**2.1 Distribution and phylogeography**

*Vanilla* Plum. ex Miller is an ancient genus in the Orchidaceae family, Vanilloideae subfamily, Vanilleae tribe and Vanillinae sub-tribe (Cameron, 2004, 2005).

*Vanilla* species are distributed throughout the tropics between the 27th north and south parallels, but are absent in Australia. The genus is most diverse in tropical America (52 species), and can also be found in Africa (14 species) and the Indian ocean islands (10 species), South-East Asia and New Guinea (31 species) and Pacific islands (3 species) (Portères, 1954). From floral morphological observations, Portères (1954) suggested a primary diversification centre of the *Vanilla* genus in Indo-Malaysia, followed by dispersion on one hand from Asia to Pacific and then America, and on the other hand from Madagascar to Africa. This hypothesis was rejected following the first phylogenetic studies of the genus (Cameron, 1999, 2000) which suggested a different scenario with an American origin of the genus (160 to 120 Mya) and a transcontinental migration of the *Vanilla* genus before the break-up of Gondwana (Cameron, 2000, 2003, 2005; Cameron et al., 1999). The genetic differentiation between New World and Old World species observed would therefore be a consequence of the further separation of the continents. Our recent molecular phylogeny using chloroplastic *psa*B, *psb*B, *ps*bC, and *rbc*L regions (Bouetard et al., 2010) supported the hypothesis of an American origin of the genus (figure 1). However, the recent discovery of a fossilized orchid pollinaria (20 Mya) (Ramirez et al., 2007) allowed the dating of Vanilloidae sub family at 72 Mya, well after the separation of Gondwana which questions the hypothesis of a vicariate evolution of the *Vanilla* genus (Bouetard et al., 2010).

Transoceanic dispersion appears more credible and would have been implied at least three times in the evolution of the *Vanilla* genus (figure 1). This was demonstrated by dating a *Vanilla* molecular phylogeny, testing these two extreme evolutionary scenarios (vicariate

Biodiversity and Evolution in the *Vanilla* Genus 3

*V. palmarum*

*V. lindmaniana*

*V. mexicana*

*V. leprieuri V. ensifolia V. pompona V. chamissonis V. odorata V. xtahitensis V. planifolia V. bahiana*

**Group β** American species

> **Group γ** Old World + Caribbean species

sect. *Tethya*

**Group α** Membranous species

*Vanilla* subgen. *Xanata*

sect. *Xanata*

*Vanilla* subgen. *Vanilla*

Fig. 1. Schematic representation of the molecular phylogeny of the *Vanilla* genus based on *rbc*L, *psa*B, *psb*B and *psb*C (Bouetard et al., 2010), distinguishing clades α, β and γ. The geographical origin of the species is indicated. Species underlined are from sect. Aphyllae, others are from sect. Foliosae (as per Rolfe's classification). Taxonomic classification as per Soto Arenas & Dressler (2010) is indicated. Flowers of representative species and their voucher number (CR) in the BRC Vatel collection are presented (photographs: M Grisoni). Estimated divergence times (in Mya) derived from Bayesian relaxed clock analyses (uncorrelated exponential relaxed molecular clock model) (Bouetard et al., 2010) are indicated for key nodes: (i) origin of *Vanilla*, (ii) separation between New and Old World *Vanilla* species; (iii) separation between African and Asian species; origin of Aphyllae species (iv) in the South West Indian Ocean area and (v) in the Caribbean-West Indies area.





*V. madagascariensis V. roscheri V. humblotii V. dilloniana*

**Asia Africa**

**America**

**CWI**

**SWIO**

*V. imperialis*

*V. crenulata V. africana*

*V. albida V. aphylla*



*Vanilla*





Upper values correspond to the Gondwanan dispersion scenario and lower values correspond to the transoceanic dispersion scenario. Blue dots on clade nodes indicate transoceanic dispersion whatever the scenario tested. World maps at different geological

In the first thorough taxonomic treatment of the genus published, Portères (Portères, 1954) described 110 species in the *Vanilla* genus. This number was reduced by different authors (Cameron et al., 1999; Soto Arenas, 1999, 2006; Soto Arenas & Dressler, 2010), but some species were not included (Hoehne, 1945) and new species have since been described (Z.J. Liu et al., 2007; Pignal, 1994; Soto Arenas, 2006, 2010; Soto Arenas & Cameron, 2003; Szlachetko & Veyret, 1995). There are to date more than 200 *Vanilla* species described (Bory et al., 2008b; Cameron, 2011b), but numerous synonymies remain and there is therefore an urgent need to thoroughly revise the taxonomic classification of the *Vanilla* species. We recently reviewed (Bory et al., 2010) the complexity of the processes involved in the evolution and diversification

times are provided.

**Late Jurassic 152 Mya**

**Late Cretaceous 94 Mya**

**GONDWANA PACIFIC OCEAN TETHYS OCEAN LAURASIA**

**N. Am.**

**Africa S. Am.**

**Middle Eocene 50.2 Mya**

**N. Am. Eur.**

**N. ATL. OCEAN S. ATL. OCEAN**

**N. Am. N. ATL. OCEAN S. Am. S. ATL. OCEAN Africa Mad. IND. OCEAN India Asia Eur.**

**PACIFIC OCEAN**

**PACIFIC OCEAN**

**PACIFIC OCEAN**

**Middle Miocene 14 Mya**

**Mad. India S. Am.**

**Africa**

**Asia**

**Mad./India**

**TETHYS OCEAN**

**Eurasia**

**IND. OCEAN**

*versus* transoceanic dispersion) (Bouetard et al., 2010) (figure 1). The Gondwanan dispersion scenario used 95 Mya as prior on the NW/OW node (the minimum age assumption for the break-up of Gondwana), whereas the NW/OW transoceanic dispersion scenario used 71 Mya as prior on the Vanilloidae node (a date estimated from fossil orchid pollinaria dating (Ramirez et al., 2007)) (figure 1). This provided evidence for at least three transoceanic dispersion events whatever the original scenario retained for the differentiation of NW versus OW species: from Africa to Asia, from Africa to the South West Indian Ocean Islands, and from Africa back to America (Carribean region) (Bouetard et al., 2010) (figure 1).

#### **2.2 Taxonomy and phylogeny**

Taxonomic classification is based on morphological variations in vegetative and floral characters. Ephemeral flowers and their scarce availability in herbarium specimens associated with the fact that vegetative characters show important intra-specific variations are responsible for the difficulties in providing a clear taxonomic classification in *Vanilla* (Bory et al., 2010).

The first classification (Rolfe, 1896) distinguished two sections in the genus: section Foliosae, and section Aphyllae with leafy or leafless species, respectively. Portères (1954) then divided section Foliosae in three sub-sections: Papillosae, with thick leaves and a labellum with fleshy hairs, Lamellosae with thick leaves and a labellum with scaly lamellae, and Membranacae with thin membranous leaves.

The *Vanilla* genus taxonomy has recently greatly beneficiated from molecular phylogenetics. The sequences used were chloroplastic *rcb*L (Cameron et al., 1999; Soto Arenas & Cameron, 2003), *psa*B (Cameron, 2004), *psb*B and *psb*C (Cameron & Molina, 2006), and the results obtained showed that Rolfe's sections and Portères' sub-sections classically used for taxonomy in *Vanilla* did not have a phylogenetic value. A recent study (Bouetard et al., 2010), based on these four markers combined, revealed three major clades in the genus, called groups α, β, et γ (figure 1). Group α is represented by *V. mexicana* and is ancestral. Separation between group β (composed of New World/American Foliosae species) and group γ (composed of Old World/African and Asian Foliosae and American, Asian and African Aphyllae species) is more recent. This study confirmed an American origin of the genus, and also showed that the sections Foliosae and Aphyllae are not monophyletic (figure 1), a statement that questions the classical taxonomic treatment of the genus proposed by Rolfe (1896) and Portères (1954).

Recently, based on phylogenetic data of 106 species, (Soto Arenas & Cribb, 2010) proposed a new taxonomic classification, differentiating two sub-genera in the *Vanilla* genus. A group contains species previously classified as sub-section Membranaceae: *V. angustipetala*, *V. martinezii*, *V. inodora*, *V. mexicana*, *V. parviflora*, *V. edwalii* and the monospecific genus *Dictyophyllaria dietschiana* now *V. dietschiana* (Bouetard et al., 2010; Cameron, 2010; Pansarin, 2010a2010b; Soto Arenas & Cameron, 2003). It was named genus *Vanilla* sub-genus *Vanilla* as it contains the typus species for the genus (*V. mexicana*). It corresponds to the ancestral phylogenetic group α (figure 1). The remaining *Vanilla* species are included in genus *Vanilla* sub-genus *Xanata*, which is further divided in two sections: section *Xanata* (corresponding to phylogenetic group β) and section *Tethya* (group γ) (figure 1). Within section *Xanata*, an early diverging group is noteworthy (figure 1) containing *V. palmarum*, *V. lindmaniana* and *V. bicolor* (Bouetard et al., 2010; Cameron, 2010; Soto Arenas & Cameron, 2003). This preliminary revised classification is a major step towards a needed complete revision of the genus based on molecular analyses.

*versus* transoceanic dispersion) (Bouetard et al., 2010) (figure 1). The Gondwanan dispersion scenario used 95 Mya as prior on the NW/OW node (the minimum age assumption for the break-up of Gondwana), whereas the NW/OW transoceanic dispersion scenario used 71 Mya as prior on the Vanilloidae node (a date estimated from fossil orchid pollinaria dating (Ramirez et al., 2007)) (figure 1). This provided evidence for at least three transoceanic dispersion events whatever the original scenario retained for the differentiation of NW versus OW species: from Africa to Asia, from Africa to the South West Indian Ocean Islands,

Taxonomic classification is based on morphological variations in vegetative and floral characters. Ephemeral flowers and their scarce availability in herbarium specimens associated with the fact that vegetative characters show important intra-specific variations are responsible for the difficulties in providing a clear taxonomic classification in *Vanilla* (Bory et al., 2010). The first classification (Rolfe, 1896) distinguished two sections in the genus: section Foliosae, and section Aphyllae with leafy or leafless species, respectively. Portères (1954) then divided section Foliosae in three sub-sections: Papillosae, with thick leaves and a labellum with fleshy hairs, Lamellosae with thick leaves and a labellum with scaly lamellae, and

The *Vanilla* genus taxonomy has recently greatly beneficiated from molecular phylogenetics. The sequences used were chloroplastic *rcb*L (Cameron et al., 1999; Soto Arenas & Cameron, 2003), *psa*B (Cameron, 2004), *psb*B and *psb*C (Cameron & Molina, 2006), and the results obtained showed that Rolfe's sections and Portères' sub-sections classically used for taxonomy in *Vanilla* did not have a phylogenetic value. A recent study (Bouetard et al., 2010), based on these four markers combined, revealed three major clades in the genus, called groups α, β, et γ (figure 1). Group α is represented by *V. mexicana* and is ancestral. Separation between group β (composed of New World/American Foliosae species) and group γ (composed of Old World/African and Asian Foliosae and American, Asian and African Aphyllae species) is more recent. This study confirmed an American origin of the genus, and also showed that the sections Foliosae and Aphyllae are not monophyletic (figure 1), a statement that questions the classical taxonomic treatment of the genus

Recently, based on phylogenetic data of 106 species, (Soto Arenas & Cribb, 2010) proposed a new taxonomic classification, differentiating two sub-genera in the *Vanilla* genus. A group contains species previously classified as sub-section Membranaceae: *V. angustipetala*, *V. martinezii*, *V. inodora*, *V. mexicana*, *V. parviflora*, *V. edwalii* and the monospecific genus *Dictyophyllaria dietschiana* now *V. dietschiana* (Bouetard et al., 2010; Cameron, 2010; Pansarin, 2010a2010b; Soto Arenas & Cameron, 2003). It was named genus *Vanilla* sub-genus *Vanilla* as it contains the typus species for the genus (*V. mexicana*). It corresponds to the ancestral phylogenetic group α (figure 1). The remaining *Vanilla* species are included in genus *Vanilla* sub-genus *Xanata*, which is further divided in two sections: section *Xanata* (corresponding to phylogenetic group β) and section *Tethya* (group γ) (figure 1). Within section *Xanata*, an early diverging group is noteworthy (figure 1) containing *V. palmarum*, *V. lindmaniana* and *V. bicolor* (Bouetard et al., 2010; Cameron, 2010; Soto Arenas & Cameron, 2003). This preliminary revised classification is a major step towards a needed complete revision of the

and from Africa back to America (Carribean region) (Bouetard et al., 2010) (figure 1).

**2.2 Taxonomy and phylogeny** 

Membranacae with thin membranous leaves.

proposed by Rolfe (1896) and Portères (1954).

genus based on molecular analyses.

Fig. 1. Schematic representation of the molecular phylogeny of the *Vanilla* genus based on *rbc*L, *psa*B, *psb*B and *psb*C (Bouetard et al., 2010), distinguishing clades α, β and γ. The geographical origin of the species is indicated. Species underlined are from sect. Aphyllae, others are from sect. Foliosae (as per Rolfe's classification). Taxonomic classification as per Soto Arenas & Dressler (2010) is indicated. Flowers of representative species and their voucher number (CR) in the BRC Vatel collection are presented (photographs: M Grisoni). Estimated divergence times (in Mya) derived from Bayesian relaxed clock analyses (uncorrelated exponential relaxed molecular clock model) (Bouetard et al., 2010) are indicated for key nodes: (i) origin of *Vanilla*, (ii) separation between New and Old World *Vanilla* species; (iii) separation between African and Asian species; origin of Aphyllae species (iv) in the South West Indian Ocean area and (v) in the Caribbean-West Indies area. Upper values correspond to the Gondwanan dispersion scenario and lower values correspond to the transoceanic dispersion scenario. Blue dots on clade nodes indicate transoceanic dispersion whatever the scenario tested. World maps at different geological times are provided.

In the first thorough taxonomic treatment of the genus published, Portères (Portères, 1954) described 110 species in the *Vanilla* genus. This number was reduced by different authors (Cameron et al., 1999; Soto Arenas, 1999, 2006; Soto Arenas & Dressler, 2010), but some species were not included (Hoehne, 1945) and new species have since been described (Z.J. Liu et al., 2007; Pignal, 1994; Soto Arenas, 2006, 2010; Soto Arenas & Cameron, 2003; Szlachetko & Veyret, 1995). There are to date more than 200 *Vanilla* species described (Bory et al., 2008b; Cameron, 2011b), but numerous synonymies remain and there is therefore an urgent need to thoroughly revise the taxonomic classification of the *Vanilla* species. We recently reviewed (Bory et al., 2010) the complexity of the processes involved in the evolution and diversification

Biodiversity and Evolution in the *Vanilla* Genus 5

In natural conditions, vanilla plant density can be extremely variable from being very high in certain areas (*V. trigonocarpa* (Soto Arenas & Dressler, 2010), *V. pompona* (Householder et al., 2010)) from very low as reported for wild *V. planifolia* in Mexico with less than one plant found per square kilometre (Soto Arenas, 1999). Some species are known to flower very frequently (*V. chamissonis,* (Macedo Reis, 2000)) to very un-frequently (*V. planifolia, V. hartii*, (Schlüter, 2002; Soto Arenas & Dressler, 2010)). A single flower per inflorescence generally opens in *Vanilla*, except 2-3 in some species (*V. odorata, V. martinezii, V. insignis*) and flowers are ephemeral (one day) except for some rare species such as *V. inodora* (2-3 days) (Soto Arenas & Dressler, 2010) or *V. imperialis* for which the flowers can be fertilized 4-5 days after opening (unpublished data). Seedlings can be found very frequently for species such as *V. bicolor* and *V. palmarum* (Householder et al., 2010) or be extremely rare as in *V. pompona* in Madre de Dios (Householder et al., 2010) or *V. planifolia* in Mexico (Schlüter, 2002). All these natural history traits will have deep effects on the levels of *Vanilla* species biodiversity that can be found in the wild. Particularly, the relative balance between vegetative and sexual reproduction and their relative efficiency will be of major importance in shaping populations genetic diversity. Exploring *Vanilla* species reproductive systems is therefore

*Vanilla* species, like other orchids, are characterized by the presence of a rostellum membrane separating female and male reproductive systems, therefore limiting selfpollination. The diverse floral morphology observed in *Vanilla* species (figure 1) suggests that they have evolved to adapt to different pollinators (Soto Arenas & Cameron, 2003).

A few *Vanilla* species are described as spontaneously self-pollinating (Householder et al., 2010; Soto Arenas & Cameron, 2003; Soto Arenas & Dressler, 2010; Van Dam et al., 2010), as suggested by their abnormally high fruit set (table 1). This is consistent with general data in orchids showing that autogamous species display a much higher fruit set (77%) than cross pollinating species for which the majority show fruit set <20% (Tremblay et al., 2005). Based on high fruit set, these suggested autogamous species are *V. palmarum*, *V. savannarum*, *V. bicolor* (American species of the *V. palmarum* group), *V. guianensis*, *V. martinezii* (American species of the *V. mexicana* group) and *V. griffithii* (an Asian species). Possible self-pollination for *V. inodora* is also reported (Soto Arenas & Dressler, 2010), due to the large fruit set

observed in some populations, although others have a fruit set as low as 2.5%.

Table 1. Suggested self-pollinating *Vanilla* species and recorded natural fruit sets.

*V guianensis* 78% (Householder et al., 2010) *V. palmarum* 76% (Householder et al., 2010) *V bicolor* 71% (Householder et al., 2010) *V. bicolor* 42.5% per raceme (Van Dam et al., 2010) *V. martinezii* 53% in a clone (Soto Arenas & Dressler,

More precise observations are available for some of these species. *V. guianensis* is supposedly self-pollinated at early anthesis, as it was observed that the stigma and the

2010)

Species Natural fruit set (self-pollination) Reference

essential in this context.

**3.1** *Vanilla* **pollination** 

**3.1.1 Self-pollinating species** 

of the *Vanilla* genus and concluded that *Vanilla* must be considered as a TCG, a "Taxonomic Complex Group" (Ennos et al., 2005). Indeed, it exhibits (i) an uniparental reproduction mode (vegetative growth) (Portères, 1954) (ii) interspecific hybridization in sympatric areas (Bory et al., 2010; Bory et al., 2008c; Nielsen, 2000; Nielsen & Siegismund, 1999) and (iii) polyploidy (Bory et al., 2010; Bory et al., 2008a; Lepers-Andrzejewski et al., 2011a; Lepers-Andrzejewski et al., 2011b). These mechanisms have profound effects on the organization of the biological diversity and have been described as responsible for the difficulty to define discrete, stable and coherent taxa in such TCGs (Ennos et al., 2005). *Vanilla* is a typical example of a genus for which the barcoding protocols (*mat*K and *rbc*L) as proposed by the CBOL (M.L. Hollingsworth et al., 2009; P.M. Hollingsworth & CBOL Plant Working Group, 2009 ; Ratnasingham & Hebert, 2007), will therefore not be sufficient to revise the species taxonomy. The lack of genetic incompatibility between most *Vanilla* species (Bory et al., 2010) and the proven occurrence of inter-specific hybridizations in the genus (Bory et al., 2010; Bory et al., 2008c; Nielsen, 2000; Nielsen & Siegismund, 1999) will necessitate the obligate survey of nuclear regions in addition to cpDNA markers to resolve introgression patterns and correctly identify *Vanilla* species (Rubinoff, 2006). As an example, the species *V. ×tahitensis* was recently shown to be a *V. planifolia* x *V. odorata* hybrid using a combined ITS and chloroplastic phylogenetic analysis (Lubinsky et al., 2008b), when chloroplastic DNA alone repeatedly identified this species as identical to its maternal donor parent *V. planifolia* (figure 1). Moreover molecular genetic diagnostics can only be useful for barcoding biodiversity when species delimitations are either subtle or cryptic but nonetheless clear-cut. In a TCG, taxon limits are themselves diffuse, therefore genetic analysis alone might fail in the identification of discrete species (Ennos et al., 2005). A typical example of expected difficulties will be within the *V. pompona* species complex which was recently described as containing subspecies *pompona*, *pittieri*, and *grandiflora* based on ITS data, although the latter two are rather paraphyletic (Soto Arenas & Cribb, 2010) . In *Vanilla*, taxonomic revision of species will therefore have to use a combination of taxonomic, morphological, ecological, reproductive biology, cytogenetic (polyploidy estimates) and genetic (nuclear and chloroplastic) assessments.
