WE redescribe and recognize Etheostoma cyanorum as a valid species, separate from Etheostoma radiosum (Orangebelly Darter; Moore and Rigney, 1952; Collette, 1965; Matthews and Gelwick, 1988), within the larger Etheostoma whipplei (Redfin Darter) species complex, or clade (Retzer et al., 1986; Lang and Mayden, 2006; Near et al., 2011). We review its taxonomic history and recent treatments, supplement and reevaluate previous morphological data, and add molecular genetic data supporting specific recognition. Etheostoma cyanorum is known only from the Blue River drainage in southcentral Oklahoma, USA, a tributary of Red River. The range of E. radiosum, minus E. cyanorum, includes tributaries of the Washita River west of Blue River, and all south-flowing drainages in Oklahoma or Arkansas east of Blue River from the Clear-Muddy Boggy drainage to the upper Ouachita River basin in Arkansas (Fig. 1). In the Saline River of the Ouachita River basin (note there are two “Saline” rivers in Arkansas) Etheostoma artesiae (Redspot Darter) represents the clade. Etheostoma whipplei occurs north of E. cyanorum and E. radiosum, limited in the region to the Arkansas River basin (Retzer et al., 1986; Piller et al., 2001; Robison and Buchanan, 1988, in press). Thus, the four species (including E. cyanorum) of the E. whipplei clade exhibit near complete allopatry, with a narrow zone of sympatry between E. radiosum and E. artesiae above the Fall Line in the Red and Ouachita river drainages (Piller et al., 2001). But in hundreds of samples from southwest Arkansas or southeast Oklahoma by WJM since 1976, no two of the species have been found together.

Fig. 1

Range map of E. cyanorum and E. radiosum in the south-central United States. Etheostoma cyanorum is found only in the Blue River drainage, Oklahoma, USA. Nominal subspecies E. r. radiosum is restricted to the Ouachita and Little River drainages to the east, and E. r. paludosum is restricted to the Kiamichi, Muddy Boggy, Clear Boggy systems, and tributaries of the Red River to the west. Drainages were identified by layering georeferenced collection data ( http://www.fishnet2.net) into GIS software. MtDNA CR (black) and COI (gray) sampling localities are superimposed on drainages.

The Redfin Darter complex, and Etheostoma radiosum cyanorum (Moore and Rigney, 1952) in particular, have a complicated taxonomic history (details in Supplemental Text A; see Data Accessibility). The first species described in the group was Boleichthys whipplii (now = E. whipplei) from Coal Creek, Indian Territory (now Pittsburg County, Oklahoma; Girard, 1859). Jordan and Gilbert (1886) collected specimens within the range of the future E. radiosum in the upper Ouachita River basin in Arkansas, reporting them as Etheostoma whipplei (amended spelling) but noting a “lack of red spots” (unlike E. whipplei). The first collections of E. r. paludosum, as “E. whipplei,” later described as Poecilichthys radiosus paludosus (Moore and Rigney, 1952), were in the Kiamichi River drainage by Seth E. Meek in 1894 (Meek, 1896), and by H. A. Pilsbry in 1903 in the Muddy Boggy drainage (Fowler, 1904). Thus by 1904, Etheostoma whipplei (or whipplii) was viewed as one species in Oklahoma (then Indian Territory), Arkansas, and extreme north Texas, but with considerable variation noted. Ortenburger and Hubbs (1926) reported Poecilichthys whipplii (spelled with two “ii”s) from a collection on 19 June 1925 in Yanubbe Creek, McCurtain County, Oklahoma, comprising the first collection of E. radiosum from the Little River basin. The Blue River drainage was not sampled by ichthyologists until the 1940s, thus P. r. cyanorum (Moore and Rigney, 1952), now E. cyanorum, remained undetected.

Hubbs and Black (1941) recognized Poecilichthys whipplei radiosus, from the Caddo River, Arkansas, as a new subspecies. However, Hubbs and Black (their map, p. 3) included specimens from the (eastern) Saline River in Arkansas and rivers in Texas, now known to have Etheostoma artesiae (Piller et al., 2001) instead of E. radiosum. Moore and Rigney (1952) elevated Poecilichthys radiosus to species and provided diagnoses for three new subspecies: P. r. cyanorum from Blue River, P. r. paludosus from the Clear Boggy and Kiamichi systems, and P. r. radiosus throughout the rest of the range to the east. Moore and Rigney (1952) and Scalet (1971) diagnosed the Blue River subspecies as having a blunter and more “decurved” snout, deeper head, a “larger, heavier body” (but see Results about validity of these traits), fewer unpored scales, and more pored scales than other subspecies. Bailey et al. (1954) placed all darters in three genera, moving Poecilichthys into the genus Etheostoma.

Echelle et al. (1975) assessed lactate dehydrogenase LDH-1 and esterase ES-3 enzymes at multiple sites within each drainage in the range of E. radiosum as then known. For both LDH-1 and ES-1 Blue River fish differed substantially from those in the nearby Muddy Boggy and Kiamichi drainages, but not from populations in southwest Arkansas (figures 1 and 2 of Echelle et al., 1975).

Retzer et al. (1986) compared E. radiosum (as then understood) to E. whipplei, reiterating color characteristics (red spots on flanks of E. whipplei) separating the two species. However, they noted a lack of any overall morphometric differences between the two species in a principal components analysis. Piller et al. (2001) recognized E. artesiae (Hay, 1881) as distinct from E. whipplei due to meristic differences and allopatry and noted the close taxonomic relationship of both to E. radiosum.

Matthews and Gelwick (1988) assessed the subspecies of Moore and Rigney (1952) with geographically dense sampling, including newly discovered E. radiosum in tributaries of the Washita River (Matthews et al., 1986). They recognized substantial morphological differences between specimens from Blue River (then considered E. radiosum) and all other populations including head depth and pored and unpored lateral line scale counts. Matthews and Gelwick (1988) recommended continued recognition of the three subspecies of Moore and Rigney (1952) pending further investigation of variation, and they considered the new populations from the Washita River drainage to align with E. r. paludosum.

Two papers from 2011 took very different approaches, but each has important implications for systematics of E. radiosum and E. cyanorum. Near et al. (2011) compared 245 North American darter (Percidae) species (a priori including as species several recognized subspecies) with one mitochondrial and two nuclear genes for a comprehensive darter phylogeny. Near et al. (2011 and their supplementary material) included as full species one individual each of Etheostoma cyanorum (from lower Blue River near Milburn, Oklahoma); “Etheostoma paludosum” (from Kiamichi River headwaters west of the Oklahoma–Arkansas border; and Etheostoma radiosum (from Caddo River south of Caddo Gap, Arkansas). Near et al. (2011, appendix) erected “Vexillapinna” as a “new clade name” that included E. whipplei, E. artesiae, E. radiosum, E. cyanorum, and E. paludosum. A tree (their figure 3) including all 245 darters, showed E. artesiae sister to a group with all other species in the Vexillapinna clade, within which E. paludosum and E. cyanorum were sister, and E. radiosum and E. whipplei (in part) were sister. However, the analyses of Near et al. (2011) were based on a single individual of each of the three nominal subspecies of E. radiosum, thus did not address geographic variation within those taxa. Nevertheless, Near et al. (2011) represents the first time in the literature that species level names were applied to the recognized subspecies of E. radiosum (Page, 1983; Matthews and Gelwick, 1988).

April et al. (2011) used “DNA barcoding,” following mtDNA protocols of the Fish Barcode of Life Campaign, for 752 species of North American fishes, to seek evidence of unrecognized “cryptic diversity.” Their study included a large number of “E. radiosum” across its range including fish from Blue River (now = E. cyanorum), but not populations in the Washita River system (Matthews et al., 1986). Even with those western populations excluded, April et al. (2011, and their supporting information) detected a high degree of cryptic diversity within E. radiosum, including five separate clusters representing genetically dissimilar taxa. They noted (p. 10606) that such clusters might represent new species but called them “unconfirmed candidate species.”

The results of Near et al. (2011) and April et al. (2011) provide contrasting views of possible systematics within the E. radiosum complex. Near et al. (2011: p. 569) accepted a priori that previously named subspecies of E. radiosum were full species, because those “allopatrically distributed subspecies exhibit morphological differences on the order of those observed between described species.” In contrast, April et al. (2011; specimen data available at  http://www.barcodinglife.com) used broad geographic sampling across most of the known range of E. radiosum. Five clusters (=their “BIN”s) from their analyses showed patterns important for taxonomy of the complex. Their most important finding relative to the current study was that their BIN (BOLD.AAA3140) was restricted to 15 specimens from multiple locations in Blue River, representing E. cyanorum, and no Blue River material was in any other BINs. Their BIN (BOLD.ABZ2860) contained 49 individuals from a geographically broad region, from the Clear and Muddy Boggy system in south Oklahoma to the Ouachita drainage near Hot Springs, Arkansas. A fourth BIN (BOLD.ABZ2859) included 46 individuals from a geographically limited area comprising the Little River basin in southeast Oklahoma and southwest Arkansas, but excluding the Cossatot River, from which all individuals were in a separate, fifth BIN (BOLD.AA3138). In summary, April et al. (2011) found all material from Blue River uniquely contained in one BIN, supporting designation of E. cyanorum as a distinct species.

In the second edition of the Peterson Field Guide to Freshwater Fishes of North America North of Mexico, Page and Burr (2011) included E. artesiae, E. radiosum, and E. whipplei as valid species, noting E. radiosum had the three named subspecies from Moore and Rigney (1952). Finally, given all available information at the time of deliberation by the Committee on Common and Scientific Names of Fishes, Page et al. (2013) recognized E. artesiae, E. radiosum, and E. whipplei as distinct species, but not E. r. cyanorum or E. r. paludosum. We now recommend recognition of E. cyanorum as a valid species but withhold comment on E. r. paludosum or the nominal subspecies until a more detailed geographic molecular study of E. radiosum is completed.

MATERIALS AND METHODS

Morphology.—Meristic counts and measurements for specimens in all drainages having E. radiosum or E. cyanorum were taken from original, paper datasheets on which counts and measurements were recorded by hand for Matthews and Gelwick (1988). We checked a computerized database against those original datasheets, made a few minor corrections detected as typos, and re-examined results from Matthews and Gelwick (1988). Ten counts and 16 measurements (Supplemental Table A; see Data Accessibility) followed Hubbs and Lagler (1964) as modified by Retzer et al. (1986). MedCalc Version 18.9.1 (MedCalc for Windows Version 18.9.1, MedCalc Software, Ostend, Belgium) was used on 331 specimens for meristics and 320 for measurements to calculate mean, standard deviation, and 95% confidence intervals for the mean (equivalent to mean62 standard errors) for E. cyanorum in the Blue River drainage and for 13 populations of E. radiosum across its entire range. Supplemental Table A (see Data Accessibility) and MedCalc calculations included 40 E. cyanorum for which all meristic counts were complete and 39 E. cyanorum for which all measurements were complete.

Additional morphological data.—In addition to counts from Matthews and Gelwick (1988), pored and unpored lateral line scales were counted for another 40 adult males in nuptial color collected in Blue River in August 2017 (BRTP1 and BRTP2) to compare to the Matthews and Gelwick (1988) data for these putative diagnostic characters (per Moore and Rigney, 1952).

Another 11 small to large (38–54 mm SL) adult male (determined by dissection) E. cyanorum collected in Blue River in 1994 by E. Marsh (ECM 94-8) were used to assess shape of the head relative to a “blunt snout,” considered diagnostic of E. r. cyanorum by Moore and Rigney (1952). We measured “angle of the muzzle” (per Hubbs and Black, 1941), which we now call “angle of anterior profile” (AAP) as a better descriptor, emphasizing the angle between the horizontal axis of the body and the predominant head profile. The AAP was measured by placing each fish on a horizontal line level with the anteriormost point of the snout and parallel to the midline of the body, then, viewing from directly above, drawing a straight line at the angle of the dominant slope of the anterior profile or “muzzle” of Hubbs and Black (1941), intersecting the horizontal line to determine the angle (AAP; Supplemental Figure A; see Data Accessibility).

Coloration.—Descriptions of coloration for specimens in all drainages having E. radiosum or E. cyanorum were taken from original color slides of freshly caught material made in the field by WJM, standardized field sheets on which color for fresh specimens was recorded on schematic outlines of fish by F. Gelwick, and comments in field notes supporting Matthews and Gelwick (1988). Another 40 adult males in nuptial color collected in Blue River in August 2017 (BRTP1 and BRTP2, OMNH 86858) were examined for details of the color pattern in the spinous dorsal fin to compare to descriptions of subspecies of E. radiosum in Moore and Rigney (1952) and Matthews and Gelwick (1988). Twelve additional adult E. cyanorum in peak nuptial color and one adult female were collected from Blue River upstream of Connerville, Oklahoma, on 30 March 2018 (WJM 3645), with the largest male (OMNH 86859) and the female (OMNH 86860) photographed in color by N. Lang, within 30 minutes of being narcotized, for the species description (below). The other 11 (OMNH 86861) were used by WJM to further assess the color pattern in the spinous dorsal fin with fresh material.

Genetic methods.—Fifty-two specimens were collected from tributaries of the Red River in southeastern Oklahoma and southwestern Arkansas, USA (Fig. 1). Sampling was focused on the Blue River and geographically adjacent drainages including the Clear Boggy River to the east, and tributaries of the Washita River to the west. Collecting was mainly in upland portions of drainages where E. radiosum is most abundant. Fishes were collected with a seine or backpack electrofisher, anesthetized in MS-222 on site, and either frozen in liquid nitrogen or preserved in 95% ethanol in the field. Quantity and quality of genomic DNA did not differ between sample preservation methods, and all samples were readily amplified by polymerase chain reaction (PCR).

Nucleic acids were isolated from white muscle or fin clips using proteinase-K digestion and phenol/chloroform extractions followed by ethanol precipitation and resuspension in distilled water (Turner, 1997). PCR was conducted with oligonucleotide primers L15926 and H16498 that target a ∼450 bp fragment of the mitochondrial control region (CR) flanked by the mtDNA proline tRNA gene (Kocher et al., 1989; Shields and Kocher, 1991). Amplification occurred in 50 µl reaction volumes containing 5 µl sample DNA template, 5 µl 10X reaction buffer, 200 µM dNTPs, 2 µM MgCl₂, 0.5 units Taq DNA polymerase, and primers added at a final concentration of 0.8 µM. Thermocycling consisted of an initial denaturation step at 94°C for 2 min, followed by 25 cycles of denaturation at 94°C for 30 seconds, annealing at 48°C for 30 seconds, and extension at 72°C for 1 minute. PCR products were sequenced directly in both directions using the BigDye^™ version 3.1 cycle sequencing kit and an ABI Prism 3130 capillary sequencer, in accordance with manufacturer's instructions. Fragments were concatenated using Sequencher version 4.9 (Gene Codes Corp., Ann Arbor, MI) and aligned manually.

In addition to CR data, we examined nucleotide sequence variation in the mitochondrial cytochrome oxidase I (COI) gene obtained from the Barcode of Life Database (BOLD) public data portal ( http://www.boldsystems.org). COI data for E. radiosum were distributed in four BINs (AAA3138, n = 5; AAA3140, n = 15; ABZ2859, n = 46; ABZ2860, n = 49), and we excluded BIN AAA3139 (n = 1). Sequences for 115 individuals were downloaded into the MEGA version 7.0.26 software environment (Kumar et al., 2016). Sequence alignments were checked and adjusted by eye and translated to corresponding amino acid sequences using the vertebrate mitochondrial DNA code in MEGA.

CR and COI datasets could not be combined into a single phylogenetic analysis because sequences were not obtained from the same individuals and the geographic coverage of the datasets differed (Fig. 1). The CR dataset included Washita River drainage samples from Glasses and Little Glasses creeks that drain into Lake Texoma in the western-most portion of the range of E. radiosum. The COI dataset had denser coverage of eastern localities including Little River tributaries in southeastern Oklahoma and southwestern Arkansas (Fig. 1) but did not include the Washita drainage.

Etheostoma whipplei was selected as outgroup based on evolutionary affinity with E. radiosum (Piller et al., 2001; Near et al., 2011). A single CR sequence (GenBank accession no. U77029) sampled from Lee Creek in western Arkansas was included in alignments of E. radiosum and E. cyanorum. A total of 58 COI sequences from Etheostoma whipplei was downloaded (BIN AAA3316, n = 58) from the BOLD data portal, aligned with sequences of E. radiosum and E. cyanorum, then translated and checked in MEGA as above. In all cases, data matrices included only unique DNA haplotypes.

Phylogenetic analysis.—Initial phylogenetic analysis of nucleotide sequence variation in the CR was conducted using the maximum-likelihood (ML) search algorithm implemented in MEGA. We selected the best-fit model of nucleotide substitution using MODELTEST (Posada and Crandall, 1998) as implemented in MEGA. Once a substitution model was selected, searches for the most likely tree were conducted relative to an initial tree obtained automatically by applying Neighbor-Joining and BioNJ algorithms to a matrix of pairwise distances, and then selecting the topology with highest log-likelihood value. Support for nodes in the CR gene tree was assessed by bootstrap resampling with replacement with 1000 replicates. Nucleotide sequence variation in the control region was visualized in a median-joining haplotype network (Bandelt et al., 1999) rendered in POPART version 1.7 (Leigh and Bryant, 2015). For COI data, we used similar ML methods, model-testing, and search strategies, with node support assessed via bootstrap resampling (1000 replicates).

In addition to analyses in MEGA, we estimated CR and COI gene trees in separate runs using MRBAYES v. 3.2.2 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003) run in the CIPRES Portal v. 3.3 cluster environment at the San Diego Supercomputer Center. Best-fit models of nucleotide substitution (identified from MEGA) were used, and analyses were run for 1x10⁷ generations with sampling the of Markov chain every 1000 generations. This procedure resulted in 100,000 trees for each analysis, of which 10,000 were discarded as burn-in. Support for nodes was determined by posterior probabilities obtained from the majority-rule consensus tree that was visualized using the program FIGTREE version 1.4.2 (Rambaut, 2016). Three replicate runs were conducted with different starting trees to ensure consistency.

RESULTS

Fig. 2

Holotype of Etheostoma cyanorum, as P. radiosus cyanorum (Moore and Rigney, 1952). Photograph courtesy of the Fish Division, University of Michigan Museum of Zoology.

Fig. 3

(A) Male E. cyanorum, 63 mm SL, in nuptial color, from mainstem Blue River, 1.7 km west-northwest of Connerville, OK, 30 March 2018 (WJM 3645; OMNH 86859); (B) female E. cyanorum, 41 mm SL, same collection (OMNH 86860). Location: 34°27.067′ N, 96°39.327′ W. Photographs by N. Lang. (C) Detail of distal color band in nuptial male from same collection, same date (OMNH 86861). Photograph by WJM.

Fig. 4

Angle of the anterior profile for 11 male E. cyanorum, from Blue River in September 1994, with significant trendline, and the holotype (indicated by arrow; not included in regression or in trendline).

Fig. 5

A median-joining haplotype network depicting nucleotide sequence variation in the mtDNA control region (CR). Circles represent unique CR haplotypes and hash marks represent the number of nucleotide differences separating distinct haplotypes. Shaded polygons represent haplotypes grouped by drainage. Haplotypes of E. cyanorum are found only in the Blue River drainage. All haplotypes observed in the Caddo and Little Missouri rivers (not shown in figure) were also observed in the Ouachita River (shown in figure). Etheostoma whipplei is the outgroup taxon (GenBank accession no. U77029).

Fig. 6

A majority-rule consensus gene tree reconstructed from mtDNA cytochrome oxidase 1 (CO1) sequence data obtained from the Barcode of Life Database (BOLD). Maximum-likelihood (ML) and Bayesian trees had identical topologies. Branch lengths are proportional to inferred mutations. Shading of lineages represents samples from E. whipplei (outgroup) in black, E. cyanorum in gray, and E. radiosum in white. ML bootstrap proportions/ Bayesian posterior probabilities are reported. See Data Accessibility for tree file.

Table 1

Meristic characters for Etheostoma cyanorum (Blue drainage), including minimum, maximum, mean, and standard deviation, compared to 13 populations of Etheostoma radiosum.

Table 2

Morphometric characters for Etheostoma cyanorum (Blue drainage), including minimum, maximum, mean, and standard deviation, compared to 13 populations of Etheostoma radiosum.

DISCUSSION

Etheostoma cyanorum and E. radiosum have a long and complicated taxonomic history. The complex has been placed in several genera, confused with similar species (E. artesiae) in collecting accounts, and has had three subspecies described. Now, its allopatry with E. radiosum, and its distinctive morphology and unique genetics, warrant recognition of E. cyanorum as a full species. Evidence has been building for years that E. cyanorum is a species distinct from E. radiosum. Moore and Rigney (1952) recognized that it was morphologically the most divergent of their three putative subspecies, with extreme values for 13 of 15 meristic and morphometric traits. Echelle et al. (1975) showed in analyses of two enzymes that populations in the Blue River drainage differed markedly from those in nearby drainages. Morphological findings of Matthews and Gelwick (1988) could have justified recognition of E. cyanorum, but they chose to withhold recognition pending more evidence. That evidence now exists in the form of increased scrutiny about details of head shape, examination herein of additional morphological or color information since Matthews and Gelwick (1988), and new genetic information. Specifically, the mtDNA sequence assessments described here, including data from April et al. (2011), indicate that E. radiosum (including E. cyanorum) could include at least four genetically distinctive and unique lineages. Similar results were recovered in a mtDNA ND2 gene tree (not shown; N. Lang, pers. comm.).

It is worth noting that mitochondrial genes are linked and effectively represent a single genetic locus. While thorough phylogeographic assessment of nuclear DNA loci would have been ideal, existing data on RAG1 and S7 gene regions showed exceptionally low nucleotide sequence divergence between E. cyanorum and E. radiosum (Near et al., 2011), suggesting that these loci would be largely uninformative at the evolutionary timescale studied here. It is also clear that the mtDNA genome in darters is subject to deep introgression, a term that reflects a mtDNA transfer event prior to diversification of species within lineage Vexillapinna, including the E. radiosum complex (Near et al. 2011). Nonetheless, both mtDNA datasets in the present study showed nearly identical patterns of evolution across the geographic range of E. radiosum and indicate that E. cyanorum is distinct and the most deeply diverged lineage in the species complex. This result is robust to deep introgression and ongoing hybridization (Matthews et al., 2016) with the congener E. pulchellum.

All available evidence supports the recognition of E. cyanorum but suggests that the named subspecies E. r. paludosum or E. r. radiosum require additional studies of geographic variation in molecular and morphological characters if validity of those subspecies is to be determined. The mtDNA gene trees indicated recent and substantive gene flow across drainages previously thought to harbor different subspecies (e.g., Ouachita and Kiamichi basins share identical haplotypes). Closely related haplotypes from the Ouachita, Caddo, and Little Missouri rivers in southeastern Arkansas and the Kiamichi and Clear-Muddy Boggy rivers in southeastern Oklahoma formed a large and geographically widespread lineage with haplotypes shared across drainages, also suggesting relatively recent gene flow. These distinct watersheds span the putative geographic ranges of E. r. radiosum and E. r. paludosum. Moreover, earlier allozyme data indicated that the Kiamichi, Muddy, and Clear Boggy River samples were genetically differentiated (Echelle et al., 1975).

Etheostoma cyanorum is endemic to one drainage, distributed allopatrically from all other populations now considered to be E. radiosum. It is recognized morphologically on the basis of scale counts and a wider interorbital width. Some measurement traits are allometric such that large adults have a shorter snout and deeper body, with a blunter and more downturned snout, as noted by Moore and Rigney (1952) and evaluated in detail by Matthews and Gelwick (1988) and herein. But because these head and body shape traits are allometric, they are not useful for separating smaller individual E. cyanorum from E. radiosum, in spite of being considered diagnostic of the subspecies E. r. cyanorum (Moore and Rigney, 1952). Finally, there is clear evidence from mtDNA analysis that E. cyanorum markedly differs from all other populations considered E. radiosum. We conclude that E. cyanorum should be recognized as a distinct species, with the recommended common name of Blue River Orangebelly Darter.

MATERIAL EXAMINED

DATA ACCESSIBILITY

Supplemental information is available at  https://www.copeiajournal.org/ci-18-054. Mitochondrial control region sequence data are available on GenBank under accession numbers: E. cyanorum, MK490957–MK490959; E. radiosum, MK481992–MK482003.