A specific role for serotonin in overcoming effort cost

Abstract

Serotonin is implicated in many aspects of behavioral regulation. Theoretical attempts to unify the multiple roles assigned to serotonin proposed that it regulates the impact of costs, such as delay or punishment, on action selection. Here, we show that serotonin also regulates other types of action costs such as effort. We compared behavioral performance in 58 healthy humans treated during 8 weeks with either placebo or the selective serotonin reuptake inhibitor escitalopram. The task involved trading handgrip force production against monetary benefits. Participants in the escitalopram group produced more effort and thereby achieved a higher payoff. Crucially, our computational analysis showed that this effect was underpinned by a specific reduction of effort cost, and not by any change in the weight of monetary incentives. This specific computational effect sheds new light on the physiological role of serotonin in behavioral regulation and on the clinical effect of drugs for depression.

Clinical trial registration: ISRCTN75872983.

Keywords: computational biology; decision-making; effort; escitalopram; human; neuroscience; reward; serotonin; systems biology.

Conflict of interest statement

FM: Supported by The French Ministère de la Recherche and the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 604102 (Human Brain Project).GMG: Holds shares in P1vital Ltd and has served as consultant, advisor or CME speaker in the last 12 months for AstraZeneca, Cephalon/Teva, Convergence, Eli Lilly, GSK, Lundbeck, Medscape, Otsuka, Servier, Sunovion, Takeda.JFWD: Currently advises or carries out research funded by Autifony, Sunovion, Lundbeck, AstraZeneca and Servier; all payment is to the University of Manchester; he has share options in P1vital Ltd.RG: Received compensation as a member of the scientific advisory board of Janssen, Lundbeck, Roche, Takeda; he has served as consultant and/or speaker for Astra Zeneca, Pierre Fabre, Lilly, Otsuka, SANOFI, Servier and received compensation, and he has received research support from Servier.The other authors declare that no competing interests exist.

Figures

Figure 1.. Task design and behavioral performance.

(A) The screenshots depict a trial as it was presented to subjects.Subjects were free to allocate their effort as they wished over the 30s corresponding to the trial duration. They were instructed that their monetary payoff would be proportional to both the monetary incentive and the effort duration, i.e. the time spent squeezing a handgrip harder than a target force level, which varied with task difficulty. Subjects were provided with on-line feedback on the payoff accumulated in the trial (score on the top) and on the instantaneous pressure exerted on the grip (fluid level in the thermometer). The force time series of an example trial is shown below the screenshots, revealing 3 effort periods, with rewarded effort (force above target) plotted in black (not gray). Two factors were manipulated across trials: (i) the incentive level, shown as a coin image (1, 2 or 5p) and (ii) the difficulty level, corresponding to the same white bar in the thermometer reached with different target force levels (70%, 80% or 90% of the maximal force). The last screen summarized the payoff cumulated over preceding trials. (B) Using a double-blind procedure, healthy subjects were assigned to one of the two treatment groups, corresponding to a daily intake of either placebo or escitalopram (10 mg during the initial phase, 20 mg during the intermediate and late phase) during 9 weeks. Each subject completed the effort allocation task three times at distinct treatment phases (initial, intermediate and late). Numbers of subjects and visits correspond to data sets included in the analysis after compliance and quality checks. (C) The three left-most graphs show task performance (as reflected in monetary payoff) sorted by treatment group (black: placebo; gray: escitalopram) and time since treatment onset. Statistical significance was assessed with two-sample, two-sided t-tests. On the right-most plot, payoff was averaged over visits at the subject level. Statistical significance was assessed with ANOVAs including treatments as between-subject factors and test phase (initial, intermediate or late) as a within-subject factor. *p<0.05; **p<0.005. Error bars indicate Student's 95% confidence intervals. DOI:http://dx.doi.org/10.7554/eLife.17282.003

Figure 2.. Computational results.

(A) The cost-evidence accumulation model assumes that effort and rest durations are respectively determined by the accumulation (mean slope Sem) and dissipation (mean slope Srm) of cost evidence between bounds (mean amplitude Am). Possible modulations of these parameters by incentive and difficulty levels were implemented in 20 distinct models. In the best model identified (#20) by Bayesian selection, increasing effort difficulty shortens effort duration by steepening the accumulation slope (a parametric effect controlled by parameter Sed and illustrated with colors from yellow to red). Increasing the incentive level has two effects: first, it shortens rest duration by speeding up the dissipation (parametric effect of Sri, illustrated by colors from dark to light blue); second, it lengthens effort duration by pushing back the bounds (parametric effect of Ai, illustrated by green scaling). (B) Plots show inter-subject means and Student's 95% confidence intervals obtained for the fitted values of model parameters (which were averaged over visits at the subject level). To facilitate visual comparison, scales and offsets were adjusted so that mean and error bars are visually equal across plots in the placebo group. Statistical significance corresponds to ANOVAs including treatment group (escitalopram vs. placebo) as a between-subject factor and treatment phase as a within-subject factor (initial, intermediate or late); **p<0.005. (C) Data in the placebo group served as a baseline to simulate effort and rest durations after imposing a 20% increase in computational parameters. In the table, each row corresponds to a simulated change in one single parameter. Colors denote the effect sizes recovered by model fitting for each parameter, as percent of change compared to baseline. Numbers indicate the percentage of 'hit' (on the diagonal) and 'false alarm' (off-diagonal) in detecting a significant change in parameter values with a paired t-test thresholded at p<0.01. (D) The graph illustrates why the effect of escitalopram, characterized at the computational level as a reduced accumulation slope of cost-evidence during effort (Sem), should translate at the behavioral level into both a longer effort duration and an increased sensitivity of effort duration to incentive level. DOI:http://dx.doi.org/10.7554/eLife.17282.005

Figure 3.. Behavioral results.

(A) Plots show inter-subject means and Student's 95% confidence intervals obtained from linear regression.Regression coefficients were averaged over visits at the subject level. To facilitate visual comparison, scales and offsets were adjusted so that mean and s.e.m. are visually equal across plots in the placebo group. Statistical significance corresponds to ANOVAs including treatment group (escitalopram vs. placebo) as a between-subject factor and treatment phase as a within-subject factor (initial, intermediate or late); *p<0.05, **p<0.005. (B) As predicted by the cost-evidence accumulation model, effort duration and its sensitivity to incentive level are correlated across subjects (one dot corresponds to one subject; values were averaged across visits for each subject). The line shows the linear regression fit obtained when pooling the two treatment groups (ρ56=0.55, p<10–5). DOI:http://dx.doi.org/10.7554/eLife.17282.010