Lise Dupont, UMR 137-IRD, Faculté des Sciences et Technologies, Université Paris 12, 61 Avenue du Général de Gaulle, 94010 Créteil cedex, France. E-mail: firstname.lastname@example.org
1Tropical rainforest fragmentation may increase the risk of extinction of limited dispersers such as soil-feeding termites which play a fundamental role in soil structure and fertility.
2We tested the hypothesis that population genetic structure of Labiotermes labralis, one of the most abundant soil-feeders in the Amazonian primary forest, may vary in response to local ecological conditions.
3As a pre-requisite, two factors that may have important consequences on population genetic structure were examined: colony social organisation and infection by Wolbachia bacterium. This cytoplasmic endosymbiont is able to manipulate arthropod reproduction and thus, to alter patterns of mtDNA variation.
4Three sites in French Guiana showing variable level of forest disturbance were studied. Three hundred and thirty-eight neuters and 14 primary reproductives from 17 colonies were genotyped using six microsatellites to analyse colony and population genetic structure. Moreover, one sequence of the COII mitochondrial gene was obtained for each nest.
5We showed that all nests were monogamous. In a single site, all nests were infected by the same Wolbachia strain. This pattern of infection was not associated to a particular mitochondrial haplotype.
6In the most disturbed site, a significant inbreeding coefficient associated with high relatedness between primary reproductives suggested that anthropogenic habitat fragmentation has induced genetic isolation of the population; a result reinforced by the mitochondrial data. Thus, habitat fragmentation might have serious consequences for the persistence of L. labralis populations in French Guiana.
Tropical rainforests support over 50% of earth's species but are experiencing massive land conversion by humans and are rapidly being transformed into a network of habitat patches (Dirzo & Raven, 2003; Fahrig, 2003; Sodhi, 2008). Understanding how forest fragmentation alters the ecological stability and evolutionary potential of the surviving flora and fauna is of fundamental importance. Molecular markers provide useful tools to investigate the ecological effects of habitat fragmentation on populations of endangered species (Freeland, 2005). Fragmented populations are expected to experience increased genetic drift, owing to small effective population sizes, and reduced gene flow, owing to decreased rates of movement among patches. Changes in the relative balance of genetic drift and gene flow are predicted to result in the erosion of genetic diversity and, simultaneously, in the increased differentiation among local populations (see Keyghobadi, 2007 for review). Populations that have been reduced in size and isolated by fragmentation are predicted to exhibit increased levels of inbreeding and, potentially, reduced fitness resulting from inbreeding depression (Keyghobadi, 2007).
In tropical ecosystems, soil-feeding termites account for more than 60% of the 2600 known termite species (Eggleton, 2000) and are ecosystem engineers playing a fundamental role in soil structure and fertility (Lavelle et al., 1997). Among termites, soil-feeders are the feeding group the most sensitive to forest disturbance (Eggleton et al., 2002). The severe decline of soil-feeders’ diversity after habitat fragmentation (Eggleton et al., 1996) may be explained by a vulnerable morphology (i.e. small size and soft body; Eggleton et al., 1998), a high degree of resource specialisation (i.e. humified substrates) and a restricted recolonisation ability (i.e. restricted dispersal) compared to wood and leaf-litter feeders (Gathorne-Hardy et al., 2000).
In social insects such as termites, the colony social organisation depends on the mode of foundation of new colonies. Three major types of colonies may be found: monogamous colonies typically founded by one pair of primary reproductives, the ‘queen’ and ‘king’; primary polygamous colonies for which the pleometrosis process (i.e. association of multiple primary reproductives during colony foundation) is typically the mode of settlement of the colony; and secondary polygamous colonies with secondary reproductives emerging from the parental colony and replacing dead or senescent primary reproductives. The type of social organisation defines the degree of relatedness between nestmates and may thus shape the genetic structure within populations. Knowledge about breeding strategy is thus essential to interpret population genetic patterns in termites.
The reproductive system of insects may be altered by Wolbachia bacteria which are cytoplasmic endosymbionts (α-proteobacteria) that infect a wide range of arthropod and nematode hosts. They are almost exclusively maternally inherited and may enhance their transmission by modifying the reproductive system of their arthropod host in various ways, that is, cytoplasmic incompatibility, male killing, feminisation and parthenogenesis induction (see Stouthamer et al., 1999 for review). Spread of a specific Wolbachia strain through a host population can have a directly analogous effect on variation in other cytoplasmically inherited markers, such as mitochondrial DNA (mtDNA; Rokas et al., 2001). Mitochondrial diversity of the population may thus be significantly reduced and geographical structure may be erased (Hurst & Jiggins, 2005). Association between Wolbachia infection and particular mitochondrial haplotypes (i.e. linkage disequilibrium) has been demonstrated for instance in a chrysomelid beetle (Keller et al., 2004a) and fire ants (Shoemaker et al., 2003). Testing for the presence of Wolbachia in a particular species is thus an essential pre-requisite when using mtDNA as marker in studies of population genetics.
In this study, we tested the impact of anthropogenic disturbance on the population genetics of Labiotermes labralis Holmgren (Termitidae, Nasutitermitinae), a soil-feeding termite from the Amazonian primary rainforest and subsisting patches (Constantino et al., 2006). Seventeen colonies from three sites in French Guiana submitted to differential anthropogenic pressures were studied using both nuclear (i.e. microsatellites) and mtDNA markers. Because patterns of mtDNA variation may reflect the evolutionary history of Wolbachia bacteria rather than their hosts (Jiggins, 2003), Wolbachia infection was investigated using a polymerase chain reaction (PCR)-based screening technique. We specifically addressed the following questions: (i) what kind of colony organisation characterises this species? (ii) does Wolbachia infection explain patterns of mtDNA variation? (ii) does human disturbance affect the genetic variability of the study populations?
Material and methods
Study area and sampling
While the inland rainforests of French Guiana are currently largely undisturbed, those of the coastal region addressed in this study are under threat. In this region, both timber extraction and agriculture are increasing due to population growth. Labiotermes labralis colonies were sampled in three sites (Fig. 1) showing various level of forest disturbance. The most preserved site, Cacao, was located in the Amazonian forest. A Hmong population is established in this site but people only cultivate the edge of the forest. The other colonies were sampled in two sites in Iracoubo, a littoral zone inhabited by a pluri-ethnic population. The Patagai site was a forest exploited by the ONF (Office National des Forêts). However, L. labralis was collected in a plot which was not exploited since a drastic cut made in 1993. The site of Rocoucoua was recently inhabited by Hmong families (1990) who practice a mixed form of agriculture combining inputs with the logic of productivity and slash and burn over the past thirty years. Interestingly, no L. labralis was found in a fourth site (Bellevue, Iracoubo), which was subjected to an Amerindian farming system using traditional slash and burn since 30 years.
At Rocoucoua and Patagai, sampling was carried out in a square of 16 points regularly spaced out by 200 m delimiting a window of landscape of 1 km2. Around each point located by GPS (GPS Magellan Gold, Santa Clara, CA, USA), the presence of termite nests was checked in a radius of 100 m. At Rocoucoua, the landscaped window was made of two sub-windows 1 km apart. An additional transect of 200 m × 1 km, which was crossed by the track, was prospected between the two sub-windows. In the site of Cacao, two transects of 200 m × 3 km distant of 6 km and situated on a footpath were surveyed. As described by Emerson and Banks (1965), L. labralis nests of about 1–3 m, with external ornamentations hanging like drops and a sculptured inside surface, were found at the base of the trees or stumps, in moist and swampy portions of the rainforest. All L. labralis nests discovered on the sites were sampled.
Neuters were collected from seven, five and five nests situated at Rocoucoua, Patagai and Cacao respectively. Reproductives were also sampled when it was possible to examine the whole nest. At Cacao and Patagai, some very large nests sometimes located higher than 3 m up the tree could not be entirely checked. Immediately after sampling, termites were placed in absolute ethanol and stored at 4 °C in the laboratory.
Microsatellite loci genotyping and nuclear genetic diversity
Total genomic DNA was obtained from the head and thorax of the workers and reproductives and only from the thorax of the soldiers using a DNeasy Tissue kit (QIAGEN). Twenty neuters (10 workers and 10 soldiers) from each colony except in N25 and N23 (N = 19; Table 1) and 14 reproductives were genotyped at six microsatellite loci: Lal1 to Lal6 following protocols detailed in Harry et al. (2007). The null hypothesis of independence between loci was tested using the program genepop version 3.4 (Raymond & Rousset, 1995). For each colony and site, genetic diversity was analysed computing allele frequencies, number of alleles, observed and expected heterozygosities using genetix version 4.05 (Belkhir et al., 2004). To take into account variation in sample size, allelic richness (AR; El Mousadik & Petit, 1996) was estimated using fstat version 2.9.3 (Goudet, 2002). Because genotypes within colonies may not be independent due to a strong family structure, a re-sampling procedure (i.e. a single neuter per colony was selected at random for a total of 20 replications) was performed for these analyses.
Table 1. Genetic characteristics of Labiotermes labralis colonies and Wolbachia infection status (Wolb.).
Number of sterile individuals genotyped (N), sampling of the queen and/or the king, relatedness between reproductives (Rp) and COII haplotypes (AN EU519314–EU519317) are indicated for each nest. Rp was calculated from inferred genotypes obtained with gerud 2.0 in colonies without sampled reproductives. *Wolbachia infection status: +, infected; −, uninfected;†t-test showed that mean Rp was significantly higher in Rocoucoua than in Patagai and Cacao.
In the colonies in which a ‘queen’ and ‘king’ were sampled, the observed genotype frequencies of the neuters were compared with frequencies expected from the parental genotypes by means of a log-likelihood ratio test (G-test) and locus-specific P-values were combined using r version 2.4 software (Stouffer's method). Moreover, the gerud version 2.0 program (Jones, 2005) which reconstructs parental genotypes even when no parents are known was used to assess the minimal number of parental genotypes in each nest and to reconstruct putative parental genotypes. The coefficient of relatedness (R, Queller & Goodnight, 1989) was estimated between reproductives from each colony using identix software (Belkhir et al., 2002). For colonies in which the queen only or no reproductive was sampled, the genotype(s) of the most-likely parent(s) generated by gerud were used for relatedness estimations.
Colony and population genetic structure
Colony and population genetic structure were investigated by estimating F-statistics using genetix version 4.05 (Belkhir et al., 2004). First, genetic differentiation among all sites and between pairs of sites was estimated by calculating , the Weir and Cockerham (Weir & Cockerham, 1984) estimator of FST (Wright, 1951). An unbiased estimate of the probability that FST values depart significantly from zero was calculated through a 2000-iteration Markov chain permutation procedure to test for differences in allelic distributions between pairs of sites using genepop version 3.4.
Because there was significant differentiation between sites, the analysis was also performed for each site separately so that F-values used to infer the population breeding system were not confounded by higher-level genetic structure, as proposed by Vargo (2003). We used the notation developed for social insects (see Vargo, 2003 and references therein). The genetic variation is partitioned among individuals (I), colony (C) and total components (T). Thus, FIT represents the standard inbreeding coefficient, FCT, the genetic differentiation among colonies, and FIC reflects the colony inbreeding coefficient. FIC is expected to be highly sensitive to the mating patterns within social groups (see Vargo, 2003). Specifically, FIC is expected to be strongly negative for colonies headed by outbred reproductive pairs and to increase towards zero with inbreeding among reproductives. The 95% CIs were constructed by bootstrapping over loci with 1000 replications. Cases where 95% CIs did not overlap with zero were considered to be significantly different at the α = 0.05 level.
Mitochondrial DNA analysis
A 625-bp fragment of the cytochrome c oxidase II gene (COII) was amplified using the primers A-tLeu and B-tLys modified by Miura et al. (2000). After purification using a DNA and Gel band Purification kit (GFX TM PCR kit, Amersham Biosciences, Piscataway, NJ, USA), sequence reactions were performed using BigDye Terminator Cycle Sequencing kit version 1.1 (Applied Biosystems, Foster City, CA, USA) and sequence data were obtained using an automatic DNA sequencer (Applied Biosystems, ABI PRISM 310). Sequences were aligned using the clustal w program (Thompson et al., 1994). A haplotypic network was constructed to analyse the COII haplotype distribution across sites using a reduced-median algorithm (Bandelt et al., 1995) as implemented in the software network 126.96.36.199 (http://www.fluxus-engineering.com/sharenet.htm).
Infection status was determined on primary reproductives when it was possible or, alternatively, on two neuters per nest. Screening for Wolbachia was performed by PCR using Wolbachia-specific primers for the coxA Cytochrome c oxidase, subunit I gene (Baldo et al., 2006). In colonies showing positive Wolbachia infection with the coxA primers, a fragment of the Wolbachia ftsZ cell-cycle gene was also amplified using the primers ftsZunif / ftsZunir (Lo et al., 2002). Successful amplifications were sequenced and Wolbachia supergroup affiliation was characterised by blast searches.
Microsatellite polymorphism and genetic diversity
In total, 338 neuters, 8 queens and 6 kings (Table 1) were genotyped at six variable microsatellite loci (Table 2). No pair of loci was found to be in linkage disequilibrium and Mendelian inheritance of all loci was confirmed. Over the whole data set, the number of microsatellite alleles ranged between two and six depending on the locus and mean expected heterozygosity ranged from 0.33 to 0.70 (Table 2). Over the 21 microsatellite alleles recorded, only 11 were shared by the three sites. No significant difference of gene diversity (HE) and allelic richness (AR) was found between sites. However, the mean observed heterozygosity was significantly lower in the Rocoucoua site (P < 0.05, Table 3).
Table 2. Variability of microsatellite loci in the study sites. Computation was carried out using 20 re-sampled data sets. NC is the number of colony per populations. Atot is the number of alleles per locus over the whole data set. The mean number of alleles (Nall) and the mean expected heterozygosity (HE) over the 20 analysis are given as well as alleles that were not present in all sites (A).
Rocoucoua (NC = 7)
Patagai (NC = 5)
Cacao (NC = 5)
Overall (NC = 17)
Table 3. Genetic diversity and F-statistics (with 95% CI) within sites. The mean allelic richness (AR), the mean expected heterozygosity (HE) and the mean observed heterozygosity (HO) across loci over the 20 re-samplings are presented. In addition, the colony inbreeding coefficient (FIC), the standard inbreeding coefficient (FIT) and the genetic differentiation among colonies within each site (FCT) are also shown. * Significant values at the α = 0.05 level.
In the six colonies in which a ‘queen’ and a ‘king’ were genotyped, all expected F1 genotypes were present among the neuters and they occurred in expected ratios (combined P-values range from 0.177 to 1). The cumulative exclusion probability computed in gerud was 0.65 when neither parent is known, 0.87 when one parent is known with certainty and one is unknown and 0.97 when both parents are known. Parentage assignments showed that all colonies had Mendelian genotypes in ratio in accordance with a single pair of reproductives. The genotypes of the sampled reproductives always appeared in the most likely paternity/maternity solution. The relatedness coefficient between reproductives (Rp) was statistically higher in Rocoucoua than in Patagai and Cacao (t = 3.819, P = 0.003 and t = 3.932, P = 0.003 respectively, Table 1).
Genetic differentiation between sites
The global test for genetic differentiation among sites revealed significant heterogeneity in allele frequencies among the three sites ( = 0.201, P < 0.001). Genetic heterogeneity was observed among each pair of sites. The higher value was obtained for the Rocoucoua-Cacao pair ( = 0.246, P < 0.001), followed by the Patagai-Cacao pair ( = 0.186, P < 0.001) and the Patagai-Rocoucoua pair ( = 0.161, P < 0.001).
Colony genetic structure
Table 3 shows the estimates of the F-statistics within each site. In Patagai and Cacao, FIC values were strongly negative indicating that colonies were headed by outbred reproductive pairs, while in Rocoucoua a near-zero FIC value was observed suggesting that the reproductives were inbred. Moreover, a significant FIT value was obtained in Rocoucoua that differed significantly from values obtained in the other sites. Such a result may be due to sub-structuring in the population (i.e. the so-called Wahlund effect) and/or to inbreeding.
The haplotypic network shows that individuals from Patagai and Cacao shared the same haplotype (i.e. H3, Fig. 2 and Table 1) except one worker from the colony N21 of Cacao displaying the haplotype H4 which is separated by only one mutational step to H3. In Rocoucoua two different haplotypes, separated by four mutational steps were observed: H1 was detected in N4 and N5 while H2 was identified in all the other Rocoucoua colonies.
Wolbachia infection was detected in all colonies of the Cacao site (Table 1). Sequencing of both the coxA gene from 11 individuals (i.e. the queen of the N21 colony and two neuters of each other five colonies) and the ftsZ gene from one individual per infected colony (i.e. six sequences) resulted in the same sequences of coxA (EU513375, 343 bp) and of ftsZ (EU513376, 292 bp) genes. blast results suggested that Wolbachia endosymbiont of L. labralis belong to the supergroup F. Indeed, the coxA and the ftsZ sequences were highly similar to sequences of two other Wolbachia endosymbionts of termite namely Apilitermes longiceps (supergroup F, EF 417914, E-value = 2 × 10−143) and Coptotermes lacteus (supergroup F, DQ 837190, E-value = 4 × 10−73), respectively.
Genetic markers are powerful tools to address the influence of habitat fragmentation on the genetic structure of populations. Recently, molecular markers have been increasingly used to study the genetic implication of habitat fragmentation for insects, for instance ground beetles (Keller & Largiader, 2003; Keller et al., 2004a,b; Keller et al., 2005), ants (Gyllenstrand & Seppa, 2003; Bickel et al., 2006), butterflies (Krauss et al., 2004; Anton et al., 2007) and bumblebee (Ellis et al., 2006). To our knowledge, there is no data available on termites from fragmented tropical ecosystems. Here we investigated the population genetic structure of L. labralis, which belong to an important group of termites (i.e. soil-feeding termites) both in terms of biodiversity and their function in the ecosystem, in a fragmented area of French Guiana. As an essential pre-requisite, two factors that may have important consequences on population genetic structure were examined: colony social organisation and infection by Wolbachia bacterium.
Colony social organisation
In social insects such as termites, the breeding system of the colonies defines the genetic relatedness among nestmates and may thus shape the genetic structure within populations. Because termite colonies are usually founded from a single reproductive pair (Nutting, 1969) they are, in theory inclined to be monogamous families. Predominance of monogamy in colonies was indeed confirmed by genetic studies in Nasutitermes nigriceps (Thompson & Hebert, 1998a,b) and Cubitermes spp. (own unpublished data) but the occurrence of polygamous societies was also revealed in some species. For instance, it was demonstrated that colonies of Nasutitermes corniger (Atkinson & Adams, 1997) and Nasutitermes costalis (Thompson & Hebert, 1998b) contain multiple queens and kings. The co-existence of numerous reproductive females in a colony was also revealed in mature colonies of the termite Macrotermes michaelseni (Hacker et al., 2005). In L. labralis, no more than two reproductives (the queen and the king) were found in the study nests and parentage analysis revealed that all colonies had Mendelian genotypes in expected ratio. Thus, despite the low variability of the microsatellite loci that limited our ability to detect multiple related reproductives, our results strongly suggest that monogamy is the general trend in L. labralis whatever the sampling site. In most genetic studies on termites, relatedness between reproductives was inferred from relatedness between neuters. Here the genetic relatedness of mating individuals could thus be directly estimated.
Alteration of the reproductive system by Wolbachia can have a profound impact on the genetic structure of their host populations (Pannebakker et al., 2004). In particular, single mtDNA haplotypes may become widespread in the host population through hitchhiking with a successful Wolbachia strain (Turelli et al., 1992). Here, investigation of Wolbachia infection revealed positive results only in the Cacao site, in which different mitochondrial haplotypes (H3 and H4) were observed. Moreover, uninfected colonies from Patagai also presented the haplotype H3. Thus, no particular link between infection and mtDNA haplotype was observed.
In social insects such as termites, males as well as sterile workers/soldiers are dead ends for Wolbachia propagation. It is thus unknown whether and how Wolbachia may manipulate their reproduction in comparison to other arthropods (see Zientz et al., 2004 for review). In addition, the Wolbachia strain detected in all Cacao nests belonged to the F supergroup. Although strains belonging to the F supergroup have been detected in several arthropod species including termites (Lo & Evans, 2007), as well as in members of the filarial nematode genus Mansonella (Casiraghi et al., 2005), the type of interaction (i.e. beneficial to the host or parasitic) of F-Wolbachia with their hosts is still unknown (Panaram & Marshall, 2007).
Given this lack of knowledge about Wolbachia's possible influence and because other symbionts might alter the population genetics of mtDNA (for instance Cardinium, Hurst & Jiggins, 2005), we would strongly recommend using both mtDNA and nuclear markers to study L. labralis population genetics.
Influence of habitat fragmentation on Labiotermes labralis genetic structure
The site of Rocoucoua has suffered recent habitat fragmentation due to intensive agricultural practices started after the installation of Hmong families from 1990. Assuming that decreasing habitat patch size and increasing patch isolation lead to concomitant changes in the size and isolation of local populations (for a review see Keyghobadi, 2007), the L. labralis population sampled in Rocoucoua site was thus expected to have experienced enhanced genetic drift owing to a reduction of effective population size. Moreover, the high genetic differentiation between sites separated by less than 15 km (i.e. Rocoucoua and Patagai) suggested restricted gene flow in this species. Increased genetic drift, in association with reduced gene flow, is predicted to result in the erosion of genetic diversity and in inbreeding (Keyghobadi, 2007).
Although the comparison of genetic diversity indices (i.e. AR and HE) between sites did not reveal any significant differences between the three sites, the high relatedness between Rocoucoua reproductives illustrated by the Rp and FIC values revealed inbreeding in this site. This result was confirmed by the significantly lower level of observed heterozygosity in Rocoucoua as compared to the other sites. Indeed, the frequency of homozygotes in a population can be taken as a general index of the level of inbreeding that the population experiences (Hedrick, 2000).
Inbreeding in Rocoucoua may illustrate the effect of forest fragmentation on sexual partner assortment. In monogamous species, the reproductives that appear at the time of swarming are full-siblings. Because soil-feeding termites are weak fliers and because different colonies may swarm asynchronously (Nutting, 1969), it is possible that new colonies may occasionally originate from sibling pairs rather than intercolony matings. The absence of inbreeding in the Patagai and Cacao sites suggests that such pairing between siblings does not occur frequently in connected habitat areas. However, in the Rocoucoua site where L. labralis colonies were found in remaining patches of forest, isolated from each other by roads or cultures, meeting of alates from different patches was unlikely. The alternative hypothesis for the pairing between closely related individuals is that colonies could be headed by secondary reproductives. However, adultoids (i.e. young imagoes remaining within the native nest instead of swarming) that are the principally secondary reproductives described in the family Termitidae (Myles, 1999), were never identified during sampling.
Mating happening at close distance around the parental nest should lead to the genetic isolation of the colonies situated in each patch. A genetic sub-structuring of the Rocoucoua population was indeed suggested by the mitochondrial data since we showed that the geographically closest nests of Rocoucoua (i.e. N4–N5 and N3–N17) shared the same mitochondrial haplotype. Moreover, the significant FIT value may be indicative of the occurrence of sub-populations in the Rocoucoua site, in addition to common mating between relatives. At Patagai and Cacao however, the lack of genetic sub-structuring and the absence of inbreeding indicate that colonies are inter-breeding units.
The results presented in this study are only preliminary because of small sample sizes and additional populations are needed to confirm our conclusions. In particular, a more complete survey on the whole distribution range would be necessary to fully understand some of the patterns observed here such as the occurrence of relatively distant mtDNA haplotypes in Rocoucoua. In addition, a high genetic differentiation was revealed by the microsatellite data between the Patagai and Cacao sites while both sites shared the same main COII haplotype (H3). Because microsatellites may trace back more recent events than mtDNA, this result may be indicative of a recent restricted connection between those sites but this needs confirmation.
Nevertheless, all analyses were congruent and showed inbreeding in the most disturbed habitat. This result is likely explained by a genetic isolation of the colonies in remnant patches. However, no strong erosion of genetic diversity was observed in this site. These observations highlight that genetic response to fragmentation is highly sensitive to the temporal scale over which fragmentation has occurred. Indeed, genetic responses to fragmentation take time to manifest themselves. As underlined by Keyghobadi (2007), an important question is whether populations would persist long enough for observable changes in genetic diversity to occur.
Landscape fragmentation is indeed expected to increase the risk of extinction in populations of limited dispersers due to the combined effects of reduced population size and increased isolation. The lack of L. labralis in some forested patches surveyed during the sampling trip might be indicative of a growing scarcity of the species in some areas. Habitat fragmentation may thus have serious consequences for the persistence of L. labralis populations if it weakens L. labralis ability for colonisation of new habitat patches or recolonisation of disturbed habitats. Such a negative influence of forest fragmentation on populations of soil-feeding termites is likely to have striking consequences for soil fertility of remnant patches.
The authors are grateful to the CIRAD and IRD staffs (Kourou, Cayenne) for field research logistic, Valery Gond and Corinne Rouland for their scientific field collaboration, Mme Kouyouri, Mr Vanq and Mr Sotak for giving us agreement to sample termites, and L. Celini for termite identification. This work was partially supported by the MEDD (Ecosystèmes Tropicaux Program, Ministère de l’Environnement et du Développement Durable, France).