Neuropeptide S (NPS) neurons: Parabrachial identity and novel distributions

Abstract Neuropeptide S (NPS) increases wakefulness. A small number of neurons in the brainstem express Nps. These neurons are located in or near the parabrachial nucleus (PB), but we know very little about their ontogeny, connectivity, and function. To identify Nps‐expressing neurons within the molecular framework of the PB region, we used in situ hybridization, immunofluorescence, and Cre‐reporter labeling in mice. The primary concentration of Nps‐expressing neurons borders the lateral lemniscus at far‐rostral levels of the lateral PB. Caudal to this main cluster, Nps‐expressing neurons scatter through the PB and form a secondary concentration medial to the locus coeruleus (LC). Most Nps‐expressing neurons in the PB region are Atoh1‐derived, Foxp2‐expressing, and mutually exclusive with neurons expressing Calca or Lmx1b. Among Foxp2‐expressing PB neurons, those expressing Nps are distinct from intermingled subsets expressing Cck or Pdyn. Examining Nps Cre‐reporter expression throughout the brain identified novel populations of neurons in the nucleus incertus, anterior hypothalamus, and lateral habenula. This information will help focus experimental questions about the connectivity and function of NPS neurons.

short-sleepers (4−5 h per night). Replacing a single copy of mouse Npsr1 with this human variant (Y206H) enhanced responsivity to NPS, increased wakefulness by more than an hour per day, and prevented a memory deficit that normally occurs after sleep deprivation (Xing et al., 2019).
Relative to this genetic and pharmacologic information about its receptor, we know little about the neurons that produce NPS. In rodents and humans, NPS neurons were found in a similar brainstem region (Adori et al., 2015;Clark et al., 2011). The initial report (Xu et al., 2004) used in situ hybridization to identify Nps-expressing neurons near the locus coeruleus (LC) in rats. Nps-expressing neurons lacked gene expression that identifies catecholaminergic neurons in the LC, and likely release the fast-excitatory neurotransmitter glutamate because they coexpress a vesicular glutamate transporter, not the enzymes that synthesize GABA (Xu et al., 2007). These reports in rats and a study in mice (Clark et al., 2011) also identified a rostral cluster of Nps-expressing neurons. Rostral NPS neurons reportedly occupy the parabrachial nucleus (PB) and principal sensory trigeminal nucleus in rats (Xu et al., 2004; and the Kölliker-Fuse nucleus in mice (Clark et al., 2011;Liu et al., 2011). It is unclear whether these neuroanatomical discrepancies reflect species differences or interpretational differences within this complex region of the brainstem.
Currently, we lack the neuroanatomical and genetic information necessary to identify NPS neurons within the molecular framework of the PB region. Such information would improve our understanding of NPS neurons, the region they inhabit, and the neural circuits connected to this region. This information would be useful for designing and interpreting genetically targeted experiments that distinguish the connectivity and function of NPS neurons from other neurons in this complex and diverse region.
The PB region contains two macropopulations of glutamatergic neurons, which derive from separate embryonic precursors (Karthik et al., 2022). Adult neurons generated from these precursors form intermingled yet mutually exclusive populations. Each macropopulation contains subpopulations distinguished by expression of neuropeptides, receptors, and other genetic markers Geerling et al., 2016;Grady et al., 2020;Huang et al., 2020;Karthik et al., 2022;Miller et al., 2012;Palmiter, 2018). Many other neuronal populations closely surround the PB. These include the LC, laterodorsal tegmental nucleus, mesencephalic trigeminal nucleus, and Barrington's nucleus, but it remains unclear how NPS neurons relate to these populations.
Based on their location and distribution in the PB region, we hypothesized that Nps-expressing neurons are a subset of Atoh1derived neurons. To test this hypothesis and better characterize NPS neurons, we used a combination of in situ hybridization, immunolabeling, and Cre fate-mapping for Atoh1. To enable genetically targeted experiments involving these neurons, we also generated mice with Cre recombinase knocked into the endogenous Nps locus and crossed these mice to a Cre-reporter strain, which allowed us to identify cells with a history of Nps expression.
For additional information about Cre-driver and -reporter mice, see Table 1. All mice were group-housed in a temperature-and humiditycontrolled room on a 12/12-h light/dark cycle with ad libitum access to water and standard rodent chow (Envigo 7013). All experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committees at the University of Iowa.
Pronuclear-stage embryos were collected using methods described in Pinkert (2002). Embryos were collected in KSOM media (Millipore; MR101D) and washed three times to remove cumulus cells.
Cas9 RNPs and single-stranded repair template were injected into the pronuclei of the collected zygotes and incubated in KSOM with amino

Perfusion and tissue sections
All mice were deeply anesthetized with ketamine-xylazine (i.

Fluorescence in situ hybridization
To label various mRNA transcripts, we used RNAscope probes listed in
We incubated these sections overnight at room temperature on a tissue shaker.

Imaging, analysis, and figures
All slides were scanned using an Olympus VS120 microscope. We began by first acquiring a 2× overview scan then using an 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.
We used cellSens (Olympus) to count cells that contained Nps mRNA labeling in four cases with high-quality labeling in rostral and caudal sections to count and measure Nps neurons. We used Abercrombie correction as an approximate way to adjust for double-counting at the tissue edges (Guillery, 2002) and then multiplied the corrected cell counts (from our 1-in-3 tissue series) by 3 for a rough estimate of the number of Nps-expressing neurons in a mouse brainstem.
Cell counts and measurements are expressed as mean ± standard deviation. We used QuPath (Bankhead et al., 2017) to count cells expressing Nps mRNA, Nmb mRNA, and/or L10GFP in sections from Nps-2A-Cre;R26-lsl-L10GFP mice. We counted every cell that contained in-focus labeling of at least 10 Nps mRNA puncta, and we did not count any lone, diffuse, or out-of-focus puncta. To identify cells with L10GFP expression across all z-stacks, we counted every cell that had cytoplasmic L10GFP and a nuclear void in focus.

Nomenclature
For rat and mouse genes, we used MGI nomenclature. For rat and mouse proteins and Cre-reporters, we used common abbreviations from the published literature. For neuroanatomical structures and cell populations, where possible, we used and referred to nomenclature defined in peer-reviewed neuroanatomical literature. In some instances, we used or referred to nomenclature derived from rodent brain atlases (Dong, 2008;Paxinos & Franklin, 2013;Paxinos & Watson, 2007;Swanson, 1992).

Nps-expressing neurons in the PB region
Overall, very few cells contained Nps mRNA. We counted an average of 119 ± 41 Nps-expressing cells (bilateral counts from n = 4 mice) in a 1-in-3 series of brainstem sections spanning the central midbrain and rostral medulla. Multiplying by 3 and Abercrombie-correcting each count (Guillery, 2002) suggested that the mouse brainstem contains approximately 300 Nps-expressing cells (278 ± 63).

Nps-expressing cells had a distinctly neuronal appearance. They
were medium-sized, slightly oblong or round, with a long-axis diameter of 17.1 ± 3.5 µm (range 9.5−27.6 µm; n = 160). Their neuronal morphology was also evident from abundant Nps mRNA content, which in many spilled out from the soma into one or two proximal dendrites.

Foxp2 colocalization
Nps-expressing neurons overlap PB subregions where many glutamatergic neurons express the transcription factor Foxp2 (Karthik et al., 2022). Using in situ hybridization to determine the relationship between this population and Nps-expressing neurons, we found uniform colocalization between Nps and Foxp2 (Figures 4 and 5).
Every Nps-expressing neuron also expressed Foxp2. Conversely, Npsexpressing neurons were a small subset of the much larger population of Foxp2-expressing neurons in this region.

Nps and other neuropeptidergic Foxp2 subpopulations
Among Foxp2-expressing neurons located caudally, the Nps-expressing subset concentrated medial to the LC. We next examined their The rostral lateral PB contains a large population of neurons that express the neuropeptide Cck, many of which also express Foxp2 Grady et al., 2020). To examine the relationship between Nps and Cck, we labeled mRNA for both genes in combination with Foxp2. Again, we found uniform colocalization between Nps and Foxp2. We also found Foxp2 expression in many Cck-expressing neurons, but Nps and Cck were mutually exclusive ( Figure 5). Caudally, near the LC, fewer neurons expressed Cck and were again mutually exclusive with Nps (not shown).

Nps-expressing PB neurons are Atoh1-derived
Many Foxp2-expressing neurons in the PB region derive from embryonic precursors that express the transcription factor Atoh1 (Gray, 2008;Karthik et al., 2022). However, this region also contains Foxp2expressing inhibitory neurons, which derive from Ptf1a-expressing precursors Gray, 2008;Karthik et al., 2022), and a rostral subset of Foxp2-expressing neurons derives from Lmx1a/bexpressing embryonic precursors (Karthik et al., 2022). To determine the origin of Nps-expressing neurons, we used Cre fate-mapping for

Atoh1.
In brainstem tissue from Atoh1-Cre;R26-lsl-L10GFP Cre-reporter mice, we used in situ hybridization to label Nps mRNA. Virtually every Across four mice with the strongest labeling, we counted between 9 and 71 NPS-immunoreactive neurons bilaterally (43 ± 29). As with Nps mRNA, NPS-immunoreactive neurons uniformly expressed the L10GFP Cre-reporter for Atoh1 (Figure 7). Also, they contained nuclear immunoreactivity for FoxP2, not Lmx1b ( Figure 8). Labeling Lmx1b and FoxP2 also confirmed that all rostral NPS-immunoreactive neurons are dorsal to the Kölliker-Fuse nucleus (Figure 8a and b).

Nps Cre-reporter in the PB region
We used Nps

Nps Crereporter in the nucleus incertus
We found a large population of L10GFP-expressing neurons spanning the midline of the pontine central gray matter, between the left and right cranial nerve VII genu (Figure 9e and f). Their distribution included a region referred to as "nucleus incertus" (Dong, 2008) or "central gray alpha/beta/gamma" (Paxinos & Franklin, 2013). Within Nps mRNA (Figure 14a-d). We found the same patterns of Nps mRNA labeling in mice without Nps-2A-Cre (Figure 14e and f). Just rostral to the hypothalamic cluster, we found a few, scattered neurons in the F I G U R E 1 1 In the rostral lateral PB, neurons expressing the L10GFP Cre-reporter for Nps contain FoxP2 but not Lmx1b. In this region, immunoreactivity for Lmx1b (red, a) and FoxP2 (blue, c) define mutually exclusive populations of neurons except in the KF, which contains neurons coexpressing both transcription factors (e). L10GFP-expressing neurons (green, b) are dorsal to Lmx1b-immunolabeled neurons (d) and contain FoxP2 immunolabeling (f). These L10GFP-expressing neurons are dorsal to the KF in the lateral PB (g). Scale bars in (f; also applies to a-e) and (g) are 500 µm. Additional mcp, middle cerebellar peduncle; DR, dorsal raphe nucleus F I G U R E 1 2 The caudal cluster of Nps Cre-reporter neurons is medial to the LC, with some neurons extending into Barrington's nucleus (Bar). Most L10GFP-expressing neurons (green, b) in this region contained FoxP2 (blue, c) and not Lmx1b (red, a). FoxP2 and Lmx1b identified mutually exclusive neuronal populations, with FoxP2-immunoreactive neurons encircling Bar and light Lmx1b immunoreactivity identifying neurons in the LC (f). (d-g) Merged images. Bar contained several L10GFP-expressing neurons that lacked FoxP2. Scale bars in (a; also applies to b-f) and (g) are 200 µm lateral preoptic area, near the bed nucleus of the stria terminalis. Immediately lateral to the hypothalamus, we found a small cluster of neurons in the medial amygdala.

Additional Nps Cre-reporter expression
Due to these unexpected sites of Cre-reporter expression, we analyzed the full brain in a series of Nps-2A-Cre;R26-lsl-L10GFP mice (n = 10). In a 1-in-3 series of tissue sections spanning the mouse brain, we counted 1770 ± 425 L10GFP-expressing neurons. After Abercrombie correction and multiplication by 3, this represents approximately 3789 ± 900 neurons in the mouse brain with a history of Nps expression. L10GFP-containing neurons had a similar size compared to Nps mRNA containing neurons, with a long-axis diameter of 16.0 ± 1.7 µm (range 13.8−19.7 µm; n = 300). There were no differences in the neuroanatomical distribution and appearance of L10GFP-containing neurons among cases, and the overall number did not differ between male (3869 ± 1015; n = 6) and female (3669 ± 827; n = 4) Nps-2A-Cre;R26-lsl-L10GFP mice (p = .75 by two-tailed t-test).
These include 2927 ± 905 (77.3%) in the brainstem, primarily in the PB, but also in the nucleus incertus and rare, scattered neurons in the rostral periaqueductal gray matter. The remaining L10GFP-expressing neurons clustered in the hypothalamus (382 ± 154; 10.1%), thalamus (414 ± 97; 10.9%), and medial amygdala (65 ± 35; 1.7%), as described above. We did not find any L10GFP-expressing neurons in the cerebral cortex, except for a lone hippocampal neuron in two cases. We did not find any L10GFP-expressing neurons in the basal ganglia, basal forebrain, or septal nuclei. We did not find any L10GFP-expressing neurons in the cerebellum. Finally, we did not find any L10GFP-expressing

NPS promotes wakefulness, and genetic polymorphisms in NPSR1
have been linked to several human diseases, but the homeostatic Additionally, we will discuss the unexpected finding of neurons with a history of Nps expression in novel neuroanatomical locations.

Developmental-genetic framework of the PB region
The brainstem tegmentum at the midbrain-hindbrain junction contains many, diverse neurons. The larger, more distinctive neurons (locus coeruleus, mesencephalic trigeminal nucleus, laterodorsal tegmental nucleus, and Barrington's nucleus) stand out from an expansive background of smaller neurons in the PB and pontine central gray matter, which have received less attention.
The neuroanatomical boundaries subdividing these regions in current brain atlases derive from efforts in rats to organize this region using cytoarchitecture (Fulwiler & Saper, 1984;Paxinos & Watson, 2007;Swanson, 1992). Subsequent patterns of connectivity, gene and protein expression, Fos activation, and Cre fate-mapping blurred these boundaries and identified neuronal populations that do not fit this framework (Gasparini et al., 2021;Karthik et al., 2022). Delineations in rats do not always translate cleanly to mice (Biag et al., 2012;Gasparini et al., 2019;Karthik et al., 2022) and may be even less helpful for translating findings from rodents to humans. Importantly, it is not possible to use position or cytoarchitecture in this region to draw borders between most neuronal populations because subpopulations with separate connections and functions intermingle extensively (Gasparini et al., 2019;Geerling et al., 2016;Geerling et al., 2011;Grady et al., 2020;Huang et al., 2020;Karthik et al., 2022;Miller et al., 2012).
We recently described a molecular ontology of transcription factors and other genetic markers that define and differentiate subpopulations of neurons in and around the PB (Karthik et al., 2022). In the rostral rhombic lip of the embryonic brain, neuroepithelial cells that express the transcription factor Atoh1 generate a wide variety of neurons, including the PB, Barrington's nucleus, and most neurons in the cerebellum (Akazawa et al., 1995;Ben-Arie et al., 1997;Machold & Fishell, 2005;Rose et al., 2009;Wang et al., 2005). Most Atoh1-derived neurons in the PB express Foxp2 and intermingle with a separate macropopulation of glutamatergic neurons, which derives from a mutually exclusive lineage marked by the transcription factors Lmx1a and Lmx1b (Karthik et al., 2022).
Identifying NPS neurons as Atoh1-derived and Foxp2-expressing places them within the molecular ontology of the PB region.

NPS neurons are a subset of Atoh1-derived, Foxp2-expressing PB neurons: distinctions and implications
Our results add Nps to a growing list of neuropeptides (including Pdyn, Cck, and Grp) that subdivide the Atoh1-derived PB macropopulation into distinct subpopulations (Karthik et al., 2022).
In previous work, we identified a rostral and caudal pair of neuronal clusters that express Foxp2, receive direct input from aldosteronesensitive neurons, and activate in response to sodium deprivation (Gasparini et al., 2019;Geerling & Loewy, 2006Geerling et al., 2011). The rostral cluster is in the lateral PB, and the caudal cluster is located near the LC (Gasparini et al., 2021). Much like these sodiumdeprivation-activated neurons, NPS neurons form distinctive rostral and caudal clusters. However, their molecular identity is distinct.
In contrast to NPS neurons, sodium-deprivation-activated neurons express Pdyn (Gasparini et al., 2021;Lee et al., 2019). Also, most of the caudal NPS cluster is medial to the locus coeruleus, while the distribution of "pre-LC" sodium-deprivation-activated neurons skews laterally, into the medial PB (Gasparini et al., 2021). Based on this genetic and neuroanatomical dissection, we predict that Nps-expressing neurons do not respond to sodium depletion and do not play a role in sodium appetite.
Considering that NPS reduces feeding (Smith et al., 2006) and that glutamatergic neurons near the LC reduce food and water intake (Gong et al., 2020;Li et al., 2019), it is tempting to speculate that NPS neurons relay meal-related information from the NTS to the hypothalamus, constraining food or fluid consumption. It remains unknown whether such viscerosensory information reaches any NPS neurons or how that information interacts with wake-promoting effects of NPS. Of note, meal timing influences circadian rhythm (Bolles & Stokes, 1965;Boulos et al., 1980;Mistlberger, 1994), and NPS neurons may serve as a bridge between meal-related viscerosensory information and the diencephalic neurons that control sleep/wake switching and other aspects of circadian physiology.
Clearly, a more complete understanding of the NPS system will require learning what activates Nps-expressing neurons. In a previous study, restraint stress and forced-swim stress activated neurons in the PB region expressing an Nps-eGFP transgene . Beyond this, we have very little information about NPS neuron activity, and we do not know what input signals they receive. Neighboring pre-LC neurons integrate input from the NTS, hypothalamus, bed nucleus of the stria terminalis, and cerebral cortex (Gasparini et al., 2019;Kelly & Watts, 1998;Lee et al., 2019;Li et al., 2019), but we do not yet know if NPS neurons receive input from these regions.
Regarding their output connectivity, learning that NPS neurons express Foxp2 and derive from Atoh1-expressing precursors provides a clue to the list of brain regions they are likely to target (Huang et al., 2020;Karthik et al., 2022). Anterograde and retrograde tracing in rats and mice defined the array of brain regions that do and do not receive input from Foxp2-expressing neurons, relative to other glutamatergic and GABAergic populations in the PB region Huang et al., 2020;. Based on this information, Npsexpressing neurons in the PB region probably do not project axons to the cerebral cortex, basal forebrain, amygdala, or most regions of the hindbrain. This is noteworthy due to the prominent expression of the NPS receptor (Npsr1) in the cerebral cortex, basal forebrain, and amygdala (Clark et al., 2011;Xu et al., 2007). These regions may represent mismatches between NPS and its receptor (Clark et al., 2011), but our identification of Nps-expressing neurons in the thalamus, hypothalamus, and medial amygdala expands the possible sources of endogenous NPS. For example, NPS release in response to stress has been detected in the amygdala using microdialysis (Ebner et al., 2011).
Based on results of conventional axonal tracing (Luppi et al., 1995;Shin et al., 2011) and Cre-conditional tracing in Foxp2-IRES-Cre mice (Huang et al., 2020), we predict that NPS neurons in the PB project their axons to the paraventricular thalamic nucleus (PVT), hypothalamic subregions, and possibly the lateral septum, ventral tegmental area, or periaqueductal gray matter. These regions are known to contain NPS-immunoreactive axons and Npsr1-expressing neurons, and particularly dense axonal labeling and Npsr1 mRNA highlight the PVT as a potentially important target (Clark et al., 2011). The expanded population of neurons with a history of Nps expression identified in our Cre-reporter mice extends far more rostrally than the PB injection sites in previous Foxp2-IRES-Cre tracing experiments (Huang et al., 2020), so we cannot rule out the possibility of additional output targets.
Planning and interpreting future experiments which evaluate possible functions of NPS neurons would benefit from a detailed, brain-wide map of their axonal projections.

Distribution of Nps-expressing neurons
Despite differences in neuroanatomical nomenclature, the patterns of Nps mRNA and NPS immunolabeling we observed in the PB region are consistent with previous reports (Adori et al., 2015;Clark et al., 2011;Xu et al., 2004). Our estimate that approximately 300 neurons contain Nps mRNA in this region resembles a previous report that approximately 500 neurons in the mouse brainstem expressed an Nps-eGFP transgene . Medial to the LC, Cre-reporter expression and FoxP2 immunolabeling revealed Nps reporter-expressing neurons in Barrington's nucleus.
This nucleus contains multiple glutamatergic subpopulations (Keller et al., 2018), and it will be important to determine the role of those with a history of Nps expression.
Further medially, we found neurons with a history of Nps expression in the nucleus incertus. Neurons here send output to the interpeduncular nucleus, septum, and several other brain regions (Goto et al., 2001). Stimulating the Nmb-expressing subset of nucleus incertus neurons increases locomotor speed and pupillary dilation (Lu et al., 2020).
Few neurons here actively expressed Nps in our cases, but a sizeable minority with the Nps Cre-reporter also expressed Nmb. It will be important to distinguish the connectivity and function of nucleus incertus neurons with a history of Nps expression.
Discovering Nps expression in the habenula was particularly surprising because a transcriptomic study of this brain region did not identify substantial Nps expression (Hashikawa et al., 2020). Lateral habenular neurons reduce dopamine neuron activity, which is important for negative reward prediction (Christoph et al., 1986;Jhou et al., 2009;Matsumoto & Hikosaka, 2007;Stamatakis & Stuber, 2012). It will be interesting to learn the function of habenular neurons that express

Nps.
In the anterior hypothalamus, we identified another population of Cre-reporter-expressing neurons ventral to the paraventricular hypothalamic nucleus. The subparaventricular region receives input from several limbic brain regions and contains GABAergic interneurons that inhibit glutamatergic neurons in the paraventricular nucleus (Herman et al., 2002). A ventral subregion referred to as the subparaventricular zone (SPZ) relays input from the suprachiasmatic nucleus to neurons implicated in circadian function (Vujovic et al., 2015). It will be important to determine the functional role of neurons in this region that express Nps.

Limitations
The L10GFP intensity and distribution were similar across Nps Crereporter mice and did not vary with mRNA labeling intensity. This discrepancy reflects expected differences between the dynamic transcription of endogenous Nps and the strong, continuous expression of L10GFP expression from the ROSA26 locus with a CAG promoter (Krashes et al., 2014). The purpose of this study was to characterize NPS neurons within the molecular framework of the PB region, and we used Nps knockin Cre-reporter mice to study adult neurons with a history of Nps expression but did not examine peripheral tissues or embryonic Nps or L10GFP expression. We also did not test the func-

CONCLUSIONS
The potential therapeutic importance of NPS makes it imperative that we learn more about neurons that produce it, but few studies have interrogated the connectivity and function of these neurons. Our results clarify the neuroanatomical distribution and molecular identity of NPS neurons within the diverse framework of the PB region. This information lays the groundwork for future experiments involving NPS neurons.

AUTHOR CONTRIBUTIONS
JCG planned the project, secured funding, helped with initial histology and microscopy, and drafted the paper with RZ and DH.