Contrasting population genetic structures of freshwater eels (Anguilla marmorata, A. bicolor, A. mossambica) from Madagascar

Abstract

Freshwater eels (genus Anguilla) are of major biological and commercial interest due to their complex catadromous life cycle and as a high‐value fishery resource, but the biology of tropical species is still poorly understood. This study investigates the genetic integrity, diversity and population structure of three species of tropical eels (Anguilla marmorata, A. bicolor and A. mossambica) in the western Indian Ocean, based on partial nucleotide sequences of the mtDNA control region (n = 107 samples) and the nuclear GTH2b gene (n = 91 samples) for individuals collected from two freshwater systems in northern Madagascar. No first‐generation hybrids between the studied species were found, despite frequent heterozygosities at the nuclear locus. For the mtDNA control region, all species were characterized by high haplotype diversity (ranging between 0.992 and 0.996) and nucleotide diversity (ranging between 0.035 and 0.059), patterns that are consistent with a historical population expansion during the late Pleistocene as inferred from demographic modelling. In line with previous studies, we confirmed distinct mtDNA haplogroups in A. marmorata and A. bicolor which can be attributed to the partial and/or past separation of populations across the Indian Ocean, whereas the western Indian Ocean endemic A. mossambica revealed no structure. Our findings align with existing evidence for spawning areas in the western Indian Ocean for all three species, with A. marmorata partially separated from eastern Indian Ocean populations and A. bicolor showing higher connectivity. Our findings will aid in the future management of local anguillid stocks.

1. INTRODUCTION

The genus Anguilla (freshwater eels) is the only extant representative of the teleost fish family Anguiliidae and comprises 19 species and subspecies distributed in and around the Indian Ocean, the western and southern Pacific and the northern Atlantic (e.g., Arai, ²⁰²⁰). Anguillids are renowned for a catadromous life history whereby mature individuals migrate from freshwater and estuarine habitats to oceanic spawning areas, from where larvae known as leptocephali are transported by currents back towards continental waters (e.g., Schmidt, ¹⁹²²; Tesch, ¹⁹⁷⁷; Aida et al., ²⁰⁰³). This leads to a distinctive structure of their populations, and for the temperate species (Anguilla anguilla, A. japonica and A. rostrata), a panmictic breeding strategy results in a single, broadly homogenous gene pool with high levels of genetic variation (Avise et al., 1986; Mank & Avise, ²⁰⁰³; Pujolar, Jacobsen, Als, Frydenberg, Munch, et al., ²⁰¹⁴; Gong et al., ²⁰¹⁹; Enbody et al., ²⁰²¹). Anguillids are also of high economic importance, with the import and export of eels and their products having a global value of several billion USD annually (FAO, ²⁰²⁴), and sharp declines of the widely harvested temperate species led to a shift in commercial interest towards eels from tropical regions (Arai, 2014; Dekker & Casselman, ²⁰¹⁴; Limburg & Waldman, ²⁰⁰⁹; Wirth & Bernatchez, ²⁰⁰³). A full picture of population genetic structures for all tropical species, including those with potentially multiple spawning areas, is still outstanding and thus remains a key research and management question (Aoyama, 2023; Faulks et al., ²⁰²⁵; Righton et al., ²⁰²¹).

A remarkable feature of the genus Anguilla is that several species regularly share spawning areas, which can facilitate interspecific hybridisation. The best‐studied example is represented by the Atlantic A. anguilla and A. rostrata spawning in the Sargasso Sea, with first‐generation hybrids predominantly found in Iceland while low levels of introgression are detectable throughout the continental range of both species (Albert et al., 2006; Pujolar, Jacobsen, Als, Frydenberg, Magnussen, et al., ²⁰¹⁴; Tengstedt et al., ²⁰²⁵; Wielgoss et al., ²⁰¹⁴). As a further example, frequencies of first‐ and second‐generation hybrids across three species (A. marmorata, A. megastoma and A. obscura) exceeded 20% for individuals caught on the island of Gaua in the southwestern Pacific, with hybrid breakdown and cytonuclear incompatibilities in introgressed individuals ensuring the cohesion of species (Barth et al., 2020; Schabetsberger et al., ²⁰¹⁵). To what extent hybridisation is pervasive in other tropical regions where several Anguilla species meet has yet to be documented.

The western Indian Ocean is characterized by a complex pattern of surface currents flowing in easterly, westerly and southerly directions (Biastoch et al., 2024; Marsh & van Sebille, ²⁰²¹), and is home to four species of freshwater eels, all of which are distributed across mainland Africa as well as Madagascar (A. bengalensis, A. bicolor, A. marmorata and A. mossambica; e.g., Arai, ²⁰²⁰). Investigations of their population genetic structures have so far focused on single species, either to infer their ocean‐wide diversification (e.g., Gagnaire et al., ^{2009, 2011}; Ishikawa et al., ²⁰⁰⁴; Minegishi et al., ^{2008, 2012}) or their local differentiation (Elmy et al., 2024; Frankowski et al., ²⁰²⁰). Comparative genetic inferences across several species sampled in shared freshwater habitats would provide additional information about their local natural history (see e.g. Robinet et al., ²⁰⁰⁸ for evidence on possible migration routes based on local species compositions) but are so far largely lacking.

The present study aims to provide further insights into the genetic structure of three freshwater eels (A. marmorata, A. bicolor and A. mossambica) sampled from northern Madagascar. More specifically, this study employs information drawn from nuclear and mitochondrial DNA sequences to (i) screen the sampled individuals for evidence of between‐species hybridisation and (ii) document their population structures and standing levels of genetic variation to infer their local natural histories as well as historic effective population sizes in a biogeographic context. By doing so, this study provides important information for the future conservation management of anguillids in the western Indian Ocean.

2. METHODS

2.1. Ethics statement

The study was conducted in accordance with all relevant guidelines and policies. The field work in Madagascar was conducted under the following permits: 036/22/MEDD/SG/DGGE/DAPRNE/SCBE.Re and 044/23/MEDD/SG/DGGE/DAPRNE/SCBE.Re (sampling, issued by the Secrétariat général, Direction générale de la gouvernance environnementale, Direction des aires protégées, des ressources naturelles renouvelables et des ecosystems); 7/2022/MEDD/SG/DREDD.ANLJ/CEF.512, 211/22/MEDD/SG/DREDD/ANLJ/SRF and 55/23/MEDD/SG/DREDD ALM/CEF AZK (transport, issued by the Secrétariat Général, Direction Générale de l'Environnement et du Developpement durable Analanjirofo, Cantonennement de l'Environnement et des Forêts Maroantsetra, the Secrétariat Général, Direction Regionale Analanjirofo, Service Regionale des Forêts and the Secrétariat Général, Direction Generale de l'Environnement et du Developpement durable Alaotra Mangoro, Cantonennement de l'Environnement et des Forets Ambatondrazaka) and 122 N‐EA06/MG22 and 196 N‐EA05/MG23 (export, issued by the Secrétariat Général, Direction Générale de la Gouvernance Environnementale, Direction des Aires Protégées, des Ressources Naturelles Renouvelables et des Ecosystems). The laboratory work (DNA sample processing) was approved by the ethics committee of the University of Salford.

2.2. Sample collection

Samples of all three species (n = 48 for A. marmorata, n = 32 for A. bicolor, n = 1 for A. mossambica) were collected from streams on the Masoala Peninsula (localities: Anaovandrano, Iketry, Fampotabe and Antalavia) in May 2022. Additional samples of A. mossambica (n = 37) were collected in April 2023 from the Maningory River (locality Ankarongana; Figure 1). Eels were captured by local fishermen with hook and line and by the research team using fykenets and electrofishing (Grassl ELT62II backpack unit). Captured individuals were anaesthetized in a clove oil bath until motionless. Total length, as well as the distances from the lower jaw to the anus and to the origin of the dorsal fin, were measured to the nearest millimetre. All individuals were weighed, and morphologically assigned to species based on body proportions, colouration and dentition of the upper jaw in the field following Ege (1939; for details on all captured individuals, including those for which no DNA are available, see Table S1). A small clip of the pectoral fin was taken and preserved in absolute ethanol for genetic analyses, after which the eels were immediately released.

FIGURE 1.

(a) Schematic distributions of the three study species (A. marmorata, A. mossambica, A. bicolor) in and around the Indian Ocean, with northern Madagascar highlighted; (b) Outline map of northern Madagascar, showing the two sampling locations; samples of A. marmorata and A. bicolor were collected on the Masoala Peninsula, and samples of A. mossambica were mainly collected in the Maningory River.

2.3. Laboratory work

Genomic DNA was extracted using Qiagen DNEasy Blood & Tissue extraction kits following the manufacturer's instructions and calibrated to a concentration of approximately 10 ng/μl. Following a previous study by Schabetsberger et al. (2015) and Gubili et al. (2019) on five Anguilla species in the south western Pacific, we sequenced one mitochondrial DNA locus (partial control region, CR) using the primers MTF (5′‐GACTTTAACTTCCATCCTCAACTCCC‐3′) and MTR (5′‐GGGAACTTGAAATAGGAGCCAAATGCCAG‐3′) to amplify a 527 bp fragment, and one nuclear locus (gonadotropin II‐beta subunit, GTH2b) using the primers GTH2bF (5′ TCAATCATCTTACATCTGCAGC‐3′) and GTH2bR (5′‐TAACCTTATTCTGTCAAGTGCG‐3′) to amplify a 280 bp fragment. Amplifications of the mtDNA CR locus were carried out in 20 μL reaction volumes with 2 μL of 10x reaction buffer NH₄, 0.6 μL of 50 mM MgCl₂, 2 μL of 1 mM deoxyribonucleotide triphosphate (dNTP), 0.4 μL of each primer and 0.5 μL of Taq (all reagents Bioline) and 14.5 μL polymerase chain reaction (PCR) water, using the following temperature profile: 5 min at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s at 58°C and 30 s at 72°C, concluded with 5 min at 72°C. Amplifications for the GTH2bF locus used the same 20 μL reaction mix and general thermal profile, with an annealing temperature of 52°C. Each PCR reaction was tested for successful amplification using 1.5% agarose gels prior to commercial Sanger sequencing (Macrogen, the Netherlands).

2.4. Population genetic analyses

All sequences were checked using ProSeq3 (Filatov, 2009), and aligned and trimmed on MEGA11 (Tamura et al., 2021). After trimming, the final size of the sequences was 450 bp for the CR and 266 bp for the GHT2b gene. The quality of sequences was checked on Geneious Prime 2025 (https://www.geneious.com), retaining sequences with >80% classified as High Quality. The resulting sequences were entered the BLAST function of the National Center for Biotechnology Information (NCBI) nucleotide database for species assignment (https://www.ncbi.nlm.nih.gov/), recording the percentage of nucleotide matching to already deposited sequences. For the CR sequence, we considered the result with the highest identification value, whereas for the GHT2b sequence, all results with 100% query cover as well as results at >98% identification with a query cover ranging between 95% and 99% were considered. For the nuclear sequence, A. mossambica presented heterozygous nucleotides in two individuals at position 99 (C/T), and in eight individuals at position 141 (A/G). Thus, nuclear‐phased haplotypes were estimated using Bayesian methods on PHASE v2.1.1 (Stephens & Donnelly, 2003).

Number of haplotypes (H), haplotype diversity (h), number of polymorphic sites and nucleotide diversity (π) were calculated using DnaSP 6.0 (Librado & Rozas, 2009). Molecular diversity indices were calculated on ARLEQUIN v3.5.2.2 (Excoffier & Lischer, 2010) using the pairwise difference option. ARLEQUIN was also used to compute neutrality tests, assessing Tajima's D (Tajima, 1989) and Fu's F _s (Fu & Li, 1993) with 1000 simulated samples, allowing to investigate the presence of non‐random changes in population sizes, levels of genetic drift and the presence of bottleneck events (Excoffier & Lischer, 2010). Phylogenetic trees for each locus were constructed using Geneious Prime 2025.

For the phylogenetic network analyses, the CR sequences documented in this study were integrated into existing datasets of the three species collected from the GenBank database (see Table S2; note that sample numbers were renamed for the network to increase simplicity). For A. mossambica, all sequences were used. For A. bicolor, only sequences of A. bicolor bicolor were considered to avoid taxonomic complications and for A. marmorata, only sequences from the Indian Ocean and the Indo‐Pacific Seas were included due to its large range and the large number of haplotypes available. For each species‐specific alignment, ambiguously aligned regions were removed using BMGE v1.12 (Block Mapping and Gathering with Entropy; Criscuolo & Gribaldo, 2010). This reduced the alignments by 74 (483 to 409) for A. marmorata, 17 (472 to 455) for A. bicolor bicolor and 6 (456 to 450) nucleotides for A. mossambica. A maximum‐likelihood phylogeny was inferred using IQ‐TREE v1.6.12 (Nguyen et al., 2015), with ModelFinder used to identify the best‐fitting DNA substitution model, which was chosen according to BIC (A. marmorata TIM3+F+I+G4; A. bicolor bicolor TPM2u+F+R3; A. mossambica HKY+F+G4). The phylogenetic networks were constructed using Hapsolutely v0.2.3 (Vences et al., 2024) based on Fitch distances, considering either both transitions and transversions or transversions only.

Mutation models for the CR gene were constructed using BEAUTi and BEAST v1.8.4 (Drummond & Rambaut, 2007). For A. mossambica all samples except MAD2/1 were used (due to its highly differentiated genotype, see below). The HKY nucleotide substitution model and the strict clock model were applied, using a uniform distribution for the clock rate with an initial value of 2.5⁻⁸, an upper value 2.51⁻⁸ and a lower value 2.49⁻⁸, at a chain length of 10⁸. Bayesian skyline plot (BSP) models were constructed on Tracer v1.7 (Rambaut, 2018) using the evolutionary models built on BEAST. Due to a lower level of polymorphism in the nuclear locus, we conducted this analysis with the mtDNA gene (CR locus) only.

3. RESULTS

3.1. Species identification and evidence for hybridisation

After accounting for PCR and sequencing failures, the sample sizes were n = 40, 32 and 35 for the mtDNA CR locus of A. marmorata, A. bicolor and A. mossambica, respectively (GenBank Accession Numbers: PZ698420‐698526), and n = 38, 19 and 34 for the GHT2b locus (GenBank Accession Numbers: PZ718737‐PZ718827), notably with 41.1% amplification failures for A. bicolor. The results from BLAST (NCBI) revealed a concordance between morphological and mitochondrial as well as nuclear genetic species identification for all samples (where available). For the mtDNA CR locus, 12 (30%) A. marmorata, 3 (9.4%) A. bicolor and no A. mossambica had 100% BLAST matches with previously documented haplotypes; all 40 A. marmorata, 26 (81%) A. bicolor and 17 (47.2%) of A. mossambica individuals found matches at >98.0% similarity, with one A. mossambica (MAD2/1) standing out with a low similarity of 91.5% at no other closer matches. For the GTH2b locus, the similarity was 99.3% for all A. marmorata except two individuals (99% and 100% similarities, respectively); no comparable sequences are available for A. bicolor and A. mossambica. The GHT2b sequence diagnostically identified A. marmorata, with identical sequences between some A. bicolor and A. mossambica (data not shown); it also contained a heterozygote nucleotide in ten (29.4%) of A. mossambica individuals, and in two (10.5%) of A. bicolor individuals.

3.2. Genetic diversity and population structure

The DNA polymorphism analyses (Table 1) revealed consistently high haplotype diversities for the mtDNA CR locus for all three species (h between 0.988 and 0.997), with nucleotide diversities ranging between 0.032 and 0.059; for the nuclear GHT2b locus, the corresponding values are 0.102–0.883 (h) and 0.001–0.015 (π). The total number of CR haplotypes (H) was 32 for A. marmorata, 30 for A. bicolor and 33 for A. mossambica; for the GHT2b sequence, 2, 7 and 13 genotypes were revealed for A. marmorata, A. bicolor and A. mossambica, respectively. Accordingly, the numbers of polymorphic sites (s) ranged between 76 and 121 for the mtDNA CR locus and between 2 and 13 for the GHT2b locus. Tajima's D and Fu's F resulted in negative values for all three species for the mtDNA CR locus (values ranging between −0.588 and −1.840 and between −18.284 and −10.615, respectively, with the latter being statistically significant). For the GHT2b sequence, the respective values ranged from −1.090 to 0.931 and from −3.490 to 0.843, respectively, without any statistical significance (Table 1).

TABLE 1.

Sequence polymorphism analyses for Anguilla marmorata, A. bicolor bicolor and A. mossambica of one mitochondrial (CR, control region) and one nuclear sequence (GHT2b). π, nucleotide diversity; D, Tajima's D; F, Fu's F; h, haplotype diversity; H, number of haplotypes; n, number of sequences; s, number of polymorphic sites.

Species	n	H	h	π	s	D	F
CR locus:
A. marmorata	40	32	0.988 (± 0.008)	0.032 (± 0.005)	76	−0.922 (p = 0.192)	−11.330 (p = 0.006)
A. bicolor	32	30	0.996 (± 0.009)	0.059 (± 0.029)	121	−0.588 (p = 0.299)	−10.615 (p = 0.006)
A. mossambica	35	33	0.997 (±0.008)	0.035 (± 0.003)	116	−1.840 (p = 0.018)	−18.284 (p < 0.001)
GHT2b locus:
A. marmorata	38	2	0.102 (± 0.065)	0.001 (± 0.001)	2	−1.090 (p = 0.175)	0.257 (p = 0.315)
A. bicolor	19	7	0.830 (± 0.054)	0.015 (± 0.008)	13	0.931 (p = 0.857)	0.843 (p = 0.680)
A. mossambica	34	13	0.8830 (± 0.020)	0.009 (± 0.006)	12	−0.214 (p = 0.461)	−3.490 (p = 0.083)

The phylogenetic tree of the mtDNA CR locus revealed an unambiguous distinction of the three species, presenting A. marmorata and A. bicolor as sister species to A. mossambica (Figure S1). Considering transversions only (Figure 2) as well as transitions and transversions (Figure S2), the Indian Ocean mtDNA CR haplotype networks for A. marmorata revealed haplogroups which partly combined eastern and western Indian Ocean populations. The haplotypes revealed by the present study appear to be panmictic with existing haplotypes from Reunion, forming their own haplogroup with however a small number of samples overlapping with a haplogroup predominantly from Indonesia (Ambon). The A. bicolor haplotypes, on the other hand, were rather uniformly distributed across two main haplogroups which accordingly comprised both eastern and western Indian Ocean populations (Figure 2, Figure S2). Anguilla mossambica is confined to the western Indian Ocean where it appeared to form one coherent haplogroup except for a single outlier individual (Figure S2, Table S2).

FIGURE 2.

Mitochondrial DNA (control region) haplotype networks for Anguilla marmorata (a), A. bicolor bicolor (b) and A. mossambica (c) in the Indian Ocean, considering transitions only. For sample details, see Table S2.

The BSPs revealed a historic increase in effective population size for all species. A. marmorata was subjected to a slight decline between 200,000 and 150,000 years ago (BP) followed by a rapid expansion plateauing around 100,000 BP. Anguilla bicolor shows a steady increase that started 900,000 BP, plateauing around 200,000 BP. A. mossambica shows a rapid increase in population size that started around 200,000 BP, plateauing around 50,000 BP (Figure 3).

FIGURE 3.

Bayesian Skyline Plots based on partial control region mtDNA sequences for Anguilla marmorata (a), A. bicolor (b) and A. mossambica (c) to model their effective population sizes over time.

4. DISCUSSION

This study focused on the population structure of A. marmorata, A. bicolor and A. mossambica sampled from freshwater habitats in Madagascar, employing the same genetic markers (sequences of the partial mtDNA control region and a nuclear sequence) which were previously used for five Anguilla species in the southwestern Pacific (including A. marmorata; Schabetsberger et al., ²⁰¹⁵; Gubili et al., ²⁰¹⁹). The amplification of the partial mtDNA control region using the MTF/MTR primer set continued to consistently yield in sequence data. The amplification of the nuclear GHT2b locus, however, was unsuccessful for 13/32 (41.1%) of A. bicolor samples, and the obtained sequences were not fully diagnostic to separate A. bicolor from A. mossambica (Speranza, 2025). We thus retained the nuclear sequence information to provide partial insights into possible between‐species hybridisation but conducted the inferences of population structures and demographic histories on mtDNA information alone.

4.1. Genetic evidence for hybridisation

Based on full concordances between morphological, mitochondrial and (where applicable) nuclear genetic species designations, we found no evidence for hybrid individuals. This finding is in stark contrast to, for example, the island of Gaua in the southwestern Pacific, where >20% of individuals across three species were found to be of hybrid origin, providing evidence for interspecific matings in a common spawning area (Barth et al., 2020; an initial study was based on the same set of genetic markers as the present investigation; Schabetsberger et al., ²⁰¹⁵). For the present study site in the western Indian Ocean, a hydrodynamic model showed that the local distribution of leptocephali and glass eels is compatible with a spawning area east of Madagascar that is shared between the study species (Pous et al., 2010; but see also Robinet et al., ²⁰⁰³). If this is indeed the case, then the lack of first‐generation hybrids observed herein could be due to a lower propensity of the studied species assemblage to undergo interspecific matings, reduced fitness of individuals resulting from interspecific matings or due to temporal segregation of reproduction (as suggested by Pous et al., ²⁰¹⁰). Notably, however, we also found a high number (n = 12) of individuals that were heterozygote in the nuclear GHT2b gene, which in the South Pacific supported the A. marmorata—A. megastoma hybrid status of individuals bearing discordances between mtDNA and morphology (Schabetsberger et al., 2015). Given their high proportions in the absence of evidence for first‐ or second‐generation hybrids and the lack of comparative sequence data from non‐overlapping parts of the global A. mossambica and A. bicolor ranges, we tentatively interpret such genotypes as within‐species polymorphisms rather than evidence for ongoing between‐species introgression. Given the inability of the nuclear sequences to diagnostically separate A. mossambica and A. bicolor, they could however be indicative of interbreeding earlier in the species' evolutionary histories.

4.2. Genetic diversity and population demography

The levels of similarity of the mtDNA haplotypes revealed by the present study with sequences available on GenBank generally reflect the amount of existing information for each study species. For example, this is the first quantitative investigation of A. mossambica based on the mtDNA CR locus (the two already available haplotypes stem from single specimens; Ishikawa et al., ²⁰⁰⁴; Minegishi et al., ²⁰⁰⁵), and all haplotypes presented in this study have not been previously recorded.

For A. marmorata, we revealed high CR haplotype diversity and a high number of polymorphic sites, similar to other studies conducted in the Indian Ocean and the western Pacific based on the same locus (Gubili et al., 2019; Minegishi et al., ²⁰⁰⁸; Nguyen et al., ²⁰²²; Schabetsberger et al., ²⁰¹⁵) as well as based on other loci (Arai & Taha, 2021; Marini et al., ²⁰²⁴). The same is the case for A. bicolor, with high haplotype diversity as previously documented in South East Asia (Marini et al., 2021). Despite being restricted to the western Indian Ocean and possibly linked to higher demographic stability, A. mossambica possessed a number of haplotypes and polymorphic sites, which were comparable to the more widespread A. marmorata and A. bicolor, with the lowest haplotype diversity and a more pronounced signature of population expansion based on Tajima's D and Fu's F. The latter test statistics have a generally higher statistical power to detect recent population expansions and were significant for all three study species, whereas a Tajima's D is further indicative of positive selection and was not significant for A. marmorata and A. bicolor (species with more partitioned gene pools and occurring in the western and eastern Indian Ocean; a similar pattern was also found by Gubili et al., ²⁰¹⁹ for Pacific eels).

Despite differential levels of significance, the generally negative Tajima's D and Fu's F are in line with the modelled effective population sizes, for A. marmorata and A. mossambica revealing an increase primarily in the last 150,000 years which coincides with sea level rises during the late Pleistocene (for other examples from the southwestern Pacific see Visram et al., ²⁰¹⁰; Hoareau et al., ²⁰¹²). Such a pattern of population increase was also observed for four eel species, including A. marmorata, in the southeastern Pacific (Gubili et al., 2019), as well as for other A. mossambica populations from the western Indian Ocean (Frankowski et al., 2020). The effective population size of A. bicolor, on the other hand, appeared to increase more steadily over a longer time frame of one million years, similar to populations from South East Asia (A. bicolor pacifica, Marini et al., 2021) and further corroborated by population rises over shorter timescales in the Indo‐Pacific (Minegishi et al., 2012; spanning 14,000 years) and East Africa (Elmy et al., 2024; spanning 250,000 years). The difference between A. bicolor compared to the other two study species could be due to its population split in the Indian Ocean approximately 140,000 years ago, followed by secondary contact thereafter (Minegishi et al., 2012). It however should be borne in mind that the inferences are based on a single locus with limited variability and should thus be treated with caution.

4.3. Local population genetic structure, with implications for spawning migrations

A main finding of the present study is that the investigated eel species differ in their population structure, representing a single (A. mossambica) or several (A. bicolor and A. marmorata) mtDNA haplogroups despite their syntopy in northern Madagascar. Two non‐mutually exclusive explanations can account for this observation. Firstly, the three species underwent differential evolutionary histories, with the existence of several haplogroups indicative of past population subdivisions. Secondly, distinct haplogroups within given species can arise through the current use of separated spawning areas followed by sympatry during the freshwater phase.

A. marmorata is currently the only anguillid which clearly lacks panmixis, leading to the assumption of multiple spawning areas across its range (for a summary see Faulks et al., ²⁰²⁵). It is distributed across the Indian as well as Pacific Ocean, with the emergence of the Sunda and Sahul shelves during the Pleistocene having potentially led to a temporary separation of populations on either side of the Indo‐Australian archipelago (as observed in a range of taxa, e.g., Hobbs et al., ²⁰⁰⁹). In addition to genetic groups partitioned within the Pacific Ocean, A. marmorata is further separated into eastern and western populations in the Indian Ocean, with suspected spawning areas west of Sumatra and east of Madagascar, respectively (Gagnaire et al., 2009, 2011; Ishikawa et al., ²⁰⁰⁴; Minegishi et al., ²⁰⁰⁸; Watanabe et al., ²⁰⁰⁸). Our finding of a dominant haplogroup in A. marmorata in addition to more distant individuals which are part of otherwise Asian populations conforms to previous studies in the western Indian Ocean (Ishikawa et al., 2004; Minegishi et al., ²⁰⁰⁸). This pattern has previously been interpreted as secondary contact of distinct eastern and western populations through occasional unidirectional larval dispersal westward along the South Equatorial Current (Gagnaire et al., 2009), with a lack of population isolation further evidenced through admixed individuals and low levels of nuclear divergence (Gagnaire et al., 2011; Minegishi et al., ²⁰⁰⁸). A western Indian Ocean spawning ground which separates the local A. marmorata populations from the eastern Indian Ocean is also in line with otolith growth rates in leptocephali that suggest a shared environment (Robinet et al., 2008), as well as a local hydrodynamic model based on the distribution of larvae (Pous et al., 2010; leptocephali of A. marmorata were already found northeast of Madagascar during the DANA expedition in 1929/30; Jespersen, ¹⁹⁴²).

For A. bicolor, the individuals from Madagascar were rather equally distributed across two previously documented distinct haplogroups without pronounced geographic signatures across the Indian Ocean (Minegishi et al., 2012; Arai & Taha, ²⁰²¹; but see also Marini et al., ²⁰²¹), leading to a less compact clustering of individuals compared to the majority of the A. marmorata individuals from Madagascar (Figure 2). The two haplogroups are best explained by ancestral population separation through the rise of the Mascarene ridge, followed by the mixing of populations through east–west migration loops facilitated by the South Equatorial and Equatorial Counter Currents, the long larval period of A. bicolor, and its almost year‐long breeding (Minegishi et al., 2012). A clear subdivision into two main haplogroups was however not detected for populations from Tanzania (Elmy et al., 2024). Additonally considering the evidence for separate migration routes for leptocephali in the western Indian Ocean based on otolith growth rates (Robinet et al., 2008), this also gives rise to the possible exististence of multiple spawning grounds or partially allochronic breeding cohorts (see also Robinet et al., ²⁰⁰³, who, based on glass eels sampled in a fishery of Réunion Island, provide evidence for an A. bicolor spawning ground which is separate from A. marmorata and A. mossambica). Future studies on A. bicolor should use multi‐locus nuclear markers to investigate its subtle local population genetic differentiation (as e.g., conducted for A. marmorata, Gagnaire et al., 2011).

In contrast to our other study species (A. marmorata and A. bicolor), the western Indian Ocean endemic A. mossambica revealed no population substructure, in line with findings by Frankowski et al. (2020) across a wider area and thus providing evidence for a single panmictic spawning area. During the DANA expedition (Jespersen, 1942), leptocephali of A. mossambica were collected in the same area as A. marmorata, but at however different developmental stages and thus suggesting a potential use of the same spawning area at different times (see also Pous et al., ²⁰¹⁰).

The conservation management of Anguilla species in the western Indian Ocean is currently hampered by incomplete information on migration, spawning areas and population structure. The main finding of the present study is that we found no indication of between‐species hybridisation despite our data conforming to previous evidence of at least partly shared spawning areas. We also show that, despite being sampled in nearby freshwater environments in Madagascar, the three study species were characterized by markedly differing haplotype diversities, likely reflecting their range‐wide evolutionary histories and confirming existing evidence from previous investigations (Ishikawa et al., 2004; Minegishi et al., ²⁰⁰⁸). These findings have important implications for the management of these populations in view of ensuring their sustainable viability. Future studies should consider a wider systematic sampling of eels along the western Indian Ocean to investigate the possibility of more localized hybrid occurrences (in the North Atlantic, for example, hybrids are only found on Iceland, and in the South Pacific, a hybridisation hotspot is represented by Gaua; Barth et al., ²⁰²⁰; Tengstedt et al., ²⁰²⁵). In view of the increasing harvest demand and thus need to manage the existing stocks, this would also facilitate a better picture of the spatial distribution of the documented A. marmorata and A. bicolor haplogroups across Madagascar and the African continent.

AUTHOR CONTRIBUTIONS

Giada Speranza: Conducted laboratory work and data analysis; co‐wrote and reviewed the manuscript. Estelle Razafindranaivo: Coordinated and conducted field sampling and data collection. Denis Miranto Andriamahefa: Conducted field sampling and data collection. Meelis Tambets: Conducted field sampling and data collection. Marc Sztatecsny: Conducted field sampling and data collection. Fanomezana Ratsoavina: Coordinated the international collaboration between the Universities of Antananarivo and Salzburg; obtained the research permits. Zafimahery Rakotomalala: Coordinated the international collaboration between the Universities of Antananarivo and Salzburg; obtained the research permits. Julia M. I. Barth: Contributed to the data analysis, reviewed the manuscript, funding acquisition (co‐PI). Michael Matschiner: Helped with the data analysis, reviewed the manuscript, funding acquisition (co‐PI). Chrysoula Gubili: Supervised the study, coordinated and contributed to the data analysis, reviewed the manuscript, funding acquisition (co‐PI). Robert Jehle: Supervised the study, co‐wrote and reviewed the manuscript, funding acquisition (co‐PI). Robert Schabetsberger: Conceived and supervised the study, coordinated international collaboration, coordinated and conducted field sampling and data collection, contributed to data interpretation, reviewed the manuscript, funding acquisition (PI). All authors approved the final version of the manuscript.

FUNDING INFORMATION

This study was funded by the Austrian Science Fund (FWF, Grant number P34091‐B; Principle Investigator R. Schabetsberger). Part of the laboratory work was additionally funded by the University of Salford. Chrysoula Gubili was supported by a Royal Society Wolfson Visiting Fellowship (RSWVF\R2\232003).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Figure S1. Phylogenetic tree for Anguilla marmorata, A. bicolor bicolor and A. mossambica based on partial mtDNA control region sequences for samples collected in Madagascar. For sample details, see Table S2.

Figure S2. Mitochondrial DNA (control region) haplotype networks for Anguilla marmorata (A), A. bicolor bicolor (B) and A. mossambica (C) in the Indian Ocean, considering transitions and transversions. For sample details, see Table S2.

Table S1. List of individuals recorded during field work and their measurements. Y: sampling year; L: sampling location (A – Anovandrano, collected 2022; M – Maningory River, collected 2023); W: weight (kg); T: total length (cm); A: snout‐anus length (cm); D: snout‐dorsal fin length (cm); G: gill opening (cm); M: mouth (cm); Eh (horizontal eye diameter, mm); Ev (vertical eye diameter, mm); P: left pectoral fin length (mm).

Table S2. Sample designations used for the network analyses based on partial control region mtDNA sequences for Anguilla marmorata, A. bicolor bicolor and A. mossambica. GenBank accession numbers reflect sample numbers detailed in Table S1, with sample designators unified for Figure 2.

Speranza, G. , Razafindranaivo, E. , Andriamahefa, D. M. , Tambets, M. , Sztatecsny, M. , Ratsoavina, F. , Rakotomalala, Z. , Barth, J. M. I. , Matschiner, M. , Gubili, C. , Jehle, R. , & Schabetsberger, R. (2026). Contrasting population genetic structures of freshwater eels (Anguilla marmorata, A. bicolor, A. mossambica) from Madagascar. Journal of Fish Biology, 109(4), 2409–2418. 10.1111/jfb.70612

DATA AVAILABILITY STATEMENT

All newly created sequences are available through GenBank (CR locus: Accession Numbers PZ698420‐698526; GHT2b locus: Accession Numbers PZ718737‐PZ718827).