Phylogenetic α- and β-diversities jointly reveal leaf-litter ant community assembly mechanisms along a tropical elevational gradient

This study was conducted along the eastern slope of the Cofre de Perote mountain, in Veracruz, Mexico. This region is located at the junction of the Trans-Mexican volcanic belt and the Sierra Madre Oriental. We selected eight study sites spanning an elevational gradient of 3500 meters of altitude. Regardless of the geographical distance, all sites were systematically separated with an elevational difference of 500 meters on average between each other. We placed our study sites at the following elevations above sea level: 30-50 m, 610-670 m, 900-1010 m, 1470-1650 m, 2020-2230 m, 2470-2600 m, 3070-3160 m and 3480-3540 m, however, for simplicity, we will refer to each site as discrete units (i.e. 0, 600, 1000, 1500, 2100, 2500, 3100, 3500 m).

Sampling sites were old-growth forests characterized by no obvious forest use and highly dominance of mature forests, except in the case of the lowest site (i.e., La Mancha), where most of its original vegetation has been transformed. To overcome the effect of perturbation in the studied patterns, we sampled La Mancha in a secondary forest with up to 30 years of regeneration. All sampling sites were closed-canopy forests in which a leaf-litter layer could be guaranteed. During the rainy season (July-September) of 2018 one 300-m transversal transect was located at each one of the eight study locations where we established 10 equidistantly sampling points (i.e., 30 meters between each other). Two independent 1-m2 samples were taken perpendicularly to each sampling point: one 10 meters on the right side and the other 10 meters from the left side. This procedure was repeated in a second transect placed during the dry season (March-May) of 2019 to increase community characterization as well as reduce any seasonality effect on our diversity patterns. Transects within an elevational site were separated at least 1 km away from each other. Thus, we obtained 320 m2 leaf-litter samples characterized the whole mountain (8 study sites x 20 m2 per transect x 2 transects = 320 m2). In each 1-m2 quadrat, we collected the leaf litter inside and sifted it through a coarse mesh screen of 1-cm grid size to remove the largest fragments and concentrate the fine litter. The concentrated fine litter from each sample was suspended in independent mini-Winkler sacks for 3 days in the laboratory. Falling arthropods were collected into a container with 95% ethanol. Ant workers were removed from each container for identification. When possible, specimens were identified at the species level. If not, we assigned a morphospecies number.

Phylogenetic tree constructions

Ideally, one would use a complete, species-level phylogeny of all ant species present in your study area to calculate phylogenetic diversity, yet our current understanding of ant relationships is still limited. As an alternative, we built a genus-level phylogeny based on the tree by Moreau & Bell, (2013), but using the phylogenetic relationships and divergence times within Myrmicinae from Ward et al. (2015). This phylogeny was then pruned to keep only a single species per genus to generate a genus-level phylogeny. To maximize taxonomic coverage, we replaced genera that were missing from those studies by closely-related lineages that were not present in our dataset using other phylogenetic studies (Borowiec, 2016; Lapolla et al., 2010; Schmidt & Shattuck, 2014). We then used the list of species (Supporting Information Table S1) in our dataset to simulate a species-level phylogeny in which the relationships within genera were obtained from a Yule (pure-birth) process using the genus.to.species.tree function in the “phytools” package (Revell, 2012). A total of 1000 simulated trees were obtained to account for phylogenetic uncertainty [see Arnan et al. (2018) and Divieso et al. (2020) for similar approach]. Additionally, we constructed a maximum clade credibility tree (hereafter MCC tree) which was used to summarize the uncertainty of the 1000 simulated trees. The MCC tree was constructed from the sample of the 1000 trees with the maxCladeCred function incorporated in the “ape” package (Paradis et al., 2019). Both the 1000 hypothetical trees and the MCC tree were used in downstream analyses (Supporting Information Fig. 1).