Hedyselmis opis : Description of the larva and its phylogenetic relation to Graphelmis ( Coleoptera : Elmidae : Elminae )

The riffle beetle genus Hedyselmis Hinton, 1976 includes two species from the Malay Peninsula, with adults with a highly deviating morphology. Its phylogenetic relationships are unclear, although it has been hypothesized to be related to Graphelmis Delève, 1968, a large genus widely distributed in the Oriental and East Palaearctic regions. In this paper the larva of H. opis Hinton, 1976 is described based on material collected in the Cameron Highlands (Malaysia) and the conspecificity with co-existing adults tested using sequences of one nuclear (5’ end of 18S rRNA) and three mitochondrial gene fragments (5’ end of the large ribosomal unit + tRNAleu + 5’ end of the NADH dehydrogenase subunit 1; 5’ end of cytochrome c oxidase subunit I; and a fragment of cytochrome b) with a total of ca. 2,600 bp. This is the first example of the use of molecular data to match different life stages within the family Elmidae. The larva of H. opis has a subcylindrical body typical of many other elmid genera; abdominal segments 1–7 with preserved pleura; and ninth segment with oval operculum. The last instar larvae have clearly visible prominent spiracles on mesothorax and abdominal segments 1–8. The phylogenetic position of Hedyselmis in relation to Graphelmis was investigated using molecular data for three species of Graphelmis plus a selection of other Elmidae genera. Hedyselmis opis is nested within Graphelmis, confirming their close relationship and suggesting that their status requires taxonomic revision.


INTRODUCTION
Other than for several well known genera, descriptions of elmid larvae, especially from tropical regions, are rarely found in the literature.The major problem lies in the high diversity of the material collected and the difficulty of assigning larvae to conspecific adults.The most reliable method of determining elmid larvae to species is to rear them under laboratory conditions, which is usually very difficult due to the habitat requirements and the length of larval development (Brown, 1987).Thus, the description of larvae is normally confined to species collected with only one representative of the family or with clearly unrelated species.An alternative way of matching adults with larvae is by using DNA sequences with the appropriate level of variation (i.e., with enough variability for there to be differences among closely related species but not enough to impede recognition).There are surprisingly few examples of this use of DNA sequences (e.g., Balke et al., 2005;Jeon & Ahn, 2005;Miller et al., 2005), despite the increasing availability of molecular facilities and their wide use in forensic science (e.g., Wells & Sperling, 2001).Previous descriptions of Elmidae larvae are based exclusively on morphological characters (e.g., Manzo & Archangelsky, 2001;Springer & Acosta, 2003) and this is the first study to use molecular data to assign larvae to specific adults within the Elmidae.Hinton (1976) described the genus Hedyselmis based on six females of one species, H. opis, from the Malay Peninsula.Only recently the males of H. opis and one additional congeneric species were described, also from the Malay Peninsula (Jäch & Boukal, 1997).Adults of both species of Hedyselmis have unusual morphology and seem to be most closely related to the genus Graphelmis (Jäch & Boukal, 1997;iampor Jr., unpubl.).In 2001 some additional material of H. opis, together with unknown elmid larvae, was collected by the junior author in the Malay Peninsula.The lack of any other species of Elmidae at that locality made it likely that the larvae belonged to the same species.
The aim of the current study was to describe the unknown Elmidae larva and using four molecular markers to test the hypothesis that it belongs to the species Hedyselmis opis.The molecular data was also used to outline possible relationship between Hedyselmis and Graphelmis, as suggested most closely related genus.

Morphological analyses
Specimens prepared for morphological study were cleaned and examined under a Nikon SMZ-1B stereo-microscope under diffuse lighting at magnifications up to 140×.Mouth parts were dissected and placed in lactic acid for a few days before examination using a Carl Zeiss transmitted light microscope at magnifications up to 600×.Drawings were made using a drawing tube.
For scanning electron microscopy, specimens were dehydrated in a graded ethanol series, air-dried from absolute ethanol, mounted on stubs using double-sided tape, sputter coated with gold and then viewed in a Hitachi S800 at 10kV.Metric characters were measured to the nearest 0.05 mm using a Nikon SMZ-1B with an ocular grid.Morphological terms follow Lawrence (1991).
DNA was extracted from whole specimens using the Qiagen DNeasy tissue kit.Four fragments were amplified using PCR; three mitochondrial [826 bp from the 3' end of the cytochrome oxidase subunit I (cox1), 358 bp of the cytochrome b apoenzyme (cob), and 835 bp comprising the 3' end of the rrnL (16S rRNA), the adjacent transfer RNA leucine 2 (tRNAleu), and part of NADH dehydrogenase subunit 1 (nad1)] and one nuclear (602 bp of the 5' end of the ribosomal 18S rRNA gene) (see Table 2 for the primers used).Amplification products were purified using Qiagen Qiaquick PCR purification columns and sequenced in both directions.Sequences were sent to GenBank and have accession numbers DQ005512-DQ005519, DQ266471-DQ266513 and DQ267445 (Table 1).

Phylogenetic analyses
Protein-coding genes were not length variable and there was little variation in the ribosomal genes within Elmidae (length of the 18S rRNA fragment was between 626 and 628 bp; of the 16S rRNA fragment between 817 and 827 bp).The sequences were aligned manually.The combined data matrix (all four genes) was analysed using parsimony in PAUP* software version 4.0b10 (Swofford, 2002), with a TBR heuristic search of 10,000 replicates and the option 'save multiple trees' activated.For comparison with the Bayesian probability results (see below), gaps in all searches were coded as missing characters, although treating them as a 5 th character state did not change the topology of the trees (not shown).Node support was measured using Bremer Support (PBS) values (Bremer, 1994) on constraint trees generated by means of TreeRot.v2(Sorenson, 1996) and non-parametric bootstrapping (Felsenstein, 1985) using 1,000 pseudoreplicates of 50 random additions each.Bayesian analyses were executed with MrBayes 3.04 (Huelsenbeck & Ronquist, 2001), using a GTR + I + G model as selected by Modeltest 3.6 (Posada & Crandall, 1998), with the parameters estimated for each partition (i.e., gene fragment).Searches were executed with default priors (uniform probabili-ties) starting with random trees with three heated and one cold Markov chains for 1,000,000 generations, sampled at intervals of 100 generations.To determine the point at which the Markov chains reached stationarity, the log-likelihood scores were plotted against generation time and the point when the likeli-

Matching of adults and larvae
The sequences of the nuclear gene (fragment of 18S rRNA) of the elmid larva and adult Hedyselmis opis were identical.Among the mitochondrial genes, the only unambiguous difference was in the cob sequence, in which in position 352 the larva had a T and the adult a C.This was a synonymous change in the third codon position.In both the sequences of the cox1 and 16S rRNA+nad1 of the adult there were some ambiguous positions (with double peaks in the chromatogram): in cox1, positions 157 and 166 had a T in the larva and T or C in the adult, positions 205 and 343 had A in the larva and A or G in the adult (all synonymous ambiguities in third codon positions); and in 16S rRNA, in positions 266 and 616, larva had G, adult A or G.All ambiguous positions were in sections of high quality sequence, without any ambiguous chromatogram in nearby positions.The total uncorrected genetic divergence between the two specimens was thus between 0.04% (if all ambiguities are resolved as matches) and 0.27% (when all ambiguities are resolved as mismatches).For the mitochondrial genes only, the divergence ranges from 0.05-0.35%.

Phylogenetic analysis
The combined aligned matrix had 2,717 characters (cox1 826, cob 407, 16S 853, 18S 631).Within the whole dataset, there were 628 parsimony informative characters (658 when gaps were treated as a 5 th character state).Heuristic searches resulted in a single most parsimonious tree (consistency index CI = 0.51, retention index RI = 0.37).The sequences of the larva and adult of H. opis were grouped together with high support (100% bootstrap) and included within Graphelmis in a well supported clade (100% bootstrap, 17 Bremer support, Fig. 44).
In the Bayesian analyses the sampled ML values reached stationarity at ca. 20,000 generations, but the first 100,000 (i.e., 1,000 trees) were discarded as a burn-in.The topology of the 50% majority rule consensus tree was almost identical to that of the parsimony tree, with the same highly supported nodes relating the larvae and adults of H. opis and the grouping of Hedyselmis within Graphelmis.

DISCUSSION
In Coleoptera almost identical sequences for the genes used in the present work are found only within the same species, usually from the same or closely linked populations (see e.g.Ribera et al., 2003).Although there are almost no available molecular data on species of Elmidae, it is very likely that the same is true for this family.Some species in very closely related species complexes of recent origin may have nearly identical or identical mitochondrial haplotypes, even though they can be separated using morphological characters (e.g.Ribera & Vogler, 2004).In our case, the genus has only two known, well-characterised species with presumably allopatric distributions (Jäch & Boukal, 1997) and it is highly unlikely that there were larvae of a very closely related undescribed species of the same genus at the same locality (with no adults).Thus, it is highly likely the described larva belongs to Hedyselmis opis.
All our phylogenetic analyses support a close relationship between Hedyselmis and Graphelmis, as suggested by Jäch & Boukal (1997).Hedyselmis and Graphelmis adults share several common characters (shape of prosternal process, fused third and fourth elytral stria, basal teeth on tarsal claws, shape of male pregenital segment), supporting this very close relationship (Jäch & Boukal, 1997 and F. iampor Jr, unpubl. observ.).The larva of Graphelmis has not been formally described, but it is mentioned in Glaister (1999).Based on this work, as well as personal communication with the author, some larvae collected in Malaysia were preliminary assigned to the genus Graphelmis, although further study is needed to support this identification.These larvae are generally similar to those of Hedyselmis, differing only in having less flattened integument scales, paired spines on dorsal side of the abdominal segment 9, slightly longer legs and lighter colouration with a yellowish pattern.Jäch & Boukal (1997) suggest that Cephalolimnius Delève, 1973 is also a close relative of Hedyselmis.This assumption is supported by the wing venation and morphology of male unpaired sclerite of spiculum gastrale.Unfortunately specimens of this genus were not available for study, so the relationship Hedyselmis -Cephalolimnius could not be tested.
Although our study suggests that Graphelmis may be paraphyletic with respect to Hedyselmis, we refrain here from introducing any formal nomenclatorial change.Our sampling of Graphelmis, as well as the knowledge of their larval stages, is insufficient to allow a definitive inclusion of Hedyselmis within Graphelmis.Union programme SYNTHESYS at the MNCN in Madrid, project ES-TAF 136.

Fig. 44 .
Fig. 44.Single most parsimonious tree obtained using the combined dataset.Above branches, posterior probability values (×100) of the Bayesian analyses (only if > 0.5); below branches, bootstrap (only if more than 50%) / Bremer support values of the parsimony analyses.

TABLE 1 .
Geographical origin, collectors and Genbank accession numbers of the taxa studied.

TABLE 2 .
Primers used in this study.