Comprehensive sampling from mitochondrial genomes substantiates the Neoproterozoic origin of land plants.

Molecular phylogenetics illustrates the evolution and divergence of green plants by employing sequence data from various sources. Interestingly, phylogenetic reconstruction based on mitochondrial genes tends to exhibit incongruence with those derived from nuclear and chloroplast genes. Although the uniparental inheritance and conservatively retained protein-coding genes of mitochondrial genomes inherently exclude certain potential factors that affect phylogenetic reconstruction, such as hybridization and gene loss, the utilization of mitochondrial genomes for phylogeny and divergence time estimation remains limited. Here, we meticulously assembled a comprehensive dataset comprising 565 mitochondrial genomes spanning all major lineages of green plants. Using multiple partitions and phylogenetic models, our phylogenies based on mitochondrial genes support paraphyly for both bryophytes and charophytes, place hornworts (Anthocerotaceae) as sister to all tracheophytes, and recover stonewort (Charophyceae) as sister to land plants. We systematically assessed potential impact factors in mitochondrial coding sequences, such as GC-content heterogeneity and codon-usage bias. Further, by rigorously testing seven different dating strategies, we examined the impact of various confounding factors that may affect divergence time estimation, including various numbers and prior settings of fossil calibrations, and rate heterogeneity among sites and across lineages. Our dating results support a Neoproterozoic origin (i.e., Crown age) of land plants and a Triassic origin of angiosperms, as also supported by nuclear data. In conclusion, we particularly underscore the significance of exploring different partitioning strategies and addressing among-lineage heterogeneity in both phylogenetic and dating analyses by extending sampling combined with meticulous data pruning to mitigate the influence of systematic errors in phylogenetic inference.