A new experimental model for the investigation of sequential hermaphroditism

The stunning sexual transformation commonly triggered by age, size or social context in some fishes is one of the best examples of phenotypic plasticity thus far described. To date our understanding of this process is dominated by studies on a handful of subtropical and tropical teleosts, often in wild settings. Here we have established the protogynous New Zealand spotty wrasse, Notolabrus celidotus, as a temperate model for the experimental investigation of sex change. Captive fish were induced to change sex using aromatase inhibition or manipulation of social groups. Complete female-to-male transition occurred over 60 days in both cases and time-series sampling was used to quantify changes in hormone production, gene expression and gonadal cellular anatomy. Early-stage decreases in plasma 17β-estradiol (E2) concentrations or gonadal aromatase (cyp19a1a) expression were not detected in spotty wrasse, despite these being commonly associated with the onset of sex change in subtropical and tropical protogynous (female-to-male) hermaphrodites. In contrast, expression of the masculinising factor amh (anti-Müllerian hormone) increased during early sex change, implying a potential role as a proximate trigger for masculinisation. Collectively, these data provide a foundation for the spotty wrasse as a temperate teleost model to study sex change and cell fate in vertebrates.

The stunning sexual transformation commonly triggered by age, size or social context in some fishes is one of the best examples of phenotypic plasticity thus far described. To date our understanding of this process is dominated by studies on a handful of subtropical and tropical teleosts, often in wild settings. Here we have established the protogynous New Zealand spotty wrasse, Notolabrus celidotus, as a temperate model for the experimental investigation of sex change. Captive fish were induced to change sex using aromatase inhibition or manipulation of social groups. Complete female-to-male transition occurred over 60 days in both cases and time-series sampling was used to quantify changes in hormone production, gene expression and gonadal cellular anatomy. Early-stage decreases in plasma 17β-estradiol (E2) concentrations or gonadal aromatase (cyp19a1a) expression were not detected in spotty wrasse, despite these being commonly associated with the onset of sex change in subtropical and tropical protogynous (female-to-male) hermaphrodites. In contrast, expression of the masculinising factor amh (anti-Müllerian hormone) increased during early sex change, implying a potential role as a proximate trigger for masculinisation. Collectively, these data provide a foundation for the spotty wrasse as a temperate teleost model to study sex change and cell fate in vertebrates.
For most vertebrates, sexual fate is genetically determined and remains fixed throughout life. However, for many teleost fishes sex is more plastic 1 . The complete sex change that sequentially hermaphroditic fishes undergo during their reproductive lives, while unique among vertebrates, is taxonomically widespread across the teleost tree [1][2][3][4] . The direction and process of sex change differ greatly among teleosts [5][6][7] and appear to have evolved multiple times 1 . The cellular and molecular processes that underpin sex change are slowly being determined for a few focal species 8 . However, the extent to which these processes might be conserved and reused across the teleosts to achieve sex change is poorly understood 9 .
Changes in the social environment and community structure often cue the timing of sex change [10][11][12] . Species where manipulation of social structure can readily induce natural sex change are convenient models to understand the mechanistic drivers of this transformation. More broadly, these species present in vivo opportunities to examine cell fate pathways, neuro-endocrine plasticity, genetic and epigenetic regulation of life-history trajectory and reproductive status.
Current sex change research mostly focuses on tropical and subtropical models within the Labridae [13][14][15][16][17][18][19] , Serranidae [20][21][22][23][24] and Gobiidae [25][26][27][28] . Unfortunately, such manipulations have frequently been limited to the wild because sex change in laboratory settings has proven challenging for many of these species. In addition, few studies have focused on temperate fishes that experience strong reproductive seasonality and a protracted period of sex change. These species, arguably, offer an extended window of graded change in which to tease out fine-scale modulation of physiological and molecular drivers.
The New Zealand (NZ) spotty wrasse, Notolabrus celidotus, is an endemic protogynous (female-to-male sex change) temperate zone (35° S-47° S) labrid that is well suited to laboratory studies. These small (< 26 cm) fish are abundant and easily caught on shallow reefs and in harbours around the NZ coastline. They have dimorphic initial phase (IP) and terminal phase (TP) colour morphs, characteristic of most wrasses 29,30 . As in other wrasses, two male sexual strategies exist with IP sneaker males displaying female mimicry and behaviourally dominant TP

Methods
Experimental set-up. Experiment 1: Induction of sex change in spotty wrasses by aromatase inhibition (AI2014). In this experiment, the aromatase inhibitor (AI) fadrozole (C 14 H 13 N 3 ) was used to induce sex change in captive IP spotty wrasse individuals between August and September 2014. Fish were captured around high tide by hook and line off the coast of Tauranga, Bay of Plenty, New Zealand (37.6878° S, 176.1651° E) and subsequently maintained at the Aquaculture Centre at Toi Ohomai Institute of Technology, Tauranga. TP males were distinguished from IP males (IPM) and females by external observation: IP fish have a large inky thumbprint spot in the middle of the body, whereas TP males have an irregularly shaped row of blackish spots and light electric blue wavy patterns on their cheeks 29 (Fig. 1). IPM and females, however, cannot be externally distinguished. Thirty IP fish ranging from 154 to 229 mm total length (TL) were distributed across three 400-L recirculating seawater systems under a natural photoperiod. Natural sex change was blocked by placing a TP male in each tank, creating a socially inhibitory environment. During the experiment, fish were fed frozen greenshell mussels (Perna canaliculus) three times per week. Pellets containing 200 μg fadrozole (Sigma-Aldrich) in a matrix of cholesterol:cellulose = 95:5 were made in-house (as described in Refs. 43,44  www.nature.com/scientificreports/ matrix only. Following an acclimation period of three weeks, on day 0 of the experiment all IP individuals were given a single intramuscular fadrozole implant (n = 16, equivalent dose = 1.04 mg/kg − 4.64 mg/kg) or a sham implant without hormone (n = 14) using a Ralgun implanter (Syndel, Ferndale, WA). All fish were removed from individual tanks on day 21 (Tank 1, n = 11), day 39 (Tank 2, n = 11) or day 60 (Tank 3, n = 11; end of experiment). Fish were anesthetised in an aerated bath containing 600 ppm 2-phenoxyethanol (Sigma-Aldrich) and blood samples were collected from the caudal vein using a 1 mL heparinised syringe. Fish were then euthanised by rapid decapitation. To standardise tissue collection between fish, samples were consistently harvested from the same region of the gonad. This corresponded to approximately the middle third of one gonadal lobe, which was then excised and fixed in either 10% neutral buffered formalin (IP individuals) or Bouin's solution (TP males) overnight before being transferred to 70% EtOH for histological processing. Body weight and length, and gonadal weight were measured for each fish. Experiment 2: Social induction of sex change in spotty wrasses within their breeding season (SI2016). Sex change was induced in captive spotty wrasses through manipulation of social groups (i.e., removal of males from the treatment tanks) between September and December 2016. Fifty IP (IPM or IP females) and ten TP individuals ranging from 150 to 215 mm TL were captured and maintained as described in Experiment 1 (AI2014). Fish were distributed into groups across ten 400-L recirculating seawater systems (five IP fish per tank) such that each tank contained a hierarchy of different sized IP fish and a single TP male (215-244 mm TL). The size difference between the largest and second largest IP fish was at least 16 mm in all tanks. After a 3-week acclimation, TP males were either removed from the treatment tanks (n = 5) or left in the control tanks (n = 5). Subsequently, the largest IP fish from each tank was terminally sampled on days 0, 30, 50, 60, 65, or 66 (end of experiment) (n total = 10 per sampling day). Blood plasma collection, anaesthetic administration, tissue dissection and recording of morphometrics were conducted as described for AI2014.

Experiment 3: Social induction of sex change in spotty wrasses outside their breeding season (SI2018).
Social manipulation was used to induce sex change in captive spotty wrasses outside the breeding season between January and April 2018. Sixty-five IP and twelve TP individuals ranging from 138 to 218 mm TL were captured and maintained as described in AI2014. Fish were distributed across twelve 400-L recirculating seawater systems such that each tank contained a hierarchy of 4-5 different-sized IP fish and a single TP male (194-220 mm TL). Three control (5 IP females + 1 TP) tanks were maintained and nine manipulated (5 IP females − 1 TP) tanks had the males removed on day 0 after a 2-week acclimation period. A further five IP females were terminally sampled on day 0 to provide a baseline indication of reproductive status. Fish were sampled over a time series as follows; day 1 (n = 5), day 11 (n = 5), day 26 (n = 10), day 36 (n = 10), day 55 (n = 10), day 92 (n = 9). Eleven mortalities occurred during the experiment.
Anaesthetic administration, blood plasma collection, and recording of morphometrics were conducted as described for AI2014. One gonadal lobe was flash frozen in ice-cold (on dry ice) isopentane (C 5 H 12 ) (Sigma-Aldrich) and stored at -80 °C for RNA analyses. The second gonadal lobe was preserved for histological analysis as described in AI2014.
Fish in all three experiments were maintained and manipulated in accordance with New Zealand National Animal Ethics Advisory Committee guidelines. Ethics applications were approved by the Animal Ethics Committee of Toi Ohomai Institute of Technology and independently reviewed by the University of Otago Animal Ethics Committee; and reporting in the manuscript follows the recommendations in the ARRIVE guidelines 45 . Gonadal tissue processing for histology. Histological analysis of gonadal tissues was used to characterise cellular changes occurring across sex change. Tissues from AI2014 and SI2016, fixed in Bouin's solution (TP males) or 10% neutral buffered formalin (IP individuals), were processed for routine embedding in paraffin (New Zealand Veterinary Pathology, Hamilton Laboratory, New Zealand). SI2018 gonadal tissues were embedded in paraffin (Otago Histology Services Unit, Department of Pathology, Dunedin School of Medicine, University of Otago, New Zealand). Sections were cut at 3-4 µm and stained with Mayer's haematoxylin and eosin.
Gonadal sections were examined under light microscope to confirm the sex of each individual and samples were subsequently classified into a series of stages to demonstrate the transition from initial stage female through to terminal phase male (Table 1).
RNA concentration was measured using a Qubit 2.0 Fluorometer (Life Technologies) and RNA integrity was evaluated on a Fragment Analyzer (Advanced Analytical Technologies Inc.). The RNA profiles of sex-changing gonads presented a strong peak of low molecular weight RNA. This is considered to be a result of massive 5S rRNA amplification in ovaries 54 , and masks the 18S and 28S rRNA peaks used to calculate RNA Integrity Number (RIN) values, making them unreliable estimates of RNA integrity. Similar patterns have been observed in ovaries and/or intersex gonads of thicklip gray mullet (Chelon labrosus 55 ), sharsnout seabream (Diplodus puntazzo 56 ) and bluehead wrasse 54 . Therefore, in spotty wrasses, RNA integrity for such samples was confirmed by visual inspection of 18S and 28S rRNA peaks.
Gene expression analysis. A probe array of 18 candidate genes was designed for spotty wrasse based on the nanoString nCounter™ technology (Supplementary Table S1). Spotty wrasse-specific transcript sequences were identified from a draft spotty wrasse transcriptome assembly (EV Todd & NJ Gemmell, 2015, unpublished data, now superseded by a fully annotated genome https:// vgp. github. io/ genom eark/ Notol abrus_ celid otus/). Reference transcripts from zebrafish (Danio rerio) were downloaded from GenBank or Ensembl for each target gene, and actb1, eef1a1a and g6pd as potential housekeeping genes (see Supplemental Information for details on the determination of the housekeeping genes' stability). Transcripts were used to identify the corresponding spotty wrasse sequences from the transcriptome via local BLASTn and tBLASTx (translated the query nucleotide sequences, and the spotty wrasse transcriptome into deduced amino acid sequences in all six possible frames, which were then compared by local alignment). For the best matching spotty wrasse contig, gene identity was confirmed using online nucleotide BLAST (BLASTn) against the NCBI database (http:// www. ncbi. nlm. nih. gov/).
Best-match spotty wrasse contig sequences were submitted to nanoString Technologies for probe design. NanoString nCounter™ CodeSet gene expression quantification was performed on a panel of 18 sex changerelated genes at the Otago Genomics Facility, Biochemistry Department, University of Otago, New Zealand. These data were generated using 100 ng gonadal RNA from the SI2018 fish [5 control females (day 0; F), 19 ET, 9 MT, 9 LT and 5 TP]. The full suite of 18 genes were subjected to PCA analysis (see below) to help verify the gonadal staging. Here, we also present gene expression profiles for six gonadal genes relating to the endocrine function of the reproductive and stress axes ( Table 2). The remaining expression data, including more novel targets, will be addressed in future work. The geometric mean of gene pair actb1|g6pd was selected as the reference gene for data normalisation (see "Results"). Normalisation calculations were performed automatically in the nanoString nSolver Analysis software (version 4.0), and normalised data were log base-2 transformed prior to analysis. Due to non-normality of the normalised nanoString data, the non-parametric Kruskal-Wallis test 50 was run separately for each target gene to determine whether stage had a significant effect on gene expression level, and post hoc comparisons using Dunn's tests 51 with Benjamini-Hochberg correction for multiple comparisons 52 were performed to determine where the significance, if any, lay between stages. All analyses were performed in R (v. 3.3.2) 53 . NanoString Expression Data Analysis Guidelines (MAN-C0011-04) were followed to determine an expression threshold, set as the log 2 of the geometric mean of the negative control counts plus two standard deviations.
Principal Component Analysis (PCA) (scaled, R (v. 1.1.453) 53 ) was used to visualise variation in the overall mRNA expression across the expanded suite of 18 sex change-related genes (Supplementary Table S1) in relation to sexual stage and assess clustering of the histological stages defined in Table 1. To identify the genes Table 1. Histological stages of gonadal sex change in New Zealand spotty wrasse.

Stage Abbreviation Characteristics
Initial phase female F Ovary is dominated by healthy oocytes. Depending on season, these may be predominantly previtellogenic stage oocytes or may also include vitellogenic and maturing oocytes. Atretic late-stage oocytes may be common post-spawning

Results and discussion
The definition and delineation of sexual stages differs between studies of protogynous fish. In this study, we define five gonadal stages to describe the sexual transition from female to terminal phase male endpoints in the temperate spotty wrasse. While these stages are based on gonadal histology the classification is also supported by the PCA analysis of the gonadal gene expression data. These samples showed clusters arranged in the sequential order of the sexual stages from female through to male (PC1, 50.7% variation explained; Fig. 2). The spotty wrasse gonadal gene set was also clustered according to the degree of their developmental commitment (PC2, 19.4% variation explained). Fully differentiated ovary and testis were separated from less differentiated transitional gonads. These patterns show a striking resemblance to whole transcriptome gonadal expression profiles from transitioning tropical bluehead wrasse 8 . This demonstrates the importance and relevance of the spotty wrasse gene panel for studying sexual transition in protogynous wrasses more broadly.

Histological analysis of gonadal sex change. Experiment 1: Induction of sex change in spotty wrasses
by aromatase inhibition (AI2014). Aromatase inhibition successfully induced sex change or changes consistent with the beginning of sex change in 93% of the surviving (2 mortalities) captive female spotty wrasses held under socially inhibitory conditions. Histological analysis confirmed that among the AI-implanted females, 12 fish reached ET stage (day 21, n = 2; day 39, n = 5; day 60 n = 5), and one reached LT stage (day 60). A single fadrozoleimplanted fish remained female. In contrast, none of the control females showed signs of ovarian atresia or sex change. None of the experimental fish were found to be IPM. Experiment 2: Social induction of sex change in spotty wrasses within their breeding season (SI2016). The manipulation of social groups, through male removal, successfully promoted sex change (81%) in female spotty wrasses during the 2016 breeding season. Histology confirmed that among the socially manipulated females, 15 fish reached ET stage (day 30, n = 2; day 50, n = 3; day 60, n = 4; day 65, n = 3; day 66, n = 3), one reached MT stage (day 50), one LT stage (day 50), and one was found to be a fully TP male (day 60). Four of the socially manipulated fish remained female (day 30, n = 2; day 66, n = 2). No conclusive evidence of sex change was found in any of the control females, although ovarian atresia was observed in four of these individuals (day 30, n = 3; day 66, n = 1). Histological analysis also confirmed that five of the starting 50 IP individuals were IPM (10.0% frequency) based on the histological features enumerated in Table 1, a slightly higher ratio than reported previously (4.1-5.7%) 30 .

Experiment 3: Social induction of sex change in spotty wrasses outside their breeding season (SI2018).
Male removal conducted outside the breeding season also induced sex change in female spotty wrasses. All socially manipulated females showed histological signs of ovarian degeneration or sex change. Histology confirmed that 18 females reached ET stage (day 1, n = 3; day 11, n = 2; day 26, n = 3; day 36, n = 3; day 55, n = 4; day 96, n = 3), 12 MT stage (day 1, n = 1; day 11, n = 3; day 26, n = 4; day 36, n = 1; day 55, n = 3), 13 LT stage (day 26, n = 2; day 36, n = 6; day 55, n = 3; day 92, n = 2) and 6 became full TP males (day 1, n = 1; day 92, n = 5). No IPM were found (Fig. 3). Unfortunately control tanks with male fish present experienced an unknown infection and reduced survival in the later stages of the experiment. This confounded the efficacy of socially induced sex change in this experiment. Nonetheless, the results of the other two experiments (AI2014 & SI2016) indicate that both chemical and social manipulation increases sex change compared to control female fish that have a male present (Fisher's exact test, p < 0.001). This is also supported by previous studies with this species (Muncaster, unpublished data).  30 . Of the socially manipulated spotty wrasses, the more mid-transitional and fully transformed males were recorded in the experiment conducted during the post-spawning period (SI2018) in summer and early autumn (January to April; the effect of seasonality on the sexual phases detected during sampling in the social induction experiments is summarised in Fig. 4). In this experiment, 55% of IP fish had MT through to TP stage gonads, whereas less than 25% of fish socially manipulated during the October-December breeding season (SI2016) presented the same stage gonads. Sex change is often seasonally biased with the greatest occurrence following the breeding season in temperate and warm water species, but occurs year-round in many tropical species 23,31,34,57,58 . Our results support observations in wild spotty wrasses, in which natural sex change has been documented during the non-reproductive months between November and May 30 . This indicates that for experimental purposes, post-spawned fish present the best candidates for socially manipulated sex change. However, we have also demonstrated that male removal can lead to sex change within the breeding season.
Steroid production in relation to sex change. Plasma E2 concentrations showed a general decreasing trend from female to male stages (Fig. 5A). Fish treated with an aromatase inhibitor (AI2014) had mean plasma E2 concentrations ranging from 0.34 ± 0.25 SD ng/mL to 0.08 ng/ml for ET and LT stages respectively. In comparison, female fish had a mean plasma E2 concentration of 0.42 ± 0.46 SD ng/mL, although this was not statistically different. Social manipulation during the breeding season (SI2016) showed significantly higher plasma E2 concentrations in females (0.38 ± 0.39 SD ng/mL) compared to males (0.07 ± 0.00 SD ng/mL, p < 0.001 for TP and 0.07 ± 0.00 SD ng/mL, p < 0.05 for IPM). However, plasma E2 concentrations were minimal in all sexual stages when social manipulation was conducted after the completion of the spawning season (SI2018). Similar reduced E2 concentrations after breeding were observed in other protogynous species 22,23,59 . Gonadal resorption and quiescence often follows reproduction in temperate fish and this is characterised by ovarian atresia and reduced sex steroid concentrations 60 . This is not surprising considering the importance of E2 in driving seasonal oocyte growth in teleosts [61][62][63][64] . In addition, confinement could also have, concurrently with ovarian atresia and a   www.nature.com/scientificreports/ been implicated as a critical initiator of sex change in many protogynous species 15,18,22,34 . Yet, despite a 2.3-fold decrease in plasma E2 concentration between F and ET (0.16 ± 0.16 SD ng/mL) fish during the breeding season (SI2016), this drop was not statistically significant. Similarly, there was no significant decrease of plasma E2 concentration between female and early transitional spotty wrasses from the other experiments (AI2014 & SI2018). While this may, in part, relate to sample size or seasonal influence, similar results were also reported in the bambooleaf wrasse (Pseudolabrus sieboldi) 67 and orange-spotted grouper (Epinephelus coioides) 24 . Undetectable E2 concentrations existed from MT to male (TP & IP) stages in nearly all fish. Therefore, while there is no evidence of a minimum plasma E2 threshold required to set off the gonadal transition in spotty wrasse, a general reduction appears to be an important initiating event of the process of sex change in this species. Indeed, E2 depletion has, in cases, been observed to lead to masculinisation in hermaphroditic and gonochoristic species alike 16,[68][69][70][71] . Elevated plasma 11KT concentrations were observed in individual fish towards the transitional and male stages in all three experiments (AI2014, SI2016 and SI2018; Fig. 5B). While variability in the data and reduced statistical power made the detection of discernible differences impossible, the timing of these observations coincides with the histological appearance of spermatogenic cysts (see Table 1). Increased 11KT concentrations were also evident from mid transition onwards in other protogynous species 18,22,72 . Fish treated with AI (AI2014) had minimal 11KT values in the earlier sexual stages. This was evident in the F (0.17 ± 0.03 ng/mL) and ET (0.22 ± 0.17 SD ng/mL) fish while a later stage LT individual presented with 1.0 ng/mL. Fish socially manipulated during the breeding season (SI2016) showed remarkably similar plasma 11KT concentrations regardless of sexual stage. These values ranged from 0.13 ± 0.02 SD ng/mL in F to 0.23 ± 0.31 SD ng/mL in TP. All of the F, transitional, and IPM had either identical mean or individual 11KT concentrations of 0.14 ng/mL. Many of the androgen concentrations of fish socially manipulated after the breeding season (SI2018) remained minimal as expected during quiescence. Reported 11KT levels for male wild caught spotty wrasse range from 0.56 ± 0.30 to 0.78 ± 0.64 ng/mL 66 , and are higher values than those recorded in our experimental manipulations. For this reason, we cannot dismiss the possibility that plasma 11KT concentrations were suppressed due to captivity 65 .
The lowest mean 11KT concentrations (0.13 ± 0.02 SD ng/mL) presented in F and were only slightly higher in subsequent stages such as ET (0.15 ± 0.03 SD ng/mL) and MT fish (0.19 ± 0.15 SD ng/mL). Highest plasma 11KT values were observed in LT (0.29 ± 0.32 SD ng/mL) fish, while slightly reduced concentrations were recorded in TP individuals (0.22 ± 0.18 SD ng/mL). This disparity of androgen concentrations in late-stage fish most likely reflects the chronology of the experiment. TP fish were removed from the socially manipulated tanks at the beginning of the experiment immediately after the breeding season. Collection of these tissue and plasma www.nature.com/scientificreports/ samples would likely have been when spermatogenesis was minimal or non-existent. The occurrence of several LT fish occurred later in the experiment, coinciding with seasonal gonadal recrudescence. The androgen profiles from the fish in this study do not show a clear statistical relationship between sexual stages. However, the role of 11KT in driving spermatogenesis is well established in teleosts 18,61,73,74 . Much of this androgen activity is likely to be paracrine in nature with steroidogenic somatic cells stimulating local germ cells both directly and indirectly within the gonadal compartment 75 . The additional role of 11KT in expressing seasonal male secondary sexual characteristics, such as morphometric and behavioural modifications, also exists 76,77 . This likely requires elevated plasma concentrations for remote, effective target cell signalling and may in part explain substantial elevations of 11KT prior to breeding in many male teleosts. The absolute concentrations of androgen required to stimulate spermatogenesis are conceivably much lower. Furthermore, prior to these peak physiological concentrations the actual concentrations of 11KT within the gonad are likely to be higher than in the plasma. It is, therefore, possible that the absolute concentrations of 11KT required to initiate spermatogenesis www.nature.com/scientificreports/ during gonadal restructure are not reflected in the spotty wrasse plasma samples. This issue may be further investigated using in vitro explant culture systems 78 . Alternatively, gonadal expression of key steroidogenic enzymes should also reflect, in proxy, the androgen activity across different sexual stages.
Sexually dimorphic expression of gonadal endocrine genes. To obtain a more sensitive assessment of the endocrine regulation of gonadal sex change, the expression of three genes encoding gene products involved in steroid biosynthesis, two steroid receptors, and anti-Müllerian hormone were measured. The temporal regulation of steroid hormones is essential for the coordination of sex differentiation, sexual maturation and behaviour in vertebrates. They are also potent mediators of gonadal sex change in teleosts 18,35,79 . As expected cyp19a1a expression was greatest in F and was gradually downregulated across sex change (Fig. 6A). Expression did not differ significantly across stages. As reflected in the E2 data, there was no evidence of an early, rapid downregulation that has been thought to trigger sex change in other species 5,6,15 . Levels of cyp19a1a expression remained similar across ET, MT and LT stages (2018) while in comparison, plasma E2 concentrations were negligible by mid transition in all three experiments (AI2014, SI2016 and SI2018). While cyp19a1a is an unlikely proximate trigger of sex change in spotty wrasse, a more distant association exists nonetheless. This is evident in the number of fish that changed sex following aromatase inhibition (AI2014) as well as the occurrence of sex change in fish socially manipulated after the breeding season (SI2018) when plasma E2 concentrations were minimal. Considering the potent feminising action of E2, a reduction of gonadal concentrations may act as a gateway to facilitate the progression of transition rather than acting as an early trigger. This action may be through the release of steroid-induced suppression on male-pathway genes 79 .
In teleosts, testosterone (T) can be bioconverted into 11KT by the enzymes 11β-hydroxylase (Cyp11c1) and 11β-hydroxysteroid dehydrogenase type 2 (Hsd11b2) 33 . Upregulation of cyp11c1 was observed across sex change (Fig. 6B), with significantly (X 2 (4) = 32.39, p < 0.001) greater expression in TP spotty wrasses than F (median 2.6-fold greater) and ET (median 2.1-fold greater) fish. Albeit less pronounced, hsd11b2 expression increased in a similar pattern across sex change (X 2 (4) = 32.13, p < 0.001; Fig. 6C). The simultaneous upregulation of both cyp11c1 and hsd11b2 from MT through to TP stages coincides with the presence of spermatogenic cysts in the gonad and likely drives the increase in gonadal 11KT production. Although this is not evident in plasma 11KT concentrations (see Fig. 5B), these expression patterns may indicate either undetected paracrine action of 11KT within the early testis or the accumulation of enzyme transcripts prior to translation.
Another contemporary hypothesis for a universal trigger of sex change involves stress. Cyp11c1 and Hsd11b2 are key players in the teleost stress response, being necessary for the interrenal biosynthesis of cortisol and for its subsequent inactivation to cortisone, respectively 7,80-82 . Cross-talk between the interrenal and reproductive axes through the upregulation of these enzymes has been implicated as influencing masculinisation in teleosts 10,15,24,81,83 , particularly through the production of weak interrenal-derived androgens, released as part of the stress response, which could conceivably serve as precursors for 11KT synthesis 66 . As cortisol exerts its effects by activating its receptors, among which Nr3c1 and Nr3c2 are best characterised, the genes encoding these receptors present interesting targets. These genes have opposing expression patterns in bluehead wrasse, with nr3c1 showing a male bias while nr3c2 is female-biased 8,32 . However, gonadal expression of both nr3c1 (X 2 (4) = 28.40, p < 0.001) and nr3c2 (X 2 (4) = 13.02, p < 0.05) was elevated in the LT to TP stages of spotty wrasse (Fig. 6D,E). Expression of these genes did not differ between F and TP fish, which is likely connected to sample size-related variability, particularly in female fish. Early-stage expression of cyp11c1, hsd11b2 and nr3c2 was thought to indicate a role for cortisol in triggering sex change in the bluehead wrasse model 8 . In contrast, upregulation of these genes occurred in later stage gonads in the spotty wrasse. This suggests a relationship between cortisol and sex change may exist but not necessarily as an initiator in this species. Chen et al. 84 report a marked transient increase of cortisol concentrations during the early stages of protogynous sex change in orange-spotted grouper. Similar to what we observed in spotty wrasse, these fish do not show a significant reduction in levels of E2 during sex change. As such, the precise role of cortisol and the stress response during sex change warrants further investigation.
Anti-Müllerian hormone, encoded by amh, is strongly linked to male vertebrate sex differentiation 39,85 . This was reflected in a linear function across sex change stages in spotty wrasse. Amh expression was greatest in TP males (2.2-fold, 1.6-fold and 1.4-fold greater in TP than F, ET and MT, respectively; X 2 (4) = 38.71, p < 0.001; Fig. 6F). The apparent upregulation of amh in ET and MT stage fish indicates its importance in the initiating stages of sex change in this species. Furthermore, amh demonstrated the greatest contribution to PC1 (ovaryto-testis transition) of the PCA analysis. This is consistent with other studies showing upregulation of amh at the onset of protogynous sex change and complementary downregulation in the early stages of protandrous sex change 15,42,86 . The expression pattern of amh provides further validation for the histological staging of ET individuals as resolving cellular differences between ET fish and post-spawned females with regressing ovaries can be difficult in sex changing fish. A recent mechanistic model for socially induced sex change in the protogynous bluehead wrasse highlights an association between early amh activation and the stress hormone, cortisol 8 . Therefore, amh may form part of the molecular trigger that initiates sex change, and its upregulation could be a useful early molecular marker for protogynous species.

Conclusion
This study validates the chemical and social induction of sex change in captivity and describes the anatomical, endocrine and molecular events during sex change to establish the New Zealand spotty wrasse as a temperatewater model for sex change research. Sex change is most readily induced following the reproductive season when plasma sex steroid concentrations are low.  www.nature.com/scientificreports/ In general, the female-related markers E2 and cyp19a1a declined during sex change while the male-related markers 11KT, cyp11c1, hsd11b2 increased. The enzymes encoded by these latter genes are also actively involved in the interrenal stress response. The gonadal expression of nr3c1 and nr3c2 increased in the later stages of sex change in spotty wrasse. This could suggest a putative role for stress during sex change but there is no clear evidence for cortisol as an early-stage activator of sex change. Similarly, E2 did not seem to be an initiator of sex change in this model, but amh showed early-stage upregulation. Apart from being an interesting target to better understand early activators of sex change, amh may prove to be a beneficial marker to resolve ET stage fish from females undergoing seasonal ovarian atresia.