The Role of Vesicular Glutamate Transporter Type 3 in Social Behavior, with a Focus on the Median Raphe Region

Social behavior is important for our well-being, and its dysfunctions impact several pathological conditions. Although the involvement of glutamate is undeniable, the relevance of vesicular glutamate transporter type 3 (VGluT3), a specific vesicular transporter, in the control of social behavior is not sufficiently explored. Since midbrain median raphe region (MRR) is implicated in social behavior and the nucleus contains high amount of VGluT3+ neurons, we compared the behavior of male VGluT3 knock-out (KO) and VGluT3-Cre mice, the latter after chemogenetic MRR-VGluT3 manipulation. Appropriate control groups were included. Behavioral test battery was used for social behavior (sociability, social discrimination, social interaction, resident intruder test) and possible confounding factors (open field, elevated plus maze, Y-maze tests). Neuronal activation was studied by c-Fos immunohistochemistry. Human relevance was confirmed by VGluT3 gene expression in relevant human brainstem areas. VGluT3 KO mice exhibited increased anxiety, social interest, but also aggressive behavior in anxiogenic environment and impaired social memory. For KO animals, social interaction induced lower cell activation in the anterior cingulate, infralimbic cortex, and medial septum. In turn, excitation of MRR-VGluT3+ neurons was anxiolytic. Inhibition increased social interest 24 h later but decreased mobility and social behavior in aggressive context. Chemogenetic activation increased the number of c-Fos+ neurons only in the MRR. We confirmed the increased anxiety-like behavior and impaired memory of VGluT3 KO strain and revealed increased, but inadequate, social behavior. MRR-VGluT3 neurons regulated mobility and social and anxiety-like behavior in a context-dependent manner. The presence of VGluT3 mRNA on corresponding human brain areas suggests clinical relevance.


Introduction
Vesicular glutamate transporter type 3 (VGluT3) is the latest transporter discovered from the solute carrier (SLC) 17 protein family (Fremeau et al., 2002), and thus, its role in behavior is still largely unknown.Homozygous VGluT3 knock-out (KO) mice are viable (Horvath et al., 2018) and their behavioral characterization is ongoing (Seal et al., 2008;Amilhon et al., 2010;Divito et al., 2015;Balazsfi et al., 2016Balazsfi et al., , 2018a;;Fazekas et al., 2019).However, detailed investigation of its contribution to social behavior is still missing, despite the importance of these interactions in everyday life as well as its disturbances in psychopathologies.
Numerous brain regions contain VGluT3-positive neurons, such as the hippocampus, cortex, and median raphe region (MRR) in the midbrain.The MRR is known as one of the serotoninergic nuclei, which provides serotoninergic innervation to the forebrain (Dahlstrom and Fuxe, 1964;Vertes et al., 1999).However, only 5.742% of the MRR neurons have been shown to express serotonin (5-HT; Sos et al., 2017).The rest of the cells express numerous other neurotransmitters and neuromodulators, such as GABA (Calizo et al., 2011;Sos et al., 2017), dopamine (Jahanshahi et al., 2013), and glutamate, characterized by VGluT2 (Szonyi et al., 2019) or VGluT3 (Fremeau et al., 2002;Herzog et al., 2004;Amilhon et al., 2010;Sos et al., 2017).According to our current knowledge, ∼10.651% of the MRR cells contain VGluT3 as a neuronal marker, while ∼3.603% of the MRR cells show serotonin and VGluT3 coexpression as well (Sos et al., 2017).These VGluT3-positive cells target key areas such as the medial prefrontal cortex (mPFC), hippocampus, thalamus, hypothalamus, and dorsal raphe (Jackson et al., 2009;Szonyi et al., 2016).Numerous studies have investigated the role of MRR in behavior (Wang et al., 2015;Bland et al., 2016;Balazsfi et al., 2017;Turcotte-Cardin et al., 2019), but they were solely concentrating on its serotonergic content.However, it was demonstrated that at least in certain areas (e.g., mPFC, hippocampus, medial septum), functional glutamate receptors are expressed in the postsynaptic sites of these MRR-VGluT3-positive projections (Szonyi et al., 2016).The optogenetic stimulation of the whole MRR decreased aggressive behavior in a phasic manner in mice, which was accompanied by increased serotoninergic, GABAergic, and glutamatergic neurotransmitter levels in the mPFC (Balazsfi et al., 2018b).While it is established that MRR regulates social behavior, the specific role of its VGluT3+ neurons is still unknown.
Our aim was to characterize the role of VGluT3 in social behavior.With the usage of VGluT3 knock-out (KO) mice, the global role of this transporter could be revealed; however, life-long, constitutive lack of a gene product could result in compensatory mechanisms, ultimately masking behavioral effects.The chemogenetic manipulation of MRR-VGluT3 neurons would shed new light on the role of this specific subpopulation giving the opportunity for both stimulatory and inhibitory short-term interventions.The comparison of the two models would provide deeper insight on the role of VGluT3.

Animals
All mice (C57BL/6J background) were obtained from the local colonies of the Institute of Experimental Medicine, Budapest, Hungary.VGluT3 wild-type (WT) and KO mice were bred in heterozygous mating pairs.Their genotype was determined by PCR from a small tail sample collected from 2-to 3-day-old animals.VGluT3-Cre mice originated from homozygous mating.During the test battery adult male mice (14-15-week-old) were housed in Makrolon cages (40 cm × 25 cm × 26 cm) under a "reversed" 12 h light/dark cycle (lights on at 7 P.M., 21 ± 1°C, 50-60% humidity), with food (standard mice chow, Charles River) and tap water available ad libitum.Thirty minutes before the start of the first experiment, the animals were separated to increase social interest during subsequent testing.The experiments were conducted between 9 and 13 h, during the subjective dark period of the mice.The KO-WT and VGluT3-Cre animals were tested in different series with minor differences in the protocol (Fig. 1A): 1. Series 1: WT (N = 10) and VGluT3 KO (N = 11) animals were compared.2. Series 2: VGluT3-Cre animals went under stereotaxic surgery.Three subgroups were formed based on the injected viral vectors containing "designer receptors exclusively activated by designer drug" (DREADD) sequences: control (N = 8), excitatory (N = 13), and inhibitory (N = 15).c-Fos activation was studied in a different batch of animals from those used in the behavioral studies.Wild-type (N = 5) and VGluT3 KO (N = 8) mice were perfused 90 min after a social interaction test (SIT).In the case of VGluT3-Cre mice (N control = 6; N excitatory = 3), separate animals were injected intraperitoneal (i.p.) with clozapine-N-oxide (CNO, ligand of DREADD; CAS No.: 34233-69-7, Tocris Bioscience, 1 mg/kg/10 ml saline) and perfused 30 (time for CNO to take effect) + 90 min (c-Fos expression peak) after injection.
All tests were approved by the local committee of animal health and care (PEI/001/33-4/2013, PE/EA/254-7/2019) and performed according to the European Communities Council Directive recommendations for the care and use of laboratory animals (2010/63/EU).

Experimental design
Thirty minutes before the experiments (except for social discrimination test; SDT), VGluT3-Cre animals were injected intraperitoneally with CNO (Chaves et al., 2022).There were 48-72 h intervals between behavior tests (except for SDT) to ensure the clearance of previously injected CNO (Fig. 1A).For comparability the timing was the same in both series.
The experimental room was dark, only lit by infrared light [except for SIT and elevated plus maze (EPM)].The experiments were recorded and analyzed later by computer-based event recorders (H77; Solomon Coder (https:// solomoncoder.com);Noldus EthoVision).The test apparatuses were cleaned with 20% ethanol between animals except for SIT and RIT (with bedding).
In the end, the animals were anaesthetized by ketamine-xylazine and fixed by transcardial perfusion with phosphatebuffered saline (PBS; Molar Chemicals) and 4% paraformaldehyde (PFA; Molar Chemicals).After 24 h postfixation in PFA, the solution was changed to azide-PBS (Na-azide from Sigma-Aldrich).Before cutting, the brains were immersed in 20% glucose (Reanal) azide-PBS for at least 24 h.Thirty-micrometer slices were cut with sledge microtome and stored in cryoprotectant (all components from Molar Chemicals) at −20°C.Immunohistochemistry was done for RFP and VGluT3 in order to validate the virus expression (Fig. 1B) and cell specificity (Fig. 1C).Staining for c-Fos was conducted to explore neuronal activity.
Open field.Open field (OF) measures the locomotor activity and anxiety-like behavior of mice.Animals were put into an empty white plastic box (40 cm × 36 cm × 15 cm; Fig. 2A).Distance traveled, frequency, and time spent in periphery and centrum (75% of the arena) were analyzed by the Noldus EthoVision XT 13.
Sociability.Right after OF, two identical wired cages (with identical weights on them) were placed into the empty plastic boxes (object habituation phase; Fig. 2A).Based on their location, they were either labeled as "left" or "right."After 5 min, an unknown, smaller, juvenile male conspecific was placed under one of the cages for 5 min (sociability phase).
Social preference index (SI) was calculated as follows: where t mouse is the time spent sniffing the stimulus mouse; t cage is the time spent sniffing the empty cage.Any other type of behavior was labeled as "other." Social discrimination.Twenty-four hours after sociability test, SDT was conducted.The experimental setting was identical to that of sociability (Fig. 2A).Under each wired cage, a conspecific was placed for 5 min.One was used during  A, Schematic figure of sociability test, which started with a 5 min of OF test for mobility and anxiety-like behavior.Right after that, 5 min of object habituation followed when two identical wired cages were put into the arena.Right after that, one unfamiliar, juvenile conspecific was put under one of the wired cages for sociability phase.One day later, a 5 min SDT was conducted, during which the already familiar and a new unfamiliar conspecific were placed under the wired cages with interchanged positions compared with the previous day.B, Distance moved during OF.In dark, there was no difference between WT and VGluT3 KO mice (t test, N WT = 9; N KO = 11).On the other hand, inhibition of VGluT3+ neurons in the MRR resulted in decreased mobility (one-way ANOVA; N control = 8; N excitatory = 13; N inhibitory = 14).C, Percentage of time in different zones during OF.Increased anxiety-like behavior (repeated-measures ANOVA) in VGluT3 KO mice is reflected by decreased time spent in the centrum (75% of the arena floor).D, Schematic representation of EPM apparatus.EPM consists of two closed arms and two open arms connected by a centrum area, elevated 67 cm above the ground.E, Open/total (%) is a locomotion independent parameter that reflects anxiety-like behavior on sociability ("old"), while the other one was unknown ("new").Here CNO was not injected; thus, the effect on memory consolidation and not on recall was tested.
Social discrimination index (SD) was calculated as follows: SD = t "old" − t "new" t "old" + t "new" × 100, where t "old" is the time spent sniffing the familiar stimulus mouse; t "new" is the time spent sniffing the unfamiliar stimulus mouse.
Social interaction.The test was conducted in a transparent Plexiglass chamber (35 cm × 20 cm × 25 cm, Geo Maxi mini aquarium) with bedding (Fig. 3C).On previous day mice were put into the respective chamber for 2 × 15 min, 4 h apart, for habituation.During the experiment, two animals from the same group (e.g., KO vs KO), roughly weighting the same, were put into the chamber.The mice could freely behave for 10 min under normal lighting.
Social (e.g., sniffing), aggressive (e.g., biting, aggressive dominance), and defensive behaviors were analyzed for both test animals.Any other type of behavior was labeled as "other." Resident intruder.An unfamiliar, physically smaller, but adult conspecific was put into the homecage of the test animals.For 10 min, the animals could freely behave (Fig. 3D).Only the behavior of the test animal was analyzed (with similar parameters as in case of SIT).
EPM.The animals were placed onto the middle, central zone of the EPM apparatus (height, 25 cm; width, 7 cm; length, 30 cm; 67 cm above the floor) with two open and two closed arms aimed to measure anxiety (Fig. 2D).The mice could behave freely for 5 min in bright light (120 lux).
Open arm entries reflecting mobility independent anxiety were calculated: The frequency of risk assessment (RA) behavior (head dipping, stretched attend posture, and rearing) and grooming were also evaluated.
Y-maze.This test detects the short-term learning and memory of rodents, reflected by continuous spontaneous alternation (Hughes, 2004).The apparatus consists of three arms (A, B, and C; 25 cm × 5 cm × 21 cm) at 120°, connected by a central zone (Fig. 4C).No extra external spatial cues were used during the experiment.The consecutive arm entries reflect intact short-term memory.Mice were placed at the end of arm A and were allowed to explore the maze freely for 5 min.
The Ni-DAB-stained slices were evaluated with Olympus DP70 light microscope (4× magnification).The virus expression was examined between −4.04 and −4.96 mm from bregma.In case of no or unilateral staining, or infected other brain regions (e.g., dorsal raphe), the test animal and associated data were excluded from the statistical analysis (17/53; Fig. 1B).
c-Fos cell counting in brain sites relevant to social behavior and well-known projection areas of MRR (prelimbic cortex, infralimbic cortex and anterior cingulate, medial septum, CA1, CA2, and dentate gyrus) as well as in MRR was also done by Ni-DAB staining.Slices were incubated in the anti-c-Fos primer solution (1:1,000, rabit, #sc-52, Santa Cruz) for two nights on 4°C.The rest of the protocol was the same as described above.The stained slices were scanned by Pannoramic MIDI II Slide scanner (3DHISTECH) and cell bodies (three slides/animal) were counted by ImageJ.

Human brain samples
The study was approved by the Hungarian Medical Research Council -Scientific and Research Ethical Committee (#40197-2/2019/EKU), in accordance with the Ethical Rules for Using Human Tissues for Medical Research in Hungary (HM 34/1999) and the Code of Ethics of the World Medical Association (Declaration of Helsinki).Postmortem brain samples were obtained from the Human Brain Tissue Bank, Semmelweis University.The brains of 11 individuals were used in  to Figure 4. KO, knock-out; SDT, social discrimination test; VGluT3, vesicular glutamate transporter type 3; WT, wild-type.## p < 0.01 versus "old" mouse; $$ p < 0.01 versus random 50; == p < 0.01 versus control; @@ p < 0.01 versus excitatory.
the study.Average ± SEM age was 56.4 ± 54.33 years, while the postmortem delay was 5.72 ± 0.71-h (for details see Extended Data Table to Fig. 5-3).
Total RNA was isolated by RNeasy Mini Kit (Biomarker) according to the manufacturer's instructions.RNA was diluted into RNase-free water.The quality and quantity of extracted RNA was determined using NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific), and only those with A260/A280 ratio between 1.8 and 2.1 were used in subsequent experiments.The isolated RNA concentration was calculated and normalized with RNase-free water and reverse transcribed into cDNA using SuperScript II reverse transcriptase kit (Invitrogen).After 10-fold dilution, 2.5 μl of the resulting cDNA was used as template in PCR reactions using SYBR Green dye (Sigma).The PCR reactions were performed on CFX-96 C1000 Touch Real-Time System (Bio-Rad Laboratories) with iTaq DNA polymerase (Bio-Rad Laboratories) in total volumes of 12.5 μl under the following conditions: 95°C for 3 min, followed by 35 cycles of 95°C for 0.5 min, 60°C for 0.5 min, and 72°C for 1 min.A melting curve was performed at the end of amplification cycles to verify the specificity of the PCR products.All the measurements were conducted in duplicates.The primers used for RT-qPCR were synthesized by Integrated DNA Technologies and used at 300 nM final concentration.Sequences of intron spanning primers were designed based on the common parts of the two splice variants (NM_001145288.2 and NM_139319.3,122nt-3834nt) to eliminate detection of genomic DNA.The forward primer was GTCTGTCCCCTCATTGTCGG, while the reversed primer was CACAATTCTGGGAGGTGGCT, with an expected product size of 280 bp.Final cDNA products were sequenced by BIOMI and showed matching to the originally used primers (available upon request).One-third of the amplified PCR product was loaded in agarose gel (1.2%) containing Eco Safe nucleic acid staining solution (1:20,000, Pacific Image Electronics) and the electrophoresis was conducted in 1× TAE (40 mM Tris-acetate, 1 mM EDTA, pH 8.0) buffer.The applied separation voltage was 100 V.After the electrophoresis, the DNA bands were visualized by UV transillumination.

Statistical analysis
StatSoft 13.4 software was used.Outliers were defined as data points outside the interval of group mean ± 2× standard deviation for the given parameter and thus, were excluded from the analysis.This outlier analysis revealed uncharacteristic behavior resulted in the exclusion of the following number of animals: one WT mouse in the OF test; three WT mice in the Y-maze test; and one WT and one KO mice in sociability and SDTs.Similarly, in the case of VGluT3-Cre mice, the following number of animals were excluded: one mouse in the inhibitory group in the Y-maze test; two mice in the excitatory group during sociability test-object habituation phase, out of which one had to be also excluded from the rest of the experiment (sociability phase, SDT) as well.In the SIT and the EPM test, 1-1 mouse from the excitatory group was excluded.Finally, two mice (one control and one inhibitory) had to be removed from the RIT analysis.
To compare WT and KO animals, we used a t test (effect of genotype), while for the behavior of VGluT3-Cre mice, oneway analysis of variance (ANOVA; effect of manipulation) was used.The c-Fos-positive cell counting was done for every brain region separately, as staining was conducted also separately.Therefore, the groups were also compared with t test (WT-KO, effect of genotype; VGluT3-Cre control-excitatory, effect of manipulation).However, post hoc correction for multiple comparisons was also done by Benjamini-Hochberg correction (FDR, false discovery rate).SI and spontaneous alteration were tested with single-sample t test against 50%, while SD values were tested against 0%.The comparison of sides (left vs right), social preferences in sociability (mouse vs cage), and social discrimination (old vs new) were conducted by mixed-design ANOVA (effect of genotype or manipulation, effect of choice as repeated measure).For aggressive behavior of WT-KO animals in SIT, contingency table was made and Pearson's chi-square test was conducted.Fisher LSD was used for post hoc analysis.Data were expressed as mean ± SEM and p < 0.05 was considered statistically significant, while 0.1 < p < 0.05 was accepted as a marginal difference.

Results VGluT3 KO studies Mobility
There was no difference between the WT and VGluT3 KO animals in the distance moved in OF (Fig. 2B; Extended Data Table to Fig. 2-1), frequency of closed arm entries on EPM (Extended Data Table to Fig. 2-2), or arm entries in Y-maze (Extended Data Table to Fig. 4-2).

Anxiety-like behavior
In the OF, the frequencies of entering different zones did not show any differences.However, the KO animals spent less time in the centrum than the WT (t (18) = 2.963; p = 0.008; t test; Fig. 2C; Extended Data Table to Fig. 2-1).This increased anxiety-like behavior was confirmed on the EPM: KO animals tended to enter the open arm less than the WT (open/ total%: t (19) = 2.019; p = 0.058; t test; Fig. 2E; Extended Data Table to Figure 2-2).Additionally, the frequency of rearing (risk assessment) was lower for the KO animals (t (19) = 2.645; p = 0.016; t test).Overall, the KO group exhibited less risk assessment behavior than WT (RA WT = 67.500± 5.584; RA KO = 49.091± 6.511; t (19) = 2.125; p = 0.047).

Social interest
During object habituation, KO spent more cumulative time with the two objects than the WT (F (1,17) = 4.378; p = 0.052) without side preferences.This is also reflected by the reduced time spent with "other" behavior in KO mice (t (17) = 2.09; p = 0.052; Extended Data Table to Figure 3-1), during which they did not investigate the two empty cages.

Social behavior
During SIT frequency of defensive and "other" behavior showed only marginal genotype effect (t (19) = −1.808;p = 0.086 and t (19) = −1.936;p = 0.068, respectively; Extended Data Table to Fig. 3-3).However, there was an increase in the percentage of time (t (19) = −2.182;p = 0.042) and frequency of aggressive behavior (t (19) = −2.762;p = 0.012) in KO compared with WT.Since the values were low compared with the full length of the test (10 min), our continuous data were transformed into discrete variables: those animals who showed aggression were marked with "yes," while those who did not were marked with "no" and a contingency table was created.Pearson's chi-square test showed a significant difference (χ 2 = 7.636; p = 0.006): 6 KO animals out of 11 exhibited aggressive behavior, while the WT group showed none (Fig. 3B,C; Extended Data Table to Fig. 3-4).
In RIT, there was no difference between groups, and only a marginal effect of genotype was found in the frequency of defensive behavior (t (19) = 1.878; p = 0.076; (Fig. 3D,E; Extended Data Table to Fig. 3-5).These results show a contextdependent role of VGluT3 in aggressive behavior.

Social discrimination
There was a marginal effect of genotype in the percentage of time spent with stimulus animals (F (1,18) = 3.142; p = 0.093) and a significant genotype × choice interaction (F (1,18) = 4.281; p = 0.053).Post hoc analysis of the interaction showed that KO mice spent more time with the already familiar, "old" mouse compared with the WT group (p = 0.015) and also compared with the unknown, "new" mouse (p = 0.031; Fig. 4A; Extended Data Table to Fig. 4-1).This was also reflected by SD: WT animals did not differ significantly from random chance 0, unlike the KO group (t (10) = −2.352;p = 0.040).Thus, the SD difference between the groups was significant (t (17) = 2.503; p = 0.023; Extended Data Table to Fig. 4-1).Nevertheless, their average SD was −19.329 ± 8.217, which reflects a preference for the "old" mouse.

Neuronal activity
Ninety minutes after, SIT neuronal activation was measured by c-Fos immunohistochemistry.There was a significant decrease in the number of c-Fos-positive cells in the medial septum (t (11) = 4.199; p = 0.001) and infralimbic cortex (t (11) = 3.402; p = 0.006) of VGluT3 KO mice compared with WT.Other brain areas did not show significant difference after FDR correction (Fig. 5A-D; Extended Data Table to Fig. 5-1).

Mobility
MRR-VGluT3 manipulation only tended to affect the distance traveled in OF (F (2,32) = 3.04; p = 0.062; Fig. 2B; Extended Data Table to Fig. 2-3).Post hoc analysis showed that the inhibitory group moved less compared with control (p = 0.082) and excitatory groups (p = 0.029).Moreover, on EPM (Extended Data Table to Fig. 2-4) and in Y-maze (Extended Data Table to Fig. 4-4), the locomotion did not differ between the groups.

Anxiety-like behavior
In the OF, neither the frequencies nor the percentage of time spent in the centrum differed between the groups (Fig. 2C; Extended Data Table to Fig. 2-3).On the other hand, on the EPM MRR-VGluT3 manipulation significantly influenced the open/total% values (F (2,32) = 3.828; p = 0.032).According to the post hoc analysis, the excitatory group entered the open arms more often than control group (p = 0.009; Fig. 2E; Extended Data Table to Fig. 2-4).The frequency of rearing showed a tendency (F (2,31) = 2.769; p = 0.078), while head dipping was significantly influenced by MRR-VGluT3 manipulation (F (2,31) = 3.572; p = 0.040).Post hoc test of head dipping revealed an increase in the excitatory group compared with both control and inhibitory groups (p < 0.05).Based on this, the excitation of MRR-VGluT3 neurons was proved to be anxiolytic.

Social discrimination
None of the groups could differentiate between the stimulus animals (Extended Data Table to Fig. 4-3).However, there was a significant effect of MRR-VGluT3 manipulation in both the frequencies of interaction (F (2,32) = 3.336; p = 0.048) and percentage of time spend with it (F (2,32) = 7.070; p = 0.003).Post hoc test showed that inhibitory group had greater interaction frequencies compared with control (p = 0.020), which differed marginally from the excitatory groups (p = 0.092; Extended Data Table to Fig. 4-3).In the meantime, post hoc test revealed that the inhibitory group spent more time with the two conspecifics together compared with both control (p = 0.003) and excitatory groups (p = 0.005; Fig. 4B; Extended Data Table to Fig. 4-3).Accordingly, the frequency (F (2.32) = 2.586; p = 0.091) and percentage of time (F (2,32) = 6.526; p = 0.004) spent with "other," nonsocial behavior showed marginal effect of MRR-VGluT3 manipulation.Post hoc analysis confirmed that the inhibitory group initiated less often (frequency) and spent less time with nonsocial behavior compared with control (p = 0.004) and excitatory (p = 0.007) groups.Interestingly, this increase in social interest (reflected by the higher time spent with both stimulus animals) was observable 24 h after MRR-VGluT3 inhibition was induced during sociability test.

Neuronal activity
Only the MRR region showed statistically significant neuronal activation.As expected, after CNO injection, we could detect increased number of c-Fos-positive cells in the stimulatory DREADD group compared with control virus injected one (t (7) = −14.987;p < 0.001; Fig. 5E,F), confirming that the excitatory DREADD receptors were functional and induced neuronal activation.However, its projection areas did not show increased neuronal activity (Extended Data Table to Fig. 5-2).

Human brain PCR results
VGluT3 mRNA levels were investigated in four human brainstem areas: lateral parabrachial nucleus, pontine reticular formation, pontine raphe nucleus, and midbrain raphe nuclei (Fig. 5G).As negative control, temporal and frontopolar cortex samples were used (Fig. 5H,I).All examined brainstem areas-but not the cortical regions-expressed VGluT3 mRNA , indicating possible clinical relevance of our study.

Discussion
We examined the role of VGluT3+ neurons in social behavior, and-as possible confounding factors-in mobility, anxiety-like behavior and short-term memory in mice.For the short summary of the main findings, see Table 1.
During previous studies, the complete lack of VGluT3 resulted in deafness (Seal et al., 2008), highly anxious behavior phenotype (Amilhon et al., 2010;Balazsfi et al., 2016;Sakae et al., 2019), altered fear, and stress reactivity (Horvath et al., 2018;Balazsfi et al., 2018a).KO mice proved to be hyperactive, which is circadian dependent and only appears during their active, dark period (Gras et al., 2008;Divito et al., 2015).However, during short observations in a new, aversive (lighted) environment, they might move even less than the controls (Horvath et al., 2018;Balazsfi et al., 2018a;Fazekas et al., 2019;Sakae et al., 2019;de Almeida et al., 2023).Compared with our previous studies (Balazsfi et al., 2018a;Fazekas et al., 2019), in this current project the experiments were done during active period.This could possibly explain why we found no hypoactivity, and the shortness of observation (5 min compared with, for example, 5 h; Gras et al., 2008) could stand behind the undetectability of hyperactivity.Nevertheless, this experimental set-up further confirms the circadian-dependent role of VGluT3 in locomotion.
It is already well accepted in the literature that the constitutive lack of VGluT3 throughout the lifetime of mice results in a highly anxious behavioral phenotype (Amilhon et al., 2010;Balazsfi et al., 2016;Sakae et al., 2019).Accordingly, we showed on EPM that VGluT3 KO mice also exhibited less risk assessment behavior.Moreover, even though OF was conducted in dark, less anxiogenic context, KO animals still preferred the periphery.
In the sociability test, KO animals proved to be more social, which we failed to observe in our previous study (Fazekas et al., 2019).This could be explained by the previously mentioned circadian difference, as well as by slight difference in the protocol.Interestingly, the KO animals showed aggressive behavior only during SIT, which might be explained by the protocol.Although the habituation to the context of SIT is aimed to reduce aggression, the test was conducted in an anxiogenic (lighted) environment, which could have facilitated aggression in the sensitive KO mice, which is known to exhibit high anxiety-like behavior (Amilhon et al., 2010).In accordance with the literature (Amilhon et al., 2010), in RIT there was no significant change in their behavior, further supporting the anxiety-induced nature of aggression during SIT.
Another group linked decreased VGluT3 protein expression to impaired memory (Cheng et al., 2011).We also found KO mice to be worse in certain aspects of learning, such as working memory and cognitive flexibility (Fazekas et al., 2019).A new study on them confirmed their proper learning and recognition memory ability but impaired pattern separation after aversive stimuli (de Almeida et al., 2023).In the SDT, the VGluT3 KO animals had impaired long-term (24 h) social memory.Similarly, in our previous study VGluT3 KO mice showed inadequate long-term (24 h) social memory in the three-chamber sociability test.However, in the Y-maze, which measures working memory, we failed to observe the previously detected impairment (Fazekas et al., 2019).Since the lack of difference was observed during the active phase without hypolocomotion in the present study, but a previous one showed marked hypolocomotion during their inactive phase (de Almeida et al., 2023), this indicates a major sensitivity of these tests to environmental factors, as well as suggest that the KO-induced changes are subtle and can be easily compensated.
Interestingly, social behavior elicited lower neuronal activation in KO than in WT on relevant brain areas (infralimbic cortex and the medial septum).These are projection target of MRR (Leranth and Vertes, 1999;Vertes et al., 1999;McKenna  and Vertes, 2001;Szonyi et al., 2016) indicating that MRR might be a key hub in VGluT3-influenced social behavior.Therefore, we conducted an MRR-focused analysis using VGluT3-Cre mice.The activity of MRR has been linked to locomotion in numerous occasions.Stimulation of the MRR resulted in a decreased locomotion and a long-term fear memory formation (Bland et al., 2016;Balazsfi et al., 2017).By using different pharmacons (Shim et al., 2014), the effect of MRR on locomotion was partly dopamine receptor (D2) dependent and partly independent.According to our results, the later pathway potentially includes VGluT3+ neuronal (in)activity.However, stimulation was ineffective and the locomotor effect was not detected in all tests.Interestingly, when all VGluT3-positive neurons of the body were excited with chemogenetics, hypolocomotion was detected in the light phase of mice (Kljakic et al., 2022).Based on these and our results, the role of VGluT3-positive neurons in locomotion is complex and MRR is just one brain area which participates in this role.
Anxiety-like behavior is strongly tied to serotonin, which is also present in the MRR.By using mRNA interference specific to the serotonin transporter in the MRR, anxiety-like behavior was increased on EPM (Verheij et al., 2018).In contrast, in our study excitation of VGluT3+ cell of MRR decreased anxiety.This is in line with the increased anxiety of KO mice and can be explained by the vesicle loading role of VGluT3 (Amilhon et al., 2010).Thus, we might assume that excitation of MRR-VGluT3 neurons increased serotonin loading, thereby reducing anxiety.However, this effect is not robust as the phenomenon was not observable in OF and inhibition was ineffective.
As of its role in social behavior, optogenetic stimulation of the whole MRR decreased aggression in the SIT paradigm, which was accompanied by not only increased levels of serotonin but also glutamate in the mPFC (Balazsfi et al., 2018b).In our present experiment, the decreased time spent with the conspecific and the increased time spent with other behavior during the sociability test in the excitatory group indicate a decreased social interest.However, inhibition increased social interest only 1 d after manipulation, during the SDT.Additionally, inhibition affected friendly social behavior in a contextspecific manner: in the anxiogenic SIT tended to increase it, while in RIT decreased it.However, despite our expectations we were unable to find any other major role of MRR-VGluT3+ cells in social behavior.
MRR has been implicated in memory numerous times.It has an effect on both theta oscillation (Viana Di Prisco et al., 2002;Bland et al., 2016) and ripple activity of the hippocampus (Wang et al., 2015), known elementary phenomena of memory formation.Moreover, it is known that MRR sends nonserotoninergic projections to the medial septum and hippocampus (Aznar et al., 2004;Fortin-Houde et al., 2023), key brain areas of learning and memory (Tsanov, 2018;Jin et al., 2020;Wu et al., 2021).As previously mentioned, excitation of the MRR elicited a fear memory without actual aversive stimulus, while inhibition of the nucleus prevented the memory formation (Balazsfi et al., 2017).Based on these results, it is expected that MRR-VGluT3+ neurons play a role in some sort of learning and/or memory formation processes.Despite this, in the two memory-related experiments (SDT and Y-maze), we failed to identify any effect.Whether they regulate more complex processes still need to be investigated.
In the MRR, the increase in c-Fos-positive neurons was significant after injection, proving two major points: (1) despite the control animals being injected as well, there is no neuronal activation by CNO and (2) the expressed excitatory DREADD receptors were functional.
Interestingly, the excitation and inhibition of the MRR-VGluT3+ neurons did not yield complementary results.However, we must take into consideration that the used DREADDs activated-inhibited different pathways (G q vs G i ), which could have caused the different outcomes.Moreover, it has been shown that the MRR-VGluT3+ cells are not homogenous.Unfortunately, with chemogenetics we are unable to differentiate between these cells and it is possible that we have seen the interaction of the activity of these cells.
The expression pattern of VGluT3 in the brain is mostly known from rodent studies.In the brainstem, abundant VGluT3 mRNA and protein levels were found in the dorsal and median raphe nuclei, but only punctate pattern in the pontine nuclei (Schafer et al., 2002;Nakamura et al., 2005;Stornetta et al., 2005).Guo et al. (2005) showed VGluT3-positive cells in the cat parabrachial nucleus.Vigneault et al. (2015) conducted in situ hybridization and immunoautoradiography to detect mRNA and protein levels, respectively, of all three types of VGluTs in the human brain on selected brain regions.However, the only brainstem regions they examined were the collective raphe nuclei, highlighting the dorsal raphe and the pons area.In accordance with the literature, they found abundant VGluT3 mRNA and protein expression in the dorsal raphe; however, in the rest of the pons, it was detected only on protein level.By using RT-qPCR on human brain samples, we confirmed the presence of VGluT3 in human brainstem on the mRNA level.Midbrain raphe nuclei, pontine raphe nuclei, pontine reticular formation, and lateral parabrachial nucleus all showed detectable transcripts, while the control cortical samples did not.This is in accordance with the literature on rodents (Fremeau et al., 2002;Gras et al., 2002;Schafer et al., 2002;Herzog et al., 2004;Guo et al., 2005;Stornetta et al., 2005;Hioki et al., 2010).
In summary, our experiments showed that the total ablation of the VGluT3 gene shifted the behavior to be socially more active.However, the fact that the KO animal showed increased aggression only during the anxiogenic SIT test, but not during the RIT, suggests that the aggression of these animals was induced by anxiety.We further confirmed the highly anxious phenotype and an impairment in long-term social memory in the VGluT3 KO animals.In the case of VGluT3+ MRR neurons, we found that they play a role in locomotion and long-term social interest.Although in our case it was not possible to differentiate between the VGluT3+ only and VGluT3+/5-HT+ neurons of the MRR, it was demonstrated that either one or both subpopulations can contribute to the anxiolytic effect of MRR but has subtle effect on social behavior.

Figure 1 .
Figure 1.Timeline of the experimental protocols and immunohistochemical validation.A, Schematic timeline of the behavioral test battery.In case of VGluT3-Cre mice, experiments started 4 weeks after the stereotaxic surgery-which included 2 weeks of habituation to "reversed" day/night cycle-in order to ensure viral expression of DREADDs.Syringes indicate the occasions when VGluT3-Cre mice were injected intraperitoneally with CNO (1 mg/ml/kg) 30 min before the start of the experiment.Two to 3 d of break were kept between the experiments in order to eliminate the carry-over effect of CNO.WT and VGluT3 KO mice went under the same procedure, except for the surgery and CNO administration.B, Example of proper (left) and improper (right) viral expression in VGluT3-Cre mice.After the experiments, brains were collected from the animals, and Ni-DAB staining was done on midbrain slices against the RFP encoded in the AAV.Only the data of those animals that had proper viral expression in the MRR were included in the statistical analysis.On the right, there is an example of unacceptable staining in the DR.C, Representative figure of viral expression specificity in the MRR.Double fluorescent immunohistochemistry was done of midbrain slices against VGluT3 (Alexa-488, green) and RFP (Alexa-594, red) in control virus injected VGluT3-Cre mice.White arrows indicate example neurons where both markers were found.Scale bar, 100 μm.AAV, adeno-associated virus vector; Aq, Aqueduct cerebri; CNO, clozapine-N-oxide; DR, dorsal raphe; DREADD, designer receptor exclusively activated by designer drugs; KO, knock-out; MRR, median raphe region; Ni-DAB, nickel-3.3′ -diaminobenzidine; RFP, red fluorescent protein; VGluT3, vesicular glutamate transporter 3; WT, wild-type.

Figure 2 .
Figure 2. Results of experiments on mobility and anxiety-like behavior.A, Schematic figure of sociability test, which started with a 5 min of OF test for mobility and anxiety-like behavior.Right after that, 5 min of object habituation followed when two identical wired cages were put into the arena.Right after that, one unfamiliar, juvenile conspecific was put under one of the wired cages for sociability phase.One day later, a 5 min SDT was conducted, during which the already familiar and a new unfamiliar conspecific were placed under the wired cages with interchanged positions compared with the previous day.B, Distance moved during OF.In dark, there was no difference between WT and VGluT3 KO mice (t test, N WT = 9; N KO = 11).On the other hand, inhibition of VGluT3+ neurons in the MRR resulted in decreased mobility (one-way ANOVA; N control = 8; N excitatory = 13; N inhibitory = 14).C, Percentage of time in different zones during OF.Increased anxiety-like behavior (repeated-measures ANOVA) in VGluT3 KO mice is reflected by decreased time spent in the centrum (75% of the arena floor).D, Schematic representation of EPM apparatus.EPM consists of two closed arms and two open arms connected by a centrum area, elevated 67 cm above the ground.E, Open/total (%) is a locomotion independent parameter that reflects anxiety-like behavior on

Figure 3 .
Figure 3. Results of experiments on social behavior.A, Percentage of time spent with empty cages and the cage containing an unfamiliar, juvenile conspecific.Both WT and VGluT3 KO mice preferred the social stimuli compared with the inanimate object, but KO animals spent more time with the conspecific (repeated-measures ANOVA; N WT = 10; N KO = 11).Similarly, all animals in the VGluT3-Cre groups preferred the social stimuli, but the excitatory group had decreased interest compared with the inhibitory group (repeated-measures ANOVA; N control = 8; N excitatory = 11; N inhibitory = 15).B, Main differences between groups in SIT.VGluT3 KO mice exhibited more aggressive behavior compared with the WT animals (t test; N WT = 10; N KO = 11).On the other hand, friendly social behavior was only marginally changed in the VGluT3-Cre mice (one-way ANOVA; N control = 8; N excitatory = 13; N inhibitory = 15).C, Schematic representation of SIT protocol: for 10 min two experimental animals from the same treatment group (e.g., KO vs KO, excitatory vs excitatory) was placed into a neutral, anxiogenic environment to freely interact.Under the figure Gantt diagrams visually represent the behavior of all experimental animals during SIT.Different behaviors are represented by different colors.D, Schematic representation of RIT protocol: for 10 min an unfamiliar, physically smaller, but adult conspecific was placed into the home cage of the experimental mouse to freely interact.Under the figure Gantt diagrams visually represent the behavior of all experimental animals during RIT.Different behaviors are represented by different colors.E, Frequency of aggressive behavior did not differ between WT and VGluT3 KO mice during RIT (t test; N WT = 10; N KO = 11).F, Frequency of friendly social behavior was decreased in the inhibitory group of VGluT3-Cre mice in RIT (one-way ANOVA; N control = 8; N excitatory = 13; N inhibitory = 14).Data are expressed as average ± SEM.Empty squares represent individual values.For more details see Extended Data Tablesto Figure 3. KO, knock-out; RIT, resident intruder test; SIT, social interaction test; VGluT3, vesicular glutamate transporter type 3; WT, wild-type.## p < 0.01 versus mouse; **p < 0.01 versus WT; == p < 0.01 versus control; @ p < 0.05 versus excitatory; -p < 0.01 versus inhibitory.

Figure 4 .
Figure 4. Results of social and working memory experiments.A, SDT conducted 1 d after sociability test (Fig. 2A) showed a preference for the familiar, "old" mouse in VGluT3 KO mice (repeated-measures ANOVA; N WT = 10; N KO = 11).B, In SDT for VGluT3-Cre mice, there were no differences between the groups regarding their social memory (repeated-measures ANOVA), but the cumulative time spent with social stimuli showed an increased social interest in the inhibitory group (one-way ANOVA; N control = 8; N excitatory = 12; N inhibitory = 15).C, Schematic figure of Y-maze test.Three identical arms placed perpendicular can be freely explored by the experimental animal for 5 min.Proper working memory is reflected by consecutive, different arm entries, called spontaneous alternation.D, All groups had intact working memory compared with random 50% except for the VGluT3-Cre control group (single sample t test; N WT = 10; N KO = 11; N control = 8; N excitatory = 13; N inhibitory = 14).Data are expressed as average ± SEM.Empty squares represent individual values.For more details see Extended Data Tablesto Figure 4. KO, knock-out; SDT, social discrimination test; VGluT3, vesicular glutamate transporter type 3; WT, wild-type.## p < 0.01 versus "old" mouse; $$ p < 0.01 versus random 50; == p < 0.01 versus control; @@ p < 0.01 versus excitatory.

Figure 5 .
Figure 5. c-Fos staining in the mouse brain and VGluT3 mRNA expression in healthy human samples.A-D, There are decreased numbers of c-Fos-positive cells in the KO (B,D) compared with WT (A,C) mice in the mPFC (A,B) and MS (C,D) after SIT (N WT = 5; N KO = 8).E,F, Functional DREADD excitation is reflected by increased c-Fos activity in the MRR of VGluT3-Cre mice (N control = 6; N excitatory = 3).Scale bar, 100 μm.G, Relative expression of VGluT3 mRNA to the measured housekeeping gene (GAPDH) in the investigated brainstem areas (LPB: N = 3; PRF: N = 10; PRN: N = 6; RN: N = 3).H, After RT-qPCR, samples containing VGluT3 cDNA were run with gel electrophoresis and sequencing (BIOMI) to validate the specificity of primers.Brain samples from pontine raphe nuclei, which contain the human equivalent of mouse MRR, all expressed VGluT3 (N = 6).I, Different cortical areas of healthy human samples were used as negative controls (N = 6).Data are expressed as average ± SEM.Empty squares represent individual values.For more details see Extended Data Tablesto Figure 5. bp, base pair; DREADD, designer receptors exclusively activated by designer drugs; KO, knock-out; LPB, lateral parabrachial nucleus; mPFC, medial prefrontal cortex; MRR, median raphe region; MS, medial septum; PRF, pontine reticular formation; PRN, pontine raphe nucleus; RN, midbrain raphe nuclei; RT-qPCR, reverse transcription quantitative polymerase chain reaction; SIT, social interaction test; VGluT3, vesicular glutamate transporter type 3; WT, wild-type.

Table 1 .
Summary of the main findings The results are presented in comparison with their respective control groups: that is, VGluT3 wild-type littermates for knock-out mice and control group (expressing only RFP in the VGluT3+ neurons in the MRR) for the excitatory (Gq DREADD + RFP) and inhibitory (Gi DREADD + RFP) chemogenetically manipulated groups.Ø, no difference compared with respective control group; -, not applicable; DREADD, designer receptors exclusively activated by designer drugs; EPM, elevated plus maze; MRR, median raphe region; OF, open field; RFP, red fluorescent protein; RIT, resident intruder test; SDT, social discrimination test; SIT, social interaction test; VGluT3, vesicular glutamate transporter 3.