Botanical Journal of the Linnean Society, 2012, 168, 43–63. With 9 figures

A combined morphological and molecular phylogenetic analysis of () and taxonomic implications

LIMIN LU1,2, JUN WEN1,3* and ZHIDUAN CHEN1*

1State Key Laboratory of Systematic and Evolutionary Botany, Institute of Botany, the Chinese Academy of Sciences, Beijing 100093, China 2Graduate School of Chinese Academy of Sciences, Beijing 100039, China 3Department of Botany, National Museum of Natural History, MRC166, Smithsonian Institution, Washington, D.C. 20013-7012, USA

Received 21 June 2010; revised 11 March 2011; accepted for publication 18 August 2011

Parthenocissus (the Virginia creeper genus, Vitaceae) consists of 13 species and shows a disjunct distribution between Asia and North America. We investigated the inflorescence structure, calyx morphology, appendages on the inner side of petals, leaf epidermis, pollen and seed characters throughout the genus. A combined phylogenetic analysis with 27 morphological and 4137 molecular characters was conducted and the result was largely congruent with that of the previous molecular work, but with higher resolution. The combined analysis identified two clades corresponding to the Asian and North American taxa. Parthenocissus feddei was resolved as closely related to the clade containing P. cuspidifera, P. heterophylla and P. semicordata. The four species share synapomorphies of having conspicuously raised veinlets, an obscurely five- (to eight-) lobed calyx, appendages on the inner side of petals covering the entire length of anthers and foveolate pollen exine ornamentation. Within the Old World clade, the pentafoliolate species were weakly supported as more closely related to species with both simple and trifoliolate leaves. Furthermore, the ancestral states of tendril apices, inflorescence structure, appendages on the inner side of petals and exine ornamentation of pollen grains were reconstructed on the molecular strict consensus tree. The appendages on the inner side of petals and exine ornamentation of pollen grains were suggested to be important characters in the of Parthenocissus, especially for species with trifoliolate leaves. Finally, the previous classifications of Parthenocissus were evaluated within the phylogenetic framework. © 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63.

ADDITIONAL KEYWORDS: morphology – phylogeny – seed morphology.

INTRODUCTION Java to northern Thailand and three in North America (Soejima & Wen, 2006; Wen, 2007; Chen, Parthenocissus Planch. (the Virginia creeper genus, Ren & Wen, 2007). Vitaceae) consists of c. 13 species and shows a dis- Parthenocissus can be easily distinguished from junct distribution between Asia and North America the other genera of Vitaceae by its highly branched (extending to Mexico and the Caribbean region). tendrils, adhesive discs at tendril apices, inconspicu- There are approximately ten Old World species dis- ous floral discs and two long ventral infolds extending tributed primarily in eastern Asia, with one species in from the apex to the base of the seed (Brizicky, 1965; the western Ghats of India and Sri Lanka, one in Li, 1998; Chen et al., 2007; Wen, 2007). Li (1990) segregated Parthenocissus thomsoni (M.Lawson) Planch. and P. austro-orientalis Metcalf from Par- *Corresponding authors. E-mail: Jun Wen, [email protected]; thenocissus and established the new genus Yua Zhiduan Chen, [email protected] C.L.Li based on their bifurcate tendrils, leaf-opposed

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 43 44 L. LU ET AL.

Table 1. Comparison of two classification systems of Parthenocissus by Galet (1967) and Li (1996). Names in parentheses indicate synonyms

Galet (1967) Li (1996)

Ser. Trifoliolae Galet Sect. Parthenocissus Ser. Trifoliolae Galet P. cuspidifera P. chinensis P. dalzielii P. cuspidifera P. heterophylla (P. neilgherriensis) P. feddei P. semicordata (P. anamallayana, P. himalayana) P. heterophylla Ser. Tricuspidatae Galet P. semicordata (P. cuspidifera var. pubifolia) Ser. Parthenocissus P. suberosa P. quinquefolia P. tricuspidata Sect. Margaritaceae C.L.Li Ser. Heterophyllae C.L.Li Ser. Quinquefoliolae Galet P. dalzielii Ser. Tricuspidatae Galet P. henryana P. suberosa P. heptaphylla P. tricuspidata P. laetevirens Sect. Tuberculiformes C.L.Li P. quinquefolia P. henryana P. vitacea P. laetevirens

compound dichasium and ventral infolds extending recognized nine species and one variety of Partheno- upward two-thirds from the base of the seed. The cissus in China: P. chinensis C.L.Li, P. cuspidifera generic status of Yua has been supported by recent (Miq.) Planch. var. pubifolia C.L.Li, P. dalzielii molecular phylogenetic analyses (Wen et al., 2007; Gagnep., P. feddei (H.Lév.) C.L.Li, P. henryana Nie et al., 2010). (Hemsl.) Diels & Gilg, P. laetevirens Rehd., P. quin- Parthenocissus was established as a genus by Plan- quefolia (L.) Planch. (as an introduced species native chon (1887). Two infrageneric classification systems to North America), P. semicordata (Wall.) Planch., have been proposed since then (Table 1). Galet (1967) P. suberosa Hand.-Mazz. and P. tricuspidata (Sieb. & recognized three series in Parthenocissus: series Zucc.) Planch. The treatment by Chen et al. (2007) Tricuspidatae Galet mainly with simple leaves, series was largely congruent with that of Li (1998), except Trifoliolae Galet mainly with trifoliolate leaves and that P. cuspidifera var. pubifolia was treated as a series Quinquefoliolae Galet mainly with palmate synonym of P. semicordata in Chen et al. (2007). penta- (to hepta-) foliolate leaves. Based on tendril Species delimitation in south-eastern Asia and the morphology, Li (1996) divided the Chinese species of Himalayan region, however, has been highly contro- Parthenocissus into three sections: (1) section Par- versial. More than six names of species and infra- thenocissus with young tendril apices slender, curving specific taxa have been proposed in these regions and slightly expanded as fist-shaped adhesive discs; (e.g. P. anamallayana Planch., P. cuspidifera (Miq.) (2) section Margaritaceae C.L.Li with young tendril Planch., P. heterophylla (Blume) Merr., P. himalayana apices expanded into ball-like structures; and (3) (Royle) Planch., P. neilgherriensis (Wight) Planch. and section Tuberculiformes C.L.Li with palmate, pentafo- P. semicordata (Wall.) Planch.). The circumscrip- liolate leaves and young tendril apices expanded into tions of these species and their distribution require tubercles. investigation. Parthenocissus has a long and complex taxonomic Family level phylogenetic analyses have not history. Rehder (1905) recognized three species of resolved the relationships within Parthenocissus. Parthenocissus from North America: P. quinquefolia Ingrouille et al. (2002), using plastid rbcL sequences, (L.) Planch., P. vitacea (Knerr) Hitchc. and P. hepta- demonstrated the monophyly of Vitaceae and, on the phylla (Buckl.) Britton ex Small. His treatment has basis of two species, Parthenocissus. The plastid data been widely accepted by later workers (e.g. Brizicky, by Soejima & Wen (2006) supported the close rela- 1965; Soejima & Wen, 2006; Wen et al., 2007). The tionship between Parthenocissus and Yua, with three classification of Chinese Parthenocissus mainly species of Parthenocissus forming a clade with Yua follows Li (1998) and Chen et al. (2007). Li (1998) austro-orientalis. The GAI1 data, with more species of

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 45

Parthenocissus sampled, resolved Parthenocissus Parthenocissus (Li, 1996). We also investigated char- as monophyletic with moderate support. Recently, acters such as veinlets, calyx morphology and append- Nie et al. (2010) presented a molecular phylogenetic ages on the inner side of petals, because these analysis of Parthenocissus based on the plastid (atpB- characters have never been systematically evaluated rbcL, rps16 and trnL-F) and nuclear GAI1 sequences. for this genus (e.g. Li, 1996). Calyx morphology and The monophyly of Parthenocissus was confirmed with appendages on the inner side of petals were observed strong support. Within Parthenocissus, two major and photographed with a stereomicroscope (a Nikon clades corresponding to their distribution in Asia and SMZ1000 with a Nikon DXM 1200F digital camera). North America were recognized. Within the Asian Under light microscopy (LM), the leaf epidermis of clade, three subgroups congruent with their leaflet all 13 Parthenocissus spp. and one species of Yua was numbers were recognized. The Asian pentafoliolate examined. Samples were taken from mature leaves of P. henryana and P. laetevirens were supported as c.1¥ 1cm2 in size, boiled in water for approximately being closely related to each other. Three taxa with 10 min and then macerated in Jeffery’s solution simple and trifoliolate leaves (P. dalzielii, P. tricuspi- (Stace, 1965). Pieces of leaf epidermis were stained in data and P. suberosa) formed a well-supported clade. a solution of 1% safranin (in 50% ethanol) and then Taxa with exclusively trifoliolate leaves (P. hetero- dehydrated in an ethanol series (50, 75, 85, 95 and phylla, P. semicordata, P. feddei and P. chinensis) 100%) before being mounted in Canada balsam. To formed a clade. Relationships within the three Asian check the consistency of epidermal characters, we clades, however, have remained unclear. The phylo- examined materials covering the distributional range genetic relationships in the trifoliolate clade were not with at least ten slides made for the same specimen. well resolved: the Himalayan P. semicordata was only The stomatal index was calculated using the formula weakly supported as sister to P. heterophylla from S/(E + S) ¥ 100% based on at least ten slides. Java; and the positions of P. feddei and P. chinensis Scanning electron microscopy (SEM) was used to were still unclear. document micromorphology of leaves and petals and The morphology of Parthenocissus has not been pollen and seed characters. Samples were dehydrated systematically examined. The molecular phylogenetic in 100% ethanol (leaf and seed materials were study (Nie et al., 2010) has clarified the main phylo- cleaned in a KQ-300DA ultrasonic cleaner for 10 min genetic relationships within Parthenocissus, but the before dehydration) and were then placed on metal position of the Asian pentafoliolate species and rela- stubs using double-sided adhesive tape and sputter tionships within the trifoliolate clade were not well coated with gold in a Hitachi E-1010 Ion Sputter resolved. The objectives of this study are thus to: (1) Coater. The materials were subsequently observed document the comparative morphology of Partheno- and photographed under a Hitachi S-4800 scanning cissus based on leaf, inflorescence, floral, pollen and electron microscope. Pollen sizes from polar view (P) seed characters; (2) conduct a combined morphologi- and equatorial view (E) were measured from 15 cal and molecular phylogenetic analysis of Partheno- grains of each sample, except for P. heterophylla, with cissus; and (3) test the previous classifications of only eight grains measured because of insufficient Parthenocissus in a phylogenetic framework. material. Data of lumina diameters were measured based on 100 replicates. Seed measurements such as thickness of anticlinal walls of seed surface cell were based on ten replicates. MATERIAL AND METHODS The terminology of Dilcher (1974) and Wilkinson COMPARATIVE MORPHOLOGY (1979) was used to describe the leaf epidermis. We studied specimens from the following herbaria: A, Descriptive terminology of pollen followed Erdtman BM, CAL, CAS, CDB, E, IBSC, K, KUN, L, LE, MO, (1952), and that of seeds followed Tiffney & Barg- NY, P, PE, PH, U, US and WU. We also examined hoorn (1976), Tiffney (1979), Chen & Manchester living material from China (Fujian, Guangdong, (2007) and Chen (2009). Hunan, Jiangxi, Sichuan, Tibet and Yunnan), Japan, Java of Indonesia, Nepal, Thailand, Vietnam and the USA. Taxon names in this study generally followed Li CHARACTER SELECTION (1996), except that we recognized P. cuspidifera as a A total of 27 vegetative and floral morphological char- distinct species and used the broadly defined P. chin- acters were applied to the morphological phylogenetic ensis as in Nie et al. (2010). analysis of Parthenocissus (Table 2). They consisted Comparative studies of inflorescence structure, of 17 binary and ten multi-state characters. Data tendril morphology and leaf characters were con- were mainly derived from our own observations with ducted. We paid close attention to tendril morphology, consultation of literature (e.g. Li, 1998; Ren et al., which was emphasized in Li’s classification system of 2003; Wen, 2007). Some characters deserve further

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 46 L. LU ET AL.

Table 2. List of the 27 characters and character states scored for the phylogenetic analysis of Parthenocissus

1. Lenticels on branches: (0) 27–60/cm2; (1) 5–17/cm2. 2. Tendril apices: (0) disc absent; (1) slender and curving, rarely expanded; (2) disc ball-like; (3) disc tuberculate. 3. Tendril branches: (0) dichotomous; (1) 3–12 branched. 4. Leaflet numbers: (0) 5; (1) 7; (2) simple and 3; (3) 3. 5. Heterophylly in vegetative shoots: (0) absent; (1) mostly three-foliolate, sometimes simple; (2) mostly simple, occasionally three-foliolate. 6. Veinlets: (0) inconspicuous or slightly raised; (1) conspicuously raised. 7. Leaflet/leaf apices: (0) acute to acuminate; (1) acute cuspidate. 8. Leaflet/leaf margin: (0) serrate; (1) remotely denticulate. 9. Tooth spacing: (0) regular; (1) irregular. 10. Shape of cells on adaxial epidermis: (0) polygonal; (1) irregular. 11. Pattern of anticlinal wall on adaxial epidermis: (0) straight or arched; (1) sinuolate. 12. Shape of cells on abaxial epidermis: (0) polygonal; (1) irregular. 13. Pattern of anticlinal wall on abaxial epidermis: (0) straight or arched; (1) sinuolate. 14. Cuticles on abaxial leaf epidermis: (0) striated or papillate; (1) striated; (2) striated and scaly; (3) almost smooth. 15. Inflorescence position: (0) on long branches; (1) on short branches. 16. Inflorescence structure: (0) dichasium; (1) loose corymb; (2) panicle; (3) compound dichasium. 17. Secondary inflorescences on short branches: (0) 1; (1) 2; (2) > 2. 18. Calyx: (0) truncate; (1) obscurely 5 (–8) -lobed. 19. Appendages on the inner side of petals: (0) inconspicuous; (1) covering approximately half the length of anthers; (2) covering the entire length of anthers. 20. Pollen shape: (0) prolate; (1) perprolate; (2) subprolate. 21. Exine ornamentation of pollen grains: (0) reticulate; (1) foveolate–reticulate; (2) foveolate. 22. Seed shape: (0) pyriform; (1) obovate. 23. Ventral infolds on seed: (0) upward 2/3 from base; (1) from base to apex. 24. Cell shape of seed surface: (0) 3–4 sided; (1) 4–5 sided; (2) 5–6 sided. 25. Periclinal walls of seed surface cell: (0) smooth; (1) uneven. 26. Anticlinal cell walls of seed surface cell: (0) raised; (1) grooved. 27. Thickness of anticlinal cell walls of seed surface cell: (0) thick (> 15 mm, 15.6–17.7 mm); (1) thin (< 10 mm, 5.7–9.9 mm).

comments. Li (1996) regarded character 2 (tendril six to 12 in P. vitacea and three to nine in P. hepta- apices) as a diagnostic character in his taxonomic phylla). We thus coded the tooth shape (serrate vs. system of Parthenocissus. Our fieldwork indicated remotely denticulate) instead of tooth numbers on that whether the tendril apex expanded into a disc leaf/leaflet margins. Character 9 (tooth spacing) and may be subject to environmental conditions. For its two states (regular and irregular tooth spacing) example, it may easily expand into an adhesive disc were defined following Ellis et al. (2009). For charac- on smooth walls, but not on rough bark of trees. ter 16 (inflorescence structure), Parthenocissus semi- Nevertheless, the disc shape was generally consistent cordata was described as having divaricate cymes in with Li’s descriptions (Li, 1996). Regarding character Shetty & Singh (2000), as were P. heterophylla and 3 (tendril branches), our survey did not recognize P. chinensis according to our observations. Brizicky discrete states in Parthenocissus. For example, it may (1965) described P. vitacea and P. heptaphylla as vary from five to 11 within the same individual in having bifurcately branched corymb-like inflores- P. quinquefolia, and from four to eight in P. henryana. cences. Although both P. semicordata and P. vitacea We thus coded tendril branching as three- to were described as having a dichasium, we observed 12-branched in Parthenocissus and dichotomous in that the inflorescence of the latter was composed of Yua. We initially treated character 8 (leaflet/leaf three to five secondary dichasia. We thus coded the margin) as a quantitative character. We counted the inflorescence of P. vitacea and P. heptaphylla as a tooth numbers on one side of the middle leaflet (each compound dichasium. species with 15 specimens examined). There were, however, no discrete states recognizable among PHYLOGENETIC ANALYSIS species (e.g. nine to 17 in Parthenocissus heterophylla, All 13 species of Parthenocissus were included in the six to 13 in P. laetevirens, five to 12 in P. quinquefolia, phylogenetic study. Yua thomsoni and Y. austro-

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 47 orientalis were chosen as outgroups based on their 1 27 close relationship with Parthenocissus (Li, 1998;

Soejima & Wen, 2006; Chen et al., 2007; Wen et al., 0 2007). Twenty-seven morphological characters were scored in Table 3. Plastid (atpB-rbcL, rps16 and trnL-F) and nuclear GAI1 sequences of 12 Partheno- ???? cissus spp. were downloaded from GenBank (Appen- ???? dix), which were initially aligned in Clustal X version 1.83 (Thompson et al., 1997) and then manually adjusted in the program Se-Al v2.0a11 (Rambaut, 2002). Phylogenetic analyses of the morphological data set (Table 3), the molecular data set and the combined data set were conducted in PAUP* 4.0b10 (Swofford, 2003) using maximum parsimony (Fitch, 1971). A heuristic search was performed with 1000 random- taxon-addition replicates, holding one tree at each step during stepwise addition, tree bisection– reconnection (TBR) branch swapping and Multrees option on and collapsing branches if maximum length was zero. All characters and character states were

weighted equally with gaps treated as missing data. s as outgroups. ‘?’ represents missing data Parsimony bootstrap analyses (Felsenstein, 1985) were subsequently performed employing 1000 heuristic-search replicates. To test congruence between the morphological and molecular data sets, we implemented the incongru- ence length test (ILD; Farris et al., 1994) with PAUP* Y. austro-orientali 4.0b10 (Swofford, 2003) under the heuristic search constraints and 1000 replicates. Combinability of the and two data sets prior to phylogenetic analysis was also assessed by visual comparison of the trees derived from the separate data partitions. Yua thomsoni CHARACTER EVOLUTION

The ancestral states of two characters (tendril with apices and inflorescence structure), upon which Li’s classification of Parthenocissus (Li, 1996) was based, were reconstructed. In addition, the evolution of two additional characters (appendages on the inner side of petals and exine ornamentation of pollen grains) Parthenocissus was also examined. These characters were optimized in the program Mesquite version 2.74 (Maddison & Characters Maddison, 2010) under a maximum parsimony cri- 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 terion onto the molecular strict consensus tree. The states of the four characters are listed in Table 2. 000000000000000000000000000 131000000000011220000112100 111000001000011220000112011 111301000001131001202111101 121220000001111110020112100 121220000001111110000112100 101000001000011320000112011 00000000000??0000000000 1310000000011112200001121101100001000011320010112011 11130100000113100120211 1 111300000000011000101112101 111301110001131001202111101 121210000001111110000112100 11130100011112110120211110 RESULTS MORPHOLOGICAL CHARACTERS The inflorescence was a dichasium in P. heterophylla (with a well-developed main axis, three to five nodes Morphological data matrix of long, Fig. 1A); it had an inconspicuous main axis, approximately two nodes long in P. semicordata Yua thomsoni P. quinquefolia P. semicordata P. suberosa P. tricuspidata P. vitacea Yua austro-orientalis P. laetevirens P. heptaphylla P. heterophylla P. henryana Table 3. Species 01 02 P. cuspidifera P. dalzielii P. feddei and P. chinensis (Fig. 1G); P. feddei possessed a loose P. chinensis

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 48 L. LU ET AL.

Figure 1. Inflorescence structure, calyx morphology and appendages on the inner side of petals. A–C, Parthenocissus heterophylla (Maxwell 95–706, PE). D–F, P. feddei (Liu 28598, IBSC). G–I, P. chinensis (Feng 4851, PE). J–M, P. dalzielii (Li 254, PE). M, simple and small leaves on long branches. Scale bars, 1 cm (A, D, G, J, M); 1 mm (B, C, E, F, H, I, K, L). Illustrations by Ai-Li Li.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 49

Figure 2. Calyx morphology and appendages on the inner side of petals in Parthenocissus (arrows indicate appendages). A–B, Parthenocissus heterophylla (Forbes 823, LE). C–D, P. feddei (Zhong 827, KUN). E–F, P. chinensis (Wu et al. 1327, PE). G–H, P. dalzielii (Lu 229, PE). Scale bars, 500 mm. corymbose polychasium with a well-developed main ined (Fig. 3F–L). Stomatal size ranged from axis (Fig. 1D); P. dalzielii, P. tricuspidata and P. sub- 31.2 ¥ 23.3 mm to 43.5 ¥ 30.5 mminP. tricuspidata, erosa had a loose corymbose polychasium with an 30.4 ¥ 23.4 mm to 39.2 ¥ 29.4 mminP. dalzielii, inconspicuous main axis (Fig. 1J); P. henryana, P. la- 36.1 ¥ 22.7 mm to 43.5 ¥ 33.2 mminP. chinensis, etevirens and P. quinquefolia possessed a paniculate 32.3 ¥ 25.3 mm to 39.7 ¥ 30.7 mminP. quinquefolia, polychasium with a well-developed main axis; and 25.6 ¥ 15.6 mm to 36.4 ¥ 25.9 mminP. vitacea, P. vitacea and P. heptaphylla had a compound dicha- 30.1 ¥ 19.2 mm to 41.8 ¥ 26.3 mminP. henryana and sium with the main axis well developed, which dif- from 31.4 ¥ 22.0 mm to 43.0 ¥ 30.0 mminP. laetevi- fered from the simple dichasium in having more than rens. These values were overlapping and no discrete two secondary dichasia in the reproductive shoot. In states were recognizable. Average stomatal index was most cases, the calyx was entire in Parthenocissus calculated under LM based on at least ten slides: 10.9 and Yua, but it was observed to be five- (to eight-) in Parthenocissus dalzielii, 9.8 in P. quinquefolia, 9.3 lobed in P. heterophylla (Figs 1B and 2A) and P. feddei in P. vitacea, 6.5 in P. tricuspidata collected from (Figs 1E and 2C). There was an appendage on the China and 4.9 in P. tricuspidata collected from Japan. inner side of each petal, which was two or three times Both stomatal size and stomatal index varied sub- divided and approximately 1.0–1.5 mm long in P. het- stantially within species; therefore, these characters erophylla (Figs 1C and 2B) and P. feddei (Figs 1F and were not used in the cladistic analysis. Under SEM, 2D), approximately 0.2–0.5 mm long in P. chinensis trichomes were found on both adaxial and abaxial leaf (Figs 1I and 2F), but inconspicuous (< 0.2 mm) in surfaces of P. suberosa (Figs 3D and 4K) and mainly other examined species (e.g. P. dalzielii, Figs 1L and on abaxial veins in P. chinensis (Fig. 4L). The adaxial 2H). epidermis was usually with an ordered arrangement Under LM, the shape of leaf epidermal cells was of epidermal cells (Fig. 4B). Sometimes the epidermal polygonal (Fig. 3A, C–E, G, I, L) or irregular (Fig. 3B, cell pattern was obscured by a layer of wax ornamen- F, H, J, K). The anticlinal cell walls were straight, tation (Fig. 4A). The cuticular membrane on the arched or sinuolate. Stomata were anomocytic and abaxial epidermis was usually striate (Fig. 4D, E), only existed on abaxial epidermis in all species exam- scaly and granular in Parthenocissus, and it was

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 50 L. LU ET AL.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 51

Figure 3. Characteristics of leaf epidermis under light microscopy (LM) in Parthenocissus. A–E, adaxial epidermis. A, P. semicordata (Wang 82816, KUN); B, P. feddei (Zhong 827, KUN); C, P. chinensis (Qinghai-Xizang Exped. 750538, PE); D, P. suberosa (Yi 02156, KUN); E, P. vitacea (Levins & Bartels 81983, PE). F–L, abaxial epidermis. F, P. feddei (Zhong 827, KUN); G, P. chinensis (Qinghai-Xizang Exped. 750538, PE); H, P. tricuspidata (Chen 12–07, PE); I, P. vitacea (Levins & Bartels 81983, PE); J, P. dalzielii (Suzuki et al. s.n., PE); K, P. henryana (Fu 5384, PE); L, P. laetevirens (Chen 845, PE). Scale bars, 50 mm. ᭣

usually striate or papillate (Fig. 4G, J) in Yua. pyriform in Yua (Fig. 6B) from the dorsal view. The However, the presence or absence of papillae varied rostrate base was usually more obtuse in Partheno- within Y. thomsoni and Y. austro-orientalis (Fig. 4J cissus (Fig. 6A, C) than that in Yua (Fig. 6B, D) from shows papillae). The outlines of the pair of guard cells both dorsal and ventral views. Furthermore, two were elliptic (Fig. 4J) in most cases, occasionally sub- lateral ventral infolds in Parthenocissus were usually orbiculate (Fig. 4I). Guard cells were often thickened curved, asymmetric and running from base to apex and were made up of the outer stomatal rims. A (Fig. 6C), whereas those of Yua were straight, sym- nearly smooth rim was common in Parthenocissus metrical and running upward two-thirds from the and Yua, but the rim was sinuolate in P. feddei, seed base (Fig. 6D). In contrast, Parthenocissus and P. chinensis and sometimes in Y. thomsoni (Fig. 4J). Yua shared suites of characters such as round to oval Detailed characteristics of leaf epidermis in Partheno- chalaza located in the centre of the dorsal side cissus and Yua under LM and SEM are summarized (Fig. 6A, B) and reticulate seed sculpturing (Fig. 6E– in Table 4. M). In general, the pattern of seed sculpturing and Pollen grains of all species in Parthenocissus and shape of epidermal cells were consistent within Par- Yua were monomorphic, radially symmetrical, isopo- thenocissus and Yua. Aspects of anticlinal and peri- lar and tricolporate in apertures. Based on the defi- clinal cell walls showed a wide range of variations nition of Erdtman (1952), the pollen grains of among the investigated taxa. Thickness of anticlinal Parthenocissus were in the medium-sized category cell walls of seed surface cell was measured based on (from 29.5 ¥ 23.6 mminP. suberosa to 49.9 ¥ 23.9 mm ten replicates and two types were recognized: thick in P. heptaphylla). The outline of pollen grains in this (> 15 mm, 15.6–17.7 mm) and thin (< 10 mm, 5.7– genus was elliptical (Fig. 5A, D, G, J) in equatorial 9.9 mm). Characteristics of seeds are provided in view and was three-lobed circular in polar view Table 6. (Fig. 5B, E, H, K). According to the P/E ratio, the pollen shape in Parthenocissus was mainly prolate (P/E range: 8/6–8/4; Fig. 5A, D) with two other rare PHYLOGENETIC ANALYSES but distinct classes: subprolate (P/E range: 8/7–8/6) The morphological matrix included 27 morphological and perprolate (P/E range: > 8/4; Fig. 5G). The pollen characters, of which 22 were parsimony informative. apertures of all species examined in Parthenocissus Twelve maximally parsimonious trees (MPTs) with a and Yua were tricolporate with the colpi extending length of 56 steps, a consistency index (CI) of 0.73 and nearly to the poles. Of all the material observed in the a retention index (RI) of 0.77 were obtained. The present study, P. heptaphylla had the longest colpus molecular matrix contained 4137 characters, of which (Fig. 5G) and P. suberosa had the shortest. In some 73 were parsimony-informative. Nine MPTs with a species, the colpi were wide enough to show dense length of 196 steps (CI = 0.94; RI = 0.91) were granulae inside (Fig. 5E). In most cases, however, the retained from the molecular analyses. Results of the colpi were narrow, smooth and linear (Fig. 5A, G, J). ILD test showed no significant incongruence between Exine ornamentation of Parthenocissus and Yua was the morphological and the molecular data (P = 0.953). foveolate or reticulate. Based on the average lumina Hence, the two data sets were combined for the phy- diameter, three pollen types may be recognized in logenetic analysis. The matrix of the combined mor- the examined taxa: foveolate (< 0.2 mm, Fig. 5C), phological and molecular data set included 4164 foveolate–reticulate (0.2–0.5 mm, Fig. 5F) and reticu- characters, of which 95 were potentially parsimony late (> 0.5 mm, Fig. 5I, L). Description of pollen varia- informative. Maximum parsimony analysis yielded 15 tion in Parthenocissus and Yua is provided in Table 5. MPTs with a length of 250 steps (CI = 0.91; RI = 0.88). Number of seeds per fruit in Parthenocissus and The strict consensus trees of the molecular and mor- Yua ranged from one to four, mostly two or three. The phological data sets are presented in Figure 7A, B, shape of seeds varied to some degree among fruits respectively. The strict consensus tree of the com- with different seed number. Nevertheless, the shape bined morphological and molecular analyses is was generally obovate in Parthenocissus (Fig. 6A) and depicted in Figure 8.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 52 L. LU ET AL.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 53

Figure 4. Characteristics of epidermal surface under scanning electron microscopy (SEM) in Parthenocissus and Yua. A–B, adaxial epidermis. A, P. heterophylla (Forbes 983, LE); B, P. dalzielii (Suzuki et al. s.n., PE). C–G, abaxial epidermis. C, P. heterophylla (Forbes 983, LE); D, P. feddei (Zhong 827, KUN); E, P. henryana (Fu 5384, PE); F, P. laetevirens (Chen 845, PE); G, Y. thomsoni (Lin 2004161, PE). H–J, characteristics of stomata. H, P. heterophylla (Forbes 983, LE); I, P. suberosa (Nie 3413, PE); J, P. thomsoni (Lin 2004161, PE). K–L, trichomes on abaxial epidermis. K, P. suberosa (Nie 3413, PE); L, P. chinensis (Qinghai-Xizang Exped. 750538, PE). Scale bars, 100 mm (A–G); 20 mm (H–J); 200 mm (K–L). ᭣

In the morphological phylogenetic analyses, three CHARACTER OPTIMIZATION main clades were recognizable within Parthenocissus: The characters of tendril apices, inflorescence struc- the Semicordata clade, the Tricuspidata clade and the ture, appendages on the inner side of petals and exine Quinquefolia clade (Fig. 7B). The Semicordata clade ornamentation of pollen grains were traced onto the was supported with a bootstrap (BS) value of 97%, molecular strict consensus tree (Fig. 9). Of the four which included P. cuspidifera, P. heterophylla, P. semi- character states for tendril apices (disc absent; cordata and P. feddei. The Tricuspidata clade (96% slender and curving, rarely expanded; disc ball-like; BS) consisted of P. dalzielii, P. tricuspidata and and disc tuberculate), the ancestral state was equivo- P. suberosa. The Quinquefolia clade (68% BS) was cal and the state ‘slender and curving tendril’ may comprised of Parthenocissus quinquefolia and a clade have evolved at least twice in Parthenocissus. Never- of P. vitacea and P. heptaphylla (82% BS). theless, states ‘disc ball-like’ and ‘disc tuberculate’ Results of the combined morphological and molecu- seem to have derived only once in Old World Par- lar analysis were broadly congruent with those of the thenocissus (Fig. 9A). The parsimony reconstruction molecular analysis (Figs 7A and 8). A similar topology of the inflorescence structure indicated its complex was obtained except that the clade containing P. cus- history: the ancestral state appeared to be equivocal pidifera, P. heterophylla, P. semicordata and P. feddei and the states ‘panicle’ and ‘loose corymb’ evolved at was collapsed in the molecular strict consensus tree least twice in Parthenocissus (Fig. 9B). The character (Fig. 7A). Within Parthenocissus, two main clades ‘appendages on the inner side of petals’ included three were recognizable, corresponding to their distribution states. ‘Inconspicuous appendage’ was inferred to be in the New and the Old World, with 100% (vs. 100% ancestral in Parthenocissus. The half-anther length in the molecular analysis) and 72% (vs. 93% in the appendage and the full-anther length appendage molecular analysis) bootstrap values, respectively. were supported as derived states (Fig. 9C). The opti- Within the New World clade, P. vitacea and P. hepta- mization of the character exine ornamentation of phylla formed a strongly supported clade (98% BS in pollen grains suggested ‘reticulate’ as ancestral state, the combined analysis vs. 96% BS in the molecular and ‘foveolate–reticulate’ and ‘foveolate’ as derived analysis). Within the Old World clade, three clades states (Fig. 9D). correlating with their leaflet numbers were supported (Figs 7A and 8). The Asian pentafoliolate P. henryana and P. laetevirens were supported as close to each DISCUSSION other (69% BS in the combined analysis vs. 88% BS in the molecular analysis). Three taxa with both simple PHYLOGENY and trifoliolate leaves (P. dalzielii, P. tricuspidata and Three main clades are supported by the morpho- P. suberosa) formed a well-supported clade (100% BS logical phylogenetic analysis: the Semicordata clade, in both molecular and combined analyses). In the the Tricuspidata clade and the Quinquefolia clade combined analysis, P. feddei was strongly supported (Fig. 7B). The combined morphological and molecular (96% BS) to be sister to the clade containing P. cus- analysis provided a robust phylogenetic hypothesis pidifera, P. heterophylla and P. semicordata (Fig. 8). for Parthenocissus, which is largely congruent with The close relationship between P. cuspidifera, P. het- the recently published molecular phylogenetic work erophylla and P. semicordata was only weakly sup- (Nie et al., 2010). Two main clades corresponding to ported in the combined analysis (BS = 55%). The their distribution in the New and the Old World are broadly defined P. chinensis was resolved as sister to recognized (Fig. 8). the clade including all the other trifoliolate species The three New World species are strongly sup- (Fig. 8). Furthermore, a weakly supported clade con- ported as a clade in the combined analysis taining the Asian pentafoliolate species and taxa with (BS = 100%), and they share the synapomorphy of simple and three-foliolate leaves was recognized in having irregular teeth on the leaflet/leaf margin, both the molecular and combined strict consensus grooved anticlinal cell walls and smooth periclinal cell trees (Figs 7A and 8). walls of the seed surface cell. Parthenocissus vitacea

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 54 L. LU ET AL.

is strongly supported as sister to P. heptaphylla,as supported by two synapomorphies: absence of discs at the tendril apex, and possessing a compound dicha- sium. Brizicky (1965) also considered the narrow endemic Parthenocissus heptaphylla from central Texas to be closely related to P. vitacea and thought it may even be treated as a variety or subspecies of P. vitacea. Parthenocissus quinquefolia has been widely cultivated as a climbing ornamental in Asia and it shares several characters with the Asian species. For instance, Li (1998) placed P. quinquefolia in section Parthenocissus together with the trifoli- olate taxa based on its tendril morphology (tendril slender and curving, rarely expanded as discs). In addition, P. quinquefolia possesses a paniculate poly- chasium similar to that of the Asian pentafoliolate P. henryana and P. laetevirens. The morphological convergences may explain why the monophyly of New World Parthenocissus is not strongly supported (68% BS) by the morphological data. Congruent with the previous molecular data (Nie et al., 2010), three groups corresponding to leaflet numbers are recognizable within the mono- phyletic Old World Parthenocissus in our combined analysis: the clade of pentafoliolate taxa (BS = 69%), the clade of taxa with simple and trifoliolate leaves (BS = 100%) and taxa exclusively with trifoliolate leaves (BS = 86%). Parthenocissus henryana and P. laetevirens form the pentafoliolate group and they share the synapomorphies of tendril apex extended as tuberculate discs and inflorescence a paniculate poly- chasium. These two species have an overlapping distribution in Central China. They differ in that

under light and scanning electron microscopy P. henryana has four-ridged young branches (vs. rounded branches in P. laetevirens) and that P. la- Yua etevirens has a bullate leaflet surface (vs. smooth

and leaflet surface in P. henryana) (Li, 1998). Parthenocissus dalzielii, P. tricuspidata and P. sub- erosa form a strongly supported clade with the fol- lowing synapomorphies: tendril apex ball-like, leaves both simple and trifoliolate, a loosely corymbose inflo- rescence and inflorescences on short branches with Parthenocissus two orders of branching (Fig. 1J). Their close relation- ship was also indicated in Li’s classification system (Li, 1996). Although all three species have two types of leaves, P. dalzielii is usually trifoliolate and only sometimes has simple leaves, whereas P. tricuspidata and P. suberosa are mainly with simple leaves, occa- sionally with trifoliolate leaves on the lower two to Adaxial epidermisShape of cellsPolygonal Pattern of anticlinalPolygonal wallPolygonal Shape ofIrregular Straight cells or archedPolygonal Straight or arched PatternPolygonal of Straight anticlinal or wall archedPolygonal SinuolatePolygonal Straight Abaxial Shape or epidermis of arched Polygonal guardPolygonal Straight cells or arched IrregularPolygonal Straight or Cuticular arched Irregular membrane Polygonal Straight or arched StraightPolygonal Straight or or arched arched Irregular StraightPolygonal Straight or or arched arched Polygonal SinuolatePolygonal Straight or Irregular arched Irregular Straight or arched Polygonal Sinuolate Straight or Suborbiculate arched Polygonalfour Straight Straight or or arched Elliptic arched Irregular Sinuolate Sinuolateshort Irregular Straight or arched Irregular Straight or arched Striate Polygonal Sinuolateshoots. Elliptic Elliptic Polygonal Sinuolate Sinuolate Elliptic Elliptic Straight Nearly or smooth arched Elliptic Straight or Elliptic Elliptic arched Striate striate Elliptic Elliptic Suborbiculate Elliptic Striate Striate Elliptic Striate Striate Nearly and smooth scaly Striate Nearly Striate smooth Striate or papillate Striate The systematic positions of P. heterophylla from tropical Asia, P. semicordata from the Sino- Leaf epidermal characters of Himalayan region, P. cuspidifera from the Khasi hills in India and P. feddei from Central China have long been controversial. Nie et al. (2010) suggested

Table 4. Taxon P. chinensis P. cuspidifera P. dalzielii P. feddei P. henryana P. heptaphylla P. heterophylla P. laetevirens P. quinquefolia P. semicordata P. suberosa P. tricuspidata P. vitacea Yua thomsoni a close relationship between P. heterophylla and

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 55

Figure 5. Equatorial view, polar view and sculpture details of pollen grains of Parthenocissus species under scanning electron microscopy (SEM). A–C, Parthenocissus heterophylla (Forbes 823, LE). D–F, P. chinensis (Zhang & Lang 4362, PE). G–I, P. heptaphylla (Wen & Xie 9788, PE). J–L, Yua thomsoni (China, Hunan, Xia 224, PE). The foveolate, foveolate–reticulate and reticulate pollen exine ornamentation are presented in C, F, and I, L, respectively. Scale bars, 10 mm (A, B, D, E, G, H, J, K); 5 mm (C, F, I, L).

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 56 L. LU ET AL.

P. semicordata (64% BS), but the phylogenetic posi- tion of P. feddei was unresolved and the Indian endemic species P. cuspidifera was not sampled. Our combined phylogenetic analysis suggests that P. feddei forms a robust clade (96% BS) with a clade containing the south-eastern Asian and Himalayan P. cuspidifera, P. heterophylla, and P. semicordata. They share the following synapomorphies: conspicu- ously raised veinlets, an obscurely five- (to eight-) lobed calyx (Figs 1B, E and 2A, C), appendages on the inner side of petals covering the entire length of m) Exine ornamentation

m anthers (Figs 1C, F and 2B, D) and foveolate pollen exine ornamentation (Fig. 5C). Parthenocissus chin- ensis is supported as sister to the clade of the other trifoliolate taxa. The position of the two Asian pentafoliolate species was not resolved in the previous molecular phyloge- netic analysis (Nie et al., 2010). In the current study, P. henryana and P. laetevirens were supported as more closely related to species with both simple and trifoliolate leaves, although this was only weakly sup- ported (Figs 7A and 8).

CHARACTER EVOLUTION In their molecular phylogenetic analyses of Partheno- cissus, Nie et al. (2010) traced the character evolution of leaflet numbers and suggested that leaflet number in this genus has a complex history and cannot be used as a character for infrageneric classification. Morphology of tendril apices was considered as an important character in Li’s classification of Partheno- cissus (Li, 1996). Our optimization of tendril apices suggests that the ancestral state was equivocal and the state ‘slender and curving tendril’ may have evolved at least twice in Parthenocissus, whereas the states ‘disc ball-like’ and ‘disc tuberculate’ seem to m) P/E Shape Lumina diameter ( m

Yua have derived only once in Old World Parthenocissus (Fig. 9A). From our field observations, we propose and that whether the tendril apices expand into discs or not may depend on the environment, but the disc shape is generally consistent within species. Morphological floral characters often play an impor- tant role in taxonomic studies (Gerrath & Posluszny,

Parthenocissus 1989). We traced three floral characters (inflorescence structure, appendages on the inner side of petals and m) E (

m exine ornamentation of pollen grains). Inflorescence

(32.9)–35.5–(41.0)(29.5)–32.0–(34.7)(31.1)–32.7–(34.6)(29.0)–30.6–(34.5) (20.3)–22.3–(24.5)(34.0)–37.7–(43.7) (20.3)–24.0–(28.0)(47.9)–49.9–(51.8) (23.1)–24.0–(24.6)(34.2)–37.2–(40.6) (20.6)–22.9–(25.6) 1.60(43.2)–43.2–(43.3) (17.7)–20.2–(25.2) 1.34(31.0)–33.2–(35.8) (21.0)–23.9–(26.2) 1.36(29.5)–32.0–(34.7) (19.4)–22.7–(28.1) 1.34 Prolate(27.1)–29.5–(32.8) (23.1)–24.0–(24.8) 1.88 Prolate(33.8)–39.8–(42.9) (22.2)–23.7–(25.1) 2.10 Prolate(37.8)–39.5–(42.3) (20.3)–24.0–(28.0) 1.66 Prolate(40.5)–41.8–(44.4) (21.7)–23.6–(25.1) 1.81 Prolate (0.12)–0.37–(1.18) (18.7)–21.3–(25.1) 1.41 Perprolate (0.05)–0.08–(0.25) (23.7)–27.4–(30.9) 1.34 Prolate (0.22)–0.55–(1.13) (19.5)–21.7–(24.0) 1.25 Prolate (0.04)–0.13–(0.32)structure 1.89 Prolate (0.33)–0.98–(2.70) (0.28)–0.71–(1.49) 1.45 Prolate Foveolate–reticulate 1.94 Subprolate Foveolate (0.04)–0.09–(0.46) Prolate Reticulate (0.10)–0.85–(2.32)was Prolate Foveolate (0.22)–0.79–(1.59) prolate (0.27)–0.74–(1.99) Reticulate used Reticulate (0.03)–0.08–(0.25) Foveolate as (0.41)–0.75–(1.55) Reticulate (0.17)–0.76–(1.63) an Reticulate (0.21)–1.16–(3.42) important Reticulate Foveolate Reticulate Reticulate taxonomic reticulate char- acter in the classification system of Li (1996, 1998). The optimization suggests a complicated evolutionary pattern of inflorescence structure in Parthenocissus. The ancestral state appeared to be equivocal and the Pollen morphology of states ‘panicle’ and ‘loose corymb’ each evolved at least twice in Parthenocissus (Fig. 9B). The character inflorescence structure may not reflect the true P. chinensis P. cuspidifera P. dalzielii P. feddei P. henryana P. heptaphylla P. heterophylla P. laetevirens P. quinquefolia P. semicordata P. suberosa P. tricuspidata P. vitacea Yua thomsoni Table 5. Taxon P ( phylogenetic relationships within Parthenocissus.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 57

Figure 6. Scanning electron microscopy (SEM) micrographs of seeds and seed coat micromorphology in Parthenocissus and Yua. A–D, mature seeds. A, dorsal view, P. dalzielii (Cheng 137, PE); B, dosal view, Y. thomsoni (Libo Exped. 1006, PE); C, ventral view, P. feddei (Yang 03069, IBSC); D, ventral view, Y. thomsoni (Libo Exped. 1006, PE). E–M, seed coat micromorphology. E, P. feddei (Yang 03069, IBSC); F, P. chinensis (Qiu 55225, PE); G, P. dalzielii (Cheng 137, PE); H, P. suberosa (Zhong 83398, PE); I, P. tricuspidata (Lu,BJ001, PE); J, P. quinquefolia (Lu BJ002, PE); K, P. vitacea (Deam 29480, MO); L, P. heptaphylla (Atha 185, NY); M, Y. thomsoni (Fan et al. 18–05, PE). Scale bars, 1 mm (A, B, C, D); 100 mm (E, F, G, H, I, J, K, L, M).

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 58 L. LU ET AL.

Table 6. Seed morphology and seed coat micromorphology of Parthenocissus and Yua (‘–’ indicates missing data)

Seed Surface cell Anticlinal Thickness of Periclinal Taxon shape Ventral infolds shape walls anticlinal walls (mm) walls

P. chinensis Obovate From base to apex 5–6 sided Raised 6.1 Uneven P. cuspidifera Obovate From base to apex 4–5 sided Raised 6.2 Uneven P. dalzielii Obovate From base to apex 5–6 sided Raised 16.1 Uneven P. feddei Obovate From base to apex 4–5 sided Raised 9.0 Uneven P. henryana Obovate From base to apex 5–6 sided Raised 5.7 Uneven P. heptaphylla Obovate From base to apex 5–6 sided Grooved 7.9 Smooth P. heterophylla Obovate From base to apex – – – – P. laetevirens Obovate From base to apex 5–6 sided Raised 17.7 Uneven P. quinquefolia Obovate From base to apex 5–6 sided Grooved 9.9 Smooth P. semicordata Obovate From base to apex 4–5 sided Raised 6.9 Uneven P. suberosa Obovate From base to apex 5–6 sided Raised 16.0 Uneven P. tricuspidata Obovate From base to apex 5–6 sided Raised 16.00 Uneven P. vitacea obovate from base to apex 5–6 sided Grooved 8.1 Smooth Yua thomsoni pyriform upward 2/3 from base 3–4 sided Raised 15.6 Smooth

For instance, the North American P. quinquefolia Li (1996) are all monophyletic (Fig. 8). Suessenguth possesses a paniculate polychasium similar to that (1953) discussed the Asian and North American of the Asian pentafoliolate species (P. henryana and species of Parthenocissus separately. However, he also P. laetevirens), but they are not closely related. The indicated that the Asian P. henryana and the North evolutionary patterns of the two floral characters, American P. quinquefolia were closely related. It appendages on the inner side of petals and exine seems that the two geographical groups of Suessen- ornamentation of pollen grains, are supported as guth were for the convenience of his discussions only, congruent with the phylogeny. The inconspicuous rather than for classification purposes. With a small appendages and reticulate pollen ornamentation are genus of only 13 species, we prefer to just illustrate each inferred as ancestral in Parthenocissus. The the phylogenetic relationships (Figs 7 and 8) and we full-anther length appendage and foveolate pollen will not formally subdivide the genus into sections ornamentation are each supported as derived states and series in this paper. (Fig. 9C, D). From our observations, the full-anther Li (1996) described P. chinensis as a new species length appendage separates the anther from the based on its smaller inflorescence and smaller leaflets stigma before anthesis, and it reflexes to expose the with fewer teeth on the leaflet margin, compared anther when the petals open fully. The appendage on with P. semicordata. Parthenocissus chinensis was the petal may thus prevent self-pollination. described by Li (1996) as endemic to south-western and western Sichuan. We noted that this smaller- leaved species is also distributed in north-western TAXONOMIC IMPLICATIONS Yunnan and Tibet in our fieldwork. Nie et al. (2010) Two infrageneric classifications of Parthenocissus broadly sampled P. chinensis and P. semicordata cov- (Galet, 1967; Li, 1996) have been proposed (Table 1). ering their range from western China to the Hima- Galet (1967) used leaf architecture to divide Par- layan region. They found that ‘P. semicordata’ from thenocissus into three series: Trifoliolae, Tricuspida- southern and western China formed a clade with tae and Quinquefoliolae (Table 1). Galet (1967) placed P. chinensis, whereas the typical P. semicordata from P. dalzielii in series Trifoliolae, but the species forms southern Himalaya formed another clade with the a clade with P. suberosa and P. tricuspidata of series tropical P. heterophylla. They proposed a broadly cir- Tricuspidatae. Furthermore, the pentafoliolate series cumscribed P. chinensis with the inclusion of the pre- Quinquefoliolae of Galet (1967) is paraphyletic. Li viously identified ‘P. semicordata’ in China as in (1996) recognized three sections and four series in Li (1996, 1998). Our comparative morphological Parthenocissus (Table 1). Li’s sections Margaritaceae study did not detect any consistent characters to and Tuberculiformes are each monophyletic (Fig. 8). distinguish P. chinensis from ‘P. semicordata’ identi- Li’s section Parthenocissus, however, is paraphyletic, fied by Li (1998), except that the former has smaller with taxa of sections Margaritaceae and Tuberculi- leaflets and shorter inflorescences, which may be an formes nested within it. The four series recognized by adaptation to the arid or semi-arid inhabit. Typical

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 59

A P.P. ccuspidiferauspidifera B P.P. hheterophyllaeterophylla 86 P.P. ssemicordataemicordata 97

94 P.P. ffeddeieddei

P.P. cchinensishinensis

P.P. ddalzieliialzielii 93 + P.P. ttricuspidataricuspidata 100 96

P.P. ssuberosauberosa 100 64 100 P.P. hhenryanaenryana 88 P.P. llaetevirensaetevirens

P.P. qquinquefoliauinquefolia

100 P.P. vvitaceaitacea 68 96 82 P.P. hheptaphyllaeptaphylla

YuaYua tthomsonihomsoni

Y.Y. austro-orientalisaustro-orientalis

—— 3-foliolate —— Old World

—— simple —— New World —— 5(-7)-foliolate

Figure 7. Maximum parsimony analysis of Parthenocissus with leaf architecture and distribution indicated. A, strict consensus tree of molecular (plastid atpB-rbcL, rps16, trnL-F regions and nuclear GAI1 gene) analysis. B, strict consensus tree of morphological analysis (27 characters). Bootstrap values (> 50%) are provided below the branches.

P. semicordata was based on Wallich’s Vitis semicor- tions support Nie et al.’s (2010) expansion of the data, originally described from the mountains of concept of P. chinensis to include most of the previ- Shivapore in Nepal (Roxburgh, 1824). From field ously identified ‘P. semicordata’ from China in Li observations and specimen examination, we define (1996, 1998). P. semicordata as broadly distributed from Bhutan, Parthenocissus heterophylla was based on Blume’s northern India, and Nepal of the Himalayan region, Ampelopsis heterophylla (Blume, 1825) from Java, through Myanmar and western China, to Thailand which was often confused with P. semicordata (e.g. and Vietnam. Morphologically, typical P. semicordata Latiff, 1982). From our comparative morphological differs from Li’s ‘P. semicordata’(=P. chinensis) from study and the phylogenetic analyses, we here support southern and western China in the following charac- the close relationship between P. heterophylla and ters: the inflorescence with a well-developed main P. semicordata, but argue that they should be recog- axis (vs. main axis inconspicuous in Li’s ‘P. semicor- nized as two distinct species. The two species differ data’, see Fig. 1G), the petal with a longer appendage in their tendril morphology, with P. heterophylla inside (1.0–1.5 vs. 0.2–0.5 mm in Li’s ‘P. semicordata’, having secondary branches of tendrils and longer see Figs 1I and 2F) and pollen grains with foveolate shoots that bear inflorescences (Fig. 1A). Parthenocis- exine ornamentation (vs. foveolate–reticulate in Li’s sus heterophylla is only narrowly distributed in ‘P. semicordata’, see Fig. 5E). Therefore, our observa- Java, Malay, India, Thailand and southern Vietnam.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 60 L. LU ET AL.

sect. Parthenocissus P.P. ccuspidiferauspidifera sect. Margaritaceae P.P. hheterophyllaeterophylla sect. Tuberculiformes 55 ser. 96 P.P. ssemicordataemicordata Trifoliolae P.P. ffeddeieddei 86 P.P. cchinensishinensis ser. P.P. ddalzieliialzielii 72 Heterophyllae P.P. ttricuspidataricuspidata 100 ser. Tricuspidatae P.P. suberosasuberosa

P.P. hhenryanaenryana 100 69 P.P. llaetevirensaetevirens

P.P. qquinquefoliauinquefolia ser. 100 P.P. vvitaceaitacea Parthenocissus 98 P.P. hheptaphyllaeptaphylla

YuaYua tthomsonihomsoni Outgroups Y.Y. aaustro-orientalisustro-orientalis

Figure 8. Strict consensus tree of combined morphological and molecular analysis of Parthenocissus with classification of Li (1996) indicated. Bootstrap values (> 50%) are provided below the branches.

Parthenocissus cuspidifera from the Khasi Hills of bined morphological and molecular phylogenetic northern India was treated as a synonym of P. semi- analyses recognize two clades within Parthenocissus cordata by Shetty & Singh (2000). Our phylogenetic corresponding to their distribution in Asia and North result did not resolve the relationships within the America. Among the Asian species, three clades are clade of P. cuspidifera, P. heterophylla and P. semicor- supported congruent with their leaflet numbers. Par- data. Parthenocissus cuspidifera clearly differs from thenocissus feddei from central China forms a clade P. heterophylla and P. semicordata in its conspicuously with species from Himalaya and south-eastern Asia. cuspidate leaf apices (vs. acute to acuminate in P. het- The Asian pentafoliolate species are weakly sup- erophylla and P. semicordata) and denticulate leaf ported as closely related to species with both simple margins (vs. serrate in P. heterophylla and P. semicor- and trifoliolate leaves. The optimization of a vegeta- data). Parthenocissus feddei was considered as con- tive and three floral morphological characters specific with P. heterophylla by Rehder (1934). Our suggests appendages on the inner side of petals and morphological survey suggests that it differs from exine ornamentation of pollen grains as important P. heterophylla in its straight lateral veins into the taxonomic characters. Previous classifications of teeth (vs. arching and branched in P. heterophylla) Parthenocissus are evaluated based on our compara- and an inflorescence of a racemose polychasium (vs. a tive morphological study within the phylogenetic dichasium in P. heterophylla; see Fig. 1A, D). framework.

CONCLUSIONS ACKNOWLEDGEMENTS We documented leaf morphology, inflorescence struc- We are grateful to the curators and staff members of ture, characters of calyx, appendages on the inner A, BM, CAL, CAS, CDB, E, IBSC, K, KUN, L, LE, side of petals, pollen grains and seeds of Parthenocis- MO, NY, P, PE, PH, U, US and WU for allowing us to sus throughout Asia and North America. The com- examine specimens. We thank F. M. B. Jacques

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 61 Yua austro-orientalis Yua thomsoni Yua austro-orientalis Yua thomsoni P. chinensis P. heterophylla P. semicordata P. feddei P. henryana P. laetevirens P. suberosa P. tricuspidata P. dalzielii P. heptaphylla P. vitacea P. chinensis P. heterophylla P. semicordata P. feddei P. henryana P. laetevirens P. suberosa P. tricuspidata P. quinquefolia P. dalzielii P. heptaphylla P. vitacea P. quinquefolia

Tendril apices Inflorescence structure discs absent dichasium slender and curving, rarely expanded loose corymb discs ball-like panicle discs tuberculate compound dichasium ABequivocal equivocal P. heptaphylla P. vitacea P. quinquefolia P. tricuspidata P. dalzielii Yua thomsoni P. chinensis Yua austro-orientalis Yua thomsoni P. chinensis P. heterophylla P. semicordata P. feddei P. henryana P. laetevirens P. suberosa P. tricuspidata P. heptaphylla P. vitacea Yua austro-orientalis P. heterophylla P. semicordata P. feddei P. henryana P. laetevirens P. suberosa P. dalzielii P. quinquefolia

Exine ornamentation Appendages on the inner side of petals of pollen grains inconspicuous reticulate covering about half the length of anthers foveolate–reticulate C covering the entire length of anthers D foveolate

Figure 9. Parsimony inference of character evolution onto the molecular strict consensus tree. A, tendril apices. B, inflorescence structure. C, appendages on the inner side of petals. D, exine ornamentation of pollen grains. for translating the French references, G. McKee for REFERENCES translating the German references and three review- ers and Dr Michael Fay for their constructive Blume CJ. 1825. Bijdragen tot de Flora van Nederlandsch comments. This research was financially supported by Indië. 1ste Stuk. Batavia: Ter Lands Drukkerij. the National Natural Science Foundation of China Brizicky GK. 1965. The genera of Vitaceae in the southeast- (grant no. 40830209), the Chinese Academy of Sci- ern United States. Journal of the Arnold Arboretum 46: ences (grant no. KSCX2-YW-R-136), the US National 48–67. Science Foundation (grant DEB 0743474 to S. R. Chen I. 2009. History of Vitaceae inferred from morphology- Manchester and J. Wen) and the Small Grants based phylogeny and the fossil record of seeds. Thesis, Program of the National Museum of Natural History, University of Florida. the Smithsonian Institution. Chen I, Manchester SR. 2007. Seed morphology of modern

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 62 L. LU ET AL.

and fossil Ampelocissus (Vitaceae) and implications for Planchon JE. 1887. Monographie des Ampélidées vrais. In: phytogeography. American Journal of Botany 94: 1534– de Candolle AFPP, de Candolle C, eds. Monographiae phan- 1553. erogamarum. Paris: G. Masson, 305–654. Chen ZD, Ren H, Wen J. 2007. Vitaceae. In: Wu ZY, Hong Rambaut A. 2002. Se-Al: sequence alignment editor. Version DY, Raven PH, eds. Flora of China. Vol. 12. Beijing: Science 2.0a11. Oxford: University of Oxford, Available at: http:// Press; and St Louis: Missouri Botanical Garden Press, 173– tree.bio.ed.ac.uk/software/seal/ 177. Rehder AA. 1905. Parthenocissus. In: Sargent CS, ed. Trees Dilcher DL. 1974. Approaches to the identification of and shrubs. Cambridge: The Riverside Press, 183–190. angiosperm leaf remains. Botanical Review 40: 91–108. Rehder AA. 1934. Notes on the ligneous described by Ellis B, Daly DC, Hickey LJ, Johnson KR, Mitchell JD, Léveillé from eastern Asia. Journal of the Arnold Arboretum Wilf P, Wing SL. 2009. Manual of leaf architecture. New 15: 1–27. York: Cornell University Press. Ren H, Pan KY, Chen ZD, Wang RQ. 2003. Structural Erdtman G. 1952. Pollen morphology and taxonomy. characters of leaf epidermis and their systematic signifi- Angiosperms. Stockholm: Almqvist and Wiksell. cance in Vitaceae. Acta Phytotaxonomica Sinica 41: 531– Farris JS, Källersjö M, Kluge AG, Bult C. 1994. Testing 544. significance of incongruence. Cladistics: The International Roxburgh W. 1824. Flora indica. Vol. 2. Serampore: The Journal of the Willi Hennig Society 10: 315–319. Mission Press. Felsenstein J. 1985. Confidence limits on phylogenies: an Shetty BV, Singh P. 2000. Vitaceae. In: Singh NP, Vohra JN, approach using the bootstrap. Evolution 39: 783–791. Hajra PK, Singh DK, eds. Flora of India. Vol. 5. Calcutta: Fitch WM. 1971. Toward defining the course of evolution: Botanical Survey of India, 246–324. minimum change for a specific tree topology. Systematic Soejima A, Wen J. 2006. Phylogenetic analysis of the grape Zoology 20: 406–416. family (Vitaceae) based on three chloroplast markers. Galet P. 1967. Recherches sur les méthodes d’identification et American Journal of Botany 93: 278–287. de classification des Vitacées temprées. II, Thesis, Univer- Stace CA. 1965. Cuticular studies as an aid to plant tax- sité de Montpellier. onomy. Bulletin of British Museum (Natural History) Gerrath JM, Posluszny U. 1989. Morphological and ana- Botany 4: 1–78. tomical development in the Vitaceae. IV. Floral development Suessenguth K. 1953. Vitaceae. In: Engler A, Prantl K, eds. in Parthenocissus inserta. Canadian Journal of Botany 67: Die natürlichen pflanzenfamilien, 2nd edn. Berlin: Duncker 1356–1365. and Humblot, 174–333. Ingrouille MJ, Chase MW, Fay MF, Bowman D, Bank Swofford DL. 2003. PAUP*: phylogenetic analysis using par- MVD, Bruijn ADE. 2002. Systematics of Vitaceae from the simony (*and other methods). V. 4.0beta10. Sunderland, MA: viewpoint of plastid rbcL DNA sequence data. Botanical Sinauer Associates. Journal of the Linnean Society 138: 421–432. Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Latiff A. 1982. Studies in Malesian Vitaceae, I–IV. Federation Higgins DG. 1997. The Clustal X windows interface: flex- Museums Journal 27: 46–93. ible strategies for multiple sequence alignment aided by Li CL. 1990. Yua C.L.Li – a new genus of Vitaceae. Acta quality analysis tools. Nucleic Acids Research 25: 4876– Botanica Yunnanica 12: 1–10. 4882. Li CL. 1996. New taxa in Vitaceae from China. Chinese Tiffney BH. 1979. Nomenclatural revision: Brandon Vita- Journal of Applied Environmental Biology 2: 43–53. ceae. Review of Palaeobotany and Palynology 27: 91–92. Li CL. 1998. Vitaceae. In: Delectis Florae Reipublicae Popu- Tiffney BH, Barghoorn ES. 1976. Fruits and seeds of laris Sinicae Agendae Academiae Sinicae Edita, ed. Flora Brandon lignite. 1. Vitaceae. Review of Palaeobotany and reipublicae popularis sinicae. Vol. 48. Beijing: Science Press, Palynology 22: 169–191. 1–177. Wen J. 2007. Vitaceae. In: Kubitzki K, ed. The families and Maddison WP, Maddison DR. 2010. Mesquite: a modular genera of vascular plants. Vol. 9. Berlin: Springer-Verlag, system for evolutionary analysis, version 2.74. Available at: 466–478. http://mesquiteproject.org Wen J, Nie ZL, Soejima A, Meng Y. 2007. Phylogeny of Nie ZL, Sun H, Chen ZD, Meng Y, Manchester SR, Wen Vitaceae based on the nuclear GAI1 gene sequences. Cana- J. 2010. Molecular phylogeny and biogeographic diversifi- dian Journal of Botany 85: 731–745. cation of Parthenocissus (Vitaceae) disjunct between Asia Wilkinson HP. 1979. The plant surface (mainly leaf). In: and North America. American Journal of Botany. 97: 1342– Metcalfe CR, Chalk L, eds. Anatomy of the dicotyledons, 2nd 1353. edn. Oxford: Clarendon Press, 97–167.

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63 PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS 63

APPENDIX VOUCHER INFORMATION AND GENBANK ACCESSION OF THE MOLECULAR DATA IN THE PHYLOGENETIC ANALYSIS OF PARTHENOCISSUS

Taxon Voucher Locality GAI1 trnL-F rps16 atpB-rbcL

P. chinensis Wen 10645 (US) China, Yunnan HM223445 HM223293 HM223350 HM223405 P. dalzielii Chen 2007120501 (PE) Japan, Okinawa HM223448 HM223299 – – P. feddei ZDG 319 (KUN) China, Hunan HM223457 HM223307 HM223359 HM223416 P. henryana Nie & Meng 359 (KUN) China, Guizhou EF141265 HM223272 HM223329 HM223383 P. heptaphylla Wen 9770 (US) USA, Texas HM223419 HM223256 HM223313 HM223366 P. heterophylla Wen 10696 (US) Indonesia, Java HM223441 HM223288 HM223345 HM223400 P. laetevirens Wen 9376 (US) China, Hunan HM223425 HM223267 HM223323 HM223378 P. quinquefolia Wen 8662 (US) USA, Virginia EF141268 HM223269 HM223325 HM223380 P. semicordata Nie & Zhu 651 (KUN) Nepal, Daman HM223443 HM223291 HM223348 HM223403 P. suberosa Nie & Meng 358 (KUN) China, Guizhou HM223429 HM223273 HM223330 HM223384 P. tricuspidata Wen 7316 (US) USA, Illinois EF141271 AB235042 AB234964 AB234928 P. vitacea Wen 7157 (US) USA, Illinois EF141267 AB235036 AB234963 AB234927 Yua austro- S. Ickert-Bond 1313 (US) China, Guangdong HM223432 AB235085 AB234977 – orientalis Yua thomsoni Nie & Meng 469 (KUN) China, Sichuan EF141298 HM223277 HM223335 HM223389

© 2011 The Linnean Society of London, Botanical Journal of the Linnean Society, 2012, 168, 43–63