Maturation of amygdala inputs regulate shifts in social and fear behaviors: A substrate for developmental effects of stress

Abstract

Stress can negatively impact brain function and behaviors across the lifespan. However, stressors during adolescence have particularly harmful effects on brain maturation, and on fear and social behaviors that extend beyond adolescence. Throughout development, social behaviors are refined and the ability to suppress fear increases, both of which are dependent on amygdala activity. We review rodent literature focusing on developmental changes in social and fear behaviors, cortico-amygdala circuits underlying these changes, and how this circuitry is altered by stress. We first describe changes in fear and social behaviors from adolescence to adulthood and parallel developmental changes in cortico-amygdala circuitry. We propose a framework in which maturation of cortical inputs to the amygdala promote changes in social drive and fear regulation, and the particularly damaging effects of stress during adolescence may occur through lasting changes in this circuit. This framework may explain why anxiety and social pathologies commonly co-occur, adolescents are especially vulnerable to stressors impacting social and fear behaviors, and predisposed towards psychiatric disorders related to abnormal cortico-amygdala circuits.

Fear and anxiety disorders are the most prevalent subtype of neuropsychiatric disorders, with a lifetime prevalence of 30% of the United States population and over 8% for posttraumatic stress disorder alone (Kessler et al., 2005; Kilpatrick et al., 2013). These fear- and anxiety-related disorders, such as generalized anxiety disorder or post-traumatic stress disorder, are characterized by heightened fear responding and social dysfunction (e.g., Berle et al., 2018; Mahan & Ressler, 2012). Fear learning and responding is at the core of these disorders, but these processes can be highly variable and are affected by several different factors, including quality of the social environment and prior stressful experience. The influence of social environment and stress changes over the course of the lifespan. Furthermore, several maturational periods overlap with increases in degree of symptom presentation and severity, and although the typical age of onset varies depending on the specific type of anxiety disorder, overall diagnosis rates of anxiety-related disorders are disproportionately high during childhood and adolescence (Hartley & Lee, 2015). This suggests a potential developmental sensitivity where younger populations are uniquely disadvantaged. These disorders have a high degree of symptom persistence despite intervention (Essau et al., 2018; Yonkers, et al., 2003) and can increase risks for additional neuropsychiatric disorders. For instance, diagnosis of an anxiety disorder during adolescence corresponds with increases in later alcohol (Dyer et al., 2019) and substance abuse (Woodward & Fergusson, 2001). Despite findings suggesting adolescents may be particularly vulnerable to fear- and anxiety-related disorders, many current behavioral therapies are based on adults and may not necessarily be effective across ages.

Differences in fear and social behaviors are hallmarks of the transition from adolescence to adulthood (Figure 1), and sensitivity of these behaviors to stressors differs in these age groups. Adolescence corresponds with both a heightened social interaction that progressively declines into adulthood (Varlinskaya & Spear, 2008) and elevated reactivity to some discrete fear cues (Pattwell et al., 2012). This progressive decline in the propensity to value and engage in general social interaction (i.e., social drive) and increase in fear- and anxiety-related disorder onset coincides with a period of cortico-BLA maturation (Figure 1A). The importance of the amygdala in the behavioral changes from adolescence to adulthood has been well established (e.g., Scherf et al., 2013). The amygdala is part of a circuit that strongly influences social and fear behaviors, and includes several frontal cortical regions and subcortical limbic regions. The prefrontal cortex (PFC), comprised of rodent ventral PFC areas of prelimbic and infralimbic cortices (mPFC) and the dorsal area of anterior cingulate cortex (ACC) have been implicated in the regulation of fear (Descalzi et al., 2012; Sotres-Bayon et al., 2012; see Gilmartin et al., 2013, for a review) and social behaviors (Jeon et al., 2010; Rudebeck et al., 2007; Van Kerkhof et al., 2013). Their projections to the amygdala undergo synaptic strengthening during development (Arruda-Carvalho et al., 2017) and have been implicated in the control of excitation and inhibition in the amygdala. Development of these cortical regions may therefore provide a mechanism for the transition in fear and social behaviors through changes in the balance between excitatory and inhibitory influences on the amygdala. This shared reliance on the maturation of cortico-amygdala circuitry may explain the similar developmental trajectory for social and emotional functions. Poor social environment or high stress exposure during adolescence commonly produces a phenotype with comorbid heightened anxiety and impaired social function. We propose that this overlap of circuitry may help explain why social and emotional function share high sensitivity to social environmental conditions and stress exposure during adolescent periods of PFC-BLA maturation. And conversely, the lasting comorbid anxiety and social deficits caused by stress or the social environment lead to behavioral changes in the way that one interacts with the social environment, such as social avoidance or self-isolation, that subsequently shape the maturation of PFC-BLA circuitry. In addition to providing insight into the pathology of comorbid social and emotional abnormalities, understanding this relationship between these factors together could provide hints towards new age-specific treatment strategies. Together, changes to the social environment, stress sensitivity, fear, anxiety and social behavior, and cortico-BLA maturation are all linked to provide insight about why adolescents are susceptible to neuropsychiatric disorder development and how this is different from other age groups. In this review, we will discuss the developmental shifts in cortico-amygdala circuitry and the interactions between stress, fear, and social behavior during rodent development, with a major focus on rat literature to provide a mechanistic insight as to how these processes mature during development. We present evidence that developmental behavioral changes during adolescence are driven partly through maturation of PFC-amygdala circuitry (Figure 1) that modulates the balance of excitation and inhibition in the amygdala.

Figure 1. Cortical-BLA maturation coincides with shifts in fear and social behaviors.

(A) Throughout adolescence, maturation of PFC inputs to BLA and the ability to regulate BLA function increases. Over adolescence, social interaction and other behaviors that reflect social motivation peaks and then declines into adulthood. In parallel, sensitivity to the social environment decreases into adulthood, for instance as measured by changes caused by isolation. At the same time, the regulation of fear responding increases, as measured with extinction learning. (B) The types of social engagement change over the course of the lifespan, with high degree of play behavior seen during adolescence, and more cautious contact or investigative behaviors commonly seen in adults. Play behavior during adolescence is essential for the development of social behaviors and responding during adolescence and adulthood. Deprivation of play behavior, as can be produced by social isolation, produces lasting changes that alter age-specific social responding.

Local amygdala and cortico-amygdala maturation

The amygdala is comprised of several different subregions. The basolateral amygdala (BLA) complex in particular has been implicated in fear learning and memory (e.g. Fanselow & Ledoux, 1999; Gale et al., 2004), social behavior (Sajdyk & Shekhar, 1997; Truitt et al., 2007), and cortical inputs to this region continue to mature through adolescence (Arruda-Carvalho et al., 2017). BLA activity and its output to other brain regions guides behavioral responding. The activity of these outputs is determined by many factors. Foremost among these factors are excitatory glutamatergic and inhibitory GABAergic processes. Primarily glutamatergic synaptic inputs from cortical and thalamic areas can directly excite BLA projection neurons or indirectly inhibit projection neurons by activation of GABA-releasing interneurons (Figure 2). These inputs are sensitive to many external conditions and internal states (e.g., stressors, experience-related learning, pain, internal drive states), leading to a shift in the balance of excitation and inhibition in the BLA, in order to promote specific behavioral responses. This suggests that a shift in this BLA balance can promote behavioral responding according to environmental demands or internal drives.

Figure 2. Maturation of cortical-BLA circuits increases fear regulation and decrease social drive.

The cellular correlates of PFC-BLA regulation have been elucidated in several studies, and are simplified here. The firing activity of BLA neurons is responsive to inputs from the PFC, displayed here with in vivo extracellular recordings during PFC stimulation in anesthetized rats. PFC stimulation can produce excitatory or inhibitory actions on BLA projection neurons. The excitatory action can be isolated in vivo (A, B, top; when GABA_A receptors are blocked with picrotoxin, 10 pmol intra-BLA). The relative degree of excitation that can be produced by PFC inputs is weaker in adolescents than adults (A, B, top). However, when GABA_A systems are intact (without picrotoxin), the excitatory effects of PFC are limited by recruitment of inhibitory elements in the BLA. The inhibitory impact of PFC inputs to BLA is weaker in adolescents compared to adults (A, B, lower). A potential explanation for this is that PFC inputs to the BLA are not functionally mature in adolescents (C, D, schematic), and the inhibitory interneuron elements in the BLA of adolescents are also not functionally mature, where ongoing PFC maturation is indicated by a dashed box and fewer contacts in adolescents when compared to adults. Therefore, PFC inputs are less able to recruit inhibition of the BLA in adolescents, and less able to regulate BLA neuronal firing activity. When compared to adults, adolescents engage in more social interaction and these interactions are more rewarding and non-specific, but adolescents show relatively poor fear inhibition. Maturation of PFC inputs to the BLA is associated with less overall social interaction, more selective social interactions, and better fear regulation, perhaps due to mature ability to recruit inhibitory processes in the BLA.

Developmental differences in the influence of inputs on the balance between BLA excitation and inhibition may promote age-specific behaviors. There are several mechanisms that can lead to changes in the balance between excitation and inhibition during adolescence. For instance, a developmental shift can be driven by increases in GABAergic tone in the BLA that regulate BLA neuronal activity in rodents. This is supported by work demonstrating preferential increases in BLA neuronal firing following GABA receptor inhibition in adults but not in adolescents (Zhang & Rosenkranz, 2016), indicative of a more potent ongoing influence of GABA over BLA neuron firing in adults. In addition, increases in the degree of GABAergic synaptic input in BLA, seen as frequency of spontaneous inhibitory postsynaptic currents (IPSCs), occur from adolescence into adulthood. In contrast, there are only minor changes in the degree of glutamatergic input to BLA projection neurons, measured as spontaneous excitatory PSCs (sEPSC), miniature EPSCs (mEPSC), or responses to glutamate application between adolescence and adulthood. This indicates that changes in BLA function might be primarily related to increases in GABAergic systems between adolescence and adulthood (Figure 2), instead of a shift in overall glutamatergic signaling (Zhang & Rosenkranz, 2016). However, despite similarities between age groups in overall glutamatergic signaling, specific glutamatergic inputs to the BLA may change during development and may be sensitive to changes in environmental demands. While this review generally focuses on glutamatergic and GABAerigc systems globally, many peptide systems change during development to regulate neuronal activity and behavior. For example, oxytocin neurons are essential for a variety of developmental changes in behavior, and their receptors are present at PFC-BLA synapses regulating social recognition (Lukas et al., 2010; Tan et al., 2019). This highlights that the discussion of excitatory and inhibitory processes may have far-reaching implications that include peptide receptors during development at PFC-BLA synapses.

PFC inputs to BLA stand out due to their later maturation. These inputs change substantially from adolescence to adulthood, characterized by shifts in the connectivity and strength at PFC-BLA synapses (Arruda-Carvalho et al., 2017). Retrograde tracing of BLA-projecting PFC neurons has shown that the total number of PFC neurons projecting to the BLA are stable from childhood to late adolescence, but this decreases by nearly 50% from late adolescence (PND 45) to adulthood (PND 90), demonstrating substantial pruning of PFC-BLA projections (Cressman et al., 2010). These results show that anatomical changes in PFC-BLA connectivity occur from early adolescence to adulthood. Further, retrograde and anterograde tracing techniques have demonstrated the continued maturation of the reciprocal BLA-PFC circuitry into late adolescence (PND 45). This approach showed that the most extensive increases in the density of amygdala fibers in the PFC occur until PND 50 and progressively slow into adulthood (Cunningham et al., 2002; Verwer et al., 1996).

Our understanding of the reciprocal connections between BLA and PFC and how they are regulated has improved greatly in the past few years due to the ability to selectivity manipulate and record individual projections and subsets of inhibitory interneurons. Inhibitory interneurons potently regulate activity of PFC and BLA output with consequences on fear, anxiety and social behaviors (Courtin et al., 2024; Truitt et al., 2007; 2009). Recent studies delineate the intraPFC GABAergic circuitry, and how it impacts specific sets of PFC outputs. In the PFC, similar to most cortical regions, inhibitory interneurons can be categorized via a number of different parameters (DeFelipe et al., 2013; Markram et al., 2004), but more broadly categorized by expression of parvalbumin (PV+), somatostatin (SOM+), or vasoactive intestinal peptide (VIP+), with or without cholecystokinin (CCK+) or calretinin (Kawaguchi & Kondo, 2002). While all categories of PFC interneurons can regulate PFC projection neurons, PV+ interneurons provide substantial perisomatic inhibition that strongly regulates projection neuron firing (Lucas & Clem, 2017). However, PV+ interneurons can be regulated by SOM+ interneurons or other PV+ interneurons to permit fear expression as a result of direct experiences or when learned through social inference (Courtin et al., 2014; Xu et al., 2019). The effects of PV+ or SOM+ activation are bidirectional, as activation of either PV+ or SOM+ neurons at specific frequencies can also facilitate social behavior (Liu et al., 2020). Layered on top of this, VIP+ interneurons can inhibit SOM+ neurons (and PV+ neurons to a lesser degree), leading to disinhibition of PFC projection neurons (Pi et al., 2013). There are variable degrees of inhibition from specific sets of PFC interneurons onto specific sets of PFC outputs (Anastasiades et al., 2018; Lee et al., 2014; Lu et al., 2017), including onto PFC neurons that project to BLA (Lu et al., 2017).

BLA inputs to PFC show some regional variability, but generally BLA inputs can activate PFC neurons that project to a variety of regions, including those that project back to BLA, to the nucleus accumbens, and to brainstem/hypothalamic regions that influence autonomic responses (Cheriyan et al., 2016; Gabbott et al., 2006; 2012; Little & Carter et al., 2013; McGarry & Carter, 2016; McGinty & Grace, 2007). In addition to promoting PFC excitation, inputs from the BLA notably activate inhibitory circuits within the PFC (Dilgen et al., 2013). Within the PFC, BLA inputs can activate PV+ and SOM+ interneurons that subsequently inhibit PFC neurons, many of which are in the IL, that project back to BLA (Gabbott et al., 2006; McGarry & Carter, 2016). However, maturation of BLA inputs to PFC continues through adolescence (Cunningham et al., 2002; Johnson et al., 2016). Paired with this, inhibition within the PFC also continues to develop into adolescence and early adulthood (Baker et al., 2017; Drzewiecki et al., 2020; Koppensteiner et al., 2019; Rinetti-Vargas et al., 2017; Vincent et al., 1995; Yang et al., 2014), and there is evidence for differences in the maturation of inhibition onto different sets of PFC projection neurons (Vandenberg et al., 2015). This can result in late maturation of BLA inputs to PFC interneurons (Cunningham et al., 2008), ultimately limiting the ability of the BLA to reduce PFC outputs back to BLA during adolescence.

BLA GABAergic interneuron populations can also be defined by expression of PV, SOM or CCK with calbindin, or a combination of calretinin and CCK or VIP (Kemppainen & Pitkänen, 2000; Mascagni & McDonald, 2003; McDonald & Mascagni, 2001; McDonald et al., 2002). PV+ interneurons make up the largest proportion of BLA interneurons and tend to target somata and proximal regions of BLA projection neurons (Muller et al., 2006; Vereczki et al., 2016), with some targeting other interneurons that can include PV+ interneurons (Muller et al., 2005; Wolff et al., 2014). Other interneurons, such as SOM+, tend to target more distal dendritic regions or other interneurons (Muller et al., 2003; 2007), while some CCK+ interneurons form perisomatic baskets around projection neurons, similar to PV+ cells, or distal inhibitory inputs similar to SOM+ cells (Mascagni & McDonald, 2003; Veres et al., 2017). Similar to that of other interneuron populations, VIP+ interneurons also target principle cells and other interneurons (Rhomberg et al., 2018; for review Lucas & Clem, 2018). In the amygdala, interneuron-interneuron regulation plays a critical role in behavior, clearly seen with fear that is expressed as a result of PV+ interneurons inhibition of SOM+ interneurons (Cummings & Clem, 2020; Wolff et al., 2014). However, the inhibition over BLA projection neurons can shift. For instance, sets BLA neurons that project to PL are active during fear expression, while BLA neurons that project to IL are preferentially activated to inhibit fear responses (Ganella et al., 2018; Senn et al., 2014). This shift may be related to activation of perisomatic inhibition onto these different sets of BLA projection neurons (Davis et al., 2017; Trouche et al., 2013; Rovira-Estaban et al., 2019; Vogel et al., 2016). On top of that, PFC inputs to BLA potently activate BLA interneurons that inhibit BLA projecting PFC neurons (Hübner et al., 2014), allowing PFC to gate BLA-IL or BLA-PL neuronal activation. Taken together, this suggests a reciprocal circuitry whereby select sets of BLA neurons can shift the balance of activity between PL and IL depending on whether a stimulus signals danger or safety.

The influence of PFC on BLA activity continues to mature during adolescence. Activity from PFC inputs can directly excite BLA projection neurons and indirectly inhibit them, via GABAergic interneurons (Rosenkranz & Grace, 2002). Additional sources of inhibition may include the medial or lateral intercalated groups of inhibitory neurons (Busti et al., 2011; Marowsky et al., 2005; Pinard et al., 2012; Royer et al., 1999; Skelly et al., 2016). The ability of the PFC to inhibit BLA projection neurons increases over the course of adolescence in rats (Selleck et al., 2018). In vivo measures of BLA neuronal activity include action potential firing and local field potentials (LFPs) that reflect coordinated activation of many synapses and the postsynaptic response of many neurons. Because LFPs can reflect postsynaptic responses to excitatory inputs and the ability of the neuron to integrate excitatory synaptic input, changes across a series of LFPs often reflect the summation of many postsynaptic responses by a group of BLA neurons. Inhibition recruited by PFC inputs to BLA in vivo can be reflected in its effects on action potential firing of BLA neurons and on the summation of LFPs. In adult rats, there is a substantial degree and duration of inhibition imposed over BLA neuronal firing and suppression of local field potential integration by PFC inputs to BLA (Selleck et al., 2018). In contrast, adolescents show less PFC regulation of BLA firing and local field potential summation during PFC train stimulation, indicating reduced influence of PFC over BLA (Selleck et al., 2018). This age dependency is likely driven by progressive increases in BLA inhibitory circuits from adolescence to adulthood, alongside stronger ability of PFC inputs to recruit these inhibitory circuits (Figure 2). Therefore, anatomical pruning of PFC inputs along with heightened ability to recruit BLA inhibitory circuits characterizes the transition from adolescent to adult BLA.

Continued maturation might not be unique to the basal and lateral nuclei of the BLA. The basomedial amygdala (BMA) is located directly ventral to BLA and is under substantial GABAergic influence in adulthood that may undergo similar maturational changes to control physiological responses (e.g., heart rate) via outputs to the hypothalamus (Mesquita et al., 2016). The BMA also receives inputs from the ventral, but not dorsal, PFC (Adhikari et al., 2015). These responses contain both excitatory and inhibitory components, with inhibitory components mediated by local GABAergic neurons (Adhikari et al., 2015). While developmental expression of ventral PFC-BMA projections has not been thoroughly studied, the process may be similar to other PFC-BLA projections, where GABAergic regulation of the BMA increases over the course of adolescence.

Prefrontal and amygdala contributions to social behavior during development

The key neural regions involved in social behaviors are distributed across the brain, and include several amygdala and extended amygdala regions, hypothalamic regions, PFC, nucleus accumbens, hippocampus, and sensory regions that play specific roles in aspects from detection to motivational, autonomic, and consummatory responses. As we will delineate below, the PFC and BLA are important components of the circuitry regulating these behaviors. BLA activity is important for a variety of social behaviors, including social learning and social interaction. These behaviors are mediated by changes in BLA excitation and inhibition, which is regulated by PFC inputs. Social behaviors comprise some of the most profound differences from adolescence to adulthood, characterized by a shift towards adult social drive and behavioral repertoire. Social interaction is a key metric for indicating social drive in both humans (Anderson et al., 1999; Blakemore, 2008) and rodents (Doremus-Fitzwater et al., 2009; Douglas et al., 2004; Kim et al., 2013). This measure can also be dissected into different types of interactions that represent distinct constructs at discrete developmental periods. Sociability (as measured with preference, duration, and type of interaction) is high during adolescence and this progressively declines into adulthood (Douglas et al., 2004; Li et al., 2019; Moy et al., 2004). Types of social behaviors can include play (e.g., chasing and pinning) and investigation (e.g., sniffing and grooming behaviors), with a greater degree of social play observed in adolescents and a shift towards more cautious investigation in adults (Figure 1B). Social play is a rough-and-tumble activity that may look like, but is distinct from, aggressive behaviors (often involving submission) in that it allows alternating positions between two rodents, and this behavior largely occurs prior to sexual maturation (Vanderschuren et al., 1997; 2016). Engagement in play behavior has rewarding value and is an essential part of social experience during adolescence that guides maturation towards interactions appropriate for adults (Vanderschuren et al., 1997). Complete absence of play opportunity by social isolation disrupts maturation of social behaviors, decreases social contacts, and produces aberrant behaviors in adulthood. Reciprocal engagement in social play is believed to be essential for maintained social drive into adulthood. Negative long-term effects of play deprivation can be mitigated by exposure to young rats that reciprocate play engagement, but not by older adult rats that do not reciprocate (Hol et al., 1999). During developmental periods of high play behavior, social interaction and engagement are also believed to be influenced by a drive to explore novel conspecifics (Kareem & Barnard, 1982; Terranova et al., 1999). Rodents form a social recognition memory, and display greater social interactions with novel conspecifics compared with cage mates, particularly in juvenile and adolescent rats between PND 28–40, (Smith et al., 2015).

This greater drive for a novel social partner may complement the role of play experience. Social play during adolescence is thought to provide exposure to new and unexpected circumstances that fosters behavioral flexibility (Panksepp, 1981). Activity in cortical regions, particularly the PFC, has been closely linked with cognitive flexibility, and this type of flexibility coincides with maturation of the cortex as well as cortical inputs to the amygdala. While several amygdala regions (like the basolateral, central medial amygdala), play an important role in regulating social behavior and memory (Christianson et al., 2010; Dumais et al., 2016; Lukas et al., 20), shifts in engagement of play and investigation behaviors correspond with changes in PFC and BLA neural circuits. Thus, play and novelty-seeking behaviors are more prevalent during adolescence and occur at a time of ongoing maturation of cortical inputs to the amygdala. PFC inputs may regulate these social behaviors by setting the balance between excitation and inhibition in the BLA and BMA. In line with this, decreased GABA function within the BLA and the PFC reduces sociability in adults (Paine et al., 2017) and inhibition of the PFC decreases frequency and duration of social play in adolescents (van Kerkhof et al., 2013). Play behavior can be enhanced with BLA cannabinoid manipulations that reduce GABAergic release (Piomelli et al., 2000; Trezza et al., 2012). Further, disinhibition of the BMA can reduce social novelty-dependent increases in heart rate (Mesquita et al., 2016; Trezza et al., 2012). These studies suggest that PFC inputs to BLA can drive adolescent social behavior, but too much or too little GABAergic regulation in amygdala can impair social play and responses to social novelty. As the PFC-amygdala circuit matures, the PFC produces more substantial inhibitory regulation of the BLA, and this may have a role in the transition away from social play behaviors towards more cautious investigative behaviors over the course of development. This increased ability of PFC to regulate BLA and BMA may also correspond with a role of this circuit in greater selectivity of adult-specific social behavior. In line with this, maturation of social behavior includes a redirection of social engagement. For example, increases in sex-specificity of social preference, as well as sexual and territorial behaviors arise over the course of development alongside decreased non-selective social engagement and decreased play, and these types of engagement may be differentially regulated by cortical-BLA maturation.

Social interactions are believed to have rewarding properties, particularly during adolescence. This has been measured with social place preference during which animals are allowed to explore chambers previously associated with a conspecific or a novel object. The time spent in a conspecific-paired chamber can indicate a greater propensity and motivation to engage in social behaviors (Varlinskaya & Spear, 2007), and subsequent preference for socially-paired chambers indicates the long-term effects and value of prior interactions (Peartree et al., 2012). As with duration of social interaction and social behaviors, there is a shift in this preference during development such that adolescents spend more time than adults in a chamber previously paired with a conspecific than a chamber paired with a neutral object (Douglas et al., 2004). While several interaction behaviors are present in adults and adolescents, engagement in play behavior is believed to be the best indicator for subsequent social place preference (Calcagnetti & Schechter, 1992; Peartree et al., 2012). This suggests the rewarding, positive valence of these interactions might be especially high during adolescence. In a similar social preference design, activity in the ACC region of the PFC is necessary to learn cues paired with social partners (Basile et al., 2020). Together with previous work, this suggests that local activity in the ACC may mediate aspects of social preference, and these preference behaviors change during development as the ACC-BLA pathway matures.

Isolation-driven changes in the PFC and BLA regulate social and nonsocial behaviors

Social interruptions that begin prior to adolescence, for instance maternal separation and post-weaning social isolation, have life-long effects. This type of isolation has considerable effects on the adolescent and adult social repertoire, fear and aggressive behavior, as well as maturation of neural systems (e.g., transmitters and peptides) (Lukas et al., 2010; Oliveira et al., 2019; Toth & Haller et al., 2011; Veenema et al., 2009). However, critical windows for social exposure have been identified during adolescence that promote cognitive flexibility and social ability in adulthood, and have been linked to psychiatric disorders following extensive deprivation (Fone & Porkess, 2008; Geyer et al., 1993). The impact of social interruption heavily depends on its duration during adolescence, as shorter duration interruptions are readily correctable, while longer duration interruptions can lead to significant impairments. Brief social isolation of several hours can have a strong, but transient influence on social interaction during development (Douglas et al., 2004; Hol et al., 1999; van den Berg et al., 1999). By analogy to other motivated drives, such as hunger and thirst, brief isolation can increase arousal or social drive through a homeostatic mechanism, where the degree of social interaction increases in order to minimize potential harmful effects of isolation (Hole, 1991; Matthews & Tye, 2019). Similarly, brief isolation (over the course of hours to a couple of weeks) typically increases or causes a rebound increase in social interaction and social place preference. This effect is dependent on reciprocation of social engagement by the partner (Douglas et al., 2004; Hole, 1991; Varlinskaya et al., 1999). The effects of brief social isolation are exaggerated in adolescents relative to adults, demonstrating an adolescent-specific sensitivity to changes in the social environment (Douglas et al., 2004; Varlinskaya & Spear, 2008) indicative of higher social drive. Further, deprivation of play behavior specifically, even without complete social isolation, increases later interaction during adolescence. For instance, social interaction was increased when mesh dividers were used to permit ongoing interaction but eliminate play behavior (Holloway & Suter, 2003). This indicates that social drive can be modulated just by limiting the nature of the social interaction, and that an element of the reciprocal opportunity of social behaviors is the target of social drive.

While brief deprivation can be readily compensated by subsequent increased interactions and preference, if prolonged isolation occurs during the period of heightened play (weeks 4–5), this can reduce frequency of social behavior in later adolescence (van den Berg et al., 1999; Figure 1B), and influence behavior and stress responses into adulthood. This includes reduced duration of social interaction, increased latency to submission following attacks by a dominant rat, and elevated physiological stress responses following social defeat in adulthood as a result of isolation during postnatal weeks 4 and 5 (Hol et al., 1999; van den Berg et al., 1999). The lasting effects of isolation during weeks 3 and 4 have also been shown to impair reversal learning and social interaction during adulthood (Han et al., 2011; Makinodan et al., 2017). However, re-socialization with socially-housed, but not previously isolated, conspecific following this two-week isolation can be sufficient to rescue decreases in social interaction (Makinodan et al., 2017). This suggests that isolation during this time period has lasting effects on interaction that can only be repaired by exposure to conspecifics with appropriately developed socially reciprocating repertoires. This behavioral effect was accompanied by partial rescue of immediate early gene expression in the PFC (Makinodan et al., 2017). Together, these results indicate that temporary isolation during a developmental period of heightened play causes preferential, long-term reductions in cognitive flexibility, social behaviors (Figure 1B), and responding to social stressors that may be mediated by changes in PFC activity.

Extended isolation rearing, occurring during the entire duration of adolescence, has been closely associated with reduced cognitive flexibility and heightened anxiety that persist throughout the lifespan. This has been demonstrated using several tasks, and impairments range from deficits in sensory-motor gating with prepulse inhibition (Geyer et al., 1993), reversal learning (Amitai et al., 2015), object recognition memory (Jones et al., 2010), time in center during open field testing (Hermes et al., 2011), reduced exploration of the elevated plus maze (Weiss et al., 2004), and increased aggressive behaviors (Oliveira et al., 2019). The effects of post-weaning isolation on reversal learning begin to emerge only four weeks after isolation begins when groups were tested during adolescence (Powell et al., 2015), demonstrating that lasting reductions in cognitive flexibility begin to emerge during adolescence. This long-duration social deprivation also reduces social interaction time in adults and impairs observational fear learning (a type of fear learning dependent on sensitivity to social influence; Hermes et al., 2011; Yusufishaq & Rosenkranz, 2013). Together, this work suggests that social exposure during adolescence is essential for the development of a range of behaviors, including social behavior, and cognitive functions including memory processes, and that heightened social drive during adolescence may have a protective role to ensure appropriate development of these functions.

The long-term effects of isolation are likely mediated by lasting changes in neural circuitry. PFC and BLA neuronal morphological changes are evident following post-weaning isolation (Wang et al., 2012), and inhibitory transmission within the PFC and BLA are altered as a result of post-weaning isolation. In the PFC, GABAergic function is reduced through increases in GAT-1, which decreases duration of GABA synaptic events and reduces tonic inhibition (Hickey et al., 2012). In several amygdala subnuclei, including the BLA, GAD67 is increased following isolation (Gilabert-Juan et al., 2012). Together, this indicates that increases in local PFC activity through hypo-GABAergic function may increase PFC input to the BLA. Because PFC input to the BLA can have an inhibitory influence, this may ultimately drive increases in inhibitory transmission and expression of GABAergic markers like GAD67. These processes can work together to decrease functional activity in the amygdala, which has been linked to decreases in social behavior. Isolation-dependent changes in inhibition also reduce maturation of BLA inputs to other amygdala subnuclei regulating social behaviors, like the medial amygdala (Adams & Rosenkranz, 2016; Twining et al., 2017), indicating that isolation-related changes within the BLA (possibly driven by PFC input) disrupt interconnectivity within the amygdala (Adams & Rosenkranz, 2016). These circuitry changes have a long-term influence on social behaviors. Furthermore, maturation of this circuit is dependent on social interaction, setting up a potential spiraling dysfunction whereby a reduction in social enrichment disrupts this circuitry, and circuitry disruptions further impair the ability to socially engage. In summary, adaptive systems are in place to enhance social engagement during adolescence by rewarding and reinforcing social interactions, through a motivated drive. This drive increases to compensate for social deprivations to promote better stress coping, cognitive flexibility, and social adeptness in adulthood. Key components of the circuitry that support this drive are sensitive to the social environment. While this review focuses on PFC-BLA circuits, it should be noted that the medial amygdala plays a key role in a wide range of social behaviors (Hong et al., 2014; Li et al., 2017; for recent reviews: Jennings & Lecea; Yang & Shah, 2016).

The need for social experience during development can have both positive and negative effects. For example, higher social drive can facilitate social experiences that promote appropriate neural circuit and behavioral maturation, but it can also introduce the potential for heightened sensitivity to social disruptions. While juvenile or adolescent isolation has robust and lasting effects (Cilia et al., 2001), isolation experiences during adulthood have little effect the magnitude of isolation-dependent changes in social interaction and preference are much smaller when compared to adolescents (Douglas et al., 2004; Varlinskaya & Spear, 2008). The contributions of the PFC-BLA pathway to this process are still relatively understudied. Increased capacity of PFC to regulate the BLA (Ferrara et al., 2020) may contribute to increased selectivity and refinement of social interactions in adults and seemingly decreased social motivation. The completed maturation of this circuit may also render adults less sensitive to brief or prolonged social isolation.

PFC-BLA circuitry and fear learning during adolescence and adulthood

Social function is not the only ability that matures during adolescence, and whose maturation is sensitive to the social environment. As described above, disruptions of the social environment during adolescence can lead to abnormal anxiety or other affective functions in parallel with abnormal social function. This shared sensitivity suggests that changes in similar neuronal circuitry may contribute to both social and affective abnormalities, but is there a similar developmental trajectory for anxiety, fear and related affective behavior, and are they reliant on maturation of PFC-BLA circuits similar to social function?

Developmental differences between adolescents and adults have been identified in Pavlovian fear conditioning, which involves the pairing of a neutral, conditional stimulus (CS+) and an aversive, unconditional stimulus (UCS), which evokes a fear response (e.g., Rescorla, 1966; see Ayres 1998, for a review). This responding can be compared to other stimuli explicitly not paired with a UCS (e.g., a CS−) using a differential fear paradigm. Typically, responding will be higher to the CS+ than the CS−, indicating good discrimination between the two (e.g., Ferrara et al., 2017; Trask et al., 2020). When tested for long-term fear retention, adolescents exhibit robust and inflexible fear responses to the CS+ in humans (Lau et al., 2011) and in studies that used both humans and rodents (Pattwell et al., 2012). This has been seen with weaker fear discrimination in adolescent mice (4–5 weeks), which show enhanced responding to a CS-following fear conditioning (Ito et al., 2009). This weaker adolescent discrimination between fear and safety cues has been associated with increased anxiety symptomology in humans (Haddad et al., 2015), suggesting that dysfunction in fear discrimination might be mediated by processes that also impact anxiety-like behavior.

Fear persistence during adolescence has also been clearly demonstrated by maintained fear responses following extinction learning, where a CS+ is repeatedly presented without the UCS. While behaviorally, extinction looks similar to “unlearning” in that fear expression to the CS+ declines in an inverse function of acquisition (e.g., Rescorla & Wagner, 1972), extinction involves the formation of a new inhibitory memory that is especially dependent on the context in which it is learned (Bouton & Bolles, 1979). Adolescents (PND 35) show a reduced ability to acquire this new inhibitory fear memory during extinction, and instead display persistent fear responding during an extinction test the following day (McCallum et al., 2010; Pattwell et al., 2012). This effect was not seen in their adult (PND 70) and preadolescent (PND 24) counterparts, suggesting a limited window of sensitivity (Kim et al., 2011; McCallum et al., 2010; Pattwell et al., 2012). The reduced ability to extinguish fear is not only during a transient adolescence period, but instead this inability extends to the late adolescent period (PND 47–48; Baker & Richardson, 2015). Differences in fear responding between adolescents and adults are in part a result of immature PFC function, which is essential for fear discrimination (Corches et al., 2019; Likhtik et al., 2014; Vieira et al., 2015) and extinction (Kwapis et al., 2012; Milad & Quirk, 2002). Activity at PFC synapses in the BLA is believed to regulate changes in fear expression, particularly after extinction, which can strengthen these synapses (Arruda-Carvalho & Clem, 2014; Cho et al., 2014). The ongoing maturation of local GABAergic processes and inputs (e.g., BLA-PFC) within the PFC undergo pruning and are integrated into PFC circuitry, and the increase in strength and inhibitory transmission at PFC inputs to the BLA underlie the ability to inhibit fear responding during development. In line with this, decreased expression of a plasticity marker, phosphorylated MAP kinase (MAPK), in the IL region of the PFC (an area crucially important for acquisition and expression of extinction learning; see Sotres-Bayon & Quirk, 2010) is seen in adolescents relative to both adult and preadolescent rodents (Kim et al., 2011). However, additional trials of extinction increased MAPK plasticity markers in adolescent IL and improved fear extinction retention. This suggests that adolescents have uniquely reduced capability on prefrontal-dependent tasks due to immature function of PFC neurons and subsequent PFC regulation of BLA responses, but this can be overcome with extended trials perhaps through the strengthening PFC-BLA connections.

Adolescent-specific differences in fear expression may be attributed to the time in which fear is learned or the developmental period that fear is retrieved. In a systematic study, Baker & Richardson (2015) trained and extinguished groups prior to, during, and following adolescence. If extinction occurred outside the adolescent window, extinction retention was higher. If conditioning and extinction occurred during adolescence extinction retention was poorer. However, if fear training occurred prior to adolescence, but extinction occurred during adolescence, extinction was similar to juvenile and adult groups. In both of these instances of extinction during adolescence, there was similar lower expression of phosphorylated MAPK in the PFC and BLA (Baker & Richardson, 2015), despite differences in extinction retention performance. This indicates that current PFC maturational state alone is not sufficient to explain extinction differences. Yet, extinction retention is similar in adults whether initial conditioning occurred in adolescence or adulthood. How can this be understood? Extinction is believed to require formation of a competing memory trace for behavioral suppression, and this new learning strengthens synaptic contacts between the PFC and BLA. These newly strengthened synapses can sufficiently suppress expression of an aversive fear memory. However, a memory acquired during adolescence may be more resistant to inhibition, combined with weaker extinction processes in adolescence related to weaker inhibition within the BLA, this can produce reduced acquisition of fear extinction. In support of an enduring adolescent fear memory, adolescent fear conditioning facilitates adult fear learning and interferes with the rate at which discrimination learning occurs (Muller et al., 2018). Because the PFC is essential for discrimination between cues and regulates BLA activity through GABAergic processes, the maturation of PFC inputs to the BLA may be an underlying factor for similar activity-dependent changes in PFC-BLA circuitry regardless of when fear was acquired. This pathway may not only guide fear learning and regulation within adolescence but can also alter the ability to learn in adulthood, demonstrating the lasting impact of adolescent experiences on later learning. While a memory acquired during adolescence will be resistant to inhibition through extinction until that developmental period has lapsed, adolescent fear learning even has lasting effects on later ability to discriminate fear and safety cues. These results add to a growing literature that demonstrate changes in fear learning are not necessarily linear throughout the lifetime (see Hartley & Lee, 2015, for a review) and pinpoint adolescence as a uniquely sensitive period for alterations in this type of learning, likely influenced by maturation of a PFC-BLA network. This points to a developmental trajectory for these affective behaviors that is similar to social function, and likewise is sensitive to maturation of a PFC-BLA network.

Interaction of social and fear behavior.

Broad changes in the social environment during adolescence can lead to enduring changes in affective behavior, as described in a prior section. However, social environment and affective behaviors can have a rapid impact on each other. This is exemplified by fear responses that can be learned through observation, an innately social form of learning (i.e., social fear learning). Social fear learning therefore represents one way in which we can examine the interaction of social drive and fear responding. During the typical social fear learning procedure, two rodents are placed in a fear conditioning chamber and separated by a mesh wall to allow for social interaction and observation of fear responses (Jeon et al., 2010; Twining et al., 2017). Social interaction is typically permitted for several minutes prior to the presentation of CS-UCS pairings to the demonstrator rat. The CS-UCS pairings result in an increase in fear responding (freezing). The observer rat never directly experienced the UCS but still acquires a freezing response during training, and this response is displayed at a long-term retention test when the CS is presented (Jeon et al., 2010; Twining et al., 2017; Yusufishaq & Rosenkranz 2013). Mimicking behaviors have been seen during social fear learning retrieval when an observer and demonstrator are tested together, with the most common response being freezing. However, when a demonstrator rat exhibits a different conditional fear response (e.g., hyperlocomotion) during retrieval sessions, this behavior is also quickly seen in observer (Allsop et al., 2018). The cues in the environment as well as the ability to engage in social interaction facilitate not only the observer’s ability to learn the fear response but also the type of response, indicating this learning is dependent on social inference (Jeon et al., 2010; Twining et al., 2017). In support of this, prolonged post-weaning social isolation, which results in reduced social interaction and social drive, impairs the ability to socially learn fear (Yusufishaq & Rosenkranz, 2013). This reduction in fear likely occurs through alterations in social processing, as the time oriented to the demonstrator rat was reduced in isolated rats (Yusufishaq & Rosenkranz, 2013). Interestingly, post-weaning isolation had no effect on directly experienced fear, suggesting a relatively social-specific change in learning during adulthood as a result of social deprivation during adolescence.

The effects of prolonged post-weaning social isolation on social fear learning provides further evidence for a developmental time window wherein the circuitry for social and affective behaviors is vulnerable. Intra-amygdala communication between the lateral and medial amygdala is essential for social fear in adults, (Twining et al., 2017), and post-weaning social isolation that overlaps with adolescence disrupts the maturation of medial amygdala and impairs the input from lateral to medial amygdala (Adams and Rosenkranz, 2016). This indicates that amygdala systems may be sensitive to changes to the social environment during adolescence and subsequently impact the ability to socially learn fear responses in adulthood (Twining et al., 2017). If the vulnerable circuitry involved PFC-BLA, one would anticipate that this circuitry is important in normal social fear learning. In adults, ongoing activity in the ACC and BLA is required for social fear learning memory formation, as synchronized activity between the ACC and BLA occurs during social fear learning and increases in immediate early gene (c-fos) expression are seen in the ACC and BLA as a result of social fear learning (Ito et al., 2015; Jeon et a., 2010), suggesting the coordinated activity between the ACC and BLA is required for fear learning through social inference. Further, this pathway may be specific to social fear learning as activity at ACC-BLA synapses enables CS encoding during social fear learning for subsequent fear recall, but activity in this pathway is not essential for directly experienced fear memory formation (Allsop et al., 2018). The mechanism through which the ACC promotes BLA changes is in part through glutamatergic receptors. NMDA-mediated transmission at ACC-BLA synapses increases relative to AMPA-mediated transmission following social fear learning, which was interpreted in those conditions to reflect formation of new silent synapses (Ito et al., 2015). The increase NMDA relative to AMPA currents is believed to prime synapses for the facilitation of future experience-dependent activity. In line with this, when passive avoidance training occurs one day following social fear learning, this learning and retention is enhanced (Ito et al., 2015). This suggests that the ACC can drive BLA activity, particularly during social fear learning, to promote social fear retention and other learning through an increase in NMDA-mediated transmission.

Social fear learning is highly dependent on the same socially motivated behaviors that change during development. PFC-BLA circuits contribute to social motivation, and developmental changes in social motivation may be mediated by PFC-BLA maturation. It is expected that adolescents will have differences in social fear learning. However, our current understanding of adolescent social fear learning performance and how social fear memories are encoded is unclear. It is possible that underdeveloped ACC inputs to the BLA may interfere with the ability to learn through social inference in adolescents, similar to weak PFC-BLA inputs that underlie immature extinction. It is also possible that the high social drive in this age group may instead preferentially promote social fear learning, as the degree of social interaction is directly correlated with the ability to learn and retain a social fear memory.

While social fear learning involves the observation of fear learning, it is possible to alter fear responding in trained and untrained groups by social interaction during fear testing, where the CS in the absence of the UCS is presented. These sessions can be completed with or without mesh barriers between a fear conditioned and unconditioned rat. The addition of a counterpart separated by a mesh barrier during testing often reduces fear in the conditioned group, referred to as social buffering (e.g., Kiyokawa & Takeuchi, 2017), but increases fear in unconditioned groups when allowed to directly interact at a test as seen in fear conditioning-by-proxy paradigms (Bruckey et al., 2010). Specifically, in fear conditioning-by-proxy, it is believed that untrained rodents develop fear responses by their investigative behavior (e.g., anogenital sniffing) of conditioned rodents during CS-elicited fear responding during the testing session (Jones et al., 2018; Sterley et al., 2018). Because ACC-BLA activity drives social fear behavior, this pathway may be essential to learn fear during fear conditioning-by-proxy procedures. While it is unclear how the ACC contributes to social buffering in adults, it is reasonable to propose that ACC-BLA activity drives socially learned fear responses during fear conditioning-by-proxy through increased social interaction and subsequent detection of conspecific behavioral, endocrine, and/or pheromonal fear and stress responses (Sterley et al., 2018).

Alternatively, when a subject who has not received conditioning is present and visible at test but separated by a divider from the conditioned subject, this often decreases fear responding as well as BLA activity in previously conditioned adults (Fuzzo et al., 2015; Kiyokawa & Takeuchi, 2017; Kiyokawa et al., 2007). This social buffering of fear may be related to increased activity (marked by increases in c-fos) in the IL region of the PFC as well as the basal nucleus of the amygdala (Kiyokawa et al., 2007). This suggests that social buffering recruits other PFC-BLA activity that is involved in suppression of fear responses (Kiyokawa et al., 2007). It may be expected that suppression of fear behavior following learning would be less sensitive to social buffering in adolescents, which relies on similar neural circuitry as extinction processes. However, it appears that some forms of social buffering of fear can be robust in infants, as maternal presence elicits a unique insensitivity to dangerous stimuli during infancy (Opendak et al, 2019; Sanchez et al, 2015). This suggests that while non-maternal social buffering may be reduced in adolescents, maternal presence may mitigate the inability to reduce fear through buffering leading to stronger fear suppression. Overall, this evidence suggests that a PFC-BLA circuit is integral in social buffering, and leads to a prediction that disruptions in the social environment during development that disrupt maturation of PFC-BLA circuits would lead to deficits in social buffering. While intuitive, this has not yet been systematically studied.

In social fear learning and fear conditioning-by-proxy paradigms, if observers had prior shock exposure they show greater fear responses during social conditioning compared to rats that were naïve to shock (Allsop et al., 2018; Kim et al., 2010). This fear response was directly correlated with 22khz USV emission recorded during a long-term test (Kim et al., 2010). The authors suggested that this prior shock exposure that elicits a 22khz USV creates an internal CS representation. At a subsequent test, the 22khz USV emitted by a demonstrator rat may then result in active retrieval of the observer’s prior fear memory. That is, the 22khz USV now serves as a CS to elicit a fear response on its own. Thus, fear by-proxy would be acquired through a mechanism similar to second-order conditioning (e.g., Helmstetter & Fanselow, 1989). While communication via USV is social in nature, these results may instead suggest that observers exposed to shock prior to social fear learning do not learn fear responses based purely on social inference, but instead acquire fear based on the retrieval of a previous directly-learned fear experience.

An alternative explanation is that prior learning experiences facilitate subsequent learning. As stated above, several preparations examining social transmission of fear rely on prior experience with shock before social fear learning or pair-testing in fear-conditioning-by-proxy paradigms (Allsop et al., 2018; Kim et al., 2010). Shock exposures prior to fear conditioning-by-proxy or social fear learning are similar to contextual fear learning (where the context in which conditioning occurs is the most accurate predictor of the UCS; Anagnostaras et al, 2001). This may allow a previously formed memory to be retrieved and modified with new information (i.e., from shock-USV to CS-shock-USV) that is predictive of an aversive event. Measurement of fear based on these experiences may be inherently different than fear learned through social inference, particularly with ACC contributions to this learning. Indeed, the ACC seems especially important in uncertainty and attentional processes during learning, and may not be entirely specific to social components of learned fear (Kim et al., 2016; Weible, 2013). The ongoing maturation of PFC inputs, including the ACC, indicates that potential developmental differences in social fear acquisition could be a result of maturation in this circuit as well as differences in USV between age groups.

Effects of stress on PFC regulation of BLA activity during development

At a similar time of high social interaction and low fear inhibition, adolescents are highly vulnerable to stressors and the effects of stressors on social and fear behaviors. Adults and adolescents display similar increased anxiety-like behaviors (e.g., changes in elevated plus maze behavior) as well as increased adrenal gland weight after repeated stress (Zhang & Rosenkranz, 2012). Despite behavioral similarities between ages, stress exposure differentially impacts cortico-amygdala circuits in adolescents and adults.

Following repeated stress exposure, in vivo BLA neuronal firing activity is elevated in adults, while adolescents show increases in the number of active neurons encountered (Zhang & Rosenkranz, 2012). There is a variety of ways that neuronal firing can be increased. Changes in BLA excitatory neuronal intrinsic excitability, excitatory or inhibitory synaptic inputs, and responses to specific PFC inputs are all potential sources for modification of BLA firing or functioning of cortico-amygdala circuits. Chronic restraint stress increases neuronal membrane excitability in adolescents throughout the BLA (Hetzel & Rosenkranz, 2014), in a similar manner to what is seen in basal nucleus of the BLA after post-weaning social isolation (Rau et al, 2015). While adults also show increased BLA excitability, this was preferential for the lateral nucleus of the BLA (Hetzel & Rosenkranz, 2014; Rosenkranz et al., 2010). Much of this increased excitability has been attributed to differences in the after-hyperpolarization (AHP; Hetzel & Rosenkranz, 2014; Rosenkranz et al., 2010). The AHP is a hyperpolarization of membrane potential across several time domains (fast, medium, and slow) that potently suppresses repeated neuronal firing and can also reduce integration of synaptic inputs (Faber & Sah, 2002; Power et al, 2011; Womble & Moises, 1993). In BLA neurons, the medium and slow AHPs are preferentially reduced after repeated stress, which increases BLA neuronal excitability and can increase the neuronal response to excitatory synaptic inputs. Targeting AHPs can partly repair effects of repeated stress (Atchley et al., 2012; Rosenkranz et al., 2010), suggesting that this contributes to stress-related increases in BLA activity.

The increases in BLA activity as a result of chronic stress exposure in adults is also accompanied by increases in excitatory synaptic inputs, and their ability to drive BLA neuronal firing. This has been seen using in vivo intracellular recordings demonstrating increases in the frequency of spontaneous EPSPs directly correlated with the number of spines in adults (Padival et al., 2013) and several studies that demonstrate dendritic hypertrophy and increased spines (site of most glutamatergic inputs to BLA principal neurons; Padival et al., 2013; Rosenkranz et al., 2010; Vyas et al., 2002; 2004). Furthermore, adults show increases in mEPSC frequency and increased responsiveness to glutamate after stress (Liu et al., 2020; Zhang & Rosenkranz, 2016). In contrast, there is less evidence for increased excitatory input in adolescents after repeated stress, and there may even be a decrease of excitatory inputs (Padival et al., 2013). Instead, adolescents show reduced inhibitory influences, such as lower mIPSC frequency, that is attributed to impairment in GABAergic processes (Zhang & Rosenkranz, 2016). Overall, evidence suggests that stress causes a greater reduction in GABAergic inhibitory influences in adolescents compared to adults, but increases glutamatergic excitatory influences in adults (Liu et al., 2020; Zhang & Rosenkranz, 2016).

Stress significantly alters function of cortical regions (reviewed in Arnsten, 2015). As described above, PFC inputs to the BLA have mixed excitatory and inhibitory actions that allow regulation of BLA neurons. Chronic stress increases excitatory components of PFC inputs in adults (Liu et al., 2020), an effect selective to PFC input in the BLA and not in reciprocally connected PFC and BLA neurons (Liu et al., 2020). In support of this, repeated stress can increase the in vivo excitatory impact of PFC inputs to BLA in adults (Figure 3), but in adolescents it more dramatically reduces the inhibitory effects in BLA recruited by PFC inputs (Figure 4), in contrast to effects in control animals (Selleck et al., 2018). Blockade of the major inhibitory source in the BLA (GABA_A receptors) further confirmed that effects of stress on PFC inputs to BLA are more directly related to increased excitation in adults, but decreased inhibition in adolescents (Figure 5). Taken in this context, along with previous literature, this suggests that stress may decrease the ability for the PFC to suppress BLA activity, which is already weaker in adolescents.

Figure 3. Repeated stress influence on excitatory PFC➜BLA.

The excitatory effects of PFC inputs to BLA can be measured and compared. In adolescents, repeated stress has relatively small effect on excitatory responses of BLA neurons to PFC stimulation [A, left; measured as an input-output response probability to single pulse stimulation of PFC (0.2 ms duration), across a range of stimulation intensities (0–0.9 mA, as Zhang and Rosenkranz, 2016), main effect of stress p>0.05, F(1,63)=3.74; n=32–33 neurons/group, 2-way RM-ANOVA], but increased excitatory responses of BLA neurons to PFC stimulation in adults (A, right; main effect of stress p=0.0013, F(1,53)=11.54; n=27–28 neurons/group, 2-way RM-ANOVA). This can be understood as stronger PFC inputs that drive BLA activity in adult rats after stress, perhaps contributing to overproduction of fear or anxiety (B).

Figure 4. Repeated stress influence on inhibitory PFC➜BLA.

The inhibitory effects of PFC stimulation on BLA neurons can be observed as suppression of BLA neuron firing (as Figure 2), and quantified as a changed from normalized baseline across a range of PFC stimulation intensities (0.2 ms duration single pulse, 0–0.9 mA). Repeated stress reduced the inhibitory effect of PFC stimulation on BLA neuron firing in adolescents (A, left; quantified as a percent change in firing rate compared to baseline firing, A, middle). Stress reduced the inhibitory influence of PFC stimulation on neurons of the BLA, and this was specific for responses measured from the basal nucleus of the BLA (A, right; basal nucleus main effect of stress p=0.0029, F(1,26)=10.82, average −31.5% change; lateral nucleus main effect of stress p>0.05, F(1,27)=0.15, average −1.3% change; n=13–15 neurons/groups). Inhibitory effects of PFC stimulation can be observed in adults as well (B, left), and can potently suppress BLA neuron firing (B, left, middle). Repeated stress moderately decreased the inhibitory effects of PFC stimulation on BLA neuron firing in adult rats (B, right; basal nucleus main effect of stress p =0.012, F(1,29)=7.22, average −18.3% change; lateral nucleus main effect of stress p=0.032, F(1,24)=5.184, average −20.8% change; n=13–16 neurons/groups). These effects can be understood as a reduction in already weak inhibition over BLA in adolescents after stress, leading to less regulation of BLA action potential firing (C, left), and a moderate reduction of inhibition in adults after stress, where a dashed box indicates an immature PFC in adolescence and a solid PFC box indicates the PFC has reached maturity in adulthood (C, right). This can lead to dysregulated activity of BLA neurons after stress in adults and adolescents.

Figure 5. Stress effects the type of response exhibited by BLA neurons.

PFC stimulation can produce excitatory or inhibitory responses in BLA neurons. Proportion of neurons that show excitation or inhibition can be assessed across age. During adolescence, there is also a substantial proportion of neurons that do not respond at all to PFC stimulation. Repeated stress decreases in the proportion of neurons that show an inhibitory response to PFC stimulation (A; by almost 30%, left, bar graph to present data numerically, right, waffle chart that also includes neurons that showed ‘no response’), with a smaller increase in the proportion of neurons excited. Adult rats have a relatively larger proportion of neurons that are inhibited by PFC stimulation. Repeated stress causes a substantial increase in the proportion of neurons that exhibit an excitatory response to PFC stimulation (B; almost 35%), and a relatively smaller decrease in the proportion of neurons that show an inhibitory response.

Acute stressors also alter BLA excitation and inhibition as well as PFC-BLA processes. For example, 30-minutes on an elevated platform blocks LTP in the BLA-PFC pathway (Maroun & Richter-Levin, 2003), and similar to chronic stress, brief restraint stress increases local glutamate in the BLA and increases BLA hypertrophy in adults (Maroun et al., 2013; Reznikov et al., 2007; Vyas et al., 2004; Yasmin et al., 2016; 2020). Brief experience on an elevated platform reduces phosphorylation of the GluA1 AMPA receptor subunit in the PFC and increases it in the BLA, while expression of GluA2 phosphorylation decreases in the BLA (Caudal et al., 2010). Because GluA2 are subunits of calcium non-permeable AMPA receptors, this suggests increased calcium-related plasticity in the BLA but not PFC following acute stressors. While this work demonstrates that glutamatergic synapses can be a target for the effects of acute or chronic stress in adults, there is substantially less known about effects of acute stress on these processes in adolescence.

Stress effects on fear and social behaviors during development

Stress during adolescence has long-lasting effects on fear learning and anxiety well into adulthood (see Romeo & McEwen, 2006, for a review), suggesting that stress experiences produce long-lasting changes in the neural circuitry that supports fear learning. This has been seen as increased anxiety-like behavior in adults and adolescents after periods of stress (Lovelock & Deak, 2019; Lukkes et al., 2009; Gresack et al., 2010; McIntosh et al., 2013; Skelly et al., 2015; Tao et al., 2017). However, in some instances, adolescent stress experiences have been reported to result in stress buffering (inoculation) later in life. For example, adolescent stress exposure leads to decreased physiological response (corticosterone release) to acute stress in adulthood (Holliday et al., 2020), and chronic social instability stress during adolescence suppresses both cued and contextual fear memory recall when measured in young adulthood (Morrissey et al., 2011). This highlights that the nature of adolescent stress experience and the measures used to assess its effect need to be considered when evaluating the presumed harmful effects of adolescent stress.

Moderate repeated stress has mixed effects on fear conditioning that depend on type of stress, and whether fear acquisition, consolidation, recall/expression or acquisition and retention of extinction is assessed. Repeated stress in adults usually enhances expression of cued conditioned fear (Atchley et al., 2012; Conrad et al., 1999). However, the effect of stress during adolescence is variable when fear conditioning is performed as adults, where enhanced fear (Yee et al., 2012) or no effect (Garcia et al., 2008; Toledo-Rodriguez & Sandi, 2007) has been reported. Stress during adolescence, however, can increase conditioned fear acquired during adolescence (Zhang & Rosenkranz, 2013), with the effects of stress on acquisition of extinction dependent on age. Namely, there is mixed evidence for slower acquisition of extinction in adults that experience repeated stress (Garcia et al., 2008; Miracle et al., 2006; Zhang & Rosenkranz, 2013), although the evidence is slightly more consistent with slower extinction or worse recall of extinction when exposed to stress during adolescence (Zhang & Rosenkranz, 2013). However, this might not be uniform across sex (Blume et al., 2019; McCormick et al., 2013). While the focus here is on moderate levels of stress, severe stressors that share features with traumatic stress can exert significant effects on fear extinction (for reviews, see Maren & Holmes, 2016; Wellman & Moench, 2019).

Stress-enhanced fear learning (Rau et al., 2005) has been recently used to study the effects of stressful experience on subsequent fear learning. Using stress-enhanced fear learning, animals are first exposed to a traumatic experience (i.e., 15 1-s 1mA unsignalled footshocks) before receiving contextual fear conditioning (using a single context-shock pairing) in a novel context. Prior acute (Pennington et al., 2017; Rau & Fanselow, 2005) or chronic (Hoffman et al., 2015) experience with stress resulted in a marked increase in freezing to the conditioning context relative to control animals that received equivalent exposure to the chamber without the shocks. This heightened response is resistant to immediate massed extinction occurring 10 minutes or 72 hours following initial learning (Long & Fanselow, 2012). Further, enhanced fear conditioning to the novel context persisted even when responding to the stress context was reduced through extinction (i.e., exposure to the chamber without presentation of the shock; see also Hassien et al., 2020). Interestingly, certain effects of stress-enhanced fear learning (i.e., increased later fear conditioning and unconditional fear to a loud noise stimulus) seem dependent on non-associative mechanisms, whereas others (i.e., reduced open field exploration) seem dependent on associative mechanisms (Hassien et al., 2020). This suggests that some behavioral effects of stress-enhanced fear learning are reflective of generalized fear responding rather than changes caused by stress or sensitization per se.

Adolescent stress not only influences fear and anxiety but also extends to social behaviors (Oehler et al., 1986). Chronic restraint stress impairs social interaction and preference in both adults and adolescents (Doremus-Fitzwater et al., 2009), and when stress occurs prenatally, there are lasting effects on social behaviors. Namely, peri-adolescent social play behavior is reduced, and this prenatal stress exacerbates physiological stress responses to restraint stress (Morley-Fletcher et al., 2003). Post-weaning social isolation impairs social fear learning and reduces social interaction, while leaving cued fear learned through direct shock exposure unaffected (Yusufishaq & Rosenkranz, 2013). Further, intermittent physical stress exposure decreases social interaction (Makinodan et al., 2012; Tao et al., 2017; Yusufishaq & Rosenkranz, 2013). This pinpoints adolescence as a period of heightened sensitivity to social disruptions, and in particular social stressors when compared to adults (Vidal et al., 2007; 2010). Taken together, this frames a potential spiraling effect, wherein stressors can have a larger impact in adolescence that disrupts social behaviors, that leads to further impairments.

The impact of stress on fear learning is dependent on a distributed system throughout the brain, including the BLA (Hoffman et al., 2015; Ponomarev et al., 2010) and the PFC (Negrón-Oyarzo et al., 2014), where reduced influence of the PFC on BLA neuronal activity and regulation of the BLA has been observed. Effects of chronic stress on extinction learning in adolescents in particular have been attributed to abnormal maturation of the PFC and its output (Makinodan et al., 2012; Negrón-Oyarzo et al., 2014). However, stress exposures can result in a similar resistance to extinction in adults in some instances (Woon et al., 2020), demonstrating that maturational susceptibility during adolescence may combine with other stress-induced impairments in cortico-amygdala function (Makinodan et al., 2012; Pattwell et al., 2012; Tottenham & Galván, 2016).

Maturation of PFC-amygdala circuits mediates sensitivity to stressors and the developmental shifts in social drive and fear

In summary, we have highlighted a role for maturation of PFC inputs in the BLA in the refinement of fear and social behaviors from adolescence to adulthood. We present evidence for a role of PFC-BLA circuitry in the developmental changes in fear and social behaviors during adolescence through a shift in the balance between GABA and glutamatergic processes. Excitation from the PFC, which can be seen from the PL, IL and ACC, to BLA primary neurons can facilitate social interaction and fear and other affective behaviors. While both excitatory and inhibitory circuits are maturing during adolescence, PFC inputs (ACC, PL, and IL) to the BLA gain a greater ability to recruit intra-BLA inhibitory networks by adulthood. A PFC-BLA circuit, largely comprised of the PL and IL, can drive or select the appropriate affective response via excitatory influences, and can also produce inhibition over the BLA to regulate fear and social behaviors. Although not an absolute distinction, PL often has a prevalent role in selection and driving of appropriate BLA-mediated responses while IL has a key role in suppression of behavioral responding during extinction. The ACC plays a major role in social-specific behaviors, suggesting that these inputs in particular play a more specific role than the PL and IL in social behavior. The increase in fear and social behavior during adolescence followed by behavioral refinement during the transition to adulthood parallels the developing cortico-amygdala circuits that refine BLA activity in adulthood and may be largely characterized by increased PFC inhibitory regulation over the BLA. The inhibitory PFC-BLA circuit that gains strength during adulthood refines social and fear behavior, resulting in more selective social interaction and increased ability to inhibit fear responses. The PFC-BLA circuits are highly sensitive to the social environment, and socially-driven plasticity at PFC-BLA synapses during the adolescent play period can be essential for aspects of social and cognitive development, and coping strategies to stressors much later in life. Exposure to stressors during adolescence has particularly harmful effects on later behavior, and this can produce several parallel affective and social abnormalities. This dual deficit may arise due to overlapping PFC-BLA circuits that are involved in social and affective behaviors, and their sensitivity to the harmful effects of stressors during adolescence. This can lead to deficient maturation of PFC-BLA circuits that no longer appropriately guide behavior due to an inability to appropriately regulate GABAergic processes in the BLA. Overall, this places maturation of cortical inputs to the BLA as a major target of pathology in guiding alterations in social behaviors and ability to control fear responses (Figure 2).

Therapeutic relevance

Depression and anxiety disorders, as well as specific symptoms across many other disorders, involve abnormality in cortico-amygdala circuits. These same circuits are sensitive to stress in humans, and stress is a factor in depression and anxiety both of which are characterized by abnormal social and fear or anxiety. Furthermore, stress during adolescence is a risk factor for later psychiatric disorders, likely by impacting overlapping cortico-amygdala circuits. Adolescence is a critical period for development, with a wide range of changes in functional organization associated with puberty, and sensitivity to stress exposure and the social environment. Because adolescence is a period of heightened stress reactivity, with experiences during this stage creating long-lasting changes in the neural circuitry underlying social and fear behaviors, targeting treatments towards adolescents might not only serve to reduce symptoms during this time but will also likely reduce the development of pathologies that stem from abnormal fear learning and social repertoire later in life. As social behavior during adolescence is affected by several classes of drugs of abuse during this time (Trezza et al., 2014), future work will need to disentangle how these drug-induced changes in social behavior affect fear responding later in life. Several of the findings discussed extend to human work, as many elements of fear and social circuits, and many effects of stress and social disruptions on fear and social behaviors are similar across species (Gewirtz & Baer, 1958; Glen et al., 2012; Lee et al., 2016; Meyer et al., 2017; Romeo, 2013). Further, experiences that result in abnormal fear and social behaviors during adolescent development have been implicated in schizophrenia, anxiety, depression, and autism (Barak & Feng, 2016; Meyer-Lindenberg et al., 2008; Jalbrzikowski et al., 2013). A better understanding of these behaviors, sensitivity to the social environment and stress, and the circuitry mediating these processes may provide valuable insight for targeted therapies.

Highlights.

• Development is characterized by changes in fear learning and social behavior

• Adolescence is a period of heightened sensitivity to stressors

• Behavioral changes during development coincide with cortico-amygdala maturation

• Cortico-amygdala maturation may underlie stressor sensitivity that impacts behavior