Accumbal D2 cells orchestrate innate risk-avoidance according to orexin signals

Abstract

Excitation of accumbal D2 cells governs vital actions including avoidance of learned risks, but the origins of this excitation and roles of D2 cells in innate risk-avoidance are unclear. Hypothalamic neurons producing orexins/hypocretins enhance innate risk-avoidance via poorly understood neurocircuits. We describe a direct orexin→D2 excitatory circuit, and show that D2 cell activity is necessary for orexin-dependent innate risk-avoidance in mice, thus revealing an unsuspected hypothalamus-accumbens interplay in action selection.

Innate ability to avoid lethal risks is key for survival in all organisms. In mammals, context-appropriate actions such as risk-avoidance are computed by the brain, but relations between the underlying neural signals are incompletely understood. Orexin/hypocretin peptides are a fundamental brain system required for context-appropriate brain state control^{1, 2}. They are made exclusively by a subset of lateral hypothalamic (LH) neurons, which are activated by diverse internal and external stresses, and evoke central and autonomic arousal by releasing orexin at brain-wide projections^{1, 2, 3, 4}. Previous studies indicated that the orexin signals are sufficient and necessary for inducing anxiety-like states, such as increased innate risk-avoidance (a hallmark of anxiety⁵) in rodents, or panic attacks in humans^{6, 7, 8, 9}. However, it remains unknown what downstream circuits are essential for innate risk-avoidance driven by natural orexin signals. Among brain areas innervated by orexin_LH cells is the nucleus accumbens (NAc)⁴, a central switch of action strategies¹⁰. The NAc contains dopamine-inhibited D2_NAc neurons¹¹ (Fig. S1A,B), whose excitation was recently found to drive avoidance of learned risk¹⁰. However, it is unclear where the D2_NAc cell excitation physiologically originates, and whether it drives innate risk-avoidance.

To test whether D2_NAc cell excitation may originate from orexin_LH neurons, we optogenetically photo-stimulated NAc orexin axons, or applied orexin peptide, while recording from D2_NAc cells in mouse brain slices (Fig. 1A-D; see Supplementary Methods). The orexin stimulations excited D2_NAc cells on a time-scale similar to the risk-avoidance-associated D2_NAc cell excitation in vivo¹⁰, and this excitation was abolished by orexin receptor antagonism (Fig. 1D,E; a low-probability fast orexin cell → D2 cell excitation was also present in some cells, Fig. S1F). Orexin stimulation did not affect most other NAc cells, such as NAc D1 cells (Fig. 1E and Fig. S2A,B; this suggests that D1 and D2 cells studied here were largely non-overlapping as also found in ¹²), or NAc fast-spiking or ACh interneurons (n = 12 and 10, respectively, data not shown). However, orexin stimulation excited NPY_NAc interneurons (Fig. 1E, Fig. S1C-E), which according to previous work may indirectly stimulate D2_NAc cells by inhibiting D1_NAc cells¹³. The LH contains melanin-concentrating hormone (MCH) neurons that are distinct from orexin neurons⁴. We found, by rabies-assisted retrograde tracing of direct inputs to D2_NAc cells, that D2_NAc–projecting LH cells are more likely to contain orexin than MCH (Fig. 2); and optogenetic stimulation of NAc MCH axons did not change D2_NAc cell excitability (Fig. S2C,D). Together, these data demonstrate cell-type-specific LH→NAc circuitry that would be expected to directly and indirectly increase D2_NAc cell activity.

Fig. 1. Functional evidence for orexin_LH→D2_NAc excitation.

A. Left: targeting ChR2 to orexin cells. Right: expression of ChR2 in LH, representative example of 5 brains. 3V = third ventricle; Arc = arcuate nucleus; f = fornix; M,L,D,V = medial, lateral, dorsal, ventral.

B. Orexin cell axons (ChR2-mCherry, red) in NAc shell, representative example of 5 brains. DS = dorsal striatum, ac = anterior commissure.

C. Top: recording scheme. Bottom: defining electrical fingerprint of a D2_NAc cell (see Supplementary Methods).

D. Top row: effect of orexin axon photo-stimulation (blue) on D2_NAc cell membrane potential (left) or current (right); representative example of n = 22 cells. Lower rows: same experiment with orexin receptor blockade (TCS/SB, see Supplementary Methods), representative example of n = 5 cells.

E. Left: effect of bath-applied orexin on a D2_NAc cell (representative example of n = 8 cells). Right: group data (raw values for individual cells and mean±sem) for different NAc cell types. Numbers above data are p values from one-sample two-tailed t-tests (t,df from left to right = 1.08,7; 4.84,7; 5.56,10; 1.16,7; 0.5423,5).

Fig. 2. Anatomical evidence for a direct orexin_LH→D2_NAc circuit.

A. Left: targeting strategy. Right: eGFP expression in D2_NAc shell neurons, zoomed images are included to confirm labelling of cell bodies in NAc shell rather than core (representative example of n = 4 brains).

B. Left histology images: Orexin and MCH immunoreactivity in LH neurons that directly innervate D2_NAc cells, representative examples from n = 4 brains. Right plot: incidence of orexin and MCH immunoreactivities D2_NAc -projecting LH neurons (values are mean±sem and raw values for individual brains). Number above bars is p value from unpaired t-test with Welch’s correction, t,df = 8.611,4.723 (analysis of 1636 cells from 4 brains).

To probe the roles of natural orexin_LH and D2_NAc signals in action selection, we quantified the intensity of innate risk-avoidance behavior in two assays (open space and predator odor avoidance, see Supplementary Methods), combined with chemogenetic and pharmacological modulation of D2_NAc and orexin signals. Our main aim was to examine behavioral roles of natural (i.e. spontaneously-generated by the brain) orexin and D2_NAc cell signals, which we isolated by comparing behavior resulting from natural vs. signal-specifically suppressed brain function. Suppression of natural orexin signaling by orexin receptor antagonism (Fig. 3B-E) decreased risk-avoidance behaviors (left plots in Figs. 3C and E, relevant statistics are shown in green; importantly, these effects were not associated with locomotor sedation, Fig. S4E). Thus, orexin is necessary as well as sufficient^{6, 7, 9} for normal innate risk-avoidance (the sufficiency of NAc orexin stimulation for risk-avoidance was further confirmed in our behavioral assays, Fig. S3). By chemogenetically silencing or stimulating D2_NAc cells (Figs. 3A and S4A,B), we also found that D2_NAc cells were necessary (left plots in Figs. 3C and E, relevant statistics are shown in purple) and sufficient (Fig. S4A-D) for normal innate risk-avoidance.

Fig. 3. Combinatorial probing of roles of D2_NAc and orexin cells in risk-avoidance.

A. Left: targeting scheme for hM4Di-mCherry to D2_NAc cells. Middle: localization of hM4Di-mCherry cells in NAc shell (red); representative example of n = 4 brains. Right: Effect of CNO on a D2_NAc–hM4Di cell (representative example of n = 12 cells; mean±sem hyperpolarization = 3.25±0.37 mV, p < 0.0001 by two-tailed one-sample t-test; t,df = 8.687,11.

B. Examples of raw data from individual mice during a 15 min open-field test.

C. Group data for experiment in B (values are means±sem and/or raw values for individual mice), numbers above data are p values from Sidak’s or Tukey’s post-tests (left plot, Two-way RM ANOVA p < 0.0001 F(3, 42) = 13.41, n = 8 mice in each group), or from two-tailed paired t-tests (middle and right plots respectively: t,df=7.692,7 and t,df=7.692,7). Control mice were D2::ChR2 mice.

D. Examples of raw data from individual mice in the fox odor risk-taking test.

E. Group data for experiment in D (values are means±sem or raw values for individual mice), numbers above data are p values from Sidak’s or Tukey’s post tests (left plot, Two-way RM ANOVA p = 0.004 F(3, 18) = 6.348, n = 4 mice in each group), or from two-tailed paired t-tests (middle and right plots respectively: t,df=3.748,3 and t,df=3.748,3).

We next examined co-dependencies of risk-avoidance arising from natural orexin or D2_NAc signals, by combinatorial silencing of global orexin receptors and local D2_NAc cells. We found that risk-avoidance driven by natural orexin signals (isolated by quantifying behavior with and without orexin antagonist in individual mice) was abolished by D2_NAc cell silencing (Fig. 3C,E: center plots). This indicates that translation of natural orexin signals into innate risk-avoidance requires D2_NAc cells, and that other orexin-excited cells¹, including NPY_NAc cells identified here, are not sufficient for this translation. Finally, we analyzed whether natural orexin signals are necessary for the behavioral output D2_NAc cells. We found that risk-avoidance driven by natural D2_NAc cell activity (isolated by quantifying behavior with and without CNO in individual D2_NAc-hM4Di mice) was substantially reduced by orexin receptor blockade (Fig. 3C,E: left plots), suggesting that D2_NAc cells regulate behavior according to orexin tone. Thus, orexin is not only sufficient to stimulate D2_NAc cells (Fig. 1D), but also necessary for driving their behaviorally-relevant output in vivo (Fig. 3C,E).

Overall, these findings reveal an excitatory drive to D2_NAc cells that emanates from LH orexin cells, and show that D2_NAc cells are essential for innate risk-avoidance mediated by spontaneously-released orexin. These findings suggest interesting hypotheses for further study. For example, orexin cell activity is implicated in many types of reward-seeking^14,15, and it is tempting to speculate that the enhanced risk-avoidance caused by orexin may help avoid danger while seeking rewards. Since dopamine inhibits but orexin excites D2_NAc cells, D2_NAc -dependent risk-avoidance may be computed from antagonistic integration of these neurochemical representations of reward and stress. Insights into neuropsychiatric disorders linked to orexin signals⁶ will benefit from understanding how this LH system governs molecularly-defined brain switches of action strategies.