Training-induced circuit-specific excitatory synaptogenesis in mice is required for effort control

Abstract

Synaptogenesis is essential for circuit development; however, it is unknown whether it is critical for the establishment and performance of goal-directed voluntary behaviors. Here, we show that operant conditioning via lever-press for food reward training in mice induces excitatory synapse formation onto a subset of anterior cingulate cortex neurons projecting to the dorsomedial striatum (ACC_→DMS). Training-induced synaptogenesis is controlled by the Gabapentin/Thrombospondin receptor α2δ-1, which is an essential neuronal protein for proper intracortical excitatory synaptogenesis. Using germline and conditional knockout mice, we found that deletion of α2δ-1 in the adult ACC_→DMS circuit diminishes training-induced excitatory synaptogenesis. Surprisingly, this manipulation does not impact learning but results in a significant increase in effort exertion without affecting sensitivity to reward value or changing contingencies. Bidirectional optogenetic manipulation of ACC_→DMS neurons rescues or phenocopies the behaviors of the α2δ-1 cKO mice, highlighting the importance of synaptogenesis within this cortico-striatal circuit in regulating effort exertion.

Fig. 1. Action sequence engages the anterior cingulate cortex during instrumental actions.

a Representation of the Skinner box used for training and testing mice on a lever-press-for-food reinforcer test. b Schematic of the fixed ratio schedules used for the operant training and testing. c Lever press (LP) rate for trained C57BL/6J mice (n = 17 mice, 8 males, and 9 females) across the 9 days of training. d Representative lever press raster plots for FR5 day 4 and FR10 day 9 of LP performance. e Flow chart and example images of the c-Fos⁺ cell segmentation and quantification. f Bar plot of c-Fos⁺ cells for the cerebral cortex regions. g Schematic representation of the ACC microdissection procedure used for the bulk RNA-seq. h Heat map of the top 20 differentially expressed genes (DEG). Bolded genes are the IEGs overexpressed in Trained mice. i Volcano plot organized based on logarithmic fold change (logFC) and P value (−log₁₀P value). Arrows signify genes of interest. Source data are provided as a Source Data file. Drawings in (a, g) are created with BioRender.com.

Fig. 2. A high ratio schedule induces excitatory synaptogenesis in the Anterior Cingulate Cortex.

a Example images of the segmented c-Fos⁺ cells across the layers of the ACC in untrained and trained groups. b Layer-specific quantification of the c-Fos⁺ cells in the dorsal and ventral ACC in trained and untrained C57BL/6J mice. c Representation of the specific cortical layers in which synaptic analysis was performed for VGlut1 and VGAT synapses (created with BioRender.com). d Representative images from Untrained and Trained mice stained with VGluT1 and PSD95 antibodies. The arrows indicate co-localized puncta. e Quantification of VGluT1/PSD95 co-localized puncta density in the ACC and DMS. f Schematic representation of the viral injections to label ACC_→DMS neurons (created with BioRender.com). g Tile scan image of a coronal brain section from a C57BL/6 J tdTomato⁺ mouse. Insets show a magnification of the ACC (i) and DMS (ii). h Schematic representation of mEPSC recording from ACC_→DMS neurons and representative traces. i Cumulative distribution and bar plots of inter-event interval. j Cumulative distribution, and bar plots of amplitude. Source data are provided as a Source Data file.

Fig. 3. VGlut1-PSD95 synapse formation in the anterior cingulate cortex is regulated by the thrombospondin receptor α2δ−1.

a Breeding scheme for the generation of α2δ−1 WT and KO mice (created with BioRender.com). b Representative images of VGluT1/PSD95 staining in the ACC of α2δ−1 WT and KO mice. The arrows in the merged channel indicate co-localized puncta. c Comparison between untrained α2δ−1 WT and KO mice. d Layer-specific comparison of untrained and trained α2δ−1 KO in the ventral ACC. e Example image of the segmented c-Fos⁺ cells across the layers of the ACC of a trained α2δ−1 KO mouse. f Quantification of the c-Fos+ cells in untrained and trained α2δ−1 KO mice compared to WT. Source data are provided as a Source Data file.

Fig. 4. Constitutive lack of α2δ−1 increases effort exertion without affecting the learning of instrumental actions.

a Schematic representation of the fixed ratio and PR schedule used for the α2δ−1 WT and KO mice. b Lever press (LP) rate for the 9 days on the FR schedule for α2δ−1 WT (n = 27 mice; 14 male and 13 female) and KO (n = 20 mice; 10 male and 10 female). c Schematic representation of the progressive ratio (PR) schedule. The value (n) of the ratio (R) increment of 5 for every received reward (i), starting with R = 1. d Representative peri-reward raster histograms of LP for α2δ−1 WT and KO mice. e Cumulative reward count over the PR session (bin=5 min) for α2δ−1 WT (n = 22; 17.2 ± 0.8 rewards) and KO (n = 19; 20.2 ± 0.6 rewards) animals. f Breakpoint for α2δ−1 WT (n = 22; Breakpoint = 79.1 ± 3.8 and KO (n = 19; Breakpoint = 99.9 ± 2.6). g Schematic representation of the extinction schedule in which the action is not reinforced. α2δ−1 WT (n = 22) and KO (n = 19). The normalized number of lever press is reported for the 2 days of testing (dashed line). h Schematic representation of the omission schedule in which the reinforcer is delayed by each press. α2δ−1 WT (n = 17) and KO (n = 8). The lever press/min are reported for the 2 days of testing (dashed line). i Schematic representation of the devaluation test schedule. j Lever press/min in valued and devalued states after pre-feeding for α2δ−1 WT and KO mice. Source data are provided as a Source Data file. Drawings (c, g, h) are created with BioRender.com.

Fig. 5. Conditional deletion of α2δ−1 from ACC_→DMS neurons increases effort exertion without affecting the learning of instrumental actions.

a Schematic representation of the injection strategy to conditionally knock out α2δ−1. b Lever press rate for the 9 days on the FR schedule for α2δ−1 (+/+) and α2δ−1(f/f) mice. c Schematic representation of the progressive ratio (PR) schedule. The value (n) of the ratio (R) increment of 5 for every received reward (i), starting with R = 1. d Representative peri-reward raster histograms of LP for both groups. e Cumulative reward count over the PR session for α2δ−1 (+/+) and α2δ−1(f/f) animals. f Breakpoint for α2δ−1(+/+) and α2δ−1(f/f) animals. g Left: Schematic representation of the Extinction schedule in which the action is not reinforced. Right: Normalized lever press number in a 3 min bins for α2δ−1 (+/+) and α2δ−1(f/f) animals. h Left: Schematic representation of the omission schedule in which the reinforcer is delayed by each press. Right: The lever press/min are reported for the 2 days of testing (dashed line). i Schematic representation of the devaluation test schedule. j Lever press/min in valued and devalued states after pre-feeding. Source data are provided as a Source Data file. Drawings in (a, c, g, h) are created with BioRender.com.

Fig. 6. Circuit-specific conditional deletion of α2δ−1 reduces the excitatory synapses number onto the ACC_→DMS neurons.

a Schematic representation of the electrophysiological recordings from ACC_→DMS tdTomato⁺ neurons in both α2δ−1(+/+) and α2δ−1(f/f) mice. b Example traces of the intrinsic excitability. c Action potential (AP) frequency as function of the injected current. d Average resting membrane potential. e Left: Schematic representation of the tdTomato⁺ neurons used to quantify the VGluT1-PSD95 synapses (created with BioRender.com). Right: Representative images. White arrows point at the puncta within the tdTomato mask. f Quantification of synaptic density between conditions. g Example traces from mEPSC recordings. h Left: Cumulative distribution of the Inter-event interval. Right: Average frequency of mEPSCs. i Left: Cumulative distribution of amplitude in pA. Right: Average amplitude. j Representative traces from mEPSC recordings from untrained and trained α2δ−1(f/f) mice. k Left: Cumulative distribution of the Inter-event interval. Right: Average frequency. l Left: Cumulative distribution of amplitude. Right: Average amplitude. Source data are provided as a Source Data file.

Fig. 7. Optogenetic excitation of ACC_→DMS neurons inhibits the lever press behavior.

a Schematic representation of the viral injections and fiber implants for the optogenetic rescue experiments in the ACC_→DMS projecting neurons of α2δ−1(+/+) and α2δ−1(f/f) mice. b Schematic representation of the FR20 schedule used for the optogenetic experiments with 5 min Off and On schedule. c Representative peri-reward raster histograms of an α2δ−1(+/+) and an α2δ−1(f/f) mouse during the light-Off and light-On periods. d Lever press/min for the α2δ−1(+/+) mice (n = 8) and α2δ−1(f/f) mice (n = 11). Paired two-tailed t test for α2δ−1(+/+) light-Off (27 ± 2.9) and light-On (20 ± 2.6) [t (7) = 6.9; P = 0.0006] and for α2δ−1(f/f) light-Off (42 ± 3.8) and light-On (32 ± 3.8) [t (10) = 6.7]. e Modulation index for CTRL, α2δ−1 (+/+), and α2δ−1 (f/f) mice. One-way ANOVA [F (2, 20) = 16.27, P < 0.001], Tukey’s multiple comparisons revealed a significant difference between CTRL and α2δ−1 (+/+) [q (20) = 7.3; P = 0.0001] and between CTRL and α2δ−1 (f/f) [q (20) = 7.0; P = 0.0002] and no differences between α2δ−1 (+/+) and α2δ−1 (f/f) [q (20) = 0.56, P = 0.916; P = 0.916]. For all graphs: data showed as mean ± s.e.m. alpha = 0.05. Source data are provided as a Source Data file. Drawings in (a–c) are created with BioRender.com.

Fig. 8. Inhibition and excitation of ACC_→DMS neurons have opposite effects on the performance of lever press bout and effort exertion.

a Left: Schematic representation of viral injections for the BLINK2 and behavioral schedule used for inhibitory experiments and Right: Example image of BLINK2 expression within ACC_→DMS neurons. b Lever press/min of the entire PR test for BLINK2 mice showing the effect of BLINK2 inhibition for the first 10 min upon stimulation. n = 10 mice. c Quantification of behavioral parameters during the first 10 min of PR for no stimulation (NoStim) and stimulation (Stim) conditions. Lever press/min during NoStim (18 ± 3.8) and Stim (30 ± 4.5) days of PR. d, Cumulative reward count during NoStim and Stim days of PR. e Breakpoint between NoStim (37 ± 4.0) and Stim (49 ± 4.4) days. f Left: Injection strategy for ChR2 expression and the protocol used for optogenetic experiments using ChR2 during PR schedule. Right: Example image of ChR2 expression in ACC_→DMS neurons. g Quantification of behavioral parameters during the first 10 min of PR for NoStim and Stim conditions. Lever press/min during NoStim (41 ± 4.4) and Stim (26 ± 2.4) days of PR. h Cumulative reward count during NoStim and Stim days of PR. i Breakpoint between NoStim (77 ± 5.2) and Stim (67 ± 3.8) days. j Modulation index of BLINK2 (0.26 ± 0.07; n = 10) and ChR2 (−0.22 ± 0.05; n = 7) mice during the PR schedule. k Left: Proposed intrinsic mechanism of regulation of effort exertion through training-induced excitatory synaptogenesis. Right: Representation of how optogenetic manipulation of ACC_→DMS neurons would affect the effort exertion. Source data are provided as a Source Data file. Drawings in (a, f, k) are created with BioRender.com.