Molecular characterization of Moenkhausia ( Pisces : Characiformes ) populations with different lateral line developmental levels

The genera Hemigrammus and Moenkhausia have been traditionally diagnosed mainly by the former having lateral line completely pored whereas the latter having a lateral line with a few pored scales. Those features have been used to diagnose species of both genera in the upper Paraná River floodplain. Specimens with the diagnostic features of Moenkhausia bonita, collected in the upper Paraná River floodplain, exhibited different developmental levels of the lateral line, making it difficult to distinguish them from specimens of Hemigrammus sp. We analyzed the gene encoding cytochrome C oxidase I (COI) and intron 1 of the nuclear gene S7 to investigate the genetic similarities between the called Hemigrammus marginatus and M. bonita and to confirm their identities. Molecular sequences of other Moenkhausia species were analyzed for genus delimitation tests. The results reveal genetic similarities of M. bonita specimens with different developmental levels of the lateral line, and also distinguish between M. bonita and Hemigrammus sp. Species delimitation tests revealed that specimens from the upper Paraná River floodplain were M. bonita and were distinct from other Moenkhausia species. The developmental level of the lateral line is not a consistent characteristic that distinguishes between Moenkhausia and Hemigrammus species.


INTRODUCTION
are among the genera of Characiformes that are not monophyletic (Mirande 2010, Oliveira et al. 2011, Mariguela et al. 2013) and feature a highly problematic taxonomy (Lima et al. 2003).Moenkhausia comprises 81 valid species of small fish widely distributed in the hydrographic basins of South America and Hemigrammus comprises 64 valid species with a wide geographic distribution (Eschmeyer and Fong 2017).There are 35 Hemigrammus species in Brazil, of which 17 are endemic (Lima et al. 2003, Bertaco andCarvalho 2005).Moenkhausia and Hemigrammus share features such as five teeth in the inner series of the premaxilla, and small scales covering the base of the caudal fin.The limits suggested by Eigenmann (1907) for distinguishing these two genera consider the lateral line, completely pored in Moenkhausia and incomplete in Hemigrammus.
Specimens attributed to Hemigrammus marginatus Ellis, 1911 have been captured in the Paraná River basin (Britto et al. 2003, Lima et al. 2003, Graça and Pavanelli 2007); however, there are several taxonomic problems regarding its genus status ( de Brito Portela-Castro andJúlio Júnior 2002, Maniglia et al. 2012).Cytogenetic divergences were registered between populations of H. marginatus from the upper Paraná River floodplain and individuals from the São Francisco River basin, suggesting that they may be distinct species ( de Brito Portela-Castro andJúlio Júnior 2002, Maniglia et al. 2012).Maniglia et al. (2012) evidenced by molecular data the existence of two species of Hemigrammus in the upper Paraná River basin, and that both species were not H. marginatus.Maniglia et al. (2012) also suggested that further studies using molecular tools should be undertaken to determine the distribution of the Hemigrammus species in the upper Paraná River basin.Ota et al. (2015) have restricted H. marginatus to rivers in northeastern Brazil, but that name was used for several years to identify specimens with incomplete lateral line in the upper Paraná River basin.Due to this statement we are herein treating specimens with complete lateral line as Hemigrammus sp.
Moenkhausia bonita Benine, Castro & Sabino, 2004 has been described from the Baía Bonita, Paraguay River basin, in Bonito, Mato Grosso do Sul State, Southwestern Brazil.Benine et al. (2004) stated that Moenkhausia bonita is similar to Hemigrammus marginatus (herein Hemigrammus sp.), but can be distinguished from the latter by having complete lateral line and a lozenge-shaped spot on the caudal peduncle.
Although the geographical distribution of M. bonita has been restricted to the Mato Grosso do Sul State (Froese and Pauly 2017), specimens with its diagnostic features have been captured in the floodplain of the upper Paraná River.However, these specimens exhibit different developmental levels of the lateral line that include complete, discontinuous, and incomplete lateral lines, which is a diagnostic feature of Hemigrammus sp.However, as the specimens have the other diagnostic feature, a lozenge-shaped spot on the caudal peduncle, even the specimens with a lateral line not completely pored have been tentatively identified as M. bonita.The upper Paraná River is an area with specific complex history and evident endemism, at least for the fish groups, because of the migratory barrier formed until recently by the Sete Quedas Falls, which isolated a great part of the icthyofauna of the Upper Paraná River Basin and the ichthyofaunas of the middle and lower stretches of the Paraná-Paraguay river basin (Britski and Langeani 1988, Langeani et al. 2007), but many species of fish are naturally shared between these basins (Froelich et al. 2017).
Once the correct taxonomic identification is an essential factor for the quantification of species richness, biodiversity patterns, composition, and structuring of communities (Gotelli 2004), the ribosomal protein S7 gene intron 1 (rpS7) were amplified using polymerase chain reaction (PCR) and primers, as described by Ward et al. (2005) and Chow and Hazama (1998), respectively.PCR mixtures (25 μl) containing Tris-KCl (20 mM of Tris-HCl, pH 8.4, and 50 mM of KCl), 1.5 mM of MgCl2, 2.5 mM of each primer, 0.1 mM of each dNTP, 1 μl Taq of DNA polymerase, and 15 ng de mold DNA and water.Temperatures for the amplification were as follows: initial cycle at 94°C, 2 min; 40 cycles at 94°C, 30 s; 50°C-58°C (according to the primer), 40 s; and 72°C, 2 min, followed by a final extension for 5 min at 72°C.All extraction and amplification products were quantified by comparison with known quantities of lambda DNA in agarose gel (0.8%) stained with ethidium bromide.Amplicons were purified according to Rosenthal et al. (1993) and these were subjected to BigDye Terminator Cycle sequencing and DNA sequencing was performed using a MegaBACE DNA Analysis System (Amersham), both according to manufacturer's instructions.

PHYLOGENETIC ANALYSES
Individual sequences of each gene were edited using BioEdit (Hall 1999) and were later aligned to MEGA 6 (Tamura et al. 2013) using the algorithm Clustal W (Thompson et al. 1994).Exclusion of identical sequences and identification of haplotypes were performed using ElimDupes (https://hcv.lanl.gov/content/sequence/HCV/ToolsOutline.html).MEGA 6 and DNAsp (Librado and Rozas 2009) were employed to characterize the COI locus.The best nucleotide substitution model was generated using jModelTest 2 (Darriba et al. 2012) based on the Bayesian Information Criterion.Partial sequences (522 bp) of the mitochondrial gene COI were obtained from 46 specimens: M. bonita with complete (n = 16), incomplete (n = 7), or discontinuous (n = 10) lateral lines; and M. gracilima (n=3) and M. intermedia (n=3) employment of molecular techniques is required.Nuclear and mitochondrial markers are highly promising because they reveal the complex evolutionary process of Neotropical fish (Mariguela et al. 2013), taxonomic distinction of the species (Benine et al. 2009) and evaluation of the genetic variability of fish (Oliveira et al. 2002, Prioli et al. 2002, Panarari-Antunes et al. 2012).
Current assays employ molecular tools to analyze the genetic similarities between populations from the type-locality of M. bonita and those with several developmental levels of the lateral line, which inhabit the upper Paraná River floodplain, in order to achieve conclusive identification of specimens attributed to M. bonita and Hemigrammus sp.Further, current analysis comprises molecular sequences of other species of the genus Moenkhausia for species delimitation tests.

BIOLOGICAL MATERIAL
Specimens previously identified as M. bonita (n=33), M. gracilima Eigenmann, 1908 (n=3) and M. intermedia Eigenmann, 1908 (n=3) were collected in the upper Paraná River basin.Specimens of M. bonita (n=7) were collected in the Baía Bonita, Paraguay River basin in Bonito, Mato Grosso do Sul State, the type-locality of the species.The specimens were killed by overdosing of clove oil, according to Griffiths (2000).Specimens were collected under a scientific collecting permit (Sisbio 52596-1, process nº 61/405817/2015), were placed in flasks with ethyl alcohol and stored at −20°C.Sample sites are shown in Figure 1.
Total DNA was extracted from muscle tissue samples using the Wizard Genomic DNA Purification Kit (Promega Corporation, USA), according to the manufacturer's instructions.Partial sequences of the mitochondrial gene encoding cytochrome C oxidase I (COI) and  River basin, wherefore, we are herein treating as Hemigrammus sp.Information on all sequences analyzed and their respective haplotypes are presented in Table SI and Table SII.An ultrametric tree was constructed using sequences of the COI and a Bayesian inference statistic method using BEAST software (Drummond et al. 2012), accounting for birth-death speciation model and relaxed molecular clock.Analyses were separately replicated at least twice, and the results were analyzed using Tracer 1.6 with an effective sample size of >200.
Three phylogenetic trees were constructed using the maximum likelihood method RaxmLGUI (Silvestro and Michalak 2012) using sequences of COI and rpS7, with a rapid bootstrap algorithm with 1,000 resamplings, using the GTR +G model.A median-joining type haplotype network (Forster and Ro 1994) was generated using PopArt (http:// popart.otago.ac.nz) from the COI sequences of samples from the floodplain of the upper Paraná River.The trees were subsequently edited using FigTree.The GMYC test employs the ultrametric test and delimitation is achieved by comparing the distances of the samples from the most recent common ancestor and speciation event, taking into account a time scale (Fujisawa and Barraclough 2013).An ultrametric tree was built using BEAST to analyze GMYC.The HKY + G model was selected using jModelTest 2 and employed two partitions of the codon bases (first + second position; third position).The GMYC test was conducted in R using a splits package (Ezard et al. 2015), with a single threshold.
The PTP test employs a phylogenetic tree and considers the number of substitutions according to the length of branches, where the probability of substitutions that lead to speciation follows the Poisson distribution (Zhang et al. 2013).A phylogenetic tree of maximum likelihood, which was used for the PTP test, was constructed from the GTR + G model, taking into account the subpartitions of codon positions (first + second; third position).The test was performed using the bPTP Webserver (http://species.h-its.org/ptp).

Moenkhausia bonita
In M. bonita, ten haplotypes were identified within the COI gene of samples from the upper Paraná River floodplain and three haplotypes of samples from the type-locality.Specimens from the typelocality and the upper Paraná River floodplain do not share haplotypes.Sequences of each haplotype are identified in Table SI.
Using the Kimura-2-parameter with 1,000 bootstrap resamplings, the genetic distance of COI sequences among the M. bonita samples from the type-locality and from the upper Paraná River floodplain was 1.25%.Intraspecies variations of the two groups were 0.05% for the samples from the Paraná River floodplain and 0.03% for samples of the type-locality.Genetic distances among all haplotypes of Moenkhausia are presented in Table SIII.The genetic distance between M. bonita and THAÍS F.M. MOTA et al.
Different mitochondrial haplotypes in the upper Paraná River floodplain were not associated with different developmental levels of the lateral line (complete, discontinuous, and incomplete).For example, the haplotype BI_9 was shared among specimens with three types of lateral lines, which is also shown by a haplotype network of the medianjoining type (Figure 2).
Phylogenetic trees of maximum likelihood demonstrated that the specimens of M. bonita with different developmental levels of the lateral line and M. bonita from the type-locality were in the same clade (Figure 3).Only two variable nucleotide positions and three haplotypes were detected in the 17 partial sequences of rpS7 from samples from the upper Paraná River floodplain.BC 13 was the most frequently detected haplotype (n=14).Further, the haplotype represents specimens with all types of lateral lines.Intraspecies variation of rpS7 among the haplotypes from the upper Paraná River floodplain was 0.0032%, and there was a 5.79% variation between M. gracilima and M. bonita.A Bayesian tree of rpS7 was generated and shows an introgression of the nuclear haplotype among the different mitochondrial haplotypes independent of the type of lateral line (Figure 3).

DELIMITATION OF OTUs
Different species of Moenkhausia (n = 107) were employed to analyze the delimitation of OTUs (46 and 61 from the present study and BOLD, respectively).The GMYC test employed to identify OTUs, contributed significant results (Table I), considering samples of M. bonita collected in the upper Paraná River floodplain and samples collected in the type-locality as a single species, and separating them the other species of Moenkhausia, with a posteriori probability superior 95%.However, sequences of Moenkhausia acquired from BOLD, which were initially identified as the same species, were grouped into different clusters (Figure 4).
The potentially identical species were indicated by the PTP test with an interval of 10-16 species.The OTUs indicated by the PTP test were the same as indicated by the GMYC test, the specimens of M. bonita collected in the upper Paraná River floodplain and samples collected in the typelocality were grouped in the same OTUs.Data that support delimitation are presented in Figure S1 (Supplementary Material).included 10 haplotypes that were not shared by the population of Hemigrammus sp. with only two haplotype.The Consortium for the Barcode of Life suggests that the rates of interspecies genetic distance should be 10 times greater than those of intraspecies genetic distance, so that species can be considered different (Hebert et al. 2004).The mean rates of interspecies genetic distance determined here were greater than those of intraspecies genetic distance (0.05% in M. bonita and 3.23% between M. bonita and Hemigrammuus sp.).Similar results were reported using a DNA barcode to analyze fish (Bellafronte et al. 2013, Pereira et al. 2013).
Although we have found that M. bonita from the upper Paraná River and Hemigrammus sp.represent different species, the percentage of nucleotide differences between them is low and reveals genetic similarity.Specimes of Hemigrammus sp., were long attributed to H. marginatus (Ota et al. 2015).In the original description of M. bonita morphological similarity with H. marginatus is evidenced (Benine et al. 2004).A phylogenetic analysis of the characteristics of the sperm of Moenkhausia and Hemigrammus species among others led Santana et al. (2012) to propose the hypothesis that M. bonita and H. marginatus represent sister taxa.The same hypothesis was proposed by Mariguela et al. (2013), in a phylogenetic analysis of osteologic characters and external morphology as well as an analysis of molecular data.
Phylogenetic analyses of COI sequences distinguishing M. bonita and Hemigrammus sp., reveals genetic similarities between M. bonita from the upper Paraná River and M. bonita specimens from its type-locality.All these specimens are grouped within the same cluster despite their intraspecies nucleotide differences.It is expected that there are high rates of intraspecies divergence among geographically separated populations (Hebert et al. 2003).April and Mayden (2011) found intraspecies variation in freshwater fish that was higher than that in other taxonomic animal groups; this may be explained by low gene flow caused by species fragmentation in water environments (Ward et al. 1994), mainly among small-sized species (April andMayden 2011). Pereira et al. (2013) found high intraspecies variation among geographically separated small-sized species.
The gene nuclear rpS7 was efficient for assigning specimens from the upper Paraná River floodplain with different developmental levels of the lateral line into the same group that shares the same haplotype, indicating that all specimens with different developmental levels of the lateral line belong to the same species.Further, the phylogenetic tree of rpS7 reveals that there is introgression of the nuclear haplotype among different mitochondrial haplotypes, regardless of the specimen's type of lateral line.Benine et al. (2009) and Marinho (2010) criticize the validity of the developmental levels of the lateral line as a feature for distinguishing Moenkhausia and Hemigrammus.Characters such as the lateral line that defines Hemigrammus and Moenkhausia may have evolved several times in different strains.Further analysis within the phylogenetic context is required to solve its true importance for delimiting Characidae groups (Marinho 2010).Moenkhausia sanctaefilomenae includes specimens with variations in the Molecular delimitation tests of OTUs were useful for distinguishing Moenkhausia species that inhabit different Brazilian hydrographic basins.However, we show here that certain sequences of Moenkhausia acquired from BOLD, which were initially identified as the same species, belong to different species; for example, M. sanctaefilomenae was represented in different clusters.In such cases, a taxonomic revision of the groups should be undertaken.According to Benine et al. (2009), M. oligolepis may represents a species complex, and further studies should be undertaken to solve the issue.
In this study, we evidenced that specimens collected in the upper Paraná River floodplain with different development levels of the lateral line belongs to M. bonita.According to the molecular analyses, this trait is not sufficient for distinguishing the species of Moenkhausia and Hemigrammus.
collected in the upper Paraná River floodplain and; M. bonita with complete lateral line (n=7) from its type-locality.We determined the partial sequences of rpS7 (420 bp) of 19 specimens: M. bonita with complete (n = 8), incomplete (n = 4), or discontinuous (n = 5) lateral lines from upper Paraná River floodplain and M. gracilima (n=2) used as outgroup.Developmental levels of the lateral line of specimens of M. bonita are presented inTable SI and Table SII (Supplementary Material).Sequences of Hemigrammus available from BOLD were added inserted to the analyses.H. marginatus (n=4) collected in São Francisco River basin and H. marginatus (n=6) collected in the upper Paraná

DISCUSSION
Our analyses of the COI gene were efficient in discriminating specimens of M. bonita, which were sampled in the upper Paraná River floodplain, with different levels of development of the lateral line and specimens of Hemigrammus sp. from the upper Paraná River basin, indicating that they are representatives of different species.The population of M. bonita from the upper Paraná River floodplain

Figure 2 -
Figure 2 -Median-joining haplotype network from samples of Moenkhausia bonita from the upper Paraná River floodplain with different developmental levels of the lateral line, as well as samples from the Baía Bonita, Mato Grosso do Sul State, with complete lateral lines.

Figure 3 -
Figure 3 -Mirrored phylogenetic trees of mitochondrial (a) and nuclear (b) haplotypes constructed from specimens of Moenkhausia populations with different developmental levels of the lateral line.Hemigrammus marginatus (HM) and Hemigrammus sp.(H.sp. 1) were used as outgroup.The circles show the bootstrap values > 70.Numbers between brackets indicate the number of specimens with each haplotype.Dotted lines show the presence of a specific nuclear haplotype of a specific mitochondrial haplotype and vice versa.Sequences of intron 1 of the nuclear gene S7 were not obtained for samples without connection.

Figure 4 -
Figure 4 -Bayesian ultrametric tree of the COI gene and delimitation of OTUs of Moenkhausia, according to the GMYC test (specimens in red represent same cluster inferred by GMYC).Circles represent a posteriori probability > 95%.Only the sequences of the COI of M. bonita specimens were obtained in this study, the others were obtained from Genbank (TableSI).Green: Moenkhausia bonita from the upper Paraná river.Grey: Moenkhausia bonita from the type-locality.

Table SII -
Genetic distances of the COI gene among all Moenkhausia haplotypes.

Table SIII -
Genetic distances of the COI gene among all Moenkhausia haplotypes.Figure S1 -Phylogenetic tree of the COI gene and delimitation of OTUs of Moenkhausia, according to the PTP test (specimens in red represent same cluster inferred by PTP).