Processing of Natural Echolocation Sequences in the Inferior Colliculus of Seba’s Fruit Eating Bat, Carollia perspicillata

Natural echolocation sequences used as acoustic stimuli. Two representative echolocation sequences recorded during a forward swing of the pendulum. A–E, Energetic, spectral, and temporal parameters characterizing the simple echolocation sequence. The sequence contains echo information from one object (acrylic glass wall positioned at the end of the swing). A, Call duration (black trace) and call interval (gray trace) over the time course. Call durations and intervals decrease toward the end of the swing. B, Call intensity is independent from object distance and varies between 67 and 82-dB SPL. Echo intensity increases during the approach from 41- to 82-dB SPL. C, Echo delays decrease over time. Oscillogram (D) and spectrogram (E) of the simple echolocation sequence. F–H, Same plots as in B–E but with physical parameters from the multiple-object sequence. During the swing the bat faced three objects. Thus, each call was followed by at least two echoes coming from different objects. Object A is overflown by the animal between 400 and 450 ms. Therefore, echolocation signals after 450 ms do not contain echo information from object A.

Tonotopy of the cIC in C. perspicillata. A, Schematic frontal view on the cIC. CFs increased with recording depth. B, Representative frequency receptive fields from six units recorded from different depths of one penetration track. Depths are indicated by roman numerals (I-VI; A). The CFs of the units are indicated by white stars in the receptive fields and increase with the recording depths. High-frequency tuned neurons typically had multipeaked frequency receptive fields. C, Scatter plot shows the increase of the CF along the recoding depth for 85 collicular units. D, Histrogram represents the distribution of CFs from 90 collicular units recorded in the present study. Units with CFs higher than 35 kHz (dashed vertical line) were classified as high-frequency tuned units and were tested further with the echolocation sequences from Figure 1.

Collicular neurons synchronize more their discharges to the stimulus envelope than cortical neurons. A, Neuronal response from a representative unit of the IC. Median (black trace), 25th, and 75th quantile (gray traces) spike wave form is shown in the left upper corner. Stimulus envelope and oscillogram of the stimulus are shown below. Two call-echo elements of the stimulus envelope are magnified on top. Neuronal responses are shown as raster plots, where one dot indicates an action potential, and as PSTHs. In the sequence situation (Seq), the animals were stimulated with the natural echolocation sequence. In the element situation (Ele) single call-echo elements of the sequence were randomly presented with a 400-ms interstimulus time interval. The time borders of the call-echo elements are indicated by gray vertical lines in the oscillogram. For visualization, the raster plot of the element situation was created by concatenating the neuronal responses to the call-echo elements. Alternating gray scales visualize which action potentials were evoked by which call-echo element. B, Neuronal response from a unit of the AC to the sequence and element situation. Raster plots and PSTHs are organized as in A, except that the binsize of the PSTH was adjusted to 5 ms. C, Boxplot and histogram of the suppression rates in the IC and AC. In the sequence situation, IC units were less suppressed than AC units (Mann–Whitney t test: p < 10⁻⁵). D, Boxplots showing the CC values calculated between PSTHs, with a binsize of 1 ms, and the stimulus envelope. In the IC (black boxplots), the CC values did not differ between the element and sequence situation (p > 0.05), indicating that subcortical suppression prevails the neuronal synchronization to the stimulus envelope. CC values from the AC (gray boxplots) were significantly smaller than in the IC (p < 10⁻⁵) and decreased further from the element to the sequence situation (p < 10⁻⁵). Wilcoxon signed rank test for testing between stimulus conditions and Kruskal Wallis one-way ANOVA and Dunns multiple comparison post hoc test for comparing between IC and AC. E, Scatter plot shows that in AC (gray circles) the suppression rate was correlated with the decrease of neuronal synchronization from the element to the sequence situation (Spearman: r = −0.45; p < 10⁻⁵, f(x) = −0.34x + 0.21). No correlation between the suppression rate and changes in neuronal synchronization was found in the IC (black circles; Pearson: r = −0.02; p = 0.87).

Collicular suppression increases signal-to-noise ratio. A, top, Oscillogram of the stimulus. Vertical gray lines define the time borders of the call-echo elements. Bottom, Population activity map in response to the element situation. Normalized PSTHs were transformed into grayscale coded activity maps. Neuronal activity from each unit is represented in a single row. B, Population activity map in response to the sequence situation. C, Population suppression map calculated by subtracting population activity map in response to the element situation (A) from the map calculated in response to the sequence situation (B). Respectively, bright and dark bins represent high and weak suppression rates. D, E, Median PSTHs calculated form the response to the element (D) and sequence (black PSTH; E) situation. The time course of the median suppression is plotted in gray. Note that strong suppression occurs during and directly after high activity rates. The latter suppression reduces the postactivity to zero (black arrows). F, Boxplots showing the increase in the signal-to-noise ratio in the sequence situation compared to the element situation. Wilcoxon signed rank test: p < 10⁻⁵. norm, normalized; SNR, signal-to-noise ratio.

Collicular suppression sharpens neuronal tuning to call-echo elements. A, Oscillogram of the echolocation sequence. Neuronal response of the example unit from Figure 3A is shown as raster plot and PSTH (binsize = 2 ms). To investigate neuronal tuning to certain call-echo elements, each spike was assigned according to its time point to a call-echo element. Time windows used for spike assignments to corresponding call-echo elements are indicated by alternatingly colored horizontal bars. The spikes assigned to a time window and thus to a call-echo element are correspondingly color coded. The activity rate was plotted against the call-echo elements which can be characterized based on their echo delay (x-axis). Note that, the depicted unit responded more strongly to long than to short delays, having its maximum response at element #8 (best delay of 20 ms). B, C, Normalized population activity maps in response to the element (B) and sequence (C) situation. Units were ordered along the y-axis according to their best delay calculated from the response to the sequence. D, E, Boxplots and histograms represent the best delay (D) and median delay shifts (E), calculated by subtracting the best or median delay in response to the element situation from the best or median delay in the sequence situation, respectively.

Neuronal response to the multiple-object sequence. A–D, Stimulus oscillograms, population activity maps (binsize = 2 ms), and median PSTHs from 77 collicular units in response to the multiple-object (A), object A (B), object B (C), and object C (D) sequence. Legends on the left side from each oscillogram define the position of the object along the swing trajectory. Note that each acoustic event, including calls and echoes, is represented in the response pattern. E, Autocorrelation functions of the PSTHs from an example unit indicated by arrows in A–D. The autocorrelation function of the PSTH in response to the multiobject sequence is wider than the one of the PSTHs in response to the single-object sequences. E, Statistical comparison of the area under the autocorrelation curves of 77 collicular units indicate that the response to the multiple-object sequence was broader than the response to the single-object sequences (Friedman one-way ANOVA and Dunn’s multiple comparison test: p < 10⁻⁵).