Monophyly of the species of Hepatozoon (Adeleorina: Hepatozoidae) parasitizing (African) anurans, with the description of three new species from hyperoliid frogs in South Africa

Haemogregarines (Apicomplexa: Adeleiorina) are a diverse group of haemoparasites reported from almost all vertebrate classes. The most commonly recorded haemogregarines to parasitize anurans are species of Hepatozoon Miller, 1908. To date 16 Hepatozoon species have been described from anurans in Africa, with only a single species, Hepatozoon hyperolli (Hoare, 1932), infecting a member of the Hyperoliidae. Furthermore, only two Hepatozoon species are known from South African anurans, namely Hepatozoon theileri (Laveran, 1905) and Hepatozoon ixoxo Netherlands, Cook and Smit, 2014, from Amietia delalandii (syn. Amietia quecketti) and three Sclerophrys species, respectively. Blood samples were collected from a total of 225 individuals representing nine hyperoliid species from several localities throughout northern KwaZulu-Natal, South Africa. Twenty frogs from three species were found positive for haemogregarines, namely Afrixalus fornasinii (6/14), Hyperolius argus (2/39), and Hyperolius marmoratus (12/74). Based on morphological characteristics, morphometrics and molecular findings three new haemogregarine species, Hepatozoon involucrum Netherlands, Cook and Smit n. sp., Hepatozoon tenuis Netherlands, Cook and Smit n. sp. and Hepatozoon thori Netherlands, Cook and Smit n. sp., are described from hyperoliid hosts. Furthermore, molecular analyses show anuran Hepatozoon species to be a separate monophyletic group, with species isolated from African hosts forming a monophyletic clade within this cluster.


Introduction
Haemogregarines (Apicomplexa: Adeleiorina) are heteroxenous, intraerythrocytic or intraleucocytic parasites, infecting a broad range of vertebrate intermediate hosts including amphibians, reptiles, fishes, birds and mammals. These parasites are possibly transmitted by an equal diversity of haematophagous invertebrate definitive hosts or vectors, such as dipteran insects, ticks, mites, leeches and even gnathiid isopods (see Smith, 1996;Davies and Johnston, 2000;Curtis et al. 2013). Haemogregarines are currently divided into four families (Barta et al. 2012), namely Dactylosomatidae Jakowska and Nigrelli, 1955, Haemogregarinidae Léger, 1911, Hepatozoidae Miller, 1908, and Karyolysidae Labbé, 1894 Within the Hepatozoidae, Hepatozoon Miller, 1908 is characterized by the presence of gamonts in erythrocytes or leucocytes, with no merogonic division occurring in the peripheral blood of the vertebrate host. Furthermore, Hepatozoon species are characterized by the pairing (syzygy) of gamonts in the definitive invertebrate host or vector following a blood meal. These paired gamonts then penetrate the gut wall and enter the haemocoel where sporogonic development and ultimately the formation of large oocysts occur. These thick-walled oocysts (also known as large multisporocystic oocysts) contain sporocysts with sporozoites, the infective stages of the parasite, which emerge upon the ingestion by the intermediate vertebrate host and give rise to merogonic stages in the liver (Desser et al. 1995;Smith, 1996;Barta, 2000).
Over the past decade several phylogenetic studies on adeleorinid parasites, using 18S rDNA sequences, have provided useful insight into the evolutionary relationships of this group, as well as the better capability to distinguish between species. However, because the 18S rRNA nuclear gene is a relatively conserved marker, it shows certain nodes to be unresolved (Barta et al. 2012;Maia et al. 2012;Haklová-Kočíková et al. 2014;Cook et al. 2016). In an effort to resolve these polytomies, a new genus Bartazoon Karadjian, Chavatte and Landau 2015, was proposed for species previously regarded as belonging to Hepatozoon parasitizing reptiles, amphibians, marsupials, birds and rodents, and was proposed to be transmitted solely by biting insects (Karadjian et al. 2015). However, the suggested life history of certain species within the proposed genus such as Hepatozoon fitzsimonsi (Dias, 1953) do not conform to the recommended characteristic defining Bartazoon (see Cook et al. 2014;Karadjian et al. 2015). Also as pointed out by Maia et al. (2016b), it is possible that Hepatozoon perniciosum Miller, 1908, the type species of the genus Hepatozoon, may, in fact, form part of the newly proposed genus Bartazoon, as most other rodent haemogregarine species do. Furthermore, increased work on the phylogenetic relationships of the haemogregarines continues to identify new genetic lineages, showing that Bartazoon is not a well-supported monophyletic group (Tomé et al. 2016;Maia et al. 2016a). Thus, to revise the deeper taxonomy (family and genus level) of haemogregarines based on their phylogenetic affinities and life histories, more studies using faster-evolving markers such as mitochondrial genes (e.g. Leveille et al. 2014), elucidating life cycles, and building larger datasets are necessary. Therefore, as suggested and used by Maia et al. (2016b) we will continue to refer to species parasitizing anuran hosts as species of Hepatozoon and not Bartazoon.
Prior to the study of Netherlands et al. (2014a) all the African anuran Hepatozoon species descriptions, ranging from the early 1900s till the late 1970s, were solely based on the morphology of the peripheral blood gamont stages. Unfortunately, many of these descriptions were scantily illustrated and incomplete, with almost 60% of the species described from the same host (Sclerophrys regularis) and in more or less the same geographical area (see Netherlands et al. 2014a, b). Thus many of these species may later need to be synonymized once more advanced and standardized methods are used to characterize these haemogregarines. In South Africa only five studies on amphibian haemogregarines have been carried out (Laveran, 1905;Fantham et al. 1942;Netherlands et al. 2014a, b;Netherlands et al. 2015). From these, only a single study was a multispecies haemoparasite survey across different anuran families (Netherlands et al. 2015), and although in that study several different haemogregarines were observed in anurans, only one hyperoliid species, Hyp. marmoratus (as mentioned above) contained a Hepatozoon species, which was not identified to species level.
Thus the objectives of the current study were (1) to establish which hyperoliid frog species in northern KZN, South Africa, contain haemogregarines.
(2) to determine the species diversity of the haemogregarine parasites observed.
(3) to ascertain if any of the haemogregarines found were previously described or reported species and (4) to compare any parasites characterized in the current study with available molecular data for anuran haemogregarines in order to determine their phylogenetic relationships.

Frog collection and study area
A total of 225 individuals representing nine hyperoliid species, were collected from several localities throughout northern KwaZulu-Natal, South Africa ( Fig. 1), following the collection methods described in Netherlands et al. (2015). Frogs were identified using Du Preez and Carruthers (2009), and identifications were confirmed by one of the authors of this guide (LdP). After processing, all specimens were released at the site of capture. This study received the relevant ethical approval from the North-West University's AnimCare ethics committee (ethics number: NWU-00372-16-A5).

Processing of samples and light microscopy screening
Blood (>0·1 mL) was taken from each frog via cardiac or femoral venipuncture and thin blood smears prepared on clean glass slides, air-dried, fixed and stained using Giemsa-stain (FLUKA, Sigma-Aldrich, Steinheim, Germany). The remaining blood was preserved in 70% ethanol for molecular work (ratio 1 : 15). Stained blood smears were screened at 1000× and images captured and measured using the imaging software NIS Elements Ver. 4 as described by Netherlands et al. (2015). Fifty mature gamonts were measured per Hepatozoon species. Measurements comprised the parasite's length (including recurved tail when present) and width within its parasitophorous vacuole (PV), and the parasite's nucleus length and width. Measurements of the PV length and width, and the length from mid nucleus to both anterior and posterior end of the parasite were also taken. Parasitaemia was calculated per 100 erythrocytes, with ∼10 4 erythrocytes examined per blood smear, following previous methods (see Cook et al. 2015a).
Conditions for PCR were as follows: initial denaturation at 95°C for 3 min, followed by 35 cycles, entailing a 95°C denaturation for 30 s, annealing at 61°C for 30 s with an end extension at 72°C for 2 min, and following the cycles a final extension of 72°C for 10 min. PCR reactions were performed with volumes of 25 µL, using 12·5 µL Thermo Scientific DreamTaq PCR master mix (2×) (final concentration: 2× DreamTaq buffer, 0·4 mM of each dNTP, and 4 mM MgCl 2 ), 1·25 µL (10 µM) of each of the primer sets mentioned above, and at least 25 ng DNA. The final reaction volume was made up with PCR-grade nuclease-free water (Thermo Scientific). Reactions were undertaken in a Bio-Rad C1000 Touch™ Thermal Cycler PCR machine (Bio-Rad, Hemel Hempstead, UK). Resulting amplicons were visualized under ultraviolet light on a 1% agarose gel stained with gel red using a Bio-Rad GelDoc™ XR+ imaging system (Bio-Rad, Hemel Hempstead, UK). PCR products from each sample were sent to a commercial sequencing company (Inqaba Biotechnical Industries (Pty) Ltd, Pretoria, South Africa) for purification and sequencing in both directions. Resultant sequences were assembled, and chromatogram-based contigs were generated and trimmed using Geneious R9·1 ( against previously published sequences using the Basic Local Alignment Search Tool (BLAST) (Altschul et al. 1990).
Sequences obtained in the current study were deposited in the NCBI GenBank database under the following accession numbers [GenBank: MG041591-MG041605].
For comparison, all 18S rDNA sequences of anuran haemogregarines, longer than 1500 nt (comprising species of Hepatozoon, Hemolivia, Babesiosoma and Dactylosoma) as well as Hepatozoon sipedon Smith, Desser and Martin, 1994, [GenBank: JN181157] from the snake Nerodia sipedon sipedon, were downloaded from GenBank and aligned to the sequences generated in the current study. Hepatozoon sipedon was selected as it was shown by Barta et al. (2012) to be sister to Hepatozoon catesbianae (Stebbins, 1904) and Hepatozoon clamatae (Stebbins, 1905), at that point the only two species of Hepatozoon of frogs for which 18S rDNA sequences were available. Furthermore, H. sipedon first makes use of a frog intermediate host in which tissue development occurs before transmission to its second intermediate snake host (see Smith et al. 1994). Thus all species included in the analysis have an anuran host in their life cycle. Although there are other sequences available from a Hepatozoon species characterized from the anurans Pelophylax perezi [GenBank: KF733812] and Leptodactylus chaquensis [GenBank: JX987775], from the Azores in the North Atlantic Ocean, and Pantanal, Brazil respectively, they were not added to our analysis because these concerned shorter fragments (see Harris et al. 2013;Leal et al. 2015). Babesiosoma stableri Schmittner and McGhee, 1961 [GenBank: HQ224961] and Dactylosoma ranarum Lankester, 1871 [GenBank: HQ224957; HQ224958] were chosen as the outgroup, as they were shown by Barta et al. (2012) to belong to a sister group to our current ingroup. Sequences were aligned using the MUSCLE alignment tool (Edgar, 2004) under the default settings and implemented in Geneious R9·1. The alignment consisted of 14 sequences with a 1497 nt conserved region selected using the Gblocks 0·91b server (Castresana, 2000). To infer phylogenetic relationships both Bayesian inference (BI) and Maximum likelihood (ML) methods were used. The BI analysis was performed using MrBayes 3·2·2 (Huelsenbeck and Ronquist, 2001) and the ML analysis was performed using RAxML Ver. 7·2·8. (Stamatakis, 2014) both implemented from within Geneious R9·1. Prior to the analyses, a model test was performed to determine the most suitable nucleotide substitution model, according to the Akaike information criterion using jModelTest 2·1·7 (Guindon and Gascuel, 2003;Darriba et al. 2012). The model with the best AICc score was the Transitional model (Posada, 2003) with estimates of invariable sites and a discrete Gamma distribution (TVM + I + Γ). However, this model was substituted by the General Time Reversible (Tavaré, 1986) model (GTR + I + Γ) in MrBayes and in RAxML, as this was the next model available with the best AICc score. For the BI analysis, the Markov Chain Monte Carlo (MCMC) algorithm was run for 10 million generations, sampling every 100 generations, and using the default parameters. The first 25% of the trees were discarded as 'burn-in' with no 'burn-in' samples being retained. Results were visualized in Trace (implemented from within Geneious R9·1), to assess convergence and the burn-in period. For the ML analysis, nodal support was assessed using 1000 rapid bootstrap inferences. Model-corrected (TVM + I + Γ) genetic distances were calculated in PAUP version 4.0a152 (Swofford, 2002), with the assumed proportion of invariable sites = 0·598 and the gamma shape parameter = 0·775.

Remarks:
Based on the morphology and morphometrics of peripheral blood stages in Hyp. marmoratus, H. involucrum n. sp. does not conform morphologically to any of the 16 currently recognized Hepatozoon species in African anurans. The only other named species infecting a member of the Hyperoliidae, is H. hyperolii, and can be distinguished from H. involucrum n. sp. based on the shape of the former parasite's gamont. The gamont of H. hyperolii is cylindrical with rounded ends and a long recurved tail folded onto itself in the absence of a prominent PV (see Fig. 6A-C). In contrast the gamont of H. involucrum n. sp. has an elongated and encased gamont, which is often recurved at both the anterior and posterior poles. The mean length and width of H. involucrum n. sp., which includes the parasite's PV, is 18·3 µm long × 8·3 µm wide. Although these mean length measurements do overlap with several species namely, H. faiyumensis, H. francai, H. moloensis and H. neireti, the mean width in combination with the length of these species does not conform. Overall the gamont measurements of H. involucrum n. sp. compare closest to those of H. moloensis (18·8 µm long × 7·8 µm wide), which was described from an unidentified Sclerophrys species in Molo, Kenya (see Hoare, 1920). However, the oval shape, recurved tail and absence of a PV in H. moloensis are distinctive and distinguishable from H. involucrum n. sp. as described above. Similarly, these distinctive characteristics of H. involucrum n. sp, which differentiate it from H. moloensis, also differentiate it from other African anuran species of Hepatozoon.
In South Africa a Hepatozoon species corresponding morphologically to H. involucrum n. sp. was reported from the same host and area in an anuran biodiversity blood parasite survey by Netherlands et al. (2015), however, this parasite was not formally described or named (see Netherlands et al. 2015, Fig. 2D).
For the phylogenetic analyses the topologies of both the BI and ML trees were similar. The analyses showed Hemolivia stellata   (Hoare, 1932), described from an unidentified Hyperolius species in Uganda. Redrawn and adapted from Hoare (1932). (D and E) Unnamed Hepatozoon species reported in Hyperolius puncticulatus, from Amani, Tanzania. Redrawn and adapted from Ball (1967). Scale bar: 10 µm.

1046
Edward C. Netherlands et al.
[GenBank: KP881349] as a well-supported sister taxon to the Hepatozoon species cluster, with H. sipedon [GenBank: JN181157] shown to be a sister species to a well-supported monophyletic clade comprising Hepatozoon species isolated from anuran hosts. The Hepatozoon species isolated from African and North American anurans formed two well-supported monophyletic clades, respectively, and were separate from the European species H. magna [GenBank: HQ224960]. The African Hepatozoon clade represents a polytomy with H. involucrum n. sp. and H. tenuis n. sp., forming a well-supported monophyletic clade and H. ixoxo and H. theileri, forming a poorlysupported clade, nested within this polytomy and separate to H. thori n. sp. (Fig. 7).

Discussion
In the present study, we screened the peripheral blood of 225 individual frogs from nine species within the Hyperoliidae. Six species (A. aureus, A. delicates, Hyp. tuberlinguis, Hyp. pusillus, K. senegalensis and P. maculatus), totalling 205 specimens were found negative for haemogregarine parasites. Only 20 frogs from three species were found positive, namely A. fornasinii (6/14), Hyp. argus (2/39), and Hyp. marmoratus (12/74). Morphological and molecular data indicate that the haemogregarines parasitising these hosts represent three distinct species of Hepatozoon, herein described as H. involucrum n. sp. parasitising Hyp. marmoratus; H. tenuis n. sp., parasitising A. fornasinii, Hyp. argus and Hyp. marmoratus; and H. thori n. sp. parasitising Hyp. argus and Hyp. marmoratus. Mature gamonts of H. involucrum n. sp. are characterized by the prominent PV encircling the large gamont, as well as the recurved ends of both poles of the gamont. When compared with H. tenuis n. sp., the overall appearance and characteristics are similar, except for a difference in size of the gamont and PV. The interspecific divergence between these two species is 1·0%. This has been shown in several studies to correspond to species-level differences in haemogregarines and for the slow evolving 18S rRNA marker (see Barta et al. 2012;Cook et al. 2015b;Borges-Nojosa et al. 2017). Hepatozoon thori n. sp. can be distinguished from both H. involucrum n. sp. and H. tenuis n. sp. based on the distinctive hammer-like shape of the gamont's PV. The interspecific divergence between H. thori n. sp., H. involucrum n. sp. and H. tenuis n. sp. was 2·0% and 1·8%, respectively.
The only other named species of Hepatozoon infecting a member of the Hyperoliidae is H. hyperolii described in an unidentified Hyperolius species by Hoare (1932), this parasite being vermicular in shape and folding over on itself within its host erythrocyte (see Fig. 6A-C) and therefore does not conform to any of the Hepatozoon species of the present study. However, Ball (1967) reported a second, but unnamed species in Hyperolius puncticulatus from Amani, Tanzania, and this species conforms both in size and shape to H. thori n. sp. (see Fig. 6D and E). In the current study, we propose that these two species are the same, despite parasitizing different hosts and possibly being geographically isolated. However, to confirm this, molecular data for this species from Amani, Tanzania is required.
In our phylogenetic analysis, Hepatozoon species isolated from anuran hosts formed a well-supported monophyly, separate to other closely related species of Hepatozoon. Furthermore, the African clade formed a monophyly, with H. thori n. sp. separate from the other species within this clade. Hepatozoon involucrum n. sp. and H. tenuis n. sp. form a well-supported monophyletic clade nested within the larger African clade. With an interspecific divergence of 1·0% (model-corrected distance), these two species are closely related, which concurs with their close morphological resemblance. Hepatozoon ixoxo and H. theileri form a less well supported (0·80/75) monophyletic group. The BI statistical information for the bipartitions of this group showed that apart from the 80% probability support, only 13% included H. thori n. sp. as part of this clade and 9% showed H. theileri formed a clade with H. involucrum n. sp. and H. tenuis n. sp., thus explaining the low support of this group. Furthermore, H. ixoxo and H. theileri differ considerably in morphological structure (see Conradie et al. 2017), and if compared with the phylogenetic and morphological relationship of H. involucrum n. sp. and H. tenuis n. sp. (as mentioned above), the former two species are not expected to be sister species. This underlines the importance of increased taxon sampling for these parasites, as the addition of more species to this dataset could result in better-supported clades and the polotomy of the African clade could be resolved. Additionally, fasterevolving markers (e.g. mtDNA) may further explain the biogeography and evolutionary history of these species globally. However, to date, only one haemogregarine, H. catesbianae isolated from the frog Rana catesbeiana has mtDNA sequence data available (see Leveille et al. 2014). Although these markers (mtDNA) may be complementary in providing an evolutionary perspective among these parasite groups, a lot more data are required if we want to use similar sized datasets such as those available for 18S rDNA sequences for haemogregarines, especially in terms of vertebrate host diversity (amphibians, reptiles, fishes, birds and mammals) and geographical distribution. This study highlights the importance of screening anurans from different families and genera in an effort to increase the known biodiversity of these parasites and types of hosts they infect. This study also shows the significance of providing detailed descriptions or reports of species, localities and host records, as we were able to link a species reported by Ball (1967) with H. thori n. sp. in the current study based on the morphological details he provided. However, although morphological details are important, the use of them in combination with molecular tools provides a richer dataset with which to work, allowing us to infer historical relationships. Furthermore, if molecular data were available for all the currently recognized species of Hepatozoon, those with close morphological characteristics could be correctly distinguished. This stresses the importance of using both of these techniques in combination when describing species, and where possible to provide molecular data for already described species. Future research should, when possible, include faster-evolving genes, identification of possible definitive hosts or vectors and life cycle studies. Fig. 7. Consensus phylogram of anuran haemogregarines based on 18S rDNA sequences. Tree topologies for both Bayesian inference (BI) and Maximum Likelihood (ML) analyses were similar (represented on the BI tree), showing the phylogenetic relationships for H. involucrum n. sp., H. tenuis n. sp., and H. thori n. sp. (represented in bold), compared to other species of anuran Hepatozoon (with the exception of Hepatozoon sipedon), Hemolivia, and three species from the Dactylosomatidae as outgroup. Clades that neither produced 0·80 posterior probability (BI) or 70 bootstrap (ML) nodal support values were collapsed. Black circles represent 100% support for both BI/ML. The scale bar represents 0·02 nucleotide substitutions per site.