Analyses of the molecular datasets revealed several distinct clades of Notropis stramineus, detailed in results. Consequently, several terms are here introduced and used from this point onward. These include Notropis stramineus sensu lato, for all individuals currently considered to belong to the species N. stramineus, N. stramineus sensu stricto, for a single clade within N. stramineus sensu lato to which the nominal taxon belongs, the N. stramineus species group, which includes N. stramineus sensu lato, N. topeka, N. procne, and N. chihuahua, and lastly the N. stramineus species complex, which includes four clades of N. stramineus sensu lato

Tissue samples of Notropis stramineus sensu lato (Fig. 2) were obtained from U.S. and Canadian museum collections and through fieldwork conducted between the years 2010 and 2020. Specimens were collected using sampling protocols approved by the Texas A&M University Institutional Animal Care and Use Committee (TAMU IACUC # 2017–0047, 2020–0033). Euthanized individuals were fixed in buffered formalin or preserved in 95% ETOH and deposited at the Biodiversity Research and Teaching Collections (TCWC) at Texas A&M University as voucher specimens. Tissues were obtained either from fin clips or a muscle subsample was taken from the right side of the body. Finally, in order to represent the genus Miniellus as proposed by Stout et al. (2022), sequences from additional species not sampled or borrowed from museum collections were obtained from GenBank.

Ingroup samples for the 3-loci dataset included 160 individuals across the distribution of N. stramineus sensu lato (including the type localities of N. s. stramineus and N. s. missuriensis), three N. topeka, three N. procne, and one N. heterodon (Fig. 2; Table S1). Outgroup samples represent 65 species of Notropis sensu lato and members of other closely related genera (of which there are 1–3 samples per species for most, and 20 for N. chihuahua; Table S1).

UCE data were collected for three to five members of each Notropis stramineus clade (including N. chihuahua) identified through preliminary analyses of Sanger-sequenced loci, and three species purported to make up the genus Miniellus sensu Mayden et al. (2006) including N. procne (type species of the genus), N. topeka, and N. heterodon. Outgroup taxa were selected to be representative of Notropis sensu lato and included N. bifenatus, N. braytoni, N. buchanani, N. chalybaeus, N. longirostris, N. mekistocholas, N. nubilus, N. sabinae, N. boops, and N. volucellus. Agosia chrysogaster, a species identified recently as the possible sister taxon to the Shiner clade (Mayden et al. 2006; Gidmark and Simons 2014; Schönhuth 2018), was also included as the most distantly related outgroup taxon. All sequences generated or obtained from GenBank are listed in Table S1 (Sanger data) and Table S2 (UCE data). Demultiplexed sequence reads have been deposited on NCBI and can be accessed through NCBI BioProject ID PRJNA1363575. Institutional abbreviations of natural history collection names throughout the manuscript, figures and tables follow Sabaj (2020).

Samples for morphological study were obtained either from museum collections or were collected from the wild. Euthanized individuals were fixed in 10% buffered formalin and deposited at the Biodiversity Research and Teaching Collections (TCWC) at Texas A&M University. The distribution of samples of Notropis stramineus sensu lato examined for morphological study by clade is shown in Figure S1. Specimens for which traditional measurements, scale counts, and meristic counts and characters were collected are listed following each species account. Male and female specimens in good condition were prioritized and include those collected near the type locality of previously described species. When possible, specimens from the localities from which genetic samples were available were analyzed for inclusion in morphological datasets. Specimens were examined and imaged using a Zeiss SteREO Discovery V20 stereomicroscope equipped with a Zeiss Axiocam MRc5 digital camera.

Although it is the case that Notropis, as recognized in earlier studies (e.g., Page and Burr 2011; Page et al. 2013), is a polyphyletic assemblage that is in urgent need of systematic study and revision (see Mayden et al. 2006), in this study the recent nomenclatural changes suggested by Stout et al. (2022) are not fully embraced in which 91 of 92 species of North American leucisids previously included in Notropis (e.g., Page and Burr 2011; Page et al. 2013) are divided among multiple genera, including Alburnops, Hydrophlox, Miniellus, Notropis sensu stricto (following Stout et al. 2022), ‘Notropis’, and Paranotropis, based on a phylogenomic study of 87 species and in reference to the results of prior studies. Although these nomenclatural changes have been quickly embraced by others (e.g., Page et al. 2023), it is possible that many of these changes may be premature, especially those that relate to Miniellus in particular, for reasons outlined in detail in this study. In an attempt to facilitate navigation of this study by non-experts (that may adopt the recently published ASIH-AFS official list of common and scientific names of fishes from the United States, Canada, and Mexico; Page et al. 2023), the generic membership scheme of Stout et al. (2022) is adopted only in a subset of the figures (Figs 5, 6, S2) and Table S1, but we continue to use the generic name Notropis as applied in earlier studies (e.g., Page and Burr 2011; Page et al. 2013) in the main body of the text. For example, Notropis (Miniellus) stramineus is used in the aforementioned figures and table (following Page et al. 2023) but we use Notropis stramineus to refer to the same taxon in the text.

DNA extraction. Genomic DNA was extracted from muscle or fin-clips using a DNeasy Tissue Extraction Kit (Qiagen, Inc., CA, USA) following the manufacturer’s protocols. Three loci, one mitochondrial (cyt b) and two nuclear (S7, RAG1) were targeted as these have been demonstrated previously to be informative for inferring relationships among groups of North American leucisids (e.g., see Schönhuth et al. 2011, 2012; Conway and Kim 2016). All polymerase chain reaction (PCR) amplifications were conducted in 25 μl reactions containing 12.5 μl EmeraldAmp GT PCR Master Mix (Takara, Japan), 10.7 μl nuclease-free water, 1 μl (300 ng) of template DNA, and 0.4 μl of each primer (forward and reverse) under the following PCR conditions: cyt b was amplified using the primer pair LA-Danio and HA-Danio (Mayden et al. 2007); cycling included an initial denaturation at 94°C for 4 min, then 35 cycles at 94°C for 30 s each, 50°C for 50 s, and 72°C for 1 min, followed by a final extension at 72°C for 10 min. S7 was amplified using the primer pair S7RPEX1F and S7RPEX2R (Chow and Hazama 1998); cycling included an initial denaturation at 94°C for 4 min, then 35 cycles at 94°C for 30 s each, 61°C for 30 s, and 72°C for 1 min, followed by a final extension at 72°C for 10 min. RAG1 was amplified using the primer pair R12533F and R14078R (Lopez et al. 2004); cycling an included initial denaturation at 94°C for 4 min, then 35 cycles at 94°C for 30 s each, 53°C for 60 s, and 72°C for 1 min, followed by a final extension at 72°C for 10 min. Amplified PCR product was cleaned using ExoSAP-IT and sequenced by either the Genomics Core Lab at Texas A&M University – Corpus Christi or Psomagen (MD, USA). Sequences have been deposited in GenBank under the accession numbers provided in Table S1. Forward and reverse sequences were assembled into contigs either manually using FinchTV (Geospiza Inc.) or in Geneious Prime 11.1.5 (www.geneious.com). Base calls in all sequences were visually checked for accuracy, and uncorrected pairwise distances between select sequences were calculated using MEGA version 11.0.13 (Tamura et al. 2021).

Sequence alignment and phylogenetic and haplotype network analyses. For each method of phylogenetic inference, individual gene datasets as well as a 3-loci concatenated matrix (cyt b, S7, and RAG1) were analyzed. Sequence alignment was performed with MAFFT vs. 7 (Katoh and Standley 2013) either on the webserver (https://mafft.cbrc.jp/alignment/server/) or on the CIPRES Science Gateway portal (Miller et al. 2010). PartitionFinder2 (Lanfear et al. 2017) was used to determine models of nucleotide substitution for each codon position of the two coding genes by selecting the model with the lowest AIC score. Two MCMC runs of 50 million generations were performed in MrBayes 3 (Ronquist et al. 2012) on the CIPRES Science Gateway portal under an uninformative prior with parameters allowed to vary. Tracer 1.7 (Rambaut et al. 2018) was used to visualize the posterior probability distribution and confirm convergence between runs. Maximum likelihood (ML) was implemented using RaXML version 8 (Stamatakis 2016) on the CIPRES portal under a GTR+ GAMMA model with support evaluated using 1000 bootstrap replications. Trees were rooted with Agosia chrysogaster, a species belonging to a monotypic genus previously recognized with morphological methods as a member of the western clade of North American minnows (Coburn and Cavender 1992) or alternatively as a member of the Shiner clade, as one of three possible sister taxa of the remaining members of this clade (Mayden 1992; Gidmark and Simons 2014; Schönhuth 2018). Trees were visualized using FigTree vs. 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree). Median-joining haplotype networks (Bandelt et al. 1999) were constructed using PopART version 1.7 (Leigh and Bryant 2015).

DNA extraction and library preparation. Genomic DNA was extracted from fin clips or muscle tissue using a DNeasy Tissue Extraction Kit following the manufacturer’s protocol, except in the final elution step, in which DNA was eluted in 75µl elution buffer for higher DNA concentration. Each extracted sample was checked for adequate DNA quality and concentration by running 2µl of extracted DNA on an agarose gel with a 100 bp ladder to select samples with high-weight bands (>500 bp).

Library preparation was performed at the Marine Genomics Lab at Texas A&M University – Corpus Christi, Corpus Christi, TX. DNA was standardized to a concentration of 20 ng/µl in 100 µl 1X Tris-EDTA (TE) buffer, then randomly sheared to approximately 800 bp (Sonolab 7.2 program: 800 bp DF 2 time = 80 s) using an M220 Focused-ultrasonicator (Covaris, MA, USA). Because the length of the core regions of UCEs are approximately 150 bp, and fragments lacking the core region would not be captured by the UCE probe set, fragments that are approximately this length (150 bp) would not be useful as they would either not be captured or consist only of an uninformative core region with little informative flanking region on either end. Therefore, to remove short fragments (< 400 bp), sonicated samples were purified with MagBind total bind NGS (Omega BioTek, GA, USA) at 0.6x concentration. DNA concentration was calculated by quantifying samples using Accublue High-Sensitivity (HS) dsDNA quantification kit (Biotium). Subsamples of the sonicated products were then electrophoresed on a 2% agarose gel to visualize band-size distribution. To repair DNA ends from the random shearing, oligonucleotide adapters were blunt-end ligated. End prep included 60 µl reactions consisting of 50 µl sonicated DNA, 3µl end-prep enzyme and 7 µl end-prep buffer. Samples were then ligated to adaptor sequences using the NEB Ultra II DNA library prep Kit for Illumina (New England Biolabs, MA, USA), with the following conditions: 93.5µl reaction included 56.5 µl Blunt End prepped DNA, 30µl Ligation MM, 1µl Ligation enhancer, 6µl Oligo adaptor incubated at room temperature for 30 minutes. Ligated samples were again purified with MagBind total bind NGS, this time at a 1x concentration (= 93.5 µl). Samples were dual indexed with Illumina compatible PCR primers using a bead in protocol with Adapterama iTru5 and iTru7 indexed PCR primers (Glenn et al. 2016); 50 µl reactions comprised 25µl MasterMix, 5 µl iTru5 and iTru7 primer, and 15 µl ligated DNA. PCR conditions included: an initial denaturation step of 98°C for 45 s, followed by 19 cycles of 98°C for 15 s, 60°C for 30 s, 72°C for 1 min 15 s, followed by a final extension of 72°C for 5 min. Following this PCR indexing step, samples were again cleaned, this time using polyethylene glycol (PEG) at 0.9x concentration (50 µl sample * 0.9 = 45 µl PEG/sample) with the following protocol: 45 µl PEG was added to each sample and incubated for five minutes, then placed onto a magnet; supernatant was removed, followed by two washes of EtOH. Samples were eluted in 26 µl water.

Cleaned and indexed samples were then quantified using a Qubit 2.0 fluorometer. Following quantification, samples were enriched for targeted UCE loci following the standard protocol with the Actinopterygian UCE bait-set (myBaits UCE Actinopterygians 0.5Kv1 by Arbor Biosciences, MI, USA; Faircloth et al. 2013) that uses 2001 baits to target 500 UCE loci. Samples were standardized to 250 ng prior to pooling and eight samples were pooled per hybridization reaction. Because each of these samples were different volumes during standardization (some needing as much as 10 µl), samples were dried using a vacufuge and then re-eluted in 7 µl per the manufacturers protocol.

UCE hybridization was performed per the manufacturers protocol with a hybridization temperature (T_H) of 65°C (Arbor Biosciences). This temperature was chosen based on the assumption that approximately 10% of data or less will be missed, and the relatively close relatedness of the study taxa. Hybridized samples were then cleaned using Streptavidin beads (Dynabeads) following standard protocol and eluted in 30 µL of water. Clean hybridized samples were further PCR amplified with 16 cycles; 50 µl reactions each contained 1X Phusion High Fidelity buffer, 1.5 mM MgCl2, 0.25 mM of dNTP, 25 pmol of each P5 and P7 primers, 2.0 U/µl Phusion Taq, and 15 µl DNA. Cycling included: an initial denaturation step at 95°C for 5 min, 98°C for 30 s, and 16 cycles of 98°C for 15 s, 65°C for 30 s, and 72°C for 1 min, 30 s with a final extension step at 72°C for 5 min.

To determine whether UCEs had been successfully enriched (at least a 50× enrichment of product compared to pre-enriched samples; Faircloth et al. 2013), quantitative PCR (qPCR) was performed using GoTaq qPCR master mix (Promega, WI, USA) and a StepOnePlus Real-Time PCR system (Applied Biosystems, inc., MA, USA). Quantitative PCR reactions of 25 µl included: water, GoTaq qPCR Master Mix, 25 pmol qPCR primer, and 2 µl standardized DNA (1 ng/µl). Quantitative PCR conditions included: an initial denaturation step at 95°C for 5 m, followed by 40 cycles of 95°C for 10 s, 60°C for 20 s, and 72°C for 30 s, and a final cooling step at 40°C for 10 s. Forty-eight successfully enriched samples were then quantified, standardized, and pooled for sequencing (paired-end, 2×150 bp) on an Illumina HiSeq 4000 by Azenta, Inc. (MA, USA).

Processing of sequence data. Demultiplexed data were downloaded and adapters and low-quality bases trimmed from sequence data using Trimmomatic (Bolger et al. 2014). Contigs were assembled using Trinity (Grabherr et al. 2011). Next, reads were clustered based on an 80% similarity threshold to create a consensus sequence per contig per individual (CD-hit-est; Li and Godzik 2006). A database of UCE contig matches to UCE probes was created (probe.matches.sqlite) at the beginning of the PHYLUCE pipeline using LASTZ; here contigs were screened for paralogs depending on the number of probe matches/mismatches (Faircloth et al. 2016). Due to the broad taxon sampling of this study, it was possible that loci that could be informative for N. stramineus and closely related species were discarded during initial filtering due to their absence in outgroup taxa. To try to account for this, an additional dataset (referred to as reduced dataset) was created that included only N. stramineus sensu lato, N. topeka, N. procne, and N. chihuahua, with N. mekistocholas, N. nubilus, and N. heterodon to root the trees. In the following outlined methods, each analysis step was performed for both the complete dataset and reduced dataset independently.

Four data matrices, representing different levels of data completeness, were created for each of the two unfiltered datasets (complete and reduced, in terms of taxa included). Each set of four ranged from complete, in which no loci are missing for any taxon; i.e., all taxa share all loci (100%), as well as increasingly inclusive datasets (75%, 50%, and 25%) in which 75%, 50% and 25% of taxa are included in each locus alignment of each dataset, respectively; this is somewhat counterintuitive to these names, as the two 25% datasets will include more data than the 100% datasets (because more loci are included as the requirement for inclusions is only that 25% of taxa retained this locus in the alignment). This resulted in eight unique data matrices (Table 2). Alignment was performed using MAFFT, and taxon names were edited to remove the names of loci for use in downstream analyses in the PHYLUCE pipeline.

Phylogenetic analyses of UCE data. Maximum likelihood analyses were conducted on concatenated matrices resulting from all eight datasets using RAxML vs. 8.2.12 (Stamatakis 2016), applying the GTR+Gamma substitution model. Individual gene trees for each data set were constructed using RAxML under the same nucleotide substitution model. Node support was estimated using bootstopping, an efficient method that determines whether sufficient bootstrap replicates have been completed based on a threshold of variation in branch support for a large percentage (99%) of the bootstrap replicates (a 0.03 or 3% threshold is the empirically recommended setting and was used here; Pattengale et al. 2010; Stamatakis 2016). Species trees were constructed by analyzing the best individual ML gene trees using the summary coalescent method in the program ASTRAL-III (AS; Mirarab and Warnow 2015). Simulation studies have shown that while empirical studies using concatenated datasets often result in well-supported topologies, in taxa with incomplete lineage sorting (ILS) concatenation methods perform relatively poorly and can be inconsistent (Kubatko and Degnan 2007; Roch and Steel 2015) because of gene tree species tree discordance. Therefore, while it is common to conduct an initial concatenated analysis, summary methods such as AS are considered more robust. Support values output by AS are “local posterior probabilities” (LPP) which represent the frequency a given taxon quartet is recovered in the individual UCE gene trees summarized by AS (Sayyari and Mirarab 2016), and the “best” tree maximizes the number of taxon quartets found most often in the gene trees. Internal branch lengths represent coalescent units while terminal branch lengths are not estimated (Mirarab et al. 2014). Discordance among gene trees within a dataset was estimated by calculating the normalized quartet score during species tree construction. This ranges in value from 0 to 1 and represents the proportion of quartets found in the gene trees that are present in the species tree. A normalized quartet score nearer 1 indicates very low gene-tree discordance while a low value, e.g., 0.4, represents very high levels of gene-tree discordance; i.e., in a species tree with a normalized quartet score of 0.4, 60% of quartets found in individual gene trees are not present in the estimated species tree.

To visualize possible discordance in the dataset, an additional analysis was performed with the reduced, 100% dataset using BEAST2 v.2.7.5 (Bouckaert et al. 2019) under uninformative priors with a run of 10 million generations on the CIPRES Science Gateway. Clades that were consistently recovered in the other analyses in this study were constrained to be monophyletic. Convergence of the Markov chain as well as estimated sample sizes were inspected using Tracer v.1.7. DensiTree v.2.2.6 (Bouckaert and Heled 2014) was used to visualize the posterior distribution of trees resulting from this analysis.

Traditional morphometrics and meristic characters. Twenty measurements were collected from 330 specimens representing 50 individuals per clade of Notropis stramineus sensu lato identified in analyses of molecular data as well as ten specimens each of the other members of the N. stramineus species group. Measurements were collected from the left side to the nearest 0.1 mm using digital calipers. Measurements, collected following Hubbs and Lagler (1958), included: (1) standard length (SL); (2) body depth at origin of dorsal fin; (3) predorsal-fin distance; (4) prepelvic-fin distance; (5) pre-anal distance; (6) preanal-fin distance; (7) length of dorsal-fin base, (8) length of anal-fin base; (9) dorsal-caudal distance; (10) caudal-peduncle length; (11) caudal-peduncle depth; (12) head length (HL); (13) head depth through orbit; (14) head depth at occiput; (15) orbit diameter; (16) interorbit distance; (17) snout length; (18) distance from snout to occiput; (19) mouth width; and (20) length of lower jaw. Measurements 2–11 were calculated as a percentage of SL and measurements 13–20 as percentage of HL.

External meristic characters were collected from 50 individuals representing each clade of N. stramineus sensu lato, as well as ten specimens each of the remaining members of the N. stramineus species group. External meristic characters follow Hubbs and Lagler (1958) except: (1) lateral line scale counts do not include those on the base of the caudal fin, which are counted separately; (2) the last dorsal-fin ray is not double-counted while the last two anal-fin rays that share a pterygiophore are both counted, and (3) scale counts (6) and (7) for which scales are counted on one side only, from dorsal midline to ventral midline. Scale counts included: (1) scales in lateral-line scale row; (2) lateral-line scales on base of caudal fin; (3) pre-dorsal scales; (4) scale rows above lateral-line scale row; (5) scale rows below lateral-line scale row (4 and 5 at dorsal-fin origin); (6) circumferential scale rows; and (7) circumpeduncular scale rows. Thirteen internal counts were taken, including: (1) total vertebrae; (2) abdominal vertebrae; (3) caudal vertebrae; (4) position of dorsal fin in relation to neural spines; (5) position of anal fin in relation to hemal spines; (6) number of ribs; (7) dorsal-fin rays; (8) anal-fin rays; (9) principal caudal-fin rays; (10) dorsal procurrent rays in caudal fin; (11) ventral procurrent rays in caudal fin; (12) pelvic-fin rays; and (13) pectoral-fin rays. Vertebral counts include the four Weberian centra and terminal compound centrum (Fink and Fink 1981). Internal meristic counts were obtained from specimens cleared and double-stained following Taylor and Van Dyke (1985).

To visualize variation within the subset of measurements identified as potentially important during geometric morphometric analysis (see below) among the five clades of Notropis stramineus s. l., violin plots combined with boxplots were constructed using a combination of the R packages ggplot2 (Wickham 2011) and ggstatsplot (Patil 2021) performed in R (R Core Team 2023). These included orbit diameter, snout length, body depth, and snout to occiput.

Multivariate statistics. Principal component analysis (PCA) of traditional measurements obtained from the measurement data collected (described above) was performed in base R using the function “prcomp()”. To account for the effect of size on measurements, each measurement was regressed against standard length, and the residuals were appended to the original data matrix and used in further principal component analysis. PCA plots were visualized using the package “ggbiplot”, a ggplot2 based biplot, in R. The broken-stick model was used to determine how many PCs were important to maintain for analysis. To determine whether a significant difference was present between species in the returned PCA data, a multivariate analysis of variance (MANOVA) was performed on the PC scores returned, grouped by clade. To further determine which clades differed significantly from one another, pairwise t-tests were performed for the scores obtained from principal components one through three.

Geometric morphometrics. Two-dimensional geometric morphometric data were collected for use in multivariate analyses of shape data. Images were taken of the left side of 20 specimens of each member of the Notropis stramineus species group (except for N. chihuahua, for which 10 specimens were photographed) with a Canon 60D DSLR camera. Only specimens without obvious preservation artifacts (e.g., bending, desiccation, damage) were photographed and used in analyses. For photographs, specimens were placed onto a small amount of putty resting within a plastic box containing Sylguard and positioned to be parallel with the surface. A pin was placed into the underlying Sylguard and used to separate the posteriormost fin rays of the dorsal and anal fin from the body. Landmarks for the head and body include: (1) anteriormost point of rostrum; (2) posteriormost point of nare; (3) anteriormost, (4) posteriormost; (5) mid-dorsal and (6) ventral points of orbit; (7) and occiput; (8) anterior and (9) posterior insertion of dorsal fin; (10) insertion of dorsal and (11) ventral procurrent rays of the caudal fin; (12) anterior and (13) posterior insertion of anal fin; (14) anterior insertion of pelvic and (15) pectoral fins; (16) posteriormost point of opercle; (17) point of greatest curvature of preopercle; and (18) posteriormost point of the jaw (Fig. 3).

Landmarks were placed using the digitizing tool available in the R package StereoMorph (Olsen and Westneat 2015). After placement, to correct for orientation, size and location, centroids were calculated and shape data for each specimen were transformed via least-squares generalized Procrustes analysis (GPA) using the function “gpagen” in the R package GeoMorph (Adams et al. 2020) to create Procrustes shape coordinates. Each specimen was assigned to its respective clade, with clade membership acting as the predictor variable in subsequent analyses. All analyses were performed for all landmarks (i.e., head and body), while a subset of analyses was additionally carried out for head landmarks only. An initial dataset included the N. stramineus species group and included all landmarks (body and head). For all analyses, an alpha value or significance threshold was set at α = 0.05. To determine whether specimen size had a significant effect on shape (allometry), an analysis of variance (ANOVA) test was performed on a complex model including size as a predictor variable for this complete dataset. Size was found to have had a significant effect (P = 0.018), suggesting the size of a specimen has some effect on shape, however, a subsequent ANOVA test (using the RRPP package, “Residual Randomization Permutation Procedure”, as implemented in Geomorph; Collyer and Adams 2018) of the suitability of a simple model (only effect of clade membership on shape) versus the complex model (effect of clade membership + size) determined that the more complex model did not significantly improve analyses (P = 0.802; fail to reject null [simple] model). Additionally, explaining the effect of size in this more complex model had no impact on the significance of shape differences explained by clade membership. Therefore, remaining analyses were performed without including size as an explanatory variable in ANOVA tests.

Pairwise comparisons of clades were analyzed for each dataset using the function “pairwise” in the package RRPP as implemented in GeoMorph. Two methods to compare clades of the Notropis stramineus species group were implemented. The first tests the statistical significance (α = 0.05) of the Euclidean distance between the least-squares (LS) mean of the Procrustes shape coordinates for each clade. The second tests the statistical significance (α = 0.05) of the difference in the morphological disparity within a clade (variance) between clades. Least squares means were considered significantly different if the Euclidean distance value fell outside of the one-tailed upper 95% confidence limit. In this case, the confidence limit is one-sided or one-tailed because the LS means are treated as an absolute value. To tease apart shape differences among clades, additional subsets of the initial all-landmark and head-only datasets were constructed to more readily visualize shape differences among species and clades of interest using point and vector diagrams, also analyzed as described above. PCA graphs were constructed using the package ggplot2. Thin-plate spline deformation grids were constructed along the principal component axes to aid in visualizing trends in shape difference.

Osteology. Select specimens were cleared and double-stained following the protocol of Taylor and Van Dyke (1985) for the examination of bone and cartilage and dissected followed Weitzman (1962). Cleared and stained specimens were examined and photographed using a Zeiss SteREO Discovery V20 stereomicroscope equipped with a Zeiss Axiocam MRc5 digital camera.

Computed tomography (CT) scans of a representative of members of the Notropis stramineus species group were obtained at the Karel F. Liem BioImaging Center (Friday Harbor Laboratories, University of Washington, WA) using a Bruker SkyScan 1173 scanner with a 1 mm aluminum filter. Scans were run at 65 kV and 123 μA on a 2048×2048 pixel CCD at a resolution of 9.9–12.4 μm. Specimens were heat-sealed into a small plastic bag and scanned while inside a 50 ml or 15 ml plastic Falcon tube stuffed with packing peanuts to prevent any movement while scanning. The resulting CT data were reconstructed using NRecon and visualized using 3D Slicer v.4.10.2 (https://www.slicer.org). Raw and reconstructed datasets have been deposited on MorphoSource (https://www.morphosource.org; Blackburn et al. 2024). Cranial osteological terminology follows Harrington (1955).

Terminology of cranial and body regions. Members of the Leuciscidae often display sexual dimorphism in the size and number of tubercles present on the surface of the head, body, and fins, and characters of tuberculation often are useful for distinguishing between closely related taxa (e.g., Cross 1953; Mayden 1989; Chen and Arratia 1996; Cashner et al. 2010). Despite this usefulness, previous studies primarily concerning or simply remarking on the location, morphology and arrangement of tubercles (and other keratinous structures) have used an inconsistent variety of terms and methods of description (e.g., “breeding stars”, “contact organs”, and other names; Roberts 1982; Wiley and Collette 1970; Pinion and Conway 2018). This makes it difficult to derive general comparisons among studies. Similarly, characters of pigmentation of the head, while often useful in species identification and discrimination, could more readily be compared if partitioned into comparable regions. In an effort to standardize descriptions of cranial (and nearby) structures for tubercles and pigmentation, we introduce a new terminology that utilizes well-known and easily observable morphological landmarks of the cranium and surrounding area. This allows distinct tubercle and pigmentation distributions to be described within the boundaries of designated regions (Fig. 4). These regions include: (1) rostral; (2) nasal; (3) lacrimal; (4) supraorbital; (5) interorbital; (6) infraorbital; (7) preopercular; (8) fronto-occipital; (9) opercular; (10) subopercular; (11) internarial; (12) gular; (13) interopercular; (14) mandibular; (15) branchiostegal; and (16) chest.

In addition to these terms relating to the cranium, several terms are used relating to body pigmentation throughout species accounts. These include: (1) cross-hatching pattern, in which the posterior margin of scales are outlined in melanophores, creating a hatching pattern, particularly when the melanophores do not exactly follow the (rounded) scale itself and instead form a diamond pattern; (2) cleithral streak, in which a diffuse line of melanophores, sometimes tightly organized, sometimes loosely, extends from the first canal ossification of the body lateral-line canal ventrally bordering the posterior part of the cleithrum to a point along the imaginary horizontal line through the center of the opercle; (3) pre-dorsal wedge, in which melanophores along the dorsal midline are concentrated heavily just anterior to the insertion of the dorsal fin, typically in a roughly triangular pattern; and (4) dorsal pre-caudal wedge, in which melanophores along the dorsal midline are concentrated just anterior to the insertion of the dorsal procurrent rays of the caudal fin.

Scanning electron microscopy. Tuberculate males (presumably collected during the peak of spawning activity) of the Notropis stramineus species group were prepared for scanning electron microscopy (SEM) following the methods of Pinion et al. (2021). Specimens were examined using a TESCAN Vega 3 environmental scanning electron microscope at the Microscopy and Imaging Center at Texas A&M University (RRID: SCR_022128). Aspects of tuberculation (tubercle distribution, size, abundance, and individual tubercle morphology) were assessed qualitatively and quantitatively from scanning electron micrographs. To compare patterns of tuberculation, the cranial regions outlined above were superimposed onto scanning electron micrographs and qualitatively assessed for differences in pattern.

A total of 262, 143, and 190 sequences of cyt b, RAG1, and the first intron of S7, respectively, was generated. Final datasets comprised 263 samples for the cyt b locus including 81 outgroup taxa, 140 samples for the RAG1 locus including 71 outgroup taxa, and 169 samples for the S7 locus including 50 outgroup taxa. The cyt b alignment was 1133 bp in length and was partitioned by codon; the RAG1 alignment was 1504 bp in length and partitioned by codon; the S7 alignment was 977 bp in length (including gaps) and was treated as a single partition. Mean, uncorrected genetic distances among and within recovered clades and related species using the fragment of the cyt b locus ranged from 16% to 2% between clades and ~0.1% to ~0.3% within clades (Table 3). Nucleotide substitution models by partition are detailed in figure insets (Figs 5, S2). Results of Bayesian and ML analyses of datasets were congruent, and therefore only trees resulting from Bayesian analyses of Sanger-sequenced loci are presented.

Summary statistics for the eight UCE datasets generated are provided in Table 2. The average locus length was ~1140 bps, with an average of ~70 informative sites per locus. Normalized quartet scores for all estimated species trees ranged from 0.68–0.71, meaning discordant gene trees are present in all datasets, and that ~29–32% of quartets found in ML analyses of individual loci (gene trees) are not present in the estimated species trees. Minor differences in interrelationships of outgroup taxa were present in most datasets but are not discussed.

Notropis stramineus sensu lato. Analyses of the concatenated Sanger-loci and UCE datasets (Figs 5, 6) as well as the cyt b (Fig. S2A) and S7 (Fig. S2B) loci provided resolution of six distinct clades of Notropis stramineus sensu lato. The topologies derived from the concatenated Sanger-loci datasets and UCE datasets were congruent with the individual Sanger-sequenced loci in clade membership of the N. stramineus species group but differed variously in the interrelationships of these clades. However, one consistency across analyses was that individuals of Notropis topeka, N. procne, and N. chihuahua (and several other members of Notropis in the cyt b gene tree and Sanger-loci concatenated analysis) were recovered inside of clades of N. stramineus sensu lato. Therefore, the taxon N. stramineus sensu lato likely comprises a paraphyletic assemblage of six clades.

The recovered six clades that make up Notropis stramineus sensu lato include the following. The first, referred to as Clade 1, contains individuals collected in the Great Lakes system, the Ohio, upper Illinois, and Tennessee River drainages, including those collected from the Detroit River in Michigan, the type locality for N. stramineus; this clade is monophyletic in all phylogenetic trees but one based on the RAG1 locus. Clade 2 includes individuals collected from the Canadian, Red, Arkansas, and Pecos Rivers; this group was not monophyletic in the individual cyt b and nuclear gene trees (Fig. S2) but was monophyletic in the resulting trees of the concatenated Sanger-loci and UCE loci analyses (Figs 5, 6). Clade 3 containing individuals distributed in some independent Gulf Slope drainages of Texas, e.g., the Nueces, San Antonio, and Guadalupe Rivers; monophyletic in the cyt b and S7 gene trees (Fig. S2A,B), as well as in the resulting trees of the concatenated Sanger-loci and UCE loci analyses (Figs 5, 6). Clade 4 contains specimens collected in the Red River of the North, the Missouri, and Mississippi Rivers, including specimens collected from tributaries of the Missouri River near St. Joseph, Missouri, the type locality for the subspecies N. s. missuriensis; also monophyletic in all phylogenetic trees but one based on the RAG1 locus. The fifth clade recovered is Clade 5, with individuals collected from the Devils River, a tributary of the Rio Grande in Texas. And finally, Clade 6 comprises individuals collected in the Colorado River basin of Texas. Clades 5 and 6 comprise a monophyletic group in the S7 gene tree and UCE phylogeny (LPP = 1; BS = 100).

These molecular results in combination with the morphological results have resulted in the recognition of five distinct lineages (Clades 1, 2, 3, 4, and 5+6) as distinct species herein; a summary of monophyly per clade and the corresponding species, either described as new or redescribed, is provided in Table 4. Putative clade distributions are illustrated in Figure 7.

The six clades above, in addition to N. procne, N. topeka, and N. chihuahua, form the herein proposed N. stramineus species group. The monophyly of this group is supported by all analyses of UCE datasets (LPP = 1, BS = 100). There were, however, several differences in the cyt b, S7 and RAG1 gene trees. In the cyt b gene tree, the N. stramineus species group is not monophyletic due to the distant placement of N. stramineus Clade 5 and N. chihuahua among outgroup taxa, and the inclusion of N. alborus as the sister taxon of N. procne. In all but the concatenated-loci analysis and cyt b, Clade 6 joins Clade 5 and forms a monophyletic clade that is the sister group to N. chihuahua.

Four clades comprise the proposed N. stramineus species complex, which was monophyletic in all UCE analyses (LPP = 1, BS = 100). The species complex was however paraphyletic with respect to either N. topeka in cyt b and S7 gene trees, N. topeka and Clade 6 in the resulting trees from the concatenated Sanger-loci and cyt b analyses, or N. procne and N. topeka in the RAG1 gene tree. Relationships among these three gene trees were highly discordant, something that was also seen in the UCE phylogenetic hypothesis, obvious at a glance by the various topologies simultaneously visualized with DensiTree (Fig. 8).

Despite high levels of gene-tree discordance in the UCE datasets, topologies were entirely congruent in all AS analyses. The topologies resulting from the ML analyses of concatenated-loci datasets (not shown) were congruent with those of the AS analyses with the following differences observed in four of the eight ML analyses of concatenated-loci datasets: first, although Notropis procne was recovered as the sister taxon to the clade containing the N. stramineus species complex + N. topeka in the majority (13/16) of analyses, in three analyses (concatenated-loci ML analysis of the 25, 50 and 75% missing data datasets using the complete dataset), N. topeka was instead placed as the sister taxon to the N. stramineus species complex + N. procne (not shown). Second, within the N. stramineus species complex, Clade 3 was recovered as the sister group to Clade 1 in 15 out of 16 analyses. However, in one of the concatenated-loci ML analyses (complete dataset, 100%), Clade 3 was recovered as the sister taxon to the remaining clades of the N. stramineus species complex (Fig. S3). Aside from these two differences, only support values differed among analyses and tended generally to be higher for recovered N. stramineus species complex interrelationships when more loci were included (25%, 50% and 75% complete datasets vs. 100% dataset), but lower for the inclusion of N. topeka within a clade including the N. stramineus species complex exclusive of N. procne with increasing numbers of loci.

Miniellus. Miniellus sensu Mayden et al. (2006), which contains Notropis procne, N. heterodon, N. stramineus and N. topeka was not monophyletic in any analysis. In UCE analyses, this was due to the exclusion of N. heterodon and inclusion of N. chihuahua. Results from analyses of the concatenated Sanger-loci dataset, cyt b, S7 and RAG1 loci were more complex; however, consistent reasons included the recovery of members of this putative genus in two clades, these not always as sister groups, and also typically due to the exclusion of N. heterodon and recovery of this species as more closely related to other members of Notropis, as well as the close relationship of N. chihuahua with N. lucifer sp. nov.

Miniellus sensu Stout et al. (2022), which is proposed to contain 21 species (see Table 1 for membership), was not monophyletic in the S7 gene tree due to the exclusion of N. sabinae, N. longirostris and N. ortenburgeri. Monophyly of this putative genus could not be evaluated using UCEs as all species included in Miniellus sensu Stout et al. (2022) were not included in the UCE dataset. Monophyly could not be rejected in the concatenated Sanger-loci, cyt b or RAG1 analyses as all members were included within an unresolved polytomy. The only species in Miniellus sensu Stout et al. (2022) without a sequence for RAG1 (or S7) available (but with a cyt b sequence available) and thus not evaluated in the two nuclear gene trees is N. melanostomus.

Haplotype network. Using a 1125 bp fragment of the cyt b gene, a total of 89 haplotypes were recovered among 164 individuals of Notropis stramineus sensu lato. Nine haplogroups were found corresponding to the six major clades recovered in the phylogenetic analyses (groups with >10 mutational steps between them were considered to be distinct haplogroups) (Fig. S4). Clades 1 and 4 were characterized by a relatively high number of haplotypes but lower number of mutations between haplotypes, while Clades 2 and 3 have both a higher number of haplotypes and a higher number of mutational steps between haplotypes. Relative to Clade 6, Clade 5 had relatively few haplotypes represented. A second analysis was performed in which the remaining members of the N. stramineus species group were included (N. procne, N. topeka, and N. chihuahua; Fig. S5). This included 189 sequences, from which 107 haplotypes were recovered.

Meristic characters. Scale, fin-ray and vertebral counts largely overlapped in all clades, with some differences discussed further under species accounts (scale counts, Table 5; fin ray and vertebrae counts, Table 6).

Traditional morphometrics. Proportional measurements for each clade can be found in Table 7. For all measurements, ranges of values were largely overlapping for all clades of Notropis stramineus (visualized in PCA plots; Figs S6, S7), though the mean values of many of the measurements differed significantly among various clades (not shown). Loadings for PCs 1–3 are presented in Table S3. Results of MANOVA tests are presented in Table S4, along with pairwise t-tests of PC scores among clades. Hybrid violin/boxplots of select measurements are shown for visualization purposes for some of the more highly significant differences between clades, which are referenced as appropriate throughout species accounts. Differences in head measurements (% HL) were typically more drastic between clades than body measurements (% SL). For example, the orbit diameter of Clade 3 was significantly larger than the remaining clades, followed by Clade 1 and Clades 5 and 6, snout length in Clade 4 was shorter than in the other clades (Fig. S8), and distance from the snout to the occiput occupied a greater percentage of head length in Clade 3, differing significantly from all other clades except for Clade 1 (Fig. S9).

Geometric morphometrics. Despite a large amount of overlap in the first few principal components, pairwise comparisons of LS means revealed statistically significant differences among many of the clades of Notropis stramineus sensu lato identified by molecular analyses, as well as between members of the N. stramineus species group (Table S5). While most pairs of clades and species were found to significantly differ in shape, several clade pairs were consistently found to not possess body or head shapes that differed significantly. For overall body shape, this includes Clade 1 and 2, Clade 1 and 4, Clade 1 and 5, Clade 2 and 4, and Clade 2 and 5. For head shape this includes only Clade 1 and 2, and Clade 1 and 4. Results of the PCA that included all members of the N. stramineus species group with vector diagrams representing extremes of principal components (minimum and maximum) are shown in Figure S10. Similar summaries for all-landmark and head-only landmark analyses including only N. stramineus sensu lato are shown in Figures 9 and S11, respectively. Pairwise analyses of those clades for which significant differences were found are shown in Figures S12 and S13.

In the all-landmark dataset with all members of the N. stramineus species group (Fig. S10), principal components one to four accounted for 73% of the cumulative proportion of variance. Principal component one (33.3%) typically separated specimens along an axis that contained deeper-bodied individuals (dorsally and ventrally) with a more anteriorly-placed anal fin and a longer caudal peduncle at the positive end of the axis, and individuals that were more terete, thinner, with a more posteriorly-placed anal fin and shorter caudal peduncle at the negative end of the axis. Principal component two (20.5%) separated individuals with a steeper dorsal profile and a longer snout (positive end of axis) from those with a shorter snout and thinner body dorsally (negative end of the axis).

In the all-landmark dataset with only Notropis stramineus sensu lato represented (Fig. 9), results closely resembled that of the previous analysis that included all members of the N. stramineus species group; however, the Euclidean distance between the LS means of Clade 4 and 5 was no longer significant (P = 0.131). Principal components one through four contained 67.5% of the cumulative proportion of variance, with principal component one and two explaining 34.2% and 12.6% of the variation, respectively. Principal component one separated individuals on the positive end of the axis from the negative end of the axis based on narrower, less steeply sloped, and shorter heads, and a dorsal profile that also slopes less steeply. Principal component two placed individuals on the positive end of the axis that have a shorter caudal peduncle, and those with the longer caudal peduncle on the negative end.

Osteology. No major osteological differences were identified among clades of Notropis stramineus sensu lato. However, differences in the degree of ossification (i.e., presence of ossified bone vs. cartilage) of the ethmoid region, and the shape of the vomer, were apparent between individuals of Clade 5 and the remaining members of N. stramineus sensu lato (Fig. 10).

The name Miniellus was first made available by Jordan (1882: 842), the type species Hybognathus procne Cope, 1865 designated later by Jordan and Evermann (1896: 254). Much like Notropis, Miniellus has had a tumultuous taxonomic history, and early on was considered by Jordan to be a subgenus of varying genera, e.g., Hudsonius (Jordan 1882: 842), Cliola (Jordan and Gilbert 1883) or as a synonym of other genera, e.g., Alburnops (Jordan 1885) and ultimately Hybopsis (Jordan et al. 1930). Later, in his type catalogue for the genus Notropis, Gilbert (1978) treated Miniellus as a synonym of Notropis, a decision that remained until two recent attempts to characterize the species comprising Miniellus and elevate it to the rank of genus. The first was Mayden et al. (2006), in which the relationships of notropin shiners and minnows (Notropis sensu lato and relatives) were inferred from analysis of the cyt b gene. Mayden et al. (2006) proposed that the species N. procne (type species), N. topeka, N. stramineus, and N. heterodon should comprise Miniellus, and recovered the topology seen in Figure 37A. However, Mayden et al. (2006) suggested that several additional species of Notropis not included in their study (viz. N. chihuahua, N. uranoscopus, N. alborus and N. albizonatus) also may need to be included in the genus based on a separate study (Warren et al. 1994). This same four-species composition of Miniellus was supported by Schönhuth et al. (2018) in which two mitochondrial (cyt b and COI) and two nuclear (RAG1 and IRBP) genes were used to construct phylogenetic hypotheses for the Leuciscidae. Schönhuth et al. (2018) included broad as well as dense taxon sampling, and despite assessing much of the family Leuciscidae, also managed to include all four original members of the genus Miniellus sensu Mayden et al. (2006). With their combined (concatenated) dataset, Schönhuth et al. (2018) recovered Miniellus as monophyletic, with the relationships of the four included species as seen in Figure 37B, a topology that differed slightly from Mayden et al. (2006). A strict consensus between the two studies would include N. stramineus, N. topeka, and N. heterodon in a trichotomy with N. procne as the sister taxon. Notably, none of the analyses in the present study recovered N. heterodon within a clade that contains N. stramineus, N. topeka, and N. procne (Figs 5, 6, S2, S3). Stout et al. (2022) also did not recover N. heterodon as closely related to N. stramineus (Fig. 37C). A possible explanation for this discrepancy is discussed in the following discussion section, “Notropis heterodon”.

Recently, Stout et al. (2022) used exon capture to study the relationships of notropins, and proposed that Miniellus be expanded to comprise the aforementioned species as well as 17 more species currently placed in Notropis. This would bring the total number of species included in Miniellus to 21 (only 10 of which are represented by sequence data in the analyses of Stout et al. 2022; see Table 1). These putative members of Miniellus include (asterisks denote those species not included in the dataset of Stout et al. 2022, as in Table 1): Notropis albizonatus Warren & Burr, 1994 (palezone shiner), N. alborus Hubbs & Raney, 1947 (whitemouth shiner)*, N. ammophilus Suttkus & Boschung, 1990 (orangefin shiner), N. anogenus Forbes, 1885 (pugnose shiner)*, N. boops Gilbert, 1884 (bigeye shiner)*, N. chihuahua Woolman, 1892 (Chihuahua shiner), N. greenei Hubbs & Ortenburger, 1929 (wedgespot shiner)*, N. heterodon (blackchin shiner), N. longirostris (Hay, 1881) (longnose shiner), N. mekistocholas Snelson, 1971 (Cape Fear shiner)*, N. melanostomus Bortone, 1989 (blackmouth shiner), N. nubilus (Forbes, 1878) (Ozark minnow), N. ortenburgeri Hubbs, 1927 (Kiamichi shiner)*, N. perpallidus Hubbs & Black, 1940 (peppered shiner)*, N. procne (Cope, 1865) (swallowtail shiner)*, N. rafinesquei Suttkus, 1991 (Yazoo shiner)*, N. sabinae Jordan & Gilbert, 1886 (Sabine shiner), N. scabriceps (Cope, 1868) (New River shiner), N. stramineus (Cope, 1865) (sand shiner), N. topeka (Gilbert, 1884) (Topeka shiner)* and N. uranoscopus Suttkus, 1959 (skygazer shiner)*.

Despite lacking over half of these species in their dataset, Stout et al. (2022) recommended their inclusion in Miniellus for a variety of reasons. These reasons vary by species, but, in general, amount to two major factors: these species have either been previously hypothesized as closely related to either one of the four originally proposed members of Miniellus (N. procne, N. topeka, N. heterodon or N. stramineus), or to species recovered as closely related to any one of these original four species in Stout et al. (2022). These previous hypotheses of relationships were often made by authors that were experts in the group (e.g., Hubbs and Raney 1947; Snelson 1971), though these were often based on overall morphological similarity (i.e., did not include formal phylogenetic methods). Stout et al. (2022) also do not test these relationships because, as mentioned above, most of the species were missing from their dataset. Notably, two of the taxa proposed to belong in Miniellus by Mayden et al. (2006) are missing from Stout et al.’s (2022) dataset, including the type species of the genus Miniellus, N. procne.

Interestingly, Stout et al. (2022: 13) state that rather than finding the expected sister group relationship between N. stramineus and N. heterodon, they recover “several” Notropis species as more closely related to either N. stramineus or N. heterodon. Stout et al. (2022) then use the fact that N. heterodon was originally included in Miniellus by Mayden et al. (2006) and Schönhuth et al. (2018) as partial justification to greatly expand Miniellus, and thusly accommodate N. heterodon within the proposed genus. In the present study, analyses including all but one of the species proposed by Stout et al. (2022) to belong to Miniellus (Notropis albizonatus, a rare and federally protected species for which sequences were unavailable) reveal a more complex scenario. The genus Miniellus sensu Stout et al. (2022) was not recovered as monophyletic in the S7 gene tree due to the placement of N. sabinae, N. longirostris and N. ortenburgeri, and in the other two gene trees, members of this putative genus were recovered within a polytomy. Additionally, one species used to justify expansion of Miniellus by Stout et al (2022), N. heterodon, was most likely included in early compositions of the genus in error (see next section titled ‘Notropis heterodon’ below).

Based on this information, the expansion of Miniellus as proposed by Stout et al. (2022), and quickly adopted by Page et al. (2023), is considered by the present authors to be premature. Further study that combines broad taxon sampling, morphological analysis (to aid generic diagnosis), and powerful genomic tools such as exon capture or UCE analysis will be required to clarify the limits and membership of Notropis and related genera, including Miniellus, which somewhat ironically has replaced Notropis sensu lato as the catch-all or waste-bin group of North American leuciscids.

Notropis heterodon, the blackchin shiner, was proposed by Mayden et al. (2006) as one of the four species to belong to Miniellus based on their analysis of the cyt b locus in which N. heterodon was placed as the sister taxon to N. stramineus (Fig. 37A). In their combined dataset, Schönhuth et al. (2018) report a similar but slightly different relationship (Fig. 37B), still with N. heterodon closely allied with the other three members of Miniellus sensu Mayden et al. (2006), but in this case as the sister taxon to N. stramineus + N. topeka. Stout et al. (2022) did not include N. procne or N. topeka in their analyses, as discussed above, and therefore one would expect to see in their results a sister-group relationship between N. stramineus and N. heterodon. However, this is not the case and Stout et al. (2022) instead recovered N. heterodon as the sister taxon to N. nubilus (Fig. 37C). In the present study, N. heterodon is placed well outside of the clade containing N. stramineus, N. topeka, and N. procne in all analyses and is most commonly placed as the sister taxon to N. boops (a species not included by Stout et al. 2022).

Because of the sharp contrast between Mayden et al.’s (2006) and Schönhuth et al.’s (2018) relationship of N. heterodon with the results of Stout et al. (2022) and the present study, we took a more detailed look at the evidence in prior studies supporting the phylogenetic placement of this species. The sequence for Notropis heterodon analyzed by Schönhuth et al. (2018) is available from GenBank (MG806610.1), and thus this sample was compared with one from the present study and others available from GenBank, which includes a partial sequence from an unpublished study (KM523298.1) as well as a sample generated by Schönhuth et al. (2003) (AY140697.1).

The Notropis heterodon sample from Schönhuth et al. (2018) was collected in the Wabash River system, and has a cyt b sequence (GenBank MG806610) nearly identical to some haplotypes belonging to N. stramineus Clade 1 recovered herein (with only a single base pair difference from the most similar haplotype, an individual of N. stramineus collected in the Great Lakes drainage). Therefore, this N. heterodon sequence likely results from a misidentification, most likely of N. stramineus. By distribution, this specimen would belong to N. stramineus Clade 1 of the present study (N. stramineus sensu stricto). The N. stramineus sample from which Schönhuth et al. (2018) obtained their IRBP sequence was collected in the Missouri River in Missouri, and so by distribution should belong to N. missuriensis. If one treats the “N. heterodon” of Schönhuth et al. (2018) as N. stramineus sensu stricto, and their N. stramineus as N. missuriensis, their tree topology (resulting from analysis of a combined dataset of both mitochondrial and nuclear loci) matches the topology resulting from the S7 gene tree in the present study (see Fig. 37B vs. Fig. S2B).

In Mayden et al. (2006), Notropis heterodon is found to be the sister taxon to N. stramineus, as mentioned above (Fig. 37A). Although their N. heterodon sample was not available to evaluate further (the sequence is not publicly available), their published relationship differs from that recovered in the present study, and potentially represents another misidentification when compared with the tree topology derived from the dataset herein and that of Stout et al. (2022). A search of the FishNet2 database (fishnet2.net) reveals the other species collected during the same collection event that yielded their tissue of N. heterodon (UAIC 7881.01). According to this database, two N. ludibundus (= stramineus of present authors) were collected with 18 specimens of N. heterodon. It seems possible (though difficult to confirm) that these authors either misidentified N. stramineus as N. heterodon or simply made a mistake when taking tissue samples. The locality of this collection is in Illinois, and given as “Cedar Lake, on the East side at island (Fox-Des Plains River).” By location, if this were a misidentification, their sample would most likely be one of N. stramineus sensu stricto. The sequences identified as belonging to N. stramineus in Mayden et al. (2006) (UAIC 8420.07) was collected in the Wabash River system in Ohio and is found to be the sister taxon of their “N. heterodon”. Because both of these samples are from localities in which only N. stramineus sensu stricto has been sampled in the present study, their N. heterodon, if indeed misidentified as N. stramineus, would be expected to belong to N. stramineus sensu stricto, and therefore a “sister-group” relationship between two individuals of N. stramineus collected from these localities would not be surprising.

It is therefore likely that the reason Notropis heterodon does not group with the Miniellus sensu Mayden et al. (2006) assemblage (N. stramineus, N. topeka, and N. procne) in the present analysis, which is based on the same gene (cyt b), is because the original placement of the species in Miniellus by Mayden et al. (2006) was based on a misidentification, and then supported later by a separate misidentification by Schönhuth et al. (2018). In support of this conclusion, the recent paper by Stout et al. (2022) using exon capture did not find a close relationship between N. stramineus and N. heterodon. As a consequence, Stout et al. (2022) reasoned that because Miniellus inclusive of N. heterodon was well supported in the previous two papers, all of the taxa that fall out between the other members of Miniellus sensu Mayden et al. (2006) and N. heterodon should be included in the genus, thus greatly expanding Miniellus (see section above: ‘A history of the genus Miniellus Jordan, 1882’).

Notropis stramineus sensu stricto. The clade comprising Notropis stramineus sensu stricto, represented by the lectotype ANSP 4131 designated by Fowler (1910), is consistently recovered and well-supported in all analyses other than that of RAG1. A member of the N. stramineus species complex, the sister taxon to this species is unclear, as the recovered sister-group relationships were poorly supported across the analyses.

This species occupies the Great Lakes drainage, the northwestern slopes of the Appalachian Mountains, and the adjacent northeastern tributaries to the Mississippi River. This distribution includes the Canadian provinces Ontario and Quebec, as well as the U.S. states New York, Pennsylvania, West Virginia, Ohio, Kentucky, Tennessee, Michigan, Indiana, Illinois, and Wisconsin. Michels (2000) did not include samples from this clade, though Pittman (2011) included four samples from a tributary of the Ohio River in West Virginia that she allocated to her “Group O” that would represent individuals from the distribution of N. stramineus sensu stricto.

Anthropogenic modifications to the Illinois River could have resulted in transfer of the sand shiner from the Great Lakes system, and therefore the presence of this species there could potentially be unnatural. In recent times, and in its natural state, nearly all of the streams and rivers in Illinois belonged to the Mississippi drainage, except for a small area near the shoreline of Lake Michigan. However, since the late 19^th and early 20^th century, modifications to the Chicago River, including the addition of a canal (the “Chicago Area Waterway System”), have redirected the waterway and connected the Great Lakes drainage to the Mississippi River through the Illinois River. This connection has been shown to be responsible for recent introductions of species not native to the Illinois and Mississippi River systems from the Great Lakes, for example, a subspecies of the banded killifish Fundulus diaphanus (Hartman et al. 2023) and the round goby Neogobius melanostomus (Irons et al. 2006). More finely grained sampling work at the population level is needed to understand the native status and range limitations of N. stramineus in the Great Lakes and Illinois River systems.

The distribution of Notropis stramineus sensu stricto in the eastern side of Tennessee in the Cumberland and Tennessee Rivers creates a disjunct part of the distribution. This is perhaps a result of the capture of headwater streams during the Pliocene and Pleistocene from the Ohio River system, as many of the fishes inhabiting the upper Cumberland River can also be found in the Kentucky River of the Ohio drainage (Starnes and Etnier 1986). For example, this distribution pattern can also be seen in the tongue-tied minnow, Exoglossum laurae (Oswald et al. 2020). An additional possible scenario is one in which a previously more widespread population existed that was distributed across both river systems, and with time intermediate populations became extirpated resulting in the current disjunct distribution.

Notropis missuriensis. The clade comprising Notropis missuriensis, including specimens collected at the type locality of Hybopsis missuriensis Cope, 1871 (a tributary of the Missouri River near St. Joseph, Missouri), represented by the lectotype designated by Fowler (1910) (ANSP 4374) is consistently recovered as monophyletic in all analyses except that of RAG1. Like N. stramineus sensu stricto, N. missuriensis is a member of the N. stramineus species complex. And like the three other members of this species complex, the sister taxon of this species is also unclear. Morphologically, N. missuriensis resembles most strongly N. multicorniculatus sp. nov., the two distinguished primarily by extent of tuberculation, with N. multicorniculatus sp. nov. possessing tubercles in a greater abundance on the head, body and pectoral fins.

The distribution of this species likely includes the north-flowing Red River of the North, a part of the Hudson Bay system, as well as tributaries of the Missouri and Mississippi Rivers in the Great Plains of the U.S. However, much of the western portion of the Missouri River basin was not sampled for the molecular portion of the present study, and therefore cannot be definitively assigned at this point to Notropis missuriensis (see grey area of uncertainty in Fig. 7). Therefore, based on the sampling herein, the distribution of this clade putatively includes the U.S. states South Dakota, Kansas, Minnesota, Iowa, Illinois, Missouri, and Tennessee, and is hypothesized to also include Colorado, Montana, Nebraska, North Dakota, Wyoming, and the Canadian provinces Manitoba and Saskatchewan as these areas include rivers that are part of the upper Mississippi River system. This distribution differs from that described for the subspecies N. s. missuriensis by Tanyolaç (1973), which includes the rivers of the southwestern U.S. (occupied instead by N. multicorniculatus sp. nov., N. oblitus sp. nov., and N. lucifer sp. nov.).

As mentioned above, future work is necessary to more finely delineate the range of this species. Some other North American leuciscids with members distributed in the eastern portion of the Missouri River basin have closely related congeners that are distributed in the western part of the Missouri River basin (e.g., Macrhybopsis, Hybognathus; Hoagstrom and Echelle 2022; Moyer et al. 2009, respectively). The possibility that this is also true for the N. stramineus species complex should be considered and tested.

While the membership of Notropis missuriensis remains consistent in all analyses, the relationships of this species are variable. While in both the cyt b gene tree and the concatenated analysis N. missuriensis is placed as the sister taxon to all remaining clades of the N. stramineus species complex + N. topeka, in the S7 gene tree it is instead the sister taxon to N. topeka, with these together as the sister taxon to N. stramineus sensu stricto. However, UCE analyses strongly support membership of N. missuriensis in the N. stramineus species complex. Therefore, the relationship of N. missuriensis and N. topeka depicted in the S7 gene tree (Fig. S2B) could stem from incomplete lineage sorting.

Notropis multicorniculatus sp. nov. Distributed in the Arkansas, Red, Canadian (native), and upper Pecos (likely non-native introductions from the Canadian River; C. Hoagstrom and M. Osborne, pers. comm; Hoagstrom et al. 2025; samples of “sand shiner” have also been reported from the lower Pecos, though these were not examined herein, and their species membership is as of yet unclear) drainages, this species is represented by the type specimen TCWC 15789.09, collected in the Canadian River in Hemphill County, Texas. Notropis stramineus sensu lato has been reported to have been found in the Trinity River, though it seems to be uncommon (Hendrickson and Cohen 2022) and no examples from this river were included in the present study. This species is not consistently recovered in analyses of individual gene trees but is well supported in the concatenated Sanger-loci analysis as well as in UCE analyses.

Notropis multicorniculatus sp. nov. contains the most genetic diversity among all recovered clades of N. stramineus sensu lato (Fig. S4; Table 3) and has largely inconsistent relationships across the different analyses. This diversity is not surprising given that this assemblage is distributed in the far south, and populations and species with southern distributions unaffected by glaciation can exhibit higher levels of genetic diversity and geographic structuring (Osborne et al. 2014). Higher levels of diversity within these clades could also be explained by vicariant events that occurred when populations entered independent Gulf Slope drainages and became isolated from source populations, even if the isolation was not permanent (Conner and Suttkus 1986). Notropis multicorniculatus sp. nov. is not monophyletic in any of the gene tree analyses. The closer relationship of some of the samples collected in the Canadian, Arkansas, Pecos, and Red rivers to N. topeka than to other samples collected in these very same rivers is most likely a result of limited phylogenetic information in the three Sanger-loci. Alternatively, this relationship could indicate incomplete lineage sorting and ancestral polymorphism between individuals within N. multicorniculatus sp. nov. and N. topeka, or mitochondrial introgression from N. topeka into populations of these rivers.

The genetic distinctiveness of Notropis stramineus sensu lato collected in the Arkansas River from those collected in the geographically proximate Missouri River system (Figs 5, S2) was identified first by Michels (2000) and then by Pittman (2011), both in unpublished dissertations (Fig. 38). Interestingly, N. multicorniculatus sp. nov. equates to two of the genetic “groups” recovered by Pittman (2011); these are her Groups A (Arkansas River) and R (Red River). Genetic distinctiveness of populations distributed in the Arkansas River and populations in the Missouri and Mississippi River systems has also been demonstrated for several additional taxa (e.g., Notropis nubilus; Berendzen et al. 2010, and Hybopsis amblops; Berendzen et al. 2008). While there are many fishes endemic to the Arkansas River system (e.g., Notropis girardi Hubbs & Ortenburger, 1929; an undescribed species in the Lepomis megalotis species complex; Kim et al. 2022), endemicity in the Arkansas River does not appear to be the case in N. multicorniculatus sp. nov., as individuals of this species from the Arkansas River system are closely related in both mitochondrial and nuclear DNA to individuals collected in the Red, Canadian and Pecos Rivers.

In some taxa, a phylogenetic affinity between populations of the Colorado River basin and the Pecos River has been demonstrated that is in line with hypothesized connections of the two systems created by paleogeographic events (e.g., the Etheostoma lepidum species complex, MacGuigan et al. 2021, Notropis amabilis and other species, Hoagstrom et al. 2025). This is, however, not what was observed herein. It is likely that populations of N. multicorniculatus sp. nov. in the Pecos River are a result of bait-bucket introduction, and the source population for the introduced individuals is the Canadian River, with native populations potentially extirpated from the Pecos (C. Hoagstrom and M. Osborne, pers. comm; Hoagstrom et al. 2025; see ‘Remarks’ section in species description); if true, this would explain the lack of an expected close phylogenetic relationship of these individuals to those of the Colorado River system.

Notropis oblitus sp. nov. This species is distributed within the Edwards Plateau ecoregion of central Texas, in the upper reaches of the Guadalupe and Nueces River basins. It may also occur in the intervening San Antonio basin but appears to be absent from the Colorado River basin based on our dataset. This species is represented by the type specimen TCWC 16882.13, collected in the Frio River in Uvalde Co., Texas. Though interrelationships of this species were not resolved in the RAG1 gene tree, N. oblitus sp. nov. forms a clade in the cyt b and S7 gene trees (though nested within the clade containing N. multicorniculatus sp. nov.), as well as in the concatenated-loci and UCE data analyses. Repeated patterns of shared ancestry among other members of the N. stramineus species group, however, suggests some level of past or semi-recent hybridization has shaped the current genetic composition of this lineage.

Notropis oblitus sp. nov. is a member of the N. stramineus species complex with an enigmatic phylogenetic position. In the mitochondrial cyt b gene tree, N. oblitus sp. nov. is recovered in a sister group relationship with N. lucifer sp. nov. collected in the Colorado river drainage, while in the S7 gene tree the species is nested within a clade comprising N. multicorniculatus sp. nov. The former could be the result of an earlier, not very recent hybridization, with divergence of this fragment of the cyt b gene of 2%. The latter likely results from a combination of incomplete lineage sorting and hybridization between N. oblitus sp. nov. and N. multicorniculatus sp. nov., as the two are recovered with high support within the N. stramineus species complex in UCE analyses, yet, like the other members of the complex (but to a greater degree), demonstrate high discordance of topologies within the group among the UCE gene trees (see Fig. 8). The most common topology found in analyses of UCEs recovered N. oblitus sp. nov. as the sister taxon to N. stramineus sensu stricto, though there is little support. An alternative topology recovered more rarely places N. oblitus sp. nov. as the sister taxon to the remaining three members of the N. stramineus species complex (Fig. S3).

Given the high levels of endemism this region is known for, it is not so surprising to uncover evolutionarily distinct lineages here within a supposedly widespread species. In fact, the Edwards Plateau ecoregion has repeatedly been shown to harbor previously unrecognized diversity in species with disjunct distributions. For example, the Etheostoma lepidum complex, which was shown to likely comprise three species, one in the eastern part of the Edwards plateau, a second in the Pecos River, and a third in the Conchos and San Saba Rivers (Colorado River tributaries; MacGuigan et al. 2021). The genus Dionda is a more extreme example, with even more granular structure in this region, likely due to the habitat preference of the genus being largely for headwater habitats and therefore more susceptible to isolation after headwater capture (Schönhuth et al. 2012).

Notropis oblitus sp. nov. is the most morphologically distinctive member of the N. stramineus species complex, with a significantly larger eye, a more silvery and translucent appearance in life, and generally smaller body size. Some of these characteristics appear to be intermediate between those of N. lucifer sp. nov. and those of the remaining members of the N. stramineus species complex. For example, N. lucifer sp. nov. from the Devils River lacks a cleithral streak, and the cross-hatching pattern dorsal to the lateral line, and has reduced pigmentation. This is true to a lesser degree in N. oblitus sp. nov. and in N. lucifer sp. nov. from the Colorado River basin. This could be a result of what appears to have been semi-recent hybridization between populations of N. lucifer sp. nov. in the Colorado River basin and N. oblitus sp. nov. but could also correlate with adaptation to similar habitat types (the clear, bedrock bottom streams these species inhabit versus the turbid, sandy-bottomed streams of the other species).

Notropis lucifer sp. nov., Notropis chihuahua, and mito-nuclear discordance. Notropis lucifer sp. nov. comprises individuals collected in the Devils River (a tributary of the Rio Grande) and the Colorado River basin of Texas. This species is distinct from other members of N. stramineus sensu lato most obviously in that it possesses reduced pigmentation and sparsely scattered large, macromelanophores along the head and body, which is shares with its apparent sister taxon, N. chihuahua (Figs 13–16E, F, I, 36).

While across all analyses Notropis lucifer sp. nov. was consistently recovered as the sister group of N. chihuahua (LPP = 1; BS=100), relationships of N. lucifer sp. nov. + N. chihuahua are inconsistent among gene trees. In the mitochondrial cyt b gene tree, specimens of N. lucifer sp. nov. from the Colorado River basin (Clade 6) are recovered as the sister taxon to N. oblitus sp. nov. while N. lucifer sp. nov. collected from the Devils River (Clade 5) is nested within a clade containing Notropis chihuahua among outgroup taxa. For the nuclear intron S7, Clade 6 is the sister taxon to Clade 5, and these together form the sister taxon to N. chihuahua. In the slower-evolving RAG1, N. lucifer sp. nov. and N. chihuahua comprise a single intermixed clade and are not reciprocally monophyletic (a single specimen belonging to Clade 6 is nested within Clade 3 in RAG1; it is possible however that this is a result of incomplete lineage sorting [ILS] given the relatively slower rate of evolution in RAG1). This close relationship of N. lucifer sp. nov. and N. chihuahua in both nuclear gene trees, as well as in the UCE loci, but distant relationship in the cyt b gene tree of N. lucifer sp. nov. Colorado populations from N. lucifer sp. nov. in the Devils River and N. chihuahua, may appear to be a straight-forward case of introgressive hybridization at first glance, but is actually more perplexing.

While Notropis lucifer sp. nov. and N. chihuahua are readily distinguished morphologically, with N. chihuahua differing most obviously in pigmentation (e.g., large “macromelanophores”; Burr and Mayden 1981: 257), the position of N. lucifer sp. nov. collected in the Devils River in the cyt b gene tree (nested within N. chihuahua), and uncorrected genetic distance between the two being only ~0.3% (Table 3), a number that would be regarded as safely conspecific, all suggests the likely scenario of mitochondrial introgression between N. chihuahua and the Devils River population of N. lucifer sp. nov. However, the origin of the putatively introgressed mitochondrial DNA is less clear. Morphologically, both N. lucifer sp. nov. and N. chihuahua appear to belong in the N. stramineus species group, and nuclear loci (S7, RAG1 and UCEs) place them within this clade. Yet, this nearly identical mtDNA shared between N. lucifer sp. nov. in the Devils River and N. chihuahua is ~15% divergent in uncorrected genetic difference (Table 3) from the N. stramineus species group, placing them instead among outgroup taxa (viz. N. ammophilus, N. boops, N. anogenus, N. heterodon, N. longirostris, N. mekistocholas, N. nubilus, N. rafinesquei), with N. uranoscopus as the sister taxon to this clade. Since the phylogenetic affinity of N. chihuahua and the N. stramineus species group appears to be genuine, the introgression of mtDNA from N. chihuahua to N. lucifer sp. nov. in the Devils River or vice versa does not explain the placement of these two taxa well outside of the group containing clades of N. stramineus and close relatives N. procne and N. topeka.

One possible explanation for this discrepancy is that the mitochondrial DNA shared by N. lucifer sp. nov. in the Devils River and N. chihuahua has entered into the mitochondrial genome of the two taxa from a species not sampled in this study. This could have occurred through one or a combination of several possible means. In one scenario, this unsampled species, whether extinct or extant, could have hybridized with either N. chihuahua or Devils River N. lucifer sp. nov., the two later then hybridizing with each other. In a second scenario, the unsampled species hybridized with both species independently, resulting in the shared mitochondrial signal. These scenarios would represent the “indeterminate” introgressive signal of Near et al. (2011), or a “mitochondrial poltergeist” of Fritz et al. (2024). The event that caused the mitochondrial introgression either did not involve Colorado River populations of N. lucifer sp. nov., or, alternatively, populations of N. lucifer sp. nov. in the Colorado River basin were involved in the introgression event, but subsequent (more recent) hybridization events involving these populations and N. oblitus sp. nov. occurred that obscured this older signal. This would be consistent with what appears to be some signals of introgression between N. lucifer sp. nov. in the Colorado River basin and N. oblitus sp. nov.

Hybridization resulting in introgression is not rare in freshwater fishes and may be a source of genetic diversity and an agent of evolutionary change rather than the evolutionary sink it is often thought to represent (Hubbs 1955; Arnold 1997; Seehausen 2004). Hybridization may be more of a general rule than an exception in many groups of North American freshwater fishes (Near et al. 2011). It has been reported from several groups, including darters (Near et al. 2011), catfishes (Egge and Simons 2006), salmonids and catostomids (Smith 1992). Recent studies of North America minnows have shown hybridization, even intergeneric hybridization, to be extremely common (Zbinden et al. 2023), and introgression resulting from hybridization has likely played a major role in shaping the mitochondrial signal seen in many groups of minnows and shiners (e.g., species of the Macrhybopsis aestivalis species complex; Echelle et al. 2018; Hoagstrom and Echelle 2022; Cyprinella in central Texas; Broughton et al. 2011).

It appears that the membership of N. chihuahua in the N. stramineus species group is well-supported by molecular analyses and morphology. Consistent placement of N. chihuahua and N. lucifer sp. nov. with the remaining members of the N. stramineus species group using both nuclear loci, particularly in the slower evolving RAG1, and UCEs, as well as highly congruent morphology, points to a close relationship and a phylogenetic affinity of these taxa. A close affinity of N. chihuahua with N. stramineus has been proposed before by Burr and Mayden (1981).

Though Notropis lucifer sp. nov. at present does not seem to be in immediate danger of conservation risk, its limited, disjunct distribution and the history of anthropogenic change in the region warrant keeping a close eye on the species in the near future.

Notropis procne , N. alborus and N. albizonatus. Notropis procne, the swallowtail shiner, is distributed widely throughout the Atlantic Slope drainages and, importantly, is the type species of the genus Miniellus (Fig. 35). This species has long been considered to be closely related to N. stramineus sensu lato, e.g., Snelson (1971) considered Notropis procne to be “an Atlantic coast derivative” of N. stramineus. In the present study, N. procne is consistently recovered as a member of the N. stramineus species group.

Snelson (1971) outlined the Notropis procne species group to include N. procne, N. alborus, N. albizonatus (at the time yet undescribed), N. heterolepis Eigenmann & Eigenmann, 1893, N. mekistocholas, N. stramineus, and N. uranoscopus. Snelson (1971) drew comparisons between N. procne and N. alborus, contrasting them with the newly described N. mekistocholas, e.g., in “… lacking herbivorous modifications, in having seven rather than eight anal rays, and in numerous other features”. Despite drawing these similarities between N. procne and N. alborus, Snelson (1971) concludes that N. alborus shares the greatest affinity with N. heterolepis, and is probably an “Atlantic coast derivative” of that species (presently, the generic placement of N. heterolepis remains unclear, with Stout et al. (2022) placing the species within the grade ‘Notropis’). Mayden (1989) tentatively placed N. alborus within the subgenus Hybopsis but cautioned that this assemblage may have been constructed based on convergent characters related to a preference for benthic habitats. Warren et al. (1994) used allozyme data to construct a hypothesis of relationships within the N. procne species group and is the only study to date to include N. alborus, N. albizonatus, N. procne, N. chihuahua and N. stramineus together in a phylogenetic study. Unfortunately, Warren et al. (1994) treat N. alborus as an outgroup a priori, a decision based on the hypothesis of Mayden (1989) that N. alborus was an outgroup of Notropis (Warren et al. 1994: 880), and thus any close relationship to N. procne may have been obscured. Schönhuth et. al. (2018) placed N. alborus as the sister taxon to a clade containing N. procne, N. heterodon (though see above discussion section “Notropis heterodon”), N. topeka and N. stramineus in a concatenated analysis. Analyses herein of the cyt b and RAG1 loci support the placement of N. alborus within the N. stramineus species group (Fig. S2A, C), likely as the sister taxon to N. procne (an S7 sequence was not available for N. alborus). Stout et al. (2022) did not include N. alborus in their analyses. All taken together, it appears as if a combination of morphological evidence, e.g., Snelson (1971), and molecular evidence (the present study, Schönhuth et al. 2018) supports the inclusion of N. alborus as a core member of the N. procne species group sensu Warren et al. (1994) and therefore also the N. stramineus species group.

Snelson (1971) referred to the undescribed Notropis albizonatus as the “stramineus-like form from the Cumberland and Tennessee drainages”. Notropis albizonatus, the palezone shiner, long recognized as distinct, was eventually described as a species endemic to the Tennessee and Cumberland River drainages (Warren et al. 1994). Warren et al. (1994) also considered N. albizonatus to be a member of the Notropis procne species group, most closely related to N. ludibundus (= stramineus sensu lato). While neither the present analysis nor those of Mayden et al. (2006) or Stout et al. (2022) included samples of Notropis albizonatus, it cannot be ruled out as a member of the N. stramineus species group, as an analysis of allozymes in Warren et. al. (1994), the only study to include all of these species (see above), placed N. albizonatus as the sister taxon to a clade containing N. chihuahua and N. uranoscopus, with N. ludibundus (= stramineus sensu lato) + N. procne as the sister group to that assemblage. In other words, N. albizonatus was nested within a clade containing those species found in the present study to be closely related to N. stramineus. Mayden et al. (2006) also discuss the possibility that N. albizonatus, N. chihuahua and N. uranoscopus could belong within their Miniellus, though they did not include these species in their analysis. As recommended by Mayden et al. (2006), further study is needed regarding the position of N. albizonatus, though material is scarce due to its extremely limited distribution and federally endangered status (Warren et al. 2000).

Notropis topeka. The Topeka shiner (Notropis topeka) is an endangered species distributed in the Midwest in the Mississippi and Missouri River system. This species is the most morphologically distinct member of the N. stramineus species group, with males developing a bright red or orange head and fins, and impressive tubercles distributed mainly on the top of the head and along the anteriormost portions of the fins (Figs 21C,F,I, 33, 34A).

Analyses of the individual cyt b, S7 and RAG1 loci placed Notropis topeka among clades of N. stramineus sensu lato, while analyses of UCEs placed the species outside rather than within the N. stramineus species complex with strong support. This result suggests a weak or misleading phylogenetic signal due to incomplete lineage sorting, or introgression, overcome by analysis of many UCE loci. Hybridization between N. topeka and N. missuriensis or N. multicorniculatus sp. nov. would not be surprising given the sympatry of N. topeka and these two species, and the fact that N. topeka appears to have arisen within the clade comprising the N. stramineus species complex + N. procne. This could represent an example of peripatric speciation, a hypothesis that has also been proposed in the past; Schmidt and Gold (1995) uncovered a sister-group relationship of N. topeka and N. stramineus (using the mitochondrial locus cyt b) and refer to the Topeka shiner as a “regional derivative” of the “widespread N. stramineus”.

Species identification in the Notropis stramineus species group. Distinguishing between members of the Notropis stramineus species group presents some challenges, due to the subtle nature of the diagnostic characters, and the fact that many of the characters relate to sexual dimorphism and are thus only useful when in the possession of adult males in breeding condition (e.g., nuptial coloration, tuberculation). Despite this, distinguishing between the species is not an impossible task, perhaps making the descriptor “cryptic” for this (and probably other similar species complexes) inappropriate.

Notropis topeka, with its bright red coloration in life and extensive tuberculation (in males), deep body, large number of scale rows, small eye and relatively small head, is perhaps the most distinctive member of the N. stramineus species group (Fig. 34A), its distinctiveness approached only by N. procne with its combination of a long, slender body, terminal mouth and characteristic dark lateral stripe (Figs 34B, 35). Notropis chihuahua (Figs 34C, 36), despite an overall similar external appearance to N. stramineus sensu lato, is also readily distinguished by its distinctive pigmentation pattern comprising large macromelanophores across the body and head, more abundant than in N. lucifer sp. nov.

In life, the blueish sheen along the bodyside of Notropis stramineus, N. missuriensis and N. multicorniculatus sp. nov., in combination with the scales dorsal to the lateral-line scale row comparatively densely outlined in melanophores, may be the most obvious way to distinguish these taxa from those distributed farther south in Texas and Mexico. In turn, the species in Texas and Mexico can be most readily distinguished from each other externally by the very large size of the eye in combination with peach-to yellow pectoral fin pigmentation and a cleithral streak in N. oblitus sp. nov. in comparison to a lack thereof in N. lucifer sp. nov. (cleithral stripe sometimes weakly developed, particularly in males, in Colorado River individuals). Interestingly, both N. oblitus sp. nov. and N. lucifer sp. nov. are quite pale in life, perhaps due to inhabiting somewhat similar habitats (clear, spring-fed streams flowing over calcareous bedrock) when compared to the silty, sandy streams in which the remaining members of the N. stramineus species complex are typically distributed (compare Figs S14 and S15 with Figs S16 and S17). Notropis stramineus can most readily be distinguished from N. missuriensis and N. multicorniculatus sp. nov. by a comparatively larger eye, more slender head, and elongate body. If fortunate enough to be comparing nuptial males, tuberculation in N. missuriensis and N. multicorniculatus sp. nov. is much more pronounced than in N. stramineus, a characteristic that appears to be quite consistent across the large distribution of the species.

Perhaps the most difficult species to distinguish from each other are Notropis missuriensis and N. multicorniculatus sp. nov. Some evidence points to a possible sister-group relationship between the two, and perhaps a geographically proximate distribution has facilitated a higher level of geneflow in recent times than has been the case in the other lineages. However, the two can be distinguished by the size and number of predorsal scales. The higher number of scales in N. multicorniculatus sp. nov. (circumferential and predorsal) by consequence results in a predorsal region that has smaller scales and is more crowded in appearance. Further, N. multicorniculatus sp. nov. tends to have a deeper head as well as a deeper body than N. missuriensis.

Comments on the N. stramineus species complex. In the present study, unraveling species limits and relationships within what we now call the Notropis stramineus species complex proved challenging due to conflicting signals in gene trees and a conserved morphology. For example, using Astral, a value of ~0.7 was calculated for the normalized quartet scores of all species trees estimated. This means that on average, ~30% of taxon quartets found in individual gene trees were not present in the resulting species trees, suggesting gene-tree discordance is present in all of the datasets examined herein. This discordance mainly pertains to the Notropis stramineus species complex, its unclear intrarelationships further displayed by low support values (Figs 5, S3) and visualization of the posterior distribution of species trees in DensiTree (Fig. 8). This suggests that the N. stramineus species complex has been affected by several factors that make estimating phylogenetic relationships of the different clades extremely difficult. These factors most likely include incomplete lineage sorting and ancient (and potentially also recent) hybridization through secondary contact. While the most common topology suggests a sister-group relationship between Notropis stramineus sensu stricto and N. oblitus sp. nov., and N. missuriensis and N. multicorniculatus sp. nov., there is likely a great deal of shared ancestry among these groups, including possible ancient and recent hybridization and gene flow, and further study is needed to more confidently estimate the evolutionary history of this group.