A new vanilla species from Costa Rica closely related to V. planifolia (Orchidaceae)

. We describe a new vanilla species growing in sympatry with Vanilla planifolia Jacks. ex Andrews (Orchidaceae) in the province of Limón, Caribbean coast of Costa Rica. The morphology of the reproductive and vegetative organs observed on vines cultivated under shade-house, the nuclear (Internal Transcribed Spacer) and plastid ( mat K) nucleotide sequences, as well as the contents of aromatic compounds measured in ripe fruits, show that this species is close to but distinct from V. planifolia. The name V. sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp. nov. is proposed for this new Vanilla species endemic in Costa Rica. It is especially distinguished from V. planifolia by a reduction of about 30% of the size of the fruits and flowers, by a divergence of ITS sequences for at least two species-conserved nucleotides compared to seven other species of the V. planifolia group, and by the presence of anisic compounds and low content of phenolic compounds (including vanillin) in the fruits. These results confirmed the extension of the area of distribution of V. planifolia southward to Costa Rica, where a recent speciation process occurred. Because of its particular agronomic and aromatic properties, V. sotoarenasii sp. nov. could represent a valuable biological resource for the vanilla industry. 2017. A new vanilla species from Costa Rica closely related to V. planifolia (Orchidaceae). European Journal of Taxonomy 284: 1–26.


Introduction
The Vanilla genus, Vanilla Plum.ex Miller (Miller 1754), belongs to the subfamily Vanilloideae, tribe Vanillinae (Cameron 2010).It is an ancient group of tropical orchids that originated in America about 70 million years ago and differentiated in America, Africa and Asia (Ramírez et al. 2007;Bouétard et al. 2010).The classification of the species of Vanilla was recently reviewed by Soto Arenas & Cribb (2010), who divided the genus into two subgenera, Vanilla and Xanata Soto Arenas & P.J.Cribb, and further split the subgenus Xanata into two sections Xanata and Tethya Soto Arenas & P.J.Cribb.Currently, about 110 Vanilla species are reported and clustered into 20 groups (Portères 1954;Soto Arenas & Cribb 2010;Cameron 2011;Pansarin et al. 2012).The V. planifolia group is the most important by the number of species ( 16) and, economically, because it contains the commercial species V. planifolia Jacks.ex Andrews (Andrews 1808) and V. × tahitensis J.W.Moore (Moore 1933), the aromatic fruits of which provide the vanilla flavor used by the food and perfumes industries.The taxonomy of the species from Mexico and Central America was reviewed by Soto Arenas & Dressler (2010) on the basis of their morphology and Internal Transcribed Spacer (ITS) DNA sequences.Recently, two new species have been described in the region, V. esquipulensis Archila & Chiron (Archila & Chiron 2012) in Guatemala and V. rivasii Molineros, R.T.González, Flanagan & J.T.Otero (Molineros-Hurtado et al. 2014) in Chocó, northern Colombia, plus two more in French Guiana, V. inornata Sambin & Chiron (Sambin & Chiron 2015) and V. aspericaulis Sambin & Chiron (Sambin & Chiron 2015).1827), V. planifolia, V. pompona Schiede (Schiede 1829), V. sarapiquensis Soto Arenas (Soto Arenas & Dressler 2010) and V. trigonocarpa Hoehne (Hoehne 1944).
During surveys carried out in Costa Rica since 2012 (Azofeifa-Bolaños et al. 2014) abundant populations of vanilla which did not fit any species previously described in the country were observed in Limón Province (Caribbean coast of Costa Rica).This vanilla, called "vanilla Limón" (VanL), had strong morphological similarities with a vanilla accession collected in Cahuita (Limón, Costa Rica) in 1993, referenced Pignal 396b (P00075132) in the herbarium of the Muséum national d'Histoire naturelle in Paris, P (herbarium acronym following Thiers continuously updated).Living specimens of this accession were preserved in the Emmanuel Liais park (Cherbourg-en-Cotentin, France) under accession number CH554, and subsequently in the Biological Resources Centre (BRC) Vatel (Saint Pierre, La Réunion) under accession number CR0068, and were tentatively classified as V. aff.planifolia.
Molecular approaches have enhanced plant taxonomy by allowing reliable genealogy-based classifications (Besse 2014).In the last decade, nuclear, mitochondrial and chloroplastic DNA sequencing has been used to study plant diversity and resolve taxonomic positions in all plant families, including the puzzling group of vanilloid orchids (Cameron 2009(Cameron , 2010;;Soto Arenas & Dressler 2010).
The aim of this study was thus to investigate the taxonomic status of VanL by using complementary approaches involving morphology of reproductive and vegetative organs, molecular barcoding and the accumulation of secondary metabolites in fruits.

Plant and DNA samples
The plant samples used in this study were obtained from field surveys carried out in Costa Rica, Mexico and Guatemala, supplemented with lyophilized plant materials preserved in the Mexican National Collection (Reyes-López et al. 2014), the BRC Vatel (Roux-Cuvelier & Grisoni 2010), and DNA from the herbarium of the Instituto Chinoin, AMO (Thiers continuously updated).
The survey in Costa Rica was conducted between November 2012 and March 2016 in the entire country.Prospections were targeted on areas where vanilla were recorded based on herbarium specimen information and consulting orchid experts, local guides, and earlier project experiences, among others.The material was collected according to permit number 061-2013 issued on 12 Jun.2013 by National System of Conservation Areas (SINAC) of the Ministry of Environment and Energy (MINAE).
All plant and DNA samples analyzed in this study are listed in Table 1.

Morphological traits
To compare the morphology of vegetative and reproductive organs of VanL to those of V. planifolia, the following traits were measured on plants cultivated in controlled conditions in BRC Vatel in La Réunion: length and greater width of the third and fourth leaves, diameter and length of the third inter-node, length and width of the tube, length of the ovary, length and width of the dorsal sepal.Measurements were carried out on two to six plants per accession grown under shade-house in November, which is the optimum vegetative growth and flowering period in La Réunion.From 14 to 30 fresh organs were measured per accession according to the availability of material.Length and weight of fruits were also measured at maturity (in July).We used the experimental tool "collaboratoire" of the national French infrastructure e-ReColNat (ANR-11-INBS-0004) for specimen comparisons.

DNA extraction and sequencing
Total DNA was extracted from lyophilized leaf samples using the Qiagen DNA plant minikit (Dusseldorf, Germany) according to manufacturer protocols.Quantity and quality of the DNA were estimated using a Nanodrop spectrophotometer (Wilmington, USA), and DNA extracts were adjusted to 20 ng µl -1 for polymerase chain reaction (PCR) amplification.
The Internal Transcribed Spacer (ITS) of nuclear ribosomal DNA and part of the plastid maturase K (matK) gene were chosen for molecular characterization because they are among the most discriminant loci for orchids (Cameron 2009;Hollingsworth et al. 2009;Xu et al. 2015) and have distinct inheritances, biparental for ITS and maternal for matK.The DNA samples were PCR amplified using GoTaq kit (Promega, USA) with the two primer pairs AB101/AB102 (Sun et al. 1994) for the ITS sequence, and matK-743F (5'-CTTCTGGAGTCTTTCTTGAGC-3') / matK-1520R (5'-CGGATAATGTCCAAATACCAAATA-3') for matK.The 25 µl PCR reaction mixture contained: PCR reaction buffer, 50 nmol MgCl 2 , 1 U Taq polymerase (GoTaq, Promega, USA), 5 nmol dNTPs, 10 nmol of each forward and reverse primer and 40 ng of genomic DNA.Amplification reactions were performed using a 96-well GeneAmp PCR System 9700 thermocycler (Applied Biosystems, USA).The annealing temperatures for primers were 60°C for ITS and 56°C for matK.Samples from Costa Rica were also amplified using the Multiplex PCR kit (Qiagen) with the same primers, and the PCR Mix and cycling conditions defined by the provider.Amplicons were sequenced in both directions as part of the Bibliothèque du Vivant project (Paris, France), and by Genwiz (Takeley, UK) and Genoscreen (Lille, France).Nucleotide sequences were aligned using the ClustalW package included in Bioedit software (Hall 1999) and cleaned manually to generate consensus sequences.The data set was complemented with reference sequences obtained previously (Soto Arenas & Dressler 2010).

Identification and quantification of aromatic precursors in mature fruits
Hand-pollinated fruits of VanL (CR0068) and three V. planifolia accessions (CR0196, CR0040 and CR0038), cultivated under shade-house in La Réunion, were harvested at 8 months after pollination, in 2013 and 2014, then freeze-dried to minimize possible enzyme degradation.The fruits were ground to a fine powder with a mortar and pestle and stored at minus 20°C until extraction.
Extraction of volatile compounds was performed using a protocol adapted from Palama et al. (2009) and Pérez Silva et al. (2011).Fifty milligrams of the ground material was suspended in 10 ml of phosphate buffer (0.1 M; pH 5).The mixture was ultra-sonicated (frequency 35 kHz; AXTOR Model CD-4820) for 10 min at ambient temperature.The mixture was rapidly heated to 80°C for 10 min to inactivate endogenous enzymes and, after cooling at 25°C, was centrifuged at 5000 rpm for 3 min and then filtrated on a Whatman no.1 paper (Sigma-Aldrich, USA).The filtrate was complemented with 0.4 ml of a glycosidase rich enzyme preparation (AR2000 ® , Sigma-Aldrich, Mexico; 70 mg•ml -1 in phosphate buffer mentioned above) and adjusted to a volume of 10 ml with the pH 5 buffer solution.The extract was vortexed and incubated for 4 h at 40°C for enzymatic hydrolysis of the glycosylated precursors.One milliliter of hydrolyzed extract was filtered at 0.45 µm prior to HPLC analysis.
For each sample, extraction and HPLC were run in duplicate.

Data analysis
All statistical analyses were performed with the R statistical software (R Core Team 2013).Multiple comparisons of means of morphological and aromatic traits were performed using the Multcomp package in R (Hothorn et al. 2008).
Phylogenetic trees were inferred using the maximum likelihood method implemented in MEGA7 (Kumar et al. 2016) after determining the best substitution model using the online version of JModelTest2 (Guindon et al. 2003;Darriba et al. 2012).Branch robustness was assessed by bootstrapping 1000 datasets and branches with less than 65% bootstrap support were collapsed.

Field observations
A total of 131 vanilla plants were collected in the survey carried out throughout Costa Rica in the time period 2013−2016.Among them, 17 were tentatively classified as V. planifolia (12) or VanL (5) according to the morphology of the leaves, stem or flower.The geographic origin and code numbers of these plants are provided in Fig. 1 and Table 2.
Morphologically, the VanL plants showed most similarity with V. planifolia and V. ×tahitensis, a species having a hybrid origin involving V. planifolia and V. odorata (Lubinsky et al. 2008).However, compared to these two species VanL plants had much smaller leaves, flowers and fruits (Fig. 2) and a distinct shape of leaves; elliptic-obovate in the case of VanL, oblong for V. planifolia and lanceolate for V. × tahitensis.
On  (Hoehne 1910) and V. schwackeana Hoehne (Hoehne 1944)) by color of sepal, color of petal, size and shape of tube, label ornamentation, and shape, thickness and texture of leaves.
Five out of the six VanL samples, CR0068, CR2180, CR2553, CR2554 and UNA-049, were collected in natural forests at three localities of the Caribbean province of Limón (Costa Rica), indicated in Table 2.
The sixth VanL accession (CR2557) was cultivated in a farm with no information on its origin.During the survey, ten V. planifolia plants were found growing wild at two localities of the Limón province.
Five of these accessions were collected in Barra Parismina, a human settlement established about 50 years ago that is still not connected by road to the rest of the country and where, according to one of the first settlers, vanilla was present prior to human occupation and no vanilla was ever introduced by man (Gabriel Taylor, Parismina, Costa Rica, pers.comm.).They are therefore considered to be occurring naturally in this area and not having escaped from local cultivation.They were morphologically indistinguishable from cultivated V. planifolia collected at farms in San Isidro (CR2559) and Mora (UNA-0230) or the three V. planifolia from Talamanca (UNA-0022, UNA-0059, CR2555).
To date, the wild populations of VanL and V. planifolia were both sampled in tropical humid forests of the southern coastal lowlands of Limón Province (Fig. 3).This biotope is characterized by plant species adapted to sandy soils such as Terminalia catappa L. The VanL plant from Guápiles (CR2557) was sampled in a vanilla plot where vanilla was associated with pepper and cinnamon.In Gandoca-Manzanillo, Parismina and Cahuita the flowering happened from October to January, as deduced from the many racemes found in mid-November bearing flowers and young fruits.Mature vanilla beans were harvested in December 2015 from the VanL population of Guápiles (CR2557), which indicated a flowering time probably ending in March.

Phylogenetic analysis
To elucidate the taxonomic position of VanL plants, the nuclear ITS DNA and the matK plastid gene were sequenced for 125 and 55 vanilla accessions, respectively.The sample set included VanL specimens, accession representatives of the diversity of the V. planifolia group, as well as outgroup species (Table 1).
The phylogenetic tree inferred from the 506 positions of the 125 aligned ITS sequences (465 to 500 nt) revealed a clade separated with 93% bootstrap support, containing the five VanL accessions (CR2180, CR2553, CR2554, CR2557 and UNA049) along with CR0068 from BRC Vatel and two accessions from the AMO, collected in Costa Rica and Honduras respectively, identified as V. planifolia cf.plan-02 and plan-03 (Fig. 4).This clade was within the V. planifolia accessions and close to the V. bahiana / V. phaeantha clade.It was more distantly related to the V. insignis and the V. helleri -V.odorata -V.× tahitensis clades.
The eight partial ITS sequences in the VanL clade differed from all the accessions of the V. planifolia clade by two conserved nucleotides at positions 390 and 479 of the alignment (Table 3).Similarly, V. bahiana differed from V. phaeantha by only two conserved nucleotides (nt 172 and 420).In contrast, the V. planifolia clade differed from the V. bahiana and V. phaeantha group by 13 conserved nucleotides, and from V. insignis by 15 conserved nucleotides.
The phylogenetic tree inferred from the 741 positions of the 55 aligned matK sequences (699 to 734 nt) separated V. planifolia with high bootstrap support from its most closely related species, including V. odorata, V. insignis, V. bahiana and V. phaeantha (Fig. 5).The eight VanL accessions sequenced fell within the V. planifolia clade, confirming their very close relationship with this species.
Phylogenetic trees inferred using the Neighbor Joining, Parsimony and Bayesian methods were congruent with the ML trees (data not shown).

Morphology of vegetative and reproductive organs of VanL
The morphology of the leaf, stem, and flower of VanL is very similar to that of V. planifolia (Fig. 2).However, the former differs from the latter by a highly significant reduction in size of the vegetative and reproductive organs (Table 4).In average, leaf and stem dimensions are 29 to 41% smaller in VanL compared to V. planifolia.
The size reduction in VanL is less important in the flower, with tube length only 7% and sepal length 12% lesser than for V. planifolia.However, in VanL, the dorsal sepal is 29% narrower and the ovary 19% shorter than in V. planifolia.The blooming period of VanL cultivated in La Réunion was approximately two weeks earlier than that of V. planifolia accessions, but they overlapped from October to November.
Mature fruits of VanL were 45% shorter and 55% lighter than those of V. planifolia (Fig. 2; Table 4).They were also indehiscent and more cylindrical, contrary to most V. planifolia fruits, which dehisce at maturity and have a more triangular section.

Aromatic content of mature fruits
Strongly contrasted aromatic profiles were observed between VanL and V. planifolia by comparing the contents of nine volatile precursors detected in mature fruits by HPLC (Table 5).The fruits of VanL had a much lower content of vanillyl compounds, particularly vanillin which did not exceed 0.26% of dry matter, while it was over 2.28% in V. planifolia beans.Conversely, VanL fruits had significant amounts of anisyl compounds which were not detected by HPLC in V. planifolia fruits, and p-hydroxybenzyl (PHB) alcohol was present at a much higher titer in VanL compared to V. planifolia.The contents of the two other p-hydroxyl compounds were not significantly different between VanL and V. planifolia.
Table 3. Inter-clade polymorphisms of within-clade-conserved nucleotides along the partial ITS sequence of Vanilla planifolia Jacks.ex Andrews, V. sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp.nov.and their most closely related species.The triangular matrix on the right indicates the number of polymorphic sites between species.Nucleotides diverging from the V. planifolia sequence are in bold.Y = C or T, K = G or T.   As a whole, the VanL fruit had a significantly lower (about 30%) level of total aromatic content (all nine molecules) and a very distinct aromatic profile compared to V. planifolia.Indeed, the volatiles of V. planifolia fruits were strongly dominated by vanillyl compounds (83.1% of all volatiles, particularly vanillin which represented 68%).By contrast, the VanL fruits contained a more equilibrated profile with a slight dominance of vanillyl compounds (41.9% of all volatiles, in which vanillyl-alcohol represented 31%), p-hydroxybenzyl and anisyl compounds (35.9% and 22.2%, respectively).

Phenology
The flowering period occurs from October to March in Costa Rica.

IUCN status -Vulnerable
The populations of V. sotoarenasii sp.nov.observed in the province of Limón showed strong vegetative development of vines and natural seed set was frequently observed.However, a small number of populations have so far been observed and only over a very limited area of Limón Province (10 populations over a coastal strip of about 50 km²), and its habitat is periodically submerged by the ocean.This makes us inclined to tentatively classify V. sotoarenasii sp.nov.as vulnerable D2 (IUCN 2012) until further data on distribution and population dynamics have been obtained.

Discussion
Using complementary approaches involving morphology of the reproductive and vegetative systems, molecular barcoding and the accumulation of secondary metabolites in fruits, we highlighted specific traits for the vanilla populations sampled on the Caribbean coast of Costa Rica which revealed a new taxon, V. sotoarenasii sp.nov.The morphology of the flowers and leaves clearly assigned V. sotoarenasii sp.nov.within the V. planifolia group (Soto Arenas & Dressler 2010).Based on flower morphology, V. sotoarenasii sp.nov.was more similar to V. planifolia and V. × tahitensis than to any other vanilla species of this group.However, V. sotoarenasii sp.nov.differs from the other two species by several characteristics (Table 6).
Firstly, our data and previous data by Costantin & Bois (1915) and Portères (1953) showed that V. sotoarenasii sp.nov.has a significantly smaller size for all organs measured (stem, leaf, flower and fruit) compared to V. planifolia and V. × tahitensis, and had a distinct shape of the leaves: elliptic to obovate in the case of V. sotoarenasii sp.nov., elliptic to oblong for V. planifolia and narrowly oblong to lanceolate for V. × tahitensis (Table 6).In addition, the flowers of V. sotoarenasii sp.nov.are more whitish, with a narrower label showing marked papillae, compared to those of V. planifolia, which are more greenish, with a wider label and smooth papillae.
On the basis of HPLC quantification of hydrolyzed volatile compounds in mature fruits, the aromatic precursors of V. sotoarenasii sp.nov.are very distinct from those of V. planifolia.In particular, fruits of V. sotoarenasii sp.nov.were characterized by less predominant vanillin content and the presence of anisyl compounds.They should develop, after over-maturation or curing, flavors extremely different from those of V. planifolia and more similar to those of V. pompona or V. × tahitensis (Pérez- Silva et al. 2006;Brunschwig et al. 2009;Maruenda et al. 2013).
On the other hand, molecular analysis of nuclear DNA sequences (ITS) unambiguously separated the V. sotoarenasii sp.nov. group of plants from the V. planifolia group including 68 accessions originating from seven countries (Comoros, Costa Rica, Guatemala, Madagascar, Mayotte, Mexico, and La Réunion).This result is corroborated by amplified fragment length polymorphism (AFLP) analyses by Bory et al. (2008) that clearly separated V. sotoarenasii sp.nov.CR0068 from 303 V. planifolia genotypes of diverse origins with an average distance comparable to the distance between V. planifolia and V. × tahitensis.The facts that i) at the plastid DNA level (partial matK sequence) V. sotoarenasii sp.nov.and V. planifolia share the same clade, and ii) in the ITS phylogeny V. sotoarenasii sp.nov. is in an internal position within the V. planifolia clade (like V. helleri within the V. odorata cluster), suggest the recent radiation of V. sotoarenasii sp.nov.from V. planifolia populations.
So far, V. sotoarenasii sp.nov.has only been observed in the Limón Province of Costa Rica where it is sympatric with V. planifolia, which has been reported as native to Costa Rica (Soto Arenas & Dressler 2010).However, recent introductions in the country of V. planifolia cultivars and hybrids have also been documented (Soto Arenas & Dressler 2010;Belanger & Havkin-Frenkel 2011;Varela Quirós 2011).
Historical, genetic and phylogenetic data are insufficient to decide whether V. sotoarenasii sp.nov.derived from natural or introduced populations of V. planifolia, nor how and when the radiation occurred.
Finally, given its original aromatic content combining vanilyl and anisic notes, its high level of resistance (at least for CR0068) to the Fusarium root rot of vanilla (Koyyappurath et al. 2015), and its early and abundant flowering in culture, V. sotoarenasii sp.nov.could be a promising genitor for breeding programs aiming to produce new vanilla varieties for the agroindustry.

Fig. 4 .
Fig. 4. Phylogenetic tree derived from the partial ITS sequences (506 positions) of the 125 accessions listed in Table 1, showing the differentiation of the Vanilla sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp.nov.clade from V. planifolia Jacks.ex Andrews and all other related species.The tree was inferred using the Maximum Likelihood method based on the Tamura-Nei model with invariant sites and Gamma distribution of evolutionary rates.The figures indicate the percentage of bootstrap support.Branches with less than 65% support were collapsed.Countries of origin: Bra = Brazil; Com = Comoros: CR = Costa Rica; FGu = French Guiana; FPo = French Polynesia; Gua = Guatemala; Hon = Honduras; Mad = Madagascar; May = Mayotte; Mex = Mexico; Run = La Réunion; na = geographic origin not available.Numbers in brackets indicate the number of similar accessions merged in one branch for outgroup species.

Fig. 5 .
Fig. 5. Phylogenetic tree derived from the partial matK sequences (725 positions) of the 55 accessions listed in Table 1, showing the Vanilla sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp.nov.accessions (red dots) within the V. planifolia clade but distinct from other related species.The tree was inferred using the Maximum Likelihood method based on the Hasegawa-Kishino-Yano model and Gamma distribution of evolutionary rates.The figures indicate the percentage of bootstrap support.Countries of origin: Bra = Brazil; CR = Costa Rica; FGu = French Guiana; FPo = French Polynesia; Mad = Madagascar; Mex = Mexico; Run = La Réunion; na = geographic origin not available.Numbers in brackets indicate the number of similar accessions merged in one branch for outgroup species.

Table 1 .
Identification and origin of the accessions of species of Vanilla Plum.ex Miller used in this study.[continued on next 3 pages]

Table 4 .
Means of different measurements on vegetative and reproductive organs of two accessions of Vanilla sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp.nov.(CR0068 and CR2180) and two accessions of V. planifolia Jacks.ex Andrews (CR0040 and CR0196).Measurements are in mm unless otherwise indicated.In each line, means with different letters are significantly different (p < 0.005, Tukey's test).StE = standard error.
11 17 16 15 The above results indicate than VanL and V. planifolia are very closely related but significantly distinct in morphological and biochemical traits, as well as in nuclear nucleotide sequences.We therefore propose to place VanL in a specific taxon distinct from V. planifolia, and to name it Vanilla sotoarenasii M.Pignal, Azofeifa-Bolaños & Grisoni sp.nov.