Novelties in Brazilian Tradescantia L. (Commelinaceae)

Abstract We present a new record to the Brazilian territory (i.e. Tradescantia boliviana), the rediscovery of a species exclusively known from the cultivated type collection (i.e. T. valida), the description of a new taxon (i.e. T. chrysophylla), synonyms for T. crassula and T. boliviana, correct the typification of T. crassula, and designation of a lectotype for T. ambigua and T. ambigua var. pilosula. Furthermore, we present illustrations, comments, distribution maps, and identification keys for the studied taxa.


Results
We update the number of species of Tradescantia known from Brazil to 14 (from 12), with the description of a new species, a new record, two new synonyms, and the rediscovery of T. valida G.Brückn in the wild. We present complete descriptions of the new species and the rediscovered taxon, and detailed diagnoses for the other three studied species. The results are organized in the treated sections (i.e. T. sect. Austrodescantia, T. sect. Campelia, T. sect. Mandonia, and T. sect. Zebrina). Furthermore, as a result of our improved knowledge of Tradescantia, we present an updated, illustrated, and more functional identification key for the sections occurring in Brazil.
Comments. Tradescantia sect. Austrotradescantia has been the subject of several recent studies (Pellegrini 2015(Pellegrini , 2016(Pellegrini , 2017Pellegrini et al. 2015Pellegrini et al. , 2016Funez et al. 2016). Species diversity in this section is centered in Southeastern and Southern Brazil, where all of the accepted species occur. However, some species in the group have a wider distribution, reaching neighboring countries like Argentina, Bolivia, Paraguay, and Uruguay (Pellegrini 2015;Pellegrini et al. 2016). Tradescantia sect. Austrotradescantia has recently been revised, and its morphology thoroughly analyzed in a yet unpublished MSc thesis (Pellegrini 2015). As part of our revision of this section, we describe a new species, report the rediscovery and inclusion of T. valida in the section, and a new synonym for T. crassula (with a correction of its typification).  Conservation status. Tradescantia crassula possesses a wide EOO (ca. 408,686.868 km 2 ). Thus, following the IUCN recommendations (IUCN 2001), it should be considered Least Concern (LC).

Tradescantia crassula
Nomenclatural notes. Funez et al. (2016) indicate that Pellegrini (2015) erroneously designated the specimen Sellow 3033 (B100521014) as the lectotype for T. crassula. However, their affirmation is incorrect according to the Code since the thesis lacks either a ISSN or an ISBN, and was never distributed to the general public (McNeill et al. 2012, Art. 29.1). The work cited by Funez et al. (2016) was a draft version of the first author's unpublished M.Sc. thesis, with many incomplete and partially incorrect data (Pellegrini, unpublished data), and therefore is not considered a effective publication by the Code. Furthermore, according to Art. 30.8 (McNeill et al. 2012), any thesis published on or after 1 January 1953 and stated to be submitted to a university for the purpose of obtaining a degree, does not constitute effective publication; unless it contains any kind of statement or other internal evidence that it is regarded as an effective publication by its author or publisher. Since no statement is made in the final version of the thesis (i.e. Pellegrini 2015), the publication does not meet any of the Code's requirements, and therefore cannot be treated as effectively published. In the final version, Pellegrini (2015) gives detailed information on the typification of T. crassula, which is effectively published here and corrects the typifications by Funez et al. (2016).
The date written on the original label, and treated by Funez et al. (2016) as the collections date, is "Dec. 1836". Also, it is possible to see in the label the names of Sellow and Humboldt. According to Stafleu and Cowan (1985), Friedrich Sellow lived from 1789-1831, and collected plant specimens in Southern Brazil and Uruguay 1819-1831, funded by Humboldt. This easily explains why both botanists are mentioned in the original label, and why the collector is to be considered as Sellow, instead of Humboldt or both botanists. Furthermore, given that Sellow died in 1831 and his expeditions were done just before his death, it would be impossible for "Dec. 1836" to represent the actual collection date. We believe this date might correspond to the date when this specimen was acquired by Kunth, and placed into his personal herbarium. Finally, Link and Otto (1828) make direct reference to their new species being based on Sellow collections. According to the Code (McNeill et al. 2012, Art. 9.2), the Sellow 3033 (B100521014) specimen is a suitable choice for a lectotype, superseding the lectotypification of the original illustration, done by Funez et al. (2016). The epitypification by Funez et al. (2016) should be disregarded entirely because the original illustration is informative enough to correctly apply the name T. crassula. All the diagnostic features of this species (see comments below) are visible and sufficient for appropriate diagnosis, in the original illustration.
Taxonomic notes. The species in this section are especially variable morphologically and when in cultivation or growing in shaded areas they can change their vegetative morphology quite drastically. Few characters in the vegetative organs were observed to be constant in the T. crassula group and thus are of little taxonomic relevance. This can be exemplified by the phyllotaxy and pubescence of the leaf-blades, which have been shown to vary greatly due to ecological features (Pellegrini 2015(Pellegrini , 2016. On the other hand, the pubescence of the internodes, leaf-sheaths, and of the margin of the leaf-blades were observed to be constant and reliable for species delimitation. As previously indicated in other Tradescantia sections (Anderson and Woodson 1935), the pubescence of the pedicels and sepals seem to be highly stable within each species, easily observable, and thus, reliable for species delimitation. As aforementioned, T. crassula is highly variable in vegetative morphology. All studied individuals always presented glabrous leaf-blades, setose sepals with long hairs along the keel, and white petals. The specimens cited by Funez et al. (2016) as representing T. schwirkowskiana, fit perfectly the circumscription adopted by us for T. crassula, showing variation only in the degree of branching of the stems, color of the leaf-blades, and degree of impression of the secondary veins. All of this morphological variation can be easily explained by the ecological features of the area where the specimens were collected (i.e. shaded moist forests in mountainous regions from the state of Santa Catarina). Aside from that, the authors state that T. crassula possesses spirally-alternate leaves and a rhizomatous base. While developing our studies for the taxonomic revision of T. sect. Austrotradescantia and a morphological phylogeny for the genus (Pellegrini 2015), we analyzed 50% of the species in the genus and observed that all species of Tradescantia produce spirally-alternate leaves when young. This feature is normally lost during development of most species of T. sect. Austrotradescantia, but is always retained by T. valida G.Brückn. (see comments below), sometimes retained by T. cerinthoides Kunth (Pellegrini 2015(Pellegrini , 2016, and rarely retained by T. crassula (Pellegrini, pers. obs.). No species in Tradescantia were observed to produce rhizomes (Pellegrini 2015). The only known drought resistance strategy observed in the genus was the production of tuberous roots; present in all species of T. sect. Mandonia, T. sect. Parasetcreasea, T. sect. Separotheca, T. sect. Setcreasea and sect. Tradescantia, and exclusively in T. commelinoides Schult. & Schult.f. from T. sect. Cymbispatha (Pellegrini 2015 Diagnosis. Similar to T. cymbispatha due to its habit with an indefinite base, creeping stems with ascending apex, sessile succulent leaves with flat blades homogeneously covered by indumenta, inconspicuous secondary veins, saccate cincinni bracts, broadly ovoid floral buds, sepals without keels, and pistil the same length as the stamens. It can be differentiated by its velutine to hispid, golden to light brown indumentum covering almost the entire plant, strongly unequal cincinni bracts, and pedicels and sepals glandular-pubescent, or with a mixture of glandular and eglandular hairs. Type. BRAZIL. São Paulo: Biritiba Mirim, Estação Biológica de Boracéia, fl., 24 Nov 1983, A. Custódio Filho 1910 (holotype: RB!; isotype: SP!).
Specimens seen ( Etymology. The epithet "chrysophylla" means golden leaves and is given after the golden hairs that cover the whole plant, but especially the leaves. Distribution and habitat. Tradescantia chrysophylla is endemic to Brazil, more precisely to the states of Rio de Janeiro, São Paulo, Paraná and Santa Catarina (Fig.  8). It can be found growing as a terrestrial, rupicolous or as an epiphyte, understory in shaded and moist forests.
Phenology. It was found in bloom and fruit from August to December, but peaking during October.

Tradescantia valida
Phenology. It was found in bloom and fruit in December and April. Conservation status. Tradescantia valida is only known from the cultivated type collection and collections in Jaguari, state of Rio Grande do Sul. Thus, in accordance with the IUCN recommendations (IUCN 2001), it should be considered as Data Deficient (DD), until further information on the species becomes available.

Discussion.
Tradescantia valida is closely related to the remaining three species in the T. crassula group, due to its erect habit, definite base, sessile, conduplicate to falcate, succulent leaves, generally with inconspicuous secondary veins, cincinni bracts non-saccate at base, pistil longer than the stamens, hilum longer than ½ the length of the seed, and for preferentially inhabiting open areas and rocky outcrops (Pellegrini 2015(Pellegrini , 2016. Tradescantia valida can be differentiated easily from all remaining species of the T. crassula group by its spathaceous cincinni bracts and by the presence of supernumerary bracts. The presence of spathaceous cincinni bracts is a character previously reported in T. sect. Austrotradescentia exclusively for T. umbraculifera Hand.-Mazz., a member of a clade named T. fluminensis group by Pellegrini (2015), and thus quite remarkable in this distantly related species. Aside from that, the presence of supernumerary bracts is unique within T. sect. Austrotradescantia, but well-known in species from T. sect. Cymbispatha (Pellegrini 2015;Pellegrini et al. 2016) and T. sect. Mandonia (Grant 2000;Pellegrini 2015).
In the T. crassula group, T. valida is similar to T. cerinthoides due to its sepals without dorsal keels. However, they can be easily differentiated due to its generally linear elliptic to linear lanceolate to lanceolate leaf-blades (vs. elliptic to broadly elliptic or ovate to broadly ovate or obovate to broadly obovate, in T. cerinthoides), glabrous with margins setose at the base or until the middle (vs. pubescent on both sides or only abaxially, rarely glabrous on both sides with ciliate margins), and pedicels and sepals glabrous or only sparsely pubescent with eglandular hairs (vs. evenly densely velutine to hispid, sometimes with a mixture of glandular and eglandular hairs). Tradescantia valida is much more similar to T. crassula and T. seubertiana M.Pell., due to their leaf-blades and floral pubescence. These species can be easily differentiated by the pubescence of the margin of their leaf-sheaths (ciliate to shortly-setose in T. crassula; glabrous in T. seubertiana; and long-setose in T. valida), the pubescence of their sepals (long-setose along the keels in T. crassula; glabrous in T. seubertiana; and glabrous or with few hairs at the apex in T. valida), and by the shape of their floral buds (broadly ovoid T. crassula; ellipsoid in T. seubertiana; and ellipsoid in T. valida).
Comments. Tradescantia sect. Campelia is monospecific and represented by T. zanonia (L.) Sw. It was considered by Hunt (1986) to be unique within the genus due to its fleshy pedicel and sepals covering the capsule, giving the fruit a berry-like appearance, which is consumed by birds and other small animals (Hunt 1986;Pellegrini, pers. obs.). Nonetheless, the morphological similarity to Tradescantia sect. Zebrina is indisputable, as pointed out by Hunt (1986) and here reaffirmed by us. Tradescantia sect. Zebrina is highly variable in the following characters: (1) the position of the inflorescence; (2) if it perforates the leaf-sheaths or not; (3) the degree of conation between the sepals, the petals; and (4) the degree of connation between the petals and stamens. Thus, both sections are distinguished solely based on the consistency of their calyx (Pellegrini, pers. obs.). Diagnosis. The section is characterized by perennial herbs, with tuberous roots, definite base, without rhizomes, leaves with symmetric or slightly asymmetric base, inflorescences mainly axillary, sessile, cincinni bracts much reduced or rarely leaf-like in the terminal inflorescences, bracteoles inconspicuous, flowers flat, rarely tubular, sepals equal, free, not keeled, petals free, sessile, stamens 6 and equal, free or epipetalous, filaments coiling at post anthesis, medially sparsely bearded with moniliform hairs, connectives quadrangular to rectangular, rarely rhomboid, anther sacs C-shaped, ovary pubescent, stigma truncate to capitulate, seeds scrobiculate to rugose, rarely costate, embryotega conspicuous and dorsal (Hunt 1980;Pellegrini 2015). Comments. Tradescantia sect. Mandonia is a poorly understood group, currently represented by 12 species (Hunt 1980(Hunt , 1986b(Hunt , 2007Grant 2000Grant , 2004Zamudio et al. 2013). This section possesses a very interesting disjunctive distribution, being restricted to Seasonally Dry Tropical Forests throughout the Neotropics. Species delimitation is especially complicated in this this group, due to great vegetative plasticity within species, and conserved reproductive features. Most species can be easily identified based on their allopatric distributions, but hardly differentiated based solely on their morphological features (Pellegrini, unpublished data). In South America, T. sect. Mandonia is currently represented exclusively by two species, T. ambigua Mart. ex Schult. & Schult.f. and T. boliviana. Tradescantia boliviana has hitherto been considered exclusive to Argentina, Bolivia, Paraguay, and Peru (Grant 2004). However, after analyzing several herbaria we came across specimens representing T. boliviana that reached the Brazilian territory, in the state of Mato Grosso do Sul. Thus, we present an identification key differentiating both species, illustrations, a distribution map, and the needed comments and typifications. Least Concern (LC). However, when taking the AOO into consideration, it is considerably reduced (ca. 216.000 km 2 ). Furthermore, The Caatinga and Cerrado domains are greatly threatened by human activities such as deforestation, cattle breeding, and various crops. Thus, following the IUCN (2001) recommendations, T. ambigua should be considered Endangered [EN,A2c+B2ab(ii,iii,iv,v) Nomenclatural notes. When describing T. ambigua, Schultes and Schultes (1830) mention that their new species is based on a Martius specimen from the Province of Bahia, but with no reference to herbaria or collector number. While consulting M we came across a specimen (Martius 140) matching perfectly the protologue, and annotated in Martius's handwriting. Thus, it is here designated as the lectotype.
When describing T. ambigua var. glabriuscula, Clarke (1881) only lists the collection Gardner 2334. However, Clarke makes no reference to the herbaria in which this specimen might be distributed. After analyzing several collections, we found specimens of Gardner 2334 at BM and K. According to Stafleu and Cowan (1976), Clarke had access to both collections, and many of the type specimens for names described by him were commonly from either of the two herbaria. The specimen at K is composed of a much longer stem, containing a greater number of leaves, inflorescences and flowers. Added to that, it is annotated on Clarke's handwriting, making it the obvious choice for a lectotype.
Morphological notes. Tradescantia ambigua, as most species in T. sect. Mandonia, presents a high degree of morphological variation. The plants vary greatly in size, branching of the stem, shape of the leaves, pubescence of the leaves, pubescence of the pedicels and sepals, and shape and color of the petals. Nonetheless, this variation has no obvious geographical pattern and seems rather random across its distribution range. Thus, don't recognize any infraspecific taxa for T. ambigua. , and based solely on this criterion it should be considered Least Concern (LC). Nonetheless, its AOO is considerably reduced (ca. 172.000 km 2 ), added to that fact that most of the studied specimens are at least more than 20 years old. Thus, following the IUCN (2001) recommendations, T. boliviana should be considered Endangered [EN,A2bcd+B2ab(ii,iii,iv,v) Nomenclatural notes. In the protologue of T. ambigua var. pilosula, Hoehne (1915) mentions two collections of his own when describing this new variety. The author makes no reference as to the herbarium in which each specimen is placed or to the existence of possible duplicates. According to Stafleu and Cowan (1979), Hoehne's types are generally at SP with duplicates in several other herbaria. However, after two visits to the SP we were unable to locate any of the specimens (Hoehne 4499,4723). Nonetheless, Stafleu and Cowan (1979) make an important remark that until 1917, Hoehne was living and working in Rio de Janeiro. After analyzing the collection of R, we came across both specimens placed in the general collection. Both specimens were annotated by Hoehne, but the specimen Hoehne 4499 possesses a beautiful illustration attached to it, showing the details of the plants' floral morphology. Thus, it is designated by us as the lectotype for T. ambigua var. pilosula.
Morphological notes. Tradescantia boliviana is a morphologically variable species across its distribution. Nonetheless, in the same way as T. ambigua, there is no obvious geographical pattern in this variation. The presence of glandular hairs in the pedicels and sepals can be observed in some of the individuals, but aside from that they don't seem to differ in any other aspect from the other specimens. This variation is peculiar, but not unrecorded in the genus, and a similar scenario is described by Pellegrini (2015Pellegrini ( , 2016 for T. cerinthoides (T. sect. Austrotradescantia) and by Faden (1993) for T. crassifolia (T. sect. Mandonia).
After analyzing the type specimens for T. ambigua var. pilosula, we noticed that the pedicels and sepals are hispid, the connectives are quadrangular, and the ovary velutine to sparsely velutine at apex. Added to that, the distribution of the specimens collected by Hoehne is congruent with the distribution of T. boliviana, but disjunctive from T. ambigua. Thus, we consider T. ambigua var. pilosula a synonym of T. boliviana.  Diagnosis. The section is characterized by perennial herbs, with thin fibrous roots, definite or indefinite base, without rhizomes, leaves with symmetric to asymmetric base, inflorescences terminal or axillary, pedunculate, cincinni bracts spathaceous, bracteoles conspicuous and linear, flowers tubular, sepals unequal, basely to completely conate, keeled or not, petals free or conate, long-clawed, stamens 6 and subequal, epipetalous, filaments straight at post anthesis, medially sparsely bearded with moniliform hairs, connectives sagittate to linearly-tapered, anther sacs round, ovary glabrous, stigma capitate, seeds rugose, embryotega inconspicuous and semilateral (Hunt 1986;Pellegrini 2015). Comments. Tradescantia sect. Zebrina is a small group, composed of ca. five species, ranging from Mexico to Venezuela. Tradescantia zebrina Heynh. ex Bosse is widely cultivated worldwide, and occurs in Brazil as an invasive species (Hunt 1986;Pellegrini 2017). As aforementioned, the section is small but morphologically diverse, being poorly differentiated from T. sect. Campelia and T. sect. Corinna. As stated by Hunt (1986), these three sections seem to blur into one another, with several species being originally assigned to one group and subsequently transferred to another.

Conclusion
Tradescantia, like many other genera in Commelinaceae, is a taxonomically complicated and morphologically diverse genus. In order to safely propose taxonomic novelties, it is necessary to possesses a broader knowledge on the group, especially regarding the morphological plasticity within each species. This can only be achieved with extensive field and herbaria research, complemented with the cultivation and observation of some individuals. Many recent studies of Brazilian Commelinaceae have been narrowly focused, and proposed new species and several typifications Hassemer et al. 2016aHassemer et al. , 2016bHassemer 2017). As demonstrated by us in the present work, this can lead to the unnecessary description of new names, causing the inflation of accepted species and their conservation assessments. Perhaps the most unfortunate result of such studies is the potential for incorrect typification and application of names (e.g. Funez et al. 2016;Hassemer et al. 2016b;Hassemer 2017). Thus, we strongly suggest that future typifications and descriptions of new taxa in Commelinaceae be carried out as part of a broader and more detailed taxonomic framework.