Although challenging, the effort to understand species boundaries and biodiversity remains one of biology's most fundamental and urgent tasks (Soltis and Gitzendanner 1998; Pettengill and Neel 2011; Paul et al. 2013; Fennessy et al. 2016); how species are delineated impacts the allocation of limited conservation funds, designation of official protection status, the selection of individuals used for conservation programs, and the general public's perception of systematics and conservation. The evolution of cryptic species that are unequivocally genetically distinct and even reproductively isolated, but are morphologically overlapping or nearly indistinguishable from other species, represent a major challenge in addressing these issues (e.g. Bickford et al. 2007; Adams et al. 2014; Shirley et al. 2014; Grbic et al. 2015). The wider occurrence of species complexes (assemblages of closely related often cryptic species that defy easy delimitation) also present a perplexing but important window into the process of speciation and diversification, adaptation, gene flow, and reticulate evolution. Achieving a balance between recognition of infraspecific phenotypic variability, identification of cryptic species, taxonomic over-splitting, and conservation is a contentious and fiercely debated topic, particularly in charismatic groups such as orchids, primates, and birds (e.g. Pillon and Chase 2007; Dueck and Cameron 2008b; Hopper 2009; Sangster 2009; Markolf et al. 2011; Zachos et al. 2013; Swarts et al. 2014). This complicated set of issues is further obfuscated by the lack of a universally accepted species concept. The task faced by systematic and conservation biologists is how to best reflect evolutionary pattern and process while simultaneously recognizing diagnosable species, developing useful classifications, and protecting the greatest amount of biodiversity. This challenge is intensified when dealing with putative species that present large degrees of phenotypic variation and plasticity.

The genus Spiranthes Rich. (Orchidaceae) has long presented a significant taxonomic and identification challenge: many of its currently accepted 35 species exhibit marked phenotypic variation at both local and continental scales, and hybridization has traditionally been thought to be pervasive. Furthermore, cryptic speciation has been documented in the group (Brown et al. 2008; Dueck and Cameron 2008a; Pace and Cameron 2016, Pace et al. 2017), a few species exhibit polyploidy and apomixis (Catling 1982; Sheviak 1982), the species status of some taxa is questionable (Walters 2005; Dueck and Cameron 2008b), incomplete lineage sorting is likely a reality in the group (Dueck et al. 2014), and taxonomic inflation by certain authors is a troublesome issue (e.g. Brown 1999a, 1999b). The group that best encompasses all of these issues is the Spiranthes cernua (L.) Rich. species complex (Figs. 1–4): a true Gordian Knot frequently described as “intractable” (Sheviak 1982, 1991; Sheviak and Brown 2002).

Taxonomic History —Linnaeus' original description of Ophrys [Spiranthes] cernua L. is minimalistic (e.g. lacking details such as measurements) and the type location is imprecise: “Virginia, Canada” (Linnaeus 1763). The lectotype (Kalm s.n. (LINN) and drawings of lectotype flowers by A. Gray in Sheviak and Catling 1980, p. 531) displays two plants 25.5–36.1 cm tall with linear to linear-lanceolate leaves and strongly nodding flowers 5.5–7.8 mm long in profile. The labellum is bluntly acute (as denoted in the original description) and slightly dilated near the apex, with two prominent basal callosities (nectar glands). As indicated by 18th and early 19th century collections, the name S. cernua was broadly applied to any autumnal flowering Spiranthes in eastern North America (e.g. Muhlenberg 180 (LINN) represents what are now recognized as five distinct species).

The species S. casei Catling & Cruise, S. magnicamporum Sheviak, S. ochroleuca (Rydb.) Rydb., S. odorata (Nutt.) Lindl., and S. parksii Correll (Figs. 3, 4), were proposed to accommodate much of the morphological variation expressed by the autumn-flowering Spiranthes. All of these autumnal North American Spiranthes traditionally have been thought to be very closely related to one another and to S. cernua, and most have been referred to, or placed under synonymy with S. cernua. This group came to be known collectively as the Spiranthes cernua species complex sensu Sheviak (Sheviak 1982, 1991; Sheviak and Brown 2002). Even with this added knowledge of species-level diversity among the autumn-flowering Spiranthes, S. cernua sensu traditum encapsulates a wide degree of morphological and reproductive variability fromwithin-population to biome level scales. The floral and reproductive phenomena of peloria, cleistogamy, and apomixis are common within S. cernua, contributing to this variability (Sheviak 1982, 1991). Some peloric individuals of S. cernua, particularly in the central and southern Great Plains, are very robust in stature, approaching S. magnicamporum in overall appearance. In some regions, only cleistogamous individuals may be found in particular populations; occasionally peloric and non-peloric, cleistogamous and open, and apomictic and sexual individuals may all be found within a single population or populations separated by only a fewkilometers.Based on thework of Catling and Sheviak (Catling 1982, 1983; Catling and Brown 1983; Sheviak 1982, 1991; Sheviak and Brown 2002), S. cernua is traditionally understood to be a facultatively agamospermic polyploid ‘compilospecies’ (2n = 45, 60; sensu Harlan and de Wet 1963) that freely hybridizes with all other species in the S. cernua species complex. Sheviak (1973) went so far as to state “a ‘typical’ S. cernua is difficult to define because of the species' intrinsic hybrid nature.” Building upon this hypothesis, Homoya (1993) wrote that hybridization has led to at least five habitat-specific “genetic races” (i.e. ecological morphs) occurring within Indiana alone. Despite its hypothesized pervasiveness, the morphological identity of potential hybrids is unclear, and no primary hybrids involving S. cernua have been formally named; the dubious S. ×borealis P.M. Br. (S. casei × S. ochroleuca) is the only named hybrid in the complex. Hybridization has come to be hypothesized as the only, or at least major, source of morphological variability within this complex, with little attention paid to other potential evolutionary forces that might favor the evolution of widespread phenotypic heterogeneity or the possibility of cryptic speciation in combination with or in the absence of hybridization.

Fig.1.

Spiranthes cernua s. s. A. New Castle Co., Delaware, sample sc1. B. Worchester Co.,Maryland, sample sc8d. C.Wayne Co., Georgia, sample sc15. D–E. Sussex Co., Delaware. F. Cape May Co., New Jersey, sample sc6. G. Scott Co., Arkansas, sample sc29; flowers reduced (sub-peloric) and partially closed. H. Towns Co., Georgia; photo: James Fowler, used with permission. I. James City Co., Virginia, sample sc9. J. Lexington Co., South Carolina. K. Pike Co., Alabama. L. Co-occurring S. ×kapnosperia (L) and S. cernua (R), Jackson Co., North Carolina; scale bars denote cm.

Fig.2.

Spiranthes cernua s. s. A. Habit. B. Inflorescence. C. Flower and floral bract in profile.D–G. Flower, dissected view.D.Dorsal sepal. E. Dorsal petal. F. Lateral sepal. G. Labellum. H. Labellum, column, and ovary in natural position. I–K. Column. I. Dorsal and ventral views. J. Profile. K. Anther. L. Pollinia. Drawn from Pace 612 and Pace 615 by Bobbi Angell.

Fig.3.

A. Spiranthes casei, Michigan; photo: Aaron Strouse, used with permission. B–C. Spiranthes ochroleuca, Rutland Co., Vermont. C. Characteristic ochroleucous-yellow coloration on abaxial labellumsurface. D. Spiranthes ovalis var. erostellata Catling,Waukesha Co., Wisconsin, sample so2. E. Spiranthes niklasii, Yell Co., Arkansas (type location). F. Spiranthes ×kapnosperia, Transylvania Co., North Carolina (type location).

Fig.4.

A. Spiranthes magnicamporum, Guadalupe Co.,New Mexico, sample sm7h. B. Spiranthes triloba, Polk Co., Florida. C. Spiranthes igniorchis, Polk Co., Florida; photo: Dave Briley, used with permission. D. Spiranthes longilabris, Brunswick Co., North Carolina, sample location of 13b.

Repeated attempts have been made to clarify species relationships and species status within the Spiranthes cernua complex, with different methods favored by various authors including breeding systems (Catling 1981), morphological discriminant analysis (Sheviak and Catling 1980; Catling 1981; Sheviak 1982), and cytology (Sheviak 1982, 1991). Most recently, DNA sequences (Dueck and Cameron 2007, 2008b; Dueck et al. 2014; Pace and Cameron 2016; Pace et al. 2017) and AFLPs and microsatellites (Walters 2005, Manhart and Pepper 2007) have been utilized to understand the systematics of these orchids. We suggest that viewing morphological variation through the integrated context of molecular phylogenetic relationships and morphometric analysis offers an attractive and promising path forward.

Phylogenetic Background —Recent molecular phylogenetic research investigating taxonomic confusion in portions of the S. cernua species complex (Dueck et al. 2014, Pace and Cameron 2016, Pace et al. 2017) reveal that circumscription of the Spiranthes cernua complex sensu Sheviak (1982; Sheviak and Brown 2002) does not properly reflect evolutionary history. Specifically, S. odorata does not belong in this grouping since it is distantly related to S. cernua, instead being sister to the clade containing most of the Eastern North American species. Additionally, S. triloba (Small) Schum., a species lost under synonymy of both S. cernua and S. odorata for more than a century, is a distinct species, forming a sister relationship with S. magnicamporum. Together, Dueck et al. (2014), Pace and Cameron (2016), and Pace et al. (2017) demonstrate that the Spiranthes cernua complex sensu nov. includes S. casei, S. cernua, S. magnicamporum, S. parksii, and S. ochroleuca (the traditionally included species), plus S. triloba, in addition to the unexpected S. longilabris Lindl., S. igniorchis M.C. Pace, and S. ovalis Lindl. (Figs. 3, 4). Furthermore, Dueck et al. (2014) found that S. cernua itself is likely polyphyletic, and S. casei is broadly embedded within S. ochroleuca. Although support values for individual taxon clades were typically high (PP > 0.90; often 1), the exact relationship among clades was not well resolved, particularly within the Spiranthes cernua complex s. s.: the S. cernua clades, S. parksii, and S. ochroleuca+ S. casei. Additionally, the morphological limits of these potential clades were not explored. This occasional deficiency of resolution has been used to advance opposing positions as exemplified by the debate over the species status of federally endangered S. parksii: Dueck and Cameron (2008b) hypothesized that its sub-peloric morphology, coupled with a lack of molecular resolution between S. cernua and S. parksii (and the fact that recognition of the latter would render the former paraphyletic), indicates the two are the same species, whereas Jacobsen et al. (2009) strongly rejected Dueck and Cameron (2008b), attributing the unresolved relationship to the neutral nature of the sampled DNA markers.

To resolve the long-standing problems of species relationships and species status within this complex, we present an expanded sampling of taxa and loci, with integrated morphological and biogeographic data. Working under an integrated history-bound phylogenetic species concept (Baum and Donoghue 1995; Dayrat 2005) in which monophyly is emphasized in concert with supporting morphological and ecological data, these data support the division of ‘S. cernua’ into five distinct but morphologically cryptic and variable species and hybrids.

Methods

Taxonomic Sampling —Approximately 700 herbarium specimens of species in the Spiranthes cernua species complex were carefully reviewed from AMES, BH, CLEM, CM, F, FLAS, FSU, LSU, MO, NY, NYS, SEL, UARK, US, USF,WILLI,WIS, and WVA(herbarium acronyms from Thiers 2017; Appendix 1). Individual flowers from the lowermost quarter of the inflorescence were rehydrated for morphological examination from select individual specimens. Fieldwork based on georeferenced herbarium specimens was conducted by M. Pace in Alabama, Arkansas, Delaware, Georgia, Florida, Illinois, Indiana, Maryland, New Jersey, New Mexico, New York, North Carolina, Ohio, Oklahoma, Pennsylvania, South Carolina, Texas, Virginia, Vermont, and Wisconsin from 2012–2016. Sampleswere collected for herbariumand spirit vouchers, morphological measurements, and DNA sequencing. For areas we were unable to visit for fieldwork, 1–10 yr old herbarium specimens were judiciously sampled with permission for inclusion in the molecular phylogenetic analysis

Morphometrics —To analyze and understand the morphological variability and differentiation of species, boxplots were created for selected species. These plots were visualized in R Studio (R Development Core Team 2012) using rehydrated labella from herbarium samples, scaled drawings in published literature (Sheviak 1982, 1991), and whole herbarium specimens. Measurements of rehydrated labella included labellum length, width below callosities, width at constriction, width of median point below constriction, and callosity height. Measurements of whole herbarium specimens include height of the plant, leaf length, leaf width at widest point, lateral sepal length, lateral sepal width at midpoint, and labellum length.

Table 1.

Summary table of molecular regions used in the phylogenetic analysis. Two nuclear: ACO and nrITS, and five chloroplast: matK, ndhJ, trnL-F, trnS-M, and ycf1 3′.

Molecular and Phylogenetic Methods —Phylogenetic analyses incorporated and expanded upon the dataset of Dueck et al. (2014), Pace and Cameron (2016), and Pace et al. (2017). For new accessions, 3–4 unopened buds or ca. 1 cm² of leaf tissuewere collected and silica-gel dried for later extraction of Total gDNA. IBI plant isolate kits (Peosta, Iowa) and Maxwell® 16 LEV plant DNA kits (Madison, Wisconsin) were used for all newly collected samples. All accessions were amplified for the chloroplast gene regions matK, ndhJ, trnL intron, trnS-fM, andycf1 3′, nuclear ribosomal ITS (internal transcribed spacers 1 and 2, and the 5.8S subunit; “nrITS”), and the low-copy nuclear region ACO (Appendix 2). PCR amplification protocols used were as follows: chloroplast (except ycf1) and nrITS: following Dueck et al. (2014); ACO: following Guo et al. (2012); ycf1: following Neubig et al. (2009). The PCR products were purified using ExoSap-It (Cleveland, Ohio), and cycle sequencing products were cleaned using Agencourt CleanSeq (Beverly, Massacusetts) magnetic beads. Direct sequencing of cleaned cycle sequencing products was performed at the University of Wisconsin — Madison Biotechnology Center. Resulting chromatogramswere edited and aligned using softwaremodules available in Geneious 7.1 including MUSCLE. Ambiguities in datasets were coded with standard IUPAC-IUB symbols for nucleotide nomenclature (Cornish-Bowden 1985). If samples failed to amplify after repeated attempts for a given locus they were coded as missing data (Table 1). Based on previous research (Dueck et al. 2014, Pace and Cameron 2016), Spiranthes odorata was used as the outgroup.

The data were analyzed as: 1) individual loci, 2) combined chloroplast data, 3) combined nuclear data, 4) combined nuclear + chloroplast data, and 5) combined nuclear + chloroplast data using a reduced dataset composed of samples with no missing data and no potential hybrid individuals. Phylogenetic analyses were performed under Bayesian Inference (MrBayes on XSEDE (3.1.2)) implemented through CIPRES Portal v. 3.3 (Miller et al. 2010). Based on Dueck et al. (2014), the GTR+G (generaltime-reversible with a gamma distribution) modelwas implemented for all datasets and partitions. Analyses were run for 10,000,000 generations, with a sample frequency of 100,000, nruns = 2, nchains = 6, temp = 0.2, and a burn-in of 500,000. Phylogenetic inference of the 50% majority-rule consensus tree was constructed using the “sumt” option based on the remaining trees. The topologies of these trees were visualized in FigTree (Rambaut 2014). The combined nuclear + chloroplast data was visualized as a network in the program Splits Tree Window (Huson and Bryant 2006). To clarify discordance between individual gene trees resulting from possible incomplete lineage sorting, we also took a multispecies coalescent approach, estimating the species tree using *BEAST (Drummond et al. 2012) under a birth-death process. The species tree was visualized in DensiTree v. 2.2 (Bouckaert 2010).

Results

Phylogenetics —Individual gene trees generally found S. cernua to be polyphyletic (trees not shown). Only the reconstruction of nrITS failed to support the polyphyly of S. cernua, and it failed to resolve the relationships between S. ochroleuca and S. cernua s. l., whereas ycf1 recovered weak to moderate support for the polyphyly of S. cernua s. s. within a broader polytomy of S. ochroleuca and S. casei. The phylogenetic hypothesis based on combined nuclear data mostly lacks resolution along the backbone of the phylogeny among otherwise moderately to fully supported clades of individual species, whereas the combined chloroplast hypothesis provides more well-supported resolution among strongly to fully supported clades of species (Fig. 5). These reconstructions, however, find S. cernua as it has traditionally been circumscribed to be strongly polyphyletic (with these polyphyletic a priori S. cernua samples now labeled as S. arcisepala, S. incurva, and S. niklasii in the trees). Discordance between chloroplast and nuclear topologies was observed, with the nuclear tree recovering S. arcisepala in a strongly supported (PP =0.98) polytomy with S. cernua s. s. and S. incurva, and the chloroplast tree fully supporting (PP =1) a sister relationship to S. ochroleuca + S. casei in (Fig. 5). When samples with missing data and presumed hybrids were excluded, the sister relationship between S. arcisepala and S. ochroleuca + S. casei is preserved, although it is weakly supported (PP=0.64) (Fig. 6). Discordance between nuclear and chloroplast trees was also observed in the relationships of S. magnicamporum and S. triloba, and S. ovalis in relation to the S. cernua species complex s. s. (Fig. 5). Additional instances of discordance that indicate hybridization are discussed below.

In nearly all of our phylogenetic hypotheses, S. parksii is consistently recovered embedded within the S. cernua s. s. clade in an unresolved relationship. The only phylogenetic reconstruction to find any resolution between S. parksii and S. cernua s. s. is the reduced sampling combined nuclear + chloroplast tree (Fig. 6), which only found weak support (PP= 0.60) for a sister relationship between a clade of S. cernua s. s. as distinct from a fully supported S. parksii clade.

Evidence for, and Instances of, Hybridization—Based on the samples and molecular regions included here, we recovered molecular evidence for geographically specific hybrid speciation. Hybridization was inferred by samples that display discordant topologies within otherwise strongly supported clades, display strong connections to other species in the network analysis, and have ambiguities in nuclear chromatograms in regions that correspond to nucleotide differences between inferred parental species (Figs. 5, 7, and 8). Cases of probable hybridization are seen in three instances: 1) samples from across the northern Interior Lowlands, Great Lakes Basin, and western and northern Appalachian Highlands which display discordance between S. cernua s. s. and S. magnicamporum, and are described below as Spiranthes incurva comb. nov.; 2) samples from the Ouachita Mountains, Boston Mountains, and Crowley's Ridge in Arkansas displaying a ridge of papillae along the labellum midvein, which switch clades between S. cernua s. s. (nuclear) and a sister clade to S. ovalis (chloroplast), described below as Spiranthes niklasii sp. nov.; and 3) samples from the southern Smoky Mountains that switch clades between S. cernua s. s. (nuclear) and S. ochroleuca + S. casei (chloroplast), and are described below as Spiranthes ×kapnosperia nothsp. nov.

Fig.5.

Bayesian phylogenetic tree reconstructions. Posterior probabilities within species clades are only indicated if they are > 0.90. The discordant positions of the three hybrid taxa are highlighted; colors the same as in Fig. 7. A. Bayesian phylogenetic tree reconstruction based on the combined five-locus chloroplast data. B. Bayesian phylogenetic tree reconstruction based on the combined ACO + ITS nuclear data.

Fig.6.

A. Bayesian phylogenetic tree reconstruction based on the reduced sampling nuclear+chloroplast dataset after removing hybrid individuals and individuals with missing data. Posterior probabilities within species clades are only indicated if they are > 0.90. B. Coalescent species tree based on the reduced sampling nuclear +chloroplast dataset after removing hybrid individuals and individuals with missing data. All trees are shown. The root canal is highlighted in dark blue and the consensus trees are highlighted in faded colors.

When we observed our nuclear data for samples of S. incurva, we found multiple base pair ambiguities in both loci; these sites of nucleotide ambiguity essentially corresponded perfectly to sites of molecular differentiation between S. cernua s. s. and S. magnicamporum (Fig. 8; e.g. an A in S. cernua, a G in S. magnicamporum, and an R — representing either A or G — in S. incurva). In nearly all cases, examination of the nuclear chromatographs found essentially identical overlapping peaks of either possible nucleotide. For S. niklasii, all chloroplast loci examined had nucleotide mutations that were unique to this species in addition to those shared with its closest relative, S. ovalis. These plants only displayed nucleotide ambiguities in the nrITS dataset (corresponding to nucleotide differences between S. cernua and S. ovalis), whereas in the ACO data set some regions shared nucleotide differences with S. cernua, and in other segments of the same locus they shared nucleotide differences with S. ovalis (Fig. 8). Individuals of S. ×kapnosperia (identified a priori as S. ochroleuca) clustered with S. ochroleuca in the chloroplast datasets, but shared several nucleotide changes with S. cernua, and clustered with S. cernua in nuclear analyses. The incongruence and connection to other species of these hybrid species is also depicted in our network analysis (Fig. 7).

Fig.7.

Phylogenetic network from Neighbor Net analysis of the S. cernua species complex s. l. plus S. odorata, using the nuclear + chloroplast dataset, including all samples of both hybrid and non-hybrid taxa. The position of species is indicated by colored ovals.

Species Tree—Our species tree reconstruction (using the reduced combined nuclear+chloroplast dataset after removing potential hybrid individuals and individual samples with missing data) found that S. arcisepala ismost closely related and sister to a clade of S. ochroleuca + S. casei, with this grouping in turn sister to S. cernua+the former S. parksii (Fig. 6). Overall, the species tree reconstruction recovered sister relationships between S. igniorchis and S. longilabris, S. magnicamporum and S. triloba, S. ochroleuca + S. casei and S. arcisepala, and hypothesizes a gradation of relationships between species of the S. cernua species complex s. l.: ((S. igniorchis, S. longilabris), (S. ovalis, (S. magnicamporum, S. triloba)), (S. cernua + S. parksii, (S. arcisepala, S. ochroleuca + S. casei))). Although this topology is broadly congruent with our reduced combined nuclear + chloroplast dataset gene tree, there are differences in the relationship of S. ovalis and the sister pair of S. magnicamporum and S. triloba. Some uncertainty in the reconstruction of overall relationships along the backbone, particularly in the topological placement of S. arcisepala and S. ovalis, is also apparent in the visualization of the background consensus trees.

Fig.8.

Examples of nucleotide alignments highlighting regions supporting hybridization between species. A. ITS. Spiranthes incurva highlighted in purple, with nucleotide ambiguities corresponding to differences between S. cernua s. s. (upper) and S. magnicamporum (lower). B. ACO. Spiranthes niklasii highlighted in gray, displaying shared nucleotide changes with both S. ovalis (upper, e.g. position 556) and S. cernua s. s. (lower, e.g. position 596).

Morphometrics —Our boxplots of the labella of S. arcisepala, S. cernua, S. incurva, S. ×kapnosperia, S. magnicamporum, S. niklasii, and S. ochroleuca yield largely overlapping plots (Fig. 9), thus highlighting the cryptic nature of these species and the confusion many botanists encounter when endeavoring to identify them. Based on the sampling we included, S. arcisepala is the least variable species, however the measured features entirely overlap the space and variability of S. cernua s. s. Similarly, the variation of S. ochroleuca is also contained within S. cernua s. s.. S. arcisepala, S. cernua, and S. ochroleuca in turn overlap with the morphological variability of S. incurva. Thus morphological measurements alone cannot clearly separate a priori groupings. A separate analysis (not shown) including measurements of the sepals, labella, leaves, and whole plants found a slight difference between a priori taxa, however all taxa were widely overlapping, essentially similar or overlapping in all of the features we measured.

Qualitative characters appear to be more useful in confidently distinguishing members of the S. cernua species complex sensu nov. The lateral sepals of S. arcisepala are typically downwardly falcate (Figs. 10, 11), whereas the lateral sepals of S. incurva typically sweep upward and occasionally outward at the apices (Figs. 12, 13); similarly, the lateral sepals of S. cernua usually sweep upwards or are more or less parallel to the ground (i.e. within the plane of the dorsal sepal and petals) (Figs. 1, 2). The flowers of S. arcisepala are also generally slightly smaller than S. cernua and S. incurva, and the flowers of S. cernua, as indicated by the specific epithet, often become strongly nodding as the inflorescence matures, whereas the flowers of S. arcisepala and S. incurva are more commonly held parallel to the ground, or only slightly nodding (the flowers of S. incurva are occasionally slightly ascending).

Taxonomic Treatment

Fig.9.

Boxplot comparisons of select labellum characters for the Spiranthes cernua species complex s. s. and hybrids. A. Callosity (nectar gland) height (mm). B. Labellum length (mm). C. Width of lower labellum (mm).

Fig.10.

Spiranthes arcisepala.A.Habit. B. Inflorescence. C. Flower and floral bract in profile.D–G. Flower, dissected view.D.Dorsal sepal. E. Dorsal petal. F. Lateral sepal. G. Labellum. H. Labellum, column, and ovary in natural position. I–K. Column. I. Dorsal and ventral views. J. Profile. K. Anther. L. Pollinia. Drawn from Pace 1005 and Pace 1008 by Bobbi Angell.

Fig.11.

Spiranthes arcisepala. A. Type location, Long Lake, Hamilton Co.,New York. B. Delaware Co., New York, sample NY1. C. Indian Lake, Hamilton Co., New York. D. Wells, Hamilton Co., New York. E. Morris Co., New Jersey. F. Fayette Co., Pennsylvania.

Fig.12.

Spiranthes incurva. A. Habit. B. Inflorescence. C. Flower and floral bract in profile. D–G. Flower, dissected view. D. Dorsal sepal. E. Dorsal petal. F. Lateral sepal. G. Labellum. H. Labellum, column, and ovary in natural position. I–K. Column. I. Dorsal and ventral views. J. Profile. K. Anther. Drawn from Pace 629, Pace 631, and Pace 1007 by Bobbi Angell.

Fig.13.

Spiranthes incurva. A. Hamilton Co., New York. B. Portage Co., Ohio, sample sc31. C. Orleans Co., Vermont, sample soch10a. D. Dane Co., Wisconsin. E. Cleistogamous, entirely apomictic population, Lake Co., Indiana. F. Potter Co., Pennsylvania. G. Co-occurring S. incurva (L) and S. ochroleuca (R) (artificially arranged), Warren Co., New York.

Fig.14.

Distribution maps of the S. cernua species complex s. s., based on herbariumspecimens and phylogenetic sampling. A. S. arcisepala, S. cernua s. s., S. incurva, S. ×kapnosperia, S. niklasii. B. S. casei and S. ochroleuca. Maps produced by Elizabeth Kiernan, New York Botanical Garden GIS Lab.

Fig.15.

A–B. Habitat of S. cernua s. s. A. Marshy pond edge, with Nyssa sylvatica Marshall, Chamaecyparis thyoides (L.) Britton, Sterns & Poggenb., Sarracenia L., and Eriocaulon L., Lexington Co., South Carolina. B. Seeping, graminoid-cyperoid and brushy roadside, mixed with S. ×kapnosperia (type location), Transylvania Co., North Carolina. C–E. Habitat of S. incurva. C. Xeric roadside slope, with Solidago nemoralis Aiton and Andropogon gerardi Vitman, Orleans Co., Vermont; sample location of soch10. D. High diversity fen, Clark Co., Ohio, sample location of sc32a. E. Sandy, xeric dune swales and back dunes, co-occurring with S. magnicamporum (with Mr. S. Martella), Lake Co., Illinois. F. Habitat of S. niklasii; xeric, prairie-like opening, Greene Co., Arkansas, sample location of AR1. G–H. Habitat of S. arcisepala. G. Seeping fen, McKean Co., Pennsylvania. H. Type location; roadside ditch and Sphagnum seep with Eriophorum virginicum L. under a power-line cut, Hamilton Co., New York.

Fig.16.

Spiranthes ×kapnosperia. A. Habit. B. Inflorescence. C. Flower and floral bract in profile. D–G. Flower, dissected view. D. Dorsal sepal. E. Dorsal petal. F. Lateral sepal. G. Labellum. H. Labellum, column, and ovary in natural position, with detail of spherical abaxial labellum glands. I–K. Column. I. Dorsal and ventral views. J. Profile. K. Anther. L. Pollinia. Drawn from Pace 1024 and Pace 1030 by Bobbi Angell.

Fig.17.

Spiranthes niklasii. A. Habit. B. Inflorescence. C. Flower and floral bract in profile. D–G. Flower, dissected view. D. Dorsal sepal. E. Dorsal petal. F. Lateral sepal.G. Labellum, with detail of central adaxial papillae.H. Labellum, column, and ovary in natural position, with detail of abaxial labellumsurface. I–K. Column. I. Dorsal and ventral views. J. Profile. K. Anther and pollinia. Drawn from Pace 1036 by Bobbi Angell.

Discussion

For the first time, the evolutionary relationships and morphological variation of the entire S. cernua species complex are put into a comprehensive molecular phylogenetic, biogeographic, and taxonomic context. Our phylogenetic hypotheses indicate that these orchids are recently evolved, very closely related, that speciation via hybridization has occurred, and that incomplete lineage sorting may be a complication to more fully understanding their evolutionary relationships. This conclusion is supported by the presence of short branch lengths and an ambiguous and sometimes discordant genetree topology within the S. cernua species complex s. l. (Figs. 5, 6, 7). Separate analyses indicate the S. cernua species complex s. l. evolved ca. 2–4.5 mya, and the S. cernua species complex s. s. shared a common ancestor ca. 1.5–2.5 mya (unpublished data). Pollinators are widely shared across species within the complex (Catling 1982, 1983), and we have documented three instances of hybrid speciation, indicating that speciation has likely occurred due to evolutionary forces other than complete reproductive isolation. It is possible that some of these species may have evolved in response to the glacial cycles of the Quaternary as populations of ancestral species fragmented and/or migrated, and that hybridization likely allowed for newly evolved species to adapt to newly available environments and microhabitats.

Owing to the pioneering work of Catling (1982) and Sheviak (1976, 1982, 1991), S. cernua s. l. had traditionally been hypothesized to engage in frequent unidirectional and geographicallyspecific hybridization as the “gene recipient” with all other members of the traditional S. cernua species complex.According to this hypothesis, all individuals of S. cernua s. l. that were found within the general range of any other complex member were presumed to be hybrid in origin. Sheviak (1982, 1991) called potential S. magnicamporum × S. cernua hybrid populations and individuals (here designated as S. incurva) triploid and tetraploid forms and “low-prairie races” of S. cernua that arose through “adaptive gene flow” whereby ecologically advantageous genes, characteristics, and morphologies were incorporated into S. cernua s. l. through repeated backcrossing. Thus S. cernua s. l., which in eastern North America inhabits moist habitats, could survive in more xeric and pyrogenic Midwestern prairies. The overarching premise of this hypothesis, that hybridization might confer some evolutionary benefit in a novel environment on the resulting plants vs. nonhybrid S. cernua through incorporation of novel genetic information and neofunctionalization, is plausible and cannot be rejected by the molecular phylogenetic data we present here. However, S. incurva grows in a wide variety of habitats, ranging from periodically inundated to xeric (Fig. 15). Furthermore, S. incurva does not occur over the entire shared range of S. cernua s. l. and S. magnicamporum as originally hypothesized by Catling (1982) and Sheviak (1976, 1982, 1991; Sheviak and Brown 2002), and S. incurva also occurs outside of the range of S. magnicamporum in regions such as Wisconsin north of the Tension Zone, the Keweenaw Peninsula of Michigan, and northern Vermont (Fig. 14). Because all of our samples from the upper Midwest, Great Lakes Basin, Interior Lowlands, western and northern Appalachians, and Saint Lawrence River Valley originally identified as S. cernua were recovered as hybrids, it appears likely that all S. cernua s. l. (minus those that are S. arcisepala) from this region actually represent S. incurva. Catling and Brown (1983), working on cooccurring populations of S. cernua s. l. and S. magnicamporumin the Eastern Peninsula of Southwestern Ontario, identified only three potential individuals of S. magnicamporum × S. cernua. Although we were unable to incorporate samples of putative S. cernua s. l. from that specific area into our study, included accessions of S. cernua s. l. from nearby Ohio and Ottawa, Ontario, were recovered as S. incurva, and it is likely that all of the plants in Catling and Brown (1983) identified as S. cernua were in fact S. incurva. Samples we included and identified a priori as S. magnicamporum from two stations in southern Ontario were recovered as non-hybrids. Individuals we included of S. cernua s. s. from Arkansas and Texas (including S. parksii), were also not recovered as hybrids, notwithstanding the occurrence of S. magnicamporum throughout those regions. Furthermore, as indicated by the taxonomically re-interpreted breeding work of Sheviak (1982), S. incurva is most likely reproductively isolated from S. magnicamporum as crosses between these species did not produce seeds, and so introgression back into S. magnicamporum appears to be impossible.

We did not recover evidence supporting hybridization between S. arcisepala and S. magnicamporum, even when these plants grow within pollination range in northern Ohio or western Virginia, nor did we find evidence of hybridization between S. arcisepala and S. incurva where these taxa are broadly sympatric in Ohio, Pennsylvania, and the Adirondack Mountains (Fig. 14). Future population genetic based approaches may reveal evidence supporting low-levels of introgression that we were unable to detect given the molecular markers we utilized for the present work. We propose, however, that S. incurva and S. arcisepala are distinct species worthy of recognition.

Taxonomy —Interpretation of published names, particularly older names, in Spiranthes is frequently challenging in the absence of accompanying images, well-preserved specimens, and genetic and geographic context. Multiple lines of evidence, however, support the conclusions we reach here. Spiranthes cernua can be placed as the correct name for our S. cernua s. s. clade based on the morphology of its lectotype. The geographic origin of this specimen cannot be definitively determined (“Virginia, Canada”), as Kalm traveled throughout the ranges of S. arcisepala, S. cernua s. s., and S. incurva, from the greater Philadelphia region, north to Montreal, and west to Niagara Falls. Furthermore, the brief formal description provided by Linnaeus could apply to either S. cernua s. s. or S. incurva (though it hews more towards S. cernua). The predominantly strongly nodding flowers of the lectotype, however, are particularly common features of plants studied and collected for phylogenetic analysis by M. Pace in northern Delaware and the Mid-Atlantic region (e.g. sample sc1), and the labellum is essentially identical to the broadly oblong, dilated labellum of plants from the Coastal Plain (i.e. S. cernua s. s.) versus the broadly to narrowly lanceolate labellum of S. incurva. The upward to horizontal position of the lateral sepals do not suggest any affinity to S. arcisepala. Thus, we are confident in assigning the Kalm lectotype specimen from LINN to our phylogenetic clade as S. cernua s. s. As discussed in Sheviak (1982), the type of Limodorum autumnale likely conforms to plants recovered within our phylogenetic clade S. cernua s. s., and should be placed in synonymy with S. cernua. The collection of that specimen from “Carolina” also supports this conclusion (but see Ward (2007), which discusses ambiguities with attributing the “Walter Herbarium” to Thomas Walter, and Ward and Beckner (2011), which does not select a neotype at BM from the Walter/Frasier herbarium for L. autumnale).

In contrast to S. cernua s. s., the detailed original description (Jennings 1906) and ample type material of S. incurva (as Ibidium Salisb. ex Small) creates a rich and detailed concept of this species, matching the morphology of individuals we recovered as being hybrid in nature from the prairies and western Appalachians. Combined with its geographic origin from the narrow strip of Central Lowlands along the Lake Erie coast in northwest Pennsylvania, a region where our phylogenetic data indicate S. cernua s. s. does not occur, this name confidently and unambiguously represents plants that we found using molecular data to be hybrid in origin between S. cernua s. s. and S. magnicamporum. As the oldest confidently placed name for hybrid S. cernua s. s. × S. magnicamporum plants, Ibidium incurvum is transferred to Spiranthes, serving as the basionym for S. incurva.

Ambiguous Names —Neottia cernua var. major Eaton was described as “stem tall, somewhat leafy: flowers very large” with no geographic origin or type designation (Eaton 1829). Given the brief description and lack of a type, it is possible N. cernua var. major represents S. magnicamporum, S. incurva, S. odorata, or large flowered S. cernua s. s., and this name cannot be placed. Sheviak (1982) tentatively placed N. cernua var. major in synonymy with S. cernua s. l., and suggested that one of his ‘New England forms’ frequently found in dry to moist sand may represent this variety. Curiously, the illustration paired with this suggestion approaches S. arcisepala, however, we would not describe S. arcisepala as particularly tall or leafy, and the flowers are typically smaller in all dimensions than S. cernua, S. incurva, S. magnicamporum, and S. odorata. We confidently state that N. cernua var. major should not be applied to S. arcisepala, although it cannot be placed in synonymy with any particular name.

The names S. brevicaulis Raf., S. flexuosa Raf. nom. illeg., and S. petiolaris Raf. all lack type specimens (likely destroyed), and their descriptions (Appendix 3) are too brief and ambiguous to assign any one name to a particular morphology or phylogenetic clade with any confidence. Sheviak (1982) reached a similar conclusion, and termed S. brevicaulis and S. flexuosa “nomina obscura.” Based on their formal description and geographic origin S. brevicaulis and S. petiolaris could represent S. cernua, S. incurva, or S. magnicamporum; all three taxa match aspects of Rafinesque's descriptions, and all three grow in the general vicinity of the Ohio River Valley in Illinois (S. petiolaris) and Kentucky (S. brevicaulis). The description of S. flexuosa, collected in the Appalachian Mountains (no state given), is so vague that it could represent S. arcisepala, S. cernua s. s., S. incurva, S. ochroleuca, or S. ovalis.As is the case withN. cernua var. major, although these are validly published names, the destruction of their type specimens and lack of descriptive detail render them unable to be placed, and thus relegated to the side lines of taxonomy in favor or more recently published and confidently placed names.

Spiranthes casei, S. parksii, and the Role of Apomixis in Questions of Species Status and Taxonomic Rank —The S. cernua species complex has a long history of study focused on the occurrence of apomixis, most commonly examining the phenomenon in S. cernua s. l. (Leavitt 1900, 1901; Schnarf 1929; Swamy 1948; Catling 1982; Sheviak 1982; Schmidt and Antifinger 1992). This literature establishes facultative polyembryony as a frequent occurrence across the geographic ranges of S. cernua s. s. (including S. parksii), S. incurva, S. ochroleuca, and S. casei. Populations of some taxa, such as S. cernua s. s., S. incurva, and S. casei, are frequently composed entirely of apomicitic individuals (e.g. the former S. parksii), and the frequency of apomixis can vary from year to year, possibly in relation to environmental factors. Apomixis is also frequently tied to instances of cleistogamy in S. incurva (Fig. 13) and/or degrees of peloria (typically a reduction of the labellum as exhibited in S. casei, some populations of S. incurva, and the former S. parksii).

Apomixis and the accompanying concept of microspecies are important and regionally significant evolutionary forces, particularly when combined with polyploidy and complex patterns of hybridization and reticulate evolution (e.g. Burgess et al. 2014; Dyer et al. 2012). However, we find the use of apomixis to support the species status of micro-endemic, mostly asexually reproducing, and morphologically minutely different individuals and micropopulations (e.g. S. parksii; see also Sorbus (Ludwig et al. 2013)), as less than ideal, especially when hybridization and reticulation occur between these “species”, such that it could be argued these microspecies actually represent an interbreeding meta-population of a single species. Accordingly, we think apomictic populations must also have additional features distinguishing them from other taxa.

Supporting the earlier work of Dueck et al. (2014), our expanded chloroplast and nuclear reconstructions recovered S. casei as broadly embedded within a largely unresolved polytomy with S. ochroleuca (Fig. 5). Additional analyses combining the chloroplast and nuclear datasets, and eliminating samples with any missing data, did not yield wellsupported resolution between these two a priori species (Fig. 6). Based on this phylogenetic topology and the morphological, ecological, and reproductive context we discuss below, we think the evolutionary history of S. casei may best be expressed as an ecologically specific subspecies of S. ochroleuca. We do not, however, propose any formal changes at this time. Spiranthes casei is clearly very closely related to S. ochroleuca, and can be distinguished from S. ochroleuca by its smaller flowers, non-recurved dorsal sepal and petal apices (Fig. 3), reduced and slightly sub-peloric cordate labellum, predominantly apomictic reproductive mode, and restriction to lichen and bracken barrens of the greater northern Great Lakes Basin and the Canadian Maritimes (Fig. 14). Spiranthes casei and S. ochroleuca, however, are both variable species, and approach or overlap in several characters, including flower and abaxial labellum color, habitat, and inflorescence spiral tightness. In the case of the former S. parksii, all of the available molecular data fail to find any well-supported reciprocally monophyletic cladogenesis or genetic distinction between this purported species and S. cernua s. s. from surrounding areas in Texas and the wider North American Coastal Plain (Figs. 5, 6). Furthermore, the plants are highly restricted in distribution, all apparent floral distinctions appear to be tied to mutant peloria (which is particularly common in the short-grass prairies of the Midwest), and there is no ecological or phenological distinction between it and S. cernua s. s. As such, we think the evidence strongly supports the position of ‘S. parksii’ as an apomictic regional floral morph and synonym of S. cernua s. s. In contrast, predominantly apomictic S. casei is distributed over a wide area and is mostly morphologically and ecologically distinct fromS. ochroleuca. It may be possible that S. casei has evolved several times as an ecotype specializing on lichen barrens and disturbed open thickets or may represent an extreme within the wider morphological variability of S. ochroleuca. Future studies that incorporate next generation DNA sequencing data and methods that address questions at the population genetic level should be conducted in order to continue to evaluate the hypotheses presented here regarding the species status of S. casei.

Conclusions —Sheviak (1982, 1991) correctly hypothesized that hybridization has played an important role in the evolution of the S. cernua species complex, but the observed patterns of hybridization and evolutionary relationships are often different than he anticipated in both scope and the species involved. Hybridization has occurred between S. cernua s. s. and S. magnicamporum (S. incurva), however, hybridization is unlikely to be ongoing and is confined to a specific biogeographic region. Hybridization has also occurred between S. cernua s. s. and S. ochroleuca (to form S. ×kapnosperia), but only in an extremely limited area, and we present evidence for an unexpected ancient hybrid between S. cernua s. s. and S. ovalis (S. niklasii). Conversely, we did not recover evidence to support hybridization between S. cernua s. s. and S. odorata or S. romanzoffiana Cham., and some of the morphological variation expressed by S. cernua s. l. actually represented cryptic speciation in the absence of detectible hybridization events (S. arcisepala).

Although the species within the S. cernua species complex may still be challenging to identify due to overlapping morphological characters, they represent an intriguing natural group through which to explore evolutionary questions and hypotheses such as how speciation occurs in the presence of ongoing hybridization and geneflow, how to delineate withinspecies and between-species phenotypical variability, the role of geography in speciation, and the phenomenon of cryptic speciation. Although we have presented solutions to several long-standing systematic problems within the complex through our taxonomic refinement of S. cernua, recognition of S. incurva, and the description of S. arcisepala, S. niklasii, and S. ×kapnosperia, more population-based research is warranted so that fine-scale levels of hybridization and population connectivity can be illuminated. It is our hope that the information we present here will facilitate conservation of the S. cernua species complex while also informing research focused on other orchid genera in which hybridization is hypothesized to be a prominent evolutionary force, such as Epidendrum (e.g. Pinheiro et al. 2010), Platanthera Rich. (e.g. Wallace 2003), and Tolumnia Raf. (e.g. Ackerman and Galarza-Pérez 1991).