Molecular recircumscription of Broussonetia (Moraceae) and the identity and taxonomic status of B. kaempferi var. australis

Background Despite being a relatively small genus, the taxonomy of the paper mulberry genus Broussonetia remains problematic. Much of the controversy is related to the identity and taxonomic status of Broussonetia kaempferi var. australis, a name treated as a synonym in the floras of Taiwan and yet accepted in the floras of China. At the generic level, the monophyly of Corner (Gard Bull Singap 19:187–252, 1962)’s concept of Broussonetia has not been tested. In recent studies of Broussonetia of Japan, lectotypes of the genus were designated and three species (B. kaempferi, Broussonetia monoica, and Broussonetia papyrifera) and a hybrid (B. ×kazinoki) were recognized. Based on the revision and molecular phylogenetic analyses, this article aims to clarify these issues. Results Herbarium studies, field work, and molecular phylogenetic analyses indicate that all Taiwanese materials identifiable to B. kaempferi var. australis are conspecific with B. monoica of Japan and China. Molecular phylogenetic analyses showed that Broussonetia sensu Corner (Gard Bull Singap 19:187–252, 1962) contains two clades corresponding to sect. Broussonetia and sect. Allaeanthus, with Malaisia scandens sister to sect. Broussonetia. Conclusions Based on our analyses, B. kaempferi var. australis is treated as a synonym of B. monoica and that B. kaempferi is not distributed in Taiwan. To correct the non-monophyly of Broussonetia sensu Corner (Gard Bull Singap 19:187–252, 1962), Broussonetia is recircumscribed to contain only sect. Broussonetia and the generic status of Allaeanthus is reinstated. Electronic supplementary material The online version of this article (doi:10.1186/s40529-017-0165-y) contains supplementary material, which is available to authorized users.

is characterized by membranous stipules, globose syncarps, drupes covered by thickly sets of slender stalked bracts of various shapes, crustaceous to ligneous endocarps, and conduplicate to plane cotyledons. Although Corner (1962)'s expanded concept has been followed by most authors (e.g., Berg 1977;Rohwer 1993;Chang et al. 1998;Zhou and Gilbert 2003;Berg et al. 2006) except for Capuron (1972) who sustained the generic status of Allaeanthus, the monophyly of Broussonetia sensu Corner (1962) has not yet been tested (Zerega et al. 2005;Clement and Weiblen 2009) and much about the taxonomy of the genus remains unsettled.
Commonly known as paper mulberry, Broussonetia papyrifera is renowned as a fibrous tree essential to the development of paper making technique in ancient China around 100 A.D. (Ling 1961;Barker 2002). Long before Linnaeus' time, paper mulberry had been cultivated widely in European gardens (Barker 2002) and, as documented during Captain James Cook's circum-Pacific voyages, clonally propagated across Remote Oceanic islands by Austronesian-speaking peoples for making bark cloth (tapa), a non-woven textile that is highly symbolic of Austronesian material culture (Matthews 1996;Whistler and Elevitch 2006;Seelenfreund et al. 2010). This fast-growing dioecious weedy tree species is most likely native to China, Taiwan, and continental Southeast Asia (Matthews 1996); however, because of its long history of utilization (Matthews 1996;Barker 2002;Chang et al. 2015), considerable discrepancies exist in the literature regarding distribution ranges of B. papyrifera (Table 1). Based on the phylogeographic analysis of chloroplast ndhF-rpl32 intergenic spacer, Chang et al. (2015) demonstrated that Pacific paper mulberry originated in southern Taiwan, providing the first ethnobotanical support for the "out of Taiwan" hypothesis of Austronesian expansion. Peñailillo et al. (2016) further showed that Pacific paper mulberries are predominately female, consolidating reports on the clonal nature and corroborating Chang et al. (2015)'s inference. In addition to its long-fiber, this fast growing weedy tree has also been introduced for erosion control worldwide (Matthews 1996). Consequently, the multipurpose paper mulberry has been naturalized Table 1 Distribution of Broussonetia papyrifera in selected literatures Kanehira (1936) Taiwan, Myanmar, Thailand, Malaysia, Pacific islands, China, Japan Chûjô (1950) Japan, Korea, China, Ryukyus, Taiwan, Philippines, Vietnam, Thailand, Myanmar, India, Malay, Sumatra, Java, Borneo, SW Pacific islands, Europe, North America, Australia Liu (1962) Taiwan, India, Thailand, Malaysia, Pacific islands, Japan, China  Taiwan, Indo-Malaysia, China, Japan to the Pacific islands, Taiwan Ohwi (1965) Cultivated for making paper in Japan (Honshu, Shikoku, Kyushu); Ryukyus, Formosa, China, Malaysia  China, Japan, the Pacific Islands, Malaysia, Thailand and India Kitamura and Murata (1980) Central and southern China, Taiwan, Vietnam, Thailand, Myanmar, India, Malaysia, Pacific Islands Yamazaki (1982) S. China, Taiwan, Indochina, Thailand, Burma and Malaysia. Cultivated in Japan Yamazaki (1989) Central and southern China, Indochina, Malaysia Liao (1991Liao ( , 1996 Taiwan, Southern China, Japan, the Pacific Islands, Indochina, Malaysia, Thailand, Burma and India Liu et al. (1994), Lu et al. (2006) Central and southern China, Taiwan, Japan, Malay, Pacific islands Matthews (1996) Japan, Korea, northern, central, and southern China, Taiwan, Vietnam, Laos, Thailand, Cambodia, Myanmar, India (Sikkim), islands Southeast Asia (excluding the Philippines and Borneo), Melanesia, and Polynesia islands Florence (1997) Native to China and Japan, widely cultivated in South East Asia, Malaysia and the Pacific in southern Europe and become invasive in Argentina, Ghana, Uganda, Pakistan, the Philippines, Solomon Islands, and USA. (Matthews 1996;Barker 2002;Morgan and Overholt 2004;Florece and Coladilla 2006;Whistler and Elevitch 2006;Marwat et al. 2010;Bosu et al. 2013).
Although paper mulberry has long been introduced to Europe (Barker 2002), it is Kaempfer (1712)'s plate ("Kampf. amoen. 471. t. 472") depicting paper mulberry (as "Morus papyrifera") in Japan cited by Linnaeus (1753) that was lectotypified (Florence 1997) for Morus papyrifera L., the basionym of Broussonetia papyrifera. In Japan where paper mulberry is known as "Kajino-ki" (Okamoto 2006), B. papyrifera has long been regarded as non-native (Schneider 1917), also introduced for paper making around ca. 610 A.D. (Matthews 1996;Barker 2002). Quite confusingly, the name Kajino-ki was taken by Siebold (1830) for B. kazinoki, a name long applied to a small 'monoecious' shrub with 'globose' staminate catkins ca. 1 cm across known as Hime-kôzo in Japan (Chûjô 1950;Kitamura and Murata 1980;Yamazaki 1989;Okamoto 2006). Elsewhere, B. kazinoki is also widely found in China Zhou and Gilbert 2003), Taiwan (Liao 1989(Liao , 1991(Liao , 1996, and Korea . The natural hybrid between Hime-kôzo and Kajino-ki known as Kôzo in Japan (as B. kazinoki × B. papyrifera; Kitamura and Murata 1980;Okamoto 2006) and Daknamu in Korea  has also been long cultivated and favored by Japanese and Korean farmers for traditional paper making for centuries (Yamazaki 1989). In 2009, this natural hybrid was further named B. ×hanjiana M. Kim . The third species, B. kaempferi, is a 'dioecious' lianascent climber with 'spicate' staminate catkins ca. 1.5-2.5 cm long distributed in Japan (known as Tsuru-kôzo), central to southern China, and Vietnam (Ohwi 1965;Yamazaki 1982;Zhou and Gilbert 2003;Okamoto 2006), with a controversial record in Taiwan (Suzuki 1934;Kanehira 1936;Liao 1989Liao , 1991Liao , 1996. In the article titled ' A speciograhical revision on Broussonetia kazinoki' , Suzuki (1934) studied a set of highly variable specimens akin to "Hime-kôzo" collected from Taiwan first identified as B. kaempferi sensu Forbes and Hemsley (1894) by Hayata (1911). After comparing with specimens collected from Japan, Suzuki (1934) concluded that B. kazinoki and B. kaempferi are different species and that all the Taiwanese specimens should be collectively recognized as a distinct taxon, which he named B. kaempferi var. australis T. Suzuki. However, Suzuki (1934)'s treatment was not cited in Kanehira (1936), the most influential pre-World War II work on the woody flora of Taiwan . Instead, Kanehira (1936) followed Hayata (1911)'s treatment, identifying the entity as B. kaempferi and stating that the species is dioecious.
Interestingly, although a majority of the treatments of Kanehira (1936)'s 'Formosan Trees' were followed in the first edition of the Flora of Taiwan  and its predecessor (Liu 1962), both Liu (1962) and  treated the species as B. kazinoki, with B. kaempferi var. australis synonymized under B. kazinoki [though mistakenly typed as B. "kazinoki" Sieb. var. australis Suzuki in Liu and Liao (1976)]. Subsequently, Yamazaki (1982) revisited the issue. Yamazaki (1982) emphasized the differences in leaf shapes, adopting Suzuki (1934)'s treatment by circumscribing B. kaempferi var. kaempferi as a variety endemic to Japan and B. kaempferi var. australis a variety distributed in southern China, Taiwan, and Vietnam. Yamazaki (1982)'s treatment was adopted by most treatments of the Chinese floras (e.g., Chang et al. 1998;Zhou and Gilbert 2003;Liu and Cao 2016) with rare exceptions such as Cao (2000) in which B. kaempferi var. australis was treated as a synonym of B. kaempferi. The taxonomic status of B. kaempferi var. australis was further complicated when Liao (1989Liao ( , 1991Liao ( , 1996, in addition to B. kazinoki, reported B. kaempferi from Taiwan, with B. kaempferi var. australis again treated as a synonym of B. kazinoki. Liao (1989Liao ( , 1991Liao ( , 1996's treatment has been followed by all subsequent works of Taiwan Yang et al. 1997;Lu et al. 2006) as well as local online blogs (e.g., Nature Campus http://nc.kl.edu.tw/bbs/index.php). In a recent assessment of the conservation status of the flora of Taiwan, B, kaempferi is listed as a 'vulnerable' species with its small and declining populations (Wang et al. 2015).
Given the complicated taxonomy of these names, it is rather surprising that none of the abovementioned authors had attempted to examine and clarify type materials of the two names described by Siebold (1830) as well as B. kaempferi var. australis. After lectotypifying Siebold's Japanese plant names ,  revised the taxonomy of Broussonetia of Japan. Surprisingly, the specimen of Siebold's collections of Japanese plants that matched best to the protologue of B. kazinoki and thus lectotypified (M-0120984) turned out to be Kôzo , the natural hybrid between Hime-kôzo and Kajino-ki cultivated for traditional paper making. Consequently, B. monoica Hance, the next valid name long synonymized under B. kazinoki (e.g., Zhou and Gilbert 2003) becomes the correct name for Hime-kôzo . For B. kaempferi, the plate of 'Papyrus spuria' in Kaempfer (1712) was lectotypified . Based Ohba and Akiyama (2014)'s treatment, the four species of Broussonetia in Japan are B. kaempferi (Tsuru-kôzo), B. ×kazinoki (Kôzo), B. monoica (Hime-kôzo), and B. papyrifera (Kajino-ki).
Because  dealt only with Japanese materials, this study attempts to clarify the distribution range of B. papyrifera and resolve controversies surrounding the name B. kaempferi var. australis based on herbarium work, field observation, and molecular data. We also sampled species of Broussonetia sect. Allaeanthus which thus far has never been sampled (e.g., Zerega et al. 2005;Clement and Weiblen 2009) to test the monophyly of Broussonetia sensu Corner (1962).

Taxon sampling
Herbarium specimens of A, BM, E, GH, HAST, K, TAI, TAIF, and TNM (herbarium acronyms according to Index Herbariorum; Thiers 2016) were examined. Specimen images of Naturalis Biodiversity Center (http://bioportal.naturalis.nl/?language=en&back), the Chinese Virtual Herbaria (http://www.cvh.org.cn/), and Global Plants on JSTOR (http://plants.jstor.org/) were consulted. Fieldtrips were conducted in Taiwan, China (Zhejiang, Fujian, Guangdong, and Guangxi), and the Philippines. All voucher specimens were deposited in HAST. To expand geographic range of our taxon sampling, herbarium collections were also sampled with the permission from E, HAST, Harvard University Herbaria (A and GH), TAIF, and TNM. The HTTP URIs of the images of important (types and vouchers) specimens examined are listed in Table 2.

Molecular phylogenetic analyses
To test the monophyly of Broussonetia sensu Corner (1962), Clement and Weiblen (2009)'s aligned DNA matrix of chloroplast ndhF and nuclear 26S (TreeBASE Study ID S2229) assembled for phylogenetic analyses of Moraceae was adopted, with morphological characters of the matrix excluded. The analyses of Clement and Weiblen (2009) sampled 76 species representing 32 Moraceae genera and B. papyrifera was shown as a sister taxon of Malaisia scandens (Lour.) Planch. in the tribe Dorstenieae. All three species of sect. Broussonetia, plus B. ×kazinoki, and three of the four species of sect. Allaeanthus were sampled (Additional file 1) for phylogenetic analyses. Conditions for PCR amplification of ndhF and 26S detailed in Clement and Weiblen (2009) were followed. Phylogenetic analyses were performed using MrBayes v3.2.6 (Ronquist et al. 2012) for Bayesian inferences (BI) and GARLIC (Bazinet et al. 2014) for maximum likelihood (ML) analyses. Based on Akaike Information Criterion implemented in jModeltest 2 (Darriba et al. 2012), the models GTR + I+Γ and TVM + Γ, which were chosen in previous study (Zerega et al. 2005), were selected for 26S and ndhF, respectively. For both BI and ML analyses, the matrix was partitioned.
For ML analysis, five independent searches and 500 replicates of bootstraps were performed and results were summarized by PAUP v. 4.0a150 (Swofford 2002). For Bayesian inferences, all parameters were unlinked and estimated independently for each data partition. Two analyses were performed in parallel, each with 4 chains of 20 million generations with temperature set to 0.1, and posterior distribution was sampled every 500 generations. Model parameters and tree statistics were summarized in MrBayes and posterior probabilities higher than 0.75 were mapped to the maximum likelihood best tree manually.

Identity of Broussonetia kaempferi var. australis
Over the past few years, we have observed several wild populations in Taiwan that matched to the protologue and paratypes of B. kaempferi var. australis described in Suzuki (1934). Figure 1 summaries their morphological variation and key characteristics. Together with observations of herbarium specimens at A, BM, E, GH, HAST, K, TAI, TAIF, and TNM, we conclude that all Taiwanese materials are monoecious with globose staminate catkins ( Fig. 1c-e), the key characteristics of B. monoica sensu . We did not find any living or herbarium collections of Taiwan bearing spicate staminate catkins (Fig. 1n) that are characteristic of B. kaempferi .

Molecular phylogenetic analyses
Topologies of BI and ML analyses were identical with differences in support values. Figure 2 depicts results of ML analysis marked with both BI and ML support values.
With the additional samples of Broussonetia sensu Corner (1962), the overall phylogenetic relationships of current analyses are congruent with Clement and Weiblen (2009), with samples of Broussonetia sensu Corner (1962) placed in tribe Dorstenieae (Fig. 2). However, although the monophyly of Broussonetia sect. Allaeanthus and sect. Broussonetia were each strongly supported, Malaisia scandens was placed as the sister clade to sect. Broussonetia, rendering Broussonetia sensu Corner (1962) paraphyletic. To correct the paraphyly of Broussonetia sensu Corner (1962), we propose to reinstate the generic status of Allaeanthus Thwaites. Alternatively, an expanded Broussonetia by including M. scandens would not only necessitate further nomenclatural changes but also generate a genus with no obvious diagnostic character.
Within the clade sect. Broussonetia, all samples of Taiwan that would be identified as B. kaempferi var. australis sensu Suzuki (1934), plus the natural hybrid B. ×kazinoki, were placed in a strongly supported clade of B. monoica (Fig. 2), supporting our observations that all Taiwanese materials are part of the highly polymorphic B. monoica. All three samples of B. kaempferi formed a strongly supported clade sister to the strongly supported clade of B. monoica, with the clade of B. papyrifera further sister to the clade composed of B. kaempferi and B. monoica.
Within the clade sect. Allaeanthus, B. kurzii and B. greveana were successively sister to the clade of B. luzonica with strongest supports. Although our sampling did not include Broussonetia zeylanica (≡Allaeanthus zeylanicus), the type species of Allaeanthus, we are confident that our analysis will sustain as morphologically B. luzonica and B. zeylanica are quite similar (Corner 1962), differing from each other merely by the length of staminate catkins (10-26 cm in B. luzonica vs. ca. 6 cm in B. zeylanica) and margins of leaves (entire vs. serrate) and stipules (entire vs. denticulate).

Taxonomic treatment
Our phylogenetic analyses revealed that species of Broussonetia sensu Corner (1962) were placed in two clades corresponding to sect. Allaeanthus and sect. Broussonetia, with Malaisia scandens placed sister to the clade of sect. Broussonetia with strongest supports. To correct the paraphyly of Broussonetia sensu Corner (1962), we propose to reinstate the generic status of Allaeanthus Thwaites. Within Broussonetia sect. Broussonetia,   Fig. 2 Maximum likelihood tree based on chloroplast ndhF and nuclear 26S sequences. Bootstrap percentage ≥50 are labeled above branches. Bayesian posterior probability values ≥0.75 are labeled under branches. Linages obtained in this study are followed by collection sites (Country: locality), collectors and original collection numbers. All Taiwanese samples of Broussonetia monoica (collection sites in green) would be identified as B. kaempferi var. australis sensu Suzuki (1934)