Material and methods
Selection of taxanext section
The data set analysed in this study consists of: the newly determined 18S rDNA sequences of Hrabeiella periglandulata, Parergodrilus heideri, Stygocapitella subterranea, Aeolosoma hemprichi, Protodrilus purpureus, Ctenodrilidae n. gen. et sp., and Arenicola marina; a compilation of all the sequences of this gene available in EMBL for polychaetes and other taxa assumed to be closely related to, or members of, ‘non-clitellate Annelida’, namely, Sipuncula, Echiura, Pogonophora (including Vestimentifera) and Aphanoneura (Rouse & Fauchald, 1995); and another eleven 18S rDNA sequences, also taken from EMBL, representing the oligochaete, branchiobdellidan, acanthobdellidan and euhirudinean Clitellata (Table I).
The choice of an outgroup to Polychaeta is problematic because recent molecular studies failed to recover the monophyly of this taxon (McHugh, 1997; Kojima, 1998; Westheide et al., 1999) and because putatively valid outgroups, such as Mollusca, proved to be scattered among annelids or even appeared as an ingroup within polychaetes (Winnepenninckx et al., 1995, 1998; Siddall et al., 1998). For this reason, not only three Mollusca but also two Arthropoda were included in the data set. Arthropods were designated as an unambiguous outgroup (Table I), in accordance with their molecular placement among the Ecdysozoa, the sister group of the Lophotrochozoa to which all others taxa herewith considered belong (Aguinaldo et al., 1997; Adoutte et al., 2000).
Collection of new specimens
The newly sequenced material was collected at the following localities (abbreviations used in Table 1): (MO) Montalbuccio, 43°20‘N 11°15‘E, Siena, Italy (in 1998; coll. E. Rota); (CA) Camugnano, 44°10‘N 11°10‘E, near Bologna, Italy (in 1999; coll. C. Jacomini and E. Rota); (SA) Costa Paradiso, 41°03‘N 8°55‘E, Sardinia (in 1999; coll. E. Rota); (DU) Monkstown, 53°18‘N 6°10‘E, Dublin, Ireland (in 1999; coll. O. Schmidt); (EL) Capo S. Andrea, Elba Island, 42°48‘N 10°09‘E, Tyrrhenian Sea (in 2000; coll. E. Rota and C. Erséus). All the material was fixed and preserved in 95-99 % alcohol.
DNA extraction, amplification and sequencing
DNA of A. hemprichi, H. periglandulata and S. subterranea was extracted according to a standard Chelex™ procedure (Singer-Sam et al., 1989; Hillis et al., 1996). DNA extraction of P. heideri and A. marina was made using “High Pure PCR Template Preparation” from Boehringer-Mannheim Biochemicals, and that of Ctenodrilidae n. gen et sp. and P. purpureus using “QIAamp DNA Mini Kit” from Quiagen, following the instructions of the manufacturers. 18S rRNA gene fragments of A. hemprichi and H. periglandulata were amplified and sequenced according to Martin (2001), using the 16 primers of Winnepenninckx et al. (1994), kindly granted us by the authors. Specimens of the other taxa were analysed following the protocol described in Erséus et al. (2000).
Since it has repeatedly been shown that the sequence alignment may influence the phylogenetic relationships inferred from ribosomal genes (Wägele & Stanjek, 1995; Winnepenninckx & Backeljau, 1996; Erséus et al., 2000; Martin et al., 2000), we studied 18S rDNA sequence data aligned in two different ways: (1) using DCSE (De Rijk & De Wachter, 1993; De Rijk, 1995), which considers a secondary structure model; (2) using Clustal W (Thompson et al., 1994), default settings, without manual corrections. Whatever the alignment method used, some hyper-variable regions of the gene (domain 23 in particular; Van de Peer et al., 1996) were virtually impossible to align and were discarded from the final alignment.
Maximum parsimony analyses of the resulting alignments (EMBL accession numbers ALIGN-000074 for ClustalW and ALIGN-000096 for DCSE) were performed using PAUP*, version 4.0b4a (Swofford, 1998) with the following settings: unweighted characters, heuristic search, random addition of sequences with 100replicates, tree-bisection-reconnection (TBR) branch-swapping algorithm, ‘Multrees’ option in effect, gaps treated as missing. Bootstrap analyses were performed using a heuristic search and TBR branch-swapping on 100 replicates. The Bremer support index was calculated with PAUP* in connection with AutoDecay 4.0 (Eriksson, 1998).