Multiple Inhibitory Mechanisms in the Basolateral
Amygdala Facilitate Fear Extinction
September 10, 2018
Literature thesis (16,525 words) Peter Saalbrink, 6102794
MSc in Brain and Cognitive Sciences, track: Behavioural Neuroscience University of Amsterdam
Multiple Inhibitory Mechanisms in the Basolateral Amygdala Facilitate Fear Extinction September 10, 2018
Literature thesis (16,525 words) Peter Saalbrink, 6102794
MSc in Brain and Cognitive Sciences, track: Behavioural Neuroscience University of Amsterdam
Supervisor: dr. Harm Krugers; Co-assessor: dr. Helmut Kessels
Abstract
The extinction of conditioned fear is facilitated by complex mechanisms, with a central role for the basolateral complex of the amygdala (BLA). Impairments in these mechanisms may lead to anxiety disorders. Here, I connect current knowledge on inhibitory amygdala
microcircuits and endocannabinoid (eCB) signaling in the context of fear extinction, to navigate the search for extinction-enhancing targets and novel treatments for anxiety disorders. First, I review literature showing that fear extinction involves depotentiation of conditioned CS-US associations as well as learning new CS-“no US” associations. Furthermore, I discuss studies showing that glutamatergic and GABAergic transmission enable fear extinction through dissociated pathways. Additionally, I explore how fear extinction can be modulated, for example by stress. Finally, I discuss evidence of eCB-mediated changes in synaptic plasticity in the BLA underlying the inhibitory microcircuits that facilitate fear extinction.
Keywords: basolateral amygdala (BLA); fear extinction; endocannabinoid (eCB) system; inhibitory microcircuits; synaptic plasticity.
Chapter 1. Introduction ...7
§1.1. Fear Memory ... 7
¶1.1.1. Aim of This Thesis ... 9
¶1.1.2. Studying Fear in Animals ... 9
§1.2. Fear Conditioning ... 10
¶1.2.1. Fear Conditioning as a Model of Fear Memory ... 10
¶1.2.2. Auditory Fear Conditioning ... 10
§1.3. Fear Extinction ... 12
¶1.3.1. Renewal, Reinstatement, and Spontaneous Recovery of Fear ... 12
¶1.3.2. Fear Extinction Is Context-Dependent ... 13
¶1.3.3. Is Fear Extinction Suppression of Fear, not Forgetting? ... 13
¶1.3.4. Can Fear Extinction Be Both Suppression and Forgetting of Fear? ... 14
Chapter 2. Mechanisms of Fear Extinction...14
§2.1. Brain Areas of Fear Conditioning and Extinction ... 14
¶2.1.1. Amygdala ... 15
¶2.1.2. mPFC ... 17
¶2.1.3. Hippocampus ... 18
¶2.1.4. Network of Fear Conditioning and Extinction ... 18
§2.2. Plasticity in Fear Conditioning ... 20
¶2.2.1. Long-Term Potentiation of Lateral Amygdala Neurons ... 20
¶2.2.2. LTP Is NMDA-Dependent ... 21
¶2.2.3. LTP-Induced Cellular Mechanisms ... 22
¶2.2.4. LTP-Induced Changes in Synaptic Structure ... 23
¶2.2.5. Plasticity Occurs in Multiple Pathways ... 23
¶2.2.6. LTP = Learning? ... 24
§2.3. Plasticity in Fear Extinction ... 24
¶2.3.2. Long-Term Potentiation of Basolateral Amygdala Neurons ... 26
¶2.3.3. Long-Term Potentiation of GABAergic Interneurons ... 27
¶2.3.4. Long-Term Potentiation of Inhibitory Synaptic Transmission ... 27
¶2.3.5. Depotentiation of Conditioning-Induced Plasticity ... 28
¶2.3.6. Extinction-Induced Cellular Mechanisms ... 29
¶2.3.7. Extinction-Induced Changes in Synaptic Structure ... 30
¶2.3.8. Plasticity Occurs in Multiple Pathways ... 31
¶2.3.9. Extinction = New Learning? ... 32
§2.4. Microcircuits in the Amygdala in Fear Conditioning ... 33
¶2.4.1. The Fear Engram ... 33
¶2.4.2. Circuits of Fear Conditioning ... 33
§2.5. Microcircuits in the Amygdala in Fear Extinction ... 34
¶2.5.1. The Extinction Engram ... 35
¶2.5.2. Circuits of Fear Extinction ... 35
¶2.5.3. Fear Extinction Involves a Network Shift ... 36
¶2.5.4. Inhibitory Activity in BLA Microcircuits ... 36
¶2.5.5. Inhibitory Activity in the mITC of the Amygdala ... 38
¶2.5.6. Interneurons in the BLA ... 40
¶2.5.7. Mechanisms of Fear Extinction ... 44
§2.6. Top-Down Control in the mPFC-BLA Pathway ... 45
Chapter 3. Modulating Fear Extinction ...47
§3.1. Stress-Related Impairments in Fear Conditioning... 47
§3.2. Stress-Related Impairments in Fear Extinction ... 48
¶3.2.1. Enhancing Extinction Through Stress ... 50
¶3.2.2. Mechanisms of Stress-Modulated Extinction ... 50
Chapter 4. Endocannabinoids and Fear Extinction ...51
¶4.1.1. The Endocannabinoid System ... 51
¶4.1.2. Endocannabinoid Functioning ... 51
¶4.1.3. Endocannabinoid Signaling ... 52
§4.2. The Role of Endocannabinoids in Fear Conditioning and Extinction ... 53
¶4.2.1. Endocannabinoids in the Fear System ... 53
¶4.2.2. Endocannabinoid Signaling Regulates Inhibition and Disinhibition ... 54
¶4.2.3. A Cellular Mechanism of Endocannabinoid Signaling ... 55
¶4.2.4. Different Effects of 2-AG and AEA in Fear Extinction ... 56
¶4.2.5. eCB Signaling Mediates Fear Extinction Through Metaplasticity... 56
¶4.2.6. Endocannabinoids are Involved in Extinction-Induced Disinhibition ... 57
§4.3. Modulation of Endocannabinoid Signaling ... 57
¶4.3.1. The Endocannabinoid System as a Target for Anxiety Disorders ... 58
Chapter 5. Discussion ...59
§5.1. Summary ... 59
§5.2. A Note on Balances and Spatiotemporal Specificity in Fear Extinction ... 59
§5.3. Immediate Extinction Is Dissociated From Delayed Extinction ... 60
§5.4. Fear Research in Humans ... 61
§5.5. Concluding Remarks ... 61
Chapter 1. Introduction
§1.1. Fear Memory
In daily life, individuals constantly have to deal with stressful situations which require risk assessment and adaptations in defensive behavior. The emotion that across species lies central to such abilities is fear (LeDoux, 2012). Fear underlies fear-motivated learning (or fear learning), a type of associative learning in which initially neutral stimuli become threatening through pairing with fearful stimuli, leading to a fear response (Maren, 2001). Behavioral responses to fear include fight, flight, or avoidance (i.e., fright or freeze)
reactions. Memory of fearful experiences serves dynamic adaptation to the environment and is essential to coping with future threats and thus to survival (Izquierdo et al., 2016). Fear memory, sometimes also called emotional or aversive memory, is very robust, because its consolidation is enhanced by emotional arousal (Roozendaal, 2000). Not surprisingly, mammals have specialized and phylogenetically conserved networks of brain areas that facilitate fear memory and related behavior (Fanselow & Poulos, 2005; Milad & Quirk, 2012). Therefore, fear learning is often used to study fear memory, behavior, and its neural substrates.
Fear memory is a form of associative memory, meaning aversive events are stored together with other stimuli from the environment (Maren & Quirk, 2004; Pape & Paré, 2010). Fear learning is, as other forms of learning, based on neuroplasticity (Maren, 2001).
Furthermore, fear memory is subject to many complex interactions, such as genetic variation and past experiences in individuals. Because of this complexity, perturbations of the fear system sometimes result in maladaptation. Under some circumstances, dysfunction of the fear system may even lead to psychopathology. Anxiety disorders, such as posttraumatic stress disorder (PTSD), are a form of generalized fear, and can result from impairments in the fear system (Mahan & Ressler, 2012; Milad et al., 2009; Shin et al., 2009). Anxiety disorders
have high prevalence amongst the general population. For example, lifetime prevalence of PTSD was 7.4% in the Netherlands in the 2000s, with higher prevalence amongst females (8.8%) than males (4.3%; De Vries & Olff, 2009).
The complexity of the fear system and the resilience of fear memory also results in difficulty of finding effective treatment methods for anxiety disorders (Fitzgerald et al., 2014; Rabinak & Phan, 2014; Singewald et al., 2015). For example, fear memory can last for long periods of time, and can generalize from one stimulus, context or environment to another (Likhtik & Paz, 2015). Meanwhile, fear reduction and fear relief are often fragile, transient and context-dependent (Riebe et al., 2012). Furthermore, the several brain networks and its related systems of neurotransmitters have not been fully identified (Krabbe et al., 2018), which makes it difficult to target anxiety disorders pharmacologically.
An often-used form of behavioral therapy, for example for PTSD, is exposure theory, which is based on the extinction of fear memory (Rothbaum & Davis, 2003). In fear
extinction, the environmental and contextual stimuli of a fear memory are repeated in absence of the fearful, aversive stimulus itself. This leads to a dismantling of the fear association, and eventually the dissolution of the fear memory (Myers & Davis, 2007). However, this is a complex process. Fear extinction is, like fear memory, controlled by specific brain mechanisms (Barad et al., 2006; Milad & Quirk, 2002). Extinction memory is fragile and context-dependent, and is subject to recovery, reinstatement and renewal of fear (Bouton, 2004). Therefore, fear extinction is not always successful. However, mechanisms are being discovered through which fear extinction can be enhanced. Although great progress has been made in the field of fear and extinction memory research, many neurobiological processes underlying fear and extinction memory are not yet fully understood.
¶1.1.1. Aim of This Thesis
This thesis addresses the concepts that currently exist in fear extinction research. First, an overview of fear memory and commonly used research methods is presented. Next,
current knowledge on the mechanisms that establish fear extinction is discussed. Specifically, the association between synaptic plasticity and activity of microcircuits in the basolateral complex of the amygdala (BLA) in the context of fear extinction will be described in detail in this thesis. Furthermore, I will discuss how fear extinction can be enhanced or impaired. For example, the impairing role of stress, as well as the enhancing role of endocannabinoid (eCB) signaling in fear extinction will be explored. Finally, I will summarize current challenges in fear extinction research. Ultimately, the goal of this thesis is to navigate the search for extinction-enhancing pharmaceutical targets and therapeutic treatments for patients with anxiety disorders (Fitzgerald et al., 2014; Singewald et al., 2015). This will be achieved through an in-depth review of studies in animals as well as in humans, combining studies from molecular and neurobiological disciplines, and an analysis of networks underlying fear behavior.
¶1.1.2. Studying Fear in Animals
Behavioral responses to fear have been conserved throughout evolution (LeDoux, 2012; Wotjak & Pape, 2013). Therefore, fear memory and its extinction are often used as a translational assay for studying treatments for anxiety disorders (Graham & Milad, 2011). For example, it has recently been shown by Gunduz-Cinar et al. (2016) using an extinction-impaired mouse model that the selective serotonin reuptake inhibitor (SSRI) fluoxetine can facilitate fear extinction. In that research study, extinction-impaired mice that have been subjected to fear conditioning (see next section) showed less freezing behavior after extinction learning when combined with administration of this drug (Gunduz-Cinar et al., 2016). Identifying therapeutic targets for fear extinction processes using laboratory research
helps in the understanding and treatment of anxiety disorders. Therefore, I will focus on animal research for a great part of this thesis.
§1.2. Fear Conditioning
¶1.2.1. Fear Conditioning as a Model of Fear Memory
Fear memory is often studied in a laboratory setting using a fear conditioning paradigm (Maren, 2001; Pavlov, 1927). Fear conditioning is a simple yet robust form of classical or Pavlovian conditioning, and involves an associative learning process. Fear conditioning involves several stages of fear memory, amongst which are the formation or acquisition, encoding and long-term stabilization or consolidation, recall or retrieval of the fear memory, extinction of the fear memory, and reconsolidation (Figure 1). As such, it is extensively studied in both humans and rodents, and thus provides us with an excellent research method to study fear learning and memory. Recently, much progress has been made with new research methods in rodents, such as in vivo monitoring of cell activity (e.g., calcium imaging and tagging of immediate early genes) and selectively activating restricted neuronal populations (e.g., optogenetics [Johansen et al., 2012] and chemogenetics [Gafford & Ressler, 2016], which involve the synthetization and delivery of receptor proteins which allow for specific neurons to be individually activated or inhibited). Thus, new tools are being developed that allow us to identify with great detail the mechanisms underlying fear memory and behavior.
¶1.2.2. Auditory Fear Conditioning
Auditory fear conditioning paradigms (also called classical or cued fear conditioning) are based on learning a stimulus-stimulus association (Pavlov, 1927). The association to be learned is the pairing of a neutral and innocuous tone or a cue, the conditioned stimulus (CS), with a noxious and aversive stimulus, such as a weak electrical footshock; the unconditioned stimulus (US). When laboratory animals, often rodents, are subjected to this experimental
design, such a footshock leads to freezing behavior, called the conditioned fear response (CR). Freezing is usually described as total immobility except for breathing-related movements. This can be easily detected and quantified, for example relative to the total amount of time spent in the experimental environment.
In auditory fear conditioning, the CS that is used for pairing often is a tone, and for every stage of the conditioning paradigm the context that is used for testing will change. For example, initial learning will take place in context A, extinction learning in context B, and subsequent testing in another context, C. In a paradigm called contextual fear conditioning, by contrast, the environment is the CS, and animals will exhibit the CR when reintroduced to the experimental context after initial fear learning (Pavlov, 1927). In this thesis, the focus will be mainly on auditory fear conditioning. However, sometimes I will explicitly mention contextual fear conditioning as well.
Fear conditioning occurs rapidly and robustly: a single pairing session is sufficient to produce a lasting fear association, which appears from re-exposure to subsequent CS
presentations resulting in fear retrieval or expression and observations of the CR. Moreover,
Figure 1. Behavioral responses during acquisition and extinction of fear memory. During fear conditioning, pairing of a US with a CS results in fast acquisition of fear memory and the expression of a fear CR (left). Subsequent CS presentations during extinction learning result in a decrease of the fear CR.
repeated CS-US pairings can generate a long-term fear association that leads to expression of the CR over an extended period of time (Pavlov, 1927).
§1.3. Fear Extinction
Fear extinction constitutes normal adaptation to experiences of anxiety and fear. If an environmental stimulus that previously signaled danger is repeated without the associated dangerous stimulus, the fear association is extinguished. That is, fear behavior is gradually decreased upon repeated or prolonged exposure to non-paired CS presentations (Orsini & Maren, 2012). Fear extinction is often used as a measure of learned inhibition of conditioned fear. For example, mice that have been fear conditioned show less freezing behavior after extinction learning (Figure 1; Milad & Quirk, 2012).
Thus, fear extinction is an active form of associative learning in which the expression of a conditioned fear response is reduced in the absence of an aversive stimulus (Furini et al., 2014; Pavlov, 1927; Quirk & Mueller, 2008). Just as fear conditioning, fear extinction can be subdivided in acquisition, consolidation and retrieval stages (Milad & Quirk, 2012; Quirk & Mueller, 2008). As mentioned above, impairments in fear extinction may lead to the
development of anxiety disorders such as posttraumatic stress disorder (Myers & Davis, 2007).
¶1.3.1. Renewal, Reinstatement, and Spontaneous Recovery of Fear
Fear is often not fully extinguished, and the initial fear memory can be persistent. Indeed, extinguished fear can return after spontaneous recovery, renewal, or reinstatement. Specifically, spontaneous recovery of the CR could occur with the passing of time, exposure to the CS in a new context (different from that of fear acquisition or extinction learning) leads to fear renewal, and renewed presentation of the US might result in reinstatement of the freezing behavior (Pavlov, 1927).
¶1.3.2. Fear Extinction Is Context-Dependent
The finding that fear renewal occurs when the CS is presented in a novel context after extinction learning, indicates that fear extinction is context-dependent (Bouton, 2004). It is thought that context modulates CS-US and CS-“no US” associations in such a manner that the CS predicts either the US or absence of the US (Orsini & Maren, 2012).
¶1.3.3. Is Fear Extinction Suppression of Fear, not Forgetting?
Findings that fear can return after extinction learning suggest that expression of the conditioned fear response is inhibited by extinction learning, rather than erased (Duvarci & Paré, 2014; Maren, 2015; Wotjak & Pape, 2013). From these behavioral observations, it is suggested that fear extinction learning entails the formation of a new, inhibitory memory trace (Furini et al., 2014; Pavlov, 1927). This inhibitory extinction memory trace then competes with the fear memory trace.
Because the original fear memory trace remains intact, extinction is a less durable and less permanent form of fear memory than a conditioned CS-US association (Barad et al., 2006; Maren & Holmes, 2015; Myers & Davis, 2007; Quirk & Mueller, 2008; Wotjak & Pape, 2013). Fear extinction is rather described as a form of relearning that leads to fear relief (Riebe et al., 2012) instead of actual erasure of a fear memory. It is therefore important to dissociate fear extinction from the forgetting of a fear memory trace (Myers & Davis, 2007).
Thus, extinction is not the same as forgetting, but reflects learning of a CS-“no US” association that competes with the original CS-US fear memory (Myers & Davis, 2007; Orsini & Maren, 2012). Additionally, it has been found that erasure vs. suppression by extinction is dependent on the time interval between fear acquisition and extinction learning, where conditioning is immediately followed by extinction learning sometimes impairs and sometimes enhances extinction memory (Maren, 2014; Myers et al., 2006).
¶1.3.4. Can Fear Extinction Be Both Suppression and Forgetting of Fear?
However, this erasure vs. suppression debate, sometimes also referred to as unlearning vs. new learning, could simply point to different stages and aspects of fear extinction. For example, specific CS-US associations could be inhibited by extinction learning, whilst others are erased (Herry et al., 2010), as supported by findings of different functional subpopulations of neurons within the basal nuclei of the amygdala (BA; Herry et al., 2008). Furthermore, it has been suggested that fear and extinction are two different emotional states triggered by a single CS representation based on different molecular signaling pathways, and thus does not necessarily involve inhibition of memory traces but rather the activation of a different pathway (Tronson et al., 2012; Krabbe et al., 2018). This is consistent with the idea that the initial fear memory trace remains intact. Indeed, it should be noted that erasure of fear memory traces following immediate extinction (Mao et al., 2006; Kim et al., 2007) reflects a depotentiation mechanism that is dissociable from other extinction mechanisms (see chapter 5). Lastly, it has been suggested that extinction learning leads to habituation of the fear response (Kamprath et al., 2006), although habituation does not account for the context-dependence of fear extinction (Orsini & Maren, 2012). All proven and putative mechanisms underlying this important aspect of fear extinction will be discussed in more detail in the next chapter of this thesis.
Chapter 2. Mechanisms of Fear Extinction
§2.1. Brain Areas of Fear Conditioning and Extinction
Several key neural loci for the acquisition, consolidation, and retrieval of fear memory and its extinction have been discovered (Bruchey et al., 2007; Izquierdo et al., 2016; Maren, 2005), amongst which the amygdala (Quirk & Mueller, 2008; Wotjak & Pape, 2013), the medial prefrontal cortex (mPFC; Santini et al., 2004), and the hippocampus (Sierra-Mercado et al., 2011).
Multiple research studies have found that the amygdala is most crucial to fear
memory; i.e., without the amygdala, fear learning could not occur (Barad et al., 2006; Bukalo et al., 2015; LeDoux, 2000; Tovote et al., 2015; Wotjak & Pape, 2013). However, the mPFC and hippocampus also have essential roles, at least in some stages of fear conditioning (Sierra-Mercado et al., 2011). For example, storage of fear memory is thought to be facilitated by the amygdala in parallel to the hippocampus (Sierra-Mercado et al., 2011). Therefore, I will discuss the roles of these brain areas in more detail in this section. First, the networks of fear memory and fear extinction will be discussed, followed by synaptic
plasticity in these networks. I will then go into more detail on the microcircuits of fear, and, finally, on the top-down control by the mPFC.
¶2.1.1. Amygdala
Fear memory, including the expression and extinction of fear, relies on the amygdala. The amygdala consists of two different functional subdivisions: the central nucleus of the amygdala (CeA) and the basolateral complex of the amygdala (BLA). Furthermore, between the CeA and the BLA lie several clusters of GABAergic inhibitory neurons called the intercalated cell masses (ITC). The ITC are innervated by the BLA and the IL, and have projections onto CeA neurons.
The CeA further consists of the lateral central amygdala (CeL), the medial central amygdala (CeM), and the capsular region of the central amygdala (CeLc). The CeA receives input from the BLA, the ITC, and the mPFC. The CeA mainly contains local inhibitory microcircuits of GABAergic inhibitory medium-spiny interneurons, and can be viewed as the output nucleus of the amygdala, with the CeM mediating passive components of the fear response and the CeL promoting active fear responses (Wotjak & Pape, 2013).
The cortex-like BLA can further be subdivided into the lateral nucleus of the amygdala (LA), which is viewed as an input region, and the basal nuclei of the amygdala
(BA), which are viewed as integration and output regions and in turn consist of the
basolateral nucleus of the amygdala (BL) and the basomedial nucleus of the amygdala (BM; Amano et al., 2011). However, many research studies only distinguish between the LA and the BA. The LA receives input mainly from sensory areas, such as the auditory cortex and the thalamus, and is the main site for CS-US convergence; while the BA receives input from associational areas, such as the mPFC and the ventral hippocampus (vHPC; Cho et al., 2013; LeDoux, 2007; Sierra-Mercado et al., 2011; Strobel et al., 2015). The BLA mainly contains two types of neurons: pyramidal cell-like principal neurons, and several types of GABAergic inhibitory interneurons (Capogna, 2014). Both types of BLA neurons are part of local
microcircuits within the BLA, but only principal neurons project to regions outside the BLA. The LA is an essential site for plasticity after fear conditioning (LeDoux, 2002). Specifically, plasticity occurs after long-term potentiation of thalamic and cortical input in the LA
(LeDoux, 2002). Furthermore, convergence of the CS and the US occurs at LA neurons (LeDoux, 2002). LA neurons project to the ITC and the BA. Principal neurons in the BA can further be functionally subdivided based on their projection target. For example, BA principal neurons that project to the prelimbic subdivision of the mPFC (PL) show enhanced activity to CSs after fear conditioning and are thought to facilitate fear expression (i.e., “fear neurons”), and can be functionally dissociated from BA principal neurons with projections to the infralimbic subdivision of the mPFC (IL), which facilitate fear extinction (i.e., “extinction neurons”; Herry et al., 2008). Additionally, the LA contains neurons that are extinction-resistant (Herry et al., 2008; Repa et al., 2001). Furthermore, this finding implies the BLA in both rapid context-dependent switching for multiple conditioning-related states as well as long-term fear memory consolidation (Herry et al., 2008; Orsini & Maren, 2012). The
different types of BLA principal neurons and interneurons will be discussed at the end of this chapter.
¶2.1.2. mPFC
The mPFC can be subdivided in the functionally different PL and IL regions (Santini et al., 2004). Both regions receive input from and have projections to the hippocampus and the BLA. However, the PL and IL work in opposing manners, promoting fear preservation and suppression, respectively.
The PL (dorsal anterior cingulate cortex [dACC] in humans) facilitates acquisition, consolidation, and retrieval of fear, but not of fear extinction (Knapska & Maren, 2009; Orsini & Maren, 2012). The PL targets and activates BLA neurons (Likhtik & Paz, 2015). In turn, the BLA gates fear in PL neurons after conditioning, while the ventral hippocampus gates fear in PL neurons after extinction (Sotres-Bayon et al., 2012). Interestingly, input of PL to BA neurons requires converge of vHPC contextual input in fear renewal after
extinction to overcome extinction-induced inhibition in the BLA (Orsini et al., 2011; Orsini & Maren, 2012).
The IL (ventromedial prefrontal cortex [vmPFC] in humans) facilitates fear extinction through enhancing connectivity with the BLA (Knapska & Maren, 2009; Orsini & Maren, 2012; Santini et al., 2008). Specifically, the IL selectively responds to extinguished CSs, and BA-projecting IL neurons are recruited during extinction retrieval (Orsini et al., 2011). The IL is thought to integrate input from BLA and vHPC to achieve this (Izquierdo et al., 2016; Milad & Quirk, 2012). Furthermore, the IL targets the inhibitory CeL and ITC, which then inhibit CeM output neurons, reducing expression of the fear response (Milad & Quirk, 2012; Wotjak & Pape, 2013). Indeed, activity of IL neurons is necessary for the formation and storage of extinction memory; however, not for its retrieval. For example, optogenetic
research has shown that activating IL neurons during extinction learning enhanced extinction memory, as witnessed by reduced fear expression; while inhibiting IL neurons during
through IL projections to the BLA. It has been shown that the IL-BLA pathway is essential in regulating BLA excitability and plasticity (Milad & Quirk, 2012; Quirk & Mueller, 2008). Thus, the IL is thought to gate between conditioning and its extinction in fear learning.
¶2.1.3. Hippocampus
The hippocampus modulates fear responses through projections to the mPFC and the BLA (Milad & Quirk, 2012). Specifically, whereas the BLA represents CSs that are
ambiguous with respect to their associative meaning, the hippocampus is involved in disambiguating the meaning of the CS using contextual cues, allowing for generalization across temporal differences in CSs to determine fear behavior (Orsini & Maren, 2012). Furthermore, the hippocampus and the BLA show synchronized activity during retrieval and expression of conditioned fear (Seidenbecher et al., 2003). The ventral CA1 region of the hippocampus (vHPC) is essential in contextual fear conditioning and in the contextual modulation of extinction of auditory fear conditioning (Milad & Quirk, 2012), and transmits discrete emotional information to the BLA during fear learning (Orsini & Maren, 2012). Additionally, the dorsal CA1 region of the hippocampus (dHPC) transmits contextual representations of the conditioning environment to the BLA during fear learning (Orsini & Maren, 2012). vHPC contextual information then converges on excitatory or inhibitory fear associations within the BLA, resulting in expression or extinction of fear, respectively. Thus, the main role of the hippocampus in fear extinction is context-dependent modulation, for example during extinction recall (Goshen et al., 2011; Milad & Quirk, 2012; Orsini & Maren, 2012).
¶2.1.4. Network of Fear Conditioning and Extinction
Above, I have discussed the brain network of fear memory, consisting of the
amygdala, the mPFC, and the hippocampus (Figure 2). In this network, the LA encodes CS-US and CS-“no CS-US” associations required for fear expression and extinction. LA “fear
neurons” and “extinction neurons” receive input from the PL and IL, respectively, and can be modulated by contextual information from the vHPC (Herry et al., 2008; Sierra-Mercado et al., 2011). Fear behavior is determined through output of the CeA upon CS activation in the LA, directly via BA and ITC activity and indirectly via the IL, and is modulated by
feedforward inhibition (Royer & Paré, 2002). Specifically, “fear neurons” excite CeM and CeL neurons that enhance passive and active components of the fear response, respectively. In the CeL, plasticity-dependent ON and OFF neurons signal the CS (Ciocchi et al., 2010;
Figure 2. The brain network of fear extinction. Upon extinction acquisition, CS afferents innervate the LA and the IL and signal safety. The IL and the BA receive contextual input from the vHPC. The BA inhibits CR output by the CeM via the CeL and via the mITC. Lines with arrowheads indicate excitatory connections, lines with circles indicate inhibitory connections. BA, basal nuclei of the amygdala; CeL, lateral central amygdala; CeM, medial central amygdala; CS, conditioned stimulus afferents; IL, infralimbic subdivision of the medial prefrontal cortex; LA, lateral nucleus of the amygdala; mITC, medial intercalated cell masses; PL, prelimbic subdivision of the medial prefrontal cortex; vHPC, ventral hippocampus.
Haubensak et al., 2010). In other words, fear conditioning entails that an aversive stimulus causes BLA and subsequent CeA activation that triggers a fear response; and when this stimulus coincides with a neutral stimulus, synaptic transmission is potentiated in such a manner that the activation pattern becomes similar to that of the aversive stimulus: the fear response now becomes triggered upon the neutral stimulus, and the fundamentals for expression of the fear memory have been established (Wotjak & Pape, 2013). The synaptic mechanisms that facilitate this will be discussed in the next section.
§2.2. Plasticity in Fear Conditioning
An important mechanism that underlies learning processes is synaptic plasticity. The best studied form of associative learning-related synaptic plasticity is long-term potentiation (LTP). In the case of fear learning, acquisition of conditioned responses arises from LTP of synapses in the LA (Fanselow & Poulos, 2005; Paré et al., 2004; Rogan et al., 1997; Rumpel et al., 2005; Sah et al., 2008). Specifically, convergence of CS and US activation patterns occurs within individual neurons in the LA and is established by associative plasticity that increases activity of LA neurons upon CS activation (LeDoux, 2000).
Importantly, plasticity in the BLA is mainly involved in acquisition of fear
conditioning. After fear learning, plasticity-dependent fear memory is consolidated, involving plasticity in the PL (Santini et al., 2004). Then, fear retrieval involves concerted activation by the PL and BLA (Herry et al., 2010). This consolidation of fear memory by distributing it across the fear network allows for strengthening of the memory trace (Orsini & Maren, 2012).
¶2.2.1. Long-Term Potentiation of Lateral Amygdala Neurons
Fear conditioning results in synaptic changes in LA neurons that indicate the occurrence of LTP (Kim & Cho, 2017; LeDoux, 2000; Rogan et al., 1997). Specifically, pairing of the CS and the US results in connections that are strengthened by LTP in both the
CS and US pathway (Orsini & Maren, 2012). Thus, LTP in the LA facilitates fear conditioning (Maren, 2001; Maren & Quirk, 2004).
LTP is induced when synaptic neurons share coincident activity, and only when a depolarization threshold is reached by the neuron. LTP is triggered by presynaptic release of the neurotransmitter glutamate. Signaling of glutamate is dependent on N-methyl-d-aspartic acid (NMDA) and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)
receptors, and of these, NMDA receptors mediate the depolarization threshold of LTP (Maren, 2005; Orsini & Maren, 2012; Walker et al., 2002).
¶2.2.2. LTP Is NMDA-Dependent
When presynaptic glutamate is first released, it binds to postsynaptic AMPA receptors, which depolarizes the neuron. After the depolarization threshold is reached, glutamate can also bind to postsynaptic NMDA receptors. NMDA receptor activation results in Ca2+ influx, in turn triggering other cellular mechanisms. For example, the AMPA/NMDA ratio is increased through AMPA receptor trafficking and insertion in the dendritic spines of the postsynaptic neuron, further enhancing LTP (Kessels & Malinow, 2009; LeDoux, 2000; Maren, 2015; Orsini & Maren, 2012; Rumpel et al., 2005). This mechanism supports robust changes in synaptic strength required for the encoding of fear memory.
Thus, when presynaptic activity coincides with sufficient postsynaptic depolarization, postsynaptic NMDA receptors are activated by extracellular glutamate and allow Ca2+ influx, inducing a long-term increase in synaptic efficacy (Pape & Paré, 2010). In this mechanism, NMDA receptors are coincidence detectors that transform correlated neuronal activity into changes in synaptic strength (Pape & Paré, 2010). Thus, conditioning-induced LTP is NMDA receptor-dependent.
However, conditioning-induced signaling relies on both NMDA and AMPA receptors, and additionally, on voltage gated calcium channels (VGCCs) in large spines on
BLA dendrites, as well as on a decrease in inhibitory synaptic transmission through
GABAergic interneurons in the amygdala (Chauveau et al., 2012; Furini et al., 2014; Orsini & Maren, 2012; Pape & Paré, 2010; Roozendaal et al., 2009; Suvrathan et al., 2014;
Zimmerman & Maren, 2010). Indeed, after fear conditioning, plasticity at excitatory synapses on inhibitory interneurons is impaired (Szinyei et al., 2007). The significance of this finding will be discussed in the next section. First, LTP induced cellular mechanisms will be
discussed.
¶2.2.3. LTP-Induced Cellular Mechanisms
LTP induces both postsynaptic modifications, for example through cellular protein signaling pathways in addition to the aforementioned AMPA receptor trafficking, as well as presynaptic modifications, such as increases in presynaptic neurotransmitter release through paired-pulse facilitation (Orsini & Maren, 2012). Cellular signaling pathways that are required for the consolidation of fear memory involve protein synthesis (Orsini & Maren, 2012). Indeed, the extracellular regulated kinase (ERK) / mitogen-activated protein kinase (MAPK) signaling pathway is implicated in conditioning- and consolidation-induced synaptic plasticity (Orsini & Maren, 2012). Furthermore, multiple kinases are found to converge upon this pathway. For example, Ca2+/calmodulin-dependent protein kinase II (CAMKII) enhances synaptic efficacy by binding to NMDA and AMPA receptors in the postsynaptic density, and there is essential for conditioning-induced AMPA receptor trafficking and insertion (Orsini & Maren, 2012). Additionally, other kinases in the amygdala that mediate LTP and synaptic plasticity, such as cAMP-dependent protein kinase (PKA), Ca2+/phospholipid-dependent protein kinase (PKC), and its isoform PKMζ, are important for fear memory consolidation (Orsini & Maren, 2012). Conditioning-induced protein synthesis is further regulated through transcription factors, such as cAMP responsive element binding protein (CREB), and
CREB expression and subsequent IEG transcription determines which BLA neurons are incorporated into the memory trace (Kim et al., 2013; Rogerson et al., 2014). Lastly, CREB induces morphological changes (e.g., dendritic spine growth) during LTP through brain-derived neurotrophic factor (BDNF) signaling, which is necessary for fear conditioning (Orsini & Maren, 2012). Thus, through activation of cellular signaling pathways, fear conditioning can result in dendritic spine formation and alterations, for example through NMDA-dependent AMPA receptor insertion during LTP (Orsini & Maren, 2012).
¶2.2.4. LTP-Induced Changes in Synaptic Structure
It has been found that fear conditioning results in increased spine density in BLA principal neurons (Heinrichs et al., 2013; Leuner & Shors, 2013; Roozendaal et al., 2009). Fear conditioning leads to changes in dendritic spine morphology, and, more specifically, increased large mushroom spine density in the amygdala (Heinrichs et al., 2013; Keifer Jr et al., 2015; Pignataro et al., 2013). It has been suggested that this process, the conversion of thin spines into mushroom spines, is LTP-dependent and indicative of learning (Bourne & Harris, 2007).
¶2.2.5. Plasticity Occurs in Multiple Pathways
A dissociation can be made between cortical and thalamic LTP at LA synapses. While thalamic LTP is mainly NMDA receptor-dependent, cortical LTP requires additional cellular signaling cascades (Ehrlich et al., 2009; Orsini & Maren, 2012), which will be described in more detail below. Additionally, LTP is gated by activity of local inhibitory circuits. LA neurons receiving thalamic and cortical input also have synapses with local fast-spiking inhibitory interneurons, which control glutamatergic LTP by modulating feedforward GABAergic transmission postsynaptically and presyntaptically, respectively (Ehrlich et al., 2009). Specifically, fear conditioning results in a decrease in GABAergic signaling in the BLA (Ehrlich et al., 2009). Because of this finding, conditioning of BLA neurons is
suggested to occur in a manner analogous to receptive field plasticity in the sensory cortex (Ehrlich et al., 2009; Johansen et al., 2012). For example, LTP can be controlled by local inhibitory interneurons in the BLA through neuromodulator signaling, and through determining integration time windows and spatiotemporal spread for incoming sensory information (Ehrlich et al., 2009; Myers & Davis, 2007). The different types of interneurons in the BLA will be discussed in more detail at the end of this chapter, and the modulators of local inhibitory activity will be discussed in chapter 4.
¶2.2.6. LTP = Learning?
The finding that LTP in the LA underlies fear conditioning has been confirmed by cellular imaging (Barot et al., 2009; Barot et al., 2008), chemogenetics (Jasnow et al., 2013; Wolff et al., 2014), and optogenetics (Johansen et al., 2010; Kim & Cho, 2017; Nabavi et al., 2014). For example, it has been confirmed using optogenetics that LTP at LA synapses combined with an US is sufficient for the formation of fear memory and, conversely, long-term depression (LTD) of these synapses leads to impairments in fear expression which can be recovered by subsequent LTP (Johansen et al., 2010; Nabavi et al., 2014). Importantly, LTD after LTP is not extinction of the fear memory, because subsequent optical LTP could not recover the CR when LTD was replaced by extinction learning; nor did subsequent optical LTP produce a CR in animals that only received behavioral fear conditioning and then extinction learning (Nabavi et al., 2014). These findings indicate that LTP indeed underlies fear learning.
§2.3. Plasticity in Fear Extinction
Synaptic plasticity also underlies extinction of fear memory. Fear extinction-induced plasticity has some similarities to conditioning-induced plasticity. For example, the BLA is also a key neural locus for plasticity after fear extinction learning (Maren, 2015; Orsini &
Maren, 2012). Furthermore, extinction-induced plasticity includes postsynaptic as well as presynaptic modifications (Orsini & Maren, 2012).
¶2.3.1. Fear Extinction Relies on Multiple Mechanisms
As discussed in the first chapter of this thesis, fear extinction involves learning of a new, inhibitory memory trace. Extinction memory traces are thought to compete with fear memory traces (Royer & Paré, 2002; Tronson et al., 2012). Extinction is viewed as a parallel NMDA-dependent learning process (Davis, 2011; Mao et al., 2006; Orsini & Maren, 2012; Royer & Paré, 2002; Sotres-Bayon et al., 2007), facilitated through mechanisms similar to fear conditioning. Additionally, findings of erasure of parts of the original fear memory suggest that fear extinction learning not only relies on LTP, but also on mechanisms like depotentiation (DP) or LTD. Indeed, Maren (2015) identifies three putative mechanisms of extinction-induced inhibitory plasticity in the BLA: LTP of synapses on BLA interneurons or GABAergic neurons of the ITC; LTP of inhibitory synaptic transmission (iLTP) from BLA interneurons on BLA principal neurons; and DP or LTD of synapses on BLA principal neurons that were formed during fear conditioning (Figure 3A; Kim et al., 2007; Maren, 2015). Note that DP is the reversal of LTP, while LTD entails de novo silencing of synapses. As fear extinction only occurs after fear conditioning and associated LTP, I will only refer to DP in the current section. Furthermore, note that LTP of glutamatergic transmission onto inhibitory interneurons and iLTP of GABAergic transmission onto other neurons both result in suppression of output, but through different mechanisms.
Fear extinction depends on plasticity within a distributed neural network, discussed at the beginning of this chapter (Orsini & Maren, 2012). Neuroplasticity induced by fear
extinction is bidirectional (Maren, 2015; Riebe et al., 2012): on the one hand, inhibition of fear memory results in the suppression of fear expression; on the other hand, fear extinction results in the formation of new memory trace. Combined, these four mechanisms enable
dynamic regulation of fear memory through plasticity in the BLA (Maren, 2015). First, I will discuss the details and triggers of these mechanisms. Then, additional and putative
mechanisms will be discussed in chapter 4.
¶2.3.2. Long-Term Potentiation of Basolateral Amygdala Neurons
Fear extinction-induced LTP is NMDA-receptor dependent (Furini et al., 2014; Maren & Holmes, 2015; Walker et al., 2002; Zimmerman & Maren, 2010), and occurs in the LA during acquisition and in the IL during consolidation of extinction memory (Sotres-Bayon et al., 2007; Sotres-Bayon et al., 2009). Extinction-induced LTP results in the formation of CS-“no US” associations (Figure 3B). Thus, the acquisition and consolidation of fear extinction is facilitated by NMDA receptor-dependent plasticity in the LA and IL, respectively (Orsini
Figure 3. Three inhibitory and one excitatory synaptic mechanisms of fear extinction. A. LTP of synapses on BLA interneurons (left), iLTP from BLA interneurons onto BLA principal neurons (middle), and DP of synapses on BLA principal neurons that were potentiated during conditioning (right). Adapted from Maren (2015) and Trouche et al. (2013). B. Possibly, excitatory CS afferents potentiate LA neurons, forming CS-“no US” associations, while IL afferents potentiate BA “extinction neurons” that suppress CeA output. Lines with arrowheads indicate excitatory connections, lines with circles indicate inhibitory connections. Solid red lines indicate strengthening of synaptic connections, dotted red lines indicate weakening of synaptic connections. CS, conditioned stimulus afferents; DP, depotentiation; IL, infralimbic subdivision of the medial prefrontal cortex; iLTP, LTP of inhibitory synaptic transmission; LTP, long-term potentiation; LTP, long-term potentiation at inhibitory interneurons; PV, parvalbumin-positive interneuron; PN, BLA principal neuron.
& Maren, 2012; Tovote et al., 2015). Furthermore, it seems plausible that BA “extinction neurons” also form a target for extinction-induced NMDA-dependent LTP, for example of IL inputs, although evidence therefor is sparse (Amano et al., 2010; Kim & Cho, 2017).
¶2.3.3. Long-Term Potentiation of GABAergic Interneurons
Increases in GABAergic synaptic transmission in the BLA have been implicated in fear extinction learning (Chhatwal et al., 2005a; Duvarci & Paré, 2014; Heldt & Ressler, 2007; Lin et al., 2009; Sangha et al., 2009), consistent with the notion that fear extinction requires inhibition of fear networks in the BLA. Indeed, GABAergic activity is required for the acquisition and context-dependent expression of fear extinction memory and the
suppression of conditioned fear (Orsini & Maren, 2012). Extinction-induced LTP at BLA interneurons is AMPA receptor-dependent (Polepalli et al., 2010). Furthermore, specific glutamatergic transmission from BLA neurons onto GABAergic cells in the ITC, modulated by neuropeptide S, is important in fear extinction (Jüngling et al., 2008).
¶2.3.4. Long-Term Potentiation of Inhibitory Synaptic Transmission
Suppression of fear memory traces by extinction is mediated during acquisition by enhanced inhibitory synaptic transmission in the BLA through upregulation of GABAergic synapses (Chhatwal et al., 2005a; Heldt & Ressler, 2007), and during retrieval by increased activation of inhibitory circuits in the BLA following stronger activation of interneurons by the IL (Ehrlich et al., 2009). A role for iLTP in extinction is supported by findings of plasticity of inhibitory synapses (Chhatwal et al., 2005a; Trouche et al., 2013; Vogel et al., 2016) and findings of spine formation on principal neurons after fear extinction (Maroun et al., 2013). Furthermore, extinction induces GABA receptor trafficking (Lin et al., 2009). Thus, extinction learning results in enhanced GABAergic synaptic transmission in the BLA.
¶2.3.5. Depotentiation of Conditioning-Induced Plasticity
Other findings imply DP in fear extinction. For example, extinction of fear reverses conditioning-induced LTP in the thalamo-BLA pathway (Kim et al., 2007; Lin et al., 2003a). This DP is dependent upon NMDA receptors and possibly VGCCs, and halts MAPK
signaling, resulting in increased intracellular calcineurin (Lin et al., 2003b; Orsini & Maren, 2012). Moreover, DP results in AMPA receptor endocytosis, possibly mediated by
calcineurin (Kim et al., 2007), reducing the conditioning-induced increase in BLA
AMPA/NMDA ratio (Lin et al., 2010). Indeed, AMPA receptor endocytosis has been shown to erase the original fear memory (Clem & Huganir, 2010; Monfils et al., 2009). Thus, extinction induces NMDA-dependent synaptic weakening (Orsini & Maren, 2012).
Thus, the AMPA/NMDA ratio of BLA principal neurons decreases after extinction learning through NMDA-mediated endocytosis of AMPA receptors, almost to a
pre-conditioning degree (Asede et al., 2015; Clem & Huganir, 2010; Kim et al., 2007; Orsini & Maren, 2012). Consequently, there is a reduction in the threshold of synaptic potentiation in the BLA after extinction learning (Lee et al., 2013). Thus, extinction learning reverses conditioning-induced increases in the synaptic expression of AMPA receptors (Mao et al., 2006). Indeed, it has been found that NMDA receptor-dependent DP aids the reversal of conditioning-induced plasticity in the BLA after extinction learning, showing that extinction is, in specific cases, at least partly accomplished by erasure of the original fear memory trace (Lin et al., 2003a; Lin et al., 2003b). However, one study found no decrease in
AMPA/NMDA ratio when conditioning was immediately followed by extinction learning (Lin et al., 2011). Additionally, it has been found that pharmacological activation of NMDA receptors in the BLA promotes fear extinction and prevents fear reinstatement, and extinction is impaired by NMDA receptor inhibitors (Mao et al., 2006; Walker et al., 2002; Zimmerman & Maren, 2010). For example, administration of the NMDA-agonist d-cycloserine was found
to facilitate extinction while reversing the conditioning-induced increase in the AMPA/NMDA ratio (Davis, 2011; Lin et al., 2010).
Additionally, NMDA receptors also mediate DP through inhibitory synapses on BA “fear neurons,” although it remains unclear how exactly (Maren, 2013; Zimmerman & Maren, 2010). One study found NMDA receptor-mediated DP of cortical synapses on BLA neurons that was associated with fear extinction (Hong et al., 2009), although there is also a role for presynaptic plasticity (Maren, 2015), which will be discussed in chapter 4. Finally, a role for DP of BLA neurons in fear extinction is supported by findings of a decrease in spine density after extinction learning (Heinrichs et al., 2013),
¶2.3.6. Extinction-Induced Cellular Mechanisms
Acquisition of fear extinction relies on the ERK/MAPK signaling cascade (Kim et al., 2007; Lin et al., 2003b; Sotres-Bayon et al., 2007) and induces c-Fos upregulation in the BA (Herry & Mons, 2004). Other extinction-induced cellular mechanisms which are important to extinction-induced LTP and DP include AMPA receptor trafficking, c-Fos upregulation, CREB overexpression, increased BDNF signaling, and ERK/MAPK activity in BLA principal neurons (Maren & Holmes, 2015). For example, vHPC-mediated BDNF signaling results in the stabilization of extinction memory in the IL (Chhatwal et al., 2006; Peters et al., 2010). These mechanisms show partial overlap with conditioning-induced cellular
mechanisms, only differentiated by downstream targets (Tronson et al., 2012) and by balancing plasticity at excitatory and inhibitory synapses (Maren, 2015; Senn et al., 2014). Furthermore, extinction-induced signaling cascades include protein kinases, such as PKA, CAMK, and phosphatidylinositol-3 (I3k), which are involved in the stabilization of extinction memory (Orsini & Maren, 2012). Indeed, the MAPK signaling pathway, as well as associated activation of transcription factors, gene expression, and protein synthesis in the BLA and the IL, is essential for successful extinction (Herry & Mons, 2004; Herry et al., 2006; Lin et al.,
2003b; Lu et al., 2001). Interestingly, MAPK signaling can also be induced by eCB signaling, which will be discussed in chapter 4.
¶2.3.7. Extinction-Induced Changes in Synaptic Structure
Fear extinction learning is associated with changes in neuronal morphology in the BLA, such as dendritic spine density and synaptic structure (Keifer Jr et al., 2015; Pignataro & Ammassari-Teule, 2015). For example, it has been found, using a mouse model, that fear conditioning and then extinction learning involves a decrease in dendritic complexity, in spine number, and in spine density in the BLA, especially on third-order branches of principal neurons in the BLA, following the increase in spine density caused by fear conditioning (Heinrichs et al., 2013). More specifically, fear extinction resets BLA spine density to baseline levels, although the size of spines remains at pre-extinction levels (Pignataro & Ammassari-Teule, 2015). Another research study showed, using a rat model, that fear conditioning and then extinction produced decreased density of thin spines on first-order branches of principal neurons in the right BLA, with rats with relatively fewer and shorter dendritic branches showing the highest freezing during extinction; but fear extinction also resulted in dendritic retraction and increased density of thin spines on fourth-order branches in the left BLA (Maroun et al., 2013).
Whereas Heinrichs et al. (2013) found a reversal of the effects of fear conditioning, confirming findings of reversible conditioning-induced plasticity by fear extinction in the lateral amygdala (Clem & Huganir, 2010; Hong et al., 2011) and in the BLA (Sierra-Mercado et al., 2011); Maroun et al. (2013) found evidence of a lateralized differentiation in fear conditioning and extinction processes, where extinction-induced alterations in morphology are specific to the left BLA. The conclusion of Maroun et al. (2013) is supported by the notion that fear extinction involves the inhibition rather than erasure of fear memory (Wotjak & Pape, 2013). Furthermore, it should be noted that pre-extinction dendritic hypertrophy in
the BLA is associated with impaired extinction in the 129S1/SvImJ mouse strain (Camp et al., 2012), and successful extinction may indeed result in dendritic retraction as found by Maroun et al. (2013). Finally, whereas Heinrichs et al. (2013) subjected their animals to a more elaborate extinction learning and spine counting method (i.e., the Sholl method), the advantage of the research study by Maroun et al. (2013) is the differentiation by hemisphere; otherwise, the methods used by both studies are quite similar. Alternatively, it is suggested that plasticity in the BLA following fear conditioning and then extinction could just reflect the stressfulness of the testing (Maroun et al., 2013), because stress promotes plasticity in the BLA (Roozendaal et al., 2009), as discussed in the next chapter. Thus, although it is clear that the BLA is a key neural locus for fear extinction, the effects of structural plasticity and
lateralization, and whether or not these are complementary, remain somewhat unclear. Concluding, fear extinction learning results in a decrease of conditioning-induced spinogenesis, and an increase in dendritic spines that mediate inhibitory transmission.
Evidence of extinction-induced spinogenesis and associated synaptic transmission in the BLA consists of possible mediation by scaffold protein syntenin-1 (Talukdar et al., 2018),
neuroligins and neurexins (for contextual fear conditioning; Liu et al., 2016; Polepalli et al., 2017; Südhof, 2008), protease-nexin-1 (Meins et al., 2010), GABA-synthesizing GAD65 (for cued fear conditioning; Sangha et al., 2009), and BDNF (Peters et al., 2010).
¶2.3.8. Plasticity Occurs in Multiple Pathways
As discussed in this section, extinction-induced plasticity occurs in multiple pathways within the BLA. Furthermore, the IL has been identified as a source of extinction-induced plasticity of ITC cells via potentiation of BLA inputs (Amano et al., 2010; Jüngling et al., 2008, Likhtik et al., 2008), albeit disynaptic (Strobel et al., 2015). ITC plasticity is essential for successful extinction (Milad & Quirk, 2012; Sierra-Mercado et al., 2011). Furthermore, it has been found that NMDA receptor-dependent plasticity-mediated changes occur in ITC and
CeL neurons after fear extinction (Meins et al., 2010), indicating that the dissociation between “fear neurons” and “extinction neurons” found in the BLA (Herry et al., 2008) has an analogy in the ITC and the CeL (Meins et al., 2010), discussed at the end of this chapter.
Additionally, fear extinction results in a increase in spine density in the IL (Moench et al., 2016). Just as in fear acquisition, plasticity in the BLA is only involved in the acquisition stage of fear extinction. Plasticity-dependent extinction memory is then consolidated,
involving plasticity in the IL (Do Monte et al., 2015; Milad & Quirk, 2002). Specifically, the BLA is thought to be essential for acquisition and storage of extinction memory, while the IL and the vHPC are thought to mediate extinction consolidation and context-dependent
extinction expression, respectively (Orsini & Maren, 2012). Lastly, extinction retrieval
involves concerted activation by the IL and BLA (Herry et al., 2010; Quirk & Mueller, 2008).
¶2.3.9. Extinction = New Learning?
Using optogenetics, it has been confirmed that fear extinction results in the formation of new, inhibitory memory traces. Specifically, it has been found that fear extinction does not affect AMPA/NMDA ratios in CS-associated inputs to the BLA (Kim & Cho, 2017),
indicating that a CS-US representation is maintained even after extinction (Maren, 2017). Thus, fear extinction is indeed a form of new learning instead of unlearning.
Summarizing the findings discussed in this section, extinction causes a redistribution of fear memory within the amygdala as well as in the amygdala-hippocampal-mPFC network. Furthermore, within the amygdala, some neurons undergo DP after fear extinction, whilst others mediate CS-“no US” associative learning, and yet others maintain suppressed fear memory. The mechanisms that facilitate DP, LTP and the two types of inhibitory LTP have been discussed. As the IL, required for extinction consolidation, sends glutamatergic projections to inhibitory neurons within the amygdala, synaptic plasticity in this pathways could facilitate those mechanisms. This possibility will be explored at the end of this chapter.
§2.4. Microcircuits in the Amygdala in Fear Conditioning
As already mentioned in this chapter, fear expression is mediated through a specific subpopulation of neurons in the BLA. Specifically, information of the CS and the US converge within individual neurons in the LA, and it is then transmitted to a specialized subpopulation of BA principal neurons termed “fear neurons” (LeDoux, 2000; Maren & Quirk, 2004). This network is modulated by so-called microcircuits, consisting of neurons and interneurons within the BLA that modulate activity, necessary for the regulation of behavioral responses.
¶2.4.1. The Fear Engram
It is thought that fear memory is stored in the LA in the form of fear engrams,
activation patterns of subsets, or ensembles, of individual neurons that facilitate adaptive and dynamic behavior to fear (Davis & Reijmers, 2018; Grewe et al., 2017; Izquierdo et al., 2016; Morrison et al., 2016). Fear engrams enable adaptations to spatiotemporal dynamics and require a precise balance of stability and flexibility. This is achieved through manipulating synaptic plasticity in response to fear learning (Bocchio et al., 2017; Grewe et al., 2017; Gründemann & Lüthi, 2015), as described in the previous section. It has been found that LA neurons are sparsely allocated to the engram, and that this occurs based on relative neuronal excitability (Morrison et al., 2016).
¶2.4.2. Circuits of Fear Conditioning
“Fear neurons” in the BA are unidirectionally innervated by vHPC neurons, which may override retrieval of extinction memory (i.e., after extinction learning has taken place) in a context-dependent manner, and thus mediate fear renewal (Herry et al., 2008). BA “fear neurons” then transmit this information to the amygdala’s output nuclei in the CeA via three pathways, with direct output establishing the fear response (Orsini & Maren, 2012; Pape & Paré, 2010). One pathway consists of unidirectional excitatory synapses from LA neurons
onto CeL neurons, which then project to the CeM. Note that, after conditioning, the CS alone elicits a conditioned fear response through LA neurons (Orsini & Maren, 2012). Another pathway consists of excitatory LA projections onto BL and BM neurons, which then also project to the CeM. Finally, a third pathway consists of excitatory LA and BA projections onto the GABAergic ITC cells, which have inhibitory synapses in the CeA. ITC cells are also innervated by the IL, and are thought to have a modulatory role in the gating of information flow between BLA and CeA (Amir et al., 2011; Royer & Paré, 2002), which will be
discussed below.
Within the CeA, inhibitory microcircuits exist which are essential to fear acquisition (Ciocchi et al., 2010; Haubensak et al., 2010). Upon an aversive CS, the inhibitory control of the CeL over the CeM is suppressed, which drives the expression of the CR (Maren, 2011). Regulation of CeL inhibition is therefore crucial to fear expression. However, the BLA remains essential for consolidation and retrieval of fear memory (Orsisi & Maren, 2012). Thus, together with the finding of inhibitory microcircuits within the CeL (Ciocchi et al., 2010; Haubensak et al., 2010), it can be concluded that the CR is generated through disinhibition of the CeM via the CeL and ITC pathways (Ehrlich et al. 2009; Paré et al., 2004).
§2.5. Microcircuits in the Amygdala in Fear Extinction
Fear extinction involves many of the same brain areas that are involved in fear conditioning, although different circuits in those areas are involved. In general, fear extinction relies on the inhibition of fear circuits (Tovote et al., 2015). This is mediated by the strengthening of inhibitory synaptic transmission, as discussed earlier in this chapter. As mentioned in the previous paragraph, fear extinction is mainly facilitated by the BLA. Specifically, the BA is required for extinction learning, while the LA is required for storage of extinction memory (Amano et al., 2011; Herry et al., 2008; Herry et al., 2010). More
specifically, fear extinction depends on complex networks consisting of multiple parallel microcircuits of excitatory and inhibitory synapses in the BLA (Duvarci & Paré, 2014). These microcircuits are differentially recruited during fear expression and fear extinction, with network shifts depending on interactions with the mPFC which are discussed at the end of this chapter.
¶2.5.1. The Extinction Engram
As discussed in the previous section, fear memory is encoded in the brain in the form of fear engrams. During and after extinction learning, the CS engram is found to become more distinctive from the US engram, but without reverting to its initial form (Grewe et al., 2017). This finding is supportive of the notion that extinction learning is suppression of the fear memory, not erasure (Grewe et al., 2017; Myers & Davis, 2007).
¶2.5.2. Circuits of Fear Extinction
Fear extinction memory is encoded in a dynamic and distributed network including the BLA, vHPC, and mPFC, in which enhanced inhibitory activity mediates the signaling of safety of a particular context as well as the suppression of conditioned fear responses (Myers & Davis, 2007; Pape & Paré, 2010; Quirk & Mueller, 2008). Circuits involving fear
extinction center around the BLA (Duvarci & Paré, 2014; Izquierdo et al., 2016; Maren & Quirk, 2004; Pape & Paré, 2010), which is interconnected with the IL (Bukalo et al., 2015; Cho et al., 2013; Do Monte et al., 2015; Herry et al., 2008; Knapska et al., 2012; Likhtik & Paz, 2015; Orsini et al., 2011; Strobel et al., 2015) and the vHPC (Herry et al., 2010; Maren, 2001; Orsini et al., 2011).
2.5.2.1. Contextual modulation by the vHPC
The vHPC gates fear in PL neurons after extinction (Sierra-Mercado et al., 2011). Communication between the vHPC and the BLA is involved in fear extinction learning, but is not essential (Orsini & Maren, 2012). Specifically, the vHPC gates BLA-based fear in the PL
(Sotres-Bayon et al., 2012), i.e., the vHPC activates inhibitory interneurons in the PL thus increasing the CR. After extinction, the vHPC reduces fear by inhibiting PL responsiveness by BLA input (Sotres-Bayon et al., 2012). Furthermore, vHPC input to the IL is a major source of BDNF signaling necessary for successful fear extinction (Orsini & Maren, 2012; Peters et al., 2010).
¶2.5.3. Fear Extinction Involves a Network Shift
As already mentioned in this chapter, extinction of fear expression is mediated
through “extinction neurons” in the BLA. “Extinction neurons” are bidirectionally connected with IL neurons that mediate consolidation but not acquisition of extinction memory (Herry et al., 2008). Importantly, some “fear neurons” are extinction-resistant and maintain increased responsiveness to the conditioned stimulus, but extinction learning suppresses their output (Amano et al., 2011; Herry et al., 2008; Tovote et al., 2015). In “extinction neurons,” rapid reduction of CS responsiveness occurs through DP of thalamic inputs (Kim et al., 2007). Thus, extinction results in a network shift in the BLA and a reorganization of fear memory, allowing for flexible and context-dependent changes (Duvarci & Paré, 2014; Grewe et al., 2017; Orsini & Maren, 2012; Pape & Paré, 2010).
¶2.5.4. Inhibitory Activity in BLA Microcircuits
As discussed, fear extinction is partly established by inhibition of BLA principal neurons (Herry et al., 2010; Maren & Quirk, 2004). Specifically, it is thought that local inhibitory microcircuits within the BLA suppress fear expression and facilitate fear
extinction. In such a microcircuit, GABAergic BLA interneurons are recruited by the mPFC for context-dependent inhibition of CS-evoked activity of BA “fear neurons” (Ehrlich et al., 2009; Herry et al., 2010).
Indeed, it has recently become clear that especially inhibitory microcircuits in the BLA are crucial for extinction processes (Krabbe et al., 2018). Moreover, it has been
suggested that specific principal neurons in the BLA facilitate extinction through inhibitory processes (Herry et al., 2010; Krabbe et al., 2018; Maren & Quirk, 2004). For example, it has been found that fear conditioning and fear extinction are regulated by balancing the activation of differential microcircuits containing BLA principal neurons that project to the PL and the IL, respectively (Senn et al., 2014; Vogel et al., 2016). Thus, fear extinction activates a distinct population of BLA principal neurons that project to the IL (Amano et al., 2011; Grewe et al., 2017; Herry et al., 2008; Krabbe et al., 2018; Livneh & Paz, 2012; Repa et al., 2001; Senn et al., 2014). Note that for the distinction between PL-projecting and
IL-projecting principal neurons in the BLA, I here use the previously discussed denomination of “fear neurons” and “extinction neurons,” respectively (Duvarci & Paré, 2014; Gründemann & Lüthi, 2015; Herry et al., 2008; Jasnow et al., 2013).
2.5.4.1. Balanced microcircuit activity regulates fear extinction
Using optogenetics, it has been confirmed that glutamatergic “extinction neurons” in the BLA, identified by the gene Thy1, upon activation indirectly inhibit a subpopulation of CeM neurons and suppress LA excitatory activity (Duvarci & Paré, 2014; Jasnow et al., 2013). Furthermore, activation of “extinction neurons” during conditioning impaired fear acquisition, and activation of “extinction neurons” during CS presentations enhanced extinction retrieval (Jasnow et al., 2013). Thus, activity of “extinction neurons” in the BLA can impair consolidation of fear memory and promotes consolidation of extinction memory (Jasnow, 2013).
As discussed, fear extinction induces pathway-specific plasticity. This enables an asymmetry in the regulation of inhibitory microcircuits, which contributes to selective activation of BLA output pathways during fear extinction (Senn et al., 2014; Vogel et al., 2016). Furthermore, within the BLA, fear conditioning and extinction induce opposing changes in the strength of inhibitory transmission and in the expression of inhibition-related
genes, enabling dynamic regulation of fear memory by distinct disinhibitory microcircuits (Chhatwal et al., 2005a; Ehrlich et al., 2009; Harris & Westbrook, 1998; Heldt & Ressler, 2007; Wolff et al., 2014). Thus, regulation of activity by inhibitory microcircuits in the BLA is essential for successful fear extinction.
¶2.5.5. Inhibitory Activity in the mITC of the Amygdala
The intercalated cell masses (ITC) are an inhibitory gate for the input and output nuclei of the amygdala that control fear expression and its extinction. The ITC can be
functionally divided into the lateral paracapsular ITC (lITC) and the medial paracapsular ITC (mITC), which contain GABAergic neurons that convey feedforward inhibition to the BLA and, in the latter case, from the BLA to the CeA (Asede et al., 2015; Ehrlich et al., 2009; Paré et al., 2004). Additionally, the mITC contains two subsets of functionally different neurons (Bienvenu et al., 2015; Ehrlich et al., 2009). Indeed, opposing plasticity at mITC neurons further suggest the existence of two functionally different types of mITC neurons. Inhibitory control over BA “fear neurons” by mITC “fear-suppressing neurons” is decreased upon fear conditioning, resulting in subsequent recruitment of mITC “fear-promoting neurons” by BA “fear neurons” that drives fear expression (Asede et al., 2015). Fear retrieval, then, relies on disinhibition of “fear-promoting neurons” and inhibition of “fear-suppressing neurons” in the mITC (Asede et al., 2015).
Specifically, extinction learning modulates mITC regulation of amygdala microcircuit activity using feedforward and feedback inhibition of amygdala input and output nuclei (Asede et al., 2015). Interestingly, in contrast to BLA principal neurons, the AMPA/NMDA ratio of mITC neurons increases upon extinction learning, showing a reversal of a
conditioning-induced decrease in synaptic strength and mITC-mediated feedforward inhibition by extinction (Asede et al., 2015). Furthermore, extinction learning leads to expression of immediate early genes (IEGs) in mITC neurons (Busti et al., 2011; Knapska &
Maren, 2009). Lastly, administration of neuropeptide S to mITC neurons increased their BLA input and enhanced extinction learning (Jüngling et al., 2008). Taken together, these findings indicate that mITC neurons are essential for successful extinction.
Thus, following extinction learning, activation of a subset of mITC neurons by BLA “extinction neurons” suppresses CeM output and expression of the fear response.
Additionally, decreased activation of another subset of mITC neurons by inhibited BA “fear neurons” reduces inhibition of CeL neurons, in turn again suppressing CeM output (Ehrlich et al., 2009; Likhtik et al., 2008). Possibly, these two different pathways are active during the acquisition and consolidation phases on the one hand, and expression and retrieval phases on the other hand of fear extinction, respectively.
2.5.5.1. The IL-mITC pathway
Another inhibitory microcircuit that is suggested to enable inhibition of fear consists of IL projections onto inhibitory neurons in the mITC (Bruchey et al., 2007; Maren, 2005). mITC neurons function as an inhibitor of fear responses (Maren, 2011; Quirk & Mueller, 2008). For example, it has been found that extinction-induced stimulation of the mITC by excitatory input from the BLA and the IL suppresses the fear response by inhibiting excitatory transmission from the BLA to the output nuclei of the amygdala (Amano et al., 2010; Asede et al., 2015; Huang et al., 2014; Likhtik et al., 2008; Lin et al., 2003a; Lin et al., 2003b; Palomares-Castillo et al., 2012; Quirk et al., 2003). Specifically, IL-induced NMDA-dependent presynaptic plasticity in the BLA-mITC pathway enhances feedforward inhibition to the CeA and facilitates extinction memory formation (Amano et al., 2010; Jüngling et al., 2008; Knapska & Maren, 2009; Royer & Paré, 2002).
Using optogenetics, it has been confirmed that both the lITC and the mITC receive auditory input and are important in fear learning (Strobel et al., 2015). The lITC, upon
