Dopaminergic Plasticity of the Human Brain in the. Expression and Inhibition of Conditioned Fear

Transcription

1 Dopaminergic Plasticity of the Human Brain in the Expression and Inhibition of Conditioned Fear Jennifer I. Lissemore Integrated Program in Neuroscience McGill University, Montreal July 2016 A thesis submitted to McGill University in partial fulfillment of the requirements of the degree of Master of Science Jennifer I. Lissemore 2016

2 ABSTRACT Fear-associated learning may play a role in the development and maintenance of some anxiety disorders. Extensive evidence from rodent studies suggests that dopaminergic plasticity within the mesocorticolimbic dopamine pathway is critical for the expression and inhibition of conditioned fear. There remains a dearth of evidence from human studies, however, implicating mesocorticolimbic dopamine in fear conditioning. We used high-resolution positron emission tomography (PET) and the 18 F-Fallypride radioligand to investigate central dopamine release in humans during the expression and inhibition of conditioned fear. Twelve volunteers (6F/6M; mean ± SD age = 24.1 ± 3.7 years) underwent three PET scans each. First, a baseline scan (PET1) was performed. Subsequently, following the acquisition of conditioned fear, a 2 nd PET scan (PET2) measured D2/3 receptor binding in response to an aversively conditioned cue. Lastly, a 3 rd scan (PET3) investigated D2/3 receptor binding in response to the same cue following the inhibition of conditioned fear. Skin conductance response and subjective ratings of anxiety associated with the conditioned cue were significantly higher during PET2 than PET3 (p = 0.006, p < 0.001, respectively), suggesting the presence of a conditioned fear response during PET2 only. Elevated dopamine release was found in the bilateral hippocampus and left amygdala during PET3 compared to baseline (p<0.001, uncorrected), suggesting the activation of emotional memory processes during the suppression of learned fear. Dopamine release was also increased in the posterior cingulate gyrus during the inhibition (PET3), compared to the expression (PET2), of learned fear (p < 0.001, uncorrected). Surprisingly, no significant changes in dopamine release were observed during the expression of learned fear. To our knowledge, these are the first findings in humans to suggest that dopaminergic plasticity within brain regions implicated in associative memory processes is involved in the suppression of conditioned fear. i

3 RÉSUMÉ Conditionnement par la peur est un processus d'apprentissage fondamental impliqué dans certains troubles anxieux, y compris le syndrome de stress post-traumatique. Les données des études sur les animaux suggèrent que l'activité de la dopamine dans les régions mésocorticolimbiques du cerveau peut être nécessaire pour l expression et l inhibition de la peur conditionnée. Il reste un manque d études humaines, cependant, impliquant la dopamine mésocorticolimbique dans le conditionnement par la peur. Nous avons utilisé la tomographie par émission de positrons (TEP) à haute résolution avec le radioligand 18 F-Fallypride pour enquêter sur la libération de la dopamine régionale dans le cerveau de l'homme sain au cours de l expression et de l inhibition de la peur conditionnée. Douze volontaires (6F/6M ; moyenne d âge = 24,1± 3,7 ans) ont subi trois scans TEP chacun. Premièrement, une analyse de base (TEP1) a été réalisée. Par la suite, après l acquisition de la peur conditionnée, un 2 ème scan TEP (TEP2) a mesuré le potential de fixation de 18 F-Fallypride pendant la présentation d un stimulus conditionné. Enfin, le 3 ème scan (TEP3) a étudié la libération de la dopamine pendant la présentation du même stimulus neutre après l inhibition de la peur conditionnée. La conductivité de la peau et les évaluations subjectives de l anxiété associée au stimulus conditionné étaient significativement plus élevés au cours de TEP2 que TEP3 (p = 0.006, p < 0.001, respectivement), ce qui suggère la présence d une réponse de peur conditionnée pendant TEP2 seulement. Nous avons trouvé la libération de dopamine élevée dans l'hippocampe bilatérale et l amygdale gauche lors de l inhibition de la peur conditionnée, par rapport à la ligne de base (p < 0.001, non corrigé), ce qui suggère l activation des processus de mémoire émotionnelle lors de la répression de la peur conditionnée. Le relâchement de la dopamine a aussi été augmentée dans le gyrus cingulaire postérieur lors de l inhibition (TEP3), par rapport à l expression (TEP2), de la ii

4 peur conditionnée (p < 0.001, non corrigé). Curieusement, aucun changement significatif (p > 0.5) de la libération de dopamine n a été observé au cours de l expression de la peur conditionnée. Ces résultats suggèrent que la plasticité dopaminergique dans les régions du cerveau impliquées dans les processus de mémoire associative est importante dans la suppression de la peur conditionnée. iii

5 ACKNOWLEDGEMENTS The work presented in this thesis required the patient and dedicated help of several individuals. First, my supervisor, Dr. Chawki Benkelfat, has been supportive throughout my degree, and provided me with countless opportunities for which I am enormously grateful. Importantly, this work was also done in close collaboration with Dr. Atsuko Nagano-Saito, who made this project possible through immense help with design, implementation and troubleshooting, and who taught me various neuroimaging analysis skills. Further, I d like to thank my committee members: Dr. Marco Leyton was an integral part of this project, allowed me to unofficially join his lab group, and always had answers to my questions. Dr. Jens Pruessner brought up some invaluable perspectives on the project, for which I am very appreciative. Kelly Smart proved to be a good friend, who I frequently turned to for support, and with whom I shared many laughs. Additionally, all of the members of the Leyton lab were incredibly helpful and friendly during my time in the Ludmer building, and the lab nurses were instrumental in the completion of this study; Dominique Allard guided me through psychiatric interviews and ECGs, and France Durand helped me to retrieve lost blood work on many occasions. I would like to acknowledge funding support from NSERC and FRQS, as well as grant funding from CIHR. Additionally, the PET Unit and cyclotron team, who supported over a hundred hours of scanning, and the MRI Unit, made scanning an enjoyable learning process. To my family, thanks for always being supportive of my work, and for giving me every opportunity to pursue the path I choose. Last but not least, Jeffrey Munzar has pushed me to try new things, always made everything sound easier than it was, and kept me laughing. iv

6 CONTRIBUTION OF AUTHORS Jennifer Lissemore was responsible for study preparation, data collection, data analysis and interpretation, and thesis preparation. Dr. Atsuko Nagano-Saito was involved in the study design, preparation, data acquisition, analysis and interpretation. Prof. Chawki Benkelfat supervised the project and contributed to study design and data interpretation. Prof. Marco Leyton was also involved in study design and data interpretation. Dr. Paul Gravel contributed to the data analysis. The study was funded by the Canadian Institute of Health Research (CIHR). Jennifer Lissemore was supported by funding from NSERC and FRQS. v

7 TABLE OF CONTENTS ABSTRACT... i RÉSUMÉ... ii ACKNOWLEDGEMENTS... iv CONTRIBUTION OF AUTHORS... v LIST OF TABLES... viii LIST OF FIGURES... ix INTRODUCTION Fear-Associated Learning Neural Circuitry of Conditioned Fear Animal Literature Human Literature Neural Circuitry Implicated in the Inhibition of Conditioned Fear Animal Literature Human Literature Dopamine and Associative Learning Dopamine and the Expression of Conditioned Fear Dopamine and the Inhibition of Conditioned Fear Measuring Dopamine Release In Vivo in Humans: 18 F-Fallypride Hypotheses METHODS Participants Procedure Fear Conditioning Paradigm Autonomic Arousal Indices of Conditioned Fear Apparatus PET and MRI PET and MRI Data Processing Statistical Analyses RESULTS Participant Characteristics Subjective Ratings vi

8 3.3 Autonomic Arousal Indices of Conditioned Fear Brain-wide, Voxel-wise Analyses Region of Interest Analyses DISCUSSION & CONCLUSIONS REFERENCES vii

9 LIST OF TABLES Table 1: Study Timeline 18 viii

10 LIST OF FIGURES Figure 1. Timeline of each PET scan. (a) PET1 with no stimulus presentations. (b) PET2 with CS+ presentations (neutral shape associated with shock). (c) PET3 with CS- presentations (neutral shape no longer paired with shock) Figure 2. An axial slice demonstrating a portion of the ROI mask used for the a priori hypothesis-based analyses, overlaid on an average T1 MRI of all participants. Shown here: the anterior hippocampus (left, pink; right, dark blue), amygdala (left, light blue; right, green), nucleus accumbens (left, white; right, red), and ventromedial prefrontal cortex (left, yellow; right, orange).. 26 Figure 3. The mask applied for voxel-wise comparisons comparing BP ND between PET scans (BP ND > 0.6, t > 4).. 27 Figure 4. Mean percentage frequency of skin conductance responses (SCRs) in each PET scan. Error bars reflect standard error of the mean. 29 Figure 5. An example from one participant of skin conductance response (microsiemens) over time during PET2 and PET3 (the expression and inhibition of conditioned fear, respectively). Each trial is 30s: the triangle indicates the onset of the 3s conditioned stimulus, which is followed by a 20s countdown, and a blank screen during which the participant either does (PET2) or does not (PET3) expect a shock 30 Figure 6. t-maps in coronal, axial and sagittal views demonstrating decreases in 18 F-Fallypride binding potential in the bilateral anterior hippocampus (the left hippocampal cluster is identified with the white arrow; MNI coordinates of peak: x, y, z = -30, -13, -24), and the left amygdala (identified with the red arrow; x, y, z = -24, 3, -23) in PET3 during the inhibition of conditioned ix

11 fear, as compared to PET1 during baseline. Coloured t-maps are thresholded for visualization purposes, and overlaid on an average T1 MRI of all participants. 31 Figure 7. t-maps in coronal, axial and sagittal views demonstrating decreases in 18 F-Fallypride binding potential in the bilateral posterior cingulate gyrus (identified with the white arrow; x, y, z = -8, -44, 35) in PET3 during the inhibition of conditioned fear, as compared to PET2 during the expression of learned fear. Coloured t-maps are thresholded for visualization purposes, and overlaid on an average T1 MRI of all participants 32 Figure 8. Mean BP ND across PET scans in the left and right anterior hippocampus. Left hippocampus: * p = 0.04 (paired t-test), p = 0.1 (paired t-test); right hippocampus: p = 0.06 (Wilcoxon Signed-Rank Test), * p = 0.02 (paired t-test), uncorrected for multiple comparisons. Error bars represent standard error of the mean. 33 Figure 9: Mean BP ND across PET scans in the left and right amygdala. Left amygdala: * p = 0.03 (Wilcoxon Signed-Rank Test), p = 0.13 (paired t-test); right amygdala: p = 0.14 (Wilcoxon Signed-Rank Test), p = 0.11 (paired-t-test), uncorrected for multiple comparisons. Error bars represent standard error of the mean. 34 Figure 10: Mean BP ND across PET scans in the ventral tegmental area (VTA), ventromedial prefrontal cortex, and left and right medial prefrontal cortex. No significant changes in BP ND between scans were observed. Error bars represent standard error of the mean x

12 INTRODUCTION Fear conditioning is a learning process putatively involved in a range of psychiatric disorders, including post-traumatic stress disorder (PTSD), panic disorder, specific phobias (1-4), obsessive-compulsive disorder (5, 6) and schizophrenia (7, 8). An understanding of the mechanisms underlying the acquisition, expression and inhibition of learned fear may therefore offer novel therapeutic insight. In classical fear conditioning, during the acquisition of conditioned fear, a neutral, conditioned stimulus (CS) is repeatedly paired with an aversive, unconditioned stimulus (US). This results in the expression of learned fear, in which the CS alone evokes a fear response (9). The CS-US association can then be inhibited using fear modulation techniques, such as extinction, in which the CS is repeatedly presented without the US, or through reversal learning, in which a different neutral stimulus becomes associated with the US. The catecholaminergic neurotransmitter dopamine (DA) is involved in the neurobiological response to rewarding and aversive stimuli, and extensive evidence from animal studies suggests that DA plays a fundamental role in fear conditioning, as reviewed by Pezze and Feldon (10). The mesocorticolimbic DA pathways, which project from the ventral tegmental area to the nucleus accumbens, amygdala, hippocampus and medial prefrontal cortex, have been of particular interest in the study of aversive learning. Although DA within this circuit has been shown to be necessary for different stages of fear conditioning through studies in rodents, there is a paucity of evidence from human studies implicating central DA in aversive learning. In human studies of associative learning with rewarding stimuli, conditioned DA release within the striatum (11), limbic regions (12) and 1

13 medial prefrontal cortex (13) has been demonstrated in response to reward-related conditioned cues. In human neuroimaging studies of associative learning with aversive stimuli, activity within mesocorticolimbic brain regions during the acquisition, expression and inhibition of learned fear has been observed, however the specific involvement of brain regional dopamine in different stages of fear conditioning has yet to be explored in the healthy human brain. 1.1 Fear-Associated Learning Fear conditioning is an associative learning process in which a neutral CS (such as a light, shape or tone) is repeatedly paired with an aversive US (such as an electric shock, air blast or loud, unpleasant tone), resulting in a conditioned fear response during presentation of the neutral, conditioned cue alone. There are different phases of fear-associated learning that putatively involve overlapping but distinct neural processes: (1) the acquisition of conditioned fear, in which the association between the CS and US is learned (i.e. the formation of a fear memory trace associated with the CS), (2) the expression of conditioned fear, which occurs in response to the CS following acquisition (i.e. the retrieval of the CS-associated fear trace, accompanied by a conditioned fear response), (3) extinction or reversal learning, in which it is learned that the CS is no longer associated with the US (i.e. formation of a safety trace associated with the CS), and (4) the inhibition of conditioned fear, in which the safety trace, rather than the fear trace, is retrieved in response to the CS. Furthermore, there is substantial variability in the types of fear conditioning paradigms that can be studied. Fear conditioning can be further subdivided into delay conditioning, in which the CS and US are presented simultaneously, and trace conditioning, in which there is a temporal gap between the CS and US. Additionally, the US can be paired with a context (context- 2

14 dependent fear conditioning), or with a discrete cue (cue-dependent fear conditioning) using single-cue (1 conditioned cue only) or differential (2 conditioned cues: CS+ paired with US, CSnot paired with US) fear conditioning paradigms. The CS and US can also be auditory, visual, cutaneous or olfactory in modality, and the CS-US contingency rate, (the fraction of CS presentations that are paired with the US), can vary from 0% (instructed fear) to 100%. Furthermore, there are many different methods for measuring conditioned fear responses, including freezing behaviour in rodents, fear potentiated startle (the presentation of a loud tone elicits a startle reflex, which can be amplified with simultaneous presentation of a conditioned fear cue), and measurements of autonomic arousal in humans. Fear conditioning is a fundamental learning process that can be adaptive, however abnormal fear conditioning processes may also contribute to the development and persistence of some psychiatric disorders. For example, in post-traumatic stress disorder (PTSD), there is evidence of delayed extinction learning and recall (3, 14). Understanding the neurobiology of associative learning with aversive stimuli is therefore valuable to the development of new treatment avenues for these disorders. 1.2 Neural Circuitry of Conditioned Fear There is a large body of literature, primarily from studies in rodents, directly or indirectly implicating mesocorticolimbic brain regions in the acquisition and expression of conditioned fear. Implicated regions include the ventral tegmental area (VTA), amygdala, hippocampus, nucleus accumbens, and medial prefrontal cortex (mpfc). 3

15 1.2.1 Animal Literature Lesion, inactivation, microstimulation and single/multi-unit recording studies in animals have identified various DA-modulated brain regions that are involved in the acquisition and expression of learned fear. For instance, illustrating the role of the VTA in fear conditioning, electrical stimulation of the VTA in rats was found to enhance fear-potentiated startle, while NMDA lesions of the VTA causing cell loss inhibited the expression of conditioned fear, as evidenced by a lack of potentiated startle (15). The VTA connects bidirectionally to the amygdala, and the amygdala has also been extensively implicated in the acquisition and expression of learned fear. Electrolytic or NMDA lesions of the amygdala have been shown to block the acquisition and expression of fearpotentiated startle in rodents (16, 17). Furthermore, mimicking the effects of a conditioned fear cue on the startle reflex, electrical stimulation of the amygdala immediately before a noise burst can enhance acoustic startle (18). Additionally, extracellular recordings from the central nucleus of the amygdala in rabbits have shown increased electrophysiological activity in the amygdala in response to a conditioned aversive stimulus, and this activity is correlated with conditioned heart rate responses (19, 20). The hippocampus is well established as a key region for learning and memory, and it has been shown to play a role in the formation and expression of learned fear. For example, rats with bilateral hippocampal lesions failed to express conditioned fear through freezing behaviour after a classical trace conditioning procedure, in which the CS and US are separated by a temporal gap (21). Similarly, muscimol-induced inactivation or electrolytic lesions of the ventral hippocampus have been shown to disrupt the acquisition and expression of conditioned fear after a classical auditory fear conditioning procedure in rats, as measured using freezing behaviour (22-24). 4

16 Lastly, the nucleus accumbens and mpfc have also been implicated in fear conditioning. Nucleus accumbens core lesions in rats, for instance, disrupted the acquisition of fear conditioning to a discrete cue (25). Similarly, inactivation of the prelimbic area of the mpfc has been shown to impair the expression of learned fear in rats (22). Additionally, single neuron recordings in the rat prelimbic subregion of the mpfc showed increased activity in response to the aversively conditioned cue in a trace conditioning paradigm (26), and microstimulations of the prelimbic subregion enhanced the expression of learned fear to a tone (27) Human Literature While the study of fear conditioning in rodents is informative, exploration of this learning process in humans is an essential next step. The advent of functional neuroimaging in the 1990 s allowed for the neural circuitry of fear-associated learning to be similarly studied in humans. A multitude of human functional magnetic resonance imaging (fmri) studies have found activation in regions of the mesocorticolimbic DA pathway during fear conditioning. Similar results have been found in human positron emission tomography (PET) studies using the H 15 2 O tracer to investigate cerebral blood flow during fear conditioning. Nevertheless, there has been considerable variability in the brain regions activated in human neuroimaging studies of fear conditioning, which could be accounted for by methodological differences between studies, as reviewed by Sehlmeyer and colleagues (28), including different CS/US modalities, participant characteristics, contingency rates, and neuroimaging modalities (fmri or H 15 2 O PET). Despite some inconsistent findings in neuroimaging studies, the amygdala is often implicated in human fear conditioning, in accordance with the animal literature. Cheng and colleagues (29), for example, found that the blood-oxygen level dependent (BOLD) response was increased in the amygdala during CS presentations accompanied by a fear response, i.e. during the expression of 5

17 learned fear as assessed using objective skin conductance response (SCR) measurements, compared to CS trials without a fear response. Neuroimaging studies have similarly shown that the hippocampus is involved in human fear conditioning, and hippocampal activation may be more prominent in studies that employ a trace conditioning procedure, as well as when a tactile US is used. For instance, in an aversive trace conditioning procedure using a 50% CS-US contingency rate and a loud sound as the US, the BOLD response to the CS+ alone (CS paired with US) was significantly higher in multiple regions including the amygdala and anterior hippocampus, compared to the CS- (the CS not paired with the US in a differential fear conditioning paradigm) (30). A similar study using concurrent delay and trace conditioning procedures, with electric shock as the US, also found a positive correlation between amygdalar and hippocampal activity and SCR measurements indicative of conditioned fear responses (31). The ventral striatum has also been implicated in some fmri studies of conditioned fear. When comparing the BOLD response to the CS+ and CS-, activation within the ventral striatum has been found at the onset of the CS+ (32). Activity in the ventral striatum and conditioned fear responses measured using SCR have also been found during fear conditioning in participants who reported learning the CS-US contingency, compared to those that did not (33). Notably, the above findings from human neuroimaging studies are correlational and do not causally implicate mesocorticolimbic brain regions in the acquisition and expression of conditioned fear in humans. Although causal evidence in humans is scarce compared with research in animals, it has been shown that unilateral temporal lobectomy, involving removal of the amygdala and hippocampus, leads to impaired fear learning (34). 6

18 1.3 Neural Circuitry Implicated in the Inhibition of Conditioned Fear After the formation of a CS-US association, it is critical that an individual be able to suppress the learned fear association once it is no longer relevant. The inhibition of learned fear can be achieved through fear extinction, in which the CS is presented repeatedly without the aversive stimulus, or through reversal learning, which is similar to extinction, except that the aversive stimulus simultaneously becomes associated with a different neutral cue. Regardless of the strategy used, the inhibition of conditioned fear is widely believed to involve the formation of a new associative memory, rather than erasure of the original associative memory (35, 36). Unfortunately, fewer studies have been conducted on the suppression of conditioned fear, relative to the acquisition or expression of conditioned fear Animal Literature While the prelimbic region of the mpfc has been implicated in the expression of learned fear, the infralimbic subregion of the mpfc, located more ventrally, has been implicated in the inhibition of learned fear in rodents. Using single-unit recordings, neuronal activity within the infralimbic subregion of the mpfc has been shown to increase during extinction recall in rats, and this activity correlated with the degree of extinction observed behaviourally (37). Further suggesting that the infralimbic subregion is involved in the active inhibition of learned fear, electrical stimulation of this area reduced conditioned fear, as indicated by decreased freezing behaviour (37), whereas inactivation impaired extinction learning and recall in rats (22). Although the infralimbic prefrontal cortex has been most extensively implicated in fear extinction in rodents, the amygdala and hippocampus have also been shown to play a role in fear extinction. For example, Sierra-Mercado and colleagues additionally found that inactivation of the basolateral amygdala or ventral hippocampus using muscimol infusions in rodents disrupted 7

19 extinction recall in an auditory fear conditioning and extinction paradigm (22). Furthermore, it has been shown that intra-amaygdala infusions of D-cycloserine, which promotes NMDA receptor activity, facilitated the extinction of conditioned fear in rats (38). Extinction is an integral component of reversal learning, in that the fear response to the CS that was initially paired with the US becomes extinguished following reversal of aversive reinforcement contingencies. Both extinction and reversal learning retroactively interfere with and modulate the initial learned fear memory (39). The neural correlates of reversal learning have been studied less extensively than those of extinction. Similarly to studies of extinction, however, lesions of the orbitofrontal cortex (OFC) in primates has been shown to impair performance on reversal learning tasks (40, 41) Human Literature Considerably fewer neuroimaging studies have been conducted on the inhibition of conditioned fear. Some fmri and PET H 15 2 O studies have, however, found activation within frontal cortices during the suppression of learned fear. For example, Molchan and colleagues (42) observed increased cerebral blood flow in the inferior frontal cortex during extinction learning, and Gottfried & Dolan (43) found activation of the OFC, among other regions, during extinction learning. Additionally, activations during extinction recall were observed in both the hippocampus and vmpfc, thought to be the human homologue to the rat infralimbic cortex, and positive correlations were found between activity in these regions, and the magnitude of extinction recall, as measured using SCR (44, 45). Along the same lines, increased activity in the vmpfc has been observed in response to the new CS- (previously paired with the US) compared to the new CS+ after a reversal learning procedure (46). 8

20 1.4 Dopamine and Associative Learning The above-mentioned brain regions, which have been implicated in the different phases of fear conditioning, are all modulated by dopamine (DA), which is an important catecholamine neurotransmitter in the brain. There are five DA receptor subtypes that fall into two classes: D1- like receptors (D1 and D5 receptors) and D2-like receptors (D2, D3 and D4 receptors). Dopaminergic neurons originate in the ventral tegmental area (VTA) and substantia nigra (SN) of the midbrain and project throughout the brain via two established pathways: the nigrostriatal DA pathway, which projects from the SN to the striatum, and the mesocorticolimbic DA pathway, which projects from the VTA to the amygdala, hippocampus, ventral striatum and prefrontal cortex. Whereas the nigrostriatal pathway has primarily been implicated in motor control, the mesocorticolimbic pathway is commonly implicated in the response to rewarding and aversive stimuli. Studies in animals have employed a variety of techniques to investigate the dopaminergic system in vivo. For example, microdialysis and voltammetry allow the study of dopamine release in response to stimuli in vivo in particular brain regions. Furthermore, dopamine can be selectively abolished and restored in particular brain regions of rodents, for example using viral vectors, in order to study the causal role of regional dopamine in certain behaviours. Unfortunately, such methods are not currently amenable to studies in humans, and most human studies to-date have explored the role of dopamine in fear conditioning using systemic pharmacological manipulations Dopamine and the Expression of Conditioned Fear A rich body of animal studies has demonstrated that dopamine is critical for the expression of conditioned fear. For example, systemic administration of the indirect DA agonist d- 9

21 amphetamine or the DA precursor tyrosine enhances the expression of conditioned fear in rats (47, 48). Moreover, Fadok and colleagues (49) demonstrated that fear-potentiated startle is abolished in DA-deficient mice, and that restoration of DA synthesis with administration of L- DOPA (3, 4-dihydroxy-L-phenylalanine) immediately after training reinstated conditioned fear. More specifically, dopaminergic activity within mesocorticolimbic circuitry has been shown to be of particular importance in the expression of learned fear. Increased levels of the dopamine metabolite DOPAC in the VTA have been demonstrated in response to both a mild footshock and a neutral auditory cue associated with footshock, as measured using high-performance liquid chromatography (50). Interestingly, Fadok and colleagues also found that viral-mediated restoration of DA synthesis specifically within the VTA of DA-deficient mice reestablished fearpotentiated startle (49). Correspondingly, it has been shown that infusion of the DA D2/3 receptor agonist quinpirole, which inhibits DA neurons, into the VTA of rats blocks the expression of learned fear (15, 51, 52). Dopaminergic activity in the amygdala, which receives projections from the VTA, is also believed to play a role in conditioned fear. In addition to decreasing conditioned fear, quinpirole injections into the VTA decreased CS-related increases in DA in the basolateral amygdala, as measured using microdialysis (53). Similarly, extracellular levels of DA in the amygdala have been found to increase during the acquisition and expression of aversive learning in rats (54). Further, injections of D2 receptor antagonists into the amygdala have been found to disrupt the acquisition and expression of learned fear (53, 55). Moreover, reflective of findings in the VTA, Fadok and colleagues (56) showed that selective viral-mediated restoration of DA synthesis in the basolateral amygdala of DA-deficient mice restored long-term conditioned fear memory. 10

22 Surprisingly, despite evidence suggesting that the ventral hippocampus is an important component of the circuitry mediating fear conditioning, dopaminergic activity within the ventral hippocampus during cued fear conditioning is an understudied topic. One recent study found that systemic injection of a D1-like receptor antagonist in mice blocked hippocampal long-term synaptic potentiation (LTP) during the acquisition of learned fear (57). By contrast, there is considerable evidence suggesting that dopaminergic mechanisms within the nucleus accumbens are involved in aversive learning. Numerous microdialysis studies in rats have found elevated extracellular DA in the nucleus accumbens in response to aversive stimuli (58, 59) and in response to aversively conditioned cues (60-62). In accordance with these findings, it was shown that recovery of DA synthesis in the nucleus accumbens of DA-deficient mice re-established long-term memory for fear-potentiated startle (56). Finally, DA transmission in the mpfc, which also receives dopaminergic projections from the VTA, is believed to be important for the expression of learned fear. As an illustration of this, Pezze, Bast & Feldon (63) showed that infusion of a D 1 /D 2 -receptor antagonist or d- amphetamine into the mpfc reduced the expression of conditioned fear to a tone in rats. Microdialysis studies have also shown that DA release in the mpfc increases in response to aversive stimuli (58, 64) and in response to aversively conditioned cues (65). There is, however, a dearth of evidence implicating dopaminergic mechanisms in human fear conditioning. An fmri study of healthy adults by Diaconescu and colleagues (66) investigated the effects of pharmacological manipulations on functional connectivity during fear conditioning. It was found that DA agonist and antagonist drugs changed the functional connectivity between regions implicated in fear conditioning, such as the ventral striatum and amygdala, in response to the CS+ compared to the CS-. Similarly, BOLD activity related to 11

23 prediction error during an aversive learning task was found to be altered by amphetamine (indirect DA agonist) or haloperidol (DA D2 receptor antagonist) (67). Nevertheless, the abovementioned findings are indirect investigations of dopamine, and do not provide a clear understanding of dopaminergic involvement in the expression of conditioned fear in humans Dopamine and the Inhibition of Conditioned Fear Less evidence exists regarding the involvement of central DA in the inhibition of learned fear. In rodents, intraperitoneal injections of d-amphetamine or a D1 receptor agonist in rats prior to non-reinforced CS presentations have been shown to impair extinction, as evidenced by persistent fear-potentiated startle in response to the CS (68). Accordingly, intraperitoneal injections of a D2 receptor agonist prior to extinction disrupted extinction memory, while a D2 receptor antagonist enhanced extinction recall in mice (69). Although systemic modulations of dopamine have been found to affect fear extinction, the involvement of brain regional dopamine in the inhibition of learned fear remains unclear. Within mesocorticolimbic circuitry, 6-hydroxydopamine lesions of the mpfc that reduced DA levels delayed extinction learning, and D4 receptor antagonist injections into the mpfc impaired extinction memory in rats (70, 71). Similarly, D2 receptor antagonist administration into the infralimbic subregion of the rat mpfc disrupted the inhibition of learned fear after extinction (72). Furthermore, novelty-induced enhancement of fear extinction has been shown to depend on dopamine D1 receptors in the hippocampus (73). Dopaminergic mechanisms in the reversal of fear learning have not yet become an active area of study. In non-fear-related reversal learning, however, it was shown that D2 receptor knockout mice and mice treated with a D2-like receptor antagonist demonstrated impaired reversal learning and had reduced expression of early growth response gene 2 in the mpfc and OFC after 12

24 the reversal learning test, but not after a different learning task (74). This suggests that D2 receptor activity in frontal regions is important for the cognitive flexibility required in reversal learning. Lastly, genetic studies in humans have shown that extinction learning is affected by a long D4 receptor exon III allele (75), and by a DAT1 polymorphism (76). However, to our knowledge, no direct investigations of brain regional dopaminergic activity in the inhibition of conditioned fear have been reported in humans Measuring Dopamine Release In Vivo in Humans: 18 F-Fallypride Using positron emission tomography (PET), it is now possible to investigate brain regional dopamine in humans in vivo. There are several radioligands currently available for the study of regional dopamine release, including the D2/3 antagonists 18 F-Fallypride, 11 C-Raclopride and 11 C-FLB F-Fallypride in particular is a highly selective, high affinity DA D2 and D3 receptor radiotracer that allows striatal and extrastriatal changes in dopamine release to be studied in humans in vivo (77). Quantification of 18 F-Fallypride binding in the human brain offers a reliable estimate of D2/3 receptor availability in striatal and extrastriatal regions (78). The distribution of 18 F-Fallypride binding has been shown to correlate with concentrations of D2/3 receptors measured in vitro in primates (79). 18 F-Fallypride and endogenous dopamine compete for D2/3 receptors, and increases in dopamine reduce the number of available D2/3 receptors. Therefore, changes in 18 F-Fallypride binding between scans are inversely related to changes in DA release in the brain (80), i.e. decreases in 18 F-Fallypride binding across scans reflect increases in dopamine release. Evidence of this relationship comes primarily from amphetamine challenge studies in humans and nonhuman primates, in which administration of d- 13

25 amphetamine, which is known to increase dopamine release (81), consistently reduces 18 F- Fallypride binding potential in striatal and extrastriatal brain regions (82, 83). There are many advantages to using the 18 F-Fallypride radioligand in human studies of mesocorticolimbic dopamine. Unlike 11 C-Raclopride, 18 F-Fallypride allows the study of D2/3 receptor binding in extrastriatal regions with low-moderate binding, including limbic regions such as the amygdala and hippocampus. Moreover, compared to the 11 C-FLB457 radiotracer, 18 F-Fallypride has a longer half-life (110 minutes) and faster uptake / washout kinetics, therefore it is currently the only tracer with the ability to image both striatal and extrastriatal regions. Disadvantages to the use of the 18 F-Fallypride tracer include lower sensitivity to d- amphetamine-induced changes in dopamine neurotransmission in cortical regions, as compared to 11 C-FLB457 (84), and less power to detect d-amphetamine-induced changes in dopamine release in the striatum, as compared to 11 C-Raclopride (80). Nevertheless, previous studies using 18 F-Fallypride PET have found significant changes in dopamine release in cortical and striatal regions in response to stress or drug-related cues (12, 85). 1.5 Hypotheses We hypothesized that following the acquisition of learned fear, 18 F-Fallypride binding potential would decrease in response to the CS+ in the VTA, amygdala, hippocampus, nucleus accumbens, and mpfc, as compared to baseline. This would indicate increased DA release in midbrain, limbic, striatal and prefrontal regions during the expression of learned fear. Correspondingly, we hypothesized that the amplitude of the skin conductance response to the CS+, indicative of a conditioned fear response, would correlate with the change in binding potential in these regions of interest (ROIs). 14

26 Second, it was hypothesized that following the subsequent suppression of the CS-US association, binding potential would decrease in the ventromedial prefrontal cortex in response to the CS-, as compared to baseline and the expression of conditioned fear. This would suggest increased DA release in the ventromedial prefrontal cortex during the suppression of learned fear. 15

27 METHODS 2.1 Participants Healthy, right-handed volunteers between the ages of 20 and 40 years were recruited from the community using an advertisement on the McGill University classifieds website. Exclusion criteria for the study included: (a) current or past Axis I disorder, including current or past substance abuse; (b) family history of Axis I disorder; (c) current or past chronic medication use, excluding birth control; and (d) significant physical illness in the past 12 months. Participants also had to be available at least one weekday (Monday-Friday) between the hours of 11:00 and 17:00. A low level of exposure to tobacco (< 5 cigarettes / day) and cannabis (< twice a month) was permitted, though they were asked to refrain from use for at least 24h and 1 week, respectively, before scanning. After completion of a brief online survey that screens for major exclusion criteria, eligible participants were ed with more information on the study. Participants then underwent a phone interview, followed by an in-person Structured Clinical Interview for DSM IV (SCID) (86), an electrocardiogram, standard blood work, a urine toxicology test (Triage, Biosite Diagnostics, San Diego, CA; sensitive to phencyclidine, tetrahydrocannabinol, amphetamine, cocaine, benzodiazepines and opiates), and a urine pregnancy test for women. Participants also completed the Zuckerman-Kuhlman-Aluja Personality Questionnaire (87) and the State-Trait Anxiety Inventory (STAI-Trait) (88). Next, participants underwent a physical examination by a medical doctor. Once participants were deemed mentally and physically healthy, a final screening session was conducted to ensure that participants showed an adequate physiological response to the aversive, 16

28 unconditioned stimulus being used in this study, consisting of a mild electric shock to the wrist. Specifically, all participants were required to show an increase in heart rate of at least 10 beats per minute, or an increase in skin conductance response of at least 10%, immediately before or after presentation of the aversive stimulus. The study was carried out in accordance with the Declaration of Helsinki, and was approved by the Research Ethics Board of the Montreal Neurological Institute. All participants provided written, informed consent. 2.2 Procedure Following screening, participants underwent 6 experimental sessions, consisting of three 3- hour PET scans, an MRI scan and two stimulus pairing sessions. The study timeline is shown in Table 1. First, a baseline PET scan (PET1) was performed, which served as a control. During this scan, participants were presented with a blank screen, and no stimuli were presented. Participants were instructed to stay awake, keep their eyes open, and relax, and it was explained that no shocks would be delivered during the scan (the shock electrode was not set up). Either immediately following PET1, or within 10 days of PET1 and prior to fear conditioning, participants also underwent an anatomical MRI scan for co-registration with PET. 17

29 Table 1. Study Timeline. Day 1 Day 6 Day 7 Day 13 Day 14 PET1: CS-US Pairing: PET2: CS-US Pairing: PET3: Baseline PET Acquisition of conditioned fear Expression of conditioned fear Reversal learning Inhibition of conditioned fear ALONE ALONE Approximately 1 week after the baseline PET scan, participants underwent a differential fear conditioning procedure in the PET room (see section 2.3 below). During this procedure, one of two neutral shapes / conditioned stimuli became associated with an electric shock / unconditioned stimulus (the CS associated with the US is termed the CS+). Next, on the business day following the acquisition of conditioned fear, a second PET scan (PET2) was performed to investigate 18 F-Fallypride binding in the brain in response to the CS+ alone, during the expression of learned fear. Immediately before PET2, participants completed a contingency awareness questionnaire to confirm intact memory of the CS-US association, and to prime the CS-US associative memory. Participants were instructed that only one of the shapes from pairing would be presented during the scan, and were shown which shape would be presented during the scan (the CS+) approximately 10 minutes before scanning, during the transmission scan (see Figure 1). Additionally, the shock electrode was set up, but no shocks were delivered. Following tracer injection, participants were presented with 60 trials of the CS+ with a 20-second countdown, in the absence of shock, during the first 30 minutes of the scan. At least 1 week after PET2, participants underwent a second CS-US pairing procedure to suppress the initial CS-US association (see section 2.3 below). The following business day, a 18

30 third and final PET scan (PET3) was performed to investigate dopamine release during the suppression of conditioned fear, in response to the CS no longer associated with shock (the CS-) (see Figure 1 for a summary of PET scan timelines). PET3 was performed identically to PET2, with the same shape being presented, and the only difference between these scans was the associative significance of the CS. As in PET 2, participants completed a contingency awareness questionnaire, confirming the correct CS-US association, immediately before PET3, and participants were shown the shape to be presented during the scan (the new CS-) for 10 minutes before tracer injection. (a) Transmission scan Tracer injection ~ 12:00pm Dynamic acquisition scan Break Dynamic acquisition scan 6 mins Arrival ~ 11:30am 90 mins 30 mins 60 mins (b) Transmission scan Tracer injection ~ 12:00pm Dynamic acquisition scan Break Dynamic acquisition scan CS+ Arrival ~ 11:30am (c) Transmission scan Tracer injection ~ 12:00pm Dynamic acquisition scan Break Dynamic acquisition scan 6 mins Arrival ~ 11:30am CS- Figure 1. Timeline of each PET scan. (a) PET1 with no stimulus presentations. (b) PET2 with CS+ presentations (neutral shape associated with shock). (c) PET3 with CS- presentations (neutral shape no longer paired with shock). 19

31 2.3 Fear Conditioning Paradigm Prior to beginning the experiment, the subjective pain threshold of each participant was established. Following administration of an electric shock to the wrist (US), participants rated the discomfort they experienced on a 5-point numerical rating scale (NRS) (0 = No Sensation; 1 = Just Noticeable; 2 = Uncomfortable; 3 = Pain Threshold; 4 = Painful; 5 = Maximum Tolerable), and on a visual analog scale (VAS) of pain (0 = No Pain; 100 = Extremely Painful) (89, 90). Pain threshold was defined as subjective ratings of 3 on the NRS and at least 20 on the VAS, and was described to participants as very uncomfortable, but not yet painful. Shock levels began at 20V and were increased in increments of 2V until pain threshold was reached. Each participant s pain threshold was re-established immediately prior to CS-US pairings, using the pain threshold established earlier as a starting point in order to minimize the number of shocks administered before CS-US pairing. Pairing sessions consisted of a cue-dependent, trace fear conditioning procedure with partial reinforcement. Two neutral, grey shapes, a circle and a triangle with equal surface areas, were presented as conditioned stimuli (CS+ or CS-), and the shape that was paired with shock was counterbalanced across participants (i.e. for half the participants, the triangle was paired with shock, and for half the participants, the circle was paired with shock). Ten CS+ and 10 CS- were presented in pseudo-random order, such that no more than 3 trials with the same cue were presented consecutively. Presentation of each CS (3 seconds) was followed by a 20-second countdown, then a blank screen in which participants either were or were not shocked. The contingency rate was 30% (i.e. a shock was administered in 3 / 10 CS+ trials). Both the low contingency rate and 20-second countdown were included to help sustain the stress response during PET2 by increasing anticipation of the unpleasant stimulus (91). Increasing the 20

32 uncertainty about receiving an electric shock using a low contingency rate has been shown to evoke a stronger anxiety response (92), and low contingency rates, as compared to high contingency rates, have been shown to elicit greater activation within fear conditioning circuitry (93, 94). Participants were instructed as follows prior to pairing: A series of two shapes will be presented; one of these shapes will sometimes be followed by a shock, and the other shape will never be followed by a shock. Participants were not informed of the contingency rate, or which CS would be paired with shock. After completion of all trials, participants rated the average level of discomfort experienced during the shocks on the NRS and VAS. Subsequently, a contingency awareness questionnaire was completed that assessed (i) which CS the participant associated with shock, (ii) the level of anxiety associated with each CS (1 = None; 2 = A Little; 3 = Moderate; 4 = Extreme), and (iii) how often each of the shapes was followed by a shock. If participants did not learn the correct association in the first 20 trials, the conditioning procedure was repeated. In the second pairing session, a reversal learning procedure was used that was identical to the first pairing session, except that the associations were reversed; the CS+ became the CS-, and vice versa. For example, if the triangle was paired with shock in the first pairing session, during the second pairing session, the triangle was presented repeatedly in the absence of shock, and the circle was instead paired with shock. The same contingency awareness questionnaire was filled out immediately after pairing to confirm that the correct CS-US association was learned. 21

33 2.4 Autonomic Arousal Indices of Conditioned Fear As objective measures of a conditioned fear response, electrodermal activity and heart rate (HR) were measured continuously throughout all pairing and PET scanning session. Electrodermal activity was analyzed as the frequency of skin conductance responses (SCRs), which reflect phasic deflections in the electrical conductivity of the skin. The threshold for detection of a SCR was set at 0.01 microsiemens (µs) OR >3SD of baseline mean. SCR amplitudes were calculated by subtracting the mean skin conductance level 2s before CS onset from the peak value in the SCR waveform during the CS trial period (95). This period included the duration of the CS, 20-second countdown and blank screen in which a shock was expected, i.e. the entire CS-US interval (30s) was analyzed (96). Plasma cortisol levels were also sampled periodically during all PET scans (6 samples / scan, as illustrated in Fig.1), through the same catheter set up for tracer injection in the left arm vein. 2.5 Apparatus All visual stimuli were presented in video glasses (EVG920D Video Eyewear; 640x480 resolution, with a virtual display equivalent to 80 at 1m with a 35 viewing angle). The presentation of visual and electrical stimuli was programmed using SuperLab 4.5 stimulus presentation software. AcqKnowledge software and BIOPAC Systems equipment (MP100, Biopac Systems, Goleta, CA) were used to record SCR and HR, and to synchronize physiological measurements with stimulus presentations. For SCR measurements, BIOPAC EL507 Ag/AgCl disposable electrodermal electrodes with 0.5% chloride salt electrolyte gel were set up on the middle 22

34 phalanges of the right hand index and middle fingers. Two BIOPAC EL503 Ag/AgCl general purpose electrodes with 7% chloride salt electrolyte gel were set up on the left and right sides of the chest for heart rate measurements. The stimulating bar electrode (Biopac convex unshielded bar electrode EL351, with 2 tin electrodes spaced 30mm apart) was secured to the inner surface of the participant s left (nondominant) wrist, over the ulnar nerve, with electrode gel applied for contact with the skin. The electrode leads were attached to a Biopac Constant Voltage Stimulator Unipolar Pulse (STM200) for the delivery of 50 ms electric pulses. 2.6 PET and MRI Participants were asked to refrain from consuming alcohol or drugs/medication for 1 day before scanning, caffeine for 4 hours before scanning, food or water for 1 hour before scanning, and social drugs for at least 1 week before scanning. Additionally, participants were asked to wake up at least 3 hours prior to scanning, and to refrain from physical activity / exercise for 1 hour before scanning. Prior to each PET scan, a urine toxicology screen for illicit drugs of abuse (Triage, Biosite Diagnostics, San Diego, CA) was performed, and a urine pregnancy test was performed in women. Subjective ratings of mood, anxiety and alertness were assessed before scanning and twice during the scan (administered verbally 30 and 150 minutes into scanning), using the Profile of Mood States (POMS) (97), state-trait anxiety inventory (STAI)-State (88) and visual analog scale (VAS) of alertness (98). 23

35 All PET scans were conducted using a Siemens high-resolution research tomograph (HRRT; CTI/Siemens; Knoxville, Tennessee) dedicated brain scanner, and began late morning/early afternoon (between 11AM and 2PM). Scans consisted of a 6-minute transmission scan for attenuation correction, followed by a 90-minute dynamic acquisition scan, then a 30-minute break, and a final 60 minutes of dynamic acquisition scanning (total scan duration of 180 minutes). It has been shown that 180-minute scan durations are sufficient to reach transient equilibrium in both striatal and extrastriatal regions (99). Prior to dynamic scanning, participants were administered a bolus injection of 18 F-Fallypride (range = mci; PET1 = 4.7 ± 0.3 mci; PET2 = 4.6 ± 0.4 mci; PET3 = 4.7 ± 0.4 mci) through an i.v. catheter placed in the left arm vein. The spatial resolution of the scans was between 2.3 and 3.4mm full width at half maximum (FWHM), and dynamic scanning consisted of the following sequence of frame durations: 3 10s, 5 30s, 4 60s, 4 120s, 5 300s, 5 600s, a 30-minute break, and 6 600s. Participants also underwent a 9-minute anatomical T1-weighted MRI scan for PET/MR coregistration. MRI scans were performed using a 3T Siemens TIM Trio Magnetom scanner equipped with a 32-channel head coil (Erlangen, Germany). The MRI measurement protocol used was the ADNI 3D MP-RAGE protocol (TR = 2300ms; TE = 2.98ms; flip angle = 9 ; voxel size = mm). 2.7 PET and MRI Data Processing PET data was reconstructed using an MLEM (maximum-likelihood expectation maximization) iterative reconstruction algorithm (100) that corrects for scattered and random coincidences, attenuation, and detector non-uniformities. 24

36 PET data were motion corrected for head movement / repositioning errors using an automated algorithm-based frame realignment method (101). The voxel-wise non-displaceable binding potential (BP ND ) map was created using the Simplified Reference Tissue Model (SRTM) (102) with the basis functions method optimized for 18 F-Fallypride from 11 C-Raclopride studies (103). The gray matter of the cerebellum, excluding the vermis, was used as a reference region in order to account for non-specific binding, since it is devoid of D2/3 receptors (78, 102). Next, using a linear transformation, the 6 th and 11 th frames of the PET emission data were summed and co-registered to each participant s anatomical MRI scan. Each participant s MRI scan was then transformed into ICBM152 space, and using the resulting transformation parameters, the co-registered BP ND map was also transformed into MNI152 space (104). In order to reduce the effect of anatomical variations, the BP ND map was then smoothed using a 6-mm Gaussian filter. Lastly, regions of interest (ROIs) were defined in the left and right amygdala, anterior hippocampus, nucleus accumbens, medial prefrontal cortex, ventromedial prefrontal cortex and ventral tegmental area. ROI masks were created with the Wake Forest University (WFU) PickAtlas toolbox (105) for SPM12, using the Automated Anatomical Labeling (AAL) atlas (right and left amygdala, hippocampus, superior medial frontal cortex, medial orbital frontal cortex, anterior cingulate cortex and insula) (106), and the IBASPM 71 library (right and left nucleus accumbens) ( Additionally, since the PickAtlas toolbox does not include a mask for the ventral tegmental area (VTA), the VTA atlas from the Adcock lab at Duke University was used (107). Since the ventral hippocampus in rodents, corresponding to the anterior hippocampus in humans, connects more densely to the amygdala (108, 109), receives stronger dopaminergic projections from the ventral tegmental area 25

37 (110), and has been more widely implicated in fear conditioning, as compared to the dorsal/posterior hippocampus, the relatively large automatically-segmented hippocampal ROI mask was manually reduced to include only an anterior portion of the hippocampus. All ROIs were checked against the MRI scans of each participant and adjusted manually if necessary. Mean and maximum BP ND values were calculated from the BP ND map for each region of interest. An example of the ROI mask is shown in Figure 2. Figure 2. An axial slice demonstrating a portion of the ROI mask used for the a priori hypothesis-based analyses, overlaid on an average T1 MRI of all participants. Shown here: the anterior hippocampus (left, pink; right, dark blue), amygdala (left, light blue; right, green), nucleus accumbens (left, white; right, red), and ventromedial prefrontal cortex (left, yellow; right, orange). 2.8 Statistical Analyses POMS scores on 6 bipolar scales (elated-depressed, composed-anxious, energetic-tired, agreeable-hostile, confident-unsure, clearheaded-confused) were transformed into population normalized t scores. 26

38 Exploratory, brain-wide comparisons were performed with minc tools (glim_image) using a height threshold of p < 0.001, uncorrected for multiple comparisons. Voxel-wise paired t-tests were performed comparing PET1 with PET2, PET1 with PET3, and PET2 with PET3. In order to reduce the number of comparisons, a mask was applied including only voxels with BP ND > 0.6 and t > 4 (voxel-wise one-sample t-tests comparing BP ND at each voxel to a BP ND of 0.6) (see Figure 3). Figure 3. The mask applied for voxel-wise comparisons comparing BP ND between PET scans (BP ND > 0.6, t > 4). A priori hypothesis-driven ROI analyses were also performed, using IBM SPSS Statistics Version 23 for Macintosh. For both mean and maximum BP ND values, a three-way repeated measures ANOVA was performed (Scan, 3 levels; Hemisphere, 2 levels; Region, 6 levels), followed by planned pairwise comparisons (two-tailed paired t-test, or Wilcoxon Signed-Rank Test in the case of a non-gaussian distribution of the difference scores, as determined by the Shapiro-Wilk test). Additionally, Pearson correlations were performed between the change in regional BP ND values and the change in the rate of SCRs between PET scans, and in the case of non-normally distributed difference scores, Spearman s rank correlations were run. 27

39 RESULTS 3.1 Participant Characteristics 15 healthy right-handed volunteers participated in the study. Three female participants did not complete the study; 1 participant withdrew due to a claustrophobic experience in the MRI scanner, 1 participant was excluded due to head pain during PET1, and 1 participant withdrew after PET1 for unknown reasons. Therefore, a total of 12 participants completed the study (6F/6M; mean ± SD age = 24.1 ± 3.7 years, range = years). 3.2 Subjective Ratings Following both CS-US pairing sessions, as well as immediately before PET2 and PET3, all participants reported the correct CS-US association. All participants reported a little to moderate anxiety related to the CS+ prior to PET2, and by contrast, no anxiety associated with the same shape (new CS-) prior to PET3. The CS-associated anxiety scores associated with the CS presented during scanning differed significantly between PET2 and PET3 (paired two-tailed t-test: t 10 = 9.8, p < ). The average shock level that was subjectively rated to be at pain threshold was 35.4V (range = 19 58V). 3.3 Autonomic Arousal Indices of Conditioned Fear Although SCRs were observed to gradually diminish over the course of CS presentations in PET2 in most participants, it has been shown that SCRs habituate over time in response to stimuli (111), despite subjective contingency ratings remaining high. In order to account for this 28

40 habituation, the frequency of SCRs was recorded from the first 10 CS presentations during scanning. The percentage frequency of SCRs was significantly higher in PET2 than in PET3 (2-tailed paired t-test, t 11 = 3.41, p = 0.006), and significantly higher than the frequency of non-specific SCRs (SCRs that occur in the absence of identifiable stimuli) occurring during the same time intervals in PET1 (2-tailed paired t-test, t 11 = -4.13, p = 0.002) (Figure 4). There was no significant difference in the frequency of SCRs between PET1 and PET3 (2-tailed paired t-test, t 11 = -1.08, p = 0.3). The difference scores of SCR frequencies between PET scans were normally distributed, as determined by the Shapiro-Wilk test (p > 0.05). An example of the observed changes in skin conductance over time in PET2 and PET3 for one participant is illustrated in Figure 5. Figure 4. Mean percentage frequency of skin conductance responses (SCRs) in each PET scan. Error bars reflect standard error of the mean. 29

41 SCR (Microsiemens) Shock Expected PET 2 Shock Expected SCR (Microsiemens) PET Time (s) Figure 5. An example from one participant of skin conductance response (microsiemens) over time during PET2 and PET3 (the expression and inhibition of conditioned fear, respectively). Each trial is 30s: the triangle indicates the onset of the 3s conditioned stimulus, which is followed by a 20s countdown, and a blank screen during which the participant either does (PET2) or does not (PET3) expect a shock. 3.4 Brain-wide, Voxel-wise Analyses Exploratory analyses revealed decreases in BP ND, which reflect increases in dopamine release, in the bilateral anterior hippocampus and left amygdala in PET3 (the inhibition of conditioned fear), as compared to PET1 (baseline) (t > 4.3, p < 0.001, uncorrected for multiple comparisons) (see Figure 6). Additionally, significant decreases in BP ND, reflecting increases in dopamine release, were observed in the bilateral posterior cingulate gyrus in PET3 (the inhibition of conditioned fear), as compared to PET2 (the expression of conditioned fear) (t > 4.3, p < 0.001, 30

42 uncorrected) (Figure 7). Surprisingly, no significant changes in BP ND were observed between PET1 (baseline) and PET2 (the expression of conditioned fear). Figure 6. t-maps in coronal, axial and sagittal views demonstrating decreases in 18 F-Fallypride binding potential in the bilateral anterior hippocampus (the left hippocampal cluster is identified with the white arrow; MNI coordinates of peak: x, y, z = -30, -13, -24), and the left amygdala (identified with the red arrow; x, y, z = -24, 3, -23) in PET3 during the inhibition of conditioned fear, as compared to PET1 during baseline. Coloured t-maps are thresholded for visualization purposes, and overlaid on an average T1 MRI of all participants. 31

43 Figure 7. t-maps in coronal, axial and sagittal views demonstrating decreases in 18 F-Fallypride binding potential in the bilateral posterior cingulate gyrus (identified with the white arrow; x, y, z = -8, -44, 35) in PET3 during the inhibition of conditioned fear, as compared to PET2 during the expression of learned fear. Coloured t-maps are thresholded for visualization purposes, and overlaid on an average T1 MRI of all participants. 3.5 Region of Interest Analyses ROI analyses largely confirmed the voxel-wise analysis findings. Decreased BP ND, reflecting elevated dopamine release, was found in the bilateral hippocampus in PET3 (the inhibition of conditioned fear), as compared to both PET1 (baseline) and PET2 (the expression of conditioned fear). This difference was significant at p < 0.05 between PET1 and PET3 in the left hippocampus, and between PET 2 and PET3 in the right hippocampus, and was otherwise trending, as shown in Figure 8. 32

44 Figure 8. Mean BP ND across PET scans in the left and right anterior hippocampus. Left hippocampus: * p = 0.04 (paired t-test), p = 0.1 (paired t-test); right hippocampus: p = 0.06 (Wilcoxon Signed-Rank Test), * p = 0.02 (paired t-test), uncorrected for multiple comparisons. Error bars represent standard error of the mean. Similarly, in the left amygdala, a significant decrease in BP ND was observed in PET3 relative to PET1 (Wilcoxon Signed-Rank Test, p = 0.03) (Figure 9). A similar decrease was observed in the right amygdala in PET3 relative to PET1, and in the bilateral amygdala in PET3 relative to PET2, however these reductions were not significant (p = 0.1). Notably, although there appears to be a slight decrease in BP ND bilaterally in the amygdala in PET2 compared to PET1, this difference was not significant (p > 0.4), and was driven by a single outlier - when the outlier was removed, there was no apparent reduction in BP ND. 33

45 Figure 9: Mean BP ND across PET scans in the left and right amygdala. Left amygdala: * p = 0.03 (Wilcoxon Signed-Rank Test), p = 0.13 (paired t-test); right amygdala: p = 0.14 (Wilcoxon Signed- Rank Test), p = 0.11 (paired-t-test), uncorrected for multiple comparisons. Error bars represent standard error of the mean. It is important to note that the pairwise comparisons reported here were not corrected for multiple comparisons, however, the reported significant findings were observed bilaterally, and were hypothesized a priori based on the literature, thus reducing the likelihood of false positives. 34

46 Figure 10: Mean BP ND across PET scans in the ventral tegmental area (VTA), ventromedial prefrontal cortex, and left and right medial prefrontal cortex. No significant changes in BP ND between scans were observed. Error bars represent standard error of the mean. Contrary to our hypotheses, no significant changes in BP ND were found in the other ROIs between scans (Figure 10). In particular, no ROIs showed significant changes in dopamine release between PET1 and PET2, and no significant changes in dopamine release were found in the ventromedial prefrontal cortex in PET3 as compared to PET1 and PET2. Furthermore, no significant correlations were observed between the change in SCR frequency and the change in regional BP ND between PET scans (p > 0.1). 35