Kappa opioid receptor availability correlates with depressive symptom burden in chronic migraine: a preliminary investigation

Abstract

Background

Chronic pain is frequently accompanied by affective disturbances, yet the neurobiological mechanisms linking pain perception and affective vulnerability remain poorly understood. The kappa opioid receptor (KOR)/dynorphin system has emerged as a key modulator of nociceptive and affective processes, in part through regulation of dopamine signaling within mesocorticolimbic circuits. However, the role of KOR signaling to pain-related affect and functional integration in migraine has not been characterized.

Methods

We measured KOR availability using positron emission tomography (PET) with [¹¹C]LY2795050, a selective KOR antagonist radioligand in 12 individuals with chronic migraine (CM; 11 F/1 M; 42.0 ± 15.1 years) and 11 healthy controls (HC; 9 F/2 M; 40.9 ± 17.0 years). PET scans included an early resting and late sustained thermal pain stimulus stress (STPTS) phase, with KOR availability quantified as non-displaceable binding potential (BP_ND). Group differences in KOR BP_ND were assessed using region-of-interest (ROI) and voxel-wise analyses focusing on corticolimbic regions. Resting-state functional MRI (rs-fMRI) was additionally acquired to explore ROI-to-ROI functional connectivity (rsFC) within predefined regions. Exploratory interaction analyses examined whether group moderated the relationship between KOR BP_ND and rsFC strength. Associations between KOR BP_ND and clinical measures of pain intensity and affective symptoms were assessed using Spearman’s rho correlation.

Results

Compared with HC, individuals with CM showed significantly elevated KOR BP_ND in the left anterior insula and lateral orbitofrontal cortex, particularly during STPTS phase. Exploratory interaction analyses revealed a differential relationship between fronto-insular KOR BP_ND and rsFC between the anterior and posterior insula. In addition, higher fronto-insular KOR BP_ND was positively correlated with greater subclinical depressive symptom scores, which were predominantly in the minimal-to-mild range, in the CM group.

Conclusions

These findings indicate altered fronto-insular KOR availability in individuals with CM and suggest that KOR signaling may contribute to affective symptom burden, even at subclinical levels. The observed associations between KOR availability and rs-fMRI measures warrant cautious interpretation and further investigation in larger samples.

Supplementary Information

The online version contains supplementary material available at 10.1186/s10194-026-02398-9.

Background

Migraine is a multisymptomatic neurological disorder characterized by debilitating headache accompanied by photophobia and phonophobia, reflecting pain hypersensitivity to normally innocuous light and sound stimuli [1]. When headache occurs on 15 or more days per month, with at least eight days meeting migraine criteria, the condition is classified as chronic migraine (CM) [2]. Individuals with CM experience substantially impaired quality of life and demonstrate elevated risk for comorbid depression [3]. Accumulating evidence indicates that emotional and motivational dysregulation represent core features of migraine pathophysiology, yet the neurochemical basis of these dimensions remain incompletely characterized [4, 5].

The kappa opioid receptor (KOR) system, endogenously activated by dynorphin, regulates both nociceptive processing and affective states [6]. Evidence from animal studies suggests that persistent pain dysregulates KOR signaling within mesocorticolimbic dopaminergic circuits, resulting in attenuated dopamine transmission and consequent negative affect, including depressive-like and aversive behavioral responses [4, 7–9]. These findings implicate maladaptive KOR signaling in the affective vulnerability associated with chronic pain [8].

Despite substantial preclinical evidence implicating KOR signaling in pain and affective processing, its role in human chronic pain disorders, including CM, remains poorly understood. To address this knowledge gap, human neuroimaging studies are critically needed to determine whether KOR availability is altered in patients with CM and such alterations relate to affective symptom burden, a prominent clinical feature of the disorder.

In the present study, using PET with the selective KOR antagonist radioligand [¹¹C]LY2795050, we examined whether KOR availability in corticolimbic regions, including the nucleus accumbens, amygdala, hippocampus, medial prefrontal cortex, rostral anterior cingulate cortex, medial and lateral orbitofrontal cortex, and anterior and posterior insula, differed between individuals with CM and healthy controls, and whether it was associated with depressive symptom severity. Pain-related variables (thermal pain threshold, self-reported average headache intensity) were examined as secondary outcomes.

We hypothesized that (1) individuals with CM would exhibit greater KOR BP_ND in corticolimbic regions relative to healthy controls, and (2) higher KOR BP_ND would correlate positively with depressive symptom severity. We additionally examined whether the relationship between KOR availability and ROI-to-ROI resting-state functional connectivity differed between groups.

Methods

Study participants

We recruited participants aged 18–65 years, including patients with CM and age- and sex-matched HCs. CM was defined according to the International Classification of Headache Disorders, 3rd edition (ICHD-3), as headache occurring on ≥ 15 days per month for more than 3 months, with at least eight days per month fulfilling migraine criteria [2]. Initial eligibility was assessed via structured telephone prescreening. Individuals who passed the prescreening were invited for an in-person screening visit within 8 weeks. During this visit, participants underwent a review of medical and headache history including monthly headache frequency, years of migraine, and average pain intensity over the past month, followed by a neurological examination.

Exclusion criteria for both groups included: (1) presence of another chronic pain disorder; (2) current or past diagnosis of psychotic disorders (e.g., schizophrenia) or severe depression; (3) ongoing disability litigation; (4) current use of opioids or dopaminergic drugs; (5) history of neurological illness (e.g., stroke, epilepsy); (6) use of investigational drugs or devices within 30 days prior to enrollment; and (7) any severe medical condition considered by the principal investigator to interfere with study procedures. The study was conducted in accordance with the Declaration of Helsinki and was approved by the University of Michigan Institutional Review Board (HUM00165387). All participants provided written informed consent prior to enrollment.

Measurement

Clinical assessment

Questionnaire-based clinical measures were collected at two study time points. At the screening visit, participants completed the Beck Depression Inventory-II (BDI-II; 21 items, range 0–63) [10], the Allodynia Symptom Checklist-12 (ASC-12; 12 items, range 0–24) [11], and the Short-Form McGill Pain Questionnaire-2 (SF-MPQ-2; 22 items, sensory and affective subscales scored as mean item ratings) [12]. The Positive and Negative Affect Schedule (PANAS, 20 items; Positive Affect and Negative Affect subscales, each range 10–50) [13] was administered at the screening, MRI, and PET sessions; the values reported in this study were collected immediately before PET scanning to capture affective state proximal to radiotracer acquisition. Among these measures, BDI-II was designated as the primary clinical outcome. PANAS and SF-MPQ-2 were included in supplementary correlation analyses, and ASC-12 was used for clinical characterization. Additional psychometric details are provided in Supplementary Table S2.

Imaging data acquisition

PET and MRI examinations were usually conducted on separate days, with an intersession interval of 0 to 31 days (median, 3.0 days). One participant completed both sessions on the same day because of scheduling constraints. The PET session consisted of a single 90-min acquisition without interruption, and the full visit lasted about 3 h, including clinical intake, radiotracer preparation, and post-scan procedures. The MRI session included T1-weighted structural imaging, resting-state fMRI, diffusion tensor imaging, and magnetic resonance spectroscopy, with a total acquisition time of about 40 min. No restrictions were placed on caffeine or food intake before scanning. All participants received compensation under the IRB-approved protocol (HUM00165387).

Participants were asked to refrain from abortive medication at least 48 h before each scan. Patients with CM were not required to be headache-free at the time of scanning. Headache status was recorded at each scan visit using a structured 5-category item based on the timing of the most recent migraine attack (1 = currently experiencing headache; 2 = within the last 12 h; 3 = within the last 24 h; 4 = within the last 48 h; 5 = more than 48 h ago). 5 of 12 CM participants reported current headache at the time of PET scanning. Individual headache status and medication information recorded at each visit are provided in Supplementary Table S1.

PET imaging consisted of a 90-min dynamic acquisition following a bolus injection of [¹¹C]LY2795050, a selective KOR antagonist radiotracer, developed at the University of Michigan PET Center. The total dose of 15 mCi (555 MBq) was administered, with a maximum mass of 0.03 µg per kilogram of body weight. Half of this dose was administered as an initial bolus, followed by a continuous infusion at a fixed rate of the remainder to rapidly reach the steady-state tracer levels. Logan plot analysis confirmed that steady-state was achieved within 5–7 min, with full equilibrium by 40 min. This allowed quantification of KOR binding under both early baseline (resting state, 10–40 min post injection) and late challenge (40–90 min post injection) conditions. A total of 28 frames were acquired over the scan duration.

The latter involved a 20-minute sustained thermal pain threshold stress (STPTS), conducted as previously described [14]. Briefly, heat pain stimulation was delivered to the trigeminal mandibular region (V3) ipsilateral to the participant’s worst headache side using a 16 mm² surface thermode (MEDOC Pathway, Ramat Yishai, Israel). Starting from 32 °C, temperature increased at 1 °C/s until the participant indicated pain onset, then returned automatically to baseline; maximum temperature was capped at 50 °C. This threshold-tracking cycle was repeated continuously over 20 min. Thermal pain threshold was defined as the mean temperature (°C) at first pain perception across STPTS cycles.

MRI data were acquired using a 3T GE MR750 scanner (GE Healthcare, Milwaukee, WI, USA). High-resolution structural images were obtained with a T1-weighted spoiled gradient recalled echo (SPGR) sequence (TR = 12 ms, TE = 5 ms, inversion time = 500 ms, flip angle = 15°, field of view = 256 mm, voxel size = 1.0 × 1.0 × 1.0 mm). Resting-state functional MRI (rs-fMRI) data were collected using a multiband echo-planar imaging sequence (TR = 800 ms, TE = 30 ms, flip angle = 52°, field of view = 216 mm, voxel size = 2.4 × 2.4 × 2.4 mm), with a total of 379 volumes acquired. During the rs-fMRI acquisition, participants were instructed to fixate on a central cross presented on a screen, relax, and refrain from engaging in focused thoughts. Head motion was minimized using foam cushions, and physiological signals, including cardiac and respiratory activity, were simultaneously recorded using a finger pulse oximeter and an abdominal pressure belt.

Preprocessing

PET-KOR availability

Dynamic KOR data were motion-corrected by realigning each image frame to a reference volume. Parametric images were generated using a modified Logan graphical analysis with the cerebellar cortex as a reference region, under the assumption of negligible KOR density [15, 16]. For each scan, three sets of parametric images were computed: (1) a tracer delivery index (K1 ratio), proportional to cerebral blood flow; and binding potential non-displaceable (BP_ND) images derived from (2) 10–40 min post-injection (baseline resting) and (3) 40–90 min post-injection (during STPTS phase). The Logan plot linearized within 10 min of tracer administration, and the slope of the linear portion yields the distribution volume ratio (DVR), from which BP_ND was calculated as DVR − 1. BP_ND is a dimensionless index of receptor availability, proportional to Bmax/KD, where Bmax is the density of receptors available to the radioligand and KD is the equilibrium dissociation constant [17]. Parametric images were co-registered to each subject’s T1-weighted structural MRI for group-level analysis in standard space.

Resting-state fMRI

MRI data were first organized in the Brain Imaging Data Structure (BIDS) format for standardized data management and reproducibility of the preprocessing and analysis pipeline [18]. Resting-state functional MRI data were preprocessed using the fMRIPrep pipeline, a robust and standardized pipeline that integrates tools from FMRIB Software Library (FSL), Advanced Normalization Tools (ANTs), and FreeSurfer for anatomical and functional imaging processing [19]. First, distortions caused by magnetic field inhomogeneities in the functional images were corrected using field maps in FSL, followed by slice-timing correction, motion correction, and co-registration to the individual’s T1-weighted structural image. Brain tissue segmentation was performed using FAST, and images were subsequently normalized to MNI152 2 mm standard space using ANTs nonlinear registration. Temporal denoising was conducted by regressing out six motion parameters and aCompCor components extracted from white matter and cerebrospinal fluid (CSF) in fMRIPrep using FSL. Additionally, a temporal bandpass filter (0.01–0.1 Hz) was applied using AFNI’s 3dBandpass to remove low-frequency drift and high-frequency noise. Finally, spatial smoothing was applied using the Smallest Univalue Segment Assimilating Nucleus (SUSAN) algorithm in FSL with a 5 mm FWHM Gaussian kernel. This kernel size approximates twice the acquired voxel size (2.4 mm isotropic), consistent with established recommendations for fMRI spatial smoothing [20].

Statistical analysis

ROI and voxel-wise analyses of KOR BP_ND

Region-of-Interest (ROI) analysis was performed on predefined corticolimbic regions. Subcortical ROIs included the nucleus accumbens (NAc), amygdala, and hippocampus, known for their roles in reward, motivation, and affective dimensions of pain [21]. Cortical ROIs included the medial prefrontal cortex, rostral anterior cingulate cortex, medial and lateral orbitofrontal cortex, and anterior and posterior insular cortex, selected for their involvement in cognitive and emotional regulation of pain [22] (Supplementary Figure S1). Prefrontal and subcortical ROIs were derived from the Brainnetome atlas [23]. Insular ROIs were derived from the Destrieux atlas [24]. All ROIs were defined bilaterally. The anterior and posterior insular ROIs were derived from the Destrieux atlas [24], because its sulco-gyral parcellation follows the central insular sulcus and provides a more anatomically direct separation of the anterior and posterior insula.

For each ROI, mean BP_ND values were extracted using FSL’s fslstats (FSL v6.0, FMRIB, Oxford, UK). Group differences between CM and HC were tested using the Mann-Whitney U test on age-adjusted residuals derived from linear regression of BP_ND on age. Results were initially interpreted at an uncorrected significance level of P < 0.05. To correct for multiple comparisons across ROIs, a Bonferroni correction was applied, yielding a corrected significance threshold of P < 0.0028 (α = 0.05/18).

To assess group differences between individuals with CM and HC and identify spatially localized effects within predefined corticolimbic ROIs, voxel-wise comparisons of KOR BP_ND were conducted using SPM12 (Wellcome Trust Centre for Neuroimaging, London, UK). Statistical maps were first thresholded at a voxel-wise level of P < 0.001 (uncorrected) with a minimum cluster extent of 10 contiguous voxels, which aligns with previous receptor PET studies in small samples [25–27]. Small-volume correction (SVC) was applied to account for multiple comparisons across voxels within each ROI, and statistical significance was defined as P_FWE < 0.05 within the corresponding ROI mask. Of note, the ROI-level analysis assessed group differences in mean BP_ND averaged across each predefined mask using a non-parametric test. In parallel, the voxel-wise SPM analysis tested for spatially localized group differences within the same ROI volumes using the standard parametric GLM framework implemented in SPM for small-volume-corrected inference.

All analyses were carried out on KOR BP_ND parametric images derived from both the early resting condition (10–40 min) and the late STPTS condition (40–90 min). Due to technical issues during PET acquisition in the late condition, data from 9 of 12 individuals with chronic migraine and 9 of 11 healthy controls were included in the STPTS analyses.

For the subgroup analysis, we divided patients into two groups based on the time elapsed since their last migraine attack before the PET scan (< 12 h vs. > 12 h). This analysis aimed to explore whether recent migraine activity, indicated by the attack-to-PET scan interval, affected the KOR availability. Group differences were tested using the Mann–Whitney U test.

Resting-state FC and its association with KOR BP_ND

From the preprocessed fMRI data, the mean BOLD time series for each ROI was extracted using the NiftiLabelsMasker function from Nilearn [28]. Pairwise Pearson correlation coefficients were computed between all ROI time series and then Fisher z-transformed. For each of the 153 unique ROI pairs, a general linear model (rsFC ~ group + BP_ND + group × BP_ND) was fitted using the ordinary least squares implementation in Statsmodels. Correction for multiple comparisons across all 153 connections was performed using the Benjamini–Hochberg false discovery rate (BH-FDR) procedure with q = 0.05 [29]. In addition, given the modest sample size, we conducted a complementary exploratory analysis to identify connectivity patterns that may warrant further investigation. Specifically, connections were ranked by the absolute t-statistic of the interaction term, and the top 10% of connections were selected for follow-up permutation testing using a max-t procedure (5,000 permutations), implemented in a custom Python script using SciPy. This approach does not provide FWE control across the full set of connections and was therefore used solely to prioritize potentially relevant effects in a hypothesis-generating framework [30].

Clinical correlation

All clinical correlation analyses were performed in Python using Spearman’s rank correlation (scipy.stats.spearmanr). In the main analysis, correlations were tested between KOR BP_ND in the combined anterior insula/lateral OFC cluster and three pre-specified clinical variables: BDI-II, average headache intensity over the previous month, and thermal pain threshold. Benjamini–Hochberg FDR correction (q = 0.05) was applied across these three tests. Additional correlations across individual bilateral ROIs and a broader set of clinical and affective variables were treated as exploratory and are reported in the Supplementary Materials.

Results

Clinical characteristics

As summarized in Table 1, participants with CM had a mean disease duration of 28.0 years. The CM group exhibited higher levels of depressive symptoms (BDI-II: 6.4 ± 6.4) than healthy controls (HCs: 2.5 ± 3.9), although the difference was not statistically significant (P = 0.16).

Table 1.

Demographic and clinical characteristics of study participants

Characteristic	Healthy controls	Chronic migraine	Group difference
n	11	12
Age, years	40.9 ± 17.0 [21, 64]	42.0 ± 15.1 [18, 61]	U = 66, p = 1.00
Sex (F/M)	9/2	11/1	p = 0.59†
BDI-II score	2.5 ± 3.9 [0, 13]	6.4 ± 6.4 [0, 16]	U = 89, p = 0.16
Frequency (days/month)	N/A	19.2 ± 4.2 [15, 30]	-
Years of migraine	N/A	28.0 ± 16.8 [2, 54]	-
Pain intensity (0–10)	N/A	5.5 ± 2.1 [2, 8]	-
ASC-12 score	N/A	5.1 ± 3.7 [0, 11]	-
SF-MPQ-2 sensory	N/A	3.8 ± 2.1 [0.8, 7.2]
SF-MPQ-2 affective	N/A	4.2 ± 1.6 [0.5, 6.0]
PET scan
PANAS-positive	33.9 ± 8.1 [24, 50]	26.9 ± 8.7 [13, 40]	U = 38, p = 0.10
PANAS-negative	11.3 ± 2.4 [10, 18]	11.8 ± 2.3 [10, 16]	U = 77, p = 0.49
Thermal threshold (°C)ᵃ	44.6 ± 4.3 [35.7, 48.7]	46.0 ± 3.4 [40.4, 49.4]	U = 53, p = 0.29

Abbreviations: F/M, Female/Male; BDI-II, Beck Depression Inventory-II; ASC-12, Allodynia Symptom Checklist-12; SF-MPQ-2, Short-Form McGill Pain Questionnaire-2; PANAS, Positive and Negative Affect Schedule

Data are presented as Mean ± SD [min, max]. Group differences for continuous variables tested using Mann–Whitney U test. † Fisher’s exact test. ᵃ n = 9 per group. SF-MPQ-2 sensory and affective scores represent the mean of 18 sensory and 4 affective descriptors, respectively (each calculated 0–10)

In addition, CM individuals reported lower positive affect scores on the PANAS scale compared to HCs (26.9 ± 8.7 vs. 33.9 ± 8.1), although the difference did not reach statistical significance (P = 0.10). No significant group difference was observed in negative affect scores (11.8 ± 2.3 vs. 11.3 ± 2.4, P = 0.49). As expected, depressive symptoms in the CM group were positively correlated with self-reported pain intensity over the last month (rho = 0.73, P = 0.007). The full Spearman correlation matrix across all clinical variables within the CM group is provided in Supplementary Table S3.

ROI and voxel-wise results

ROI-based analyses revealed that, during the late STPTS condition, KOR BP_ND in the left and right anterior insular cortices was greater in the CM group than in HCs (age-adjusted P < 0.05, uncorrected across ROIs). However, these effects did not survive conservative Bonferroni correction for the number of ROIs examined (Fig. 1).

Fig. 1.

KOR availability across PET sessions by diagnostic group. (A) average group-level maps of KOR BP_ND during early resting and late sustained thermal pain threshold stress (STPTS) challenge conditions for healthy controls (HC) and chronic migraine (CM) participants. (B) Region-of-interest quantification of BP_ND across mesocorticolimbic network regions during early resting (left panel) and late STPTS (right panel) conditions. Bars represent the group median ± 95% confidence interval. CM participants demonstrated consistently elevated KOR BP_ND relative to HCs across most examined regions. During the late STPTS condition, bilateral anterior insula showed greater KOR BPnd in the CM group after controlling for age (* P < 0.05, uncorrected across ROIs); however, these differences did not survive Bonferroni correction for the number of ROIs

Complementing these ROI findings, voxel-wise analysis identified a cluster showing 26.9% greater KOR BP_ND in the CM group relative to HCs, located across the lateral OFC and anterior insula (peak MNI coordinates: X = -38, Y = 22, Z = -16) at a voxel-wise threshold of uncorrected P < 0.001 and k > 10 (Fig. 2). The peak within this cluster survived small-volume FWE correction for both the left lateral OFC (peak P_FWE = 0.015; cluster P_FWE = 0.055) and the left anterior insula (peak P_FWE = 0.018; cluster P_FWE = 0.016). Detailed cluster-level statistics, including cluster extent, peak MNI coordinates, peak T-values, and FWE-corrected p-values, are provided in Table 2. No group differences were observed during the early resting condition, and no regions showed lower BP_ND in the CM group compared to controls.

Fig. 2.

Group differences in KOR BP_ND during late sustained thermal pain threshold stress (STPTS) condition. This analysis revealed a significant cluster showing greater KOR BP_ND in patients with chronic migraine (CM) compared to healthy controls (HC). The cluster was located in the left anterior insular cortex and lateral orbitofrontal cortex (OFC). Results are shown at P < 0.005 (red) and P < 0.001 (yellow), uncorrected, for visualization purposes. Right bar graph displays mean KOR BP_ND within the significant OFC/insular cluster for each group, with individual data points overlaid. Error bars represent standard error of the mean

Table 2.

Voxel-wise group differences in KOR BP_ND during late STPTS condition (CM > HC)

SVC ROI	Hemi	k	Peak T	Peak MNI (x, y,z)	PFWE (peak)	PFWE (cluster)
Lateral OFC	L	21	5.45	−38, 22, − 16	0.015	0.055
Anterior insula	L	29	4.68	−36, 18, − 10	0.018	0.016

Voxel-level threshold: P < 0.001 uncorrected, k ≥ 10. Small-Volume Correction applied within each predefined ROI mask. No clusters survived whole-brain FWE correction. CM: n = 9; HC: n = 9

To explore whether proximity to migraine attack influenced KOR availability, CM participants were stratified based on time elapsed since their last attack (< 12 h vs. >12 h). Of the 12 patients with CM, 7 were scanned within 12 h of their last migraine attack and 5 were scanned more than 12 h after (early resting phase). For the late STPTS phase, complete data were available for 5 and 4 patients in the < 12-hour and > 12-hour groups, respectively.

In the early resting phase, patients with CM scanned during or within 12 h of their last attack (ictal and immediate post-ictal) showed higher KOR BP_ND in the left amygdala, right hippocampus, left anterior insula, and posterior insula, compared with those scanned at later time points (uncorrected P < 0.05). No significant differences were detected during the late STPTS phase between these groups (Supplementary Fig. 2).

Resting-state FC associated with KOR BP_ND

No resting-state functional connectivity (rsFC) connections survived correction for multiple comparisons using the BH-FDR procedure. As an exploratory follow-up analysis, we examined the subset of connections ranked within the top 10% by the absolute t-statistic of the group × KOR BP_ND interaction term. Within this subset, permutation-based testing identified a significant interaction for rsFC between the left posterior insula and right anterior insula (t = − 4.23, uncorrected P = 0.0008, adjusted P = 0.013). Two additional connections showed marginal effects, including the left anterior insula–left posterior insula (t = − 3.41, uncorrected P = 0.004, adjusted P = 0.060) and the left lateral OFC–left hippocampus connection (t = − 3.49, uncorrected P = 0.004, adjusted P = 0.052) (Fig. 3).

Fig. 3.

Interactions between group and KOR BP_ND on resting-state functional connectivity (rsFC) within mesocorticolimbic region pairs. The interactionsare displayed for three ROI pairs: left anterior insula–left posterior insula (left), left posterior insula–right anterior insula (center), and left lateral orbitofrontalcortex–left hippocampus (right): Shaded areas indicate 95% confidence intervals. Abbreviations: aINS, anterior insula; pINS, posterior insula; lOFC, lateralorbitofrontal cortex

Across these connections, the interaction effects reflected a consistent pattern whereby higher KOR BP_ND was associated with lower rsFC in the CM group, whereas the association was positive in HCs. Given the absence of effects surviving whole-network correction and the limited statistical power of this preliminary sample, these rsFC findings are considered exploratory and should be interpreted as hypothesis-generating pending replication in larger cohorts.

Clinical correlation

Among individuals with CM, elevated KOR BP_ND in the anterior insula and lateral OFC was positively correlated with depressive symptom severity, as measured by the BDI-II (rho = 0.77, P = 0.016, adjusted P = 0.048). In addition, KOR BP_ND was positively associated with pain intensity scores over the last month (rho = 0.68, P = 0.045. adjusted P = 0.068). However, this association did not survive FDR correction and should be interpreted cautiously (Fig. 4).

Fig. 4.

Associations between KOR BP_ND and clinical variables in the chronic migraine group. (A) KOR BP_ND in the anterior insula and lateral orbitofrontal cortex was positively correlated with depressive symptom severity (rho = 0.77; P = 0.016). (B) Positive correlation was observed between KOR BP_ND and pain intensity over the past month (rho = 0.68; P = 0.045)

When examining individual bilateral ROIs separately, rather than the combined anterior insula/lateral OFC cluster used in the main analysis, pain-related variables did not show consistent associations with BP_ND across ROIs, whereas BDI-II showed a broader positive pattern across multiple regions (Supplementary Tables S4–S5).

Discussion

The present study identified elevated KOR BP_ND in participants with CM compared to HCs, with notable differences observed in the left anterior insula and lateral OFC during the STPTS paradigm (thermal challenge). This elevation correlated with lower corticolimbic rsFC and greater depressive symptoms in individuals with CM. Our findings extend prior evidence that the KOR system contributes to affective disturbances in chronic pain state [6, 7]. Whether elevated KOR availability reflects a peri-ictal state, a more stable trait-related characteristic of CM, or a combination of both remains uncertain. BP_ND was numerically higher in participants scanned within 12 h of their last attack (n = 7) than in those scanned later (n = 5), although this comparison should be interpreted with caution given the small subgroup sizes. Notably, the near-attack subgroup also reported higher BDI-II scores (10.0 ± 5.9 vs. 1.4 ± 2.6; P = 0.01), which makes it difficult to separate possible effects of attack proximity from concurrent affective burden. Because [¹¹C]LY2795050 BP_ND cannot distinguish receptor density from endogenous dynorphin occupancy, the neural basis of this pattern cannot be resolved from the current data.

While the dynorphin/KOR system mediates adaptive stress responses under normal conditions, chronic pain promotes affective dysregulation, contributing to the association between pain and psychiatric disorders [6, 8]. Theoretically, elevated endogenous dynorphin and hyperactivation of KOR in chronic pain should compete with PET radioligands, yielding reduced KOR BP_ND [31]. Contrary to this prediction, we observed significantly increased KOR BP_ND in the insula and OFC of individuals with CM. This finding suggests that KOR alterations in CM reflect complex maladaptive neuroplasticity rather than simple receptor upregulation or occupancy.

Functionally, the anterior insula and lateral OFC are core nodes of the salience and reward networks, respectively, and are increasingly recognized as key modulators of affective state and internal monitoring [32]. The anterior insula serves as a central hub for interoceptive processing and plays a crucial role in integrating affective information into cognitive and motivational processes [33]. In migraine, altered anterior insular connectivity has been reported, linking this region to abnormal sensory gain and interoceptive processing [34, 35]. Preclinical evidence indicates that KOR directly modulates excitability within insular circuits through a dynorphin/KOR microcircuit [36], suggesting a mechanism through which altered KOR availability could affect insular function.

In addition, orbitofrontal dysfunction, as indexed by impaired decision-making and white matter microstructural changes, has been documented across the migraine spectrum [37, 38]. The lateral OFC has been specifically implicated in encoding non-reward and aversive outcome [39]. Thus, elevated KOR availability in this subregion may reflect altered affective valuation processes in CM.

The voxel-wise analysis localized the peak group difference to the left hemisphere. In a meta-analytic study, chronic pain conditions are characterized by dysregulated left anterior insular reactivity [40]. In migraine, reduced gray matter volume in the left dorsal anterior insula has been reported and correlated with migraine frequency and duration [41]. The present left-lateralized finding is partly consistent with this pattern, though larger samples will be needed to determine whether it reflects a reproducible feature of CM.

As synthesized in a recent review [42], the KOR system functions as a homeostatic gatekeeper, limiting excessive excitation of motivational circuits. Clinically, this mechanism aligns with the behavioral phenotype of migraine attacks, which exhibit pronounced sensory hypersensitivity (e.g., allodynia) and aversion (e.g., photophobia, phonophobia), necessitating withdrawal from novel stimuli and cessation of environmental exploration to prevent sensory overload. This interpretation is supported by recent findings that KOR antagonism boosts the motivational value of novel stimuli [43]. The elevated KOR availability observed in our CM group likely inhibits these adaptive behavioral responses.

Regarding the molecular mechanisms underlying this observation, we hypothesized two non-exclusive mechanisms to explain our findings. First, the recurrent nature of migraine attacks may trigger a compensatory upregulation of KOR density, a phenomenon documented in other conditions involving stress exposure [44]. This increase in receptor density could outweigh the competitive effects of endogenous dynorphin against the radiotracer, resulting in a net increased BP_ND. Second, chronic high-frequency migraine attacks may cause the dynorphin release rate to exceed synthesis and transport from the cell body [45]. This could lead to depletion of the releasable dynorphin pool at the synapse, a state consistent with allostatic load models in chronic stress conditions [46].

This dynorphin depletion hypothesis is further supported by our subgroup analysis. Individuals with a shorter interval since the last migraine attack (< 12 h) exhibited higher insular BP_ND at rest compared to those scanned later. In specific animal model studies, dynorphin was found to be more susceptible to depletion than other opioid peptides, such as beta-endorphin and enkephalin [47, 48]. Given that CM patients exist along a continuum between pre-ictal, ictal, and post-ictal states, we propose that ictal and immediate post-ictal dynorphin depletion may transiently increase KOR availability by reducing endogenous competition.

Interestingly, we found that KOR availability in the anterior insula and lateral OFC was positively correlated with depressive symptom scores. Notably, this pattern aligns with recent findings by Slifstein et al. [49] who demonstrated that cortical KOR availability was associated with anhedonia in schizophrenia. This suggests that KOR alterations may be more closely linked to motivational deficits and reward insensitivity than to mood disturbance per se. Broadly, these results corroborate preclinical evidence showing that intense or repeated stress leads to concurrent dynorphin/KOR system upregulation and aversive emotional responses [50]. This phenomenon has been consistently reported across models of depression, addiction vulnerability, and stress-induced anxiety, and is reversible by KOR antagonists [51].

Although participants with CM in our study exhibited only minimal-to-mild depressive symptoms, we hypothesize that increased KOR availability in these affect-related regions may contribute to depression vulnerability or represent an early neurochemical marker preceding explicit mood disorders. However, our interpretation requires caution due to the unexpected observation that patients scanned closer to an attack also reported higher BDI-II scores (10.0 ± 5.9 (n = 7) vs. 1.4 ± 2.6 (n = 5); Mann-Whitney U: 33.0; P = 0.01). Therefore, increased KOR BP_ND likely represents a composite signal reflecting both temporal proximity to the last attack and state-dependent affective distress.

Additionally, we observed differential associations between KOR availability and functional integration in CM vs. HC across three anatomically distinct pathways including the right anterior insula and left posterior insula, the left anterior and posterior insula, and the left lateral OFC and hippocampus. These findings are interesting given the functional specialization along the posterior-to-anterior insular axis, where the posterior insula processes interoceptive and somatosensory information, while the anterior insula integrates these signals with affective valence and coordinates salience detection [52, 53]. The opposing directions of KOR BP_ND-FC associations between groups suggest that KOR system alterations in CM may relate to reorganized functional architecture within networks supporting interoceptive awareness, affective processing, and salience attribution [54].

Despite these novel findings, several methodological limitations should be acknowledged. First, the small sample size limits statistical power and precludes robust examination of potential moderating factors, such as sex. Importantly, previous preclinical studies have demonstrated distinct sex-dependent mechanisms in KOR-mediated aversion and analgesia [55, 56]. Since our sample was predominantly female, our findings cannot be generalized to male patients without further validation. Second, our FC analysis employed an exploratory thresholding strategy (top 10% connections). While this approach was chosen to minimize Type II errors in this preliminary investigation, findings should be interpreted as hypothesis-generating rather than confirmatory. Third, the BP_ND represents net receptor availability, a composite measure influenced by multiple factors including receptor density, affinity, and endogenous ligand occupancy, which precludes definitive conclusions about the specific molecular mechanisms underlying observed group differences. Lastly, although participants using opioid or dopaminergic medications were excluded, several CM participants reported use of other CNS-acting agents (e.g., alprazolam, lasmiditan). Medication-related confounding cannot be fully excluded, particularly for agents with GABAergic mechanisms, and future studies with larger samples should incorporate medication status as a covariate.

Conclusions

This first in vivo neuroimaging study demonstrates that individuals with CM exhibit regionally elevated KOR availability within corticolimbic circuitry, specifically in the anterior insula and lateral OFC, which correlates with depressive symptom burden. While these findings are consistent with the established role of KOR in mediating stress-induced negative affect, they may also point to a role in modulating sensory and affective salience. Collectively, these findings provide neurobiological evidence for understanding opioid system contributions to CM and underscore the need for mechanistic studies to determine whether such KOR alterations reflect compensatory adaptations to sensory overload, pathological dysregulation, or trait-like vulnerability factors.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Abbreviations

aCompCor

Anatomical component-based noise correction

AFNI

Analysis of Functional NeuroImages

ANTs

Advanced Normalization Tools

ASC-12

Allodynia Symptom Checklist-12

BDI-II

Beck Depression Inventory-II

BIDS

Brain Imaging Data Structure

BP_ND

Non-displaceable binding potential

CM

Chronic migraine

CNS

Central nervous system

CSF

Cerebrospinal fluid

EPI

Echo-planar imaging

FDR

False discovery rate

fMRI

Functional magnetic resonance imaging

FSL

FMRIB Software Library

FWE

Family-wise error

FWHM

Full width at half maximum

HC

Healthy control

ICHD-3

International Classification of Headache Disorders, 3rd edition

KOR

Kappa opioid receptor

MNI

Montreal Neurological Institute

MRI

Magnetic resonance imaging

OFC

Orbitofrontal cortex

PANAS

Positive and Negative Affect Schedule

PET

Positron emission tomography

ROI

Region of interest

rsFC

Resting-state functional connectivity

SPM

Statistical Parametric Mapping

STPTS

Sustained thermal pain threshold stress

SVC

Small-volume correction

SUSAN

Smallest Univalue Segment Assimilating Nucleus

Author contributions

R.K., P.S., and A.D. were responsible for the conceptualization and design of the study. D.K., M.S., T.K., and X.S. performed the PET and MRI data acquisition and processed the neuroimaging data. D.K. was a major contributor in performing the statistical analyses and drafting the manuscript. D.K., F.P., and A.D. critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

Funding

This research was supported by the grant R01 NS094413 from the National Institute of Neurological Disorders and Stroke (NINDS) awarded to Dr. DaSilva and the New Faculty Research Support Grant from Changwon National University in 2025 exclusively awarded to Dr. Kim. The funding agency had no role in the design or conduct of the study.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and was approved by the University of Michigan Institutional Review Board (HUM00165387). All participants provided written informed consent prior to enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.