Two new species of Siphocampylus (Campanulaceae, Lobelioideae) from the Central Andes

Abstract Two species of Siphocampylus (Campanulaceae: Lobelioideae) from the Central Andes of Peru and Bolivia are described, illustrated, and discussed with reference to related species. One species, Siphocampylus antonellii, is endemic to high elevation grasslands of Calca, Peru, while the second, Siphocampylus siberiensis, is endemic to cloud forests of Cochabamba, Bolivia. Both species are robust shrubs that produce tubular pink flowers that are likely pollinated by hummingbirds.


Introduction
While the subfamily Lobelioideae Burnett of Campanulaceae Juss. is cosmopolitan in its distribution, more than half of its ~1200 species are restricted to the Neotropics (Lammers 2007). Most of these species (~550/680) belong to the Andean-centered centropogonid clade, which comprises Burmeistera Triana (~120 species), Centropogon C. Presl (~210 species), and the mainland species of Siphocampylus Pohl (~220 species). Th is clade is unique within Lobelioideae for its combination of Neotropical distribution, woody habit, entire corolla tube (i.e., neither fenestrate nor dorsally cleft), and fl oral adaptation to pollination by either hummingbirds or nectar bats (Lagomarsino et al. 2014;Lammers 2002;Muchhala and Lammers 2005). Within the centropogonid clade, Siphocampylus can further be distinguished by its capsular (vs. baccate) fruit. Th e genus, however, is not monophyletic (Antonelli 2008;Knox et al. 2008): in addition to the Caribbean species, which are distantly related to the centropogonid clade, its mainland species form at least 11 subclades that are polyphyletic with respect to the berry-producing Centropogon . Th is result is not surprising given the high degree of character overlap between Centropogon and Siphocampylus, which has frequently resulted in the incorrect fi ling of material in herbaria (Gleason 1921;Lammers 1998) and the occasional description of a new species to the incorrect genus or of uncertain placement (e.g., C. dubius [Zahlbr.] E. Wimm. for both scenarios) (Gleason 1921). Th e non-monophly of these genera is largely due to the dynamic evolution of fruit-type within the clade: fl eshy fruits have evolved approximately eight times from dry-fruited ancestors, likely as a result of migration to densely forested habitats (Givnish et al. 2005(Givnish et al. , 2009Lagomarsino et al. 2014). While further phylogenetic information is necessary to begin to re-delimit generic boundaries to refl ect evolutionary relationships within Neotropical Lobelioideae, Siphocampylus species are easily identifi ed to genus within the confi nes of the current classifi cation system, especially when fruiting material is available. Despite its non-monophyly, Siphocampylus remains a conspicuous, if poorly studied, component of the cloud forests and high elevation grasslands that comprise much of the tropical Andean global biodiversity hotspot (Myers et al. 2000). Future taxonomic and phylogenetic work focused on the centropogonid clade will undoubtedly result in the description of many new species and discovery of new clades defi ned by synapomorphies.
Here we describe two new species of hummingbird-pollinated Siphocampylus. Type specimens for both species were included in the most recent molecular phylogeny of the centropogonid clade, which is based on fi ve plastid markers and includes relatively dense taxon sampling that spans all recognized taxonomic divisions, geographical occurrences, and morphological variation within Neotropical Lobelioideae . Th e phylogeny provides the information necessary to discuss these new species in relation to their closest evolutionary relatives. We additionally discuss similarities and diff erences of the species placed closely to their relatives in the most recent monograph of Siphocampylus (Wimmer 1943(Wimmer , 1953(Wimmer , 1968.

Siphocampylus antonellii
Distribution and habitat. Siphocampylus antonellii is endemic to Peru, where it grows on rocky slopes in puna habitat at ~3800 m in elevation. It is only known from the type collection.
Phenology. Individuals were collected in fl ower in December; the rest of the phenology of this species remains unknown.

Figure 1. Siphocampylus antonellii.
A Flowering branch B Leaf, abaxially, including detail of leaf margin and stellate hairs that cover surface C Staminate-phase fl ower, including bibracteolate pedicel, with detail of sepal and stellate hairs that cover the outer corolla surface D Longitudinal section of a pistillate-phase fl ower, showing the insertion of staminal tube to corolla, style and stigma as situated relative to the stamens, and bilocular ovary with axile placentation E Corolla lobe detail F Detail of anther tube, including apical hairs on ventral anthers, and stigma. Drawing by Bobbi Angel from the type.
Etymology. It is an honor to name this attractive species for Dr. Alexandre Antonelli (1978-), a biogeographer and phylogeneticist at the University of Gothenburg. Antonelli has made many important contributions to our understanding of Neotropical biodiversity through space and time, and to the evolution of various taxa, including Lobelioideae. His eff orts in the latter brought the second author to the fi eld in Costa Rica in 2005 in search of Campanulaceae, and helped to inspire the fi rst author to study Centropogon, Siphocampylus, and Burmeistera.
Conservation status. Siphocampylus antonellii is endemic to a narrow stretch of high-elevation grassland (puna) in Calca, Peru, where it is locally abundant. Only a single population of this species is known, from which the type collection was made. Due to its small area of occurrence and the threat of future deforestation in its habitat, we tentatively consider this species to be Vulnerable (IUCN 2014). Its vulnerable status is further justifi ed by its roadside occurrence, near major construction eff orts.
Discussion. Siphocampylus antonellii is most similar to S. elfriedii E. Wimm. (Fig.  3C) and S. parvifolius E. Wimm., which are both also endemic to Peru. Th ese species share a shrubby habit ( Fig. 2A) Table 1 and by the agreeable lemon-like smell that it emits.

Siphocampylus siberiensis
Distribution and habitat. Siphocampylus siberiensis is endemic to Bolivia, where it has been collected at the edge of the road at ca. 2700-2900 m in elevation in cloud forest.
Phenology. Individuals were collected in fl ower and fruit in December and in fl ower only in April; the rest of the phenology of this species remains unknown.
Etymology. Th e specifi c epithet of this species refers to the type locality, the Serranía de Siberia, a mountain range at the limit between the Cochabamba and Santa Cruz departments in Bolivia.
Conservation status. Siphocampylus siberiensis is known only from a single population in Serranía de Siberia in central Bolivia; this population is represented by the two cited collections. Th is species appears to be locally rare, and only one individual was encountered during our fi eldwork. Due to its small area of occurrence and the threat of future deforestation in its habitat, we tentatively consider this species to be Vulnerable (IUCN 2014). Its vulnerable status is further justifi ed by its roadside occurrence in montane cloud forest, a habitat type known to be particularly sensitive to human encroachment.

Discussion.
Siphocampylus siberiensis can be recognized by its shrubby habit with multiple stems arising from a single point; leaves aggregated at the apex of branches (Fig. 4A); conspicuous venation (especially on the abaxial leaf surface) (Figs 4A, 5C); solitary fl owers borne in the axil of leaves (Fig. 4A); light pink corolla with a tube that is cylindrical for its entire length (i.e., not basally constricted) (Figs 4B, D, 5F); anther tube that is densely pubescent in the sutures between individual anthers (Figs 4E, 5F); and fruits that are both ribbed and lobed (Figs 4F, 5E).
Molecular phylogenetic analysis places S. siberiensis in a clade that includes S. tunarensis Zahlbr., S. tunicatus Zahlbr., and S. umbellatus (Kunth) G. Don; this clade is closely related to S. boliviensis Zahlbr. and S. sparsipilus E. Wimm.   (Figs 5, 6). Th ese species are all restricted to the Central Andes of Peru and Bolivia, with the exception of S. umbellatus, whose range also extends to Brazil. Th is clade is composed of robust shrubs or trees that are exceptionally tall for the centropogonid clade (Figs 5A, 6E-F), or rarely scandent subshrubs (S. sparsipilus and some collections of S. boliviensis), with ebracteolate pedicels, a shallow, hemispherical hypanthium (turbinate in S. tunarensis), and leaves that leave prominent scars after abscission (Fig.  4A) and have dentate margins and reticulate venation that is conspicuous on both surfaces, but especially the abaxial surface (Figs 4A, 5C). Both bright pink (S. siberiensis, S. boliviensis, S. sparsipilus) and dull colored (S. tunarensis, S. tunicatus, S. umbellatus) corollas are represented in this clade. Th is color variation, which is associated with diff erent gross corolla morphologies, likely refl ects adaptation to diff erent pollinators (hummingbirds and bats, respectively) (Figs 5F-G, 6A-D).
Even though they are not the most closely related species, the pink, narrow fl owers of S. siberiensis most closely resemble those of S. boliviensis and S. sparsipilus. However, the latter two species can be easily distinguished by their corollas that are constricted at the base (vs. not constricted) and much rounder hypanthium (vs. fl attened at top). Th e other species in the immediate clade that includes S. siberiensis diff er in their dull reddish (S. tunarensis  (Fig. 6C). Siphocampylus tunarensis can be separated by its linear, revolute corolla lobes (Fig. 6B) Fig. 5F]). Furthermore, while S. siberiensis is a robust shrub 3-4 m tall, S. tunarensis can grow to be a very tall tree (>10 m) with a diameter of more than 30 cm and is possibly one of the largest species of Campanulaceae in the Americas (Fig. 6E).
Th e species that is most superfi cially similar to S. siberiensis, S. boliviensis, is placed in the same couplet as S. macrostemon A. DC. in the dichotomous key to the members of the genus in Wimmer (1953). Th is markedly diff erent species, which has not yet been sampled in molecular phylogenetic analysis, can be distinguished by its subsessile  leaves (vs. pedicels 0.3-0.7 cm long) that are smaller (5-8 × 1.31.5 cm vs. 10.5-19.5 × 2.9-4.6 cm) and sparsely pubescent on the adaxial surface (vs. tomentose), minutely dentate leaf margins (vs. doubly dentate and ciliate), shorter pedicels (2.6-4.2 cm vs. 4.5-9.4 cm) that are bracteolate (vs. ebracteolate), and glabrous corolla (vs. pubescent). Th e other species most closely related to S. siberiensis fall into many disparate taxonomic units within the current classifi cation scheme of the genus. Th is makes it diffi cult to place this new species in the context of Wimmer's taxonomy; this is likely due to this treatment's reliance on single, often arbitrary characters to designate groups.
Th e measurements of the calyx and corolla in parentheses correspond to E. Fernández et al. 3583 (MO). Th is specimen apparently has a white corolla, but otherwise corresponds to the species concept for S. sibieriensis presented here.