Diversification of Prochilodus in the eastern Brazilian Shield: Evidence from complete mitochondrial genomes (Teleostei, Prochilodontidae)

1 INTRODUCTION

The freshwater fish order Characiformes includes approximately 2260 species, distributed in 24 families with about 90% being endemic to the Neotropical region with the others found in Africa (Betancur-R et al., 2018; Fricke et al., 2020; Oliveira et al., 2011). One of the Neotropical endemic clades is the family Prochilodontidae, popularly known as curimbatás, bocachicos, or flannel-mouthed characiforms, that is comprised of three genera: Ichthyoelephas Posada 1909 (two species), Prochilodus Agassiz 1929 (13 species), and Semaprochilodus Fowler 1941 (six species) (Castro & Vari, 2004). Prochilodus encompasses bottom-feeding, detritivorous species of medium-size and is morphologically characterized by highly specialized lips, teeth, and jaws (Castro & Vari, 2004; Melo & Sidlauskas, 2017). They occur in all major South American basins and are commercially important for industrial and regional fisheries due to their natural abundance (Batista & Petrere Jr, 2003; Castro & Vari, 2004).

The taxonomic revision of the family delimited 13 species of Prochilodus, with five being from southern and southeastern South America: (i) P. argenteus Spix & Agassiz 1829, (ii) P. costatus Valenciennes 1850, (iii) P. harttii Steindachner 1875, (iv) P. lineatus (Valenciennes 1837), and (v) P. vimboides Kner 1859, but their interspecific relationships were not resolved entirely (Castro & Vari, 2004). A recent molecular phylogeny revealed that P. vimboides from coastal Brazilian rivers is sister to all remaining Prochilodus, and the other four species were placed in two well-supported clades: one composed by P. argenteus (Rio São Francisco) sister to P. harttii (Rio Pardo and Rio Jequitinhonha), and another composed by P. lineatus (Paraná-Paraguay system and Rio Paraíba do Sul) nested within the paraphyletic P. costatus (Rio São Francisco) (Melo et al., 2016). Further analyses using a mtDNA dataset and a larger taxon sampling supported the non-monophyly of P. costatus and P. lineatus (Melo et al., 2018), but a recent study using the mitochondrial control region obtained the monophyly of these two species (Chagas et al., 2020). However, all these studies used a single or few loci resulting in low node support in many instances along the topologies (Chagas et al., 2020; Melo et al.,2016, 2018), and thus some clades in the phylogeny of Prochilodus remain to be explored by a large-scale molecular analysis.

Both complete and nearly complete mitochondrial genomes are available for characiform fishes (e.g., Landínez-García et al., 2016; Nakatani et al., 2011; Saitoh et al., 2003), but a combined phylogenetic analysis has not been conducted yet. For Prochilodus, complete mitogenomes have been generated for P. argenteus, P. costatus, and P. lineatus, all presenting similar gene annotations and phylogenetic analyses supporting the monophyly of the genus (Chagas et al., 2016; Carmo et al., 2016). However, these studies did not include more than one specimen to test the monophyly of the species nor did they use other prochilodontids. Thus, gathering mitogenome data from multiple specimens from each species is crucial to test the monophyletic status of P. costatus and P. lineatus and to assess the timing for diversification and biogeography of Prochilodus.

Although recent advances have been achieved in understanding the evolutionary relationships of Prochilodus (Chagas et al., 2020; Melo et al., 2016, 2018), the biogeographic history of the genus remains uncertain, especially in the eastern Brazilian Shield. A large-scale phylogeographic analysis among Prochilodus species from three hydrographic basins using mtDNA data showed that the phylogenetic structure is not associated with geography (Sivasundar et al., 2001). However, it is unclear which biogeographic events are related to the diversification of Prochilodus from the Rio São Francisco and those from adjacent basins (i.e., Paraná-Paraguay system and Pardo/Jequitinhonha rivers) or the geologic factors potentially driving speciation processes. Although some studies have tested the influence of geomorphological structures (e.g., faults and lineaments) in the current distribution of several freshwater fish taxa (Abreu et al., 2019; Buckup, 2011; Ribeiro, 2006), few empirical studies have addressed biogeographic hypotheses along the eastern Brazilian Shield (Camelier & Zanata, 2014; Pereira et al., 2013; Wendt et al., 2019), and none of these studies have directly combined time-calibrated phylogenies and geological evidence to explain species diversification in the eastern Brazilian Shield.

In this study, we characterized and analyzed mitochondrial genomes of five Prochilodus species from the eastern Brazilian Shield: P. argenteus, P. costatus, P. lineatus, P. harttii, and P. vimboides, with the mitogenomes of the latter two being described for the first time. Our aims were to (i) test the monophyly of P. costatus and P. lineatus, (ii) estimate the divergence time for Prochilodus diversification, and (iii) associate the timing of interspecific diversification with the influence of recent geologic reactivations of faults and lineaments on the distribution and speciation processes throughout the eastern Brazilian Shield.

2 MATERIAL AND METHODS

3 RESULTS

3.2 Comparative mitogenome analyses

We analyzed two different and complete mitochondrial genomes for Prochilodus argenteus and P. costatus, and three for P. lineatus, comparing the newly generated mitogenomes with those previously available at GenBank. The number of differences (substitutions and indels) in the nucleotide sequences among different mitogenomes from the same species varied between 26 and 102 (Alignment S1). The mitogenomes of the sister species P. harttii and P. argenteus differed by 330 nucleotides. Complete mitogenome intraspecific variation ranged from 31 variable sites for P. costatus to 102 for P. lineatus. The pairwise K2P distances for all analyzed mitogenomes varied between 0.2% and 6.4% (Table S2) and p-distances varied between 0.16% and 6.9% (Table S3). The overall K2P divergence of mitogenomes from the same species ranged from 0.002 to 0.05, while the divergence among species varied from 0.009 to 0.064. As expected, pairwise distances for each region compared to the base genome (P. lineatus) revealed that the control region D-loop presented the highest number of polymorphisms followed by the ND4 gene (Figure 1). Results indicated a similar amount of variation in most genes (e.g., CYTB, ND1, and COX1) but very low variation in the 12S, 16S, COX2, and the rRNA ribosomal genes (Figure 1).

3.3 Phylogenetic relationships and timing of diversification

The aligned matrix contained 11,641 bp with 6909 variable sites (59%). DAMBE did not detect any saturation in transitions and transversions, indicating that the matrix was useful for phylogenetic analyses. The characiform tree based on mitogenomes shows high posterior probabilities for most nodes (Figure 2; Figure S1). The phylogeny reveals Prochilodontidae as sister to Curimatidae and this two-family clade being sister to Anostomidae and Chilodontidae, forming the traditional and well-established superfamily Anostomoidea (Vari, 1983) with Parodontidae as immediate sister group. It also shows the divergence time for the analyzed species of Characiformes (Figure 2; Figure S1). All represented families appear as monophyletic, and the time-calibrated analysis shows the split of Prochilodontidae and Curimatidae during the Eocene at around 42 MYA (61–25 MYA, highest posterior density, HPD), the earliest divergence in Prochilodontidae, that is, the split of Ichthyoelephas and Prochilodus, occurring at around 33 MYA (52–14 MYA, 95% HPD) (Figure 2). The split of P. vimboides is estimated to have occurred in the Miocene at around 11 MYA (20–5 MYA, 95% HPD), and the divergence between the clades from the eastern Brazilian Shield (P. argenteus+P. harttii and P. costatus+P. lineatus) occurring more recently in the Pliocene at approximately 3.3 MYA (5.6–1.7 MYA, 95% HPD) (Figures 2, 3). Within the two clades, we estimated that the interspecific divergences occurred contemporaneously during the Quaternary, more specifically in the Early Pleistocene: P. argenteus (Rio São Francisco) split from P. harttii (Rio Jequitinhonha) at around 2.1 MYA (3.8–0.9 MYA, 95% HPD), and P. costatus (Rio São Francisco) split from P. lineatus (Rio Grande/Paraná) at around 1.1 MYA (2.1–0.5 MYA, 95% HPD). Figure 2 shows the relationships among related characiform families, and Figure 3 depicts only Prochilodus relationships.

4 DISCUSSION

This study presents six new complete mitochondrial genomes for five Prochilodus species, with two being novel to science: P. harttii and P. vimboides. The combined mitogenomes allow the comparison of intraspecific features for the first time, as only one mitochondrial genome for each species was previously available. For example, the mitogenomes of P. lineatus presented more nucleotide variation than those of P. costatus, suggesting that the broad occurrence of the former in the Paraná-Paraguay system might help to increase intraspecific genetic variability. Among species, the highest interspecific divergence observed here (0.064) is still lower than that of the COX1 approach exclusively (0.103) (Melo et al., 2018), suggesting that some mtDNA regions other than COX1 might promote a reduction in the phylogenetic signal. In fact, our analyses indicate that all mitochondrial regions, except D-loop, present a genetic distance of no more than 0.10 among the five species (Figure 1). These results strengthen a previous hypothesis suggesting a very low genetic variation among Prochilodus species (Melo et al., 2018), likely due to their exceptional ability to perform long-distance migrations (Carolsfeld et al., 2003; Godinho & Kynard, 2006).

The phylogeny generated by new data of complete mitochondrial genomes exactly matches the relationships presented by the multilocus phylogeny of the family (Melo et al., 2016) by revealing Prochilodus vimboides as the sister clade to all other Prochilodus and the presence of two species pairs: P. argenteus sister to P. harttii, and P. costatus sister to P. lineatus (Figures 2, 3). The position of P. vimboides in the phylogeny of Prochilodus is biogeographically intriguing since the species has a restricted distribution in southeastern Brazil, while the other clade has species broadly distributed across South America from northern Colombia to the La Plata estuary (Castro & Vari, 2004; Melo et al., 2016). The time-calibrated phylogeny indicates that P. vimboides split from the remaining species during the Middle-Late Miocene boundary at around 11 MYA and suggests that the putative events related to this speciation process were certainly not the same as those for the other two species pairs from the eastern Brazilian Shield, which occurred later in the Early Pleistocene (Figure 3). However, the lack of mitogenomes for species of the remaining clades of Prochilodus from other Neotropical regions prevents any further exploration of biogeographic factors that may have promoted the diversification of P. vimboides relative to other species, leaving this topic an open question in the evolutionary history of the genus.

The cladogenetic event during the Pliocene at around 3.3 MYA splits the most recent common ancestor (MRCA) of all four species into two major clades: Prochilodus argenteus sister to P. harttii and P. costatus sister to P. lineatus. At first glance, this intuitively suggests that the MRCA lived in the Rio São Francisco region and had two subsequent dispersal paths: one toward the east to Rio Pardo and Rio Jequitinhonha (P. harttii) and another one southerly to Rio Grande/Paraná basin (P. lineatus). However, the phylogeny including all species of Prochilodontidae (Melo et al., 2016) indicates that these two clades appear in a successive arrangement with the clade P. costatus and P. lineatus being actually sister to the P. nigricans group, which is comprised of species mostly from Greater Amazonia (Amazon, Orinoco, and Guianas) and north-eastern Brazil (Melo et al., 2016). Furthermore, recent evidence from other Neotropical freshwater fish groups such as pimelodids (Tagliacollo et al., 2015), loricariids (Silva et al., 2016), bryconids (Machado et al., 2018), and curimatids (Melo et al. in review) indicates that most dispersal events occurred from the proto-Orinoco-Amazonas region to the craton drainages of the Brazilian and Guiana shields. In addition, the dispersal events to the Brazilian Shield were likely catalyzed by mega river captures between the wetlands of the proto-Orinoco-Amazonas and the proto-Paraguay basin (Albert et al., 2011, 2018; Tagliacollo et al., 2015). Therefore, the biogeographic scenario involving P. lineatus in the Paraná-Paraguay system and Rio Paraíba do Sul is probably more complex than we expected, and further studies including species of the P. nigricans group (Melo et al., 2016) from northern South America will reveal additional and complementary evidence to address this question.

We provide evidence that P. argenteus of the Rio São Francisco split from P. harttii of the Rio Jequitinhonha during the Early Pleistocene at around 2.1 MYA (Figure 3). The Serra do Espinhaço mountain chain divides the São Francisco basin of the several Atlantic-draining systems such as the Pardo, Jequitinhonha, Mucuri, and Doce rivers. The most recent geologic activities in the region are related to the Rio Araçuaí Fault during the Quaternary <1.6 MYA (Saadi et al., 2002), thus coinciding approximately with the split of P. argenteus and P. harttii in the time-calibrated phylogeny (Figure 3). Geologic evidence also suggests that post-Pliocene events in the Planalto Setentrional da Serra do Espinhaço reactivated thrust faults and, as a consequence, the upper reaches of the eastern tributaries of the São Francisco basin were captured by the upper Rio Jequitinhonha (Saadi, 1995). This is also supported, for example, by the relationship between Megaleporinus reinhardti (Lütken, 1875) from the São Francisco basin and M. garmani (Borodin, 1929) from Rio Jequitinhonha (Ramirez et al., 2017), and by the shared presence of native species in both basins such as Leporinus taeniatus Lütken, 1875 and Steindachnerina elegans (Steindachner, 1875) (Pugedo et al., 2016). In this context, we hypothesize that river captures promoted by Early Pleistocene reactivations of the Rio Araçuaí Fault in the Serra do Espinhaço may have promoted lineage split with subsequent speciation in this clade. Recent introductions have later established populations of P. argenteus into the Rio Jequitinhonha (Castro & Vari, 2004; Pugedo et al., 2016) that now hybridizes with its sister species P. harttii (Sales et al., 2018).

The mitogenome phylogeny supports the monophyly of both Prochilodus costatus from the Rio São Francisco and P. lineatus from the Rio Grande of the Paraná-Paraguay system (Figure 3). These results reject recent hypotheses using multilocus and barcoding studies that questioned their monophyly (Melo et al., 2016, 2018). Another recent study using the control region also found their monophyly (Chagas et al., 2020), although a marker such as this has extremely high levels of polymorphism (see Figure 1) meaning it is rather useful at a population level (e.g., Sivasundar et al., 2001; Machado et al., 2017). With this result, we corroborate the taxonomic revision and morphological phylogeny supporting their validity and monophyly (Castro & Vari, 2004) and reveal that complete mitochondrial genomes can provide distinct results compared to single or multilocus approaches. It is also possible that mitochondrial introgression and/or hybridization in Prochilodus (e.g., Sales et al., 2018) might affect the phylogenetic reconstruction. Finally, the lack of samples from the Rio Paraíba do Sul prevents a test for the monophyly of P. lineatus sampled in distinct drainages, although this has been addressed by the multilocus phylogeny (Melo et al., 2016).

Prochilodus costatus and P. lineatus split from each other during the Early Pleistocene at approximately 1.1 MYA (Figure 3). Prochilodus lineatus occurs broadly across immense regions of the Paraná-Paraguay and Paraíba do Sul basins, whereas P. costatus is endemic to the Rio São Francisco (Castro & Vari, 2004). These two basins are separated by the Serra da Canastra mountain range that divides the Rio Grande in the south, the upper Rio São Francisco and Rio Abaeté in the north/northeast and the Rio Araguari of the Paranaíba basin in the northwest (Figure 3). This region has the geologic influence of the Upper Rio São Francisco crustal discontinuity (DCARSF) dated to the Paleoproterozoic in the southern portion of the São Francisco craton (Saadi et al., 2002). According to Saadi et al. (2002), the most recent movements are dated to the Quaternary (<1.6 MYA) with several earthquakes occurring in the southeast margin of DCARSF, an area with the largest concentrations of seismic events in southeastern Brazil. In addition, the geomorphology of the Upper Rio São Francisco is also controlled by the Estrela Fault dated to the Neoproterozoic (Saadi et al., 2002). The intense geologic reactivations of these two faults may have promoted biotic exchange between the Rio Grande and the upper São Francisco systems with subsequent reduction of gene flow leading to speciation.

This river capture hypothesis is supported by (i) patterns of species richness with estimates of 63 species shared between the Upper Rio Paraná and Rio São Francisco (Buckup, 2011), by the (ii) presence of several genetic lineages in common (e.g., Pereira et al., 2011; Rossini et al., 2016; Serrano et al., 2019), and by (iii) instances of sister relationships. For example, Ramirez et al. (2020) found an Early Pleistocene divergence (~2 MYA) of the anostomids Schizodon nasutus Kner, 1858 from the Rio Paraná and S. knerii (Steindachner, 1875) from the Rio São Francisco. Machado et al. (2018) detected an intraspecific divergence among populations of the bryconid Salminus hilarii Valenciennes, 1850 from both basins at approximately 1.9 MYA. Therefore, the combined evidence reinforces our hypothesis of Early Pleistocene river capture in the Serra da Canastra region as the most plausible explanation for the cladogenesis of P. costatus and P. lineatus.

Overall, this study provides robust evidence for the monophyly of both Prochilodus costatus and P. lineatus, and reveals that interspecific diversification in the genus occurred during the Early Pleistocene when tectonic reactivations, associated especially with the Rio Araçuaí Fault in the Serra do Espinhaço and the Upper Rio São Francisco crustal discontinuity in the Serra da Canastra, may have promoted river captures between the Rio São Francisco and the adjacent drainages of the Jequitinhonha and Upper Paraná basins. We acknowledge, however, that earlier tectonic events surely occurred in the eastern Brazilian Shield and may have been important for other diversification processes (e.g., P. vimboides), but the available mitogenome data suggest that recent Quaternary reactivations had a more important role in the interspecific diversification of Prochilodus. Future studies will have the opportunity to (i) integrate mitogenomes of other prochilodontid species, (ii) investigate biogeographic and geomorphologic events associated with the diversification of P. vimboides along coastal Atlantic rivers such as those in the Mucuri, Doce and Paraíba do Sul basins, and (iii) test geodispersal events including other Prochilodus species, especially members of the P. nigricans group as well as other prochilodontids (Ichthyoelephas and Semaprochilodus), and the putative association of species from lowlands of the Amazon basin with the species of the Brazilian and Guiana shields.