Selective corticostriatal plasticity during acquisition of an auditory discrimination task

Abstract

Perceptual decisions are based on the activity of sensory cortical neurons, but how organisms learn to transform this activity into appropriate actions remains unknown. Projections from the auditory cortex to the auditory striatum carry information that drives decisions in an auditory frequency discrimination task. To assess the role of these projections in learning, we developed a channelrhodopsin-2-based assay to probe selectively for synaptic plasticity associated with corticostriatal neurons representing different frequencies. Here we report that learning this auditory discrimination preferentially potentiates corticostriatal synapses from neurons representing either high or low frequencies, depending on reward contingencies. We observe frequency-dependent corticostriatal potentiation in vivo over the course of training, and in vitro in striatal brain slices. Our findings suggest a model in which the corticostriatal synapses made by neurons tuned to different features of the sound are selectively potentiated to enable the learned transformation of sound into action.

Extended Data 1

Corticostriatal projections from auditory cortex to striatum. (a) Coronal view for the location in the striatum that receives auditory cortical inputs. (b) Confocal image of auditory cortical axon terminals expressing ChR2-Venus in the striatum. Scale bars: 2 mm.

Extended Data 2

Slope measurement for ChR2-LFP and GABAergic synaptic transmission does not contribute to the CHR2-LFP slope in vivo. (a) Raw ChR2-LFP traces (left panel) were normalized to the amplitude of their corresponding early component (A_i). The normalization factor A_i was determined as the peak of the raw trace in the time window (W1) between 0.5 ms to 1.2 ms after light stimulation onset. (b) The rising phase of the late component of ChR2-LFP (in a time window W2 defined by rise from 10% to 90% of the peak P) was fit linearly, and the slope of the fit was used for the quantification of ChR2-LFP. (c) left panels: ChR2-LFP before (black traces) and after (orange traces) picrotoxin application (20 mM, 5 ul). Raw traces are averaged traces from 60–80 trials at each condition (upper row). Normalized traces are raw traces normalized to their peaks of first components (as illustrated in a). Right panel: slopes measured from normalized traces in control and picrotoxin conditions for each recording before and after picrotoxin application (p=0.8, paired signed-rank test). Data are presented as mean ± s.e.m.

Extended Data 3

ChR2-LFP depends on presence of ChR2-expressing axons. To rule out the possibility that the TTX-insensitive component of the light-evoked response resulted from a photoelectric or other artifact, rather than from ChR2-evoked currents, we assessed light-evoked responses in brain regions that did not express ChR2. (a) Four independent recordings in the auditory striatum (red traces) which receives auditory cortical input (ChR2-expressing axons are present), and the overlying somatosensory cortex (black traces) which lacks auditory cortical input (ChR2-expressing axons are absent). Each pair of recordings is from the same tetrode/fiber bundle. The recordings indicate that the light artifact is negligible under our conditions. (b) Comparison of the first component amplitude from each recording pair.

Extended Data 4

Normalization procedure corrects for variation in light power in vivo (for in vitro data see Fig. 4 d&e). (a) Example of ChR2-LFP recorded at different light levels. (b) Normalized ChR2-LFP the same example in a. (c) Slopes from 5 example recordings across 1–10 mW light level range (colored symbols are examples shown in a & b). Grey lines are drawn from the mean values of each group. Together with the data shown in Fig. 4e, the normalization procedure thus minimizes fluctuations in the response arising from artifactual changes in the number of recruited fibers, but preserves changes arising from actual increases or decreases in synaptic efficacy.

Extended Data 5

Quantification of corticostriatal projection topography. (a) Normalized red and green fluorescence intensities measured across the tonotopic axis from image shown in Figure 4a. (b) Mean red/green intensity ratio across the tonotopic axis, n = 3 sections from 2 rats. Shading, s.e.m.

Extended Data 6

ChR2-LFP slope does nor vary systematically across tonotopic axis in naïve rats. (a) ChR2-LFP slope map from 3 striatal slices (n = 3 rats). (b) Quantification of ChR2-LFP slope across tonotopic axis. Data are presented as mean ± s.e.m.

Extended Data 7

Gradient of ChR2-LFP across the dorsoventral (non-tonotopic) axis showed no difference between the two training groups. (a) Averaged ChR2-LFP slopes with position along tonotopic axis for LowRight and LowLeft (7 rats from each group). (b) Individual gradients of ChR2-LFP across dorsoventral aixs from LowRight and LowLeft groups (p=0.22, paired t-test).

Extended Data 8

Model showing how corticostriatal potentiation could mediate task acquisition. (a) In the naïve rat, the strength of corticostriatal connections does not depend on their frequency preference. (b) Training to associate low stimuli with rightward choices and high stimuli with leftward choices (LowRight) selectively potentiates corticostriatal synapses tuned to low frequencies in the left hemisphere and corticostriatal synapses tuned to high frequencies in the right hemisphere. Thus in the trained rat, low stimuli drive rightward choices and high stimuli drive leftward choices.

Extended Data 9

To exclude the possibility that spiking responses affected the ChR2-LFP measurement, we analyzed the data after median or lowpass filtering. (a) Single trial (upper rows) and average (bottom row) examples of unfiltered, median filtered and Butterworth lowpass filtered responses. Average traces are presented as mean values (black traces) with 95% confidence intervals (grey shading). (b) ChR2-LFP examples in Figure 2a with different filter settings. (c) ChR2-LFP measurements from examples shown in Figure 2a at different filter settings.

Extended Data 10

Changes in ChR2-LFP could result from variation in response timing precision. To rule out this possibility we compared slopes measured from single trial and mean responses. (a) Single trial responses (left panel) and slopes measured from individual trials and mean response (right panel) at a weakly light-responsive site. (b) An example robustly responsive site. (c) Comparison of mean slopes from single trial responses and slopes quantified from mean responses.

Figure 1. Dissection of ChR2-LFP in vivo

(a) Cloud-of-tones task. (b) Example spectrograms of cloud-of-tones stimuli. (c) Recording paradigm to examine corticostriatal synaptic strength in vivo. (d) ChR2-LFP recorded from auditory striatum under control conditions (black trace) and after application of picrotoxin (orange), CNQX and APV (pink) and tetrodotoxin (light gray). The slope of the CNQX/APV-sensitive component was used to quantify corticostriatal synaptic strength (dotted line). Scale bars: 20 μV, 5 ms.

Figure 2. Frequency-selective potentiation of corticostriatal ChR2-LFP slope during learning

(a–b) ChR2-LFP (LFP slope: see Methods) before (black) and during (red) training at example sites tuned to low (a) and high frequency (b). Session 1 is defined as the first session in which the animal performed the full task (see Methods). Scale bars: 2 ms. (c) Population average of normalized (see Methods) ChR2-LFP slope during learning for sites tuned to low (<14 kHz, n=16 sites, closed circles) and high (>14 kHz, n=6 sites, open circles) frequencies. Mean ± s.e.m. (d) Potentiation is restricted to sites tuned to low (<14 kHz) frequencies (23 recording sites from 8 rats; least squares regression of potentiation against frequency, p=0.011).

Figure 3. Potentiation of ChR2-LFP slope is modality-specific

(a) Visual 2-AFC task. (b) ChR2-LFP from an example auditory striatum site during visual and auditory task learning, analyzed as in Fig. 2a. Scale bar: 2 ms. (c) Population average of normalized ChR2-LFP slope during visual and auditory task training. (d) Visual training fails to potentiate auditory striatal input (12 recording sites from 4 rats; least squares regression, p=0.192).

Figure 4. Gradient of corticostriatal ChR2-LFP slopes encodes the association between stimulus and action

(a) Tonotopy of projections from auditory cortex to striatum. Scale bar, 1 mm. (b) Recording paradigm. Light spot (blue) activates a subset of ChR2-expressing corticostriatal axons (green) near recording site. (c) Pharmacological dissection of ChR2-LFP in a striatal slice. Scale bars, 50 μV and 5 ms. (d) Paired in vitro EPSC and LFP at different external divalent ion concentrations. Slopes of LFP measured from raw traces (upper row) and normalized traces (lower row) changed linearly with EPSC amplitudes (R² = 0.96 & 0.99 for circles, R² = 0.94 & 0.81 for squares in upper and lower row respectively. Grey lines in right panels are linear regression fit for each recording pair). Scale bars for d&e, 100 μV and 5 ms. (e) Paired recording at different light levels. Slopes of the LFPs measured from raw traces (upper row) changed monotonically with EPSC amplitudes (R² = 0.94 for solid circles, R² = 0.91 for squares and 0.77 for diamonds). Slopes of the normalized LFPs remain constant (lower row). (f) Normalized ChR2-LFP recorded at many sites within a striatal slice. Sample waveforms (1–3) shown above. ChR2-LFP slope increases with position along tonotopic axis (lower panel). (g) Population data for LowRight (n=7 rats) and LowLeft (n=7 rats). Error bars are s.e.m. (h) Gradient correctly identifies learned association in 14/14 individual rats (binomial test p=0.00006). Slope of example shown in f is indicated in purple. Bars: mean values.