Introductory Paragraph: Seeds are a key lifecycle stage for many plants. Seeds are also the basis of agriculture and the primary source of calories consumed by humans ( 1 ). Here, we employ single-nucleus RNA-sequencing to generate a transcriptional atlas of developing Arabidopsis thaliana seeds, with a focus on endosperm. Endosperm, the primary site of gene imprinting in flowering plants, mediates the relationship between the maternal parent and embryo ( 2 ). We identify transcriptionally-uncharacterized nuclei types in the chalazal endosperm, which interfaces with maternal tissue for nutrient unloading ( 3 , 4 ). We demonstrate that the extent of parental bias of maternally expressed imprinted genes varies with cell cycle phase, and that imprinting of paternally expressed imprinted genes is strongest in chalazal endosperm. Thus, imprinting is spatially and temporally heterogeneous. Increased paternal expression in the chalazal region suggests that parental conflict, which is proposed to drive imprinting evolution, is fiercest at the boundary between filial and maternal tissues.

Early male and female function indicate that N. thermarum is predisposed towards self-pollination, a phenomenon that is likely to have evolved multiple times within Nymphaea. While formation of distinct micropylar and chalazal developmental domains in the endosperm, along with a copious perisperm, characterize the seeds of most members of the Nymphaeales, seed ontogenies vary between and among the constituent families. Floral biology, life history traits and small genome size make N. thermarum uniquely promising as an early-diverging angiosperm model system for genetic and molecular studies.

For more than 225 million y, all seed plants were woody trees, shrubs, or vines. Shortly after the origin of angiosperms ∼140 million y ago (MYA), the Nymphaeales (water lilies) became one of the first lineages to deviate from their ancestral, woody habit by losing the vascular cambium, the meristematic population of cells that produces secondary xylem (wood) and phloem. Many of the genes and gene families that regulate differentiation of secondary tissues also regulate the differentiation of primary xylem and phloem, which are produced by apical meristems and retained in nearly all seed plants. Here, we sequenced and assembled a draft genome of the water lily Nymphaea thermarum, an emerging system for the study of early flowering plant evolution, and compared it to genomes from other cambium-bearing and cambium-less lineages (e.g., monocots and Nelumbo). This revealed lineage-specific patterns of gene loss and divergence. Nymphaea is characterized by a significant contraction of the HD-ZIP III transcription factors, specifically loss of REVOLUTA, which influences cambial activity in other angiosperms. We also found the Nymphaea and monocot copies of cambium-associated CLE signaling peptides display unique substitutions at otherwise highly conserved amino acids. Nelumbo displays no obvious divergence in cambium-associated genes. The divergent genomic signatures of convergent loss of vascular cambium reveals that even pleiotropic genes can exhibit unique divergence patterns in association with independent events of trait loss. Our results shed light on the evolution of herbaceousness—one of the key biological innovations associated with the earliest phases of angiosperm evolution.