Differential Ascending Projections From the Male Rat Caudal Nucleus of the Tractus Solitarius: An Interface Between Local Microcircuits and Global Macrocircuits

To integrate and broadcast neural information, local microcircuits and global macrocircuits interact within certain specific nuclei of the central nervous system. The structural and functional architecture of this interaction was determined for the caudal nucleus of the tractus solitarius (NTS) at the level of the area postrema (AP), a relay station of peripheral viscerosensory information that is processed and conveyed to brain regions concerned with autonomic-affective and other interoceptive reflexive functions. Axon collaterals of most small NTS cells (soma <150 μm2) establish excitatory or inhibitory local microcircuits likely to control the activity of nearby NTS cells and to transfer peripheral signals to efferent projection neurons. At least two types of cells that constitute efferent pathways from the caudal NTS (cNTS) were distinguished: (1) a greater numbers of small cells, seemingly forming local excitatory microcircuits via recurrent axon collaterals, that project specifically and unidirectionally to the lateral parabrachial nucleus; and (2) a much smaller numbers of cells likely to establish multiple global connections, mostly via the medial forebrain bundle (MFB) or the dorsal longitudinal fascicle (DLF), with a wide range of brain regions, including the ventrolateral medulla (VLM), hypothalamus, central nucleus of the amygdala (ACe), bed nucleus of the stria terminalis (BNST), spinal cord dorsal horn, brainstem reticular formation, locus coeruleus (LC), periaqueductal gray (PAG) and periventricular diencephalon (including the epithalamus). The evidence presented here suggests that distinct cNTS cell types distinguished by projection pattern and related structural and functional features participate differentially in the computation of viscerosensory information and coordination of global macro-networks in a highly organized manner.

The population of cNTS cells is heterogeneous with respect to morphological features such as soma size, shape, subnuclear location and axo-dendritic arborization Negishi and Kawai, 2011). The relationship between cell morphology and projection sites in cNTS efferent systems has not yet been studied using neural tract tracers, except for studies describing the topographical subnuclear arrangement of NTS-PB pathways . We have investigated correlations between distinctive morphological features of cNTS neurons and functional properties such as patterns of spontaneous excitatory or inhibitory postsynaptic activity . A quantitative correlation is evident not only between each morphological feature but also between morphological and functional parameters. For example, the soma size of cNTS neurons is an excellent marker for other morphological and functional features: (1) small cells (soma <150 µm 2 ) have mostly local axon collaterals and are divided into excitatory (glutamatergic) and inhibitory (GABAergic) neurons, of which the former are more concentrated in the dorsal region of the cNTS possessing an extranuclear projection axon in addition to intranuclear local collaterals; and (2) larger cells (soma >150 µm 2 ), which lack local axon collaterals, receive robust inhibitory synaptic inputs and are concentrated specifically in the ventral subregion Senba, 1996, 1999;Okada et al., 2006Okada et al., , 2008Yoshioka et al., 2006;Negishi and Kawai, 2011). The purpose of the present study was to assess differences in projection targets or pathways of projection axons. Such local structure/function correlates of NTS neurons are essential for understanding the role of the cNTS as a relay station for peripheral visceral sensory inputs.

Animal Preparation
All surgical and experimental procedures were approved by the Institutional Committee for the Care and Use of Experimental Animals at the Jikei University School of Medicine and performed in accordance with Guidelines for Proper Conduct of the Animal Experiments by the Science Council of Japan. The tracer study was carried out using 32 male Sprague-Dawley rats (weight range, 280-310 g). Animals were anesthetized with an intraperitoneal (i.p.) injection of ketamine (30 mg/kg) and xylazine (24 mg/kg) and placed in a stereotaxic instrument for tracer injections. In some cases, 0.5% isoflurane was additionally administered through a nose mask to obtain sufficient depth of anesthesia during tracer injections. For disector analysis of the number of neurons and somal areas in the cNTS, five male Sprague-Dawley rats weighing 290-306 g were deeply anesthetized with urethane (1 g/kg i.p.) and processed for immunocytochemistry.

Tracer Study
A glass micropipette (tip diameter, 1-15 µm) was filled with either the anterograde tracer biotin dextran amine (BDA; 2%, MW 10,000; Molecular Probes, Eugene, OR, USA) or the retrograde tracer cholera toxin B subunit (CTB; 2%, Sigma, St. Louis, MO, USA) dissolved in phosphate-buffered saline (PBS; 0.01 M, pH 7.4). After making an incision in the atlanto-occipital dural membrane, either iontophoretic or pressure injections of BDA were administered through the glass micropipettes inserted into the exposed left dorsal medulla at the level of the area postrema (AP) under a stereoscopic microscope (Bregma −13.7, Lateral 0.6, Depth 8.2 in mm; Paxinos and Watson, 1998). The depth was 100-600 µm from the brain surface. Iontophoretic injections of CTB were made through the glass micropipettes inserted into the left side of the brain, targeting several restricted regions as defined in the stereotaxic atlas of Paxinos and Watson (1998). Iontophoretic delivery of BDA or CTB was achieved by administering a 0.1-0.3 µA current in a cycle of 5 s on and 5 s off for 20 min using an intracellular pre-amplifier (IX1, Dagan Corp, Minneapolis, MN, USA). Pressure injection of the tracer solution (volume, 100-300 nL) was made under direct visual inspection using a pneumatic microinjector (IM-11-2, Narishige, Tokyo, Japan). After completion of the injection procedure, the pipette was removed, the skin was sutured with surgical glue, and the rats were replaced into their home cages. After a survival period of 7-10 days, the rats were deeply anesthetized with urethane (1 g/kg i.p.) and processed for histological procedures.

Histology
After deep anesthesia with urethane, the animals were perfused transcardially with 50 mL saline, followed by 300 mL of a fixative consisting of 0.1 M phosphate buffer containing 4% paraformaldehyde. Brains were removed and postfixed overnight in the same fixative, followed by immersion in 30% sucrose in 0.1 M phosphate buffer for cryoprotection. Cryostat sections (50 µm thickness) were serially collected in PBS.
For stereological analysis, cryostat sections were first incubated overnight in PBS containing rabbit polyclonal primary antibody against neuronal-specific-enolase (NSE; 1:1, CA#N-0649, Antigen: γ/γ isoenzyme of NSE from human brain, Sigma, St. Louis, MO, USA) with 5% bovine serum albumin and 0.3% Triton X-100, followed by an incubation in PBS Axon terminals in lateral parabrachial nucleus (PBL) of each case are shown in Figure 4. Based on the visual examination of BDA axonal labeling in each brain region of tissue sections: "−" extremely few or no potential boutons or varicose axons; "+" lightly scattered boutons and/or varicose axons, "++" moderately dense boutons and varicose axon, "+ + +" dense axon terminal field with abundant branches and boutons.

Analysis of Cell Size and Number
The soma size and number of neuronal profiles stained by CTB or NSE, respectively, were measured and counted based on photographed images of neuron profiles. For CTB-immunostained profiles, the number containing a clear nucleus was counted in the dorsal and ventral subregions of the cNTS at the level of the AP. For NSE-immunostained profiles, a disector method of stereological analysis (West, 1999;Guillery, 2002) was applied for counting the number of neuronal cell bodies in each subnuclear region of the cNTS at the level of the AP, using a frame size of 100 µm × 100 µm . The number of soma profiles was expressed as numerical density (ND). The method is described in detail elsewhere .
The contour of each cell soma immunostained for CTB or NSE was digitally traced, and the area was measured on enlarged photographic images using image analysis software (ImageJ v.1.34n) 1 . The size distribution graph was normalized and constructed using statistical software (Microcal OrignPro9.1J, Malvern Instruments, Malvern, Worcestershire, UK).

BDA as an Anterograde Tracer and the Injection Sites
Each BDA injection usually resulted in a demarcated injection site whose core varied in diameter from 50 to 900 µm, depending on the BDA volume injected and the resultant labeled injection size. Only cases in which the maximum labeled extent of the injection site was within the medial portion of the NTS (often with some spread into the dorsal motor nucleus of the vagus, dmnX) at the level of the AP were analyzed in the present study (n = 16; Table 1). Densely labeled neurons with Golgi-like dendrites were observed within the injection core and its vicinity. Outside the injection site, no cell bodies or dendrites were labeled, and only axon fibers and boutons were found in multiple brain regions either in isolation or en masse. Maximum injections (Case 1a, b) involving most of the medial portion of the NTS resulted in wide-ranging axon labeling, reaching the diencephalic and telencephalic structures. Smaller injections produced a variety of patterns of axon labeling within the brainstem structures, depending on the size and density of BDA in each injection site (Cases 2, 3 a-e; Table 1; Figure 1). ). This case presented the most extensive and widespread distribution of labeled axons in this study. The projection was dominantly ipsilateral, with minor contralateral labeling in the pons (Figure 4). In the caudal medulla, thickets of labeled axons from the injection site traversed first ventrolaterally to the VLM ( Figures 1I,H), with most of them then turning sharply in a rostral direction and ascending to the pons, with a small contingent descending to the dorsal horn of the spinal cord (DH; Figure 1J). In the pons (Figure 1G), BDA-labeled axons coursed sharply in a dorsal or dorsomedial direction to form moderate to dense terminal fields in the lateral parabrachial nucleus (PBL), locus coeruleus (LC; A6 catecholamine cell group), and subcoeruleus nucleus. Along the ascending course in the ventrolateral regions of the medulla through pons, boutons and branched varicose axons were found apparently associated with the A1, C1, A5 and A7 catecholamine cell group regions in the reticular formation and somatic or visceral motor nuclei including the ambiguus, facial, trigeminal motor, and superior salivatory nuclei. In the caudal midbrain at the level of the inferior colliculus ( Figure 1F), two groups of loose or scattered labeled axons with boutons were seen. The first group, dorsolateral fascicle (DLF or Schütz fascicle), coursed dorsomedially into the PAG, with moderate boutons and varicose axons, especially in the ventrolateral portion of the PAG. This group seemed to be associated with the laterodorsal tegmental (LDT) and pedunculopontine tegmental (PPT) cholinergic cell group regions. The second and larger group of labeled axon fibers of a loose texture coursed dorsally along the lateral lemniscus and formed axon pathways ascending to the rostral midbrain. In the rostral midbrain at the level of the superior colliculus ( Figure 1E), the first group of labeled axons (DLF) ascended rostrally in the PAG, with few boutons in the dorsal portion of the PAG. The second group formed an ascending axon bundle further rostrally in the lateral tegmental tract and gave off a small group of labeled axons ventromedially along the medial lemniscus to the interpeduncular nucleus (IP), with few boutons or varicose axons associated with A8, 9, 10 catecholamine cell group regions. In the caudal diencephalon at the level of the mammillary nucleus ( Figure 1D), the second main group of labeled axons primarily coursed sharply in the ventromedial direction to form the authentic medial forebrain bundle (MFB; Nieuwenhuys et al., 1982). The DLF traveled further rostrally in the vicinity of the third ventricle. In the more rostral diencephalon (Figures 1B,C), the DLF gave off few boutons in the periventricular thalamic nucleus and the habenular nuclei. The second group of axons (MFB) gave off moderate boutons or varicosities with branched axons in regions surrounding the MFB: the lateral, dorsomedial, and ventromedial hypothalamic nuclei (LH, DMH, VMH), the median eminence (ME), the paraventricular hypothalamic nucleus (PVH), and the supraoptic and suprachiasmatic nuclei. In the telencephalon (Figures 1A,B), moderate numbers of boutons and varicosities with branched axons were observed in the BNST and ACe, being dispatched by the MFB. Few varicose axons were found in the ventral pallidum (VP), the septal nucleus, the nucleus of the horizontal limb of the diagonal band (HDB), the medial preoptic nucleus (MPO), and further up to the piriform and insular cortices.

Anterogradely Labeled Axons From the cNTS
The maximum injections within the cNTS of the four rats (Case 1a, #1-4) generally presented similar overall distribution FIGURE 3 | Drawing of anterogradely labeled axons with boutons in representative brain regions following the injection of BDA in the cNTS. Both types of boutons en passant and terminaux were differentially observed according to the brain regions. Dense aggregates of fine boutons are noted consistently in the nucleus of the tractus solitarius (NTS) and PBL, while larger boutons were observed less frequently in the BNST. Each example of axons and boutons was traced and drawn en bloc as an interconnected structure sampled from preparations of 50-µm thick sections.
patterns of labeled axons and boutons in a wide range of brain regions, from the telencephalon to the spinal cord (Figure 1), with site-specific features of each axon bouton arrangement (Figure 3). The highest density of boutons and varicose axons was seen consistently in the NTS (injection site) and the PBL (Figures 2, 3). The high density of varicose axons with boutons was characterized by concentrated aggregations of a greater number of boutons and/or varicosities of a small diameter (0.5-1.2 µm) interconnected by branched finer axons with shorter intervals in these two regions (Figure 3). A similar feature was noted in axon bouton aggregates in other brain regions, including the BNST, PVH and LC; however, each bouton was much larger and only seen in a minor portion of the axons in every region (Figure 3). An overall high relative density of varicose axons was seen in the LC (Figure 2), possibly due to the compactness of the cell population characteristic of this nucleus. Moderate-to-few boutons with varicose axons sharing similar morphologic features were seen globally in the PVH, ACe, DMH, PAG, A5 area, VLM and many other brain regions (Figures 2, 3). These varicose axons were characterized by their boutons en passant with longer lengths of a thicker interconnected axon or/and boutons terminaux attached to an unbranched, short axon emitting from the parent axon (Figure 3).
When the injection volume was smaller, varied patterns of labeled axon distribution were seen, as shown in Table 1. In general, the smaller the injection volume, the fewer labeled axons in particular target regions and the shorter the maximal distance between the labeled axons and the injection site. A salient feature of labeled axons according to injection size was observed to be associated with the PBL in particular, as described below. Figure 4 shows several patterns of labeled axons in the PBL following BDA injections in the cNTS at the level of the AP, depending on differences in the size and density of the tracer deposits. With maximum injection volume of BDA into the cNTS (a photograph in Figure 4C1), the PBL, especially the inner layer (PBLi), received by far the densest innervation in the whole brain (Figures 4A,C1). This region corresponds precisely to the external and part of the dorsal, central PBL . The dense terminal boutons were seen to be provided by thick ascending axon bundles running dorsally from the pontine ventral reticular formation (Figures 4B,D). A compact bundle (arrows in Figure 4A) coursing along the ventral spinocerebellar tract (VSC), gave off several smaller bundles of axon fibers piercing the outer layer (PBLo) to the dense terminal field in the PBLi (Figures 4A,C1). In addition to the axon boutons in the FIGURE 4 | Qualitative characteristics in the differential appearance of anterogradely labeled axon fibers and boutons in the PBL depending on the size and density of BDA deposits in the cNTS. (A) Photograph of maximal axon labeling in the PBL following a large injection in the cNTS (See also C1,D). Case 1a #3. A particularly dense field of BDA-labeled axon boutons was noted in the inner portion of the PBL (PBLi, Bernard et al., 1993). The bouton aggregate appeared to be provided by several smaller bundles of axon fibers piercing the outer layer (PBLo, Bernard et al., 1993) from parent axons traveling along the ventral spinocerebellar tract (VSC: arrows in A) and more numerously by those of axons entering from the ventral aspect (B). In addition to the dense terminal field, moderate-to-few boutons and varicose axons were scattered throughout the PBL. See also (C1,3). (C1-3) Drawings of characteristic axon fibers and boutons in the PBL following injections of BDA into the cNTS shown in each photograph. (C1) Maximal axon labeling. Case 1a #2. (C2,3) Weak-to-moderate labeling was observed differentially in the inner and outer layers of the PBL, according to the size and density of BDA deposits. Case 2 #1 and Case 3c #1, respectively ( Table 2). (D) Mapping of each area of (A-C). Note labeled axons in the contralateral parabrachial nucleus (PB) and LC via axons traveling in the superior medullary velum (arrowheads). See also Figure 1G. (E) Ascending axons of both passing and terminating types from the cNTS in the PB and surrounding regions at the level of ponto-midbrain juncture. They seem to contain at least three axon bundle systems, indicated here by different colors: blue, MFB; green, DLF; red, those terminating in the PBL.

Parabrachial Nucleus
PBL ipsilateral to the injection side, this compact axon bundle provided boutons and varicose axons in the contralateral LC and PBL after traveling in the superior medullary velum (arrowheads in Figure 4D).
Two examples of smaller injections were drawn in Figures 4C2,3. In the former case (Case 2 #1 in Table 1), a tracer deposit of ∼50 µm in diameter was concentrated in the dorsal NTS at the level of the AP. Very few labeled neurons with dendrites were outside the injection core. In this case, no labeled ascending axons were found beyond the pons ( Table 1).
In the PBL, labeled varicose axons with boutons terminaux were found specifically in the PBLi, corresponding to a highly dense varicose axon region or bundled axon tracts in the maximum injection cases. On the other hand, in the latter case (Case 3b #1 in Table 1), in which several scattered labeled cells with larger soma and long dendrites were seen associated with a small injection core (Figure 4C3), scattered varicose axons with boutons were seen more widely in the PBL (both PBLi and PBLo) and the superior cerebellar peduncle (SCP). Few-tomoderate axons of similar features were seen in the LC and PAG; however, no axons were found beyond the midbrain ( Table 1). The results underscore the presence of ascending axons of both passing and terminating types from the cNTS in the PB and surrounding regions at the level of the ponto-midbrain juncture. We concluded that the PB and surrounding regions appear to contain at least three ascending axon systems, as shown by different colors in Figure 4E: the MFB in blue, the dorsal longitudinal fascicle (DLF) in green, and a tract terminating in the PBL in red. These three types of axon tracts might provide differential varicose axons and boutons in the PB and the surrounding regions, in terms of anatomical features such as density, subnuclear distribution and bouton type.
To discriminate more clearly the three possible tracts ascending from the cNTS, the retrograde tracer cholera toxin B subunit (CTB) was injected into several brain regions belonging to each system, and the morphologic features of the labeled neurons were quantitatively analyzed in the cNTS at the level of the AP.

Retrogradely Labeled Neurons in the cNTS
In brain regions in which labeled axons were relatively dense following BDA injection into the cNTS, six regions of the PBL, PAG, PVH, ACe and BNST were selected for injection of the retrograde tracer CTB to allow characterization of each ascending pathway and to identify any differences in location or cell size between origin neurons in the cNTS at the level of the AP. Figure 5A shows the CTB injection sites (PBL, PB, PAG, PVH, ACe and BNST), all of which appear in the G, E, B, or A plane levels in Figure 1. For all CTB injections, retrogradely labeled neurons were most numerous in the NTS at the level of the AP (Figure 5B), especially in the medial portion. The retrograde labeling was dominant ipsilateral to the injection side. The most rostral NTS (top sections in each panel in Figure 5B) showed few or no retrograde neuronal labeling except for a case in which a retrograde tracer was injected into the PB that included the medial PB, a terminal field of the gustatory pathway (Norgren, 1978). Figure 6 presents photographs at low and higher magnifications showing retrogradely labeled neurons in the cNTS at the level of the AP, following CTB injections in discrete brain regions of the PBL, PAG, PVH, ACe and BNST. In one case of PBL (a, top row), retrogradely labeled cells were densely distributed evenly in the medial portion of the cNTS, including the AP. In contrast, in other cases of CTB injections, retrogradely labeled cells were distributed more frequently in the ventral region of the cNTS, and the number of labeled cells was far smaller than that of the PBL case. The size, density, and location of retrogradely labeled neurons in the cNTS following CTB injection was not uniform with respect to injection site (Figure 6). The apparent difference in these anatomical features according to injection site was quantitatively addressed to analyze possible differences in efferent systems ascending from the cNTS.

Cell Size and Intranuclear Location
In most cases of CTB injection, the location of retrogradely labeled neurons was uneven along the dorso-ventral orientation in the medial portion of the cNTS at the level of the AP. Therefore, we first examined the neuronal cytoarchitecture of this region to compare it to that of retrogradely labeled cells using a stereological method with neuron-specific-enolase (NSE)-immunocytochemistry . NSE immunoreactivity was detected in perikaryal cytoplasm of virtually all neurons in the cNTS (Figure 7B). The dorsal region of the cNTS corresponds to the dorsomedial subnucleus and subpostrema zone of the commissural subnucleus , which are reported to receive specifically peripheral gastrointestinal and baroreceptor afferent terminals (Kalia and Sullivan, 1982;Chan et al., 2000). The ventral region of the cNTS corresponds to the medial subnucleus and ventral portion of the commissural subnucleus , which are reported to receive specifically central afferent terminals from the PVH and several telencephalic regions (van der Kooy et al., 1984).
Using a disector frame of 100 µm × 100 µm applied to the dorsal and ventral subregions of the cNTS (Figures 7A,B), the cell number in (ND/mm 3 ) and cell area in µm 2 were measured (Figures 7C1,2; Table 2). A different cell-size distribution was noted between the subregions of the cNTS at the level of the AP (Figures 7C1,2). The dorsal cNTS consisted of a homogeneous population of small neurons (99.1%; 94.5 µm 2 in mean soma area; Figure 7C1), whereas the ventral cNTS consisted of at least two population of small cells (61.4%; 104.8 µm 2 in mean soma area) and larger cells (39.6%; 161.9 µm 2 in mean soma area; Figure 7C2). The mean soma area of cells in respective subregions was 106.5 µm 2 and 127.6 µm 2 ( Figure 7C, Table 2). Based on stereology using NSEimmunocytochemistry, ND of each subregion was calculated as 1.9 ×10 5 /mm 3 (dorsal cNTS), 2.3 × 10 5 /mm 3 (ventral cNTS), respectively ( Table 2). Since the volume of each subregion was about 5.2 × 10 −3 mm 3 (50 µm thick section), the dorsal and ventral cNTS contained about 1000 and 1200 neurons, respectively ( Table 2).
The quantitative stereological results based on NSE immunocytochemistry of cNTS cells served as a reference for comparison with those of the retrogradely labeled cNTS neurons following CTB injections into discrete brain regions ( Table 2). Table 2 shows the soma size (µm 2 in mean soma area), number of cell profiles per section, and subnuclear bias of cell location (as percentage) of the cNTS neurons projecting into each brain region (PBL, PAG, PVH, ACe and BNST). With regard to soma size, dorsal cNTS neurons, PBL-and PAG-projecting neurons shared a similar value of about 105 µm 2 in mean area ( Table 2) and a size distribution graph mostly fitted by a Gaussian curve with a single peak ( Figure 8B). A closer look at the PAG graph ( Figure 8B), however, reveals a small hump at about 150 µm 2 in soma area, suggesting that the PAG-projecting cNTS cells contained a very small fraction of larger neurons. On the other hand, the ventral cNTS neurons, PVH-, ACe-and BNST-projecting cells consisted of rather heterogeneous populations of cells in terms of soma size. Each group of cells contained varied proportions of larger neurons of different sizes (Table 2, Figure 8B). Consequently, their size distribution graphs could not be fitted by a simple Gaussian curve (Figure 8B), and their mean soma sizes shifted to larger values of 128-158 µm 2 ( Table 2). The distribution graph shapes and subnuclear locations were similar between PVH-and BNST-projecting cells. These groups of projection cells were the largest cells of ventral cNTS neurons, having a soma area of more than ∼200 µm 2 ( Figure 8B). Collateralization of projection axons between the two groups might occur, considering the similarities in overall shape of the size distribution graphs and the subnuclear location of labeled cells. In contrast,  Table 2.
ACe-projecting cells generally had a smaller soma area than PVH-or BNST-projecting neurons, suggesting differential ascending pathways of parallel rather than collateral systems, if any, through the MFB. The results demonstrate that 61% of PBL-, 89% of PAG-, and more than 98% of PVH-, ACe-, BNST-projecting cells of small and larger sizes were localized in the ventral cNTS (Figure 8, Table 2). In addition, small GABAergic neurons are reported to be exclusively concentrated in the ventral cNTS (Okada et al., 2008). Thus, the ventral cNTS neurons seem to consist of extremely varied populations of locally and globally projecting cells containing at least those elucidated in the present study (Figures 8, 9). The conspicuous heterogeneity of the ventral cNTS neurons produced a size distribution graph of complex Gaussian curves with multiple peaks (Figures 7, 8).

DISCUSSION
The results of this study indicate that ascending efferent systems from the cNTS establish a wide-ranging network that involves multiple brain regions, including the PBL, PAG, PVH, ACe and BNST (Figure 9). In addition to the discrete brain nuclei, the network appears to be associated with not only multiple catecholaminergic and cholinergic cell groups but also several circumventricular organs such as the subfornical organ (SFO), organum vasculosum lamina terminalis (OVLT), ME, and AP ( Figure 9). The ascending pathway from the cNTS courses through the ventrolateral portion of the reticular formation of the pons and medulla oblongata and diverges abruptly in the dorsolateral pons at the pons-midbrain junction into at least three channels: the cNTS-PBL tract, MFB and DLF. The cNTS-PBL tract is the densest pathway of the ascending system from the cNTS. The MFB and DLF diverge near the PB. The MFB reaches telencephalic structures such as the piriform cortex, insular cortex, septal nucleus, VP, HDB, lateral preoptic area and BNST after giving rise to small branches of axon bundle at various levels of the diencephalon and midbrain. The DLF courses through the PAG and ascends near the third ventricle to reach thalamic and epithalamic structures, such as the periventricular thalamic nuclei and the habenular nuclei. The longer ascending pathways of the MFB and DLF are provided by different neuronal populations of the cNTS, as judged by the soma size of labeled cells (Figure 8). The former originates from larger NTS neurons, while the latter from small cells. Furthermore, each single NTS cell belonging to the same pathway seems to give rise to multiple short collateral axons to more than one brain area along each pathway (MFB or DLF), judging by the number gap between origin NTS cells and axon fibers/boutons in the destination. The cNTS-PBL pathway is the third system, which differs from both the MFB and DLF as judged by the soma location, size, and number of originating cell populations (Figures 8, 9). The three distinct ascending pathways originate,  Table 2. (B) Normalized size distribution of the target-specific projection neurons (colored lines, upper) and region-specific cNTS cells (dotted or lined lines, lower). Note the projection-specific difference in intranuclear location and cell-body size between the cNTS neurons. Also note the similarity in morphological features between PVH-and BNST-projecting neurons.
in substantial percentage, from differential neuronal populations of the cNTS. It is coincidental that a collateral long axon pathway appears to play a minor contribution to the wide-ranging efferent system originating from the cNTS (Hermes et al., 2006).

Comparison to Data and Methodologies in the Literature Using Anterograde Tracers
Autoradiography using radioactive amino acids as an anterograde tracer has elucidated a wide-ranging efferent system originating in the NTS. The possible destinations include the parabrachial nucleus (PB), PAG, hypothalamic nuclei, periventricular thalamic nuclei, and several forebrain structures including the ACe and BNST (Norgren, 1978;Ricardo and Koh, 1978;Beckstead et al., 1980). In particular, Ricardo and Koh (1978) identified the MFB and DLF as two neural channels that provide wide-ranging axons in the diencephalon and telencephalon. Recently, a more sensitive anterograde tracer, Phaseolus vulgaris leucoagglutinin (PHAL), was introduced and applied to studies of several brain regions, including the rostral and cNTS, to describe detailed efferent axon terminal structures in specific brain regions, including the hypothalamus, PAG and PB (Gerfen and Sawchenko, 1985;ter Horst et al., 1989;Herbert and Saper, 1992).
In the present study, the qualitative features of the projections ascending from the cNTS are confirmed to allow analysis of the more comprehensive architecture of this wide-ranging, multipleefferent system. Indeed, this study is the first to address which pathway is the main projection and the comparative weights of each pathway. It was also uncertain whether and to what extent the multiple-projection system is established by axon collaterals. For this purpose, studies have used retrograde double labeling with different fluorescent tracers; however, over-or underestimation may be inescapable due to methodological limitations (Hermes et al., 2006). Moreover, this method may be difficult to apply to the NTS efferent system, since the PB seems to be particularly critical for the assessment of collateralization of efferent pathways. Therefore, a careful quantitative comparison was made between these multiple projections using a sensitive retrograde CTB tracer combined with stereological estimation of cNTS neuron size and number.
A recently introduced tracer, BDA (10 K only), was confirmed to be more suitable than the widely-used PHAL as an anterograde tracer for study of the system efferent from the cNTS for several reasons (Sarhan et al., 2005). BDA can be injected via 128.8 ± 32.6 (52.0-229.5, n = 234) 18.6 ± 6.0 (11-29, n = 15) 98.5 (% in the ventral region) BNST 3 153.2 ± 58.6 (58.2-417.7, n = 255) 21.6 ± 5.0 (16-31, n = 9) 98.0 (% in the ventral region) Numbers and areas were counted and measured for NSE-immunopositive somal profiles identified stereologically in the dorsal and ventral cNTS ( Figure 7) and for retrogradely labeled somal profiles in 50 µm thick sections, following injection of cholera toxin B subunit (CTB) in PBL, periaqueductal gray (PAG), paraventricular hypothalamic nucleus (PVH), amygdaloid central nucleus (ACe) and bed nucleus of the stria terminalis (BNST; Figure 5). * Values were calculated for 50 µm thick section volumes of respective portions of cNTS, based on the numerical density (ND). Statistical values are presented as the mean ± SD (range, number of samples). both electrophoresis and air pressure. In addition, the volume of the BDA injection can be strictly controlled by changing either the diameter of the glass pipette tip or the injection duration (Sarhan et al., 2005;Lanciego and Wouterlood, 2011;Ling et al., 2012). Taking advantage of the merits of this tracer, differential injections of BDA into the cNTS were made to gain a more comprehensive and comparative understanding of the cNTS-PBL pathway.

Parabrachial Nucleus
The present study shows that the PBL is the densest innervation site from the cNTS. The medial PB is reported to be heavily innervated by the rostral NTS (gustatory pathway), which sends very few axons beyond the midbrain in rats (Norgren, 1978;ter Horst et al., 1989). A robust NTS→PB connection, not reciprocal, was confirmed (Fulwiler and Saper, 1984; and is considered to terminate mostly in the PB (Ito and Seki, 1998). The PB has been divided into several distinct subnuclei, each of which is reported to make specific afferent and efferent connections with various brain regions (Fulwiler and Saper, 1984;. A field of particularly dense innervation in the PBL from the cNTS (Figures 4A,C1) corresponds precisely to the external lateral subnucleus and part of the central and dorsal lateral subnuclei of the PB . Aside from the terminal fields of a dense innervation from the NTS, several other subnuclei of the PB, together with the surrounding reticular formation including the Kölliker-Fuse nucleus (Figure 4C1), also received few-tomoderate axon boutons and varicose fine axons from the cNTS. The PB and surrounding regions are known to send wide-ranging efferent axons mostly through ascending pathways to various brain regions, including the zona incerta (ZI), hypothalamic nuclei (dorsomedial, lateral, paraventricular hypothalamic nuclei and median preoptic nucleus), BNST, ACe (via MFB system) and periventricular thalamic nuclei (via DLF system; Saper and Loewy, 1980;Fulwiler and Saper, 1984;Bernard et al., 1993;Alden et al., 1994;Bester et al., 1997Bester et al., , 1999. In the present study, most of these regions were also confirmed to be directly innervated by the cNTS. The PB appears to be a discrete relay station for multiple major ascending projections from the lower brainstem and spinal cord, with only a minor contingent of descending projections; part of the dorsal and lateral PB projecting to the VLM and the Kölliker-Fuse nucleus projecting to the ventrolateral subnucleus of the NTS and spinal cord (Fulwiler and Saper, 1984). The present study indicates the presence of at least three ascending projections (the cNTS→PBL tract, MFB, and DLF) from the cNTS that are differentiated in and around the PB (Figure 9). Each projection is distinct not only in density and location of varicose axons in the PB subnuclei but also in density, location, and size of neurons in the cNTS subregion.
The PB is considered to be a relay station, not only of viscerosensory afferent information from the NTS but also of somatosensory afferents conveying nociceptive and temperature information (Cechetto et al., 1985;Slugg and Light, 1994;Feil and Herbert, 1995). The spino-and trigeminoparabrachial pathways are topographically organized in the PB and Kölliker-Fuse nucleus (Feil and Herbert, 1995). The topographically intimate organization of terminal fields of varicose axons are also evident between the visceral and somatosensory afferents Feil and Herbert, 1995), underscoring the anatomical substrate of the PBL as autonomic and somatosensory integration. However, it remains to be determined whether varicose axons seen in the PB and surrounding reticular formation (including the Kölliker-Fuse nucleus provided by the spino-and trigemino-parabrachial pathways) are primarily of the en passant or terminaux type. A particularly interesting question is whether there is a pathway corresponding to the cNTS→PBL tract in the somatosensory system.
Most of the small neurons in the dorsal subregion of the cNTS are non-GABA cells and have a single projection axon with branched local varicose axons Senba, 1996, 1999;Paton et al., 2000;Yoshioka et al., 2006). These non-GABAergic local axons are thought to establish an excitatory local microcircuit, producing robust excitatory postsynaptic activities in the neighboring small cells, even without peripheral inputs (Fortin and Champagnat, 1993;Senba, 1996, 1999). The present result suggests that a majority of the peripheral signals entering the dorsal cNTS are, to a large extent, modulated in the robust local microcircuit activity and conveyed to the PBL. In other words, the dense cNTS→PBL tract could transfer an ongoing activity of the local network in the dorsal NTS according to peripheral modulation.
The ventral cNTS is, in contrast, quite heterogeneous in many respects, as described above. The most conspicuous feature of this subregion is that it consists of heterogeneous populations of cells establishing a multiple macrocircuit that encompasses nearly all regions of the brain. The synaptic connection in this subregion of the cNTS is relatively sparse but involves every pattern of afferent and efferent synaptic connections: local GABAergic and non-GABAergic microcircuit inputs and outputs Senba, 1996, 1999;Yoshioka et al., 2006), the peripheral and central afferent inputs (Kalia and Sullivan, 1982;van der Kooy et al., 1984;Housley et al., 1987;Jasmin et al., 1997;Chan et al., 2000;Geerling et al., 2010), and multiple global outputs reaching all brain areas, including the circumventricular organs. The ventral cNTS is likely an interface between the local microcircuit and global macrocircuit.
The dorsal vagal complex is a composite structure of the cNTS and the dmnX, an aggregate of the parasympathetic preganglionic neurons. It is strategically interesting that dendrites of the dmnX neurons are arranged preferentially dorsally in the ventral cNTS (Shapiro and Miselis, 1985;Rinaman et al., 1989). The dorsal vagal complex, in this sense, consists of three layers, each of which is strategically intimate in functional and anatomically interrelated to other layers to perform a complicated computation of autonomic-affective reflexes for homeostatic control of the whole body (Zhang et al., 1998;Nieuwenhuys et al., 2008).

Whole-Brain-Ranging Global Unmyelinated Network
The global network originating from the cNTS was visualized by chromogen coloring following avidin-biotin-peroxidasecomplex (ABC) reaction with the transported anterograde tracer BDA. Since the ABC reagents are of relatively high molecular weight, sufficient penetration into myelinated axon fibers seems unlikely, even after Triton treatment of tissue sections. For visualization of myelinated axons, PHAL immunohistochemical staining is likely much less effective because one further step (the immunoglobulin reaction) is required. Indeed, in our anterograde-tracing section materials, no preganglionic axon bundles of the vagus nerve in the medulla were visualized, even when the dmnX cell bodies and dendrites were densely stained with injected BDA. These results indicate that axon fibers visualized by PHAL or BDA histochemistry are likely to be unmyelinated or thinly myelinated (Nieuwenhuys et al., 1982).
Descending axon systems from the hypothalamus (PVH and LH) were demonstrated to critically involve the cNTS by PHAL anterograde tracing studies (ter Horst et al., 1984;Luiten et al., 1985;Zheng et al., 1995;Geerling et al., 2010;Hahn and Swanson, 2010). The described tracts mainly coursed through the MFB and DLF in the diencephalon and midbrain. The courses of these pathways were similar to those ascending from the cNTS. In the pons and medulla oblongata, a third descending contingent that was not seen in the ascending tract from the cNTS emerged and continued from the DLF in the PAG and coursed ventrally in the raphe regions, reaching the lamina X around the central canal in the spinal cord. In the spinal cord, varicose axons were also seen in the lateral funiculus in addition to the dorsal horn. In contrast to the PB, which sends mostly ascending axons, PHAL studies of the midbrain showed that the PAG sends diverse descending axons that reach the spinal cord (Cameron et al., 1995a,b;Krout and Loewy, 2000). Both ascending and descending long-and mid-ranging pathways originating from different nuclei, including the PVH, LH, PAG and PB, overlap each other to a certain extent and cover an extremely wide range of brain regions to allow autonomic-affective behavioral control (Krout et al., 1998;Hahn and Swanson, 2010). This global macro-network also seems to be critically associated with monoaminergic and cholinergic diffuse projection systems that connect efferently to the whole brain by both synaptic and nonsynaptic neurotransmission and with the circumventricular organs that exert humoral and fluid control of homeostasis (Nieuwenhuys et al., 2008;Geerling et al., 2010).
Ascending axons from the spinal dorsal horn have been demonstrated to travel through the MFB and DLF to terminate in numerous telencephalic, diencephalic and brainstem structures, in addition to somatosensory thalamic nuclei (Cliffer et al., 1991;Giesler et al., 1994;Gauriau and Bernard, 2004;Willis and Coggeshall, 2004). In the brainstem, it was shown that varicose axons from the superficial spinal horn were moderately distributed in the PBL, PAG and other structures (Cliffer et al., 1991;Craig, 1995;Gauriau and Bernard, 2004). The distribution of axons and boutons in the brainstem originating from the superficial dorsal horn is very similar to that from the cNTS, as revealed by our present study. Taken together, these observations indicate that the autonomic-affective and parts of the somatosensory system likely share a similar diffuse and loose network. This network consists of mostly unmyelinated varicose axons that interact with common multiple brain regions located throughout the brain, possibly performing an extremely wide-ranging spectrum of hemostatic control tasks, including autonomic regulation, motivation, emotion, attention, arousal, learning, memory and sensory-motor-autonomic integration (Nieuwenhuys et al., 2008;Geerling et al., 2010).

An Interface Between the Local Microcircuit and Global Macrocircuit
Gateways from the periphery to this central global network are the cNTS (viscerosensory) and superficial layers of the spinal or trigeminal dorsal horn (somatosensory) that receive dense peripheral afferent axons consisting mostly of unmyelinated C and partly Aδ fibers (Jänig, 1996;Willis and Coggeshall, 2004;Nieuwenhuys et al., 2008). Nociceptor, chemosensitive and mechanosensitive visceral sensory information (such as smooth muscle spasms) are considered to be conveyed via both vagal and spinal (or trigeminal) afferents into the central nervous system (Jänig, 1996;Willis and Coggeshall, 2004). It follows that the same signals occurring in a certain specific visceral organ could enter into the central nervous system via two channels of the cNTS and superficial layers of the spinal or trigeminal dorsal horn. Understanding how the seemingly same information in the periphery is differentially processed and transferred within the central nervous system is essential. Several similarities between the cNTS and superficial layers of the spinal and trigeminal dorsal horn are evident: (1) both regions contain concentrated small neurons of both glutamatergic and GABAergic phenotypes (Willis and Coggeshall, 2004); (2) they establish dense local microcircuits in situ via local axon collaterals and transfer the peripheral signals to small or larger projection neurons localized in the neighboring layers (Willis and Coggeshall, 2004); and (3) the latter group of neurons seems to organize global macrocircuits. A prominent difference is the presence of a dense cNTS→PBL tract in the viscerosensory system, while small cells in the superficial layers of the spinal or trigeminal dorsal horn seem to send very few projection axons far beyond their own segment (Willis and Coggeshall, 2004). The differential architecture of the interface of local microcircuits and global macrocircuits between the cNTS and superficial layers of the spinal or trigeminal dorsal horn might underlie a huge difference in, for example, pain signal computation in a conscious state.

Functional Considerations and Concluding Remarks
Discussions of the possible functional significance of brain studies customarily reference electrophysiological or pharmacological articles related to specific anatomical sites or connections. For example, recent literatures have highlighted chemistry of NTS-PVH and NTS-LC connections (Affleck et al., 2012), central and peripheral chemoreception mechanism involving several brainstem regions such as NTS, LC, respiration-related areas and raphe (Kumar and Prabhakar, 2012;Nattie and Li, 2012;Guyenet, 2014;Lopes et al., 2016), and food intake/energy expenditure, appetite control by cholecystokinin concerned with NTS-PVH projection (Rinaman, 2010;D'Agostino et al., 2016). However, this is not the case in the present study, as the focus is not analytical or reductionist but rather a gestalt or holistic view of homeostatic control of the whole body to describe a possible underlying network system. Large networks of unmyelinated or thinly myelinated neurons were present to a greater extent than expected, exerting both conscious and unconscious functions including autonomic and non-autonomic activities in both the peripheral and central nervous systems. However, the significance of the overall activity, interaction, and architecture has not been fully addressed because such investigation may involve a large spectrum of relevant structures and functions. The cNTS might be one of the key stations serving as a sensory gateway from the periphery to the central macro-network and is expected to be critically involved in both somatic and visceral (autonomic) motor systems.