It has been suggested that ecosystem fragmentation has important consequences for animal populations by reducing and dividing the distribution areas and by limiting connections between the fragments 12. As a result, this natural or anthropological phenomenon is described as a process promoting the genetic differentiation of isolated populations and could be involved in speciation. Insects are interesting models for studying the effects of fragmentation owing to their abundance, limited dispersal, short generation time and sensitivity to disturbances. In tropical ecosystems, termites (Isoptera) may represent up to 95% of soil insect biomass; humivorous termites contribute to tropical soil fertilization 3456. Since soil-feeders are very sensitive to changes in their environment, they constitute interesting bio-indicators in landscape fragmentation studies, when the evolution of their specific richness is surveyed 789.

Studying the link between genetics and fragmentation requires evaluation of breeding structure, gene flow and genetic differentiation among populations. This implies unambiguous identification of species. For a long time, termite systematics was based on morphological and\or morphometrical character sets for individuals belonging to various castes (alates, soldiers or workers). During the past decade, an increasing number of taxonomic studies have shown that molecular methods constitute fast and reliable diagnostic systems, which complement morphological identification. Indeed, several studies based on mitochondrial genome sequences such as the cytochrome oxidase genes, the AT-rich region and the 16S rDNA have thrown a great deal of light on termite taxonomy and on phylogenetic and phylogeographic analyses of the Reticulitermes 101112131415, Nasutitermes 161718 and Heterotermes 19 genera. Because mitochondrial gene trees do not necessarily reflect species trees owing to their maternal inheritance, the addition of nuclear markers in molecular studies is useful for confirming the organismal phylogeny 20. Among the nuclear sequences, the internal transcribed spacers (ITS) of rDNA are more polymorphic between than within species. It is generally assumed than concerted evolution homogenizes individual rDNA repeats and produces a mostly uniform sequence in all repeats in a given species. Although there is intra-individual variation in some taxa, the ITS2 region sequences are considered phylogenetically informative and able to distinguish closely-related species 21. Such an evolutionary pattern has allowed sibling or cryptic species in the Reticulitermes genus to be discriminated 222324. Microsatellites are also very useful nuclear polymorphic markers and have contributed widely to the resolution of colony and population genetic structures in Isoptera 252627. Many surveys have concerned the xylophageous genus,Reticulitermes, in non-natural fragmented areas such as urban ecosystems 2829. To date, however, microsatellite markers have not been used to investigate termite phylogeny or species discrimination, as has been done for other groups e.g. vertebrates 30, ticks 31, wasps 32 and ants 33.

In the present survey, we studied the humivorous afro-tropical Cubitermes genus (Termitidae, Termitinae), which shows remarkable ecological plasticity, colonizing savannahs or forests according to species. Among the soil-feeding Termitidae, the Cubitermes genus is one of the main builders; its nest can shelter up to 10 genera of true inquiline or optional builder termites 34. In spite of its undeniable ecological interest, the taxonomy of this genus is poorly resolved and a complete revision appears necessary 35. Indeed, at least 64 species of Cubitermes have been described on the basis of morphometrical characters, but it is strongly suspected that some of these are synonymic. It is important to note that extensive taxonomic work is necessary for the West African species since the sole currently available key concerns the East African species 36. Molecular data are also lacking for the Cubitermes genus, as only one sequence from cytochrome oxidase I 37 and two from 12S mtDNA [GenBank: AF475037, AF475001] are registered.

The aim of this work is to clarify the taxonomic status of the Cubitermes colonies from the Lopé Reserve region (Gabon), in order to obtain a better understanding of tropical termite diversity in fragmented areas. The Lopé Reserve is typically characterized by a mosaic of forests and savannahs and constitutes an ideal setting for studies of natural fragmentation.

On the basis of morphological comparisons with type specimens in the collections of the British Museum (London) and the Royal Museum for Central Africa (Tervuren), and because they showed no diagnostic morphometrical variation or molecular divergence in mitochondrial 12S and 16S rDNA (Harry, unpublished data), the Cubitermes colonies were affiliated to a single species, Cubitermes sp. affinis subarquatus (Sjöstedt). Here we attempt to reconstruct a phylogeny based on three types of polymorphic and independent molecular markers. We sequenced a portion of the mitochondrial COII gene and the ITS2 region, and determined the genotypes at five microsatellite loci isolated from Cubitermes subarquatus 38, for 19 Cubitermes colonies from four different sites (Figure 1). Three of these sites were in the same geographical scale within the Lopé Reserve and corresponded to forest patches of different ages, including small savannah patches: Okoumé (75 years old), Rocher (800 years old) and Chameau (800 years old). A fourth site, Doda, was an isolated gallery-forest outside the Reserve within a savannah landscape.

In tropical soil-feeding termites, it is impractical to apply a Biological Species Concept 39 because of the difficulties in realizing experimental crosses and observing natural hybridizations. Thus, our purpose here was to apply a Phylogenetic Species Recognition (PSR) based on the Genealogical Concordance Concept 40 to diagnose the Cubitermes species or to detect cryptic species. Such a concept has often been applied in bacteria, fungi and animals. In termites, research involving DNA-based taxonomy of structural and agricultural pest species is important 10. Comparison of multiple molecular markers has allowed the taxonomic status of new species in Reticulitermes to be assessed 6061 and has provided evidence for species synonymy in this genus 121423. Many DNA-based species recognitions are corroborated by evidence of morphological/chemical differences or geographically defined groups. However, PSR can be a powerful tool for diagnosing otherwise undistinguishable species, because genetic changes occurring in recently-isolated species may be observed before morphological or behavioral changes arise 41. As found recently in other social insects, genetic isolation is not always accompanied by evident morphological differentiation. In fire ants belonging to the Solenopsis genus, genetic analyses based on allozyme and mitochondrial markers demonstrated the occurrence of sympatric and indistinguishable cryptic species 424344.

The genetic results obtained in our study from the combined mitochondrial, nuclear and microsatellite markers unequivocally show deep separation among four groups of genotypes in the Cubitermes sp. affinis subarquatus colonies from the Lopé Reserve region. The congruence of these three types of unlinked molecular markers strongly supports the existence of differentiated genetic pools in this limited area. The mitochondrial and nuclear sequences show very little variation within the four groups but there is considerable variation among them. Although the allele ranges overlap for all the loci, possibly because of size homoplasy, the microsatellite data confirm the occurrence of four differentiated groups. These unexpected levels of subdivision are unlikely to have arisen under gene flow, so these four newly-detected groups can be seen as good evidence for the existence of cryptic species previously unrecognized by morphological techniques.

The lack of detectable morphological differences among Cubitermes species is not surprising in view of the data in the taxonomic literature, which imply that the genus is highly complex (great number of species, synonymy, missing data, etc.) 35. Furthermore, the boundaries between nominal Cubitermes species are often concealed by intra-specific morphometrical variability. In his revision of the East African Cubitermes species, Williams 36 mentioned that most of the specific characters vary greatly in size and can also vary markedly in shape and proportion within a particular species.

In addition, our genetic data inform the current debate about the use of comparative phylogenetic methods for studying present-day species distributions. These distributions do not necessarily reflect the geographical range of the ancestral species at the time of speciation, because geographical distributions are often labile owing to climate fluctuations, territory expansion or extinction of competitors. In particular, we have certainly not sampled the entire species range and enabled definite conclusions to be drawn about biogeographic differentiation and speciation. However, the repartition of the cryptic Cubitermes species is quite interesting in relation to the species distributions of poorly dispersive insects.

In our study, we can consider two landscape units: (1) the Lopé Reserve south of the Ogooué River, where Cubitermes spA, C. spB and C. spC live sympatrically in the closely apposed sites of Okoumé, Chameau and Rocher, the only exception being that Cubitermes spA is absent from the Chameau site; (2) the Doda zone to the north, where Cubitermes spD is restricted to the distant and isolated gallery-forest of Doda and represents the only Cubitermes species in the Doda zone across 10 km. The ecological differentiation of the two zones results from paleogeographic events. Indeed, since the last glaciations, the North Ogooué has progressively run dry, resulting in the regression of all forest types except riparian. In the South Ogooué, a wet environment has been maintained by the well-developed hydro-geographical network and this has allowed the forest to be preserved.

The contemporary sympatric distributions of three of the Cubitermes species (Cubitermes spA, C. spB and C. spC) in the Reserve zone could reflect their ability to disperse within a mosaic of forests of variable ages (e.g. differing in biotic and edaphic parameters) and savannah "buffer-zones". Indeed, their dispersion seems not to be affected by fragmentation on this small scale. Little is known about the dispersion modalities and the reproductive strategies of the Cubitermes species. It is very likely that swarming (i.e. alate dispersal flight) is the main mode of dispersal in this genus, since budding (i.e. local secondary reproduction initiated by the differentiation of neotenic reproductives, derived from the nymphs or workers of the colony) is not as common in Termitidae as in lower termites 45. It has been suggested that although the active flight of winged termites is very limited (a few hundreds of meters), the meteorological conditions accompanying dispersal flights could strongly influence the distance covered by these sexual alates. Actually, a recent genetic study of Macrotermes michaelseni (Termitidae) suggested that some winged termites can travel considerable distances (50 km), most likely by passive drift 46. Furthermore, the effectiveness of dispersal obviously depends upon the number of alates and the rate of predation 47.

Finally, the particular distribution pattern of Cubitermes spD raises the question of the link between the history of the forest fragmentation and the modalities of speciation. One can indeed wonder whether the history of successive modifications of habitats in this geographical zone has not contributed to the isolation of the termite species populating it. Areas such as isolated gallery forests, like Doda, may constitute refuges for fauna, where some species are led to disappear while others begin to differentiate under the influence of genetic drift.

The combination of mitochondrial and nuclear markers provides a reliable diagnostic method for separating Cubitermes species and offers a complement to the morphometrical diagnostic. Furthermore, these molecular markers could reveal useful information about their phylogenetical relationships. Similarly, such methods could be extended to termites for which species taxonomy is ambiguous.

VR and CD carried out the molecular genetic studies, participated at the data analysis and the manuscript draft. AL intensively helped VR and CD in DNA sequencing. MH initiated the study, realized the sampling, contributed to the analysis of the results and to the writing of the paper. All authors read and approved the final manuscript.