Despite many attempts in the Sanger sequencing era, the phylogeny of fig trees remains unresolved, which limits our ability to analyze the evolution of key traits that may have contributed to their evolutionary and ecological success. We used restriction-site-associated DNA sequencing (c. 420 kb) and 102 morphological characters to elucidate the relationships between 70 species of Ficus. To increase phylogenetic information for higher-level relationships, we targeted conserved regions and assembled paired reads into long loci to enable the retrieval of homologous loci in outgroup genomes. We compared morphological and molecular results to highlight discrepancies and reveal possible inference bias. For the first time, we recovered a monophyletic subgenus Urostigma (stranglers) and a clade with all gynodioecious Ficus. However, we show, with a new approach based on iterative principal component analysis, that it is not (and will probably never be) possible to homogenize evolutionary rates and GC content for all taxa before phylogenetic inference. Four competing positions for the root of the molecular tree are possible. The placement of section Pharmacosycea as sister to other fig trees is not supported by morphological data and considered a result of a long-branch attraction artefact to the outgroups. Regarding morphological features and indirect evidence from the pollinator tree of life, the topology that divides Ficus into monoecious versus gynodioecious species appears most plausible. It seems most likely that the ancestor of fig trees was a freestanding tree and active pollination is inferred as the ancestral state, contrary to previous hypotheses. However, ambiguity remains on the ancestral breeding system. Despite morphological plasticity, we advocate restoring a central role to morphology in our understanding of the evolution of Ficus, as it can help detect systematic errors that appear more pronounced with larger molecular datasets.

中文翻译：

---

探索系统偏差、生根方法和形态学证据，揭开榕属（桑科）的进化史

尽管在桑格测序时代进行了许多尝试，但无花果树的系统发育仍未得到解决，这限制了我们分析可能有助于其进化和生态成功的关键性状进化的能力。我们使用限制性位点相关 DNA 测序 ( c . 420 kb) 和 102 个形态特征来阐明 70 种榕树之间的关系。为了增加更高层次关系的系统发育信息，我们针对保守区域并将配对读取组装到长基因座中，以检索外群基因组中的同源基因座。我们比较了形态学和分子学结果以突出差异并揭示可能的推理偏差。我们第一次发现了一个单系亚属Urostigma（扼杀者）和一个拥有所有雌雄同株榕的进化枝。然而，我们通过基于迭代主成分分析的新方法表明，在系统发育推断之前，不可能（并且可能永远不会）使所有分类群的进化速率和 GC 含量均质化。分子树根的四个竞争位置是可能的。将Pharmacosycea部分作为其他无花果树的姐妹的位置没有得到形态学数据的支持，并且被认为是对外群的长分支吸引人工制品的结果。关于来自生命传粉树的形态特征和间接证据，将榕树分为雌雄同株与雌雄同体的物种似乎最合理。似乎最有可能的是，无花果树的祖先是一棵独立的树，主动授粉被推断为祖先状态，这与之前的假设相反。然而，祖传育种系统仍然存在模棱两可的问题。尽管形态具有可塑性，但我们主张恢复形态在我们理解无花果进化过程中的核心作用，因为它可以帮助检测在更大的分子数据集上显得更加明显的系统错误。