Amygdala physiology in pain

Abstract

The amygdala has emerged as an important brain area for the emotional-affective dimension of pain and pain modulation. The amygdala receives nociceptive information through direct and indirect routes. These excitatory inputs converge on the amygdala output region (central nucleus) and can be modulated by inhibitory elements that are the target of (prefrontal) cortical modulation. For example, inhibitory neurons in the intercalated cell mass in the amygdala project to the central nucleus to serve gating functions, and so do inhibitory (PKCdelta) interneurons within the central nucleus. In pain conditions, synaptic plasticity develops in output neurons because of an excitation-inhibition imbalance and drives pain-like behaviors and pain persistence. Mechanisms of pain related neuroplasticity in the amygdala include classical transmitters, neuropeptides, biogenic amines, and various signaling pathways. An emerging concept is that differences in amygdala activity are associated with phenotypic differences in pain vulnerability and resilience and may be predetermining factors of the complexity and persistence of pain.

Interest in a role of the amygdala in pain stems from at least two factors. On the one hand, pain has a strong emotional-affective dimension and is characterized, if not defined, by its unpleasantness (Merskey et al., 1979), and the amygdala has long been known as a key player in emotions and associated disorders. On the other hand, anatomical and functional evidence provided a direct link to the “pain system” through nociceptive inputs (Gauriau & Bernard, 2004) and projections to pain modulatory centers (Heinricher, Tavares, Leith, & Lumb, 2009). Research over the past two decades has identified amygdala processing of nociceptive information, plasticity in pain conditions, and behavioral consequences (Neugebauer, Li, Bird, & Han, 2004; Neugebauer, 2015; Veinante, Yalcin, & Barrot, 2013; Thompson & Neugebauer, 2017). The analysis of cell type and synapse specific mechanisms is an ongoing area of research.

Our current overall concept of amygdala function in pain can be described as follows (Fig. 1). In pain conditions, increased nociceptive input (and/or “stress signals” in so-called functional pain conditions without any tissue pathology) drives hyperexcitability of amygdala output neurons. One consequence of increased amygdala output is the facilitation of spinal, and perhaps peripheral, nociceptive processing. Another effect is the deactivation of (medial) prefrontal cortical control centers, resulting in the well-documented cognitive deficits associated with pain conditions (Moriarty, McGuire, & Finn, 2011; Apkarian et al., 2004b; Ji et al., 2010) and in a loss of cortical control of amygdala processing (Kiritoshi & Neugebauer, 2018). The combination of these vicious cycles of gain and loss of function allows the persistence of pain-related neuroplasticity in the amygdala and drives pain behaviors and pain persistence (Neugebauer, 2015; Thompson & Neugebauer, 2017).

Figure 1. Current concept of amygdala function in pain.

See text for details.

Amygdala neurocircuitry of “pain processing”

Inputs

The amygdala receives pain-related information mainly through two lines of input (Fig. 2). The external lateral parabrachial area (PB) in the brainstem provides highly preserved nociceptive information (also referred to as the “direct pathway” (Liu et al., 2011)), whereas multimodal sensory information reaches the amygdala from thalamic nuclei and cortical areas (Neugebauer et al., 2004; Thompson & Neugebauer, 2017). The discovery of the spino-parabrachio-amygdala pain pathway to the lateral and capsular divisions of the central nucleus of the amygdala (Bernard, Peschanski, & Besson, 1989; Gauriau & Bernard, 2004) led to the identification of neurons in these amygdala regions (CeLC) that were activated by orthodromic stimulation in the parabrachial area and responded exclusively or predominantly to noxious stimuli (Bernard, Huang, & Besson, 1992; Neugebauer & Li, 2002). The term “noxious” is defined as actually or potentially tissue damaging, and refers to a stimulus that results in withdrawal reflex responses and/or is perceived as painful. The presumed nociceptive input from the PB (Bernard, Alden, & Besson, 1993) was localized in brain slices as the fiber tract dorsomedial to the central nucleus and ventral to, but outside, the caudate-putamen; and synaptic responses of CeLC neurons to electrical stimulation of these fibers demonstrated its functional significance (Neugebauer, Li, Bird, Bhave, & Gereau, 2003). These findings have since been confirmed by others (Lopez de Armentia & Sah, 2007; Miyazawa, Takahashi, Watabe, & Kato, 2018; Cheng et al., 2011; Ikeda, Takahashi, Inoue, & Kato, 2007) and validated definitively with optogenetic approaches (Sugimura, Takahashi, Watabe, & Kato, 2016). The PB input is highly peptidergic and the sole source of calcitonin gene-related peptide (CGRP) in the amygdala (Han, Li, & Neugebauer, 2005; Dobolyi, Irwin, Makara, Usdin, & Palkovits, 2005; Shinohara et al., 2017).

Figure 2. Neurocircuitry of amygdala pain mechanisms.

See text for details. CeA, central nucleus; LA-BLA, lateral-basolateral nuclei; ITC, intercalated cells; Glu, glutamate.

CeLC neurons with PB input also receive excitatory and feedforward inhibitory inputs from the lateral-basolateral amygdala (LA-BLA) (Fig. 2). The LA-BLA network receives and integrates multimodal sensory, including nociceptive, information from thalamic nuclei (midline, posterior intralaminar and posterior regions) and cortical regions, including anterior cingulate, insular and sensory association cortices (Pape & Pare, 2010; Lanuza, Moncho-Bogani, & LeDoux, 2008; Janak & Tye, 2015; Duvarci & Pare, 2014). Systems electrophysiology studies suggested that LA-BLA and PB inputs converge onto multireceptive CeLC neurons that respond not only to noxious but also to innocuous stimuli (Neugebauer & Li, 2002). Feedforward inhibition of CeLC neurons from the LA-BLA network involves a cluster of GABAergic interneurons in the intercalated cell mass (ITC), which are also a direct or indirect target of medial prefrontal cortical influences (Ren & Neugebauer, 2010; Ren et al., 2013; Kiritoshi & Neugebauer, 2018) and have been implicated in fear extinction (Likhtik, Popa, pergis-Schoute, Fidacaro, & Pare, 2008; Sotres-Bayon & Quirk, 2010; Amano, Unal, & Pare, 2010; Amir, Amano, & Pare, 2011; Pare & Duvarci, 2012; Duvarci & Pare, 2014). PKCdelta interneurons in the central nucleus also receive LA-BLA input and may serve gating functions for central amygdala output neurons (Janak & Tye, 2015).

Cell types

CeLC neurons are GABAergic and contain distinct populations that express neuropeptides (corticotropic releasing factor (CRF), somatostatin and/or dynorphin) or PKCdelta. PKCdelta neurons are distinct from peptidergic CeLC neurons, but a neuron may express more than one peptide. For example, CRF neurons co-express somatostatin and/or dynorphin peptides in rats (Pomrenze et al., 2015), and neurons in the lateral division of the central nucleus co-express CRF, somatostatin, neurotensin and tachykinin 2 (encoding neurokinin B) mRNA in mice (McCullough, Morrison, Hartmann, Carlezon, Jr., & Ressler, 2018), although a stricter segregation has been reported for mice by others (Fadok et al., 2017; Hunt, Sun, Kucukdereli, Klein, & Sah, 2017; Sanford et al., 2017). While CRF has been shown to play an important role in amygdala pain mechanisms (Ji & Neugebauer, 2007; Ji, Fu, Ruppert, & Neugebauer, 2007; Ji & Neugebauer, 2008; Fu & Neugebauer, 2008), the CeLC cell types involved in pain-related functions remain to be determined. Importantly, nearly half of the CRF neurons receive CGPR input from the PB (Harrigan, Magnuson, Thunstedt, & Gray, 1994).

Projections

CRF and somatostatin rather than PKCdelta neurons are known to form long-range projections to brainstem (e.g., periaqueductal grey, PAG; parabrachial area, PB), basal forebrain and hypothalamic areas involved in behavioral expression and pain modulation (Fig. 2) (Pomrenze et al., 2015; Penzo, Robert, & Li, 2014). It remains to be determined, however, if output from “pain-related” CeLC neurons directly or indirectly (via the medial division of the central nucleus or bed nucleus of the stria terminalis) regulates these target regions to modulate pain associated behaviors. Cell type-specific projections to targets involved in pain modulation are currently an area of intense research (see for example Li & Sheets, 2018). In contrast to the central nucleus, BLA neurons project strongly to various cortical regions, including the infra- and pre-limbic and anterior cingulate cortices, for direct synaptic excitation and feedforward inhibition of cortical pyramidal cells (Kiritoshi, Ji, & Neugebauer, 2016; McGarry & Carter, 2016; Cheriyan, Kaushik, Ferreira, & Sheets, 2016). There is evidence to suggest that these amygdalo-cortical interactions play an important role in cognitive deficits and loss of cortical control of amygdala processing in pain conditions (Ji et al., 2010; Bushnell, Ceko, & Low, 2013; Vachon-Presseau et al., 2016b; Kiritoshi & Neugebauer, 2018) (see “Amygdala neuroplasticity in pain models and pain conditions“).

Amygdala neuroplasticity in pain models and pain conditions

Activity changes

Preclinical studies.

Electrophysiological single-unit recordings in anaesthetized rats found increased ongoing (background) activity and responsiveness of CeLC neurons to peripheral stimuli in an arthritis pain model within a few hours after induction (Neugebauer & Li, 2003). This study identified neurons as receiving input from the parabrachial area (PB) and found activity changes only in multireceptive CeLC neurons that responded to noxious stimuli and, to a lesser degree, also to innocuous stimuli, but not in nociceptive-specific neurons that were activated exclusively by noxious stimuli. The interpretation was that additional input besides from PB is required for pain related changes, and subsequent studies focused on these multireceptive neurons (Li & Neugebauer, 2004b; Li & Neugebauer, 2004a; Han et al., 2005; Li & Neugebauer, 2006; Ji & Neugebauer, 2007; Ji & Neugebauer, 2009; Ji & Neugebauer, 2014; Kim, Thompson, Ji, Ganapathy, & Neugebauer, 2017). Changes in background activity such as irregular and burst firing, and increased evoked responses of multireceptive CeLC neurons were also observed in a chronic neuropathic pain model (spinal nerve ligation, SNL) (Goncalves & Dickenson, 2012; Goncalves, Friend, & Dickenson, 2015; Ji et al., 2017).

Levels of neuronal activity markers in the CeLC also increased in different pain models. Increased c-Fos expression in the central nucleus has been shown in the formalin pain model (Miyazawa et al., 2018), in visceral pain models of noxious colorectal distension (Traub, Silva, Gebhart, & Solodkin, 1996), esophagitis (Suwanprathes, Ngu, Ing, Hunt, & Seow, 2003) and cystitis (Bon, Lanteri-Minet, Michiels, & Menetrey, 1998), and in the chronic constriction injury (CCI) model of neuropathic pain (Seno et al., 2018). Activation of extracellular signal-regulated kinase (ERK) in the CeLC was increased in models of acute inflammatory pain (Carrasquillo & Gereau, 2007; Kolber et al., 2010) and muscle pain (Cheng et al., 2011). CRF protein and mRNA in the central nucleus were increased in cystitis (Nishii, Nomura, Aono, Fujimoto, & Matsumoto, 2007) and colitis (Greenwood-Van Meerveld, Johnson, Schulkin, & Myers, 2006) models of visceral pain and in neuropathic pain models (Rouwette et al., 2011; Ulrich-Lai et al., 2006). Glucocorticoid receptor mRNA expression in the central nucleus also increased in a neuropathic pain model (Ulrich-Lai et al., 2006), and metabotropic glutamate receptor mGluR1 and mGluR5 protein in the central nucleus increased in an arthritis pain model (Neugebauer et al., 2003). This work provided an initial framework for the site and neurochemical changes that occur in the amygdala during pain.

Studies in humans.

Neuroimaging studies in humans confirm and validate preclinical evidence for amygdala activation by painful stimuli and increased activation in pain conditions (Simons et al., 2014a; Apkarian, Bushnell, Treede, & Zubieta, 2005; Vachon-Presseau et al., 2016a; Bingel & Tracey, 2008). In normal subjects (healthy volunteers), activation of the amygdala was found in response to noxious mechanical, thermal and chemical (capsaicin) stimuli (Simons et al., 2014a). Thulium-YAG laser-evoked heat pain stimuli without a concomitant tactile component also produced amygdala activation (fMRI signal changes) in healthy volunteers (Bingel et al., 2002; Bornhovd et al., 2002). It should be noted that some studies also reported amygdala deactivation in response to painful stimuli such as rectal distension (Berman et al., 2006) and others (reviewed in Neugebauer et al., 2004; Apkarian et al., 2005). Amygdala activation has also been linked to top down pain modulation through enhanced functional connectivity with the rostral anterior cingulate cortex (Bingel & Tracey, 2008); and there is evidence for activation of the μ-opioid system in the amygdala detected by radioligand positron emission tomography (PET) during sustained muscle pain induced by infusion of hypertonic saline into the jaw muscle (Zubieta et al., 2005; see also Apkarian et al., 2005).

Importantly, electrophysiological studies in patients with epilepsy (Liu et al., 2011) found evidence for a direct nociceptive pathway and an indirect ventral to dorsal pathway to the central nucleus (dorsal region in the human amygdala) by measuring local field potentials in response to thulium-YAG laser-evoked heat pain stimuli (Liu et al., 2010; Liu et al., 2011). These clinical neuroimaging and electrophysiology data validate the results of preclinical work on the pain-related amygdala circuitry processing direct nociceptive input from PB and multimodal thalamo-cortical input from LA-BLA as discussed earlier (see “Inputs” in “Amygdala neurocircuitry of ‘pain processing’“).

A number of neuroimaging studies also reported amygdala activation in patients with pain conditions such as irritable bowel syndrome (IBS), fibromyalgia, complex regional pain syndrome (CRPS), migraine, and osteoarthritis (Simons et al., 2014a; Boadas-Vaello, Homs, Reina, Carrera, & Verdu, 2017; Apkarian et al., 2005). Evidence for activation of the amygdala in so-called functional pain syndromes such as IBS is particularly strong. Regional cerebral blood flow measurement with positron emission tomography (PET) consistently showed increased amygdala activation in IBS patients during visceral stimulation (rectal distension) (Mayer et al., 2005; Tillisch, Mayer, & Labus, 2011; Naliboff et al., 2003). Functional MRI studies in IBS patients found greater positive resting-state functional connectivity between the amygdala and various brain areas, including insula, other cortical regions, and midbrain, compared to healthy controls, and this evidence for increased neural synchrony or activity correlated positively with IBS symptom severity (Qi et al., 2016). IBS patients with visceral hypersensitivity showed increased positive resting-state functional connectivity of the amygdala within the default mode network compared to “normosensitive” IBS patients (Icenhour et al., 2017). In patients with CRPS, pain intensity correlated positively with amygdala volume in a structural MRI (T1) study (Barad, Ueno, Younger, Chatterjee, & Mackey, 2014). Resting-state functional connectivity (fMRI study) from amygdala to cortical (including prefrontal and anterior cingulate cortices) and subcortical (including basal ganglia) regions was increased in pediatric patients with CRPS, correlated with pain-related fear, and was decreased after rehabilitation treatment (Simons et al., 2014b). Migraine patients showed increased amygdala activity (fMRI) in response to negative, but not positive or neutral, emotional stimuli (International Affective Picture System) (Wilcox et al., 2016) and enhanced functional connectivity between amygdala and (visceroceptive) insular cortex (Hadjikhani et al., 2013). In osteoarthritis patients, pain evoked by pressure applied to the most sensitive part of the knee correlated with amygdala activity (fMRI) (Baliki et al., 2008). Chronic back pain patients showed reductions in endogenous activation of the μ-opioid system in the amygdala detected by radioligand PET (Martikainen et al., 2013).

Importantly, an emerging new concept links preexisting structural and functional properties of the cortico-limbic brain circuitry to the risk for developing chronic pain (Vachon-Presseau et al., 2016a). Anatomical (T1), diffusion tensor imaging (DTI), and fMRI data from patients with subacute back pain suggest that the white matter network and functional connectivity of dorsal medial prefrontal cortex, amygdala and nucleus accumbens, and bilateral total amygdala volume were independent predictors of pain persistence or recovery after one year (Vachon-Presseau et al., 2016b). These clinical studies validate the amygdala activity changes observed in preclinical studies in a number of pain models.

Synaptic plasticity

The analysis of synaptic mechanisms of pain-related amygdala activity changes in brain slice physiology studies identified an excitation-inhibition imbalance in CeLC neurons in several pain models. Increased excitatory synaptic transmission at the PB-CeLC synapse was recorded in models of acute inflammatory pain (Adedoyin, Vicini, & Neale, 2010; Miyazawa et al., 2018; Shinohara et al., 2017; Sugimura et al., 2016), arthritis pain (Neugebauer et al., 2003; Han et al., 2005; Bird et al., 2005; Fu & Neugebauer, 2008; Fu et al., 2008; Ren & Neugebauer, 2010), visceral pain (Han & Neugebauer, 2004) and muscle pain (Cheng et al., 2011), and in a chronic neuropathic pain model (SNL) (Ikeda et al., 2007; Nakao, Takahashi, Nagase, Ikeda, & Kato, 2012). Excitatory synaptic transmission from the BLA to CeLC neurons was also increased in models of arthritic pain (Neugebauer et al., 2003; Ren & Neugebauer, 2010; Ren, Palazzo, Maione, & Neugebauer, 2011; Ren et al., 2013; Fu & Neugebauer, 2008) and neuropathic pain (SNL) (Ji et al., 2017; Ikeda et al., 2007), but not in a colitis model of visceral pain (Han & Neugebauer, 2004). Pain-related processing in the LA-BLA regions, their inputs and projection targets within the amygdala, including CeLC, are less well understood, but there is evidence for increased activity of BLA neurons and enhanced excitatory transmission at the LA-BLA synapse in an arthritis pain model (Ji et al., 2010). It remains to be determined which specific cell types in the central nucleus undergo synaptic changes. The original assumption was that non-accommodating or regular spiking neurons represent peptidergic projection neurons (Schiess, Callahan, & Zheng, 1999), which would largely correspond to PKCdelta negative neurons (Haubensak et al., 2010) (see “Cell types“ in “Amygdala neurocircuitry of ‘pain processing’“). Therefore, initial work on pain-related changes focused on the presumed CRF neurons (Neugebauer et al., 2004), but electrophysiological firing pattern and neurochemical signature do not necessarily identify the same cell type. Addressing these questions is an ongoing area of amygdala pain research.

Importantly, synaptic changes that developed in pain models in vivo are preserved in the ex vivo brain slice preparation that is disconnected from the site of injury and from peripheral and spinal nociceptive processing. These changes are therefore maintained, in part at least, independently of continued nociceptive drive hence justifying their designation as “pain-related neuroplasticity”. The concept of amygdala neuroplasticity in pain is further supported by evidence for increased cell proliferation [bromodeoxyuridine (BrdU)-positive cells] in the central and basolateral nuclei in a neuropathic pain model (spared nerve injury, SNI), and these cells co-labeled for neuronal markers, but not for differentiated glial cells, perhaps suggesting the generation of new amygdala neurons in chronic pain (Goncalves et al., 2008).

Excitatory synaptic plasticity can develop because inhibitory control mechanisms fail. CeLC neurons are modulated by glutamate-driven feedforward inhibition that involves inhibitory interneurons in the intercalated mass (ITC) and/or within the central nucleus (PKCdelta cells) (Neugebauer, 2015; Thompson & Neugebauer, 2017) (see Fig. 2). Feedforward inhibition of CeLC neurons is decreased in brain slices from arthritic rats, permitting the increase of synaptically evoked spiking, a measure of neuronal output (Ren & Neugebauer, 2010; Ren et al., 2011; Ren et al., 2013). Decreased GABAergic inhibition and increased excitability and burst activity have also been found in CeLC neurons in brain slices from neuropathic rats (SNL model) (Jiang et al., 2014). An important driver of feedforward inhibition of CeLC neurons is the output from medial prefrontal cortical pyramidal cells (Kiritoshi & Neugebauer, 2018). Glutamatergic projections of these pre- and infra-limbic pyramidal cells target inhibitory interneurons such as ITC cells either directly or indirectly via BLA (Kiritoshi & Neugebauer, 2018)(see Fig. 2). Prefrontal cortical feedforward inhibition of amygdala output neurons plays an important role in fear extinction and fails in conditions of impaired fear extinction (Duvarci & Pare, 2014; Likhtik et al., 2008; Chang & Maren, 2010; Hefner et al., 2008; Kim, Jo, Kim, Kim, & Choi, 2010; Sierra-Mercado, Padilla-Coreano, & Quirk, 2011).

Impaired feedforward inhibition of CeLC neurons in pain has been linked to decreased output from infralimbic pyramidal cells (Kiritoshi et al., 2016; Kiritoshi & Neugebauer, 2018) as the consequence of medial prefrontal cortical deactivation through enhanced amygdala (BLA)-driven feedforward inhibition (Kiritoshi et al., 2016; Ji et al., 2010; Ji & Neugebauer, 2014) (see Fig. 2). Using optogenetic activation of infralimbic axon terminals, brain slice physiology studies found decreased glutamate-driven synaptic inhibition of CeLC neurons in an arthritis pain model (Kiritoshi & Neugebauer, 2018) and increased glutamate-driven synaptic inhibition of infralimbic pyramidal cells (Kiritoshi et al., 2016). BLA neurons are known to project heavily to pre- and infra-limbic cortical regions where they form direct excitatory contacts not only with pyramidal cells but also with parvalbumin and somatostatin positive inhibitory GABAergic interneurons that target mainly the somatic and proximal axonal regions to regulate synaptic integration and neuronal output (Gabbott, Warner, & Busby, 2006; McGarry & Carter, 2016).

Deactivation of the medial prefrontal cortex has been well documented in models of acute arthritis (Ji et al., 2010; Ji & Neugebauer, 2011; Ji & Neugebauer, 2014; Kiritoshi et al., 2016) and chronic neuropathic pain (SNI and CCI) models (Zhang et al., 2015; Dale et al., 2018; Radzicki, Pollema-Mays, Sanz-Clemente, & Martina, 2017; Wang et al., 2015; Metz, Yau, Centeno, Apkarian, & Martina, 2009; Kelly, Huang, Meltzer, & Martina, 2016). Whereas in the acute model enhanced glutamate-driven feedforward inhibition from the BLA plays a critical role in the decrease of infra- and pre-limbic pyramidal cell activity (Ji et al., 2010; Sun & Neugebauer, 2011; Kiritoshi et al., 2016), structural and functional changes contribute to decreased excitability of prelimbic pyramidal cells in the chronic pain model. These changes include reduced branching of apical dendrites (Metz et al., 2009; Kelly et al., 2016), impaired glutamatergic transmission (Metz et al., 2009; Kelly et al., 2016) presynaptically at the ventral hippocampal input and postsynaptically at the medial dorsal thalamic input though unchanged inhibition/excitation ratio (Kelly & Martina, 2018), loss of excitatory cholinergic modulation due to M1 receptor internalization (Radzicki et al., 2017), reduced intracortical glutamatergic signaling (Cheriyan & Sheets, 2018), and increased feed-forward inhibition mediated by parvalbumin-expressing GABAergic interneurons (Zhang et al., 2015). Functional and structural abnormalities such as altered connectivity and gray matter atrophy have also been reported in human pain patients with chronic back pain (Apkarian et al., 2004a; Mayer et al., 2005; Baliki, Geha, Apkarian, & Chialvo, 2008; Geha et al., 2008).

In summary, pain-related neuroplasticity in the amygdala output region constitutes an excitation-inhibition imbalance due to medial prefrontal cortical deactivation that leads to impaired (infralimbic) feedforward inhibition of CeLC neurons.

Behavioral consequences of amygdala activation and neuroplasticity

The significance of pain-related amygdala activation and neuroplasticity has been determined in studies linking neural activity causally to pain-like behaviors through pharmacological or optogenetic manipulations that are known to increase or decrease amygdala activity. The overall finding is that increasing amygdala activity can generate or facilitate certain pain-like behaviors under normal conditions in the absence of any pain-producing tissue pathology, whereas deactivation has inhibitory behavioral effects in models of acute and chronic pain (Neugebauer et al., 2004; Neugebauer, 2015; Veinante et al., 2013; Thompson & Neugebauer, 2017). An emerging new concept supported by clinical data (Vachon-Presseau et al., 2016b; Vachon-Presseau et al., 2016a) holds that differences in amygdala neuronal activity are associated with phenotypic differences in pain vulnerability and resilience and may be predetermining factors of the complexity and persistence of pain (Ji, Yakhnitsa, Kiritoshi, Presto, & Neugebauer, 2018).

Amygdala activation

Interventions that increase amygdala output, even in the absence of tissue injury, can generate or facilitate pain-like behaviors. Optogenetic (channelrhodopsin-2-mediated) activation of neurons in the central amygdala nucleus induced mechanical allodynia and increased visceromotor responses to urinary bladder distension (Sadler et al., 2017). Pharmacological activation of G_q/11-coupled metabotropic glutamate receptors (mGluRs) in the central nucleus with (R,S)-3,5-dihydroxyphenylglycine (DHPG) induced mGluR5-mediated peripheral mechanical hypersensitivity (Kolber et al., 2010) and increased mGluR5-mediated visceromotor responses (Crock et al., 2012). Pharmacological ERK activation in the central nucleus with phorbol 12,13-diacetate (PDA) also induced peripheral mechanical hypersensitivity (Carrasquillo & Gereau, 2007). Activation of the mGluR5-ERK system has been shown to increase activity of amygdala CeLC neurons (Li & Neugebauer, 2004b; Li, Ji, & Neugebauer, 2011). Stereotaxic administration of neuropeptides (CGRP and CRF) into the central nucleus induced mechanical hypersensitivity and increased audible and ultrasonic vocalizations evoked by innocuous and noxious stimuli; the pronociceptive effects were blocked by their respective receptor antagonists (Han, Adwanikar, Li, Ji, & Neugebauer, 2010; Ji, Fu, Adwanikar, & Neugebauer, 2013). CRF effects persisted when HPA axis function was suppressed by pretreatment with dexamethasone (subcutaneously). In contrast, CGPR injection into the BLA was antinociceptive, decreasing mechanical and thermal sensitivity (Li et al., 2008), which may be due to an action on BLA-driven feedforward inhibition of central amygdala neurons (see “Inputs“ in “Amygdala neurocircuitry of ‘pain processing’“). Finally, disinhibition of the central nucleus with a GABA_A receptor antagonist (bicuculline) produced anxiety-like behavior in normal rats (Jiang et al., 2014). A metabotropic glutamate receptor mGluR7 agonist (AMN082), shown to disinhibit CeLC neurons in brain slices (Ren et al., 2011), increased mechanical sensitivity and audible and ultrasonic vocalizations, and induced anxiety-like behavior in the elevated plus maze in normal naïve rats (Palazzo, Fu, Ji, Maione, & Neugebauer, 2008). The effects of manipulations to decrease GABAergic transmission suggest that amygdala output normally is under tonic inhibitory control.

Amygdala deactivation

Increasing the inhibitory tone in the central nucleus had antinociceptive and anxiolytic-like effects in neuropathic pain models. A GABA_A receptor agonist (muscimol) administered into the central nucleus inhibited anxiety-like behavior and CeLC neuronal activity in the SNL model (Jiang et al., 2014). In the CCI model, muscimol in the central nucleus also reversed mechanical hypersensitivity and depression-like behavior in the forced swim test and attenuated escape/avoidance behavior, while HPA axis function remained unchanged (Pedersen, Scheel-Kruger, & Blackburn-Munro, 2007; Seno et al., 2018). Antinociceptive and anti-depressive-like effects were also seen with muscimol injected into the neighboring BLA in the CCI model (Seno et al., 2018). Restoring GABAergic feedforward inhibition of CeLC neurons with neuropeptide S (NPS) acting on NPS receptors (NPSR) on ITC cells, decreased amygdala (CeLC) activity and inhibited vocalizations to noxious stimuli and anxiety-like behaviors in the elevated plus maze in arthritic rats without affecting mechanosensitivity (Ren et al., 2013; Medina, Ji, Gregoire, & Neugebauer, 2014). In these studies, NPS was injected into the ITC or applied nasally, and these effects were blocked by stereotaxic application of NPSR antagonists ([D-Cys(tBu)⁵]NPS or SHA68) into the ITC area lateral to the CeLC.

Blockade of NMDA receptors in the central nucleus with MK-801 inhibited nocifensive (hindlimb withdrawal reflex) and affective (place avoidance test) behaviors in a neuropathic pain model (SNI) (Ansah, Bourbia, Goncalves, Almeida, & Pertovaara, 2010). Antagonists for NMDA (AP5) or non-NMDA (CNQX) receptors administered into the central nucleus also inhibited emotional responses (footshock-evoked vocalization afterdischarges) under normal conditions without affecting mechanosensitivity (Spuz & Borszcz, 2012). Antagonists for G_q/11-coupled mGluR 1 (CPCCOEt) and mGluR5 (MPEP) in the central nucleus inhibited audible and ultrasonic vocalizations evoked by noxious mechanical stimuli in an arthritis pain model (Han & Neugebauer, 2005) and aversive behaviors (place avoidance) in a neuropathic pain model (SNI) (Ansah et al., 2010). In these studies in rats, mechanical hypersensitivity was also inhibited by CPCCOEt but not by MPEP. Other studies in mice found that MPEP or mGluR5 conditional knock-out in the central nucleus had inhibitory effects on mechanosensitivity in the formalin pain model (Kolber et al., 2010) and on visceromotor reflexes in the bladder distension pain model (Crock et al., 2012). Activation G_i/o-coupled mGluR8 in the central nucleus with DCGP decreased mechanical hypersensitivity, inhibited vocalizations and had anxiolytic-like effects in the elevated plus maze in an arthritis pain model, but had no effect under normal conditions (Palazzo et al., 2008). DCPG in the central nucleus also decreased thermal hypersensitivity in an inflammatory pain model (intraplantar carrageenan), and mGluR8 expression was increased in this model (Palazzo et al., 2011). Importantly, these glutamate receptor-targeting interventions also inhibited pain-related neuronal activity and/or synaptic excitation of CeLC neurons (reviewed in (Neugebauer, 2015; Thompson & Neugebauer, 2017), establishing a mechanistic link between amygdala output and pain-like behaviors.

Neuropeptide receptors in the CeA have also been targeted for beneficial behavioral effects in pain models. A CGRP1 receptor antagonist (CGRP_8–37) administered into the central nucleus decreased mechanical hypersensitivity and inhibited audible and ultrasonic vocalizations in an arthritis pain model, but had no effects in normal animals (Han et al., 2005). A CRF1 receptor antagonist (NBI27914) administered into the central nucleus or BLA also decreased mechanical hypersensitivity, inhibited audible and ultrasonic vocalizations, and had anxiolytic-like effects in the elevated plus maze in a model of arthritic pain without any effects in normal animals (Ji et al., 2007; Ji et al., 2010; Fu & Neugebauer, 2008). A CRF2 receptor antagonist (astressin-2B) had no significant behavioral effect (Fu & Neugebauer, 2008). A nonselective CRF receptor antagonist (CRF_9–41) in the central nucleus had no significant effects in a neuropathic pain model (SNI) (Bourbia, Ansah, & Pertovaara, 2010) but had antinociceptive effects on morphine withdrawal-induced thermal hyperalgesia (McNally & Akil, 2002). Again, blockade of CGPR1 or CRF1 receptors also inhibited amygdala activity and neurotransmission (reviewed in (Neugebauer, 2015; Thompson & Neugebauer, 2017).

The serotonergic system plays an important role in pain modulation but a novel concept has emerged that the 5-HT_2C receptor subtype in the amygdala mediates undesirable side effects and weak or inconsistent efficacy of selective serotonin reuptake inhibitors (SSRIs) (see Thompson & Neugebauer, 2017). A 5-HT_2C receptor antagonist (SB242084) administered into the BLA did indeed restore the ability of a systemically applied SSRI (fluvoxamine) to inhibit audible and ultrasonic vocalizations and anxiety-like pain behaviors, but not mechanical hypersensitivity, in an arthritis pain model (Gregoire & Neugebauer, 2013). Viral vector mediated 5-HT_2C receptor knockdown in the BLA inhibited mechanical hypersensitivity and audible and ultrasonic vocalizations, and decreased anxiety-like and depression-like behaviors in a neuropathic pain model (SNL) (Ji et al., 2017). 5-HT_2C receptor blockade or knockdown in the central nucleus had no effect in these studies. 5-HT_2C receptor knockdown in the BLA blocked the increase in neuronal activity of CeLC neurons in the SNL model through a mechanism that involved decreasing BLA-driven synaptic excitation (Ji et al., 2017). These studies indicate that 5-HT_2C receptors in the amygdala can facilitate affective components of pain processing.

Small-conductance calcium-activated potassium (SK) channels play an important role in the regulation of neuronal excitability neurons in the central nucleus (see Thompson & Neugebauer, 2017). Activation of SK channels in the central nucleus, but not in BLA, by stereotaxic administration of riluzole inhibited audible and ultrasonic vocalizations (emotional responses) in models of arthritis and neuropathic (SNL) pain without affecting mechanical hypersensitivity (Thompson, Yakhnitsa, Ji, & Neugebauer, 2018; Thompson, Ji, & Neugebauer, 2015). Riluzole also decreased neuropathic pain-induced depression-like behavior in the forced swim test (Thompson et al., 2018). The effects of riluzole were eliminated by an SK channel blocker (apamin) but not by a blocker of large-conductance calcium-activated potassium BK channels (charybdotoxin), implicating SK channels in the riluzole effects (Thompson et al., 2018; Thompson et al., 2015). Underlying neuronal mechanisms included the increase of SK channel-mediated afterhyperpolarization and synaptic inhibition of CeLC neurons by riluzole (Thompson et al., 2018).

Inhibition of signal transduction pathways such as ERK or PKA in the central nucleus also inhibited pain-like behaviors. Stereotaxic injection of a MEK inhibitor (U0126) into the central nucleus decreased peripheral mechanical hypersensitivity in the formalin pain model but had no effect under normal conditions (Carrasquillo & Gereau, 2007). U0126 or an inhibitor of PKA (KT5720), but not PKC (GF109203x), administered into the central nucleus inhibited mechanical hypersensitivity and audible and ultrasonic vocalizations in an arthritis pain model, but had no effects in normal animals (Fu et al., 2008). Inhibition of ERK and PKA, but not PKC, also decreased synaptic excitation of CeLC neurons in brain slices from arthritis rats (Fu et al., 2008), linking amygdala function to pain-like behaviors.

Finally, increasing medial prefrontal cortical activity and output to inhibit neuronal activity of CeLC neurons in the amygdala also inhibited pain-like behaviors. A combination of agonists for mGluR5 (VU0360172) and cannabinoid receptor CB1 (ACEA) increased infralimbic pyramidal cell output in an arthritis pain model (Ji & Neugebauer, 2014) by increasing their excitatory drive and removing abnormal synaptic inhibition (Kiritoshi et al., 2016). This pharmacological strategy decreased the enhanced CeLC neuronal hyperactivity in the arthritis pain condition (Ji & Neugebauer, 2014) and inhibited mechanical hypersensitivity and audible and ultrasonic vocalizations, and mitigated cognitive decision-making deficits in a rodent gambling task in the arthritis pain model(Kiritoshi et al., 2016). These results provide a causal link between cortical control of amygdala output and pain-like behaviors.

Conclusions

Convergence of direct nociceptive input from the lateral parabrachial area in the brainstem and multimodal thalamo-cortical inputs through the lateral-basolateral amygdala network positions the central nucleus with its known projection neurons well to participate in the processing of pain-related information and to contribute to pain behaviors and their modulation. Amygdala activity changes have been documented in various preclinical pain models (with an emphasis on neurons in the lateral and capsular divisions of the central nucleus) and in clinical pain conditions. Pain-related increase of amygdala output is generated by neuroplastic changes manifested as an excitation-inhibition imbalance largely due to loss of cortically driven control (feedforward inhibition) of amygdala neurons. Optogenetic and pharmacological manipulations have linked amygdala activity changes causally to pain-like behaviors. A unifying view is that amygdala (central nucleus) activation can generate or facilitate pain-like behaviors under normal conditions in the absence of any tissue pathology, whereas deactivation inhibits certain behaviors in acute and chronic pain models (Neugebauer et al., 2004; Neugebauer, 2015; Veinante et al., 2013; Thompson & Neugebauer, 2017). Recent evidence from preclinical and clinical studies further suggests that differences in amygdala activity and connectivity may predict vulnerability or resilience to pain (Ji et al., 2018; Vachon-Presseau et al., 2016b; Vachon-Presseau et al., 2016a). A current focus of research on amygdala pain mechanisms is on cell type-, synapse-, input- and projection-specific functions related to pain. Another area of investigation is the hemispheric lateralization of pain-related amygdala function as there is accumulating evidence to suggest that the right amygdala undergoes activity changes and generates pain behaviors whereas the function of the left amygdala remains unclear (Carrasquillo & Gereau, 2008; Kolber et al., 2010; Ji & Neugebauer, 2009; Sadler et al., 2017). These future studies will help elucidate the circuits and factors that determine the bidirectional relationship between the amygdala and pain processing. Shifting amygdala function from driving pain aspects to engaging the endogenous pain control system through cell- and synapse-specific targeting may have therapeutic implications for pain management.