Introduction

The Amazon giant Arapaima gigas (Schinz, 1822) is the largest freshwater scaled fish in the world with adults reaching up to 250 kg and measuring over 2.5 m in total length (Nelson et al., 2016). This emblematic air-breather fish is a promising candidate species for aquaculture development and has a high-valued market in South America (Oliveira et al., 2012). The natural geographical distribution of A. gigas includes the basins of the Amazon, Tocantins-Araguaia and Essequibo rivers which cover Brazil, Ecuador, Guyana and Peru, and the species has already been introduced into several non-indigenous systems (Castello & Stewart, 2010). A. gigas is a dioecious and iteroparous species with sexual maturity reached after three to five years of age (Godinho et al., 2005).

Reproduction involves nest building by males and females in the sandy bottom of lentic habitats during the rainy season from November onwards (Castello, 2008a; Castello, 2008b). External fertilisation generally involves a single female, but often participation of more than one male, suggestive of sneaking behaviours which could help increase genetic diversity in the species (Farias et al., 2015). After spawning the nest is guarded by both parents until egg hatching, whereupon intense male parental care continues and a characteristic lateral migration towards flooded food-rich areas ensues (Castello, 2008a). Females normally reproduce multiple times within the reproductive season (Godinho et al., 2005) with a mean fecundity estimated to be c. 11,000 fingerlings per spawning event and an equal sex ratio at hatch (Neves, 1995; Queiroz, 2000). In the dry season, from June onward, water levels in the rivers decrease, marking the end of parental care (Castello, 2008a). At this stage, adults and offspring migrate back to the low lands reaching the river canals and várzea lagoons, when dispersal through the main rivers occurs (Castello, 2008a; Araripe et al., 2013). Overall, adults are not believed to migrate long distances, they are solitary and well adapted to hypoxic conditions during the

droughts (Castello, 2008a; Araripe et al., 2013). In some regions the dry season can be severe, resulting in mass mortalities of Arapaima within dried up lagoons. At such times rescuing operations are sometimes carried out for conservation reasons (Vitorino et al., 2015).

Given its obligate air-breathing behaviour, A. gigas is an easy target for fishermen and natural populations have been historically depleted or even eradicated close to the main cities (Hrbek et al., 2005). It is estimated that populations today represent only 13

% of historic levels (Castello et al., 2011) and since 1986 A. gigas has been included in the CITES red list (Castello & Stewart, 2010). While occasionally successful, reproduction of A. gigas in captivity is not a routine practice, this requiring further development of technologies for gender identification and control of spawning (Núñez et al., 2011). Therefore, fingerlings are high-priced in the aquaculture market, and their illegal capture from the wild is another challenge for conservation. Translocations of animals from one site to another has been regarded as another issue of concern because morphotypes and novel species have been described, suggesting patterns of allopatric differentiation across hydrographic basins (Stewart, 2013a; Stewart, 2013b; Watson et al., 2016).

To date, several studies have been conducted aiming to characterise the genetic diversity and structure of natural arapaima populations. These have involved the use of mitochondrial markers (mtDNA), microsatellite or inter-simple sequence repeats (ISSR) markers to study eight populations from the Amazon, Solimões and Tocantins river systems (Hrbek et al., 2005; Hrbek et al., 2007; Hamoy et al., 2008; Araripe et al., 2013), four populations from Tocantins and Araguaia (Vitorino et al., 2015) and five populations from Essequibo and Branco rivers (Watson et al., 2016). Overall, these studies found higher levels of genetic diversity within the large Amazon river basin compared to other

systems with the population structure suggesting minimal genetic flow and high genetic differentiation between populations. So far, molecular data has failed to confirm a multispecies scenario for Arapaima, which today is supported only by morphological analyses of very few specimens (Stewart, 2013a; Stewart, 2013b; Watson et al., 2016).

Further genetic research on A. gigas has focused on acquiring tools for gender identification since this species is not sexually dimorphic, a factor that has impeded captive reproduction and aquaculture development to date. First, the species karyotype was characterised (2n=56) though no obvious sex chromosome dimorphism was observed (Marques et al., 2006; Rosa et al., 2009), while later a bulked segregant analysis failed to find sex-related markers (Almeida et al., 2013). Up-to-date, next generation sequencing (NGS) technologies have not yet been applied to investigate genomic variability and diversity in populations of A. gigas.

Technologies involving the sequencing of restriction-site-associated DNA (RAD) markers allows the simultaneous discovery and genotyping of thousands of single nucleotide polymorphism (SNP) markers (Baird et al., 2008). A range of genotyping by sequencing (GBS) technologies which can be readily applied to non-model organisms that lack extensive genomic resources have become ever more popular for SNP genotyping (Zhanjiang, 2011; Hohenlohe et al., 2012; Andrews et al., 2016). A variant of the original RAD-seq protocol, double digest RAD (ddRAD, Peterson et al., 2012), provides a more scalable methodology which also reduces both time and costs of library preparation. In aquaculture, RAD sequencing has been exploited in a range of studies, including identification of sex-related markers(Palaiokostas et al., 2013; Palaiokostas et al., 2015; Brown et al., 2016), construction of linkage maps (Kai et al., 2014; Manousaki et al., 2016) and investigations of different populations and species within a genome-scale view of evolutionary processes (Saenz-Agudelo et al., 2015; Bian et al., 2016).

The appropriate RAD-seq approach to use to genotype a novel non-model species needs to be derived empirically, being dependent on many factors, including the number of informative markers desired, the quality of extracted DNA samples, the specific restriction sites present and the natural level of polymorphism within the genome, and the size of budget available for the project. A pilot ddRAD-based quantitative trait loci (QTL) study involving two families from a captive Arapaima stock, using the same restriction enzymes combination used for other studies (SbfI and SphI; i.e. Brown et al.

(2016); Manousaki et al. (2016); Taslima et al. (2017), gave disappointing low numbers of SNPs (< 200) compared to the 1-2K SNPs identified in other fish species. It was not possible to establish from this one case whether the low level of polymorphism was due to selecting inbred captive individuals or was the norm for current A. gigas populations.

Thus, this study aimed to investigate the degree of SNP polymorphism within and amongst natural populations of A. gigas from the Amazon, Solimões, Tocantins and Araguaia rivers, with the goal to identify a SNP panel that would prove useful to characterise genetic diversity and structure in these populations.

2.2. Material and Methods