Study Site

Fieldwork took place in the Nouragues Nature Reserve, French Guiana (N 04°05’, W 52°41’) between 12 and 26 January 2010. The site experiences about 3000mm of rainfall per year, mostly distributed between January and June and with a drier season from September to November. The mean annual temperature is 26°C. The Nouragues is uninhabited and almost completely free from human disturbances.

In order to maximise the diversity of sampled habitats, we collected termites in 10 parcels of 1ha around the station. The plots were scattered on the slope of the Nouragues granite outcrop, from topmost rocky savannah to flat riverbank. The substrates exploited by termites were variable and included: poor aggregations of organic matter bellow stones in the rocky savannah (leptosols), richer soils (dystric nitisols), and organic matter-rich substrates represented by abandoned nests. We also sampled rotten wood or organic matter accumulations below trees, and poor sandy soils repeatedly washed by the river during seasonal floods (dystric nitisols with gleyic properties).

Parcels A-C and E-G were lowland primary rainforest growing on well-drained soil. Parcel D was very similar to the former six, but differed in the high density of lianas. Parcel H consisted of a forest growing beside a creek on poorly drained soil. Parcel J was a low forest growing on the foothills of the Nouragues inselberg, transitional between lowland rainforest and the rock savannah growing on the granitic outcrop. Finally, parcel I was located at the frontier of a low forest and a typical primary lowland rainforest, representing an intermediate type between the two types [32].

Ethics Statement

The research permits required to work in the Nouragues Nature Reserve were included in the Nouragues project we got from Le Centre national de la recherche scientifique (CNRS). Our work did not involve any endangered or protected species.

Sampling

Sampling effort was about three men/day for each parcel and did not follow a systematic protocol. We instead searched for termites of the Anoplotermes-group in all suitable microhabitats, namely: in the soil, in rotten pieces of wood or in abandoned termite nests. Six types of samples were collected: five categories of foraging substrates and one category of controls, i.e., soil samples without termite activity. Three controls were collected in each parcel, always from bare ground selected randomly for their absence of any termite activity. The five types of foraging substrates were: (1) abandoned nest = abandoned material from termite nests, devoid of its original builders, usually built by Labiotermes labralis and sometimes possibly by Embiratermes neotenicus, but often modified by other termite species; (2) rotten wood = pieces of decayed wood on which termites were feeding; (3) bare soil = soil not taken from near the base of any tree, palm or other type of vegetation; (4) soil (palm) = soil situated at the base of a palm tree; (5) soil (tree) = soil situated at the base of a tree, generally between buttress roots. For each foraging party encountered, we collected a voucher sample composed of 20 to 50 termite workers in 80% alcohol for species identification, recorded the type of substrate and sampled the foraging substrate for subsequent analyses. Species identification was carried out based on worker morphology and DNA barcoding [32]. Foraging substrate was collected right at the place where foraging termites were found. To ensure that samples were foraging parties and not galleries, only samples composed of more than 20 individuals were collected. 50 to 100g of soil or abandoned nest material were stored at -20°C, dried at 50°C and shipped to Belgium where analyses were carried out. Altogether, the foraging substrate of 173 termite parties was recorded, and 118 samples of termites were collected with their associated foraging substrate (S1 Table). Dead wood was not analysed in the present work because it is fully organic and cannot be compared to soil in many aspects such as its texture. Thirty additional control soil samples were analysed.

Soil Analyses

Twenty-seven physico-chemical variables were measured on each soil sample: total carbon and nitrogen, exchangeable cations, exchangeable acidity, available phosphorus, pH (H₂O and KCl extracts), organic matter content and soil texture (clay, silt and sand).

Soil analyses were carried out following standard protocols [33]. Soil samples were air-dried and sieved at 2 mm. Total carbon (C) and nitrogen (N) content was determined on crushed samples with a dry combustion C-N analyzer. Exchangeable cations and plant available phosphorus were extracted with 1 M ammonium acetate EDTA at pH 4.65 and measured with ICP-OES (Ca²⁺, Mg²⁺, K⁺, Fe³⁺, Zn²⁺, Cd²⁺, Pb²⁺, Cu²⁺, Ni²⁺). Exchangeable acidity and exchangeable aluminium were extracted with 1 M KCl and determined by derivative titration curves (H⁺ and Al³⁺). The cation exchange capacity (CEC) was calculated as the sum of base cations and the exchangeable acidity expressed in cmol_c kg^-1. Soil pH-H₂O and KCl were measured using a combination of glass electrodes in a 1:4 (v/v) slurry after 2 h of contact. Organic matter content (OM) was determined by loss on ignition at 550°C. Soil texture (i.e. proportions of clay, silt and sand) was determined using 10 g of sample after H₂O₂ pre-treatment and dispersion with Na citrate. The sand fraction was separated by wet sieving, and the clay and silt fractions were determined using a sedimentation method.

Data Analyses

We used a detrended correspondence analysis (DCA) to visualise species association with type of substrate. The analysis was carried out using the CANOCO software with the option detrending by segment, using 26 segments, that allows rescaling each segment to mean value of zero on the second axis [34]. Species that were encountered on fewer than four occasions were discarded from the analysis as the detection of species-site association is dubious if it is based on too few records. The correlation between species scores along the first DCA axis and the mean δ¹⁵N isotopic values previously measured for 14 species [30] was investigated using Pearson correlation (note that the δ¹⁵N isotopic values were obtained from different samples). Differences between feeding substrates (abandoned nest, bare soil, soil (palm) and soil (tree)) and control were investigated using Kruskal-Wallis tests on each soil variable separately, testing for differences between substrate types. Multiple comparisons of mean ranks, implemented in Statistica, were used to test for differences between groups. Dunn-Šidák corrections were used to avoid errors resulting from multiple comparisons. To investigate differences between species in terms of feeding substrate composition, Kruskal-Wallis tests followed by multiple comparisons of mean ranks and Dunn-Šidák corrections were computed using each soil variable separately for all species with four measurements or more, and only from bare soil samples. We computed a principal component analysis (PCA) in order to dimensionally reduce the dataset and facilitate its visualization. We used only the samples for which the complete series of soil variables measured was available (88 out of the 160 soil samples). Other samples with lacking data for one or more variables were discarded from the analysis. The analysis was carried out on log-transformed data using a covariance matrix, implemented in the CANOCO software. To give the same weight to each variable, data were standardized and centered prior to the analysis, so that data average was zero with a variance of one. Differences between substrate types along the first four PCA axes were investigated by comparing scores with Kruskal-Wallis tests followed by multiple comparisons of mean ranks and Dunn-Šidák corrections. Differences between species and parcels were not investigated using component scores because there were too few records to reach significance.

Do soil-feeding termite species differ in the foraging substrate?

Our results suggest that the niche partitioning in Neotropical soldierless termites, as previously revealed by δ¹⁵N isotopic analyses [30], is caused by preferential feeding on specific substrates and that further subcategories within a foraging substrate cannot be recognized. Examples of species-specialization include Aparatermes cingulatus and Ap. sp A, which are characterized by a low score along the first axis of the DCA (Fig 1), indicating that they feed on substrates rich in organic matter such as rotten wood or abandoned nests; other species, such as Longustitermes manni, An. jheringi or An. sp AD, obtained high scores along the first axis of the DCA, showing their association with soil poor in organic matter content. Interestingly, species position along the first axis, reflecting their type of foraging substrate, was positively correlated with δ¹⁵N isotopic values. A factor explaining interspecific differences previously measured with δ¹⁵N isotope ratios could therefore be species specialization toward a particular substrate type, which was not taken into account previously [30]. By contrast, no differences were observed among samples of bare soil exploited by six different species, although including additional species could give different results.

The fact that several species of the Anoplotermes-group feed on the same type of soil may suggest they share the same feeding niche, which would contrast with other soil arthropods such as oribatid mites [7, 8] and springtails [9, 10]. In this study, the substrates we analysed contained at least tens of termites, and therefore represented, with few possible exceptions, foraging sites rather than galleries connecting the nest to foraging area; the collected soil samples thus most likely represented the actual substrate termites fed on. However, the scale at which our samples were collected, namely spoonfuls of soil around termite foraging parties, may not match the scale at which individual termites actually select food particles.

The suggested absence of feeding niche differentiation among members of the Anoplotermes-group feeding on bare soil also seems surprising because of the highly variable morphology of the worker digestive tract in this group. One structure of special interest is the enteric valve, which often bears a sclerotized armature whose shape is highly diverse among species [32, 35, 36, 37]. The anatomy of the enteric valve and other digestive tract structures undoubtedly plays a role in digestion, but it might not be as appropriate to distinct feeding groups as previously suggested [25].

Although soil-feeding termites are frequently found within items of dead wood, this substrate was not analysed in the present work, because it is fully organic and cannot be compared to soil in many aspects such as its texture. Anoplotermes-group termites occurring in the wood seem to feed exclusively on the already decomposed fraction of the wood as suggested by stable isotope data [26, 30].

Do soil-feeding termites select their foraging sites to enhance nutrients intake?

We measured 27 variables to determine the influence of soil composition and structure on termite communities. The 27 variables were highly correlated together as shown by a PCA whose first three axes explained 67% of the variance (Fig 3, S1 Fig). Rich soil patches have consistently high content in all nutrients and poor soil patches are poor in all nutrients (Fig 3, S1 Fig), suggesting that targeting rich soil patches would enhance intake of all nutrients simultaneously. However, we found no evidence that termites feeding in bare soil target nutrient rich patches.

Palm tree crowns work as a funnel gathering large amounts of leaf litter which accumulates on the ground underneath and is degraded by decomposers. Similarly, abandoned termite nests are commonly inhabited by secondary occupants, which use them as shelter and food source [38]. In agreement with previous studies [39, 40], our results show that abandoned nests are richer in organic matter and nutrients than bare soil, but not than soil (palm) and soil (tree), which are of comparable richness in organic matter and nutrients (Table 1, Fig 3, S1 Fig). High nutrient content makes nests attractive to a large number of termite species among which some, like Inquilinitermes spp., are obligatorily associated to the nest of other species.

Many species were also collected only from the bare soil and never occurred in richer substrates. We tested whether bare soils (with termites) were in nutrient content richer compared to controls (bare soils without termites), but our results did not reveal any significant difference. Soil-feeding termites thus did not select richer soil patches. As also the case for tropical forest ants, soil structure and composition therefore seem to be unimportant for the structuring termite communities [41]. Previous studies found that sodium enhances litter termite abundance in some areas [15], but no such influence was detected here. One possible explanation is that sodium and other nutrients are available in sufficient amounts, and therefore do not limit population size in the Nouragues reserve, which is located about 100 km away from the sea and therefore receives sea-water aerosols [15]. Additionally, many parameters differ between our study and Kaspari et al. [15]'s, including the sampling layer (soil vs. litter) and treatments (random sampling in natural conditions vs. high experimental enrichment of small patches), making comparisons difficult. Overall, our results indicate that unlike abandoned nests or soil (palm) which are targeted by many species, termites foraging on bare soil do not preferentially select richer soil patches and are indifferent to organic matter and mineral content, at least within the range of variation naturally observed in the Nouragues forest.

Our results indicate that niche partitioning occurs among the so-called “soil-feeding” termites through the selection of species-specific feeding substrates (e.g., organic matter-rich substrates such as abandoned nests or soil at the base of palms). However, any further subcategories among Anoplotermes-group species feeding on soil sensu stricto could not be recognized, which suggests that these species do not select foraging patches according to their texture or composition. Additional studies are needed to test whether or not these species differ by selection of food items at a higher scale of resolution. If they occupy the same trophic niche, a further question is whether other ecological or behavioural features (nest site, colony size, reproductive strategies) contribute to niche separation. Alternatively, a broad overlap of niches would suggest, according to neutral theory, the importance of random drift effects in shaping the termite community.