Introduction

Rutaceae comprises 154 genera and approximately 2100 species distributed worldwide, but mainly in tropical and subtropical regions (Kubitzki et al. 2011). Members of the family are characterized by the presence of secretory cavities or oil glands within the tissues of the shoot system that appear as pellucid dots on the leaves (Engler 1931; Metcalfe and Chalk 1950; Kubitzki et al. 2011). Phylogenetic and morphoanatomical studies within the family are of particular importance in order to better target the exploitation of the group (Groppo et al. 2008).

The traditional classification of Rutaceae was proposed by Engler (1931), who recognized six subfamilies according to the degree of connation of carpels, the structure of fruits, and anatomic features of oil glands. Engler’s taxonomic treatment is the most extensive ever performed on Rutaceae and includes detailed information about the morphology and geographical distribution of all groups. However, the classification has been questioned on the basis of evidence obtained from studies on chemosystematics (Silva et al. 1988) and, more recently, molecular phylogenetics (Chase et al. 1999; Poon et al. 2007; Groppo et al. 2008, 2012; Bayly et al. 2013; Morton and Telmer 2014). It is important to accurately investigate morphological features of members of Rutaceae, looking for further evidence toward a better understanding of the relationships within the family. Thus, it is necessary to find morphological synapomorphies supporting the clades recently recognized using molecular data.

Metrodorea A.St.-Hil. is a Neotropical genus within Rutaceae that contains six described species that are distributed mainly in the Brazilian rainforests and seasonal forests, but also in Suriname and Bolivia (Kaastra 1982; Kubitzki et al. 2011; Dias et al. 2013, 2015). The phyllotaxy is decussate with peculiar intrapetiolar stipules, which are united by intertwined hairs, and so cover the shoot apex, thus protecting developing leaf primordia (Cruz et al. 2015). Most species of the genus have leaves comprising one to three leaflets on the same plant. Cruz et al. (2015) observed that heterophylly results that even though all leaves start out as trifoliate, one or two leaflets may abort during development, probably associated with the reduction in space caused by the tight structure covering the shoot apex. Such limitation of space may have also have led to the selection of individuals that produce a single terminal leaflet rather than more leaflets, generating a strictly unifoliolate architecture, as observed in M. maracasana.

Metrodorea has been classified with Esenbeckia Kunth, Raulinoa R.S.Cowan, and Pilocarpus Vahl, which together made up the subtribe Pilocarpinae of the tribe Galipeeae in subfamily Rutoideae (Engler 1931; Kaastra 1982; Pirani 1999). This subtribe is characterized by globular floral buds, and yellowish or violet petals that are nearly regular and free (Kaastra 1982). However, phylogenetic analyses based on molecular data (Groppo et al. 2008, 2012) indicate that the clade formed by Balfourodendron Méllo ex Oliv. and Helietta Tul., traditionally included in the subfamily Toddalioideae (Engler 1931; Pirani 1998) is strongly supported as the sister group of a clade containing Metrodorea and Esenbeckia. These two clades form a polytomy with Pilocarpus and other genera, while Raulinoa was not present in those analyses.

A recent molecular phylogeny of Metrodorea based in the ITS nuclear ribosomal spacers and trnS-G intergenic spacer (Dias et al. 2015) revealed that the sister group of Metrodorea could be Raulinoa and/or the clade Helietta + Balfourodendron, depending on the use of parsimony or bayesian criteria. Strong floral similarities between Helietta and Esenbeckia, which belong to different subfamilies according to Engler’s classification, have been highlighted by Kaastra (1982) and Pirani (1998). Kaastra (1982) suggested that it might be appropriate to remove all genera but Pilocarpus from the Pilocarpinae because this genus is distinctly different from the other in its morphological and chemical aspects. Kubitzki et al. (2011) proposed a number of informal alliances, based on the phylogeny of Groppo et al. (2008), which include the “Esenbeckia alliance” (Metrodorea, Raulinoa, and Esenbeckia) and the “Balfourodendron alliance” (Balfourodendron and Helietta, already transferred to the subfamily Rutoideae). Furthermore, these authors stated that the current phylogenetic evidence makes it very difficult to establish associations between Pilocarpus and the other genera.

Previous studies on the leaf anatomy of members of Pilocarpus and Raulinoa have identified a wide array of structures potentially useful for the understanding of the relationships between the members of Rutoideae especially those belonging to Galipeeae (Arioli et al. 2008; Muntoreanu et al. 2011; Cruz et al. 2015). However, additional information concerning the stipules of Metrodorea together with other novel anatomical features may constitute key characters for elucidating the species relationship within this genus and might also assist in understanding its relationships with other genera.

Within this context, the aims of the present study are: (1) to provide a detailed examination of leaves and reunite available reproductive structures data of Metrodorea and closely related genera; (2) to perform a combined analysis of morphological and available molecular sequences from databases in order to analyze whether morphological data can add resolution to previous phylogenies, contributing to elucidate the phylogenetic relationships within this group; and (3) to get insights on the structural evolution based on molecular trees and possible interpretations.

Materials and methods

Taxon sampling

Twenty species belonging in Rutoideae were analyzed (Table 1), including all the six described species of Metrodorea, five species of Esenbeckia [5/30 spp. of the genus, according to Kubitzki et al. (2011)], two species of Pilocarpus (2/17 spp.), two of Helietta (2/8 spp.), Raulinoa echinata (monospecific genus), Balfourodendron riedelianum (1/2 spp.), Conchocarpus heterophyllus (1/45 spp.), and Galipea trifoliata (1/15 spp.). All these species are representative from the “American Clade” of Rutoideae (Groppo et al. 2008, 2012). As an outgroup, we used Zanthoxylum rhoifolium, a representative from “Old World and Oceania Clade” (Groppo et al. 2008, 2012) that share many chemical similarities with some of the Australian genera (Kubitzki et al. 2011). Voucher specimens (Table 1) are deposited in the Universidade de São Paulo herbarium (SPF).

Table 1 Classification of analyzed species and their voucher numbers
Full size table

Morphological and anatomical studies

Leaf characters were directly examined by us. Five leaves between the fifth and tenth nodes of different branches were collected from plants of each species, fixed in formalin-acetic acid–ethanol 50% (FAA) for 48 h and stored in 70% ethanol. Samples selected based on their structural integrity were submitted to conventional ethanol/tert-butanol gradient (50–100%) dehydration and embedded in paraffin (Johansen 1940). Longitudinal and transverse sections were obtained with a rotary microtome, stained with Safranin and Astra Blue dyes according to the method of Bukatsch (1972), and mounted permanently on slides with synthetic resin. Leaves epidermis were dissociated for surface viewing following Franklin (1945). Scanning electron microscopy (SEM) was used to observe trichomes on the leaf surface and in the inner region of protective intrapetiolar stipules. For SEM analysis, shoot apices and leaves fragments fixed in FAA were dehydrated through a gradient series of ethanol, critical point dried, mounted on metal stubs, and sputter-coated with gold as described by Silveira (1989), and then analyzed on a Zeiss DSM 940 microscope. Leaves were cleared and stained according to Strittmater (1973), and the venation pattern was classified according to the guidelines established by Ellis et al. (2009).

All reproductive structures characters used in the analyses were compiled from Kaastra (1982, for former Pilocarpinae); Pirani (1998, for Helietta and Balfourodendron), Pirani (1999, for all species, except Metrodorea concinna); Kubitzki et al. (2011, for all genera), El Ottra et al. (2013, for Conchocarpus heterophyllus), and Dias et al. (2015, for Metrodorea).

Forty-five morphological characters (see Table 2) showed some variation and were used for the construction of a matrix of taxa and morphological characters (Table 3), following the guidelines of Sereno (2007).

Table 2 Characters and their states used in the compilation of the matrix
Full size table
Table 3 Matrix of taxa and characters
Full size table

Retrieving of molecular sequences and alignment

We used the nuclear ribosomal spacers ITS1 and ITS2, and the gene 5.8S (here referred as ITS/5.8S region) and the cpDNA trnS-G intergenic spacer obtained by Dias et al. (2015) for their Metrodorea phylogeny (accession numbers for sequences in GenBank available in their article). We aligned the sequences using the same method used by them (ClustalX software, Thompson et al. 1997) and trimmed the alignment based on the shortest sequence as they did.

Phylogenetic trees and reconstruction of ancestral character states

The matrix of taxa and morphological characters (Table 3) was added to the molecular alignment. For species with just morphological data available, we considered each molecular state of character as missing (“?”). Models and partitions were detected in PartitionFinder 2.1.0 (Lanfear et al. 2016) for molecular data and used for Bayesian analysis (BA) together with morphological data. We treated ITS/5.8S, trnS-G, and morphological data as three different partitions, with the models GTR + Gamma, HKY + Gamma, and Standard Discrete Model + Gamma, respectively. The analysis was conducted with MrBayes 3.2 (Ronquist et al. 2012), with 100 million generations and 0.25 burn-in fraction, and aims to detect whether adding morphological data with new taxa will improve the resolution of the BA tree of Dias et al. (2015) that could not solve the sister group of Metrodorea.

For the reconstruction of ancestral character states in Metrodorea, we used the molecular data (alignments of ITS/5.8S and trnS-G) for the 13 species studied by Dias et al. (2015) and the same model and partitions used by them (GTR-Gamma-I, two partitions). We performed a maximum likelihood (ML) analysis using the software RAxML 8 (Stamatakis 2014) to obtain an ML tree. The morphological data concerning these species were optimized in this ML tree using the Mesquite software (Maddison and Maddison 2015), with the maximum likelihood criterion and the Markov k-state one-parameter model. Character states that appeared in nodes with proportional likelihood ≥0.65 were interpreted as likely (as in Castroviejo-Fisher et al. 2014) and considered a potential synapomorphy.

Results

The matrix (Table 3) of morphological characters and states (Table 2) is descriptive and provides a summary of the examined characters. Below, we give more details for some of the characters that may be of difficult interpretation (e.g., constrictions and swellings) or that deserve better details, with supporting images.

Leaf morphology (characters 1–9)

Leaves of the studied species are simple in Esenbeckia grandiflora, E. leiocarpa, E. pilocarpoides, and Metrodorea maracasana; pinnate in Pilocarpus spp. and Zanthoxylum rhoifolium; and trifoliolate or seldom bifoliolate (some Metrodorea samples) in the remaining species. There are constrictions between the base and the blade in Metrodorea spp., Esenbeckia grandiflora, E. pilocarpoides, E. pumila, Pilocarpus spp., Balfourodendron riedelianum, Helietta spp.¸ and Zanthoxylum rhoifolium. Swellings (tumescences) are present on the leaf bases of E. grandiflora and E. leiocarpa. The presence of such constrictions and swellings at the base of species with a non-divided blade usually imply the reference to them as “unifoliolate leaves,” but they are identical to simple leaves with respect to the development and blade morphology. The only remarkable exception is the mostly trifoliolate species of Metrodorea (all but M. maracasana, which is consistently simple-leaved), whose leaf primordia bear three leaflet primordia each, but some of them may be aborted during development resulting in mature leaves bearing only one or two leaflets.

Anatomy of the leaf base (characters 10–18)

The dermal system of the petiole shows in some regions a periderm with a subepidermal phellogen (Fig. 1a, b) and, sometimes, lenticels (Fig. 1b). The exceptions are Metrodorea mollis and Esenbeckia leiocarpa, which bear papillose epidermis with trichomes (Fig. 1c, d) that is not replaced by a periderm. In Metrodorea and Raulinoa, the proximal and adaxial part of the petiole exhibits a region with long intertwined multicellular trichomes. However, only Metrodorea species present capitate glandular trichomes (Fig. 1e) in the proximal region of the petiole.

Fig. 1
figure 1

Leaf base. a Metrodorea nigra. Distal cross section of petiole. Closed vascular cylinder surrounded by perivascular fibers and periderm. b Esenbeckia grandiflora. Cross section of petiole. Lenticel and perivascular fibers (arrows). c M. mollis. Proximal cross section of petiole. Vascular bundles (arrows) are associated with intrapetiolar stipule. d M. mollis. Proximal cross section of stipules. Interlocking cells in margins. e M. maracasana. SEM. Inner region of intrapetiolar stipule with glabrous margin protecting a pair of leaf primordia. Close to the margin, long intertwined non-glandular trichomes (black arrow) and below them, two of the glandular trichomes pointed by white arrows. Bars a, c 400 µm; b, d 100 µm; e 200 µm; le lenticel; is intrapetiolar stipule

Full size image

Along the axis of the petiole, the vascular system consists of three bundles at the proximal region (Fig. 1c) and a closed cylinder at the distal region (Fig. 1a), except in Metrodorea stipularis, that does not present a region with a closed cylinder in its leaf base, neither a free petiole (its leaflets are sessile, the blades directly attached to the stipule). Most species present perivascular fibers forming an interrupted cylinder (Fig. 1a, b), except Metrodorea mollis, Esenbeckia leiocarpa, Galipea trifoliata, Helietta apiculata, and Balfourodendron riedelianum, which lack such fibers. The intrapetiolar stipules in Metrodorea species are vascularized, except in M. nigra. Adhesion between the margins of the stipules is provided by interlocking epidermal cells in M. mollis (Fig. 1d), M. flavida, and M. concinna, while in other species the stipule margins are glabrous (Fig. 1e).

Dermal system of the leaf blade (characters 19–24)

Unicellular non-glandular trichomes occur on both adaxial and abaxial surface of most species, but do not occur in Conchocarpus heterophyllus, Esenbeckia pilocarpoides, and most of Metrodorea spp. In Metrodorea, this type of trichomes is present only in M. mollis (Fig. 2a). Esenbeckia irwiniana has these trichomes distributed more densely on the abaxial surface of the leaves (Fig. 2b), while other species does not exhibit a visible difference of distribution between the two surfaces. Multicellular non-glandular trichomes on the abaxial surface occur only in Conchocarpus heterophyllus (Fig. 2c). Glandular trichomes are present in all studied species with the exception of Esenbeckia grandiflora and Zanthoxylum rhoifolium. Capitate trichomes with a multicellular peduncle occur in Conchocarpus heterophyllus, E. irwiniana, and E. pilocarpoides (Fig. 2d–f), whereas peltate trichomes are present in Metrodorea spp. (Fig. 2g), Raulinoa echinata, and Helietta spp.

Fig. 2
figure 2

Leaf blade. a Metrodorea mollis. Cross section. Uniseriate epidermis, unicellular non-glandular trichomes (arrows), and peltate glandular trichome. Secretory cavity in the midrib region. b Esenbeckia irwiniana. SEM of cross section. Higher density of trichomes on abaxial side when compared with adaxial side. c Conchocarpus heterophyllus. Multicellular non-glandular trichome. d C. heterophyllus. Capitate glandular trichome with multicellular peduncle and unicellular head. e–f Capitate glandular trichomes with multicellular peduncle and head; e E. pilocarpoides; f E. irwiniana; g M. mollis. Cross section. Peltate trichome associated with secretory cavity in the midrib region. h–l Epidermal cells in front view; h E. grandiflora, abaxial surface with anomocytic stomata, unicellular non-glandular trichome (arrow), and straight anticlinal cell walls; i M. stipularis, abaxial surface with peltate trichome (arrow) and sinuous anticlinal cell walls; j M. mollis, abaxial surface with sinuous anticlinal cell walls and trichomes (arrow); k M. maracasana, adaxial surface with sinuous anticlinal cell walls and peltate trichome (arrowed); l M. stipularis, adaxial surface with sinuous anticlinal cell walls. Bars a, b, h 100 µm; c, f, g, i–l 50 µm; d, e 20 µm; pt peltate trichomes; sc secretory cavities

Full size image

In most of the studied species, anticlinal walls of epidermal cells are straight in frontal view (Fig. 2h), while in Metrodorea spp. they are sinuous (Fig. 2i–l), as well as in Esenbeckia pilocarpoides and Galipea trifoliata. Anomocytic stomata (Fig. 2h–j) occur only on the abaxial surface, except in E. irwiniana. This species has this type of stomata on both surfaces, but they are rare on the adaxial side.

Mesophyll and midrib region (characters 25–31)

All species have dorsiventral mesophyll with uniseriate palisade parenchyma (Fig. 3a–c), except Esenbeckia pumila and Helietta spp., which present multiseriate palisade parenchyma. Vascular bundles are collateral and involved by a sheath (Fig. 3c, d). Extensions of the bundle sheath occur in E. pumila and E. irwiniana oriented toward the adaxial and abaxial surface (Fig. 3e).

Fig. 3
figure 3

Cross sections of the leaf blade. a Conchocarpus heterophyllus. Druse (arrow), uniseriate palisade parenchyma, and spongy parenchyma containing small intercellular spaces. b Metrodorea mollis. Uniseriate palisade parenchyma and spongy parenchyma with loosely arranged arm cells. c M. nigra. Bundle sheath (arrow). d Esenbeckia grandiflora. Detail of bundle sheath (arrow). e E. irwiniana. Extensions (arrows) of bundle sheath oriented toward adaxial and abaxial surfaces. f M. maracasana. Vascular cylinder in the midrib region formed by two arches (arrows). Secretory cavities can be observed close to the abaxial surface. g E. irwiniana. Midrib region with medullary arch (arrow); h M. maracasana. Detail of midrib with secretory cells (arrows). i M. nigra. Crystals of palisade parenchyma in compartments separated by cell wall (arrow). Secretory cavity surrounded by epithelial cells under polarized light. j E. pilocarpoides. Raphides bundle under polarized light. k M. stipularis. Palisade parenchyma with prismatic crystals in compartments separated by walls (arrows) transversal to the length of the cell. l M. nigra. Prismatic crystals under polarized light, some of them concentrated in the hypodermis of the midrib region. Bars a–d, h 50 µm; e 100 µm; f, g, l 200 µm; j–k 20 µm; pp palisade parenchyma; sp spongy parenchyma; sc secretory cavities

Full size image

In the midrib region, the vascular system consists of two vascular arches, one in the adaxial side and the other on the abaxial side, forming a vascular cylinder (Fig. 3f). In Esenbeckia irwiniana, a central vascular arch occurs in the medullary parenchyma of this cylinder (Fig. 3g). Secretory cells (Fig. 3h) and secretory cavities with epithelial cells (Fig. 3i) are present in the mesophyll and in the region of the midrib.

Different types of crystals occur in the midrib region, mesophyll and vascular tissues. They can be raphides (Fig. 3j), druses (Fig. 3a), or prismatic crystals that may appear in a cubic or in a rod-like form (Fig. 3k). Prismatic crystals (when they occur) and Pilocarpus druses are found in compartments in the palisade parenchyma created by walls transversal to the length of the cell. However, in Esenbeckia pilocarpoides and Metrodorea mollis, such crystals are less frequent in all tissues and were not observed in the palisade parenchyma. In M. concinna, M. nigra, M. maracasana, and M. flavida, some crystals are concentrated in the hypodermis of the midrib region (Fig. 3l).

Venation (characters 32–36)

The venation pattern of the studied species is pinnate and simple brochidodromous, as in Metrodorea mollis (Fig. 4a). Connections of the secondary veins with the midrib are excurrent in Balfourodendron riedelianum, Esenbeckia grandiflora (Fig. 4b), E. pumila, Galipea trifoliata, and Helietta puberula; decurrent only in the proximal secondary veins in Raulinoa echinata, Metrodorea mollis (Fig. 4a), M. nigra, M. concinna, and M. stipularis; and decurrent in the other studied species, as in Conchocarpus heterophyllus (Fig. 4c). Intersecondary veins are absent only in E. leiocarpa. Tertiary veins are all mixed percurrent (Fig. 4d), whereas quaternary veins are irregular reticulate in E. irwiniana, E. pilocarpoides (Fig. 4e), E. pumila, Pilocarpus jaborandi, and Zanthoxylum rhoifolium and freely ramifying in the other species, as in Metrodorea maracasana (Fig. 4f). The veins of all studied species have branched ends (Fig. 4e–g) and exhibit areolation (Fig. 4g–i). The areolas show poor development (i.e., irregular sizes and shapes with more than seven sides) in B. riedelianum, Helietta spp. (Fig. 4g), R. echinata, Pilocarpus spp., and Zanthoxylum rhoifolium and good development (i.e., regular sizes and shapes with three to six sides) in the other species (Fig. 4h, i). The marginal ultimate venation is incomplete in B. riedelianum, E. grandiflora (Fig. 4j), E. leiocarpa, E. pumila, Helietta spp., and Pilocarpus spp., and it is looped in all other species (Fig. 4k).

Fig. 4
figure 4

Leaf venation. a Metrodorea mollis. Primary vein forms a pinnate pattern together with secondary veins. Connections are decurrent at the base of the leaf and excurrent at the apex. Intersecondary veins, tertiary veins, and quaternary veins are shown. b Esenbeckia grandiflora. Excurrent connections between primary and secondary veins (arrows). c Conchocarpus heterophyllus. At the leaf apex, connections between primary vein and secondary veins are decurrent (arrows), like those of the leaf base. d M. nigra. Tertiary veins with opposed (black arrows) or alternate (white arrows) patterns forming mixed percurrent arrangement. e E. pilocarpoides. Irregular quaternary reticular veins with branched ends. Meeting points between quaternary veins pointed by arrows. f M. maracasana. Quaternary veins freely branched. g Helietta apiculata. Poor development of areoles. h M. stipularis. Good development of areoles. i E. irwiniana. Good development of areoles. j E. grandiflora. Incomplete terminal venation. k M. nigra. Looped terminal venation. Bars a, c, d, h 3 mm; b 1 mm; e, f, i, k 400 µm; g 900 µm; j 500 µm; 1 primary veins; 2 secondary veins; i2 intersecondary veins; 3 tertiary veins; 4 quaternary veins; A areoles

Full size image

Reproductive structures (characters 37–45)

Although the literature used to compile these results presents different details for different species in some cases (e.g., types of imbricative estivation of the corolla and endocarp anatomy) or may lack data for some characters (e.g., polyembryony), no inconsistency was found in reproductive structures descriptions.

The flowers of Balfourodendron riedelianum and Raulinoa echinata are tetramerous, while those of the other species studied are pentamerous. Some species of Helietta other than H. puberula present some tetramerous flowers among the prevailing pentamerous ones, though. Corollas are red to purple in Raulinoa echinata and Metrodorea nigra, and they are white to cream in all other species. The estivation of the corolla is valvate in Metrodorea and imbricate in the other species. Complete connation of the carpels occurs in all studied species except for Pilocarpus microphyllus and Esenbeckia pilocarpoides, in which the carpels are connected through the base, and Conchocarpus heterophyllus, in which they are connected at the central axis and at the base.

All Metrodorea species, Esenbeckia pilocarpoides, E. pumila, E. irwiniana, and Zanthoxylum rhoifolium present muricate fruits; E. grandiflora bears echinate fruits; and other species present smooth fruits. Dorsal expansions are present and may be dorsal apophyses in Esenbeckia, Metrodorea, Pilocarpus, and Raulinoa; or winged structures in Balfourodendron and Helietta. All of the studied species have been described as having biovulate locules, although they may differ regarding the number of seeds per locule. In Metrodorea nigra, for example, there are normally two seeds per locule, although variations on this pattern may occur, having also been described by Souza et al. (2008). The occurrence of polyembryony is reported in E. grandiflora and E. pilocarpoides (Kaastra 1982).

Phylogenetic inferences

The result of combined BA (Fig. 5), compared with the Bayesian tree made exclusively with molecular data by Dias et al. (2015), shows a similar topology for Metrodorea species, with a high posterior probability support value (PPS = 100%) corroborating the monophyly of the genus. Otherwise, the addition of morphological data reduced the support of the clade M. maracasana + M. concinna + M. flavida + M. mollis (PPS from 94 to 80%).

Fig. 5
figure 5

Majority-rule consensus Bayesian tree of Metrodorea and related species. Numbers above branches are posterior probability percentages

Full size image

Raulinoa is the sister group of Metrodorea in our tree (PPS = 86%). The clade formed by Metrodorea + Helietta + Balfourodendron + Raulinoa echinata in Dias et al. (2015) has the posterior probability of 64%, lower than the 84% value obtained by us with combined data.

Relationships of other studied species, most of them devoid of molecular information, showed a lower resolution (polytomies or lower PPS value). For example, Esenbeckia grandiflora, E. leiocarpa, and E. pilocarpoides species devoid of molecular data appeared in a polytomy together with Galipea trifoliata and a clade formed by other 13 species. However, the larger data set (with morphological and molecular data) allowed us to retrieve Raulinoa and the “Balfourodendron alliance” as close relatives of Metrodorea with a better support.

ML analysis (Fig. 6) with the molecular data set of Dias et al. (2015) recovered the same topology of their Parsimony bootstrap tree. For Metrodorea, we have found similar bootstrap values. Notwithstanding, the relations between Metrodorea, Raulinoa, and the “Balfourodendron alliance” presented bootstrap values under 50, so it is not possible to point the sister group of Metrodorea in this analysis since these values do not represent accurate clades (Hillis and Bull 1993).

Fig. 6
figure 6

Maximum likelihood tree of Metrodorea and related species. Numbers above branches are bootstrap values based on 1000 replicates (values below 50% are not shown)

Full size image

Concerning the ML optimizations of morphological characters within the ML tree generated with molecular data, possible synapomorphies of the genus Metrodorea are: leaflet abortion, vascularized intrapetiolar stipule (with loss of vascularization in M. nigra), sinuous cell walls contour in surface view of the epidermis, glandular trichomes on the adaxial and proximal region of the petiole, good development of areoles, muricate fruits, and valvate estivation of the corolla (Fig. 7). Also, the presence long non-glandular trichomes concentrated in the adaxial-proximal region (character 12) and the presence of looped veins (character 36) would be synapomorphic for Metrodorea with 0.98 and 1.00 proportional likelihood (PL), respectively. These features are found in Raulinoa echinata and appear with 0.38 PL (character 12) and 0.2 PL (character 36) in the common ancestor of Raulinoa and Metrodorea that is shared with Balfourodendron and Helietta. Since the relationship between Metrodorea and these genera is not solved with a good support in our phylogeny, it is not possible to confirm that this feature appeared in the ancestral of Metrodorea.

Fig. 7
figure 7

Optimizations of characters that states changes represent possible synapomorphies in Metrodorea. Black and white balls in the nodes are probabilities graphs of which state these ancestral present. Striped balls represent non-applicable characters related to equivocal nodes (plain gray balls). At the right bottom corner, a compilation of all detected possible synapomorphies in the Metrodorea phylogeny

Full size image

Our topology shows an early divergence of Metrodorea stipularis, with the remaining species forming a clade supported by the concentration of crystals in the hypodermis, but this feature was not observed in M. mollis. Nevertheless, it can be just a result of the rarity of crystals in all tissues of this species leaves rather than a reversion. Within this clade, M. nigra emerged as sister group of a clade formed by the other four species, and M. maracasana as the sister group of the clade formed by M. concinna, M. flavida, and M. mollis, but no morphological synapomorphies could be identified that support these relationships. Otherwise, the crown group, composed by M. concinna, M. flavida, and M. mollis, is corroborated by the presence of margins of the stipules containing short intertwined trichomes and papillary cells that enhance adhesion.

Discussion

“Unifoliolate leaves” are a condition usually applied to several Rutaceae-bearing leaves with a joint or articulation at the base of the blade (e.g., Kaastra 1982; Kallunki and Pirani 1998; Pirani 1999; Kubitzki et al. 2011). This is mostly upon the reasoning that in a phylogenetic sense these plants would have reduced the number of leaflets during evolution to one. Any trying to find an anatomical or ontogenetic correlation between articulations and the division of the leaf blade has been unsuccessful since there is no evidence that these structures represent some residual feature from ancestral extra leaflets, as previously shown (Muntoreanu et al. 2011; Cruz et al. 2015). Thus, unifoliolate leaves reported for the Rutaceae members studied herein are identical to simple leaves with respect to development, except for some Metrodorea species leaves (in all described species, except M. maracasana) whose leaf primordia bear three leaflet primordia each (two of them may be aborted during development), so we classify them as early trifoliolate leaves (as in Cruz et al. 2015).

It is remarkable that some species like Esenbeckia leiocarpa, even though presenting a small swelling at the petiole, are classified as simple-leaved (Kaastra 1982). E. grandiflora also bears a swelling but associated with a constriction and is classified as unifoliolate by the same author. On the other hand, Kaastra (1982) classified as unifoliolate the leaves of E. pilocarpoides, which bear no conspicuous swelling but instead a distinct articulation at the petiole apex. Here, we conclude that it is more accurate to distinguish the possession of articulations and characterize associated constrictions or swellings, instead of mingling such character with the states of simple blade versus divided in leaflets. Anyway, we recovered the abortion of leaflet primordia as a possible synapomorphy of Metrodorea (Fig. 7), while the evolution of swellings and constrictions at the petiole apex of several among the sampled taxa have not shown a clear pattern. A similar condition of abortion of leaflet primordia may occur also in E. pilocarpoides subsp. maurioides Kaastra (not sampled here), since it presents in the same specimen simple leaves and leaves with two or three distinct leaflets (the lateral ones are smaller and sometimes partially connate to the terminal one). This diversity of leaf development and of leaf base structures is certainly promising to be investigated in further more extensive studies in the family.

Regarding other correlations with the division of the leaf, Duval (1903) associated the number of palisade parenchyma layers to compound and simple leaves in Pilocarpus. According to this author, species with simple leaves do not have palisade parenchyma with one single layer of cells. The simple-leaved species Raulinoa echinata and Esenbeckia leiocarpa (according to Kaastra 1982) have one layer of palisade parenchyma, so it was not possible to expand the generalization of Duval (1903) to other genera. Also, Muntoreanu et al. (2011) described one layer of palisade parenchyma in simple leaves of P. spicatus and P. pauciflorus, contradicting Duval (1903).

Metcalfe and Chalk (1950) reported that hairs may be infrequent in many members of Rutaceae. This may apply for Metrodorea, for example, that lacks non-glandular hairs in the leaf blades of most of its species. Kaastra (1982) and Pirani (1998) observed that the occurrence or absence of these trichomes can vary within the same species, but we did not observe such variation within the analyzed material. It is possible that the reported variation can be observed in younger or older leaves than the ones that we analyzed and it may be related to dehiscence process of these trichomes. More development studies are necessary to investigate this possibility.

Secretory cavities are the most characteristic feature of the Rutaceae and are described as schizogenous or lysigenous (Metcalfe and Chalk 1950; Kaastra 1982). However, cell lysis is often an artifact of the fixation procedure employed during the preparation of plant material (Turner et al. 1998), and our anatomical data with the applied technics do not bring any further evidence to contribute to this discussion. Machado et al. (2016) present a detailed study of the histochemistry, ultrastructure, and ontogeny with careful fixation procedures of the oil cavities in Metrodorea nigra, describing a schizo lysigenous and seasonal development for the species.

The central vascular arch observed in the midrib region of Esenbeckia irwiniana is a feature that had been described in some plants as a medullary plate (Radford et al. 1974), and it is the first report for this structure in the vascular system of Rutaceae.

The types of crystals observed in the present study have already been described in some Rutaceae, and here, we register their occurrence for more species. Prismatic crystals found in compartments created by walls transversal to the length of the cell in palisade parenchyma are similar to Pilocarpus druses that also are organized in this way in the species studied by us and by other authors (Metcalfe and Chalk 1950; Muntoreanu et al. 2011).

The venation pattern typically pinnate and simple brochidodromous observed in all studied species of the present paper was also cited by Muntoreanu et al. (2011) in other species of the family. This pattern seems to be very old, as it was described in fossilized specimens of the Rutaceae (Vepris and Clausena) dating from the late Oligocene period (Pan 2010). It may be the only venation pattern of Rutaceae species for primary and secondary veins, or at least the most common. Dede (1962) proposed a classification of veins patterns in Rutaceae based on the position of secretory cavities in 80 genera. Despite he presented an extensive study, this classification does not allow a comparison with other families and it is hard to be made with regular clearing technics, and with herbarium or fossilized material, as they do not properly preserve secretory cavities. So, we recommend for new Rutaceae venation studies a detailed description according to Ellis et al. (2009), as we did. More details about the venation, like areolation and terminal veins, may be of big importance in new studies with other species and may provide potential phylogenetically informative characters, as we state below.

Although reuniting morphological data is essential to systematics studies, when used to make phylogenetic analyses, it usually generates matrixes much smaller than matrixes made with molecular data. So, when we prepared the combined matrix, morphological data were clearly underrepresented, except when molecular data were not available. This does not necessarily imply in a smaller accuracy, even with a lot of missing data, as Jenner (2004) stated in his good discussion about the use of combined morphological and molecular data in phylogenetic reconstructions.

The adhesion promoted by intertwined trichomes in the stipules of Metrodorea species was already explored by us in our previous study (Cruz et al. 2015), and here, we detected that when present in the margin of stipules, these trichomes represent a good synapomorphy for the crown clade, composed by M. concinna, M. flavida, and M. mollis.

There are discrepancies between the relationships of Raulinoa with the Helietta + Balfourodendron clade and with Metrodorea in the topologies recovered by us and by Dias et al. (2015). Considering a polytomy containing these three groups, there are three possible rooted trees as resolutions (Fig. 8). The clade Raulinoa + Metrodorea would find support based on the following structural characters: the presence of long intertwined trichomes connecting the base of the leaves, the absence of glandular trichomes on the leaf blade, the presence of decurrent veins only at the base of the leaf blade, and presence of looped terminal veins. In the second proposition, Helietta + Balfourodendron is considered sister group of Metrodorea based on the existence of constrictions between the leaf base and blade. The third possibility, in which Raulinoa is sister group of Helietta + Balfourodendron, would find support based on the synapomorphy of poor development of areoles. The acceptance of any one of these solutions implies the presence of homoplasies in character states that sustains the other two proposals. Clearly, the first assumption minimizes the number of necessary homoplasies to explain the relationship.

Fig. 8
figure 8

Possible relationships between Metrodorea, Helietta, Balfourodendron and Raulinoa, and potential synapomorphies. a Metrodorea and Raulinoa are sister groups. This hypothesis is supported by a larger number of synapomorphies. b “ Balfourodendron alliance” is the sister group of Metrodorea. c Metrodorea is the sister group of a clade formed by the three other genera

Full size image

The findings presented herein allowed some robust hypotheses to be drawn about the evolution of some anatomical aspects of the studied species. Thus, the relationship of Metrodorea and Raulinoa postulated previously by Cowan (1960) and Kaastra (1982) is now additionally supported by a large number of possible structural synapomorphies. However, it is important to emphasize that the relationship between the stipules of Metrodorea and the thorns of Raulinoa (which are modified branches) postulated by Cowan (1960) has no grounds of homology and so cannot be taken as an appropriate evidence of the close relationship of the two genera, a fact previously pointed out by Kaastra (1982).

In summary, our study shows morphological features which are phylogenetically informative characters in Metrodorea and related Rutoideae, by using new leaf anatomy and venation data, including areolation and terminal veins, and from available data for reproductive structures. Further investigation on their structure certainly will bring additional evidence to keep on improving the knowledge about the systematics and evolution of the group.