Mitochondrial evidence for a new evolutionary significant unit within the Gila eremica lineage ( Teleostei , Cyprinidae ) in Sonora , Northwest Mexico

espanolPresentamos las afinidades filogeneticas y el codigo de barras del ADN de Gila cf. eremica, una poblacion morfologicamente divergente y geograficamente aislada de G. eremica DeMarais, 1991. Los analisis filogeneticos mitocondriales de cyt–b, cox1 y nd2 muestran la existencia de un patron de clados dentro del linaje de G. eremica, que situa a G. cf. eremica en un clado de identidad especifica que comparte un supuesto ancestro comun con G. eremica, originario de la cuenca del rio Matape. El analisis de codigo de barras, en que se utilizo un metodo basado en caracteres del programa informatico CAOS, mostro siete caracteres puros que diferencian a G. eremica de su congenere regional G. purpurea y un caracter fijo en G. cf. eremica que lo diferencia de G. eremica. Estos resultados y la reciente deteccion de diferencias morfologicas diagnosticas entre G. cf. eremica y G. eremica sostienen la hipotesis de que Gila cf. eremica es una unidad evolutivamente significativa dentro del linaje de G. eremica. EnglishWe present the phylogenetic affinities and DNA barcode of Gila cf. eremica, a geographically isolated and morphologically divergent population from G. eremica DeMarais, 1991. Mitochondrial phylogenetic analyses of cyt–b, cox1 and nd2 show a clades pattern within the G. eremica lineage, placing G. cf. eremica in a clade of specific identity to and sharing a putative common ancestor with G. eremica from the Matape River basin. The barcoding analysis using a character–based approach of CAOS showed seven single pure characters discriminating G. eremica from its regional congener G. purpurea, and one fixed character in G. cf. eremica discriminating it from G. eremica. These results and the recent detection of diagnostic morphological differences between G. cf. eremica and G. eremica support the hypothesis of Gila cf. eremica as an significant evolutionary unit within the G. eremica lineage.


Introduction
The genus Gila, one of the most widespread groups of the family Cyprinidae in North America, includes morphologically heterogeneous fishes inhabiting waters of arid and semiarid regions of the western United States (USA) and northwestern Mexico (Miller et al., 2005).Using both mitochondrial and nuclear markers it has been observed that recent phylogenetic analyses of cyprinids in North Americaalign Gila with ten other nominal genera within a so-called Revised Western Clade (RWC) (Schönhuth et al., 2012), and suggest that Gila comprises an evolutionary lineage involving at least 18 species, including species currently recognized taxonomically in the monotypic genera Acrocheilus and Moapa (Schönhuth et al., 2014).Nevertheless, Schönhuth and colleagues termed he composition of the Gila lineage incomplete because of phylogenetic affinities of G. coerulea and Ptychocheilus lucius that were only resolved within Gila by using mitochondrial marker cyt-b rather than the concatenated nuclear genes rag1, rhod, and s7 or concatenated mitochondrial and nuclear markers (Schönhuth et al., 2012(Schönhuth et al., , 2014)).
Molecular and morphological analyses of Gila species occurring in northern México and western USA suggest that nominal species of the G. robusta complex in the Colorado River basin show both allopatric and sympatric distributions, with probable hybrid origins at least in part (DeMarais et al., 1992;Dowling and DeMarais, 1993;Gerber et al., 2001;Schönhuth et al., 2014;Dowling et al., 2015;Page et al., 2017).Gila species in Mexican waters of the Atlantic slope Chihuahuan Desert region, and those in the Pacific slope, however, apparently show mainly allopatric distributions associated with major river drainages, suggesting peripatric speciation events (Wiley, 1981;Schönhuth et al., 2014).
Phylogenetic affinities and current distributions of all known species of Gila show that those in Mexico include an Atlantic-slope lineage referred to as the Chihuahuan Desert Group, which includes G. pulchra, G. conspersa, G. nigrescens, G. brevicauda, an undescribed species, and a lineage composed of G. modesta nested within G. pandora (Schönhuth et al., 2014).Their phylogenetic affinities, morphological characteristics, and geographical distributions suggest that species in this group share a single common ancestor, unrelated to any species belonging to the G. robusta complex of the Colorado River system (Uyeno, 1960;Schönhuth et al., 2014).The remaining nominal species of Gila in México (G. ditaenia, G. minacae, G. purpurea, G. eremica) occur in Pacific-slope drainages.These four species were not resolved in analyses by Schönhuth et al. (2014) as part of their Chihuahuan Desert Group.However, they were resolved as monophyletic within the greater Gila lineage, and G. eremica and G. purpurea were corroborated as sister species (Schönhuth et al., 2012(Schönhuth et al., , 2014)), as proposed by previous morphological analysis (DeMarais, 1991).
The study of the evolution of Gila is important because it is an opportunity to understand the evolution of freshwater fish because it relates to the geological history of western North America.Evaluations of Gila species in USA and México suggest a current-day lack of understanding regarding the diversity of the genus (Schönhuth et al., 2014).This is substantiated by paraphyletic groupings obtained by Schönhuth et al. (2012Schönhuth et al. ( , 2014) ) and recent records of undescribed populations of the genus in several drainages of central-north and northwest Mexico (DeMarais, 1991;Varela-Romero, 2001;Norris et al., 2003;Minckley and Marsh, 2009;Bogan et al., 2014;Schönhuth et al., 2014).
The rapid development of molecular taxonomic and systematic methods in recent years has provided several tools to study biodiversity.Nowadays, in addition to molecular phylogenetic methods, the DNA barcoding technique (Hebert et al., 2003a(Hebert et al., , 2003b) ) has been applied as a molecular taxonomic tool to support species identifications and species discoveries (Hebert et al., 2003a(Hebert et al., , 2003b;;Witt et al., 2006;Hubert et al., 2008;Rach et al., 2008;Lara et al., 2010;Li et al., 2011;Zou and Li, 2016;Yi et al., 2017).Current analytical methods to assign DNA barcodes to taxa can be divided into distance-based, phylogeny-based, and character-based approaches (Hebert et al., 2003a;Pons et al., 2006;Sarkar et al., 2008;Puillandre et al., 2012;Taylor and Harris, 2012).Although genetic distance-based methods for DNA barcoding have been considered useful tools in species discrimination and cryptic species discovery (Ward et al., 2005;Hubert et al., 2008;Lara et al., 2010;April et al., 2011;Lakra et al., 2015, Zou andLi, 2016), this approach has been questioned because of the relatively high rates of evolution of mitochondrial DNA between and within species (and between different groups of species) that can result in overlaps of intra-and interspecific distances, thus suggesting an uncertain existence of a barcoding gap for all species (Kipling and Rubinoff, 2004;Rubinoff, 2006;Rubinoff et al., 2006).
Recent records of unstudied populations of Gila in northwestern México include two populations inhabiting a series of large spring-fed pools (pozas) of the Arroyo El Tigre sub-basin (Varela-Romero, 2001;Bogan et al., 2014).This sub-basin pertains to the Mátape River basin to the east and includes the adjacent low-elevation subtropical canyons La Balandrona and La Pirinola, both located in the southeastern sector of the Sierra El Aguaje coastal mountain range, near the towns of San Carlos and Guaymas (fig. 1,inset).The proximate geographical location of these newly discovered populations suggest they are part of the Gila eremica lineage (Varela-Romero, 2001;Bogan et al., 2014) and thus are referred to herein as Gila cf.eremica.However, both populations are geographically isolated from populations of the lineage of G. eremica inhabiting the Mátape and Sonora River basins to the east and northeast.This isolation was probably promoted by volcanic events occurring in the area during the Miocene (Mora-Álvarez and McDowell, 2000), causing a geographical disconnection of the Arroyo El Tigre sub-basin from the Mátape River sub-basin.Recent morphological evaluations of the Gila eremica lineage revealed the G. cf.eremica populations as distinct compared to all G. eremica and other selected congeners analyzed, and showed at least 16 morpho-linear and two meristic characters that distinguish G. cf.eremica from other G. eremica populations (Ballesteros-Córdova et al., 2016).These mensural and meristic differences detected in the G. cf.eremica populations, as well as their isolated geographic occurrence, suggest a potential evolutionary isolation event within the G. eremica lineage (Ballesteros-Córdova et al., 2016), similar to that proposed for G. eremica and G. purpurea in Sonora (DeMarais, 1991;Schönhuth et al., 2014).The proposal of a Gila eremica lineage comprised of populations from the Sonora and Mátape River basins, plus the G. cf.eremica from the isolated Arroyo El Tigre sub-basin in La Balandrona and La Pirinola canyons, calls for development of molecular analyses to further investigate the evolutionary affinities of G. cf.eremica within the entire Gila lineage, and to potentially detect character attributes that may discriminate it from related groups.Knowledge of the evolutionary history of Gila can contribute to elucidating speciation mechanisms involved in this taxonomically problematic genus and other related fishes from arid and semiarid regions in North America.Also, the potential recognition of an evolutionary significant unit within Gila in México would enable the development of management strategies for its conservation.

Sample collection and DNA extraction
A total of 178 specimens of five species of the genus Gila occurring in four river basins (seven sub-basins) Mátape River basin

DNA extraction and PCR amplification
Total DNA was obtained from fin tissue of all collected specimens following protocols of an extraction kit, the QIAamp DNA Mini Kit (QIAGEN).Amplification reactions were performed in a total volume of 50 µl using GoTaq Colorless Master Mix (Promega).
The mitochondrial gene cyt-b was totally amplified (1140 bp) using the primers FW-L15058 5'-TGA CTT GAA AAM CCA CCG TTG-3' and RV: H16249 5'-TCA GTC TCC GGT TTA CAA GAC-3' as reported by Kocher et al. (1989).The total sequence (1047 bp) of mitochondrial gene nd2 was amplified with primers ND2F: 5'-AAC CCA TRC YCA AGA GAT CA-3' and ND2R: 5'-ACT TCT RCT TAR AGC TTT GAA GG-3', designed for other sequences of the genus as reported in GenBank.Conditions for the amplification of cyt-b and nd2 comprised an initial denaturation for 5 min at 94 ºC followed by 35 cycles of 50 s at 94 ºC, 50 s annealing temperature at 50 ºC for cyt-b and 60 ºC for nd2, and then 2 min at 72 ºC.The final extension was performed at 72 ºC for 7 min.A region of 651 bp of the mitochondrial gene cox1 was amplified with the primers FishF2_t1: 5'-TCT ACA AAY CAC AAA GAC ATT GGT AC-3' and FishR2_t1: 5'-ACC TCT GGG TGR CCA AAG AAT CAG AA-3', modified of Ivanova et al. (2007) to make them more specific to Gila.Conditions for amplification of cox1 comprised an initial denaturation for 5 min at 94 ºC followed by 34 cycles of 50 s at 94 ºC, 50 s at 50 ºC, and 1 min at 72 ºC.The final extension was performed at 72 ºC for 7 min.The PCR products were sent to Macrogen, Inc., Seoul, South Korea for purification and bidirectional sequencing according to the company's specifications.

Phylogenetic inferences
Obtained sequences were edited and assembled by overlapping using Chromas Pro 1.6 (Technelysium Pty Ltd, South Brisbane, Queensland, Australia).Each gene was identified using BLAST searches (Altschul et al., 1990) against GenBank data.Sequence divergence for the members of the G. eremica lineage, including G. cf.eremica and G. purpurea, was analysed using MEGA v5 (Tamura et al., 2011).Phylogenetic relationships of Gila cf.eremica from both La Balandrona and La Pirinola canyons were first evaluated via mitochondrial gene cyt-b, with sequences used by Schönhuth et al. (2014) for their phylogenetic inferences of the genus, including other members of their Revised Western Clade, plus our sequences of the G. eremica lineage.
The evolutionary relationships of the two G. cf.eremica populations were corroborated using concatenated sequences of the mitochondrial genes cyt-b, nd2 and cox1 of all specimens obtained in this study plus sequences of congeners available from GenBank.Phylogenetic trees using Maximum Likelihood (ML) for cyt-b and the concatenated mitochondrial gene dataset (cyt-b, nd2 and cox1) were estimated using RAxML-HPC2 on XSEDE 8.0.24 (Stamatakis, 2014).The JModeltest2 software (Darriba et al., 2012) was used to find the best nucleotide substitution models for each dataset separately.We defined data blocks based on genes, and used the Akaike information Criterion (Posada and Buckley, 2004), and the best-fit model was used for the subsequent analyses.ML trees were performed on the CIPRES Science Gateway 3.3 (Miller et al., 2010), using GTRGAMMA model, and 1,000 bootstraps pseudoreplications (Felsenstein, 1985) to estimate the node reliability.Bayesian inference (BI) analyses were conducted for each gene data set using MrBayes v3.2.6 (Ronquist and Huelsenbeck, 2003).The Akaike information criterion (AIC) implemented in JModelTest2 (Posada and Buckley, 2004;Darriba et al., 2012) was used to identify the optimal molecular evolutionary model for each partition block on each sequence data set of the analysis.For BI, ten million cycles were implemented in four simultaneous Monte Carlo Markov chains; sampling the Markov chain at intervals of 1,000 generations.Log-likelihood stability was attained after 100,000 generations; the first 1,000 trees were discarded as burn-in in each analysis.The remaining trees were used to compute a 25 % majority rule consensus tree in PAUP* (Swoford, 2002).Support for BI tree nodes was determined based on values of Bayesian posterior probabilities.Final trees of ML and BI were edited using FigTree 1.4.2(Rambaut, 2014).
DNA barcode and cox1 sequence analysis DNA barcode analyses included samples of nominal Gila eremica populations from the Sonora and Mátape River basins, samples of G. cf.eremica from the Arroyo El Tigre sub-basin's La Balandrona and La Pirinola canyons, G. purpurea from Arroyo San Bernardino, and G. ditaenia as outgroup.Cox1 sequences were edited and assembled by overlapping using Chromas Pro v1.6 (Technelysium Pty Ltd, South Brisbane, Queensland, Australia).The values of haplotype (H), nucleotide diversity (π) (Nei, 1987), and the polymorphic/variable sites were estimated with DnaSP v5.0 (Librado and Rozas, 2009).Genetic uncorrected p distance analysis between groups was performed using MEGA v5 (Tamura et al., 2011).A neighbor-joining (NJ) 50 % majority-rule consensus tree was constructed in PAUP* (Swoford, 2002) over 1,000,000 bootstrap replicates, using K2P (Kimura, 1980).The obtained tree was incorporated into the NEXUS file of the DNA data matrix of Gila species using Mesquite v3.10 (Maddison and Maddison, 2016), according to the specifications of Jörger and Schrödl (2014).The NEXUS/TREE file was carried out in CAOS software package (Sarkar et al., 2008) to identify diagnostic single pure characters (sPu) which are present in all members of a clade but absent from all members of another clade.These sPu characters were used for taxa discrimination of our groups of interest.

Phylogenetic relationships within the Gila eremica lineage
Amplified sequences of the mitochondrial gene cyt-b for all specimens of the G. eremica lineage analyzed had a length of 1,140 bp with no insertions or deletions.Thirty-four positions were variable sites, 15 of which were parsimony informative.Individuals of the G. eremica lineage using mitochondrial gene cyt-b were classified into 19 haplotypes.Haplotypes 8 and 10 were shared between individuals from the Sonora and San Miguel Rivers sub-basins (Sonora River drainage), and none of the remaining haplotypes were shared with any other population.Sequence divergence (uncorrected p distance) within the G. eremica lineage ranged between 0.26 % and 1.11 % with an average of 0.7 %.Genetic divergence of G. eremica populations from the Sonora, San Miguel and Mátape Rivers subbasins compared to G. cf.eremica from La Balandrona and La Pirinola canyons (Arroyo El Tigre sub-basin) was 0.44-1.14%.The mean distance between all individuals of G. eremica and G. cf.eremica was 0.83 %.The genetic divergence for individuals of G. eremica (but not including G. cf.eremica) against G. purpurea was 1.93-2.72%.Genetic divergence between G. cf.eremica and G. purpurea ranged from 2.11-2.46%.Mean nucleotide frequencies within nominal G. eremica populations were 26.30 % A, 29.57% T, 27.48 % C, and 16.65 % G. Mean nucleotide frequencies for G. cf.eremica populations were 26.58 % A, 29.59 % T, 27.56 % C, 16.37 % G.The estimated Transition/Transversion bias (R) for both groups was 2.27.Substitution pattern and rates were estimated under the General Time Reversible model (GTR) (Nei and Kumar, 2000).
Phylogenetic analyses of cyt-b using ML and BI with all species analyzed of the Revised Western Clade of Schönhuth et al. (2014), and including our samples of G. eremica and G. cf.eremica, showed the same topology with variations in nodal support of bootstrap probabilities (BP) for ML and posterior probabilities (PP) for BI (fig.2).The tree topology was consistent with that obtained by Schönhuth et al. (2014) for members of the Gila lineage (excluding G. cf.eremica).Members of the G. eremica lineage, including G. cf.eremica of the present study, were always resolved as monophyletic, with G. purpurea as the sister species (BP = 98 %, PP = 100 %) (fig.2).The G. eremica lineage was resolved with a clade for the Sonora River sub-basins (Sonora and San Miguel Rivers), and a sister clade for the Mátape River sub-basins (Mátape River and Arroyo El Tigre) (fig.2).Individuals of G. cf.eremica from La Balandrona and La Pirinola canyons of the Arroyo El Tigre sub-basin were nested together sharing a putative common ancestor, corroborating these two isolated populations as unequivocal members of the G. eremica lineage (fig.2).
Phylogenetic analyses by ML and BI, using the concatenated results of mitochondrial genes cyt-b, nd2, and cox1, included Gila robusta, G. ditaenia, G. purpurea, members of the G. eremica lineage (including G. cf.eremica), with G. minacae as outgroup.The tree topology resulting from the analyses was the same for both criteria, with variations in the nodal support values of BP for ML and PP for BI (fig.3).The analyses showed that members of the G. eremica lineage are monophyletic (BP = 100 %, PP = 100 %) with G. purpurea as sister species (BP = 100 %, PP = 100 %), and corresponded with our results from cyt-b in regards to monophyly within the G. eremica lineage (figs.2, 3).The geographical clade for all members of the Mátape River basin, including those from the two isolated Arroyo El Tigre canyons, was better supported in the concatenated genes' analyses (fig.3, BP = 76 %, PP = 95 %) compared to using cyt-b alone (fig.2).Individuals of G. cf.eremica from both canyons in the Arroyo El Tigre sub-basin were supported in a clade of specific identity (fig.3, BP = 94 %, PP = 100 %) and indicating relationship with G. eremica from the Mátape River sub-basin as putative closest relative (fig.3).

DNA barcoding analysis of the Gila eremica lineage
The analyses of 82 cox1 sequences of several Gila species showed a total of nine haplotypes: five for G. eremica, one for G. cf.eremica, one for G. purpurea, and two for G. ditaenia.The G. eremica lineage, including G. cf.eremica, showed a haplotype and nucleotide diversity of 0.6093 (SD = 0.034) and 0.00125 (SD = 0.00112), respectively.Individuals of G. eremica from the Sonora, San Miguel and Mátape Rivers sub-basins shared at least one haplotype, and also showed unique haplotypes for the Sonora and San Miguel rivers sub-basins.The analysis did not detect shared haplotypes among G. cf.eremica and all the G. eremica populations.The sequences of taxa included in the DNA barcoding analyses showed 46 polymorphic sites (table 2).The genetic uncorrected p distances analysis between groups produced a value of 1.27 % for populations of G. eremica and G. purpurea.The distance value between G. cf.eremica and G. purpurea was 1.38 %, and the value between nominal G. eremica and G. cf.eremica was 0.20 %.
The DNA barcoding analysis using the characterbased approach with CAOS showed eight single pure characters (sPu) to discriminate G. eremica from G. purpurea, nine sPu to discriminate G. cf.eremica from G. purpurea, and one fixed sPu in the 29 analyzed sequences of G. cf.eremica, discriminating it from G. eremica (table 2).

Discussion
The monophyly of the G. eremica lineage (Sonora and Mátape River basins) and G. purpurea (Bavispe River in northern headwaters of the Yaqui River basin) obtained in the present study was highly supported by both ML and BI criteria, as previously suggested by morphological analyses by DeMarais (1991) and molecular data by Schönhuth et al. (2014).Phylogenetic relationships inferred here for both ML and BI using the cyt-b gene alone and the concatenated set of genes cyt-b, nd2 and cox1, support with high values of posterior probabilities, the monophyly of the G. eremica lineage for all its members, and corroborated G. cf.eremica as a member of the lineage (figs.2, 3).The monophyly, geographical clades, and low genetic divergence detected here within the G. eremica lineage may be explained by relatively recent isolation of once-connected drainages inhabited by this lineage, as suggested for other nominal species of Gila occurring in México (Schönhuth et al., 2014).In addition, we provide evidence for the existence of two geographical clades for the Sonora and Mátape river basins, with high scores for both ML and BI criteria (figs. 2,3).
The close relationship and low genetic divergence between G. eremica from the Mátape River and G. cf.eremica from Arroyo El Tigre sub-basin (figs.2, 3) suggest a putative common ancestor for these populations and indicate a relatively recent connection between the two sub-basins.The phylogenetic analysis also supports G. cf.eremica as a clade of specific identity, apart from other members of the lineage, but closely related to populations of G. eremica in the Mátape River sub-basin.The morphological differences detected between G. cf.eremica and G. eremica (Ballesteros-Córdova et al., 2016), along with its phylogenetic position resolved in the present study, suggest that G. cf.eremica is an evolutionary significant unit within the G. eremica lineage that requires additional species delimitation methods such as DNA barcoding for further discrimination.
The effectiveness of character-based methods (e. g., CAOS) for taxa detection relies on the use of diagnostic characters, as those used in traditional taxonomy.The character-based method is based on the premise that members of a given taxon share a distinctive combination or combinations of diagnostic character attributes (e. g., polymorphisms) that are absent in related groups, and these attributes can be used for species discrimination (Rach et al., 2008;Sarkar et al., 2008;Bergmann et al., 2009;Zou et al., 2011;Jörger and Schrödl, 2013;Zou and Li, 2016).Despite the proposal to use more than three characters as a DNA barcoding gap to separate natural taxonomic groupings (Rach et al., 2008, Yassin et al., 2010;Zou et al., 2011;Yu et al., 2014), Jörger and Schrödl (2013) argue that CAOS does not Fig. 2. Recovered tree of phylogenetic relationships via Maximum Likelihood (GTR + G + I model) for haplotypes of all members of the Gila lineage, including other species of the Revised Western Clade (Schönhuth et al., 2012(Schönhuth et al., , 2014)), using mitochondrial gene cyt-b.Numbers on branches represent ML bootstap probabilities (BP > 65 %)/BI posterior probabilities (PP > 70 %): SS, Sonora River sub-basin; SMS, San Miguel River sub-basin; MS, Mátape River sub-basin; LB, La Balandrona Canyon; LP, La Pirinola Canyon (* specimens from Arroyo El Tigre sub-basin).
Fig. 2. Árbol recuperado de relaciones filogenéticas mediante el método de la máxima verosimilitud (Modelo GTR + G + I) de los haplotipos de todos los miembros del linaje de Gila, incluidas otras especies del Revised Western Clade (Schönhuth et al., 2012(Schönhuth et al., , 2014))   possess an objective criterion with which to delimit a threshold number of distinguishing nucleotides that would indicate a species boundary (i.e., to delimit a probable new species) or an independent population belonging to the same species.On the other hand, the supposition that a characteristic attribute has been fixed within a population gains more confidence if a higher number of samples is analyzed (Rach et al., 2008).According to Zhang et al. (2010) 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 2 2 2 2 3   0 2 2 3 3 4 4 8 8 2 4 6 7 7 7 9 9 3 6 7 2   4 0 3 4 7 0 6 2 5 4 5 6 0 5 8 0 3 2 2 4 G. eremica obtained in our several molecular analyses bolsters the morphological differences previously detected (Ballesteros-Córdova et al., 2016).Such differences may reflect a general phenotypic plasticity of freshwater fishes to adapt to environmental alterations, or variation in stream size, flow and substrate (Hubbs, 1940), leading in our case to a morphological variant within the Gila eremica lineage.However, the morphological differences seen in the G. eremica lineage members (Ballesteros-Córdova et al., 2016) are consistent with the establishment of a fixed polymorphism in the geographically isolated G. cf.eremica.Moreover, none of our phylogenetic analyses nested individuals of G. cf.eremica with samples of G. eremica, thus revealing this potentially ESU as a natural group.Similarly, the character-based method using cox1, and sequences analyses of cyt-b and its concatenation with nd2 and cox1 showed a different pattern of variation in G. cf.eremica compared with G. eremica, and with G. purpurea.However, current results will need to be further tested using nuclear data.
The process of species identification through DNA barcoding has often been confused with species discovery (DeSalle, 2006).Species identification has been considered a valid use for the DNA barcode, which does not rely on any particular species concept (Rach et al., 2008).This appears to be because species identification using DNA barcode is consistent with any concept of species a taxonomist may recognition of an isolated taxon of Gila in the Arroyo El Tigre sub-basin of the Sierra El Aguaje reveals a potential microendemism for the genus in subtropical canyons of this region of Northwest Mexico.The evidence presented here calls for further studies aimed at clarifying the biology, origin and history of G. cf.eremica populations and contributes to increased understanding of the evolution and conservation of fish species inhabiting arid and semiarid regions in Mexico and the USA.
The phylogenetic data and DNA barcoding results obtained here, coupled with those from the morphological analyses for G. cf.eremica (Ballesteros-Córdova et al., 2016) and its geographic isolation supports it as a natural evolutionary significant unit within G. eremica.The low genetic distance detected between G. cf.eremica and nominal G. eremica, along with their phylogenetic affinities indicates a relatively recent disruption within this lineage.Our data thus contribute to the knowledge of systematics and evolution of the greater Gila lineage.The

Fig. 1 .
Fig. 1.Collection sites for Gila specimens analyzed.Open circles represent towns, and numbers in solid black circles are locations detailed in table 1. Hydrographic drainage divides are indicated by thick lines.Dashed lines indicate contemporary intermittent drainage courses.Dotted lines are state boundaries.

Fig. 3 .
Fig. 3. Recovered phylogenetic tree via maximum likelihood and posterior probabilities of Bayesian inference of the concatenated genes cyt-b, cox1, and nd2 for populations of the Gila eremica lineage plus other three species examined in this study (SS, Sonora River sub-basin; SMS, San Miguel River sub-basin; MS, Mátape River sub-basin; LB, La Balandrona Canyon; LP, La Pirinola Canyon.Numbers on branches represent values of ML bootstap probabilities (BP > 65 %)/BI posterior probabilities (PP > 70 %).(* specimens from Arroyo El Tigre sub-basin).
3 C C G G A T C G T T A G G C A G T A C T A A A C C 15 C C G G A T C G T T A G G C A G T A C T A A A C C 14 -T ---------A -------------27 -

Table 2 .
Rach et al. (2008)le size for a DNA barcoding analysis should range from 9.5 to 216.6, which is within the range of samples of G. cf.eremica and G. eremica analyzed here.Our analysis revealed a single fixed polymorphism in all 29 examined sequences of G. cf.eremica that was absent from all analyzed specimens of the G. eremica populations.The unique polymorphism detected here in G. cf.eremica with respect to G. eremica represents a sPu character and thus reveals an apomorphy within the G. eremica lineage.This indicates a different pattern of genetic variation for G. cf.eremica compared with G. eremica and also with the close congener G. purpurea.The study byRach et al. (2008)identified less than three diagnostic characters in ten very closed related sister taxa of the insect order Odonata using mitochondrial gene ndh1, showing similar results to those obtained here between populations of G. cf.eremica and G. eremica.The close relationship and low genetic divergence of the two G. cf.eremica populations with respect to Polymorphic sites and composite of attribute characters obtained by CAOS in the first 651 bp of the mitochondrial gene cox1 for G. cf.eremica compared to nominal G. eremica populations, G. purpurea and G. ditaenia.Numbers at the top indicate variable sites of the fragment studied in this study.The far right column shows the number of individuals sharing each haplotype.Pure diagnostic characters among G. cf.eremica, G. eremica populations and G. purpurea are in bold.Single pure characters between G. cf.eremica and G. eremica are shaded and in bold: LB, La Balandrona Canyon; LP, La Pirinola Canyon; S, Sonora River; SM, San Miguel River; M, Mátape River.