Genetic characterization of native and introduced populations of the neotropical cichlid genus Cichla in Brazil

A molecular phylogenetic analysis based on mitochondrial 16S ribosomal DNA and Control Region sequences from native and introduced populations was undertaken, in order to characterize the introduction of Cichla (peacock bass or tucunaré) species in Brazil. Mitochondrial DNA haplotypes found in introduced fish from Minas Gerais state (southeastern Brazil) clustered only with those from native species of the Tocantins River (Cichla piquiti and C. kelberi), thereby suggesting a single or, at most, few translocation acts in this area, even though with fish from the same source-population. Our study contributes to an understanding of the introduction of Cichla in regions of Brazil outside the Amazon basin, and adds phylogenetic data to the recently describe Cichla species, endemic from the Tocantins-Araguaia basin.


Introduction
The introduction of non-native species is considered to be the second greatest threat to native biodiversity, after habitat loss (UNEP 2005). Pimentel et al. (2005) estimate that non-native or indigenous species represent a cost of approximately $120 billion per year in damages and control for the USA alone. In addition to the economic impact, invasive species imply severe negative consequences for native fish species in southern Brazil. One particularly important example of this is the Amazonian cichlid genus Cichla, the species of which are collectively known in Brazil as tucunaré. Introduced tucunaré appear to have been responsible for the local extinction of the characoid fish species Metynnis cf. roosevelti at sites in northeastern Brazil (Molina et al., 1996) and approximately half the native fish species of a natural lake in the Doce River valley in southeastern Brazil (Godinho et al., 1994), where it was introduced together with the red piranha (Pygocentrus nattereri). Kullander and Ferreira (2006) revised the taxonomy of the genus based on morphological characteristics, thereby identifying nine new species. These include the two from the Tocantins basin, namely C. kelberi and C. piquiti, which these authors also considered to be the very species introduced into the river basins of Minas Gerais state.
Cichla has frequently been introduced into river basins outside their native range, both in Brazil and other countries, whereupon they have often become invasive, being implicated in the local extinction of at least nine fish species (Zaret and Paine, 1973). Little is known, however, on the origins of these populations, their taxonomic identity or the number of individuals introduced in the first place. While Cichla were apparently introduced into northeastern Brazil in the late 1940s by government agencies for the purpose of installing fish breeding farms (Fontenele, 1948;Fontenele and Peixoto, 1979), no information is available as regards their river or rivers of origin. Subsequently, the fish also became well established in southeastern Brazil (Alves et al., 2007), although it is not known whether these populations were established by animals that escaped from aquaculture farms or were deliberately released. As Cichla is a highly prolific fish, which adapts well to lentic conditions, its populations have expanded rapidly in hydroelectric reservoirs, floodplains and lagoons (Godinho et al., 1994;Pompeu and Godinho 2003).
Invasive populations of Cichla were first registered in southeastern Brazil during the 1980s (Agostinho et al., 1994), and since then, several impacts on indigenous fish populations have been reported (Pompeu and Godinho 2003). In the hydroelectric reservoirs of Furnas and Marimbondo, in the upper Paraná basin, the tucunaré, together with a second Amazonian species, the croaker (Plagioscion squamosissimus), have become the predominant members of the fish community (Santos et al., 1994), and are already an important resource for local fisheries, being widely appreciated both for food and as a sport fish (Torloni, 1993).
Determining the invasion of introduced species is an important step in understanding and controlling the spread of invasive species (Chandler et al., 2008). For example, by identifying the source of an invasive population, the most important transport vectors for introducing novel species into new areas, and which therefore need reinforced regulation, can be determined (Carlton, 2001). Specifying the invasion pathway can also lead to a deeper understanding of the criteria and mechanisms required for successful invasion (Vermeij, 1996).
The present study aims at specifying the genetic origin of the invasive tucunaré populations in four major Minas Gerais river basins by means of the molecular phylogenetic analysis of partial mtDNA regions (16s and Control Region -CR), besides providing additional molecular information on those native species not analyzed by Willis et al. (2007). The genetic data generated herein were compared with those from previous studies on Cichla from other South American sites (Oliveira et al., 2006;Renno et al., 2006;Willis et al., 2007), in an attempt to elucidate the history of Cichla invasions.

Samples
Tissue samples (fins) were collected from indigenous (Amazon basin, n = 41) and non-indigenous (non-Amazonian rivers, n = 103) specimens of tucunaré and preserved in 90% ethanol. Vouchers were preserved in a 30% formaldehyde solution and then transferred into an 80% ethanol solution. In the Amazon basin, fishes were collected from the Tucuruí reservoir (site 6 - Figure 1) on the Tocantins River -the eastern limit of the Cichla range -and from the Solimões River in the western Amazon basin. Non-indigenous specimens were collected from four different river systems in southeastern Brazil (Table 1, Figure 1) and one site in northeastern Brazil. DNA sequences from GenBank (Table 1) were included in the analyses, in order to compare our data with those available for other regions and species.

Taxonomic identification
All specimens were identified, according to Kullander and Ferreira (2006). Both C. kelberi and C. piquiti were collected from the Tucuruí and Itumbiara reservoirs (Figure 1: sites 3 and 6), whereas only C. kelberi was collected at sites 1 and 4, and C. piquiti was the only species found at sites 2 and 5. Specimens collected in Catu lake (site 7) could not be reliably identified based on Kullander and Ferreira (2006).

Genetic procedures and data analysis
Tissue samples were digested using proteinase K, and DNA isolated by phenol/chloroform purification (Sambrook et al., 1989). Two mitochondrial markers were amplified, namely 16S rDNA (16S) and Control Region (CR), as both had previously been used in studies on population genetics of Cichla (Oliveira et al., 2006;Willis et al., 2007). A portion of 16S was amplified with the primers 16S-L1987 (5'-GCC TCG CCT GTT TAC CAA AAA C-3') and 16S-H2909 (5'-CCG GTC TGA ACT CAG ATC ACG T -3'; Palumbi et al., 1991), and a portion of the CR was amplified with the primers L (5'-AGAGC GTCGGTCTTGTAAACC-3', Cronin et al., 1993, and H16498 (5'-CCT GAA GTA GGA ACC AGA TG-3', Meyer et al., 1990). These primers amplify PCR products that are approximately 500 bp (16S) and 460 bp (CR) in length. After the PCR reaction (16S: 94°C for 3 min, 30 cycles at 94°C for 1 min, 55°C for 1 min and 72°C for 3 min, followed by an extension of 72°C for 7 min; CR: 94°C for 4 min, 50°C for 30 s, 72°C for 2 min, 30 cycles at 94°C for 602 Carvalho et al. 15 s, 56°C for 30 s and 72°C for 2 min, with an extension of 72°C for 10 min), 1-2 mL of the PCR product were used in the DNA sequence reaction, according to manufacturer's instructions (Big Dye Terminator Mix). The reactions were analyzed in an automated DNA sequencer (ABI 377 or ABI 310, Perkin Elmer) and the DNA sequenced in both directions. The nucleotide sequences generated in this paper were deposited in GenBank under the accession numbers DQ779579-DQ779586, FJ904286-FJ904291 for 16S rDNA and Control Region FJ890798-FJ890816.
The chromatograms were checked by manual inspection and aligned using default parameters in the ClustalW software included in the Bioedit program (Hall, 1999).
Phylogenetic trees were generated in a Bayesian framework MrBayes (Huelsenbeck et al., 2001(Huelsenbeck et al., , 2002. Confidence for each node was verified by posterior probabilities for MrBayes. A PAUP 4* version 4.0b10 for Unix was also used for parsimony analysis (Swofford, 2003). The 16S and CR data sets were analyzed separately in order to incorporate as much information as possible from previ-Phylogenetics of introduced Cichla 603  (Guindon and Gascuel, 2003). The evolutionary models used for rRNA 16S and Control Region were HKY (Hasegawa et al., 1985) and GTR (Rodriguez et al., 1990), respectively, both combined with the assumption that a proportion of sites are invariable (I), and that rate variation is modeled by gamma distribution (Yang, 1994). The posterior distribution of trees was approximated by using the Markov chain Monte Carlo as implemented in MrBayes. We ran four independent runs with four parallel chains, where each run started from a randomly chosen tree with 4 million generations each, and was sampled every 1000th generation. The first 1 million generations were discarded in each run as being a burn-in stage. Trees and parameter values from the four runs were pooled and a majority-rule consensus tree was calculated. The major haplotypes identified by Oliveira et al. (2006), Renno et al. (2006) and Willis et al. (2007), were used to determine the genetic origin and species of each introduced population (non-indigenous), as well as the taxonomic distinctness of the Tocantins population, as suggested by Kullander and Ferreira (2006). Geophagus brasiliensis was used as outgroup in both analyses.

Results
After trimming ambiguous ends, sequences of 410 bp and 359 bp were obtained, respectively, for 16S and the Control Region. The CR was more variable than 16S, with 157 variable sites, of which 102 were informative for parsimony analysis, as opposed to 50 variable sites with 17 informative for 16S. As topology and branch confidences in maximum parsimony were very similar to Bayesian analysis, only the later is shown in the present work.
Despite fewer specimens analyzed in relation to the Control Region, the Bayesian cladograms of the two markers were, in general, congruent.

16S analysis
The indigenous species from the Tocantins River formed two distinct clusters, C. piquiti, and C. kelberi significantly supported by Bayesian posterior probabilities (PP) of 0,95 and 0,99, respectively. The tucunarés from a marginal lake of the São Francisco and the Itumbiara and Furnas reservoirs (Figure 1; sites: 2, 3 and 5) are all included within the C. piquiti clade.
On the other hand, the data also indicated that tucunarés from the Itumbiara, Elvécio and Três Marias reservoirs (Figure 1: sites: 1, 3 and 4) were highly associated (PP = 0,92) to C. kelberi from the Tocantins River. In addition, the C. kelberi group was strongly connected to nonindigenous Cichla sp. from Catu lake, as well as to other unresolved lineages, such as C. monoculus from the Negro and Solimoes Rivers, C. ocellaris and C. orinocensis (PP = 0.92). Conversely, C. temensis from the Negro River was not associated to any other group, appearing as an independent indigenous lineage (Figure 2).

Control region analysis
In the control region analysis, the indigenous species from the Tocantins River also proved to be separated into two distinct clusters, C. piquiti and C. kelberi, supported by high posterior probabilities of 1.0 and 0.98, respectively. Furthermore, specimens from the C. piquiti clade were grouped together with tucunarés from the Itumbiara, Furnas, Marginal and Lajeado reservoirs (PP = 1.0). Interestingly, sustained by high posterior probability (PP = 0.97), C. temensis from the Negro River, which was unresolved by 16S analysis, appeared as a sister group of the C. piquiti group. Control Region data also suggested a close relationship among C. intermedia from the Orinoco, Cichla sp. from the Mamoré, and a group formed by Cichla sp from the Amazonas and Xingu (two unidentified native specimens), as well as the Cichla sp. farm species. This latter species (Cichla sp Farm), which is firmly connected (PP = 1.0) to native tucunarés from the Xingu and Amazonas, was denominated C. temensis by Oliveira et al. (2006). The unidentified native specimen from the Mamoré River is a basal specimen in this group. Cichla ocellaris and C. orinocensis appeared as independent unresolved lineages. On the other hand the C. kelberi specimens were all strongly grouped (PP = 0.98) with several non-indigenous tucunarés, such as those from Itumbiara, Paraná, Três Marias (Morada Nova and Felixlândia) and the Dom Elvécio lake. The data also showed a very well supported subgroup (PP = 0.94), this including a nonindigenous specimen of C. kelberi from the Tocantins, grouped together with introduced specimens from Três Marias (Felixlândia and Morada Nova; Figure 1: site 1), the Doce River (Dom Elvecio lake) (Figure 1: site 4) and 604 Carvalho et al.  Table 1) were also used for comparison. Numbers above branches represent Bayesian posterior probabilities. Stars means non-indigenous specimens.
Itumbiara (Figure 1: site 3). Only one haplotype was detected in all the introduction-sites, with the exception of Itumbiara, with 3 distinct CR haplotypes for C. kelberi (Figure 3). The unidentified specimens from Ceará (Catu Lake) clustered with C. monoculus haplotypes from the Madeira and Amazonas Rivers, with high posterior probabilities (Figure 3: PP = 0.96).

Discussion
The phylogenetic results of this study confirm that the introduced populations in the river basins of southeastern Brazil (Minas Gerais) all originated from the Tocantins River. In addition, we confirmed the genetic distinctness of the two Cichla species from the Tocantins (C. piquiti and C. kelberi), from those of the Amazon basin. Fontenele (1948) and Fontenele and Peixoto (1979) reported that government agencies had translocated tucunarés to northeastern Brazil in order to improve the frail fish fauna in this region, this process then being repeated in southeastern Brazil. Our genetic data contradict this statement, given that the haplotypes of the introduced population from Ceará (Catu lake; northeastern Brazil) clustered with samples from the Amazon and Madeira Rivers, thus indicating the taxonomic identification of this species as C. monoculus, and the occurrence of at least two independent translocations. Additional samples from Ceará and other northeastern sites are required to confirm this difference.
The fact that the Itumbiara reservoir was the only locality in Minas Gerais where two species of Cichla occurred in sympatry (Table 1), as well as the only site where three different CR haplotypes for C. kelberi were detected (Figure 3), suggests that this was the target of more intensive translocations, and thus may represent the place of origin of the stocks of other localities in Minas Gerais.
Unexpectedly, different species, C. kelberi and C. piquiti, were recorded at sites 1 and 2, respectively, despite being located in the same river basin (São Francisco), thereby suggesting that Cichla is not widely dispersed in this region, possibly through initially introducing only a few specimens at each site, which would also account for only one species being recorded in most sites in Minas Gerais.
The success of introduced species has been generally attributed to their capacity of adapting to novel environments (Stockwell et al., 2003;Alcaraz et al., 2005). For example, the invasive freshwater fish Lepomis gibbosus has managed to adapt to different habitat-zones in reservoirs (Bhagat et al., 2006). Likewise, its accentuated environmental plasticity and aggressive behavior may have enabled Cichla to successfully colonize new habitats throughout Brazil, in particular where habitats have been modified by dams. For instance, following the extinction of small native teleosts, the diet of tucunarés in Dom Elvécio Lake (site 4) shifted primarily to freshwater prawns (personal observation by the authors). Reflecting this, local anglers now prefer to use prawns as bait, rather than the usual techniques used elsewhere. Extirpation of the introduced stocks of tucunaré is probably impossible in most cases, but understanding the dynamics of these invasions might be helpful in developing management strategies for these and other non-indigenous species.
With the exception of a review by Kullander and Ferreira (2006), the introduced Cichla in Minas Gerais are still being misidentified as C. temensis, C. ocellaris and C. monoculus (e.g., Alves et al., 2007). None of these species were, however, detected at any of the four sites from Minas Gerais analyzed here. C. temensis from the Negro river seems to be a sister species of tucunaré from the Tocantins Phylogenetics of introduced Cichla 605 Figure 3 -Phylogenetic tree based on control region (CR) sequences in Cichla spp, and on haplotypes from the Tocantins River and several sites in southeastern Brazil, as well as Catu lake in northeastern Brazil. Selected haplotypes from the Amazon drainage (GenBank -see Table 1) were also used for comparison. Numbers above branches represent Bayesian posterior probabilities. Stars means non-indigenous specimens.
river. However, the Tocantins population represents an isolated stock, which reinforces the finding that this river basin is inhabited by two of the new species (C. piquiti and C. kelberi), as proposed by Kullander and Ferreira (2006). In the present study, the analysis of two mitochondrial markers revealed that the two Cichla species endemic to the Tocantins River (Kullander and Ferreira, 2006) have been translocated extensively to sites in southeastern Brazil. Our study also provides additional phylogenetic information, since Willis et al. (2007) did not analyze C. piquiti or C. kelberi, and also elucidates certain aspects of the introduction of this highly invasive species.