Although phylogenetic distance between native and exotic species has a close link with their interactions, it is still unclear how environmental stresses and species interactions influence the relationship between phylogenetic distance and biological invasions. Here we assessed the effect of invader-native phylogenetic distance on the growth of the invader (Symphyotrichum subulatum) under three levels of drought (no, moderate or intensive drought). Under no drought, interspecific competition between close relatives was the dominant process, and native communities more closely related to the invader showed higher resistance to invasion, supporting Darwin’s naturalization hypothesis. In contrast, under intensive drought, facilitation between close relatives by mutualism with arbuscular mycorrhizal fungi (AMF) became more important, and the invader became more successful in their more closely related native communities, supporting the pre-adaptation hypothesis. The colonization rate of AMF of the invader was higher in more closely related native communities regardless of the drought treatment, but it was only positively related to invader biomass under intensive drought. Therefore, the shift of species interactions from competition to facilitation may be ascribed to the promotion of AMF to invasion occurring under intensive drought, which leads to the effect of closely related natives on the invader shifting from negative to positive. Our results provide a new angle to resolve Darwin’s naturalization conundrum from the change of species interactions along a stress gradient, and provide important clues for invasion management when species interactions change in response to global climatic change.

Experimental plant communities with five levels of species richness (1, 2, 4, 8 and 16 species) were constructed at Taizhou University (28°39′N, 121°23′E) in Taizhou, Zhejiang Province, China. The species pool consisted of 16 native herbaceous species (Antenoron filiforme, Achyranthes aspera, Solanum nigrum, Penthorum chinense, Sesbania cannabina, Patrinia scabiosaefolia, Eclipta prostrata, Polygonum chinense, Bidens pilosa, Perilla frutescens, Artemisia migoana, Justicia procumbens, Persicaria lapathifolia, Lolium perenne, Cichorium intybus, and Medicago sativa) were selected to stimulate natural communities. Patrinia scabiosaefolia (Valerianaceae), Achyranthes aspera (Amaranthaceae), Solanum nigrum (Solanaceae) and Artemisia migoana (Asteraceae) are usually dominant species. We constructed a total of 46 communities with different species compositions: one monoculture for each of the 16 species, 10 two-species mixtures, 10 four-species mixtures, 9 eight-species mixtures and one 16-species mixture. Each of the two-, four- or eight-species mixtures had a different species composition, and the species were randomly chosen from the species pool. All the 45 communities of the one-, two-, four- and eight-species mixture were replicated six times, and the community of the 16-species mixture was replicated 30 times, resulting in a total of 300 communities. The six replicates of the 45 communities were randomly assigned to the six combinations of three drought intensity treatments (no, moderate and intensive drought) and two invasion treatments (with or without an exotic plant invasion; as described in the next two sections); the 30 replicates of the 16-species mixture were also randomly assigned to the six treatment combinations of drought and invasion, with five replicates each. 

We harvested the communities in all containers in October 2016. The biomass of most species almost reached to the highest value in October. In each container, aboveground living plants were sorted to species, dried to constant mass at 80°C and weighed. As some pathogens seemed to specifically infect Medicago sativa, no plants of this species survived at the end of the experiment. As the monocultures of M. sativa had no living plants, these containers (one in each of the six treatment combinations of drought and invasion) were not included in data analysis. Consequently, there was a final sample of 49 plots in each of six treatments.