Control of defensive behavior by the nucleus of Darkschewitsch GABAergic neurons

ABSTRACT

The nucleus of Darkschewitsch (ND), mainly composed of GABAergic neurons, is widely recognized as a component of the eye-movement controlling system. However, the functional contribution of ND GABAergic neurons (ND_GABA) in animal behavior is largely unknown. Here, we show that ND_GABA neurons were selectively activated by different types of fear stimuli, such as predator odor and foot shock. Optogenetic and chemogenetic manipulations revealed that ND_GABA neurons mediate freezing behavior. Moreover, using circuit-based optogenetic and neuroanatomical tracing methods, we identified an excitatory pathway from the lateral periaqueductal gray (lPAG) to the ND that induces freezing by exciting ND inhibitory outputs to the motor-related gigantocellular reticular nucleus, ventral part (GiV). Together, these findings indicate the ND_GABA population as a novel hub for controlling defensive response by relaying fearful information from the lPAG to GiV, a mechanism critical for understanding how the freezing behavior is encoded in the mammalian brain.

INTRODUCTION

Freezing, a defensive response common in rodents and other animals, has been employed as an important index for different fear studies including conditioned fear [1–6] and innate fear [7–11]. Previous studies have identified several brain regions involved in freezing regulation [12]. In particular, the periaqueductal gray (PAG) is a common output for defensive behaviors and is involved in the control of various defensive behaviors [13]. Accumulating evidence suggests that the circuitry that mediates fear responses is complex and involves multiple independent circuits that process different types of fear [12]. However, how the fear response output center selectively generates distinct defensive responses such as freezing, flight or hiding remains largely unknown.

The nucleus of Darkschewitsch (ND), as one of the accessory oculomotor nuclei, was first described in 1889 and was postulated as being involved in eye movement control via zone second cervical spinal cord of the flocculus [14–17]. Immunocytochemical and in situ hybridization studies have identified that the majority of neurons in the ND are positive for glutamate decarboxylase (GAD) [18,19]. However, the anatomical outputs and inputs and the physiological functions of GABAergic neurons of the ND remain unclear.

Surprisingly, we observed that the ND_GABA neurons were selectively activated by the mouse predator odorant trimethylthiazoline (TMT) and foot shocks. We speculated that the ND_GABA neurons might be involved in fear-related-behavior regulation. If so, how the ND_GABA populations integrate fear information and control defensive responses remains unknown. To test this hypothesis, we combined electrophysiological recordings, optogenetics, chemogenetics and fiber photometry to investigate the ND_GABA neurons and found that ND_GABA neurons specifically influenced freezing-like behavioral responses, but not hiding- or fleeing-like responses.

We further identified ND_GABA neurons as a downstream target of the lateral PAG (lPAG) that relay fear information to the gigantocellular reticular nucleus, ventral part (GiV) and control the freezing behavior. Lastly, we found that the somatostatin (SOM) but not parvalbumin-positive (PV) GABAergic cell population of ND plays a key role in controlling freezing-like behavior. Together, these findings suggest that the SOM⁺ ND_GABA population might be an important behavioral output center that encodes freezing-like defensive behavior.

RESULTS

ND_GABA neurons selectively respond to fearful stimuli

Previous studies showed that foot shocks (painful stimulation) and predator-odorant trimethylthiazoline (TMT) successfully provoke a freezing behavior in free-behaving mice [1,3,8,10,11]. To explore the common freezing center in the mouse brain, we first examined Fos protein expression in the whole brain with the existence of freezing induced by foot shocks or TMT. Intriguingly, we observed massive Fos protein expression in the anterior ND (Fig. 1A–D), which has not been linked to fear before. In addition, our in situ hybridization studies confirmed that the majority of Fos protein positive cells in the ND were positive for the vesicular GABA transporter (Vgat) (∼70%; Fig. 1A–D), suggesting that GABAergic neurons in the ND might be involved in foot-shock- and TMT-induced fear regulation.

To further examine whether ND_GABA neurons specifically respond to different types of fearful stimuli, selectively targeted GABAergic cells of the ND with calcium (Ca²⁺) indicator GCaMP6m by injecting a Cre-dependent adeno-associated virus (AAV) vector carrying the GCaMP6m, and implanted an optical fiber into the ND of Vgat-Cre mice (Fig. 1E and F; Fig. S1A and B). We observed a large increase in Ca²⁺ activity when mice were exposed to TMT (5 s; Fig. 1G) or foot shocks (2 s; Fig. 1H), whereas no significant change in emitted fluorescence signals was observed with control stimulation (i.e. the lemon odor and 1% acetic acid; Fig. 1I and J) or in control-virus-injected mice (Fig. 1G–J).

Since fiber photometry does not provide information about the activity of individual cells, to further explore the correlation between individual neuron activity and defensive response, we next performed in vivo stereotrode recordings of the ND neurons by implanting a drivable electrode into the anterior ND (Fig. 2A). Although we found some cells were inhibited, the majority of active ND neurons were activated by TMT (33% vs. 4%; Fig. 2B–F). In particular, some of the activated cells dramatically increased firing frequency when mice were in a freezing state (Fig.2B, G–I). Moreover, some of the freeze-responding cells remained active for a long time when the animal showed freezing behavior (Fig. 2B). Together, these findings demonstrate that some ND_GABA neurons are highly responsive to fearful stimuli.

ND_GABA neurons control freezing behavior

To further investigate the potential causal role of ND_GABA neurons in underlying defensive responses, we injected an AAV-carrying Cre-dependent channelrhodopsin-2 (ChR2) into the ND nucleus of Vgat-Cre mice and implanted an optical fiber above the ND (Fig. 3A; Fig. S1C and D). Indeed, in ChR2-expressing mice, when compared to eYFP-expressing mice, optogenetic excitation (20 Hz) of the ND_GABA neurons immediately halted their movement (Fig. 3B and C; Video S1). However, activation of ND_GABA neurons did not influence other defensive responses, such as hiding behavior (Fig. 3D). Previous studies have shown that fearful stimulation rapidly induces defensive behavior and alters the autonomic nervous system in mice [11,20,21]. The activation of ND_GABA neurons indeed resulted in changes in autonomic functions, including a significant increase in pupil size (Fig. 3E) and a decrease in heart rate (Fig. 3F). Conditioning paradigms were used to demonstrate that animals avoid contextual cues previously associated with fear [22–26]. We also observed that optogenetic excitation of ND_GABA neurons promotes aversion by using the conditioning paradigm, whereas eYFP-expressing mice spent equal time in both chambers (Fig. 3G). In addition, we repeated this experiment by chemogenetic stimulation of ND_GABA neurons (Fig. S2). Compared with the control viral-vector injected animals (AAV-DIO-mCherry), upon clozapine-N-oxide (CNO, 1.0 mg/kg) administration, mice with AAV-DIO-hM3Dq injection showed high-level immobility and anxiety (Fig. S2B–E).

Although our in situ hybridization studies confirmed that the majority of neurons in the anterior ND of mice were Vgat positive, many vesicular-glutamate transporter 2 positive cells were observed in the caudal of the ND (Fig. S3A and B). However, activation of glutamatergic neurons in the ND had a much weaker effect on freezing behavior control (Fig. S3C and D).

To investigate the necessity of ND_GABA neurons in freezing behavior control, we examined whether optogenetic inactivation of ND_GABA neurons by light stimulation (473 nm constant light) of Guillardia theta anion channel rhodopsin 1 (GtACR1) caused a decrease in freezing response. Indeed, inactivation of ND_GABA neurons (Fig. 3H and Fig. S4) caused a significant decrease in immobility time induced by TMT (Fig. 3I and J), by an auditory cue in mice with previous fear-conditioned learning (Fig.3K and L), or by looming stimulation (Fig.3M–O), but without affecting the general locomotion in the open field test (Fig. S4B–D) and the hiding behavior provoked by TMT (Fig. S4E–G).

In summary, ND_GABA neurons are sufficient and necessary for regulation of freezing behavior.

Optostimulation of ND_GABA neurons did not promote sleep or seizure response

Since the ND has been postulated to be involved in controlling motor behavior, we performed a forced-swimming test with light stimulation of ND_GABA neurons and found that light stimulation failed to stop mice attempting to escape from the container, thus the immobility observed above was not a result of a loss of motor ability. However, immobility was observed when placing the animal in a container with less water so that it can stand at the bottom with hind paws (Fig. S5A–C; Video S2). Additionally, a simultaneous electroencephalogram (EEG) recording showed that light-induced brain waves and power intensity are different in rapid-eye-movement (REM) sleep, non-REM sleep and epilepsy [27–31] (Fig. S5D–F). Therefore, the cessation of movement did not result from falling asleep or having a seizure.

Mapping inputs of ND_GABA neurons

To identify how ND_GABA neurons integrate into the brain's ‘encephalic aversion system’ [32] to support freezing behavior, we used a rabies-virus-based (RV) trans-synaptic tracing strategy [33] to map whole-brain monosynaptic inputs onto ND_GABA neurons (Fig. 4A–C). Histological analysis verified the localization of starter cells in the ND (Fig. 4A). We observed the vast majority of RV-labeled cells located in fear-related brain regions such as the zona incerta (ZI) [34,35], the lateral hypothalamus (LH) [36,37], the superior colliculus (SC) [38–40], the lateral parabrachial nucleus (LPB) [41,42] and the lPAG [43] (Fig. 4B and C).

The PAG is an essential common output center for defensive behavior, as demonstrated before [12,13,44]. In particular, the glutamatergic population of the ventrolateral periaqueductal gray (vlPAG_Glu) is involved in freezing-like behavior control [45–48]. So, we hypothesized that the ND_GABA neurons might receive lPAG inputs and encode freezing-like defensive behavior. To investigate whether the lPAG_Glu→ND pathway is involved in any defensive behavior control, we optogenetically activated lPAG glutamatergic terminals in the ND and found that photostimulation immediately induced a freezing response (Fig. 4D). Additionally, to exclude the influence of the ND passing fibers from the PAG to other brain regions evoke the freezing behavior, we used dual-switch AAV-borne molecular payload technology [49] to express ChR2 genetically in lPAG_Glu neurons that specifically project to the ND. As expected, excitation of ND-projecting lPAG_Glu neurons promoted freezing behavior as well, whereas the control viral-vector injected animals showed no effects (Fig. 4E). These results verified that the ND, as a downstream target of lPAG_Glu neurons, controls freezing behavior.

ND_GABA neurons relay threat information from the PAG to the medulla

Next, to investigate the downstream target of ND_GABA neurons underlying immobility, we injected anterograde AAV-DIO-eYFP into the ND of Vgat-Cre mice and examined the anatomical organization of the ND GABAergic projection by brain-wide imaging of the ND neuronal axons expressing eYFP. We observed that GABAergic neurons project primarily to brain regions involved in movement control, such as the SC and GiV, and also project to several regions involved in fear regulation, including the laterodorsal tegmental nucleus (LDT), the PAG, the LH and the ZI (Fig. 5A–D). Given that previous studies have shown that the GiV is involved in immobility control [46,50], we speculated that the GiV as the downstream of ND_GABA neurons controls freezing-like behavior. We then examined the effects of optogenetically exciting ND_GABA neuronal terminals in the GiV. Compared to eYFP-expressing mice, ChR2-expressing mice showed immediately halted movement (Fig. 5E), an effect similar to that of direct stimulation of GABAergic neurons in the ND. To exclude the possibility of backpropagation, we also selectively manipulated the GiV-projecting GABAergic neurons in the ND and observed a robust freezing-like response (Fig. 5F). To further test whether the GiV is one of the downstream targets of ND_GABA neurons in freezing control, we directly manipulated GABAergic and glutamatergic neurons in the GiV and found that optogenetic inhibition of GiV_Glu neurons but not GiV_GABA neurons reliably promoted the freezing response (Fig. S6). More importantly, when selectively manipulating the GiV projecting GABAergic neurons in the ND, which receive lPAG inputs, we observed a significant freezing-like response as well (Fig. 5G).

In summary, we identified ND_GABA neurons, as a downstream target of the lPAG, relaying threatening information to the GiV to produce freezing behavior (Fig. 5H).

Somatostatin GABAergic neurons in the ND control freezing-like behavior

Fluorescence in situ hybridization (FISH) showed that most venus-labeled ND_GABA neurons in the Vgat-venus mice expressed SOM or PV (Fig. 6A and B; Vgat + SOM+: ∼37%; Vgat + PV+: ∼30%) and few cells with SOM and PV co-labeling (∼4.5%), suggesting the presence of at least two subtypes of GABAergic cells. Next, to test the role of PV and SOM neurons in freezing control, we injected a Cre-dependent AAV-vector carrying the ChR2 or GtACR1 into the ND of PV-Cre or SOM-Cre mice (Fig. 6C, G and K). Optogenetic excitation with ChR2 of the PV neurons neither enhanced freezing behavior (Fig. 6D) nor reduced freezing induced by TMT or conditioning cue with previous fear-conditioned learning (Fig. 6E and F). However, activation of the SOM neurons caused a strong increase in freezing response (Fig. 6G and H), pupil size (Fig. 6I) and a decrease in heart rate (Fig. 6J). Moreover, light inhibition of SOM neurons in the ND strongly decreased the freezing response induced by TMT or conditioning cue in mice with previous fear-conditioned learning (Fig. 6K–M). Whereas, in eYFP-expressing mice without ChR2 or GtACR1, light stimulation had no such effects. Together, these results indicate that SOM neurons in the ND are particularly important for defensive behavior control.

DISCUSSION

One of the major goals of neuroscience research is to understand how fear results in the expression of a range of adaptive or defensive behaviors which have a profound role in the survival of a species in a constantly changing environment. Indeed, it has long been assumed that an ‘encephalic aversion system’ [32] exists in the mammalian brain and that this system, in the presence of a threat, generates defensive responses such as freezing, fleeing or hiding [21]. However, fear is an incredibly complex emotion, and although a brain structure called the amygdala is often considered the ‘fear center’ in the brain, it is certain that the ability of the brain to generate and control defensive reactions involves many different brain regions [12]. Although different types of threats are detected by the brain via different sensory modalities, we believe that there are common threat integration and fear response output centers in the brain [12,13,51,52]. The PAG, as a common output for defensive behaviors [13,53]. In particular, the glutamatergic outputs to pre-motor targets in the medulla play a critical role in encoding freezing-like behavior [46] and the lPAG glutamatergic neurons mainly innervate medulla GABAergic neurons [54]. This suggests that lPAG glutamatergic neurons promote immobility by directly activating medullary GABAergic neurons. We unexpectedly found that the ND, as a downstream target of the lPAG, specifically mediates freezing-like defensive behavior through its inhibitory projections to the GiV (a subregion of the medulla), in part at least. We discovered that selective activation of ND_GABA neurons is sufficient to induce freezing-like behavior evaluated by previous studies [55,56]: (i) defensive behaviors such as freezing [21]; (ii) altering autonomic function such as changing heart rate during fear and anxiety [11]; (iii) pupil size, which is greatly increased during fear and aggression [57,58], but is strongly reduced during drowsiness and sleep [59,60]; and (iv) EEG (electroencephalogram) and EMG (electromyography)recording [11,27,28,31]. Moreover, the ND_GABA neurons have connections with many fearful-behavior-related brain regions (Figs 4 and 5; i.e. the PAG).

lPAG glutamatergic outputs can directly regulate the medulla to encode freezing-like defensive behavior. Here, we identified that medulla-projecting ND_GABA neurons, as a downstream target of lPAG glutamatergic neurons, encode freezing-like defensive behavior as well. This supports the existence of multiple parallel neuronal circuits in the brain to control a defensive response in order to facilitate more flexible switching between different defensive behaviors.

Medullary glutamatergic neurons are essential to support high-speed locomotion [61]. Thus, we hypothesized that ND_GABA neurons encode immobility, at least in part, by inhibiting medullary glutamatergic neurons, and result in halted locomotion. In addition, the GiV directly sends projections to the spinal cord and modulates cardiac response, thus the autonomic response evoked by manipulating the ND_GABA neurons could also result from the medulla [50,62–64]. We cannot exclude the possibility that other resources [65] in the medulla mediate ND_GABA projections and evoke defensive behavior and autonomic responses.

The GABAergic population in the ND contains both SOM⁺ and PV⁺ GABAergic neurons (Fig. 6), at least. The freezing-promoting effect is partly attributable to the SOM⁺ GABAergic neurons (Fig. 6G–M). Although previous studies have shown the reciprocal connections between SOM⁺ and PV⁺ neurons within the cortex or brainstem nucleus and their opposite roles in modulating conditioned social fear or innate fear [11,66], activation of PV⁺ GABAergic neurons in the ND had no effect on freezing regulation (Fig. 6C–F). However, we cannot exclude the possibility that the two populations have reciprocal connections in the ND, and they have distinct roles in controlling different defensive responses; it will be an important future research topic.

Fear is highly correlated with anxiety, panic disorder and post-traumatic stress disorder. However, there are currently no effective clinical interventions or treatments for these disorders. In the study, we identified an inhibitory subtype of ND neurons as a novel hub in the control of freezing behavior. Our results advance the current understanding of how threats selectively trigger freezing, a specific defensive response, via lPAG_Glu-ND_GABA-GiV_Glu circuitry, and provide precise anatomical and functional information that is important for the discovery and development of new therapeutic interventions for mood disorders.

MATERIALS AND METHODS

Mouse line

Adult C57BL/6J, Vgat-IRES-Cre (Jackson, strain name Slc32a1^tm2(cre)Lowl/J, stock number 016962), Vglut2-IRES-Cre (Jackson, strain name B6J;129S6-(FVB)-Slc17a6^tm2(cre)LowI/MwarJ, stock number 028863), SOM-Cre (Jackson, strain name Sst^{tm2.1(cre)Zjh}/J, stock number 013044), PV-Cre (Jackson, strain name B6.129P2-Pvalb^tm1(cre)Arbr/J, stock number 017320) and Vgat-Venus mice were used for the experiments. C57BL/6J mice were purchased from Vital River Experiment Animal Co., Ltd. Mice were housed 3–5 mice per cage on a 12 h light/dark cycle with lights on from 7 : 00 a.m. to 7 : 00 p.m. Mice had free access to food and water ad libitum. The mice were maintained at the animal facility of Zhejiang University, and all animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC).

Detailed materials and methods are available in the supplementary data.

DATA AVAILABILITY

All custom code used for analysis in this manuscript is available on request. All data are reported in the main text and supplementary materials, stored at Zhejiang University and available upon request.

ACKNOWLEDGEMENTS

We thank Dr. Chenyan Ma (SMART) and Dr. Xunyi Yan (East China Normal University) for critical reading of the manuscript.

FUNDING

This work was supported by grants from STI2030-Major Projects (2021ZD0203400 to Y.Q.Y.), the National Natural Science Foundation of China Major Project (T2293733, T2293730, 31970939 to Y.Q.Y.; 82288101, 82090033, U20A6005 to S.D.), the Natural Science Foundation of Zhejiang Province (LZ22H090001 to Y.Q.Y.; LR24C090001 to H.Y.), the Key R&D Program of Zhejiang Province (2022C03034 to Y.Q.Y.), the CAMS Innovation Fund for Medical Sciences (2019-I2M-5-057 to S.D.), and the Non-Profit Central Research Institute Fund of the Chinese Academy of Medical Sciences (2023-PT310-01 to Y.Q.Y.).

AUTHOR CONTRIBUTIONS

Conceptualization, H.Z., Y.Q.Y., H.Y. and S.D.; methodology, H.Z., J.L., Y.S. and Y.Q.Y.; analysis, H.Z., J.L., Y.S., L.S., Y.L., L.Z. and H.Y.; investigation, H.Z., J.L., Y.S., X.F. and B.H.; visualization, H.Z., Y.S. and H.Y.; writing, H.Z., H.Y., Y.Q.Y. and S.D.; funding acquisition, Y.Q.Y. and S.D.; resources, Y.Q.Y., S.D. and X.M.L.; supervision, Y.Q.Y., S.D. and H.Y.

Conflict of interest statement. None declared.

References

Author notes