Biogeography and Distribution of the Cryptic Species Rosyface Shiner Notropis rubellus and Carmine Shiner Notropis percobromus in Illinois
Conservation and management of cryptic species is particularly challenging since it can be difficult to determine their exact distributions. Knowledge of species distribution is required to recognize management units based on taxonomy and whether there are any pertinent biogeographic patterns that could be relevant to the development of a management plan. Such a situation exists in Illinois with the Rosyface Shiner Notropis rubellus and Carmine Shiner Notropis percobromus. There are no reliable anatomical characters to distinguish between the two species. Instead, they are distinguished by genetic data. Samples were obtained for these two species from across the state from each major watershed in which they are found. Based on cytochrome b sequence variation the two species can be distinguished by 44 species-distinct differences. Notropis percobromus is restricted to the Rock River watershed and nearby Mississippi River tributaries in northwest Illinois. Notropis rubellus is found in the Illinois River basin and Vermilion-Wabash watershed. This is a reflection of historical drainage patterns from the geologic period when eastern Illinois was connected to the Great Lakes, as opposed to today when all the rivers in this study are part of the Mississippi basin. Using a previously published rate of divergence of the cytochrome b gene, we estimated a divergence time for these two species that was consistent with a previous estimate—2.8–2.6 MYA. Statistically, there was no genetic difference among populations within N. rubellus or N. percobromus. Haplotype networks and a phylogenetic analysis do provide some evidence for potential bottlenecks/founder effects and/or haplotype-specific selection within each species.
CRYPTIC species are anatomically identical (as far as we know) but are distinguished based on other characteristics, such as molecular data (Bickford et al., 2007). Since the collection of molecular data are usually more expensive and time-consuming than physically examining an organism, we do not always know about a cryptic species. Even if we are aware of the existence, we may not know their geographic distribution. Until we know the details of a species' distribution, it is problematic to create and implement any conservation-management plans.
Examples of cryptic species in Illinois for which there is little insight regarding distribution are Rosyface Shiner, Notropis rubellus, and Carmine Shiner, Notropis percobromus. Notropis rubellus was originally described by Agassiz in 1850 based on specimens from Lake Superior near Sault St. Marie (Hubbs and Brown, 1929). Notropis percobromus was described by Cope in 1871 based on specimens from either Kansas or Missouri in the vicinity of St. Joseph, MO (Gilbert, 1978). Over the years, several other similar looking species were described but were eventually placed in the synonymy of N. rubellus (Eschmeyer et al., 2018). Taxonomists recognized that N. rubellus was probably a mix of species, but resolving individual species was difficult because they all looked nearly identical, and there were debates about the influence of ecophenotypic plasticity (Bailey and Allum, 1962; Humphries and Cashner, 1994). With only a couple exceptions, such as Notropis micropteryx (Cope, 1868) and Notropis suttkusi (Humphries and Cashner, 1994), there was little consensus on how to differentiate the taxa within the complex (Gilbert, 1978; Humphries and Cashner, 1994). The end result was one widely distributed species ranging from the upper Mississippi east through the Great Lakes to the edge of New England, then south to Tennessee, and extending back west to the Ozarks (Page and Burr, 1991; Boschung et al., 1998).
Wood et al. (2002) examined allozyme data in order to clarify the diversity within the N. rubellus species group. They studied 37 gene loci with samples from 33 populations across the entire range of N. rubellus. Three of these populations were within Illinois. Several methods of analysis were used, and the phylogenetic position of the Illinois populations changed from tree to tree, depending upon the method used to analyze the data (Wood et al., 2002).
The final interpretation of the results was only one species in Illinois, N. percobromus, which also ranged northwest into the upper Mississippi River basin and east into northern and central Indiana, as well as into the Ozarks (Wood et al., 2002). To the northeast of Illinois was N. rubellus. However, Wood et al. (2002) also commented that although the Illinois basin and Wabash basin populations were characterized as N. percobromus, they shared affinities with N. rubellus. The intermediate position and paraphyletic pattern of these populations in the phylogenies could be due to past mixing events between the species in the region or to the fact that the historical biogeography of the Wabash watershed is more complex than currently thought. The results for Illinois and Indiana were considered tentative and requiring additional data and study (Wood et al., 2002).
Berendzen et al. (2008) carried out such an additional phylogenetic study, combining analysis of mtDNA cytochrome b with analysis of the Wood et al. (2002) allozyme data set. The cytochrome b analysis included data from 122 specimens from 46 localities across the entire range of the N. rubellus species group, including four localities within Illinois. Notropis rubellus and N. percobromus were recovered as sister species that diverged an estimated 3.05–2.5 million years ago. For Illinois, populations within the Rock and upper Mississippi watersheds were assigned to N. percobromus, and the Illinois and Wabash watersheds were considered N. rubellus (Berendzen et al., 2008).
However, the most recent revision of the Peterson Field Guide to Freshwater Fishes of North America (Page and Burr, 2011) follows Wood et al. (2002) with just one species, N. percobromus, in Illinois. Simon (2011) also follows Wood et al. (2002) in regard to the distributions of N. percobromus and N. rubellus in adjacent Indiana, and the International Union for Conservation of Nature uses the same distribution for the two species in its global assessment of N. rubellus (NatureServe, 2013), not including it within Illinois.
This convoluted taxonomic history and present uncertainty has created a great deal of confusion among Illinois ichthyologists. The fact that previous genetic/allozyme studies were based on only three or four Illinois localities provides little geographic resolution. The purpose of this study is a) to clarify whether both N. rubellus and N. percobromus are within Illinois, or if there is only one, which species is present, b) if there are two species in the state, to determine their exact distributions, and c) to consider what the genetic patterns may tell us about the historical biogeography of these species and the region.
MATERIALS AND METHODS
Samples
Tissue samples were obtained for individuals collected from localities across the known range of the N. rubellus species group within Illinois (Fig. 1). Localities were chosen to represent each of the major watersheds and sub-watersheds. Individuals were collected with seines and backpack electrofisher, or by Illinois Department of Natural Resources and Illinois Natural History Survey staff as they conducted various fish surveys. Fish were preserved in 95% ethanol.
Citation: Ichthyology & Herpetology 106, 3; 10.1643/CI-17-668
DNA extraction, PCR, and sequencing
DNA was extracted from muscle tissues of fish preserved in 95% ethanol according to the manufacturer's protocol using the DNeasy Blood and Tissue Kit from Qiagen with the following modifications. Prior to the initial incubation with lysis buffer and proteinase K, the cut pieces of muscle tissue were rinsed and vortexed three times in 300 μl volumes of 1X phosphate buffered saline (PBS). After each rinse, a microcentrifuge was used to pellet the tissue fragments to make removal of the PBS easier. The incubation in lysis buffer and proteinase K was carried out at 56°C for 1 hour and 30 minutes. At the end of the DNA extraction process, the DNA was eluted using only 100 μl of the elution buffer and a single elution step.
PCR reactions to amplify a region of the cytochrome b gene spanning more than 1100 base pairs were conducted using the primers HA-danio and LA-danio (Mayden et al., 2007) under the following conditions. Each 50 μl reaction contained approximately 80 ng of DNA (although smaller amounts often produced good results), 2 U of GoTaq Flexi DNA Polymerase (Promega), 0.4 mmol of each dNTP, 0.5 μmol of each primer, 2.5 mmol of magnesium chloride, and 1X GoTaq Flexi buffer. The PCR cycling conditions were as follows: 2–3 minutes at 95°C (all reactions were moved from ice directly to a hot PCR block within the first minute of this condition to prevent false priming at temperatures lower than the optimal annealing temperatures), followed by 35 cycles of 94°C for 1 minute, 48°C for 1 minute, and 72°C for 2 minutes. This was followed by a final extension step at 72°C for 5 minutes.
After verifying the production of a product of the expected size using an agarose gel, PCR products were cleaned according to the manufacturer's protocol using the DNA Clean and Concentrator™-5 kit from Zymo Research. Instead of using the elution buffer provided with the kit, a very small volume of water (ca. 10–14 μl) was used for elution.
Samples were sequenced at the University of Chicago Comprehensive Cancer Center DNA Sequencing Facility using the primers HA-danio and LA-danio. Two internal primers were designed to ensure that the extreme ends and internal portions of the sequence were accurately determined. The two internal primers are Notr-int-L (sequence: 5′–TTGCCCGTGGCCTATACTAC–3′), which can be used together with HA-danio to sequence one half of the PCR product, and Notr-int-R (sequence: 5′–GGATGGATCGGAGAATAGCA–3′), which can be used together with LA-danio to sequence the other half of the PCR product.
These data were supplemented with previously deposited GenBank sequences from an additional nine individuals from four Illinois locations (accession numbers EU084777, EU084778, EU084779, EU084780, EU084813, EU084814, EU084815, EU084816, and EU084817) and five individuals from four locations in Indiana and Michigan (GenBank accession numbers EU084801, EU084802, EU084803, EU084804, and EU084806; Berendzen et al., 2008). The sequences from Indiana and Michigan were included because they were the most similar sequences in GenBank to the sequences from Illinois.
DNA sequence analysis
For the current study, each sequence was trimmed to generate a sequence of 1119 bases that spanned the same region of the cytochrome b gene. In each case 11 bases at the beginning of the coding sequence were removed because of potential ambiguities in this region, and 10 bases at the end of the sequence were removed for the same reason. As a means of preliminary characterization, each sequence was analyzed using the BLAST algorithm on the NCBI website (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch#) to assess whether it appeared to be a sequence from N. percobromus or from N. rubellus on the basis of previous research (Berendzen et al., 2008). Unique haplotypes were identified, and haplotype networks were derived using the program TCS (Clement et al., 2000).
To compare various sequences to one another (to identify characteristic patterns in nucleotide differences etc.), either BioEdit (Hall, 1999) or the sequence alignment program Multiple Alignment using Fast Fourier Transform (MAFFT) was employed (http://www.ebi.ac.uk/Tools/msa/mafft/).
To predict and compare the amino acid sequences for each haplotype, the program EMBOSS Transeq (https://www.ebi.ac.uk/Tools/st/emboss_transeq/) was used to translate the DNA sequences into predicted amino acids, and MAFFT was used to compare the resulting amino acid sequences. Arlequin version 3.5 (Excoffier and Lischer, 2010) was used to assess the number of private haplotypes in each watershed, to determine ratio of transitions to transversions, to conduct a pairwise analysis of the FST analogue ΦST, to run AMOVA analyses, and to conduct Mantel tests for isolation by distance. DnaSP version 5 (Librado and Rozas, 2009) was used to calculate haplotype diversity, average nucleotide difference, and nucleotide diversity. A Maximum Likelihood (Tamura-Nei model; Tamura and Nei, 1993) phylogenetic tree was constructed with MEGA version 7 (Kumar et al., 2016) in which all the sequences from the current study were analyzed together with all the sequences from the study of Berendzen et al. (2008). A bootstrap test of phylogeny was used with 1000 replications. A cytochrome b sequence from N. stilbius (accession number AF 352286) was used as the outgroup sequence. Initial trees for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. MEGA was also used to compute within group distances and between group mean distance for the two species.
RESULTS
Sequence alignment and species distribution
A total of 52 sequences from the protein coding region of the cytochrome b gene ranging in length from 1121 to the full length of 1140 nucleotides were obtained and deposited in GenBank (accession numbers MF541779–MF541813 and MF572914–MF572930). With the addition of the nine samples from Illinois previously identified by Berendzen et al. (2008), the total number of samples from Illinois considered was 61. This included samples from 24 locations, with 21 samples identified as N. percobromus and 40 as N. rubellus. The samples of N. percobromus all came from eight locations in the northwest region of the state, while the samples of N. rubellus all came from 16 locations in eastern and central Illinois. Although some populations of the two species are located relatively close to one another, no overlap in the geographic ranges of the species was detected.
Identification of haplotypes and other aspects of sequence characterization
A total of 11 unique haplotypes were found among the 21 samples of N. percobromus from Illinois (three of these, designated as NP9–11 in Figures 1 and 2 represent previously identified sequences EU084777, EU084779, and EU084780, respectively) and 21 unique haplotypes among the 40 samples of N. rubellus from Illinois (two of these, designated as NR20–21 in Figures 1 and 2 represent previously identified sequences EU084813 and EU084817, respectively). Figure 2 shows the likely relationships between the various haplotypes from Illinois, as well as four closely related sequences from Indiana that were previously deposited in GenBank (designated an NR-IN1 through 4 that represent previously identified sequences EU084801, EU084802, EU084803, and EU084804, respectively) and one from Michigan (designated as NR-MI1 that represents a previously identified sequence EU084806). Some haplotypes (e.g., NR1 and NR2) are abundant, while many represent single samples. There are many possible affinities represented, including a branch in which NR18 and NR19 appear more closely related to previously isolated samples from Indiana and Michigan than to those of other samples collected in this study.
Citation: Ichthyology & Herpetology 106, 3; 10.1643/CI-17-668
The 32 haplotypes representing samples from Illinois were compared by sequence alignment in order to detect bases that are always the same within a species and different between the two species. A total of 44 such species-distinct variations were identified (data not shown). Among these variations, 41 represent transitions and three represent transversions. All but one of the species-distinct changes in bases are changes that occur at the third base in a codon, and none result in a change at the level of amino acids. Other non-species-distinct variations in base sequences occur in some haplotypes, and among those, only three result in a single amino acid change. Presumably, these three changes occur in codons that encode amino acids that are somewhat freer to vary than are other amino acids in the cytochrome b protein. One such variant was found in NR12 and a sequence from N. rubellus previously deposited in GenBank (accession EU084815, which is identical to NR12 except for several ambiguities and some missing nucleotides at the beginning). Another variant found in NR14 does not appear in the cytochrome b genes of other samples of Notropis previously deposited in GenBank; however, it does occur in this gene in other fishes, including some in the family Cyprinidae as determined via a BLAST search. The third variant occurs in a sequence of N. rubellus (EU084817) previously deposited in GenBank but does not occur in other cytochrome b genes of Notropis found in GenBank. The only other cytochrome b sequence with this variant that was revealed by a BLAST search occurred in species in the order Salmoniformes (accession number NP_008302).
The between group genetic distance at the species level was 5.1% when a model based on the number of nucleotide differences was used. Other models gave somewhat higher values with 5.6% being the highest (when the Tajima-Nei or Maximum Composite Likelihood model was used). Within group mean distances were 0.4% for N. rubellus and 0.3% for N. percobromus regardless of which model was used. The mean number of nucleotide differences was 4.7 for N. rubellus and 3.4 for N. percobromus, and the haplotype diversity indices were 0.91 and 0.78, respectively.
Comparisons of haplotypes found within watersheds
We considered various parameters that relate to the nucleotide sequences of the haplotypes within each watershed (Table 1). All of the watersheds exhibited relatively high haplotype diversity (0.5 or greater). Another indication of genetic diversity is the presence of private haplotypes in almost all the watersheds, with the Leaf River watershed being the only one to lack at least one private haplotype. The values for average nucleotide differences, nucleotide diversity, and the ratio of transitions to transversions are also presented in Table 1. The values for nucleotide diversity are low but consistent with such values for some other freshwater fish populations (Alves et al., 2001; Mesquita et al., 2001; Perdices et al., 2004; Ma et al., 2010). The extreme bias towards transitions compared to transversions is a common phenomenon, especially in mitochondrial genes (e.g., Wakeley, 1996) and is expected to be greatest among more closely related sequences (Moritz et al., 1987). We also characterized the nucleotide composition for each haplotype and found a pattern of low G content (ranging between 16.9% and 17.3% in both species; data not shown) that is typical of the cytochrome b gene in many animal taxa (Johns and Avise, 1998).
We calculated and compared the FST analogue ΦST (a statistic developed by Excoffier et al., 1992 for use with DNA sequence data) and conducted a pairwise analysis among the watersheds. After implementing a Bonferroni correction, we found no comparisons that required rejection of the null hypothesis that there were no differences between watersheds. An AMOVA analysis for each species showed that most of the variation occurred within populations (91.6% for N. rubellus and 100% for N. percobromus) rather than between populations. A Mantel test (Mantel, 1967) for each species indicated that there was no evidence of isolation by distance (p = 0.22 for N. rubellus and p = 0.26 for N. percobromus).
A phylogenetic analysis (Fig. 3) that included the sequences from this study plus those from the study of Berendzen et al. (2008) placed all of the sequences from this study into two clades previously identified by Berendzen et al. (2008) as either N. rubellus or N. percobromus. This was in agreement with our initial BLAST results (Note, the other clades that were not specifically associated with N. rubellus or N. percobromus by Berendzen et al. [2008] were also resolved, but they are not shown here). The analysis also indicates relationships between some sequences from this study and various sub-clades identified among the sequences from Berendzen et al. (2008) that will be described below.
Citation: Ichthyology & Herpetology 106, 3; 10.1643/CI-17-668
DISCUSSION
Sequence variation between species
The sequence comparison and the haplotype analyses presented here should be of value for the purposes of species management in Illinois, and they may also provide some insight into the evolutionary history of the two species. Of particular significance for efforts aimed at species conservation and management is the enhanced insight regarding species distribution for the two species that our study has provided via the gene sequencing approach that we have presented. The comparison of each sequence in this study with other sequences previously deposited in GenBank gave unambiguous matches with one species or the other. Furthermore, comparing the estimates of between group vs. within group mean genetics distance further validates the ability of this method to distinguish between the two species. If the most conservative estimates of these values are used, a barcoding gap (e.g., Hebert et al., 2003; Meyer and Paulay, 2005) of 4.7% is revealed, a value as great or greater than that of most vertebrate sister species when cytochrome b sequences are compared (Avise and Walker, 1999).
Although our data set is not extensive at the species level, trends in sequence divergence can be used to generate estimates regarding the phylogenetic history of these two species in Illinois. Based on the rate of genetic divergence for the cytochrome b gene assumed by Berendzen et al. (2008) of 2.0% per one million years, and the between group genetic distance values for N. percobromus and N. rubellus, the average genetic divergence time estimated is 2.8–2.6 MYA. This is an estimate that is highly consistent with the estimated range of 3.05–2.50 MYA calculated by Berendzen et al. (2008) for these two species.
Species distribution and biogeography
The current study represents the most comprehensive population genetics study of the Notropis rubellus-percobromus species complex in the state of Illinois. It both verifies and extends the basic insights regarding the distribution in Illinois of the species N. percobromus and N. rubellus provided by Berendzen et al. (2008). While the previous study suggested a likely separation of the two species in Illinois, with N. percobromus in Stephenson (n = 2) and Ogle counties (n = 2) and N. rubellus in La Salle (n = 2) and McLean counties (n =3), the number of individuals (n = 9) and locations (n = 4) from Illinois was relatively low.
There is a remarkably clear distribution pattern between the two species in Illinois. The current study suggests that N. percobromus is restricted to northwestern Illinois (Fig. 1), in the Rock River watershed (which includes Willow Creek, Franklin Creek, Leaf River, and Crane Grove Creek) and Apple River. These are all tributaries of the upper Mississippi. Notropis rubellus occupies the eastern portion of the state in the Illinois and Wabash River basins.
Today all three primary watersheds (upper Mississippi, Illinois, and Vermilion-Wabash) are part of the Mississippi basin, but for significant periods of time since the retreat of the glaciers Lake Michigan drained through a Des Plaines–Illinois River channel (Willman, 1971; Wiggers, 1997). Hence the current distribution is a reflection of an older Great Lakes basin that supported N. rubellus and a Mississippi basin that supported N. percobromus. Other species, e.g., Rainbow Darter, Etheostoma caeruleum, have similar distributions (Ray et al., 2006). There is no evidence at this time for overlap between the ranges of N. rubellus and N. percobromus. This pattern has been observed with other taxa, with allopatry more common among cryptic species as compared to congeneric non-cryptic species (Vodă et al., 2015).
Sequence variation within species and phylogenetic associations
Statistically, there was no difference among populations within N. rubellus or N. percobromus according to pairwise comparisons of the FST analogue ΦST, indicating panmixia across the sampled ranges. This lack of structure in the current day populations was confirmed by the results of the AMOVA analyses and the Mantel tests for isolation by distance. To assess the evolutionary history of the two species, patterns associated with the haplotype network and the phylogenetic analysis were considered.
The placement and relative abundance of the haplotypes NR1 and NP1 in the haplotype networks (Fig. 2) suggest that they could be the founding haplotypes for the two species in areas of the sampled ranges; however, the phylogenetic analysis challenges this interpretation. That analysis indicates that the sequences NR2, NR9, NR11, and NR14 from N. rubellus (all of which are clustered together in the haplotype network) are all less derived than NR1, and similarly that the sequences NP4 and NP6 from N. percobromus (which are also clustered together in the haplotype network) are less derived than NP1. Since NR1 and NP1 are now the most common haplotypes, it appears that they may have originated from the less derived haplotypes and become the most common haplotypes either via an event that resulted in a population bottleneck that randomly favored these haplotypes or by these haplotypes conveying some selective advantage. In support of this interpretation, it is interesting to note that while NR1 and NP1 are the most common haplotypes in this study, Berendzen et al. (2008) did not detect these haplotypes outside of Illinois.
In Figure 3, haplotypes NR18 and NR19 are less derived than NR1; they are also in a different sub-clade than NR1. Both observations suggest that NR18 and NR19 are not derived from the NR1 haplotype as suggested by the associations shown in Figure 2. Consistent with the maximum likelihood results, NR18 and NR19 are from the Wabash watershed and are grouped in a sub-clade whose other members were sampled exclusively from locations in Indiana or Michigan.
Monitoring and management
Monitoring cryptic species has unique challenges. Besides routine status surveys, it is also possible that their ranges may change over time. These changes could be difficult to detect if we are reliant upon the relatively more time-consuming and expensive process of DNA analyses to recognize species. Environmental DNA (eDNA) may help monitoring in the future. This is a method in which DNA is extracted from a relatively easy to obtain water sample instead of from a tissue sample from an organism (e.g., de Souza et al., 2016), but the method still involves the associated lab work and costs.
Regardless of the exact method of taxon identification, monitoring cryptic species has unique challenges that need to be anticipated and incorporated into any conservation and management plans. The species examined in this study can be divided into smaller taxonomic units based on watersheds, and management plans should reflect this biogeography and the particular threats found in particular regions.
Distribution map of Notropis rubellus (blue squares) and Notropis percobromus (red circles) haplotypes used to determine the range of the two species in Illinois. Numbers in parentheses indicate how many individuals have the same haplotype at the same location. Rivers shown in dark blue are in the Upper Mississippi watershed, those in light blue are in the Illinois River watershed, and those in purple are in the Wabash River watershed.
Haplotype networks for sequences from this study and all others in GenBank from Illinois, as well as some from Indiana or Michigan (shown as shaded ovals) that were highly similar to sequences in this study (see text for details). The relative size of each oval is roughly proportional to the number of samples represented by the respective haplotypes with the smallest representing single samples. Nodes between ovals represent predicted but unsampled haplotypes. Each haplotype is different from the one above it by one nucleotide. (A) Haplotype network for sequences identified as Notropis rubellus. (B) Haplotype network for sequences identified as Notropis percobromus.
Maximum likelihood phylogenetic tree including all sequences from this study combined with all sequences from a previous phylogenetic analysis conducted by Berendzen et al. (2008). The tree with the highest log likelihood is shown. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Values shown along branches are bootstrap values. The scale bar represents the number of substitutions per site. Only the portion of the tree that included clades representing N. rubellus and N. percobromus are shown here. The sequences from GenBank from Indiana and Michigan that are shaded in Figure 2 are also shaded in this figure. See text for other details.
Contributor Notes
Associate Editor: M. T. Craig.