Molecular Ontology of the Nucleus of Solitary Tract

ABSTRACT

The nucleus of the solitary tract (NTS) receives visceral information and regulates appetitive, digestive, and cardiorespiratory systems. Within the NTS, diverse processes operate in parallel to sustain life, but our understanding of their cellular composition is incomplete. Here, we integrate histologic and transcriptomic analysis to identify and compare molecular features that distinguish neurons in this brain region. Most glutamatergic neurons in the NTS and area postrema co‐express the transcription factors Lmx1b and Phox2b, except for a ventral band of neurons in the far‐caudal NTS, which include the Gcg‐expressing neurons that produce glucagon‐like peptide 1 (GLP‐1). GABAergic interneurons intermingle through the Lmx1b+Phox2b macropopulation, and dense clusters of GABAergic neurons surround the NTS. The Lmx1b+Phox2b macropopulation includes subpopulations with distinct distributions expressing Grp, Hsd11b2, Npff, Pdyn, Pou3f1, Sctr, Th, and other markers. These findings highlight Lmx1b–Phox2b co‐expression as a common feature of glutamatergic neurons in the NTS and improve our understanding of the organization and distribution of neurons in this critical brain region.

Using histologic and transcriptomic analysis, we found that co‐expression of the transcription factors Lmx1b and Phox2b identifies glutamatergic neurons in the nucleus of the solitary tract (NTS), except for a caudal‐ventral band of neurons that include Gcg‐expressing neurons, which produce glucagon‐like peptide 1 (GLP‐1). The Lmx1b+Phox2b macropopulation includes subpopulations with distinct distributions expressing Grp, Hsd11b2, Npff, Pdyn, Pou3f1, Sctr, Th, and other markers. This information improves our understanding of the organization and distribution of neurons in this vital brain region.

1. Introduction

The first site in the brain that receives visceral information is the nucleus of the solitary tract, commonly referred to by the abbreviation of its original Latin name, nucleus tractus solitarii (NTS). This interoceptive region of the dorsomedial medulla surrounds the solitary tract, which is a distinctive bundle of myelinated axons delivering viscerosensory inputs to diverse neurons in and around the NTS (Cutsforth‐Gregory and Benarroch ²⁰¹⁷). NTS neurons integrate these peripheral inputs with top–down input from a variety of other brain regions (Gasparini et al. 2020; Holt, Pomeranz, et al. ²⁰¹⁹; van der Kooy et al. ¹⁹⁸⁴; Whitehead, Bergula, and Holliday ²⁰⁰⁰) and send output to forebrain and brainstem regions that regulate affective, appetitive, cardiorespiratory, and neuroendocrine functions (Beckstead, Morse, and Norgren ¹⁹⁸⁰; Geerling and Loewy ²⁰⁰⁶; Herbert, Moga, and Saper ¹⁹⁹⁰; Holt ²⁰²²; Loewy and Burton ¹⁹⁷⁸; Quillet et al. ²⁰²³; Rinaman ²⁰¹⁰; Ter Horst et al. ¹⁹⁸⁹).

The general organization of NTS neurons has been studied in a variety of species, including cats (Loewy and Burton 1978), humans (McRitchie and Tork ¹⁹⁹³), hamsters (Whitehead 1988), rats (Altschuler et al. 1989; Herbert, Moga, and Saper ¹⁹⁹⁰; Kalia and Sullivan ¹⁹⁸²), and mice (Ganchrow et al. 2014). Previous tracing studies identified a mosaic pattern of inputs, with separate visceral structures providing input to overlapping NTS subregions (Altschuler et al. 1989; Bassi et al. ²⁰²²; Contreras, Beckstead, and Norgren ¹⁹⁸²; Hamilton and Norgren ¹⁹⁸⁴; Kalia and Sullivan ¹⁹⁸²). Separate visceral stimuli activate neurons in separate NTS subregions (Ran et al. 2022), and intermingled clusters of NTS neurons express diverse molecular markers (Ludwig et al. 2021; Riche, De Pommery, and Menetrey ¹⁹⁹⁰; Zhang et al. ²⁰²¹), but our ability to distinguish which subpopulations control which specific functions remains limited.

One limiting factor is a lack of knowledge about the overall spatial and molecular organization of NTS neurons. Neurons with separate connections and functions can be distinguished by differences in gene expression, and using molecular‐genetic features to distinguish neurons is a useful strategy in both humans and experimental animals (Siletti et al. 2023; Yao et al. ²⁰²³; Zeisel et al. ²⁰¹⁸). For example, clarifying the molecular ontology of the parabrachial nucleus, which receives output from the NTS, revealed that intermingled populations of neurons have different connectivity patterns and functions (Karthik et al. 2022; Pauli et al. ²⁰²²). Clarifying the molecular ontology of the NTS could lay the foundation for linking specific neurons to their specific functions.

Transcription factors control gene expression, and the transcriptome of each neuron influences the pattern of synaptic connections it establishes (Hirsch et al. 2021). This property makes transcription factors particularly useful for identifying and distinguishing neurons in the hindbrain (Gray 2013; Karthik et al. ²⁰²²). Transcription factors with expression patterns that overlap the NTS include Lmx1b, which is important for the development of certain monoaminergic and glutamatergic neurons in the brainstem (Dai et al. 2008), and Phox2b, which is necessary for the development of specific catecholaminergic and glutamatergic populations, including the NTS (Dauger et al. 2003; Gray ²⁰¹³; Kang et al. ²⁰⁰⁷). In this study, we combined histological analysis with publicly available transcriptomic data to investigate the spatial distributions of these two transcription factors and a variety of other molecular markers that identify and distinguish neuronal subpopulations in and around the NTS.

2. Materials and Methods

2.1. Mice

All mice were group‐housed in a temperature‐ and humidity‐controlled room on a 12/12 h light/dark cycle and with ad libitum access to water and standard rodent chow. Overall, we used n = 40 (31 males and 9 females) mice ranging in age from 8 to 16 weeks (20–40 g body weight). Included among these were knockin‐Cre and Cre‐reporter strains detailed in Table 1. These mice were hemizygous and maintained on a C57BL6/J background. All experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee at the University of Iowa.

TABLE 1.

Cre‐driver and ‐reporter mice.

Strain	References	Source information	Key gene
R26‐LSL‐L10GFP Reporter	Krashes, M. J., B. P. Shah, J. C. Madara, et al. 2014. “An Excitatory Paraventricular Nucleus to AgRP Neuron Circuit That Drives Hunger.” Nature 507, no. 7491: 238.	Available from originating investigators http://www.informatics.jax.org/allele/MGI:5559562	Floxed transcription STOP cassette followed by EGFP/Rpl10 fusion reporter gene under the control of the CAG promoter targeted to the Gt(ROSA)26Sor locus
Vgat‐IRES‐Cre (Slc32a1‐IRES‐Cre)	Vong, L., C. Ye, Z. Yang, B. Choi, S. Chua, B. B. Lowell. 2011. “Leptin Action on GABAergic Neurons Prevents Obesity and Reduces Inhibitory Tone to POMC Neurons.” Neuron 71, no. 1: 142–154.	Jax 028892 https://www.jax.org/strain/028862	IRES‐Cre inserted after the Slc32a1 stop codon
Vglut2‐IRES‐Cre (Slc17a6‐IRES‐Cre)	Vong, L., C. Ye, Z. Yang, B. Choi, S. Chua, B. B. Lowell. 2011. “Leptin Action on GABAergic Neurons Prevents Obesity and Reduces Inhibitory Tone to POMC Neurons.” Neuron 71, no. 1: 142–154.	Jax 016963 https://www.jax.org/strain/016963	IRES‐Cre inserted downstream of the stop codon of Slc17a6 on chromosome 7

2.2. Perfusion and Tissue Sections

Mice were anesthetized with a mixture of ketamine–xylazine (i.p. 150‐15 mg/kg, dissolved in sterile 0.9% saline), then perfused transcardially with phosphate‐buffered saline (PBS), followed by 10% formalin–PBS (SF100‐20, Fisher Scientific). After perfusion, brains were removed and fixed overnight in 10% formalin–PBS. We sectioned each brain into 40‐µm‐thick coronal slices using a freezing microtome and collected tissue sections into separate, one‐in‐three series. Sections were stored in cryoprotectant solution at −20°C until further processing.

2.3. Immunofluorescence Labeling

For immunofluorescence labeling of mouse brain tissue, we removed from cryoprotectant 6–9 evenly spaced axial tissue sections from the caudal medulla for n = 3 cases for each marker combination. We rinsed all sections in PBS, then loaded them into a solution containing one or more primary antisera (Table 2). These antisera were added to a PBS solution of 0.25% Triton X‐100 (BP151‐500, Fisher), 2% normal donkey serum (NDS, 017‐000‐121, Jackson ImmunoResearch), and 0.05% sodium azide (14314, Alfa Aesar) as a preservative (PBT–NDS–azide). We incubated sections overnight at room temperature on a tissue shaker. The following morning, the sections were washed 3× in PBS and incubated for 2 h at room temperature in PBT–NDS–azide solution containing species‐specific donkey secondary antibodies. These secondary antibodies were conjugated to Cy3, Cy5, Alexa Fluor 488, Brilliant Violet, or biotin (Jackson ImmunoResearch #s 703‐545‐155, 705‐545‐147, 705‐165‐147, 706‐165‐148, 706‐065‐148, 713‐065‐147, 711‐165‐152, 711‐545‐152, 711‐675‐152, 711‐175‐152, 715‐545‐003, 715‐165‐150, 715‐065‐151, 715‐545‐150; each diluted 1:1000 or 1:500). If a biotinylated secondary antibody was used, sections were again washed 3× in PBS, then incubated for an additional 2 h in streptavidin‐Cy5 #SA1011; Invitrogen). Tissue sections were then washed 3× in PBS, mounted on glass slides (#2575‐plus; Brain Research Laboratories), dried, and then coverslipped using Vectashield (Vector Labs). All slides were stored in slide folders at 4°C until imaging. For immunofluorescence labeling of the nuclear transcription factor Pou3f1 and the cytoplasmic enzyme neuronal nitric oxide synthase (nNOS), available antisera both were rabbit primaries. Thus, we incubated tissue in Pou3f1 antibody overnight, followed by 3× washes in PBS then Cy3‐donkey‐anti‐rabbit for 2 h on the second day. Then, we washed the tissue in PBS and incubated it overnight in the nNOS primary antibody overnight, followed by 3× washes in PBS then Alexa Fluor 488‐donkey‐anti‐rabbit secondary antibody for 2 h on the third day. This protocol allowed us to use both primaries made in the same host, labeling one protein in the nucleus and the other in the cytoplasm.

TABLE 2.

Antisera used in this study.

Antigen	Immunogen description	Source, host species, RRID	Concentration
Choline acetyltransferase (ChAT)	Human placental choline acetyltransferase	Millipore, goat polyclonal, #AB144P; RRID: AB_2079751	1:1000
Green fluorescent protein (GFP)	Full length green fluorescent protein from the jellyfish Aequorea Victoria	Thermo Fisher Scientific, Chicken, #A10262; RRID: AB_2534023	1:3000
11‐beta‐hydroxysteroid dehydrogenase type 2 (HSD2)	Fusion protein Ag5146 (amino acids 1–405 encoded by human HSD11B2; GenBank BC036780)	Protein Tech, rabbit polyclonal, cat# 14192‐1‐AP; RRID: AB_2119643	1:2000
11‐beta‐hydroxysteroid dehydrogenase type 2 (HSD2)	Amino acids 261–405 from the C‐terminus of human 11β‐HSD2	Santa Cruz, rabbit polyclonal, cat# sc_20176; RRID: AB_2233199	1:700
Lmx1b	Full‐length LIM homeobox transcription factor 1 beta protein from mouse	C. Birchmeier, Max Delbruck Center for Molecular Medicine, Berlin; guinea pig polyclonal, RRID: AB_2314752	1:8000
Neuronal nitric oxide synthase (nNOS)	A recombinant protein consisting of 195 amino acids from the N‐terminus of rat nNOS protein	Invitrogen, Rb polyclonal, RRID:AB_2313734	1:1000
Paired‐like homeobox 2b (Phox2b)	Mouse monoclonal antibody raised against amino acids 11–70 mapping near the N‐terminus Phox2b of human origin	Santa Cruz mouse, monoclonal, # sc‐376997, RRID: AB_2813765	1:1000
Paired‐like homeobox 2b (Phox2b)	BSA‐coupled 15mer corresponding to the C terminus of the Phox2b protein with an added N‐terminal tyrosine	H. Enomoto, School of Medicine at Kobe University, Japan; RRID: AB_2895590	1:12,000
Oct6 (Pou3f1)	Synthetic peptide within human Oct6	Abcam, rabbit monoclonal: #ab126746; RRID: AB_11130256	1:2000
Tyrosine hydroxylase (TH)	Native tyrosine hydroxylase from rat pheochromocytoma in sheep host	Millipore, sheep polyclonal, cat#: Ab1542; RRID: AB_90755	1:2000

2.4. Immunohistochemistry

For brightfield labeling of mouse brain tissue, we removed a full‐brain series of tissue sections from cryoprotectant for n = 3 cases, rinsed them in PBS, then incubated them in 0.3% hydrogen peroxide (#H325‐100, Fisher), in PBT for 30 min to quench endogenous peroxidase activity. After three washes in PBS, we loaded sections into PBT–NDS–azide containing one primary antiserum overnight at room temperature on a tissue shaker. After 3× PBS washes the following morning, we incubated sections for 2 h in a 1:500 solution of biotinylated specie‐specific donkey antibody (#711‐065‐152; Jackson) in PBT–NDS–azide. Sections were washed three more times and placed for 1 h in biotin–avidin complex (Vectastain ABC kit PK‐6100; Vector), washed 3× in PBS, and incubated in nickel–diaminobenzidine (NiDAB) solution for 10 min. Our stock DAB solution was prepared by adding 100 tablets (#D‐4418, Sigma, Saint Louis, MO) into 200 mL ddH₂O, then filtering it. We then used 1 mL of this DAB stock solution, with 300 µL of 8% nickel chloride (#N54‐500, Fisher Chemical) per 6.5 mL PBS. After 10 min in NiDAB, we added hydrogen peroxide (0.8 µL of 30% H₂O₂ per 1 mL PBS‐DAB) and swirled sections for 2–5 min until observing black (nickel–DAB) color change. After two rapid PBS washes, we wet‐mounted one or more sections, checked them in a light microscope to ensure optimal staining, and in rare cases replaced sections for up to one additional minute for additional enzymatic staining. Finally, after washing an additional 3× in PBS, we ordered sections and mounted them on glass slides. Slides were air‐dried then dehydrated in an ascending series of alcohols and xylenes, then coverslipped with Cytoseal 60 (#8310‐16 Thermo Scientific).

2.5. Fluorescence In Situ Hybridization

To label mRNA, we used a variety of different RNAscope probes, detailed in Table 3, along with the RNAscope Fluorescent Multiplex Reagent Kit (#320851; Advanced Cell Diagnostics) or RNAscope Fluorescent Multiplex Detection Kit v2 (#323100; Advanced Cell Diagnostics). To label tissue using the original Fluorescent Multiplex Reagent Kit, we first mounted a one‐in‐three series of brainstem tissue slices containing the NTS on glass slides, then left slides to dry overnight. Sections were outlined using a Super‐HI PAP pen (Research Products Incorporated) to form a hydrophobic barrier, then we washed with PBS 2 × 2 min at room temperature. Then, we incubated the slices with Protease IV in glass petri dishes floating in a 40°C water bath for 30 min and washed with PBS 2 × 2 min. Next, we incubated sections in combinations of probes from Table 3 for 2 h at 40°C. After that, we incubated slides in AMP1‐4‐FL for 15–30 min each at 40°C and washed with RNAscope wash buffer (#320058; diluted 1:50 in ddH₂O) 2 × 2 min between steps.

TABLE 3.

RNAscope probes used for fluorescence in situ hybridization.

Probe	Common name for target gene/mRNA	Channel	ACD catalog #	Lot #
Mm‐Npff	Neuropeptide FF	C1	479901	20290A
Mm‐Gcg‐C3	Glucagon/Glucagon‐like peptide‐1	C3	400601‐C3	24172E
Mm‐Grp‐C2	Gastrin‐releasing peptide	C2	317861‐C2	19211A
Mm‐Pdyn‐C3	Prodynorphin	C3	318771‐C3	17290A
Mm‐Phox2b‐C2	Phox2b	C2	407861‐C2	19179B
Mm‐Sctr‐C2	Secretin receptor	C2	508511	23163

To label tissues using the Multiplex Fluorescent Reagent Kit v2, we removed tissue from cryoprotectant, mounted them on glass slides, and allowed them to dry for 1 h, then washed with PBS for 5 min, baked for 30 min at 60°C, and then postfixed in 10% neutral buffered formalin (NBF) for 15 min at 4°C. Slides were then dehydrated in a series of 50% ethanol, 70% ethanol, and 100% ethanol solutions, for 5 min each. After repeating the 100% ethanol step, slides were allowed to dry at room temperature. Sections were incubated in hydrogen peroxide (#322380; Advanced Cell Diagnostics) for 10 min at room temperature, then washed with distilled water in two 2‐min washes. We next outlined sections to form a hydrophobic barrier using an ImmEdge pen (Vector Laboratories). The slides were then covered with Protease III, incubated for 30 min at 40°C, then washed for 2 min, twice using ultrapure water. The sections were then incubated in RNAscope probes (Table 2) for 2 h at 40°C. After that, AMPs 1−3 were added, in series, for 15−30 min each, at 40°C, with two 2‐min rinses in 1× RNAscope Wash Buffer between each step. Sections were covered by channel‐specific HRP for 15 min at 40°C, then 150–200 µL Opal dye (cat# FP1496001KT; Akoya Biosciences) for 30 min at 40°C, then HRP blocker for 15 min at 40°C, with two 2‐min RNAscope Wash Buffer rinses between each step. After HRP, we either rinsed sections with PBS 2× or performed an immunofluorescence protocol to label HSD2 or L10GFP. For this, we incubated sections in a PBT–NDS–azide solution with a primary antiserum (either rabbit anti‐HSD2 or chicken anti‐GFP) overnight at 4°C. After washing 3× with PBS the next morning, we incubated sections at 4°C for an additional 2 h in a PBT–NDS–azide solution containing a fluorophore‐conjugated secondary antibody (either Cy3‐conjugated donkey anti‐rabbit or Alexa Fluor 488‐conjugated anti‐chicken). After a final set of PBS washes, we coverslipped all slides using Vectashield with DAPI.

2.6. Imaging, Analysis, and Figures

All slides were scanned using an Olympus VS120 microscope. We began by first acquiring a 2× overview scan and then using a 10× objective to scan all tissue sections. We then acquired 20× and in some cases 40× z‐stacks encompassing all regions of interest for this study. For each slide, this produced a virtual slide image (VSI) file containing a 10× whole‐slide layer, plus separate layers with 20× and/or 40× extended‐focus images in regions of interest. For brightfield images of NiDAB cases immunolabeled for Phox2b or Lmx1b, we used a 20× objective with extended focal imaging (EFI) to collect and combine in‐focus images from 15 focal planes through the tissue.

After reviewing data in OlyVIA (Olympus), we used cellSens to crop full‐resolution images and Adobe Photoshop to adjust brightness and contrast. We used Adobe Illustrator to make drawings, plot cells for figures, arrange images, and add lettering for figure layouts. Scale bars were traced in Illustrator atop calibrated lines from cellSens to produce clean white or black lines in each figure.

2.7. Nomenclature

For mouse genes, we used MGI nomenclature. For proteins and Cre‐reporters, we used common abbreviations from the published literature. For rostrocaudal levels and major subdivisions of the NTS in mice, we adapted the general schema conceived in rats by Saper and colleagues, who subdivided the NTS into three levels (rostral, intermediate, and caudal) and two subdivisions (medial and lateral; Herbert, Moga, and Saper ¹⁹⁹⁰). For other neuroanatomical structures and cell populations, we used and referred to nomenclature defined in peer‐reviewed neuroanatomical literature.

3. Results

3.1. Distinguishing NTS From Surrounding Neurons

Partly distinct distributions of neurons express the nuclear transcription factors Phox2b and Lmx1b (Dai et al. 2008; Kang et al. ²⁰⁰⁷), but plotting the adult distribution of NiDAB immunolabeling for each transcription factor revealed the NTS as a prominent region of overlap. In the dorsomedial medulla, a dense concentration of neurons expressing each transcription factor covered all subdivisions of the NTS at all rostrocaudal levels, as well as the area postrema (black arrows in Figure 1). In contrast, these two transcription factors identified largely separate neuronal distributions outside the NTS. For example, Phox2b‐expressing neurons filled the dorsal vagal motor nucleus (DMV) and nucleus ambiguus (green arrows in Figure 1a–d), while Lmx1b‐expressing neurons were prominent in the medullary raphe and spinal trigeminal nuclei (blue arrows in Figure 1e–h). Both transcription factors identified scattered neurons in the medullary reticular formation, rostral and ventrolateral to the NTS.

FIGURE 1.

Distribution of neurons expressing the transcription factors Phox2b (a–d) and Lmx1b (e–h) at four rostrocaudal levels of the mouse brainstem. Each red dot represents one neuron with nuclear immunolabeling. Black arrowheads highlight dense concentrations in the dorsomedial medulla, nucleus of the solitary tract (NTS), and area postrema. The green arrowheads highlight Phox2b‐immunoreactive neurons in the nucleus ambiguus (a) and in the dorsal vagal motor nucleus (DMV, b–d). Blue arrowheads Lmx1b‐immunoreactive neurons in the medullary raphe and parapyramidal nuclei (f) and in the outer layers of the spinal trigeminal nucleus (h).

Immunofluorescence labeling confirmed the extensive Phox2b distribution in the NTS, area postrema, and DMV and revealed its nearly uniform co‐localization with Lmx1b in neurons across the NTS and area postrema (Figure 2a–d), plus a small number of neurons extending through the ventrolateral margin of the NTS and into the reticular formation. Both Phox2b and Lmx1b were absent from the large, ChAT‐immunoreactive somatomotor neurons of the hypoglossal nucleus (Figure 2a). Dorsal to the hypoglossal nucleus and immediately ventral to the NTS, we did not find any Lmx1b in DMV neurons, which contained nuclear immunoreactivity for Phox2b and cytoplasmic immunoreactivity for choline acetyltransferase (ChAT, blue in Figure 2e–g), nor was Lmx1b expressed in any Phox2b‐expressing and ChAT‐immunoreactive neurons of the nucleus ambiguus (not shown). Thus, co‐expression of Lmx1b identifies a large, noncholinergic subset of Phox2b‐expressing neurons in the NTS and area postrema.

FIGURE 2.

Immunofluorescence labeling for Phox2b (green), Lmx1b (red), and ChAT (blue) in the dorsomedial medulla (a). The expanded inset from the NTS (b–d) highlights the extensive co‐expression of Lmx1b and Phox2b. The expanded inset from the DMV (e–g) highlights the uniform co‐expression of Phox2b in cholinergic DMV neurons, which have cytoplasmic immunoreactivity for choline acetyltransferase (ChAT). In this region, as reported previously (Gasparini et al. 2019; Kang et al. ²⁰⁰⁷), Phox2b and ChAT co‐localized exclusively in the DMV, and ChAT‐immunoreactive (blue) neurons in the hypoglossal nucleus lacked both transcription factors. Scale bar in (a) is 200 µm. Scale bars in (d) and (g) are 20 µm and apply to panels in the same rows.

3.2. Fast‐Neurotransmitter Phenotype of NTS Neurons

To determine the fast‐neurotransmitter phenotype of Lmx1b‐expressing NTS neurons and understand the overall proportions of NTS glutamatergic or GABAergic neurons expressing this transcription factor, we used Cre‐reporter mice for the vesicular GABA transporter (Slc32a1/Vgat) or the vesicular glutamate transporter (Slc17a6/Vglut2). In Slc32a1/Vgat reporter mice, GABAergic neurons intermingled with Lmx1b‐expressing neurons throughout the NTS and area postrema, but none expressed Lmx1b (Figure 3a–c). GABAergic neurons also formed dense clusters in the parasolitarius nucleus (between the dorsal NTS and gracile nucleus) and in the ventrolateral NTS. In Slc17a6/Vglut2 reporter mice, a large majority of glutamatergic neurons in the NTS and area postrema expressed Lmx1b (Figure 3d–f).

FIGURE 3.

Immunofluorescence labeling for the transcription factor Lmx1b (red) in mice expressing an L10GFP Cre‐reporter (green) for Slc32a1/Vgat (a‐c) or Slc17a6/Vglut2 (d‐f). Scale bars in (d) and (e) are 200 µm and apply to remaining panels of similar size.

3.3. Some Caudal, Ventral Neurons Lack Phox2b and/or Lmx1b

Plotting neurons across successive rostrocaudal levels (Figure 4a–d) revealed a ventral band of glutamatergic neurons lacking Lmx1b in the far‐caudal NTS (Figure 4d). Although some of these glutamatergic neurons did express Phox2b, many of them lacked both transcription factors (Figure 5). Thus, Lmx1b co‐expression with Phox2b identifies a noncholinergic, glutamatergic majority of neurons in the NTS and area postrema, except for far‐caudal, ventral neurons that express Phox2b alone (without Lmx1b), and an intermingled subset that lacks both transcription factors. Far‐caudal neurons lacking both Phox2b and Lmx1b are larger, with horizontally elongated somata and dendrites, and their distribution extends laterally, into the reticular formation.

FIGURE 4.

Rostral‐to‐caudal (a–d) plots of L10GFP Cre‐reporter expression (for Slc17a6/Vglut2) and nuclear immunolabeling for Lmx1b in the dorsomedial medulla. Blue arrowheads in panel (d) highlight a band of putatively glutamatergic neurons that lack Lmx1b in the ventral, far‐caudal NTS. Scale bar is 100 µm and it applies to all panels.

FIGURE 5.

Phox2b (red) and ChAT (blue) immunofluorescence in mice expressing the L10GFP Cre‐reporter for Slc17a6/Vglut2 (green), at a level caudal to the area postrema. This marker combination confirmed that Phox2b‐expressing neurons form separate glutamatergic and cholinergic subsets in the NTS and DMV (a–e) and revealed that many glutamatergic neurons lack Phox2b in the caudal‐ventral NTS and adjoining reticular formation. (f) Plotting Lmx1b‐ and Phox2b‐immunoreactive neurons at this caudal level, along with the L10GFP Cre‐reporter for Slc17a6/Vglut2, revealed a dorsal subset of glutamatergic neurons expressing both transcription factors (green with brown), along with a glutamatergic subset expressing only Phox2b (green with red), a moderate number of glutamatergic neurons lacking both transcription factors (green), and a smaller, dorsal subset with only Lmx1b (green with cyan), along with nonglutamatergic Phox2b‐expressing neurons in the DMV (pink dots). Scale bars are 200 µm and apply to remaining panels of similar size.

3.4. Subpopulations of NTS Neurons

Next, we labeled markers for three well‐established subpopulations of NTS neurons to clarify their relationships to the intersectional Lmx1b+Phox2b macropopulation of NTS glutamatergic neurons and to one another.

First, we immunolabeled the enzyme 11‐beta‐hydroxysteroid dehydrogenase type 2 (HSD2). This glucocorticoid‐metabolic enzyme identifies a population of approximately 200 neurons that sense aldosterone and selectively increase sodium appetite (Gasparini et al. 2024, 2019; Jarvie and Palmiter ²⁰¹⁷; Resch et al. ²⁰¹⁷). Across the characteristic V‐shaped distribution of HSD2‐immunoreactive neurons, between the subpostremal NTS and the DMV, a large majority contained Lmx1b (Figure 6a,b), and all HSD2 neurons contained nuclear Phox2b immunoreactivity, as reported previously (Gasparini et al. 2019; Geerling, Chimenti, and Loewy ²⁰⁰⁸; Resch et al. ²⁰¹⁷).

FIGURE 6.

(a–e) Immunofluorescence labeling for the enzymes 11‐beta‐hydroxysteroid dehydrogenase type 2 (HSD2, ice‐blue) and tyrosine hydroxylase (TH, green), in combination with Lmx1b nuclear immunolabeling (red). Scale bars in (c) and (e) are 100 µm, and scale bar in (c) applies to remaining panels.

Second, we immunolabeled tyrosine hydroxylase (TH). This rate‐limiting enzyme for catecholamine synthesis identifies multiple subsets of glutamatergic‐catecholaminergic neurons in the NTS and area postrema (Stornetta, Sevigny, and Guyenet ²⁰⁰²; Zhang et al. ²⁰²¹). TH‐immunoreactivity was mutually exclusive with HSD2, and TH‐immunoreactive neurons had a broader overall distribution in the NTS and area postrema (Figure 6). Many, small TH‐immunoreactive neurons distributed through the area postrema and extended laterally and ventrally into the subpostremal NTS (“A2d” population). All of these smaller neurons contained intense nuclear immunoreactivity for Lmx1b. Further ventrally, larger and medium‐sized TH‐immunoreactive neurons clustered along the dorsolateral edge of the DMV (“A2” population) and extended into far‐caudal levels of the NTS, behind the area postrema. This ventral subpopulation of larger TH‐immunoreactive neurons had lighter and more variable Lmx1b expression, with some members in the far‐caudal NTS lacking this transcription factor entirely. All TH‐immunoreactive neurons had strong nuclear immunoreactivity for Phox2b (not shown).

Third, we immunolabeled Pou3f1, a transcription factor reported to identify neurons in a subregion of the NTS receiving vagal mechanosensory input from the esophagus (Lowenstein et al. 2023). Pou3f1 co‐localized with a prominent subset of Lmx1b‐immunoreactive neurons in the NTS (Figure 7). At and immediately rostral to the obex, this round, dense cluster contained nuclear immunoreactivity for Pou3f1 and cytoplasmic immunoreactivity for nNOS (Figure 8), which is a marker for esophageal premotor neurons (Wiedner, Bao, and Altschuler ¹⁹⁹⁵) that are located in a subregion referred to as the central NTS subnucleus and that send output to the rostral (esophageal) subregion of the nucleus ambiguus (Cunningham and Sawchenko 1989; Herbert, Moga, and Saper ¹⁹⁹⁰). Outside this prominent cluster, we found lighter and more variable Pou3f1 in additional neurons distributed caudally through the medial NTS. None of these contained HSD2 (not shown) or TH (Figure 7).

FIGURE 7.

(a–f) Immunofluorescence labeling for the transcription factors Pou3f1 (blue) and Lmx1b (red), along with TH (green) in the intermediate NTS, rostral to the obex. Scale bar is 200 µm and applies to all panels.

FIGURE 8.

(a–f) Immunofluorescence labeling for Pou3f1 (blue, nuclear immunoreactivity) and neuronal nitric oxide synthase (nNOS, red, cytoplasmic immunoreactivity). Scale bar in (d) is 50 µm. Scale bars in (a) and (f) are 100 and 50 µm, respectively, and apply to remaining panels of similar size.

3.5. Publicly Available Transcriptomic Data

Next, we sought markers for additional subpopulations within the Lmx1b+Phox2b glutamatergic macropopulation by reviewing single‐cell RNA sequencing (scRNA‐seq) data in the Allen Brain Cell Atlas (ABCA; https://knowledge.brain‐map.org/abcatlas). This publicly available dataset includes transcriptomic data from four million aggregate cells collected from adult mouse brains (Yao et al. 2023). Before seeking novel markers, we began by assessing the alignment between these transcriptomic data and the core ontology of molecular markers shown above.

Consistent with our findings, and with previously published information (Gasparini et al. 2019; Geerling, Chimenti, and Loewy ²⁰⁰⁸; Kang et al. ²⁰⁰⁷; Stornetta et al. ²⁰⁰⁶), the ABCA brain‐wide UMAP includes a large subset of neurons expressing Phox2b (Figure 9a–c). Within this Phox2b macropopulation, small subsets express cholinergic marker genes, including Chat (blue arrow in Figure 9d), identifying them as members of the DMV, nucleus ambiguus, and hindbrain motor nuclei (Gasparini et al. 2019; Kang et al. ²⁰⁰⁷; Pattyn, Hirsch, et al. ²⁰⁰⁰). Most other neurons in the Phox2b macropopulation co‐express the vesicular glutamate transporter Slc17a6/Vglut2 (Figure 9e), and in the ABCA taxonomy, these neurons form a large subset of a class designated “24 MY Glut” (medullary glutamatergic neurons). Within this Slc17a6+Phox2b macropopulation, one tiny island of cells expresses the vesicular GABA transporter (Slc32a1/Vgat; red arrow in Figure 9f) with little to no co‐expression of the GABA synthetic enzymes Gad1 and Gad2 (not shown).

FIGURE 9.

Single‐cell transcriptomic analysis from the Allen Brain Cell Atlas (ABCA). (a) Uniform manifold approximation and projection (UMAP) of the approximately four million combined neurons in this public dataset, with medullary glutamatergic neurons (ABCA class “24 MY Glut”) highlighted in pale lilac. (b–p) See Results text for details.

Also consistent with previously published results (Dai et al. 2008; Gasparini et al. ²⁰¹⁹; Gray ²⁰¹³; Guo et al. ²⁰⁰⁸; Karthik et al. ²⁰²²; Z. R. Liu et al. ²⁰¹⁰; Miller et al. ²⁰¹²; Zhao et al. ²⁰⁰⁶), the ABCA brain‐wide UMAP includes a large subset of neurons expressing Lmx1b (Figure 9g). In contrast to Phox2b, Lmx1b is expressed in not only “24 MY Glut” but also the ABCA classes designated “14 HY Glut” (hypothalamic glutamatergic neurons in subclasses 134, 135, and 136, which include previously identified neurons in the subthalamic and parasubthalamic nuclei; Dai et al. ²⁰⁰⁸), “21 MB Dopa” (midbrain dopaminergic neurons; Dai et al. ²⁰⁰⁸; Guo et al. ²⁰⁰⁸), “22 MB‐HB Sero” (midbrain and hindbrain serotonergic neurons; Dai et al. ²⁰⁰⁸; Zhao et al. ²⁰⁰⁶), and “23 P Glut” (pontine glutamatergic neurons, including a large subset of neurons in the parabrachial nucleus; Karthik et al. ²⁰²²). All Lmx1b neurons outside the dopaminergic and serotoninergic subsets are glutamatergic (all express Slc17a6/Vglut2, and none express the vesicular GABA transporter Slc32a1/Vgat).

A distinct subset of these Lmx1b glutamatergic neurons co‐expresses Phox2b. This intersectional subset (Lmx1b+Phox2b) includes most of the Phox2b macropopulation (Figure 9h), but none of the cholinergic Phox2b neurons. This Lmx1b co‐expressing subset also excludes several glutamatergic subsets of Phox2b‐expressing neurons (ABCA clusters 4475–4483 and 4485–4489), which presumably include the Lbx1/Phox2b co‐expressing neurons distributed along the ventral hindbrain surface (Cui et al. 2024; Hernandez‐Miranda et al. ²⁰¹⁸). All the neurons we identified histologically (above) are readily identifiable in the ABCA dataset as mutually exclusive subpopulations of the Lmx1b+Phox2b intersectional macropopulation. Specifically, we found a solitary “peninsula” of neurons expressing Hsd11b2, which encodes HSD2 (green arrow in Figure 9i). We also found a separate swath of neurons expressing Pou3f1 (arrow in Figure 9j), and multiple clusters of neurons expressing the norepinephrine transporter Slc6a2 (Figure 9k), all of which are mutually exclusive with Hsd11b2 and Pou3f1.

3.6. Additional Subpopulations in the Subpostremal, Dorsal, and Lateral NTS

Having confirmed that these ABCA transcriptomic data align with our histologic results, we explored this dataset for additional neurotransmitter markers, neuropeptides, transcription factors, and other genes expressed within the Lmx1b+Phox2b intersectional macropopulation. We began by examining genes and proteins familiar from our previous work (Gasparini et al. 2019; Geerling, Kawata, and Loewy ²⁰⁰⁶) and neuroanatomical literature involving the NTS (Armstrong et al. 1981; C. Chen et al. ¹⁹⁹⁹; Herbert and Saper ¹⁹⁹⁰; Kivipelto ¹⁹⁹¹; Riche et al. ¹⁹⁹⁰; Yamazoe et al. ¹⁹⁸⁴; Zheng et al. ²⁰¹⁵). We also examined genes annotated in the ABCA taxonomy as potential markers for clusters of glutamatergic neurons in the hindbrain.

Overall, neurons in the intersectional Lmx1b+Phox2b macropopulation do not express serotonin or dopamine reuptake transporter genes (Slc6a2, Slc6a4). However, overlapping subsets of this macropopulation do express catecholaminergic markers, including dopa decarboxylase (Ddc), dopamine beta‐hydroxylase (Dbh), and the vesicular monoamine transporter (Slc18a2), in addition to the norepinephrine transporter (Slc6a2, above). All subsets expressing these catecholaminergic marker genes co‐express the vesicular glutamate transporter (Slc17a6/Vglut2) and lack the vesicular GABA transporter (Slc32a1/Vgat).

Several other genes, including Adcyap, Cartpt, Cck, Gad2, Gal, Nmu, Nts, Penk, Phox2a, Slc17a8, Sst, and Tac1, had complex and often overlapping patterns of expression that we will not describe in detail here. However, a small number of neuropeptide genes appear to distinguish mutually exclusive subsets of NTS neurons. For example, strong expression of Npff (neuropeptide FF) identifies a distinct island of neurons (Figure 9m), and this Npff island includes an upper peninsula with strong co‐expression of Grp (Figure 9n).

We localized Npff mRNA to a subpopulation of neurons in the subpostremal NTS, whose distribution extends into the dorsomedial NTS, just lateral to the rostral edge of the area postrema, where a small subset co‐expressed Grp (Figure 10a–c). This distribution of Npff‐expressing neurons resembled previous descriptions of NPFF immunoreactivity in rats (Kivipelto 1991) and Cre‐reporter expression in mice (Quillet et al. 2023). Also, the subpostremal component of this subpopulation resembled the distribution of a Cre‐reporter for Sctr (secretin receptor) reported in mice (figure 3G of Y. Liu et al. ²⁰²³), and we observed in the ABCA dataset that Sctr identified the remaining (non‐Grp) part of the Npff island (Figure 9o). Therefore, we labeled Sctr in combination with Npff, followed by HSD2 immunolabeling, which confirmed that both populations are mutually exclusive with and dorsal to the HSD2 neuron distribution (Figure 11). Sctr‐expressing neurons formed a “V” shape in the subpostremal NTS, and virtually all of these neurons co‐expressed Npff (Figure 11). NTS neurons co‐expressing Npff and Sctr extended rostrally, up to the level of the obex, and we found additional, strong expression of Sctr in a subset of DMV neurons and in the ependymal cells lining the fourth ventricle and central canal of the spinal cord (Figure 12).

FIGURE 10.

(a–c) Npff (green) and Grp (magenta) mRNA labeling in subpostremal and dorsomedial parts of the NTS. Arrowheads highlight the subset of neurons co‐expressing Npff and Grp. All scale bars are 100 µm, and the scale bar in (c) also applies to (b).

FIGURE 11.

Npff (green) and Sctr (blue) mRNA labeling followed by HSD2 immunofluorescence labeling in the medial NTS, ventral to the area postrema. Many neurons co‐express Sctr and Npff (b), but neither are co‐expressed with HSD2 (c, g, h). Scale bar in (a) is 100 µm. Scale bar in (d) is 50 µm and applies to all remaining panels (b–h).

FIGURE 12.

(a–g) Npff (green) combined with Sctr (magenta) and Pdyn (red) mRNA labeling at the level of the obex. Many neurons co‐express Npff and Sctr. Additional Sctr expression was evident in the ependymal cells lining the fourth ventricle and in a small number of DMV neurons. At this level, most Pdyn neurons clustered near the solitary tract in the dorsolateral NTS, separate from neurons co‐expressing Npff and Sctr. Scale bar in (a) is 200 µm. Scale bar in (d) is 200 µm and applies to remaining panels (b–g).

At the same level, labeling Pdyn mRNA revealed an additional subpopulation of neurons that were mutually exclusive with Npff and Sctr and had a more lateral distribution (Figure 12). Caudal to this level, we found fewer and more diffuse neurons, including a small number of Pdyn‐expressing neurons in the DMV (identified previously in Tao et al. ²⁰²¹), and a small number of ependymal cells expressing Pdyn in the midline of the fourth ventricle. However, at this relatively rostral level, the prominent cluster of Pdyn‐expressing neurons occupied a more lateral subregion, near the solitary tract, which included subregions previously described in rats as “interstitial,” “intermediate,” “dorsolateral,” and “ventrolateral” subnuclei (Herbert, Moga, and Saper ¹⁹⁹⁰; Kalia and Sullivan ¹⁹⁸²).

3.7. Caudal NTS

A well‐known population of neurons in the caudal hindbrain produces glucagon‐like peptide 1 (GLP‐1), which is encoded by the gene Gcg. Gcg‐expressing neurons distribute through the caudal reticular formation and into the lateral edge of the caudal NTS, and they reduce food intake (Holt, Richards, et al. ²⁰¹⁹; Kreisler and Rinaman ²⁰¹⁶; Ly et al. ²⁰²³). They are often referred to as NTS neurons, so we were surprised to find in the ABCA that Gcg expression identifies an island of neurons that lack both Lmx1b and Phox2b (Figure 9p). Intrigued by this finding, and having identified a ventral, far‐caudal population of glutamatergic neurons lacking these transcription factors (Figures 4 and 5), we labeled Gcg in combination with Phox2b. At levels caudal to the area postrema, extending ventrolaterally into the caudal medullary reticular formation, Gcg expression distinguished a small subset of medium‐large neurons along the ventrolateral fringe of the NTS. In both the NTS and the reticular formation, Gcg‐expressing neurons intermingled with neurons expressing Phox2b, but we did not find any neurons co‐expressing Gcg and Phox2b (Figure 13a–d). Gcg‐expressing neurons co‐express Slc17a6/Vglut2 in the ABCA dataset (consistent with previous findings in the rat hindbrain; Zheng et al. ²⁰¹⁵), and we found uniform co‐localization between a Cre‐reporter for Slc17a6/Vglut2 and Gcg mRNA (Figure 13e–g). Nearby in the caudal NTS, a subset of noradrenergic (Slc6a2‐expressing) neurons co‐express the neuropeptide Prlh (prolactin‐releasing hormone; C. Chen et al. ¹⁹⁹⁹; Holt and Rinaman ²⁰²²; Ly et al. ²⁰²³) and the calcitonin receptor Calcr (W. Cheng et al. ²⁰²¹). Unlike Gcg, Prlh‐expressing neurons do co‐express Phox2b (Figure 9l), but their Lmx1b expression is low and variable in the ABCA dataset. Thus, the caudal‐ventral band of glutamatergic neurons identified here as lacking either Lmx1b or both Lmx1b and Phoxb2 (Figures 4 and 5) include a Prlh‐expressing subset that lacks Lmx1b and a Gcg‐expressing subset that lacks both Lmx1b and Phox2b.

FIGURE 13.

(a–d) Gcg (red) and Phox2b (white) mRNA labeling in the caudal medulla. No neurons in the NTS or reticular formation co‐express Gcg and Phox2b. (e–g) Gcg mRNA expression localizes to relatively large neurons expressing the L10GFP Cre‐reporter for Slc17a6/Vglut2. Scale bar in (a) is 250 µm. Scale bar in (d) is 50 µm and applies to (b–c). Scale bar in (e) is 50 µm. Scale bar in (g) is 50 µm and applies to (f).

4. Discussion

Our results establish the groundwork for a molecular ontology of NTS neurons. We present a curated overview of molecular markers that identify and distinguish neurons in this region, including novel and previously identified subpopulations (Figure 14). Information presented here has immediate practical implications and opens opportunities for neuroscience research aimed at linking specific NTS neurons to specific interoceptive functions.

FIGURE 14.

Illustration of molecular markers in the NTS.

4.1. Molecular Markers Identify NTS Neurons and Distinguish Subpopulations

Hindbrain neurons develop along rostrocaudal segments known as rhombomeres. Cells in each segment express combinations of transcription factors that, along with positional cues provided by secreted factors along the dorsoventral axis, determine the specification, migration, differentiation, and circuit connections of all hindbrain neurons (Hernandez‐Miranda, Müller, and Birchmeier ²⁰¹⁷; Hirsch et al. ²⁰²¹). We know very little about which transcription factor combinations are responsible for the development of specific NTS subpopulations, but pivotal findings in this region include observations that the NTS fails to develop in mice with deletion of either Phox2b or Rnx Tlx3 (Dauger et al. 2003; Qian et al. ²⁰⁰¹). Phox2b is also necessary for the development of noradrenergic neurons, peripheral autonomic ganglia, parasympathetic premotor neurons, and a subset of motor neurons and glutamatergic interneurons in the hindbrain (Pattyn, Goridis, and Brunet ²⁰⁰⁰; Pattyn, Hirsch, et al. ²⁰⁰⁰; Pattyn et al. ¹⁹⁹⁹).

Beyond this, we know little about factors specifying connections and functions of NTS neurons. Learning that Lmx1b expression together with Phox2b selectively identifies glutamatergic NTS neurons extends previous neuroanatomical studies that highlighted either Phox2b (Kang et al. 2007) or Lmx1b (Dai et al. 2008) in this region. Complementing the necessity of Phox2b for noradrenergic neuron development, Lmx1b plays a critical role in the specification and maintenance of serotoninergic and dopaminergic neurons (Ding et al. 2003; Guo et al. ²⁰⁰⁸; Smidt et al. ²⁰⁰⁰; Zhao et al. ²⁰⁰⁶). Lmx1b is also important for maintaining the glutamatergic identity of neurons in the dorsal horn of the spinal cord and principal trigeminal sensory nucleus (Szabo et al. 2015; Xiang et al. ²⁰¹²). A transcription factor that is co‐expressed with Lmx1b, Tlx3 (Dai et al. 2008), is also necessary for specifying glutamatergic over GABAergic identity in the dorsal horn of the spinal cord (L. Cheng et al. ^{2004, 2005}). Tlx3 is also required for the development of Phox2b‐expressing neurons in the NTS but not in the DMV (Qian et al. 2001). ABCA data confirm broad co‐expression of Tlx3 with Lmx1b, and we hypothesize that these two transcription factors specify and maintain the glutamatergic identity of Phox2b‐expressing neurons in the NTS.

In contrast, Phox2b‐expressing neurons in the nucleus ambiguus and DMV lack Lmx1b and maintain a cholinergic identity. We also identified a small population of Phox2b‐expressing glutamatergic neurons that lack Lmx1b. These neurons, in the caudal NTS, are sandwiched between the larger Lmx1b+Phox2b macropopulation dorsally and the cholinergic DMV Chat+Phox2b population ventrally. They include a Prlh‐expressing subset of A2 noradrenergic neurons (C. Chen et al. ¹⁹⁹⁹), which play a role in feedback‐inhibition of appetite (W. Cheng et al. ²⁰²¹; Holt and Rinaman ²⁰²²; Ly et al. ²⁰²³). We do not know the developmental lineage of these neurons or the functional implications of lacking Lmx1b, nor do we know the lineage of neighboring neurons that lack both Lmx1b and Phox2b—including the Gcg‐expressing (GLP‐1) neurons, which play a related role in appetite feedback (Gaykema et al. 2017; Holt, Richards, et al. ²⁰¹⁹; Holt and Rinaman ²⁰²²; Kreisler and Rinaman ²⁰¹⁶; Vrang et al. ²⁰⁰³). Nonetheless, this information has practical value. For example, planning experiments in Lmx1b‐Cre mice (or any other Lmx1b‐driven recombinase strain) to access NTS neurons would exclude these caudal‐ventral subsets. Likewise, experiments using Phox2b‐Cre mice would exclude Gcg/GLP‐1 neurons and surrounding glutamatergic neurons that lack Phox2b in this caudal region. Overlapping subpopulations of neurons linked to appetite control express the calcitonin receptor Calcr or leptin receptor Lepr (W. Cheng, Gonzalez, et al. ²⁰²⁰; Qiu et al. ²⁰²³), and it is tempting to speculate that a distinct developmental trajectory, revealed by the lack of Lmx1b (or lack of both Lmx1b and Phox2b), specifies patterns of adult gene expression and connectivity that allow a neuron to regulate appetite, rather than one of the autonomic reflex functions classically associated with the caudal NTS. We hypothesize that smaller neurons in the Lmx1b+ macropopulation control autonomic reflexes via local connections in the hindbrain, while larger, non‐Lmx1b/non‐Phox2b neurons at the caudal‐ventral fringe of this region control appetite via long‐range projections to the upper brainstem and forebrain. The molecular ontology detailed here lays a foundation for raising and testing this and similar questions about connectivity and functions of NTS neurons.

In preliminary studies, we have examined adult expression of a variety of other transcription factors, including those highlighted in previous studies of hindbrain development (Gray 2013; Hernandez‐Miranda, Müller, and Birchmeier ²⁰¹⁷), without finding any expression patterns as broadly selective for NTS neurons as the intersection of Phox2b with Lmx1b/Tlx3, and without yet identifying more specific markers for individual NTS subpopulations. Additional subpopulations that are demarcated by more complex combinations of genetic markers may be identified using spatial transcriptomic analysis (K. H. Chen et al. ²⁰¹⁵). Given the vital and clinically important functions of neurons in this intricate and diverse region, we hope that the curated molecular ontology provided here will help accelerate work in the NTS, while also highlighting facets where our understanding remains incomplete.

4.2. Implications for Studying Functionally Distinct NTS Subpopulations

The present results have additional practical implications for studying the connectivity and functions of NTS neurons. As noted above, knowing which genes are co‐expressed (and which are not) should guide the selection of recombinase‐conditional experimental approaches for targeting specific neurons of interest. For example, the total absence of Phox2b mRNA in Gcg‐expressing (GLP‐1) neurons raises the possibility that previous experimental approaches using transgenic Phox2b‐Cre mice to activate or alter gene expression in this population (Scott et al. 2011; Wang et al. ²⁰¹⁵) may not have involved GLP‐1 neurons. Both transcriptomic evidence in the ABCA and our in situ evidence confirmed the mutual exclusivity of Phox2b and Gcg, and these findings complement a previous report that only NTS neurons lacking Gcg/GLP‐1 expression could be transduced by a virally injected GFP vector with the Phox2a/Phox2b‐driven promotor PRSx8 (Holt, Richards, et al. ²⁰¹⁹). The developmental lineage of Gcg/GLP‐1 and neighboring glutamatergic neurons that lack Phox2b is unknown, so the same consideration may apply to several subtypes of neurons in this ventral‐caudal region. Whether or not Phox2b‐negative populations should be considered part of the NTS is open to debate, but learning more about their (likely distinct) developmental lineage could uncover experimentally and clinically useful information.

More importantly, our results provide a useful framework for the debate over whether and to what extent molecular identity defines functionally discrete subpopulations of neurons in the NTS. As detailed above, we can demarcate many subpopulations on the basis of single‐gene expression in the adult NTS, and in many other brain regions, molecular identity dictates distinct patterns of connectivity and function. However, a recent study questioned the link between genetic identity and connectivity in the NTS. That study showed that the established mosaic of input from different cranial nerves and visceral systems (Altschuler et al. 1989; Bassi et al. ²⁰²²; Contreras et al. ¹⁹⁸²; Hamilton and Norgren ¹⁹⁸⁴; Kalia and Sullivan ¹⁹⁸²) corresponds to a similar topography of NTS neurons responding to stimulation of different locations along the gastrointestinal tract (Ran et al. 2022). In an attempt to study a link between this viscerotopic pattern and genetically defined subtypes of NTS neurons, the authors injected viral vectors that delivered a non–Cre‐dependent calcium indicator and Cre‐dependent GFP in several knockin‐Cre mouse strains (Calcr‐, Cartpt‐, Chrh2‐, Gcg‐, Penk‐, Pdyn‐, Sst‐, Tac1‐, and Th‐Cre). Many of these Cre‐driver genes—particularly Cartpt, Penk, Sst, and Tac1—are expressed broadly, in multiple subpopulations scattered across several NTS subregions, and most of the uniquely expressed genes highlighted above were not used. The authors found responses to mechanical distention of both the stomach and duodenum in GFP‐expressing neurons of each strain and interpreted this finding as evidence that the position of a neuron within the NTS may hold more useful information about its input connectivity than the molecular identity of that neuron (Ran et al. 2022). Arguing against this perspective are several examples of marker‐defined NTS subpopulations with highly specific connections and distinct functions.

First, the aldosterone‐sensitive HSD2 neurons project axons from the NTS to highly specific targets in the upper brainstem and forebrain (Gasparini et al. 2019; Geerling and Loewy ²⁰⁰⁶; Jarvie and Palmiter ²⁰¹⁷; Resch et al. ²⁰¹⁷). The present results confirm our previous reports in rats and mice showing that this population is distinct from other identified subpopulations of NTS neurons (Gasparini et al. 2019; Geerling, Kawata, and Loewy ²⁰⁰⁶). Activating Hsd11b2‐expressing neurons selectively increases sodium appetite (Jarvie and Palmiter 2017; Resch et al. ²⁰¹⁷), and removing Hsd11b2‐expressing neurons from the NTS eliminates aldosterone's ability to increase sodium appetite (Gasparini et al. 2024). Importantly, HSD2 neurons do not receive direct synaptic input from the vagus nerve and do not project axons to regions of the hindbrain that control autonomic reflexes, and activating these neurons has no effect on blood pressure (Gasparini et al. 2018; Resch et al. ²⁰¹⁷). They surround the area postrema, which has an opposing effect on sodium appetite, and molecular‐genetic techniques are necessary to study HSD2 neurons, arguing against the utility of topography for determining cell‐type/function relationships in this subregion.

Second, separate subpopulations of neurons in the caudal, ventral NTS potently reduce food intake. These neurons express an overlapping variety of molecular markers, including Calcr, Cartpt, Cck, Gcg, Lepr, and Prlh. Stimulating Cck‐expressing NTS neurons that send output to the lateral parabrachial nucleus has an aversive impact and reduces food intake (Qiu et al. 2023; Roman, Sloat, and Palmiter ²⁰¹⁷), while Gcg/Lepr and Calcr/Prlh neurons reduce appetite without an aversive effect, suggesting that their function may be to promote postprandial satiation (W. Cheng, Gonzalez, et al. ²⁰²⁰; W. Cheng, Ndoka, et al. ²⁰²⁰; W. Cheng et al. ²⁰²¹; Gaykema et al. ²⁰¹⁷). Further, Prlh‐ and Gcg‐expressing populations appear to relay separate oral or gastroduodenal nutrient and stretch‐sensory information, and activating them produces either prolonged inhibition of (Gcg) or brief pauses in (Prlh) food intake (Ly et al. 2023). Stimulating a catecholaminergic Dbh/Npy‐expressing subset of NTS neurons increases food intake (J. Chen et al. ²⁰²⁰), consistent with a separate report that rostral catecholaminergic (Th‐expressing) NTS neurons increase hunger, while caudal Th‐expressing neurons (presumably including the Calcr/Prlh subset) reduce food intake (Aklan et al. 2020; Sayar‐Atasoy et al. ²⁰²³). Supporting the value of topographical information is the finding that neurons in the area postrema expressing the GLP‐1‐receptor (Glp1r) inhibit appetite with adverse hedonic effects, while those in the underlying NTS lack the averse hedonic affect (Huang et al. 2024).

Third, the distribution of Pdyn‐expressing neurons we identified above, surrounding the solitary tract at intermediate levels of the lateral NTS subdivision, bears striking resemblance to the terminal distribution in rats of vagal axons arriving from the larynx (figure 12B of Altschuler et al. ¹⁹⁸⁹). This afferent/cell‐type correspondence was not investigated in the calcium imaging experiments described above (Ran et al. 2022), and it strongly suggests a role of these lateral, Pdyn‐expressing neurons in gag, swallow, or cough reflexes. The cough reflex was linked to Tac1‐expressing neurons in a subsequent study (Gannot et al. 2024), but Tac1 is expressed in multiple clusters of NTS neurons, and it is unclear which subset promotes cough‐like reflex behavior.

Fourth, one of the most clear‐cut input–output relationships in the entire brainstem is the close correspondence between vagal axons arriving from the esophagus and esophageal premotor neurons located in the prominent, round “central subnucleus” of the NTS (Cunningham and Sawchenko 1989; Herbert, Moga, and Saper ¹⁹⁹⁰; Wiedner, Bao, and Altschuler ¹⁹⁹⁵). We replicated the previously identified distribution of neurons expressing the nuclear transcription factor Pou3f1 in this location and showed that a prominent, round subset of these neurons expresses Nos1/nNos, which was previously shown to identify esophageal premotor neurons in rats (Wiedner, Bao, and Altschuler ¹⁹⁹⁵). Pou3f1‐expressing NTS neurons receive input from vagal sensory neurons that express Prox2/Runx3 and innervate the upper gastrointestinal tract (Lowenstein et al. 2023). While awaiting functional confirmation, these neuroanatomical features strongly suggest that these Pou3f1/Nos1 neurons play a highly specific role in reflex control of esophageal motility.

To our knowledge, no subpopulation markers have been identified for gustatory‐relay neurons in the rostral NTS. Similarly, no cell‐type–specific markers have been identified for NTS neurons controlling several well‐known autonomic reflexes, including the baroreflex, chemoreceptor reflex, gastric vago‐vagal reflex, gastro‐colic reflex, Hering–Breuer reflex, and vasovagal reflex. Our new findings provide several new candidates as a template for identifying and testing new subpopulations in future work. For example, while we do not yet have a subpopulation‐specific marker for these neurons, the prominent population of large GABAergic (Slc32a1/Vgat‐expressing) neurons in the ventrolateral NTS may be the “pump cells” that receive slowly adapting stretch receptor input from the lungs and mediate the Hering–Breuer reflex (Ezure and Tanaka 2004). Future work may also narrow down a more specific subpopulation responsible for the “sickness behaviors” and “retching‐like behavior” that have been attributed to widespread populations of NTS neurons that express the neuropeptide Adycap1 and the calcium‐binding protein Calb1, respectively (Huo et al. 2024; Ilanges et al. ²⁰²²).

4.3. Limitations

Our approach emphasized reproducibility by focusing on genes with stable expression in adult mice. However, some genes are differentially expressed, for example, in a fed vs. fasted condition (Dowsett et al. 2021). One well‐known example of variable gene expression is the activity‐induced gene Fos, which can distinguish a variety of different neurons after specific experimental perturbations that selectively activate a neuron of interest. If any subpopulation of NTS neurons activates its unique pattern of gene expression only under a specific condition, it is unlikely that our study or any other attempt to categorize NTS neurons in a limited number of baseline conditions would detect that subpopulation.

Another limitation is that we deliberately focused on the NTS and not the area postrema, which contains neurons that detect toxins and evoke nausea, because neuronal subpopulations in the area postrema were described in a previous study (Zhang et al. 2021). Also, while we focused on transcription factors, neuropeptides, and a limited number of other genes, additional gene‐families may be useful for distinguishing NTS neurons. In particular, G‐protein–coupled receptors can be useful targets for drug therapies, and we and others have identified interesting expression patterns for a handful of G‐protein–coupled receptors, including the angiotensin II receptor Agtr1a and the secretin receptor Sctr in the medial NTS (Gasparini et al. 2019; Gonzalez et al. ²⁰¹²; Y. Liu et al. ²⁰²³; Resch et al. ²⁰¹⁷), as well as Calcr, Gfral, and Glp1r the area postrema (W. Cheng, Gonzalez, et al. ²⁰²⁰; W. Cheng et al. ²⁰²¹; Huang et al. ²⁰²⁴; Ludwig et al. ²⁰²¹; Qiu et al. ²⁰²³; Zhang et al. ²⁰²¹). These receptors may allow circulating peptide hormones and drugs to modulate the activity of specific subsets of neurons in the subpostremal NTS and area postrema (Resch et al. 2017; Zhang et al. ²⁰²¹).

In addition, we focused on glutamatergic NTS neurons without identifying subtype‐specific markers for GABAergic neurons in this region. Most GABAergic neurons in the NTS probably function as local interneurons, without long‐range axonal projections outside the NTS, but some may project axons to the ventrolateral medulla and other sites, and these projection neurons likely include the large GABAergic “pump cells” in the ventrolateral NTS, which control the pulmonary Hering–Breuer reflex (Tanaka and Ezure 2004). The developmental lineage of these and other GABAergic neurons in the NTS remains unknown, but it is presumably different from that of Lmx1b‐ and Phox2b‐lineage glutamatergic neurons. Further work is needed to determine this lineage and whether this region includes functionally distinct GABAergic subpopulations demarcated by specific genetic markers.

Finally, beyond the core ontology of Lmx1b and Phox2b expression, our approach focused on subpopulation boundaries demarcated by mutually exclusive, single gene expression. This approach successfully distinguished NTS subpopulations expressing Gcg, Grp, HSD2/Hsd11b2, Npff, Pou3f1/Nos1, and TH/Slc6a2, but some subpopulations may not express a single, unique marker. Isolating those subpopulations will require multidimensional analysis. Depending on the chosen algorithm and other analytic design choices, clustering multidimensional data from many thousands or millions of single‐cell transcriptomes can generate divergent results, even from the same dataset, and previous scRNA‐seq studies have reported several different neuronal clustering schemas (Dowsett et al. 2021; Ilanges et al. ²⁰²²; Lowenstein et al. ²⁰²³; Ludwig et al. ²⁰²¹; Reiner et al. ²⁰²²; Su et al. ²⁰²⁴; Tao et al. ²⁰²¹; Zhang et al. ²⁰²¹). Future work using spatial transcriptomic methods (K. H. Chen et al. ²⁰¹⁵) will extend our present findings and help clarify discrepancies among previous scRNA‐seq reports. Regardless of the analytic methods used, combined labeling of individual genes and proteins in brain tissue remains the gold‐standard way to confirm or refute any newly hypothesized molecular demarcations derived from algorithmic cluster analysis.

4.4. Conclusion

Our study provides a useful framework for targeting specific subpopulations of NTS neurons. Future work in this region should add scaffolding to this curated foundation of molecular markers, which will help guide translational work in other species, including humans.

Author Contributions

J.C.G. planned the project, secured funding, and drafted the manuscript with S.G. S.G., G.A.P., and A.S.P.M. performed histology and microscopy. S.G. and G.A.P. performed high‐sensitivity in situ hybridization. G.A.P. and S.G. drafted the figures. G.A.P., S.G., and J.C.G. edited the figures and figure legends together. S.G. and J.C.G. edited and finalized the text together. All authors approved the text and figures in this manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/cne.70004.

Data Availability Statement

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