Systematics of stalked jellyfishes (Cnidaria: Staurozoa)

Staurozoan classification is highly subjective, based on phylogeny-free inferences, and suborders, families, and genera are commonly defined by homoplasies. Additionally, many characters used in the taxonomy of the group have ontogenetic and intraspecific variation, and demand new and consistent assessments to establish their correct homologies. Consequently, Staurozoa is in need of a thorough systematic revision. The aim of this study is to propose a comprehensive phylogenetic hypothesis for Staurozoa, providing the first phylogenetic classification for the group. According to our working hypothesis based on a combined set of molecular data (mitochondrial markers COI and 16S, and nuclear markers ITS, 18S, and 28S), the traditional suborders Cleistocarpida (animals with claustrum) and Eleutherocarpida (animals without claustrum) are not monophyletic. Instead, our results show that staurozoans are divided into two groups, herein named Amyostaurida and Myostaurida, which can be distinguished by the absence/presence of interradial longitudinal muscles in the peduncle, respectively. We propose a taxonomic revision at the family and genus levels that preserves the monophyly of taxa. We provide a key for staurozoan genera and discuss the evolution of the main characters used in staurozoan taxonomy.


INTRODUCTION
Staurozoa is a class of benthic cnidarians, the so-called stalked jellyfishes (Figs. 1 and 2), represented by approximately 50 species (Clark, 1863;Kramp, 1961;Daly et al., 2007). However, from the first stauromedusan species described (Lucernaria quadricornis Müller, 1776) until their proposition as the fifth class of Cnidaria (Marques & Collins, 2004), Figure 2 General external anatomy of stalked jellyfishes. Craterolophus convolvulus: (A) lateral view, (B) oral view. Abbreviations: am, arm; cl, calyx; gd, gonad; mn, manubrium; pd, pedal disk; pe, peduncle; tc, tentacle cluster. Photo credit: David Fenwick. the group has had a long history of classifications, being labeled as a "puzzling group" (Gwilliam, 1956). While one species was erroneously first placed among sea cucumbers (Manania auricula as Holothuria lagenam referens Müller, 1776), most assessments prior to the 1850's assumed that they were closely related to sea anemones (Cuvier, 1817;Cuvier, 1830) until Sars (1846) noted that the presence of gastric cirri suggested that they were allied with the jellyfishes. Reflecting this thinking, Goette (1887) included Stauromedusae as a suborder within Scyphozoa, a position that was only recently challenged. Marques & Collins (2004) proposed the class based on a phylogenetic analysis of morphological and life cycle traits, as the clade uniting the fossil group Conulatae and the Stauromedusae. In light of further evidence from the fossil record, a subsequent analysis of a similar dataset contradicted the hypothesis that Conulatae and Stauromedusae form a clade, and proposed the composition of Staurozoa to consist exclusively of the extant Stauromedusae (Van Iten et al., 2006). The same analysis suggested that Staurozoa is the sister group to all other medusozoans (Cubozoa, Hydrozoa, and Scyphozoa), a result corroborated by analyses of nuclear ribosomal data ; see also Van Iten et al., 2014). In contrast, however, analyses of complete mitochondrial genome data (Kayal et al., 2013) suggest that Staurozoa may be the sister group of Cubozoa, and more recent phylogenomic analyses support a clade formed by Staurozoa, Cubozoa, and Scyphozoa (Zapata et al., 2015), demonstrating that more studies are necessary to reach a stable topology for Cnidaria. Although evolutionary studies have supported monophyly of the class (Collins & Daly, 2005;Collins et al., 2006;Kayal et al., 2013), comparatively little effort has been applied toward determining the systematic relationships among species of Staurozoa, with rare exceptions (Collins & Daly, 2005;Lutz et al., 2006). The current classification of Staurozoa is mainly based on the proposals of Clark (1863), Haeckel (1879), Uchida (1929) and Carlgren (1935), and is completely focused on anatomical features. Uchida (1973) proposed a hypothesis of relationship among families of stalked jellyfishes based on the characters that he regarded as important, but this analysis was not derived from specific evolutionary methods. A recent molecular inference, with limited taxon sampling, demonstrated the need for reassessing suprageneric clades, because several were found not to be monophyletic (Collins & Daly, 2005). Additionally, many characters used in the taxonomy of the group have ontogenetic and intraspecific variation, and demand consistent assessments and clarifications to establish their correct homologies (Miranda, Morandini & Marques, 2009). Consequently, staurozoan classification and taxonomy is subjective, based on phylogeny-free inferences, and families and genera may be commonly defined by homoplasies (Collins & Daly, 2005). Therefore, Staurozoa is in need of a thorough systematic revision.
Inferences about the relationships among staurozoan species are especially important because of the phylogenetic status and position of Staurozoa, as a distinct clade separate from the other major cnidarian groups (Anthozoa, Cubozoa, Hydrozoa, and Scyphozoa) Van Iten et al., 2006;Kayal et al., 2013;Zapata et al., 2015). The peculiar life cycle of staurozoans (Wietrzykowski, 1912;Kikinger & Salvini-Plawen, 1995;Miranda, Collins & Marques, 2010) is tightly connected to their unique anatomy, in which characters of polypoid and medusoid stages are present in the same stauromedusa (Miranda, Collins & Marques, 2013). Our expectation is that a better understanding and interpretation of the character evolution within the group will provide crucial information for inferences in cnidarian evolution.
Therefore, it is of the utmost importance to carry out an evolutionary analysis encompassing a large number of species of Staurozoa. This study presents the most comprehensive phylogenetic hypothesis for Staurozoa yet proposed and provides the first phylogenetic classification for the group. Further, we provide a key for staurozoan genera and discuss evolution of the main characters used in staurozoan taxonomy.

MATERIAL AND METHODS Molecular
Twenty-four species from ten genera, plus eight non-identified species (identified to genus level), from different regions of the world, were used in the molecular analyses (Table 1). Tissue samples from the tentacle clusters (or marginal lobes for Lipkea spp.) were removed and preserved in 90-100% ethanol, and stored at -20 C. DNA extractions were carried out with InstaGene (Bio-Rad) at the Universidade de São Paulo, Instituto de Biociências (IB-USP, Brazil), or using an organic phenol-chloroform method on the automated DNA isolation system, AutoGenPrep 965 (AutoGen Inc., Holliston, MA, USA) at the Smithsonian's Laboratories of Analytical Biology (LAB, USA), following the manufacturers' protocols. Genes were amplified using PCR, then purified with AMPure Ò (Agencourt Ò ) or ExoSAP. Different molecular markers (mitochondrial COI and 16S; nuclear ITS-ITS1+5.8S+ITS2, 18S, and 28S) were targeted for analyses (Tables 2  and 3). These markers were previously adopted and have been shown to be efficient for evolutionary studies in medusozoans (Dawson, 2004;Collins et al., 2006;Collins et al., 2008;Miranda, Collins & Marques, 2010;Nawrocki et al., 2013;Cunha, Genzano & Marques, 2015). DNA sequencing was done using the BigDye Ò Terminator v3.1 kit (Applied Biosystems) and the same primers used for PCR ( Table 2). The procedure was carried out on an ABI PRISM Ò 3100 Genetic Analyzer (Hitachi). Samples were extracted, amplified and sequenced at LAB (USA) and IB-USP (Brazil). Outgroup sequences (Anthozoa, Cubozoa, Hydrozoa, and Scyphozoa) were obtained in GenBank (Table 4).
The following models were used in the Maximum Likelihood and Bayesian analyses: COI-HKY+I+G; 16S-TIM2+I+G; ITS-K80+I+G; 18S-TIM2+I+G; 28S-TIM3+I+G; combined-GTR+I+G (no partitioned analyses were conducted). Maximum Likelihood analyses (ML) were performed with individual and combined dataset, using PhyML Table 3 Polymerase chain reaction (PCR) conditions for the different molecular markers used in the phylogenetic analyses.   3.0 (Guindon et al., 2010). Branch support was estimated by bootstrapping (Felsenstein, 1985) with 1,000 replicates for the PA (PAUP Ã 4.1) and ML (PhyML) analyses. The Bayesian inference (BA) was also performed with individual and combined dataset, in MrBayes v3.2 (Ronquist & Huelsenbeck, 2003), with 5,000,000 generations sampled every 1,000 generations, four chains, and four independent runs. One fourth of the topologies were discarded as burnin, and the remaining used to calculate the posterior probability. Following MrBayes v3.2 manual, convergence was assessed by ensuring that the average standard deviation of split frequencies was less than 0.01 after 5,000,000 generations, and that the convergence statistic (PSRF = Potential Scale Reduction Factor) was close to 1.0 for all parameters. FigTree (http://tree.bio.ed.ac.uk/software/figtree/) was used to visualize and edit the resulting trees. The alignments and trees are available in the repository of phylogenetic information TreeBASE at: http://purl.org/phylo/treebase/ phylows/study/TB2:S18971. Selected morphological characters generally used in the taxonomy of Staurozoa were optimized by using ACCTRAN (accelerated transformation) in the combined molecular phylogenetic tree at the generic level, using TNT 1.1 (Goloboff, Farris & Nixon, 2008).

Phylogeny
The PA, ML, and BA topologies based on combined markers are similar (Figs. [3][4][5]. The main difference is the relationships among Lucernariopsis vanhoeffeni, Lucernariopsis campanulata, and Kishinouyea sp.  and the relationships among Kishinouyea corbini, Lucernariopsis tasmaniensis, and Kishinouyea sp. SAF. Single-gene topologies under PA, ML, and BA show varying levels of correspondence to the combined topology (Figs. S1-S15 and 6). At least one molecular marker individually supports each main group observed in the PA, ML, and BA results ( Fig. 6). This is the most comprehensive molecular phylogenetic hypothesis that has been presented for Staurozoa, which consequently allows us to carry out a comparative analysis of trait distribution across clades, as well as to provide a major revision for the classification of the class (Figs. 7 and 8; Table 7).
The class Staurozoa has traditionally been divided into the subgroups Cleistocarpida and Eleutherocarpida ( Fig. 8), based on the presence and absence, respectively, of an internal structure called the claustrum ( Fig. 9; Table 8). However, a preliminary phylogenetic analysis for the class (Collins & Daly, 2005) suggested that these groups, proposed by Clark (1863) (Fig. 8), were not monophyletic. Our study, with better taxon sampling, corroborates this preliminary result, and refutes the suborders Eleutherocarpida and Cleistocarpida ( Fig. 8). Instead, our working hypothesis based on our combined set of molecular data ( Fig. 7) shows that staurozoans are divided into two well-supported groups, which can be distinguished one from the other by the absence/presence of interradial longitudinal muscles in the peduncle (or stalk) (Figs. 10 and 11; Table 8). We propose two new suborders for the Staurozoa: Myostaurida (from the Greek myos: muscle; stauro: cross) and Amyostaurida composed of species with and without interradial muscles in the peduncle, respectively (Figs. 7, 8F, 10 and 11; Table 8). Presence of longitudinal muscles in the peduncle (Figs. 10A and 11) is a character easily recognizable with a cross-section of the middle region of the peduncle (Uchida, 1929;Ling, 1937;Ling, 1939;Berrill, 1963;Miranda, Collins & Marques, 2013), and consequently a useful feature for distinguishing the two major subgroups of stalked jellyfishes (see discussion about character evolution below).
We followed Daly et al. (2007) and elevated Craterolophinae to the family level, as Craterolophidae (Figs. 7, 8E and 8F), including only the genus Craterolophus (Figs. 7 and 8; Table 7). We included specimens of C. convolvulus from Europe (Germany and the United Kingdom) and from the U.S.A. (Table 1) in our analysis. However, there was no specimen available of C. macrocystis; the species is very rare, having been recorded only twice (Hutton, 1880;von Lendenfeld, 1884). Therefore, the monophyly of the genus and, consequently, the family, remains to be tested.
The monophyly of the family was tested and corroborated in our analysis . However, the two traditional genera Lucernariopsis and Kishinouyea did not resolve as monophyletic . According to current taxonomy, the distinction between the three genera of this family is subtle. Kishinouyea and Sasakiella differ by the absence and presence, respectively, of primary tentacles (Ling, 1937). Both Kishinouyea and Lucernariopsis do not have primary tentacles in adults, but they are thought to differ in the internal anatomy of the peduncle. Whereas species of Kishinouyea (and Sasakiella) have four chambers basally and one chamber in the middle of the peduncle, species of Lucernariopsis have just one chamber throughout the peduncle (Uchida, 1929;Kramp, 1961). However, these characters change during development (Uchida, 1929;Figure 5 Bayesian phylogenetic hypothesis. Analysis based on combined data of mitochondrial markers COI and 16S, and nuclear markers ITS, 18S (SSU), and 28S (LSU). Posterior probability at each node. ANT, Antarctica; AUS, Australia; EPR, East Pacific Rise; GER, Germany; JAP, Japan; NZ, New Zealand; SAF, South Africa; UK, the United Kingdom; USA, the United States of America. Hirano, 1986). Additionally, a cross-section at the very base of the peduncle is rarely reported in the description of species; most only include information concerning the middle region of the peduncle (e.g., Kishinouyea hawaiiensis in Edmondson, 1930;Lucernariopsis capensis in Carlgren, 1938;Miranda et al., 2012), or do not mention where the peduncle was sectioned (e.g., Corbin, 1978), causing some doubt about whether this distinction is reliable in defining these genera. Recently, Lucernariopsis tasmaniensis was described with "a single cruciform chamber that becomes four-chambered basally within pedal disc" (Zagal et al., 2011), a character that corresponds to the genera Kishinouyea and Sasakiella (Kramp, 1961). Not surprisingly, our phylogenetic hypothesis (Figs. [3][4][5] indicates that the traditional distinctions between these genera are not robust. We suggest that the three genera of Kishinouyeidae be synonymized due to the lack of characters to differentiate them. Kishinouyea Mayer, 1910 would have priority over Lucernariopsis Uchida, 1929 andSasakiella Okubo, 1917. However, there is a further nomenclatural problem in Uchida's (1929) proposal of the genus Lucernariopsis based on Lucernaria campanulata (Lamouroux, 1815;Gwilliam, 1956: 10). Previously, Clark (1863) had recognized Lamouroux' species as not assignable to Lucernaria, since the species does not have interradial muscles in the peduncle, and proposed the new genus name Calvadosia (non Calvadosia Cossmann 1921; junior synonym of Calvadosiella Wenz 1939; Mollusca, Gastropoda) to accommodate it. Thus, following the rule of priority, the proper generic name of Lucernariopsis Uchida, 1929 would be Calvadosia Clark, 1863. Consequently, Calvadosia has priority over Kishinouyea Mayer, 1910, and we therefore synonymize Kishinouyea, Sasakiella, and Lucernariopsis within Calvadosia. The name of the family remains the same, according to ICZN, article 40. 1.
Our phylogenetic analyses show a close relationship between Haliclystus, Stenoscyphus, Depastromorpha, and Manania (Figs. [3][4][5]. Based on this evidence and on morphological similarities (see below, Table 8), we propose that these genera should be assigned to the family Haliclystidae (Figs. 7 and 8F; Table 7). We also include in this family the not yet sampled genera Depastrum and Halimocyathus, but this needs to be tested in future studies.
According to the phylogeny, Stenoscyphus inabai is closely related to Haliclystus borealis and Haliclystus tenuis (Figs. [3][4][5], and deeply nested within Haliclystus spp. In order to keep Haliclystus monophyletic, and since the name Haliclystus Clark, 1863 has priority over the name Stenoscyphus (a monospecific genus) Kishinouye, 1902, we synonymize Stenoscyphus with Haliclystus (Figs. 7 and 8F; Table 7). Some limited developmental data has already suggested a close relationship between these two genera (Hirano, 1986). The main difference between the former genus Stenoscyphus and Haliclystus is an entire and divided coronal muscle, respectively (Kramp, 1961;Hirano, 1986). Therefore, Haliclystus inabai is the only described Haliclystus with an entire coronal muscle (Table 8).
Based on morphological evidence, we include Depastrum Gosse, 1858 and Haliclystus Clark, 1863 in the same family (Figs. 7 and 8F; Table 7). However, there is a nomenclatural issue related to these genera. Haeckel (1879) proposed both the subfamilies Depastridae and Haliclystidae in the same book (Fig. 8B). Both names were used by Uchida (1929), but Carlgren (1935), Kramp (1961) and Daly et al. (2007) used only Depastrinae/Depastridae, and replaced Haliclystidae by Lucernariidae (Fig. 8). Consequently, the prevailing name would be Depastridae. However, there are two caveats: (1) Depastrum cyathiforme, the single species of the genus (Table 7), is not sampled in this study and consequently its position in the phylogeny (i.e., its relationship with other genera) is more tentative (Fig. 7; based only on morphological similarities); and (2) the last report of D. cyathiforme in the literature was about 40 years ago (den Hartog, 1976). Therefore, we believe it is better for nomenclatural stability to use the name Haliclystidae over Depastridae, and as first revisers refer to the International Code on Zoological Nomenclature (ICZN), article 24. 2

.2.
Family Kyopodiidae Larson, 1988 Type genus: Kyopoda Larson, 1988 The Kyopodiidae is a monospecific family proposed by Larson (1988) as part of Eleutherocarpida. Kyopoda lamberti Larson, 1988 has an unusual morphology: its calyx is reduced and the gonads and gastric cavity reside at the base of the peduncle (Larson, 1988).
There was no specimen available of K. lamberti to be included in our phylogenetic analyses. In addition, its particular morphology hampers attempts to identify a relationship with other genera of Staurozoa, which makes future study focusing on the homologies of K. lamberti with other Staurozoa especially interesting. Therefore, we presently retain the monogeneric family Kyopodiidae and assign it to the suborder Myostaurida (Figs. 7 and 8F; Table 7) because K. lamberti has interradial longitudinal muscles associated with the infundibula (Larson, 1988).

Family Lucernariidae Johnston, 1847
Type genus: Lucernaria Müller, 1776 The family Lucernariidae was proposed by Johnston (1847), including only the genus Lucernaria. Whereas Clark (1863) used the name Lucernariae for all of Stauromedusae, Haeckel (1879) was actually the originator of the name Stauromedusae, in which he placed the family Lucernariidae, divided into two subfamilies: 1) Haliclystidae, including the genera Haliclystus and Lucernaria; and 2) Halicyathidae, including Halicyathus (=Halimocyathus) and Craterolophus (Fig. 8B). Carlgren (1935) proposed Lucernariinae as a subfamily of Clark's (1863) family Eleutherocarpidae, including Lucernaria, Haliclystus, and Stenoscyphus (Fig. 8D), and a similar classification was used by Kramp (1961). Kikinger & Salvini-Plawen (1995), and then Daly et al. (2007), used Lucernariidae as a family of suborder Eleutherocarpina and suborder Eleutherocarpida, respectively, including the genera Haliclystus, Stenoscyphus, Lucernaria, and Stylocoronella (Fig. 8E). However, the topologies presented by Collins & Daly (2005) contradicted monophyly of this grouping (cf. Haeckel, 1879, i.e., when including at least Lucernaria and Haliclystus),  Miranda. a pattern corroborated in our results . Accordingly, we propose that Lucernariidae be limited to the genera Lucernaria and Stylocoronella (Figs. 7 and 8F; Table 7). This hypothesis has to be tested further because Stylocoronella has not yet been available for inclusion in our molecular-based phylogenetic analysis (Fig. 7), but it is consistent with the morphological similarities of Lucernaria and Stylocoronella (Table 8). Kikinger & Salvini-Plawen (1995) superficially remarked that Stylocoronella spp. appear to be congeneric with Lucernaria, although they presented a fundamental difference concerning the fate of the primary tentacles. In Lucernaria, the primary tentacles reduce to absent through development (Berrill, 1962), whereas in Stylocoronella the primary tentacles are retained (Table 8) and become integrated among the adradial clusters of the secondary tentacles (Kikinger & Salvini-Plawen, 1995). However, this developmental difference cannot be distinguished in adults, making its application difficult. Additionally, the coronal muscle seems to be vestigial in Stylocoronella (Table 8) (Kikinger & Salvini-Plawen, 1995), but this information needs further observations.

Character state evolution
Stalked jellyfishes have relatively few external characters useful for taxonomy (Hirano, 1997). Consequently, some internal features are also employed to differentiate these animals (Uchida, 1929;Ling, 1937;Ling, 1939;Miranda, Collins & Marques, 2013). However, most of these characters vary intraspecifically and ontogenetically and they have to be assessed and cautiously employed to differentiate species (Miranda, Morandini & Marques, 2009). We review the main characters used in the traditional taxonomy of Staurozoa (Table 8) and interpret their significance based on the new phylogenetic hypothesis for the class (Figs. 3-5 and 7; Table 7).
A preliminary phylogeny based on nuclear and mitochondrial molecular markers suggested that neither Cleistocarpida nor Eleutherocarpida are monophyletic and that the claustrum "is a more labile feature than suspected and that it may have been lost on more than one occasion," and should not be used to diagnose subgroups within the class Staurozoa (Collins & Daly, 2005: 229). These conclusions are corroborated by our analysis Table 8). Most of the genera in the family Haliclystidae (suborder Myostaurida) have claustrum (Depastromorpha, Depastrum, Halimocyathus, and Manania), except the type genus Haliclystus (Fig. 8F; Tables 7 and 8). In addition, species of Craterolophus, family Craterolophidae (suborder Amyostaurida), also have claustrum (Tables 7 and 8), indicating a homoplastic character (Fig. 11).
Claustrum has also been described in the medusa stage of Cubozoa (Thiel, 1966). However, the internal organization of this structure is different between Staurozoa and Cubozoa (gonads associated with the exogon in Cubozoa; Thiel, 1966), and the existence of a typical staurozoan claustrum in Cubozoa is doubtful (Thiel, 1966). Therefore, if the claustrum in Staurozoa is not homologous to the structure in Cubozoa, claustrum appeared at least twice in the evolution of stalked jellyfishes, and it was lost in Haliclystus (Fig. 11, ACCTRAN). Alternatively, if considered a symplesiomorphy of Staurozoa (Collins & Daly, 2005), claustrum was lost in Calvadosia, Haliclystus, and in the clade Lucernaria + Lipkea (most parsimonious reconstruction).

Interradial longitudinal muscles in the peduncle
The stalked jellyfishes can have four interradial longitudinal muscle bundles, formed by epitheliomuscular cells, in the peduncle (Fig. 10A) (Miranda, Collins & Marques, 2013). These muscles have been generally used to distinguish genera and families of Stauromedusae (Table 8). Clark (1863), for example, distinguished the genus Calvadosia from Lucernaria based on the absence and presence of these muscles, respectively. Uchida (1929) separated stauromedusae without claustrum into three families, one of them (Kishinouyeidae) without muscles in the peduncle. At the same time, Uchida (1929) divided stauromedusae with claustrum into two subfamilies, Depastrinae with muscles in the peduncle, and Craterolophinae without these muscles (Fig. 8C).
Additionally, Uchida (1929) proposed using the shape of the muscle in the peduncle as seen in cross-section as a specific character of Haliclystus stejnegeri in relation to its congeners. Gwilliam (1956: 7) accepted the use of the muscular system to differentiate higher hierarchical levels (e.g., genera and families), but considered it virtually impossible to apply at the specific level due to considerable intraspecific variation, and because the shape depends on both the size (age) and degree of contraction of a given specimen.
Accordingly, the muscles in the peduncle have been treated inconsistently in classification schemes for Staurozoa. For instance, Uchida (1929) assigned Kishinouyea and Sasakiella to the family Kishinouyeidae, but incongruously assigned Lucernariopsis to the Haliclystidae, where it stands out by being the only other genus in the family without muscles in the peduncle (Fig. 8C). Finally, Uchida (1973) clearly considered the presence of claustrum as more important than the muscles in the peduncle in classification.
Our phylogenetic hypothesis reveals that Staurozoa can be divided into two main clades : one only with species possessing the four interradial longitudinal muscles in the peduncle, and the other exclusively formed by species without interradial longitudinal muscles in the peduncle (Table 8). Accordingly, we propose two new suborders for class Staurozoa, order Stauromedusae based on the presence and absence of interradial longitudinal muscles in the peduncle, suborder Myostaurida and Amyostaurida,respectively (Figs. 7 and 8F; Table 7). Collins et al. (2006) inferred that four interradial, intramesogleal longitudinal muscles associated with peristomial pits (infundibula) were symplesiomorphic in Staurozoa, and shared by the ancestral staurozoan with some (but not all) other medusozoans, a hypothesis we have used in our reconstruction (Fig. 11). Four intramesogleal muscles are characteristic of polyps of scyphozoans (Thiel, 1966;Marques & Collins, 2004;Collins & Daly, 2005). Cubopolyps also possess intramesogleal muscles, though the number is not fixed (Chapman, 1978;Marques & Collins, 2004). In hydropolyps, the musculature consists of a layer of longitudinal epidermal muscular fibers and circular gastrodermal fibers (Marques & Collins, 2004). According to this hypothesis, the longitudinal interradial muscles in the peduncle were lost in the clade Amyostaurida (Fig. 11). Additional clues to understand the likely evolutionary polarity of this character could come from detailed examination of its ontogenetic origins across Staurozoa. However, few stauropolyps have ever been studied (Wietrzykowski, 1912;Kikinger & Salvini-Plawen, 1995), and there is no information concerning the presence/absence of interradial longitudinal muscles in developing stauropolyps of Amyostaurida.

Chambers in the peduncle
The peduncle of stauromedusae can have four perradial chambers delimited by gastrodermis (Fig. 14A) (Miranda, Collins & Marques, 2013), which are connected apically to the gastrovascular system of the calyx (Berrill, 1963). The number of chambers in the peduncle has been one of the characters most used in the literature to distinguish staurozoan genera (Clark, 1863;Mayer, 1910;Uchida, 1929;Kramp, 1961). The animals can either have one chamber in the peduncle (e.g., Lucernaria; Kramp, 1961); four chambers (e.g., Haliclystus; Kramp, 1961); four chambers in lower section of the peduncle, which fuse to form one chamber medially (e. g., Kishinouyea;Mayer, 1910); or one chamber in lower position with four chambers medially (e.g., some Manania, Larson & Fautin, 1989) (Table 8). When animals have four chambers in the medial position of the peduncle, these chambers fuse apically at the transition between peduncle and calyx (Uchida & Hanaoka, 1933;Miranda, Collins & Marques, 2013). Also, the number of chambers in the peduncle appears to vary during development of different species (Mayer, 1910;Uchida, 1929;Hirano, 1986), which makes its interpretation more complex. For instance, Wietrzykowski (1911) and Wietrzykowski (1912) observed Haliclystus octoradiatus with one chamber until the stage of 32 tentacles, when, progressively, four independent chambers are formed upward. This pattern was later observed in different species of Haliclystus, whose juveniles have a single-chambered peduncle, later divided into four chambers from the base to the top of the peduncle (Hirano, 1986).
Manania is probably the taxon with the widest variation concerning the number of peduncular chambers (Table 8): four chambers were reported throughout the peduncle in Manania distincta, Manania gwilliami, and Manania handi (Kishinouye, 1910;Larson & Fautin, 1989); four chambers medially and one chamber basally (the lower portion of the peduncle) in Manania atlantica and Manania uchidai (Naumov, 1961;Berrill, 1962); and one chamber throughout the peduncle in Manania auricula (Clark, 1863) and Manania hexaradiata (Broch, 1907;Kramp, 1961;Naumov, 1961). However, as the number of chambers in the peduncle in some Manania species is known to vary with ontogeny (Uchida, 1929;Hirano, 1986), the number of chambers is not a robust character to differentiate species and even staurozoan genera. For example, Clark (1863) considered Halimocyathus sufficiently different from Manania, both taxa described by him. One important difference in his descriptions is the four-chambered peduncle in the former, and single-chambered in the latter. However, different species of Manania were also later described with a four-chambered peduncle (Larson & Fautin, 1989). Therefore, as a general rule, even though the number of chambers in the peduncle seems to be an important character, it should be cautiously employed in the taxonomy of staurozoans (Uchida, 1929;Hirano, 1986).
There have also been some misinterpretations of the number of chambers in the peduncle, making it more difficult to employ this character in taxonomy. Calvadosia nagatensis (Mayer, 1910) and Calvadosia hawaiiensis (Edmondson, 1930) were reported with a four-chambered peduncle, but in fact they have one cruciform chamber throughout the peduncle and only at the level of the pedal disk can the four chambers be observed, sometimes separated by an axial canal (Uchida, 1929;Ling, 1939;Larson, 1980). In another example, Haliclystus was suggested to be closely related to Lucernaria  Miranda. because Haliclystus antarcticus and species of Lucernaria were reported to have a single chamber in the peduncle (Mayer, 1910: 536). In actuality, H. antarcticus has four chambers in the peduncle (Pfeffer, 1889;Carlgren, 1930;Miranda, Collins & Marques, 2013).
Ontogenetic data led Uchida (1929: 153) to hypothesize that "the single-chambered condition of the peduncle is more primitive than the four-chambered one." However, there is a broad occurrence of four chambers in peduncles of Staurozoa, present at least in Craterolophus, Depastromorpha, Depastrum, Haliclystus, Halimocyathus, and some Manania, and this state would be a potential synapomorphy of Staurozoa (Fig. 11,  ACCTRAN), as the four perradial chambers in the peduncle of stalked jellyfishes are not found in any other cnidarian life history stage (Collins & Daly, 2005).
In many species, primary tentacles are present in juvenile stauromedusae, but disappear during development (Uchida, 1929;Berrill, 1962;Larson, 1980). This suggested that an "erratic occurrence of these primary tentacles ( : : : ) indicates that they are negligible as diagnostic characters and of small significance" (Elmhirst, 1922: 221, also highlighted by Uchida, 1929. There is fragmented information about this character, at least partly for a widespread lack of observation of young specimens of most species: Lamouroux (1815) reported that primary tentacles are sometimes observed in C. campanulata, probably in juveniles and in abnormal individuals; Uchida (1929), Ling (1939) and Larson (1980) reported the presence of rudiments of primary tentacles in very young specimens of Calvadosia nagatensis and Calvadosia corbini, as was also observed in Craterolophus convolvulus (Gross, 1900;Carlgren, 1935) and in species of Lucernaria (Berrill, 1963;Collins & Daly, 2005).
In some cases, the eight primary tentacles can also be retained throughout the life of the specimen (Fig. 13) and this condition was distinctive for the former genus Sasakiella (Ling, 1937), which comprised two species, presently Calvadosia tsingtaoensis and Calvadosia cruciformis (Table 7). These two species are differentiated by the number of primary tentacles retained, four in perradial positions in C. tsingtaoensis, and eight, in both the perradii and interradii, in C. cruciformis (Ling, 1937: 15). There may be, however, intraspecific variation for the character, probably related to development: in "a few extreme cases examined the four perradial primary tentacles [of C. cruciformis] are clearly seen but the four interradial ones are reduced to short rudiments. In young specimens all eight of them are well developed" (Ling, 1937: 19).
The development of Stylocoronella riedli and Stylocoronella variabilis shows that the primary filiform tentacles persist in these species, but are transformed into capitate tentacles and clustered together with the secondary tentacles at the tips of the adradial arms (Kikinger & Salvini-Plawen, 1995), a condition never reported in other genera of stalked jellyfishes (Table 8).
Particular marginal structures are also found in Lipkea, a genus morphologically quite distinct from all other stauromedusae (Uchida, 1929: 151) (Fig. 1N). Species of Lipkea have a variable number of lobes (or lappets) at the margin of the calyx (Pisani et al., 2007). Lipkea ruspoliana was described with perradial and interradial lobes, which were suggested to be homologous to the eight primary tentacles, not to the arms of other stauromedusae that are normally adradial (Uchida, 1929). According to this hypothesis, lobes would be highly metamorphosed primary tentacles (Uchida, 1929). However, L. sturdzii and L. stephensoni were described with adradial lobes (Antipa, 1893;Carlgren, 1933). The homology between lobes and primary tentacles was then questioned by Carlgren (1933), who referred to the lobes as modified arms, which was subsequently followed by the description of Lipkea with adradial marginal lobes and without perradial and interradial anchors (Kramp, 1961). Recently, the lobes of L. ruspoliana have been interpreted to be modified tentacles, with an adradial position (Pisani et al., 2007). We consider that the homology of these structures is still under debate, demanding further investigation, particularly of their development.
The shape of anchors has frequently been used in the taxonomy of Haliclystus (Gwilliam, 1956;Miranda, Morandini & Marques, 2009;Kahn et al., 2010). However, their morphology has intraspecific and ontogenetic variation, and consequently it must be carefully assessed when employed to differentiate species of the genus (Miranda, Morandini & Marques, 2009;Kahn et al., 2010).

Pad-like adhesive structures
Pad-like structures can be present individually in the outermost secondary tentacles of the tentacular cluster (Larson & Fautin, 1989), or as a broad structure on the tip of each arm (Larson, 1980;Miranda et al., 2012) (Fig. 15; Table 8). Apparently, the pads help the animal to adhere to its substrate. Calvadosia corbini was observed in situ attached to algae by the pedal disk or by the pad-like adhesive structures on the arms' tips (Larson, 1980). In aquaria, C. corbini mainly use the pads to attach to the substratum, and the relatively large size of the pad compared to the pedal disk makes the importance of this structure for attachment clear (Larson, 1980). The glandular pads located on the anchor and on the abaxial tentacles of Kyopoda lamberti were hypothesized to temporarily serve to reattach the stauromedusae if it becomes detached (Larson, 1988).
There is only scattered information on the ontogeny of the pad-like adhesive structures. They apparently appear in the outermost tentacles late in development of C. cruciformis (Hirano, 1986: 197). Also, the broad adhesive pad-like structure on the tip of each arm hypothetically results from the fusion of several secondary outermost tentacles in C. corbini (Larson, 1980). Pad-like adhesive structures in the outermost tentacles and on the tips of the arms were considered to be homologous by Corbin (1978), but this requires more rigorous study.
This character has already been used to diagnose subfamilies (Carlgren, 1935). However, Carlgren (1935) overlooked the occurrence of pad-like adhesive structures in the outermost tentacles of some species of Haliclystus, which emphasizes the variation of this character within genera (Gwilliam, 1956). The pads in Haliclystus (especially in Haliclystus californiensis; Gwilliam, 1956;Kahn et al., 2010) are never as large as those found in Manania and Calvadosia, but their presence in Haliclystus should be taken into account in considering the relevance of this character for taxonomy.
The presence of these adhesive structures has been used in species descriptions. For instance, Larson (1980) included the pad-like adhesive structures on the tips of the arms as a distinguishing feature of C. corbini. However, he probably overlooked the presence of the structure in C. hawaiiensis because the character is neither well illustrated nor Figure 15 Pad-like adhesive structures. Calvadosia tasmaniensis: (A-B) pad (pa) on the tip of an arm separate from the secondary tentacles (tc); Calvadosia cruxmelitensis: (C) pad (pa) on the tip of an arm, with secondary tentacles (tc) arising directly from it; Craterolophus convolvulus: (D) pads (pa) in the outermost secondary tentacles (tc); Calvadosia vanhoeffeni: (E) pads (pa) in the outermost secondary tentacles (tc); Calvadosia campanulata: (F) pads (pa) in the outermost secondary tentacles (tc). Photo credit: Lucília Miranda. described in the original description by Edmondson (1930), but nevertheless present (Grohmann, Magalhães & Hirano, 1999).

Coronal muscle
The coronal or marginal muscle is a band of epitheliomuscular cells at the calyx margin of stauromedusae (Gwilliam, 1956;Miranda, Collins & Marques, 2013). It is considered a synapomorphy of Medusozoa, probably lost in Hydrozoa , often associated with the swimming movement of jellyfishes (Arai, 1997). In the benthic medusae of Staurozoa, the contraction of the coronal musculature, along with contraction of the longitudinal muscles, considerably reduces the total volume of the animal, probably making its adherence to substrate more efficient in highly hydrodynamic habitats (Hyman, 1940;Miranda, Collins & Marques, 2013).
The position of coronal muscle in relation to the anchor/primary tentacles has also been used in the taxonomy of staurozoans (Carlgren, 1935;Gwilliam, 1956). In Manania, for example, the coronal muscle lies on the exumbrellar (external) side of the anchors (Gwilliam, 1956) (Fig. 12C), whereas in Depastromorpha the coronal muscle lies on the subumbrellar side (internal) of the anchors (Fig. 12D) (Carlgren, 1935). According to Carlgren (1935), only Manania and Depastrum have an external coronal muscle in relation to anchor/primary tentacles, but the phylogenetic signal of this character still has to be tested, specifically when specimens of D. cyathiforme become available for molecular study.
Family Craterolophidae Uchida, 1929 Diagnosis: No interradial longitudinal muscles in peduncle. Peduncle with four perradial chambers. Claustrum present. Without perradial and interradial anchors (rhopalioids) between arms. Individual pad-like adhesive structures can be present in outermost secondary tentacles. Coronal muscle divided.
Family Kishinouyeidae Uchida, 1929 Diagnosis: No interradial longitudinal muscles in peduncle. Peduncle with one central gastric chamber and some species with four chambers at base of peduncle (pedal disk). Claustrum absent. No perradial and interradial anchors (rhopalioids) between arms (C. cruciformis with 4 interradial and 4 perradial primary tentacles, and C. tsingtaoensis with 4 perradial primary tentacles only). Species can have individual pad-like adhesive structures in outermost secondary tentacles or broad pads along tips of arms. Coronal muscle divided.
Our molecular results suggest the probable existence of new species of the genus ( Fig. 7; Calvadosia sp. 1 NZ, Calvadosia sp. 2 NZ, Calvadosia sp. 3 Moorea, Calvadosia sp. 4 SAF), which are being properly collected and/or morphologically analyzed in order to be tested and adequately described.

Genus Depastrum Gosse, 1858
Type species: Depastrum cyathiforme (Sars, 1846) Diagnosis: Four interradial longitudinal muscles in peduncle. Peduncle with four perradial chambers. Claustrum present. No perradial and interradial anchors (rhopalioids) between arms, but one or more primary tentacles on perradius and interradius. No pad-like adhesive structures at secondary tentacles. No discernible arms, but eight (vestigial) sinuosities. Tentacles on each of the eight adradial groups arranged in one or several rows around calyx margin. Coronal muscle entire.
The molecular results show a possible new species from Australia, Haliclystus sp. AUS (Fig. 7), previously identified as Stenoscyphus inabai (McInnes, 1989;Falconer, 2013), which is being collected and morphologically analyzed in order to be properly described.
Halimocyathus platypus was described based on only one specimen (Clark, 1863), and its validity and relationship with Manania spp. still has to be tested in light of molecular and morphological data whenever new material becomes available.

Genus Manania Clark, 1863
Type species: Manania auricula (Fabricius, 1780) Diagnosis: Four interradial longitudinal muscles in peduncle. Peduncle with four perradial chambers, or one central gastric chamber, or one chamber in lower position with four chambers medially. Claustrum present. Perradial and interradial anchors between arms. Adhesive (glandular) cushions surrounding bases of eight anchors, which have knobbed remnants of primary tentacles. Eight short arms. Individual pad-like adhesive structures in outermost secondary tentacles. Entire coronal muscle, external to anchors.

Genus Lipkea Vogt, 1886
Type species: Lipkea ruspoliana Vogt, 1886 Diagnosis: Same as family. Diversity: Three valid species: Lipkea ruspoliana Vogt, 1886; Lipkea sturdzii (Antipa, 1893), and Lipkea stephensoni Carlgren, 1933. The molecular results suggest a possible new species from Japan, Lipkea sp. Japan (Fig. 7), which is being morphologically analyzed in order to be properly described. Unidentified specimens of Lipkea have also been observed in Australia and in New Zealand (Zagal et al., 2011) and the species affinities of these stauromedusae requires further studies.
Daniel Jones (National Oceanography Centre of the UK) for sharing specimens of Lucernaria spp.; and to Kensuke Yanagi (Coastal Branch of Natural History Museum and Institute, Chiba) for sharing information and providing specimens from Japan (Lipkea sp. JAP). We thank Nat Evans (University of Florida) and Robert Wilson (US Geological Survey) for assisting in the generation of some genetic sequences at earlier stages of their careers. We thank Ronald Shimek, Richard Lutz (Rutgers University), Mat Vestjens, and Anne Frijsinger for kindly providing images of stalked jellyfishes (Fig. 1). The Moorea Biocode Project is acknowledged for supporting fieldwork leading to the fortuitous discovery of the first staurozoans known from French Polynesia. We are also grateful to André Morandini for discussions about Medusozoa, and to Tim Collins, Marymegan Daly, and an anonymous reviewer who helped to improve the quality of the manuscript. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Grant Disclosures
The following grant information was disclosed by the authors: NSF: AToL EF-0531779.