POLYCHAETES HAVE EVOLVED FEEDING LARVAE NUMEROUS TIMES. G. W. Rouse

BULLETIN OF MARINE SCIENCE, 67(1): 391 409, 2000 POLYCHAETES HAVE EVOLVED FEEDING LARVAE NUMEROUS TIMES G. W. Rouse ABSTRACT Trochophore is commonly used to designate larvae with opposed-band feeding and a particular set of ciliary bands. The trochophore has been proposed to represent the ancestral larval form for a group of metazoan phyla as well as for the Annelida. The name trochophore is also often applied to larvae that do not conform to the above definition. A cladistic analysis of spiralian taxa (with special reference to polychaetes), based on a suite of adult and larval characters, is used to assess whether the trochophore (sensu stricto) is a plesiomorphic form for Spiralia or taxa such as Polychaeta. The results favor the conclusion that the trochophore, if defined as a feeding larval form using opposed bands, should not be regarded as plesiomorphic for Spiralia, or any other large taxon such as Polychaeta. The trochophore is therefore re-defined as a larval form with a prototroch. This broad definition covers a wide variety of larvae, and matches the current usage more accurately than the restricted term. The evolution of feeding larvae in a more general sense is also assessed. It is concluded that lecithotrophy is the plesiomorphic condition for polychaetes and that feeding larvae have evolved multiple times. This conclusion is supported by the fact that among polychaete larvae there a number of different feeding mechanisms. Some of these novel feeding modes have yet to be adequately described. Hatschek (1878) was the first to use the term trochophore for a polychaete larva, Polygordius sp. He described this larva as having an apical organ, pre-oral and post-oral ciliated bands flanking a groove, a complete gut, and a pair of protonephridia. Hatschek (1878) proposed that Polygordius represented an ancestral annelid and erected the group Archiannelida, from which the other annelid groups had evolved. The larval form of Polygordius thus also represented the ancestral larval form of the annelids. More recently Nielsen (1995) defined a trochophore as a planktotrophic protostome larva with prototroch and metatroch of compound cilia, surrounding the adoral ciliary zone of separate cilia; gastrotroch (= neurotroch) and telotroch may be present. This is very similar to the original formulation by Hatschek (1878) and Nielsen (1995) also regarded the strictly defined trochophore as plesiomorphic for a large clade of the Metazoa. The suggestion that a trochophore, sensu Hatschek and Nielsen, is plesiomorphic for Annelida has been evaluated a number of times (Shearer, 1911; Segrove, 1941; Ivanova- Kazas, 1985a,b,c; Heimler, 1988; Popkov, 1993; Haszprunar et al., 1995). These authors have pointed out the wide diversity of polychaete larval forms and suggested that the trochophore (sensu Hatschek and Nielsen) is a derived larval type. Nielsen (1995, 1998) has rejected these criticisms on the grounds that convergence in ciliary bands of these larval forms is very improbable. Recently Nielsen (1998) restated his hypothesis about evolution within Metazoa without explicit reference to his Trochaea theory. His use of the word trochophore implied that the larval form was primitively planktotrophic. He also applied it to some lecithotrophic larvae and larvae that lacked bands found in the strictly defined trochophore. In this paper the word trochophore will be unambiguously defined and the hypothesis of Nielsen (1995, 1998) and Hatschek (1878) will be assessed. 391

392 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 The opposed-band method of downstream-feeding, involving two ciliary bands beating in opposition (usually a prototroch and metatroch), has been documented in the larvae of gastropod and bivalve molluscs (Strathmann, 1987), serpulid and oweniid polychaetes (Strathmann et al., 1972; Emlet and Strathmann, 1994) and adult rotifers (Strathmann et al., 1972). The larvae of some entoprocts also apparently feed using an opposed-band system (Strathmann et al., 1972; Strathmann, 1987; Nielsen, 1995). Strathmann (1978) discussed whether the opposed-band system may have evolved independently in these taxa, but accepted the hypothesis that it (and hence feeding larvae), was primitive for the taxa concerned, thus corresponding with the Trochaea hypothesis of Nielsen (1995) (see Rouse, 1999). Strathmann (1978) suggested that the loss of feeding in annelid larvae could still leave structures needed for swimming. These structures could subsequently be available for the evolution of new feeding mechanisms. The evidence for this hypothesis was provided by the fact that many polychaete larvae do not have an opposed-band system yet are still planktotrophic (Strathmann, 1978, 1987, 1993). Regardless of Strathmann s caveats, it is often assumed that the opposed-band system and feeding larvae is a plesiomorphic feature of Spiralia (e.g., Wray, 1995), a position recently maintained by Nielsen (1998). In this paper the evolution of larval forms in Annelida is discussed with particular reference to the hypothesis that planktotrophic larvae represent a plesiomorphic condition in the group. This outline is based on the cladistic analyses by Rouse (1999, 2000). METHODS The analyses by Rouse (1999, 2000) were based on the study by Rouse and Fauchald (1997), with the addition of new characters on larval development, and some additional taxa. The characters based on larval morphology are discussed in detail in Rouse (1999), but are briefly outlined here with special reference to polychaetes. For detailed sources of references, consult the Appendices in Rouse (1999) and Rouse (2000). Within Polychaeta, families were used by Rouse and Fauchald (1997) as terminal taxa, largely because this allowed the most heuristic assessment of relationships based on present knowledge. This is in spite of the artificial nature of using any Linnean rank (de Quieroz and Gauthier, 1994). Rouse and Fauchald (1997) regarded attempts at analysis below the family level, say at species or individual level, as problematic, in terms of computational capabilities, time and available information. For analyses of the evolution of larval morphology, Rouse (1999, 2000) used the same terminals and characters as Rouse and Fauchald (1997) with some additional taxa and a number of new, mainly larval, characters. The decision made in Rouse (1999, 2000), in terms of coding taxa for the larval characters, was that if features had been documented in any member of the taxon concerned, then they were coded as present for the character, unless there was a detailed phylogenetic hypothesis providing evidence to the contrary (e.g., larval feeding in Mollusca). This method assumed that the various larval characters were therefore plesiomorphic in a given group and so biased the analyses towards possibilities that feeding larvae and the strict trochophore larvae were primitive. The features based on larval ciliation are shown in Figure 1 and are called the apical organ, meniscotroch, akrotroch, prototroch, ciliated feeding groove, oral brush, metatroch, neurotroch and telotroch. These ciliary bands or patches were used as nine separate characters, though the grouping of some of the bands was also used as one character, Strict trochophore (see below). The issue of larval feeding was treated differently in Rouse (2000) than in Rouse (1999). Rouse (1999) used larval feeding in a restricted sense to involve downstream feeding using an opposed-band

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 393 system (see below). Rouse (2000) used a broader definition involving larval feeding that used the prototroch to generate a downstream-feeding current. APICAL ORGAN. The apical organ is a group of cells (at the anterior end of many larvae) thought to be sensory. They subsequently become part of the brain (Nielsen, 1995). The apical organ can often be identified by the presence of a group of cilia known as the apical tuft (Fig. 1). The distribution of the apical organ appears to be widespread. Among Polychaeta, it is present where detailed anatomical studies have been done, but Rouse (1999, 2000) largely relied on whether or not an apical tuft was present. However, scoring based on the apical tuft is potentially misleading since Polygordiidae, which have an apical organ, lack an apical tuft, except in the very earliest stages (Woltereck, 1904). The apical tuft also tends to disappear relatively early in development of other polychaetes. AKROTROCH AND MENISCOTROCH. Häcker (1896) introduced the term akrotroch to describe a ciliated region between the prototroch and the apical organ. An akrotroch is defined here as any ring of cilia that lies anterior to the prototroch (i.e., on the episphere), and is not associated with the apical organ. An akrotroch can be distinguished from a meniscotroch (see below) in being a complete ring around the episphere (Fig. 1). Akrotrochs have only been described in the larvae of annelids and have been found in seven polychaete groups; Cirratulidae, Dorvilleidae, Eunicidae, Lumbrineridae, Onuphidae, Orbiniidae and Syllidae. They are absent from 42 families and unknown in the larvae of the remaining 34. Bhaud and Cazaux (1982) defined a meniscotroch as a crescent-shaped area of short cilia on the episphere. They further distinguished it by the cilia of the central part of the meniscotroch being longer, bent and forming a pointed brush (Bhaud and Cazaux, 1982). In contrast, Lacalli (1980) distinguished this particular ciliated patch as being separate from the meniscotroch, and called it the frontal organ. This is not accepted here. Based on their more restricted definition Bhaud and Cazaux (1982) regarded a meniscotroch as being absent in taxa such as Hesionidae, Polynoidae, Pholoidae and Sigalionidae, even though they have a ciliated patch or patches on the episphere. Rouse (1999) used the more general definition with a further refinement in that the meniscotroch is always on the ventral surface of the episphere, in line with the mouth (Fig. 1). A meniscotroch is found in 13 polychaete families, all of them members of the clade Phyllodocida. PROTOTROCH. The term prototroch, coined by Kleinenberg (1886), is defined as a ring of (usually compound) cilia that is derived from a group of specific cells, called trochoblasts (Damen and Dictus, 1994). While there is variation in the specific cellular composition of the prototroch, it is essentially the same in all annelids, echiurans, entoprocts, molluscs, and sipunculans (Rouse, 1999). Embryological studies have identified at least primary trochoblasts (embryonic cells 1a 2-1d 2 ) in 12 polychaete families (Rouse 1999). For most polychaetes no detailed cell lineage studies have been performed to assess whether a true prototroch exists. It was then assumed by Rouse (1999) that a distinct ciliated band in the immediately preoral region of a larva corresponds to a prototroch (Fig. 1). The prototroch beats in a downstream direction, with the effective stroke from anterior to posterior. This beat is used for swimming in pelagic larvae and also for food gathering in some larvae. The only polychaetes where a prototroch can said to be absent are Aeolosomatidae and Histriobdellidae. METATROCH. A metatroch has been defined as a post-oral band of cilia that beats in a posterior to anterior direction, in opposition to the prototroch (see Strathmann, 1993; Nielsen, 1995). This was accepted by Rouse (1999) and the definition was expanded, with reference to annelids, to encompass any ciliary band that lies behind the mouth (Fig. 1), but is still presegmental (i.e., lies on the peristomium). The implications of this emphasis on a positional definition were that the metatroch did not necessarily have to be associated with feeding. However, this expanded delineation of homology does allow the greatest possibility for the metatroch to be a plesiomorphic feature. Members of five polychaete families have been shown to have a metatroch in the traditional sense (i.e., associated with food gathering). These are Amphinomidae, Opheliidae, Oweniidae, Polygordiidae, and Serpulidae, though it should be noted that larvae of some members of the Opheliidae lack a metatroch and there is some doubt about the Amphinomidae (Rouse, 2000). It

394 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 Figure 1. Diagrammatic representation of a ventral view of a larva showing all ciliary bands described in this paper. Note that no larval form has ever been described with all of these bands. Abbreviations: a. apical tuft, ak. akrotroch, b. buccal opening (actual or incipient), f, ciliated feeding groove, m. metatroch, me. meniscotroch, n. neurotroch, o. oral brush, p. prototroch, t. telotroch. would appear that the metatrochs in all of these taxa are peristomial structures. Saccocirridae may have members that use the metatroch for feeding (Pierantoni, 1906), but further investigation is required. Additionally, there are a number of other polychaete groups that have a metatroch, but it is not used in feeding. The families are: Ampharetidae, Chaetopteridae, Dorvilleidae, Eunicidae, Lumbrineridae, Myzostomidae, Onuphidae, Orbiniidae, Protodrilidae, Sabellidae and some Serpulidae. In most of these cases the larvae are lecithotrophic, but some are planktotrophic without using the metatroch (Chaetopteridae, Protodrilidae). Other polychaete families have erroneously been cited as having a metatroch, or an opposed-band ciliary feeding system, and hence implying the presence of a metatroch. These families are Capitellidae, Sabellariidae, Siboglinidae (= Pogonophora), and Syllidae (Rouse, 2000). CILIATED FEEDING GROOVE. Where the metatroch has been shown to be involved in food gathering, there is often a third ciliary band between the prototroch and metatroch (Fig. 1). This is commonly called the ciliated food groove (Nielsen, 1995; Strathmann, 1993) and acts to transport food. Among polychaetes it is present in larval oweniids and serpulids. Amphinomidae and Polygordiidae may also have this feature. ORAL BRUSH. An oral brush is a bundle of long cilia at the posterior base of the prototroch, on the left side of the mouth (Fig. 1), of larvae of the scaleworm polychaete families Pholoidae, Polynoidae and Sigalionidae (Phillips and Pernet, 1996). The oral brush extends posteriorly, and is used to help ingest particles up to 60 µm in diameter. It appears that the brush is an alternative way of facilitating feeding in polychaete larvae that are planktotrophic, but lack a metatroch (Phillips and Pernet, 1996). TELOTROCH. A telotroch is a ring of cilia lying at the posterior of a larva (Fig. 1), that Strathmann (1993) believes to have a locomotory function. A telotroch is present in the Echiura and among the Polychaeta it is generally present where larval development is known (unknown for 29 families), though it often appears in larvae at a rather late stage of development. It has not been shown in any

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 395 developmental studies on Aeolosomatidae, Chaetopteridae, Glyceridae, Hesionidae, Histriobdellidae, Lopadorhynchidae, Magelonidae, Pisionidae, Sabellidae, Serpulidae, Sternaspidae and Tomopteridae. NEUROTROCH. The term neurotroch was introduced by Gravely (1909) to identify a ventral longitudinal band of cilia running from behind the mouth to near the anus (Fig. 1). It is commonly referred to as a gastrotroch (e.g., Nielsen, 1985), but this term is used to describe ventral transverse ciliary rings on segments in polychaete larvae (Bhaud and Cazaux, 1982) and is not used here. Among polychaetes the occurrence of a neurotroch is sporadic. It occurs in at least some members of 32 families, including Siboglinidae (= Pogonophora) and has been shown to be absent in 18 others. STRICT TROCHOPHORE. This was a character used by Rouse (1999) that directly assessed the homology argument involved in Nielsen s (1995) definition of the trochophore. Though Nielsen (1995: table 10.1) indicated that the Rotifera have a trochophore, development in this group is always direct. Taxa that have at least some members with feeding larvae, an apical plate, a prototroch, a metatroch, a ciliated feeding groove, and a pair of larval protonephridia are Echiura, Entoprocta, Polygordiidae (though they arguably lack a ciliated feeding groove), and Serpulidae. The presence of a telotroch (not in Entoprocta, Serpulidae or in early trochophores of Polygordiidae) and neurotroch (not in Polygordiidae) was not required, since these were not in Nielsen s most restrictive definition. LARVAL FEEDING. A planktotrophic larval form using an opposed-band system is fundamental to the restricted definition of a trochophore, according to Nielsen (1995). This requires the use of the prototroch to generate a current for locomotion and feeding, and the metatroch creates a beat in the opposite direction (Strathmann, 1993). The ciliated food groove, made up of short cilia between the prototroch and metatroch, moves captured food to the mouth. The possibility that opposed-band larval feeding is a plesiomorphic condition was assessed by Rouse (1999), using this feature as a binary (absent/present) character. The presence of a prototroch and metatroch (i.e., opposed-bands) has led Strathmann (1987, 1993) to suggest that the Echiura, Amphinomidae, Oweniidae, Polygordiidae, Serpulidae and possibly Opheliidae, may feed in a similar manner and this was accepted by Rouse (1999), even though some of these taxa do not have strict trochophores. All other polychaete groups do not have opposed-band larval feeding, with the possible exception of Saccocirridae. Where planktotrophic larvae occur in other polychaete groups, such as Capitellidae, Chaetopteridae, Magelonidae (as young larvae), Pectinariidae, Phyllodocida (sensu Rouse and Fauchald, 1997), Poecilochaetidae, Protodrilidae, Sabellariidae, Saccocirridae (some), and Spionidae, they utilize other forms of particle capture. The larvae of Chaetopteridae, Poecilochaetidae and, from within the Phyllodocida, Pisionidae use some form of mucus feeding (Åkesson, 1961; Strathmann, 1987; Nozais et al., 1997). The larvae of Protodrilidae and some Saccocirridae use an eversible pharynx (Jägersten, 1952; Sasaki and Brown, 1983) and are encounter predators. Where larval feeding has been found in Capitellidae, Magelonidae (young), Pectinariidae, Phyllodocida, Sabellariidae and Spionidae, it involves the use of the prototroch. It cannot be classified as opposed-band downstream larval feeding, since members of all these taxa lack a metatroch. Rouse (2000) used a more general definition of downstream larval feeding that accomodated these other feeding methods. This assumed that any feeding method that involved a downstream current generated by the prototroch for feeding was homologous. By expanding the concept of downstream feeding beyond that of just opposed-band feeding, the global homology of this larval feeding mode is greatest. This presents the most stringent assessment of Strathmann s and Nielsen s hypothesis that feeding larvae are primitive. The families that have been found to only have downstream-feeding larvae are Amphinomidae, Chrysopetalidae, Glyceridae, Nephtyidae, Oweniidae, Pectinariidae, Polynoidae, and Sabellariidae. On the other hand, Capitellidae, Echiura, Hesionidae, Magelonidae, Opheliidae, Pholoidae, Phyllodocidae, Serpulidae, and Spionidae also have members with larvae that do not feed (or are lecithotrophic until becoming encounter predators), and hence are lecithotrophic (Wilson, 1991 and Rouse, 1999). Rouse (1999, 2000) assumed lecithotrophy to be secondary in each case; i.e., the

396 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 presence of downstream-feeding is assumed to be plesiomorphic in each polymorphic group. This assumption that feeding is always plesiomorphic will have to be tested using more detailed cladistic analyses at the appropriate levels and may well be shown to be wrong in some cases. For example, in a preliminary analysis on the evolution of reproduction in Serpulidae, Rouse and Fitzhugh (1994) suggested that feeding larvae in this group may be secondary. What will perhaps be surprising to some is the incidence of groups in Polychaeta where no feeding larvae have been found at all. These taxa are Alvinellidae, Ampharetidae, Arenicolidae, Cirratulidae, Ctenodrilidae, Dorvilleidae, Eunicidae, Flabelligeridae, Goniadidae, Histriobdellidae, Lumbrineridae, Maldanidae, Myzostomatidae, Nereididae, Onuphidae, Orbiniidae, Protodriloididae, Psammodrilidae, Questidae, Sabellidae, Siboglinidae, Sternaspidae, Syllidae, Sphaerodoridae, Terebellidae, Tomopteridae and Trichobranchidae (Rouse, 2000). Most of these groups are well studied and feeding larval forms, if they occur, would presumably have turned up in detailed plankton surveys such as those of Thorson (1946). This list of taxa with no planktotrophic larvae given above differs with the findings of Wilson (1991) and Giangrande (1997). For instance, Wilson (1991) erroneously listed Cirratulidae, Dorvilleidae, Eunicidae, Goniadidae, Nereididae, Sabellidae and Syllidae as having planktotrophic members. Some of these mistakes are simple taxonomic errors, e.g., a serpulid Pomatoceros triqueter has planktotrophic larvae and is listed under the Sabellidae, but most appear to be misinterpretations of planktonic existence for planktotrophy. ANALYSES. Rouse and Fauchald (1997) prepared two scoring matrices for their analysis of the Polychaeta. Rouse (1999) used only one of the types of analysis used by Rouse and Fauchald (1997), the A/Pw matrix, and applied it to a restricted set of taxa (with four more polychaete families than the restricted taxon set in Rouse and Fauchald 1997). Rouse (2000) used the same restricted set of taxa as Rouse (1999) but used the APe, APw and M forms of coding. The overall conclusions about the evolution of larval forms and larval feeding were largely unaffected by the different coding methods. The data matrices used in Rouse (1999, 2000) and Rouse and Fauchald (1997) are available from the following sites: TreeBASE (http://herbaria.harvard.edu/treebase/) and the author s website (http://www.wallace.bio.usyd.edu.au/papers/gregr/). Rouse (1999, 2000) performed cladistic analyses using PAUP 3.1.1 (Swofford, 1993) and most parsimonious trees were rooted using Rotifera, based on arguments by Aguinaldo et al. (1997), Nielsen (1995), and Winnepenninckx et al. (1995). Analyses were subject to successive approximations character weighting (SACW) (Farris, 1969; Carpenter, 1988, 1994), though this resulted in different trees than the original most-parsimonious trees. The results of Rouse (1999, 2000) were not substantively affected by the use of SACW and transformations studied on the initial trees provide essentially the same conclusions. Analysis of character state distributions and optimisations were performed using MacClade 3.07 (Maddison and Maddison, 1992). ACCTRAN transformation were favored since these tend to preserve the initial homology hypotheses (Pinna de, 1991) though in fact transformations that maximize reversals should be used in this regard (see Rouse 2000). RESULTS The major methodological difference between the analyses of Rouse (1999) and Rouse (2000) was that Rouse (1999) used a restricted larval feeding character (opposed-band) and Rouse (2000) used a much broader character (general downstream feeding). There were also a few errors in the matrix of Rouse (1999) that were corrected in Rouse (2000). Nevertheless, the resulting most parsimonious trees were basically similar. For the purposes of this paper only the A/Pw analysis of Rouse (1999, 2000) will be discussed here, though the varies other analyses give essentially the same conclusions. One of the six most parsimonious trees derived after SACW from the matrix in Rouse (2000) is shown in Figures 2 4. Basal taxa such as the Rotifera, Platyhelminthes and Nemertea were pruned

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 397 Figure 2. Basal portion of one of six most-parsimonious trees (after SACW) from the A/Pw analysis of Rouse (2000). Some basal taxa such as Entoprocta, Nemertea, Platyheminthes, and Rotifera have been pruned. Taxon names shown in bold are represented by a larval diagram. The sister group to the polychaete clade seen in this figure (Scolecida in Rouse and Fauchald 1997) is the Palpata (in Rouse and Fauchald 1997) and is shown in Figure 3 and 4. Taxon names that are underlined have members (at least some) that show downstream feeding, except for the Mollusca where feeding larvae are derived within the clade. Terminals marked with? have larvae that are unknown. Sources of the modified larval diagrams are as follows: Mollusca, Neomenia carinata (Jägersten, 1972); Sipuncula, Phascolosoma vulgar (Gerould, 1907); Arenicolidae, Arenicola marina (Newell, 1948); Maldanidae Clymenella torquata (Lacalli, 1980); Capitellidae, Mediomastus fragile (Rasmussen, 1956); Opheliidae Ophelia bicornis (Wilson, 1948); Orbinidae, Leitoscoloplos pugettensis (Blake, 1980); Paraonidae, Aricidea sp. (Fewkes, 1883).

398 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 399 from this tree. Six most parsimonious trees were found in Rouse (1999) that differed from Rouse (2000) slightly in the relationship among the Phyllodocida. Results using the other matrices were similar but for details see Rouse (2000). For detailed outlines on the evolution of larval ciliary bands see Rouse (1999). STRICT TROCHOPHORE. Trochophore larvae, according to the definition of Nielsen (1995), are only found in four disparate taxa, Echiura, Entoprocta, Polygordiidae and Serpulidae, and would appear to have evolved independently in each case. The argument by Nielsen (1995) in favor of the plesiomorphic status of the feeding trochophore larva is predicated on the idea that the various ciliary bands involved would not evolve independently. However, the more relevant question concerning the trochophore is one concerning the co-occurrence of ciliary bands in larval forms, not the evolution of the bands themselves. CILIARY BANDS. The apical plate (or apical tuft) was found to be a uniform condition, found in all ingroup taxa, except where it has been lost independently in Clitellata, (Orbiniidae, Paraonidae), and Sabellidae. The lack of any data for the larvae of Questidae means that, owing to their cladistic proximity to the clade (Orbiniidae, Paraonidae), they could have lost this feature as well. An akrotroch appeared independently four times across Polychaeta with 2 possible transformations. The four occurrences are in Cirratulidae, Syllidae, the clade (Dorvilleidae, Lumbrineridae, Eunicidae, Onuphidae) and Orbiniidae. The character based on the meniscotroch had a length of one step in Rouse (1999) was a synapomorphy for the members of Phyllodocida (except Paralacydoniidae, Syllidae and Sphaerodoridae). In Rouse (2000) the tree topologies were different from Rouse (1999) with respect to Phyllodocida (see Fig. 4), and the meniscotroch was inferred to have been lost in Paralacydoniidae. The presence of a prototroch was a synapomorphy for all ingroup taxa in the analysis, except the Nemertea and Platyhelminthes. Chaetopteridae was the only taxon of the 14 annelids coded with a? where there was some knowledge about embryonic cell lineage. It appears that trochoblasts (and hence a prototroch) have been transformed from an easily recognizable state in Chaetopteridae. There were various possible transformations for the metatroch character in Rouse (1999, 2000). Using an accelerated transformation (ACCTRAN, accelerating changes towards the root, thus favoring reversals) a metatroch appears below the node joining Entoprocta with most other spiralians, and then is lost below the node joining Clitellata and Polychaeta, Figure 3. (Opposite page) Portion of one of six most-parsimonious trees (after SACW) from the A/ Pw analysis of Rouse (2000) showing one of the two major clades of Palpata, the Canalipalpata (as in Rouse and Fauchald 1997). Taxon names shown in bold are represented by a larval diagram. The sister group to Canalipalpata is Aciculata (Rouse and Fauchald 1997) and is shown in Figure 4. Taxon names that are underlined have members (at least some) that shown downstream feeding. Terminals marked with? have larvae that are unknown. Sources of the modified larval diagrams are as follows: Flabelligeridae Flabelliderma commensalis (Spies, 1977); Cirratulidae Cirriformia spirabrancha (Blake, 1975b); Ampharetidae Melinna palmata (Grehan et al., 1991); Pectinariidae Pectinaria koreni (Sveshnikov, 1978); Trichobranchidae Terebellides stroemi (Willemöes-Suhm, 1871); Terebellidae Amphitrite ornata (Mead, 1897); Spionidae Scolecolepis fuliginosa (Day, 1934); Magelonidae Magelona alleni (Wilson, 1982); Chaetopteridae, Chaetopterus variopedatus (Cazaux, 1965); Sabellariidae, Sabellaria alveolata (Wilson, 1929); Sabellidae, Chone duneri (Yun and Kikuchi, 1991); Serpulidae, Galeolaria caespitosa (Andrews and Anderson, 1962); Siboglinidae, Siboglinum fiordicum (Bakke, 1974); Polygordiidae, Polygordius sp. (Hatschek, 1878); Protodrilus adhearens (Jägersten, 1952); Saccocirridae Saccocirrus papillocerus (Pierantoni, 1906); Oweniidae Owenia fusiformis (Wilson, 1932).

400 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 401 and also in Mollusca. A metatroch is then acquired independently seven times in Polychaeta: i.e., separately in Opheliidae; Orbiniidae; the clade (Amphinomidae, Dorvilleidae, Eunicidae, Euphrosinidae, Lumbrineridae, Onuphidae); the clade (Oweniidae, Polygordiidae, Protodrilidae, Protodriloididae, Saccocirridae); Ampharetidae; Chaetopteridae, and the clade (Sabellidae, Serpulidae). If a transformation was used that maximized reversals then a metatroch was found to be a plesiomorphic condition for all ingroup taxa with eight subsequent losses in the Annelida and an independent acquisition only in the Ampharetidae. The character based on the ciliated feeding groove appeared independently in Entoprocta, Echiura, Amphinomidae, Oweniidae, Polygordiidae and Serpulidae. The oral brush character had a length of two steps. It arises below the node joining Pholoidae with the other scaleworm taxa. Sigalionidae and Polynoidae show this feature, which is subsequently lost in Aphroditidae. The unknown condition for Acoetidae and Eulepethidae means there are various possible reconstructions for this character in the scaleworm clade. The transformation for the telotroch differed slightly between Rouse (1999) and Rouse (2000). However, under ACCTRAN in each study, the presence of a telotroch appears below the node joining the clade (Mollusca, Sipuncula) with the remaining ingroup taxa. It was then lost in the Sipuncula, Clitellata, and 8 9 times in the Polychaeta: in Chaetopteridae, Magelonidae, Oweniidae, Protodriloididae, and several times in Phyllodocida. Rouse (2000) corrected errors for the Spionidae and Poecilochaetidae which had been scored as present for a neurotroch (Rouse, 1999), when in fact it appears to be absent. This resulted in different transformations for this character in the two studies, though in both the neurotroch, under ACCTRAN, appears below the node joining Entoprocta with other most other spiralians. It is then lost in the clade (Mollusca Sipuncula), and a number of times in Polychaeta. In Rouse (1999) the neurotroch, under ACCTRAN, shows 10 losses among polychaetes with subsequent reappearance in Phyllodocida, Protodrilidae and Saccocirridae. In Rouse (2000) the ACCTRAN transformation shows five losses from the initial present condition in Polychaeta. A neurotroch then reappears and is lost in several clades of Canalipalpata. Various possible transformations occur, in part, because of the unknown states for Acrocirridae and Alvinellidae and their sister group relationships (to Flabelligeridae and Ampharetidae respectively) with taxa that have lost the neurotroch. Also, there are various transformations through the grade Nereididae Figure 4. (Opposite page) Portion of one of six most-parsimonious trees (after SACW) from the A/ Pw analysis of Rouse (2000) showing one of the two major clades of Palpata, Aciculata (as in Rouse and Fauchald 1997). Taxon names shown in bold are represented by a larval diagram. The sister group to Canalipalpata is Aciculata (Rouse and Fauchald 1997) and is shown in Figure 3. Taxon names that are underlined have members (at least some) that shown downstream feeding. Terminals marked with? have larvae that are unknown. Sources of the modified larval diagrams are as follows: Polynoidae, Harmothoe longisetis (Cazaux, 1968); Sigalionidae, Sthenelais boa (Cazaux, 1968); Pholoidae, Pholoe synopthalmica (Cazaux, 1968); Chrysopetalidae, Paleonotus belli (Blake, 1975a); Glyceridae, Glycera convoluta (Fuchs, 1911); Paralacydoniidae, Paralacydonia paradoxoa (Bhaud, 1967); Phyllodocidae, Anaitides williamsi (Blake, 1975a); Pisionidae, Pisione remota (Aiyar and Alikunhi, 1940); Hesionidae, Ophiodromus pugettensis (Blake, 1975a); Nephtyidae, Nephtys hombergi (Fuchs, 1911); Nereididae, Platynereis bicanaliculata (Blake, 1975a); Syllidae, Syllis variegata (Cazaux, 1984); Amphinomidae, Eurythoe complanata (Kudenov, 1974); Dorvilleidae, Dorvillea rudolphi (Blake, 1975a); Lumbrineridae, Lumbrinereis impatiens (Cazaux, 1972); Eunicidae, Eunice valens (as E. kobiensis) (Åkesson, 1967); Onuphidae, Nothria elegans (Blake, 1975c).

402 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 (Pilargidae Nephtyidae) and in the clade (Oweniidae, Polygoridiidae, Protodrilidae, Protodriloididae, and Saccocirridae). As can be seen from the above summary the various ciliary bands have had varying histories in polychaete larvae (for illustrated transformations, see Rouse, 1999). To provide some sense of this variability one of the most parsimonious trees from Rouse (2000) has representative figures of many of the polychaete larval forms (Figs. 2 4). LARVAL FEEDING. In Rouse (1999) the character based on opposed-band larval feeding (under ACCTRAN) evolved six times independently; in Echiura, Entoprocta, Opheliidae, Serpulidae, the clade (Amphinomidae Euphrosinidae), and arises below the node joining Oweniidae with the clade comprised of Polygordiidae, Protodrilidae, Protodriloididae and Saccocirridae. There is a reversal in that clade for the clade comprised of Protodrilidae and Protodriloididae; the former uses an eversible buccal organ to catch prey and the latter has direct development. In Rouse (2000) the more general character of downstream larval feeding also showed a high level of homoplasy. Under an ACCTRAN transformation, downstream feeding appeared independently 10 times; in Entoprocta, Echiura, the clade consisted of Oweniidae, Polygoridiidae, Protodrilidae, Protodriloididae, and Saccocirridae (with subsequent loss in the Protodrilidae and Protodrilidae clade), the clade (Arenicolidae, Capitellidae, Maldanidae, Opheliidae, Scalibregmatidae) [though with subsequent loss in the clade (Arenicolidae, Maldandidae)], the clade (Sabellariidae, Sabellidae and Serpulidae) [with subsequent loss in the Sabellidae], the Spionida [with subsequent loss in the Poecilochaetidae], the Pectinariidae, the Amphinomida, and twice in the Phyllodocida excluding the Syllidae [with losses in each of Nereididae, Pisionidae, Sphaerodoridae and Goniadidae]. The cladistic occurrence of downstream larval-feeding among polychaetes is shown in Figures 2 4 (for illustrated transformations see Rouse, 2000). DISCUSSION The results of Rouse (1999) outlined here clearly reject the homology of the strict trochophore, as defined by Hatschek (1878) and Nielsen (1995). Such a feeding larval form appears to have evolved several times independently. The argument by Nielsen (1995) in favor of the plesiomorphic status of the strict trochophore is predicated on the idea that the various ciliary bands involved would not evolve independently. However, the more relevant question about the trochophore concerns the co-occurrence of ciliary bands in larval forms, not the evolution of the bands themselves. The homology of the apical tuft, prototroch, neurotroch, and telotroch is largely supported here. These various larval features are either plesiomorphic for taxa included in this analysis (e.g., apical tuft), are a synapomorphy for a large suite of taxa (e.g., the prototroch), or are labile features that have been lost (or possibly evolved independently) a number of times (e.g., metatroch, neurotroch and telotroch). Thus, there are numerous combinations of these various features (plus the ciliated food groove, akrotroch, meniscotroch and oral brush) that result in a suite of larval forms, of which the strictly defined trochophore is only one. This is clearly seen in Figures 2 4. The various taxa with a strict trochophore larva (Echiura, Entoprocta, Polygordiidae and Serpulidae) actually have differing components anyway, supporting the conclusion that they are the result of convergence. For instance, Polygordiidae do not have a neurotroch,

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 403 arguably lack a ciliated feeding groove, and only develop a telotroch after segmentation has begun. Entoprocta and Serpulidae also lack a telotroch. The only larval form that has all of the features of the strict trochophore is that of some Echiura, argued to be a derived polychaete group (Nielsen, 1995; Nielsen et al., 1996; McHugh, 1997). In accordance with the results of his study, and to stablilize terminology, Rouse (1999) recommended that the use of the word trochophore in a strict sense be abandoned. It was redefined as a larval form that has a prototroch originating from a specific group of cells, the trochoblasts. LARVAL FEEDING. The hypothesis that opposed-band larval feeding is plesiomorphic for Spiralia, and hence also within Annelida, was not supported in Rouse (1999), which instead suggested it has evolved multiple times; in Entoprocta, Echiura and in several polychaete groups. This is in addition to the body of evidence that suggests opposed-band larval feeding is derived within Mollusca (Haszprunar, Salvini-Plawen, and Rieger, 1995; Ponder and Lindberg, 1997). The more general hypothesis that downstream larval feeding is plesiomorphic was also rejected in Rouse (2000). The implication of this is, of course, that lecithotrophy is the plesiomorphic condition for polychaetes, and that feeding larvae have evolved multiple times. This rejection of the general homology hypothesis of feeding by a current generated by the prototroch is supported when a closer examination of the actual feeding mechanisms involved is made. There appears to be several types of downstream feeding when this is considered in relation to the occurrence of ciliary bands and what is known about larval feeding. The broad definition, assumption of homology of downstream feeding made in Rouse (2000), and the fact that polymorphic terminals were scored as having downstream feeding, amount to a biases in favor of this mode being primitive. Close examination of the actual feeding methods suggests the initial homology hypothesis is incorrect. This ultimately is supported by the disparate occurrence of downstream larval feeding shown on Figures 2 4. Though little is known about the actual downstream feeding of certain larval forms, it is clear that there are several modes of feeding, and up to 12 separate types can be identified among polychaetes. These are summarized as follows: FEEDING WITH MENISCOTROCH (ONCE OR TWICE?). Strathmann (1987, 1993) noted that several taxa in Phyllodocida with planktotrophic larvae lack a metatroch (and food groove). These were Nephtyidae, Phyllodocidae, Polynoidae and Sigalionidae. The feeding mechanism in the latter two families has now been elucidated by Phillips and Pernet (1996) and involves the prototroch, meniscotroch and the oral brush. This mechanism allows the capture of a much greater range of particle sizes (2 60 µm in diameter) than opposedband feeding. Phyllodocidae and Nephtyidae lack an oral brush (but for Nephtyidae see Wilson, 1936) but have been recorded as ingesting large particles (Strathmann, 1987) and do possess a meniscotroch. Other polychaete families within Phyllodocida also have a meniscotroch and have planktotrophic larvae. This includes Chrysopetalidae (Blake, 1975a), Glyceridae (Simpson, 1962; Cazaux, 1967), Hesionidae (Haaland and Schram, 1982; 1983), and possibly Paralacydoniidae (Bhaud, 1967), though the latter taxa seems to lack a meniscotroch. The feeding mechanisms of these taxa have yet to be described, but may involve the use of the prototroch and menisotroch (where present). It is possible that the use of an oral brush for feeding as seen in scaleworms evolved after the evolution of feeding larvae in Phyllodocida, where just the prototroch and meniscotroch can be used. Further resolution of the relationships among the Phyllodocida is required in this regard.

404 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 OPPOSED-BAND LARVAL FEEDING (SIX KINDS?). Opposed-band larval feeding was scored as present here for Echiura, Amphinomidae, Oweniidae, Polygordiidae, Serpulidae, and Opheliidae. The cladistic placement of these taxa (Figs. 2 4) suggests that this mode evolved independently in each taxon (with the exception of Oweniidae and Polygordiidae, where it may be homologous). When the actual morphology of the structures providing this opposed-band feeding is studied, it appears that this convergence can be reasoned out. For instance, the opposed-band larval feeding in the oweniid Owenia fusiformis is unique in that it is based on simple cilia (Emlet and Strathmann, 1994), whereas other opposed-band systems appear to be based on compound cilia. In Amphinomidae, the diagram of particle flow by Jägersten (1972) suggests a ciliated food groove is present and that opposed-band larval feeding is occurring on the tentacles, something unique to this group. Note however, that Kudenov (1974) and Marsden (1960) show early stage larvae of the amphinomid Eurythoe complanata and that there is only a prototroch present. Further investigation is clearly required on Amphinomidae and Euphosinidae. Echiura, Serpulidae, and Polygordiidae possess the requisite bands, though the Polygordiidae food groove appears to not be ciliated, and proper feeding experiments are needed for both Echiura and Polygordiidae. Hermans (1978) described opposed-band larval feeding in Armandia brevis. However, no mention of ciliated feeding groove was made by Hermans (1978). Also, Opheliidae have various larval forms and some species have planktotrophic larvae that lack a metatroch, such as Armandia cirrosa described by Guérin (1973). The lecithotrophic opheliid larvae of Ophelia bicornis (Wilson, 1948) also lack a metatroch. FEEDING MECHANISMS UNKNOWN (FOUR KINDS?). Among other polychaetes, Capitellidae (Rasmussen, 1956; Eckelbarger and Grassle, 1987; Hansen, 1993), Magelonidae (Wilson, 1982), Pectinariidae (Wilson, 1936), Sabellariidae (Wilson, 1929) and Spionidae (Hannerz, 1956; Daro and Polk, 1973) may also have downstream-feeding larvae, though the actual mechanisms have yet to be described. Capitellidae and Sabellariiidae have been regarded as having opposed-band feeding (Nielsen, 1998). Strathmann (1978) refers to Thorson (1946) for data on capitellid larvae with opposed-band feeding but this paper contains no such information. Hansen (1993) suggested that the larvae of the capitellid Mediomastus fragile feed with a prototroch and metatroch. While this species may be a downstream-feeder, the larvae of M. fragile clearly lack a metatroch, as shown by Rasmussen (1956) (as H. filiformis) and no other studies of capitellid larvae have described a metatroch (e.g., Eisig, 1899; Wilson, 1933). Strathmann (1987) mentions a metatroch is present in Sabellariidae, but this is based on unpublished observations and is not supported by other studies. A metatroch was not shown by Wilson (1929), Dales (1952) or Smith and Chia (1985) in their detailed studies of sabellariid larvae (contra Nielsen, 1998). Wilson (1929) referred to posterior extensions of the prototroch that overhang the lateral edges of the mouth as lip folds and it is conceivable that they act in a similar fashion to a metatroch by beating food back towards the mouth. However, there is clearly no ciliated band behind the mouth (Wilson, 1929; Dales, 1952; Smith and Chia, 1985). Feeding by larval spionids and magelonids has been described briefly (Daro and Polk, 1973; Wilson, 1982). Daro and Polk (1973) state that lateral circular currents are produced by the larvae of Polydora ciliata larvae that are 3 8 segments long and that they ingest particles less than 20 µm in diameter. They suggest that abdominal cilia are responsible for generating this feeding current. However, since a neurotroch is lacking in spionid larvae, the only abdominal bands that could generate this current are the

ROUSE: EVOLUTION OF POLYCHAETE LARVAE 405 gastrotrochs (Wilson, 1928) and it seems from the diagram by Daro and Polk (1973) that the prototroch could also be involved. How food particles are actually captured by young spionid larvae has yet to be explained. Wilson (1982) fed young magelonid larvae on various flagellates and diatoms. Presumably the young larvae fed with the aid of their prototroch, which Wilson (1982) described as being somewhat similar to that of sabellariids in being expanded ventrally over the mouth. With the development of their feeding palps, the magelonid larvae became encounter predators on bivalve larvae, but Johnson and Brink (1998) provide new views on encounter predation by polychaete larvae. The larvae of all pectinariids described to date are planktotrophic, but how they feed is unknown. Wilson (1936) described very early development, although he was unable to get the larvae to survive for more that a few days. Wilson s (1936) study, and that of Lagadeuc and Retière (1993) show that the ciliation around the mouth of pectinariid larva is very elaborate and involves the prototroch and large ciliated lips lateral to the mouth. This is reminiscent of the lips that develop in sabellariid larvae and possible homologies should be investigated. CONCLUSIONS The trochophore, if defined as a feeding larval form using opposed-bands cannot, on current evidence, be regarded as a plesiomorphic larval form. The evidence suggests that the various ciliary bands have differing evolutionary histories, and only the Echiura (possibly an annelid group) has members with the classical trochophore. The name trochophore should refer to larval forms that have a prototroch. This broad definition covers a wide variety of larval forms and matches the current usage more accurately than the restricted term. Features such as the neurotroch and telotroch and opposed-band feeding show convergence and reversals and can be associated with trochophores but this is not requisite. There are clearly different larval feeding modes in polychaetes, and opposed-band feeding represents but one of the existing feeding modes. Various larval-feeding modes have evolved independently from a lecithotrophic condition. In each case it appears that the prototroch, which has a primarily locomotory role, has become involved in larval-feeding in association with other ciliary bands such as the meniscotroch, metatroch and/or oral brush. It is possible that larval feeding has evolved independently more than 10 times among polychaetes. ACKNOWLEDGMENTS Many thanks to I. Fine (Badham Library University of Sydney) who tracked down many essential references, as did L. A. Ward (Smithsonian Institution). D. J. Patterson provided facilities and advice. This work was supported by an Australian Research Council QEII Research Fellowship. LITERATURE CITED Aguinaldo, A. M. A., J. M. Turbeville, L. S. Linford, M. C. Rivera, J. R. Garey, R. A. Raff and J. A. Lake. 1997. Evidence for a clade of nematodes, arthropods and other moulting animals. Nature 387: 489 493. Aiyar, R. G. and K. H. Alikunhi. 1940. On a new pisionid from the sandy beach, Madras. Rec. Indian Mus. 42: 89 107.

406 BULLETIN OF MARINE SCIENCE, VOL. 67, NO. 1, 2000 Åkesson, B. 1961. On the histological differentiation of the larvae of Pisione remota (Pisionidae, Polychaeta). Acta Zool., Stockh. 42: 177 225.. 1967. The embryology of the polychaete Eunice kobiensis. Acta Zool., Stockh. 48: 141 192. Andrews, J. C. and D. T. Anderson. 1962. The development and settling of the polychaete Galeolaria caespitosa Lamarck (Fam. Serpulidae). Proc. Linn. Soc. N.S.W. 87: 185 188. Bakke, T. 1974. Settling of the larvae of Siboglinum fiordicum Webb (Pogonophora) in the laboratory. Sarsia 56: 57 70. Bhaud, M. 1967. Étude du développement de quelques larves d annélides polychètes à Banyulssur-Mer. Vie Milieu 18: 531 558. and C. Cazaux. 1982. Les larves de polychètes des côtes de France. Oceanis 8: 57 160. Blake, J. A. 1975a. The larval development of Polychaeta from the northern California coast. III. Eighteen species of Errantia. Ophelia 14: 23 84.. 1975b. The larval development of Polychaeta from the northern California. I. Cirriformia spirabrancha (family Cirratulidae). Trans. Am. Microsc. Soc. 94: 179 188.. 1975c. The larval development of Polychaeta from the nothern California coast. II. Nothria elegans (Family Onuphidae). Ophelia 13: 43 61.. 1980. The larval development of Polychaeta from the northern California coast. IV. Leitoscoloplos pugettensis and Scoloplos acmeceps (family Orbiniidae). Ophelia 19: 1 18. Carpenter, J. M. 1988. Choosing among multiple equally parsimonious cladograms. Cladistics 4: 291 296.. 1994. Successive weighting, reliability and evidence. Cladistics 10: 215 220. Cazaux, C. 1965. Développement larvaire de Chaetopterus variopedatus (Renier). Act. Soc. Linnéenne de Bordeaux 102: 1 16.. 1967. Développement larvaire de Glycera convoluta Keferstein. Vie et milieu 18: 559 571.. 1968. Étude morphologique du developpement larvaire d annelides polychetes (Basin d Arcachon). I. Aphroditidae, Chrysopetalidae. Arch. Zool. exp. gén. 109: 477 543.. 1972. Développment larvaire d annélides polychètes (Bassin d Arcachon). Arch. Zool. exp. gén. 113: 71 108.. 1984. Reproduction et développement larvaire du Syllidae Syllis variegata Grube, 1860 (Annelide, Polychete). Cah. Biol. Mar. 25: 123 139. Dales, R. P. 1952. The development and structure of the anterior region of the body in the Sabellariidae, with special reference to Phragmatopoma californica. Quart. J. Microsc. Sci. 93: 435 452. Damen, P. and W. J. A. G. Dictus. 1994. Cell lineage of the prototroch of Patella vulgata (Gastropoda, Mollusca). Develop. Biol. 162: 364 383. Daro, M. H. and P. Polk. 1973. The autecology of Polydora ciliata along the Belgian coast. Netherlands J. Sea Res. 6: 130 140. Day, J. H. 1934. Development of Scolecolepis fuliginosa (Clpd.). J. Mar. Biol. Ass. U.K. 19: 633 654. de Quieroz, K. and J. Gauthier. 1994. Toward a phylogenetic system of biological nomenclature. TREE. 9: 27 31. Eckelbarger, K. J. and J. P. Grassle. 1987. Interspecific variation in genital spine, sperm and larval mophology in six sibling species of Capitella. Bull. Biol. Soc. Wash. 7: 62 76. Eisig, H. 1899. Zur Entwicklungsgeschichte der Capitelliden. Mitt. Zool. Stat. Neapel 13: 1 292. Emlet, R. B. and R. R. Strathmann. 1994. Functional consequences of simple cilia in the mitraria of oweniids (an anomalous larvae of an anomalous polychaete) and comparisons with other larvae. Pages 143 157 in W. H. Wilson, S. A. Stricker and G. L. Shin, eds., Reproduction and development of marine invertebrates. Johns Hopkins Univ. Press, Baltimore, Maryland. Farris, J. S. 1969. A successive approximations approach to character weighting. Syst. Zool. 18: 374 385.