On a new species and genus in the Cypridini (Crustacea, Ostracoda, Cyprididae) from South Africa, with a phylogenetic analysis of the tribe and a discussion on the genus concept in this group

The present paper describes Mnementh brennei n. sp. n. gen., a new non‐marine ostracod species and genus from temporary wetlands in the Western Cape Province (South Africa). The new genus is the fifth in the subfamily Cypridinae and a phylogenetic analysis using morphological characters demonstrates that it is most closely related to its putative sister taxon Pseudocypris. Apart from this cluster, the phylogenies are not at all robust and this is to all probability due to the incongruent (mosaic) evolution amongst the different characters in valve and soft part modules. Implications of such incongruent evolution of characters for the concept of genera in this group are discussed.


Introduction
Ostracods are small (average length 51 mm), bivalved crustaceans which abound in most types of aquatic and (semi-) terrestrial environments. Up to 80% of all non-marine species are in the family Cyprididae. This family consists of more than 20 subfamilies, and their phylogenetic relationships are highly obscure. Some attempts have been made to deduce the phylogenetic positions of different genera within one subfamily, e.g. the Megalocypridinae (Martens and Coomans 1990) and Scottiinae ), but such analyses are hampered by the lack of a sufficient number of synapomorphic features. This is primarily due to the fact that ostracod morphology is most conservative, especially in soft parts (the limbs). Here, I describe a new genus within the Cypridinae and analyse the phylogenetic position of this new taxon within the subfamily. In addition, I test if the evolution in valve and soft part characters has been congruent, or if both groups of characters have had independent evolutions.

Material
Material for the present paper was collected in September 2001 from a series of temporary water bodies in the surroundings of Van Rhynsdorp (Figure 1), with the assistance of Liz Hoenson (South African Museum, Cape Town). The field trip was provoked by the analysis of ostracod material from that area, offered by Jenny Day and Genevieve Jones (University of Cape Town, Cape Town). Only nine water bodies were sampled during 2 days, but the collection turned out to be very diverse. Apart from the present new genus and species, several other new and rare taxa were found. The giant ostracod, Liocypris grandis, which was presumed extinct, was rediscovered for the first time since its original description by Sars (1924); the presence of additional appendages in the adult female necessitated the erection of a new subfamily, the Liocypridinae (Martens 2003;Matzke-Karasz and Martens 2005). The new genus and species of Cypridopsinae and the new species of Heterocypris and of Megalocypris will be described elsewhere. An overview of the different types of wetlands in the Van Rhynsdorp area is given in Jones and Day (2001).

Phylogenetic analysis
A generic phylogenetic analysis of the tribe Cypridini was performed based on 17 morphological characters (see Appendix 1), using the ostracod genus Chlamydotheca s.l. as outgroup. Chlamydotheca s.l. is thought to be the sister group to the Cypridini. The data matrix (see Appendix 2) was analysed using PAUP 4.06 (Swofford 1998). Input order of taxa was as in the matrix in Appendix 2. All characters were allocated the type ''ordered'' (Wagner parsimony). Trees were built using Maximum Parsimony (MP) with the branch-and-bound routine (furthest taxon input) and with Neighbour Joining (NJ) methods (mean character difference). Bootstrapping used the full heuristic method in MP; for both MP and NJ 10,000 replicates were performed. Testing the (in-)congruence of evolution in soft part and valve characters was done by allocating higher weights to either set of characters and assessing the effect on topology, robustness (bootstrap support) and various descriptive indices (Consistency Index, CI; Retention Index, RI; Homoplasy Index, HI) of the resulting trees.

Derivation of name
Named after the bronze dragon Mnementh, mate to Ramoth, last remaining golden queen dragon on the planet of Pern. I herewith renew my tribute to their creator, Mrs Anne McCaffrey.

Diagnosis
A cypridinid genus, globular and large. RV with frontal and caudal selvage marginally inwardly displaced, and with anterior valve margin not ventrally protruding; ventro-caudal part of calcified inner lamella linearly elevated, not running parallel to valve margin. LV without inwardly displaced selvage and without large inner list.
T1 with penultimate segment divided and with seta d 1 two to three times as long as seta d 2 . Hemipenis with ventral protuberance on the medial shield.

Differential diagnosis
Mnementh n. gen. clearly belongs in the Cypridini, as it has three to five additional ''8''-shaped loops of the post-labyrinthal spermiductus in the hemipenis. It can be distinguished from the other four genera in this tribe by a mixture of features: by the divided penultimate segment of the T2 from Cypris and Pseudocypris; by the sub-marginal selvages in both valves from Cypris, Ramotha, and Globocypris; by the absence of a large inner list on the LV from Pseudocypris; by the fact that seta d 1 is more than twice the length of d 2 from Cypris and Pseudocypris; and by the absence of the additional lateral setae on the T1-palp in the female from Globocypris. The new genus, Mnementh, shows morphological characteristics which classify it in between Pseudocypris and Ramotha, but it also has autapomorphic features, e.g. the additional ventral protuberance on the medial shield of the hemipenis which is thus far unique in the tribe. This ventral protuberance might actually be valid at the specific level only. Only the discovery of a second species in this genus will decide on this issue.

Type material
Holotype: male (OC.2913): soft parts dissected in glycerine in a sealed slide, valves stored dry in a micropalaeontological slide. Allotype: female (OC.2914), dissected and stored as the holotype. Paratypes: two males and one female dissected and stored as the holotype (OC.2915-2917), two males (OC.2920, 2921) and three females (OC.2918, 2919, 2922) stored in toto dry in a micropalaeontological slide after use for SEM; ca 50 males and females stored in toto in 80% alcohol (OC.2923) in one tube, ca 20 males and females stored in toto in 80% alcohol (SAM.A45530).

Repository
Holotype, allotype, dissected and dried paratypes, and one tube with wet paratypes are in the Ostracod Collection of the Royal Belgian Institute of Natural Sciences (Brussels, Belgium). One tube with wet paratypes is in the South African Museum (Cape Town, South Africa).   Sars, 1924.

Derivation of name
The species is named after my friend and colleague, Luc ''brenne'' Brendonck (University of Leuven, Belgium), in recognition of his major contribution to the knowledge of Anostracan biology, especially in southern Africa, and in fond memory of years of friendship and rewarding collaboration.

Diagnosis
Large species, with globular carapace, very hirsute and pitted. Anterior margin of both valves produced towards the ventral side. Anterior calcified inner lamellae wide. Anterior part in dorsal view pointed like a rostrum, greatest width (. half the length) in the middle. Prehensile palps (male) asymmetrical, with terminal segment in right palp larger than in the left palp. T1-palp in female short and wide. Hemipenis with dorsal lobe of the lateral shield large, ventral lobe of the lateral shield long and rectangular, distally produced towards the ventral side. T2 stoutly built. Caudal ramus slender.

Additional description of male
Carapace ( Figure 2N) in dorsal view rounded, with greatest width situated in the middle and being more than half of the length; posterior part rounded, anterior part pointed, in both dorsal and ventral view like a rostrum ( Figure 2K), RV slightly overlapping LV. Ventrally ( Figure 2M), RV overlapping LV in the middle. Valves in external view ( Figure 2E, F) highly arched, greatest height situated slightly behind the middle; anterior margin ventrally produced, posterior margin nearly straight; ventral margin strongly sinuous; valve surface very hirsute, except in the centre, and pitted to varying degrees.
RV ( Figure 2D) with anterior margin ventrally produced, posterior margin broadly rounded, nearly straight; dorsal margin highly arched; ventral margin strongly sinuous in the middle. Calcified part of the inner lamella very wide anteriorly, narrower along the posterior side. Anterior valve margin with clear, but submarginal, selvage ( Figure 2I), antero-ventrally with some traces of inner lists; posterior margin without selvage, but with striking inner list, not running parallel to the valve margin.
LV ( Figure 2C) with outline similar to that of RV. Anterior calcified inner lamella wide, with remnants of inner lists, but no selvage; posterior calcified inner lamella narrower, also with large inner list not running parallel to valve margin.
Central muscle scars with mandibular scars large and with scar pattern of adductor muscles conforming to that of the subfamily. A1 ( Figure 3C) seven-segmented and conform to the subfamily. Terminal segment about as long as wide, all natatory setae long. Aesthetasc Ya less than twice as long as the short seta on this segment. Rome organ minute. A2 ( Figure 3B) with natatory setae reaching well beyond tip of terminal claws; aesthetasc Y short. Apical chaetotaxy ( Figure 3A) with typical sexual dimorphic characters, but with claw G 1 short; seta of y 3 about twice as long as the aesthetasc and seta z3 as long as claw z1 (seta is longer in Cypris).
Md with coxa elongated ( Figure 3E). Md-palp ( Figure 4A) four-segmented. First segment with large respiratory plate and a group of four apical setae: one long and smooth, one short, narrow and smooth (alpha-seta), the latter flanked by one large ''s''seta, set with a double row of setulae, and by one short ''s''-seta. Second segment with two groups of apical setae: an internal group, consisting of a short and hirsute beta-seta, three long and smooth and one long and barbed setae, external group consisting of two long and one shorter setae. Penultimate segment with four dorsal, subapical setae; apically with one narrow and hirsute gamma-seta of intermediate length, two centrally situated smooth setae and three subequal and subapical setae on the internal side. Terminal segment (not shown) with four claw-like and with two or three short and slender setae, all apically inserted.
Mx1 with three endites, a two-segmented palp ( Figure 3D) and a large respiratory plate. Second palp segment about three times as long as basal width. Third endite with two smooth apical claws and with a long and stout lateral seta, apart from the normal apical setae.
Right prehensile palp ( Figure 4D) with elongated basal segment, carrying one apical sensory seta; terminal segment three-dimensionally curved; when flattened in a slide broad, with a narrow apex and with distal margin showing a blunt angle, apically with one broad sensory organ.
Left prehensile palp ( Figure 4E) with basal segment slightly shorter than that of the right palp; terminal segment with broad base, rapidly narrowing towards the tip, the latter bearing a single sensory organ. T2 ( Figure 5B) stout, with seta d 1 more than three times as long as d 2 ; all segments short, wide and hairy. Second segment with one short apical seta. Penultimate segment divided; segment 3A with one subapical seta; segment 3B with one large and one minute apical setae. Fourth segment with one lateral seta and, apart from the apical claw, with one other apical clawlike seta; apical claw short and stout and only in its distal half set with a double row of spines. T3 ( Figure 5A) a cleaning limb with an apical pincer and without further special features. Caudal ramus ( Figure 5D) with ramus narrow and curved, proximally swollen, distally carrying two claws and two setae and with ventral margin serrated with minute setulae. Attachment of the caudal ramus ( Figure 5C) distally split into a short ventral and a longer dorsal branch.  Hemipenis ( Figure 3F, G) with a broadly rounded medial shield, asymmetrically expanded towards the ventral side and with a ventral protuberance; dorsal lobe of the lateral shield symmetrically rounded, ventral lobe of the lateral shield long and distally expanding (but see Figure 3G with pointed vls). Internal anatomy with the normal labyrinth, consisting of the elongated parts ''a'' and ''c'' and the rounded hinge-joint ''b'', followed by the three to five ''8''-shaped coils of the inner spermiductus, situated distally from the labyrinth, the sclerotized semi-circular loop and the various hollow trabecules, leading to the bursa copulatrix.

Additional description of female
Valves (Figure 2A, B, G, H, J) basically as in the male, but in general somewhat larger and even higher. Carapace in dorsal and ventral view ( Figure 2K, M) with greatest width situated in the middle; RV frontally and ventrally overlapping LV. A2 ( Figure 3A) basically as in the male, apart from the normal sexual dimorphism in the apical chaetotaxy. T1 ( Figure 4C) with palp undivided, carrying the normal unequal three apical setae and no lateral setae.
T2 somewhat plumper and heavier than in the male. Genital lobe undivided and without specific characteristics. Ovaria on both sides curved upwards.
For measurements see Table I.

Phylogenetic analysis
When all characters are given equal weight (51), the Maximum Parsimony (MP) method can only partly resolve the trees. In addition, in both MP and distance methods, bootstrap values supporting the Cypris/Globocypris and the Pseudocypris/Mnementh clusters are very low ( Figure 6A: 53 and 63, respectively). Allocating higher weight to the valve characters offers strong bootstrap support for the Cypris/Globocypris clade (100), but the Pseudocypris/Mnementh cluster remains only weakly supported ( Figure 6B: 60). Allocating higher weight to the soft part features dramatically changes the tree topology, with Cypris now clustering with Pseudocypris (bootstrap 580) instead of with Globocypris, which now becomes a basal taxon. In this ''weighted soft part'' tree, nodes have high bootstrap support in both MP and distance methods  ( Figure 6C). CI, HI and RI are lowest in the analyses with equal weight for all characters, they score better in the ''weighted valve'' trees and are best in the ''weighted soft part'' tree (Table II).

Taxonomic history of Cypridinae
The taxonomic history of the Cypridinae is confused. Hartmann and Puri (1974) Wouters (1983) created Tanganyikacypridinae for the genus Tanganyikacypris. This subfamily was later lowered to the rank of tribe and lodged in the Megalocypridinae by Wouters et al. (1989). Martens (1990Martens ( , 1992 grouped four genera in the nominal tribe Cypridini: Cypris, Pseudocypris, Globocypris Klie, and a new genus Ramotha Martens. The latter genus comprised a number of large African species, previously referred to either Cypris, Eucypris, or Strandesia, i.e. to no less than three different subfamilies within the Cyprididae. This illustrates the taxonomic confusion in cypridid ostracods, even amongst relatively large species. Matzke-Karasz and Martens (2007) meanwhile discovered additional appendages in females of Afrocypris barnardi, much like in Liocypris grandis and therefore transferred Afrocypris to the Liocyprinidinae. The positions of the other genera presently lodged within the Cypridinae (Riocypris, Afrocypris, Chlamydotheca, and Bennelongia) are still largely uncertain.
Position of Mnementh n. gen.
In all of the phylogenetic analyses performed here, Mnementh n. gen. clusters closest to Pseudocypris and, to a lesser extent, to Ramotha. It seems fairly clear that Mnementh and Pseudocypris are sister-taxa, but their mutual relationship to Ramotha remains less clear. The weighted soft part tree indicates that Mnementh n. gen. is in fact derived from a lineage intermediate to those of Ramotha and Pseudocypris. But in the weighted valve tree, the three genera form a separate clade (albeit without support), while in the general tree Ramotha is situated basally to the four other genera in the subfamily. Apart from the seemingly strong association between Pseudocypris and Mnementh n. gen., not much can be deduced at present regarding the phylogenetic position of the genera in the Cypridinae.

Morphological mosaic evolution in Cypridinae?
Of the two ''weighted'' analyses, the valve tree is closest in topology to the general tree, and it thus seems that the phylogenetic reconstruction is most influenced by valve characters, such as presence or absence of selvages and inner lists. The topology of the resulting tree in the weighted soft part analysis is different, with Cypris being closest to Pseudocypris and Globocypris being allocated a basal position within the clade. Characters used here are mostly associated with the stout walking leg (T2), where the penultimate segment can be fused or divided and the setae d 1 and d 2 can have different relative lengths. Such characters are also used in other subfamilies of the Cyprididae (e.g. the Eucypridinae) and have demonstrated their merit to characterize genera. Apparently, character evolution has not been congruent, and this high level of uncertainty in the dataset is expressed by the weak support of the tree resulting from the general analyses. It basically means that if the evolution of the valves is assumed to have dominated cladogenesis in this group, then several reversals in soft part characters must have occurred.
On the other hand, if it is assumed that soft part characters are more conservative, then it follows that many reversals in valve characters (such as loss and re-appearance of selvages and inner lists) must have occurred during the history of this clade. It has been suggested that soft parts are more conservative than valves in most Cyprididae (Martens 1998), whereas the inverse is true in, for example, Candonidae (Pinto et al. 2005). In this view of character evolution within the Cyprididae, the weighted soft part tree in Figure 6C would best reflect the reality of the evolutionary history of these genera. This is also the most robust tree with the highest CI and RI and the lowest HI. However, apart from this rule of thumb on conservative evolution of soft parts (which remains to be quantitatively demonstrated), there is no criterion that can be used to discriminate between the different topologies in Figure 6.
The difference in topologies between the two ''weighted'' trees, however, unequivocally demonstrates that the valve and soft part ''modules'' (Danielopol et al. 1990) can evolve independently from each other. This in itself does not have to be a surprising result, as previous studies have already shown that different structures in ostracods can have independent developmental programmes. For example, Gonzalez Mozo et al. (1996) showed that the five natatory setae on the A2 in Herpetocypris can evolve independently from each other, while Yin et al. (1996) found that different setae in clones of Limnocythere inopinata react differently to similar environmental changes.

Genus concepts
In this particular case of the Cypridinae, however, the incongruence in evolution between the two sets of characters could have an effect on the question of what constitutes a genus in Ostracoda in general.
Ever since the first attempt to integrate palaeontological and neontological ostracod classifications by Hartmann and Puri (1974), there is a strongly held belief by ostracod taxonomists that new taxa (species, genera, or above) need to have uniting features in both valves and soft parts. Incongruence in the evolution of both modules could force changes in practical ostracod taxonomy. There would be problems in the application of the classic genus concepts.
Two of the classical genus concepts are the empirical concept (practicality) and the phenetic concept (similarity), both widely applied in the first half of the 20th century. The phylogenetic concept (viewed more narrowly as ''cladistic concept'' by Dubois (1988)) adds biological relevance to genera, in that it requires a group of species to be more than morphologically similar and practically applicable; a genus in this view constitutes a monophyletic clade of species. Dubois' synthetic concept (1988) requires a genus to be a genetic, phylogenetic, and ecological unit, which can be identified by a single criterion, that of hybridizability (species belong to the same genus if they can produce viable hybrids).
To test validity of genera, then, one of the following criteria can be used: (1) to rely on morphological similarity of species; (2) to attempt phylogenetic analyses; or (3) to carry out hybridization experiments.
Morphological similarity is applied in the first two concepts and generally lacks biological relevance, as both concepts are prone to produce polyphyletic genera. Dubois' hybridizability concept, intellectually appealing as it may be, is seldom applicable as hybridization experiments on invertebrates are difficult: most species are indeed known from preserved museum specimens only. The phylogenetic concept is best embedded in present-day systematic research, as it requires application of the same phylogenetic techniques used at all taxonomic levels, and using all types of data (morphological, molecular,…). A major problem in this concept is the definition of a general cut-off point that defines genera. In the case of distance-criteria in tree building, for example, does one define genera at 5, 10 or 15% of similarity? This subjective decision caused some taxonomists to abandon all Linnean hierarchical levels altogether and to only work with non-hierarchical clades (see http:// www.ohio.edu/phylocode/ but also Sluys et al. (2004)). However, this phylocode approach also relies on robust phylogenies. If morphological phylogenies are not robust, because of incongruent evolution between different sets of characters, than validity of genera as well as of phylocode clades in such groups can be seriously weakened.
Taking the Cypridini as an example, the possible solutions are: 1. Collapse all genera within this tribe into one, large genus. This is not advisable, as a lot of information would be lost. In the present model group, Cypris, Ramotha, and Pseudocypris are monophyletic clades comprising several species each and this hierarchical clustering would be lost if they are no longer considered separate genera. 2. Decide on priority amongst modules of characters. If soft parts in Cyprididae are indeed (generally) more conservative than valves, then it would be possible to increase robustness of phylogenies by adding more weight to such characters. It is possible to test this assumption by quantitatively comparing the degree of morphological evolution of both character modules at a higher taxonomic level. 3. Improve phylogenies, either by incorporating new morphological characters, or by combining morphology with molecular markers. This seems rather straightforward, but again the problem is that most taxa are represented by preserved museum specimens from which DNA is (thus far) difficult to extract.

Conclusions
The present phylogenetic analysis of a small group of cypridinid ostracods (five genera, ca 30 species) has shown that character evolution might have occurred in a mosaic way, with incongruence between soft part and valve modules. This poses problems for phylogenetic reconstructions of the evolutionary history of the group, but also for modern classification, as at least phylogenetic genus-concepts depend on robust phylogenies.