Phylogenetic relationships of a new species of seed-beetle infesting Cercis siliquastrum L. in China and in Europe (Coleoptera: Chrysomelidae: Bruchinae: Bruchini)

Abstract Bruchidius siliquastri n. sp. is described. This seed–beetle develops in pods of Cercis siliquastrum, a widely grown ornamental tree. Several recent reports reveal that this species is well established in Southern France. The presence of this species is also reported from Hungary. The relationships of this new species are investigated using both morphological data and molecular phylogenetic analyses. Our results indicate that this species is likely related to a mostly Paleotropical group, which includes members of genera Bruchidius and Conicobruchus. Hypotheses on the geographic origin of this new species are also discussed.

T he genus Cercis belongs to the family Leguminosae, subfamily Caesalpinioideae, tribe Cercideae; its various species are commonly known as "redbuds". Up to ten species are currently recognized (Isely 1975;Tralau 1981;Salatino et al. 2000;Davis et al. 2002). Two of them (Ce. canadensis L. 1753 and Ce. occidentalis Torr. ex A. Gray 1850) are native to North America, while six species originate from China, one species (Ce. griffi thii Boiss. 1873) originate from Central Asia, and Ce. siliquastrum L. 1753 from Central Asia to Europe (ILDIS, 2006). Cercis siliquastrum (Mediterranean redbud, Judas tree, "arbre de Judée") is grown in all temperate regions of the world for its profusion of pink to magenta fl owers. Chinese species are less commonly found in gardens, but are quite frequently planted in parks and arboretums.
Most Bruchinae recorded from Cercideae feed on seeds of various species of Bauhinia; they belong to four diff erent genera: Caryedon, Caryedes, Gibbobruchus and Pseudopachymerus. A single species, Gibbobruchus mimus (Say 1831), is known to predate on the two Nearctic species of Cercis (Cushman 1911;Zacher 1952;Whitehead & Kingsolver 1975;Center & Johnson 1976, Johnson 1977Hetz & Johnson 1988). Th e record of Callosobruchus maculatus (Fabricius 1775) on Cercis canadensis by Zacher (1952) is very doubtful and has not been confi rmed. No seed-beetle has been so far recorded on any Old World species of Cercis.
In 2003, three specimens of seed-beetle emerged from a small sample of Ce. siliquastrum pods collected in Montpellier (Southern France, Hérault). Th ese specimens proved to be new to Science and without known relative in Europe. Th ey were found to be conspecifi c with specimens bred in China from an unidentifi ed species of Cercis (K.-W. Anton, pers. comm.). Th ree years later, the species was collected in the Rhône valley (Valence, and as far north as Villefranchesur-Saône). Th is seed-beetle was also reported from Southern France in the following localities: Cadalen (Tarn) (A. Sergent, pers. comm.), Gémenos (Bouchesdu-Rhône), La Roquebrussanne (Var) (P. Ponel, pers. comm.) and Nice (Alpes-Maritimes) (M. Clément, pers. comm.). It was also recently reported from Hungary (O. Merkl, pers. comm.) In this study, the new species is described and its relationships with other bruchines are investigated using both morphological data and molecular phylogenetic analyses.

Material and Methods
Samples of recently matured pods of Cercis siliquastrum were fi rst collected in Montpellier in June 2003. Additional material was collected from the same location during the summer of 2004, from Valence (Drôme) in 2005 and from Villefranchesur-Saône (Rhône) in early 2006. Pods were kept in aerated plastic vials until emergence of adults.

Morphological analysis
Male and female genitalia were mounted on microscope slides in water-soluble DMHF (dimethyl hydantoin formaldehyde). Figures of genitalia were drawn from microphotographs of slide preparations. Male and female habitus were captured using a Canon EOS 350D digital single lens camera with a MPE 65 mm MACRO lens. Body length was measured from apex of pronotum to apex of elytra. Th e terminology of genital parts follows Kingsolver (1970).

Molecular phylogenetic analyses
DNA was extracted from an ethanol preserved specimen using the Qiagen DNeasy tissue kit (Qiagen, GmbH, Germany). Partial sequences from three mitochondrial genes (12S rRNA, cytochrome b and cytochrome c oxidase subunit I) were obtained using standard protocol described elsewhere (Kergoat et al. 2004). In order to minimize the computational time of the phylogenetic analyses, a two-steps strategy was followed. First, we have conducted preliminary analyses on a large data set of 152 species to identify the closest relatives of the new species. Genbank sequences (as well as unpublished data from G.J. Kergoat) from four genes (12S rRNA, cytochrome b, cytochrome c oxidase subunit I and 28S-D2 rDNA) were concatenated and aligned using ClustalX (Th ompson et al. 1997) with default settings. MrBayes 3.12 (Huelsenbeck & Ronquist 2001) was used to perform partitioned Bayesian analyses (one partition per gene; two distinct runs of 10 million generations; burnin period of 100,000 generations). For each gene, the best-fi t substitution model was determined using Modeltest (Posada & Crandall 1998). In all analyses, Pachymerus cardo (Fahraeus 1839) (Chrysomelidae: Bruchinae: Pachymerini) and Gibbobruchus sp. (Chrysomelidae: Bruchinae: Bruchini) were used as outgroups, following previous studies (e.g. Kergoat et al. 2005). More thorough analyses were subsequently performed on the data set determined by preliminary analyses, using two inference methods (Bayesian inference and parsimony). Bayesian inference analyses were carried out with the settings used beforehand. For parsimony, the program PAUP* 4.0b10 (Swoff ord 2002) was used to perform unweighted analyses with the following options: heuristic search with 1,000 random-addition replicates; Tree-bisection-reconnection (TBR) algorithm; Maxtrees set to 1,000; Random addition sequences. Since a preliminary incongruence length diff erence test (ILD; Farris et al. 1994) was not signifi cant, the whole data set was analyzed simultaneously. In all parsimony analyses, gaps were considered as a fi fth character following Giribet & Wheeler (1999). Th e robustness of trees was assessed by non-parametric bootstrap procedures (1,000 replicates; Felsenstein 1985) for parsimony analyses and by clade posterior probabilities (CPP) for Bayesian inference analyses (Huelsenbeck et al. 2002). Description. Length (pronotum-pygidium): 2.8-3.7 mm. Body (fi g. 1) short ovate, integument, including legs and antennae, black, abdomen yellowish-red, except ventrite 1 partly black basally. Often extreme apex of femora and ventral part of hind tarsi very slightly reddish; 4 anterior (specially fore) legs sometimes not black, but rather of a dark reddish brown. Vestiture made of thin and short setae, rather dense, though not completely covering integument, recumbent, except on basal part of ventrite 1; dorsally composed of whitish setae, less dense on sides and apex of elytra, denser on disc and scutellum; on pronotum, setae longer on disc and basal lobe, becoming shorter and less dense on sides; pygidium almost uniformly covered with short and thin setae; setae moderately long on ventral side. Male. Head short; eyes bulging, maximum head width 1.43 times width behind eyes; eyes separated by 0.35 times head width including eyes; distance between posterior rim of eyes and apex of clypeus / distance between eyes = 1.76; eye cleft to about half of its length, width at bottom of sinus composed of 7-8 ommatidia; maximum width of postocular lobes equal to 1/3"' eye width at sinus; carina on frons absent, no interocular tubercule. Antenna of moderate length, slightly surpassing elytral base; antennal segments 2-3 subcylindrical, almost equal in length, segments 5-11 widened at apex, but always longer than wide, 11 oval (L/W = 2.1). Length of antennomeres: 1.4 : 1 : 1.4: IA: 1.5 : 1.6: 1.6 : 1.7: 1.7 : 13: 2.6. Pronotum strongly convex, with greatest width at base (W/L = 1.57),subtrapezoidal, its sides convex at about 2/3 of its length from base; without oblique impression on sides of basal lobe. Its disc with strong, well separated punctures, becoming stronger and slightly coriaceous laterally; punctuation suddenly interrupted on posterior half of sides, where integument is alutaceous, with a slight carina-like separation between the two areas. Elytra short, 1.04 times longer than combined width, their sides strongly convex, maximum width at middle; disc fl attened; base of interstriae 2-4 margined, with a very small tooth at base of stria 3. Striae on disc well defi ned; interstriae fl at, strongly alutaceous, with irregular lines of larger punctures. Hind femora incrassate, at their widest 1.8 times wider than mid femora; mesoventral margin carinate, with small preapical denticle, immediately followed by a much shorter one (visible only at higher magnifi cation); hind tibiae widened apically, with dorsomesal and ventral carinae complete and strong, lateral not reaching base; apex of tibia with mucro shorter than width of tarsomere 1 at base; lateral denticle as long as mucro, and dorsal denticles 1/2 of lateral denticle. First tarsomere ventrally without apical denticle. Abdomen with ventrite 5 emarginate, medially as long as sternite 4; ventrite 1 basally with a large patch of short erect setae reaching beyond middle of ventrite; these setae slightly scaly, 2-3 times shorter and much denser than those on surrounding integument (fi g. 2). Pygidium subcircular (W/L = 1.01), with apex turned under (see fi g. 1). Genitalia (fi g. 3): Median lobe of moderate length (maximum width excluding basal hood / total length = 0.27); ventral valve subtriangular, moderately sclerotized, bearing numerous sensillae and on each side a row of 10 to 13 short setae; valve fl anked by a pair of large hinge sclerites; internal sac with two or three large pointed denticles. Basal strut without keel. Lateral lobes cleft to 0.6 their length; apex of parameres with numerous long setae. Female. Similar to male, but last abdomina1 ventrite not emarginate, longer than 4 th ventrite, pygidium slightly narrower than in male (L/W = 1.05), bearing a pair of elongated foveae (see fi g. 1); foveae 0.3 to 0.4 mm long, 0.1 wide, moderately deep, unmargined, their bottom without setae, slightly alutaceous. Genitalia: spiculum gastrale as in fi g. 4; spermathecal body narrow and strongly curved, with an elongated apical diverticulum; opening of spermathecal gland

Molecular phylogenetic analyses
In all preliminary analyses, B. siliquastri was found clustered within a clade constituted by species belonging to genera Bruchidius Schilsky 1905, Conicobruchus Decelle 1951and Megabruchidius Borowiec 1984. Based on this result, a subset of 21 species (including the two outgroup species) was defi ned (see tab. 1) and analyzed following the methods described above.
A mean sequence divergence (K2P distance) of 16.68 ± 0.72% is found between B. siliquastri and the other 19 ingroup species (the closest species being Conicobruchus strangulatus (Fahraeus 1839), with 15.51% of divergence). Th is value is consistent with the mean sequence divergence of 15.53 ± 3.70% that was found when considering the whole subset of 21 species. Bayesian inference and parsimony analyses yield topologies that are quite distinct (fi g. 5). Under

Figure 5
Phylogenetic relationships of the sampled bruchine species. Th e tree of the left corresponds to the result of the Bayesian inference analyses with numbers at nodes indicating CPP values (only values >50% are fi gured). Th e tree on the right corresponds to one of the two most-parsimonious trees from the parsimony analysis of the combined data set with numbers at nodes indicating bootstrap values (only values >50% are fi gured). Th e three clades that are recovered by both analyses are indicated by capital letters (A, B and C).
Male morphology is characterized by the presence of a large setiferous patch on the fi rst ventrite. Abdominal setiferous areas were described in a number of members of the subtribe Acanthoscelidina: Acanthoscelides lobatus (Fall 1910) (Johnson 1970), A. quadridentatus (Schaeff er 1907, Bruchidius aurivillii (Blanc 1889), B. uberatus (Fahraeus 1839), Megabruchidius tonkineus (Pic 1904) (Faustini & Halstead 1982), Scutobruchus ceratiobruchus (Philippi 1859) (Kingsolver 1968), and is present in a large number of tribes, including the basal tribe Pachymerini. Th e presence of large hinge sclerites in the median lobe of the new species is a feature that is rather common in various African and Asian species presently known as Bruchidius or Conicobruchus spp., particularly in B. bimaculatopygus, B. confusus and B. lineatopygus. Th e particular conical sclerites of the internal sac are also of a rather uncommon type. Similar sclerites are found in Bruchidius danilevskyi Anton 1999, B. sapamoroensis Kingsolver 1984, B. savitskyi Anton 1999and B. tuberculicauda Lukjanovitch & Ter-Minassian 1957. But the latter four species share a ventral valve that is defi nitely more pointed than in the present species. Female morphology is characterized by the presence of a pair of longitudinal depressions located on the sides of the pygidium. Pygidial depressions were so far known only in the B. centromaculatus group (Faustini & Halstead 1982;Anton & Delobel 2003); foveae in B. siliquastri are however quite distinctive because they are not limited by any sort of carina.
Molecular analyses indicate a close relationship between B. siliquastri and an assemblage of species belonging to the genera Bruchidius and Conicobruchus (clade B in both trees). Th is grouping is also suggested by the fact that a member of this group, Co. strangulatus, exhibits the lowest level of sequence divergence with B. siliquastri. Th e level of support for the clade B is not high, as indicated by a moderately high CPP value (75%) and by low bootstrap values (below 50%). Having said that, the results of molecular analyses are consistent with the morphological evidence that brings together B. siliquastri with the species that possess similar large hinge sclerites in the median lobe. Interestingly the latter character is shared by all species belonging to clade B, with the exception of B. lichenicola.
Th ough our initial sampling of 152 species was quite dense (for instance, B. aurivillii and members of B. centromaculatus group with pygidial depressions were included), we only partly manage to get some of the key-species that possess one of the characteristic features of B. siliquastri (i.e., particular sclerites of the internal sac, pygidial depressions or setiferous areas). In the absence of a denser sampling, the hypothesis of a close relationship between B. siliquastri and the species related to the other members of clade B thus appears as the most likely. In addition, our analyses reveal the paraphyly of the genus Conicobruchus, and the paraphyly of the group cisti. Th ese preliminary results highlight the need of pursuing studies in the taxonomy and systematics of Old-World seed-beetles in order to clarify their phylogenetic relationships. Th ey justify our present conservative decision not to create a new genus, and to ascribe the new species to the genus Bruchidius.
Th e fact that this species was also reported from unidentifi ed species of Cercis in China, and its relatedness with non-European bruchines allows us to discuss of the likely alien status of this species in Europe. Given the rather conspicuous appearance of B. siliquastri (at least for a seed-beetle) and the wide native area of distribution of Ce. siliquastrum (from Europe and Central Asia) it seems diffi cult to explain why this seed-beetle was not discovered earlier. A recent shift of B. siliquastri from Oriental species of Cercis toward Ce. siliquastrum thus appears as the most likely explanation. We can postulate that this species is now present in many other countries (and perhaps in the New World too), due to the popularity of the European redbud as an ornamental tree. To conclude, we suggest surveying the possible development of this seed-beetle species on other commonly grown Cercis species, and especially on the two New World species that appear to be closely related with Ce. siliquastrum according to recent molecular analyses (Davis et al. 2002).