A randomized controlled experimental medicine study of ghrelin in value-based decision making

BACKGROUND The stomach-derived hormone ghrelin stimulates appetite, but the ghrelin receptor is also expressed in brain circuits involved in motivation and reward. We examined ghrelin effects on decision making beyond food or drug reward using monetary rewards. METHODS Thirty participants (50% women and 50% men) underwent 2 fMRI scans while receiving i.v. ghrelin or saline in a randomized counterbalanced order. RESULTS Striatal representations of reward anticipation were unaffected by ghrelin, while activity during anticipation of losses was attenuated. Temporal discounting rates of monetary reward were lower overall in the ghrelin condition, an effect driven by women. Discounting rates were inversely correlated with neural activity in a large cluster within the left parietal lobule that included the angular gyrus. Activity in an overlapping cluster was related to behavioral choices and was suppressed by ghrelin. CONCLUSION This is, to our knowledge, the first human study to extend the understanding of ghrelin’s significance beyond the canonical feeding domain or in relation to addictive substances. Contrary to our hypothesis, we found that ghrelin did not affect sensitivity to monetary reward anticipation, but rather resulted in attenuated loss aversion and lower discounting rates for these rewards. Ghrelin may cause a motivational shift toward caloric reward rather than globally promoting the value of reward. TRIAL REGISTRATION EudraCT 2018-004829-82. FUNDING Swedish Research Council (2013-07434), Marcus and Marianne Wallenberg foundation (2014.0187) and National Institute on Drug Abuse/National Institute on Alcohol Abuse and Alcoholism Intramural Research Program.


Reaction Time and Number of hits
Reaction times did not differ between ghrelin and placebo sessions, neither overall (grand mean: 223.7±2.2 vs. 226.1±2.3, mean±SEM; paired t-test: p=0.26), nor within any individual trial category [RM-ANOVA:s with Intervention (ghrelin vs. placebo) as within subject factor, and session order and sex as covariates: reward trials: p=0.12; loss trials: p=0.26; neutral trials: p=0.19].
Similarly, the total number of hits did not differ between ghrelin and placebo sessions, neither overall (grand mean: 43.6±0.64 vs. 43.0±0.67; p=0.47), nor within any individual trial category [RM-ANOVA:s with Intervention (ghrelin vs. placebo) as within subject factor, and session order and sex as covariates: reward trials: p=0.48; loss trials: p=0.12; neutral trials: p=0.57].

Sensitivity analyses -Intervention x Outcome anticipation whole-brain analysis
Compared to the MVM model controlling for sex reported in the main text (Figure 2), additional controls for BMI and age (also controlling for sex) only marginally reduced the main effect of Anticipation on striatal activity from 1734 voxels to 1689 voxels, using the same Intervention (placebo/ghrelin) x Anticipation (high loss/low loss/neutral/low reward/high reward) approach, and no main effects of either age nor BMI (or sex) were identified.

Reward anticipation
Whole-brain MVM analysis of the anticipation phase showed the expected main effect of reward anticipation [F(1,40)=45.91, p<0.001, ηp 2 =0.63], with a large cluster in the ventral striatum (1188 voxels). Ventral striatal activation was proportional to reward magnitude, with significantly higher activation for maximum reward anticipation vs lower reward and neutral anticipation (all ps<0.001; Bonferroni corrected). Low reward also resulted in increased activity vs neutral anticipation (mean difference ± SEM: 0.039±0.007, p<0.001), illustrating that striatal activity scaled with reward magnitude ( Figure S4). Activity in this cluster was unaffected by ghrelin (main effect: p=0.49; interaction: p=0.77). Sex did not have any significant effects on the results.
The specific analysis of activations associated with losses rather than rewards is provided in the next section. However, we noted that, when analyzing extracted betas within the cluster identified by the analysis of reward anticipation, there was also a robust main effect of loss anticipation [F(2,54)=19.10, p<0.001, ηp 2 =0.41]. This effect showed a similar parametric increase for each amount, with the highest activation for high loss compared to lower loss (mean difference ± SEM: 0.019±0.007, p=0.026) and neutral (mean difference ± SEM: 0.048±0.009, p<0.001). Lower loss anticipation was also significantly different from neutral (mean difference ± SEM: 0.029±0.008, p=0.003). We also found a main effect of Intervention on loss anticipation in this cluster [F(1,27)=5.56, p=0.026, ηp 2 =0.17], with stronger deactivation during ghrelin compared to placebo (mean difference ± SEM: 0.019±0.008, p=0.026). As shown in Figure S4, there was a trend towards an Intervention x Anticipation interaction (p=0.091), illustrating the parametric nature of the relationship. Sex did not have a significant effect.
Further inspection of the whole-brain analysis also revealed a small main effect of Intervention, with higher activation in the right calcarine gyrus during ghrelin compared to placebo (see Figure S10).

Loss Anticipation
When activations associated with loss anticipation were specifically analyzed, we found striatal activations similar to those observed during reward anticipation (see Figure S5), with the largest striatal cluster in the right putamen (262 voxels). A main effect of Anticipation [F(2,54)=32.84, p<0.001, ηp 2 =0.55] revealed a similar parametric increase in activation for losses, with significantly higher activation for maximum loss compared to low loss (mean difference ± SEM: 0.030±0.008, p<0.001) and neutral anticipation (mean difference ± SEM: 0.057±0.008, p<0.001). Low loss activation was significantly higher compared to neutral (mean difference ± SEM: 0.028±0.007, p=0.001). There was no main nor interaction effect of Intervention (p=0.147 and p=0.458, respectively). Sex did not have any significant effects.
Similar to the ventral overlap between loss and reward, there was an increase in reward activity in this dorsal cluster, with a strong main effect of Anticipation [F(1,38)=30.74, p<0.001, ηp 2 =0.53], showing a significantly higher activation for high reward compared to low reward (mean difference ± SEM: 0.016±0.005, p=0.02) and neutral (mean difference ± SEM: 0.056±0.009, p<0.001), while low reward also was significantly different compared to neutral (mean difference ± SEM: 0.040±0.007, p<0.001). There was no main nor interaction effect of Intervention (p=0.835 and p=0.268, respectively). There was no effect of Intervention. Sex did not influence the results.
Only two small additional clusters influenced by Intervention were identified across the whole brain, with stronger deactivation of the left postcentral gyrus (13 voxels) and increased activation in the right calcarine gyrus (12 voxels) during ghrelin compared to placebo ( Figure S11).

Feedback phase
During feedback, the MVM results showed a significant Intervention x Feedback interaction, F(1, 26)=9.63, p=0.005, ηp 2 =0.27, revealing a (whole-brain corrected) cluster of bilateral activation in the right posterior MCC [MNI:2,14,41]. Compared to placebo, ghrelin elicited increased posterior MCC activation in response to successfully obtained rewards relative to neutral feedback (mean difference ± SEM: 0.123±0.048, p=0.016; Bonferroni corrected). MCC activation in response to reward was also coupled to a stronger deactivation in response to neutral feedback during the ghrelin infusion.

Motion Analyses
A paired t-test showed a numerically small but statistically significant difference in average motion per TR between the placebo and ghrelin session (mean difference: 0.007 mm, p=0.034; placebo > ghrelin). Closer inspection showed that this was mainly driven by females (p=0.044; placebo > ghrelin), while average motion per TR was not significantly different between sessions in males (p=0.15). Max motion displacement and censor fraction per stimulus did not differ between the two sessions overall (p=0.22 and p=0.077, respectively). At trend-level, censor fraction per stimulus was slightly higher during placebo, mainly driven by two subjects with 27% and 35% of stimuli censored.
These subjects were removed from all statistical analyses.

Delay Discounting (DD) Task
Average response time did not differ between ghrelin and placebo sessions [1.87s±0.06 vs.

Adverse events
No treatment-emergent adverse events were recorded. The adverse events recorded were all mild, transient, related to the inconvenience of MR scanning, and evenly distributed between ghrelin and placebo sessions. Please see Table S7 for all reported side effects.

Participant characteristics
Neither age nor BMI differed between men and women, as confirmed by independent samples t-test (p=0.64 and p=0.66, respectively; see Table S8 for mean and SEM values).          Table S3. Activation locations for Anticipation High reward > Neutral in the MID task. All regions surpassed the threshold for whole-brain significance, determined as described in Methods.

Location
Cluster -21, -28, 60 Note. * = deactivation (high loss < neutral). Table S4. Activation locations for Anticipation High loss > Neutral in the MID task. All regions surpassed the threshold for whole-brain significance, determined as described in Methods.
Note. * = deactivation (reward < neutral). Table S5. Activation locations for Feedback Reward > Neutral in the MID task. All regions surpassed the threshold for whole-brain significance, determined as described in Methods.

Location
Cluster  Table S6. Activation locations for main effects of Choice and Intervention in the Delay Discounting task. All regions surpassed the threshold for whole-brain significance, determined as described in Methods. There was no brain-wide significant region for Choice x Intervention interaction. MNI: Montreal Neurological Institute.

Ghrelin session Saline session
Muscle twitching during MRI. Pain in the lower back at the end of the MRI. Common cold symptoms one day after the session.
Common cold symptoms one week after the session.
Experienced warmth in the beginning of the infusion. Nothing unpleasant.
Sprained toe during training 2 days after the session.
Numbness of the left buttock at the end of MRI.
Left hand went numb during the MRI for about 30 seconds.
Short muscle cramp in the left thigh at the end of MRI.
Muscle twitching during MRI.
Common cold symptoms one day after the session.
Two days after the session noticed redness in left index fingernail.

Dizziness.
Feelings of discomfort during MRI. Cold sweats.
Headache. Bright light in the field of vision before MRI, disappeared after ca 30 min.
Increased blood pressure after fMRI. Frightened by the muscle twitches.
Common cold symptoms six days after the session.
Drop in blood pressure during blood draw.
Pressure on the nose from fMRI glasses .
Headache at the end of the session.
Pain from needle insertion site four days after the session.
10 days after the session right hand injury at work.
Nausea a few hours after the session.