Seasonal flooding regime and ecological traits influence genetic structure of two small rodents

Although codistributed species are affected by the same abiotic factors, such as rivers and seasonal flooding regimes, ecological traits, such as locomotion habits and habitat preferences, may also influence differences in levels of genetic diversity and differentiation. We examined population genetic structure and diversity of Hylaeamys megacephalus and Oecomys aff. roberti, two cricetid rodent species from the mid-Araguaia River in central Brazil, using mitochondrial DNA sequence data. Specifically, we aim to test whether the Araguaia River acts as a barrier to the gene flow of these two species and to assess how ecological traits, such as locomotion habits and habitat preferences, may influence differences in levels of genetic diversity and differentiation. As both species occur in flooded forests, neither showed genetic differences related to river banks. Oecomys aff. roberti showed stronger population structure that appears to be associated with isolation by distance. This arboreal species maintained stable populations in the Araguaia River, while the terrestrial H. megacephalus was more affected by seasonal floods, resulting in a genetic signature of population expansion. Our initial predictions were largely supported by our results given that locomotion habits and habitat preferences of each species appears to have played a role on the genetic structure of these two sympatric rodent species.


Introduction
Genetic differentiation is the result of the interaction of both historical and ongoing evolutionary processes (e.g., Moritz et al. 2000;Patton et al. 2000). Landscape features such as rivers and mountains have been considered important factors in shaping the genetic structure of Neotropical species (Antonelli et al. 2010;Leite and Rogers 2013). However, the ability to disperse across these landscape features and to recover from bottlenecks is also related with species ecology and life history (Matocq et al. 2000). Indeed, dispersal tendencies, habitat preferences, life span, and mating system and success determine population sizes and gene flow rates among populations, which in turn will determine levels of genetic diversity and differentiation (Allendorf and Luikart 2007).
The Araguaia River in central Brazil is one of the main drainage basins of the Cerrado biome and lies in the transition to Amazonia (Latrubesse et al. 2009). This river represents a relatively stable geographical barrier due to its anabranching pattern, consisting of multiple channels separated by vegetated semipermanent alluvial islands, with low sinuosities (Latrubesse 2008), which are expected to limit gene flow between population of forest dwellers from opposite river banks (e.g., Aleixo 2004;Bates et al. 2004;Antonelli et al. 2010;Faria et al. 2013;Leite and Rogers 2013). River geomorphology has been argued as a possible cause of differences in the genetic differentiation of populations by river bank, and anabranching rivers like the Araguaia are probably stronger barriers than meandering rivers, such as the Juru a (e.g., Moritz et al. 2000;Patton et al. 2000;and Antonelli et al. 2010). Indeed, previous studies have shown that the Araguaia is an important barrier to sister species of the climbing rats of the genus Rhipidomys (Bonvicino et al. 2008;Rocha et al. 2011), but it is not known to what extent this finding is consistent across other cricetid rodents with different ecological traits.
Here, we examine two species, Hylaeamys megacephalus and Oecomys aff. roberti, that were the two most abundant cricetid rodents captured in the mid-Araguaia River basin . Hylaeamys megacephalus is a terrestrial and generalist species with high habitat tolerance (Emmons and Feer 1997;Percequillo et al. 2008), occurring in both flooded and unflooded forests of the Araguaia River . Oecomys aff. roberti (treated as Oecomys sp. in Rocha et al. 2011) is so far known only from the Araguaia River, where it presents arboreal habits and prefers flooded forests Ramos Pereira et al. 2013), like its closely related congener O. roberti in the Juru a River .
We used two mitochondrial molecular markers (cytochrome B and D-loop) to compare population genetic structure and diversity of these two cricetid rodent species from the mid-Araguaia River in central Brazil. Specifically, we aim to test whether the Araguaia River acts as a barrier to the gene flow of these two species and to assess how ecological traits, such as locomotion habits and habitat preferences, may influence differences in levels of genetic diversity and differentiation. These two similarsized species co-occur along the Amazonia-Cerrado ecotone, but they show distinct ecological traits, leading to different predictions regarding levels of genetic diversity and differentiation (Table 1). As both species occur in flooded forests, we hypothesize that both will exhibit low differentiation across river banks, showing haplotype sharing between river banks and islands. However, H. megacephalus will likely present lower overall genetic differentiation than O. aff. roberti, as the former is a more generalist species in terms of habitat. We also hypothesize that the ground-dwelling species, H. megacephalus, is more prone to reveal population size fluctuations following seasonal flooding regime, while the arboreal species, O. aff. roberti, is more likely to maintain stable populations in the Araguaia River throughout the year (Table 1).

Study area and sample collecting
Specimens of Hylaeamys megacephalus and Oecomys aff. roberti were collected from 26 sampling points in the mid-Araguaia River and its surroundings (Fig. 1, see gazetteer in Supplementary material for locality names and geographical localization). The Araguaia River is an anabranching river consisting of multiple channels separated by vegetated semipermanent alluvial islands excised from preexisting floodplain or formed by within-channel or deltaic accretion (Narson and Knighton 1996;Latrubesse 2008). The landscape is characterized by a mosaic of seasonally flooded and unflooded gallery forests located along the margins of the streams and channels. In welldrained areas, vegetation physiognomies more typical of the Cerrado occur (Oliveira-Filho and Ratter 2002). The climate is tropical, with a rainy season lasting from October to April and a dry season from May to September (INMET 2011). Table 1. Summary of ecological traits and genetic predictions for Hylaeamys megacephalus and Oecomys aff. roberti in mid-Araguaia River, central Brazil.

Species
Ecological traits Genetic predictions H. megacephalus Terrestrial, more prone to local extinction due to seasonal flooding regime Unimodal mismatch distribution; evidence of population expansion in neutrality tests; proportionally less haplotypes in islands and shared between river banks and islands due to higher rate of extinction Habitat generalist Weaker genetic structure (lower F ST ) and no evidence of isolation by distance Occurs in both seasonally flooded and unflooded forests Low differentiation across river banks, shared haplotypes between river banks and islands O. aff. roberti Arboreal, less prone to local extinctions due to seasonal flooding regime Multiple peaks in mismatch distribution; no evidence of population expansion in neutrality tests; proportionally more haplotypes present in islands and shared between river banks and islands due to lower rate of extinction Habitat specialist Stronger genetic structure (higher F ST ) and isolation by distance Prefers seasonally flooded forests Low differentiation across river banks, shared haplotypes between river banks and islands Sampling was carried out between June 2007 and November 2008, using a standardized trapping protocol to sample small nonvolant mammals in upland and floodplain gallery forests. Live traps (Sherman and Tomahawk) and pitfalls (30 L and 60 L) were set in each sampling point. Three sampling periods averaging seven nights (between five and nine nights) were performed in each sampling point (see Rocha et al. 2011 for detailed information on habitat, sampling, and specimens). Tissue samples have been deposited at Universidade Federal do Esp ırito Santo (UFES), Vit oria, Brazil.

DNA extraction and amplification
DNA was extracted from liver and ear tissue samples preserved in ethanol using the salt-extraction method (Bruford et al. 1992). An 801-bp fragment of the cyt B was amplified by polymerase chain reaction (PCR) using the primers MVZ05 and MVZ16 (Smith and Patton 1993) for both species. 860-bp fragments of D-loop were amplified by PCR using the primers L15774M and H651B (Fumagalli et al. 1997) for H. megacephalus. 496-bp fragments of D-loop were amplified by PCR using the primers L0 and E3 (Douzery and Randi 1997;Huchon et al. 1999) for O. aff. roberti. Amplifications were performed using the following PCR profiles: (1) cyt B: initial denaturation at 94°C for 5 min, followed by 39 cycles with denaturation at 94°C for 30 s, annealing at 48°C for 45 s, polymerization at 72°C for 45 s, and a final extension at 72°C for 5 min; (2) D-loop: initial denaturation at 94°C for 2 min, followed by 35 cycles with denaturation at 94°C for 30 s, annealing at 54°C (H. megacephalus) or 45°C (O. aff. roberti) for 90 s, polymerization at 68°C for 1 min, and a final extension at 72°C for 10 min. Mitochondrial fragments were purified using ExoSap-IT â (USB Corporation, Cleveland Ohio, US) and sequenced using an automatic sequencer ABI 3130-XL (Perkin Elmer, Applied Biosystems, Foster City, California), with the above-listed primers.
Sequence alignment was performed using CLUSTALW algorithm implemented in MEGA version 5 (Tamura et al. 2011), with posterior manual edition. All sequences generated in this study have been deposited in GenBank (Table 2).

Genetic diversity indices and population genetic structure
Number of haplotypes and polymorphic sites, haplotype and nucleotide diversity values were estimated with DnaSP v.5 (Librado and Rozas 2009) for both molecular markers from each species. Phylogeographic relationships among haplotypes (concatenated cyt B and D-loop loci) from both species were estimated through Bayesian inference (BI) in MrBayes v.3.1.2 (Ronquist and Huelsenbeck 2003). The model of nucleotide substitution was selected using MrModeltest (Nylander 2004). Haplotype trees were sampled every 100 of 10 7 generations until Markov chain became stationary, that is, when standard deviation of split frequencies was below 0.01 (Ronquist et al. 2009). A 50% majority rule consensus tree was obtained after "burn-in" of 25% of the sample points to generate Bayesian posterior probabilities (BPP). Trees were then edited in FigTree v1.4 (available at http://tree.bio.ed.ac.uk/software/figtree/).
Median-joining (MJ) networks were produced for both markers separately using NETWORK (Bandelt et al. 1999) and using only variable nucleotide sites. We divided the samples into three subpopulations based on geography (one from each river bank and another from the islands) to test the role of the river in population structuring. Spatial analyses of molecular variance were performed for both markers separately using SAMOVA 1.0 (Dupanloup et al. 2002) to identify groups of populations, which accounted for genetic differentiation and geographic distances. Analyses were performed for K = 2 to 10 to select the optimal number of groups (K), with the 26 sampling points for each species. K was selected based on the maximum (or plateau) value of fixation index F CT (variance among groups relative to the total variance) and on the minimum number of groups with single populations.
The genetic differentiation between sampling points was assessed with pairwise F ST estimates, using ARLE-QUIN 3.5.1.3 (Excoffier and Lischer 2010). The P-value was estimated using 10,000 permutations. Isolation by distance was assessed using a simple Mantel test. The correlation between matrices of pairwise estimates of genetic differentiation (F ST ) and matrix of geographic distances was estimated with the Mantel test, using Isolation by Distance web service v.3.23 (Jensen et al. 2005).

Demographic history
Neutrality and demographic history were evaluated for both markers separately from the entire sample distribution of each species using DnaSP v.5 (Librado and Rozas 2009). Populations under expansion are expected to exhibit smooth and unimodal mismatch distributions, while populations at demographic equilibrium are characterized by ragged and erratic mismatch distributions (Harpending 1994). Deviations from the sudden population expansion model were further tested using the Harpending's raggedness statistics, which quantify the smoothness of the observed distribution (Harpending 1994). We also used statistics to detect demographic events based on frequency spectrum of mutations, Tajima's (1989) D and Ramos-Onsins and Rozas's (2002) R 2 , and based on haplotype frequencies, Fu's (1997) Fs. Coalescence simulations with 1000 replicates were applied to determine the P-value of each statistics.

Genetic diversity indices
Partial sequences of cyt B and D-loop were obtained from 49 and 38 individuals of H. megacephalus, respectively, resulting in 21 haplotypes and 36 polymorphic sites for cyt B, and 23 haplotypes and 79 polymorphic sites for D-loop (Table 3). Partial sequences of cyt B and D-loop were obtained from 35 and 45 individuals of O. aff. roberti, respectively, resulting in nine haplotypes and 16 polymorphic sites for cyt B, and 11 haplotypes and 18 polymorphic sites for D-loop (Table 3). Haplotype and nucleotide diversities were relatively higher for D-loop than cyt B for both species (Table 3).

Population genetic structure
Median-joining networks from both markers revealed intraspecific haplotype sharing across the mid-Araguaia River and also on islands (Figs 2 and 3). Two H. megacephalus cyt B haplotypes (HHc1 and HHc5) and one D-loop haplotype (HHd1) were shared by specimens from both river banks; and one haplotype of each marker (HHc4 and HHd4) was also shared with specimens from the islands (Fig. 2). One O. aff. roberti cyt B haplotype (HOc1) was shared by specimens from both river banks, along with specimens from the islands; one cyt B haplotype (HOc2) and three D-loop haplotypes (HOd1, HOd2, and HOd6) were shared by specimens from the eastern river bank and the islands; and one D-loop haplotype (HOd3) was shared by specimens from both river banks (Fig. 3). Bayesian inference (BI) trees obtained for concatenated haplotypes resulted in a basal polytomy (Fig. S1). The BI tree for H. megacephalus lacked a clear geographic structure and no haplogroups could be identified (Fig.  S1A). In O. aff. roberti, on the other hand, two haplogroups were identified, but without clear geographic structure (Fig. S1B).
Using SAMOVA, we found three (K = 3) and four (K = 4) groups of populations in both species based on cyt B and D-loop, respectively (Table 4). These groups Table 3. Number of individual sequences (n), base pairs (bp), haplotypes (h), polymorphic sites (S), haplotype diversity (Hd), nucleotide diversity (p), raggedness statistics (r), and deviation from neutrality tests (D,Fs,R 2 (Table 4). Most F ST values were not significant in either species (Table S1 and S2). Only sampling points 21 and 23 revealed moderate-to-high levels of differentiation in H. megacephalus, and sampling points 12 and 20 revealed high levels of differentiation in O. aff. roberti (Table S1 and S2).

Demographic history
We found no evidence of population expansion in the mismatch distributions of either species (Fig. 4). Consistently, the observed raggedness values were not significantly lower than the expected in an expansion model for both markers (Table 3). The only significant negative value was Fu's Fs for the cyt B dataset of H. megacephalus (Table 3). All other neutrality tests showed nonsignificant  Table 1); white corresponds to the eastern river bank, black to the western river bank, and gray to the islands of the Araguaia River. values (Table 3), therefore rejecting the population expansion model.

Discussion
Although previous studies have shown that the Araguaia River is an important barrier to some small mammals (Rocha 2011;Rocha et al. 2011;Faria et al. 2013), we found no evidence of this riverine barrier on the two rodent species analyzed. Molecular phylogenies showed no geographic structure, and there are shared haplotypes across river banks and throughout the islands of the Araguaia River. This pattern of haplotype sharing indicates that these islands are important stepping stones facilitating gene flow between river banks.
Because the Araguaia River has not been acting as an effective abiotic barrier to gene flow among populations of O. aff. roberti and H. megacephalus, differences in genetic structure and diversity between these two species may be related to biotic factors, like their ecological traits. Differences in overall patterns of genetic diversity at a local scale can be explained by the combination of historical and ongoing processes of differentiation, but also by the ecological traits of each species (Patton et al. 1996). For example, Matocq et al. (2000) compared the genetic structure of two sympatric and ecologically distinct echimyid rodents, Proechimys steerei and P. simonsi, and found that the former, which inhabits the seasonally flooded forest, exhibits high gene flow, while the latter, which occupies upland, unflooded habitats, shows little or no evidence of ongoing gene flow among populations.
Hylaeamys megacephalus is a generalist species with high habitat tolerance (Percequillo et al. 2008), thus allowing higher gene flow among populations, even across long distances. The lack of correlation between genetic and geographic distances, even for samples that are very distant (e.g., sampling points 25 and 26), corroborates a scenario of gene flow instead of isolation by distance for this species. In addition, H. megacephalus is terrestrial, and we therefore expect population declines during the flooding season, followed by expansions in the dry season when more ground habitat is available. Although an  Table 1); white corresponds to the eastern river bank, black to the western river bank, and gray to the islands of the Araguaia River. erratic mismatch distribution may indicate a population in equilibrium, significant negative values in neutrality tests, such as shown by the cyt B data from H. megacephalus, as well as star-like genealogies showing very short and very long branch lengths, suggest population expansion (Ray et al. 2003). Maruyama and Fuerst (1985) also argue that population bottlenecks followed by rapid expansion increase the number of rare alleles in natural populations. Indeed, a higher number of rare haplotypes were recorded for H. megacephalus in the Araguaia River. Additionally, the maintenance of high haplotype diversity of populations that experience demographic declines during the seasonal floods is possible due to high reproductive potential of the species, allowing rapid expansion after the flooding, and to the intermixing of the surviving populations (Matocq et al. 2000).
In contrast, O. aff. roberti is more specialized, showing arboreal habits, and preference for flooded forests, thus influencing gene flow among distant populations and resulting in higher genetic differentiation. Although most F ST values were not significant, this pattern is corroborated by a significant positive correlation between genetic and geographic distances, suggesting isolation by distance. Additionally, O. aff. roberti is less prone to be affected by the flooding regime in the Araguaia River because of its arboreal habits. Indeed, O. aff. roberti showed nonsignificant values for the neutrality tests and also an erratic mismatch distribution, which indicate stable rather than expanding populations.
When comparing several small mammal species within Araguaia River basin, we corroborate previous hypothesis by Moritz et al. (2000), which pointed out that upland forest specialists are more prone to exhibit riverine diversification than specialists of floodplain forests. Indeed,    Rocha et al. 2011) genetically differentiated by the Araguaia River banks (see Rocha 2011;Rocha et al. 2011 andFaria et al. 2013).
Our initial genetic predictions based on ecological traits were largely supported by the results of both H. megacephalus and O. aff. roberti from the mid-Araguaia River. We showed that species with different ecological adaptations have different responses to physical and temporal events, and such events can leave fingerprints on the genetic diversity and differentiation patterns. Although the two rodent species co-occurring in flooded forests are not genetically structured by river bank, the fact that H. megacephalus is terrestrial and O. aff. roberti is arboreal made them more or less prone to local extinction due to the seasonal flooding regime of the Araguaia River. Additionally, differences in habitat tolerance are also important in differentiating populations from distant locations. Further comparative studies will help elucidate in which ways differences in ecological traits may influence the geographic distribution of genetic diversity. The combination of molecular and ecological studies is a powerful tool in understanding evolutionary processes that determine species diversification in the Neotropics. Tajima, F. 1989

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Bayesian inference tree of concatenated haplotypes of Hylaeamys megacephalus (A) and Oecomys aff. roberti (B).