Introduction
Research has shown that individuals with high psychopathic traits form a heterogeneous population. For instance, Karpman postulated the presence of two clinical entities, both characterized by high levels of psychopathic traits but distinct levels of emotional lability, suggesting that psychopathy may result from multiple etiological pathways (Karpman,
1941). More recently, this distinction has been extended to children with high callous-unemotional traits (CU) with and/without severe levels of anxiety. Indeed, children with the primary variant (i.e., high CU traits and low anxiety) are thought to display low emotional arousal and a hypo-reactivity to social cues (e.g., facial expression of fear), whereas emotional hyperarousal and high sensitivity to negative emotions may represent the core features of the secondary variant (i.e., high CU traits and anxiety) (Craig et al.,
2021). In their recent literature review, Craig et al. (
2021) showed that 83.3% of the included studies reported that the secondary variants had significantly higher levels of childhood adversities (e.g., abuse, traumas) compared to their counterparts in the primary variant and controls. There is also evidence that children with the secondary variant also show more severe hyperactivity/impulsivity traits, internalizing traits, irritability, aggressive behaviors, substance misuse and suicidal behaviors (Cecil et al.,
2018; Fanti et al.,
2013; Goulter et al.,
2017; Huang et al.,
2020; Kimonis et al.,
2012; Meehan et al.,
2017). Although studies found significant differences between variants at a clinical level, the neurobiological markers of these variants remain largely understudied. Nevertheless, it has been postulated that variants of CU traits (and psychopathy in adults) may mainly differ in amygdala reactivity during fear processing, given that psychopathic traits and anxiety are linked to opposite activity in such region (decreased and increased, respectively) (Ashworth et al.,
2021; Blackford & Pine,
2012; Dugré et al.,
2020; Poeppl et al.,
2019). Indeed, recent studies showed consistent differences between variants in the amygdala (as a predefined region-of-interest) during fear processing (Fanti et al.,
2020; Meffert et al.,
2018; Sethi et al.,
2018). Moreover, Motzkin and colleagues (
2011) showed that the functional connectivity between the amygdala and the ventromedial prefrontal cortex differentiated both variants in adults. Aside from the interests for fear processing, the differences between variants regarding the neurobiological mechanisms underpinning motivation, reward processing and decision-making remains largely understudied.
Preclinical research pursued in the last decades has provided substantial evidence that the dopaminergic neurons projecting from the ventral tegmental area to the nucleus accumbens (NAcc) / ventral striatum (VS) and the ventro-medial prefrontal cortex (vmPFC) play a key role in motivation (Haber & Knutson,
2010; Wise,
2002). Coherently with these findings, past meta-analyses of functional neuroimaging studies in humans have consistently showed that the NAcc is involved in reward processing (Diekhof et al.,
2012; Flannery et al.,
2020; Liu et al.,
2011; Sescousse et al.,
2013), subjective valuation (Bartra et al.,
2013; Clithero & Rangel,
2014) and reward prediction error (Corlett et al.,
2022). In adults, the VS shows positive connectivity with the medial prefrontal cortex (including the ventromedial prefrontal cortex and orbitofrontal cortex), subcortical structures (e.g., amygdala and hippocampus), posterior cingulate cortex and insular cortex (i.e., anterior to posterior), and negative connectivity with the anterior midcingulate cortex, the supplementary motor area, the superior temporal gyrus and superior parietal lobule (Di Martino et al.,
2008; Janssen et al.,
2015; Zhang et al.,
2017). From childhood to adulthood, the resting-state functional connectivity between the NAcc and frontal regions (including perigenual and subgenual anterior cingulate cortex, ventromedial prefrontal cortex, and orbitofrontal cortex) linearly decreases, whereas its connectivity with the posterior insula shows a quadratic effect (Fareri et al.,
2015), highlighting its potential role in the development of various psychopathologies. In fact, a growing body of literature show that functional connectivity of the NAcc is associated with numerous psychopathologies during adolescence such as anxiety and depressive symptoms (Dorfman et al.,
2016; Pan et al.,
2017), impulsive decision-making (Costa Dias et al.,
2013), substance misuse (Huntley et al.,
2020; Morales et al.,
2021), and social problems (Fareri et al.,
2017). Thus, the maturational deficits in motivational processes are thought play a major role in our understanding of externalizing problems in children and adolescents (Bjork & Pardini,
2015).
Prior work has shown that adolescents with CU traits (Blair et al.,
2001; Scerbo et al.,
1990) and adults with psychopathic traits (Blair et al.,
2006; Mitchell et al.,
2002; Newman & Kosson,
1986) may exhibit atypical reward processing, that is, they are more likely to persist in a previously rewarded response even when the risk for punishment/losses increase. In the neuroimaging literature, the effect of CU traits on brain activity during reward fMRI tasks yields inconsistent results across studies (Byrd et al.,
2014; Murray et al.,
2018). For example, in a community sample of healthy adolescents, CU traits correlated with activity of the VS during reward anticipation, but the effect was no longer significant when controlling for severity of externalizing problems (Huang et al.,
2019). Some have found that CU traits were negatively associated with the medial prefrontal cortex but not the ventral striatum during reward anticipation (Veroude et al.,
2016), whereas others found that CU traits were unrelated to reward anticipation (Murray et al.,
2023). When receiving rewards, youths with disruptive behavior disorder (DBD) and elevated CU traits showed reduced activity of the dorsal striatum (but not ventral) as a function of prediction error when receiving reward (White et al.,
2013). Similarly, Zhang and colleagues (Zhang et al.,
2023) found that CU traits were negatively associated with activity of the dorsal striatum (but not the ventral part) in response to reward relative to punishment. These conflicting results may be partially explained by the relatively small sample sizes used to detect significant effect of the ventral striatum. For example, in a recent study of 995 youths with DBDs, Hawes and colleagues (Hawes et al.,
2021) found that those with high CU traits (DBD + CU) were characterized by reduced activity in dorsal anterior cingulate cortex (as well as those with low CU traits [DBD-CU]) compared to their counterparts in the typically developing group during reward anticipation. Children with DBD-CU additionally exhibited reduced ventral and dorsal striatal activity during reward anticipation (Hawes et al.,
2021). When receiving rewards, both DBD + CU and DBD-CU groups showed greater activation of the NAcc and OFC, compared to controls (Hawes et al.,
2021). Other studies found limited evidence of differences in brain activity during reception of reward between children with high CU traits and high conduct problems and controls (Byrd et al.,
2018; Finger et al.,
2011). In adults, some studies showed that the severity of psychopathic traits correlated with VS activity when anticipating rewards (Bjork et al.,
2012), whereas others showed no such effect when viewing drug cues (Cope et al.,
2014) or a greater effect in loss rather reward reception (Pujara et al.,
2014). Across the limited number of studies using the NAcc (or VS) as a seed of interest during resting-state, similar divergence across results is observed. Indeed, Hosking et al. (Hosking et al.,
2017) found that the functional connectivity between the NAcc and the medial prefrontal cortex was negatively associated with severity of psychopathic traits in incarcerated adults (PCL-R). Moreover, Factor 2 of the PCL-R (but not Factor 1) positively correlated with functional connectivity between the NAcc and dorsolateral prefrontal cortex and negatively correlated with functional connectivity between the NAcc and the postcentral gyrus (Korponay et al.,
2017). However, other studies found no significant difference in VS functional connectivity between adult offenders with psychopathy and those without psychopathy (Motzkin et al.,
2014) or between adults with an antisocial personality disorder (ASPD) and elevated psychopathic traits (PCL-R ≈ 25) and those without ASPD (Kolla et al.,
2018).
It is noteworthy to mention that these discrepancies may principally originate from the large heterogeneity in population with high CU/psychopathic traits (e.g., variants). Indeed, recent results indicate that at low levels of social adversities (e.g., Foster Home, Divorced Parents, Welfare Food Stamps), high CU traits were associated with reward hypo-responsivity (i.e., less pre-ejection period shortening), whereas higher CU traits were associated with reward hyper-responsivity at high levels of social adversities (Gao & Zhang,
2021). In addition, some preliminary results also suggest that individuals with the primary variant (but not those on the secondary variant) may be unable to integrate socio-affective information into decision-making to select the appropriate behaviors (Koenigs et al.,
2010,
2012). While adolescents with high CU traits and adults with high psychopathic traits may show aberrant reward processing and decision-making, the neurobiological differences between the primary variant (hypo-arousal) and those with the secondary variant (hyper-arousal) remain to be elucidate.
To our knowledge, no studies have examined the NAcc functional connectivity between variants of CU traits, leaving unknown whether they may be characterized by specific neurobiological impairments. Despite that variants are well described at a clinical level, searching for neurobiological markers of variants in childhood and adolescence is of utmost importance to gain insight of their underlying mechanisms and better characterize their developmental route. To achieve this goal, we conducted a latent profile analysis (LPA) to extract data-driven subgroups in a large sample of children and adolescents using callousness and anxiety as dimensions of interests. We subsequently conducted seed-to-voxel analyses using the bilateral NAcc as seeds of interest to examine differences in functional connectivity between variants. Given that some evidence suggests that primary variant (but not the secondary variant) may show similar utilitarian decision-making and clinical presentation as patients with lesions to the vmPFC (Koenigs et al.,
2010), we further hypothesized that this hypo-arousal group may be characterized by decreased functional connectivity within the mesocorticolimbic system (i.e., NAcc and vmPFC, as similarly found in (Hosking et al.,
2017), whereas the secondary variant (hyper-arousal group) may rather be characterized by decreased connectivity between the NAcc and regions involved in regulatory mechanisms (e.g., ventro- and dorso-lateral PFC, aMCC/pre-SMA, see meta-analyses on emotion regulation: (Kohn et al.,
2014; Zilverstand et al.,
2017) given their potential hyper-responsivity to reward (Gao & Zhang,
2021). In addition, considering that some effects found in reward processing are also related to severity of impulsivity/antisocial factor and are observed in adolescents and adults with Conduct Disorder/Antisocial Personality Disorder (Buckholtz et al.,
2010; Carré et al.,
2013; Hawes et al.,
2021; Huang et al.,
2019; Murray et al.,
2018; Rubia et al.,
2009) (Bubenzer-Busch et al.,
2016; Crowley et al.,
2010; Völlm et al.,
2007), we conducted supplemental analyses controlling for the severity of hyperactivity/impulsivity symptoms as well as conduct problems.
Discussion
In our study, we aimed to investigate differences between variants of psychopathy in NAcc functional brain connectivity using a large sample size of adolescents. Latent Profile Analysis using CU traits and anxiety revealed 3 homogeneous subclasses (ANX, TD, CU/ANX+) but failed to identify the expected variants. These groups did not statistically differ on functional connectivity of the NAcc when using a stringent statistical threshold across the whole-brain (p < 0.001 uncorrected with pFWE < 0.05). However, when using a more liberal threshold at a cluster level (> 20 voxels), we observed that groups differed on NAcc connectivity to the pINS, lOFC, BA19 as well as AG, SMA, and SPL. Secondary analyses using only the Callousness subscale of the ICU successfully identified the primary and the secondary variants. However, the four groups only statistically differed in NAcc functional connectivity when using a more liberal threshold (> 20 voxels), replicating the pINS and SMA findings and additionally showing a potential specific dysconnectivity between the NAcc and the STG in the primary variant. These results highlight the importance of studying subgroups of children and adolescents exhibiting high levels of callousness and offer novel insight about the potential neurobiological differences between variants.
Despite that individuals with high psychopathic traits are traditionally characterized by an absence of anxiety (Cleckley,
1951; Karpman,
1941) and fearlessness (Lykken,
1995), a non-negligible percentage of them actually report high levels of anxiety. Indeed, the secondary variant is thought to show a more severe clinical presentation compared to the prototypical one. In our data-driven analysis using the ICU total score and SCARED, we failed to identify the primary variant. While others have been unable to identify the primary (Euler et al.,
2015; Lee et al.,
2010) or the secondary variant (Colins et al.,
2018; Goulter et al.,
2017), one possible explanation is that the primary variant may be more easily identified through justice-involved sample including only males, whereas the secondary variant may be more prevalent in clinical settings including both sexes (Craig et al.,
2021). Here, the community-referred recruitment model and the inclusion of both sexes may have explained the inability to find a primary variant across the sample. Also, of the data-driven studies aiming to identify variants of CU traits (Craig et al.,
2021), majority uses other co-occurrent features (e.g., CP, physical, emotional, and/or sexual abuse, trauma) which raise the question whether the variants depend on other features rather than solely on levels of anxiety and CU traits. Still, some failed to identify the secondary variant even after adding other clustering features such as maltreatment and negative affect (Colins et al.,
2018). Yet, another possibility is that some of the subconstructs of CU traits may blur the ability to adequately capture the inter-individual variability underpinning variants. In the current study, variants were successfully found when using the callousness score of the ICU, but not the total score. This may be partially explained by the fact that variance in subscales of the ICU may largely reflects variance from the general factor (Ray & Frick,
2020), but they remain only moderately correlated, as observed in the current study (
r ranging from 0.40 to 0.66). Similarly, fear and anxiety are poorly distinguished in research on psychopathy (Hofmann et al.,
2021; Hoppenbrouwers et al.,
2016), leaving unknown whether deficits in threat detection or responsivity may improve the identification of variants compared to the usual subjective measure of trait anxiety. Unequivocally, future studies should specifically aim to identify the core features (the most optimal set of clustering variables) delineating the primary and secondary variants in order to provide a more standardized way to identify these children in research but also in clinical practice.
Individuals with co-occurrent psychopathic traits and high levels of anxiety are typically characterized by a dysregulated clinical profile which include borderline personality features (Blackburn & Coid,
1999; Goulter et al.,
2019; Skeem et al.,
2003,
2007). On a neurobiological level, we found that this particular group significantly differ from TD, and ANX, in functional connectivity between the NAcc and pINS, lOFC, AG, and SMA. However, when comparing the secondary to the primary variants (found in the subsequent analyses), only the NAcc-pINS and NAcc-SMA connectivity replicated, suggesting important deficits in the secondary group. In the TD group, we found a significant connectivity between the NAcc and the pINS, but not with the SMA, suggesting that the latter connectivity may be aberrant in the secondary variant. While the interpretation of this aberrant connectivity remain elusive, further investigation is necessary to identify whether this functional connectivity may be related to specific symptoms not found in TD, may reflect a brain reorganization, or a spurious result. Across neuroimaging literature, the pINS appear to be implicated in processing sensory information (i.e., interoceptive processes, (Kurth et al.,
2010; Tian & Zalesky,
2018), whereas the SMA is often linked to motor planning, sensory and memory tasks (Chung et al.,
2005; Sheets et al.,
2021), future studies should aim to examine the functional roles of these connectivity in the specific symptomatology of children with the secondary variant that may distinguish them from the primary variant.
Prior work suggested that individuals with the primary variant may be characterized by abnormal decision-making including utilitarian moral decision (Koenigs et al.,
2010,
2012). We thus hypothesized that this group may be characterized by decreased functional connectivity between the NAcc and vmPFC. However, groups did not significantly differ in NAcc-vmPFC connectivity. However, we found an increased connectivity between NAcc and STG in the primary variant compared to other groups. In healthy subjects, both NAcc and STG are co-activated during reward processing (Arsalidou et al.,
2020; Lopez-Gamundi et al.,
2021; Wilson et al.,
2018) but also during social cognition including self-agency (Sperduti et al.,
2011) and personal perspective during moral reasoning (Boccia et al.,
2017). Deficits in activity of this particular region was observed in offenders with an antisocial personality disorder and psychopathy during reversal learning (i.e., rewarded responses > punished errors) (Gregory et al.,
2015). Although we did not find any difference in functional connectivity between the core regions of reward processing, the NAcc-STG connectivity highlights the interaction between reward and other potential networks (e.g., social cognition) that may underpin behaviors that are specific to the primary variants.
Limitations
The current study aimed to examine differences in NAcc functional connectivity between variants of callous traits using a large sample of children and adolescents. Nevertheless, some limitations need to be acknowledged. First, our sample comprised children and adolescents with psychopathologies, recruited using a community-referred recruitment model. It is thus difficult to interpret our results given the absence of a true control group with no psychopathologies. Furthermore, the absence of such group could have reduced the ability to detect significant differences between groups. We still found significant between-group differences using a large sample. Studies should seek to examine whether the functional connectivity differences found in our study significantly discriminate between variants of callous traits and healthy controls. Secondly, neuroimaging suffers from a replicability crisis, which increase concerns about spurious results due to limited sample size and methodologies. In our study, the length of resting-state fMRI was relatively short (10 min). However, the UK Biobank include only 6 min resting-state scanning session and show similar results to those with > 20 min (i.e., ABCD & HCP) (Marek et al.,
2022). A longer scanning session from 10 to 20 min (Anderson et al.,
2011; Birn et al.,
2013; O’Connor et al.,
2017) is then preferred to a single 5-6 min, however gains in intersession reliability reduce after 9–12 min (Birn et al.,
2013). In addition to the scan length (i.e., 10 min total), the Healthy Brain Network include a TR = 0.8 with a multiband of 6, which inherently increases the number of acquired volumes (see (Jahanian et al.,
2019; Liao et al.,
2013). We acknowledge that a longer scanning session would have been optimal, the scan length, the sample size and the number of volumes acquired meet the current recommendations in the resting-state neuroimaging literature; suggesting that they should provide reliable estimates. Studies aiming to replicate our findings are strongly encouraged. Thirdly, prior work using data-driven techniques to identify variants with the ICU also include other variables such as childhood maltreatment (Craig et al.,
2021). In our study, childhood maltreatment was not assessed. Since adverse childhood events are not equivalent to childhood maltreatment, it remains unknown whether the secondary variant found in our study reported higher childhood maltreatment compared to the primary variant group. Lastly, we did not use IQ as a potential confounder given that theoretical framework of the brain structures underpinning IQ does not involve the NAcc (Jung & Haier,
2007). However, it is possible that including IQ as a covariate may have provided a more precise estimate of the NAcc-lateral PFC.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.