Evolutionary relationship between plants and insects: Insights from island communities

Interactions between plants and insects have long fascinated scientists. While some plants rely on insects for pollination and seed dispersal, insects rely on plants for food or as a habitat. Despite extensive research investigating pair-wise species interactions, few studies have characterized plant and insect communities simultaneously, making it unclear if diverse plant communities are generally associated with diverse insect communities. This work aims to better understand the historical and evolutionary relationships between plant and insect phylogenetic diversity (PD) on islands. We hypothesized that phylogenetically diverse plant communities (i.e. high PD) support diverse insect communities, with the relationship varying with island isolation, area, age, and latitude. Species lists for plants and insects were compiled from the published literature, and plant PD was calculated using ´standardized mean pairwise distance´ (SES.MPD) and ´standardized mean nearest taxon distance´ (SES.MNTD). For insects, PD was estimated using the number of genera, families, and orders. We found that plant diversity in evolutionary recent times (SES.MNTD) is associated with recent insect diversity (number of genera), but no relationship was found between plant and insect diversity across whole phylogenies (plant SES.MPD vs. number of insect families). Distant islands generally support high PD of plants (high SES.MPD and SES.MNTD) and insects (low number of genera). Plant and insect PD was generally high on small islands, except for plant SES.MPD revealing no relationship with island size. Insect PD was somewhat higher on young islands (low number of families), whereas there was no relationship between island age and plant PD. Plant SES.MPD was higher on high latitude islands, yet we did not find significant relationships between the latitude and the metrics of insect PD, or plant SES.MNTD. These findings suggest that protecting high plant PD may also help conserve high insect PD, with a focus on small and distant islands as potential hotspots of phylogenetic diversity across multiple taxa.

Methods
The data were collected using the databases Google Scholar (for papers published from 1945) and JSTOR (for papers published between 1700-1945). We considered all papers containing species lists of plants and insects collected in the same locations. We used the following search terms: 

"Species list" OR “SPECIES RICHNESS” OR “SPECIES DIVERSITY” OR “INVENTORY” OR “CHECKLIST” OR “SURVEY” AND (("Plant diversity" OR "Plant survey" OR "Plant community") AND ("Insect diversity" OR "Insect survey" OR "Insect community")).

Papers focusing on pest insects and/or agricultural studies were excluded as they are not representative of the natural environment. A total of 81 pairs of plant-insect species lists were compiled from 53 publications and book chapters (Supplementary file 1). The papers contained full species lists, with no native and non-native species separated. We excluded one location (Rose Atoll) where both plant and insect species richness is <5.

The final dataset represented a diverse range of islands, both oceanic (n=55) and continental (n=26), with a wide span of island parameters such as distance from mainland, area and age. We measured the island's minimal distance to the mainland using Google Maps, whereas the area was determined using Global Island Explorer (Sayre et al. 2019). The estimated age of the islands was determined by the information in the published literature (available in Supplementary file 1).

Care was taken that only above-ground insects were sampled in all studies, diminishing the threat of any sampling bias across studies. The authors used the most appropriate methods for above-ground insect sampling such as Malaise traps (e.g. Early 1995), sweep nets (Hwang et al. 2022, Meads & Fitzgerald 2001), hand searching (Meads & Fitzgerald 2001, Early 1995), black light traps (Lim et al. 2003, Ryu et al. 2021), pitfall traps (e.g. Ryu et al. 2021), or yellow pan traps (e.g. Early 1995).

Plant species names were updated and standardized using The Plant List (The Plant List 2013) using “GBOTB.extended.TPL” in the package “U.Taxonstand” (Zhang & Qian 2022). We constructed a phylogeny of plant species based on the megaphylogeny for seed plants (Smith & Brown 2018, updated by Jin & Qian 2019), using the package “V.PhyloMaker2” (Jin & Qian 2022).

We used the package “picante” to calculate the indices of plant phylogenetic diversity: ‘mean pairwise distance’ (MPD) and ‘mean nearest taxon distance’ (MNTD) across the plant species in each community (Webb 2000, Webb et al. 2002, Pavoine & Bonsall 2011). MPD measures PD across the whole phylogeny (including both ancient and recent diversification), whereas MNTD measures PD among the tips of the phylogeny (including only recent diversifications).

To make the indices independent of species richness, we used the standardized effect sizes of both indices (Miller et al. 2017): 

SES.MPD = (MPDobs – MPDnull) / sd(MPDnull),
SES.MNTD = (MNTDobs – MNTDnull) / sd(MNTDnull),

where MPDobs/MNTDobs is the observed value in a community, MPDnull/MNTDnull is the average of the expected value in the randomized communities (n = 1000 randomizations), and sd(MPDnull)/sd(MNTDnull) is the standard deviation of 1000 null values. Negative values of SES.MPD and SES.MNTD indicate that species tend to be more closely related in a community, whereas values greater than zero indicate that species tend to be more distantly related.

Due to the lack of a global phylogeny for insect species, we were not able to calculate the indices of phylogenetic diversity for insects. We, therefore, used the number of taxonomic groups as a proxy for phylogenetic diversity in insects (Zou et al. 2020). We calculated the average number of insect families per order as a proxy for more ancient diversifications (similar to plant SES.MPD), and the average number of insect genera per family as a proxy for more recent diversifications (similar to plant SES.MNTD). Though simple and coarse, the numbers of families and genera have been used as proxies for evolutionary relationships when complete phylogenies are not available (e.g. Keith et al. 2005, Brehm et al. 2013). A low value of these proxies indicates a phylogenetically diverse community (i.e., high PD) whereas a high value indicates a phylogenetically poor community (i.e., low PD).

We performed Moran's I test on the plant and insect diversity metrics to detect spatial autocorrelation. We studied the relationships between plant and insect phylogenetic diversity (plant SES.MPD vs. number of insect genera per family, and plant SES.MNTD vs. number of insect families per order) across all islands, using a spatial autoregressive model, and calculating correlations between the spatially detrended variables. We studied the effect of island parameters (distance, area, age, and type) on plant and insect phylogenetic diversity by using a linear regression model with generalized least squares (GLS) that accounts for spatial autocorrelation. We chose the most suitable model based on the Akaike Information Criteria (AIC). All data analysis was conducted in R Studio (R Core Team 2022).