Multigene phylogeny of the scyphozoan jellyfish family Pelagiidae reveals that the common U.S. Atlantic sea nettle comprises two distinct species (Chrysaora quinquecirrha and C. chesapeakei)

Background Species of the scyphozoan family Pelagiidae (e.g., Pelagia noctiluca, Chrysaora quinquecirrha) are well-known for impacting fisheries, aquaculture, and tourism, especially for the painful sting they can inflict on swimmers. However, historical taxonomic uncertainty at the genus (e.g., new genus Mawia) and species levels hinders progress in studying their biology and evolutionary adaptations that make them nuisance species, as well as ability to understand and/or mitigate their ecological and economic impacts. Methods We collected nuclear (28S rDNA) and mitochondrial (cytochrome c oxidase I and 16S rDNA) sequence data from individuals of all four pelagiid genera, including 11 of 13 currently recognized species of Chrysaora. To examine species boundaries in the U.S. Atlantic sea nettle Chrysaora quinquecirrha, specimens were included from its entire range along the U.S. Atlantic and Gulf of Mexico coasts, with representatives also examined morphologically (macromorphology and cnidome). Results Phylogenetic analyses show that the genus Chrysaora is paraphyletic with respect to other pelagiid genera. In combined analyses, Mawia, sampled from the coast of Senegal, is most closely related to Sanderia malayensis, and Pelagia forms a close relationship to a clade of Pacific Chrysaora species (Chrysaora achlyos, Chrysaora colorata, Chrysaora fuscescens, and Chrysaora melanaster). Chrysaora quinquecirrha is polyphyletic, with one clade from the U.S. coastal Atlantic and another in U.S. Atlantic estuaries and Gulf of Mexico. These genetic differences are reflected in morphology, e.g., tentacle and lappet number, oral arm length, and nematocyst dimensions. Caribbean sea nettles (Jamaica and Panama) are genetically similar to the U.S. Atlantic estuaries and Gulf of Mexico clade of Chrysaora quinquecirrha. Discussion Our phylogenetic hypothesis for Pelagiidae contradicts current generic definitions, revealing major disagreements between DNA-based and morphology-based phylogenies. A paraphyletic Chrysaora raises systematic questions at the genus level for Pelagiidae; accepting the validity of the recently erected genus Mawia, as well as past genera, will require the creation of additional pelagiid genera. Historical review of the species-delineating genetic and morphological differences indicates that Chrysaora quinquecirrha Desor 1848 applies to the U.S. Coastal Atlantic Chrysaora species (U.S. Atlantic sea nettle), while the name C. chesapeakei Papenfuss 1936 applies to the U.S. Atlantic estuarine and Gulf of Mexico Chrysaora species (Atlantic bay nettle). We provide a detailed redescription, with designation of a neotype for Chrysaora chesapeakei, and clarify the description of Chrysaora quinquecirrha. Since Caribbean Chrysaora are genetically similar to Chrysaora chesapeakei, we provisionally term them Chrysaora c.f. chesapeakei. The presence of Mawia benovici off the coast of Western Africa provides a potential source region for jellyfish introduced into the Adriatic Sea in 2013.


INTRODUCTION
Scyphozoan jellyfishes (Cnidaria, class Scyphozoa), which include the conspicuous moon, lion's mane and sea nettle jellyfishes, exhibit significant and widespread economic and ecological impacts on a wide array of marine and estuarine communities. Jellyfish aggregations, blooms, and swarms damage economically important fisheries, close tourist beaches by stinging swimmers, clog intakes of coastal power and desalination plants, invade ecosystems, and can affect oxygen levels when mass numbers of carcasses are deposited (Arai, 1997;Purcell, Uye & Lo, 2007;Richardson et al., 2009;Bayha & Graham, 2014;Qu et al., 2015). On the other hand, jellyfish serve important roles as major prey items for some fish and sea turtles, in carbon capture and advection to the Deep Ocean, as important microhabitat for fish, invertebrates, and symbiotic algae, and as economic resources for humans (as food and therapeutic compounds) (Doyle et al., 2014, Omori & Nakano, 2001Castro, Santiago & Santana-Ortega, 2002;Arai, 2005;Houghton et al., 2006;Lynam & Brierley, 2007;Ohta et al., 2009;Lebrato et al., 2012;Diaz Briz et al., 2017). Recent attention given to large medusae blooms has led to speculation that anthropogenic events are driving global increases in jellyfish bloom magnitudes, though long-term data sets are still equivocal on this point (Richardson et al., 2009;Brotz & Pauly, 2012;Condon et al., 2013).
Despite their importance, evolutionary and taxonomic relationships of even some of the most recognizable scyphozoan species remain unsettled, which can impede our abilities to effectively study, predict and mitigate the ecological and economic effects of these nuisance species. Recent systematics studies have directly challenged taxonomic relationships at all levels. A mitogenomic analysis recently challenged the placement of the order Coronatae, such as Periphylla, within Scyphozoa (Kayal et al., 2013; but see Kayal et al., 2017) and the new family Drymonematidae was created based on morphological, molecular, and life history data . Studies employing molecular and/or morphological data have revealed novel species in multiple scyphozoan genera, including the moon jellyfish Aurelia (Dawson & Jacobs, 2001;Schroth et al., 2002;Dawson, 2003), the genus Drymonema , the upside down jellyfish Cassiopea (Holland et al., 2004), and the lion's mane jellyfish Cyanea (Dawson, 2005;Kolbasova et al., 2015). Many of these studies have uncovered unrecognized jellyfish invasions and clarified evolutionary relationships in the group (from order to species level) vital to understanding their ecological and economic impacts, and elucidating the evolution of traits that permit these impacts.
The scyphozoan family Pelagiidae (Gegenbaur, 1856), currently made up of four genera (Pelagia, Chrysaora, Sanderia, and Mawia), contains some of the world's most notorious blooming jellyfish. The geographically widespread mauve stinger (Pelagia noctiluca) forms dense aggregations that heavily impact aquaculture, fisheries, and tourism along the North Sea and Mediterranean Sea (Canepa et al., 2014). Recently, a species found for the first time in the Mediterranean was described and assigned first to the genus Pelagia , but later to the novel genus Mawia, based on molecular and morphological data (Avian et al., 2016). Blooms of the jellyfish Chrysaora fulgida (previously identified as Chrysaora hysoscella) have increased over past decades in the Northern Benguela current on the west coast of Africa, coinciding with decreased fish catches and general breakdown of beneficial trophic interactions as compared to nearby ecosystems not jellyfish-dominated (Lynam et al., 2006;Flynn et al., 2012;Roux et al., 2013). Likewise, blooms of very large Chrysaora plocamia medusae form off the coast of Peru, interfering with fisheries, aquaculture, and power plants by clogging nets, seines, and water intakes (Mianzan et al., 2014).
A species of special note is the U.S. Atlantic sea nettle Chrysaora quinquecirrha (Desor, 1848), one of the most recognizable, well-studied, and ecologically important jellyfish along the U.S. Atlantic and Gulf of Mexico coasts Larson, 1976). Because its predation pressure shows ecosystem-wide, controlling influence on zooplankton dynamics (Feigenbaum & Kelly, 1984;Purcell, 1992;, Chrysaora quinquecirrha has been termed a keystone predator for the Chesapeake Bay ecosystem . The jellyfish negatively impacts economically important fisheries by feeding on eggs and larvae Purcell, 1997) and blooms impact tourism by stinging swimmers (Cargo & Schultz, 1966;Schultz & Cargo, 1969;Cargo & King, 1990). As a result, a program was developed to predict both real-time occurrences of sea nettle blooms  and year-to-year bloom magnitudes using past data on environmental conditions (salinity, temperature, etc.) that favor jellyfish populations .
Not long after Agassiz erected Dactylometra, Dactylometra quinquecirrha served to cast doubt on pelagiid generic discrimination. Bigelow (1880) recognized that some brackish water (e.g., Chesapeake Bay) Dactylometra quinquecirrha matured at 24 tentacles (a character of Chrysaora) rather than 40 (a character of Dactylometra), something , saw as the "Chrysaora" stage in their development to the "Dactylometra" stage. Stiasny (1930) also cast doubt on the ability to effectively differentiate Chrysaora and Dactylometra. As a result, Kramp (1955) reasoned Dactylometra and Kuragea to be merely developmental stages and subsumed both within the genus Chrysaora (Eschscholtz, 1829), since it has taxonomic priority.  determined that Chrysaora quinquecirrha went through stages of one to more than seven tentacles per octant, often in the same geographic region, supporting the contentions of  and Kramp (1955). A morphology-based phylogeny of the Pelagiidae (Gershwin & Collins, 2002) indicated two groups coinciding with the previous genera Dactylometra and Chrysaora, but noted that the weak phylogenetic support would make resurrecting the genus Dactylometra premature. Another morphology-based phylogeny (Morandini & Marques, 2010) found support for a Dactylometra clade based on tentacle and lappet number, but noted that this would require many Chrysaora species to have their own genera. A robust phylogenetic hypothesis of relationships within Pelagiidae based on comprehensive taxon sampling is an important step toward removing taxonomic confusion at the genus and species-levels, including assessing the taxonomic status of the new genus Mawia (Avian et al., 2016) and clarifying taxonomic questions related to Chrysaora quinquecirrha.
In order to examine evolutionary relationships and taxonomic boundaries in the family Pelagiidae, with special focus on the genus Chrysaora and the species Chrysaora quinquecirrha, we collected nuclear (large subunit ribosomal rDNA) and mitochondrial (cytochrome c oxidase I and large subunit ribosomal rDNA) sequence data from individuals representing all four extant genera (Chrysaora, Mawia, Pelagia, and Sanderia), including 11 currently recognized species of Chrysaora and one species each of Mawia (Mawia benovici), Pelagia (P. noctiluca), and Sanderia (S. malayensis). To further examine the taxonomy of the U.S. Atlantic sea nettle Chrysaora quinquecirrha, specimens were included from its entire range along the U.S. Atlantic and Gulf of Mexico coasts (estuarine and coastal), taking care to sample all recognized morphotypes, with representatives also examined morphologically (macromorphology and cnidome).

Sample collection
Specimens were collected in the field or at public aquaria husbandry facilities, either by the authors or others with extensive knowledge of Scyphozoa, in an effort to collect as many species of Chrysaora as possible, as well as representative species of Pelagia, Mawia, and Sanderia (Table 1; Fig. 1). An unknown and unidentified pelagiid specimen was collected from Dakar, Senegal and was accompanied by a photograph that did not allow for specific identification (Fig. S1). For Chrysaora quinquecirrha, samples were collected from 10 different sites along the Atlantic and Gulf of Mexico coasts (Table 1; Fig. 2), covering both coastal and estuarine environments, with the intention of capturing as many structural and color morphotypes as possible (Fig. 3). Both white (Table 1: NF1-NF3) and red-striped (Table 1: NF4-NF5) color morphs (Figs. 3C and 3D) were collected from Norfolk, VA (NF). In all cases, a small piece of gonad, tentacle or oral arm tissue was excised and preserved in 80-99% ethanol or DMSO-NaCl solution (Dawson, Raskoff & Jacobs, 1998). Where possible for some sites (Table S1), individuals were also preserved in 4% buffered formalin and seawater for later morphological analyses. Additional published pelagiid sequences were included in the final data set (Table 2).

DNA extraction, PCR amplification and DNA sequencing
Genomic DNA was extracted from preserved tissue samples by CTAB (cetyltrimethylammonium bromide) methods (Ausubel et al., 1989) and stored at -20 C. Polymerase chain reaction (PCR) amplifications targeted three genetic regions: mitochondrial large subunit ribosomal DNA (16S), cytochrome c oxidase subunit I (COI), and nuclear large subunit ribosomal DNA (28S) using primers shown in Table S2. We chose genetic regions that have been useful in examining species boundaries and/or examining genus and family level relationships in the Scyphozoa (Dawson & Jacobs, 2001;Schroth et al., 2002;Holland et al., 2004;Dawson, 2005;Dawson, Gupta & England, 2005;

DI TB
A. RI

Atlantic Ocean
Nantucket Figure 2 Collection locations of Chrysaora quinquecirrha s.l. medusae used in this study. Abbreviations all refer to Table 1 and Table S1. (A-C) are enlargements of rectangular inset regions. The star at Nantucket harbor indicates the type locality of C. quinquecirrha (Desor, 1848). Diamonds represent important museum collection sites (

Phylogenetic reconstruction
For all analyses, Cyanea capillata (Blomsterdalen, Norway) was used as the outgroup because it was shown to be among those scyphozoans least diverged from Pelagiidae . COI sequences were aligned using CLUSTALX v2.1 (Larkin et al., 2007) under default parameters, and checked by eye using their amino acid translations as a guide. 16S and 28S sequences were aligned using MAAFT v7.245 employing the E-INS-I strategy (Katoh & Standley, 2013), since this strategy has been demonstrated to be effective for loci containing conserved motifs embedded within hypervariable regions (Katoh & Toh, 2008). Hypervariable regions of questionable alignment were removed from the MAAFT alignments using GBlocks v0.91b (Castresana, 2000) under default parameters, except that gapped positions were set to half. Phylogenetic analyses were run under maximum likelihood (ML) and Bayesian inference (BI) frameworks for COI, 16S, 28S, and a combined dataset. ML phylogenetic trees were constructed using PhyML v3.0 , employing the best-fit substitution models assessed using jModelTest v2.1.7 (Darriba et al., 2012) under Akaike (AIC) and Bayesian (BIC) information criteria, as well as decision theory performance-based selection (DT). For COI (TPMmf+I+G), 16S (TIM2+I+G), and combined (GTR+I+G) datasets, selection criteria were unanimous, while BIC and DT chose TrNef+I+G for 28S. A 1,000 bootstrap replicate analysis was performed in PhyML to obtain node support values. BI of gene phylogenies was carried out using MrBayes v3.2.6 (Ronquist et al., 2012). The same model of nucleotide evolution (GTR+I+G, with gamma distribution approximated by four discrete categories) was assumed for all analyses, since it is not possible to implement the less complicated models used in the ML tree searches (in the cases of 16S and COI). For each dataset, two independent MCMC runs were conducted until the standard deviation of split frequencies decreased to less than 0.01 (16S: 6,481,000; COI: 19,608,000; 28S: 1,390,000; combined: 1,002,000) generations, sampling every 1,000. The number of generations was determined by assessment of convergence using the minimum estimated sample size and potential scale reduction factor, as implemented in MrBayes. Posterior probabilities were calculated using all trees other than the first 25%, which were discarded as "burnin". All trees were visualized using Figtree v1.4.2 (Rambaut, 2014) and redrawn for presentation using Adobe Illustrator CC v19.1.0 (Adobe Systems, Inc., San Jose, CA, USA). Mean interclade and intraclade, as well as minimum interclade sequence divergence values (Kimura 2-parameter) were determined using MEGAv7.0.14 (Kumar, Stecher & Tamura, 2016) and nucleotide statistics calculated in Seaview v4.6 (Gouy, Guindon & Gascuel, 2010).

Morphological analysis of Chrysaora quinquecirrha
While our study did not include a family-wide morphological analysis, we did perform morphological analyses on jellyfish identified as Chrysaora quinquecirrha from the U.S. Atlantic and Gulf of Mexico coasts. We examined a total of 57 formalin-preserved samples we collected from Charlestown Pond (RI), Cape Henlopen (DE), Rehoboth Bay (DE), York River (VA), Charleston (SC), and Dauphin Island (AL) ( Table S1). In addition, we examined a total of 63 individuals housed at the Smithsonian Institution National Museum of Natural History (USNM) that were collected from the U.S. Atlantic and Gulf of Mexico coasts and identified as Chrysaora quinquecirrha or Chrysaora sp. (Table S1). We examined morphological characters (and their states) previously employed for Pelagiidae (Gershwin & Collins, 2002) that pertained to the medusa stage, with the addition of maximum oral arm length, where preservation state allowed for its measurement (Table 3). In addition, a total of 35 individuals that were examined morphologically, but not included in the phylogenetic analyses, were assigned to molecular species/clades using mitochondrial 16S sequence data collected using the established procedure described above (Table S1).
In agreement with Morandini & Marques (2010), we use the term heterotrichous microbasic rhopaloid, recognizing that there are likely at least two nematocysts that cannot be effectively delineated based on basic light microscopy, as shown in other previous work (Sutton & Burnett, 1969). In all cases, formalin-preserved tentacle tissue was homogenized in distilled water in 1.5 mL microcentrifuge tubes and nematocysts were examined using differential interference contrast microscopy (DIC). A small piece of formalin-fixed tentacle tissue was homogenized in 100 mL of distilled water in a 1.5 mL tube using a plastic microcentrifuge pestle until little visible intact tissue remained. A small drop was then placed on a slide under cover slip and examined at 60Â in DIC using an Olympus BX63 microscope, with photographs taken using an Olympus DP80 camera run by CellSens Dimension 1.13 (Olympus Life Science, Inc., Waltham, MA, USA).  Fig. 8C and Fig. S1. * If two outlier specimens are included, the upper range is six tentacles/octant. Although maximum bell diameter for Chrysaora quinquecirrha has been recorded as great as 40 mm (Gershwin & Collins, 2002;Morandini & Marques, 2010), no animals >20 mm were observed in this study.
A total of 15 individuals were examined for nematocyst size measurements (Table S1). In all cases, 10 samples of each nematocyst type were photographed and later measured using CellSens Dimension 1.13 computer program for length and width. Linear discrimination analysis (LDA) was used to determine whether species could be distinguished on the basis of nematocyst measurements using the lda routine in the R package MASS (Venables & Ripley, 2002).
A total of 10 individuals were examined for nematocyst diversity (Table S1). Since initial estimates indicated that nematocyst diversity varied by tentacle region, nematocyst counts were done from three tentacle regions for each individual: proximal (near the base of the tentacle), medial (in the middle of the tentacle), and distal (at the end of the tentacle). For each region, the first 200 nematocysts were photographed and categorized according to nematocyst type. Only lone nematocysts were enumerated, with any nematocysts still adhering to epithelial tissue ignored, since smaller nematocysts (e.g., a-isorhizas) could be obscured. In order to examine any differences in nematocyst diversity between different tentacle regions (distal, medial, proximal), a mosaic plot showing the relative proportions of nematocyst types in the various regions was made using the R package vcd version 1.4-3 (Meyer, Zeileis & Hornik, 2016). In order to visualize differences in proportions of nematocyst types (four types, three regions) between the two species we conducted non-metric multidimensional scaling of the Euclidean distance matrix using the isoMDS routine in the R package MASS (Venables & Ripley, 2002).

Sequence data characteristics and phylogenetic inference
The COI dataset consisted of 73 sequences, 59 of which are new. All sequences were 616 bp in length. The 16S data set was made up of 67 sequences, including 60 new sequences and 7 published sequences. New complete sequences varied in length from 598 base pairs (bp) for Chrysaora lactea to 608 bp (Chrysaora chinensis). The MAAFT-aligned data set (included published sequences) was 628 bp, but the dataset was truncated to 582 bp (95.7%) after treatment with GBlocks. The 28S dataset included 35 sequences, including 24 new sequences and 11 published sequences. New sequences ranged in size from 998 (Chrysaora chinensis) to 1,018 bp (Chrysaora africana). The MAFFT alignment (which included published sequences) was 1,027 bp, but the final data set was 1,015 bp (98.8%) after removal of regions via GBlocks.
[5.28 ± 0.48] 95% CI tentacles per octant ( Fig. 8A; Table 3). Animals collected from the estuarine Atlantic and all Gulf of Mexico sites (Table S1) had oral arms that were on average 3.00 ± 0.39 (95% CI) times as long as the bell diameter ( Fig. 8B; Table 3). Animals collected from coastal regions of the U.S. Atlantic (Table S1) had oral arms that were on average [1.24 ± 0.27] 95% CI times as long as bell diameter ( Fig. 8B; Table 3). Of the animals that were examined morphologically, a total of 38 individuals were also sequenced for 16S to see which Chrysaora clade they fell into (K2P sequence divergence <1.5%). Medusae examined morphologically that fell into the Chrysaora chesapeakei phylogenetic clade had an average of 2.99 ± 0.03 tentacles per octant and oral arms that were [2.80 ± 0.78] 95% CI times as long as bell diameter on average, while all those that fell in the Chrysaora quinquecirrha clade had an average of 5.63 ± 0.78 tentacles per octant and oral arms that were on average 0.93 ± 0.18 (95% CI) times as long as bell diameter on average (Figs. 8A and 8B). We also studied the cnidome of medusae identified as Chrysaora quinquecirrha, examining the measurements of individual nematocyst types ( Fig. 8C; Fig. S1), as well as the representation of each type overall. Nematocyst measurements indicated significant grouping for holotrichous A-isorhizas, but not for other types. A-isorhiza measurements (length vs. width) showed two distinct groups, with one group containing only animals from U.S. Atlantic estuaries and the Gulf of Mexico and the other containing coastal Atlantic animals (Fig. 8C). All sequenced animals in the smaller group (coastal Atlantic) were genetically similar to Chrysaora quinquecirrha for 16S, while all jellyfish from the larger group (estuarine Atlantic and Gulf of Mexico) that were sequenced for 16S were genetically similar to Chrysaora chesapeakei (Fig. 8C). For animals identified as Chrysaora chesapeakei (based on habitat, macromorphology, and/or genetics), LDA analysis (A) Tentacle counts. Graph represents tentacles per octant against bell diameter (mm) for field collected and museum specimens. Squares represent animals taken from estuarine Atlantic and Gulf of Mexico regions (Chrysaora chesapeakei), while circles represent animals taken from coastal Atlantic regions (Chrysaora quinquecirrha). All animals with 16S sequences matching the Chrysaora chesapeakei clade appear in red, while those whose sequences matched the Chrysaora quinquecirrha clade appear in blue. (B) Maximum oral arm measurements. Graph represents maximum oral arm length against bell diameter (mm) for field-collected and museum specimens. Squares represent animals taken from U.S. Atlantic estuaries and the Gulf of Mexico (Chrysaora chesapeakei), while circles represent animals taken from coastal Atlantic regions (Chrysaora quinquecirrha). Only animals with fully intact and extended oral arms were included. All animals with 16S sequences matching the Chrysaora chesapeakei clade appear in red, while those whose sequences matched the Chrysaora quinquecirrha clade appear in blue. (C) Average size measurements for holotrichous A-isrohiza nematocysts (length vs. width), based on 10 nematocysts per. Error bars represent 95% CI (2 Ã standard error). Squares represent nematocysts from estuarine Atlantic and Gulf of Mexico medusae (Chrysaora chesapeakei), while circles represent nematocysts from coastal Atlantic medusae (Chrysaora quinquecirrha). Photograph of an average sized A-isorhiza from Chrysaora quinquecirrha appears on the left and a photograph of an average size A-isorhiza from Chrysaora chesapeakei appears on the right. Scale bars = 10 mm. Photographs have been resized so that all error bars are the same size on the page to allow size comparisons. All animals with 16S sequences matching the Chrysaora chesapeakei clade appear in red, while those whose sequences matched the Chrysaora quinquecirrha clade appear in blue. Triangles represent average values from  for morphs identified as Dactylometra quinquecirrha (gray) or Dactylometra quinquecirrha var. chesapeakei (white).
indicated that individual A-isorhiza measurements correctly identified species 97.8% of the time (2.2% of the time, they were incorrectly identified at Chrysaora quinquecirrha), while they were correctly identified 100% of the time using the mean of 10 nematocyst measurements. For animals previously identified as Chrysaora quinquecirrha (based on habitat, macromorphology, and/or genetics), LDA correctly identified them 100% of the time, whether one or 10 nematocysts were used. Figure S2 (A-C) shows measurement graphs for a-isorhiza, O-isorhiza, and heterotrichous microbasic rhopaloids, all of which indicate no significant groupings of measurements.

Genus-level systematic inference
Our most robust phylogenetic hypothesis for Pelagiidae (Fig. 7), based on the combined data set, directly contradicts current generic definitions, as well as earlier morphologicalbased phylogenies of the Pelagiidae. Both Gershwin & Collins (2002) and Morandini & Marques (2010) considered Chrysaora to be reciprocally monophyletic with respect to both Sanderia and Pelagia, with Sanderia in a basal position (Figs. 9A and 9B). In contrast, our analyses indicate that Chrysaora is paraphyletic with respect to Pelagia, Sanderia, and the newly erected Mawia . Mediterranean Mawia benovici is not in the combined analysis, but our Senegal pelagiid (SEN) can be treated as Mawia benovici, based on COI (Fig. 4) and 28S (Fig. 6) phylogenies (see below). Paraphyly of Chrysaora is not supported in morphological or genetic analyses in Avian et al. (2016) (Figs. 9C and 9D), but this is almost certainly a result of incomplete taxon sampling. For example, their analysis based on combined morphological and genetic data only included Chrysaora  hysoscella (Mediterranean), while the 28S dataset included a subset of sequences published at the time, (Chrysaora hysoscella, Chrysaora lactea, and Chrysaora c.f. chesapeakei [see below]), all of which occur in a single clade in our analysis (Figs. 7 and 9E). Including fewer published sequences gave the appearance of Chrysaora monophyly, which may have biased the establishment of Mawia. For instance, throughout Avian et al. (2016), Chrysaora is often used as a singular entity (i.e., monophyletic), such as an entire section that examines characters at the "genus level". This more readily allows for the conclusion of a novel genus Mawia, as it sidesteps the difficult taxonomic questions raised by the paraphyly of Chrysaora. That notwithstanding, in agreement with both Piraino et al. (2014) and Avian et al. (2016), our analyses show Mawia benovici to be a close relative of S. malayensis . Given the stark morphological differences between Sanderia and Mawia Avian et al., 2016), this relationship is more than a bit surprising. Our working hypothesis for the relationships within Pelagiidae (Figs. 7 and 9), especially the paraphyletic Chrysaora, raises serious systematic questions for the genus level. To accept the validity of Mawia, as well as previously established Pelagia and Sanderia, each of which can be easily distinguished morphologically from those currently classified as Chrysaora, additional genera would have to be erected within Pelagiidae in order to maintain monophyly of these generic groupings. An initial matter would be to which clade should the genus Chrysaora should be limited. Because the type species of Chrysaora is Chrysaora hysoscella, the genus would best be limited to the clade containing Chrysaora hysoscella, Chrysaora fulgida, Chrysaora lactea, Chrysaora plocamia, Chrysaora quinquecirrha, and Chrysaora chesapeakei (see below). This then would leave three other lineages in need of new genera: (1) Chrysaora africana plus Chrysaora melanaster; (2) Chrysaora chinensis; and (3) Chrysaora achlyos, Chrysaora colorata, and Chrysaora fuscescens. The latter grouping (Chrysaora achlyos, Chrysaora colorata, and Chrysaora fuscescens) has a close relationship to P. noctiluca (except for COI) and there is genetic support for generic designation. Unfortunately, none of the morphological characters employed in this study clearly diagnose this clade or other Chrysaora lineages, as has been the case in other studies seeking to reconcile jellyfish taxonomy based on morphology and molecular data. (Dawson & Martin, 2001;Dawson, 2003;. Future study will benefit from more detailed morphological analyses to identify additional characters that could then be mapped onto molecular phylogenies (e.g., Fig. 7), as well as greater taxonomic sampling (e.g., two additional Chrysaora species accepted and two declared nomen dubium in Morandini & Marques (2010), more geographic samples of Pelagia and Sanderia). Both would allow for better resolution to define genera and better explain their evolutionary relationships.

Interspecific evolutionary relationships and geographic patterns
While our molecular phylogenies bear almost no resemblance to the morphology-based phylogenies within the currently defined genus Chrysaora (Gershwin & Collins, 2002;Morandini & Marques, 2010) (Fig. 9), there are some relationships that occur in all phylogenies. All phylogenies agree on a close relationship between Chrysaora achlyos and Chrysaora colorata (Figs. 9A,9B and 9E). Our phylogeny is in general agreement with Morandini & Marques (2010) in delineating their basal "Pacific" group (Chrysaora achlyos, Chrysaora colorata, Chrysaora fuscescens, Chrysaora melanaster, and Chrysaora plocamia), except that our Chrysaora plocamia samples came from the Atlantic and occur in an "Atlantic" group (Table 1; Fig. 1). Morandini & Marques (2010) reasoned that this basal group may have provided ancient species that then invaded the Atlantic, splitting into various Atlantic groups. Our combined phylogeny (Fig. 7) is in general agreement, with Pacific Chrysaora species generally occupying a more basal position in the tree compared to the Atlantic species. Major disagreements with Morandini & Marques (2010) include the placement of Chrysaora chinensis and Chrysaora pacifica (both Pacific jellies) as closely related to Chrysaora quinquecirrha and Chrysaora lactea, with the Chrysaora pacifica placement also a disagreement with Gershwin & Collins (2002). Likewise, the very close relationship among Chrysaora fulgida, Chrysaora hysoscella, and Chrysaora plocamia was not found in any of the morphological phylogenies ( Fig. 9), though Chrysaora hysoscella and Chrysaora plocamia were closely related in Gershwin & Collins (2002).
One item of note here is our use of aquarium samples, which may be problematic where they are not confirmed with field-collected specimens. Aquarium collected specimens of Chrysaora pacifica (originally Chrysaora melanaster-see below) and Chrysaora chinensis are genetically confirmed, based on published sequences from field-collected specimens of known geographical origin (Figs. 4 and 5). In addition, our aquarium-collected Chrysaora fuscescens is identical to published 16S sequence of field-collected animals from Vancouver Island, Canada (NCBI JX393256). However, Chrysaora colorata, Chrysaora achlyos, and S. malayensis are represented only by aquarium specimens and, therefore, conclusions based on these sequences should be made with care, given questions surrounding geographic provenance and any unnatural interbreeding that might occur in an aquarium system. Future studies incorporating field-collected specimens are necessary for confirming or refuting relationships shown here.
It appears that Bigelow (1880) was correct that Chesapeake Bay Chrysaora that matured at 24 tentacles represented a distinct taxon from Dactylometra quinquecirrha. Our data refute the hypothesis that these individuals represent a growth stage toward the fivetentacled Chrysaora quinquecirrha described from the coast . However, an important point is that it has been claimed that individuals only reach the "five-tentacled" stage after 130 mm , when small tentacles emerge between the secondary tentacles and the rhopalia  Plate 64), termed Stage 5 in . In our data set, only a single individual larger than 130 mm was encountered and collected from the estuarine Atlantic or Gulf of Mexico (Dauphin Island, AL) and it had exactly three tentacles per octant (Fig. 8A). However, it is possible that within the estuarine Atlantic and Gulf of Mexico, these Chrysaora may develop small tertiary tentacles at very large sizes, though they likely never develop fully, as was observed in some animals examined here. Furthermore, in one case,  may have collected Chrysaora from an area (Broadkill River, DE) that experiences both species, albeit at different times of the day, seemingly supporting the hypothesis of development stages. The mouth of the Broadkill River experiences tidal inflows capable of pulling coastal Chrysaora into the inlet during high tide and outflows capable of pulling estuarine Chrysaora from the intercoastal waterway during low tide (K. M. Bayha, 1994Bayha, -2004. In any case, the growth of small tertiary tentacles in large estuarine Atlantic and Gulf of Mexico Chrysaora, along with the dependence on a single morphological character (tentacle number), likely led to the historical taxonomic uncertainties we are clarifying here.
Several lines of evidence support the U.S. Atlantic coastal Chrysaora group retaining the species name Chrysaora quinquecirrha and the estuarine Atlantic/Gulf of Mexico group requiring a different name. First, Pelagia quinquecirrha (=C. quinquecirrha) (Desor, 1848) was described from a coastal zone region (Nantucket Harbor, MA) as having 40 tentacles and our coastal Atlantic animals were characterized by possessing 40 or more tentacles. Furthermore, one of our sampling sites and a museum specimen were from coastal waters (Buzzard's Bay, MA) near the Chrysaora quinquecirrha type locality. Assigning a species name to the U.S. Atlantic estuaries/Gulf of Mexico group is more problematic, owing to inconsistencies in .  compared two color morphs found within the Chesapeake Bay, a small, white morph (e.g., Fig. 3D) and a larger red-striped morph (e.g., Fig. 3E), which the author assumed to be Dactylometra (=Chrysaora) quinquecirrha.  assigned the white morph to the new subspecies Dactylometra quinquecirrha var. chesapeakei, based on very small differences in holotrichous a-isorhiza measurements, though without statistical support. However, for our Norfolk (VA) samples, white (NF1-NF3) and red-striped (NF4-NF5) morphs occurred in the same genetic clades for 16S and COI (Figs. 4 and 5) and we found no overall pattern of differentiation in our holotrichous a-isorhiza measurements (Fig. S2A). Furthermore, for holotrichous A-isorhiza measurements, both morphs from  are consistent with our U.S. Atlantic estuary/Gulf of Mexico group (Fig. 8C). In summary, evidence from nematocyst measurements (Fig. 8C), locality (Chesapeake Bay), and phylogenetic data (Figs. 4 and 5) support the U.S. Atlantic estuarine/Gulf of Mexico group and both morphs from  as representing the same species. Even though  may have been mistaken in describing Dactylometra quinquecirrha var. chesapeakei, that name is taxonomically available based on Article 45.6.4 of the International Code of Zoological Nomenclature (ICZN, 1999). As such, all animals from the U.S. Atlantic estuary/Gulf of Mexico lineage should be assigned to the elevated species name Chrysaora chesapeakei . The placement of Gulf of Mexico medusae in Chrysaora chesapeakei differs from Morandini & Marques (2010), who placed them in the species Chrysaora lactea, based on similarities in octant tentacle orientation (2-3-1-3-2). However, our genetic data clearly separate these animals from the distantly related Chrysaora lactea  and the number of tentacles (approximately three) and lack of tertiary tentacles in the Gulf of Mexico animals observed here and in Morandini & Marques (2010) (USNM 49733 and USNM 53826) make accurate determination of tertiary tentacle orientation problematic.
In addition to their taxonomic value, it is possible that some of the morphological characters that delineate Chrysaora quinquecirrha and Chrysaora chesapeakei may be related to adaptations to different predominant prey items, especially for feeding on the ctenophore Mnemiopsis leidyi. In general, Mnemiopsis leidyi, which is a major prey item for Chrysaora (Feigenbaum & Kelly, 1984), exhibits an inshore, estuarine preference and a seasonal spread from estuarine to coastal waters (Costello et al., 2012;Beulieu et al., 2013). As such, Mnemiopsis leidyi may be a more frequent prey item for estuarine Atlantic Chrysaora than for coastal animals. Larger oral arms, as exhibited in Chrysaora chesapeakei (Fig. 8B), have been argued to be an adaptation for scyphozoans that feed on gelatinous prey . In addition, the larger and more numerous A-isorhiza nematocysts found in estuarine Chrysaora might be better suited to efficiently attaching to and feeding on very soft-bodied organisms such as Mnemiopsis leidyi. Since different nematocyst types are assumed to have different functions based on morphological and discharge characteristics (Rifkin & Endean, 1983;Purcell, 1984;Heeger & Möller, 1987;Purcell & Mills, 1988;Colin & Costello, 2007), it has been proposed that nematocyst diversity within an organism can be correlated to dietary preferences, at least in a coarse sense (Purcell, 1984;Purcell & Mills, 1988;Carrette, Alderslade & Seymour, 2002). In particular, isorhiza nematocysts, which typically serve to entangle hard prey or penetrate soft tissue (Purcell & Mills, 1988;Colin & Costello, 2007), are likely important for feeding on gelatinous prey, since they are the only types found in some jelly-feeding medusae, such as hydrozoan narcomedusae (Purcell & Mills, 1988) and the scyphozoan Drymonema larsoni (KM Bayha, personal observation). A-isorhizas are about twice as numerous in Chrysaora chesapeakei as in Chrysaora quinquecirrha (16.5 ± 3.4% vs. 7.8 ± 3.4%) and are significantly larger (Fig. 8C) in Chrysaora chesapeakei. It is possible that the more numerous A-isorhizas, possessing longer tubules, could penetrate farther into the extremely soft-bodied Mnemiopsis leidyi, resulting in greater capture efficiency.

Chrysaora in the Caribbean
Chrysaora medusae collected from the Caribbean Sea are genetically very similar to Chrysaora chesapeakei. Chrysaora in the Caribbean have historically been included in the species C. lactea Morandini & Marques, 2010), C. quinquecirrha (Perry & Larson, 2004), or Chrysaora sp. (Persad et al., 2003). Our Caribbean samples, limited only to Jamaica and the Bocas del Toro region of Panama, appear to be two lineages (both found in JAM) slightly diverged from each other (4.4-5.1% for COI) and from Chrysaora chesapeakei (6.2-7.7% for COI) from the U.S. east coast estuaries and the Gulf of Mexico. These animals cannot be assigned to Chrysaora lactea (type locality = Rio de Janiero, Brazil), as was previously done by  and Morandini & Marques (2010), since these animals are distantly related to Chrysaora lactea for most genetic regions examined . At present, it is unclear if the Caribbean forms represent distinct or incipient species and further study of them from across the region is necessary. For the time being, we advocate referring to Caribbean animals as Chrysaora c.f. chesapeakei ahead of a formal systematic redescription based on genetic and careful morphological examination.

Chrysaora melanaster and Chrysaora pacifica
Our phylogenetic data confirm the morphological conclusions in Morandini & Marques (2010) that Japanese Chrysaora historically identified as Chrysaora melanaster, and labeled as such in public aquaria worldwide for decades, are actually the distinct species Chrysaora pacifica. Kramp (1961) synonymized the Pacific Chrysaora species C. melanaster (Brandt, 1835) and the Japanese jellyfish C. pacifica (Goette, 1886) to Chrysaora melanaster. This identification convention made it into jellyfish identification books (Wrobel & Mills, 1998) and subsequently Japanese Chrysaora labeled as Chrysaora melanaster became a mainstay in early jellyfish exhibits, such as the Monterey Bay Aquarium (MBA), and then in aquaria throughout the world (W Patry, personal communication). Morandini & Marques (2010) separated Chrysaora melanaster and Chrysaora pacifica on morphological grounds (tentacle and lappet number) and deemed all aquarium specimens of Japanese origin to be Chrysaora pacifica. Our data (Figs. 4 and 5) confirm this, as aquarium-collected jellyfish previously labeled Chrysaora melanaster (MBA) are distantly related to wild-caught Chrysaora melanaster (BER) from its type locality (Bering Sea), but are nearly genetically similar (sequence divergence: COI: 0.5%; 16S: 0.6%) to wild-caught Chrysaora collected from the Eastern Korean coast (KOR), where this jellyfish was recently redescribed as C. pacifica (Lee et al., 2016) and Kyoto, Japan (KYO), both near the type locality of Nagasaki, Japan (Goette, 1886).

Chrysaora africana/fulgida
Our phylogenies support the resurrection of Chrysaora species along the southwestern coast of Africa. Three species of Chrysaora were previously identified from the southwestern coast of Africa: Chrysaora hysoscella (Kramp, 1955), C. fulgida (Reynaud, 1830) and C. africana (Vanhöffen, 1902). However, Kramp (1961) deemed Chrysaora africana a variant of Chrysaora fulgida, and Morandini & Marques (2010) placed all of these sightings within the species Chrysaora fulgida. All phylogenies indicate two distantly related species of Chrysaora from Namibian waters , with those appearing superficially similar to Chrysaora fulgida (brown striped) or to Chrysaora africana (red tentacles) placed provisionally into these species. These designations are consistent with upcoming redescriptions of Chrysaora fulgida and Chrysaora africana of S. Neethling, 2014, unpublished data based on morphological and genetic analyses. Chrysaora has increased over recent years in this area, with concomitant ecological perturbations (Lynam et al., 2006;Flynn et al., 2012;Roux et al., 2013), underscoring the importance of correct species identification.

Mawia benovici
In addition to revealing higher level phylogenetic relationships, our data add to our knowledge regarding the distribution of Mawia benovici, indicating a possible source region for the introduced species. Piraino et al. (2014) hypothesized that Mawia benovici (then Pelagia benovici), likely arrived into the Adriatic Sea via ballast water. Our data indicate that two small pelagiid jellyfishes collected from the beach near Dakar, Senegal are Mawia benovici based on COI and 28S phylogenies (Figs. 4 and 6) (there are no published 16S sequences for Mawia benovici). While this is not definitive evidence that Mediterranean Mawia benovici populations originated from the western coast of Africa, it raises the possibility. While many West African species have arrived in the Mediterranean through the Strait of Gibraltar or occasionally inhabit the Western Mediterranean (Gofas & Zenetos, 2003;Antit, Gofas & Azzouna, 2010), there are examples of animals introduced via shipping or fishing practices from West Africa to the Mediterranean (Ben Souissi et al., 2004;Antit, Gofas & Azzouna, 2010;Luque et al., 2012;Zenetos et al., 2012). If Mawia benovici did originate from the western coast of Africa, it is more likely that it was a result of shipping or fishing practices, since there are no records of Mawia benovici between Gibraltar and the Adriatic Sea to our knowledge.
Description of other specimens: Bell diameter up to approximately 20 cm (observed 17.0-175.0 mm), almost hemispherical but flattened in small individuals. Exumbrella finely granulated with small, inconspicuous marks (papillae); exumbrellar color varies considerably, varying from all white to a completely brown or red colored bell, to a bell with radial lines of red/brown with a spot in the center of the bell. Radial lines may be relatively inconspicuous without a noticeable spot in the center. Tentacles typically number 24 or 3 per octant (average 3.07 ± 0.07), with primary tentacle surrounded by two secondary tentacles (2-1-2), primary tentacle longer than secondary tentacles, up to 3-4 times bell diameter. In some rare cases, small tentacles may occur laterally to secondary tentacle, occurring between the secondary tentacle and rhopalium. In almost all cases, this tentacle is similar in size to or smaller than the lappets surrounding it. In very rare cases (twice observed), about five or more tentacles per octant have been seen, though these medusae had aberrant tentacle patterns overall (e.g., more than one tentacle emerging from same spot, tentacles emerging below lappet). Tentacles are white, slightly pinkish. Marginal lappets rounded and typically 32 or 4 per octant (average 4.08 ± 0.06). Tentacle clefts of varied depth with primary clefts deeper than secondary clefts, which are deeper than rare tertiary clefts. Radial and ring musculature not obvious. Brachial disc circular. Pillars evident. No quadralinga. Subgenital ostia rounded, approximately 1/10 of bell diameter. Oral arms v-shaped with frills emanating from tube-like structure; straight without spiral; curved, frilled edges taper toward proximal end of oral arms. Oral arms long, approximately three times bell diameter on average (as much as 5.6 times bell diameter). Oral arms vary in color, from transparent white, to red or brown colored tubule surrounded by pinkish frilled edges. Four semi-circular gonads, white, pinkish or slightly orange, well developed within pouch outlining gastric filaments. About 16 stomach pouches bounded by 16 septae. Septae bent at 45 angle distally toward the rhopalia terminating near tentacle in rhopalar lappet, resulting in tentacular pouches being somewhat larger than rhopalar pouches distally.