GABAergic Neurons in the Rat Medial Septal Complex Express Relaxin-3 Receptor (RXFP3) mRNA

The medial septum (MS) complex modulates hippocampal function and related behaviors. Septohippocampal projections promote and control different forms of hippocampal synchronization. Specifically, GABAergic and cholinergic projections targeting the hippocampal formation from the MS provide bursting discharges to promote theta rhythm, or tonic activity to promote gamma oscillations. In turn, the MS is targeted by ascending projections from the hypothalamus and brainstem. One of these projections arises from the nucleus incertus in the pontine tegmentum, which contains GABA neurons that co-express the neuropeptide relaxin-3 (Rln3). Both stimulation of the nucleus incertus and septal infusion of Rln3 receptor agonist peptides promotes hippocampal theta rhythm. The Gi/o-protein-coupled receptor, relaxin-family peptide receptor 3 (RXFP3), is the cognate receptor for Rln3 and identification of the transmitter phenotype of neurons expressing RXFP3 in the septohippocampal system can provide further insights into the role of Rln3 transmission in the promotion of septohippocampal theta rhythm. Therefore, we used RNAscope multiplex in situ hybridization to characterize the septal neurons expressing Rxfp3 mRNA in the rat. Our results demonstrate that Rxfp3 mRNA is abundantly expressed in vesicular GABA transporter (vGAT) mRNA- and parvalbumin (PV) mRNA-positive GABA neurons in MS, whereas ChAT mRNA-positive acetylcholine neurons lack Rxfp3 mRNA. Approximately 75% of Rxfp3 mRNA-positive neurons expressed vGAT mRNA (and 22% were PV mRNA-positive), while the remaining 25% expressed Rxfp3 mRNA only, consistent with a potential glutamatergic phenotype. Similar proportions were observed in the posterior septum. The occurrence of RXFP3 in PV-positive GABAergic neurons gives support to a role for the Rln3-RXFP3 system in septohippocampal theta rhythm.


INTRODUCTION
Intrinsic neural circuits within, and projections from, the MS subserve various roles of this important brain area in different functions ranging from arousal, attention and spatial working memory (Givens and Olton, 1990;Sweeney et al., 1992;Osborne, 1994). Much research on the MS has centered on characterizing its projections to the hippocampus [see (Zaborszky et al., 2012(Zaborszky et al., , 2014 for review], in addition to descending projections from the MS to the hypothalamus, raphe nuclei and the NI (Borhegyi and Freund, 1998;Leranth et al., 1999;Sánchez-Pérez et al., 2015). Modulation of septal function has been traditionally viewed to derive strongly from ascending projections from the posterior hypothalamus and brainstem, including the raphe nuclei, which have been described as modulators of hippocampal theta rhythm via activation of the septohippocampal projection system (Vertes and Kocsis, 1997;Vertes, 2005). In addition, descending projections from the somatostatin-positive GABA projection neurons of the hippocampus provide a descending feedback regulation of the MS (Toth et al., 1993;Gulyas et al., 2003;Yuan et al., 2017).
However, the less well-studied projection from the NI in the pontine tegmentum also strongly modulates the MS (Goto et al., 2001;Olucha-Bordonau et al., 2003. Specifically, NI projections to the MS are associated with modulation of hippocampal theta rhythm. Electrical stimulation of the NI increased theta rhythm band power of the CA1 hippocampal field potential and NI lesions attenuated the increased hippocampal theta rhythm power induced by stimulation of the nucleus reticularis pontis oralis (RPO) in urethane-anesthetized rats (Nuñez et al., 2006).
A major population of GABA neurons in the NI co-express the neuropeptide, Rln3 (Ma et al., 2007) and NI projections and Rln3-positive fibers are in close contact with cholinergic and GABAergic neurons in the MS (Olucha-Bordonau et al., 2012). Moreover, infusion of a Rln3 analog into the MS increased hippocampal theta rhythm, whereas infusion of a Rln3 receptor antagonist impaired the theta rhythm produced by novel environment exploration or RPO stimulation (Ma et al., 2009). Different approaches in recent years have confirmed and extended these observations regarding the role of the NI and its associated peptide Rln3 in subcortical modulation of hippocampal theta rhythm, with an observed synchrony between the firing of NI neurons and different phases of hippocampal theta rhythm (Ma et al., 2013;Martínez-Bellver et al., 2015. The cognate receptor for Rln3 is the G i/o -protein-coupled receptor, RXFP3. In in vitro studies in Chinese hamster ovary cells transfected with RXFP3, bath application of Rln3 results in inhibition of cAMP synthesis and increased ERK phosphorylation (Liu et al., 2003;van der Westhuizen et al., 2005van der Westhuizen et al., , 2007Bathgate et al., 2013). In agreement with a potential inhibitory effect of neuronal RXFP3 activation, Rln3 and a selective RXFP3 agonist, RXFP3-A2 (Shabanpoor et al., 2012), hyperpolarized RXFP3-expressing magnocellular neurons in the rat paraventricular and supraoptic hypothalamic nuclei (Kania et al., 2017). Furthermore, following intracerebroventricular (icv) infusion of RXFP3-A2, we observed increased phospho-ERK levels in the MS and disruption of spatial working memory in a spatial alternation test (Albert-Gascó et al., 2017), although the precise relationship between these effects is not known.
The MS is composed of a heterogeneous population of neurons and each neuronal type participates in a different way in septo-hippocampal interactions (Sotty et al., 2003). For example, slow firing cholinergic neurons facilitate hippocampal activity (Sotty et al., 2003), while PV GABAergic projection neurons inhibit hippocampal interneurons (Toth et al., 1997). Somatostatin positive neurons are concentrated in the HDB (Köhler and Eriksson, 1984), but to our knowledge, no functional role has been assigned to these neurons. Different types of calcium-binding protein-expressing neurons and neurons expressing choline acetyltransferase (ChAT) are targeted by NI axons/terminals in the rat (Olucha-Bordonau et al., 2012), but it is not clear which of these neurons express RXFP3. Thus, we explored the distribution of Rxfp3 mRNA expression in different neuronal types of the rat septal area using multiplex in situ hybridization and specific probes for Rxfp3, ChAT, vGAT (slc32a1), PV, and somatostatin (SOM) transcripts.

Animals
Experiments were conducted with approval from The Florey Institute of Neuroscience and Mental Health Animal Ethics Committee, in compliance with guidelines of the National Health and Medical Research Council of Australia. Adult male Sprague-Dawley rats weighing 300-320 g were maintained on a 12-12 h light-dark cycle with lights on at 0700 h. Rats were provided free access to food and water.

Multiplex in Situ Hybridization (ISH)
The distribution of septal Rxfp3 mRNA-positive neurons and their GABAergic or cholinergic phenotype was assessed using RNAscope multiplex in situ hybridization. RNAscope R is a commercial method provided by Advanced Cell Diagnostics (ACD, Newark, CA, United States), which involves the incubation of post-fixed, fresh-frozen brain sections with up to three custom probes. Standard probes contain 20 ZZ pairs (25 base pairs/Z) which cover a total of ∼1000 base pairs of the target mRNA. In silico verification of the probes is performed and validated to select oligonucleotides with compatible melting temperature for optimal hybridization under RNAscope assay conditions and minimal cross-hybridization to off-target sequences. There is a verification procedure conducted following each major step during the probe design to guarantee accuracy, according to previously described rules (Wang et al., 2012).

Imaging and Quantification of Co-expression of Transcripts
Fluorescence images were taken with an LSM 780 Zeiss Axio Imager 2 confocal laser scanning microscope (Carl Zeiss AG, Jena, Germany). The system is equipped with a stitching stage, and Zen software (Carl Zeiss AG) was used to stitch tiled images taken with a 20 × objective. Quantification of cellular colocalization of transcripts (one section/bregma level, rat and probe combination) was conducted manually using Fiji (Schindelin et al., 2012) [Note: Results consistent with those observed in the sections assessed, were also observed in adjacent brain areas, and in other rat brains, for all probes]. The total number of positive neurons for each region was counted separately, relative to DAPIstained nuclei, to avoid bias. The percentage co-expression of transcripts was related to the total number of Rxfp3 mRNApositive neurons in each of the septal areas. Higher-power, (inset) images to illustrate co-localization were taken using a 40 × objective.

RESULTS
In these experiments, we assessed the rostrocaudal distribution of Rxfp3 mRNA-positive neurons in the MS/diagonal band, LS, triangular septal nucleus, and SFi, and determined whether these Rxfp3 mRNA-positive neurons co-expressed ChAT or vGAT mRNA or PV and/or vGAT mRNA (or SOM and/or vGAT mRNA). All these neurotransmitter-related transcripts and their related proteins or peptides have been described as clear markers of the onion-like structure of the septum. According to Wei et al. (2012), the onion-like MS can be described as a five-layer structure with layers determined by their highest density marker (MS-1-MS-3, LSv, and LSi). Layers are distributed from the midline to the LSi with MS-1 on the midline, rich in PV neurons; followed by MS-2, rich in ChAT neurons; followed by MS-3, rich in nNOS; followed by CR (LSV), and CB (LSi). The following results illustrate a high level of co-localization of Rxfp3 and vGAT mRNA in neurons in most septal regions. In contrast, in caudal septal regions and diagonal band, no co-localization of Rxfp3 with vGAT mRNA occurred, suggesting an alternative non-GABAergic phenotype ( Table 1).

Triangular Septal Area, and Septofimbrial and Dorsolateral Septal Area Contain Heterogeneous Populations of Rxfp3 mRNA-Positive Neurons
In the most caudal region of the septum analyzed (bregma −0.24 mm), the distribution and phenotype of Rxfp3 mRNApositive neurons varied within the different nuclei. In the LSD, Rxfp3 mRNA-positive neurons were widely and evenly distributed (Figures 9A,B) and were mainly vGAT mRNApositive (∼86%; 12/14 neurons), with a small number of neurons located near the SFi that were vGAT mRNA negative (∼14%; 2/14 neurons; Figures 2, 9K-N).
Rxfp3 mRNA-positive neurons in the SFi were mainly distributed in the most dorsal part of the nucleus near the corpus callosum (cc). In the ventral SFi, Rxfp3 mRNA-positive neurons were lower in number (Figures 9A,B). Throughout the dorsal and ventral SFi, the majority of Rxfp3 mRNA-positive neurons co-expressed vGAT mRNA (∼75%; 75/100 neurons; Figures 2, 9L-O).
The triangular septal area (TS) contained three Rxfp3 mRNApositive neuron populations based on their differential phenotype and distribution. Dispersed Rxfp3 mRNA-positive neurons were present in the most dorsal portion near the midline, while in the most ventral TS, a large, densely packed population of Rxfp3 mRNA-positive neurons were distributed alongside the border with the SFO (Figures 9A,B). The ventral TS area was rich in vGAT mRNA-expressing neurons, while the dorsal TS was not. Rxfp3 mRNA-positive neurons in the dorsal TS were generally vGAT mRNA-negative, while in the ventral TS, Rxfp3 mRNA-positive neurons were generally vGAT mRNApositive. In the lateral part of the ventral TS, there was a population of Rxfp3 mRNA-positive neurons which were vGAT mRNA-negative. Overall, ∼79% of Rxfp3 mRNA-positive neurons in TS co-expressed vGAT mRNA (193/245 neurons), while the remainder were negative (∼21%; 52/245 neurons; Figures 2, 9C-K).

DISCUSSION
In this study, we have employed RNAscope multiplex in situ hybridization (Wang et al., 2014a,b;Li and Kim, 2015) to characterize the neurochemical phenotype of Rxfp3 mRNApositive neurons in the rat septal area. The highly specific nature of the method means that these data represent a more accurate estimation of the distribution of RXFP3 than studies using putative antisera against the receptor protein (Meadows and Byrnes, 2014), although this powerful approach does not provide information about the subcellular location of RXFP3, which might be available with alternative protein detection methods. In the septal area, neurons expressing Rxfp3 transcripts were concentrated in the MS complex, including the diagonal band nuclei, and in the posterior septum, including the SFi and TS nuclei. Some Rxfp3 mRNA-positive neurons were also detected in LS divisions. Our findings are consistent with previous studies of the presence and distribution of Rxfp3 mRNA in the rat septal area detected using radioactive oligonucleotide probes. Specifically, MS and HDB displayed moderate to high levels of Rxfp3 mRNA, while in LSI and VDB, expression was moderate (Sutton et al., 2004;Ma et al., 2007). These findings are consistent with concurrent studies of these and other transcripts in rat hippocampus (Ma and Gundlach, unpublished data). Therefore, in light of the strong innervation of the rat septal region by Rln3-positive nerve fibers, the presence of septal RXFP3 binding sites, and functional studies (Ma et al., 2007(Ma et al., , 2009, we conclude that the detection of Rxfp3 mRNA reflects the expression of functional RXFP3 protein by these neurons. In the MS, VBD, and HDB, the vast majority of Rxfp3 mRNA-positive neurons co-expressed vGAT mRNA. Furthermore, a population of these presumed GABAergic Rxfp3 mRNA-positive neurons are PV mRNA-positive (Wei et al., 2012). Septal PV/GABA neurons are the main source of the GABAergic projections to the hippocampus and specifically target hippocampal interneurons (Freund and Antal, 1988;Freund and Gulyas, 1997). A number of studies have demonstrated that PV/GABA neuron activity is crucial for hippocampal theta rhythm (Borhegyi et al., 2004;Bassant et al., 2005;Simon et al., 2006). The modulation of the GABAergic inter-neuronal inhibition of hippocampal pyramidal neurons has been reported to be a source for hippocampal theta rhythm synchronization (Toth et al., 1997). In addition, septal PV/GABA neurons expressing cyclic nucleotide activated, non-selective cation channels play a role in driving hippocampal theta rhythm (Varga et al., 2008;Hangya et al., 2009). Notably, RXFP3 activation results in inhibition of cellular cAMP synthesis in cell-based assays in vitro (Liu et al., 2003;van der Westhuizen et al. 2007van der Westhuizen et al. , 2010, consistent with a similar interaction in vivo (see further discussion below).
In contrast to the strong association with GABAergic neurons, only a small number of cholinergic (ChAT mRNA-positive) neurons co-expressed Rxfp3 mRNA. However, anterograde neural tract-tracing and immunohistochemical studies suggest that cholinergic (ChAT-positive) septal neurons receive a robust innervation from the Rln3 rich NI (Olucha-Bordonau et al., 2012). Thus, the influence of NI neurons on the septal cholinergic system might be mediated by NI neurons that contain GABA only or other peptides, such as cholecystokinin, which is expressed in the NI (Kubota et al., 1983;Olucha-Bordonau et al., 2003) (Ma and Gundlach, unpublished data).
The discovery that Rxfp3 mRNA is absent from MS cholinergic neurons provides new insights into the nature of the coordinated neural actions that result in the generation and modulation of hippocampal theta rhythm, and since RXFP3 activation often produces neuronal inhibition in vitro Kania et al., 2017), it is possible that pERK activation in MS cholinergic neurons occurs via RXFP3-mediated inhibition of non-PV, GABAergic interneurons (Leranth and Frotscher, 1989). In this regard, optogenetic activation of cholinergic septohippocampal neurons suppressed ripple sharp waves and enhance theta rhythm oscillations (Vandecasteele et al., 2014) and local circuit inhibitory actions on cholinergic neurons are a primary process in the generation of septal rhythmicity (Leão et al., 2015). Furthermore, icv infusion of an RXFP3 agonist (RXFP3-A2; Shabanpoor et al., 2012) resulted in increased phosphorylation of ERK in the MS, mainly in ChAT-immunoreactive neurons (Albert-Gascó et al., 2017). Given the observed absence of Rxfp3-and ChAT mRNA-positive neurons in the MS in the present study, and the observation that RXFP3 activation routinely induces neuronal inhibition (Kania et al., 2017), there is a possibility that the pERK activation within the cholinergic neurons occurs via a reduction in local circuit inhibition within the MS.
In addition, ∼25% of the Rxfp3 mRNA-positive neurons in the MS were non-GABAergic, non-cholinergic in nature. Although further studies are required to better identify the phenotype of these neurons, it is presumed that some or many are glutamatergic neurons, since they constitute ∼25% of the total MS neuron population (Colom et al., 2005;Gritti et al., 2006). Glutamatergic neurons provide both local and septohippocampal projections (Manseau et al., 2005;Henderson et al., 2010;Huh et al., 2010) and interestingly, optogenetic activation of MS glutamatergic neurons produces strong theta rhythm synchronization, mainly mediated by local septal circuits (Robinson et al., 2016).
Considerable data suggest a strong link between RXFP3 activation in the MS and modulation of hippocampal theta rhythm. Hippocampal theta rhythm has been traditionally associated with arousal mechanisms which are directly involved in attentional mechanisms of memory (Vinogradova, 1995). The NI, along with other brainstem areas, the hypothalamus and the basal forebrain, promote arousal and fast electroencephalographic (EEG) rhythms (Brown and McKenna, 2015;Korotkova et al., 2018). Moreover, stimulation of the NI promotes arousal and is associated with cortical EEG desynchronization, increased locomotor activity, and headscanning vigilance behavior during fear recall (Ma et al., 2017). In addition, ipsilateral NI stimulation induces locomotion and rotation at latencies consistent with a role in the modulation of premotor areas like the basal forebrain (Farooq et al., 2016). Furthermore, Rln3 and Rxfp3 gene knockout mice display reduced voluntary running wheel activity during the dark, active phase (Smith et al., 2012;Hosken et al., 2015) providing further evidence for a likely role for this signaling system in sustained arousal and related locomotor and exploratory activity.

CONCLUSION
The strong expression of Rxfp3 mRNA by GABAergic neurons in the rat MS and adjacent nuclei, is consistent with the central role of these neurons in the control of hippocampal theta rhythm by actions on local septal circuits. In turn, these actions may indirectly influence septal cholinergic neurons/circuits and hippocampal interneurons via septohippocampal projections. Notably, independent studies have revealed a strong Rln3 innervation of the hippocampus and identified Rxfp3 mRNA expression by hippocampal GABA neurons in the rat (Ma and Gundlach, unpublished data), consistent with direct actions of Rln3/RXFP3 signaling on these circuits. Therefore, further studies of the neurotransmitter and neurochemical phenotype of septal and hippocampal neurons that express Rxfp3 mRNA and their precise functional roles are warranted in both normal adult rats and mice, and in models of neuropathology and cognitive and psychiatric disorders.

AUTHOR CONTRIBUTIONS
HA-G and SM designed and performed the experiments, prepared the figures and drafted the manuscript. FR-B and AMS-P contributed to some experiments, figure preparation and corrections of the draft manuscript. ALG and FEO-B planned the research, supervised the experiments, edited the figures, and prepared the final version of the manuscript.