Molecular Phylogenies of Leeches and Haemoparasites Infecting Freshwater Turtles in Aquatic Ecosystems of Northern Africa Suggest Phylogenetic Congruence between Placobdella costata Sensu Lato and Haemogregarina stepanowi Sensu Lato

Haemogregarines are blood parasites with a life-cycle involving a vertebrate as the intermediate host and an invertebrate as the definitive host and vector. Extensive phylogenetic investigations based on 18S-rRNA gene sequences have shown that Haemogregarina stepanowi (Apicomplexa: Haemogregarinidae) is able to infest a large diversity of freshwater turtle species, including the European pond turtle Emys orbicularis, the Sicilian pond turtle Emys trinacris, the Caspian turtle Mauremys caspica, the Mediterranean pond turtle Mauremys leprosa, and the Western Caspian turtle Mauremys rivulata, among others. From the same molecular markers, H. stepanowi is further considered to be a complex of cryptic species predisposed to infect the same host species. While Placobdella costata is known to be the unique vector of H. stepanowi, it is only recently that independent lineages within P. costata have been illustrated—suggesting the presence of at least five unique leech species across Western Europe. The aims of our study were therefore to investigate from mitochondrial markers (COI) the genetic diversity within haemogregarines and leeches infecting freshwater turtles of the Maghreb, in order to identify processes of parasite speciation. We showed that H. stepanowi consists of at least five cryptic species in the Maghreb, while two Placobella species were identified in the same area. Although an Eastern–Western speciation pattern was apparent for both leeches and haemogregarines, we cannot make definitive conclusions regarding co-speciation patterns between parasites and vectors. However, we cannot reject the hypothesis of a very strict host–parasite specificity within leeches.


Introduction
Haemogregarines are blood parasites of the phylum Apicomplexa with a life-cycle involving a vertebrate as an intermediate host and an invertebrate as a definitive host and vector. They are currently classified into four families: the Haemogregarinidae which common ectoparasites of freshwater vertebrates. It has been shown that host specificity differs among Placobdella spp. Certain leeches exhibit an opportunistic generalist behavior with a wide range of host species: for instance, Placobdella montifera Moore 1906 [22,23] and Placobdella phalera (Graf 1899) [24] which infest a great many species; Placobdella multilineata Moore 1953 which has been reported in 17 species and subspecies of reptiles, such as alligators, crocodiles, snakes, and turtles, as well as in amphiumas [25]; and Placobdella ornata (Verrill 1872) which infests turtles, as well as fish, amphibians, birds, and humans [22]. In contrast, certain other leeches exhibit a very strict host specificity. This is the case for Placobdella cryptobranchii (Johnson and Klemm 1977) [26] which has been reported from a single amphibian host species, i.e., the Ozark hellbender Cryptobranchus alleganiensis bishopi Grobman [22,24,26], although Moser et al. [27] claimed an additional host species, i.e., the Mudpuppy Necturus maculosus (Rafinesque). This was also the case for Placobdella appalachiensis Moser and Hopkins 2014 which was reported from a single host species, i.e., the Eastern hellbender Cryptobranchus alleganiensis (Daudin) [28]. Placobdella costata (Fr Müller 1846), the type species of the genus Placobdella, was considered as the only European representative of the genus until Soors et al. [29] reported Placobdella ornata (Verril 1872) in Europe following the introduction of its host. While it was first thought to infect a single turtle species, i.e., E. orbicularis across its entire geographical distribution, from Western Europe to Russia [30], as well as in Algeria [31] and Tunisia [15], P. costata is now known to infect certain other turtles. This has indeed been reported from M. caspica in Iran [32,33], from E. trinacris in Sicilia [34], and from M. leprosa in Spain [35,36], Morocco [37], and Tunisia [15]. While it is now well received from the investigation of molecular markers that Placobdella is a monophyletic genus, certain cryptic diversity was evidenced within several species, among which are Placobdella mexicana Moore 1898 and Placobdella ringueleti López-Jiménez and Oceguera-Figueroa 2009 [38]. Based on COI and ITS1 genetic variations, Kvist et al. [39] also illustrated seven independent lineages within P. costata, suggesting at least five unique leech species across Central and Eastern Europe, as well as in Algeria. Independent phylogenetic relationships were shown among Italian populations of P. costata and their two host species E. orbicularis and E. trinacris [40]; thus, Kvist et al. [39] concluded that species diversity and speciation within P. costata were likely not the result of speciation, isolation, or dispersal of the host species.
Hence, considering the low specificity of H. stepanowi towards its intermediate chelonian hosts, the high diversity of Haemogregarina spp. within turtles, as well as the cryptic species diversity within P. costata, we aimed to investigate the genetic diversity within haemogregarines and leeches infecting North African freshwater turtles. Our objectives were to compare phylogenetic patterns between hosts and their parasites in order to identify processes of parasite speciation.

Materials and Methods
Field-work investigations were carried out on 2019 and 2020, mainly in freshwater aquatic environments of Morocco, Algeria, and Tunisia suitable for E. orbicularis and M. leprosa (see Figure 1 and Table 1 for GPS localities). Sampling sessions were conducted during the peak activity period of both turtles, from March to November. Traps were baited with fish and set in waters with plastic bottles to maintain them at the surface water and to prevent turtles' drowning. They were firmly attached to the bank of the water bodies with a solid rope and checked every day, usually between one to three days, before being removed. Trapped turtles were marked individually on the marginal scutes for further capture-mark-recapture procedures. They were sexed, measured, weighted, and checked for ectoparasites. Leeches usually occur at the base of the anterior and/or posterior limbs. When present, they were collected and preserved in 70% alcohol for molecular systematic studies. Blood was also collected from each specimen from the dorsal coccygeal vein running in the midline of the turtle's tail with the help of 1-mL insulin syringes. A drop of blood was immediately spread out on a glass slide, fixed with a few drops of methanol after air drying, and stored temporarily until staining in the lab. Of the remaining blood, one half was stored in 90% ethanol for molecular systematic studies of the host and their haemoparasites, and the second half was preserved into two distinct Microtainer ® tubes with anticoagulant that were immediately frozen in liquid nitrogen before being transported to the lab where they were finally stored at −80 • C for further biochemical analyses. Investigated turtles were released at the same place of capture in the field.
Blood smears on slides were stained with 10% Giemsa for 20 min in the laboratory, air dried, and examined by optical microscopy using a ×40 objective for the first screening. A Leica digital camera using a ×100 objective was used to capture images for biometric measurements and developmental-stage identification according to Dvoráková et al. [12,41]. A total of 422 blood smears were prepared and examined, including 406 from M. leprosa collected at 12 localities of Morocco, six of Algeria, and 14 of Tunisia, and 16 from E. orbicularis collected at five localities of Algeria.
DNA extractions were performed with the E.Z.N.A Tissue DNA Kit (Omega biotek, Norcross, GA, USA) following recommendations of the supplier for blood and tissue samples preserved in ethanol. DNA extracts were resuspended in approximately 200 µL of Tris-EDTA 1X elution buffer before use for PCR. The cytochrome c oxidase I (COI) of leeches was amplified using the forward LCO Plac 5 -AYTCAACTAATCAYAAAGAYATTGG-3 and reverse HCO Plac 5 -TADACTTCWGGRTGACCAAAAAATCA-3 primers which were designed for Placobdella spp. [38]. The COI of haemogregarines was amplified using the forward HemoFor4 5 -TGGACATTATACCCACCTTTAAG-3 and reverse HemoRev4 5 -ATACAACCCATAGCTAGTATCAT-3 primers which were designed for Haemogregarina spp. [16]. PCR assays of COI were conducted in a final volume of 25 µL comprising 1X buffer, 1.5 mM MgCl 2 , 0.2 mM dNTPs, 0.4 mM primers, 0.75 units GoTaq Polymerase (Promega, Charbonnières-les-Bains, France), and DNA (3 µL) under the following conditions: an initial step of 5 min at 95 • C for long denaturation; followed by 35 cycles of 30 s at 95 • C for denaturation, 30 s at 48 • C for annealing, and 1 min at 72 • C for elongation; with a final step of 10 min at 72 • C for terminal extension. The success of the PCR amplifications was controlled following electrophoresis in gels of 1% agarose stained with ethidium bromide. When amplifications were successful, they yielded PCR fragments of approximately 700 bp for the COI of leeches and 465 bp for the COI of haemogregarines which were subsequently sequenced by the Genoscreen Company (Lille, France).
COI chromatograms were edited with Chromas 2.2.6 (Technelysium Pty Ltd., Brisbane, Australia), and the resulting new sequences obtained for leeches and haemogregarines were respectively grouped in two distinct files with certain others that were extracted from GenBank. The final COI leech data set included 78 COI sequences corresponding to 23 distinct Placobdella species on the one hand and two Helobdella species used for outgroup comparisons on the other. At this stage, P. costata was considered as a single species and referred as P. costata sensu lato. Conversely, the final COI protozoan data set included 86 COI sequences corresponding to seven species of Piroplasmida, 17 species of Haemosporida, and 30 species of Eucoccidiorida, including nine putative Haemogregarina species from two South American turtles, and Scrippsiella sweeneyae Balech 1965 used for outgroup comparisons [16]. At this stage, H. stepanowi was considered as a single species and referred as H. stepanowi sensu lato. The COI sequences included in both data sets were aligned independently using Clustal W [42] implemented in MEGA7 [43] under default parameters.
For the Bayesian analysis, a GTR + I + G model was selected independently for the two COI partitions following the hierarchical likelihood ratio tests (hLRTs) implemented in Modeltest 3.06 [44]. Therefore, six types of substitutions and invariable-gamma rates with four gamma rate categories were applied for each partition, evolutionary parameters being estimated independently. The leech COI partition included 689 characters, while the protozoan COI partition included 448 characters. Bayesian analyses were run using MrBayes 3.04b [45], with four chains running for five million generations and sampled every 100 cycles. Consensus trees for both data sets were drawn, after removing the first 5000 trees as the burn-in phase, and opened with TreeGraph 2 [46]. These were converted into a phylogram, and only nodes supported by more than 0.75 posterior probabilities were shown.
For the parsimony approach, a bootstrap test with 1000 replicates was applied for each data set following a heuristic search under PAUP* Version 4.0 b10 [47] on the parsimonyinformative characters, gaps being treated as missing data. The nearest neighbor interchange (NNI) branch-swapping option was used to find trees of minimal length. In order not to weight down the illustrated trees, only bootstrap proportions higher than 75% were indicated next to the branches within the clades P. costata sensu lato and H. stepanowi sensu lato.
For the distance analysis, a bootstrap test with 5000 replicates was applied for each data set. Minimum evolution (ME) trees were searched using the close-neighbor-interchange (CNI) algorithm [48] at a search level of one for each replication based on Kimura twoparameter distances [49] that were computed under the pairwise-deletion option. Evolutionary analyses were conducted in MEGA7 [43]. Regarding parsimony, only bootstrap proportions higher than 75% were indicated next to the branches for the two clades of interest.
A total of 22 COI sequences from Placobdella rugosa (Verrill 1874) were retrieved from GenBank in order to investigate the molecular threshold within Placobdella. All of these sequences correspond to distinct leech specimens that were sampled at separate sites across Canada, which represents most of the geographic range of that species; thus, we expected that the intra-average distances measured within all these specimens may reflect the average genetic variations within each Placobdella species. Pairwise p-distances were calculated within P. rugosa and the average distance with its standard deviation (SD) was then estimated and given with the minimum and maximum distances between all investigated specimens. Finally, pairwise p-distances were calculated within and between haploclades of P. costata sensu lato to explore the species diversity.
The interspecific average pairwise COI p-distance was estimated to about 2.1% between two closely related valid distinct species, namely Eimeria lancasterensis Joseph 1969 and Eimeria sciurorum Galli-Valerio 1922, and further considered as the molecular threshold within Haemogregarina [16]. Therefore, pairwise p-distances were calculated within and between haploclades of H. stepanowi sensu lato to explore the species diversity based on the molecular threshold. vein running in the midline of the turtle's tail with the help of 1-mL insulin syringes. A drop of blood was immediately spread out on a glass slide, fixed with a few drops of methanol after air drying, and stored temporarily until staining in the lab. Of the remaining blood, one half was stored in 90% ethanol for molecular systematic studies of the host and their haemoparasites, and the second half was preserved into two distinct Microtainer ® tubes with anticoagulant that were immediately frozen in liquid nitrogen before being transported to the lab where they were finally stored at −80 °C for further biochemical analyses. Investigated turtles were released at the same place of capture in the field.  Table 1.   Table 1. Note: Numbers in brackets refer to localities mapped in Figure 1.

Results
Following field-work investigations, M. leprosa was sampled in 12 localities of Morocco, among which five showed the presence of leeches attached to turtles. Following blood observations, haemogregarines were observed within turtles in five localities: only three of them had leeches, the other two did not. The prevalence of haemogregarine infection varied from approximately 11% to 80% in localities where turtles were found to be infected (Table 1). In Algeria, M. leprosa was sampled in six localities, among which three showed the presence of leeches attached to turtles. Following blood observations, haemogregarines were observed within turtles in five localities: only three of them had leeches, the other two did not. The prevalence of haemogregarine infection varied from approximately 25% to 75% in localities where turtles were found to be infected. Emys orbicularis was also sampled in five localities of Algeria-three times in syntopy with M. leprosa-among which three localities showed the presence of leeches attached to turtles. Haemogregarines were observed within turtles in three localities, a single one had leeches, the other two did not.
The prevalence of haemogregarine infection varied from approximately 50% to 75% in localities where turtles were found to be infected (Table 1). In Tunisia, M. leprosa was sampled in 14 localities, among which three showed the presence of leeches attached to turtles. Following blood observations, haemogregarines were observed within turtles in three localities: only two of them had leeches, the last one did not. The prevalence of haemogregarine infection varied from approximately 58% to 67% in localities where turtles were found to be infected (Table 1).
In total, 31 leeches were processed for their COI, 11 collected from M. leprosa sampled from four localities of Morocco, 15 collected from M. leprosa and E. orbicularis sampled from six localities of Algeria, and five collected from M. leprosa sampled from three localities of Tunisia (Table 2). They were characterized by six distinct haplotypes, C8a to C8c in Morocco, C5a to C5c in Algeria, and C5a to C5b in Tunisia. According to the phylogenetic reconstruction shown in Figure 2, P. costata sensu lato constitutes a robust clade that can be split into six distinct lineages, namely haploclades C1 + C7, C3, C4, C5, and C6 based on [39], and Haploclade C8 based on the present results. Conversely, due to technical difficulties in amplifying the COI of haemogregarines from blood samples, only 13 sequences were obtained from M. leprosa sampled from two localities of Morocco, from one locality of Algeria, and from two localities of Tunisia (Table 3). These were characterized by nine distinct haplotypes: 4a to 4d in Morocco, 1c in Algeria, and 3g to 3i and 5a in Tunisia. According to the phylogenetic reconstruction shown in Figure 3, H. stepanowi sensu lato constitutes a robust clade that can be split into five distinct lineages, namely haploclades 1 to 3 based on Attia El Hili et al. [16], and haploclades 4 and 5 based on the present results. Note: Numbers in brackets refer to localities mapped in Figure 1; * sequences used for phylogenetic analyses.     Note: Numbers in brackets refer to localities mapped in Figure 1; * sequences used for phylogenetic analyses.
Regarding the average p-distance that was estimated within P. rugosa, the COI molecular threshold can be considered as approximately 0.90% pairwise substitutions ± 0.0042 within Placobdella (Table 4). Therefore, according to the average p-distances that were estimated within haploclades of P. costata sensu lato on the one hand, which varied from 0.30% for the minimal average distance to 1.44% for the maximal average distance (Table 4), and between haploclades of P. costata sensu lato on the other, which varied from 4.19% for the minimal average distance to 8.15% for the maximal distance (Table 5), each haploclade can be considered as a distinct species, which is in agreement with Kvist et al. [39]. If Haploclade C6 can be considered as P. costata sensu stricto according to Kvist et al. [39] (see Figure 2), leeches that were collected from M. leprosa and E. orbicularis of Algeria and Tunisia can be regarded as the undescribed species corresponding to Haploclade C5 that was previously identified [39]. Similarly, leeches that were collected from M. leprosa of Morocco can be regarded as another undescribed species corresponding to Haploclade C8 identified here for the first time.
Considering the importance of molecular COI of 2.1% pairwise substitutions [16], the average p-distances that were estimated within haploclades of H. stepanowi sensu lato on the one hand, which varied from 0.43% for the minimal average distance to 1.61% for the maximal average distance (Table 6), and between haploclades of H. stepanowi sensu lato on the other, which varied from 6.71% for the minimal average distance to 8.21% for the maximal distance (Table 7), suggest that each haploclade can be considered as a distinct species. While it is still difficult to assign one haplotype to H. stepanowi sensu stricto, haploclades 1 and 3 characterizing haemogregarines sampled from M. leprosa of Algeria and Tunisia, respectively, were previously identified in Attia El Hili et al. [16]. Conversely, haploclades 4 and 5 characterizing haemogregarines sampled from M. leprosa of Morocco and Tunisia, respectively, were identified here for the first time.

Discussion
Regarding the genetic diversity of P. costata leeches sampled in Morocco, Algeria, and Tunisia, we can consider that two species are well differentiated in the Maghreb. According to Kvist et al. [39], P. costata is phylogenetically split into seven independent lineages in the Northern Mediterranean basin, Tunisia, and Algeria-at least five of which correspond to distinct species based on COI genetic variations. Among the mitochondrial haploclades that were illustrated in [39], the C5 haploclade, which groups three different haplotypes characterizing leeches collected in Algeria and Tunisia, was also recovered in our study from leeches sampled in the same area (Haplotypes C5a-c), C5a-b being found in Algeria and Tunisia, and C5c being restricted to Algeria. In addition, we identified a new distinct haploclade that for convenience we named Haploclade C8. The latter also contained three distinct haplotypes (C8a-c) that characterized leeches that were collected in four distant localities of Morocco: C8a being found in three distinct localities, and C8b-c being restricted in the fourth locality. While the levels of COI variation clearly indicated that the C8 haploclade could also be considered as a distinct species, the phylogenetic relationships between all haploclades did not allow the resolution of the deepest nodes in the tree as already illustrated in [39], thus suggesting a rapid diversification of leeches. Based on phylogeographic studies of several amphibian and reptile species across their geographical distribution area in the Maghreb (Morocco, Algeria, and Tunisia), an Eastern-Western speciation pattern was evidenced for most of the investigated species [50]. This speciation pattern was very similar to the phylogeography of M. leprosa that supported its northwestern origin before the species colonized Europe [51]. It was impossible to consider a sister group relationship between the C5 and C8 haploclades, and at least two leech species infect M. leprosa in its home range of Africa; thus, it is very unlikely that leeches in the Maghreb co-diverged along with their chelonian hosts. Furthermore, C5b was found in both turtle species, i.e., M. leprosa and E. orbicularis, thus strengthening the conclusions of Vecchioni et al. [40] who clearly showed that the speciation of P. costata leeches infecting E. orbicularis and E. trinacris in Italy was not the result of co-divergence processes.
Regarding the COI genetic diversity of H. stepanowi revealed during this study on the blood of freshwater turtles from Morocco, Algeria, and Tunisia, and that already evidenced in [16], we can assume that at least five haemogregarine species are well differentiated in the Maghreb. Based on our phylogenetic tree, H. stepanowi sensu lato was phylogenetically split into five independent lineages in the Southern Mediterranean basin: Haploclades 2, 3, and 5 being restricted to Tunisia, Haploclade 1 to Tunisia and Algeria, and Haploclade 4 to Morocco. From the same mitochondrial marker, three independent lineages within the host P. vogli of South America were also reported [17]. Surprisingly, a fourth molecular lineage infecting the Rio Magdalena River turtle Podocnemis lewyana Duméril was found nested within that complex of haploclades [17], thus suggesting the occurrence of at least four undescribed species infecting two distinct host species (see Figure 3). While levels of COI genetic variations clearly indicate the occurrence of several haemogregarine species within the same host species in sympatry, even in syntopy, which is illustrated by numerous cases of coinfections, it is still extremely difficult to link a particular molecular lineage to a specific morphological species, as also reported in [17]. While the first molecular investigations on Haemogregarina have shown from nuclear markers that H. stepanowi possesses a wide distribution though Europe, Turkey, and the Middle East to Iran with a very low hostspecificity [12], the most recent advents in molecular systematics of H. stepanowi have shown that this species could be actually split into a great number of distinct cryptic haemogregarine species [15,19,52,53], infecting E. orbicularis and M. leprosa on the one hand and additional hosts such as M. rivulata and the invasive turtle species Trachemys scripta (Thunberg) [19] on the other. Therefore, one may wonder whether the haemogregarine distribution is correlated to that of its definitive host, namely the Placobdella leech.
Investigating the links between host and parasite diversities to assess evolutionary processes requires well-resolved host and parasite phylogenies. It is also of utmost importance to set up exhaustive sampling procedures as far as possible in order to reduce the biases due to missing data in the field. Regarding leech and haemogregarine phylogenies, while a single leech species was found in Morocco (Haploclade C8), a single haemogregarine species was also evidenced in the same geographical area (Haploclade 4). Neither of these two haploclades were found in Algeria and Tunisia, suggesting a close correspondence between the parasite and its definitive host in Morocco. However, given the technical difficulties encountered in amplifying the COI of the haemoparasites, we did not obtain any information about parasite genotypes infecting turtles in Oued Zat and Oued Sebou localities-while we did for leeches. Additional data would therefore be needed to make conclusions. Concerning the eastern part of the investigated area, Algeria and Tunisia, while a single leech species was found across the sampled turtles (Haploclade C5), three distinct haemogregarine species were evidenced in the same geographical area, i.e., Haploclade 1 in Lac Noir (Algeria), Haploclade 3 in Ghayada (Tunisia), and Haploclade 5 in El Garia (Tunisia). Although none of these parasite haploclades were ever evidenced in Morocco, such as the host haploclade that was restricted to Algeria and Tunisia, it is still difficult to make conclusions regarding a close correspondence between the parasite and its definitive host in the eastern part of Maghreb because of the occurrence of the three distinct parasite lineages. However, it should be noted that no leeches were recovered from El Garia (Tunisia), which prevented linking Parasite Lineage 5 to a particular leech haploclade. Similarly, and as explained above, the difficulties encountered in amplifying the COI of haemogregarines did not allow for the genotyping of parasites in the Algerian Madjen Belhriti, Brabtia, and Lac Tonga, or Tunisian El Hania localities, while the genotypes of leeches were recovered. Additional data should help to provide conclusions regarding strict host (leech) and parasite (haemogregarine) specificity.
Although an Eastern-Western speciation pattern is apparent for both leeches and haemogregarines, with a certain level of geographical host fidelity for parasites, we cannot make definitive conclusions regarding co-speciation patterns between haemoparasites and leeches. The diversity of haemogregarines is the highest in the eastern part of the Maghreb; thus, we may also expect a higher leech diversity in that area, which is currently not the case. For this reason, deeper sampling should be conducted to collect more leeches on the one hand, and in as many localities as possible on the other. Lastly, the genotyping of haemogregarines was performed directly from blood samples of turtles; thus, we recommend that future studies genotype haemogregarines directly from leech blood extracts. Indeed, while host-parasite specificity seems to be very low within turtles, we still cannot reject the hypothesis of a very strict host-parasite specificity within leeches.

Conflicts of Interest:
The authors declare no conflict of interest.