Evolution has shaped the mammalian brain during millions of years, endowing it with redundant and inter-related neurotransmitter networks to manage and administer stress. The characteristics of the “berserk,” the ultimate warrior—superhuman physical strength, insensitivity to pain, lack of concern for the consequences of his actions—are possibly the display of an extreme state, an upper limit of human physical and mental condition at the core of norepinephrine (NE)-induced states.

Although a variety of hormones may turn on neuronal circuits for the execution of energetically demanding behavioral tasks, only a fistful of neurotransmitters have the capability to actually regulate the global state of activation of the whole brain, managing effectively and parsimoniously the necessarily limited energy/power capability of the brain and of the whole organism. The NE-releasing Locus Ceruleus (LC) is anatomically and functionally intertwined with the brain area which is arguably the major recipient of stress-related information: the paraventricular nucleus of the hypothalamus (PVN, Figure 1). Other hypothalamic nuclei also impinge upon the LC. Among them the hypocretin-expressing nuclei in the lateral hypothalamus (Henny et al., 2010; Carter et al., 2012). The hypothalamus-LC axis controls input and output information from and to the autonomic system through the brainstem, to and from the neuroendocrine system through the pituitary gland and numerous gland-to-brain biochemical feedback loops, as well as all the rest of the central nervous system (CNS), through brain and spinal cord volume transmission (Figure 1).

Further hints of the biological importance and pervasiveness of central adrenergic function come from the analgesic properties of NE and its agonists (Simpson and Lin, 2007), by its important role in the control of body temperature, like during inflammatory response (Bencsics et al., 1995; Ordway et al., 2007; Osaka, 2009), as well as from the observation that other relevant input to the LC originate from nuclei coordinating vital functions like sex/reproduction (Nucleus Paragigantocellularis), respiration (Parabrachial and Solitary Tract Nuclei), and vestibular balance control (Simpson and Lin, 2007).

The control of the above functions is likely retained in the evolution from lower to higher mammals (Tohyama et al., 1974), but the increase in brain size and complexity associated with its disproportionate anatomical development makes the mammalian CNS particularly vulnerable to sudden surges in energy consumption caused by stressful situations. This latter evolutionary purpose might have further strengthened the importance of LC as master energy hub (Berridge and Waterhouse, 2003; O'Donnell et al., 2012), enhancing—particularly in humans—its role in the etiology of stress-related conditions.

Many hormones have a potential for global control of energy expenditure and activity regulation. Among them—for instance—the corticosteroid system is well placed for exerting a global and sophisticated biochemical regulation of energy demand and distribution (De Kloet, 2004), but lacks the property of anatomical and functional contiguity that the LC possesses. For similar reasons, the cytokine network, which also has the potential to control the brain (and bodily) global energy distribution (Guijarro et al., 2006), also does not seem to qualify as “central energy master.”

Central cholinergic fibers made up a highly divergent and almost ubiquitous release system (McKinney and Jacksonville, 2005; Smythies, 2005a). However, the existence of a large number of nuclei and brain areas that independently control the release of acetylcholine (Ach; Nucleus Basalis of Meynert, medial septum, latero-dorsal tegmentum, etc.), each toward or within their respective anatomical targets suggests that the cholinergic system hardly exerts a genuinely centralized control of energy expenditure.

The central histaminergic system stemming from the tuberomammillary nucleus of the hypothalamus appears to play a powerful and genuinely centralized role in triggering an emergency and alert maintenance response (Wada et al., 1991; Sakata et al., 1997; Shan et al., 2015). To our present knowledge, though, the histaminergic system does not appear to display a repertoire of cellular and synaptic actions paralleling the complexity and flexibility of the LC/adrenergic system, which is perhaps rivaled in its pervasiveness and variety of effects only by the Raphe/5HT system (Heisler et al., 2003; Zhou et al., 2005). In this respect, only serotonergic projections from the Raphe nucleus and the cholinergic fibers from the basal forebrain (Nucleus Basalis of Meynert) reach the extent and density of LC adrenergic projections throughout the CNS (Smythies, 2005b). The importance of the serotoninergic system in the coordination of the stress response has been reviewed elsewhere (Waselus et al., 2011).

While we highlight the importance of developing a comprehensive theory integrating the roles of the many neurotransmitters involved in the stress response, we will henceforth limit our discussion on the role of NE. In the following sections we will discuss experimental evidence relating central NE function to activities related to stress (stress perception, elaboration, and execution of a stress-ridding plan, as well as storage—or deletion—of related memories), and will make an attempt to integrate previous literature into a qualitative model in which increasing levels of NE co-ordinate the activity of different brain areas, inducing global brain states with increasing energy consumption and stress levels. We will only briefly mention the effects of NE on long-term processes, which we have recently reviewed elsewhere (Salgado et al., 2016).

It is worthwhile emphasizing the genuinely global nature of NE, differing from its chemical precursor dopamine, whose cortical projecting axons target more selectively the prefrontal cortex (Robbins and Arnsten, 2009). For the sake of clarity, we would like to highlight that stress activates two distinct pools of NE: a central one and a peripheral one, the latter associated with sympathetic nervous system activation. While the interaction between the two pools is essential to the understanding of the systemic effects of stress, only the former will be considered in the present discussion.

For the purpose of this discussion, we will broadly define stress as any situation in which an organism increases its energy consumption beyond an expected or biologically bearable range (which greatly varies even among individuals of the same species), and as stressor its objective or perceived source. We will get back to this definition of stress in Section LC-CNS Interactions. In the presence of most types of stressors the brain carries out the following (conscious or subconscious) functions: (1) evaluation of the stressor characteristics (short- or long-term intensity, duration, and consequences), (2) elaboration of a strategy to eliminate the stress(or), (3) execution of such plan, and (4) long-term storage (or in some case erasing) of stressor-related memories. Among these memories are the inner representations of the stress as a measure of potential danger, as well as the representation of one or more actual stress exit strategies, and their perceived effectiveness (or lack thereof).

In the historical context of the studies of the stress response, a critical element of adrenergic effects had been recognized early in the interaction between the hypothalamus-pituitary-adrenal gland (HPA) neuroendocrine axis and LC reviewed in Gold and Chrousos (2002) and Gold (2015). In fact, NE-releasing neurons of the LC are important targets of corticotropin-releasing hormone (CRH)-producing hypothalamic neurons from the PVN (Nicolaides et al., 2015), as well as from other limbic areas including the amygdala (Ordway et al., 2007), whose activity in turn stimulates LC leading to NE release. The importance of the PVN-LC axis is underscored by the observation that the inactivation of glucocorticoid receptors in the LC induces depression-like symptoms in a mouse model (Chmielarz et al., 2013).

A critical feature of the well-studied HPA response to stress is the negative feedback between the production of glucocorticoids and the activation of the HPA axis, which occurs both at the level of CRH-producing neurons in the PVN of the hypothalamus as well as in pituitary ACTH-producing corticotrophs (Figure 2). An energetically meaningful consequence of the elevation of glucocorticoid levels is the parallel shut-down or at least decrease of the high-energy consuming immune adaptive system, which in turn may increase the chance of infection and cancer in chronically stressed individuals (Reiche et al., 2004).

Importantly—unlike the glucocorticoid negative-feedback on HPA axis—the activation of the PVN/CRH/NE/LC branch of the stress response not only does not produce a negative feedback (Gold, 2015), but produces a positive feedback which opposes and jeopardizes the closure of the HPA loop associated with glucocorticoids (Figure 2). The presence of a feedforward loop between CRH-producing areas of the cortex and of the hypothalamus and the LC is a risk factor in the induction of maladaptive plasticity of the stress system, which greatly enhances its vulnerability to intense and/or chronic challenge. Systemic inflammation can be considered as the opposite phenomenon, whereby a combined action of pro-inflammatory cytokines induces a temporary state of physical apathy and inaction (Haroon et al., 2012; Miller et al., 2013). A consequence of such “sick response” is to spare systemic energy and promote a prompt recovery of the organism affected by a viral or bacterial infection.

Decade-long seminal work from the group of Aston-Jones and Waterhouse provides a solid ground for assessing the basic functions and activity dynamics of the LC (Rajkowski et al., 1994; Aston-Jones et al., 1998; Usher et al., 1999; Aston-Jones and Cohen, 2005; Aston-Jones and Waterhouse, 2016). Using mostly in vivo electrophysiological recordings from both primate and rodent models, this body of work has shown that LC displays virtually no activity during the sleep phase, whereas during the wake state it displays two emergent firing patterns: a tonic one, associated with the arousal level of the animal, and a phasic one, related with decision making and attention. Importantly, the extent of LC phasic firing appears to follow an inverse-U shape function with respect to the levels of tonic LC firing. In fact, while low levels of tonic firing—as during low arousal level—are insufficient to elicit a consistent behavioral response, intermediate tonic levels yield optimal phasic firing, whereas high levels of tonic firing—associated with high arousal and limbic activation—yield low phasic response as LC units firing activity gets closer to saturation. The eventual overall effect of NE depends on the spatio-temporal pattern of its volume release (Fuxe et al., 2010), on the nature of the cellular and synaptic targets affected, and on the type of receptor activated (summarized in Figure 3).

NE is released in the brain by axonal varicosities via volume transmission (Grzanna, 1980; Jones and Yang, 1985; Agnati et al., 1995). NE receptors have been first studied in the periphery and subsequently identified throughout the whole CNS, with different densities and regional specializations as reviewed earlier (Ramos and Arnsten, 2007). In terms of molecular effects, they can be categorized into three main groups, in descending order of affinity: α₂Rs, (≈50 nM), α₁Rs (≈300 nM), and βRs (≈0.7–0.8 μM) reviewed in Ramos and Arnsten (2007). Both α₂- and β-Rs are known to activate guanosine-dependent (G–)protein receptors, each affecting adenylyl cyclase in opposite directions, namely by decreasing (α₂Rs) or increasing (βRs) the intracellular concentrations of cyclic adenosine monophosphate (cAMP). In contrast, α₁Rs activate phospholipase C (PLC), thus triggering the synthesis of intracellular diacylglycerol and activation of protein kinase C as well as of inositol phosphate, which in turn releases Ca²⁺ from intracellular stores (Ramos and Arnsten, 2007).

The existence of widespread families of high-affinity neurotransmitter receptors (NE α₂Rs, M₂ muscarinic, D₂ dopaminergic) whose activation decreases adenylyl cyclase activity (G_i) suggests that basal (tonic) cytosolic levels of cAMP are not zero, and that they concur to the regular maintenance cellular processes active during cell rest. As a corollary, we hypothesize that the inactivity of the CNS noradrenergic system—similar to that of the other monoaminergic and the cholinergic system—during sleep is associated with a non-zero level of cAMP (Figure 4), and a tonic level of cellular energy expenditure. Such ground-level of cellular metabolic activity is possibly necessary to carry out a number of functions including a temporary enhancement of immune function during the resting phase (Kamath et al., 2015) and memory consolidation (Wilson and McNaughton, 1994; Barnes and Wilson, 2014; Figure 3). A slight increase in the concentration of neurotransmitters activating a G_i (in the case of NE up to 100 nM) could be sufficient to shut down such sleep-associated maintenance cellular processes and re-direct cellular metabolic energy to the quiet-wake related activities. Only relatively higher NE concentrations (around or above 0.4 μM) would be able to solidly activate the PLC cascade and increase cAMP levels above its resting levels (Figure 3), increasing the cellular supply for more energetically demanding biological activities (Figure 4).

Cellular and synaptic adrenergic modulation (Salgado et al., 2016) suggests that the activation of the cortical branch of the LC/adrenergic system could simultaneously perform two tasks: (1) single neuron activation by modulation of intrinsic conductances, with consequent local mobilization of large amounts of metabolic energy, and (2) temporary shut down or depression of the activity of other cortical areas that are unnecessary or irrelevant to the resolution of a particular behavioral contingence. This end could be achieved by a combination of selective depression of excitatory and enhancement of inhibitory synaptic transmission (Waterhouse et al., 1991; Sessler et al., 1995; Salgado et al., 2016).

Heterogeneous mechanisms of adrenergic modulation in different cortical areas like sensory vs. prefrontal cortices (Salgado et al., 2011, 2012b; Roychowdhury et al., 2014) together with activity-dependent modulation may concur to a functional selective enhancement or depression of neuronal activity in specific areas (Hains and Arnsten, 2008; Arnsten et al., 2012; Edeline, 2012; Roychowdhury et al., 2014). In the following sections we will review experimental evidence of effects of the noradrenergic receptor families mentioned above, from clinics and animal models, in an attempt to condense the related information in an integrated view. In many cases it will be hard to guess how a particular cellular or synaptic phenomenon participates into a behavioral function. In the tables we will either report the author's interpretation of their experimental finding, or will formulate a plausible one, keeping in mind that the same cellular experimental data may play different roles in a systemic function.

Based on pharmacological observations in clinics and in animal models, the activation of α₂Rs is deemed necessary for optimal performance in working memory and other tasks carried out with a strong prefrontal cortex (PFC) component (Gamo and Arnsten, 2011; Arnsten and Jin, 2012, 2014). Along with promoting working memory—and possibly related to it—α₂Rs also appear to contribute to a plethora of functions such as attention and impulse inhibition (Brennan and Arnsten, 2008; Robbins and Arnsten, 2009). The activation of α₂Rs decreases the inhibitory synaptic drive onto the tuberomammillary nucleus of the hypothalamus, contributing to alertness (Nakamura et al., 2013). α₂R activation in the medial septum and hippocampus increases theta rhythm (Kitchigina et al., 2003), presumably enhancing cognition. The increase in spontaneous inhibitory postsynaptic currents (sIPSCs) in the PVN following α₂R agonist application (Chong et al., 2004), together with the α₂R-induced decrease in glutamatergic drive onto the ventral tegmental area (VTA) may contribute at least part to α₂Rs anxiolytic properties.

Ample evidence exists that α₂Rs directly modulate PFC activity (Kovács and Hernádi, 2003; Wang et al., 2010), reviewed in Arnsten and Li (2005). Most of this literature indicates a beneficial effect of α₂R activation for working memory, although in some studies beneficial effects of α₂R blockers have been described (Brown et al., 2012; Bari and Robbins, 2013). The α₂R-induced block of N-methyl-D aspartate receptor (NMDAR)-mediated current (Liu et al., 2006), would corroborate a role in PFC learning for this receptor, but it could also represent a faster “clearance” of PFC reverberant circuits. Particularly remarkable is the evidence from multiple studies, of the beneficial effects of α₂R-induced block of dendritic hyperpolarization-activated cyclic nucleotide (HCN) channels (Wang et al., 2011; Zhang Z. et al., 2013), whose age-related decline is considered an important cause of cognitive deterioration (Wang et al., 2011).

α₂R activation does modulate motor activity (Villégier et al., 2003), although not always in the same direction (Cathala et al., 2002; Carey and Regehr, 2009). α₂R activation appears to modulate cerebellar activity, necessary for fine timing and control of distal movement (Hirono and Obata, 2006; Di Mauro et al., 2013; Lippiello et al., 2015). Little evidence is reported of α₂R modulation of sensory areas, mainly in auditory (Leão and Von Gersdorff, 2002; Salgado et al., 2011) and olfactory (Nai et al., 2009) areas.

Different—sometimes contradictory—evidence about the global effects elicited by α₂R ligands may perhaps be explained as the result of two contrasting actions on excitatory (pro-convulsive) presynaptic α₂Rs and inhibitory (anti-convulsive) postsynaptic α₂Rs, as revealed by an epilepsy study (Szot et al., 2004). It is worth mentioning that the use of tricyclic medication, used as antidepressant, may induce α₂R internalization (Cottingham et al., 2015), perhaps indicating that depression may be associated with or even caused by an increase in α₂R expression, possibly elicited by a high NE tone associated with prolonged stress.

Overall, the activation of α₂Rs by an increased but moderate NE tone (possibly ≤ 100 nM), appears to increase alertness, improve working memory, attention, PFC function in general, and enhance fine motor control and sensory processing, possibly acting on pre- and post-synaptic receptors carrying out opposite functions. Table 1 reports a series of α₂R-mediated effects grouped per putative function.

In general, α₁Rs activation promotes wake and activity by directly affecting neurons (Schmeichel and Berridge, 2013; Igata et al., 2014), and possibly also by activating astrocytes (Pankratov and Lalo, 2015). Activation of α₁Rs also concurs to the anorexigenic effect of NE and amphetamines (da Silva et al., 2014).

A specific and consequential effect of α₁R activation is emotion control. The decrease of the inhibitory drive onto the VTA may indicate an increase in motivation (Velásquez-Martínez et al., 2015). α₁Rs are also clearly involved in the stress response, as revealed by acute restraint stress (Alves et al., 2014), predatory stress (Rajbhandari et al., 2015), and maternal separation (Coccurello et al., 2014) studies. In agreement with its role in the stress response, block of α₁Rs impairs HPA activation (Yang et al., 2012).

α₁R positive modulation of working memory and other PFC function also seems to be solidly established by a wealth of data. For instance, α₁R activation improves working memory deficit induced by applications of the GABA_AR agonist muscimol (Hvoslef-Eide et al., 2015), while α₁R block disrupts the “go” performance in a “go-no-go” task (Bari and Robbins, 2013). At the synaptic basis of these effects could lay an enhancement in glutamatergic function (Luo et al., 2014b, 2015a), which may, in turn, yield a general increase in firing frequency in the PFC (Zhang Z. et al., 2013). Other cellular and synaptic effects of α₁R activation in the PFC, like an increase in inhibitory drive (Luo et al., 2015b) or a specific decrease in NMDAR-mediated response are not necessarily prone to similarly straightforward interpretations.

Motor effects of α₁R activation appear to be associated with a generalized increase in motor activity (Villégier et al., 2003), accompanied by a reduced fine motor control (Aono et al., 2015), suggested also by improved rotarod performance of KO α₁R mice (Mishima et al., 2004). A decreased glutamatergic cerebellar drive may concur to a reduced distal motor control (Lippiello et al., 2015).

Even more puzzling are the effects of α₁R activation on sensory activity. While the effectiveness of α₁R activation on sensory areas appears to be well established, its overall function remains enigmatic, possibly due to opposite effects on excitatory and inhibitory synaptic systems, as well as to a genuine heterogeneity of the response to different sensory modalities. For instance, α₁R activation increases firing in the somatosensory cortex (Devilbiss and Waterhouse, 2000), but decreases firing frequency and responses to glutamate in the visual (Terakado, 2014) and in the auditory cortex (Manunta and Edeline, 1997; Dinh et al., 2009). In the latter—in turn—activation of α₁Rs elicits opposite responses on electrically-evoked GABAergic transmission originating from different cortical layers (Salgado et al., 2011, 2012a). αRs (possibly α₁Rs) are involved in auditory cortex activity-dependent plasticity evoked by electric or optogenetic stimulation of LC (Martins and Froemke, 2015).

The olfactory bulb is also not exempt from displaying apparently contrasting α₁R-induced effects, like an increase of GABAergic response (Zimnik et al., 2013) and membrane depolarization (Nai et al., 2009). The effects of α₁R activation on sensory areas may be related to sensory modality selection after adrenergic activation, and/or maintenance of the excitatory/inhibitory balance following intense activation.

On the other hand, α₁R activation elicits clearly positive effects on memory and learning (Doze et al., 2011), as corroborated by studies on constitutively active α_1AR mentioned earlier (Collette et al., 2014), and by a worsened learning and working-memory related performance in KO α₁R mice (Spreng et al., 2001). An increase in rebound excitation and neuronal ensemble synchronization associated with an α₁R–mediated increase in GABA release in the entorhinal cortex (Lei et al., 2007; Cilz et al., 2014) may be at the root of at least some of the α₁R–induced improvements in learning and memory.

Depression, psychosis, and numerous treatments for stress-related neuropsychiatric disease appear to modulate importantly α₁Rs expression and function, although the direction of such modulation is not always consistent with illness or therapeutic effects. For instance, long-term administration of imipramine or electroconvulsive therapy increase the expression of α₁Rs (Nalepa et al., 2002), but the antidepressant effects of other tricyclic antidepressants (TCAs) (Ramakrishna and Subhash, 2012) or quetiapine (Nikiforuk, 2013) reduce α₁Rs expression. The interpretation of these results is further complicated by the age-dependence of α₁Rs function (Deupree et al., 2007). Table 2 summarizes some of the systemic and cellular effects associated with α₁Rs activation.

Many functions identified for αRs are also brought about by βRs activation. One of them is wake and alertness (Schmeichel and Berridge, 2013). Especially interesting is the effect of βRs activation on astrocytes (Song et al., 2015; Dienel and Cruz, 2016; Sherpa et al., 2016), which induces a decrease in extracellular brain volume. βR are also neuroprotective (Laureys et al., 2014), and decrease endotoxin-induced toxicity (Jiang et al., 2015), possibly by eliciting process retraction in resting microglia (Gyoneva and Traynelis, 2013), in contrast with the induction of neurite growth in cultured cortical primary neurons (Day et al., 2014).

βRs exert their effects in many sensory areas including the somatosensory cortex (Devilbiss and Waterhouse, 2000), the visual cortex (Terakado, 2014), the auditory cortex (Manunta and Edeline, 2004; Salgado et al., 2011), cochlear nucleus, lateral lemniscus, inferior colliculus (Wanaka et al., 1989), and the olfactory bulb (Shakhawat et al., 2015). Activation of βRs impairs sustained attention (Bari and Robbins, 2013), and increases the power (but not the frequency) of γ-oscillations (Haggerty et al., 2013), apparently without impairing cognitive flexibility (Steenbergen et al., 2015).

Similar to the sensory effects of αRs described in the previous sections, the effects of βRs do not necessarily appear to converge onto an unequivocal single function, representing either genuine differences between sensory areas, or recovery of the excitation/inhibition balance through adjustment of synaptic strength or other cellular mechanisms.

The body of knowledge concerning the effects of βRs on a variety of limbic functions is remarkably consistent with the hypothesis that a high concentration of tonic NE is critical for eliciting or modulating emotion. Fear memory—for instance—is impaired after administration of βR blockers (Fitzgerald et al., 2015; Zhou et al., 2015), and βR activation interferes with fear extinction induced by novel stimuli (Liu et al., 2015). Interestingly, social stress generates microRNA which decreases fear response acting on βRs (Volk et al., 2014). These data indicate that βRs activation is unequivocally associated with fear and fear memory, most likely because of their high expression in the amygdala. βRs are also causally related to anxiety generation, as suggested by the anxiogenic effect of βR-agonist administration (Hecht et al., 2014), and corroborated by elegant experiments where βRs were activated via optogenetic means (Siuda et al., 2015). Interestingly, βR blockers also reduce the anxiogenic effect of cocaine intake (Wenzel et al., 2014), while βR agonist administration within the pre-Botzinger complex increases spontaneous sigh frequency (Viemari et al., 2013).

Perhaps not surprisingly, βRs are important mediators of stress effects. For instance, restraint stress reduces dopaminergic effects but not after blocking βRs (Chang and Grace, 2013). Along the same line, stress elevates LC release of NE, leading to desensitization of βRs, an effect that—if chronic—may give rise to a depressive behavioral phenotype (Porterfield et al., 2012). Motor function appears to be improved by βR activation. βR agonists increase cerebellar GABA input (Di Mauro et al., 2013), increase excitatory input to Purkinje cells, and decrease the threshold for cerebellar LTD (Lippiello et al., 2015), although worsening spatial orientation performance (Robinson et al., 2015).

A wide experimental database supports a positive effect of βRs in learning and memory (Salgado et al., 2016). For instance, βR activation increases long-term potentiation (LTP) in the hippocampus and in the neocortex (Laing and Bashir, 2014; Hansen and Manahan-Vaughan, 2015; O'Dell et al., 2015) and memory retrieval, possibly by shutting down an after hyperpolarization activated (AHP) current (Zhang L. et al., 2013; Zhou et al., 2013), by “unsilencing” of silent synapses (Rozas et al., 2015), but also by inducing hippocampal long term depression (LTD; Goh and Manahan-Vaughan, 2013; Lethbridge et al., 2014).

βRs involvement in long-term synaptic plasticity is further indicated by the increased predisposition to long-term changes in both GABAergic synapses (Inoue et al., 2013) and glutamatergic synapses (Maity et al., 2015) following exposure to βRs or stress, in a βR-dependent manner (Grigoryan and Segal, 2013; Grigoryan et al., 2015).

The role of βRs in neurologic and psychiatric disease is an example of bell-shaped curve: on one hand, a βR deficit is associated with CNS malfunction and impairment, on the other one, it is a βRs hyperactivation causes distress and neurodegeneration. The former instance is epitomized by the βRs deficits associated with decreased LC-adrenergic function in aging (Santulli and Iaccarino, 2013). The finding of antibodies against βRs in the plasma of chronic fatigue syndrome (Loebel et al., 2016), and the improvement in memory (Dang et al., 2014) and cognitive performance (Phillips et al., 2016) in Down syndrome patients treated with βR agonist highlight the global importance of the βR-dependent component of noradrenergic transmission. The involvement of βRs in Alzheimer disease (AD) symptomatology is somehow controversial. For instance, βR activation appears to increase tau-protein phosphorylation—one of the hallmarks of AD (Wang et al., 2013), while routine presentation of novel stimuli is reported to protect from the toxicity from β-amyloid—another important AD marker—through βR activation (Li et al., 2013).

A βR-dependent increase in excitatory transmission in the bed nucleus stria terminalis (BNST) has been interpreted as distress factor in drug addiction seeking behavior (Egli et al., 2005), while βR block has been proposed as treatment for depression-related allodynia (Barrot et al., 2009). A possible general interpretation of the body of work related to the function of βRs in the context of stress is that short-term, acute, activation of βRs promotes demanding performances, whereas their chronic stimulation may lead to detrimental consequences of the same functions promoted by βRs short-term action. Table 3 summarizes recent experimental work on βR central function.

The high expression and high functional relevance of βRs in the amygdala and overall in the initiation of stress response would prompt them as target for pharmacological intervention in the treatment of stress-related psychiatric illness. Administration of the βR blockers, indeed, does decrease the behavioral and biochemical effects of social stress (Wohleb et al., 2011), of restraint stress (Tamburella et al., 2010), and shock-probe defensive burying response (Bondi et al., 2007), possibly by inhibiting cytokine release from microglia, among other effects (Wohleb et al., 2011). Promising results in the treatment of acute effects of stress come from the development of the blood-brain permeable β₃R agonist amibegron (Stemmelin et al., 2008). Activation of βRs may be an important therapeutic component of the antidepressant effect of mirtazapine (Rauggi et al., 2005).

An old hypothesis positing that the therapeutic effect of antidepressant was due to downregulation of βRs (as elicited by TCA treatment; Peet and Yates, 1981) has long been discarded (Charney et al., 1986). βRs agonists have been proposed also in the treatment of the memory impairment associated with psychotic schizophrenia, but the detrimental effects of βR-agonists on working memory and general cognitive flexibility, prevent their routine use (Friedman et al., 2004). In spite of a clear involvement of βRs (and CRH receptors) in amygdala activation in the etiology of PTSD, less or no effective has been the use of βR blocker in the long-term treatment of post-traumatic stress disorder (PTSD; Amos et al., 2014) and schizophrenia (Wahlbeck et al., 2000). βR blockers lack of effectiveness may perhaps be explained by the occurrence of βR internalization induced by their persistent activation. βR internalization was one of the first β-arrestin mediated processes to be described (reviewed in DeWire et al., 2007). Neurons have among the highest expression of non-visual β-arrestin in the whole mammalian body (Gainetdinov et al., 2004). Stress activates β-arrestin mediated internalization of βRs as well as internalization of corticotropin-releasing hormone (CRH) type 1 receptors (Hauger et al., 2009). Either process is an important mechanism of neuronal desensitization to stress response. A related third neuronal desensitization process is the G-protein receptor kinase (GRK)—mediated switch of G-protein functioning from its classic pathway (adenylyl cyclase activation through Gs, in the case of βRs) to a ERK-only pathway (Hauger et al., 2009). This process prevents short-term action of βRs (G_s-induced activation of adenylyl cyclase) but potentially triggers longer-term mechanism like mitogen activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK), synaptogenesis, and, possibly, maladaptive synaptic plasticity.

Good evidence exists for gap-junction mediated synchronous LC activation (Ishimatsu and Williams, 1996; Rash et al., 2007). Computational modeling supports the hypothesis of simultaneous activation of LC neurons, and simultaneous increase in brain NE (Gao and Holmes, 2007; Patel and Joshi, 2015), although alternative hypothesis have been proposed (Chandler et al., 2014a,b; Chandler, 2015). These observations suggest that different brain states may be elicited by increasing NE concentrations progressively activating ARs from high to low affinity for NE. In each of these states a different combination of tonic and phasic NE levels would give rise to regional differences in the brain activity, as well as to specific patterns of global function. In this section we will describe a largely speculative proposal for a sequence of cortical states in the order of progressively higher energy demand (Constantinople and Bruno, 2011), higher CNS NE levels, and progressive binding to the sequence of adrenoceptors α₂R, α₁R, and βRs in the order of affinity from the highest to the lowest (Figure 6). The intensity of gray indicates the level of physiological (non-pathological) activation, the thickness of the arrows indicates the strength of the connectivity between areas.

Sleep is the lowest-energy state, corresponding to a virtual absence of adrenergic tone. The whole organism—particularly the CNS—refills metabolic energy stores depleted during the previous wake phases. During this metabolic stage the organism is able to carry out important low-energy demand functions which do not require behavioral performance, like maintenance immune functions, rehearsal of mnemonic segments (Foster and Wilson, 2006) or motor sequence silent replay (Barnes and Wilson, 2014), aimed to synaptic stabilization and episodic or procedural memory consolidation.

The second state in the energy-demand ladder (quiet wake) is associated to low LC tonic firing, an active reward system, and mostly positive emotion in limbic areas. In this circumstance, the VTA releases dopamine that stabilizes the motivation axis represented by LC-Nucleus Accumbens-PFC receiving further input from the limbic cortices, and producing an optimal balance in the activity and reciprocal interaction among cortical areas and between cortical and subcortical regions. This state is characterized by optimal and flexible exchange of information between cortical areas and a relatively low physical and mental energetic load. In terms of LC activation/NE release, is associated with low but not nihil LC tonic firing and a high dynamic range for phasic LC responses to novel or salient stimuli, while, in terms of adrenoceptor activation, corresponds to tonic activation of α₂ adrenoceptors, sporadic activation of α₁Rs, and memory-promoting activation of βRs during phasic LC activation.

Tonic NE concentrations in the α₁R activation range would promote a third condition (active wake) represented by a series of states characterized by selective activity-dependent enhancement of energy consumption in particular cortical areas. Such areas would be selected depending on the specific demands of the circumstance, driven by a relatively high emotional tone in limbic areas. For instance, during strenuous physical activity, strong sensory engagement, or critical behavioral planning, NE released from LC would selectively depress the activity in non-critical cortical areas in an α₁R-dependent fashion, through depression of glutamatergic synapses (Dinh et al., 2009; Roychowdhury et al., 2014). At the same time, NE release would enhance energy consumption in critical cortical area(s) (i.e., motor controlling, sensory areas, working memory, or other brain areas) in an activity-dependent fashion, through a combination of phasically activated α₁- and β-AR activation. This state would be associated with a relatively high tonic level of LC firing, a limited range of phasic LC responses to salient or new stimuli, and a decreased but still functional communication between different cortical areas. In the “active wake” state, an in-built circuitry, prepared by evolution for automatic processing, would promote strong but not overwhelming activity in sensory areas, motor areas, as well as in the PFC. The latter would elaborate variable strategies to resolve the specific demands of the contingency for which automatic processing is not effective. In social mammals, extinguishing a stressor may require an additional inter-individual interaction (social) component that overburdens the limbic system and is therefore especially vexing on the individual.

At the physiological highest level of energy consumption, the LC would display the strongest tonic activation, high tonic NE release, and a limited or inexistent range of phasic NE release, corresponding to the “fight-or-flight” (FoF) response described in the pioneering work of Cannon reviewed in Fee and Brown (2002), at the extreme of which could lay the berserk condition. This state is characterized by strong activation of cortical βARs, strong negative or positive emotion, impulsive response, deficient sensory activity, shut-down of the planning (PFC) areas, and scant or inefficient intra-cortical communication (Holmes and Wellman, 2009). This condition would be terminated with either of two outcomes: (1) the elimination of the stressor, resetting of LC activity to low tonic state, reactivation of a temporarily inhibited dopaminergic system, and return of the system to a low-energy state (sleep or idle wake), or, on the opposite end (2) failure to eliminate the stressor, depletion of organic energy reserve, and, possibly onset of long-term deficit or even death. In humans, this condition may give rise to neuropsychiatric disorder including epilepsy, burn-out syndrome, psychosis, depression, or anxiety, depending on the stressor pattern, individual genetic predisposition, and previous life history.

The maximum duration and intensity of high-energy states (active wake and distress/FoF) bearable by a specific individual would display significant inter-individual differences related to genetics, previous training/experience, and motivational state, and strongly depends on the history of the subject, to the point that periodic and controlled incursions into the stress state may be beneficial to increase the probability of successfully extinguishing unexpected future stressors. The hypothesis is graphically summarized by the sketches in Figures 5, 6.

An oft found bell-shaped dependence of specific PFC-dependent performance on adrenergic activation (Roychowdhury et al., 2012; Sapolsky, 2015) could be interpreted as a transition from a low-energy state (state 2: quiet wake), to state 3 (active wake), associated with a larger energy mobilization, stronger engagement, and improved cognitive performance (left part of the bell shape curve). At the right end of the curve would lay the transition between state 3 (stress) and state 4 (FoF, right part of the bell shape), with a massive engagement of βRs in the cortex as well as in subcortical nuclei—particularly the amygdala—and consequent dysfunctional working memory, in favor of an optimal impulsive, automatic, motor response and full-fledged autonomic sympathetic response (Bouret and Sara, 2005; Hains and Arnsten, 2008; Gamo and Arnsten, 2011). Needless to say, a comprehensive theory of energy mobilization in high-demanding states (active wake and distress/FoF) should include the role of other global transmitters, particularly acetylcholine, histamine, and 5HT. Such discussion is left for important future work, and falls outside the scope of this review.

It is tempting to speculate further that the regional pattern of energy consumption in this condition may have changed in the course of mammalian evolution, and possibly along mankind history, such that the effects of stress on motor and sensory cortices used to be a lot more severe during early history/evolution/developmental stages, compared with the effects on the PFC, while the latter has become (is becoming) the major subject—and potential victim—of stress in modern society, particularly for adolescent and adult humans.

Stress notoriously impairs the dynamic balance between sympathetic and parasympathetic autonomic branches, affecting sleep, digestion, endocrine function, by altering the balance between peripheral parasympathetic and sympathetic tones (Grippo and Johnson, 2009; Silvani et al., 2016). Even more consequential, in the CNS, physiological stress of high intensity and/or prolonged duration may lead to β arrestin-mediated internalization of adrenergic receptors, studied in detail for βRs (Stone and Quartermain, 1999), leading in turn to a de-sensitization of βR-mediated central adrenergic pathways (Fu and Xiang, 2015). A possible consequence of intense or prolonged stress could be a decrease in effectiveness of the LC adrenergic system, leading to a decrease in the expression of the β₁R type (Porterfield et al., 2012) in limbic areas, and to a change in the expression of βR-related effectors in other parts of the limbic system like the hippocampus (Benes et al., 2004).

We represented the two poles of LC/NE function with two examples each, in Figure 7 (LC hypofunction) and Figure 8 (LC hyperfunction). In these figures, red and yellow represent pathologically hyper- or hypo-active areas, respectively. The intensity of the blue stripes inside LC sketch (light in Figure 7 and strong in Figure 8) represents the level of tonic LC activation. As example of LC hypofunction we represented attention deficit disorder with hyperactivity (ADHD), in which a monoaminergic hypofunction yields a hypofunctional PFC, which in turn fails to exert a satisfactory inhibitory control on automatic motor activity, associated with sensory distractibility (Figure 7, left). While other monoaminergic deficits (principally dopaminergic) are most likely also involved in ADHD, the efficacy of the NE-reuptake inhibitors like atomoxetine in ADHD treatment corroborates the notion that an LC/NE deficit is a critical component of this condition.

While the causes and mechanisms of clinical depression involve factors other than the LC/NE system, a similar pharmacological argument—the efficacy of selective serotonin/norepinephrine reuptake inhibitors (SNRIs)—also indicates that a deficient noradrenergic system plays a critical role in the treatment of this affliction. In our hypothesis, depression—like ADHD—is also associated with a LC/NE and PFC deficit, but, compared to ADHD, is associated with opposite roles of limbic (hyperactive in depression) and motor (hypoactive in depression) areas (Figure 7, right).

As examples of conditions associated with long-term consequences (maladaptive plasticity) of LC hyper-function we selected anxiety disorder (Figure 8, left) and psychosis (Figure 8, right). Anxiety disorders are characterized by hyperactivation of the limbic-sensory axis, with a prevalence of a reality-detached negative mood. Different anxiety disorders may be associated with different degrees of motor activation, ranging from aggression (like in post-traumatic stress disorder), to freezing (like in a rodent response to a predator). Remarkably, anxiety and depression would only differ in terms of motor areas (in)activation, suggesting that further maladaptive plasticity may quickly convert an anxious state into a full-fledged depression, and that the same subject may oscillate between two conditions, which could even take place simultaneously. The large co-morbidity of anxiety and depression (van Tol et al., 2010) corroborates our hypothesis.

An example of even more severe mental condition associated with LC hyperfunction is psychosis (Figure 8, right). This condition is associated with hyperfunction of most cortical areas, promoted by a high monoaminergic tone, leading, in turn, to severe PFC functional impairment. Signs of generalized cortical hyper-function are pathognomonic symptoms of psychosis, like paranoia and hallucinations, as well as aggression. In this condition, prolonged and/or intense stress elicits a type of maladaptive plasticity that sensitizes limbic and sensory areas leading to loss of touch with reality, and—in the most dramatic cases—aggression and gross working memory impairment. Clinical support for a strong involvement of the LC/NE system—together with other monoaminergic systems—in psychoses, is the precipitation of psychotic episodes after intake of drugs including legal or illegal NE and other monoaminergic re-uptake blockers. Up-regulation of α₁Rs may be a component of PFC impairment observed in the ventral-hippocampal lesion model of schizophrenia (Al-Khairi et al., 2009).

Depression and psychosis share the traits of working memory impairment, some level of detachment from reality, and hypersensitivity of limbic and sensory system, all of which can be triggered by prolonged or intense stress. Clinically, these two conditions may thus represent the result of a parallel process of stress-induced maladaptive plasticity landing on opposite poles of motor drive because of genetics or cultural factor. The similarities between depression and psychosis may explain the presence of both conditions (plus bipolar disease) associated with genes including DISC1 and neuregulins (Blackwood et al., 2007). This hypothesis is further supported by the finding of a reduced monoaminergic drive in a DISC1 animal model displaying depressive symptoms (Lipina et al., 2013).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.