Searching for Diamonds in the Apomictic Rough: A Case Study Involving Boechera lignifera (Brassicaceae)

The genus Boechera is one of the most difficult species complexes in North America, with about 70 sexual diploids and hundreds of apomictic taxa representing diverse combinations of nearly every known sexual genome. In this study, we set out to clarify the taxonomy of Boechera lignifera, which currently includes a small number of sexual diploid populations in addition to the widespread apomictic diploid upon which the name is based. Using data from cytological studies, microsatellite DNA analyses, geography, and morphology, we demonstrate that the apomictic populations are genetically quite divergent from the sexual diploids. We propose the name Boechera kelseyana to accommodate the sexual diploid taxon, which occurs entirely south of the geographic range of B. lignifera. Boechera kelseyana is consistently separable from B. lignifera based on pollen and seed morphology, the length and proximal orientation of fruiting pedicels, differences in the branching and orientation of trichomes on the lowers stems, and the number of flowers and cauline leaves on unbranched fertile stems. Keywords—Apomixis, Cytology, microsatellites, PCO-MC, species delimitation. Genera exhibiting widespread apomixis are anathema to many taxonomists; the mere mention of their names (e.g. Antennaria Gaertn., Crataegus L., Crepis L., Hieracium L., Potentilla L., Rosa L., Rubus L., and Taraxacum F. H. Wigg.) leaves botanists looking for ways to change the topic of conversation. The most common form of apomixis found in these groups is of the gametophytic type, in which an embryo arises directly from the megagametophyte without fertilization (Asker and Jerling 1992). This particular type of apomixis is strongly correlated with polyploidy (Carman 1997; Whitton et al. 2008; Hojsgaard et al. 2014), and can arise from within a single diploid species (autopolyploidy), through hybridization between species (allopolyploidy) or, at higher ploidy levels, a combination of the two processes (see Ramsey and Schemske 1998). In the few genera that express gametophytic apomixis at the diploid level (e.g. Boechera Á. Löve & D. Löve, Erigeron L.), diploid apomicts represent an added layer of complexity, providing a basis for the production of an even greater diversity of polyploids (Alexander et al. 2015). As species complexes age, the sexual diploids giving rise to the apomictic taxa become increasingly hard to detect. With their genomes represented in a diversity of apomictic derivatives, the sexual diploids become more difficult to distinguish using the characters and tools commonly employed by taxonomists. Over time, successful apomictic populations often become more numerous and widespread, colonizing habitats unoccupied by their sexual diploid progenitors. This pattern is so pervasive in both plants and parthenogenetic animals that it has given rise to a substantial literature under the rubric “geographical parthenogenesis” (Vandel 1928; Lynch 1984; Haag and Ebert 2004; Kearney 2005; Hörandl 2006, 2009; Vrijenhoek and Parker 2009; Verhoeven and Biere 2013). Within regions of geographic overlap, destabilizing hybridization and competition with their apomictic relatives can lead to severe reductions in the geographic range and effective population sizes of the sexual taxa (Lynch 1984; Whitton et al. 2008), substantially increasing the risk of extinction (Lynch et al. 1995; Coron et al. 2013; Coron 2014). The processes involved in the evolution and maturation of apomictic species complexes can greatly complicate efforts by systematists to locate and sample sexual diploids. But the importance of this task cannot be overstated. It is the sexual diploids that are the products of divergent evolution (cladogenesis); only secondarily do they give rise to apomicts through reticulate processes. Thus, the sexual diploids are the foundational “pillars” upon which apomictic complexes are built (Stebbins 1950; Hörandl et al. 2009), and even the most cursory understanding of such complexes must begin with a full accounting of sexual diploid diversity. For the past ten years, our labs have collaborated in an NSF-funded effort to identify and describe all sexual diploids in the genus Boechera (Brassicaceae). As currently circumscribed, the genus includes about 70 such entities, plus hundreds of apomictic taxa involving diverse combinations of nearly every known sexual genome. A few of the sexual species, such as the widespread and well-studied B. stricta (Graham) Al-Shehbaz, are so distinctive morphologically that their genetic contribution to any apomictic hybrid is immediately apparent. However, most sexual diploids belong to clades containing several grossly similar species (Alexander et al. 2013, 2015), and their relationships to apomictic taxa cannot be determined on the basis of morphology alone. By combining cytological and molecular data with in-depth morphological studies, we have made substantial progress in parsing these complexes and identifying the sexual diploids involved. Here we describe the results of our initial investigations of the B. lignifera (A. Nels.) W. A. Weber complex. Boechera lignifera was described in 1899 as a new species of Arabis L., typified based on specimens collected by Aven Nelson near Green River, Wyoming. Most subsequent authors have treated it as a distinct species, either within Arabis (Rollins 1941, 1993; Mulligan 1995; Welsh 2003) or, more recently, as a member of the largely North American Boechera (e.g. Holmgren 2005; Al-Shehbaz and Windham 2010). Based on broad-scale morphological and palynological studies of Boechera, Windham and Al-Shehbaz (2006) reported notable variability among plants assigned to B. lignifera. First, it became clear that the two plants on the “holotype” sheet at the Rocky Mountain Herbarium (RM 12591) represented different taxa, one with sparsely pubescent fruits and the other with glabrous fruits. Overall, the plant with glabrous fruits was a better match for Nelson’s (1899) protologue, and this was designated the lectotype of A. lignifera by Windham and Al-Shehbaz (2006). Most of the pollen remaining on the anthers of this plant was malformed, though a few ovoidspheroid pollen grains suggestive of apomictic reproduction were present as well. Studies of microsporogenesis in plants from the type locality that are morphologically similar to the lectotype indicate that B. lignifera sensu stricto is an apomictic diploid (Windham and Al-Shehbaz 2006). Beck et al. (2012) and Alexander et al. (2015) have shown that diploid apomixis in Boechera is strongly associated with hybridity, and thus it is quite likely that B. lignifera s. s. is of hybrid origin. Based on morphological and ecological similarities, previous authors have suggested a number of potential close relatives that could function as progenitors. These include Arabis (Boechera) selbyi (Rydb.) W. A. Weber (Rollins 1941), one or more of the sexual diploids belonging to the polyphyletic B. holboellii (Hornem.) Á. Löve & D. Löve complex (Welsh 2003; Windham and Al-Shehbaz 2006), or B. cobrensis (M. E. Jones) Dorn (Al-Shehbaz and Windham 2010). However, systematic studies of pollen morphology suggest another possible sexual diploid progenitor; B. lignifera itself. Plants assigned to this taxon are known to vary with respect to pollen type and reproductive mode (Windham and Al-Shehbaz 2006). Whereas most populations in the northern part of the geographic range were inferred to be apomictic in earlier pollen and/or chromosome studies, those at the southern edge (New Mexico and Arizona) produced an abundance of narrowly ellipsoidal, symmetrically tricolpate pollen grains usually associated with normal meiosis and sexual reproduction. In this paper, we explore the taxonomic significance of the observed variation in reproductive mode within Boechera lignifera s. s. Based on a broad sampling of collections scattered across the geographic range of the taxon, we combine data from analyses of microand megasporogenesis, pollen and trichome morphology, microsatellite DNA, geography, and macromorphology to address one central question: Do the apomictic and sexual populations included within B. lignifera by previous authors represent a single, reproductively-variable species or two distinct taxa? Materials and Methods Sampling—Sixty plants were included in the microsatellite analyses that provided the genetic foundation for this effort (Appendix 1); all of these were included in the geographic and morphologic studies, with various subsets subjected to analyses of microsporogenesis (15), pollen morphology (39), and megasporogenesis (3). Samples were chosen to provide broad geographic sampling of both the presumed apomictic (northern) and sexual (southern) populations of B. lignifera s. l. Special emphasis was placed on sampling type specimens and vouchers for published chromosome counts. All plants are represented by voucher specimens deposited at the herbaria listed in Appendix 1. Cytology—Microsporogenesis was studied in flower buds of wild or greenhouse-grown plants fixed in Farmer’s solution (3 parts 95% ethanol: 1 part glacial acetic acid). Fixed materials were stored (for up to 15 yrs) at −20°C and transferred to 70% ethanol before preparing slides. Flower buds were dissected under 20 × magnification on a Leica MZ7.5 Stereozoom dissecting microscope; sampling targeted flowers in which the petals were 1⁄8-1⁄4 the length of the enclosing sepals (a measurement providing a proxy for peak microsporogenesis in Boechera). Anthers from these flowers were macerated in a drop of 1% acetocarmine stain, allowed to stain for at least ten minutes (adding to the stain droplet as needed to prevent drying), then mixed 1:1 with Hoyer’s solution prior to placing the cover slip and squashing. Slides were examined with a Meiji MT5310L phase contrast microscope, and representative cells were photographed using a Canon EOS Rebel T3i digital camera. Pollen was examined from naturally dehiscing anthers of all flowering specimens included in the microsatellite study. For light microscopy, pollen samples were mounted in glycerol, and immediately examined and photographed at 400 × magnification using the phase microscope/camera set-up described in the Cytology section. For our scanning electron microscopy studies of pollen morphology, anthers shedding pollen were mounted on standard SEM stubs using double-sided tape and then sputter coated with gold using a Quorum Q-150T ES Sputter Coater. Samples were examined and photographed on a FEI/Phillips XL-30 S-FEG ESEM at an accelerating voltage of 15 kV. For studies of megasporogenesis, clusters of floral buds at the late pre-anthesis stage and younger were fixed, either in formalin acetic acid alcohol (FAA) for 48 h or in Farmer’s fixative (see Cytology section) for at least 24 h, then transferred to 70% ethanol for storage (at room temperature or at −20°C respectively depending on the fixative used). Buds were cleared in 2:1 benzyl benzoate dibutyl phthalate as in Crane and Carman (1987). Pistils were then excised, and those ranging from 0.7–3.0 mm in length were mounted such that pistils of similar length occurred in separate columns on the slides. Pistils were mounted in a small volume of clearing solution, up to 16 per slide, and covered with a coverslip. Ovaries inside the cleared pistils were studied using a BX53 microscope equipped with differential interference contrast (DIC) optics, and the pistils were photographed using an Olympus MicroFire 599809 camera. Microsatellite Analyses—Genomic DNA was extracted from 60 airdried herbarium specimens (Appendix 1) using a CTAB protocol modified for 96-well plates (Beck et al. 2012), Qiagen DNeasy plant mini kits (Qiagen, Germantown, Maryland), or the modified Qiagen DNeasy protocol described in Alexander et al. (2007). Microsatellite allele variation was assessed at 14 previously published loci [ICE3, ICE14 (Clauss et al. 2002); a1, a3, b6, c8, e9 (Dobeš et al. 2004); BF3, BF9 BF11, BF15, BF18, BF20, Bdru266 (Song et al. 2006)]. Forward primers for each locus were labeled with 6-FAM or HEX, and sets of three loci were simultaneously amplified using a multiplex PCR protocol. Reactions (8 ul) contained 2.5 μl 2x Qiagen Multiplex PCR master mix, 0.2 μM each primer, and approx. 20 ng DNA template. Cycling conditions included denaturing at 95°C (15 minutes), 30 cycles of 94°C denaturing (30 sec), 53°C annealing (90 sec), and 72°C extension (60 sec), followed by a final extension step at 60°C (30 minutes). Amplicons were sized using 500 ROX on an Applied Biosystems 3730xl DNA Analyzer and alleles were determined using GeneMarker 1.9 (SoftGenetics, State College, Pennsylvania). Samples exhibiting duplicate within-speciesmultilocus genotypes were then excluded. In order to visualize major genetic contrasts among the samples, the 14-locus allele matrix was subjected to a principal coordinates analysis (PCoA) in GenAlEx 6.0 (Peakall and Smouse 2006) using a standardized covariance matrix derived from a binary genotypic genetic distance (Huff et al. 1993). The matrix was then subjected to multidimensional clustering using principal coordinates analysis with modal clustering (PCOMC) (Reeves and Richards 2009, 2011). This approach identifies the most cohesive groups in a dataset by simultaneously considering information on all informative PCoA axes, ranking each group by a stability value that reflects the density of the group in multidimensional space. Geography and Morphology—The distribution map for apomictic and sexual populations of B. lignifera s. l. was generated with QGIS v. 2.0.1-Dufour software (QGIS Development Team 2014), using state and county shape files obtained from the DIVA-GIS website (www.divagis.org/Data; Hijmans et al. 2004). Latitude and longitude for specific SYSTEMATIC BOTANY [Volume 40