The genus Basilichthys ( Teleostei : Atherinopsidae ) revisited along its Chilean distribution range ( 21 o to 40 o S ) using variation in morphology and mtDNA

There is still doubt as to the number of species of the freshwater Chilean ichthyofauna, 64 % of which have conservation problems. One of the groups is that of the silversides of the genus Basilichthys. Three morphological species of this genus have been described in Chile with disjoint distributions: Basilichthys semotilus, B. microlepidotus and B. australis; the latter two overlap in distribution only in the Aconcagua River and are not easily distinguishable by morphological and meristic characters. In order to evaluate the effi cacy of identifi cation of these species by molecular techniques, we analyzed the sequence of 9 % of the mitochondrial DNA (Control Region and COI) of individuals from the Loa River (21o41’ S) to the Valdivia River (39o50’ S), adding meristic features for B. microlepidotus and B. australis in order to study population variation to clarify the taxonomy of the native species of the genus. The phylogenetic analysis showed that the individuals of Basilichthys semotilus form an haplogroup separated from the other species of the genus; however, B. australis and B. microlepidotus form a monophyletic group that shares the most common haplotypes. An analysis of meristic information showed no statistically signifi cant differences in the number of lateral line scales or number of rays in the fi ns between B. microlepidotus and B. australis. These results do not support the current classifi cation for the latter two species; there appears to be one group in the extreme north of the country (Basilichthys semotilus) and a second group in central Chile which should be called B. microlepidotus. This information will be useful to review the conservation status of the Chilean fauna.


INTRODUCTION
The freshwater fauna of Chile, because of its geographical isolation, is rather different and has a low diversity compared to the other American regions.In order to formulate effective conser vation plans for this fauna, it is necessar y to determine the real diversity of species present in specific places and their geographic distributions.In Chile, this has begun fi rst with foreign expeditions and more recently in conser vation programs by making lists of species with conser vation problems (Eigenmann 1927, Ministerio del Medio Ambiente 2011).However, there are still limitations imposed by species identifi cation, both from the presence of as yet undescribed species and in the assignment of species level to the geographically isolated populations.
Even though 64 % of the species of Chilean freshwater fi shes have been reported to have conservation problems (Habit et al. 2006, Vila et al. 2006), the low number of morphological dif ferences in characters among species has caused difficulties with this assessment (Campos 1982, Gajardo 1985, Arratia 1990, Dyer 2000b).The study of fi shes by Eigenmann (1927) contributed notably to their classifi cation and a reduction in the number of described species; however, debates about the validity of some of them have long existed and still exist (Dyer 2000a).
In terms of abundance and distribution, one of the most representative Chilean freshwater fi sh genera is Basilichthys (Girard, 1855).This genus belongs to the South American subfamily Atherinopsidae, characterized by having a nonprotractile mouth with the skin interrupted over the middle of the snout (Eigenmann 1927).Three allopatric species of this genus have been described: Basilichthys semotilus (Cope, 1874) in the Loa River (21º41' S), B. microlepidotus (Jenyns, 1841), found from the Huasco River (28º30' S) to the Aconcagua River (32º20' S) (Dyer 2000a) and B. australis (Eigenmann, 1928) from the Aconcagua River to Chiloé Island (42º18' S) (Campos et al. 1984).
In a complete phylogenetic revision of the Atherinopsinae, Dyer (1997) remarked that the species composition of Basilichthys has been problematic since its description.Dyer (1997) recognized two species groups: Semotilus, which includes the species from southern Perú and northern Chile B. semotilus, B. beardsleei (Abbott, 1899) andB. archaeus (Cope, 1878) and the Microlepidotus group, with two species: B. microlepidotus and B. australis.Within the Microlepidotus group, the species do not have clear diagnostic morphological characters that allow their identification.Eigenmann (1927) found that these species are scarcely distinguishable, dif fering only in the scale numbers of the lateral line.Later, Gajardo (1985), by using large sample sizes, showed that the number of scales overlaps; no other clear and signifi cant morphological or meristic dif ferences have been found between the species.Furthermore, molecular analysis did not fi nd differences in allele composition for 37 allozymic loci (Gajardo 1988).At present the only evidence that supports these species is a difference in the modal number of chromosomes (2n = 46 for B. microlepidotus and 2n = 48 for B. australis) (Gajardo 1992).Furthermore, both species have been indicated to coexist in the Aconcagua basin (Gajardo 1985).
In recent years new molecular techniques have been developed that are useful to complement the morphological determination of species and/or the presence of isolated populations of the same species.These kinds of techniques have proved successful in different groups of organisms (i.e.Hebert et al. 2003a, 2003b, Ward 2009), providing precision and repeatability (Hebert & Gregory 2005).In the case of freshwater fi shes, molecular markers have proved ver y useful because they tend to show clear phylogenetic patterns and their relation to the histor y of each watershed (Ber natchez & Wilson 1998, Bermingham & Martin 1998, Avise 2000).In recent years, sequencing of the cytochrome c oxidase I gene (COI) has been used in the taxonomy of freshwater fi sh; this gene has proved to be useful in identifying species in accordance with morphology (e.g., Ward et al. 2005, Hubert et al. 2008, Lara et al. 2010).
The objective of this study is to evaluate the taxonomy of the Basilichthys species in Chile.In order to determine the species diversity of this group, samples from ten geographically isolated rivers from 27º to 40º S were analyzed using two fragments of mitochondrial DNA (COI and Control Region), and the meristic counts used currently in Basilichthys taxonomy.With this information the taxonomic status of these species will be updated, providing additional information for better conservation plans.

Sampling
Specimens of Basilichthys were caught by netting, angling and electrofi shing from different rivers ranging from the Loa River (21º41' S; 69º35' W) to the Valdivia River (39º50' S; 72º44' W).Geographical locations and putative species names used in this study are listed in Table 1 and shown in Fig. 1.Voucher specimens were deposited in the Museo Nacional de Historia Natural (MNHN, Chile; Table 1).Humane standards were followed in the handling of the fi sh; all fi sh were euthanized using 100 mg L -1 tricaine methanosulfonate.

DNA extraction, amplifi cation and mtDNA sequencing
Total genomic DNA was extracted from ethanolpreser ved fin clips using the salt-extraction method (Aljanabi & Mar tínez 1997).Pure DNA was stored at -20 ºC in 50 μL of water until analysis.Using the mitochondrial sequence of Hypoatherina tsurugae (GenBank AP004420; Miya et al. 2003), silversidespecifi c primers for the control region were designed as follows: forward (5'-CCT AAC TCC CAA AGC TAG GAT-3') and reverse (5'-TGC GGT ACT TGC ATG TGT AA-3').Amplifi cation from the template DNA used the following conditions: 1x buffer, 3.2 nM MgCl 2 , 0.2 U μL -1 dNTP (Invitrogen), 5 pmol forward and reverse primers, and 0.1 U μL -1 Taq DNA polymerase (Invitrogen).The PCR reaction (in 25 μL fi nal volume) included a denaturing step of 94 ºC for 3 min followed by 30 cycles of 94 ºC for 30 sec, 60 ºC for 90 sec, and 72 ºC for 90 sec with a fi nal elongation step of 10 min at 72 ºC.The primers and PCR conditions for the COI amplifi cation followed Folmer et al. (1994).PCR products of both the COI and the Control Region were cleaned using QIAQuick columns (QIAGen, Mississauga, Ontario, Canada) and sequencing in both directions was performed in Macrogen Inc. (http://www.macrogen.com).Sequences were aligned using ProSeq software (Filatov 2002) and checked using Multalign online software (Corpet 1988).

Analysis
In order to determine genetic relationships among specimens sampled from the 10 dif ferent rivers we performed two analyses: (a) a neighbor joining-based relationship (NJ) performed in Mega 4.0 (Tamura et al. 2007) for haplotypes of both COI and the Control Region separately.(b) Using the same sequences, a Maximum Parsimony (MP) analysis run using PAUP

Meristic counts analysis
To evaluate possible meristic dif ferences among populations of B. microlepidotus and B. australis, samples from nine rivers (Fig. 1 and Table 2) were analyzed.For each specimen we counted the number of lateral line scales and rays of the fi ns.An ANCOVA analysis was performed for each measured character by using the GLM procedure of SAS (SAS Institute 1998) with the standard length as a covariate in all analysis.Normality and homocedasticity was tested by using the Shapiro and Bartlett tests, respectively.When the assumptions were violated, the Box Cox analysis was used to fi nd the best transformation to the data.Finally, when the ANCOVA was significant the LSMeans pair wise analysis was performed as the a posteriori test.The α = 0.01 was used to control a possible type I error.

RESULTS
For the COI gene, 687 bp were sequenced, obtaining 10 haplotypes in the 31 individuals analyzed.No inser tions or deletions were detected, so alignment was straightfor ward.There were 97 polymorphic sites of which 19 were parsimony informative characters.The MP COI analysis retained one most parsimonious tree with the following parameters: length = 20, consistency index (CI) = 1.00 and retention index (RI) = 1.00.
The NJ and MP analyses gave similar results, separating the haplogroup of B. semotilus from the other species of the genus.There was a mean difference of 19.5 bp (SD = 1.26) between the haplotypes of B. semotilus and the group with B. microlepidotus and B. australis.These analyses showed that B. australis and B. microlepidotus belong to a single haplogroup with bootstrap support > 99 %.Also, the latter two species shared haplotype 3 (H3), which was found in most of the studied watersheds (Fig. 2A and 2B, Table 3).
For the 731 bp of the control region sequenced we found 15 haplotypes in the 31 organisms analyzed.This sector also did not present alignment problems; only one insertion was present, in the individuals from the Valdivia River.145 sites were polymorphic and 138 were parsimony informative characters.The MP COI analysis retained 17 trees and the most parsimonious tree presented the following parameters: length = 164, consistency index (CI) = 0.91 and retention index (RI) = 0.93.
Both the NJ and MP analyses produced results similar to those found for the COI region.The analyses showed one haplogroup composed of B. semotilus, clearly separated from a second haplogroup composed of B. australis and B. microlepidotus by an average of 29.4 bp differences (SD = 2.23) (2.5 %).This region of mitochondrial DNA showed high statistical support (bootstrap support 100 %) for one haplogroup containing all the sequences of B. australis and B. microlepidotus.These two species shared 4 of the 13 haplotypes found in the haplogroup (haplotypes 3, 5, 6 and 7; Fig. 2C and 2D, Table 3).
The meristic analyses showed that the inclusion of the standard length in all ANCOVA analyses was statistically signifi cant (P < 0.001).The number of scales ranged from 50 (Choapa  Fish length, number of scales of the lateral line and fi n rays by river for Basilichthys group microlepidotus.Table includes average, standard deviation and minimum-maximum range for the character obtained for each site.Means sharing the same letter are not signifi cantly different (P < 0.01).N = number of fi sh analyzed.
The pectoral fi n rays ranged from 8 to 16, showing statistical dif ferences among sites (F 8,317 = 3.81; P =0.0003).Pair wise analyses revealed that the Choapa River had silversides with a mean ray number larger than that observed in the samples from Maipo and Maule (P < 0.001).All other paired comparisons did not show statistical differences.The ventral fi n rays ranged from fi ve to six for all sites studied.For this character, the ANCOVA analysis found differences among sites (F 8,317 = 3.95; P = 0.0002); specimens from the Aconcagua River were larger than those of all other sites (P < 0.01).The rays of the anal fi n ranged from 11 to 17 and the means were not different among sites (F 8,295 = 1.96;P = 0.051).The dorsal fi n showed between 8 and 12 rays and the ANCOVA analysis detected statistical differences among sites (F 8,316 = 3.29; P = 0.001).The pairwise analyses showed that samples from Maipo River had a lower mean number of rays than samples from Choapa (P = 0.004) and that the Aconcagua River sample had fewer rays than the Choapa River (P < 0.001), Rapel River (P = 0.001), Maule River (P < 0.001) and Valdivia River (P = 0.003).Rays in the caudal fi n ranged from 15 to 23.In the ANCOVA analysis this character showed differences in means among sites (F 8,317 = 3.29; P = 0.001).Pair wise analyses showed that Huasco and Biobío had fewer caudal fi n rays than Limarí (P = 0.003 and P = 0.001, respectively) and Valdivia (P = 0.003 and P = 0.001, respectively).Also, this analysis detected that fish from Limarí more rays than the sample from the Rapel River (P = 0.006).Overall, all meristical analysis per formed for specimens obtained from dif ferent drainage basins showed no statistical differences in their means between B. micrilepidotus and B. australis, as most fi sh usually have a wide range in their meristics characters (e.g., Parenti 1984).

DISCUSSION
Historically the recognition and determination of freshwater fi sh species has been especially dif ficult; few morphological characters have been used to discriminate between species.This has been recognized as a problem in the Trichomycteridae (e.g., Ar ratia 1990), Cheirodontidae (e.g., Campos 1982) and Atherinidae (e.g., Dyer 1997, 2000a).Our results in the genus Basilichthys indicate clear genetic differences between the semotilus and microlepidotus groups proposed by Dyer (1997).Specifically, within the microlepidotus group the meristic and genetic information indicated the presence of a monophyletic group without apparent morphological differences.
Since the work of Gajardo (1985Gajardo ( , 1988Gajardo ( , 1992) ) and recent discussion by Dyer (1997), classical morphological description has been controversial for B. microlepidotus and B. australis.Although organisms from the extremes of their geographic distribution show a few morphological differences, individuals from the central zone of the country have very similar characteristics where they also have been described living in sympatr y (Gajardo 1985, Dyer 2000b, 2006).A recent analysis indicated possible migration after the formation of four watersheds (Limarí, Choapa, Aconcagua and Maipo) in central Chile (Quezada-Romegialli et al. 2010), which also suggests that the two species previously described in this area may really be only one.
One of the most relevant characters described is the number of lateral line scales, whose range is from 74 to 96 (mean = 85.70) in B. microlepidotus and from 86 to 115 (mean = 99.46) in B. australis (Gajardo 1985).This author pointed out that most of the variance was within localities and he did not fi nd signifi cant dif ferences between the species.By using samples from nine rivers, the present study found similar results to those obtained for two rivers by Gajardo (1985), indicating that the differentiation of B. microlepidotus and B. australis does not have meristical support.
Further analysis using the gene products of 37 allozyme loci did not show differences between the individuals of B. microlepidotus and B. australis (Gajardo 1988).Probably the only supporting evidence that both are valid species is a kar yological analysis performed by Gajardo (1992).In this analysis, he found that B. microlepidotus from the Petorca River (32º20' S) had a modal number of 2n = 46 (NF = 62), while B. australis from the Angostura River (33º06' S) had a modal number of 2n = 48 (NF = 57) (Gajardo 1992).Although for some authors a difference in chromosome number is a good character to distinguish species, it may also be indicative only of chromosomal races.There are examples of chromosome races in various groups of Chilean organisms such as the catfi sh Trichomycterus areolatus (Valenciennes, 1840) which has 2n = 54, 55 and 56 in different rivers in the south of the country (Colihueque et al. 2006).These variations have also been reported in mice (e.g., Nachman & Searle 1995) and reptiles (e.g., Lamborot et al. 2003, Olmo 2005).In these terms, chromosome differences may not imply reproductive isolation in some vertebrata taxa.
All of the previously compiled information with respect to recognizing both species has not been conclusive.For this reason, the present study complements the meristic information usually utilized in the group with mtDNA sequences to present more evidence to help elucidate their taxonomy.
Mitochondrial DNA studies, especially those of the COI region, have proved to be a standardized method that also helps in species identifi cation in cases where taxonomic diversity is undetected (Hebert et al. 2004, Pegg et al. 2006).mtDNA analysis in fi sh biodiversity has been increasing in the last years (Ward et al. 2005, Hubert et al. 2008, Lara et al. 2010), showing that this technique concurs with previous morphological analyses.Ward & Holmes (2007) showed in a study performed in 338 fi sh species that this region gave conclusive results except for one pair of species that may hybridize (Ward & Holmes 2007).Ward (2009) demonstrated, using sequences of COI of 1088 fi sh species, that the probability that two identical sequence samples belong to the same species is 98-99 %.Indeed, in this case the presence of a haplotype (H3) in most of the studied rivers is evidence that these two species must be only one specifi c unit.
The variation of mitochondrial DNA in our analysis does not support the separation of B. microlepidotus and B. australis as dif ferent species.The statistical suppor t in the tree analyses and the presence of shared haplotypes for both the Control Region and COI indicate that there is one species with a wide distribution.Based on this evidence and according to the zoological nomenclature code, the name B. microlepidotus (Jenyns, 1841) has priority and should be used in the future.
On the other hand, it is interesting that B. semotilus, with nor thern distribution, is genetically different from the organisms found in the central area of Chile, with an average difference of 19.5 base pair differences in the COI region and 29.4 bp dif ferences in the Control Region compared to B. microlepidotus.B. semotilus is found in Peru from the Reque River, Lamberque (7º S) to the Sama River, Tacna (18º S), and in Chile it has been found in Codpa (19º S) and Loa River (22º S).Dyer's study (2000a) indicated that the large distance between the Sama River, Perú and the Loa River, Chile (about 1000 km) might indicate that the Chilean individuals belong to a different species and this is presently under study (Dyer 2009, personal communication).
The ichthyic fauna of Chile has a low richness (N = 44 species) compared to other South American regions (Arratia 1997), the principal characteristic being its high endemism (Arratia et al. 1981, Ruiz & Berra, 1994, Vila et al. 1999, Habit et al. 2006).Because of this high endemism, better knowledge of the real number of species, their geographic distributions and population sizes are presently necessar y in order to develop effective conser vation plans and a new conser vation classifi cation status.In the case of the silverside B. microlepidotus, individuals may currently be moving between watersheds using marine roots (Quezada-Romegialli et al. 2010), thus this species may be able to re-colonize watersheds naturally after ecological disasters.However, this process of natural re-colonization may be reduced by barriers to the fi sh dispersal.Due to the presence of dams which limit the natural requirements of reproduction, feeding and migration of fl uvial species, restoration plans for populations will have to include planned and directed translocation of individuals.Other aspects such as habitat deterioration and the introduction of invasive species such as salmonids in the country's freshwater systems have also had an important negative impact on native fi shes (Habit et al. 2005, Pardo et al. 2009).Overall, our research shows that conser vation plans must be performed case by case.While B. microlepidotus could be translocated from dif ferent central Chilean drainages to allow the survival of the species under a perturbation event, B. semotilus has low alternatives of translocation and therefore requires a specifi c plan of protection.
The new molecular tools, together with morphological characters, will help to bring up to date taxonomic information of species to unify criteria between the investigators who work in conservation.These methods that are not invasive (if samples are obtained from fi n pieces), complemented with morphological characters will be the key to identify adults, larval stages and eggs, as well as forensic work looking for habitat and its fauna conservation.

Fig. 2 :
Fig. 2: Trees constructed with COI and Control Region sequences for species of the genus Basilichthys.(A) Neighbor joining analysis and (B) Maximum Parsimony for COI sequences.(C) Neighbor joining analysis and (D) Maximum Parsimony for Control Region sequences.In all cases, numbers above branches are bootstrap values (only values > 50 % are shown).See Table 2 for location of each haplotype.Árbol realizado con secuencias del gen COI y Región Control para las especies del género Basilichthys.(A) análisis Neighbor joining y (B) Maximum Parsimony realizado con secuencias del gen COI.(C) análisis Neighbor joining y (D) Maximum Parsimony realizado con secuencias de la Región Control.En todos los casos, el número sobre los brazos representa el valor del boostrap (se muestran solo valores sobre 50 %).Ver Tabla 2 para la ubicación geográfi ca de cada haplotipo.

Fig. 3 :
Fig. 3: Distribution of scale number by specimen in the lateral line.Each graph represents a sampled River.

TABLE 1
Geographic locations, putative species and sample size for genetic analyses.Sitios de estudio, nombre putativo de las especies y número de muestras para el análisis genético.Swofford 2002).For both NJ and MP, the consistency of branches was tested using a bootstrap re-sampling with 1000 replicates.All sequences used in this analysis were published in Genbank with the following accession numbers: Control Region: FJ380091 to FJ380105 and COI: FJ380197 to FJ380116.Sequences of Odontesthes regia (Humboldt 1809) sampled from Iquique (20º42' S; 70º11' W) were used as outgroup for these analyses (Genbank Accession Number: COI: FJ380117, Control Region: FJ380106).

TABLE 3
Name of each haplotype found in each River.