Modulation of anxiety circuits by serotonergic systems

Abstract

Anxiety is a complex emotional state associated with sustained heightened autonomic and behavioral arousal and an increase in avoidance behavior. Anxiety-related behavior is a form of risk assessment behavior that is associated with a level of uncertainty or unpredictability regarding the outcome of emotionally salient events, often when both rewarding and aversive outcomes are possible. In this review, we highlight recent advances in our understanding of the neural circuits regulating anxiety states and anxiety-related behavior with an emphasis on the role of brainstem serotonergic systems in modulating anxiety-related circuits. In particular, we explore the possibility that the regulation of anxiety states and anxiety-related behavior by serotonergic systems is dependent on a specific, topographically organized mesolimbocortical serotonergic system that originates in the mid-rostrocaudal and caudal parts of the dorsal raphe nucleus.

• 5-Hydroxytryptamine
• 5-HT
• serotonin
• anxiety
• basolateral amygdala
• dorsal raphe nucleus

Introduction

Aims and scope of review

Anxiety-related physiological and behavioral responses are mediated by specific neural circuits in the brain. Serotonin is an important modulator of these anxiety-related circuits but the mechanisms through which serotonin modulates these circuits and the associated physiological and behavioral responses are not fully understood (Iversen 1984; Handley 1995; Graeff et al. 1996; Graeff et al. 1997; Maier and Watkins 1998; Zhuang et al. 1999Gray and McNaughton 2000; Graeff 2002; Millan 2003; Maier and Watkins 2005). In this review, we have three aims. Our first aim is to present operational definitions of anxiety states and anxiety-related behavior in animal models. Our second aim is to briefly define neural systems involved in modulating anxiety- related behaviors. Our third aim is to consider how serotonergic systems specifically modulate these anxiety-related circuits. Anatomical evidence suggests that serotonergic systems are topographically organized within the brainstem raphe nuclei, with serotonergic neurons in different regions of the brainstem raphe nuclei projecting to different functional systems (Imai et al. 1986; Lowry 2002; Abrams et al. 2005). Consequently, we argue that in order to understand the role of serotonergic systems in the regulation of anxiety-related behavior and anxiety states, it will be important to characterize the anatomical distribution and the cellular and functional properties of serotonergic neurons that project to neural structures regulating anxiety-related behaviors. As it is not possible, within the constraints of this review, to consider the effects of serotonergic systems on all components of the neural circuits underlying anxiety-related behaviors in animal models and humans, we will focus on serotonergic systems that project to the basolateral amygdala, a nodal structure in the neural circuits modulating anxiety-related behavior in rodents (Millan 2003). In this review, we will not discuss anxiety disorders in humans, which are thought to involve dysfunction of neural systems, including serotonergic systems, underlying anxiety- or fear-related states. For this, the reader is directed to previous reviews on this topic (Nutt et al. 1990, 1998, 1999; Handley 1995; Bell and Nutt 1998; Nutt 2001, 2005).

Operational definition of anxiety states and anxiety-related behavior

Anxiety is a complex emotional state associated with sustained heightened autonomic and behavioral arousal and a sustained increase in avoidance behavior in environmental situations characterized by a level of uncertainty, unpredictability, or uncontrollability (Walker et al. 2003; Herry et al. 2005; Maier and Watkins 2005), including situations where a conflict between approach and avoidance exists (Millan 2003). An important component of anxiety states is that the associated increases in anxiety-related physiological and behavioral responses are not tied to specific environmental cues but are trans-situational (Maier and Watkins 2005), meaning that they can be elicited under diverse environmental contexts. Anxiety-related behavior is associated with a level of uncertainty or unpredictability regarding the outcome of emotionally salient events, often when both rewarding and aversive outcomes are possible. Anxiety-related behavior in animals, which includes increased vigilance and scanning of the environment, has been referred to as “risk assessment” behavior (Blanchard and Blanchard 1989). In a complex unpredictable environment, the individual may perceive simultaneously the potential for both rewarding and aversive outcomes of specific behavioral strategies. For example, a black-tailed prairie dog in the shortgrass prairie may desire to exit its burrow to forage on young grass shoots or explore the environment in search of other potential rewards; if upon emerging from the burrow it detects the odor of a fox, a natural predator, then a conflict arises. This conflict, between approaching a potential rewarding stimulus and avoiding a potential aversive stimulus, is central to many anxiety states.

Anxiety states may therefore involve unique neural mechanisms mediating vigilance and scanning behaviors in situations characterized by a level of uncertainty or unpredictability with respect to the outcome of events, leading to the identification of the valence of potential outcomes (positive or negative), and predictions as to the probability of rewarding or aversive outcomes based on different behavioral strategies. A cardinal feature of anxiety-related behaviors as they have classically been defined is that they are benzodiazepine-sensitive; i.e. benzodiazepine drugs can shift the balance between approach and avoidance behaviors, reducing avoidance behaviors and/or increasing approach behaviors (Gray 1982; Gray and McNaughton 2000). Other behavioral responses that have been associated with anxiety states, such as enhanced startle in a brightly lit open field environment, are also benzodiazepine-sensitive (Walker et al. 2003). Behavioral paradigms that involve purely avoidance responses (e.g. escape responses on an elevated T-maze), and therefore, do not involve a conflict between approach and avoidance behaviors or choices between different behavioral strategies, are not sensitive to benzodiazepine drugs (Graeff et al. 1998; Zangrossi, Jr. et al. 1999).

In the case of anxiety states that involve conflicts between approach and avoidance behaviors, there are multiple mechanisms though which increases in anxiety-related behavior can occur. For example, in an open field test, a commonly used measure of anxiety-related behavior in rodents, there is a conflict between a natural tendency to explore the central region of the novel environment (approach behavior) and the tendency to remain in the relative safety of the sides and corners of the apparatus (avoidance behavior). An increase in avoidance behavior (or a decrease in approach behavior) may be observed as a result of (1) activation of neural mechanisms regulating avoidance behaviors (e.g. increasing the perceived aversive nature of a stimulus or the perceived probability of an aversive outcome), (2) inhibition of neural mechanisms regulating approach behaviors (e.g. decreasing the perceived rewarding nature of a stimulus or the perceived probability of a rewarding outcome), (3) disruptions of the ability to accurately predict potential rewarding or aversive outcomes of specific behavioral strategies, or (4) disruptions of the ability to attach emotional salience to potential rewarding or aversive outcomes. Thus, anxiety-related behavior can be viewed as the output of neural mechanisms involved in assessing the valence, relative value, and emotional salience of stimuli, and the relative probability of the outcomes of specific behavioral strategies.

Experimental models of anxiety-related behavior

Many experimental models of anxiety-related behavior in rodents involve a conflict between approach and avoidance behaviors. In some cases, the conflict between approach and avoidance is immediately apparent. For example, in the Geller-Seifter Conflict Test and the Vogel Conflict Test, food or water deprived rats are simultaneously punished with a mild shock when performing behaviors to obtain a reward (e.g. food, or drink, respectively). In other cases, such as those based on exploration of a novel object or novel environment, the conflict is less obvious. A conflict exists between exploring a novel environment, a conspecific, or an object which is a potential source of reward and the real or perceived risks associated with exploring a novel environment. The balance between approach and avoidance behaviors can be manipulated by the experimenter by altering parameters such as light intensity (Valle 1970; Cunha and Masur 1978; Hall et al. 2000). For rodents, which are nocturnal animals, bright light is more aversive than low light and results in more avoidance behaviors (e.g. due to an increase in the perceived aversive nature of the stimulus). Conversely, in the Geller-Seifter or Vogel Conflict tests, the rewarding properties of food or water, respectively, are increased by food or water deprivation before the test, increasing baseline approach behaviors. The latter manipulations shift the baseline of responding by altering motivational aspects of behavior as opposed to specifically modulating anxiety-related mechanisms.

Although, animal models involving a conflict between approach and avoidance behaviors have been very useful in characterizing the anxiolytic and anxiogenic properties of drugs and in defining the neural systems underlying anxiety states and anxiety-related behaviors (Gray and McNaughton 2000; Millan 2003), this does not mean that a conflict between approach and avoidance is an essential feature of anxiety states or anxiety-related behaviors. It may be that uncertainty, unpredictability, or uncontrollability in the presence of a potential aversive outcome is sufficient to evoke an anxiety state (Walker et al. 2003; Maier and Watkins 2005), or that unpredictability per se is the essential stimulus leading to the development of an anxiety state (Herry et al. 2005). Recent studies in rats and humans suggest that an unpredictable, relative to a predictable, auditory stimulus leads to activation of the basolateral amygdala, a nodal structure in the neural circuits regulating anxiety-related behavior (see below), and an increase in anxiety-related behavior as measured in the elevated plus-maze (Herry et al. 2005). Activation of the afferents from the basolateral amygdala results in direct monosynaptic excitatory postsynaptic responses in layer V pyramidal neurons in the medial prefrontal cortex (mPFC; Orozco-Cabal et al. 2005), neurons that are known to project directly to the dorsal raphe nucleus and to control serotonergic neuronal activity (Hajos et al. 1998). Both the mPFC and the dorsal raphe nucleus have been implicated in regulating the behavioral consequences of unpredictable or uncontrollable stress (Amat et al. 2005; Maier and Watkins 2005). Thus, the basolateral amygdala, mPFC, and dorsal raphe may constitute an important circuit determining the behavioral consequences of unpredictable or uncontrollable stress, including increases in avoidance behaviors in anxiety tests (Amat et al. 2005; Maier and Watkins 2005).

Neural circuits modulating anxiety-related behavior

As defined above, anxiety-related behavior is a form of risk assessment behavior associated with a level of uncertainty or unpredictability regarding the outcome of emotionally salient events, often when both rewarding and aversive outcomes are possible. This suggests that neural circuits underlying anxiety-related behaviors may encompass neural circuits associated with approach or reward-related behaviors, avoidance behaviors, and decision-making based on the predicted probabilities of the outcomes of specific behavioral strategies.

The neural systems underlying approach behaviors or reward pathways are well characterized, and there are several excellent reviews on this topic (Berridge and Robinson 1998; Kalivas and Nakamura 1999; Dayan and Balleine 2002; Berridge 2004; Kelley 2004; Morgane et al. 2005). An important component of neural circuits underlying reward pathways and approach behaviors is the mesolimbic dopaminergic projection from the ventral tegmental area in the midbrain to the nucleus accumbens. The reward circuitry is modulated by important input from the prefrontal cortex, hippocampus, and amygdala (Kelley 2004; Morgane et al. 2005).

Attempts also have been made to understand the neural systems involved in the modulation of the conflict between approach and avoidance that is inherent in many forms of anxiety states (Gray 1982; Gray and McNaughton 1987; Gray and McNaughton 2000; Millan 2003). Intuitively, it seems that neural systems mediating approach and avoidance behaviors (particularly passive avoidance behaviors; c.f. Gray and McNaughton; Gray 1982) must be involved, but that higher order processes are likely to be involved in predicting outcomes of specific behavioral responses and resolving conflicts between two opposing behavioral strategies. For example, Gray has suggested that the septo-hippocampal system plays a critical role in the neurobiology of anxiety states and in particular that it plays a critical role in predicting the outcome of specific behavioral responses, (Gray 1982; Gray and McNaughton 2000), an idea elaborated in what is called the “comparator model of hippocampal function” (Gray 1983; Gray and McNaughton 2000; Vinogradova 2001). These ideas are consistent with recent findings that hippocampal “place cells” encode not only instantaneous place but also encode information about past and future events [e.g. prospective and retrospective encoding (Ferbinteanu and Shapiro 2003; Jeffery 2004)]. For example, some hippocampal place cells are modulated (either switched ‘on’ or ‘off’) depending on whether the rat is going to turn right or left in a maze to obtain a reward (prospective coding). This function of the hippocampus may be particularly relevant to risk assessment behavior where predictions about the potential outcome of different behavioral strategies (approach or avoidance) are required. In contrast, the basolateral amygdala may be involved in assigning emotional salience to both potential rewarding and potential aversive outcomes (see below), while the bed nucleus of the stria terminalis (BNST) is thought to be part of a lateral or specific emotional motor system (Holstege 1995) that is specifically involved in behavioral responses to diffuse aversive signals such as bright light (Davis 1998; Walker et al. 2003), and the mPFC is thought to be involved in behavioral suppression (Lacroix et al. 2000) and essential for the effects of controllability on dorsal raphe serotonergic systems and behavior (Amat et al. 2005). Finally, the anterior cingulate cortex, at least in humans, is thought to have a fundamental role in relating actions to their consequences, both positive reinforcement outcomes and errors, and in making choices between different behavioral strategies, based on their potential outcomes (Rushworth et al. 2004). Indeed, a distributed neural system is acknowledged by most investigators to play an important role in the regulation of anxiety-related behavior and anxiety states, including the septohippocampal system, basolateral amygdala, BNST, mPFC, and cingulate cortex (Davis 1998; Gray and McNaughton 2000; Millan 2003; Walker et al. 2003). Thus, neural circuits mediating anxiety-related behaviors are likely to encompass aspects of reward-related approach pathways and avoidance pathways but also to involve additional processes associated with predicting outcomes of specific behavioral strategies and making decisions on behavioral strategies based on the predicted outcomes.

The hypothesis that changes in anxiety states involve a highly distributed but anatomically interconnected neural system is supported by recent studies using immunohistochemical staining for the protein product of the immediate-early gene, c-fos. Immunohistochemical staining of Fos is a useful method for identifying neural pathways involved in responses to specific stimuli (Morgan and Curran 1991; Kovacs 1998). In a series of studies in rats, Singewald and Sharp demonstrated that injections of anxiogenic drugs (i.e. drugs that increase the subjective experience of anxiety in humans and increase avoidance behaviors in rodents in a variety of anxiety tests) with different pharmacological properties increase Fos expression in a distributed neural system (Singewald and Sharp 2000; Singewald et al. 2003). In the forebrain, all four anxiogenic drugs tested increased Fos expression in the central amygdaloid nucleus, the BNST, lateral septum, paraventricular nucleus of the hypothalamus, lateral hypothalamus, and the infralimbic and prelimbic regions of the mPFC. All drugs, excluding the serotonin agonist m-chlorophenylpiperazine (mCPP), also increased Fos expression in the basolateral and medial amygdala, the dorsomedial hypothalamus, cingulate cortex, and parts of the motor cortex (Singewald et al. 2003). Within the brainstem, all drugs increased Fos expression in the periaqueductal gray, locus coeruleus, lateral parabrachial nucleus, and nucleus of the solitary tract and all drugs except the cholecystokinin analogue BOC-CCK4 increased Fos expression in the dorsal raphe nucleus (Singewald and Sharp 2000). Most components of this distributed neural system had previously been identified as part of the neural network of brain structures involved in the integration and induction of anxiety states based on other methodological approaches (Gray 1982; Gray and McNaughton 2000; Millan 2003). Similar results have been observed in studies evaluating the effects of exposure to anxiogenic stimuli, such as the chemical component from fox feces, 2,5-dihydro-2,4,5-trimethylthiazoline (TMT) on c-fos mRNA expression in rats (Fendt et al. 2003; Day et al. 2004) (cf. (Duncan et al. 1996).

The basolateral amygdala as a nodal structure in anxiety-related circuits

Although it is recognized that the regulation of anxiety-related behavior will be the product of a distributed neural system, it will be instructive to look in detail at specific components of the system that appear to play a central role in these behaviors. As many anxiety-related behaviors involve conflict between approach and avoidance, it would be particularly instructive to focus on components of anxiety-related neural pathways that are involved in the attribution and/or association of emotional salience (of either a positive or negative nature) to specific environmental stimuli and that are connected anatomically and functionally to neural pathways mediating both approach and avoidance behaviors. One such structure is the basolateral amygdala. Indeed, based on lesion studies the basolateral amygdala has been identified as part of a system responsible for voluntary instrumental choice behavior based on emotional events. The lateral/basolateral nucleus appears to be critical for the sensory relay of the stimulus (LeDoux et al. 1990; Campeau and Davis 1995) and association and/or attribution of emotional salience to a variety of rewarding (Meil and See 1997; Ledford et al. 2003; Jackson and Moghaddam 2004) as well as aversive cues (Miserendino et al. 1990; Mayford et al. 1996; Rogan et al. 1997; Davis 1997; Shumyatsky et al. 2002). In simple terms, it has been suggested that the involvement of the basolateral amygdala in reward-related or avoidance-related processes is best characterized as relating to aspects of “liking” (Blundell et al. 2003) or “not liking” and the use of this information to modulate voluntary or intentional behavior (Killcross et al. 1997). There is also evidence that the basolateral amygdala is integral to emotional learning and disruption of this nucleus disrupts the memory enhancing effects of emotionally salient cues (Roozendaal et al. 1999; Schafe et al. 2000; Roozendaal et al. 2001; Ferry et al. 1999; McGaugh 2002; Goosens and Maren 2003; LaLumiere et al. 2003; Roesler et al. 2003). It is possible that different neurotransmitters act within the basolateral amygdala to differentially regulate approach and avoidance behaviors. There is accumulating evidence that the stress-related and anxiogenic neuropeptide corticotropin-releasing factor (CRF) acting within the basolateral amygdala may be critical for aversive or negative-cue induced behavioral responses (Lee et al. 1994; Heilig et al. 1994; Gray and Bingaman 1996; Sajdyk et al. 1999; Roozendaal et al. 2002; Heimer 2003). In contrast, dopamine in this nucleus may be more involved in the emotional salience elicited by reward cues (Ledford et al. 2003; Jackson and Moghaddam 2004). As would be expected based on these functional studies, the basolateral amygdala is anatomically integrated with neural pathways involved in both reward (Kelley 2004) and avoidance (Dong et al. 2001; Millan 2003). Consequently, we and others have focused on the potential role of the basolateral amygdala in regulation of anxiety-related behavior and anxiety states (Pesold and Treit 1995; Sanders and Shekhar 1995; Campbell and Merchant 2003; Etkin et al. 2004; Rainnie et al. 2004; Sajdyk et al. 2004; Abrams et al. 2005; Adamec et al. 2005).

Modulation of anxiety-related circuits by serotonin

In order to understand the role of serotonin in the regulation of anxiety-related behavior, it will be important to determine the effects of serotonin on the neural circuits that mediate anxiety-related behaviors. This represents a significant challenge for investigators. The effects of serotonin on these neural circuits are likely to depend on several variables. These include the time course of serotonin action, the extracellular concentrations of serotonin (which may determine which serotonin receptor populations are affected based on anatomical proximity to the site of release or the affinity of specific serotonin receptors for the ligand), the type and presynaptic or postsynaptic location of serotonin receptors within each specific brain region, systems level interactions among different nodes in the neural circuitry [e.g. interactions between the mPFC and the amygdala (Berretta et al. 2005; Pezawas et al. 2005), or the mPFC and the dorsal raphe nucleus (Amat et al. 2005)], differential proteolytic processing and co-release of peptidergic co-transmitters in serotonergic neurons (Blackmore et al. 2001), or context dependent changes in presynaptic and post-synaptic mechanisms regulating serotonergic signaling [initiated, for example, by glucocorticoid hormones (Stutzmann et al. 1998; Tronche et al. 1999)].

Despite this complexity, several lines of evidence suggest that serotonergic systems play an important role in modulating anxiety-related behaviors. In animal models of anxiety that involve conflict, drugs or lesions that reduce serotonergic activity can have anxiolytic effects that are qualitatively similar to the anxiolytic effects of benzodiazepines (Kahn et al. 1988; Graeff 1990; Griebel 1995; Graeff et al. 1996; Gray and McNaughton 2000). Microinfusions directly into the dorsal raphe nucleus of anxiogenic drugs that increase serotonergic activity, like the benzodiazepine receptor partial inverse agonists N-methyl-beta-carboline-3-carboxamide (FG-7142) (Lista et al. 1990), methyl 6,7-dimethoxy-4-ethyl-beta-carboline-3-carboxylate (DMCM) (Maier et al. 1995), and the anxiety-related neuropeptide urocortin 2 (Hammack et al. 2003; Amat et al. 2004; Maier and Watkins 2005) increase anxiety-related behavioral responses. In contrast, microinfusions of 5-HT or 5-HT_1A agonists into the dorsal raphe decrease anxiety-related behavioral responses, presumably via activation of somatodendritic 5-HT_1A autoreceptors on serotonergic neurons resulting in decreased 5-HT neurotransmission (Higgins et al. 1988; Higgins et al. 1992; Schreiber and De 1993). Likewise, intra-dorsal raphe microinfusions of other drugs expected to decrease serotonergic activity (e.g., the selective neurotoxin 5,7-dihydroxytryptamine, benzodiazepines, and the inhibitory neurotransmitter GABA) have all been reported to reduce behavioral measures of anxiety [see for review, Maier and Watkins (1998)]. Consistent with these findings, selective lesion of serotonergic systems in the amygdala results in an anxiolytic effect in a punished drinking test (Sommer et al. 2001). Together, these findings are consistent with the classic hypothesis that serotonergic systems can facilitate anxiety-related behavior (Wise et al. 1972).

Effects of serotonin within the basolateral amygdala

Evidence suggests that the basolateral amygdala is an important site for the effects of serotonin on anxiety circuits. Electrical stimulation of the rat dorsal raphe serotonergic pathways or direct iontophoretic application of serotonin inhibits spontaneous neuronal activity within the cortical and basolateral amygdala in vivo (Wang and Aghajanian 1977) and activation of serotonin 5-HT_1A and 5-HT₂ receptors within the basolateral amygdala alters the electrophysiological properties of identified basolateral amygdala interneurons and projection neurons in vitro (Rainnie 1999; Stein et al. 2000). These findings suggest that serotonin modulates basolateral amygdala function. Evidence for both anxiogenic and anxiolytic effects of serotonin receptor activation within the basolateral amygdala has been described. For example, bilateral intra-basolateral amygdala injection of mCPP, a mixed 5-HT₂ agonist, reduces exploratory behavior in an open field while increasing the latency to investigate a novel object. These responses are attenuated by prior treatment with SB-242084, a 5-HT_2C receptor-specific antagonist (Campbell and Merchant 2003). 5-HT_2C receptor activation within the basolateral amygdala may have long-term effects on anxiety-related behaviors as in vitro electrophysiological studies have demonstrated that 5-HT₂ receptor activation initiates long-term potentiation that can be prevented by a selective 5-HT_2C antagonist (Chen et al. 2003). Similarly bilateral intra-basolateral amygdala injection of 8-OH-DPAT, a 5-HT_1A receptor agonist, increased anxiety-like behavior in the social interaction test (Gonzalez et al. 1996). This response was reversed by treatment with a 5-HT_1A receptor antagonist. In contrast, intra-basolateral amygdala injections of 8-OH-DPAT had no effect on inhibitory avoidance behavior in the elevated plus-maze (Gonzalez et al. 1996), and had the opposite (anxiolytic) effect in the elevated T-maze test (Zangrossi, Jr. et al. 2001). These disparate findings might be expected as the basolateral amygdala is involved in assigning emotional salience to both potential rewarding and potential aversive stimuli and altering basolateral amygdala function is likely to modulate both approach and avoidance behaviors. In addition, it is possible that the actions of serotonin in the basolateral amygdala are involved in both the induction and the termination of anxiety-related behaviors via actions on different serotonin receptors or different postsynaptic targets. In summary, evidence suggests that 5-HT receptor activation within the basolateral amygdala does alter anxiety-related behavior, but additional studies will be required to develop a clear picture of the mechanisms involved. An important step in this direction is to understand the properties of serotonergic neurons that innervate anxiety-related circuits, including the basolateral amygdala. This includes their location within the brainstem raphe complex, their afferent regulatory mechanisms, and their unique molecular, cellular, and functional properties.

Functional subsets of serotonergic neurons

Serotonergic systems are topographically organized

Serotonergic systems within the brainstem raphe complex give rise to topographically organized projections to forebrain targets (Imai et al. 1986; Lowry 2002; Insel et al. 2003; Abrams et al. 2005). In other words, each region of the forebrain that is innervated by brainstem serotonergic systems has its own unique representation in the brainstem raphe complex. For example, neurons that project to the caudate putamen and substantia nigra are located within rostral regions of the dorsal raphe nucleus while those that project to the amygdala are concentrated in the mid-rostrocaudal region of the dorsal raphe nucleus, and those that project to the hippocampus are located in the most caudal regions of the dorsal raphe nucleus, the interfascicular part of the dorsal raphe, and the median raphe nucleus (Steinbusch et al. 1981; Köhler and Steinbusch 1982; Imai et al. 1986; Abrams et al. 2005). dorsal raphe neurons that project to the caudate putamen do not project to the hippocampus and vice versa (Imai et al. 1986). Individual dorsal raphe neurons projecting to the caudate putamen also project to other forebrain and brainstem structures involved in motor function including the substantia nigra (van der Kooy and Hattori 1980), and nucleus reticularis gigantocellularis pars α (Li et al. 2001). These findings suggest that individual serotonergic neurons give rise to collateral projections to distributed but functionally related targets. If we make the assumption that the subset of serotonergic neurons projecting to anxiety-related forebrain circuits plays a unique role in modulating anxiety-related behaviors and anxiety states, then these anatomical findings suggest that it may be possible to identify unique anatomical, morphological, and molecular properties of these serotonergic neurons.

Subsets of serotonergic neurons have different behavioral correlates

The presence of multiple subsets of serotonergic neurons with unique behavioral correlates within the dorsal raphe nucleus (Rasmussen et al. 1984; Sakai and Crochet 2001) is consistent with the hypothesis that subsets of serotonergic neurons may have functional properties linked to context-dependent forms of behavioral arousal (e.g. ambulatory locomotion versus anxiety-related behaviors). For example, a subpopulation (approximately 25% of those studied) of serotonergic neurons predominantly localized in ventral portions of the dorsal raphe nucleus is strongly activated in association with oral-buccal movements during chewing, licking, and grooming behaviors (Fornal et al. 1996). Subsequent studies in behaving rats have identified six different subtypes (types I-A, I-B, I-C, II-A, II-B, and II-C) of presumed serotonergic neurons based on their behavioral correlates (Sakai and Crochet 2001), several of which show distinct patterns of topographical organization within the dorsal raphe nucleus. The finding that anxiety-related behaviors are associated with increases in extracellular serotonin concentrations in the basolateral amygdala (Gonzalez et al. 2004), hippocampus, cortex, and septum, but not in the striatum (Beekman et al. 2005), is consistent with the hypothesis that mesolimbocortical and mesostriatal serotonergic systems can be independently regulated (Daugherty et al. 2001; Lowry 2002; Insel et al. 2004)

Distribution of serotonergic neurons projecting to anxiety-related circuits

Imai and colleagues have reported a rostrocaudal topography within the dorsal raphe with neurons projecting to the caudate putamen and substantia nigra located in the rostral region of the dorsal raphe nucleus, neurons projecting to the amygdala located in the mid-rostrocaudal dorsal raphe nucleus, and neurons projecting to the locus coeruleus and hippocampus in the caudal dorsal raphe nucleus (Imai et al. 1986). Although, it is unlikely that functional subsets of serotonergic neurons will be confined to subdivisions of the dorsal raphe nucleus that have been defined based on cytoarchitectonic criteria, these findings and others (Kohler et al. 1980; Ottersen 1981; Köhler and Steinbusch 1982; Köhler et al. 1982; Petrov et al. 1992, 1994; Van Bockstaele et al. 1993; Commons et al. 2003; Abrams et al. 2005) suggest that serotonergic neurons projecting to structures associated with neural circuits modulating anxiety states are concentrated within the mid-rostrocaudal and caudal dorsal raphe nucleus (Figure 1).

Figure 1 Diagrammatic illustration of the topographical distribution of serotonergic neurons projecting to the caudate putamen and substantia nigra (blue) and serotonergic neurons projecting to brain structures implicated in the regulation of anxiety-related behavior (yellow) in rat brain (for references, see text). Although this is an oversimplification, these cell populations have a differential distribution in the dorsal raphe nucleus (DR), with serotonergic cells projecting to the caudate putamen and substantia nigra located in more rostral regions of the DR. In some cases, individual serotonergic neurons project to functionally related targets [e.g. CP/ SN (van der Kooy and Hattori 1980), mPFC/ Acb (Van Bockstaele et al. 1993)]. In some cases, serotonergic neurotransmission of the dorsal part of the mid-rostrocaudal and caudal DR can be independently regulated (Macedo et al. 2005), suggesting that although the innervation arises in part from the same brain region, it could also arise in part from different populations of serotonergic neurons. In the mid-rostrocaudal DR, serotonergic neurons projecting to the CP are highly concentrated within compact cell clusters in the DRD “core” and DRV “ellipse” regions. Numbers below illustrations of rostral, mid-rostrocaudal, and caudal DR indicate stereotaxic coordinates with reference to Bregma. Scale bar, 500 μm. Abbreviations, Acb, nucleus accumbens; BL, basolateral amygdala; BNST, bed nucleus of the stria terminalis; CE, central amygdaloid nucleus; CP, caudate putamen; DA, dorsal hypothalamic area; DRD, DR, dorsal part; DRI, DR, interfascicular part; DRV, DR, ventral part; DRVL, DR ventrolateral part; LC, locus coeruleus; LPB, lateral parabrachial nucleus; mPFC, medial prefrontal cortex; SN, substantia nigra.

In order to increase our understanding of the distribution of dorsal raphe neurons projecting to neural circuits modulating anxiety-related behaviors, we conducted a detailed analysis of the distribution of dorsal raphe neurons that project to the basolateral amygdala (Abrams et al. 2005). Although, dorsal raphe neurons projecting to the basolateral amygdala were widely distributed throughout the rostrocaudal extent of the dorsal raphe nucleus, these neurons were concentrated in the dorsal part of the middle and caudal dorsal raphe nucleus. Consistent with this observation, following a small injection of retrograde tracer into the anterior part of the basolateral amygdala, retrogradely labeled cells were largely restricted to the dorsal part of the mid-rostrocaudal dorsal raphe nucleus (Figure 2). Comparison of the distribution of dorsal raphe neurons projecting to the basolateral amygdala with the distribution of dorsal raphe neurons projecting to other components of the neural circuits implicated in the regulation of anxiety-related behavior (particularly those neurons with collateral projections to the prefrontal cortex and nucleus accumbens) reveals that these neurons are concentrated within the dorsomedial part of the mid-rostrocaudal dorsal raphe nucleus (Figure 3).

Figure 2 Labeling of dorsal raphe neurons following injection of the retrograde tracer FluoroGold into the anterior part of the rat basolateral amygdala. (a) illustration of the distribution of retrogradely labeled cells within the dorsal raphe nucleus (coronal sections). Box in -8.00 mm illustration indicates the region illustrated at higher magnification in the photograph in (c). (b) photomicrograph illustrating the injection site restricted to the anterior part of the basolateral amygdaloid nucleus (-3.30 mm Bregma). (c) photomicrograph illustrating dorsal raphe neurons retrogradely labeled with FluoroGold. Retrogradely labeled dorsal raphe neurons were concentrated in the dorsal part of the mid-rostrocaudal dorsal raphe nucleus [gray shading in panel (a), levels -8.00 and -8.24 mm]. Numbers below illustrations in panel (a) indicate stereotaxic coordinates with reference to Bregma. Arrows in (c) indicate FluoroGold labeled neurons in the dorsal raphe nucleus. Abbreviations: BLA, basolateral amygdaloid nucleus, anterior part; BLP, basolateral amygdaloid nucleus, posterior part; BLV, basolateral amygdaloid nucleus, ventral part; BMP, basomedial amygdaloid nucleus, posterior part; CeC, central amygdaloid nucleus, capsular part; CeL, central amygdaloid nucleus, lateral division; LaDL, lateral amygdaloid nucleus, dorsolateral part; LaVL, lateral amygdaloid nucleus, ventrolateral part; LaVM, lateral amygdaloid nucleus, ventromedial part. Scale bar, (a) 873 μm, (b) 800 μm, (c) 50 μm.

Figure 3 Diagrammatic illustration of the distribution of dorsal raphe neurons projecting to anxiety circuits. (a) illustration (coronal sections) of the location of injections of retrograde tracer into the prelimbic (PrL) and infralimbic (IL) parts of the medial prefrontal cortex (mPFC, left), nucleus accumbens (middle), and the basolateral amygdala (right). (b) illustration of the distribution of retrogradely labeled neurons in the dorsal raphe nucleus. Open circles represent single-labeled neurons following injections of retrograde tracer in the mPFC; open triangles represent single-labeled neurons following injections of retrograde tracer in the nucleus accumbens (AcbC and AcbSh regions); grey squares represent double-labeled neurons following injections of retrograde tracers into the mPFC and nucleus accumbens; black circles represent single-labeled neurons following injections of retrograde tracer in the basolateral amygdala. The dorsal part of the mid-rostrocaudal dorsal raphe nucleus (gray shading) contains concentrations of neurons with collateral projections to both the mPFC and nucleus accumbens (grey squares), and neurons projecting to the basolateral amygdala (black circles). Abbreviations: ac, anterior commissure; AOP, anterior olfactory nucleus, posterior part; BL, basolateral amygdala; AcbC, nucleus accumbens, core; AcbSh, nucleus accumbens, shell; CPu, caudate putamen; f, fornix; mPFC, medial prefrontal cortex; LV, lateral ventricle; LS, lateral septum; mt, mammillothalamic tract; opt, optic tract. Scale bar, (a) 264 μm, (b) 1 mm. Adapted from Van Bockstaele et al. 1993 and Abrams et al. 2005, with permission.

Based on the knowledge that there is a topographical organization throughout the rostrocaudal axis of the dorsal raphe nucleus, we predicted that anxiety-related stimuli may have selective actions on topographically organized subsets of serotonergic neurons in regions of the dorsal raphe nucleus with projections to anxiety circuits. We investigated the effects of different anxiety-related stimuli on the expression of Fos throughout the dorsal raphe nucleus (Abrams et al. 2005; Staub et al. 2005). Increased behavioral arousal and vigilance behaviors following treatment with different anxiogenic drugs including the adenosine receptor antagonist caffeine, the serotonin 5-HT_2A/2C receptor agonist m-chlorophenyl piperazine (mCPP), and the benzodiazepine receptor partial inverse agonist N-methyl-beta-carboline-3-carboxamide (FG-7142) were associated with increases in Fos expression in the dorsal part of the mid-rostrocaudal and caudal dorsal raphe nucleus but not other areas of the dorsal raphe nucleus studied (Abrams et al. 2005). Likewise, treatment with the anxiety-related neuropeptide, urocortin 2 (a selective CRF type 2 receptor agonist) (Staub et al. 2005), or exposure to social defeat (Gardner et al. 2005), also selectively increased Fos expression in the same region (Staub et al. 2005). It should be pointed out that the studies using anxiogenic drugs did not verify that the drug-induced increases in Fos expression in brainstem serotonergic neurons were associated with increases in anxiety-related behaviors in validated behavioral tests of anxiety. Therefore, it is uncertain if the anxiogenic drug-induced increases in Fos expression are associated with (1) the performance of anxiety-related behaviors, (2) the development of an anxiety-like emotional state (see below), or (3) physiological or behavioral responses unrelated to the performance of anxiety-related behaviors or the development of a drug-induced anxiety-like emotional state. Future studies should be able to resolve these issues.

The dorsal part of the mid-rostrocaudal dorsal raphe nucleus

Although the neural mechanisms underlying this anatomical specificity are not clear, retrograde tracing studies have revealed that the dorsal part of the mid-rostrocaudal dorsal raphe nucleus is neurochemically distinct (Commons et al. 2003) and receives unique afferent input (Peyron et al. 1998). Following retrograde tracer injections in the dorsal part of the mid-rostrocaudal dorsal raphe nucleus, large numbers of retrogradely-labelled cells were seen in the lateral, ventral and medial parts of the BNST, whereas only a small to moderate number was visualized after injections in the other dorsal raphe subdivisions. Several lines of evidence suggest that the bed nucleus of the stria terminalis plays an important role in mediating anxiety-related behavior (Davis 1998; Gewirtz et al. 1998; Walker et al. 2003; Levita et al. 2004) and therefore the very specific afferents from the BNST to the dorsal part of the mid-rostrocaudal dorsal raphe region may account for the topographical specificity of the effects of anxiety-related drugs, the anxiety-related neuropeptide urocortin 2, or social defeat on Fos expression in serotonergic neurons within this region. Although, it is not possible, using Fos immunohistochemical methods, to establish cause and effect relationships, several lines of evidence suggest that serotonergic neurons within the dorsal part of the mid-rostrocaudal and caudal dorsal raphe may play an important role in the modulation of anxiety states.

Experimental models of anxiety states

As defined above, anxiety is a complex emotional state associated with sustained heightened autonomic and behavioral arousal and a sustained, trans-situational increase in avoidance behavior, often in environmental situations where a conflict between approach and avoidance exists. Although anxiety-related behavior is an innate behavioral response, a number of stimuli have been identified that result in sustained, trans-situational increases in avoidance behaviors. These include a number of pharmacologically distinct anxiogenic drugs, anxiogenic neuropeptides, aversive stimuli such as bright light, and inescapable shock. All of these stimuli have been shown to increase anxiety-related behavior in diverse environmental contexts and therefore, are thought to induce a generalized anxiety state as opposed to specific anxiety-related behaviors. It should be emphasized that although much of what we know about neural mechanisms underlying anxiety-related behavior is based on animal models involving conflict between approach and avoidance behaviors [e.g. (Millan 2003)], changes in anxiety states are likely to result in changes in many other physiological and behavioral responses. The distinct temporal profiles of different anxiety-related states and the potential role of serotonergic systems in the development of these anxiety states are outlined below.

Models of anxiety states with distinct temporal profiles

An important component of anxiety states is that the sustained increases in anxiety-related physiological and behavioral responses are not tied to specific environmental cues but are trans-situational, meaning that they can be elicited under diverse environmental contexts. Several models of anxiety states with very distinct temporal profiles have been developed, suggesting that the mechanisms underlying the development and maintenance of these anxiety states may be different. For example, bright light, which is aversive to rats and increases anxiety-related behaviors in a variety of experimental contexts (Valle 1970; Cunha and Masur 1978; Hall et al. 2000), also increases the magnitude of startle responses following exposure of rats to startle-eliciting noise bursts [reviewed in Walker et al. (2003)]. This effect is disrupted by both benzodiazepine and non-benzodiazepine anxiolytics (Walker and Davis 1997; de Jongh R. et al. 2002; Walker and Davis 2002), supporting the possibility that the effects of bright light on startle responses are due to development of an anxiety state. The effect of bright light takes several minutes to develop in order to evoke a maximal response and, following a 20 min exposure to bright light, is present for at least another 20 min (Walker et al. 2003). Thus, as pointed out by Walker and colleagues, light-enhanced startle requires a long-duration stimulus and the effect of the stimulus outlasts its duration. The behavioral effects of the stress and anxiety-related neuropeptide CRF on startle responses and other anxiety-related behaviors develop within 30 min and persist for several hours (Dunn and Berridge 1990; Walker et al. 2003), a similar time course compared to several anxiogenic drugs (Millan 2003). In contrast, exposure to inescapable stress appears to result in an anxiety-like emotional state that lasts for a period of approximately 24-48 h (Maier and Watkins 2005), while other experimental models of anxiety are associated with increased anxiety-related behavioral responses that are sustained for several weeks (Gewirtz et al. 1998; Sajdyk and Shekhar 2000; Sajdyk et al. 2004). The different time courses involved in these different animal models of anxiety states suggest that they are associated with different underlying neural mechanisms. There is evidence supporting the hypothesis that at least some of these anxiety-related states are dependent on activation of serotonergic systems in the dorsal raphe nucleus.

A role for serotonergic systems in the development of anxiety states

Currently, the most convincing evidence for a role for serotonergic systems in the development and maintenance of anxiety states is based on studies involving the administration of anxiogenic drugs (see discussion above) and exposure of rats to inescapable stress. Exposure of rats to inescapable stress results in a constellation of behavioral consequences 24–48 h later, including increases in anxiety-related behavior in the social interaction and circular elevated maze tests of anxiety [reviewed in Maier and Watkins (2005)]. Several lines of evidence suggest that the anxiety state that follows exposure to inescapable stress is dependent on activation of serotonergic neurons within the dorsal raphe nucleus, and these arguments have been outlined recently in an excellent review by Maier and Watkins (2005). Determining the role of serotonergic systems in the development and maintenance of chronic anxiety states with different temporal profiles is an important objective for future studies.

Summary and future directions

Although a clear understanding of the neural mechanisms underlying anxiety states and the essential stimulus parameters for the development of anxiety states have proven elusive, evidence suggests that elements of uncertainty, unpredictability, or uncontrollability play a central role (Walker et al. 2003; Herry et al. 2005; Maier and Watkins 2005). Interestingly, a neural circuit including the basolateral amygdala, mPFC, and dorsal raphe may play a central role in detecting unpredictability or uncontrollability as a stimulus parameter in rodents and humans (Herry et al. 2005; Amat et al. 2005; Maier and Watkins 2005). Evidence suggests that serotonergic neurons within the dorsal part of the mid-rostrocaudal and caudal dorsal raphe nucleus are selectively activated by a variety of anxiety-related stimuli. This region of the dorsal raphe nucleus, compared to other regions of the dorsal raphe nucleus, is neurochemically distinct and receives unique afferent input from the BNST, a structure that has been implicated in the regulation of anxeity-related behaviour. This region of the dorsal raphe nucleus in turn gives rise to collateral projections to multiple nodes within a distributed neural system modulating anxiety-related behaviors including the basolateral amygdala and mPFC. These findings suggest that different subsets of serotonergic neurons may differentially influence neural processes underlying the initiation, maintenance, or termination of anxiety-related behaviors and anxiety states. In future studies it will be important to determine which subpopulations of serotonergic neurons within the dorsal part of the mid-rostrocaudal and caudal dorsal raphe are affected by anxiogenic drugs or anxiogenic stimuli (e.g. identified based on their projections to other brain regions, or their unique molecular, morphological, or physiological properties). It will also be important to determine the specific roles that these serotonergic neurons play in the modulation of anxiety states, and the time course of these effects. Identification of the molecular, cellular, and functional properties of different subsets of serotonergic neurons may lead to a better understanding of the role of serotonergic systems in modulation of complex emotional states including anxiety.

Acknowledgements

We are grateful to Dr Miles Orchinik, Dr Katharine Semsar, Dr Paul Gasser, Andrew Evans, and Daniel Staub for critical comments on this manuscript. This work was supported by NIMH RO1 MH065702 and RO1 MH52619 to A.S.; C. A. Lowry is a Wellcome Trust Research Fellow (RCDF 068558/Z/02/Z).