Genomic imprinting in an early‐diverging angiosperm reveals an ancient mechanism for seed initiation in flowering plants

The evolution of the seed habit marks a pivotal innovation of the spermatophytes. Angiosperms further refined this trait by coupling the development of seed accessory structures to fertilization, optimizing resource allocation. Here, we show that post-fertilization auxin production is an evolutionarily conserved paternal trigger for the development of seed accessory structures in angiosperms. We also provide evidence that this pathway likely operates in the nourishing structure of gymnosperm seeds, but rather under maternal genetic control. We therefore propose a model in which the coupling of seed development to fertilization in the angiosperm common ancestors likely coincided with a change in parental expression of auxin biosynthesis genes. This evolutionary innovation likely contributed to the emergence of the endosperm, a defining feature underpinning the rapid diversification and ecological success of flowering plants. The plesiomorphic seed habit of spermatophytes is the most successful method of sexual reproduction in vascular plants. However, there are key differences between the seeds of its two clades, gymnosperms and angiosperms. In gymnosperm seeds, the embryo is the only biparental structure and it is nourished by a haploid maternal tissue derived from the proliferation of the megagametophyte (Linkies et al., 2010; Sakai, 2013). However, in angiosperms, the development of the seed nourishing structure, the endosperm, is tightly linked to fertilization. This is because in angiosperms, both the embryo and the endosperm are biparental, and result from the fertilization of two maternal gametes, the egg cell and the central cell, by two paternal sperm cells. These two fertilization products interact with each other, and the formation of the endosperm is indispensable for embryo viability (Lafon-Placette & Köhler, 2014). Moreover, these structures are surrounded by a seed coat of maternal sporophytic tissues, which provides various functions (Radchuk & Borisjuk, 2014), and whose development is dependent on signals originating in the biparental endosperm (Weijers et al., 2003; Roszak & Köhler, 2011; Figueiredo et al., 2016). This coupling to fertilization has obvious advantages, as it prevents the allocation of nutrients to ‘empty seeds’, which do not carry an embryo, and is therefore a major innovation of angiosperms. Using the model dicot species Arabidopsis thaliana, we previously demonstrated that the formation of the biparental endosperm is linked to the paternal expression of auxin biosynthesis genes (Figueiredo et al., 2015). This phenomenon, known as genomic imprinting, involves the parent-of-origin specific allele expression of certain genes (Batista & Köhler, 2020). The main auxin biosynthesis pathway is a two-step process, the first being catalyzed by TRYPTOPHAN AMINOTRANSFERASES (TAA/TARs) and the second by YUCCA (YUC) flavin monooxygenases (Mashiguchi et al., 2011). Before fertilization, the expression of genes encoding these enzymes is repressed in the maternal gametophyte by H3K27me3 repressive marks (Figueiredo et al., 2015; Moreno-Romero et al., 2019), which prevents auxin production in the absence of fertilization. However, the paternal alleles of TAA/TAR and YUC genes are not labeled with these repressive epigenetic marks and are thus expressed in the fertilized central cell (Figueiredo et al., 2015). This means that upon gamete fusion, auxin biosynthesis initiates in a paternal-dependent fashion, triggering the development of the endosperm and of the surrounding sporophytic tissues (Figueiredo et al., 2015, 2016). Consistently, exogenous applications or ectopic production of auxin in unfertilized Arabidopsis ovules lead to the formation of asexual endosperms and seed coats, also known as autonomous seeds (Figueiredo et al., 2015, 2016). Notably, genes involved in auxin biosynthesis are imprinted in the endosperm of other eudicots and of monocots (Luo et al., 2011; Waters et al., 2013; Hatorangan et al., 2016), and auxin biosynthesis seems to be a hallmark of seed nourishing structures (Figueiredo & Köhler, 2018; Florez-Rueda et al., 2024). However, the degree to which these mechanisms are conserved remains to be assessed. Here, we asked if the imprinting of auxin biosynthesis genes is evolutionarily conserved in angiosperms and found that, indeed, auxin drives seed development in early-diverging angiosperms which diverged c. 120–150 million years ago (Ma) from the mesangiosperms (which include the monocot and eudicot clades) (Zuntini et al., 2024). We propose that auxin biosynthesis in the seed nourishing structure was likely a feature of the gymnosperm ancestors and became coupled to fertilization in the angiosperm's most recent common ancestor. If the mechanisms of seed initiation are conserved in early-diverging angiosperms and in mesangiosperms, then these mechanisms most likely coupled seed development to fertilization at the origin of angiosperms. To test the hypothesis that auxin biosynthesis is an evolutionarily conserved mechanism for seed initiation, we generated the endosperm imprintome of the early-diverging water lily species Nymphaea caerulea (order Nymphaeales; Fig. 1a, Supporting Information Fig. S1). We used low-coverage genome sequencing to screen individuals of N. caerulea at the Botanical Gardens of Berlin and Potsdam, Germany, and identified two diploid self-incompatible individuals with a significant level of DNA polymorphisms (henceforth referred to as Plant 1 and Plant 2; Fig. S1; Table S1). We used these individuals for reciprocal crosses and collection of endosperm material using laser capture microdissection; we generated three libraries for each direction of the cross, corresponding to independent pollination events. Sequencing of endosperm-purified RNA-seq libraries allowed us to estimate allele-specific expression for 5441 genes in the N. caerulea genome (29% of the total genes expressed in the seed), for both directions of the reciprocal crosses (Fig. 1b; Tables S1, S2). The mean inferred maternal proportions were 0.5192 for Plant 1 and 0.4814 for Plant 2 (Table S1). Very close to 0.5, as expected for a diploid endosperm (Williams & Friedman, 2002). We thus identified 144 candidate maternally expressed genes (MEGs), as those that exceed our imposed threshold of 0.75 maternal proportion in both directions of the cross (Fig. 1b,c; Tables S1, S2). Symmetrically, we inferred 170 paternally expressed genes (PEGs), which show < 0.25 maternal proportion in both cross directions (Fig. 1b,c; Tables S1, S2). Of the 314 imprinted genes identified in the N. caerulea endosperm, we found 57 to be imprinted in at least one other species whose endosperm had been examined for imprinting (Fig. 1d; Table S2). Notably, there was an average of seven commonly imprinted genes between species. Such a low number of conserved imprinted genes is expected as imprinting has been shown to evolve rapidly (Hatorangan et al., 2016). Interestingly, we found comparatively more PEGs compared to a recent study on Nymphaea imprintomes (Povilus et al., 2025). However, there are two main differences between the two studies which may account for this discrepancy: (1) our analysis was done in an intraspecific cross, rather than interspecific; and (2) our endosperm samples were collected at an earlier time point. The endosperm in Nymphaea seeds is relatively small and devoid of nutrients, and the seed nourishing function is carried out by a maternal sporophytic tissue derived from the ovule nucellus called the perisperm (Fig. 1e; Povilus et al., 2015). It is thus reasonable to expect that paternal effects are increasingly reduced as the water lily seed develops, due to the limited role of the endosperm at later stages of development. In such a scenario, PEGs would mostly be detectable in early stages of endosperm development, as shown here, and become less prevalent in later stages of development, as described (Povilus et al., 2025). This points to genomic imprinting in the endosperm not being static throughout seed development, but rather a very dynamic phenomenon. We then asked which gene ontology (GO) terms were enriched in the set of imprinted genes in N. caerulea (Fig. S2). We found enriched GO terms pointing to pathways with known roles in seed development, including those involved in epigenetic processes (Table S2). Predicted protein interaction networks can be found in Fig. S2. Notably, genes involved in hormonal metabolism were also found in enriched functional categories, and this included genes involved in auxin biosynthesis, such as YUC10 (Table S2). In fact, the YUC10 is one of the only three genes that we found imprinted in endosperms of both N. caerulea and in species of monocots and eudicots (Table with our hypothesis that paternal expression of auxin biosynthesis genes is a conserved feature of To that processes are enriched seed development in we also of seed tissues of N. caerulea and three as embryo, endosperm and coat (Fig. on of the seed tissues et al., Tables The with using allowed us to the which in gene expression rather than time or identified three such originating from a the perisperm and one in the embryo (Fig. and two in of the endosperm (Fig. We that gene expression with the inferred of genes this more a functional of auxin seed the perisperm expressed the genes and the endosperm, the auxin biosynthesis YUC10 in also expression of key auxin and including and and the embryo, auxin genes such as and and were expressed Table This analysis that the expression of YUC10 is specific to Nymphaea a role of auxin in the development of three of its seed these the of a the genes expression that from the maternally derived perisperm the embryo and endosperm which increasingly in and This is also the first of the of gene expression at the level in an early To test water lily seeds auxin fertilization, we in and seed tissues of species of Nymphaea (Fig. early seeds to those of tissues, later stages of seed development coincided with an of auxin (Fig. Table We also these in water and Fig. Table that this feature is conserved in the In fact, auxin in seeds was shown in years ago et al., and this has been in both eudicots and monocots & et al., Figueiredo et al., 2015; et al., the that post-fertilization auxin is a feature of angiosperm To further our hypothesis that auxin is an evolutionarily conserved of seed development, we if exogenous of auxin is to pollination and lead to seed formation in early-diverging unfertilized N. caerulea ovules with auxin in than fertilized (Fig. Table and show of tissue of the sporophytic This the formation of a perisperm and the of the seed coat (Fig. However, fertilized seeds, these autonomous seeds are and the endosperm structure is also not and by only of it are (Fig. We thus seed as a for seed with a role of auxin in seed development in N. auxin applications show a in ovule and the of of auxin biosynthesis et al., 2011; et al., the of its fertilized seeds in a (Fig. Table to out that this role of auxin is specific to N. we the auxin applications to other species of with (Fig. Table The was for the water of the (Fig. we then our to which is a self-incompatible species of of the (Fig. we that auxin autonomous seed formation in which of the nucellus and of the surrounding sporophytic structures This means that the of auxin is not specific to the but is also for other early-diverging angiosperm To further the by exogenous auxin those by fertilization, we compared the of ovules and fertilized seeds of N. caerulea to those of ovules (Fig. Table seeds a set of expressed genes compared to unfertilized with the of endosperm development in with to an genes were commonly in both ovules and in fertilized seeds of pathways (Fig. Table This included of auxin biosynthesis, and such as and as as and and genes linked to seed coat and perisperm of these including and were also in the genes tissue of our of fertilized seeds (Table Fig. These that exogenous auxin the by fertilization, those to auxin and sporophytic tissue our points to post-fertilization auxin biosynthesis in the endosperm being a conserved of seed development in the angiosperms. This is by the that auxin biosynthesis genes are endosperm PEGs in Nymphaea and in species from which the diverged et al., 2011; et al., 2011; et al., 2011; Waters et al., 2013; et al., Figueiredo et al., 2015; et al., 2018; et al., This is as there is limited in imprinting (Hatorangan et al., et al., 2016), and auxin biosynthesis genes are thus the to that We thus that those genes a common evolutionary origin and to the & on our of seed nourishing tissues et al., we identified two and which the YUC10 and (Table with our of these are expressed in the endosperm of each angiosperm species and a by in Fig. et al., 2025). genes to these been described as PEGs in angiosperm endosperms by in Fig. our study of early-diverging angiosperm also an role for auxin in angiosperm endosperm development in those et al., in and N. the YUC10 and are expressed in and not in other seed or tissues (Fig. Table to the fertilization of the angiosperm megagametophyte is an innovation of the angiosperms, to the formation of the biparental embryo and However, in the seed nourishing function is by the haploid maternal which of fertilization. the development of the ovule in the gymnosperm which the of the is tightly linked to auxin et al., 2024). However, in angiosperms auxin are by fertilization, in these processes are linked rather to in the ovule et al., 2024). We thus that the development of the megagametophyte a nourishing structure be linked to the expression of auxin biosynthesis genes in both of if its parent-of-origin would be maternal in the gymnosperm and paternal in the angiosperm If that is then of the angiosperm endosperm auxin biosynthesis PEGs be expressed in the gymnosperm To test this we our previously of collected et al., and found that this was the the megagametophyte the and two YUC10 and (Fig. Table those genes the as angiosperm and (Table and to the YUC and TAA/TAR et al., 2025). us to that auxin production is a feature of at least gymnosperm that a role for auxin in the development of the tissues of the seed the emergence of the angiosperm and the origin of the endosperm the gymnosperm structure is of maternal the expression of auxin biosynthesis genes would be maternal in the and that the parent-of-origin expression to the in the angiosperms. These genes thus likely became PEGs the evolution of fertilization in the angiosperm common ancestor. However, the expression of auxin biosynthesis genes is of maternal origin in the gymnosperm seed nourishing structure, we out that and the also be involved in auxin the signals are that in the and trigger the development of the nourishing structure be an of We thus propose a model auxin is an evolutionarily conserved of seed development in spermatophytes (Fig. in the maternal expression of auxin biosynthesis genes in the nourishing seed structure is by by in angiosperms auxin biosynthesis in the endosperm is paternal in coupling the development of the seed nourishing structure and accessory tissues to fertilization. This in likely a post-fertilization biparental resource allocation to the seeds, thus to the emergence of the endosperm as a pivotal innovation that the evolutionary success of flowering plants. Nymphaea were in the Botanical Gardens of and Germany, in the of to self-incompatible species were for and was by the genome to genome of diploid and species the et al., 2013). and us to two individuals of the species Nymphaea caerulea in the of the Botanical of were and seed production was or which is of This was in to other species of the which seeds on the of Nymphaea caerulea was one of the Nymphaea species that we and in which of seeds were in the in we to imprinting, an was the of between we to at least two individuals that had polymorphisms between that allowed for specific expression and imprinting We therefore three individuals of Nymphaea and two of N. caerulea from the Berlin and Botanical We collected material and which was used for genome on the with We then the number of between the (Fig. S1; Table S1). These genome were used for imprinting as described on and were using individuals at the Berlin and Botanical To test the of auxin as an of seed development, we a number of species with a The were carried out in the of Berlin and the of N. N. N. N. and were used for the We also used a that but seeds This is called Nymphaea the we identified on first of were a by the of in the we used and which be and the to pollination in the pollination the were on as a control. the auxin were and then was to the on The of was to the of the to to the of to a of to the small of N. were carried out at the in Berlin on of as of an early angiosperm are and only one in the collection of the Berlin Botanical only and were carried This was also the for the of the as were not successful in of our in our pollination with the Nymphaea to in the of were collected a and by in The material was in the further as described set of N. caerulea was then in the of the of Plant in were carried out on these two throughout and to as 1 and We our to an earlier time and also our a the of a threshold for the of we a of auxin from to In to test the effects of auxin in the we exogenous applications of the auxin et al., and et al., to the We two of each and In were on the of the first and at least each was to the this set of were collected and at of the was and for as described seed the seeds were from tissue and to seed or ovules were in with and using a hormonal were in differences between were using the test on by from were in and are in Table To further the by auxin those by fertilization, we a analysis on N. were collected from at a and a the of the described each three were from seeds, which were from the and in was using and to for and were using and on an of the generated are in Table were to in The were at on a were on which were further and as were by in or for were a to tissue structure, were with for and with in for 1 a water the were rapidly a to to the the were in were on an using and seed was by the most central of seeds using of the set of of of N. caerulea at 1 pollination were for We to of the main of the Nymphaea endosperm, embryo, and We used three and were collected from the of the Botanical a These were later on the of Plant in and used for imprinting be described was as were a was in and The material was in the which in less than a were to in the were on a at were as previously described et al., In capture was using a the endosperm from the embryo and surrounding perisperm and seed coat tissue was laser using a at the and was or in the the were at was using the to the The and of total were with that was used for further the was using the for Sequencing was by DNA successful with from to libraries were with on an at in the of the generated are in Table crosses were on the two individuals of N. caerulea in the the of direction of the cross were used for the study and were collected 1 these endosperm tissue was and its was generated as described were previously at the to the of individuals for this in the and of the was with were by and with using the We used a to further from to maternal proportions gene of the was done the N. genome at using We generated three independent endosperm libraries cross direction the to capture allele-specific expression in the was with et al., 2011). We to only with a allele and allele-specific expression to imprinting in the endosperm the to not only of reciprocal but also of in as for the imprintome et al., In between the parental were using genome of both parental Using a we the of the alleles in each maternal proportions gene in each direction of the a gene was for each direction of the cross, we used for maternal proportions in reciprocal crosses to a gene as imprinted and and imprinted as PEGs < and PEGs < maternal These are and with the of a maternal proportion of gene expression in the diploid endosperm of genes candidate and PEGs significant from the expected 0.5, as by with of N. caerulea imprinted genes with can be found in Table and of the in Table of imprinted genes in other were for with the of imprinted genes in the N. caerulea endosperm from et al., 2011; et al., & et al., 2016), et al., et al., et al., 2016), et al., et al., species of et al., and an cross of the water Nymphaea and Nymphaea (Povilus et al., were and for RNA-seq generated in this study were using the and analysis with expression to each were to the N. genome at using et al., and were with et al., and were generated for with than in at least two samples were the gene expression were using a with using et al., in these the endosperm libraries generated for imprinting in were also included in the We for expression between of seed and and and and were as those with < and To a set of genes expressed in each of the major seed embryo, endosperm, and we identified the set of genes expressed in a tissue as the of those in both including the were generated using the et al., in of the total and the of in each tissue are in Table Fig. a of the including those used for imprinting the hormonal on N. caerulea seeds a fertilization, and a auxin were between unfertilized fertilized seeds and unfertilized seeds in et al., 2014). were using the as < change To between ovules and fertilized seeds, the in was and its was by a test in for both and are in Table from functional was carried out in et al., and et al., using the expressed gene as the genes for the and genes for the hormonal enriched terms had < To further genes functional the in was with an of were generated with the in to expression of gene To further the seeds of N. we analysis of seed were collected 1 by which time had successful We the cell and the seeds to from the perisperm and to the endosperm and The and surrounding perisperm tissue were then to seeds were with at least in to were from the tissue using the the with were on and were The was with and and The tissue was in the for less than 2 and a This was by at for at the of was and the were a second was and the was reduced to were using a was using to the and sequencing was by libraries were generated from the N. caerulea sequencing were using and et al., to and were with et al., To we et al., and further for with & 2020). was to et al., from three libraries were using to for effects et al., was with the three genes for each were identified (Table and of the genes were using et al., the genes in this and cell were using et al., to dynamic gene expression and of seeds for was in the Botanical Gardens of and the two of were used for the one in and one in were but the success of the was very were because of the of the seeds as were collected at an early time to had and was later time at 2 was only and only a seeds were collected for were and and for at and tissue were points to seeds in each of the the Nymphaea seed a of species of the Nymphaea from the Botanical Berlin was used for in were and seeds the for but were on to be used for were and three time points of collection were an early time being to a by a in the seeds in most of the species and a later time at a time in which the seeds are and are or or tissue were and and for at points to independent of seeds species. We a that allowed for the of auxin and of its et al., 2020). In was using a The of water the of the the for further The of auxin and of its was with a previously described method et al., 2020). In of the samples was using 1 of an by a with of water with of of the the was and in of The samples were in a of a and a The of a with a The of a of water and The was as for from to from to at from to and to by to the was and to The was using a in The was using a for the of The were previously using the of was in the The of samples was on the total for this in each were in and are in Table We and at the Botanical of and the at the Botanical of Berlin for us to the and plants. We for with the for with the auxin and and and for with the seed This was by the to and a to and by and the the and the and the imprinting the and the with from generated is at under Fig. Nymphaea individuals used in this Fig. terms and interaction networks for imprinted genes in Nymphaea Fig. of genes on the of expression in tissue Fig. of cell on expression Fig. and analysis of gene expression in the seed of Nymphaea on genes. Fig. of auxin genes at the cell in the Nymphaea caerulea Fig. is fertilization in Fig. in an of including species of and in the Fig. gene expression in and Fig. of RNA-seq Table expression Table of imprinted genes and Table and Table gene and with seed Table and of Table auxin Table biosynthesis genes and Table of is not for the or of Supporting Information by the than be to the The is not for the or of by the than be to the corresponding for the The remains with to in and in