Size, content, and arrangement of plastomes
After the low-quality reads and adaptor sequences were filtered out, 20,453,560–28,744,390 clean reads were obtained for six Wisteria and Millettia japonica. The NOVOPlasty assembly produced a long contig representing the whole plastome. These seven newly sequenced plastomes ranged from 130,116 to 132,547 bp in length, and the GC contents were from 34.20% to 34.50% (Fig. S1; Table 1). All plastomes contained 110 unique genes arranged in same order, with 76 protein-coding genes, 30 tRNA genes, and four rRNA genes (Table 1), and there was a single inverted repeat. In all seven sequences, ten protein-coding genes (atpF, clpP, ndhA, ndhB, petB, petD, rps12, rpl2, rpl16, rpoC1) and six tRNA genes (trnA-UGC, trnG-UCC, trnI-GAU, trnK-UUU, trnL-UAA, trnV-UAC) possessed a single intron, whereas two genes (rps12, and ycf3) contained two introns (Table S3). All seven sequenced species had one intron in clpP gene, which was not absent in Glycyrrhiza glabra. For rps12 gene, Wisteria had 2 introns, whereas G. glabra just had one. All seven plastome sequences have been submitted to the GenBank (accession numbers: MN311163, MN311167-MN311171; Table 1).
Gene rearrangements were not found in Glycyrrhiza glabra, Millettia japonica, Wisteria sinensis, Caragana kozlowii Kom., Carmichaelia australis R. Br., Astragalus mongholicus Bunge, Cicer arietinum L., or Medicago truncatula Gaertn.. On the contrary, large segments of the plastomes were either reversed and lost in Trifolium subterraneum L., Pisum sativum L., Lens culinaris Medik., Lathyrus sativus L. and Vicia faba L., (Fig. 1). In addition, plastome structure did not change among species of Wisteria (Fig. S2). The MISA analysis identified a total of 540 SSRs across the six Wisteria plastomes. The number of SSRs ranged from 81 (W. villosa) to 99 (W. floribunda). Mono-nucleotide SSRs (A/T/G) took the largest percentage, followed by di-nucleotide SSRs (AG/AT/TA/TC) (Fig. 2A). The combined total of tri-, tetra-, penta- and hexa-nucleotide SSRs was no more than 15%. Besides, only W. floribunda and W. venusta contained hexa-nucleotide SSRs, and W. frutescens did not have penta-nucleotide SSRs (Table S4).
The forward, palindromic and reverse repeats were detected in W. floribunda, W. venusta and W. villosa, while only forward and palindromic repeats were found in W. brachybotrys, W. frutescens and W. sinensis (Fig. 2B, Table S4). The total number of repeats ranged from 43 (W. villosa) to 70 (W. venusta) and most repeats were in no-coding regions (82.09%-87.14%, except 65.85% for W. villosa). Among the six plastomes of Wisteria, the proportion of forward repetition was the highest, ranging from 31 in W. villosa to 65 found in W. venusta (Fig. 2C, Table S5).
In Wisteria, nucleotide diversities (Pi) of all genes, intergenic spacers, and introns ranged from 0.0008 (ndhD) to 0.04648 (trnN-ycf1) (Table S6). There were seven hypervariable regions with Pi greater than 0.025 including trnN-ycf1, rpl33-rps18, trnS-trnG, ndhF-trnH, rps18, clpP-psbB, and rps8-rpl14 (Fig. 3).
Phylogenetic relationships
The matrix of 28 plastome CDS sequences was 71,545 bp in length. In both the ML and BI trees Glycyrrhiza glabra, Millettia japonica and Wisteria formed a clade (BS = 90, PP = 0.5), which was sister to the remaining genera. M. japonica was the closest taxon to Wisteria (BS = 1, PP = 1). Caragana was sister to (Astragalus + Carmichaelia), they then were sister to the clade containing Cicer L., Medicago, Trifolium, Pisum L., Lathyrus, Lens Mill. and Vicia (BS = 1, PP = 1).
The interspecific relationships of Wisteria were well resolved (BS = 100, PP = 1 for all nodes, Fig. 4). Wisteria frutescens was sister to the clade of Asian species (BS = 1, PP = 1). There were two subclades within the Asian clade (BS = 1, PP = 1): W. brachybotrys and W. venusta, and the other three species. Two individuals of W. floribunda formed a clade, which was sister of W. sinensis and W. villosa (BS = 1, PP = 1). However, relationship between W. villosa and W. sinensis was still unresolved.
Substitution rates
The mean dS of SC genes was 0.022 ± 0.001 in the IRLC pairs and 0.009 ± 0.000 in the non-IRLC pairs. For IR genes, they were 0.014 ± 0.003 and 0.002 ± 0.000 respectively. All herbaceous pairs had significant higher dS than the woody pairs (P < 0.01). The herbaceous pairs in the IRLC had 2.07-fold dS in SC region than woody pairs, whose average dS of the SC genes was 0.010 ± 0.000 (Table S2).
Divergence time estimation
The estimated divergence time of IRLC was 40.11 (35.73–44.36) Ma (node c, Fig. 5, Table 2). The stem age and crown age of Wisteria were 23.85 (23.26–24.86) Ma and 8.66 (2.40–17.38) Ma, respectively (node d, node 2). Then, the clade of W. brachybotrys and W. venusta was separated from other Asian wisteria at 2.22 (0.75–5.67) Ma (node 3), and the split of these two species at 0.06 (0.00–0.23) Ma (node 6). The crown age of W. floribunda, W. sinensis and W. villosa was estimated to be 0.87 (0.25–2.15) Ma (node 4). The crown ages of other genera in IRLC were listed in Table 2.
Ancestral area reconstruction
The result of BMM analysis indicated that the most recent common ancestor (MRCA) of Wisteria was distributed in China with 33.57% probability, in Japan with 33.22% probability, meanwhile, in eastern North America with 12.22% probability (Fig. 5). There may be a dispersal event from China to eastern North America, then followed by a vicariance event, with the probability of 12.78%. For Asian wisterias, the ancestral range was indistinct, with 37.09% chance in China, a 31.18% chance in Japan and 31.43% probability in China and Japan. Vicariance events were also identified between W. brachybotrys and W. venusta with 37.83% probability, between W. floribunda and W. sinensis-W. villosa with 34.26% probability.