Comparative analysis of monoaminergic cerebrospinal fluid‐contacting cells in Osteichthyes (bony vertebrates)

Abstract Cerebrospinal fluid‐contacting (CSF‐c) cells containing monoamines such as dopamine (DA) and serotonin (5‐HT) occur in the periventricular zones of the hypothalamic region of most vertebrates except for placental mammals. Here we compare the organization of the CSF‐c cells in chicken, Xenopus, and zebrafish, by analyzing the expression of synthetic enzymes of DA and 5‐HT, respectively, tyrosine hydroxylase (TH) and tryptophan hydroxylase (TPH), and draw an evolutionary scenario for this cell population. Due to the lack of TH immunoreactivity in this region, the hypothalamic CSF‐c cells have been thought to take up DA from the ventricle instead of synthesizing it. We demonstrate that a second TH gene (TH2) is expressed in the CSF‐c cells of all the three species, suggesting that these cells do indeed synthetize DA. Furthermore, we found that many CSF‐c cells coexpress TH2 and TPH1 and contain both DA and 5‐HT, a dual neurotransmitter phenotype hitherto undescribed in the brain of any vertebrate. The similarities of CSF‐c cells in chicken, Xenopus, and zebrafish suggest that these characteristics are inherited from the common ancestor of the Osteichthyes. A significant difference between tetrapods and teleosts is that teleosts possess an additional CSF‐c cell population around the posterior recess (PR) that has emerged in specific groups of Actinopterygii. Our comparative analysis reveals that the hypothalamus in mammals and teleosts has evolved in a divergent manner: placental mammals have lost the monoaminergic CSF‐c cells, while teleosts have increased their relative number.


| I N T R O D U C T I O N
Cerebrospinal fluid-contacting (CSF-c) cells are a peculiar cell population with a bipolar morphology, one process protruding into the CSF.
DA and 5-HT are synthesized from the aromatic amino acids tyrosine and tryptophan, respectively, by rate-limiting enzymes, tyrosine hydroxylase (TH), and tryptophan hydroxylase (TPH) (reviewed in . TH and TPH belong to the same family of aromatic amino acid hydroxylase (AAAH). The corresponding metabolites, L-3,4-dihydroxyphenylalanine (L-DOPA) and 5-hydroxytryptophan wileyonlinelibrary.com/journal/cne the amino acid decarboxylase (AADC). Thus, the difference between DA-and 5-HT-synthesizing cells solely depends on the presence of either TH or TPH, which determines the neurotransmitter phenotype.
Our previous study showed that two TH-encoding genes, TH1 and TH2, are present in Osteichthyes, except in placental mammals that have secondarily lost the TH2 gene (Yamamoto, Ruuskanen, Wullimann, & Vernier, 2010). In the zebrafish CSF-c cell populations, th2 is much more abundantly expressed than th1, and the localization of the th2 1 /th1cells well matches that of so-called DA-accumulating cells (Yamamoto, Ruuskanen, Wullimann, & Vernier, 2011). Hypothalamic CSF-c cells in nonmammals has been long considered to accumulate DA instead of synthesizing it (Smeets & Gonzalez, 1990;Smeets & Reiner, 1994), because they are DA-immunopositive (DA 1 ) but THimmunonegative (TH -). We found that the commercially available TH antibodies do not detect TH2 protein in immunohistochemistry . These observations raise the possibility that the socalled DA-accumulating cells indeed synthetize DA using TH2 in nonmammalian vertebrates.
The current study addresses the evolution of the monoaminergic CSF-c cells, as well as the evolution of the surrounding brain area, focusing on the two paralogous genes of DA and 5-HT synthetic enzymes. Since the CSF-c cells are located along the ventricular wall, their organization largely depends on the configuration of the hypothalamic recess. We thus took into account the ventricular morphology in our analysis, as we recently did to study the regionalization of the optic recess region (ORR) .
We demonstrate that, as it is the case in zebrafish, TH2 is expressed in the CSF-c cells known as the "DA-accumulating cells" in chicken and Xenopus. Moreover, we found that many of these TH2 cells coexpress TPH1, also in all three species. Our study suggests that this dual phenotype in the monoaminergic CSF-c cells is inherited from the common ancestor of Osteichthyes. Nonetheless, a large difference is also observed between tetrapods and teleosts. The CSF-c cells in tetrapods are organized around one hypothalamic recess, while those in teleosts are organized around two hypothalamic recesses.

| Animals and tissue preparation
In this study we used three vertebrate species: chicken (Gallus gallus), Xenopus (Xenopus tropicalis), and zebrafish (Danio rerio). All experimental protocols and handling, use, and care of laboratory animals were conducted in compliance with the current normative standards and Zebrafish were kept in our own colony. Adult brains (3 months to 2 years old animals; n 5 30; both males and females) were used in all the experiments except for the whole brain imaging, in which 6-weekold zebrafish brains were used. Animals were either wild-type (AB) or ETvmat2:GFP, an enhancer trap transgenic line in which the reporter gene is inserted into the second intron of vesicular monoamine transporter 2 (vmat2) (Kawakami et al., 2004;Wen et al., 2008). Zebrafish were deeply anesthetized using MS222 diluted in fish water (0.2%).
Brains were then dissected and fixed in ice-cold 4% PFA in PBST over-

| Immunohistochemistry
Histological sections were rehydrated in ethanol gradient series and extensively washed in PBST at room temperature. Sections were incubated for 1 hr at room temperature with 4% normal goat serum (Bethyl Laboratories Inc., Montgomery, TX) in PBST containing 0.3-0.6% Triton X-100 (Sigma-Aldrich), and then incubated with primary antibodies ( Following the general formatting of gene symbols, chicken genes will be abbreviated with upper-case letters (e.g., TH1), while those in Xenopus and zebrafish will be abbreviated with lower-case letters (e.g., th1).
In order to refer to the homologous genes throughout the three species or throughout vertebrates, we will use upper-case (e.g., TH1) in this article.

| In situ hybridization
Fluorescent in situ hybridization was performed as described previously (see Fontaine et al., 2013) with modifications. Briefly, brain sections After extensive washes, samples were counterstained with DAPI.
In some cases, in situ hybridization was coupled to immunohistochemistry, as described above.

| Image acquisition
Brain sections were imaged using a Zeiss LSM700 laser scanning con-

| 3D whole brain reconstruction
The 3D-visualizations of zebrafish specimens were generated using amira6 3D visualization and reconstruction software platform (FEI, Hillsboro, OR) on a HP Z720 workstation.
In the process of direct volume rendering, the brightness value of each voxel (volume pixel) is converted into an opacity value. Bright voxels therefore are rendered very opaque, while dark voxels are ren- To visualize the relationship of the vmat2-GFP expression pattern (green) with the ventricle system (magenta), we applied manual-and threshold-based segmentation and surface reconstruction. While we were able to define the volume of the vmat2-GFP expression pattern using amira's build-in thresholding and surface-simplification algorithms, the ventricle system was segmented manually on the basis of our reference staining (DiI). Since DiI in our application stains all the membranes, we were able to reconstruct the morphology of the ventricles by following the high-density staining (presumably tight junctions) of the cells facing the lumen of the ventricle as well as the void of the ventricles itself. Subsequently the manual segmentation was refined by thresholding and supervised intensity-based region growing.
Surfaces were computed (SurfaceGen) on the basis of the defined volumes. The segmentation and resulting surfaces were smoothed and    Figure S2).

| TH2 is expressed in the hypothalamic CSF-c cells in chicken and Xenopus
Previous results have shown that TH2 is found in the genome of various nonmammalian vertebrate species .
In agreement with this hypothesis, we show that TH2 transcripts are found in CSF-c cells in the chicken hypothalamus (Figure 4a,c,f). In chicken, TH2 is expressed in the CSF-c cells all along the hypothalamic recess, rostrally in the area identified as the PVO and extending caudally toward the infundibulum (Inf) (Figure 4a,c). The CSF-c cells are devoid of TH1 expression (Figure 4b,d,g) and of TH immunoreactivity (Figure 4e,h).
We observed low levels of TH2 transcripts in cells located dorsolaterally to the CSF-c cells (Figure 5a; asterisk). This cell population is described as the A11 dopaminergic cells in pigeon  Figures 1c,4b-h,5b; asterisks). Besides, weak TH2 labeling is observed In previous works , we showed that the zebrafish th2 is expressed in CSF-c cell populations in the hypothalamic ventricular areas. We here show that tph1a, one of the two tph1 paralogs duplicated in the teleost lineage (Bellipanni et al., 2002;Gaspar & Lillesaar, 2012), is also expressed in these CSF-c cell populations (Figure 8), some of them colocalizing with th2 transcripts (Figure   8; arrows).  (Kaslin & Panula, 2001).
In agreement with the coexpression of th2 and tph1a genes, we observed that some CSF-c cells contain both DA and 5-HT in zebrafish
In our previous publications describing the th2 1 DA cell populations in zebrafish , we used the nomenclature proposed by Rink and Wullimann (2001), in which the term  Neary & Northcutt, 1983), while the PR is only present in some groups of Actinopterygii including teleosts and nonteleost fishes such as gar (Parent & Northcutt, 1982) and sturgeon (Kotrschal et al., 1985). Since PR is neither found in Polypterus (  Optical sectioning by confocal microscopy shows that TH2 (d2) and TPH1 (d3) are expressed in the soma (arrowheads) and in the processes of CSF-contacting cells (arrow). Orthogonal views of optical sections confirm the overlap of TH2 and TPH1 signals. In Xenopus, th2 (e, i; magenta) and tph1 (f, j; green) are expressed both in the rostromedial (e-h) and the caudolateral (i-l) parts of the PVO. Some CSF-c cells coexpress both genes (g, k). The coexpression is confirmed by orthogonal views of optical sections (h1) of the area depicted by a dashed rectangle in (g), showing that th2 (h2) and tph1 (h3) are colocalized in the same cells (arrowheads). D 5 dorsal; L 5 lateral; LR 5 lateral recess; v 5 ventricle. Scale bar 5 100 mm in (a) (applies to a, b, c) and in (e) (applies to e, f, g); 50 mm in (i) (applies to i, j, k, l) In any case, the abundance of monoaminergic CSF-c cells is reduced in the tetrapod lineage as compared to Actinopterygii, with the extreme case of complete loss in mammals. The expression of TH1 in the CSF-c cells is also reduced in tetrapods, with no TH1 expression in the avian CSF-c cells. Since TH 1 (which is presumably TH1 1 ) CSF-c cells are abundant in many teleosts and Chondrichthyes (Smeets & Reiner, 1994), it is likely that the ancestor of jawed vertebrates possessed abundant CSF-c cells which express both TH1 and TH2.
Although the functional significance is not clear yet, it is likely that they played an important role in the brain of ancestral vertebrates.

| DA/5-HT dual phenotype of the hypothalamic CSF-c cells
It has been long known there are DA 1 cells in the CSF-c cells around the hypothalamic area, although these cells are not TH immunoreac-tive. Thus, these DA 1 cells were considered to be DA-accumulating cells (Smeets & Reiner, 1994). As a matter of fact, the presence of the TH2 transcripts in these cells strongly suggests that DA is synthesized by the TH2 enzyme in this cell population.
Moreover, we found the CSF-c cells containing both DA and 5-HT. The localization of th2 in 5-HT 1 cells devoid of TH immunoreactivity prompted some authors to propose that TH2 is functionally equivalent to TPH, synthesizing 5-HT instead of DA (Ren, Li, Zhong, & Lin, 2013). In their hypothesis, TH1 expressing cells synthesize DA while TH2 expressing cells synthesize 5-HT in zebrafish. However, later studies do not support this idea (McPherson et al., 2016;Semenova, Chen, Zhao, Rauvala, & Panula, 2014). Based on our results revealing the presence of TH2 (but not necessarily TH1) in the DA 1 CFS-c cells in three vertebrate groups, it is more likely that TH2 is able to synthesize DA in this area.  (Kang et al., 2007(Kang et al., , 2010.
Similarly to the case of the TH2 gene, which has been long ignored due to its loss in the mammalian genome, the importance of TPH1 in the brain has been also underestimated. This is because TPH1 expression is found only in the pineal gland that synthesizes MEL in mammals.
Studies on nonmammalian species suggest that the presence of dual neurotransmitter phenotypes of catecholamines and indolamines in the forebrain is common in Osteichthyes except for mammals.

| Physiological roles of the monoaminergic CSF-c cells
A study in quail revealed that 5-HT cells in the avian PVO express opsin 5, a UV sensitive opsin, suggesting that the PVO is involved in , and (f) respectively. Both monoamines were observed in a few cell bodies (arrow in f) and endfeet (arrowhead in f). In teleosts, some of the DA 1 CSF-c cells (g) are also immunoreactive for TH (h; orange). Higher magnification images of the PVO are shown in (i-l). In CSF-c cells, intense DA immunoreactivity is present in the cell soma (j; arrows), processes, and the endfeet contacting the ventricle (k; arrowheads). In contrast, intense TH immunoreactivity is mostly observed in the soma and processes (l; arrows), but not in the endfeet. D 5 dorsal; L 5 lateral; v 5 ventricle. Scale bar 5 50 mm in (a) (applies to a, b, c); 200 mm in (g) (applies to g, h); 100 mm in (i) (applies to i, j, k, l)  applies to a, b, c, d); 50 mm in (f) (applies to f, g, h, i); 50 mm in (k) (applies to k, l, m, n) deep brain photoreception that regulates seasonal reproduction (Nakane et al., 2010). Since many TPH1 1 cells coexpress TH2 in the chicken brain, it is possible that the dual DA/5-HT phenotype plays a significant role in this function. Indeed, it has long been proposed that hypothalamic CSF-c cells play a role in deep brain photoreception to synchronize periodicity in nonmammals (Vigh et al., 2002). The loss of CSF-c cells may thus be correlated to the loss of deep brain photoreception in the mammalian forebrain. Osteichthyes. Thus, these two animal groups must be compared with caution.
Based on a one-to-one comparison of zebrafish and mammalian DA cell populations, it has been proposed that DA cells in the PVO (PVOa) are homologous to the A12/A14 DA population (Filippi, Mueller, & Driever, 2014). This is unlikely, since mammalian A12/A14 do not comprise CSF-c cells. Furthermore hypothalamic A12 cells in amniotes project to the median eminence and regulate pituitary functions via portal blood system, while the teleost pituitary receives direct Some comparable monoaminergic cell groups are plotted on schematic drawings of brain sections of mouse, chicken, Xenopus, and zebrafish. The sagittal plane is close to the midline (rostral to the left). The diamonds represent CSF-c cells, and the circles represent non-CSF-c cells. TH2/TPH1-expressing CSF-c cells (red diamonds) are commonly found along the hypothalamic recess throughout vertebrates, while placental mammals have lost the CSF-c cells in the hypothalamic region. Below the sagittal section of chicken, Xenopus, and zebrafish, frontal sections around the hypothalamic recess (corresponding to the level of gray lines; a,b,c) are shown. In chicken and Xenopus, there is only one hypothalamic recess (named either 3rd ventricle or lateral recess; LR), while in zebrafish, there is an additional recess (posterior recess; PR). Note that in Xenopus, the caudally located CSF-c cells are not observable in the sagittal section close to the midline, because it is located at the lateral end of the LR. The A11-like TH1-expressing cell population (blue dots; non-CSF-c cells) projecting to the spinal cord is commonly found dorsolateral to the CSF-c cells in all the vertebrate groups. More caudally, the A9/10-like DA cell population projecting to the telencephalon (green dots) is found in tetrapods, while this cell population is lacking in teleosts. L 5 lateral recess; M 5 mesencephalon; P 5 prosencephalon; PR 5 posterior recess; R 5 rhombencephalon DA inputs from the anterior optic recess region (ORR) (Fontaine et al., 2015). Thus, it is possible that teleosts do not have DA cells homologous to the mammalian A12 group.
Dorsolateral to the CSF-c TH2 1 /TPH1 1 cells, non-CSF-c TH1 1 cells are consistently found throughout Osteichthyes (Figure 12; blue circles). They are characterized by a distal projection to the spinal cord ( Figure 12; blue arrows; Medina, Puelles, & Smeets, 1994;S anchez-Camacho, Marín, Smeets, Donkelaar, Gonz alez, 2001;Tay, Ronneberger, Ryu, Nitschke, & Driever, 2011) as the mammalian A11 cell group does (Bj€ orklund & Skagerberg, 1979;Skagerberg, Bj€ orklund, Lindvall, & Schmidt, 1982). Nonetheless the homology of this cell population needs to be determined carefully, because there are some discrepancies. First, in teleosts, which lack mesencephalic DA cells, these non-CSF-c TH1 cells were also suggested to be similar to A9/A10, based on their projection to the subpallium (Figure 12; green arrow in zebrafish; Rink & Wullimann, 2001). Second, the regional identity of the corresponding area is not clear. In amniotes it is often considered to be the hypothalamus (which is a part of the secondary prosencephalon), while in teleosts it is considered to be the posterior tuberculum (which is a part of the diencephalon). A recent publication in zebrafish (Biran, Tahor, Wircer, & Levkowitz, 2015) proposes to reconsider this part of the posterior tuberculum as a part of the hypothalamus, as defined in mammals (Puelles & Rubenstein, 1993. A boundary between the hypothalamus and the diencephalon remains to be more precisely defined.
The current model of the hypothalamic organization is largely based on mammalian studies. However, our study demonstrates that the mammalian hypothalamus is an exceptional case lacking a whole population of CSF-c cells. To better understand the general organization in vertebrates, models of hypothalamic organization need to be refined by including a larger range of vertebrate species.

ACKNOWLEDGMENTS
We thank Ingrid Colin, Charlotte Bureau, Manon Thomas, and S ebastien Bedu, for technical support in the experiments. We also thank members of TEFOR Core Facility, specially Elodie de Job and Laurie Riviere for the whole brain tissue cleaning. We thank Muriel Perron's team (in particular Albert Chesneau for his kind help) for providing Xenopus and Sophie Creuzet for chicken embryos, as well as Laure Bally-Cuif's team for zebrafish tph1a probe and practical support. We also thank Serina Khater and Catherine Pasqualini for English corrections, and Lydie Collet for her help with Figure 5.
Finally we thank Drs. Florian Razy-Krajka and Christina Lillesaar for discussion at an early stage of the work.