Cycad phylogeny spans 346 accessions and all 10 genera

A species-level phylogenomic analysis of cycads recovered all 10 genera as monophyletic, while showing that the apparent certainty of many branches should be interpreted alongside considerable disagreement among individual gene trees. The study analyzed 346 cycad accessions, including approximately 326 accepted species, or about 86% of the 380 species recognized by the World List of Cycads. All 10 genera and both families were represented, with Ginkgo biloba included as the outgroup. Researchers also produced a time-calibrated tree, individual gene trees and a concatenated amino-acid alignment for use in systematics, biogeography, comparative biology and conservation analyses.

The project combined publicly available RNA sequencing data with the reference genome of Cycas panzhihuaensis. Paired-end Illumina reads were assembled into transcriptomes, from which protein-coding regions were predicted and orthologous gene families were identified. The researchers applied a relaxed single-copy strategy, retaining loci present in at least 75% of the 346 cycad accessions and allowing no more than 10% to carry a second copy. This produced 1,409 nuclear loci with average taxon occupancy of 94.9%, compared with only 115 loci under a strictly single-copy approach.

Each locus was aligned independently, and regions that were poorly aligned or rich in gaps were removed. IQ-TREE then inferred a gene tree for every locus, with phylogenetic pruning used where accessions contained multiple copies. A partitioned maximum-likelihood analysis treated all 1,409 loci as independent partitions, each with its own best-fit substitution model. The concatenated alignment contained 411,345 amino-acid sites. Support was assessed with 1,000 ultrafast bootstrap replicates, while gene and site concordance factors measured how individual gene trees and alignment sites supported each branch.

The resulting backbone placed Dioon as sister to the remaining Zamiaceae. The other zamiads formed two principal clades: Encephalartos, Lepidozamia and Macrozamia in one, and Bowenia, Stangeria, Ceratozamia, Microcycas and Zamia in the other. Every genus had 100% ultrafast bootstrap support and crown-node gene concordance factors of 62% to 81%. The Microcycas–Zamia relationship had the strongest genomic support among the intergeneric nodes, with a gene concordance factor of 66.7% and a site concordance factor of 85.5%.

Support was less uniform for other relationships. Although 240 of 344 scored internal branches, or 69.8%, had bootstrap support of at least 95%, and the median bootstrap value was 100%, the median gene concordance factor was only 3.2%. A total of 272 nodes, or 79.1% of the scored internal branches, had gene concordance below 10%. Site concordance was generally higher, averaging 40.3%, and approximately 71% of scored nodes exceeded the 33.3% expectation under random quartet resolution. The placements of Bowenia and Stangeria were comparatively weak, with gene concordance factors of 11.6% and 16.1% and site concordance factors of 33.8% and 36.8%.

The study links the contrast between high bootstrap values and low gene concordance to the kind of signal contained in the individual loci. The dominant pattern of discordance was polyphyly, with a mean gene-tree failure proportion of 89.0%, rather than strong support for an alternative resolution, whose mean values were 1.6% for the first alternative and 1.7% for the second. The authors note that slow molecular evolution characteristic of gymnosperms, together with recent diversification within cycad genera, can leave short internal branches with too few substitutions for individual loci to resolve confidently. Concatenation across 1,409 loci can still provide strong support for the overall topology.

Divergence times were estimated by penalized likelihood in TreePL after Ginkgo was pruned from the tree, with 12 minimum–maximum age constraints. Five constraints were derived from a fossil-calibrated total-evidence analysis, while seven were secondary genus-level calibrations taken from confidence intervals in recent molecular dating studies. Gene-wise bootstrap resampling was performed 100 times, with the 1,409 loci resampled with replacement and branch lengths re-optimized on the fixed maximum-likelihood topology, to calculate 95% confidence intervals for all 345 internal nodes. The mean width of these intervals was 3.97 million years.

The time-calibrated tree placed the Cycadales crown at approximately 359 million years, with a 95% confidence interval of 320–359 million years. This estimate reached the upper calibration bound and was consistent with the Carboniferous–Permian age inferred from the fossil record. The Zamiaceae crown was estimated at approximately 194 million years, with a 95% confidence interval of 165–196 million years. Backbone divergence estimates generally occupied 28% to 66% of their permitted calibration intervals, indicating that molecular branch lengths contributed to those ages rather than the bounds alone determining them.

Estimates for the crowns of individual genera were less independent of their constraints. Six of the seven genera with crown calibrations converged on the maximum permitted age: Dioon at 57 million years, Cycas at 40 million years, Lepidozamia at 34 million years, Zamia at 33 million years, Macrozamia at 29 million years and Encephalartos at 27 million years. Ceratozamia was the exception. Its estimated crown age of 32 million years, with a 95% confidence interval of 25–33 million years, fell within its 13–36 million-year calibration range and was interpreted as more directly informed by the molecular data. The authors caution that rate smoothing can push ages toward older values across short branches associated with recent radiations.

All trees and supporting data have been archived in Zenodo and linked to the World List of Cycads, where the phylogram and timetree can be explored through an interactive display. The public resources include the maximum-likelihood tree with branch lengths, bootstrap and concordance values; the dated tree with confidence intervals; 1,409 individual gene trees; and the concatenated alignment, partition definitions, support values and calibration data. Tip labels use accepted taxonomic names, allowing tree tips to be connected directly to the World List’s species records. The framework can support taxonomic evaluation, identification of species groups needing denser population sampling, trait mapping, biogeographic analysis and conservation calculations based on phylogenetic diversity or evolutionary distinctiveness.

Subject of Research: Species-level phylogenomic relationships and divergence times of the 10 extant cycad genera

Article Title: A comprehensive phylogenomic framework for cycads ( Cycadales )

Article References: Calonje, M., Clugston, J. A. R., & Coiro, M. (2026). A comprehensive phylogenomic framework for cycads (Cycadales). PhytoKeys, 275, 81-95. https://doi.org/10.3897/phytokeys.275.194283

Image Credits: AI Generated

DOI: 10.3897/phytokeys.275.194283

Keywords: cycads, Cycadales, phylogenomics, phylotranscriptomics, molecular dating, gene concordance, Zamiaceae, Cycadaceae, plant evolution, open genomic data, comprehensive, phylogenomic