Although the pollinarium and pistil characters have been widely studied in the family, a detailed study of the structures composing the pollinarium and their different states remains a necessity. Most of the references on this subject (e.g. Dressler and Dodson 1960; Schill and Pfeiffer 1977; Dressler 1981, 1993; Rasmussen 1982; Blackman and Yeung 1983; Burns-Balogh and Funk 1986; Kurzweil 1998; Freudenstein and Rasmussen 1996, 1997; Johnson and Edwards 2000; Rasmussen and Johansen 2006; Singer et al. 2008; Pedersen et al. 2013; Kant et al. 2013; Freudenstein and Chase 2015) do not show detailed images and/or schemes relating to types, shapes, and organization of the accessory structures of the pollinarium. A detailed study in a large group, such as Epidendroideae, allows better discrimination and interpretation of their evolution, affinities, origins, and possible functions. Our study corroborated the great variability in the Epidendroideae highlighted previously by several authors (Freudenstein and Rasmussen 1999; Szlachetko and Margońska 2002; Szlachetko 2003; Cameron 2003; Rasmussen and Johansen 2006; Singer et al. 2008; Nieto and Damon 2008; Szlachetko and Mytnik-Ejsmont 2009; Damon and Nieto 2012). Most of the characters are very variable within tribes and subtribes. The results of the ancestral states reconstruction highlighted this.
On the other hand, the main gynostemium traits were used to perform classifications, but there is very little information about structure and ontogeny (formation and types) of ovules. Therefore, studies like this one will always be a contribution, because knowledge about the states of the characters of the pollinarium and pistil allows a better understanding of the great variability of the subfamily Epidendroideae. It is especially important to work with fresh material that permits a better visualization and interpretation of the morphology, avoiding possible errors derived from the deterioration of the most fragile structures.
The reconstruction of the ancestral character states in Epidendroideae with maximum parsimony and maximum likelihood suggested a MRCA of the Epidendroideae, the complete pollinarium, with granulate pollinium, caudicle, tegular stipe, and viscidium. In resolving the most parsimonious ancestral character state when the analyses excluded or included representatives of other subfamilies (results not shown), some results vary. This could be related to the many changes that have occurred in the characters over time, probably caused by the strategies of dispersal of the pollinium and the pollinator type. The tribes that still preserve full pollinarium are Epidendreae (Calypsoinae) and Cymbidieae (Catasetinae, Coeliopsidinae, Eulophiinae, Eriopsidinae, Maxillariinae, Oncidiinae and Stanhopeinae). There are many similarities in the pollinarium of Oncidiinae and Aeridinae; both subtribes share the number of pollinia (two or four), the existence of a suture (in the double type), the short and irregular freniculae caudicle, and tegular stipe. The exception is the viscidium, which is very similar in almost all Oncidiinae (removable viscidium, shaped platform), but highly diversified in Aeridinae. The nature of the pollinarium is a principal factor in understanding the evolution of the family Orchidaceae. It has been used as a relevant character in the classification and species identification because its features are taxonomically informative (Dressler 1986; Singer and Koehler 2004; Ramírez et al. 2007). However, some authors suggested parallel or convergent trends in pollinarium/pollinium morphology in the orchid groups (Williams 1970; Stenzel 2000; Dathe and Dietrich 2006) and that pollinarium types must have evolved several times independently (Singer and Koehler 2004; Dathe and Dietrich 2006).
The compaction of pollen in the pollinia in Epidendroideae may probably be caused by the modification of the strategies of pollen dispersal, in comparison with the usual dispersal known in other plants. It could be due to the fact that the masses of pollen in variable quantity and weight turned into ballast for pollinators whose body structure would not support such dimensions (because of pollinia size and weight). Probably, as part of the evolution in the family, the attached or accessory structures of the pollinia (caudicle, stipe, and viscidium) were formed to facilitate transport and together with the pollinia, constitute the pollinarium. Previously, other studies have mentioned that the caudicle, stipe, and viscidium attached to pollinium have different origins (Dressler 1993; Johnson and Edwards 2000; Singer and Koehler 2004). It has also been inferred that the association of the accessory structures with the pollinium is closely related to the strategy of transfer of pollen (Dressler 1986; Pacini and Hesse 2002; Singer and Koehler 2004) and that the structure and function of the pollinarium cannot be studied isolated from other structural modifications which occurred during the evolution in Orchidaceae (Dressler 1986; Johnson and Edwards 2000).
The Epidendroid caudicles are of two types: (1) Freniculae, elastic and sticky, that are considered to be the apomorphic state (Freudenstein and Rasmussen 1997). This occurs in Cymbidiinae, Catasetinae, Maxillariinae, Stanhopeinae, Coeliopsidinae, Oncidiinae and Aeridinae. (2) Appendiculae caudicle composed by elastoviscin and pollen occurs in Bletiinae, Collabiinae, Coelogyninae, Sobralieae, Laeliinae, and Pleurothallidinae. Both types were proposed by Mansfeld (1935) and accepted by Freudenstein and Rasmussen (1997). The two types vary in length; the longest occurs in some Bletiinae, Collabieae, Sobralieae, and Laeliinae, which coincide with the absence of the rest of the accessory structures. The appendiculae caudicles are considered as an extension of the pollinia (Freudenstein and Rasmussen 1997). Probably the absence of viscidium, for which the main function is to stick the pollinium to the pollinator’s body, losing this function, wherein the caudicle then functions as the breaking point to facilitate deposition of the pollinia on the stigma (Johnson and Edwards 2000). There is no relationship between the size and the number of the pollinia and their evolution. This is evident in some Laeliinae (Brassavola, Laelia, Schomburgkia, and Sophronitis) having eight large pollinia. One would expect that the more the pollinia, the smaller they will be. Their reduction would make transport of the pollinarium by the pollinating vectors easier and allows us to hypothesize that probably because of the difficulty in transporting the pollinarium of these taxa, it has a completely different morphology from the rest: Their joined caudicles form a mat with the pollinia attached to both ends (Fig. 2a). A study in Laelia rubescens revealed that only four of the eight pollinia are attached as a unit through the column of viscous material. They are transported by the animal vector to rather large receptive flowers, while still forming a package of four large pollinia relative to the stigmatic surface, making it difficult and mechanically unlikely that pollinia from different sources may be deposited on the stigma (Trapnell and Hamrick 2006).
The stipe of Epidendroideae is usually individual and tegular, but is variable in length and shape. The pollinarium is hamular only in the Tropidieae. Both types, tegular and hamular pollinia, originate from the rostellum (Szlachetko and Rutkowski 2000). The last accessory structure composing the pollinaria is the viscidium, that in Epidendroideae can be diffuse (Kurzweil 1988; Freudenstein and Rasmussen 1997; Rothacker 2007; Pedersen et al. 2013) or detachable (Kurzweil 1988; Freudenstein 1994; Freudenstein and Rasmussen 1997; Rothacker 2007) and plays a very important role in the dispersion of the pollinarium. Some authors highlight the functional importance of the stipe and viscidium in the pollination process (Dressler 1981, 1993). The stipe, by its length, facilitates the approach of the pollinia to the stigma (Hidayat et al. 2006), and the viscidium, by its viscous and sticky consistency, adheres the pollinium to the body of the pollinator (Johnson and Edwards 2000). However, it was observed in this study that in some cases in the absence of both structures, the caudicle assumes the function of adherence of the pollinia to the pollinator, as occurs in Laeliinae (Oerstedella, Prosthechea, Scaphyglottis, Brassavola, Sophronitis, and Laelia). Laelia is a good example because most of them lack viscidium and its role is played by the caudicle. The much better sample is Pleurothallidinae because most members lack viscidium and its role is played by the caudicle. Finally, some members of Arethusinae (Arundina) and Dendrobiinae (Dendrobium) studied have sessile pollinia.
As expected, the reconstruction of ancestral states in the subfamily suggested for the MRCA of Epidendroideae pollinium texture that the granulate state is the most parsimonious. Nevertheless, a detailed study of the whole family (Mosquera-Mosquera et al. unpublished data) revealed that the pollinia texture has also changed over time in different directions. In this study, the granular state (Fig. 7) is only observed in basal tribes and subtribes belonging to Epidendroideae (Neottieae, Sobralieae, Arethusinae, Xerorchideae, and Agrostophyllinae) agreeing with previous authors (Dressler 1981; Hesse et al. 1989; Rothacker 2007; Chase et al. 2015). In Orchidaceae, only Neuwiedia and Apostasia belonging to the subfamily Apostasioideae possess powdery pollen not forming pollinia and without elastoviscin. Therefore, the occurrence of the granular pollinia with easily friable tetrads in Epidendroideae indicates that very little elastoviscin is present inside the pollinium reducing the cohesion of the tetrad, this condition being very similar to that of members of Apostasioideae (Dressler 1993; Mosquera-Mosquera et al. unpublished data).
As an intermediate condition, the sectile pollinia (Fig. 7) with the pollen organized in massulae occurs in Nervilieae, Tropidieae, Wullschlaegelieae, and Thaieae. Sectile pollinia are present in about half of the Orchidoideae species and are characteristic of the tribe Orchideae and subtribe Goodyerinae (Dressler 1993; Chase et al. 2003; Ramírez et al. 2007). It is also characteristic in the Epidendroideae subtribes Gastrodiinae, Epipogiinae, Nerviliinae, Arethusinae, Thelasiinae and Laeliinae (Freudenstein and Rasmussen 1997), Neottieae (Rothacker 2007) and Tropidieae (Singer et al. 2008). In relation to this state, Freudenstein and Rasmussen (1997) recognize two states in the whole family that they termed Orchidoid and Epidendroid. The advantage of the sectile pollinia is that a single pollinium can pollinate several flowers (Freudenstein and Rasmussen 1997; Johnson and Edwards 2000; Pacini and Hesse 2002; Harder and Johnson 2008).
Finally, most of the taxa studied show a compact or hard pollinium (Fig. 7), very common in Epidendroideae, representing the highest degree of cohesion of the pollen that is deposited as a unit on the stigma (Harder and Johnson 2008). The presence of a hard pollinium was a very important innovation for radiation of the subfamily which became the most diverse, successful and widely distributed within Orchidaceae as previously suggested by Dressler (1993). The sectile and hard pollinia seem to be apomorphic within Orchidaceae (Freudenstein and Rasmussen 1997). The sectile pollinia arose at least four or five times in the evolution of the family (Burns-Balogh and Funk 1986; Dressler 1993; Pansarin and Estanislau do Amaral 2008), particularly in Epidendroideae.
Within the Epidendroideae, the diversity of the pollen texture is such that this subfamily presents all types reported for the Orchidaceae. It seems that texture is the result of the gradual aggregation of pollen (Harder and Johnson 2008). In this process, elastoviscin has played an important role allowing the existence of different dispersion units, from monads to tetrads (Pacini and Hesse 2002), and soft, sectile, or hard pollinia (Dressler 1986; Johnson and Edwards 2000). The pollen in monads, found in the most basally diverging orchid genera, belonging to Apostasioideae is the primitive condition, but it has evolved several times in the orchid family (Dressler 1993; Cameron et al. 1999; Pansarin and Estanislau do Amaral 2008). Only Apostasioideae lacks this viscous and sticky substance, and the pollen is dispersed individually (Mosquera-Mosquera et al. unpublished data). The aggregation of the pollen in angiosperms has happened at least 39 times and independently; we think that the above-mentioned changes improve the efficiency in the process of dispersion of the pollen (Harder and Johnson 2008).
There is also much diversity in the number of pollinia in Epidendroideae (Fig. 7). Although parsimony analysis does not define which of the states (two, four, six, or eight) is the most parsimonious for the MRCA of the subfamily, the four pollinia reached the highest value (0.5880) in ML analyses. Therefore, this analysis would support the suppositions to interpret this character (Freudenstein and Rasmussen 1996). We recognized four types in the subfamily. The plesiomorphic state is the presence of four pollinia (Rasmussen 1986; Rothacker 2007). It occurs in Maxillariinae, Oncidiinae, Laeliinae, Pleurothallidinae, Zygopetalinae, Calypsoinae, Neottieae, Ponerinae, Agrostophyllinae, Podochileae, Coelogyninae, Polystachyinae, Malaxideae, and Dendrobiinae. The presence of six pollinia was observed in this study only in the genus Appendicula (Podochileae), besides being a rare character state in the family. The sectile texture corresponds to an intermediate state, not only in the number of pollinia but also in the trend to compaction of the pollen in Orchidaceae. It is interpreted that the compaction of the pollen has evolved from granular to sectile and then to compact (Dressler 1986; Freudenstein and Rasmussen 1996, 1997; Harder and Johnson 2008). Six pollinia have already been mentioned in the genus Appendicula (Keng et al. 1998) and for two Pleurothallidinae: Chamelophyton and Brachionidium (Stenzel 2000). The third case originated by the division of the 4 pollinia resulting in 8 (Freudenstein and Rasmussen 1996), as we observed in some members of Agrostophyllinae, Collabieae, Bletiinae, Sobralieae, Laeliinae, Arethusinae (e.g. Arundina (Szlatchetko and Margonska 2002), Xerorchideae and Podochileae. The last one, contrary to the previous one, occurred after the reduction from four to two pollinia or the fusion of the initial four pollinia. This is the most frequent case in most taxa of Epidendroideae. Subtribes having this state are Oncidiinae, Catasetinae, Cyrtopodiinae, Stanhopeinae, Eulophiinae, Nervilieae, Thaieae, Calypsoinae, Wullschlaegelieae, Tropidieae, Coeliopsidinae, Cymbidiinae, Pleurothallidinae, Eriopsidinae, Aeridinae, and Angraecinae. This change in the number of pollinia occurred especially in five subtribes with different trends, in Aeridinae, Oncidiinae, and Pleurothallidinae (reduction from 4 to 2), Agrostophyllinae and Laeliinae (a division from 4 to 8). A hypothetical development model of the pollinia number revealed that the shape and type of pollinia are produced by the fusion or the septation of meristematic regions in the anther tissue in the subfamily Epidendroideae. The meristem septation produces four or eight pollinia per anther, while the lack of septation resulted only in two pollinia per anther (Freudenstein and Rasmussen 1996). The anther formed by four pollinia seems to be primitive in the family, and the anthers with two or eight pollinia derived from them (Freudenstein and Rasmussen 1996; Cameron et al. 1999). These changes in the number of pollinia disagree with the hypothesis of a reduction pattern from eight to six, four, or two pollinia (Dressler 1993) in many groups of Epidendroideae.
Undoubtedly, there is a relationship between the number and orientation of pollinia, with some exceptions (e.g. Dendrobium and Schomburgkia). In most cases, there are four superposed pollinia, whereas in the case of two, six, or eight pollinia, they are juxtaposed, although in another study two pollinia were treated as superposed (Freudenstein and Chase 2015). A differential overgrowth of the thecae and the flattening of the anthers (by a lack of full reorientation of the juxtaposed thecae) give rise to the superposed pollinia (Freudenstein and Rasmussen 1996; Freudenstein et al. 2002; Freudenstein and Chase 2015). The subtribes Oncidiinae, Pleurothallidinae, and Aeridinae have two or four pollinia when they are double and are juxtaposed. These changes of orientation have happened in Epidendroideae on many occasions from the most basal taxa to the most recent (Fig. 7). The most parsimonious state for this character is the pollinia juxtaposed.
Freudenstein and Chase (2015) treated the double pollinia as superposed even in the cases the pollinium have a cleft or a bilobed appearance, and suggested would be an evidence of a partition that have been lost or are still being developed. It was also observed in some double pollinia studied in this research and occurred mainly in the subtribes Stanhopeinae and Oncidiinae. The presence of a suture that can vary in length and location in the pollinium, and contrary to that proposed by Freudenstein and Chase (2015), are pollinia juxtaposed because they are arranged in the theca next to each other.
The orientation of the pollinium is a condition widely related to the rotation of the thecae and the anther dehiscence (Kurzweil 1987a, b; Dressler 1993; Freudenstein and Rasmussen 1996; Freudenstein et al. 2002), associating them this way: latrorse/superposed, introrse/juxtaposed. Studies on the anther ontogeny showed that the plesiomorphic state is the introrse anthers with juxtaposed pollinia and there have been several derived states; nevertheless, the Vandeae tribe that has latrorse anthers and superposed pollinia is believed to have experienced little or no reorientation. Therefore, it is likely that the current pollinia orientation states are due to heterochronic changes in the anther (Freudenstein et al. 2002).
The most parsimonious state for the MRCA of Epidendroideae is the entire pollinium (Fig. 7) without suture (ML = 0.9929). Therefore, the presence of a suture in the pollinium must be considered apomorphic as it is present in most of the double pollinium. (It is the advanced/apomorphic state.) The suture is present only in a few groups and has variations in length and location in the pollinium of the subtribes Angraecinae, Catasetinae, Eulophiinae, Cyrtopodiinae, Maxillarinae (Cryptocentrum sensu Chase et al. 2003), Coeliopsidinae, Stanhopeinae and Oncidiinae and the tribe Sobralieae. Probably, one of the persistent signs, which allow inferences about the reduction of the pollinium, is the presence of the suture. That is not exclusive to the double pollinia, as it was also observed in the four pollinia in Cryptocentrum and eight pollinia of Elleanthus (Sobralieae). In the pollinia, the suture is the result of the union of the pollen content of two cores in a late stage of ontogeny; therefore, it should be considered as a different state (Freudenstein and Rasmussen 1996).
Unlike the pollen, the pistil in the subfamily Epidendroideae has very little variation. It is unilocular, but in Huntleya (Zygopetalinae) and Peristeria (Coeliopsidinae), the pistil appears to be trilocular because the tissue of the placenta intrudes and makes contact in the centre of the locule creating three false locules. In the literature about Orchidaceae, one locule has been reported for Epidendroideae and three for Apostasioideae and Cypripedioideae (Garay 1960; Atwood 1984; Freudenstein and Rasmussen 1999; Cameron 2003). The last is the plesiomorphic state in the family. The same occurs in taxa belonging to families related to the Orchidaceae—e.g. Hypoxidaceae (Rudall and Bateman 2002; Kocyan 2007; Kocyan et al. 2011). However, some taxa such as Cattleya, Laelia, Sophronitis, Scaphyglottis (Laeliinae), Vanda (Aeridinae), and Angraecum (Angraecinae) have an extension of placental tissue, suggesting a hypothetical intermediate state. In a phylogenetic context, the further retraction of the placenta to form the unilocular ovary has occurred at least four times in orchids (Cameron 2003).
In Orchidaceae, non-fertilized flowers have ovules with poorly differentiated tissues (Sagawa and Israel 1964; Zhang and O’Neill 1993; Li et al. 2016). This was verified in most of the samples of Epidendroideae studied. Only two genera present well-differentiated ovules. One is Stanhopea, having primordial ovules presenting an erect position and differentiated teguments clearly bitegmic in the initial stages (Fig. 6a). The second is Epipactis, having an anatropous ovule (Fig. 6b). It has been asserted that before the integuments are differentiated, the curvature of the apex row cells occurs in the primordial anatropous or campylotropous ovules (Endress 2011). It defines the location of the chalaza and the micropyle position. However, more evidence is required to consider it as an orthotropic ovule given that there are no references to the existence of this type in the Orchidaceae family.
Orchidaceae belongs to the group of angiosperm families with larger numbers of ovules, which incidentally coincides with the presence of pollinia that can bind between 5000 and 4,000,000 pollen grains (Schill et al. 1992). There are two ways to interpret this phenomenon; the first is the possibility that the selection has favoured the formation of numerous small seeds and second that selection promoting, as secondary adaptation for a large number of ovules, the agglutination of pollen grains that ensure their fertilization (Johnson and Edwards 2000).
During the evolution of Epidendroideae, the unilocular condition of the ovary (Fig. 6) is probably very advantageous for pollination, given the high number of ovules (usually over 100) attached to the placenta. The presence of a spacious area within the carpel facilitates the growth of the many pollen tubes formed by the pollen grains agglutinated in the pollinia (Rudall et al. 2002). It is expected that genera having the highest values of ovules, and a diffuse placenta have much more success in the fertilization (e.g. Epipactis, Corallorhiza, Spathoglottis, Dryadella, Masdevallia and Trisetella) than other Epidendroideae having pistils with linear placenta (one or two rows) in an area along the placenta of the ovary.
Another interesting characteristic is the presence of intra-ovarian trichomes. In this study, intra-ovarian trichomes were observed in subtribes Coeliopsidinae and Aeridinae. The intra-ovarian trichomes appeared on the placenta of ovules with poorly differentiated cells located in the mid-ovary region of Vanda. To describe the type of trichomes, we considered the theories proposed by some authors that indicate that the development of the ovary occurs after fertilization (Clifford and Owens 1990; Zhang and O’Neill 1993; Tsai et al. 2008; Mayer et al. 2011). In addition, Zhang and O’Neill (1993) described the development of hair cells emerging from the endocarp after pollination during the development of the fruit. Similarly, Mayer et al. (2011) also described the presence of endocarpic trichomes in fruits of Oncidium flexuosum only after pollination. However, in this study, only ovaries at different floral stages were studied and not fruits.
In Orchidaceae, the subject of intra-ovarian and endocarpic trichomes has scarcely been addressed so far; they have been described in Aeridiinae (Hallé 1986), in Angraecinae (Freudenstein and Rasmussen 1999), and in Oncidium flexuosum (Mayer et al. 2011). They have also been studied in other monocotyledons, such as Asteliaceae belonging to Asparagales (Rudall et al. 1998). In this last case, they have been associated with the production of mucilage that covers the entire cavity of the ovary. Likewise, the trichomes located in the placenta have been reported for both monocotyledons (Araceae (French 1987)) and dicotyledons [e.g. Salicaceae, Sapotaceae, Urticaceae, and Euphorbiaceae (Maheshwari 1950; Nagaraj 1952; Steyn et al. 1991)].
The results of our investigation indicate that pistils in the subfamily Epidendroideae are formed by three carpels divided into six valves, but only half of the six alternating valves are fertile. However, the pistils of Cattleya and Sophronitis have a very different pattern from the other Epidendroideae. Only the fertile carpels are well distinguished. We interpret that small sterile valves are located between the fertile valves and it seems to have pistils with three carpels (Fig. 5a). Probably, this uncommon type of pistil represents an intermediate state between pistils with three carpels and pistils with three carpels divided into six valves. Further investigation into this aspect is necessary to determine if this organization is the result of the decrease in the number of the valves or the division of the carpels. The number of carpels forming the pistil in Orchidaceae has been controversial. The presence of six carpels has been mentioned as a characteristic of the family (Vermeulen 1966; Puri 1951), but it has also been indicated that there is no evidence that the gynoecium consists of six carpels (Kurzweil 1998). Rasmussen and Johansen (2006) showed that in some cases the pistil is divided into six valves and, when this happens, the three sterile valves correspond to a sepal base and the three fertile ones to a petal base. In other cases such as subfamily Apostasioideae only three fertile carpels have been reported (Kocyan and Endress 2001; Mosquera-Mosquera et al. unpublished data).
All these carpel structural changes in Epidendroideae are clear evidence of the evolutionary trends in the ontogeny of Orchidaceae, from three isomorphic carpels, in some Cypripedioideae and Apostasioideae, to six well-defined heteromorphic valve carpels in most of the Epidendroideae taxa studied here. Probably, the division of the ovary into these six valves is the result of a late development in ontogeny (Kurzweil 1998).
The study of the anatomy of the pistil in Epidendroideae revealed variations in the vascular tissue organization. Poorly vascularized carpels (with a single ventral bundle) are predominant. On the contrary, Trichocentrum, Stanhopea, and Trigonidium present highly vascularized carpels (with ventral, dorsal, and lateral vascular bundles). Swamy (1948) asserted something similar about the more primitive Apostasioideae and Cypripedioideae orchids: Six vascular bundles from the inflorescence axis enter and cross the pistil longitudinally. In other subfamilies, such as Epidendroideae, only three vascular bundles of the inflorescence axis reach the floral axis, then branching in each genus or species in a different way (Swamy 1948; Rao 1974).
Another component of the vascular system is represented by the lateral vascular bundles, which are located on the periphery of the carpels or are dispersed without following a regular pattern. These are smaller than the dorsal and ventral bundles and arise from the branching of the dorsal vascular bundles (Hardy and Stevenson 2000). It can be inferred that the abundant vascularization in taxa presenting ventral, dorsal, and lateral vascular bundles (e.g. Catasetum, Cattleya, Anguloa, Lycaste, Maxillaria, Sievekingia) is different from the plesiomorphic state—high vascularization lacking lateral vascular bundles—present in the most basal orchid genera (Apostasia, Neuwiedia, and Selenipedium) suggested by Rao (1974). On the contrary, abundant vascularization in Epidendroideae is a derived condition related to carpels with thick parenchyma. The detailed study of the vascularization of the Cypripedioideae, Apostasioideae, and Vanilloideae and also Orchidoideae and Epidendroideae clearly showed that there is no direct relationship between the first and the second groups, since in the latter there are patterns of high complexity associated with a large diversity that is reflected in the floral characters (Swamy 1948).
Trends in character evolution
The four reconstructed pollinia characters (number, orientation, texture, and the presence of suture in the pollinium) vary in Epidendroideae tribes and subtribes (Online Resource 2). The distribution of the character states across the subfamily could probably result from having evolved more than once in the subfamily.
Our results are consistent with the currently accepted systematic basis for the tribes and subtribes within Epidendroideae. The current taxonomic proposal is based on molecular data (Whitten et al. 2000; Pridgeon et al. 2001, 2003; Borba et al. 2002; van den Berg et al. 2005; Kułak et al. 2006; Sandoval-Zapotitla et al. 2010; Freudenstein and Chase 2015). In other cases, anatomical information of plant organs (Stern and Whitten 1999; Luo and Chen 2000; Stern and Judd 2001; Stern et al. 2004; Chase et al. 2008, 2009; Stern and Carlsward 2008) and palynological studies (Ackerman and Williams 1980; Hesse et al. 1989; Stenzel 2000) are also taken into account in the classification. Furthermore, in this study, we found that changes in the pollinarium and pistil are not linear and represent a contribution to understanding the evolution of the reproductive organs and a reconstruction of the pollinium character states permitting their evolution to be understood more clearly (Figs. 7, 8).
It has been hypothesized that the rapid radiation of Epidendroideae is probably associated with a specialization in pollination syndromes. As a result, this group has a mosaic of primitive and derivative morphological and molecular characters, with few autapomorphies and fewer synapomorphies (Cameron et al. 1999; Chase et al. 2003; Rothacker 2007; Whitten et al. 2014). The continuous structural changes in the pistil and orchid pollinarium establish a much narrower connection with specific animal pollinators, which could lead to the elimination of competition between and within taxa of this family. Although the impressive floral diversity of orchids has been attributed to the adaptation to specific pollinators the diversity is higher in species sharing a pool of generalist pollinators (e.g. Johnson et al. 1998; Cozzolino et al. 2004).