Co-transmitting neurons in the lateral septal nucleus exhibit features of neurotransmitter switching

The lateral septal nucleus (LSN) is a highly interconnected region of the central brain whose activity regulates widespread circuitry. As such, the mechanisms that govern neuronal activity within the LSN have far-reaching implications on numerous brain-wide nuclei, circuits, and behaviors. We found that GABAergic neurons within the LSN express markers that mediate the release of acetylcholine (ACh). Moreover, we show that these vGATLSN neurons release both GABA and ACh onto local glutamatergic LSN neurons. Using both short-term and long-term neuronal labeling techniques we observed expression of the cholinergic neuron marker Choline Acetyltransferase (ChAT) in vGATLSN neurons. These findings provide evidence of cholinergic neurotransmission from vGATLSN neurons, and provide an impetus to examine dynamic co-neurotransmission changes as a potential mechanism that contributes to neuronal and circuit-wide plasticity within the LSN.


Introduction
Co-transmission of multiple neurotransmitters from the same neuron provides an adaptable means to regulate neural circuit activity. This is especially true in cases in which both an excitatory and an inhibitory neurotransmitter are released from the same neuron. Maintaining excitatory/inhibitory balance from multiple sources is crucial for proper functioning of neural circuits (Tatti et al., 2017). Consequently, when multiple different neurotransmitters are released onto the same postsynaptic cell, neurons and their respective circuits are afforded an increased spatiotemporal precision in regulating postsynaptic neuronal activity (Root et al., 2014;Shabel et al., 2014). This is opposed to when each neurotransmitter is released onto different postsynaptic neurons, which provides co-releasing neurons the unique ability to activate certain nodes of the circuit while simultaneously inhibiting others (Granger et al., 2016). The flexibility of such a signaling paradigm enables an expansion of signaling diversity, while also facilitating the ability of neural circuits to quickly respond to changes in their environment. Mechanisms that drive neural circuit plasticity provide even further flexibility, allowing circuits and the neurons within them to adapt to changing environmental cues. The co-transmission of acetylcholine (ACh) and GABA has been well documented within multiple nuclei and brain regions (Lee et al., 2010;Saunders et al., 2015bSaunders et al., , 2015aGranger et al., 2016Granger et al., , 2020Case et al., 2017;Takács et al., 2018). However, whether there is a continuous capability for plasticity within these circuits into adulthood, and the role that this plasticity plays in co-transmission is not well understood.
In response to activity-dependent cues, certain populations of neurons alter transcription or release of new or different neurotransmitters (Spitzer, 2015(Spitzer, , 2017. Neurotransmitter switching often changes inhibitory neurons to excitatory neurons, or vice-versa, allowing key neurons within a circuit the ability to change the sign of their effect. Cellular state plasticity is capable of altering behavior (Dulcis et al., 2013;Spitzer, 2015Spitzer, , 2017 through transcription-based changes in ACh and GABA release . Though neurotransmitter switching might help explain the co-transmission of multiple neurotransmitters from the same neuron (Spitzer, 2015;Granger et al., 2016), it is not known whether neurotransmitter switching affects co-transmission events recorded throughout the brain. Moreover, only a limited number of brain regions have been identified in which dynamic neurotransmitter switching occurs (Spitzer, 2017).
The Lateral Septal Nucleus (LSN) is a centrally located and highly interconnected nucleus within the forebrain that lies just medial to the lateral ventricles. Traditionally this nucleus has been implicated in the regulation of anxiety and aggression (Albert and Richmond, 1976;Sheehan et al., 2004;Singewald et al., 2011;Leroy et al., 2018), however recent reports have uncovered previously overlooked roles for the LSN Yang, 2015, 2016). Studies of the LSN have primarily focused on GABAergic (Zhao et al., 2013;Sweeney and Yang, 2016;Leroy et al., 2018;van der Veldt et al., 2021) and glutamatergic neurons (Kocsis et al., 2003;Lin et al., 2003;Sheehan et al., 2004). However, much remains unknown regarding other cell types within the LSN, including their mode of neurotransmission, their projection targets, and how these features dictate behavioral regulation. Though LSN neurons develop from the same neurogenic niche as the cholinergic neurons of the medial septum (Verney et al., 1987;Sheehan et al., 2004), cholinergic neurons within the LSN have not been previously investigated (Ichikawa et al., 1997;Lein et al., 2007;Rossi et al., 2011;Saunders et al., 2015a).
Here we present evidence of cholinergic transmission from the LSN. Moreover, we identified ACh/GABA co-expressing and co-transmitting neurons within the LSN that exhibit features of neurotransmitter switching. We find that these co-transmitting neurons project onto local glutamatergic neurons within the LSN, thereby regulating local neural circuit activity.

Fluorescent in situ hybridization
All brain sections analyzed by FISH were prepared by the In Situ Hybridization core at the Jan and Dan Duncan Neurological Research Institute at Texas Children's Hospital. ChAT RNA was stained for using the following probe: TTGCTTGGTGTGGAACAGTGCCGGTTCGGTGCGTAA-CAGCCCAGGAGAGCAGGTCGGCAGCTCTGCTACTCTGGATTAAGAA TCGCTAGGATGCCTATCCTGGAAAAGGTCCCCCCAAAGATGCCTGTA-CAAGCTTCTAGCTGTGAGGAGGTGCTGGACTTACCTAAGTTGC-CAGTGCCCCCACTGCAGCAAACCCTGGCCACCTACCTTCAGTGCATG-CAACACCTGGTACCTGAAGAGCAGTTCAGGAAGAGCCAGGCCATT GTGAAGCGGTTTGGGGCCCCTGGTGGCCTGGGTGAGACCCTGCAGGA AAAGCTCTTGGAGAGACAGGAGAAGACAGCCAATTGGGTCTCTGAA-TACTGGCTGAATGACATGTATCTAAACAACCGCCTGGCCCTGC CAGTCAACTCTAGCCCTGCTGTGATCTTTGCTCGGCAGCACTTCCAA-GACACCAATGACCAGCTAAGGTTTGCAGCCAGCCTCATCTCTGGT GTGCTTAGCTACAAGGCTCT (Lein et al., 2007). vGAT RNA was stained for using the following probe: GCCATTCAGGGCATGTTCG (Lein et al., 2007). Adult mouse brains were collected and immediately embedded and frozen in Tissue-Tek OCT compound. All coronal sections were collected sequentially from the forebrain at 25 µm thickness. In situ hybridization was performed as previously described (Yaylaoglu et al., 2005). Stained sections containing the LSN were imaged using a Leica TCS SP8 STED microscope.

Immunohistochemistry and microscopy
Animals were deeply anesthetized using isoflurane and were transcardially perfused with PBS followed by 4% PFA (diluted from 16% Paraformaldehyde EM Grade No. 15710 Electron Microscopy Sciences). Brains were dissected and post-fixed in 4% PFA overnight at 4 degrees Celsius. Brains were cryoprotected in 30% sucrose/PBS solution for 1 more day at 4 degrees Celsius. Brains were then embedded and frozen in O.C.T. (Fisher HealthCare No. 4585) and stored at − 80 degrees Celsius until sliced. Brains were sliced using a cryostat (Leica CM1860) in coronal 40 µm sections. For anti-ChAT immunohistochemistry, 40 µm freefloating sections were blocked for 1 hr at room temperature in 10% donkey serum blocking solution made in PBS-T (1X PBS, 1% Triton-X 100, pH 7.35). Sections were incubated for 14 days at 4 degrees Celsius at 1:500 dilution of block solution containing goat anti-ChAT primary antibody (Millipore AB144P). Sections were washed 5 times for 10 min in plain PBS-T. Sections were then incubated in secondary antibodies (1:500 donkey anti-goat Alexafluour-488) for 2 hr at room temperature. Sections were then washed 5 times at 10 min each in PBS-T. All sections were mounted using DAPI Fluoromount-G (SouthernBiotech, 0100-20). Sections from ChAT-Cre;vGAT-Flp;Tri-cistronic mice, and mice injected with AAV were washed with PBS 5 times for 10 min, blocked at 1 hr at room temperature in 10% donkey serum blocking solution, washed 5 times with PBS-T for 10 min, and then mounted with DAPI Fluoromount-G. Detection of fluorescent expression was performed using a Leica TCS SPE confocal microscope, or Leica TCS SP8 STED microscope.

Channelrhodopsin electrophysiology
Brain slices containing the LSN were prepared from 12 to 16 weekold animals that had been virally injected with AAV-DIO-ChR2-EYFP and AAV-fDIO-mRuby2 between 2 and 4 weeks prior to recording. Cells were patched using voltage clamp internal solution. Cells were recorded for a minimum of 10 sweeps for each pharmacological condition. Each sweep consisted of 2 s of recording, with an excitation duration of 1 ms of blue light (470 nm) exposure with 30 s between each sweep. ACSF with the following pharmacological concentrations in the slice chamber were used: 1 µM TTX (Tocris, 0.5 mM 4-AP (Tocris), 20 µM AP-5 (Tocris), 10 µM CNQX (Tocris).

Quantification
The number of cells expressing viral, or FISH signal, as well as the amount of EYFP signal expressed following AAV-nEF-COn-FOn-ChR2-EYFP injections were quantified using the Imaris (Bitplane AG, Zürich, Switzerland) software.

Contact for reagent and resource sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Benjamin Arenkiel (Arenkiel@bcm.edu).

LSN neurons co-express cholinergic and GABAergic markers
To identify the distribution of ACh/GABA co-transmitting neurons throughout the brain, we first sought to label all neurons that express markers for the release of these two neurotransmitters. We identified the acetylcholine synthetic enzyme, ChAT, as well as the vesicular GABA transporter, vGAT, as reliable and faithful markers of cholinergic and GABAergic neurons respectively (Armstrong et al., 1983;Chaudhry et al., 1998;Rossi et al., 2011;Vong et al., 2011;Gizowski and Bourque, 2020;Graybuck et al., 2021). To identify putative co-transmitting neurons, we generated mice with three transgenic alleles to facilitate cell type-specific labeling. These included ChAT-Cre (JAX Stock No. 031661), vGAT-Flp (JAX Stock No. 029591), and a tri-cistronic fluorescent label that is Cre-and Flp-dependent ( Supplementary Fig. 1A) (Lusk et al., 2022). In these animals, all neurons that are both ChAT-positive and vGAT-positive express the tri-cistronic allele, and are thus labeled with a nuclear BFP, a cytosolic GFP, and a membrane-bound tdTomato ( Fig. 1A and Supplementary Fig. 1). From this cross, we identified tri-cistronic-positive neurons in the LSN ( Fig. 1B-C), revealing a population of cholinergic and GABAergic co-expressing neurons in this area.
Certain neuronal populations switch the expression of neurotransmitter-related genes during development (Kotak et al., 1998;Nabekura et al., 2004;Spitzer, 2017). To confirm that these fluorescently labeled neurons are indeed co-expressing neurons rather than neurons that only expressed these markers at some point in development, we sought to verify the expression of cholinergic and GABAergic markers in the adult mouse brain. Towards this, we injected a Cre-dependent mVenus adeno-associated virus (AAV-DIO-mVenus) and a Flp-dependent mRuby2 adeno-associated virus (AAV-fDIO-mRuby2) into the LSN of adult ChAT-Cre; vGAT-Flp mice (Fig. 1D). We identified a large number mVenus-positive cholinergic neurons, mRuby2-positive GABAergic neurons, and dually labeled co-expressing neurons ( Fig. 1E-F). These findings confirmed the existence of ChAT/vGAT co-expressing neurons within the adult LSN and suggested that labeled cells identified via genetic reporter expression were not just a remnant of neurotransmitter switching during development.

FISH and antibody-based immunohistochemistry detect co-expressing LSN neurons
Noting the paucity of literature describing either cholinergic neurons or ChAT/vGAT co-expressing neurons within the LSN (Ichikawa et al., 1997;Rossi et al., 2011;Saunders et al., 2015a), we performed fluorescent in situ hybridization (FISH) to label ChAT and vGAT transcripts throughout the basal forebrain. Consistent with what we observed via genomic and viral labeling experiments, indeed we identified ChAT/v-GAT co-expressing neurons in the LSN ( Fig. 2A-B). ChAT expression in these neurons appeared lower when compared to nearby ChAT positive neurons in the vertical domain of the diagonal band of Broca (VDB) (Fig. 2A, bottom left). Indeed, when examining data from the Allen Brain Atlas, we found a similar pattern of low ChAT expression within the vGAT-rich region of the LSN (Supplementary Fig. 2) (Lein et al., 2007). Importantly, we found ChAT expression only within the lateral portion of the LSN, juxtaposing the lateral ventricle (+1.1 to +0.38 mm from Bregma). This excluded the most anterior and posterior portions of the LSN, which extends from Bregma + 1.7 to − 0.45 mm (Franklin and Paxinos, 2008). To further characterize these neurons, we performed immunohistochemistry (IHC) to stain for ChAT expression, which allowed us to directly observe ChAT-positive LSN neurons. These neurons had less fluorescent intensity when compared to ChAT-positive neurons within the VDB (Fig. 2C, bottom left), and consistent with FISH data, we only identified these neurons within the lateral LSN ( Fig. 2C-D).

Number of co-expressing LSN neurons increases with time
Co-expressing neurons have been shown to exist in a transcriptionally dynamic state, in which one or more of the neurotransmitterspecific markers changes expression over time (Spitzer, 2015;Dulcis et al., 2017;. This type of expression profile provides insight into the function of the co-transmitting neuron, while also providing a potential explanation as to why these populations have not received previous attention. To address the possibility of dynamic ChAT and/or vGAT expression within lateral LSN neurons, we leveraged a Creand Flp-dependent viral channelrhodopsin-EYFP reporter (AAV-nEF--COn-FOn-ChR2-EYFP) (Fenno et al., 2014) to label both the cell bodies and membranes of ChAT-Cre and vGAT-Flp co-expressing neurons within the LSN (Fig. 3A). We selected a membrane bound reporter to increase the detection of labeled neuronal surface area, thereby allowing us to measure a greater dynamic range in reporter expression. Using this vector, we confirmed potent expression of the conditional AAV reporter 8 wks following injection (Fig. 3B). After this period, we harvested tissue and imaged samples at three additional time points 16 wks apart (Fig. 3C). We hypothesized that if neurotransmitter switching was occurring, by allowing longer times for the viral label to incubate, we would detect increased numbers of co-expressing neurons. Indeed, through this approach we found increased expression of the Cre-and Flp-dependent viral label over time, exhibiting both increased signal intensity and increased number of EYFP-positive cell bodies (Fig. 3D). To ensure that the injected viral reporter marked only cells that expressed both Cre and Flp, we also injected this same construct into control wildtype animals and incubated the marker for 24 weeks. After this incubation period, we found no expression of the reporter (Supplementary Fig. 3), suggesting that increased viral reporter expression observed in LSN neurons (Fig. 3E) was due to ChAT and vGAT expression dynamically changing within that population over time. To determine whether the increased co-expression of ChAT and vGAT was due to aging, we performed FISH analysis in wildtype mice at 24 and 44 weeks of age, comparing the number of ChAT/vGAT co-positive neurons in the LSN at these two ages. This analysis found no difference in the number of ChAT/vGAT co-expressing neurons between the two groups (Supplementary Fig. 4).

LSN neurons co-release ACh and GABA onto glutamatergic LSN neurons
To validate the functional expression of ACh and GABA in LSN neurons, we next tested whether the identified co-expressing LSN neurons release both ACh and GABA onto postsynaptic targets via electrophysiological recordings. Due to the large number of GABAergic neurons within the LSN ( Fig. 2A,D, Supplementary Fig. 2B) (Lein et al., 2007;Zhao et al., 2013), we targeted the light-gated ion channel, channelrhodopsin (ChR2) to vGAT LSN neurons, and subsequently measured for cholinergic and GABAergic co-transmission onto postsynaptic targets. Assuming that subpopulations of vGAT LSN neurons would have connections within the LSN, we labeled local glutamatergic neurons by injecting a combination of Cre-dependent ChR2 (AAV-DIO-ChR2-EYFP) and Flp-dependent mRuby2 (AAV-fDIO-mRuby2) viruses into the LSN of vGAT-Cre; vGlut2-Flp (JAX Stock No. 030212) mice (Fig. 4A-B). While performing whole-cell recordings from red labeled vGlut2 LSN neurons, and simultaneously photostimulating vGAT LSN inputs in the presence of both TTX and 4-AP, we found that 100% of patched vGlut2 LSN neurons received fast, monosynaptic, GABAergic currents (n = 15/15 cells), and 20% of these vGlut2 LSN neurons also received small (5-20 pA) metabotropic, cholinergic currents (n = 3/15 cells), as evidenced by light-evoked currents that were blocked by the application of the mAChR-specific antagonist atropine (Fig. 4B,C). Application of the nicotinic receptor antagonist mecamylamine did not affect photo-evoked currents.
Additionally, we injected a mixture of AAV-DIO-ChR2-EYFP and AAV-fDIO-mRuby2 viruses into the LSN of ChAT-Cre; vGlut2-Flp mice, to record cholinergic and GABAergic release from ChAT-Cre-positive LSN neurons onto vGlut2-Flp-positive LSN neurons. Recording 17 cells from 4 separate animals, we found neither cholinergic, nor GABAergic responses in mRuby2-positive cells (data not shown), suggesting the numbers of co-transmitting neurons is likely small.
Together, these data show that a subpopulation of vGAT LSN neurons indeed release both GABA and ACh onto postsynaptic target neurons, demonstrating that ACh/GABA co-transmission occurs within the LSN.

Discussion
Understanding how neurons function together to manifest emergent circuit properties of the brain remains largely an ongoing challenge. This is due, at least in part, to our lack of a full mechanistic understanding of the detailed modes of neurotransmission between diverse populations of brain cells. Recent advances in neural circuit research have uncovered an incredible diversity of cell types and synaptic connections throughout the brain (Saunders et al., 2018;Huang et al., 2020). However, many of these advances account only for the expression and connectivity patterns of neural circuits statically in time and are largely unable to define how neurons within these circuits dynamically change.
Evidence demonstrating circuit and organism-level transcriptional changes throughout development, and in response to diverse environmental stimuli, illustrates the necessity of understanding how neural circuit activity changes in different contexts. Assays including in situ staining techniques and sequencing approaches account for the state of neuronal populations only statically due to the short temporal nature of these assays. However, genetic and conditional viral labeling techniques allow for the incubation of a conditional marker within defined populations of neurons, thereby labeling any neuron that satisfies the condition of the reporter over the course of incubation. By combining both short-and long-timescale vital labeling techniques we have uncovered a population of ACh/GABA co-transmitting neurons within the lateral LSN, that changes the expression of neurotransmission related genes over time. We have found that ChAT expression in this neuronal population appears low using FISH and IHC. This may explain why the presence of these neurons has been largely overlooked.
To explore whether ChAT/vGAT co-expressing neurons within the LSN exhibited dynamic transcriptional regulation, we leveraged a conditional viral reporter that expressed a membrane bound label only in the presence of both Cre and Flp. The observed increase in expression of this fluorescent reporter is congruent with ChAT expression in vGAT LSN neurons, which changes over time. Here we propose that vGAT-Flppositive neurons lowly express ChAT-Cre at distinct periods throughout time. This low expression of Cre, in the presence of vGATdriven Flp, is sufficient to permanently activate the Cre and Flp dependent conditional reporter, resulting in the increased labeling seen at later time points. The increased number of EYFP-positive cell bodies suggests that an increasing number of vGAT LSN neurons are turning on expression of the ChAT-Cre allele at some point in time. This experimental approach provides no evidence of cessation of ChAT expression, as the EYFP marker is permanently expressed once it has been activated in these neurons (Fig. 3C-E). However, these findings illustrate a dynamic expression of ChAT in vGAT LSN neurons. Importantly, though this dynamic expression in the context of neural co-transmission demonstrates features of neurotransmitter switching, this observation does not strictly meet the criteria to be considered a bona fide neurotransmitter switching event (Spitzer, 2017), as there is no evidence of loss of vGAT expression in these neurons. Additionally, many of the reported cases of neurotransmitter switching have been associated with an inciting stimulus that drives the observed transcriptional changes. At the time of this report, we have not identified a conserved stimulus, if any, that is able to selectively drive the observed expression changes. However, based on our findings, we reason that whatever factor causes the changes in expression is experienced by the mice in an activity-dependent manner over time in their home cages.
The function of such dynamic transcriptional changes is not obvious. That ChAT is responsible for the synthesis of ACh suggests that these observed transcriptional changes might give neurons the ability to release ACh. Indeed, we find that, along with GABA, vGAT LSN neurons release ACh onto local glutamatergic neurons, implying that the observed transient expression of ChAT is at least sufficient to drive ACh production and release onto a subset of downstream circuitry. During these experiments we observed a residual bicuculline-, mecamylamineand atropine-resistant current. Upon further analysis, we found that this current has a reversal potential near − 55 mV, suggesting that it is likely a chloride-linked current. However, the current is not sensitive to bath application of the GABA C receptor blocker, TPMPA at 20 μM (data not shown). Thus, it is possible that this current is the result of incomplete inhibition by the competitive antagonist bicuculline. However, further experimentation is necessary to confirm this hypothesis. That we observed evidence of co-transmission postsynaptically in only a small fraction of the vGlut2 LSN neurons likely reflects the rare and transient expression of ChAT in the vGAT LSN population. However, this might also reflect variation in the expression of muscarinic and nicotinic receptors by vGlut2 LSN neurons. Dynamic changes in postsynaptic receptor expression that pair with pre-synaptic neurotransmitter expression changes have been described in other regions of the rodent brain (Kotak et al., 1998;Korada and Schwartz, 1999). Future work will be necessary to determine the receptor expression profile of vGlut2 LSN neurons, as well as the frequency and magnitude of ChAT expression changes in vGAT LSN neurons over time, and to what stimulus.
The low intensity of ChAT staining within LSN neurons using traditional staining techniques is likely the reason why this population has not garnered previous attention. This could explain why leveraging conditional genetic methods to label these neurons provides better visualization, as the high recombination efficiency of Cre, even at low concentrations, allows for strong expression of the conditional reporter. However, this hypothesis cannot fully explain the increased signal observed when incubating a Cre-and Flp-dependent marker over longer periods of time. Rather, this observed increase in the intensity of signal and in the number of EYFP-positive neurons is more consistent with dynamic changes in ChAT expression within the vGAT LSN population. Moreover, electrophysiological evidence suggests that the observed ChAT-expression is sufficient for the release of ACh from this neuronal population.
We identified local vGlut2 LSN neurons as a postsynaptic target of both ACh and GABA release from co-transmitting LSN neurons, and the muscarinic nature of the cholinergic component of this co-transmission suggests a role for this concerted release in regulating the activity of postsynaptic targets. In such a model, muscarinic signaling may modulate the activity of postsynaptic targets in conjunction with synaptic GABAergic inhibition. Similar synergistic regulation has been reported in other areas of the brain, including the olfactory bulb (Pressler et al., 2007;Zhan et al., 2013;Smith et al., 2015), and striatum (Suzuki and Momiyama, 2021). Alternatively, ACh release may be exerting a presynaptic effect on vGAT LSN neurons similar to what has been previously reported (Grilli et al., 2009;Nakamura and Jang, 2012). However, future experimentation is needed to assess these mechanistic possibilities.
Finally, we report that transcriptional changes in adult mice, previously known to be the primary mechanism driving neurotransmitter switching, also constitutes a candidate mechanism to facilitate neural co-transmission. Attention in these two fields of studyneurotransmitter switching and neural co-transmission -is rapidly growing. However, to date, only a few examples have linked the two neural phenotypes (Furshpan et al., 1976;Spitzer, 2015). This work provides further rationale to examine both neural states together. Whether previously reported examples of neural co-transmission reflect neurons in a transient transcriptional state needs to be more closely examined. The results of these future experiments will help us better understand both neural circuit biology, and the plasticity mechanisms that govern the astounding dynamics that define neuronal circuits.

Funding sources
This work was supported by the American Heart Association (AHA 20PRE35040011 to PH), the McNair Medical Institute, and the National Institutes of Health (R01 EB027145-01A1/04 and R01 NS078294-06A1 to BA). This project was also supported in part by the RNA In Situ Hybridization Core facility at Baylor College of Medicine, which is supported by a Shared Instrumentation grant from the NIH (1S10OD016167) and the NIH IDDRC grant P50 HD103555 from the Eunice Kennedy Shriver National Institute of Child Health & Human Development. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health & Human Development or the National Institutes of Health.