Parasite phylogeny

The partial LSU rDNA sequences included 20 Cichlidogyrus species, 3 Scutogyrus species and 7 remaining species as outgroup (Enterogyrus coronatus, E. sp. 1, E. sp. 2, Onchobdella aframae, O. bopeleti, Protogyrodactylus alienus and P. hainanensis). LSU sequences of Cichlidogyrus and Scutogyrus species varied from 637 bp (C. ergensi) to 844 bp (C. arthracanthus). No variability was observed among the individuals of C. tilapiae found in four different host species, i.e. H. fasciatus, O. niloticus, S. galilaeus and T. guineensis, and between Cichlidogyrus sp. 2 from S. galilaeus and T. guineensis. However, some nucleotide variability (i.e. p-distance corresponding to 0.037) was found between the individuals of C. halli 1 found in S. galilaeus and T. guineensis, and C. halli 2 found in O. niloticus. An unambiguous alignment spanned 533 positions. Information on the LSU rDNA alignment, as well the model selected by ModelTest and its parameters, are shown in Table 1. The MP analysis provided 56 equally parsimonious trees with 669 steps (CI = 0.661, RI = 0.757). All phylogenetic analyses yielded a similar tree topology (Figure 1).

Based on the analyses of LSU rDNA sequences, Cichlidogyrus with Scutogyrus species (i.e. Cichlidogyrus/Scutogyrus group) parasitizing Cichlidae formed a strongly supported monophyletic group, with Cichlidogyrus pouyaudi in basal position relative to other Cichlidogyrus and Scutogyrus species, which was strongly supported by bootstrap values (ME and MP analyses) and moderately supported by Bayesian posterior probabilities.

The second data set included the sequences composed of partial SSU rDNA and the entire ITS1 region obtained for Cichlidogyrus and Scutogyrus species. Partial SSU rDNA sequences varied from 423 bp (S. bailloni, C. yanni) to 483 bp (C. ergensi, C. falcifer, C. tilapiae, S. longicornis), and ITS1 sequences ranged between 330 bp (S. bailloni, S. minus) and 498 bp (C. pouyaudi). The partition homogeneity test implemented in PAUP*4b10 was used to test the congruence of the phylogenetic signal in partial SSU rDNA and ITS1 alignment. No significant difference was found (p = 0.1), sequences were combined, and the concatenated alignment spanned 542 positions (see Table 1 for details). Phylogenetic analyses based on partial SSU rDNA and ITS1 sequences included 26 Cichlidogyrus species and 3 Scutogyrus species. Cichlidogyrus pouyaudi was selected for rooting the tree following the results of phylogenetic analyses based on partial LSU rDNA sequences. Information about the data sets used in the analyses of combined SSU rDNA and ITS1 sequences and the model selected by ModelTest are included in Table 1. The MP analysis provided 16 equally parsimonious trees with 367 steps (CI = 0.586, RI = 0.674). ML, MP, ME analyses yielded congruent trees (Figure 2).

All phylogenetic analyses supported the monophyly of Scutogyrus, in spite of the slight topological differences between trees; this brings support to the validity of this genus. Overall, six clades of gill monogeneans were recognized using phylogenetic reconstructions based on combined SSU rDNA and ITS1 sequences. Clade 1, strongly supported by all phylogenetic analyses, included three Cichlidogyrus species parasitizing Hemichromis species i.e. C. longicirrus, Cichlidogyrus sp.1 and C. falcifer. Clade 2, with weak or moderate support values resulting from different phylogenetic analyses, included 4 Cichlidogyrus species parasitizing three different cichlid fish species. Scutogyrus minus, S. longicornis and S. bailloni formed a monophyletic group (clade 3). The position of Cichlidogyrus tilapiae parasitizing four different cichlid species (see Table 2) was unresolved. Clade 4, with strong support values, included three Cichlidogyrus species parasitizing Tilapia. Clades 2, 3 and 4 and C. tilapiae formed a weakly supported group using BI analysis. The large clade 5 with weak to good nodal support depending on the phylogenetic method applied (see Material and Methods, part “Phylogenetic analyses of parasite species”, for the definition of nodal support values) was formed by two groups strongly supported by BI; this clade included 10 Cichlidogyrus species, among them 8 are parasites of Tilapia guineensis. The strongly supported clade 6 included Cichlidogyrus halli collected from three different cichlid hosts, Sarotherodon galilaeus, Tilapia guineensis and Oreochromis niloticus.

Host phylogeny

An unambiguous alignment of cytochrome b sequences from cichlids spanned 342 positions. All analyses yielded congruent topologies among the phylogenetic trees. The MP analysis resulted in 2 equally parsimonious trees of 777 steps (CI = 0.319, RI = 0.447). In ML reconstructions, the use of the codon model produced a tree with a low resolution (although congruent with other, more resolved, phylogenetic hypotheses for cichlids obtained here). We then kept the well-resolved tree obtained using the codon partition scheme (Figure 3). Parameters of codon partition model are included in Table 1.

Cophylogenetic analysis

For parasites, only the tree based on SSU rDNA and ITS1 sequence data (Figure 2) was used because of the much higher clade support values than in the LSU tree (Figure 1). The tanglegram showing associations between Cichlidogyrus and Scutogyrus monogenean species and their cichlid fish hosts, based on ME phylogenetic trees, is presented on Figure 4. We used two methods to assess the level of cophylogenetic signal in these host-parasite associations: 1. ParaFit, a method which compares patristic distance between host pairs and their corresponding parasites across the whole association and is able to take into account multiple parasites/hosts per hosts/parasites if any, and 2. Jane, a method comparing the two tree topologies (considering branch lengths) that optimally fits the parasite tree onto the host tree by mixing different kinds of coevolutionary events with predefined costs. The optimal fit is found by minimizing the global cost of the reconstruction. In both approaches, the observed structure is tested against a distribution generated from random associations to assess if it is statistically significant, and ParaFit tests the effect of individual host-parasite associations (“links”) on the global fit (see Material and Methods section for additional details and references). Using ParaFit, the overall cophylogenetic structure was significant (with ME (p = 0.001) or ML (p = 0.047) trees). Nine individual links out of 34 contributed significantly to this global fit (see Figure 4) using ME trees, but only 3 with ML trees (Hemichromis fasciatus – Cichlidogyrus longicirrus, H. fasciatus – C. falcifer, H. letourneuxi – Cichlidogyrus sp. 1).

Using different cost schemes in Jane generated different results (Table 3), but the significance of the global cost (P = 0.01) was only attained in the TreeFitter default model (cost settings 0 for cospeciation, 0 for duplication, 2 for host-switch, 1 for loss, 1 for failure to diverge (added to the original TreeFitter cost only based on the four first types of coevolutionary events)). Among the cost sets tested here, this set of costs has the highest host-switching cost. In all reconstruction, the number of duplications (i.e. parasite speciation without host speciation, or intrahost speciation) was very high. All other cost schemes used resulted in a non significant fit between parasite and hosts phylogenies.

Mapping of the morphology of the attachment apparatus onto the parasite phylogeny

The mapping of anchor morphology suggests that similar shape for both anchor pairs is their ancestral character state, and then that anchors with different shapes represent a derived condition (Figure 5A). The mapping of marginal hooks shape onto the parasite phylogenetic tree (Figure 5B) suggests that having all pairs of small marginal hooks (i.e. 2.2) is the ancestral state in Cichlidogyrus. From this ancestral state, two different character states of marginal hooks may have derived (representing character states 2.1 in Cichlidogyrus and 2.3 found in Cichlidogyrus and Scutogyrus). Mapping also suggests the occurrence of two independent changes in the shape of marginal hook toward a state where all pairs of hooks are large (character state 2.4), one in Cichlidogyrus sp. 1 parasitizing Hemichromis letourneuxi and the other in Cichlidogyrus arthracanthus parasitizing Tilapia guineensis. The ancestral state of the ventral transverse bar (Figure 5C) could not be hypothesized from this analysis but the morphological type with membranous extension (i.e. character state 3.1) was observed in the majority of species. Two changes in the shape of the ventral transverse bar toward a massive bar with membranous extension (character state 3.3) were inferred in Cichlidogyrus tilapiae and Cichlidogyrus sclerosus, as well as one change in the ventral transverse bar with membranous extension toward the bar supporting one large oval plate (character state 3.4) in Scutogyrus species. We could not identify the ancestral state of the dorsal transverse bar (Figure 5D), but dorsal bars with well developed auricles (character state 4.1) were observed in the majority of Cichlidogyrus species. A bar with two small auricles on the anterior face was found in C. pouyaudi and Cichlidogyrus parasitizing Hemichromis species. One change in this state toward the dorsal bar with two long auricles and lateral outgrowths was inferred in Scutogyrus.

Phylogenetic status of gill monogeneans parasitizing cichlid fishes

Based on phylogenetic analyses of LSU rDNA using as outgroup specific gill parasites of Onchobdella, endoparasitic Enterogyrus (found in Sarotherodon and Tilapia species) and Protogyrodactylus (a parasite genus selected following Mendlová et al. [31], but not included among cichlid parasites), we investigated whether or not the Cichlidogyrus/Scutogyrus group is monophyletic, and confirmed this was the case. Previously published molecular phylogenetic analyses suggested Cichlidogyrus to be a polyphyletic taxon [26], [31], [32] and pointed to the different origins for endoparasitic Enterogyrus and ectoparasitic Onchobdella (specific to Hemichromis species) compared to gill monogeneans Cichlidogyrus and Scutogyrus [31]. Moreover, the non-monophyletic origin of Scutogyrus was supported by phylogenetic analyses of ribosomal DNA sequences [32]. It has been proposed that the Scutogyrus genus arose from Cichlidogyrus, according to the morphology of dorsal and ventral transverse bars [33]. In the present study, based on the phylogenetic analyses using LSU, SSU rDNA and ITS1 sequences, we suggest that Scutogyrus species form a monophyletic group contrary to Wu et al. [32]. However, we confirmed the polyphyletic origin of Cichlidogyrus, suggesting the need for a taxonomical revision of this genus.

Cichlidogyrus/Scutogyrus phylogeny: a link to behavioral strategies of cichlid fish

The phylogenetic analyses using LSU rDNA sequences performed in this study placed in basal position Cichlidogyrus pouyaudi parasitizing Tylochromis intermedius, which suggests that this parasite diverged earlier than the other Cichlidogyrus and Scutogyrus species. Cichlidogyrus pouyaudi was originally described from Tylochromis jentinki in West Africa [34], where it was observed that the structure of the dorsal transverse bar of this parasite species is different of the other Cichlidogyrus species. Pariselle and Euzet [34] suggested that such haptor morphology represents an archaic feature in Cichlidogyrus species living on ancient cichlid fish such as Tylochromis species.

In this study, six clades of gill parasites within Cichlidogyrus and Scutogyrus were identified using phylogenetic analyses based on SSU and ITS1. Two clades only contain strictly host specific Cichlidogyrus parasites, i.e. clade 1 parasitizing Hemichromis species and clade 4 parasitizing Tilapia guineensis. Contradictory to the prediction of Pouyaud et al. [26], we found that three Cichlidogyrus species parasitized both mouthbrooders and substrate-brooders. Clade 2 included Cichlidogyrus parasitizing mouthbrooder cichlids (i.e. Oreochromis and Sarotherodon), except Cichlidogyrus sp. 2 found on the mouthbrooder Sarotherodon galilaeus as well as the substrate-brooder Tilapia guineensis. The absence of other Cichlidogyrus parasitizing Tilapia guineensis in this clade suggests a secondary host transfer of Cichlidogyrus sp. 2 from mouthbrooders to substrate-brooders. Cichlidogyrus thurstonae and C. douellouae, both parasites of mouthbrooders (in clade 5 of phylogenetic trees), probably colonized their mouthbrooder host species through lateral transfers (i.e. host switch), as suggested Pouyaud et al. [26]. Our study was limited to Senegal, but we cannot rule out that the host range of generalist Cichlidogyrus species infecting two cichlid groups with different reproductive behavioral strategies spans a broader geographical area (i.e. all cichlid species living in Africa). Concerning the generalist Cichlidogyrus species reported in our study, prevalence and abundance of C. halli 1 and Cichlidogyrus sp. 2 were higher in Sarotherodon galilaeus than in Tilapia guineensis, and higher prevalence and abundance were observed for C. tilapiae in Oreochromis niloticus than in Sarotherodon galilaeus, Tilapia guineensis and Hemichromis fasciatus. This suggests that generalist monogeneans display a level of host preference, i.e. a generalist selects preferentially one host species within its host range, such as a “common” host species compared to “additional” host species [35]. This supports the hypothesis of a mouthbrooder origin for Cichlidogyrus in clade 2 as well as a mouthbrooder origin for the C. halli clade.

Cichlid phylogeny

The cichlid phylogeny based on cytochrome b sequences supported three monophyletic groups of African cichlid species: substrate-brooders Hemichromis, mouthbrooders Sarotherodon and Oreochromis, and substrate-brooder Tilapia. The third group contains East African cichlid species. This separation of Sarotherodon and Oreochromis from Tilapia species has been previously reported based on the mitochondrial tRNA^Pro gene and the control region sequences [36]. This finding agrees with the hypothesis that the mouthbrooding behavior of Oreochromis and Sarotherodon genera evolved from a substrate-brooding behavior [36]. Mayer et al. [13] suggested that West African cichlids of the genera Tylochromis and Hemichromis diverged from the common cichlid stock first and then followed the divergence of Tilapia and Oreochromis. The separation of mouthbrooders and substrate-brooders is then supported by our phylogenetic analyses based on cytochrome b. Following Pouyaud et al. [26], the split between mouthbrooders and substrate-brooders is hypothesized to be linked to the separation of their specific gill parasitofauna. This idea is supported by the observation that Scutogyrus parasitizes only mouthbrooders (Sarotherodon and Oreochromis). Further, Hemichromis species possess specific gill monogeneans from the Onchobdella genus, not shared by other cichlid species. However, some Cichlidogyrus species are able to parasitize both mouthbrooders and substrate-brooder cichlid species (see above). In the present study Tylochromis intermedius was found at a basal position relative to the other African cichlids. This basal position of Tylochromis among African cichlids supports the observation of Streelman et al. [37] using sequences of the nuclear locus Tmo-4C4, and of Zardoya et al. [38] based on microsatellite data. Morphological analyses place Tylochromis as a sister group to African tilapiines [12], but this assumption is not supported by molecular studies [39], [40].

Structural evolution of the haptor

Pouyaud et al. [26] defined four morphological groups within Cichlidogyrus/Scutogyrus parasites, “halli”, “scutogyrus”, “tiberianus” and “tilapiae”, using cluster analysis on morphometrical data from haptoral sclerotized parts. Our phylogenetic analyses did not support the monophyly of the “tiberianus” or “tilapiae” groups, because the most diversified clade 5 in our phylogenetic reconstructions included all species classified as “tiberianus” by Pouyaud et al. [26] but also C. cubitus classified as “tilapiae”. The mapping of the haptor morphological characters performed in the present study relies on the hypothesis that C. pouyaudi diverged early compared to the other Cichlidogyrus and Scutogyrus species (see above) and thus, the characters of its haptor (two morphologically similar pairs of anchors, all pairs of small marginal hooks, a dorsal bar with small auricles and a ventral bar without membranous extension) are considered to form the ancestral Cichlidogyrus haptor type. Moreover, following previous studies on cichlid phylogenies (see above) and Pouyaud et al. [26], we expected to find derived structural characters of haptor in Cichlidogyrus from Tilapia, Oreochromis and Sarotherodon compared to Cichlidogyrus of Hemichromis and Tylochromis (because these latter genera display a basal position in the African cichlid phylogeny). The mapping of structural characters in Cichlidogyrus and Scutogyrus species suggests that the haptor evolved from the simplest type toward the more complex. It also suggests a trend towards a clade-specific morphology with respect to marginal hooks, even if a few changes toward more complicated characters and one reversion to the ancestral state (for C. cubitus) were inferred in this analysis. Nevertheless, mapping does not support a different evolution of structural parts of the haptor in mouthbrooder and substrate-brooder cichlids and thus, it does not suggest any morphological adaptation of Cichlidogyrus species to the cichlids displaying different reproductive strategies. However, Cichlidogyrus/Scutogyrus species of only six cichlid fish species were considered here, and further studies (e.g. taking into account parasites from congeneric Tilapia or Sarotherodon host species) are needed to confirm these hypotheses. Pouyaud et al. [26] compared two dendrograms, based on morphological data from sclerotized parts of respectively the haptor and the reproductive organ in Cichlidogyrus/Scutogyrus parasites, with a phylogeny reconstructed from ribosomal DNA sequences. The two dendrograms were different and only the dendrogram computed from haptor data was congruent with the phylogenetic tree. This suggests that the morphology of the haptor is more suitable for inferring phylogenetic relationships than the morphology of reproductive organs, maybe due to a faster rate of evolutionary change in the morphology of reproductive structures.

Cophylogenetic analysis of Cichlidogyrus/Scutogyrus parasites and their cichlid hosts

Because monogeneans are parasites with a direct life cycle and are highly host specific, they have long been considered to cospeciate with their hosts [15], [16], [41]. Recent studies have shown that to the contrary, monogeneans rarely display any significant cospeciation signal with their hosts, and that host-switching and duplication were thought to be important evolutionary events in parasite diversification, e.g. in Lamellodiscus [20], Gyrodactylus [22], [24], Polystoma [25], and Dactylogyrus [21], [35]. In the latter studies, Dactylogyrus diversification was explained in a large part by intrahost speciation (parasite duplication). All these studies suggest that the high host specificity of monogeneans is not linked to cospeciation. In the present study, distance-based analysis suggests that the global cophylogenetic structure in the Cichlidogyrus/Scutogyrus-cichlid system is significant. Tree-based analyses, however, indicate that this global structure is not significant, unless if the cost of host-switching is high. That supports previous hypotheses that host-switching, followed by speciation which results in the maintenance of high host specificity, is an important component of monogenean diversification. In all reconstructions, the number of duplications is high, which is also coherent with previously published hypotheses on monogenean evolution. Duplication is then suggested to be the main coevolutionary event explaining the diversification of gill monogeneans living on West African cichlid fish. This has been observed in other gill monogeneans-freshwater fish systems such as Dactylogyrus-Cyprinidae [21] and Thaparocleidus-Pangasidae (our unpublished data). The fact that global fit is significant with distance-based analysis only suggests that parasites switch to not too distantly related hosts (but not necessarily sister-species). This is supported by the fact that only few individual associations (related to putative cospeciation events, see Legendre et al. [42] and Desdevises et al. [20]) significantly explain this global fit. This confirms the opportunistic behavior and the evolutionary plasticity of monogeneans, which can certainly easily duplicate on hosts, switch hosts and speciate on their new host species, then diversifying at a high rate and maintaining their tremendous diversity.

Parasite sampling and identification

A total of 28 parasite species belonging to four dactylogyridean genera (Cichlidogyrus, Onchobdella, Scutogyrus and Enterogyrus) were collected from the gills and stomachs of cichlid species (Table 2) during field studies in the Niokolo Koba National Park, (Senegal, Africa). Eighty-six cichlid specimens, belonging to six species (Hemichromis fasciatus, Hemichromis letourneuxi, Oreochromis niloticus, Sarotherodon galilaeus, Tilapia guineensis and Tylochromis intermedius), were caught in the Gambie River (Gue de Damantan: 13°1′37″N, 13°11′33″W; Simenti: 13°0′50″N, 13°10′24″W; Camp de Lion: 13°0′53″N, 13°8′41″W), in the Niokolo River (Passage Koba: 13°2′21″N, 13°6′5′W; Pont Suspendu: 13°0′54″N, 13°7′55″W), in the Mare de Simenti (13°1′4″N, 13°10′33″W) and Mare de Wouring (13°7′56″N, 13°11′9″W) and used for this study. Sarotherodon galilaeus individuals from Ivory Coast were also investigated for Scutogyrus species (because these parasites occur on S. galilaeus in Senegal) but none were found during this study.

All cichlids sampled were examined using standard parasitological methodology described in Ergens and Lom [43]. Monogeneans were removed from the gills of freshly killed fish, placed in a drop of water on slides covered by a coverslip, and identified using a light microscope equipped with phase contrast and digital image analysis (Micro Image 4.0 for Windows, Olympus Optical Co., Hamburg, Germany). Parasite determination was performed according to the morphology and size of the sclerotized parts of the haptor (dorsal and ventral anchors, dorsal and ventral transverse bars, marginal hooks) and the reproductive organs (vagina and copulatory organ) following original descriptions [33], [34], [44]–[61]. Parasite specimens were individually preserved in 95% ethanol before DNA extraction. Some specimens from each species were fixed on slides in a mixture of glycerine and ammonium picrate [62].

Molecular analyses of parasites

Parasites were removed from ethanol and dried, and genomic DNA was extracted using DNeasy™ Tissue Kit (QIAGEN) following the manufacturer's instructions. The LSU rDNA region was amplified using C1 and D2 primers [63]. The amplification reaction was performed using 2 units of Taq polymerase (Fermentas), 1x PCR buffer, 1.50 mM MgCl₂, 0.2 mM of dNTP, 0.50 µM of each primer, 0.1 mg/ml BSA and an aliquot of 30 ng of genomic DNA in a total volume of 30 µl. The polymerase chain reaction (PCR) was carried out using the following steps: 2 min at 94°C followed by 39 cycles of 20 sec at 94°C, 30 sec at 56°C and 1 min 30 sec at 72°C, and then 10 min of final elongation at 72°C. The partial SSU rDNA region and the entire ITS1 were amplified in one round using S1 and IR8 primers [64]. The amplification reaction was performed using 1.5 units of Taq polymerase (Fermentas), 1x PCR buffer, 1.50 mM MgCl₂, 0.2 mM of dNTP, 0.8 µM of each primer, 0.1 mg/ml BSA and an aliquot of 30 ng of genomic DNA in a total volume of 30 µl. PCR was carried out in the Mastercycler ep gradient S (Eppendorf) with the following steps: 2 min at 94°C followed by 39 cycles of 1 min at 94°C, 1 min at 53°C and 1 min 30 sec at 72°C, and 10 min of final elongation at 72°C.

The PCR products were electrophoresed on a 1% agarose gel and then purified by either Wizard® SV Gel and PCR Clean-Up System (PROMEGA) or QIAquick PCR Purification Kit (QIAGEN). Sequencing was performed on an ABI 3130 Genetic Analyzer (Applied Biosystems) using Big Dye Terminator Cycle Sequencing kit version 3.1 (Applied Biosystems). Sequences were analyzed using Sequencher software (Gene Codes Corp) and deposited in EMBL under Accession numbers (see Table 2).

Phylogenetic analyses of parasite species

DNA sequences were aligned using Clustal W multiple alignment [65] in BioEdit v. 7.0.9 [66]. Gaps and ambiguously aligned regions were removed using GBlocks [67] with the less stringent parameters available in the software. We applied the following criteria “Allow smaller final blocks”, “Allow gap positions within the final blocks” and “Allow less strict flanking positions”. First, phylogenetic analyses using LSU rDNA sequences including Cichlidogyrus and Scutogyrus as ingroup and Enterogyrus, Onchobdella and Protogyrodactylus as outgroup (following Mendlová et al. [31]) were performed. Next, phylogenetic analyses using partial SSU rDNA and ITS1 sequences of Scutogyrus and Cichlidogyrus species were performed. The list of parasite species used in the LSU rDNA and/or SSU rDNA and ITS1 alignments is shown in Table 2.

Phylogenetic analyses based on minimum evolution (ME), maximum parsimony (MP) and maximum likelihood (ML) were performed in PAUP*4b10 [68]. Bayesian inference of phylogeny (BI) was computed using MrBayes 3.1.2 [69]. MP analyses were performed using a heuristic search using 10 random searches with a stepwise random addition sequence running on unweighted informative characters and TBR branch swapping. ModelTest [70] was applied to select the most appropriate substitution model of nucleotide evolution for each data set using hierarchical likelihood ratio tests (hLRTs), to be applied in ME, ML (also using heuristic search and TBR) and BI tree reconstructions. ME analysis [71] was performed using heuristic search with a distance optimality criterion. Support for internal nodes were estimated using a bootstrap resampling procedure [72] with 1000 replicates for MP and ME and 100 replicates for ML. Bayesian inference analyses were performed using four Monte Carlo Markov chains running on 3,000,000 generations for LSU rDNA and 1,000,000 for the SSU rDNA and ITS1 data set, with trees being sampled every 100 generations. The “burn-in” asymptote was estimated by plotting the number of generations against the log likelihood scores for the saved trees, and all the trees before stationarity were discarded. In resulting phylogenetic trees, clade support indicated by bootstrap values/posterior probabilities was considered as follows: weak support 50–63%/0.5–0.69, moderate support 64–75%/0.7–0.84, good support 76–88%/0.85–0.94 and strong support 89–100%/0.95–1.00 [73].

Host phylogeny

The phylogeny of cichlid fishes was previously investigated using mitochondrial DNA sequences [9], [74], [75], nuclear DNA sequences [13], [39], [76] and microsatellite data [37]. However, no previously published study included all the cichlid species investigated in the present study.

Fin clips from cichlid species were preserved in 95% ethanol before DNA extraction. Mitochondrial DNA was isolated with DNeasy Blood and Tissue Kit (QIAGEN) following the manufacturer's instructions. The partial region of cytochrome b (434 bp) of Tilapia guineensis, Hemichromis fasciatus, H. letourneuxi, Oreochromis niloticus and Sarotherodon galilaeus was amplified using forward primer L14725 (5′-CGAAGCTTGATATGAAAAACCATCGTTG-3′) designed by Farias et al. [14] and reverse primer Cichlidae_cytb_1R (5′-WRACKGYAGCVCCTCAGAATGAYA-3′) designed in this study. The partial region of cytochrome b (452 bp) of Tylochromis intermedius was amplified using forward primer (5′-TTTTACCAGGACTCTAACCAGGA-3′) and reverse primer (5′-GCYCCTCARAATGATATTTGTCC-3′), both of them designed in this study. The PCR reaction mixture consisted of 1.5 units of Taq polymerase (Fermentas), 1x PCR buffer, 2.50 mM MgCl₂, 0.2 mM of dNTP, 0.3 µM of each primer and an aliquot of 30 ng of genomic DNA in a total volume of 30 µl. The PCR was carried out in the Mastercycler ep gradient S (Eppendorf) with the following steps: 4 min at 95°C followed by 30 cycles of 30 sec at 95°C, 30 sec at 50°C and 45 sec at 72°C, and 10 min of final elongation at 72°C. Electrophoresis was performed on 1% agarose gel and the PCR product was purified by Wizard® SV Gel and PCR Clean-Up System (PROMEGA) or QIAquick PCR Purification Kit (QIAGEN). Sequencing was performed on an ABI 3130 Genetic Analyzer (Applied Biosystems) using Big Dye Terminator Cycle Sequencing kit version 3.1 (Applied Biosystems). The sequences of cytochrome b were deposited in EMBL under Accession numbers (Table 4). DNA sequences were aligned using Clustal W multiple alignment [65] in BioEdit v. 7.0.9 [66].

Phylogenetic analyses were based on partial cytochrome b sequences including 24 fish species as ingroup (i.e. Cichlidae of Africa, Madagascar and South America), and Cymatogaster, Etheostoma and Halichoeres as outgroup (Cichlidae of North America and Western Pacific selected following Farias et al. [14]). Because these coding sequences were highly conserved and of the same size (342 positions after aligning the cytochrome b sequences of our cichlid species with the sequences available in GenBank, see Results section), no insertion or deletion event nor trimming was needed to improve the alignment, that was carefully checked visually. The list of fish species and accession numbers for cytochrome b sequences are shown in Table 4. Bayesian inference analyses were performed using four Monte Carlo Markov chains running on 2,000,000 generations with trees being sampled every 100 generations. Cytochrome b DNA sequences were considered with an evolutionary model designed for coding sequences taking the genetic code into account [77], [78], as well as a codon partition scheme considering independently each position within the codon [79]. The low variability within the alignment precluded the use of translated sequences in phylogenetic reconstructions.

Cophylogenetic analyses

Two methods of coevolutionary analyses were used: a distance-based method called ParaFit [42] implemented in CopyCat [80] and a tree-based method implemented in Jane 3.0 [81]. Note that Jane 3.0 was designed for a good handling of widespread parasites (i.e. using multiple hosts), as it is the case in this study, and that ParaFit was tested for such situation [42], for which it was shown to be efficient. A useful review of existing methods for cophylogenetic studies is given in Light and Hafner [82]. A tanglegram representing the host-parasite associations was reconstructed using TreeMap 1.0 [83].

Distance-based methods focus only on the fit between host and parasite distances and do not test for the presence of any coevolutionary events. These methods use distance matrices and host associations to determine if hosts and parasites are randomly associated. An advantage of this is that they do not require fully resolved phylogenetic trees and can account for parasites associated with multiple hosts. In this study, patristic distances were calculated in CopyCat for each host and parasite phylogeny. The global fit between trees is computed and tested by randomizing individual host-parasite associations (links). ParaFit was also used to test whether a particular host-parasite link contributed to this global fit. Tests of significance were performed using 999 permutations.

Tree-based methods use topologies and branch lengths to assess the fit between host and parasite phylogenies. These methods attempt to reconstruct the shared evolutionary history between hosts and their parasites with the smallest “cost” or smallest number of hypothesized historical events. A disadvantage of tree-based methods is that they require fully resolved phylogenies, and then do not account for phylogenetic uncertainty. Some of them, as TreeMap 1.0, do not appropriately account for parasites associated with multiple hosts in certain cases and therefore may underestimate host switching [83], [84]. To overcome this problem, we conducted analyses with Jane [81], using different event costs schemes. In addition to the four types of coevolutionary events classically used in such studies i.e. cospeciation, duplication (parasite speciation without host speciation), host switching, and sorting, Jane uses a fifth type named “failure to diverge”, referring to the instances when a host speciation is not followed by parasite speciation, which remains as the same species on the newly produced host species. Each type of event is attributed a cost and the algorithm searches the reconstruction with the lowest global cost. In our study, we used the fully resolved minimum evolution parasite tree inferred from the analysis of combined SSU rDNA and ITS1 data. Seven models with different event costs scheme were used for the cophylogenetic analyses performed in Jane (see Table 3), using 500 generations and a population size of 50 as parameters of the genetic algorithm. Statistical tests were computed using 999 randomizations with random parasite trees.

Mapping of the morphology of attachment apparatus onto the parasite phylogenetic tree

We investigated whether morphological evolution, i.e. evolution of the attachment apparatus, is linked to the parasite phylogeny. Categorical coding was used for character states, which were unordered because no relevant hypotheses could be applied for character polarization. Morphological characters of the haptor were mapped onto the fully resolved minimum evolution parasite tree inferred from the analysis of combined SSU rDNA and ITS1 data using MacClade version 4.0.1 with Farris optimization [85]. The following morphological characters were evaluated: shape of the anchors, shape of marginal hooks, shape of the ventral transverse bar, and shape of the dorsal transverse bar (see Table 5 for character states). The morphological characters of the sclerotized parts of the haptor and character states are shown on Figure 6. Two character states were defined for anchors: similar shape of both pairs of anchors and different shape of the first (ventral) and second pair (dorsal) of anchors. Four character states were defined for marginal hooks: first pair of large hooks and 3rd to 7th pair of small hooks, all pairs of small hooks, first pair of small hooks and 3rd to 7th pair of large hooks, and all pairs of large hooks. The morphology of the second pair of hooks was not considered in this analysis because of its small size in all parasite species. For the ventral bar, four morphological types were defined: bar with membranous extension, bar without membranous extension, massive bar with membranous extension, and arched bar supporting one large, thin, oval plate marked by fan-shaped median thickenings. Finally, three character states were defined for the dorsal bar: bar with two well developed auricles on the ventral face of the bar, bar with two very long auricles and lateral outgrowths, and bar with two small auricles on the anterior face of the bar.