Social interaction has intrinsic rewarding and motivational properties

To assess the motivation to approach and interact with a conspecific, we developed a social operant task using self-paced instrumental paradigm (Figure 17. A, B). In this task, during the shaping phase (see Material and Methods) animals learn to associate a lever-press (considered as the conditioned stimulus CS) with the opening of a gridded-door (7 sec) that allow interaction with a conspecific (considered as the unconditioned stimulus US; see Material &

Methods for more details, Figure 17. B). We first analyzed the locomotor activity of the mice. By aligning and centering all the trials on the lever-presses, we constructed a peri-event time histogram (Figure 17. C, D). Interestingly, while velocity does not change before association learning, a transient peak of velocity right after the lever-press occurs once the mice learned the task (Figure 17. E).

This higher velocity may reflect the motivation of the animals to interact with the conspecific. Moreover, we observed that the number of lever-presses increases across the days (Figure 17. F, G). This increase translates the associative learning between the lever-press and the opening door, and the ability of the animals to work to interact with unfamiliar conspecific. These results show that conspecifics per se promote motivation.

Representative heatmaps of the velocity and the VTA DA firing rate at different time points around the lever-press show the correlation between the behavior and the neuronal activity (Figure 18. A). Furthermore, defining a lever-press zone (around the lever) and a social area (around the door) allow to define relevant places in the chamber to look at the behavior and the VTA DA activity (Figure 18. A).

Before learning the task, the VTA DA neuron activity increases when the door is open, i.e. when the animals can interact with the conspecific (Figure 18. B, C).

This increase is consistent with the results previously described in the free social interaction task. Remarkably, the significant decrease of the VTA DA activity observed right after the closing door could reflect the unexpected end of social interaction (Figure 18. C).

averaged velocity centered on the lever-presses. RM two-way ANOVA (Time main effect: F(1,28) = 14.9274, P = 0.0006; Learning main effect: F(1,28) = 47.5395, P < 0.0001; Interaction Time x Learning:

F(1,28) = 17.3262, P = 0.0003) followed by Bonferroni-Holm post-hoc test correction. (F) Raster plot example of one animal across the sessions. Each bar represents a lever-press within one session. (G) Number of lever-presses for all the animals during the Shaping phase and the Instrumental phase.

RM one-way ANOVA (Time main effect: F(24,480) = 11.22, P < 0.0001). Error bars report s.e.m.

Interestingly, after learning, VTA DA neuron activity transiently and strongly increases during the CS (lever press) and not anymore during the US (social interaction) (Figure 18. C). To better understand the signification of this increase, we plotted the VTA DA neuron activity during the lever-press depending on the animal’s time to transit to the social zone (around the auto-guillotine door;

Figure 18. D). Remarkably, higher is the CS-evoked peak of VTA DA activity, shorter is the transition between the lever and the social area (Figure 18. D).

Thereby, isolating the trials between transitions (lower and higher than 2 seconds between the CS and the US), we show a significant and transient peak of VTA DA activity at the CS (Figure 18. E). Interestingly, the trials where the transitions are lower than 2 seconds correspond to the trials in which we observe the highest increase in CS-evoked VTA DA activity. Extracting the missing trials (the animal does a transition less than 2 seconds between the lever-press and the social areas but without pressing the lever), we notice a decrease in the VTA DA activity when the mouse was supposed to interact with the conspecific (Figure 18. F, G). These trials can be thought as an unexpected omission of the social interaction and are consistent with similar paradigm done with natural rewards. Quantification of VTA DA neuron activity before and after learning, but also during omission trials, reveals changes in the neuronal activity during the task (Figure 19. A, B). We show an increase of the VTA DA activity in the social zone before learning, while the increase occurs in the lever zone after learning (Figure 19. A). As control, we quantified the VTA DA activity in the same zone when the door is closed (i.e without interaction). After learning, when the animals perform the task properly, by pressing the lever and doing a transition between the zones, the increase is still present in the lever zone. Interestingly, while the transition is made without lever-press, while the CS-evoked increase of VTA DA activity is still present, there is a decrease of activity in the social zone (Figure 19. A).

effect: F(6,58) = 4.1730, P = 0.0015) followed by Bonferroni-Holm post-hoc test correction. Pearson Correlation. (E) Peri-event time histogram of the VTA DA neurons firing rate after learning the task.

Trials when the transitions between the lever-press and the social zone are lower and higher than 2 seconds. Unpaired t test (t(39) = 2.2314). (F) Representative heatmap of all the VTA DA cells during the task, aligned on the exit of the lever zone after learning, when the transitions between the lever and the social areas are less than 2 seconds. (G) Peri-event time histogram of the VTA DA neurons firing rate after learning the task. The trials are aligned on the exit lever zone when the animals did not press the lever. The transition is lower than 2 seconds. These trials are counted as omission.

Paired t test (t(21) = 2.7570). Shaded error bars report s.e.m.

The shift of the VTA DA activity from the US to the CS has been previously described during natural reward and substance of abuse consumptions, and is called reward prediction error (RPE). Our findings strongly suggest that the increase of the VTA DA activity during the lever-press is a prediction about the future conspecific interaction (Figure 18. C). Furthermore, the decrease during omission trials and the correlation between the CS-evoked peak of VTA DA activity and the delay to interact, strongly suggest the usual RPE formalism.

Thereby, positive conspecific interaction would have intrinsic rewarding and motivational properties.

To better understand the role of VTA DA neurons in social behaviors and the neurobiological circuits involved, we then manipulated VTA DA neurons and observed the impact of this manipulation on conspecific interaction.

Figure 19: Quantification of VTA DA neurons during social operant task

(A) Left: VTA DA activity quantification before learning in social and lever zones, when the door is closed or opened. RM two-way ANOVA (Zone main effect: F(1,105) = 2.182, P = 0.1426; Door main effect: F(1,105) = 8.285, P = 0.0048; Interaction Zone x Door: : F(1,105) = 1.084, P = 0.3003). Right: VTA DA activity quantification after learning in social and lever zones, when the door is closed or

opened. RM two-way ANOVA (Zone main effect: F(1,80) = 1.012, P = 0.3174; Door main effect: F(1,80) = 8.171, P = 0.0054; Interaction Zone x Door: : F(1,80) = 1.28, P = 0.2613). (B) Left: VTA DA activity quantification after learning in social and lever zones, with or without transitions between the zones, when the lever-press is made. RM two-way ANOVA (Zone main effect: F(1,84) = 1.09, P = 0.2994;

Transition main effect: F(1,84) = 7.031, P = 0.0096; Interaction Zone x Transition: : F(1,84) = 4.778, P = 0.0316). Right: VTA DA activity quantification after learning in social and lever zones, with or without transitions between the zones, when there is no lever-press. RM two-way ANOVA (Zone main effect: F(1,82) = 3.986, P = 0.0492; Transition main effect: F(1,82) = 0.5139, P = 0.4755; Interaction Zone x Transition: : F(1,82) = 9.893, P = 0.0023). Error bars report s.e.m.

III. Role of VTA DA neurons and neuroligin 3 in sociability