Arginine Vasopressin-Containing Neurons of the Suprachiasmatic Nucleus Project to CSF

Visual Abstract


Introduction
The suprachiasmatic nucleus (SCN) of the hypothalamus, comprised of ;20,000 neurons, functions as a circadian clock, coordinating and synchronizing daily rhythms in numerous bodily functions including metabolism, physiology, behavior, and hormone secretion. These neurons contain a substantial array of peptides (Antle and Silver, 2005), among which the best studied synthesize arginine vasopressin (AVP; van den Pol and Tsujimoto, 1985;Kalsbeek et al., 2010). AVP-expressing neurons form the major population of neurons in the SCN and are estimated to constitute 37% of the nucleus (Moore, 2013). Through most of its rostrocaudal extent, AVP perikarya lie within the shell region (Moore, 2013;Lokshin et al., 2015;Varadarajan et al., 2018). It is widely accepted that AVP neurons delineate the boundaries of the SCN and it is thought that AVP neurons provide the major output signal from the SCN (Cassone et al., 1988;Daikoku et al., 1992;Dai et al., 1997;Abrahamson and Moore, 2001;Ramanathan et al., 2006;Moore, 2013).

Efferent connections of the SCN
Output from the SCN influences visceral function (Ueyama et al., 1999), REM sleep (Lee et al., 2009), and timing of the LH surge (Williams et al., 2011). SCN efferents have been extensively studied (Moore, 2013;Morin, 2013) and more specifically, the efferents of AVP neurons have also been explored (LeSauter and Silver, 1998;Kalsbeek et al., 2010). The medial preoptic area, the periventricular and sub-paraventricular nucleus, dorsomedial hypothalamus and paraventricular nucleus of the thalamus have been identified as targets of AVP neurons. However, the extent to which all of these connections originate from SCN or AVP neurons of other origins is not known.

AVP expression in the SCN and CSF is rhythmic
The transcriptional machinery of the core clockwork directly regulates rhythmic expression of AVP in the SCN shell, but not in other AVP-producing neurons (Jin et al., 1999;Silver et al., 1999). The concentration of AVP in the CSF, thought to be of SCN origin, fluctuates daily with a peak in the morning (Schwartz and Reppert, 1985;Söderstein et al., 1985;Jolkkonen et al., 1988;Stark and Daniel, 1989;Kalsbeek et al., 2010). Samples from rat CSF and blood definitively demonstrated circadian rhythms in AVP levels in the CSF but not in the blood (Schwartz et al., 1983). This work established the importance of the SCN in producing circadian rhythms in CSF AVP as SCN lesions abolished rhythmicity and reduced measurable levels of the peptide but left open the possibility that neurons of the SCN may trigger AVP release from some other neural sites. Also unknown was the route whereby AVP of SCN origin might reach the CSF, a question in the present study.

The present study
While rhythmicity in CSF AVP levels is well established, the question of how AVP of SCN origin might reach the CSF led to the present study. Much of the work on the SCN has been done in coronal sections and this focus neglects the rostral and caudal-most aspects of the nucleus, both of which are rich in AVP containing neurons (Riddle et al., 2017;Varadarajan et al., 2018). Additionally, in coronal sections it is difficult to visualize the close and extended contact of the SCN with the wall of the third ventricle. Here, we first sought to better understand the full extent of AVP neurons in the mouse SCN. This work pointed to dense AVP fibers projecting toward the CSF, a previously unidentified target. To better characterize their axonal and dendritic processes, we filled AVP neurons with biocytin and characterized their afferent and efferent processes using confocal microscopy. To assess communication of AVP neurons with CSF, immunochemistry was used to label the ventricular wall and the tracer cholera toxin b subunit (CTb ), was injected into the lateral ventricle and SCN AVP cells bearing the tracer were identified.

Animals and housing
Two strains of mice were used to optimize detection of markers of interest with our multi-label protocols. For detection of PER2, we used adult C57BL/NJ mice (The Jackson Laboratory), for cell filling with biocytin, laboratory bred 15-d-old Per1-GFP mice (Kuhlman et al., 2000) were used. Mice were group-housed (four per cage) in translucent propylene cages (48 Â 27 Â 20 cm) in a colony room with a 12/12 h light/dark cycle at 21 6 2°C and provided with free access to food and water. All procedures were approved by the Columbia University Institutional Animal Care and Use Committee.

Localization of AVP neurons and efferents
Mice were anesthetized with 100 mg/kg of ketamine and 10 mg/kg of xylazine and perfused transcardially with 50 ml of saline followed by 75 ml of 4% paraformaldehyde. Brains were postfixed overnight at 4°C and then cryoprotected in 20% sucrose in 0.1 M phosphate buffer 0.9% saline (PBS). For immunochemistry (ICC), sections (50 mm) were cut in the sagittal, horizontal or coronal planes on a cryostat (Microm HM 500 M) and collected into PBS. The brain sections were processed as free-floating sections in 24-well plates. For single label ICC, sections were washed in PBS 1 0.1% Triton X-100 (PBST) and then blocked with normal donkey serum for 1 h on a shaker at room temperature (RT). Sections were placed in a 0.3% PBST solution containing rabbit anti-AVP antibody at 4°C for 48 h. Sections were washed (3Â) and incubated for 2 h at RT in 0.3% PBST containing the appropriate secondary antibody. Sections were washed in PB, mounted, dehydrated in ethanol solutions, cleared in CitriSolve (Fisher Scientific) and coverslipped with Krystalon (EM Diagnostics). For double label ICC, the same procedure was used except primary antibodies were either mouse anti-AVP and rabbit anti-PER2; mouse anti-AVP and rabbit anti-S100b or rabbit anti-AVP and goat anti-CTb . The appropriate fluorescent secondaries were used (Table 1).

AVP in relation to the ependymal wall
To assess AVP fibers in supra-and sub-ependymal zones of the third ventricle, we observed AVP and S100b labeled sections (n = 5 female mice). S100b stains ependymal cells (Didier et al., 1986;Carlén et al., 2009;Lavado and Oliver, 2011).

Biocytin filling of patched AVP cells to determine morphology of AVP neurons
To determine the morphology of AVP neurons and their axonal and dendritic projections, Per1-GFP mice were sacrificed by cervical dislocation between zeitgeber time (ZT)12 and ZT15 (when PER1-GFP expression is high), and cells were patched and filled with biocytin as described for GRP-GFP cells (Drouyer et al., 2010). For ICC of sections bearing biocytin-filled cells tissue was incubated overnight at 4°C with Cy2 streptavidin (Jackson ImmunoResearch) in PBST, then processed as above except primary antibodies were guinea pig anti-AVP and rabbit anti-GFP, stained with the appropriate secondary antibodies.

Analysis of biocytin filling
Each filled cell was examined throughout the z stack using confocal microscopy to assess whether it was single, double or triple labeled for biocytin (green), PER1-GFP (magenta), and AVP (blue). Of 73 cells filled with biocytin in nine animals, 53 were PER1-GFP positive and 20 were PER1-GFP negative. Of the 53 PER 1-GFP positive cells, 30 were AVP1 and 23 were AVP-. The 20 PER1-GFP negative cells were also AVP negative. Axons originate from the soma of AVP neurons or from a proximal dendrite and are very fine and beady. Dendrites are larger and possess spines, swellings, and/or appendages (described in van den Pol, 1980).

Measuring cell size
In the initial study, confocal microscopy was used to measure the perimeter of distinct populations of AVP neurons, enabling comparison with previous studies of this population (Campos et al., 2014;Moore et al., 2002). Images were taken from the rostral-most and caudalmost aspects, at the largest extent of the SCN, the midregion in sagittal sections (two sections/SCN, n = 8 SCNs). The cells were measured at their greatest extent, where a distinct nucleus was seen on the 1 mm z-axis optical section. The perimeter was determined using the Annotations and Measurements tool on NIS-Elements Advance Research software. Seven to 21 cells were measured/region for each brain section (total N = 76 rostral, 89 caudal cells). For volumetric measures, a new set of confocal images of AVP stained SCNs were imported into the imaging software Aivia (version 9, DRVISION Technologies LLC). The segmentation output in the pixel classifier tool allows for visualization of the cells in 3D and was used to measure volume of individual rostral (N = 88) and caudal (N = 101) cells.

CTb tracer, lateral ventricle injection site
While CTb is an anterograde and retrograde tracer, it behaves primarily as a retrograde tracer at the doses and sites described here (see Leak and Moore, 2012 for extended discussion and explanation). CalB-GFP mice were anesthetized with 100 mg/kg of ketamine and 10 mg/kg of xylazine at ZT14 and prepared for aseptic surgery. The head was shaved and the mouse was positioned in a stereotaxic apparatus (David Kopf Instruments). Injections were aimed at the lateral ventricle (AP 11.1, ML 10.7 from bregma, DV À3.0 from the top of the skull). Pressure injections of 4 ml of CTb (1% low salt solution, List Biologicals) were made with a 10-ml syringe (Hamilton Co) at a rate of 2 ml/min. The cannula was left in the brain at the end of the infusion for an additional 2 min and removed slowly over 1 min to minimize tracer diffusion upon needle withdrawal. The animals were killed 7 d following surgery. Mice (n = 8) were injected with CTb . The injection penetrated the lateral septum in three brains but was strictly localized to the lateral ventricle in five injections (Fig. 1).

Microscopy
Immunofluorescent labeling was examined on a Nikon Eclipse E800 (Morrell Instruments) microscope using the filters 480 6 40 nm for Cy2, 560 6 40 nm for Cy3 and 620 6 40 nm for Cy5. Images were captured with a USB-3 DMK33UX174 camera (Morrell Instruments) using the Nikon NIS-Elements Basic Research software. To identify AVP dendrites and axons in biocytin-filled cells, and to determine their relation to the CSF, images were taken on a Nikon Eclipse Ti2E confocal microscope (Nikon) equipped with a LUNV Laser unit and an argon-krypton laser using the excitation wavelengths 488 nm for Cy2, 561 nm for Cy3, and 638 nm for Cy5. NIS-Elements Advanced Research software was used to visualize the zstacks. Confocal images of sections from CTb -injected mice were analyzed to determine the colocalization of CTb with AVP. The images were imported into Photoshop (Adobe Photoshop CS). Single labeled CTb and AVP and double labeled cells were counted by two experimenters (agreement: 93 6 2%).

Image analysis
For visualization of cells and/or fibers, we used Photoshop. To recover lost details in dark and bright areas of the image Shadows and Highlights were adjusted in the Image-Adjustment-Levels dialogue box in Photoshop, by dragging the black and white Input Levels to the edge of the first group of pixels on either end of the histogram (Figs. 4, 5). The function of Shadows and Highlights is to brighten areas of shadow and darken areas of brightness based on surrounding pixels; it does not simply lighten or darken an image (Adobe Photoshop CS6). For color images, Shadows and Highlights were manipulated for each color separately. Grayscale images were inverted before these visual optimizations.

Statistical analyses
Data for neuronal perimeter and volume of AVP neurons were measured using both statistical inference and estimation statistics. For statistical inference, we used Student's t test for normally distributed data with equal variance, Welch's t test for normally distributed data with unequal variance and Mann-Whitney U test for data lacking normal distribution. For these analyses, outliers were removed (for perimeter, 6 rostral and 4 caudal; for volume, 7 rostral and 2 caudal) using the Tukey method [quartile 1 and 3 6 (1.5*interquartile range)]. For both perimeter and volume distribution normality was analyzed by the Kolmogorov-Smirnov test and equality of variances was calculated by F test.

AVP in SCN
Because AVP-containing neurons are often used to delineate the extent of the SCN, we investigated the accuracy of this protocol by comparing the area of PER2 expression at its peak (ZT12-ZT14) to that of AVP-ir. In sagittal sections, we were surprised to find that at mid and lateral levels of the nucleus, the area of PER2 expression extends well beyond that delineated by AVP (Fig. 2). Most medially PER2 expression does correspond to the area delineated by AVP neurons (data not shown).
As previous studies suggest a relationship between the size of SCN neurons and their location within the nucleus (Campos et al., 2014), we also examined the size of AVP neurons in the rostral versus the caudal SCN measuring both perimeter and volume. The results indicate that neurons lying caudally are larger than rostral ones for both measures and this held true whether statistical inference or estimation statistics were applied (Fig. 3).
The precise localization of AVP neurons in the main body of the SCN is shown in seven serial sagittal sections (50 mm) through the extent of the SCN from the medialmost to the lateral-most aspect, in two animals sectioned at slightly different angles. Additionally, a substantial rostral prong is shown in both sagittal and horizontal views (Fig. 4).   While the SCN is embedded within the hypothalamus for most of its extent, most medially, it is separated from the rest of the hypothalamus by the third ventricle (Fig. 4, animals #1 and #2, images 1 and 2). At its dorsal aspect the SCN abuts the third ventricle, most evident when the section is cut perfectly parallel to the midline (Fig. 4, animal #1, images 1 and 2). Here the medial SCN is located right below the third ventricle. At the mid-level (animal #1, images 3-5), the rostral pole of the SCN abuts the third ventricle and the main body is caudal to the ventricle. Laterally (animal #1, image 7), the SCN lies entirely within the hypothalamic parenchyma and is no longer in contact with the third ventricle. In animal 2, in sections cut at a different angle, the third ventricle is seen medially (images 1-3), but no longer visible at images 4-7.
At the medial aspect of the SCN, AVP neurons occur throughout the SCN (Fig. 4A, animal #1, image 1; animal #2, images 1 and 2), and these extend in a rostral prong that originates in the ventral part of the nucleus and lies below the third ventricle, just above the optic chiasm (sagittal view in Fig. 4A, animal #1, image 1, animal #2, images 1 and 2; and horizontal views in Fig. 4B, animals #3 and #4, black arrows). The rostral prong extends ;250 mm rostrocaudally, 20-30 mm dorsoventrally, and ;80 mm mediolaterally. At the mid-SCN at its greatest extent in the sagittal view, AVP neurons lie rostrally and caudally, but are sparse in the core and in the mid-dorsal aspect (animal #1, image 5; animal #2, images 5,6; see also horizontal view animal #3 {white arrows}). In addition, AVP neurons are present along the base of the SCN (Fig. 4, both animals, image 4; animal #2, image 5, arrowheads). At the lateral aspect, the AVP neurons are densely packed (both animals, image 7).

Relation of AVP fibers to ependymal wall and CSF in coronal and sagittal sections
In coronal sections, at its greatest extent, the SCN is rich in AVP neurons (Fig. 5A), and many dendrites and axons course toward the CSF (Fig. 5B). In sagittal sections, projections of AVP neurons travel rostrally toward the third ventricle (Fig. 5C, arrow). Close to the midline, the SCN lies directly above the optic chiasm and its dorsal aspect abuts the third ventricle (Fig. 5D). Here, AVP fibers, appear to cross the ependymal wall to reach the CSF in the third ventricle.

Morphology of axons and dendrites of AVP neurons
To identify the axons and dendrites of AVP neurons, we patched and filled SCN neurons with biocytin. Of the 30 AVP1 neurons studied, 22 had distinguishable axons and dendrites. As previously described (van den Pol, 1980;Abrahamson and Moore, 2001), axonal processes of SCN origin are very thin and exhibit bulbous, irregularly spaced varicosities along their entire length (Fig. 6). Dendrites are larger and possess spines, swellings and/or appendages.

SCN dendrites and/or axons crossing the ependymal wall
To determine the terminal regions of SCN AVP neurons, we labeled sagittal sections for AVP and S100b , an ependymal wall marker. Some processes appear to terminate subependymaly while others cross into the third ventricle. Both AVP axons (Fig. 7) and dendrites (Fig. 8) extend through the ventricular wall both at the level of the SCN (Figs. 7A-C, 8A,C) or near the SCN (Figs. 7D, 8B,D). Terminal boutons can be seen within the third ventricle (Fig. 7B',C').

CTb tracer in lateral ventricle labels SCN neurons
As noted by Leak and Moore (2012): "The CSF contains neuroactive substances that affect brain function and range in size from small molecule transmitters to peptides and large proteins. CSF-contacting neurons are a well-known and universal feature of non-mammalian vertebrates..." Given the evidence of AVP1 fibers directed to the CSF, we next examined whether AVP cells would uptake the tracer CTb from the ventricle. For this, we injected CTb into the lateral ventricle. Figure 9 shows the cells that took up the CTb tracer and colocalized with AVP (inset) in the mid SCN. The distribution pattern of CTb in the SCN was similar in all five animals where injection was restricted to the lateral ventricle.
Of the 4122 AVP and 2488 CTb cells evaluated, 15.3 6 0.9% AVP were co-labeled with CTb and 25.5 6 1.9% CTb cells were co-labeled with AVP. The results point to extensive input from CSF to SCN neurons bearing AVP.

CTb in SCN medial to lateral
Cells containing CTb are located mostly within the SCN shell where AVP cells are present (Fig. 10). There are a few CTb cells in the very medial SCN (image 1). They are located dorsally in the medial SCN (image 2). In the mid and lateral SCN, they are located mostly within the dorsomedial region and are sparse in the SCN core  Fig. 4) showing many fibers crossing the ventricular wall (z-axis: 2.6 mm). A', Image showing the two branches of an axon reaching the CSF, caudal to the SCN (arrows; z-axis: 1 mm). B, C, Sections (at level 2 in Fig. 4) show AVP fibers extending dorsally from the SCN, crossing the ependymal wall (z-axis: B, 2.6 mm; C, 3.5 mm). B', C', Images showing axons crossing the ependymal wall with terminal boutons ending in the third ventricle (arrows; z-axis, B', C', 1.6 mm). D, Section showing the SCN and the third ventricle (z-axis: 1 mm). D', High-magnification images of the area dorsal to the SCN showing an AVP axon coursing on the ventricular side of the ependymal wall (z-axis: 0.5 mm). SCN: suprachiasmatic nucleus; 3V: third ventricle.
(images 5-7). They are also numerous in the dorsal area where AVP is sparser (image 5).

AVP fibers reach the CSF
A major finding of the present study is that SCN AVP neurons course toward and even terminate in the fluid space of the third ventricle. This communication is most obvious in sagittal sections through the SCN, especially near the midline. And while AVP neurons have often been used to demarcate the outer borders of the SCN (Vandesande et al., 1975;Sofroniew and Weindl, 1980;Card and Moore, 1984;van den Pol and Tsujimoto, 1985), the sagittal view makes it clear that there are rhythmically expressing PER2 neurons that lie beyond Taken together the detailed description of AVP neurons in the SCN make three distinct points: AVP cells are at the base of the core region of the SCN (Fig. 4). Dorsomedial region of the SCN shell is devoid of AVP neurons, AVP neurons extend rostrally along the midline along a narrow protrusion that originates in the medial-most aspect of the ventral SCN. While each of these details is modest, taken together they can correct incomplete or erroneous anatomic information in previous less detailed work and can provide the groundwork for fine-tuning future studies. Among the questions raised is the nature and function of rhythmic SCN neurons that lie around the outer borders of the SCN and whose existence was not previously realized.
As noted in the Introduction, efferent projections of the SCN and more specifically, efferents of AVP neurons of the SCN have been extensively studied (Berk and Finkelstein, 1981;Stephan et al., 1981;Hoorneman and Buijs, 1982;Kalsbeek et al., 1993;Morin, 1994;Teclemariam-Mesbah et al., 1999;Leak and Moore, 2001;Moore and Silver, 1998). Although there are some differences among species (Morin, 1994), AVP is a major component of the SCN in most species that have been studied (Morin and Allen, 2006). Furthermore, an important contribution of SCN to rhythmic AVP levels in CSF has been known since the mid-1980s, although evidence of the route of communication was lacking. To our knowledge however, there are no prior reports of SCN processes reaching the CSF, though there is light and electron microscopic evidence of (unidentified) somal processes inserted in the ependymal layer or directly contacting the CSF space (Xiao et al., 2005). This may be a consequence of the difficulty of detecting the very fine SCN fibers that course in a rostro-caudal direction (Figs. 4, 5) especially when they lie right near the midline. The results are consistent with evidence of volume transmission by the SCN and with prior studies. Specifically, paracrine signaling with a role for AVP, has been demonstrated in culture preparations of the SCN [Honma, 2020 (see Fig. 5), Maywood et al., 2011;Tokuda et al., 2018]. Furthermore, SCN grafts encapsulated in a polymer material can restore circadian activity rhythms through volume transmission (Silver et al., 1996). Finally, the SCN is densely labeled with CSF injections of .20 nl of a tract tracer CTb (Leak and Moore, 2012). More specifically, they showed that within 2-10 d of injection, the CTb label appears in specific groups of neuronal perikarya in the telencephalon, thalamus, hypothalamus and brainstem, many at  a considerable distance from the ventricles. They hypothesize that these observations support the view that ventricular CSF is a significant channel for volume transmission and identify those brain regions most likely to be involved in this process.

AVP neurotransmitters
The role of AVP in rhythmicity has been puzzling in view of the fact that AVP-deficient Brattleboro rats display attenuated but normal circadian rhythms (Groblewski et al., 1981;Brown and Nunez, 1989;Kalsbeek et al., 2010). AVPcontaining neurons express GABA and prokineticin 2 (Masumoto et al., 2006;Welsh et al., 2010) pointing to the possibility that these neurons play a broader role in the circadian timing system than does the AVP protein itself. Mice in which the essential core clock gene Bmal1 was deleted in AVP neurons (Avp-Bmal1 À/À mice), have impaired circadian activity rhythms (Mieda et al., 2015) and a gradually expanded interval activity time on being moved from LD to DD. Bmal1 restoration in AVP neurons of the SCN using a recombinant adeno-associated virus vector reversed the circadian impairment of Avp-Bmal1 À/À mice. The results were interpreted to mean that the circadian oscillation persists at the cellular level, but that mutual coupling among SCN neurons that regulate activity onset and offset was impaired.

Rhythms in SCN-regulated functions
As noted by van den Pol (2012) peptide release is difficult to study, as it is seen independent of synaptic specializations, in dendrites, axonal boutons and axon shafts. This is consistent with the possibility that axons and dendrites described here (Figs. 7,8) both contribute to the rhythms of AVP in CSF. AVP has a role in the regulation of hydration, blood pressure, body temperature, corticotropin release, and has been implicated in depression, memory, social and sociosexual behavior and in sex differences (for review, see de Vries and Miller, 1998;Kalsbeek et al., 2010). The stress system in the human brain has been implicated in depression and neurodegeneration (Swaab et al., 2005). The relationship of CSF levels of AVP to plasma levels is of interest in studies of autism, suicide (Brunner et al., 2002) and personality disorders (Coccaro et al., 1998;Oztan et al., 2018) and the two compartments seem to be separately regulated (Kagerbauer et al., 2013). Much remains to be investigated as understanding the function of this compartment in the context of circadian rhythms requires unravelling a sequence of steps involved in volume transmission. AVP is a phylogenetically ancient molecule of immense versatility with very diverse functions in both the brain and in the periphery (Zelena and Engelmann, 2018). Within the SCN this peptide is important not only as an output signal but also in maintaining synchrony of the SCN network (Shan et al., 2020). A broad view of its potential was suggested decades ago when Gillette and Reppert (Gillette and Reppert, 1987) hypothesized that "Oscillations in AVP secretion and neuronal firing rate potentially represent separate modes by which the SCN transmit time information to other brain regions. The electrical oscillation may relay time-of-day by efferent neuronal output along a limited number of specialized circuits. The AVP secretory rhythm potentially modulates daily rhythm in physiology by a generalized neurohormonal effect." The present work directs attention to the latter possibility.