Response of Olfactory Sensory Neurons to Mercury Ions in Zebrafish: An Immunohistochemical Study

Abstract Olfactory sensory neurons (OSNs) of fish belong to three main types: ciliated olfactory sensory neurons (cOSNs), microvillous olfactory sensory neurons (mOSNs), and crypt cells. Mercury is a toxic metal harmful for olfaction. We exposed the olfactory epithelium of zebrafish to three sublethal Hg2+ concentrations. Molecular markers specific for the different types of OSNs were immunohistochemically detected. Image analysis of treated sections enabled counting of marked cells and measurement of staining optical density indicative of the response of OSNs to Hg2+ exposure. The three types of OSNs reacted to mercury in a different way. Image analysis revealed that mOSNs are more susceptible to Hg2+ exposure than cOSNs and crypt cell density decreases. Moreover, while the ratio between sensory/nonsensory epithelium areas is unchanged, epithelium thickness drops, and dividing cells increase in the basal layer of the olfactory epithelium. Cell death but also reduction of apical processes and marker expression could account for changes in OSN immunostaining. Also, the differential results between dorsal and ventral halves of the olfactory rosette could derive from different water flows inside the olfactory chamber or different subpopulations in OSNs.

Mercury is a global, bio-accumulative persistent and toxic metal, which causes a very high risk to environmental and human health (for references, see Bernhoft, 2012;Syversen & Kaur, 2012;Bjørklund et al., 2017;Ha et al., 2017;Zhu et al., 2018). It is present in the environment in consequence of both natural processes (volcanic events, forest fires, rock weathering) and human activities (Risher & De Rosa, 2007). Anthropogenic sources include sources of nonoccupational exposure before mercury was replaced with safe substances (home, office, school, and health care devices: thermometers, barometer, sphygmomanometers, fluorescent bulbs, batteries, paints, inks), fossil fuel combustions, and occupational exposures (dental practice, mining activities, various industrial applications) (Bhan & Sarkar, 2005;Risher & De Rosa, 2007). Mercury exists in three chemical forms: elemental mercury, known as metallic mercury, organic mercury, mainly methyl mercury, and inorganic mercury compounds, mercury salt, mainly mercuric chloride. Each is characterized by a specific profile of toxicity (Guzzi & La Porta, 2008). In addition to elemental mercury which is slightly water soluble and represents only a small portion of the total mercury in superficial freshwater, inorganic mercury together with organic mercury are the chemical forms primarily found in water, sediment, soil, and living organisms (Zhu et al., 2018). In water, both inorganic and organic mercury are found in dissolved, colloidal, and suspended phases. In nature, the different mercury chemical forms undergo complex cycling and transformations (Krabbenhoft & Rickert, 1995;Gonzalez-Raymat et al., 2017). Several environmental models for mercury transport and fate in aquatic systems have been developed, which represent valuable decision-making tools in mercury pollution control and management (see Zhu et al., 2018). Due to mercury ubiquitous environmental presence, freshwater and seawater fish may contain large amounts of it taken through the food chain. That is why eating large amounts of fish is hazardous especially for children and the developing fetus in pregnant women (Goldman & Shannon, 2001).
Mercury is a diffuse neurotoxic agent affecting fish olfaction: in the Atlantic salmon, it was observed accumulation of mercuric chloride around the borders of OSNs (Tierney et al., 2010), while it caused cilia degeneration and cell death (both sensory and nonsensory) in the Indian major carp after extended exposure (15 and 30 days) to sublethal doses (Ghosh & Mandal, 2014). Unlike copper (Razmara et al., 2021), the mechanisms by which mercury affects olfaction are less known (Ribeiro et al., 1995). Indeed, nothing is reported about behavioral responses or adverse effects after short-term exposure (less than 7 days) in adult fish, so it is impossible to compare mercury effects with other metal toxicity.
In this study, we analyzed the effect of exposure to sublethal doses of mercury ions (Hg 2+ ) in the three different types of OSNs in zebrafish (Danio rerio Hamilton, 1882). Zebrafish is a suitable model because immunohistochemical properties of its olfactory organ are extensively described in the literature (Hill et al., 2005;Germanà et al., 2007;Gayoso et al., 2011;Braubach et al., 2012), and, moreover, we can compare mercury data with those concerning other metals (Lazzari et al., , 2019.

Animals
The study was performed on 40 adult zebrafish. Fish of both sexes, ∼6 months old and about 4 cm long were purchased from the Coral Aquarium, Bologna, Italy. They were acclimatized in tanks filled with a 1:2 mixture of dechlorinated tap and distilled water for 30 days, at 25°C in a 12:12 light-dark cycle. Automatic fish feeders distributed commercial food twice a day. All procedures were in accordance with the guidelines of European Communities Council Directive (86/609/CEE), the current Italian legislation regarding the use and care of animals, and the guidelines issued by the US National Institutes of Health. This study was approved by the Scientific Ethics Committee of the University of Bologna (protocol no. 17/79/2014). Cruz et al. (2013) listed some studies reporting the mercury concentration of 20 μg/L found in the aquatic environment. In their study on zebrafish larvae, Sun et al. (2018) applied the environmentally relevant concentration of 16 μg/L Hg 2+ . In the present study, we decided to use Hg 2+ concentrations based on the lethal concentration 50 (LC 50 ) and not possible environmental values because they have a large range of variability (see Fleck et al., 2016 for an estimation of the total concentration of mercury in different aquatic sites in Western North America). In the US EPA ECOTOX database (http://cfpub.epa.gov/ecotox/), LC 50 for mercury chloride measured in adult zebrafish (6-month old) after 48 h exposure is >60-100 μg/L (Bresch, 1982). In the present study, we used three Hg 2+ concentrations made from mercuric chloride (Merck, Darmstadt, Germany) dissolved in acclimatization water: two concentrations were supposed to be sublethal: low quantity (LQ), 0.25 μM of mercuric chloride, corresponding to 50 μg/L of Hg 2+ ; medium quantity (MQ), 0.5 μM of mercuric chloride, corresponding to 100 μg/L of Hg 2+ . The third concentration was over the LC 50 range found by Bresch (1982) and Wang et al. (2013a): high quantity (HQ), 1 μM of mercuric chloride, corresponding to 200 μg/L of Hg 2+ . However, Ung et al. (2010) determined that a 4-day exposure to 300 μg/L of mercuric chloride (∼220 μg/L of Hg 2+ ) induces mortality in less than 50% of adult zebrafish (30%). Due to a certain discrepancy in lethal values in literature (Vutukuru & Basani, 2013;Wang et al., 2013a;Macirella et al., 2016;Amorim et al., 2017), we decided to take into account also the concentration of 1 μM HgCl 2 . Each of three 10 L tanks containing 50, 100, and 200 μg/L of Hg 2+ respectively, housed 10 fish. According to a static/renewal system, fish were moved to new tanks containing freshly prepared Hg 2+ solutions every 24 h. The effect of the three concentrations of Hg 2+ on olfactory mucosa was evaluated in fish sacrificed after 4-day (96 h) exposure, according to standard procedure in acute toxicity tests. Ten control fish were hosted in a tank containing acclimatization water, changed every 24 h. Control animals were sacrificed at the same time as mercury-exposed fish. Table 1 reports representative values of water parameters for acclimatization, treatment, and control tanks.

Tissue
Zebrafish, previously anesthetized with 0.1% 3-aminobenzoic acid ethyl ester (MS-222, Sigma, St. Louis MO, USA), were sacrificed by decapitation. Brains were exposed after dorsal cranium removal. Heads were fixed by immersion in a modified Bouin's fixative solution (see Lazzari et al., 2019) for 24 h at room temperature. After multiple extended 0.1 M phosphate buffer washes, specimens were decalcified in 0.25 M buffered EDTA, pH 7.4, for 9 days at room temperature, dehydrated, and then embedded in Paraplast Plus (Leica Biosystems, Richmond, IL, USA; melting point 55-57°C). Five μm-thick serial frontal sections were collected on silanized slides (Sigma). Some sections were colored with hematoxylin and eosin. Immunohistochemical detection was carried out on adjacent slides.

Image Acquisition, Quantification, and Statistical Analysis
Except for PCNA, only sensory areas of the lamellae were sampled. Comparisons between fish were made selecting central lamellae from similar sections of the olfactory organ as it appears in frontal sections of the head (Fig. 1). Bright-field images of the olfactory organ were taken using an Olympus BH-2 microscope (Olympus Italia, Segrate, Italy) equipped with a BEL BlackL 5,000 digital camera (BEL Engineering, Monza, Italy). Light conditions were the same throughout acquired images. Micrographs were assembled with Adobe Photoshop CS3 (Adobe Systems, San Jose, CA, USA). Image processing was realized without content alteration. ImageJ 1.52o software (https://imagej.nih.gov/ij/) was employed in measurements of sensory lamellae. In anticalretinin immunostained sections, the thickness was measured from the epithelial basal lamina to the apical surface, excluding cilia and microvilli. Both labeling pattern and the presence of long cilia on the surface of nonsensory areas were used as characteristics to distinguish between sensory and nonsensory epithelial areas. We measured the areas (μ 2 ) covered by sensory and nonsensory epithelia in three lamellae per section using ImageJ 1.52o software, then we calculated and averaged the ratios of sensory to nonsensory epithelium areas.
A physical disector method was applied for evaluating the densities of TrkA-positive cells (Bettini et al., 2016) and PCNA-positive cells. Proliferating cells were counted in the central rosette, from the interlamellar curves to the sensory/nonsensory borders of each lamella ( Fig. 1), while the outer region was not considered, being not involved in neurogenesis (Bayramli et al., 2017). Anti-G α olf , anti-TRPC2, and anti-HuC/D immunostaining patterns did not permit cell discrimination and counting (see also Bettini et al., 2016;Lazzari et al., 2017Lazzari et al., , 2019, so we evaluated the Optical Density (OD) since the immunostaining intensity can be considered as an indirect index of the number of immunopositive cells (Iqbal & Byrd-Jacobs, 2010). In olfactory lamellae, average gray values of sensory regions and background-unstained zones were obtained by ImageJ analysis. The OD was then calculated as the logarithm of the ratio between gray values of background and sensory Region Of Interest.
In all measurements, data were obtained and averaged from three lamellae per section. For each antibody tested, mean values of fish groups were compared by ANOVA and LSD post hoc test. Using Excel 2016 (Microsoft Corporation, Redmond, WA, USA), bar graphs of mean values ± standard errors were obtained.

Results
Exposure of zebrafish to the highest Hg 2+ ion concentration for 96 h caused one fish death out of a sample of ten. At the highest concentrations, the only appreciable micro-anatomical/histological effect was an increase in the unicellular mucous glands opening onto the luminal surface of the olfactory lamellae (Fig. 2). Instead, tissue disruption and metaplasia were not detected in the olfactory lamellae that also maintained their spatial arrangement. Olfactory rosettes retained their morphology. In hematoxylin eosin-stained sections, light microscopy analysis of the lamellae with high magnification revealed that even the highest mercury concentration applied did not cause removal or significant reduction of cilia (Fig. 3).
Incubation with anti-calretinin antibody that stains cell bodies and dendrites of both mature cOSNs and mOSNs, resulted in clear immunopositivity ranging from the basal to the apical layers of the olfactory epithelium . When the whole olfactory rosette was examined, the OD of control was statistically greater than ODs related to the three Hg 2+ concentrations. Conversely, in the comparison between the three metal concentrations, only OD for LQ was significantly greater than OD for HQ ( Fig. 4e). In the dorsal half of the olfactory rosette, OD analysis revealed a condition similar to the whole rosette, with the addition of a significantly higher OD value for MQ compared to HQ (Fig. 4f). In the ventral half of the olfactory rosette, OD resembled what is found in the whole organ ( Fig. 4g). Image analysis of calretinin immunostained sections permits to evaluate two additional morphometric parameters in the olfactory lamellae: the thickness of the sensory epithelium and the ratio between sensory and nonsensory epithelium areas. Image analysis of the entire olfactory rosette, as well as of each dorsal and ventral half, showed no significant differences in the ratio sensory/nonsensory areas of the olfactory lamellae among the four groups of fish (Figs. 5a-5c).
Regarding the epithelium thickness, the value at HQ appeared significantly decreased compared to control, LQ, and MQ not only in the entire rosette but also in each dorsal and ventral half (Figs. 6a-6c). Moreover, in the whole rosette, the thickness of the MQ group was significantly decreased compared to control (Fig. 6a). In the dorsal half of the olfactory rosette, epithelium thickness of LQ, MQ, and HQ groups appeared significantly lower than in control ( Fig. 6b). At the free surface of the olfactory epithelium, anti-G α olf antibody marked apical dendrite endings and cilia of cOSNs (Figs. 7a-7d). When the entire olfactory rosette was evaluated, HQ and MQ showed OD values significantly decreased compared to control and LQ (Fig. 7e). In the dorsal half of the olfactory rosette, OD values repeat the condition found in the entire organ (Fig. 7f). In the ventral half of the olfactory rosette, OD values showed no statistically relevant difference (Fig. 7g).
Anti-TRPC2 antibody-stained dendritic apical knobs and microvilli of mOSNs at the apical region of the sensory epithelium of the olfactory rosettes (Figs. 8a-8d). In the whole olfactory rosette, as well as in both its dorsal and ventral halves, mercury exposure resulted in lower OD values than control (Figs. 8e-8g). In the entire rosette and its dorsal half, OD for HQ was statistically lower than values for both MQ and LC (Figs. 8e,8f). In the ventral half of the olfactory rosette, OD at HQ was significantly lower than LQ only (Fig. 8g).
Anti-TrkA antibody clearly stained crypt cells, big neurons placed in the olfactory epithelium region close to the olfactory lamella surface (Figs. 9a-9d). In the whole olfactory rosette, crypt cell number for MQ and HQ was lower compared to control and LQ (Fig. 9e). In the dorsal half of the olfactory rosette, both MQ and HQ groups show a crypt cell density significantly smaller than in the LQ group (Fig. 9f). Ventrally, cell density showed a similar situation, with values for MQ and HQ lower than control and progressively decreasing with more severe exposure conditions (Fig. 9g).
Anti-HuC/D antibody, specific for both mature and immature neurons, stained the olfactory epithelium from the basal to the apical layer (Figs. 10a-10d). In the whole rosette, as well as in both its dorsal and ventral halves, no significant change in OD was observed between the four fish groups (Figs. 10e-10g).
In the olfactory lamellae, anti-PCNA stained cells undergoing mitotic events reside specifically in the regions bordering the sensory epithelium and also in its basal layer . In the whole rosette, as well as in its dorsal half, exposure to Hg 2+ showed no significant differences between the four groups of fish (Figs. 11e,11f). When the statistical analysis was restricted to the ventral half of the rosette, Hg 2+ exposure resulted in a significant increase of mitotic cell density between HQ and the three other groups in the basal layer of the sensory region, while interlamellar curves and the border between sensory and nonsensory epithelium appeared not to be affected (Fig. 11g).

Discussion
In the olfactory epithelium of zebrafish, the immunohistochemical markers investigated allowed the identification of three main populations of OSNs with differential response to Hg 2+ ions at different concentrations. In the whole olfactory rosette, the OD of G α olf , indicative for cOSNs, results significantly reduced compared to control only for MQ (22%) and HQ (30%). When compared to each other, the OD of cOSNs at the three used Hg 2+ concentrations shows values for MQ and HQ significantly lower than LQ (17 and 26%, respectively), indicative for dose dependency. The response of cOSNs to MQ and HQ exposure results dose-independent since OD differences between MQ and HQ turn out to be statistically nonsignificant. The response of cOSNs appears dose-independent also for Ni 2+ exposure, as previously reported (Lazzari et al., 2019). When the dorsal and ventral half of the olfactory rosette are separately evaluated after Hg 2+ exposure, the condition of the OD for the dorsal region resembles the outcome for the entire organ. On the contrary, the ventral region of the olfactory rosette shows no significant differences between the four groups of fish. The reasons, as well as the functional relevance of these regional differences in the olfactory organ, are still unclear and require detailed studies. A hypothesis could involve different dynamics of water circulation inside the olfactory chamber, which could differentially expose OSNs to pollutants dissolved in the water. Indeed, in the ventral region of the olfactory rosette, cOSNs could be somehow less exposed to pollutants, as, for example, the water flow entering the olfactory chamber encounters first the lamellae of the dorsal half of the olfactory rosette. Also, different subpopulations of cOSNs, having different sensitivity to Hg 2+ ions, may be present in different regions of the olfactory organ. By comparing the results on Ni 2+ and Hg 2+ exposures, we may suppose that cOSNs have higher Ni 2+ susceptibility in the ventral zone of the olfactory rosette, whereas Hg 2+ susceptibility prevails in the dorsal zone (Lazzari et al., 2019). Subpopulations of olfactory neurons appear to exist also in other fish species as in the case of crypt cells in Poecilia reticulata .
In the whole olfactory rosette after Hg 2+ exposure, OD for TRPC2, indicative for mOSNs, shows a statistically significant decrease at each metal concentration tested, with a mean value of 49% compared to control. OD values for LQ and MQ are statistically equivalent, but both of them show relevant dose effects compared to HQ. Therefore, similarly to that obtained for G α olf (cOSNs), the response for TRPC2 (mOSNs) is greater in the ventral region. The comparison between mean OD values for G α olf and TRPC2 suggests that mOSNs are more sensitive to Hg 2+ exposure than cOSNs. Since, instead, cOSNs are more susceptible than mOSNs when exposed to Cu 2+ and Ni 2+ (Lazzari et al., , 2019, this new finding suggests a difference in the response of OSNs to different heavy metals. Hentig & Byrd-Jacobs (2016) studied the effects of zinc on zebrafish OSNs with both morphological and functional approaches, showing differential impairment of fish OSN classes. In particular, ultrastructural analysis of epithelial surface with scanning electron microscopy showed that cOSNs are heavily affected, whereas mOSNs generally have no or smaller damage. Behavioral findings were in accordance with morphological evidence. In fact, after zinc exposure, the sensibility to amino acids, detected by mOSNs and initiating feeding behavior, was restored more quickly than the response to bile salts, detected by cOSNs and related to social cues. As suggested by Hentig & Byrd-Jacobs (2016), mOSNs may be considered a highly protected component of the olfactory system because they are connected to feeding response and satisfy an immediate need for the individual; instead, cOSNs, that mediate social behavior, may be seen to meet a less urgent necessity for individual survival and may therefore be more subject to damage. Thus, the observed variation can be seen as the outcome of diverse evolutionary forces affecting the OSN populations differently.
Differential impairment of fish OSN classes, even if with opposite findings, was reported applying a functional approach based on electro-olfactograms (EOG) after nickel exposure . EOG study on fathead minnows (Pimephales promelas) and yellow perch (Perca flavescens) shows that, in both species, nickel impairs mOSN function whereas there was no significant alteration of cOSN activity. The observed impairment caused by Ni on mOSN function could still represent a success for the individual if it forced fish to swim away from polluted areas in search of food. Based on the above reported functional studies, we may suggest that the preservation of mOSN activity cannot be considered a general feature but can vary after exposure to different metals and/or if considering different fish species.
The results with EOG on nickel exposure of Dew et al. (2014) are in contrast with our findings in zebrafish (Lazzari et al., 2019), since immunohistochemistry revealed that both mOSNs and cOSNs are affected by Ni exposure, with OD reduction for G α olf (specific for cOSNs) greater than OD reduction for TRPC2 (specific for mOSNs) (Lazzari et al., 2019). As a possible explanation for the observed difference between immunohistochemical and EOG results, we may suggest that, after metal exposure, OD reduction for each antigen can have a specific threshold beyond which a functional damage occurs.  (b) Fish exposed to 200 μg/L Hg 2+ . Exposure to high mercury concentration caused an increase in the number of unicellular mucous glands opening into the luminal surface of the olfactory lamellae. Olfactory lamellae retain their structural integrity in fish exposed to high mercury concentration. Scale bar = 50 μm. reduction of their OD values, but, after Ni 2+ exposure, OD reduction for TRPC2 might exceed its damage-threshold value thus that EOG could indeed reveal alterations in mOSN activities. On the contrary, at the same Ni 2+ concentrations, the reduction of OD for G α olf might remain below its damage-threshold with consequently no effects detected in cOSN functions by EOG.
In immunohistochemical analysis, we reported that cOSNs appear more susceptible to Ni 2+ than mOSNs because cOSNs have a greater percentage decrease of OD, compared to their control value, than mOSNs. Nevertheless, this percentage shift might be not sufficient to trigger functional alterations in cOSNs. Instead, in mOSNs, a smaller reduction of OD for G α olf could Fig. 4. Calretinin immunohistochemistry in zebrafish olfactory epithelium before and after 50, 100, and 200 μg/L Hg 2+ exposure. (a) Heavy anti-calretinin immunoreaction stains the sensory area (S) of the olfactory lamellae of control fish. The nonsensory region (NS) of the lamellae appears unstained, except for some scattered cells (even if the two areas are well distinguished, there is no barrier between them, so some sensory cells can migrate and differentiate in the adjacent nonsensory region). (b) Representative micrograph of lamellae exposed to 50 μg/L of Hg 2+ . (c) Representative micrograph of lamellae exposed to 100 μg/L of Hg 2+ . (d) Representative micrograph of lamellae exposed to 200 μg/L of Hg 2+ . Scale bar = 50 μm. (e) Optical density analysis of the olfactory epithelium in the whole olfactory rosette. (f) Optical density analysis of the olfactory epithelium in lamellae of the dorsal half of the olfactory rosette. (g) Optical density analysis of the olfactory epithelium covering lamellae of the ventral portion of the olfactory rosette. C, control, 0 μg/L Hg 2+ ; LQ, low quantity, 50 μg/L Hg 2+ ; MQ, medium quantity, 100 μg/L Hg 2+ ; HQ, high quantity, 200 μg/L Hg 2+ . Asterisks indicate significant differences: *, p < 0.05; **, p < 0.01. be sufficient to exceed the damage-threshold value and generate functional damage. Interestingly, EOG recordings on copperexposed fish  are in accordance with the histological analysis in goldfish (Kolmakov et al., 2009) and zebrafish , with copper specifically affecting cOSNs. Taking all these observations into consideration, we believe that, in studies on metal toxicity, attention must be paid when comparing data obtained with the two types of investigations, since the changes detected by immunohistochemical analysis cannot be always directly compared to functional results. For completeness, it should be mentioned that incongruities among EOG studies also exist (Sandahl et al., 2007;Baldwin et al., 2011). Differences in technical approaches, ions concentrations, exposition times, water chemistry, and intrinsic differences among fish species can explain the discrepancy reported in different electrophysiological investigations ( Another problem concerns the significance of the decrease in immunostaining intensity observed by densitometric analysis at the level of the apical processes of OSNs. This decrement may be indirect evidence of the reduction in the number of OSNs, therefore indicative of cell death. However, OD decrement could also derive from both a reduction in apical process number and a decrease in the expression of the protein chosen as a molecular marker. In the Indian major carp (Labeo rohita) exposed to sublethal concentrations of HgCl 2 , Ghosh & Mandal (2014) analyzed the apical surface of the olfactory epithelium by scanning electron microscopy. These authors report toxicity responses in fish olfactory epithelium, in particular clumping and loss of cilia. Damage and loss of cilia with retention of microvilli were also described by Ribeiro et al. (1995) in Trichomycterus brasiliensis after Hg 2+ exposure and by Kolmakov et al. (2009) in goldfish after Cu 2+ treatment. However, ultrastructural studies on zebrafish olfactory epithelium exposed to Hg 2+ ions are still missing. These morphological studies are important since they could prove whether, also in zebrafish, exposure to mercury leads to cilia and/or microvilli reduction or complete disappearance of apical cell processes with probable cell death. In our immunohistochemical analysis, cell cilia are clearly present and stained in the apical/ luminal surface of the olfactory epithelium. Therefore, in zebrafish exposed to the Hg 2+ concentration tested, a reduction in G α olf and TRPC2 expression could explain the decrement in OD values for these two markers on the still visible apical processes. This hypothesis would be in line with studies reporting that Cu 2+ exposure could cause epigenetic modifications (Tilton et al., 2008) and down-regulation of transcription of genes involved in olfactory signal transduction (Wang et al., 2013b) without morphological damages and cell death. However, since, in the present study, Hg 2+ exposure is also linked to a reduction in epithelium thickness and cell proliferation activity, we suggest that the reduction in the marker expressions can be in part related to cell death, as we hypothesized in a previous study on Cu 2+ exposure . Crypt cells are the third main population of OSNs located in the sensory epithelium covering the olfactory lamellae of fish Representative micrograph of lamellae exposed to 50 μg/L of Hg 2+ . (c) Representative micrograph of lamellae exposed to 100 μg/L of Hg 2+ . (d) Representative micrograph of lamellae exposed to 200 μg/L of Hg 2+ . Scale bar = 50 μm. (e) Optical density analysis of the olfactory epithelium surface in the whole olfactory rosette. (f) Optical density analysis of the olfactory epithelium surface in the lamellae of the dorsal half of the olfactory rosette. (g) Optical density analysis of the olfactory epithelium surface in the lamellae of the ventral half of the olfactory rosette. C, control, 0 μg/L Hg 2+ ; LQ, low quantity, 50 μg/L Hg 2+ ; MQ, medium quantity, 100 μg/L Hg 2+ ; HQ, high quantity, 200 μg/L Hg 2+ . Asterisks indicate significant differences: *, p < 0.05; **, p < 0.01. olfactory rosette (Hansen & Zeiske, 1998;Hansen & Finger, 2000;Ferrando et al., 2006;Lazzari et al., 2007;Gayoso et al., 2011Gayoso et al., , 2012. This cell type is not present in the olfactory epithelium of tetrapods. For the localization of crypt cells, TrkA is the molecular marker now recognized as being specific for crypt cells of zebrafish (Catania et al., 2003;Ahuja et al., 2013;Bettini et al., 2016). Therefore, we used anti-TrkA in our studies on the cytological effects of heavy metals on crypt cells (Bettini et al., 2016;Lazzari et al., 2017Lazzari et al., , 2019. In the whole olfactory rosette, exposure to MQ and HQ Hg 2+ causes a great decrease in crypt cell density, respectively of 49 and 67% compared to control. In the comparisons between each other, the cell density at the three Hg 2+ concentrations applied shows values at MQ and HQ statistically lower than LQ (51 and 69%, respectively), indicating dose dependence. Conversely, when MQ and HQ exposure were compared, crypt cell density appears not significantly different, thus excluding dose dependence. Mercury concentration between LQ and MQ, and higher than HQ would be useful to investigate the existence of other dose-dependence conditions. When the dorsal and ventral regions of the olfactory rosette are examined separately, Hg 2+ exposure affects both regions. However, a greater reduction appears in the ventral zone, with crypt cell densities at MQ and HQ of 56 and 75% lower than control, respectively. These density values, even lower than the measures for the whole olfactory rosette, indicate that the ventral half of the olfactory organ is severely affected by Hg 2+ ions. Previous studies evaluated the response of crypt cells to Cu 2+ and Ni 2+ exposures (Lazzari et al., , 2019. In both studies, the statistical analysis does not support any significant difference after metal exposure. Therefore, the conclusion was that Cu 2+ and Ni 2+ , at least at the tested concentrations, do not seem to affect crypt cells (Lazzari et al., , 2019. In these studies on Cu 2+ and Ni 2+ exposure, the authors hypothesized that the specific morphology of crypt cells can account for the lack of harmful effects on these cells. cOSNs and mOSNs have their cilia and microvilli arising from the dendritic apical region of the receptor neurons and are directly exposed to the water flux at the surface of the olfactory epithelium. This condition may account for the great susceptibility of cilia and microvilli to water pollutants. On the contrary, in crypt cells, both cilia and microvilli are present, but not directly immersed in the water flux inside the olfactory chamber (Lazzari et al., 2007). In crypt cells, their cytoplasmic processes are included in a deep cavity at the apical zone of the cell. The narrow crypt hole opening into the olfactory chamber could be responsible for limiting the passage of substances, pollutants included, towards the inner cilia and microvilli. In the present study, a clear effect on crypt cells density could be related to a higher toxicity of mercury compared to copper and nickel. In vertebrate olfactory epithelium, OSNs have a limited life span. Throughout life, mitotic proliferation takes place in the basal layer of the olfactory epithelium and next cell differentiation gives rise to new neurons substituting dead OSNs (Graziadei & Graziadei, 1979;Bettini et al., 2006a). Under normal conditions, OSN populations remain unvaried because mature OSNs downregulate proliferation and differentiation that would substitute deceased elements (Wu et al., 2003). In the olfactory epithelium exposed to Hg 2+ , the OD for HuC/D, indicative for the total neuronal population including undifferentiated elements, remains statistically constant revealing that substitution of dead cells is still working properly. On the other hand, the density of proliferating cells measured by anti-PCNA immunohistochemistry shows a tendency to increase that becomes statistically relevant only in the ventral half of the olfactory rosette where OD of TRPC2 and TrkA show a significant reduction. These immunohistochemical analyses appear to support each other in suggesting that a certain degree of cell death occurs in the populations of mature OSNs after Hg 2+ exposure.
Another consideration regards the regions involved in active neurogenesis. Bayramli et al. (2017) and Calvo-Ochoa et al. (2021) reported that the main sites of the proliferation of new olfactory neurons are the interlamellar curve and the sensory/ nonsensory border, while the sensory epithelium does not contribute to constitutive cell turnover. mOSNs and cOSNs originate from both marginal sites; nonetheless, cOSNs seem to be prevalently generated in the interlamellar site, while mOSNs in the sensory/nonsensory border (Bayramli et al., 2017). As a consequence, we could expect some density variations in the two parts of the olfactory organ, to regenerate the cOSNs lost in the dorsal rosette and the mOSNs in the ventral one. However, we did not observe a significant difference among treatments in the proliferative activity in these regions, while the increment is statistically evident in the basal layer of the sensory epithelium. Our finding is in line with Calvo-Ochoa et al. (2021), who reported that, under regenerative conditions in response to damage, the horizontal basal cells (HBCs) in the basal layer of the olfactory sensory epithelium, normally dormant, are activated and start proliferating to expand the population of globose basal progenitor cells (GBCs).
We analyzed three types of OSNs that are intermingled in the OE. However, in zebrafish, it is known that different olfactory receptors segregate into distinct spatial subdomains (Weth et al., 1996), and, recently, Ahuja et al. (2018) showed also distinctly different expression zones for some zebrafish V2R-related OlfC receptors. It could be interesting, even if beyond the scope of the present study, to analyze the effects of toxicants on the different subdomains, examining also if the migration of progenitors from the proliferative zones follows different pathways, as observed by Bayramli et al. (2017).
According to Ghosh & Mandal (2014), we found that mercury exposure causes a proliferation and enlargement of round mucous cells in the olfactory epithelium of zebrafish. Most of them are opening to the luminal surface of the epithelium, showing that an enhancement in mucous secretion occurs. Acetylcholine secretion may be responsible for the upregulation of mucous secretion (Inglis et al., 1997). Mucous film over the olfactory epithelium surface has the function of binding odorants and ions involved in the transduction of olfactory signals. However, when the mucous layer becomes excessive, it prevents pollutants from directly contacting the epithelial surface and decreases toxicity effects (Ghosh & Mandal, 2014), and the mucous can also bind and eliminate heavy metals from the system (Ray et al., 2012).
In the olfactory organ of Indian major carp, another manifestation of toxic effect is the significant increase of the olfactory epithelium thickness due to hyperplasia of basal cells and the appearance of vacuoles (Ghosh & Mandal, 2014). At the higher concentration tested (132 μg/L for 30 days), Ghosh & Mandal (2014) found severe decrement of OSNs and tissue alteration with epithelium thickening due to hyperplasia of epithelial basal cells unable to differentiate in new OSNs and cyst formation. In the olfactory epithelium of zebrafish, as indicated by a limited increase in the density of anti-PCNA immunopositive cells, we suggest that mercury exposure affects less the normal cell cycle regulation. In zebrafish, the decrease in the OD for calretinin, which is indicative for a heterogeneous population of mature cOSNs and mOSNs, is in agreement with the decrease in OD for G α olf and TRPC2.
Although no specific study is at present available on the effects of Hg 2+ on nonneuronal cells of the olfactory epithelium, the lack of severe histological alterations throughout tissue appears in favor of a predominant involvement of OSNs, at least at the Hg 2+ concentration tested. This is in agreement with previous studies on the effect of Cu 2+ and Ni 2+ on zebrafish olfactory organs (Lazzari et al., , 2019. As for Ni 2+ exposure, in contrast to Cu 2+ treatment, zebrafish olfactory epithelium does not show any increase in sensory area compared to control by the action of Hg 2+ . Behavioral and electrophysiological studies of Cu 2+ action in Colorado pikeminnow (Ptychocheilus lucius) (Beyers & Farmer, 2001) and fathead minnow (Pimephales promelas) (Dew et al., 2012) suggested that sensory area growth was associated with functional recovery. In this respect, new specific studies could clarify existence, extension, and process of neuronal recovery following Hg 2+ treatment.
In our experiments, according to the sublethal level of the tested concentrations, only HQ exposure caused a fish mortality rate of 10%. This is consistent with a 30-35% death rate at 220 μg/L HgCl 2 reported by Ung et al. (2010).
Little is known about the mechanism of olfaction impairment by mercury. Passow et al. (1961) attributed the inhibition of active transport and other cell functions to mercury affinity for sulfhydryl groups on the cell membrane. Following mercury chloride exposure, Richetti et al. (2011) hypothesize a signal transmission impairment through alterations in cholinergic transmission and in the antioxidant competence of nervous tissue.

Conclusions
The immunohistochemical analysis of the olfactory epithelium shows that Hg 2+ exposure clearly affects the expression pattern of the molecular markers of the different OSN types.
Interestingly, the different OSNs respond differentially to Hg 2+ exposure at the sublethal concentration tested. In particular, mOSNs are the most susceptible with a mean reduction of TRPC2 OD of 49% compared to control, against a 19% decrease of G α olf OD for cOSNs, and a 37% drop of crypt cell density measured by anti-TrkA staining. These variations show some dosedependent effects, in particular for TRPC2 OD values. Therefore, immunohistochemical analysis differentiates OSNs by Hg 2+ sensitivity: mOSNs, crypt cells, and cOSNs, going from the most susceptible to the least. These immunohistochemical results differ from studies on fish exposure to other metals, in which cOSNs resulted more susceptible than mOSNs (Kolmakov et al., 2009;Dew et al., 2014;Ghosh & Mandal, 2014;Lazzari et al., 2017Lazzari et al., , 2019. Moreover, in zebrafish, crypt cells appeared not statistically modified by Cu 2+ and Ni 2+ exposures (Lazzari et al., , 2019, whereas the effect of Hg 2+ on crypt cells is evident also at the lowest concentration. In zebrafish, Hg 2+ exposure revealed that cOSNs were more susceptible in the dorsal region, whereas mOSNs were more affected in the ventral area; this is different from the response to Ni 2+ exposure that was greater in the ventral half of zebrafish olfactory organ for both cOSNs and mOSNs (Lazzari et al., 2019). We can hypothesize that different subpopulations of receptor cells inside each main type of OSNs may be responsible for the detected regional differences. This hypothesis needs further investigation.
Despite the changes in tissue staining described in this study, immunohistochemical methods, per se, do not provide information on functional and ecological outcomes of metal exposure. Different functions characterize cOSNs, mOSNs, and crypt cells (Sato & Suzuki, 2001;Hamdani & Døving, 2006;Vielma et al., 2008;Bazáes & Schmachtenberg, 2012;Ahuja et al., 2013;Biechl et al., 2016). Various studies concern behavioral effects resulting from Cu 2+ exposure (Scott & Sloman, 2004;Tilton et al., 2011;Svecevičius, 2012;Poulsen et al., 2014;Sovová et al., 2014;da Silva Acosta et al., 2016;Simonato et al., 2016) and Ni 2+ treatment (Dew et al., , 2016. It should also be pointed out that, at present, no molecule acting as a specific activator of crypt cells is known, and this complicates the set up of behavioral investigations for this cell type. Also, when considering the behavioral implications of contaminants on the overall olfactory sensing of fish, we should remember that olfactory function impairment could reasonably derive from a combination of effects of pollutants on OSN types, on sensory integration circuits, and on effector organs (Abu Bakar et al., 2017;Amorim et al., 2017). Further studies analyzing transcriptomic profiles of the olfactory mucosa under different exposure conditions, as well as electro-olfactographic and behavioral effects resulting from mercury exposure, will significantly increase our knowledge about the mechanism of action of this metal in fish olfaction impairment.
Financial support. This work was supported by national public funds grant RFO2017FRANCESCHINI from the Italian Ministry of University and Research (MUR).
Conflicts of interest. The authors declare that they have no conflicts of interest.