Prefrontal cortical trkB, glucocorticoids, and their interactions in stress and developmental contexts

Abstract

The tyrosine receptor kinase B (trkB) and glucocorticoid receptor (GR) regulate neuron structure and function and the hormonal stress response. Meanwhile, disruption of trkB and GR activity (e.g., by chronic stress) can perturb neuronal morphology in cortico-limbic regions implicated in stressor-related illnesses like depression. Further, several of the short- and long-term neurobehavioral consequences of stress depend on the developmental timing and context of stressor exposure. We review how the levels and activities of trkB and GR in the prefrontal cortex (PFC) change during development, interact, are modulated by stress, and are implicated in depression. We review evidence that trkB- and GR-mediated signaling events impact the density and morphology of dendritic spines, the primary sites of excitatory synapses in the brain, highlighting effects in adolescents when possible. Finally, we review the role of neurotrophin and glucocorticoid systems in stress-related metaplasticity. We argue that better understanding the long-term effects of developmental stressors on PFC trkB, GR, and related factors may yield insights into risk for chronic, remitting depression and related neuropsychiatric illnesses.

1. Introduction

Human cortical brain development is a non-linear process, with individual subregions maturing according to temporally distinct trajectories. The prefrontal cortex (PFC) is among the last brain regions to structurally mature (Giedd et al., 1999; Sowell et al., 2001; Gogtay et al., 2004) and undergoes significant synaptic remodeling during adolescence. Following an initial phase of spinogenesis and synaptogenesis during childhood and early adolescence is a protracted period of dendritic and synaptic pruning in the PFC that persists until young adulthood (Rakic et al., 1994; Huttenlocher, 1979; Anderson et al., 1995). Such dramatic structural modification may open a window of vulnerability to insults like stressor exposure or drugs of abuse. Indeed, some evidence suggests that stressor exposure during adolescence can impact long-term behavioral outcomes and psychiatric disease risk. For example, a stress-induced depressive episode in adolescence increases the risk of depression recurrence and treatment resistance throughout life (Thapar et al., 2012).

Adolescence is also a period of vulnerability to the development of neuropsychiatric illness; for example, up to 50% of “adult” psychiatric disorders, such as depression or schizophrenia, initially present during adolescence (Kessler et al., 2005, 2007), and many are triggered by stress. Understanding why adolescence is a period of vulnerability to the development of stress-induced psychopathology may be advanced by examining molecular mechanisms that control dendritic spine dynamics and neuroplasticity. This review examines the literature regarding developmental and stress-induced changes in receptors for two potent modulators of dendritic spines and synaptic plasticity – Brain-derived Neurotrophic Factor (BNDF) and glucocorticoids.

BDNF and signaling through its high-affinity tyrosine receptor kinase B (trkB) are involved in neural development and synaptic plasticity, are implicated in the etiology of major depression, and are differentially expressed and regulated throughout development and between brain regions (Bartkowska et al., 2010; Luberg et al., 2010). Cortisol, which binds to glucocorticoid receptors (GRs) and mineralocorticoid receptors (MRs), is a principal mediator of the stress response, and also regulates brain development and neuronal plasticity (McEwen, 2000; Liston & Gan, 2011). Prolonged exposure to high levels of glucocorticoids (as occurs with chronic stress) can have effects on neuronal architecture that are typically perceived as deleterious (e.g., Wellman, 2001; Radley et al., 2004, 2006; Anderson et al., 2016).

In the following sections, we review the structural maturation of the PFC during postnatal development (section 2) and the impact of stressor exposure on PFC neuronal morphology (section 3). Then we review the mechanisms by which trkB- and GR-mediated signaling events regulate dendritic spine densities and morphologies (sections 4 & 5), and how the expression and activity of trkB and GR in the PFC change across postnatal development (section 6), may be impaired in depression (section 7), and are modulated by stress (section 8). Finally, the role of BDNF/trkB and GR systems, and their interactions, in stress-related metaplasticity is reviewed in section 9. We conclude by discussing the implications of these findings for understanding the pathogenesis of depression (section 10).

2. Structural maturation of the PFC

Considerable evidence from longitudinal imaging studies indicates that the PFC is among the last brain regions to structurally mature (Giedd et al., 1999; Sowell et al., 2001; Gogtay et al., 2004). In the frontal cortex, gray matter volume increases during childhood and into adolescence, peaking at approximately age 11–12 (Lenroot & Giedd, 2006), then declines steadily until young adulthood (Giedd et al., 1999; Sowell et al., 1999; Gogtay et al., 2004). Moreover, this cortical gray matter loss progresses in a back-to-front fashion in the frontal lobe, ending with the PFC (Gogtay et al., 2004). Region-specific trajectories of gray matter maturation may be associated with synaptic pruning, increased myelination of intra-cortical axons, or changes in programmed cell death (Huttenlocher, 1979; Juraska & Markham, 2004; Markham et al., 2007). Postmortem histological findings indicate that the general pattern of synaptic changes in the human cortex is similar across regions, with an initial overproduction of synapses peaking in early childhood and synapse elimination continuing throughout adolescence (Huttenlocher, 1997). Nevertheless, the PFC shows the highest degree of dendritic spine proliferation, reaching spine densities in childhood that are twice as high in some areas as in adults (Petanjek et al., 2011). Rates of dendritic spine loss are also slowest in the PFC, with dendritic spine pruning continuing until the third decade of life (Petanjek et al., 2011).

This process of dendritic spine and synapse overproduction, followed by a prolonged period of pruning during adolescence, is conserved across mammalian species (e.g., Gourley et al., 2012a; Bourgeois et al., 1994; reviewed Shapiro et al., 2017a) and is thought to support species-typical adolescent behavior. These changes refine synaptic connections between subregions of the PFC and with other limbic and subcortical structures such as the hippocampus, amygdala, and basal ganglia (including by way of thalamic relays). Additionally, axons in the frontal cortex continue to be myelinated during adolescence (Barnea-Goraly et al., 2005; Sowell et al., 2003; Giedd et al., 1999). Synaptic overproduction, pruning, and myelination in the adolescent PFC are thought to enhance the efficiency of communication between regions, optimizing executive functions in the transition to adulthood. “Executive function” refers to a cluster of cognitive processes necessary for goal-directed decision making, working memory, attention, planning, and reasoning. Not surprisingly – considering the protracted structural maturation of the PFC – these cognitive skills continue to develop throughout adolescence (Blakemore & Choudhury, 2006; Conklin et al., 2007; Best & Miller, 2010).

3. Stressor exposure impacts neuronal morphology in the PFC

Abundant evidence indicates that chronic stress remodels PFC neuronal structure and induces deficits in PFC function. Rodent studies have focused almost exclusively on the medial prefrontal cortex (mPFC) – typically referring to the anterior cingulate, prelimbic (PL), and infralimbic (IL) cortices – in mature rodents. These investigations reliably report dendritic retraction on pyramidal neurons and reduced spine densities following chronic stress (see for discussion, Anderson et al., 2016). Stress-induced structural remodeling of mPFC neurons is associated with behavioral impairments in attentional set-shifting (Liston et al., 2006) and failures in goal-directed decision making (Dias-Ferreira et al., 2009). Chronic exposure to the primary stress hormone corticosterone (CORT; cortisol in humans) in adult rodents can also decrease spine densities on mPFC pyramidal neurons (Radley et al., 2008; Liu & Aghajanian, 2008; Anderson et al., 2016; but see Seib & Wellman, 2003). Exogenous CORT exposure also leads to dendritic remodeling, with investigations largely (but not always) revealing atrophy of apical arbors on pyramidal neurons (Wellman, 2001; Cerqueira et al., 2007; Liu & Aghajanian, 2008). Simplification of apical dendrites in the PL cortex has also been reported in adolescent male and female rats exposed to chronic stress beginning in pre-adolescence (Eiland et al., 2012).

Fewer studies have examined whether morphologic effects persist after stressor exposure ends. Of those that have, many focus on the mPFC (PL and/or anterior cingulate cortex), revealing somewhat mixed results. Multiple reports suggest that dendrites and dendritic spines can recover with 1–3 weeks of rest (Radley et al., 2005; Bloss et al., 2010, 2011; Moench & Wellman, 2017). In agreement, dendritic spine densities on pyramidal neurons in the IL cortex also recover with a 1-week washout period following excessive CORT exposure (Gourley et al., 2013). Long after elevated CORT, a “rebound” may occur – on layer V neurons, PL dendritic spine densities appear to increase as mice enter their fourth week of recovery following oral CORT (Swanson et al., 2013). By contrast, Anderson et al. (2016) reported that chronic exposure to CORT (via subcutaneous pellets) induced dendritic spine loss and shrinkage in the PL, which persisted for 3 full weeks. This discrepancy relative to Swanson et al. (2013) may be attributable to the inclusion of layer II/III neurons in the Anderson report (suggesting cell type-specific recovery patterns) or to the fact that subcutaneous CORT pellets disrupt diurnal glucocorticoid rhythmicity by clamping CORT at the peak circadian levels. Meanwhile, oral CORT exposure (via the drinking water) leaves circulating CORT at normal or even low levels during the daytime (when mice are sleeping and thus not consuming CORT) and high when they are awake and drinking (Gourley et al., 2008a; Barfield et al., 2017a). Further, it is important to note that single-time-point analyses cannot dissociate whether stress/CORT impacts dendritic spine formation or elimination rates or survival. For this, one requires in vivo two-photon microscopy, for example, which is considerably more labor intensive.

In another region, the orbitofrontal cortex (OFC), chronic stress induces hypertrophy of apical dendrites on layer II/III pyramidal neurons (Liston et al., 2006; Dias-Ferreira et al., 2009), but dendritic spine loss (Xu et al., 2016a) and depressive-like behaviors and decreased post-synaptic density-95 (PSD-95) and kalirin-7, proteins essential for spine function, stabilization, and maturation (Ehrlich et al., 2007; Ma et al., 2003). Gourley et al. (2013) also reported that chronic CORT reduces dendritic spine counts on neurons in deep-layer OFC and suggested that stress-induced dendrite elaboration in the OFC may be an adaptive response to stress.

In the context of postnatal development, adolescent CORT eliminates dendritic spines the PL cortex, and remaining spines are larger in volume, but not head size, suggesting aberrantly large necks (Barfield et al., 2017a). This spine dysmorphia persists for several weeks after CORT. Other groups report cortical spine loss and smaller head diameters with preadolescent (from postnatal day (P) 21–31) CORT, stress, and trkB inhibition (Arango-Lievano et al., 2015, 2016; see also spine loss in Jeanneteau et al., 2018). Stress-related head atrophy is also observed in the mature PL (Radley et al., 2013; Anderson et al., 2014), but we caution that dendritic spine density and head size do not necessarily positively co-vary. As an example, early-life (preadolescent) GR inactivation causes cortical spine loss and dendritic spine head enlargement (Arango-Lievano et al., 2015, 2016), a potentially adaptive response, given that large dendritic spine heads house synapses and are necessary for certain forms of learning and memory (Hayashi-Takagi et al., 2015).

Rats exposed to 5 days of footshock stress in pre-adolescence (P21–25) display despair-like behavior in the forced swim test (FST) in adulthood, as well as decreased cortical thickness in the PL and IL cortices and reduced dendritic spine densities and dendrite lengths on pyramidal neurons in the IL cortex (Lyttle et al., 2015). Notably, these morphological alterations are evident >11 weeks after the last footshock stressor, and are blocked by the selective serotonin reuptake inhibitor (SSRI), fluvoxamine. Fluvoxamine also blocked stress-induced depressive-like behavior. In line with these findings, social isolation during early adolescence (P30–35) reduces synaptophysin, a pre-synaptic marker, in the IL cortex of adult rats (Leussis et al., 2008). Synaptophysin loss was blocked by MK-801 (an NMDA receptor antagonist) or adinazolam (a benzodiazepine derivative) in adolescence (P40–55), suggesting that dampening glutamatergic activity may be an effective strategy in mitigating some of the enduring neurobehavioral effects of social adversity.

The OFC also appears to be vulnerable to long-term effects of stress or glucocorticoid exposure in adolescence. For example, chronic exposure to CORT in rodents’ drinking water during adolescence (P35–56) induces anhedonic-like behavior in adulthood and loss of dendritic spines in hippocampal CA1 and on deep-layer pyramidal neurons in the OFC and IL cortex (Gourley et al., 2013). Spine densities in the basal amygdala also increase (Gourley et al., 2013). However, following a 1-week washout period (when rodents are returned to regular drinking water following the cessation of CORT), dendritic spine densities in all regions normalized, except in the OFC. Thus, stress-related structural modifications of the OFC can persist beyond the period of stressor or glucocorticoid exposure, and may be associated with some of the long-lasting behavioral consequences of adverse experiences in adolescence.

The OFC is implicated in adolescent-emergent Major Depressive Disorder (MDD) and vulnerability to psychopathology following early-life adversity. For example, adolescents with MDD exhibit hypoactivity of the OFC during reward-related decision-making tasks (Shad et al., 2011), and early-life adversity is associated with reduced gray matter volume (Hanson et al., 2010; De Brito et al., 2013; Dannlowski et al., 2012) and cortical thickness (McLaughlin et al., 2014; Lim et al., 2017) in the OFC. And, decrements in OFC gray matter following childhood maltreatment may mediate vulnerability to depression (Edmiston et al., 2011). Ansell et al. (2012) found that an interaction between cumulative adverse events over the lifetime and greater subjective experience of chronic stress is associated with less gray matter volume in the OFC. Thus, stress-related changes in OFC volume may increase risk for developing psychopathology in the face of adverse life events.

Glucocorticoids and neurotrophins are key regulators of dendritic spine plasticity, and their dysregulation is associated with synapse loss and stress-related psychopathology. Meanwhile, the actin cytoskeleton forms the structural lattice supporting dendritic spines. Disrupting cytoskeletal dynamics can have profound effects on spine structure, communication between neurons, and ultimately, on behaviors, including those relevant to depression (Licznerski & Duman, 2013; Wong et al., 2013). The filamentous (F-) actin meshwork within dendritic spines is constantly being polymerized and depolymerized by an array of regulatory proteins. The long-term maintenance of spines requires a balance of actin polymerization and depolymerization. Synaptic scaffolding proteins and pre- and post-synaptic adhesion systems also help provide structural support for spines. Activity-dependent neurotrophin signaling is generally thought to strengthen and help maintain active synapses by activating spine stabilization pathways. These pathways are also targeted by glucocorticoids, which are necessary for spine development and maintenance (Liston et al., 2013; Haditsch et al., 2009; Shapiro et al., 2017b). However, chronic stress or excess glucocorticoid levels lead to dendritic spine loss, at least in part, by GR-mediated disruption of the molecular mechanisms that stabilize spines. These topics will be discussed in further detail in the next sections.

4. BDNF-TrkB regulation of dendritic spines

BDNF signaling through trkB is involved in the development, maintenance, and plasticity of synapses throughout life. The trkB receptor exists in two forms – a full-length form (designated trkB.FL) and a truncated form (designated trkB.T1). TrkB.FL is expressed on neurons, while TrkB.T1 is expressed on both neurons and glia (Ohira & Hayashi, 2009). The truncated receptor contains the same extracellular and transmembrane domains and initial 12 intracellular amino acid sequences as the full-length receptor, but lacks the tyrosine kinase intracellular domain (Middlemas et al., 1991). Upon ligand binding, the full-length trkB receptor dimerizes and is autophosphorylated at tyrosine residues in the intracellular domains, initiating mitogen-activated protein/extracellular signal-regulated kinase (MAPK/ERK), phosphatidylinositol-3-kinase (PI3K), and phospholipase C gamma (PLCγ) signaling cascades (Reichardt, 2006; Carvalho et al., 2008). Lacking the tyrosine kinase-containing domain, TrkB.T1 is unable to activate these signaling cascades; however, this isoform may have distinct functions via alternative signaling pathways (see for further discussion, Ohira & Hayashi, 2009).

TrkB.FL-mediated signaling regulates the structural and functional plasticity of dendritic spines through effects on gene expression, neurotransmitter release, trafficking of synaptic proteins, and activities of membrane receptors and actin cytoskeleton regulatory proteins. These effects are mediated by the MAPK/ERK, PI3K, and PLCγ signaling cascades, which regulate overlapping but also distinct molecular processes. These pathways have been extensively reviewed elsewhere and are beyond the scope of this review (Reichardt, 2006), but here we briefly discuss the pertinent findings.

Phosphorylation at Tyr817 (Tyr816 in mice) on trkB.FL leads to the recruitment and activation of PLCγ, which hydrolyses phosphatidyl inositides to generate diacylglyercerol (DAG) and inositol 1,4,5-trisphosphate (IP3). DAG activates protein kinase C (PKC), while IP3 stimulates the release of Ca2+ from intracellular stores, increasing the activation of Ca2+/calmodulin-dependent protein kinases and other Ca2+-regulated targets, such as adenylyl cyclase (AC) and the transcription factor cyclic adenosine monophosphate (cAMP) response element binding protein (CREB) (Minichiello, 2009; Yoshii & Constantine-Paton, 2007). TrkB signaling through PLCγ also activates the plasma membrane transient receptor potential canonical subfamily 3 (TRPC3) channel that generates sustained cationic currents at synapses (Amaral & Pozzo-Miller, 2007).

TrkB.FL phosphorylation at Tyr516 in humans (Tyr515 in mice) recruits and phosphorylates the adaptor protein, Src homologous and collagen-like protein (Shc). Activated Shc recruits a complex consisting of the adaptor Growth factor receptor-bound protein 2 (Grb2) and the Ras exchange factor son of sevenless (SOS), which activates Ras. Upon activation by Ras, the protein kinase Raf phosphorylates mitogen-activated/extracellular signal-regulated kinase 1/2 (MEK1/2), which subsequently phosphorylates ERK1/2 (English et al., 1999). ERK1/2 influences protein synthesis-dependent plasticity by activating CREB (Ying et al., 2002; Shaywitz & Greenberg, 1999) and enhancing signaling through mammalian target of rapamycin (mTOR), a kinase that regulates protein translation and long-term synaptic changes. Ras homologue enriched in brain (Rheb), an activator of mTOR, is normally inhibited by the tuberous sclerosis complex 1/2 (TSC1/2) of proteins (Dwyer & Duman, 2013). Activated ERK1/2 phosphorylates and inhibits TSC1/2, thereby activating mTOR.

Phosphorylation of Shc at Tyr516 of trkB.FL also leads to PI3K activation through Ras-dependent and -independent pathways. Ras-dependent activation of PI3K occurs via Shc/Grb2/SOS/Ras. Ras-independent activation involves Shc/Grb2 and subsequent recruitment of Grb2-associated binding protein 1 (Gab1), which binds and activates PI3K (Holgado-Madruga et al., 1997). Phosphoinositides generated by PI3K and phosphoinositide-dependent kinases cooperatively activate the protein kinase Akt (also called Protein kinase B). Akt enhances mTOR-mediated protein translation by phosphorylating TSC1/2. Signaling through PI3K/Akt also regulates the trafficking of synaptic proteins (Yoshii & Constantine-Paton, 2007).

BDNF and trkB.FL are expressed in both pre- and post-synaptic compartments of synapses. BDNF is rapidly released from both compartments in an activity-dependent manner (Hartmann et al., 2001; Kohara et al., 2001; Kojima et al., 2001). The activation of trkB.FL on pre-synaptic sites by BDNF released from post-synaptic sites leads to stabilization of the pre-synaptic site (Lin & Koleske, 2010). Likewise, activation of trkB.FL on post-synaptic sites by pre-synaptic BDNF release contributes to stabilization of the post-synaptic site. Moreover, rapid glutamate-dependent release of BDNF from post-synaptic sites can activate trkB.FL on that same spine (Harward et al., 2016). This autocrine BDNF-trkB.FL signaling mechanism is necessary for structural LTP. The mechanisms by which post-synaptic trkB.FL discriminates the source of BDNF is unclear, but may involve differences in the BDNF transcripts localized to the pre- and post-synaptic sites and differences in the storage and release of BDNF (e.g., vesicle type).

Activity-dependent secretion of BDNF is one mechanism by which active synapses are selectively modulated (Snider & Lichtman, 1996; Boulanger & Poo, 1999). However, because BDNF is a diffusible molecule, additional mechanisms to achieve synaptic specificity of activity-dependent plasticity likely exist. Indeed, one mechanism involves spine-autonomous BDNF-trkB signaling and coincident activation of 3 Rho GTPases (Hedrick et al., 2016). Another mechanism involves the second messenger, cAMP, which regulates synaptic responses to BDNF. Specifically, Ji and colleagues (2005) report that cAMP gates BDNF-induced trkB phosphorylation and enhances the mobilization of trkB into synapses by facilitating the association of trkB with the glutamate receptor scaffolding protein PSD-95. Enhanced localization of trkB at the post-synaptic density is thought to potentiate the response to BDNF (Meyer-Franke et al., 1998). Synaptic activity increases intracellular cAMP levels – by enhancing Ca2+ influx and activation of Ca2+-stimulated AC (Wang et al., 2003) or through activation of AC by neurotransmitter-stimulated G protein-coupled receptor (GPCR) signaling. Thus, activity/cAMP-dependent recruitment of trkB to stimulated synapses and activity-dependent release of BDNF selectively potentiates BDNF-trkB signaling at active synapses.

BDNF-trkB-mediated signaling is involved in the induction, maintenance, and consolidation phases of synaptic long-term potentiation (LTP; Park & Poo, 2013; Panja & Bramham, 2014), a critical component of activity-dependent synaptic strengthening. For example, BDNF activation of trkB enhances pre-synaptic glutamate release (Gottschalk et al., 1998; Pereira et al., 2006; Matsumoto et al., 2006) by trkB-MAPK/ERK-mediated phosphorylation of the synaptic vesicle protein synapsin I, which modifies the interaction of synaptic vesicles with the actin cytoskeleton to facilitate exocytosis and neurotransmitter release (Jovanovic et al., 1996, 2000). Enhancement of glutamate release by BDNF also occurs via trkB-PLCγ-induced release of Ca2+ from pre-synaptic intracellular stores, which increases the number of docked synaptic vesicles for exocytosis (Numakawa et al., 2002, 2009).

BDNF-trkB is also involved in post-synaptic mechanisms of LTP. Several studies have shown that BDNF modulates the activity and trafficking of ionotropic glutamate receptors (Caldeira et al., 2007a,b). For example, BDNF induces translocation of the GluA1 subunit of AMPA receptors (Nakata & Nakamura, 2007) and GluA1-containing AMPA receptors (Fortin et al., 2012) to the postsynaptic membrane, and this depends on trkB-PLCγ-induced Ca2+ signaling. Furthermore, BDNF enhances NMDA-mediated synaptic currents in a trkB-dependent manner (Kolb et al., 2005) and augments the probability of NMDA receptor channel opening (Levine et al., 1998). This effect is mediated by the protein tyrosine kinase Fyn, which, when activated by trkB (Iwasaki, 1998), phosphorylates the NR2B subunit of the NMDA receptor (Xu et al, 2006). BDNF-trkB signaling also increases AMPA and NMDA receptor components (Caldeira et al., 2007b) and synaptic proteins needed to support increased synaptic activity (Kumamaru et al., 2008; Matsumoto et al., 2006) through the MAPK/ERK and PI3K-Akt-mTOR pathways.

BDNF signaling through trkB is also linked with the structural alterations associated with LTP, including the formation of nascent spines and the enlargement of existing spines. BDNF stimulation of trkB increases dendritic spine density in the hippocampus (Tyler & Pozzo-Miller, 2001; Amaral & Pozzo-Miller, 2007), and this process is dependent on MAPK/ERK1/2 activation (Alonso et al., 2004) and activation of TRPC3 channels by the trkB-PLCγ pathway (Yoshii & Constantine-Paton, 2010). BDNF also modulates spine morphogenesis in the presence of synaptic activity (Tanaka et al., 2008) through trkB-mediated signaling events, which promote the transformation of immature spines into stabile mushroom spines. For example, activity-dependent secretion of BDNF and subsequent stimulation of trkB-PI3K-Akt signaling facilitates the trafficking of PSD-95 from the soma to dendrites and synapses (Yoshii & Constantine-Paton, 2007). Subsequent work suggested that BDNF increases synaptic PSD-95 by enhancing microtubule invasion into spines (Hu et al., 2011), since microtubules can serve as highways for delivery of synaptic proteins. PSD-95 functions as a scaffold protein at the post-synaptic membrane, binding glutamate receptors, adhesion molecules, cytoplasmic signaling enzymes, and cytoskeletal regulatory elements required to support the structural and functional properties of active, stabile spines. For example, PSD-95 binds kalirin, a Rho guanine nucleotide exchange factor (GEF) that activates Rac1 to promote actin polymerization and spine stability (Ma et al., 2003). Thus, enhanced trafficking of PSD-95 to spines following BDNF stimulation of trkB likely facilitates stabilization of nascent spines or activity-dependent growth or existing spines.

Additional mechanisms by which BDNF-trkB impact activity-dependent dendritic spine remodeling involve regulation of proteins that control actin cytoskeleton dynamics. These include the Rho family of small GTPases, such as RhoA, Rac1, and Cdc42 and other small GTPases. For example, trkB-mediated activation of the Rac-GEF, Tiam1, can act synergistically with pathways downstream of NMDA receptor activation to stabilize spines (Rex et al., 2007). Phosphorylation of Tiam1 by trkB following stimulation by BDNF (Miyamoto et al., 2006) or by CAMKII following NMDA receptor activation (Tolias et al., 2005) activates Rac1, which promotes F-actin dynamics through activation of p21-activated protein kinase (PAK). PAK phosphorylates LIM motif-containing protein kinase 1 (LIMK1), which phosphorylates and thereby inactivates the actin depolymerizing factor, cofilin (Lai & Ip, 2013). Inactivation of cofilin allows for actin polymerization and spine growth. Moreover, the interaction of trkB and Tiam1 requires the phosphorylation of trkB at S478 by cyclin-dependent kinase 5 (Cdk5) (Cheung et al., 2007). And, Cdk5-mediated phosphorylation of trkB is required for activity-dependent spine remodeling (Lai et al., 2012). TrkB activation also increases cortactin localization to spines (Koleske, 2013). Cortactin, an F-actin binding protein, stabilizes and promotes branching of actin filaments.

4.1. Neuroanatomical, developmental investigations

Forebrain-specific loss of Bdnf and Trkb leads to similar outcomes in the mouse brain, including dendritic spine loss and simplification of dendrite arbors (Gorski et al., 2003; Xu et al., 2000), suggesting that BDNF acts via trkB in the healthy brain to promote dendritic spine and dendrite stability. (Here, “stability” refers to the process by which dendritic spines and dendrites not destined for pruning are retained.) Mice with early-onset forebrain-specific knockdown of Trkb in cortical pyramidal neurons (trkB loss occurs between P14–P28) exhibit dendritic retraction as early as P28, followed by neuronal loss between P42–P72 (Xu et al., 2000). TrkB-deficient neocortical neurons transplanted into the cerebral cortices of wild-type mice similarly display impaired dendritic growth and reduced neuronal survival (Gates et al., 2000). Furthermore, blocking trkB-mediated signaling with 1NaPP1 from P21–31 in mice expressing a Trkb mutant that is sensitive to chemical inactivation reduces dendritic spine densities on apical dendrites of layer II/III excitatory neurons in primary sensory cortex at P31 (Arango-Lievano et al., 2015, 2016).

In mice with early-onset forebrain-specific knockdown of Bdnf, pyramidal neurons develop normally until about P21, when neurons begin to shrink, and dendritic arbors atrophy (Gorski et al., 2003). Because the structural atrophy of neurons in Bdnf knockdown mice occurs during the period when BDNF in the cortex normally increases (P21–35, reported by Gorski et al., 2003; and see section 6.1 for discussion of PFC in particular), these findings hint at an essential role for developmentally-typical elevations in BDNF in the maintenance of neuronal morphology. Moreover, late-onset forebrain-specific Bdnf knockdown (with progressive knockdown beginning at P21 and peaking by P56) does not impact dendritic spine density at P35, but results in a 30% loss of layer II/III dendritic spines in visual cortex at P84 (Vigers et al., 2012). This reduction in spine density following progressive loss of BDNF in adolescence and adulthood suggests that continuous BDNF is required for the ongoing stability of cortical dendritic spines.

5. Glucocorticoid regulation of dendritic spines

Activation of the hypothalamic-pituitary-adrenal (HPA) axis by stressor exposure stimulates the synthesis and release of glucocorticoids, which bind to steroid hormone receptors throughout the body and brain to initiate a cascade of behavioral and physiological changes that enable an organism to effectively cope with stress. Glucocorticoids regulate the termination of the stress response through negative feedback at the level of the hypothalamus and pituitary, as well as other brain regions such as the PFC, hippocampus, and amygdala. Glucocorticoids act by binding to type I mineralocorticoid receptors (MR) and type II glucocorticoid receptors (GR). GRs are present on neuronal cell bodies, dendrites, and dendritic spines, and they regulate gene transcription upon activation by ligand binding. GRs also localize to plasma membranes, where they mediate intracellular signaling via non-genomic mechanisms.

Glucocorticoids exert biphasic effects on dendritic spine density and morphology; the dose and duration of exposure critically determine the degree and type of neuronal remodeling. For example, acute GR blockade reduces dendritic spine head diameters on pyramidal neurons in deep-layer mPFC, suggesting that GR tone is important for the maintenance of large, synapse-containing spines (Swanson et al., 2013). Insufficient or excessively high levels of GR activity can have detrimental effects on neuronal morphology and behavior, in part by regulating trkB expression and activity.

Although the mechanisms underlying the modulation of dendritic spine structure by glucocorticoids remain incompletely understood, rapid effects are thought to involve glucocorticoid-induced regulation of the actin cytoskeleton. For example, acute CORT application rapidly (within 1 hr) increases spine density and spine head diameter in adult rat hippocampal organotypic slice cultures in a GR- and NMDA receptor-dependent manner (Komatsuzaki et al., 2012). Moreover, CORT-induced spinogenesis is abolished by inhibiting MEK1/2, protein kinase A (PKA), PKC, or PI3K. Phosphorylation of actin-binding proteins by these kinases can regulate spine morphology by influencing actin cytoskeleton dynamics (Lin & Koleske, 2010). For example, ERK1/2 phosphorylates cortactin, which stabilizes and promotes branching of actin filaments. CORT rapidly increases phospho-ERK1/2 (p-ERK1/2; i.e., active ERK1/2) in PC12 cells (Qiu et al., 2001), and activation of GR by the synthetic glucocorticoid dexamethasone (DEX) has the same effect in hippocampal slice cultures (Jafari et al., 2012).

Jafari et al. (2012) report that acute DEX rapidly increases levels of phosphorylated cofilin in whole hippocampal homogenates, but reduces phosphorylated cofilin in hippocampal CA1 dendritic spines. Meanwhile, Liston et al. (2013) report that acute CORT-induced activation of GR in cortical cultures increases spine formation via cofilin phosphorylation. Cofilin severs filamentous actin (F-actin) in its active, de-phosphorylated state, allowing for actin polymerization and depolymerization. During the initial phases of LTP, when F-actin is being remodeled, cofilin must be active (de-phosphorylated) to translocate to the spine and create free barbed ends for actin polymerization (Bosch et al., 2014). Cofilin must be phosphorylated to be retained in the spine and for the maintenance of spine enlargement. A spine-selective loss in p-cofilin following DEX treatment, reported by Jafari et al. (2012), may be due to the fact that double immunostaining for PSD-95 and p-cofilin was measured. Because p-cofilin would accumulate in newly formed spines before PSD-95 (Bosch et al., 2014), this quantification method may underestimate glucocorticoid-induced increases in total spine content of p-cofilin. While the directionality of changes in glucocorticoid-induced p-cofilin remains to be clarified, it can be assumed that GRs are able to stimulate rapid changes in dendrite and dendritic spine morphology through modulation of actin cytoskeleton regulatory proteins.

GRs can modulate glutamatergic and BDNF-mediated neurotransmission in a ligand-dependent and -independent fashion, respectively, which would be expected to impact dendritic spine morphology and plasticity. Numakawa et al. (2009) suggest that the GR functions as an adaptor protein, forming a complex with trkB at the membrane. This trkB-GR interaction is critical for BDNF-stimulated glutamate release via activation of trkB/PLCγ signaling. Notably, in cortical cultures, BDNF/trkB/PLCγ-stimulated glutamate release was enhanced by overexpression of GR and reduced by GR down-regulation. GRs also modulate synaptic transmission when bound by acute stress-elicited or exogenous glucocorticoids (Popoli et al., 2012). For example, acute stress rapidly enhances stimulus-evoked glutamate release from pre-synaptic terminals in the PFC through a GR-dependent mechanism that involves an increase in the readily releasable pool of synaptic vesicles (Musazzi et al., 2010). Additionally, acute stress or CORT increases the insertion of AMPA and NMDA receptors into the post-synaptic membrane in the rat PFC, potentiating the response to glutamate (Yuen et al., 2009, 2011). Glucocorticoids can also enhance glutamate neurotransmission indirectly by recruiting endocannabinoid signaling (Hill et al., 2011). Thus, acute stressor exposure, through GR-mediated mechanisms, enhances synaptic transmission, an initial step in LTP induction.

Recently, Liston and Gan (2011) reported that a low dose of CORT enhances dendritic spine turnover within several hours of application in the cortex of adolescent and adult mice, increasing spine formation and elimination to a similar degree. In addition, circadian glucocorticoid oscillations regulate spine plasticity, and these modifications are associated with learning and memory (Liston et al., 2013). Specifically, learning-associated changes in spine formation and elimination were assessed using transcranial two-photon imaging of pyramidal neurons in the motor cortex of mice before and after rotarod training. Circadian glucocorticoid peaks enhanced spine formation after motor skill learning through a non-genomic, GR-dependent mechanism involving phosphorylation of LIMK1 and cofilin. Circadian glucocorticoid troughs promoted the stabilization of a subset of newly formed learning-associated spines and the elimination of a subset of pre-existing spines through an MR-dependent transcriptional mechanism.

Unsurprisingly, glucocorticoids regulate dendritic spine dynamics during adolescence. Administration of an MR antagonist at P30 in mice reduces rates of spine formation and elimination over a 24h period in the developing barrel cortex, suggesting that MR binding is critical for spine remodeling during adolescent development (Liston & Gan, 2011). Further, GR antagonism at P30 reduces spine formation over 24h, revealing an important role for typical, healthy GR occupancy in dendritic spine proliferation during adolescence. In agreement, mutation of specific GR phosphorylation sites throughout early development decreases cortical dendritic spine density in mice (Arango-Lievano et al., 2015; see also Arango-Lievano et al., 2016). In contrast to acute CORT conditions, in which rates of spine formation and elimination were roughly equivalent, spine elimination rates exceed formation rates with longer (or high dose) CORT exposure (Liston & Gan, 2011). This phenomenon could account for stressor-related dendritic spine loss, discussed at the beginning of this review.

Notably, increased dendritic spine turnover elicited by acute CORT exposure in adolescent mice involves the elimination of recently formed spines; meanwhile, chronic CORT results in the loss of both recently formed spines and stable spines established early in development (Liston & Gan, 2011, Liston et al., 2013). This loss of otherwise stably maintained spines suggests that prolonged exposure to excess glucocorticoids may disrupt mechanisms of spine stability.

Chronic glucocorticoid exposure down-regulates BDNF and trkB in the PFC. For instance, chronic CORT in rodents decreases BDNF protein in mPFC (Gourley et al., 2012b), Bdnf mRNA in OFC (Gourley et al., 2009), Bdnf exon IV mRNA (associated with activity-dependent transcription) in PFC (Dwivedi et al., 2006b), trkB protein in frontal cortex (Kutiyanawalla et al., 2011) and mPFC (Gourley et al., 2012b), and Trkb mRNA in PFC (Kutiyanawalla et al., 2011). Prolonged CORT exposure also reduces signaling proteins downstream of trkB, including components of the PI3K-Akt-mTOR pathway (Howell et al., 2011), relevant because trkB activity impacts GR phosphorylation. BDNF stimulation of trkB and downstream activation of ERK1/2 and JNK induces the phosphorylation of GR at 3 sites (Arango-Lievano et al., 2015). Mutation of these sites or blockade of trkB-mediated signaling from P21–31 decreases dendritic spine densities on apical dendrites of excitatory neurons in primary sensory (S1) cortex. Further, chronic unpredictable stress or CORT exposure from P21–31 reduces BDNF protein, GR phosphorylation, and spine densities (Arango-Lievano et al., 2015, 2016). Fluoxetine treatment by contrast increases GR phosphorylation and ameliorates spine loss, but not in mice expressing GR with mutated BDNF-sensitive phosphorylation sites, indicating that the ability of fluoxetine to reverse stress-induced spine loss depends on trkB-mediated GR phosphorylation (Arango-Lievano et al., 2015). Collectively, these findings suggest that trkB-mediated signaling regulates cortical dendritic spine formation in part by modulating GR function. Importantly, disruption of this trkB-GR interaction (for example, by chronic stress) leads to dendritic spine loss.

6. Developmental trajectories of trkB and GR

6.1. TrkB in the PFC across postnatal development

Several studies have characterized trkB (full-length and truncated isoforms) and BDNF levels in the rodent, macaque, and human PFC at various developmental time points, but findings are not always congruent. Fryer and colleagues (1996) report that levels of Trkb.fl mRNA in the rat frontal cortex increase dramatically following birth, and remain relatively static into adulthood. Meanwhile, Trkb.t1 mRNA levels increase sharply around P10–15 and remain static into adulthood, except for another slight increase in mRNA levels at P20. The temporal pattern of trkB protein in the frontal cortex mirrors that of mRNA levels, with trkB.FL predominating at earlier stages of development, and trkB.T1 increasing in late postnatal development. However, mRNA levels were assessed in this report at embryonic day 15 (E15), E18, P0, P5, P10, P15, P20, P30, and adulthood, and protein levels at P0, P10, P20, P30, and adulthood, thus precluding any conclusions about changes in trkB throughout adolescent development in rats (a period from P28–56; see Spear, 2000).

In the mouse visual cortex, BDNF increases dramatically from P21–35 (Gorski et al., 2003), and Jeanneteau et al. (2010) reveal robust levels in the PFC by P17. Following this period, Bdnf mRNA in the PFC appears relatively stable across adolescent development (measured at P20, P35, P40, and P60 by Andersen and Sonntag, 2014). However, some Bdnf transcripts change throughout adolescence (Andersen and Sonntag, 2014). At least 8 promoters within the Bdnf gene in rodents drive the transcription of alternative mRNA transcripts (Aid et al., 2007), allowing for temporal and spatial control of Bdnf (Tao et al., 1998; Timmusk et al., 1993; Baj et al., 2011). Activity-dependent expression of Bdnf in the cortex is primarily mediated by promoter IV (originally referred to as promoter III) (Tao et al., 1998, 2002; Lu, 2003; Sakata et al., 2009). In rat PFC, the Bdnf transcript containing exon IV decreases modestly from pre-adolescence to early adolescence (P20–35), then steadily increases thereafter until adulthood (P35–60) (Andersen & Sonntag, 2014). By contrast, exon IIc modestly increases in early adolescence, decreases sharply from P35 to P40, then slightly increases from P40 to adulthood. In this study, PFC dopamine D3 receptor (Drd3) mRNA was positively correlated with PFC Bdnf (total) mRNA, but negatively correlated with Bdnf exon IIc mRNA. Given that expression of dopamine receptors in the PFC peaks during late adolescence in rodents and humans (Naneix et al., 2012; Weickert et al., 2007), these findings suggest that Bdnf alternative transcripts may have different functional roles in the maturation of the PFC across adolescence.

In macaques, trkB.FL protein levels in the PFC are similar at E140, birth, and adulthood, but increased at P60 (Ohira et al., 1999). TrkB.T1 increases dramatically between birth and P60, and is modestly higher at adulthood. The late postnatal increase in trkB.T1 is in agreement with prior work in rat frontal cortex (Fryer et al., 1996) and mouse prefrontal cortex (Shapiro et al., 2017a). Hayashi and colleagues (2000) assessed trkB.FL immunoreactivity in the PFC of macaques at E140, P7, P6 months (P6m), and adulthood, and found that immunoreactivity was highest at P6m. This is also the age at which synapse density peaks in the PFC (Bourgeois et al., 1994); thus, Hayashi and colleagues (2000) speculated that trkB.FL signaling may be important for synaptogenesis in the PFC. However, few reports examine developmental changes in trkB in the PFC of non-human primates. Because adolescence in macaques typically occurs from 2 to 5 years of age, expression of the trkB isoforms throughout adolescent development of the non-human primate PFC remains to be determined.

Generally speaking, BDNF mRNA and protein in the macaque PFC also varies across development, and the pattern of change resembles that reported for trkB.FL. Mori and colleagues (2004) assessed BDNF protein levels in the PFC of macaques at E120, E140, birth, P2m, and adulthood. BDNF levels increase across development, peaking at P2m, then decline in adulthood. Subsequent work by this group more extensively characterized the developmental changes in BDNF (Mori et al., 2006). Specifically, BDNF mRNA and protein levels were measured at E120, E140, birth, P3 weeks (P3w), P2m, P3m, P6m, P2 years (P2y), P4y, and adulthood in the PFC and several other brain regions. In the PFC, BDNF protein increases steadily throughout development, reaching peak levels at P2m, which remains stable until P6m, then declines steadily until adulthood. Moreover, the decline in BDNF protein levels in primary sensory and motor cortices begins earlier in development than in the PFC, mirroring region-specific trajectories of synaptic remodeling during childhood and adolescence. This pattern suggests that BDNF signaling may be an important regulator of spine plasticity and activity-dependent refinement of synaptic connections during cortical development.

Although the studies discussed thus far have examined developmental changes in BDNF or trkB in the PFC of rats and macaques, sub-regions within the PFC may exhibit distinct patterns of BDNF/trkB. Developmental trajectories of structural maturation vary between sub-regions of the PFC (van Eden & Uylings, 1985), suggesting that the temporal profile of trkB may also vary between PFC sub-regions. Indeed, recent work in female mice has shown that trkB.FL in the OFC and mPFC increases across adolescent development (timepoints include P35, P42, and P56) but levels are modestly higher at P42 (mid-adolescence) in the OFC relative to the mPFC (Shapiro et al., 2017a). Furthermore, trkB.T1 increases across adolescent development in the OFC, but does not change in the mPFC. These findings suggest that characterizing trkB within distinct PFC sub-regions may improve the ability to detect changes across development.

At least four studies in humans report levels of trkB and BDNF in the dorsolateral PFC (DLPFC), including in adolescents and young adults (summarized in Fig. 1). Romanczyk and colleagues (2002) examined the developmental profile of TRKB.FL and TRKB.T1 mRNA expression. TRKB.FL mRNA levels increase from infancy to adolescence and young adulthood, drop in adulthood, and decrease further in the aged brain. Notably, TRKB.FL mRNA peaks in young adulthood, most prominently in superficial layers (II and III), the site of cortico-cortical projections. Further, TRKB.T1 mRNA modestly increases from infancy to adulthood. Webster and colleagues (2002) report that, like TRKB.FL, BDNF mRNA increases in young adulthood. Specifically, BDNF mRNA is relatively low in infants and adolescents, but approximately 1/3 higher in young adults and adults. Furthermore, cortical layers III and V show the greatest age-dependent increase in BDNF mRNA.

Figure 1. A graphical summary of developmental changes in trkB.FL and BDNF in human DLPFC.

(A) Levels of full-length trkB (TRKB.FL) mRNA in layers III and V reported by Romanczyk et al. (2002) and TRKB.FL mRNA reported by Luberg et al. (2010). trkB.FL protein levels reported by Luberg et al. (2010). (B) Levels of BDNF mRNA alternative transcripts containing exons II and IV reported by Wong et al. (2009) and total BDNF mRNA in layers III and V reported by Webster et al. (2002). Mature BDNF (mBDNF) protein levels reported by Wong et al. (2009). Abbreviations: dorsolateral prefrontal cortex, DLPFC.

In the PFC, synapses are overproduced in childhood, then eliminated throughout adolescence (Huttenlocher & Dabholkar, 1997). Thus, the peak in TRKB.FL and BDNF mRNA in young adulthood reported by Romanczyk et al. (2002) and Webster et al. (2002) occurs at the tail-end of a protracted period of major synaptic remodeling and refinement of cortical circuits. The presumed increase in BDNF-trkB signaling at this time may be involved in the stabilization and long-term maintenance of the synapses that are not subject to pruning. Moreover, the prominent elevation in TRKB.FL and BDNF mRNA in layer III may reflect an important role for BDNF-trkB signaling in the refinement or stabilization of cortico-cortical connections. Layer V, which projects to subcortical areas including the striatum, also experiences a prominent age-dependent increase in BDNF mRNA (Webster et al., 2002). Structural MRI suggests that in addition to the PFC, striatal grey matter density also decreases throughout late adolescent development (Sowell et al., 1999). Because BDNF is anterogradely transported and released onto striatal neurons (Altar et al., 1997; Kokaia et al., 1998), an increase in BDNF within layer V of DLPFC during late adolescence/young adulthood may contribute to the maturation of cortico-striatal synapses.

The pattern of developmental changes in trkB reported by Romanczyk et al. (2002) somewhat differs from that reported by Luberg and colleagues (2010), who examined TRKB mRNA and protein in neonates, infants, toddlers, school-age children, teenagers, young adults, and adults. TRKB.FL mRNA is relatively similar across all age groups except for toddlers and adults, with expression significantly higher in toddlers compared to adults. However, there does seem to be a trend for increased TRKB.FL mRNA in teenagers and young adults relative to school-age children and adults, which is more in line with the two reports discussed above. Luberg et al. (2010) also found that trkB.FL protein increases from the neonatal period to infancy, then decreases with age. Further, TRKB.T1 mRNA levels are similar in neonates, infants, young adults, and adults, but decreased in toddlers, school-age children, and teenagers. Protein levels of trkB.T1 steadily increase from neonates to teenagers, then decrease slightly in young adults and adults.

Importantly, as in rodents, several alternative BDNF transcripts exist in humans (Pruunsild et al., 2007; Cattaneo et al., 2016a), and their expression varies across postnatal development in distinct regional and temporal patterns (Timmusk et al., 1993, 1994; Sathanoori et al., 2004; Wong et al., 2009). The generation of alternative transcripts appears to regulate the laminar and subcellular localization of BDNF mRNA (Pattabiraman et al., 2005). Thus, quantification of total levels of BDNF mRNA may mask region- and layer-specific patterns of alternative transcript expression across postnatal development.

Wong and colleagues (2009) characterized the laminar distribution of four major BDNF alternative transcripts across development in DLPFC. Transcripts I, IV and VI are highest in infants and decline with age. Transcript II peaks in toddlers, then drops in school-age children and does not change thereafter. BDNF protein peaks in infants and decreases across development. Wong et al. (2009) also used in situ hybridization to examine laminar differences in BDNF mRNA. They report that BDNF mRNA hybridization is elevated in layer IV in young adults. However, the authors suggest that this pattern likely reflects increased BDNF mRNA in the apical dendrites of layer V pyramidal neurons, which extend into layer IV. Additionally, with increasing age, the distribution of BDNF mRNA in layer V and VI becomes more diffused around pyramidal neurons, suggesting that BDNF mRNA may be more targeted to dendritic processes as the cortex matures.

In summary, in the human DLPFC, some evidence suggests that TRKB.FL mRNA is highest in young adulthood (Romanczyk et al., 2002), and BDNF mRNA increases between adolescence and young adulthood (Webster et al., 2002). However, the TRKB.FL mRNA and protein levels reported by Luberg et al. (2010) were not significantly increased in young adulthood, and BDNF protein declined from infancy to adulthood in another report (Wong et al., 2009). Nevertheless, TRKB.FL mRNA levels are somewhat higher in teenagers and young adults relative to infants, school-age children, and adults (Luberg et al., 2010), and dendritic BDNF mRNA content in layers V and VI increases with age (Wong et al., 2009).

6.1.1. Sex differences

One reason why findings discussed above from Luberg et al. (2010) and Wong et al. (2009) may not be consistent with those of Romanczyk et al. (2002) and Webster et al. (2002) is that the cohort examined by Romanczyk et al. (2002) and Webster et al. (2002) consisted primarily of males, except for the infant age group. The cohort examined by Luberg et al. (2010) and Wong et al. (2009) contained both sexes in all age groups. Given the modulatory effects of sex steroid hormones on trkB-mediated signaling (Hill, 2012a), it is possible that BDNF or trkB is developmentally regulated in a sex-dependent manner. Indeed, at least one study in mice suggests that expression of BDNF and trkB from early adolescence to adulthood may be sexually dimorphic (Hill et al., 2012b).

Hill and colleagues (2012b) report a trend for increased trkB.FL protein levels in the frontal cortex of male mice during early- to mid-adolescence (postnatal weeks 3–7), but levels of phosphorylated trkB (p-trkB) are unchanged across adolescence and adulthood. BDNF protein levels peak in young adulthood (weeks 8–10). Moreover, changes in trkB.FL and BDNF levels from adolescence to adulthood in the striatum are similar to those in the frontal cortex. This pattern of changes is different for females; there are no significant changes in trkB.FL (but see subregion-specific analyses in Shapiro et al., 2017a) and BDNF across adolescence and adulthood in the frontal cortex and striatum. However, levels of p-trkB in the frontal cortex and striatum increase starting in early adolescence, peak at mid-adolescence, then drop in young adulthood. These patterns suggest that changes in the activities and levels of BDNF and trkB across development differ between males and females. Grouping sexes together in experimental analyses may hinder the ability to detect these temporal changes.

6.2. GR in the PFC and HPA axis activity across postnatal development

Stress-induced activation of the HPA axis and subsequent release of glucocorticoids varies throughout postnatal development (McCormick & Mathews, 2007). In rodents, adolescents exhibit a more prolonged CORT response following acute stress than adults (Romeo et al., 2004, 2006a; McCormick & Mathews, 2007; Foilb et al., 2011; Lui et al., 2012). An early study in rats revealed that adults exhibit a greater reduction in stress-induced CORT levels following pre-treatment with DEX than adolescents (Goldman et al., 1973), suggesting that glucocorticoid-mediated negative feedback is weaker in adolescents. Furthermore, following repeated exposure to a stressor, adult rodents show habituation of CORT responses, but adolescents exhibit a sensitized response; peak CORT levels are higher immediately following the stressor, but return to baseline faster (Doremus-Fitzwater et al., 2009; Romeo et al., 2006b). Together, these findings suggest that negative feedback on the HPA axis may be attenuated in adolescent animals relative to adults, but also more prone to sensitize.

Given differential stress responses between rodents of differing ages, one might hypothesize that limbic and cortical regions modulating stress reactivity exhibit decreased GR during adolescence. However, multiple studies examining the developmental regulation of GR protein and mRNA in the hippocampus, paraventricular nucleus (PVN) of the hypothalamus, pituitary, and mPFC report similar levels in pre-, mid- and post-adolescent animals (Dziedzic et al., 2014; Romeo et al., 2008, 2013; Vazquez, 1998). Further, Perlman and colleagues (2007) report a significant effect of age on GR mRNA levels in the human DLPFC, with peak, rather than low, expression during adolescence. Moreover, they find that GR mRNA in various hippocampal subfields does not vary with age. Sinclair and colleagues (2011) report that protein and mRNA expression of GR-α, the predominant GR isoform that translocates to the nucleus upon ligand binding to alter gene transcription (Oakley et al., 1999), increases threefold during postnatal development of the human DLPFC to peak in teenagers, then decrease in young adults and adults. Consistent with these findings, NR3C1, the gene that encodes the GR, in the human PFC peaks between ages 15 and 25 (Harris et al., 2009). Notably, several splice variants and translational isoforms of GR exist (Vandevyver et al., 2014). Because these isoforms can differ in transcriptional activities, trafficking patterns, coactivator recruitment, and subcellular distribution (Lu & Cidlowski, 2005; Oakley & Cidlowski, 2013), differences in the relative levels of GR isoforms in adolescents vs. adults may contribute to differences in GR function across development.

7. Altered PFC trkB and GR in depression

Depressed patients often exhibit disrupted circadian glucocorticoid cycling, resting hypercortisolemia (Yehuda et al., 1996), and impaired negative feedback control of the HPA axis (Holsboer, 2000; Barden, 2004). Furthermore, the Stress Hypothesis posits that HPA axis dysregulation may be an etiologic factor in depression (De Kloet et al., 1998; Holsboer, 2000). Indeed, successful antidepressant treatment is associated with normalization of HPA activity in depressed patients (Barden, 2004), and risk of relapse is higher in patients who do not show normalization of baseline plasma cortisol levels and feedback inhibition of the HPA axis with antidepressant treatment (O’Toole et al., 1997; Holsboer & Ising, 2010). Research in humans and animal models indicates that disruption of forebrain GR-mediated signaling contributes to dysregulation of the HPA axis and possibly, the pathophysiology of depression (Calfa et al., 2003; Boyle et al., 2005; Cattaneo & Riva, 2016b). Specifically, post-mortem studies report decreased GR mRNA in the frontal cortex of depressed patients (Webster et al., 2002) and PFC of teenage suicide victims (Pandey et al., 2013), potentially reflecting a compensatory down-regulation of GRs in response to elevated glucocorticoids. By contrast, antidepressant treatments increase the expression and function of GRs (Calfa et al., 2003; Pariante & Miller, 2001). Furthermore, polymorphisms in the human GR gene (NR3C1) are associated with altered basal cortisol levels, negative feedback inhibition, and stress-induced cortisol responses (Wüst et al., 2004; DeRijk et al., 2008), as well as increased risk for depression (Van West et al., 2006).

Desensitization of GRs and impaired negative feedback of the HPA axis can elevate glucocorticoids, which could impact neuronal morphology via neurotrophic factors including BDNF. According to the Neurotrophic Hypothesis of Depression and Antidepressant Efficacy (Duman et al., 1997), down-regulation of BDNF by prolonged exposure to elevated glucocorticoids contributes to structural and functional alterations in cortico-limbic regions associated with depression, and chemically distinct antidepressants act by stimulating BDNF. In line with this perspective, mRNA and protein expression of BDNF (Karege et al., 2005; Dwivedi et al., 2003a; Qi et al., 2015) and trkB (Dwivedi et al., 2003a; Qi et al., 2015), as well as p-trkB levels (Dwivedi et al., 2009a), are reduced in the PFC of postmortem MDD/suicide subjects. Several studies also report alterations in signaling proteins downstream of trkB in postmortem MDD/suicide subjects, including decreased ERK1/2 signaling (Dwivedi et al., 2009b, 2006a, 2001) and increased MAPK phosphatase 1 and 2 (Duric et al., 2010; Dwivedi et al., 2001) in the PFC; decreased PI3K-Akt signaling and increased PTEN levels (phosphatase and tensin homolog deleted on chromosome 10, which inhibits phosphorylation of Akt) in the ventral PFC (Karege et al., 2007, 2011); decreased mTOR and its downstream signaling targets in the PFC (Jernigan et al., 2011); decreased protein and mRNA levels of CREB (a transcription factor downstream of trkB) in the OFC (Yamada et al., 2003) and Brodmann’s area 9 (consisting of DLPFC and mPFC) (Dwivedi et al., 2003b); and decreased p-CREB levels in the OFC (Yamada et al., 2003). Notably, CREB is decreased in PFC, but not hippocampus, of teenage suicide victims (Pandey et al., 2007). Furthermore, BDNF-trkB signaling is likely involved in the therapeutic efficacy of a variety of chemically-distinct antidepressant treatments (see Saarelainen et al., 2003; Duman & Monteggia, 2006; Rantamäki et al., 2007).

8. Stress-induced alterations in PFC trkB and GR

Chronic stress-induced disruption of trkB-mediated signaling and GR function in key cortico-limbic regions is postulated to increase vulnerability to depression, as discussed above. Also as discussed, experiments using animal models with reduced trkB starting in early adolescence suggest that BDNF/trkB are critical for normative dendritic spine maturation and the maintenance of stabile spines throughout life. GR is also involved in adolescent spine maturation, promotes spine plasticity across the lifespan, and regulates the activity of the HPA axis. Thus, loss of neurotrophic support and impaired GR function may impact risk for psychopathology by increasing basal and stress-induced release of glucocorticoids and altering neuronal morphology and/or structural plasticity. During adolescence, disruption of trkB or GR may be particularly impactful for the still-maturing PFC. In this section, we review evidence that chronic stressors and glucocorticoid exposure modulate trkB and GR activity. We also discuss evidence from rodent models implicating dysregulation of both BDNF-trkB and GR systems in the development of depressive-like phenotypes, and restoration of these systems by antidepressants.

8.1. Effects of acute/chronic stressors or glucocorticoid exposure on trkB

Transient activation of the HPA axis by acute stress enhances neurotransmission and synaptic plasticity in the PFC, adaptive effects thought to enhance emotional memory processes and promote resilience to future stressors. By contrast, chronic stress impairs neuroplasticity in the PFC and dysregulates the HPA axis, maladaptive effects that may contribute to allostatic load and increase risk for the development of psychopathology (McEwen & Gianaros, 2011). Likewise, BDNF-trkB signaling is differentially regulated by acute and chronic stressors. Acute stress increases p-ERK2 levels in the frontal cortex and PFC of mice and rats (Galeotti & Ghelardini, 2012; Meller et al., 2003; Shen et al., 2004). In cultured cortical neurons, acute CORT or DEX increases phosphorylation of trkB, PLCγ, Akt, and ERK1/2, with peak levels occurring after 2–4 hrs (Numakawa et al., 2009). Further, Jeanneteau and colleagues (2008) report that glucocorticoids can stimulate p-trkB through the genomic actions of GR, independent of effects on BDNF synthesis and release. What GR-responsive gene products are responsible for the phosphorylation of trkB remains to be determined.

8.1.1. Changes associated with depressive-like phenotypes

In contrast to the stimulatory effects of acute stressor or glucocorticoid exposure on trkB-mediated signaling, chronic stressors and prolonged exposure to glucocorticoids reduce levels and activity of trkB in the PFC. These biochemical changes are associated with depressive-like behaviors, and can be blocked or reversed with antidepressant treatment (table 1). For instance, chronic forced swim stress in male rats induces anhedonic-like behavior and decreases p-ERK2 and p-CREB in the PFC, and these changes are blocked by fluoxetine treatment (Qi et al., 2006, 2008). Moreover, p-ERK2 in the PFC is positively correlated with body weight in stressed animals during stressor exposure, suggesting that the degree of disruption in PFC trkB-ERK2 signaling correlates with the severity of stress-induced physiological impairments (Qi et al., 2006).

Table 1. Summary of studies examining short-term effects (within 1 week) of repeated stressors or CORT exposure on prefrontal cortical BDNF-trkB. If depressive-like behaviors or other longer-term neurobiological measures were also collected, they are noted.

“Immobility” refers to immobility in the forced swim test or tail suspension test. Abbreviations: antidepressant, ADT; corticosterone, CORT; euthanasia, euth.; infralimbic cortex, IL; medial prefrontal cortex, mPFC; orbitofrontal cortex, OFC; prefrontal cortex, PFC; prelimbic cortex, PL; full-length trkB, trkB.FL; truncated trkB, trkB.T1.

Reference	Species, age	Stress/CORT	Duration	Endpoint (timing relative to end of stressor/CORT)	Reported brain region	Protein/gene changes	Behavioral changes	ADT-like effects if applicable
Qi et al., 2006	adult male rats	forced swim	14 days	behavior: <1 day euth.: 1 day	PFC	↓ p-ERK2/ERK2	↓ saccharin preference
Qi et al., 2008	adult male rats	forced swim	21 days	behavior: <1 day euth.: ~4 days	PFC	↓ p-ERK2, p-CREB	↓ saccharin preference	Fluoxetine (concurrent with stressor exposure) blocked all
Zhou et al., 2017	adult male mice	Unpredictable stress	35 days	behavior: 4–7 days euth.: 7 days	PFC	↓ BDNF ↓ p-trkB/trkB	↓ sucrose preference ↑ immobility	Fluoxetine or biperiden (days 28–37 of the stress period) blocked all; Biperiden actions blocked by trkB antagonist
Wang et al., 2015	adult male mice	Unpredictable stress	56 days	behavior: 2–3 days euth.: 3 days	PFC	↓ p-ERK1/2/ERK1/2 ↓ p-Akt/Akt ↓ p-CREB/CREB	↓ sucrose preference ↑ immobility	Acute alarin (1 and 3 days after stressor) reversed all changes except sucrose preference; alarin actions blocked by trkB antagonist
Zhang et al., 2016b	adult male mice	Unpredictable stress	56 days	behavior: 1 day euth.: 3 days	PFC	↓ BDNF ↓ p-trkB/trkB ↓ PSD95 ↓ synaptophysin	↓ sucrose preference	7,8-DHF (days 29–56 of the stress period) blocked all (trkB antagonist blocked the effects of 7,8-DHF)
Wang et al., 2016	adult male rats	restraint stress	21 days	behavior: 2 days euth.: 1 day	PFC	slight ↓ BDNF ↓ Bdnf mRNA ↓ p-ERK1/2/ERK1/2	↓ sucrose preference ↑ immobility	Fluoxetine or resveratrol (concurrent with stressor exposure) blocked all changes
Pesarico et al., 2017	adult male mice	social defeat	10 days	behavior: 1 day euth.: 1 day	PFC	↓ p-ERK1/2/ERK1/2 ↓ p75 receptor ↑ proBDNF ↑ p-CREB/CREB	↓ social interaction	FDPI (synthetic isoquinoline compound) (concurrent with stressor exposure) blocked all changes
Ma et al., 2016	adult male mice	social defeat (susceptible mice)	10 days	behavior: 1–3 days euth.: 4 days	PFC	↓ mBDNF ↓ p-trkB/trkB --- proBDNF	↓ sucrose preference ↑ immobility	Acute fluoxetine + brexpiprazole (2 days after stress) reversed all (trkB antagonist blocked)
					PL	↓ spine density		Fluoxetine + brexpiprazole reversed
					IL	--- spine density
Wang et al., 2018	adult male rats	social defeat	14 days	behavior: <1 day-5 days euth.: ~6 days	mPFC	↓ BDNF --- p-ERK1/ERK1 ↓ p-ERK2/ERK2 ↓ p-CREB	↓ sucrose preference (BDNF in mPFC correlated with preference) ↓ cognitive flexibility ↓ reversal learning (BDNF in OFC correlated with better reversal learning)
					OFC	↓ BDNF --- p-ERK1/2/ERK1/2 ↓ p-CREB
Barfield et al., 2017a	Adolescent male mice	oral CORT	11 days	behavior: >14 days euth.: <1 day	mPFC	↓ trkB.FL/trkB.T1 --- p-ERK1/2/ERK1/2	↓ goal-directed decision making ↑ immobility (after acute stressor) ↓ reward-related motivation	7,8-DHF (P39–47) blocked decision-making impairments and motivational deficits
					PL	↓ spine density ↑ spine volume
				83 days	PL	↑ spine volume
de Sousa et al., 2015	adult female mice	CORT injections	21 days	behavior: <1 day euth.: 1 day	PFC	↓ BDNF	↓ sucrose preference ↑ immobility	Desvenlafaxine (DVS) or DVS+Alpha-lipoic acid (during last week of CORT) blocked all changes
Karisetty et al., 2017	adult male, female mice	Unpredictable stress	21 days	behavior: 2 days euth.: 3 days	PFC	--- BDNF (males) ↓ BDNF (females)	↓ sucrose preference (females) ↑ immobility (both sexes)
Shirayama et al., 2015	adult male rats	Inescapable electric shock	2 days	behavior: 6 days euth.: 2 or 6 days	mPFC	↑ proBDNF ↓ mBDNF ↓ p-trkB/trkB	↑ “helplessness” (failure in conditioned avoidance test)	7,8-DHF (2 days after shock) restored p-trkB (trkB antagonist blocked); 7,8-DHF infusion in the IL (but not PL) reduced “helplessness”
Bai et al., 2016	adult male rats	Unpredictable stress	21 days	behavior: 2–4 days euth.: 5 days	neocortex	↑ proBDNF ↑ p75 receptor (protein & mRNA) ↑ sortilin (protein & mRNA) ↓ trkB (protein & mRNA) ↓ Bdnf mRNA ↓ spine length	↓ sucrose preference ↑ immobility	Anti-proBDNF antibody (1 day after stress) reversed stress-induced behavioral changes and normalized spine lengths
Chiba et al., 2012	adult male rats	restraint stress	28 days	behavior: <1 day-6 days euth.: 1 day	PFC	--- BDNF --- trkB ↓ BDNF-stimulated glutamate release	↓ sucrose preference ↑ immobility
Yan et al., 2016	Adolescent male mice	CORT injections	21 days	behavior: 1 day euth.: 2 days	PFC	↓ BDNF ↓ p-trkB/trkB ↓ p-CREB/CREB	↓ sucrose preference, ↑ immobility	Fluoxetine (concurrent with stressors) blocked all (trkB antagonist blocked effects of fluoxetine)

Five weeks of chronic unpredictable stress in male mice also induces anhedonic-like and depressive-like behavior, with concomitant reductions in BDNF and p-trkB/trkB in the PFC (Zhou et al., 2017). Meanwhile, fluoxetine during the last week of stress exposure blocks these stress-induced changes. Decreased levels of p-ERK1/2/ERK1/2, p-Akt/Akt and p-CREB/CREB in the PFC following 8 weeks of chronic unpredictable stress has also been reported (Wang et al., 2015). In this study, Alarin (a member of the galanin family of neuropeptides) in stressed mice reverses depressive-like behavior and restores trkB signaling proteins in the PFC; furthermore, this effect depends on trkB activation. Similarly, Zhang et al. (2016b) report that the putative trkB agonist, 7,8-dihydroxyflavone (7,8-DHF), blocks anhedonic-like behavior and loss of BDNF, p-trkB/trkB, PSD95, and synaptophysin in the PFC of mice exposed to 8 weeks of unpredictable stress. The authors of the present review also find that 7,8-DHF has antidepressant-like effects, in this case in adolescent mice (Barfield et al., 2017a).

Studies utilizing other chronic stress models, including chronic restraint, repeated social defeat, and inescapable electric shock (learned helplessness model) report largely consistent findings of reduced BDNF (protein and mRNA), trkB (protein and mRNA), p-trkB/trkB and downstream signaling proteins in the PFC when brain tissue is collected within a week after the last stress exposure (table 1) (Wang et al., 2016, 2018; Pesarico et al., 2017; Ma et al., 2016; Shirayama et al., 2015). Moreover, the restoration of BDNF-trkB levels by antidepressant treatments in stressed animals is concurrent with the reversal of stress-induced behavioral alterations. At least one study also reports concomitant stress-induced alterations in neuronal morphology in the mPFC (Ma et al., 2016). For example, in the repeated social defeat stress model used by Ma et al. (2016), ‘susceptible’ mice (i.e., those displaying social avoidance after the stress exposure) show increased anhedonic-like and despair behavior, reduced mature BDNF and p-trkB/trkB in the PFC, and lower dendritic spine densities in the PL cortex compared to non-stressed mice. A combination of the atypical antipsychotic brexpiprazole and a sub-threshold dose of fluoxetine produces rapid antidepressant-like effects and increases BDNF, p-trkB/trkB, and spine densities in stress-susceptible mice; moreover, pre-treatment with a trkB antagonist blocks the effects of fluoxetine + brexpiprazole. Exposure to elevated CORT levels during early adolescence in male mice decreases the ratio of trkB.FL/trkB.T1 in the mPFC, but does not chronically alter p-ERK1/2/ERK1/2 levels (Barfield et al., 2017a). Adolescent CORT-exposed mice nevertheless develop long-term deficits in goal-directed decision making and decreased motivation, which can be blocked by the trkB agonist 7,8-DHF.

Despite evidence for sex differences in the prevalence of stress-related psychopathologies (Kessler, 2003; Holbrook et al., 2002), relatively few animal studies examining neurotrophic mechanisms of stress vulnerability have used females. Nevertheless, de Sousa et al. (2015) report that chronic CORT exposure in female mice decreases BDNF protein levels in the PFC and induces anhedonic-like and depressive-like behavior. This pattern is consistent with our own findings that the trkB agonist 7,8-DHF can correct PFC-dependent decision-making abnormalities in CORT-exposed female adolescent mice (Barfield et al., 2017a; Barfield & Gourley, 2017b). In another report, chronic unpredictable stress decreases PFC BDNF in adult female, but not male, mice. And while both stress-exposed males and females develop increased immobility in the FST, only females also develop decreased sucrose preference (Karisetty et al., 2017). Considering the increased incidence of depression in women (Kessler, 2003), additional studies exploring sex differences in the vulnerability to, and persistence of, stress-induced changes in PFC BDNF-trkB are certainly warranted.

Chronic stress alters levels of the precursor form of BDNF, proBDNF, which binds to the p75^NTR and sortilin receptors. proBDNF consists of an N-terminal prodomain and a C-terminal mature domain, and is cleaved to yield a prodomain and mature BDNF (mBDNF, which binds to trkB). In contrast to mBDNF, uncleaved proBDNF enhances apoptosis, inhibits neurite outgrowth, and promotes spine pruning (Teng et al., 2005; Koshimizu et al., 2009; Sun et al., 2012; Orefice et al., 2016). Shirayama et al. (2015) report that inescapable electric shock increases proBDNF in the mPFC of male rats. Repeated social defeat stress in mice also increases proBDNF in the PFC (Pesarico et al., 2017). However, in this study, p75^NTR was reduced, which the authors suggest may be a compensatory response to increased proBDNF. Nevertheless, an isoquinoline compound (FDPI, a putative modulator of monoaminergic activity) prevents stress-induced changes in proBDNF and p75^NTR, as well as stress-induced social avoidance (Pesarico et al., 2017).

proBDNF may contribute to the behavioral and morphological consequences of chronic stress. Rats exposed to chronic unpredictable stress have higher levels of proBDNF (protein), p75^NTR (protein and Ngfr mRNA) and sortilin receptor (protein and Sort1 mRNA), and lower levels of Bdnf mRNA and trkB (protein and mRNA) in the neocortex (Bai et al., 2016). Stressed rats also exhibit anhedonic- and depressive-like behavior and decreased dendritic spine lengths in the neocortex. Furthermore, administration (i.c.v. or i.p.) of an anti-proBDNF antibody increases sucrose preference, decreases immobility in the FST, and restores spine lengths. Thus, chronic stress may “shift the balance” between two opposing pathways [proBDNF and its receptors (p75^NTR and sortilin) vs. BDNF and trkB] towards proBDNF-mediated signaling, favoring neurodegeneration and spine loss. Of note, because stress reduced Bdnf mRNA, the stress-induced increase in proBDNF protein is likely due to inhibition of proteolytic cleavage of proBDNF to mBDNF, and not increased transcription of Bdnf. Interestingly, proBDNF is elevated in the serum and plasma of patients with depression (Yoshida et al., 2012; Zhou et al., 2013), and antidepressant treatments increase the enzymes that cleave proBDNF (Sartori et al., 2011; Segawa et al., 2013). In sum, chronic stress impacts the balance of proBDNF vs. mBDNF, which may mediate the neurobehavioral response to adversity.

In the absence of chronic stress- or CORT-induced changes in BDNF or trkB levels, BDNF-trkB function can nevertheless be impacted. Chiba et al. (2012) report that chronic restraint stress in rats does not alter trkB or BDNF levels in the PFC, but decreases BDNF-stimulated glutamate release in PFC slices, down-regulates GR, induces immobility in the FST, and decreases sucrose preference. Numakawa et al. (2009) suggest that this impairment in BDNF-trkB function may be due to decreased trkB-bound GR levels following chronic stress. As discussed above (section 5), the interaction of GR with trkB is critical for BDNF-trkB activation of PLCγ signaling that stimulates glutamate release. In cultured neurons, prolonged CORT exposure decreases GR, resulting in reduced trkB-GR interaction and attenuated BDNF-induced glutamate release (Numakawa et al., 2009). Together, these patterns suggest that stress-induced glucocorticoids can impair BDNF-trkB function, at least in part, by reducing the interaction of trkB and GR in the PFC, which may contribute to stress-induced behavioral alterations. The effects of chronic stress on GR are discussed in detail in section 8.2.

Numerous reports indicate that the antidepressant-like efficacy of a variety of drugs can be blocked by co-administering a trkB antagonist (Wang et al., 2015; Zhang et al., 2016b; Ma et al., 2016; Zhou et al., 2017). Interestingly, multiple studies also report that administration of the trkB antagonist ANA-12 alone has antidepressant-like effects (Cazorla et al., 2011; Zhang et al., 2014, 2015). This phenomenon is thought to be mediated by antagonism of trkB receptors in the nucleus accumbens (NAc) – a brain region in which stress-induced changes in BDNF-trkB are, generally speaking, opposite those in the PFC. For example, several days after the end of a 10-day social defeat period in mice, mBDNF is decreased in the PFC and increased in the NAc (Zhang et al., 2015). Increased mBDNF and p-trkB/trkB, as well as decreased proBDNF, are also observable in the NAc of rats exposed to inescapable electric shock (Shirayama et al., 2015). In this study, infusion of the trkB agonist 7,8-DHF (but not ANA-12) in the PFC reduces escape failures and latency to escape in a new situation with a controllable shock (conditioned avoidance test) – considered antidepressant-like effects. The same behavioral patterns are evident with infusion of ANA-12 (but not 7,8-DHF) into the NAc. Antidepressant-like effects of BDNF blockade in the NAc have also been reported (discussed in Nestler and Carlezon, 2006).

Some evidence suggests that antidepressants correct stress-induced alterations in BDNF-trkB and neuronal morphology simultaneously in both the PFC and NAc. For example, Ma et al. (2016) report that fluoxetine + brexpiprazole blocks depressive- and anhedonic-like behaviors in mice susceptible to chronic social defeat. This treatment also boosts BDNF, p-trkB/trkB and dendritic spine density in the PFC, while normalizing (reducing) BDNF, p-trkB/trkB and dendritic spine density in the NAc. These findings suggest that the therapeutic-like efficacy of antidepressants may depend on a nuanced modification of BDNF-trkB systems throughout the brain.

8.1.2. Persistence of stress-related changes on prefrontal cortical BDNF-trkB

The studies discussed thus far assessed short-term effects of stress on trkB-mediated signaling (determined ≤1 week after stressor cessation). Identifying and characterizing persistent alterations may yield further insight into the chronic nature of depression. Far fewer studies have examined long-term changes in BDNF or trkB following chronic stress exposure, especially in the PFC, but initial findings are summarized in table 2.

Table 2. Summary of studies examining long-term effects of repeated stressors or CORT exposure on prefrontal cortical BDNF-trkB and depressive-like behaviors.

In the case of stressor/CORT exposure during adolescence (studies below the bold line), ages are indicated, with “P” referring to postnatal day. See also the report of Barfield et al., 2017a, listed in table 1. “Immobility” refers to immobility in the forced swim test or tail suspension test. Note that, relative to the immediate effects of stressor/CORT exposure summarized in table 1, effects are variable and some compensation/adaptation in BDNF may be detectable. Abbreviations: anterior cingulate, AC; antidepressant, ADT; corticosterone, CORT; euthanasia, euth.; female, F; infralimbic cortex, IL; lateral orbitofrontal cortex, lOFC; male, M; medial orbitofrontal cortex, mOFC; medial prefrontal cortex, mPFC; orbitofrontal cortex, OFC; prefrontal cortex, PFC; prelimbic cortex, PL.

Reference	Species, age	Stress/CORT	Duration	Endpoint (timing relative to end of stressor/CORT)	Reported brain region	Protein/gene changes	Behavioral changes	ADT-like effects if applicable
First et al., 2011	adult male rats	Unpredictable stress	35 days	behavior: 1–13 days euth.: 14 days	frontal cortex	--- BDNF --- trkB ↓ p-ERK1/2/ERK2	--- spatial learning	Fluoxetine (concurrent with stressors) blocked changes in p-ERK1/2
Yang et al., 2017	adult male mice	Unpredictable stress	28 days	behavior: 14 days euth.: 15 days	cortex	↓ Bdnf mRNA	↓ sucrose preference ↑ immobility	Anti-proBDNF injection in AC (on last day of stressor exposure) reversed all changes
Fanous et al., 2010	adult male rats	intermittent social defeat	every 3rd day for 10 days	2 hours	PL	↑ BDNF --- Bdnf mRNA
					IL	--- BDNF ↑ Bdnf mRNA
					AC	↑ BDNF ↑ Bdnf mRNA
				28 days	PL, IL, AC	--- BDNF --- Bdnf mRNA
Lin et al., 2009	adult male, female rats	footshock	21 days + every other day for 21 days	1 day	PL	--- BDNF - M ↓ BDNF - F
					IL	--- p-CREB - M, F
					AC	↓ p-CREB - M --- p-CREB - F
			21 days	22 days	PL, IL, AC	--- p-CREB - M, F --- BDNF - M, F
Dong et al., 2017	adult male mice	social defeat (susceptible mice)	10 days	behavior: 2–9 days euth.: 10 days	PFC	↓ BDNF ↓ p-trkB/trkB ↓ PSD-95	↓ sucrose preference ↑ immobility	Ketamine (2 days after stress) reversed all changes
					PL	↓ spine density
					IL	--- spine density
Yang et al., 2015	adult male mice	social defeat (susceptible)	10 days	behavior: 2–8 days euth. 8 days	PFC	↓ BDNF ↓ p-trkB/trkB --- trkB	↓ sucrose preference ↑ immobility	R-Ketamine and S-Ketamine (1 day after stress) reversed all (trkB antagonist blocked ADT-like behavioral effects)
					PL	↓ spine density
					IL	--- spine density
Yang et al., 2016	adult male mice	social defeat (susceptible)	10 days	behavior: 2–9 days euth.: 10 days	PFC	↓ BDNF --- proBDNF ↓ p-trkB/trkB --- trkB ↓ PSD-95	↓ sucrose preference ↑ immobility	R-Ketamine (2 days after stress) reversed all changes
Leem et al., 2014	adult male mice	restraint stress	21 days	behavior: 14 days euth.: 18 days	mOFC	↓ p-ERK1/2 ↓ p-MEK1/2	↓ social interaction ↑ immobility	Imipramine (concurrent with stressors) blocked behavioral changes and modifications in the mOFC
					Cingulate cortex	↓ p-ERK1/2
					mPFC	↓ p-ERK1/2
Gourley et al., 2008b	adult male mice	oral CORT	20 days	behavior: 11 days euth.: 20 days	mPFC	--- p-ERK1/2/ERK1/2	↓ reward-related motivation ↑ immobility	Amitriptyline (for 1 week following CORT) restored motivation; amitriptyline or fluoxetine (for 2 weeks following CORT) blocked immobility
Gourley et al., 2008c	adult male mice	oral CORT	20 days	21 days			↓ sucrose intake	Amitriptyline (for 2 weeks following CORT) or fluoxetine (for 3 weeks following CORT) restored sucrose intake
				behavior: 11 days euth.: 20 days	mPFC	--- p-ERK1/2/ERK1/2	↓ reward-related motivation (scores covaried with BDNF and p-ERK1/2)	Amitriptyline (for 1 week following CORT) restored motivation
Gourley et al., 2012b	adult male rats	oral CORT	20 days	>14 days			↓ goal-directed decision making	Amitriptyline (for 10 days following CORT) restored decision-making
				7 days	mPFC	↓ trkB
	adult male mice			behavior: >14 days euth.: following behavioral testing	mPFC	↓ BDNF	↓ goal-directed decision making ↓ reward-related motivation (scores covaried with BDNF in CORT mice)	Riluzole (for 3 weeks following CORT) blocked effects of CORT on motivation and BDNF. Local BDNF infusion restored motivation
Gourley et al., 2009	adult male rats	oral CORT	20 days	behavior: 14 days euth.: 21 days	lOFC	↓ Bdnf mRNA	↓ conditioned fear extinction ↓ sucrose preference
					IL	--- Bdnf mRNA
Zhang et al., 2017	adolescent male rats	individual housing, P21-end + unpredictable stress, P28–41	14 days	1 day			↓ sucrose preference
				behavior: 21 days euth.: 32 days	mPFC	↓ BDNF ↓ p-ERK1/2/ERK1/2 ↓ p-CREB	--- sucrose preference ↑ immobility ↓ cognitive flexibility
Xu et al., 2016b	adolescent male mice	social defeat, P28–37, then individual housing until euth.	10 days	behavior: 7 days euth.: 8 days	mPFC	↑ Bdnf mRNA	--- cognitive flexibility
				behavior: 42 days euth.: 49 days	mPFC	↓ BDNF ↓ Bdnf mRNA	↓ cognitive flexibility	Duloxetine (P65–79) reversed all changes
					OFC	--- BDNF --- Bdnf mRNA
Xu et al., 2017	adolescent male mice	social defeat, P28–37, then individual housing until euth.	10 days	behavior: 42 days euth.: 49 days	mPFC	↓ Bdnf total mRNA ↓ Bdnf IV mRNA	↓ cognitive flexibility	Tranylcypromine (P65–78) reversed changes in behavioral measures and Bdnf IV
Desbonnet et al., 2012	adolescent male mice	individual housing, P31-end + social defeat, P35–45	10 days	5 days			↓ sucrose preference
				behavior: 25 days euth.: 40 days	PFC	--- Bdnf mRNA	↑ agonistic behavior --- sociability

Chronic mild stress in adult male rats decreases p-ERK1/2/ERK2 in the frontal cortex, detectable 2 weeks following stressor exposure, and daily fluoxetine throughout the duration of stress normalizes p-ERK1/2/ERK2 levels (First et al., 2011). However, BDNF and trkB in the frontal cortex, as well as spatial learning and memory in the Morris water maze test, were apparently not affected by stress exposure in the same report. In another study, chronic unpredictable stress reduces Bdnf in the cortex of adult mice, and Bdnf deficiency is detectable 2 weeks following stress, in tandem with decreased sucrose preference and despair-like behavior (Yang et al., 2017). Infusion of an anti-proBDNF antibody into the anterior cingulate cortex reverses all stress-induced behavioral and molecular alterations (for discussion of proBDNF, see section 8.1.1).

While the long-term effects reported by Yang et al. (2017) are generally consistent with the short-term effects of chronic stress discussed earlier, studies that have directly compared short- and long-term changes in BDNF-trkB reveal less consistent results. For example, 2h following the last intermittent social defeat stressor, adult male rats have either increased or unchanged levels of BDNF and Bdnf mRNA in the IL, PL, and anterior cingulate cortex (summarized table 2) (Fanous et al., 2010). Meanwhile, no changes in BDNF or Bdnf mRNA are evident 4 weeks following the last stressor. Similarly, adult male and female rats experience sex-dependent changes in p-CREB and BDNF in the PL and anterior cingulate cortex (but not IL) 1 day following chronic footshock stress, but no changes are evident 3 weeks after stress (Lin et al., 2009). Depressive-like behaviors were not assessed in these studies, precluding the ability to draw conclusions about the molecular mechanisms of stress vulnerability or resilience.

Mice susceptible to social defeat stress (i.e., show social avoidance) have decreased levels of BDNF, p-trkB/trkB, and PSD-95 in the PFC 8–10 days after the last stressor, along with increased immobility and decreased sucrose preference (Yang et al., 2015, 2016; Dong et al., 2017). These stress-related behavioral and molecular changes can be reversed by a single dose of ketamine or one of its stereoisomers (R-ketamine and S-ketamine) (Dong et al., 2017; Yang et al., 2015). A trkB antagonist blocks the antidepressant-like effects of R- or S-ketamine on depressive-like behaviors, suggesting that ketamine’s sustained effects are mediated by modulation of trkB. This conclusion is consistent with evience that the rapid antidepressant-like effects of ketamine require rapid elevations in BDNF translation (Autry et al., 2011). Also of note, ketamine, R-ketamine, and S-ketamine restore decreased dendritic spine densities in the PL cortex of stress-susceptible mice (Dong et al., 2017; Yang et al., 2015).

In line with the hypothesis that a chronic stress-related depressive-like phenotype may be characterized by alterations in both BDNF-trkB signaling and GR function, chronic glucocorticoid exposure in rodents produces an antidepressant-sensitive persistent depressive-like state (Gourley et al., 2008a,b), desensitizes GRs (Chiba et al., 2012), and decreases BDNF and trkB in mPFC (Gourley et al., 2012b). Recent work by our group has shown synergistic effects of chronically diminished mPFC BDNF and GR signaling on goal-directed decision making (Gourley et al., 2012b), which is disrupted in depression (Dickson & Moberly, 2013; Griffiths et al., 2014). Specifically, knockdown of Bdnf in the PL cortex (mimicking the effects of chronic CORT exposure) does not by itself impair goal-directed action selection, but co-administration of a subthreshold dose of the GR antagonist RU38486 impairs the ability of mice to select actions based on the likelihood that they will be rewarded. Thus, deficient BDNF in the PL cortex increases vulnerability to the effects of reduced GR binding on goal-directed decision-making processes. Because BDNF-trkB primes GR phosphorylation, necessary for cortical spine stability (Arango-Lievano et al., 2015), reduced BDNF in the PL may have amplified dendritic spine destabilization caused by GR blockade, thereby disrupting GR-dependent learning and memory. Consistent with this notion of synergy, acute stimulation of cortical cultures with BDNF and the synthetic glucocorticoid DEX increases dendritic spine density, which can be blocked by selective inactivation of GR phosphorylation (Arango-Lievano et al., 2015).

The vast majority of studies examining persistent stress-induced changes in prefrontal BDNF/trkB systems have focused on the mPFC, but emerging evidence suggests that the OFC may also be vulnerable to long-term change. For example, adult male mice exposed to chronic restraint stress have decreased p-ERK1/2 in the medial OFC, cingulate cortex and mPFC, and decreased p-MEK1/2 in the medial OFC 2.5 weeks following the last restraint stressor (Leem et al., 2014). The antidepressant imipramine blocks stress-induced depressive-like behavior [social avoidance, immobility in the tail suspension test (TST)] and reductions in MEK/ERK1/2 signaling, but only in the medial OFC, strongly suggesting that trkB-ERK1/2 signaling in this region influences social interaction and stress coping.

In adult male mice, chronic CORT exposure does not alter p-ERK1/2 in the mPFC when measured 3 weeks after CORT, but does impact depressive-like behavior, which is sensitive to antidepressant treatment (Gourley et al., 2008b,c). While these findings suggest that CORT-induced depressive-like behavior may not be accompanied by gross changes in mPFC p-ERK1/2, Gourley et al. (2008c) also report that reward-related amotivation covaries with BDNF and p-ERK1/2 levels in the mPFC of CORT-exposed mice, notable because tissue extracts may be expected to include medial OFC tissue. Additional work by this group indicated that CORT reduces Bdnf mRNA in the lateral OFC, but not in the IL, detectable 3 weeks following CORT, and concomitant with impairments in conditioned fear extinction and sucrose preference (Gourley et al., 2009). These findings are interesting, considering CORT also decreases dendritic spine densities on pyramidal neurons in the lateral OFC, and unlike in some other brain regions, spine densities fail to recover after a washout period (Gourley et al., 2013). Hence, the OFC may be particularly susceptible to long-lasting stress-related alterations in BDNF-trkB and neuronal morphology, but further studies are warranted.

Adolescent rodents may be particularly vulnerable to developing stress-related depressive-like behavior, behavioral inflexibility, and PFC BDNF-trkB deficiencies in adulthood. Three weeks following chronic mild stress exposure in early-adolescent (P28–41) male rats, BDNF, p-ERK1/2/ERK1/2, and p-CREB in the mPFC is decreased, immobility in the FST increases, and behavioral flexibility in the (mPFC-dependent) attentional set-shifting task is impaired (Zhang et al., 2017). Deficits correlated with p-ERK1/ERK1. Interestingly, when tested in adolescence, rats also display anhedonic-like sucrose neglect, but this phenomenon does not persist into adulthood. A separate study exposed male mice to social defeat in early adolescence (P28–37) then singly housed them. Subsequent testing suggested that certain deficits in cognitive flexibility and mPFC BDNF may be delayed, in this case, evident 6 weeks, but not 1 week, after the last social defeat (Xu et al., 2016b). Duloxetine (serotonin and norepinephrine reuptake inhibitor) in adulthood (P65–79) reverses both behavioral and molecular alterations. The authors also note that no stress-induced changes in BDNF within the OFC are evident at the 6-week time point. A follow-up study by this group revealed that total Bdnf mRNA and Bdnf IV (but not transcripts I and VI) mRNA are lower in the mPFC of adult male mice with a history of adolescent social defeat stress (P28–37) (Xu et al., 2017). Adolescent stress also increases histone 3 dimethylation at a region downstream of the Bdnf IV promoter – an epigenetic modification that represses Bdnf IV gene expression through chromatin remodeling (for review of epigenetic modifications following early-life stress, see Jawahar et al., 2015 and Silberman et al., 2016). The monoamine oxidase inhibitor, tranylcypromine, in adulthood (P65–78), reverses the stress-induced changes in Bdnf IV mRNA and epigenetic modifications.

In contrast to the findings of Zhang et al. (2017) and Xu et al. (2016b, 2017), Desbonnet et al. (2012) report that social defeat stress in male mice from P35–45 (and individual housing from P31-end of experiment) does not alter Bdnf mRNA levels in the PFC when measured nearly 6 weeks following the last defeat. Desbonnet et al. (2012) exposed mice to social defeat stress during a slightly later period of adolescence, which may account for this apparent resilience to stress. Consistent with this notion, Zhang et al. (2016a) found that mice exposed to social defeat (then singly housed) from P28–37, but not P38–47 or P70–79, develop deficits in cognitive flexibility and reversal learning 6 weeks later. The authors of the present review have similarly reported that oral CORT exposure in mice during early adolescence (P31–42), but not adulthood (P70–81), disrupts goal-directed decision making several weeks after the cessation of CORT (Barfield et al., 2017a). Additionally, CORT, repeated forced swimming, or trkB antagonism during the same early-adolescent period (P31–42) impairs OFC-dependent behavioral flexibility in adulthood (Barfield & Gourley, 2017b). Although studies directly comparing the long-term effects of stress or CORT during different periods of adolescence/adulthood on PFC BDNF-trkB and function are somewhat lacking, current evidence suggests that stressors during early adolescence may be more impactful than those in adulthood. Indeed, chronic stressors in adulthood often need to be twice as long in duration as stressors in adolescence to produce comparable long-term behavioral and molecular consequences (see table 2).

8.2. Effects of chronic stressor exposure on GR

A well-documented consequence of chronic stressor exposure is the desensitization of GRs in the PFC and hippocampus, leading to impaired negative feedback of the HPA axis. Upon glucocorticoid binding, the GR dissociates from a chaperone protein complex, dimerizes, and translocates from the cytosol to the nucleus, where it regulates gene transcription by interacting with glucocorticoid response elements on DNA or by modulating the activity of other transcription factors (Vandevyver et al., 2012; Oakley & Cidlowski, 2013). The genomic actions of GRs facilitate the termination of the stress response, but under conditions of chronic stress, alterations in the expression, trafficking, and transcriptional activity of GRs may contribute to dysregulation and improper termination of the HPA axis (Pariante, 2006; Pariante & Lightman, 2008).

Chronic stress in rodents decreases GR levels in the PFC (Mizoguchi et al., 2003; Chiba et al., 2012; Chen et al., 2016; Zhuang et al., 2016). Furthermore, in chronically stressed rats, DEX infusion into the PFC fails to suppress plasma CORT levels, indicating that stress-induced changes in GRs in the PFC likely contribute to the disrupted negative feedback on the HPA axis (Mizoguchi et al., 2003). In addition to regulating GR levels, chronic stress alters GR function by modulating the expression of GR co-chaperones. Guidotti and colleagues (2013) report that chronic stress increases cytosolic GR and the co-chaperone FKBP5 (FK506 binding protein 51) in the PFC. Because FKBP5 reduces the affinity of the GR for its ligand and restrains the translocation of GRs to the nucleus (Wochnik et al., 2005), its up-regulation by chronic stress may contribute to dysregulation of the HPA axis by impairing nuclear translocation of the GR and GR-mediated transcription. In line with this perspective, polymorphisms in the human FKBP5 gene that are associated with higher protein/mRNA expression decrease negative feedback inhibition of the HPA axis (Binder et al., 2008).

Impaired GR function following chronic stress may also be mediated by changes in receptor phosphorylation. GR phosphorylation regulates GR transcriptional activity through modification of protein-protein interactions, which can affect stability of the receptor and the association and recruitment of co-factors (Ismaili & Garabedian, 2004). GR phosphorylation also modifies the subcellular localization of the receptor (Rogatsky et al., 1998) and plays a role in the non-genomic actions of the GR by modulating GR-activated cytoplasmic signaling pathways (Ismaili & Garabedian, 2004).

Chronic stress in adult rats alters phosphorylation of GRs at specific residues (Adzic et al., 2009; Papadopoulou et al., 2015), and patterns of stress-induced GR phosphorylation are sexually dimorphic (Mitic et al., 2013). Several specific phosphorylation sites on the rodent GR are now associated with effects of sex, stress, and antidepressants on GR function (Ismaili and Garabedian, 2004; Arango-Lievano et al., 2015). Differential phosphorylation at these sites can positively and negatively regulate GR transcriptional activation. Phosphorylation at S232 by various cyclin-dependent kinase (Cdk) complexes stimulates GR translocation to the nucleus (Adzic et al., 2009; Davies et al., 2008) and promotes transcriptional activation in a gene-specific manner (Rogatsky et al., 1998; Chen et al., 2008). Phosphorylation at S246 by c-Jun N-terminal kinase (JNK) inhibits GR transcriptional activity (Rogatsky et al., 1998) by promoting nuclear export of the receptor (Adzic et al., 2009; Davies et al., 2008). Following trkB activation by BDNF, GR is phosphorylated at S155, S287, and to a lesser extent, S246, by ERK and JNK, thus providing one avenue for cross-talk between trkB and GR signaling (see section 9 for further discussion of BDNF-trkB and GR interactions).

Chronic unpredictable stress in mice decreases p-GR at S232 in the PFC but not hippocampus (Papadopoulou et al., 2015). Other adversities, like chronic isolation, decrease the ratio of p-GR at S232 vs. S246 in the PFC and hippocampus of female, but not male, rats (Mitic et al., 2013). Loss of p-GR at S232 and/or increase at S246 may stimulate GR export from the nucleus and decrease GR transcriptional activation, since GR transcriptional activity is highest when levels of p-GR at S232 exceed that of p-GR at S246 (Chen et al., 2008). These changes may contribute to impaired negative feedback control of the HPA axis. In humans, the ratio of p-GR at S211 (human S211, rat S232) vs. S226 (human S226, rat S246) is low in MDD patients (Simic et al., 2013b). Interestingly, in women but not men, this ratio negatively correlates with current reports of depression, anxiety and stress (Simic et al., 2013a). Thus, differential patterns of GR phosphorylation between males and females may contribute to sex differences in the effects of chronic stress on GR function and the HPA axis.

Some of the effects of chronic stress in impairing negative feedback of the HPA axis may be associated with ovarian steroids, regulators of the GR and its co-regulators (Bourke et al., 2012, 2013; Malviya et al., 2013). Certain forms of chronic stress during adolescence can disrupt behavioral, HPA axis, and GR functioning in female, but not male, rats (Bourke et al., 2013; Bourke & Neigh, 2011), and behavioral vulnerabilities are also reported in adolescent female mice (Barfield & Gourley, 2017b). Greater susceptibility to stress-induced GR dysfunction in females may have implications for understanding increased incidence of depression in women (Kessler, 2003).

8.3. Restoration of both BDNF-trkB and HPA systems by antidepressants

Deficits in neurotrophin signaling and hormonal stress responses are implicated in the pathophysiology of depression, yet historically, studies utilizing animal models have focused on either BDNF-trkB or HPA axis systems. In this section, we discuss investigations that concurrently assessed the sensitivity of both to stress and importantly, behaviorally-efficacious antidepressant treatments.

Several studies employing the unpredictable chronic mild stress model in adult male rats and mice report largely consistent effects – anhedonic- and despair-like behavior, reduced BDNF in the PFC, increased blood serum CORT levels, and adrenal hypertrophy (indicative of an overactive HPA axis) (Jin et al., 2015; Abdul Shukkoor et al., 2016; Pytka et al., 2017; Zu et al., 2017) (table 3). These patterns are also seen in female mice (Filho et al., 2015). At least one study additionally reports decreased p-ERK, p-CREB, and GR (protein and mRNA) in the PFC following chronic stress (Zu et al., 2017). In all of these investigations, chronic administration of traditional, novel, or experimental antidepressants reversed stress-induced molecular, neuroendocrine, and behavioral alterations.

Table 3. Summary of studies examining effects of chronic stressors on both prefrontal cortical BDNF-trkB and HPA measures.

“Immobility” refers to immobility in the forced swim test or tail suspension test. Abbreviations: antidepressant, ADT; corticosterone, CORT; euthanasia, euth.; glucocorticoid receptor, GR; hippocampus, Hipp; hypothalamic pituitary adrenal, HPA; postnatal day, P; prefrontal cortex, PFC; serotonin, 5-HT

Reference	Species, age	Stress/CORT	Duration	Endpoint (timing relative to end of stressor/CORT)	Brain region	Protein/gene changes	HPA changes	Behavioral changes	ADT-like effects if applicable
Chang et al., 2016	adult male rats	Unpredictable stress	56 days	euth.: 2 days			↑ CORT	↓ sucrose intake at day 28 of stress exposure period	7,8-DHF (concurrent with last 4 weeks of stress) dose-dependently blocked changes in CORT
Jin et al., 2015	adult male mice	Unpredictable stress	28 days	<1 day	PFC	↓ BDNF	↑ CORT ↑ adrenal weight	↓ sucrose preference	Fluoxetine or oleoylethanolamide (concurrent with last 3 weeks of stress) blocked all
Zu et al., 2017	adult male mice	Unpredictable stress	35 days	behavior: 3–6 days euth.: 7 days	PFC	↓ BDNF protein, mRNA ↓ p-ERK1/2/ERK1/2 ↓ p-CREB/CREB	↑ CORT ↓ GR protein,mRNA	↓ sucrose preference ↑ immobility	Fluoxetine or higher doses of bacopaside I (concurrent with last 2 weeks of stress) blocked all
Pytka et al., 2017	adult male mice	Unpredictable stress	28 days	1 day	PFC	↓ BDNF	↑ CORT ↑ adrenal weight	↓ sucrose preference ↑ immobility	Fluoxetine or higher doses of HBK-15 (5-HT receptor antagonist) (concurrent with stressors) blocked all
Réus et al., 2012	adult male rats	Unpredictable stress	40 days	behavior: 1–7 days euth.: 7 days	PFC	-- BDNF	↑ CORT ↑ adrenal weight	↓ sucrose intake	Memantine (for 1 week following stress) reversed all and increased BDNF in PFC
Shukkoor et al., 2016	adult male rats	Unpredictable stress	42 days	<1 day	PFC	↓ BDNF	↑ CORT	↓ sucrose preference ↑ immobility	Fluoxetine (concurrent with last 4 weeks of stress) blocked all
Filho et al., 2015	Adult female mice	Unpredictable stress	28 days	behavior: <1 day euth.: 2 days	PFC	↓ BDNF	↑ CORT	↓ sucrose preference, ↑ immobility	Fluoxetine or chrysin (concurrent with stressors) blocked all
Shilpa et al., 2017	adult male rats	Immobilization stress	10 days (2hr/day)	14 days	Frontal cortex	--- BDNF	--- GR	↓ spatial learning & Memory ↓ sucrose preference ↑ immobility	Environmental enrichment (6hr/day for 2 weeks following stress) reversed all
					Hipp	↓ BDNF	↓ GR

Interestingly, Réus et al. (2012) found that while chronic unpredictable stress induces anhedonic-like behavior and increases circulating CORT and adrenal gland weights in adult male rats, BDNF in the PFC is not altered. Memantine (an NMDA receptor antagonist) reverses all stress-induced changes and increases BDNF in the PFC; thus, enhancing BDNF was associated with antidepressant-like efficacy, even though BDNF was not affected by stress in this particular report. The lack of an effect of stress on BDNF may be due to more mild stress conditions relative to other reports. For example, in the experiments by Abdul Shukkoor et al. (2016) and Zu et al. (2017) (see table 3 for comparisons), animals were exposed to 6h of restraint, while animals in Réus et al. (2012) were exposed to 1–3h of restraint or 1.5–2h of restraint at 4°C. Additionally, in experiments by Abdul Shukkoor et al. (2016) and Zu et al. (2017), animals undergoing chronic stress were singly housed. By contrast, animals were group housed for 30 days out of a 40-day stress exposure period by Réus et al. (2012), which could buffer some stress effects.

The studies discussed in this section thus far focused on short-term (≤1 week) effects of chronic stress experienced during adulthood. We identified one investigation that examined neurobiological consequences in adult male rats 2 weeks after repeated (10 days) immobilization stress (Shilpa et al., 2017). BDNF and GR are not significantly affected in the frontal cortex, but are decreased in the hippocampus. Environmental enrichment for 2 weeks reverses anhedonic- and despair-like behaviors, ameliorates spatial learning and memory impairments, normalizes hippocampal BDNF, and partially restores hippocampal GR. Although the authors report no significant stress-induced changes in BDNF and GR in the frontal cortex, different PFC subregions may be variably affected by immobilization stress. Thus, lumping all subregions together as the “frontal cortex” may mask subregion-specific changes in BDNF and GR.

Yan et al. (2016) assessed PFC BDNF-trkB, blood serum CORT levels, and depressive-like behavior in male mice exposed to CORT during adolescence (P35–56), though these variables were measured within a few days following the cessation of CORT injections. CORT decreases BDNF, p-trkB/trkB, and p-CREB/CREB in the PFC, as well as sucrose preference, and increases immobility in the FST. Blood serum CORT levels are also elevated in mice given exogenous CORT, as expected. Treatment with fluoxetine during the CORT exposure period blocks the effects of CORT on all measures, including serum CORT levels, and a trkB antagonist abolishes these effects of fluoxetine. The ability of fluoxetine to normalize serum CORT levels, in a trkB-dependent manner, suggests that fluoxetine facilitates negative feedback on the HPA axis, which may be mediated by effects on GR (Barden, 2004). Importantly, Lee et al. (2016) find that chronic, but not acute stress, reduces hippocampal GR activity. In addition, the therapeutic action of fluoxetine involves hormone-independent activation of GR.

Collectively, these studies demonstrate that putative core features of depression in humans – dysregulated BDNF-trkB and HPA systems – can be recapitulated in chronic stress or chronic glucocorticoid exposure rodent models. Disruption of both systems is associated with the development of depressive-like behavior, while restoration of both systems by antidepressants is associated with the reversal of depressive-like behavior. Taken together, these findings suggest that antidepressants may treat depressive-like behaviors by modulating GR function and activating BDNF-trkB signaling in the PFC. However, in depressed humans, cognitive/behavioral symptoms, mood changes, putative deficits in neurotrophin signaling and hormonal stress response abnormalities persist well beyond the period of chronic stress exposure. Thus, it is important for future work using rodent models to determine whether stress-induced dysregulation of BDNF-trkB and HPA systems and depressive-like behavior persists in the weeks and months following stressor exposure. This is especially pertinent to studies involving prolonged stressor or glucocorticoid exposure during adolescence, since initial evidence suggests that some depressive-like behaviors, cognitive deficits, and biochemical changes may not emerge until adulthood (Xu et al., 2016b; Zhang et al., 2017).

9. BDNF-trkB and GR systems in stress-related metaplasticity

A key factor in understanding the long-term effects of stressor exposure is the likelihood that stressful life events will alter the molecular, cellular, or behavioral response to stressors later in life (Schmidt et al., 2013). According to this view, chronic stressors may increase susceptibility to developing psychopathology in response to subsequent stressor exposure. This priming of neurobiological systems by stressful experiences constitutes a form of stress-induced metaplasticity. Furthermore, the concept of metaplasticity may be particularly relevant for understanding the impact of adverse experiences during adolescence on behavior and stressor vulnerability and resilience later in life. In this section, we discuss evidence implicating BDNF-trkB and GR systems in stress-induced metaplastic modifications.

9.1. HPA axis reactivity to adolescent experience

Considerable evidence indicates that chronic stress during key developmental periods can alter subsequent function of the HPA axis, with effects potentially persisting throughout the lifespan (Meaney et al., 1996). Thus, the concept of stress-induced metaplasticity applies to the effects of early environmental stressors on the HPA response to subsequent stressors. Furthermore, stress-induced programming of the HPA axis may occur beyond prenatal, neonatal, and perinatal development (McCormick et al., 2010). Chronic stress during adolescence can also have long-term effects on subsequent stressor reactivity. For example, adult male rats with a history of chronic variable stress exposure during adolescence (P28–56) exhibit a greater and more prolonged acute stress-induced CORT response, as well as decreased levels of GR protein in hippocampus, than control counterparts (Isgor et al., 2004). Additionally, rats exposed to a triple stressor on P28 and re-exposed to swim stress on P35 and P60 show higher basal CORT levels and reduced GR in the hippocampus at P61 (Uys et al., 2006), suggesting that trauma exposure in early adolescence may impair HPA axis function through down-regulation of GR. Meanwhile, chronic stress exposure in late adolescence/young adulthood (as opposed to early adolescence) blocks the stimulation of PFC Bdnf, Trkb, and p-ERK1/2 by cocaine later in life (Fumagalli et al., 2009). These effects may be associated with blunted CORT, given that cocaine triggers an acute stress response, which typically intensifies BDNF-trkB levels and activities.

Several studies suggest that the impact of adolescent experience on HPA axis activity in adulthood may be sex-specific. Chronic mild stress in rats during early adolescence results in an exaggerated stress-induced CORT response in adulthood in females, but not males (Pohl et al., 2007). Similarly, social subjugation stress in rats from P28–38 potentiates stress-induced CORT levels and increases adrenal gland weight in adult females relative to controls and adult males (Weathington et al., 2012). Chronic social stress (daily 1h social isolation, then housing with a new cagemate) of rats during adolescence (P33–48) exaggerates the CORT response to acute restraint stress 3 weeks later in females, but not males (McCormick et al., 2005). Collectively, these findings suggest that stressful life events during adolescence can alter the hormonal response to stressors later in life, and females may be particularly vulnerable to such long-term changes.

The cellular and molecular mechanisms responsible for long-term changes in HPA axis function following adverse experiences in adolescence are not well known. Nevertheless, decreased hippocampal GR expression and an exaggerated CORT response to acute stress in adult animals with a history of early-life stress may be mediated by hyper-methylation of the glucocorticoid receptor gene (Nr3c1) early in life, which persists into adulthood (Weaver et al., 2004). Suicide completers with a history of childhood abuse exhibit increased methylation of NR3C1 at the exon 1_F promoter (human homologue of the rat exon 1₇ promoter) and reduced GR in hippocampus (McGowan et al., 2009). Methylation of NR3C1 at exon 1_F is also associated with impaired recovery of the cortisol stress response in adolescents exposed to a social stress task (van der Knaap et al., 2015a). Furthermore, exposure to stressful life events in adolescence, but not childhood or the perinatal period, is associated with higher methylation of NR3C1 at exon 1_H in adolescents (van der Knaap et al., 2014). Methylation at the exon 1_H promoter at age 16 is also associated with an increased risk of an internalizing disorder diagnosis at a 3-year follow up (van der Knaap et al., 2015b). However, it remains to be determined whether methylation at exon 1_H in adolescents alters GR and/or HPA function. Nevertheless, it is possible that epigenetic modifications induced by major life stressors in adolescence may contribute, at least in part, to long-term reduction of GR, exaggerated stress-induced glucocorticoid release, and impaired negative feedback control of the HPA axis.

The majority of studies examining the effects of chronic adolescent stress on HPA axis function in adulthood have focused on the hippocampus, but the PFC also regulates HPA reactivity. For instance, mPFC (anterior cingulate and IL cortex) lesions prolong the CORT response to acute restraint stress (Diorio et al., 1993), suggesting that the mPFC participates in inhibition of the HPA axis. However, the role of GRs appears to vary between PFC subregions. In naïve male rats, viral-mediated (sh-RNA) Gr knockdown in both the IL and PL cortex enhances the CORT response to acute stress (McKlveen et al., 2013). Only knockdown of Gr in the IL cortex potentiates the acute stress-induced CORT response in chronically stressed animals, however, and increases immobility in the FST. By contrast, only knockdown of Gr in the PL cortex increases baseline CORT levels in chronically stressed animals. Thus, disruption of GR function in the PFC may contribute to reported alterations in HPA axis reactivity following chronic stressor exposure in adolescence – namely, increased basal glucocorticoid levels, impaired negative feedback inhibition of the HPA axis, and a heightened CORT response to acute stress.

Notably, GR levels in the PFC, but not hippocampus, drop following re-exposure to a restraint stressor in rats previously subjected to chronic restraint stress (Gadek-Michalska et al., 2013). Thus, PFC GRs may be especially sensitive to stress-induced metaplasticity. Future work examining the effects of chronic stressor exposure during adolescence on PFC GR function in adulthood may provide new insight into mechanisms by which adverse experiences alter stress responsivity later in life.

9.2. GR and BDNF-trkB interactions in stress contexts

Patients with MDD can exhibit dysregulation of the HPA axis, decreased GR in the PFC, and decreased trkB and downstream signaling proteins in the PFC, suggestive of a relationship between these alterations and the pathophysiology of depression. Empirical support for this perspective comes from studies utilizing mutant mice with altered Gr, Bdnf, or Trkb expression. Moreover, trkB and GR signaling pathways intersect, and we will discuss evidence that their coordinated actions regulate neurobehavioral responses to stress.

9.2.1. GR disruption

Using a Cre/LoxP system, Boyle and colleagues (2005) generated mice with forebrain-specific knockout of Gr (FBGRKO). In this case, progressive loss of GR does not begin until ~P21, with complete deficit by ~P120–180. Consistent with a role for forebrain GR in the regulation of HPA axis activity, FBGRKO mice exhibit higher basal and peak CORT levels and impaired negative feedback inhibition of the HPA axis. FBGRKO mice also develop despair-like behavior in the FST and TST and anhedonic-like sucrose neglect. Interestingly, depressive-like behavior and altered CORT release in FBGRKO mice are observed at P180 (when GR is nearly absent) but not at P60 (when GR levels are reduced by ~50%). Furthermore, the tricyclic antidepressant, imipramine, normalizes alterations in circadian CORT levels and depression-like behavior, suggesting that imipramine may act on systems impacted by forebrain GR loss (such as BDNF-trkB, discussed below). However, imipramine does not restore negative feedback on the HPA axis, suggesting that the up-regulation of GRs by antidepressants may be required for their ability to restore negative feedback inhibition of the HPA axis (Barden, 2004).

In contrast to mice lacking forebrain GR, mutant mice expressing 50% less GR protein than typical mice (i.e., Gr+/− mice) do not exhibit alterations in basal or peak CORT levels, depressive-like behavior in the FST, abnormal context- and cue-dependent fear conditioning, or anxiety-like behavior (Ridder et al., 2005). However, Gr+/− mice exhibit a higher and more prolonged CORT response to acute restraint stress and impaired suppression of CORT in the DEX suppression test, consistent with impaired negative feedback inhibition of the HPA axis. Gr+/− mice also display increased “helplessness” following 2 days of inescapable and uncontrollable foot shocks. Thus, GR levels may influence gradients of HPA axis dysregulation and susceptibility to stress-induced depression; i.e., as GR levels decrease, the intensity or duration of stressors sufficient to induce depressive-like behaviors and HPA axis disturbances also decreases.

Ridder et al. (2005) also report that hippocampal BDNF protein is diminished in Gr+/−mice, suggesting that vulnerability to depressive-like behavior in GR-deficient mice may be related to impaired BDNF-trkB signaling. Molteni and colleagues (2010) find that acute stressor exposure up-regulates Bdnf mRNA in the PFC of wild-type mice (an adaptive neuronal response), but not Gr+/− mice, suggesting that GR dysfunction may impair the adaptive neurochemical reaction to acute stressors involving BDNF-trkB. In support of this idea, acute stress enhances the post-translational processing of proBDNF to mBDNF (which activates trkB) in the hippocampus of wild-type, but not Gr+/− mice (Molteni et al., 2010). These findings indicate that GR signaling interacts with BDNF/trkB systems to modulate neurobehavioral responses to stress.

9.2.2. BDNF disruption

The investigations using GR mutant mice discussed above suggest that defective GR expression or function impacts BDNF. Likewise, studies utilizing mutant Bdnf mice suggest that BDNF influences HPA axis reactivity and susceptibility to stress-related depressive-like behavior. Mice with a knock-in of the human Val66Met single nucleotide polymorphism (SNP) in the BDNF gene, which decreases the activity-dependent secretion of BDNF (Egan et al., 2003; Chen et al., 2004), do not differ from wild-type (WT) mice in baseline plasma CORT levels (Yu et al., 2012). After exposure to repeated bouts of restraint stress for 7 days, however, BDNF^Val/Met mice exhibit a greater stress-induced elevation in CORT than WT mice. PFC BDNF protein levels are lower in BDNF^Val/Met mice, and restraint stress causes a greater loss of Bdnf mRNA in the PFC of BDNF^Val/Met mice compared to WT mice. Furthermore, BDNF^Val/Met potentiates the stress-induced development of depression-like behaviors and working memory impairments and dendritic spine loss on apical dendrites in the mPFC. Interestingly, Yu et al. (2012) also report a positive correlation between Bdnf mRNA in the mPFC and apical dendritic spine density in the PFC with working memory. At least one study in male rats reports that chronic CORT produces long-term impairments in conditioned fear extinction, a PFC-dependent process, in Val/Met, but not WT, rats (Gururajan et al., 2015). Thus, impairments in the activity-dependent secretion of BDNF may increase vulnerability to HPA axis hyperactivity and stress-related disruption in PFC Bdnf expression, neuronal morphology, and function.

These conclusions differ from those of Notaras et al. (2017), who compared the effects of late-adolescent CORT exposure (P42–56; in the drinking water) in BDNF^Val/Val, BDNF^Val/Met and BDNF^Met/Met mice (male and female) on depressive-like behavior and mPFC BDNF-trkB 2 weeks after the cessation of CORT. Control Met/Met mice (expressing the least amount of BDNF) are more immobile in the FST than Val/Val mice (expressing the most BDNF), and CORT exposure increases immobility in Val/Val mice. By contrast, neither genotype nor late-adolescent CORT exposure impacts BDNF or trkB.FL protein levels in the mPFC in adulthood. CORT decreases trkB.T1 in Val/Val, but not Met/Met, mice, and increases phosphorylated trkB in both genotypes. Thus, it appears that Met allele carriers are not more vulnerable to CORT, as one might expect, during late adolescence. Interestingly, in late-adolescent mice, a longer duration of CORT exposure (i.e., 3 weeks, extending into adulthood) induces different long-term behavioral and molecular changes (e.g., see table 3). Meanwhile, in early-adolescent mice, less than 2 weeks of CORT exposure is sufficient to induce long-term depression-like behaviors and deficiencies in trkB throughout cortico-limbic regions (Barfield et al., 2017a). Determining whether early adolescence and late adolescence in mice are periods of stressor vulnerability and resilience, respectively, is important because it could reveal mechanisms of stressor vulnerability/resilience.

Other studies utilizing different methods of reducing BDNF report sex-dependent vulnerability to stress. For example, chronic unpredictable stress exposure induces anhedonic-like sucrose neglect and increases stress-induced CORT release in female, but not male mice with forebrain-specific inducible knockdown of Bdnf (beginning in adulthood) (Autry et al., 2009). Monteggia et al. (2007) report that, even without stress exposure, Bdnf knockdown beginning in adolescence induces anhedonic-like behavior and behavioral despair in female, but not male, mice in adulthood (Monteggia et al., 2007). Whether female rodents carrying the Bdnf Met allele would be more vulnerable than males to HPA axis disturbance, alterations in PFC neuronal morphology, and behavioral/cognitive impairments following stress exposure in adolescence vs. adulthood would be an interesting focus for future investigations.

9.2.3. Importance of individual differences

Blugeot et al. (2011) suggest that stress-induced disruption of BDNF alters hippocampal neuronal morphology and increases vulnerability to depressive-like behavior upon re-exposure to stress later in life. While this study focuses on molecular and morphological alterations in the hippocampus, it is likely that many of the findings could apply to the PFC as well and are worthy of discussion here. After a stress sensitization paradigm consisting of 4 days of social defeat, male rats exhibited HPA axis hyperactivity, reduced hippocampal BDNF, and morphological changes in the hippocampus, such as dendritic retraction and reduced spine density. Following a 4-week recovery period, two subpopulations of rats were identified: “vulnerable” rats that continued to exhibit diminished BDNF and morphological alterations in hippocampus, and “non-vulnerable” rats that recovered. Additionally, “vulnerable” but not “non-vulnerable” rats developed chronic mild stress-induced depressive-like and anhedonic-like behavior and adrenal hypertrophy. Intracerebroventricular administration of the trkB agonist, 7,8-DHF, during chronic mild stress blocked all neurochemical, morphological and behavioral alterations in “vulnerable” and “non-vulnerable” rats, including increased CORT levels in both groups. Long-lasting impairments in BDNF-trkB signaling following a priming stressful event may increase vulnerability to depressive-like behavior when an organism is re-stressed later in life, but organisms can differ significantly from one to another.

A noteworthy point regarding the study of Blugeot et al. (2011), discussed in the prior paragraph, is that persistent hippocampal BDNF deficits and morphological alterations resulting from the priming social defeat stress were related to vulnerability to depressive-like behavior, and not to a depressive-like state, per se. Furthermore, in “vulnerable” rats, HPA axis activity normalized following a recovery period after the priming stressor, but not following the subsequent chronic mild stress exposure protocol that ultimately induced depression-like behavior. Together, this pattern suggests that simultaneous dysregulation of both BDNF-trkB signaling and HPA axis activity (or GR function) may trigger depression-like phenotypes in some rodent models.

While the work of Blugeot et al. (2011) does not reveal whether the persistent BDNF deficits and chronic mild stress-related disruption of HPA axis activity in “vulnerable” rats are causally related or simply co-occurring, Taliaz and colleagues (2011) indicate that neurotrophin signaling in adolescence can alter HPA axis activity in adulthood. Specifically, viral-mediated Bdnf knockdown in the dorsal dentate gyrus of young rats (surgery at P21, with expected maximum gene knockdown by ~P35) induces anhedonic-like behavior when rats are tested in adulthood and elevates both baseline and acute stress-induced CORT. When control rats are exposed to chronic mild stress during adolescence (from P31–59), only those that exhibit anhedonic-like behavior in adulthood (behaviorally “non-resilient”) also develop deficient hippocampal BDNF and elevated baseline and acute stress-induced CORT levels. Furthermore, unlike in young rats, Bdnf knockdown in adult rats does not cause long-lasting elevations in CORT levels. These findings indicate that stress-related anhedonic-like behavior is associated with reduced hippocampal BDNF and persistently elevated CORT levels, and further suggest that chronic stress-induced disruption of BDNF-trkB-mediated signaling during adolescence may contribute to dysregulation of HPA axis activity.

10. Conclusions

Chronic stress is a well-known risk factor for depression. Abundant evidence from animal studies indicates that exposure to chronic stress recapitulates many of the core behavioral symptoms of depression in humans, as well as key structural and neuroendocrine alterations, including neuronal remodeling and synaptic loss in the PFC and impaired negative feedback inhibition of the HPA axis. Glucocorticoids and neurotrophins are critical regulators of dendritic spine structure and function, and their dysregulation is implicated in synaptic loss associated with stress-related psychopathology. As reviewed here, prefrontal trkB and GR activities are modulated by stressor exposure. Disruption of trkB- and GR-mediated signaling events by chronic stressor exposure may set the stage for the development of psychopathology by impairing dendritic spine stability in the PFC and altering HPA activity.

Neurobiological consequences of, and behavioral outcomes following, chronic stress vary depending on the developmental timing of stressor exposure. This may be due to temporal and regional differences in the trajectory of synaptic maturation and neurotrophin and glucocorticoid receptors across postnatal development, as well as the grouping of neurobiological “units” of proteins (see Zeisel et al., 2018), which together, may contribute to windows of vulnerability to adverse experiences. Indeed, trkB and GR signaling are critical for normal trajectories of brain development, and abnormal adolescent PFC development may contribute to psychiatric disease onset.

Studies using mutant Bdnf/Trkb mice indicate that decreased BDNF-trkB signaling during adolescence results in decreased cortical dendritic spine densities in adulthood, indicating that BDNF-trkB signaling is required for the ongoing stabilization of cortical dendritic spines throughout life. Additionally, some studies report that trkB and BDNF levels in the PFC markedly increase in adolescence or young adulthood, suggesting that BDNF-trkB-mediated signaling may be especially important for activity-dependent refinement of synaptic connections during adolescence. Moreover, adolescents exhibit greater and more prolonged glucocorticoid release following exposure to acute stressors, and HPA reactivity does not habituate with repeated exposure to the same stressor, unlike in adults. Increased expression of GR in the PFC in late adolescence/young adulthood may also render the PFC more susceptible to down-regulation of BDNF or trkB by excessive GR activation during this developmental period. Thus, disruption of trkB-mediated signaling by chronic stressor exposure (or certain gene variants; see Giza et al., 2018) may be particularly impactful in the adolescent PFC, potentially disrupting the trajectory of dendritic spine maturation, resulting in long-lasting structural alterations that may underlie persistent behavioral impairments and increase risk for depression onset.

A key step in understanding mechanisms of susceptibility to stress-related psychopathology may be to identify long-term effects of stressor exposure, particularly stressor exposure during putative vulnerability periods like early adolescence. Initial findings in rodents, as reviewed here, suggest that persistently reduced BDNF-trkB activity following stressor or CORT exposure contributes to long-lasting structural alterations in cortico-limbic regions that underlie vulnerability to depressive-like behavior (Blugeot et al., 2011; Barfield et al., 2017a). Further, effects of chronic stress or CORT exposure often vary by cellular layer and even cellular compartment (e.g., apical vs. basal dendrites), suggestive of circuit-specific effects. Future work examining persistent, circuit-level alterations in molecular regulators of neuronal morphology following stressor exposure may yield critical insight into factors impacting risk for depression.

Importantly, adverse experiences can also impact the neurobehavioral consequences of subsequent stressor exposure, an example of metaplasticity. TrkB and GR systems, and their interaction, may be involved in the effects of stressful life events on the molecular, cellular, neuroendocrine, and behavioral response to stressors later in life. The coordinated actions of trkB and GR regulate neurobehavioral responses to stress, and disruption of one system may increase susceptibility to stress-induced disruption of the other system. A depressive-like state may be characterized by structural defects in cortico-limbic brain regions and both impaired BDNF-trkB-mediated signaling and dysregulated HPA activity. Thus, treatment strategies that target both neurotrophin and glucocorticoid systems may be most effective in reversing structural deficits and associated cognitive/behavioral impairments in depression. Although stress-related psychopathologies, such as depression, are complex and multi-faceted, characterization of the long-term effects of chronic stress on trkB-glucocorticoid interactions in the PFC may facilitate the identification of risk factors and biomarkers, and may critically inform the development of novel treatments and early intervention strategies.