Fig 1.
CaMKII blockers (KN93 left; tatCN21 right) act through a whole-cell, two-fold division of the GABAA single channel conductance in a subset of cells from trained rats.
A, E: Depicted for each cell is the effect of CaMKII blocker on the average and standard deviation of the event amplitudes. The effect of the CaMKII blocker clearly exhibits two clusters, where one cluster exhibits a large blocker effect on the average and standard deviation of the event amplitude (KN93: trained, n = 16; pseudo, n = 8; naïve, n = 10. tatCN21: trained, n = 10; pseudo, n = 7; naive, n = 5) B, F: Similarly, the effect of CaMKII blocker on the average event amplitude and single channel current exhibits two clusters (KN93: trained, n = 12; pseudo, n = 6; naive, n = 5. tatCN21: trained, n = 10; pseudo, n = 4; naïve, n = 5). C, G: An amplitude distribution curve averaged for all cells in the affected cluster (all cells in the blocker affected cluster for which the single channel current could be calculated; KN93: n = 5; tatCN21: n = 4) is shown for the event amplitudes recorded before (purple) and after (orange) CaMKII blocker. The blocker effect was computationally reversed (black curve) by applying the reverse calculation on the data recorded after applying the CaMKII blocker: The fraction of events that were affected by the drug was calculated (Materials and Methods). This fraction of events was randomly selected from the events recorded after the blocker was applied and the amplitude of these events was multiplied by a factor of two. A new amplitude distribution curve was calculated from the whole set of data, using both the multiplied and the non-multiplied events, exhibiting a fully computationally reversed CaMKII blocker effect. The P values were calculated using Cramer Von Mises two-sample test. D, H: the blocker effect could be reversed in each cell separately. Left: an example of one cell where the histogram of events amplitudes before CaMKII blocker (blue) matched the histogram of the calculated events amplitude after the blocker effect was computationally reversed (red), right the P value was calculated for each cell and the average of these P values was calculated from all cells in the highly affected cluster.
Fig 2.
The CaMKII blocker effect cannot be reversed by an additive model (left) or a 3-fold multiplicative model (right).
A, D: Amplitude distribution curve before and after CaMKII blocker (as in Fig 1C). The black curve in A was calculated assuming a model in which the amplitude of each event was increased by a random number ranging from zero to twice the difference between the average event amplitude before the blocker and after the blocker. The black curve in D was calculated assuming a model of multiplication by a factor of 3, where the fraction of events that were affected by the blocker was calculated for each cell and the calculated amplitude distribution curve was reconstructed accordingly. In both models the reverse calculation assuming these models did not yield an effective reversal of the blocker effect on the amplitudes distribution curve. B, E The effect of the blocker could not be reversed on a single cell basis, yielding a high average P value as was shown in the C, F.
Fig 3.
The cells in the CaMKII blocker affected cluster were affected by the task learning through whole-cell twofold multiplication of the GABAA single-channel current.
A. The average single channel current of the cells in the CaMKII affected cluster from the trained group (T-reduced) was approximately twice that of the average single channel current in the naïve group and the non affected trained group (T–not reduced). B. The amplitude distribution curve of the cells in the affected cluster in the trained group could be reconstructed by multiplying all events in the naive group by a factor of two (T-reduced, n = 9; T-not reduced, n = 12; naïve, n = 12). C. The trained group deviated from the control group mainly by the presence of the sub-group of cells that exhibited a large single channel current. D. The cumulative frequency distribution of the trained group could be explained by multiplying by two the average event amplitude of a randomly selected 35% of the cells in the naïve group (P<0.34 Kolmogorov-Smirnov test). Each point represents the average event amplitude in a neuron (x-axis).
Fig 4.
CaMKII blockers (KN93 left; tatCN21 right) act through a whole-cell twofold division of the AMPA single channel conductance in a subset of cells from trained rats.
A, E For each cell, the effect of CaMKII blocker on the average and standard deviation of the AMPA mediated event amplitudes is shown. The effect of the CaMKII blocker exhibits a clear division into two clusters (KN93: trained, n = 16; pseudo, n = 6; naïve, n = 14. tatCN21: trained, n = 7; pseudo, n = 5; naïve, n = 7). B, F Similarly, the effect of CaMKII blocker on the AMPA mediated average event amplitude and AMPAR single channel current exhibits two clusters (KN93: trained, n = 10; pseudo, n = 6; naïve, n = 10. tatCN21: trained, n = 6; pseudo, n = 3; naïve, n = 4). C, G A distribution curve of the AMPA mediated event amplitude averaged for all cells in the affected cluster recorded before and after the CaMKII blocker (KN93: n = 5; tatCN21: n = 4). The black curve computes the reverse of the blocker effect under the multiplication model after calculating for each cell the fraction of events that were affected by the blocker (see Fig 1C and 1G). D, H the blocker effect could be reversed for each cell separately. Left: an example of one cell; right: the P value was calculated for each cell and the average of these P values was calculated from all cells in the affected cluster.
Fig 5.
The cells in the CaMKII blocker affected cluster were affected by the task learning through whole-cell twofold multiplication of the AMPA single-channel current.
A. The average single channel current of the cells in the CaMKII affected cluster from the trained group (T-reduced, n = 6) was approximately twice that of the average single channel current in the naive group (n = 13) and the non-affected cells in the trained group (T-not reduced, n = 11). B. The amplitude distribution curve of the cells in the more affected cluster could be reconstructed by multiplying all events in the naive group by a factor of two (T-reduced, n = 6; T-not reduced, n = 11; naïve, n = 13). C. The trained group deviated from the control group mainly by the sub-group of cells that exhibited a high single channel current. D. The cumulative frequency curve of the cell’s average distribution curve could not be fully explained by two-fold multiplication of the average event amplitude of a randomly selected 40% of the cells in the naive group.
Fig 6.
The amplification mechanism multiplied the net synaptic current.
A. The net synaptic current is the sum of the inhibitory and excitatory currents. Multiplication of both the inhibitory (upper panel) and excitatory currents (middle panel) multiplies the net synaptic current (lower panel). Therefore, if the net synaptic current is positive this multiplication will further depolarize the cell and if negative will further hyperpolarize the cell. B. The ratio of the activation frequency of inhibitory to excitatory synapses was varied by decreasing the activation frequency of each inhibitory synapse from11Hz to 3Hz, and increasing the activation frequency of each excitatory synapses from2Hz to 10Hz by 0.5Hz (x-axis). Each point on the graph is the average of 10 simulations. The different ratios of activation frequencies (inhibitory activation frequency divided by the excitatory activation frequency x-axis) resulted in different net synaptic currents (y-axis). The amplification mechanism increased the net-synaptic current while maintaining its polarity. A significant multiplicative effect was mainly observed at positive net synaptic currents, thus primarily when the effect of the net synaptic current was excitatory. This figure is a result of one simulated cell. C. Multiplication factors of the net synaptic current as a function of the average membrane potential. The simulation protocol is as described in B, where each color indicates a different set of simulations (in total 10 different sets of simulations). The sets of simulations differed in terms of intrinsic characteristics (resting potential: -90 - -60 mV) and the average strength of the synapses (excit/inhib: 0.6–1.3). Each dot in a set of simulations is a result of a different activation frequency of the inhibitory and excitatory synapses, as was described in B.
Fig 7.
Whole-cell balanced amplification increased the slope of the input-output curve through an increase of the average and the stdev of the membrane potential.
A. The number of spikes measured during an interval of 500 ms, before (yellow) and after (green) whole-cell balanced amplification is shown for increasing fractions of active strong synapses (red: 15%; orange: 30%; purple: 45% and blue: 60%) as evidenced by the increased net synaptic current (X-axis). Whole cell balanced amplification increased the gain of the spiking response to the same input. B. Slopes of the regression lines before (left) and after (right) whole-cell balanced amplification. In all simulations (20 different sets of simulations) the data matched a linear fit (r>0.95) C. Membrane potential before (left) and after (right) the amplification mechanism was applied.D. Distribution of the membrane potential before and after the amplification mechanism was applied. Whole-cell balanced amplification increased both the average and the standard deviation of the membrane potential. Black lines indicates the fit to a normal distribution where the amplification shifted the average by 3.7 mV and multiplied the stdev by 1.6.
Fig 8.
In the balanced state, whole cell balanced amplification functions as a stable and robust linear amplifier.
A. For each level of intrinsic excitability (indicated by the color bar) the multiplication factor induced by whole-cell balanced amplification was calculated for different strengths of net synaptic input (, x-axis) using Eq 2. We examined the sensitivity of the multiplication factor to the net synaptic current by normalizing it to its mean value (Y -axis). For all intrinsic excitability levels (
), indicated by the color bar), the value was close to one, indicating an almost constant multiplication factor. The image above the color bar helps to visualize
, by indicating for each value of
(and thus for the aligned color) the proportion of the voltage above the threshold when the mean synaptic current was zero. B. The mean multiplication factor (Y-axis) was calculated for each level of intrinsic excitability (
, X-axis).
Fig 9.
Whole-cell amplification functions as an instantaneous amplifier.
A. At the peaks of the fluctuations the effect of the amplification was larger. The synaptic activity exhibited the same pattern before and after the amplification thus demonstrating the instantaneous effect of whole-cell balanced amplification. B. The probability of the excitatory synapses was increased according to a time course described by an alpha function (tau = 24 ms), with a multiplication factor of 5, resulting in voltage transients that were considerably larger than the ongoing voltage fluctuations with a half time around 90 ms. Despite a different synaptic activity pattern, and the relatively short duration of these transients the amplification induced by whole-cell balanced amplification was very evident. In both traces the sodium conductance was set to zero to eliminate the action potentials. C. Whole-cell balanced amplification induced robust event based amplification where in the large majority of the synchronized events (74%) the number of spikes increased. Simulation were performed at 30 different simulation sets at which both resting potential and synaptic strength were varied (see Materials and Methods) D. Due to the multiplicative nature of whole-cell amplification, the increase in the number of spikes was pronounced during the increased voltage transients while the increase in the number of spikes during the baseline was negligible.
Fig 10.
Whole-cell balanced amplification had a considerable impact on the probability that very short voltage transients would elicit a spike.
A. An alpha function-like time course of increased rate activation of excitatory synapses was induced (tau = 12 ms, multiplication factor = 5) resulting in brief (half time ~ 30 ms) excursions of voltage bumps. B. For each trace (30 bumps of voltage elevations) the probability of eliciting a spike was calculated before and after whole cell balanced amplification. The simulation parameter space was spanned by varying the resting membrane potential, the synaptic strength, and the activation frequency of the synapses as indicated in the Materials and Methods. Binning the probabilities in all conditions resulted in highly non-linear amplification, where bumps that had a low likelihood of eliciting a spike were amplified to threshold values with probabilities of 50%. Using Eq 2 we tested whether the multiplication of σ and by a factor of 1.7 could account for the simulation results. β was set such that the maximum frequency was 100 Hz or 150 Hz [58] at values where the average current was 3σ more depolarized than threshold (a range in which sodium channels still do not undergo pronounced inactivation). The parameter space (35 different sets of simulations) was spanned by varying
(0.4–0.9, similar to the values obtained in the simulations) and the threshold
(0.5–3.7), yielding results that spanned the same area as the simulations. The values of
are indicated by color coding, and the values of β are indicated by solid lines (max firing frequency of 150 Hz) and double lines (max firing frequency of 100 Hz). C. Binning these calculated values based on the probability before whole-cell amplification yielded a curve that was similar to the simulated curve.
Table 1.
Values for simulation parameters.