Reproductive patterns, genetic diversity and inbreeding depression in two closely related Jumellea species with contrasting patterns of commonness and distribution

Abstract

Backgrounds and Aims Theory predicts that the long-term persistence of plant populations exposed to size reduction can be threatened by a loss of genetic diversity and increased inbreeding. However, several life-history and ecological traits can influence the response to population size reduction. The reproductive patterns, levels of genetic diversity and magnitude of inbreeding depression of the rare and fragmented Jumellea fragrans and of its widespread congener J. rossii were studied. The aim was to evaluate the effects of over-collection and fragmentation on J. fragrans and to enhance our knowledge of the biology and ecology of the two species, used for their aromatic and medicinal properties on Réunion.

Methods Hand pollination experiments were conducted to determine the breeding system and to evaluate the potential for inbreeding depression in both species. Nuclear microsatellite markers were used to investigate selfing rates and levels of genetic diversity.

Key Results Jumellea rossii revealed a mixed-mating system, and inbreeding depression at the germination stage (δ = 0·66). Levels of genetic diversity were relatively high [allelic richness (A_R) = 8·575 and expected heterozygosity (H_e) = 0·673]. In J. fragrans, selfing rates suggest a mainly outcrossing mating system. Genetic diversity was lower than in J. rossii, but not yet critically low (A_R = 4·983 and H_e = 0·492), probably because of the mainly outcrossing mating system and the relatively high density of individuals in the studied population. Jumellea fragrans did not show inbreeding depression, and it is hypothesized that the population had progressively purged its genetic load during successive fragmentation events.

Conclusions Even if the persistence of the J. fragrans population is not threatened in the short term, its genetic diversity has probably been reduced by fragmentation and over-collection. In situ conservation actions for J. fragrans and ex situ cultivation of both species are recommended in order to meet the demand of local people.

INTRODUCTION

Many oceanic islands possess extraordinarily high levels of plant biodiversity and endemism relative to their surface area, and for this reason they are of great interest for biological and ecological studies (Groombridge, 1992). However, oceanic islands also harbour much of the world’s threatened biodiversity, mainly due to habitat degradation and fragmentation (Caujapé-Castells et al., 2010). Habitat degradation and fragmentation reduce plant abundance and density, resulting in a decrease in population sizes and an increase in spatial isolation between populations. Theoretically, these physical changes are thought to affect reproduction, gene flow and genetic diversity (Young et al., 1996; Aguilar et al., 2008) and are likely to have a negative impact on the persistence and evolutionary potential of the concerned plant populations in the face of environmental change (Young et al., 1996).

Plants of aromatic, medicinal or horticultural interest are at additional risk, because they also frequently face over-collection, which contributes to population size reduction. The magnitude of the impact of unregulated harvesting on plant populations depends on the part of the plant collected, the plant life form and its range and distribution (Sheldon et al., 1997), but is considered as one cause of mortality which directly affects the population vital rates and may increase the extinction risks (Mondragón, 2009). According to Vance (2002), populations of harvested plants may become extinct more rapidly due to over-collection than they do from fragmentation and habitat destruction alone. However, it has become clear that the response to population size reduction caused by habitat fragmentation and over-collection is likely to be species specific and depends on particular life-history traits (Young et al., 1996).

The breeding system, and especially the compatibility system, is one of the main attributes that can influence the degree of susceptibility to fragmentation and over-collection (Aizen et al., 2002; Aguilar et al., 2006). Plant breeding systems range from obligate cross-fertilization (self-incompatible species) to pollinator-mediated or obligate self-fertilization (self-compatible species) (Lloyd and Schoen, 1992). Whereas reproduction in obligate outbreeders will require the presence of other mates and a pollen vector, reproduction of inbreeders will have a lower dependence on the presence of mates and, in the case of autopollinated species (sensu Catling, 1990), of pollen vectors. In this regard, self-incompatible species are considered to be more vulnerable to population size reduction compared with self-compatible or autopollinated species, because they can experience reproductive loss caused by the scarcity of mates and/or by low pollinator visitation rates (Aguilar et al., 2006). Likewise, the mating system (i.e. the average proportion of cross- vs. self-fertilization), because it determines the spatial distribution of genetic variation within and among populations, may influence the effects of population size reduction on genetic diversity (Loveless and Hamrick, 1984). Habitat fragmentation and individual destruction affect mating by reducing the availability of mates and the genetic diversity of the remaining population via a genetic bottleneck. This can reduce the opportunities for outcrossing and increase inbreeding through self-fertilization or mating between related individuals (e.g. biparental inbreeding) (Ellstrand and Elam, 1993). An increase in selfing rates may result in an increase of homozygosity and in the expression of inbreeding depression (i.e. the relative reduction of fitness of selfed vs. outcrossed progeny) due to the homozygous expression of deleterious recessive alleles (Charlesworth and Charlesworth, 1987; Carr and Dudash, 2003), reducing progeny fitness and survival (Husband and Schemske, 1996; González-Varo et al., 2010). Inbreeding depression is an important parameter to consider in a conservation context, because it is one of the most important threats to the persistence of rare or fragmented taxa (Hedrick and Kalinowski, 2000). Also, the estimation of its magnitude via manual crossing is a likely predictor of persistence of species experiencing population size reduction (Barrett and Kohn, 1991). Finally, some ecological processes such as initial distribution and abundance of a species, initial size and density of populations and longevity of a species can also influence the degree of susceptibility to fragmentation and over-collection (Young et al., 1996).

In the context of worldwide biodiversity loss, essentially due to habitat destruction and exploitation, the establishment of sustainable management programmes is needed. To be efficient, the management of threatened plant species has to deal with a range of different tools, including ecological and genetic studies of plants, and requires an understanding of their life-history characteristics. Among these life-history traits, breeding system, mating system and inbreeding depression are key components, because they are determinant factors for abundance, distribution, genetic diversity and, therefore, for evolutionary dynamic and long-term viability of populations in a context of population size reduction (Barrett and Harder, 1996; Hamrick and Godt, 1996; Husband and Schemske, 1996; Duminil et al., 2007; Castro et al., 2008). Moreover, these data are of great importance for plant reintroduction and plant production in cases where cultivation can be a solution permitting both the reduction of collecting pressure on natural populations and the meeting of the demand of the plant of interest (Sarasan et al., 2011).

Orchidaceae are a family of considerable economic importance, particularly in horticulture and floristry, but numerous species are also used in traditional herbal medicine (Mahendran and Bai, 2009). However, the use of these plants often remains unregulated and results in over-collection. This massive extraction of plants from their natural environment, together with the destruction, modification and fragmentation of their habitats make Orchidaceae one of the plant families with the highest proportion of endangered species (Salazar, 1996)

Jumellea rossii and Jumellea fragrans are sister orchid species endemic to the Mascarene archipelago (Rakotoarivelo et al., 2012). The species are close morphologically, but differ in their habitat and pollination syndrome (Mallet et al., 2012). Traditionally, leaves of both species are harvested for their aromatic and medicinal properties (Dorvault, 1886; Lavergne, 1990; Shum Cheong Sing and Smadja, 1992). Jumellea rossii still occurs in large populations and across a wide habitat range (Mallet, 2014), whereas J. fragrans populations have been severely reduced due not only to excessive harvesting, but also to the destruction of its habitat. In this context, both species, and particularly J. fragrans, present strong conservation issues. In order to decrease collection pressure and to strengthen the existing populations of these aromatic and medicinal species, a global programme of conservation and promotion has been initiated (Mallet, 2014).

In the present study, we assess and compare the reproductive patterns, levels of genetic diversity and levels of inbreeding depression of two sister orchid species differing in their commonness. More precisely in one population of J. rossii and one population of J. fragrans, we aim to: (a) determine the breeding systems by a hand pollination experiment; (b) estimate the mating systems and levels of genetic diversity with nuclear microsatellite markers; (c) quantify pollen removal and deposition rates; and (d) measure the potential for inbreeding depression. These data will enhance our knowledge of the biology and ecology of these two species and will also help us to evaluate the eventual effects of over-collection and fragmentation on J. fragrans.

MATERIALS AND METHODS

Study site and species

Réunion (55°39'E, 21°00'S) is a small, young oceanic island (about two million years) in the Mascarene archipelago (Indian Ocean). It is a volcanic island dominated by two volcanos: the Piton de la Fournaise that is still active (2619 m) and the Piton des Neiges, the highest point of the island (3071 m). This complex topography results in the existence of a wide variety of habitats (Strasberg et al., 2005). Four major habitat types can be distinguished on Réunion (Strasberg et al., 2005): lowland (<500 m a.s.l.), sub-mountain (between 500 and 1000 m a.s.l.), mountain (between 1000 and 1800 m a.s.l.) and sub-alpine habitats (between 1800 and 3069 m a.s.l.). The native flora includes about 150 orchid species in 30 genera, including 25 % that are endemic to Réunion, occurring in all habitat types.

This study was focused on two epiphytic orchid species, Jumellea fragrans and J. rossii, which are sister species endemic to the Mascarene Islands (Micheneau et al., 2008a; Rakotoarivelo et al., 2012). These species differ in their commonness, distribution and pollination syndromes. Jumellea rossii is relatively common in the wet forests of Réunion between 500 and 1800 m a.s.l. Plants are characterized by clumps of stems producing 1–5 creamy white flowers, which are odoriferous at dusk. The flowers have a short 16–20 mm nectariferous spur. The pollinarium consists of two yellow and small pollinia, attached to a one-piece viscidium, and is separated from the stigma by the rostellum (Fig. 1). Pollination is effected when at least one pollinium is deposited in the stigmatic cavity of the column. The flowering period lasts from December to March depending on altitude. Jumellea rossii is mainly pollinated at night by species of Noctuidae (Mallet, 2014). Jumellea fragrans is essentially found in the remaining lowland rain forest at < 500 m a.s.l. and, because of the considerable loss of this habitat due to human activities, has a fragmented and limited range. The species is morphologically similar to J. rossii, but has larger leaves and flowers, and generally blooms in March–April. The flowers have a longer spur than J. rossii (approx. 39 mm), and are pollinated by a hawk moth (Sphingidae; Mallet, 2014). The study was conducted at two different localities on Réunion: in the mountain windward rain forest of Bélouve for J. rossii and in the lowland rain forest of Mare Longue for J. fragrans.

Fig. 1.

Floral morphology of Jumellea fragrans. ds, dorsal sepal; pt, petal; ls, lateral sepal; lb, labellum; ov, ovary; pl, pollinarium; rs, rostellum; sc, stigmatic cavity; sp, spur. Scale bar = 1 cm.

Compatibility system determination and inbreeding depression estimation

Crossing treatments

To characterize the breeding system of J. fragrans and J. rossii, hand pollination experiments were conducted during two consecutive flowering seasons (2011/2012 and 2012/2013). In both years, at each site, between eight and 22 plants per species were selected on the basis of having several unopened buds and/or fresh unpollinated flowers. Unpollinated flowers and unopened buds were enclosed in fine-mesh cotton bags, to prevent future pollinator visitation. On each plant, less than half of the opened flowers were used to apply one or both of the following treatments (for a total of about 30 flowers per treatment). (a) Cross-pollination: flowers were pollinated with pollinia from a different plant, situated >5 m away. (b) Self-pollination: flowers were pollinated with their own pollinia. Additional plants with unmanipulated flowers were also enclosed, and were used to test for autopollination. At fruit maturity (just before dehiscence), fruit set was recorded, and the capsules produced were collected. For each pollination treatment and each species, the mean individual fruit set was calculated according to eqn (1) (Micheneau et al., 2008c):

$$P=\left[\left(1/N\right)\times {\displaystyle \sum _{i=1}^N{\left(n_{\text{p}}/n_{\text{t}}\right)}_i}\right]\times 100$$ (1)

with P = mean percentage of fruit set, N = number of plants, n_p = number of hand-pollinated flowers setting fruit and n_t = total number of hand-pollinated flowers. Standard errors were also calculated.

The capsules were weighed (to the nearest 0·01 g) and measured (length and width, to the nearest 0·1 mm). Mean percentages of fruit set and mean fruit size were compared between pollination treatments and species using Mann–Whitney tests (Sokal and Rohlf, 1995). All the statistics were computed on R 2.15.1 (R Development Core Team, 2012).

Embryo viability and germination procedures

The viability of the embryos produced by cross- and self-pollination was investigated for both species. Samples of 5 mg of seeds per fruit were evaluated with a 2,3,5-triphenyl tetrazolium chloride (TTC) viability test (Lakon, 1949). Seed samples were pre-conditioned by placing in 1·5 mL microtubes filled with a 10 % (w/v) sucrose solution. After incubation for 24 h at room temperature, sucrose solution was removed, seeds were washed twice with distilled water and a solution of 1 % TTC was added. Seeds were incubated in the dark at 40 °C for 48 h. A minimum of 100 seeds was counted per sample under a light microscope at × 100 magnification. Embryos with any degree of red or pink coloration were scored as viable, and unstained embryos were scored as non-viable. The percentage of viable seeds for each treatment and each species was calculated using eqn (1), with P = mean percentage of viable seeds, N = number of capsules, n_m = number of viable seeds in the sample and n_t = total number of seeds in the sample.

To test whether pollination mode influenced seed germination rates, seeds were sterilized in a 3·4 % solution of H₂O₂ for 20 min and then sown in aseptic conditions on Petri dishes containing half-strength Murashige and Skoog basal medium (Murashige and Skoog, 1962) supplemented with 0·1 mm of the phytohormone 6-benzylaminopurine (6-BAP). The cultures were incubated at 25 °C with a 16 h photoperiod. The percentage of seed germination was scored after 11 weeks of culture and calculated for each treatment and each species according to eqn (1), with P = mean percentage of germinated seeds, N = number of capsules, n_m = number of germinated seeds per capsule and n_t = total number of seeds sown per capsule. Mean percentages of viable and germinated seeds were compared between pollination mode and species using Mann–Whitney tests (Sokal and Rohlf, 1995).

Potential for inbreeding depression

Inbreeding depression (δ) was calculated on the following variables: fruit production, fruit weight, length and width, seed viability rate and germination rates according to eqn (2) (Charlesworth and Charlesworth, 1987) :

$$\delta =\left(p_{\text{cross}}–p_{\text{self}}\right)/p_{\text{cross}}$$ (2)

where p_cross and p_self represent the performance of cross- and self-pollination treatments for each variable measured, and a value of 0·5 is generally interpreted as evidence of a mixed-mating system with intermediate levels of inbreeding depression (Wallace, 2003).

Pollen deposition vs. removal rates

During three flowering seasons (2009/2010, 2011/2012 and 2012/2013), 22–74 individuals per species were examined at least once a week in the two study sites to determine male and female pollination success (removal and deposition of pollinia, respectively). The pollen removal rate for each species was calculated according to eqn (1), with P = mean percentage of pollen removal, N = number of plants, n_m = number of flowers with removed pollinia and n_t = total number of flower per plant. The pollen deposition rate was calculated following the same formula. Pollination success variations between years and species were analysed using non-parametric Kruskal–Wallis χ² approximation and Mann–Whitney tests (Sokal and Rohlf, 1995). The analyses were computed on R 2.15.1 (R Development Core Team, 2012).

Genetic diversity and estimated selfing rates

The two populations studied were sampled between January and July 2012 on Réunion (sampling permit code from the Parc National de La Réunion: DIR/I/2012/002). Collected leaves were dried in silica gel and DNA was extracted using a DNeasy® Plant mini kit (Qiagen, Hilden, Germany). Individuals were genotyped for 12 nuclear microsatellite loci (P2G7, P1A9, P2E3, P2H10, P2G6, P2G11, P2G2, P2E12, P2D1, P1B10, P2E2 and P2G4) isolated from J. rossii and successfully amplified in J. fragrans (Molecular Ecology Resources Primer Development Consortium et al., 2010). PCR multiplexes, electrophoresis and allele size determination were performed as described by Mallet et al. (2014).

FreeNA (Chapuis and Estoup, 2007) was used to estimate null allele frequencies for each locus in each population, according to the expectation maximization (EM) algorithm of Dempster et al. (1977). The mean observed number of alleles per locus (A_L) and the mean number of rare alleles per locus (R_A, frequency <0·05) for each population were computed using GenAlex 6.5 (Peakall and Smouse, 2006, 2012), and allelic richness (A_R; El Mousadik and Petit, 1996) was calculated as implemented in the software FSTAT 2.9.3 (Goudet, 2001) based on 37 diploid individuals. Allelic richness can be used to make comparisons of the mean number of alleles among species with different sampling size. The observed heterozygosity across all loci (H_o), the expected heterozygosity across all loci (H_e) and the fixation index (F_IS), estimated through the fixation index of Weir and Cockerham (Hedrick, 2005), were calculated using Genepop 4·0 (Raymond and Rousset, 1995). Comparisons of genetic diversity indexes between the two populations studied were assessed using Wilcoxon’s signed-ranks tests (Sokal and Rohlf, 1995). Departure from Hardy–Weinberg equilibrium (HWE) was tested using exact tests of Guo and Thompson (1992), based on Markov chain iteration (1000 iterations) using Genepop 4·0 (Raymond and Rousset, 1995).

To estimate selfing rates in each population, we used three complementary methods. The first method is based on the fixation index (F_IS), using eqn (3) (Ritland, 1990):

$$S_{\left(F\right)}=\text{2}{F}_{\text{IS}}/\left(\text{1}+F_{\text{IS}}\right)$$ (3)

We used two other estimation of selfing rates: these methods are multilocus methods implemented in the software RMES (robust multilocus estimates of selfing; David et al., 2007). These methods are not biased by the presence of null alleles and mis-scorings that may occur when estimates of selfing are based on F_IS. The first method, S_(g2), is a point estimation of the two-locus heterozygosity disequilibrium and the second, S_(ML), is based on maximum likelihood of the whole distribution (David et al., 2007).

RESULTS

Compatibility system and inbreeding depression

Jumellea rossii set similar numbers of fruits in the self- and cross-pollination treatments (Table 1, P = 0·71, Mann–Whitney test), and none of the flowers tested for autonomous self-pollination produced fruit. The same pattern was observed for J. fragrans (P = 0·86, Mann–Whitney test). These results clearly indicate that both species are self-compatible, but are fully dependent on a pollen vector to set fruit.

Table 1.

Mean ± s.e. of fruit set, fruit and seed quality for plants of Jumellea fragrans and J. rossii that were cross- or self-pollinated

Character	J. fragrans				P-value	J. rossii				P-value
	Cross		Self			Cross		Self
	n	Mean ± s.e.	n	Mean ± s.e.		n	Mean ± s.e.	n	Mean ± s.e.
Fruit set (%)	35	46·90 ± 7·90	33	49·98 ± 7·87	0·86	18	72·80 ± 9·86	17	83·65 ± 6·60	0·71
Fruit weight (g)	7	0·80 ± 0·10	8	0·80 ± 0·10	1·00	8	1·17 ± 0·13	8	1·12 ± 0·13	0·80
Fruit length (mm)	7	57·90 ± 3·30	8	57·27 ± 2·81	0·87	8	59·06 ± 2·85	8	58·07 ± 2·68	0·57
Fruit width (mm)	7	4·55 ± 0·20	8	4·56 ± 0·26	0·73	8	5·76 ± 0·27	8	5·58 ± 0·24	0·57
Viable seeds (%)	11	44·49 ± 11·28	8	34·78 ± 11·51	0·72	9	77·77 ± 2·70	12	67·77 ± 4·04	0·13
Germination (%)	8	48·89 ± 10·94	6	34·75 ± 12·05	0·49	7	15·34 ± 6·62	9	5·19 ± 0·93	0·03*

The significance of Mann–Whitney tests between pollination mode for each species are given: *P < 0·05.

Within each species, no significant difference in fruit set, fruit length, fruit width, fruit weight or embryo viability was observed between the outcrossed and self-treatments (Table 1). Consequently, the inbreeding depression coefficient was not significant for these five measured stages (Fig. 2). However, for J. rossii, the germination percentage was significantly higher (P = 0·03, Mann–Whitney test) for seeds from cross-pollination (15·34 ± 6·62 %) than for seeds from self-pollination (5·19 ± 0·93 %), resulting in a significant inbreeding depression coefficient of 0·66 (Fig. 2).

Fig. 2.

Inbreeding depression (δ) magnitude for each stage recorded in J. fragrans and J. rossii. The bars give the value of δ, with above the significance found from associated crossed and selfed comparison (Mann–Whitney tests; *P < 0·05, **P < 0·01, ***P < 0·001). At each stage, the significance for species comparison is indicated across the top of the graph (Mann–Whitney tests; n.s., not significant, P-values as for individual bars). Stages are indicated as follows: fset, fruit set; wght, fruit weight; lgth, fruit length; wdth, fruit width; viab; seed viability; germ, seed germination.

Jumellea fragrans and J. rossii differed for all the measured variables, except for fruit length (Table 1; Fig. 2): J. rossii had higher fruit set, heavier and wider fruits and more viable seeds than J. fragrans. In contrast, germination rates were significantly higher for J. fragrans than for J. rossii.

Pollen deposition vs. removal rates

Pollen removal and pollen deposition percentages showed variation between years for J. rossii (Fig. 3, Table 2). This variation was mainly caused by the 2009 flowering season in which the removal rate was low and no deposition was observed in the population. For J. fragrans, only pollen removal percentages showed variation between years (P = 0·004, Kruskall–Wallis test), ranging from 14·23 ± 4·49 % to 42·52 ± 6·50 % over the 3 years studied (Fig. 3;, Table 2).

Fig. 3.

Mean ± s.e. pollen removal (male reproductive success) and pollen deposition (female reproductive success) in J. fragrans (A) and J. rossii (B) for the 3 years studied. For each year, the numbers of flowers and individuals studied are given in parentheses, and the significance for removal vs. deposition comparisons is indicated accross the top of the graph (Mann–Whitney tests; n.s., not significant, *P < 0·05, **P < 0·01, ***P < 0·001).

Table 2.

χ² values from Kruskal–Wallis tests for pollen removal and deposition in Jumellea fragrans and J. rossii for the 3 years studied

Factors		χ2 values
	Variables	Pollen removal		Pollen deposition
J. fragrans	Years	11·23	**	3·39	n.s.
J. rossii	Years	15·65	***	20·6	***
2009	Species	7·73	**	12·52	***
2011	Species	3·72	n.s.	0·01	n.s.
2012	Species	2·03	n.s.	0·56	n.s.

n.s., not significant; **P < 0·01; ***P < 0·001.

Pollen removal and pollen deposition rates within a year showed no significant variation between species, except for 2009 when removal and deposition rates for J. rossii were dramatically low (P = 0·005 for pollen removal, P < 0·001 for pollen deposition, Kruskal–Wallis tests; Table 2). Pollen removal was always significantly higher than pollen deposition, except in 2011 for J. fragrans when no significant difference was observed (P = 0·19, Kruskal–Wallis test; Fig. 3; Table 2).

Genetic diversity and estimated selfing rates

All 12 microsatellite loci were polymorphic for the two species. The average frequency of null alleles was 0·19 ± 0·07 for J. rossii and 0·11 ± 0·09 for J. fragrans. The level and pattern of estimated genetic variation differed substantially between the two studied populations. Jumellea rossii had a higher genetic diversity than J. fragrans with a higher mean number of alleles per locus (P = 0·009, Wilcoxon’s signed-ranks test; Table 3) and allelic richness (P = 0·021, Wilcoxon’s signed-ranks test; Table 3). Jumellea rossii also had a higher mean number or rare alleles per locus than J. fragrans (P = 0·012, Wilcoxon’s signed-ranks test; Table 3). For both species, mean observed heterozygosity was lower than mean expected heterozygosity, suggesting heterozygote deficiency. F_IS estimates were 0·455 for J. rossii and 0·303 for J. fragrans, and exact tests showed a significant deviation from HWE due to the heterozygote deficiency in both populations (Table 3).

Table 3.

Genetic diversity for 12 nuclear microsatellites in Jumellea fragrans and J. rossii

Parameter	J. fragrans	J. rossii	P-value
n	37·75 ± 0·45	58·75 ± 2·14	–
AL	5·17 ± 2·62	9·25 ± 3·67	0·009**
AR	4·98 ± 2·61	8·57 ± 3·17	0·021*
RA	2·00 ± 2·00	4·67 ± 2·84	0·012*
Ho	0·350 ± 0·217	0·372 ± 0·178	0·791
He	0·492 ± 0·193	0·673 ± 0·217	0·042*
FIS	0·303	0·455	–

Values are given as the mean ± s.d.

n, sample sizes; A_L, alleles per locus; A_R, allelic richness; R_A, number of rare alleles per locus; H_o, observed heterozygosity; H_e, expected heterozygosity; F_IS. fixation index.

The significance of Wilcoxon’s signed-ranks tests between the two species is given: *P < 0·05, **P < 0·01.

Whatever the method, J. rossii expressed high and significant estimates of selfing rates ranging from 0·39 to 0·63 (Table 4). For J. fragrans, estimates of selfing rates were lower than for J. rossii and ranged from 0·16 to 0·47 (Table 4). Only the F_IS and the g2 methods gave significant selfing rate estimations.

Table 4.

Estimated selfing rates of Jumellea fragrans and J. rossii by F_IS, g2 and maximum likelihood (ML) methods

Species	n	FIS method		g2 method		ML method
		S(F)	P(F)	S(g2)	P(g2)	S(ML)	P(ML)
J. fragrans	38	0·47	***	0·31	**	0·16	n.s.
J. rossii	61	0·63	***	0·51	***	0·39	***

n, sample sizes; S_(F), selfing rates estimated with the F_IS method; P_(F), significance of S_(F) based on Guo and Thompson’s exact tests (Guo and Thompson, 1992) testing departure from Hardy–Weinberg equilibrium; S_(g2), selfing rates estimated using the g2 method (David et al., 2007); P_(g2), significance of the P-value for the test of H_o: S_(g2) = 0. S_(ML), selfing rates estimated with the ML method (David et al., 2007); P_(ML), significance of the P-value for the test of H_o: S_(ML) = 0.

n.s., not significant; **P < 0·01; ***P < 0·001.

DISCUSSION

Compatibility system

The hand pollination experiment performed to test the compatibility system of the two species showed that both species are fully self-compatible. This is the case for most other species of sub-tribe Angraecinae for which breeding systems have been investigated (Nilsson et al., 1985, 1987; Nilsson and Rabakonandrianina, 1988; Martins and Johnson, 2007; Micheneau et al., 2008c) and, more generally, for most orchid species (van der Pijl and Dodson, 1966; Dressler, 1981). However, none of the bagged flowers set fruit in the studied populations, so neither species is autopollinating, and both are fully dependent on a pollen vector to set fruits. Autofertility is widespread in Orchidaceae (5–20 % of the total species; Catling, 1990), and in general in insular ecosystems where it is often explained as an evolutionary consequence of lack of pollinators or specialized pollinators (Barrett, 1985; Barrett and Harder, 1996; Jacquemyn et al., 2005). On Réunion, almost half of orchid species are thought to be self-pollinating (Jacquemyn et al., 2005), including five of the nine Jumellea species. Autofertility in Jumellea seems to be linked to pollination syndrome, and especially to spur length, as the five autopollinating species have a sphingophilous pollination syndrome with long spurs (spur length >9 cm), and is thought to be linked to the absence of long-tonged specialized pollinators (Jacquemyn et al., 2005; Micheneau et al., 2008b). Jumellea rossii and J. fragrans, which have short spurs (spur length <5 cm), maintain interactions with pollinators.

Pollen deposition vs. removal rates

For J. fragrans, over the 3 years of the study, the mean pollen removal percentage was 28 %, while the mean pollen deposition percentage was 12 %. For J. rossii, the same trend was observed: the mean pollen removal percentage was 26 %, whereas the mean pollen deposition percentage was 6 %. First, those results reflect the presence of effective pollinators in each studied population. Except in 2009, when pollen removal and pollen deposition rates were dramatically low in the population of J. rossii, pollen movements and pollination efficiency are globally the same in both populations. The values of pollen removal and deposition are similar to tropical orchids, in which median fruit set is about 17 % (Tremblay et al., 2005). Numerous studies relate negative effects of fragmentation on plant–pollinator interactions. For example, in a review by Ferreira et al. (2013) based on 155 studies, the authors found that 92 % of the studies showed significant effects of landscape changes on the diversity, frequency and movement patterns of pollinators and/or on the diversity, reproductive systems, reproductive success and productivity of plants. Despite this general trend, the reproduction of the studied population of J. fragrans seems to be unaffected by habitat size reduction and maintains efficient plant–pollinator interactions with similar pollen removal and deposition percentages to the J. rossii population. Secondly, the study of pollen movements indicate that a large proportion of removed pollinia never encountered stigmas in the two studied populations and the low fruit sets observed suggest that pollinators can be a limiting factor in both populations. This seems to be a current phenomenon in orchids, and particularly in tropical orchids (Tremblay et al., 2005).

Mating systems

The heterozygote deficiency, the F_IS values and the estimated selfing rates observed in both populations studied suggest a proportion of selfing in the mating system of both species. An important feature of the floral biology of the two species is the stigma–anther separation by the rostellum, which prevents autonomous self-pollination. Moreover, bagging experiments confirmed the absence of autofertility in these species. Hence, inbreeding in J. fragrans and J. rossii appears to be due to facilitated inbreeding (geitonogamy, pollinator-mediated autogamy or biparental selfing). In many self-compatible species, the selfing component of a mixed-mating system is considered to be non-adaptative and is an unavoidable consequence of the large floral display needed to attract pollinators, resulting in geitonogamous selfing (Lloyd and Schoen, 1992; Jarne and Charlesworth, 1993; de Jong et al., 1993; Harder and Barrett, 1995). The main pollinators reported for J. rossii are two species of Noctuidae (Mallet, 2014), which generally visit several flowers on the same plant, and may have short flight ranges. This behaviour may promote geitonogamy, and could explain the selfing rates observed in J. rossii. The only pollinator reported for J. fragrans is a member of Sphingidae, Agrius convolvuli (Mallet, 2014). Despite their potentially longer flight ranges compared with Noctuidae, hawk moths are also known to visit several consecutive flowers per plant. Although they belong to distinctive pollinator groups, the known pollinators of J. rossii and J. fragrans have foraging patterns promoting geitonogamous selfing. As both species can have many flowers blooming simultaneously on the same plant, the selfing component of J. fragrans and J. rossii might be principally caused by geitonogamy. Both species experience inbreeding, but the values of the estimated selfing rates found for J. rossii are high and indicative of a mixed-mating system, whereas the lower and less significant values found for J. fragrans suggest that the species is predominantly outcrossing. Population size reduction is known to increase selfing, because of the reduction of the pool of available mates and/or the alteration of pollinator behaviour which may increase biparental inbreeding and/or geitonogamy (Jump and Peñuelas, 2006; Rusterholz and Baur, 2010; Aguilar et al., 2012). For these reasons, we might have expected to detect higher selfing rates in the J. fragrans population. Levels of geitonogamy can be impacted by factors such as floral display or plant density in the population, increasing with larger floral display (Hodges, 1995; Harder and Barrett, 1996; Snow et al., 1996; Karron et al., 2004; Mitchell and Karron, 2004) and higher plant density. In the two studied populations, floral display and plant density seem to be the same (data not shown) and could not explain the lower selfing rates observed in J. fragrans. However, this could be explained by the low efficiency of the pollinator. Indeed, pollinator species differ in their effectiveness in depositing and removing pollen from flowers. Generally, in moth-pollinated orchids, pollination is more efficient when the pollinator proboscis is slightly shorter than the flower spur, the head of the pollinator pressing firmly against the orchid column, enhancing pollination (Nilsson, 1988). In the case of J. fragrans, the pollinator proboscis is longer than the flower spur (Mallet, 2014) and this may result in some non-efficient pollination events. In such a case, pollen removal and deposition may be non-systematic, and pollen distribution may be done more randomly. This could explain the relatively low selfing rates of J. fragrans, despite the pollinator behaviour promoting geitonogamy.

Levels of genetic diversity

In this study, genetic variation within populations of each studied species was assessed with 12 nuclear microsatellite markers. In the J. rossii population, these markers showed a medium-high level of genetic variation (A_R = 8·575 and H_e = 0·673) in the light of the few other studies assessing genetic diversity in orchid species through microsatellite markers (Gustafsson, 2000; Gustafsson and Sjögren-Gulve, 2002; Swarts et al., 2009; Chen et al., 2014). Compared with J. rossii, the J. fragrans population exhibited a moderate level of genetic diversity (A_R = 4·983 and H_e = 0·492). Outcrossing plants typically show higher genetic variation within populations than selfing plants (Loveless and Hamrick, 1984; Hamrick and Godt, 1996), but, despite the fact that J. fragrans seems to show a higher level of allogamy than J. rossii, J. fragrans has a lower diversity. The levels of genetic diversity in J. fragrans might reflect the effects of population size reduction and fit the general opinion that widespread species have higher genetic variation than fragmented and narrowly distributed species (Hamrick and Godt, 1990). In a study comparing levels of genetic diversity of the rare orchid Gymnadenia odoratissima with the more common G. conopsea, Gustafsson and Sjögren-Gulve (2002) found a similar pattern. The successive fragmentation events of the J. fragrans habitat (Strasberg et al., 2005) could have resulted in the loss of allelic richness or gene diversity in the studied population. This can occur through population bottlenecks at the time of disturbance and genetic drift afterwards (Barrett and Kohn, 1991; Ellstrand and Elam, 1993). However, genetic diversity in J. fragrans is relatively high despite its restricted geographical distribution. Some other studies have found similar results (Jump and Peñuelas, 2006; Mix et al., 2006; Vargas et al., 2006; Moreira et al., 2009; Gonzales et al., 2010; Takahashi et al., 2011).

First, the mainly outcrossing mating system of J. fragrans can contribute to the maintenance of a certain level of genetic diversity. Secondly, the medium level of heterozygosity in J. fragrans in this study could be explained by the fact that the reduction of heterozygosity occurs more slowly than the loss of allelic diversity. As a consequence, the observed level of heterozygosity in a population whose size has been reduced by fragmentation can be larger than the level of heterozygosity which can be expected from the observed allelic richness (Cozzolino and Noce, 2003). It seems to be the case in J. fragrans, as the allelic richness appears to be more affected by population size reduction than the heterozygosity. Finally, J. fragrans had a wider geographical range until relatively recently, and its distribution has been rapidly declining as a result of destruction of its habitat (Strasberg et al., 2005). Assuming a long generation time of individuals of J. fragrans, the time span should not be long enough to permit the evolution of detectable strong fragmentation effects. So, if we consider that the effect of fragmentation increases with the number of generations elapsed (Aguilar et al., 2008), we can expect a future decrease in the genetic diversity in the studied population of J. fragrans. Moreover, the negative effects of fragmentation and over-collection on genetic diversity have probably been delayed by the relatively high density and total number of individuals occurring in the studied population. Indeed, species with high density are exposed to less potential risk of negative effects of population size reductions, because it results in remnants of smaller size containing many individuals. In such a case, reduction in genetic diversity can be neither large or immediate (Chung et al., 2014).

Potential for inbreeding depression

The magnitude of inbreeding depression detected by manual crosses is closely related to genetic diversity and levels of inbreeding of the population, and is a crucial parameter in the evolution of the mating system (Darwin, 1876; Charlesworth and Charlesworth, 1987). Jumellea rossii expressed inbreeding depression for seed germination (δ = 0·66). This result is in the range of the values observed for a mixed-mating system (Winn et al., 2011), so is consistent with the mating system observed in this species, and is considered as the fitness cost associated with geitonogamy (Eckert and Barrett, 1994). Conversely, we found no indication of inbreeding depression in J. fragrans in this study, as there was no significant difference between outcrossing and self-treatment for any of the measured parameters. Numerous factors can influence levels of inbreeding depression, including heterozygote frequency, mating history and life stage, but studies suggest that most inbreeding depression is caused by the homozygous expression of deleterious recessive alleles after inbreeding (Charlesworth and Charlesworth, 1987; Carr and Dudash, 2003). However, repeated selfing reduces the magnitude of inbreeding depression by purging these deleterious alleles over time (Husband and Schemske, 1996; Pico et al., 2007). Thus, in natural populations, selfing rates and the magnitude of inbreeding depression are generally negatively correlated (Charlesworth and Charlesworth, 1987; Husband and Schemske, 1996; Culley et al., 1999); historically, outcrossing species usually express higher levels of inbreeding depression than selfing species (Husband and Schemske, 1996; Pico et al., 2007). For this reason, we expected to detect inbreeding depression in the mainly outcrossing population of J. fragrans.

A potential explanation for this absence of significant levels of inbreeding depression is linked to the history of the species, which would have allowed it to purge a large part of its genetic load. Indeed, J. fragrans experienced many events of fragmentation and overcollection, resulting in successive genetic bottlenecks (Young et al., 1996; Aguilar et al., 2008). First, repeated bottlenecks are expected to reduce overall levels of genetic variation due to the loss of rare alleles during the sampling process (Jump and Peñuelas, 2006). Because deleterious recessive alleles that underlie inbreeding depression are often rare, part of these alleles are expected to be lost via the sampling process (Barringer et al., 2012). The increase in inbreeding which usually follows population bottlenecks, and the associated reduction of genetic diversity, can result in expression of inbreeding depression via the expression of the remaining recessive deleterious alleles. However, as inbreeding persists in fragmentation conditions, these deleterious alleles should be progressively purged. Finally, populations that have experienced repeated bottlenecks and in which fragmentation conditions persist over successive generations are likely to exhibit reduced levels of inbreeding depression due to a reduction in genetic load by the sampling process and purging (Ægisdóttir et al., 2007), particularly in self-compatible species (Husband and Schemske, 1996). This scenario is consistent with the medium genetic diversity observed in J. fragrans, and could explain the absence of inbreeding depression. An absence of potential for inbreeding depression allows us to think that reproduction (until the germination stage) is not limited by inbreeding depression, and could be positive for the short-term persistence of the population.

However, the literature on inbreeding depression reveals that its magnitude increases throughout the life cycle of many plant species (reviewed by Husband and Schemske, 1996), so inbreeding depression can sometimes be first detected in later stages of the life cycle, such as seedling growth (Karron, 1989). The absence of inbreeding depression at the studied life stages does not mean that inbreeding depression will not be detected in later stages for J. fragrans. Consequently, in order to treat the question of inbreeding depression rigorously, data from later stages of the life cycle are needed.

Intraspecific vs. interspecific variation in reproductive patterns

The study was conducted on one population of each species and does not allow the assessment of intraspecific vs. interspecific variation in magnitude in the studied parameters. However, a study based on ten other populations representing the three main habitats of J. rossii revealed that all the populations showed a heterozygote deficiency and F_IS values suggesting a proportion of selfing in their mating system (Mallet et al., 2014). F_IS values were statistically similar in the three habitat types, ranging from 0·389 to 0·414. Given these results and those of the present study, all the J. rossii populations investigated seem to express a mixed-mating system. We can suggest that there is probably no major intraspecific variation in the reproductive pattern in J. rossii. Concerning J. fragrans, only one other relatively large population is known. This population presented a similar pattern to the studied population, with a heterozygote deficiency and a F_IS value of 0·336 (Mallet, 2014), suggesting that both J. fragrans populations have a similar mating system. All the other known populations consist of few isolated individuals, making comparative study difficult.

These results suggest that the interspecific variation in the reproductive patterns highlighted in this study is larger than the intraspecific variation.

Conclusions

In this study, our results demonstrated that the two Jumellea species were self-compatible, but required a pollen vector to set fruit. Mating system estimates indicated that J. rossii had a mixed-mating system, whereas J. fragrans was mainly outcrossing, but in both species the selfing component was probably due to pollinator-mediated geitonogamy. The population of the widespread J. rossii had a higher level of genetic diversity than the J. fragrans population, which has been fragmented and is also illegally harvested. Given the current situation of rapid decline of its population size and its restricted distribution, conservation efforts should focus on J. fragrans. Indeed, although not yet critical, the genetic diversity of the species could still decrease assuming that negative effects of fragmentation increase with the number of generatios elapsed, and could be further reduced by the continuation of an excessive harvest. A continued reduction of J. fragrans genetic diversity could severely affect its evolutionary potential. So, it is necessary to carry out both in situ and ex situ conservation for this species. Even if not currently endangered, J. rossii is also of concern due to unsupervised collection. So, as for J. fragrans, it is important to conserve its natural populations and their genetic diversity. In this context, ex situ cultivation of both species seems to be a solution to decrease collecting pressure and meet the demand for this valuable resource.