Cell-Type-Specific Contributions of Medial Prefrontal Neurons to Flexible Behaviors

Abstract

Behavioral flexibility and impulse control are necessary for successful execution of adaptive behavior. They are impaired in patients with damage to the prefrontal cortex (PFC) and in some clinically important conditions, such as obsessive-compulsive disorder. Although the medial prefrontal cortex (mPFC) has been investigated as a critical structure for behavioral flexibility and impulse control, the contribution of the underlying pyramidal neuron cell types in the mPFC remained to be understood. Here we show that interneuron-mediated local inactivation of pyramidal neurons in the mPFC of male and female mice induces both premature responses and choice bias, and establish that these impulsive and compulsive responses are modulated independently. Cell-type-specific photoinhibition of pyramidal deep layer corticostriatal or corticothalamic neurons reduces behavioral flexibility without inducing premature responses. Together, our data confirm the role of corticostriatal neurons in behavioral flexibility and demonstrate that flexible behaviors are also modulated by direct projections from deep layer corticothalamic neurons in the mPFC to midline thalamic nuclei.SIGNIFICANCE STATEMENT Behavioral flexibility and impulse control are indispensable for animals to adapt to changes in the environment and often affected in patients with PFC damage and obsessive-compulsive disorder. We used a probabilistic reversal task to dissect the underlying neural circuitry in the mPFC. Through characterization of the three major pyramidal cell types in the mPFC with optogenetic silencing, we demonstrated that corticostriatal and corticothalamic but not corticocortical pyramidal neurons are temporally recruited for behavioral flexibility. Together, our findings confirm the role of corticostriatal projections in cognitive flexibility and identify corticothalamic neurons as equally important for behavioral flexibility.

Keywords: corticostriatal; corticothalamic; mouse optogenetics; prefrontal cortex; probabilistic reversal task; pyramidal neuron.

Figure 1.

Design of probabilistic reversal task and baseline behavioral performance. A, Top, Sequence of events in the probabilistic reversal task. After an intertrial interval, the LED light on the center port was turned on (Initiation cue), and mice initiated a trial by poking the center port (Trial start). Subsequently, LED lights in both left and right ports were turned on, and mice freely made a choice between the two peripheral ports (Choice). Correct choices were followed by delivery of a food pellet at 75% probability, whereas incorrect choices were never rewarded. Bottom, Performance in an example session of 6 trial blocks. Top, Magenta represents right reward blocks. Cyan represents left reward block. Long ticks indicate rewarded trials. Short ticks indicate unrewarded trials. Left choice probability for each trial was calculated from the number of left choices in the recent 10 trials. B, Choice accuracy around reversal: when mice achieved ≥80% choice accuracy in the past 10 trials, the location of the correct side was switched to the other side at 15% probability at the beginning of each trial. C, Stay probability after two consecutive choices to the same peripheral port (blue boxes) or to alternate ports (red boxes). Horizontal black bars inside boxes represent medians. Food reward or no reward in the previous two trials are indicated below the graphs. n = 68 mice, consecutive choices; main effect of outcome type, χ₍₃₎² = 182.56, p = 2.5e-39, alternate choices; main effect of outcome type, χ₍₃₎² = 121.83, p = 3.1e-26, Kruskal–Wallis test. D, Action initiation and response inhibition were measured by the number of premature responses scored when mice directly entered the side choice port without responding to the initiation cue by poking into the center port. The frequency of premature responses gradually decreased during 10 consecutive testing sessions (n = 67 mice, main effect of session, χ₍₉₎² = 102.04, p = 6.1e-18, Friedman test).

Figure 2.

Interneuron-mediated mPFC inactivation impairs impulse control and behavioral flexibility. A, Left, Scheme of the bilateral placement of fiberoptics in the mPFC of VGAT-ChR2 mice for optogenetic activation of cortical interneurons and consequent suppression of pyramidal neuron activity. Right, Time periods of optogenetic stimulation during the probabilistic reversal task. Blue light laser (5 ms pulse at 40 Hz) was delivered in 10% of the trials in one session at three different times during the task: (1) laser during the pre-start period (magenta, 1.1, 3.2, and 8.3 s for 25, 50, and 75 percentile) was turned on between the onset of the center LED light and the center port poke; (2) laser during the pre-choice period (cyan, 1.2, 1.6, and 2.3 s for 25, 50, and 75 percentile) was turned on between the center port poke and the side port poke; and (3) laser during the post-choice period of the previous trial (yellow) was kept on for 3 s after choice. B, Photoinhibition of pyramidal neurons during the pre-start period induced a significant increase in the frequency of premature responses (WT, n = 11 mice, p > 0.05; VGAT-ChR2, n = 12 mice). ***p < 0.001 (Wilcoxon signed-rank test). C, Behavioral flexibility was quantified by the choice bias index (difference between left and right choice probability for each mice). Photoinhibition of pyramidal neurons during the pre-choice period induced a significant increase in choice bias in VGAT-ChR2 mice (WT, n = 11 mice, p > 0.05; VGAT-ChR2, n = 12 mice). **p < 0.01 (Wilcoxon signed-rank). D, E, Photoinhibition of pyramidal neurons during the post-choice period did not affect either premature responses or choice bias in the following trials (D, E): WT, n = 11 mice, p > 0.05; VGAT-ChR2, n = 12 mice, p > 0.05 (Wilcoxon signed-rank test). For the position of fiber placements and estimate of the area receiving photoinhibition, see Figure 2-1.

Figure 3.

Effect of mPFC on choice bias depends on outcome history, but mPFC inactivation does not affect the evaluation of outcomes. A, Choice bias indices calculated separately for different past outcomes. Choice bias indices in pre-choice inactivation trials were significantly larger than corresponding choice bias indices for laser off trials if unrewarded trials occurred during the past two trials but not if the two past trials were both rewarded (left). VGAT-ChR2, n = 12 mice, Trial (n − 2) (n − 1) = (Rewarded, Rewarded); median = 0.15 (Laser OFF); 0.29 (Pre-choice), W = 49, p = 1.00 (Trial (n − 2) (n − 1)) = (Unrewarded, Rewarded); median = 0.11 (Laser OFF); 0.13 (Pre-choice), W = 70, p = 0.048 (Trial (n − 2) (n − 1)) = (Rewarded, Unrewarded); median = 0.15 (Laser OFF); 0.29 (Pre-choice), W = 70, p = 0.048 (Trial (n − 2) (n − 1)) = (Unrewarded, Unrewarded); median = 0.12 (Laser OFF); 0.31 (Pre-choice), W = 71, p = 0.037, Wilcoxon signed-rank test for Laser Off versus Pre-choice. p values were multiplied by 4 for Bonferroni's correction of p values. B, Choice was modeled using a logistic regression model with independent variables for outcomes and inactivation types. Regression coefficients for rewarded or unrewarded outcomes were unchanged in pre-start (magenta) and pre-choice inactivation (cyan) compared with corresponding coefficients in laser-off (gray) trials. This indicated that the increased choice bias (squares) induced by pre-start or pre-choice photoinhibition was independent from the processing of reward information. For inactivation conditions, coefficients of choice plus those of choice × laser interaction were plotted (a + c, a + e, a + g, b + d, b + f, b + h; for the labeling of variables, see Materials and Methods). Absolute values of coefficients of bias terms were significantly increased in pre-start and pre-choice inactivation (VGAT-ChR2, n = 12 mice, OFF vs Pre-start, OFF vs Pre-choice). *p < 0.05 (Wilcoxon signed-rank test with Bonferroni's correction of p value).

Figure 4.

Interneuron-mediated mPFC photoinhibition induces premature responses and choice bias through different cognitive or motor deficits. mPFC photoinhibition during the initiation cue (pre-start: magenta bars) increased premature responses for both biased and anti-biased sides. The biased side corresponds to the side with higher choice probability in each inactivation condition. For all inactivation conditions, VGAT-ChR2: n = 12 mice. **p < 0.01, Laser OFF versus Pre-start inactivation (Wilcoxon signed-rank test). ns: p > 0.05 for bias versus nonbiased direction of pre-start trials (Wilcoxon signed-rank test). All p values were corrected with Bonferroni's method.

Figure 5.

Cre-expressing transgenic mouse lines that target pyramidal neurons in different cortical layers in the mPFC. A, C, E, Sagittal sections of Cre-expressing mouse lines crossed with the reporter Rosa26-floxed-EGFP reporter line. A, Sepw1-Cre (NP39), (C) Colgalt2-Cre (NF111), and (E) Syt6-Cre (KI109). B, D, F, Left, Location of Cre-positive neurons in the mPFC of each mouse line was visualized by unilateral injection of AAV-FLEX-L10-EGFP virus in Sepw1-Cre (B), Colgalt2-Cre (D), and Syt6-Cre (F). Middle column, Magnified views of images in the left column. Right column, Number of cells along cortical depth in the areas shown in the middle column.

Figure 6.

Axonal projections of mPFC cortical pyramidal cells. A, Tracing axonal projections by unilateral injection of AAV-FLEX-ArchT in the mPFC of Colgalt2-Cre (NF111) mice showed dense projections of Cre-positive neurons in the striatum (Inset 1). B, Tracing of AAV-FLEX-ArchT-labeled projections in Syt6-Cre (KI109) mice showed corticothalamic axon bundles in the striatum (Inset 1) and axon terminals in the thalamus (Inset 2), including mediodorsal, ventromedial, and parafascicular nuclei. C, Tracing of AAV-FLEX-ArchT-labeled projections in the mPFC of Sepw1-Cre (NP39) mice showed mostly ipsilateral and contralateral cortical projections (Inset 1): anteroposterior 1.18 mm and anteroposterior −1.70 mm. BLA, Basolateral amygdala; Cg, Cingulate cortex; CPu, caudate-putamen; Th, thalamus; M1, primary motor cortex; M2, secondary motor cortex; ZI, zona inserta.

Figure 7.

Photoinhibition of mPFC corticostriatal and corticothalamic neurons impaired behavioral flexibility without affecting impulsivity. A, AAV-FLEX-ArchT bilateral injection and optical fiber placement in the mPFC of Colgalt2-Cre mice. B, Premature responses were not affected by photoinhibition in ArchT expressing Colgalt2-Cre mice (GFP: n = 8 mice, p > 0.05, ArchT: n = 10 mice, p > 0.05, Wilcoxon signed-rank test). C, Inactivation during the pre-choice period induced a significant increase in choice bias index (GFP: n = 8, p > 0.05, ArchT: n = 10 mice, OFF vs Pre-choice). **p < 0.01 (Wilcoxon signed-rank test). D, Diagram depicting the optical stimulation of axon terminals in the dorsal striatum of Colgalt2-Cre mice expressing AAV-FLEX-ArchT in the mPFC. E, Premature responses were not affected by photoinhibition (ArchT axon terminals: n = 9 mice, p > 0.05, Wilcoxon signed-rank test). F, Photoinhibition at axon terminals in the striatum during the pre-choice period induced a significant increase in choice bias index (ArchT axon terminals: n = 9 mice, OFF vs Pre-choice). **p < 0.01 (Wilcoxon signed-rank test). G, Location of AAV-FLEX-ArchT virus injection and optical fiber implants in the mPFC of Syt6-Cre mice. H, Premature responses were not affected by photoinhibition of ArchT-positive neurons in Syt6-Cre mice (Syt6-Cre: n = 9 mice, p > 0.05, Wilcoxon signed-rank test). I, Photoinhibition of ArchT-positive neurons in Syt6-Cre mice during pre-choice period induced a significant increase in choice bias index (Syt6-Cre: n = 9 mice, OFF vs Pre-choice). *p < 0.05 (Wilcoxon signed-rank test). J, Location of virus injection and fiber implants in the mPFC of Sepw1-Cre mice. K, Premature responses were not affected by photoinhibition of ArchT-positive neurons in Sepw1-Cre mice (Sepw1-Cre: n = 9 mice, p > 0.05, Wilcoxon signed-rank test). L, Choice bias index was not affected by photoinhibition of ArchT-positive neurons in Sepw1-Cre mice (Sepw1-Cre: n = 9 mice, p > 0.05, Wilcoxon signed-rank test). For the virus expression and optical fiber placements, see Figure 7-1.

Figure 8.

Summary of anatomical and genetic results and underlying circuits for behavioral flexibility. A, Summary of mouse lines for expression of ChR2 or ArchT in the mPFC, indicating the dominant projection area and the behavioral phenotypes induced by optogenetic stimulation during the probabilistic reversal task. B, Model for the cortico-striato-thalamocortical circuit underlying behavioral flexibility. Adapted with permission from Pauls et al. (2014). Behavioral flexibility was impaired when corticostriatal and corticothalamic deep layer pyramidal neurons (red arrows) were inactivated. GPe, External globus pallidus; GPi, internal globus pallidus; SNr, substantia nigra pars reticulata.