The taxonomic identification of Physospermopsis species was by field investigations and specimen examinations from herbaria BM, BNU, CDBI, CVH, E, HITBC, ILL, K, KUN, LBG, LE, MW, NAS, NHW, P, PE, PEY, SABG, SM, SZ, UC, WU, WUK.

In the field investigations, we sampled three populations of P. delavayi, two populations of P. rubrinervis, two populations of P. shaniana, one population of P. obtusiuscula (1938: 231) and one population of P. nana (2000: 538) in Yunnan Province. We sampled one population of P. kingdon-wardii and one population of P. obtusiuscula in Tibet. One population of P. alepidioides (H.Wolff & Hand.-Mazz.) R.H.Shan (1941: 187) and one population of P. muliensis were sampled in Sichuan Province. All populations were collected from the type locality and adjacent regions, and the features of every species were closely matched with the types and original descriptions (de Candolle 1830; Franchet 1894; Diels 1912; Wolff 1929; Wolff et Handel-Mazzetti 1933; Shan et Liou 1979). The specific collection information are listed in Appendix 1.

Fruits, leaf segments and specimens from these eight species of Physospermopsis were collected in the field for morphological study. Morphological analyses of leaves and fruits based on herbarium specimens or formaldehyde-acetic acid-alcohol (FAA) preserved material were photographed by a stereomicroscope Nikon SMZ25 (Japan). The morphological data were measured using KaryoType (Altnordu et al. 2016).

We sampled 13 populations, representing eight species of Physospermopsis in our phylogenetic analysis, and obtained 31 sequences of other Apiaceae species from the NCBI (Appendix 1). Based on previous research (Zhou et al. 2009), Bupleurum krylovianum B.K.Schischkin (1935: 2010) and Bupleurum rockii H. Wolff (1929: 187) were selected as the outgroup for studying the phylogenetic position of Physospermopsis. We chose Pl. franchetianum W.B.Hemsley (1892: 307) and Pl. wrightianum H.Boissieu (1903: 847) as the outgroup for studying interspecific relationships within Physospermopsis. The DNA sequences of two chloroplast loci (rpl16, rps16) and one nuclear region, the internal transcribed spacers of ribosomal DNA (ITS), were used for phylogenetic analyses. According to the research to date (Zhou et al. 2008, 2009; Downie et al. 2010; Guo et al. 2018; Panahi et al. 2018), these three markers should be sufficient to obtain the general information about relationships within the genus and its phylogenetic position within the family Apiaceae.

The fresh leaves were collected from field specimens in Yunnan, Sichuan and Tibet, China. Voucher specimens were deposited in the Herbarium of Sichuan University (SZ). Total genomic DNA was extracted from silica-dried leaves with plant genomic DNA kit (Cwbio Biosciences, Beijing, China). The universal primers ITS4 (5’-TCC TCCGCT TAT TGA TAT GC-3’) and ITS5(5’-GGA AGT AAA AGT CGT AAC AAG G-3’; White et al. 1990) were used to amplify the entire internal transcribed sequences. The rpl16 intron region was amplified using primers F71(5’-GCT ATG CTT AGT GTG TGA CTC GTT G-3’) and R1516 (5’-CCC TTC ATT CTT CTA TGT TG-3’) (Jordan et al. 1996; Kelchner and Clark 1997). The rps16 sequences were amplified with primers rps16 3’exon (5’-CCT GTA GGY TGN GCN CCY TT-3’) and rps16 5’exon (5’-AAA CGA TGT GGN AGN AAR CA-3’)(Downie and Katz-Downie 1999). PCR amplification was implemented in a 30 μL volume reaction, including 3 μL total DNA, 1.5 μL forward primer, 1.5 μL reverse primer, 15 μL 2×Taq MasterMix (Cwbio, Beijing, China), and 9 μL ddH2O. The amplification of the ITS region was obtained by initial denaturation for 3 min at 94 °C, followed by 30 cycles of 45 s at 94 °C, 70 s at 54 °C, and 90 s at 72 °C, then final extension of 10 min at 72 °C. Amplification of cpDNA intron regions was obtained by initial denaturation for 3 min at 94 °C, followed by 36 cycles of 45 s at 94 °C,70 s at 58.5 °C, and 90 s at 72 °C, then final extension of 10 min at 72 °C. All PCR products were separated using a 1.5% (w/v) agarose TAE gel and sent to Sangon (Shanghai, China) for sequencing. New sequences obtained for this study have been deposited in GenBank. GenBank accession numbers are provided in the Appendix 1.

We used SegMan7 (Burland 2000) to assemble ITS and cpDNA sequences. ClustalX (Jeanmougin et al. 1998) was used to align DNA sequences with manual adjustment. We then used MEGA7 (Kumar et al. 2016) to manually adjust and obtain ITS and cpDNA datasets. Gaps were positioned to minimize nucleotide mismatches. Bayesian inference (BI) and Maximum likelihood (ML) methods were used for phylogenetic analyses, using MrBayes v3.2 (Ronquist et al. 2012) and RAxML v8.2.4 (Stamatakis 2014), respectively. Before undertaking BI analyses, MrModeltest version 2.2 (Nylander 2004) was used to determine the best model of nucleotide substitution and the GTR+G model under the Akaike Information Criterion (Akaike 1974) was selected. Bayesian analyses were performed over 20 million generations with a variant of Markov Chain Monte Carlo (MCMC) method and the trees were saved to a file every 1,000 generations. The first 20% trees were discarded as “burn-in” and the remaining 80% trees were used to build a majority-rule consensus tree based on analysis of the program Tracer v1.4 (Drummond and Rambaut 2007). ML analyses were performed using RAxML v8.2.4 with the GTR+G model and 1,000 bootstrap replicates. We constructed the BI tree with ITS data from all 44 taxa to test the systematic position of Physospermopsis. And we mapped some valuable morphological characteristics of Physospermopsis on phylogenetic tree, including leaves, bracts and bracteoles, ribs of fruits. The BI and ML trees were constructed for analysis of interspecific relationships within Physospermopsis using ITS and plastid datasets from the 13 Physospermopsis populations we sampled, one Physospermopsis species and the two Pleurospermum species downloaded from NCBI. Detailed information on the investigated taxa can be found in the Appendix 1.

Through observations in the field, the most important characteristic to identify Physospermopsis species was prominent bracts and bracteoles. Physospermopsis shaniana, P. nana, P. muliensis, P. rubrinervis, P. obtusiuscula and P. kingdon-wardii usually have leaf-like bracts and bracteoles (Fig. 1A3–F3). While P. alepidioides and P. delavayi possess lanceolate or oblong bracts and bracteoles with a 2–3-lobed apex and dark purple margin (Fig. 1G3, H3). Furthermore, leaf shape varies with species and can be obovate-lanceolate (e.g. P. alepidioides), triangular (e.g. P. rubrinervis), obovate-orbicular (e.g. P. delavayi) or linear-lanceolate (e.g. P. nana) segments (Fig. 1B2, D2, G2, H2). Besides, the leaves of P. kingdon-wardii and P. obtusiuscula are 2-pinnate and ovate-oblong, and have 2–6 pairs of ovate pinnae with pinnatisect margin. Physospermopsis muliensis and P. shaniana possess 3–5 pairs pinnae with pinnatifid margin, narrowly winged petioles and narrow and purple-red sheaths. Fruit morphology was recorded prior to alcohol preservation because the alcohol altered the color slightly (as is seen in photographs). The fruits of Physospermopsis were emerald green or chartreuse, ovoid to broadly ovoid, and typically had a slightly cordate base, a gradually narrowed and laterally flattened apex, with filiform or prominent ribs. Fruit shape and size of all Physospermopsis species were similar except that P. kingdon-wardii had fruit half the size of other species and very prominent and sinuate ribs. Physospermopsis nana and P. muliensis fruits had relatively prominent and filiform ribs, but P. muliensis fruits had scattered warts especially on the ribs and P. nana had smaller fruit. Physospermopsis obtusiuscula fruits were ovoid with narrowly winged and sinuolate ribs. The fruit of P. delavayi had an obvious cordate base, and filiform and less prominent ribs. Physospermopsis alepidioides, P. rubrinervis and P. shaniana had ovoid, verucose fruits with prominent ribs, but P. alepidioides did not have a cordate base, while the other two species had a slightly cordate base. The fruit of P. shaniana had many small warts distinguishing it from P. rubrinervis. For easy reading and comparison, the main morphological characteristics were listed in Table 1.

Figure 1. 

Morphological characters of Physospermopsis A1–H1 habit A2–H2 basal leaf A3–H3 umbel A4–H4 mericarps A1 habit of P. shaniana B1 habit of P. nana C1 habit of P. muliensis D1 habit of P. rubrinervis E1 habit of P. obtusiuscula F1 habit of P. kingdon-wardii G1 habit of P. delavayi H1 habit of P. alepidioides A2 basal leaf of P. shaniana B2 basal leaf of P. nana C2 basal leaf of P. muliensis D2 basal leaf of P. rubrinervis E2 basal leaf of P. obtusiuscula F2 basal leaf of P. kingdon-wardii G2 basal leaf of P. delavayi H2 basal leaf of P. alepidioides A3 umbel of P. shaniana B3 umbel of P. nana C3 umbel of P. muliensis D3 umbel of P. rubrinervis E3 umbel of P. obtusiuscula F3 umbel of P. kingdon-wardii G3 umbel of P. delavayi H3 umbel of P. alepidioides A4 mericarps of P. shaniana B4 mericarps of P. nana C4 mericarps of P. muliensis D4 mericarps of P. rubrinervis E4 mericarps of P. obtusiuscula F4 mericarps of P. kingdon-wardii G4 mericarps of P. delavayi H4 mericarps of P. alepidioides.

Through comprehensive sampling, the ITS analyses indicated that the 13 populations of Physospermopsis we sampled and P. muktinathensis M.A.Farille & S.B.Malla (1985: 512) formed an individual clade. Physospermopsis was confirmed to be a monophyletic group and nested in Pleurospermeae. Trachydium roylei Lindl. (1835: 232) and Pl. wilsonii H.Boissieu (1906: 433) were the closest relatives of Physospermopsis (Fig. 2).

Figure 2. 

Bayesian tree inferred from the analysis of the 44 samples of ITS data. Branch lengths are proportional to the amount of character changes, scale = 0.02 substitutions per character. The tree is rooted with Bupleurum. The names of the clades identified are those of Zhou et al. (2008, 2009).

The ITS dataset tree topologies generated from BI and ML analyses were consistent. Therefore, only the BI tree with posterior probabilities (PP, 0–1) and bootstrap support values (BS, 0–100%) is illustrated in Fig. 3A. The first to differentiate from Physospermopsis was P. muktinathensis, which is distributed in Nepal. Three populations of P. delavayi and one of P. alepidioides united as a strongly supported (BI–PP = 1; ML–BS = 100%) group. Physospermopsis obtusiuscula was supported as a sister group to P. kingdon-wardii (BI–PP = 1; ML–BS = 100%). Physospermopsis rubrinervis, P. muliensis, P. nana and P. shaniana were allied in all trees (BI–PP = 1; ML–BS = 100%). However, clear interspecific relationships between P. rubrinervis, P. muliensis, P. nana and P. shaniana were not strongly supported by ML or BI analyses.

Figure 3. 

Bayesian trees of Physospermopsis and its related genus inferred from ITS (A) and plastid rpl16+rps16 (B). Values on the branches indicate their support (Bayesian posterior probability/ Maximum-likelihood bootstrap). Branch lengths are proportional to the amount of character changes, scale = 0.01 (A), 0.005 (B) substitutions per character.

The cpDNA dataset tree topologies inferred by BI and ML analyses were consistent (Fig. 3B). However, results of a partition homogeneity test for the ITS and cpDNA datasets indicated that these genomes provide significantly different phylogenetic estimates. The taxa involved in this conflict are highlighted in Fig. 3. There was no chloroplast data for P. muktinathensis. The first to differentiate were P. obtusiuscula and P. kingdon-wardii (BI–PP = 1; ML–BS = 99%). The relationships of the three P. delavayi populations differed from the ITS dataset tree topology, although this cpDNA dataset relationship was not strongly supported. The cpDNA dataset tree topologies indicated that LYG population was closer to the HB population (BI–PP = 0.33; ML–BS = 47%), while LYG was closer to LGH in the ITS dataset tree topologies (BI–PP = 0.34; ML–BS = 70%). Additionally, the relationships between P. rubrinervis, P. muliensis, P. nana were not consistent with the ITS tree, where P. nana allied with P. rubrinervis in the cpDNA tree (BI–PP = 1; ML–BS = 72%), whereas P. nana allied with P. muliensis in the ITS tree (BI–PP = 1; ML–BS = 94%).

Physospermopsis is monophyletic. The reasons for previous designations as a polyphyletic genus were likely attributable to the misidentification of several species (e.g. P. rubrinervis, P. kingdon-wardii, P. cuneata). Besides, P. cuneata is a poorly known species and unusual within the genus for its lack of conspicuous bracts and bracteoles, and therefore the phylogenetic placement of it is highly controversial. However, the most recent consensus is that P. cuneata should not be placed in Physospermopsis (Zhou et al. 2008; Zhou et al. 2009; Pimenov 2017). So previous molecular studies only involved five physospermopsis species which were widely accepted; we added another three physospermopsis species in this study, including P. alepidioides, P. obtusiuscula, and P. nana. Evidence obtained through more precise checking of generic type, infrageneric types and extensive herbarium specimens, literature and field investigations, analyzing morphological characters, and ITS and cpDNA evidence. This comprehensiveness allows us to be confident that Physospermopsis is monophyletic and nested in Pleurospermeae. In addition, we propose that the Physospermopsis clade should be replaced by the East Asia Clade.

The molecular results indicated that Physospermopsis is closest to Pleurospermum. Morphologically, Pleurospermum usually possess numerous bracts and bracteoles with white scarious margins, conspicuous or obsolete calyx teeth, white or purple-red petals with clawed base and narrow apex, prominent, acute ribs (Pan and Watson 2005). However, we found that Physospermopsis differed from Pleurospermum by less prominent and even inconspicuous fruit ribs, and the bracts and bracteoles did not have white scarious margins, resulting in an obvious, diagnostic boundary between Pleurospermum and Physospermopsis. The closeness of the two genera is also evidenced in pollen morphology. Wang and Pu (1992) found P. alepidioides and P. muliensis pollen to be rhomboidal and similar to several Pleurospermum species whereas other Physospermopsis species (P. rubrinervis and P. delavayi) have more advanced rectangular types. In addition, Pan and Watson (2005) identified several Physospermopsis species (e.g. P. obtusiuscula) with morphological similarities to Pleurospermum species, including having long fruit ribs and bulgy fruit walls, while other species had flattening of fruit and reduced wall thickness. Consequently, Physospermopsis is sister to Pleurospermum.

The morphological characteristics mapped on the phylogenetic tree indicated that most closely related species have similar morphological characteristics. For example, P. rubrinervis, P. muliensis and P. shaniana are highly consistent on leaves, bracts and bracteoles, ribs on fruits (Fig. 2). Similarly, these species are the geographically sympatric species (Fig. 4). Resolution of the relationships between these species will only be achieved through continued studies, which may be difficult due to their geographic and morphological similarities. However, we can learn that P. nana, P. rubrinervis, P. muliensis and P. shaniana are the more advanced species in Physospermopsis. The morphological characters of P. nana are the most particular; these might be caused by hybridization with Pleurospermum species.

Figure 4. 

Geographic distribution of the eight Chinese Physospermopsis species in China. The altitude, scale, name of provinces and provincial capitals are also showed on the map.

The interspecific relationships between certain species within Physospermopsis are evident based on the consistencies between ITS and cpDNA trees. For instance, P. alepidioides showed a close affinity to P. delavayi in phylogenetic tree, and they have similar bracts and bracteoles (entire or 2–3-lobed at apex, with dark purple margin) (Figs 1, 2). However, differing leaf shapes can be used to easily distinguish these two species because P. alepidioides has an undivided leaf with sparsely serrated margin (Fig. 1H2) and P. delavayi has a pinnate leaf (Fig. 1G2). Physospermopsis kingdon-wardii is sister to P. obtusiuscula, which is consistent with their geographic closeness. Physospermopsis kingdon-wardii appears more morphologically similar to P. obtusiuscula (including the leaves, bracts and bracteoles), but differs in its reduced stem, small stature and small fruits with prominent and sinuate ribs (Fig. 1).

The topologies of the ITS and cpDNA trees differed in the positioning of P. delavayi, P. muliensis and P. nana (Fig. 3). This inconsistency between nrDNA ITS and cpDNA data has been reported in some studies of Apiaceae (Lee and Downie 2006; Zhou et al. 2008, 2009; Spalik et al. 2009; Bone et al. 2011; Yi et al. 2015; Panahi et al. 2018). This difference generally has been caused by incomplete lineage sorting, hybridization, homoplastic substitutions and introgression. Since we did not sample by lineage and execute gene flow analysis, what caused the inconsistency cannot be determined. Previous studies have indicated that Pleurospermeae occupies a relative position in the base of the Apioideae (Zhou et al. 2008, 2009; Downie et al. 2010), the differentiation time should be earlier. Thus, for Physospermopsis, we infer the more effective reason for the inconsistency between nrDNA ITS and cpDNA data is hybridization. A further study based on widely sampling and deeper analysisis needed. However, several diagnostic characteristics can be utilized in the field and laboratory to separate them. P. rubrinervis can easily be recognized by purple-red nerves on the leaves, bracts and bracteoles with purple-red margin (Fig. 1). Physospermopsis muliensis possesses a slender, branched stem and narrowly winged basal petioles with narrow sheaths (Fig. 1). Physospermopsis nana has bracts and bracteoles with white scarious margins and linear-lanceolate segments with membranous-margined sheaths (Fig. 1). The stem of P. shaniana was reduced and branched at the base, and had prominent bracts 1–2-pinnate with developed, broad sheaths (Fig. 1).