Quadriacanthus species (Monogenea: Dactylogyridae) from catfishes (Teleostei: Siluriformes) in eastern Africa: new species, new records and first insights into interspecific genetic relationships

Background African catfishes of the families Bagridae and Clariidae are known to be parasitized with monogeneans of Quadriacanthus Paperna, 1961 (Dactylogyridae). The genus remains taxonomically challenging due to its speciose nature and relatively wide host range representing two fish orders, i.e. Siluriformes and Osteoglossiformes, in Africa and Asia. Here, we investigated diversity of Quadriacanthus spp. parasitizing Clarias gariepinus (Burchell), Heterobranchus bidorsalis Geoffroy Saint-Hilaire, and Bagrus docmak (Forsskål) collected in the Lake Turkana (Kenya) and Nile River Basin (Sudan). The interspecific relationships among Quadriacanthus spp. parasitizing catfishes inferred from ribosomal DNA sequences were investigated for the first time. Methods A combined morphological and molecular approach was used for description of the new species and for a critical review of the previously described Quadriacanthus spp., by means of phase contrast microscopic examination of sclerotized structures, and assessing the genetic divergence among the species found using rDNA sequences. Results Seven species (including four new) of Quadriacanthus were identified. These were as follows: Quadriacanthus aegypticus El-Naggar & Serag, 1986, Quadriacanthus clariadis Paperna, 1961, Quadriacanthus fornicatus n. sp., Quadriacanthus pravus n. sp., and Quadriacanthus zuheiri n. sp. from Clarias gariepinus (Clariidae); Quadriacanthus mandibulatus n. sp. from Heterobranchus bidorsalis (Clariidae); and Quadriacanthus bagrae Paperna, 1979 from Bagrus docmak (Bagridae). For both 18S-ITS1 and 28S rDNA regions, Q. clariadis from a clariid fish was found to be most closely related to Q. bagrae from a bagrid host. Quadriacanthus mandibulatus n. sp. was observed to be the most distant species from the others. The separation of Q. mandibulatus n. sp. from the other species corresponds with the different morphology of its copulatory tube. The copulatory tube is terminally enlarged in Q. mandibulatus n. sp., while the tube in all other congeners studied is comparatively small and with an oblique tapering termination. Conclusions This study contributes to a better understanding of African dactylogyrid diversity and provides the first molecular characterization of Quadriacanthus spp. The observed interspecific genetic relationships among Quadriacanthus spp. from clariids and Q. bagrae from a bagrid host suggest a possible host-switching event in the evolutionary history of the genus. Our records extend the currently known geographical range for Quadriacanthus spp. to Kenya and Sudan.


Background
Monogenea is a diverse group of mostly ectoparasitic flatworms showing great potential as model organisms to study the ecological and evolutionary processes that drive diversification and speciation. The high host specificity shown by most monogeneans enables searches for links between the ecological characteristics of the hosts and the diversity of their parasites [1].
The genus was proposed by Paperna [4] for Q. clariadis Paperna, 1961 from the gills of Clarias gariepinus (Burchell) (syn. C. lazera) collected in Israel and characterized, in part, by having two unequal bars, each with a solid base. Kritsky & Kulo [5] subsequently emended the diagnosis of Quadriacanthus and recognized that the ventral bar is composed of two components articulating medially. Despite the work of these authors, Dubey et al. [6] established Anacornuatus Dubey, Gupta & Agarwal, 1992 for those species of Quadriacanthus that possess a two-piece ventral bar instead of a single-piece ventral bar, as indicated by Paperna [4]. They were evidently unaware of the work of Kritsky & Kulo [5] and hence erred in proposing the new genus. Consequently, Lim et al. [2], who listed 24 species of Quadriacanthus parasitizing Clariidae (Clarias spp. and Heterobranchus spp.) and African Bagrus spp., plus one Quadriacanthus species of doubtful validity, infecting tilapia (see also [5]), synonymized Anacornuatus with Quadriacanthus. However, these authors were not able to ascertain the validity of the two species assigned to Anacornuatus and considered them as species inquirendae. Tripathi et al. [7] added generic characters, redefined the dorsal bar as "T or Y-shaped with mid-posterior process", and limited the taxon to 25 species (including Anacornuatus postbifidus Dubey, Gupta & Agarwal, 1992 as a new combination within Quadriacanthus). The recent descriptions of three new species from Clarias submarginatus Peters by Bahanak et al. [8] brings the number of Quadriacanthus species from siluriform hosts to 28. Nack et al. [3] revealed the presence of a species of Quadriacanthus on a fish host belonging to the Notopteridae (Osteoglossiformes). This finding extends the host range of species of Quadriacanthus to a new family (Notopteridae) and even a new order (Osteoglossiformes).
In a recent survey of monogeneans parasitizing catfishes from Kenya and Sudan, we recovered three described (Q. aegypticus El-Naggar & Serag, 1986, Q. clariadis Paperna, 1961 and Q. bagrae Paperna, 1979) and four new species of Quadriacanthus. We thus aimed to describe these four species and determine their relationships to congeners based on the partial 18S, entire ITS1, and partial 28S rDNA sequences.

Fish collection
Catfish hosts Bagrus docmak (Forsskål) (Bagridae), Clarias gariepinus (Burchell) and Heterobranchus bidorsalis Geoffroy Saint-Hilaire (Clariidae) (all autochthonous fishes) were collected by hook-and-line or beach seine net, or purchased at local fish markets in five localities in Kenya (Lake Turkana) and Sudan (White and Blue Nile River) during the period 2008-2014 (Table 1; Fig. 1). Fish hosts were identified using the keys given by Bailey [9] and Hopson & Hopson [10]. Scientific and common names of fishes are those provided in FishBase [11] and verified in Eschmeyer et al. [12].

Parasite collection and identification
The gills of freshly killed fishes were extracted and examined in bottled water under a dissecting microscope. Live monogeneans were individually picked from the gills with fine needles and immediately processed. Some specimens were prepared for morphological studies following Musilová et al. [13]. Briefly, they were flattened using coverslip pressure in order to best expose their hard parts, and fixed with a mixture of glycerine and ammonium picrate (GAP). Specimens collected for DNA analyses were bisected using fine needles under a dissecting microscope. Subsequently, one half of the body (either the posterior part with haptoral sclerites or anterior part containing the male copulatory organ) was fixed in 96% ethanol for later molecular analysis; the other body half was completely flattened under coverslip pressure and fixed with GAP for species identification. The body half in GAP was deposited (one per species) as a hologenophore, i.e. a voucher specimen from which a molecular sample is directly derived (see [14] for terminology). Parasite specimens collected in Kenya were not used for molecular analysis. The mounted monogenean specimens (or their parts) were studied using an Olympus BX 61 microscope equipped with phase contrast optics, and drawings were made with the aid of a drawing attachment. Measurements, all in micrometres, were taken using digital image analysis (Stream Motion, version 1.9.2) and are presented as the range followed by the mean and the number (n) of specimens measured in parentheses. The dimensions of the body and haptor were obtained from unflattened specimens as the longest measurements in dorsoventral view; measurements of the sclerotized structures (the haptoral and reproductive hard parts) were taken from specimens flattened under coverslip pressure, facilitated by the blotting of excess water with a filter paper. The schemes of measurement for the hard structures are shown in Fig. 2; in essence, the method of measuring the anchors follows the procedures outlined by Řehulková & Gelnar [15]. The numbering of hook pairs (in Roman numerals) follows that recommended by Mizelle [16]. The male copulatory organ is henceforth abbreviated to MCO. For comparative purposes, specimens of some previously described species were examined:  [17]).

DNA extraction, PCR amplification and sequencing
DNA was extracted from 2 to 6 individuals of each collected species using a DNeasy® Blood & Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer's instructions. DNA was stored in AE buffer at -20°C. Two nuclear ribosomal DNA fragments were used in our analysis: fragment spanning partial 18S rDNA (18S) and entire internal transcribed spacer 1 (ITS1), and fragment of partial nuclear 28S rDNA (28S). Until now, only two 28S sequences for Quadriacanthus kobiensis Ha, 1968 (EF100545, AY841874) and one 18S-ITS1 fragment for Quadriacanthus sp. (HG491496) had been deposited in GenBank. The partial 28S fragment was amplified using primers C1 (forward; 5′-ACC CGC TGA ATT  TAA GCA T-3′) and D2 (reverse; 5′-TGG TCC GTG  TTT CAA GAC-3′) [18]. The 18S-ITS1 fragment was amplified in one round using primers S1 (forward, 5′-ATT CCG ATA ACG AAC GAG ACT-3′) [19] and IR8 (reverse, 5′-GCT AGC TGC GTT CTT CAT CGA-3′), that anneal to the 18S and 5.8S rDNA genes, respectively [20]. PCRs were performed according to Mendlová et al. [21]. The PCR products were electrophoresed on a Gold View strained agarose gel (2%) and then successful PCRs, in which a single fragment was amplified, were purified using High Pure PCR Product Purification Kit (Roche, Mannheim, Germany). The purified PCR products were sequenced for both strands with the same primers as used in the amplification. Sequencing was carried out using BigDye® Terminator v3.1 Cycle Sequencing Kit and an Applied Biosystems 3130 Genetic Analyzer. Nucleotide sequences of the 18S-ITS1 and partial 28S regions were assembled and edited using Sequencher software (Gene Codes Corporation, Ann Arbor, MI, USA). The final sequences were deposited in the GenBank database under accession numbers KX713993-KX713998 and KX685951-KX685956.

Sequence and phylogenetic analysis
Because no significant differences were found between 18S and ITS1 sequence data (partition homogeneity test, P = 1.00), further analyses were performed based on concatenated 18S-ITS1 sequences. Two datasets (18S-ITS1 and 28S) were used to estimate the interspecific relationships among the Quadriacanthus species. All corrected 18S-ITS1 and 28S sequences were aligned using ClustalW [22] and improved manually using the program BioEdit version 7.1.11 [23]. Alignments were then trimmed automatically using TrimAl v.1.3 [24]. Calculations of genetic distances (Kimura 2-parameter [25]) among sequences of Quadriacanthus species were carried out in MEGA 7 [26].
Maximum likelihood analyses were conducted using MEGA 7 [26] with 1000 rapid bootstrap replicates. Data were modelled according to the K2 + G model. Phylogenetic trees were rooted by including available data from GenBank: Schilbetrema sp. (HG491495) isolated from Schilbe intermedius Rüppell in Africa for the 18S-ITS1 dataset, and Schilbetrema sp. (KP056243) isolated from Pareutropius debauwi (Boulenger) in West Africa for the 28S dataset. Quadriacanthus sp. (HG491496) isolated from Heterobranchus bidorsalis Geoffroy Saint-Hilaire in Senegal (West Africa) was also included in the 18S-ITS1 phylogenetic analysis, and Q. kobiensis Ha, 1968 (AY841874) isolated from Clarias batrachus (Linnaeus) in China was used in the 28S phylogenetic analysis.

Differential diagnosis
The present specimens closely fit the characters of the original description of Q. aegypticus by El-Naggar & Serag [27] as well as the drawings/measurements provided by Kritsky & Kulo [5] and Douëllou & Chishawa [28] and there is little doubt that they are conspecific. However, it should be mentioned that in the original paper by Douëllou & Chishawa [28] the illustrations of Q. aegypticus were mixed with those of another Quadriacanthus species; their figure 2 (identified as the haptoral structures, MCO and vagina of Q. aegypticus) clearly represents the sclerotized structures (actually without depiction of a vagina) of another species of Quadriacanthus, most probably those of Q. clariadis, although the morphometrical characteristics of the sclerotized structures correspond to those originally described for Q. aegypticus. Also, Douëllou & Chishawa's [28] figure 1 (identified as the sclerotized structures of Q. bagrae) and figure 2 are identical (see also remarks to Q. bagrae); indeed, figure 2 was later replaced with the correct version (i.e. illustrating the sclerotized structures of the real Q. aegypticus) as an erratum to Douëllou & Chishawa's [28] original paper.

Differential diagnosis
This species was elevated from subspecies status under Q. clariadis and adequately redescribed by Kritsky & Kulo [5]. The morphology and measurements of our specimens generally correspond with the redescription and later characterization of this species by Tripathi et al. [7]. Small differences were observed in the morphology of the hooks. However, the shapes of the hooks fall within the variation observed in our series. In individual specimens, the shanks of hooks appear to be more or less robust. The report of Q. bagrae by Douëllou & Chishawa [28] is erroneous because their drawings and measurements of the haptoral structures and MCO suggest that these authors were dealing with another Quadriacanthus species, most probably with Q. clariadis. Their depiction of the sclerotized structures of "Q. bagrae" (figure 1) shows a ventral bar with elongated arms (each component is more than twice longer than the total length of the ventral anchor, while it is less than twice longer in Q. bagrae) and a dorsal anchor with an elongated bent shaft and short point (in Q. bagrae, the dorsal anchor has a short curved shaft and a moderate point), all of which are characters consistent with specimens identified as Q. clariadis by El-Naggar & Serag [34], Kritsky & Kulo [5], Tripathi et al. [7], and the present study. Moreover, Douëllou & Chishawa [28] themselves supported our opinion by stating that the hooklets (= hooks) of their specimens were similar to those of Q. clariadis rather than to those of Q. bagrae. According to our observations, the morphology of Q. bagrae and Q. clariadis male copulatory organ is very similar; the morphology of haptoral sclerites, however, clearly differs between the two Quadriacanthus species.

Molecular characterization
The sequence of the 18S-ITS1 region of Q. bagrae was 921 bp long, of which 515 bp corresponded to the partial  Table 2).

Differential diagnosis
This species was adequately redescribed by Kritsky & Kulo [5]. Examination of the voucher specimen (MRAC 37.160) showed that our specimens are conspecific with this material. The morphology of the   . Accessory piece basally articulated to the copulatory tube in the form of a spike-like structure; medial part lightly sclerotized; distal part a hook-like structure with broader base.

Differential diagnosis
Quadriacanthus fornicatus n. sp. could be confused with Q. simplex, a species described on Heterobranchus isopterus in Ivory Coast by N'Douba et al. [36], by having nearly identical haptoral sclerites. However, these species are easily differentiated by the comparative morphology of the MCO. The accessory piece of the MCO in Q. simplex is noticeably simpler than that in the new species.

Molecular characterization
The
The new species differs from the latter species by possessing a lightly sclerotized (poorly differentiated or needle-like) distal part of the supporting membrane of the dorsal bar (the distal part of the supporting membrane is fimbriated in Q. thysi) and from all other congeneric species by having a comparatively broad copulatory tube with subterminal flange.

Molecular characterization
The sequence of the 18S-ITS1 region of Q. mandibulatus n. sp. was 879 bp long, of which 500 bp corresponded to the partial 18S rDNA region and 379 bp corresponded to the ITS1 region. The sequence of the partial 28S region was 777 bp long. No intraspecific variability was found in the 18S-ITS1 and 28S sequences.

Description
[Based on 4 flattened and 2 unflattened specimens in GAP; Fig. 8]. Body length 580-589 (585; n = 2); greatest width   [5,36]. It differs from Q. gourenei and Q. papernai by the ventral bar possessing longer (rapidly tapering) components, and is easily differentiated from Q. ashuri by having a ventral anchor with a longer shaft. The new species most closely resembles Q. numidus in the morphometry of the haptoral sclerites; in particular, the ventral anchor of both species is characteristic by having a relatively small base and markedly long evenly curved shaft, and by lacking a sclerotized vagina. However, Q. pravus n. sp. differs from Q. numidus in the shape of the accessory sclerite of the dorsal anchor (triangular in Q. pravus vs wing-shaped in Q. numidus), and in having an MCO characterized by an accessory piece with a terminal hook and subterminal pestle (an accessory piece lamellate in Q. numidus). Douëllou & Chishawa [28] reported that the accessory piece of the MCO of their specimens, identified as Q. numidus, was slightly different from that described by Kritsky & Kulo [5]. According to Douëllou & Chishawa's [28] characterization and depiction, it seems that their specimens are conspecific with our specimens (Q. pravus n. sp.) rather than with the Q. numidus specimens of Kritsky & Kulo [5]. However, because of the poor condition of the slide (MNHN 146 HF), the MCO could not be observed in any of the two voucher specimens. Thus, we hesitate to formally synonymize Q. numidus of Douëllou & Chishawa [28] with Q. pravus n. sp. at this time.

Molecular characterization
The sequence of the 18S-ITS1 region of Q. pravus n. sp. was 919 bp long, of which 514 bp corresponded to the partial 18S rDNA region and 405 bp corresponded to the entire ITS1 region. The sequence of the partial 28S region was 799 bp long. No intraspecific variability was found in the 18S-ITS1 and 28S sequences.   . Copulatory tube straight to slightly curved; base with thickened margins; l = 28-30 (30; n = 6). Accessory piece basally articulated to the copulatory tube; medial part formed as a clamp jaw; hookshaped termination serving as a guide for distal portion of the copulatory tube.
sp. differs from Q. aegypticus by having a noticeably smaller MCO composed of a copulatory tube without basal flange (with flange in Q. aegypticus) and simpler accessory piece (i.e. without two medial diverticula and distal hooks). Examination of the holotype and two paratypes of Q. agnebiensis showed that Q. zuheiri n. sp. differs from the latter species by possessing: (i) a longer ventral anchor with less arched shaft; (ii) a larger accessory sclerite on the part of the dorsal anchor; (iii) shorter and less robust hooks VI and VII; and (iv) an accessory piece with more complex medial part (formed as a lightly sclerotized clamp jaw) and hooked (double hooked in Q. agnebiensis) distal termination.

Molecular characterization
The sequence of the 18S-ITS1 region of Q. zuheiri n. sp. was 877 bp long, of which 469 bp corresponded to the 18S rDNA region and 408 bp corresponded to the ITS1 region. The sequence of the partial 28S region was 772 bp long. No intraspecific variability was found in the 18S-ITS1 and 28S sequences.

Interspecific genetic relationships within genus Quadriacanthus
No intraspecific variability was detected for the 18S-ITS1 and 28S regions. The overall K2P mean genetic distance was 10.34% for the 18S-ITS1 sequences and  Table 3. Among the Quadriacanthus species, Q. clariadis exhibited the lowest genetic divergence from Q. bagrae (1.89% for 18S-ITS1, 0.92% for 28S). Q. mandibulatus n. sp. and Q. fornicatus n. sp. exhibited the greatest genetic distances (5.87%) for 28S rDNA sequences, and Q. mandibulatus n. sp. and Q. bagrae represented the most divergent species pair for 18S-ITS1 sequences (13.95%; Table 3). An unambiguous alignment of 18S-ITS1 sequences spanned 799 positions, of which 275 positions were variable. The 28S alignment contained a total of 725 bp with 246 variable characters. The phylogenetic trees of Quadriacanthus species parasitizing East African freshwater siluriform fishes inferred from 18S-ITS1 and 28S fragments had very similar topologies (Fig. 10). In both gene trees, Q. bagrae showed a sister relationship to Q. clariadis and Q. fornicatus n. sp.; Q. zuheiri n. sp. was sister to Q. pravus n. sp. For 28S, Q. mandibulatus n. sp. formed a separate clade (occupying a basal position); for 18S, Q. mandibulatus n. sp. formed one clade with Q. zuheiri n. sp. and Q. pravus n. sp. Moreover, the phylogenetic analysis of 18S-ITS1 rRNA gene sequences revealed identity between Q. mandibulatus n. sp. and Quadriacanthus sp. retrieved from GenBank (they differed in one nucleotide). Therefore, we consider this Quadriacanthus sp. with the HG491496 sequence, isolated from the airbreathing clariid Heterobranchus bidorsalis in Senegal (Šimková, pers. com.), as a representative of Q. mandibulatus n. sp.

Discussion
The geographical distributions and host preferences of species of Quadriacanthus suggest an interesting evolutionary history of the group. Species of Quadriacanthus have been confirmed as parasites of fishes representing three families, namely the Clariidae, Bagridae (Siluriformes), and Notopteridae (Osteoglossiformes) [2,3]. Clariid catfishes most likely originated in Asia 40-50 MY ago but contemporary African and Asian species originated from a common ancestor that was present on the Arabian plate about 15 MY ago [38]. From that moment, the ancestral species came back to Asia and colonized Africa probably through brackish water bridges like lagoons [39]. Species of Quadriacanthus infesting clariids occur in the freshwaters of Africa, India, Malaysia, Thailand, China and Vietnam [2,7]. Inasmuch as members of dactylogyrid genera are generally considered highly host-specific (usually confined to members of a single host family), the wide geographical distribution of Quadriacanthus spp. on clariid hosts suggests comparatively old host-parasite relationships, i.e. lasting at least 15 MY. On the other hand, formulating a hypothesis on the origin of Quadriacanthus species on bagrids (B. bajad, B. docmak and B. orientalis) in Africa is more problematical. Species of Quadriacanthus have not been found on bagrids in Asia, although these fishes have occasionally been examined for gill parasites [2]. The family Bagridae was poorly defined until its revision by Mo [40] and de Pinna [41], who established the families Austroglanididae, Claroteidae and Auchenoglanididae for all African genera (except Bagrus!) previously considered members of the Bagridae [42]. The wellknown Farenholz' rule states that the natural classification of some parasite groups usually corresponds directly with the natural relationships of their hosts [43]. Indeed, species of claroteids and auchenoglanidids are known to harbour species of Protoancylodiscoides Paperna, 1969 and Bagrobdella Paperna, 1969, respectively, while those of Bagrus are known to be infected with one species of Quadriacanthus, i.e. Q. bagrae [2]. Some authors (e.g. Brooks & McLennan [44]) believe that monogeneans possess characteristics that perfectly adapt them for surviving numerous host-switching events. Assuming that members of the Clariidae are the ancestral hosts of species of Quadriacanthus, the occurrence of Q. bagrae (while clearly a member of the genus) on African bagrid hosts probably resulted from host switching. Our phylogenetic reconstruction indicates that Q. bagrae is phylogenetically nested within the parasites from Clarias gariepinus at a derived position of the tree (Fig. 10). More specifically, Q. bagrae from Bagrus docmak is a sister species to Q. clariadis from C. gariepinus. The clade is located at a derived position of the tree, suggesting that Q. bagrae (or its ancestor) transferred from clariids to species of Bagrus and not conversely. Several studies suggested that such lateral transfer (host switch) can occur both between related host species (e.g. [45]) and even between phylogenetically distant host species [46][47][48].
Recently, Nack et al. [3] hypothesized that the presence of Quadriacanthus euzeti Nack, Pariselle & Bilong Bilong, 2015 on Papyrocranus afer (Notopteridae, Osteoglossiformes) is probably the result of a lateral transfer from species belonging to Clarias or Bagrus which live sympatrically with P. afer in Lake Ossa (South Cameroon). Although more data are needed to resolve phylogenetic relationships within Quadriacanthus, the occurrence of Q. bagrae on Bagrus docmak may represent a similar lateral transfer from a species of Clarias, probably C. gariepinus. Bagrus docmak inhabits, among other locations, the Nile River, where it lives in sympatry with Clarias gariepinus [9]. Although we cannot verify the accuracy of the identification, Q. bagrae was also recorded on C. gariepinus by some authors [5,7]. Because the drawings of the MCO provided by these authors are insufficient for detailed     comparison with our specimens, confirmation of the records of Q. bagrae on C. gariepinus will depend on the collection and evaluation (morphological and molecular) of new parasite material from C. gariepinus. It will be interesting to see whether Q. bagrae on C. gariepinus is a genuine Q. bagrae (sensu stricto). If they represent two different species of Quadriacanthus, then the occurrence of Q. bagrae on Bagrus docmak may suggest, at this time, a case of host switching with speciation.
Until now, there were no studies on the genetic characteristics of Quadriacanthus spp.; thus, the molecular data presented here represent an important advance in the molecular identification and differentiation of this genus. In our study, molecular characterization is presented for six Quadriacanthus species (i.e. for all the species recorded in our study, except Q. aegypticus). The interspecific genetic relationships among Quadriacanthus spp. observed in this study are congruent with the similarity of the basic morphology of the sclerotized structures, especially of those of the MCO (Figs. 11,12,13). The separation of Q. mandibulatus n. sp. from the other species corresponds with the different morphology of its copulatory tube. The copulatory tube is terminally enlarged and with a subterminal flange in Q. mandibulatus n. sp., while the corresponding structure in all other congeners studied is comparatively small and with an oblique tapering termination (Figs. 13, 14).

Conclusions
This study suggests that species of Quadriacanthus parasitizing catfishes in the Old World provide useful models for the study of biogeography and coevolution. However, future studies are needed that would have to involve the examination of dactylogyrids from a greater number of host individuals and host species from a larger geographical area, the utilization of other monogenean taxa, and the incorporation of a homologous series of host features into the matrix derived from the parasite cladogram.