Hidden genes rewrite the family tree of Taiwan’s elusive forest herbs

A group of small, shadow-dwelling herbs that botanists have struggled to classify for decades has just yielded its secrets to thousands of genes at once. In a study published in Botanical Studies, researchers led by Min-Xin Luo and Pei-Chun Liao sampled every known Aspidistra taxon on the island of Taiwan and reconstructed their evolutionary history from transcriptome data. Their conclusion overturns a long-standing classification: the two varieties of Aspidistra daibuensis, long treated as forms of a single species because they look alike, do not share an exclusive common ancestor. Instead, one of them has been formally raised to full species rank as Aspidistra longkiauensis.

The genus Aspidistra, a member of the asparagus family, comprises more than 200 species scattered across eastern and southeastern Asia. Most live in lowland forest understories below 1,500 meters, flowering close to the ground where litter often hides their blossoms. That habit makes them easy to overlook, and new species keep turning up. Most have narrow ranges, and many are endemic to single locations. Taiwan harbors exactly five taxa: Aspidistra attenuata, Aspidistra daibuensis var. daibuensis, Aspidistra daibuensis var. longkiauensis, Aspidistra mushaensis var. mushaensis, and Aspidistra mushaensis var. longiconnectiva. All five are endemic to the island, and all were included in the new analysis.

The team collected nine field samples and grew them in a common garden before extracting RNA from young shoots and root apical meristems. After sequencing on an Illumina NovaSeq 6000 platform, they assembled transcripts de novo and identified 1,029 single-copy orthologous genes. A coalescent-based species tree built with ASTRAL-II from maximum likelihood gene trees placed Aspidistra attenuata and the longkiauensis variety together as sister groups, with the two mushaensis varieties forming another sister pair, while daibuensis sat on a more distant branch. A Bayesian multispecies coalescent analysis using 189 genes whose topologies matched the species tree confirmed the same arrangement.

Molecular dating placed the divergence of Taiwanese Aspidistra from its sister species, the widespread Aspidistra elatior, at roughly 0.42 million years ago, with a 95 percent confidence interval of 0.36 to 0.48 million years. A complementary relaxed-clock analysis calibrated with four secondary points from a broader sampling of 341 single-copy genes pushed the estimates further back, putting the crown age of the genus at 3.27 million years and the origin of the Taiwanese lineage at 1.71 million years. Either way, the Taiwanese species are evolutionarily young, and that youth has consequences for how their genes are sorted.

When the researchers interrogated individual gene trees against the species tree using an approximately unbiased topological test, only 18.37 percent of the 1,029 genes, or 189 of them, failed to reject the species tree hypothesis. More than 80 percent of gene trees told a different story, a pattern the authors attribute largely to incomplete lineage sorting, the retention of ancestral genetic polymorphisms that fail to coalesce before successive speciation events. A QuIBL analysis of twelve triplet topologies found that only four showed a statistically significant contribution from introgression, involving just 1, 4, 6, and 12 genes respectively, less than 2 percent of the dataset. The overwhelming majority of discordance therefore reflects incomplete sorting rather than hybridization, a signature of rapid speciation that left hundreds of genes without time to sort cleanly into species-specific lineages.

The morphological puzzle centered on daibuensis and longkiauensis. Although the two look strikingly similar, 20.80 percent of genes, 214 out of 1,029, did not reject a topology grouping them as sisters. Within that set, 19 genes turned out to be chloroplastic, pointing to photosynthesis-related functions as a driver of the resemblance, and at least seven were ubiquitin-related, involved in cell signaling, stress responses, immunity, and morphogenesis. A PAML branch model test then flagged four of these genes as under positive selection, with ratios of nonsynonymous to synonymous substitution rates exceeding one: the early light-inducible protein ELIP1, the chloroplastic gene Iojap, the proteasome biogenesis-associated chaperone PBAC5, and the protein arginine methyltransferase PRMT11. Three of the four share a connection to light response and plant coloration, hinting that the two taxa converged physiologically as they adapted to the dim forest floor.

Approximate Bayesian computation comparing nine evolutionary scenarios gave the strongest support, a posterior probability of 0.6711, to the species tree scenario, which clusters longkiauensis with attenuata rather than with daibuensis. The morphological resemblance scenario scored only around 10 percent, and a hybrid origin for the mushaensis varieties received just 4 percent. Parameter estimates from the best model suggest attenuata diverged from longkiauensis about 8,800 years ago, from the mushaensis lineage about 18,600 years ago, and from daibuensis about 93,100 years ago, assuming a one-year generation time. The top two scenarios both reject the idea that daibuensis and longkiauensis belong together as varieties of one species.

Among 21 quantitative morphological traits measured in an earlier study by Lu and colleagues, only one carried a phylogenetic signal strong enough to track the species tree: stigma width. Blomberg’s K statistic exceeded one for this character, indicating evolutionary conservatism driven by shared ancestry rather than random drift, while every other trait showed no significant deviation from a Brownian motion model. Because the anthers of Aspidistra sit below the stigma, stigma width may filter which pollinators can reach them, potentially reinforcing reproductive barriers between species. The authors argue that this makes stigma shape, and width in particular, a reliable diagnostic character for future taxonomic revisions, whereas traits governed by environmental plasticity can produce homoplasious similarities that mislead classification.

The genetic evidence also resolved the status of the mushaensis varieties. NeighborNet network analysis and principal component analysis both clustered the longiconnectiva variety firmly with mushaensis, supporting its recent reduction from a full species to a variety. Hierarchical STRUCTURE analysis of simple-sequence repeat data showed that apparent hybrid signals between attenuata and mushaensis fade at higher clustering levels, suggesting that what looked like admixture may instead reflect shared ancestral polymorphism. The researchers note that the three western taxa, attenuata, longkiauensis, and mushaensis, all lie on the western side of Taiwan’s Central Mountain Range, while daibuensis occupies the east, a geographic split that may have reinforced their early divergence.

On the strength of these results, the team formally elevated Aspidistra daibuensis var. longkiauensis to species rank as Aspidistra longkiauensis, distinguishing it from daibuensis by a shorter leaf, a smaller leaf blade ratio, thinner perianth lobes, a longer stigma curvature, and a narrower stigma width. The authors caution that their conclusions rest entirely on Taiwan’s endemic species and that with more than 200 relatives across Asia, broader sampling will be needed to confirm whether these patterns hold across the whole genus. Still, the study offers a clear demonstration that morphology alone can mislead when rapid radiation and convergent adaptation are at work, and that only by reading the full genealogical record, gene by gene, can taxonomists untangle species that look alike but evolved apart.

Subject of Research: Phylogenomics and taxonomy of the endemic Taiwanese Aspidistra species

Article Title: Impact of incomplete lineage sorting and natural selection on the phylogenetic and taxonomic uncertainties of Aspidistra in Taiwan

Article References: Luo, M.-X., Yang, M.-J., Lu, C.-T., & Liao, P.-C. (2025). Impact of incomplete lineage sorting and natural selection on the phylogenetic and taxonomic uncertainties of Aspidistra in Taiwan. Botanical Studies, 66(1), Article 35. https://doi.org/10.1186/s40529-025-00482-y

Image Credits: AI Generated

DOI: 10.1186/s40529-025-00482-y

Keywords: Aspidistra, phylogenomics, incomplete lineage sorting, convergent evolution, transcriptomics, taxonomy, Taiwan, species delimitation, stigma morphology, molecular dating, gene tree conflict, natural selection