Ketamine restores escape behavior by re-engaging dopamine systems to drive cortical spinogenesis

Abstract

Escaping aversive stimuli is essential for complex organisms, but prolonged exposure to stress leads to maladaptive learning. Stress alters neuronal activity in distributed networks, plasticity, and neuromodulatory signaling; yet, the field lacks a unifying framework for its variegated sequelae. Here we build this framework using learned helplessness paradigm, where ketamine restores escape behavior after aversive learning. Lowdimensional optical readout of dopamine (DA) neuron activity across learning predicts acute behavioral responses, transitions through learning phases, and future sensitivity to ketamine treatment. Ketamine's effects are blocked by chemogenetic inhibition of DA signaling and mimicked by optogenetic activation. We use 2-photon glutamate uncaging/imaging to interrogate structural plasticity in medial prefrontal cortex, revealing that dendritic spinogenesis on pyramidal neurons is regulated by aversive experience and recovered by ketamine in a DA-dependent manner. Together, these data describe recurrent circuits that causally link neuromodulatory dynamics, aversive learning, and plasticity enhancements driven by a therapeutically promising anti-depressant. 3 Ketamine and its S-enantiomer esketamine, which act as antagonists of the glutamatergic N-methyl-Daspartate (NMDA) receptors, demonstrate rapid onset anti-depressant effects in clinical trials 1,2 . Previous studies have shown that ketamine ameliorates depressive-like behaviors in animal models following acute or prolonged stress [3][4][5][6][7] . Accumulating evidence implicates the enhancement of synaptic plasticity in ketamine's behavioral effects 5,6,[8][9][10][11][12] . However, which neural circuit dynamics mediate the effects of this rapidly acting antidepressant on behavior and plasticity are currently unknown.Changes in reward-based and aversive learning represent features of depressive disorders that can be modeled in rodents [13][14][15] . One established model of aversive learning is learned helplessness [16][17][18] . Following prolonged inescapable stress exposure, animals learn that outcomes are independent of their behavioral actions; this learning eventually diminishes attempts to escape from avoidable stressful stimuli 19 . Specific activity patterns of dopaminergic (DA) neurons in the ventral tegmental area (VTA), and downstream DA release, encode reward and aversion [20][21][22][23][24] . This activity is differentially modulated by acute and chronic stress [25][26][27] , regulating depressive-like behaviors [28][29][30] . A recently published meta-analysis suggests that acute sub-anesthetic doses of ketamine may increase DA levels in the cortex, dorsal striatum, and nucleus accumbens 31,32 , but the causal relationship between the DA system and ketamine's effects on neural plasticity and behavior remains to be elucidated. Low dimensional readout of DA neuron activity dynamics in aversive learningTo define the function of midbrain DA neurons during aversive learning, we used a modified model of learned helplessness 5,32 ...

GPCR-Based Dopamine Sensors—A Detailed Guide to Inform Sensor Choice for In Vivo Imaging

“…Moreover, several studies have found that somatodendritic DA release in the midbrain is less dependent on calcium as compared to axonal release in the striatum, making GCaMPs a less appropriate tool to study such somatodendritic release mechanisms. Nonetheless, combined monitoring of calcium transients (GCaMPs) and DA release (dLight/GRAB-DA) can represent a powerful platform to understand how neural activity shapes release dynamics, as demonstrated in recent preprints for acetylcholine [ 97 ] or DA [ 98 ].…”

“…In Table 1 you will find a summary of currently available DA biosensors as well as their main properties, which are further detailed in Section 5 . DA biosensors have already generated key findings in the basic understanding of reward behavior [ 101 , 102 , 103 , 104 , 105 , 106 , 107 ], thirst regulation [ 108 ], feeding behavior [ 109 ], addiction [ 38 , 110 , 111 ], aversive learning [ 112 ], depressive-like behavior [ 98 ], sleep-wake cycle [ 113 ] or to dissect neuromodulator mechanisms in disease models [ 114 , 115 ] using a variety of in vivo imaging modalities shown in Figure 1 . DA biosensors can also be used to understand DA release dynamics in vitro or ex vivo [ 58 , 59 , 60 , 61 ], as shown for example in Reference [ 116 ] where dLight1 was used to understand the metabolic demands and bioenergetic roles of the mitochondria in governing phasic DA release.…”

“…Another major advantage is the ability to express GPCR sensors in specific cell types using cre-driver lines and flexed vectors. For instance, a flexed AAV for dLight1.1 was previously used to target DA 1 receptor (DRD1)-expressing neurons in the mPFC [ 98 ]. This may be particularly useful to compare the responses to DA between different cell types which express different DA receptor subtypes or for example between astrocytes and neurons.…”

“…In Supplementary Table S2 we present an overview of previously published in vivo applications of DA biosensors from the dLight, RdLight and GRAB-DA families in rodent animal models [ 38 , 58 , 60 , 61 , 98 , 101 , 102 , 103 , 104 , 105 , 106 , 107 , 108 , 109 , 110 , 111 , 112 , 113 , 114 , 115 , 116 , 224 , 226 , 227 , 228 ]. The large majority of experiments was performed in either the dorsal striatum or the NAc where the DA concentration reaches its maximal levels.…”

“…DA biosensors have also successfully been put to use in brain regions with medium to low DA, albeit with reduced signal to noise ratio. This includes the frontal/motor cortex (dLight1.1 & 1.2) [ 58 ], the mPFC (dLight1.1) [ 98 ], the basal (dLight1.1 & 1.2) and central (GRAB-DA2m) amygdala [ 38 , 98 , 103 ] and the bed nucleus of the stria terminalis (GRAB-DA2m) [ 38 ]. These studies used sensors with medium to high affinities (90–765 nM) for DA and maximal dFF between 230–340%.…”