1. Neuroscience
Download icon

Opioid antagonism modulates wanting-related frontostriatal connectivity

  1. Alexander Soutschek  Is a corresponding author
  2. Susanna C Weber
  3. Thorsten Kahnt
  4. Boris B Quednow
  5. Philippe N Tobler
  1. Department of Psychology, Ludwig Maximilian University, Germany
  2. Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Switzerland
  3. Department of Neurology, Northwestern University Feinberg School of Medicine, United States
  4. Experimental and Clinical Pharmacopsychology, Department of Psychiatry, Psychotherapy and Psychosomatics, Psychiatric Hospital, University of Zurich, Switzerland
  5. Neuroscience Center Zurich, University of Zurich and Swiss Federal Institute of Technology Zurich, Switzerland
Research Article
  • Cited 0
  • Views 252
  • Annotations
Cite this article as: eLife 2021;10:e71077 doi: 10.7554/eLife.71077

Abstract

Theoretical accounts distinguish between motivational (‘wanting’) and hedonic (‘liking’) dimensions of rewards. Previous animal and human research linked wanting and liking to anatomically and neurochemically distinct brain mechanisms, but it remains unknown how the different brain regions and neurotransmitter systems interact in processing distinct reward dimensions. Here, we assessed how pharmacological manipulations of opioid and dopamine receptor activation modulate the neural processing of wanting and liking in humans in a randomized, placebo-controlled, double-blind clinical trial. Reducing opioid receptor activation with naltrexone selectively reduced wanting of rewards, which on a neural level was reflected by stronger coupling between dorsolateral prefrontal cortex and the striatum under naltrexone compared with placebo. In contrast, reducing dopaminergic neurotransmission with amisulpride revealed no robust effects on behavior or neural activity. Our findings thus provide insights into how opioid receptors mediate neural connectivity related to specifically motivational, not hedonic, aspects of rewards.

Editor's evaluation

The authors measured the effects of the opioid receptor antagonist naltrexone (50mg), and the dopamine D2/3 antagonist amisulpiride (400mg), on self-reported reward wanting vs. liking and functional connectivity between the prefrontal cortex and striatum (using functional magnetic resonance imaging) in healthy human participants, using a between-subjects design. Naltrexone led to lower wanting, but not liking, and these changes were associated with greater frontostriatal connectivity. Amisulpiride also tended to increase connectivity on wanting trials but did not affect either wanting or liking scores. The results raise the possibility that both opioid and dopamine transmission influence reward wanting, with the former more closely related to conscious processes. This manuscript will be of broad interest to neuroscientists interested in the brain mechanisms underlying reward processing, both at the circuit and molecular levels.

https://doi.org/10.7554/eLife.71077.sa0

Introduction

Rewards are central for goal-directed behavior as they induce approach behavior toward valued outcomes (Schultz, 2015). Theoretical models distinguish between behavioral dimensions of rewards, such as the motivational drive to obtain rewards (‘wanting’) versus the hedonic pleasure associated with reward consumption (‘liking’), whereby ‘wanting’ and ‘liking’ refer to preconscious, rather than conscious, mental states (Berridge, 1996; Berridge and Kringelbach, 2015; Berridge et al., 2009). Dysfunctions in wanting and liking of rewards belong to the core symptoms to addiction, which can be conceptualized as a wanting-dominated state with deficits in switching to liked non-drug rewards (Berridge and Robinson, 2016; Berridge et al., 2009). It is thus important to obtain a better understanding of the human brain mechanisms underlying wanting and liking. Previous animal research suggested that wanting and liking relate to dissociable neurochemical mechanisms (Berridge and Kringelbach, 2015; Berridge and Valenstein, 1991): Dopaminergic activity is thought to modulate the wanting component of rewards, but not liking (Berridge and Valenstein, 1991). In contrast, the opioidergic system has later been associated with both wanting and liking (Berridge and Kringelbach, 2015). Human studies support the hypothesized link between dopaminergic activation and cue-triggered wanting (Hebart and Gläscher, 2015; Soutschek et al., 2020b; Weber et al., 2016) as well as the motivation to work for rewards (Cawley et al., 2013; Chong et al., 2015; Korb et al., 2020; Skvortsova et al., 2017; Soutschek et al., 2020a; Venugopalan et al., 2011; Westbrook et al., 2020; Zénon et al., 2016). Noteworthy, two of these studies suggest that dopamine changes only experimental measures of wanting (motivation to work for rewards), but not self-reported wanting ratings (Korb et al., 2020; Venugopalan et al., 2011).

Consistent with animal findings, pharmacological manipulations of the opioid system affected both wanting and liking aspects of rewards in humans (Buchel et al., 2018; Chelnokova et al., 2014; Eikemo et al., 2016). Less is known, however, about the neuroanatomical basis of human wanting and liking. Both dopaminergic and opioidergic neurons project to reward circuits in the striatum as well as to the prefrontal cortex (Delay-Goyet et al., 1987; Lidow et al., 1991), and recent neuroimaging findings suggest that these regions indeed play a role in processing of wanting and liking (Weber et al., 2018). In particular, the ventral striatum appears to encode the currently behaviorally relevant reward dimension and dynamically switch functional connectivity with wanting- and liking-encoding prefrontal regions (Weber et al., 2018; such frontostriatal gating constitutes one possibility how dopamine [and by extension opioids] can affect functions involving prefrontal cortex and the striatum; Cools, 2011). This is also in line with the view that wanting and liking are preferentially represented in partially distinct subregions of the striatum (which are in turn connected to different input and output regions; Peciña, 2008). Thus, processing of wanting and liking appears to be dissociable on both a neurochemical and an anatomical basis. However, it remains unknown how pharmacological and connectivity-related brain mechanisms interact. We investigated whether frontostriatal connectivity related to motivational and hedonic judgements is modulated by dissociable neurotransmitter systems.

To test this hypothesis, the current study investigated the impact of pharmacologically manipulating dopaminergic and opioidergic systems on the neural processing of wanting and liking information. This study was part of a larger project investigating also the roles of opioidergic and dopaminergic activity for reward impulsivity (Weber et al., 2016). Here, we administered a task that allows distinguishing between wanting and liking dimensions of valued goods and assessed how pharmacologically reducing dopaminergic (using the dopamine antagonist amisulpride) or opioidergic neurotransmission (with the opioid antagonist naltrexone) changes frontostriatal connectivity related to parametric wanting and liking judgements. We hypothesized that, on a behavioral level, reducing opioid receptor activation will reduce both wanting and liking (Buchel et al., 2018; Chelnokova et al., 2014; Eikemo et al., 2016), whereas reduced dopaminergic neurotransmission will selectively affect wanting (Cawley et al., 2013; Korb et al., 2020; Venugopalan et al., 2011). We further hypothesized that on a neural level the behavioral effects of the pharmacological manipulations are mirrored by changes in frontostriatal connectivity related to wanting and liking information (Weber et al., 2018). In particular, we expected that frontostriatal connectivity during wanting judgements is modulated by opioidergic and dopaminergic activation (as these neurotransmitters have been related to the processing of wanting), whereas frontostriatal connectivity during liking judgements should be reduced after reduction of opioidergic neurotransmission.

Results

Opioid antagonism reduces wanting ratings

We analyzed the data of healthy young volunteers who rated how much they wanted or liked everyday items in the MRI scanner. We collected wanting and liking ratings for all items in the MRI scanner twice, once before (pre-test session) and once after (post-test session) participants played a game on the computer where they won or lost 50% of the items (in order to have equal numbers of won and lost items for the statistical analysis). This allowed us to assess whether participants behaviorally distinguished between wanting and liking ratings, because based on our previous findings we expected that winning and losing items has dissociable effects on wanting and liking (Weber et al., 2018). Participants actually received the won items at the end of the experiment (i.e., after the post-test session). We therefore selected everyday items (e.g., batteries or candles – for the full list of items, see Weber et al., 2018) that should be both wanted and liked by the majority of our participants from the Zurich student population. To test the impact of pharmacologically manipulating dopaminergic and opioidergic receptor activation on wanting and liking, participants received either naltrexone (N = 37), amisulpride (N = 40), or placebo (N = 39) prior to performing the task in the scanner.

First, we performed a sanity check whether participants distinguished between wanting and liking ratings by assessing the impact of winning versus losing items on wanting and liking ratings in the post-test session. As recommended for pre-test/post-test designs (Dugard and Todman, 1995), we regressed ratings in the post-test session on item-specific pre-test ratings. Moreover, we included predictors for Judgement (wanting versus liking), Item type (lost versus won), and the interaction terms. Contrary to our previous study (Weber et al., 2018), we observed no significant Judgement× Item type interaction, β = 0.50, t(111) = 1.61, p = 0.11, which does not replicate our previous result that winning versus losing items has dissociable effects on wanting versus liking of the items (Figure 1B and Table 1). However, separate analyses for wanting and liking ratings revealed no significant difference in wanting ratings between won and lost items, β = 0.29, t(115) = 0.64, p = 0.52, whereas liking was more strongly reduced for lost than for won items, β = 01.84, t(115) = 3.16, p = 0.002, with the latter effect replicating our result of decreased liking of lost versus won items (Weber et al., 2018).

Task procedure and behavioral results.

(A) Participants rated in the MRI scanner how much they wanted or liked objects before (pre-test) or after (post-test) they won or lost these items in a game between the scanning sessions. (B) On each trial, a cue indicated whether a wanting or liking rating was required, followed by the presentation of the current object (here: a pick-up sticks game). Participants had to rate how much they wanted or liked the presented object within 3.5 s, then the next trial started after a variable inter-trial interval (mean = 3 s). (C) Liking ratings were significantly reduced for objects that were lost relative to won in the gamble, while wanting ratings did not significantly differ between lost versus won items. (D) The opioid antagonist naltrexone significantly reduced wanting ratings relative to placebo, while liking ratings were unaffected by naltrexone or the dopamine antagonist amisulpride. For illustration purposes, participant-specific mean wanting/liking ratings are plotted on a scale from 0 to 10, while the statistical analyses are conducted on the participant- and item-specific wanting and liking ratings. Error bars indicate standard error of the mean, black dots represent individual data points. *p < 0.05, ***p < 0.001.

Table 1
Results of mixed general linear model 1 (MGLM-1) on wanting and liking ratings in the post-test as function of Judgement (wanting versus liking), Item type (lost versus won), and Pre-test ratings.

Standard errors of the mean (SE) are in brackets.

Beta (SE)t-Valuedfp-Value
Intercept0.21 (0.66)0.321190.75
Judgement–2.27 (0.64)3.54125<0.001
Item type0.07 (1.10)0.07980.95
Pre-test79.87 (0.50)158.30124<0.001
Judgement × Item type2.01 (1.24)1.621090.11
Judgement × Pre-test2.41 (0.61)3.93289<0.001
Item type × Pre-test–1.62 (0.71)2.261330.03
Judgement × Item type × Pre-test0.91 (0.90)1.012880.31

Next, we assessed the impact of reducing dopamine and opioid receptor activity on wanting and liking judgements. We analyzed ratings (pre- and post-test) with predictors for Amisulpride (versus placebo), Naltrexone (versus placebo), Judgement, Session (pre-test versus post-test), and the interaction terms. This analysis provided evidence that reducing opioid neurotransmission differentially affected wanting and liking ratings, Naltrexone × Judgement, β = 7.02, t(125) = 2.36, p = 0.02, while we observed no significant effects for amisulpride, β = 3.79, t(126) = 1.30, p = 0.20 (Figure 1C and Table 2). Judgement type-specific analyses suggested that wanting ratings were significantly reduced under naltrexone (mean = 4.5, standard deviation [sd] = 1.0) relative to placebo (mean = 4.9, sd = 1.0), β = –13.85, t(115) = 2.12, p = 0.04, Cohen’s d = 0.47, whereas amisulpride (mean = 4.9, sd = 1.0) showed no significant effects on wanting ratings relative to placebo, β = –1.39, t(116) = 0.22, p = 0.83, Cohen’s d = 0.05. Neither naltrexone (mean = 5.2, sd = 0.9) nor amisulpride (mean = 5.4, sd = 0.8) showed significant effects on liking relative to placebo (mean = 5.2, sd = 0.8), for both t < 1.17, p > 0.24, Cohen’s d < 0.27. Taken together, our findings provide evidence for involvement of opioidergic neurotransmission in wanting judgements.

Table 2
Results for mixed general linear model 2 (MGLM-2) assessing drug effects on wanting and liking ratings as function of Drug (amisulpride versus placebo and naltrexone versus placebo), Judgement (wanting versus liking), and Session (pre-test versus post-test).

Standard errors of the mean (SE) are in brackets.

Beta (SE)t-Valuedfp-Value
Intercept3.62 (5.50)0.66980.51
Amisulpride2.37 (5.09)0.471140.64
Naltrexone–6.80 (5.19)1.311140.19
Judgement4.38 (2.08)2.111250.04
Session–3.30 (1.98)1.6716120.10
Amisulpride × Judgement3.79 (2.92)1.301260.20
Naltrexone × Judgement7.02 (2.98)2.361250.02
Amisulpride × Session0.02 (2.79)0.0016290.99
Naltrexone × Session2.70 (2.84)0.9516180.34
Judgement × Session–1.09 (1.97)0.5518450.58
Amisulpride × Judgement × Session–1.50 (2.78)0.5418560.59
Naltrexone × Judgement × Session–2.95 (2.83)1.0418530.30

Amisulpride can show both pre-synaptic and post-synaptic effects depending on the administered dose. To control for the possibility that the effective dose of amisulpride might vary between participants due to differences in body weight, we added the predictor body weight (as well as its interactions with all other factors) to the above reported regression model. While the Naltrexone × Judgement interaction remained significant, β = 7.99, t(124) = 2.69, p = 0.008, there were still no significant amisulpride effects, all t < 1.46, all p > 0.14. There was thus no evidence for dose-dependent effects of amisulpride on wanting or liking ratings.

To assess the robustness of these findings, we conducted also a non-hierarchical analysis of pharmacological effects on wanting and liking ratings using the mean wanting and liking ratings across all items, determined separately for each participant and session (pre-test versus post-test). The analysis of wanting ratings replicated the significant main effect of naltrexone versus placebo, t(128) = 2.16, p = 0.03, while amisulpride showed no significant effect on mean wanting ratings, t(128) = 0.23, p = 0.82. Mean liking ratings were neither affected by naltrexone, t(126) = 0.03, p = 0.98, nor amisulpride, t(126) = 1.22, p = 0.23. Thus, also the non-hierarchical analysis of aggregated mean data provided no evidence for significant amisulpride effects.

Opioid antagonism reduces wanting-related frontostriatal connectivity

Next, we investigated the neural mechanisms underlying the impact of opioid antagonism on wanting. Following the procedures from our previous study (Weber et al., 2018), we first determined the neural correlates of wanting and liking by computing GLM-1 in which onset regressors for wanting and liking judgements were modulated by non-orthogonalized parametric modulators for wanting and liking ratings. Wanting ratings (independently of the required judgement type) correlated with activation in ventromedial prefrontal cortex (VMPFC; z = 7.32, whole-brain FWE-corrected, p < 0.001, peak = [0 44–7]), dorsolateral prefrontal cortex (DLPFC; z = 6.63, whole-brain FWE-corrected, p < 0.001, peak = [–21 38 44]), and posterior cingulate cortex (PCC; z = 5.29, whole-brain FWE-corrected, p = 0.002, peak = [–3 –37 38]) (Figure 2A and Table 3). Liking ratings correlated with BOLD signal changes in more posterior parts of PCC (z = 4.37, whole-brain FWE-corrected, p = 0.02, peak = [–9 –64 38]) (Figure 2B and Table 4). Moreover, we also replicated our previous finding that liking ratings correlate with activity in orbitofrontal cortex (OFC) when applying small-volume correction (SVC; anatomical mask for the OFC based on the wfupickatlas; z = 3.20, small-volume FWE-corrected, p = 0.046, peak = [–21 50–4]). Together, these data replicate our previous findings that wanting and liking are correlated with activation in VMPFC and OFC, respectively. However, we observed no significant effects of naltrexone or amisulpride (relative to placebo) on these neural representations of wanting or liking in these regions (or at the whole-brain level), even at lenient statistical thresholds (p < 0.001 uncorrected, cluster size >20 voxels).

Neural correlates of (A) wanting and (B) liking independently of behavioral relevance.

Wanting correlated with activation in dorsolateral prefrontal cortex (DLPFC), ventromedial prefrontal cortex (VMPFC), and posterior cingulate cortex (PCC) (whole-brain FWE-corrected). Liking correlated with activation in dorsal PCC (whole-brain FWE-corrected) and orbitofrontal cortex (small-volume FWE-corrected). (C) Wanting ratings significantly correlated with activation in the striatum during wanting judgements (small-volume FWE-corrected). Images are thresholded at p < 0.001 uncorrected.

Table 3
Anatomical locations and MNI coordinates of the peak activations correlating with wanting ratings in general linear model 1 (GLM-1).

We report activations surviving whole-brain FWE correction at peak level (p < 0.05). Hem = Hemisphere (L = left, R = right); BA = Brodmann area.

MNI coordinates
RegionHemBAXYZkZ
VMPFCR/L10044–74397.32
DLPFCL8–213844396.63
L8–33234445.11
PCCL23-3–3738395.29
R/L230–133514.79
Anterior cingulate cortexR326351135.27
Frontopolar cortexL10–12652014.82
Table 4
Anatomical locations and MNI coordinates of the peak activations correlating with liking ratings in general linear model 1 (GLM-1).

We report activations surviving whole-brain FWE correction at peak level (p < 0.05). Hem = Hemisphere (L = left, R = right); BA = Brodmann area.

MNI coordinates
RegionHemBAXYZkZ
Dorsal PCCL31–9–643814.87

GLM-1 revealed no significant wanting- or liking-related striatal activation, which may appear surprising given the canonical role of the striatum for reward processing (Bartra et al., 2013). However, this might be due to the fact that the parametric modulators for wanting and liking only explained unique variance (regressors were not orthogonalized), while striatal activation might be shared by wanting and liking. To test this, we computed two further GLMs, one (GLM-3) where we orthogonalized liking with respect to wanting (such that the regressor for wanting explained the variance shared by wanting and liking) and one where we orthogonalized wanting with respect to liking (GLM-4). In GLM-3, we observed bilateral wanting-related activation in the striatum (z = 7.09, whole-brain FWE-corrected, p < 0.001, peak = [–9 11 –1]), PCC (z = 12.22, whole-brain FWE-corrected, p < 0.001, peak = [0 –28 25]), VMPFC (z = 11.13, whole-brain FWE-corrected, p < 0.001, peak = [0 47 –7]), posterior parietal cortex (z = 8.42, whole-brain FWE-corrected, p < 0.001, peak = [–45 –67 35]), and DLPFC (z = 5.59, whole-brain FWE-corrected, p < 0.001, peak = [24 32 47]). Likewise, in GLM-4 liking ratings (including the variance shared with wanting) correlated with activation in striatum (z = 6.19, whole-brain FWE-corrected, p < 0.001, peak = [–9 14 –1]), PCC (z = 10.95, whole-brain FWE-corrected, p < 0.001, peak = [0 –31 35]), DLPFC (z = 7.77, whole-brain FWE-corrected, p < 0.001, peak = [–18 32 50]), VMPFC (z = 7.69, whole-brain FWE-corrected, p < 0.001, peak = [–3 50 –4]), and posterior parietal cortex (z = 6.13, whole-brain FWE-corrected, p < 0.001, peak = [–45 –67 35]). Thus, both wanting and liking correlated with activation in regions belonging to the neural reward system. However, also in the GLMs with orthogonalized parametric modulators, we observed no effects of naltrexone or amisulpride (relative to placebo) on activations related to wanting (GLM-3) or liking (GLM-4) ratings even at lenient statistical thresholds (p < 0.001 uncorrected, cluster size >20 voxels).

Previous research showed that wanting-related prefrontal activation is functionally coupled with the ventral striatum depending on the behavioral relevance of wanting judgements (Weber et al., 2018). Consistent with our previous finding, striatal activation was significantly correlated with wanting ratings when those were behaviorally relevant (wanting ratings on wanting trials in GLM-2), z = 4.46, p = 0.003, peak = [–6 11 –1], small-volume FWE-corrected with anatomical mask for the striatum (Figure 2C). We note that wanting-related striatal activation did not survive FWE correction at the whole-brain peak level (p = 0.14, although it did survive whole-brain FWE correction at the cluster level, p < 0.001), such that this effect appears to be somewhat weaker than in our previous study (Weber et al., 2018).

We next assessed whether wanting-related prefrontal regions are functionally connected with the striatum by conducting a psychophysiological interaction (PPI) analysis with the striatum as seed region. To test whether the pharmacological manipulations changed the functional connectivity of the striatum with specifically wanting-related brain regions, we applied SVC within a mask of significant wanting-correlated voxels in DLPFC and VMPFC in GLM-1 (thresholded with FWE at peak level, k = 478). On wanting trials, we observed enhanced functional coupling between striatum and DLPFC as a function of increasing wanting ratings (wanting ratings on wanting trials: z = 3.87, small-volume FWE-corrected, p = 0.02, peak = [–21 41 41]), as to be expected given that both the DLPFC and the striatum showed significant wanting-related activity. Moreover, DLPFC-striatum connectivity on wanting trials was stronger for wanting than for liking ratings (wanting > liking ratings on wanting trials: z = 3.64, small-volume FWE-corrected, p = 0.04, peak = [–21 41 41]) (Figure 3A). The wanting-dependent DLPFC-striatum coupling is consistent with previous findings that connectivity between the stratum and prefrontal correlates of wanting depends on whether wanting judgements are behaviorally relevant (Weber et al., 2018).

Effects of Judgement type and drug on parametric striatal connectivity.

(A) On wanting trials (collapsed across drug groups), dorsolateral prefrontal cortex (DLPFC)-striatum connectivity was enhanced for wanting relative to liking aspects of rewards (image thresholded at p < 0.001 uncorrected). (B) Wanting-related DLPFC-striatum coupling was significantly stronger under naltrexone compared with placebo (image thresholded at p < 0.001 uncorrected). (C, D) Extracted parameter estimates for DLPFC (as defined by the significant cluster in general linear model 1 [GLM-1]), separately for wanting and liking judgements. (C) If wanting judgements were behaviorally relevant, naltrexone increased wanting relative to liking-related DLPFC-striatum connectivity. (D) No significant drug effects on DLPFC-striatum connectivity were observed on liking trials. Error bars indicate standard error of the mean, black dots represent individual data points. *p < 0.05.

Next, we tested how our pharmacological manipulation changed functional connectivity between the striatum and wanting-related cortical regions. Compared with placebo, naltrexone increased DLPFC-striatum coupling for wanting relative to liking ratings on wanting trials (naltrexone > placebo for wanting > liking ratings on wanting trials: z = 3.81, small-volume FWE-corrected, p = 0.02, peak = [–18 35 38]) (Figure 3B). Moreover, the impact of naltrexone on DLPFC-striatum connectivity was significantly stronger on wanting than on liking trials, ((wanting > liking ratings)wanting trials > (wanting > liking ratings)liking trials-related connectivity in the naltrexone relative to the placebo group: z = 3.55, small-volume FWE-corrected, p = 0.05, peak = [–18 35 38]). Thus, the effects of naltrexone on frontostriatal connectivity were specific for wanting judgements. We observed no further regions showing significantly reduced wanting-related connectivity under naltrexone relative to placebo, and we also observed no significant differences between amisulpride and placebo as well as naltrexone and amisulpride. Thus, reducing opioid neurotransmission strengthened the functional connection between the striatum and prefrontal cortex when wanting judgements were behaviorally relevant.

The observed drug effects on the neural level raise the question whether the impact of naltrexone on behavioral wanting ratings can statistically be explained by its influence on DLPFC-striatum connectivity. For this purpose, we extracted parameter estimates from the significant DLPFC cluster for the (wanting > liking ratings)wanting trials-contrast and tested whether the impact of naltrexone on wanting ratings is mediated by the naltrexone effects on DLPFC-striatum connectivity (using quasi-Bayesian confidence intervals as implemented in the mediation package for R; Tingley et al., 2014). However, the test of the significance of the indirect path (which is the decisive criterion for the presence of a mediation effect; Zhao et al., 2010) showed no significant effect, ACME = –0.18, p = 0.06. Thus, the data do not provide sufficient evidence that naltrexone effects on behavior can be explained with modulation of frontostriatal connectivity. We also note that the main effect of naltrexone on wanting ratings remained significant when controlling for DLPFC-striatum connectivity, β = –8.48, t(76) = 2.54, p = 0.01, while DLPFC-striatum connectivity was stronger for more highly wanted items in the placebo group, β = 586, t(73) = 2.30, p = 0.02. We must therefore be careful with interpreting the naltrexone effects on brain connectivity as the potential cause for the behavioral drug effects.

For completeness, despite having observed no significant drug effects on liking in the behavioral analysis, we also performed whole-brain analyses assessing which brain regions show enhanced functional connectivity as function of liking ratings. On liking trials, connectivity with the striatum was stronger for liking than for wanting ratings in OFC, (z = 3.16, small-volume FWE-corrected, p = 0.05, peak = [–33 50–7]), replicating our previous findings. However, no brain regions showed significant effects of naltrexone or amisulpride (relative to placebo) on liking-related connectivity with the striatum even at low, exploratory statistical thresholds (p < 0.001, cluster size >20 voxels).

To assess the robustness of the naltrexone effects on wanting-related DLPFC-striatum connectivity, we extracted parameter estimates from the significant wanting-related DLPFC cluster in GLM-1 and regressed the parameter estimates on predictors for Drug, Judgement, Relevance, and the interaction terms (using the lmer function in R). The significant Naltrexone× Judgement × Relevance interaction, β = 6.0e-03, t(452) = 2.46, p = 0.01, replicated the finding that naltrexone had dissociable effects on wanting and liking as a function of the behavioral relevance of these reward components. We also observed a significant Amisulpride × Relevance interaction, β = 4.1e-03, t(452) = 2.42, p = 0.02. Separate analyses for wanting and liking judgements revealed that in wanting trials DLPFC-striatum connectivity was enhanced for wanting compared with liking ratings under both naltrexone, Naltrexone× Relevance interaction, β = 5.6e-03, t(226) = 3.19, p = 0.002, and amisulpride, Amisulpride × Relevance interaction, β = 4.1e-03, t(226) = 2.35, p = 0.02 (Figure 3C). In contrast, liking judgements showed no significant effects of naltrexone or amisulpride relative to placebo, all t < 1.34, all p > 0.18 (Figure 3D). This result supports our findings based on SVC according to which naltrexone increases wanting-related relative to liking-related DLPFC-striatum connectivity on wanting trials, and hints at a similar function for amisulpride (though this was not evident in the SVC-based analysis).

Discussion

In animal models, the wanting and liking dimensions of rewards are processed by partly distinct brain regions and neurotransmitter systems, but in humans it remained unclear so far how opioidergic and dopaminergic systems orchestrate the processing of wanting and liking. The current findings provide evidence for dissociable roles of opioidergic neurotransmission in processing the two dimensions of rewards on both a behavioral and a neural level. Behaviorally, lowering opioidergic activation with naltrexone selectively reduced wanting, not liking, ratings for non-consumable goods. On a neural level, this reduction in wanting was reflected by changes in DLPFC-striatum connectivity: When wanting judgements were required, DLPFC-striatum connectivity was significantly stronger for the behaviorally relevant wanting ratings than for the irrelevant liking dimension of rewards.

Importantly, wanting-related functional coupling between DLPFC and striatum was significantly stronger under naltrexone than under placebo. This is consistent with recent findings relating opioid receptor inhibition with increased connectivity between the prefrontal control system and reward circuits (Elton et al., 2019; Lim et al., 2019) and suggesting prefrontal kappa opioid receptors to mediate the impact of naltrexone on drug craving in alcohol use disorder (de Laat et al., 2019). Through corticostriatal loops, the striatum receives input from several cortical regions and can prioritize processing of behaviorally relevant information (Frank, 2011). DLPFC provides inhibitory input to the striatum and was shown to reduce wanting-related activation in the striatum (Dong et al., 2020; Koob and Volkow, 2010), consistent with the importance of frontostriatal loops for self-control (van den Bos et al., 2014). According to this view, the observed positive relationship between wanting ratings and DLPFC-striatum connectivity might indicate that DLPFC exerts top-down control over striatal representations of wanting predominantly for highly wanted items, whereas there may be less need for inhibitory top-down control for less desired goods (van den Bos et al., 2014). By strengthening DLPFC-striatum connectivity, naltrexone enhances top-down processes predominantly for highly wanted items, which would explain why lower wanting under naltrexone is associated with stronger DLPFC-striatum connectivity even though in the placebo group functional DLPFC-striatum coupling is increased for highly wanted items. In any case, because we observed no significant brain-behavior mediation effect (p = 0.06), the stronger DLPFC-striatum connectivity under naltrexone should not be interpreted as the cause for the changes in behavioral ratings under naltrexone. It is further worth noting that in the current study we observed significant effects predominantly in the left hemisphere. In the literature on frontostriatal connectivity during reward processing, both left- and right-lateralized effects were reported (van den Bos et al., 2014; van den Bos et al., 2015; Yuan et al., 2017). We therefore do not make any claims regarding whether this result pattern represents just a power issue or a truly lateralized effect.

Contrary to our hypotheses, we did not observe effects of naltrexone on liking or amisulpride effects on wanting. Interestingly, however, a recent study observed no influences of opioid and dopamine antagonists on self-report wanting and liking ratings but only on experimental measures of these reward dimensions (Korb et al., 2020). In fact, previous studies reporting effects of dopaminergic manipulations on wanting operationalized wanting with experimental measures rather than self-report (Soutschek et al., 2020a; Soutschek et al., 2020b; Weber et al., 2016), while studies using self-report ratings observed no or only weak effects of pharmacological manipulations (Case et al., 2016; Ellingsen et al., 2014; Løseth et al., 2019). In line with our previous study (Weber et al., 2018), we had decided to use self-report ratings to avoid the issue that implicit measures of liking such as face muscle activity are more open to alternative interpretations (Pool et al., 2016), but we acknowledge that implicit measures might be more sensitive to pharmacological interventions. Moreover, given that in our study participants had to provide liking ratings without being able to actually consume or handle the items, our measurements might have been less sensitive than those of other studies assessing the liking of consumed rewards (Buchel et al., 2018; Chelnokova et al., 2014; Eikemo et al., 2016). It is also worth noting that it has recently been suggested that amisulpride shows, if any, only weak effects on BOLD signal changes in the reward system (Grimm et al., 2020). We note though that in the same sample of participants amisulpride showed significant effects on tasks for cue reactivity and delay discounting (Weber et al., 2016), which were administered 2.5 hr after drug intake (while the rating task started 1 hr after drug intake). Due to this difference in timing, it is thus not possible to decide whether the different amisulpride effects on these tasks can be explained by different sensitivities of these tasks to dopaminergic manipulations or by the time course of amisulpride effects. In any case, one should thus be careful with interpreting these unexpected null findings as being inconsistent with previous pharmacological results manipulating dopaminergic activity with different compounds than amisulpride.

Interestingly, the ROI analysis provided some evidence for amisulpride effects on wanting at the neural level, as amisulpride increased wanting-related DLPFC-striatum connectivity, similar to the findings for naltrexone. However, the impact of amisulpride on wanting-related frontrostriatal connectivity (gating) needs to be interpreted with caution, given the lack of significant amisulpride effects on behavior.

Our findings have important implications for clinical research, given that dysfunctions in wanting and liking are prevalent in several psychiatric disorders. Substance use disorders, for example, are characterized by increased wanting of drugs as reflected in craving symptoms (Berridge, 2012; Edwards, 2016), and craving has been linked to impairments in prefrontal top-down control over the striatum (Feil et al., 2010). Naltrexone is approved in several countries for the treatment of alcohol use disorder (Krystal et al., 2001; Srisurapanont and Jarusuraisin, 2005) and opioid dependence (Johansson et al., 2006) and was shown to reduce relapse risk and craving specifically in alcohol use disorder. Consistent with the view that naltrexone reduces the salience of drug cues by strengthening prefrontal activation (Courtney et al., 2016), we speculate that the beneficial effects of naltrexone on alcohol use and craving might be explained by increased top-down control of DLPFC over striatal wanting signals as a consequence of opioid receptor inhibition (but see Nestor et al., 2017). Our results may thus improve the understanding of neural mechanisms underlying pharmacological treatments of dysfunctional wanting in substance use disorders.

Several limitations are worth mentioning. First, we did not assess wanting and linking prior to drug administration, such that we cannot control for potential baseline differences in wanting and liking between drug groups. Thus, it remains possible that the non-significant effects of amisulpride on wanting and of naltrexone on liking are caused by such pre-existing baseline differences, or that the sample size was not sufficient to detect these effects in a between-subject design. We also note that the doses for amisulpride and naltrexone might not have been pharmacologically equivalent. In fact, while 50 mg naltrexone produces 95% μ-opioid receptor occupancy (Weerts et al., 2008), 400 mg amisulpride leads to a lower dopamine receptor occupancy of 85% (Lako et al., 2013), which might be a further reason for why naltrexone showed stronger effects on behavior and brain activation than amisulpride. Lastly, while high doses of amisulpride (≥400 mg) reduce postsynaptic dopaminergic signaling, lower doses of amisulpride increase dopaminergic activity via presynaptic mechanisms (Schoemaker et al., 1997), but higher doses may also increase signaling at D1 receptors and thereby counteract the inhibitory effects on D2 neurotransmission. As the effective dose of amisulpride might differ between participants (Sescousse et al., 2016), one might argue that presynaptic and postsynaptic effects of dopamine might have canceled out across participants, leading to the observed null effect of amisulpride on behavior on the group level. However, contrary to this view, we observed no significant amisulpride effects even when controlling for body weight as proxy for effective dose, and we note that in previous studies we had observed effects of 400 mg amisulpirde on behavior (Burke et al., 2018; Soutschek et al., 2017) and multivariate neural data (Kahnt et al., 2015). It seems thus unlikely that the null effects of amisulpride on the rating task can be explained solely by the chosen dosage.

Taken together, our findings deepen our understanding of the neurochemical mechanisms mediating the impact of wanting of rewards on behavior. Opioid receptors are involved in the modulation of the strength of inhibitory prefrontal input to the striatum encoding the behavioral relevance of the wanting dimension of rewards. These insights into the interactions between neuroanatomical and neurochemical brain mechanisms implementing wanting-driven approach behavior advance our understanding of the mechanisms underlying pharmacological treatments of substance use disorders.

Materials and methods

Participants

A total of 121 healthy volunteers (58 females; Mage = 21.8 years, range = 18–30), recruited via email from the internal pool of the Laboratory for Social and Neural Systems Research (which includes mainly students from the University of Zurich and ETH Zurich), participated in the study. According to power analysis assuming the effect size (Cohen’s d = 0.65) from a previous study in our lab on the impact of amisulpride on value representations in the neural reward system (Kahnt et al., 2015), 38 participants per group allow detecting a significant effect (alpha = 5%) with a power of 80%. The goal of the power analysis was to optimize the sample size for finding drug effects on neural reward signals. However, given the differences between the current study design and the study by Kahnt et al., 2015, we note that the power might not have been optimal for all statistical tests in the current investigation (e.g., drug effects on explicit ratings or functional connectivity). Three participants were excluded from the analysis due to response omissions in more than 30% of all trials in the rating task (see below), two further participants were excluded because of excessive head movement (>5 mm in one of the six head motion parameters) in the scanner. Thus, the final sample comprised 116 participants (placebo: N = 39; naltrexone: N = 37; amisulpride: N = 40). Drug groups were matched with regard to age (p = 0.40), sex (p = 0.34), years of education (p = 0.45), and BMI (p = 0.29). Participants were screened prior to participation for exclusion criteria including history of brain disease or injury, surgery to the head or heart, and neurological or psychiatric diseases (including alcohol use disorder, depression, schizophrenia, bipolar disorders, claustrophobia, or Parkinson symptoms) via paper-pencil questionnaires. Further exclusion criteria were a severe medical disease such as diabetes, cancer, insufficiency of liver or kidneys, acute hepatitis, high or low blood pressure, any cardiovascular incidences, epilepsy, pregnancy or breastfeeding, past use of opiates or other drugs that may interact with amisulpride or naltrexone (such as stimulants). A qualitative drug urine screening test (M-10/5-DT, Diagnostik Nord, Schwerin, Germany) was performed to rule out illicit drug use prior the test session (amphetamines, barbiturates, buprenorphine, benzodiazepines, cannabis, cocaine, MDMA, methadone, and morphine/opiates). All participants provided written informed consent. For their participation, they received 40 Swiss francs per hour. The study was approved by the Ethics Committee of the Canton of Zurich and was part of a larger project where we investigated also pharmacological effects on Pavlovian-to-instrumental transfer and delay discounting (published in Weber et al., 2016). These tasks were administered after the rating task reported in the current manuscript (3 hr after drug intake). The larger project (though not the rating task) was preregistered on https://www.clinicaltrials.gov/ (NCT02557984).

Procedure

Request a detailed protocol

Participants received a pill containing either placebo (N = 40), 400 mg amisulpride (N = 41) or 50 mg naltrexone (N = 40) in a randomized and double-blind manner 3 hr before performance of the experimental tasks. Amisulpride is a selective dopamine D2/D3 receptor antagonist, whereas naltrexone is an unspecific opioid receptor antagonist that acts primarily on the μ- and κ-opioid receptors, with lesser and more variable effects on δ-opioid receptors (Rosenzweig et al., 2002; Weerts et al., 2008). We asked participants to fast for 6 hr before arrival at the lab. One hour after drug intake, participants started the wanting/liking rating task (see below) in the MRI scanner, which took approximately 90 min. The (first) peak in plasma concentration for amisulpride is after 60 min (Rosenzweig et al., 2002), whereas for naltrexone the peak is after 120 min (Verebey et al., 1976), such that participants performed the rating task around peaks in average plasma concentration. After task completion, participants answered post-experimental questionnaires, which probed whether they thought they had received a drug or placebo, and measured their mood (one rating was not recorded in the placebo group). We determined amisulpride and naltrexone blood plasma levels immediately before and after the behavioral tasks with high-performance liquid chromatography–mass spectrometry in order to control for the pharmacokinetics of the drugs.

Task design

Request a detailed protocol

Participants performed a task in which they had to rate how much they wanted or liked 40 non-consumable everyday items (Weber et al., 2018). Before performing the rating task in the scanner, we familiarized participants with the items by physically presenting all items to them. The rating task was implemented in Matlab (The MathWorks, Natick, MA) and the Cogent 2000 toolbox. We asked participants to rate each item according to how much they wanted to have it, as well as how much they liked the item at that moment. In each trial, participants first saw a cue indicating the type of rating (wanting or liking) (1 s), followed by an image of the item (3 s), and finally the rating screen (3.5 s). Ratings were provided on a continuous scale using a trackball. Trials were separated by a variable intertrial interval (mean 3 s). Each item was rated twice for wanting and twice for liking, resulting in 160 trials split into four runs before the game (pre-test) and four runs after the game (post-test). Between the pre-test and post-test experimental sessions, participants played a game inside the scanner in which they could win the items. The game consisted of a perceptual task in which participants had to indicate whether the item was presented to the left or the right of the midpoint of the screen. Participants won items that they classified correctly. The difficulty of the game was calibrated such that participants won and lost 50% of the items.

MRI data acquisition and preprocessing

Request a detailed protocol

Whole-brain scanning was performed with a Philips Achieva 3T whole-body MRI scanner equipped with an eight-channel head coil (Philips, Amsterdam, The Netherlands). For each of the eight scanning runs, 227 T2*-weighted whole-brain EPI images were acquired in ascending order. Each volume consisted of 33 transverse axial slices, using field of view 192 mm × 192 mm × 108 mm, slice thickness 2.6 mm, 0.7 mm gap, in-plane resolution 2 mm × 2 mm, matrix 96 × 96, repetition time (TR) 2000 ms, echo time (TE) 25 ms, flip angle 80°. Additionally, a T1-weighted turbo field echo structural image was acquired for each participant with the same angulation as applied to the functional scans (181 slices, field of view 256 mm × 256 mm × 181 mm, slice thickness 1 mm, no gap, in-plane resolution 1 mm × 1 mm, matrix 256 × 256, TR 8.4 ms, TE 3.89 ms, flip angle 8°).

Preprocessing was performed with SPM 12 (https://www.fil.ion.ucl.ac.uk/spm/). The functional images of each participant were motion corrected, unwarped, slice-timing corrected (temporally corrected to the middle image), and co-registered to the anatomical image. Following segmentation, we spatially normalized the data into standard MNI space. Finally, data were smoothed with a 6 mm FWHM Gaussian kernel and high-pass filtered (filter cutoff = 128 s).

Behavioral data analysis

Request a detailed protocol

Behavioral data in the rating task were analyzed with mixed general linear models (MGLMs) using the lme4 package in R. The alpha threshold was set to 5% (two-tailed). Degrees of freedom and p-values were computed using the Satterthwaite approximation with the lmerTest package. To replicate our previous findings that winning versus losing items has dissociable effects on wanting and liking ratings, we regressed item-specific ratings in the post-test session on fixed-effect predictors for Judgement (wanting versus liking), Item type (lost versus won), z-transformed item-specific ratings in the pre-test, and all interaction terms (MGLM-1). All these predictors were also modeled as random slopes in addition to participant-specific random intercepts. We also performed separate analyses for wanting and liking ratings (MGLM-2) where post-test item-specific ratings were predicted by Item type and Pre-test ratings.

To assess drug effects on wanting and liking ratings, we regressed session- and item-specific ratings on fixed-effect predictors for Drug (amisulpride versus placebo and naltrexone versus placebo), Judgement, Session (pre-test versus post-test), and the interaction effects (MGLM-3). All fixed effects varying on the individual level (i.e., Judgement, Session, and Judgement × Session) were also modeled as random effects in addition to participant-specific intercepts. Again, we performed separate analyses for wanting and liking (MGLM-4), which were identical to MGLM-3 but left out all predictors for Judgement.

MRI data analysis

Request a detailed protocol

To investigate drug effects on neural activation related to wanting and liking ratings, we computed two GLMs, following previous procedures (Weber et al., 2018). GLM-1 included an onset regressor for the presentation of the current item and the rating bar (duration = 7 s). This onset regressor was modulated by three mean-centered parametric modulators, that is, mean session-specific and item-specific wanting and liking ratings as well as decision times (to control for choice difficulty). The mean item-specific correlation between wanting and liking ratings was r = 0.71. We did not orthogonalize the parametric modulators, such that the results for the regressors reflect the unique variance explained by wanting or liking ratings. A separate regressor modeled all items for which no session- and item-specific value could be computed due to response omissions. GLM-2 was identical to GLM-1 with the only difference that it included separate onset regressors for wanting and liking trials. In GLM-2, the onset regressors for wanting and liking trials were modulated by parametric modulators for both wanting and liking ratings, which allowed assessing judgement-specific (e.g., wanting ratings on wanting trials) and judgement-unspecific (e.g., liking ratings on wanting trials) neural correlates of wanting and linking (Weber et al., 2018). Finally, we computed two further models, one where the liking regressor in GLM-1 was orthogonalized with respect to wanting (such that the regressor for wanting contained the variance shared by wanting and liking; GLM-3) and one where wanting was orthogonalized with respect to liking (GLM-4). In all models, the regressors were convolved with the canonical hemodynamic response function in SPM. We also added six movement (three translation and three rotation) parameters as covariates of no interest.

For statistical analysis, we first computed the following participant-specific contrasts: For GLM-1, we computed parametric contrasts for wanting ratings and liking ratings (independently of judgement type) in GLM-1. For the second-level analysis, we entered the contrast images from all participants in a between-participant, random effects analysis to obtain statistical parametric maps. First, we investigated the neural correlates of wanting and liking independently of administered drug and conducted whole-brain second-level analyses using one-sample t-tests. To assess drug effects, we employed second-level independent t-tests for naltrexone versus placebo as well as amisulpride versus placebo. For these analyses, we report results that survive whole-brain family-wise error corrections at the peak level. In the figures, we set the individual voxel threshold to p < 0.001 with a minimal cluster extent of k ≥ 20 voxels. Results are reported using the MNI coordinate system.

PPI analysis

Request a detailed protocol

To examine how our pharmacological manipulations modulated the frontostriatal connectivity of wanting and liking, we conducted a whole-brain PPI analysis with the striatum as seed region. We defined the seed region by building a sphere (diameter = 6 mm) around the coordinates of wanting-related striatum activation in GLM-2 (MNI coordinates: x = –6, y = 11, z = –1). To create the regressors for the PPI analysis, we first extracted the average time course from the seed region for each individual participant (physiological regressor). We then multiplied the physiological regressor with psychological regressors for (i) wanting ratings on wanting trials, (ii) liking ratings on wanting trials, (iii) liking ratings on liking trials, and (iv) wanting ratings on liking trials. Next, we computed a GLM (PPI-1) that included the interaction terms, the physiological regressor, and the psychological regressors. We also added separate onset regressors for wanting and liking trials as well as movement parameters as regressors of no interest. For the statistical analysis, we computed contrasts for wanting ratings on wanting trials, wanting > liking ratings on wanting trials, liking ratings on liking trials, and finally liking > wanting ratings on liking trials. We submitted these contrasts to a second-level analysis to yield statistical parametric maps with a one-sample t-test. Because GLM-1 revealed wanting ratings to correlate with activation in DLPFC and VMPFC, we tested whether the striatum seed region shows functional connectivity with DLPFC and VMPFC. For this purpose, we applied SVC with a mask that included the significant wanting-correlated voxels in bilateral VMPFC and DLPFC from GLM-1 (thresholded with FWE correction at the voxel level). Additionally, we also performed exploratory whole-brain analyses.

Data availability

The behavioral data that support the findings of this study are available on Open Science Framework https://osf.io/6cevt/.

The following data sets were generated
    1. Soutschek A
    (2021) Open Science Framework
    ID 6cevt. Wanting_liking_pharma_fmri.

References

    1. Schoemaker H
    2. Claustre Y
    3. Fage D
    4. Rouquier L
    5. Chergui K
    6. Curet O
    7. Oblin A
    8. Gonon F
    9. Carter C
    10. Benavides J
    11. Scatton B
    (1997)
    Neurochemical characteristics of amisulpride, an atypical dopamine D2/D3 receptor antagonist with both presynaptic and limbic selectivity
    The Journal of Pharmacology and Experimental Therapeutics 280:83–97.

Decision letter

  1. Jonathan Roiser
    Reviewing Editor; University College London, United Kingdom
  2. Michael J Frank
    Senior Editor; Brown University, United States
  3. Guillaume Sescousse
    Reviewer; Inserm, France

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

Decision letter after peer review:

Thank you for submitting your article "Opioid antagonism reduces wanting by strengthening frontostriatal connectivity" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Michael Frank as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Guillaume Sescousse (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted the text below to help you prepare a revised submission.

The reviewers agreed that this was an ambitious study and that with appropriate analysis and revisions the results are likely to interest a broad audience due to the translational interest in how dopamine and opioid systems affect liking and wanting of rewards across species. They felt that the observation that blocking opioid receptors specifically reduces the wanting of rewards, possibly via modulating frontostriatal coupling, is a potentially important result; and that the methodology underpinning the results was generally appropriate and solid. However, they also commented that the current version of the manuscript lacks much necessary methodological information, and that a substantial rewrite including considerable additional information and some further analysis is necessary.

We are therefore issuing a revise and resubmit decision, but with no guarantee that the revised submission will be accepted as the reviewers' confidence in the findings will in large part rest on your responses to their queries.

Essential revisions:

Introduction

1) The paper appears to rest heavily on a single neuroimaging paper from the same group, and some prior findings from that study are replicated, albeit with a very small effect size. It neglects much of the rest of the relevant literature however, e.g. in the numerous unreferenced claims in the Introduction.

Methods

2) Please specify the urine tox screen used. Please explain how participants were recruited and identified as being free of exclusion criteria. For example, was a formal, semi-structured clinical interview for psychiatric disorders conducted?

3) The authors write that wanting and liking ratings were elicited twice (before and after fMRI), yet analyses appear to indicate that these judgements were also elicited during fMRI? A number of other details about the task are missing, such as the rationale for selection of items, whether these were actually won or hypothetically won, why the 'difficulty' was set to 50%, and why only explicit ratings were included.

4) The authors report that they did not orthogonalize the parametric modulators modelling the wanting and liking ratings. While this makes sense as a way to ensure that each modulator captures unique variance, presumably wanting and liking ratings are tightly correlated. Was that the case? Could the authors provide some details about the average correlation between the ratings?

5) The reviewers appreciated the power analysis. However some details are missing so that the numbers provided can be fully appreciated by the readers. In particular, the study cited as a reference for effect sizes (Kahnt et al., 2015) used a different task and different analyses, which makes it unclear how the authors were able to derive a relevant effect size estimate. Also, since there are many effects that can be tested in the present study (i.e. various behavioral and brain-related effects and correlations), it is important to report explicitly for which of these effects sample size was optimized.

6) The reviewers also appreciated the pre-registration on clinicaltrials.gov. However again some details seem to be missing. The pre-registration mentions a PIT task and a delay discounting task, but no rating task. This discrepancy requires explanation. Furthermore, the pre-registration implies that this study was embedded in a larger project involving more tasks than the one reported here. It is important to report this information explicitly, and explain how the rating task took place in the context of this larger protocol.

Results

7) The reviewers queried the statistics presented in the analysis of the behavioural data. In Figure 1, the amisulpride group shows clearly elevated ratings of wanting and liking, at a magnitude similar to the effect of naltrexone on wanting, but the difference between amisulpride and placebo is reported as non-significant for liking ratings. The statistics reported do not appear to match the means – e.g. in panel D. The difference between naltrexone and placebo for wanting looks similar (with even larger between-subject variance), so it seems surprising that the effect of amisulpride on liking ratings is non-significant. Can the authors verify there is not a mistake? It also seems from Figure 1D that amisulpride might have a main (increasing) effect on ratings: have the authors checked for such a main effect?

8) The authors should explain how the degrees of freedom reported for the contrasts were determined – they often do not appear to match either the number of participants or the number of observations.

9) The description of the Judgement × Item type interaction as "marginally significant" is inappropriate. A quick check with http://statcheck.io/ shows that the two-tailed p-value is 0.11, so the one-tailed p-value is 0.055. Following recent recommendations in the field (e.g. Gibbs and Gibbs, 2015; Otte et al., 2021), especially in a context where some scientists have called for lowering the statistical threshold of p<0.05 for significance (Benjamin et al., 2017), the authors should report this result as non-significant, and/or substantiate it with a complementary statistical approach.

10) On a related point, on p5 a p-value of 0.52 cannot support a claim of the form "wanting ratings did not differ between won and lost items". Such a p-value > 0.05 merely suggests that the data at hand don't provide sufficient evidence to reject the null. Bayesian statistics could potentially be used to make the sort of claims made by the authors. The same issue applies on p. 6: "amisulpride did not change wanting ratings relative to placebo, β = -1.39, t(116) = 0.22, p = 0.83".

11) Ideally one would like to establish a causal mechanism from the present results: is reduced wanting following opioid receptor blockade mediated by a direct action of naltrexone on frontostriatal connectivity (as suggested in the title of this submission)? Have the authors have considered performing a mediation-style analysis?

12) On p9 it would informative to present the striatal activation observed for the correlation with relevant wanting ratings. The statistics are reported for only one unilateral peak voxel, so it is hard to assess how robust this result is. This is particularly important since this result is supposed to be a replication of the authors' previous work (Weber et al., 2018).

13) The authors note that DLPFC-striatum connectivity was associated with self-reported wanting and that naltrexone increased this connectivity further. Since naltrexone also decreased self-reported wanting, are these results contradictory? This needs clarification.

Discussion

14) The authors used an intermediate dose of amisulpride (400 mg), which has been suggested to exert a mixture of pre- and post-synaptic effects depending on the participant, with pre-synaptic blockage likely to increase D1 receptor activation. Therefore any group-wise effect may cancel out across participants (see e.g. van der Schaaf et al., 2012; Sescousse et al., 2016; Eisenegger et al., 2014). Could this be a reason for the limited effects of amisulpride observed in the present study?

15) Some of the key results appear to be unilateral, i.e. specifically in the left hemisphere (striatal activity scaling with wanting ratings and DLPFC-striatum connectivity). Such a unilateral effect could reflect limited statistical power and fragile results, but also a truly lateralized effect. It would be worth commenting on this in the Discussion.

16) Based on previous literature it is surprising that the striatum was not part of the neural correlates of wanting and liking (Figure 2) – striatal activation is only identified when the ratings are restricted to the relevant ones. Yet, based on studies like Barta et al. (2013) or Lebreton et al. (2009) that show that the striatum is at the core of a robust and automatic valuation system, one might expect activity in this region to correlated with ratings even when these are not explicitly required. Could the authors comment on this?

Reviewer #2:

There is still an ongoing debate regarding the exact role of these neurotransmitters in the brain, and this study is very timely given that this debate has been mostly fueled by animal work. The present study builds on previous work from the authors (Weber et al., 2018), replicating some of their fMRI results, and adding a pharmacological manipulation. Using a between-subject, placebo-controlled, double-bling protocol, the authors show that blocking opioid receptors with naltrexone specifically reduces wanting of rewards, while blocking dopamine D2 receptors with amisulpride did not have noticeable effects. Importantly, the authors show that this pharmacological effect was accompanied by an increase in frontostriatal coupling. These results complement the previous study of the authors and strengthen their hypothesis of a frontostriatal gating of motivational – but no hedonic – aspects of reward processing.

This study has a variety a strengths in my opinion. The authors provide some evidence (even if sometimes limited) that some of their previous behavioral and fMRI results replicate. Sample sizes are relatively large for a pharmaco-fMRI study, and the methods are generally sound and well-justified, providing confidence that conclusions are well-supported by the data. The study was preregistered, which also increases faith in the results.

Nonetheless, some aspects could be improved. In particular, some statistical procedures would gain in being clarified or strengthened. Some important information is also sometimes missing, regarding methods and embedding of this study in a larger project. Finally, several points could be discussed in more detail in the Discussion, especially regarding striatal activity and drug dosage.

Reviewer #3:

This labor-intensive study tested contributions of dopamine and opioid systems to reward wanting vs. liking in healthy humans. Strengths include an impressive sample size for a single-center fMRI / pharmacological challenge study (N=116), carefully constructed hypotheses with suitable methods and statistical analyses, and an interesting question. The primary limitation was the use of a selective D2/3 receptor antagonist and dose that likely has complex effects, partially decreasing post-synaptic D2/3 transmission, increasing dopamine release (via blocked autoreceptors), and increasing D1 receptor transmission.

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

Thank you for resubmitting your work entitled "Opioid antagonism modulates wanting-related frontostriatal connectivity" for further consideration by eLife. Your revised article has been reviewed by 3 peer reviewers and the evaluation has been overseen by Michael Frank as the Senior Editor, and a Reviewing Editor.

The reviewers and editors feel that the manuscript has been improved but there are several remaining major issues that need to be addressed, as outlined below. Please note that if these issues are not satisfactorily addressed in your revised submission then unfortunately we will not be able to consider the manuscript further, as it is not editorial practice to issue multiple revise resubmit decisions at eLife.

1) The most important issue is that there remains a discrepancy between similar effect sizes of naltrexone/amisulpride on wanting (albeit in opposite directions) and the corresponding pattern of P-values obtained from the hierarchical analysis. A similar issue is also present in relation to liking ratings. The authors need to explore this discrepancy in considerably more detail and resolve it, as follows:

a) Conduct a non-hierarchical analysis using the mean ratings for wanting and liking (in two separate models, one for wanting, one for liking). The reason for this is that it appears from the data depicted in Figure 1D that amisulpride may increase wanting and also liking (where the effect may actually be even greater). The reviewers noted that for liking ratings the mean difference is ~0.5 points and the SDs are actually lower than for wanting ratings at 1.0/1.2, which is suggestive of a larger effect than the effect of naltrexone on wanting which is significant in the hierarchical model. It would also be useful to provide the standardised effect sizes (Cohen's d) for the 4 comparisons against placebo (2 for wanting, 2 for liking).

b) Assuming that the above analyses using the mean ratings provide a discrepant pattern of significance to the hierarchical analysis, this then needs to be investigated thoroughly and explained in the manuscript, both for wanting and for liking ratings. The authors need to dig into the data carefully and figure out why this discrepancy arises. For example, if amisulpride makes participants more variable in their ratings (or naltrexone make them more consistent), this would be important for interpretation. Or perhaps some assumptions of the hierarchical model have been broken? Or perhaps the covariance structure requires amending? Or perhaps the model did not converge? Without resolution of this discrepancy it will not be possible to consider the manuscript further.

2) The authors now report the magnitude of the correlation between wanting and liking ratings, which is unsurprisingly high (r = 0.71). Since they have not serially orthogonalized the parametric regressors in the main analysis, this means that much of the variance of these ratings is simply removed. For this reason it is not clear how much we can infer from the non-significant drug effects, considering that these were assessed using only a fraction of the ratings variance, which may result in an insensitive analysis. Therefore further analyses are required here to substantiate the conclusion that there were no drug effects (as reported on the top of p10 – it is assumed that currently this refers to the model without serial orthogonalisation, although this should be stated explicitly for clarity).

The authors do provide some results from an analysis using serial orthogonalisation, with liking orthogonalised against wanting (p9), yielding the expected striatal activation for the parametric effect of wanting (which then carries the shared variance), which is reassuring – as they note this suggests that the striatal signal is substantively affected by the collinearity between wanting and liking ratings. Please additionally report the drug effects in this analysis. The authors should also report the effects from an analysis in which the serial orthogonalisation is performed in the alternate order (i.e. wanting against liking, such that the liking regressor now carries the shared variance), including both the main parametric effect (this time of liking) and drug effects.

3) It is helpful that the authors added the information that previous data from the same study were published in a 2016 paper by Weber et al. Oddly they do not mention the results of that paper, even in the discussion of the (apparent – see point 3 below) non-significant effects of amisulpride. The 2016 findings are highly relevant, since the amisulpride group was found to suppress cue-based responding and reward impulsivity. Similar results, but weaker, were reported for naltrexone. Both groups also reported lower mood than the placebo group.

The authors explain that the PIT and delay discounting tasks were completed after the end of scanning, i.e. after 60 minutes absorption time + 90 minutes fmri rating task = minimum 2.5 hours after drug administration. Hence, it seems highly relevant for the interpretation of the present data that, in the exact same participants, the same dose of amisulpride reported to show a null during 1-2.5 hours, showed what are (presumably) expected effects after 2.5 hours. Therefore it is necessary to mention this prior publication from the same study in the Introduction, and discuss the results, especially with respect to dose timing, in the Discussion.

https://doi.org/10.7554/eLife.71077.sa1

Author response

Essential revisions:

Introduction

1) The paper appears to rest heavily on a single neuroimaging paper from the same group, and some prior findings from that study are replicated, albeit with a very small effect size. It neglects much of the rest of the relevant literature however, e.g. in the numerous unreferenced claims in the Introduction.

Following the reviewers’ recommendations, we provide more literature in the introduction section, such that all claims are now backed-up by references and that the current study appears less based exclusively on our previous neuroimaging study.

In particular, we now provide references for the following claims:

– p.3: “Previous animal research suggested that wanting and liking relate to dissociable neurochemical mechanisms (Berridge and Kringelbach, 2015; Berridge and Valenstein, 1991)”

– p.3:“Both dopaminergic and opioidergic neurons project to reward circuits in the striatum as well as to the prefrontal cortex (Delay-Goyet, Zajac, Javoy-Agid, Agid, and Roques, 1987; Lidow, Goldman-Rakic, Gallager, and Rakic, 1991)”

– p.4: “In particular, the ventral striatum encodes the currently behaviorally relevant reward dimension and dynamically switches functional connectivity with wanting- and liking-encoding prefrontal regions accordingly (Weber, Kahnt, Quednow, and Tobler, 2018; switching between frontostriatal loops according to behavioral relevance constitutes one possibility how dopamine (and by extension opioids) can affect functions involving prefrontal cortex and the striatum: Cools, 2011). This is also in line with the view that wanting and liking are preferentially represented in partially distinct subregions of the striatum (which are in turn connected to different input and output regions; Peciña (2008)).”

– p.4: “We hypothesized that, on a behavioral level, opioid receptor blockade will reduce both wanting and liking (Buchel, Miedl, and Sprenger, 2018; Chelnokova et al., 2014; Eikemo et al., 2016), whereas reduced dopaminergic neurotransmission will selectively affect wanting (Cawley et al., 2013; Korb et al., 2020; Venugopalan et al., 2011)”

Methods

2) Please specify the urine tox screen used. Please explain how participants were recruited and identified as being free of exclusion criteria. For example, was a formal, semi-structured clinical interview for psychiatric disorders conducted?

We now clarify which urine drug test we used in our study (p.20):

“A qualitative drug urine screening test (M-10/5-DT, Diagnostik Nord, Schwerin, Germany) was performed to rule out illicit drug use prior the test session (amphetamines, barbiturates, buprenorphine, benzodiazepines, cannabis, cocaine, MDMA, methadone and morphine/opiates).”

Regarding participant recruitment, we clarify that participants were recruited via email from the internal participant pool of the Laboratory for Social and Neural Systems Research, which includes mainly students from the University of Zurich and ETH Zurich (p.19):

“A total of 121 healthy volunteers (58 females; Mage = 21.8 years, range = 18-30), recruited via email from the internal pool of the Laboratory for Social and Neural Systems Research (which includes mainly students from the University of Zurich and ETH Zurich), participated in the study.”

The exclusion criteria were checked by a paper-pencil questionnaire rather than by a formal clinical interview. Participants had to indicate whether they had been diagnosed with a psychiatric disorder in the past. We clarify the procedure on p.19:

“Participants were screened prior to participation for exclusion criteria, including a history of brain disease or injury, surgery to the head or heart and neurological or psychiatric diseases (including alcohol use disorder, depression, schizophrenia, bipolar disorders, claustrophobia or Parkinson symptoms) via paper-pencil questionnaires.”

3) The authors write that wanting and liking ratings were elicited twice (before and after fMRI), yet analyses appear to indicate that these judgements were also elicited during fMRI? A number of other details about the task are missing, such as the rationale for selection of items, whether these were actually won or hypothetically won, why the 'difficulty' was set to 50%, and why only explicit ratings were included.

We apologize for the lack of clarity here. Wanting and liking ratings were collected during the acquisition of fMRI data in the scanner. There were two fMRI sessions. One before and one after a game in which participants won half of the items. We now formulate this more clearly in the manuscript on p.5 (see below). In addition, we clarify that participants actually won the items in the game, i.e., they could take the won items home after the experiment. We therefore selected everyday items that should (at least to some extent) be wanted and liked by the majority of our participants stemming from the student population in Zurich. The “difficulty” of the game was set to 50% in order to have equal numbers of won and lost items for the statistical analysis. We clarify these issues on p.5:

“We collected wanting and liking ratings for all items in the MRI scanner twice, once before (pre-test session) and once after (post-test session) participants played a game on the computer where they won or lost 50% of the items (in order to have equal numbers of won and lost items for the statistical analysis). […] We therefore selected everyday items (e.g., batteries or candles – for the full list of items, see Weber et al. (2018)) that should be both wanted and liked by the majority of our participants from the Zurich student population.”

Lastly, concerning the inclusion of only explicit self-report measures for wanting and liking, we followed our previous study (Weber et al., 2018) and decided to avoid the issue that implicit measures like face muscle activity are arguably harder to interpret than explicit measures. Having said this, we also recognize the possibility that implicit measures might be more sensitive to drug effects than explicit measures of wanting and liking (see Korb et al., 2020). We therefore extended the discussion of this issue on p.16:

“In line with our previous study (Weber et al., 2018), we had decided to use self-report ratings to avoid the issue that implicit measures of liking, such as face muscle activity are more open to alternative interpretations (Pool, Sennwald, Delplanque, Brosch, and Sander, 2016), but we acknowledge that implicit measures might be more sensitive to pharmacological interventions.”

4) The authors report that they did not orthogonalize the parametric modulators modelling the wanting and liking ratings. While this makes sense as a way to ensure that each modulator captures unique variance, presumably wanting and liking ratings are tightly correlated. Was that the case? Could the authors provide some details about the average correlation between the ratings?

In the revised manuscript on p.23, we now report the mean item-specific correlation between wanting and liking ratings, which was r = 0.71.

“The mean item-specific correlation between wanting and liking ratings was r = 0.71. We did not orthogonalize the parametric modulators, such that the results for the regressors indicate the unique variance explained by wanting or liking ratings.”

5) The reviewers appreciated the power analysis. However some details are missing so that the numbers provided can be fully appreciated by the readers. In particular, the study cited as a reference for effect sizes (Kahnt et al., 2015) used a different task and different analyses, which makes it unclear how the authors were able to derive a relevant effect size estimate. Also, since there are many effects that can be tested in the present study (i.e. various behavioral and brain-related effects and correlations), it is important to report explicitly for which of these effects sample size was optimized.

In the revised manuscript, we provide more details regarding the power analysis. For the current power analysis, we had used the impact of amisulpride (versus placebo) on the representation of reward in the OFC reported on page 4107 in the study by Kahnt et al. (2015). However, we agree with the reviewers that the current study used different tasks and additional pharmacological interventions (naltrexone) compared with the Kahnt et al. study. While power analyses assume that a study tries to replicate a previously observed effect, the novelty of our investigation made it impossible to conduct a proper power analysis including all effects on the behavioral and neural level. Instead, we decided to collect a sample size that should be sufficient to observe pharmacological effects on neural reward presentations, but we acknowledge that this represents a best guess for the true required sample size rather than a proper power analysis based on a replication of previous effects.

We discuss this issue in the revised manuscript on p.19:

“According to power analysis assuming the effect size (Cohen’s d = 0.65) from a previous study in our lab on the impact of amisulpride on value representations in the neural reward system (Kahnt, Weber, Haker, Robbins, and Tobler, 2015), 38 participants per group allow detecting a significant effect (α = 5%) with a power of 80%. […] However, given the differences between the current study design and the study by Kahnt et al. (2015), we note that the power might not have been optimal for all statistical tests in the current investigation (e.g., drug effects on explicit ratings or functional connectivity).”

6) The reviewers also appreciated the pre-registration on clinicaltrials.gov. However again some details seem to be missing. The pre-registration mentions a PIT task and a delay discounting task, but no rating task. This discrepancy requires explanation. Furthermore, the pre-registration implies that this study was embedded in a larger project involving more tasks than the one reported here. It is important to report this information explicitly, and explain how the rating task took place in the context of this larger protocol.

We thank the reviewer for pointing to this important issue. The current study was indeed part of a larger project where we had investigated also the effects of naltrexone and amisulpride on a PIT task and a delay discounting task. These two tasks were administered outside the scanner, after the wanting/liking rating task in the scanner, such that they cannot have affected performance in our rating task. The results for the PIT and delay discounting tasks have been published (Weber et al., 2016, Translational Psychiatry). Re-checking the preregistration on clinicaltrials.gov, we realized that indeed only the PIT and delay discounting tasks had been preregistered, not the rating task. We apologize for this mistake and now clarify in the revised manuscript that the current sub-project was not mentioned in the pre-registration (p.20).

“The study was approved by the ethics committee of the canton of Zurich and was part of a larger project where we investigated also pharmacological effects on Pavlovian-to-instrumental transfer and delay discounting (published in Weber et al. (2016)). These tasks were administered after the rating task reported in the current manuscript. The larger project (though not the rating task) was preregistered on www.clinicaltrials.gov (NCT02557984).”

Results

7) The reviewers queried the statistics presented in the analysis of the behavioural data. In Figure 1, the amisulpride group shows clearly elevated ratings of wanting and liking, at a magnitude similar to the effect of naltrexone on wanting, but the difference between between amisulpride and placebo is reported as non-significant for liking ratings. The statistics reported do not appear to match the means – e.g. in panel D. The difference between naltrexone and placebo for wanting looks similar (with even larger between-subject variance), so it seems surprising that the effect of amisulpride on liking ratings is non-significant. Can the authors verify there is not a mistake? It also seems from Figure 1D that amisulpride might have a main (increasing) effect on ratings: have the authors checked for such a main effect?

We thank the reviewers for these important comments. We checked the statistics and found them to be correct as reported in the manuscript. However, we also understand the reviewers’ impression that e.g. Figure 1D suggests an influence of amisulpride on liking, even though this effect is not significant in the statistics. This can be explained by the fact that Figure 1 shows the mean wanting and liking ratings for each participant (collapsed across items), while the mixed linear models statistically analyze ratings for each item and participant (with participant-specific and item-specific random intercepts). The means and standard errors in the figures might therefore occasionally give an impression that deviates from the actual statistical results. To avoid misunderstanding, we added a statement to the legend for Figure 1 clarifying that the figure displays the mean wanting and liking ratings for illustration purpose, while the statistical analyses are based on participant- and item-specific ratings.

“For illustration purposes, participant-specific mean wanting/liking ratings are plotted on a scale from 0 to 10, while the statistical analyses are conducted on the participant- and item-specific wanting and liking ratings.”

8) The authors should explain how the degrees of freedom reported for the contrasts were determined – they often do not appear to match either the number of participants or the number of observations.

We now clarify that degrees of freedom were computed using the Satterthwaite approximation (p.22), which is the default option in the lmerTest package. We note that with the Satterthwaite approximation the degrees of freedom depend on the variance within samples, such that degrees of freedom can differ even for tests including the same number of observations or participants.

“Degrees of freedom and p-values were computed using the Satterthwaite approximation with the lmerTest package.”

9) The description of the Judgement × Item type interaction as "marginally significant" is inappropriate. A quick check with http://statcheck.io/ shows that the two-tailed p-value is 0.11, so the one-tailed p-value is 0.055. Following recent recommendations in the field (e.g. Gibbs and Gibbs, 2015; Otte et al., 2021), especially in a context where some scientists have called for lowering the statistical threshold of p<0.05 for significance (Benjamin et al., 2017), the authors should report this result as non-significant, and/or substantiate it with a complementary statistical approach.

We apologize for the lack of clarity here. It is correct that the precise two-tailed p-value is 0.1087 (0.054 one-tailed). We used the term “marginally significant” to indicate that the p-value is close to the statistical threshold (other fields, e.g., economics use p=0.1), but following the reviewers’ recommendations we now refer to this effect as non-significant trend-level effect (p.6):

“we observed a non-significant but trend-level effect for the Judgement × Item type interaction, β = 0.50, t(111) = 1.61, p = 0.054, one-tailed, suggesting that winning versus losing items tended to have dissociable effects on wanting versus liking of the items”.

10) On a related point, on p5 a p-value of 0.52 cannot support a claim of the form "wanting ratings did not differ between won and lost items". Such a p-value > 0.05 merely suggests that the data at hand don't provide sufficient evidence to reject the null. Bayesian statistics could potentially be used to make the sort of claims made by the authors. The same issue applies on p. 6: "amisulpride did not change wanting ratings relative to placebo, β = -1.39, t(116) = 0.22, p = 0.83".

We apologize for these unclear formulations. We re-formulated these sentences and clarify that non-significant results represent only absence of evidence for meaningful differences rather than evidence for the null hypotheses. We agree that in principle Bayesian statistics could be used to corroborate claims in favor of the null hypothesis, but as we did not mean to make such claims we opted to formulate more carefully.

p.6:

“Separate analyses for wanting and liking ratings revealed no significant difference in wanting ratings between won and lost items, β = 0.29, t(115) = 0.64, p = 0.52”.

p.6:

“whereas amisulpride (mean = 5.2, sd = 1.7) showed no significant effects on wanting ratings relative to placebo, β = -1.39, t(116) = 0.22, p = 0.83.”

11) Ideally one would like to establish a causal mechanism from the present results: is reduced wanting following opioid receptor blockade mediated by a direct action of naltrexone on frontostriatal connectivity (as suggested in the title of this submission)? Have the authors have considered performing a mediation-style analysis?

We thank the reviewers for this interesting suggestion. We conducted a mediation analysis by adding the individual wanting-related DLPFC-striatum connectivity strength during wanting judgements as additional predictor to the behavioral model assessing the impact of naltrexone on wanting. When we assessed the significance of the indirect path (i.e., whether naltrexone effects on wanting can statistically be explained via modulation of DLPFC-striatum connectivity) using the mediation package in R, the indirect path showed only a trend-level effect, p = 0.06. Thus, the data provide no clear evidence that reduced wanting after opioid receptor blockade can causally be explained by naltrexone effects on frontostriatal connectivity. We therefore weakened our statements throughout the manuscript (including the title; see below) to avoid the impression that the naltrexone effect on behavior can causally be explained by changes in frontostriatal connectivity.

We report the mediation analysis on p.12:

“The observed drug effects on the neural level raise the question whether the impact of naltrexone on behavioral wanting ratings can statistically be explained by its influence on DLPFC-striatum connectivity. […] We must therefore be careful with interpreting the naltrexone effects on brain connectivity as the potential cause for the behavioral drug effects.”

We added a discussion of the non-significant mediation analysis on p.15:

“because we observed no significant brain-behavior mediation effect (p = 0.06), the stronger DLPFC-striatum connectivity under naltrexone should not be interpreted as the cause for changes in behavioral ratings under naltrexone.”

12) On p9 it would informative to present the striatal activation observed for the correlation with relevant wanting ratings. The statistics are reported for only one unilateral peak voxel, so it is hard to assess how robust this result is. This is particularly important since this result is supposed to be a replication of the authors' previous work (Weber et al., 2018).

In the revised manuscript, we present the wanting-related striatum activation during wanting judgements in Figure 2C and we also report more details for the wanting-related striatum activation. We clarify that the activation does not survive FWE-correction at the voxel-wise whole-brain level (only FWE-correction at the cluster level), suggesting that wanting-related striatum activation might be weaker in the current study compared with our previous work. However, given the clear a priori hypothesis based on our previous findings, it is justified to use SVC to test for wanting-related striatum activity. We report the stats for the whole-brain correction on p.11:

“We note that wanting-related striatal activation did not survive FWE-correction at the whole-brain level (p = 0.14, although it did survive whole-brain FWE-correction at the cluster level, p < 0.001), such that this effect appears to be somewhat weaker than in our previous study (Weber et al., 2018).”

13) The authors note that DLPFC-striatum connectivity was associated with self-reported wanting and that naltrexone increased this connectivity further. Since naltrexone also decreased self-reported wanting, are these results contradictory? This needs clarification.

Thank you for this interesting comment. One possibility for reconciling these two findings, in line with previous research (van den Bos et al., 2014), is that the DLPFC exerts top-down (inhibitory) influence over the striatum to reduce impulsivity. Assuming that impulsivity is associated with higher levels of wanting, such an influence may be important only for highly wanted items, whereas there is no need for an inhibition of striatal wanting signals for less desired items. From this perspective, the enhanced wanting-related DLPFC-striatum connectivity under naltrexone can be interpreted as stronger prefrontal top-down control over the striatum particularly for highly wanted items (i.e., when top-down impulse control is actually required). Because the study followed a between-group design, we cannot determine the strength of the pharmacological intervention on the individual level and it is thus not possible to test whether the naltrexone effects on behavior and brain activity are correlated. Moreover, as described above, also the mediation analysis showed no significant effect (p=0.06). In the added discussion of top-down control on p.15, we are thus cautious with interpreting the stronger functional connectivity under naltrexone as the cause for the naltrexone effects on behavioral wanting ratings:

“According to this view, the positive relationship between wanting ratings and DLPFC-striatum connectivity might indicate that DLPFC exerts top-down control over striatal representations of wanting predominantly for highly wanted items, whereas there may be less need for inhibitory top-down control for less desired goods (van den Bos et al., 2014). In any case, because we observed no significant brain-behavior mediation effect, the stronger DLPFC-striatum connectivity under naltrexone should not be interpreted as the cause for changes in behavioral ratings under naltrexone.”

Discussion

14) The authors used an intermediate dose of amisulpride (400 mg), which has been suggested to exert a mixture of pre- and post-synaptic effects depending on the participant, with pre-synaptic blockage likely to increase D1 receptor activation. Therefore any group-wise effect may cancel out across participants (see e.g. van der Schaaf et al., 2012; Sescousse et al., 2016; Eisenegger et al., 2014). Could this be a reason for the limited effects of amisulpride observed in the present study?

We agree with the reviewers that amisulpride can have both presynaptic and postsynaptic effects, and that these effects might cancel out on the group level, which might contribute to the weak to absent effects of amisulpride in the current study. In this case the effects of amisulpride should vary as a function of body weight as proxy of effective dose. However, adding body weight as predictor to the behavioral MGLMs did not change the overall result pattern, the impact of naltrexone remained significant, while we observed no significant effects of amisulpride. We report this analysis on p.7:

“Amisulpride can show both pre-synaptic and post-synaptic effects depending on the administered dose. […] There was thus no evidence for dose-dependent effects of amisulpride on wanting or liking ratings.”

Moreover, we note that in other studies using the same dose of amisulpride and a similar delay between drug intake and task performance we did observe significant amisulpride effects on decision-making (Burke et al., 2018; Soutschek et al., 2017). This further speaks against the possibility that the weak effects of amisulpride in the rating task can be explained by the administered dose. We elaborated the discussion of the non-significant amisulpride effects on p.17-18:

“Lastly, while high doses of amisulpride (≥400 mg) reduce postsynaptic dopaminergic signaling, lower doses of amisulpride increase dopaminergic activity via presynaptic mechanisms (Schoemaker et al., 1997). […] However, contrary to this view, we observed no significant amisulpride effects even when controlling for body weight as proxy for effective dose, and we note that in previous studies we had observed effects of 400 mg amisulpirde on behavior (Burke et al., 2018; Soutschek et al., 2017) and multivariate neural data (Kahnt et al., 2015). It seems thus unlikely that the null effects of amisulpride on the rating task can be explained solely by the chosen dosage.”

15) Some of the key results appear to be unilateral, i.e. specifically in the left hemisphere (striatal activity scaling with wanting ratings and DLPFC-striatum connectivity). Such a unilateral effect could reflect limited statistical power and fragile results, but also a truly lateralized effect. It would be worth commenting on this in the Discussion.

Good point. In the literature, reward-related functional connectivity with the striatum was reported for both the left and the right DLPFC. In agreement with this notion, a direct comparison between left and right hemisphere revealed no significant difference. We therefore avoid any claims regarding lateralization in the current manuscript (p.15).

“It is worth noting that in the current study we observed significant effects predominantly in the left hemisphere. […] We therefore do not make any claims regarding whether this result pattern represents just a power issue or a truly lateralized effect.”

16) Based on previous literature it is surprising that the striatum was not part of the neural correlates of wanting and liking (Figure 2) – striatal activation is only identified when the ratings are restricted to the relevant ones. Yet, based on studies like Barta et al. (2013) or Lebreton et al. (2009) that show that the striatum is at the core of a robust and automatic valuation system, one might expect activity in this region to correlated with ratings even when these are not explicitly required. Could the authors comment on this?

We agree with the reviewers that there is evidence that the striatum is part of an automatic valuation system. While GLM-1 (modelling wanting and liking ratings irrespective of their relevance) did not show significant wanting-related striatum activation, it is important to keep in mind that the parametric modulators in GLM-1 were not orthogonalized, explaining thus only the unique contributions of wanting and liking to differences in neural activity. In fact, when we computed a further GLM where liking was orthogonalized with respect to wanting, then wanting ratings significantly correlated with activation in the striatum, with the peak in the left striatum (z = 7.09, whole-brain FWE-corrected, p < 0.001, peak = [-9 11 -1]) but the cluster extended to both hemispheres. Thus, our findings are consistent with the well-described role of the striatum for reward processing, though this had not been evident in the GLM with non-orthogonalized parametric modulators due to the shared variance between wanting and liking.

We now report the results for the GLM with orthogonalized parametric modulators on p.8-9:

“GLM-1 revealed no significant wanting- or liking-related striatal activation, which may appear surprising given the canonical role of the striatum for reward processing (Bartra, McGuire, and Kable, 2013). […] In fact, when we orthogonalized liking with respect to wanting (such that the regressor for wanting explained the variance shared by wanting and liking), we observed bilateral wanting-related activation in the striatum (z = 7.09, whole-brain FWE-corrected, p < 0.001, peak = [-9 11 -1]).”

References:

Berridge, K. C., and Kringelbach, M. L. (2015). Pleasure systems in the brain. Neuron, 86(3), 646-664. doi:10.1016/j.neuron.2015.02.018

Berridge, K. C., and Valenstein, E. S. (1991). What psychological process mediates feeding evoked by electrical stimulation of the lateral hypothalamus? Behav Neurosci, 105(1), 3-14. doi:10.1037//0735-7044.105.1.3

Buchel, C., Miedl, S., and Sprenger, C. (2018). Hedonic processing in humans is mediated by an opioidergic mechanism in a mesocorticolimbic system. eLife, 7. doi:10.7554/eLife.39648

Cawley, E. I., Park, S., aan het Rot, M., Sancton, K., Benkelfat, C., Young, S. N.,... Leyton, M. (2013). Dopamine and light: dissecting effects on mood and motivational states in women with subsyndromal seasonal affective disorder. J Psychiatry Neurosci, 38(6), 388-397. doi:10.1503/jpn.120181

Chelnokova, O., Laeng, B., Eikemo, M., Riegels, J., Loseth, G., Maurud, H.,... Leknes, S. (2014). Rewards of beauty: the opioid system mediates social motivation in humans. Mol Psychiatry, 19(7), 746-747. doi:10.1038/mp.2014.1

Cools, R. (2011). Dopaminergic control of the striatum for high-level cognition. Curr Opin Neurobiol, 21(3), 402-407. doi:10.1016/j.conb.2011.04.002

Delay-Goyet, P., Zajac, J.-M., Javoy-Agid, F., Agid, Y., and Roques, B. (1987). Regional distribution of μ, δ and κ opioid receptors in human brains from controls and parkinsonian subjects. Brain Res, 414(1), 8-14.

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

Kahnt, T., Weber, S. C., Haker, H., Robbins, T. W., and Tobler, P. N. (2015). Dopamine D2-receptor blockade enhances decoding of prefrontal signals in humans. J Neurosci, 35(9), 4104-4111. doi:10.1523/JNEUROSCI.4182-14.2015

Korb, S., Gotzendorfer, S. J., Massaccesi, C., Sezen, P., Graf, I., Willeit, M.,... Silani, G. (2020). Dopaminergic and opioidergic regulation during anticipation and consumption of social and nonsocial rewards. eLife, 9. doi:10.7554/eLife.55797

Lidow, M. S., Goldman-Rakic, P. S., Gallager, D., and Rakic, P. (1991). Distribution of dopaminergic receptors in the primate cerebral cortex: quantitative autoradiographic analysis using [3H] raclopride,[3H] spiperone and [3H] SCH23390. Neuroscience, 40(3), 657-671.

Pool, E., Sennwald, V., Delplanque, S., Brosch, T., and Sander, D. (2016). Measuring wanting and liking from animals to humans: A systematic review. Neurosci Biobehav Rev, 63, 124-142. doi:10.1016/j.neubiorev.2016.01.006

Venugopalan, V. V., Casey, K. F., O'Hara, C., O'Loughlin, J., Benkelfat, C., Fellows, L. K., and Leyton, M. (2011). Acute phenylalanine/tyrosine depletion reduces motivation to smoke cigarettes across stages of addiction. Neuropsychopharmacology, 36(12), 2469-2476. doi:10.1038/npp.2011.135

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

Weber, S. C., Kahnt, T., Quednow, B. B., and Tobler, P. N. (2018). Frontostriatal pathways gate processing of behaviorally relevant reward dimensions. PLoS Biol, 16(10), e2005722. doi:10.1371/journal.pbio.2005722

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

The reviewers and editors feel that the manuscript has been improved but there are several remaining major issues that need to be addressed, as outlined below. Please note that if these issues are not satisfactorily addressed in your revised submission then unfortunately we will not be able to consider the manuscript further, as it is not editorial practice to issue multiple revise resubmit decisions at eLife.

1) The most important issue is that there remains a discrepancy between similar effect sizes of naltrexone/amisulpride on wanting (albeit in opposite directions) and the corresponding pattern of P-values obtained from the hierarchical analysis. A similar issue is also present in relation to liking ratings. The authors need to explore this discrepancy in considerably more detail and resolve it, as follows:

a) Conduct a non-hierarchical analysis using the mean ratings for wanting and liking (in two separate models, one for wanting, one for liking). The reason for this is that it appears from the data depicted in Figure 1D that amisulpride may increase wanting and also liking (where the effect may actually be even greater). The reviewers noted that for liking ratings the mean difference is ~0.5 points and the SDs are actualy lower than for wanting ratings at 1.0/1.2, which is suggestive of a larger effect than the effect of naltrexone on wanting which is significant in the hierarchical model. It would also be useful to provide the standardised effect sizes (Cohen's d) for the 4 comparisons against placebo (2 for wanting, 2 for liking).

Following the reviewers’ advice, we conducted a non-hierarchical analysis of pharmacological effects on wanting and liking ratings. For this, we computed the mean wanting and liking ratings across all items, separately for each participant and session (pre-test versus post-test). The analysis of wanting ratings replicated the significant main effect of naltrexone versus placebo (p = 0.03), while amisulpride showed no significant effect on mean wanting ratings (p = 0.82). Mean liking ratings were neither affected by naltrexone (p = 0.98) nor amisulpride (p = 0.23). Thus, also a non-hierarchical analysis of aggregated mean data provides no evidence for significant drug effects. We currently did not include it in the manuscript as it just replicates the results from the hierarchical analysis; however, we would be happy to add it to the manuscript if the reviewers prefer us to do so. As suggested by the reviewer, we now report Cohen’s d for the drug effects on wanting and liking on p.6-7:

“Judgement type-specific analyses suggested that wanting ratings were significantly reduced under naltrexone (mean = 4.5, standard deviation (sd) = 1.0) relative to placebo (mean = 4.9, sd = 1.0), β = -13.85, t(115) = 2.12, p = 0.04, Cohen’s d = 0.47, whereas amisulpride (mean = 4.9, sd = 1.0) showed no significant effects on wanting ratings relative to placebo, β = -1.39, t(116) = 0.22, p = 0.83, Cohen’s d = 0.05. […] Taken together, our findings provide evidence for involvement of opioidergic neurotransmission in wanting judgements.”

Still, this leaves open the question how the perceived discrepancy between the figure and the statistics can be explained. We carefully checked our scripts for creating the plots and, embarrassingly, discovered an error in the calculation of the mean ratings used for the figures (wrong assignment of individual values to drug groups in the plotting script). We now corrected the figures as well as the means/standard deviations reported in the manuscript, and in our view the discrepancy between figures and statistics has now resolved. We apologize for this confusion and are grateful to the reviewers whose careful inspection of the figures allowed us to correct this mistake.

b) Assuming that the above analyses using the mean ratings provide a discrepant pattern of significance to the hierarchical analysis, this then needs to be investigated thoroughly and explained in the manuscript, both for wanting and for liking ratings. The authors need to dig into the data carefully and figure out why this discrepancy arises. For example, if amisulpride makes participants more variable in their ratings (or naltrexone make them more consistent), this would be important for interpretation. Or perhaps some assumptions of the hierarchical model have been broken? Or perhaps the covariance structure requires amending? Or perhaps the model did not converge? Without resolution of this discrepancy it will not be possible to consider the manuscript further.

Given that the non-hierarchical analysis shows converging results with the hierarchical analysis and we corrected the error in the script for the plots, there is no need for further digging into the data as far as we can see.

2) The authors now report the magnitude of the correlation between wanting and liking ratings, which is unsurprisingly high (r = 0.71). Since they have not serially orthogonalized the parametric regressors in the main analysis, this means that much of the variance of these ratings is simply removed. For this reason it is not clear how much we can infer from the non-significant drug effects, considering that these were assessed using only a fraction of the ratings variance, which may result in an insensitive analysis. Therefore further analyses are required here to substantiate the conclusion that there were no drug effects (as reported on the top of p10 – it is assumed that currently this refers to the model without serial orthogonalisation, although this should be stated explicitly for clarity).

The authors do provide some results from an analysis using serial orthogonalisation, with liking orthogonalised against wanting (p9), yielding the expected striatal activation for the parametric effect of wanting (which then carries the shared variance), which is reassuring – as they note this suggests that the striatal signal is substantively affected by the colinearity between wanting and liking ratings. Please additionally report the drug effects in this analysis. The authors should also report the effects from an analysis in which the serial orthogonalisation is performed in the alternate order (i.e. wanting against liking, such that the liking regressor now carries the shared variance), including both the main parametric effect (this time of liking) and drug effects.

We thank the reviewers for this suggestion. As recommended, we now report the results for two further GLMs, one where liking was orthogonalized against wanting, and one where wanting was orthogonalized against liking. These GLMs suggest that both wanting and liking ratings correlate with activity in the neural reward system (striatum, VMPFC, and PCC). In addition, we also tested for drug effects on the neural correlates of wanting or liking, but no effect survived correction for multiple comparisons.

In the revised manuscript, we report these analyses on p.9-10:

“GLM-1 revealed no significant wanting- or liking-related striatal activation, which may appear surprising given the canonical role of the striatum for reward processing (Bartra, McGuire, and Kable, 2013). […] However, also in the GLMs with orthogonalized parametric modulators, we observed no effects of naltrexone or amisulpride (relative to placebo) on wanting (GLM-3) or liking (GLM-4) ratings even at lenient statistical thresholds (p < 0.001 uncorrected, cluster size > 20 voxels).”

We describe these models in the Materials and methods section on p.25-26:

“Finally, we computed two further models, one where the liking regressor in GLM-1 was orthogonalized with respect to wanting (such that the regressor for wanting contained the variance shared by wanting and liking; GLM-3) and one where wanting was orthogonalized with respect to liking (GLM-4).”

3) It is helpful that the authors added the information that previous data from the same study were published in a 2016 paper by Weber et al. Oddly they do not mention the results of that paper, even in the discussion of the (apparent – see point 3 below) non-significant effects of amisulpride. The 2016 findings are highly relevant, since the amisulpride group was found to suppress cue-based responding and reward impulsivity. Similar results, but weaker, were reported for naltrexone. Both groups also reported lower mood than the placebo group.

The authors explain that the PIT and delay discounting tasks were completed after the end of scanning, i.e. after 60 minutes absorption time + 90 minutes fmri rating task = minimum 2.5 hours after drug administration. Hence, it seems highly relevant for the interpretation of the present data that, in the exact same participants, the same dose of amisulpride reported to show a null during 1-2.5 hours, showed what are (presumably) expected effects after 2.5 hours. Therefore it is necessary to mention this prior publication from the same study in the Introduction, and discuss the results, especially with respect to dose timing, in the Discussion.

We agree with the reviewers that the results of the already published data inform the interpretation of the current findings. As recommended, we now mention that the study was part of a larger project already in the introduction section (p.4):

“This study was part of a larger project investigating also the roles of opioidergic and dopaminergic activity for reward impulsivity (Weber et al., 2016).”

When discussing the non-significant amisulpride effects in the Discussion section, we now clarify that in the same participants amisulpride showed significant effects in other tasks. We discuss that this discrepancy might be explained either by different sensitivities of the tasks to dopaminergic manipulations or by the time course of amisulpride effects (p.17-18).

“We note though that in the same sample of participants amisulpride showed significant effects on tasks for cue reactivity and delay discounting (Weber et al., 2016), which were administered 2.5 hours after drug intake (while the rating task started 1 hour after drug intake). Due to this difference in timing, it is not possible to decide whether the different amisulpride effects on these tasks can be explained by different sensitivities of these tasks to dopaminergic manipulations or by the time course of amisulpride effects.”

https://doi.org/10.7554/eLife.71077.sa2

Article and author information

Author details

  1. Alexander Soutschek

    Department of Psychology, Ludwig Maximilian University, Munich, Germany
    Contribution
    Formal analysis, Writing - original draft
    For correspondence
    alexander.soutschek@psy.lmu.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8438-7721
  2. Susanna C Weber

    Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Zürich, Switzerland
    Contribution
    Conceptualization, Formal analysis, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Thorsten Kahnt

    Department of Neurology, Northwestern University Feinberg School of Medicine, Chicago, United States
    Contribution
    Conceptualization, Writing - review and editing
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3575-2670
  4. Boris B Quednow

    1. Experimental and Clinical Pharmacopsychology, Department of Psychiatry, Psychotherapy and Psychosomatics, Psychiatric Hospital, University of Zurich, Zurich, Switzerland
    2. Neuroscience Center Zurich, University of Zurich and Swiss Federal Institute of Technology Zurich, Zürich, Switzerland
    Contribution
    Conceptualization, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Philippe N Tobler

    1. Zurich Center for Neuroeconomics, Department of Economics, University of Zurich, Zürich, Switzerland
    2. Neuroscience Center Zurich, University of Zurich and Swiss Federal Institute of Technology Zurich, Zürich, Switzerland
    Contribution
    Conceptualization, Funding acquisition, Supervision, Writing - original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4915-9448

Funding

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (Grants 10001C_188878 and 100014_165884)

  • Philippe N Tobler

Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung (100019_176016)

  • Philippe N Tobler

Velux Stiftung (981)

  • Philippe N Tobler

Deutsche Forschungsgemeinschaft (SO 1636/2-1)

  • Alexander Soutschek

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

Acknowledgements

We thank Karl Treiber for expert support with data collection as well as Beatrice Beck Schimmer for medical support.Funding and financial disclosure: PNT received funding from the Swiss National Science Foundation (Grants 10001C_188878, 100019_176016, and 100014_165884) and from the Velux Foundation (Grant 981). AS received an Emmy Noether fellowship (SO 1636/2-1) from the German Research Foundation.

Ethics

Clinical trial registration https://www.clinicaltrials.gov/ (NCT02557984).

Human subjects: All participants provided written informed consent. The study was approved by the ethics committee of the canton of Zurich (KEK-ZH-NR2012-0347).

Senior Editor

  1. Michael J Frank, Brown University, United States

Reviewing Editor

  1. Jonathan Roiser, University College London, United Kingdom

Reviewer

  1. Guillaume Sescousse, Inserm, France

Publication history

  1. Received: June 8, 2021
  2. Preprint posted: June 21, 2021 (view preprint)
  3. Accepted: November 10, 2021
  4. Accepted Manuscript published: November 11, 2021 (version 1)
  5. Version of Record published: November 17, 2021 (version 2)

Copyright

© 2021, Soutschek 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

  • 252
    Page views
  • 47
    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. Neuroscience
    Rawan AlSubaie et al.
    Research Article Updated

    Projections from the basal amygdala (BA) to the ventral hippocampus (vH) are proposed to provide information about the rewarding or threatening nature of learned associations to support appropriate goal-directed and anxiety-like behaviour. Such behaviour occurs via the differential activity of multiple, parallel populations of pyramidal neurons in vH that project to distinct downstream targets, but the nature of BA input and how it connects with these populations is unclear. Using channelrhodopsin-2-assisted circuit mapping in mice, we show that BA input to vH consists of both excitatory and inhibitory projections. Excitatory input specifically targets BA- and nucleus accumbens-projecting vH neurons and avoids prefrontal cortex-projecting vH neurons, while inhibitory input preferentially targets BA-projecting neurons. Through this specific connectivity, BA inhibitory projections gate place-value associations by controlling the activity of nucleus accumbens-projecting vH neurons. Our results define a parallel excitatory and inhibitory projection from BA to vH that can support goal-directed behaviour.

    1. Cell Biology
    2. Neuroscience
    Angela Kim et al.
    Research Article Updated

    Insulin-induced hypoglycemia is a major treatment barrier in type-1 diabetes (T1D). Accordingly, it is important that we understand the mechanisms regulating the circulating levels of glucagon. Varying glucose over the range of concentrations that occur physiologically between the fed and fuel-deprived states (8 to 4 mM) has no significant effect on glucagon secretion in the perfused mouse pancreas or in isolated mouse islets (in vitro), and yet associates with dramatic increases in plasma glucagon. The identity of the systemic factor(s) that elevates circulating glucagon remains unknown. Here, we show that arginine-vasopressin (AVP), secreted from the posterior pituitary, stimulates glucagon secretion. Alpha-cells express high levels of the vasopressin 1b receptor (V1bR) gene (Avpr1b). Activation of AVP neurons in vivo increased circulating copeptin (the C-terminal segment of the AVP precursor peptide) and increased blood glucose; effects blocked by pharmacological antagonism of either the glucagon receptor or V1bR. AVP also mediates the stimulatory effects of hypoglycemia produced by exogenous insulin and 2-deoxy-D-glucose on glucagon secretion. We show that the A1/C1 neurons of the medulla oblongata drive AVP neuron activation in response to insulin-induced hypoglycemia. AVP injection increased cytoplasmic Ca2+ in alpha-cells (implanted into the anterior chamber of the eye) and glucagon release. Hypoglycemia also increases circulating levels of AVP/copeptin in humans and this hormone stimulates glucagon secretion from human islets. In patients with T1D, hypoglycemia failed to increase both copeptin and glucagon. These findings suggest that AVP is a physiological systemic regulator of glucagon secretion and that this mechanism becomes impaired in T1D.