The breeding and experimental protocols involved in using zebrafish were approved by the IACUC of the Model Animal Research Center, Nanjing University. All methods were performed in accordance with the relevant guidelines and regulations.

The studies involving human participants were reviewed and approved by the local ethics committee of the Affiliated Nanjing Brain Hospital, Nanjing Medical University (2017-KY017). Written informed consent to participate in this study was provided by the legal guardian/next of kin of the participants.

Two sgRNAs were designed according to the website.¹ The sequences of sgRNAs are GAUAAAAUCUACAUGUACGGGUU UUAGAGCUAGAAAUAGCAAGUUAAAAUAAGGCUAGUCC GUUAUCAACUUGAAAAAGUGGCACCGAGUCGGUGCUU UUUUU (atrn-sgRNA1), GGAGGGAAGAUUGACUCCACGU UUUAGAGCUAGAAAUAGCAAGUUAAAAUAAGGCUAGU CCGUUAUCAACUUGAAAAAGUGGCACCGAGUCGGUGC UUUUUUU (atrn-sgRNA2). Two sgRNAs were mixed with Cas9 mRNA and microinjected into the embryos at one-cell stages. The microinjection amount was 1 nl of solution containing 0.05 ng of two sgRNAs and 0.25 ng of Cas9 mRNA. When screening the mutation, the primers were designed to identify the genotypes. atrn −4 F: 5′-CAGGATAAAATCTACATGTA-3′; atrn −4 F(34): 5′-CAGGATAAAATCTACACGGA-3′; atrn −22 F: 5′-TAAAATCTACATGTACGGAG-3′; atrn −22 F(34): 5′-TAAAATCTACATGTAGGAAA-3′; atrn R: 5′-ATATTGTACTG CTGCACTTG-3′.

Digoxigenin-UTP-labeled antisense RNA probes were synthesized in vitro using a linearized plasmid or PCR product as template. The template of atrn probe was amplified with the following primers: 5′-CCTCTCAAAGCTGGATGACATTA AAC-3′ (forward) and 5′-CACACTTCTCTCCATCTGCATAC-3′ (reverse). Whole mount in situ hybridization (WISH) on examining the expression patterns of atrn was performed as described previously (Yue et al., 2018). The embryos were raised to adults as founders. The F1 mutant zebrafish were screened using the method as described before (Kantari et al., 2007). Briefly, F1 generation zebrafish were genotyped by the PCR method using the genomic template prepared from the caudal fins clipped from zebrafish older than 6 weeks with a commercial kit (YSY, China). The primers used for amplifying the genomic sequences were 5′-TATGAGACTCGGTGCTGCAG-3′ and 5′-CCTAAAACATGTCCTGTACTGTATG-3′. The PCR conditions were 95°C for 2 min, 35 cycles (94°C for 30 s, 56°C for 30 s, and 72°C for 30 s), and a final extension of 10 min at 72°C. F1 mutant zebrafish carrying the atrn knockout allele were selected and inbred to get atrn null progeny (F2).

Real-time RT-PCR assay (qRT-PCR) was performed to examine the relative expression levels of the genes of interest. For the coding genes in this study, we used ABI Step One Plus (Applied Biosystems), as reported previously (Yue et al., 2018). Transcript levels of the examined coding genes were normalized to the gapdh mRNA level according to standard procedures. We extracted total RNA from at least 30 embryos, three brains or testis in each group using Direct-zol RNA Mini-Prep (Zymol Research). The venous samples of six healthy people and 20 patients suffering from mental disorder with a high testosterone level were extracted total RNA by using Direct-zol RNA Mini-Prep (Zymol Research). Zebrafish and human venous cDNA at indicated stages were prepared by HiScript II 1st Strand cDNA Synthesis (Vazyme). The PCR was performed as follows: 50°C for 2 min, 95°C for 10 s, 40 cycles (95°C for 5 s, 60°C for 20 s), 95°C for 15 s, 60°C for 30 s, 95°C for 15 s. The primers used are listed as below: cyp51-rt-F1: 5′-CACACGGAGAAACACACAACCAC-3′, cyp51-rt-R1: 5′-CT AACAATGTGCAACTGTAGTG-3′; hsd17b7-rt-F1: 5′-CCGA CCAAGCAAGATGGATCTTG-3′, hsd17b7-rt-R1: 5′-CTCCAT GAGCAGTTTATAAATGACC-3′, zgapdh-qF: 5′-CGCTGGCAT CTCCCTCAA-3′, zgapdh-qR: 5′-TCAGCAACACGATGGCTG TAG-3′. hATRN-RT-F1: 5′-TCCAGACGTAGAGAGCAACT TC-3′, hATRN-RT-R1: 5′-TTGTTGCCAAAACACGGCTC-3′, hGAPDH-Q-F1: 5′-CACTAGGCGCTCACTGTTCTC-3′, hGAPDH-Q-R1: 5′-CCAATACGACCAAATCCGTTGAC-3′.

To observe the morphology, we treated embryos with 0.04 mg/ml MS-222 (Sigma) at room temperature for 5 min. Embryonic pictures were taken by an Olympus MVX10 microscope (Japan), and the photos were improved by Photoshop. The transgenic zebrafish lines Tg(hb9:eGFP), atrn^nju70 and Tg(olig2:dsRed), and atrn^nju70 were to confirm the development of motor neurons and oligodendrocytes. All confocal images were acquired using a Zeiss LSM880 confocal microscope. Then when measuring the weight and length of zebrafish, we treated them with 0.04 mg/ml MS-222 (Sigma), then blot the water, placed it in a Petri dish, measured it with an analytical balance, and recorded the data.

Live video tracking of zebrafish larvae was performed by Zebralab Video-Track system (Viewpoint, France). The mature zebrafish (4 months postfertilization) were transferred to a box (well size: 11-cm length × 7-cm width × 7-cm height) with 400 ml of E3 medium, and larvae (aged 5 dpf, 6 dpf) were transferred to a 24-well culture plate with 1 ml of E3 medium. In larvae, the detection threshold value for movement was set at 2–20 mm/s. Locomotor activity was measured per 1 min, and videos were recorded for 5 or 10 min. In adult zebrafish, the detection threshold value for movement was set at 20–75 mm/s. Locomotor activity was measured per 1 min, and videos were recorded for 1 h.

At 4 mpf, a mirror attack test was carried out using a box (well size: 11-cm length × 7-cm width × 7-cm height) with a one-sided mirror. Individual zebrafish was transferred to the apparatus with 400 ml of E3 medium. The movement was recorded for 10 min with data collection every 30 s. The detection parameter for movement was set at 20–75 mm/s, and the movement in the entire space and the movement in the mirror space were recorded.

Total RNA was isolated from the single brain of a 4-months postfertilization (mpf) wild-type and atrnnju70 zebrafish using TRIzol reagent (Invitrogen). The transcriptome sequencing was performed by Novel-Bio Bio-Pharm Technology Co. (Shanghai, China). Gene expression levels were quantified by RPKM (reads per kilobase of transcript per million mapped reads) arithmetic. The MapSplice software was used for data alignment, and EB-Seq arithmetic was used for the screening of differential expression genes.

After the samples were separated, they were quick frozen with liquid nitrogen, 250 μl of 1 × PBS was added in the ice, the tissue was sufficiently broken with a grinder, and it was centrifuged at 12,000 × g at 4°C for 1 min. The next specific steps follow the protocol. A standard curve was created using a computer software, mELISA, which is capable of generating a four-parameter logistic (4-PL) curve-fit.

This study was approved by the Ethics Committee of the Department of Psychiatry, Affiliated Brain Hospital of Nanjing Medical University, Jiangsu, China. From 2018 to 2020, all men and women aged 20 and above, their pathological information, including testosterone content, clinical diagnosis, etc., were all used for analysis. All individuals (including 3,588 males and 5,667 females) were required to have an empty stomach at least 8 h before venous sampling. The method of Elecsys Testosterone II, based on the principles of competition, was applied to measure concentrations of serum testosterone. Serum testosterone samples were analyzed by electrochemiluminescence using a Cobas e601 analyzer in the clinical laboratory of the Affiliated Nanjing Brain Hospital, Nanjing Medical University, and the kit was provided by Roche Diagnostics GmbH.

Experiments were performed two or three times independently. Data are shown as mean ± SD. Statistical analysis was carried out with GraphPad Prism 7 (GraphPad Software, La Jolla, CA, United States). Data were first tested for normality using the Kolmogorov–Smirnov’s test. If data sets exhibit normal distribution, we employed Student’s t-test for equal variances or Welch’s t-test for unequal variances. If data sets are found not to exhibit normal distribution, Mann–Whitney test was applied. Testosterone level differences between the patients with different mental disorders, and healthy controls were tested using Chi square. A value of p < 0.05 (^∗) was considered statistically significant, and p < 0.01 (^∗∗) and p < 0.001 (^∗∗∗) were considered very statistically significant.

Previous studies have shown that Atrn mutant affects motor behavior in mice and rats (Khwaja et al., 2006; Sakakibara et al., 2008; Izawa et al., 2010; Cheng et al., 2014). To better understand the function of Atrn in neural behavior, we use zebrafish as a model animal. Colinear analysis showed that human ATRN, and mouse and zebrafish atrn were all colinear with GFRA4 and SLAC4A11, indicating that zebrafish atrn and mamalian atrn are orthologous genes (Supplementary Figures 1A,B). Unlike human and rat ATRN, the zebrafish atrn gene only encodes a transmembrane form of protein, but it has two transcripts, one is 1,345 amino acids (aa) and the other is 1,383 aa (Sun et al., 2017), both of them containing crucial domains.

The developmental expression of atrn mRNA was examined by semiquantitative PCR relative to actb2 as an endogenous control (Supplementary Figure 1C). From zygote to dome stage, atrn mRNA was widely expressed in animal poles (Sun et al., 2017), and maternal atrn mRNA was decreasing at this stage. By 24 h postfertilization (hpf), the zygote atrn mRNA was detected extensively in the head. This result suggests that Atrn may be involved in the development of the nervous system. In adult zebrafish, a high expression of atrn was mainly found in the brain, eyes, testes, and ovaries (Supplementary Figure 1D), in line with the widespread expression pattern of ATRN in human tissues (Tang et al., 2000).

Atrn spans a total of 29 exons and encodes a protein with a conserved CUB domain, kelch repeats, C-type lectin-like domain, and laminin EGF domain (Figure 1A). To further analyze the function of Atrn in zebrafish, we designed two sgRNAs to target the eighth exon (Figure 1B). After screening, we obtained two atrn mutant lines, nju70 and nju71, with deletion of 22 bases and 4 bases in the genomic sequence, respectively. The protein product prediction indicates that both encode nonfunctional truncated proteins (Figure 1C). The whole-mount in situ hybridization results showed that the expression of atrn in the maternal zygote (MZ) atrn^–/– is reduced (Supplementary Figure 1B). The mutants were raised to adulthood and found to grow normally for breeding. Since the results produced by the two mutant lines were consistent, the atrn^nju70 mutant line was chosen for the following study.

Morphology analysis of embryos produced by heterozygotes (atrn^nju70/+) in-crosses showed no developmental defects in homozygous mutant embryos. When the embryos develop to adults, the MZ mutant is not different from the wild type (Supplementary Figure 1A). By crossing to the transgenic Tg (hb9: eGFP) and Tg (olig2: dsRed), we found that atrn^–/– did not affect the development of motor neurons and oligodendrocytes (Supplementary Figure 1C).

Compared with wild-type (WT) littermates, atrn^+/– and atrn^–/– zebrafish showed reduced body weight in both genders at 4 months postfertilization (4 mpf) (Supplementary Figures 1D,E). This result is in line with studies in mice (Kuwamura et al., 2002). Body lengths of these mutant zebrafish were the same as WT littermates (Supplementary Figures 1F,G). Therefore, they appear slimmer (Supplementary Figure 1A).

We then analyzed their motion using ZebraLab (Viewpoint, French) during day (10:00–14:00) and night (0:00–4:00). A single adult zebrafish was placed in a box containing 400 ml of egg water and placed in Zebralab. The test time was 1 h, and the data were recorded every 1 min. The speed to stay inactive (black)/moderate motion (green) for adult fish was set to 25 mm/s, and the speed for moderate motion/large motion (red) is 70 mm/s. Ten males and females of different genotypes were tested each, and the analysis found that during the day, the trajectory distribution map of mutant males showed more large movements (Figure 1D). By counting the average velocity of each fish in each minute, and drawing the speed-time diagrams (Figures 1E,F), the average speed of the mutant males was significantly higher than that of the sibling wild type, and there were no differences among the three genotypes of females during the day. At night (Figure 1G), the average velocity of mutant females is significantly lower than that of the wild type, and they preferred to be quiet, while there is no significant difference between the three genotypes of males (Figures 1H,I). During the day, the behavior of mutant male zebrafish is consistent with the behavior of Atrn mutant mice.

In order to confirm the effect of atrn knockout on zebrafish’s aggressive behavior, we designed a mirror attack experiment to test. A mirror was placed on the left side of the box, and ZebraLab was used to record the trajectory of the zebrafish within 10 min (Figure 2A). During the test, the movement distance of the female and male atrn^–/– in the entire box was significantly higher than that of the wild type (Figure 2C), and the velocity–time distribution graph is also shown in Figure 2D. From the trajectory diagram of the zebrafish near the mirror (the attack space; the observer can determine whether it is a female or a male fish), atrn^–/– zebrafish have more middle (velocity between 20–75 mm/s) and quicker (velocity higher than 75 mm/s) trajectories, while the wild type have more quiet trajectories (velocity lower than 20 mm/s) (Figure 2B). This confirms that the motility of atrn^–/– is significantly more than that of the wild type. Finally, by counting the movement distance in the attack space, the attack movement distance of the atrn^–/– is significantly higher than that of the wild type of the same sex (Figure 2E).

In order to understand the underlying mechanism of the effect of Atrn on behavior, brains were isolated from 4 mpf wild-type and mutant male and female adult zebrafish. Three brains were taken from each group and sent for transcriptome sequencing analysis (Figure 3A). The steroid synthesis pathway was enriched in results of the male and female mutant brains (Figures 3B,C; Kuwamura et al., 2002). The expression levels of related genes in the steroid synthesis pathway were upregulated, including cyp51, encoding the key enzyme for cholesterol synthesis, and hsd17b7, encoding the key enzyme for testosterone synthesis (Figure 3D).

Testosterone is a steroid hormone synthesized from cholesterol (Hong et al., 2004). About 90% of testosterone comes from Leydig cells, and the rest is produced in the adrenal cortex and other tissues (Geniole et al., 2017; Guan et al., 2019). In the Leydig cell cytoplasm, lanosterol acts as a precursor to cholesterol synthesis, catalyzed by 14α-demethylase, which is encoded by cyp51 (Geniole et al., 2017). Cholesterol enters the mitochondria with the help of StAR protein, produces pregnenolone under the catalysis of P450scc, and then synthesizes testosterone under the catalysis of 17β-hydroxylase, which is encoded by hsb17b7 (Forgacs et al., 2005; Oyola and Handa, 2017).

qRT-PCR results confirmed that the expression levels of cyp51 and hsd17b7 were significantly increased in the brain and testis of the male mutants (Figures 3E–H). In the brain of female mutants, the expression levels of cyp51 and hsd17b7 were also increased (Figures 3I,J), while in the ovaries, cyp51 was unaltered (Figure 3K), and hsd17b7 was decreased (Figure 3L). Testosterone ELISA Assay (Parameter, United States and Canada) was then used to test the testosterone content in brain tissues and gonads of male and female zebrafish of different genotypes, 10 samples per group. The levels of testosterone in the brain tissue of heterozygous and mutant males are significantly higher than those of the wild type (Figure 3M). Similarly, the testosterone content of mutant testis is higher than that of the sibling wild type (Figure 3N).

The adult male atrn MZmut exhibits more pronounced active behavior when compared with the wild type and feature higher testosterone levels (Figures 3E–H). The increase in testosterone levels was detected in both the brain and the testis. We sought to determine which source drives the change in locomotion. To answer these questions, we collected embryos produced by the wild type and the homozygous mutant in-crossing, and raised them to 28.5°C to 6 dpf. Twenty-four larvae were randomly selected and placed in a 24-well cell culture plate. One larva was placed in each well, and 1 ml of egg water was added. A 24-well cell culture plate was placed in Zebralab, the test time was 10 min, and data were recorded every 1 min. The speed of the inactive (black)/moderate motion (green) was set to 2 mm/s, and the speed of moderate motion/large motion (red) was 10 mm/s. The results show that the mutants are significantly more active than the wild type (Figures 4A,B). The speed–time diagram shows that during the entire test, the atrn MZmut had a significantly higher rate of movement than the wild type (Figure 4D). The average speed of each larva within 10 min was calculated, respectively (Figure 4C). At a statistical level, the average velocity of the mutants was significantly higher than that of the wild type (Figure 4D).

Similarly, wild-type and mutant embryos were collected at the same time period, and three samples in parallel were collected from each group, with 30 larvae per sample. The qRT-PCR results showed that the expression level of cyp51 in the mutant was not different from that of the wild type, while the expression level of hsd17b7 was significantly higher than that of the wild type (Figure 4E). At the same time, 12 genotypes of larvae were collected, each of which contained 10 larvae. After removing the egg water, 250 μl of 1 × PBS was added and milled. Testosterone ELISA Assay was used to test the overall testosterone content of one larva. The testosterone content in the mutant larvae was significantly higher than that of the wild type and heterozygotes (Figure 4F).

Does elevated testosterone have a direct effect on zebrafish locomotion? We exposed wild-type embryos to different concentrations of testosterone for treatment from the development of 24 hpf. The egg water was replaced every 24 h. The same concentration of testosterone was added and processed to 5 dpf. At this time, the zebrafish bladder was fully developed, and zebrafish were able to normally swim normally. The testosterone treatment group was divided into 1, 10, and 100 ng/ml. We then randomly selected 24 larvae from each group, placed them into a 24-well cell culture plate, and proceeded with the tracking analysis (Zebralab). The trajectory distribution chart clearly shows that the zebrafish juvenile group featured a significantly higher locus of movement than did the untreated group (Supplementary Figures 3A–D). Analysis of exercise rates after treatment with exogenous testosterone reveals an increase in average velocity of zebrafish juveniles. With the increase in testosterone concentration, the average velocity of larvae shows a downward trend in a dose-dependent manner (Supplementary Figures 3E,F). The above results indicate that the treatment with exogenous testosterone can mimic the phenotype observed in atrn^–/–, but with the increase in testosterone concentration, due to its drug toxicity, it inhibits the behavior of zebrafish larvae.

In the atrn MZmutant embryos, human wild-type ATRN (hATRN wt) was overexpressed, and we found that 25 and 50 pg of hATRN wt mRNA can efficiently reduce the swimming speed (Figures 4H,I). Similarly, hATRN wt mRNA also can rescue the higher testosterone level (Figure 4J) and the expression level of cyp51, hsd17b7 in atrn MZmut zebrafish embryos (Figure 4G). The results indicated that human wild-type ATRN can significantly rescue the phenotype in atrn MZmut. These data support the interplay of testosterone and ATRN in shaping behaviors in the animal model.

Schizophrenia is characterized by disturbances in perception and thinking, in behavior, emotional, and social symptoms (including lack of motivation, lack of social interaction and apathy), and cognitive impairment (Brennan, 2011; Seibt et al., 2011). The above results have shown that ATRN has influence on the testosterone level and resulted to abnormal locomotion, which is similar to the positive symptoms (abnormally active), but the correlation between behavior and testosterone is still unknown. Some researchers have illustrated that more hormones can affect the development of mental disorder, for example, steroid hormones, especially testosterone, are related to emotional processing (Derntl et al., 2009; van Wingen et al., 2009).

The information of patients suffering from mental disorder in the hospital was collected to analyze the relationship between the expression level of ATRN and testosterone level. The clinical standard for testosterone content in the human body was determined in adults aged 20–49. The normal level for men is 8.64–29.00 nmol/L, and the normal level for women is 0.29–1.67 nmol/L. The normal testosterone level in adults aged over 50 is 6.68–25.70 nmol/L (men) and 0.10–1.42 nmol/L (women).

In light of the known testosterone effects on emotional processing, we postulated that levels of circulating hormone may serve as a surrogate marker. To put this information into perspective, both the expression level of ATRN and the testosterone level were both examined in subjects diagnosed with schizophrenia. Manufacturer data obtained from healthy adults served as a reference.

When stratified by age and gender, we counted 2,579 men aged 20–49 years, 3,926 women aged 20–49 years, 1,009 men aged over 50 years, and 1,741 women aged over 50 years. In the age group 20–49 years, men presenting with a diagnosis of schizophrenia were overrepresented relative to controls and featured lower testosterone levels (Figure 5A and Supplementary Table 1).

In contrast, women with schizophrenia were more likely to have higher testosterone levels. Likewise, women with a diagnosis of bipolar disorder also featured elevated testosterone levels (Figure 5C and Supplementary Table 2). In addition, among the test population over the age of 50, men with anxiety, insomnia, and depression tend to cluster in the category “high testosterone levels” (Figure 5B and Supplementary Table 3). Only a small group of women suffering from anxiety and insomnia exhibited higher testosterone levels (Figure 5D and Supplementary Table 4). Thus, a dimorphic pattern was observed primarily among patients with schizophrenia. Depending on gender, testosterone levels were either lower or higher than expected.

To establish the contribution made by ATRN expression level to testosterone levels and dimorphic effects, we then purposefully selected 14 patients with ATRN mutations who had been diagnosed with schizophrenia, including five men and nine women (Supplementary Table 5). At different points in time, testosterone levels were determined. We noted that in male patients, testosterone levels were still within the normal range or lower. In contrast, the testosterone levels of female patients were above the normal range (Supplementary Table 5). Moreover, male subjects with ATRN mutations were more likely to be classified as “low testosterone level.” Woman with ATRN mutations were more likely to be classified as “high testosterone level” (Figures 5E,F).