Kappa opioid receptor availability in borderline personality disorder: An in-vivo investigation with [¹¹C]EKAP PET imaging

Abstract

Background:

Borderline personality disorder (BPD) is a chronic mental health condition associated with marked elevations in disability. Limited pharmacological treatment options are currently available, and kappa opioid receptors (KOR) have been identified as a promising treatment target. However, the KOR system has not yet been directly examined in BPD. This study examined KOR in individuals with BPD and investigated links between KOR and BPD’s clinical features.

Methods:

Twenty individuals with BPD and 20 demographically matched healthy control (HC) volunteer participants were scanned with [¹¹C]EKAP positron emission tomography (PET) and completed self-report and interview measures. KOR availability (volume of distribution; V_T) was measured in brain regions of interest (ROIs) relevant to the pathophysiology of BPD.

Results:

No differences in age or sex were observed between HCs (Mean age: 32.94; SD=8.45; 60% female), and individuals with BPD (Mean age: 33.30; SD=9.34; 60% female). In individuals with BPD, KOR V_T was significantly lower in the amygdala (M_diff=−8.04; 95% CI[11.40; −4.69]), anterior cingulate (M_diff=−2.95; 95% CI [−5.22; −0.68]), hippocampus (M_diff=−2.12; 95% CI [−3.37; −0.88]), insula (M_diff=−3.58; .95% CI[−5.71; −1.46]), and frontal ROIs (M_diff=−1.83; 95% CI[−3.28; −0.38]). V_T correlated negatively with specific clinical features including impulsivity (r(28)=−.45 p=.013, 95% CI[−.72; −.11]) and depressive symptoms (r(28)=−.43, p=.021, 95% CI[−.70, −.07]), but not trauma symptoms.

Conclusions:

Findings demonstrate differences in KOR availability in BPD and relationships between KOR availability and clinically relevant features. Results highlight the relevance of KOR in BPD pathophysiology and further implicate the opioid system as a promising target for treatment development in BPD.

Introduction

Borderline personality disorder (BPD) is a severe and disabling psychiatric condition linked to emotional dysregulation and affective instability, disruption of interpersonal relationships and self-image, and engagement in impulsive and high-risk behaviors (1). BPD represents a significant public health concern: despite prevalence estimates ranging from 1–6%(1, 2), BPD accounts for up to 25% of required inpatient hospitalizations (3) in the United States (US), and is associated with markedly elevated rates of suicide attempts and deaths (1). BPD is notoriously resistant to pharmacological treatment (4); while available medications have demonstrated some efficacy in targeting specific BPD traits (e.g., impulsivity) (4), they typically do not effectively address overall severity and associated functional impairment (5). A deeper understanding of the pathophysiology of BPD is necessary to inform the development of targeted therapies that can reduce risk and symptom severity, and improve well-being.

The endogenous opioid system has long been considered as a potential target for mood and affect regulation difficulties (6). Opioid receptors, including mu (MOR), delta (DOR), and kappa (KOR), are densely concentrated in the frontolimbic network (7). In BPD, fMRI studies have demonstrated abnormal functioning within this same frontolimbic circuit, including hyperactivity in limbic areas (amygdala; insula) coupled with hypoactivity in frontal areas (8). Opioid receptors play key roles in modulating features central to BPD including physical and affective pain (9), emotion regulation (6, 10), social attachment (11), and impulsivity (6). Finally, impulsive, high-risk, and sensation-seeking behaviors commonly observed in BPD may result from efforts to compensate for opioid system dysfunction (10, 12) by stimulating endogenous opioid release (10). Lower levels of some opioid peptides (β-endorphin and met-enkephalin) have been observed in plasma (10) and cerebrospinal fluid (12) in individuals with BPD relative to healthy controls (HCs) (10) and individuals with psychiatric disorders with no history of self-injurious behaviors (12). Additionally, individuals with BPD taking opiate medications report euthymia rather than euphoria following ingestion (10).

To date, no controlled studies have examined the effects of opioid agents on overall BPD symptomology (13). However, interventions targeting the opioid system have reduced some of the most dangerous, intractable symptoms of BPD. Clinical case studies have shown reductions in self-injurious behaviors (13) and dissociation (4) following treatment with opioid antagonists (e.g., naltrexone (4, 13)). Additionally, research implicates the opioid system in the efficacy of ketamine (14). Ketamine is one of few medications shown to reduce suicide risk (15), but is potentially problematic for use in conditions often characterized by dissociative symptoms (15). Currently, many available opioid agents primarily target MOR, which plays a demonstrated role in reward processing and mood regulation (11). More limited research has supported the utility of DOR agonists as antidepressant (16). Unfortunately, enthusiasm for continued MOR-targeted drug development has been dampened by limited efficacy (4) and concerns about abuse liability (6). Similarly, development of DOR targeted agents has been limited by concerns regarding observed convulsive side-effects (16).

Alternatively, growing literature has implicated KOR as a potential target for intervention in BPD (17). KOR antagonists pose a lower risk for abuse potential (18) and adverse side effects. While administration of KOR agonists can result in dysphoric (19) and dissociative effects (20), both animal models ((e.g., 21)) and recent human studies ((e.g., 22)) have linked KOR antagonism to antidepressant and anxiolytic effects. Notably, KOR antagonists offer antidepressant and anxiolytic effects together (17), potentially conferring distinct clinical benefits in individuals with BPD, whose presentations are often characterized by high complexity and comorbidities (1). Moreover, KOR and associated endogenous peptide dynorphin play key roles in emotional dysregulation (17, 23), and activation of KOR is implicated in development of mood, cognitive, and dissociative symptoms consistent with BPD (24). Additionally, in nonhuman primates, elevated dynorphin levels are associated with dissociation and impaired cognitive functioning (20). Preclinically, KOR activation has also been implicated in the maintenance of interpersonal relationships (25) and shown to mediate the relationship between pain and depressed mood (9) by depressing mesolimbic dopamine release. Both human and preclinical work implicate KOR antagonists as a viable treatment target for BPD. The efficacy of KOR antagonists may be associated in part with KOR’s ability to reduce the impact of stress on mood (26). While trauma exposure is not required to meet diagnostic criteria for BPD, past work estimates that up to 90% of individuals with BPD experience early life trauma (27). Strong preclinical evidence supports a relationship between stress, dynorphin release, and KOR activation, with relationships also demonstrated for MOR and DOR subtypes (23). Finally, extensive preclinical evidence suggests that KOR mediates the relationship between stress exposure and the development of psychopathology (26).

Together, robust evidence suggests that KOR antagonism represents a highly promising potential treatment target for BPD. Using [¹¹C]EKAP positron emission tomography (PET), it is possible to quantify KOR availability in vivo in the living human brain. However, no studies have directly investigated KOR availability in individuals with BPD. In the present study, we aimed to examine KOR availability in participants with BPD relative to demographically matched HCs. We hypothesized that due to increased occupancy of KOR by dynorphin (28), individuals with BPD would have significantly lower KOR availability relative to HCs in frontal and limbic brain regions relevant to the pathophysiology of BPD (8). Second, we aimed to examine whether KOR availability is linked to endophenotypic correlates of BPD. Specifically, we hypothesized that KOR availability would be negatively correlated with clinical features including emotion dysregulation and impulsivity, depressive and anxiety symptoms, BPD symptoms, and trauma exposure.

Methods and Materials

Participants

Forty participants were selected from ongoing studies based on demographic matching (age and sex) across groups. The BPD group included 20 individuals meeting diagnostic criteria for DSM-5 BPD recruited from the local community around New Haven, CT, between 2021–2025 to complete PET scanning with [¹¹C]EKAP. Initial screening procedures included physical (e.g., electrocardiography, hematology, chemistry, thyroid/liver function), psychiatric, and neurological examinations to establish diagnoses and rule out any major medical or neurological illnesses. Exclusion criteria for participants with BPD included diagnoses of psychotic or pervasive developmental disorders, current substance use disorder of any severity, or a lifetime history of moderate-to-severe substance use disorder, positive urine pregnancy test or toxicology for exclusionary substances (e.g., opiates), history of loss of consciousness for more than 5 minutes, or positive urine pregnancy test. Psychiatric medications were permitted to increase ecological validity; individuals taking medications with direct effects on the opioid system (e.g., naltrexone) were not eligible. Figure S1 displays a diagram of BPD participant recruitment. Individuals in the HC group recruited as part of this study (n=8) followed identical screening procedures and could not meet criteria for any current or past psychiatric disorder. Remaining HCs were selected from a data repository utilizing comparable scanning procedures (scanned between 2015–2016) (29) (Table 1). Detailed screening procedures including eligibility criteria for HCs from the data repository are detailed elsewhere (29). This study was approved by the Yale University Human Investigation Committee and Radiation Safety Committee (Radioactive Investigational Drug Committee; Radioactive Drug Research Committee). All participants provided written informed consent.

Table 1.

Participant Demographic and Clinical Characteristics

	BPD	HCs	Statistical Test (T, F, or chi2, p-value)
Variable	N = 20 +	N = 20+
Sex (M:F) (%F)	8:12 (60%)	8:12 (60%)	0.00, .99
Age (years) (M ± SD)	33.30 (9.34)	32.95 (8.45)	0.12, .90
BMI (M ± SD)	29.06 (5.05)	26.48 (4.63)	0.72, .10
Psychiatric Medication (Y:N)	14:6	1:19 ^	18.03, <.001
Comorbid PTSD Diagnosis (N, %)	6 (30%)	N/A	N/A
Smoking Status (Y:N)	5:15	0:20	5.71, .02
BDI (M ± SD)	20.60 (10.64)	0.63 (1.41)	8.22, < .001
STAI-S (M ± SD)	40.95 (10.27)	27.00 (6.14)	3.57, .001
STAI-T (M ± SD)	53.55 (10.92)	28.38 (6.78)	6.03, < .001
BIS (M ± SD)	72.20 (11.18)	51.13 (6.83)	4.94, < .001
DERS (M ± SD)	48.30 (12.51)	27.75 (4.50)	6.39, < .001
BSL (M ± SD)	1.12 (0.73)	0.07 (0.07)	6.37, < .001
PCL (M ± SD)	37.75 (18.07)+	N/A*	N/A
CTQ (M ± SD)	59.67 (23.29) +	28.40 (9.67)	5.54, < .001

Note: Valid percentages were used for descriptive statistics. Significant differences between group differences are bolded. Data were not available for all participants on some variables. Thus, N’s may not reflect total number of cases.

^*An insufficient number of HC were administered a PCL to facilitate group comparisons.

BDI = Beck Depression Inventory; STAI = State Trait Anxiety Inventory; BIS = Barratt’s Impulsiveness Scale; DERS = Difficulties in Emotion Regulation Scale; BSL = Borderline Symptom List; PCL = PTSD Checklist; CTQ = Childhood Trauma Questionnaire.

^{^}One HC participant reported use of fluoxetine for treatment of gastrointestinal illness.

⁺= variables with missing data for BPD group; N=18 for CTQ; 12 for PCL. Clinical measures for HCs were available for N=8, except for CTQ, where data were available for N=5.

Clinical Assessments

Participants were administered the Structured Clinical Interview for DSM-5 (SCID-5),(30) McLean Screening Instrument (31), and Clinician-Administered PTSD Scale for DSM-5 (CAPS-5) (32) to establish diagnostic status. Symptom and clinical measures included the Beck Depression Inventory (BDI-II) (33), State-Trait Anxiety Inventory (STAI) (34). Borderline Symptom List (BSL-23) (35), Barratt Impulsiveness Scale (BIS) (36), Childhood Trauma Questionnaire (CTQ) (37), PTSD Checklist for DSM-5 (38), and Difficulties in Emotion Regulation Scale (DERS-18) and its subscales (lack of emotional awareness, lack of emotional clarity, difficulties with impulse control, regulating emotions, and engaging in goal directed behaviors) (39).

MRI and PET Procedures

T1-weighted MRI scans were acquired to examine potential structural abnormalities and facilitate co-registration with PET data. MRI scans were performed on a 3-Tesla Siemens Prisma MR scanner. [¹¹C]EKAP was synthesized by the Radiochemistry Lab at the Yale PET Center as previously reported (40). Pet imaging data were acquired on state-of-the-art, high resolution Biograph Vision PET/CT (Siemens) and high-resolution research tomograph (HRRT) (Siemens) scanners. Dynamic PET data were collected for 120 minutes following single intravenous bolus administration of ≤20mCi of [¹¹C]EKAP. There were no differences between groups in [¹¹C]EKAP injected dose or mass (Table 2).

Table 2.

Radiotracer Characteristics and VT Values

	BPD	HCs	Statistical Test (T, F, or chi2, p-value)
Variable	N = 20	N = 20
Scanner: HRRT/VISION (N: HRRT)	14	12	0.44, .51
Injected dose of EKAP (mCi) (M ± SD)	12.65 (4.88)	15.03 (3.23)	1.82, .08
Injected mass of EKAP (μg) (M ± SD)	1.13 (0.36)	1.11 (0.39)	0.20, .85
Amygdala VT (M ± SD)	17.62 (4.39)	25.67 (5.98)	23.50, < .001
ACC VT (M ± SD)	13.92 (3.48)	16.88 (3.60)	6.94, .012
Frontal VT (M ± SD)	9.80 (2.36)	11.63 (2.16)	6.54, .015
Hippocampus VT (M ± SD)	7.44 (1.93)	9.57 (1.96)	11.96, .001
Insula VT (M ± SD)	13.86 (3.35)	17.45 (3.30)	11.62, .002

PET images were reconstructed with corrections for attenuation, scatter, randoms, deadtime. For HRRT, reconstruction used the MOLAR algorithm (29) with motion correction applied using Polaris Vicra data. For Vision, reconstruction used the manufacturer’s software (version: VG80C), incorporating time-of-flight information. Attenuation correction was based on a 6-min transmission scan for HRRT and a low-dose CT scan for Vision. Each frame was realigned to 0–10 min summed PET images to correct for motion, and visual inspection was performed for all data. The summed images were then coregistered to each participant’s T1-weighted MR image using a six-parameter mutual information algorithm (FLIRT, FSL 3.2), and the T1-weighted MR image was nonlinearly registered to the MR template (Bioimage Suite). ROIs were defined using the automated anatomical labeling template for SPM2.(41) Using the combined transformations from the template to the individual participant’s PET space, regional time-activity curves (TACs) were generated. TACs were masked using segmentations derived from Computational Anatomy Toolbox (CAT12).

A metabolite-corrected arterial input function was utilized to derive volume of distribution (V_T: the ratio of radioligand concentration in ROI to the concentration in plasma at equilibrium) (42). V_T was selected as the outcome measure given the lack of a suitable reference region for KOR targets in the human brain (29). f_p (plasma free fraction) values ranged from .10-.28 in the study sample, with means of .142 (range .09-.23) in the BPD group and .188 (range .10-.20) in HC. However, f_p calculated using arterial blood was not available for 4 of 20 HC participants. Across all participants imaged to date (n=64) and in keeping with findings from prior studies, f_p did not differ as a function of group (BPD vs. HC; F(1,63)=2.537, p=.116). As such, we elected to use V_T as outcome measure. V_T was estimated at the regional level using multilinear analysis1 (MA1) model applied to TACs (43).

Statistical Analyses

Statistical analyses were completed using SPSS Statistics v29 (IBM Corp.). Correlational analyses, independent samples t-tests, and chi-square analyses were used to examine demographic differences between groups. Groups were demographically matched 1:1 regarding age, sex, and BMI (where possible). Follow-up analyses were conducted to covary for demographic variables showing significant relationships with KOR availability.

Bilateral frontolimbic ROIs relevant to the pathophysiology of BPD (8) were selected for analyses (hippocampus; amygdala; insula; ACC; frontal cortex). Figure S2 displays an ROI map demonstrating that there was no overlap between the ACC and frontal cortex. Multivariate analysis of variance (MANOVA) assessed differences between HCs and individuals with BPD in these regions. Diagnosis was included as the fixed factor, with KOR availability in the five ROIs included as dependent variables. Differences for specific ROIs were examined with follow up ANOVAs. Percent differences were computed to quantify the magnitude of differences in KOR availability between groups. Multivariate (MANCOVA) and univariate (ANCOVA) analyses of covariance were used for follow-up analyses controlling for demographic characteristics.

Pearson’s bivariate correlations examined relationships between KOR availability and clinical correlates of BPD. To provide a comprehensive and dimensional examination of these relationships, correlations were conducted first across the combination of the 20 BPD subjects and the 8 HCs for whom clinical measures were available. This combined group will be referred to as the full sample when describing correlational results. Data from clinical measures (e.g., DERS; STAI) were available for the subset of HCs recruited as part of this study (n=8) but not for HCs analyzed from the data repository. False discovery rate (FDR) corrections for multiple comparisons were applied in planned (a-priori) correlational and between-group analyses. Follow-up exploratory correlational analyses were also conducted within each group separately; specifically, within the BPD group, and among the HCs with available clinical measures.

Results

Demographic Differences

Preliminary analyses demonstrated no significant differences in age (t(38)=0.124, p=.902, M_diff=0.35, SE=2.82, 95% CI[−5.35, 6.05]), sex (X² = 0.00, p>.99), or BMI (t(38)=1.69, p=.100, M_diff=2.58, SE=1.5311, 95% CI[−0.52, 5.68]) between individuals with BPD and HCs. No significant differences were observed in dose (t(32.94)=−1.82, p=.078, M_diff=−2.38, SE=1.31, 95% CI[−5.05, 0.28]) or injected mass of [¹¹C]EKAP (t(38)=0.20, p=.846, M_diff =.023, SE=.12, 95% CI[−0.22, 0.26]). Psychiatric medication status and smoking status differed significantly between groups, with more frequent medication usage (e.g., SSRIs; [X²=7.92, p=.005]) and smoking (X²=5.17, p=.017) among individuals with BPD relative to HCs.

Across the combined sample, significantly lower KOR availability was observed in males relative to females in the insula t(38)=−2.12, p=.040, M_diff=−2.46, SE=0.16, 95% CI[−4.81, −0.11]), ACC t(38)=−2.49, p=.017, M_diff =−2.87, SE=1.15, 95% CI[−5.21, −0.54]), and frontal cortex (t(38)=−2.71, p=.010, M_diff=−1.96, SE=0.72, 95% CI[−3.43, −0.49]). KOR was negatively correlated with age in the ACC (r(40)=−.33, p=.036, 95% CI[.02, .58]), insula (r(40)=−.32, p=.046, 95% CI[.01, .57]), and frontal ROIs (r(40)=−.34, p=.033, 95% CI[.03, .59]). KOR availability as negatively correlated with BMI in the amygdala (r(40)=−.42, p=.007, 95% CI[−.65, −.12]), ACC (r(40)=−.51, p<.001, 95% CI[−.71, −.23]), frontal cortex (r(40)=−.56, p<.001, 95% CI[−.74, −.30]), insula (r(40)=−.49, p=.001, 95% CI[−.70, −.22]), and hippocampus (r(40)=−.51, p<.001, 95% CI[−.71, −.23]).

Differences in KOR Availability Between Groups

Results demonstrated a significant overall effect of diagnostic group on KOR availability (F(5, 34)=6.94, p<.001, np²=0.51). Significantly lower KOR availability was observed in the BPD group relative to HCs in the amygdala (F(1, 38)=23.50, p<.001, d=1.53; M_diff =−8.04; SE=1.66; 95% CI[11.40; −4.69]), hippocampus (F(1, 38)=11.96, p=.001, d=1.09; M_diff=−2.12, SE=0.61; 95% CI [−3.37; −0.88]), ACC (F(1, 38)=6.94, p=.012, d=0.83; M_diff=−2.95, SE=1.12; 95% CI [−5.22; −0.68]), insula (F(1, 38)=11.62, p=.002, d=1.08; M_diff=−3.58, SE=1.05; .95% CI[−5.71; −1.46]), and frontal cortex (F(1, 38)=6.54, p=.015, d=0.81; M_diff= −1.83, SE=0.72; 95% CI[−3.28; −0.38]) (see Table 2, Figures 1A–B). All findings survived FDR corrections. Follow-up MANCOVAs remained significant when controlling for sex, age, medication, BMI, and smoking status (Supplemental Table 1). Exploratory analyses stratified participants with BPD by presence of comorbid for posttraumatic stress disorder (PTSD). Findings indicated no significant differences in KOR availability between BPD alone and those with BPD and PTSD (all p’s> 05).

Figure 1.

(A) MANOVA demonstrating significantly lower KOR availability (V_T) in participants with BPD (purple) relative to HCs (gray). Amy = amygdala; Hipp = Hippocampus; ACC = anterior cingulate cortex. (B) Averaged parametric image showing KOR availability (V_T) in HC (top) relative to individuals with BPD (bottom).

Clinical Correlations

Table 3 displays all uncorrected correlations, including 95% confidence intervals. Uncorrected correlations are reported below. FDR corrections were applied across all 70 planned correlational analyses. Correlations that survived FDR corrections are noted below and flagged in Table 3. For all analyses, correlations were conducted both across the full sample and within each of the groups (BPD and HC).

Table 3.

Correlations between KOR Availability and Clinical Measures in the Full Sample

		DERS A	DERS C	DERS G	DERS I	DERS N	DERS S	DERS T	BDI	BIS	BSL	STAI T	STAI S	PCL	CTQ
Amy	r	−.40 *	−.63 ** +	−.19	−.12	−.34	−.25	−.43 *	−.43 *	−.47 *	−.43 *	−.37	−.51 **	−.44	−.18
	95% CI	[−.67, −.03]	[−.81, −.34]	[−.53, −.19]	[−.47, .27]	[−64, .04]	[−.57, .14]	[−69, −.07]	[−.70, −.07]	[−.72, −.11]	[−.69, −.07]	[−.65, .01]	[−.74, −.17]	[−.79, .12]	[−.55, .25]
ACC	r	−.36	−.51 **	−.16	−.08	−.28	−.20	−.35	−.32	−.34	−.31	−.26	−.38 *	−.38	−.03
	95% CI	[−.65, .01]	[−.74, −17]	[−.50, .23]	[−.44, .31]	[−.59, .11]	[−.54, .18]	[−.64, .03]	[−.62, .06]	[−.63, .04]	[−.61, 07]	[−.57, .13]	[−.66, −.01]	[−.76, .18]	[.39, .44]
FC	r	−.39 *	−.56 ** +	−.12	.03	−.23	−.14	−.30	−.28	−.29	−.26	−.22	−.38 *	−.34	.02
	95% CI	[−.66, −.01]	[−.77, −.24]	[−.47, .27]	[−.35, .40]	[−.56, .16]	[−.49, .24]	[−.60, .09]	[−.59, .10]	[−.60, .01]	−[.58, .12]	[−.55, .17]	[−.66, −.01]	[−.74, 24]	[39, .43]
Hipp	r	−.39 *	−.50 **	−.16	−.12	−.32	−.30	−.40 *	−.36	−.34	−.34	−.29	−.40 *	−.32	−.02
	95% CI	[−.67, −.02]	[−.74. −.15]	[−.50, .23]	[−.47, .27]	[−.62, .06]	[−.60, .09]	[−.67. −.03]	[−.65, .01]	[−.63, .04]	[−.63, .04]	[−.60, .09]	[−.67, −.03]	[−.72, .26]	[−.43, .39]
Insula	r	−.41 *	−.60 ** +	−.18	−.06	−.26	−.23	−.38 *	−.40 *	−.41 *	−.36	−.35	−.47 *	−.38	−.06
	95% CI	[−.68, −.05]	[−.79, −.29]	[−.52, .21]	[−.42, .32]	[−.58, .12]	[−.56, .16]	[−.66, −.01]	[−.67, −.03]	[−.68, −.04]	[−.65, .02]	[−.64, .03]	[−.72, −.12]	[−.76, .19]	[−.46, .37]

Note: DERS = Difficulties in Emotion Regulation Scale; A = Awareness; C = Clarity; G = Goals; I = Impulse Control; N = Nonacceptance; S = Strategies; T = Total. BDI = Beck Depression Inventory; BIS = Barratt Impulsiveness Scale; BSL = Borderline Symptom List; STAI = State Trait Anxiety Inventory; T = Trait; S = State; PCL = PTSD Checklist; CTQ = Child Trauma Questionnaire.

^*= p < .05

^**= p < .01.

Amy = amygdala; FC = frontal cortex; Hipp = hippocampus; ACC = anterior cingulate cortex.

⁺indicates correlations that survived FDR corrections across all 70 tests. Correlations include participants with BPD and HC with clinical data (n=8).

Emotion Dysregulation and Impulsivity

In the full sample, impulsivity was negatively correlated with KOR in the amygdala and insula (r(28)=−.41–45 p’s=.031–013; Figure 2D). Overall emotion dysregulation was negatively correlated with KOR in the hippocampus, insula, and amygdala (r(28)=−.38-.43, p’s=.047-.023). Specific emotion dysregulation subscales also showed relationships with KOR availability. Emotional clarity was negatively correlated with KOR in all ROIs (r(28)=−.50-.63, p’s=.007-<.001; Figure 2A–C), with associations in the amygdala, frontal cortex, and insula surviving FDR corrections. Emotional awareness was negatively correlated with all ROIs except for the ACC (r(28)=−.39-.41, p’s=.043-.029).

Figure 2.

Correlations showing negative associations between emotional clarity and KOR availability (V_T) in the amygdala (A), insula (B), and frontal ROIs (C), and impulsivity and KOR availability in the amygdala (D). Participants with BPD are represented by purple data points; HCs (participants recruited for this study, n = 8) with grey dots. Amy = amygdala.

In the BPD group, emotional awareness was negatively correlated with KOR availability in the amygdala (r(20)=−.58, p=.007, 95% CI[ −.82, −.19]), frontal cortex (r(20)=−.47, p=.038, 95% CI[ −.75, −.03]), hippocampus (r(20)=−.49, p=.028, 95% CI[ −.77, −.06]), and insula (r(20)=−.50, p=.024, 95% CI[ −.77, −.08]). Emotional clarity was negatively correlated with KOR availability in the amygdala (r(20)=−.57, p=.008, 95% CI[ −.81, −.17]), ACC (r(20)=−.51, p=.021, 95% CI[ −.78, −.09]), frontal cortex (r(20)=−.59, p=.007, 95% CI[ −.82, −.20]), and insula (r(20)=−.55, p=.012, 95% CI[ −.80, −.14]). Correlations were not significant when examining the HC group alone. Within group correlations are displayed in Tables S2–3.

Depressive and Anxiety Symptoms

In the full sample, state anxiety was negatively correlated with KOR availability across all ROIs (r(28)=−.38-.51, p’s=.049-.006; Figure 3A–B). Depressive symptoms were also examined across the full sample (amygdala and insula (r(28)=−.40-.43, p’s=.037-.021; Figure S3). However, limited variability in depressive symptoms was observed among HCs. When examining each group separately, correlations did not reach statistical significance in either the BPD group (BDI: r(20)= −.01-.19, p’s=.98-.42; STAI-S: r(20)= −.20–24, p’s=.40-.31) or in the in the HC group (BDI: r(8)= −.39-.50, p’s=.35–21; STAI-S: r(20)= −.51–62, p’s=.19-.10). These correlations are presented in Table S2–3.

Figure 3.

Correlations showing negative associations between KOR availability (V_T) and anxiety symptom severity in the amygdala (A) and insula (B). Participants with BPD are represented by purple data points and HCs (participants recruited for this study, n = 8) with grey dots. Amy = amygdala.

BPD Symptoms

Average BPD symptom severity (BSL scores) were examined in the full sample, and were negatively correlated with KOR availability in the amygdala (r(28)=−.43, p=.023; Figure S3). However, limited variability in BSL scores were observed among the HCs. When examining within-group correlations, the directions of the relationships remained negative but did not reach statistical significance in either the BPD (r(20)=−.11-.21, p’s=.64-.39) or HC (r(8)=−.45–63, p’s=.26-.09) groups. These correlations are presented in Table S2–3. Exploratory analyses with specific BPD symptoms (using the full sample) are presented in Table S4.

Discussion

To our knowledge, this is the first study to directly examine KOR availability in BPD and its relationship to clinical symptom presentation. Consistent with hypotheses, KOR availability was significantly lower in individuals with BPD relative to demographically matched psychiatrically healthy adults in a priori frontal and limbic ROIs. Moreover, large effect sizes were observed, with 17–37% lower KOR availability in those with BPD relative to HCs (d’s=0.83–1.53). Notably, these differences survived multiple comparisons correction, indicating widespread differences in KOR functioning in brain areas densely concentrated with opioid receptors (7). In accordance with hypotheses, KOR availability was negatively correlated with emotion dysregulation and impulsivity (e.g., emotional awareness and clarity). Notably, emotion dysregulation was linked to KOR availability across the full sample, and within the BPD group specifically. State levels of anxiety were also negatively correlated with KOR availability across the full sample. While depressive and BPD symptoms were linked to KOR availability across all participants, reduced variability among those without BPD limits interpretations of these findings. When the groups were examined separately, correlations did not reach statistical significance for either group.

The observed differences in KOR availability between individuals with BPD and HCs is consistent with past findings suggesting that life stressors/trauma activate the dynorphin/KOR system, introducing a diathesis or vulnerability for negative health outcomes. Indeed, trauma exposure is thought to precipitate the development of BPD and is documented in up to 90% of BPD cases (27). Increased dynorphin release has been shown in response to chronic social defeat stress exposure in rodents (44); this effect could result in increased occupancy of KOR, yielding fewer receptors available for the radiotracer to bind (28) and lower KOR availability. Accordingly, in a pilot PET study with a first generation KOR agonist radiotracer, individuals with the greatest laboratory stress responses had the lowest KOR availability (45). Alternatively, dynorphin/KOR activation may lead to downregulation of KOR over time, brought about by receptor internalization, which has been epigenetically linked to glucocorticoid regulation of KOR expression in postmortem work (46).

In the full sample, we observed moderate to large negative associations between KOR availability and state anxiety levels, as well as cardinal features of BPD: emotion dysregulation, and impulsivity. Further, when examining the BPD group separately, KOR availability was robustly and negatively correlated with emotional unawareness and lack of emotional clarity. These findings are consistent with past work implicating dynorphin/KOR in emotional dysregulation (17, 23) and development of mood, cognitive, and dissociative symptoms (24) and further suggest the relevance of KOR functioning to BPD pathophysiology. Notably, observed relationships with clinical symptoms were most robust in limbic regions including amygdala, implicated in emotional processing and social cognition (47). The roles of KOR and dynorphin in regulating amygdalar circuity, including inhibition of dopamine and glutamate release in downstream projection regions (48), may help to explain the strength of these relationships.

Prior in vivo human studies have found evidence for correlations between KOR availability and depression severity independent components analysis (ICA)-derived regions that encompassed portions of the frontal lobe and insula (49). Endogenous dynorphin release/activation has also been linked to stress-induced addictive behaviors in humans (50). Recent clinical findings in individuals with schizophrenia demonstrated positive associations between KOR availability and anhedonia in cortical brain regions (51). Anhedonia was measured in the context of negative symptoms of schizophrenia. Conversely, in the present study, anhedonia was captured on the BDI, but was not assessed as a separate construct. Further, while anhedonia is a transdiagnostic symptom, differences in its presentation have been observed in depression relative to schizophrenia (52); these factors may have contributed to the discrepant findings.

Contrary to our hypotheses most clinical associations were present in the full sample but not when examining the BPD subsample. As documented in preclinical literature, KOR activation mediates stress-induced development of depressive symptomatology (21), specifically anhedonia and dysphoria, through stress’ induction of dynorphin release (19). Additionally, pretreating rodents with the selective KOR antagonist norBNI before repetitive blast trauma can block resultant aversive and anxiety-like behavioral dysfunction (53). Thus, these non-significant findings were unexpected. Restricted range of responses in the BPD group may have contributed to the absence of significant findings; the added variance in clinical symptom scales by including the HC group may have provided additional power to detect effects. Future larger studies should formally test for the presence of threshold effects.

Similarly, it was unexpected that no correlations were observed with PTSD symptomatology or childhood trauma. However, this lack of effect replicates prior null findings between KOR binding and childhood trauma in individuals with MDD (45). It may be that childhood trauma introduces a vulnerability for development of BPD, but the distal nature of such trauma precludes our ability to detect an associative effect. As early life trauma is strikingly common in BPD (27), it may be difficult to separate the neurobiological contributions of early life trauma from those attributable to the pathophysiology of the disorder. Further work examining individuals with BPD with and without early life trauma will be necessary to further elucidate relationships between early life stressors and KOR availability. Moreover, while the PTSD Checklist (PCL) assesses the severity of PTSD symptoms, it does not stratify responses by the specific nature of the trauma. As demonstrated in previous literature (54), trauma type can affect links between trauma exposure and biological processes. Future work should consider the impact of distinct trauma types on KOR availability.

The current study has several limitations. The sample size (N = 40 overall; 20 per group) was adequately powered for planned analyses; however, future studies should utilize larger samples. Two scanners were used in this study. Chi square analyses revealed that groups were well balanced across scanners, and no differences in VT were observed as a function of scanner. Thus, it is unlikely that scanner type affected the results. However, we were unable to perform within-subjects comparisons of data across scanners, representing a methodological limitation. Despite matching, we observed effects of sex and age on KOR availability as previously reported (29). Further, the BPD group included six individuals with comorbid PTSD; the phenotypic differences in these disorders may be differentially reflected in the dynorphin/KOR system and future, larger studies should investigate this possibility. Finally, longitudinal studies tracking onset and progression in BPD with repeated PET imaging with [¹¹C]EKAP would be of immense value to determine causality of the dynorphin/KOR system’s role in BPD to aid in the development of novel treatment targets. Future studies may also expand beyond frontolimbic brain regions examined in this initial analysis.

In conclusion, this well-powered PET study provides preliminary evidence for alterations in the KOR system in individuals with BPD relative to demographically matched HCs. Lower KOR availability was observed across frontolimbic ROIs associated with the pathophysiology of the disorder, and strongly associated with clinical correlates of BPD. These findings have promising implications for BPD treatment development, specifically surrounding KOR antagonists. Evidence of lower KOR availability in BPD builds upon past work demonstrating KOR antagonists’ ability to act as antidepressants and anxiolytic agents (17), and upon findings showing relationships between opioid system modulation and reduction in self-injurious behaviors (13). Given the centrality of these clinical features to BPD, trials exploring KOR antagonists in this population may yield significant clinical benefits. In summary, these results highlight the relevance of KOR in BPD pathophysiology and suggest the opioid system may serve as a promising target for treatment development in this debilitating and severely undertreated disorder.

Supplementary Material

Figures S1–S3, Tables S1–S4

KEY RESOURCES TABLE.

Resource Type	Specific Reagent or Resource	Source or Reference	Identifiers	Additional Information
Add additional rows as needed for each resource type	Include species and sex when applicable.	Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.	Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.	Include any additional information or notes if necessary.
Deposited Data; Public Database	4835; 3771	NIMH Data Archive	SCR_004434
Software; Algorithm	Bioimage Suite	Neuroimageing Tools & Resources Collaboratory	SCR_002986
Software; Algorithm	FMRIB's Linear Image Registration Tool	Analysis Group, FMRIB, Oxford, UK	SCR_024922	https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FLIRT
Software; Algorithm	SPSS V. 29.0.2	IBM SPSS Statistics	SCR_016479
Other	Biograph Vision PET/CT	Siemens	N/A
Other	High-resolution research tomograph	Siemens	N/A
Other	3-Tesla Siemens Prisma MR scanner	Siemens	N/A
Radiotracer	11C-METHYL-(R)-4-(2-(3,4-DICHLOROPHENYL)ACETYL)-3-((DIETHYLAMINO)METHYL)PIPERAZINE-1-CARBOXYLATE	Radiochemistry Center, Yale PET Center	PubChem CID: 74539703

Data availability:

Subsets of data used in this manuscript are available through NIMH Data Archive (#4835; #3771).