Skip to main content
NCBI home page
Annals of Botany logoLink to Annals of Botany
. 2020 Apr 18;126(5):849–863. doi: 10.1093/aob/mcaa073

Phenotypic diversity assessment within a major ex situ collection of wild endemic coffees in Madagascar

Aurore Rimlinger 1, Nathalie Raharimalala 2, Véronique Letort 3, Jean-Jacques Rakotomalala 2, Dominique Crouzillat 4, Romain Guyot 5, Perla Hamon 5, Sylvie Sabatier 1,✉
PMCID: PMC7539352  PMID: 32303759

Abstract

Background and Aims

Like other clades, the Coffea genus is highly diversified on the island of Madagascar. The 66 endemic species have colonized various environments and consequently exhibit a wide diversity of morphological, functional and phenological features and reproductive strategies. The trends of interspecific trait variation, which stems from interactions between genetically defined species and their environment, still needed to be addressed for Malagasy coffee trees.

Methods

Data acquisition was done in the most comprehensive ex situ collection of Madagascan wild Coffea. The structure of endemic wild coffees maintained in an ex situ collection was explored in terms of morphological, phenological and functional traits. The environmental (natural habitat) effect was assessed on traits in species from distinct natural habitats. Phylogenetic signal (Pagel’s λ, Blomberg’s K) was used to quantify trait proximities among species according to their phylogenetic relatedness.

Key Results

Despite the lack of environmental difference in the ex situ collection, widely diverging phenotypes were observed. Phylogenetic signal was found to vary greatly across and even within trait categories. The highest values were exhibited by the ratio of internode mass to leaf mass, the length of the maturation phase and leaf dry matter content (ratio of dry leaf mass to fresh leaf mass). By contrast, traits weakly linked to phylogeny were either constrained by the original natural environment (leaf size) or under selective pressures (phenological traits).

Conclusions

This study gives insight into complex patterns of trait variability found in an ex situ collection, and underlines the opportunities offered by living ex situ collections for research characterizing phenotypic variation.

Keywords: Phenology, LMA, LDMC, functional traits, Rubiaceae, ex situ collection

INTRODUCTION

Tropical rainforests of Africa and Madagascar harbour huge biodiversity characterized by great species richness and endemicity (Myers et al., 2000; Yoder and Novak, 2006; Vences et al., 2009). In such a context, for genera that have experienced radiative speciation phenotypic divergence can be important while the associated genetic differentiation could remain low, resulting in low congruency between genetic and phenotypic diversity and in the recognition of morphological units as species (Shaffer and Thomson, 2007). For instance, 58 and 47 Coffea species (excluding the ex-Psilanthus genus and the Baracoffea grouping) from Madagascar and Africa, respectively, have been described (Davis et al., 2006; Couturon et al., 2016), although three and seven species from Madagascar and Africa, respectively, are not yet formally recognized.

For Coffea, clear relationships between species was not possible by analysis conducted with two internal transcribed spacer (ITS) regions plus four plastid sequences (Maurin et al., 2007) or with a limited number of nuclear microsatellite markers (Razafinarivo et al., 2012). A completely resolved molecular phylogenetic tree was obtained using 28 800 concatenated single-nucleotide polymorphism (SNP) markers obtained by genotyping-by-sequencing (GBS) methodology (Hamon et al., 2017). The necessity to use GBS for assessment of between-species genetic divergence could be explained by two non-exclusive mechanisms: (1) the absence of a strong genetic barrier and therefore the maintenance of gene flow between species even at a low rate (Hey, 2006; Niemiller et al., 2008; Nosil, 2008) and (2) recent speciation coupled with rapid morphological divergence (Janssen et al., 2008). Regarding speciation in Madagascar, the general agreement is that ‘the importance of dispersal to the assembly of the Malagasy flora cannot be denied’ (Yoder and Nowak, 2006). The high level of speciation seems to be associated with the existence of different small niches side by side. Therefore, the resulting species should be highly differentiated at the phenotypic level but not, or less so, at the genetic one (Yoder and Nowak, 2006).

In addition to environmental selective pressures, traits (i.e. individual heritable features, as defined by Garnier and Navas, 2012) are also shaped by developmental (i.e. structural limitation on phenotypic variability, as defined by Smith et al., 1985), functional and genetic constraints.

A common evolutionary origin or, on the other hand, evolutionary convergence could be the result of similar trait values for two species, or of correlations between traits. Hence, functionally similar species may belong to the same clade or may be phylogenetically distant. Leaf functional traits such as leaf mass per area (LMA) and leaf dry matter content (LDMC, ratio of dry leaf mass to leaf fresh mass) were also considered in this study, to test the possible compensatory effect between these two traits and whether this correlation still exists in the case of evolutionary trade-off. Indeed, high LMA values characterize plants with slow growth that produce large leaves but whose longevity is low (Westoby et al., 2002); high LDMC values characterize plants whose foliar organs accumulate nutrients and constitute important storage pools with low tissue turnover.

The most recent Coffea molecular phylogeny highlighted independent and parallel diversification in the two landmasses of Africa and Madagascar (Hamon et al., 2017), resulting in total endemism in these main regions. More widely, none of the species occurring in West and Central Africa is found in East Africa, in the Western Indian Ocean Islands or in Asia (broad sense) and vice versa. Another important characteristic concerns the great vulnerability of Coffea species, as their distribution is generally restricted to one or a few forests (Davis et al., 2006; http://publish.plantnet-project.org/project/wildcofdb). In such a context, forest cutovers and burns, human activities and natural hazards (cyclones, fires) could dramatically decrease Coffea richness at both the population and the species level. As a consequence, more than 80 % of Coffea species are present on the Red List (from Near Threatened to Critically Endangered) of endangered species of the International Union for Conservation of Nature (IUCN; Davis et al., 2006).

The ex situ collection at Kianjavato is the one that groups the highest number of Madagascan species. Such a collection is especially important as very few Madagascan Coffea species are present in living collections outside Madagascar. So, this collection represents unique Coffea biological resources. In terms of traits of interest, nearly all Madagascan wild coffees are caffeine-free (for review see Hamon et al., 2015). This collection is also of interest since many Madagascan species are endangered (Davis et al., 2006), and some of them have already disappeared from their natural habitat (Krishnan et al., 2013). The study of genetic diversity of Madagascan and African Coffea species showed a good genetic diversity (Razafinarivo et al., 2012). A comparison of four species and two populations per species in in situ and ex situ conditions showed a good representation of the allelic diversity preserved in the ex situ plant material (Andrianasolo et al., 2013). Yet Madagascan Coffea species show different growth habits, growing as ligneous shrubs or trees (66 species in total; Davis et al., 2006; Couturon et al., 2016; http://publish.plantnet-project.org/project/wildcofdb). Excepting the baracoffea group, (Davis and Rakotonasolo, 2008), they are mainly plants with monopodial growth, axillary inflorescences and evergreen leaves.

Like most Coffea species, they conform to Roux’s architectural model with rhythmic growth and continuous branching, and lateral flowering on the plagiotropic branches (Hallé et al., 1978). In Coffea species, stem growth occurs phytomer by phytomer, with a phyllochron that varies depending on seasonality (Varossieau, 1940). Another striking example of adaptation lies in their phenological traits. For all species, inflorescence and flower development require the dry season to be followed by adequate rainfall (Portères, 1946).The number of days from this triggering rain to the opening of the flower bud (anthesis) is genetically determined and depends on the species considered (Le Pierrès, 1995; Noirot et al., 2016). A better knowledge of the reproductive strategies is an important first step for the development of appropriate conservation strategies.

Within the past few years, as more phylogenetic data have become available, a number of studies have connected phylogenetic analyses with phenotypic, physiological or phenological traits (Bacon et al., 2016; Reginato and Michelangi, 2016). Several descriptive statistics have been introduced to gauge the phylogenetic signal as a quantity that could thus help compare different traits across phylogenies (Pagel, 1999; Blomberg et al., 2003). Not all traits behave similarly for closely related species: while some traits stay stable throughout the phylogeny (Davies et al. 2013), others are highly labile (Cavender-Bares et al., 2009). Owing to shared ancestry, non-independence caused by evolutionary history is expected for species traits. Phylogenetic signal measures this tendency for related species (Blomberg and Garland, 2002): the stronger the signal, the larger the influence of the phylogenetic structure on the species traits.

In this study, the availability of Coffea molecular phylogeny and phenotypic data for a large set of Madagascan wild coffees allows us to address the following questions: How is the phenotypic variation distributed among Madagascan wild coffee species and what are their phenotypic relationships? Are there different reproductive strategies? Are there phenotypic traits associated with species evolution, and thus expressing phylogenetic signal?

MATERIALS AND METHODS

Plant material

Plant material came from the ex situ Coffea collection at the Kianjavato Coffee Research Station (KCRS) in Madagascar (coordinates 21°22′28″ S, 47°52′02″ E). The collection is situated 550 km from Antananarivo, at altitudes ranging from 50 to 400 m above sea level. The collection is mainly constituted by Madagascan wild Coffea species (Chevalier, 1942) excluding the Baracoffea group from the Western coast described by Davis and Rakotonasolo (2008). They are representatives of the five botanical series from the Great Island, which were classified by Charrier (1978) according to floral and leaf traits. These series are named Garcinioides, Millotii complex, Multiflorae, Subterminales and Verae.

The coffee trees sampled were adult trees mainly collected between the 1960s and the 1980s, growing under natural forest cover. Details of plant material (such as species name, botanical series, geographical origin and coordinates) are provided in Table 1.

Table 1.

Coffee species examined, with population code, species name, botanical series (from Charrier, 1978), origin, province, GPS coordinates and habitat class (1, <2 dry months; 2, 2–6 dry months; 3 , >6 dry months)

Population code Species Botanical series Origin/locality Province Latitude Longitude Habitat class
A964 C. dubardii Jum. Garcinioides Anivorano Nord Antsiranana (North) 12°44′ S 49°09′ E 3
A969 C. dubardii Jum. Garcinioides North of Vohemar (Diégo-Suarez) Antsiranana (North) 13°20′ S 49°57′ E 3
A516 C. heimii J.-F.Leroy Garcinioides Diégo-Suarez (Sahafary) Antsiranana (North) 12°35′ S 49°26′30″ E 3
A40 C. mogenetii Dubard Garcinioides Montagne d’Ambre Antsiranana (North) 49°12′ S 12°26′ E 3
A252 C. tetragona Jum. & H.Perrier Garcinioides Behangony (Est de Maromandia) Mahajanga (Northwest) 14°13′30″ S 48°09′ E 3
A601 C. abbayesii J.-F.Leroy Millotii complex Fort-Dauphin (Isaka-Ivondro) Toliara (Southeast) 24°45′15″ S 46°51′45″ E 2
A572 C. ambodirianaensis Portères Millotii complex reserve Ambodiriana Toamasina (East) 18°27′08″ S 48°56′36″ E 1
A206 C. dolichophylla J.-F.Leroy Millotii complex Nosy Varika Vohimanoro), Ambodilafa Fianarantsoa (Southeast) 20°29′ S 48°10′30″ E 1
A208 C. farafanganensis J.-F.Leroy Millotii complex Farafangana (Amboangibe) Fianarantsoa (Southeast) 22°53′ S 47°48′ E 1
A212 C. millotii J.-F.Leroy Millotii complex Vatovavy Fianarantsoa (Southeast) 21°24′3″ S 47°56′32″ E 1
A219 C. millotii J.-F.Leroy Millotii complex Ambatovaky Fianarantsoa (Southeast) 21°22′48″ S 47°53′06″ E 1
A222 C. millotii J.-F.Leroy Millotii complex Mananjary (Tolongoina) Fianarantsoa (Southeast) 21°31′ S 47°25′ E 2
A721 C. millotii J.-F.Leroy Millotii complex Ampasinambo (Nosy Varika) Fianarantsoa (Southeast) 20°31′15″ S 48°01′75″ E 1
A575 C. richardii J.-F.Leroy Millotii complex Fenerive-Est (Tampolo) Toamasina (East) 17°17′ S 49°25′ E 1
A817 C. richardii J.-F.Leroy Millotii complex Soanierana Ivongo Toamasina (East) 16°45′ S 49°35′ E 1
A227 C. andrambovatensis J.-F.Leroy Multiflorae Tolongoina Toamasina (East) 21°46′15″ S 47°57′07″ E 1
A310 C. andrambovatensis J.-F.Leroy Multiflorae Vatovavy Fianarantsoa (Southeast) 21°24′3″ S 47°56′32″ E 1
A529 C. ankaranensis J.-F.Leroy ex A.P.Davis & Rakotonas. Multiflorae Standoko (Diégo-Suarez) Antsiranana (North) 12°45′13″ S 49°08′48″ E 3
A808 C. ankaranensis J.-F.Leroy ex A.P.Davis & Rakotonas. Multiflorae Ankarana (Diégo-Suarez) Antsiranana (North) 12°50′57″ S 49°32′36″ E 3
A403 C. arenesiana J.-F.Leroy Multiflorae Moramanga (Ambodivato) Toamasina (Centre) 18°56′ S 48°12′ E 2
A303 C. bertrandii A.Chev. Multiflorae Italy (Fort-Dauphin) Toliara (Southeast) 25°02′ S 46°59′ E 3
A5 C. bertrandii A.Chev. Multiflorae Mahampoana (Fort-Dauphin) Toliara (Southeast) 25°01′ S 46°38′ E 3
A956 C. costei sp. nov. Ined. Multiflorae Route côtière Sambava Antsiranana (North) 13°45′13″ S 50°0′48″ E 2
A570 C. coursiana J.-F.Leroy Multiflorae Tamatave (Betampona RNI n°1) Toamasina (East) 17°53′ S 49°13′ E 1
A315 C. leroyi A.P.Davis Multiflorae Ifanadiana (Ambodiarafia) Fianarantsoa (Southeast) 21°20′ S 47°45′ E 1
A18 C. mangoroensis Portères Multiflorae Ambatondrazaka Toamasina (East) 19°16′3″ S 48°24′09″ E 1
A401 C. mangoroensis Portères Multiflorae Mangoro (Moramanga) Toamasina (East) 18°55′ S 48°14′ E 2
A402 C. mangoroensis Portères Multiflorae Mangoro (Moramanga) Toamasina (East) 18°56′ S 48°13′ E 2
A321 C. montis-sacri A.P.Davis Multiflorae Vatovavy Fianarantsoa (Southeast) 21°24′35″ S 47°56′32″ E 1
A12 C. perrieri Drake ex Jum. & H.Perrier Multiflorae Fort-Dauphin (Amboasary-Atsimo Toliara (Southeast) 24°50′ S 46°33′55″ E 3
A421 C. perrieri Drake ex Jum. & H.Perrier Multiflorae Ihosy Fianarantsoa (Southeast) 22°33′03″ S 46°07′40″ E 3
A730 C. perrieri Drake ex Jum. & H.Perrier Multiflorae Tsaratanàna Mahajanga (Northwest) 16°47′ S 47°39′ E 3
A71 C. resinosa (Hook.f.) Radlk. Multiflorae Maroantsetra (Forêt Farakaraina) Toamasina 15°26′ S 49°44′ E 1
A8 C. resinosa (Hook.f.) Radlk. Multiflorae Nosy Varika Fianarantsoa (Southeast) 20°36′26.6″ S 48°31′56.5″ E 1
A827 C. resinosa (Hook.f.) Radlk. Multiflorae Fananehana Toamasina 15°53′26″ 49°42′19″ E 1
A906 C. resinosa (Hook.f.) Radlk. Multiflorae Tanambao Antsiranana (North) 17°43′38″ S 48°49′59″ E 2
A910 C. resinosa (Hook.f.) Radlk. Multiflorae North Soanierana Ivongo Toamasina (East) 16°55′17″ S 49°35′07″ E 1
A913 C. resinosa (Hook.f.) Radlk. Multiflorae Manompana Toamasina (East) 16°40′58″ S 49°45′17″ E 1
A915 C. resinosa (Hook.f.) Radlk. Multiflorae Ivontaka Toamasina (East) 16°18′ S 49°49′00″ E 1
A538 C. sahafaryensis J.-F.Leroy Multiflorae Sahafary (Diégo-Suarez) Antsiranana (North) 12°53′18″ S 49°22′ E 3
A978 C. sahafaryensis J.-F.Leroy Multiflorae North of Vohémar Antsiranana (North) 13°17′20″ S 49°57′ E 3
A20 C. vianneyi J.-F.Leroy Multiflorae Ampasinambo (Nosy Varika) Fianarantsoa (Southeast) 20°32′53″ S 48°07′50″ E 1
A946 C. vianneyi J.-F.Leroy Multiflorae Moramanga Toamasina (Centre) 18°56′ S 48°15′ E 2
A966 C. augagneurii Dubard Subterminales Anivorano Nord Antsiranana (North) 12°43′ S 49°10′ E 3
A973 C. boiviniana (Baill.) Drake Subterminales Nafokovo (Nord Vohémar) Antsiranana (North) 13°22′13″ S 49°58′12″ E 3
A980 C. boiviniana (Baill.) Drake Subterminales Vohémar Antsiranana (North) 13°21′ S 49°52′48″ E 3
A535 C. bonnieri Dubard Subterminales Diego-Suarez (Forêt d’Ambre) Antsiranana (North) 12°29′30″ 49°10′40″ E 3
A974 C. jumellei J.-F.Leroy Subterminales Nafakovo (North Vohémar) Antsiranana (North) 13°22′13″ 49°58′12″ E 3
A957 C. pervilleana (Baill.) Drake Subterminales Route côtière Sambava-Diégo Suarez Antsiranana (North) 13°49′17″ 50°02′04″ E 2
A958 C. pervilleana (Baill.) Drake Subterminales Route côtière Sambava-Diégo Suarez Antsiranana (North) 13°49′17″ S 50°02′04″ E 2
A528 C. ratsimamangae J.-F.Leroy ex A.P.Davis & Rakotonas. Subterminales Sandokoto (Diégo-Suarez) Antsiranana (North) 13°48′30″ S 49°17′39″ E 2
A967 C. ratsimamangae J.-F.Leroy ex A.P.Davis & Rakotonas. Subterminales Anivorano Nord Antsiranana (North) 12°44′ S 49°14′ E 3
A307 C. sakarahae J.-F.Leroy Subterminales Iakora (Ihosy) Fianarantsoa (South) 23°06′ S 46°39′ E 3
A515 C. tsirananae J.-F.Leroy Subterminales Diégo-Suarez (Cap d’Ambre) Antsiranana (North) 11°58′ 49°16′ E 3
A308 C. vatovavyensis J.-F.Leroy Subterminales Vatovavy Fianarantsoa (Southeast) 21°24′ S 47°56′ E 2
A830 C. vatovavyensis J.-F.Leroy Subterminales Unknown Unknown Unknown Unknown NA
A977 C. vohemarensis A.P.Davis & Rakotonas. Subterminales North of Sambava Antsiranana (North) 14°14′ 50°8′30″ E 2
A574 C. homollei J.-F.Leroy Verae Réserve Ambodiriana (Tamatave) Toamasina (East) 17°59′13″ S 49°17′48″ E 1
A213 C. kianjavatensis J.-F.Leroy Verae Vatovavy Fianarantsoa (Southeast) 21°24′30″ S 47°56′30″ E 1
A602 C. kianjavatensis J.-F.Leroy Verae Fort-Dauphin Toliara (Southeast) 24°45′15″ S 46°51′45″ E 2
A320 C. lancifolia A.Chev. Verae Mananjary (Madiorano/Tolongoina) Fianarantsoa (Southeast) 21°31′ S 47°29′ E 1
A405 C. lancifolia A.Chev. Verae Tolongoina (Kararaika forest) Fianarantsoa (Southeast) 22°57′02″ S 47°22′56″ E 2
A571 C. lancifolia A.Chev. Verae Reserve Ambodiriana Toamasina (East) 18°27′08″ S 48°58′36″ E 1
Open in a new tab

Phenotypic records

Records were made during the years 2010–11 and 2013–14 on a total of 364 plants from 63 populations corresponding to 36 species. Descriptive traits (such as leaf shape and colour, domatia position, aperture shape and pubescence) and quantitative traits (numbers of leaves, fruits, seed size and phenology, number of days between the triggering rain and blossoming, number of days between flowering and ripe fruits [referred to hereafter as the maturation phase] and number of flowers per inflorescence) were recorded. Except for leaf, fruit and seed measurements, the traits were recorded on a maximum of ten trees (depending on availability) per population.

Mature leaf measurements were performed on five leaves (picked on the fifth branching from the top) and five trees per population. To avoid any bias due to artificial desiccation, fresh mass was measured directly after sampling on site at Kianjavato research centre.

Measurements were done on 20 fruits (including fruit disc diameter) and 40 seeds (two seeds per fruit, including 100-seed weight) per population. Four additional variables were calculated: (1) leaf shape as the ratio of length to width; (2) leaf area as length × width; (3) volume of fruits; and (4) seeds as length × width × thickness. The 28 variables and range of variation are listed in Supplementary Data Table S1.

The number of days for maximum seed germination was obtained after sowing between 68 and 120 seeds per harvest in nurseries at Kianjavato. This trait was recorded for 47 populations corresponding to 28 species.

Two functional attributes, leaf mass per area (LMA, recorded for 39 populations either 33 species) and leaf dry matter content (LDMC, i.e. the ratio of leaf dry mass to leaf fresh mass), which was recorded for 30 populations either 29 species, were added, as well as two physiological attributes: stomata density and stomata length, as measured in Razafinarivo et al. (2012) were recorded for 56 populations either 35 species. The Leaver module was used to extract leaf area with the plugin Toaster (Tree and Plant Organs and Structures Analyzer, Borianne, 2012) run in ImageJ software. Leaf dry mass was obtained after drying for 4 d in an oven at 60 °C. For all species, in order to compare even-aged organs, measurements of LMA and LDMC were made on the youngest metamer at the branch extremity and in full light. The diameter of metamers was also measured on these metamers, which were green.

Environmental parameters

Climatic data were extracted from GPS coordinates of each species (Table 1) and from WorldClim information (http://www.worldclim.org). The habitat class (Table 1) was obtained from the distribution of the number of dry months in the original habitat of each species.

Among the 19 available bioclimatic variables, four (annual mean temperature, maximum temperature of warmest month, minimum temperature of coldest month and annual precipitation) were used in this study. Additional climate data, such as annual potential evapotranspiration, climatic water deficit and number of dry months, were extracted from MadaClim (https://madaclim.cirad.fr/). An index of aridity was also calculated as annual precipitation minus annual potential evapotranspiration (Table 2).

Table 2.

Physical parameters of habitats corresponding to habitat classes 1–3 as defined by the number of dry months

Habitat class Number of dry months Annual precipitation (mm) Temperature of the coldest month (°C) Temperature of the warmest month (°C) Annual potential evapotranspiration (mm) Aridity index (mm)
1 0–1 2051–3360 11.6–19.9 26.3–30.9 945–1289 1081–2142
2 2–6 1260–2265 10.6–20.7 23.1–30.9 814–1370 175–1111
3 >6 552–1829 10.2–22.8 28–33.1 950–1627 −691 to 379
Open in a new tab

Statistical analyses

Relationships among Madagascan wild coffees.

Given that the recorded traits are quantitative and qualitative, quantitative traits (in all 14) were transformed into either similar size classes or clearly distinct range classes depending on the distribution of the trait. For qualitative traits, each modality was treated as a new trait. Therefore, from the 28 initial variables, a total of 107 new variables were obtained and encoded 1 for presence and 0 for absence. To assess relationships between individuals within populations, populations within species and between species, a neighbour-joining tree based on a matrix of Dice’s dissimilarity indices and 200 bootstraps was constructed using DARwin v0.5 (Perrier and Jacquemoud-Collet, 2006).

For quantitative traits (reproductive, vegetative, phenological, physiological and functional leaf traits) we used standardized data to visualize for each trait and each species the deviation from the mean of the values obtained for all species. The results for each species were ordered following the rooted molecular phylogenetic tree extracted from Hamon et al. (2017).

Relationships between phenotypic and bioclimatic traits.

Three habitat classes of species were defined according to the number of dry months at the original population location: 0 or 1 dry month for S1, 2–6 dry months for S2 and >6 dry months for S3. Comparisons were performed on seven traits recorded for all populations investigated (63 populations corresponding to 36 species). For 15 traits, data were obtained for 54–61 populations. Six traits were recorded for 30–39 populations corresponding to at least 29 species. Mean comparisons between subsets were performed with ANOVA using R (https://www.r-project.org/).

Analysis of interspecific and intraspecific (population) diversity.

We summarize the interspecific and intraspecific (population) trait variation in Supplementary Data Table S2, showing the mean trait values for a set of species. Differences between populations were analysed by one-way ANOVA and the Tukey test. We also verified the contribution of population level to total trait variation by decomposing the variance of each trait using a linear mixed model that took into account the nested structure of the sampling design (package lme4, Bates et al., 2015).

Phylogenetic signal.

The phylogenetic signal allows a better understanding of evolutionary traits within a group by looking at trait relatedness between species in the light of their phylogeny. It can be measured using different indices: Pagel’s λ (Pagel, 1999) and Blomberg’s K (Blomberg et al., 2003) were selected after assessment of their respective performance and shortcomings. Pagel’s λ, estimated by a maximum likelihood approach, provides the most probable description of the observed trait distribution given a Brownian model for trait evolution. A value of λ equal to 1 indicates that the trait evolves exactly as suggested from the original topology of phylogeny while a value equal to 0 means an absence of phylogenetic signal. Blomberg’s statistic (K) quantifies the phylogenetic signal. It is a descriptive statistical parameter to describe the degree of difference between the F-statistic of simulated data and the observed F-statistic distribution as the ratio of the observed mean squared error derived from a phylogenetically corrected mean and the expected mean square error obtained from the analysis by considering tree topology and branch length information. A K value inferior to 1 suggests that the trait’s evolution was independent of phylogeny (i.e. the phylogenetic signal is low). A K value equal to 1 suggests that the trait has evolved at a constant rate during species evolution, while K > 1 indicates a high phylogenetic signal for the considered trait (i.e. the trait is more similar for related species than expected by the evolutionary model). The significance of K is obtained from 999 permutations. The phylogenetic signal was only considered significant for a trait when Pagel’s λ and Blomberg’s K gave consistent results with the following thresholds for the P values: Pλ < 0.05 and PK < 0.1. Phylogenetic signals were estimated with the function phylosig in the package phytools (Revell, 2012) from mean values per species for 29–36 species depending on the trait. Statistical tests (normality, independence and homoscedasticity of residuals) for validation of the linear model were performed using the package lmtest (Zeileis and Hothorn, 2002). Calculations were performed with the function pgls of the R package caper (Orme et al., 2012). Figure drawings were made within R 3.2.5.

RESULTS

Structuring of wild coffees based on traits

Individual phenotypic relationships showed that, globally, intraspecific variation was lower than interspecific variation, which can be seen in Supplementary Data Table S2, with ANOVA results on interspecific variation compared with intraspecific (population) variation. The deviation from the mean was lower for intraspecific variation, and populations were generally pooled into fewer groups than were species. The contribution of intraspecific diversity to the variance of species traits (Supplementary Data Table S3) varied between traits, being fairly high (leaf length, number of flowers per inflorescence), moderate (shape index, petiole length) or low (length of the maturation phase).

The estimation of dissimilarity indices within populations was lower than between species (Fig. 1). As an example, C. resinosa individuals were grouped close to each other, even if they exhibited considerable phenotypic variability.

Fig. 1.

Fig. 1.

Open in a new tab

Phenotypic relationships among Madagascan Coffea species. Neighbour-joining tree based on Dice’s dissimilarity indices calculated from 107 variables encoded for presence/absence for 364 individuals and 200 bootstraps using DARwin v5 (Perrier and Jacquemoud-Collet, 2006). Only bootstrap values higher than 80 % are shown. The botanical series to which each species belongs to is given in yellow for Garcinioides, black for Millotii complex, blue for Multiflorae, pink for Subterminales and gold for Verae.

Only two Charrier’s series (the Millotii complex and the Garcinioides) appeared very homogenous but no overall structuring according to botanical series classification is obvious. Using standardized data for each trait made it possible to discern some trends following phylogenetic species relationships (Fig. 2). The ratio between internode mass and leaf mass was higher for three species that are highly branched and bear small leaves: C. vohemarensis and C. vatovavyensis (Subterminale series) and C. bertrandii (Multiflorae series).

Fig. 2.

Fig. 2.

Open in a new tab

Phenotypic tendencies. Standardized trait values are shown in columns for each species following their phylogenetic relationships based on Hamon et al. (2017), with large black squares indicating the highest values. The series Millotii complex is coloured in red and species with a longer length maturation phase are shown in gold. Leaf-related and vegetative traits are in orange, phenological traits and inflorescence trait are in yellow, fruit and seed traits are in light blue, functional leaf trait are in dark blue and stomata traits are in purple.

Another grouping of sister species (Fig. 2, in red) was characterized by long petioles and leaves, thick and long internodes, large leaf area and large fruits and seeds (Fig. 3A). Similarly, a grouping of northern species (Fig. 2, in yellow) was characterized by a long maturation phase and included two species easy to identify from their fruit shape (C. tsirananae, Fig. 3B) or leaf shape (C. heimii, Fig. 3C). The latter species also exhibited the highest value for LMA.

Fig. 3.

Fig. 3.

Open in a new tab

Internodes, fruits and seeds of C. richardii (A), fruit of C. tsirananae (B) and leaf of C. heimii (C).

Reproductive traits and strategies

Madagascan wild coffee trees exhibited a wide range of number of days between the triggering rain and flowering (from 4 to 13 d). Two series, Garcinioides and Millotii complex, were characterized by a long duration from rainfall to flowering (10–13 d) while Verae exhibited a short to medium duration (4–7 d). Regarding the three bioclimatic subsets of species, S1, S2 and S3, the range of the number of days between the triggering rain and blossoming was large for the three subsets: from 4 to 10 (S2) or 13 (S1) and from 7 to 12 for S3 (Fig. 4A). Multiflorae appeared as the most diverse series, with durations from short (4 d) to long (10 d) (Fig. 4B).

Fig. 4.

Fig. 4.

Open in a new tab

(A) Variation in time between the triggering rain and blossoming according to the number of dry months (0–1 for S1 or habitat class 1, 2–6 for S2 or habitat class 2 and >6 for S3 or habitat class 3) for each population. (B) Dates of blossoming according to the botanical series of each population. GAR, Garcinioides; MIL, Millotii complex; MUL, Multiflorae; SUB, Subterminales; VER, Verae.

Maturation phase lengths were also diverse and ranged from 59 to >350 d. In very humid habitats (0 or 1 dry month) as well as in intermediate habitats (2–6 dry months), species ripen in 2–6 months. In very dry habitats (>6 dry months), species ripened in contrasting times, either 2–3 months (for the majority) or around 1 year (Fig. 5A). Dates of harvest varied through the year from January to December (Fig. 5B). Early species (all Verae and most Subterminales and Multiflorae) were harvested in January or February. Medium species (mainly Millotii) were harvested from March to May and very late species (all Garcinioides) from mid-November to December. The sequence of harvest was nearly invariable whatever the year. Considering seed germination, there was generally either no dormancy or germination after a couple of months.

Fig. 5.

Fig. 5.

Open in a new tab

(A) Variation in length of the maturation phase (in days) according to the length of the dry season (in number of dry months). (B) Dates of fruit ripening according to botanical series. GAR, Garcinioides; MIL, Millotii complex; MUL, Multiflorae; SUB, Subterminales; VER, Verae.

In humid (S1) and intermediate (S2) habitats, the maturation phase and seed germination occurred during the hot rainy season with a short time between the two physiological steps. However, in dry habitats (S3) two behaviours were found. Some species, such as C. boiviniana, had a very short maturation phase followed by seed germination few weeks later, and other species, such as C. tsirananae, had a long maturation phase (around 1 year) followed by immediate seed germination (occurring during the rainy season at year n + 1 after flowering). An overview of the timing of the different reproductive phases considered as reproductive strategies is given in Fig. 6.

Fig. 6.

Fig. 6.

Open in a new tab

Reproductive strategies of Madagascan wild coffee trees. Diagram showing the flowering (blue colours) and fructification (grey) periods and the germination period (yellow colour) according to the length of the rainy/dry seasons during the year. The length of the rainy season is indicated by clouds. The number of dry months in the dry season is indicated by suns.

Environmental effect

Given the distribution of the number of dry months in their natural habitats, three habitat classes of species were defined from the bioclimatic parameters of their natural habitat. The main features of these habitat class of species are given in Table 2.

The higher the number of dry months, the greater the range of the average temperature of the coldest month and the greater the annual potential evapotranspiration as a result of warm temperatures.

Regarding biological traits, mean comparisons between habitat classes for each trait showed significant differences between the three habitat classes, as described in Table 3 for vegetative traits (leaf length, leaf width and leaf area, two-leaf mass and internode diameter perpendicular to leaves, internode mass/two-leaf mass ratio, internode length/diameter ratio, ratio of length of internode to internode diameter parallel to leaves), reproductive traits (number of flowers per inflorescence, maturation phase length and seed thickness). The ratio between the mass of the internodes and the mass of the two leaves varied significantly with habitat (Table 3). This ratio was higher for habitat classes 2 and 3 than for class 1. The size of leaves of species from habitat class 1 and two-leaf mass were higher. In contrast, seed thickness was significantly higher for species of habitat class 3.

Table 3.

Mean comparison between the three bioclimatic subsets of species defined on the number of dry months in the natural habitat of the studied population/species. Mean for each trait according to the number of dry months in the natural habitat of coffee populations. Difference in means between subsets was tested with ANOVA. ***P < 0.001, **P < 0.01, *P < 0.05; values <0.05 are considered as evidence of trait divergence

Habitat class 1 2 3 P-value (ANOVA)
Leaf length (mm)*** 113.05185 76.3 68.15909 0.000185
Leaf width (mm)*** 48.34444 35.02857 28.92727 2.16E−05
Leaf area (mm2)*** 6150.185 3156.571 2299.262 0.00012
Two leaves mass (g)** 1.465 0.453 0.5341176 0.00925
Internode diameter (perpendicular to leaves, mm)** 3.720833 2.293 2.195625 0.00454
Maturation phase length (d)** 104.6296 105.0714 178.3182 0.00308
Ratio internode mass/two-leaf mass (g)* 0.1275 0.259 0.2175 0.0487
Ratio length of internode/internode diameter parallel to leaves* 12.78 24.117 21.7975 0.039
Number of flowers per inflorescence* 4.403704 2.8 3.177273 0.0123
Seed thickness (mm)* 3.885926 4.053077 4.673 0.0422
Leaf shape 2.362 2.17 2.588 0.412
Internode diameter (parallel to leaves, mm) 2.058333 1.587 1.636875 0.123
Petiole length (mm) 8.12 6.45 6.34 0.151
Internode length (mm) 23.16667 32.1 33.35 0.056
Internode mass (g) 0.2143 0.0878 0.101 0.121
Fruit length (mm) 15.86148 14.69077 15.5035 0.795
Number of days from triggering rain to flowering 7.648148 7.321429 8.475 0.13
Fruit thickness (mm) 12.69296 11.46154 11.442 0.445
Fruit shape 1.225926 1.169231 1.2535 0.707
Fruit volume (mm3) 4836.13 3041.622 2910.932 0.212
Pedicel length (mm) 3.498889 3.811538 4.293 0.108
Disc diameter (mm) 3.391111 3.181667 2.784444 0.3
Seed length (mm) 9.183704 8.683846 10.3285 0.237
Seed width (mm) 5.692593 5.041538 6.57 0.228
Seed volume (mm3) 295.5733 219.8423 352.434 0.549
100-seed weight (g) 9.677778 8.8175 14.291333 0.138
LMA mean (g m−2) 175.7083 137.78 202.3235 0.235
LDMC (mg g−1) 357.1 344.4143 382.1583 0.388
Mean stomata density 203.5238 152.2308 171.3182 0.0797
Stomata length (µm) 8.12 8.378923 8.225682 0.845
Open in a new tab

The bold type numerals in this table correspond to the significant different traits between the three environments.

Phylogenetic signal

The phylogenetic signal was assessed using Pagels’ λ and Blomberg’s K parameters to describe the degree of difference between the F-statistic of simulated data obtained from 999 permutations and observed F-statistic distributions.

We found that ten traits (internode mass/leaf mass ratio, petiole length, parallel internode diameter, internode length/diameter ratio, leaf length, length of the maturation phase, seed shape index, LDMC, stomata length and stomata density) evolved in relation to species evolution (Fig. 7).

Fig. 7.

Fig. 7.

Open in a new tab

Phylogenetic signal in coffee trees traits using Blomberg’s K parameter. A K value close to 1 (black dashed line) and higher than the lowest threshold value of 0.52 (grey dashed line) indicates a strong phylogenetic effect for a trait. The significance data (*P < 0.05) were consistent with those from Pagel’s λ (P < 0.05). Leaf-related and vegetative traits are in orange, phenological traits in yellow, fruit and seed traits in light blue, functional leaf traits in dark blue and stomata traits in purple.

DISCUSSION

The data were acquired in an ex situ collection of endemic wild Coffea species. This study was only possible because this collection is the only one that includes such a large number of Madagascan wild species. Still, the data collected do not come from the species’ natural habitats. This is an important point in our study, as trait differentiation exhibited by species under ex situ cultivation compared with natural populations has been reported theoretically (Ensslin and Godefroid, 2019) and experimentally (Rauschkolb et al., 2019). Nevertheless, trait changes have been mostly reported for annual species (Enßlin et al., 2011), which have had the time to evolve through multiple generations. Ex situ collections of perennial species have been shown to conserve highly differentiated populations (Barth et al., 2009) and to be widely used for phenotypic and genetic characterization (Migicovsky et al., 2019). Indeed, great morphological and phenological variability is still observed between the species in the Kianjavato ex situ collection. Trees are generally the same age and grow in a similar environment and thus are not subject to trait variation caused by different developmental stages and environments (Fortunel et al., 2020). We discuss below the different phenotypic differences observed in these trees.

Reproductive strategies and limitation of gene flow

Our study showed that phenology parameters (the number of days to blossoming after the triggering rain and the maturation phase length; Figs 4 and 5) in the ex situ collection did not correlate with original natural habitat type; a similar range of variation was observed in the three types. Nonetheless, the phenological traits varied considerably among Coffee species, suggesting that ex situ collections could be useful in investigating species’ phenology.

For coffees, flowering requires two important steps. The first is the floral induction that is set up after a dry period. It is climate-dependent, as reported in New Caledonia by Gomez et al. (2016). Indeed, flowering patterns for the three cultivated Coffea species (C. canephora, C. liberica and C. arabica) introduced at the beginning of the 19th century followed the timing and sequence of precipitation in the study area. The second step corresponds to the development of the flower buds following a triggering rain (of at least 10 mm; Portères, 1946). It is genetically determined, since the average number of days between the triggering rain and blossoming is fixed for each species and ranged between 5 and 13 d (Noirot et al., 2016). These values remained invariable whatever the growing place, as was observed in New Caledonia for the three introduced species (Gomez et al., 2016). In such conditions, climate changes or a new climate growing environment should have affected only floral induction and then, as a consequence, the quality of blossoming, which depends on the intensity of the triggering rain. Under climatic changes, flowering could be disturbed. Erratic flowerings might open the way for flowering synchronization of usually non-synchronous flowering species, thus favouring unusual gene flow. However, the loss of species integrity also involves cross-species hybridization success and capability to produce fertile progeny. This phenomenon is probably the origin of the natural cross between two wild species, producing the famous Arabica coffee (Carvalho, 1952; Lashermes et al., 1999).

Within tropical forests, a broad spectrum of flowering patterns has been found (Sakai, 2001) and observed variations in phenology may arise from phenotypic plasticity or from genetically based local adaptations or a combination of these (e.g. Zalamea et al., 2011; Anderson et al., 2012). Coffee flowering phenology has been characterized as highly dependent on climatic factors such as precipitation (Crisosto et al., 1992) and temperature (Lin, 2008), and is genetically controlled (Le Pierrès, 1995). The time between the triggering rain and the flowering date was well conserved. For Madagascan wild coffees, flowering starts in September with Coffea sp. ‘Sambava-Diego’ (described in Couturon et al., 2016 and in http://publish.plantnet-project.org/project/wildcofdb) and C. vatovavyensis. They are followed in October by some members of Subterminales, Millotii complex and Multiflorae species, followed in November by the late-flowering species belonging to Millotii complex and Multiflorae, all the Garcinioides and C. boiviniana, C. alleizettii, C. augagneuri, C. ratsimamangae and C. sakarahae (Subterminales) and finishing in December with the very late-flowering C. tsirananae (Subterminales).

Flowering patterns in Coffea thus appear to have a strong genetic component, causing the flowering times of each species to remain distinct (Akaffou et al., 2014). Partly asynchronous responses among species to key environmental factors would thus be expected to act as prezygotic barriers. The lack of a well-marked dry season is considered to influence flower bud formation by creating different levels of induction (Crisosto et al., 1992) and favouring multiple flowering events during a season (Allan and Pannell, 2009). However, we do not know whether adaptive evolution will allow populations to reach new phenotypic optima rapidly enough to keep pace with climate change (Anderson et al., 2012).

Reproductive barriers due to no overlap of flowering is very important for sympatric species, especially since cross-fertilizations of two different species were successfully tested (Louarn,1992; Charrier, 1978). Therefore, in natural conditions, species integrity in sympatry should be mainly maintained by the absence of flowering overlap, highlighting the necessary differences in the interval between the triggering rain and blossoming. However, this physiological barrier is not absolute. Floral induction is highly dependent on climate parameters (alternating dry and rainy periods). Reproductive strategies thus involve mechanisms that favour species integrity (physiological flow gene barrier), and species survival in climate changes and during new niche colonization.

For species survival, legitimate seed germination must benefit from rains to produce plantlets that can skip the dry season without damage. In Madagascar, northern species among Subterminales have either a very short or a very long maturation phase. This indicates two opposite strategies to avoid the dry season. Three behaviours have been observed in seasonally dry tropical forests: (1) flowering and fruiting exclusively in the rainy season, (2) flowering in the rainy season and fruiting in the dry season, and (3) flowering and fruiting in the dry season (Luna-Nieves et al., 2017). The seasonal variation of irradiance influenced seed development times (Zimmerman et al., 2007).

Climate effect on traits related to plant development

Our results showed that the climatic factors of the original natural habitat affect leaf size and mass of trees from the ex situ collection. The diameter of the youngest internodes, which is a proxy for the size of the meristem, decreased with increasing aridity (Table 3). The size of leaves is a trait adaptive to drought (Westoby et al., 2002; Wright et al., 2017): in a dry climate the leaves are small and thick with a high LMA (Niinemets, 2001).

Generally, Coffea species of humid regions have large leaves, short internodes and large internode diameter, therefore having a low ratio of internode length to diameter. The shape of the stem is more massive. On the contrary, species of dry regions have long internodes and small internode diameter and thus a higher ratio of internode length to diameter (i.e. a slender stem shape), especially for habitat class 2. Our results are in line with Corner’s rules on leaf/stem size: species with large leaves have a thick stem and large fruits and weak branching while species with small leaves have a thin stem, small fruits and strong branching (Corner, 1949). But it is not the species with the biggest fruits that have proportionally the biggest seeds (Aarssen, 2005). Our results show that seed thickness is significantly higher for species of habitat class 3. Thus, in a dry habitat the seeds were thicker.

The shape of the stem (i.e. the ratio of its internode length to its diameter, also called slenderness) affects its biomechanical properties (Lauri, 2019). The greater its slenderness, the lower its mechanical stability.

Trees with small leaves have a higher number of leaves compared with trees with large leaves (Kleimann and Aarssen, 2007). There is a trade-off between leaf mass and leaf number. In coffee trees, it could have an influence on the phyllochron (the time between the production of two successive metamers) as observed in African coffee trees (unpubl. data, Sylvie Sabatier). Small-leaved species have a shorter phyllochron and thus many more leaves per axis for a similar growing period.

Climatic factors influence the value of the ratio between internode mass and leaf mass, which are higher for species growing in intermediate and dry regions (Table 3).These species therefore allocate more to the stem than to the leaves. In dry environments, species with small leaves are favoured because they have a high LMA and thus better photosynthetic capacity (Westoby et al., 2002). The mechanical cost is lowest for species with small leaves (Givnish, 1987). Small leaves are dense and thick (Niinemets, 2001) with low transpiration (Westoby et al., 2002).

Our results show, for instance, that seed thickness is significantly higher for species of habitat class 3, which is the driest habitat. This opens a perspective for future analyses: in the context of the challenges posed by climate change, our data set could be used to study ex situ phenotypic variation of species in relation to their original environment.

Phylogenetic signal

Some of the results we obtained are in good agreement with other studies on phylogenetic signal in plants: the strong conservatism of LDMC had already been thoroughly described (Purschke et al., 2013; Kazakou et al., 2014; Fort et al., 2015; Poorter et al., 2018). Some other traits less studied, such as the ratio of internode mass to leaf mass (illustrating the variability in stem versus leaf biomass allocation between species), also expressed a highly significant signal (Fig. 7). Stomatal length and stomata density have a significant phylogenetic signal. In the Coffea genus generally, species with smaller genomes have smaller leaves with larger stomata and grow in dry environments, while species with larger genomes from humid habitats have longer leaves with a higher stomata density (Razafinarivo et al., 2012).

Leaf mass per area reflects the cost of leaf construction, light interception and plant fitness with a low phylogenetic signal (Flores et al., 2014). It is linked to leaf lifespan and to photosynthetic capacity per mass unit (Am). Generally, species with low LMA have fast growth rate (Flores et al., 2014).

The phylogenetic signals of leaf petiole length, leaf length, leaf area and leaf width were weakly significant. The size of the leaves is mainly controlled by global climatic factors (Wright et al., 2017). The phylogenetic signals of the ratio of internode length to diameter or slenderness were also weakly significant. Stem size is species-dependent within Coffea. The ratio between internode mass and leaves mass shows a higher phylogenetic signal. These relationships have been studied with the aim of better understanding axialization processes (Lauri and Normand, 2017). Our results show that this trait would be related to species evolution within the Coffea genus. For future research, it should be checked whether these results are found in other groups of plants.

The length of the maturation phase shows a high, significant phylogenetic signal. It is well known that the phenology is an adaptive trait (Chuine, 2010). These results highlight the assumption that phenological sensitivity was a significant predictor of species adaptation to climate change (Cleland et al., 2012).

Conclusions

The ex situ collection of Madagascan wild coffees proved to be a repository harbouring a wide diversity of phenotypic variation among Coffea species. It helped us improve our understanding of coffee tree biology and the complexity of associations between morphological or phenological traits and climatic conditions or genetic factors. We showed that the phenology of Madagascan wild coffees was not correlated to natural habitat type, but rather influenced both by climatic conditions and genetic parameters at the species level. The maturation phase was highly variable in length (from 59 to 360 d) and its phylogenetic signal was strong. We also found a strong phylogenetic signal in other traits that are rarely described (stem slenderness and the ratio between stem dry mass and leaf dry mass). It is also interesting to see that, even in these ex situ conditions, the climatic factors of the natural habitat still affect leaf and growth traits. Given these results, our study seems to also highlight the role of ex situ collections, which gather large numbers of accessions living under common conditions, not only as repositories of species diversity but also as potential sources of common garden experiments, offering an important opportunity to describe phenotypic and inter-annual variation.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: 28 traits with range of variation. Table S2: mean trait values for eight traits, shown for 15 species and 14 populations of five different species and their pooled standard error of the mean. Table S3: contribution of intraspecific diversity (different populations) to the variance of eight different traits.

mcaa073_suppl_Supplementary_Table_S1
Click here for additional data file. (2.7MB, docx)
mcaa073_suppl_Supplementary_Table_S2
Click here for additional data file. (32KB, docx)
mcaa073_suppl_Supplementary_Table_S3
Click here for additional data file. (14.6KB, docx)

ACKNOWLEDGEMENTS

This work is a part of a collaborative project with the National Center of Applied Research and Rural Development (FOhibem-pirenena momba ny FIkarohana ampiharina amin’ny Fampandrosoana ny eny Ambanivohitra, FOFIFA), in Madagascar and two French research institutes (Cirad and IRD, Montpellier, France). We thank the FOFIFA staff at the Kianjavato Coffee Research Station (KCRS), Madagascar. We thank Yves Caraglio (UMR AMAP) for his critical and valuable comments on the manuscript.

FUNDING

This work was supported by the Agropolis Foundation (grant 0902-009) and by funds from the Botany and Computational Plant Architecture joint research unit (UMR AMAP) and by funds from the Diversity, Adaption and Plant Development joint research unit (UMR DIADE).

LITERATURE CITED

  1. Akaffou DS, Konate I, Sié RS, et al.  2014. Flowering phenology and yield-related traits in the interspecific cross between Coffea pseudozanguebariae Bridson and C. canephora Pierre. Australian Journal of Crop Science  8: 1272–1280. [Google Scholar]
  2. Allan E, Panell JR. 2009. Rapid divergence in physiological and life-history traits between northern and southern populations of the British introduced neo-species, Senecio squalidus. Oikos  118: 1053–1061 [Google Scholar]
  3. Anderson JT, Inouye DW, McKinney AM, Colautti RI, Mitchell-Olds T. 2012. Phenotypic plasticity and adaptive evolution contribute to advancing flowering phenology in response to climate change. Proceedings. Biological Sciences  279: 3843–3852. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Andrianasolo D, Davis AP, Razafinarivo NJ, et al.  2013. High genetic diversity of in situ and ex situ populations of Madagascan coffee species: further implications for the management of coffee genetic resources. Genetics and Genomes  9: 1295–1312. [Google Scholar]
  5. Aarssen LW. 2005. Why don’t bigger plants have proportionally bigger seeds?  Oikos  111: 199–207. [Google Scholar]
  6. Bacon CD, Velásquez-Puentes F, Flórez-Rodríguez A, et al.  2016. Phylogenetics of Iriarteeae (Arecaceae), cross-Andean disjunctions and convergence of clustered infructescence morphology in Wettinia. Botanical Journal of the Linnean Society  182: 272–286. [Google Scholar]
  7. Barth S, Forneck A, Verzeletti F, Blaich R, Schumann F. 2009. Genotypes and phenotypes of an ex situ Vitis vinifera ssp. sylvestris (Gmel.) Beger germplasm collection from the Upper Rhine Valley. Genetic Resources and Crop Evolution  56: 1171–1181. [Google Scholar]
  8. Bates D, Mächler M, Bolker B, Walker S. 2015. Fitting linear mixed-effects models using Lme4. Journal of Statistical Software  67. [Google Scholar]
  9. Blomberg SP, Garland T Jr. 2002. Tempo and mode in evolution: phylogenetic inertia, adaptation and comparative methods. Journal of Evolutionary Biology  15: 899–910. [Google Scholar]
  10. Blomberg SP, Garland T Jr, Ives AR. 2003. Testing for phylogenetic signal in comparative data: behavioral traits are more labile. Evolution  57: 717–745. [DOI] [PubMed] [Google Scholar]
  11. Borianne P. 2012. Automated valuation of leaves area for large-scale analysis needing data coupling or petioles deletion. In: Guo Y, Kang MZ, Dumont Y, eds. Plant growth modeling, simulation, visualization and applications. Beijing: IEEE Press, 50–57. [Google Scholar]
  12. Carvalho A. 1952. Taxonomia de Coffea Arabica L. Caracteres morfologicos dos haploides. Bragantia  12: 201–212. [Google Scholar]
  13. Cavender-Bares J, Kozak KH, Fine PV, Kembel SW. 2009. The merging of community ecology and phylogenetic biology. Ecology Letters  12: 693–715. [DOI] [PubMed] [Google Scholar]
  14. Charrier A. 1978. La structure génétique des caféiers spontanés de la région malgache (Mascarocoffea). Leurs relations avec les caféiers d’origine africaine (Eucoffea). PhD Thesis, Université Paris-Sud Orsay, France. [Google Scholar]
  15. Chevalier A. 1942. Les caféiers du globe. Caféiers sauvages et cultivés. Encyclopédie biologique, XXII Fascicule II. Paris: Lechevalier, 1–33. [Google Scholar]
  16. Chuine I. 2010. Why does phenology drive species distribution?  Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences  365: 3149–3160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Cleland EE, Allen JM, Crimmins TM, et al.  2012. Phenological tracking enables positive species responses to climate change. Ecology  93: 1765–1771. [DOI] [PubMed] [Google Scholar]
  18. Corner EJH. 1949. The Durian theory or the origin of the modern tree. Annals of Botany  13: 367–414. [Google Scholar]
  19. Couturon E, Raharimalala NE, Rakotomalala JJ, et al.  2016. Caféiers sauvages – un trésor en péril au cœur des forêts tropicales! Wild coffee-trees – a threatened treasure in the heart of tropical forests!  Montpellier: Association Biodiversité, Ecovalorisation et Caféiers. [Google Scholar]
  20. Crisosto CH, Grantz DA, Meinzer FC. 1992. Effects of water deficit on flower opening in coffee (Coffea arabica L.). Tree Physiology  10: 127–139. [DOI] [PubMed] [Google Scholar]
  21. Davies TJ, Wolkovich EM, Kraft NJB, et al.  2013. Phylogenetic conservatism in plant phenology. Journal of Ecology  101: 1520–1530. [Google Scholar]
  22. Davis AP, Rakotonasolo F. 2008. A taxonomic revision of the baracoffea alliance: nine remarkable Coffea species from western Madagascar. Botanical Journal of the Linnean Society  158: 355390. [Google Scholar]
  23. Davis AP, Govaerts R, Bridson DM, Stoffelen P. 2006. An annotated taxonomic conspectus of the genus Coffea (Rubiaceae). Botanical Journal of the Linnean Society  152: 465–512. [Google Scholar]
  24. Ensslin A, Godefroid S. 2019. How the cultivation of wild plants in botanic gardens can change their genetic and phenotypic status and what this means for their conservation value. Sibbaldia  17: 51–69. [Google Scholar]
  25. Enßlin A, Sandner TM, Matthies D. 2011. Consequences of ex situ cultivation of plants: genetic diversity, fitness and adaptation of the monocarpic Cynoglossum officinale L. in botanic gardens. Biological Conservation  144: 272–278. [Google Scholar]
  26. Flores O, Garnier E, Wright IJ, et al.  2014. An evolutionary perspective on leaf economics: phylogenetics of leaf mass per area in vascular plants. Ecology and Evolution  4: 2799–2811. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Fort F, Jouany C, Cruz P. 2015. Hierarchical traits distances explain grassland Fabaceae species’ ecological niches distances. Frontiers in Plant Science  6: 1–11 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Fortunel C, Stahl C, Heuret P, Nicolini E, Baraloto C. 2020. Disentangling the effect of environment and ontogeny on tree functional dimensions for congeneric species in tropical forest. New Phytologist  226: 385–395. [DOI] [PubMed] [Google Scholar]
  29. Garnier E, Navas ML. 2012. A trait-based approach to comparative functional plant ecology: concepts, methods and application for agroecology. A review. Agronomy for Sustainable Development  32: 365–399. [Google Scholar]
  30. Givnish TJ. 1987. Comparative studies of leaf form assessing the relative roles of selective pressures and phylogenetic constraints. New Phytologist  106: 131–160. [Google Scholar]
  31. Gomez C, Despinoy M, Hamon S, et al.  2016. Shift in precipitation regime promotes interspecific hybridization of introduced Coffea species. Ecology and Evolution  6: 3240– 3255. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Hallé F, Oldeman RAA, Tomlinson PB. 1978. Tropical trees and forests: an architectural analysis. New York: Springer. [Google Scholar]
  33. Hamon P, Rakotomalala JJ, Akaffou S, et al.  2015. Caffeine-free species in the genus Coffea. In: Preedy V, ed. Coffee in health and disease prevention. San Diego: Academic Press, 39–44. [Google Scholar]
  34. Hamon P, Grover CE, Davis AP, et al.  2017. Genotyping-by-sequencing provides the first well-resolved phylogeny for coffee (Coffea) and insights into the evolution of caffeine content in its species. Molecular Phylogenetics and Evolution  109: 351–361. [DOI] [PubMed] [Google Scholar]
  35. Hey J. 2006. Recent advances in assessing gene flow between diverging populations and species. Current Opinion in Genetics & Development  16: 592–596. [DOI] [PubMed] [Google Scholar]
  36. Janssen T, Bystriakova N, Rakotondrainibe F, Coomes D, Labat JN, Schneider H. 2008. Neoendemism in Madagascan scaly tree ferns results from recent, coincident diversification bursts. Evolution  62: 1876–1889. [DOI] [PubMed] [Google Scholar]
  37. Kazakou E, Violle C, Roumet C, et al.  2014. Are trait-based species rankings consistent across data set and spatial scales?  Journal of Vegetation Science  25: 235–247. [Google Scholar]
  38. Kleiman D, Aarssen LW. 2007. The leaf size/number trade-off in trees. Journal of Ecology  95: 376–382. [Google Scholar]
  39. Krishnan S, Ranker TA, Davis AP, Rakotomalala JJ. 2013. The study of genetic diversity patterns of Coffea commersoniana, an endangered coffee species from Madagascar: a model for conservation of other littoral forest species. Tree Genetics & Genomes  9: 179–187. [Google Scholar]
  40. Lashermes P, Combes MC, Robert J, et al.  1999. Molecular characterisation and origin of the Coffea arabica L. genome. Molecular & General Genetics  261: 259–266. [DOI] [PubMed] [Google Scholar]
  41. Lauri PE. 2019. Corner’s rules as a framework for morphology, architecture and functioning issues and steps forward. New Phytologist  221: 1679–1684. [DOI] [PubMed] [Google Scholar]
  42. Lauri PÉ, Normand F. 2017. Are leaves only involved in flowering? Bridging the gap between structural botany and functional morphology. Tree Physiology  37: 1137–1139. [DOI] [PubMed] [Google Scholar]
  43. Lin BB. 2008. Microclimate effects on flowering success in coffee agroforestry systems. American-Eurasian Journal of Agriculture and Environmental Science  3: 148–152. [Google Scholar]
  44. Louarn J. 1992. La fertilité des hybrides interspécifiques et les relations génomiques entre caféiers diploïdes d’origine africaine (Genre Coffea L. sous-genre Coffea). PhD Thesis, University of Montpellier,  France. [Google Scholar]
  45. Luna-Nieves AL, Meave JA, Morellato LPC, Ibarra-Manriquez, 2017. Reproductive phenology of useful seasonally dry tropical forest trees: guiding patterns for seed collection and plant propagation in nurseries. Forest Ecology and Management  393: 52–62 [Google Scholar]
  46. Maurin O, Davis AP, Chester M, Mvungi EF, Jaufeerally-Fakim Y, Fay MF. 2007. Towards a phylogeny for Coffea (Rubiaceae): identifying well-supported lineages based on nuclear and plastid DNA sequences. Annals of Botany  100: 1565–1583. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Migicovsky Z, Warschefsky E, Klein LL, Miller AJ. 2019. Using living germplasm collections to characterize, improve, and conserve woody perennials. Crop Science  59: 2365–2380. [Google Scholar]
  48. Myers N, Mittermeier RA, Mittermeier CG, da Fonseca GA, Kent J. 2000. Biodiversity hotspots for conservation priorities. Nature  403: 853–858. [DOI] [PubMed] [Google Scholar]
  49. Niemiller ML, Fitzpatrick BM, Miller BT. 2008. Recent divergence with gene flow in Tennessee cave salamanders (Plethodontidae: Gyrinophilus) inferred from gene genealogies. Molecular Ecology  17: 2258–2275. [DOI] [PubMed] [Google Scholar]
  50. Niinemets U, 2001. Globe-scale climatic controls of leaf dry mass per area, density, and thickness in trees and shrubs. Ecology  82: 453–469. [Google Scholar]
  51. Noirot M, Charrier A, Stoffelen P, Anthony F. 2016. Reproductive isolation, gene flow and speciation in the former Coffea subgenus: a review. Trees  30: 597–608. [Google Scholar]
  52. Nosil P. 2008. Speciation with gene flow could be common. Molecular Ecology  17: 2103–2106. [DOI] [PubMed] [Google Scholar]
  53. Orme D, Freckleton RP, Thomas GH, et al.  2012. caper: comparative analysis of phylogenetics and evolution in R. Methods in Ecology and Evolution  3: 145–151. [Google Scholar]
  54. Pagel M. 1999. Inferring the historical patterns of biological evolution. Nature  401: 877–884. [DOI] [PubMed] [Google Scholar]
  55. Perrier X, Jacquemoud-Collet J-P. 2006. DARwin software http://darwin.cirad.fr
  56. Le Pierrès D. 1995. Etude des hybrides interspécifiques tétraploïdes de première génération entre Coffea arabica L. et les caféiers diploïdes. PhD Thesis, University Paris XI, France. [Google Scholar]
  57. Poorter L, Castilho CV, Schietti J, Oliveira RS, Costa FRC. 2018. Can traits predict individual growth performance? A test in a hyperdiverse tropical forest. New Phytologist  219: 109–121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Portères R. 1946. Action de l’eau, après une période sèche, sur le déclenchement de la floraison chez Coffea arabica L. L’Agronomie Tropicale  1: 148–158. [Google Scholar]
  59. Purschke O, Schmid BC, Sykes MT, et al.  2013. Contrasting changes in taxonomic, phylogenetic and functional diversity during a long-term succession: insights into assembly processes. Journal of Ecology  101: 857–856. [Google Scholar]
  60. Rauschkolb R, Szczeparska L, Kehl A, Bossdorf O, Scheepens JF. 2019. Plant populations of three threatened species experience rapid evolution under ex situ cultivation. Biodiversity and Conservation  28: 3951–3969. [Google Scholar]
  61. Razafinarivo NJ, Rakotomala JJ, Brown SC, et al.  2012. Geographical gradients in the genome size variation of coffee trees (Coffea) native to Africa and Indian Ocean islands. Tree Genetics & Genomes  8: 1345–1358. [Google Scholar]
  62. Reginato M, Michelangeli FA. 2016. Diversity and constraints in the floral morphological evolution of Leandra s.str. (Melastomataceae). Annals of Botany  118: 445–458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Revell LJ. 2012. phytools: an R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution  3: 217–223. [Google Scholar]
  64. Sakai S. 2001. Phenological diversity in tropical forests. Population Ecology  43: 77–86. [Google Scholar]
  65. Shaffer HB, Thomson RC. 2007. Delimiting species in recent radiations. Systematic Biology  56: 896–906. [DOI] [PubMed] [Google Scholar]
  66. Smith JM, Burian R, Kauffman P, et al.  1985. Developmental constraints and evolution. Quarterly Review of Biology  60: 265–287. [Google Scholar]
  67. Varossieau WW. 1940. On the development of the stem and the formation of leaves in Coffea species. Annales du Jardin Botanique de Buitenzorg  50: 115–198. [Google Scholar]
  68. Vences M, Wollenberg KC, Vieites DR, Lees DC. 2009. Madagascar as a model region of species diversification. Trends in Ecology & Evolution  24: 456–465. [DOI] [PubMed] [Google Scholar]
  69. Westoby M, Falster DS, Moles TA, Vesk PA, Wright IL. 2002. Plant ecological strategies: some leading dimension of variation between species. Annual Review of Ecology and Systematics  33: 125–159. [Google Scholar]
  70. Wright IJ, Dong N, Maire V, et al.  2017. Global climatic drivers of leaf size. Science  357: 917–921. [DOI] [PubMed] [Google Scholar]
  71. Yoder AD, Nowak MD. 2006. Has vicariance or dispersal been the predominant biogeographic force in Madagascar? Only time will tell. Annual Review of Ecology Evolution, and Systematics  37: 405–431. [Google Scholar]
  72. Zalamea PC, Munoz F, Stevenson PR, et al.  2011. Continental-scale patterns of Cecropia reproductive phenology: evidence from herbarium specimens. Proceedings. Biological Sciences  278: 2437–2445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Zeileis A, Hothorn T. 2002. Diagnostic checking in regression relationships. R News  2: 7–10. [Google Scholar]
  74. Zimmerman JK, Wright SJ, Calderon O, Pagan MA, Paton S, 2007. Flowering and fruiting phenologies of seasonal and a seasonal neotropical forest: the role of annual changes in irradiance. Journal of Tropical Ecology  23: 231 [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

mcaa073_suppl_Supplementary_Table_S1
Click here for additional data file. (2.7MB, docx)
mcaa073_suppl_Supplementary_Table_S2
Click here for additional data file. (32KB, docx)
mcaa073_suppl_Supplementary_Table_S3
Click here for additional data file. (14.6KB, docx)

Articles from Annals of Botany are provided here courtesy of Oxford University Press

RESOURCES