Abstract
Variation in leaf morphology in plant species predicts long-term growth and yield. Consequently, we aim to understand the genetic basis of natural variation in poplar leaf morphology as an avenue to maximize biomass accrual. Multilocus GWAS and deep learning genomic prediction were used to investigate the genetic architecture of twelve correlated traits representing leaf size and shape in a diverse population of 313 Populus balsamifera L. genotypes. 94 significant associations were detected, with 70 associations unique to a single trait, and 24 were detected in association with more than one trait. We developed genomic selection models to predict leaf morphology in novel genotypes using a strategy called GWADL (Genome-Wide Association enriched Deep Learning). We detected significant SNP-trait associations in the poplar TOR orthologue and likely upstream activating kinases SnRK3 and SnAK1. The most significant polymorphism, explaining variance in tip angle, leaf mass-per-area, and serration density, was detected in association with TERPENE SYNTHASE5, PbTPS5. Exogenous application of sesquiterpenes β-eudesmol and 1αH,5αH-Guaia-6-ene-4β,10β-diol in developing young poplar leaves resulted in significantly smaller mature leaves. This study provides a genetic and mathematical foundation for improving poplar performance by optimizing leaf morphology, and importantly identified a novel role for the sesquiterpene synthase PbTPS5 in normal plant growth and development.