Serotonergic innervation of the amygdala: targets, receptors, and implications for stress and anxiety
- First Online:
- Cite this article as:
- Asan, E., Steinke, M. & Lesch, KP. Histochem Cell Biol (2013) 139: 785. doi:10.1007/s00418-013-1081-1
- 1.6k Downloads
The amygdala is a core component of neural circuits that mediate processing of emotional, particularly anxiety and fear-related stimuli across species. In addition, the nuclear complex plays a key role in the central nervous system stress response, and alterations in amygdala responsivity are found in neuropsychiatric disorders, especially those precipitated or sustained by stressors. Serotonin has been shown to shape and fine-tune neural plasticity in development and adulthood, thereby allowing for network flexibility and adaptive capacity in response to environmental challenges, and is implicated in the modulation of stimulus processing and stress sensitivity in the amygdala. The fact that altered amygdala activity patterns are observed upon pharmacological manipulations of serotonergic transmission, as well as in carriers of genetic variations in serotonin pathway-associated signaling molecules representing risk factors for neuropsychiatric disorders, underlines the importance of understanding the role and mode of action of serotonergic transmission in the amygdala for human psychopathology. Here, we present a short overview over organizational principles of the amygdala in rodents, non-human primates and humans, and review findings on the origin, morphology, and targets of serotonergic innervation, the distribution patterns and cellular expression of serotonin receptors, and the consequences of stress and pharmacological manipulations of serotonergic transmission in the amygdala, focusing particularly on the extensively studied basolateral complex and central nucleus.
KeywordsBasolateral amygdala Central amygdala nucleus Serotonin receptors Emotion Rodent Primate
Bed nucleus of the stria terminalis
Calcium/calmodulin-dependent protein kinase II
Lateral central nucleus
Lateral capsular subdivision of the central nucleus
Medial central nucleus
Central nervous system
Corticotropin releasing factor
Dorsal raphe nuclei
Extracellular signal-regulated kinase(s)
Gamma-amino butyric acid
Glutamic acid decarboxylase
Paracapsular intercalated cell islands
In situ hybridization
Median raphe nuclei
Selective serotonin reuptake inhibitor
Tryptophan hydroxylase 2
Vasoactive intestinal polypeptide
The amygdala, first described in the human brain as an almond-shaped mass of gray substance in the anteromedial temporal lobe by Karl Friedrich Burdach in 1819, has long been recognized as a key component of brain emotion circuits across species (LeDoux 2007). Experimental studies mainly carried out in rats, and, more recently, in inbred or genetically modified mice displaying emotional dysregulation and serving as animal models for studies on the neurobiological mechanisms related to affective disorders in humans (Fernandez and Gaspar 2011; Holmes 2008; Narayanan et al. 2011; Sartori et al. 2011; Stinnett and Seasholtz 2010), have documented a seminal function of the amygdala for the specific processing of aversive emotional stimuli eliciting fear and anxiety-related behavior (LeDoux 2000, 2007, 2012; Lowry et al. 2005; Mahan and Ressler 2012; Sierra-Mercado et al. 2011; Wang et al. 2011). In particular, the role of malleable synaptic networks of the amygdala for the acquisition, expression, retention, and extinction of conditioned fear has been extensively analyzed (for review see Barad et al. 2006; Pape and Pare 2010). Connections between the amygdala and the ventral striatum have additionally been proposed to convey functions in reward learning and motivation (Everitt et al. 1999; LeDoux 2007; Man et al. 2012). Moreover, the amygdala appears to be involved in the modulation of attention, perception, and explicit memory, possibly via its general role in attaching emotional significance to external stimuli, particularly pain and other aversive but also appetitive stimuli (LeDoux 2007; Rea et al. 2009). Investigations in non-human primates corroborated findings from rodents (Belova et al. 2008; Centeno et al. 2007; Gothard et al. 2007; Ichise et al. 2006; Murray and Izquierdo 2007; Paton et al. 2006; reviewed in Hikosaka et al. 2008), and functional imaging techniques have delivered an increasing body of evidence showing that the amygdala plays a crucial role for emotional stimulus processing also in humans. Consequently, alterations in amygdala responses have been documented in psychiatric disorders, particularly those characterized by heightened anxiety (Boll et al. 2011; Drevets et al. 2008; Levens et al. 2011; Phelps and LeDoux 2005; Shin and Liberzon 2010; Simon et al. 2006; Toyoda et al. 2011; Vizueta et al. 2012). In fact, as suggested in a recent review of studies which contributed to elucidating factors modulating the responsiveness of the amygdala and its effect on physiological and behavioral reactions to emotional stimuli, genetically and experience-influenced deviations in amygdala responsivity may promote inadequate processing of both internal states perceived as threatening or environmental adversity, resulting in anxiety disorders and, ultimately, depression (Canli and Lesch 2007).
Stress, either acute or chronic, is one of the prime factors which has experimentally and clinically been shown to modulate amygdala reactivity. Given its nodal position in the fear/anxiety neurocircuitry, it is not surprising that the amygdala plays a critical role in the stress response of the central nervous system (CNS), presumably orchestrating the complex set of physiological (behavioral, neuroendocrine, and neurochemical) and cognitive responses (attentional processes, memory formation, consolidation, and retrieval) elicited by stressful experience (Berretta 2005; Shin and Liberzon 2010; Stinnett and Seasholtz 2010; van Marle et al. 2009). Neuronal and synaptic plasticity in the amygdala is observed after acute and chronic exposure to stressors (Mahan and Ressler 2012; McEwen et al. 2012; Mitra et al. 2005, 2009; Leuner and Shors, 2012; Nietzer et al. 2011; Roozendaal et al. 2009; Vyas et al. 2002, 2004, 2006). In the short term, these alterations may be essential for adaptive changes in behavior in a hostile or life-threatening environment (Compan 2007). However, if stress-related structural and functional plasticity particularly in the amygdala leads to permanent homeostatic imbalances in anxiety circuits even after cessation of threat, emotional dysregulation, i.e., exaggerated or inadequate reactions to intrinsic or extrinsic signals, may be the consequence. Stress-induced deviant amygdala reactivity thus might underlie the fact that many affective disorders are precipitated or aggravated and sustained by stress (Mahan and Ressler 2012; McEwen et al. 2012; Stinnett and Seasholtz 2010). In accordance with this suggestion, alterations in amygdala responsivity to emotional stimuli are particularly prevalent in stress-related anxiety syndromes in humans such as the posttraumatic stress disorder, phobias, and panic disorder (Damsa et al. 2009; Furmark 2009; LeDoux 2007; Mahan and Ressler 2012). The growing interest in amygdala function and dysfunction, especially with respect to neuropsychiatric disorders, has incited a large number of investigations aimed at elucidating how amygdala processing of emotional stimuli is organized, both in terms of connectivity, and on a cellular and molecular level. With increasing knowledge about interconnections of amygdaloid nuclei, and about the neurochemistry and signal transduction properties of intrinsic and extrinsic neurotransmitter and neuromodulator systems, it has become evident that it is essential to understand how the different systems interact in order to reveal the basis of amygdaloid processing under various conditions (LeDoux 2007; Mahan and Ressler 2012).
Among the neurotransmitter systems modulating information flow through amygdala circuits, particularly the influence of serotonin (5-Hydroxytryptamin, 5-HT) transmission has received attention. Pharmacobehavioral studies in experimental animals have documented that the serotonergic system is strongly implicated in modulation of anxiety-like behavior, conditioned fear and stress responses (Graeff 2002; Graeff et al. 1996; Jasinska et al. 2012; Lowry et al. 2005), and is also involved in the regulation of reward-related behavior (Hayes and Greenshaw 2011). In all species from rodents to primates, the amygdala receives dense serotonergic innervation, and experimental manipulations of serotonergic transmission in the amygdala cause altered anxiety-like behavior in experimental animals (Bauman and Amaral 2005; Christianson et al. 2010; Fallon and Ciofi 1992; Freedman and Shi 2001; Hensler 2006; Lowry et al. 2005; Menard and Treit 1999; Smith and Porrino 2008). Stress is associated with differential serotonergic transmission in the amygdala (Amat et al. 1998; Ichise et al. 2006; Jasinska et al. 2012; Kawahara et al. 1993; Mitsushima et al. 2006; Rueter and Jacobs 1996; Zanoveli et al. 2009), and stress-induced anxiety-like behavior in experimental animals can be modulated by targeted pharmacological manipulation of amygdaloid serotonergic signal transduction (Christianson et al. 2010; Lowry et al. 2005; Menard and Treit 1999). Evidence implicating serotonergic transmission in the amygdala in emotion regulation also in humans comes from functional imaging studies documenting changes in amygdala activity after pharmacological manipulation of serotonergic subsystems in healthy volunteers (Cools et al. 2005; Evers et al. 2010; Norbury et al. 2009). Elevated serotonin turnover in the brain (Barton et al. 2008; Esler et al. 2007) and alterations of serotonin signalling molecules in the amygdala (Murrough et al. 2011a, b; Stockmeier 2003) in patients with affective disorders implicate serotonergic influence on amygdaloid circuits also in human psychopathology.
Moreover, research of the last 15 years has documented that human carriers of genetic variations in serotonin system-associated transmission molecules display altered amygdala activation patterns upon presentation of emotional stimuli (Canli and Lesch 2007; Hariri et al. 2002; Holmes 2008; Kilpatrick et al. 2011; Lee and Ham 2008). Convincing evidence has also been provided that such variations exert a major influence on individual differences in the stress response and on the risk for stress-related neuropsychiatric disorders (Holmes 2008; Jasinska et al. 2012; Mahan and Ressler 2012). Experimental studies in animal models with genetically modified expression of components of the serotonin signaling pathway subsequently documented that variations in genes modulating serotonin transmission have a significant impact on structural development, neurochemical properties, and functions of the amygdala (Fernandez and Gaspar 2011; Holmes 2008; Homberg 2012; Li et al. 2003; Mahan and Ressler 2012; Nietzer et al. 2011; Wellman et al. 2007). In view of the obvious importance of serotonergic transmission for amygdala functions, it is essential to unravel the cellular and molecular mechanisms of serotonin effects on amygdala circuits. In particular, precise information concerning the cellular targets of the serotonergic innervation and analysis of the influence of serotonergic transmission on identified cell types in the amygdala is of prime importance. Therefore, after a short introduction into organization and connectivity of the amygdala in rodents and primates, and into origins and morphology of the serotonergic amygdala innervation, the present review focuses on compiling data concerning the interrelationships of serotonergic afferents with identified amygdala neurons, the regional and cellular expression of serotonin receptors (5-HTR), and the serotonergic transmission in the amygdala in stress and anxiety. Since most studies have been carried out in rats and mice, discussion of findings in these species will form the basis of the review, with data from primate and human studies mentioned where appropriate. Reviewed literature for the different aspects discussed below, as for the introduction above, will include original articles and, additionally, numerous expert reviews of different aspects which should be consulted for further reading.
Organization of the amygdala
Subdivisions (Fig. 1a, b)
In addition to superficial and deep nuclei, another major nucleus, the central nucleus (Ce), is found at mid to caudal levels in the dorsal amygdala adjacent to the junction of the stria terminalis with the complex. According to Alheid (2003), the Ce is part of the “central extended amygdala”, a basal forebrain macrostructure consisting additionally of the lateral bed nucleus of the stria terminalis (BNST), which shares cytoarchitectural and histochemical features with the Ce, and of a continuous chain of cell islands linking the Ce with the BNST. In rodents, primates, and humans, the Ce can be subdivided into two major divisions, namely the medial (CeM) and lateral (CeL) Ce. Additional subdivisions, in particular a lateral capsular Ce (CeLc), have been described in rodent (Fig. 1a) and monkey amygdala with somewhat varying denominations (Entis et al. 2012; Freedman and Shi 2001; Fudge and Tucker 2009; McDonald 1982a; Pitkänen et al. 1998).
Besides the large amygdaloid nuclei, small dense groups of gamma-amino butyric acid (GABA)-containing cells, the intercalated cells (IC), are found in rat and primate amygdala. IC islands are mostly embedded in fiber tracts encapsulating and separating the main nuclei, and/or are situated between deep and superficial nuclei. The paralaminar nucleus (PL), a narrow sheet of small, densely packed cells covering the ventral surface of the La and BL, is recognizable as a nuclear entity only in the primate amygdala (Fig. 1b) (Amaral et al. 1992; deCampo and Fudge 2012). Small cell groups between the La and BL and the external capsule in the rat amygdala, which have been included in the group of paracapsular intercalated cell islands (ICp; Fig. 1a) by some authors (Millhouse 1986; Palomares-Castillo et al. 2012), have been suggested to constitute the PL equivalent in rodents, although while most ICp cells are GABAergic, this is true only for a minority of PL cells (deCampo and Fudge 2012). In addition, not readily categorizable parts of the amygdala are the anterior amygdaloid area and the amygdalostriatal area interposed between the major amygdalar nuclei and their neighboring structures at rostral and dorsal poles of the complex.
Based on cytoarchitectonic and neurochemical characteristics and developmental gene expression patterns, it has been suggested that the deep nuclei are of pallial origin and represent a nuclear extension of the frontotemporal cortex, while the Ce and Me are subpallial and represent ventral extensions of the striatum (Medina et al. 2004; Puelles et al. 2000; Swanson and Petrovich 1998). Phylogenetically, although all amygdaloid nuclei found in rodents are present also in monkey and human, the deep nuclei are particularly progressive and are significantly increased in size compared to the Ce and Me in primates (Fig. 1a, b) (O’Rourke and Fudge 2006; Pitkänen and Kemppainen 2002). In the present review, we shall preferentially use the nomenclature according to Alheid et al. (1995), since it is commonly applied in studies of rodent amygdala. Our focus will be primarily on the La, BL and Ce, whose roles in emotional stimulus processing and in stress response have been investigated extensively (e.g., Bhatnagar et al. 2004; Cardinal et al. 2002; Charney 2003; Davis 1992; Goldstein et al. 1996; LeDoux 2003; Pape and Pare 2010; Shekhar et al. 2003; Shors and Mathew 1998; Wills et al. 2010).
Tracing studies in rodents and non-human primates have indicated that the amygdala possesses similar hodological features in these species, with cortical connections more elaborate in primate than in rodent (Price 2003). In a simplified scheme of the information flow through the amygdala mediating fear reactions (Ehrlich et al. 2009; LeDoux 2007) (Fig. 1c), the La serves as a “gatekeeper” to the amygdala receiving sensory cortical and thalamic input (McDonald 1998; Turner and Herkenham 1991). The La projects to the BL, which receives additional input from the hippocampal formation and polymodal association areas (Pitkänen et al. 1997; Sah et al. 2003). BL and La project back to cortical areas, and to the striatum, and possess numerous subdivisions, which are interconnected by a multitude of inter- and intranuclear projections (Pape and Pare 2010; Pitkänen et al. 1997). La and BL send projections to the Ce, which receives some cortical input (Fudge and Tucker 2009; McDonald et al. 1999) in addition to visceral, humoral, and nociceptive information via a broad range of inputs from the hypothalamus, brainstem, midline thalamus, and spinal cord (Cliffer et al. 1991; Menetrey and De Pommery 1991; Ottersen 1980, 1981; Ottersen and Ben-Ari 1979; Rinaman and Schwartz 2004; Saper and Loewy 1980; Veening 1978). The Ce serves as the principal output station of the amygdala to brainstem and hypothalamic regions. Information relayed to the Ce from other nuclei is extensively processed in Ce intrinsic circuits (Ciocchi et al. 2010; Duvarci et al. 2011; Ehrlich et al. 2009; Haubensak et al. 2010). In addition, Ce activity is modulated by inhibitory input from the IC, which are targeted by excitatory input from the BL, La, and medial prefrontal cortex areas, particularly the infralimbic cortex (Pape and Pare 2010). Ce amygdalo-fugal projections arise mainly from the CeM and orchestrate neuroendocrine and autonomic consequences, and ultimately motor action effects of amygdala processing (Krettek and Price 1978; LeDoux et al. 1988; Petrovich and Swanson 1997; Pitkänen et al. 1997; Ulrich-Lai and Herman 2009; Veening et al. 1984).
Cyto- and Chemoarchitecture
BL and La in rodents contain a majority of glutamatergic pyramidal cells (~85 %) (McDonald 1982b) which project to targets outside the amygdala and to other amygdaloid nuclei, and a minority of GABAergic interneurons (~15 %) forming local inhibitory circuits (Fig. 1c–e) (Ehrlich et al. 2009; McDonald 1982b; Millhouse and DeOlmos 1983; Pape and Pare 2010; Swanson and Petrovich 1998). Apical and basal dendrites of typical pyramidal neurons possess numerous spines which are targeted by extrinsic or internuclear excitatory afferents and are key players in synaptic plasticity (Hering and Sheng 2001; Koch and Zador 1993; Leuner and Shors 2012; Segal 2005). Based on correlative investigations of anxiety-like behavior and neuronal morphology in stressed or genetically modified rodents, it has been proposed that increased spine number and density on pyramidal neurons of the La and BL are the structural manifestation of stress- and anxiety-induced strengthening of synaptic connectivity in this amygdaloid region, representing a morphological correlation of an increased excitatory drive and/or hyperexcitability of output neurons, and presumably resulting in specific anxiety-related behavior (Leuner and Shors 2012; Mozhui et al. 2010; Nietzer et al. 2011; Roozendaal et al. 2009; Vyas et al. 2006). GABAergic interneurons of the BLC (Fig. 1e) are a morphologically, neurochemically, and electrophysiologically heterogeneous population of local circuit neurons which are part of both feed-back and feed-forward inhibitory circuits and control excitability of La and BL output neurons (Fig. 1c) (Ehrlich et al. 2009; Muller et al. 2003; Pape and Pare 2010; Rainnie et al. 2006). In the rat, these interneurons possess sparsely spiny to aspiny dendrites (Millhouse and DeOlmos 1983), and express, in separate and sometimes overlapping subpopulations, the calcium-binding proteins parvalbumin (PV), calbindin (CB), and calretinin (CR), and/or numerous neuropeptides such as Met- and Leu-enkephalin (Met-, Leu-ENK), somatostatin (SOM), cholecystokinin (CCK), vasoactive intestinal polypeptide (VIP), and neuropeptide Y (NPY), with NPY-producing interneurons representing a subgroup of SOM-immunoreactive (ir) neurons (Fig. 1f, g) (Asan 1998; McDonald 1989; McDonald and Pearson 1989; Muller et al. 2007a; Pitkänen and Kemppainen 2002; Rainnie et al. 2006). Interestingly, while PV-ir interneurons target both perisomatic and distal compartments of pyramidal cells, the subgroup of SOM-ir interneurons contacts preferentially distal dendrites (Muller et al. 2006, 2007a). Both subgroups additionally innervate other interneurons, and PV-ir neurons apparently form inhibitory networks coupled by gap junctions (Muller et al. 2005).
While the majority of CeL neurons are typical medium spiny neurons resembling striatal principal neurons, most CeM neurons are larger with lower spine density (McDonald 1982a). Virtually all Ce neurons are GABAergic, and express CB, CR, and numerous neuropeptides such as corticotropin releasing factor (CRF), SOM, NPY, CCK, Met- and Leu-ENK, Substance P, VIP and many others in partly overlapping subpopulations (Fig. 1c, e–g; Akmaev et al. 2004; Asan 1997, 1998; Cassell et al. 1986; McDonald 1997, 1982a; McDonald and Betette 2001; McDonald and Mascagni 2001; Pitkänen and Kemppainen 2002; Yilmazer-Hanke et al. 2002). In mouse amygdala, similar subgroups of GABAergic and peptide/calcium-binding protein producing neurons exist (Avila et al. 2011; Real et al. 2009; Rose et al. 2006). Recent findings indicate that CB, PV, and GABA dynamically colocalize in cells of the mouse deep nuclei throughout development (Davila et al. 2008). However, differences in the distribution of CRF-ir neurons and fibers apparently exist between rats and mice in the Ce (Asan et al. 2005). The localization of SOM- and NPY-ir neurons and fibers is very similar to rat in monkey amygdala (McDonald et al. 1995), but differences in (sub)nuclear density again exist for CRF-ir elements (Bassett and Foote 1992). Distribution of CB, CR, and PV profiles in monkey and rat amygdala is, in principle, again similar to findings in rats, with some differences in the densities of neurons and fibers in nuclear subdivisions (Mascagni et al. 2009; Pitkänen and Kemppainen 2002; Sorvari et al. 1995; 1996a, b).
Serotonergic forebrain innervation: general remarks
Localization of serotonergic neurons and distribution of serotonergic fibers in rodents, primates, and humans have been documented using immunocytochemistry for serotonin or for the serotonin transporter (5-HTT), which is specifically expressed at high levels in the majority of serotonergic afferents to the forebrain (Brown and Molliver 2000; Hale and Lowry 2010; Hornung 2003; Smith and Porrino 2008; Steinbusch and Nieuwenhuys 1981; Wilson and Molliver 1991; Yamamoto et al. 1998). More recently, detection of the rate-limiting enzyme of serotonin synthesis in brain, tryptophan hydroxylase 2 (Tph2) (Walther and Bader 2003), has been used to specifically identify serotonergic brainstem neurons (e.g., Abrams et al. 2005; Commons et al. 2003; Gardner et al. 2009; Gutknecht et al. 2009).
Localization, morphology and electrophysiological properties of serotonergic forebrain projection neurons
Numerous investigations have documented that the DR consists of multiple topographically organized, hodologically and functionally heterogeneous subpopulations of serotonergic neurons with unique behavioral correlates (Amat et al. 2004; Lowry et al. 2005). Electrophysiological investigations in cats additionally identified typical and atypical serotonergic neurons within these subpopulations based on differences in firing patterns during wake-sleep states (Sakai and Crochet 2001). Typical serotonergic neurons represent the majority of DR serotonergic neurons and fire at a regular frequency of ~3 spikes/s during active waking (Jacobs and Fornal 1999; Sakai and Crochet 2001). In anesthetized rats, spiking frequencies of 10–14/10 s were noted in serotonergic DR neurons, with male rats displaying significantly higher frequencies than freely cycling or ovariectomized female rats (Klink et al. 2002). Jasinska et al. (2012) suggested that the regular activity pattern results in a steady serotonergic transmission in target structures including the amygdala during active wakefulness, and proposed that the steady synaptic concentration of serotonin and a sustained activation of the postsynaptic 5-HTR in the target regions enables active, goal-directed motor and cognitive functions.
Modes of serotonergic transmission
In contrast to glutamate, which is rapidly cleared from the synaptic cleft by binding to glutamate transporters situated at high density in the membranes of astroglial processes enveloping excitatory synapses (Lehre and Danbolt 1998), serotonin released from junctional (synaptic) or non-junctional sites has been shown to escape the synaptic cleft and/or to diffuse in the (extrasynaptic) extracellular space for considerable distances (Bunin and Wightman 1998) until removed by 5-HTT, which is localized in the extrasynaptic membranes of serotonergic terminals and axons (Zhou et al. 1998). Moreover, extrasynaptic neuronal and glial localization has been documented for instance for serotonin receptors, such as 5-HT1A (Azmitia et al. 1996; Kia et al. 1996a, b), which exhibits an affinity for its ligand in the low nanomolar range corresponding to concentrations measured for extrasynaptic serotonin after release (Bunin and Wightman 1998). Thus, in addition to its function as a synaptic transmitter, serotonin may mediate its effect in a paracrine mode, also termed “volume” transmission as opposed to “wiring” (synaptic) transmission in the concept introduced by Agnati et al. (1986; for further review see 1995, 2010). According to Fuxe et al. (2007), volume and wiring transmission of serotonin may represent complementary modes of communication with wiring transmission rather subserving short-term actions and volume transmission involved in long-lasting modulation of cellular networks. Furthermore, volume transmission presumably enables communication between all brain cell types including neuron-glia interactions and transactivation processes between 5-HTR and other, for instance neurotrophic factor, receptors presumably mediating neurotrophic and further effects of serotonin (Fuxe et al. 2007).
Serotonergic amygdala innervation
Tracing studies have shown that amygdaloid serotonergic afferents in the rat arise mainly from the DR (Fig. 2a) (Abrams et al. 2005; Halberstadt and Balaban 2006; Ma et al. 1991), while only scarce afferents from the MR have been found (Vertes et al. 1999). These findings are corroborated by the fact that type M axons are practically absent from the La and relatively sparse in the BL and Ce (Fig. 2b) (Bonn et al. 2013; Vertes et al. 1999). Similar axonal morphologies and distribution patterns were described in the squirrel and macaque monkey amygdala (Bauman and Amaral 2005; Sadikot and Parent 1990). Within the rat DR, serotonergic neurons projecting to the amygdala are specifically localized to the dorsal part of midrostrocaudal and caudal portions (Abrams et al. 2005; Imai et al. 1986; Jacobs et al. 1978; Lowry et al. 2008). Apparently, separate populations of serotonergic neurons within these DR subregions target specific amygdala nuclei. Thus, anterograde tracing from the dorsomedial DR, which contains a cluster of CRF-producing serotonergic neurons, yielded labeling of afferents within the Ce (Commons et al. 2003), while retrograde tracing identified serotonergic neurons projecting to the BL throughout the dorsal midrostrocaudal DR, particularly within a dorsal “shell” subregion of the DR overlapping the anatomical border between dorsal, ventral, and ventrolateral DR (Abrams et al. 2005). Independent regulation of the activity of DR neuron subpopulations projecting to the Ce and BL was suggested by the finding that aversive stimulation of the inferior colliculus elicits increases in extracellular serotonin concentration in the BL but not in the Ce (Macedo et al. 2005). Interestingly, recent investigations on mice deficient for Pet-1, a transcription factor required for the specification of serotonergic identity in a majority of neurons in the raphe nuclei, showed that serotonergic axons innervating the BL and CeM apparently originate in a subset of DR neurons distributed throughout the midbrain raphe complex which is independent of Pet-1 for acquisition of serotonergic identity (Kiyasova et al. 2011). At least some of the amygdala-projecting DR neurons co-innervate additional brain regions. Thus, Ce-projecting serotonergic DR neurons send collaterals to the hypothalamic paraventricular nucleus, providing coordinated modulation of specific neural circuits such as those implicated in central autonomic and endocrine control, anxiety, and conditioned fear (Abrams et al. 2004; Petrov et al. 1994; Waselus et al. 2011).
Immunohistochemical and retrograde tracing studies in rats and monkeys documented that serotonergic fibers reach the amygdala via the ventral amygdalofugal fiber bundle and stria terminalis (Fallon and Ciofi 1992; Sadikot and Parent 1990). In rats, labeling of serotonergic afferents using serotonin and 5-HTT antibodies leads to largely identical immunolabeling patterns. Only few type M fibers not detected using 5-HTT IHC are found in serotonin immunoreactions in the CeM and BL (Fig. 2b). Serotonergic innervation of rat amygdaloid nuclei is heterogeneous: whereas La and BL display a moderately dense to very dense meshwork of serotonergic fibers, respectively, with particularly intense innervation in the posterior part of the BL, the CeL and CeLc receive only scarce innervation, and the CeM contains a moderate density of serotonergic fibers (Fig. 2c, e) (Asan et al. 2005; Bonn et al. 2013; Fallon and Ciofi 1992; Smith and Porrino 2008; Sur et al. 1996). The main ICs and the BM display moderate to dense fiber plexus, as does the Me. The serotonergic innervation of mouse amygdala appears similar to that in rats, with slight differences in fiber densities in some subnuclei (Asan et al. 2005). Thus, the density of serotonergic afferents appears higher in the CeLc in mice than in rats (Fig. 2d) (Asan et al. 2005).
In contrast to the findings in rodents, a very dense serotonergic innervation is found in the Ce, particularly in the CeL, in the non-human primate amygdala, documented using serotonin/5-HTT immunoreactions and 5-HTT binding studies (Fig. 2f) (Bauman and Amaral 2005; Freedman and Shi 2001; O’Rourke and Fudge 2006; Sadikot and Parent 1990; Smith and Porrino 2008). Subtle differences were described between serotonin and 5-HTT immunoreactions in the Ce subnuclei (Freedman and Shi 2001). Moderate to dense and low to moderate densities of serotonin- and 5-HTT-ir fibers and of 5-HTT binding were observed in the primate BL and La, respectively (Bauman and Amaral 2005; O’Rourke and Fudge 2006; Sadikot and Parent 1990; Smith and Porrino 2008), and a particularly high density of serotonergic fibers was found in the PL (Bauman and Amaral 2005; O’Rourke and Fudge 2006). Serotonergic fiber plexus in the ICs were found to be moderately and very dense using serotonin and 5-HTT antibodies, respectively (Bauman and Amaral 2005; O’Rourke and Fudge 2006). Systematic immunohistochemical investigations of the nuclear and subnuclear serotonergic innervation in the human amygdala are lacking. Early autoradiographic studies of 5-HTT binding showed highest concentrations of [3H]-cyanoimipramine and [3H]-paroxetine binding sites in cortical and anterior amygdaloid nuclei, moderate concentrations in the BL and Ce, and lower concentrations in the La (Gurevich and Joyce 1996). In a recent study comparing [3H]-citalopram binding sites in amygdalae from postmortem brains from control individuals and from patients with alcohol dependence, the concentration of binding sites in the ventral amygdala including the BLC amounted to only about half of the concentration detected in the dorsal amygdala including the Ce (Storvik et al. 2007). The obvious species differences noted for the serotonergic innervation of the amygdala, particularly of the Ce, are not found to the same extent for the dopaminergic innervation, which is more similar in rats and monkeys with densest innervation in the CeL and ICs, very dense innervation of the BL and rather scarce fibers in the La (Asan 1997, 1998; Asan et al. 2005; Eliava et al. 2003; Smith and Porrino 2008). On the other hand, similar innervation pattern differences as described for the serotonergic innervation appear to exist also between rat and primate/human noradrenergic innervation (Smith and Porrino 2008). Thus, interactions between these monoamine systems in the different amygdaloid nuclei presumably also vary between species.
Transmission electron microscopic studies of serotonergic afferents and terminals in the amygdala have, to date, been carried out almost exclusively in the rat La and BL. In a detailed study of the anterior BL, Muller et al. (2007b) documented that serotonin-ir terminals were round to ovoid, often linked by a slender axonal segment, and contained densely packed, small round vesicles. In addition, dense core vesicles were observed in most terminals. Quantitative analyses documented that 76 % of serotonergic terminals reconstructed from serial sections formed small synaptic junctions to target structures, 20 % of these were observed to make more than one synapse. The vast majority of synaptic contacts displayed characteristics of symmetric synapses. In a recent qualitative study in the rat La and BL, serotonergic terminals were found to form predominantly small, symmetric junctions, supporting the previous findings (Bonn et al. 2013). In a comparative study of 5-HTT-ir terminals in wild type and Pet-1-deficient mice (Kiyasova et al. 2011), reconstructions of serial ultrathin sections of the BL revealed structural features of labeled terminals very similar to those described by Muller et al. (2007b) for the rat in both genotypes. In the wild type, asymmetric synaptic contacts were found in 38 % of serotonergic terminals while 95 % of 5-HTT-ir terminals in Pet-1-deficient mice formed asymmetric synapses, indicating that high incidence of (asymmetric) synaptic contacts might be a characteristic of the subset of Pet-1 independent serotonergic raphe neurons projecting to the BL and Ce (Kiyasova et al. 2011).
Morphologically identified interrelations of serotonergic afferents with specific target neurons
The identity of serotonergic targets and, in particular, the ultrastructural morphology of interrelations have been more extensively studied in the rat BLC. A comprehensive light microscopic analysis of serotonergic innervation of neuronal subpopulations was presented by Muller et al. (2007b). Contacts were found between serotonin-ir and calcium/calmodulin-dependent protein kinase II (CaMK)-ir pyramidal neurons, and between serotonin-ir and PV- (Fig. 3c), CB-, CR-, SOM- and CCK-ir interneurons. Ultrastructural analyses in the anterior BL showed that the great majority of serotonergic terminals contacted dendritic spines and small-caliber dendritic shafts, indicating that serotonergic input to this nucleus primarily targets the distal compartment of pyramidal cells. Documentation of immunoreactivity for CaMK in many of these target structures supported this observation. In the mouse BL, 5-HTT-ir terminals also formed synapses preferentially on spines and small dendrites, supporting pyramidal cells as postsynaptic targets also in this species (Kiyasova et al. 2011). Serotonergic afferents in the rat anterior BL additionally formed appositions and, occasionally, synapses on perikarya and dendrites of PV- and VIP-ir interneurons (Muller et al. 2007b). In a recent study, membrane appositions and synapses on NPY-ir interneurons were documented in the rat La and BL (Fig. 3d–f) (Bonn et al. 2013). In addition to contacts with neuronal perikarya, dendrites and spines, appositions of serotonergic axons with other axons were frequently observed, and unlabeled and adjacent serotonergic terminals were often found to contact the same postsynaptic structure (Muller et al. 2007b).
Serotonin receptors: general remarks
To date, 14 5-HTR grouped into seven families (5-HT1A-F, 5-HT2A-C, 5-HT3, 5-HT4, 5-HT5, 5-HT6, and 5-HT7 receptors) have been identified. In addition, pre-mRNA splicing and editing variants were documented for 5-HT2C, 5-HT3, 5-HT4, and 5-HT7. Characteristics of the receptors have been reviewed previously (e.g., Barnes and Sharp 1999; Bjork et al. 2010; Charnay and Leger 2010; Hannon and Hoyer 2008; Puig and Gulledge 2011) and will only briefly be recalled here in some detail for those receptor subtypes studied most extensively (5-HT1A, B; 5-HT2; 5-HT3). With the exception of 5-HT3 receptors, which are ligand-gated ion channels, 5-HTR are metabotropic, G protein-coupled receptors.
The 5-HT1 receptor class is composed of five receptors (5-HT1A, 5-HT1B, 5-HT1D, 5-HT1E, and 5-HT1F) which, in humans, share 40–63 % overall sequence identity and couple preferentially to Gi/o to inhibit cAMP formation. 5-HT1A activation additionally leads to hyperpolarization of the cell membrane and thus to a decreased firing rate due to opening of G protein-coupled inwardly rectifying potassium channels (Luscher and Slesinger 2010; Polter and Li 2010). Additional evidence links the 5-HT1A receptor to extracellular signal-regulated kinases (ERK) and Akt signaling pathways (Lesch and Waider 2012; Polter and Li 2010). 5-HT1A is found as an inhibitory autoreceptor both somatodendritically on DR neurons and as an inhibitory heteroreceptor in pyramidal and interneurons in cortical areas (Andrade 2011; Azmitia et al. 1996; Aznar et al. 2003; Bonn et al. 2012; Descarries et al. 2006). Ultrastructural studies indicate a localization of the receptor protein in synaptic and extrasynaptic membranes (Riad et al. 2000). Presynaptic localization of 5-HT1A on GABAergic terminals (Fink and Gothert 2007) has been proposed based on electrophysiological studies. For 5-HT1B, electrophysiological evidence has indicated that it is an inhibitory autoreceptor on serotonergic terminals and may also serve as an inhibitory heteroreceptor on other, for instance glutamatergic or cholinergic terminals, regulating transmitter release (for reviews see Barnes and Sharp 1999; Descarries et al. 2006; Fink and Gothert 2007; Riad et al. 2000).
The class of 5-HT2 receptors comprises three members, 5-HT2A-C, which exhibit 46–50 % overall sequence identity and couple preferentially to Gq/11 activating phospholipase C to increase inositol phosphates and cytosolic calcium (Hannon and Hoyer 2008). In addition, 5-HT2A/C-activation has been shown to activate ERK via the β-arrestin/Src/dynamin pathway (Yuen et al. 2008). Electrophysiologically, 5-HT2 receptors have been found to mediate neuronal excitation in most brain areas (Andrade 2011; Barnes and Sharp 1999). 5-HT2 receptors may also couple to G12/13 which are known to mediate long-term structural changes in cells (Hannon and Hoyer 2008). 5-HT2C activation has additionally been shown to facilitate cell depolarization by inhibition of specific potassium channels promoting potassium outflow (Weber et al. 2008). 5-HT2A receptor protein has been immunohistochemically detected in cortical and hippocampal pyramidal and GABAergic interneurons, with somatodendritic staining associated with the cytoplasm and both synaptic and extrasynaptic membranes of preferentially small dendritic branches (Andrade 2011; Becamel et al. 2004; Cornea-Hebert et al. 1999; Descarries et al. 2006; Hafizi et al. 2011; Miner et al. 2003; Rodriguez et al. 1999). Occasionally, immunolabeling for the receptor in axons was observed (Cornea-Hebert et al. 1999; Miner et al. 2003; Peddie et al. 2008; Rodriguez et al. 1999). 5-HT2A trafficking, targeting and signaling is regulated by scaffolding proteins and kinases, processes which have been recently reviewed by Allen et al. (2008). 5-HT2B has been detected immunohistochemically in a somatodendritic localization in the cerebellum, septum, hypothalamus, and medial amygdala in rodent brain (Duxon et al. 1997). The 5-HT2C receptor was one of the first 5-HTR cloned, with extremely high expression in the choroid plexus, and very high levels in limbic areas (Hannon and Hoyer 2008). Electrophysiologically, the receptor mediates neuronal excitation in many brain areas (Barnes and Sharp 1999). In both rodent and human, transcripts encoding the 5-HT2C can undergo mRNA editing, leading to various isoforms of the protein with region- and possibly even cell-specific expression in the rat brain (Burns et al. 1997; Drago and Serretti 2009; Englander et al. 2005; Hackler et al. 2006; Marion et al. 2004; Moya et al. 2011; Werry et al. 2008). In vitro experiments indicate a negative correlation between editing and constitutive activity/ligand binding affinity for the different isoforms, with some edited isoforms possessing 10–15 fold lower efficiency of G protein-coupling than the non-edited form (Burns et al. 1997; Marion et al. 2004). Less edited isoforms, on the other hand, are more readily internalized than more highly edited ones, which consequently show less desensitization (Marion et al. 2004). Perturbations of serotonin levels appear to elicit changes in RNA editing, with lower levels decreasing and higher levels increasing edited isoforms, but results of different studies are not entirely consistent (Gurevich et al. 2002; Werry et al. 2008). Also, in mouse strains displaying differential emotionality, consistent differences in brain region-specific 5-HT2C mRNA editing were documented (Englander et al. 2005; Hackler et al. 2007). Immunohistochemical labeling for 5-HT2C was found in numerous cortical and subcortical brain regions (Clemett et al. 2000; Descarries et al. 2006), and double labeling indicated that 5-HT2C-ir neurons in cortical areas were both GABAergic interneurons, particularly the subpopulations producing PV, and presumably pyramidal cells (Liu et al. 2007).
The 5-HT3 receptor is a ligand-gated ion channel, which triggers rapid depolarization due to the opening of non-selective cation channels (Na+ and Ca2+ influx, K+ efflux). The receptor has been localized to the perikaryal cytoplasm, dendritic postsynaptic membranes and subsynaptic cytoplasm of telencephalic inhibitory interneurons and, possibly, pyramidal cells (Miquel et al. 2002; Morales et al. 1996). Additionally, electrophysiological and ultrastructural studies point to a significant presynaptic localization of these receptor subtypes on cholinergic, dopaminergic, noradrenergic, and GABAergic terminals (Fink and Gothert 2007; Koyama et al. 2000; Miquel et al. 2002).
5-HT4, 5-HT6, and 5-HT7 receptors all couple preferentially to Gs and promote cAMP formation by activation of various adenylyl cyclases. Numerous splice variants have been documented for 5-HT4 and 5-HT7 receptors. Both receptors display widespread localization in peripheral tissues and a heterogeneous distribution in the CNS (Hannon and Hoyer 2008). 5-HT6 receptor appears to be restricted to various regions of the CNS (Hannon and Hoyer 2008), with recent studies in human postmortem brain indicating a localization to PV neurons in the striatum (Marazziti et al. 2012). The 5-HT5 receptor is the least well characterized of the receptors. Two subtypes have been identified in rat and mouse, which presumably couple to Gi/o proteins and inhibit cAMP formation. In humans, only the 5-HT5A variant appears to be expressed (Thomas 2006).
Colocalization of different subtypes of 5-HTR and of 5-HTR with other receptors
In recent years, numerous studies have documented coexpression, colocalization, and functional coupling between different 5-HTR subtypes and between individual 5-HTR subtypes and other receptors. Thus, electrophysiological studies indicated colocalization of 5-HT1A with 5-HT2A or 5-HT4 on cortical pyramidal neurons (Andrade 2011; Roychowdhury et al. 1994), and of 5-HT1A with 5-HT2C and 5-HT4 on subthalamic nucleus neurons (Stanford et al. 2005), mediating inhibitory and excitatory effects, respectively. Colocalization of 5-HT2A with glutamate receptor subunits [N-methyl-d-aspartate (NMDA) and amino-3-hydroxy-5-methyl-4-isoxazoleproprionic acid (AMPA) receptor subunits NR1 and mGluR2, respectively] was observed in dendrites and spines, respectively, of neurons in various brain areas such as the striatum and dentate gyrus (Peddie et al. 2008; Rodriguez et al. 1999; Yuen et al. 2008). Yuen et al. (2008) documented that 5-HT1A and 5-HT2A/C regulated NMDA receptor currents in individual prefrontal cortical pyramidal cells in a counteractive manner. They showed that activation of 5-HT2A/C receptors, by activating ERK via the β-arrestin-dependent pathway, significantly attenuated the microtubule-disrupting effect of 5-HT1A-mediated signal transduction on NMDA receptor trafficking to the membrane. Interestingly, in animals exposed to acute stress, the enhancing effect of 5-HT2A/C on cortical neuron firing was lost, while the decreasing effect of 5-HT1A on firing was intact, indicating that the complex effects of 5-HTR on neuronal excitability are selectively altered under stressful conditions (Zhong et al. 2008). Gonzales-Maeso et al. (2008) found colocalization of 5-HT2A and mGluR2 receptor immunoreactivity in mouse somatosensory cortical neurons, and established that 5-HT2A receptors form functional complexes with mGluR2 receptors which triggered unique cellular responses when targeted by hallucinogenic drugs. Indeed, recent evidence indicates that changes in Gi/Gq activity mediated by the mGluR2/5-HT2A heterocomplex might predict psychoactive behavioral effects of a variety of pharmacological compounds (Fribourg et al. 2011). Coexpression and functional interactions of 5-HT2 and CRF1 receptors in mouse cortical neurons were proposed by Magalhaes et al. (2010), and crosstalk and heteromerization of 5-HT2A and dopamine D2 receptors in HEK 293 cells by Albizu et al. (2011). Further coexpression of different subtypes of 5-HTR and of 5-HTR with other receptors were documented in the amygdala (see below). The list of receptor interactions is continuously growing, and the reciprocal influence on signal transduction between coupled 5-HTR or between 5-HTR and other receptors adds another dimension of complexity to the problem of unraveling serotonin functions.
Implications of receptor/innervation mismatch and of subcellular localization of receptor proteins
As mentioned above, serotonin may act via paracrine or volume transmission in target areas due to the fact that transporters mediating reuptake are localized preferentially in non-terminal segments of serotonergic axons, enabling an extrasynaptic diffusion of the transmitter. In this respect, it is interesting to note that virtually all available ultrastructural studies have documented localization of the different receptors on non-synaptic or extrasynaptic membranes. Additionally, in some brain areas, a mismatch between the local density of 5-HTR and of serotonergic terminal axons, and a relative paucity of synaptic contacts of serotonergic terminals has been noted, leading to the suggestion that in these areas serotonergic transmission may be mainly mediated by volume transmission (Descarries et al. 2006).
Serotonin receptors in the amygdala
Using ISH and IHC, it was previously shown that 5-HT1A mRNA and protein are produced at relatively low levels in the rat amygdala (Miquel et al. 1991). Receptor binding studies showed low but significant levels of this receptor in deep nuclei of the monkey and postmortem human amygdala (Hall et al. 1997; Law et al. 2009; Varnas et al. 2004). Our own ISH studies showed low levels of 5-HT1A expression in numerous cells of the rat CeL and CeM (unpublished observations). Low expression levels were also found in cells in the La, and low to moderate expression in numerous cells of the BL (Fig. 4b). Occasionally, coexpression of 5-HT1A and 5-HT2C mRNA was observed (Fig. 4c–c’’) (Bonn et al. 2012, 2013). Saha et al. (2010) found immunolabeling for 5-HT1A in neurons of the rat CeL and CeM, and documented that about 30 % of Ce cells projecting to the caudal dorsomedial medulla oblongata were positive for this receptor. Additionally, numerous cells of the rat BL and La were 5-HT1A-ir (Saha et al. 2010). The identity of 5-HT1A-expressing amygdaloid cells was addressed in studies in the rat using markers for pyramidal cells and for the different interneuronal populations. Thus, in immunohistochemical studies which did not differentiate between amygdaloid nuclei, 71.3 % of principal cells, 89.1 % of CB-ir cells, and 88.1 % of PV-ir cells were found to be 5-HT1A-ir (Aznar et al. 2003). Colocalization studies indicated the presence of 5-HT1A in about a third of neurokinin-1 receptor (NK-1)-ir neurons of the La and BL (Hafizi et al. 2011). Specific lesioning of these neurons, which represent about 40 % of large CCK-ir and 88 % of NPY-ir BL interneurons, increased anxiety-like behavior in a social interaction test (Truitt et al. 2009), indicating that these interneuron subclasses have anxiolytic properties. Our own double ISH studies documented that 50–60 % of NPY-producing interneurons of the La and BL express 5-HT1A mRNA (Fig. 4d–d’’), and that some of these additionally coexpress 5-HT2C mRNA (Fig. 4g–g’’’) (Bonn et al. 2013). Immunolabeling for 5-HT1A in rodents was generally shown to be somatic with some labeling of dendrites (Aznar et al. 2003; Hafizi et al. 2011; Morrison et al. 2011). The fact that this localization was seen in naïve animals without prior intracerebroventricular injection of colchicine, a method used to enhance somatic staining by blocking microtubule-dependent transport (e.g., Bombardi 2011; McDonald and Mascagni 2007), indicated that somatodendritic immunolabeling represents the physiological localization of the receptor in these neurons. Additionally, electrophysiological studies on isolated BL neurons suggested the presence of functional 5-HT1A on terminals of GABAergic neurons (Kishimoto et al. 2000; Koyama et al. 2002).
Early ISH studies documented high levels of 5-HT2A mRNA in the rat amygdala (Pompeiano et al. 1994). McDonald and Mascagni (2007), using three different antibodies, investigated the presence of the receptor in the rat BLC in specimen from both naïve and colchicine-injected rats. Although the immunolabeling patterns achieved using the various antibodies differed considerably, the authors proposed that moderate somatodendritic staining for 5-HT2A in pyramidal neurons particularly after colchicine injections represented specific staining. Among the non-pyramidal neurons, a subgroup previously shown to project to the mediodorsal thalamus was intensely labeled for 5-HT2A, and colocalization studies showed that the majority of PV-ir and a smaller subpopulation of SOM-ir interneurons colocalized the receptor, while large CCK-ir and CR-ir interneurons did not display significant colocalization. The authors additionally mentioned labeling of principal cells in the Ce and lack of immunoreactivity in the ICs. In a more recent detailed immunohistochemical study, Bombardi (2011) documented immunoreactivity for the receptor in the neuropil of the Ce and ICs of naïve animals. After colchicine injection, intensely labeled 5-HT2A-ir cells were observed in the CeLc, and numerous cells with light somatodendritic staining were found in the other Ce subregions. Additionally, numerous lightly and individual strongly stained cells were observed in the ICs. In the BLC, some neuropil labeling was present in non-colchicine-injected rats, and somatodendritic labeling for 5-HT2A was confirmed in pyramidal cells and GABAergic interneurons after colchicine application. Jiang et al. (2009) reported almost complete overlap of 5-HT2A- and PV-immunoreactivities in the BLC, with 87.3 % of 5-HT2A-ir cells colocalizing PV-immunoreactivity. Of the NK-1-ir cells in the BL, which represent another subgroup of interneurons in the BLC (Truitt et al. 2009) (see above), only a minority (8.1 %) colocalized 5-HT2A (Hafizi et al. 2011). Using electrophysiological analyses of adolescent rat BLC slices, Jiang et al. (2009) conclusively documented that serotonergic facilitation of GABA release was primarily mediated by 5-HT2A. 5-HT2A immunolabeling was shown also in the amygdala of other rodent species: numerous immunoreactive cells were documented in the hamster BLC (Morrison et al. 2011). Studies in mouse amygdala showed significant 5-HT2A ligand binding particularly in the BLC (e.g., Huang et al. 2004; Li et al. 2003; Schiller et al. 2003), but analysis of the cellular expression of the receptor has not been reported. Although 5-HT2A mRNA and binding was documented in the monkey BLC (Lopez-Gimenez et al. 2001), and 5-HT2A expression was found in the human amygdala (Guest et al. 2000), detailed nuclear and cellular localization studies for the receptor are also lacking in primate.
As for 5-HT2A, high 5-HT2C mRNA expression was shown in the rat amygdala in early investigations (Pompeiano et al. 1994). Immunohistochemical studies documented high numbers of 5-HT2C-reactive neurons in the BLC and cortical amygdaloid nuclei, and particularly in the Me and IC in rat amygdala without prior colchicine injections (Clemett et al. 2000). Using radioactive probes, highest expression levels of 5-HT2C mRNA in mouse amygdala were found in the La and somewhat lower levels in the BL and Me. 5-HT2C ligand binding was higher in the BL than in the La and Me (Li et al. 2003). Analysis of the autoradiographs shown in this article indicates very low levels of 5-HT2C mRNA and binding in the Ce. Recent radioactive ISH studies on rat amygdala showed high levels of 5-HT2C mRNA in the La and Me, and intermediate levels in the BL (Greenwood et al. 2012). Similar results were obtained by non-radioactive ISH studies in our laboratory (Fig. 4a), which documented most numerous 5-HT2C mRNA-reactive cells with high expression levels in the La, BM, Co, and Me. Particularly in the La, 5-HT2C mRNA-reactive cells apparently represented the majority of cells present in this nucleus. In the BL, comparatively few 5-HT2C mRNA-positive cells with high reactivity were observed (Bonn et al. 2012, 2013). CeM and CeLc contained some and numerous moderately mRNA-reactive cells, respectively, while in the CeL only few reactive cells were found. Weakly reactive cells were observed in the ICs (Fig. 4a). Additionally, the study by Bonn et al. (2013) showed that 5-HT2C mRNA was expressed in 30–40 % of NPY mRNA-producing interneurons in La and BL (Fig. 4e), and was coexpressed with 5-HT1A in individual NPY mRNA-producing neurons (Fig. 4g–g’’’; see above). 5-HT2C binding and/or ISH studies revealed presence of the receptor in the human and non-human primate amygdala, with high concentrations in the dorsomedial region in the human (Pasqualetti et al. 1999). In contrast to cortical regions, where somatodendritic 5-HT2C immunolocalization was documented in GABAergic and other, possibly pyramidal neurons (Liu et al. 2007), the subcellular localization of the receptor protein in amygdaloid cells has not been described in detail in the IHC studies.
Early immunohistochemical studies showed intensely 5-HT3-ir neurons in the amygdala (Morales et al. 1996), and combined IHC/ISH documented expression of 5-HT3 mRNA in GABAergic neurons (Morales and Bloom 1997). Subsequently, the receptor protein was found to be exclusively localized in GABAergic interneurons of the amygdala, particularly in the La, BL, BM, and cortical nuclei (Mascagni and McDonald 2007). A small percentage (8–16 %) of 5-HT3-ir neurons colocalized CCK, and apparently belonged to the subtype of large CCK-ir interneurons; very few 5-HT3AR-ir neurons expressed CR, VIP or PV, and none were SOM- or CB-ir (Mascagni and McDonald 2007). Around 50 % of La and BL 5-HT3-producing interneurons showed expression of cannabinoid-receptor 1 (CB1) (Morales et al. 2004). CB1-immunoreactivity has previously been shown to be associated mainly with large CCK-ir interneurons, again suggesting that this type of interneuron is susceptible to 5-HT3-mediated serotonergic modulation. However, a significant number of 5-HT3-ir interneurons did not colocalize any of the markers applied in the cited studies. Additionally, Bonn et al. (2013) documented that NPY mRNA-producing neurons do not colocalize 5-HT3 mRNA (Fig. 4f). Thus, the majority of interneurons expressing this 5-HTR remain to be identified. Concerning the subcellular localization of the receptor, all IHC studies showed an exclusively somatic localization of 5-HT3 immunoreaction product without prior colchicine treatment (Mascagni and McDonald 2007; Morales and Bloom 1997). However, electrophysiological investigations on dissociated BL neurons indicated a presynaptic localization of 5-HT3 in GABAergic interneurons in the amygdala facilitating GABA release (Koyama et al. 2000), which can be counteracted by presynaptic 5-HT1A activation via a G protein-mediated mechanism (Koyama et al. 2002).
5-HTR equipment of identified amygdaloid neuron types
Knowledge of the receptor equipment of the different neuron types in the amygdala is a necessary basis for interpreting effects of intra-amygdaloid serotonin or 5-HTR agonist/antagonist injections. Although the picture is far from clear, and some findings are controversial, a short synopsis is attempted which summarizes the available data described above. It appears that most principal and interneurons in the different amygdaloid nuclei, at least in the rat, produce several subtypes of 5-HTR. Pyramidal cells of the La are positive for 5-HT1A and 2A, and most likely express also 5-HT2C. The majority of pyramidal cells of the BL, on the other hand, also apparently produce 5-HT1A and 5-HT2A, but the majority is negative for 5-HT2C mRNA. All receptors appear to be localized somatodendritically in pyramidal cells. The enhancement of somatodendritic staining for 5-HT2A after colchicine injections argues for some transport of the receptor protein into neuronal processes, i.e., distal dendrites or axons. Of the GABAergic interneurons in the La and BL, different subclasses express different receptors. Thus, most PV- and CB-ir interneurons produce 5-HT1A, and the majority of PV-ir neurons are 5-HT2A-reactive. A small number of SOM-ir interneurons also produces 5-HT2A. The subgroup of NPY-producing SOM-ir interneurons falls into two large groups, with more than half expressing 5-HT1A mRNA, and about a third producing 5-HT2C mRNA. These groups apparently do not overlap to a significant extent. Somatodendritic localization in interneurons is probable for all receptors, and lack of enhancement of 5-HT2A labeling using colchicine indicates that this receptor is not transported into neurites to a significant extent. On the other hand, electrophysiological evidence indicates an additional possible localization of 5-HT1A in preterminal GABAergic axons. A subpopulation of large CCK- and very few PV-, CR- and VIP-ir neurons are 5-HT3-ir. Neither pyramidal cells nor CB-, SOM-, and NPY-ir interneurons produce 5-HT3. Immunoreactivity for 5-HT3 is restricted to the somata, although preterminal axonal presence of the receptor has been indicated by electrophysiological means.
For the Ce, the studies indicate moderate somatodendritic presence of 5-HT1A and 5-HT2A but lack of 5-HT2C in CeL neurons. Numerous CeM and CeLc neurons, on the other hand, apparently possess all three receptor subtypes, with CeLc neurons showing particularly strong labeling for 5-HT2A after colchicine indicating that these cells export the receptor, possibly into their projections to target areas.
Effects of genetic variations in serotonin transmission molecules on the amygdala
In 1996, a common variable number tandem repeat polymorphism in the transcriptional control region of the human 5-HTT gene (SLC6A4) was reported (Lesch et al. 1996). The short (s) allelic version of this polymorphism leads to reduced expression of 5-HTT, and numerous subsequent studies have shown that individuals carrying the s-allele display an increased risk for anxiety disorders and depression particularly if they are additionally subjected to early-life adversity or repeated stressful experience (Holmes 2008). S-allele carriers exhibit greater amygdala neuronal activity in response to fearful stimuli compared with individuals homozygous for the long allele (Hariri et al. 2002). To date, association studies have led to the identification of further genetic variants of serotonin transmission molecules which might represent candidate genes for affective disorders. Numerous genetically modified mouse models with targeted disruption, or over- or underexpression of these genes have been generated, and have been subjected to behavioral and other experimental studies, as comprehensively reviewed by Holmes (2008). In those cases where functional imaging studies were carried out in individuals bearing risk genes, alterations of amygdala activity in response to different kinds of emotional stimuli and stress have been found (e.g. Canli and Lesch 2007; Holmes 2008; Kilpatrick et al. 2011; Lee and Ham 2008).
Although investigations addressing specific alterations in the amygdala, particularly on a cellular level, in the various mouse models are surprisingly few, some alterations in neurochemical, structural, and functional parameters were reported. Thus, differential alterations in the density and/or signal transduction efficiency were shown for various receptors in 5-HTR knockout mice as reviewed by Compan (2007). In 5-HTT-deficient mice, which display increased anxiety-like behavior in a variety of behavioral paradigms, a reduced 5-HT1A binding density was found in the amygdala particularly of females of this genotype (Li et al. 2003). Also, the density of 5-HT2C binding in the BLC was significantly increased (Li et al. 2003). On the other hand, Moya et al. (2011) detected an increase in the more edited, less effective isoforms of 5-HT2C in the amygdala of 5-HTT-deficient mice. Morphological studies documented a significantly increased dendritic spine density of BLC pyramidal neurons in 5-HTT-deficient compared to wild-type mice (Nietzer et al. 2011; Wellman et al. 2007). Interestingly, while wild-type mice displayed an increased spine density after a repeated loser experience, a further increase in spine density was absent in 5-HTT-deficient mice indicating a conspicuous lack of neuroplastic responses in this social stress paradigm (Nietzer et al. 2011). In a recent study on the effects of a complete lack of neuronal serotonin in Tph2-deficient mice, a significant reduction was found for the number and density of GABAergic neurons in the BL and La compared to wild-type animals (Waider et al. 2012).
Effects of stress and anxiety-related stimuli on serotonergic transmission in the amygdala
Numerous studies, primarily conducted in rats, have suggested that acute stress- and anxiety-related stimuli as well as systemic or intra-DR application of anxiogenic drugs and anxiety-related peptides such as CRF or its analogue, urocortin, preferentially alter the activity of subpopulations of serotonergic neurons located within the mid-rostrocaudal and caudal DR (Abrams et al. 2005; Amat et al. 2005; Evans et al. 2009; Gardner et al. 2005; Grahn et al. 1999; Hale et al. 2008; Lowry et al. 2005; McEuen et al. 2008; Spannuth et al. 2011; Staub et al. 2005, 2006) including serotonergic neurons projecting to the amygdala (Abrams et al. 2005; Commons et al. 2003; Hale et al. 2008). Early microdialysis studies assessing extracellular serotonin levels after stress in the amygdala of rats were carried out without specific reference to individual nuclei and yielded inconsistent results. Thus, while Rueter and Jacobs (1996) showed elevated amygdaloid serotonin concentrations upon stressful experiences such as tail pinch, cat exposure, and swimming, Kirby et al. (1995) observed reduced serotonin concentrations in the amygdala after forced swimming. Interestingly, intrinsic sex differences in the stress reactivity of the serotonergic amygdala innervation were reported by Duchesne et al. (2009). The authors documented that in rats with a history of mild stress experience, females exhibited higher serotonin levels in amygdala tissue extracts while males showed higher amygdaloid 5-HT metabolism, which was additionally upregulated by stress in postnatally handled animals.
More recent studies analyzed serotonin levels in specific amygdaloid nuclei in relation to stressful experience or after application of anxiogenic peptides to the DR. Mo et al. (2008) showed that restraint stress in awake male rats during the dark cycle induced significant increases of extracellular serotonin in the Ce, but not in the Me. The stress-induced enhanced serotonin release could be blocked by intracerebroventricular injections of CRF1/2 receptor antagonist. Accordingly, earlier studies had shown that infusion of CRF into the DR induced freezing behavior which was positively correlated with an immediate increase in serotonin release in the Ce (Forster et al. 2006) supporting a functional interplay between the CRF-mediated modulation of DR neuron activity, serotonergic modulation of Ce neuron activity, and behavioral alterations. Commons et al. (2003) documented that serotonergic neurons in the DR projecting to the Ce produce CRF and target CRF-ir neurons. Ce-CRF neurons, in turn, have been shown to project to the DR (Gray 1993), and presumably influence the activity of serotonergic neurons projecting to the medial frontal cortex via CRF2 receptors (Forster et al. 2006). Interestingly, the increased serotonin release in the Ce elicited by CRF-application to the DR via CRF2 receptors was significantly enhanced in chronically d-amphetamine-treated rats, suggesting a contribution of the DR-CRF-Ce circuit in the dysphoric symptoms of drug withdrawal (Scholl et al. 2010).
In the BLC, as in the Ce, increases in extracellular serotonin concentrations were observed after application of anxiogenic peptides to the DR, and in experimental animals subjected to specific stress paradigms, particularly inescapable or uncontrollable stress (Amat et al. 1998; Christianson et al. 2010), and/or after induction of anxiety states (Amat et al. 1998; 2004; Christianson et al. 2010; Kawahara et al. 1993; Maier and Watkins 2005; Mitsushima et al. 2006; Zanoveli et al. 2009). Mitsushima et al. (2006) showed that serotonin release in the BL was higher in males than in females under control conditions, and was increased after restraint stress in both sexes with significantly higher elevations in females. These findings were proposed to represent neurochemical correlates of sex differences in the response to stress, suggesting that sex hormones might influence amygdaloid emotion circuits through hormone-dependent regulation of serotonergic innervation. In fact, estrogen levels have been shown to influence Tph2 production in different subregions of the DR including the amygdala-projecting caudal dorsomedial region (Hiroi et al. 2006).
Apparently inconsistent results concerning the electrophysiological, neurochemical and behavioral consequences of stress-induced alterations of serotonin levels in the various nuclei have led to divergent interpretations of serotonin functions. While some studies on serotonin in the Ce concluded that stress-induced increased release of serotonin is necessary for eliciting endocrine, autonomous and behavioral responses to stress and fear (Forster et al. 2006; Scholl et al. 2010), its stimulatory effect on the hypothalamo-pituitary-adrenocortical axis presumably being mediated via inhibitory 5-HT1A receptors (Feldman et al. 2000), others found that application of 5-HT1A receptor agonist into the Ce reduces conditioned fear responses (Groenink et al. 2000), and that lentivirally mediated downregulation of this receptor in the Ce in mice is anxiogenic (Li et al. 2012). Similar apparently inconsistent results were found for behavioral effects of 5-HT1A-mediated transmission in the BL. While injections of selective 5-HT1A agonists into the BLA impaired the acquisition and expression of conditioned fear and conditioned defeat, respectively, in rats and syrian hamsters (Li et al. 2006; Morrison and Cooper 2012), activation of BL 5-HT1A receptors appeared to be ineffective or even enhanced anxiety-like behavior in social interaction tests (Gonzalez et al. 1996) and for escape behavior in the elevated T-maze (Zangrossi et al. 1999), indicating that pharmacological activation of 5-HT1A receptors in the BL differentially affects conditioned and unconditioned emotional responses (Morrison and Cooper 2012).
The observation that inescapable or uncontrollable stress led to exaggerated serotonin release in the BLC during subsequent stressful experience (e.g. footshocks: Amat et al. 1998; juvenile social exploration: Christianson et al. 2010) and to concomitantly enhanced anxiety-related behavior (reduced social exploration: Christianson et al. 2010), led to the suggestion that serotonergic transmission in deep nuclei may play a critical role in mediating the potentiation of fear and anxiety produced by uncontrollable stressors. Generally, anxiety- and fear-like behaviors are correlated with an increase in output of the BLC to limbic structures that mediate specific behaviors (Christianson et al. 2010). Pharmacobehavioral manipulations in these stress paradigms indicated that anxiety-increasing serotonin effects in the BLC were mediated by the 5-HT2C receptor (Campbell and Merchant 2003; Christianson et al. 2010). Recent findings by Vicente and Zangrossi (2011) supported 5-HT2C-mediated anxiogenic actions of elevated serotonin in the BL by showing that inhibitory avoidance acquisition was specifically facilitated after intra-BLA administration of serotonin and of a 5-HT2C agonist (MK-212). Electrophysiological studies using whole-cell recordings documented enhanced action potential firing of La principal neurons in rat amygdala slices prepared 1 h after in vivo intraperitoneal (i.p.) administration of the selective serotonin reuptake inhibitor (SSRI) fluoxetine, a time point at which the animals displayed enhanced anxiety-like behavior in the elevated plus maze test. Activity increase of La neurons was also found after in vitro fluoxetine application directly onto amygdala slices (Ravinder et al. 2011). Based on earlier findings which had shown that conditional stimulus-evoked spike firing in La neurons is increased after fear conditioning (Maren and Quirk 2004), and that SSRI treatment increased extracellular serotonin in the amygdala (Bosker et al. 2001), Ravinder et al. (2011) proposed that a serotonin-dependent increase in neuronal excitability was one possible mechanism leading to the increased anxiety-like behavior following acute fluoxetine treatment. Vicente and Zangrossi (2011) delivered conclusive evidence that this acute SSRI effect is mediated by 5-HT2C receptors in the BLC. On the other hand, an inhibitory effect of serotonin on glutamate-evoked La neuron activity was shown in single-unit recordings in the La of rats by Stutzmann et al. (1998). The authors additionally documented that inhibition of glutamatergic activity by iontophoretically applied serotonin occurred in part through activation of GABAergic interneurons (Stutzmann and LeDoux 1999). In acute BL slices, serotonin application directly activated GABAergic interneurons, presumably via excitatory 5-HT2A receptors, and increased the frequency of inhibitory synaptic events in projection neurons. The serotonin effect was dose- and time-dependent, with high doses or prolonged application reducing excitation of inhibitory interneurons via presynaptic inhibition of glutamate release (Rainnie 1999). Also, intra-BL injections of the SSRI citalopram reduced freezing induced by conditioned fear stress (Inoue et al. 2004), and application of serotonin or of 5-HT1A and/or 5-HT2A agonists into the BL reduced the tonic immobility induced by restraint stress in guinea pigs (Leite-Panissi et al. 2006). These findings were interpreted to indicate that enhanced serotonergic transmission in the amygdala, particularly in the BL, decreased conditioned fear and anxiety (Inoue et al. 2004; Leite-Panissi et al. 2006). In a recent review, Jasinska et al. (2012) hypothesized that it might even be a reduction of serotonin release in the amygdala which contributes to behavioral and physiological reactions to severe and uncontrollable stressors. The authors argued that such stressors, which produce long-lasting behavioral changes such as learned helplessness in experimental animals (Maier and Watkins 2005), specifically activate serotonergic neurons in the ventrolateral DR (Crawford et al. 2010; Gardner et al. 2009; Hale et al. 2011; Roche et al. 2003), initiating raphe-raphe interactions which inhibit forebrain-projecting DR neurons via serotonin volume transmission acting on inhibitory autoreceptors. The resulting reduction of serotonin release in target regions then presumably leads to a lack of serotonergic activation of inhibitory interneurons controlling the activity of BL pyramidal neurons according to the findings described above (Rainnie 1999; Stutzmann and LeDoux 1999), to consequent hyperexcitation of the BLC output, and thus to an imbalance of emotion circuits (Jasinska et al. 2012).
A dual activity-increasing effect of serotonergic transmission on different neuron types in the rat La and BL was also indirectly suggested by Hale et al. (2010), who documented that i.p. administration of anxiogenic drugs induced correlated increases in the expression of the activity marker c-fos in serotonergic neurons of the mid-rostrocaudal DR and in PV-ir inhibitory interneurons of the BL. The activity marker was also strongly induced in non-PV-ir neurons, among them most likely pyramidal cells which have been shown to be activated by acute and repeated restraint stress (Reznikov et al. 2008). Hale et al. (2010) suggested that c-fos expression in glutamatergic projection neurons might be involved in the anxiety response itself while activation of PV-expressing interneurons, potentially via serotonin acting at 5-HT2A receptors on these PV-expressing interneurons (see below), may occur after activation of glutamatergic projection neurons and contribute to termination of the anxiety response.
Few experimental studies on serotonin availability and function in the amygdala exist for non-human primates. Rainnie (Rainnie 2003) reported species differences in the responses elicited by serotonin on BL slices: while in the rat acute slice preparation, as described above, application of serotonin lead to inhibition of >85 % of BLA projection neurons via excitation of local GABAergic interneurons (Rainnie 1999; Stutzmann and LeDoux 1999), exogenous application of serotonin to primate slice preparations directly inhibited >60 % of projection neurons and evoked a direct postsynaptic excitation in 20 % of neurons tested. Direct measurements of serotonin in the human amygdala are lacking, but results from imaging studies after different manipulations of the serotonergic system again yield an inconsistent picture: tryptophan depletion and 7-day or subchronic (21-day) SSRI administration, which should theoretically lead to decreased and enhanced availability of serotonin, respectively, led to an increase and a reduction in amygdala reactivity, respectively (Arce et al. 2008; Harmer et al. 2006; van der Veen et al. 2007); on the other hand, acute SSRI administration induced increased amygdala reactivity to emotional stimuli (Bigos et al. 2008).
Summary, conclusions and remaining questions
Although the findings reviewed yield an enormously complex and occasionally puzzling picture of the role and mode of action of serotonin for information processing in the amygdala, they also provide a basis for thought-provoking and pertinent conclusions, and, ultimately, offer direction to further research. Thus, there is no doubt that the amygdala receives dense serotonergic innervation across species, that serotonergic transmission in the amygdala is relevant for emotional stimulus processing in rodents and in primates including man, and that imbalances in this transmission accompany or even underlie emotional dysregulation phenomena. However, there are significant differences in the serotonergic innervation density of specific amygdaloid subregions, particularly the Ce, between the species. This implies that interactions between the serotonergic and other extrinsic or intrinsic amygdaloid systems in this nucleus, and their effect on the output of the amygdala to Ce target areas, also vary. It would be interesting to address whether these differences might reflect species-specific variations in the modulation of behavioral and endocrine responses elicited upon the activation of core circuits ensuring survival in a threatening environment. These common “survival circuits”, in which the amygdala features as a central component across species, have been proposed by LeDoux in his recent enlightening paper (LeDoux 2012). Accordingly, in spite of the noticeable differences in details, there are numerous morphological and functional analogies of serotonergic parameters in the amygdala among rodents and primates. Therefore, investigations of these parameters in rodents, particularly in mouse models for genetic variations causing emotional dysregulation, appear suited to further our understanding of serotonin–amygdala interactions also in humans. It is, however, also obvious from the reviewed literature that it is essential to carefully design and perform detailed analyses of the regulation of serotonin transmission in different amygdaloid nuclei, so that experimental paradigms chosen for analyses provide controlled and reproducible conditions which allow addressing the problem of changes in serotonin release and/or action on different receptors over time, in different (sub)nuclei, and in answer to different environmental challenges. From the short overview on effects of stress and anxiety on serotonergic transmission in the amygdala given above, it is obvious that release and effect of serotonin in the amygdala depend on the concrete paradigms tested, which underlines the proposition that differential serotonergic mechanisms, in the amygdala and other areas of the “survival circuits”, enable flexible and alternative behavioral responses to variable environmental necessities (Compan 2007; LeDoux 2012). In accordance with this suggestion, and although differences in methodology and subnuclear targets analyzed preclude direct comparisons even between studies carried out in specific nuclei such as the Ce and BLC, the analysis of the literature data cited suggests that stress-induced changes in serotonergic tone in the amygdala mediate differential effects via dose- and time-dependent selective activation of distinct receptor subtypes localized on specific types of target neurons. Of course, effects in other nuclei, which have not been studied in such detail, add to the complexity of the amygdala output outcome of serotonergic transmission. Clearly, understanding of the various serotonin effects in the amygdala requires more detailed knowledge of the cellular targets and molecular mechanisms of serotonergic transmission than available to date. Thus, it is necessary to further elucidate the cellular and subcellular distribution of all relevant 5-HTR, so that effects of serotonin via wiring or volume transmission, or via pre- or postsynaptic actions on specific receptors, can be considered and addressed, for instance in further electrophysiological studies. Results of these experiments could then form a platform on which pharmacobehavioral analyses of rodent models can be performed which may deliver insights into the possibility of preventive and/or therapeutic interventions in human disorders.
The writing of this article and the authors’ related research were supported by the Deutsche Forschungsgemeinschaft (SFB 581/B9 and Z3, RTG 1253, SFB TRR 58/A1 and A5, KFO 125).