The two sympatric ant-plant sister species Barteria fistulosa and B. dewevrei have hollow branches (called domatia) in which ants nest. Each plant is colonized by a single colony of one of three ant species, Tetraponera aethiops, T. latifrons or Crematogaster sp. In addition, two sister strains of Chaetothyriales fungi can also be hosted within domatia of these Barteria. To test for association specificity between the three types of partners we investigated the association pattern in two localities, Kessala and Bigoundou, in Gabon. Plants and ants were identified by morphology and fungi by molecular typing. In Kessala we found a preferential association between B. fistulosa, T. aethiops and the fungal strain Y1 on the one hand, and B. dewevrei, T. latifrons and the fungal strain Y9 on the other hand. Crematogaster ants were found in B. dewevrei, in plants that had smaller domatia than those colonized by Tetraponera. In Bigoundou we found mostly B. fistulosa, which were mostly colonized by T. aethiops. However, the two fungal strains were equally represented. This tripartite symbiosis shows association specificity, the pattern of which can vary across sites. Association specificity and site variation are probably related to ecological conditions but these still need to be identified.

Ant-plants produce hollow structures called domatia to host protecting ants. Although size variation in domatia is well documented between related species, intraspecific variation is little explored. The central African ant-plant Barteria dewevrei exibits strong variation in domatium size, giving the opportunity to explore the mechanism underlying variation in a mutualistic trait. We showed that domatium size in Barteria dewevrei varies between sites. We transplanted individual plants between two sites in Gabon where plants have different domatium sizes. Domatium size of transplanted plants changed, revealing that variation in this mutualistic trait is driven by phenotypic plasticity. The two sites differed in their environmental conditions: highland open savanna on sandy soil vs lowland closed tropical rain forest on sandy-loam soil. However, as stomatal density and δ13C of leaves did not differ between sites or between branches produced before and after transplantation, we have no cue on the role of abiotic stress (such as light intensity and water availability) in domatium size variation. As the obligate Tetraponera ant symbionts are too large to fit in the small domatia, variation of the mutualistic trait in response to environmental change through phenotypic plasticity may impact this specialized mutualism.

Barteria fistulosa and B. dewevrei, central African rain-forest trees, provide nesting cavities for Tetraponera aethiops and T. latifrons ants, respectively, which protect them against herbivores. To compare protection efficiency between these two symbioses, for 20 plants of each species in two sites in Gabon we measured the time elapsed before ants reached a focal leaf, for host leaves that were undisturbed, damaged (cut with scissors) or subjected to slight vibration (mimicking such damage), and for damaged leaves of the non-host Barteria species. Tetraponera aethiops displayed stronger protective behaviour than did T. latifrons. Time to reach a damaged host leaf (4.5 ± 2.6 min, mean ± SD) did not differ significantly from time to reach a leaf subjected to slight vibration (5.2 ± 3.0 min) for T. aethiops, but response to a leaf subjected to slight vibration (9.5 ± 1.9 min) was significantly slower than that to a damaged leaf (7.8 ± 1.9 min) for T. latifrons. The faster response of T. aethiops to slight vibration may have masked a response of this species to chemical signalling. Both ants reached damaged host leaves faster than damaged leaves of the non-host Barteria sp., indicating host plant specificity in ant responses.