Population genetics data help to guide the conservation of palm species with small population sizes and fragmented habitats in Madagascar

Study sites and target species

The Itremo Massif, officially designated in 2015 as a protected area (Madagascan Ministry of Environment, Ecology, Sea and Forests, 2015) and part of the SAPM network created out of the Durban Vision, is located in the Central Highlands of Madagascar, to the west of Ambositra in the Amoron’i Mania region. The massif is characterised by an outcrop of metamorphic rocks dominated by quartzite and schist over a distance of approximately 120 km, lying north to south, interrupted frequently by open plains (Cox, Armstrong & Ashwal, 1998). Ranging in elevation between 1,500 and 2,100 m above sea level, the geomorphology of the whole area is generally abrupt and its bioclimatic type is warm subhumid, characterised by ca 1,500 mm of annual precipitation, a mean average temperature of 25 °C, and up to seven dry months (Cornet, 1974). Given these physical settings, Itremo’s biodiversity is recognised as particularly unique with a flora including many narrow endemics recorded from several distinct vegetation types, dominated by xerophytic rocky outcrops, grassland, Tapia woodland and gallery forest (Moat & Smith, 2007).

Four palm species have been recorded from this area: Dypsis ambositrae Beentje, D. baronii (Becc.) Beentje & J Dransf., D. decipiens (Becc.) Beentje & J Dransf., and Ravenea madagascariensis Becc. Two of these species, D. ambositrae and D. decipiens, have been assessed as Critically Endangered and Endangered respectively, using IUCN Red List categories and criteria (version 3.1) (Vié et al., 2008; Rakotoarinivo & Dransfield, 2012a; Rakotoarinivo & Dransfield, 2012b; Rakotoarinivo et al., 2014). Dypsis baronii and R. madagascariensis were assessed using the same categories and criteria and found to be of Least Concern, so were not included in this study (Rakotoarinivo & Dransfield, 2012c; Rakotoarinivo & Dransfield, 2012d).

Dypsis decipiens is found across the Central High Plateau of Madagascar between Andilamena and Fianarantsoa. In the Itremo Massif Protected Area, the species occurs in large stands of hundreds of individuals, along riverine valleys and hillsides, in open grasslands, and it has been estimated that at least 2,700 individuals may grow here (Fig. 2A) (Rakotoarinivo & Rajaovelona, 2013). In contrast, D. ambositrae is restricted to much smaller numbers. Fewer than 50 mature individuals were known in the wild at the start of this work, all restricted to gallery forest in the Itremo and Ambositra areas (Fig. 2B) (Rakotoarinivo & Rajaovelona, 2013).

Sampling strategy

Fieldwork took place between July 2012 and March 2013 as part of a Conservation Leadership Project managed by Kew Madagascar Conservation Centre, under research permits issued by the Madagascan Ministry of Environment and Forests (permit number: 151/12/MEF/SG/DGF/DCB/SAP/SCB, issued to Mijoro Rakotoarinivo). Samples of leaf tissue were dried in silica gel to preserve their DNA. The majority of specimens were collected from the Itremo Massif region, where the population sizes of the two species are largest; other known sites for the two species visited and sampled were Ankazobe, Ilaka, Ambalamanakana, and Ankafina Tsarafidy (Fig. 3). There are some unconfirmed reports of small numbers of D. decipiens also being found east of Antananarivo, but these sites were not sampled during this study. For the large populations of D. decipiens in the Itremo Massif (Fig. 4), a 1 km² sampling grid was overlain on satellite imagery of the region. Only this one large species of palm is found in the open grassland of this region, so individual palms could be mapped accurately and, using the grid, leaf material could be collected from individuals from each grid cell in the field.

DNA extraction

DNA was extracted from the silica-dried leaf samples using a modified CTAB method (Doyle & Doyle, 1990)—washing the DNA pellet with 70% ethanol, drying it at 37 °C, and resuspending in TE buffer (20 mmol Tris-HCl, 0.1 mmol EDTA) (Rivers et al., 2011; Devey et al., 2013)—and the DNA concentration was quantified using a Nanodrop spectrophotometer.

AFLP methods

AFLP methods followed those described by Vos et al. (1995). EcoRI and MseI restriction enzymes were used to digest each DNA sample, and double stranded EcoRI and MseI adaptors were ligated onto the ends of the digested fragments. A primer trial was conducted using 12 primer pairs, to select three combinations that would provide a level of variation appropriate for this study. Primer combinations MseI-CTC + EcoRI-AC, MseI-CTG + EcoRI-AAG, and MseI-CAC + EcoRI-AAC, were used in the analysis, the selective EcoRI primer being labelled with a fluorescent dye, and a 250 ROX size ladder being used. Resulting fragments were analysed using an ABI 3730 Genetic Analyser. GeneMapper (Applied Biosystems, 2005) was used to automatically score the presence and absence of fragments, ranging between 40 bp and 400 bp, in a binary matrix for all samples. Scoring of every electropherogram was manually checked to confirm the presence or absence of additional fragments that the programme was unable to unambiguously score automatically.

DNA samples collected in the field which did not produce consistent, scoreable AFLP bands, were excluded from subsequent scoring and analyses (see ‘Discussion’).

AFLP data analysis

Descriptive statistics for the D. ambositrae and D.decipiens datasets were calculated using AFLPsurv (Vekemans, 2002), using values of the within-population inbreeding coefficient F_IS calculated in I4A (Chybicki, Oleksa & Burczyk, 2011) and GenAlEx v.6.3.1 (Peakall & Smouse, 2006; Peakall & Smouse, 2012).

Relationships among populations of both species were investigated by calculating Nei’s genetic distances to generate genetic distance matrices, visualising the relationships between populations and individuals using Principal Coordinates Analysis (PCoA) (Gower, 1971) in MVSP (Kovach, 2013) and GenAlEx (Peakall & Smouse, 2006; Peakall & Smouse, 2012) to assess the distinctiveness of populations.

The significance of the partitioning of AFLP variation among populations was tested using Analysis of Molecular Variance (AMOVA) in GenAlEx using 999 permutations, and a Mantel test was implemented in GenAlEx to test for Isolation by Distance (IBD), comparing genetic and geographic distance matrices for both species.

AFLP analysis

AFLP profiles with data from all three primer combinations were produced and analysed for 20 D. ambositrae and 50 D. decipiens individuals. Descriptive statistics of the results for both species are shown in Table 1.

Genetic structure

The PCoA results using Euclidean genetic distances between individuals show that there is some clear separation of the genetic variation in the two species into clusters. The two populations of D. ambositrae separate into two separate clusters using the first two PCoA axes (Fig. 5). The clusters visualised for D. decipiens using PCoA axes 1 and 2 (Fig. 6) do not separate the seven populations into seven separate groups, but there is some pattern—the two clusters seen comprise populations D2 and D4 in one cluster, and D5, D6 and D7 in the other. Individuals from populations D1 and D3 are divided between the two groups. The proportion of variation represented by each of the first two axes is shown in Figs. 5 and 6 (D. ambositrae: 18.3% and 13.5% respectively, D. decipiens: 7.7% and 5.7%). The third axis for each species (not shown) adds an additional 9.5% and 4.3% of the variation but reveals no more clustering of the individual points. Oddly, for both species, there is a single outlying individual. The outlying individual for D. ambositrae is a sample collected from the A2 population, outside the Itremo protected area. The outlying individual of D. decipiens is one of just two individuals collected from the D7 population north of Antananarivo (the other individual is found within the main D5-7 cluster of individuals in the PCoA).

The AMOVA test (Table 2) shows that in both species the majority of the genetic diversity is evident within populations rather than among them (79% and 93% of the variation for D. ambositrae and D. decipiens respectively); thus, more of the genetic variation in D. ambositrae is related to among population variation than in D. decipiens (21% compared with 7%), though it is recognised that the number of subpopulations studied for each species differ substantially. The pairwise genetic distances among sample sites (ϕPT) was much higher for D ambositrae (0.211) compared with those for D. decipiens (0.074).

Table 2:

Results of AMOVA conducted on variation among and within populations of Dypsis ambositrae and D. decipiens.

The data include the percentage of variation explained by the different hierarchical levels, and Mantel test results comparing linear genetic distance and geographic distance matrices between individuals of each species.

		Analysis of Molecular Variance (AMOVA)							Mantel
		df	SS	MS	Est. var.	%	ϕPT	P	Rxy	P
D. ambositrae	Among pops	1	61.400	61.400	4.600	21
	Within pops	18	310.250	17.236	17.236	79
Total		19	371.650		21.837		0.211	0.001**	0.344	0.010*
D. decipiens	Among pops	6	305.866	50.978	2.778	7
	Within pops	43	1500.874	34.904	34.904	93
Total		49	1806.740		37.682		0.074	0.001**	0.562	0.010*

DOI: 10.7717/peerj.3248/table-2

Notes:

Tests of significance

*P < 0.05.

**P < 0.01, based on 999 permutations.

Genetic diversity

This study has shown that overall genetic diversity and levels of heterozygosity are fairly low in both of the study species. Comparatively higher levels of genetic diversity are known to enhance species’ resilience and persistence in the wild, so these results raise concerns about the future ability of these two palms to survive in the future in the face of climate change and encroaching human-induced habitat destruction. Interestingly, the more widespread species D. decipiens actually has lower genetic diversity than D. ambositrae, the species with the dramatically smaller and geographically restricted population.

The low proportion of polymorphic loci revealed in the two palm species (46.5% and 36.6%) is comparable with the low proportion found in a similar AFLP study of Madagascan legume tree species, where an average of 40% of loci were found to be polymorphic, and the more threatened species were found to have fewer polymorphic loci than the less threatened taxa (Rivers et al., 2011). Low genetic diversity has also been reported in several threatened palms elsewhere in the world (Dowe, Benzie & Ballment, 1997; Shapcott, 1998; González-Pérez, Caujape-Castells & Sosa, 2004), but in other studies substantially higher genetic diversity has been revealed (Shapcott, 1999; Henderson, Billotte & Pintaud, 2006; Shapcott et al., 2007).

Compared with another study of Madagascan palm population genetics using AFLPs, the study of Beccariophoenix madagascarienis (Shapcott et al., 2007), the measures of genetic diversity revealed in here seem to be low. For B. madagascariensis, mean genetic differentiation among populations (F_ST) is much higher (at 0.599, versus 0.095 and 0.081 respectively for D. ambositrae and D. decipiens), and similarly, the expected heterozygosity (H_e) is 0.595–0.691 for B. madagascariensis, versus 0.233 and 0.239 for D. ambositrae and D. decipiens. In the Madagascan legume AFLP study of Rivers et al. (2011), species considered to be of higher threat status were also shown to have lower genetic diversity indices, including the Shannon Diversity Index, as seen in this study for both D. ambositrae and D. decipiens). A study of two Madagascan palms, Voanioala gerardii and Lemurophoenix halleuxii (Shapcott et al., 2012), showed contrasting patterns—demonstrating high genetic diversity in the former but low genetic diversity in the latter, but used microsatellites rather than AFLPs.

Genetic and geographic structure

There is clear evidence for a degree of genetic and geographic clustering in both species. Clustering is more clearly defined in D. ambositrae, with two genetically distinct populations, which correspond to the two geographically distinct populations (A1 and A2), being revealed by the AFLP analysis. In D. decipiens, there seems to be some genetic structuring dividing the sampled individuals into two main groups, but these do not correspond completely with the geographically distinct populations. There could be said to be a core Itremo genetic group, comprising individuals from D1 (west Itremo), D2 (north Itremo), D3 (central Itremo) and D4 (south Itremo), and a separate genetic group that follows the road network through Itremo from the west to the centre, and northward along routes RN 4 and 7. The distribution of this latter group could reflect the anthropogenic movement of the species along the road system.

Conservation recommendations

The population sizes and genetic and geographic groupings revealed in this study can be used to inform conservation decision making for the two species studied, and already are being taken into account in current conservation efforts.

The overall genetic diversity represented in both species is so low that extensive, time- and funding-consuming efforts to try to conserve representatives from across their entire distributions (e.g., in comprehensive seed collections) would not be the best use of available resources, as most of the limited genetic diversity would be captured by relatively few individuals. Although we might predict threatened species to represent low genetic diversity compared with less threatened species, and the results of this study are congruent with this theory, the results of these other studies suggest that this rule does not always apply and cannot be assumed for all species. Further study of biological and ecological traits of threatened taxa, including life history traits, population distributions and trends, geographic and habitat preferences, may help to interpret these conclusions further.

The results of this study suggest that it should be a priority to collect seed from both D. ambositae populations in order to capture the maximum genetic diversity. This should result in ex situ seed and living collections being that which represent as many unique alleles as possible, improving the genetic basis upon which future re-introduction and supplementation work would be based. In situ efforts should aim to protect both of the remaining populations from future degradation—although one population is protected within Itremo NPA, the population near route RN7 is highly threatened.

The situation with D. decipiens is perhaps less urgent than is the case with D. ambositrae, which has a much larger total population size. Moreover, most of the plants are found within the Itremo NPA, a legally designated site managed for species and habitat conservation objectives by RBG Kew with strong community engagement and development work. Interestingly, this study has shown that the genetic groupings of the species are limited to just two main clusters, and these do not correspond well with the geographic sites from across the full distribution of the species, from which the plants were sampled. Resource-intense efforts could in theory be made to collect D. decipiens seed from across its full distribution in order to capture as much genetic diversity as possible, but the results of this study show that increased sampling effort may well result in limited gains in terms of genetic diversity represented in the collected material.

The small populations of D. decipiens occurring outside the Itremo Massif area do not represent significant genetic diversity beyond that represented in the main populations within Itremo NPA, and so are less important for conservation either in situ or ex situ. We hypothesis that these small populations, reliably found near main roads, may be the result of humans deliberately moving these aesthetically attractive plants.

Since most individuals of D. decipiens grow within the Itremo NPA, and therefore are conserved in situ, our main recommendation would be ensure that the species is well represented in ex situ collections, ideally by progeny from a range of mother plants. As with other horticulturally desirable species, D. decipiens is widely grown around the world. However, in the case of species such as this, confined in the wild to remote islands like Madagascar, cultivated plants are likely to be the result of just a small number of original exported seed collections–and therefore represent a very limited gene pool from the species in the wild. This certainly applies to the Critically Endangered palm Tahina spectabilis, although represented in many ex situ collections around the world, virtually all plants are progeny of a single seed collection taken from just one individual in the wild (J Dransfield, pers. comm., 2016; Gardiner et al., in press). In the case of D. decipiens, collections of seed and/or seedlings from a greater number of individuals from within the Itremo NPA would likely improve the genetic diversity represented in ex situ collections. The lack of clear genetic groupings each directly corresponding to particular geographic populations means that seed collections cannot (and need not) be gathered from across the entire geographic distribution of the species, and less structured and more opportunistic collecting may well result in the collection of as much genetic diversity as a more systematic, geographically even but labour-intensive approach.