Acetylcholine potentiates glutamate transmission from the habenula to the interpeduncular nucleus in losers of social conflict

Switching behaviors from aggression to submission in losers at the end of conspecific social fighting is essential to avoid serious injury or death. We have previously shown that the experience of defeat induces a loser-specific potentiation in the habenula (Hb)-interpeduncular nucleus (IPN) and show here that this is induced by acetylcholine. Calcium imaging and electrophysiological recording using acute brain slices from winners and losers of fighting behavior in zebrafish revealed that the ventral IPN (vIPN) dominates over the dorsal IPN in the neural response to Hb stimulation in losers. We also show that GluA1 α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor subunits on the postsynaptic membrane increased in the vIPN of losers. Furthermore, these loser-specific neural properties disappeared in the presence of an α7 nicotinic acetylcholine receptor (nAChR) antagonist and, conversely, were induced in brain slices of winners treated with α7 nAChR agonists. These data suggest that acetylcholine released from Hb terminals in the vIPN induces activation of α7 nAChR followed by an increase in postsynaptic membrane GluA1. This results in an increase in active synapses on postsynaptic neurons, resulting in the potentiation of neurotransmissions to the vIPN. This acetylcholine-induced neuromodulation could be the neural foundation for behavioral switching in losers. Our results could increase our understanding of the mechanisms of various mood disorders such as social anxiety disorder and social withdrawal.


INTRODUCTION
Fighting behavior in conspecific animals has commonly evolved in social animals as a way of adapting to conflict over limited resources. 1 It enables the survival of social species of wildlife by leading to the construction of territorial habitats and/or hierarchical social relationships for resource sharing and allocation. 2 Recent research has found that the initiation of conspecific fighting in specific genders and situations requires activity in sub-regions of the ventromedial hypothalamus and/or prefrontal cortex. [3][4][5][6][7][8][9] Fighting behavior is usually initiated with the appearance of mutual agonistic or aggressive behaviors and terminated with loser switching behaviors from aggression to submission. 10,11 Namely, it is essential to stop aggressive behaviors to determine social relationships without inflicting serious injury or death on the opponents. 2 Both inappropriate social aggression and persistent social withdrawal will disrupt our social relationships and social life. While studies of psychiatric patients show that social deficits such as social anxiety disorder 12 are challenging to find the neural basis of those symptoms, the detailed mechanism for regulation of status in social conflict has not been understood well. Switching behaviors from aggression to submission in losers of conspecific fighting could be a model for studies of dysregulation of social behaviors in humans, and revealing its neural foundation will help our understanding of social deficits in psychiatric disorders.
The habenula (Hb) is a common structure among vertebrates and its glutamatergic output regulates downstream monoamine systems. Previously, we showed that the Hb-interpeduncular nucleus (IPN) pathway controls the outcomes of fighting behavior in zebrafish. The zebrafish Hb-IPN pathway consists of three parallel pathways: lateral subnucleus of the dorsal Hb (dHbL) to the dorsal/intermediate IPN (d/iIPN), medial subnucleus of the dorsal Hb (dHbM) to the ventral/intermediate IPN (v/iIPN), and ventral Hb (vHb) to the median raphe (MR) [13][14][15][16] (Figures 1A and 1B). Silencing of the dHbL or dHbM outputs resulted in a decreased or increased rate of winning, respectively, while the experience of defeat decreased the dHbL-d/iIPN transmission and tended to potentiate the dHbM-v/iIPN transmission. [17][18][19] The plasticity of two parallel transmissions in the Hb-IPN pathway induced by the experience of defeat could be the neural foundation of behavioral switching from aggression to submission.
One key candidate molecule for plasticity in the Hb-IPN transmission pathway is acetylcholine (ACh), which is a classical transmitter commonly used in the nervous system. In addition to being a neurotransmitter in typical synaptic connections, ACh modulates synaptic activity mediated by other transmitters, such as dopamine release in the striatum 20 and glutamatergic plasticity in the hippocampus. 21 The Hb also has a cholinergic neural population, which uses ACh as a co-transmitter with glutamate. [22][23][24][25] Cholinergic neurons are localized in the dHbM in zebrafish Hb 26 and are thought to modulate glutamatergic dHbM-v/iIPN transmission by ACh release. In this study, we revealed that pharmacological manipulation of the a7 nicotinic ACh receptor (a7 nAChR) can switch neural states in the Hb-IPN pathway in brain slices from defeated fish by GluA1-mediated potentiation of the dHbM-v/iIPN transmission downstream of a7 nAChR.

RESULTS
Alpha7 nAChR is expressed in neurons of the v/iIPN and is responsible for the slow current induced by highfrequency stimulation of the fasciculus retroflexus To investigate the effects of ACh on the dHbM-v/iIPN neurotransmission, we first identified functional nAChR expressed in the vIPN. The IPN consists of several sub-regions, 25 and it is thought that neurons in the cell-dense center of the iIPN (ciIPN) receive inputs from the dHbM terminals with their dendrites distributed in the vIPN ( Figures 1C and 1D). Of 106 ciIPN neurons stained by biocytin loading successfully with their dendrites, most of them (103 cells, 97.2%) had dendrites in the vIPN, and only 3 ciIPN neurons (2.8%) had dendrites in the MR. Strong expression of various nAChR subunits has also been reported in the ciIPN, 26 and in situ hybridization for the a7 nAChR gene shows that it is strongly expressed in the ciIPN ( Figure 1E). Excitatory slow currents mediated by AChRs can be recorded in the IPN by high-frequency stimulation of cholinergic afferents in mice and larvae of zebrafish. 23,26 As such, we stimulated the fasciculus retroflexus, which is composed of afferents from the Hb nucleus, at high frequency in adult zebrafish brain slices and recorded ACh-dependent slow inward currents from ciIPN neurons ( Figure 1F). Most ACh-dependent currents were suppressed by the a7 nAChR antagonist methyllycaconitine (MLA), with no difference in amplitude observed with the addition of all-nAChR antagonists, mecamylamine and hexamethonium (F 2,12 = 25.089, p < 0.0001, one-way ANOVA with Tukey's post hoc test; Figures 1G and 1H). This suggests that these slow currents are mainly mediated by a7 nAChR. Therefore, we Article investigated the effects of a7 nAChR agonists and antagonists on the potentiation of the dHbM-v/iIPN neurotransmission.
Pharmacological manipulations of a7 nAChR switch dominance of the dHbM-v/iIPN transmission Two male zebrafish bred in isolation for more than 24 h were placed in the same tank for 30 min, and their fighting behavior was observed (Figure 2A). Acute brain slices of the winners and losers were prepared within 2 h and subjected to intracellular calcium imaging using a calcium-sensitive dye, Oregon green. The calcium response of the IPN to electrical stimulation of the Hb was evaluated ( Figures 2B-2E). To detect the relative difference in peak amplitude between the dIPN and the vIPN, each amplitude was normalized against the peak amplitude of response across the entire IPN.
Comparing the amplitude of calcium response between naive, winner, and loser groups, the loser group showed significantly lower values in the dIPN and significantly higher values in the vIPN, compared with the other groups (dIPN, F 2,30 = 5.50, p = 0.0092; vIPN, F 2,30 = 5.755, p = 0.0077, one-way ANOVA with Tukey's post hoc test; Figures 2F and 2G). These results are consistent with our previous paper, 17 indicating that we could detect the potentiation of the Hb-IPN transmission induced by the experience of fighting, again here. To compare the dominance of the dHbM-v/iIPN neurotransmission between groups, we calculated the vIPN/dIPN ratio, which is the value of the vIPN amplitude divided by the value of the dIPN amplitude, and obtained a significantly larger value in the loser group (F 2,30 = 6.819, p = 0.0036, one-way ANOVA with Tukey's post hoc test; Figure 2H). These data indicate that the potentiation of the Hb-IPN transmission induced by the experience of defeat in social fighting causes dominance of the dHbM-v/iIPN transmission over the dHbL-d/iIPN transmission.
After treatment of brain slices from the winner group with the a7 nAChR agonist PHA 543613 (PHA), we measured the calcium response of the IPN upon electrical stimulation of the Hb. We found that the amplitude of the calcium response in the vIPN became significantly larger than that observed in the dIPN, indicating that the dHbM-v/iIPN neurotransmission became dominant (winner, p = 0.359; winner + 1 mM PHA, p = 0.0010, paired t test; Figures  2I-2K), as observed in the loser group. Conversely, treatment of brain slices from the loser group with the a7 nAChR antagonist, MLA, abolished the dominance of the dHbM-v/iIPN neurotransmission, as observed in the winner and naive slices (loser, p < 0.0001; loser + 50 nM MLA, p = 0.840, paired t test; Figures  2L-2N). Nicotine (Nic) treatment of brain slices from the winner group produced similar results to PHA treatment, and co-administration of MLA with Nic abolished the effect. Conversely, the a3b4* and a4* nAChR antagonists (SR 16584 [SR] and dihydro-b-erythroidine [DHbE]) could not inhibit this effect (5 mM Nic, p < 0.0001; Nic + MLA, p = 0.722; Nic + 20 mM SR, p = 0.0006; Nic + 10 mM DhbE, p = 0.0009; paired t test; Figure 3A). These results indicated that the effect of Nic treatment was mediated by a7nAChR.
Comparing the vIPN/dIPN ratio between groups, the Nictreated and PHA-treated winner groups showed significantly larger values, which were comparable to those of the loser group, indicating that the dHbM-v/iIPN neurotransmission had become dominant (F 2,27 = 11.08, p = 0.0003, one-way ANOVA with Tukey's post hoc test; Figure 3B). Conversely, the MLAtreated loser group had a small ratio comparable to the winner group, indicating that the dHbM-v/iIPN neurotransmission dominance had disappeared (p = 0.0055, Student's t test; Figure 3C). In co-administration experiments of Nic and nAChR antagonists, the vIPN/dIPN ratio was significantly smaller only in the presence of MLA, again indicating that the effect of Nic treatment was mediated by a7 nAChR (F 3,47 = 9.079, p < 0.001, one-way ANOVA with Tukey's post hoc test; Figure 3D).
This trend was also consistent in inter-group comparisons of calcium response amplitudes, with Nic-and PHA-treated groups significantly lower in the dIPN and significantly higher in the vIPN than untreated winners (dIPN, Increased GluA1 subunits in the postsynaptic membrane of ciIPN neurons in losers and its regulation by a7 nAChR Next, we investigated the mechanism of potentiation observed in the vIPN of the loser group. To investigate changes in Values are presented as mean ± SEM. Statistical significance was defined as *p < 0.05; **p < 0.01; ***p < 0.001; NS, not significant, p R 0.05.

OPEN ACCESS
presynaptic release probability, we recorded a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptormediated evoked excitatory postsynaptic currents (AMPAR-eEPSCs) from ciIPN neurons in response to paired electrical stimulation with a 50-ms interval in the presence of 50 mM D-APV, an N-methyl-D-aspartic acid (NMDA) receptor blocker ( Figure 4A). A paired-pulse ratio (PPR) was calculated from the AMPAR-eEPSCs recorded from the brain slices of naive, winner, and loser groups. If the presynaptic release probability is increased in this potentiation, the PPR value in losers must be smaller than the other groups. However, there was no difference among the groups, indicating no increase in presynaptic release probability in losers (F 2,27 = 0.515, p = 0.6033, one-way ANOVA; Figure 4B). On the other hand, administration of a GluA1 and GluA3 homomeric AMPARs antagonist, philanthotoxin 74 (PhTx), resulted in a higher sensitivity of AMPAR-eEPSCs in the loser group (F 2,33 = 7.696, p = 0.0018, one-way ANOVA with Tukey's post hoc test; Figures 4C and 4D).
In addition, when AMPAR-mediated miniature EPSCs (AMPAR-mEPSCs) were recorded at 100 events/cell and their duration was examined, we found that short and long events were mixed, with a tendency for winners to have long events and losers to have short events ( Figures 4E and 4F). Comparing the time to peak and half-decay times for each group, we found that the winners showed longer values, and the losers showed shorter values compared with that in the other groups (time to peak, F 2,3397 = 333.3, p < 0.0001; half-decay time, F 2,3397 = 454.2, p < 0.0001; one-way ANOVA with Tukey's post hoc test; Figures 4G-4I, S1A, and S1B).
We also tested the sensitivity of these AMPAR-mEPSCs against PhTx by monitoring their frequencies before and after the PhTx application. For comparison, the AMPAR-mEPSC frequency in the presence of PhTx was normalized by the value in the pre-drug condition. The AMPAR-mEPSC frequency in the loser group showed a higher sensitivity to PhTx than that the winner group (p = 0.048, Student's t test; Figures 4J and 4K). (E) Comparison of normalized peak responses in the dIPN and vIPN among a7 nAChR agonist-treated and untreated winner groups using the same datasets as shown in Figure 2K and (A).
(F) Comparison of normalized peak responses in the dIPN and vIPN among nicotine and nAChR antagonists treated winner groups using the same datasets as shown in (A).
(G) Comparison of normalized peak responses in the dIPN and vIPN between a7 nAChR antagonist treated and untreated loser groups using the same datasets as shown in Figure 2N. Values are presented as mean ± SEM. Statistical significance was defined as **p < 0.01; ***p < 0.001; NS, not significant, p R 0.05. These results suggest a higher PhTx sensitivity in the loser group because of the higher representation of the AMPAR-mEPSCs with a shorter duration. Therefore, we examined the relationship between AMPAR-mEPSC duration and PhTx sensitivity by linear regression analysis. Averaged values of time to peak or halfdecay time before application of PhTx and normalized AMPAR-mEPSC frequency in the presence of PhTx were plotted for each neuron in a field, and linear fitting was applied for them. The results showed a significant positive correlation between the normalized frequency and duration of AMPAR-mEPSCs, and that cells with a higher PhTx sensitivity show a higher representation of shorter-duration AMPAR-mEPSCs (time to peak, R = 0.828, p = 0.0031; half-decay time, R = 0.892, p < 0.0001; Figure 4L), indicating that the shorter-duration event is more sensitive to PhTx. Considering that only GluA1 or GluA4-mediated currents have a shorter duration than currents mediated by other subunits, 27,28 and the effect of PhTx, a GluA1 and GluA3 antagonist, we concluded that the expression of GluA1 subunits is increased in postsynaptic membranes in loser ciIPN neurons. Next, we investigated the effects of a7 nAChR agonists and antagonists on the duration of AMPAR-mEPSCs. AMPAR-mEPSCs were recorded from each ciIPN neuron after brain slices were treated with Nic or PHA. Duration analysis showed that Nic or PHA treatment shortened the time to peak and halfdecay times in the winner group (time to peak, F 2,2897 = 460.2, p < 0.0001; half-decay time, F 2,2897 = 727.1, p < 0.0001; oneway ANOVA with Tukey's post hoc test; Figures 5A, 5B, S1C, and S1D). Furthermore, the effect of Nic in the winner group was suppressed by MLA, but not by the antagonists of other nAChR subunits, except for a partial inhibition of decay time with DHbE (F 3,2696 = 92.33, p < 0.0001, one-way ANOVA with Tukey's post hoc test; Figures 5C, 5D, S1E, and S1F), indicating that it is mediated by a7 nAChR. Conversely, MLA treatment prolonged the time to peak and half-decay times of AMPAR-mEPSCs in the loser group (time to peak, p < 0.0001; half-decay time, p < 0.0001; Student's t test; Figures 5E, 5F, and S1G). These findings indicate that a7 nAChR agonists and antagonists can switch the proportion of GluA1 subunits on the postsynaptic membranes of ciIPN neuron dendrites, as well as the dominance of the dHbM-v/iIPN neurotransmission.
Enhanced synaptic transmission to MR-projecting neurons in the ciIPN of losers It has been reported that trafficking of GluA1 into the postsynaptic site prior to other AMPAR subunits is necessary for induction of long-term potentiation in mice hippocampus 29,30 and that GluA1 is transported to silent synapses and unsilences them during long-term potentiation of synaptic transmission. 31 This may explain how increased GluA1 subunits cause dominance in the dHbM-v/iIPN neurotransmission. An increase in the number of GluA1 subunits in the postsynaptic membrane of ciIPN neurons may increase the number of active synapses, resulting in enhanced dHbM-v/iIPN neurotransmission. The analysis of mEPSCs is the conventional way to judge which property of the synapse has changed for the induction of synaptic plasticity regardless of presynaptic excitability. Differences in the amplitude of mEPSCs reflect changes in the conductance and/or the number of receptor channels in the postsynaptic membrane. Differences in frequency reflect presynaptic release probability and/or the number of active synapses. 32 We recorded AMPAR-mEPSCs from ciIPN neurons in acute brain slices at À60 mV of holding potential, reversal potential for Cl À channels, to exclude inhibitory currents. Conversely, GABA A receptor-mediated miniature inhibitory postsynaptic currents (GABA A R-mIPSCs) were recorded at 0 mV of holding potential, reversal potential for AMPARs ( Figure 6A). We could confirm successful recordings of mEPSCs distinctively from failures in patch-clamping by successful recordings of mIPSCs from the same neuron, even when we detected extremely low mEPSC frequency ( Figure 6A, bottom).
No statistically significant differences were detected in the amplitudes and the frequencies of AMPAR-mEPSCs between naive, winner, and loser groups, but the frequencies tended to be higher in losers, although not statically significant (amplitude, F 2,66 = 2.496, p = 0.090, one-way ANOVA; frequency, H = 5.6231, p = 0.060, Kruskal-Wallis; Figure S2A). To further analyze AMPAR-mEPSC frequencies in ciIPN neurons, we classified neurons according to their mEPSC frequency, and the relative distributions of neurons are summarized in Figure 6B. We found that the distribution of the low-frequency population (below 0.4 Hz) in the loser group was about 2.6 times lower than that in the winner group and that the distribution of the high-frequency population (above 1.6 Hz) was about 6.7 times higher. This result showed that although the change in the AMPAR-mEPSC frequencies of all analyzed neurons was moderate, the high-frequency population was significantly expanded in the loser, suggesting that the activities of specific cell groups in the ciIPN were switched.
For the classification of neural populations in the ciIPN, we examined where ciIPN neurons project and confirmed that they project to the MR ( Figure 6C). Axonal collaterals of MR-projecting ciIPN neurons terminate in the MR ( Figure 6C, double arrowheads), while the other branch further descends to the dorsal tegmentum area ( Figure 6C, double arrows). Of all ciIPN neurons that were successfully stained by biocytin loading from the whole-cell recording electrode, more than half of them had axons terminating in the MR, which was approximately four times the number of neurons whose axons descend only to the dorsal tegmentum area ( Figure 6D Values are presented as mean ± SEM. Statistical significance was defined as *p < 0.05; **p < 0.01; NS, not significant, p R 0.05. See also Figure S1. Values are presented as mean ± SEM. Statistical significance was defined as *p < 0.05; NS, not significant, p R 0.05. See also Figure S3. neurobiotin was injected into the MR for retrograde labeling of MR-projecting neurons ( Figure 6E), multiple somata of ciIPN neurons were stained ( Figure 6F, asterisks). For the authenticity of the retrograde staining, see Figure S2B.
To verify whether the neural activity of MR-projecting neurons is switched by the experience of defeat, we performed dye injections into the MR on acute slices and recorded neural activity from retrograde-labeled MR-projecting neurons in the ciIPN. We selectively recorded AMPAR-mEPSCs from MR-projecting neurons ( Figure 6G). The retrograde-labeled cells were visually confirmed, then stained with biocytin loaded from the recording electrode using a different color for further confirmation of recorded cells after neurophysiological experiments ( Figure 6H). Comparing the amplitude and frequency of AMPAR-mEPSCs recorded from MR-projecting neurons between naive, winner, and loser groups showed significantly higher frequencies in the loser group, but there was no significant difference in amplitude (frequency, F 2,12 = 4.982, p = 0.027, one-way ANOVA with Tukey's post hoc test; amplitude, F 2,12 = 3.017, p = 0.087, one-way ANOVA; Figures 6I-6K). These results indicate that synaptic transmission to MR-projecting neurons located in the ciIPN is potentiated in the loser group. In addition, considering the results from the analysis of PPR and AMPA-mEPSCs in ciIPN neurons (Figure 4), active synapses of MR-projecting neurons may increase in this potentiation.

Lower contribution of cvIPN neurons to potentiation of the dHbM-v/iIPN neurotransmission
We also recorded AMPAR-mEPSCs and GABA A R-mIPSCs from neurons in the center of the vIPN (cvIPN), a cell-dense region elongated vertically in the vIPN ( Figure S3A). Dendritic morphologies indicate that these neurons also receive input from the dHbM terminals in the vIPN ( Figure S3B). Most of cvIPN neurons (91.7%) with biocytin loading after recordings showed no efferent axons, and when an efferent axon was observed (8.7%), its target was not the MR.
Recorded miniature events were analyzed in the same way as for ciIPN neurons, but a comparison of amplitude and frequency among naive, winner, and loser groups found no significant difference or trend in AMPAR-mEPSCs (amplitude, F 2,44 = 0.083, p = 0.92, one-way ANOVA; frequency, H = 1.0546, p = 0.590, Kruskal-Wallis; Figure S3C). Further analysis of the relative distributions of neurons for each frequency band found no activity switching, as seen in ciIPN neurons. Furthermore, neurons with frequencies below 0.4 Hz accounted for over 70% of all cells within each group (Figure S3D), indicating a lower contribution of cvIPN neurons to the dHbM-v/iIPN transmission and its potentiation. Collectively, these results indicate that MR-projecting neurons in the ciIPN, not the cvIPN, are playing a key role in the potentiation of the dHbM-v/iIPN neurotransmission.
We also compared the amplitude and frequency of GABA A R-mIPSCs recorded from ciIPN and cvIPN neurons. Although there was no significant difference among naive, winner, and loser groups in any cases, the mIPSC frequency tended to be lower in ciIPN neurons of losers, and the amplitude tended to be smaller in cvIPN neurons of losers (amplitude in ciIPN neurons,  Figures S3E and S3F). Inhibitory input would not work as the main mechanism for the potentiation of the dHbM-v/iIPN neurotransmission.
The excitability of the Hb does not differ between winners and losers Finally, we assessed the contribution of the excitability of Hb neurons to the potentiation in the dHbM-v/iIPN transmission regulated by a7 nAChR. We recorded the spiking activities of Hb neurons induced by current injection during patch-clamp recordings in the dHbL and the dHbM in brain slices. The number of spikes in dHbL neurons induced by +20 pA and 500 ms of current injection was not different between the winner and the loser groups and was significantly smaller only in the naive group (F 2,30 = 8.68, p = 0.0011; one-way ANOVA with Tukey's post hoc test; Figures 7A and 7B). However, as the electrical stimulation on the Hb caused strong forced excitation of Hb neurons, it was likely that the same level of calcium response was evoked in the IPN of the naive group as that in the winner group ( Figures 2F and 2G).
We also recorded spiking activities from dHbL neurons in slices from the PHA-treated winner and the MLA-treated loser groups and found that the number of spikes in those neurons was not different from that in the drug-untreated winner or loser group (winner versus winner + PHA, p = 0.688; loser versus loser + MLA, p = 0.407, Student's t test; Figure 7B). No significant difference was detected for the number of spikes in dHbM neurons among groups (among naive, winner, and loser, F 2,32 = 0.93, p = 0.405; one-way ANOVA; winner versus winner + PHA, p = 0.371; loser versus loser + MLA, p = 0.349, Student's t test; Figures 7C and 7D). These results indicated that the experience of fighting behavior does not make any difference in the excitability of Hb neurons between winner and loser fish and that the pharmacological manipulations of a7 nAChR in brain slices do not affect the excitability of Hb neurons. Therefore, the excitability of Hb neurons does not contribute to the loser-specific neural properties in the IPN.

DISCUSSION
Using calcium imaging of acute brain slices from winners and losers of social fighting behavior, this study shows that the dHbM-v/iIPN neurotransmission is dominant in the Hb-IPN neural circuit of losers. We found that this dominance can be pharmacologically manipulated via a7 nAChR, where agonists and antagonists can switch dominance on and off, respectively. Furthermore, analysis of AMPAR-mEPSCs recorded from ciIPN neurons revealed that the experience of defeat induced an increase in GluA1 AMPAR subunits on the postsynaptic membrane of ciIPN neuron dendrites located in the vIPN and that this effect on GluA1 subunits can also be regulated by pharmacological manipulation of a7 nAChR. In addition, a combination of tract tracing and physiological recordings showed that the potentiation of transmission from the dHbM terminals to MRprojecting ciIPN neurons was induced in losers, potentially due to an increase in active synapses. Therefore, we suggest that ACh released from Hb terminals in the vIPN causes activation of a7 nAChR, resulting in an increase in GluA1 on the postsynaptic membrane of MR-projecting ciIPN neurons, which ll OPEN ACCESS increases active synapses on those neurons, resulting in potentiation of the dHbM-v/iIPN neurotransmission. This switching between neural states may be the neural foundation of behavior switching from aggression to submission at the end of fighting behavior.
Cholinergic transmission in the dHbM-v/iIPN pathway ACh-dependent excitatory slow inward currents were recorded from ciIPN neurons in adult zebrafish brain slices following a high-frequency stimulation of the fasciculus retroflexus. Similar to slow inward currents induced by frequent stimulation of cholinergic afferents in mouse brain slices and zebrafish larval dissected brains, 23,26 the slow currents recorded in this study were nAChR dependent. Application of a receptor subtype-specific antagonist showed mediation of a7 nAChR. Since a7 nAChR is highly permeable to calcium, 33 these cholinergic slow currents are expected to be accompanied by calcium influx. In this study, pharmacological activation of a7 nAChR increased GluA1 on the postsynapse. In addition, it has previously been reported that stabilization and aggregation of GluA1 on the postsynapse by Nic requires an a7 nAChR-mediated calcium influx in the mouse hippocampus. 34 As such, we suggest that the influx of calcium activates calcium-dependent second messengers that promote the transport and anchoring of GluA1 to the postsynapse. Phosphorylation of the C terminus of GluA1 35 and binding of GluA1 to the actin filament at the postsynaptic membrane 36 may occur, but further studies are needed for confirmation.
In addition, in this study, the ACh-dependent slow current in this paper showed a sudden appearance and shorter duration than those recorded from other model systems. 23,26 If the cholinergic slow currents are due to the volume transmission of ACh discussed in studies of mouse brain slices, 23 the shorter duration may be due to slow accumulation and rapid washout of ACh as our zebrafish IPN samples were very small and located on open brain slices. On the other hand, the homo-pentamer structure of a7 nAChR equips 5 agonist binding sites, which requires occupation of over 3 sites with agonists to open the channel. The greater the number of bound agonists, the greater the current produced, 37 and the relationship between agonist concentration and channel conductance will not be linear in the initial phase of a7 nAChR activation. These suggest that the slow accumulation of ACh in our slices enabled us to see the initial non-linear increase of a7 nAChR current conductance which is caused by the nature of agonist bindings. In the closed in vivo environment, it is plausible that cholinergic slow currents with a longer duration are generated. Therefore, ACh accumulated locally in the IPN is thought to induce a sustained calcium influx in ciIPN neurons via a7 nAChR and efficiently activate calcium-dependent second messengers.
We assume that aversive information such as pain from being bitten by an opponent induces the firing of cholinergic Hb neurons. The Hb receives input from the hypothalamus which conveys information about aversive stimuli in mice, 38,39 and the hypothalamus-Hb connection has also been confirmed in zebrafish. 18,40 The main target of the hypothalamus is thought to be the lateral Hb (equivalent to the vHb in zebrafish), but possibly some population of the hypothalamic afferents activates cholinergic Hb neurons. The repeated bites will cause the bursting of cholinergic Hb neurons, accumulation of ACh in the vIPN, and potentiation of the dHbM-v/iIPN in losers. This is consistent with the fact that fish, which are given more bite by the opponent during fighting behavior, tend to become losers. 17 We need further research such as observation of Hb activities during fighting behavior for the validation of this idea.
Contributions of the GABAergic system and AChRs in the Hb-IPN transmission to switching from aggression to submission The mechanism described above could underlie the switching of potentiation of the dHbM-v/iIPN transmission by the pharmacological activation/inactivation of a7 nAChR. Although the released acetylcholine possibly affects presynaptic Hb terminals in the vIPN by direct or indirect mechanism, the results from analysis of PPR and Hb neuron excitability proved that it does not change presynaptic release probability or presynaptic excitability. Therefore, we concluded that a7 nAChR-mediated potentiation is a postsynaptic event. Also at behavior level, administration of Nic by intraperitoneal injection reduced aggressive behavior in mice, and this effect was suppressed by co-administration of MLA. 41 Such switching of neural state underlying behavior by ACh is also seen in the hippocampus, which is necessary for contextual memory. When the ACh concentration is high, a phase lock with theta oscillation occurs, and afferent input is enhanced, resulting in a coding state where memory acquisition is possible. Conversely, when the ACh concentration is low, the phase lock weakens, leading to a retrieval state where consolidation occurs synchronously with sharp wave ripples. 42 This coding/retrieval state switching is enabled by the action of GABAergic transmission and muscarinic ACh receptors (mAChRs) receptors in the hippocampus together with nAChRs, including a7 nAChR. 42,43 Although this study focused on the role of a7 nAChR, mAChR and the GABAergic system may also be involved in neural state switching in the IPN in a different aspect. In zebrafish, the dHbMv/iIPN transmission suppresses a competing parallel pathway, the dHbL-d/iIPN transmission, via GABA B receptors on the dHbL axon terminals. 44 In addition to potentiation of one side of the parallel pathway, such trans-inhibition of parallel pathways would also be necessary for efficient switching of the neural states. In mice, GABA released from IPN neurons enhances synaptic transmission from Hb cholinergic axons to the IPN via GABA B receptors located at cholinergic presynaptic terminals in the IPN. 45,46 This phenomenon may be induced simultaneously with the postsynaptic regulation involving a7 nAChR, as shown here. Those reports in fish and mice indicate that the GABAergic system in the IPN acts inhibitorily in the dHbL-d/ iIPN transmission, but also acts excitatorily in the dHbM-v/iIPN transmission when the dHbM-v/iIPN pathway is activated. In our results, partial decreases of GABA A R-mIPSCs in both the ciIPN and the cvIPN were found without statistical significance, which suggests that disinhibition of IPN neurons also happened simultaneously with the potentiation of the dHbM-v/iIPN transmission. Therefore, we also need to pay attention to the GABAergic system for further investigation of the Hb-IPN transmission.
In this study, an analysis of nAChRs other than a7 nAChR found that the a3b4* nAChR antagonist tended to reduce the dIPN/vIPN ratio of calcium response even without statistical significance ( Figure 3D) and that the effects of Nic on AMPAR-mEPSCs were partially suppressed by the a4* nAChR antagonist ( Figure 5C). These results suggest that other unknown mechanisms are possibly involved in the ACh-mediated potentiation of the dHbM-v/iIPN neurotransmission. Even if such a mechanism exists, we still concluded that it will not be the main system, since the a7 nAChR agonist/antagonist alone was sufficient to switch loser-specific properties in the dHbMv/iIPN transmission.

Possible roles of the dHbM-v/iIPN-MR pathway in the switching of behaviors
We identified MR-projecting ciIPN neurons as cells that play a key role in the potentiation of the dHbM-v/iIPN transmission. In mammals, it has been reported that suppression of serotonergic transmission by serotonin autoreceptors, 5-HT 1A and 5-HT 1B , on serotonin (5-HT) neurons causes a reduction of aggressive behaviors, where aggressive behaviors are accompanied by excitation of 5-HT neurons. [47][48][49] Similarly, administration of a 5-HT 1A receptor antagonist promoted aggression in zebrafish. 50 Considering ciIPN neurons are GABAergic, 51 potentiation of the dHbM-v/ iIPN transmission may suppress the activity of 5-HT neurons in the MR, and play a role in stopping aggressive behavior.
On the other hand, stress resilience decreases when brain 5-HT levels decrease, making it difficult to select active responses in mammals. 52 Inhibition of 5-HT neurons by potentiation of the dHbM-v/iIPN transmission in losers possibly induces a reduction of resilience, which may contribute to stopping aggressive behavior and inducing passive submission. It is supportive that the dHbL-d/iIPN silenced zebrafish larvae, where the dHbM-v/iIPN transmission was more active than the dHbL-d/ iIPN, shows helpless behavior 53 In humans, selective serotonin reuptake inhibitors (SSRIs), which enhance serotonergic transmission, are the first-line pharmacologic treatment for social anxiety disorder 12 and successfully treated patients who showed social withdrawal. 54 On the other hand, socio-economic status in a community correlates with the responsivity of the serotonergic system to fenfluramine administration 55 and 5-HT transporter-linked polymorphic region genotypes. 56 Taking the evolutionary conservation of the Hb-IPN-MR pathway into account, these studies suggest that the dHbM-v/iIPN-MR pathway also controls social behaviors in humans through the regulation of 5-HT transmission.
It has also been reported that the dHbL-d/iIPN silenced fish do not exhibit flight behavior but freeze against conditioned stimulus in cued fear conditioning 15 and show an inability to adapt to new rules in conditioned place-avoidance learning. 57 In addition, some neural populations in the dHbM respond to aversive olfactory cues and are thought to mediate aversive behavioral responses. 58 60 As such, the Hb-IPN possibly has a function in controlling the switching of multiple behaviors in various situations, and it should be further investigated how such switching of behavior in these various situations is interrelated.
In conclusion, physiological investigation of the Hb-IPN transmission after social fighting revealed that the experience of defeat induces a loser-specific neural state where ACh switches on the potentiation of the dHbM-v/iIPN transmission and that the potentiation will cause inhibition of 5-HT neurons in the MR to make fish prone to give up social fighting by reducing resilience in social conflict behaviors. This ACh-induced neuromodulation would be the neural foundation for switching from aggression to submission in losers at the end of fighting behavior. These findings may promote our understanding of the mechanisms of various mood disorders.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
day, fish were paired and habituated in the same test tank for over 15 min with an opaque partition. At the beginning of the test, the partition was removed to allow for interaction, and the fish were left for 30 min in total. Behaviors were recorded on video and determination of the winner and loser were checked offline. Fish pairs that did not show or end their fighting during the 30 min were excluded. After the behavior test, the fish were returned and kept in the home tank until the start of further experiments. For the naive group, fish were subjected to isolation only under the same conditions as above.

Slice preparation
Fish were anesthetized in cold water, and whole brains were isolated in a cold cutting solution (118 mM N-methyl-D-glucamine, 2.9 mM KCl, 20 mM HEPES, 25 mM Glucose, 1.2 mM N-acetyl Cysteine, 10 mM MgSO 4 ,7H 2 O, 0.5 mM CaCl 2 ,2H 2 O, pH = 7.8, saturated with O 2 ). Brains were embedded in 1.3%-3% low-melting-temperature agarose (Agarose XP, NIPPON GENE) in a cutting solution and sagittally sliced at 120 mm for the IPN electrophysiology and 300-400 mm for the Hb electrophysiology and calcium imaging using a vibrating microtome (LinearSlicer PRO7, DOSAKA EM). The whole left Hb and its connection to the IPN were kept intact for calcium imaging. Brain sectioning was performed in the oxygenated cutting solution on ice.
We carefully prepared brain slices in the same manner every time to minimize the differences among individual slices caused by sectioning. Most of the FR and the ciIPN were kept intact by adjusting the cut surface of slices under the stereo microscope so that the slice contains 2/3 to 3/4 of the entire IPN, and the amount of the IPN remaining in the slice was estimated from the size of the vIPN exposed on the ventral side of the slice ( Figure S4A). The bipolar electrode for the Hb stimulation was designed for the insertion into the dHbL with one tip and into the dHbM with the other tip to stimulate both subnuclei ( Figure S4B). The bipolar electrode for the FR stimulation was always placed on the slices so that the FR was positioned between two tips of the electrode. The position of the FR and the boundary between the dHbL and the dHbM which could be recognized as a dent on the surface ( Figure S4B arrow) were distinguished visually under the microscope. The typical shape of the left Hb, a dent on its surface, was checked for each fish to find and exclude the rare case of situs inversus, left-right inversion of the internal organ.

Intracellular calcium imaging
Slices containing the Hb-IPN connection were loaded with 3-4 mM Oregon green BAPTA-1, AM (OGB, molecular probes) in recovery solution with 0.02% Pluronic F-127 and 0.01% Cremophor EL for 1 h, and washed in recovery solution without the dye for 30-50 min. For drug treatments, drugs were also included in the solution. OGB-loaded slices were transferred to a recording chamber on a largefield microscope (MVX10, Olympus). Fluorescent signals were detected with a high-speed charge-coupled device camera system (MiCAM02, Brainvision) at 27 C-29 C. Images were acquired by BV_Ana (Brainvision) and binned to 384 x 256 pixels at 10 Hz. Calcium responses were evoked by the application of electrical stimuli (3.5-4 mA, 100 ms duration, 40 ms intervals, 5 times) on the Hb with a bipolar tungsten electrode (self-made) connected with an isolator unit (ISO-Flex, A.M.P.I.) and controlled by a stimulator (SEN-8203, Nihon Kohden). Data for a single response comprised an average of six repeated recordings with 20 s intervals.
Each set of imaging data was normalized to the average fluorescence levels over 300 ms before electrical stimulation and filtered with a median filter (3 x 3 pixels). To make figures for each representative sample, data were normalized with the maximum value in the first frame, and a Cubic filter (3 x 3 pixels) was further applied for pseudo-color image data. The ROIs for the dIPN and vIPN were manually selected so that dIPN ROI would cover D-i, D-ii, and vIPN ROI would cover V-i, V-ii, V-iii. 25 The peak values of averaged fluorescence signals within ROIs for the dIPN and the vIPN were normalized with the peak value of that within the ROI for the whole IPN area. One frame of data (1 frame = 100 ms) at the peak of the response was always the second frame after the Hb stimulation and was used for amplitude comparisons.

Electrophysiological recording
Slices were transferred to a recording chamber on an upright microscope (BX51WI, Olympus) equipped with a charge-coupled device camera. Whole-cell patch-clamp recordings were performed from ciIPN, cvIPN, dHbL, or dHbM neurons. We identified the ciIPN and the cvIPN by the difference in the density of distributed cell bodies. They can be distinguished from adjacent sub-regions, in which neurons are located sparsely. We defined the boundary of the ciIPN and the cvIPN ( Figure S4C, arrow) at the same level in the dorsoventral axis with the boundary of the iIPN and the dorsal part of the vIPN (indicated as I and V-ii in Figure S5C, respectively). Cell bodies were visualized with differential interference contrast microscopy and approached by recording pipettes using a visual guide. An extracellular solution containing (in mM): 134 NaCl, 2.9 KCl, 10 HEPES, 10 glucose, 1.2 MgCl 2 , and 2.1 CaCl 2 , pH = 7.8, saturated with O 2 was perfused continuously at 27 C-29 C.
For recordings of GABA A R-mIPSCs and AMPAR-mEPSCs from the same IPN neuron, 0.2 mM tetrodotoxin (Nacalai tesque) and 50 mM D-APV (Alomone Labs) were added to the extracellular solution. For AMPAR-mEPSCs recordings, 10 mM SR 95531 (Alomone Labs) was further added. For AMPAR-eEPSCs recordings, 50 mM D-AP5 and 10 mM SR 95531 were added to the extracellular solution. The stimulation of the Hb efferent bundles in the fasciculus retroflexus was applied by a bipolar tungsten electrode (self-made) connected with an isolator unit (ISO-Flex, A.M.P.I.) and controlled by a stimulator (Master-8, A.M.P.I.) (100 ms, 0.5-4 mA square