Connectivity and molecular profiles of Foxp2- and Dbx1-lineage neurons in the accessory olfactory bulb and medial amygdala

Abstract

In terrestrial vertebrates, the olfactory system is divided into main (MOS) and accessory (AOS) components which process both volatile and non-volatile cues to generate appropriate behavioral responses. While much is known regarding the molecular diversity of neurons that comprise the MOS, less is known about the AOS. Here, focusing on the vomeronasal organ (VNO), the accessory olfactory bulb (AOB) and the medial amygdala (MeA) we reveal that populations of neurons in the AOS can be molecularly subdivided based on their ongoing or prior expression of the transcription factors Foxp2 or Dbx1 which delineate separate populations of GABAergic output neurons in the MeA. We show that a majority of AOB neurons that project directly to the MeA are of the Foxp2-lineage. Using single-neuron patch-clamp electrophysiology, we further reveal that in addition to sex-specific differences across lineage, the frequency of excitatory input to MeA Dbx1- and Foxp2-lineage neurons differs between sexes. Together, this work uncovers a novel molecular diversity of AOS neurons and lineage- and sex-differences in patterns of connectivity.

Graphical Abstract

The rodent accessory olfactory system (AOS) processes pheromonal cues that are detected by sensory neurons in the vomeronasal organ (VNO). The VNO projects to the accessory olfactory bulb (AOB), whose output neurons in turn project to the medial amygdala (MeA), a subcortical limbic structure that processes pheromonal information to generate innate behaviors through its downstream targets. We show that expression of the transcription factor genes Dbx1 (green cells) and Foxp2 (red cells), marks largely distinct subsets of neurons in the VNO, AOB and MeA, in varying proportions. Embryonic or postnatal expression of these transcription factors is observed specifically in subsets of vomeronasal sensory neurons (VSNs); glomerular layer (GlA) and mitral/tufted (M/T) neurons in the AOB; and projection neurons in the MeA, alongside neurons that express neither transcription factor (gray cells). Projections of Foxp2-lineage M/T neurons (red arrow) constitute a majority of the input from the AOB to the MeA. Foxp2-lineage and Dbx1-lineage neurons in the MeA have diverse neuropeptide profiles, with CARTPT, Tac2 and Npy being relatively enriched in the Dbx1-lineage while the reverse is true of Ucn3. We further uncover sex differences in the electrophysiological inputs to these MeA populations.

1. Introduction

1.1. Rodent Olfactory Circuitry

The olfactory system is divided into two functionally distinct components: the main olfactory system (MOS) and the accessory olfactory system (AOS). Odorants that activate the MOS bind to receptors expressed by olfactory sensory neurons (OSNs) in the main olfactory epithelium (MOE) in the nose, which projects directly to the main olfactory bulb (MOB) in the brain (Dulac & Wagner, 2006; Mombaerts et al., 1996; Vassar et al., 1994). From here information is sent to olfactory cortical areas for higher order processing (Hintiryan et al., 2012; Igarashi et al., 2012; Shipley & Adamek, 1984). In contrast, odorants that stimulate the AOS bind directly to receptors expressed by vomeronasal sensory neurons (VSNs) in the vomeronasal organ (VNO), located in the lower part of the nasal septum in proximity to the roof of the mouth. The AOS processes both volatile and non-volatile cues that are mainly dedicated for innate behaviors such as mating, territorial defense and predator avoidance (Papes et al., 2010). Non-volatile cues include pheromones, such as those released by anal and lacrimal glands and those present in urine (Cavaliere et al., 2020; Stowers & Liberles, 2016). These cues impart information regarding the hormonal state and sex of a conspecific (Ben-Shaul et al., 2010; Del Punta et al., 2002; Dulac & Wagner, 2006; Kimchi et al., 2007; Stowers et al., 2002; Thoß et al., 2019; Wysocki & Lepri, 1991). VNO olfactory sensory neurons project directly to the accessory olfactory bulb (AOB), which sits apart from the MOB in the posterior-dorsal aspect of the olfactory bulb (OB) (Wagner et al., 2006). The AOB projects directly to the medial (MeA) and cortical (CoA) nuclei of the amygdala and the bed nucleus of stria terminalis (BNST), all of which are interconnected and send robust projections to the hypothalamus (Davis et al., 1978; Dulac & Wagner, 2006; Dwyer et al., 2022; Gutiérrez-Castellanos et al., 2014; Lischinsky et al., 2023; Meurisse et al., 2009; Pardo-Bellver et al., 2012). The direct input from the MOB and AOB to higher-order processing centers in the brain is unique to the olfactory system as all other sensory modalities (touch, taste, vision, hearing) are first relayed through the thalamus. Thus, the main and accessory olfactory systems work in parallel and in tandem to allow an animal to rapidly interpret a complex olfactory world without thalamic processing (Hintiryan et al., 2012; Igarashi et al., 2012; Kang et al., 2011a; Kang et al., 2011b; Murakami et al., 2005; Pro-Sistiaga et al., 2007; Shepherd, 2005; Shipley & Adamek, 1984).

1.2. Olfactory System Neuronal Diversity

The diversity of olfactory cues that an animal senses in its environment is vast and complex. This is reflected in the large number of receptors; hundreds and thousands, that are present in the sensory neurons of the VNO and MOE, respectively (Buck & Axel, 1991; Dulac & Axel, 1995; Herrada & Dulac, 1997; Matsunami & Buck, 1997, Ryba & Tirindelli, 1997). The logic by which this complex sensory information is processed in the brain is currently much better understood in the MOS than the AOS. In the MOS, there appears to be regionalization of olfactory cue identification in the MOE and subsequent sorting in the MOB. The MOE is subdivided into anatomically and molecularly distinct zones which recognize different odorant classes (Hallem & Carlson, 2006; Nara et al., 2011; Ruiz Tejada Segura et al., 2022). This segregation is maintained in the MOB, which is also subdivided based on the nature of the cue (Burton et al., 2022; Fishilevich & Vosshall, 2005; Ma et al., 2012; Oka et al., 2006; Rubin & Katz, 1999; Sakano, 2010; Wachowiak & Cohen, 2001). Based on cell morphology, anatomic localization and physiological properties (Nagayama et al., 2014), the mitral and tufted (M/T) output neurons of the MOB are heterogeneous. More recent RNA-seq studies have uncovered a deeper level of diversity of MOB output neurons, identifying at least 8 distinct molecular subtypes (Zeppilli et al., 2021). This molecular coding in the MOB appears to predict neuronal subtype-specific patterns of inputs to higher-order neurons in the olfactory cortex, although this remains an area of intense investigation (Adam et al., 2014; Uchida et al., 2014).

Although less well-characterized than the MOS, neurons within the VNO and AOB can also be characterized by several criteria. The VNO is subdivided into anatomically and molecularly distinct apical and basal layers, which project to the anterior (aAOB), and the posterior (pAOB) AOB, respectively (Dulac, 2000; Dulac & Axel, 1995; Herrada & Dulac, 1997; Ichikawa et al., 1994; Imamura et al., 1985; Jia & Halpern, 1996; Knöll et al., 2003; Matsunami & Buck, 1997; Ryba & Tirindelli, 1997; Sugai et al., 1999). Interestingly, the aAOB and pAOB are also molecularly distinct from each other (Huilgol et al., 2013), and may regulate different types of innate behaviors (e.g. reproductive versus aggressive) (Kumar et al., 1999; Montani et al., 2013; Nunez-Parra et al., 2011). Similar to the MOB, M/T neurons comprise the output neuronal populations of the AOB. Based on morphological criteria, there appear to be at least 3 subtypes of AOB output neurons (Larriva-Sahd, 2008; Yonekura & Yokoi, 2008). However, the molecular diversity of both VNO and AOB output neurons remains unexplored.

In the MeA, a direct synaptic target of AOB output neurons, up to 20 different neuronal subtypes exist as characterized by their molecular and/or intrinsic electrophysiological properties (Bian, 2013; Carney et al., 2010; Chen et al., 2019; Keshavarzi et al., 2014; Lischinsky et al., 2017; Matos et al., 2020). In addition to possessing a variety of local interneuronal subtypes, the MeA comprises both excitatory and inhibitory output neurons (Bian et al., 2008; Choi et al., 2005; Wu et al., 2017). These MeA output neurons are subdivided based on their diverse morphological, electrophysiological and/or molecular properties (Keshavarzi et al., 2014; Lischinsky et al., 2017; Matos et al., 2020), which correlate with the innate behaviors regulated by these neurons (Lischinsky et al., 2023). One criterion which defines molecular diversity in the MeA is the differential expression of transcription factors (e.g. Otp, Foxp2 and Dbx1) in embryonic progenitors that give rise to different MeA projection neuronal populations (Lischinsky et al., 2017). These transcription factor-defined populations express lineage-specific patterns of select sex hormone-related proteins (e.g. Aromatase, estrogen and androgen receptors) (Lischinsky et al., 2017) and ion channels (e.gs. Kv7.1, Kir5.1, Kir2.1, Slo2.2) (Matos et al., 2020). This parcellation of MeA identity by molecular expression raises the question of whether other components of the interconnected AOS, such as the VNO and AOB, are also subdivided by the expression of the same molecular identifiers.

1.3. Sex differences in the AOS

One of the major functions of the AOS is to process information as to the reproductive state of conspecifics (Ben-Shaul et al., 2010). This information appears to be encoded differently in male and female brains. For example, cues from an estrus female elicit behavioral responses that are very different depending on whether the cue is being detected by a female or a male mouse (Yao et al., 2017). Although much remains to be understood, the first level of distinguishing between same or opposite sex cues is likely initiated in the VNO, where there are dedicated receptors for sex-specific cues (Isogai et al., 2011; Li & Dulac, 2018). Beyond this peripheral parsing of male/female cues, the processing of sex-specific information also occurs in higher-order brain regions, such as the MeA and BNST (Bergan et al., 2014; Li et al., 2017; Rigney et al., 2019). However, how this information is relayed to and processed in the MeA remains an open question.

The MeA displays extensive sex differences in cell morphology, dendritic complexity, cell size, intrinsic biophysical properties, and gene expression (Cooke et al., 2007; Cooke & Woolley, 2005; Hines et al., 1992; Matos et al., 2020). Using cFos staining and in vivo neuronal population recordings, our work (Lischinsky et al., 2017) and the work of others (Bergan et al., 2014; Li et al., 2017) revealed that the MeA also displays sex differences in neuronal population responses to olfactory cues. Patch-clamp single-neuron electrophysiological and tracing studies have also revealed sex differences in patterns of inputs to the MeA. For example, neurons in the male MeA have more excitatory input than females (Billing et al., 2020; Cooke & Woolley, 2005). However, the MeA neuronal subtype target of these inputs remains unknown. Here, in addition to revealing lineage diversity in the AOS, we took advantage of our ability to specifically tag two major populations of MeA GABAergic output neurons, Foxp2- and Dbx1-lineage neurons, to further address putative sex differences in synaptic inputs. Such information is necessary to piece together a neuronal subtype-level understanding of how male and female brains differentially process olfactory information for appropriate behavioral outputs.

1.4. Summary

The identification of neuronal diversity by transcription factors that are expressed during embryogenesis has provided a roadmap to unraveling the functionality of different neuronal subclasses across the nervous system. For example, this approach has been highly informative in elucidating the function of cortical interneurons (Batista-Brito et al., 2008; Mayer et al., 2018; Mi et al., 2018). Here, we used transcription factor expression, specifically ongoing or prior expression of Foxp2 or Dbx1 as an entry-point to parse neuronal diversity in the AOS, and further explore sex-specific patterns of input to the MeA, a main locus of convergence for olfactory cues with innate behavioral relevance. Our findings also provide a platform for future exploration of the specific behavioral roles played by transcription-factor identified AOS circuits.

2. Materials and Methods

2.1. Animals

Mice were housed on a 12h light/dark cycle and had ad libitum access to food and water. All mice used here were considered adult at > 2 months, with specific ages for each experiment indicated below. Mice used were: Foxp2^cre (JAX #030541), Dbx1^cre (Bielle et al., 2005), C57BL/6 (JAX #000664), LSL-FlpO (JAX #028584), and Rosa26YFP (JAX# 006148). Mice were genotyped using a commercial genotyping service (Transnetyx Inc., Cordova, TN). Foxp2^cre and Dbx1^cre mice were maintained as heterozygotes on the C57BL/6 background and crossed to Rosa26YFP mice as described in the Results Section.

2.2. Viruses and Stereotaxic Surgery

All experimental and surgical procedures were conducted in accordance with and approved by the Institutional Animal Care and Use Committee at Children’s National Hospital. For all surgeries, adult (2- to 5-month-old) animals were anaesthetized with isoflurane and placed into a stereotaxic apparatus (Stoelting Co. #51600). Body temperature was maintained with a heating pad during surgery and recovery, and 1.5–2% isoflurane was delivered continuously through a nose port. Animals were treated with analgesic buprenorphine (0.09 mg/kg body weight, of 0.03 mg/mL buprenorphine prepared in sterile saline) prior to surgery, and every 12 hours afterwards, as needed. Mice were monitored daily and sacrificed 2–5 weeks post viral injection.

Accessory Olfactory Bulb

Virus obtained from the University of North Carolina (UNC) Vector Core, mixed 1:1 (v/v) with blue fluorescent 1% solid polymer microspheres (Thermo Fisher Scientific, Cat#B0100), was injected unilaterally (400 nL, at 30 nL/min) into the accessory olfactory bulb (AOB) (AP: +4.1, ML: ±1.0, DV: −1.5 from Bregma) with a Hamilton syringe, into either Foxp2^cre (Virus: rAAV5-hSyn-Con/Foff.EYFP.WPRE, Titer: 2.6 × 10¹² GC/mL) or Dbx1^cre;FlpO (Virus: rAAV5-hSyn-Coff/Fon.EYFP.WPRE, Titer: 3.6 × 10¹² GC/mL) adult mice. Following virus delivery, the syringe was left in place for 3 min to prevent backflow and then slowly withdrawn.

Medial Amygdala

AAV2-retro virus (Virus: rAAV2-retro-Ef1a-DO_DIO-TdTomato_EGFP-WPRE-pA, Titer: 1.0 × 10¹³ GC/mL), obtained from Janelia Viral Tools Facility (now Addgene Plasmid #37120, RRID: Addgene_37120) was injected bilaterally (200 nL, at 15 nL/min) into the MeA (AP: −1.6, ML: ±2.2, DV: −4.8 from Bregma) of adult Foxp2^cre mice.

2.3. Immunohistochemistry

Adult animals were anaesthetized and transcardially perfused with 10 mL of 1X Phosphate Buffered Saline (PBS) followed by 10 mL of 4% paraformaldehyde (PFA) in 1X PBS. After perfusion, the brains were extracted and incubated overnight in the same fixative and cryoprotected in phosphate-buffered 30% sucrose solution for >48h at 4°C. After cryoprotection, brains were embedded in O.C.T Compound (Fisher HealthCare Cat No. 23–730-571). For experiments described in Figures 4 & 8, serial 40–60 μm coronal cryosections of the AOB and MeA (Figure 4) or MeA alone (Figure 8) were cut using a cryostat (CM3050S, Leica) and collected in PBS containing 0.02% (w/v) sodium azide. For immunofluorescence staining in the AOB (Figure 2), 30 μm-thick coronal cryosections were used, sampled every 120 μm. For experiments in Figures 2 & 4, the full antero-posterior range of the AOB (Bregma +3.56 mm to Bregma +2.96 mm) based on Franklin & Paxinos (2008), was sampled. For experiments in Figure 4 & 8, the full antero-posterior range of the MeA from Bregma −1.34 mm to Bregma −2.18 mm was sampled. Sections were excluded if found to be outside region boundaries or damaged in the region of interest. Sections were then incubated in blocking buffer (5% – 10% normal donkey serum (NDS), 0.5% Triton X-100, 1X PBS) for 1h at room temperature (RT) and subsequently incubated in the primary antibody mixture diluted in blocking buffer overnight at 4°C. Sections were then rinsed 3 × 10 min with PBST (1X PBS with 0.5% Triton X-100) and incubated in appropriate secondary antibodies for 1h at RT. Sections were then washed 3 × 10 min with PBST and 1 × 10 min with 1X PBS, and finally mounted with Fluoromount G^® containing DAPI (Thermo Fisher, Cat No.0100–20).

Figure 4: Retrograde tracing of AOB inputs to MeA neurons.

(a) Schematic illustrating the injection strategy of dual switch retrograde AAV transduction in the medial, cortical and posterior nuclei of the amygdala (MeA-CoA-PA) in Foxp2^cre mice. (b) Representative immunofluorescence image of a coronal section at the injection site shows strong viral expression in the posterior MeA (Bregma −1.94 mm). (c-e) Immunofluorescence images of the ipsilateral anterior AOB (coronal section, Bregma +3.56 mm) show Foxp2+ (recombined EGFP+, green, c) and Foxp2− (non-recombined tdTomato+, red, d) neurons. Merged image (e) reveals very minimal overlap of EGFP and tdTomato. (f-h) Immunofluorescence images of the ipsilateral middle AOB (coronal section, Bregma +3.20 mm), show similar recombination in the middle AOB. Crosses indicate section orientation (D – dorsal, M – medial, L – lateral). (i) Pie chart showing quantification of the percentage of Foxp2+ (green) and Foxp2- (red) ipsilateral AOB output neurons projecting to the MeA-CoA-PA. Abbreviations: CoA – Cortical Amygdala cp – cerebral peduncle, EplA – External plexiform layer of the AOB, MiA – Mitral cell layer of the AOB, MeA – Medial Amygdala and opt – optic tract. Scale bar equals 200 μm in b and 100 μm in all other panels. n = 7 mice (4 females and 3 males, 10–17 sections/mouse).

Figure 8: MeA Foxp2-lineage neurons receive excitatory and inhibitory inputs, with no sex differences.

(a) Posterior MeA Foxp2-lineage neurons (EYFP+, yellow), identified by immunofluorescence in coronal sections from Foxp2^cre;RYFP mice and co-immunostained for the excitatory pre- and postsynaptic markers, VGLUT2 (magenta) and PSD95 (cyan), respectively. (b-e) Higher magnification images of boxed region in a, with EYFP (b, yellow), VGLUT2 (c, magenta) and PSD95 (d, cyan) channels shown separately. Merged image (e) shows Foxp2-lineage neurons expressing PSD95 and surrounded by VGLUT2 puncta (solid white arrowheads in b-e), indicating putative excitatory synapses. (f) Box-and-whisker plots quantifying Foxp2-lineage neurons positive for both VGLUT2 and PSD95; plotted as a percentage of total Foxp2-lineage (EYFP+) neurons in the posterior MeA, comparing females versus males. (g) Posterior MeA Foxp2-lineage neurons (EYFP+, yellow), co-immunostained for the inhibitory pre- and postsynaptic markers, VGAT (magenta) and Gephyrin (cyan), respectively. (h-k) Higher magnification images of boxed region in g, with EYFP (h, yellow), VGAT (i, magenta) and Gephyrin (j, cyan) channels shown separately. Merged image (k) shows Foxp2-lineage neurons expressing Gephyrin and surrounded by VGAT puncta (solid white arrowheads in h-k), indicating putative inhibitory synapses. (l) Box- and-whisker plots quantifying Foxp2-lineage neurons positive for both VGAT and Gephyrin; plotted as a percentage of total Foxp2-lineage (EYFP+) neurons in the posterior MeA, comparing females versus males. Crosses in a, g indicate section orientation (D – dorsal, M – medial, L – lateral). Abbreviations: cp – cerebral peduncle, itc – intercalated nucleus of the amygdala, MeA – Medial Amygdala and opt – optic tract. Scale bars in a, g equal 250 μm. In all other panels, scale bars equal 20 μm. n.s. indicates not significant (p > 0.05, Mann-Whitney U test). n = 5 mice per sex (3 sections/mouse, bilateral counts).

Figure 2: Dbx1-lineage and Foxp2+ neurons in the AOB.

(a, h, p) Low magnification images of unilateral, coronal sections from the anterior (a), middle (h) and posterior (p) olfactory bulb (OB) from a Dbx1^cre;RYFP mouse. Foxp2+ (red) and EYFP+ (green, Dbx1-lineage) neurons identified by immunofluorescence and co-labeled with DAPI (blue). White boxes in a, h, p highlight the accessory olfactory bulb (AOB), shown at higher magnification in b-e, i-l and q-t. (b-e) Higher magnification images of boxed region in a, show Foxp2+ neurons (red, c) and EYFP+ Dbx1-lineage neurons (green, d), in both the external plexiform (EplA) and mitral cell (MiA) layers of the anterior AOB. DAPI (blue) shown in e. Many Foxp2+ neurons (red) are present throughout the AOB, with fewer Dbx1-lineage neurons (green, EYFP+). Merged image (b) shows minimal overlap of the two populations. (f, g) Box-and-whisker plots quantifying number, per unilateral section of Foxp2+ (red circles, left Y-axis), EYFP+ (green squares, right Y-axis) and co-expressing (double positive, orange diamonds, right Y-axis) neurons in the EplA (f) and MiA (g). n = 19 unilateral sections from 5 adult mice (4 males, 1 female). (i-l) Higher magnification images of the middle AOB (boxed region in h). (i) Glomerular layer (GlA), EplA and MiA of the AOB are shown. (m-o) Quantification of neuron number in each layer of the middle AOB. n = 41 unilateral sections from 5 adult mice (4 males, 1 female). (q-t) Higher magnification images of the posterior AOB (boxed region in p). GlA and MiA are much smaller and the EplA is absent at the posterior level of the AOB. Quantification of each layer in the posterior AOB layer is shown in u, v. n = 10 unilateral sections from 5 adult mice (4 males, 1 female). Solid white arrowheads in c, d, j, k, r, highlight examples of quantified neurons. In a, h and p, crosses indicate section orientation (D – dorsal, M – medial, L – lateral). Abbreviations: EplA – External plexiform layer of the AOB, GlA – Glomerular layer of the AOB, MiA – Mitral cell layer of the AOB. Scale bars in a, h, p equal 200 μm, in b-t: 100 μm. *** p < 0.001, **** p < 0.0001 (Mann-Whitney U tests).

All primary and secondary antibodies were obtained commercially (Table 1), and validated by the manufacturer or by prior studies. Immunogen information is available on the RRID portal and manufacturers’ webpages. We optimized the dilutions for our assays. For primary antibodies, we further verified their staining patterns by comparing expression patterns to published literature and/or expected expression patterns from the Allen Brain Atlas (Lein et al, 2006), when available. Secondary antibodies were also routinely tested by us for specificity using negative controls without primary antibodies.

Table 1:

List of antibodies used

Antibody	Host species & Dilution range used	Vendor & Catalog No.	RRID	Immunogen	Validation
Anti-Foxp2	Rabbit, 1:500–1:1000	Atlas Antibodi es, HPA000 382	AB_1078908	Recombinant Protein Epitope Signature Tag (PrEST) antigen sequence (from human FOXP2) Human antigen and its mouse ortholog (ENSMUSG0 0000029563) have 99% sequence identity	By manufacturer: Affinity purified using PrEST antigen as affinity ligand Reacts with a band of approximately 100 kDa in Western Blot using lysates of Human Cell Line RH-30. In previous literature: Human Protein Atlas: https://www.proteinatlas.org/ENSG00000128573-FOXP2 (Enard et al., 2009; Kast et al., 2019; Kondabolu et al., 2023; Reimers-Kipping et al., 2011) By authors: Positive controls in wildtype adult mouse brains - staining pattern compared to known expression pattern of Foxp2 in other brain regions, reported previously in the literature (Co et al., 2020; Druart et al., 2020; Kast et al., 2019; Lischinsky et al., 2017, 2023).
Anti-Foxp2	Goat, 1:200	Santa Cruz Biotechn ology, sc-21069	AB_2107124	Human FOXP2 N-terminus	In previous literature: (French et al., 2007; Lischinsky et al., 2017; Reimers-Kipping et al., 2011) French et al., (2007) demonstrate antibody reactivity with a band between 100 kDa and 70 kDa in Western blot using striatal precursor lysates from E16.5 mouse embryos and further validate the antibody using lysates from a conditional Foxp2 knockout that shows no band. Reimers-Kipping et al., (2011) compare results between this and the above (HPA000382, Atlas) antibody using IHC - fluorescence in adult and embryonic mouse brain.
Anti-GFP (to detect EGFP and EYFP)	Rat, 1:500–1:1000	Nacalai-Tesque 04404–84	AB_10013361	His-GFP (full length) fusion protein	In previous literature: (Lischinsky et al., 2023; Liu et al., 2021; Nakamura et al., 2021) By authors: Positive controls - staining pattern compared to known expression pattern of Foxp2 and Dbx1- derived neurons in the wildtype adult mouse brain and vomeronasal organ, reported previously in the literature (Causeret et al., 2023; Lischinsky et al., 2017, 2023). Internal negative controls using the uninjected hemisphere in viral transduction experiments with retro-AAV2 dual switch reporter virus show no signal or background, in contrast to strong signal on the injected side, confirming specificity.
Anti-CART	Rabbit, 1:500	Phoenix Pharmace utical H003–62	AB_2313614	Not available	In previous literature: (Fenwick et al., 2006; Kirouac et al., 2006; Lee et al., 2020)
Anti-PDE4A	Rabbit, 1:500	Abcam, ab14607	AB_301375	C-terminal region of Human PDE4A As per manufacturer: Labels all known PDE4A variants (A1, A5, A8, Ax and testisspecific A).	By manufacturer: Immunogen affinity purified In previous literature: (Katrancha et al., 2019; Munch et al., 2018; Yang et al., 2018)
Anti-OMP	Goat, 1:5000	Wako, 54410001	AB_2315007	Rodent OMP	In previous literature: (Henkel et al., 2017; Nguyen & Imamura, 2019; Treloar et al., 1999)
Anti-PSD95	Rabbit, 1:50	Proteinte ch, 20665–1-AP	AB_2687961	Peptide	By manufacturer: Antigen affinity purified Western Blot analysis using HeLa cell lysates shows reactivity with a single band of size between 70 kDa and 100 kDa that is absent when using HeLa cell lysates pretransfected with shRNA directed against PSD95. Positive controls for immunohistochemistr y using paraffin embedded mouse and rat brain slices demonstrate strong, specific staining in the dentate gyrus with protein localized near cell surface as expected from known function at synapses. In previous literature: (Chen et al., 2016; X. Li et al., 2021; Wang et al., 2021)
Anti- VGLUT2	Guinea Pig, 1:100	Millipore , AB2251- I	AB_2665454	KLH-conjugated linear peptide corresponding to 18 amino acids from the C-terminal, cytoplasmic domain of rat Vesicular Glutamate Transporter 2 (VGluT2)	By manufacturer: Unpurified. Representative lot data shows Western blot reactivity with a band of approximate size 56 kDa in rat brain membrane extract, smaller than calculated size of 64.6 kDa consistent with reported apparent molecular weight. Representative lot data shows punctate, cytoplasmic staining in mouse and rat formalin fixed paraffin embedded brain tissue section. In previous literature: (Li et al., 2020; Madeira et al., 2020; Modol et al., 2020) Li et al. (2020) use shRNA approach to knockdown VGLUT2 and demonstrate consequent lack of immunoreactivity.
Anti-Gephyrin	Rabbit, 1:200	Thermo Fisher, PA5– 55157	AB_2641960	Recombinant protein corresponding to human Gephyrin, amino acids 291 to 430 100% sequence identity with mouse ortholog	By manufacturer: Affinity purified using antigen Western Blot analysis of PC-3 cell lysates demonstrates reactivity with a single band of approximately 93 kDa which is nearly absent in lysates of PC-3 cells transfected with shRNA directed against GPHN. Immunohistochemistr y shows significant staining in human cerebral cortex and weak or minimal staining in human lymph node concordant with RNA-Seq TPM values in the two tissues. In previous literature: Not available
Anti-VGAT	Mouse, 1:1000	Invitroge n, MA5– 24643	AB_2637258	Recombinant protein corresponding to human VGAT amino acids 13–103 95% sequence identity to mouse otholog	By manufacturer: Immunohistochemistr y in mouse and rat brain slices demonstrate VGAT immunoreactivity in brain regions known to have GABAergic neurons e.g. substantia nigra pars reticulata, cerebellum and globus pallidus. In previous literature: (Kaczmarek-Hajek et al., 2018; Sammoura et al., 2023; Torz et al., 2022)
Anti-tdTomato	Goat, 1:500	LSBio, LS-C340696	AB_2819022	Purified recombinant peptide (Discosoma tdTomato) produced in E. coli	By manufactuer: Immunoaffinity purified Western Blot analysis of HEK293 cell lysates shows reactivity with a single 55 kDa band Does not cross-react with GFP In previous literature: (Beier et al., 2021; Co et al., 2020; Zhang et al., 2021) By authors: Internal negative controls using the uninjected hemisphere in viral transduction experiments with retro-AAV2 dual switch reporter virus show no signal or background, in contrast to strong signal on the injected side, confirming specificity.
Anti-Rabbit secondary (Cy3-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 711–165–152	AB_2307443	Rabbit IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Rabbit secondary (Cy5-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 711–175–152	AB_2340607	Rabbit IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Rat secondary (Alexa Fluor 488 conjugated)	Donkey, 1:5001:1000	Thermo Fisher Scientific, A-21208	AB_2535794	Rat IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Mouse secondary (Cy3-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 711–165–150	AB_2340813	Mouse IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Mouse secondary (Cy5-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 715–175–150	AB2340819	Mouse IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Goat secondary (Cy3-conjugated)	Donkey, 1:1000–1:1500	Jackson Immuno Research, 705–165–147	AB_2307351	Goat IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Goat secondary (Cy5-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 705–175–147	AB_2340415	Goat IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Guinea Pig secondary (Cy3-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 706–165–148	AB_2340460	Guinea Pig IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.
Anti-Guinea Pig secondary (Alexa Fluor 647-conjugated)	Donkey, 1:1000	Jackson Immuno Research, 706–605–148	AB_2340476	Guinea Pig IgG (H+L)	Negative control by authors with identical immunohistochemistr y conditions but no primary antibody, shows no discernible background staining.

2.4. Multiplexed fluorescent in situ hybridization

8-week old male Dbx1^cre;RYFP mice were transcardially perfused with 10 mL ice-cold 1X PBS, followed by 10 mL freshly-prepared ice-cold 4% paraformaldehyde in 1X PBS. Whole brains were collected, incubated in 4% PFA/1X PBS at 4°C for 24h, followed by cryoprotection in 30% sucrose in 1X PBS for 48h at 4°C. Brains were then embedded in O.C.T. compound and stored at −80°C until cryosectioning. 10 μm slide-mounted cryosections of the medial amygdala from Bregma −1.58 mm to Bregma −1.94 mm (as identified in Franklin & Paxinos, 2008) were obtained using a Thermo Scientific HM525 NX cryostat at −22°C. Slides were stored at −80°C with dessicant until staining. RNAscope^™ HiPlex-12 Mouse kit (ACDBio, Cat #324106) was used to detect the following Mus musculus mRNA targets, available from ACDBio: Tac2 (#446391), Tacr1 (#428781), Trh (#436811), Foxp2 (#428791), Avp (#401391), Ucn3 (#464861), Npy1r (#427021), Htr2c (#401001), Npy (#313321), Ecel1 (#475331) and EYFP (#312131). Protocol for HiPlex-12 sample preparation, pretreatment and staining was followed exactly as per ACDBio’s user manual for fixed, frozen tissue sections.

2.5. Microscopy

To analyze labeled neurons in the AOB and output projections to the MeA (Figures 3 & 4), 10X (N.A. 0.40) and 20X (N.A. 0.75) images were acquired using an Olympus BX63 epifluorescence microscope or a Nikon A1 confocal microscope (0.7 – 3.0 μm step size, 1.2 AU pinhole, 20X/N.A. 0.75 or 10X/N.A. 0.45). All off-target regions were excluded from analysis. For the synaptic labeling experiment in Figure 8, images were acquired on an Olympus FV1000 or a Nikon A1 confocal microscope. For immunofluorescence analysis in the AOB (Figure 2), images were acquired as Z-stacks (2 μm step size, 1.2 AU pinhole, 20X/N.A. 0.75 or 10X/N.A. 0.45 objectives) on a Nikon A1 confocal microscope. For RNAScope in situ hybridization, all samples were imaged on a Leica DMi8 THUNDER deconvolution microscope using 10X (N.A. 0.32) and 63X oil (N.A. 1.40) objectives. 395, 470, 550 and 640 nm excitation lines and corresponding emission filter sets (440, 510, 590 and 700 nm respectively) were used. 20–200 ms camera exposure time was used for each channel, varied depending on observed signal-to-noise in that channel and maintained uniformly across samples. THUNDER large-volume adaptive deconvolution with computational clearing was performed on acquired widefield images. Deconvolution settings: 1.47 refractive index, Good’s Roughness method of regularization with parameter 0.05, medium optimization, number of iterations and cutoff gray value set to ‘Auto’, feature scale of 535 nm (minimum possible) for all channels except DAPI, feature length of 2673 nm (software default) for DAPI, 98% strength for all channels. Single-Z plane tile-scan images of the MeA were acquired with automatic linked shading correction, deconvolved, stitched (statistical blend) and exported as TIFFs for further image processing and quantification.

Figure 3: AOB Foxp2-lineage neurons project to the MeA.

(a) Schematic of anterograde viral transduction into the AOB of Foxp2^cre mice. Boxed regions within images (b, e) from The Mouse Brain in Stereotaxic Coordinates (Franklin & Paxinos, 2008) denote the AOB and MeA shown in (c, d) and (f, g), respectively. (c, d) Representative images from unilateral coronal sections of virally transduced AOBs show recombined EYFP+ neurons (green) from two different mice (c, d) in the AOB (middle level, Bregma +3.08 mm). (f, g) EYFP+ green fibers (white arrows) emanating from transduced AOB mitral cells (MiA) are observed along the input pathway to the posterior MeA (Bregma −1.46 mm) – ventral tissue boundary of the posterior MeA. Crosses in c, d indicate section orientation (D – dorsal, M – medial, L – lateral). Abbreviations: AOB – Accessory Olfactory Bulb, Hyp – hypothalamus, ic – internal capsule, MeA – Medial Amygdala, MiA – Mitral cell layer of the AOB, opt – optic tract. In all panels, scale bar equals 100 μm and DAPI-labelled nuclei are shown in blue. n = 6 mice (3 females and 3 males).

2.6. Image Processing, Quantification and Statistical Analyses

All image processing of immunohistochemistry data, including linear brightness/contrast adjustment, was performed using Fiji (Schindelin et al., 2012). For data presented in Figure 2, unilateral Foxp2+ neuronal counts on confocal Z-stacks were performed semi-automated, in Imaris 10.0.0 using Spots Model (Estimated Diameter – 3.8 μm), manually refined by quality threshold and machine learning using visual inspection, and finally filtered by “Distance to Surface” option to obtain counts within each AOB layer (which are first defined as Imaris Surface objects using boundaries demarcated in Franklin & Paxinos, 2008). Accuracy of semi-automated counting was verified by comparison to manual counting of 25% of the dataset in Fiji. Unilateral Dbx1 (EYFP) and co-expression counts were performed manually. For Foxp2 and EYFP, a neuron was counted only if it also had a visible DAPI+ nucleus. Data were first sorted according to antero-posterior position by visual inspection and comparison to Franklin & Paxinos (2008), and then quantified by AOB layer (absolute neuron counts), also following the layer nomenclature used by Franklin & Paxinos (2008) and previously published studies (Cádiz-Moretti et al., 2014; Gascuel et al., 2012; Gentier et al., 2015; Kuteeva et al., 2004; Stanic et al., 2006). The glomerular layer (GlA) was identified by its distinctive clustering of cell bodies in middle and posterior AOB sections.

The external plexiform layer (EplA) was identified in middle AOB sections as a distinct region (band) of sparseness in cell bodies immediately ventral to the GlA and dorsal to the mitral cell layer (MiA). The MiA in middle AOB sections was demarcated as the AOB layer immediately dorsal to the dorsal lateral olfactory tract (dlo); and has a visibly denser cell body distribution and is thicker than the EplA. In anterior AOB sections, the GlA is entirely absent, the MiA was first demarcated as the crescent-shaped region dorsal to the dlo, with a thick, dense distribution of cell bodies and the EplA as the remainder of the AOB. In posterior AOB sections, the EplA and MiA are visually indistinguishable and were consequently considered together as a single sublayer (the MiA). All counts were analyzed in GraphPad Prism 9 using independent Mann-Whitney U tests, α = 0.05 with number of unilateral sections as sample size and without correction for multiple comparisons. Percentages reported in Figures 1 & 4, were calculated by averaging individual percentages obtained from each datapoint (individual mouse) within an experimental group; whereas the absolute numbers reported in the text in parentheses are summations of individual cell counts obtained from all datapoints (all mice) within an experimental group. For in situ hybridization data, linear brightness/contrast adjustments were made for all channels in Fiji similarly across tissue sections, ROIs for the MeA were drawn based on Franklin & Paxinos (2008) using 10X tile-scanned DAPI image of whole section; and counting labels were added to Foxp2 mRNA+ and EYFP mRNA+ neurons (DAPI+ nuclei with at least 6 mRNA puncta on or clustered tightly around nucleus) in the MeA using the Cell Counter plugin. These were aligned individually to images of each investigated mRNA target using DAPI channel for transparency-based manual image registration in Adobe Photoshop. MeA neurons expressing each target (6 or more mRNA puncta on or tightly clustered around nucleus) were quantified within Foxp2 mRNA+ and EYFP mRNA+ neurons and statistically analyzed (Mann-Whitney U tests, α = 0.05, followed by FDR correction for multiple comparisons using Benjamini-Hochberg procedure with 10% FDR) and plotted using GraphPad Prism 9.

Figure 1: Dbx1-lineage chemosensory neurons in the VNO.

(a) Triple immunofluorescence for EYFP (green), PDE4A (red) and DAPI (blue) in a unilateral coronal section of the VNO from a Dbx1^cre;RYFP mouse. b-d and f-h correspond to the image shown in a. (b-d) Recombined EYFP+ neurons (green) are shown within the total DAPI+ (blue) population across the sensory epithelium (SE). Solid white arrowheads in d highlight EYFP+ neurons. (e) Quantification of EYFP+ neuron number as a percentage of the total DAPI+ population. (f-h) EYFP+ neurons are observed in both the PDE4A+ apical layer (A) and the PDE4A− basal layer (B), demarcated by the white lines in g & h. Solid white arrowheads in h highlight EYFP+ PDE4A+ neurons (yellow) in the apical VNO. Empty white triangles highlight EYFP+ PDE4A− neurons (green) in the basal VNO. (i) Quantification of the proportions of PDE4A+ (apical) and PDE4A− (basal) neurons within the EYFP+ (Dbx1-lineage) population. (j-l) Double immunofluorescence for EYFP (green, j) and OMP (red, k), a marker of mature olfactory neurons. Solid white arrowheads in l highlight EYFP+ OMP+ (yellow) mature Dbx1-lineage VNO neurons and empty white triangles highlight EYFP+ OMP− (green) immature Dbx1-lineage neurons in the neurogenic niche. (m) Quantification of mature EYFP+ neuron number as a percentage of the total EYFP+ population. n = 18 mice (9 males and 9 females, 4–6 unilateral sections/mouse) in e, i, m. Abbreviations: BV – Blood Vessel, L – Lumen, NN – Neurogenic Niche, NSE – Non-Sensory Epithelium, SE – Sensory Epithelium. Scale bars equal 100 μm.

2.7. Patch-clamp electrophysiology recordings

Sexually naive, adult mice (P50–P90) were deeply anaesthetized with isoflurane and sacrificed. Brains were removed and immediately immersed in an ice-cold carbogenated (95% O₂ and 5% CO₂) sectioning solution (75 mM sucrose, 10 mM D-glucose, 25 mM NaHCO₃, 87 mM NaCl, 2.5 mM KCl, 1.0 mM NaH₂PO₄, 1.0 mM MgCl₂, and 0.5 mM CaCl₂; pH 7.3; 295–300 mOsm/kg. 300 μm coronal slices were sectioned on a vibratome (Leica VT1200S) at the level of posterior MeA (Bregma −1.56 to −1.94 mm; Franklin and Paxinos, 1997). Slices were collected and placed in oxygen-equilibrated artificial CSF (ACSF) composed of the following: 125.0 mM NaCl, 3.5 mM KCl, 1.0 mM MgCl₂, 1.25 mM NaH₂PO₄, 2.0 mM CaCl₂, 26.0 mM NaHCO₃, and 10.0 mM D-glucose; pH 7.3; 295–300 mOsm/kg. Dbx1^cre;RYFP-positive or Foxp2^cre;RYFP-positive neurons were visualized using a Nikon FN1 epifluorescence microscope with a 450 to 490 nm filter. Whole-cell patch-clamp recordings from EYFP-positive fluorescent neurons were performed at RT with continuous perfusion of carbogenated ACSF. Signals were acquired on a patch-clamp amplifier (Multiclamp 200B) and digitized at 250 kHz with an A/D converter (DigiDATA1550B). Recordings were performed with glass electrodes pulled on a Sutter P-2000 pipette puller (Sutter Instruments), with 3–5 MΩ resistance and filled with a potassium gluconate-based intracellular solution containing the following: 119.0 mM K-gluconate, 2.0 mM Na-gluconate, 6.0 mM NaCl, 2.0 mM MgCl₂, 10.0 mM HEPES, 0.9 mM EGTA, 4.0 mM Mg-ATP, 14.0 mM Tris-creatine PO₄, and 0.3 mM Tris-GTP; pH 7.3; 285–295 mOsm/kg. Neurons were first recorded in current clamp configuration as reported in Matos et al. (2020). Subsequently, they were switched from current clamp to voltage clamp configuration. Each seal was tested to have maintained access resistance at < 30 MΩ throughout the current and voltage clamp recordings, verified at the beginning and end of a recording session. Neurons that did not meet these criteria were excluded from all experiments. Voltage clamp recordings were done for 3 minutes at −60 mV gap free holding. Only the 3^rd minute of recording for each stage was analyzed. sEPSC events were detected using Clampfit Software 10.7 (Molecular Devices) by employing “template search” – several events were initially selected and saved as a template and then used to detect similar events from the entire recording. Statistical analysis and plotting of event frequency data were performed in GraphPad Prism 9. Data were determined to have a lognormal distribution using D’Agostino-Pearson’s test, and then log transformed for subsequent analyses with 2-way ANOVA using sex and lineage as independent variables. Homoscedasticity and normality of residuals of log transformed data were verified by Spearman’s test and D’Agostino-Pearson omnibus (K2) tests respectively. ANOVA interaction and main effects were inspected (α=0.05) using F test, followed by Sidak’s test for multiple pairwise planned comparisons with α=0.05.

3. Results

3.1. Foxp2+ and Dbx1-lineage neurons in the VNO and AOB

Our previous studies revealed that Foxp2+ and Dbx1-lineage neurons comprise non-overlapping populations of inhibitory output neurons in the MeA (Lischinsky et al., 2017, 2023). In the MeA, Foxp2 is first expressed during embryogenesis and remains on through adulthood. In contrast, Dbx1 is expressed only during embryogenesis in mitotic forebrain ventricular zone (VZ) progenitors and is turned off when they transition to the subventricular zone (SVZ) (Hirata et al., 2009). To mark Dbx1-lineage neurons we used an anti-GFP antibody on tissue from previously validated (Bielle et al., 2005) Dbx1^cre mice crossed to RYFP reporter mice (Figures 1, 2, 5, 6 & 7). For gene expression studies (Figures 2, 5 & 6), where we assessed both Dbx1-lineage and Foxp2+ neurons, Foxp2+ neurons were identified using a well-characterized antibody or by RNAscope^™ in situ hybridization. For viral cre-based connectivity tracing, electrophysiology and synaptic experiments (Figures 3, 4, 7 & 8), we used previously validated Foxp2^cre mice (Rousso et al., 2016) crossed to RYFP reporter mice or injected with an EYFP- and/or tdTomato-carrying reporter virus.

Figure 5: Distribution of Dbx1-lineage and Foxp2+ neurons in the MeA in relation to CARTPT expression.

(a) Image from The Mouse Brain in Stereotaxic Coordinates (Franklin & Paxinos, 2008) showing the posterior MeA (red box) in coronal view at the same antero-posterior level as the sections shown in panels (b-h). (b) Brightfield immunohistochemistry image showing CARTPT expression localized in the posterodorsal MeA. (c-e) Immunofluorescence image showing Dbx1-lineage neurons (EYFP+, green, c) and CARTPT+ cell bodies and fibers (red, d) in the MeA of a Dbx1^cre;RYFP mouse. CARTPT expression matches the distribution of Dbx1-lineage neurons which are observed embedded within a zone of CARTPT immunoreactivity (e). (f-h) Complementary distribution of Foxp2+ neurons (green, f), to the expression pattern of CARTPT (red, g), with Foxp2+ neurons predominantly observed outside the zone of CARTPT expression (h). Cross in b indicates section orientation (D – dorsal, M – medial, L – lateral), same followed in c-h. Abbreviations: cp – cerebral peduncle, Hyp – hypothalamus, MeA – Medial Amygdala and opt – optic tract. All scale bars equal 200 μm.

Figure 6: Fluorescent in situ hybridization for neuropeptides and receptors in Foxp2+ and Dbx1-lineage neurons in the MeA.

(a-a”) Expression and quantification of Tac2 mRNA (cyan)-expressing neurons within Foxp2 mRNA+ (yellow, a) and Dbx1-lineage (EYFP mRNA+, magenta, a’) neurons, as shown in coronal sections from the posterodorsal MeA. (a”) Box-and-whisker plots show percentage of co-expression of Tac2 mRNA within each population. (b-h”) Expression and quantification of the following other mRNAs: Ucn3 (b-b”), Npy (c-c”), Tacr1 (d-d”), Avp (e-e”), Ecel1 (f-f”), Htr2c (g-g”) and Trh (h-h”). Solid white arrowheads indicate neurons double positive for Foxp2 or EYFP mRNA, and the candidate gene mRNA; empty white triangles indicate neurons positive for Foxp2 or EYFP mRNA but not the candidate gene mRNA. Scale bars equal 25 μm for all images. * FDR-adjusted p value < 0.05 after multiple Mann-Whitney U-tests. All other comparisons not significant (FDR-adjusted p > 0.05). n = 4 male mice (2–4 sections/mouse, bilateral counts).

Figure 7: Sex and lineage differences in the frequency of spontaneous excitatory postsynaptic currents (sEPSCs).

(a) Representative epifluorescence and (b) DIC images show a representative recombined EYFP+ neuron (white arrows) in the posterior MeA targeted for ex vivo patch electrophysiology. (c) Representative current traces in whole-cell voltage clamp mode (−60 mV holding potential) from Dbx1-lineage (top) or Foxp2-lineage (bottom) neurons with traces from females and males shown separately. Orange circles indicate counted sEPSC events. (d) Quantification of frequency (events per second plotted as mean ± s.e.m.) of sEPSCs in Dbx1-lineage (left) or Foxp2-lineage (right) neurons in ACSF, from female (circles, white bar) or male (triangles, black bar) mice. * p < 0.05, ** p < 0.01, **** p < 0.0001, other pairwise comparisons not significant, α = 0.05 (2-way ANOVA & Sidak’s posthoc test). n = 10–13 neurons from 5–7 mice/group. Scale bars equal 50 μm in a, b.

We first determined if Foxp2+ or Dbx1-lineage neurons are present in the VNO (Figure 1) and AOB (Figure 2) of 1- to 3-month-old mice. The adult VNO contains mature vomeronasal neurons (VSNs) and immature newly generated VSNs in a neurogenic niche located at the edge of the VNO (Katreddi & Forni, 2021). While no Foxp2+ neurons were found in the adult VNO (data not shown), we observed numerous Dbx1-lineage neurons, comprising ~7% of the total DAPI+ population (5589 EYFP+ neurons out of 78382 DAPI+ cells) across the VNO sensory epithelium (Figure 1a–e). The adult VNO comprises a lumen-facing apical layer and a non-lumen-facing basal layer. These layers project to the anterior and posterior AOB, respectively (Jia & Halpern, 1996; Knöll et al., 2003; Sugai et al., 1999). The apical layer is marked by PDE4A, a member of the cAMP-specific family of phosphodiesterases (Lau & Cherry, 2000). We found Dbx1-lineage VSN cell bodies almost evenly distributed across the PDE4A+ (~43%, 2496 PDE4A+ neurons out of 5589 EYFP+ neurons) apical layer and the PDE4A- (~57%, 3093 PDE4A-neurons out of 5589 EYFP+ neurons) basal layer (Figure 1a, f–i). Co-labeling of EYFP+ neurons with OMP, a marker for mature VNO neurons (Farbman & Margolis, 1980) revealed that ~42% of Dbx1-lineage neurons are mature (2174 OMP+ neurons out of 5589 EYFP+ neurons, Figure 1j–m). Thus, in the VNO, Dbx1-lineage neurons comprise both immature and mature VSNs located in both the apical and basal layers.

In the olfactory bulb (OB), we next examined if Foxp2+ and Dbx1-lineage neurons were present in the MOB and the AOB. Co-immunostaining for Foxp2 and EYFP in Dbx1^cre;RYFP mice revealed neurons from both populations across the OB (Figure 2a,h,p), with an apparent greater number of Foxp2+ neurons. Focusing on the AOB, which receives direct input from the VNO, we observed Foxp2+ and Dbx1-lineage neurons located within the glomerular (GlA), external plexiform (EplA) and mitral cell (MiA) layers (Figure 2b–g, i–o, q–v). Across all layers and the antero-posterior extent of the AOB, we found that Foxp2+ neurons were a greater population than Dbx1-lineage neurons (Figure 2f–g, m–o, u–v). We further found that in all layers, the majority of Foxp2+ and Dbx1-lineage neurons were separate populations with minimal overlap in expression (Figure 2f–g, m–o, u–v), a population segregation that mimics what we previously observed in the MeA (Lischinsky et al., 2017).

3.2. Foxp2+ neurons comprise the majority of MeA-projecting AOB output neurons.

The MeA receives direct projections from the AOB (Zheng et al., 2020). We therefore wanted to examine whether Foxp2- and Dbx1-lineage neurons in the AOB project to the MeA. To accomplish this, we injected an anterograde AAV5-hSyn-Con/Foff.EYFP.WPRE virus into the AOB of Foxp2^cre mice (Figure 3a) and an anterograde AAV5-hSyn-Coff/Fon.EYFP.WPRE virus into Dbx1^cre;FlpO mice. The use of Dbx1^cre;FlpO mice to trace Dbx1-lineage neurons is necessitated by the fact that cre is no longer expressed when Dbx1 expression ceases during later embryogenesis (Bielle et al., 2005; Hirata et al., 2009; Medina et al., 2005). Constitutive flp expression after cre-driven recombination in Dbx1^cre;FlpO mice enables tracing using Flp-dependent reporters. Viral transduction of AAV5-hSyn-Coff/Fon.EYFP.WPRE in Dbx1^cre;FlpO mice resulted in recombination very few Dbx1-lineage neurons, thus precluding our ability to follow their projections (data not shown). This low number of recombined AOB neurons is likely a reflection of the lower number of Dbx1-lineage M/T neurons compared to the number of Foxp2+ neurons (Figure 2). In contrast, in injected Foxp2^cre mice, we observed large numbers of recombined AOB neurons (Figure 3b–d), with robust EYFP+ projections emanating from the AOB to the MeA (Figure 3e–g).

As we observed strong projections from Foxp2-lineage AOB M/T neurons to the MeA, we next wanted to determine if this population represented most of the projections to the MeA. To accomplish this, we targeted the MeA of Foxp2^cre mice with a retrograde dual-switch reporter ‘Retro-AAV2’ virus (rAAV2-retro-Ef1a-DO_DIO-TdTomato_EGFP-WPRE-pA) (Saunders et al., 2012; Tervo et al., 2016) (Figure 4a–b). In addition to the MeA, we also typically observed a low level of recombination in the adjacent cortical (CoA) and posterior (PA) nuclei of the amygdala due to viral spread. Retro-AAV2 is taken up by presynaptic neuronal terminals and translocated retrogradely to cell bodies where presynaptic cre-expressing cell bodies are EGFP+/tdTomato− while cre-negative cell bodies are EGFP−/tdTomato+ (Figure 4b). In the AOB, we found both EGFP+ and tdTomato+ neurons (Figure 4c–h) indicating that both Foxp2+ and Foxp2− AOB M/T neurons project to the amygdala. Quantification of the numbers of EGFP+ and tdTomato+ neurons revealed that the EGFP+ population represented the majority of recombined neurons (2324 EGFP+ neurons out of 3426 EGFP+ or tdTomato+ neurons, Figure 4i). We observed a few neurons expressing both EGFP and tdTomato, likely representing a low level of cre-independent background recombination reported by Tervo et al. (2016). Overall, these results indicated that in the AOB, a majority of amygdala-projecting neurons are of the Foxp2-lineage.

This raises an intriguing question about the relative contribution of Dbx1-lineage versus non-Dbx1-lineage neurons to the output projections from the AOB to the MeA. Retrograde tracing of the Dbx1-lineage from the MeA in adults is however currently precluded by the lack of a Flp-dependent version of the dual switch retrograde AAV tracer virus.

3.3. MeA Foxp2+ and Dbx1-lineage neurons express different cohorts of neuropeptides

Our previous studies revealed that MeA Foxp2+ and Dbx1-lineage neurons express different sex hormone-related proteins and ion channels (Lischinsky et al., 2017; Matos et al., 2020). Furthermore, a recent study has revealed that these lineages control different innate behaviors (Lischinsky et al., 2023). The MeA also expresses a variety of neuropeptides and receptors that likely play a neuromodulatory role in regulating innate behaviors such as mating, aggression, feeding, maternal care and social interaction, based on their known roles in other limbic nuclei. To explore whether the two populations of our interest express different combinations of neuropeptides, we conducted immunohistochemistry (Figure 5) and multiplex RNAscope^™ in situ hybridization (Figure 6) in sections from Dbx1^cre;RYFP mice. Candidates were chosen based on the following criteria: The gene of interest: 1) is expressed in the MeA as shown either in prior published studies or the Allen Brain gene expression atlas (Lein et al., 2006), 2) plays a known role in MeA function or innate social behavior, and/or 3) was observed in a previous RNA-seq screen of the adult MeA (Chen et al., 2019). Following these criteria, we generated a list of 10 top candidate genes (Table 2). Of the probes examined, we observed the expression of CARTPT (Cocaine- And Amphetamine-Regulated Transcript Protein), a neuropeptide implicated in feeding, reward and stress (Carpenter et al., 2020; Funayama et al., 2022; Kristensen et al., 1998; Lee et al., 2022), to be the most lineage segregated. We found high expression of CARTPT in the MeA (Figure 5a) and in a pattern strikingly resembling the distribution of Dbx1-lineage neurons (Figure 5b–e), and complementary to the distribution of Foxp2+ neurons (Figure 5f–h). Dual immunofluorescence revealed that the majority of CARTPT+ cell bodies and projections were embedded within regions of Dbx1-lineage neurons (Figure 5e). In contrast, CARTPT+ cell bodies and projections did not co-localize with Foxp2+ neurons (Figure 5h). We next assessed the expression of the other 9 candidates by multiplexed RNAscope^™ in situ hybridization in the MeA (Figure 6). Of these, we confirmed the expression of all candidates except Npy1r. Of the markers expressed in the MeA, most were expressed only in a small subset of either Dbx1-lineage or Foxp2+ neurons. Of these markers, we found three Tac2, Ucn3 and Npy were expressed significantly more in neurons of one lineage versus the other. We found both Tac2 and Npy mRNA were modestly but significantly enriched in Dbx1-lineage neurons relative to Foxp2 mRNA+ neurons (Figure 6a–a”, c–c”). In contrast, we found Ucn3 mRNA which encodes the Urocortin 3 peptide, enriched in Foxp2 mRNA+ neurons in the MeA (Figure 6b–b”).

Table 2:

List of neuropeptide and receptor candidate genes expressed in the MeA and previously implicated in social or other behaviors.

Gene Symbol	Gene name	Behaviors implicated in	References
Tac2	Tachykinin 2	Aggression, Freezing (fear)	(Asahina et al., 2014; Zelikowsky et al., 2018)
Ucn3	Urocortin 3	Aggression, Social Novelty Preference	(Autry et al., 2021; Shemesh et al., 2016)
Npy	Neuropeptide Y	Aggression, Social Interaction/Social Anxiety, Maternal behavior	(Karl et al., 2004; Muroi & Ishii, 2015; Sajdyk et al., 1999, 2002)
Tacr1 (Nk1r)	Tachykinin receptor 1 (Neurokinin 1 receptor)	Aggression, Mating	(Berger et al., 2012; Halasz et al., 2009)
Avp	Arginine vasopressin	Aggression, Social Interaction, Pair Bonding	(Donaldson et al., 2010; Gutzler et al., 2010; Koolhaas et al., 1990; Terranova et al., 2017; Whylings et al., 2020)
Ecel1	Endothelin converting enzyme-like 1	Aggression, Social Behavior	(Delprato et al., 2018; Xu et al., 2012)
Htr2c	Serotonin (5-HT) receptor 2C	Social Interaction/Social Anxiety, Social Novelty, Freezing (fear), Aggression	(Martin et al., 2012; Sejourne et al., 2015)
Trh	Thyrotropin releasing hormone	Social Interaction, Aggression	(Crowley & Hydinger, 1976; Kwon et al., 2021; Puciklowski et al., 1988)
Npy1r	Neuropeptide Y receptor Y1	Aggression, Social Interaction/Social Anxiety	(Padilla et al., 2016; Sajdyk et al., 1999)
Cartpt	Cocaine-and amphetamine-regulated transcript protein prepropeptide	Feeding, Cocaine/Reward-seeking, Stress resiliency, Social dominance and Aggression	(Carpenter et al., 2020; Funayama et al., 2022; Kristensen et al., 1998; Lee et al., 2022)

3.4. Sex differences in inhibitory and excitatory input to MeA Foxp2- and Dbx1-lineages

In addition to molecular differences described above and by Lischinsky et al. (2017) between MeA Foxp2- and Dbx1-lineage neurons, our previous studies revealed lineage differences in intrinsic biophysical properties (Matos et al., 2020). Prior studies from others revealed male/female differences in total inputs to the MeA, with males displaying more excitatory input (Billing et al., 2020; Cooke & Woolley, 2005). However, the MeA neuronal subtype targets of these inputs remains unknown. Therefore, we next conducted patch-clamp electrophysiology and measured spontaneous postsynaptic currents (sPSCs), a measure of synaptic activity, in EYFP+ MeA neurons in Foxp2^cre;RYFP and Dbx1^cre;RYFP mice of both sexes (Figure 7a–b). Gap-free recordings in artificial cerebrospinal fluid (ACSF) at a holding potential of − 60 mV largely measured spontaneous AMPA/kainate receptor-mediated currents (sEPSCs, Figure 7c). We found a significantly higher frequency of total sEPSCs (Figure 7d) in male Dbx1-lineage neurons compared to male Foxp2-lineage neurons. We additionally observed that Foxp2-lineage neurons in females displayed a significantly higher event frequency than those in males, with the reverse true of Dbx1-lineage neurons (Figure 7d). Further, Foxp2-lineage neurons receive lower excitatory input than Dbx1-lineage neurons in males (Figure 7d), a finding consistent with our previous observations (Lischinsky et al., 2017).

As we observed sex differences in sPSC recordings within MeA Foxp2-lineage neurons (Figure 7d), we next wanted to assess whether there were corresponding differences in the expression of either excitatory or inhibitory synaptic markers on MeA Foxp2-lineage neurons. To accomplish this, we assessed the expression of the excitatory pre- and postsynaptic markers, PSD95 and VGLUT2, and the inhibitory pre- and postsynaptic markers Gephyrin and VGAT, in Foxp2^cre;RYFP mice by immunohistochemistry (Figure 8). We quantified both the proportion of Foxp2-lineage (EYFP+) neurons receiving excitatory and inhibitory inputs as well as the number of colocalized pre- and postsynaptic excitatory or inhibitory puncta on EYFP+ neurons. We established a threshold of 5 or more puncta on an EYFP+ neuron as positive for the corresponding marker. We found that a subset of Foxp2-lineage neurons in both males and females received putative direct excitatory (Figure 8a–e) or inhibitory (Figure 8g–k) input. However, we found no differences in the proportion of neurons receiving excitatory (Figure 8f) or inhibitory (Figure 8l) input in males compared to females, nor in the number of puncta on each EYFP+ neuron (data not shown).

4. Discussion

4.1. Summary of findings

Using a combination of gene expression analyses and viral circuit mapping approaches, we revealed the molecular diversity and patterns of connectivity of neurons within core structures of the accessory olfactory system (AOS), the VNO, AOB and MeA. We find that two transcription factor-expressing populations, Foxp2+ neurons and neurons derived from the Dbx1 lineage define molecularly distinct neuronal populations across brain nuclei that comprise the AOS. We further find that Foxp2+ neurons in the AOB comprise the overwhelming majority of outputs to the MeA. Interestingly, we further find sex differences in the electrophysiologically recorded frequency of inputs to MeA neurons. Our findings suggest that subpopulations of neurons identified by ongoing or prior expression of select transcription factors may define distinct subcircuits within the AOS, and uncover a sexual dimorphism in their input connectivity.

4.2. Lineage diversity across an interconnected circuit

The AOS is dedicated to processing innate behaviors such as mating, aggression and predator avoidance. These behaviors are considered ‘hardwired’; meaning that they manifest without prior behavioral training. Although shaped by hormonal influences, the patterns of wiring of these circuits are likely in large part pre-determined by developmental genetic programs. However, these genetic programs remain unknown. Our previous studies, and the studies of others linking MeA embryonic development to neuronal diversity (Aerts & Seuntjens, 2021; Carney et al., 2010; García-Moreno et al., 2010; Lischinsky et al., 2017) revealed that embryonic expression of the transcription factors Otp, Foxp2 and Dbx1 define separate populations of neural progenitors that later give rise to non-overlapping populations of either excitatory (Otp) or inhibitory (Foxp2, Dbx1) MeA output neurons. Across the developing nervous system, expression of discrete subclasses of transcription factors in neural progenitors directs the emergence of later neuronal subtype identity (Aydin et al., 2019; Heavner et al., 2020; Hörmann et al., 2020; Sagner et al., 2021). The early endowment of MeA neuronal diversity as defined by transcription factor expression (Otp, Foxp2, Dbx1), suggests a potential molecular code for how the MeA is assembled. Here, our gene expression analysis revealed that this molecular coding may be at least partially conserved in the VNO and AOB which lie one and two nodes upstream of the MeA, respectively. We found that in addition to the MeA, Dbx1-lineage neurons mark subsets of sensory neurons in the VNO and output neurons in the AOB. In the VNO, Dbx1-lineage neurons comprise ~7% of the entire sensory neuron population across both apical and basal layers. A recent study revealed that Dbx1 is expressed as early as E12.5 in the developing VNO (Causeret et al., 2023). Thus, similar to other regions of the nervous system, Dbx1 also marks early developing progenitors in the VNO. As the VNO is the first site of sensory processing in the AOS, it is interesting to speculate that Dbx1-lineage sensory neurons may express specific subclasses of olfactory receptors implicated in select innate behaviors. If this is the case, it also raises the intriguing possibility that neurons of the same transcription factor identity/lineage (in this case Dbx1), directly connect with each other to form a transcription factor labeled line for the processing of select olfactory cues. However, the case for a potential labeled-line circuit is not as strong for neurons marked by Foxp2. In contrast to the Dbx1-lineage, Foxp2 marks a much larger population of AOB neurons, comprising more than 2/3^rds of the output neurons projecting directly to the MeA. The lower number of Dbx1-lineage neurons in the AOB precluded our ability to trace Dbx1-lineage AOB projections to their final destinations using anterograde viral tracing. However, it would be reasonable to assume that AOB Dbx1-lineage neurons in part comprise the MeA-projecting Foxp2-negative population. Regardless, combined with our previous findings (Lischinsky et al., 2017), our current findings reveal that within and outside the MeA, expression of the same transcription factors define regions of a known accessory olfactory circuit.

4.3. MeA neuropeptide expression

Our previous studies revealed that Foxp2- and Dbx1-lineage neurons in the MeA express cohorts of sex hormone-related proteins and ion channels in a lineage-specific manner. Aromatase and ER-α, and the action potential regulating ion channels Kir5.1, Kir6.1, KChip4.1, Cav1.2 and Kv7.1, are all expressed in a greater proportion of Dbx1-lineage neurons, whereas the ion channel Kir2.1 is expressed in a greater proportion in the Foxp2-lineage (Lischinsky et al., 2017; Matos et al., 2020). Here, we extended this prior knowledge by assessing expression of neuropeptides known to have a function in MeA-regulated behaviors such as feeding, aggression, and mating. We show that expression of CARTPT, Tac2 and Npy are enriched in the Dbxl-lineage while Ucn3 is expressed in more Foxp2 mRNA+ neurons than Dbx1-lineage neurons. Of these, the most interesting pattern was that of CARTPT, whose expression pattern strikingly mimics the distribution of Dbx1-lineage neurons. As the CARTPT antibody marks both cell bodies and CARTPT+ fibers, it was difficult to discern if Dbx1-lineage neurons are producing CARTPT or receiving dense CARTPT+ input. Regardless, the strong expression overlap implicates MeA Dbx1-lineage neurons in aspects of feeding, homeostasis and/or reward. In addition, our multiplexed RNAscope^™ in situ hybridization analyses revealed a higher expression of Tac2 and Npy in Dbx1-lineage neurons, implicating MeA Dbx1-lineage neurons in aggression, social novelty or sexual arousal. Tac2 encodes a neuropeptide, tachykinin isoform 2, that promotes aggressive behavior in fruit flies and mice (Asahina et al., 2014; Zelikowsky et al., 2018). Npy encodes Neuropeptide Y which, in rodents, modulates aggression through Y1 receptors in the medial amygdala (Karl et al., 2004), is implicated in reduced social anxiety (Sajdyk et al., 1999, 2002) and regulates maternal behavior, a critical sex-specific social behavior (Muroi & Ishii, 2015). In contrast, of the 10 neuropeptides explored, we found only Ucn3 to be enriched in the Foxp2 mRNA+ population. Urocortin 3 has been shown to promote preference for social novelty through its action in the MeA (Shemesh et al., 2016), infant-directed aggression through its function in the perifornical area of the hypothalamus (Autry et al., 2021) and feeding (Stengel & Taché, 2014). This suggests a role for Foxp2+ neurons in these behaviors and is consistent with the role of the Foxp2 gene and Foxp2+ neurons in aggression (Herrero et al., 2021; Lischinsky et al., 2017, 2023). Aside from CARTPT, it is important to note that although lineage restricted, Tac2, Npy and Ucn3 are only expressed in a small subset of neurons. This, however, does not preclude a putative important lineage-specific role in behavior for the following reasons: First, our expression analysis was conducted using tissue from home-cage animals. As levels of neuropeptide expression are typically state-dependent, it is likely we are observing only a baseline of expression, which may change over time. Second, our prior studies of cFos activation patterns in the MeA revealed that even during behavioral tasks which elicit a strong behavioral response, cFos is expressed only in a subset of neurons within each lineage (Lischinsky et al., 2017). This indicates that perhaps only a handful of neurons within a given population needs to be active to be engaged in a given behavior. Third, these (and other) neuropeptides may act in concert in overlapping or different populations to modulate behavior. While these lineage-specific expression patterns of neuropeptides importantly extend our knowledge of molecular diversity of MeA neurons, it remains to be determined how different behaviors are regulated by each lineage, as there are many mechanisms beyond neuropeptide expression that influence behavior. These include, for examples, patterns and types of input/output connectivity and intrinsic neuronal excitability and other biophysical parameters. Future transcriptomic analysis of Foxp2- and Dbx1-lineage neurons will be highly informative in providing a fuller picture of the molecular diversity of these populations.

4.4. Sex differences in connectivity

The MeA has been long recognized as a highly sexually dimorphic brain region, with known differences in structural properties such as cell morphology, dendritic complexity, and cell size (Cooke et al., 2007; Cooke & Woolley, 2005; Hines et al., 1992). Our prior patch-clamp electrophysiology studies further revealed sex differences in intrinsic biophysical properties, including action potential firing dynamics in both Foxp2- and Dbx1-lineages (Matos et al., 2020). Moreover, recent gene expression and single-cell RNA-seq transcriptomic studies revealed sex differences at the molecular level in the MeA (Chen et al., 2019). Interestingly, the sex differences in gene expression in the MeA were most prominent in inhibitory GABAergic neurons as opposed to the excitatory glutamatergic population.

While our previous work revealed that Dbx1-lineage neurons receive greater excitatory input than Foxp2-lineage neurons (Lischinsky et al., 2017), we did not explore whether there were sex differences in inputs to these two GABAergic output populations. To examine this, here we took both an electrophysiological approach by measuring the frequency of spontaneous inputs and an immunohistochemical approach to assess putative synaptic connectivity using well characterized markers of excitatory and inhibitory synapses. Via patch-clamp analyses, we uncovered sex differences in total inputs to Foxp2-lineage neurons and Dbx1-lineage neurons. We found that Foxp2-lineage neurons in females have more spontaneous inputs than in males, while the reverse is true of Dbx1-lineage neurons. There are several possible mechanisms that can account for these observations. First, the number of excitatory and/or inhibitory synaptic inputs may be sexually dimorphic. However, our immunohistochemical analysis of synaptic markers in Foxp2-lineage neurons suggests that this may not be the case. Although we cannot rule out that ultrastructural analysis of synapses at a higher resolution would uncover differences, our data suggest other mechanisms may be occurring. These include sex differences in presynaptic firing rate and/or number of presynaptic neurotransmitter release events or postsynaptic membrane excitability via differences in ion channel expression. Further clues to the underlying mechanism maybe inferred from recent MeA RNA-seq studies (Chen et al., 2019), which revealed major transcriptomic sex differences in MeA GABAergic neurons in genes implicated in synaptic function and communication. Although we do not know whether these differences come from within the Foxp2- and/or Dbxl-lineages, our electrophysiological findings provide an entry point to link sex differences in gene expression with sex differences in synaptic input that we observe in MeA Dbx1- and Foxp2-lineage neurons.

Interestingly, prior electrophysiological studies revealed that the MeA receives more excitatory input in males than in females (Cooke & Woolley, 2005). A more recent anatomical study showed that MeA Aromatase+ neurons in males receive more inputs from the AOB than in females (Dwyer et al., 2022). While in potential contrast to our findings of greater synaptic input to Foxp2-lineage neurons in females, it is likely that the male/female pattern of inputs to the MeA varies from population to population; with some neuronal subtypes receiving more (or stronger) inputs in males and others in females. In addition to the Foxp2-and Dbx1-lineage neurons, the MeA is populated by a vast array of interneurons, excitatory (Otp+) output neurons (Chen et al., 2019; Lischinsky et al., 2017) and likely other inhibitory and/or excitatory output neurons not of the Foxp2-, Dbx1-, or Otp-lineages. The MeA receives strong input from not only the AOB, but also from the posterior amygdala, cortical amygdala, BNST as well as lesser input from other brain regions (Cádiz-Moretti et al., 2014; Lischinsky et al., 2023). Thus, there are likely sex differences based not only on which MeA population is innervated, but also by source of input. Regardless of the underlying mechanism, our study introduces an additional layer of refinement in the analysis of physiological inputs to the MeA via developmental transcription factor-defined subpopulations.

Although we found robust differences across sex in synaptic input as uncovered by patch-clamp electrophysiology, it is important to note that in our analysis we did not segregate females based on estrus cycle. Prior work in the MeA (Dalpian et al., 2019) and the hypothalamus (Dias et al., 2021; Yin et al., 2022), revealed estrus state-dependent changes in the strength of neuronal connectivity. Moreover, hormones greatly shape brain development at several levels. In addition to the estrus state-dependent short-term hormonal cycling which transiently affects circuitry, developmental hormonal surges also have an impact on how these circuits are initially wired together (Simerly, 2003). As the MeA plays a central role in regulating sex-specific innate behaviors such as aggression and mating, there may also be non-hormonally driven genetic programs that are involved in the establishment of male and female differences in wiring patterns during early postnatal development. Exploration of both these intrinsic and extrinsic influences on induction and maintenance of sexually dimorphic patterns of MeA connectivity and ultimately behavior, will continue to be an interesting area of investigation.

Data Availability Statement

The data that support the findings of this study are available from the corresponding authors upon reasonable request.