Conserved and divergent gene regulatory networks for crop drought resistance

Understanding crop drought resistance mechanisms is critical for enhancing resilience to intensifying climate change. However, the conserved and divergent drought resistance mechanisms under the genetic basis across species remain unclear. Here, using an innovative graph-based deep learning framework, we construct drought-responsive large-scale gene regulatory networks (GRNs) across transcriptional, proteomic, and metabolic layers in major Poaceae crops integrating over 5000 bulk RNA-seq datasets to map 3.3 million interactions among 130,000 genes. Other crops in Fabaceae and Solanaceae are also included for further validation. Large-scale data and advanced algorithms provide rigorous, genome-wide evidence and insights into drought GRNs. We find robustly conserved interaction patterns of TCP–PP2C and ERF–2OGD linked to abscisic acid and redox regulation. Divergent mechanisms are identified, including SPL–PELP in rice/wheat lineage and ERF–Psb28 in maize/sorghum lineage from large-scale comparative network analysis. We propose the hypothesis that drought resistance in crops is phylogenetically constrained by the topological structure of GRNs, rather than just a few key genes. Drought resistance of representative C3 crops, with inherited growth-related drought escape and avoidance strategies, is governed by a concentrated network with drastic reprogramming, while drought resistance of representative C4 crops, with genetically determined drought tolerance strategy, is governed by a distributed and stable network. Our findings offer deep insights into the evolutionary dynamics of GRNs in crops, revealing how gene network architecture has been shaped by natural selection to drive genetic adaptation in response to climate challenges. Deng et al. leverage AI and large-scale transcriptome data to map drought-response gene networks, uncovering conserved and lineage-specific regulatory patterns, and revealing that drought resistance is shaped by network architecture, not only individual genes.