The Phylogeny of Carangiform Fishes: Morphological and Genomic Investigations of a New Fish Clade

Taxon sampling.—

The choice of taxa is a critical component of any phylogenetic study, particularly for one focusing on a species-rich and morphologically diverse group like the Carangiformes. Many researchers have argued that the “groundplan” or “exemplar” approach is most appropriate when conducting a study such as the one herein, particularly when morphological data are included (e.g., Yeates, 1995; Prendini, 2000, 2001). Our focus is the interrelationships of the families of carangiform fishes exclusive of the species-rich flatfishes (≈73% of carangiform species diversity). As the monophyly of the Pleuronectoidea has been consistently recovered in morphological and molecular analyses (e.g., Chapleau, 1993; Harrington et al., 2016; Byrne et al., 2018), we only sampled two families from this clade to reduce the noise in the morphological matrix that would be created by densely sampling the asymmetric and highly aberrant flatfishes. Therefore, we took an exemplar approach when selecting taxa for our study, emphasizing morphological diversity and taking previous studies into account. Once selected, taxa were used in two matrices: one matrix composed of discrete morphological characters and a second matrix comprising DNA sequences of ultraconserved element loci (e.g., Harrington et al., 2016; hereafter, UCEs).

A total of 35 taxa, which includes eight outgroup taxa and 27 representatives of the carangiform fishes, were included in our morphological matrix. To provide a thorough test of carangiform monophyly and relationships, eight outgroup taxa were chosen based on the results of previous morphology-based (Starks, 1899; Johnson, 1984, 1993; Kang et al., 2017) and DNA-based studies (e.g., Wainwright et al., 2012; Betancur-R. et al., 2013a, 2013b; Near et al., 2013; Harrington et al., 2016; Rabsoky et al., 2018). These outgroups included taxa from the Centrarchidae, Channidae, Mugilidae, Nandidae, Percidae, Polycentridae, Sciaenidae, and Scombridae. Within the Carangiformes, all non-flatfish carangiform families, which include Carangidae, Centropomidae, Coryphaenidae, Echeneidae, Istiophoridae, Lactariidae, Latidae, Leptobramidae, Menidae, Nematistiidae, Polynemidae, Rachycentridae, Sphyraenidae, Toxotidae, and Xiphiidae, were sampled in our study. We included at least one representative of each of the four subfamilies of Carangidae to provide a limited test of the monophyly of the family, as members of the Echeneoidea have been found nested within the Carangidae in previous DNA-based hypotheses (e.g., Smith and Wheeler, 2006; Harrington et al., 2016; Rabosky et al., 2018). However, the limits and relationships within the Carangidae (sensuNelson, 2006) were beyond the focus of this study. Additionally, many studies have recovered the non-monophyly of flatfishes (e.g., Smith and Wheeler, 2006; Li et al., 2009; Near et al., 2012, 2013; Betancur-R. et al., 2013b; Campbell et al., 2013), recovering Psettodidae as a separate and non-sister lineage from the Pleuronectoidea. Despite the recovery of flatfish non-monophyly, recent analyses based on either morphology (Chapleau, 1993; Hoshino, 2001) or DNA-sequence data (e.g., Betancur-R. et al., 2013a; Betancur-R. and Ortí, 2014; Campbell et al., 2014a; Harrington et al., 2016; Byrne et al., 2018; Rabosky et al., 2018) have recovered flatfishes as monophyletic. We sampled three flatfish families in our study, including members of the Achiridae, Psettodidae, and Scophthalmidae to provide a limited test of the monophyly of the Pleuronectoideo, but, as noted above, testing the relationships within the Pleuronectoidea was outside of the focus of this study.

The molecular dataset in this study sampled 33 taxa, including eight outgroup taxa and 25 carangiform fishes. Tissue samples of Lactarius lactarius and Psammoperca waigiensis were not available for analysis and were represented in the combined matrix by morphological data exclusively. Every effort was made to match morphological and molecular vouchers at the species level. However, some congeners were used when building these datasets due to the difficulties in obtaining species-specific samples. Four congener taxa, two among the outgroup taxa (Nandus, Scomber) and two among carangiform fishes (Trachinotus, Sphyraena), were represented by different species of the same genus between the morphological and molecular matrices. Taxa represented by congeners were listed by only their genus in the analysis and resulting phylogeny. Lists of taxa used in both the morphological and molecular components of this study can be found in the Material Examined section and Supplemental Tables 1 and 2 (see Data Accessibility), respectively, with symbolic codes for institutional resource collections following Sabaj (2016). Both the morphological and molecular datasets were rooted on the scombroid Scomber.

Collection of morphological data.—

We constructed a novel morphological dataset for this study that included 201 hard and soft tissue characters coded for carangiform and outgroup taxa sampled. Of the 201 characters included in the morphological dataset, approximately 80 characters were either explicitly coded from or modified from the following sources: Bridge, 1896; Gregory and Conrad, 1937; Smith and Bailey, 1962; Gosline, 1968; Smith-Vaniz and Staiger, 1973; Johnson, 1975; Greenwood, 1976; Freihofer, 1978; Fink, 1981; Nakamura, 1983; Collette et al., 1984; Johnson, 1984; Smith-Vaniz, 1984; Feltes, 1986; Johnson, 1986; Bannikov, 1987; Van Neer, 1987; Gushiken, 1988; Chapleau, 1993; Johnson and Patterson, 1993; Roberts, 1993; Leis, 1994; O'Toole, 2002; Otero, 2004; Springer and Smith-Vaniz, 2008; Hilton et al., 2010; Bräger and Moritz, 2016; Harrington et al., 2016; and Kang et al., 2017. An abbreviated version of these character descriptions is listed in Appendix 1, with a more detailed version listed in the Supplemental Appendix (see Data Accessibility). The final morphological matrix (Supplemental Table 1; see Data Accessibility) includes 6,922 of 7,035 possible entries and is thus 98% complete at the level of individual character states. Morphological characters were coded from whole ethanol specimens, disarticulated dry skeletons, and dissected cleared and stained specimens. Cleared and stained specimens prepared by the authors of this study were cleared and double-stained for bone and cartilage following the methods of Potthoff (1984) with the following modifications: Our cartilage staining solutions contained 10 mg of alcian powder per 100 mL of solution for specimens under 80 mm SL and 15 mg of alcian powder per 100 mL of solution for specimens over 80 mm SL. We also used 0.5 g of trypsin powder per 100 mL of solution for digestion. Cleared and doubled stained specimens were dissected following the protocol of Weitzman (1974) as it pertains to the circumorbital series, suspensorium, branchial basket, and shoulder girdle. All specimens coded in the morphological dataset, including specimen preparations and associated catalog numbers, can be found in the Material Examined section.

Imaging of morphology.—

Morphological features were examined primarily with a Leica M205 C microscope, but were occasionally examined with a Nikon SMZ-18 stereomicroscope that has a P2-SHR plan apo 0.5X objective lens. Specimens and morphological features were visualized via digital photography using a variety of imaging techniques. Images were taken using a Nikon D500 with either a Venus Optics Laowa 60 mm f/2.8 2X Ultra-Macro lens or a Venus Optics Laowa 25 mm f/2.8 2.5–5X Ultra-Macro lens. Lighting for imaging was provided by either daylight LED lighting (≤5000 K) or high-energy Royal Blue lighting (440–460 nm) emitted from two NIGHTSEA BlueStar flashlights. To view the stained autofluorescing anatomical features under Royal Blue lighting, we followed the protocol established by Smith et al. (2018b) with the following modification: a 60 mm B+W Dark Red MRC 091M filter was attached to the camera and 60 mm lens combination noted above in order to view the alizarin-stained autofluorescing features. In order to overcome difficulties in positioning specimens for photography and the small depth of field in high magnification images, several images were photographed via z-stacking (also known as focus stacking; see Smith et al., 2018b). Images for z-stacking were taken using the camera, lens, and optional filter combinations listed above with the camera attached to a WeMacro 100 mm focus stacking rail controlled by a Cognisys Stackshot Controller and Helicon Remote v3.9.5. Digital images at different focal distances were then algorithmically combined into a single composite image using Helicon Focus v6.7.1.

DNA extraction, locus amplification, and sequence alignment.—

Prior to the extraction of DNA sequenced for UCE loci, muscle or fin clips were preserved in 95% ethanol, RNAlater Stabilization Solution (Thermo Fisher Scientific), or frozen (fresh) and stored at either –20°C, –80°C, or in liquid nitrogen. Either a DNeasy Tissue Extraction Kit (Qiagen) or a Maxwell RSC Blood DNA Kit and Instrument (Promega) was used to extract DNA from tissue samples following manufacturers' extraction protocols (with the exception of the replacement of the Blood DNA Kit's lysis buffer with Promega's tissue lysis buffer). For Qiagen DNeasy Kit extractions only, the first and second elution from a Qiagen filter were combined and dried to a volume of 102 μL using a Savant DNA120 SpeedVac Concentrator (Thermo Fisher Scientific). For Maxwell RSC extractions only, extractions were eluted into a volume of 102 μL. For both types of preparations, 2 μL of the raw DNA extracts were quantified using a Qubit Fluorometer 2.0 (Thermo Fisher Scientific) using the dsDNA BR Assay Kit (Thermo Fisher Scientific). When insufficient DNA was collected, multiple samples from the same specimen were extracted, combined, dried, and quantified again using the same methods stated above. Final quantified samples (100 μL in volume) were sent to Arbor Biosciences for library preparation (e.g., DNA shearing, size selection, cleanup), target capture, enrichment, sequencing on an Illumina HiSeq 2500 (Illumina), and demultiplexing. Target capture for UCE loci used the 500 UCE actinopterygian-loci probe set (Faircloth et al., 2013).

Demultiplexed sequence data from multiple runs were received in compressed FASTQ format from Arbor Biosciences. These data were uncompressed and combined into two read files per taxon. Data were then cleaned of adapter contamination and low-quality bases using the parallel wrapper illumiprocessor v2.0.7 (Faircloth, 2013) around trimmomatic v0.36 (Bolger et al., 2014). These cleaned sequencing reads were submitted to GenBank and have been assigned SRA Accession Numbers SRR11016325–SRR11016348 under BioProject PRJNA604383 (Supplemental Table 2; see Data Accessibility). Cleaned reads were then combined with previously published UCE data obtained from Harrington et al. (2016; BioProject Accession Number PRJNA341709; Supplemental Table 2; see Data Accessibility) and Alfaro et al. (2018; BioProject Accession Number PRJNA348720; Supplemental Table 2; see Data Accessibility). Assembly of all clean reads was completed using a python script (assemblo_abyss.py) within PHYLUCE v1.5.0 (Faircloth et al., 2012; Faircloth, 2016) in conjunction with ABySS v1.3.7 (Simpson et al., 2009). The k-mer parameter in ABySS was set to a value of 60. In order to identify taxon-specific contigs within the assembled UCE loci, contigs were aligned and assembled into a relational database containing all probes using a python script (match_contigs_to_loci.py, PHYLUCE) and LASTZ v1.02.00 (Harris, 2007). Minimum coverage and minimum identity for identifying UCEs were set to 80%. The PHYLUCE script get_match_counts.py was then used to search the relational database and generate a list of UCE loci shared among all taxa. This list was input into the PHYLUCE script get_fastas_from_match_counts.py to create a single FASTA file containing all UCE sequence data for all taxa. The data in this file were divided by locus using explode_get_fastas_file.py within PHYLUCE and then aligned using MAFFT v7 (Katoh and Standley, 2013). Each locus alignment that contained data from a minimum of 21 taxa (≈65% complete) was converted into PHYLIP-format files and prepared for analyses.

Partitioning schemes and nucleotide substitution models.—

A total of 463 aligned UCE loci were analyzed in our study. Across all UCE loci, mean sequence fragment length was approximately 1,000 bp, with a range of 110–5,649 bp (Supplemental Table 2; see Data Accessibility). All UCE loci were concatenated into a single matrix. The resulting matrix was minimally 65% complete at the locus level with a final length of 409,406 bps, including 76,837 parsimony informative sites. This matrix was partitioned using the Sliding-Window Site Characteristics–Entropy method (hereafter, SWSC-EN; Tagliacollo and Lanfear, 2018) to split each UCE locus into left and right flanking regions and the ultraconserved core (i.e., center segment) by rate of evolution. The resulting left, central, and right UCE segments from SWSC-EN were then used as input for PartitionFinder v2.1.1 (Lanfear et al., 2014, 2017; Stamatakis, 2014) to find the best-fitting nucleotide substitution model for each data partition. PartitionFinder selected among models using AICc and the rclusterf search method with the setting -raxml (Lanfear et al., 2014). PartitionFinder designated 515 subsets with associated models for these regions. A list of the subsets of UCEs, partitions, and associated models can be found in Supplemental Table 3 (see Data Accessibility).

Combined analysis of morphological and molecular data matrices.—

Once the morphology-based and UCE-based matrices were assembled and partitioned, tree inference was performed by simultaneously analyzing both matrices using IQ-Tree v1.6.11 (Nguyen et al., 2015; Chernomor et al., 2016). Both the morphology-based and UCE-based matrices, along with an independent partition model file, were used as inputs for executions of the IQ-Tree software. The model file consisted of the 515 PartitionFinder designated subsets and associated models for the UCE regions and a single partition and model designated for the morphological data (see Supplemental Table 3; see Data Accessibility). The total number of partitions used in the combined analysis was 516. Tree inference was performed by ten independent executions of the software with the settings perturbation strength (-pers) set to 0.2 and number of unsuccessful iterations to stop (-nstop) set to 1,000 using the above inputs and partitioning scheme. Support for the resulting topology with the highest likelihood score was assessed by generating and analyzing 500 standard bootstrap replicates (-bc). Taxa represented by exclusively morphological data (i.e., Lactarius, Psammoperca) were excluded from the support analysis due to their lack of sequence data. The best-fitting phylogeny (Fig. 2) was then reconciled with the resulting bootstrap replicates (Supplemental Fig. 1; see Data Accessibility). The resulting phylogeny was visualized with FigTree v1.4.3 (Rambaut, 2012).

View Figure

• Download as Powerpoint
• Download Figure

Character optimization.—

With the inferred phylogeny of carangiform fishes, we used the resulting tree topology and our morphological matrix (Supplemental Table 1, Supplemental Fig. 1; see Data Accessibility) as input data for WinClada v1.00.08 (Nixon, 2002) and Mesquite v3.6 (Maddison and Maddison, 2018) to view the evolution of morphological transformations. In WinClada, only unambiguous morphological character state optimizations were visualized using parsimony, and in Mesquite, morphological character-state transformations were optimized using parsimony. The optimization of these characters from WinClada is shown in Figure 3 and Supplemental Figure 2 (see Data Accessibility) and will be discussed in the following sections.

View Figure

• Download as Powerpoint
• Download Figure

Centropomoidei.—

The first and earliest diverging clade of carangiform fishes includes Centropomidae, Lactariidae, Latidae, and Sphyraenidae. The families Centropomidae and Latidae have a complex history of being considered as a single family or in distinct, somewhat related, lineages in morphology-based and DNA-based studies (e.g., Otero, 2004; Li et al., 2011). The traditional Centropomidae (sensuGreenwood, 1976) included four genera, Centropomus, Lates, Psammoperca, and †Eolates, which were allied together based on two anatomical synapomorphies: expansion of the second abdominal neural spine and pored lateral-line scales extending onto the caudal fin. Since Greenwood's (1976) study, the monophyly of the Centropomidae (sensuGreenwood, 1976) has been refuted in both morphology-based (e.g., Mooi and Gill, 1995; Otero, 2004) and DNA-based (e.g., Li et al., 2011; Near et al., 2013) studies. Mooi and Gill (1995) separated Lates and Psammoperca into the Latidae based on differences in the association between the epaxial musculature and the skeletal supports of the dorsal fin. Mooi and Gill (1995) also suggested that Hypopterus and †Eolates are members of the Latidae, but their epaxial musculature was not examined in that study. Subsequently, Otero (2004) supported the separation of Centropomidae and Latidae, noting multiple osteological differences between the taxa in these two families. In addition to these anatomical studies, multiple DNA-based studies have recovered the non-monophyly of Centropomidae (sensuGreenwood, 1976), with Latidae being recovered sister to Lactariidae (Mirande, 2016), Leptobramidae + Nematistiidae (Smith and Wheeler, 2006), Polynemidae (Campbell et al., 2014a), Psettodidae (Near et al., 2013), and Toxotidae (Li et al., 2009; Rabosky et al., 2018). However, a majority of DNA-based studies have found a sister-group relationship between Centropomidae (sensu stricto) and Latidae (e.g., Li et al., 2011; Near et al., 2012; Campbell et al., 2013; Betancur-R. et al., 2013a, 2013b, 2017; Betancur-R. and Ortí, 2014; Harrington et al., 2016). In light of these studies, the Centropomidae is currently regarded as the family of snooks, with all species of Centropomus occurring in waters off the New World. In contrast, the Latidae is composed of the extant Old World genera: Hypopterus, Lates, and Psammoperca.

We recover the Latidae as the earliest diverging lineage of Centropomoidei sister to Centropomidae, Lactariidae, and Sphyraenidae in our combined analysis. While the recovery of these four families in a clade is a novel finding, a subset of these families have been hypothesized as close allies in both morphological (e.g., Greenwood, 1976) and molecular (e.g., Betancur-R. and Ortí, 2014; Harrington et al., 2016; Rabosky et al., 2018) hypotheses. Centropomoidei is supported by seven morphological characters and DNA-sequence data. Some morphological characters supporting this node include: fewer than six spines on horizontal arm of the preopercle (character 23₀), rostral extension of external process on maxilla (character 36₁; Fig. 4A–D), enlargement of the second abdominal neural spine (character 162₁; Fig. 4E–G), procurrent spur present with proximal base of caudal-fin ray preceding procurrent ray shortened (characters 188₁ and 189₁), the presence of a bifurcated gas bladder with anterior extensions (character 193₁), and an enlarged pelvic axial ‘scale' at the point of insertion of the pelvic fin (character 197₁). When present, the external process lies on the dorsal or lateral aspect of the maxilla, originating posteriorly to the rostral head of the maxilla. While we believe that this process serves as a contact point for up to three ligaments, which have been discussed in great detail by Datovo and Vari (2013), little information about the process itself was found in previous studies. Not only does this process vary in its presence or absence among the taxa examined, but it also varies in its size and direction of extension. In a subset of taxa with an external process on the maxilla, the process extends rostrally, forming a shelf-like projection that merges with the rostral head of the maxilla (Fig. 4A–D). We observed this rostral extension of the external process in all of the taxa included in Centropomoidei (Centropomus [Fig. 4A, B], Lactarius, Lates, Psammoperca, and Sphyraena [Fig. 4C, D]), as well as two other included taxa, Lepomis and Mene. The morphology exhibited by Lepomis is similar to what is observed among centropomoids; however, the shelf-like extension of the external process in Mene is directed dorsally in comparison to what is observed in Lates and allies. This shelf-like expansion of the external process is a complex character, particularly in light of the variation in size and orientation of the external process of the maxilla, and deserves a more broad assessment among percomorph fishes to fully assess the distribution of this feature. Another morphological character that optimizes in support of the Centropomoidei is the presence of an enlarged second abdominal neural spine (character 162₁; Fig. 4E–G), which was observed in Centropomus (Fig. 4F), Lactarius (Fig. 4G), Lates (Fig. 4E), and Psammoperca but absent in Sphyraena. This character was first noted in centropomids and latids by Greenwood (1976) and was used as one of the two characters supporting the monophyly of the Centropomidae (sensu lato). In 1995, Mooi and Gill refuted the homology of this character, noting that the second neural spine of Centropomus is broadly expanded over most of its length while the second neural spine of latid species is not closely applied to the first neural spine and is only expanded proximally. We interpreted this character differently or our material varies from that of Mooi and Gill (1995) as it relates to latids. We find that the second neural spine is enlarged throughout its length and closely applied to the first neural spine among latids examined in this study (Fig. 4E). We also observed this expanded second abdominal neural spine in Lactarius (Fig. 4G). Leis (1994) was the first to note the enlargement of the second abdominal neural spine in Lactarius, also noting the close application between the enlarged first and second neural spines. Despite being present in Centropomus, Lactarius, Lates, and Psammoperca, an enlarged second neural spine was not observed in Sphyraena. Sphyraenids do exhibit a robust and enlarged first neural spine, but the second neural spine is not any more enlarged than the neural spine on the third or more posterior abdominal vertebrae. We interpret this as a reversal in this character for the sphyraenids. Despite this reversal in one genus, our analysis recovers the presence of an enlarged second abdominal neural spine supporting the monophyly of Centropomoidei. A final morphological feature optimizing in support of Centropomoidei is the presence of a bifurcated gas bladder with anterior extensions (193₁), which are present in all centropomoid taxa examined except Centropomus armatus. Meek and Hildebrand (1923) were the first to note a bifurcated gas bladder in all five members of Sphyraena they examined, but they did not note the nature of this bifurcation among these taxa. We observed one pair of robust anterior gas bladder extensions, each positioned between the shoulder girdle and the base of the neurocranium, in Sphyraena idiastes. While anterior extensions were observed in S. barracuda, they were less robust than those observed in S. idiastes. Meek and Hildebrand (1925) also surveyed the gas bladder shape in eight species of Centropomus, finding anterior extensions of the gas bladder in half of the centropomids surveyed. They reported that these extensions varied in length from rudimentary, flanking the posterior aspect of the neurocranium (e.g., C. robalito), to elongated appendages, curving posteriorly, lying along lateral sides of the gas bladder (e.g., C. undecimalis). Among the remaining taxa recovered in Centropomoidei, anterior extensions of the gas bladder were found in Lates and Psammoperca, as noted by Greenwood (1976), and Lactarius, as noted by Leis (1994). Anterior gas bladder extensions have been observed in other carangiform and outgroup taxa sampled in this study (i.e., Micropogonias, Mene, and Nematistius) as well as many other taxa we did not sample (see Tominaga et al., 1996; Webb et al., 2006). While anterior gas bladder extensions have likely evolved numerous times across teleosts, the presence of these extensions supports the relationship among taxa in the Centropomoidei.

View Figure

• Download as Powerpoint
• Download Figure

Within Centropomoidei, the recovery of Lactariidae sister to Sphyraenidae is novel from a morphological perspective, but the relationship was recovered in two previous DNA-based studies (Betancur-R. and Ortí, 2014; Rabosky et al., 2018). Also known as the False Trevally, Lactariidae is a monotypic family native to coastal Indo-Pacific waters (Leis, 1994). While this family has been traditionally regarded as a percoid, it has garnered little attention overall in systematic studies. Only one comparative study has focused on this species (Leis, 1994), finding it a difficult taxon to place based on morphological characters. Leis (1994) listed eight larval and adult characters that suggested a close relationship between a subset of carangoids and Lactarius, as well as four additional characters that suggested a close relationship between a subset of carangoids, Mene, and Lactarius. Leis (1994) also noted that, while these morphological characters were not unique, unreversed synapomorphies, they were strong indicators allying Lactarius to these fishes. Since the initial study by Leis (1994), Lactarius has been included in only a few DNA-based studies, which have recovered Lactarius sister to Menidae (Sanciangco et al., 2016; Betancur-R. et al., 2017), Sphyraenidae (Betancur-R. and Ortí, 2014; Rabosky et al., 2018), or a clade consisting of eight carangiform families (Campbell et al., 2013). The barracudas (Sphyraenidae) have had a comparatively rich systematic history across morphology-based and DNA-based hypotheses. All 27 species of Sphyraena are found in tropical to temperate marine waters throughout the world. Traditional morphological hypotheses have allied the Sphyraenidae with the Mugilidae and Polynemidae (e.g., Starks, 1899; Regan, 1912), with some studies also including the Atherinidae and Phallostethidae as close allies (e.g., Myers, 1928; Hubbs, 1944; Gosline, 1962). These studies primarily ally sphyraenids with these other families based on the posterior displacement of the pelvic girdle and the lack of an interaction between the elements of the pectoral and pelvic girdles. This hypothesis largely persisted in ensuing investigations by Cockerell (1913), Myers (1935), Hubbs (1944), Gosline (1962, 1968, 1971), Rosen (1964), Greenwood et al. (1966), and McAllister (1968) until an alternative hypothesis was suggested by Johnson (1986). In his study, Johnson (1986) allied Sphyraenidae with Scombroidei, based on five characters from the neurocranium, upper branchial arches, oral jaws, and hyoid arch. Following these morphological studies, hypotheses based on DNA-sequence data have consistently recovered Sphyraenidae as a member of the carangiform fishes (e.g., Li et al., 2011; Near et al., 2013; Harrington et al., 2016). However, these DNA-based studies have recovered a number of sister groups to the Sphyraenidae, including all other carangiform fishes (Harrington et al., 2016), Centropomidae (Near et al., 2013; Mirande, 2016; Alfaro et al., 2018), Menidae (Smith and Wheeler, 2006; Campbell et al., 2014a), and Psettodidae (Li et al., 2011).

We recovered Lactariidae sister to Sphyraenidae in our combined analysis. This relationship was supported by 13 morphological characters, including, but not limited to, the ventral process of the coracoid reaching to a similar point or past the ventral plane of cleithrum (character 113₁), the ventral process of coracoid being broadened by laminae at the ventral aspect (character 115₁), and more than two anal-fin pterygiophores anterior to first hemal spine (character 147₂), among others. Additionally, two morphological characters that optimize onto this node relate to the characteristics of the oral teeth: possession of elongated, fang-like teeth in the oral jaws (character 26₁; Fig. 4H, I) and oral teeth that are ankylosed to the premaxilla and dentary (character 27₁; Fig. 4H–J). We define a fang as an elongated and distally tapered caniniform tooth present in the oral jaws. Fangs are found across a diversity of teleosts (e.g., Anoplogaster, Esox, Meiacanthus, Stomias;Springer, 1968; Fink, 1981), but the distribution of fang-like teeth was limited among taxa examined in this study. One of the most striking features of sphyraenids is the series of fangs in the oral jaws, which have been noted by numerous authors in both extant and extinct species of sphyraenids (e.g., Meek and Hildebrand, 1923; Nishimoto and Ohe, 1982; de Sylva, 1984a; Johnson, 1986; Santini et al., 2015; Fig. 4H). A few authors have also characterized the overall shape of the fangs (de Sylva, 1984a), fang attachment to the oral jaw (Fink, 1981), and replacement of these fangs (Johnson, 1986) in previous studies. In his survey of tooth attachment types across actinopterygians, Fink (1981) found that the attachment of sphyraenid fangs to the oral jaws was dissimilar to the modes of attachment commonly found among percomorphs. The functional part of the fang was directly connected to the attachment bone on the oral jaws (i.e., ankylosed; Type 1; Fig. 4H) in sphyraenids rather than separated from the attachment bone in the oral jaws by a band of collagen that vary in shape (Types 2–4 [Fink, 1981]). Johnson (1986) characterized this morphology further within acropomatids, sphyraenids, and some scombroids, finding the replacement of sphyraenid fangs similar to the replacement of oral teeth in scombroids and allies. Fangs with a type 1 tooth attachment are also found in the oral jaws of False Trevally (Lactarius; Fig. 4I, J) and spiny turbot (Psettodes) among carangiform fishes (characters 26₁ and 27₁). Leis (1994) noted a cluster of two or three enlarged canine teeth surrounding the symphyses of both the dentary and premaxilla in larval and adult stages of Lactarius. He noted that these fang-like teeth are attached to the oral jaws via a type 1 attachment, with the posterior, non-fang-like teeth attached via a type 2 tooth attachment (Leis, 1994). However, Leis (1994) did not note if a type 2 tooth attachment was found in exclusively the larval or adult stages or found throughout the ontogeny of Lactarius. In the adult specimens of Lactarius we examined, we did not observe a collagen band between the functional and attachment parts of the non-fang-like oral teeth of Lactarius and classify all of the oral teeth as attaching via a type 1 tooth attachment (Fig. 4I, J). As larval stages were not examined in this study, we cannot rule out that type 2 tooth attachment occurs prior to the adult stage of Lactarius, and further investigation is needed. Fang-like teeth with a type 1 tooth attachment were also observed in Psettodes among taxa sampled in this study. While fang-like teeth and a type 1 tooth attachment have evolved numerous times throughout the evolution of fishes (Fink, 1981), the distribution of these characters is limited among carangiform fishes. Both the presence of fang-like oral teeth and a type 1 attachment between oral teeth and oral jaws are some characters supporting a relationship between Lactariidae and Sphyraenidae.

All carangiform fishes except Centropomoidei.—

The clade recovered sister to Centropomoidei includes members of the Carangoidei, Menoidei, Nematistioidei, Polynemoidei, and Toxotoidei, which is supported by one morphological character and DNA-sequence data. While previous DNA-based studies have recovered a subset of these families in a clade (e.g., Betancur-R. and Ortí, 2014; Harrington et al., 2016; Alfaro et al., 2018), the recovery of this clade is a novel hypothesis. Despite the extensive amount of external and internal morphological variation seen across the carangiform families not included in Centropomoidei, one morphological character, a dorsally directed external process on the maxilla (character 34₁; Fig. 5A–E), supports this clade. As noted above, the external process may be present on the dorsal or lateral aspect of the maxilla, originating posteriorly to the element's rostral head. Variation occurs in the orientation of this process as it relates to the maxilla, with some fishes exhibiting an external process directed towards the lateral aspect of the maxilla (Fig. 4A–D). Notably, the external process was dorsally directed in all carangiform taxa not included in the Centropomoidei (Fig. 5A–E) with the exception of Xiphioidea, Leptobrama, and Psettodes, which did not possess an external process (character 34₁). Outside of this lineage, a dorsally directed external process was only found in a distantly related sciaenid (Micropogonias). Additional investigations of this process are needed to fully understand the phylogenetic significance and distribution of this character.

View Figure

• Download as Powerpoint
• Download Figure

Polynemoidei.—

Despite pleuronectoids being the species-dominant lineage when compared to Polynemidae, there is a long history of the family Pleuronectidae serving as the root for the order-, suborder- and suprafamily-level designations, referring to all or a subset of flatfish families (i.e., Pleuronectiformes, Pleuronectoidei, Pleuronectoidea). Because of this, and the recovery of Polynemidae as sister to all or a subset of flatfishes in this and other studies (e.g., Harrington et al., 2016; Alfaro et al., 2018; Shi et al., 2018), we have elected to use Polynemoidei to refer to the pleuronectoid–polynemid clade so it is differentiated from other exclusively-flatfish clade names. Accordingly, we have designated Pleuronectoideo to refer to all flatfish families at the infraordinal level following the infraorder endings used by Tyler (1980).

There are approximately 800 species of flatfishes in roughly 15 families that are notable for the migration of either the left or right eye onto or across the dorsal midline of the neurocranium, resulting in an asymmetric baüplan. Despite this striking character, the monophyly of flatfishes has been debated. Early morphological studies by Kyle (1923) were some of the first to suggest multiple origins of flatfishes, allying members of Symphurus with members of the rattails (Macrouridae) and ribbonfishes (Trachipteridae), “Rhomoids” allied with pomfrets (Pampus), and “bothids” placed near the Psettodidae (Kyle, 1923). Similar suggestions of independent flatfish origins have also been made by Chabanaud (1949) and Amaoka (1969). In contrast, most authors (e.g., Rosen and Patterson, 1969; Rosen, 1973; Hensley and Ahlstrom, 1984) have included all flatfishes in the monophyletic Pleuronectiformes (sensuHensley and Ahlstrom, 1984). Flatfish monophyly has been recovered in recent morphology-based studies as well, with the cladistic study of flatfishes conducted by Chapleau (1993) recovering a monophyletic flatfishes based on three characters: migration of one eye across the midline, anterior extension of the dorsal fin above the neurocranium, and presence of the recessus orbitalis. Three subsequent studies by Hoshino (2001), Chanet (2003), and Chanet et al. (2004) also recognized the monophyly of flatfishes based on morphological data. Studies based on DNA sequence data have also recovered flatfishes as a clade (e.g., Berendzen and Dimmick, 2002; Betancur-R. and Ortí, 2014; Campbell et al., 2014a; Harrington et al., 2016; Byrne et al., 2018; Rabosky et al., 2018). However, the non-monophyly of flatfishes has been hypothesized in a number of DNA-based studies (Chen et al., 2003; Smith and Wheeler, 2006; Li et al., 2009; Near et al., 2012, 2013; Betancur-R. et al., 2013b; Campbell et al., 2013; Shi et al., 2018). These DNA-based hypotheses typically recover flatfishes in two independent clades (Psettodoidea and Pleuronectoidea) among carangiform taxa.

We recover flatfishes in a monophyletic group, supported by 16 morphological characters and DNA-sequence data. Many of the morphological features supporting flatfish monophyly are those that have been described in previous studies, namely asymmetry of the neurocranium (our character 13₁; Chapleau, 1993 character 1), presence of the pseudomesial bar (our character 14₁; Hoshino, 2006; Harrington et al., 2016, character 3), and asymmetric lateral pigmentation (our character 198₁; Harrington et al., 2016 character 4). One additional morphological character supporting the monophyly of flatfishes is the absence of posterior expansion on the dorsalmost element of the postcleithrum (character 112₁). In all other members of the carangiform fishes sampled, the dorsalmost element of the postcleithrum possessed a posterior lamellar expansion that was limited to the dorsoposterior margin of the element. However, members of the Psettodidae and Scophthalmidae were the only carangiform taxa examined lacking posterior expansion on the dorsalmost element of the postcleithrum. The postcleithrum is lost in Achirus and was coded as inapplicable (-) for this character. While further examination is needed to assess the distribution of this character among the diversity of pleuronectoids, we find the absence of posterior expansion on the dorsalmost element of the postcleithrum to be another morphological feature supporting the monophyly of Pleuronectoideo.

In addition to the recent debate regarding flatfish monophyly, the placement of Pleuronectoideo as a clade has been difficult from a morphological perspective due to the unique morphological features found in these fishes (Munroe, 2015). While flatfishes have been traditionally placed in their own order, Pleuronectiformes, the placement of this order within the evolution of acanthomorphs has been debated. Kyle (1923) placed a subset of flatfishes with other families of fishes (e.g., Stromateidae, Macrouridae) and Psettodidae as an ancestor of an extinct percoid. Rosen (1973), Hensley and Ahlstrom (1984), Chapleau (1993), and others have placed flatfishes either sister to or nested among the percoids. Studies based on DNA-sequence data have consistently recovered flatfishes among the carangiform fishes (e.g., Betancur-R. et al., 2013a, 2013b; Near et al., 2013; Harrington et al., 2016; Rabosky et al., 2018), finding a number of carangiform lineages sister to the clade. Previous allies of a monophyletic flatfishes include all other members of the Carangiformes (Campbell et al., 2014a; Smith et al., 2016), Centropomidae (Sanciangco et al., 2016), and Polynemidae (Harrington et al., 2016; Alfaro et al., 2018).

Fishes in Polynemidae (42 species) are a morphologically distinctive group commonly called threadfins for their numerous, elongated, and thread-like fin rays positioned on a ventral expansion of their pectoral girdle. Similar to sphyraenids, traditional hypotheses of threadfin interrelationships were primarily influenced by the posterior placement of their pelvic fins relative to other perch-like fishes (e.g., Starks, 1899; Gosline, 1962, 1968, 1971). Interestingly, Gill (1861) not only classified Polynemidae based on this posterior placement of the pelvic fins, but he also put forth an alternative hypothesis which allied polynemids with the drums and croakers (Sciaenidae) based on squamation, weakness of fin spines, and other characters. Later, Freihofer (1978) allied polynemids and sciaenids together based on both groups possessing an anterior extension of the nasal canal rarely observed in other fishes. De Sylva (1984b) noted similarities between the larvae of threadfins and drums, specifically in eye size and myomere count. Later, Johnson (1993) provided corroborative evidence for this relationship noting that polynemids share five of the 21 putative synapomorphies for sciaenids identified by Sasaki (1989), including a character rarely found in other fishes, the interdigitation between the metapterygoid and quadrate. The most recent morphological work on threadfins (Kang et al., 2017) supported the earlier polynemid–sciaenid grouping based largely on the characters put forth by Johnson (1993). Despite the morphological support for grouping polynemids with sciaenids, DNA-based studies have neither recovered these two families as sister lineages nor found these families near each other. While DNA-based studies have recovered polynemids among the carangiform fishes, numerous sister groups to the Polynemidae have been hypothesized, including Latidae (Campbell et al., 2014a; Smith et al., 2016), Menidae (Betancur-R. et al., 2013b; Mirande, 2016), and Sphyraenidae (Dettaï and Lecointre, 2004, 2005). Among the larger and more recent studies, one of the more commonly recovered sister groups to polynemids has been all or a subset of pleuronectoids (Harrington et al., 2016; Alfaro et al., 2018; Shi et al., 2018); we corroborate this hypothesis in this study.

We recover Polynemidae sister to a monophyletic flatfishes and support this relationship with four morphological characters and DNA sequence data, including, but not limited to, a basihyal that is shorter than hypobranchial one (character 62₀), and distal tips of posterior abdominal parapophyses directed medially and joined with the opposing parapophysis on the same vertebra, often forming a single spine-like projection with a bifurcating distal tip (character 163₁). An additional morphological character supporting this sister-group relationship is a prominent anterodorsal expansion of the parasphenoid found in all polynemids and flatfishes examined (character 8₁; Fig. 6A–C). The polynemid parasphenoid includes a prominent anterodorsal extension that begins underneath the region where the belophragm of the basisphenoid contacts the parasphenoid in the posterior half of the orbit (Fig. 6A, C). This expansion reaches dorsally and rostrally toward the frontal, which is ventrally bent in polynemids, and the lateral ethmoid in the anterior part of the orbit. A variety of expansions may be present in carangiform fishes and outgroup taxa, but we have not found a morphology that was similar to what is observed in polynemids outside of the pleuronectoids. There are slight expansions of the parasphenoid in Centropomus, Lates, and Scomber, but these expansions do not extend to a similar height dorsally or interact with the lateral ethmoids or frontal (Fig. 6D). While a large dorsal expansion of the parasphenoid can be found in Mene, this expansion is restricted to the posterior aspect of the parasphenoid, and it does not extend rostrally or interact with the neurocranial elements in the anterior region of the orbit. It should also be noted that various expansions of the parasphenoid, pterosphenoid, and ethmoids can be found among some members of the Sciaenidae (Taniguchi, 1969; Sasaki, 1989, his character 28). However, the limited number of groups that have been found to exhibit these expansions (e.g., Argyrosomus, Johnius, Nibea) occupy a derived position among the Sciaenidae across morphology-based and DNA-based studies (Sasaki, 1989; Lo et al., 2015). Upon first observation, the prominent parasphenoid extension found in polynemids was thought to be unique among the Carangiformes; however, Chapleau's (1993) study on the relationships among flatfishes showed (his figs. 4 and 5) that the neurocranium of Psettodes erumei (Fig. 6B) has a prominent anterodorsal extension of the parasphenoid that interacts with the blind-side lateral ethmoid. This extension is readily seen in both the ventral view and lateral view of the blind side of the neurocranium. While this prominent anterodorsal expansion of the parasphenoid was seen in all polynemids and flatfishes examined in this study, we did not observe this expansion in any other fishes sampled. This character provides support for the relationship between pleuronectoids and polynemids.

View Figure

• Download as Powerpoint
• Download Figure

Carangoidei + Menoidei + Nematistioidei + Toxotoidei.—

We recover a clade consisting of Carangoidei, Menoidei, Nematistioidei, and Toxotoidei that is supported by four morphological characters and DNA-sequence data. This clade has not been previously recovered in morphology-based studies, with only a few DNA-based studies that have sampled all of these lineages recovering them as a clade (i.e., Betancur-R. et al., 2013b; Betancur-R. and Ortí, 2014). One morphological character in support of the recovered clade relates to the amount of dentition present on the basihyal (character 65₁, Fig. 7) and branchial arches within the oral cavity. With the exception of Istiophorus, Nematistius, and Trachinotus, all members of this clade exhibit dentition on the basihyal (character 65₁). Basihyal dentition occurs in a number of lineages of fishes, both carangiform (e.g., Eleutheronema) and non-carangiform (e.g., Moronidae, Gill and Leis, 2019) alike. Given our analysis, the presence of basihyal dentition, as well as three other morphological characters and DNA-sequence data, supports the monophyly of the Carangoidei, Menoidei, Nematistioidei, and Toxotoidei.

View Figure

• Download as Powerpoint
• Download Figure

Menoidei + Nematistioidei + Toxotoidei.—

The recovery of a clade consisting of Menoidei, Nematistioidei, and Toxotoidei is novel hypothesis. However, only a few DNA-based studies have included representatives from all of these lineages (i.e., Betancur-R. and Ortí, 2014; Betancur-R. et al., 2017; Rabosky et al., 2018). This clade is supported by DNA-sequence data and two morphological characters, an increased number of circumorbital elements (character 16₁; Fig. 8A–C) and the absence of pored lateral-line scales extending onto the caudal fin (character 196₀). In their review of the subocular shelf of fishes, Smith and Bailey (1962) noted that the number of elements in the circumorbital series is highly conserved throughout teleosts, with six elements typically found in the series. However, these authors also noted that some fishes have undergone reduction or division of these elements, thereby decreasing or increasing the total number of elements in the series, respectively. Evidence of reductions or increases in the number of these elements can be found throughout percomorphs including members of the Carangiformes (e.g., members of Cottidae and Psychrolutidae possess fewer circumorbitals [Yabe, 1985; Girard and Smith, 2016], members of Acanthuridae possess additional circumorbitals [Tyler et al., 1989]). While six or fewer circumorbital elements were found in most of the taxa sampled in this study (character 16₀), the presence of seven or more circumorbital elements (character 16₁; Fig. 8A–C) supports the monophyly of the six families in this clade. Seven circumorbital elements were observed in Toxotoidei (Fig. 8A), while eight or more elements were found in Nematistioidei (Fig. 8B) and Menoidei (Fig. 8C). This is a different condition than what is reported for Nematistius by Rosenblatt and Bell (1976), who report six elements in the series. Among our specimens of Nematistius, we observed upwards of ten circumorbital elements in the series. As noted above, there are percomorphs that exhibit more than six circumorbital elements, but this character supports the relationships among the Menoidei, Nematistioidei, and Toxotoidei. In addition to the number of circumorbital elements, the absence of pored lateral-line scales extending onto the caudal fin (character 196₀) supports the monophyly of these three suborders. We observed pored lateral-line scales extending onto the caudal fin in all carangiform families sampled in this study except representatives of the Menoidei, Nematistioidei (H. J. Walker, pers. comm.), and Toxotoidei. The loss of pored lateral-line scales that extend on the caudal fin supports a relationship within the taxa in Menoidei, Nematistioidei, and Toxotoidei.

View Figure

• Download as Powerpoint
• Download Figure

Toxotoidei.—

The beachsalmons (Leptobramidae) include two species, Leptobrama muelleri and L. pectoralis, which are found in the waters surrounding southern New Guinea and northern Australia (Kimura et al., 2016). These fishes exhibit a set of notable external morphological characters, such as a dorsal-fin base shorter than the anal-fin base, medial fins covered with scales, and a baüplan similar to pempherids and carangids, that has led ichthyologists to ally Leptobrama with a number of percoid families in previous morphology-based studies (e.g., Cockerell, 1913; Tominaga, 1965). Early on, Steindachner (1878) noted Leptobrama was similar in overall shape to Lichia (Carangidae), while dorsal-fin shape and scale characteristics aligned the beachsalmons with the genus Brama. In subsequent studies (e.g., Cockerell, 1913; Ogilby, 1913; Greenwood et al., 1966), Leptobrama was considered a member of the Pempheridae. In contrast, Tominaga (1968) excluded Leptobrama from the Pempheridae and he described a new family, Leptobramidae, that he placed near the Carangidae based on their morphological similarity. Though numerous characters were shared between Carangidae and Leptobramidae, Tominaga (1965) also listed a few other families that exhibited similar morphological traits to Leptobrama (e.g., Scombridae, Toxotidae). Since these morphological studies, virtually all hypotheses based on DNA-sequence data that have sampled Leptobrama have recovered the family as a member of the carangiform fishes (e.g., Smith and Craig, 2007; Betancur-R. et al., 2013a, 2013b). While these DNA-based studies have recovered a number of sister groups to Leptobramidae, including Menidae and Xiphioidea (Rabosky et al., 2018), Nematistiidae (Smith and Wheeler, 2006), and Sphyraenidae (Smith and Craig, 2007), the most common sister group recovered to the leptobramids are the toxotids (Toxotidae; Betancur-R. et al., 2013a, 2013b, 2017; Betancur-R. and Ortí, 2014; Harrington et al., 2016; Sanciangco et al., 2016).

The ten species of archerfishes (Toxotidae) can be variously found throughout the fresh- and brackish-water environments of southeast Asia, northern Australia, and the Indo-Pacific archipelago (Kottelat and Hui, 2018). These fishes are notable for their ability to shoot water from their mouths, knocking terrestrial prey items off balance and into the water to be consumed (Dill, 1977). Similar to the leptobramids, toxotids exhibit a set of morphological characters that have led authors to ally the family with many other families (e.g., Kyphosidae, Monodactylidae, Mok and Shen, 1983). Günther (1860) suggested a number of allies to Toxotidae, from Chaetodontidae to Microcanthidae, based on a diversity of morphological features. Since Günther's work, two explicit groups have been allied with toxotids based on morphological data: McAllister (1968) and Mok and Shen (1983) allied Toxotidae with various squamipennes (sensuMok and Shen, 1983) based on a number of gill arch and scale characteristics, and the analysis by Springer and Orrell (2004) recovered Ambassidae and Centrolophidae allied to Toxotidae based on dorsal gill-arch musculature. Studies based on DNA-sequence data have consistently recovered Toxotidae among the carangiform fishes (e.g., Chen et al., 2007), with allies ranging from Latidae (Li et al., 2009; Rabosky et al., 2018) to Nematistiidae (Li et al., 2011). However, many DNA-based studies have recovered Leptobramidae as the sister group of Toxotidae (Betancur-R. et al., 2013a, 2013b; Betancur-R. and Ortí, 2014; Harrington et al., 2016; Sanciangco et al., 2016).

Our combined analysis recovers Leptobramidae sister to Toxotidae within the Toxotoidei; this clade is supported by 14 morphological characters and DNA-sequence data. Some morphological characters supporting this relationship include, but are not limited to, ventral margin of lachrymal serrated (character 15₁), posteriorly placed basihyal that inserts dorsal to basibranchial one (character 63₁; Fig. 8D vs. 8E), absence of a medial-extrascapular (character 103₀), the ventral processes of the coracoid and cleithrum are distinctly separate (character 114₁), and an anterior dorsal pterygiophore without an associated dorsal spine (character 135₁; Fig. 8F). Aside from members of the Toxotoidei, taxa sampled in this study possessed a basihyal placed rostral to the first basibranchial (e.g., Lates, Echeneis, Sphyraena) or rostral and slightly dorsal to the first basibranchial element (e.g., Coryphaena, Polydactylus, Trachinotus; Fig. 8E). However, Leptobrama and Toxotes exhibit a different condition, where the basihyal lies directly dorsal to basibranchial one, covering the element, and in the case of Toxotes, covering the rostral margin of the second basibranchial as well (Fig. 8D). Accompanying the posterior placement, the basihyal has an elevated position when viewing the branchial basket laterally. We did not find any indication in previous studies that this feature has been found in any other lineages of fishes. Given the results of our combined analysis, the posterior placement of the basihyal supports the monophyly of the Toxotoidei. One of the defining characteristics of acanthomorph fishes are the prominent spine(s) in the anterior portions of the dorsal and anal fins (Johnson and Patterson, 1993). In many percomorph fishes, the first two dorsal spines are borne on the anteriormost compound dorsal pterygiophore (e.g., Anthiadidae, Epinephelidae, Serranidae, Johnson, 1983, 1988). In his study of Leptobrama, Tominaga (1965) noted nine predorsals in the dorsal fin of leptobramids that did not possess an associated external ray or spine. We find that the nine predorsals listed by Tominaga (1965) correspond to three supraneurals (sensuMabee, 1988) and six spineless pterygiophores in Leptobrama. Anterior pterygiophores lacking a spine or ray are rarely found among percomorph fishes. Examples of groups with rayless anterior pterygiophores include Bathyclupea (≈6 spineless pterygiophores; Prokofiev, 2014), some members of the carangid subfamilies Scomberoidinae and Trachinotinae (≈1 spineless pterygiophore; Springer and Smith-Vaniz, 2008), Kurtus (≤3 spineless pterygiophores; Tominaga, 1965), and Toxotes (1–3 spineless pterygiophores, pers. obs.; Fig. 8F). Given the limited distribution of this character among percomorphs and the results of our combined analysis, the presence of one or more spineless anterior pterygiophores that lack an associated spine (character 135₁) is a feature that supports the sister-group relationship between Leptobramidae and Toxotidae.

Menoidei.—

The two families of billfishes, Istiophoridae (10 species) and Xiphiidae (1 species), can be found in pelagic environments throughout the oceans (Nakamura, 1983). These fishes exhibit a suite of morphological features that are rarely found in other lineages of fishes, including an elongated premaxillary rostrum and the physiological ability for regional endothermy (Brill, 1996). Studies based on comparative morphology have largely allied the xiphioids among the traditional scombroids, as well as these scombroids plus Acropomatidae, Dinolestidae, Malakichthyidae, Pomatomidae, Sphyraenidae, and Synagropidae (Johnson, 1986). Despite commonly placing these fishes among scombroids, morphological hypotheses have disagreed on the placement of billfishes among these families. Xiphioidea has been placed as the sister group to Scombridae (Collette and Russo, 1984; Collette et al., 1984) or within the Scombridae and sister to Acanthocybium (Johnson, 1986; Carpenter et al., 1995). Similar to the morphological hypotheses of Nakamura (1983), Potthoff et al. (1986), and van der Straten et al. (2006), hypotheses based on DNA-sequence data have recovered the xiphioids outside of the Scombroidei (e.g., Orrell et al., 2006; Miya et al., 2013) and among the carangiform fishes (e.g., Smith and Wheeler, 2006; Near et al., 2012; Rabosky et al., 2018). Like most members of the Carangiformes, previous hypotheses have recovered a diversity of lineages sister to the xiphioids (e.g., Smith and Wheeler, 2006; Li et al., 2009). However, a majority of DNA-based hypotheses that have included both xiphioid families have recovered this clade sister to the Menidae (Betancur-R. et al., 2013a; Campbell et al., 2013; Betancur-R. and Ortí, 2014; Harrington et al., 2016; Rabosky et al., 2018; Shi et al., 2018).

Represented by one extant species and numerous extinct species, the moonfishes (Menidae) are easily recognized by their laterally compressed, disc-like bodies, uniquely shaped medial-fin rays, and narrow pelvic fins. The extant species, Mene maculata, can be found near the bottom of coastal waters throughout the Indian Ocean, Indo-Pacific archipelago, Taiwan, and Japan (Carpenter et al., 1997). While one early study based on comparative morphology placed Mene near a variety of fishes (Beryciformes and Ophidiiformes, Springer and Johnson, 2004; see Friedman and Johnson, 2005 for review), Leis (1994) allied menids with the Carangoidei based on six morphological characters shared between carangoids and menids. Leis (1994) also noted eight additional characters shared by Lactarius, Mene, and a subset of carangoids, and he tentatively placed Lactariidae sister to Menidae and a subset of carangoids. While early DNA-based studies recovered Menidae sister to all or a subset of families within Carangoidei (e.g., Chen et al., 2003; Dettaï and Lecointre, 2004, 2005), more recent hypotheses based on DNA-sequence data have recovered Menidae sister to the Xiphioidea (e.g., Harrington et al., 2016; Rabosky et al., 2018; Shi et al., 2018).

Our combined analysis recovers Menidae sister to Xiphioidea supported by 13 morphological characters and DNA-sequence data. Some morphological characters supporting this relationship include the absence of vomerine teeth (character 5₁), lateral flaring of the parasphenoid (character 11₁; Fig. 9), lack of a projection on hypobranchial one (character 76₁), more than one accessory ossification associated with extrascapular bones of supratemporal canal (character 106₁), fusion of hypurals and urostyle into a singular plate in adults (characters 185₁, 186₁, and 187₁; Fig. 10B), and the presence of hypurostegy (character 190₁; Fig. 10A, B). In their survey of xiphioid morphology, Gregory and Conrad (1937, their fig. 8) noted and illustrated that the parasphenoid is a dorsoventrally flattened element with laterally flaring lamellar bone among the xiphioids. Nakamura (1983) described a flattened parasphenoid in all billfishes, noting variation in the width of the anterior part of the parasphenoid in various species (his character 7). The parasphenoid is typically a rod-like medial element that interacts with the base of the neurocranium posteriorly and the vomer anteriorly. While the parasphenoid is typically a simple element, the parasphenoid is dorsoventrally flattened into a plate-like element, with laterally extending lamellar bone in members of the Menoidei (Fig. 9A, B). Flaring of the parasphenoid was not observed in any other fishes examined aside from Echeneis, where the parasphenoid was a broad plate merging with the vomer. In the illustrations of Collette and Russo (1984), Acanthocybium and some species of Scomberomorus exhibit a wide parasphenoid when viewed ventrally (their figs. 17 and 19). In lateral view, it would appear that the parasphenoid in these fishes is robust throughout, without the lamellar plate-like bone expanding laterally (Collette and Russo, 1984, their figs. 15 and 16). Collette and Russo (1984) noted that the wide parasphenoid appears “independently” in several other lineages of scombroids, including Cybiosarda, Gymnosarda, Orcynopsis, and Sarda. While we were limited to viewing the osteology of Acanthocybium and Scomberomorus from published illustrations, we currently do not find the morphology exhibited by these taxa to be the same as what is observed in menoids. Further examination is needed to clarify the homology of this character across disparate percomorph clades. A number of caudal skeletal characters also support a relationship among the Menoidei, including the fusion of hypurals and urostyle into a singular plate (characters 185₁, 186₁, and 187₁) and the presence of hypurostegy (character 190₁; Fig. 10A, B). The first through fourth hypurals in Xiphioidea and Mene are fused into a plate along with the urostyle. As noted by Johnson (1986), most percomorphs possess five autogenous hypural elements, with differential fusion occurring between the upper and lower hypurals in some fishes (e.g., scombroids). Fusions between hypurals have been used to support hypotheses of relationships as they relate to scombroids, with Collette et al. (1984) finding the fusion of the upper and lower hypural plates as a synapomorphy for all scombrids and xiphioids except Gasterochisma, Grammatorcynus, and his Scombrini. Alternatively, Johnson (1986) found the fusion of the upper and lower hypural plates as a synapomorphy of Xiphioidea, the Sardini and Thunnini of Collette et al. (1984), as well as Acanthocybium, Grammatorcynus, and Scomberomorus. While fusions among the upper and lower hypurals occur in a number of taxa (see characters 183 and 184), the second and third hypurals were fused together, forming a singular plate, in Menoidei (character 185₁). Further, this hypural plate, which consists of hypurals 1–4, was fused to the urostyle in menoid taxa exclusively (characters 186₁ and 187₁). While other taxa sampled did exhibit fusion between the third hypural, fourth hypural, and urostyle (e.g., Mugil, Nandus, Scophthalmus; character 187₁), fusion was not detected between the first hypural, second hypural, and urostyle in any other lineages examined besides Xiphioidea and Mene. Hypurostegy is the anterior extension of the proximal bases of the caudal-fin rays such that they partially or completely cover the lateral aspects of the hypurals (Fig. 10A–C). Similar to hypural fusion, hypurostegy was found by Collette et al. (1984) and Johnson (1986) to support a clade of Scombridae + Xiphioidea (Fig. 10A, C). Various amounts of hypurostegy can be found within a diversity of fast-swimming carangiform fishes (e.g., Caranx) and percomorphs outside of the Carangiformes (e.g., Scomber [Fig. 10C], Luvarus [Nakamura, 1983; Johnson, 1986]). Given the distribution of hypurostegy among fishes, it is likely that the presence of this character is the result of convergence among large and fast-swimming acanthomorphs (Nakamura, 1983; Johnson, 1986). While we recognize that this feature has likely evolved in unrelated groups of fishes, the results from our combined analysis suggest that the presence of hypurostegy supports the relationship between taxa in the Menoidei.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Carangoidei.—

The amberjacks, trevally, pompanos, leatherjackets, and allies are classified in the diverse family Carangidae, which includes approximately 150 species in four subfamilies. While morphology-based studies have repeatedly recovered the monophyly of Carangidae (e.g., Smith-Vaniz, 1984; Gushiken, 1988) united largely by the posteroventral elongation of the first proximal anal-fin pterygiophore and gap between the second and the third anal-fin spines, only a subset of DNA-based hypotheses that have included representatives from three or more of the four subfamilies have recovered a monophyletic Carangidae (e.g., Reed et al., 2002; Sanciangco et al., 2016; Betancur-R. et al., 2017). Hypotheses based on DNA-sequence data mostly recover the carangid subfamilies Caranginae and Naucratinae as sister lineages with the remaining carangid subfamilies, Trachinotinae and Scomberoidinae, sister to the Echeneoidea (e.g., Smith and Wheeler, 2006; Near et al., 2012; Mirande, 2016). Given the repeated non-monophyly of Carangidae in recent analyses, we refer to the carangid subfamilies Trachinotinae and Scomberoidinae as members of the “Carangidae” (Figs. 2, 3) as future studies targeting the non-monophyly of Carangidae will likely place these two subfamilies in one (Trachinotidae) or two (Trachinotidae and Scomberoididae) separate families that, combined, are sister to Echeneoidea. The Echeneoidea has long been allied with carangid taxa in morphology-based studies, with Smith-Vaniz (1984) describing three osteological synapomorphies for the Carangoidei as defined above.

Although testing carangid monophyly was not an explicit goal of this study, we recovered the Echeneoidea nested within the Carangidae, sister to a clade of Trachinotinae and Scomberoidinae, in our combined analysis. This relationship is supported by two morphological characters and DNA-sequence data, including the possession of a small supramaxilla (character 25₀) and paired toothplates on basibranchial three (character 73₀). One of these morphological characters is the possession of a small supramaxilla (character 25₀) found in echeneoid and trachinotine taxa examined. This character is coded as inapplicable for Echeneis and Oligoplites, as the supramaxilla has been lost in these taxa. While we recognize that the size of the supramaxilla varies across many unrelated groups of fishes (e.g., Lepomis), the results from our combined analysis suggest that a small supramaxilla supports the relationship among the taxa in the Echeneoidea, Scomberoidinae, and Trachinotinae. Additional investigations are needed into the non-monophyly of Carangidae to fully assess the homology of the characters we highlight above and to test the limits of the family.

Patterns of convergence in relation to carangiform fishes.—

The phylogeny of carangiform fishes we recover highlights that morphological characters provide critical insights and help resolve the relationships among these newly recovered DNA-based clades of fishes. The combination of this novel morphological dataset with the topology from the combined analysis depicts a tremendous amount of morphological convergence seen between these and other percomorph fishes that some carangiforms have been historically allied with. For example, it has been hypothesized that Polynemidae and Sciaenidae are closely related based on a number of characters including the interdigitation between the metapterygoid and quadrate (Johnson, 1993; Kang et al., 2017). Many morphological characters were coded similarly between the Polynemidae and Sciaenidae in our dataset, with 22 of 201 characters coded as the same derived character state between these two families. These shared characters include the interdigitation between the metapterygoid and quadrate (Fig. 10D) discussed by previous authors. However, we find that interdigitation between the metapterygoid and quadrate does not support a relationship between Polynemidae and Sciaenidae (character 46), and it is also present in Leptobrama (Fig. 10E). Hypotheses based on DNA-sequence data recover Polynemidae and Sciaenidae as divergent lineages, with a number of hypotheses recovering the Pleuronectoideo as the sister group to Polynemidae (e.g., Harrington et al., 2016; Alfaro et al., 2018; Shi et al., 2018; current study). Making a similar comparison between the character states coded for Polynemidae and Psettodidae, only five characters with derived states are shared between these two lineages. These clades were not considered close allies in previous morphology-based studies. Members of the Pleuronectoideo were traditionally regarded in their own percomorph order due to their highly derived morphology. It is not surprising then that few derived morphological characters are shared between these lineages. However, given our datasets and analysis, the overwhelming morphological similarity between Polynemidae and Sciaenidae is best explained by convergence. Despite this convergence, there are no uniquely derived features shared between the Polynemidae and Sciaenidae in our dataset, and it is only by examining the morphology of these new clades that we recognize the wider distributions of rare features such as the interdigitation of the metapterygoid and quadrate (character 46). In another example, xiphioids have been included within the Scombroidei based on morphological data (Collette et al., 1984; Johnson, 1986) and 13 of 201 characters were coded as having the same derived character state between these groups, highlighting areas of morphological similarity between Scomber and Xiphioidea. However, studies based on DNA-sequence data have repeatedly recovered Menidae as the most recent ally of the Xiphioidea (e.g., Betancur-R. and Ortí, 2014; Harrington et al., 2016; Rabosky et al., 2018; Shi et al., 2018). Mene has been allied to members of the Carangiformes, specifically Carangidae and Lactariidae, based on morphological data (Leis, 1994). When comparing Menidae and Xiphioidea based on the characters coded in this study, 22 characters support the sister-group relationship between these two lineages despite them not being proposed as allies in previous morphology-based studies. It is noteworthy that eight of these characters are shared among all three lineages—Xiphioidea, Mene, and Scomber. One possible explanation for these convergences is that all three of these lineages live in similar oceanic environments and are under similar environmental pressures. As such, it is not surprising that these fishes exhibit numerous convergent morphological features (Fig. 10A–C).

The combined morphological and molecular phylogeny presented in this study provides an opportunity to arbitrate the conflict between morphology-based and DNA-based data, test the phylogenetic implications of different traits present in these fishes, and examine the evolution of morphological characters within this newly recovered clade of fishes. We hope that our findings not only highlight the utility of morphological characters as they relate to the evolution of carangiform fishes but that they also invite subsequent authors to test additional evolutionary hypotheses within this clade. Finally, we encourage other authors to integrate morphological characters with DNA-sequence data to further test the relationships of and explore the evolution of the widely diverse percomorph fishes.