Almeidea A. St.-Hil. Belongs to Conchocarpus J.C. Mikan (Galipeinae, Rutaceae): Evidence from Morphological and Molecular Data, with a First Analysis of Subtribe Galipeinae

Subtribe Galipeinae (tribe Galipeeae, subfamily Rutoideae) is the most diverse group of Neotropical Rutaceae, with 28 genera and approximately 130 species. One of its genera is Almeidea, whose species are morphologically similar to those of the genus Conchocarpus. Species of Almeidea occur in the Atlantic Rain Forest of Eastern Brazil, with one species (Almeidea rubra) also present in Bolivia. The objective of this study was to perform a phylogenetic analysis of Almeidea, using a broader sampling of Galipeinae and other Neotropical Rutaceae, the first such study focused on this subtribe. To achieve this objective, morphological data and molecular data from the nuclear markers ITS-1 and ITS-2 and the plastid markers trnL-trnF and rps16 were obtained. Representatives of eight genera of Galipeinae and three genera of Pilocarpinae (included also in Galipeeae) and Hortia (closely related to Galipeeae) were used. Five species of Almeidea and seven of Conchocarpus were included, given the morphological proximity between these two genera. Individual (for each molecular marker) and combined phylogenetic analyses were made, using parsimony and Bayesian inference as optimization criteria. Results showed Galipeinae as monophyletic, with the species of Almeidea also monophyletic (supported by the presence of pantocolporate pollen) and nested in a clade with a group of species of Conchocarpus, a non-monophyletic group. Additionally, C. concinnus appeared in a group with Andreadoxa, Erythrochiton, and Neoraputia, other members of Galipeinae. As a result, Conchocarpus would be monophyletic only with the exclusion of a group of species related to C. concinnus and with the inclusion of all species of Almeidea with the group of species of Conchocarpus that includes its type species, C. macrophyllus. Thus, species of Almeidea are transferred to Conchocarpus, and the new combinations are made here.

Almeidea Belongs to Conchocarpus and a First Analysis of Galipeinae monophyly of Conchocarpus and its relationship with Almeidea. The results obtained will provide a framework for taxonomic decisions about the genera cited above, a base to support discussion of putative morphological synapomorphies of the groups, and a first step to a better understanding of the phylogeny of the Galipeinae and the Galipeeae.

Taxa studied
All taxa included in this study belong to subfamily Rutoideae (as circumscribed in [14]). The Galipeinae were represented by all five currently recognized species of Almeidea [20,21,23], seven of Conchocarpus, and one each of Erythrochiton, Galipea, Neoraputia, Rauia, and Ravenia (the list of all species names and authors are listed in S1 and S2 Text). Given the wide geographic range of Almeidea rubra, individuals from three populations (two from Espírito Santo and Minas Gerais states, both in Brazil, and one from Bolivia) were included. The Pilocarpinae were represented by two species of Esenbeckia and one each of Metrodorea and Pilocarpus. In addition, one species of Hortia (a genus assigned traditionally to the tribe Toddaliinae of the subfamily Toddalioideae, nowadays included in Rutoideae) was also included because it was clustered with Adiscanthus (not sampled here), a genus of the Galipeinae, in a previous study [14]. One species of Zanthoxylum (tribe Zanthoxyleae) was used as outgroup in all analyses. The newly generated sequences were deposited in GenBank and the respective accession numbers are cited in S1 Text. Voucher specimens from which molecular sequences and morphological data were obtained were deposited at SPF and SPFR herbaria (herbarium acronyms according to [24]) and are summarized in S1 and S2 Text, respectively.
The pollen characteristics of many taxa included in the analysis had been described previously (e.g., [18,20,38,39]). New data from pollen were obtained from Almeidea and six species of Conchocarpus. Pollen grains were taken from mature buds of different collections deposited at SPFR. Grains were acetolysed according to the methodology described in [40] and then mounted in glycerine jelly on glass slides. Palynological terminology followed [41] and [42].
Sequences were analyzed and edited using the Biological Sequence Aligment Editor (BioEdit), v.7.1.3 [50]. Each fragment was carefully examined to verify concordance among the sites. Limits of the trnL-trnF region, the rps16 intron, and ITS-1 and ITS-2 regions were determined by comparison with sequences deposited at GenBank. The automated alignments of the sequences were made with MUSCLE v.3.6 [51]. Refinements of the resulting alignment were made by eye using BioEdit software again, including trimming of initial and final blocks of the alignments in order to delete primer sequences. Indels were coded as missing data.
Parsimony analyses were made using PAUP Ã version 4.0b10 [43]. All characters were unordered and equally weighted (Fitch parsimony, [44]). Heuristic searches were performed with the tree-bisection-reconnection (TBR) branch-swapping algorithm with "steepest descent" and "multrees" options off, with 10,000 random-taxon addition replicates, and with 1 tree held in each replicate. Bootstrap analyses [45] were implemented to verify support for the clades using TNT 1.1 [52], with 1,000 pseudoreplicates, taxon random addition, and TBR branch-swapping algorithm. TNT software is computationally more efficient than PAUP, making a bootstrap analysis with a higher number of replicates in a shorter time. The support values for bootstrap percentage (BP) were categorized in the same way as those used in the morphological analysis: strong (!88%), moderate (76-87%), weak (63-75%), and ambiguous (<63%).
Searches were made in two independent runs, each with four Markov chains simultaneously, started from random trees. Ten million generations were performed, sampling a tree at every 1,000 generations and the temperature coefficient of the chain-heating scheme set to 0.20. After the exclusion of 25% of the trees corresponding to the burn-in, a majority-rule consensus tree was calculated from remaining trees. The stationary and convergence of runs, as the effective sample size (ESS) was monitored using the program Tracer v.1.6 [55]. Robustness of clades was estimated as Bayesian posterior probabilities in the same software. It was considered a well-supported clade only when PP value was higher than 0.95.

Morphological plus molecular dataset analysis
The Incongruence Length Difference Test (ILD) [56], implemented in PAUP Ã with 1,000 replicates and the same parameters used for parsimony searches, was used to test first, the congruence between the two nuclear marker datasets and between the two plastid marker datasets, and then the congruence between the morphological and all molecular datasets. Analysis of the complete morphological and molecular dataset was performed using parsimony with the same parameters described for individual analyses.

Mapping of morphological characters
The history of some morphological characters (7,11,13,28,29,33) was traced in the majorityrule consensus tree from Bayesian analysis of the combined molecular data using Mesquite v.2.75 [57] with the ancestral reconstruction of states inferred by parsimony criterion.

Morphological Analysis
Parsimony analysis of the 35 morphological characters produced 270 equally parsimonious trees, with 128 steps (CI = 0.43, RI = 0.64, see Table 1 for a summary of all parsimony results). The strict and majority-rule (Fig 2) consensus trees resulted in similar topologies, but the clades are unsupported or weakly supported. Subtribe Galipeinae emerged as monophyletic (BP = 55%), and possible synapomorphies of the subtribe are indicated in Fig 2. Almeidea also appeared as monophyletic with low support (BP = 69%), with the pantocolporate (with numerous apertures) pollen grain as a synapomorphy for the genus (Fig 2). Other character state taxonomically important in Almeidea is the presence of glandular structure at apex of calyx lobes (present also in Conchocarpus heterophyllus, C. pentandrus and C. minutiflorus), indicating a possible case of homoplasy in the group (see Fig 2). However this character is discussed later with the results from molecular data, because the relationships of the morphological analysis are unsupported. Conchocarpus appeared as non-monophyletic, with one group of species closer than the other to Almeidea, to which C. heterophyllus was sister. Representatives of tribe Pilocarpinae (Esenbeckia, Metrodorea, and Pilocarpus) formed an unsupported clade (BP = 53%).

Sequence characteristics
Sequences of the trnL-trnF region and rps16 intron were much longer (1083 and 837 base pairs, respectively, see Table 1) than those of the nuclear regions ITS-1 and ITS-2 (456 and 352 base pairs, respectively). However, ITS-1 and ITS-2 sequences provide more polymorphic characters, with 23.2% and 30.4% of Parsimony Informative Characters (PICs), whereas trnL-trnF region and rps16 intron yield 6% and 8.8% of PICs, respectively (Table 1).
Plastid markers (trnL-trnF and rps16). Given the congruence given by the result of the ILD test, the plastid markers were combined in a single matrix, as in previous studies (e.g., [13,14]). In the majority-rule consensus trees resulting from parsimony analysis and the Bayesian analysis (Fig 3), the subtribe Galipeinae and the species of Almeidea appear as monophyletic groups with strong support (BP = 100%, PP = 1). The relationships among species of Almeidea are not well resolved, with exception of the pair, A. limae and A. lilacina, that is strongly supported in the Bayesian analysis (BP = 64%, PP = 1). Conchocarpus appears as non-monophyletic in both analyses (parsimony or Bayesian); six of the seven species of Conchocarpus in the analyses appear in a clade with Almeidea (BP = 99%, PP = 1), while the seventh (C. concinnus) appears with Andreadoxa, Erythrochiton, and Neoraputia in a strongly supported clade (PP = 1). Galipea and Rauia appear as a group (BP = 79%, PP = 1), sister to Ravenia (BP = 64%, PP = 0.97). Congruent with the morphological analysis, Pilocarpinae appear as a strongly supported clade (BP = 100%, PP = 1), with Pilocarpus sister to Esenbeckia and Metrodorea. Hortia appears as sister to the Galipeinae only in the Bayesian analysis, however this clade is unsupported (PP = 0.7).
Nuclear markers (ITS-1 and ITS-2). Results of the analyses of the nuclear markers sequences are highly congruent with those obtained from the analyses of plastid markers sequences, with higher support values in most of the clades (Fig 4). The monophyly of Galipeinae and of Almeidea is unsupported (PP = 0.78 and PP = 0.90, respectively). On other hand, Almeidea and Conchocarpus form a clade with strong support (PP = 1), but Conchocarpus is non-monophyletic in relation to Almeidea. In the clade comprising the species of Almeidea, A. limae is more related to specimens of A. rubra (individuals from MG and ES, states of Brazil) than with A. lilacina, differently of the plastid markers analysis. However this relationship was supported only in the BA analysis (PP = 1), without support in the MP analysis. Congruent with the results of plastid sequences analysis, Hortia oreadica appears as sister to Galipeinae, and Pilocarpinae appear as monophyletic. However, Ravenia spectabilis appears as sister to all other Galipeinae, in an unsupported clade (PP = 0.78, without BP value), whereas in the analysis of plastid markers sequences, it appears as sister to (Galipea, Rauia), with strong  Combined molecular analysis. The ILD test did not find incongruences in the datasets (P > 0.05), making combining them reasonable. Besides some incongruences, as commented before, the visual inspection also revealed major congruence among the topologies resulting from the analyses of the plastid (Fig 3) and nuclear markers (Fig 4). Considering this, we decided to combine the four molecular datasets. As in the analyses of plastid marker sequences and nuclear marker sequences, Galipeinae (PP = 1) and Almeidea (PP = 1) appear as monophyletic (Fig 5). A clade comprising the species of Almeidea and six of the seven species of Conchocarpus (C. concinnus was not sampled for ITS) has strong support (BP = 99%, PP = 1). Ravenia appears in an unresolved position inside Galipeinae, Hortia oreadica appears as sister to Galipeinae in an unsupported clade (PP = 0.9), and the subtribe Pilocarpinae is monophyletic (BP = 100%, PP = 1).  Combined molecular and morphological datasets analysis. The ILD test revealed that morphological and molecular datasets differed significantly (P < 0.05), indicating that these datasets should not be combined. Given, however, that some authors argue that combined datasets (whether molecular or morphological) provide the best explanatory power in analyses that involve parsimony (see [58,59]), we decided to combine those datasets, perform a parsimony analysis, and explore the results with special caution.
Topology of the strict consensus tree from the parsimony analysis (Fig 6) is largely congruent with that obtained from the combined molecular analysis (Fig 5), including the monophyly of Galipeinae (BP = 100%), Pilocarpinae (BP = 100%), and Almeidea (BP = 100%); also shows the close relationship of Almeidea with Conchocarpus (the latter a non-monophyletic genus) in a strongly supported clade (BP = 100%) and the position of C. concinnus in an unsupported clade with other genera of Galipeinae (BP = 59%). The pairing of Galipea and Rauia is strongly supported (BP = 98%) and they emerged as sister-group (BP = 73%) of the clade formed by Andreadoxa, Erythrochiton, and Neoraputia and C. concinnus. The position of Ravenia follows the results of the nuclear analysis, as a sister-group of the other clades in Galipeinae (BP = 100%). As in majority of other analyses, Hortia emerged as sister-group of Galipeinae, however without support.

Utility of the molecular markers used
In the parsimony analyses, both plastid markers were more conservative than the nuclear markers and had a lower percentage of informative characters (maximum of 13% vs. 31.7%, respectively, see Table 1). On the other hand, the higher substitution rates in the nuclear markers than in the plastid markers made alignment of the sequences more subjective, not allowing the manual editing in some cases. Despite this subjectivity, found mainly for ITS-2 in this work, the trees from the analysis of the nuclear markers are better resolved at the tips of the branches and, thus, better at resolving relationships among species when compared to the trees from the plastid markers. The combined analysis of the four molecular regions includes a larger dataset and shows resolution at all levels of the trees. The strict consensus tree resulted from parsimony analysis of the combined molecular and morphological data (Fig 6) resulted in a topology highly congruent to that from the combined analysis of the four molecular regions (Fig 5). Considering the amount of data included and the resolution of the branches, we chose reconstructing the history of some important morphological characters in the majority-rule consensus tree using Bayesian inference from the combined molecular data (Fig 5) and discuss the results with basis on this tree. However, we are aware that Bayesian methods sometimes inflate support values and sometimes produce spurious clades (see [60,61]).

Subtribe Galipeinae
All analyses showed the monophyly of Galipeinae, and some morphological characters support this clade here: the presence of coherent petals (Fig 7A), an adherence between staminal filaments and petals (Fig 7B), a basifixed attachment of anther to filament, a curved embryo ( Fig  8A), conduplicate cotyledons (Fig 8B), and plicate cotyledons. Some of these characteristics was pointed previously as possible synapomorphies for the subtribe [13,16], but were not tested before in phylogenetic analyses with focus in Galipeinae. These states are variable within Galipeinae, with some reversions occurring in some clades or taxa (e.g. free petals in Almeidea and Andreadoxa, Fig 7A), and also with some states occurring outside of the subtribe, as the curved embryo, that is also present in Decatropis, Megastigma, Choisya and other genera (see [5]). The variation of character states in Galipeinae also difficults the identification of Almeidea Belongs to Conchocarpus and a First Analysis of Galipeinae morphological synapomorphies for internal clades. As an example, curved embryo with conduplicate and plicate cotyledons can be absent in some groups, as in some Conchocarpus (for example C. concinnus) with straight embryo and not conduplicate and not plicate cotyledons ( Fig  8A and 8B). In this study, however, it is clear that appendages at the base of the anthers is synapomorphic for the clade formed by the species of Andreadoxa, Erythrochiton, Galipea, Neoraputia, Rauia, and Conchocarpus concinnus, but appendages are lacking in the two latter (perhaps through a reversion). Additionally, the clade formed by Galipea and Rauia is characterized by fruits that are almost fully syncarpic (also present in Neoraputia) rather than apocarpic as in all other Galipeinae.
Galipeinae exhibit a great variation in the degree to which petals and staminal filaments are united. Petals can be free without adherence to filaments (as in Almeidea and Spiranthera, not sampled here), coherent to each other and adherent to filaments (as in some Conchocarpus), or congenitally fused to each other and to the filaments (as in Erythrochiton). Fig 7A and 7B show the distribution of these characteristics in the taxa analyzed in the present study. A previous study [62] of six species of three genera of Galipeinae revealed three different patterns of fusion within and between petals and filaments: a floral tube formed by postgenital coherence and adherence of petals and filaments by interwining trichomes (as in Conchocarpus heterophyllus and C. minutiflorus), a tube formed by congenital fusion of petals and filaments (Erythrochiton brasiliensis), and a tube formed distally with the first pattern and proximally with the second pattern (Galipea jasminiflora and C. macrophyllus). That C. macrophyllus, Erythrochiton and Galipea appear in different clades (Fig 7A and 7B) suggests that their floral tubes, although formed by congenital fusion in all cases, are from different origins. In Ravenia spectabilis (not analyzed by [62]), that appears in an unresolved position with the other genera of Galipeinae, the pattern of development and fusion of petals and filaments may be of a different origin than that encountered in C. macrophyllus, Erythrochiton or Galipea or homologous to one of these, depending of its position in the subtribe, that is unresolved so far. Because this study included only nine of the 28 genera of the Galipeinae, analyses of a broader sample of the taxa are required to understand floral evolution and synapomorphic characteristics within the Galipeinae as a whole.

Almeidea and Conchocarpus
The phylogenetic proximity between Almeidea and Conchocarpus was supported in this work by both morphological and molecular data. As said in the introduction, Almeidea shares many characters with several species of Conchocarpus, differing from them only by its free, rather than coherent or connate, petals. Although petal color was treated before as a distinguishing character, it is not an absolute; petals in Almeidea are usually pink or lilac but rarely white, and those of Conchocarpus are usually white but rarely pink or yellow.
Due to their morphological similarity, species of Almeidea have been traditionally recognized as a natural genus with no doubts about its circumscription (see [9,11,21]). Results of this molecular study have corroborated this view by showing that the group is monophyletic. Its monophyly is further supported by the presence of pantocolporate pollen grains (Fig 9B), a unique feature among the Neotropical Rutaceae. Another feature common to all species of Almeidea is the presence of a glandular structure, interpreted as a nectary [21], in the apex of each calyx lobes ( Fig 9A); this feature was observed also in Conchocarpus heterophyllus, C. minutiflorus, and C. pentandrus. This character possibly represents a synapomorphy of the clade comprising Almeidea and its sister species of Conchocarpus as showed in Fig 9A. Free petals and filaments that are not adherent to the petals are present not only in the species of Almeidea, but also in many genera of Rutaceae outside of the Galipeinae, and are here considered to be reversions in Almeidea (Fig 7A and 7B). However it is important to mention that a slightly adherence between petals and filaments was observed in some individuals of species of Almeidea (with exception of A. albiflora), what can only be adequately tested with anatomical studies.
The relationship between the species of Almeidea is unresolved in morphological analysis and some incongruences occurred between the molecular datasets analyses. However it become clear the phylogenetic proximity between A. limae, A. lilacina and A. rubra, this last one non- monophyletic because of the specimen from Bolivia, that emerged separated from Brazilian specimens. These results are in agreement with the morphological similarities between these three species, observed especially in vegetative and fruiting materials. These similarities will be detailed in another work (Bruniera et al., unpublished), with the formal recognition of a single species for A. limae, A. lilacina and A. rubra.
Contrasted with the nomenclatural stability in Almeidea, the species now recognized as Conchocarpus (see [16]) have a history of taxonomic and nomenclatural changes, as can be seen in [16] (p. 257-258), who considered Conchocarpus polymorphic. The position of C. concinnus in a clade separate from that of the other species of the genus corroborates the view of Conchocarpus as a non-monophyletic group. C. concinnus shares with four other species of Conchocarpus (not sampled here), which are also endemic to the Brazilian Atlantic Rain Forest, the combination of anthers broadly attached to the filaments, almost free sepals still overlapping at anthesis, a glabrous ovary, and plano-convex (not plicate) cotyledons [16]. These may best be recognized as a separate genus.

Nomenclatural implications
Given that the species of Almeidea are nested in a clade of six species of Conchocarpus including the type species, C. macrophyllus, we propose to transfer Almeidea to Conchocarpus. The new nomenclatural combinations are presented at the end of the text. The relationships of C. concinnus and its relatives are being investigated in a major study of Galipeinae by the authors (unpublished), and a revision of the former species of Almeidea (now part of Conchocarpus) is also underway.

Taxonomic changes
New combinations in Conchocarpus. Accepted names are in bold-faced type, synonyms in italics.