP. Interestingly, we observed that there were two groups of the fast rising somatic events, strongly and weakly propagated dendritic spikes (which is presumably due to large impedance drops at branch point). Based on these data, we again analyzed 5,200 (13 cells) somatic voltage responses to TBS induction in Figure 1. As it turns out, we identified weak (dV/dt < 10 mV/ms) and strong (dV/dt > 10 mV/ms) putative dendritic spikes that were distinguishable from EPSPs without any local regenerative events (without dendritic spikes, black, 1.3 ± 0.01 mV/ms; weak dendritic spikes, 4.1 ± 0.4 mV/ms, n = 18; Strong dendritic spikes, 14.5 ± 0.8 mV/ms; Kruskal-Wallis test: P < 0.0001; now presented in Figure 1D). Indeed, we found the strong correlation between the number of putative dendritic spikes observed during TBS and the magnitude of LTP (r = 0.77; P < 0.005; n = 13; now presented in Figure 1F). Nevertheless, we also noticed that the number of putative dendritic spikes associated with LTP induction seems low in some cells. We think that it is likely that the induction of LTP might be occurred by a single local dendritic spikes (Remy and Spruston, 2007, PNAS 104(43); p. 17192-17197).

2) Some of the supplemental data is key to the story in my opinion. In particular, I am referring to the data in Figure 1—figure supplements 1 and 2. I suggest parts (or all) of Figure 1—figure supplement 1 is included in Figure 1, and parts (or all) of Figure 1—figure supplement 2 is included in Figure 2.

We have moved previous Figure 1—figure supplements 1 and 2 into the main body (Figure 1D and 2B, respectively) of the manuscript as suggested.

3) I still think the rational for Figure 4 is weak. This figure does not add much to the paper. The observed bAP conduction velocities are not that different from that seen in pyramidal cells where standard +10/-10 ms EPSP-AP timing protocols can evoked STDP at proximal synapses.

We agree with the reviewer that Figure 4 in the original manuscript could be confusing to readers that the conduction velocity of bAPs would cause the different plasticity form in GCs. Following the suggestion of the reviewer, we decided to remove this Figure 4. We also have revised the text accordingly (subsection “LTP by TBS at PP–GC synapses requires NMDARs and Na+ channels”).

[Editors’ note: the author responses to the re-review follow.]

Specific points:

1) There is a little confusion in the first section of the Results ("TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPS"). The authors first state "To ensure that no axosomatic AP initiation and backpropagation occur during TBS, we locally applied tetrodotoxin (TTX) to the GC axon, soma, and proximal dendrites in a subset of experiments (6 out of 13 experiments)", then later you say "To test the contribution of bAPs in this form of LTP, we applied strong synaptic stimulation without perisomatic TTX application". It would be better to first discuss and show the control case in the absence of TTX, describe as you do the presence of putative dendritic spikes during synaptic stimulation and how these are correlated with the magnitude of LTP, and only then introduce the idea that bAPs are not required showing both that the magnitude of LTP is not related to the presence of APs under control conditions and that LTP persists when APs were blocked.

Following the suggestion of the reviewer closely, we presented the data of TBS-induced LTP experiments in two separate sets upon whether axosomatic APs are present or absent during TBS (Figure 1B, with axonal firing, n = 8; Figure 1C, without axonal firing, n = 13). In Figure 1C, we either used perisomatic TTX application or adjusted the stimulus intensity so that high-frequency PP stimulation would not trigger axosomatic APs during TBS.

Accordingly, we have moved the data in previous Figure 1—figure supplement 1 (showing that APs are not required for LTP) into the main Figure (now in Figure 1B in the revised manuscript), and have substantially revised Figure 1 and the corresponding the section. We hope that the reviewer will find the revised section improved.

2) The finding that APV "abolished LTP”, but had little if any impact on putative dendritic spikes (Figure 2B and subsection “LTP by TBS at PP–GC synapses requires NMDARs and Na+ channel”: "fast rising events remained unchanged") questions the causal role of these putative dendritic spikes in LTP induction. The authors address this by using QX-314 to block Na+ channels, finding that this "abolished both plateau potentials and TBS-induced LTP". Do the authors mean "abolished both putative dendritic spikes and TBS-induced LTP"? Assuming this is the case some quantification of the effect of QX-314 on putative dendritic spikes is warranted (E.g. data on the number of putative dendritic spikes in control versus QX-314). Note, also the figure reference in the aforementioned subsection should be Figure 2C, D not Figure 2B-D.

We have corrected the sentence as suggested. We have added the new bar plots comparing the baseline EPSP amplitudes in two experimental groups (EPSPcontrol: 7.05 ± 0.54 mV, n = 7; EPSPQX-314: 9.24 ± 1.21 mV, n = 7; P = 0.10). We also have provided the data on the number of putative dendritic spikes in control versus QX-314, indicating that intracellular application of QX-314 abolished dendritic spike initiation during TBS (control: 4.14 ± 1.06, n = 7; QX-314: 0, n = 7; P < 0.005). It is now documented in Figure 3D and discussed the revised manuscript.

3) The authors state that "Pooled data demonstrated that a moderate density of Na+ channels is distributed over the dendritic membrane", yet then indicate dendritic Na+ current densities of between 136 pS/um^2 (proximal) and 206 pS/um^2 (distal). These are high not moderate densities. One model of AP backpropagation in granule cells Krueppel et al. (2011) used dendritic Na+ channel densities of around 2 mS/cm^2 (or 20 pS/um^2). That is, almost a factor of 10 lower than estimated by the authors. The estimated densities of A-type and delayed rectifier type K+ channels are also very high. Given some uncertainty in the accuracy of estimating the patch membrane area based on pipette capacitance, it might therefore be better to simply state the peak current amplitude rather than current density.

We acknowledge the suggestion of the reviewer and have changed all the conductance density values to the peak current amplitudes, which now presented in the new Figure 5D, E, and F in the revised manuscript. As we used similar glass micropipettes (i.e., resistance and geometry) in both somatic and dendritic recordings, these data will provide reliable comparisons of ion channel densities in the soma and the dendrites. We also have revised the Materials and methods section accordingly.

4) Addition of the data showing the association between putative dendritic spikes observed at the soma (based on dV/dt) and the direct observation of dendritic spikes during dendritic recordings is most welcome. In my view this data (Figure 5—figure supplement 3) is critical to the paper and therefore should not be "buried" in the supplemental data, but should be included as part of Figure 5 of the manuscript. I would argue the data in Figure 1—figure supplement 1, showing that APs are not required for LTP, is also critical to the story and therefore should also be presented as a main figure rather than supplemental data.

We appreciate the suggestion of the reviewer. We agree with the reviewer that the association between putative dendritic spikes observed at the soma and the direct observation of dendritic spikes during dendritic recordings is important information of the manuscript and therefore should be presented in the main body of the manuscript. Accordingly, we have moved previous Figure 5—figure supplement 3 into Figure 6 in the revised manuscript. We also have revised the text accordingly. Please see our answer #1 about the data in previous Figure 1—figure supplement 1.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are a few remaining issues that need to be addressed before acceptance, as outlined below:

1) There is some confusion regarding the figure labeling in the Results text, especially around new Figure 5 (formerly Figure 4, subsection “Ionic mechanisms of AP backpropagation”). Please check carefully to ensure that figures are cited correctly.

We apologize for this error. We have corrected this problem as suggested (subsection “Ionic mechanisms of AP backpropagation”).

2) Although strong dendritic spikes are defined as > 10 mV/ms in the subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs”, the definition of weak dendritic spikes is more ambiguous, especially given that some thresholds must have been set to distinguish between EPSPs and weak dendritic spikes. This discriminant belongs in the text. The text table in Figure 7B suggests it is on the order of 3 mV/ms as the minimum somatic detection of a DS.

We thank the reviewer for raising this important point. EPSPs with or without dendritic spikes could be readily distinguished by the spikelet waveforms. However, we agree with the reviewer that our definition of weak putative dendritic spikes is not clear. To avoid ambiguity, it would be much helpful to set the threshold level of dV/dt to be counted as dendritic spikes. Therefore, we have once again carefully analyzed the dV/dt of somatic voltage responses, and found that dV/dt of the EPSPs that show a distinct spikelet waveform is larger than 2.5 mV/ms. We have added this value as the criteria for somatic detection of weak putative dendritic spikes as requested (subsection “TBS-induced LTP at the PP-GC synapses does not require postsynaptic bAPs”, first paragraph).

Once again, we would like to thank the reviewers for their time and their careful analysis of our manuscript, which helped us to further improve our paper.

https://doi.org/10.7554/eLife.35269.040

Article and author information

Author details

  1. Sooyun Kim

    1. Department of Physiology, Seoul National University College of Medicine, Seoul, Korea
    2. Neuroscience Research Institute, Seoul National University College of Medicine, Seoul, Korea
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    sooyun.kim@snu.ac.kr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2035-3247
  2. Yoonsub Kim

    Department of Physiology, Seoul National University College of Medicine, Seoul, Korea
    Contribution
    Formal analysis, Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Suk-Ho Lee

    1. Department of Physiology, Seoul National University College of Medicine, Seoul, Korea
    2. Neuroscience Research Institute, Seoul National University College of Medicine, Seoul, Korea
    Contribution
    Resources, Writing—review and editing
    Competing interests
    No competing interests declared
  4. Won-Kyung Ho

    1. Department of Physiology, Seoul National University College of Medicine, Seoul, Korea
    2. Neuroscience Research Institute, Seoul National University College of Medicine, Seoul, Korea
    Contribution
    Resources, Supervision, Funding acquisition, Writing—review and editing
    For correspondence
    wonkyung@snu.ac.kr
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1568-1710

Funding

National Research Foundation of Korea (NRF-2015R1C1A1A02037776)

  • Sooyun Kim

Ministry of Education (Brain Korea 21 PLUS Program)

  • Sooyun Kim

National Research Foundation of Korea (NRF-2010-0027941)

  • Won-Kyung Ho

National Research Foundation of Korea (NRF-2017R1A2B2010186)

  • Won-Kyung Ho

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Jose Guzman and Hua Hu for critically reading the manuscript.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the Seoul National University. All of the animals were handled according to approved institutional animal care and use committee (IACUC) of the Seoul National University. The protocol (Approval #: SNU-090115-7) was approved by the Committee on the Ethics of Animal Experiments of the Seoul National University. Animals were anesthetized by inhalation of 5% isoflurane before sacrifice, and every effort was made to minimize suffering.

Reviewing Editor

  1. John Huguenard, Stanford University School of Medicine, United States

Publication history

  1. Received: January 21, 2018
  2. Accepted: March 25, 2018
  3. Accepted Manuscript published: March 26, 2018 (version 1)
  4. Version of Record published: April 9, 2018 (version 2)

Copyright

© 2018, Kim et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,364
    Page views
  • 464
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)