Neurexins in serotonergic neurons regulate serotonin transmission and complex mouse behaviors

Extensive serotonin (5-HT) innervation throughout the brain corroborates 5-HT’s modulatory role in numerous cognitive activities. Volume transmission is the major mode for 5-HT transmission but mechanisms underlying 5-HT signaling are still largely unknown. Abnormal brain 5-HT levels and function have been implicated in autism spectrum disorder (ASD). Neurexin (Nrxn) genes encode presynaptic cell adhesion molecules important for the regulation of synaptic neurotransmitter release, notably glutamatergic and GABAergic transmission. Mutations in Nrxn genes are associated with neurodevelopmental disorders including ASD. However, the role of Nrxn genes in the 5-HT system is poorly understood. Here, we generated a mouse model with all three Nrxn genes disrupted specifically in 5-HT neurons to study how Nrxns affect 5-HT transmission. Loss of Nrxns in 5-HT neurons impaired 5-HT release in the dorsal raphe nucleus and dorsal hippocampus and decreased serotonin transporter distribution in specific brain areas. Furthermore, 5-HT neuron-specific Nrxn knockout reduced sociability and increased depressive-like behavior. Our results highlight functional roles for Nrxns in 5-HT neurotransmission and the execution of complex behaviors.


Introduction
Serotonin (5-hydroxytryptamine, 5-HT) neurons in the raphe nuclei project their axons throughout the brain and modulate social interactions, stress responses, and valence among other processes.
In this study, we investigated the functions of Nrxns in the 5-HT system by assessing signaling properties and behavior in 5-HT neuron-specific Nrxn triple knockout (TKO) mice.We demonstrated that the loss of Nrxn genes reduced 5-HT release and serotonin transporter (SERT)-labeled 5-HT fibers in the mouse brain.Moreover, the lack of Nrxns in 5-HT neurons altered social behavior and depressivelike phenotypes.Our findings highlight Nrxns as functional regulators of neurotransmission and complex behaviors in the 5-HT system.

Reduced 5-HT release in Fev/RFP/NrxnTKO mice
While numerous studies indicate that Nrxns regulate fast neurotransmitter release including that of glutamate and GABA, no studies have tested Nrxn function in central neuromodulatory systems.To directly examine the role of Nrxn on 5-HT release, we measured 5-HT transients in the DRN and hippocampus using fast-scan cyclic voltammetry (FSCV) (Figure 2).5-HT release was recorded with a voltage ramp delivered through carbon-fiber electrodes (Figure 2A).Current-voltage (CV) plots displayed the expected currents for 5-HT oxidation and reduction peak potentials, indicating specificity for 5-HT (Figure 2B).5-HT transients were recorded in the DRN where 5-HT neurons are highly clustered in WT and Fev/RFP/NrxnTKO mice (Figure 2C-F).Electrical stimulation was applied at two different stimulus strengths to evoke 5-HT release in acute brain slices containing the DRN.To confirm that electrically evoked FSCV transients were mediated by 5-HT release, the selective serotonin reuptake inhibitor fluoxetine (FLX) and action potential inhibiting sodium channel blocker tetrodotoxin (TTX) were applied (Figure 2C, D, F) (Carboni & Di Chiara, 1989).Importantly, 5-HT peak amplitude was significantly reduced in Fev/RFP/NrxnTKO mice (Figure 2E).FLX caused a similar increase in 5-HT transient area in each genotype, indicating that Nrxn TKO did not change transporter activity (Figure 2F).Next, we performed FSCV recordings in the dorsal hippocampal CA3 region to determine whether differences in 5-HT release could be detected in a distal region receiving 5-HT fiber projections (Figure 2G-J).We observed robust suppression of 5-HT currents in Fev/RFP/NrxnTKO mice and no genotype-specific differences in response to FLX.Taken together, these findings indicate that Nrxns are important for 5-HT release.
We found that SERT fibers were reduced in the dorsal hippocampus and the DRN of Fev/RFP/NrxnTKO mice relative to controls.Interestingly, no differences were seen in the projections to the nucleus accumbens and vCA1 indicating that SERT inputs are not globally altered.These findings suggest that Nrxns selectively mediate SERT-positive fiber area depending on the innervated circuit.Although Nrxn TKO in cerebellar granule cells were found to cause cell death (T.Uemura et al., 2021), there were no changes in 5-HT neuron number in Fev/RFP/NrxnTKO mice indicating that reduced SERT fiber density is not due to the loss of 5-HT neurons (Figure 3 Fev/RFP/NrxnTKO underwent a direct social interaction test to examine naturally occurring interactions between a subject mouse and a juvenile stimulus mouse.In trial 1, the stimulus mouse was unfamiliar to the subject mouse.After 24 hours, the subject mouse was re-exposed to the same stimulus mouse (trial 2) (Figure 4A).Social investigation was measured across both trials and as a reduction in time that the subject mouse spent investigating the stimulus mouse in trial 2. We found that Fev/RFP/NrxnTKO mice spent less time exploring the stimulus mouse in trial 1 (Figure 4B) and differed in their investigation of the stimulus mouse across trials (Figure 4C) compared with WT mice.These results suggest that Fev/RFP/NrxnTKO mice have deficits in sociability.Interestingly, one of the depression tests, the forced swim test but not tail suspension test, revealed increased immobility behavior in Fev/RFP/NrxnTKO compared with WT mice (Figure 4D-E).Importantly, other tests addressing learning and memory and repetitive behaviors displayed no abnormalities in Fev/RFP/NrxnTKO mice (Figure 4-Figure Supplement 2).These results demonstrate that the absence of Nrxns in 5-HT neurons impairs social behavior and moderately influences depression-related behavior.

Discussion
Nrxns regulate the release of fast neurotransmitters such as glutamate and GABA by coupling Ca 2+ channels to presynaptic release machinery (Sudhof, 2017).However, their roles in central neuromodulatory systems have never been addressed.Here we provide evidence that Nrxns control neuromodulatory 5-HT release.We found that the DRN and hippocampus displayed > 40% reduction in 5-HT release in Fev/RFP/NrxnTKO mice.Fev expression begins in the embryonic stage (Scott et al., 2005), therefore the impact of Nrxn TKO during development should be noted.However, compared with the robust functional deficit observed in Fev/RFP/NrxnTKO mice, the structural deficit identified by SERTpositive fiber density was moderate (< 25%) suggesting that the primary role of Nrxns in the 5-HT system is the formation of functional components important for 5-HT release.
It is important to consider the mechanisms through which Nrxns influence release events and the specific sites that express Nrxns to control 5-HT neurotransmission.In the hippocampus, approximately 80% of 5-HT varicosities are extra-synaptic, while the remaining 20% form synapses (Oleskevich, Descarries, Watkins, Seguela, & Daszuta, 1991).Given the predominance of non-junctional specializations, we speculate that Nrxns reside at 5-HT release sites that lack a direct postsynaptic target.
The ability of Nrxns to couple with release machinery triggering 5-HT vesicle exocytosis and their roles in postsynaptic differentiation at synapses are yet to be explored.Decreased SERT innervation suggests that 5-HTergic Nrxns contribute to fiber formation, regulate the abundance of SERT itself, or that inefficient 5-HT release requires less SERT as a compensatory mechanism.Indeed, sparse pan-Nrxn deletion has been shown to blunt inferior olive neuron climbing fiber projections in the cerebellum while complete removal of Nrxns at climbing fiber synapses did not alter climbing fiber axons but impaired synaptic transmission (Chen, Jiang, Zhang, Gokce, & Sudhof, 2017).
The observed deficits in sociability and depressive-related behaviors are relevant to ASD which often presents with co-occurring conditions.The forced swim test was performed following the tail suspension test and it is possible that Fev/RFP/NrxnTKO are more susceptible to stress rather than despair-associated coping responses.Overall, the behavioral phenotypes of Fev/RFP/NrxnTKO mice are milder than that of null Nrxn KO mouse lines.Both αNrxn1 KO and αNrxn2 KO mice display social behavior deficits, elevated anxiety, and increased stereotypic behaviors (Born et al., 2015;Dachtler et al., 2014;Etherton, Blaiss, Powell, & Sudhof, 2009;Grayton, Missler, Collier, & Fernandes, 2013).In contrast, no abnormalities in repetitive behaviors were found in Fev/RFP/NrxnTKO mice suggesting that Nrxn TKO in 5-HT neurons has more selective effects on 5-HT mediated behaviors.
Our results reveal that Nrxns expressed in midbrain 5-HT neurons are important for maintaining the presynaptic molecular function of 5-HT release sites.Further investigations are necessary to decipher Nrxn-mediated 5-HT release machinery, examine the consequences of Nrxn deletion in raphe nuclei-innervated circuits in other brain regions, and address whether 5-HT therapeutics can improve behavioral deficits.
Mice were group housed (2-5 per cage) and maintained in ventilated cages with ad libitum access to food and water on a standard 12-hour light/12-hour dark cycle (lights ON at 7 A.M.) in a temperaturecontrolled (20-23°C) facility.One to two weeks prior to experimentation, mice were acclimated to a reversed light/dark cycle (lights ON at 7 P.M.).

Single-cell RNA extraction and RT-qPCR
The whole procedure was done based on our recently developed protocol (Mao et al., 2018;Uchigashima, Konno, et al., 2020;Uchigashima, Leung, et al., 2020).Briefly, cytosol from RFP+ 5-HT neurons in the dorsal and median raphe nuclei were harvested from Fev/RFP and Fev/RFP/NrxnTKO mice using the whole-cell patch-clamp approach.A SMART-Seq® HT Kit (TAKARA Bio) was used to prepare the cDNA libraries following the manufacturer's instructions.Single-cell cDNA libraries were prepared by reverse transcription at 42°C for 90 min followed by PCR amplification (Uchigashima, Konno, et al., 2020;Uchigashima, Leung, et al., 2020).To assess Nrxn expression in individual 5-HT neurons of control and Fev/RFP/NrxnTKO mice, the following TaqMan gene expression assays (Applied Biosystems) were used: (Mm00557715_m1) and Gapdh (Mm99999915_g1).The PCR reactions and analyses were performed blind.The relative expression of Nrxns or Tph2 were calculated as: Relative expression = 2 Ct,Gapdh /2 Ct,Nrxns or Tph2 ; Ct, threshold cycle for target gene amplification, and presented as fold changes relative to that of WT.

Transcriptome analysis
Single-cell RNA-seq data was obtained from a recent publication (Ren et al., 2019).See Ren et al., 2019 for the single-cell isolation and sequencing (Ren et al., 2019).
Data processing and Clustering: Datasets were downloaded from NCBI Gene Expression Omnibus (GSE135132).
Reads were aligned to mouse reference transcriptome (Mus_musculus.GRCm38.cdna.all.fa) using kallisto (Bray et al., 2016).Tximport R package (Soneson, Love, & Robinson, 2015) was used to summarize to the gene-level.Each isoform was summarized manually to account for inclusion of spliced exons in the α or β Nrxn isoforms.The manually curated transcript IDs are provided in Table 1.The gene count data was analyzed using Seurat R package v4.0.1 (Hao et al., 2021).After excluding the cells with low sequencing depth (50,000 reads) and low number of detected genes (cut-off was set at 7,500 genes), the remaining 945 cells were assigned to clusters according to the Ren et al. paper (Ren et al., 2019).Counts were normalized for each cell using the natural logarithm of 1 + counts per 10,000 [ln(1+counts/10k)].Cells were visualized using a 2-dimensional t-distributed Stochastic Neighbor Embedding (tSNE) and violin plots.The R code is provided as codeR (Source Code File 1).

SERT density analysis
We analyzed SERT innervation to the nucleus accumbens core and shell (NAcc, NAcSh; Bregma 1.18 ± 0.3 mm), stratum oriens of the CA1, CA2, and CA3 subregions of the dorsal hippocampus (dCA1, dCA2, dCA3; Bregma -1.46 ± 0.4 mm), stratum oriens of the CA1 subregion of the ventral hippocampus (vCA1; -3.16 ± 0.4 mm), and dorsal raphe nucleus (DRN; Bregma -4.56 ± 0.4 mm).To assess the density of SERT fiber inputs, four stained sections from each of three or four WT and Fev/RFP/NrxnTKO brains containing the nucleus accumbens, hippocampus, or raphe nuclei were imaged (1024 x 1024 pixels) using a laser scanning confocal microscope (LSM700, Zeiss) with a 63x oil-immersed objective (NA 1.4) at an optical zoom of 1.6 and Zen black image acquisition software (Zeiss).For each brain, six randomly chosen 100x fields of view within the region of interest were acquired with seven z-stack steps at 0.35 µm spacing to generate maximum intensity projections (MIPs) of the z-stacks.Images from all brains for a particular region were acquired using identical settings and data analyses were performed using ImageJ as previously described (Werneburg et al., 2020).The six images from each region per animal were averaged to generate a mean for that region in each animal, with n=3-4 animals per genotype.A consistent threshold range was determined by subjecting images, blinded to genotype, to background subtraction and manual thresholding for each MIP within one experiment (IsoData segmentation method, 15-225).Using the analyze particles function, the thresholded images were used to calculate the total area of SERT fiber inputs.

5-HT neuron density analysis
We analyzed the number of 5-HT-positive neurons in the DRN (Bregma -4.6 ± 0.3 mm) and MRN (Bregma -4.5 ± 0.4 mm).To assess the density of 5-HT neurons, four stained sections from each of four or five WT and Fev/RFP/NrxnTKO brains containing the DRN or MRN were imaged (1024 x 1024 pixels) using a laser scanning confocal microscope (LSM700, Zeiss) with a 20x water-immersed objective (NA 1.0) at an optical zoom of 0.5 and Zen black image acquisition software (Zeiss).Images from all brains for a particular region were acquired using identical settings and data analyses were performed using ImageJ as previously described (Uchigashima, Konno, et al., 2020).
Fast scan cyclic voltammetry 5-HT measurements were performed in the radiatum of dorsal CA3 and DRN.All experiments and analyses were performed blind to genotype.To detect 5-HT release, carbon fiber electrodes were prepared as previously described (Hashemi, Dankoski, Petrovic, Keithley, & Wightman, 2009;Matsui & Alvarez, 2018).Carbon fiber electrodes consisted of 7-µm diameter carbon fibers (Goodfellow) inserted into a glass pipette (A-M Systems, cat# 602500) with ~150-200 µm of exposed fiber.The exposed carbon fibers were soaked in isopropyl alcohol for 30 min to clean the surface.Next, the exposed fibers were coated with Nafion solution (Sigma) to improve detection sensitivity by inserting the carbon fiber into Nafion solution dropped in a 3 mm diameter circle of twisted reference Ag/AgCl wire for 30 sec with constant application of +1.0 V potential.The carbon fiber electrodes were air dried for 5 min and then placed in a 70°C oven for 10 min.A modified 5-HT voltage ramp was used, in which the carbon fiber electrode was held at +0.2 V and scanned to +1.0 V, down to -0.1 V, and back to +0.2 V at 1,000 V/s delivered every 100 ms.Prior to recording, the electrodes were conditioned in aCSF with a voltage ramp delivered at 60 Hz for 10 min.
5-HT release was evoked with electrical stimulation (30 pulses, 30 Hz, 150 or 250 µA, 1 ms) from an adjacent custom-made bipolar tungsten electrode every 10 min.The stimulating electrode was placed ~100-200 µm away from the carbon fiber electrode (John, Budygin, Mateo, & Jones, 2006).Recordings were performed using a Chem-Clamp amplifier (Dagan Corporation) and Digidata 1550B after low-pass filter at 3 kHz and digitization at 100 kHz.Data were acquired using pClamp10 (Molecular Devices) and analyzed with custom written VIGOR software using Igor Pro 8 (32-bit; Wavemetrics) running mafPC (courtesy of M.A. Xu-Friedman).Carbon fiber electrodes were calibrated with 1 µM 5-HT (Serotonin HCl, Sigma) at the end of the experiment to convert peak current amplitude of 5-HT transients to concentration.
Three consecutive traces were averaged from each recording condition for analysis.Background subtracted peak 5-HT transients and area under the curve were determined by subtracting the current remaining after TTX (tetrodotoxin citrate, Hello Bio) application from the maximum current measured.
Dopamine HCl and fluoxetine HCl were obtained from Sigma.

Behavioral assays
All behavioral experiments were performed on male mice aged 8 weeks or older.Animals were habituated to the testing room for at least 30 minutes before each experiment and all tests were conducted under dim red-light conditions and white noise to maintain a constant ambient sound unless otherwise noted.
All experiments and analyses were performed blind to genotype.Animals were used in only one behavioral paradigm for the direct social interaction test and fear conditioning.Mice underwent tests for locomotion, anxiety, repetitive behaviors, and depression in the following order: locomotor activity, open field, elevated plus-maze, grooming, marble burying, tail suspension test, forced swim test.At least two days of rest were given in between all tests except for the tail suspension test and forced swim test, during which mice were allowed to rest for at least 7 days in between.Another cohort of mice completed the object interaction test followed by at least 2 days of rest before undergoing rotarod.Behavioral testing apparatuses were cleaned with 0.1% Micro-90 (International Products Corporation) between each mouse.
Locomotor activity: Locomotor activity of each mouse was tracked in photobeam activity chambers (San Diego Instruments) for 90 minutes.Total horizontal movement was measured in 5-minute bins.
Rotarod: Motor coordination and balance were evaluated on a rotarod apparatus (San Diego Instruments) with an accelerating rotarod test.In each trial, mice were habituated to a rod rotating at 6 rpm for 30 sec, then the rotation was increased to 60 rpm over 5 min.The latency to fall was measured over five trials with an interval of 10 min between each trial.Any mice that remained on the apparatus after 5 min were removed and their time was scored as 5 min.Direct social interaction test: The test was adapted from Hitti and Siegelbaum, 2014.Each mouse was placed individually into a standard mouse cage and allowed to habituate for 5 minutes followed by the introduction of a novel male juvenile mouse.The activity was monitored for 10 min and social behavior initiated by the subject mouse was measured by an experimenter sitting approximately 2 meters from the testing cage with a silenced stopwatch.Scored behaviors were described previously (Kogan, Frankland, & Silva, 2000): direct contact with the juvenile including grooming and pawing, sniffing including the anogenital area and mouth and close following (within 1 cm) of the juvenile.After 24 hours, the 10 min test was run again with the previously encountered mouse.Any aggressive encounters observed between animals led to exclusion of the subject mouse from analysis.
Tail suspension test: Mice were placed in a rectangular TST apparatus (28 x 28 x 42 cm) and suspended by their tails which were wrapped in red lab tape at around 3/4 the distance from the base.Movement was monitored for 6 min and the last 4 min were scored for immobility behavior (absence of righting attempt).
Forced swim test: Mice were placed in a plexiglass cylinder (20 cm diameter, 40 cm height) containing 22±1°C water at a depth of 20 cm to prevent them from escaping or touching the bottom.Immobility, measured as floating in the absence of movement except for those necessary to keep the head above water, was measured during the last 4 min of a 6 min session.Following the test, mice were gently dried with a clean paper towel and placed in a fresh cage on top of a heating pad for around 10-15 minutes after which they were returned to their home cage (Yankelevitch-Yahav, Franko, Huly, & Doron, 2015).
Fear conditioning: The fear conditioning paradigm was adapted from Herry et al. (Herry et al., 2008).
Animals were not habituated in the testing room to avoid untimely association with auditory cues.Using the ANY-maze fear conditioning system (Ugo Basile SRL), mice were placed in a fear conditioning cage (17 x 17 x 25 cm) in a sound-attenuating box.The paradigm was performed under no light conditions using two different contexts (context A and B).Mice underwent four phases with 24 hours in between each session: habituation, acquisition, auditory recall, and contextual recall.On day 1 (context A), mice were habituated to five 30 sec presentations of the CS+ and CS-(white noise) at 80 dB sound pressure level.The inter-cue interval was pseudorandomized and each session with the CS+ or CS-was 10 min.
The presentation order of the CS+ and CS-trials were counterbalanced across animals.On day 2 (context A), discriminative fear conditioning was performed by pairing the CS+ with a US (1 sec foot shock, 0.75 mA, 5 CS+/US pairings; intertrial interval: 22-125 s).The onset of the US coincided with the last second of the CS+.On day 3, auditory recall was measured in context B with 5 presentations of CS+ and CS-.On day 4, the contextual recall was measured in context A for 10 min.ANY-maze software was used to analyze freezing behavior (no movement detected for 1 sec), which was scored automatically with an infrared photobeam assay in the fear conditioning cage.
Object interaction test: The test was adapted from Molas et al. (Molas et al., 2017).The apparatus consisted of a custom-made white Plexiglass T-shaped maze (three arms, each 9 x 29.5 x 20 cm, connected through a central 9 x 9 cm zone).Mice were placed in the start arm to habituate to the apparatus for 5 min.Following habituation, they were presented with identical inanimate objects located at opposite ends of the T-maze arms for 5 min/day on two consecutive days.On day 3, one of the inanimate objects was replaced with a novel inanimate object placed in the same location (counterbalanced) for 5 min.The preference ratio was calculated from day 3 data as: (total novel stimulus investigationtotal familiar stimulus investigation) / (total investigation).
Marble burying: 15 sterilized 1.5-cm glass marbles evenly spaced 2 cm apart in three rows of five were placed in a standard mouse cage with a layer of bedding at a depth of 5-6 cm.A mouse was placed in the cage for 30 min, then returned to its home cage.The number of marbles buried (2/3 of their depth covered with bedding) was counted.
Grooming: Self-grooming behavior was scored as previously described (McFarlane et al., 2008;Yang, Zhodzishsky, & Crawley, 2007).Mice were habituated for 5 min in an empty mouse cage with no bedding, then grooming behavior was observed for 10 min by an experimenter sitting approximately 2 meters from the testing cage.Cumulative time spent grooming during the 10 min session was recorded using a silenced stopwatch.
-Figure Supplement 1).Impaired social behavior in Fev/NrxnTKO/RFP mice We investigated the behavior of adult Fev/RFP/NrxnTKO mice in a variety of assays.Basic activities, evaluated by locomotor activity, rotarod performance, and open field, did not differ between Fev/RFP/NrxnTKO mice and Cre-negative controls (Figure 4-Figure Supplement 1).To examine the role of Nrxns in 5-HT system-related behavior, we next assessed social behavior.WT and Open field: Mice were placed in the center of an open arena (41 x 38 x 30.5 cm) facing the furthest wall and allowed to freely explore the arena for 10 min.Time spent in the center of the arena (20.5 x 19 cm) was automatically tracked with EthoVision XT 11.5 (Noldus).Elevated plus-maze: The apparatus (Med Associates) consists of four arms, two enclosed with black walls (19 cm high) and two open (35 x 6 cm), connected by a central axis (6 x 6 cm) and elevated 74 cm above the floor.Mice were placed in the intersection of the maze facing the furthest open arm and allowed to freely explore the maze for 5 min.Time spent in the open and closed arms (index of anxiety-like behavior) and total entries into the open and closed arms (index of locomotor activity) were automatically measured with MED-PC IV software.

Figure 1
Figure 1 Nrxn expression in the raphe nuclei and confirmation of Nrxn deletion in Fev/RFP/NrxnTKO mice.

Figure 2
Figure 2 The lack of Nrxns in 5-HT neurons reduces evoked 5-HT release in the dorsal raphe

Figure 3
Figure 3 The absence of Nrxns in 5-HT neurons decreases 5-HT fiber density in specific brain regions.

Figure 4
Figure 4 The absence of Nrxns in 5-HT neurons impairs social behavior.(A) Direct social interaction test using the same juvenile stimulus across two trials.(B) WT (gray, n = 13) and Fev/RFP/NrxnTKO