Methods

Plant material. Leaves of 88 wild specimens of Vanilla spp. from the MYP were considered in this study (Table 1). These samples were deemed wild because they were collected in natural settings without evidence of human management. Hereafter, these samples will be referred to as MYP, followed by an identifying consecutive numbering plus a key to the region where they were collected (Table 1). Of all specimens studied, only two of them were found fertile and positively confirmed as V. planifolia (“MYP1-CY” and “MYP2-CY”, both from the same population) (Figure 1-A). The other specimens were infertile and only leaves were collected from them. However, we could observe morphological differences in leaf and stem in some of these samples. The detailed information about the collecting site of the samples is not provided for protection reasons. Sterile vouchers were collected by location and deposited at herbarium CICY.

Figure 1 Flowers and vegetative structures of the Vanilla species reported for the Yucatan Peninsula and used in the analysis with ISSR molecular markers. A) V. planifolia wild (both photographs by Sara Villanueva [SV]). B) V. insignis (flower photography by Leon Ibarra González, and leaves photography by SV). C) V. odorata (flower photography by Manfred Speckmaier, and leaves photography by SV). D) V. sp. nov. aff. V. phaeantha (both photographs by SV). E) V. pompona (both photos by Verónica Borbolla). 

In the ISSR analysis, we included 11 cultivated individuals as controls (standard samples). These individuals belong to the four species of Vanilla reported for the MYP and distributed as follows:

Also, a single individual of Vanilla pompona originally from the state of Puebla (Figure 1-E), a species that could potentially occur in the MYP.

DNA extraction and ISSR technique. Total genomic DNA was extracted from fresh leaves following a modification of the ^{Dellaporta et al. (1983)} procedure, which involved a single protein washout using phenol:chloroform:isoamylic. DNA was quantified with an ND-1000 spectrophotometer (Nanodrop, USA) and its quality assessed by 0.8 % agarose gel electrophoresis (^{Sambrook et al. 1989}) and by PCR amplification of the 18S constitutive gene.

Three of the 10 ISSR primers specifically designed for Vanilla by ^{Verma et al. (2009)} were employed in this study: C07, C09 and T06. Amplifications were performed in a total volume of 20 μl, including 20 ng/μl of ADN, 200 mM of dNTPs (Invitrogen), 1.5 mM MgCl₂, 60 pg. of primer, 2.5 μl buffer 10x of Taq ADN polymerase, and 1.5 U Taq DNA polymerase (Invitrogen). Amplification conditions were as follows: an initial step of 5 minutes at 94°C, followed by 35 cycles, each consisting of 30 seconds at 94°C for denaturing, 90 seconds at 50-54°C (depending upon primer used) for aligning and 90 seconds at 72°C for elongation, and 5 minutes at 72°C for final extension. Amplifications were performed at least twice and only reproducible products (bands) were considered for data analyses. Amplified PCR products were screened by agarose gel electrophoresis with 1.5% buffer TBE 0.5X (pH 8.3) and finally stained with ethidium bromide (^{Sambrook et al. 1989}). Resulting gels were photographed and visually read with a 1 kb molecular marker as a reference for fragment length assessment.

Genetic relationships based on cluster analysis of the ISSR. Both monomorphic and polymorphic bands were considered for analyses and only well-defined bands (color intensity and reproducibility) included. The basic assumption was that bands of identical molecular weight regardless of whether belonging to individuals of the same or different population/species, represented the same allele. We compiled a database for each primer, recording the presence/absence of particular bands as 1or 0, respectively.

We performed a clustering analysis of both the standard samples as well as the MPY samples. Three different approaches followed for analyses:

Phylogenetic inference based on ITS. With the goal of obtaining an independent corroboration of the clustering patterns emerging from the ISSR analyses, we performed a phylogenetic analysis of the nuclear ITS (ITS 1 and 2) region. This study included all the individuals considered in the ISSR analysis and relevant outgroups (see below), representing both tribes (Pogonieae and Vanilleae) of Subfamily Vanilloideae (^{Chase et al. 2015}). Taxon sampling, consisting of 121 terminals, putatively representing 25 species, was devised to test the following three hypotheses:

To polarize phylogenetically informative character states, a species of the distantly related genus Cleistes Rich. ex Lindl. (tribe Pogonieae), Cleistes paranaensis Schltr. was employed as a functional outgroup. This taxon was identified as basal in a recent phylogenetic analysis of the genus (^{Pansarin et al. 2008}). The outgroup of the phylogenetic analysis consisted of the following taxa (Table 2):

The ingroup of the analysis consisted of several different Vanilla species, representing both morphological variation and geographical distribution of the genus, although Neotropical taxa thought to be related to V. planifolia constitute the bulk of the sample. Species included are:

The ITS region was amplified following ^{Poeaim et al. (2011)}, and using primers ITS1 and ITS4 (^{White et al. 1990}). Amplification reactions performed in 30 μl containing 50 ng/μl of ADN, 1.25 nm of dNTPs, 3 mM of MgCl₂, 20 pmol of each primer, 1.5 U Taq polymerase, and the buffer provided by the manufacturer (Qiagen). The PCR protocol was as follows: an initial denaturalizing at 95°C for 5 minutes, followed by a sequence of 30 cycles consisting each of 1.5 minutes of denaturalizing at 94 °C, 2 minutes of annealing at 55 °C, and 1 minute of extension at 72 °C, ending in a 10 minute final extension period at 72 °C. PCR products amplified were screened and evaluated in an agarose gel stained with ethidium bromide, using a molecular weight marker of 100 pb. PCR products were later sent for sequencing at Macrogen (http://www.macrogen.com).

DNA sequences were edited with Sequencher 5.3 <http://www.genecodes.com>. Alignments were performed with Muscle (^{Edgar 2004}) with the default parameters as available at the CIPRES Science Gateway (^{Miller et al. 2010}). Afterward, the resulting alignment was visually refined in BioEdit v. 5.0.6 (^{Hall 2001}). Average sequence length was 746 bp. To select the model of nucleotidic substitution that best fit the data, jModelTest2 (^{Darriba et al. 2012}, ^{Guindon & Gascuel 2003}) was employed. The model selected was GTR + G.

Nucleotide sequence data of the ITS region used for phylogenetic inferences were analyzed under three different paradigms: Bayesian Inference, Maximum Likelihood, and Maximum Parsimony. The first two were implemented through the CIPRES Science Gateway (^{Miller et al. 2010}). Conditions for each analysis were as described below:

The primary objective of the three analyses was the qualitative assessment of the degree of congruence between the topologies yielded by them. Consensus trees were visualized in Figure 3 and edited with Paint of Microsoft Windows version 1,511 to include clade support values 80 % or higher of posterior probabilities, parametric bootstrap, and Jackknife.

Figure 2 Assignment test of 88 wild individuals of Vanilla spp. collected in the Mexican Yucatan Peninsula and 11 standard samples of Vanilla, using ISSR markers and the Structure program. A) K=5 based in the Evanno’s method. B) K = 7 (second K optimal). CY: Center of Yucatan; CQ: Center of Quintana Roo; SQ: Southern Quintana Roo; SC: South Campeche; and SET: Southeastern Tabasco. 

Figure 3 Neighbor-Joining dendrogram based on Jaccard similarity index (^{Jaccard 1908}) of 88 wild individuals of the Mexican Yucatan Peninsula and 11 standard samples of Vanilla, using ISSR markers. The color in the branches is according to with the Figure 2A. 

Results

This section will present the results of the ISSR analyses first and then the phylogenetic ITS analysis, along with a description of their congruence or lack thereof.

Genetic relationships based on cluster analysis of the ISSR. 84 loci amplified with the three ISSR primers. Evanno’s method estimated a K = 5, which correlate with the following color-coded aggregations (Figure 2A):

According to ^{Meirmans (2015)}, considering different values of K may reflect different genetic and demographic processes and thus ensure a better biological interpretation of the data. In this study, we found that the second-highest K value was 7 (Figure 2B):

The Neighbor-Joining analysis aggregated the genotypes analyzed in two (A and B) highly supported (100 % bootstrap) major clusters (Figure 3), within which several subgroups were retrieved; these are color coded according to what was observed in the Structure analysis (Figure 2A).

The PCoA analysis identified four principal clusters (Figure 4). Also, color coded according to with the Figure 2A. These clusters were mostly “pure-colored”; except an assorted-color (yellow/fuchsia) cluster. The YELLOW/FUCHSIA group included the “Mansa” and “Variegata” cultivars of Vanilla planifolia and 44 wild-collected MYP specimens, the MYP1-CY and MYP2-CY among them. The GREEN group comprised the standard samples of V. planifolia var. “Oreja de Burro” and 23 additional wild MYP individuals. The BLUE group contained the standard samples of Vanilla sp. nov. aff. V. phaeantha and seven wild MYP individuals. Lastly, the RED group included V. pompona, V. odorata, and V. insignis with the remaining (14) wild individuals. These groups feature morphological coherence, except for the red group. The principal coordinates 1 and 2 explained 38.69 % and 12.77 % of total variation, respectively.

Figure 4 Principal Coordinates Analysis (PCoA) of 88 wild individuals of Vanilla of the Mexican Yucatan Peninsula and 11 standard samples of Vanilla, using ISSR markers. The color points are according to with the Figure 2A. 

In summary, the three clustering methods identify genetic relationships with relatively congruent patterns where there are at least four genetic groups and Vanilla planifolia isolated from the other species that occur wild in the MYP. It also identifies a genetic group that appears to correlate with morphologically distinctive plants and that will eventually be proposed as a new taxon (Vanilla sp. nov. aff. V. phaeantha).

Phylogenetic inference of Vanilla based on ITS. Including gaps, the matrix analyzed consisted of 1,018 sites with an average of 698 pb per taxon. We show here the Bayesian inference majority (80 %) consensus tree (Figure 5), including the bootstrap values from the ML analysis and jackknife values exceeding 80 from the parsimony analysis (589 informative sites [57.7 %], L: 3775, CI: 35, RI: 54). Basal nodes were fully resolved and strongly supported. However, there was not resolution within the American Vanilla taxa, due to collapse/lack of support for several clades.

Figure 5 Vanilliodeae phylogeny. Tree with the highest posterior probability resulting from the Bayesian phylogenetic inference. ML bootstrap values to the left of the diagonal; Parsimony jackknife values to the right of the diagonal; and posterior probability values above the branches. The phylogram in the lower left corner shows the length of the branches. Trees were obtained from multiple sequence alignments of nucleotides of Vanilla species using ITS. 

In our analyses, the most highly supported topology retrieves Vanilla as monophyletic (1.0:87/88) and sister to Clematepistephium (1.0), being both members of tribe Vanilleae, in congruence with previous analyses (e.g. ^{Cameron 2009}, ^{Bouetard et al. 2010}). The Afro-Caribbean clade, represented in this analysis by Vanilla aphylla, V. shenzhenica, V. imperialis, V. siamensis, V. africana, V. roscheri and V. barbellata, is strongly supported (1.0:97/100). This clade is sister to the remainder of the genus, composed in this study only of American species, hereafter referred to as the American Clade (AC). The Afro-Caribbean clade includes West Indian taxa as well as Paleotropical, mainly from Africa and tropical Asia, some of which are aphyllous.

The AC resolved as a trichotomy in our analyses, including Vanilla bahiana in an isolated position, along with a highly supported (p.p. 1.0) metaphyletic complex (i.e., monophyletic but lacking internal resolution) composed of V. grandiflora, V. pompona (Puebla and clone 8) and several MYP wild, sterile genotypes most likely attributable to V. pompona (1.0: 93/95). This clade is vegetatively characterized by succulent plants with large, massive leaves. It roughly corresponds with the “RED” group identified by the ISSR analysis. It is noteworthy that V. pompona has not been reported from the MYP, and if eventually confirmed by flowering material, it would constitute a first report for the area.

The third clade contains all the remaining species, represented by one or several samples, in the analysis. Within the third clade, two sister groups are readily identifiable. One of these clades is formed only by V. dressleri + V. cribbiana (DC-clade), composed of tropical rainforest, shade-loving species with thinly textured leaves and narrow stems. The DC-clade is sister to a poorly supported, metaphyletic group, consisting of four clades, one lacking support, whereas the other three have posterior probabilities equal or smaller than 0.85. Each of these clades roughly corresponds to additional four species of Vanilla known from the MYP. Similarly, these clades correspond with the color-coded groups recovered by the ISSR analyses. They are described below:

Discussion

There is currently a void in our knowledge of the genus Vanilla. This vacuum ranges from basic biology including aspects of its systematics, phylogenetics, ecology, and physiology through more applied aspects such as the conservation status of its species (^{Gigant et al. 2011}, ^{Herrera-Cabrera et al. 2012}, ^{Azofeifa-Bolaños et al. 2014}). At the most basic level, the identity, distribution, and circumscription of Vanilla taxa constitute the cornerstone to addressing problems related to the successful conservation and sustainable exploitation of the species of the genus.

Recently, it has become apparent that DNA molecular markers are powerful to help in the characterization of Vanilla species. These markers provide a tool from the population to the species-level identification of individual plants (^{Nagy et al. 2012}). This characteristic is particularly useful in this genus where plants are most frequently collected sterile, are intrinsically variable and phenotypically plastic according to growing conditions, are difficult to the get established and flowered in cultivation, and many species are vegetatively very similar or practically identical.

In Vanilla, only 47 of about 110 species recognized in the genus, have been included in any molecular analysis, and most of them only in phylogenetic studies (^{Soto-Arenas 2003}, ^{Minoo et al. 2008}, ^{Soto-Arenas & Cribb 2010}; ^{Soto-Arenas & Dressler 2010}). A few studies have included purported Vanilla species (e.g. V. tahitensis; ^{Besse et al. 2004}) that were later more correctly identified as hybrid taxa (e.g. ^{Lubinsky et al. 2008b}). In other cases, these hybrids had proposed as new species only to be later more correctly recognized as nothotaxa (e.g. V. hirsuta; ^{Soto-Arenas & Dressler 2010}).

Molecular characterization for Vanilla species is particularly relevant for V. planifolia, the most commonly cultivated taxon of the genus and upon which most of the world’s commercial production of vanilla based (^{Bory et al. 2008a}; ^{Lubinsky et al. 2008a}). Furthermore, this characterization is very important because natural populations of this species have become severely threatened due to illegal extraction of wild individuals for introduction into plantations, both historically and recently (^{Soto-Arenas 2006}). Moreover, destruction of natural habitats has recently become an even larger threat to the continued wild survival of V. planifolia. This study is the first to employ molecular markers to characterize a relatively large number of wild V. planifolia individuals, along with its sympatric congeners in an area of its natural distribution, the MYP.

Clustering pattern of Vanilla species in the MYP. The several approximations used in this study with ISSR markers retrieved largely congruent clustering patterns. Vanilla planifolia and Vanilla sp. nov. aff. V. phaeantha were clearly identified with the ISSR markers. The standard samples of V. pompona, V. odorata, and V. insignis were closely related to each other, probably because only a few individuals per species were tested. Three individuals of V. odorata identified by vegetative morphological features (mainly the narrow, falciform leaves), upon analyses, demonstrated to have a close genetic relationship with the group constituted by V. pompona - V. insignis - V. odorata. These individuals were included in the clade of V. odorata in the ITS phylogeny. The small number of wild-collected individuals in the MYP referable to these species suggest they are extremely rare in the area.

It is interesting to notice that the second highest delta K value obtained revealed a K = 7. These seven genetically different groups of individuals are ascribable to (and contain standard samples of) species previously known from the MYP (V. planifolia, V. insignis, V. odorata, and Vanilla sp. nov. aff. V. phaeantha). In the fifth group, a few wild MYP individuals clustered together with the V. pompona standard sample, and if morphologically confirmed, it would constitute the first record for this species that is mainly known from the western drainage of Megamexico. Vanilla planifolia var. “Oreja de Burro” was retrieved in the sixth group. And finally the seventh ancestral genotypic contribution (20–35 %, “green”, as a yet unidentified wild genotype) is present in three individuals (MYP82, MYP83, and MYP84) from southern Campeche. These individuals include in their genetic makeup about 50 % V. odorata and a smaller genetic contribution from V. insignis. However, in the PCoA, these individuals are not genetically close to V. odorata even though featuring an ancestry coefficient exceeding 50 %. Similarly, this K = 7 brings together two wild individuals to V. insignis (MYP86-SC y MYP87-SC), which are placed by the PCoA closest to V. pompona. Interestingly, although there is no clear evidence that V. pompona occurs natively in the MYP (^{Carnevali et al., 2001}, ^{Soto-Arenas 2009}), seven MYP wild individuals were identified as this species by the K = 7 (Figure 2B). We identified these sterile individuals as V. planifolia relying upon their vegetative structures. These two species are vegetatively similar, especially when juvenile, but eventually, V. pompona develops larger, heavier, much thicker leaves than those commonly found in V. planifolia (Figure 1-A and 1-E). Floral features of course, readily diagnose both species.

Regarding Vanuilla planifolia, the finding of three subgroups within this species is probably attributable to the inclusion of a larger number of samples with a similar vegetative morphology, and therefore a greater probability of identifying an increased number of genotypes referable to the species and thus of detecting genetic or phylogenetic structure within the species. It is known that both self-compatible and self-incompatible individuals occur in V. planifolia. The last type, includes the entity commonly referred to as “Oreja de Burro” (^{Soto-Arenas 2009}). Although the literature describes significant morphological features for the “Mansa” and the “Oreja de Burro” entities (^{Castillo & Engleman 1993}), we were unable to identify wild MYP morphotypes unambiguously referable to them. However, the standard sample “Oreja de Burro” morphotype was retrieved widely apart from the standard samples of the “Mansa” and “Variegata” morphotypes in the Structure and the PCoA analyses, thus suggesting phylogenetic structure within V. planifolia.

A noteworthy finding from both the ISSR and the phylogenetic analyses was the precise identification of a group of individuals morphologically referred to Vanilla sp. nov. aff. V. phaeantha. This entity, albeit closely related to the West Indian V. phaeantha and initially identified as such (^{Soto-Arenas & Dressler 2010}) it is easily differentiated by the presence of several prominent, deep yellow rows of teeth emerging from the labellum throat and reaching the labellum margin, which is also slightly fimbriated (Figure 1-D). This new taxon is in the process of being formally proposed.

Although the ISSR markers displayed high efficiency for the discrimination of individuals as belonging to clades or genetic groups associated with the standard samples of species involved in the study, some individuals were difficult to assign. A preliminary classification based on stem and leaf macro-morphology indicated that some wild individuals possibly belong to V. insignis but the genetic analysis indicates they are best referred (or genetically closer) to Vanilla sp. nov. aff. V. phaeantha, and secondly a cluster of eight individuals similar to V. planifolia but genetically related to the group comprised by V. pompona-V. odorata-V. insignis. These examples reveal how the comparison of individuals at different stages of development can be misleading in this genus, where plants are phenotypically plastic according to age and exposition, as previously pointed out by ^{Putz &Holbrook (1986)} and ^{Ray (1990)}.

Phylogenetic relationships of the Vanilla species in the MYP. The most probable cladograms obtained in our phylogenetic inferences retrieved clades whose composition was entirely coherent with patterns of aggregation identified in the ISSR analysis. An example of this is the recovery of clades consisting of Vanilla sp. nov. aff. V. phaeantha and V. planifolia (Figs. 4 and 5). This finding reinforces the notion that the ISSR markers are robust as species-level discriminators of taxa, in this particular case in Vanilla.

Only a fraction of the species of the genus have been included in phylogenetic inferences, and there has not been an exhaustive phylogenetic analysis of Vanilla to date (^{Poeaim et al. 2011}), the closest being the one by ^{Soto-Arenas & Dressler (2010)}, which included only 15 species-level taxa. Furthermore, novelties within the genus are still being reported (e.g. V. paludosa Pansarin JM, Aguiar & A. C. Ferreira (^{Pansarin et al. 2012}), V. rivasii Molineros, R. T. González, Flanagan & J. T. Otero (^{Molineros-Hurtado et al. 2014}), V. labellopapillata A.K. Koch, Fraga, J.U. Santos & Ilk.-Borg. (^{Koch et al. 2013}).

The resolution of the clades with the different methods evaluated agree with that reported by ^{Cameron & Molina (2006)}, ^{Pansarin et al. (2008)}, ^{Cameron (2009)}, and ^{Soto-Arenas & Dressler (2010)}. It is important to note that Vanilla hirsuta is synonymous with V. tahitensis J.W. Moore, which has a recent hybrid origin (V. planifolia x V. odorata; ^{Soto-Arenas & Dressler, 2010}). Therefore, its proximity to the individual genetic relationship of the standard sample of V. odorata and wild individuals identified as belonging to this species, it is not surprising. Moreover, V. bahiana described in Brazil, it appears to be very similar both morphologically and genetically with V. phaeantha (^{Soto-Arenas 2009}). However, in each of the methods evaluated in this study, this species showed a closer relationship with V. pompona, which strengthens the hypothesis that Vanilla sp. nov. aff. V. phaeantha is a different species to V. phaeantha.

Except for V. planifolia “Oreja de Burro” (in the parsimony analysis), all the specimens referred by us to V. planifolia on morphological grounds constituted a monophyletic assemblage. Flowers of this entity have yet to be analyzed by us to confidently assess its relationships. Other individuals remained unidentified at this stage and placed at intermediate positions. The phylogenetic analysis retrieved a clade consisting mainly of specimens of Vanilla planifolia. However, nested within it there were three individuals assigned to the V. pompona-V. odorata-V. insignis group (red group in K = 5) by the STRUCTURE analysis of the ISSR data. This same analysis (but with K = 7) grouped two of these individuals with V. odorata (red group). The remaining genotype is possibly a complex hybrid containing in its ancestry V. odorata (50% aqua of its genetic makeup), V. sp. (35 % green), and V. insignis (15 % yellow), where the unknown taxon is likely to be a wild genotype of V. planifolia. From this perspective, individuals that were retrieved in intermediate or conflicting positions could be interpreted either as belonging to additional Vanilla species. Yet to be formally identified (with floral material) in our area, or interspecific hybrids.

Natural hybridization has long been known to be frequent in several groups of the Orchidaceae (e.g. ^{Adams & Anderson 1958}) including Vanilla (^{Lubinsky et al. 2006}). Examples include introgression of V. claviculata (W. Wright) Sw. with V. barbellata Rchb.f in Puerto Rico (^{Nielsen & Siegismund 1999}, ^{Nielsen 2000}). Also, AFLP and SSR markers have suggested the possibility of interspecific hybrids between Neotropical species of Vanilla such as V. bahiana, V. planifolia, or V. pompona (^{Bory 2007}, ^{Bory et al. 2008b}). So, interspecific hybridization may be common whenever closely related taxa occur in sympatry (^{Bory et al. 2008}). Nevertheless, not all morphological variation (or polymorphisms) within Vanilla species is to be attributed to interspecific hybridization. As in other species, this variation could be due to genetic drift, disruptive selection, or even directional selection resulting from an evolutionary response to local conditions (^{Ackerman et al. 2011}). Also, there is a component of this variation associated with phenotypic plasticity arising from differences in life stages and different growth conditions (^{van Kleunen & Fischer 2005}).

Implications for the conservation and sustainable use of Vanilla. Wild relatives of crops constitute an important source of genes associated with the resistance to pests and diseases or with higher yields under a variety of conditions (^{Hajjar & Hodgkin 2007}, ^{Hunter & Heywood 2011}, ^{Fisher, 2012}). They are also relevant as a source of nutrients, medicine and other traditional uses for communities and for the maintenance ecosystemic services (^{Hoyt & Brown 1992}). However, before these wild relatives can be utilized in a sustainable manner, it is first necessary to determine which species belong.

In concordance with ^{Verma et al. (2009)}, our study showed that the ISSR can be used to identify Vanilla species present in the MYP. We found higher levels of genetic variation in wild samples than commonly found in vanilla plantations and that might be a result of sexual reproduction and generation from seed, as indicated by the presence of pods in some of the wild populations studied. This point is an indicative of the relevance of these reservoirs of potentially important alleles for the improvement of V. planifolia as a crop, and as such, they should be protected.

Thus, the MYP, where the species is still plentiful (yet occurring as small, widely dispersed populations) should be considered of utmost importance for the conservation of V. planifolia as a genetically healthy wild taxon, and as a source of alleles for its improvement as a crop.