It was noted that “the circumventricular organs (CVO) are peculiar brain structures that are located in the walls and often protrude in the lumen of the third and fourth ventricles (Hofer, 1958)” and that “no clear agreement can be found in the literature on the number of these organs in mammals” (Duvernoy and Risold, 2007). The adult CVOs are highly vascularized, but the development of vascularisation is not studied in detail. Unlike that in the rest of the brain their blood-brain barrier (BBB) is semitransparent. This allows the specialized cells of CVOs to directly sense the chemical composition of blood and secrete hormones into systemic circulation. Thus, the CVOs were nicknamed “windows to the brain” (Weindl and Sofroniew, 1981; Johnson and Gross, 1993). They perform diverse functions, including, but not limited to, the exchange of information between blood, brain and cerebrospinal fluid (CSF), generation of CSF with all its specialized proteins, etc. (Cottrell and Ferguson, 2004; Joly et al., 2007). More recently, it has been found that the CVOs, including the choroid plexus (CP), act as brain-immune interfaces mediating the transfer of immune cells from blood to brain (Shimada and Hasegawa-Ishii, 2017). This is in line with recent attempts to systematize the CVOs while considering microglia to be an important component of these structures (Oldfield and McKinley, 2014; Miyata, 2015; Kaur and Ling, 2017).

The whole range of CVOs' functions remains not fully understood even in model animals. Nevertheless, it includes many functions crucial for life support, as regulation of body fluids, temperature and energy balance, pain, brain detoxification, and so on. In mammals, there are nine CVOs: the pineal gland (PIN) or epiphysis [with the separate parapineal organ (PP) in some species, including zebrafish], subfornical organ (SFO), organum vasculosum laminae terminalis (OVLT), paraventricular organ (PVO), median eminence (ME), neurohypophysis (NH), subcommissural organ (SCO), and area postrema (AP). Some authors also add to this list the CP (Joly et al., 2007; García-Lecea et al., 2008; Wilson et al., 2010). At another extreme an extensive study described “about 17 different CVOs” in 31 species belonging to various groups of vertebrates, from cyclostomes to mammals (Tsuneki, 1986). This analysis suggested that the NH, ME, SCO, and PIN, found in almost all vertebrate species examined, could be the oldest CVOs, whereas the SFO and AP could have appeared in evolution relatively recently. It is thought that some CVOs are specialized to sense the chemical composition of blood. These were defined the “sensory” CVOs (OVLT, SFO, and AP). Several other CVOs coordinate brain responses by secreting hormones into peripheral blood stream. These were defined the “endocrine” or secretory CVOs (PIN, PP, NH, ME, and SCO; Ganong, 2000; Duvernoy and Risold, 2007). Despite the grouping of CVOs as specialized organs or their separation into different functional groups, the molecular mechanisms underlying such classifications remain insufficiently studied. As the first attempt in this direction the transcriptome of PIN, SCO, and SFO was analyzed using laser capture microdis section with an eye on a potential role of these CVOs as sites of periventricular tumors (Szathmari et al., 2013) and transcriptomics of the telencephalic and hindbrain CPs detected their regional specificity (Lun et al., 2015). Hence future studies may provide evidence helpful to characterize and/or classify CVOs.

The developmental analysis of CVOs in terrestrial vertebrates until now has been rather limited for the following reasons: the CVOs are small and inconspicuous, develop relatively late and their analysis is largely based on fixed material. Nevertheless, the development of CVOs has been described in different species, including the PIN (Calvo and Boya, 1981), OVLT (Szabó, 1983), SFO (Castaneyra-Perdomo et al., 1992), NH and ME (Ugrumov et al., 1989; Sasaki et al., 2003), PVO (Vigh and Vigh-Teichmann, 1998), CP (Sturrock, 1979; Dziegielewska et al., 2001), and AP (Borison, 1989; Castaneyra-Perdomo et al., 1992). The CP development was described in mice (Louvi and Wassef, 2000; Awatramani et al., 2003; Landsberg et al., 2005; Hunter and Dymecki, 2007).

The zebrafish is a popular model of developmental studies and, notably, in vivo analyses pave the way for addressing development of CVOs. The enhancer-trap transgenics represent useful tools of in vivo analysis of developing brain. Several enhancer-trap screens yielded a significant number of transgenic lines expressing various markers, including the cytosolic green fluorescent protein (GFP) (Parinov et al., 2004; Kondrychyn et al., 2009) or membrane-tethered KillerRed (Teh et al., 2010). In the zebrafish, development of some of CVOs has been studied, including the PIN (Masai et al., 1997), PIN and PP (Concha and Wilson, 2001), NH and ME (Bassi et al., 2016), PVO (Xavier et al., 2017), SCO (Fernández-Llebrez et al., 2001), AP (Ma, 1997; Holzschuh et al., 2003), and CP of the IIIrd and IVth ventricle (Bill et al., 2008; García-Lecea et al., 2008; Bill and Korzh, 2014). The study of CP development relied on in vivo analysis of the developing brain in the Gateways transgenics (ET33-E20). In this transgenic line, the GFP is also expressed in migratory microglia as well as brain regions other than CP (García-Lecea et al., 2008). Mapping of these unidentified regions based upon published literature and neuroanatomical landmarks suggested that in addition to the CP and migratory microglia the GFP expression domains may represent at least six other CVOs of zebrafish–the OVLT, SFO, AP, ME, PVO, PIN, i.e., most of CVOs of zebrafish, whereas in two other lines GFP expression was detected in the PP and SCO. In parallel, one of the transgenic lines expressing KR-KR19 was shown to express this marker in the CP (Teh et al., 2010; Korzh et al., 2011) and several other regions reminiscent of those in Gateways in indication of a similar developmental regulation of these independent markers. This illustrates the utility of the enhancer trap zebrafish transgenics for the neuroanatomical and developmental analyses of small brain structures. Given the scarcity of genetic information related to development of CVOs and the possibility to perform in vivo developmental analysis, these transgenics represent a very useful resource, which potential still remains to be exploited in more detail.

The embryos and larvae of Gateways express GFP in all three “sensory” CVOs-OVLT, SFO and AP, and “specialized” CVO-CP as well as in some, but not all, “endocrine” CVOs-ME, PVO, and PIN. It also expresses GFP in some regions that are not considered to be CVOs-the main longitudinal signaling center of the dorsal neural tube-roof plate, the regional signaling center–MHB, olfactory pits, optic nerve and migratory microglia. This establishes the “Gateways”-regulatory transcriptional machinery as the developmental integrator of most of CVOs. Interestingly, GFP expression in Gateways embraces CVOs of different subgroups and different cell lineages. In the CP GFP expression encompasses at least a couple of cell types, i.e., epithelial cells of the medial cluster and astrocyte-like cells of the lateral clusters (García-Lecea et al., 2008; Figure 5A), but not the vasculature. The fact that there is similar developmental regulation acting upon the AP, PIN, ME, OVLT, SFO, PVO, on one hand, and CP, on the other hand, further supports the notion that CP should be considered as one of CVOs, perhaps, the most divergent one. Intriguingly, while analysis of GFP expression in Gateways at the stages studied suggests some similarity of the CP and other CVOs, it, for reasons currently unknown, segregates the GFP-negative SCO and NH from the rest of CVOs. Perhaps, it indicates that the classification of CVOs must be more complex.

The expression of Gateways transgene seems to reveal developmental events common to the CVOs and roof plate, i.e., it may reflect a more general regulation compared to those events behind the separation of CVOs and roof plate. Thus, an enhancer (still to be cloned and sequenced) may play a role of a transcriptional node upon which expression of genes active in a context of CVO of different functional classes may rely upon. None of several hundreds of transgenic lines analyzed except KR19, which to a large extent phenocopies the expression pattern of Gateways (Teh et al., 2010; Korzh et al., 2011; Yan et al., 2012) show such broad expression in CVOs (Kondrychyn et al., 2009). The Gateways “enhancer” seems to be active early or very sensitive or both since in the CPIV it reveals the very early events of morphogenesis of the CP primordium-tela choroidea. If this idea is correct, this enhancer could be the early acting one that regulates the roof plate, MHB, most of CVOs, olfactory pits and migratory microglia. To be validated this idea should be studied further by subjecting Gateways along with KR19 to more detailed analysis aiming to isolate genomic regions regulating development of CVOs and study their role during development and disease of these structures.

The CVOs of adults are highly vascularized (Duvernoy and Risold, 2007). In contrast, the embryonic CVOs remain avascular in zebrafish development despite capillaries pass close by the GFP areas. The tela choroidea connects to a capillary, but it is not penetrates inside this area well before the morphologically recognized CP is formed (García-Lecea et al., 2008). Notably, early domains of GFP expression in Gateways look like preferred areas for capillaries to penetrate, whereas our detailed analysis demonstrated that during embryogenesis and early postnatal development vasculature develops in close proximity, but outside of GFP expression domains, which remain avascular (Figures 2, 3, 7). It could be due to various reasons, which may include, but not be limited to, a requirement for CVO to differentiate relying largely on local cues. Further to that, being small zebrafish embryos rely on oxygen diffusion and survive without circulation for several days. Hence the developing CVOs being even smaller may develop due to diffusion from closely located capillaries.

The CVOs share their origin with cells of signaling glia, which constitute the dorsal axial morphogenetic center of the neural tube-the roof plate (Chizhikov and Millen, 2005; Korzh, 2007). In this context it is of interest that the CVOs also is a source of periventricular tumors (Szathmari et al., 2013). In view of scarcity of information about genes regulating CVO development, it might be useful to consider as candidates genes near to the transgene insertions site in transgenic lines analyzed. These genes could be expressed in the CVOs and play a role in their development (Table 1). For example, the analysis of genes found in the vicinity of the Gateways transgene insertion on Chr. 24 (García-Lecea et al., 2008; Kondrychyn et al., 2009) brought into limelight sulf1. In parallel, this gene was found to be expressed in the CP and elsewhere in the brain and plays a role in establishing the VegfA-mediated arterial venous identity (Gorsi et al., 2010, 2014). Importantly, the sulf1 loss-of-function causes deficiency of the IVth brain ventricle (Gorsi et al., 2014), which could be attributed to deficiency of the CPIV. This opens a possibility that sulf1 could be one of the genes regulated by the activity of the same enhancer that drives GFP expression in Gateways. Indeed, the roof plate-derived chondroitin sulfate may be involved in establishing developmental patterning in the neural tube (Butler and Dodd, 2003). Henceforth, sulf1 could play an early role in the developmental program of the roof plate and at least some CVOs. The expression pattern similar to that of the Gateways was observed in the transgenic line KR19 expressing membrane-tethered KillerRed fluorescent protein, which unlike Gateways (Chr.24) has been mapped to Chr.8 (Teh et al., 2010; Korzh et al., 2011; Yan et al., 2012).

In this study, the Gateways transgenic line was used to analyze development of several CVOs, including the OVLT, SFO, AP, PIN, PVO, and ME. Several CVOs (PIN, PP, CP, SCO, AP) studied previously are localized in the dorsal neural tube, which makes their microscopic observation more convenient. This was a useful feature in looking for similarity in expression pattern of Gateways and KR19. Our analysis of transgenics demonstrates that the AP and SCO represent rather obvious thickenings of the roof plate at the brain boundaries (Figures 3, 6, Figure S2) and could be correctly mapped based both on available neuroanatomical landmarks, specific antibodies (SCO) and prior observations (Ma, 1997; Fernández-Llebrez et al., 2001; Holzschuh et al., 2003). Holzschuh et al. (2003) Mapping of other CVOs-OVLT, SFO, PVO, ME–found in more ventral position is more complicated due to their small size (ME), opacity of surrounding tissues (OVLT) or both (SFO) and is more reliable using more complex combination of microscopic techniques. The adult rat OVLT is located anterior to the IIIrd ventricle (Prager-Khoutorsky and Bourque, 2015), whereas in the developing zebrafish it is localized initially posterior to the ventral-most extent of the IIIrd ventricle, where it forms several closely connected clusters. It looks like while formed as individual cell groups, the OVLT and SFO in later development fuse to form one complex similar to that of the PIN-PP complex (Figure S2). Of interest are different signaling requirements regulating development of these two CVOs. The OVLT being a larger of the two seems to be born earlier and spared of the deleterious effect that the Wnt and Notch deficiencies have on the SFO (Figure S1). Being sensitive to Notch is consistent with the idea that the SFO is formed later that OVLT, when a pool of neural progenitors is depleted (Itoh et al., 2003; García-Lecea et al., 2008). In contrast, the Hh and FGF8a signaling are critical for development of both CVOs (Figures S1C,D,I,J). Such developmental time table is also clear from morphological analysis.

The AP and SFO are considered to be the evolutionarily “young” CVO (Tsuneki, 1986). It is rather gratifying that the SFO, although is rather minute initially, could be identified since it expresses the same transgene as two other “sensory” CVO-AP and OVLT. Unlike the closely related OVLT, until now SFO remained unidentified even in adult zebrafish. This could be due to a fusion with the larger OVLT during the period from 3 to 6 days postfertilisation. Later analysis of genes known to be expressed in the SFO of larger animals (Hindmarch and Ferguson, 2015) may link their expression to the site where the SFO is found in the brain of developing Gateways embryos/larvae.

Despite the ME being found in a very ventral position, counterintuitively it may share its origin with the dorsal midline cells due to the cephalic flexure, which brings cells found during neural plate stage in the anterior and dorsal position into the ventral forebrain later on. In the Xenopus much of the ventral forebrain derives from the anterior neural ridge (Eagleson and Harris, 1990). Similarly, in zebrafish the adenohypophysis, which the ME is adjacent to, is of the anterior placode origin and derives from the most anterior tip of the neural plate (Glasgow et al., 1997; Wang et al., 2001). ME progenitors may derive from a site very close to that of the adenohypophysis. Same could be said about the origin of the OVLT/SFO found anteriorly to the ME and adenohypophysis. This means that the OVLT/SFO progenitors during neural plate stage may reside somewhat posterior to the more anterior progenitors contributing to the adenohypophysis and ME. This hypothesis relies solely on the observation of developmental changes in the expression of Gateways transgene. Its experimental validation is beyond the scope of this descriptive study and should wait further experiments.

Conceived and designed the experiments: MG-L, IK, CT, and VK. Performed the experiments: MG-L, IK, CT, EG, JJ, and M-SY. Analyzed the data: MG-L, IK, CT, EG, JJ, and VK. Wrote and approved the paper: VK.