Revisiting the North American freshwater mussel genus Quadrula sensu lato (Bivalvia Unionidae): Phylogeny, taxonomy and species delineation

Freshwater mussels (Bivalvia, Unionidae) have suffered strong declines over the last century. High morphological plasticity of Unionidae causes disturbances in their systematics and taxonomy, hampering conservation efforts. Species that have historically been placed under the North American genus Quadrula have suffered from numerous taxonomic and species delineation problems since its inception. Four genera are presently recognized within Quadrula sensu lato, that is, Cyclonaias, Quadrula, Theliderma and Tritogonia, but their phylogenetic basis remains incompletely tested. In the present study, we reconstructed several two‐marker (mtDNA cytochrome c oxidase subunit I—COI and NADH dehydrogenase subunit 1—ND1) phylogenies with newly collected specimens and all previously available sequences covering most species within this group. We then delineated the species within the group using an integrative approach with the application of molecular statistical methods, morphometric (Fourier Shape) analyses and geographic distribution data. Four clades corresponding to these genera were consistently recovered in all phylogenies. To validate the generic status of these clades, molecular analyses were complemented with morphological, anatomical and ecological data compiled from the literature. Several revisions are here proposed to the current systematics and taxonomy of these genera, including the synonymization of Cyclonaias asperata under Cyclonaias kieneriana; the inclusion of Quadrula apiculata and Quadrula rumphiana under Quadrula quadrula; the placement of Quadrula nobilis under Tritogonia; and finally the separation of the Mobile River basin populations of Theliderma metanevra as a new species, that is, Theliderma johnsoni n. sp. The conservation implications of the proposed changes are then discussed.


| INTRODUCTION
Conservation programs and strategies are largely based on species as conservation units, making species delineation extremely important as a basic conservation tool (Prié, Puillandre, & Bouchet, 2012). However, taxon-based conservation strategies dedicated to the freshwater mussel family Unionidae, one of the world's most endangered taxa, are hindered by phylogenetic and taxonomic uncertainties (Lopes-Lima et al., 2017). This is especially true within the most species-rich Unionidae subfamily, the North American Ambleminae. Across the most recent systematics studies, the Ambleminae is divided in five tribes (Pfeiffer et al., 2019). However, polyphyly and inappropriate species boundaries have been revealed in some of these tribes, including the Quadrulini (Lydeard, Minton, & Williams, 2000;Pfeiffer et al., 2016;Serb, Buhay, & Lydeard, 2003). The quadruline freshwater mussels are distinctive animals producing thick quadrate shells, some of which are heavily sculptured. Shell morphology is highly variable within some species from this group, hindering unambiguous species identification or generic assignment. As shell morphology has been the original basis for Quadrulini systematics and taxonomy to date, the systematics and composition of this tribe have suffered a series of changes since its first description in the early 1900s (see Supporting Information Appendix S1 for an extensive taxonomic history of the Quadrulini). From the beginning of the 20th century, species that had been historically placed within the genus Quadrula sensu lato have been divided into four main species groups, that is, the Quadrula sensu stricto, the pustulosa, the metanevra and the Tritogonia species groups (Supporting Information Appendix S1). A molecular phylogeny of these taxa by Serb et al. (2003) largely confirmed these groupings and recovered four clades: Quadrula sensu strictu, the pustulosa species group, the metanevra species group and a fourth clade including Tritogonia verrucosa and Quadrula nobilis. Although these four clades are commonly referred to as genera in regional checklists (Howells 2013;Parmalee & Bogan 1998;Williams, Bogan, & Garner, 2008) the molecular, morphological and ecological evidence supporting these groups remains limited.
The present study is focused on re-examining the phylogeny, systematics and taxonomy of Quadrula sensu lato, here defined as including the species from the genera Quadrula, Theliderma, Cyclonaias and Tritogonia (Williams et al., 2017). In detail, this study aims to: (a) estimate the phylogenetic relationships of specimens collected in Texas with all published Quadrulini sequences, using a two-marker approach (COI and ND1); (b) perform a comparative shell morphometry evaluation to complement the molecular results; (c) define species boundaries with a taxonomic revision of all analysed taxa; (d) test the four classical generic constructs and their evolutionary significance; and (e) describe the conservation implications of the obtained results.

| Sample collection and materials examined
Specimens of quadruline mussels were collected from 50 sites across the state of Texas during 2003-2011 ( Figure 1). A total of 89 specimens were collected and placed in 99% ethanol for molecular analyses. Voucher specimens were labelled and deposited in the SUNY Buffalo State College Great Lakes Center collections, Buffalo, New York (BSGLC). The field work was carried out with an appropriate Scientific Research Permit SPR-0503-300 issued by the Texas Parks and Wildlife Department. Additionally, dry shell specimens of the target nominal species were selected for morphometry from specimens deposited at the North Carolina Museum of Natural Sciences (NCMS) and BSGLC (See Supporting Information Table S1 for the examined lot numbers).

| Sequencing, alignments and phylogenetic analyses
A total of 31 quadruline specimens, including all nominal taxa across the state of Texas, were selected for molecular analyses (Table 1). For each sample, genomic DNA extraction (Froufe et al., 2014), amplification and bidirectional sequencing were carried out for the F-type mtDNA cytochrome c oxidase subunit I (COI) and NADH dehydrogenase subunit 1 (ND1) genes. For COI, the primers LCO_22me and HCO_700dy (Walker et al., 2006) were used with an annealing temperature of 50°C and polymerase chain reaction (PCR) conditions as described in Froufe et al. (2014). ND1 was amplified using the PCR conditions and primers (Leu-uurF and LoGlyR) of Serb et al. (2003). Sequences were obtained with the BigDye sequencing protocol (Applied Biosystems 3730xl) by Macrogen Inc., Korea. Forward and reverse sequences were edited and assembled using ChromasPro 1.7.4 (Technelysium, Tewantin, Australia). All new sequences have been deposited in GenBank (Table 1 and Supporting  Information Tables S2 and S3). Three datasets were constructed as follows: one for COI, another for ND1 and a third concatenating COI and ND1. The COI and ND1 datasets included all newly sequenced individuals and all Quadrulini sequences available in GenBank database for each gene (Supporting Information Tables S2-S4). The COI + ND1 dataset included all individuals sequenced for both COI and ND1 plus GenBank Quadrulini specimens with sequences available for the two genes (Supporting Information Table S4). For each of the three datasets, sequences of additional specimens were downloaded from Genbank and/or newly sequenced as outgroup (details in Supporting Information Tables S2-S4). The three datasets were aligned with the MAFFT multiple sequence alignment algorithm (Katoh & Standley, 2013). Each individual gene alignment was then restricted to its unique haplotypes, retrieved using DnaSP v5.1.0.1 (Librado & Rozas, 2009).
Phylogenetic analyses were then performed on the three datasets using Bayesian inference (BI) and maximum likelihood (ML). For the BI analyses, the best-fit models of nucleotide substitution were selected using JModelTest 2.1.10 (Darriba, Taboada, Doallo, & Posada, 2012) under the Bayesian information criterion. For each individual gene dataset, a three partition scheme was applied, one per gene codon, with the following selected models: COI (GTR + I + G, HKY, HKY + G), and ND1 (HKY + G, HKY + G, GTR + I + G). For the COI + ND1 dataset, a six partition scheme was applied for the three codons of both COI and ND1 with the same models selected for the individual gene datasets. BI analyses were performed in MrBayes v3.2.6 (Ronquist et al., 2012) implemented in CIPRES Science Gateway (Miller, Pfeiffer, & Schwartz, 2010). BI analyses were initiated with program-generated trees and four Markov chains with default incremental heating. Two independent runs of 30 × 10 6 generations were sampled at intervals of 1,000 generations producing a total of 30,000 trees. Burn-in was determined upon convergence of log likelihood and parameter values using Tracer 1.6 (Rambaut, Suchard, Xie, & Drummond, 2014).
For the ML analyses, the same partitioning scheme was applied for each dataset with the same model (GTR + G) for all partitions, and sequences were then analysed in RaxML 8.2.10 HPC Black Box (Stamatakis, 2014) with 1,000 bootstrap replicates. Haplotype networks were calculated using TCS 1.21 (Clement, Posada, & Crandall, 2000) with a threshold of 95%.

| Molecular based species delineation methods
Five distinct molecular methods were applied to determine the number of molecular operational taxonomic units (MOTUs). All methods were applied to the COI, ND1 and concatenated (COI + ND1) datasets, with the exception of the BIN system that relies only on COI. The first two are distance based, that is, the BIN system implemented in BOLD (Ratnasingham & Hebert, 2013) and the Automatic Barcode Gap Discovery (ABGD) . For the BINs system, the COI dataset without the outgroups was analysed with the Cluster Sequences tool implemented in BOLD 4 (http://v4.boldsystems.org) (Ratnasingham & Hebert, 2013). The ABGD species delineation tool was applied to all three datasets without outgroup using its online version (http://wwwabi.snv.jussieu.fr/public/abgd/abgdweb. html) with the default settings and the Kimura-2-parameter distance matrix . Two tree-based molecular species delineation methods were applied to all datasets, that is, the single threshold Generalized Mixed Yule Coalescent (GMYC) model (Fujisawa & Barraclough, 2013) and the Bayesian implementation of the Poisson Tree Processes model (bPTP) (Zhang, Kapli, Pavlidis, & Stamatakis, 2013). For the GMYC method, a Bayesian ultrametric phylogenetic tree was first generated in BEAST 2.4.6 (Bouckaert et al., 2014) with the previously selected models for each partition and four independent runs of 20 × 10 6 Markov chain Monte Carlo (MCMC) generations, sampled every 1 × 10 3 generations. Convergence of the parameters was evaluated using Tracer 1.6 software (Rambaut et al., 2014). The consensus tree was annotated using TreeAnnotator 2.4.6 (Bouckaert et al., 2014). The consensus tree was loaded into the R software package "Species Limits by Threshold Statistics" (Ezard, Fujisawa, & Barraclough, 2009) in R 3.2.0 (R Core Group available via http://www.r-project.org) and analysed using the single threshold model. For the bPTP, the BI phylogenetic trees previously obtained were used as input trees in the bPTP web server (available at: http://species.h-its.org/) with 1 × 10 6 iterations of MCMC and 20% burn-in. Finally, a 95% statistical parsimony connection limit was used, by using TCS 1.21 (Clement et al., 2000). Sequence divergences (uncorrected p-distance) were assessed using MEGA 7 (Kumar, Stecher, & Tamura, 2016).

| Morphometry
For a detailed analysis of inter-and intraspecific variation in shell shape within the quadruline genera Cyclonaias, Quadrula and Theliderma, we used Fourier Shape Analysis, as developed and explained by Crampton and Haines (1996). This method decomposes xy-coordinates of a shell outline into a number of harmonics, each of which is in turn explained by two Fourier coefficients. Xy-coordinates of the sagittal shell outline of 1,222 specimens from BSGLC and NCMS collections (739 specimens of Cyclonaias spp., 254 specimens of Quadrula spp. and 229 specimens of Theliderma spp.; Supporting Information Table S1) were obtained from digital photographs using the program IMAGEJ (Rasband, 2008) and subjected to fast Fourier transformation using the program HANGLE, applying a smoothing normalization of 3 to eliminate high-frequency pixel noise. Preliminary analysis indicated that the first 10 harmonics described the outlines with sufficiently high precision. Discarding of the first harmonic, which does not contain any shape information, resulted in a set of 18 Fourier coefficients per individual. Outlines of all specimens within each of the three genera were then rotated to maximum overlap by program HTREE, resulting in the final set of 18 Fourier coefficients per individual.
For visual examination of variation in shell shape within and between true and nominal species, principal component analysis were performed on the 18 Fourier coefficients of (a) all true species (recognized by the molecular species delineation methods, see results) of Cyclonaias, including a maximum of 50 specimens per species; (b) all nominal species of Cyclonaias pustulosa; (c) only Cyclonaias kieneriana and Cyclonaias kleiniana; (d) all nominal species of Quadrula; (e) all true species (recognized by the molecular species delineation methods, see results) of Theliderma; and (f) only Theliderma metanevra and Theliderma johnsoni n. sp. (see Supporting Information Appendix S2 for a detailed description of T. johnsoni n. sp.). Synthetic outlines of extreme and average shell shapes were drawn using program HCURVE as explained in Crampton and Haines (1996).
We assessed the rate of accurate identification of true and nominal species based on shell shape using linear discriminant analysis (LDA) on the 18 Fourier coefficients. To test for statistically significant differences in sagittal shell shape between species, multivariate analyses of variance (MANOVA) were run on the 18 Fourier coefficients. Pairwise Hotelling's post hoc tests were performed to identify significant differences between each pair of true/nominal species. Statistical analyses were performed in PAST v.3 (Hammer & Harper, 2006).

| Ecological, morphological and anatomical traits
An extensive bibliographic review of selected ecological, morphological and anatomical traits was accomplished for all species within Quadrula s.l. (Table 2; Supporting  Information Table S5).

| Alignments and phylogenetic analyses
The COI dataset spanned 582 nucleotides (nt) and included 289 unique haplotypes (232 polymorphic and 192 parsimony informative sites). The ND1 dataset covered 619 bp with 339 unique haplotypes (297 polymorphic and 257 parsimony informative sites). Finally, the combined COI + ND1 dataset was 1,192 nt long and included 325 individual sequences (501 polymorphic and 427 parsimony informative sites). No insertions or deletions, and no stop codons were observed in any of the datasets after translating all sequences to amino acids.
The results of the BI and ML phylogenetic analyses for the three datasets presented similar topologies (Table 3), thus only BI phylogenetic trees are shown in Figures 2-4. In the COI phylogeny, the Quadrulini clade is monophyletic and well supported in the BI analyses. Within the Quadrulini clade, the Megalonaias + Uniomerus clade is sister to a clade including three well-supported subclades corresponding to the genera Quadrula, Tritogonia, and Theliderma, and a clade including all Cyclonaias sequences ( Figure 2).
The ND1 phylogeny recovered similar phylogenetic patterns to that obtained with COI. However, in these analyses, the Quadrulini is not monophyletic, with the remaining Ambleminae tribe clades, that is, Amblemini, Pleurobemini and Lampsilini clustering within the Quadrulini tribe clade (Figure 3). The Uniomerus clade is sister to a clade containing the four remaining Quadrulini genera (i.e., Quadrula, Tritogonia, Theliderma and Cyclonaias; Figure 3). While Cyclonaias, Quadrula and Tritogonia are well supported, Theliderma has a low support value (Figure 3). The COI + ND1 phylogeny shows Quadrulini as monophyletic with Uniomerus being sister to a clade comprising four well-supported clades (Quadrula, Tritogonia, Theliderma and Cyclonaias; Figure 4).

| Quadrula
All sequences from the nominal species Quadrula quadrula, Quadrula apiculata and Quadrula rumphiana cluster within the Q. quadrula clade in all phylogenies ( Figures  2-4). However, both nominal species Q. apiculata and Q. rumphiana were found to be nested within Q. quadrula (Figures 2-4). Both the COI and ND1 95% threshold haplotype networks of the Q. quadrula clade reveal a low number of mutations among the nominal taxa Q. quadrula, Q. apiculata and Q. rumphiana (Figure 5a,b).

| Theliderma
Not many COI sequences of Theliderma are represented in the COI dataset, and therefore in the COI and COI + ND1 phylogenies (Figures 2 and 4). Nevertheless, in these phylogenies two distinct clades were obtained in sequences from specimens of T. metanevra: one corresponding to specimens from the Tennessee basin, and the other with specimens from the Mobile basin (Figures 2 and 4). The ND1 phylogeny is better represented with all species recognized to date except for Theliderma stapes (Figure 3).

| Tritogonia
The sequences of specimens originally identified as Q. nobilis cluster together with those from T. verrucosa in all phylogenies forming a well-supported clade here assigned to Tritogonia (Figures 2-4).

| Theliderma
The pairwise uncorrected p-distance among all the nominal Theliderma species ranged between 4.0% and 10.6% for ND1 T A B L E 4 Pairwise genetic distance matrixes of nominal quadruline species of the genera Cyclonaias, Quadrula, Theliderma, and Tritogonia,  Notes. Left: mean uncorrected p-distance within putative species for cytochrome oxidase subunit I (COI) and for NADH dehydrogenase, subunit 1 (ND1) genes. Right: mean uncorrected p-distance among putative species of COI (below the diagonal) and ND1 (above the diagonal) genes.
( Table 4). The higher within p-distance recorded value was reached for T. metanevra, 1.7% for COI and 2.1% for ND1 (Table 4). All originally described Theliderma species are here recognized as MOTUs with T. metanevra being further divided in two distinct MOTUs, that is, T. metanevra for specimens from the Tennessee River basin and T. johnsoni n. sp. from the Mobile River basin ( Table 5). The p-distance values among the recognized Theliderma MOTUs varied between 3.5% and 10.1% for ND1 (Table 6). The p-distance within each of the recognized MOTUs was ≤0.9% for ND1 (Table 6).

| Cyclonaias
Linear discriminant analysis on the 18 Fourier coefficients extracted through Fourier Shape Analysis for all Cyclonaias species recognized in this study assigned 75% of individuals to the correct species (Figure 6a). Species that are particularly difficult to separate by shell shape are C. kieneriana and C. pustulosa (16% misidentifications), and Cyclonaias infucata and C. kleiniana (10%). In addition, most true species differed significantly from each other in their shell shape as approximated by 18 Fourier coefficients, with the exception of C. infucata and C. kleiniana (MANOVA, pairwise Hotelling's test p = 0.742), and C. infucata and Cyclonaias necki (MANOVA, pairwise Hotelling's test p = 0.138).
The proportion of C. pustulosa specimens correctly identified to the original nominal species within the C. pustulosa complex exceeded that of Cyclonaias specimens correctly identified to species level (see above), with 80% correct  identifications ( Figure 6b). All nominal species differed significantly from each other in their shell shape as approximated by 18 Fourier coefficients (MANOVA, pairwise Hotelling's tests p < 0.05). Using only the nominal species C. kieneriana and Cyclonaias asperata, the Fourier coefficients differed significantly between C. kieneriana and its synonym C. asperata (MANOVA: F 18,82 = 2.094, p = 0.013), and 95% of specimens were classified correctly based on shell shape through LDA (Figure 6c).

| Quadrula
Fourier coefficients differed significantly between the nominal species of Quadrula (MANOVA, pairwise Hotelling's tests p < 0.05; Figure 6c). Seventy-six per cent of specimens were assigned to the correct nominal species, with 21% and 11% of misidentifications between Q. apiculata versus Q. quadrula and Q. rumphiana, respectively.

| Theliderma
Within the genus Theliderma, 91% of specimens were identified to the correct species (as they are here recognized) by LDA of Fourier coefficients (Figure 6e). Theliderma cylindrica, characterized by its typical elongated-rectangular shape, was 100% correctly identified. Considerable difficulties in separation by shell shape were present for Theliderma sparsa versus T. johnsoni n. sp. (21% misidentifications) and T. metanevra (13%), respectively. Most true Theliderma species pairs differed significantly from each other in their shell shape with the exception of T. sparsa versus T. johnsoni n. sp. (MANOVA, pairwise Hotelling's test: p = 0.525), T. sparsa versus T. metanevra (p = 0.227) and T. stapes and T. johnsoni n. sp. (p = 0.427; p-value could not be computed for the pair T. sparsa vs. T. stapes due to low replicate number).
When including the whole Theliderma dataset in LDA, only 5% of specimens of the pair T. metanevra/T. johnsoni n. sp. were assigned to the wrong clade (Figure 6e). When using only the T. metanevra dataset, 11% of specimens were misidentified (Figure 6f), though Fourier coefficients were significantly different between the two species (MANOVA: F 18,46 = 3.097, p = 0.001).

| Diagnostic characters of the classical genera within Quadrula s.l
Species within Quadrula and Tritogonia share a number of ecological and morphological traits but are distinct from those within Cyclonaias and Theliderma (Table 2; Supporting Information Table S5). Quadrula and Tritogonia species exhibit a marked sulcus that is absent in Cyclonaias and Theliderma, with the exception of T. sparsa and T. stapes that may display shallow sulci (Table 2; Supporting Information Table S5). Quadrula and Tritogonia glochidial size index is ten times smaller than in Cyclonaias and Theliderma (Table 2; Supporting  Information Table S5). Quadrula and Tritogonia also seem to share similar morphological and behavioural patterns of the mantle displays, also known as mantle magazines. While Quadrula and Tritogonia seem to exhibit large mantle displays with a non-reflexive glochidia release strategy when disturbed, Cyclonaias and Theliderma mantle displays are small and more inconspicuous and immediatelly expell their glochidial content when physically disturbed ( Table 2; Supporting Information Table S5). However, some caution has to be taken when interpreting this data since mantle displays were only studied in a small number of species.
Within Quadrula s.l. some of the analysed characters are exclusive and can be used to recognize some of the classical recognized genera Cyclonaias, Quadrula, Theliderma and Tritogonia (Table 2; Supporting Information Table S5).
The presence of dark chevrons imprinted in the periostracum of shells is a trait that is exclusive of Theliderma species and can be used to recognize the genus within Quadrulini (Table 2; Supporting Information Table S5).
The stomate-shaped morphology of the mantle displays seems to be a diagnostic character for Cyclonaias, but laboratory studies on C. asperata (= C. kieneriana) did not observe any mantle display for this species (Haag & Staton, 2003).
Theliderma hosts are mainly composed of small cyprinids while catfishes are the main hosts of the other three Quadrula s.l. genera (Table 2; Supporting Information  Table S5). The mantle displays and glochidia of Theliderma are smaller than those of Cyclonaias (Table 2; Supporting  Information Table S5).
Tritogonia verrucosa and T. nobilis are sexually dimorphic in shell shape, a trait that is unique within the Quadrulini and therefore diagnostic of the genus (Table 2; Supporting  Information Table S5). In addition, the mantle display mechanism of T. verrucosa, which involves the mantle to completely cover both the incurrent and excurrent aperture, is very distinct from those of all of the other Quadrula s.l. species (Supporting Information Table S5). However, this trait needs to be verified for T. nobilis in order to be considered diagnostic of the genus.

Quadrula and generic support
The three BI and ML phylogenies (COI, ND1, and COI + ND1) obtained in the present study revealed a well-supported Quadrula sensu lato clade subdivided into four clades (mainly in the BI analyses), corresponding to the genera Quadrula, Cyclonaias, Theliderma, and Tritogonia (Figures 2-4; Table 3). Furthermore, taxa in these clades exhibit coherent combinations of traits that in our opinion support the validity of their generic status as recently recognized by Williams et al. (2017) (Figures 2-4; Tables 3 and 5, Supporting Information Table S5).
The current molecular phylogenies cannot strongly support any suprageneric relationships (probably due to insufficient genetic marker representation) within Quadrula s.l. However, the morphological and ecological data here presented suggest common evolutionary origins for the genera Quadrula and Tritogonia, and for Cyclonaias and Theliderma (Table 2; Supporting Information Table S5). While Quadrula and Tritogonia include large reflexive mantle displays, miniaturized shell glochidia, and marked shell sulci, Cyclonaias and Theliderma species have small non-reflexive mantle displays, larger glochidia, and absent or shallow shell sulci (Table 2; Supporting Information Table  S5).
The series of traits shared by Quadrula and Tritogonia are likely associated with maximizing attachment success to their main hosts, the catfishes (Table 2). These traits include large conspicuous mantle displays that do not respond to mechanical disturbance (but probably to another type of stimulus, for example, chemical, that might capitalize on the acute olfactory sense of their hosts) and miniaturized glochidia. Tritogonia species are the only Quadrula s.l. species that exhibit marked shell sexual dimorphism. This is probably a result of the presence of mantle displays that completely cover the incurrent and excurrent apertures T A B L E 6 Pairwise genetic distance matrixes of quadruline species of the genera Cyclonaias, Quadrula, Theliderma, and Tritogonia, as recognized in the present study  of females, resulting in a distortion of their shells ( Table  2, Supporting Information Table S5). On the other hand, a specialization in attracting small cyprinids and percids seems to have driven reproductive behaviour and morphology in Theliderma towards females that are generally completely buried with only the mantle display being visible (Sietman, Davis, & Hove, 2012). The displays of Theliderma are also more conspicuously displayed during the day favouring the visual predatory habits of cyprinids, which is in contrast to the other three Quadrula s.l. genera who are generally displaying at night when feeding activity in catfishes is highest (Hove et al. 2011). Theliderma species are unique within Quadrulini in the production of mucoid conglutinates (Haag, 2012) and by presenting dark chevrons in the shells periostracum (Table 2; Supporting  Information Table S5). The glochidia of Theliderma are also much bigger than those of Tritogonia and Quadrula and more similar in size to most of the other species within the Ambleminae (Table 2; Barnhart, Haag, & Roston, 2008). The large size of Theliderma glochidia can be related to the much lower fecundity of this genus when compared with the other Quadrula s.l. genera (Haag, 2012).
Cyclonaias presents a set of reproductive features that are similar to those in Theliderma species. However, glochidial size in Cyclonaias is always larger than in Theliderma, and Cyclonaias exhibit a prevalence to catfish hosts rather than cyprinids and percids (Table 2). Adaptation to catfish hosts again is likely associated with the unique stomate-shaped mantle displays exhibited by Cyclonaias species (Table  2). The miniaturized glochidia shared by Quadrula and Tritogonia probably represent a derivation from the more primitive glochidial size of most amblemines (Barnhart et al., 2008). On the other hand, preference for and related adaptations to catfish hosts seem to be ancestral for the Quadrulini, while preference for small cyprinids and percids in Theliderma is probably the derived state. A more robust multi-marker molecular approach is needed in order to get a clearer view on the evolutionary aspects of these interesting adaptations and to resolve the suprageneric relationships among Quadrula s.l. genera.

| Phylogeny and systematics implications
within the four Quadrula sensu lato genera

| Cyclonaias
The present results confirm the results of a recent study on this genus (Johnson et al., 2018) recognizing nine of the 14 Cyclonaias species listed by Williams et al. (2017) as valid species (Table 7). However, we here consider C. asperata as a synonym of C. kieneriana due to the residual genetic divergence between these two taxa (ND1 p-distance <1%) and the fact that C. kieneriana (Lea, 1852) has priority over C. asperata (Lea, 1861). In contrast, Williams et al. (2017) recognized both species based on their morphological distinctiveness and the fact that molecular evidence for synonymy was based on only one marker (ND1) from a single specimen. However, ND1 has been shown to be a highly representative marker of overall mtDNA evolution in unionoid mussels (Fonseca, Lopes-Lima, Eackles, King, & Froufe, 2016). In addition, divergence between C. asperata and C. kieneriana sequences was very low. As a result, both ND1 (BI and ML) analyses were unable to resolve both species' phylogenies, and all ND1 species delineation methods were unable to separate the two species (Table 5), indicating that C. asperata should be synonymized under C. kieneriana. The morphometry results supported the distinct morphology of the two nominal species but very few C. kieneriana shells (n = 4) were available, preventing a comprehensive analysis (Figure 6c). Although C. asperata has been reported from a much wider geographic range than C. kieneriana, both species are sympatric in the whole range of C. kieneriana suggesting that specimens previously described as C. kieneriana are particular smooth forms of the same species (Figure 7). Until recently, Cyclonaias archeri has been considered a subspecies of C. asperata (Turgeon et al., 1998). However, since no sequences, tissues or shell specimens of C. archeri were available for this study, we rely on Williams et al. (2008Williams et al. ( , 2017 and recognize this species as separate from C. asperata, based on its distinct morphology. Cyclonaias necki has recently been separated from Cyclonaias petrina based on molecular data (COI) and morphology (Burlakova, Karatayev, Froufe, Bogan, & Lopes-Lima, 2018;Johnson et al., 2018). The specific rank of C. necki is here confirmed by all species delineation methods used on each of the datasets ( Table 5). The shell shape is also significantly different between C. petrina and C. necki (Figure 6a), confirming observations of Burlakova et al. (2018) and Johnson et al. (2018) that C. necki shells are thinner, more compressed and more rectangular in shape with a more distinct and prominent posterior ridge. Distribution ranges of the two species are exclusive, with C. necki being present only in the San Antonio/Guadalupe River basins, while C. petrina is restricted to the Colorado basin ( Figure 8; Burlakova et al., 2018).
The present paper confirms the inclusion of four nominal species, that is, C. aurea, C. houstonensis, C. mortoni, C. refulgens, within C. pustulosa (Table 7) and Cyclonaias succissa, as a related but distinct species, as proposed by Johnson et al. (2018). None of the phylogenies resolved them as monophyletic, and p-distance values among these taxa were very low (Table 4). All nominal species here synonymized with C. pustulosa have distinct and exclusive geographic distributions ( Figure 9). The molecular results suggest that C. pustulosa is divided into several morphotypes each in a distinct geographic area. These morphotypes are clearly visible in the morphometry results and explain why these populations used to be considered distinct species (Figure 7b).
The remaining Cyclonaias species recognized in the present study, that is, C. infucata, C. kleiniana, C. kieneriana, Cyclonaias nodulata and Cyclonaias tuberculata, were always retrieved as well supported, divergent clades ( Figures  2-4), and recognized by all species delineation methods (Table 5). Furthermore, the shell shape is different among all of these latter species, except for the pair C. infucata and C. kleiniana, which might be explained by their closer genetic relationship (Figures 2-4; Table 6).

| Quadrula
In the absence of genetic data and shell materials for Quadrula couchiana and Quadrula fragosa, the first being most likely extinct (Williams et al., 2017) and the second on the verge of extinction (Sietman, 2003), we make no considerations about their systematics and accept both as valid within the Quadrula genus following Williams et al. (2017).   We here synonymize Q. apiculata and Q. rumphiana under Q. quadrula. Although only a small number of sequences were available for Q. apiculata and Q. rumphiana, the level of divergence among these three nominal species is low for both markers (Table 4). Furthermore, in all phylogenies, Q. quadrula is paraphyletic, with Q. apiculata and Q. rumphiana falling inside the clade (Figures 2-4). The level of divergence between these three nominal taxa is actually lower than the divergence between the distinct clades of COI within Q. quadrula sensu stricto identified by Mathias, Hoffman, Wilson, and Zanatta (2018) and also retrieved here in the COI phylogeny and haplotype network (Figures 2 and 6a). A specific rank for each of these divergent clades was rejected in that study due to the existence of gene-flow among them as shown by their microsatellite dataset (Mathias et al., 2018). The nominal species Q. apiculata, Q. rumphiana and Q. quadrula sensu stricto presented distinct shell shapes but only 76% of specimens were assigned to the correct nominal species ( Figure  6d). The slightly distinct shell morphology again suggests that distinct nominal species were assigned to regional forms despite the relative overlap in distribution range of Q. apiculata with both Q. quadrula and Q. rumphiana ( Figure 10) that may also be related to the considerable overlap among shell shape forms (Figure 6d).

| Theliderma
Only two shells and no genetic material were available for T. stapes, since the species is very endangered and possibly extinct (NatureServe, 2018). Until new evidence emerges, we therefore accept it as valid within the Theliderma genus following Williams et al. (2017). Based on the molecular phylogenies and all species delineation methods, we recognize five additional species within Theliderma, that is, T. cylindrica, Theliderma intermedia, T. metanevra, T. johnsoni n. sp. and T. sparsa (Figures 2-4; Tables 1 and 5). The nominal species T. metanevra is here divided in two distinct species, the T. metanevra sensu stricto with a Mississippi basin distribution and T. johnsoni n. sp. distributed within the Mobile basin ( Figure  11). The two species show high genetic divergence (3.2% for COI and 3.5% for ND1; Table 6). They also differ morphologically, presenting distinct shell shape with only 5%-11% of specimens being misidentified by Fourier analysis (Figure  6e,f) as well as other morphological features (see Supporting Information Appendix S2).

| Tritogonia
The position of T. nobilis could not be resolved in a previous single marker approach (Serb et al., 2003) but in the F I G U R E 1 1 Distribution maps of (a) Theliderma metanevra before the present study and (b) after the present study divided in T. metanevra and Theliderma johnsoni n. sp [Colour figure can be viewed at wileyonlinelibrary.com] present study, all phylogenies reveal a well-supported clade comprising T. nobilis and T. verrucosa. We therefore move the nominal species Q. nobilis into Tritogonia as T. nobilis. Until the end of the 20th century, T. nobilis was not recognized by most authors as a separate species from Q. quadrula (Williams et al., 2008). However, its placement under Tritogonia is not new as Simpson (1914) already used this designation. Both T. nobilis and T. verrucosa exhibit marked sexual dimorphism (Simpson, 1914;Williams et al., 2008), which is a synapomorphy of the genera within the Quadrulini.

| Cyclonaias
As C. asperata is here synonymized under C. kieneriana, future conservation status assessment of C. kieneriana should include the distribution of C. asperata sensu stricto ( Figure  7), which would be expected to decrease the extinction risk of the species under the currently recognized systematics. The separation of C. necki from C. petrina will likely increase the extinction risk of both species as their distributions are even smaller than previously believed (Figure 8) but see Johnson et al. (2018) for detailed conservation implications. In contrast, the secure conservation status of C. pustulosa (Supporting Information Table S6) is here strengthened by the inclusion of the nominal taxa C. aurea, C. houstonensis, C. mortoni and C. refulgens (Figure 9; Table 7). However, due to their genetic uniqueness, the populations from Eastern Texas (originally identified as C. mortoni) should be managed independently.

| Quadrula
Synonymization of the nominal species Q. rumphiana and Q. apiculata under Q. quadrula does not affect the conservation status of Q. quadrula due the wide distribution areas and low extinction risk of the three forms. That said, subtler potential genetic differences between populations originally assigned to these species are likely to be revealed in future studies applying faster evolving markers.

| Theliderma
The conservation status of T. metanevra is currently considered as secure mainly based on the species' wide distribution range. However, considering that the Mobile basin populations in fact represent a separate species (Figure 11), T. johnsoni n. sp., the conservation statuses of T. metanevra and T. johnsoni n. sp. need to be re-assessed separately, and the two species need to be managed independently.