Endogenous opioid receptor system mediates costly altruism in the human brain

The study examines whether individual differences in the endogenous μ-opioid receptor (MOR) system contribute to variability in vicarious pain processing and altruistic, financially costly helping. Prosocial behavior is widespread in humans and many animals, and empathy is hypothesized to drive prosociality by transforming others’ pain into self-referential motivational signals to alleviate it. Prior neuroimaging shows that witnessing others’ pain recruits a core network including anterior insula, mid/anterior cingulate cortex, amygdala, and somatosensory cortices, with broader engagement of motor, frontal, parietal, temporal, thalamic, and striatal regions depending on context. The opioid system, especially MORs, is central to pain and affect regulation and social motivation. Placebo analgesia and pharmacological studies implicate opioidergic mechanisms in both first-hand and vicarious pain and in prosocial tendencies. The authors therefore tested the hypotheses that: (1) participants would engage in costly helping by donating to reduce a confederate’s pain; (2) MOR availability would be negatively correlated with BOLD responses to witnessing others’ pain; and (3) higher MOR availability would be associated with greater altruistic tendencies at neural (and potentially behavioral) levels.

The literature identifies a vicarious pain network (anterior insula, ACC, amygdala, SI/SII) engaged when observing others’ pain, with extensions to motor, prefrontal, temporal, parietal, thalamic, and striatal regions. Studies that allow participants to help (e.g., trading money or own pain to reduce others’ pain) link helping with activity in insula, ACC, amygdala, dLPFC, OFC, SI, and vmPFC, and with functional connectivity patterns between insula and ACC. Financially costly helping paradigms have proven effective in quantifying brain–behavior relations in altruism. On the molecular level, μ-opioid receptors mediate effects of endogenous opioids and have widespread expression in emotion circuits. MORs modulate nociception, positive/negative affect, and social motivation in humans and nonhuman primates. Placebo analgesia (opioidergic) reduces first-hand and vicarious pain and can reduce helping; chronic opioid use relates to reduced perceived pain in others. PET with [11C]carfentanil shows endogenous opioid release during pain, links lower MOR availability to higher pain sensitivity, and negative associations between MOR availability and haemodynamic responses to vicarious pain. MOR availability also correlates with prosocial motivation (attachment). These findings motivate testing MOR involvement in costly altruism in vivo.

Participants: Thirty healthy Finnish women (mean age 24.7 ± 5.65 years, range 19–42) with normal or corrected vision. Only females were included due to sexual dimorphism in MOR system and known sex differences in empathy/pain. Exclusions: CNS-affecting medications, psychiatric/neurological conditions, substance abuse, standard MRI/PET contraindications. Structural abnormalities were screened out by neuroradiologist. All 30 underwent fMRI; 14 also underwent PET on a separate day (PET before fMRI; mean interval 84 ± 62 days, range 4–190). Ethics approval 51/1801/2019; informed consent obtained.

fMRI costly helping task: Adapted from Gallo et al. Participants met a confederate (author KS) and believed they watched her live via CCTV receiving painful electric shocks. In reality, prerecorded videos were shown. Each trial: fixation (1–3 s); first video showing shock (2 s; expressed pain intensity 2–6/10); fixation (1.5–3 s); donation phase (self-paced) to reduce intensity of second shock in that trial using two buttons (increase/decrease donation); fixation (1–3 s); second video (2 s) whose displayed intensity equals first video intensity minus donated amount; inter-trial interval (7–10 s). Endowment per trial: 6€. Donation rule: every 1€ donated reduces second shock’s intensity by 1 point on 10-point scale; undonated total/10 paid as bonus. Two runs of 15 trials each. Debriefing at end.

MRI acquisition and preprocessing: 3T GE SIGNA Premier. fMRI: T2* EPI, 45 slices, slice thickness 3 mm, TR 2600 ms, TE 30 ms, flip angle 75°, FOV 24 cm, voxel 3×3×3 mm3. Structural T1: 1×1×1 mm3. Preprocessing with fMRIPrep 1.3.0.post2: T1w bias correction, skull-strip, surfaces, masks, tissue segmentation, normalization to ICBM 152 2009c via ANTs; functional preprocessing included reference generation and skull-strip, coregistration to T1w, slice-time correction, 6 mm FWHM smoothing, ICA-AROMA motion denoising, resampling to MNI152NLin2009cAsym, high-pass filtering and CompCor components estimation.

PET acquisition and analysis: GE Discovery MI PET/CT at rest. Radiotracer [11C]carfentanil (μ-opioid receptor agonist), injected dose 250.6 ± 10.9 MBq. Dynamic acquisition 51 min with frames: 3×60 s, 4×180 s, 6×360 s. In-plane resolution 3.77 mm FWHM, tangential 4.00 mm. Corrections: dead-time, decay, attenuation. Preprocessing with MAGIA pipeline. Binding potential (BPND) estimated voxelwise using the simplified reference tissue model (SRTM) with occipital cortex as reference; interpreted as receptor availability (density of receptors unoccupied by endogenous ligand). Parametric images normalized to MNI space via T1 segmentation/normalization and smoothed (8 mm FWHM). High test–retest stability noted for [11C]carfentanil.

Regions of interest (ROIs): 17 a priori ROIs implicated in vicarious pain/empathy: amygdala, caudate, cerebellum, dorsal ACC, inferior temporal gyrus, insula, middle temporal gyrus, nucleus accumbens, orbitofrontal cortex, pars opercularis, posterior cingulate cortex, putamen, rostral ACC, superior frontal gyrus, superior temporal gyrus, temporal pole, thalamus. Subject-specific FreeSurfer parcellations used to extract average BPND.

Statistical analysis: fMRI first-level GLMs modeled three events (1st video, donation phase, 2nd video) with boxcar regressors convolved with HRF. Donation size (trial-wise) entered as parametric modulator for the 1st video. Contrasts: main effects (1st video, donation, 2nd video), and 1st > 2nd video. Group-level random-effects analyses with cluster-level FWE correction (initial Punc < 0.001, cluster-extent threshold determined to achieve FWEc; final maps thresholded at Punc < 0.001 and k = FWEc). For MOR–BOLD interactions (N=14 PET): PCA on ROI BPND to reduce dimensionality (first three PCs explained ~61%, 22%, 7%). The first PC used to predict voxelwise BOLD responses to donation-modulated 1st video and to donation phase in second-level models. Additionally, affective vicarious pain signature (AVPS) was dot-multiplied with first-level beta maps (1st video with donation modulator) to obtain a scalar per participant and correlated with MOR PCs. Finally, whole-brain linear regressions used ROI-wise BPND to predict voxelwise BOLD responses (FWE cluster-level p < 0.05). Behavioral analysis: linear mixed model (LMM) tested donations as a function of first video shock intensity with subject as random effect; for PET subsample, per-subject slope and intercept (donation = slope × intensity + intercept) were related to MOR PCs via multiple regression and Bayesian correlations.

Behavioral: Participants donated in all intensity conditions (mean donation per trial M = 2.826€, SD = 1.964). Donations increased with shock intensity in the first video (LMM: β = 0.877, SE = 0.062, t = 14.049, p < 0.001; intercept = 0.196). Random effects: intercept variance 1.943 (SD 1.394); slope variance 0.104 (SD 0.323). In the PET subsample (N=14), individual slopes/intercepts relating donation to shock intensity did not associate with MOR PCs (multiple regression for slope: F(3,10) = 0.036, p = 0.990; intercept: F(3,10) = 0.552, p = 0.658). Bayesian correlations supported the null (Table shows BF10 generally < 1; e.g., slope vs PC1 r = 0.075, p = 0.800, BF10 = 0.339).

BOLD during observing pain (1st video): Robust activation in anterior insula, ACC, thalamus, with additional activations in amygdala and striatum, as well as sensorimotor (M1, S1) and other cortical regions, when witnessing the confederate’s pain. Parametric modulation by donation size during the 1st video showed stronger activation with increasing donations in insula and ACC (and related regions).

1st versus 2nd video: Neural responses decreased during the 2nd video compared to the 1st in vicarious pain and sensorimotor regions (insula, thalamus, ACC, striatum, M1, S1). This decrease persisted even when analyses were restricted to low-intensity trials (levels 2–3) to better match content across videos, suggesting reductions not solely due to stimulus intensity differences.

MOR–BOLD associations during observing pain: Individual differences in MOR availability (PC1) showed a generally negative correlation with the sensitivity of BOLD responses to donation size during the 1st video, notably in amygdala, striatum, insula, hippocampus, thalamus, ACC, and PCC. AVPS scores correlated negatively with MOR PC3 (r = -0.659, p = 0.010, BF10 = 6.5), indicating lower MOR availability relates to stronger affective vicarious pain-related responses. Cumulative maps replicated negative associations in vicarious pain areas (insula, ACC, thalamus), with additional somatosensory, temporal, limbic, and frontal involvement.

MOR–BOLD associations during donation phase: In contrast, during the donation (decision) phase, MOR availability positively correlated with BOLD responses, particularly in ACC and hippocampus (also involving striatum and lingual regions), suggesting higher MOR tone enhances neural engagement during helping decisions.

Overall: MOR availability did not predict how much participants donated behaviorally, but it systematically modulated neural responses during pain observation (negative association) and during helping decisions (positive association), indicating a neuromolecular role for the MOR system in altruistic processes.

The findings address whether the endogenous μ-opioid receptor system modulates empathy for pain and costly helping. Participants exhibited altruistic behavior by forfeiting money to reduce another’s pain, with donations tracking perceived pain intensity. Observing others’ pain engaged canonical vicarious pain and emotion circuits (anterior insula, ACC, thalamus, amygdala, striatum), and the degree of engagement predicted larger subsequent donations. Critically, baseline MOR availability was negatively related to BOLD sensitivity to donation size during pain observation, consistent with the role of the opioid system in buffering pain and distress: lower MOR availability corresponded to stronger neural responses to others’ suffering. Conversely, during the donation phase, MOR availability positively related to BOLD responses in ACC and hippocampus (and striatum), suggesting that higher MOR tone enhances neural processes during prosocial decision-making. The reduction in activation from the 1st to the 2nd video aligns with the negative-state relief account of altruism (helping reduces vicarious distress), although alternative explanations such as task relevance/attention and predictability differences between the two videos may also account for reduced responses. Together, results suggest that MOR tone shapes where and when in the decision process the brain is most engaged—buffering affective responses during passive observation while enhancing engagement during active helping. This neuromolecular mechanism extends prior work linking opioids to pain, affect, and social bonding, and provides in vivo evidence that MORs contribute to individual differences in altruistic processes.

This study shows that individual differences in μ-opioid receptor availability modulate neural processing of others’ distress and the neural implementation of costly helping. While participants generally helped more as perceived pain increased, MOR availability did not predict donation amounts. Instead, MOR availability was negatively related to vicarious pain-related BOLD responses during observation and positively related to BOLD responses during the helping decision (notably in ACC and hippocampus). These results implicate the MOR system as a key neuromolecular pathway in altruistic behavior. Future work should include larger and more diverse samples (including males), incorporate control conditions to dissociate observation from helping-specific processes, assess endogenous opioid release capacity or use pharmacological challenges, consider simultaneous PET-fMRI, and examine interactions with other neuromodulators (e.g., oxytocin).

A single baseline PET scan cannot disentangle specific molecular mechanisms (e.g., receptor density vs. endogenous ligand occupancy) or capture endogenous opioid release capacity. The sample included only females, limiting generalizability across sexes. PET subsample size was small (N=14), which reduces power for brain–behavior associations. Some participants expressed doubts about the cover story, potentially introducing confounds, though complete disbelief was not reported. The experiment lacked a control condition to separate effects of observing pain from those of costly helping. PET and fMRI were conducted on different days (though [11C]carfentanil shows strong test–retest reliability), and thus do not provide simultaneous receptor–BOLD coupling.