The maintenance of diversity in Amazonia: a subtle combination of chance and determinism
Two main theories are evoked to explain community assembly and the maintenance of high diversity across tropical rainforest landscapes: Neutralism and Determinism. The contrast between the neutralist and the determinist theories of community assembly is quite comparable to the contrast between neutral and adaptive (molecular) evolution of populations.
Under the unified neutral theory of biodiversity (Hubbell 2001), meta-community dynamics is governed by the speciation-extinction equilibrium in which the size of populations changes randomly (‘ecological drift’), eventually leading to extinction, and populations exchange individuals according with dispersal distance between them (Ricklefs, 2006). Thus, species assemblages are random subsets of the available pool of species able to spread in a given area (Tuomisto & Ruokolainen 1997). Even if this model is often unrealistic (Ricklefs, 2006), it accounts for most of the observed patterns of species abundance in tropical communities, suggesting that neutral process play a crucial role in community assembly (Chaves et al., 2003). From an evolutionary point of view, populations may evolve neutrally (under the combination of random mutation, migration, genetic drift and demographic events). In theory, populations may diverge into separate species if gene flow is restricted, either by a biogeographic barrier, or by the geographic distance between populations (also called isolation-by-distance). In such cases of allopatric speciation, the probability to observe a given species in a given area is thus a function of the dispersal abilities of the neighborhood populations of this species (Latimer et al. 2005). The great diversity observed in Amazonia, by comparison with temperate forests, is commonly explained by differences in speciation-extinction rates that are themselves dependent on the size of the climatically similar area. The main hypothesis is that there is a positive relationship between an ecoclimatic zone and the geographic range size of a species. Subsequently, two main hypotheses could explain the great diversity of the tropics: ‘museum’ and ‘cradle’ (Chown & Gaston 2000, Mittelbach et al. 2007, Arita & Vazquez-Domingez 2008). The ‘museum’ hypothesis postulates that there is a negative relationship between the geographic range size of a species and its likelihood of extinction. This is because large ranges should buffer species against extinction by reducing the probability of range wide catastrophes and because large population sizes would minimize the chance of extinction due to stochastic reasons. Because large species range sizes are typical of the tropics, tropics should act as a museum of diversity with low extinction rates with older taxa by comparison with temperate zones. The ‘cradles’ hypothesis postulates that there is a positive relationship between the geographic range size of a species and the likelihood of its speciation. This is because species with larger ranges are more likely to undergo allopatric speciation resulting from isolations-by-distance or isolations by biogeographic barriers. Tropic may thus be viewed as cradles of diversity, with high speciation rates.
In the ‘environmental filtering’ theory, species assemblages are controlled by determinist factors involving abiotic and biotic interactions (Wright 2002). In particular, habitat heterogeneity (Terborgh et al. 2002) and local interactions (mainly competition and predation) are commonly evoked as important drivers of diversity in tropical landscapes. Environmental filtering exerted by both abiotic and biotic factors would have led to niche partitioning and habitat specialization in tropical rainforest trees. From an evolutionary point of view, the evolution of populations and the divergence between species may have been driven by selective pressures exerted by environmental heterogeneity (sympatric speciation). Moreover, habitat heterogeneity is associated to disturbance gradients (particularly logging, and tree-fall gaps). Under the disturbance hypothesis, species diversity is enhanced by intermediate levels of disturbance, as observed in French Guiana (Molino et al. 2001). Another determinist hypothesis evokes density-dependant mortality around mother trees. This hypothesis was formulated by Janzen (1970) who observed a decrease in seedlings mortality with the distance to the mother trees, probably due to allelopathic chemical compounds or to density-dependent predation. This process leads to ‘gaps of regeneration’ around mother trees, allowing the installation of other tree species and preventing mono-specific assembly. However, this process remains poorly understood and documented.
Neutrality and determinism probably act in pair in governing species evolution and assembly structuring (Gravel et al. 2006, Jabot et al. 2008), and their relative effects probably vary across geographical scales and study areas (Gravel et al. 2006, Jabot et al. 2008). In the following section, I will focus on spatial heterogeneity in tropical landscapes (and particularly that observed at local scales) without, however, excluding the existence of neutral processes.
Spatial heterogeneity in the Amazonian rainforest
Environmental heterogeneity across forests landscapes
At continental and regional scales, both precipitations and the intensity of the dry season are the main causes of climatic variations across the Amazonian forest landscape: while temperatures are quite homogeneous, precipitations show large variations among regions (ranging from 1000 to 3000 mm per year, figure 17) with a precipitation gradient that increases from Southeast to Northwest Amazonia (Mayle & Power 2008). Moreover, the intensity of the dry season is more pronounced at the extreme of the gradients, where precipitations are the most abundant. French Guiana also exhibits a large gradient of precipitations that increases from west to east, figure 18 (Wagner 2011).
At local scale, large environmental variations are caused by soil factors related to topography (figure 19). Despite its apparent homogeneity, the tropical landscape of Amazonia displays complex habitat patchiness due to the alternation of water-logged bottomlands and terra-firma. Local topography causes strong differences in environmental factors (including water, light, and nutrient availability) among local micro-habitats.
In bottomlands, plant communities are established on hygromorphic soils submitted to seasonal or permanent water-logging and frequent flooding events. As in temperate ecosystems, water-logging is a major constraint for tree regeneration and growth. Water-logging decreases the solubility and transfer of o2 in the soils. Due to root and soil microbial respiration, oxygen quickly decreases in soils; leading to hypoxia and accumulation of CO2 (Ponnamperuma 1972, Kozlowski 1997) that in turn affects root and microbial respiration (Epron et al. 2006). Moreover, water-logging leads to production of reactive oxygen species by roots that causes oxidative stress (mainly, H2O2 is produced by mitochondria when respiration slow down), Perata et al. 2011. In parallel, hypoxia causes a decreases in the root permeability that subsequently affect water and nutrient uptake from the soil, causing stomatal closure and a decrease in photosynthesis (Perata et al. 2011). On the contrary, terra-firme (slopes and hilltops) are display ferralitic and well-drained soils allowing important vertical and lateral drainage. Thus, terra-firme soils usually display lower water content than bottomlands. Tree communities, particularly seedlings unable to directly uptake water from the ground water table, may experience seasonal drought stress due to the depletion of water from at least the upper soil layers (Bonal et al. 2000, Daws et al. 2002, figure 20).
Figure 19: Soil properties along topography gradients (from Sabatier et al. 1997). Mainly, DVD=deep vertical drainage, Alt=red alloteriet, SLD= superficial lateral drainage and SH= hydromorphic soil.
Moreover, soil fertility varies from hilltops to bottomland. Reductions in soil respiration affect nitrogen cycling in bottomlands (Luizao et al. 2004) that commonly contain less nitrogen than hilltops or slopes (figure 21) but frequently contain more phosphorous than hilltops (Ferry et al. 2010). Last, topography gradients are associated with variations in irradiance transmitted below the canopy. As the soil is instable in slopes and water-logged soils, tree-fall gaps occur more frequently in slopes and bottomlands (Marthews et al. 2008, Ferry et al. 2010), figure 22.
Consequences of spatial heterogeneity on plant communities:
At regional scale, the structure and composition of plant communities may vary along rainfall gradients, as proposed by numerous studies (Givnish, 1999, Engelbrecht & Kursar, 2003, Condit et al. 2004). However, discerning whether adaptive or neutral processes are involved is a complex issue at such large scales. In the particular case of Amazonia, rainfall is supposed to exert a small effect on species diversity, whereas a strong effect of local patchiness is evident (ter Steege & Hammond 2001).
At local scale, large variations of plant community composition and diversity vary along topographic gradients. The most obvious variation of plant communities is the large increase in palm biomass in bottomlands (Kahn 1987, figure 23) and variations in tree species composition. Indeed, numerous palm and tree species are significantly associated to a particular habitat-type (Clark et al. 1999, Vormisto et al. 2004, Baraloto et al. 2007). This statement is commonly invoked as a result of adaptive radiations caused by topography leading to niche partitioning and habitat specialization. However, several studies suggested that the majority of species is generalists regarding to local habitat (figure 24, Webb & Peart 2000, Valencia et al. 2004) and their distribution is probably constrained by dispersal without being influenced by habitat heterogeneity.
Figure 23: Schematic representation of plant communities along
Figure 24: Venn diagram showing associations of tree a topography gradient. From Kahn et al. 1987. species to three local habitats (from Webb & Peart 2000). Circle intercepts show species encountered in different habitats.
Figure 25: Left: Canonical correspondence analysis for environmental variables: soil types, topographical positions, slope and elevation (from Clark et al. 1999); Right: Vegetation ordination after correspondence analysis: symbols indicates different soil types differing in drainage and hygromorphy: (from Sabatier et al. 1997)
Several topographic and soil variables are however particularly relevant for explaining tree community composition and structuring (ter Steege et al. 1993, Clark et al. 1999, Sabatier et al. 1997, Kanagaraj et al. 2011), including slope, elevation, soil water availability, drainage, and water logging, figures 25.
Even if a majority of studies focus on one or several environmental factors or topographic variables, the structure of plant communities probably results from a complex superposition of factors (among which local irradiance, nutrient availability, water-logging and drought). Thus significant habitat-associations are commonly explained by species sensitivity to the underlying constraints: Engelbrecht et al. (2005, 2007) and Poorter et al. (2008) proposed drought, Paliotto et al. (Palmiotto et al. 2004) suggested irradiance, Lopez et al. (Lopez & Kursar 2003) proposed both flood and drought, whereas Baraloto et al. (Baraloto et al. 2005) proposed both nutrients and light. For example, a field experiment revealed a reversal of performance ranking among species between local situations (Baraloto et al. 2005), suggesting different degrees of sensitivity to constraints among species. Thus, adaptation to a particular habitat may partly explain the differences in community composition and species abundance among micro-habitat.
Local habitat patchiness is also associated with large variations of tree biomass and functional traits. In bottomlands, tree biomass is lower than in terra-firma (Kahn 1987, Ferry et al. 2010), probably because soil instability constrains a more superficial root anchorage and limits tree growth. Moreover, Kraft et al. (Kraft et al. 2008) found a significant structuring of functional traits at the community level in Ecuador, which is also consistent with a role of habitat filtering, figure 26. Another kind of phenotypic structuring commonly observed in tropical rainforest is the ability of trees to develop morphological particularities, particularly in bottomlands. For example, buttress or stilt roots prevent constraints due to soil instability, whereas adventitious roots, lenticels, and aerenchyma tissues allow partial maintenance of root respiration in water-logged habitats, by allowing oxygen uptake directly from the air and oxygen transport to roots (Kozlowski 1997, Parelle 2010).
Figure 26: Distribution of SLA (expressed as a deviance from null distribution) in relation with local topography. From Kraft et al. 2008.
The entire forest dynamics vary along topographic gradients: canopy opening events created by frequent tree-fall gaps are also proposed as a driver of diversity in meta-communities (Schnitzer 2001, Robert 2003), by allowing establishment of light-demanding pioneer species and thus, creating patches of regenerations in the middle of mature communities composed by a majority of shade-tolerant tree species (Denslow et al. 1987, Schnitzer 2001, Ferry et al. 2010). Quesada et al. (2009) categorized forest dynamics according to a function of disturbance from soils, see figure 27.
Figure 27: Variations in forest dynamics in relation with soil type in Amazonia. From Quesada et al. 2009.
Consequences of spatial heterogeneity on population evolution & species divergence
As quickly evoked previously (‘3.The maintenance of diversity in Amazonia’), population evolution is driven by a combination of neutral (mutation, recombination, genetic drift, migration, reproduction, demography) and adaptive (natural selection) processes. Populations may diverge into new species, either due to isolation-by-distance that may be caused by populations isolation into refuges or biogeographic barriers (allopatric speciation), or by local adaptation to habitat heterogeneity (sympatric speciation). However, the drivers of populations evolution and speciation processes in tropical rainforest trees are poorly known, partly because the boundaries of species are often confused, and many species are organized in species complexes, with incomplete reproductive isolation between species and cryptic species (Cavers & Dick 2013).
At regional scale, many phylogeographic analyses revealed patterns of genetic divergence structured by the biogeographic history of the species, and mainly dispersal constraints that occurred during tertiary and quaternary. For example, Jacaranda copaia is widespread in the Amazon basin and comprises two sub-species: one subspecies widespread from Central America to Bolivia and another one distributed in the Guiana shield. In a recent study, Scotti-Saintagne et al. (2012) showed that the geographical patterns of genetic diversity in these two Jacaranda copaia sub-species were largely shaped by Pleistocene climatic changes that isolated ancestral species into refuges, with a center of diversification in Central Amazonia probably due to a secondary contact zone. Moreover, the absence of cross-Andean disjunction suggested that the Andean uplift was not a barrier to dispersal, probably because Jacaranda copaia is a wind-dispersed pioneer species, favored by canopy gaps and disturbances, and able to tolerate relatively dry conditions. Another example is provided by the Carapa species complex (Duminil et al. 2006, Scotti-Saintagne et al. 2012). Scotti-Saintagne et al. suggested that the biogeographic history of two Carapa species was a combination of tertiary and quaternary events, including Pliocene Andean uplifts, and then late Miocene development of Amazon drainage, but was also influenced by hybridization and introgressions during the Quaternary, figure 28.
Figure 28: Bayesian clustering analysis for the tree Genus Carapa in the Neotropics (from Scotti-Saintagne et al. 2012).
Maps indicate the structuring of genetic diversity at continental and regional scales.
In an original study (Fine & Kembel 2010), Fine et al. evoked the large influence of specialization to habitat type in driving the phylogenetic divergence between species. They analyzed the phylogenetic structure of Amazonian communities involving 1972 taxa across habitat types in Peru (white-sands that were widespread before Andean uplift and terra-firme forests composed by Cretaceous sediments that were laid down during Miocene). They compared the relative effects of habitat type and geographic distances between communities on the phylogenetic distances between taxa. They concluded that both dispersal limitation and habitat specialization influenced species divergence in tropical forests, but the effect of habitat specialization was greater than distance between communities, figure 29. They remained, however, cautious about the age of divergence: both biogeographic history of habitat types and recent in situ adaptive radiations governed by habitat heterogeneity would be involved in clade divergence.
Taken together, these results reveal that the biogeographic history of species is often insufficient to catch all the processes that structured the genetic diversity and induced speciation in Amazonian landscapes. In particular, more recent specialization to constraints would also be involved in species evolution and divergence, particularly at local scale.
At local scale, several studies revealed strong evidence of habitat specialization among closely related species. Baraloto et al. analyzed the distribution of four pairs of species from the same genus and observed divergent local habitat-associations between closely-related species (Baraloto et al. 2007). They proposed that specialization to local habitat may explain patterns of adaptive radiation in many tree genera. Similarly, Tuomisto et al. (Tuomisto 2006) observed strong evidence of niche specialization to local edaphic constraints (soil texture, soil cation content, inundation) between species of the Polybotrya genus in northwestern Amazonia.
Even if numerous studies evoked the influence of local variations in shaping the genetic diversity of tropical plants and in driving sympatric speciation, no study yet provided molecular evidences of local adaptation at intra-specific level in Amazonia. In temperate and boreal plant communities, local adaptation has been largely investigated and provides a wide range of examples: local adaptation to altitudinal gradients (Savolainen 2011), to water-logging (Parelle et al. 2010) etc… (see section ‘Molecular evolution’). In tropical rainforests, however, the relative influence of local adaptation and neutral processes in structuring the genetic diversity over short spatial scales remains largely misunderstood and requires much attention, particularly in the current context of climate change.
Tree species model, research questions and study sites
In this study, I address the question of population evolution at local scale within continuous populations of a dominant tree species widespread in French Guiana: Eperua falcata (a complete description of the species is given page 35). I addressed two main questions:
1) How is the genetic diversity of Eperua falcata structured in the forest landscapes of French Guiana?
2) Which evolutionary drivers are relevant to explain the structure of genetic diversity at local scale?
3) Does local adaptation contribute to structure the genetic diversity at local scale within continuous populations?
I analyzed the patterns of genetic diversity distribution within continuous forest landscapes of French Guiana through a global approach integrating both ecophysiological (phenotypic) and population genetics (molecular) approaches that are treated separately.
Figure 30 (page 34) provides a complete overview of the methods, the specific questions and future prospects.
The section ‘Molecular evolution’ aims at (i) analyzing patterns of genetic differentiation among local habitats, (ii) identifying which evolutionary drivers structure the local genetic diversity of Eperua falcata, and (iii) testing for local adaptation by (iiia) detecting outlier loci under diversifying selection among local habitats and (iiib) estimating the extent of (divergent) natural selection in the genome of Eperua falcata. This section involves two main approaches:
– a candidate gene approach in which targeted genes of known function (potentially involved in adaptive genetic differentiation among local habitats) were sequenced: aquaporins, catalase, farnesyltransferase, etc…
– a genome-scan approach in which I genotyped a large number of (anonymous) AFLP markers spread over the genome.
The candidate gene approach was developed during the PhD of Delphine Audigeos. I participate to this work during my Master degree by developing genetic markers and by contributing to genetic analyses. The AFLPs approach was set-up during this PhD.
In parallel to population genetics, I worked on creating a large database of Eperua falcata expressed sequences (cDNA) that were sequenced by 454-pyrosequencing prior to this PhD. I realized the bioinformatics assembly and post-processed it to characterize genes and identify polymorphism. Such a database will be useful for further high-throughput re-sequencing or genotyping of candidate loci.
The different results obtained are detailed in the research articles, but the main results are summarized into this synthesis (‘orange boxes’). The prospects of the study are discussed in the section ‘Discussion’.
Table of contents :
INTRODUCTION – Evolution in Amazonia
1. Short overview of the Amazonian rainforest
2. The building of biodiversity in Amazonia
3. The maintenance of diversity in Amazonia: a subtle combination of chance and determinism
4. Spatial heterogeneity in the Amazonian rainforest
5. Tree species model, research questions and study sites
PART 1 – Molecular evolution: population genetics and genomics
1. Population evolution
2. Population differentiation: the complex interplay between gene flow, selection, and drift.
3. Neutral differentiation
4. Adaptive differentiation
5. Next generation sequencing / genotyping and new opportunities
PART II – Phenotypic evolution and quantitative genetics
1. Causes of phenotypic variation
2. Phenotypic evolution in populations
3. Phenotypic differentiation
1. Neutralism and adaptation in Eperua falcata
2. Open questions and perspectives
3. Importance of assessing genetic diversity in a changing world
PhD RESULTS & DEVELOPMENT OF BIOINFORMATIC TOOLS
Article n°1 Molecular divergence in tropical tree populations occupying environmental mosaics