Molecular footprint of parasite co-introduction with Nile tilapia in the Congo Basin

Nile tilapia, one of the most popular aquaculture species worldwide, has been introduced into the Congo Basin several times for aquaculture purposes. Previous studies based on morphological features showed that some of the monogenean gill parasites were co-introduced with Nile tilapia and some spilled over to native Congolese cichlids. In this study, we genetically investigated the co-introduced monogeneans of Nile tilapia from three major parts of the Congo Basin: Upper, Middle and Lower Congo. We sequenced 214 specimens belonging to 16 species of Monogenea, collected from native and introduced tilapia species from Congo, Madagascar and Burundi. We evaluate their position in a phylogeny including 38 monogenean species in total. Our results confirm the co-introductions in the Congo Basin and suggest one unreported parasite transmission from introduced Nile tilapia to native Mweru tilapia in Upper Congo, which was undetectable with a morphological study alone. Shared parasite COI haplotypes between Madagascar and the Congo Basin illustrate how anthropogenic introduction events homogenize parasite communities across large geographical distances and thereby disrupt isolation by distance patterns. Contrary to our expectation, the parasite populations co-introduced in the Congo Basin reveal a high COI diversity, probably resulting from multiple Nile tilapia introductions from different geographic origins. Additionally, we tested the barcoding gap and the performance of mitochondrial COI and nuclear ribosomal ITS-1, 28S and 18S markers. We found a significant barcoding gap of 15% for COI, but none for the other markers. Our molecular results reveal that Cichlidogyrus halli, C. papernastrema, C. tiberianus, C. cirratus and C. zambezensis are in need of taxonomic revision.


Introduction
Human-mediated species translocations are ubiquitous and form a major challenge to global biodiversity today (Pimentel et al., 2001;Tollefson, 2019). As a result of invasive species, the abundance of native plants, insects and other animals has fallen by an estimated 20% since 1900 (Tollefson, 2019). Of these translocated species, tilapias are among the most widely introduced aquaculture species and are now found in over 140 countries (Deines et al., 2016). Cultured tilapia comprises species of African cichlids mainly from Oreochromis Günther 1889, Coptodon Gervais 1848, Tilapia Smith 1840 and Sarotherodon Rüppel 1852 and is estimated to produce over 5 million tonnes yearly globally (Deines et al., 2016;FAO, 2017). Of these, Nile tilapia, Oreochromis niloticus (Linneaus, 1758), is the most popular and makes up over 75% of the cultured tilapias in Sub-Saharan Africa (FAO, 2017). This species has been deliberately introduced in many African countries, despite the presence of ample native tilapias. In the DRC, culture of non-native Nile tilapia started after the Second World War (Micha, 2013). Production spread rapidly throughout the country until 1960 after which production almost halted (Micha, 2013). In 1964, the first feral Nile tilapia population was found in the North-East of the basin (Thys van den Audenaerde, 1964). From 1996 onwards, Nile tilapia production increased steadily to 3000 tons/year in 2010 (El-Sayed, 2013;Toguyeni, 2004).
A previous study identified parasite co-introductions with Nile tilapia and spillover to native cichlids in the Congo Basin (Jorissen et al., 2020). The respective parasites belong to Cichlidogyrus Paperna 1960 andScutogyrus Pariselle &Euzet, 1995. They are monogenean flatworms (Dactylogyridea Bychowksy, 1937) that mainly infect the gills of African mainland cichlids (Pariselle & Euzet, 2009), where they feed on mucus, skin and possibly blood of the hosts (Gonçalves et al., 2009). Currently, 130 valid species of Cichlidogyrus and seven of Scutogyrus have been described (WoRMS, 2021) and DNA sequences of 31 and three, respectively, are available on GenBank (NCBI, 2021). These sequences are generally limited to partial fragments of 28S or 18S + ITS1 (internal transcribed spacer 1 region) rDNA. The genes coding for 18S and 28S are the most conservative markers currently available (Vanhove et al., 2013) and are used for phylogenetic reconstructions (Mendlová & Šimková, 2014;Mendlová et al., 2012;Pouyaud et al., 2006;Wu et al., 2007). However, there is a lack of more variable molecular markers for barcoding flatworms (Littlewood, 2008;Moszczynska et al., 2009;Vanhove et al., 2013) and a lack of sequenced species. For example, only five Cichlidogyrus and Scutogyrus species have been sequenced for the mitochondrial cytochrome c oxidase subunit I (COI) gene.
The first goal of this study is to aid in filling this knowledge gap by adding sequences of Congolese native and introduced species and to evaluate the position of these previously unsequenced species in a phylogenetic tree. Additionally, we will discuss marker performance for mitochondrial COI, ITS-1 rDNA, 18S rDNA and 28S rDNA.
Secondly, because of this molecular knowledge gap, these monogeneans are predominantly diagnosed based on the morphology of their sclerotized hard parts from the reproductive organs and the attachment organ (haptor). However, out of the five morphospecies of Cichlidogyrus and one of Scutogyrus that were co-introduced with Nile tilapia into the Congo Basin (Jorissen et al., 2020), we suspect that three are in need of taxonomic revision: Cichlidogyrus halli (Price and Kirk 1967), C. cirratus Paperna 1964 and C. tilapiae Paperna 1960. A subspecies of C. halli has been described in the past (Paperna, 1979), but was later revoked (Pariselle & Euzet, 2009). Additionally, we discovered a morphotype of C. halli native to Bangweulu-Mweru, in the Upper Congo Basin, infecting the native Oreochromis mweruensis Trewavas, 1983(Jorissen et al., 2018a. However, genetic information on this morphotype is missing and would be helpful to decide on its taxonomic status. Pouyaud et al. (2006) found three molecular variants of C. halli based on 18S and 28S rDNA, but the low molecular variation did not indicate the need of taxonomical revision at the time. The morphological and molecular variation within C. cirratus could also be underestimated as this species was recently reported in new areas and on new hosts (Jorissen et al., 2020;Zhang et al., 2019 andunpublished data). Finally, Pouyaud et al. (2006) suggested that C. tilapiae might constitute a species complex of morphologically closely resembling taxa. They found larger intraspecific distances based on 18S and 28S rDNA within C. tilapiae than between some specimens of C. tilapiae and C. cubitus Dossou 1982. Our second goal is therefore to evaluate the species status of these three species through molecular data.
Thirdly, genotyping the co-introduced monogeneans of Nile tilapia could offer a more precise picture of the invasive history of the fish. Indeed, parasites can, due to their faster generation time compared to their host, shed more light on the evolutionary and biogeographical history of their hosts, as predicted by the magnifying glass concept (Nieberding & Olivieri, 2007). Moreover, genotyping can reveal biological phenomena like hybridization, as was the case for catfish parasites, which in turn implied historical contact between the respective host species that currently live in allopatry (Barson et al., 2010). Also the lack of genetic variation can be informative on host biogeography (Hayward et al., 2003). Gyrodactylus anguillae Ergens, 1960 collected from eel populations from three different continents share identical ribosomal DNA sequences, as the result from recent intercontinental live eel trade (Hayward et al., 2001).
In this study, we focus on the Congo Basin because it is the largest African basin where Nile tilapia was introduced and because the country has a historical tradition of tilapia culture (Welcomme, 1988;Micha, 2013). Additionally, molecular data from monogenean parasites within the Congo Basin, apart from Lake Tanganyika, is largely lacking. Our expeditions took place in Upper Congo (Bangweulu-Mweru) in the southeast of the basin, Middle Congo around Kisangani (DRC) and Lower Congo downstream of Boma and the tributaries of the Congo around Mbanza-Ngungu (DRC). The boundaries between these three parts are Pool Malebo around Kinshasha and Boyoma Falls upstream from Kisangani ( Fig. 1; Alter et al., 2015). We included parasite populations from introduced Nile tilapia from Madagascar because Nile tilapia is wellestablished there, and mostly the same monogenean species have been co-introduced there as in the Congo Basin (Šimková et al., 2019). Lastly, we sampled a native population of Nile tilapia from a pool next to Lake Tanganyika (Burundi) because it is geographically the closest native population of Nile tilapia to the DRC. According to the concept of isolation by distance (IBD), the genetic similarity between populations should decrease with increasing geographic distance (Avise, 1994;Poulin & Krasnov, 2010). However, introduction events can blur this signal and can also lead to lower genetic diversity in introduced populations compared to their source populations. This can lead to potential founder effects (Avise, 1994;Mayr, 1942). Therefore, we expect a low diversity in the introduced parasite populations from Congo and Madagascar, compared to the native parasite population from Burundi. Also, if Nile tilapia in Congo and Madagascar originate from a common source of introduction, they will share parasite haplotypes and no signal of IBD will be found.

Data collection
A total of 214 specimens of parasites belonging to 16 described and one undescribed species, of which 15 of Cichlidogyrus and two of Scutogyrus, were collected from seven host species (Table 1) (Jorissen et al., 2018b); one to Burundi in September 2013 (Rahmouni et al., 2017); and one to Madagascar in April 2016 (Šimková et al., 2019). Fish from Lake Kariba in Zimbabwe were caught in September 2016. Fish were collected in the wild, from aquaculture stations or bought at local fish markets and killed with an overdose of MS222 (tricaine mesylate). Specimens and sample localities are listed in Addendum 1 and shown in Fig. 1. The gill arches of the right side were dissected in the field and stored in pure ethanol. The left side of the fish was left intact for ichthyological research. Parasites were isolated in the lab using an entomologi-  (Teugels & Thys van den Audenaerde, 2003;Trewavas & Teugels, 1991). Localities 1-4 are in Lower Congo (1, Tondé; 2, Monzi; 3, Ndimba Leta; 4, Pond near Kila Kindinga), 5 in Middle Congo (Djugu-Djugu ponds), 6-12 in Upper Congo (6, Futuka Farm; 7, Lake Kipopo; 8, Zoo Lubumbashi; 9, Bumaki Farm; 10, Kiswishi River off Futuka Farm; 11, Luapula River off Kashiobwe; 12, Lake Tshangalele); 13 Pond adjacent to Lake Tanganyika, Burundi; 14-16 are in Madagascar (14, Lake Andranotapahina; 15, Ankafarantsika National Park, Lake Ravelobe; 16 Anjingo River near Antsohihy) and 17 is in Lake Kariba, Zimbabwe. Pool Malebo and Boyoma Falls (black bar) divide the Congo Basin in the above-mentioned sections. More info about the sampling localities in Addendum 1 cal needle and a Wild M5 stereomicroscope (Heerbrugg, Switzerland). Parasites were cut in half with a scalpel; the anterior body part was fixed in Hoyer's medium (Humason, 1979) and sealed on a slide with glyceel (Bates, 1997) for morphological identification. The pictures of C. cf. halli 'Burundi' were taken with a Zeiss Axio Imager Z1 microscope at a magnification of 100 × (oil immersion, 10 × ocular) under differential interference contrast, with an AxioCamMR3 camera and AxioVision v.4.2.8 software. The posterior body part of the parasite was placed in an Eppendorf tube filled with 180 µl of T1 buffer, Nucleospin Kit, Macherey Nagel, Düren, Germany, for DNA extraction and stored at − 21 °C if extraction was not carried out immediately (see Addendum 1 for the collection numbers of fish hosts, parasite vouchers and GenBank accession numbers of the parasite DNA sequences generated in this study). Parasite slides and fish hosts were both stored in the collections of the Royal Museum for Central Africa, Tervuren, Belgium (RMCA). Fish from the Madagascar expedition and Upper Congo

Phylogenetic analyses
A total of 38 species were included to build the phylogeny from ribosomal markers; for 12 of these, we present the first sequences. All sequences are submitted to GenBank and accession numbers are available in Addendum 1. Our tree was rooted on Cichlidogyrus pouyaudi Pariselle and Euzet (1994) as previous phylogenetic research found that it is the most basal taxon of the group (Mendlová & Šimková, 2014;Mendlová et al., 2010;Pouyaud et al., 2006;Wu et al., 2007). All sequence chromatograms were visually inspected for sequencing errors and blasted individually on the NCBI website (http:// www. ncbi. nlm. nih. gov). The resulting sequences were aligned with MUSCLE (Edgar, 2004a(Edgar, , 2004b under default settings, edited in MEGA 7.0.18 (Kumar et al., 2016) and cleaned-up with Gblocks 0.91b under default parameters (Castresana, 2000;Talavera & Castresana, 2007). The 28S and 18S + ITS-1 sequences were concatenated using SequenceMatrix (Vaidya et al., 2011). To test whether a concatenation was possible we performed a partition-homogeneity test in PAUPUP 1.0.3.1 (Swofford, 2003). From some specimens, either 28S or 18S + ITS-1 was successfully amplified and these sequences were aligned with the concatenated sequences to include the maximum number of species and specimens. JModelTest2 (Darriba et al., 2012) was used to determine the most appropriate model of nucleotide evolution with default parameters, employing the Akaike Information Criterium (AIC). The GTR + G + I model was found as the optimal model for the concatenated dataset with gamma being 0.65 and the proportion of invariant sites 0.29. For the COI dataset TIM1 + G + I was the optimal model, but we used GTR + G + I as it was the most similar model being available in MrBayes (Ronquist & Huelsenbeck, 2003) and RAxMLHPC2 (Stamatakis, 2014). Gamma was 0.2 and the proportion of invariant sites 0.2. Maximum Likelihood (ML) analyses were carried out with RAxML-HPC2 (Stamatakis, 2014) via the CIPRES webserver with the GTR + G + I model and 1000 iterations. Branch support was estimated by bootstrapping with 1000 replicates. The Bayesian inference of phylogeny (BI) was performed in MrBayes 3.2.7 (Ronquist & Huelsenbeck, 2003) with 4.000.000 generations sampled every 1000 generations and four chains (three cold and one hot). One fourth of the topologies were discarded as burn-in as the standard deviation of split frequencies was below 0.01 for both the concatenated dataset and for the COI dataset. Figtree 1.4.4. (Rambaut, 2018) and TreeGraph 2 (Stöver & Muller, 2010) were used to visualize and edit the trees. Additionally, we identified phylospecies and evolutionary species with a coalescent tree-based Poisson tree process model and its Bayesian implementation (bPTP, Zhang et al., 2013) on the web server (http:// speci es.h-its. org/ ptp/) under default parameters and 0.2 burnin.

Results
The sequencing of the partial COI gene resulted in 67 sequences of 314 bp of 13 species after trimming and clean-up in Gblocks; none showed stop codons. We opted for the nested fragment to include the most specimens in the analyses. The partition homogeneity test allowed the concatenation of 18S, ITS-1 and 28S fragments with a p-value of 0.95 (anything above 0.05 was sufficient for concatenation). Our concatenated sequences consisted of 1362 bp divided in fragments of 429, 275 and 658 bp of 18S, ITS-1 and 28S rDNA, respectively. We generated 31 sequences of 10 morphospecies for 18S + ITS-1 rDNA and 85 sequences of 17 morphospecies for 28S rDNA. The ITS-1 fragment was the most variable (60.2% maximum variance, Addendum 2) followed by COI (44.3%, Addendum 3), 28S (11.9%, Addendum 4) and 18S (4.8%, Addendum 5-7). For the COI alignment, the topology of both the ML and Bayesian trees was not well supported, especially the deeper branches lacked support. We therefore used the COI dataset to compare haplotypes through median-joining haplotype networks ( Fig. 2a-c). For C. halli the genetic distance between specimens was too large for a median-joining haplotype network. The COI dataset included 15 morphospecies and was divided in 18 species by ASAP and between 22 and 28 by bPTP. The concatenated dataset included 37 morphospecies and was divided in 34-38 species by ASAP and between 22 and 47 by bPTP. The species divisions in the COI dataset between ASAP and bPTP corresponded well and were largely well-supported with bPTP. However, for the concatenated ribosomal dataset with bPTP, the species divisions were largely unsupported.

Co-introduced parasites of Nile tilapia
Four of the co-introduced species of Cichlidogyrus, C. halli, C. thurstonae, C. tilapiae, and C. sclerosus showed instraspecific variation in COI (Addendum 3). The variation within the halli group was so high (0.6-30.1%, 104 polymorphic sites, Pi = 0.1463) that it did not allow a median-joining haplotype network. Within C. halli, we observed nine groups, each separated by more than ten to 30 mutations. The locality Bumaki in the Upper Congo has the most haplotypes of C. halli and these cluster in three groups. Furthermore, the native C. cf. halli 'Burundi' and C. halli morphotype 2 sensu (Jorissen et al., 2018a) showed the highest distance to the introduced specimens (0.2-3.2% 18S, 3.2-10.1% ITS-1, 1.1-1.9% 28S, 24.7-36% COI), with large distances within these native representatives for COI also, but not for the rDNA fragments (0.2% 18S, 2.4% ITS-1, 1% 28S, 23.1-29.5% COI). The COI sequence of C. cf. halli 'Burundi' has three non-synonymous mutations compared to introduced C. halli and the two specimens of C. halli morphotype 2 have two and four non-synonymous mutations respectively. One non-synonymous mutation was shared between these three native specimens and two non-synonymous mutations between the two specimens of C. halli morphotype 2. Within the introduced specimens of C. halli in Congo, three of Upper Congo and one from Middle Congo shared the same nonsynonymous mutation. Other co-introduced species, C. sclerosus (1-2.7% COI, 16 polymorphic sites, Pi = 0.01776), C. tilapiae (0.3-3.4% COI, 10 polymorphic sites, Pi = 0.01409) and C. thurstonae (0.3-3.4% COI, 13 polymorphic sites, Pi = 0.01706) show more modest variation (Addenda 3, 6). For C. sclerosus, each locality was characterized by a unique parasite haplotype, with the highest variation within Upper Congo. Only in one locality in Upper Congo a nonsynonymous mutation was found. The haplotype from the Middle Congo was more closely related to the one from Madagascar than to the haplotypes from Upper Congo (Fig. 2a). For C. thurstonae on the other hand, the highest diversity was found in Lower Congo and haplotypes are shared by parasites from Madagascar and Upper Congo (Fig. 2b). Specimens from the Upper and Middle Congo cluster furthest apart, each separated by at least five mutations respectively from the nearest C. thurstonae specimen from Lower Congo, of which three were non-synonymous. Thus, the three regions of the Congo Basin do not share any C. thurstonae haplotype. This in contrast to C. tilapiae, where apart from a few unique sequences, haplotypes are shared between the Lower and Middle Congo and between Lower Congo and Madagascar (Fig. 2c). The native parasite specimens from Burundi were clearly distinct from the introduced specimens in Congo and Madagascar. However, the amino acid sequence of these Burundese specimens is identical to all other specimens of C. tilapiae, except for one specimen from Middle Congo, which has two non-synonymous mutations.

Phylogenetic relationships among native and introduced tilapia parasites and evaluation of species status
A Bayesian phylogram constructed with 28S and 18S + ITS-1 sequences of Cichlidogyrus/Scutogyrus representatives of native and introduced tilapia parasites and sequences from GenBank is shown in Fig. 3. The topologies of the BI and ML trees were very similar (ML tree not shown). Minor differences were on the level of poorly supported clades (bootstrap support < 70), or unresolved intraspecific relationships. The basal topology is unclear from our analyses as we observe six monophyletic groups, between which the relations are unresolved. To facilitate the interpretation of the phylogram, we named each of these monophyletic groups after the oldest described species within it, following Pouyaud et al. (2006). These are artificial 'species groups' and the name of these groups does not infer anything to the evolutionary relationships or taxonomic status of the taxa within.
The "halli" group has a reference sequence of C. halli from Senegal (Addendum 1) at the base followed by a polytomy of 18 haplotypes of co-introduced C. halli from Nile tilapia in the Congo basin and Madagascar (0-2.9% 18S, 0.4-7.2%ITS-1, 0-1.4% 28S 1.6-21.9% COI, see Addenda 2-7). More derived of this group are three sequences of native specimens of Nile tilapia from Burundi, Mweru tilapia from Upper Congo and a hybrid Oreochromis host (Vanhove et al., 2018). Cichlidogyrus halli forms its own clade and the closest relative was not revealed by phylogenetic analysis.
The "papernastrema" group houses apart from Cichlidogyrus papernastrema Price, Peebles and Bamford 1969 also Cichlidogyrus zambezensis (Douëllou, 1993). Our two specimens of C. papernastrema (intraspecific distance of 1.8% 28S, 29.9% COI) do not cluster together. Cichlidogyrus zambezensis, on the contrary, is monophyletic and the distance (0.6% 28S) between specimens on different host species is larger than between two specimens of the same host species (identical 28S), although they all are from the same geographic region. Furthermore, both ASAP and bPTP split C. zambezensis in two species.
The "tiberianus" group has C. cubitus at the basal position followed directly by C. tiberianus Paperna 1960 (reference sequence from Senegal). The group then splits up in three lineages, one of which includes a polytomy of 23 haplotypes of co-introduced C. thurstonae from Nile tilapia, the reference of C. thurstonae from Madagascar and C. ergensi Dossou 1982. Distances within this lineage are 0-1.7% 18S, 0-4.4% ITS-1, 0-0.6% 28S, 0.3-4.8% COI (Addenda 2-7). However, the support for this group is very low (53 posterior probability and < 50 bootstrap value). Directly related to C. thurstonae and C. ergensi are Cichlidogyrus dossoui (Douëllou, 1993) and C. tiberianus from Upper Congo. Our specimens of C. tiberianus from Upper Congo do not cluster with the reference of C. tiberianus from Senegal. The bPTP analysis confirms this result. The distance of C. tiberianus between Upper Congo and Senegal is 0.2-0.5% 18S, 8.5-9% ITS-1 and 1.4% 28S.
In the "tilapiae" group we find all sequences of C. tilapiae from both native and introduced hosts (n = 24, with p-distances ranging from 0 to 1.7% 18S, 0.7 to 2.7% ITS-1 and 0 to 1% Fig. 3 Phylogram based on Bayesian inference using a concatenated 18S, ITS-1 and 28S dataset (1362 bp). Posterior probabilities are shown above branches and bootstrap values from the maximum likelihood analyses below branches. Nodes with probabilities lower than 50 are collapsed. Newly generated sequences in this study are in bold with mention of the host species; other sequences are GenBank reference sequences listed in Addendum 1 ◂ 28S, 0.3-3.4% COI), with no apparent structure other than that the reference sequence from Senegal is the earliest diverging (Addendum 1).

Discussion
Through molecular identification, we were able to confirm the parasite introductions into the Congo Basin and Madagascar observed in Jorissen et al. (2020). Additionally, we discovered a parasite transmission (spillover) of C. tilapiae from introduced Nile tilapia to native Mweru tilapia in Upper Congo. In Jorissen et al. (2020) we considered all C. tilapiae on Mweru tilapia native to the Upper Congo, but looking at the haplotype network (Fig. 2c) we find that all sequenced specimens of C. tilapiae belong to an introduced strain, even the ones on native hosts. This is the first record of a 'cryptic invasion' that we propose. Whether the native strain of C. tilapiae persists or is completely replaced on native hosts in the Congo Basin is unknown as we did not find the native strain.

Co-introduced parasites of Nile tilapia
The COI haplotype networks constructed for the monogenean parasites visualize the high diversity for each of the three species. There are many unique haplotypes, but with some notable exceptions. For example, The Lower and Middle Congo share a haplotype for C. tilapiae, but this could possibly be explained by natural gene flow, given the connectivity of both sites despite the large distance. More striking is the sharing of parasite haplotypes between Madagascar and Congo, which are geographically far apart. Specimens of C. thurstonae from Madagascar appeared identical to a specimen from Upper Congo (Fig. 2b) whilst Madagascar and the Lower Congo also share identical C. tilapiae haplotypes (Fig. 2c). Also, the C. thurstonae haplotype from Upper Congo that is identical to the one from Madagascar is highly divergent from specimens collected in the Lower and Middle Congo (Fig. 2b). This apparent lack of isolation by distance (IBD) typically reflects recent introduction events, which blurs geographic signals, as found in the study of Hayward et al. (2001). In that study, the monogenean eel parasite G. anguillae had identical rDNA sequences (ITS-1, 5.8S, ITS-2) in North America, Europe and Australia as a result of live eel trade that started forty to fifty years ago (Hayward et al., 2001). Similarly, Gyrodactylus cichlidarum Paperna, 1968 was co-introduced into Mexico with Nile tilapia after fish introductions started in the 1940s (García-Vásquez et al., 2017). The ITS-1 sequences of G. cichlidarum specimens that spilled over to Mexican poeciliids were almost completely identical to G. cichlidarum from Nile tilapia in Ghana (García-Vásquez et al., 2017). Our results, therefore, strongly point to an identical or geographically close source of introduction of Nile tilapia in Madagascar and Nile tilapia in the Upper and Lower Congo.
The high haplotype diversity of C. sclerosus within the Upper Congo (Fig. 2a) can partly be explained by the sampling bias (relatively more specimens were sequenced from this locality), but the divergent haplotypes strongly suggest that multiple introductions have taken place in this area, from different geographic source populations. Similarly, also C. tilapiae and C. thurstonae show high haplotype diversity. For C. tilapiae, it appears that the introduced populations in Congo show a higher diversity compared to the population of C. tilapiae on native Nile tilapia in Burundi. Even though more samples are needed to validate these trends, our results appear to refute our initial hypothesis that introduced parasite populations suffer bottlenecks. Similar scenarios have been described for other biological invasions, where introduced populations could maintain a high diversity because of multiple introductions from different source populations (Genton et al., 2005;Kolbe et al., 2004), sometimes followed by intraspecific hybridization (Rosenthal et al., 2008).

Marker performance and barcoding gap
Based on our histograms (Fig. 4), we find a significant (p < 0.05) barcoding gap for COI at 15% (Fig. 4a) but none for 28S, 18S or ITS-1 (Fig. 4b-d). Visually, there is a second gap between 7 and 11% for COI (Fig. 4a) but this was not found significant by ASAP. It should be noted that our fragment of the COI gene covers just less than a quarter of the total COI gene and constitutes the most variable part (unpublished data). Additionally, the COI dataset itself was the smallest of the four markers because hardly any references were available on GenBank and the amplification success of COI was lower than that of the rDNA markers. Whether 15% is representative for Cichlidogyrus/Scutogyrus should be investigated in the future by including more species.
We can find some clues in the literature to a possible barcoding gap for the rDNA markers, since we did not find a significant one within our dataset. Within Cichlidogyrus from Lake Tanganyika, Rahmouni et al. (2022) found Fig. 4 Histogram of K2P model-corrected distances for COI, 18S, ITS-1 and 28S made with ASAP. Red line indicates a significant (P < 0.05) barcoding gap. X-axis represents the genetic distance; Y-axis the number of distances intraspecific variation of ITS-1 up to 1.1% and interspecific variation starting at 3.5%. The 18S sequences of the same species were identical and the 28S sequences differed 0.13% between the species. COI intraspecific variation amounted to 12.2%. This distance for COI roughly corresponds with what we find in our dataset, but the observed distances for the three rDNA fragments are higher in our study. Representatives of Trinigyrus Hanek, Molnar and Fernando, 1974 (Dactylogyridae: Monogenea), which infect siluriforms, have interspecific variation < 6% for COI and around 1% for 28S (Franceschini et al., 2020). In Dactylogyrus Diesing, 1850 (Dactylogyridae: Monogenea), which infect European cyprinids, the cut-off value was set at 1.4% for a fragment consisting of partial 18S, complete ITS-1 and partial 5.8S (Šimková et al., 2004). Rahmouni, Řehulková, et al. (2017) found 1% for 28S, 0.4% for 18S and 4.3% for ITS-1 of Dactylogyrus parasitizing North-African congeneric cyprinids. From all these, we learn that the barcoding gap for 28S may be around 1% and for 18S below 1%. For ITS-1, this is likely higher than 1% but most probably lower than the 15%, which we found for COI of the included species of Cichlidogyrus.
For 28S sequences specifically, we observe that they can be identical over large geographic distances (C. thurstonae, C. sclerosus and C. falcifer) between introduced and native populations (C. tilapiae) and even between species (C. berradae and C. yanni; Scutogyrus gravivaginus (Paperna & Thurston 1969), S. bailloni Euzet, 1995 andS. longicornis (Paperna &Thurston, 1969) (Addendum 4). However, whether 28S can be identical between species of Cichlidogyrus/Scutogyrus is uncertain as the obtained references from GenBank can be morphologically misidentified. New sequences of C. berradae, C. yanni, S. gravivaginus, S. longicornis and S. bailloni are needed to verify this. In other groups of closely related monogeneans, the rDNA fragments can be conserved also, between different hosts (Kmentová et al., 2016a, b) or between closely related species (Marchiori et al., 2015). Therefore, there is potential for the rDNA fragments to be used for species delineation in Cichlidogyrus/Scutogyrus, as shown for Gyrodactylus spp. (Matĕjusová et al., 2001;Mendoza-Palmero et al., 2019) and Dactylogyrus spp. (Benovics et al., 2018). However, at this time, we need to include other methods such as bPTP to interpret results from the rDNA fragments about species status. What we can conclude so far is that the conservative rDNA fragments (18S and 28S) are good for constructing higher-level phylogenies and the COI fragment is suited for population level studies and species delineation, which has also been stated in previous studies.
Cichlidogyrus halli forms a species complex (see Jorissen et al., 2018a) supported by molecular data in the present study. ASAP suggests the "halli" group to consist of at least three species: firstly, the native specimen from Burundi, secondly the native specimens from Mweru tilapia from Upper Congo together with the specimen from the O. niloticus × mweruensis hybrid from Upper Congo, and thirdly all introduced specimens of C. halli. The genetic distances within the "halli" group are indeed large (2.1% for 28S; 3.2% for 18S; 10.1% for ITS-1 and 36% for COI). This variation is higher than all other intra/interspecific boundaries stated above. The bPTP method is inconclusive for the rDNA markers, where C. halli is divided in six species, including morphotype 2 and C. cf. halli 'Burundi', but the support for these divisions is very low. The divisions in COI are better supported and correspond with ASAP.
Morphotype 2 of C. halli (sensu Jorissen et al., 2018a) from Upper Congo, as drawn and discussed by Jorissen et al. (2018a), corresponds in locality and host species with C. halli ex O. mweruensis on the tree (Fig. 3). Therefore, we suggest that morphotype 2 should be elevated to species level. Similarly, the specimens of C. cf. halli 'Burundi' from Lake Tanganyika, Burundi, belong to the third species within C. halli, as suggested by ASAP. From the same Burundese population, we found a specimen of C. cf. halli 'Burundi' with elongated and thickened hooklets pair I compared to C. halli (Addendum 10). However, we strongly belief that the species delineation within the "halli" group should be based on morphology and genetics together. Therefore, new morphological material is needed to resolve this. For species within Cichlidogyrus/Scutogyrus, the genital sclerites are important for species identification (see diagnosis in Pariselle & Euzet, 2009). Therefore, for our study, we decided to only keep the body part with the genital sclerites and use the body part with the haptor for genetic analysis (Jorissen et al., 2018a). However, recent work on Kapentagyrus and Cichlidogyrus shows that closely related species might first diverge in haptor morphology before genital sclerites (Kmentová et al., 2016a;Messu Mandeng et al., 2015). This implicates the evolution of these parasites is related strongly to microhabitat (attachment site) and host species (Messu Mandeng et al., 2015, Gobbin et al., 2020. We conclude that morphological features of the haptor are important in this complex for species delimitation. Cichlidogyrus zambezensis and C. papernastrema together form a clade. However, both species belong to different groups within the genus based on the morphology of the haptoral hooklets. Cichlidogyrus zambezensis has seven pairs of small hooklets (group A sensu Vignon et al., 2011), whilst in C. papernastrema the first pair is thick and elongated (group B sensu Vignon et al., 2011). Pariselle and Euzet (2003) suggested a division of species of Cichlidogyrus in three groups based on the morphology of haptoral hooklets (uncinuli in the source). Additionally, Vignon et al. (2011) found a high congruence between these morphological groups and the molecular phylogeny, meaning that hooklet morphology is phylogenetically constrained. However, our result suggests that the division between groups A and B might not be supported by phylogenies. This is possible because Pariselle and Euzet (2003) and Vignon et al. (2011) included a subset of species in their analyses. Vignon et al. (2011) even omitted C. arthracanthus from their analysis because it did not fit any of the three groupings. In conclusion, this could mean that firstly, haptor morphology is not as phylogenetically constrained as previously thought (see the 'halli' group above) and secondly, that the division in three groupings -whilst useful for morphological identification (see Pariselle & Euzet, 2009) -does not cover the morphological evolution of the haptor within Cichlidogyrus fully.
Both C. papernastrema and C. zambezensis have a copulatory tube with a bulbous thickening in the middle and this could be a synapomorphy for the "papernastrema" group instead of characters of haptor morphology. Other species with a bulbous thickening of the copulatory tube and thus possibly belonging to this group are Cichlidogyrus halinus Paperna 1969, Cichlidogyrus sanjeani Pariselle and Euzet 1997, Cichlidogyrus philander (Douëllou, 1993), Cichlidogyrus bulbophallus Geraerts and Muterezi Bukinga 2020, Cichlidogyrus pseudozambezensis Geraerts and Muterezi Bukinga 2020 and Cichlidogyrus ranula Geraerts and Muterezi Bukinga 2020. Within these candidate species are several representatives from Haplochromine cichlids and others from southern Africa. Cichlidogyrus zambezensis is monophyletic, but we observe variation between specimens of different host species. Douëllou (1993) reports intraspecific morphological variation between specimens from different hosts in C. zambezensis. Cichlidogyrus zambezensis is known from four cichlid hosts, belonging to three cichlid lineages (Douëllou, 1993;Jorissen et al., 2018a;Vanhove et al., 2013). Additionally, the bPTP analysis of COI splits our samples of C. zambezensis and the reference as different species. Therefore, C. zambezensis is in need of further study and might consist of multiple species. The monophyly of C. papernastrema is not supported. Even more, the genetic distance between the two specimens of C. papernastrema is larger than between this species and C. zambezensis and above 1% for 28S and 15% for COI. Therefore, it is likely that both specimens belong to different species. Jorissen et al., (2018aJorissen et al., ( , 2018b redescribed C. papernastrema and noted large variation in thickness of the copulatory tube between specimens. It would be worthwhile to check whether this variation is a good diagnostic character to delineate species in tandem with genetic distances. In the "tiberianus" group, C. tiberianus from Bangweulu-Mweru does not cluster with the reference sequence from Senegal (Mendlová et al., 2012) and the bPTP analysis suggests both sequences belong to different species. Cichlidogyrus tiberianus infects representatives of Coptodon ranging from Senegal to Zimbabwe (Douëllou, 1993;Jorissen et al., 2018a;Mendlová et al., 2012;Pariselle & Euzet, 1995, 1996, 2009). This is a native range of over 7000 km, spanning different ichthyographic provinces and bassins. The genetic distances between C. tiberianus of Senegal and Upper Congo are above 1% for 28S and well below 1% for 18S. Cichlidogyrus tiberianus requires a species status re-evaluation backed by genetic data from across its native range and different host species. Fannes et al. (2017) used SEM to investigate the sclerotized parts of C. dossoui and C. tiberianus from Upper Congo because both species are morphologically quite similar and share hosts. The COI genetic distances are smaller between C. tiberianus and C. dossoui from Upper Congo than within C. tiberianus. Therefore, it is not surprising that C. dossoui from Bangweulu-Mweru appears as the sister species to C. tiberianus from Bangweulu-Mweru.
Furthermore, in the "tiberianus" group, C. ergensi is situated within C. thurstonae, but without support (53 posterior probability and not supported in the ML analysis), thus we do not make inferences to this result. All species in the "tiberianus" group belong to group C based on the morphology of the haptoral hooklets (Pariselle & Euzet, 2003;Vignon et al., 2011), except for C. arthracanthus and C. sp. 2, which fall outside of the classification in three main groups. Here again, the division by Pariselle and Euzet (2003) is not completely supported.
Cichlidogyrus cirratus was found to be monophyletic (100 posterior probability and 90 bootstrap support value, Fig. 3). However, the branch lengths within C. cirratus are much longer than for example between the different species of Scutogyrus. The genetic distance between C. cirratus from Bangweulu-Mweru and Senegal is 0.5% 18S, 9.8% ITS-1 and 2.8% for 28S (Addenda 2, 4, 5). This indicates that our samples might represent two separate species. It is also debated whether C. cirratus and C. mbirizei Muterezi Bukinga, Vanhove, Van Steenberge, Pariselle, 2012, are conspecific (Zhang et al., 2019). Scanning electron microscopy revealed that the distinguishing characters between C. cirratus and C. mbirizei (Muterezi Bukinga et al., 2012) on specimens of C. cirratus from China (introduced) could be transformed by turning the specimens over (Zhang et al., 2019). In conclusion, we deem it likely that C. cirratus and possibly C. mbirizei consist of multiple species and that this should be investigated further genetically. Subsequently, an evaluation of the morphological characters within C. cirratus and C. mbirizei is needed.
In C. tilapiae, the reference sequence from Senegal appears basal to all other specimens of native and introduced hosts in the Congo Basin (Fig. 3). We do not find any evidence to contest the species status of C. tilapiae as opposed to Pouyaud et al. (2006) who suggested it is a species complex based on ribosomal DNA.

Conclusion
Our results strongly point to an identical or geographically similar source of introduction of Nile tilapia in Congo and Madagascar, as both regions share identical COI parasite haplotypes. The high haplotype diversity of C. sclerosus within the Upper Congo strongly suggests that multiple introductions have taken place in this area, from different geographic source populations. This refutes our initial hypothesis that introduced parasite populations would suffer genetic bottlenecks. Also, the strong differentiation between parasites from the Upper Congo compared to those from the Middle and Lower Congo suggests different sources of introduction for the latter two regions. Finally, shared parasite haplotypes between the Lower and Middle Congo suggest that natural gene flow is possible at this scale, or it could point to a shared source of introduction or human-assisted dispersal.
Considering the genetic markers, we find a barcoding gap at 15% variation for COI, but not for the other markers. This value is quite high compared to other dactylogyrid monogeneans, but it aligns with the only other available study within Cichlidogyrus, which suggests a barcoding gap above 12% (Rahmouni et al., 2022). However, we want to stress that sequences of more species are needed to have a more complete overview of the phylogeny of the group and to estimate a barcoding gap more precisely. Based on our study, we suggest the need of taxonomic re-evaluation for C. halli, C. papernastrema, C. zambezensis, C. tiberianus and C. cirratus as they all potentially represent at least two species. Additionally, within C. halli, we find that closely related species can first diverge in haptor morphology and later differentiate in the genital sclerites. Lastly, the grouping of C. papernastrema with C. zambezensis, C. arthracanthus and C. sp. 2 within the "tiberianus" group shows that the division of the genus by Pariselle and Euzet (2003) based on haptor configuration does not explain the variation within the group fully and that this division is not always phylogenetically supported. We utter the need for a revision of morphological features corresponding with the larger clades in the phylogenetic tree of Cichlidogyrus/Scutogyrus.