Learning and attention increase visual response selectivity through distinct mechanisms

Abstract

Selectivity of cortical neurons for sensory stimuli can increase across days as animals learn their behavioral relevance and across seconds when animals switch attention. While both phenomena occur in the same circuit, it is unknown whether they rely on similar mechanisms. We imaged primary visual cortex as mice learned a visual discrimination task and subsequently performed an attention switching task. Selectivity changes due to learning and attention were uncorrelated in individual neurons. Selectivity increases after learning mainly arose from selective suppression of responses to one of the stimuli but from selective enhancement and suppression during attention. Learning and attention differentially affected interactions between excitatory and PV, SOM, and VIP inhibitory cells. Circuit modeling revealed that cell class-specific top-down inputs best explained attentional modulation, while reorganization of local functional connectivity accounted for learning-related changes. Thus, distinct mechanisms underlie increased discriminability of relevant sensory stimuli across longer and shorter timescales.

Keywords: GABAergic interneurons; attention; learning; neural circuits; plasticity; visual cortex.

Figure 1

Visual discrimination learning and attention switching in mice (A) Top, schematic showing virtual reality and imaging setup. (B) Experimental timeline. (C) Schematic of behavioral tasks. Top, visual discrimination: mice were rewarded for licking the reward spout when vertical gratings were presented and not when angled gratings were presented. Olfactory discrimination: mice were rewarded for licking when odor 1 was presented and not when odor 2 or vertical or angled gratings were presented. (D) Behavioral discrimination performance (behavioral d′) across learning and during attention switching (N = 9 mice, 7 of which were tracked across both learning and attention). Connected closed points indicate visual discrimination in individual mice. Open circles indicate olfactory discrimination. See also Figure S1.

Figure 2

Similar changes in stimulus response selectivity across 4 cell classes during learning and attention switching (A) Two example regions of in vivo image planes with GCaMP6f-expressing neurons and the same regions after post hoc immunostaining for PV, SOM, and VIP (orange, blue, and magenta, respectively) following image registration. Identified interneurons are indicated by arrowheads. (B) Example cells from the 4 cell classes, average responses to vertical (blue line), and angled (red line) grating stimuli before (pre) and after (post) learning. Shaded area represents SEM. Gray shading indicates 0–1 s window from stimulus onset used to calculate stimulus selectivity. (C) Stimulus selectivity of the same cells (rows) before and after learning (columns). Cells were ordered by their mean pre- and post-learning selectivity. (D) Average absolute selectivity of the 4 cell classes before and after learning. Error bars represent SEMs. Sign test, ^∗∗p < 0.001. Selectivity distribution in Figure S5A. (E–G) Same as (B)–(D) for attention-switching task. Cells in (C), (D), (F), and (G) were tracked both pre- and post-learning and during the attention task. N = 1,469 PYR, 166 PV, 74 SOM, and 198 VIP cells. See also Figures S2, S4, and S5.

Figure 3

Changes in stimulus selectivity during learning and attention are uncorrelated (A) Relationship between ΔSelectivity with learning (positive values indicate increased selectivity after learning) and ΔSelectivity with attention (positive values indicate increased selectivity with attention) for PYR cells (N = 1,469 cells). (B) Relationship between post-learning selectivity and selectivity in the attend condition for PYR cells. (C and D) Same as (A) and (B) for the 3 interneuron classes (N = 166 PV, 74 SOM, and 198 VIP cells). See also Figure S3.

Figure 4

Increased stimulus selectivity through selective response suppression during learning but enhancement and suppression during attention (A) Difference in calcium responses to the rewarded vertical grating stimulus, post- minus pre-learning (left) or attend minus ignore conditions (right) for all recorded PYR cells (difference-PSTHs). Responses are baseline corrected (subtraction of baseline ΔF/F −0.5 to 0 s before stimulus onset) and aligned to grating onset (dashed line). Cells are sorted by their average amplitude 0–1 s from stimulus onset. N = 1,469 matched PYR cells, in (A)–(E), N = 7 mice. (B) First principal component (PC) of the difference-PSTHs from the learning (left) and attention data (right). Circles indicate the time points (0–1 s) used to determine the PCs. (C) Percentage of variance explained by each PC during learning (left) and attention (right). (D) Distribution of weights from each cell onto the first PC during learning and attention. (E) Relationship between the weights of cells on the first PC during learning and attention. Values greater than the axis limits are pegged to the maximum displayed value. (F) Average PSTHs of all recruited cells—in other words, cells that changed from non-selective to selective stimulus responses during learning; N = 332 and 263 cells recruited with preference for vertical stimulus or angled stimulus, respectively. (G) Average PSTHs of all recruited cells during attention; N = 703 and 690 cells recruited with preference for vertical stimulus or angled stimulus, respectively. Shaded area represents SEM. Gray shading indicates 0–1 s window from stimulus onset used for analysis. See also Figure S6.

Figure 5

Distinct changes in interactions between excitatory and inhibitory cells during learning and attention (A) Top, relationship between the selectivity of individual PV cells and the mean selectivity of the local PYR population within 100 μm of each PV cell, before (pre) and after (post) learning. N = 193 PV cells. Bottom, same comparison for the ignore and attend conditions of the attention-switching task. N = 427 PV cells. (B) Average noise correlations between cell pairs belonging to the same or different cell classes, before and after learning (top) or in the ignore and attend conditions (bottom). Only cells with significant responses to the grating stimuli were included. The number of cell pairs in each cell class combination was as follows: pre-, post-learning, PYR-PYR 153,347, 84,119; VIP-VIP 1,519, 1,046; SOM-SOM 281, 128; PV-PV 2,935, 1,628; PV-VIP 1,390, 920; PV-PYR 36,652, 19,704; PYR-VIP 22,131, 4,368; SOM-PV 1,673, 798; SOM-PYR 11,374, 6,158; SOM-VIP 771, 519. Ignore/attend conditions, PYR-PYR 57,179; VIP-VIP 58; SOM-SOM 380; PV-PV 750; PV-VIP 126; PV-PYR 10,656; PYR-VIP 2,993; SOM-PV 792; SOM-PYR 6,354; SOM-VIP 134. Error bars represent SEMs. The full data distribution can be seen in Figure S5B. (C) Changes in noise correlations (shown in B) due to learning (top) or attention (bottom) as indicated by line thickness and color code. Shorter line segments indicate change in noise correlations between cells of the same type. See also Figure S5.

Figure 6

A circuit model can distinguish between different patterns of top-down attentional modulation (A) The model architecture, indicating connectivity between different cell classes and possible sources of shared external fluctuations. (B) Simulated responses of the 4 cell types to the preferred stimulus. Inset: experimentally obtained average responses of all of the cells in each cell class aligned to the vertical grating stimulus onset. Shading indicates SEM. (C) Changes in stimulus selectivity and noise correlations (NCs) obtained from models with attentional modulation applied to different combinations of cell populations. Both additive and multiplicative modulations were tested. The arrow indicates the condition that best replicated the experimental changes in selectivity and noise correlation. (D) Absolute selectivity of different cell classes without (ignore) and with (attend) attentional modulation provided to PYR and SOM populations, with PYR receiving 0.7 times the modulation of SOM (see Figures S7D and S7E). (E) Changes in noise correlations (NC change) with attentional modulation as in (D) between and within the 4 cell classes, as indicated by line thickness and color code. See also Figure S7.