Genomic signals of local adaptation across climatically heterogenous habitats in an invasive tropical fruit fly (Bactrocera tryoni)

Abstract

Local adaptation plays a key role in the successful establishment of pest populations in new environments by enabling them to tolerate novel biotic and abiotic conditions experienced outside their native range. However, the genomic underpinnings of such adaptive responses remain unclear, especially for agriculturally important pests. We investigated population genomic signatures in the tropical/subtropical Queensland fruit fly, Bactrocera tryoni, which has an expanded range encompassing temperate and arid zones in Australia, and tropical zones in the Pacific Islands. Using reduced representation sequencing data from 28 populations, we detected allele frequency shifts associated with the native/invasive status of populations and identified environmental factors that have likely driven population differentiation. We also determined that precipitation, temperature, and geographic variables explain allelic shifts across the distribution range of B. tryoni. We found spatial heterogeneity in signatures of local adaptation across various climatic conditions in invaded areas. Specifically, disjunct invasive populations in the tropical Pacific Islands and arid zones of Australia were characterised by multiple significantly differentiated single nucleotide polymorphisms (SNPs), some of which were associated with genes with well-understood function in environmental stress (e.g., heat and desiccation) response. However, invasive populations in southeast Australian temperate zones showed higher gene flow with the native range and lacked a strong local adaptive signal. These results suggest that population connectivity with the native range has differentially affected local adaptive patterns in different invasive populations. Overall, our findings provide insights into the evolutionary underpinnings of invasion success of an important horticultural pest in climatically distinct environments.

Predicting species invasiveness with genomic data: Is genomic offset related to establishment probability?

“…Among the native range, population 1 exhibited the highest heterozygosity (Figure S4). This aligns with B. tryoni's historical records (Parvizi et al, 2023;Popa-Báez et al, 2020), so this population was selected as the source population to calculate GO.…”

“…BayPass analysis was performed here from the allele count file already formatted by Parvizi et al. (2023). To ensure comparability across the different estimators in the context of real data including missing genotypes, $$$ {\mathrm{gGO}}_{\mathrm{lfmm}} $$$ was estimated using the allele frequencies obtained from the allele count file, missing data (for three different SNPs in three different populations) being replaced by the mean of allele frequencies across the 28 populations.…”

“…For the LFMM analysis (and $$$ {\mathrm{gGO}}_{\mathrm{lfmm}} $$$ estimation), we included K = 3 latent factors following Parvizi et al. (2023). For the GF analysis, neutral population structure was accounted for by using the residuals of the LFMM analysis as input variable, similar to what was done in the simulation study (Note S2).…”

“…We used publicly available data on 28 populations of Bactrocera tryoni including 15 native and 13 non‐native populations (Figure S13 ), which were previously analyzed by Parvizi et al. ( 2023 ) and Popa‐Báez et al. ( 2020 ).…”

“…To ensure comparability across the different estimators in the context of real data including missing genotypes, gGO lfmm was estimated using the allele frequencies obtained from the allele count file, missing data (for three different SNPs in three different populations) being replaced by the mean of allele frequencies across the 28 populations. For the LFMM analysis (and gGO lfmm estimation), we included K = 3 latent factors followingParvizi et al (2023). For the GF analysis, neutral population structure was accounted for by using the residuals of the LFMM analysis as input variable, similar to what was done in the simulation study (Note S2).…”