Lineage and role in integrative taxonomy of a heterotrophic orchid complex

Abstract

Lineage-based species definitions applying coalescent approaches to species delimitation have become increasingly popular. Yet, the application of these methods and the recognition of lineage-only definitions have recently been questioned. Species delimitation criteria that explicitly consider both lineages and evidence for ecological role shifts provide an opportunity to incorporate ecologically meaningful data from multiple sources in studies of species boundaries. Here, such criteria were applied to a problematic group of mycoheterotrophic orchids, the Corallorhiza striata complex, analysing genomic, morphological, phenological, reproductive-mode, niche, and fungal host data. A recently developed method for generating genomic polymorphism data-ISSRseq-demonstrates evidence for four distinct lineages, including a previously unidentified lineage in the Coast Ranges and Cascades of California and Oregon, USA. There is divergence in morphology, phenology, reproductive mode, and fungal associates among the four lineages. Integrative analyses, conducted in population assignment and redundancy analysis frameworks, provide evidence of distinct genomic lineages and a similar pattern of divergence in the extended data, albeit with weaker signal. However, none of the extended data sets fully satisfy the condition of a significant role shift, which requires evidence of fixed differences. The four lineages identified in the current study are recognized at the level of variety, short of comprising different species. This study represents the most comprehensive application of lineage + role to date and illustrates the advantages of such an approach.

Keywords: ISSRseq; Orchidaceae; SNP data; integrative species delimitation; species boundaries; species concept.

Figure 1.

A. Corallorhiza striata var. vreelandii (green). B. C. striata var. striata (red). C. C. striata from the Sierra Nevada of California, USA (blue). D. C. striata from the Coast Ranges/Cascades of California and Oregon, USA (magenta). E. Map showing the geographic range of the C. striata complex. Black empty circles are GBIF records based on herbarium collections. Colored, filled circles are sampling localities included in the ISSRseq analysis (see Fig. 2) for each of the four groupings in A.

Figure 2.

Relationships and genetic differentiation of the C. striata complex based on SNP data from ISSRseq. A. Relationships based on maximum likelihood analysis in RAxML-NG with an ascertainment bias model. B. Relationships based on analysis in SVDQuartets. Total weight of compatible quartets = 0.927. C. Relationships based on maximum likelihood analysis in IQtree2 with an ascertainment bias model. D. Principal Components Analysis of linkage disequilibrium-thinned SNP data, showing PCs 1-3. E. Population structure analysis in DyStruct for K=2-4. Colors correspond to legend in d).

Figure 3.

Morphological, phenological, and reproductive mode analyses of the C. striata complex. A. Principal Components Analysis (PCA) of 14 log-transformed floral morphological characters, showing PCs 1-2. B. Biplot showing the loading scores of each character on PCs 1-2. C. PCA of PC axes 1 and 3. D. Biplot of character loadings on PCs 1-3. E. Linear Discriminant Analysis (LDA) of the same data, showing LDA axes 1 and 2. F. Biplot of the morphological characters on LDA axes 1 and 2. G. Phylomorphospace representation of PC scores for PC axis 1, with relationships from the RAxML-NG tree superimposed. H. Two-dimensional phylomorphospace representation with the RAxML-NG tree on PCs 1 and 2. I. Density plots of the six most informative characters differentiating Sierra Nevadan and Coast/Cascade accessions. Scale is in mm. J. Violin plots of specimen records by flowering date. Note that Cascades accessions were split from Coast Range accessions here to investigate differences in flowering time between accessions from these two regions. K. Boxplots of the proportion of flowers in a raceme with clear evidence of swollen ovaries. Letters ‘a’ and ‘b’ are Tukey post-hoc comparison values that are significantly different (p < 0.0001).

Figure 4.

A. Phylogenetic analysis of >1,400 Tomentella fungal accessions, including GenBank reference sequences and members of the C. striata complex. Scale bar = 0.2 substitutions/site. B. Closeup of the clade occupied by all accessions of the C. striata complex, with closely related reference sequences from Tomentella fuscocinerea and T. patagonia. Branch lengths are scaled proportionally. Blue and magenta arrows point to associates of the Sierra Nevadan C. striata and the Coast/Cascade C. striata, respectively. Roman numerals indicate three principal sub-clades.

Figure 5.

Integrative analyses of SNP and ‘extended’ data. A. SNPs only analyzed in assignPop. B. Combined SNP and ‘extended’ data (morphology, biotic niche, and fungal hosts). C. ‘Extended’ data only. ‘LDA’ = linear discriminant analysis; ‘SVM’ = support vector machine; ‘RF’ = random forest. D. Redundancy Analysis (RDA) and variance partitioning using four- and three-species delimitations as explanatory variables with ‘extended’ data as response variables (* = Corallorhiza striata var. striata, C. striata var. vreelandii, Sierra Nevada + Coast/Cascades).