1. Neuroscience
Download icon

Dopaminergic and opioidergic regulation during anticipation and consumption of social and nonsocial rewards

  1. Sebastian Korb  Is a corresponding author
  2. Sebastian J Götzendorfer
  3. Claudia Massaccesi
  4. Patrick Sezen
  5. Irene Graf
  6. Matthäus Willeit
  7. Christoph Eisenegger
  8. Giorgia Silani  Is a corresponding author
  1. Department of Psychology, University of Essex, United Kingdom
  2. Department of Cognition, Emotion and Methods in Psychology, University of Vienna, Austria
  3. Department of Clinical and Health Psychology, University of Vienna, Austria
  4. Department of Psychiatry and Psychotherapy, Medical University of Vienna, Austria
Research Article
  • Cited 0
  • Views 691
  • Annotations
Cite this article as: eLife 2020;9:e55797 doi: 10.7554/eLife.55797

Abstract

The observation of animal orofacial and behavioral reactions has played a fundamental role in research on reward but is seldom assessed in humans. Healthy volunteers (N = 131) received 400 mg of the dopaminergic antagonist amisulpride, 50 mg of the opioidergic antagonist naltrexone, or placebo. Subjective ratings, physical effort, and facial reactions to matched primary social (affective touch) and nonsocial (food) rewards were assessed. Both drugs resulted in lower physical effort and greater negative facial reactions during reward anticipation, especially of food rewards. Only opioidergic manipulation through naltrexone led to a reduction in positive facial reactions to liked rewards during reward consumption. Subjective ratings of wanting and liking were not modulated by either drug. Results suggest that facial reactions during anticipated and experienced pleasure rely on partly different neurochemical systems, and also that the neurochemical bases for food and touch rewards are not identical.

eLife digest

Studies in rats and other species have shown that two chemical messengers in the brain regulate how much an animal desires a reward, and how pleasant receiving the reward is. In this context, chemicals called opioids control both wanting and enjoying a reward, whereas a chemical called dopamine only regulates how much an animal desires it. However, since these results were obtained from research performed on animals, further studies are needed to determine if these chemicals play the same roles in the human brain.

Korb et al. show that the same brain chemicals that control reward anticipation and pleasure in rats are also at work in humans. In the experiment, 131 healthy volunteers received either a drug that blocks opioid signaling in the brain, a drug that blocks dopamine signaling, or a placebo, a pill with no effect. Then, participants were given, on several occasions, either sweet milk with chocolate or a gentle caress on the forearm. Participants rated how much they wanted each of the rewards before receiving it, and how much they liked it after experiencing it. To measure their implicit wanting of the reward, participants also pressed a force-measuring device to increase their chances of receiving the reward. Additionally, small electrodes measured the movement of the volunteer’s smiling or frowning muscles to detect changes in facial expressions of pleasure.

Volunteers taking either drug pressed on the device less hard than the participants taking the placebo, suggesting they did not want the rewards as much, and they frowned more as they anticipated the reward, indicating less anticipatory pleasure. However, only the volunteers taking the opioid-blocking drug smiled less when they received a reward, indicating that these participants did not get as much pleasure as others out of receiving it. These differences were most pronounced when volunteers looked at or received the sweet milk with chocolate.

This experiment helps to shed light on the chemicals in the human brain that are involved in reward-seeking behaviors. In the future, the results may be useful for developing better treatments for addictions.

Introduction

Rewards are salient stimuli, objects, events, and situations that induce approach and consummatory behavior by their intrinsic relevance for survival or because experience has taught us that they are pleasurable (Schultz, 2015). Today, our understanding of the neurochemical basis of reward processing rests on 30 years of animal research, and on preliminary confirmatory findings in humans, which led to the identification of two distinct components: wanting, that is, the motivation to mobilize effort to obtain a reward, and liking, that is, the hedonic experience evoked by its consumption (Berridge, 1996; Berridge, 2018; Berridge and Kringelbach, 2015; Berridge and Robinson, 1998). This conceptual division is paralleled in cognitive theories of economic decision making (Kahneman et al., 1997; Berridge and O’Doherty, 2014) that similarly distinguish between decision utility (how much the value attached to an outcome determines its choice or pursuit) and experienced utility (referring to the hedonic experience generated by an outcome).

In animal research, the ‘taste reactivity test’, a method to assess eating-related pleasure by observing facial and bodily reactions of animals and human infants to palatable and aversive tastes, has played a fundamental role in the identification of discrete reward systems in the brain (Barbano and Cador, 2007; Berridge, 2000; Dolensek et al., 2020; Grill and Norgren, 1978; Steiner et al., 2001). Indeed, it has been shown that neither pharmacological disruption nor extensive lesion of dopaminergic neurons affects facial liking reactions (e.g. relaxed facial muscles and licking of the lips) to the consumption of sweet foods in rats (Berridge et al., 1989; Treit and Berridge, 1990), and that greater mesolimbic dopamine release induced by electric stimulation of the hypothalamus results in greater food intake without modulating hedonic reactions (Berridge and Valenstein, 1991). On the other hand, (facial) hedonic reactions to sensory pleasure are amplified by opioid, orexin, and endocannabinoid stimulation of various ‘hedonic hotspots’ of the brain, including the nucleus accumbens (NAc) shell and limbic areas such as the insula and the orbitofrontal cortex (Berridge and Kringelbach, 2015). These stimulations not only increase liking but also result in food approach and feeding behavior (Peciña and Smith, 2010; Taha, 2010).

Evidence of similar neurochemical parsing of reward processing in humans is mainly derived from research in clinical populations and a handful of recent pharmacological studies in healthy volunteers. For example, stimulation of D2/D3 receptors through dopamine agonists can induce compulsive medicament use, gambling, shopping, hypersexuality, and other addictive activities in some patients with Parkinson's disease, often without corresponding changes in subjective liking (Callesen et al., 2013; Evans et al., 2006; Weintraub et al., 2010; but see Meyer et al., 2019 for an account of the complexity of compulsive disorders in Parkinson's disease). Evidence for disrupted motivation to gain immediate rewards has been observed after dopamine D2/D3 receptors blockade in healthy volunteers, in both a pavlovian-instrumental-transfer task and a delay discounting task (Weber et al., 2016). Administration of µ-opioid receptor agonists in healthy individuals has been associated with changes in subjective feelings and motivational responses to different types of rewards, as indicated, for example, by higher pleasantness of the highest-calorie (but least palatable) food option available (Eikemo et al., 2016), greater effort to view, and liking of the most attractive opposite-sex faces (Chelnokova et al., 2014), stronger preference for stimuli with high-reward probability (Eikemo et al., 2017), and enhanced emotional ratings to positive and negative images (Atlas et al., 2014). Furthermore, administration of the non-selective opioid receptor antagonist naloxone to healthy men decreased subjective pleasure associated with viewing erotic pictures and reduced the activation of reward related brain regions such as the ventral striatum (Buchel et al., 2018).

Despite the progress made, the animal research is only partly informative to comprehend reward processing in humans. While animal research allows to investigate the activity of neurons and neurotransmitters in a much more targeted way, it is also limited to certain measures of liking (i.e. behavior and facial expressions, while humans can also provide subjective reports), and has mainly focused on food rewards in the past. Moreover, human and animal research about the neurochemical regulation of reward processing remain difficult to compare, as human pharmacological studies have struggled to adopt translational paradigms and an operationalization of reward that resembles the one used in animal research, that is, measuring decision utility and experienced utility in the same task, providing primary rewards on a trial-by-trial basis, and/or using objective hedonic reactions to consumed rewards, in addition to relying on subjective verbal report (Der-Avakian et al., 2016; Pool et al., 2016).

Including the recording of hedonic facial reactions, the ‘gold standard’ in animal research on the neurochemical regulation of reward processing in human studies seems to be a promising avenue in this regard. Recently, the use of facial electromyography (fEMG) has gained increased attention in the context of human reward processing. Results suggested that human adults relax the corrugator muscle (involved in frowning), and to a lesser extent activate the zygomaticus muscle (involved in smiling), during both anticipation and consumption of different types of pleasurable stimuli, although differences between types of rewards exist (Bershad et al., 2019; Franzen and Brinkmann, 2016; Korb et al., 2020; Mayo et al., 2018; Pawling et al., 2017; Rasch et al., 2015; Ree et al., 2019; Sato et al., 2020; Wu et al., 2015). Notably, to the best of our knowledge, no study has yet investigated implicit hedonic facial reactions to different types of rewards after pharmacological drug challenge in humans.

To fill this knowledge gap, we pharmacologically manipulated the dopaminergic and opioidergic systems in humans via oral administration of the highly selective D2/D3 dopamine receptor antagonist amisulpride (400 mg), the non-selective opioid receptor antagonist naltrexone (50 mg), or placebo, in a randomized, double-blind, between-subject design in 131 healthy volunteers (group sizes were 42, 44, and 45, respectively, for amisulpride, naltrexone, and placebo), and investigated the effects with a recently developed experimental paradigm (Korb et al., 2020), in which reward processing is operationalized similarly to animal research. Explicit subjective ratings of wanting and liking, physical effort (squeezing of an individually thresholded hand-dynamometer to obtain rewards) and implicit hedonic reactions (fEMG) during anticipation and consumption of primary social and nonsocial rewards of similar magnitude were obtained on a trial-by-trial basis (Figure 1). Sweet milk with different concentrations of chocolate flavour served as nonsocial food rewards. Gentle caresses to the forearm, delivered by a same-sex experimenter at different speeds, resulting in different levels of pleasantness (Ackerley et al., 2014; Löken et al., 2009; McGlone et al., 2014), served as non-sexual social rewards.

Figure 1 with 1 supplement see all
Main elements in each trial for the food (top) and touch (bottom) reward types.

Before the main task, participants experienced and ranked three food stimuli and three touch stimuli, based on liking (Figure 2). In the main task (here depicted), the highest-ranked (‘high’) reward was announced in half of the trials and the second-highest ranked (‘low’) reward was announced in the other half of trials. The probability of obtaining the announced reward was determined linearly by participants’ hand-squeezing effort, which was indicated in real-time. Participants knew that they would obtain the announced reward if they reached the top of the displayed vertical bar, which corresponded to their previously measured maximum voluntary contraction (MVC). The gained reward (which was either the one announced at the beginning of the trial, or – in the case of lower probability due to less squeezing – the least-liked ‘verylow’ reward) was then announced and delivered. To assess reward anticipation, EMG data was analyzed during the Pre-Effort anticipation period (3 s) at the beginning of the trial, when a possible reward was announced, as well as during the Post-Effort anticipation period (3 s announcement) preceding reward delivery. To investigate reward consumption, EMG data was analyzed during reward Delivery (5 s for food and 6.5 s for touch), and in the immediately following Relax phase (5 s). Rating slides stayed on screen indefinitely, or until participants’ button press. For a representation of all trial elements see Figure 1—figure supplement 1.

By adopting a translational approach, which makes human research comparable to animal research (e.g. measuring both real effort and hedonic facial reactions to primary rewards), we investigated two fundamental yet unresolved research questions: (1) to what extent do motivational and hedonic implicit and explicit responses during the anticipation and consumption of rewards rely on the dopaminergic and opioidergic systems in humans, and (2) do food and touch rewards share the same neurochemical basis in humans.

We made the following hypotheses based on the literature. First, because liking relies heavily on the opioidergic but not the dopaminergic system (Berridge and Kringelbach, 2015), subjective ratings of liking, and hedonic facial reactions during reward consumption, were expected to be lower after administration of the opioid antagonist naltrexone, compared to placebo, but not after administration of the dopamine antagonist amisulpride, particularly for the most preferred rewards (Eikemo et al., 2016; Smith and Berridge, 2007). Second, because wanting is believed to be regulated by the dopaminergic and opioidergic systems (Peciña and Berridge, 2013), we expected subjective ratings of wanting, and physical effort applied to obtain the preferred announced reward, to be lower after administration of both naltrexone and the D2/D3 receptor antagonist amisulpride. Third, because facial responses during reward anticipation – previously shown to occur to learned cues for rewards in rats (Delamater et al., 1986), and humans (Korb et al., 2020) – may reflect anticipated pleasure during a period commonly associated with wanting, they were expected to be affected by naltrexone, as well as by amisulpride, compared to placebo. Finally, based on fEMG results showing similar hedonic facial reactions to food and touch rewards, such as relaxation of the corrugator supercilii muscle and in some cases activation of the zygomaticus major muscle (Bershad et al., 2019; Korb et al., 2020; Mayo et al., 2018; Pawling et al., 2017; Ree et al., 2019; Sato et al., 2020), and on evidence from neuroimaging studies that supports the ‘common currency hypothesis’ of reward processing (Berridge and Kringelbach, 2015; Ruff and Fehr, 2014), we expected the same pattern of results for both types of rewards.

Results

Matching of drug groups

In order to rule out eventual group differences that were not of interest, we conducted a series of statistical tests to verify the matching of the three groups.

The three drug groups did not differ significantly in rankings of rewards before the main task, as shown by the absence of a significant Drug X Reward Level interaction for both food and touch rewards (Figure 2). Only a significant main effect of Reward Level was found for food (X2 (2)=78.1, p<0.001) and for touch rewards (X2 (2)=115.71, p<0.001), confirming the expected pattern of preferred food rewards (milk with greater chocolate content being preferred to milk with lower chocolate content), and of touch rewards (slower caresses being preferred to faster caresses).

Figure 2 with 1 supplement see all
Pre-experimental mean (SE) rankings of rewards by preference.

This ranking occurred just before the main task, which adapted to these preferences by using for each Reward Type (food, touch) the highest ranked stimulus as ‘high’ reward, the second-highest ranked stimulus as ‘low’ reward, and the lowest-ranked stimulus as ‘verylow’ reward.

In the main task, the level of reward (high, low, verylow) received in each trial depended on both the announcement cue at the beginning (high or low) and the force exerted to obtain it (verylow rewards were only obtained when participants exerted low effort, which linearly converted into low probability to obtain the announced reward). The number of trials in which high, low, and verylow rewards were obtained did not differ significantly across groups. Only a significant main effect of Reward Level was found (F(2, 763)=27.84, p<0.001), due to a greater number of high (M = 33.07, SD = 4.97) than low (M = 29.85, SD = 6.03) and verylow (M = 17.14, SD = 8.99) trials received, across all three drug groups and both reward types.

As expected, ratings of wanting and liking, as well as effort exerted to obtained the cued rewards, decreased over the course of the experiment due to habituation and fatigue (Figure 2—figure supplement 1). This decrease was similar across drug groups. Nevertheless, the covariate Block (recoded into first and second block separately by Reward Type) was included in the analyses of behavioral and fEMG measures (see below), to control for habituation and fatigue.

The three groups of participants did not differ significantly in their maximum voluntary contraction (MVC) of the hand dynamometer, which was measured right before the main task and at the end of the main task, nor in their positive and negative mood measured with the PANAS at time of pill intake and 3 hr later (all b < 0.6, all t < 0.8, all p>0.4; see Table 1). Finally, the three groups of participants did not differ significantly in terms of possible side effects, which were self-reported at time of pill intake and 3 hr later (see Table 1 for nausea scores).

Table 1
Participants’ characteristics across groups, as tested with linear regression (the t and p value refer to the main effect of Group).

BMI = Body Mass Index; MVC = Maximum Voluntary Contraction; PANAS = Positive and Negative Affective Schedule; M = Mean; SD = Standard deviation.

AmisulprideNaltrexonePlaceboGroup differences
N (male, female)42 (14, 28)44 (14, 30)45 (15, 30)
Age M (SD)23.7 (4.1)22.9 (2.8)23.1 (3.7)t = −0.73, p=0.46
BMI M (SD)22.7 (2.5)23.0 (2.3)22.2 (2.5)t = −0.99, p=0.32
MVC M (SD)211.9 (86.3)208.7 (81.8)215.3 (73.1)t = 0.19, p=0.85
PANAS pos T1 M (SD)30.5 (5.4)29.7 (7.3)29.4 (6.7)t = −0.8, p=0.42
PANAS neg T1 M (SD)12.1 (3.2)14.3 (7.5)11.5 (2.1)t = −0.7, p=0.52
PANAS pos T2 M (SD)27.1 (6.3)24.7 (8.0)26.7 (7.4)t = −0.3, p=0.80
PANAS neg T2 M (SD)10.1 (2.8)12.1 (5.5)10.5 (0.9)t = −0.5, p=0.58
Nausea T1 M (SD)1.05 (0.2)1.02 (0.1)1.00 (0.0)t = −1.5, p=0.14
Nausea T2 M (SD)1.00 (0.0)1.20 (0.6)1.00 (0.0)t = −0.1, p=0.93

Explicit measures: ratings of wanting, ratings of liking, and physical effort

Drug effects were investigated on ratings of wanting provided at the beginning of each trial, on physical effort to obtain an announced reward, and on ratings of liking provided after having obtained a reward. Interactions with the factor Drug were only found for physical effort.

Behavioral analyses on ratings of wanting (Figure 3—figure supplement 1, A–B) resulted in an expected significant main effect of Reward Level (F(1, 128.0)=119.28, p<0.001), due to higher ratings of wanting for high reward (M = 4.83, SD = 4.31) compared to low reward (M = 1.14, SD = 4.46), and a significant main effect of Block (F(1, 9653.7)=54.30, p<0.001) due to decreasing wanting from the first (M = 3.19, SD = 4.70) to the second block (M = 2.85, SD = 4.83). All other effects were not significant (all F < 2.3, all p>0.11).

To verify the lack of drug effects on ratings of wanting, we ran the same LMM using the full Bayesian method with the brms package (Bürkner, 2017). A normal prior with M = 0 and SD = 5 was defined for population-level (fixed) effects, and a half student-t prior with 3 degrees of freedom, M = 0, and scaling parameter = 4.7, was set for the standard deviation of subject-specific (random) effects. Results showed that neither amisulpride (βmean = 1, 95% Bayesian credible interval [−0.03, 2.07]), nor naltrexone (βmean = 0.07, 95% Bayesian credible interval [−0.94, 1.09]) had credible main effects on ratings of wanting (based on their respective 95% Bayesian credible interval crossing zero), nor did they interact with Reward Level or Reward Type (all βmean < 0.3, all 95% Bayesian credible interval crossing zero). In addition, a second Bayesian LMM was fitted without the main and interaction effects for the predictor Drug. The two models were compared using a Bayesian leave-one-out cross-validation (LOO-CV; Vehtari et al., 2017). This revealed a greater predictive ability for the null model (weight = 0.67, averaging via stacking of predictive distributions) than the full model (weight = 0.33), meaning that the null model is expected to be two times more accurate in predicting new data. One can thus conclude, that taking into account the drug administered does not improve the ability to predict participants’ ratings of wanting.

The LMM on effort (Figure 3—figure supplement 1, C–D) resulted in the expected significant main effect of Reward Level (F(1, 128.5)=54.41, p<0.001), due to stronger force applied for high (M = 80.49, SD = 22.35) than low rewards (M = 71.74, SD = 25.42); a significant main effect of Block (F(1, 7527.4)=175.49, p<0.001) due to decreasing effort from the first (M = 78.27, SD = 23.79) to the second block (M = 74.02, SD = 24.65); and a significant Reward Type X Drug interaction (F(2, 128.4)=4.71, p=0.01; Figure 3A) reflecting lower effort for food in the amisulpride (M = 74.98, SD = 26.57) and naltrexone (M = 73.51, SD = 24.43) groups compared to the placebo (M = 80.20, SD = 22.41) group, but similar force across drug groups for touch (amisulpride: M = 78.34, SD = 25.14; naltrexone: M = 73.78, SD = 23.15; placebo: M = 76.11, SD = 23.51). The Reward Level X Drug interaction (Figure 3B) reflected lower effort for low rewards in the amisulpride (M = 71.67, SD = 27.60) and naltrexone (M = 67.90, SD = 24.45) groups compared to Placebo (M = 75.65, SD = 23.60), but failed to reach significance (F(2, 128.5)=2.95, p=0.056). All other effects were not significant (all F < 0.9, all p>0.4).

Figure 3 with 1 supplement see all
Marginal means (and 95% CIs) for interactions with the predictor Drug in behavioral analyses.

Physical effort was lower in the amisulpride and naltrexone groups compared to placebo (A) for food but not touch rewards, and (B) non-significantly (p=0.056) for low but not high rewards. This suggests lower wanting after inhibition of both the dopaminergic and the opioidergic systems, specifically for high and low food rewards and for low-level rewards of both reward types. These null effects were confirmed with Bayesian analyses. See Figure 3—figure supplement 1 for all behavioral results.

The same LMM on ratings of liking (Figure 3—figure supplement 1, E–F) resulted in the main effect of Reward Level (F(1, 126.4)=150.55, p<0.001), with greatest liking of high rewards (M = 5.02, SD = 4.10), followed by low rewards (M = 1.79, SD = 4.15), and verylow rewards at the bottom (M = −1.19, SD = 3.89). Decrease of liking over time was shown by a significant main effect of Block (F(1, 9400.7)=129.40, p<0.001), due to a decrease in liking from the first (M = 2.90, SD = 4.61) to the second block (M = 2.25, SD = 4.79). All other effects were not significant (all F < 1.1, all p>0.34).

The lack of an effect of drug on liking was confirmed by fitting a Bayesian LMM (the same priors were set as for the models on wanting). Neither amisulpride (βmean = 0.78, 95% Bayesian credible interval [−0.20, 1.77]), nor naltrexone (βmean = 0.25, 95% Bayesian credible interval [−0.71, 1.18]) had main effects on ratings of liking, nor did they interact with Reward Level or Reward Type (all βmean < 0.30, all 95% Bayesian credible interval crossing zero). The full model had lower predictive ability (weight = 0.002) than the model without main and interaction effects of the predictor Drug (weight = 0.998), as shown with LOO-CV. These Bayesian analyses strengthen the view, already conveyed by the frequentist LMMs, that neither drug affected explicit wanting or liking of both types of rewards in this study.

Implicit measures: facial EMG

To investigate drug effects on reward anticipation and reward consumption, facial EMG was analyzed in relation to trial-by-trial subjective ratings and effort (as continuous predictors) in four periods of interest (see Figure 1). In short, the following results were found. In the Pre-Effort anticipation period (Figure 4) the corrugator was, as expected, relaxed for greater wanting and effort, and was more activated to food in the amisulpride and naltrexone groups compared to the placebo group. In the same time window, the zygomaticus muscle showed, as expected, stronger activation for greater wanting, however only in food trials. In the Post-Effort anticipation period, a non-significantly greater zygomaticus activation for greater wanting was found. In the Delivery phase, a Liking X Drug interaction was found in the zygomaticus muscle (Figure 5), reflecting the expected zygomaticus activation for greater liking in the placebo and (to a lesser extent) amisulpride group, while the opposite pattern of lower zygomaticus contraction for greater liking was found in the naltrexone group. Importantly, this interaction did not survive FDR correction (p=0.09) but seems credible based on a Bayesian LMM. Finally, in the Relax window immediately following reward administration, the corrugator significantly relaxed for the most liked food rewards, but not touch rewards.

EMG during Pre-Effort anticipation.

The corrugator relaxed in trials with greater wanting (A) and greater effort (B). Zygomaticus activation to food (C) showed the opposite pattern of activation for trials with greater wanting. This is in line with the literature and was also expected for touch. A significant Drug X Reward Type interaction was found in the corrugator analyses by wanting (D), and by effort (not shown). The anticipation of food rewards resulted in greater corrugator activation in the two drug groups compared to placebo, suggesting a reduction in hedonic facial responses. Plots A-C show marginal means and 95% CIs; plot D shows marginal means with standard errors and averages by subject.

Zygomaticus during Delivery (marginal means and 95% CIs).

A Liking x Drug interaction (not significant after FDR correction, p=0.09, but credible according to a Bayesian LMM) reflected zygomaticus activation for greater liking in the placebo group, and to a lesser extent also in the amisulpride group, but zygomaticus relaxation in the naltrexone group. This suggests that blocking of the opioidergic system resulted in an inverted effect of liking on zygomaticus activation, with less smiling for the most liked rewards.

Pre-Effort anticipation

For the corrugator muscle by Wanting, significant main effects of Reward Type (F(1, 267.9)=10.31, p=0.01), Wanting (F(1, 238.5)=7.75, p=0.01), and Block (F(1, 7867.4)=7.76, p=0.01) were found. Activation of the corrugator was greater for food (M = 116.35, SD = 84.72) than touch (M = 110.21, SD = 60.63) and decreased, as expected, with increasing ratings of wanting (slope b = −2.80; Figure 3A). A significant Drug X Reward Type interaction (F(2, 267.8)=4.08, p=0.04) reflected (Figure 4D) greater corrugator activation to food than touch in the amisulpride group (p=0.006; food: M = 119.12, SD = 95.82; touch: M = 109.46, SD = 53.44) and naltrexone group (p=0.001; food: M = 120.00, SD = 101.29; touch: M = 109.66, SD = 67.25), while the placebo group had similar activations across both reward types (p=0.66; food: M = 110.30, SD = 48.29; touch: M = 111.44, SD = 59.87). Corrugator activation to food was also significantly greater in the amisulpride and naltrexone groups compared to the placebo group (p=0.03 and. 01).

For the corrugator muscle by Effort, we found a significant main effect of Reward Type (F(1, 277.2)=11.04, p=0.008), with greater activation for food (M = 116.35, SD = 84.72) than touch (M = 110.21, SD = 60.63), a significant main effect of Effort (F(1, 207.9)=6.38, p=0.04) due to greater corrugator relaxation with increasing levels of Effort (b = -2.59; Figure 4B) a significant main effect of Block (F(1, 7662.4)=5.72, p=0.04) due to greater corrugator activation in the second (M = 115.55, SD = 87.5) compared to the first block (M = 110.98, SD = 56.4), and a significant Drug X Reward Type interaction (F(2, 277.2)=4.03, p=0.04) reflecting greater corrugator activation for food than touch in the amisulpride group (p=0.007; food: M = 119.12, SD = 95.82; touch: M = 109.46, SD = 53.44) and naltrexone group (p<0.001; food: M = 120.00, SD = 101.29; touch: M = 109.66, SD = 67.25), while the placebo group had similar activations across both reward types (p=0.72; food: M = 110.30, SD = 48.29; touch: M = 111.44, SD = 59.87). Corrugator activation for food was also significantly greater in the naltrexone group compared to the placebo group (p=0.03).

For the zygomaticus muscle by Wanting (random slopes for the Reward Type X Effort interaction were removed to allow model convergence), a significant main effect of Reward Type (F(1, 125.9)=13.78, p=0.001) was found, reflecting greater zygomaticus activation for food (M = 138.79, SD = 145.48) than touch (M = 122.98, SD = 130.67). Moreover, greater wanting predicted zygomaticus contraction in food trials (b = 5.6) but not in the touch trials (b = −2.73), as shown by a significant Reward Type X Wanting interaction (F(1, 2925.7)=6.62, p=0.04; Figure 4C). All other effects were not significant (all F < 0.8, all p>0.68).

For the zygomaticus muscle by Effort (random slopes for the Reward Type X Effort interaction were removed to allow model convergence), only a significant main effect of Reward Type was found (F(1, 129.2)=14.70, p=0.001), with greater zygomaticus activation to food (M = 138.79, SD = 145.48) compared to touch (M = 122.98, SD = 130.67). All other effects were not significant (all F < 1.2, all p>0.81).

Post-Effort anticipation

No significant effects were found for the corrugator muscle, neither by Wanting nor by Effort (all F < 1.7, all p>0.78).

For the Zygomaticus, a greater contraction for increasing levels of Wanting (b = 6.82) was observed, but the effect fell short of significance (F(1, 186.9)=6.55, p=0.08). All other effects were not significant (all F < 1.5, all p>0.51).

Reward delivery

Analysis of the corrugator resulted in a significant main effect of Reward Type (F(1,121.9) = 8.21, p=0.04), due to greater corrugator activation in response to food (M = 153.25, SD = 216.96) than touch (M = 117.23, SD = 298.61). All other effects were not significant (all F < 2.4, all p>0.48).

For the zygomaticus a significant main effect of Reward Type (F(1, 126.1)=77.97, p<0.001), was observed, due to greater zygomaticus activation in response to food (M = 264.83, SD = 233.77) than touch (M = 125.21, SD = 211.39). A Liking X Drug interaction, falling short of significance after FDR correction (F(2, 132.2)=4.22, p=0.09, Figure 5), was also found. Greater liking resulted, as expected, in greater zygomaticus activation in the placebo group (b = 5.74), and to a lesser extent also in the amisulpride group (b = 0.56). By contrast, the opposite pattern was found in the naltrexone group, with a negative slope (b = −7.86) that was significantly different from the placebo group (p=0.006) but not from the amisulpride group (p=0.09). The amisulpride and placebo group did not differ significantly between each other (p=0.3).

To further probe the Liking X Drug interaction, we also ran a Bayesian LMM with the same predictors (dummy coding was applied to the drug groups amisulpride and naltrexone, to compare them to placebo). A normal, and a half student-t prior were chosen for, respectively, population-level (fixed), and group-level (random) effects. Results confirmed a credible difference, compared to placebo, in the effect of liking on the zygomaticus activation in the naltrexone group (βmean = 13.62, 95% Bayesian credible interval [−23.37,–3.68]), but not in the amisulpride group (βmean = -4.92, 95% Bayesian credible interval [−14.99, 5.09]).

Relax phase

For the corrugator by Liking (the random slope for the Reward Type X Liking interaction was removed to allow model convergence), significant main effects of Reward Type (F(1, 155.0)=20.36, p<0.001) and Liking (F(1, 184.6)=12.41, p=0.001), and a significant Reward Type X Liking (F(1, 9231.8)=7.66, p=0.01) interaction were found. The interaction reflected a significant corrugator decrease with greater liking for food rewards (b = −23.2) but not for touch rewards (b = −5.6).

For the zygomaticus by Liking (the random slope for the Reward Type X Liking interaction was removed to allow model convergence), a significant main effect of Reward Type was found (F(1, 126.7)=144.25, p<0.001), reflecting greater zygomaticus contraction to food (M = 211.17, SD = 162.63) than touch (M = 132.05, SD = 208.32). Zygomaticus contraction was overall greater in the second (M = 166.22, SD = 148.85) compared to the first block (M = 175.58, SD = 225.67), as indicated by a significant main effect of Block (F(1, 9472.0)=7.44, p=0.03).

No significant main or interaction effects for the factor Drug were found in the CS and ZM data (all F < 1.7, all p>0.19).

Discussion

By adopting a newly developed experimental paradigm (Korb et al., 2020), in which reward processing is operationalised similarly to animal research, in combination with a dopaminergic and opioidergic drug challenge, we aimed to address two fundamental and as of yet unresolved research questions: (1) to what extent do motivational and hedonic responses in adult humans rely on shared or separate neurochemical systems, and (2) does the neurochemical basis of human reward processing differ for touch and food rewards.

Analyses of the behavioral (subjective wanting and liking ratings and effort) and physiological data (fEMG during anticipation and consumption of the reward) in relation to drug administration led to the following main results: (1) neither ratings of wanting nor liking were modulated by the pharmacological challenge (as confirmed with Bayesian analyses); (2) participants under dopaminergic or opioidergic antagonists produced significantly lower effort to obtain food rewards (Figure 3A), and non-significantly (p=0.056) lower effort to obtain low rewards of both touch and food (Figure 3B); (3) during the Pre-Effort anticipation of food, significantly higher corrugator activation was found in both the amisulpride and naltrexone groups (Figure 4D), suggesting lower hedonic anticipatory pleasure; and (4) during reward Delivery a Drug X Liking interaction was found (p=0.09 after FDR correction, but confirmed by Bayesian analyses), which reflected greater zygomaticus activation for liked rewards (and thus greater hedonic pleasure) in the placebo and to a lesser extent in the amisulpride groups, but weaker zygomaticus activation for liked rewards (less hedonic pleasure) in the naltrexone group (Figure 5). These findings are now discussed in relation to our main research questions.

Separate neurochemical systems underlie motivational and hedonic responses

In line with animal models and recent human pharmacological studies, indicating that both the dopaminergic and the opioidergic systems underlie the motivation to obtain rewards (Chelnokova et al., 2014; Peciña and Smith, 2010; Weber et al., 2016), we observed a similar effect of the D2/D3 antagonist amisulpride and the non-specific opioid receptor antagonist naltrexone on the effort produced to obtain the announced reward, resulting in a reduction of applied force. Notably, and differently from our original hypothesis, the effect was most pronounced for the second-preferred (low) rewards, as indicated by a Drug x Reward Level interaction (which however was not significant, p=0.056). A possible explanation for this finding is that our food stimuli did not vary in caloric content (i.e. the three reward levels were matched for fat and sugar). Therefore, individual preferences were derived from other mechanisms than energy value, possibly leading to a different effect of the drug on food reward processing (Barbano et al., 2009; Salamone et al., 2007). Another possible explanation is that high rewards are less susceptible to changes in their incentive salience when more options are available. Indeed, the majority of the studies in animals and humans have used only two reward levels, and have found that interference with dopaminergic or opioidergic transmission alters the outcome of cost/benefit analyses involving work-related response costs for the most valuable option (Salamone et al., 2007). Our finding suggests a similar shift in cost/benefit that is possibly sensitive to a different experimental set-up (Barbano et al., 2009).

A similar effect of both amisulpride and naltrexone during food anticipation was observed in the implicit measure of fEMG: corrugator activation was greater in both drug groups, compared to placebo. Because frowning typically reflects a more negative (or less positive) reaction to rewards and emotional stimuli (Fernández-Dols and Russell, 2017; Heller et al., 2011; Lang et al., 1993; Larsen et al., 2003), the observation that dopamine and opioid antagonists led to greater frowning might be interpreted as a reflection of less anticipated pleasure, independently of reward level, in these groups of participants.

During reward consumption, only the opioidergic antagonist naltrexone had an effect on implicit hedonic facial reactions. Lower zygomaticus activation for greater liking was found in the naltrexone group (with both frequentist and Bayesian LMMs), and this effect was significantly different from the placebo group, who instead showed the expected pattern of higher zygomaticus activation for greater liking. The amisulpride group showed the same pattern as the placebo group, although to a lesser extent. We interpret this finding as less smiling to liked rewards after administration of the opioid receptor antagonist naltrexone, as it parallels observations of fewer orofacial hedonic reactions to most preferred foods after opiodergic blockage in animals (Smith and Berridge, 2007).

Zygomaticus activation is also known to increase for highly negative stimuli, in addition to positive stimuli (Lang et al., 1999; Larsen et al., 2003), which could suggest that greater zygomaticus activation with decreasing liking in the naltrexone group is due to disgust expressions, or other negative facial reactions to the least-liked rewards, rather than to a reduction in positive hedonic responses (smiling). This, however, seems unlikely, as we only administered positive rewarding stimuli, and because drug groups did not differ in (1) initial reward preferences (Figure 2); (2) the number of high, low, and verylow rewards received; (3) the ratings of wanting and liking of these rewards; (4) the change of ratings of liking over time (Figure 2—figure supplement 1); and (5) amounts of nausea or other side-effects (Table 1). Moreover, this finding is unlikely to be explained by mouth movements related to food ingestion, because participants were instructed to swallow the food after the Delivery window, and no interaction with the factor Reward Type was found.

Taken together, and partially in line with the behavioral results (where the effects of the drugs were only observable for effort, but not for subjective wanting and/or liking ratings), fEMG data suggest a differential action of dopaminergic and opioidergic drug manipulation during the anticipation and consumption of rewards, with an effect of both drugs during the anticipation of rewards, but only of naltrexone during subsequent reward consumption. Interestingly, while the corrugator showed a general increase due to drug administration, zygomaticus activity was only affected in its relationship to subjective pleasure, and not in terms of overall activation. This differential pattern suggests that the two muscles, despite both tracking changes in hedonic value (Lang et al., 1993; Larsen et al., 2003), do not necessarily behave in a complementary, but rather in an independent way. This might also explain the heterogeneity of findings in previous studies, which often report an effect of reward valence on only one of the two muscles.

Interestingly, drug effects were found on effort levels and fEMG, but not on subjective ratings of wanting and liking. While this may come as a surprise, it is in line with several previous studies, which have reported either null or weak effects of pharmacological interventions on pleasantness likings of affective touch (Case et al., 2016; Ellingsen et al., 2014; Løseth et al., 2019; Trotter et al., 2016). However, several studies have reported significant effects of naltrexone/morphine on food liking/consumption (Bertino et al., 1991; Eikemo et al., 2016; Yeomans and Gray, 1996; Yeomans and Gray, 1997; but see Hetherington et al., 1991 for a null effect). One possibly relevant difference between some of the previous work, and our study, is that we kept calory intake constant across food stimuli, and thus across participants. More research will be needed to clarify if drug-induced changes in reward pleasantness can be reliably assessed with explicit measures (ratings) for some types of rewards (food), but instead require more implicit measures (facial responses, effort) for other types of rewards (affective touch).

Distinct neurochemical bases for food and touch rewards

Inclusion of both food and touch rewards allowed us to indirectly address the yet unresolved question (Ruff and Fehr, 2014), whether different types of rewards are processed by the same neurobiological systems (as proposed by the ‘common currency hypothesis’), or if representations coding for different rewards occur in distinct neural circuits, albeit on a common scale (Grabenhorst and Rolls, 2011). In particular social rewards, like affective touch, may constitute a separate class of stimuli, with a dedicated neural circuitry (Rademacher et al., 2010), which can be specifically impaired, for example in people with autism spectrum disorders (Chevallier et al., 2012; Haggarty et al., 2020). Although the magnitude of the two types of rewards in terms of subjective ratings and effort was carefully matched (Korb et al., 2020), most drug effects were either stronger or restricted to food trials, as indicated by significant Drug x Reward Type interactions for measures of effort to obtain the announced reward, and for corrugator activation during Pre-Effort anticipation. This suggests that the decision utility of touch and food rewards may not rely on the same neurochemical brain systems. However, fEMG responses to food were also stronger to begin with, as indicated by significant main effects of Reward Type for both muscles during the Pre-effort anticipation, Delivery, and Relax analysis windows. This might explain why only reactions to food rewards were modulated by opioidergic and dopaminergic antagonists. Another possible explanation for the less pronounced drug effects for touch is that responses to social rewards, including touch, might also depend on oxytocin and serotonin, in addition to dopamine and opioids (Fischer and Ullsperger, 2017; Tang et al., 2020; Walker and McGlone, 2013). This is also suggested by the finding of higher pleasantness ratings and greater zygomaticus activation to touch after administration of 3,4-methylenedioxymethamphetamine (MDMA), a drug that modulates serotonin, dopamine, and possibly oxytocin levels (Bershad et al., 2019; de Wit and Bershad, 2020).

Of note, drug effects on the activity of the zygomaticus muscle during reward consumption were similar for touch and food rewards, as indicated by the absence of a Drug X Reward Type interaction during the Delivery and Relax periods. To the best of our knowledge, this is the second study (after Bershad et al., 2019) to report a pharmacological modulation of hedonic responses to experienced touch (for a weaker effect see also Case et al., 2016). Major differences in our study compared to previous work (Ellingsen et al., 2014; Løseth et al., 2019) are the delivery of affective touch with the hand instead of a brush (Ellingsen et al., 2014 included touch by hand but wearing a glove), and allowing participants to select their preferred touch speed. Regarding the way touch was delivered, it is possible that the social saliency of the touch stimuli delivered through direct skin contact was enhanced, compared to when the touch is delivered with a brush, allowing us to detect subtle effects of the drug. Regarding the selection of the preferred touch speed, even if the majority of our subjects selected the slower speed as the preferred one, inter-individual differences were observed, and the implementation of a task that could account for those, may have helped to detect the effect of the drug. Further studies should investigate how such factors modulate drug responses to perceived affective touch.

The current study is characterized by a number of limitations. First, only two types of stimuli (food and touch) were used to define social and non-social rewards, and only two neurochemical systems (dopaminergic and opioidergic) were challenged. However, from the animal and human literature, we know that other systems (e.g. endocannabinoids, orexin, benzodiazepine, etc.) also contribute to the motivational and hedonic components of reward processing (Berridge and Kringelbach, 2015). Future studies should therefore broaden the neuropharmacological investigation of social vs. nonsocial reward processing, by using other compounds and different rewarding stimuli, to allow for the generalizability of the findings to other social and nonsocial rewards, and to better understand their neurochemical basis. Furthermore, computational approaches will most likely be useful to reveal hidden psychological states subtending motivation and experienced pleasure, allowing to refine how drug administration acts on these two components (Meyer et al., 2019).

Second, we used a relatively low dose of amisulpride (400 mg). Amisulpride can both increase dopaminergic neurotransmission by blocking presynaptic autoreceptors when given at low doses (50–300 mg), and decrease it by blocking postsynaptic D2/D3 receptors when given at higher doses of 400–1200 mg (Racagni et al., 2004; Schoemaker et al., 1997). The used dose of 400 mg is at the lower end of postsynaptically active high doses (Rosenzweig et al., 2002) and occupies ∼70% of D2 receptors when given for 2 weeks (Meisenzahl et al., 2008). We decided on 400 mg of amisulpride based on previous studies in humans (e.g. Weber et al., 2016), to obtain postsynaptic dopaminergic effects, while ensuring the safety and well-being of our participants, as well as allowing both participants and experimenters to remain in the dark about the type of compound or placebo administered (double blinding). The effects we observed (e.g. less effort) are in line with amisulpride’s antagonistic action on postsynaptic D2/D3 receptors. Upon availability of drug compounds that more strongly modulate the dopaminergic systems with minimal side effects, future studies should, however, explore dose-dependent changes in human reward processing.

Third, we used a cross-sectional design for drug/placebo administration. A within-subjects design would certainly have resulted in greater statistical power. However, this would have come with the cost of even greater habituation to the rewards.

Fourth, the study suffered from a lack of power to detect small effects. We had modeled the sample size on a previous study using the same drugs and doses (Weber et al., 2016). However, Weber et al., 2016 only found relatively small drug effects, and several other studies have failed to show effects of a pharmacological modulation of the opioid, serotonin, or oxytocin systems on the liking of affective touch (Ellingsen et al., 2014; Løseth et al., 2019; Trotter et al., 2016). This reveals the difficulty of uncovering the neurochemical basis of reward processing in humans and suggests that larger sample sizes should be used in future pharmacological studies to investigate the neurochemical bases of touch and other rewards.

Conclusion

We report pharmacological evidence in healthy human volunteers, across several measures including the monitoring of facial expressions with fEMG, about the role of the dopaminergic system for the motivational component and of the opioidergic system for both motivational and hedonic components of reward processing. The effort to obtain a reward and valenced facial reactions during reward anticipation were both modulated by the administration of dopaminergic or opioidergic antagonists. By contrast, facial reactions during reward experience were only altered by the opioidergic antagonist, suggesting neurochemical differences underlying hedonic expressions during anticipation and experience of pleasure. Explicit ratings of reward wanting and liking were not modulated by either drug. This constitutes the first demonstration of this kind in adult humans, using an operationalization of reward closely resembling previous animal research, and it suggests that the neurochemical regulation of pleasure (as indicated by hedonic facial reactions) is phase-specific, depending on whether the reward is anticipated or experienced. The finding that most drug effects were either stronger for, or restricted to, food trials may indicate different neurochemical brain mechanisms for social and nonsocial rewards. This point however requires further investigation via brain imaging or more direct measures of brain activity in addition to pharmacological challenges tailored to investigate the role of different neurochemical systems in the processing of social versus nonsocial rewards.

Materials and methods

Subjects

Based on previous work that had used the same compounds and doses (Weber et al., 2016), we aimed at collecting data from 40 participants per group or more. The final study sample included 131 volunteers (88 females) aged 18–35 years (M = 23.3; SD = 3.5). In the amisulpride group, blood concentrations of the drug (measured 5 hr after intake) were in or above the therapeutic range (blood samples missing for six people). Specifically, the minimum was 212 ng/mL, and 19 participants were above 604 ng/mL. All participants reported being right-handed, to smoke less than five cigarettes daily, to have no history of current or former drug abuse, to like milk and chocolate, not to suffer from diabetes, lactose intolerance, lesions or skin disease on the left forearm, and to be free of psychiatric or neurological disorders. Participants’ average Body Mass Index (BMI) was 22.6 (SD = 2.5, range 17.7–29.3). To reduce the chances that social touch would be perceived as a sexual reward, the touch stimulation was always carried out by a same-sex experimenter (see Procedure), and only participants who reported to be heterosexual were included. The study was approved by the Ethical Committee of the Medical University of Vienna (EK N. 1393/2017) and was performed in line with the Declaration of Helsinki (World Medical Association, 2013). Participants signed informed consent and received monetary compensation of 90€.

Stimuli

Request a detailed protocol

Three stimuli with identical fat and sugar content (1.5 g fat, 10 g of sugar per 100 g) were used as food rewards: milk, chocolate milk, and a 4:1 mix of milk and chocolate milk. Tap water served for rinsing at the end of each trial. The initial stimulus temperature of these liquids was kept constant (~4°C) across participants. Stimulus delivery was accomplished through computer-controlled pumps (PHD Ultra pumps, Harvard Apparatus) attached to plastic tubes (internal ø 1,6 mm; external ø 3,2 mm; Tygon tubing, U.S. Plastic Corp.), which ended jointly on an adjustable mount positioned about 2 cm in front of the participant’s mouth. In each trial, 2 mL of liquid was administered for 2 s. Overall, including stimulus pretesting (see Procedure), participants consumed 196 mL of liquids, composed of 98 mL of water, and 98 mL of sweet milk with different concentrations of chocolate aroma (depending on effort, see below).

Touch rewards consisted of gentle caresses over a previously-marked 9-cm area of the participant’s forearm (measurement started from the wrist towards the elbow). Three different caressing frequencies, chosen based on the literature and pilot testing, were applied for 6 s by a same-sex experimenter: 6 cm/s, 21 cm/s, and 27 cm/s. To facilitate stroking, the stimulating experimenter received extensive training and, in each trial, heard rhythmic sounds, indicating the rhythm for stimulation, through headphones.

EMG

Request a detailed protocol

After cleansing of the corresponding face areas with alcohol, water, and an abrasive paste, reusable Ag/AgCl electrodes with 4 mm inner and 8 mm outer diameter were attached bipolarly according to guidelines (Fridlund and Cacioppo, 1986) on the left corrugator supercilii (corrugator) and the zygomaticus major (zygomaticus) muscles. A ground electrode was attached to the participants' forehead and a reference electrode on the left mastoid. The EMG data were sampled at 1200 Hz with impedances below 20 kOHM using a g.USBamp amplifier (g.tec Medical Engineering GmbH) and the software Matlab (MathWorks, Inc).

Procedure

Request a detailed protocol

A monocentric, randomized, double-blind, placebo-controlled, three-armed study design was used. The study took place in the Department of Psychiatry and Psychotherapy at the Medical University of Vienna. Participants visited the laboratory for the first visit (T0) in which they received a health screening, followed by a second visit (T1) that included oral drug intake and the experiment described here. Pharmacological dosage, and length of waiting time after drug intake (3 hr), were modeled on previous work (Weber et al., 2016), and on the drug’s pharmacodynamics. Amisulpride reaches the first peak in serum after 1 hr, and a second (higher) peak after approximately 4 hr. The elimination half-life is 12 hr (Rosenzweig et al., 2002). At doses of 400 mg or higher, amisulpride acts as a postsynaptic D2/D3 receptor antagonist and thus results in lower dopaminergic action (Racagni et al., 2004; Schoemaker et al., 1997). Naltrexone reaches maximal concentration in plasma after 1 hr, has an elimination half-life in plasma of approximately 4 hr, and is completely cleared from plasma after 96 hr (Meyer et al., 2019). Importantly, up to 90% of mu-opioid receptors in the brain remain blocked by naltrexone after 48 hr, and partial receptor blockade could be shown up to 168 hr after intake (Lee et al., 1988).

Participants came to T1 with an empty stomach (it was morning, and they had been instructed not to eat in the preceding 6 hr), filled out the PANAS questionnaire, tested negative (or were excluded) on a urine drug screen sensitive to opiates, amphetamine, methamphetamine, cocaine (among other things), and then received a capsule filled with either 400 mg of amisulpride (Solian), 50 mg of naltrexone (Dependex), or 650 mg of mannitol (sugar) from the study doctor. All capsules looked identical from the outside, and neither participants nor the experimenters were informed of their content. Drug intake was followed by a waiting period, EMG preparation, and task instructions.

The experiment comprised two tasks following procedures described elsewhere (Korb et al., 2020). The main task started 3 hr after pill intake. Participants were seated at a table and comfortably rested their left forearm on a pillow. A curtain blocked their view of the left forearm and the rest of the room. This was particularly relevant for touch trials, in which one of two same-sex experimenters applied the touch rewards to the participant’s left forearm. Two experimenters were always present during testing, to limit the influence of participants’ experimenter preferences, and to allow participants to better concentrate on the (touch) stimuli.

Participants first completed a short task, in which they experienced and individually ranked three food rewards, and separately three touch rewards, presented randomly in sets of three of the same reward type. In the main task, which started 3 hr after pill intake, the previously most liked stimuli were used as ‘high’ rewards, the stimuli with medium liking as ‘low’ rewards, and the least liked stimuli were used as ‘verylow’ rewards. To calibrate the dynamometer, the MVC was established right before the short task, by asking participants to squeeze the dynamometer (HD-BTA, Vernier Software and Technology, USA) with their right hand as hard as possible three times, each lasting 3 s . The average MVC (peak force in newtons across all three trials) was 212 (SD = 80.4) and did not differ significantly between drug groups, as tested by linear regression (β = 1.6, SE = 8.68, t = 0.19, p=0.85).

After calibration of the dynamometer, EMG electrodes were attached, participants received detailed instructions, and completed four practice trials (two per reward type). The main task included four experimental blocks with 20 trials each. Each block contained either food or touch trials, and the blocks were interleaved (ABAB or BABA) in a counterbalanced order across participants. Each trial included the following steps (Figure 1; see Figure 1—figure supplement 1 and Supporting Information for all elements of a trial): (1) a picture announcing the highest possible reward (high or low, 3 s), (2) a continuous scale ranging from ‘not at all’ to ‘very much’ to rate (without time limit) wanting of the announced reward (ratings were converted to a Likert scale ranging from −10 to +10), (3) a 4 s period of physical effort, during which probability of receiving the announced reward was determined by the amount of force exerted by squeezing the dynamometer with the right hand, while receiving visual feedback (sliding average of 1 s, as percentage of the MVC), (4) a picture announcing the obtained reward (3 s for food, 7.3 s for touch), which could be high, low, or – if insufficient effort had been exerted – verylow (the greater participants’ effort, the higher the probability of obtaining the announced reward), (5) a phase of reward delivery (2 s for food, 6.5 s for touch – this difference in timing was necessary to obtain sufficiently long tactile stimulation, while keeping the overall trial duration similar across reward types), (6) for food trials instructions to lean back and swallow the obtained reward (duration 3 s), (7) a relaxation phase (5 s), and (8) a continuous scale to rate the liking of the obtained reward. In food trials, participants then received water for mouth rinsing. In both reward types, trials ended with a blank screen for 3–4 s. The last four trials in each block did not require pressing of the dynamometer. These trials were added to the design, in case participants would never press at all, which did not happen for any participant. Trials without pressing were kept in the data, as removing them from analyses reduced power but did not change the pattern of results. After each block participants were allowed to take a short break.

Both tasks were run on a desktop computer with Windows seven using MATLAB 2014b and the Cogent 2000 and Cogent Graphics toolboxes and presented on an LCD monitor with a resolution of 1280 × 1024 pixels. The positive and negative affect schedule (PANAS; Watson et al., 1988), and a questionnaire assessing nausea and 50 other side effects, was filled out twice at the main laboratory visit: just before pill intake, and 3 hr later. Levels of amisulpride (ng/mL) were measured in blood samples taken 5 hr after pill intake (after both tasks).

Analyses

Request a detailed protocol

Data and analysis scripts are available online (https://osf.io/vu8dz). Group comparisons for age, BMI, MVC, PANAS scores, and side effects, were made with linear regressions using the lm() function. Differences in the ranking of rewards across drug groups were tested with separate ordinal regressions by Reward Type (food, touch), using the package ordinal.

All other analyses were done with linear mixed-effects models (LMMs), fitted through restricted maximum likelihood (REML) estimation, using the lmer() function of the lmerTest package in R (which adds p values to the lme4 output; Bates and Maechler, 2014; R Development Core Team, 2019), and with Helmert contrast coding. In comparison to ANOVAs, LMMs reduce Type-I errors and allow for a better generalization of findings (Judd et al., 2012). To control for the effect of time – possibly inducing fatigue and/or habituation (Figure 2—figure supplement 1) – the four blocks were recoded to two blocks by Reward Type and entered as covariates to the LMMs. Figures (except Figure 1) were created in R using the packages ggplot2, ggpirate, and cowplot.

Behavioral data were analyzed in the following manner. Outlier trials were defined as those with a rating of wanting, rating of liking, or amount of exerted force, which was greater/smaller than the subject’s mean +/- 2 times the subject’s standard deviation. This led to an average rejection of 6.56 trials per participant (SD = 3.71). The total number of excluded trials did not differ significantly between groups (t(133) = −1.28, p=0.20). For each behavioral dependent variable (ratings of wanting and liking, and effort), a LMM was fitted with the fixed effects Reward Type (food, touch), Reward Level (high, low, verylow), Drug (amisulpride, naltrexone, placebo), their interactions, and with Block (first, second) as a covariate. Categorical predictors were centered through effect coding, and by-subject random intercepts and slopes for all within-subjects factors and their interactions were included as random effects (unless the model did not converge, in which case the random-effects structure was gradually simplified, e.g. by first dropping the interaction among within-subjects factors). Type-III F-tests were computed with the Satterthwaite degrees of freedom approximation. We report all statistically significant (p<0.05) effects, and non-significant effects with p<0.1 that are of interest because related to the main hypotheses, as Anova() outputs. Model tables showing all fixed and random effects can be found in the Supporting Information.

Due to technical failure, one participant lacked EMG data entirely, and another participant lacked the EMG for half of the trials. The EMG data were pre-processed in Matlab R2018a (www.themathworks.com), partly using the EEGLAB toolbox (Delorme and Makeig, 2004). A 20 to 400 Hz bandpass filter was applied, then data were rectified and smoothed with a 40 Hz low-pass filter. Epochs were extracted focusing on periods of reward anticipation (Pre-Effort and Post-Effort anticipation) and reward consumption (Delivery and subsequent Relax). EMG was averaged over time-windows of one second, with exception of the 6.5-seconds-long period of touch Delivery, which was averaged over five windows of 1.3 s each, to obtain the same number of windows as for food delivery. We excluded for each participant trials on which the average amplitude in the baseline period (1 s during fixation) of the corrugator or zygomaticus muscles was lower than M−2*SD, or higher than M+2*SD (M = average amplitude over all trials' baselines for the respective muscle and participant). On average, this led to the rejection of 7.7% of trials per participant (SD = 2.5). EMG analyses were carried out in four periods of interest: Pre-effort anticipation during reward announcement at the beginning of each trial (3 s), Post-effort anticipation during the announcement of the gained reward (3 s), Delivery (5 s for food and 6.5 s for touch, both averaged to five 1 s time windows), and Relax (5 s). For each trial, values in these epochs were expressed as percentage of the average amplitude during the fixation cross at the beginning of that trial. For the Pre- and Post-Effort anticipation periods, separate LMMs were fitted by muscle, with the fixed effects Drug (amisulpride, naltrexone, placebo), Reward Type (food, touch), and either trial-by-trial Wanting or Effort (these were continuous predictors), and all interactions. During the Post-Effort anticipation period, participants could receive the information that they were going to obtain the verylow reward, to which the preceding ratings of wanting and effort did not apply. Because this may have been frustrating for participants, we also carried out analyses excluding trials, in which verylow rewards were obtained. As the results did not change, we kept all trials. For the Delivery and Relax periods, separate LMMs on all trials were fitted by muscle, with the fixed effects Drug, Reward Type, and Liking. In all LMMs Wanting, Effort, and Liking were centered and scaled by subject, and Block (first, second) was added as a covariate to control for the effects of fatigue or habituation. We controlled for the false discovery rate (FDR) associated with multiple testing of the EMG data using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). Model tables with un-corrected p-values can be found in the Supporting Information.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
    From Experienced Utility to Decision Utility
    1. KC Berridge
    2. JP O’Doherty
    (2014)
    In: P. W Glimcher, E Fehr, editors. Neuroeconomics (Second Edition). Academic Press. pp. 335–351.
    https://doi.org/10.1007/978-981-10-6439-5_2
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
    Duration of occupancy of opiate receptors by naltrexone
    1. MC Lee
    2. HN Wagner
    3. S Tanada
    4. JJ Frost
    5. AN Bice
    6. RF Dannals
    (1988)
    Journal of Nuclear Medicine : Official Publication, Society of Nuclear Medicine 29:1207–1211.
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
    R: A language and environment for statistical computing, version 2.6.2
    1. R Development Core Team
    (2019)
    R Foundation for Statistical Computing, Vienna, Austria.
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
    Neurochemical characteristics of Amisulpride, an atypical dopamine D2/D3 receptor antagonist with both presynaptic and limbic selectivity
    1. H Schoemaker
    2. Y Claustre
    3. D Fage
    4. L Rouquier
    5. K Chergui
    6. O Curet
    7. A Oblin
    8. F Gonon
    9. C Carter
    10. J Benavides
    11. B Scatton
    (1997)
    The Journal of Pharmacology and Experimental Therapeutics 280:83–97.
  68. 68
    Reward Brain Mapping
    1. W Schultz
    (2015)
    In: A. W Toga, editors. An Encyclopedic Reference. Academic Press. pp. 643–651.
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83

Decision letter

  1. Mary Kay Lobo
    Reviewing Editor; University of Maryland, United States
  2. Christian Büchel
    Senior Editor; University Medical Center Hamburg-Eppendorf, Germany
  3. Siri Leknes
    Reviewer
  4. Guillaume Sescousse
    Reviewer; INSERM, France

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This study sheds light into the neurochemical basis of human reward processing by providing a translation approach, inspired by three decades of animal studies, to study the neurochemical basis of wanting and liking rewards in humans. Of interest, this study shows that physical reactions that occur when anticipating or consuming a reward rely on dopamine and/or opioid systems. Further, these studies suggest that the type of reward, social or non-social, can involve different neurochemical systems.

Decision letter after peer review:

Thank you for submitting your article "Role of dopamine and opioids on facial reactions in anticipation and consumption of social and nonsocial rewards" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Christian Büchel as the Senior Editor The following individuals involved in review of your submission have agreed to reveal their identity: Siri Leknes (Reviewer #1); Guillome Sescousse (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

This study translates key findings on the dopamine and opioid neurochemical dissociation of liking and wanting in rodents, into healthy humans. Using a well-controlled protocol closely inspired from animal literature the authors show that dopamine receptor and opioid receptor antagonists impact subjective ratings and some measures of facial EMG activity during reward anticipation, whereas only a opioid receptor antagonist affected facial EMG activity during reward consumption. Overall this study provides a significant contribution to the understanding of the neurochemical modulation underlying different aspects of behavioral reinforcement in humans.

Revisions:

The reviewers conclude that the study is of interest for eLife and that additional experiments are not needed. However, the reviewers request substantial revisions focused on the analysis and interpretations of the study.

1) The authors modeled their sample size on a previous study that showed small drug effects (Weber et al., 2016). Please clearly acknowledge the lack of power to detect small effects and provide interpretations of the results relevant to these small effects.

2) Please clarify why the main tests for the relationship between implicit and explicit responses were conducted in the full sample, rather than first being demonstrated only in the placebo group. Additionally, please clarify why the results are presented as pooled means instead of showing the mean and variance data for each drug, reward type and reward level. On the basis of the interaction effects with p's lower than 0.05, it is concluded that "low rewards" are affected by opioid and dopamine D2/D3 antagonists. Please demonstrate that this result is indeed comparable for touch and taste rewards.

3) Please clarify if side effects of the drugs were measured? For instance, nausea could affect responses to food stimuli.

4) Outlier removal is commonly applied with psychophysiological data, but not for VAS ratings. Please clarify why the outlier values were removed for these measures in the present study. Additionally, do the results differ with and without the excluded datapoints?

5) Please provide clarity into the details of the task and the outcomes:

a)It appears that the 3 reward levels for food reward were milk, chocolate milk, and a 4:1 mix of milk and chocolate milk? Was the ranking of these 3 reward levels consistent across participants? It is plausible that some participants would prefer chocolate milk while others would prefer milk, and that the strength of this preference (i.e. how far apart liking ratings are) would widely vary between participants. It would be informative to report (in supplementary material) the pre-task liking ratings for the 3 reward levels, both for the food and touch rewards. In particular, it would be nice to show that these pre-task ratings did not differ between groups, as this would strengthen the interpretation of the Drug x Reward type interaction reported in Figure 2C as a genuine effect of amisulpride, rather than a mere effect of pre-existing differences between groups.

b) Expression "if squeezing was not sufficient" and "insufficient effort" might be misleading as it implies that a minimum amount of squeezing (i.e. a threshold) was required to obtain a reward. Please clarify that there is a linear (i.e. continuous) relationship between squeezing and the probability of gaining a reward.

c) Please clarify it there were as many high reward as low reward cues.

6) Based on previous literature, when is the typical peak of maximal efficiency for amisulpride and naltrexone, and does this peak coincide with the 3h delay used in the current study?

7) There are 5 dependent variables (effort, wanting and liking ratings, EMG of CS and ZM muscles) tested at 4 different times (Pre-effort anticipation, Post-effort anticipation, Delivery and Relax), as a function of 3 main conditions (Drug, Reward type, and Reward level). As a result, there is a large number of statistical tests being performed, thus a correction for multiple comparisons would be warranted (see e.g. Cramer et al., Psychon Bull Rev 2016).

8) Due to the repetitive nature of the ratings on every trial, while there's no incentive for the participants to provide faithful ratings, the current procedure may run the risk of producing stereotyped ratings with little variability. Could the authors demonstrate that this is not the case? Further, there are possible satiety effects (i.e. ratings going down gradually, especially for the food reward). Please clarify if this occurred.

9) The authors have demonstrated a dissociation between drug effects on wanting and liking measures. However, Point 2 (“does the neurochemical basis of human reward processing differ for social and nonsocial rewards”) is overstated. The results from one social and one nonsocial stimulus type is insufficient to allow generalization to a class of rewards. Moreover, several drug studies have reported being unable to alter liking and/or wanting measures of stroking touch. This includes morphine, naltrexone, tryptophan and oxytocin. To the reviewers' knowledge, the only drug that alters perception of stroking touch is MDMA. Please provide discussion of the current understanding of the difference between drug effects on food and touch reward.

10) The decreased effort levels observed for food rewards in the amisulpride and naltrexone groups compared to the placebo group convincingly argue in favor of a role of the dopamine and opioid systems in motivation. However, and in contrast to the hypothesis stated in the Introduction, the same pattern wasn't observed for touch rewards. This discrepancy limits the generalizability of the conclusion, and warrants discussion. This also applies to the effect of amisulpride and naltrexone on CS muscle contraction (Figure 4): the fact that these effects are restricted to food reward limits the generalizability of the conclusion that the dopamine and opioid systems are involved in anticipated pleasure.

11) Additionally, for Figure 4, please clarify why reward magnitude was not included as a fixed factor similar to other analyses? Might dopamine and opioid antagonism be dissociable effects in terms of their effect on reward magnitude? Similarly, for effort analysis.

12) In the Discussion the statement "the results speak for a genuine alteration of the incentive salience of the low reward, whose rewarding value approached that of the verylow reward, due to the pharmacological manipulation" might be overstated and not supported by the findings. Subjects were never asked to exert effort to pursue verylow reward, hence, the rewarding/motivational value of verylow reward is unknown and it is inaccurate to conclude that the low reward's "rewarding value approached that for verylow reward". Moreover, Figure 2C showed that subjects liked the low rewards much more than verylow rewards; although this is not motivational salience, it might be reasonable to expect them to be associated to some degree. An elegant way to extract these motivational values would be to apply computational modelling, which has been conducted in other similar effort-based decision-making tasks.

13) In the Discussion the findings do not appear to "corroborate the hypothesis that the dopaminergic system underlies the motivational but not the hedonic component of rewards, while the opioidergic system underlies both." Figure 2C shows that the liking result for placebo is similar to naltrexone and not amisulpride. Using placebo as the reference for comparison, the findings appear to show instead that dopamine antagonism has an influence on increasing hedonism, but opioid antagonism has no effect, which is surprising in the context of existing literature.

14) In the Discussion please clearly state how the results align or do not align with the hypotheses. For instance, the discussion of the naltrexone group's zygomaticus “liking” responses gives the appearance as if the result was expected. However, it is surprising that opioid blockade (1) did not reduce subjective liking for any stimuli, not even high-calorie foods, and (2) appears to have reduced not zygomaticus activity in general, but more specifically its association with ratings of liking.

15) Please provide in the Discussion an interpretation of the effects of naltrexone and amisulpride at the doses chosen.

16) The authors suggest that the CS and ZG measures are complementary, but perhaps they are actually more different than similar? CS assesses frowning, which indicates negativity, whereas ZG measures smiling, which indicates positivity. CS is a proxy for negative valence axis, while ZG measures only positive coordinates on the valence axis; neither is a continuous assessment across the whole spectrum from the negative to positive end. This has implications for the interpretations in the study. For example, the authors state that "greater frowning during reward anticipation might be interpreted as a reflection of less anticipated pleasure", but this could be more appropriately interpreted as greater frustration, which is different from positive anticipation (see e.g. Bremhorst et al., 2019, PMID: 31848389). To further this point, if CS and ZG are indeed complementary measures, one would also expect ZG effect during the Pre-Effort anticipation of food – but that is not seen. Similarly, for reward delivery, while there is a CS effect, there is no ZG effect. What is the association between the two measures? A more in-depth discussion would help to improve the interpretations of the study.

17) How do the authors interpret the finding that during reward delivery, the ZM-liking slope for amisulpride did not differ from placebo nor from naltrexone? Comparisons between naloxone and placebo were interpreted to mean opioid antagonism reduced smiling, but no discussion was made for dopamine antagonism.

18) It is unclear from the main text or Materials and methods what feature of the force-time curve is used as the proxy of effort exerted – later in the legend of Figure 3—figure supplement 1, it is stated that "exerted force is the maximum value reached in the 4-sec period". If that is correct, please state this more clearly within the manuscript and also explain, what is the rationale for adopting the peak in a single infinitestimal timepoint within the 4-sec window as the measure of exerted force? Based on experience with hand-dynamometers in physical effort paradigms, it is not easy to control the force one intends to exert. Hence, there is typically an initial sharp spike in the force-time curve where the peak occurs (but not necessarily reflecting what the subject intends to exert), before stabilizing to a plateau that more accurately reflects the subjects intended force exerted. Thus, it is more typical in other studies to require subjects to sustain their force for say, 2s out of a 4-s window, and then take that sustained force level as a proxy of effort exerted.

19) Further, how is force exerted converted to probability? What constitutes as insufficient effort? The probability component seem an unnecessary manipulation thus what was the authors' rationale for rewarding probabilistically, rather than delivering the reward as long as sufficient effort is exerted?

20) A substantial number of subjects have a mean force exertion much higher than 100% of their MVC, especially in the amisulpride group (Figure 3—figure supplement 1) – does that mean the baseline measurement of MVC is not accurate since it is not actually maximum? Alternatively, it might also reflect issues associated with taking peak at the spike of the force-time curve as the measure of force exerted.

21) The authors do not account for the effect of fatigue in their analyses, which is presumably substantial given the large number of trials they have to squeeze the hand-grip dynamometer. The authors might include a factor of block?

22) Please provide rationale behind the choice of variables in the LMM, which includes all fixed factors and their interactions, as well as subject random intercepts and random slopes for all within-subject factors and their interactions. This appears as overparameterization. It is not easy to detect overparameterization since each random effect uses only one df, but the estimation of variance among small numbers of groups can be numerically unstable. The approach of fitting a model as complex as possible is problematic as it makes it very easy to fit a model too complex for the data. If the model fails to converge, that is fine since inference is not possible. But if the model "works", then that risks false inference (see e.g. Zurr et al., 2010).

23) The authors considered only one possible model. From a statistical point of view, however, the full range of possible interactions, from the full model (including all possible interactions between factors) to just the main effects, should be considered and the best model analyzed according to e.g. AIC/BIC.

24) Please provide more details on how the LMM analyses were conducted. For example, did they use ML or REML, and why?

25) Please provide tables denoting all the parameter estimates, SE, F-statistic, p-value etc. for each LMM.

26) In Results paragraph one: There are 80 trials in total across four blocks, but the total number of rewards obtained do not seem to add up to close to 80? Please clarify that the mean rewards reported are collapsed across condition.

Additional major comments to improve the clarity of the study:

27) The Introduction is very long. A more concise Introduction would improve the overall clarity. Further, please address the following in the Introduction:

a) Wanting and liking have been linked primarily to dopamine and opioids, but it's useful to extend this rule of thumb to encompass evidence on endocannabinoids, orexin and even benzodiazepines in enhancing liking responses to a comparable degree as mu, kappa and delta opioid microstimulation. This perspective should also be considered in the conclusion section which should be rephrased to avoid the impression that you have tested every relevant neurotransmitter.

b) Please provide an explanation for the interpretation that opioid-induced increases in motivation for high-calorie foods as evidence that opioids primarily affect wanting through liking.

c) Please consider changing to past tense when reviewing the evidence from single/small numbers of studies. Also, the effects reported in Chelnokova et al., 2014 on wanting were not explicitly subjective, rather a count of button presses which prolonged or shortened the duration an image was shown on the screen. Other relevant references on opioids and liking/wanting in humans include Gospic et al., 2008, Psychopharmacol; Atlas et al., 2013; Eikemo et al., 2017; Chelnokova et al., 2016, SCAN. Importantly, Gospic, Atlas and colleagues tested both men and women, whereas the cited evidence from Leknes' lab and Büchel's erotic image study were conducted exclusively in men. (Note that the affective image stimuli used by Gospic, Atlas and colleagues could also be described as social, and studies from Harriet de Wit's lab also indicate effects of opioid drugs on social reward responses.)

d) The following claim is surprising given the abundance of behavioural economics and facial EMG/facial emotion recognition studies in the field: "human research has struggled to adopt an operationalization that resembles the one used in animal research, e.g. measuring behavior and facial reactions instead of relying on subjective verbal report". Could a better explanation be that researchers testing human participants have much less precise ways of altering and measuring activity of opioid, dopamine and other neurotransmitters?

e) fEMG results for positive stimuli often show either zygomaticus increases or corrugator decreases, in fact Ree et al., Mayo et al. and Bershad et al. all found only corrugator decreases during touch. A reanalysis of the data from Pawling et al. available at PloS ONE suggests that the zygomaticus effect no longer meets statistical significance when only the time receiving touch is considered, and the interval after touch is removed. Please rephrase the Introduction section that currently indicates that both muscle potentials are altered in such gentle touch paradigms.

28) The end paragraph of the Discussion provides clarity to some issues noted above. It would be helpful to provide this clarity early on as the comments in the early part of the Discussion section give an impression of overgeneralizing some of the findings.

29) In the interpretation of the statistical findings please provide careful descriptions of the data and not just the p values? This includes addressing the following:

a) Please go through the manuscript carefully to avoid terminology such as "did not differ", when what is meant is "did not significantly differ". It is only correct to say that two measures do not differ if they are identical.

b) It is recommended to moderate the statement that an influence of drug via mood can be excluded.

c) P values above the α level are not accurately described as "trends"; instead please describe the pattern of the data and state clearly that the required significance level was not reached.

d) “Importantly, this finding cannot be explained by mouth movements that might have occurred during food delivery (although instruction to swallow followed the Delivery window), as statistics did not reveal an interaction with the factor Condition.”: a null effect (interaction) is interpreted as evidence that a potential confound did not influence the results. This is inappropriate with frequentist statistics.

30) Please replace the CS and ZM abbreviations with corrugator and zygomaticus. CS is commonly used to denote conditioned stimulus thus causes some confusion, especially when used in the graphs. Further, RewardType gives the impression that it denotes drink versus touch, rather than reward level. In Discussion paragraph six, it is suggested to clarify the announcement of possible reward or similar, since participants only received the announced rewards they chose to work for. Further, please edit phrases including "the fact that" when describing the results.

31) Please provide acknowledgement of limitations in the study, which certainly exists e.g. cross-sectional design (rather than within-subjects). This can be noted when discussing the interpretation of the findings.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for re-submitting your article "Dopaminergic and opioidergic regulation during anticipation and consumption of social and nonsocial rewards" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Christian Büchel as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Siri Leknes (Reviewer #1); Guillaume Sescousse (Reviewer #2); Diego Pizzagalli and Yuen-Siang Ang (Reviewer #3).

While the manuscript has been improved, there are a few outstanding issues remaining.

Revisions:

The authors were very responsive and provided a thorough revision of their manuscript, which is a very solid and valuable contribution to the literature. However, there are still some concerns that should be addressed.

1) The corrugator and zygomaticus data is a strength of the manuscript, but the current manuscript emphasizes facial responses more than is warranted by the study design and the hypotheses. Please address the following:

a) Please specify the lack of significant reductions in subjective ratings of chocolate milk and stroking touch in the Abstract.

b) The phrasing in the Introduction still appears to indicate that animal studies provide the best answers, whereas human subjective report is less valuable. However, the animal studies are severely hampered by their use of an indirect measure of liking, which has not been generalized beyond food reward but instead their strength in animal models is the ability to manipulate specific circuits/neurons. Please edit the text to reflect these limitations and strengths in animal models

c) The numerous figures of fEMG data patterns do not seem warranted, i.e. do they present the answers to the most important questions? It seems as though all effects below p=0.1 are illustrated.

d) The comparable subjective wanting and liking ratings in the placebo, naltrexone and amisulpride are a main result and should be listed as such in the Abstract, Results and Discussion. Further, please clarify that face muscle responses to high-reward stimuli were also comparable after drugs. Perhaps equivalence testing or Bayesian stats could be used to address the credibility of these interesting null results?

e) Referring to reduced activity of a frowning muscle as an indication of “hedonic anticipatory pleasure” is unwarranted and likely to confuse readers with respect to what the data shows. The corrugator is not a good measure of pleasure. Similarly, the discussion of drug effects on “hedonic responses” to touch gives the impression that the drugs altered touch pleasantness, instead of a face muscle response, which was not measured in most other drug studies.

2) The presentation of the behavioral wanting data to illustrate the two significant two-way interactions is confusing. It's better to depict non-pooled data for each reward type and level, and use lines to indicate the significant interactions. It is challenging to determine from the figure whether there was an effect of drug for wanting of gentle caresses, for instance. Please address the following:

a) The authors now provide this information in the supplement. However, showing one chart per drug instead of one per reward type/level makes it hard to discern any drug-related differences. The x axis labeling in the supplementary file was also confusing. It is recommended to remove Figure 3 and replace it with the reorganized supplementary to clarify any drug differences.

3) Please change the terminology to accurately depict the data and statistics

a) If the authors use 0.05 as their α level it is incorrect to describe results above 0.05 as trends or marginal effects. The higher force levels should be described as non-significantly higher. Similarly, the legend of Figure 5 makes too strong a case given the data. 1. This terminology is statistically inappropriate and maintains some unnecessary ambiguity in the literature. Many experts in statistics have pointed out this problem in recent years (see e.g. Gibbs and Gibbs, 2015, https://academic.oup.com/bja/article/115/3/337/312358). It is encouraged that the authors remove these terms.

b) It is also recommended to use more accurate descriptors of group differences, given the between-groups design, i.e. use “higher” instead of “increased”, “lower” instead of “reduced” etc.

4) The Results section is very hard to read especially since there are multiple periods of time analyzed. Please explain in the beginning of each section what information should be gained from the analyzed data rather than providing some of this information is in the summary of EMG data. This would help the readers to better understand the data.

5) 50 mg naltrexone is not a low drug dose. It is the standard clinical dose and blocks 90-100% of mu opioid receptors. It is recommended that the authors revisit the PET literature and take into account that 50 mg naltrexone is considered sufficient to induce so-called “full” blockade of mu receptors, when interpreting their results.

https://doi.org/10.7554/eLife.55797.sa1

Author response

Revisions:

The reviewers conclude that the study is of interest for eLife and that additional experiments are not needed. However, the reviewers request substantial revisions focused on the analysis and interpretations of the study.

1) The authors modeled their sample size on a previous study that showed small drug effects (Weber et al., 2016). Please clearly acknowledge the lack of power to detect small effects and provide interpretations of the results relevant to these small effects.

We have added the following paragraph at the end of the Discussion: “… the study may have suffered from a lack of power to detect small effects. We had modelled the sample size on a previous study using the same drugs and doses (Weber et al., 2016). However, Weber et al., 2016, only found relatively small drug effects, and several other studies have failed to show effects of a pharmacological modulation of the opioid or oxytocin systems on liking of affective touch (Ellingsen et al., 2014; Løseth et al., 2019). This reveals the difficulty of uncovering the neurochemical basis of reward processing in humans, and suggests that larger sample sizes should be used in future pharmacological studies to investigate the neurochemical bases of touch and other rewards.”

2) Please clarify why the main tests for the relationship between implicit and explicit responses were conducted in the full sample, rather than first being demonstrated only in the placebo group.

Similar, albeit smaller, effects are also found when using placebo participants only. However, we have now opted for removing these analyses all together, as they were also contained in the later analyses that included the factor Drug, and thus they were redundant.

Additionally, please clarify why the results are presented as pooled means instead of showing the mean and variance data for each drug, reward type and reward level.

Assuming that the reviewer is referring to Figure 3 in the previous version of the manuscript: EMG by Drug was not shown because the factor Drug was not included in these initial analyses, which had the goal to investigate the relationship between each muscle and wanting, effort, and liking. As written above, we have now removed these analyses, as they were redundant with the ones also containing the factor Drug.

On the basis of the interaction effects with p's lower than 0.05, it is concluded that "low rewards" are affected by opioid and dopamine D2/D3 antagonists. Please demonstrate that this result is indeed comparable for touch and taste rewards.

The LMM on effort included the 3 predictors Drug, Reward Level, and Reward Type. It resulted in a Drug X Reward Level interaction (Figure 3B), and there was no statistically significant 3-way interaction with Reward Type. Based on this, we conclude that there is not sufficient statistical evidence to suggest that this interaction differs between Reward Types.

3) Please clarify if side effects of the drugs were measured? For instance, nausea could affect responses to food stimuli.

We appreciate the reviewer’s question, and are sorry for not having reported this in the previous version of the manuscript. A self-report questionnaire had been used to measure nausea and 50 other possible side effects on a 4-point Likert scale (with the anchors 1 = “not at all” and 4 = “very much”). The side-effects questionnaire was filled out by participants at time of pill intake (T1) and three hours later (T2). This side-effects data was missing for one participant at T1 and for four participants at T2.

The side-effects did not differ between groups, neither at T1, nor at T2. We have added this information for nausea to Table 1 in the manuscript, and in Author response image 1 we show all the side effects across the 3 groups at T2.

Author response image 1
Nausea and other possible side effects were measured through self-report (4 options: “not at all”, “slightly”, “moderately”, and “very much“) three hours after pill intake.

All reported side effects were generally low, and did not differ between groups. The list of side effects (from top left to bottom right in the figure; original in German) was: headache; unrest; nervousness; belly ache; abdominal cramps; nausea; vomiting; joint/muscle pain; weakness; anxiety; thirst; swindle; dizziness; chills; sweating; tiredness; tears; tachycardia; chest pain; shortness of breath; diarrhoea; rush; itch; loss of appetite; increased energy; depressed; irritability; hallucinations; confusion; persecution mania; disorientation; tremor; eye irritation/swelling; light sensitivity; eye pain/fatigue; color vision deficiency; ear problems; ear pain; blushing; nasal problems; sneezing; coughing; increased yawning; flatulence; dry mouth; cold arms/legs; hot flashes; increased appetite; visual disturbance; speech disturbance; tinnitus.

4) Outlier removal is commonly applied with psychophysiological data, but not for VAS ratings. Please clarify why the outlier values were removed for these measures in the present study. Additionally, do the results differ with and without the excluded datapoints?

The goal of artefact removal was to reduce variability. We have followed the same procedure for outlier removal as in the previous publication with the same task (Korb et al., 2020). Differences in results with and without outliers are small at best, as can be appreciated by the forest plots comparing the models in Author response images 2, 3 and 4.

Author response image 2
Author response image 3
Author response image 4

5) Please provide clarity into the details of the task and the outcomes:

a) It appears that the 3 reward levels for food reward were milk, chocolate milk, and a 4:1 mix of milk and chocolate milk? Was the ranking of these 3 reward levels consistent across participants? It is plausible that some participants would prefer chocolate milk while others would prefer milk, and that the strength of this preference (i.e. how far apart liking ratings are) would widely vary between participants. It would be informative to report (in supplementary material) the pre-task liking ratings for the 3 reward levels, both for the food and touch rewards. In particular, it would be nice to show that these pre-task ratings did not differ between groups, as this would strengthen the interpretation of the Drug x Reward type interaction reported in Figure 2C as a genuine effect of amisulpride, rather than a mere effect of pre-existing differences between groups.

The 3 levels of food rewards were exactly as described by the reviewer. Participants indicated their preferences at the beginning of the experiment (see Procedure section for a description of the short task during which participants experienced and ranked all stimuli).

The variability in stimulus preferences was small and did not differ across Drug groups. We have conducted additional analyses, the results of which are described in the manuscript:

“In order to rule out eventual group differences that were not of interest, we conducted a series of statistical tests to verify the matching of the three groups.

The three drug groups did not differ in rankings of rewards before the main task, as shown by the absence of a significant Drug X Reward Level interaction for both food and touch rewards (Figure 2). Only a significant main effect of Reward Level was found for food (Χ2 (2) = 78.1, p <.001) and for touch rewards (Χ2 (2) = 115.71, p <.001), confirming the expected pattern of preferred food rewards (milk with greater chocolate content being preferred to milk with lower chocolate content), and of touch rewards (slower caresses being preferred to faster caresses).”

Please see Figure 2 for participants’ initial reward preferences by Reward Type and Reward level.

Moreover, we dealt with eventual differences in participants’ preferences by (1) adapting the task to participants’ initial choices, and (2) using for the analyses of the EMG ratings of wanting and liking on a trial-by-trial basis, which allowed us to even track changes from initial preferences.

b) Expression "if squeezing was not sufficient" and "insufficient effort" might be misleading as it implies that a minimum amount of squeezing (i.e. a threshold) was required to obtain a reward. Please clarify that there is a linear (i.e. continuous) relationship between squeezing and the probability of gaining a reward.

We have modified the manuscript to clearly indicate that there was a linear relationship between effort and probability of obtaining the announced reward.

Legend of Figure 1: “The probability of obtaining the announced reward was determined linearly by participants’ hand-squeezing effort, which was indicated in real-time. Participants knew that they would obtain the announced reward if they reached the top of the displayed vertical bar, which corresponded to their previously measured maximum voluntary contraction (MVC). The gained reward (which was either the one announced at the beginning of the trial, or – in the case of lower probability due to less squeezing – the leastliked “verylow” reward) was then announced and delivered.”

“In the main task, the level of reward (high, low, verylow) received in each trial depended on both the announcement cue at the beginning (high or low) and the force exerted to obtain it (verylow rewards were only obtained when participants’ exerted low effort, which linearly converted into low probability to obtain the announced reward).”

“which could be high, low, or – if insufficient effort had been exerted – verylow (the greater participants’ effort, the higher the probability of obtaining the announced reward)”.

c) Please clarify it there were as many high reward as low reward cues.

Half of the trials contained cues for high rewards, and the other half for low rewards. We have modified parts of the manuscript, including legend of Figure 1, to make this clearer: “In the main task (here depicted), the highest-ranked (“high”) reward was announced in half of the trials, and the second-highest ranked (“low”) reward was announced in the other half of trials”

The number of high and low (or verylow) rewards received could vary between participants. As indicated in the text: “In the main task, the level of reward (high, low, verylow) received in each trial depended on both the announcement cue at the beginning (high or low) and the force exerted to obtain it (verylow rewards were only obtained when participants’ exerted low effort, which linearly converted into low probability to obtain the announced reward). The number of trials in which high, low, and verylow rewards were obtained did not differ across groups. Only a significant main effect of Reward Level was found (F(2, 763) = 27.84, p <.001), due to a greater number of high (M = 33.07, SD = 4.97) than low (M = 29.85, SD = 6.03) and verylow (M = 17.14, SD = 8.99) trials received, across all three drug groups and both reward types.”

6) Based on previous literature, when is the typical peak of maximal efficiency for amisulpride and naltrexone, and does this peak coincide with the 3h delay used in the current study?

We have added the following to the Procedure:

“Amisulpride reaches a first peak in serum after one hour, and a second (higher) peak after approximately four hours. The elimination half-life is of 12 hours (Rosenzweig et al., 2002). At doses of 400 mg or higher, amisulpride acts as a postsynaptic D2/D3 receptor antagonist, and thus results in reduced dopaminergic action (Racagni et al., 2004; Schoemaker et al., 1997). Naltrexone reaches maximal concentration in plasma after one hour, has an elimination half-life in plasma of approximately four hours, and is completely cleared from plasma after 96 hours (Meyer et al., 1984). Importantly, up to 90% of mu-opioid receptors in the brain remained blocked by naltrexone after 48 hours, and partial receptor blockade could be shown up to 168 hours (Lee et al., 1988).”

The drug doses and the time delay between drug ingestion and starting of the main task were chosen based on a previous study that combined 50 mg of naltrexone and 400 mg of amisulpride (Weber et al., 2016). They result in comparable neurochemical responses, that were well tolerated in healthy participants.

7) There are 5 dependent variables (effort, wanting and liking ratings, EMG of CS and ZM muscles) tested at 4 different times (Pre-effort anticipation, Post-effort anticipation, Delivery and Relax), as a function of 3 main conditions (Drug, Reward type, and Reward level). As a result, there is a large number of statistical tests being performed, thus a correction for multiple comparisons would be warranted (see e.g. Cramer et al., Psychon Bull Rev 2016).

We have taken the following steps to limit the risk of committing Type I errors, while retaining sufficient power and mitigating the risk of Type II errors.

First, we used linear mixed models (LMMs) including random slopes. This type of analyses is more conservative (and thus reduces the risk of Type I errors) than ANOVA.

Second, we now control for the false discovery rate (FDR) associated with multiple testing of the EMG data using the Benjamini-Hochberg method (Benjamini and Hochberg, 1995). This correction is also recommended by Cramer et al., 2016. We have indicated so in the Analyses section of the Materials and methods. To be noted that, as a result of the correction (and of the inclusion of the covariate Block to control for changes over time), the Liking X Drug interaction for the zygomaticus muscle changed from significant (p = .02) to now marginally significant (p = .09).

8) Due to the repetitive nature of the ratings on every trial, while there's no incentive for the participants to provide faithful ratings, the current procedure may run the risk of producing stereotyped ratings with little variability. Could the authors demonstrate that this is not the case? Further, there are possible satiety effects (i.e. ratings going down gradually, especially for the food reward). Please clarify if this occurred.

Figure 2—figure supplement 1 shows wanting, force, and liking of the 3 Reward Levels (high, low, and for liking also verylow) across the 2 Reward Types (Food, Touch) and the 3 Drug groups. Wanting, force, and liking decreased over the course of the experiment, in agreement with the literature. The fact that ratings and force were not constant throughout, seems to contradict the reviewer’s suggestion that ratings were stereotypical and non-faithful.

Importantly, we have controlled for changes over time by including the factor Block (recoded by order and reward type) as a covariate to all analyses.

9) The authors have demonstrated a dissociation between drug effects on wanting and liking measures. However, Point 2 (“does the neurochemical basis of human reward processing differ for social and nonsocial rewards”) is overstated. The results from one social and one nonsocial stimulus type is insufficient to allow generalization to a class of rewards. Moreover, several drug studies have reported being unable to alter liking and /or wanting measures of stroking touch. This includes morphine, naltrexone, tryptophan and oxytocin. To the reviewers' knowledge, the only drug that alters perception of stroking touch is MDMA. Please provide discussion of the current understanding of the difference between drug effects on food and touch reward.

We agree that our conclusions are limited to the here-used types of rewards, and that more research is required to verify them in different social and nonsocial rewards. We have added this limitation to the Discussion: “The current study is characterised by a number of limitations. First, only two types of stimuli (food and touch) were used to define social and non-social rewards, and only two neurochemical systems (dopaminergic and opioidergic) were challenged. […] Future studies should therefore broaden the neuropharmacological investigation of social vs. nonsocial reward processing, by using other compounds and different rewarding stimuli, to allow the generalizability of the findings to other social and nonsocial rewards, and to better understand their neurochemical basis.”

We have referred to the finding that MDMA administration increases ratings of pleasantness and zygomaticus activation to touch. This work is clearly relevant, as MDMA alters the levels of several neurotransmitters, such as serotonin, dopamine, noradrenaline, and possibly oxytocin: “This is also suggested by the finding of higher pleasantness ratings and greater zygomaticus activation to touch after administration of 3,4-methylenedioxymethamphetamine (MDMA), a drug that modulates serotonin, dopamine, and possibly oxytocin levels (Bershad et al., 2019; de Wit and Bershad, 2020).”

We have discussed our findings in relationship to existing drug challenges studies on stroking touch: “To the best of our knowledge, this is the second study (after Bershad et al., 2019) to report a pharmacological modulation of hedonic responses to experienced touch (for a trend like effect see also Case et al., 2016). Major differences in our study compared to previous work (Ellingsen et al., 2014; Løseth et al., 2019) are the delivery of touch with the hand instead of a brush (Ellingsen et al., 2014 included touch by hand but wearing a glove), and the initial selection of the preferred touch speed by the participants. Regarding the delivery human touch vs. brushing, it is possible that the social saliency of the touch stimuli delivered through direct skin contact was enhanced, allowing us to detect subtle effects of the drug. Regarding the selection of the preferred touch speed, even if the majority of our subjects selected the slower speed as the preferred one, inter-individual differences were observed, and the implementation of a task that could account for those, may have helped to detect the effect of the drug. Further studies should investigate how such factors modulate drug responses to perceived affective touch.”

10) The decreased effort levels observed for food rewards in the amisulpride and naltrexone groups compared to the placebo group convincingly argue in favor of a role of the dopamine and opioid systems in motivation. However, and in contrast to the hypothesis stated in the Introduction, the same pattern wasn't observed for touch rewards. This discrepancy limits the generalizability of the conclusion, and warrants discussion. This also applies to the effect of amisulpride and naltrexone on CS muscle contraction (Figure 4): the fact that these effects are restricted to food reward limits the generalizability of the conclusion that the dopamine and opioid systems are involved in anticipated pleasure.

We fully agree, that the results on effort and partly also on the EMG were either stronger for food, or only present for food trials. We have indicated this on several occasions in the Discussion, including the following section:

“Although the magnitude of the two types of rewards in terms of subjective ratings and effort was carefully matched (Korb et al., 2020), most drug effects were either stronger or restricted to food trials, as indicated by significant Drug x Reward Type interactions for measures of effort to obtain the announced reward, and for corrugator activation during PreEffort anticipation. […] This point however requires further investigation through brain imaging, or through more direct measures of brain activity, in addition to pharmacological challenges tailored to investigate the role of different neurochemical systems in the processing of social vs. nonsocial rewards.”

11) Additionally, for Figure 4, please clarify why reward magnitude was not included as a fixed factor similar to other analyses? Might dopamine and opioid antagonism be dissociable effects in terms of their effect on reward magnitude? Similarly, for effort analysis.

Reward magnitude was not entered in the statistical models for EMG analyses. Instead, the EMG data was analysed using the continuous predictors consisting in trial-by-trial ratings of wanting and liking, as well as trial-by-trial exerted effort. We consider this approach as superior to using the categorical predictor Reward Level (with 2 or 3 levels depending on the time window). Former Figure 4, now Figure 4D, only shows the reward type (food, touch) X drug interaction, because the wanting X drug and the 3-way reward type X wanting X drug interactions were not significant. Thus, to answer the reviewer’s question, we do not have any statistical grounds to argue that dopamine or opioid antagonists have dissociable effects on corrugator activation in the pre-effort anticipation period, depending on reward magnitude (i.e. reward level).

Behavioural data, on the other hand, was analysed with the categorical predictor RewardType, because it is not possible to do otherwise (i.e. you cannot use continuous ratings as both dependent variable and predictor).

12) In the Discussion the statement "the results speak for a genuine alteration of the incentive salience of the low reward, whose rewarding value approached that of the verylow reward, due to the pharmacological manipulation" might be overstated and not supported by the findings. Subjects were never asked to exert effort to pursue verylow reward, hence, the rewarding/motivational value of verylow reward is unknown and it is inaccurate to conclude that the low reward's "rewarding value approached that for verylow reward". Moreover, Figure 2C showed that subjects liked the low rewards much more than verylow rewards; although this is not motivational salience, it might be reasonable to expect them to be associated to some degree. An elegant way to extract these motivational values would be to apply computational modelling, which has been conducted in other similar effort-based decision-making tasks.

We agree that we cannot, based on effort alone, claim that the reward value of low rewards approached that of verylow rewards in the 2 drug groups. We have modified the corresponding paragraph and now discuss the finding in the following terms: “Notably, and differently from our original hypothesis, the effect was most pronounced for the second-preferred (low) rewards, as indicated by a significant Drug x Reward Level interaction. A possible explanation for this finding is that our food stimuli did not vary on caloric content (i.e. the three reward levels were carefully matched for fat and sugar). Therefore, individual preferences were derived from other mechanisms than energy value, possibly leading to a different effect of the drug on food reward processing (Barbano et al., 2009; Salamone et al., 2007). Another possible explanation is that high rewards are less susceptible to changes in their incentive salience, when more options are available.”

We also agree that it would be worthwhile to apply computational modelling to this data, which however is not part of our current skill set, and therefore cannot be included into the paper at this time. A sentence highlighting the added value of using computational models in pharmacological challenge studies on reward processing has been added to the Discussion: “Furthermore, computational approaches will most likely be useful to reveal hidden psychological states subtending motivation and experienced pleasure, allowing to refine how drug administration acts on these two components (Meyer et al., 2019).”

13) In the Discussion the findings do not appear to "corroborate the hypothesis that the dopaminergic system underlies the motivational but not the hedonic component of rewards, while the opioidergic system underlies both." Figure 2C shows that the liking result for placebo is similar to naltrexone and not amisulpride. Using placebo as the reference for comparison, the findings appear to show instead that dopamine antagonism has an influence on increasing hedonism, but opioid antagonism has no effect, which is surprising in the context of existing literature.

The new analyses (with block as covariate) did not confirm the initial findings of Druginduced differences in liking (Drug X Reward Type interaction not significant, p = .4). At the behavioural level, amisulpride and naltrexone only affected effort levels. We therefore feel more confident in claiming such differential effects of opioidergic and dopaminergic drugs on wanting vs. liking.

14) In the Discussion please clearly state how the results align or do not align with the hypotheses. For instance, the discussion of the naltrexone group's zygomaticus “liking” responses gives the appearance as if the result was expected. However, it is surprising that opioid blockade (1) did not reduce subjective liking for any stimuli, not even high-calorie foods, and (2) appears to have reduced not zygomaticus activity in general, but more specifically its association with ratings of liking.

We have added the following section to the Discussion: “Interestingly, drug effects were found on effort levels and fEMG, but not on subjective ratings of wanting and liking. While this may come as a surprise, it is in line with several previous studies, which have reported either null or weak effects of pharmacological interventions on pleasantness likings of affective touch (Case et al., 2016; Ellingsen et al., 2014; Løseth et al., 2019; Trotter et al., 2016). It thus appears that more implicit measures are required, to succesfully observe drug-induced changes in wanting and liking of at least these two types of rewards.”

The reviewer is right in pointing out that we had specified a reduction of hedonic facial reactions (i.e. less zygomaticus, more corrugator). Specifically, in the Introduction, we say: “subjective ratings of liking, and hedonic facial reactions during reward consumption, were expected to be reduced after administration of the opioid antagonist naltrexone, compared to placebo”. However, we realised that we had not well formulated this hypothesis, even though we had written about it elsewhere in the text. Indeed, the literature frequently suggests the effect of opioidergic modulation is observable on the most palatable food or most attractive faces (Eikemo et al., 2016; Smith and Berridge, 2007). We have therefore specified better the initial hypotheses, and we refer now to effects in “the most preferred rewards”.

The analyses of fEMG data performed with subjective ratings as continuous predictors allowed us to assess two things: (1) overall changes in the activity of the muscle (main effect) and (2) changes in the association between subjective ratings and facial activity (which is a proxy for assessing the effects of a drug on different levels of the reward). We observed that naltrexone indeed altered this association, with zygomaticus activating less to preferred than non-preferred rewards. We now discuss these findings based on the adjusted hypotheses: “Notably, and differently from our original hypothesis, the effect was most pronounced for the second-preferred (low) rewards, as indicated by a significant Drug x Reward Level interaction.”

15) Please provide in the Discussion an interpretation of the effects of naltrexone and amisulpride at the doses chosen.

We have added to the Discussion a section on limitations of this study. There, we also discuss the findings in relation to the drug doses used: “Second, we used low drug doses (50 mg of naltrexone, and 400mg of amisulpride). […] Upon availability of drug compounds that more strongly modulate the opioidergic and dopaminergic systems with minimal side effects, future studies should, however, explore dose-dependent changes in human reward processing.”

16) The authors suggest that the CS and ZG measures are complementary, but perhaps they are actually more different than similar? CS assesses frowning, which indicates negativity, whereas ZG measures smiling, which indicates positivity. CS is a proxy for negative valence axis, while ZG measures only positive coordinates on the valence axis; neither is a continuous assessment across the whole spectrum from the negative to positive end. This has implications for the interpretations in the study. For example, the authors state that "greater frowning during reward anticipation might be interpreted as a reflection of less anticipated pleasure", but this could be more appropriately interpreted as greater frustration, which is different from positive anticipation (see e.g. Bremhorst et al., 2019, PMID: 31848389). To further this point, if CS and ZG are indeed complementary measures, one would also expect ZG effect during the Pre-Effort anticipation of food – but that is not seen. Similarly, for reward delivery, while there is a CS effect, there is no ZG effect. What is the association between the two measures? A more in-depth discussion would help to improve the interpretations of the study.

The reviewer seems to raise two points: first, whether each muscle is valence specific (i.e. zygomaticus = positive, CS = negative) and second, whether the two muscles are complementary.

Regarding the first point, we disagree with the view that the zygomaticus and corrugator only indicate, respectively, positivity and negativity. Instead, pleasure and positive affect typically activate the zygomaticus and relax the corrugator, while negative affect activates the corrugator and relaxes the zygomaticus. This has been shown in response to pictures, sounds, and words covering a large valence spectrum from negative to positive (Lang et al., 1993; Larsen et al., 2003). There is no reason to expect different muscle responses to rewards than other affective stimuli. Importantly, the inverse linear relationship between corrugator activation and valence is not specific to negative affect, such as frustration. On the contrary, Larsen et al., 2003, found that, “positive and negative affect have reciprocal effects on activity over corrugator supercilii” (Abstract), and “even when negative affect is held constant at minimal levels, positive affect inhibits activity over corrugator supercilii”. In other words, corrugator activation does not only increase in the range of neutral-tonegative affect, but relaxes even more in the neutral-to-positive affect range. We have added the references of Lang et al., 1993 and Larsen et al., 2003 to the manuscript.

Certainly, it remains unclear why the drug effects on corrugator and zygomaticus were so different. We have added the following to the Discussion: “Interestingly, while the corrugator showed a general increase due to drug administration, zygomaticus activity was only affected in its relationship to subjective pleasure, and not in terms of overall activation. This differential pattern suggests that the two muscles, despite both tracking changes in hedonic value (Lang et al., 1993; Larsen et al., 2003), do not necessarily behave in a complementary, but rather in an independent way. This might also explain the heterogeneity of findings in previous studies, which often report an effect of reward valence on only one of the two muscles.”

Although it is true that our task could have induced some frustration when the cued reward was not obtained, this frustration must have been mild at best, as even in those trials a good-tasting reward (and not a punishment) was delivered. Moreover, the earliest time point at which frustration could have appeared in each trial is after (and not before) the physical effort, i.e. from the Post-Effort anticipation phase onwards. Therefore, the inverse linear relation between wanting/effort and corrugator activation in Pre-Effort anticipation period cannot be explained by frustration due to not obtaining the cued reward (since this was decided later in the trial). Alternatively, frustration may have increased over the course of the experiment, due to the repeated administration of a limited number of rewards, and could therefore have affected all time windows of analyses, including the Pre-Effort anticipation and Relax phases, for which we found changes in the corrugator. This is unlikely to have been the case, however, since in the now-reported analyses we controlled for the factor Block (and thus eventual time- and repetition-induced effects of habituation and fatigue).

As to the zygomaticus muscle, it is true that previous research has reported not only a linear, but also a quadratic effect of valence, reflecting greater zygomaticus contraction for both extremely positive and extremely negative affect (Lang et al., 1993; Larsen et al., 2003). Since we only administered positive, rewarding stimuli, we believe that it is unlikely that our effects on the zygomaticus reflected increased activation to very negative valence. However, it is true that we cannot fully exclude the possibility that the inverse relationship of valence and zygomaticus activation, which was found for the naltrexone group during reward delivery, is due to smirking or disgust reactions to the less-liked rewards. This remains unlikely, however, because the number of trials with high, low, and verylow rewards did not differ between drug groups, and because the ratings of wanting and liking of these rewards did not differ across drug groups, and finally because the change of ratings of liking over time was comparable across drug groups (as shown by results remaining when controlling for the factor block, see also Figure 2—figure supplement 1).

To explain why we interpret the reversed liking-to-zygomaticus pattern in the naltrexone group as a reduction of hedonic facial responses to the most liked rewards (and not as a reflection of negative facial responses to the least liked rewards), we have added the following paragraph to the Discussion:

“Does zygomaticus activation for less liked rewards in the naltrexone group reflect negative facial reactions? […] Nevertheless, it is important to keep in mind that the drug effects on zygomaticus activation during reward delivery were only marginally significant, and should therefore be interpreted with caution.”

17) How do the authors interpret the finding that during reward delivery, the ZM-liking slope for amisulpride did not differ from placebo nor from naltrexone? Comparisons between naloxone and placebo were interpreted to mean opioid antagonism reduced smiling, but no discussion was made for dopamine antagonism.

In the previous version of the manuscript, we had not discussed the effects of the dopamine-receptor antagonist amisulpride on ZM activation slope, because precisely the slope was not significantly different from the slope for the other groups. However, reanalysis (with FDR-corrections and the covariate Block) now result in a slightly different picture. As written in the text: “the interaction was explained by greater liking resulting, as expected, in greater zygomaticus activation in the placebo group (b = 5.74), and to a lesser extent also in the amisulpride group (b = 0.56). In contrast, the opposite pattern was found in the naltrexone group, with a negative slope (b = -7.86) that was significantly different from the placebo group (p = .006) and marginally significantly different from the amisulpride group (p = .09). The amisulpride and placebo group did not differ between each other (p = .3).”

Thus, as in our previous discussion, naltrexone seems to alter and reverse the typical link between liking and smiling, but amisulpride does not. In fact, the slopes for amisulpride and placebo are not significantly different from each other. But the slope of amisulpride, which is positive like the one for placebo, is marginally different from the naltrexone group.

18) It is unclear from the main text or Materials and methods what feature of the force-time curve is used as the proxy of effort exerted – later in the legend of Figure 3—figure supplement 1, it is stated that "exerted force is the maximum value reached in the 4-sec period". If that is correct, please state this more clearly within the manuscript and also explain, what is the rationale for adopting the peak in a single infinitestimal timepoint within the 4-sec window as the measure of exerted force? Based on experience with hand-dynamometers in physical effort paradigms, it is not easy to control the force one intends to exert. Hence, there is typically an initial sharp spike in the force-time curve where the peak occurs (but not necessarily reflecting what the subject intends to exert), before stabilizing to a plateau that more accurately reflects the subjects intended force exerted. Thus, it is more typical in other studies to require subjects to sustain their force for say, 2s out of a 4-s window, and then take that sustained force level as a proxy of effort exerted.

As written in subsection “Procedure” the force was displayed as percentage of the MVC, averaged with a sliding average of one second. The maximum force was also taken based on this sliding average. Therefore, we did not take an infinitesimal timepoint of exerted force, but the average of 1 second. This value was chosen based on pilot studies, and was also used for our previous publication (Korb et al., 2020).

19) Further, how is force exerted converted to probability? What constitutes as insufficient effort? The probability component seem an unnecessary manipulation thus what was the authors' rationale for rewarding probabilistically, rather than delivering the reward as long as sufficient effort is exerted?

Force (i.e. maximum in a 4-sec period, as percentage of MVC, and after passing a 1second sliding average) was converted linearly into probability to obtain the announced reward. An insufficient effort thus constitutes either no effort at all (resulting in zero probability of receiving the announced reward, and conversely in 100% probability of obtaining the verylow reward), or a relatively low effort (e.g. 10% of the MVC only results in a 10% probability of receiving the announced reward).

We chose to reward probabilistically based on effort, as we wanted a continuous measure of wanting. Had we instead delivered a reward based on a specific effort threshold, we probably would have obtained less variability in the amount of exerted effort between trials. Furthermore, the conversion to probability implicated a certain amount of uncertainty, which allowed us to induce an anticipated response in the post-effort anticipation (otherwise completely explained by the pre-effort anticipation).

A linear conversion of effort into the probability of obtaining a reward might not be the most frequently used paradigm, but it has been used before in the relevant literature (Lopez-Persem et al., 2017).

20) A substantial number of subjects have a mean force exertion much higher than 100% of their MVC, especially in the amisulpride group (Figure 3—figure supplement 1) – does that mean the baseline measurement of MVC is not accurate since it is not actually maximum? Alternatively, it might also reflect issues associated with taking peak at the spike of the force-time curve as the measure of force exerted.

There is unfortunately no perfect way of calibrating the dynamometer, explaining why forces > the MVC can sometimes be obtained. We have explicitly measured the MVC before explaining the task to participants, in order to keep them as naïve as possible about the role that their effort would later play in the task. Nevertheless, participants may sometimes press less than their maximum force during MVC measurement. For the experimenter, it is very hard to detect and correct this. This is a recurrent problem in this type of research.

Importantly, trials with effort equal to or greater than 100% of MVC had the same probability (i.e. 100%) of obtaining the announced reward, and this was explained to participants.

Again, we took the peak of effort in each trial but only after averaging with a 1-second sliding average.

21) The authors do not account for the effect of fatigue in their analyses, which is presumably substantial given the large number of trials they have to squeeze the hand-grip dynamometer. The authors might include a factor of block?

We have included the factor block (recoded into first and second block for each reward type) as a covariate to all analyses. Doing so, we made sure that the results remained significant while controlling for the effects of habituation and fatigue. We have also added Figure 2—figure supplement 1 to show wanting, liking, and effort across trials, Reward Types, and Reward Levels.

22) Please provide rationale behind the choice of variables in the LMM, which includes all fixed factors and their interactions, as well as subject random intercepts and random slopes for all within-subject factors and their interactions. This appears as overparameterization. It is not easy to detect overparameterization since each random effect uses only one df, but the estimation of variance among small numbers of groups can be numerically unstable. The approach of fitting a model as complex as possible is problematic as it makes it very easy to fit a model too complex for the data. If the model fails to converge, that is fine since inference is not possible. But if the model "works", then that risks false inference (see e.g. Zurr et al., 2010).

We have chosen the fixed effects based on the experimental design and hypotheses (for example, we explicitly wanted to investigate if EMG varies by Reward Type, Drug, and ratings of wanting/liking and effort). We have put all fixed effects as random slopes, as this is often advised (Barr et al., 2013). According to simulations, fitting a maximal model does indeed have a cost, which however is not the increase of Type-I errors, but rather the loss of power (Matuschek et al., 2017).

23) The authors considered only one possible model. From a statistical point of view, however, the full range of possible interactions, from the full model (including all possible interactions between factors) to just the main effects, should be considered and the best model analyzed according to e.g. AIC/BIC.

We have designed the experiment to test the effects of several predictors in interaction, and have therefore fitted models that included them all – up to a limit, because otherwise the models do not converge. We favour a theory/hypothesis-driven approach, while the reviewer seems to suggest a more data-driven approach. As far as we are aware, both are legitimate procedures.

24) Please provide more details on how the LMM analyses were conducted. For example, did they use ML or REML, and why?

As we have now specified in the text (under Materials and methods, Analyses), LMMs were fitted with REML, as it provides unbiased estimates of the random effect variance-covariance matrix (https://en.wikipedia.org/wiki/Restricted_maximum_likelihood), and it is the default used by both lmer() and lmerTest(). Using the anova() function to compare models would automatically refit them with ML.

25) Please provide tables denoting all the parameter estimates, SE, F-statistic, p-value etc. for each LMM.

We have added the tables for all the statistical models to the supplementary material (using the function tab_model from the package sjPlot). Please be aware that these do not include F values, dfs, and p values as described in the manuscript – those were obtained with the function anova(), as stated in the manuscript. Please be also aware, that the pvalues for fEMG provided in the manuscript were corrected for multiple comparisons using Benjamini-Hochberg FDR, while tables include uncorrected p values.

26) In Results paragraph one: There are 80 trials in total across four blocks, but the total number of rewards obtained do not seem to add up to close to 80? Please clarify that the mean rewards reported are collapsed across condition.

We thank the reviewer for pointing out that we had erroneously provided the number of trials by Reward Type and Reward Level (because the statistical model also included the predictor Reward Type). We have now modified the corresponding text passage accordingly. To clarify: there were 80 trials per subject, separated in 4 blocks of 20 trials each. Of these half (40) were social rewards, and the other half (40) included food rewards. Each reward condition included half (20) of the trials announcing a high reward, and the other half (20) announcing a low reward.

Across both Reward Types, the average numbers of high, low, and verylow rewards were, respectively, 33.07, 29.85, and 17.14.

Additional major comments to improve the clarity of the study:

27) The Introduction is very long. A more concise Introduction would improve the overall clarity.

The Introduction has been considerably shortened.

Further, please address the following in the Introduction:

a) Wanting and liking have been linked primarily to dopamine and opioids, but it's useful to extend this rule of thumb to encompass evidence on endocannabinoids, orexin and even benzodiazepines in enhancing liking responses to a comparable degree as mu, kappa and delta opioid microstimulation. This perspective should also be considered in the conclusion section which should be rephrased to avoid the impression that you have tested every relevant neurotransmitter.

We have extended the literature evidence on other neurochemicals’ effect on liking and wanting both in the Introduction:

“On the other hand, (facial) hedonic reactions to sensory pleasure are amplified by opioid, orexin and endocannabinoid stimulation of various “hedonic hotspots” of the brain, including the nucleus accumbens (NAc) shell and limbic areas such as insula and orbitofrontal cortex (Berridge and Kringelbach, 2015).”

and in the Discussion:

“From the animal and human literature, we however know that other systems (e.g. endocannabinoids, orexin, benzodiazepine, etc.) also contribute to the motivational and hedonic components of reward processing (Berridge and Kringelbach, 2015). Future studies should therefore broaden the neuropharmacological investigation of social vs. nonsocial reward processing, by using other compounds and different rewarding stimuli, to allow the generalizability of the findings to other social and nonsocial rewards, and to better understand their neurochemical basis.”

b) Please provide an explanation for the interpretation that opioid-induced increases in motivation for high-calorie foods as evidence that opioids primarily affect wanting through liking.

After rewriting the Introduction, the sentence has been removed.

c) Please consider changing to past tense when reviewing the evidence from single/small numbers of studies. Also, the effects reported in Chelnokova et al., 2014 on wanting were not explicitly subjective, rather a count of button presses which prolonged or shortened the duration an image was shown on the screen. Other relevant references on opioids and liking/wanting in humans include Gospic et al., 2008, Psychopharmacol; Atlas et al., 2013; Eikemo et al., 2017; Chelnokova et al., 2016, SCAN. Importantly, Gospic, Atlas and colleagues tested both men and women, whereas the cited evidence from Leknes' lab and Büchel's erotic image study were conducted exclusively in men. (Note that the affective image stimuli used by Gospic, Atlas and colleagues could also be described as social, and studies from Harriet de Wit's lab also indicate effects of opioid drugs on social reward responses.)

We thank the reviewer for the suggested references. We have added them in the Introduction with the exception of Gospic which is only used neutral and negative pictures:

“Administration of µ-opioid receptor agonists in heathy individuals has been associated with changes of subjective feelings and motivational responses to different type of rewards, as indicated for example by increased pleasantness of the most palatable food option available (Eikemo et al., 2016), increased effort to view, and liking of the most attractive opposite sex faces (Chelnokova et al., 2014), increased preference for stimuli with high-reward probability (Eikemo et al., 2017), and enhanced emotional ratings to positive and negative images (Atlas et al., 2014). Furthermore, administration of the non-selective opioid receptor antagonist naloxone to healthy men decreased subjective pleasure associated with viewing erotic pictures and reduced the activation of reward related brain regions such as the ventral striatum (Büchel et al., 2018).”

d) The following claim is surprising given the abundance of behavioural economics and facial EMG/facial emotion recognition studies in the field: "human research has struggled to adopt an operationalization that resembles the one used in animal research, e.g. measuring behavior and facial reactions instead of relying on subjective verbal report". Could a better explanation be that researchers testing human participants have much less precise ways of altering and measuring activity of opioid, dopamine and other neurotransmitters?

We agree with the reviewer that the use of facial EMG is abundant in research addressing emotional reactions to different types of stimuli, and in the context of reward processing. We have reformulated the paragraph and hopefully made clear that the above-mentioned claim refers to pharmacological studies on human reward. The text in the Introduction reads as follow:

“In spite of the progress made, human and animal research about the neurochemical regulation of reward processing remain difficult to compare, as human pharmacological studies have struggled to adopt translational paradigms and an operationalization of reward that resembles the one used in animal research, i.e. measuring decision utility and experienced utility in the same task, providing primary rewards on a trial-by-trial basis, and/or using objective hedonic reactions to consumed rewards, instead of relying on subjective verbal report (Der-Avakian et al., 2016; Pool et al., 2016). Hedonic facial reactions, the “gold standard” in animal research on the neurochemical regulation of reward processing, seems a promising research tool in this regard. Recently, the use of facial electromyography (fEMG) has gained increased attention in the context of human reward processing. Results suggested that human adults relax the corrugator muscle (involved in frowning), and to a lesser extent activate the zygomaticus muscle (involved in smiling), during both the anticipation and the consumption of different types of pleasurable stimuli, although differences between types of rewards exist (Bershad et al., 2019; Franzen and Brinkmann, 2016; Korb et al., 2020; Mayo et al., 2018; Pawling et al., 2017; Rasch et al., 2015; Ree et al., 2019; Sato et al., 2020; Wu et al., 2015).”

e) fEMG results for positive stimuli often show either zygomaticus increases or corrugator decreases, in fact Ree et al., Mayo et al. and Bershad et al. all found only corrugator decreases during touch. A reanalysis of the data from Pawling et al. available at PloS ONE suggests that the zygomaticus effect no longer meets statistical significance when only the time receiving touch is considered, and the interval after touch is removed. Please rephrase the Introduction section that currently indicates that both muscle potentials are altered in such gentle touch paradigms.

We thank the reviewer for pointing out that most studies have found corrugator relaxation for pleasant touch, while zygomaticus increase for touch appears more rarely. We have modified the Introduction accordingly:

“Results suggested that human adults relax the corrugator muscle (involved in frowning), and to a lesser extent activate the zygomaticus muscle (involved in smiling), during both the anticipation and the consumption of different types of pleasurable stimuli, although differences between types of rewards exist (Bershad et al., 2019; Franzen and Brinkmann, 2016; Korb et al., 2020; Mayo et al., 2018; Pawling et al., 2017; Rasch et al., 2015; Ree et al., 2019; Sato et al., 2020; Wu et al., 2015).”

28) The end paragraph of the Discussion provides clarity to some issues noted above. It would be helpful to provide this clarity early on as the comments in the early part of the Discussion section give an impression of overgeneralizing some of the findings.

We have largely rewritten the Discussion, hopefully improving clarity and avoiding overgeneralizations.

29) In the interpretation of the statistical findings please provide careful descriptions of the data and not just the p values? This includes addressing the following:

a) Please go through the manuscript carefully to avoid terminology such as "did not differ", when what is meant is "did not significantly differ". It is only correct to say that two measures do not differ if they are identical.

We hope that this revised manuscript provides more careful descriptions of the data. We have changed all instances where a difference was described, now clearly stating that the difference was or was not “significantly different”. Certainly, it would be even more correct to specify each time that the differences were/weren’t “statistically significantly different”.

However, this should be fairly clear to the readers of eLife (especially since the wording appears most of the time in proximity of statistical values such as p values), and was skipped to facilitate text understanding.

b) It is recommended to moderate the statement that an influence of drug via mood can be excluded.

We have removed that sentence.

We now introduce the initial group comparisons with the sentence “In order to rule out eventual group differences that were not of interest, we conducted a series of statistical tests to verify the matching of the three groups.”

The presence or absence of significant differences between groups on mood and other measures is then prescribed in detail.

The passage about drug effects on mood now reads: “The three groups of participants did not differ significantly in their maximum voluntary contraction (MVC) of the hand dynamometer, which was measured right before the main task and at the end of the main task, nor in their positive and negative mood measured with the PANAS at time of pill intake and three hours later (all b < 0.6, all t < 0.8, all p > 0.4; see Table 1).”

c) P values above the α level are not accurately described as "trends"; instead please describe the pattern of the data and state clearly that the required significance level was not reached.

All p values above.05 are described as trends or as marginally significant, as long as they are also below.1. When statistically non-significant results are found, we clearly state so.

d) “Importantly, this finding cannot be explained by mouth movements that might have occurred during food delivery (although instruction to swallow followed the Delivery window), as statistics did not reveal an interaction with the factor Condition.”, a null effect (interaction) is interpreted as evidence that a potential confound did not influence the results. This is inappropriate with frequentist statistics.

We totally agree and have reformulated the sentence now stating that the absence of the interaction makes an influence unlikely: “this finding is unlikely to be explained by mouth movements related to food ingestion, because participants were instructed to swallow the food after the Delivery window, and no interaction with the factor Reward Type was found.”

30) Please replace the CS and ZM abbreviations with corrugator and zygomaticus. CS is commonly used to denote conditioned stimulus thus causes some confusion, especially when used in the graphs. Further, RewardType gives the impression that it denotes drink versus touch, rather than reward level. In Discussion paragraph six, it is suggested to clarify the announcement of possible reward or similar, since participants only received the announced rewards they chose to work for. Further, please edit phrases including "the fact that" when describing the results.

We have replaced the abbreviations CS and ZM with the words corrugator and zygomaticus (both in text and figures).

RewardType was changed to Reward Level in both text and figures.

Condition was changed to Reward Type.

In the Discussion we now distinguish between anticipation and consumption of reward.

We have removed “the fact that” from the manuscript.

31) Please provide acknowledgement of limitations in the study, which certainly exists e.g. cross-sectional design (rather than within-subjects). This can be noted when discussing the interpretation of the findings.

We have added a larger section to the Discussion, acknowledging and discussing a series of imitations, including the use of a cross-sectional design:

”The current study is characterised by a number of limitations. First, only two types of stimuli (food and touch) were used to define social and non-social rewards, and only two neurochemical systems (dopaminergic and opioidergic) were challenged. […] This reveals the difficulty of uncovering the neurochemical basis of reward processing in humans, and suggests that larger sample sizes should be used in future pharmacological studies to investigate the neurochemical bases of touch and other rewards.”

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Revisions:

The authors were very responsive and provided a thorough revision of their manuscript, which is a very solid and valuable contribution to the literature. However, there are still some concerns that should be addressed.

1) The corrugator and zygomaticus data is a strength of the manuscript, but the current manuscript emphasizes facial responses more than is warranted by the study design and the hypotheses. Please address the following:

a) Please specify the lack of significant reductions in subjective ratings of chocolate milk and stroking touch in the Abstract.

To specify the null-effects of Drug on ratings of wanting and liking, we have added this sentence to the Abstract:

“Subjective ratings of wanting and liking were not modulated by either drug.”

b) The phrasing in the Introduction still appears to indicate that animal studies provide the best answers, whereas human subjective report is less valuable. However, the animal studies are severely hampered by their use of an indirect measure of liking, which has not been generalized beyond food reward but instead their strength in animal models is the ability to manipulate specific circuits/neurons. Please edit the text to reflect these limitations and strengths in animal models

In the Introduction we have added the following text: “In spite of the progress made, the animal research is only partly informative to comprehend reward processing in humans. While animal research allows to investigate the activity of neurons and neurotransmitters in a much more targeted way, it is also limited to certain measures of liking (i.e. behaviour and facial expressions, while humans can also provide subjective reports), and has mainly focused on food rewards in the past.”

c) The numerous figures of fEMG data patterns do not seem warranted, i.e. do they present the answers to the most important questions? It seems as though all effects below p=0.1 are illustrated.

Our intention was to show the EMG in relation to ratings of wanting/liking and effort, even when no drug effects were found. The reader might indeed be interested in seeing that the corrugator and zygomaticus activation/relaxation patterns follow expected patterns, with greater wanting and effort resulting in corrugator relaxation, and greater liking resulting in zygomaticus activation.

However, following the reviewer’s request, we have now removed former Figures 5 (Post-Effort anticipation) and 7 (Relax), which showed EMG effects with .5 > p <.1.

d) The comparable subjective wanting and liking ratings in the placebo, naltrexone and amisulpride are a main result and should be listed as such in the Abstract, Results and Discussion. Further, please clarify that face muscle responses to high-reward stimuli were also comparable after drugs. Perhaps equivalence testing or Bayesian stats could be used to address the credibility of these interesting null results?

To address the credibility of the null results on ratings, we ran Bayesian LMMs for the ratings of wanting and liking, using the package brms in R (Bürkner, 2017), and also compared the full models to identical models without the predictor Drug, using leave-one-out cross-validation (LOO-CV) based on the posterior likelihoods (Vehtari et al., 2017). This confirmed the data to be more likely under the model without the predictor Drug (weights based on averaging via stacking of predictive distributions were 0.67 and 0.998 for wanting and liking) than under the full model (weights 0.33 and 0.002 for wanting and liking). These Bayesian analyses strengthen the view, already conveyed by the frequentist LMMs, that neither drug affected wanting or liking of both types of rewards in this study. We have modified the Abstract, Results and Discussion sections, accordingly. The corresponding R scripts are available on OSF, and the model result tables have been added to the supplementary materials.

It is unclear to us, what effect of muscle responses to high-reward stimuli the reviewer is referring to.

e) Referring to reduced activity of a frowning muscle as an indication of “hedonic anticipatory pleasure” is unwarranted and likely to confuse readers with respect to what the data shows. The corrugator is not a good measure of pleasure. Similarly, the discussion of drug effects on “hedonic responses” to touch gives the impression that the drugs altered touch pleasantness, instead of a face muscle response, which was not measured in most other drug studies.

We respectfully disagree with the reviewer’s comment that “the corrugator is not a good measure of pleasure”, based on the fact that many past studies investigating hedonic responses to different types of rewards have found significant relaxation of the corrugator to more pleasant rewards, but, importantly, no reliable changes in zygomaticus activation.

This is especially the case for studies on affective touch, as in Bershad et al., 2019, Mayo et al., 2018, and Ree et al., 2019.

Studies that have administered food rewards also found effects that were either exclusively associated with, or stronger/more stable in, the corrugator compared to the zygomaticus muscle (Rasch et al., 2015; Sato et al., 2020).

Our own previous study, which used the same paradigm and the same food and touch stimuli (Korb et al., 2020), also found significant pleasure responses during anticipation and consumption of rewards only in the corrugator, but not in the zygomaticus muscle.

Finally, a study by Larsen et al., 2003, looked at corrugator and zygomaticus responses to sounds, pictures, and words representing a large range in valence. They found that: “not only that very negative and mildly negative pictures both potentiated activity over corrugator supercilii, but also that very positive pictures significantly diminished activity”, and that “even when negative affect is held constant at minimal levels, positive affect inhibits activity over corrugator supercilii”. They also report that “valence had a substantially stronger effect on activity over corrugator supercilii. In addition, more participants showed a reliable linear effect of valence on activity over corrugator supercilii.” The zygomaticus major muscle, on the other hand, shows a more complex relationship to stimulus valence, as “not only very positive, but also mildly positive and very negative pictures elicited greater activity over zygomaticus major than did neutral pictures”.

We conclude that corrugator relaxation seems, on the basis of the literature including our own previous work, a good measure of pleasure, and possibly a more reliable measure of hedonic responses during the consumption (and anticipation) of rewarding stimuli, compared to the zygomaticus muscle.

If we are missing some relevant information/published papers, we would be very grateful if the reviewer could indicate them. Unfortunately, the reference that was suggested in the previous review round (Bremhorst et al., 2019) does not seem – at least to our understanding – very relevant to this regard, as it concerns FACS coding in dogs in response to frustration, and the corrugator muscle is never mentioned.

As to the discussion of drug effects on “hedonic responses” to touch, we would like to clarify that facial responses to rewards are typically considered the “gold standard” in animal research on the neurochemical regulation of reward processing, and that it was an explicit goal of this project to measure implicit facial reactions to rewards in humans, in addition to explicit ratings of wanting/liking. We consider changes in facial EMG to be a proxy of hedonic responses (as also indicated in the manuscript), and to add valuable information complementing explicit ratings of pleasantness. Indeed, as our results (as well as other null effect findings of pharmacological manipulations on pleasantness of affective touch) suggest, pharmacological manipulation of the dopamine and/or opioid system in humans may only result in subtle effects on anticipatory and consummatory hedonic reactions, which can be seen in the type of measures taken in animal research (physical effort and facial responses) but not in the subjective ratings that are more frequently used in human research.

2) The presentation of the behavioral wanting data to illustrate the two significant two-way interactions is confusing. It's better to depict non-pooled data for each reward type and level, and use lines to indicate the significant interactions. It is challenging to determine from the figure whether there was an effect of drug for wanting of gentle caresses, for instance. Please address the following:

a) The authors now provide this information in the supplement. However, showing one chart per drug instead of one per reward type/level makes it hard to discern any drug-related differences. The x axis labeling in the supplementary file was also confusing. It is recommended to remove Figure 3 and replace it with the reorganized supplementary to clarify any drug differences.

Figure 3—figure supplement 1 now shows one chart per Reward Type (instead of one chart per Drug). As it remains difficult to flag the effects of Drug X Reward Type and Drug X Reward Level, which were found for Effort, we would like to keep this figure as a supplement to Figure 3.

3) Please change the terminology to accurately depict the data and statistics

a) If the authors use 0.05 as their α level it is incorrect to describe results above 0.05 as trends or marginal effects. The higher force levels should be described as non-significantly higher. Similarly, the legend of Figure 5 makes too strong a case given the data. 1. This terminology is statistically inappropriate and maintains some unnecessary ambiguity in the literature. Many experts in statistics have pointed out this problem in recent years (see e.g. Gibbs and Gibbs 2015, https://academic.oup.com/bja/article/115/3/337/312358). It is encouraged that the authors remove these terms

We have replaced all prior mentioning of “marginally significant” or “trends for” effects, with “non-significant”.

b) It is also recommended to use more accurate descriptors of group differences, given the between-groups design, i.e. use “higher” instead of “increased”, “lower” instead of “reduced” etc.

We have changed these terms accordingly.

4) The Results section is very hard to read especially since there are multiple periods of time analyzed. Please explain in the beginning of each section what information should be gained from the analyzed data rather than providing some of this information is in the summary of EMG data. This would help the readers to better understand the data.

To improve readability of the Results section, we have summarised the main behavioural findings at the beginning of the section “Explicit measures: ratings of wanting, ratings of liking and physical effort”, and we have summarised the main EMG findings at the beginning of the section “Implicit measures: facial EMG”. We have also removed two figures, to shorten the Results section.

5) 50 mg naltrexone is not a low drug dose. It is the standard clinical dose and blocks 90-100% of mu opioid receptors. It is recommended that the authors revisit the PET literature and take into account that 50 mg naltrexone is considered sufficient to induce so-called “full” blockade of mu receptors, when interpreting their results.

We are aware that 50mg of naltrexone blocks 90% of mu-opioid receptors, and have indicated this in the Materials and methods section: “up to 90% of mu-opioid receptors in the brain remain blocked by naltrexone after 48 hours, and partial receptor blockade could be shown up to 168 hours after intake (Lee et al., 1988).”

We have now slightly changed the paragraph in the Discussion, to only say that 400mg of amisulpride may be a low dose:

“Second, we used a relatively low dose of amisulpride (400mg). […] Upon availability of drug compounds that more strongly modulate the dopaminergic systems with minimal side effects, future studies should, however, explore dose dependent changes in human reward processing.”

References

Atlas, L. Y., Wielgosz, J., Whittington, R. A., & Wager, T. D. (2014). Specifying the non-specific factors underlying opioid analgesia: Expectancy, attention, and affect. Psychopharmacology, 231(5), 813–823. https://doi.org/10.1007/s00213-013-3296-1

Barr, D. J., Levy, R., Scheepers, C., & Tily, H. J. (2013). Random effects structure for confirmatory hypothesis testing: Keep it maximal. Journal of Memory and Language, 68(3). https://doi.org/10.1016/j.jml.2012.11.001

Benjamini, Y., & Hochberg, Y. (1995). Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 57(1), 289–300. JSTOR.

Bershad, A. K., Mayo, L. M., Van Hedger, K., McGlone, F., Walker, S. C., & de Wit, H. (2019). Effects of MDMA on attention to positive social cues and pleasantness of affective touch. Neuropsychopharmacology, 44(10), 1698–1705. https://doi.org/10.1038/s41386-019-0402-z

Chelnokova, O., Laeng, B., Eikemo, M., Riegels, J., Loseth, G., Maurud, H., Willoch, F., & Leknes, S. (2014). Rewards of beauty: The opioid system mediates social motivation in humans. Molecular Psychiatry, 19(7), 746–747. https://doi.org/10.1038/mp.2014.1

de Wit, H., & Bershad, A. K. (2020). MDMA enhances pleasantness of affective touch. Neuropsychopharmacology, 45(1), 217–239. https://doi.org/10.1038/s41386-019-0473-x

Eikemo, M., Biele, G., Willoch, F., Thomsen, L., & Leknes, S. (2017). Opioid Modulation of Value-Based Decision-Making in Healthy Humans. Neuropsychopharmacology, 42(9), 1833–1840. https://doi.org/10.1038/npp.2017.58

Eikemo, M., Loseth, G. E., Johnstone, T., Gjerstad, J., Willoch, F., & Leknes, S. (2016). Sweet taste pleasantness is modulated by morphine and naltrexone. Psychopharmacology, 233(21–22), 3711–3723. https://doi.org/10.1007/s00213-016-4403-x

Korb, S., Massaccesi, C., Gartus, A., Lundstrom, J. N., Rumiati, R., Eisenegger, C., & Silani, G. (2020). Facial responses of adult humans during the anticipation and consumption of touch and food rewards. Cognition, 194, 104044. https://doi.org/10.1016/j.cognition.2019.104044

Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993). Looking at pictures: Affective, facial, visceral, and behavioral reactions. Psychophysiology, 30(3), 261–273. Scopus. https://doi.org/10.1111/j.1469-8986.1993.tb03352.x

Larsen, J. T., Norris, C. J., & Cacioppo, J. T. (2003). Effects of positive and negative affect on electromyographic activity over zygomaticus major and corrugator supercilii. Psychophysiology, 40(5), 776–785. https://doi.org/10.1111/1469-8986.00078

Lee, M. C., Wagner, H. N., Tanada, S., Frost, J. J., Bice, A. N., & Dannals, R. F. (1988). Duration of occupancy of opiate receptors by naltrexone. Journal of Nuclear Medicine: Official Publication, Society of Nuclear Medicine, 29(7), 1207–1211.

Lopez-Persem, A., Rigoux, L., Bourgeois-Gironde, S., Daunizeau, J., & Pessiglione, M. (2017). Choose, rate or squeeze: Comparison of economic value functions elicited by different behavioral tasks. PLOS Computational Biology, 13(11), e1005848. https://doi.org/10.1371/journal.pcbi.1005848

Matuschek, H., Kliegl, R., Vasishth, S., Baayen, H., & Bates, D. (2017). Balancing Type I error and power in linear mixed models. Journal of Memory and Language, 94, 305–315. https://doi.org/10.1016/j.jml.2017.01.001

Meyer, M. C., Straughn, A. B., Lo, M. W., Schary, W. L., & Whitney, C. C. (1984). Bioequivalence, dose-proportionality, and pharmacokinetics of naltrexone after oral administration. The Journal of Clinical Psychiatry, 45(9 Pt 2), 15–19.

Rosenzweig, P., Canal, M., Patat, A., Bergougnan, L., Zieleniuk, I., & Bianchetti, G. (2002). A review of the pharmacokinetics, tolerability and pharmacodynamics of amisulpride in healthy volunteers. Human Psychopharmacology, 17(1), 1–13. https://doi.org/10.1002/hup.320

Weber, S. C., Beck-Schimmer, B., Kajdi, M.-E., Muller, D., Tobler, P. N., & Quednow, B. B. (2016). Dopamine D2/3- and μ-opioid receptor antagonists reduce cue-induced responding and reward impulsivity in humans. Translational Psychiatry, 6(7), e850. https://doi.org/10.1038/tp.2016.113

https://doi.org/10.7554/eLife.55797.sa2

Article and author information

Author details

  1. Sebastian Korb

    1. Department of Psychology, University of Essex, Colchester, United Kingdom
    2. Department of Cognition, Emotion and Methods in Psychology, University of Vienna, Vienna, Austria
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    sebastian.korb@essex.ac.uk
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3517-3783
  2. Sebastian J Götzendorfer

    Department of Cognition, Emotion and Methods in Psychology, University of Vienna, Vienna, Austria
    Contribution
    Investigation, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Claudia Massaccesi

    Department of Clinical and Health Psychology, University of Vienna, Vienna, Austria
    Contribution
    Investigation, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0519-6324
  4. Patrick Sezen

    Department of Psychiatry and Psychotherapy, Medical University of Vienna, Vienna, Austria
    Contribution
    Investigation, Medical testing and drug administration
    Competing interests
    No competing interests declared
  5. Irene Graf

    Department of Psychiatry and Psychotherapy, Medical University of Vienna, Vienna, Austria
    Contribution
    Investigation, Medical exams and drug administration, curation of analysis of blood samples
    Competing interests
    No competing interests declared
  6. Matthäus Willeit

    Department of Psychiatry and Psychotherapy, Medical University of Vienna, Vienna, Austria
    Contribution
    Conceptualization, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Christoph Eisenegger

    Department of Cognition, Emotion and Methods in Psychology, University of Vienna, Vienna, Austria
    Contribution
    Conceptualization, Funding acquisition
    Competing interests
    No competing interests declared
  8. Giorgia Silani

    Department of Clinical and Health Psychology, University of Vienna, Vienna, Austria
    Contribution
    Conceptualization, Supervision, Funding acquisition, Project administration, Writing - review and editing
    For correspondence
    giorgia.silani@univie.ac.at
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4284-3618

Funding

Vienna Science and Technology Fund (CS15-003)

  • Christoph Eisenegger
  • Giorgia Silani

Vienna Science and Technology Fund (VRG13-007)

  • Christoph Eisenegger

University of Vienna (Open Access Publishing Fund)

  • Giorgia Silani

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The study was supported by the Vienna Science and Technology Fund (WWTF) with a grant (CS15-003) awarded to Giorgia Silani and Christoph Eisenegger, a grant (VRG13-007) awarded to Christoph Eisenegger, and the Open Access Publishing Fund of the University of Vienna. Funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. We thank Prof. Mathias Pessiglione and his group for help with coding the dynamometer output in Matlab, and Prof. Boris Quednow for commenting on an earlier version of the manuscript. We thank Lukas Lengersdorff, Annabel Losecaat Vermeer, Nace Mikus, Lei Zhang, and Matteo Lisi for providing statistical advice. Many thanks to Mumna Al Banchaabouchi and the interns and master students whose help was crucial for data acquisition: Mani Erfanian Abdoust, Anne Franziska Braun, Raimund Bühler, Lena Drost, Manuel Czornik, Lisa Hollerith, Berit Hansen, Luise Huybrechts, Merit Pruin, Vera Ritter, Frederic Schwetz, Conrad Seewald, Carolin Waleew, Luca Wiltgen, Stephan Zillmer. We are extremely grateful to the reviewers Siri Leknes, Guillome Sescousse, Diego Pizzagalli, and Yuen-Siang Ang, whose comments have helped us to improve the manuscript.

Ethics

Human subjects: The study was approved by the Ethical Committee of the Medical University of Vienna (EK N. 1393/2017) and was performed in line with the Declaration of Helsinki (World Medical Association, 2013). Participants signed informed consent.

Senior Editor

  1. Christian Büchel, University Medical Center Hamburg-Eppendorf, Germany

Reviewing Editor

  1. Mary Kay Lobo, University of Maryland, United States

Reviewers

  1. Siri Leknes
  2. Guillaume Sescousse, INSERM, France

Publication history

  1. Received: February 6, 2020
  2. Accepted: September 14, 2020
  3. Version of Record published: October 13, 2020 (version 1)

Copyright

© 2020, Korb et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 691
    Page views
  • 83
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Developmental Biology
    2. Neuroscience
    Athene Knüfer et al.
    Research Article Updated

    In the vertebrate central nervous system, groups of functionally related neurons, including cranial motor neurons of the brainstem, are frequently organised as nuclei. The molecular mechanisms governing the emergence of nuclear topography and circuit function are poorly understood. Here we investigate the role of cadherin-mediated adhesion in the development of zebrafish ocular motor (sub)nuclei. We find that developing ocular motor (sub)nuclei differentially express classical cadherins. Perturbing cadherin function in these neurons results in distinct defects in neuronal positioning, including scattering of dorsal cells and defective contralateral migration of ventral subnuclei. In addition, we show that cadherin-mediated interactions between adjacent subnuclei are critical for subnucleus position. We also find that disrupting cadherin adhesivity in dorsal oculomotor neurons impairs the larval optokinetic reflex, suggesting that neuronal clustering is important for co-ordinating circuit function. Our findings reveal that cadherins regulate distinct aspects of cranial motor neuron positioning and establish subnuclear topography and motor function.

    1. Neuroscience
    Anthony JE Berndt et al.
    Research Article

    Retrograde BMP signaling and canonical pMad/Medea-mediated transcription regulates diverse target genes across subsets of Drosophila efferent neurons, to differentiate neuropeptidergic neurons and promote motor neuron terminal maturation. How a common BMP signal regulates diverse target genes across neuronal subsets remains largely unresolved, although available evidence implicates subset-specific transcription factor codes rather than differences in BMP signaling. Here, we examine the cis-regulatory mechanisms restricting BMP-induced FMRFa neuropeptide expression to Tv4 neurons. We find that pMad/Medea bind at an atypical, low affinity motif in the FMRFa enhancer. Converting this motif to high affinity caused ectopic enhancer activity and eliminated Tv4 neuron expression. In silico searches identified additional motif instances functional in other efferent neurons, implicating broader functions for this motif in BMP-dependent enhancer activity. Thus, differential interpretation of a common BMP signal, conferred by low affinity pMad/Medea binding motifs, can contribute to the specification of BMP target genes in efferent neuron subsets.