Interplay between population firing stability and single neuron dynamics in hippocampal networks

  1. Edden Slomowitz
  2. Boaz Styr
  3. Irena Vertkin
  4. Hila Milshtein-Parush
  5. Israel Nelken
  6. Michael Slutsky
  7. Inna Slutsky  Is a corresponding author
  1. Tel Aviv University, Israel
  2. Hebrew University, Israel
  3. Mantis Vision, Israel
10 figures and 1 additional file

Figures

Figure 1 with 1 supplement
Quantifying GABABR-mediated synaptic inhibition at synaptic, single neuron and network levels.

(A) Representative ΔF images obtained by single (30 APs@0.2 Hz) and burst (30 APs@6 bursts, each burst contained 5 APs; inter-spike-interval, 10 ms; inter-burst-interval, 5 s) stimulations before and 10 min after 10 µM baclofen application. Scale bar: 5 μm. (B) Dose–response curve for the inhibitory presynaptic effect of baclofen during single vs burst stimulation patterns. Note, shift in the apparent IC50 from 0.33 µM in single stimulation (n = 7–12) to 7.6 µM for burst stimulation type (n = 8–10). (C) Short-term plasticity detected by FM method (calculated as Sburst/Ssingle) was in the range of 2.0 ± 0.01 to 2.3 ± 0.01 for baclofen concentrations in the range of 0.1–10 µM (n = 10–14) comparing to plasticity of control (1.4 ± 0.07, n = 13). (D) Top: image of dissociated hippocampal culture plated on MEA. Black circles at the end of the black lines are the recording electrodes. Scale bar: 200 µm. Bottom: representative trace of recording from one MEA channel. Spikes are detected based on a set threshold (blue dashed line). Inset is a zoom to one detected spike. Scale bars: 8 µV, 35 ms (6 ms for insert). (E) Example of spike sorting for one channel. Bottom: each waveform is represented in principal component space forming three distinct clusters. Top: mean waveforms for each cluster. Scale bar: 10 µV, 1 ms. (F) Representative raster plots of MEA recording before and immediately after application of 10 µM baclofen (***p < 0.0001, unpaired, two-tailed Student's t-test). (G) Acute effect of baclofen on firing rate at the single-unit level. Left: 1 µM baclofen (n = 5, 260 units). Right: 10 µM baclofen (n = 5, 314 units). (H) Acute effect of baclofen on the mean firing rate of the population (the same data as in G). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.003
Figure 1—figure supplement 1
Properties of baclofen-induced changes in synaptic dynamics and single-unit firing in hippocampal neurons.

(A) EPSC (−60 mV holding potential) traces evoked by low-frequency stimulation and by high-frequency spike bursts under control and 10 min after 10 µM baclofen application in acute hippocampal slices. Scale bars: Top: 20 pA, 20 ms; Bottom: 50 pA, 50 ms. (B) Baclofen increases short-term synaptic facilitation calculated based on EPSC measurements (1 µM—n = 5, p < 0.001; 10 µM—n = 4, p < 0.001) in acute hippocampal slices. (C) Distribution of per unit firing rate during 1 hr of control MEA recording (490 units). Firing rates are heavily skewed towards low frequencies. (D) Representative raster plots of MEA recording before and immediately after application of 1 µM baclofen. (E) Per unit correlation between baseline firing rates and the percent change in firing rates after 2 days of control recording showing significant negative correlation (Spearman r = −0.29, p < 0.001, 467 units). (F) Per unit correlation between baseline firing rates and the percent change in firing rates after 2 days in the presence of baclofen showing significant negative correlation (Spearman r = −0.47, p < 0.001, 311 units).

https://doi.org/10.7554/eLife.04378.004
Figure 2 with 4 supplements
Homeostatic regulation of firing rates is more precise at the network, than the single-unit level.

(A) Analysis of the firing rate of each unit in a representative MEA experiment over the course of 2 days of recording in the presence of 10 µM baclofen. Representative units that precisely returned (green), increased (red), and decreased (blue) relative to baseline are highlighted. (B) Analysis of the firing rate of each unit in a representative MEA experiment over the course of 2 days of recording under control conditions. Representative units that didn't change (green), increased (red), and decreased (blue) relative to baseline are highlighted. (C) Mean firing rate of 48 hr control (grey, n = 7) and baclofen (blue, n = 5) MEA recordings. 3 hr of baseline rate are shown for baclofen experiments. (For clarity, only every other hour is shown.) Error bars represent SEM. (D) Statistical comparison of the representative time points (the same data as in C). Error bars represent SEM. (***p < 0.0001, baclofen compared to baseline; all control hours were not significantly changed; repeated-measures ANOVA with Bonferroni's multiple comparison test.) (E) Distribution of unit firing rates (log scale) during baseline and after 2 days in the presence of baclofen. (F) Per unit correlation between baseline firing rates and firing rates after 2 days: Left: control (n = 7; 490 units); Right: in the presence of baclofen (n = 5; 314 units). Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping (see ‘Materials and methods’ for details). Note log scale of both axes. (G) Summary of data in F (*p < 0.05; unpaired, two-tailed Student's t-test). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.005
Figure 2—figure supplement 1
Prolonged exposure to baclofen does not cause reduction in the sensitivity of synapses and neurons to the GABABR-mediated inhibition.

(A) Baclofen (10 µM) was added to a culture dish for 2 days. The medium from this dish was applied to a new MEA. The reduction of firing rate by pre-incubated baclofen was 4 ± 1% (n = 50, p < 0.0001), similar to the one observed by fresh baclofen (Figure 1G). (B) Washout of baclofen causes an increase in mean firing rate relative to both baseline and after 2 days baclofen incubation (relative to baseline: 306% ± 67, relative to 2 days baclofen: 296% ± 38; n = 77 units). (C) Per unit correlation between firing rates before and after washout of baclofen. Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping. Note log scale of both axes. Same units as in (B). (D) Summary of data in (C). (E) Experimental protocol used for determining acute effect of baclofen after 2 days incubation with 10 µM baclofen and subsequent washout. (F) Baclofen (1 µM) decreased FM staining to 52 ± 5% (n = 18) and 49 ± 8% (n = 8) relative to control with and without 2 days incubation with baclofen, respectively (p = 0.89).

https://doi.org/10.7554/eLife.04378.006
Figure 2—figure supplement 2
Firing rate homeostasis is not precise at the level of multi-units.

(A) Per unit correlation between baseline firing rates and firing rates after 2 days: Left: control (n = 7; 192 multi-units); Right: in the presence of baclofen (n = 5; 133 multi-units). Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping (see ‘Materials and methods’ for details). Note log scale of both axes. (B) Summary of data in (A) (unpaired, two-tailed Student's t-test). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.007
Figure 2—figure supplement 3
Effect of GABA uptake inhibitor on mean firing rate.

Mean firing rate is decreased by 10 μM SKF89976A, a selective GAT-1 blocker, and fully recovers after 2 days (n = 46 units).

https://doi.org/10.7554/eLife.04378.008
Figure 2—figure supplement 4
MEA analyses are robust over different parameters.

(A) Histogram detailing the relationship between the mean firing rate of 20 min time segments and the mean firing rate of the full hour they represent. No representative time segment is more than 20% different than the full hour and 90% are within 10% (n = 792 segments). (B) CV was calculated over 8 hr using time segments between 10 s and 1 hr. There was no significant difference when time segments greater than 10 min were used. (C, D, E) Per unit MFR during baseline and after 2 days of baclofen incubation were compared using bootstrapping statistics. No differences were observed when using different durations of time segments (C, p > 0.3), different bin sizes (D, p > 0.9), or different numbers of iterations (E, p > 0.9).

https://doi.org/10.7554/eLife.04378.009
Figure 3 with 1 supplement
Calcium imaging confirms more precise homeostatic regulation of firing rates at the network, than the single-unit level.

(A) Representative traces (ΔF/F) showing the same 16 neurons before, 10 min after 10 µM baclofen application, and after 2 days in the presence of baclofen. Bar scales: 50% ΔF/F, 5 s. (B) Pseudo-color coded image showing cultured neurons expressing GCaMP6f. Scale bar: 50 µm. (C) Summary of the mean rate (averaged peak amplitude per min) before (green), after 10 min (black) and following 2 days (purple) of baclofen application. Population mean rates were restored after 2 days (n = 4, 192 neurons). (D) Distribution of neuron ΔF/F rates (log scale) during baseline and after 2 days in the presence of baclofen. (E) Per neuron correlation between baseline firing rates and firing rates after 2 days in the presence of baclofen (n = 4; 192 neurons). Colors represent neurons that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping. Note log scale of both axes. (F) Summary of data in E (54.3% ± 6 cells were significantly changed). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.010
Figure 3—figure supplement 1
Calcium imaging using GCaMP6f sensor in cultured hippocampal neurons.

(A) Fluorescence change in response to one action potential (AP). Single sweeps (grey) and averages of 10 sweeps (blue) are overlaid. Scale bars: 10% ΔF/F, 500 ms. (B) Fluorescence change (average of 10 sweeps) in response to 1—10 APs (3 trials). Scale bars: 200% ΔF/F, 2 s. (C) Peak fluorescence change as a function of number of APs (normalized to the peak fluorescence changed induced by 1 AP). A linear relationship between ΔF/F and number of APs (slope of linear fit is 1.928 ± 0.078, r2 = 0.99, n = 5). Recurrent activity was suppressed with 20 µM DNQX and 50 µM AP5 in (AB). (D) Example of cultured neurons before and 2 days following baclofen incubation show no notable change in cell appearance. Scale bar: 50 µm. (E) Baseline firing rates 1 hr and after 2 days of control recording show stability with no evidence of phototoxicity (62 neurons). (F) Per unit correlation between baseline ΔF/F mean rates and the percent change in ΔF/F rates after 2 days in the presence of baclofen showing significant negative correlation (Spearman r = −0.33, p < 0.0001, 192 neurons).

https://doi.org/10.7554/eLife.04378.011
Figure 4 with 2 supplements
Temporal firing patterns are homeostatically regulated.

(A) Representative raster plot of MEA recording before, 4 hr and 2 days after application of baclofen. (B) Baclofen causes a transient increase in fraction of spikes participating in network bursts (hours 2–4, p < 0.01; hours 4–10, p < 0.001; hour 12, p < 0.01; hour 14, p < 0.05; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). The number of spikes that are part of network bursts was divided by the total number of spikes. (C) Baclofen causes a transient, short-lived, increase in duration of network bursts (hours 2–4, p < 0.001; hour 6, p < 0.05; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). (D) Baclofen causes a transient increase in fraction of spikes participating in single-unit bursts (hours 2–4, p < 0.01; hours 4–12, p < 0.001; hour 14, p < 0.01; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). The number of spikes that are part of single-unit bursts was divided by the total number of spikes of that unit. (E) Baclofen causes a transient, short-lived, increase in duration of single-unit bursts (hours 2, p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). (F) Per unit correlation of fraction of spikes in single-unit bursts between baseline and 2 days after baclofen application: Top: control (n = 7; 279 units); Bottom: in the presence of baclofen (n = 5; 234 units). Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping (see ‘Materials and methods’ for details). Note log scale of both axes. (G) Summary of data in (D) (*p < 0.05, unpaired, two-tailed Student's t-test). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.012
Figure 4—figure supplement 1
Number of spikes in network- and single-unit bursts in the presence of baclofen.

(A) Baclofen causes a transient increase in the number of spikes per network burst (hours 2–4, p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). (B) Baclofen causes a transient increase in the number of spikes per unit burst (hour 2, p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values).

https://doi.org/10.7554/eLife.04378.013
Figure 4—figure supplement 2
Single-unit burst characteristics are stable across a large range of thresholds.

(A) Single-unit burst data for experiments from Figure 4D–G after using a burst threshold of 100 Hz, minimum three spikes per burst. (B) Single-unit burst data for experiments from Figure 4D–G after using a burst threshold of 20 Hz, minimum three spikes per burst.

https://doi.org/10.7554/eLife.04378.014
Dynamics of I/E ratio following chronic GABABR-mediated inhibition.

(A) Image of patched neuron. Alexa-fluor 488 (10 µM) was added to the patch pipette for imaging. Scale bar: 20 µm. (B) Representative traces of sEPSCs (−65 mV holding potential, bottom) and sIPSCs (+10 mV holding potential, top) for control, 4 hr and 2 days preincubation with baclofen. Measurements of sEPSCs following baclofen pre-incubation were done in the presence of baclofen. (C) Mean integrated excitatory conductance (GE) in control (n = 32), following acute (n = 10), 4 hr (n = 11) and 2 days (n = 16) baclofen application. Excitatory conductance is completely restored following 2 days of exposure to baclofen. Error bars represent SEM. (*p < 0.05, ***p < 0.0001; Kruskal–Wallis test with Dunn's multiple comparison test). (D) Mean integrated inhibitory conductance (GI, same cells as B). Inhibitory conductance is completely restored as well following 2 days of exposure to baclofen. Error bars represent SEM. (***p < 0.001; Kruskal–Wallis test with Dunn's multiple comparison test). (E) Mean I/E ratio per neuron (same cells as in C, D). The I/E ratio of each cell was calculated and the resulting ratios were averaged. *p < 0.05, **p < 0.01, ***p < 0.001; one-way ANOVA with Dunnett's multiple comparison test. Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.015
Figure 6 with 1 supplement
Chronic GABABR-mediated inhibition triggers an increase in mEPSC frequency and amplitude.

(A) Representative traces of mEPSCs for control, 4 hr and 2 day incubations in baclofen. Scale bar: 40 pA, 200 ms. Measurements of mEPSCs were done in the presence of baclofen. (B) Cumulative histograms of mEPSC amplitudes in control (n = 30) and following 4 hr (n = 11) and 2 days (n = 15) of incubation with baclofen. The mean of mEPSC amplitude increased from 25.4 pA in control to 27.7 and 30.8 pA following 4 hr and 2 days of baclofen application, respectively. (C) Summary of data in B. Mean mEPSC amplitude is significantly elevated 1.25-fold (p < 0.05) only after 2 days in baclofen. Error bars represent SEM. *p < 0.05; one-way ANOVA with Dunnett's multiple comparison test. (D) Cumulative histogram of mEPSC inter-event intervals showing a gradual shift to smaller values from control through 4 hr baclofen to 2 days baclofen incubation (the same experiments as in C). (E) Summary of data in D. mEPSC frequency is increased twofold after 4 hr (p < 0.05) and 4.6-fold after 2 days (p < 0.01) incubation in baclofen. *p < 0.05, **p < 0.01; one-way ANOVA with Dunnett's multiple comparison test. Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.016
Figure 6—figure supplement 1
Baclofen does not affect mEPSC frequency and amplitude acutely.

(A) mEPSC amplitude is not effected by acute application of baclofen (Cnt: 25.4 ± 1.5 pA, n = 31; acute Bac: 24.4 ± 2.9, n = 14, p = 0.72, unpaired, two-tailed Student's t-test). (B) mEPSC frequency is slightly, but not significantly, lowered by acute application of baclofen (Cnt: 2.1 ± 0.7 Hz, n = 31; acute Bac: 1.7 ± 0.6 Hz, n = 14, p = 0.7, unpaired, two-tailed Student's t-test).

https://doi.org/10.7554/eLife.04378.017
Intrinsic excitability is increased in response to activity suppression.

(A) Representative traces of voltage responses evoked by 20 pA step of current injections after control, 4 hr and 2 days baclofen incubation, elicited from RMP (scale bars: 40 mV, 100 ms). (B) F–I relationship after control, 4 hr and 2 days baclofen incubation. After 2 days incubation there is a significant leftward shift of the curve showing greater excitability (control, n = 18; 4 hr, n = 18; 2 days, n = 16; *p < 0.05) following long-term GABABR activation. (C) RMP is depolarized after short baclofen incubation (control, n = 19; 4 hr, n = 21; 2 days, n = 14). (D) I–V curve (same cells as C). (E) Rin is significantly increased following 2 day baclofen incubation (same cells as C).

https://doi.org/10.7554/eLife.04378.018
Figure 8 with 1 supplement
Short-term synaptic plasticity is not preserved in networks with similar firing properties.

(A) Cumulative histogram of fluorescence intensity of FM stained puncta following 30 stimuli given at 1 Hz. (ΔFsingle control, n = 15; 6099 puncta; 4 hr, n = 15, 6269 puncta; 2 days, n = 13; 5551 puncta.) (B) Summary of mean ΔFsingle from (A). Mean ΔFsingle is increased already after 4 hr of incubation with baclofen and remains high after 2 days of baclofen incubation. ***p < 0.001; one-way ANOVA with Tukey's multiple comparison test. (C) Experimental protocol used for STP experiments. (D) Representative images of FM1-43 staining from STP experiments. Note the increase in fluorescence intensity after 1 Hz stimulation following baclofen incubation. Scale bar: 5 µm. (E) The mean burst-to-single ratio of S is significantly decreased following baclofen incubation (control, n = 15; 4 hr, n = 15; 2 days, n = 13; p < 0.0001). ***p < 0.001; one-way ANOVA with Tukey's multiple comparison test. Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.019
Figure 8—figure supplement 1
Synaptic vesicle exocytosis evoked by 1 Hz stimulation is increased after 2 days baclofen incubation.

(A) FM1-43 destaining rate curves of 50 synapses under control and 2 days after 10 μM baclofen application. (B) Averaged destaining rate constants in control (n = 5) and in cultures pretreated for 2 days with baclofen (n = 6).

https://doi.org/10.7554/eLife.04378.020
Figure 9 with 3 supplements
Effects of chronic AMPAR blockade on spontaneous network firing.

(A) Representative raster plot of MEA recording before and 2 days after application of CNQX. (B) Changes in mean firing rate following 10 µM CNQX application utilizing MEA recordings (n = 4). 3 hr of baseline rates are shown. There is an immediate and prolonged reduction of firing rate (p < 0.001 for all hours compared to baseline; one-way ANOVA with Bonferroni's multiple comparison test). For clarity, only every other hour is shown. Error bars represent SEM. (C) Distribution of unit firing rates (log scale) during baseline and after 2 days in the presence of CNQX. (D) Per unit correlation between baseline firing rates and firing rates after 2 days in the presence of CNQX (n = 4, 128 units). Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping (see ‘Materials and methods’ for details). Note log scale of both axes. (E) Summary of data in (D). ***p < 0.001; one-way ANOVA with Tukey's multiple comparison test. (F) Cumulative histogram of fluorescence intensity of FM1-43 stained puncta following 30 stimuli given at 1 Hz (ΔFsingle control, n = 9, 3631 puncta; 2 days CNQX, n = 9, 5251 puncta). (G) Summary of mean ΔFsingle from (F). Mean ΔFsingle is increased already after 4 hr of incubation with baclofen and remains high after 2 days of baclofen incubation. ***p < 0.001; one-way ANOVA with Tukey's multiple comparison test. (H) Normalized fraction of spikes in network bursts (the same experiments as in B). There was a significant reduction from the fourth hour onward (p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test). (I) Normalized fraction of spikes in single-unit bursts (the same experiments as in B). There was a significant reduction from the sixth hour onward (hours 6–10, p < 0.05; hours 12–36, p < 0.01; hours 36–48, p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test). Error bars represent SEM.

https://doi.org/10.7554/eLife.04378.021
Figure 9—figure supplement 1
Effect of AMPAR and NMDAR blockers on mean firing rate measured by MEA in hippocampal cultures.

Addition of 10 µM CNQX together with 50 µM AP5 reduced mean firing rate by 80%.

https://doi.org/10.7554/eLife.04378.022
Figure 9—figure supplement 2
CNQX washout reveals an increase in the mean firing rate.

(A) Washout of CNQX causes an increase in MFR relative to both baseline and after 2 days CNQX incubation (relative to baseline: 212% ± 28, relative to 2 days CNQX: 442% ± 67; n = 51 units). (B) Per unit correlation between firing rates before and after washout of CNQX. Colors represent units that significantly increased (red), decreased (blue), or remained stable (green) as determined by bootstrapping. Note log scale of both axes. Same units as in (A). (C) Summary of data in (C).

https://doi.org/10.7554/eLife.04378.023
Figure 9—figure supplement 3
Characteristics of network- and single-unit bursts in the presence of CNQX.

(A) The mean number of spikes in each network burst is reduced by CNQX with no return to baseline (hour 6, p < 0.05; hour 8, p < 0.01; hours 10–48, p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). (B) CNQX causes an acute increase in duration of network bursts (p < 0.001; repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values) followed by a slow, non-significant reduction in duration. (C) The mean number of spikes in single-unit bursts was not changed by CNQX (p > 0.05, repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values). (D) CNQX causes a small non-significant increase in duration of single-unit bursts (repeated-measures ANOVA with Bonferroni's multiple comparison test, compared to baseline values).

https://doi.org/10.7554/eLife.04378.024
Author response image 1

Additional files

Source code 1

Contains MATLAB software and code.

https://doi.org/10.7554/eLife.04378.025

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)

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

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

  1. Edden Slomowitz
  2. Boaz Styr
  3. Irena Vertkin
  4. Hila Milshtein-Parush
  5. Israel Nelken
  6. Michael Slutsky
  7. Inna Slutsky
(2015)
Interplay between population firing stability and single neuron dynamics in hippocampal networks
eLife 4:e04378.
https://doi.org/10.7554/eLife.04378