Twelve specimens of the seawater teleost gilthead seabream (S. aurata, 25 g mean weight) were maintained at the Marine Fish Facilities of the University of Murcia. The fish were kept in 200-L aquaria of running seawater (28‰ salinity) at 24°C ± 2°C and an artificial photoperiod of 12 h light:12 h dark. The animals were fed 1% of their body weight daily with a commercial pelleted diet (Skretting, Burgos, Spain) and acclimatised for 30 days prior to the experiments. All experimental procedures were evaluated and approved by the Ethical-Scientific Committee for Animal Experimentation of the University of Murcia (protocol CEEA 701/2021) in accordance with the European Directive 2010/63/UE on the protection of animals used for scientific purposes. The animals were randomly placed into four aquaria (three animals per aquarium). Six specimens from two aquaria were then injected intraperitoneally (i.p.) with 50 µl of sterile phosphate-buffered saline (PBS, Invitrogen) (control group). The other six fish (in two different aquaria) were injected with 50 µl of 10 mg ml⁻¹ iron dextran (Sigma-Aldrich) diluted in sterile PBS (iron overload group) (Rodrigues and Pereira, 2004; Neves et al., 2015). At 72 h post-injection, five fish were randomly selected from each of the experimental groups and were anaesthetised and sacrificed by a benzocaine overdose (0.21 mM). Skin mucus samples were collected by gently scraping the surface of the gilthead seabream specimens with a plastic cell scraper, avoiding contamination of the samples with blood, urinogenital, or intestinal excretions (Guardiola et al., 2014). Blood samples were collected from the caudal vein with heparin-treated insulin syringes; after dissection, liver, head kidney (HK), anterior and posterior intestine, gill, eye, skin, spleen, and muscle samples were collected in ice and stored at −80°C until used.

Macroscopic images of the livers were obtained and immediately, liver samples (from both the control and iron overload group) were fixed in 10% formalin (Panreac) at room temperature for 24 h. After serial dehydration steps with ethanol, the samples were embedded in Paraplast (Thermo Fisher Scientific). Liver blocks were sectioned at 5 μm (Microm), and the sections were routinely stained with hematoxylin-eosin (H-E) and mounted in DPX (Sigma). Images were acquired with a Leica DFC280 digital camera coupled to a light microscope (Leica 6000B).

Frozen samples of blood, liver, HK, anterior and posterior intestine, gill, eye, skin, spleen, and muscle were lyophilised, and 100–200 mg of the resulting powder were placed in Teflon vessels with 3 ml of water, 2 ml of concentrated H₂O₂, and 5 ml of concentrated HNO₃ acid solution (Thermo Fisher Scientific). Sample digestion was performed with a Milestone ETHOS Plus Microwave system operating on a standard program (85°C, 200°C, 210°C, and 0°C during 2, 8, 10, and 20 min, respectively). Finally, the solution was diluted to 50 ml and used to determine iron concentration by flame atomic absorption spectrometry (FAAS) (VARIAN, SpectrAA 220Z ZEEMAN). The accuracy of the present results was evaluated by analysing two reference materials (DOLT-2 dogfish liver and DORM-2 dogfish muscle). Data are presented as parts per million (ppm) of Fe (milligrammes of Fe per kilogramme of dry-weight tissue), and the proportion (%) of each tissue compared to total iron in the body was determined.

The marine pathogenic bacterium Vibrio angillarum (strain R-82) was used to determine the bactericidal activity present in the skin mucus samples. Bacteria were cultured in liquid tryptic soy broth (TSB, Sigma) under agitation (24 h, 25°C, 200–250 rpm). Bacteria were adjusted to 10⁸ colony-forming unit (CFU) ml⁻¹. Bactericidal activity was determined following the method of Stevens et al. (1991) with some modifications using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay (García Beltrán et al., 2020). Samples of 20 μl of skin mucus or PBS (positive control) were placed in a 96-well U-bottom plate. Aliquots of 100 μl of the previously adjusted bacteria were added per well, and the samples were incubated (5 h, 25°C). Next, 25 μl of MTT (1 mg ml⁻¹; Sigma) was added to each well and incubated for another 10 min. The plates were centrifuged (10,000 rpm, 10 min), and the supernatant was removed. The pellet was then resuspended in 200 μl of dimethyl sulfoxide (DMSO; Sigma), which was added to each well. Finally, 100 μl of the DMSO was transferred to a 96-well flat-bottom plate, and the absorbance of dissolved formazan was measured at 570 nm. Bactericidal activity was expressed as percentage of nonviable bacteria, calculated as the difference between absorbance of bacteria surviving in test samples compared to the absorbance of bacteria from positive controls (100% growth or 0% bactericidal activity). Samples without bacteria were used as blanks (negative control). Samples without skin mucus were used as positive controls (100% growth or 0% bactericidal activity). All samples were assayed in triplicate.

Total RNA was isolated from frozen liver and HK samples with TRIzol reagent following the manufacturer’s instructions. One microgram of total RNA was treated with DNAse I (Invitrogen) to remove genomic DNA, and first-strand cDNA was synthesised by reverse transcription using ThermoScriptTM RNAse H− Reverse Transcriptase (Invitrogen) with an oligo-dT12–18 primer (Invitrogen), followed by treatment with RNAse H (Invitrogen). Fast qPCR was performed with a QuantStudio™ 5 Real-Time PCR System (Applied Biosystems) using SYBR Green PCR Core Reagents (Applied Biosystems), cDNA samples, and designed primers ( Tables 1 , 2 ) (Cordero et al., 2016a). The reaction mixtures were incubated for 10 min at 95°C, followed by 40 cycles of 15 s at 95°C, 1 min at 60°C, and finally, 15 s at 95°C, 1 min at 60°C, and 15 s at 95°C. The PCR products were subjected to melting curve analysis. Ct values of hepcidin mRNAs were normalised to the 18S housekeeping gene, and iron overload samples were compared to control cDNA for expression ratio (Livak and Schmittgen, 2001).

Protein sequences were obtained from the UniProtKB database (https://www.uniprot.org/), using the entry Pfam PF06446 (hepcidin) as a query. Of the 811 obtained (September 2022), we focused on the 196 Eupercaria sequences. From these, incomplete sequences and duplicates were removed, resulting in the sequences used in the study ( Supplementary Table S1 ). Mature hepcidin sequences were obtained after analysis of the signal peptide cleavage sites with SignalP 6.0 and mature peptides after furin action with ProP 1.0 (https://services.healthtech.dtu.dk/). Calculations and estimations on the physicochemical properties of the peptides were performed with PepCalc (https://pepcalc.com/), taking into account the formation of disulfide bonds between the cysteines present in the mature hepcidins. Sequences were aligned to default parameters using the MAFFT server (https://mafft.cbrc.jp/alignment/server/) and visualised with ESPript 4.0.10 (https://espript.ibcp.fr/ESPript) and Weblogo (https://weblogo.berkeley.edu/). Subsequently, the mature sequence tree was constructed using Mega 11 (https://www.megasoftware.net/) with the Neighbour-Joining (NJ) method, the JTT substitution model, heterogeneity between ignored sites (α = ∞) and a bootstrap of 1,000 replicates to increase the reliability of the obtained tree. The tree was visualised with iTOL (https://itol.embl.de/).

Results are presented in the figures as means ± standard error of the mean (SEM) or colour intensity in the heat maps. Statistical differences between control and iron overload groups were evaluated by a Student’s t-test in SPSS for Windows (version 15.0, SPSS Inc.). The significance level was 95% in all cases (p < 0.05).

Since the liver is a critical organ in the regulation of iron metabolism in vertebrates, not only the iron concentration in this organ was determined but also its possible alterations at the macroscopic and histological levels. At the macroscopic level, the liver of gilthead seabream appeared as a reddish organ (control group), but in fish injected with iron dextran, the livers were always darker brown, with respect to the colour observed in the liver of fish in the control group ( Figures 1A, B ). Microscopically, the hepatocyte cords in the control livers showed the typical morphology of large, rounded polygonal cells. Also, the hepatocyte cords appeared interspersed with sinusoids filled with erythrocytes in their lumen. In addition, hepatocytes showed a central, clear, spherical nucleus (with a patent nucleolus) and eosinophilic cytoplasm ( Figure 1C ).

On the other hand, the liver of iron dextran-injected gilthead seabreams showed remarkable morphological alterations, with the appearance of the organ parenchyma being totally changed only 72 h postinjection. More specifically, the organisation of the liver parenchyma was lost, including the previously described corded distribution of hepatocytes ( Figure 1D ). These cells showed their nuclei displaced towards the cell periphery and a large vacuolation was evident in the cytoplasm of the cells. As for hepatic sinusoids, the presence of blood vessel congestion and the large vacuolation of the hepatocytes made it difficult to locate sinusoid-like structures in the liver of the iron dextran-injected specimens. However, the most notable difference between the cytoplasm of hepatocytes from the control and the iron overload group was the large amount of refractile brown spherical deposits formed by hemosiderin in the cytoplasm of hepatocytes from the iron overload group, giving rise to homochromatic livers ( Figure 1D ).

Overall bactericidal activity was measured in the skin mucus of iron-overloaded fish because this constitutes the physical and chemical barrier against invading pathogens. A clear decrease in mucus protection against pathogenic Vibrio was observed with respect to the control, resulting in a 35% overgrowth of the pathogen, although a remarkable variability between samples was observed ( Figure 2 ).

The quantification of iron in several tissues and organs was performed by atomic absorption spectroscopy. The iron concentration of control fish was high in the anterior intestine (1,363.9 ppm, 54%), medium in the blood (571 ppm, 22.6%) and gills, an oxygen-retaining organ (445.1 ppm, 17.6%), low in the posterior intestine (3.6%), and very low in the liver and spleen (0.02%) ( Figures 3A, B ). In contrast, iron accumulated in the liver (20,481.5 ppm, 61%) and spleen (12,119 ppm, 36%), of the iron dextran-injected gilthead seabreams ( Figure 3 ). Furthermore, it is noteworthy that the predominant iron storage organs in the control fish were completely relegated to the liver and spleen in the iron dextran-injected fish ( Figure 3 ). In fact, no statistically significant changes in iron levels were found in either the skin or HK of control or iron dextran-injected fish.

The effect of iron overload on the expression of genes involved in iron metabolism, including the 15 hepcidin genes, was studied in both the liver (organ involved in iron accumulation and metabolism) and head kidney (the main haemopoietic organ in this fish species) of fish. While the expression of some hepcidin genes was significantly altered (p < 0.05) in the liver of fish injected i.p. with iron dextran with respect to the values recorded in control fish ( Figure 4 ), no significant change in those genes could be detected in the HK ( Figure 5 ). Thus, in the liver of iron dextran-injected fish, the expression of type I hepcidin (hamp1) was strongly upregulated (14-fold, Figure 4B ), whereas type II hepcidins maintained their mRNA levels unchanged or were even downregulated (four-fold) as in the case of hamp2.4 and hamp2.5 ( Figures 4C, D ). In contrast, the expression of some genes related to iron storage did show remarkable changes in HK ( Figure 5A ), significantly decreasing the expression of ferroportin (slc40a1) (4-fold or four-fold., p < 0.05) ( Figure 5B ), transferrin (tf) (fivefold, p < 0.05) ( Figure 5C ), and the ferritin gene fth1b (more than sixfold, p < 0.05) ( Figure 5D ). These results show that the liver and HK exhibit differential expression patterns in hepcidin and iron storage modulator genes following iron overload.

Due to the large number of Sparus aurata hepcidin sequences found in UniProt (24) with respect to the previously described genes (Serna-Duque et al., 2022), we performed a detailed bioinformatics analysis not only of these sequences but also of the hepcidin sequences deposited in that database. The result showed the existence of 806 sequences in Gnathostomata (jawed vertebrates), of which more than half (468) are found in percomorph fishes (subdivision Percomorphaceae). Within the nine well-supported series (supraordinate groups) that group the diversity of percomorphs (Sanciangco et al., 2016), Ovalentaria (191) and Eupercaria (196) have the most sequences. By manually curating the sequences of the latter supraordinal group and removing duplicate or incomplete sequences, a total of 188 sequences were obtained ( Supplementary Table S1 ). In these, the most representative orders were Perciformes (52.7%), Spariformes (21.3%), Tetraodontiformes (14.4%), and Centrarchiformes (5.9%) ( Supplementary Figure S1 ). The Sparidae (21.3%), Tetraodontidae (12.8%), Sciaenidae (9.0%), and Nototheniidae (8.5%) were the most representative among the 29 families found in this study ( Figure 6 ).

Given that hepcidins are expressed as a precursor (preprohepcidin) (Barton and Acton, 2019), which must first be processed by removing its endoplasmic signal peptide from the N-ter (prohepcidin) and subsequently converted into the mature peptide by the action of furin (a subtilisin-like endopeptidase), it was necessary to determine bioinformatically which peptide is the mature peptide for each of the sequences ( Supplementary Figure S2A ). This allowed us to determine for the first time which were the consensus sequences for both signal peptide cleavage (SSA ^↓ xP/S; Supplementary Figure S2B ) and furin cleavage (RxKR ^↓ xx; Supplementary Figure S2B ). However, in the case of this protease, not only the basic amino acids of its cleavage site must be taken into account but also its clear preference for certain amino acids at positions −7 (Y), −6 (N), and −5 (N) with respect to the cleavage site ( Supplementary Figure S2B ).

The phylogenetic tree obtained from the mature hepcidin sequences clearly shows two main clades, one corresponding to type I hepcidins (29) ( Figure 6 , red branch) and the other corresponding to type II hepcidins ( Figure 6 ). Clade 1 consists of hepcidins that bind to the ferroportin receptor, triggering its internalisation and degradation of the receptor–ligand complex, and whose consensus sequence is clearly conserved (Q-S-x-[IL]-S-[FLM]-C-x(2)-C-x(0,2)-C-x(0,1)-C-x(4)-G-x-G-x-C-C-C-C-[KR]-F) ( Supplementary Figure S3 ), unlike the nine subclades (2.1–2.9) into which type II hepcidins are divided. Thus, subclade 2.1 has only the eight highly conserved cysteine characteristics of the hepcidin motif (C-x(2)-C-C-C-C-x-C-C-C-x(3,6)-C-x(2)-C-C) ( Supplementary Figure S4 ), which stabilise the hairpin motif formed by the two antiparallel β-strands ( Supplementary Figure S2 ). This minimal motif is reinforced in subclade 2.2 with another conserved area (GVCG) in the loop upstream of the second β-strand (C-x(2)-C-C-C-C-C-x-C-C-C-x(2)-G-V-C-G-x-C-C-C) ( Supplementary Figure S5 ), which in turn is extended in clade 2.3 by the appearance of basic amino acids distributed throughout the mature peptide (G-[IFV]-K-C-C-[RK]-[FV]-C-C-x-C-C-C-x(2)-G-V-C-G-G-x-C-C-C-C-[RK]-[FK]-R-[RF]-G) ( Supplementary Figure S6 ). Subclade 2.4 has a characteristic signature in its N-ter represented by SPAD (S-P-[AK]-[DK]-C-[EQR]-F-C-x-C-C-C-[PT]-[DEN]-x(2)-G-C-[GN]-x-C-x-x-[FHY]) ( Supplementary Figure S7 ), whereas in subclade 2.5 this signature is in the C-ter with the sequence GCGTCCKF (C-R-F-C-C-C-C-x-C-C-C-x(0,1)-P-x-M-x-G-C-G-G-T-C-K-F) ( Supplementary Figure S8 ). Subclade 2.6 is arginine-rich, with a highly conserved central core that includes both the β-strands and the loop connecting them (C-[HKR]-[FL]-C-C-R-C-C-C-x(1,2)-M-x-G-G-G-C-G-[ILV]-C-C), although, within this subclade, the QGSPAR sequence in the N-ter of members of the family Sciaenidae and Tetraodontidae particularly stands out ( Supplementary Figure S9 ).

Finally, the next three subclades (2.7–2.9) are quite similar to each other (C-[KR]-F-C-C-C-x-C-C-C-P-[GDN]-M-x-G-C-G-G-G-[LV]-C-C-R-F), except for the N-ter, which ranges from the S-P-x-x of subclade 2.7 ( Supplementary Figure S10 ) to the S-P-A-G of subclade 2.9 ( Supplementary Figure S11 ), whereas subclade 2.8 ( Supplementary Figure S12 ) is highly conserved with only three positions with small mutations, belonging basically to the nine type II hepcidins of the European sea bass Dicentrarchus labrax and the single one of the hybrid striped bass Morone chrysops x Morone saxatilis.

In general, the tree shows, with some exceptions as in the case of European sea bass, that there is no family pattern within the type II hepcidins, but rather, depending on the number of mature sequences that each fish has, there is a dispersion of sequences in several subclades, implying a great variability of sequences that use as a base the same tertiary structure (two-strand antiparallel β-sheet stabilised by disulfide bridges) that gives rise to peptides with a great diversity of properties. This is clearly seen in the diversity of the 12 mature S. aurata type II hepcidin peptides ( Supplementary Table S2 ), ranging from poor to good water solubility and from a negative (−2 in A0A671WFX4) to a positive charge (+6 in B2X7F7), through the neutrality shown by A0A671WIJ8 ( Supplementary Table S2 ).