Sequence data from 94 terminals (91 species), belonging to 20 putatively neotropical Vaccinieae genera, were analyzed. The sampling strategy followed that of Pedraza-Peñalosa et al. (2013), but with emphasis on taxa of Andean origin. Eighteen species endemic to Colombia were newly sequenced and for five other species that also grow outside Colombia, a population from Colombia was chosen. The sampled taxa exemplify different aspects of reproductive and vegetative morphology. They also represent the major clades recovered within Neotropical Vaccinieae in previous phylogenetic analyses, and also include species from the Caribbean/Mesoamerican clade. Emphasis was placed on sampling Satyria, Macleania, Psammisia, Thibaudia and Cavendishia Lindl. For polymorphic species hard to identify or those whose variation is insufficiently known, more than one specimen was sequenced. For the plastic Satyria grandifolia Hoerold, two specimens recently collected in LONP (NW Colombia), each representing a separate morphospecies, were sequenced. These were analyzed along with the sequence already available in GenBank and originally collected in SW Colombia. Two specimens were also sequenced of the very rare and morphologically insufficiently understood Psammisia mediobullata Luteyn & Sylva, endemic to a small region of NW Antioquia (Colombia). Trees were rooted with Gaylussacia baccata (Wangenh.) K.Koch, a species from a genus of extra-neotropical origin that is basal with respect to all New World taxa (Kron et al. 2002).

A combination of nuclear (nrITS, 651 aligned bp.) and plastid (5’ end of ndhF, ca. 1225 aligned bp.; matK, 1331 aligned bp.) markers were selected because of their number of phylogenetically informative characters in previous studies (Kron et al. 2002, Powell and Kron 2003, Pedraza-Peñalosa 2009, 2010a, Pedraza-Peñalosa et al. 2013). All procedures used during the DNA extraction and sequencing have been published previously (Pedraza-Peñalosa 2009). Sequences were edited with Sequencher 5.2.3 (Gene Codes Corporation). For this study, 93 new molecular sequences were produced (permits DTSA 033 SFF Galeras y otros; Acceso a Recursos Genéticos Res. 734 de 30 de Abril de 2007; 35-2005-INRENA-IFFS-DCB), the rest were gathered from GenBank; all accession numbers are provided in Table 1.

A multiple sequence alignment was produced using MUSCLE (Edgar 2004) and the model of sequence evolution was estimated with jModelTest 2 (Guindon and Gascuel 2003, Darriba et al. 2012). Maximum Likelihood (ML) estimation of the phylogenetic relationships was conducted using RAxML (Stamatakis 2006), employing one thousand replicates with stepwise random taxon addition, starting MP trees, the GTRGAMMA model of evolution, and 1000 bootstraps (BS) inferences. The best ML tree was visualized with FigTree 1.3.1 (http://tree.bio.ed.ac.uk/). The ML analyses were performed for each individual loci, partition (nuclear vs. plastid), and lastly, for the entire dataset.

This HtAP subclade includes the majority of the sampled Satyria and is sister to Thibaudia costaricensis Hoerold. Satyrias. s. is strongly supported (100% bootstrap; Fig. 1); it includes the type of the genus, S. warszewiczii. Satyria leucostoma Sleumer forms a tritomy with the rest of Satyrias. s. The only well-supported groupings within Satyrias.s. are species pairs and a small subclade made up by taxa from Central America (96% bootstrap). Two of the sampled new species from LONP, S. sp. nov. 1 and S. sp. nov. 2, are placed within Satyrias. s.

In this clade (100% bootstrap; Fig. 1), Cavendishia martii (Meisn.) A.C.Sm. is strongly supported as sister to the rest. The only other well supported clade is the C. micayensis–C. lindauiana clade (≥ 94% bootstrap). Thibaudia jahnii S.F. Blake, from Venezuela, is included, and is sister to the Ecuadorean C. grandifolia Hoerold (support weak).

Within this clade (100% bootstrap; Fig. 1), Thibaudia macrocalyx–T. pachyantha form a cluster (74% bootstrap) sister to a clade (91% bootstrap) that contains three species of Satyria mixed with other species of Thibaudia.

Within the Andean clade B, Psammisia is split in two clusters (Fig. 1). The largest of the groupings, Psammisia I clade (< 50% bootstrap), is part of a tritomy that also contains Ceratostema (< 50% bootstrap) and the Psammisia II + Macleania clade (86% bootstrap). Psammisia II (97% bootstrap) (including Thibaudia inflata Luteyn) is sister to Macleania (96% bootstrap). Neither Ceratostema nor Macleania are monophyletic as C. reginaldii (Sleumer) A.C.Sm. is derived within Macleania. The other two sampled new species from LONP are resolved within Psammisia I.

This strongly supported group (99% bootstrap; Fig. 1) is among the largest Andean Vaccinieae subclades with an estimated 170 species, most of them Cavendishia. Heteroandrous taxa are very distinctive within Vaccinieae and are characterized by having strongly unequal stamens in which the filaments and/or anthers of adjacent stamens (each corresponding to one of the two staminal whorls) alternate in length. When there are differences in anther size, dimorphism can be also expressed in anther shape (Satyrias.s.) and orientation of the tubule aperture (Satyrias.s., Cavendishia). However, differences in this latter feature are poorly documented in species descriptions and are at times slight, and neither its extent nor its consistency is clear.

A close evolutionary relationship between Satyrias.s. and Cavendishia was first proposed by Smith (1932) who suggested that even though the genera are distinct, they form a separate group within Vaccinieae having staminal dimorphism in common. However, the nature of the dimorphism is quite different in each group, as pointed out below, and it should be further examined. Smith included the also dimorphic Orthaea within this group, but Andean species of Orthaea were not included in this analysis as it is currently being studied in detail (revision in preparation by N. R. Salinas).

Lastly, the HtAP species are characterized by having the pedicel articulate with the calyx (seen as a constriction at the point of attachment), a homoplasic feature that is traditionally used as part of the key characters useful to tell genera apart.

Thibaudia costaricensis is sister to Satyrias.s., a clade that is in agreement with the molecular circumscription of Satyria by Powell and Kron (2003). Both taxa have connate filaments (Fig. 1). Satyrias.s. species have in common markedly alternately unequal stamens with rigid dimorphic anthers, the longer of which have flaring tubules that are often ornamented or recurved into hooks; their filaments are equal in length (Powell and Kron 2003). Powell and Kron (2003) sampled 11 out of the 22 then recognized species and noted that the Colombian Satyria (10 species, 9 of them endemic) needed to be included in future studies to better elucidate evolutionary patterns. Seven Colombian species, including two new to science, are added to this analysis for a total of 18 Satyrias. l.

Fifteen morphospecies are placed within Satyrias.s., mostly South American (Fig. 1). However, despite the increased taxonomic coverage, support at the basal nodes is weak and the only well-supported major subclade is that of the Central American species (S. allenii–S. meiantha, 96% bootstrap). The derived position of these Central American species indicate they are the result of relatively recent dispersals from a South American ancestor. However, in the case of S. panurensis (Benth. ex Meisn.) Hook. f. ex Nied., a species found from Mexico to southern Bolivia (including the Guianese Shield to the east) and not included within the Central American clade of Satyrias.s., the direction of the migration of is not clear.

Although the new species are endemic to the same general region, have similar corolla colors and shape, and are the only Satyrias.s. known to have an ornamented calyx (winged and/or lobed), they are not sister species. Satyria sp. nov. 1, sister to a clade containing species from both Central and South America, is easily differentiated from all other Satyrias.s. because of its pseudoverticillate leaves. S. pilosa A.C.Sm., another newly sequenced species, present in Antioquia but also beyond, being endemic to the greater Chocó biogeographic region, is sister to S. sp. nov. 2 (87% bootstrap).

Some of the molecular-based relationships here obtained using a larger sampling of Satyrias.s. do not agree with some of the taxonomic rearrangements of a recent monographic study (Powell 2005). Although the proposed changes have not been formally published, the unpublished names and combinations have already appeared in taxonomic and record-based public databases, as well on annotations of herbarium specimens from several American and European herbaria. Hence the comments below.

Satyria warszewiczii is a species thought to be confined to Central America (southern Mexico to Panama), with a broad altitudinal gradient [(100–)300–2500 m] and consequently morphological variation (Luteyn and Wilbur 2005). Based on Principal Components Analyses of the morphological variation, Powell (2005) suggests that 11 species, mostly from South America, are indistinguishable from it and therefore should be synonymized.

Molecular data for Satyria latifolia A.C.Sm. and S. ventricosa, as well as two morphospecies of the variable S. grandifolia (Colombia–Peru) from NW Colombia, all putative synonyms of S. warszewiczii according to Powell (2005), were sampled. A third collection of S. grandifolia from SW Colombia, which was used in previous phylogenetic analyses (see Powell 2005, Powell and Kron 2003), was also analyzed.

The Central American Satyria meiantha, S. warszewiczii, and S. ventricosa form a clade (95% bootstrap; Fig. 1), in agreement with Powell’s taxonomic proposal, and they may as well be conspecific, although terminal branch lengths are very long. However, they are more closely related to S. allenii A.C.Sm., which was considered by Powell (2005) a distinct species, than to either S. latifolia or S. grandifolia.

On the other hand, the two newly sequenced specimens of Satyria grandifolia from NW Colombia form a well-supported clade sister to the Central American Satyrias. s., but with little support, while the S. grandifolia from SW Colombia is placed with species of extra-Andean distribution, although again with little support (Fig. 1). As for Satyria latifolia, this species is part of a poorly supported tetratomy and is not immediately related to either S. warszewiczii or any accession of S. grandifolia, as hypothesized. Moreover, there are no big differences between the samples of S. grandifolia from NW Colombia, whereas considerable changes have accumulated on the branches and terminals in their sister group, S. warszewiczii and relatives.

Altogether, the results suggest that the Satyria grandifolia from NW and SW Colombia are not conspecific. The herbarium vouchers of the S. grandifolia from NW Colombia were collected in the same biogeographic region (Chocó) where the type species was procured, about 200 km from the type locality. To rule out contamination of our sample, some molecular markers were independently re-sequenced and identical results were obtained. The herbarium voucher of the S. grandifolia from SW Colombia (Luteyn 15204) was collected much farther away from the type locality, but still within the Chocó biogeographic region. Unfortunately, because Luteyn 15204 does not have flowers (only fruits), it is not possible to reassess its taxonomic identity.

In the same general area of Luteyn 15204 there are specimens very similar to those from NW Colombia, however, others have floral and vegetative characteristics that subtly diverge from them and which have not been observed in other studied S. grandifolia collections. Without doubt Powell was right at pointing out that species delimitation within Satyrias.s. is complicated and that more fieldwork in western Colombia was advised.

It was also (Powell 2005) suggested that Satyria arborea A.C.Sm., endemic to Colombia, should be synonymized with S. allenii, endemic to Panama. However, in this analysis they are not sister species, with S. allenii placed at a more derived position within a well-supported clade that includes other Central American Satyrias. s. (Fig. 1). The newly sequenced S. arborea was collected relatively close to the type locality and it is easily differentiated from S. allenii by anther length (about twice as long in S. arborea). Anther length may be one of the best characters, beyond geographic distribution, to differentiate among these species, as other traditionally used features such as length of the petiole and pedicel and life form have been shown to be inadequate (Powell 2005).

Filaments and anthers of adjacent stamens are of different lengths in Cavendishia, a diagnostic character that has been invoked by most of the taxonomic classifications of the neotropical Vaccinieae of the last century (Luteyn 1983, Smith 1932, Sleumer 1941). However, the Cavendishia clade also includes Thibaudia jahnii (but < 50% bootstrap) toward its base (Fig. 1), a taxon that lacks the characteristic heteroandrous morphology of the genus Cavendishia and of the entire HtAP clade. Besides sharing free stamens, there are no other apparent morphological characters that satisfactorily explain the placement of T. jahnii within this clade

Cavendishia has approximately 130 species, most of which are native/endemic to Colombia. The 15 species of Cavendishia here analyzed include representatives from the two currently recognized subgenera of Cavendishia: Chalmydantha and Cavendishia, as well as of four of the five sections of subgenus Cavendishia (Foreroa is missing), all but one of the series of section Cavendishia (Uniflorae is missing), and all but series Lactiviscidae of section Engleriana (see Table 1). Unfortunately, none of the subgenera, sections, or series of Cavendishia is resolved as monophyletic. However, it must be cautioned that the phylogenetic relationships recovered within Cavendishia are too poorly supported to draw strong conclusions and a more extended taxonomic/molecular sampling of it is needed.

From a more general point of view, taxa present in Central America have a more derived position, but unlike Satyrias.s. in which Central American species are clustered together, in Cavendishia the Central American species are dispersed throughout the clade [C. lindauiana Hoerold, C. quereme (Kunth) Benth. & Hook. f., C. pubescens (Kunth) Hemsl., C. capitulata Donn.Sm., C. bracteata Ruiz & Pav. ex J.St.-Hil.].

The relationships recovered for this clade (Fig. 1) are consistent with those found by Pedraza-Peñalosa et al. (2013). Taxa in this clade all have equal stamens and their pedicels are continuous with the calyx, unlike the HtAP clade, the sister group. That being said, the sampled species of Demosthenesia A.C.Sm. have slightly unequal filaments, but not as markedly as in the HtAP clade.

The HmCP clade unites groups with diverse morphologies. The only monophyletic genera within it are relatively small (up to 12 spp.) and have contrasting geographic patterns: Anthopterus is widely distributed in the neotropics, while Pellegrinia and Demosthenesia are both endemic to a small area of the Peruvian and Bolivian Andes (Fig. 1). However, many subclades within HmCP are well supported and the two largest stand out. The four species of the Anthopterus revolutus–Thibaudia floribunda clade (77% bootstrap) have in common winged or angulate calyces and corollas, whereas the seven species of the Diogenesia alstoniana–Demosthenesia spectabilis clade (96% bootstrap) in contrast all have terete calyces and corollas.

Only one species of Themistoclesia Klotzsch was included in this analysis. However, Themistoclesia with articulate calyces have recently been described, but unfortunately, none of them was available for sequencing. These taxa also have other characteristics not previously thought to occur in the genus and it has been hypothesized they may represent a geographically and morphologically distinct clade (Pedraza-Peñalosa and Luteyn 2010). Thus, future analyses should sample in more detail the morphological diversity of Themistoclesia, as well as that of other members of the HmCP clade.

All species in the Thibaudiap.p. clade have pedicels articulated with the calyx; it is precisely the presence of such articulation that initially defined Thibaudia section Eurygania, currently a synonym of Thibaudia. However, the staminal characters that seem to be important in defining other larger clades are absent here. This analysis supports Powell and Kron’s (2003) assessment that species of Satyria in this clade, all endemic to S Peru and N Bolivia, should be segregated from Satyrias.s. (Fig. 1). The anther tubules of these Satyria do not diverge much distally, their sides being more parallel, and their tips lack ornamentations, so differing from Satyrias.s. (Powell 2005). The Thibaudia and Satyria species that cluster together all have equal or slightly unequal stamens and connate filaments.

On the other hand, although the T. macrocalyx–T. pachyantha subclade (74% bootstrap) also includes species with equal stamens, their filaments are free. Other characters that are alos shared by some of the members of the Thibaudiap.p. are thick corolla and calyx limb, anthers with poor distinction between tubules and thecae, thecae papillose and dehiscence by ventral clefts.

Psammisias.l. contains species with terete to winged calyces; fused, free or coherent staminal filaments; short to long corollas; and pinnate or plinerved laminas. It also includes perhaps the greatest variety of corolla shapes of any neotropical Vaccinieae (tubular, obconic, urceolate, turbinate and depressed, hemispheric); and also a great variety of corolla colors (yellow, magenta, vermilion, dark wine, red, white, green, etc.) and color combinations (solid, bicolor, multicolor).

The unifying staminal features of Psammisias.l. include stout anthers, free tubules, and connectives, the region where filaments adhere to the anthers, that are 2-spurred, alternately spurred (i.e. only one staminal cycle is spurred), or rarely unspurred. It is precisely because of the presence of spurs that Smith (1932) linked Psammisia to a group of Macleania whose connectives are faintly thickened distally. These thickened connectives were interpreted as a step toward the spurred condition observed in some Psammisias. l. A morphological connection between members of Psammisia and Macleania is suggested by these results, as Psammisias.l. is resolved in two clades, one of them sister to Macleania. However, the morphological basis of this relationship it is not yet known. As for the second grouping of Psammisias.l., it is part of a more basal tritomy. Unfortunately, because taxon sampling is still inadequate and the type species of the genus was not available for sequencing, it is uncertain which Psammisia clade will retain the name.

Most of the sampled Psammisias.l. are found within the Psammisia I clade (P. grandiflora–P. pedunculata, < 50% bootstrap), which is dominated by species from the N Andes (Fig. 1). Psammisia dolichopoda A.C.Sm. is the only species that is also present in Central America. However, ongoing studies suggest that it has considerable morphological variation and may represent more than one species (Pedraza-Peñalosa pers. obs.).

Two of the new species from LONP are placed here, within a clade (77% bootstrap) dominated by Colombian taxa; the only exception is the rare Psammisia oreogenes Sleumer, which was earlier thought to be exclusive to Ecuador but is now known to also occur in the Colombian portion of the Chocó biogeographic region (Fig. 1). The new species from LONP portray morphological characters that are either unusual in Psammisias.l. (e.g., P. sp. nov. 1 has large leaves with pinnate venation) or previously unknown (i.e., P. sp. nov. 2 is the only known taxon with pseudoverticillate leaves). In general, the novel taxa are morphologically different and molecular sequence data indicate they are not immediately related.

It is difficult to find unifying morphological characters for Psammisia I. Moreover, the clade lacks support and has a tritomy at its base. Within it, only the P. sp. nov. 1–P. sp. nov. 2 subclade has moderate support (77% bootstrap). All its species share chartaceous to subcoriaceous leaves, pinnate venation, racemes with short rachises typically less than 1.6 cm long making the inflorescences look fasciculate, and medium sized corollas 8–22 mm long.

Sequence data shows that Psammisia II taxa are more closely related to Macleania than to other congeneric species (Fig. 1). Unifying characteristics of the Psammisia II group comprise coriaceous leaves, plinerved venation, racemes with usually conspicuous rachises (1–26 cm long), and long corollas (17–40 mm). The inclusion of Thibaudia inflata within Psammisia II in a relatively derived position, deserves further scrutiny as its pedicel and calyx are continuous, whereas in most Macleania, and in all Psammisias.l. it is articulated.

Furthermore, other morphological features seem to also help to differentiate between the Psammisia I and II clades. The leaves of Psammisia II have laminar glands at the base of the abaxial side, while the leaves of Psammisia I do not have basal glands, or when laminar glands are present, they are then spread through the entire abaxial surface (e.g., P. sodiroi, see Pedraza-Peñalosa et al. 2013). Anther morphology also seems to be helpful, with Psammisia II possessing well-developed tubules of a diameter similar to that of the theca, while in Psammisia I tubules are smaller and narrower, of a basal diameter only a fraction of the theca width.

Macleania is sister to Psammisia II and there is good support for this clade and this relationship; the former clade includes M. floridunda Hook., the type species of the genus (Fig. 1). Because of its morphological range, Smith (1932) considered Macleania a coherent genus from which other genera were in the process of being derived. However, the subgenera Aponema and Macleania, which were thought to be morphologically well delimited, are not monophyletic. Moreover, although most sampled Ceratostema form a clade (but unsupported) placed within the basal tritomy of the Psammisia–Ceratostema–Macleania clade, C. reginaldii is resolved within Macleania.

As for Ceratostema, although as currently circumscribed is morphologically recognizable, it is not monophyletic and the sampled species fail to form a well supported clade. Ceratostema was one of the first genera to be erected, and because Ericaceae is a predominantly montane group with a high number of endemics, it is not difficult to imagine the challenges earlier taxonomists faced to procure sufficient specimens for their studies. Limited collections and field observations led to problematic generic classifications. Consequently, not only many species have been transferred out of Ceratostema, but also entire genera have been segregated from within it such as Demosthenesia and Pellegrinia. Both genera are here resolved as monophyletic and not closely related to any sampled Ceratostema. Unfortunately, both taxon (16 out of ca. 75 spp.) and molecular sampling are still insufficient to better discern the evolutionary relationships between Macleania and Ceratostema.