Brain Aromatase Modulates Serotonergic Neuron by Regulating Serotonin Levels in Zebrafish Embryos and Larvae

Teleost fish are known to express two isoforms of P450 aromatase, a key enzyme for estrogen synthesis. One of the isoforms, brain aromatase (AroB), cyp19a1b, is highly expressed during early development of zebrafish, thereby suggesting its role in brain development. On the other hand, early development of serotonergic neuron, one of the major monoamine neurons, is considered to play an important role in neurogenesis. Therefore, in this study, we investigated the role of AroB in development of serotonergic neuron by testing the effects of (1) estradiol (E2) exposure and (2) morpholino (MO)-mediated AroB knockdown. When embryos were exposed to E2, the effects were biphasic. The low dose of E2 (0.005 µM) significantly increased serotonin (5-HT) positive area at 48 hour post-fertilization (hpf) detected by immunohistochemistry and relative mRNA levels of tryptophan hydroxylase isoforms (tph1a, tph1b, and tph2) at 96 hpf measured by semi-quantitative PCR. To test the effects on serotonin transmission, heart rate and thigmotaxis, an indicator of anxiety, were analyzed. The low dose also significantly increased heart rate at 48 hpf and decreased thigmotaxis. The high dose of E2 (1 µM) exhibited opposite effects in all parameters. The effects of both low and high doses were reversed by addition of estrogen receptor (ER) blocker, ICI 182,780, thereby suggesting that the effects were mediated through ER. When AroB MO was injected to fertilized eggs, 5-HT-positive area was significantly decreased, while the significant decrease in relative tph mRNA levels was found only with tph2 but not with two other isoforms. AroB MO also decreased heart rate and increased thigmotaxis. All the effects were rescued by co-injection with AroB mRNA and by exposure to E2. Taken together, this study demonstrates the role of brain aromatase in development of serotonergic neuron in zebrafish embryos and larvae, implying that brain-formed estrogen is an important factor to sustain early development of serotonergic neuron.

functions. Fish brain is characterized by having much higher aromatase expression in brain compared to mammals (5). At the same time, fish brain has been reported to exhibit elevated neuroregenerative capacity compared to mammals (6)(7)(8). Wide spread proliferation zones are detected in zebrafish brain (6,9), while only limited areas such as subependymal and subgranular zones exhibit proliferation in mammals (7). Such high neurogenic activity in teleost fish may be attributed to increased synthesis of estrogen due to the elevated expression of brain aromatase. Indeed, expression of brain aromatase is localized in radial glial cells (RGCs), which differentiate into neurons and other glial cells contributing to adult neurogenesis as well as developmental neurogenesis (10)(11)(12). Developmental studies in zebrafish show that expression of brain aromatase in embryos increases rapidly after 12 hour postfertilization (hpf), and is regulated by posi tive feedback loop through its own product, estrogen, acting on estrogen response element of cyp19a1b (3,13,14). Therefore, the zebrafish model expressing elevated levels of brain aromatase in early development is suitable to investigate the functional sig nificance of aromatase and neural estrogen in developing brain.
Serotonin (5HT), a neurotransmitter produced by multiple enzymatic steps including a ratelimiting action of tryptophan hydroxylase (TPH), plays a major role in a number of physio logical processes and pathological conditions, such as depression (15,16), stress (15,17), cardiac function (18), reward seeking behavior (19), and anxiety (15,20). In addition, serotonergic neuron is known to be involved in neurogenic activities (21). It has been reported that 5HT is critically involved in the brain plasticity, neural trafficking, synapse formation, and network construction during development (22,23). Serotonergic neurons in raphe nuclei extend their axons to the forebrain possibly modulating the differentiation of neuronal progenitors (24). Early ontogeny of serotonergic system may further suggest its role in brain development (25). Raphe 5HT populations in human brain are considered as the earliest to be identified (24,26).
Serotonergic neurons in mammalian brain are localized mainly in raphe nuclei of brain stem, which project into accumbens, hypothalamus, substantia nigra, and periaqueductal gray (22,23). On the other hand, 5HTpositive cell bodies are detected mainly in three populations in adult fish brain: pretectal area, posterior tuberculum/hypothalamus, and raphe (27,28). Interestingly, distributions of serotonergic populations and their fibers overlap with highly proliferative areas of fish brain, which may indicate serotonergic regulation in adult neurogenesis in fish (27). In adult zebrafish, serotonin has been shown to promote regeneration of motor neurons by acting on progenitor cells (29).
It is well documented that serotonergic neuron is one of the targets of estrogen in mammals (30,31). In macaques, estrogen increases gene expression and protein contents for TPH (32), and decreases gene expression of the serotonin reuptake trans porter and the 5HT1A autoreceptor (33,34). In mammals, both ERα and ERβ are expressed in 5HT neurons with differential distributions depending on species and sex (35-37). ERβ has been shown to regulate tph2 expression in serotonergic neurons (38,39). Similarly in teleost fish, effects of ovarian steroids on serotonin system have been reported in some species. In tilapia, the response of 5HT content in brain to E2 treatment was dependent on developmental stages. Treatment between days 7 and 10 posthatching decreased 5HT content, while the treat ment at later stages increased it (40). Similar result was obtained in Japanese sea bass, which shows a significant decrease in brain 5HT content in fingerlings after E2 treatment, while the content increased in juvenile group (41). Indeed, overlapping distribu tions of ER with raphe 5HT innervation in telencephalon and diencephalon of adult zebrafish brain implies close association of ER and serotonergic neurons (27,42). It has been reported that ERβ exhibits broad distribution along the brain ventricles of telencephalon and diencephalon in adult zebrafish (43), though colocalization of ER in serotonergic neurons has yet to be docu mented in fish.
Therefore, in this study, we tested the hypothesis that brain aromatase modulates serotonergic neuron in early development of zebrafish. In order to elucidate a possible role of brainformed estrogen, we first examined the effects of exogenous E2 and then MOmediated knockdown of brain aromatase on parameters such as 5HT contents, relative tph expression levels, heart rate, and thigmotaxis in zebrafish embryos and larvae.

Fish Maintenance and embryo culture
Adult zebrafish (Danio rerio) were obtained from the local pet shop and reared in a 60L tank. Water temperature was maintained at 26-30°C, and the light regime was 14 h of light starting at 10:00 fol lowed by 10 h of dark. Fish were fed with TetraMin (Tetra Japan Inc.) twice a day. Fertilized eggs were collected within 15 min after fertilization and washed in embryo medium (EM) (0.004% CaCl2, 0.163% MgSO4, 0.1% NaCl, and 0.003% KCl) to remove debris. Embryos were transferred to a 6well plastic plate (30 embryos in 8 mL of EM per well), and incubated at 28 ± 0.5°C. The medium was changed daily. All experimental procedures and maintenance of fish were conducted in accordance with the Guide for Care and Use of Laboratory Animals published by the US National Institutes of Health.

exposure experiments
Stock solutions of 17βestradiol (E2) (SigmaAldrich) at 10 mM, ICI 182,780 (ICI) (Tocris Bioscience) at 10 mM, and dexa methasone (DEX) (Wako) at 100 mM were prepared in dimethyl sulfoxide (DMSO), and diluted with EM to the final concentra tions indicated in the experiments. Quipazine maleate salt (Q) (SigmaAldrich) and fluoxetine hydrochloride (FLX) (Wako) were dissolved in ethanol at 100 and 10 mM, respectively, which were further diluted with EM to the final concentrations used in the experiments. Control embryos were cultured in 0.1% DMSO or ethanol. Exposure started at 2 hpf and continued till embryos and larvae were subjected to the assays. The media were changed daily.

Morpholino (MO) Microinjection
Morpholino antisense oligos were purchased from Gene Tool. MO sequences are shown in Table 1. MOs were dissolved in dis tilled water to 50 mg/mL and stored at −20°C. Before injection,  Step PrimeScript RTPCR Kit (Takara) using total RNA from 7dpf zebrafish larvae and AroB primers ( Table 2). Amplified products were purified with NucleoSpin Gel and PCR Cleanup (MacheryNagel) and subcloned into pGEMT Easy Vector (Promega). Nucleotide sequences and ori entation of the inserts were verified by DNA sequencing analysis carried out using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) and ABI 3130 xl genetic analyzer (Applied Biosystems). Plasmid DNA was linearized with SalI and the full length AroB mRNA was transcribed in vitro by MAXIscript T7 Kit (Ambion).

Western Blot and Dot Blot analysis
The antiserum to brain aromatase was produced in a rabbit against the synthetic peptide, CNSNGETADNRTSKE of zebrafish AroB (SigmaGenosys). This peptide sequence has been used to raise the specific antibody as previously described (47). To confirm the specificity of the antiserum, Western blot of brain extract was conducted. Adult female zebrafish were exposed to E2 (5 and 25 ng/L) or vehicle alone (0.00025% DMSO) for 24 h (three fish per group). Brains were pooled and homogenized in HBST buffer (100 mM NaCl, 10 mM HEPES, 0.5% TritonX 100, 0.01% TPCK, and 0.01% TLCK). After centrifugation at 10,000 g for 10 min, protein concentrations in supernatant were measured using BCA protein assay kit (Thermo Scientific). Extracts (30 µg protein/ sample) were separated on 12.5% SDSPAGE and transferred to a PVDF membrane. Precision Plus Protein Unstained Standards (BioRad) were used for size reference. After blocking by 1% skim milk in PBS for 1 h, the membrane was incubated with the AroB antiserum (1:500) for 2 h, and then with the secondary antibody conjugated with alkaline phosphatase (AP) (Abcam) (1:1,000) for 1 h. After washing, the membrane was incubated in AP buffer (0.1 M Tris-HCl, pH 9.5, 0.1 M NaCl, 1 M MgCl) for 1 min. Signals were developed for 2-3 min in BCIP/NBT substrate (Roche) diluted at 1:50 in AP buffer, and the reaction was stopped by 0.5 M EDTA. All the incubation steps were done at RT. The antiserum to ovarian aromatase was raised in a rabbit using a syn thetic peptide, CKPDVYFRLDWLHKKHKRD of zebrafish AroA (SigmaGenosys). Similarly, Western blot with the antiserum (1:500) was performed using the ovarian extract prepared with HBST buffer from three adult fish (30 μg/lane).
To examine the effect of MOmediated AroB knockdown, dot blot analysis using 120 larvae at 6 dpf collected from 4 separate MO injection experiments were pooled and extracted similarly as described for the brain extract. Extracts containing 40 µg protein (3 µL) were spotted onto nitrocellulose membrane (GVS Life Science). The membrane was treated similarly as in Western blot except for the concentration of the secondary antibody at 1:2,000. Density of the blots were analyzed with NIH ImageJ software. Blots of embryo extracts treated with the preimmunized rabbit serum were used as a negative control to subtract from the density obtained with the AroB antiserum. No changes were observed among controls (uninjected, standard control MO, and inverted AroB MO) (data not shown). The effect of MOmediated AroA knockdown was also examined by dot blot analysis. Briefly, pooled 120 embryos at 2 dpf collected from 5 separate embryo cohorts were extracted. Extracts containing 30 µg protein (3 µL) were spotted onto nitrocellulose membrane and subjected to immunostaining using AroA antiserum at 1:500. No changes were observed among controls including inverted AroA MO (data not shown).

5-hT immunohistochemistry
Wholemount immunohistochemistry for 5HT was carried out according to the previous studies (48, 49). 2dpf embryos were fixed in 4% paraformaldehyde in PBS overnight at 4°C. Fixed embryos were rinsed in PBS, bleached in 3% H2O2 for 30 min and stored in methanol at −20°C until use. For immunostaining, embryos were washed in PBS containing 0.1% Tween20 and 0.5% Triton X100 (PBSTX), and then permeabilization was achieved by incubation in deionized H2O for 60 min at RT followed by 100% acetone for 8 min at −20°C. Nonspecific binding was blocked by incubation in 10% normal goat serum (NGS) and 3% BSA for 3 h at RT. After several washes with PBSTX, embryos were incubated in rabbit polyclonal anti 5HT (ImmunoStar) diluted at 1:500 in 10% NGS/ PBS containing 0.3% TritonX 100 for 2 days at 4°C. After rinsing in PBSTX for 4 h, embryos were incubated in the goat antirabbit IgG Alexa Fluor 488 (Molecular Probes Invitrogen Detection Technologies) diluted at 1:100 in 10% NGS/PBS overnight at 4°C. After thorough washing in PBSTX, embryos were mounted in 0.5% agarose and observed under the fluorescence microscope (Leica M165 FC). Negative controls processed by omitting incu bation with the primary antibody or by replacing the primary antibody with normal rabbit serum showed no positive signals. For measurement of 5HTpositive area, focus was adjusted on the field with the largest positive area, and NIH ImageJ software was used to quantify manually outlined areas. Immunostaining was performed using five to eight embryos per group, and the experiments were done in triplicate.

rT-Pcr
Total RNA was extracted from larvae at 4 and 7 dpf (25 larvae/ group) using ISOGEN II (Nippon Gene) and treated with DNase free (Ambion). cDNA was synthesized from 1 µg total RNA using Reverse Transcription System (Promega). A total reaction vol ume of 25 µL containing 2× GoTaq Green Master Mix (Promega), 10 µM of each primer, and 1 µL cDNA was subjected to PCR using Program Temp Control System PC708 (Astec). βActin was used as an internal control. Amplification conditions and primer sequences are listed in Table 2. The amplified products were separated on a 2% agarose gel. Levels of mRNAs expression were analyzed by NIH ImageJ software and normalized by the expression level in the control group at each developmental time. Experiments were done in triplicate.

heart rate Measurement
Embryos at 2 dpf were individually placed in a well of a 12well culture plate containing 500 µL of corresponding experimental medium and kept for 15 min to allow heartbeats to resume a steady rate. Heart beats were counted manually for 15 s under a stereo microscope (Leica 58APO). Ten embryos were used for each group. Experiments were repeated three times with eggs collected from different spawns.

Thigmotaxis assay
Assay was performed according to the protocols described previ ously (49, 50). Briefly, zebrafish larvae at 6 dpf were transferred into a 6well tissue culture plate with one fish per well containing 4 mL EM. The bottom of each well was divided into two portions designated as inner and outer zones. After habituation at 28°C for 2 h followed by acclimation under the video camera for 5 min, swimming activity was recorded for 5 min. For each group, 12 larvae were used. Data from 36 larvae from three different spawns were pooled and analyzed to express % of time a fish spent in the outer zone.

statistical analysis
Data are presented as mean ± SEM. Statistical differences between groups were evaluated by oneway ANOVA followed by Tukey's or least significant difference post hoc test using IBM SPSS statistics version 19. Unpaired Student's ttest was used for the dot blot analysis. Kruskal-Wallis test and Mann-Whitney U test were used for thigmotaxis assays, as the data did not meet the assump tions required for parametric testing. Significant differences were accepted when p < 0.05. resUlTs effect of e 2 exposure on serotonergic neuron 5HTpositive neurons were detected in the embryos at 2 dpf by wholemount fluorescent immunohistochemistry ( Figure 1A). Positively stained neurons were located in pretectal and thalamic  complex and raphe as reported previously (48, 51-54). 5HTpositive areas were significantly increased when exposed to low doses E2 (0.001 and 0.005 µM) but decreased in high dose (1 µM) exposure ( Figure 1B). Effects of both low and high doses E2 were significantly reversed by addition of 1 or 10 µM ICI, respectively ( Figure 1C).
Relative expression levels of tph isoforms at 4 dpf were analyzed by semiquantitative PCR. Significant increase in expression was detected in all tph isoforms when embryos were exposed to lowdose E2 (0.005 µM). Addition of 1 µM ICI completely reversed the decreased expression of tph2, while expressions of tph1a and tph1b were reversed partially ( Figure 1D). Conversely, highdose E2 exposure significantly decreased expression levels of all isoforms, which was partially reversed by addition of 10 μM ICI (Figure 1E).
When the embryos were exposed to E2, the heart rate at 2 dpf was significantly increased in the 0.005 µM group, while signifi cant decrease was found in the 0.1 and 1 µM groups (Figure 2A). Addition of 1 and 10 µM ICI significantly reversed the effects caused by low or high dose of E2, respectively (Figure 2A). To verify the role of serotonergic signaling in regulation of heart rate, effects of Q (5HT agonist) and FLX (5HT selective reuptake FigUre 2 | Effect of E2 on heart rate and thigmotaxis. Heart rate was measured in 2-dpf embryos; (a) exposure to low and high doses of E2 and co-incubation with ICI; (B) co-incubation of E2 and Q or FLX to examine involvement of serotonin signaling. Thigmotaxis assay was performed to evaluate anxiety level in 6-dpf larvae; (c) exposure to low and high doses of E2 and co-incubation with ICI; (D) anxiolytic effect of low dose of E2 when larvae were exposed to 200 µM DEX. (e) Measurements of 5-HT-positive neuron area in 2-dpf embryos exposed to DEX and low dose of E2. (F) Effects of co-incubation of E2 and Q or FLX on thigmotaxis to examine involvement of serotonin signaling. Data are presented as a mean ± SEM. Different letters in each graph indicate significant differences (p < 0.05). inhibitor) were tested. Coincubation with 0.1 µM E2 significantly reversed the decreased heart rate caused by E2. Heart rate was sig nificantly increased when exposed to Q or FLX alone ( Figure 2B).
Thigmotaxis assay was performed using 6dpf larvae. Exposure to 1 µM E2 increased the time fish spent in outer zone, suggest ing that anxiety was increased, but no significant difference was (a) Western blots of brain extracts from control and E2 exposed fish and ovarian extract from untreated fish were stained with the antiserum to AroB and AroA, respectively, showing a positive band at the expected size for each aromatase as indicated by arrowheads. (B) Dot blots of 6 and 2 dpf larval extracts using the antiserum to AroB and AroA, respectively, were analyzed. Representative blots are shown in each graph. Data are presented as a mean ± SEM. Different letters in each graph indicate significant differences (p < 0.05). observed in 0.005 µM E2 group (Figure 2C). Addition of 10 µM ICI partially reversed the increase caused by 1 µM E2 (Figure 2C). To further examine the effect of lowdose E2, embryos were exposed to 200 µM DEX and subjected to the assay. DEX alone increased the time, but coincubation with 0.005 µM E2 significantly reduced the increase (Figure 2D). Immunostaining for 5HT showed that coincubation of lowdose E2 and DEX increased the positive staining, although DEX alone had no effect (Figure 2E). To verify the role of serotonergic signaling in thigmotaxis assay, effects of Q and FLX were tested. Both Q and FLX significantly decreased the time fish spent in outer zone caused by 1 µM E2 exposure ( Figure 2F).

Validation of MO-Mediated Knockdown
Using the specific antisera Specificity of the antisera to AroB and AroA was examined by Western blot. The antiAroB revealed a single band at the expected size of 50 kDa in brain extract from the fish exposed to E2 at 25 ng/L (47) (Figure 3A). The antiAroA detected a single band at the expected size of 75 kDa in the ovarian extract ( Figure 3A), which is in agreement with the previous study (55). In addition, immunohistochemistry of the ovary showed similar localization of AroA as previously described (56) (data not shown).
The dot blot analysis of the larval extracts showed that both AroB and AroA MO injections significantly decreased immunore activity compared to the uninjected control, indicating decreased translation of AroB and AroA, respectively ( Figure 3B). Std MO, InvB, or InvA MO did not show any significant difference com pared to the uninjected control (data not shown).

effect of MO-Mediated Knockdown of aroB on serotonergic neuron
When AroB MO was injected, 5HTpositive area was signifi cantly decreased in the 5 ng/L group and partially decreased in the 2.5 ng/nL group compared to the uninjected control ( Figure 4A). Injections of Std MO and InvB MO did not show any significant difference in 5HTpositive areas compared to the uninjected control ( Figure 4A). Moreover, the injection of AroA did not show any changes ( Figure 4A). The decrease in 5HTpositive area caused by AroB was completely rescued by coinjection of 30 pg/nL AroB mRNA ( Figure 4B) and partially rescued by E2 exposure at 0.1 µM (Figure 4C). When p53 MO was coinjected with AroB MO to examine offtarget effect, no significant differ ence in 5HTpositive area was observed, suggesting that decrease in 5HTpositive area caused by AroB MO is not due to apoptosis caused by p53 activation (Figure 4D).
The effect of AroB MO injection on relative expression of tph isoforms was evaluated by semiquantitative PCR using 7dpf larvae. While expression levels of tph1a and tph1b showed no significant changes, expression of tph2 isoform was signifi cantly decreased and partially rescued by E2 exposure at 0.1 µM (Figure 4E).
Heart rate of AroB MO injected embryos was significantly decreased compared to the uninjected or Std MO and InvB MO injected controls (Figure 5A). The decrease caused by AroB MO was rescued either by coinjection of AroB mRNA or by exposure to 0.1 µM E2 (Figures 5B,C). Exposure to 100 µM Q as well as to 5 µM FLX ( Figure 5D) reversed the decrease to the control level.
In thigmotaxis assay, AroB MO injection caused significant increase in time fish spent in outer zone compared to the uninjected, Std MO injected, and InvB MO injected controls ( Figure 5E). This effect of AroB MO was rescued either by coinjection of AroB mRNA or by exposure to 0.1 µM E2 (Figures 5F,G). Exposure to 100 µM Q as well as to 5 µM FLX reversed the time fish spent in outer zone increased by AroB MO injection (Figure 5H).

DiscUssiOn
The aim of this study is to elucidate the role of estradiol and brain aromatase in modulation of serotonergic neurons in early development of zebrafish, as the early ontogeny of serotonergic system may be one of the important factors for neuronal growth and brain development. We demonstrated that exogenous administration of E2 biphasically affected parameters such as 5HTpositive areas, relative expression of tph isoforms, heart rate and thigmotactic behavior with stimulation and suppres sion of serotonin system at the low dose and the high dose,   (57), which sup ports that our MO experiments reflect the reduction of estrogen production. Nonmonotonic dose responses of hormones and endo crinedisrupting chemicals have been widely documented (58). Estrogen among other hormones is known to exhibit biphasic dosedependent effects in various physiological processes (59)(60)(61)(62)(63)(64)(65)(66). However, only limited information is available in regards to serotonin system. There is one study in fish showing that low dose of E2 stimulated monoamine oxidase activity and decreased 5HT content in hypothalamus in ovariectomized catfish, while the result was opposite for high dose (67). Our study demonstrates that biphasic dosedependent effects of E2 on serotonergic neuron in fish, and shows that the effects or both low and high doses are mediated through ER, indicat ing physiological relevance. The effect of the low dose of E2, stimulating serotonergic neuron, is likely to reflect the role of endogenous E2 in embryos, as AroB MOmediated effects demonstrate that brainformed estrogen is necessary to main tain activity of serotonergic neuron in embryos. Mechanisms of biphasic responses are complex, but may be in part controlled by downregulation and desensitization of receptors (57,68). Thus, effects of high doses of E2 on serotonergic neuron in this study may be due to downregulation/desensitization of ERs. Sequence analysis of the promoter region of zebrafish tph isoforms shows the presence of 1/2 ERE in the upstream of transcription start site in all isoforms, suggesting possible nuclear action of estrogen, though their functional analysis is yet to be reported. In human serotonergic cell line, binding of E2 and ERβ has been shown to directly interact with 1/2 ERE of tph2 promoter to elicit gene expression (39). In addition to the classical action of E2 on nuclear receptors membrane ERs plays an important role in brain (69,70). Interaction between mem brane ERs and the metabolic glutamate receptor in the brain provides a rapid and transient E2 action (71,72). Membrane bound Gproteincoupled ER, GPER/GPR30, also known to be involved in modulating rapid nongenomic action of E2, plays a role in several brain areas (73). Estrogen action through GPR30 has been suggested in regulation of serotonergic neuron in mammals (74). Further studies are required to elucidate the mechanisms by which estrogen regulates serotonergic neuron in zebrafish. (B,c) co-injection of AroB mRNA and exposure to E2, respectively, to rescue the effect of AroB MO; (D) injection of AroB MO with and without exposure to Q and FLX to examine involvement of serotonin signaling. Thigmotaxis assay was conducted using 6-dpf larvae; (e) injection of AroB MO and control MOs; (F,g) co-injection of AroB mRNA and exposure to E2, respectively, to rescue the effect of AroB MO; (h) injection of AroB MO with and without exposure to Q and FLX to examine involvement of serotonin signaling. Data are presented as a mean ± SEM. Different letters in each graph indicate significant differences (p < 0.05). Attenuation of serotonergic neuron by AroB MOmediated knockdown clearly demonstrated that brainformed estrogen is necessary to maintain the serotonin system to control heart rate and anxiety behavior in early development of zebrafish. Validity of AroB knockdown was supported by several lines of evidence. Immunoreactivity to the antiserum specific to AroB was decreased in AroB MO injected embryos. In addition to no significant effects found in the controls including standard MO, inverted AroB MO and AroA MOinjected embryos, AroB MOmediated effects were rescued by coinjection of AroB mRNA and exposure to E2. Offtarget effect of MO injection was also examined by knockdown of p53, showing that the decreased 5HTpositive area caused by AroB MO is not through activa tion of p53. The decrease in 5HTpositive area by AroB MO injection indicates that brainformed estrogen stimulates 5HT synthesis, which is in accordance with the stimulatory effect of lowdose E2. When the relative expressions of tph isoforms were examined in AroB MO injected embryos, only tph2 expression was significantly decreased by AroB MO, which is well supported by the previous studies showing tph2 but not tph1 is expressed in raphe 5HT neurons (54,75,76). Expression of tph2 in 5HT neurons in pretectal and hypothalamic complex starts to appear at 60 hpf (76). On the other hand, whereas in the exposure experiments, expressions of all isoforms were affected by E2; increased by low dose and decreased by high dose. The results support the previous studies reporting that tph2 expressed in brain is responsible for 5HT synthesis in the zebrafish (27,28,54). Thus, we provide the evidence that brainformed estrogen stimulates tph2 expression to maintain 5HT content in the serotonin neuron. The effects of E2 exposure on tph isoforms indicate E2 also modulates serotonin biosynthesis in tissues outside the brain. 5HT has been reported to be produced in various organs including intestine which is the major source of 5HT in the body and TPH1 is responsible for its synthesis (54,77). Investigation of estrogen regulation of serotonin produc tion in intestine during development would be a future research interest.
The parameters of physiological functions of serotonin system, heart rate, and thigmotactic behavior were measure to verify the activity of serotonergic neuron. The results were in accordance to the changes in 5HT levels in the neurons; the increased 5HT levels are accompanied by the increased heart rate and decreased thigmotactic behavior, while the contrary was true for the decreased 5HT levels. Serotonin is known to be involved in cardiovascular function, and the effect of central serotonergic neuron is mediated through autonomic nervous system in mammals (18). Our result of the lowdose (0.005 µM) E2 which increased heart rate corroborates the effect of MOmediated AroB knockdown, indicating that nanomolar level of brainformed estrogen, or even lower level in the tis sue, stimulates serotonergic neuron to increase the heart rate. Exposure to quipazine (serotonin agonist), or fluoxetine (selec tive serotonin reuptake inhibitor, SSRI) completely reversed the decreased heart rate caused by the high dose (0.1 µM) E2, or AroB MO injection confirming that heart rate is under the control of serotonin signaling. Taken together with a recent study showing that GPER in the pituitary of zebrafish embryo regulates heart rate through thyroid hormone (78), estrogen in brain centrally regulates heart rate through various mechanisms. On the other hand, cardiac functions are directly regulated by estrogen (79) and aromatase has been detected in the heart tissues such as myocardium in mice (80)(81)(82). Therefore, it is possible that AroB MO injection may affect aromatase expression in the heart and locally produced estrogen modulates heart rate. In some teleost fish, both ovarian and brain aromatases are expressed in the heart (83)(84)(85), but in ricefield eel only brain aromatase is detected (86), while only ovarian aromatase is present in spotted scat (87). These difference may be due to technical difference as well as differences in species and developmental and physiological status. Our preliminary analysis indicated the expression of ovarian aromatase but not the brain form in adult zebrafish heart (data not shown), suggesting that our result of MO injection is likely to be mediated through knockdown of brain aromatase expressed in the brain not in the heart. However, expression of aromatase in the heart during develop ment needs to be verified.
Thigmotaxis is an evolutionally conserved behavior associ ated with fear and has been shown to be affected by anxiolytic and anxiogenic compounds (88); thus, it has been used to meas ure anxiety levels in animals including fish (89)(90)(91)(92). Our present study shows the high dose E2 (1 µM), which decreased the 5HT level, significantly increased anxiety (increased time spent in outer zone), and this increase was abolished by addition of 5HT agonist (Q) or SSRI (FLX), indicating the effect of high dose E2 is mediated through serotonin signaling. Similarly, increased anxiety by AroB MO was also abolished by Q or FLX, which supports our hypothesis that brainformed estrogen modulates serotonergic neuron. Despite our expectation, the lowdose E2 (0.005 µM), which increased the 5HT level, did not cause reduction of anxiety. Therefore, we further examined to see if the lowdose E2 exerts anxiolytic effect in the larvae exposed to DEX to induce stress, and indeed, lowdose E2 decreased the anxiety level. Thus, our data demonstrate a negative correlation between anxiety behavior and 5HT level, which is in accord ance with previous studies. In mammals, depletion of 5HT level in rat brain induces anxiety (93) and acute reduction of trypto phan increases the anxiety level in patients of a social anxiety disorder (94). The role of 5HT in anxiety is also reported in zebrafish (20,95). Buspirone, partial agonist for 5HT1A recep tor, exerts anxiolyticlike effect in zebrafish (96). The phenotype of zebrafish leopard strain, which is characterized by increased anxietylike behavior, is rescued by acute treatment with FLX (97). Taken together, we provide the evidence that brainformed E2 has an important role in modulating anxiety through sero tonergic transmission.
In contrast to mammalian brain, where aromatase is expressed in both neuron and glia (98,99), it is well documented that brain aromatase in fish is exclusively expressed in RGCs along the ventricles of forebrain, midbrain, and hindbrain serving as neural progenitors (10,11,47). While most RGCs are trans formed into astrocytes by the time of adulthood in mammalian brain (100), presence of RGCs persists throughout the lifespan of zebrafish, which is considered to be one of the contributing factors for high capacity of neuronal proliferation (101). On the other hand, serotonin is known to play a role in neurogenesis (102). In adult zebrafish, it has been reported that projection of 5HT neurons in raphe to ventricular surface of the brain, where highly proliferative cells are found. In addition, expression of 5HT receptors are localized in ventricular surface in larval and adult zebrafish (27,103). Thus, it may be possible that RGCs in ventricular surface are innervated by 5HT neurons in raphe and modulated for neurogenesis. Interestingly, it has been reported that AroBpositive RGCs in PVO area in adult zebrafish has an ability to differentiate into serotonergic neuron (104). Taken together with our present study providing the evidence that brainformed estrogen is necessary to maintain the levels of 5HT in neurons in raphe, we can hypothesize that differentia tion of AroBexpressing RGCs in serotonin neurons is regulated by serotonin neuron in raphe, whose activity is modulated by estrogen produced by AroB. It has been shown that placenta aromatase activity and expression are stimulated by serotonergic 5HT2A receptor signaling (105). In goldfish, AroB expression in RGCs in vitro is upregulated by dopamine with modulation by E2 (106). Nonetheless, estrogen biosynthesis and homeostasis in CNS are regulated and finetuned by multiple factors like neurotransmitters and hormones, so that diverse functions of estrogen can be coordinated.
In conclusion, this study demonstrates that estradiol exhi bits a biphasic effect on serotonergic neuron, and that brain aromatase, thus brainformed estrogen plays a significant role in modulating serotonin levels to sustain appropriate develop ment and functions of serotonergic neurons which regulate heart rate and anxiety behavior in zebrafish embryos and larvae. Considering the role of serotonergic neurons in neural development and neurogenesis, it is possible to postulate that one of the mechanisms of brain aromatase and brainformed estrogen to regulate neurogenesis in teleost brain may be through modulation of serotonergic system, which awaits future investigation.

eThics sTaTeMenT
All experimental procedures and maintenance of fish were conducted in accordance with the Guide for Care and Use of Laboratory Animals published by the US National Institutes of Health.