An in vivo examination of the relationship between metabotropic glutamate receptor 5 and suicide attempts in people with borderline personality disorder

Abstract

Background:

Borderline Personality Disorder (BPD) is a serious psychiatric condition, associated with a high risk for suicide attempts and death by suicide. However, relatively little is known about the pathophysiology of BPD. The metabotropic glutamate receptor type 5 (mGlu5) has been specifically implicated in the pathophysiology of BPD and suicide attempts, with more general roles in emotion regulation, social and cognitive functioning, and pain processing. Here, we examined the relationship between mGlu5 availability, BPD, and suicide attempts in vivo for the first time.

Methods:

Eighteen individuals with BPD, and 18 age-, sex-, and smoking-status matched healthy (HC) and 18 clinical comparison controls with major depressive disorder (MDD) completed comprehensive clinical assessments and participated in an [¹⁸F]FPEB positron emission tomography (PET) scan to measure mGlu5 availability. Volume of distribution (V_T) in the frontolimbic circuit implicated in BPD pathophysiology was the PET outcome measure.

Results:

We observed significantly higher frontolimbic mGlu5 availability in BPD compared to both HC (p=.009, d=0.84, 18.43% difference), and MDD (p=.03, d=0.69, 15.21% difference). In the BPD, but not MDD group, higher mGlu5 availability was also associated with history of suicide attempts (SA; 19–25% higher, p’s=.005-.02). Further, mGlu5 availability was positively correlated with risk factors for suicide (e.g., sexual victimization, perceived burdensomeness) in BPD-SA group.

Conclusions:

Results show higher mGlu5 availability in BPD and suicide attempt for the first time. Our preliminary findings suggest mGlu5 may be a critical treatment target for BPD symptoms, including suicide attempts, and warrant further investigation in larger samples.

Introduction

Borderline Personality Disorder (BPD) is a debilitating psychiatric condition and significant source of both personal and public health burden(1, 2) associated with up to a 75% lifetime suicide attempt (SA) rate and 10% suicide mortality.(3) In addition to suicide attempts, BPD is characterized by extreme distress and emotion dysregulation, impulsivity, aggression, executive dysfunction, altered pain perception, and self-injury. Additionally, BPD presents a significant burden on interpersonal relationships. Despite the seriousness of the disorder, neurobiological research focused on BPD remains limited,(4) and currently available pharmacological agents do not have meaningful impacts on overall symptom severity.(5, 6) Although no pharmacological agents are currently FDA approved to treat BPD, use of a polypharmacy approach is unfortunately common (7, 8) leading to harmful side-effects and decreased treatment adherence. Novel targeted treatment development for BPD and suicide risk is urgently needed.

Glutamate, the primary excitatory neurotransmitter in the brain, has recently been implicated in the pathophysiology of suicide,(9) and interest has increased in its role in BPD.(10) Specifically, glutamate plays a key role in synaptic and neuroplasticity, and is associated with cardinal features of BPD including emotion regulation, executive function/impulsivity, and cognition.(10) Magnetic resonance spectroscopy (MRS) studies have demonstrated higher glutamate levels in BPD subjects in the anterior cingulate cortex (ACC) and orbitofrontal cortex (OFC),(11)^,(12, 13) and two studies reported a positive relationship between ACC glutamate and impulsivity in BPD.(12, 13) Postmortem evidence suggests that increased expression of multiple genes that encode for glutamate receptors is associated with death by suicide.(14) Both preclinical(15) and postmortem(16) evidence suggests that chronic stress exposure (thought to precipitate the development of BPD(10)) contributes to glutamate dysregulation (e.g. over-activation, mGlu5 receptor upregulation). This dysregulation can result in decreased neurogenesis in areas including the amygdala and hippocampus,(17) potentially contributing to their dysfunction in BPD.(10) Further, multiple glutamate receptors including N-methyl-D-aspartate (NMDA; an ionotropic glutamate receptor critical to synaptic plasticity(10, 18) and emotional learning (19, 20)) neurotransmission guide the ‘critical period’ for neuronal development in frontolimbic brain regions that are implicated in BPD pathophysiology.(10) Importantly, interventions targeting the glutamate system have been shown to reduce self-injurious behavior in clinical case studies,(21) and medications that influence glutamatergic processes (e.g., ketamine and lithium), have been shown to effectively reduce suicide ideation across clinical populations (22–27), with early evidence supporting similar effects in BPD.(28, 29) Unfortunately, they also influence other neurotransmitter systems and often result in unwanted side-effects.(30, 31)

The specific role of the metabotropic glutamate 5 receptor (mGlu5) in glutamate signaling pathways related to BPD and SA has not been previously tested. However, mGlu5 is an exciting potential target for the development of novel treatments for BPD and SA. Research in multiple fields (postmortem, preclinical, and genetics) substantiates the relationship between mGlu5 and core BPD symptoms. Specifically, mGlu5 has been shown to play an important role in emotion regulation,(32) and is implicated in the development of mood and anxiety symptoms in both human(33, 34) and preclinical studies.(32, 35–37) Additionally, mGlu5 is implicated in pain experience and social functioning; abnormal pain sensitivity(38) and social deficits (e.g., vulnerability to social stress)(37) are both central to BPD pathology, and powerful predictors of suicide attempts.(38–40) For example, preclinical evidence has suggested that together, mGlu5 and Homer 1 (a scaffolding protein linked to mGlu5 function (32)) mediate stress reactivity following negative social interactions,(41) and downregulation of mGlu5 appears to improve deficits in social functioning, such as normalization of social behaviors.(37) Upregulation in amygdala mGlu5 is associated with chronic visceral pain,(42) and preclinical evidence has implicated mGlu5 as a common neurobiological mechanism underlying the relationship between chronic pain and depressed mood.(43)

Evidence for links between mGlu5 and suicide risk also spans multiple fields. Postmortem work has linked the gene that encodes for mGlu5 (GRM5) to suicide mortality,(44) and epigenetic studies have observed a relationship between the gene that encodes for Homer 1 (HOMER1) and SA.(45, 46) Additionally, NDMA is implicated in both the pathophysiology of BPD(10, 47) and suicide.(9, 48–50) Preclinical research suggests that mGlu5 activation mediates NDMA function,(51–53) and a polymorphism in GRIN2B has been associated with both impulsivity(54) and SA.(49) Further, recent evidence suggests that one of the downstream effects of the rapid-acting anti-depressant and anti-suicidal agent ketamine (an NMDA antagonist medication) is downregulation in mGlu5 availability(55–57) likely via a surge in glutamate. This downregulation in mGlu5, which our work suggests is associated with an antidepressant effect(57) may be an important step in suicide-reducing properties. Thus, dysregulation in mGlu5 may affect development and maintenance of BPD symptoms and suicide attempts both directly and through downstream effects. Finally, our in vivo studies have demonstrated elevated mGlu5 availability in post-traumatic stress disorder (PTSD;(34, 58) another form of trauma-related psychopathology), and one study showed that higher mGlu5 was associated with suicide ideation in PTSD.(58) These studies led us to consider mGlu5 as a novel unexplored treatment target in BPD with the potential to address suicide risk.

Here, we examine mGlu5 availability in vivo in BPD and its relationship to suicide attempt history for the first time. Specifically, using the well-validated radiotracer [¹⁸F]FPEB (which binds with high selectivity and specificity to mGlu5) and positron emission tomography (PET),(34, 58) we aim to evaluate three hypotheses. First, we hypothesize that mGlu5 availability will be higher in people with BPD relative to matched groups of healthy controls (HC) and clinical controls with major depressive disorder (MDD). Second, we hypothesize that individuals with a history of SA will have higher mGlu5 availability (e.g., mGlu5 as a trait marker for SA risk), but this effect will be more pronounced among individuals with BPD. Finally, we hypothesize that mGlu5 availability will be associated with clinically relevant variables including impulsivity, trauma history, and interpersonal connectedness among individuals with BPD.

Methods and Materials

Participants

Eighteen individuals with BPD and comorbid MDD (mean ± SD age = 30.00±10.39 years; 5 male, 13 female), 18 individuals with MDD only (without BPD; age = 34.38±12.84 years; 5 male, 13 female), and 18 healthy controls (age = 30.66±9.13 years; 5 male, 13 female) were included in this study. Of note, some participants included in the current MDD and HC samples were recruited for, and therefore overlap with, previous studies by our group (e.g., where we sought unmedicated individuals specifically).(59–61) Participants in these groups were selected out of our large sample of MDD and HC participants such that they were matched to the BPD group for sex, age, and smoking status (given the demonstrated potential influence of these variables on mGlu5 availability(59, 62)). Demographic, clinical, and radiotracer characteristics are provided in Table 1. Ten individuals in the BPD sample reported a history of SA (6 reported multiple attempts), with a mean of 21 months (by self-report estimate) since the most recent attempt. Of the demographically matched MDD sample, only three participants had a history of SA. As such, for the analyses examining SA, an additional four individuals with MDD and an SA history were included, resulting in 22 total participants in the MDD groups for these analyses (7 SA; 15 NSA). Of note, the four individuals included were the only additional participants with an SA history and primary MDD diagnoses; thus, they were not demographically matched to the BPD sample. However, no demographic differences were observed between the MDD and BPD groups (See Table S1). Information on attempt latency was not available in MDD. Comorbid diagnosis of PTSD, pervasive developmental disorder, psychosis, and substance use disorder (current, lifetime if severe) were exclusionary for MDD and BPD groups. Use of one psychotropic medication, including SSRIs, was permitted for medications with no expected direct action on the glutamate system. No HC participants were taking psychiatric medication. BPD and MDD individuals were matched for medication use as much as possible, and analyses were repeated controlling for potential medication effects. Participants underwent physical and neurological examinations to rule out major medical or neurological illness. History of head injury including concussion was assessed. Screening involved electrocardiography, complete blood counts, serum chemistries, thyroid function test, liver function test, urinalysis and urine toxicology screening, and urine pregnancy tests (for females). Psychiatric diagnoses (and therefore eligibility) were confirmed using the Structured Clinical Interview for DSM-5 (SCID-5)(63) and SCID for Personality Disorders (SCID-5-PD).(64) BPD symptoms were assessed using the McLean(65) and Borderline Symptom-List, 23 (BSL-23).(66) Depressive symptoms were assessed using the Montgomery-Asberg Depression Rating Scale (MADRS)(67), the Hamilton Rating Scale for Depression (HAM-D),(68) and the Beck Depression Inventory (BDI).(69) Participants in the BPD group completed self-report measures relevant to BPD and suicide risk (Difficulties in Emotion Regulation Scale (DERS(70)); Acquired Capability for Suicide Scale (ACSS-FAD(71)); Interpersonal Needs Questionnaire (INQ(72)); Painful and Provocative Events Scale (PPES(73)). The study was approved by the Yale University Human Investigation Committee and the Radioactive Drug Research Committee. All participants provided written informed consent.

Table 1.

Demographic and PET parameters for the study population

	HC (n=18)	BPD (n=18)	MDD (n=18)	Statistical Test (F or chi2, p value)
Age, years (mean±SD)	30.66±9.13	30.00±10.39	31.61±10.30	0.930, .401
Sex (%female (F:M))	72.20 (13:5)	72.20 (13:5)	72.20 (13:5)	0,1.00
Tobacco users (% (N))	22.20 (4)	22.20 (4)	22.20 (4)	0, 1.00
BMI, scan day (mean±SD)	27.39±5.78	29.77±6.64	27.08±6.27	0.93, .40
MADRS (mean±SD)	0.67±.86*^	16.50±11.39	22.11±6.44	19.97, <.001
BDI (mean±SD)	1.11±1.69*^	19.66±10.69	21.50±14.96	15.12, <.001
mGlu5 Circuit VT Value (mean±SD)	18.46±3.14*	22.20±5.40^	19.06±3.53*	4.24, .02
Psych medication (%taking medication (N))	NA	38.80 (7)	38.80 (7)	0, 1.00
Injected dose, MBq (mean±SD)	170.24±32.0	164.82±22.06	165.78±21.66	0.46, .51
Injected mass, μg/kg (mean±SD)	0.37±0.17	0.44±0.62	0.39±0.38	0.60, .56

Note: HC = healthy control; BPD = borderline personality disorder; MDD = major depressive disorder. MADRS = Montgomery-Asberg Depression Rating Scale; BDI = Beck Depression Inventory.

^*= sig. differences from BPD;

^{^}= sig. differences from MDD.

MRI and PET scanning

T1-weighted MRI scans were acquired on a 3T scanner (Prisma; Siemens Medical Systems) to exclude structural abnormality and for co-registration with PET. [¹⁸F]FPEB was synthesized onsite as described previously.(74) [¹⁸F]FPEB was administered by intravenous bolus plus infusion for 120 min, with a K_bol of 190 min.(74) (75) Based on previous work assessing equilibrium, emission data were acquired 90–120 min after the start of injection on the high-resolution research tomograph (Siemens/CTI), which has intrinsic spatial resolution of ~3 mm FWHM. We have previously shown that venous and arterial concentrations are well matched at equilibrium, allowing venous sampling to be collected for metabolite correction as opposed to the more invasive arterial sampling.(74) Our group has previously showed good test–retest reliability (mean TRV of 12%) of [¹⁸F]FPEB binding using this approach.(74) Venous samples were acquired at 15, 20, 25, 30, 40, 50, 60, 75, 90, 105, and 120 min post-injection for calculation of a metabolite-corrected venous input function as validated previously.(74) A 6-min transmission scan was obtained for attenuation correction. Head motion was tracked using the Polaris Vicra optical tracking system (Vicra; NDI System). Injected dose and mass did not differ across groups (Table 1). Dynamic scan data were reconstructed with corrections for attenuation, normalization, randoms, scatter, dead time, and motion using the ordered subset expectation maximization-based MOLAR algorithm.(76, 77) Additionally, we evaluated all imaging data (PET and MR) to ensure quality (e.g., motion effects, image distortion, abnormalities in time activity curves). No subjects were excluded from analyses due to quality issues.

Image analysis

PET images were co-registered to each participant’s T1-weighted MRI images using a six-parameter mutual information algorithm (FLIRT; FSL 3.2, Analysis Group, FMRIB), which was then co-registered to the magnetic resonance template by nonlinear transformation using the Bioimagesuite software (version 2.5). Regions of interest (ROIs) were delineated using the Anatomical Automatic Labeling (AAL) atlas. The primary ROIs were regions of the prefrontal cortex (PFC; dorsolateral PFC (dlPFC), ventromedial PFC (vmPFC), OFC), limbic regions (amygdala, insula), and the ACC, selected for their relevance to the pathophysiology of both BPD and MDD. Gray matter segmentation was conducted using FSL-FAST, and a gray matter mask was applied to the ROIs to minimize partial volume effects. Given the lack of an appropriate reference region in the human brain for mGlu5 targets,(78) volume of distribution (V_T; equilibrium ratio of parent radioligand concentration in tissue relative to that in blood) was used as the outcome measure. V_T was estimated by the equilibrium analysis method described previously(75) and a venous plasma input function.(74)

Statistics

One-way Analyses of Variance (ANOVAs) and chi-square analyses (as appropriate for categorical variables) were used to assess differences between demographic and radiotracer characteristics across groups. Group differences in mGlu5 availability ([¹⁸F]FPEB V_T) were first assessed using a One-way ANOVA with frontolimbic mGlu5 (averaged V_T from the OFC, dlPFC, vmPFC, ACC, amygdala, and insula) as the dependent variable, and diagnosis as the independent variable (Hypothesis 1). Sphericity of the data was checked by Mauchly’s test and corrected by the Greenhouse – Geisser procedure if violated. Homogeneity of variances was checked by Levene’s test (no violations were detected). The relationship between diagnosis, SA history (analyzed as a binary variable), and mGlu5 availability was assessed with a second ANOVA with frontolimbic mGlu5 (averaged V_T from the OFC, dlPFC, vmPFC, ACC, amygdala, and insula) as the dependent variable, and diagnosis, SA history, and their interaction as independent variables (Hypothesis 2). In both cases, post-hoc pairwise comparisons were conducted using False Discovery Rate (FDR) to correct for multiple comparisons. Percent differences were calculated to determine extent and magnitude of V_T differences across groups. Finally, to evaluate associations between mGlu5 availability and clinical and demographic variables (Hypothesis 3) we used correlational analyses (Pearson’s r), following evaluation of normality with Shapiro–Wilk tests. Findings were considered significant at the p < 0.05 level. Cohen’s d and percent differences (for post-hoc t tests) and Pearson’s r (correlational analyses) are included as measures of effect size. Presence of outliers was evaluated using a standard of 3*standard deviation for univariate and Mahalanobis distance (p<.05 relative to the relevant chi distribution) for multivariate. No outliers were detected in any presented analyses. For planned analyses, FDR was used to correct for multiple comparisons.

Results

Between group differences in mGlu5 availability

The ANOVA evaluating group differences in frontolimbic mGlu5 availability ([¹⁸F]FPEB V_T) was significant: F_2,51=4.24, p=.02. Post-hoc tests indicated significantly higher mGlu5 availability in BPD relative to both HC (p=.009, d=0.84, 18.43% difference) and MDD (p=.03, d=0.69, 15.21% difference) (See Figure 1, Panel A and Figure 2). Results did not change when controlling for age (F_2,50=4.29, p=.02) or medication status (F_2,50=4.02, p=.02). Further, no relationship between study variables of interest and psychiatric medication use (analyzed as a binary variable; i.e., medicated vs. not medicated) were observed.

Figure 1. Mean frontolimbic mGlu5.

Panel A: Mean mGlu5 availability ([¹⁸F]FPEB V_T) was higher in BPD compared to both HC and MDD. Panel B: Mean mGlu5 availability was higher in people in the BPD group who had past suicide attempts (SA) compared to both people in the BPD group without past SA (labelled BPD (NSA)) and people in the MDD group with (MDD-SA) and without past SA (MDD-NSA). Data are also shown for healthy controls to facilitate visual comparison. However, HC were not included in this statistical analysis. P-values and % differences for pairwise comparisons are marked. HC=Healthy Control; MDD=major depressive disorder; BPD=borderline personality disorder; V_T=volume of distribution, averaged mGlu5 availability.

Figure 2.

Parametric images from representative participants (i.e., participants with V_T values closest to their respective group means) in each group. HC=Healthy Control; MDD=major depressive disorder; BPD=Borderline Personality Disorder; V_T=volume of distribution, mGlu5 availability.

Group x SA differences in mGlu5 availability

To evaluate differences in mGlu5 availability as a function of SA history (past suicide attempt vs. no past attempt), we conducted a second ANOVA. The omnibus F-test was significant (F_3,32=4.57, p=.008), as was the main effect of SA (F_1,32=6.65, p=.01). With the addition of SA, the main effect of diagnostic group was no longer significant (p=.10), while the interaction group*SA approached significance (p=.07). Post-hoc tests showed that individuals with BPD and a history of SA (BPD-SA) had higher mGlu5 availability than individuals with BPD with no history of SA (NSA; p=.005, d=2.15, 25.42% difference), or individuals with MDD with (p=.02, d=1.69, 19.69% difference) and without (p=.009, d=2.53, 24.26% difference) a history of SA (See Figure 1, Panel B). Neither recency of SA nor number of SAs was significantly related to frontolimbic mGlu5 availability in BPD (this information was not available in MDD).

Associations with clinical variables

Data on BPD specific measures (e.g., BSL-23, DERS) were not collected for MDD subjects. The following analyses were conducted only within the BPD sample. Consistent with hypotheses, we observed significant positive correlations between frontolimbic mGlu5 availability and trauma history in BPD, including history of sexual victimization (frequency of abuse; r(11)=.77, p=.006; assessed by the PPES). Significant correlations were observed between interpersonal connectedness and mGlu5 availability, including perceived burdensomeness, specifically that others would be “happier” without them (r(11)=.75, p=.008), and wished they could “be rid” of them (r(11)=.62, p=.04; assessed by the INQ)). With the exception of the last noted association, all correlations survived corrections for multiple comparisons. Our hypothesis concerning impulsivity was not sustained; there was no significant association between mGlu5 availability and impulsivity (BIS scores) in BPD, (r(18)=.08, p=.77). Exploratory analyses assessing additional clinically relevant variables demonstrated significant correlations between mGlu5 availability and self-criticism on the BDI (r(18)=.49, p=.04; see Figure 3), and specific BPD symptoms including feeling “far away” from oneself (r(14)=.57, p=.03; see Figure 3) and fears of losing control (r(14)=.68, p=.008; assessed by the BSL). No significant relationships were observed between frontolimbic mGlu5 availability and emotion dysregulation (DERS total: r(14)=.18, p=.51), overall MDD (MADRS total: r(18)=.03, p=.89) or BPD symptom severity (BSL total: r(14)=.31, p=.28).

Figure 3.

Exemplar clinical correlations within the BPD (borderline personality disorder) group sample as assessed by the BSL (Panel A) and BDI (Panel B). BPD=Borderline Personality Disorder; VT=volume of distribution, averaged mGlu5 availability.

Finally, exploratory analyses were performed to examine the relationship between clinical variables and frontolimbic mGlu5 availability in individuals reporting a history of SA cross-diagnostically (MDD and BPD groups, n=17). In individuals with a SA history, only intensity/frequency of derealization reported on the HAM-D was positively correlated with mGlu5 availability (r(14)=.67, p=.008). See Figure S1 for all significant correlations and Table S2 for descriptive statistics for measures included in correlational analyses.

Discussion

To our knowledge, this is the first study to investigate in vivo mGlu5 availability in BPD, and the first to report the relationship between mGlu5 and SA in any population. Consistent with study hypotheses, we observed significantly higher mGlu5 availability in BPD overall relative to both MDD only and HC. Further, and both consistent with hypotheses and in keeping with our previous work on suicidal ideation,(58) significantly higher mGlu5 was associated with suicide attempt history. More specifically, individuals in the BPD-SA subgroup had significantly higher mGlu5 availability with notably robust effect sizes (d’s >1.0). Finally, examination of relationships between clinical variables and frontolimbic mGlu5 availability in individuals with BPD revealed strong positive correlations with history of sexual victimization and perceiving oneself as a burden: powerful risk factors for suicide attempts and mortality in this population. However, the hypothesized relationship between mGlu5 and impulsivity in BPD was not observed.

Both BPD and suicide are clinically complex issues, presenting in heterogenous populations and requiring highly individualized treatment approaches. Beginning to uncover neurobiological underpinnings common to these issues can ultimately advance development of such treatments. These findings warrant replication and extension; however, their robust nature holds promise, and possible explanations should be considered. Dysregulation of the glutamatergic system has been directly observed in BPD,(12, 13, 79) and glutamatergic agents have been associated with reduced self-injury.(21) Similarly, recent postmortem studies have supported a relationship between glutamate dysregulation,(80, 81) including altered mGlu5,(44) and death by suicide. Our previous work has suggested the possibility that upregulated mGlu5 may result from abnormal glucocorticoid signaling in both PTSD- another form of trauma related psychopathology- and nicotine use disorders, leading to higher mGlu5 availability when the two conditions present together.(34, 59) Though less consistent, similar findings concerning HPA axis abnormality have been observed in BPD, and in relation to suicide risk.(82–86) More work is needed, including work directly investigating mGlu5 availability in individuals with suicide attempts (recent and past) in PTSD and other psychiatric populations so that the nature and specificity of the observed relationships can be better understood.

Importantly, we also observed significant relationships between mGlu5 and several clinically meaningful indices of suicide risk in BPD. Specifically, we observed positive correlations with large effect sizes between mGlu5 availability and history of sexual trauma and perceived burdensomeness. These findings are particularly notable given mounting empirical evidence supporting the interpersonal psychological theory of suicide, which proposes that perceived burdensomeness, thwarted belongingness, and acquired capability are proximal risk factors for suicide attempts, and in combination, for death by suicide.(87) Observation of positive associations between variables reflecting these constructs (such as the belief that others would be “happier” and wish to be “rid of” them) and mGlu5 availability adds weight to the overall findings. However, given the complexities of both BPD and suicide, multifaceted, individualized treatment approaches are likely warranted for these populations.

Of note, not all expected relationships were observed. MGlu5 availability was not significantly associated with impulsivity, nor with depressive or BPD severity. Mood state variability on scan day may have affected reporting on symptom measures.(88) Regarding impulsivity, future studies should include more in depth explorations of the relationships between mGlu5 availability and impulsivity in this population, for instance, by using a multidimensional measure of impulsivity (e.g., the UPPS(89)). Importantly, no significant relationship was observed between mGlu5 and recency of SA, suggesting mGlu5 may serve as a trait and not a state marker of SA in BPD, predicting risk overall but not necessarily differentiating acute risk. Consistent with previous work across several studies,(58, 90) there were no absolute differences in mGlu5 availability in MDD relative to HC group. As noted previously, the possibility that mGlu5 could function as a state marker of distress in MDD, potentially altering acutely in reaction to anxiety or stress exposure, warrants consideration. Another possibility is that the heterogeneity of MDD is relevant in this case; mGlu5 may play a different role in MDD subtypes (e.g. melancholic, anxious, atypical).

Several limitations should be considered when interpreting results. First, although our sample was large for a first PET study, it was limited in diversity and size, especially with regard to the BPD-SA group. This potentially affected statistical power. Thus, these promising findings should be regarded as preliminary. Future studies with larger, more diverse samples should aim to confirm and extend findings. Second, we observed limited variability in suicide ideation (endorsed on scan day by 16/18 individuals in the BPD group). Thus, we were unable to replicate our previous finding showing higher mGlu5 availability in individuals with PTSD reporting suicide ideation. Future work should consider a more transdiagnostic approach, including individuals below diagnostic threshold with BPD traits and a history of suicide attempts, or with comorbid PTSD and BPD. Considering recent changes to the diagnosis of personality pathology with the ICD-11, and movement toward a more dimensional conceptualization of personality disorder diagnosis broadly (though the ‘borderline pattern’ was highlighted and distinctly codified even in the ICD-11 dimensional system(91)), a move toward cross-diagnostic study may be particularly advantageous. Based on the findings in this study and others, the idea that borderline features, and their associated neurobiology, hold cross-diagnostic predicative power for suicide risk may warrant consideration. Finally, across clinical groups, just under 40 percent reported psychotropic medication use. Medicated individuals were included in the interest of ecological validity. Though this could have potentially impacted our results, analyses controlling for medication did not yield different results.

In conclusion, this study represents the first in vivo examination of frontolimbic mGlu5 availability in BPD, as well as the first direct investigation of the relationship between mGlu5 and SA. Findings support the idea that mGlu5 may represent a promising trait marker and treatment target for BPD, specifically for people with history of suicide attempt. Our results implicate mGlu5 marker of SA risk in BPD, an assertion further bolstered by strong positive correlations between mGlu5 and key risk factors for suicide attempts both in the full BPD sample, and in participants with a SA history. We hope these results can inform the development of mGlu5 targeted treatment for BPD and suicide behavior, which to our knowledge has not yet been trialed. Specifically, results suggest that antagonism or negative allosteric modulation of mGlu5 could impact BPD symptom expression, and potentially suicide risk in this population. Development and trial of such compounds in BPD should be prioritized. Future studies can build on this work to test mGlu5 availability as a transdiagnostic target for treatment of people with suicide attempts.