To reconstruct the phylogenetic relationship of Phtheirospermum, 68 accessions from Orobanchaceae (including Rehmannia) were sampled, representing all five recognized tribes, 22 genera and 30 taxa (Supplementary Table S1). For Ph. japonicum, we sampled nine individuals from nine populations in Yunnan, Sichuan, Henan, Zhejiang, and Liaoning Provinces in China, and from Kanagawa Prefecture in Japan. We sampled 13 individuals of Ph. tenuisectum, three individuals of Ph. muliense, and four and eight individuals of two undescribed taxa, respectively. Geographic sampling of the Ph. tenuisectum complex is presented in Figure 2. We sampled the three recognized species of Pterygiella including seven individuals to test the relationship between the Ph. tenuisectum complex and Pterygiella. We included sequences for Ps. glandulosa, and the samples T8 (J. S. Ying 4144), L1 (S. G. Wu 3582), and H1 (L. Lu LJ377) from the study of Dong et al. (2013). Forty-nine samples were sequenced for this study. Thirty-six sequences of four DNA barcoding loci (nrITS, matK, rbcL, and trnH-psbA) from two Pedicularis samples (Yu et al., 2011) and seven Pterygiella samples (Dong et al., 2011) have been published, as well as two trnL-F sequences from Pedicularis (Yu et al., 2015a), and one from Pterygiella (Dong et al., 2011). Forty-six sequences extracted from seven published plastomes [Bartsia inaequalis Benth., Castilleja paramensis F. González & Pabón-Mora, Cistanche phelypaea (L.) Cout., Lathraea squamaria L., Phelipanche purpurea Soják, Schwalbea americana L., and Striga hermonthica (Delile) Benth.], and the available nrITS sequences, were included for phylogenetic analyses. Voucher information or GenBank accessions are presented in Supplementary Table S1.

Total genomic DNA was extracted from silica-gel-dried leaves using a modified CTAB protocol (Doyle and Doyle, 1987). We amplified and sequenced one nuclear (ribosomal internal transcribed spacer, nrITS) and seven plastid markers (matK, rbcL, rps2, rps16, trnK-matK, trnH-psbA, and trnL-F). There was no overlap between matK and the trnK-matK regions. Primer information is listed in Supplementary Table S2. Polymerase chain reaction (PCR) amplification conditions were: one cycle at 94°C for 3min; 35 cycles of 94°C for 45 s, 53–55°C for 60 s and 72°C for 60 s; followed by 72°C for 5 min (Yu et al., 2011). PCR products were purified using ExoSAP-IT (Affymetrix, Santa Clara, CA, United States). Sequencing reactions were performed using the ABI Prism BigDye Terminator Kits (Applied Biosystems, Inc.), following a modified protocol (Yu et al., 2011). Automated sequencing was performed on ABI 3730xl DNA sequencer (Applied Biosystems) in Kunming Institute of Botany, Chinese Academy of Sciences.

The newly obtained raw sequences were assembled and edited using Geneious (Biomatters, Inc.) version 7.1 (Kearse et al., 2012). Assembled sequence consensuses were automatically aligned using MAFFT version 7.2 (Katoh and Toh, 2010), then adjusted manually in Geneious. The aligned matrices were concatenated using SequenceMatrix version 1.73 (Vaidya et al., 2011). Sequence characteristics were calculated using MEGA version 6.0 (Tamura et al., 2013).

Maximum Likelihood (ML) and Bayesian Inference (BI) methods were used to reconstruct phylogenetic trees. The nrITS and the concatenated plastid datasets were analyzed separately. No nucleotide positions were excluded from analyses. The ML tree searches and bootstrap estimation of clade support were conducted with RAxML version 8.2.10 (Stamatakis et al., 2008). These analyses used the GTR substitution model with gamma-distributed rate heterogeneity among sites and the proportion of invariable sites estimated from the data. The concatenated plastid dataset was partitioned by gene. Support values for the node and clade were estimated from 1000 bootstrap replicates. Bootstrap supports (BS) ≥ 70 were considered well supported (Hillis and Bull, 1993). Partitioned BI analyses were performed using MrBayes version 3.2.6 (Ronquist and Huelsenbeck, 2003), with DNA substitution models selected for each gene partition by the Bayesian information criterion (BIC) using jModeltest version 2.1.10 (Guindon and Gascuel, 2003; Darriba et al., 2012). Markov Chain Monte Carlo (MCMC) analyses were run in MrBayes for 10,000,000 generations for each dataset, with two simultaneous runs, and each run comprising four incrementally heated chains. The BI analyses started with a random tree, and sampled every 1000 generations. The number of generations for the datasets was considered sufficient when the average standard deviation of split frequencies was lower than 0.005, and Potential Scale Reduction Factor (PSRF) of Convergence Diagnostic (Gelman and Rubin, 1992) was 1.00. The first 25% of the trees was discarded as burn-in, and the remaining trees were used to generate a majority-rule consensus tree. Clades recovering posterior probability values (PP) ≥ 0.95 were considered as well supported (Alfaro et al., 2003; Erixon et al., 2003; Kolaczkowski and Thornton, 2007). Both ML and BI analyses, as well as jModeltest, were performed at the CIPRES Science Gateway¹.

Phylogenetic incongruence between ML and BI analyses of each dataset was visually compared using TreeGraph version 2.12 (Stover and Muller, 2010). The incongruence length difference test (ILD) (Farris et al., 1995) was not used to assess topological conflict between the nuclear and the concatenated plastid datasets because this analysis has been shown to be misleading (Baker et al., 2011). We used a conservative PP ≥ 0.95 and BS ≥ 70 as thresholds for identifying strongly incongruent clades between nrITS and the plastid datasets. In addition, topological incongruence between nrITS and the plastid datasets was also investigated using the Shimodaira–Hasegawa (SH) test (Shimodaira and Hasegawa, 1999) and the approximately unbiased (AU) test (Shimodaira, 2002). Topologies were constrained using Mesquite version 3.2 (Maddison and Maddison, 2017). The SH and AU tests were performed using PAUP version 4.10 (Swofford, 2003).

The nrITS and the plastid datasets of the Ph. tenuisectum complex were combined to recover a phylogenetic network using SplitsTree version 4.14.3 (Huson and Bryant, 2006). Pterygiella samples were included as outgroups. The Neighbor-net model was chosen using the Kimura 2-parameter (K2P) distance and Ordinary Least Square Method. Bootstrap values for the respective splits were estimated from 1000 bootstrap replicates.

There is no reliable fossil in Orobanchaceae to calibrate the phylogeny; divergence times were indirectly estimated from calibrating fossils from other families of flowering plants (Bremer et al., 2004; Magallón et al., 2015; Wang et al., 2015; Wikström et al., 2015). Due to inconsistency in taxonomic sampling and marker choice in different studies, the mean crown ages of Orobanchaceae varied from 26.0 Mya (Wikström et al., 2015) to 74.54 Mya (Wang et al., 2015). To improve the molecular dating on Orobanchaceae and the Ph. tenuisectum complex, we selected 24 families and 102 genera from Lamiales, and Solanum tuberosum as the outgroup (Supplementary Table S3). Two chloroplast DNA regions, rps2, and trnK were concatenated for molecular dating. Dating analyses were conducted using Markov Chain Monte Carlo (MCMC) methods in BEAST version 2.4 (Bouckaert et al., 2014), which was performed at the CIPRES Science Gateway². For setting parameters of BEAUti, we chose “BEAST model test” for “Site model,” “Relaxed Clock Log Normal” for “Clock model,” and “Yule Model” for speciation. Meanwhile, we selected four fossils to calibrate stem nodes of four families using CladeAge package (Matschiner et al., 2017). Of them, Acanthaceae (Acanthus rugatus Reid & Chandler), Bignoniaceae (Radermachera pulchra Reid & Chandler), Lamiaceae (Melissa parva Reid & Chandler) have reliable fossils from the Bembridge Flora (England) (Reid and Chandler, 1926) of the Late Eocene (Collinson et al., 2010), and Oleaceae (Fraxinus wilcoxiana Berry) from the Claiborne Formation of western Tennessee (United States) of the Middle Eocene (Call and Dilcher, 1992). The stem age of Acanthaceae (C4), Bignoniaceae (C3), and Lamiaceae (C5) was set as 33.9–37.8 Mya, and that of Oleaceae (C2) as 41.2–47.8 Mya. In addition, we used a normal model to constrain the crown age for Lamiales (C1) and Orobanchaceae (C6) as 84 ± 10 Mya and 56 ± 10 Mya, respectively. These ages were adopted from the TimeTree website³ (Hedges et al., 2006). Three independent MCMC runs with the same parameters were conducted. Each MCMC ran 100,000,000 generations, and was sampled every 10,000 generations. The first 5,000 generations were removed as “Pre Burnin.” Three Log outputs of the BEAST analyses were jointly evaluated using Tracer version 1.6. Effective sample sizes (ESS) of all parameters were more than 200, indicating that MCMC sampling was adequate (Lanfear et al., 2016). Trees of three runs were combined using LogCombiner by burning 20% of the initial trees. The Maximum Clade Credibility (MCC) tree was generated using TreeAnnotator by setting “Mean heights” for the “Node heights”. The MCC tree was visualized using FigTree version 1.4.2⁴.

To evaluate sequence matrix characteristics, we classified the samples of Phtheirospermum spp. and Pterygiella spp. in several groups. First, four major groups were adopted, i.e., Phtheirospermum, Pterygiella, the Ph. tenuisectum complex and the Ph. tenuisectum complex + Pterygiella. Then, every species with two or more samples in Phtheirospermum was treated as an independent group. Five species groups in Phtheirospermum were recognized, i.e., four groups in the Ph. tenuisectum complex, and Ph. japonicum.

Sequence characteristics of nrITS, seven plastid DNA loci, and the concatenated plastid datasets are presented in Table 1. The nrITS dataset was the most informative marker (except in the Ph. muliense and Phtheirospermum sp. 1), followed by two chloroplast intergenic spacers (trnH-psbA and trnL-F) and two introns (matK-trnK, and rps16). Of the three coding genes, matK was the most informative. Sequences of Ph. japonicum had no informative variation across seven plastid markers throughout nine populations. Only four deletions/insertions were found in the Japanese sample (J9), i.e., indels of a single nucleotide for rps16 and trnK, and two nucleotides in trnL-F. In contrast, the sequences of Ph. tenuisectum showed the highest variation at the species level across all eight DNA regions. The remaining three taxa of the Ph. tenuisectum complex were restricted to the Hengduan Mountains region, and showed fewer variations. Though 4.40% of nrITS sites of Pterygiella were variable, only 0 to 0.29% of the sites across seven plastid loci were variable.

Phylogenetic trees using nrITS and the plastid datasets are presented in Figures 3, 4, respectively. Pterygiella + Phtheirospermum + Xizangia (hereafter referred to as the Pterygiella group) was well supported as monophyletic (nrITS: BS/PP = 61/0.99; plastid: BS/PP = 100/1.00), and not nested in any other formerly recognized tribe. The plastid dataset supported Brandisia spp. as sister to this group (BS/PP = 69/0.96). None of the analyses resolved Pterygiella group close to tribe Rhinantheae.

Both nrITS and the plastid datasets recovered Ph. japonicum nested within Pedicularideae (nrITS and plastid: BS/PP = 100/1.00). The Ph. tenuisectum complex and Pterygiella spp. also formed a well-supported clade in both analyses (nrITS: BS/PP = 99/1.00; plastid: BS/PP = 100/1.00). The four morphotypes of the Ph. tenuisectum complex were recovered as reciprocally monophyletic, well-supported lineages. The Ph. tenuisectum complex was strongly supported as monophyletic by the plastid dataset (BS/PP = 99/1.00), and phylogenetic relationships for the four morphotypes were fully resolved (Figure 4). In contrast, Pterygiella spp. were nested within the Ph. tenuisectum complex in nrITS analyses (BS/PP = 100/1.00). In this analysis, Ph. tenuisectum was supported as sister to Phtheirospermum sp. 2 (BS/PP = 59/0.92), and Phtheirospermum sp. 1 and Pterygiella spp. formed a clade (BS/PP = 60/0.60).

Based on support values, we found hard topological conflicts between nrITS and the plastid datasets in the placement of tribes Orobancheae and Buchnereae. Additionally, these placements disagree with those recovered by McNeal et al. (2013). Further topological comparisons using SH and UA tests are presented in Table 2. In constrained analyses using nrITS dataset, both SH and UA tests indicated that the monophyly of the Ph. tenuisectum complex and Brandisia spp. sister to Pterygiella group were supported (P > 0.2). On the other hand, sister relationship between tribes Orobancheae and Buchnereae, and tribes Rhinantheae and Pedicularideae were rejected (P < 0.05). The constrained analyses using the plastid data rejected the paraphyly of the Ph. tenuisectum complex by the AU test (P < 0.05) and the sister relationship between tribes Buchnereae and Rhinantheae by both SH and UA tests. Therefore, nrITS and the plastid datasets should be analyzed separately.

To further explore phylogenetic incongruence within the Ph. tenuisectum complex, a phylogenetic network was inferred from the total dataset (Figure 5). A well supported split separated the Ph. tenuisectum complex from Pterygiella spp. (BS = 99). Each of the three species of Pterygiella were well resolved as monophyletic (BS = 100). The four morphotypes of the Ph. tenuisectum complex were also recovered as separate clusters. This pattern of relationships is similar to the plastid phylogeny, in which that Ph. tenuisectum was sister to Phtheirospermum sp. 2 + Ph. muliense (BS = 96). In addition, Phtheirospermum sp. 1 had connections with Pt. duclouxii + Pt. nigrescens (BS = 73), likely based on signal from the nrITS dataset.

The MCC tree is shown in Figure 6 and Supplementary Figure S1, including mean ages and 95% HPD interval bars. The exact values for six calibrated and 13 annotated nodes are summarized in Table 3. The mean substitution rate was 1.10 × 10^-3 per site per million years (95% HPD: 9.04 × 10^-4 - 1.29 × 10^-3). The Yule speciation rate was 4.65 × 10^-2 (95% HPD: 3.63 × 10^-2 - 5.67 × 10^-2). The coefficient of variation was 1.08 (95% HPD: 0.91 - 1.27), indicating that a high degree of rate heterogeneity was observed across the tree and that a relaxed-clock model was suitable for this dataset (Drummond et al., 2006).

Familial relationships in Lamiales were not fully resolved, and only some nodes were well supported (PP ≥ 0.95). Paulowniaceae and Phrymaceae were strongly supported as sister to Orobanchaceae (PP = 1.00). In Orobanchaceae, nine clades were well resolved (PP ≥ 0.99). Of these, seven clades have been recognized as tribes, and Brandisia as an independent lineage (Stevens, 2001 onward; McNeal et al., 2013). The Pterygiella group was recovered as monophyletic. Tribe Rehmannieae was the most basal clade, followed by splits marking the divergences of tribes Lindenbergieae and Buchnereae from the remainder of the lineage. The following splits mark the divergences of tribes Orobancheae and Cymbarieae successively, followed by Pedicularideae and Rhinantheae, leaving Brandisia and the Pterygiella group as a strongly supported clade (PP = 1.00). Brandisia was weakly supported as sister to the Pterygiella group. Within the Pterygiella group, the Ph. tenuisectum complex was strongly supported as monophyletic, and phylogenetic relationships of four morphotypes were well resolved.

The divergence time estimate for Orobanchaceae was in the early Eocene (54.56 Mya). Tribe Pedicularideae separated from tribe Rhinantheae – the Pterygiella group at the mid-Eocene (45.26 Mya), then Ph. japonicum diverged from the new world genera of tribe Pedicularideae at the late Oligocene (25.51 Mya), and its subsequent diversification occurred at the mid-Pleistocene (1.72 Mya). Diversification of the Pterygiella group was estimated to be around the early Miocene (19.64 Mya). The Ph. tenuisectum complex diverged from Pterygiella spp. during the later Miocene (7.52 Mya), with the four morphotype lineages fully established by the early Pleistocene (2.78 Mya).

The current delimitation of Orobanchaceae includes Rehmannia and Trianeophora as the basal tribe, i.e., Rehmannieae (Stevens, 2001 onward), which was strongly supported by phylogenetic analyses (Xia et al., 2009; Zeng et al., 2017). To date, seven tribes and Brandisia have been recognized in Orobanchaceae (Stevens, 2001 onward; McNeal et al., 2013). Although PHYA and PHYB nuclear data strongly supported Pterygiella as sister to the remaining genera of tribe Rhinantheae; both nrITS and the combined rps2 and matK datasets poorly supported Pterygiella as sister to Brandisia, and only the plastid dataset weakly supported the clade Pterygiella + Brandisia as sister to the remaining genera of tribe Rhinantheae (McNeal et al., 2013). In this study, both nrITS and the plastid datasets strongly resolved the Pterygiella group as monophyletic. The plastid dataset strongly supported Brandisia spp. as sister to the Pterygiella group; however, nrITS weakly resolved Brandisia as sister to a clade comprising tribes Buchnereae, Pedicularideae, and Rhinantheae. Therefore, Brandisia and the Pterygiella group should be treated as two separate groups or tribes in Orobanchaceae.

All phylogenetic analyses recovered polyphyly of Phtheirospermum, e.g., Ph. japonicum grouped with tribe Pedicularideae and the Ph. tenuisectum complex was close to Pterygiella spp. in the clade including Pseudobartsia and Xizangia (Figures 3, 4). This is not surprising, given that previous analyses place Ph. japonicum within Pedicularideae (Bennett and Mathews, 2006; McNeal et al., 2013; Yu et al., 2015a; Pinto-Carrasco et al., 2017). In this tribe, Pedicularis is the most basal group, followed by Ph. japonicum. The remaining genera are mainly distributed in the New World (Bennett and Mathews, 2006; McNeal et al., 2013), including Castilleja, Orthocarpus, and Triphysaria, another important model for parasitic plant genetics, genomics, and evolution (Tomilov et al., 2005; Westwood et al., 2010; Bandaranayake et al., 2012; Yang et al., 2015). The Ph. tenuisectum complex is clearly excluded from tribe Pedicularideae, forming a lineage including Pterygiella, Xizangia, and Pseudobartsia. The taxonomic confusion of Phtheirospermum can be ascribed to J. D. Hooker who described the second species Ph. parishii in Phtheirospermum, and transferred Euphrasia glandulosa Benth. to this genus (Hooker and Thomson, 1884). Based on the ellipsoid seeds and the minutely reticulated seed surface, C. B. Clarke suggested treating Ph. parishii as a separate genus (i.e., “Emmenospermum”), but Hooker did not adopt this treatment, because seed morphology of Scrophulariaceae was high variable within and among genera (Hooker and Thomson, 1884). Based on Hooker’s delimitation of Phtheirospermum, Ph. tenuisectum and Ph. muliense were placed in this genus (Bureau and Franchet, 1891; Tao, 1996). However, the phylogenetic analyses demonstrate that the Ph. tenuisectum complex should be separated from Ph. japonicum. Furthermore, this separation is justified by morphology; compared to Ph. japonicum, members of the Ph. tenuisectum complex bear linear rather than ovate to orbicular pinnae, smaller flowers with yellow rather than red corollas, and smaller fruits and seeds.

Though the Ph. tenuisectum complex was strongly supported as monophyletic by the plastid dataset (Figure 4), nrITS resolved Ph. tenuisectum complex as paraphyletic, with Pterygiella nested within it (Figure 3). Pinto-Carrasco et al. (2017) have combined the Ph. tenuisectum complex with Pterygiella in a single taxon, creating new combinations Pt. muliensis (C.Y.Wu & D.D.Tao) Pinto-Carrasco, E.Rico & M.M.Mart.Ort., Pt. parishii (Hook. f.) Pinto-Carrasco, E.Rico & M.M.Mart.Ort., and Pt. tenuisecta (Bureau and Franch.) Pinto-Carrasco, E.Rico & M.M.Mart.Ort.. This taxonomic decision was justified on the basis of phylogenetic results, and morphological evidence including five-toothed calyces and pollen characters. However, we suggest that the Ph. tenuisectum complex and Pterygiella should be retained as separated groups. First, in the constrained analyses, monophyly of the Ph. tenuisectum complex was not rejected (Table 2). Moreover, the Ph. tenuisectum complex is morphologically quite distinct from Pterygiella. Pterygiella spp. are characterized by winged stems (except Pt. cylindrica P. C. Tsoong), lanceolate and entire leaves, and by a calyx that is broadly campanulate and 5-veined (Supplementary Figure S2). Conversely, the Ph. tenuisectum complex has cylindrical stems, pinnatisect leaves, and a tubular calyx. In addition, seed morphology and capsule indumentum also differ between Pterygiella and the Ph. tenuisectum complex (Dong et al., 2013, 2015). Therefore, we suggest treating the Ph. tenuisectum complex as an independent genus. So far, there is no available name for this lineage. The name “Emmenospermum” cannot be selected because it would be a later homonym of Emmenosperma F. Mueller (Rhamnaceae). A new genus name needs to be published in accordance with the International Code of Nomenclature for algae, fungi, and plants (McNeill et al., 2012), including comprehensive morphological comparison with Pterygiella spp. and other relatives.

The genus Phtheirospermum was established on the basis of Ph. chinense Bunge, which was collected in north China (Fischer and Meyer, 1835). Both the genus and species names are ascribed to A.A. Bunge in F.E.L. Fischer’s and C.A. Meyer’s edited monograph, Index Seminum [St. Petersburg]. However, an early name Gerardia japonica Thunberg had been described based on a Japanese specimen. Therefore, Ph. japonicum is the correct name to replace Ph. chinense (Kanitz and Weiss, 1878). Our phylogenetic results showed the polyphyly of Phtheirospermum, the current delimitation of Phtheirospermum include Ph. japonicum alone, and Ph. tenuisectum complex needs to be transferred to a new genus. Currently, Ph. japonicum has been established as the experimental model for the study of genetics and development of the haustorium (Cui et al., 2016; Ishida et al., 2016). Maintaining the name Ph. japonicum will therefore benefit molecular biologists using this model in publications or communications with public audiences and plant specialists. This taxonomic decision would forestall confusion as has arisen in the nomenclature of the monkeyflower (Mimulus guttatus DC.), a model organism for studies of evolution and ecology that was shown to be polyphyletic. The correct name of the monkeyflower was changed to Erythranthe guttata (DC.) G. L. Nesom (Barker et al., 2012). However, some recent publications still use the old name M. guttatus (Lee et al., 2016; Ferris et al., 2017; Puzey et al., 2017).

Phtheirospermum japonicum is a widely distributed species in Eastern Asia (Hong et al., 1998). In this study, we sampled nine individuals from southwestern (Yunnan: J2, J3, J7, and J8; Sichuan: J6), through central China (Henan: J5), to eastern (Zhejiang: J1) and northeastern China (Liaoning: J4), and extending to Japan (Kanagawa: J9). Both nrITS and the plastid evidence indicated that genetic variation in Ph. japonicum was low, and current distribution range might be a result of a recent population expansion. The results of molecular dating showed that a recent radiation occurred at around 1.72 Mya (95% HPD: 0.29–3.88 Mya), and the Japanese sample was derived from a Yunnan sample (J2) at around 0.23 Mya (95% HPD: 0.0–0.73 Mya).

The Ph. tenuisectum complex can be classified as four species lineages, each strongly supported by both nrITS and the plastid datasets (Figures 3, 4), and diagnosable by morphological characters (Figure 1). Phylogenetic relationships among the four lineages were inconsistent between nrITS and the plastid datasets. It is likely that because resolution of the nrITS tree was low, constrained analyses using nrITS dataset did not reject the constrained plastid topology (Table 2). The plastid dataset strongly supported that Phtheirospermum sp. 1 was the sister to the three other lineages, in which Phtheirospermum sp. 2 was sister to Ph. muliense. As an ancestral lineage in the Ph. tenuisectum complex, Phtheirospermum sp. 1 might share plesiomorphic characters with Pterygiella spp. (i.e., incomplete lineage sorting), so that the nrITS dataset weakly supported Pterygiella spp. as sister to Phtheirospermum sp. 1 (Figure 3), and phylogenetic network showed a well-supported genetic connection between Phtheirospermum sp. 1 and Pt. duclouxii + Pt. nigrescens (Figure 5). Meanwhile, ancient hybridization/introgression cannot be excluded between Pterygiella spp. and Phtheirospermum sp. 1. In addition, ancient hybridization/introgression and incomplete lineage sorting may also have occurred between Phtheirospermum sp. 2 and Ph. tenuisectum.

The last two uplifts of the Qinghai–Tibet Plateau happened at around 8.0–7.0 Mya and 3.6–1.5 Mya (Zheng et al., 2000; Garzione et al., 2003; Molnar, 2005; Ge et al., 2006; Li et al., 2015; but see Renner, 2016). Many species groups diversified rapidly with these uplifts in the Hengduan Mountains (Wen et al., 2014; Xing and Ree, 2017). In this study, the estimated ages for the origin of Ph. tenuisectum complex and its subsequent divergences fell into the range of the two uplifts (Table 3 and Figure 6). The uplift during 8.0–7.0 Mya may have driven divergence of the Ph. tenuisectum complex from Pterygiella. The center of diversity for both the Ph. tenuisectum complex and Pterygiella is in northwestern Yunnan and southwestern Sichuan, perhaps the place of origin of the Ph. tenuisectum complex. Overlapping species distributions among these genera out of the Hengduan Mountains may be the results of later population expansion. Of the four lineages in the Ph. tenuisectum complex, Ph. muliense, Phtheirospermum sp. 1, and Phtheirospermum sp. 2 grow along/near the dry valleys of Jinsha River and its tributaries, with only the sample H5 of Phtheirospermum sp. 2 extend to Lancang River (Figure 2). In contrast, Ph. tenuisectum mainly occurs in meadows and has a wider distribution. The valleys of Jinshan and Lancang Rivers and their tributaries in the Hengduan Mountains region experience a specialized dry-hot valley climate (Yang et al., 2016), which may have contributed to divergence of Ph. muliense, Phtheirospermum sp. 1, and Phtheirospermum sp. 2. To address the evolutionary history of Ph. tenuisectum complex in the future, a phylogeographic approach may be applied using more dense population samplings and multiple individuals per population.