Developing a species concept for Bulinus

As discussed throughout this introductory chapter, the taxonomy of Bulinus has long suffered due to difficulties in applying a typical species concept to delimiting species within the genus. A biological species concept for example, bound by sexual reproduction and parentage (Dobzhansky, 1970), can not apply to a hermaphroditic organism. A morphological species concept, relying on consistently and persistently distinct and distinguishable features, as has been attempted by the works of Mandahl-Barth (1957, 1965) and Brown (1994), is rendered useless in the microtaxonomy of Bulinus due to the unusual forms and overlapping morphology encountered. With the ever-increasing abundance of genetic data being produced for Bulinus, could a phylogenetic species concept, i.e. identifying the smallest biological entities that are diagnosable and/or monophyletic (Mayden, 1997), be applied to Bulinus?

A few things need to be considered in regard to inferring taxonomy through phylogeny, starting with the fundamental schools of systematics, cladistics and phenetics. Cladistics concerns the pathways of evolution, through using synapomorphies and emphasising holophyly to form classifications and reconstruct rooted phylogenetic trees (cladograms),

phenetics on the other hand has its emphasis on the quantitative assessment of overall similarity between taxa regardless of evolutionary relation (Stuessy, 2013). Due to the ignorance of phenetics in considering ancestry, gene trees are often constructed using commonly used tree-building algorithms that have their foundations in cladistic approaches (such as maximum likelihood, parsimony and Bayesian inference), although many still exist that originate in phenetic principles (neighbour-joining and UPGMA). However, many of these tree-building algorithms originate from a combination of both cladistic and phenetic approaches, and biologists are therefore more concerned with the statistical fit of the tree and the computational time/power required to analyse the dataset, rather than the classification of the tree (Stuessy, 2013).

Following the development of Markov chain Monte Carlo (MCMC) methods (a class of algorithms that sample from a probability distribution) in Bayesian statistics, it has been possible to compute much larger genetic datasets, increasing the popularity of this method and its use in investigating Bulinus relationships (Chibwana et al., 2015; Tumwebaze et al., 2019; Clewing et al., 2020; Pennance, Allan, et al., 2020). Although sophisticated, Bayesian phylogenies can still be confounded (as for other methods) due to genetic peculiarities, such as: false similarity (homoplasy), locus duplication (gene paralogy), incomplete lineage sorting, allopatric lineages, mutational saturation effects, polyploidy and introgressive hybridisation that create complex reticulate relationships between taxa, all of which can mask a correct phylogeny (Stuessy, 2013). These complications are likely to have an effect when considering Bulinus phylogenies since allopatric lineages of Bulinus are uncovered (see Chapter 4 and 7) and polyploidy is known to exist for species of Bulinus particularly in the B. truncatus/tropicus complex (Brown and Wright, 1972). Polyploidy in Bulinus may have occurred as an adaptation to altitude (Brown, 1976) and hybridisation (Goldman et al., 1983), either way acting as a reproductive barrier between Bulinus species of differing ploidy which may not be apparent from a gene tree. Introgressive hybridisation also occurs in some of the S. haematobium group species that Bulinus transmit (Léger et al., 2016; Platt et al., 2019), although whether these hybrids relate directly to the origination of a new lineage (i.e. reticulate evolution) or are a result of introgression and repeated backcrossing within the same population needs to be inspected further. Quantitative tools for examining these kinds of networks and reticulate relationships are required (Huson et al., 2010). As for tree building algorithms within phylogenetic analyses, evolutionary nucleotide substitution models are also determined based on statistical tests in packages such as MrModelTest (Nylander, 2004), jModelTest (Darriba et al., 2012) and PartitionFinder (Lanfear et al., 2016). Allowing computational algorithms and models to dictate the phylogenetic analysis being performed isn t necessarily due to a careless or naïve systemicist, but as stated by Stuessy (2013) is due to the overwhelming challenge

of volumes of DNA data, we are using whatever algorithm we need to find patterns of relationship .

Another advantage for following a phylogenetic species concept for Bulinus, is that as is the same for a lot of freshwater pulmonates, Bulinus are poorly represented in the fossil record. Interpreting the evolution of species and ancestral forms to infer phylogeny is therefore possible predominantly from extant taxa. An extensive fossil record of Bulinus could allow for investigation into the morphological changes and potentially infer speciation over time providing a basis to determine extant taxa. The available fossil record is however very geographically limited for Bulinus, and the most recent and detailed descriptions of Bulinus fossils are from Central Europe (Harzhauser et al., 2012; Neubauer et al., 2017), adding to some older records of Bulinus outside of Africa (Neumayr, 1883) (Figure 2.10). As discussed by Neubauer et al. (2017), these Bulinus of Central Europe are likely an extraordinary introduction of an ancestral species via long distance dispersal, since there was no hydrological connection between Africa and Europe during this early Miocene period when Bulinus were known to already be present in Africa. It is still considered therefore that Bulinus very likely originated from Africa, where the earliest fossil records date back to 19-20 Ma (Pickford, 2008). It is however extremely likely that the Bulinus genus is much older than this considering the possibility of B. obtusispira and B. bavayi on Madagascar, which separated from East Africa ~158 Ma and again from India ~84 Ma (Briggs, 2003), being relic Bulinus species here (Wright, 1971; Stothard et al., 2001; Jørgensen et al., 2011), Unfortunately, the fossil record is likely to hold few answers for inferring speciation due to Bulinus being a relatively thin shelled gastropod that will not fossilize well like many of the other more robust snails found in Africa (West et al., 1991), Secondly, living specimens have proven problematic based on shell characteristics to describe, and therefore soft tissue anatomy would be required which is not preserved during fossilisation anyway (Neubauer et al., 2013). Thirdly, we know from the current distribution and habitat preferences that Bulinus have a propensity to inhabit either quickly changing or even temporary freshwater bodies, with only particular species inhabiting the long-lived lakes that offer a better setting for preservation through lacustrine deposits. The fossil record may be quite biased by these factors, and overall not provide the required information to develop time calibrated trees and resolve potential sources of error in phylogenetic trees (Forest, 2009).