A comprehensive structural, lectin and immunohistochemical characterization of the zebrafish olfactory system

Fish chemosensory olfactory receptors allow them to detect a wide range of water-soluble chemicals, that mediate fundamental behaviours. Zebrafish possess a well-developed sense of smell which governs reproduction, appetite, and fear responses. The spatial organization of functional properties within the olfactory epithelium and bulb are comparable to those of mammals, making this species suitable for studies of olfactory differentiation and regeneration and neuronal representation of olfactory information. The advent of genomic techniques has been decisive for the discovery of specific olfactory cell types and the identification of cell populations expressing vomeronasal receptors. These advances have marched ahead of morphological and neurochemical studies. This study aims to fill the existing gap in specific histological, lectin-histochemical and immunohistochemical studies on the olfactory rosette and the olfactory bulb of the zebrafish. Tissue dissection and microdissection techniques were employed, followed by histological staining techniques, lectin-histochemical labelling (UEA, LEA, BSI-B4) and immunohistochemistry using antibodies against G proteins subunits αo and αi2, growth-associated protein-43, calbindin, calretinin, glial-fibrillary-acidic-protein and luteinizing-hormone-releasing-hormone. The results obtained enrich the available information on the neurochemical patterns of the zebrafish olfactory system, pointing to a greater complexity than the one currently considered, especially when taking into account the peculiarities of the nonsensory epithelium.

www.nature.com/scientificreports/ Furthermore, our study aimed to address two specific issues. First, in the light of existing studies comparing the olfactory transcriptomes between zebrafish and mouse, which have revealed a high degree of molecular conservation, we aimed to phenotypically characterize two markers of the olfactory sensory transduction chain that are characteristic of the mammalian vomeronasal system, Gαi2 and Gαo, to examine their expression pattern in zebrafish mirrors that in mammals. Second, we aimed to investigate the possible involvement of the nonsensory zone of the olfactory rosette epithelium in chemoreception.

Material and methods
Ten wild-type 1-year-old zebrafish (Danio rerio, wild-type) were used in this study. They were maintained at 28.5 °C in 30 L aquaria at a rate of 1 fish per liter of dechlorinated water, with reverse osmosis purified, and under a light-dark cycle of 14:10. Fishes were euthanized by tricaine overdose (MS-222, Sigma, St. Louis, MO). We followed the ARRIVE guidelines to ensure that all experiments were performed under good conditions. Whole heads were promptly immersed in modified Bouin's fixative solution. After 24 h, the samples were transferred into 70% ethanol. The samples were not decalcified. In all cases paraffin embedding was used to perform the histological procedures. All samples were cut with a Leica Reichert Jung microtome with a thickness of 4-8 μm.
Gallego's trichrome. This stain allows for the differentiation of components of the connective tissue. It stains erythrocytes green, muscle fibers and collagen light blue, epithelium and glandular tissue red, bone dark blue and cartilage purple. The protocol used was described in detail in 54 as follows: sections were first stained with Ziehl acetic fuchsin for 2 min. After several washes they were introduced into formalin-acetic acid solution for 5 min. After two more washes, the sections were finally introduced into picroindigocarmine for 3-5 min.
Histochemical and immunohistochemical staining. The histochemical and immunohistochemical protocols followed by the authors have been fully described in previous contributions 54,55 . Briefly: Histochemical labelling (HQ) with lectins. We have used (1) a lectin that comes from gorse, the Ulex europaeus agglutinin (UEA), α-l-fucose specific, (2) the α-galactose-specific BSI-B 4 that comes from Bandeiraea simplicifolia, and (3) Lycopersicon esculentum agglutinin (LEA), coming from tomato with a high affinity for N-acetyl-β-d-glucosamine oligomers (Table 1). These stains selectively recognise the different components of the olfactory and vomeronasal pathways in some species 55 .
The protocol for the UEA is as follows. (i) blocking the endogenous peroxidase activity of the sample by incubation in 3% H 2 O 2 solution for 10 min; (ii) incubation for 30 min in 2% bovine serum albumin (BSA), to prevent nonspecific binding; (iii) incubation with the UEA lectin for 1 h; (iv) 3 × 5 min washes in 0.1 M phosphate buffer (PB, pH 7.2), and (v) incubating for 12 h in a peroxidase-conjugated immunoglobulin against the UEA. Finally, (vi) the sections were washed with PB and developed by (vii) incubation in 0.05% diaminobenzidine (DAB) and 0.003% H 2 O 2 for 5 min.
The protocol for the LEA and BSI-B 4 begins with the same two steps. Next, (iii) the incubation of the sections was done overnight in biotinylated lectins diluted in 0.5% BSA. The next day, the samples were (iv) 1.5 h Immunohistochemistry (IHQ) techniques. This protocol also began by (i) blocking the endogenous peroxidase. Then, (ii) non-specific binding was blocked with 2.5% horse normal serum from the ImmPRESS reagent kit Anti-mouse IgG/Anti-rabbit IgG (Vector Laboratories, CA, USA) for 30 min. (iii) The primary antibody was added at the corresponding dilution (Table 1) and incubated overnight. The next day, (iv) the samples were incubated for 20 min with the ImmPRESS VR Polymer HRP Anti-Rabbit IgG Reagent. (v) After rinsing in Tris-buffer (pH 7.61) for 10 min, (vi) the samples were finally developed using DAB in the same way as for the lectins 54,55 . All immunohistochemical protocols were checked with the appropriate controls. Samples for which the primary antibody was omitted were used as negative controls. Table 2 gives references to previously published use in zebrafish and other fishes of the antibodies here employed against the same antigens.
Acquisition of images and digital treatment. Digital images were captured using the Karl Zeiss MRc5 digital camera attached to a Zeiss Axiophot microscope. Adobe Photoshop CS4 (Adobe Systems, San Jose, CA, USA) was used to adjust parameters such as brightness, contrast and balance light levels for presentation in this work. No features of the image were enhanced in any way, moved, or introduced. Some photomicrographs were formed as a mosaic of several photographs merged with an image-stitching software (PTGui Pro, The Netherlands). In no case was this software used to generate images that do not correspond to the actual images presented in the manuscript.

Results
The zebrafish olfactory rosette occupies an anterodorsal position, slightly rostral to the orbits (Figs. 1 and 2A, Suppl. Fig. S1A-C). In a parasagittal section at the orbital level, the outlet nostril coincides with a zone of lesser development of the lamellae (Suppl. Fig. S1C). In a sagittal section, which includes the central part of the OB (Fig. 2B, Suppl. Fig. S1B), the rosette can be observed at its maximum expression, divided into two symmetrical zones by the presence of a wide raphe that serves to support the lamellae. The olfactory nerve associated with each lamella constitutes a single branch that reaches the OB from the ventromedial side (Fig. 2G, Suppl. Fig. S1D). The staining of the sensory pseudostratified epithelium with Gallego's trichrome reveals the organization and allows for the differentiation of basal cells from the olfactory sensory cells. The three primary olfactory cell types are densely intermingled but can be distinguished by their characteristic shapes and spatial positions: a slender dendrite and a basal soma for ciliated neurons, a rounded cell body and an intermediate soma position for microvillous neurons, and a large globose soma in an apical position for crypt neurons (Fig. 2C). The luminal surface of the nonsensory epithelium is covered by cilia (Fig. 2D,E). Alcian Blue staining shows the presence of acidic mucopolysaccharides on the epithelial surface of the lamella, mainly in the luminal surface of the sensory area, whereas the nonsensory epithelium border is free of acid mucins (Fig. 2F).
The zebrafish OB is diffusely laminated, but its three layers can be identified from the periphery toward the center: olfactory nerve layer, glomerular layer, and granular layer (Fig. 2G,H). The olfactory nerve layer is formed by the axonal endings of olfactory receptor neurons (ORNs) (Fig. 2G). The glomeruli do not resemble the distinct spheres observed in mammals due to the few numbers of periglomerular cells and glial elements found in zebrafish. The cell bodies of the mitral cells are intermingled in the boundaries between the glomerular and granular layers (Fig. 2H). The latest is the deeper layer, which is primarily formed by granule cells (Fig. 2G).
Immunohistochemical staining of the olfactory rosette. The immunohistochemical study with anti-Gαi2 and anti-Gαo produce two differentiated patterns (Fig. 3). Intense immunoreactivity was noticed in the central and medial portions of the OE when employing Anti-Gαi2 ( Fig. 3A-C,E,F). The immunopositive cells are distributed across the entire thickness of the epithelium, but appear mostly concentrated in its superficial half (Fig. 3F). Contrastly, anti-Gαo produces a diffuse immunoreactivity circumscribed to cells present on the apical surface of the OE (Fig. 3D,G-I). Interestingly, we also found large isolated cells with stronger anti-Gαo immunopositivity in the OE (Fig. 3H). Additionally, the anti-Gαo labelling also stains the apical surface of the nonsensory area of the hair cells (Fig. 3G). Both markers, anti-Gαi2 and anti-Gαo label the olfactory nerves ( Fig. 3E,I). www.nature.com/scientificreports/ The results of the immunohistochemical study with anti-calbindin (anti-CB) and anti-calretinin (anti-CR) are depicted in Fig. 4A,B,D,E and C,F respectively. Both markers label a subpopulation of olfactory sensory neurons. Sagittal and transverse sections of the rosette show how the anti-CB inmmunolabelling is mostly located in the deeper part of the neurosensorial epithelium (Fig. 4D), whereas the nonsensory epithelium is Anti-CB immunonegative (Fig. 4B). Anti-CR immunolabelling produces a stronger labelling than anti-CB in neuroepithelial cells, which is mostly concentrated in the medial part of the lamellae (Fig. 4C), and their deeper neuroepithelial layers (Fig. 4F). Very rarely superficial cells are immunolabelled. The nonsensory cripts do not show anti-CR labelling.  www.nature.com/scientificreports/ Anti-GFAP immunolabels isolated big cell bodies in the apical part of the whole OE ( Fig. 4G-J). These cell bodies are present in both sagittal and transverse sections. Anti-GAP-43 immunolabelling is located in the apical part of the nonsensory epithelium and in individual cell bodies in the crypts. (Fig. 4K,L). Anti-LHRH produces a light immunolabelling, mainly located in the cell processes of both the sensory and nonsensory epithelia (Fig. 4M,N).
Lectin histochemical staining of the olfactory rosette. The three lectins employed in this study labelled the olfactory system following individual patterns (Fig. 5). The labelling with LEA of the whole OE clearly delineates the limits between the olfactory and the nonsensory epithelium -this latter unstained (Fig. 5A,B). Additionally, big oval neuron-like cell bodies in the crypts of the nonsensory epithelium are also stained, and in some of them thin dendritic processes can be clearly appreciated (Fig. 5C,D,G). UEA only labelled secretory material, mainly in the crypts, but also in the sensory epithelium but in lesser extent (Fig. 5E,F). BSI-B 4 marks  www.nature.com/scientificreports/ individual cells in the crypts and in the nonsensory part of the lamella and very occasionally cells belonging to the sensory part of the OE. In all cases they are scattered cells, smaller than those marked by LEA (Fig. 5H-J).
Immunohistochemical and lectin histochemical staining of the olfactory bulb. The immunohistochemical labelling of the olfactory bulb ( Fig. 6) primarily reveals two patterns. The labelling for the calcium-binding proteins (CB and CR) is concentrated in the dorsal and lateral regions. Using the terminology established by Braubach et al. 61,81 , calcium-binding proteins were identified in the dorsal glomeruli (dG), clusters of the dorsolateral (dlG) and lateral glomeruli (lGx), and the large lateral glomerulus 1 (lG 1 ). Additionally, anticalbindin stained glomeruli belonging to the mediodorsal (mdG) and the ventromedial parts of the bulb were stained with reduced intensity (ventral posterior glomeruli, VpG). Both G protein antibodies showed similar patterns, with a higher concentration of labelling in the ventromedial glomerular cluster 1-6 (VmG 1-6 ) and the large ventromedial glomerulus 7 (VmG 7 ), which appeared stronger for anti-Gαo. Both G proteins faintly stained the mediodorsal glomeruli (mdG), and Gαi2 marks glomeruli belonging to the ventral posterior cluster (VpG). Finally, LEA exemplifies a second pattern, in which the labelling was mainly circumscribed to the ventral area. This lectin also labels the dorsolateral area but with less intensity. LHRH is expressed primarily in the dorsolateral glomeruli (dlG). Anti-GAP-43 and the other two lectins examined in this study did not show significant results in the OB.

Discussion
Under an appearance of simplicity, in recent decades the olfactory system of fish has revealed a very high structural complexity, according to cell types and receptors involved. The advent of genomic techniques such as RT-PCR, ISH, genomic and transcriptomic analyses, has been decisive for the discovery in fish of specific olfactory sensorial cell types such as kappe or crypt cells. These techniques have also played a significant role in the identification and characterization of large cell populations expressing vomeronasal receptors 42,44 , thus putting an end to the longstanding controversy about the existence of an AOS in fish 32 . In zebrafish, genomic advances have been so rapid that they have unavoidably marched ahead of morphological and neurochemical studies. Instead, studies in mammals, have formed a solid basis on which to build knowledge of the olfactory and vomeronasal systems 82,83 . For this reason, there is a lack of specific histological, lectin-histochemical and immunohistochemical studies of the olfactory rosette and OB of the zebrafish. To discuss our results, we must therefore contextualize them in a higher taxonomic order including other fish families.
Although the presence of a vomeronasal organ is a tetrapod evolutive innovation, the vomeronasal receptor genes have been identified in fish and even in the lamprey 84 . Taking into account that each vomeronasal type receptor, V1R and V2R, is associated in mammals to a unique G protein, Gαi2 and Gαo respectively, the present study aimed to determine whether there is a correlation between such G-proteins and the zebrafish olfactory cell morphology. It is known from the literature that the receptor molecules and the G-protein specific for each receptor are detectable not only in the dendritic process of the neuroreceptor cell, but also along the axons and their termination in the glomeruli of the OB 85 . For this reason, we have extended the study of these molecules of the chain transduction to the olfactory bulb.
Regarding Gαo, Hansen et al. 78 correlated in goldfish Carassius auratus the receptor cell morphology and the cell types distribution, with the expression of G-proteins, demonstrating that anti-Gαo immunoreactivity was present on microvillar ORNs located in the upper half of the OE. This happens similarly in the Gαo neurons identified by us in the zebrafish, pointing to the reliability of anti-Gαo as a marker of microvillar V2R-like cells in this species. Other studies in different fish species, such is the case of Chondrichthyes, support this view. Thus, immunohistochemical studies of G protein α subunits in the olfactory organ of Scyliorhinus canicular (Elasmobranch) and Chimaera monstrosa (Holocephali) found the presence of Gαo in virtually all ORNs, which was consistent with the presence of V2Rs 76,77 .
In our case, additionally to the profuse microvillar Gαo-positive neurons, we have also found a subpopulation of large, oval-shaped Gαo positive cells, always located on the apical surface of the OE. Although their morphology is reminiscent of crypt cells, the study by Ahuja et al. 36 in zebrafish demonstrated that these cells constitute a new olfactory cell type, the kappe cells. Their immunofluorescence study showed that kappe neurons  www.nature.com/scientificreports/ are identified by their anti-Gαo immunoreactivity, demonstrating a scattered spatial distribution within the OE, similar to, but significantly different from that of crypt neurons. Ahuja et al. 36 also found that kappe neurons project to a single identified target glomerulus within the OB, belonging to the mediodorsal cluster. This observation coincides with the anti-Gαo pattern of labelling found by us in the OB, showing immunopositivity to Gαo in the mediodorsal part of the bulb. We also found an intense immunoreactivity to anti-Gαo in the cluster of ventromedial glomeruli 1-6 (VmG 1-6 ) and the large ventromedial glomerulus 7 (VmG 7 ), an area which has been attributed to an important projection of fibers from microvillar cells 30,85 . However, this observation contrasts with that obtained by Braubach et al., who did not find any anti-Gαo positivity in the ventral glomeruli and only detected Gαo in the mediodorsal glomerulus 5 (mdG 5 ) 61 .
Our observations confirm the validity of the anti-Gαo antibody as a reliable marker of V2R-like receptor cells in zebrafish, pointing to the presence in zebrafish of a large population of microvillar and kappe cells whose transduction chain is analogous to that present in mammals V2R vomeronasal cells. Therefore, this appears to be an ancient trait conserved through the vertebrate evolution 78 .
Regarding crypt olfactory receptor cells, none of the antibodies and lectins employed by us have labelled specifically these cells. Nonetheless, the study by Catania et al. 86 characterized immunohistochemically these crypt cells in zebrafish employing an antibody against the neurotrophin receptor Trk-A.
To our knowledge, the inhibitory subunit Gαi2 has not been studied in the fish olfactory system. This fact is surprising since Gαi2 is part of the vomeronasal V1R receptor transduction chain in mammals. Mammalian Vmn1r genes show a rather dynamic evolution, in striking contrast to the highly conserved fish orthologous, the ORA gene family. In zebrafish, six ORA receptors have been identified 46 , and only in one of them, the ORA4, the precise location of its expression has been studied. Thus, ORA4 receptor was found to be expressed in crypt www.nature.com/scientificreports/ neurons, but not associated with Gαi2 but with the inhibitory G protein, Gi1b 51 . As for its ligands, it is only known that the ORA1 gene recognizes with high specificity and sensitivity the 4-hydroxyphenylacetic acid 87 , which might function as a pheromone for reproductive behaviour in zebrafish. ORA1 is ancestral to mammalian V1Rs, and its putative function as pheromone receptor is reminiscent of the role of several mammalian V1Rs as pheromone receptors. The anti-Gαi2 pattern of labelling is very different from that found with anti-Gαo, as it covers a wider thickness of the epithelium, although it rarely reaches the deepest cell layers. Moreover, unlike anti-Gαo immunolabelling, it comprises the entire extension of the olfactory epithelium clearly demarcating it from the nonsensory epithelium. The immunohistochemistry of the olfactory bulb using anti-Gαi2 results in the labelling of a huge glomeruli subpopulation, mainly belonging to the ventromedial and ventral posterior clusters, confirming that the Gαi2-positive cells in the olfactory rosette are sensory neurons that convey information to the brain. If ORA receptors coincide with the V1Rs in having the protein subunit Gαi2 in their transduction chain, it is surprising that such a small number of receptors are expressed on such a high number of olfactory cells as those detected in our zebrafish olfactory rosette and bulb immunolabelling.
Calcium-binding proteins contribute to calcium homeostasis by buffering the intracellular free calcium concentration 88 . Both CR and CB protect sensory neurons against calcium increases during periods of high frequency discharge as well as in pathological conditions 89 . Moreover, calretinin and calbindin immunoreactive (CR-ir and CB-ir, respectively) neurons in the cerebral cortex are resistant to degenerative processes in Alzheimer's disease 90 .
The distribution of CR in the olfactory system of the zebrafish was investigated for the first time by Castro et al. 62 , by using immunocytochemical techniques. Our CR immunoreactivity coincide essentially with their own observations. Accordingly, it is remarkable the presence of numerous CR-ir bipolar cells in the neuroepithelium and an intense immunopositivity in the olfactory nerve. Parisi et al. 64 performed immunofluorescence against CR on zebrafish crypt cells, finding immunolabelling in the OE, primarily in the intermediate cells, but also in the superficial layer. Morphologically, immunopositive cells resembled to them both microvillous and crypt cells. However, our light microscopy anti-CR immunopositive cells featured a slender dendrite and an elongated soma; a morphology closer to olfactory than to microvillous cells. Our observations agree with the immunoelectron microscopy investigation by Gayoso et al. 29 who found a consistent anti-CR immunopositivity in the ciliated cells and only very rarely in the microvillar cells. Additionally, neither Castro et al. 62 , Gayoso et al. 29 nor us have found anti-CR immunopositive crypt cells.
Calretinin immunolabelling of the olfactory rosette has been used in toxicity studies examining the effects of various chemicals, such as copper, zinc, urea, and detergents [91][92][93][94] . Although the labelling patterns observed in the control individuals of these studies are similar to ours, in some cases 93,94 , higher staining intensity was obtained in these studies than was obtained in our samples. Both of these toxicity studies were performed using samples decalcified with EDTA, an agent that can act as an antigenic retrieval agent. In addition, these studies have used a polyclonal antibody from a different source than ours. Therefore, the results of these studies should not be compared to our results in terms of intensity. The reduced staining intensity observed in our study allows for the better discernment of cell morphology; however, to facilitate comparisons between morphometric studies, the standardization of staining methods should be considered. Finally, studies in other fish species such as the one carried out in the olfactory rosette of guppy 59 have proved similar observations to those found by us in zebrafish.
In the OB, Castro et al. 62 and Braubach et al. 61 each presented comprehensive studies of the entire olfactory glomeruli population and found that the dorsal, dorsolateral, and lateral glomeruli clusters, the large lateral glomerulus 1, and the ventromedial glomerular fields exhibited strong anti-CR labelling, whereas the dorsomedial area exhibited only faintly CR-ir fibers. Our observations are mostly comparable to those found by them, with the exception that we did not find immunopositive ventromedial glomeruli, likely due to a difference in the levels chosen for our study.
Regarding the expression of calbindin in the olfactory system of the zebrafish, there is a lack of information. However, there have been studies in other fish species such as the chondrostean, Acipenser baerii 56 and the cladistian fish Polypterus senegalus 57 , in both cases finding a very faint expression in the olfactory rosette. Our study shows a very high expression of calbindin in the zebrafish olfactory organ, which is accompanied by a parallel expression in the OB. The immunolabelling is comparable to that produced by anti-calretinin, but wider in the case of calbindin, as it is extensible to the ventral posterior glomeruli and the mediodorsal glomeruli, bigger than the dorsal and dorsolateral glomeruli.
The expression of GFAP has been very little studied in the olfactory system of fish. Notwithstanding, there is a specific study in zebrafish by Lazzari et al. 72 about olfactory ensheathing cells (OECs) with employed different markers (antibodies against GFAP, S100, NCAM, p75). OECs are unique glial cells with axonal growth-promoting properties, involved in the regenerating capability of ORNs throughout life. These cells sustain the continuous axon extension and successful topographic targeting of the olfactory receptor neurons. They are present in the OE and the OB and are also expressed along the entire length of the olfactory nerve. Lazzari et al. 72 reported slight immunostaining in the OE, moderate staining in the OB olfactory nerve layer, and faint positivity in both the glomerular layer and the inner bulbar zone of zebrafish. Our study identified stronger labelling in both the olfactory nerve and the glomerular layers of the bulb. Although we have used the same fixative and commercial antibody as those used by Lazzari et al. 72 , their samples were treated for decalcification by EDTA, whereas we did not subject the tissue to any types of chelating treatment. This difference in sample treatment might explain the observed differences in the immunolabelling intensity observed between the two studies. Other studies performed in zebrafish have reported an anti-GFAP immunolabelling pattern in the OB similar to ours. For example, Byrd and Brunjes 25 reported that the olfactory nerve and glomerular layers feature predominant GFAP immunoreactivity, and Scheib and Byrd-Jacobs 95 reported numerous processes concentrated in the nerve and glomerular layers (GL). Additionally to the previous reports, we found the presence of immunopositive cells www.nature.com/scientificreports/ in the apical surface of the epithelium, which vary in shape and size, but they are predominantly globose. The meaning of such immunopositive cell bodies, previously undescribed, should be further studied.
Anti-LHRH has been used in fish and mammals to characterize the terminal nerve 80 ; a ganglionated extrabulbar nerve, independent of the olfactory nerve, but close enough to be identified by classical histological methods. Its function is uncertain, although its fibres facilitate migration of LHRH cells to the hypothalamus, thus participating in the development of the hypothalamic-gonadal axis 96 . Although extrabulbar elements have been characterized in the forebrain of the zebrafish 29 , they differed from the terminal nerve as they were not immunoreactive to the most widely employed marker for this nerve, FMRFamide 97 . We were not able to identify LHRH fibers in the olfactory rosette, but we observed immunoreactivity in neuronal elements of the dorsolateral OB. This result is consistent with that observed in fish by Münz and Class 98 and mammals by Witkin and Silverman 99 .
Histochemical labelling with lectins has been widely used in fish, but to a lesser extent in the olfactory system, and with only a few references to the specific case of zebrafish. UEA-I, specific for l-fucose, has been studied in trout 100 finding intense labelling in cell processes located in the apical region of the OE, in some elements of the basal layer and a few cells in the nonsensory epithelium. However, regarding the OB the authors only found positive glomerular fields in five of seven trouts, and in a heterogeneous shape. Pastor et al. 101 , studied two Teleostei, Sparus auratus and Dicentrachus labrax finding negative reaction in the olfactory rosette. Our own findings highlight the interspecific diversity in UEA marking, as we have found a positive reaction on the luminal surface of the neuroepithelium and a high concentration of l-fucose in the crypts of nonsensory epithelium, probably due to the mucosal secretion of these cells. Regarding the OB we did not find labelling in our specimens.
LEA has been more widely used to characterize the olfactory system of fish than UEA, but surprisingly there are no specific studies in zebrafish, where we have seen that it is an excellent marker of the sensory epithelium. LEA labels all its cellular elements, establishing a clear border with the unlabelled nonsensory epithelium. The striking presence in the crypts of individualized cells with a clear neuronal morphology, oval shape and fine dendritic processes, has not previously described and should be object of future studies. Interestingly, the neuronal features of these cells, would coincide with the observations by Amato et al. 102 who found TRPV4 immunoreactive "unknown cells" in the nonsensory epithelium of the zebrafish olfactory rosette. TRPV4 is a nonselective cation channel that belongs to the vanilloid subfamily of transient receptor potential ion channels. These authors suggest that these TRPV4 cells might be involved in olfactory sensation. Moreover, Parisi et al. 64 verified that TRPV4 cells did not colocalized with calretinin, which is consistent with the lack of calretinin immunopositive cells in the crypts, reported by us. Moreover, our immunolabelling with anti-GAP-43 produced a similar pattern in the crypt epithelium when compared to LEA. All these results together point for the first time to a chemosensory nature of the crypts LEA positive cells.
All mammal and fish species studied till date have shown positivity to LEA in their olfactory sensory epithelia, apart from the case of Pleuronectiformes 103 , in which surprisingly and as an exception, LEA negative staining in the OE was reported. Regarding the OB, our results are consistent with those observed in the lungfish Protopterus annectens 104 where LEA positivity was found in the ventral part of the OB, a region associated with reproductive behaviour 105 .
Our marking with BSI-B 4 stained both the olfactory and the nonsensory epithelium, respectively producing a striking pattern containing scattered cells in the olfactory sensory epithelium, and a widespread labelling of neuronal-like cells similar to those identified with LEA in the nonsensory epithelium. There are no references to compare this striking result in zebrafish, since studies with this lectin existing in other fish species, such as eels and sharks 106,107 have been restricted to the sensory epithelium, where these authors described a diffuse labelling.
Overall, we have performed the comprehensive characterization of the zebrafish olfactory system at the morphological, lectin, and immunohistochemical levels, providing new information for a wide range of previously unexplored markers and further confirming the characterization of previously examined markers using independent methodological approaches. Moreover, our results showed that specific markers of the transduction chain previously associated with mammalian vomeronasal receptors are also effectively expressed in the rosette and olfactory bulb of zebrafish and, additionally, suggest the possible involvement of the nonsensory zone of the olfactory rosette epithelium in chemoreception. These results significantly add to the currently available information regarding the neurochemical profile of the zebrafish olfactory system, indicating that the zebrafish system displays greater complexity that is currently acknowledged.