FM velocity selectivity in the inferior colliculus is inherited from velocity-selective inputs and enhanced by spike threshold

Abstract

Frequency modulation (FM) is computed from the temporal sequence of activated auditory nerve fibers representing different frequencies. Most studies in the inferior colliculus (IC) have inferred from extracellular recordings that the precise timing of nonselective inputs creates selectivity for FM direction and velocity (Andoni S, Li N, Pollak GD. J Neurosci 27: 4882-4893, 2007; Fuzessery ZM, Richardson MD, Coburn MS. J Neurophysiol 96: 1320-1336, 2006; Gordon M, O'Neill WE. Hear Res 122: 97-108, 1998). We recently reported that two additional mechanisms were more important than input timing for directional selectivity in some IC cells: spike threshold and inputs that were already selective (Gittelman JX, Li N, Pollak GD. J Neurosci 29: 13030-13041, 2009). Here, we show that these same mechanisms, selective inputs and spike threshold, underlie selectivity for FM velocity and intensity. From whole cell recordings in awake bats, we recorded spikes and postsynaptic potentials (PSPs) evoked by downward and upward FMs that swept identical frequencies at different velocities and intensities. To determine the synaptic mechanisms underlying PSP selectivity (relative PSP height), we derived sweep-evoked synaptic conductances. Changing FM velocity or intensity changed conductance timing and size. Modeling indicated that excitatory conductance size contributed more to PSP selectivity than conductance timing, indicating that the number of afferent spikes carried more FM information to the IC than precise spike timing. However, excitation alone produced mostly suprathreshold PSPs. Inhibition reduced absolute PSP heights, without necessarily altering PSP selectivity, thereby rendering some PSPs subthreshold. Spike threshold then sharpened selectivity in the spikes by rectifying the smaller PSPs. This indicates the importance of spike threshold, and that inhibition enhances selectivity via a different mechanism than previously proposed.

Fig. 1.

Frequency modulation (FM) selectivity for direction, velocity, and intensity: responses to FMs that swept the same 1-oct frequency range with direction and velocity varied as shown in the postsynaptic potential (PSP) plots. A1, B1, and C1: spike selectivity. Spikes (left) are individual traces labeled with spikes/trial. For graphic representation of each cell (right), spike probability was normalized to the maximum evoked probability. A2, B2, and C2: PSP selectivity. PSPs (left) are average responses of 10 trials with spikes filtered. For graphic representation of each cell (right), PSPs were quantified in terms of height and then normalized to the largest PSP. Color scale applies to spikes and PSPs. Labels in A apply to B and C. A: cell 1 preferred slow-velocity, low-intensity sweeps in either direction (best sweep: 5 oct/s, 25 dB SPL). B: cell 2 preferred downward, medium-velocity, medium-intensity sweeps (best sweep: downward, 120 oct/s, 45 dB SPL). C: cell 3 preferred downward, fast-velocity, high-intensity sweeps (best sweep: downward, 320 oct/s, 75 dB SPL).

Fig. 2.

Spike threshold sharpens velocity and intensity selectivity. A: distribution of preferred velocities (the preferred velocity evoked the most spikes; n = 22) NS, nonselective. B: number of velocities (incremented in factors of 2) tested that evoked spikes compared with the number that evoked PSPs; 18 of 22 cells fell below the unity line, indicating that spike threshold sharpened velocity selectivity in most cells. The 3 nonselective cells plus 1 FM-selective cell fell on the line. In 16 of 22 cells we did not test the full range of velocities that evoked PSPs, so this is a minimum estimate of how much spike threshold sharpens velocity selectivity. C: number of intensities tested that evoked spikes compared with the number that evoked PSPs (intensity range, 30–60 dB SPL; no. of intensities tested, 2–5); 8 cells fell below the unity line. In 15 of 18 cells we did not test the full range of intensities that evoked PSPs, so this is a minimum estimate of how much spike threshold sharpens intensity selectivity.

Fig. 3.

PSP height is more sensitive to FM velocity than to direction or intensity. We measured and compared the average (left) and maximum (right) change in PSP height as a function of changing FM velocity, direction, or intensity. In many cells, changing velocity changed PSP height more than changing either direction (A, B) or intensity (C, D). E and F: changing intensity or direction generated similar PSP height changes in the population of cells. Open circles, 3 nonselective (NS) cells.

Fig. 4.

Derived conductance pairs predict membrane responses. A: responses (black) from one cell to six 1-oct FM sweeps at rest and while hyperpolarizing current was injected (16- to 32-kHz sweeps at 45 dB SPL). Velocity and direction are shown below bottom traces and steady-state membrane potential in color on left; spikes (gray, single trials) truncated, firing probability as shown. PSPs are average of 10 trials; same cell as in Fig. 1B (resting membrane potential = −42 mV). Red traces were computed in a model using the conductance traces shown in B and C. B and C: conductance pairs [excitation (ge) and inhibition (gi)] derived from the membrane responses in A: ge and gi aligned with the membrane responses (B) and scaled to show differences in integral, shape, and delay between ge and gi (C). Gray shows 95% confidence interval (bootstrapping). D: predicted membrane potentials plotted against measured membrane potentials at each time point for all downward and upward FMs at the 3 velocities shown. Points fit by linear regression, R² ≥ 0.98.

Fig. 5.

Sweep parameters affect conductance size, shape, and delay between excitation and inhibition. Varying sweep velocity (A–D) or intensity (E–H) altered multiple conductance parameters. A and E: measured PSPs (black; spikes removed by filtering, firing probability as shown) and PSPs predicted (dashed gray) from derived conductance waveforms (bottom). Sweeps were downward; bars show stimulus times. A: velocities varied as shown. Intensity was 45 dB SPL (same cell as Fig. 4). E: intensity varied as shown. Velocity was 150 oct/s. B and F: effect on the delay between excitation and inhibition. Traces normalized to peak height. B: note that the null (15 oct/s) time scale is longer than the other two. C and G: effect on conductance size. Conductances aligned at peaks. D and H: effect on conductance shape. Traces aligned at peaks and normalized to maximum peak height.

Fig. 6.

Determining the conductance parameters underlying PSP selectivity. A: control PSPs computed from a hypothetical control conductance set. Conductance pairs differed in the delay between ge and gi, ge integral, and ge shape; there were no differences in gi shape or integral. B: parameter differences were eliminated by averaging the g waveforms to eliminate size and shape differences and then aligning the ge and gi peaks to eliminate delay differences. C–E, left: experimental PSPs (top) computed from experimental conductance sets (bottom). Right: experimental PSP heights plotted against control PSP heights and fit by linear regression. C: testing ge integral; average g waveforms with ge integrals scaled to be equal to the integrals of the corresponding controls. Control and experimental PSP heights are positively correlated (R² = 0.95; slope = 1.7), indicating that ge integral was important for PSP selectivity. D: testing the delays between ge and gi: average g waveforms with gi latency shifted to match the control delays (compare delays to A). Control and experimental PSP heights are not correlated (R² = 0.14, slope = 0.1), indicating that the delay was relatively unimportant for PSP selectivity. E: testing ge shape: control ge shapes with the integrals scaled to be equal to the average ge integral. Control and experimental PSP heights are negatively correlated (R² = 0.88, slope = −1), indicating that ge shape reduced PSP selectivity.

Fig. 7.

ge integrals predict PSP heights. A–F: format same as Fig. 6. A: control PSPs (black) computed from the control conductance set derived from the responses of an IC cell to 6 FM sweeps (14–28 kHz at 15 dB SPL; velocity and direction as shown). Dashed vertical lines mark excitatory peaks. Rest = −48 mV. Measured PSPs (gray) shown for comparison; spikes removed by filtering, firing probability as shown. B–F: PSPs computed from experimental conductance sets in which the differences in 1 parameter were preserved as shown. Right: experimental PSP heights plotted against control PSP heights, fit by linear regression. G: R² values plotted against the slopes from each of 10 cells. Symbols indicate which parameter was preserved. For ge integral, we distinguished the cells in which FM velocity was varied (filled red circles) from those in which intensity was varied (red +).

Fig. 8.

Rate vs. time: rate parameters predict PSP height. Format same as Fig. 6. A: PSPs computed from the control conductance set (same cell as Fig. 7). B and C: PSPs computed from the experimental conductance set with the integral differences preserved (both ge and gi) (B) and the timing differences preserved (ge shape, gi shape and delay between ge and gi) (C). D: preserving the integral differences resulted in stronger positive correlations between the control and experimental PSP heights than preserving the timing differences.

Fig. 9.

Inhibition adjusts PSP gain, thereby narrowing response selectivity. In the 10 cells with derived synaptic conductances, we modeled the effects of blocking inhibition by comparing PSPs computed with inhibition (gray, dashed) and without inhibition (black) for downward and upward FMs with velocity or intensity varied as shown. A and B: responses of 2 example cells to the preferred sweep direction, velocity varied. The control PSPs computed with excitation (ge) and inhibition (gi) were either subthreshold or near threshold (spike threshold ± 1 SD, dashed line ± gray box). The PSPs computed without inhibition reached or exceeded threshold. Cell 1 (A) is from Fig. 4. Cell 2 (B) is from Fig. 7. C–F: compared with controls, eliminating inhibition increased the absolute PSP heights without changing the height sequence (largest to smallest) except for 4 of 50 PSPs (D and E, circled). C and D: 5 of 10 cells fired spikes to sweeps (firing probability as shown); 10 of 14 subthreshold control PSPs reached or exceeded threshold when inhibition was eliminated, suggesting that blocking inhibition would reduce selectivity. C: velocity varied. D: intensity varied. E and F: 5 cells that did not fire to sweeps. E: 3 cells had voltage-gated channel blockers in the recording pipette. F: 2 cells fired to current steps, but the sweeps tested evoked only subthreshold PSPs. (For each cell, all FMs swept 1 oct with either velocity or intensity varied as shown on plot and intensity or velocity fixed as listed: cell 1, 45 dB SPL; cell 2, 15 dB SPL; cell 3, 15 dB SPL; cell 4, 150 oct/s; cell 5, 175 oct/s; cell 6, 35 dB SPL; cell 7, 75 dB SPL; cell 8, 75 dB SPL; cell 9, 45 dB SPL; cell 10, 50 oct/s).