Animal capture, handling and sampling followed all bioethical standards of the Spanish Government (Real Decreto 1201/2005, 10^th October 2005) and the Regional Government Xunta de Galicia (REGA ES360570202001/15/FUN/BIOL.AN/MPO01). Additionally, the captive breeding program for H. hippocampus and H. guttulatus (Project HIPPONUTRE, reference 16-02-01-FMP-54) was approved by the ethics committee from the Portuguese Veterinary Medicines Directorate for the Ministry of Agriculture, Rural Development and Fisheries. Under this approval, the program was performed in compliance with the Guidelines of the European Union Council (86/609/EU) and Portuguese legislation for the use of laboratory animals. Frozen samples of wild H. guttulatus, were donated by CCMAR. The protocol for fish sacrifice allowed to immediately and rapidly euthanize H. guttulatus specimens with an excess of anesthetic (2-phenoxyethanol solution, 0.40 mg/L). Individuals were placed in the anesthetic solution for at least 20 minutes and removed not less than 10 minutes after ventilation stopped.

Seven wild H. guttulatus were sampled from Ria Formosa coastal lagoon (36°59’ N; 7°51’ W; southern Atlantic coast; Portugal), in 2001. Seven H. hippocampus and six H. guttulatus were cultured at CCMAR (University of Algarve, Faro, Portugal) and died, from natural causes, in the year 2020. These seahorses were maintained in seawater from Ria Formosa coastal lagoon and physicochemical parameters, temperature and salinity, followed seasonal patterns with annual mean values of 20°C and 35.7, respectively. In the present study, the Practical Salinity Scale (PSS) was adopted to report salinity (IAPSO, 1985). Therefore, salinity is expressed as a conductivity ratio and has no units. Hippocampus guttulatus and H. hippocampus were fed on wild mysids Mesopodopsis slabberi and/or Diamysis lagunaris., depending on natural availability. Eight H. guttulatus and nine H. reidi and H. kuda (five replicates of H. reidi and four replicates of H. kuda), were cultured at IIM (CSIC, Vigo, Spain), presenting death dates between 2013 and 2019. These seahorses were cultured in seawater from Ría de Vigo (42°21 N; 8°36’ and 8°54’ W; southwestern Atlantic coast; Galicia, Spain), being maintained at seasonal temperatures of 15°C (winter) and 19°C (summer), in the case of H. guttulatus and at a constant temperature of 26 ± 0.5°C, in the case of H. reidi and H. kuda. The salinity mean value (37 ± 1) was the same for the three species. Specimens cultured at IIM were fed on mixtures of adult enriched Artemia (Iberfrost, Spain), commercially frozen mysids Neomysis sp. (Ocean Nutrition, Spain) and mysids collected from the wild (Siriella armata and Leptomysis sp.).

In the laboratory, seahorse samples frozen at -20°C, were rapidly rinsed with distilled water and freeze-dried (CoolSafe 55-9L Pro, Labogene, Lillerød, Denmark). The dorsal fins were separated from the rest of the body ( Figure 1 ) with the help of stainless-steel tweezers and cut into smaller portions with ceramic blades, before performing isotopic analysis. Tweezers and ceramic blades were cleaned with 99% (w/v) ethanol, between replicates of each species, to avoid cross contamination.

The whole-body replicates from species H. kuda and H. reidi (n = 9) ( Figure 1 ), were cut into smaller portions and manually macerated with liquid nitrogen, using a ceramic mortar and pestle. Samples were then placed in a mortar grinder (RM 200, Retsch, Hann, Germany) and homogenized. The mortar grinder and the ceramic mortar and pestle were cleaned with quartz powder and alcohol (70%), between samples, to avoid cross contamination (Ricardo et al., 2020). For acidic digestion, a subsample from the whole-body, of approximately 0.5 g was employed.

Fin samples, used for determining δ¹³C and δ¹⁵N, were transferred to tin capsules and weighed (2.22 ± 0.66 mg). Sample analysis was performed at Servizos de Apoio á Investigación (SAI) from University of A Coruña, by a continuous-flow isotope ratio mass spectrometry equipment (FlashEA1112 elemental analyzer, Thermo Finnigan, Italy) coupled to a Delta Plus mass spectrometer (FinniganMat, Germany) through a Conflo II interface. Carbon and nitrogen isotopic values were expressed as permil (‰), relative to Vienna Pee Dee Belemnite (VPDB) and atmospheric air, as described in the following equation:

X corresponds to ¹³C or ¹⁵N and R to ratios ¹³C/¹²C or ¹⁵N/¹⁴N. International reference materials for ¹³C (NBS 22, IAEA-CH-6 and USGS24) and ¹⁵N (IAEA-N-1, IAEA-N-2 and USGS25) were employed in analytical sample batch runs. To evaluate the precision in the analysis of δ¹³C and δ¹⁵N, acetanilide was used, resulting in a standard deviation of ± 0.15‰ (1-sigma, n = 10). Standards were run in every 10 samples. The isotopic analysis procedure fulfils the requirements of the ISO 9001 standard. The laboratory is submitted to annual intercalibration exercises (e.g., Forensic isotope ratio mass spectrometry scheme - FIRMS, LGC Standards, UK).

The acidic digestion of the samples consisted in the addition of 3 mL of concentrated nitric acid (HNO₃) and 1 mL of hydrochloric acid. After overnight digestion (14 - 16 h), the solutions were placed in a digestion block (DigiPrep, SCP Science; Montreal, QC, Canada) and three heating cycles were performed. These consisted in the increase, for 10 minutes, from room temperature to 50°C, which was kept stable during a period of 15 minutes. Temperature was then increased, in 15 minutes, from 50°C to 85°C and stabilized for another period of 15 minutes. The three heating cycles were interspersed with the addition of 2 mL of hydrogen peroxide. The drying process was performed in the digestion block at a temperature of 50°C. After evaporation, samples were diluted with Millipore water (Milli–Q) to a final concentration of 1-2% HNO₃, to reduce acid concentration and avoid damaging the inductively coupled plasma mass spectrometry (ICP-MS) equipment.

Total concentrations of 21 elements, aluminum (Al), arsenic (As), boron (B), barium (Ba), calcium (Ca), cobalt (Co), chromium (Cr), copper (Cu), iron (Fe), potassium (K), lithium (Li), magnesium (Mg), manganese (Mn), sodium (Na), nickel (Ni), phosphorus (P), rubidium (Rb), selenium (Se), strontium (Sr), vanadium (V) and zinc (Zn), were determined in an Agilent 7700 ICP-MS (Agilent Technologies, Santa Clara, CA, USA) equipped with an octopole collision cell and autosampler. Calibrations with standard solutions of each analyte were performed for element quantification. A quality program, with reagent blanks, replicate samples and two certified reference materials, Febs-1 (NRC Canada) and ERM-BB422 (JRC - European Commission), was employed in order to evaluate the accuracy of the ICP-MS method. Precision was estimated by the relative standard deviation (RSD) of replicate samples, being ≤ 10%. The elements’ concentrations were expressed as ratios of Ca (mmol/mol Ca) for posterior statistical analyses (Ricardo et al., 2017).

Statistical analyses were performed using the software R v.4.3.2 (R Core Team, 2019) and PRIMER v6 (Primer-e, Auckland, New Zealand) with the add-on PERMANOVA+. Niche regions and niche overlap of seahorses cultured in the same research centers (IIM or CCMAR), cultured H. guttulatus from different origins and wild and cultured H. guttulatus, were analyzed through a multidimensional δ¹³C and δ1⁵N bi-plot. The region of the niche was considered the probability of the density function, of both isotopes, at a confidence interval of 95%. The bi-plot accounts for species-specific distribution and bivariate projections of the niche regions (Swanson et al., 2015). The package Siber v.2.1.4, from R software, was used to obtain total convex hull areas (TA), which consist in a polygon that is drawn around the outermost points in the data, as well as trophic niche areas, corrected for small sample sizes (SEAc) (Jackson, 2023).The probability of trophic overlap, at a confidence interval of 95%, was estimated through NicheRover v.1.1.0 package, which uses Bayesian corrections (Swanson et al., 2015).

To assess the existence of significant differences (p-value ≤ 0.05) in the isotopic profiles (δ¹³C and δ¹⁵N) of fin samples from the four cultured seahorse species (H. reidi, H. kuda, H. guttulatus, and H. hippocampus), the groups originating from the same center (IIM or CCMAR), were compared through non-parametric analysis of similarities (ANOSIM). This analysis was performed on a resemblance matrix, which used normalized Euclidean distances. Additionally, isotopic fingerprints of wild and cultured H. guttulatus, cultured H. guttulatus specimens from CCMAR and IIM, as well as combined isotopic and elemental profiles of two cultured seahorse species (H. kuda and H. reidi), were analyzed using ANOSIM. The R statistic of this test can range from 0 to 1, which allows to measure differences between groups (Anderson et al., 2008).The ANOSIM tests were run with 9999 permutations. Due to the low number of replicates and after performing a Shapiro-Wilk test, for normality assessment, a non-parametric version of the t-test for unpaired samples (Mann -Whitney U) was performed in the raw data. This statistical analysis allowed to compare pairs of sample groups, in terms of stable isotopes of cultured seahorse species and cultured and wild H. guttulatus, as well as in terms of element composition of two cultured species. The Benjamini - Hochberg correction was applied whenever multiple comparisons were performed.

To evaluate the potential of using combined elemental and isotopic profiles for species discrimination, Principal Component Analysis (PCA) was performed on normalized data. Both stable isotopes (δ¹³C and δ¹⁵N), as well as elements that differed significantly (p-value ≤ 0.05, Mann-Whitney U) between two cultured seahorse species (H. kuda and H. reidi) were selected for PCA. The packages factoMineR (Husson et al., 2020), factoextra v1.0.7 (Kassambara, 2020) and corrplot v0.8.4 (Wei et al., 2017), from R software, were used for PCA.

All graphical representations were constructed with ggplot2 v3.3.0 (Wickham, 2016) package for R.

The δ¹³C mean values of dorsal fin samples from four seahorse species ranged between -20.38 ± 0.71‰ (H. kuda from IIM) and -14.37 ± 0.34‰ (H. hippocampus from CCMAR) ( Table 1 ). In the case of δ¹⁵N, the mean values ranged from 8.81 ± 0.31‰ (H. hippocampus from CCMAR) to 16.34 ± 0.82‰ (H. kuda from IIM) ( Table 1 ).

Hippocampus guttulatus from CCMAR presented the largest standard ellipse, corrected for small sample sizes, SEAc (2.98), while H. hippocampus from CCMAR occupied the smallest SEAc (0.39) ( Table 1 ; Figure 2 ). The centroids of the ellipses corresponding to species cultured at the same research center were positioned closer to each other, except in the case of H. guttulatus from IIM ( Figure 2 ).

Niche overlap estimates of seahorse species from IIM showed the highest value (55.65%) for the probability of H. kuda being found within the niche of H. reidi, while the lowest probability (0.12%) was registered for H. kuda being found within the niche of H. guttulatus ( Table 2 ). In the case of seahorse species from CCMAR, H. hippocampus’ niche almost fully overlapped (95.49%) with the one of H. gutttulatus. When comparing H. guttulatus, cultured at different research centers (IIM and CCMAR), the highest probability value, even if low (2.05%), was registered for niche overlap of specimens from IIM and specimens from CCMAR. Probability of wild H. guttulatus niche overlaying the one from H. guttulatus, cultured at CCMAR, was higher than the other way around, with a value of 59.18% ( Table 2 ).

The stable isotopic profiles (δ¹³C and δ¹⁵N) of seahorse’s (H. kuda, H. reidi and H. guttulatus) dorsal fins from IIM were significantly different (ANOSIM, p-value< 0.001). However, H. kuda and H. reidi did not differ significantly in terms of isotopic composition (ANOSIM, p-value = 0.095), while H. guttulatus and H. reidi, as well as H. guttulatus and H. kuda were significantly different (ANOSIM, p-value = 0.002 and 0.003, respectively) ( Table 3 ). Moreover, H. guttulatus and H. reidi differed significantly in δ¹³C and δ¹⁵N compositions (Mann- Whitney U, p-value = 0.003 for both isotopes), as was the case of H. guttulatus and H. kuda (Mann- Whitney U, p-value = 0.007 for both isotopes) ( Table 3 ). The two species cultured at CCMAR (H. hippocampus and H. guttulatus) were significantly different in terms of the isotopic profile (ANOSIM, p-value = 0.011), even though there were no significant differences for δ¹³C and δ¹⁵N (Mann- Whitney U, p-value = 0.317 and 0.306, respectively) ( Table 3 ). Cultured H. guttulatus sourced from CCMAR and IIM differed significantly in the isotopic profile (ANOSIM, p-value< 0.001), with δ¹⁵N as the variable contributing for significant differences (Mann- Whitney U, p-value = 0.004). Wild H. guttulatus from Ria Formosa and cultured H. guttulatus from CCMAR, revealed significant differences in the isotopic profiles (ANOSIM, p-value = 0.006), with δ¹⁵N as the variable significantly different (Mann- Whitney U, p-value = 0.008) between the two groups ( Table 3 ).

Hippocampus reidi and H. kuda were not significantly different concerning their stable isotope profile of dorsal fin samples (ANOSIM, p-value = 0.095). Thus, the potential of combining elemental fingerprints from the whole-body and isotopes from dorsal fins, was assessed. Phosphorous (P) was the element that revealed higher mean ratio values (mmol/mol Ca) in the whole-body of both species (771.782 ± 54.188 and 657.125 ± 17.534 mmol/mol Ca for H. kuda and H. reidi, respectively) ( Table 4 ). Arsenic and V were the elements with lower mean Ca ratios in both H. kuda (0.002 ± 0.004 e^-1 and 0.001 ± 0.009 e^-2 mmol/mol Ca, respectively) and H. reidi (0.001 ± 0.004 e^-1 and 0.001 ± 0.968 e^-4 mmol/mol Ca, respectively) ( Table 4 ). The combined isotopic and elemental profiles of H. kuda and H. reidi were significantly different (ANOSIM, p-value = 0.008). The elements that most contributed to the significant differences recorded between the two species were P, K, Na, Mg, Rb, Co, Sr and Ba (p-value = 0.035, Mann-Whitney U for all elements) ( Table 4 ).

The elements that revealed significant differences in the whole-body of H. kuda and H. reidi, as well as the stable isotopes from dorsal fins, were included in the PCA. The first two components explained 90.2% of group variability ( Figure 3 ). The centroids for species were distanced from each other and groups were well discriminated. The variables Mg and Na were strongly correlated since the corresponding lines were overlapped ( Figure 3 ). These two variables also revealed a strong relationship with K, Rb, P and Ba appearing, however, less correlated to the remaining variables. The two elements Sr and Co were also positively correlated, while the stable isotopes revealed a negative correlation ( Figure 3 ). Hippocampus kuda replicates presented higher values for K, Rb, Mg, P, Na and Ba than those of H. reidi, with the opposite being true for Co and Sr, which revealed higher values for H. reidi replicates ( Figure 3 ).

Isotopic profiles in animals’ whole bodies reflect dietary intake and composition (DeNiro and Epstein, 1978, 1981; Hobson and Clark, 1992; Fry, 2006; Martínez del Rio et al., 2009). However, isotopic profiles (δ¹³C and δ¹⁵N) can vary depending on tissue type and diet (DeNiro and Epstein, 1978, 1981). Valladares and Planas (2012) reported similar δ¹³C and δ¹⁵N values in dorsal fins and muscle tissues of H. guttulatus, validating both matrices for SIA. This analysis has already been applied to fish migration studies, using matrices such as otoliths (Huijbers et al., 2013) and muscle tissues (Davidsen et al., 2017). However, to the authors best knowledge, there are no studies addressing the use of dorsal fins’ isotopic fingerprints for tracing Hippocampus spp. geographic origin or for species discrimination.

The dorsal fins of H. kuda and H. reidi from IIM exhibited similar mean δ¹³C and δ¹⁵N values, as was the case for H. guttulatus and H. hippocampus from CCMAR ( Tables 1 , 3 ), expressing the dietary sources of both sites. Nevertheless, H. guttulatus from IIM displayed significantly different δ¹³C and δ¹⁵N values to the other two cultured species, H. kuda and H. reidi ( Tables 1 , 3 ). The results of a study performed with freshwater fish species, revealed that phylogeny, as well as morphology, could be associated with δ¹³C and δ¹⁵N, even though the correlation between isotopic data and phylogeny was not considered very strong (Keppeler and Winemiller, 2020). Hippocamus kuda and H. reidi are phylogenetically closer than H. guttulatus (Teske et al., 2004), which may partially explain the similarities in isotopic fingerprints recorded in H. kuda and H. reidi. Additionally, it has been described that temperature may affect δ¹³C values in H. guttulatus juveniles (Valladares and Planas, 2020), which could also be related with the differences reported, as H. kuda and H. reidi are both tropical species and, thus, were cultured at higher water temperatures than temperate seahorse species H. guttulatus. The ellipses referring to δ¹³C and δ¹⁵N showed a trend for species from the same research center being placed closer to each other, except for H. guttulatus from IIM ( Figure 2 ). Trophic overlap evidenced dietary similarities between H. reidi and H. kuda from IIM and between H. hippocampus and H. guttulatus from CCMAR ( Table 2 ). Nevertheless, H. guttulatus from CCMAR displayed a low probability of niche overlap with H. hippocampus, potentially due to intra-specific variability, increasing niche area ( Table 1 ). Factors, such as specimens’ size, may contribute to variability in δ¹³C and δ¹⁵N values (Davidsen et al., 2017). In seahorses, size can influence δ¹³C and δ¹⁵N values due to changes in prey preferences with growth (Valladares et al., 2016).

Hippocampus guttulatus cultured at IIM exhibited significantly higher δ¹⁵N values than those of conspecifics cultured at CCMAR ( Tables 1 , 3 ). These differences likely originated from dissimilarities in the diet provided at each site. A study on the pipefish Syngnathus acus revealed that juveniles, with a higher habitat dispersion than adults, exhibited isotopic heterogeneity, mainly at δ¹⁵N level, due to high dietary variability (Planas, 2022). Moreover, in the study developed by Planas et al. (2020b), dorsal fins from H. guttulatus fed on different diets (Artemia and/or a mixture of frozen and captured mysidaceans) displayed differences in the stable isotopic profiles. Nonetheless, it is noteworthy that the similarities in δ¹³C of H. guttulatus from IIM and CCMAR may be due to a similar assimilation of dietary items in both groups, while differences in δ¹⁵N likely reflect variations in trophic levels of consumed prey (Planas et al., 2020b).

Wild H. guttulatus from Ria Formosa coastal lagoon exhibited significantly higher mean δ¹⁵N values than those of cultured specimens from CCMAR ( Tables 1 , 3 ). This pattern aligns with findings for wild Sparus aurata, which also displayed higher δ¹⁵N values than cultured conspecifics (Guerra-García et al., 2023). These differences could be due to a higher trophic level occupied by fishes that feed in the marine environment (Arechavala-Lopez et al., 2013). Additionally, the relationship between δ¹⁵N values in estuarine fish and urbanization levels has been documented (Morris et al., 2015). In fact, between 2001 and 2002, when studied H. guttulatus were sourced from the wild, Ria Formosa coastal lagoon experienced changes in seagrass meadows due to abnormal wastewater discharges (Cabaço et al., 2008).

Previous studies have resorted to the combination of different analyses to improve geographic origin discrimination (Ricardo et al., 2024). Dorsal fin isotopic profiles, more specifically δ¹⁵N, effectively differentiated cultured H. guttulatus from different origins, and distinguished between wild and cultured specimens. However, this approach proved insufficient for discriminating between phylogenetically close species ( Table 3 ). As such, the potential of combining both isotopic profiles from dorsal fins, as well as elemental fingerprints from the whole body, was assessed. The elements that significantly differed between H. kuda and H. reidi from IIM were P, K, Na, Mg, Rb, Co, Sr and Ba ( Table 4 ). Seahorses are known to be rich in Na, K, P (Lin et al., 2008) and Mg (Lall and Lewis-McCrea, 2007), a finding in line with the results recorded for cultured H. kuda and H. reidi ( Table 4 ).

There are several factors that influence elemental composition of fish calcified structures, such as temperature, diet, water chemistry, physiology and genetics (Clarke et al., 2011; Kerr and Campana, 2014; Sturrock et al., 2015). In some cases, environmental factors are assumed to outweigh physiological aspects, when it comes to elemental composition, as is the case of Sr (Sturrock et al., 2014). This element has been pointed out as a marker of calcified structures, influenced by environmental conditions, such as water chemistry (Bath et al., 2000; Campana, 2005), even though physiological aspects can also exert a strong influence on its regulation (Sturrock et al., 2014).

Elements under rigorous physiological control, such as P, Na and K, could be interesting for species discrimination, since they are less easily shaped by environmental factors (Campana, 2005; Kerr and Campana, 2014). Magnesium present in calcified structures was previously used as a fish stock marker (Avigliano et al., 2019), even though the effects of environmental factors on Mg/Ca from calcified structures remains unclear (Nishimoto et al., 2010). It has also been described that Ba/Ca in estuarine fish otoliths may be negatively related with salinity, being the opposite true for temperature (Nelson et al., 2018). According to Barnes and Gillanders (2013), stock genetics also affected Ba/Ca incorporation in marine species Argyrosomus japonicus fingerlings. The differences recorded in Mg, Sr and Ba composition in H. reidi and H. kuda might be attributed to physiological regulation, as seawater parameters, such as salinity and temperature, were kept identical and stable for both species. In the case of Co, a study on the marine flatfish Scophthalmus maximus has shown that abiotic factors such as pH and temperature, do not have a significant effect on its assimilation (Pouil et al., 2018). Additionally, a study by Sturrock et al. (2014) found that the concentration of Rb in blood plasma of the marine flatfish Pleuronectes platessa was not largely influenced by physiological aspects. It has been suggested that this element’s uptake by fish is influenced not only by its environmental abundance, but also by concentrations of other metallic and alkali elements in the marine environment (Campbell et al., 2005).

It is important to highlight that the analysis of different biological matrices may result in different elemental profiles, due to aspects related with metabolic pathways (Sadeghi et al., 2020). Indeed, a study led by Uncumusaoğlu et al. (2012), revealed that cadmium (Cd), Cu, lead (Pb) and Zn compositions in the muscle and liver tissues of wild H. hippocampus differed significantly. The cultured adults of H. kuda and H. reidi surveyed in the present study were maintained under the same physicochemical conditions, with whole-body elemental analyses suggesting that inter-specific variability was the primary driver of elemental fingerprint differences. Thus, combining significantly different elements from the whole body and stable isotopes from dorsal fins could enhance seahorse species identification in apprehensions from illegal, unreported and unregulated (IUU) fisheries and trade. As seafood samples may undergo different processing methods, which sometimes may damage DNA quality (Marín et al., 2018), geochemical tools could offer an alternative for discriminating dry specimens from different seahorse species.