REVIEW ARTICLE | FOCUS https://doi.org/10.1038/s41593-019-0503-3REVIEW ARTICLE | FOCUS https://doi.org/10.1038/s41593-019-0503-3 in circuits of aversive emotional

Ekaterina Likhtik! !1,2* and Joshua P. Johansen! !3,4*

Emotional learning and memory are functionally and dysfunctionally regulated by the neuromodulatory state of the . While the role of excitatory and inhibitory neural circuits mediating emotional learning and its control have been the focus of much research, we are only now beginning to understand the more diffuse role of neuromodulation in these processes. Recent experimental studies of the , noradrenaline and systems in fear learning and extinction of fear respond- ing provide surprising answers to key questions in neuromodulation. One area of research has revealed how modular organiza- tion, coupled with context-dependent coding modes, allows for flexible brain-wide or targeted neuromodulation. Other work has shown how these neuromodulators act in downstream targets to enhance signal-to-noise ratios and gain, as well as to bind distributed circuits through neuronal oscillations. These studies elucidate how different neuromodulatory systems regulate aversive emotional processing and reveal fundamental principles of neuromodulatory function.

euromodulation shapes emotional learning in distributed of signal-to-noise dynamics and context-dependent coding coor- circuits across the brain. Studies using tracing, lesions, dinated by modular organization of neuromodulatory nuclei. We Nrecordings and pharmacology have provided important describe how these neuromodulatory functions are fundamentally information about projection pathways and neuromodulator con- different from neuro-regulation: the latter acts via feedback loops to tributions to shifting behavioral state during learning1–14. This field maintain , whereas neuromodulators are geared more continues to benefit from genetic and molecular developments in toward shifting circuits to a new baseline during learning. cell-specific circuit techniques that build on the existing foundation to create a more detailed analysis of how neuromodulatory systems Roles of Ach in emotional learning connect with and influence different cell types at target structures Acetylcholine modulates , and hip- during learning. Although a wide array of and pocampus in emotional learning. Ach is a critical neurotransmit- neuropeptides can act as neuromodulators, here we focus on ace- ter for several functions in the central nervous system, including tylcholine (Ach), noradrenaline (NA) and dopamine (DA) in setting sleep–wake rhythms, attention, cue detection, working and mammals, and we highlight their actions in sculpting circuit-level spatial memory1,20–23. Ach signals through nicotinic (ionotropic) processing to change fear learning, extinction and fear-discrimina- and muscarinic (Gq- or Gi-coupled) receptors. There are two main tion learning. Although not discussed here, other neuromodulatory cholinergic centers that innervate the brain: the basal forebrain systems, such as , also exhibit some degree of projection (BF) and the pedunculopontine and laterodorsal tegmental nuclei specificity and regulate aversive emotional learning15,16. We high- (Fig. 1)3,23–26. The role of BF cholinergic projections in process- light recent findings in fear-related functions (for example, learned ing emotional learning and memory has been mostly extensively passive Pavlovian defensive responses, as opposed to other types of studied. The BF is an umbrella term for a collection of telencepha- aversive learning or conscious emotional processes17), as a great deal lon nuclei that include the septum, the of Meynert of progress has been made in understanding general principles of (NBM), the (SI), the ventral pallidum (VP) neuromodulation using these behavioral approaches, and previous and the diagonal band2,3. Notably, the BF contains more GABAergic reviews have covered classical lesion and pharmacological find- and than cholinergic neurons27,28. All three ings18–20. This emerging literature on neuromodulation of fear learn- cell types project to the cortical mantle, including medial prefron- ing and extinction reveals fundamental features of Ach, NA and DA tal cortex (mPFC), and to subcortical regions that partake in fear that are applicable to how they generally regulate brain function. and extinction learning5,29–33. BF innervation is roughly organized We first discuss how Ach sharpens the signal-to-noise ratio, cue along the anterior–posterior axis34,35 and has modular connectiv- processing and communication in the amygdala, cortex and hip- ity: anatomically distributed cholinergic BF cell groups innervate pocampus during emotional learning. Next, we examine how NA targets that contribute to different aspects of emotional learning. exerts its effects on target microcircuitry and molecular signaling During fear conditioning, neural spiking in the dorsal during fear and extinction learning. We then discuss recent work (dHPC) coincides with context encoding, in the lateral and basolat- demonstrating how the anatomical organization of the locus coe- eral amygdala (LA/B) it coincides with encoding the conditioned ruleus–NA system allows it to flexibly control the balance between cue, and in the mPFC it coincides with both cue and context36,37. these opposing states at anatomically distributed sites. Finally, we The modular pattern of BF cholinergic innervation of these regions discuss the role of the DA system in fear learning and examine how allows for its multipronged impact on emotional learning, whereby it acts as a detector of when fear responding is no longer adaptive in the amygdala, cholinergic input modulates cue-based aversive to switch behavioral strategies to extinction. Overall, we identify learning; in the dHPC, spatial processing; and in the mPFC, cue several mechanisms by which neuromodulators adjust circuit-level encoding and consolidation of extinction (Fig. 2). Accordingly, opto- communication during emotional learning, including synchro- genetic activation of cholinergic BF afferents to the BLA enhances nization of distal areas, gain control of sensory stimuli, shifting acquisition and retention of cued fear38, whereas inactivation

1Biology Department, Hunter College, City University of New York, New York, NY, USA. 2The Graduate Center, City University of New York, New York, NY, USA. 3RIKEN Center for Brain Science, Laboratory for Neural Circuitry of Learning & Memory, Wako, Japan. 4Department of Life Sciences, Graduate School of Arts and Sciences, University of Tokyo, Tokyo, Japan. *e-mail: [email protected]; [email protected]

1586 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE

Glu PV+ Ach Ctx

HPC Basal forebrain

mPFC PAG Str ppn Ch1/2 module θ HPC LT (MS/DBB) LC Context encoding VTA and SN BF NTS in fear conditioning and extinction Amy PT Ch4 module θ? (VP/SI, NBM) Cue discrimination; NA DA Ach cue encoding in LA/B fear conditioning ? Fig. 1 | Neuromodulatory projections overlap at their target sites. θ , Regions containing cell bodies that synthesize the neuromodulatory θ γ θ–γ neurotransmitters NA (solid orange), DA (solid red), and Ach (solid yellow) Wake are shown in a sagittal cross-section of the adult mouse brain. Regions that partake in aversive emotional learning and are innervated by these Ach Glu neuromodulatory systems are shown in the ovals, which mPFC denote the strength of innervation by the corresponding neuromodulatory PV+ system. Multiple systems innervate each region with different strength. Amy, amygdala; Ctx, cortex (sensory); HPC, hippocampus; LT, laterodorsal θ, theta oscillation tegmental nucleus; ppn, pedunculopontine nucleus; PT, pontine γ, gamma oscillation – , theta–gamma coupling ; Str, . θ γ

Fig. 2 | Basal forebrain connectivity with the aversive emotional learning network. Cholinergic (Ach, yellow), GABAergic from PV+ cells (blue) of muscarinic cholinergic receptors in the dHPC before contextual and glutamatergic (Glu, green) cells in the medial septum–diagonal fear conditioning impairs recall of the fear-associated context but limb of the diagonal band (MS/DBB)—also known as the cholinergic does not freezing to the cue (Fig. 3)39. group 1/2 (Ch1/2) region of the BF—innervate the hippocampus (HPC) Cholinergic signaling also plays an important role in encoding and contribute to the generation and pacing of HPC theta oscillations. spatial information during fear extinction, and lesions of BF cho- Cholinergic and GABAergic projections from the MS/DBB to the HPC linergic cells impair extinction acquisition40. Furthermore, block- also contribute to contextual fear conditioning and extinction. There is ing Ach muscarinic receptors before extinction training eliminated a separate module of cholinergic, glutamatergic and GABAergic cells in fear renewal in a nonextinguished context, suggesting that down- the cholinergic group 4 region (Ch4) of BF, which is comprised of the regulating cholinergic signaling during extinction training blocks ventral pallidum, substantia innominata (VP/SI) and the nucleus basalis hippocampus-dependent encoding of contextual information, of Meynert (NBM). The Ch4 region projects to the LA/B and mPFC. rendering extinction less context-dependent (Fig. 3)41. Similarly, Cholinergic projections to the LA/B enhance neural activity during BF-mediated cholinergic signaling in the mPFC plays a role in con- cued fear conditioning. The role of PV+ GABAergic and glutamatergic solidation of extinction learning: deleting the p75 neurotrophin projections to the LA/B is less well studied, and it is not yet known receptor from BF cholinergic cells altered BF connectivity with the whether BF projections to the LA/B contribute to theta generation in mPFC, impaired extinction consolidation and diminished activ- the amygdala during fear conditioning. Ach and PV+ projections to the ity in the infralimbic cortex (IL) of the mPFC, an area known for mPFC upregulate cue discrimination during attention-based tasks and its role in extinction42. Thus, cholinergic innervation from the BF contribute to prefrontal theta and gamma oscillations and to theta–gamma modulates emotional learning circuitry at multiple nodes, and coupling during cue discrimination. How these inputs and glutamatergic depending on the time and target site of Ach release, cholinergic inputs contribute to aversive learning and extinction is not yet well input shapes responses to aversive cues, extinguished cues and/or understood. Likewise, the role of the BF in mPFC–LA/B or mPFC–HPC to the context of such cues (Fig. 2). synchronization during cued or contextual fear conditioning, respectively, is not well understood (black broken lines). Inset: local interactions in Acetylcholine modulates circuit-level communication in emo- the BF. Cholinergic activation of PV+ GABAergic cells increases cortical tional learning. Generation of theta oscillations (4–12 Hz), which desynchronization and promotes the awake state. Glutamatergic activation largely reflect pacing of inputs but also local firing, has been stud- of cholinergic cells and PV+ GABAergic cells also promotes the awake state. ied extensively in the dHPC. Cellular activity in the dHPC is orga- nized by theta oscillations during spatial exploration and contextual fear memory retrieval, when oscillations synchronize neural activ- The interactions between BF cell types are not yet well under- ity across the rostrocaudal span of the dHPC and between dHPC stood; however, their dynamics prominently contribute to the and distal sites, such as the ventral hippocampus and the mPFC43,44. aroused state, which in turn modulates learning. One notable During spatial encoding, cholinergic group 1/2 and GABAergic microcircuit in the BF is cholinergic excitation of local parvalbu- medial septum and diagonal limb of the diagonal band (MS/DBB) min-expressing (PV+) GABAergic neurons22, cells that promote projections to the dHPC, along with entorhinal inputs and the arousal and project to the LA/B, the dHPC and the mPFC6,22,29,53–56. internal dynamics of hippocampal circuitry, pace dHPC neurons GABAergic projections from the BF target GABAergic and gluta- to oscillate at the theta frequency7,36,45–50. Likewise, theta oscillations matergic cells in dHPC, LA/B and mPFC, constituting a critical coordinate BF neurons during a spatial attention task51,52, suggest- component for pacing downstream membrane oscillations that ing that BF theta oscillations organize BF activity during different arise during cognition29,30,50,55,57,58. Furthermore, BF recordings cognitive demands. show a variety of population firing dynamics that are modulated

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1587 REVIEW ARTICLE | FOCUS NATURE NEUROSCIENCE

a Cholinergic receptors HPC LA/B mPFC

nAchR ear rn α imination fe

Ionotropic α

mAchR ear , imination iz Metabotropic

b i Network modulation iii Cellular modulation wn InhibitoAch e Ach I I

Vm Vm

rpiz ii Microcircuit modulation Ach Ach nAchR mAchR mAchR nAchR

Vmw Vm Inter Inter

Fig. 3 | Acetylcholine affects emotional learning via nicotinic and muscarinic receptors that enhance signal-to-noise ratios across regions. a, A summary of how nicotinic (ionotropic) and muscarinic (metabotropic) cholinergic receptors are involved in different aspects of aversive emotional learning. The contributions of different receptor types to fear conditioning, extinction and discrimination learning are shown. Receptor contributions are based on primary research and reviews, as referred to in the text. b, Mechanisms for cholinergic mediation of signal-to-noise ratio. At the network level (i), in low cholinergic states, cortical neurons have synchronized membrane potentials whereby they fluctuate between Up and Down states that are characteristic of sleep. Cortical Ach release stabilizes cortical membrane potentials into a unimodal state, which allows for more efficient signal processing. At the microcircuit level (ii), Ach drives a disinhibitory circuit in the visual and primary auditory cortices, whereby GABAergic interneurons (black circles) located in L1 are depolarized via nAchR and inhibit GABAergic interneurons located in L2/3, thereby disinhibiting pyramidal cell firing (red triangle). The Ach-mediated disinhibition in auditory cortex is necessary for fear conditioning. (iii) Top: at the cellular level, BF Ach input to cortical and amygdala cells shifts the excitatory–inhibitory balance toward excitation. Bottom: amygdala neurons are inhibited by Ach when they are at rest, but become excited when they are more depolarized via a combination of nicotinic and muscarinic receptor activation. Prolonged after-depolarizing potentials increase long- term potentiation of cortical input to the amygdala. E, excitatory; HPC, hippocampus; I, inhibitory; E-I, excitatory–inhibitory balance; mAChR, muscarinic acetylcholine receptor; nAChR, nicotinic acetylcholine receptor; L1, cortical layer 1; L2/3, cortical layers 2 and 3; Vm, membrane potential. by selective-attention tasks, indicating that local communication in the dHPC and from the VP–SI–NBM to pyramidal and GABAergic this region contributes to cognitive performance51, suggesting an neurons in the LA/B8,31,64, where activation of BF cholinergic ter- important area for further investigation. minals enhances oscillatory fluctuation in membrane potentials When animals show defensive freezing during aversive cues, in vitro and theta power in vivo24. exploration-related theta in the dHPC is low, but there are promi- Recordings in vitro and in vivo show that BF VP–SI–NBM cells nent aversive-cue-evoked lower-frequency theta oscillations in the discharge in high-frequency bursts at the theta frequency (peak 6–7 LA/B and mPFC (peak 4–7 Hz)59–62. Critically, defensive freezing Hz) and that cholinergic inputs to the mPFC are associated with doesn’t modulate dHPC theta in this range59, and silencing the MS/ increased prefrontal theta and gamma power during cue detec- DBB impairs dHPC movement-related theta but leaves defensive tion55,65,66. Given that detection of rewarding cues is modulated by freezing and aversive-cue-driven mPFC theta intact63. This leaves a combination of cholinergic and non-cholinergic BF inputs to the open the question whether a parallel projection from the VP–SI– mPFC, it will be important to determine whether these projections NBM cell group could pace aversive-cue evoked theta oscillations in are also active during differential fear conditioning, when animals the mPFC and/or LA/B (Fig. 2). Much of the key cholinergic group learn to discriminate aversive from non-aversive cues. Notably, 1/2–dHPC connectivity that paces hippocampal theta is similar in parallel to detection of rewarding cues, mPFC gamma power also the VP–SI–NBM–LA/B module. For example, PV+ neurons project increases during discrimination of non-aversive cues, whereas from the MS/DBB and the VP–SI–NBM to GABAergic cells in the theta–gamma coupling in the mPFC and in the LA/B is enhanced dHPC and the LA/B, respectively6,30,49,53. Further, cholinergic cells during detection of aversive cues59,65,67. However, it remains unknown project from the MS/DBB to pyramidal and GABAergic neurons in whether cholinergic BF inputs mediate these changes. Likewise, the

1588 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE potential synchronizing role of the BF in mPFC–LA/B communica- during aversive discrimination learning remains to be explored, its tion during emotional learning needs further investigation. Human role in cue discrimination overall suggests that it is likely to play a imaging data demonstrate that when the BF is active, functional role. Overall, Ach amplifies cue encoding during increased atten- mPFC–LA/B connectivity increases68, suggesting that the BF may tion at sensory cortical sites that communicate with the mPFC76,78 synchronize these regions. However, the mPFC–LA/B is also a and by acting at postsynaptic mPFC targets directly65,66. The effect of highly reciprocally connected circuit, in which communication is Ach on the mPFC during aversive learning and cue discrimination, key for aversive and extinction learning. Additionally, the mPFC when attention is high, will be an important area for further study. is reciprocally connected with the BF, and prefrontal inactivation decreases Ach release at other cortical sites during sensory stimu- Roles of noradrenaline in emotional learning lation69, indicating that bidirectional mPFC–BF and mPFC–LA/B The noradrenergic system is comprised of a number of brain- communication constitutes an important pathway for learning. stem nuclei, including the (LC), pontine teg- mentum and nucleus tractus solitarius (NTS; Fig. 1) and signals Acetylcholine upregulates the signal-to-noise ratio in amygdala through β-adrenergic (β-AR, Gs-coupled), α1-adrenergic (α1-AR, and cortex. Ach enhances cue detection across multiple regions via Gq-coupled) and α2-adrenergic (α2-AR, Gi-coupled) receptors. NA a combination of inhibitory and excitatory currents that amplify sig- was traditionally thought of as an arousal system79,80, but new evi- nal-to-noise ratios. In the amygdala, in vitro recordings show that dence suggests that it serves a variety of functions due to its anatom- at resting membrane potentials, cholinergic input strongly inhib- ical organization and flexible, context-dependent coding modes81–84. its principal neurons via a combination of feedforward inhibition In this section, we will illustrate the multifunctional aspects of NA and direct postsynaptic hyperpolarization via muscarinic receptors in fear learning and extinction, which arise through the anatomical (Fig. 3)24,33,70. However, when amygdala principal cells are depolar- organization of a prominent NA center, the LC. ized, as during fear conditioning when sensory inputs drive the LA/B, cholinergic input enhances firing and elongates after depolar- Noradrenergic regulation of amygdala during emotional learn- izing potentials for multiple seconds at a time33,38. These prolonged ing and memory. Aversive experiences produce a surge in NA levels depolarizing currents create the right conditions for associative in the brain, including the amygdala85. This promotes aversive asso- synaptic plasticity (Figs. 2 and 3). Accordingly, stimulation of BF ciative learning and memory through activation of β-AR signaling cholinergic inputs to the amygdala strengthens long-term potentia- in BLA86,87. β-AR activation in BLA neurons enhances excitability tion of cortical input onto LA/B pyramidal neurons38, and blocking through recruitment of protein kinase A (PKA) and subsequent muscarinic receptors in the LA/B impairs trace fear conditioning, downregulation of SK channels, which normally reduce excitabil- a paradigm in which the cue and aversive outcome are separated ity, at excitatory synapses (Fig. 4)88. Furthermore, NA boosts syn- by several seconds71. Notably, stimulation of cholinergic BF inputs aptic plasticity in LA/B pyramidal cells by reducing excitability of amplifies signatures of amygdala physiology in the presence of fear GABAergic interneurons89. In behavioral studies, β-AR activation conditioned stimuli, such as increased LA/B low-frequency theta is necessary during, but not after, auditory fear learning, modu- oscillations24 and responses of LA/B principal neurons to theta- lating both its acquisition and consolidation through facilitation frequency stimulation of cortical afferents38,59,61,67. of Hebbian plasticity mechanisms and AMPA receptor insertion Aversive learning occurs during an enhanced state of arousal that as well as recruitment of MAP kinase86,87. Conversely, for contex- shifts sensory cortical processing, alters sensory perceptual thresh- tual aversive learning or for emotional modulation of non-aversive olds and widens stimulus generalization72. The mPFC receives input forms of spatial memory, NA activity and its interaction with cir- about the aversive cue from the visual and auditory cortices that culating corticosterone is required specifically during the memory process it first. The role of Ach in shifting cortical sensory process- consolidation period occurring after learning90,91. ing during increased attention is due to a convergence of neural Anatomical studies indicate that the amygdala is innervated by mechanisms that include disinhibition and shifting signal-to-noise noradrenergic neurons in the LC, and NTS, any dynamics. For example, in the visual cortex, Ach strengthens sig- of which could drive adrenergic receptor activity in BLA or regu- nal transmission relative to background activity by shifting neural late the central amygdala (CeA; Fig. 1)92,93. However, recent studies membrane potentials away from the synchronous up and down using optogenetic manipulations demonstrated the importance of oscillations characteristic of sleep toward a unimodal low-amplitude LC projections to the BLA in aversive learning and behavior84,94,95. steady state (Fig. 3)73,74. This is proposed to be a population-level Specifically, inhibiting the activity of a distinct population of LC effect that underlies the cholinergic role in enhancement of visual neurons that project to the BLA or inhibiting their synaptic ter- cue processing during attention. Cortical Ach release also enhances minals in the BLA during the shock period of auditory fear condi- the presence of excitatory relative to inhibitory postsynaptic cur- tioning reduces learning84. However, stimulating LC inputs to the rents and disinhibits pyramidal cells during sensory stimulation. In BLA enhances fear learning and anxiety84,94. These findings suggest the primary auditory cortex, a short period of BF stimulation paired a linear contribution of the LC-to-amygdala NA pathway in aver- with a particular frequency tone can disinhibit pyramidal neural sive learning. Furthermore, LC neurons are phasically activated by response and shift preferred frequency firing toward the paired tone aversive stimuli, and manipulations of LC neurons or their inputs for up to several hours75. Likewise, during fear conditioning, cholin- to the BLA change amygdala neuronal firing rates84,94,96. While this ergic activation of the primary auditory cortex disinhibits cortical suggests that aversive-stimulus-evoked LC engagement modulates pyramidal cells during tone–shock associations, a process that was LA/B to enhance fear learning, it is also possible that this occurs shown to be necessary for fear conditioning (Fig. 3)76. through more persistent NA activity, occurring through sustained Cholinergic modulation of the mPFC has been primarily studied evoked volume transmission in amygdala or heightened basal fir- during tasks that require attention to cues during decision mak- ing rates of LC neurons. Together, these studies suggest that dur- ing, when Ach release drives a combination of nicotinic and mus- ing aversive experiences, LC neurons release NA in the amygdala, carinic receptor-mediated responses that enhance signal-to-noise which binds to β-ARs there, enhancing excitatory signaling and ratios, theta–gamma coupling and cue discrimination (Fig. 2)65,66,77. neuronal plasticity, thereby modulating aversive memory acquisi- Notably, mPFC theta power also increases when animals discrimi- tion and consolidation mechanisms (Fig. 4). nate an aversive cue during differential fear conditioning, whereas Different LC projections, as well as other noradrenergic nuclei, mPFC gamma power increases when animals discriminate the non- also participate in aversive processes. LC projects strongly to the aversive cue59,61,67. Thus, although cholinergic impact on the mPFC hippocampus (Fig. 4), where it can release both NA and DA, both of

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1589 REVIEW ARTICLE | FOCUS NATURE NEUROSCIENCE

abFear conditioning Early extinction, high fear c Later extinction, lower fear

mPFC mPFC mPFC CeA CeA CeA

LA/B LA/B LA/B Extinction

LC LC LC

Strong aversive Aversive tone Extinguished tone experience (shock) (early extinction) (late extinction)

α1-AR β-AR β-AR α1-AR β-AR β-AR α1-AR β-AR β-AR

Opposes Enhances Opposes Opposes Enhances extinction fear learning extinction extinction extinction mPFC LA/B mPFC LA/B mPFC LA/B

Fig. 4 | Context-dependent control of fear learning and extinction by the LC. a, During fear conditioning, most LC neurons are strongly activated (inset: example spike rasters during shock period (gray bar)), causing high NA release (red dots) in amygdala and mPFC. Through activation of β-ARs in the BLA, this enhances amygdala function and fear learning and impairs the long-term retention of extinction. Higher NA levels in the mPFC impair the retention of fear extinction through activation of α1-ARs, possibly on GABAergic interneurons. b, During early extinction, amygdala-projecting LC neurons are mildly activated (inset: example spike rasters during tone period), which initially restrains extinction through the activation of β-ARs on BLA pyramidal cells. c, During later extinction, some mPFC-projecting LC cells are moderately activated (inset: example spike rasters during tone period) and enhance mPFC function by activating β-ARs on mPFC pyramidal cells, thereby facilitating extinction. Red triangles depict fear-learning-promoting, extinction-opposing cells; blue triangles depict extinction-promoting cells; green ovals depict interneurons.

which are important for contextual and spatial forms of learning97,98. is supported by findings showing that stimulation of LC inputs Furthermore, pharmacological stimulation of NTS enhances aver- to BLA enhances fear learning and that NA receptor activation sive memory consolidation as well as NA levels in amygdala99. increases excitatory signaling in BLA pyramidal neurons84,87,88,94. In addition to responding to primary aversive events, LC neu- Another possibility is that β-AR activity helps to select the BLA cell rons exhibit enhancement of sensory cue responsiveness following populations that encode a particular fear memory. Memory allo- fear conditioning100, and increases in NA are evident in amygdala cation studies report that CREB phosphorylation and excitability following fear conditioning85. Consistent with a pro-fear function of levels, both of which are modulated by β-AR signaling, control the LC-to-amygdala projections, inhibition of LC NA terminals in the selection of cells that participate in memory encoding106. Thus, amygdala facilitates fear extinction, while their excitation enhances higher expression levels or sensitivity of β-AR signaling in certain fear learning and responding84,95. Relatedly, extinction retention is BLA subcircuits could bias learning-induced plasticity to distinct impaired if extinction training occurs just after fear conditioning, LA/B cell populations. Possibly related to this, NA activity in BLA and this immediate extinction deficit is blocked if β-AR antagonists could set an aversive network context that modulates other inputs 101 are injected into the BLA . Further, blockade of α1-ARs in amyg- and biases BLA ensembles toward encoding aversive learning. dala before fear conditioning enhances subsequent extinction102. In Determining the effect of NA signaling on BLA network compu- addition to effects on fear extinction, β-AR activation in BLA is also tations and subnetwork cellular selection will be important future required for fear memory reconsolidation and for enhancing the research directions. expression of fear-related behaviors95,103. Together, these findings suggest that in addition to facilitating fear learning, NA release in Noradrenergic regulation of mPFC during extinction of aver- the amygdala in response to aversive cues opposes extinction learn- sive memories. Contrasting with the pro-aversive learning and ing and enables fear-memory reconsolidation (Fig. 4). However, memory function of NA in the amygdala, NA in the IL regulates it is still possible that subtler variations in NA levels could facili- fear extinction learning. For example, pharmacological blockade of tate extinction learning in some conditions. It will be important in β-AR signaling in IL before fear extinction learning reduces reten- future studies to determine the circuit and molecular mechanisms tion of extinction 24 h later107. One possible mechanism for this is occurring in specific amygdala cell types underlying these effects altered IL neuronal excitability, which is enhanced in IL neurons and to use newly developed genetically encoded molecular sensors by β-AR activation. The IL receives NA innervation from a specific to monitor NA activity during extinction learning104,105. population of LC neurons, which also project to other mPFC sub- There are several potential consequences of NA release on BLA regions but have limited collateralization to other brain regions84. network encoding of fear learning and memory. One possibility is Inhibition of these IL-projecting LC neurons or their terminals in that it serves a purely gain-control function by signaling the inten- the IL reduces fear extinction84. However, unlike the linear gain- sity of a threat and regulating the strength of fear learning. This control function of LC NA in the amygdala, the effects of NA in

1590 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE

Global coding mode Discrete coding mode

Broadly Broadly projecting projecting

Specific Specific connectivity connectivity

LC LC

Specific Specific connectivity connectivity

Broadly Broadly projecting Broad projecting Broad connectivity connectivity

Specific Specific connectivity connectivity

Broad Broad connectivity connectivity State-dependent State-dependent global modulator global modulator

Fig. 5 | Potential network architecture for state-dependent modular coding in LC. Distinct coding modes result from a mixture of specific and distributed afferent connectivity to different classes of LC neurons defined by their efferent projections (‘broadly projecting’ to many brain regions versus those exhibiting more ‘specific connectivity’ with distinct efferent brain region targets). In certain conditions, such as under arousal, , strong aversive stimuli or task disengagement, this allows for a ‘global coding mode” (left) driven both by coordinated activity in extrinsic inputs that are broadly connected with all LC cell types and by neuromodulators (the ‘state-dependent global modulator’ in the figure, for example, corticotropin-releasing factor) that coordinate LC subnetworks and enhance global phasic or tonic changes in LC activity. Though not shown here, local pericoeruleus GABAergic networks that modulate LC can add another layer of input processing. A ‘discrete coding mode’ (right) could result from more selective activity in inputs that are specifically connected with LC cells that themselves have specific efferent connectivity. This could be coupled with modular gap-junction coupling within specifically connected cell populations or with modular effects of local NA release (not shown).

the mPFC on fear-extinction learning (and other PFC-dependent Context-dependent modular LC coding for global and specific functions) vary depending on NA concentration and the activated neuromodulation. Delineating how LC NA operates in specific receptor subtypes108. Accordingly, blocking β-AR activity reduces target brain regions has provided valuable insights into how it can normal fear extinction107. Conversely, extinction is impaired when modulate individual functions, but it does not fully explain how LC inputs to IL are overstimulated, an effect which is dependent this small population of cells regulates global brain activity in some 84 on α1-AR activation . α1-ARs, but not β-ARs, increase GABAergic instances and regulates specific, even opposing, functions in other signaling in mPFC109, suggesting that different receptor subtypes situations14,118. Classical models postulated that LC neurons were regulate distinct subnetworks in the IL to impair or enhance extinc- homogeneous, responding to attention-arousing events in a simi- tion, respectively. lar way, with each cell maintaining broad, divergent connectivity While it isn’t known how NA and LC innervation affect IL neu- throughout the brain. Although some LC cells are indeed broadly ronal processing, physiology studies examining their effects on collateralized and project extensively throughout the brain81,82, single-unit and local field potential responses offer some clues. recent work has shown that distinct populations of LC neurons can IL neuronal responses to auditory cues increase following extinc- encode information in specific ways and that some individual cell tion, and stimulation of LC NA neurons reduces this increase110. populations are more precisely connected with their target struc- Furthermore, there are dynamic changes in IL neuronal activity tures and have distinct behavioral functions81–84,119,120. This is sup- during fear learning and extinction; and coordination of oscilla- ported by a recent in vivo electrophysiology study in anesthetized tory activity between mPFC, amygdala and hippocampus occurs animals, which found that spontaneous and stimulus-evoked LC during fear extinction111,112. Although it is unclear whether or how neuronal spiking tended to be non-correlated, but that a small pro- LC activity regulates this neuronal processing and synchrony, LC portion of LC neurons formed cell assemblies through correlated does modulate mPFC oscillatory activity. For example, in both spiking activity121 (but see ref. 122 showing more correlated LC neu- mPFC and hippocampus, stimulation of LC in anesthetized animals ronal activity in other brain and behavioral states). Related specifi- increases gamma and decreases theta oscillations, desynchronizes cally to fear and extinction learning, a recent study reported that neuronal activity and produces biphasic effects on neuronal firing separate populations of LC neurons innervate BLA or IL84. These rates113,114. While the effect of LC stimulation on IL oscillations dur- cells are not broadly collateralized, but maintain somewhat specific ing waking has not been studied, some work suggests that the effects connectivity with amygdala or mPFC. As discussed above, amyg- of LC stimulation on hippocampal oscillations may be state-spe- dala-projecting cells linearly facilitate fear learning and oppose cific. In awake animals, LC stimulation enhances theta and reduces extinction learning, whereas IL-projecting LC neurons are nec- gamma115, whereas during sleep it reduces theta oscillations116. LC essary for extinction learning, but their overstimulation reduces modulation of rhythmic activity could be important in understand- extinction (Fig. 4). Notably, LC neurons display distinct firing ing distributed oscillatory coupling between IL, amygdala and hip- modes during fear and extinction learning. During fear condition- pocampus during fear learning, discrimination and extinction. As ing, most cells respond robustly to aversive shocks, resembling the NA has been implicated in sleep dysregulation in post-traumatic classic LC arousal signal. By contrast, during fear extinction, firing stress disorder (PTSD)117, this research direction could have impor- rate responses are more modest and heterogeneous. Subpopulations tant clinical implications. of LC neurons respond to fear-inducing cues early in training, and

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1591 REVIEW ARTICLE | FOCUS NATURE NEUROSCIENCE

a receptors b C D1,5-R D2,3–4-R

CeA NAc

HPC LTP LA/B

LA/B waLTP ex waLTP VTA

mPFC wa PAG LTP

Fig. 6 | Projection-specific dopaminergic regulation of fear learning and extinction. a, Summary of dopaminergic receptor-mediated contributions to aversive emotional learning in the hippocampus (HPC), BLA and mPFC. The contributions of different receptor types to fear conditioning, extinction and discrimination learning are shown. Receptor contributions are based on primary research and reviews, as referred to in the text. b, Neurons in the VTA detect when an expected aversive shock does not occur and facilitate extinction of fear, partially through projections to the NAc, whose activity during shock omission contributes to the retention of extinction learning. Red cells participate in aversive processes and blue cells participate in extinction, while green cells are interneurons. Dopaminergic projections from VTA to LA and CeA and from PAG to CeA facilitate fear learning and discrimination, whereas VTA–mPFC projections facilitate aversive processing. LTP, long-term potentiation.

a different LC cell population responds later in extinction when well as context-dependent neuromodulation, could underlie the fear responses are being reduced. Importantly, amygdala-projecting flexible, context-specific LC neural computations. LC cells respond early in extinction, and mPFC projecting cells respond more during later extinction. Roles of dopamine in emotional learning Together, these findings support a model of context-dependent Dopaminergic regulation of fear learning in the amygdala. neural coding in LC, in which different activation modes (global Dopamine neurons are found in the (VTA) vs discrete) that incorporate precise control of activity level and and (SN) as well as in the breadth of activation (number of cells) are coordinated with projec- (PAG), and retrorubral field, and they signal through tion-specific cell populations (Figs. 4 and 5). Related to emotional the DA receptor 1,5 family (D1-R, Gs-coupled) and DA receptor processing, this provides a brain-wide NA arousal signal during 2,3,4 receptor family (D2-R, Gi-coupled) (Fig. 6a). The DA sys- more intense aversive states or more-targeted, dynamic and moder- tem is important for many aspects of behavior, including reward ate release during different phases of extinction learning. Coupled learning, , decision-making, motor control and work- with the linear facilitating effect of NA in the amygdala and the ing memory. While much research has focused on the role of DA inverted-U function of NA in the mPFC, this could explain how in reward-related behaviors, how DA regulates aversive processes higher brain-wide levels of NA facilitate aversive learning (and dis- is only now being clarified. Early pharmacological studies demon- engagement of mPFC extinction networks), whereas during extinc- strated the importance of D1-R and D2-R activation in the amyg- tion more moderate levels of NA in amygdala or mPFC opposes dala on fear learning, as well as for the expression of behavioral fear or facilitates extinction at different time points. Other behavioral responses (Fig. 6a)18. Through actions on BLA pyramidal neurons conditions also engage either global or discrete changes in LC firing and microcircuits, DA enhances signal-to-noise ratios and synaptic rates14, suggesting that this framework may offer a general mecha- plasticity. Similarly to NA, D1-R and D2-R activation reduces feed- nistic model for understanding LC NA function (Fig. 5). forward and feedback inhibition and enhances long-term poten- How this context-dependent modular coding arises is not clear, tiation125,126. Counterintuitively, both receptor subtypes produce but several possibilities exist (Fig. 5). First, the degree to which a similarly inhibitory effects on GABAergic inhibition in BLA despite particular afferent input is specifically or broadly connected with LC the fact that D1-Rs are normally Gs-coupled. Possible explanations neurons might produce discrete or global coding, respectively, when for this could be that (i) in some conditions, D1-R signaling is non- these inputs are active. This could be coupled with context specific canonical127, or (ii) these effects could be mediated through more neuromodulatory signals, such as CRF or other neuropeptides that elaborated local microcircuits. Conversely, DA agonists increase the alter basal firing rates or local circuit interactions, to change tonic excitability of inhibitory interneurons in the LA128. This suggests levels of LC activity more globally. Alternatively, neuromodulatory that DA increases signal-to-noise ratios by generally augmenting inputs to LC could alter the sensitivity or responsivity of LC neurons inhibition in the BLA while enhancing firing in activated cell popu- to external sensory afferent drive and help select phasic global or lations by reducing stimulus-evoked GABAergic inhibition. In vivo, discrete coding modes. A final, not necessarily mutually exclusive, DA agonists have a more nuanced effect, differentially regulating possibility is that local circuit interactions facilitate context-depen- distinct afferent inputs to BLA. D1-R activation reduces evoked dent LC coding. This could occur through differential connectiv- activity and afferent drive from both auditory cortex and , ity or gap-junction coupling within distinct networks of LC cells whereas D2-R activity increases thalamic input-evoked responses or through specific connectivity of pericoeruleus GABAergic net- while decreasing auditory cortex-evoked responses129. This sug- works (which ring the LC) with different LC modules123,124. Thus, gests that in addition to cellular effects on inhibitory and excitatory combinations of specific or promiscuous afferent connectivity, as transmission, DA can also differentially modulate specific afferent

1592 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE inputs. This type of circuit-selective neuromodulatory mechanism in task design or because this only occurs in a subset of VTA DA could differentiate the role of distinct neuromodulators on amyg- neurons. Supporting the latter idea, DA release and the activity of dala function. VTA DA axonal terminals in the NAc medial shell region increase An important source of DA to the BLA and CeA is the VTA and when expected aversive outcomes are omitted132,142, while this signal SN (Fig. 1). Similarly to the LC NA and ACh systems, distinct popu- is not apparent in other VTA DA projection fields. The omission- lations of DA neurons project to individual target sites10. Notably, related activity in DA neurons is necessary for aversive extinction separate populations of VTA DA neurons project to the amygdala, learning and is mediated partially by projections to the NAc medial mPFC and striatum (Fig. 6b). DA neurons in VTA and SN encode shell, which contribute during learning to the long-term retention reward prediction error: they respond more strongly to unexpected of extinction143,145 (Fig. 6b). This is consistent with prior pharmacol- rewards, which functions as an instructive signal for appetitive ogy studies demonstrating the importance of DA-receptor signaling learning9,11. However, some DA neurons respond to aversive events in the NAc146,147. Notably, activity in DA projections to mPFC and as well and may encode saliency of reward or aversive outcomes and amygdala during shock omission does not facilitate extinction145. In sensory predictive cues130–132. Demonstrating a causal role of VTA fact, optogenetic inhibition of VTA DA projections to the mPFC DA in fear conditioning, knocking down NMDA receptors in VTA and pharmacological blockade of D1-Rs, which are less sensitive to neurons reduces shock-evoked activity in putative DA neurons DA, in mPFC enhances extinction. Conversely, blockade of the more and reduces stress-induced anxiety133. Furthermore, DA-synthesis- sensitive D2-Rs in mPFC impairs extinction. Because of their differ- deficient mice are impaired in fear conditioning, and this can be ent sensitivities, D1-Rs and D2-Rs are thought to detect phasic and partially rescued by restoring DA function specifically in BLA- tonic DA levels, respectively. Together with the differential effects of projecting DA cells134. To more specifically isolate the time period in D1 and D2 antagonists, this suggests unique roles for phasic vs tonic which DA neurons contribute to fear learning, a recent study opto- release modes in mPFC during fear extinction. In humans, enhanc- genetically stimulated DA neurons during the early tone cue period ing tonic DA levels also modulates extinction of fear148, indicating and found that this manipulation facilitates fear discrimination135. the therapeutic potential of targeting of this system for anxiety dis- Furthermore, the enhancement of discrimination depends on pro- orders and PTSD. Together, these findings demonstrate that the jections to the CeA135. The CeA is also modulated by dopamine VTA DA system and its projections to the NAc detect the absence of neurons in the PAG, and the projections of these cells to CeA are an expected aversive event and thereby promote extinction learning necessary for fear learning as well as for fear conditioning-induced and memory. Future studies may define the role of the NAc in aver- synaptic strengthening of BLA-to-CeA connections136. Interestingly, sive extinction learning and determine how the func- this connection is reciprocal, as PAG dopamine neurons receive tions in concert with mPFC and amygdala during extinction—this direct inhibitory innervation from CeA cells. This reveals a poten- may provide valuable insights into how aversive emotional learning tial mechanism for the learning induced inhibition of aversive pro- and motivated, instrumental behaviors are coordinated. cessing apparent in PAG neurons136,137. Collectively, these studies demonstrate that DA projections from different regions Context-dependent modular coding in the DA system?. The regulate fear learning through projections to BLA and CeA (Figs. 1 fact that specific populations and projections of DA cells respond and 6) and suggest local synaptic, cellular and microcircuit mecha- to aversive events, while the majority of the population encodes nisms through which DA can regulate this learning function. reward prediction error, suggests that DA neurons may also exhibit In addition to modulating fear learning through projections to a form of context-dependent modular coding. In fact, a recent study the amygdala, separate populations of projection-defined DA cells reported that distinct populations of VTA and SN neurons respond also regulate aversive learning132,138. More specifically, VTA DA to different sensory, motor and contextual variables, while the larger input to the mPFC and DA-receptor signaling there facilitates aver- population encodes reward more homogeneously149. Notably, the sive learning and behavior139, possibly by temporally coupling neu- smaller subsets of response-defined neurons were topographically ral oscillations or firing between mPFC and amygdala. Accordingly, clustered, and their localization may correspond to distinct popu- 4-Hz oscillations coordinate neuronal assemblies across mPFC, lations of projection-defined cell populations149,150. Together, this hippocampus and amygdala to facilitate fear behaviors61. Further, supports the idea that, like the LC, the DA system can multiplex the VTA, mPFC and hippocampus display coordinated 4-Hz oscil- local and global signals that may be coordinated with distinct sets of lations during working memory tasks140. Other populations of VTA projection-target-specific output neurons. DA neurons that participate in aversive processes include those pro- jecting to the ventromedial shell of the (NAc) Themes in neuromodulation of aversive emotional learning and SN DA cells projecting to the tail of the striatum131,132. These Neuromodulators play a critical role in regulating neural activity dopaminergic populations respond to unexpected aversive events during emotional learning. Fear conditioning, extinction and dis- and cues that predict them. DA-receptor signaling in the hippocam- crimination learning are highly dependent on the general state of pus is also important for contextual fear learning141, but the VTA the organism, including arousal, cognitive and motivational states. and SN projections to hippocampus are sparse (Fig. 1). However, These factors alter sensory processing and in turn change how sur- recent work suggests that the LC releases DA in the hippocampus to roundings are perceived and remembered. For example, during modulate some forms of contextual learning98. periods of high arousal, neuromodulators shift neural physiology away from baseline, thereby fundamentally altering the way cells Dopamine modulation of fear-extinction learning. Fear extinc- respond to incoming stimuli and communicate at the systems level. tion occurs when predictive cues induce an anticipation of an Accordingly, one common theme for all three neuromodulatory aversive outcome and this outcome does not occur. One intrigu- systems discussed here is their causal role in determining context- ing idea proposed by computational models of DA function is dependent coding at their postsynaptic targets. Another common that the omission of an expected aversive experience is similar to characteristic of these systems is their distributed, brain-wide a rewarding event in its recruitment of the DA system. Providing projection patterns, as well as the specificity of projections from direct support for this idea, the DA system is not only activated by individual subpopulations of neuromodulatory cells. This pattern reward but also by the omission of expected aversive events130,142,143. creates functional spheres of influence that allow for modulation of However, some studies have failed to detect increases in VTA DA specific regions or wider domains, depending on how much of the neural activity in response to omission of expected aversive out- neuromodulatory system is engaged. We do not yet know whether comes (for example, see ref. 144). This could be due to differences activation thresholds or microcircuit connectivity varies for

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1593 REVIEW ARTICLE | FOCUS NATURE NEUROSCIENCE neuromodulatory cells that are wired into different pathways, which 10. Fields, H. L., Hjelmstad, G. O., Margolis, E. B. & Nicola, S. M. Ventral would provide an additional layer of control for their differential, tegmental area neurons in learned appetitive behavior and positive . Annu. Rev. Neurosci. 30, 289–316 (2007). state-dependent activation. In this respect, uncovering how neu- 11. Keifin, R. & Janak, P. H. Dopamine prediction errors in reward learning romodulatory systems are themselves modulated at the local level and : from theory to neural circuitry. 88, 247–263 (2015). will be a fruitful area for research. Nevertheless, the known distrib- 12. Sara, S. J. Te locus coeruleus and noradrenergic modulation of . uted—but modular—anatomical layout, coupled with variations in Nat. Rev. Neurosci. 10, 211–223 (2009). strength and breadth of activation across a given neuromodulatory 13. Berridge, C. W. & Waterhouse, B. D. Te locus coeruleus-noradrenergic system: modulation of behavioral state and state-dependent cognitive neuronal pool, allows for global versus more-restricted downstream processes. Brain Res. Brain Res. Rev. 42, 33–84 (2003). modulation. Thus, phasic or tonic stimuli that drive small subpopu- 14. Uematsu, A., Tan, B. Z. & Johansen, J. P. Projection specifcity in lations of neuromodulatory neurons produce a more modular and heterogeneous locus coeruleus cell populations: implications for learning localized effect downstream, whereas activity across many different and memory. Learn. Mem. 22, 444–451 (2015). subpopulations results in more widespread changes. 15. Sengupta, A. & Holmes, A. A discrete dorsal raphe to basal amygdala 5-HT circuit calibrates aversive memory. Neuron 103, 489–505.e7 (2019). In all cases, however, the net result of neuromodulatory activity 16. Burghardt, N. S. & Bauer, E. P. Acute and chronic efects of selective is to shift downstream neurons away from baseline, changing their serotonin reuptake inhibitor treatment on fear conditioning: implications mode of encoding. Notably, this is a fundamentally different mode for underlying fear circuits. Neuroscience 247, 253–272 (2013). of action than via regulatory systems, such as the hypothalamic– 17. LeDoux, J. E. & Pine, D. S. Using neuroscience to help understand fear and anxiety: a two-system framework. Am. J. Psychiatry 173, 1083–1093 (2016). pituitary–adrenal axis, which in response to stress has feedback 18. Pezze, M. A. & Feldon, J. Mesolimbic in fear loops geared toward maintaining homeostasis and keeping neural conditioning. Prog. Neurobiol. 74, 301–320 (2004). activity close to baseline. During emotional learning in a highly 19. Giustino, T. F. & Maren, S. Noradrenergic modulation of fear conditioning aroused state, such as in conditions of stress, the neuromodulatory and extinction. Front. Behav. Neurosci. 12, 43 (2018). and regulatory systems work together to shift the neural state and to 20. Ballinger, E. C., Ananth, M., Talmage, D. A. & Role, L. W. Basal forebrain cholinergic circuits and signaling in cognition and cognitive decline. establish a new homeostasis baseline, respectively. Notably, the way Neuron 91, 1199–1218 (2016). in which different neuromodulator systems affect encoding in dif- 21. Lee, M. G., Hassani, O. K., Alonso, A. & Jones, B. E. Cholinergic basal ferent states varies. For instance, whereas the cholinergic and dopa- forebrain neurons burst with theta during waking and paradoxical sleep. minergic systems are geared toward altering the signal-to-noise J. Neurosci. 25, 4365–4369 (2005). ratio through a plethora of changes at the cellular and network lev- 22. Xu, M. et al. Basal forebrain circuit for sleep-wake control. Nat. Neurosci. 18, 1641–1647 (2015). els, NA modulation has context-dependent effects. NA increases the 23. Boucetta, S., Cissé, Y., Mainville, L., Morales, M. & Jones, B. E. Discharge gain of incoming stimuli by enhancing excitability in target neurons profles across the sleep-waking cycle of identifed cholinergic, GABAergic, in some situations, whereas it disrupts network function in others. and glutamatergic neurons in the pontomesencephalic tegmentum of the Therefore, regulatory systems work to re-establish homeostasis in rat. J. Neurosci. 34, 4708–4727 (2014). the face of a wide range of changes that alter neural activity. An 24. Aitta-Aho, T. et al. Basal forebrain and brainstem cholinergic neurons diferentially impact amygdala circuits and learning-related behavior. important area for further research is to uncover how the differ- Curr. Biol. 28, 2557–2569.e4 (2018). ent systems work in concert at downstream targets and how they 25. Cissé, Y. et al. Discharge and role of acetylcholine pontomesencephalic impact each other during emotional learning (Fig. 1). Furthermore, neurons in cortical activity and sleep-wake states examined by optogenetics a deeper understanding of how the regulatory and neuromodula- and juxtacellular recording in mice. eNeuro 5, ENEURO.0270-18.2018 (2018). tory systems interact during different states will provide an impor- 26. Wu, H., Williams, J. & Nathans, J. Complete morphologies of basal forebrain cholinergic neurons in the mouse. eLife 3, e02444 (2014). tant avenue for developing novel therapeutic approaches to alleviate 27. Gritti, I., Manns, I. D., Mainville, L. & Jones, B. E. Parvalbumin, calbindin, stress-induced disruption in emotional learning. or calretinin in cortically projecting and GABAergic, cholinergic, or glutamatergic basal forebrain neurons of the rat. J. Comp. Neurol. 458, Received: 10 June 2019; Accepted: 21 August 2019; 11–31 (2003). Published online: 24 September 2019 28. Gritti, I. et al. Stereological estimates of the basal forebrain cell population in the rat, including neurons containing choline acetyltransferase, glutamic References acid decarboxylase or phosphate-activated glutaminase and colocalizing 1. Everitt, B. J. & Robbins, T. W. Central cholinergic systems and cognition. vesicular glutamate transporters. Neuroscience 143, 1051–1064 (2006). Annu. Rev. Psychol. 48, 649–684 (1997). 29. Gielow, M. R. & Zaborszky, L. Te input-output relationship of the 2. Mesulam, M. M., Mufson, E. J., Levey, A. I. & Wainer, B. H. Cholinergic cholinergic basal forebrain. Cell Rep. 18, 1817–1830 (2017). innervation of cortex by the basal forebrain: cytochemistry and cortical 30. McDonald, A. J., Muller, J. F. & Mascagni, F. Postsynaptic targets of connections of the , diagonal band nuclei, nucleus basalis GABAergic basal forebrain projections to the basolateral amygdala. (substantia innominata), and hypothalamus in the rhesus monkey. Neuroscience 183, 144–159 (2011). J. Comp. Neurol. 214, 170–197 (1983). 31. Muller, J. F., Mascagni, F. & McDonald, A. J. Cholinergic innervation of 3. Mesulam, M. M., Mufson, E. J., Wainer, B. H. & Levey, A. I. Central pyramidal cells and parvalbumin-immunoreactive interneurons in the rat cholinergic pathways in the rat: an overview based on an alternative basolateral amygdala. J. Comp. Neurol. 519, 790–805 (2011). nomenclature (Ch1-Ch6). Neuroscience 10, 1185–1201 (1983). 32. Lin, S. C., Brown, R. E., Hussain Shuler, M. G., Petersen, C. C. & Kepecs, 4. Carlsen, J., Záborszky, L. & Heimer, L. Cholinergic projections from the A. Optogenetic dissection of the basal forebrain neuromodulatory basal forebrain to the basolateral amygdaloid complex: a combined control of cortical activation, plasticity, and cognition. J. Neurosci. 35, retrograde fuorescent and immunohistochemical study. J. Comp. Neurol. 13896–13903 (2015). 234, 155–167 (1985). 33. Unal, C. T., Pare, D. & Zaborszky, L. Impact of basal forebrain cholinergic 5. Zaborszky, L., Pang, K., Somogyi, J., Nadasdy, Z. & Kallo, I. Te basal inputs on basolateral amygdala neurons. J. Neurosci. 35, 853–863 (2015). forebrain corticopetal system revisited. Ann. NY Acad. Sci. 877, 34. Li, X. et al. Generation of a whole-brain atlas for the cholinergic system and 339–367 (1999). mesoscopic projectome analysis of basal forebrain cholinergic neurons. 6. Freund, T. F. & Antal, M. GABA-containing neurons in the septum Proc. Natl. Acad. Sci. USA 115, 415–420 (2018). control inhibitory interneurons in the hippocampus. Nature 336, 35. Bloem, B. et al. Topographic mapping between basal forebrain cholinergic 170–173 (1988). neurons and the medial prefrontal cortex in mice. J. Neurosci. 34, 7. Lee, M. G., Chrobak, J. J., Sik, A., Wiley, R. G. & Buzsáki, G. Hippocampal 16234–16246 (2014). theta activity following selective lesion of the septal cholinergic system. 36. Zelikowsky, M., Hersman, S., Chawla, M. K., Barnes, C. A. & Fanselow, M. S. Neuroscience 62, 1033–1047 (1994). Neuronal ensembles in amygdala, hippocampus, and prefrontal cortex track 8. Frotscher, M. & Léránth, C. Cholinergic innervation of the rat hippocampus diferential components of contextual fear. J. Neurosci. 34, as revealed by choline acetyltransferase immunocytochemistry: a combined 8462–8466 (2014). light and electron microscopic study. J. Comp. Neurol. 239, 237–246 (1985). 37. Easton, A., Fitchett, A. E., Eacott, M. J. & Baxter, M. G. Medial septal 9. Watabe-Uchida, M., Eshel, N. & Uchida, N. Neural circuitry of reward cholinergic neurons are necessary for context-place memory but not prediction Error. Annu. Rev. Neurosci. 40, 373–394 (2017). episodic-like memory. Hippocampus 21, 1021–1027 (2011).

1594 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE

38. Jiang, L. et al. Cholinergic signaling controls conditioned fear behaviors 65. Howe, W. M. et al. Acetylcholine release in prefrontal cortex promotes and enhances plasticity of cortical-amygdala circuits. Neuron 90, gamma oscillations and theta-gamma coupling during cue detection. 1057–1070 (2016). J. Neurosci. 37, 3215–3230 (2017). 39. Gale, G. D., Anagnostaras, S. G. & Fanselow, M. S. Cholinergic modulation 66. Parikh, V., Kozak, R., Martinez, V. & Sarter, M. Prefrontal acetylcholine of pavlovian fear conditioning: efects of intrahippocampal scopolamine release controls cue detection on multiple timescales. Neuron 56, infusion. Hippocampus 11, 371–376 (2001). 141–154 (2007). 40. Knox, D. & Keller, S. M. Cholinergic neuronal lesions in the medial septum 67. Stujenske, J. M., Likhtik, E., Topiwala, M. A. & Gordon, J. A. Fear and and vertical limb of the diagonal bands of Broca induce contextual fear safety engage competing patterns of theta-gamma coupling in the memory generalization and impair acquisition of fear extinction. basolateral amygdala. Neuron 83, 919–933 (2014). Hippocampus 26, 718–726 (2016). 68. Gorka, A. X., Knodt, A. R. & Hariri, A. R. Basal forebrain moderates the 41. Zelikowsky, M. et al. Cholinergic blockade frees fear extinction from its magnitude of task-dependent amygdala functional connectivity. Soc. Cogn. contextual dependency. Biol. Psychiatry 73, 345–352 (2013). Afect. Neurosci. 10, 501–507 (2015). 42. Boskovic, Z. et al. Cholinergic basal forebrain neurons regulate fear 69. Rasmusson, D. D., Smith, S. A. & Semba, K. Inactivation of prefrontal extinction consolidation through p75 neurotrophin receptor signaling. cortex abolishes cortical acetylcholine release evoked by sensory or sensory Transl. Psychiatry 8, 199 (2018). pathway stimulation in the rat. Neuroscience 149, 232–241 (2007). 43. Sigurdsson, T., Stark, K. L., Karayiorgou, M., Gogos, J. A. & Gordon, J. A. 70. Pidoplichko, V. I., Prager, E. M., Aroniadou-Anderjaska, V. & Braga, M. F. Impaired hippocampal-prefrontal synchrony in a genetic mouse model of α7-Containing nicotinic acetylcholine receptors on interneurons of the . Nature 464, 763–767 (2010). basolateral amygdala and their role in the regulation of the network 44. Siapas, A. G., Lubenov, E. V. & Wilson, M. A. Prefrontal phase locking to excitability. J. Neurophysiol. 110, 2358–2369 (2013). hippocampal theta oscillations. Neuron 46, 141–151 (2005). 71. Baysinger, A. N., Kent, B. A. & Brown, T. H. Muscarinic receptors in 45. Zhang, H., Lin, S. C. & Nicolelis, M. A. A distinctive subpopulation of amygdala control trace fear conditioning. PLoS One 7, e45720 (2012). medial septal slow-fring neurons promote hippocampal activation and 72. Resnik, J., Sobel, N. & Paz, R. Auditory aversive learning increases theta oscillations. J. Neurophysiol. 106, 2749–2763 (2011). discrimination thresholds. Nat. Neurosci. 14, 791–796 (2011). 46. Vinogradova, O. S., Kitchigina, V. F. & Zenchenko, C. I. Pacemaker neurons 73. Minces, V., Pinto, L., Dan, Y. & Chiba, A. A. Cholinergic shaping of neural of the forebrain medical septal area and theta rhythm of the hippocampus. correlations. Proc. Natl. Acad. Sci. USA 114, 5725–5730 (2017). Membr. Cell Biol. 11, 715–725 (1998). 74. Polack, P. O., Friedman, J. & Golshani, P. Cellular mechanisms of brain 47. Huh, C. Y., Goutagny, R. & Williams, S. Glutamatergic neurons of the state-dependent gain modulation in visual cortex. Nat. Neurosci. 16, mouse medial septum and synaptically drive 1331–1339 (2013). hippocampal pyramidal cells: relevance for hippocampal theta rhythm. 75. Froemke, R. C., Merzenich, M. M. & Schreiner, C. E. A synaptic J. Neurosci. 30, 15951–15961 (2010). memory trace for cortical receptive feld plasticity. Nature 450, 48. Vandecasteele, M. et al. Optogenetic activation of septal cholinergic neurons 425–429 (2007). suppresses sharp wave ripples and enhances theta oscillations in the 76. Letzkus, J. J. et al. A disinhibitory microcircuit for associative fear learning hippocampus. Proc. Natl. Acad. Sci. USA 111, 13535–13540 (2014). in the auditory cortex. Nature 480, 331–335 (2011). 49. Joshi, A., Salib, M., Viney, T. J., Dupret, D. & Somogyi, P. Behavior- 77. Tikhonova, T. B., Miyamae, T., Gulchina, Y., Lewis, D. A. & Gonzalez- dependent activity and synaptic organization of septo-hippocampal Burgos, G. Cell type- and layer-specifc muscarinic potentiation of GABAergic neurons selectively targeting the hippocampal CA3 area. excitatory synaptic drive onto parvalbumin neurons in mouse prefrontal Neuron 96, 1342–1357.e5 (2017). cortex. eNeuro 5, ENEURO.0208-18.2018 (2018). 50. Hangya, B., Borhegyi, Z., Szilágyi, N., Freund, T. F. & Varga, V. GABAergic 78. James, N. M., Gritton, H. J., Kopell, N., Sen, K. & Han, X. Muscarinic neurons of the medial septum lead the hippocampal network during theta receptors regulate auditory and prefrontal cortical communication during activity. J. Neurosci. 29, 8094–8102 (2009). auditory processing. 144, 155–171 (2019). 51. Tingley, D. et al. Task-phase-specifc dynamics of basal forebrain neuronal 79. Nakamura, S. & Iwama, K. Antidromic activation of the rat locus coeruleus ensembles. Front. Syst. Neurosci. 8, 174 (2014). neurons from hippocampus, cerebral and cerebellar cortices. Brain Res. 99, 52. Tingley, D., Alexander, A. S., Quinn, L. K., Chiba, A. A. & Nitz, D. 372–376 (1975). Multiplexed oscillations and phase rate coding in the basal forebrain. 80. Room, P., Postema, F. & Korf, J. Divergent collaterals of rat locus Sci. Adv. 4, r3230 (2018). coeruleus neurons: demonstration by a fuorescent double labeling 53. Mascagni, F., Muly, E. C., Rainnie, D. G. & McDonald, A. J. technique. Brain Res. 221, 219–230 (1981). Immunohistochemical characterization of parvalbumin-containing 81. Kebschull, J. M. et al. High-throughput mapping of single-neuron interneurons in the monkey basolateral amygdala. Neuroscience 158, projections by sequencing of barcoded RNA. Neuron 91, 975–987 (2016). 1541–1550 (2009). 82. Schwarz, L. A. et al. Viral-genetic tracing of the input-output organization 54. Henny, P. & Jones, B. E. Projections from basal forebrain to prefrontal of a central noradrenaline circuit. Nature 524, 88–92 (2015). cortex comprise cholinergic, GABAergic and glutamatergic inputs to 83. Chandler, D. J., Gao, W. J. & Waterhouse, B. D. Heterogeneous organization pyramidal cells or interneurons. Eur. J. Neurosci. 27, 654–670 (2008). of the locus coeruleus projections to prefrontal and motor cortices. Proc. 55. Lin, S. C., Gervasoni, D. & Nicolelis, M. A. Fast modulation of prefrontal Natl. Acad. Sci. USA 111, 6816–6821 (2014). cortex activity by basal forebrain noncholinergic neuronal ensembles. 84. Uematsu, A. et al. Modular organization of the brainstem noradrenaline J. Neurophysiol. 96, 3209–3219 (2006). system coordinates opposing learning states. Nat. Neurosci. 20, 56. Yang, C. et al. Cholinergic neurons excite cortically projecting basal 1602–1611 (2017). forebrain GABAergic neurons. J. Neurosci. 34, 2832–2844 (2014). 85. Quirarte, G. L., Galvez, R., Roozendaal, B. & McGaugh, J. L. 57. Freund, T. F. & Meskenaite, V. gamma-Aminobutyric acid-containing basal release in the amygdala in response to footshock and forebrain neurons innervate inhibitory interneurons in the . peptidergic . Brain Res. 808, 134–140 (1998). Proc. Natl. Acad. Sci. USA 89, 738–742 (1992). 86. Bush, D. E., Caparosa, E. M., Gekker, A. & Ledoux, J. Beta-adrenergic 58. Unal, G. et al. Spatio-temporal specialization of GABAergic septo- receptors in the lateral nucleus of the amygdala contribute to the hippocampal neurons for rhythmic network activity. Brain Struct. Funct. acquisition but not the consolidation of auditory fear conditioning. Front. 223, 2409–2432 (2018). Behav. Neurosci. 4, 154 (2010). 59. Likhtik, E., Stujenske, J. M., Topiwala, M. A., Harris, A. Z. & Gordon, J. A. 87. Schif, H. C. et al. β-Adrenergic receptors regulate the acquisition and Prefrontal entrainment of amygdala activity signals safety in learned fear consolidation phases of aversive memory formation through distinct, and innate anxiety. Nat. Neurosci. 17, 106–113 (2014). temporally regulated signaling pathways. Neuropsychopharmacology 42, 60. Seidenbecher, T., Laxmi, T. R., Stork, O. & Pape, H. C. Amygdalar and 895–903 (2017). hippocampal theta rhythm synchronization during fear memory retrieval. 88. Faber, E. S. et al. Modulation of SK channel trafcking by beta Science 301, 846–850 (2003). adrenoceptors enhances excitatory synaptic transmission and plasticity in 61. Karalis, N. et al. 4-Hz oscillations synchronize prefrontal-amygdala circuits the amygdala. J. Neurosci. 28, 10803–10813 (2008). during fear behavior. Nat. Neurosci. 19, 605–612 (2016). 89. Tully, K., Li, Y., Tsvetkov, E. & Bolshakov, V. Y. Norepinephrine enables the 62. Paré, D. & Collins, D. R. Neuronal correlates of fear in the lateral amygdala: induction of associative long-term potentiation at thalamo-amygdala multiple extracellular recordings in conscious cats. J. Neurosci. 20, synapses. Proc. Natl. Acad. Sci. USA 104, 14146–14150 (2007). 2701–2710 (2000). 90. Roozendaal, B. et al. Basolateral amygdala noradrenergic activity mediates 63. Courtin, J. et al. Prefrontal parvalbumin interneurons shape neuronal corticosterone-induced enhancement of auditory fear conditioning. activity to drive fear expression. Nature 505, 92–96 (2014). Neurobiol. Learn. Mem. 86, 249–255 (2006). 64. Takács, V. T. et al. Co-transmission of acetylcholine and GABA regulates 91. McGaugh, J. L. Making lasting memories: remembering the signifcant. hippocampal states. Nat. Commun. 9, 2848 (2018). Proc. Natl. Acad. Sci. USA 110(Suppl 2), 10402–10407 (2013).

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1595 REVIEW ARTICLE | FOCUS NATURE NEUROSCIENCE

92. Robertson, S. D., Plummer, N. W., de Marchena, J. & Jensen, P. 120. Hirschberg, S., Li, Y., Randall, A., Kremer, E. J. & Pickering, A. E. Developmental origins of central norepinephrine neuron diversity. Functional dichotomy in spinal- vs prefrontal-projecting locus coeruleus Nat. Neurosci. 16, 1016–1023 (2013). modules splits descending noradrenergic analgesia from ascending aversion 93. Usunof, K. G., Itzev, D. E., Rolfs, A., Schmitt, O. & Wree, A. Brain stem and anxiety in rats. eLife 6, e29808 (2017). aferent connections of the amygdala in the rat with special references to a 121. Totah, N. K., Neves, R. M., Panzeri, S., Logothetis, N. K. & Eschenko, O. projection from the parabigeminal nucleus: a fuorescent retrograde tracing Te locus coeruleus is a complex and diferentiated neuromodulatory study. Anat. Embryol. (Berl.) 211, 475–496 (2006). system. Neuron 99, 1055–1068.e6 (2018). 94. McCall, J. G. et al. Locus coeruleus to basolateral amygdala noradrenergic 122. Usher, M., Cohen, J. D., Servan-Schreiber, D., Rajkowski, J. & Aston-Jones, projections promote anxiety-like behavior. eLife 6, e18247 (2017). G. Te role of locus coeruleus in the regulation of cognitive performance. 95. Soya, S. et al. modulates behavioral fear expression through the Science 283, 549–554 (1999). locus coeruleus. Nat. Commun. 8, 1606 (2017). 123. Breton-Provencher, V. & Sur, M. Active control of arousal by a locus 96. Chen, F. J. & Sara, S. J. Locus coeruleus activation by foot shock or coeruleus GABAergic circuit. Nat. Neurosci. 22, 218–228 (2019). electrical stimulation inhibits amygdala neurons. Neuroscience 144, 124. Aston-Jones, G., Zhu, Y. & Card, J. P. Numerous GABAergic aferents to 472–481 (2007). locus ceruleus in the pericerulear dendritic zone: possible interneuronal 97. Heath, F. C. et al. Dopamine D1-like receptor signalling in the pool. J. Neurosci. 24, 2313–2321 (2004). hippocampus and amygdala modulates the acquisition of contextual fear 125. Bissière, S., Humeau, Y. & Lüthi, A. Dopamine gates LTP induction in conditioning. Psychopharmacology (Berl.) 232, 2619–2629 (2015). lateral amygdala by suppressing feedforward inhibition. Nat. Neurosci. 6, 98. Takeuchi, T. et al. Locus coeruleus and dopaminergic consolidation of 587–592 (2003). everyday memory. Nature 537, 357–362 (2016). 126. Marowsky, A., Yanagawa, Y., Obata, K. & Vogt, K. E. A specialized subclass 99. Clayton, E. C. & Williams, C. L. Adrenergic activation of the nucleus of interneurons mediates dopaminergic facilitation of amygdala function. tractus solitarius potentiates amygdala norepinephrine release and enhances Neuron 48, 1025–1037 (2005). retention performance in emotionally arousing and spatial memory tasks. 127. Lorétan, K., Bissière, S. & Lüthi, A. Dopaminergic modulation of Behav. Brain Res. 112, 151–158 (2000). spontaneous inhibitory network activity in the lateral amygdala. 100. Rasmussen, K. & Jacobs, B. L. Single unit activity of locus coeruleus Neuropharmacology 47, 631–639 (2004). neurons in the freely moving cat. II. Conditioning and pharmacologic 128. Kröner, S., Rosenkranz, J. A., Grace, A. A. & Barrionuevo, G. Dopamine studies. Brain Res. 371, 335–344 (1986). modulates excitability of basolateral amygdala neurons in vitro. 101. Giustino, T. F. et al. β-Adrenoceptor blockade in the basolateral amygdala, J. Neurophysiol. 93, 1598–1610 (2005). but not the medial prefrontal cortex, rescues the immediate extinction 129. Chang, C. H. & Grace, A. A. Dopaminergic modulation of lateral defcit. Neuropsychopharmacology 42, 2537–2544 (2017). amygdala neuronal activity: diferential D1 and D2 receptor efects on 102. Lucas, E. K., Wu, W. C., Roman-Ortiz, C. & Clem, R. L. Prazosin during thalamic and cortical aferent inputs. Int. J. Neuropsychopharmacol. 18, fear conditioning facilitates subsequent extinction in male C57Bl/6N mice. pyv015 (2015). Psychopharmacology (Berl.) 236, 273–279 (2019). 130. Matsumoto, M. & Hikosaka, O. Two types of dopamine neuron distinctly 103. Debiec, J. & Ledoux, J. E. Disruption of reconsolidation but not convey positive and negative motivational signals. Nature 459, consolidation of auditory fear conditioning by noradrenergic blockade in 837–841 (2009). the amygdala. Neuroscience 129, 267–272 (2004). 131. Menegas, W., Akiti, K., Amo, R., Uchida, N. & Watabe-Uchida, M. 104. Feng, J. et al. A genetically encoded fuorescent sensor for rapid and specifc Dopamine neurons projecting to the posterior striatum reinforce avoidance in vivo detection of norepinephrine. Neuron 102, 745–761.e8 (2019). of threatening stimuli. Nat. Neurosci. 21, 1421–1430 (2018). 105. Patriarchi, T. et al. Ultrafast neuronal imaging of dopamine dynamics with 132. de Jong, J. W. et al. A neural circuit mechanism for encoding aversive designed genetically encoded sensors. Science 360, eaat4422 (2018). stimuli in the mesolimbic dopamine system. Neuron 101, 106. Yiu, A. P. et al. Neurons are recruited to a memory trace based on relative 133–151.e7 (2019). neuronal excitability immediately before training. Neuron 83, 133. Zweifel, L. S. et al. Activation of dopamine neurons is critical for aversive 722–735 (2014). conditioning and prevention of generalized anxiety. Nat. Neurosci. 14, 107. Mueller, D., Porter, J. T. & Quirk, G. J. Noradrenergic signaling in 620–626 (2011). infralimbic cortex increases cell excitability and strengthens memory for 134. Fadok, J. P., Dickerson, T. M. & Palmiter, R. D. Dopamine is necessary for fear extinction. J. Neurosci. 28, 369–375 (2008). cue-dependent fear conditioning. J. Neurosci. 29, 11089–11097 (2009). 108. Arnsten, A. F. Stress signalling pathways that impair prefrontal cortex 135. Jo, Y. S., Heymann, G. & Zweifel, L. S. Dopamine neurons refect the structure and function. Nat. Rev. Neurosci. 10, 410–422 (2009). uncertainty in fear generalization. Neuron 100, 916–925.e3 (2018). 109. Kawaguchi, Y. & Shindou, T. Noradrenergic excitation and inhibition 136. Groessl, F. et al. Dorsal tegmental dopamine neurons gate associative of GABAergic cell types in rat frontal cortex. J. Neurosci. 18, learning of fear. Nat. Neurosci. 21, 952–962 (2018). 6963–6976 (1998). 137. Ozawa, T. et al. A feedback neural circuit for calibrating aversive memory 110. Giustino, T. F., Fitzgerald, P. J., Ressler, R. L. & Maren, S. Locus coeruleus strength. Nat. Neurosci. 20, 90–97 (2017). toggles reciprocal prefrontal fring to reinstate fear. Proc. Natl. Acad. Sci. 138. Lammel, S. et al. Input-specifc control of reward and aversion in the USA 116, 8570–8575 (2019). ventral tegmental area. Nature 491, 212–217 (2012). 111. Fitzgerald, P. J., Giustino, T. F., Seemann, J. R. & Maren, S. Noradrenergic 139. Stubbendorf, C., Hale, E., Cassaday, H. J., Bast, T. & Stevenson, C. W. blockade stabilizes prefrontal activity and enables fear extinction under Dopamine D1-like receptors in the dorsomedial prefrontal cortex regulate stress. Proc. Natl. Acad. Sci. USA 112, E3729–E3737 (2015). contextual fear conditioning. Psychopharmacology (Berl.) 236, 112. Harris, A. Z. & Gordon, J. A. Long-range neural synchrony in behavior. 1771–1782 (2019). Annu. Rev. Neurosci. 38, 171–194 (2015). 140. Fujisawa, S. & Buzsáki, G. A 4 Hz oscillation adaptively synchronizes 113. Marzo, A., Totah, N. K., Neves, R. M., Logothetis, N. K. & Eschenko, O. prefrontal, VTA, and hippocampal activities. Neuron 72, 153–165 (2011). Unilateral electrical stimulation of rat locus coeruleus elicits bilateral 141. Broussard, J. I. et al. Dopamine regulates aversive contextual learning and response of norepinephrine neurons and sustained activation of medial associated in vivo synaptic plasticity in the hippocampus. Cell Rep. 14, prefrontal cortex. J. Neurophysiol. 111, 2570–2588 (2014). 1930–1939 (2016). 114. Neves, R. M., van Keulen, S., Yang, M., Logothetis, N. K. & Eschenko, O. 142. Badrinarayan, A. et al. Aversive stimuli diferentially modulate real-time Locus coeruleus phasic discharge is essential for stimulus-induced gamma dopamine transmission dynamics within the nucleus accumbens core and oscillations in the prefrontal cortex. J. Neurophysiol. 119, 904–920 (2018). shell. J. Neurosci. 32, 15779–15790 (2012). 115. Walling, S. G., Brown, R. A., Milway, J. S., Earle, A. G. & Harley, C. W. 143. Salinas-Hernández, X. I. et al. Dopamine neurons drive fear extinction Selective tuning of hippocampal oscillations by phasic locus coeruleus learning by signaling the omission of expected aversive outcomes. eLife 7, activation in awake male rats. Hippocampus 21, 1250–1262 (2011). e38818 (2018). 116. Swif, K. M. et al. Abnormal locus coeruleus sleep activity alters sleep 144. Tian, J. & Uchida, N. lesions reveal that multiple mechanisms signatures of memory consolidation and impairs place cell stability and underlie dopamine prediction errors. Neuron 87, 1304–1316 (2015). spatial memory. Curr. Biol. 28, 3599–3609.e4 (2018). 145. Luo, R. et al. A dopaminergic switch for fear to safety transitions. Nat. 117. Hendrickson, R. C. & Raskind, M. A. Noradrenergic dysregulation in the Commun. 9, 2483 (2018). pathophysiology of PTSD. Exp. Neurol. 284 Pt B, 181–195 (2016). 146. Holtzman-Assif, O., Laurent, V. & Westbrook, R. F. Blockade of dopamine 118. Zerbi, V. et al. Rapid reconfguration of the functional connectome afer activity in the nucleus accumbens impairs learning extinction of chemogenetic locus coeruleus activation. Neuron https://doi.org/10.1016/j. conditioned fear. Learn. Mem. 17, 71–75 (2010). neuron.2019.05.034 (2019). 147. Abraham, A. D., Neve, K. A. & Lattal, K. M. Activation of D1/5 119. Loughlin, S. E., Foote, S. L. & Grzanna, R. Eferent projections of nucleus dopamine receptors: a common mechanism for enhancing extinction of locus coeruleus: morphologic subpopulations have diferent eferent targets. fear and reward-seeking behaviors. Neuropsychopharmacology 41, Neuroscience 18, 307–319 (1986). 2072–2081 (2016).

1596 NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience NATURE NEUROSCIENCE FOCUS | REVIEW ARTICLE

148. Haaker, J. et al. Single dose of L-dopa makes extinction memories Additional information context-independent and prevents the return of fear. Proc. Natl. Acad. Sci. Correspondence should be addressed to E.L. or J.P.J. USA 110, E2428–E2436 (2013). Reprints and permissions information is available at www.nature.com/ 149. Engelhard, B. et al. Specialized coding of sensory, motor and cognitive reprints. variables in VTA dopamine neurons. Nature 570, 509–513 (2019). 150. Ikemoto, S. Dopamine reward circuitry: two projection systems from the Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in ventral midbrain to the nucleus accumbens- complex. published maps and institutional affiliations. Brain Res. Rev. 56, 27–78 (2007). © Springer Nature America, Inc. 2019

NATURE NEUROSCIENCE | VOL 22 | OCTOBER 2019 | 1586–1597 | www.nature.com/natureneuroscience 1597