Genetic markers of stomatal cluster development in Begoniaceae revealed through trait analysis assisted by interactive deep-learning.

Stomata of plants track the immediate demand for carbon dioxide for photosynthesis while limiting transpirational water loss. Solitary stomatal patterns are common, yet some land plants develop non-contiguous stomatal clustering where two or more stomata occur in groups and overlay a single air cavity. Clustering improves stomatal efficiency, reduces plant water-use and increases resilience to environment stress. How cluster development and physiology interact and integrate with the environment are open questions. Here we used TESSERA, a deep-learning platform for stomatal detection with an interactive interface for data review. Tracking Begonia stomatal clustering patterns across various Begonias we uncovered correlations for stomatal clustering traits. The stomatal parameter data was applied to identify genetic loci involved in Begonia stomatal development using quantitative trait locus (QTL) analysis. Combined with differential gene expression (DEG) to refine the candidate list, our analysis reveals known and potential new Begonia candidates in stomatal development. As a test of this knowledge, we cloned Begonia SPEECHLESS (BegSPCH), a loci identified in this screen and an established development-related gene in Arabidopsis. Unexpectedly, Arabidopsis spch-3 mutants transformed to express BegSPCH, but not AtSPCH, developed stomatal clusters. Thus, various molecular and environmental factors likely overlay transcriptional regulation in stomatal development.