Similar to the most reported plastomes of angiosperms [17, 40], all the plastomes in our study exhibited relatively conservative genome structure, gene content, and gene order [41, 42]. However, when compared to H. morsus-ranae and L. laevigatum, H. dubia (159,698 bp) and H. chevalieri (158,066 bp) had larger plastomes. One reason for plastome size variation is the expansion or contraction of IRs [43, 44]. Indeed, H. dubia and H. chevalieri had one (ycf1) and three more genes (ycf1, rps15, and ndhH) in IRa, respectively, than the other two species. In addition, typical patterns of the plastome evolution associated with the fluctuation of plastome length include the gain/loss of genes, pseudogenization, and variations in intergenic regions [44–46]. Our results revealed the loss of gene rps16 in African endemic H. chevalieri. This gene has been reported missing from plastome in many plant groups, e.g., Oxalidaceae [47, 48], Podostemaceae , and Violaceae . It seems that rps16 has been frequently transferred to the nuclear genome [40, 50]. Additionally, a small inversion of the trnN-GUU gene near the IR boundary was detected in H. dubia. The inversion was likely related to the shift in the IR boundary (Fig. 2), as identified and demonstrated by Zhu et al..
Previous studies [4, 8] have applied only five plastid genes (rbcL, matK, rpoB, rpoC1, trnK) to resolve the phylogeny of the genus Hydrocharis. However, there was only low nucleotide diversity( Pi < 0.06) within these five genes, which might have affected the accuracy and resolution of phylogenetic reconstruction. In our study, a number of regions (Fig S2), including five plastid genes (ycf1, rps12, rps18, rps3, accD) and five intergenic spacers (ycf1-trnN-GUU, rps16-trnQ-UUG, trnI-CAU-ycf2, ycf2-trnI-CAU, trnK-UUU-rps16), were detected as hotpots and may be helpful even in population genetic studies in the genus.
The high frequency of repeat sequences has been demonstrated to be one of the leading causes of plastome rearrangement and divergence . Our analysis identified many repetitions (> 350) with more than 30 bp length in all newly assembled plastomes. Among these, short repeat sequences (30-45 bp) were dominant, similar to other plant plastomes that have not undergone large-scale structural variation. SSRs are widely distributed in plant plastomes and exhibit relatively high polymorphism, which can be used in population genetics . The majority of SSRs were mono- and di- nucleotides in four newly assembled plastomes, which have been reported in other angiosperm plastomes, e.g., Primula , Dendrosenecio , and Oxalis . The cpSSRs reported here could be used as genetic markers for future studies into the genetic diversity of Hydrocharis.
Phylogeny and biogeographical reconstruction
Prior to this study, no molecular information was available for the African endemic H. chevalieri. Here we have clarified the relationships within Hydrocharis by assembling and reporting the plastomes of all three Hydrocharis species and Limnobium laevigatum. Unlike previous studies [4, 8], we recovered non-monophyletic Hydrocharis. The genus Limnobium represented here by L. laevigatum was nested in Hydrocharis as sister to H. dubia with robust support (Fig. 3, BS = 100/100; PP = 1/1). This contradicting result might be because previous studies did not include H. chevalier and used only a limited number of molecular markers [4, 8]. Furthermore, a series of morphological features support the current phylogenomic relationships. Vegetatively, Hydrocharis and Limnobium are indistinguishable, but H. chevalieri has stout, erect petioles, and laminas with a large number of primary veins, some of which originate from the lower half of the midrib . The remaining species have slender petioles and primary veins originating from the point of petiolar attachment . Limnobium differs from Hydrocharis by its rudimentary petals . Given the morphological similarity and small size of the genera, only two species in Limnobium and three species in Hydrocharis, the most reasonable taxonomic solution to the non-monophyly of Hydrocharis is to treat Limnobium as the synonym of Hydrocharis. Nomenclatural combinations already exist, i.e., Hydrocharis spongia Bosc and Hydrocharis laevigata (Willd.) Byng & Christenh.
Our time-calibrated tree indicated stem node age of about 53 Ma for Hydrocharis (Fig. 3), which is similar to the previously reported median age of 54.7 Ma  and in agreement with the oldest known Hydrocharis fossils from the Eocene . The crown age of Hydrocharis was estimated to be approximately 31 Ma, much older than the 15.9 Ma reported in Chen et al.  based on analysis that lacked H. chevalieri, the sister species of the remaining genus. Based on our results, Hydrocharis mainly diversified in the Miocene, which is consistent with the time inferred for many other extant plant species [57, 58]. Rapid climate change during the Miocene may have contributed to the speciation of Hydrocharis as well as extinctions, given that the fossil diversity is high in comparison to the limited extant diversity [59, 60]. The recent analysis of the distribution patterns of Hydrocharis indicated that the mean annual temperature is the main factor to impact the distribution in this genus . Furthermore, paleoclimatic changes may have induced such changes in water bodies that some groups were driven to extinction . This might explain why there are only three extant species, and a majority of fossil groups only occurred in Europe during the Miocene .
The area of origin remained uncertain for the genus because several alternative areas with low probabilities were recovered for the deeper nodes (Fig. 3). This may, of course, also reflect a widespread ancestral distribution. However, our model with the highest probabilities for the ancestor of the genus in Europe, Central Asia, and Central African regions (Fig. 3) contrasted the biogeographical model of Chen et al. , which suggested origination in the Oriental area. Additionally, our results indicated that at least three vicariance and two dispersal events had shaped the current distribution of the genus (Fig. 3). The Bering Land Bridge (BLB) has been proved to play an important role in the dispersal of different plant lineages between Asia and America [62–64]. The former genus Limnobium, only recorded in America, may have dispersed to America through the BLB during the Miocene [64, 65].