Affective reactivity during adolescence: Associations with age, puberty and testosterone

Abstract

Adolescence is a period of heightened social engagement that is accompanied by normative changes in neural reactivity to affective stimuli. It is also a period of concurrent endocrine and physical changes associated with puberty. A growing body of research suggests that hormonal shifts during adolescence impact brain development, but minimal research in humans has examined the relationship between intra-individual changes in puberty and brain function. The current study examines linear and nonlinear changes in affective reactivity in a longitudinal sample of 82 adolescents who underwent three fMRI sessions between the ages of 9 and 18 years. Changes in response to affective facial stimuli were related to age, pubertal stage, and testosterone levels. Using multilevel modeling, we highlight extensive nonlinear development of socio-emotional responsivity across the brain. Results include mid-pubertal peaks in amygdala and hippocampus response to fearful expressions, as well as sex differences in regions subserving social and self-evaluative processes. However, testosterone levels exhibited inverse patterns of association with neural response compared to pubertal stage in females (e.g., U-shaped relationship with the amygdala and hippocampus). Findings highlight potentially unique roles of age, pubertal stage and testosterone on socio-emotional development during adolescence, as well as sex differences in these associations.

Introduction

Adolescence is characterized by shifting social motivations that are accompanied by heightened emotional reactivity and increased sensitivity to social interactions. It is also a period of concurrent biological maturation, including endocrine and physical changes associated with puberty. Animal research indicates that hormonal shifts during this period underlie development of brain regions that subserve social and emotional processes (Juraska and Willing, 2017, Sisk and Foster, 2004). However, longitudinal research with adequate power to detect these developmental processes is limited in humans. Predominant theories of adolescent neurodevelopment historically propose that pubertal maturation drives alterations in subcortical processing of affective stimuli around early adolescence, while age- or experience-related maturation is responsible for more protracted changes in frontal brain regions important for cognitive regulation (Casey et al., 2008, Nelson et al., 2005, Steinberg, 2008). As such, characterizing pubertal associations with brain development also provides an important test of these heuristic “mismatch” models of adolescent neurodevelopment.

One line of investigation on adolescent socio-emotional functioning has focused on the ability to understand the emotional states of others – a crucial skill for successful social interactions. Facial expressions are an important medium for emotional communication, and consequently, research has commonly examined changes in neural responses to affective facial stimuli during adolescence. Early studies revealed greater subcortical activation in the amygdala when passively viewing fearful faces in adolescents compared to adults (Guyer et al., 2008, Monk et al., 2003), while others have shown decreased activation with age over this period (Gee et al., 2013, Killgore et al., 2001, Swartz et al., 2014). Age-related increases in nucleus accumbens (NAcc) response to happy faces between early and mid-adolescence (Pfeifer et al., 2011) also suggest there might be differential processing of both negatively- and positively-valenced social stimuli. There is also some evidence supporting adolescent “peaks” in limbic reactivity to affective facial stimuli (primarily using emotional control paradigms) compared to children and adults (Hare et al., 2008, Somerville et al., 2011), although studies examining neural responses when passively observing affective facial stimuli have largely failed to identify such non-linear trends (Flannery et al., 2017, Guyer et al., 2008).

Beyond limbic changes in socio-emotional responsivity, a number of age-related changes have been reported in the prefrontal cortex (PFC), including linear increases in lateral PFC response to fear faces over adolescence (Killgore et al., 2001, Killgore and Yurgelun-Todd, 2006) as well as a nonlinear mid-adolescent “peak” to affective facial expressions when considering longer age spans (Flannery et al., 2017). Comparatively, converging evidence supports age-related reductions in neural response to affective facial expressions in the medial PFC, including the anterior cingulate cortex [ACC] (Monk et al., 2003, Nelson et al., 2003), orbitofrontal cortex [OFC] (Flannery et al., 2017, Monk et al., 2003) and ventromedial [vm]PFC (Flannery et al., 2017, McRae et al., 2012), although increases may be present when considering early adolescence alone (Pfeifer et al., 2011). While many of these studies conducted region-of-interest analyses that focused on the PFC, others using a whole brain approach have additionally noted developmental changes in responsivity of the insula, posterior superior temporal sulcus, temporal pole, and fusiform cortices (Deeley et al., 2008, Gee et al., 2013, McRae et al., 2012, Pfeifer et al., 2011, Pine et al., 2001). In summary, there appears to be widespread changes across limbic, prefrontal, temporal and occipital cortices to affective facial stimuli over adolescence, which leads to the question of what biological processes may be driving these developmental changes.

Following on from research on age-related development, a smaller number of studies have investigated pubertal associations with neural processing of affective facial stimuli. Many of these studies have focused on limbic structures, and in particular the amygdala, given the prevalence of sex hormone receptors in the subcortex (Abdelgadir, Roselli, Choate, & Resko, 1999). Animal research has also shown pubertal changes in neurons and supporting processes, such as pruning of dendrites and synapses, in the amygdala and hippocampus (Cooke and Woolley, 2005, Drzewiecki et al., 2015, Zehr et al., 2006, Zehr et al., 2008). Human research using cross-sectional designs have identified reductions in amygdala activation to emotionally neutral facial stimuli with pubertal stage (Ferri, Bress, Eaton, & Proudfit, 2014) as well as pubertal timing (i.e., pubertal stage relative to same-aged peers, Forbes, Phillips, Silk, Ryan, & Dahl, 2011). Others have noted reduced striatal activation to fearful and angry faces with earlier pubertal timing (Whittle et al., 2015). However, longitudinal studies have shown increased amygdala and hippocampus response with pubertal stage for both positively- and negatively-valenced emotions (Moore et al., 2012), as well as increased amygdala and NAcc response for threat-related processes with rising testosterone levels (Spielberg, Olino, Forbes, & Dahl, 2014). Taken together, there appears to be some inconsistencies in limbic changes identified by cross-sectional and longitudinal studies. Moreover, despite specific hypotheses about puberty-driven peaks in functional reactivity of subcortical structures to affective stimuli (Casey et al., 2008, Nelson et al., 2005, Steinberg, 2008), there have been no longitudinal empirical investigations of nonlinear changes in limbic function elicited by affective faces in relation to pubertal maturation.

Even less is known about puberty-related changes in affective reactivity across the cortex. Cross-sectional findings suggest that adolescents in later pubertal stages (controlling for age) exhibit less ventrolateral (vl)PFC activation to fearful faces, but more activation to angry faces (Forbes et al., 2011). Comparatively, a prior investigation of (a subsample of) the current dataset found increased vlPFC responses to fearful, happy and neutral expressions at later pubertal stages in 13 year olds, as well as increased dorsomedial (dm)PFC response to happy facial expressions, and temporal pole response to happy, sad, fearful and neutral expressions (Moore et al., 2012). A generalized pattern of increased engagement of occipital cortices to emotional expressions at later pubertal stages has also been reported (Moore et al., 2012, Pagliaccio et al., 2014). Based on these limited studies, it appears that pubertal effects may lie within similar regions in the frontal, occipital and temporal cortices that exhibit age-related changes. However, further research is needed to corroborate the pattern of cortical changes identified in this limited literature, as well as better understand the role of pubertal hormones in these changes.

The current study aimed to extend the literature on pubertal associations with neural processing of affective facial stimuli, using a single-cohort longitudinal sample that spanned 9–18 years of age. While age- and puberty-related trajectories have previously been examined in this typically developing cohort (Moore et al., 2012, Pfeifer et al., 2011), the current investigation incorporated an additional (third) wave of assessments that enabled the modeling of nonlinear developmental trajectories, and also conducted novel analyses involving pubertal hormones. At each assessment, adolescents completed an affective faces fMRI task, where they passively observed exemplars of five different basic emotional expressions in a rapid event-related design. They also completed a questionnaire assessing their pubertal stage and provided a sample of their saliva that was assayed for testosterone levels.

Using multilevel modeling, we investigated linear and quadratic trajectories of subcortical and cortical activation to the affective faces task, using both age and pubertal stage as indices of maturation. We specifically tested for the presence of mid-pubertal “peaks” (i.e., inverted-U shaped trajectories) in neural response to affective facial stimuli. We hypothesized subcortical changes in the amygdala, hippocampus and NAcc, as well as cortical changes in the prefrontal, temporal and occipital regions. However, it was difficult to make hypotheses with further specificity or directionality based on prior research. We also examined the relationship between neural activity and testosterone levels, hypothesizing that hormones may partly underlie puberty-related changes in neural responsivity. Across all analyses, differences in sex and emotional condition were examined. Although these were also largely exploratory in nature, we did speculate that females may exhibit greater adolescent peaks in subcortical and medial PFC activation given they have more intense subjective experiences of, and greater biological reactivity to, affective stimuli compared to their male counterparts during this period (Garber et al., 2002, Nolen-Hoeksema, 2016, Ordaz and Luna, 2012, Yang et al., 2017, Yuan et al., 2014).