Loss of the habenula neuromodulator Kisspeptin1 disrupts learning in larval zebrafish

Learning how to actively avoid a predictable aversive stimulus involves two steps: recognizing the cue that predicts upcoming punishment, and learning a behavioral response that will lead to avoidance. In zebrafish, ventral habenula (vHb) neurons have been proposed to participate in both steps by encoding the expected aversiveness of a stimulus. vHb neurons increase their firing rate as expectation of punishment grows, but reduce their activity as avoidance learning occurs. How the change in vHb activity occurs is not known. Here, we ask whether the neuromodulator kisspeptin1, which is expressed in the ventral habenula, could be involved. Kiss1 mutants were generated with Cas9 using guide RNAs targeted to the signal sequence. Mutants, which have a stop codon upstream of the kisspeptin1 peptide, have a deficiency in learning to avoid a shock that is predicted by light. Electrophysiology indicates that kisspeptin1 has a concentration-dependent effect on vHb neurons: depolarizing at low concentrations and hyperpolarizing at high concentrations. These data suggest that as the fish learns to cope with a threat, kisspeptin1 may differentially modulate vHb neurons. This implies that learning a behavioral strategy to overcome a stressor is accompanied by physiological change in habenula neurons. Significance statement Learning to deal with adversity can positively affect one’s ability to cope with challenges in the immediate future. Control thus causes short-term change in the brain. Here, we show that the neuromodulator kisspeptin1 is required to learn to avoid a punishment. Expression and electrophysiological recordings suggest that this molecule functions by controlling the ventral habenula, a region of the brain that mediates fear by regulating serotonin release. Kisspeptin1 may be a potential player in resilience developed as a result of control, extending previous findings that it can reduce fear.


Introduction
When faced with an aversive stimulus, animals respond in a manner that is dependent on the context and experience. Upon first encountering such a threat in a novel environment, there may be panic and poorly directed attempts at escape. If they repeatedly encounter the threat, and become familiar with a safe escape route, they will be able to quickly remove themselves from danger. Better still, if the animals are able to recognize a cue that reliably predicts the impending threat, they will be able to escape before the aversive stimulus is present. This, in essence, is the phenomenon of active avoidance. Several theories have been proposed to explain the mechanism underlying active avoidance. In the two-factor theory, the animal first develops a fear of the conditioning stimulus (CS) that is paired with the aversive stimulus, by Pavlovian conditioning. Termination of the CS and the threat (unconditioned stimulus; US) then drives learning. Expectation has a critical role, and actions that lead to better than predicted outcomes are reinforced (1). The first time an aversive stimulus is encountered, the predicted outcome would be negative. However, if an escape route has been learnt, then the predicted outcome becomes positive.
How is active avoidance implemented in the brain? One structure that appears to be involved is the lateral habenula, which receives reward information from the globus pallidus of the basal ganglia (2). As first shown in monkeys, unexpected punishment leads to increased activity in lateral habenula neurons, while an unexpected reward leads to inhibition. In zebrafish, unexpected punishment leads to phasic activity in the ventral habenula (vHb) (3), which is the homolog of the mammalian lateral habenula. As the animal learns to associate a CS with the threat, there is tonic firing to the CS in vHb neurons. This causes excitation of serotonergic neurons in the dorsal raphe. As the animal learns to escape, there is decreased tonic firing. Tonic activity in the vHb has thus been proposed to encode aversive reward expectation value.
What is the mechanism of change in vHb activity as learning occurs? In zebrafish, Amo et al (3) have proposed that excitation is regulated via feedback from serotonergic neurons in the raphe to the entopeduncular nucleus, which is the teleost homolog of the basal ganglia. Whether additional mechanisms are involved is unknown. Here, we examine the possibility that a change in habenula neurons accompanies the learning process. In particular, we examine the potential involvement of the neuromodulator kisspeptin1, which is exclusively expressed in the vHb, together with its receptor (4). In zebrafish, two paralogs of kiss1 have been identified. These have non-overlapping expression, with kiss1 being restricted to the habenula, while kiss2 is expressed in the hypothalamus and posterior tuberculum (4). This allows a specific test of the role of kiss1 in the habenula via genetics.
In mammals, kiss1 is expressed in the hypothalamus, and is well studied in the context of reproduction. Burst firing of hypothalamic neurons leads to the release of Kisspeptin1, which causes immediate depolarization of gonadotropin releasing hormone (GnRH) neurons. Kisspeptin1 is also expressed in the hippocampus, where it causes an increase in EPSC and contributes to increased excitability (5). In the zebrafish habenula, kisspeptin1 has been proposed to depolarize vHb neurons, based on c-fos expression (6). However, delivery of Kisspeptin1 decreases fear, which is inconsistent with evidence that excitation of vHb is aversive (3). Moreover, lesioning of Kisspeptin1 receptorexpressing neurons in ventral habenula neurons mimics the effect of Kisspeptin1 delivery (6), which would not be expected if Kisspeptin1 causes depolarization. Hence, although Kisspeptin1 is well placed to alter responses of the habenula to aversive stimuli, how it functions is unclear. Additionally, whether it is involved in instrumental learning is unknown. Here, we address both these questions.

Results
As in adult zebrafish (7,8), kisspeptin1 is present in the ventral habenula of larval zebrafish, together with its receptor, kiss1rb (Fig. 1A, B). The kisspeptin system is thus expressed in an appropriate manner to locally regulate vHb neurons even at an early stage. To test whether kisspeptin1 is required for avoidance learning, we generated mutations in the kiss1 locus using CRISPR/Cas9. Two guide RNAs were designed to the signal peptide region of kisspeptin1 (Fig. 1C). These were injected into embryos at the 1cell stage, together with mRNA for Cas9. High-resolution melt analysis of genomic DNA derived from injected embryos indicated that the guide RNAs were effective. This was confirmed by sequencing: 8/8 injected embryos contained mutations at the target site.
Injected siblings were thus grown up. Sequencing of F1 fish indicated a transmission rate of 100%. Three alleles were obtained: one consisted of a 20 base pair deletion (kiss1 sq1sj ; Fig. 1D), while the other two were insertions of 7 and 8 base pairs (kiss1 sq2sj and kiss1 sq3sj ; Fig. 1D); kiss1 sq3sj contained a one base pair deletion, resulting in the same reading frame as kiss1sj sq2sj . All alleles gave rise to premature stop codons, upstream to the kisspeptin1 peptide (Fig. 1E). No label could not be detected in mutant fish by immunofluorescence with an antibody to the C-terminus of prepro-kisspeptin1 (7) (Fig. 1F), further indicating that the mutants lacked the peptide.
To test their ability to learn to avoid an aversive stimulus, juvenile mutants and wildtype animals (5-6 weeks of age) were tested individually in a tank with two compartments ( Fig. 2A). This is similar to a previously described apparatus (9), with the addition of a partial separator between the compartments as well as full automation (10).
Each compartment contained a red light (the CS) and electrodes that delivered an aversive shock. The light was turned on for 8 seconds in the compartment containing the fish, and co-terminated with the shock if the fish still stayed in the CS side. If the fish moved to the non-CS side and stayed there until the end of the CS presentation, no shock was delivered. Fish were exposed to 10 training trials. The cross score was calculated from the number of times an individual fish swam to the non-CS chamber before the light was turned off. Kiss1 mutants showed a lower cross score when compared to wild types (Fig. 2B), suggesting that kisspeptin1 is involved in learning active avoidance.
To determine how Kisspeptin1 affects habenula neurons, we performed whole cell patch clamp recordings from these cells and applied K-10, a conserved 10 amino acid peptide of Kisspeptin1. 1 µM TTX was added to the bath solution to globally block network activity. Cells were recorded in voltage clamp mode and were taken through a series of 500 ms voltage step protocol (Fig. 3A, lower panel). The cellular response (Fig.   3A, upper panel) was recorded before and after bath application of K-10. The recording was allowed to stabilize for 5 minutes in normal saline before K-10 was applied. Input resistances and holding currents did not change significantly with application of K-10 (p>0.05, SignTest). Difference currents were then calculated by subtracting the current values after K-10 application from the one before K-10 application (control). Difference currents calculated this way, showed that 5 µM K-10 application induced an outward current at depolarized potentials (Fig. 3B, n=4 cells from four 7dpf larvae). The mean peak amplitude of this current at 25 mV from four cells was 13.3 ± 8.9 pA. Next, to determine the specificity of this current, we bath applied a kisspeptin antagonist, 5 µM K-234 (11) along with 5 µM K-10, in a different set of experiments. The outward current was blocked in cells exposed to K-234 (Fig. 3D), confirming that this current was indeed

Discussion
We have investigated the role of the neuromodulator kisspeptin1 in active avoidance learning in zebrafish. Kisspeptin1 is expressed only in the habenula of zebrafish, primarily in ventral habenula neurons that project to the raphe (7). The kisspeptin1 receptor Kiss1rb is expressed in habenula neurons and not in downstream neurons. This, together with the defect seen in kiss1 mutants, suggests that active avoidance learning involves kisspeptin1 signaling within the habenula. However, given that loss of Kiss1 did not completely eliminate learning in all fish, other mechanisms must also be involved. These may include a change in input from the basal ganglia (entopeduncular nucleus), via feedback from the raphe (3).
A function for kisspeptin1 in regulating fear responses was suggested previously based on the finding that injection of the peptide into brain ventricle of adult zebrafish led to depolarization of habenula neurons, as assessed by c-fos expression, and a reduction of innate fear (6). Surprisingly, destruction of cells containing the kisspeptin1 receptor, including vHb neurons, had the same effect as administering kisspeptin1 peptides. This raises a conundrum: how can stimulating a neuron have the same effect as killing the neuron? The present results provide one way to resolve this contradiction, which is that Kisspeptin1 has a concentration-dependent effect on ventral habenula neurons: lowconcentrations lead to depolarization, while high concentrations lead to hyperpolarization.
A concentration-dependent effect for kisspeptin has been reported before, for example in GnRH neurons of the medaka (12), with only low concentrations causing depolarization.
Also, Ogawa et al reported that 10 -11 mol/g body weight of kisspeptin1 increased c-fos expression in the habenula, whereas a higher concentration of 10 -9 mol/g did not (13).
Successful learning may be accompanied first by kisspeptin1-mediated excitation, and then by inhibition of vHb neurons. This would increase aversive expectation whilst the CS is being associated with US, and reduce aversive expectation as the strategy to avoid the US is learnt. What controls the release of kisspeptin1 as a function of learning is unclear.
Kisspeptin may modulate neurons in the mammalian habenula, given that the kisspeptin receptor is expressed in the lateral habenula of mice and rats (14,15).
However, the role of this expression has not been addressed. Intriguingly, experiments in mammals demonstrate that the habenula and the downstream raphe (16) have a role in the sustained effects of a stressor on subsequent response to challenges (17). It is tempting to speculate that kisspeptin may be involved in this phenomenon. For example, if the animal has learned active avoidance, kisspeptin1 may be at a level that reduces subsequent activation of the raphe. The finding that delivery of kisspeptin1 peptides blocks fear in zebrafish (6), in a serotonin-dependent manner (18), is consistent with this idea. It would be interesting to test the effect of delivery kisspeptin to the habenula of mammals, to see if high levels of this peptide can reduce the effects of uncontrollable stress.

Materials and Methods
Animals. Experiments were carried out on the AB strain of zebrafish, Danio rerio, in accordance with protocols approved by the Institutional Animal Care and Use Committee.
Antibody labeling. Standard methods were used. Briefly, fish were fixed overnight at 4˚C, in 4% para-formaldehyde in PBS. A solution of PBS with 1% bovine serum albumin (Fraction V; Sigma), 1% DMSO and 0.1% Triton X-100 was the used to permeabalize the tissue and to dilute primary antibodies. The antibody to kisspeptin1 and the receptor kissr1b have been described previously (7,8). Alexa488-conjugated goat anti-rabbit antibodies (Invitrogen) were used at 1:1000 dilution, in PBS. Imaging was carried out using a Zeiss LSM510 confocal microscope, with a 40x water immersion objective. Active avoidance conditioning. Conditioning was carried out using a two-way chamber, essentially as described in (10), with the addition of a separator, made of matte black cardboard, between the two compartments. The conditioning stimulus (CS), a red LED, was delivered for 8 seconds, while the unconditioned stimulus (US), a 25 V pulse, was delivered for 100 milli-seconds. All fish were genotyped after the assay by sequencing.

Mutagenesis of the
Electrophysiology. Whole cell patch clamp recordings were done as described in (22) from vHb neurons in 6-8 dpf larvae. Briefly, the larvae were anesthetized in 0.01%