Genomic imprinting is a fascinating and complex epigenetic phenomenon that occurs in mammals and flowering plants, leading to a parentally biased allelic gene expression (Batista and Köhler, 2020). This phenomenon results from epigenetic changes, such as DNA methylation or histone modifications, that occur on single genes or gene clusters (Kawashima and Berger, 2014; Batista and Köhler, 2020). In essence, these epigenetic marks also called imprints are deposited in the gametes and silence the maternal or paternal allele after fertilization, therefore producing a parental effect during offspring development. The parental conflict theory, also known as the kinship hypothesis, is the most widely accepted explanation for imprinting evolution among various hypotheses (Haig and Westoby, 1989). The parental conflict theory posits that imprinting has evolved as a mechanism for parents to optimize their reproductive success by differentially investing in offspring according to their relatedness. In this context, paternally expressed imprinted genes are inclined to enhance the allocation of nutrients to the developing embryo, whereas maternally expressed imprinted genes act to restrain the potentially selfish behavior of paternally expressed genes by suppressing the excessive growth. The predictions of the parental conflict theory are broadly supported by the effect of parental dosage imbalance in flowering plant seeds. In reciprocal interploidy crosses, an excess of the maternal genome leads to small seeds, while an excess of the paternal genome has opposite effects (Scott et al., 1998; Feil and Berger, 2007).

In angiosperms, seed formation is initiated by double fertilization events. The fusion of one sperm with the egg gives rise to the embryo, while concurrently, another sperm fused with the central cell generates the endosperm (Li and Berger, 2012). Endosperm, an extraembryonic tissue, plays a pivotal role in the transfer of nutrients from the mother to the developing embryo, resembling the function of the placenta in mammals (Feil and Berger, 2007; Bleckmann et al., 2014). The embryo and endosperm are enveloped by the seed coat originating from the maternal integument. Remarkably, it is the endosperm, rather than the embryo, that plays a key role in coordinating the growth of three genetically distinct tissues of developing seeds (Li and Berger, 2012; Xiong et al., 2021). Genome imprinting is proposed to be involved in the regulation of seed growth, particularly through the endosperm (Scott et al., 1998; Li and Berger, 2012). Although hundreds of genes preferentially expressed from maternal or paternal alleles have been identified in the endosperm (Haig, 2013; Rodrigues and Zilberman, 2015; Batista and Köhler, 2020), most candidate imprinted genes have remained largely uninvestigated or do not exhibit a seed phenotype.

The HAIKU (IKU) pathway that controls early endosperm growth, might also control imprinted genes. The IKU pathway mainly comprises three core genes in Arabidopsis, namely IKU1, IKU2, and MINI3, encoding a VQ domain protein, a leucine-rich repeat kinase, and transcription factor WRKY10, respectively (Garcia et al., 2003; Garcia et al., 2005; Luo et al., 2005; Wang et al., 2010). Loss of IKU genes leads to early cellularization of endosperm and small seed size, which perfectly mimics the phenotype of seeds from maternal genome excess (Garcia et al., 2003; Xiao et al., 2006). Until now, none of the IKU genes have been identified as having imprinting characteristics. The regulation of cytokinin oxidase 2 (CKX2), a gene located downstream of the IKU pathway, is subject to epigenetic mechanisms that include DNA methylation and histone modification. However, CKX2 shows a biallelic expression pattern in early endosperm (Li et al., 2013).

In this study, we have successfully characterized a pair of maternally expressed genes, ICF1 and ICF2, which encode two closely related E3 ligases. These two genes originate from a recent duplication, and their promoter carries both CG and non-CG methylation that silences their paternal allele in endosperm. We showed that DNA demethylation in the central cell is required for the expression of the maternal allele of these genes but their transcriptional expression in endosperm depends directly on the IKU transcription factor WRKY10. We demonstrated that overexpression of these ICF genes partially rescues the small seed phenotype seen in iku mutants. As a result, our discovery of this pair of closely related imprinted genes, both involved in seed size regulation, offers insights into the regulatory aspects of genomic imprinting and their potential utility as seed size regulators in crops.