A two-way molecular dialogue between embryo and endosperm is required for seed development

See allHide authors and affiliations

Science  24 Jan 2020:
Vol. 367, Issue 6476, pp. 431-435
DOI: 10.1126/science.aaz4131

Filling in the gaps

In a plant seed, the embryo lies dormant surrounded by nutritive endosperm while awaiting suitable conditions to germinate. A hydrophobic cuticle around the embryo protects it from catastrophic water loss during the early days of growth. Doll et al. identified a back-and-forth signaling pathway that ensures an intact cuticle. The precursor of a signaling peptide is made in the embryo and transferred to the endosperm, where it is processed into an active form. The activated peptide diffuses back into the embryo to activate receptor-like kinases that drive cuticle development. Serve and return continues until all leaks in the cuticle are filled in and the peptide can no longer cross the barrier.

Science, this issue p. 431


The plant embryonic cuticle is a hydrophobic barrier deposited de novo by the embryo during seed development. At germination, it protects the seedling from water loss and is, thus, critical for survival. Embryonic cuticle formation is controlled by a signaling pathway involving the ABNORMAL LEAF SHAPE1 subtilase and the two GASSHO receptor-like kinases. We show that a sulfated peptide, TWISTED SEED1 (TWS1), acts as a GASSHO ligand. Cuticle surveillance depends on the action of the subtilase, which, unlike the TWS1 precursor and the GASSHO receptors, is not produced in the embryo but in the neighboring endosperm. Subtilase-mediated processing of the embryo-derived TWS1 precursor releases the active peptide, triggering GASSHO-dependent cuticle reinforcement in the embryo. Thus, a bidirectional molecular dialogue between embryo and endosperm safeguards cuticle integrity before germination.

In angiosperms, seeds comprise three genetically distinct compartments: the zygotic embryo, the endosperm, and the maternal seed coat. Their development must be tightly coordinated for seed viability. In this work, we have elucidated a bidirectional peptide-mediated signaling pathway between the embryo and the endosperm. This pathway regulates the deposition of the embryonic cuticle, which forms an essential hydrophobic barrier separating the apoplasts of the embryo and endosperm. After germination, the cuticle—one of the critical innovations underlying the transition of plants from their original, aqueous environment to dry land—protects the seedling from catastrophic water loss (1, 2).

Formation of the embryonic cuticle has previously been shown to depend on two receptor-like kinases (RLKs)—GASSHO1/SCHENGEN3 (hereafter named GSO1) and GSO2—and on ALE1, a protease of the subtilase family (25). gso1 gso2 and (to a lesser extent) ale1 mutants produce a patchy and highly permeable cuticle (2). Mutant embryos also adhere to surrounding tissues, causing a seed-twisting phenotype (6). Because subtilases have been implicated in the processing of peptide hormone precursors (79), we hypothesized that ALE1 may be required for the biogenesis of the elusive intercompartmental peptide signal required for GSO1/2-dependent cuticle deposition.

CASPARIAN STRIP INTEGRITY FACTORs (CIFs), a family of small sulfated signaling peptides, are ligands for GSO1 and GSO2 (1012). CIF1 and CIF2 are involved in Casparian strip formation in the root endodermis (10, 11). The function of CIF3 and CIF4 is still unknown. To assess the role of CIF peptides in cuticle development, the quadruple mutant (cif1 cif2 cif3 cif4) was generated (fig. S1A). Neither cuticle permeability nor seed twisting phenotypes were observed in this quadruple mutant (fig. S1, B to E). However, reduction [in the leaky sgn2-1 allele (10)] or loss [in the tpst-1 mutant (13)] of tyrosyl-protein sulfotransferase (TPST) activity results in seed-twisting and cuticle-permeability phenotypes resembling those observed in ale1 mutants (Fig. 1, A to D, and fig. S2, A to D). These data suggest that a sulfated peptide may act as the ligand of GSO1/2 during seed development.

Fig. 1 TPST and ALE1 are required for maturation of the TWS1 peptide.

(A to C, E to H, and J, and K) Toluidine blue tests on etiolated cotyledons. Scale bars, 200 μm. (D and I) Quantification of toluidine blue uptake by the aerial parts of young seedlings, normalized to chlorophyll content. N = 6, 10 seedlings per repetition. *** indicates statistical differences with one-way analysis of variance (ANOVA) followed by a post hoc Scheffé multiple comparison test (P < 0.01) in (D) and (I). Error bars represent standard deviations. (J and K) Toluidine blue permeability of tws1-4 compared with Col-0. Scale bars, 400 μm (L to N) Transmission electron micrographs of the embryo (emb) to endosperm (end) interface at the heart stage. Scale bars, 200 nm. Genotypes are indicated, and gaps in the cuticle (cut) are shown by white arrows. (O) The predicted TWS1 active peptide sequence and alignment with four other known GSO ligands (CIF1, CIF2, CIF3, and CIF4). The site of predicted sulfation is indicated with a red asterisk. (P) Anti-His Western blot of protein extracts from N. benthamiana leaves, agro-infiltrated to express TWS1::GFP(His)6 (TWS1) or the empty vector (−). Coexpression of ALE1::(His)6 or the empty-vector control are indicated by + and −, respectively. (Q) Coomassie-stained SDS-PAGE showing recombinant GST-TWS1 and the indicated site-directed mutants digested in vitro with (+) or without (−) ALE1-(His)6 purified from tobacco leaves. Arrows indicate specific cleavage products. (R) The full length TWS1 precursor. Sulfation and ALE1 cleavage sites are indicated. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; N, Asn; P, Pro; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Consistent with the hypothesis that TPST acts in the same pathway as GSO1 and GSO2, no difference was observed between the phenotype of tpst-1 gso1-1 gso2-1 triple and gso1-1 gso2-1 double mutants (fig. S2E). In contrast, TPST and ALE1 appear to act synergistically, as a phenotype resembling that of gso1 gso2 double mutants was observed in tpst-1 ale1-4 double mutants (Fig. 1, E to I, and fig. S2, F to J). This result supports the hypothesis that TPST and ALE1 act in parallel regarding their roles in embryonic cuticle formation, possibly through independent posttranslational modifications that contribute to the maturation of the hypothetical peptide signal.

Identification of the peptide signal was facilitated by a study of TWISTED SEED1 (TWS1) (14), which reported a loss-of-function phenotype that was notably similar to that of gso1 gso2 double mutants. Because existing alleles of TWS1 are in the Wassilewskija (WS) background, we generated new CRISPR alleles (tws1-3 to tws1-10) in the Col-0 background and confirmed the phenotype of resulting mutants (Fig. 1 and fig. S3). No additivity was observed when loss-of-function alleles of TWS1 and of other pathway components (GSO1, GSO2, TPST, and ALE1) were combined, providing genetic evidence for TWS1 acting in the GSO signaling pathway (fig. S4). Furthermore, gaps in the cuticle of embryos and cotyledons, similar to those observed in ale1 and gso1 gso2 mutants (2), were detected in both the tws1 mutants and tpst mutants (Fig. 1, J to N, and fig. S5). Inspection of the TWS1 protein sequence revealed a region with limited similarity to CIF peptides, including a DY motif that marks the N terminus of the CIFs (Fig. 1O) and is the minimal motif required for tyrosine sulfation by TPST (15). Corroborating the functional importance of the putative peptide domain, the tws1-6 allele (deletion of six codons in the putative peptide-encoding region) and the tws1-5 allele (substitution of eight amino acids, including the DY motif) both showed total loss of function of the TWS1 protein (fig. S3).

We tested whether TWS1 is a substrate of ALE1 by coexpression of ALE1:(His)6 and TWS1:GFP-(His)6 fusion proteins in tobacco (Nicotiana benthamiana) leaves. A specific TWS1 cleavage product was observed upon coexpression of ALE1 but not in the empty-vector control, suggesting that TWS1 is processed by ALE1 in planta (Fig. 1P). Likewise, recombinant TWS1 expressed as GST-fusion in Escherichia coli was cleaved by purified ALE1 in vitro. (Fig. 1Q). Mass spectroscopy analysis of the TWS1 cleavage product purified from tobacco leaves showed that ALE1 cleaves TWS1 between His54 and Gly55 (fig. S6). These residues are important for cleavage site selection, as ALE1-dependent processing was not observed when either His54 or Gly55 was substituted by site-directed mutagenesis (Fig. 1Q). His54 corresponds to the C-terminal His or Asn of CIF peptides (Fig. 1O). Thus, the data suggest that ALE1-mediated processing of the TWS1 precursor marks the C terminus of the TWS1 peptide. Because the CIF1 and CIF2 peptides are located at the very end of their respective precursors, C-terminal processing could represent a mechanism of peptide activation operating in the developing seed but not in the root. A summary of TWS1 modifications is provided in Fig. 1R.

To test the biological activity of TWS1, the predicted peptide encompassing the conserved N-terminal DY motif and the C terminus defined by the ALE1 cleavage site was custom-synthesized in tyrosine-sulfated form. As synthetic TWS1 cannot easily be applied to developing embryos, a root bioassay for CIF activity was used. In wild-type roots, TWS1 induced ectopic endodermal lignification, as previously observed for the CIF1 and CIF2 peptides (12). TWS1 activity was GSO1-dependent, suggesting that processed TWS1 peptide can replace CIF1 and CIF2 as a ligand for GSO1 during Casparian strip formation (Fig. 2A and fig. S7). Supporting this, TWS1 application complemented the cif1 cif2 mutant, albeit with reduced activity compared with CIF2 (Fig. 2B and fig. S8). TWS1 activity in this assay was reduced when sulfation on the DY motif was missing (Fig. 2B). Versions of TWS1 in which Y33 was mutated to either F or T only partially complemented the mutant phenotype of tws1-4 (fig S9), consistent with a residual but weak activity for nonsulfated TWS1 in vivo and with the weak loss-of-function phenotype of the tpst-1 mutant.

Fig. 2 The TWS1 peptide is a functional GSO1/GSO2 ligand.

(A) Root overlignification following treatment with the active CIF2 or TWS peptide in Col-0 and in the gso1 (sgn3-3) background. Lignin is stained in purple and CASP-GFP fusion protein, marking the Casparian strip domain, in green. Scale bar, 5 μm. (B) Complementation of cif1-2 cif2-2 Casparian strip integrity phenotype by peptide treatments. Number of gaps in CASP1-GFP signal counted after treatment with CIF2 sulfated peptide, TWS1 sulfated peptide, TWS1 nonsulfated peptide. N = 10. a, b, and c correspond to classes statistically supported by one-way ANOVA analysis, followed by Tukey tests (P < 0.05). (C to F) Grating-coupled interferometry (GCI)–derived binding kinetics. Shown are sensorgrams with raw data in red and their respective fits in black. ka, association rate constant; kd, dissociation rate constant; KD, dissociation constant. (C) Data for the GSO1 extracellular domain in the presence of the sulfated TWS1 peptide. (D) Data for the GSO2 extracellular domain in the presence of the sulfated TWS1 peptide. (E) Data for the GSO1 extracellular domain in the presence of the nonsulfated TWS1 peptide. (F) Data for the GSO2 extracellular domain in the presence of the nonsulfated TWS1 peptide. n.d., not determined.

To confirm TWS1 as a ligand of GSO1 and GSO2, the interaction of the synthetic peptide with the leucine-rich repeat (LRR) ectodomains of the receptors was analyzed in grating-coupled interferometry binding assays. GSO1 bound sulfated TWS1 with a KD (dissociation constant) of ~30 nM (Fig. 2C). The observed binding affinity is ~1/10 that of the CIF2 peptide (KD = 2.5 nM) (fig. S10), which is consistent with the reduced ability of TWS1 to complement the root phenotype of the cif1 cif2 double mutant (Fig. 2B). Sulfated TWS1 also bound to the LRR domain of GSO2, albeit with slightly reduced affinity (KD ~ 100 nM) (Fig. 2D). As previously shown for other CIF peptides (11), tyrosine sulfation was critical for the interaction of TWS1 with GSO1 and GSO2 in vitro (Fig. 2, E and F). Technical issues at high peptide concentrations may explain discrepancies between in vitro binding assays and the in vivo activity of nonsulfated TWS1. In vivo activities for nonsulfated versions of other normally sulfated peptides, including CIF2, have been reported (11, 1618). Adding a 3AA C-terminal extension to the sulfated TWS1 peptide reduced binding affinity to both GSO1 and GSO2 (fig. S10), consistent with the need for ALE1-mediated C-terminal processing for efficient signaling.

Taken together, our results suggest the sulfated TWS peptide as the missing link in the intercompartmental signaling pathway for embryonic cuticle formation. The activities of ALE1 and TPST both contribute to the formation of the bioactive peptide (Fig. 1R), which is perceived by GSO1 and GSO2 to ensure appropriate cuticle deposition.

To understand how the elements of the signaling pathway cooperate to ensure the formation of a functional cuticle, we analyzed their spatial organization. In silico data indicate that the TPST gene is expressed in all seed tissues (fig. S11) (19, 20). To investigate in which compartment TPST [which acts cell autonomously (13)] is required for TWS1 maturation, reciprocal crosses and complementation assays using tissue-specific promoters were performed. No cuticle permeability defects were observed when homozygous mutants were pollinated with wild-type pollen, confirming their zygotic origin. (Fig. 3, A to C). Expressing TPST under the ubiquitously active RPS5A promoter (21) or the PIN1 promoter [which is embryo-specific in seed (fig. S12)] complements tpst-1 cuticle defects. In contrast, no complementation was observed using the endosperm-specific RGP3 promoter (22), indicating that TPST activity is required for TWS1 sulfation specifically in the embryo to ensure cuticle integrity (Fig. 3D and fig. S13). Consistent with this observation and with a previous report (14), the TWS1 promoter was found to drive expression specifically in the developing embryo from the early globular stage onwards (Fig. 3E and fig S14). The TPST promoter (10) drove expression throughout the embryo proper at the onset of embryo cuticle establishment (the globular stage) before becoming restricted to the root tip (fig. S11). We conclude that the TWS1 peptide is both sulfated and secreted specifically in the embryo.

Fig. 3 Spatial separation of ALE1 and TWS1 expression is critical for pathway function.

(A to C) F1 seedlings from reciprocal crosses stained with Toluidine blue. (D and N) Toluidine blue quantification as in Fig. 1. a to d indicate statistical differences with one-way ANOVA followed by a post hoc Scheffé multiple comparison test (P < 0.01). (D) Complementation of tpst-1 mutant with endosperm-specific expression of TPST (pRGP3::TPST), embryo-specific expression of TPST (pPIN1::TPST), and ubiquitous expression of TPST (pRPS5a::TPST) compared with tpst-1 and Col-0. Three independent lines were analyzed. (E) Confocal images of pTWS1::mCitrine::NLS-mCitrine reporter lines, signal in yellow, autofluorescence in red. Scale bars, 50 μm. (F and G) Dry seeds (scale bars, 400 μm) and chloral hydrate cleared seeds (9 DAP) (scale bars, 100 μm), respectively, from a line expressing ALE1 in the embryo in the tws1-4 background (pTWS1::ALE1 line#7). (H and I) Seeds from crosses of Col-0 pollen onto line#7. (J and K) Self-fertilized tws1-4 seeds as a control. (L and M) Seeds from a cross of Col-0 pollen on a tws1-4 pistil as a control. Results for three further independent transgenic lines are shown in figs. S18 and S19. (N) Complementation of tws1-4 mutants by expression of TWS1 in the endosperm. Four independent lines were analyzed. (O) Model for embryonic cuticle integrity monitoring. Left shows the wild-type situation before gap-filling (nascent cuticle), illustrating the diffusion and processing of TWS1 across the embryo-endosperm interface. Right shows the wild-type situation when the cuticle is intact, spatially separating signaling components and, thus, attenuating signaling.

However, production of mature TWS1 requires a C-terminal cleavage event that we have shown to be mediated by ALE1. ALE1 is expressed only in the endosperm (4, 23), on the opposite side of the nascent cuticle to the GSO1 and GSO2 receptors, which are localized on the membranes of the epidermal cells that produce the cuticle (figs. S15 to S17) (2). Our data therefore support a model in which activation of the GSO signaling pathway depends on the diffusion of the TWS1 peptide precursor to the endosperm, where it is cleaved and activated by ALE1 before diffusing back to the embryo to trigger GSO1/2-dependent cuticle deposition. An intact cuticle would separate the subtilase from its substrate, terminating signaling.

Expressing ALE1 in the embryo, under the control of the TWS1 promoter, provided support for this model. Multiple transformants were obtained in tws1 mutants, but not in the wild-type background. When tws1 plants from four independent plants carrying the pTWS1:ALE1 transgene were pollinated with wild-type pollen—introducing a functional TWS1 allele into the zygotic compartments and thus inducing colocalization of TWS1 precursors with ALE1, GSO1, and GSO2 in the embryo—premature embryo growth arrest was observed in all seeds. This leads to severe shriveling of all seeds at maturity (Fig. 3, F to M, and figs. S18 and S19). A proportion of seeds could, nonetheless, germinate to give developmentally normal plants (fig. S20), indicating that coexpression of all signaling components in the embryo—although detrimental to embryo development—does not lead to a complete loss of viability. Growth arrest may be due to constitutive embryonic activation of the GSO1/GSO2 signaling pathway, and stress-responsive genes shown to require GSO1/GSO2 signaling for expression in the seed (2) were upregulated in seeds coexpressing GSO1, GSO2, TWS1, and ALE1 in the embryo (fig. S21). We thus postulate that the spatial separation of the TWS1 precursor and the GSO receptors from the activating protease by cuticle is required for signaling attenuation.

We next tested if CIF1, CIF2, and TWS1 could complement tws1 and ale1 mutants when expressed in the endosperm (under the RGP3 promoter). All three peptides complemented tws1 mutants, confirming that retrograde peptide movement from endosperm to embryo is sufficient to allow integrity monitoring (Fig. 3N and fig. S22). Lack of full complementation could reflect suboptimal N-terminal processing or sulfation in the endosperm. CIF1 and CIF2 (lacking C-terminal extensions) complemented ale1 mutants much more efficiently than TWS1 (fig. S23). Weak complementation of ale1 by TWS1 may reflect the presence of redundantly acting subtilases in the endosperm, as suggested by the weak phenotype of ale1 mutants.

The proposed bidirectional signaling model allows efficient embryo cuticle integrity monitoring. The sulfated TWS1 precursor is produced by the embryo and secreted (probably after N-terminal cleavage of the pro-peptide) to the embryo apoplast. In the absence of an intact cuticular barrier, it can diffuse to the endosperm and undergo activation by ALE1 (and potentially other subtilases). Activated TWS1 peptide then leaks back through cuticle gaps to bind the GSO1 and GSO2 receptors and activate local gap repair (Fig. 3O). When the cuticle is intact, proTWS1 peptides are confined to the embryo where they remain inactive.

Our results demonstrate a role for a subtilase in providing spatial specificity to a bidirectional peptide signaling pathway. In contrast, the related CIF1-, CIF2-, and GSO1-dependent signaling pathway controlling Casparian strip integrity is unidirectional, negating the need for C-terminal cleavage-mediated peptide activation (10, 12). Both pathway components and their spatial organization differ between the two systems, suggesting an independent recruitment of the GSO receptors to different integrity monitoring functions within the plant.

Supplementary Materials

Materials and Methods

Figs. S1 to S23

References (2439)

References and Notes

Acknowledgments: We thank L. Lepiniec for providing the tws1-1 and tws1-2 seeds; C. Galvan Ampudia, Y. Jaillais, and L. Armengot for materials and helpful discussions; A. Creff, U. Glück-Behrens, A. Lacroix, P. Bolland, J. Berger, I. Desbouchages, and H. Leyral for technical assistance; A. Patole, B. Martin Sempore, C. Vial, and S. Maurin for administrative assistance; and B. Würtz and J. Pfannstiel (Core Facility Hohenheim) for mass spectrometric analyses. Transmission electron microscopy images were acquired at the Centre Technologique des Microstructures, Université Lyon 1. Funding: The study was financed by joint funding (project Mind the Gap) from the French Agence National de Recherche (ANR-17-CE20-0027) (to G.I.) and the Swiss National Science Foundation (NSF) (to N.G., supporting S.F.). N.M.D. was funded by a Ph.D. fellowship from the Ministère de l’Enseignement Supérieur et de la Recherche. Funding was also provided by NSR grant no. 31003A_176237 (to M.H.) and an International Research Scholar grant from the Howard Hughes Medical Institute (to M.H.). S.O. was supported by a long-term postdoctoral fellowship by the Human Frontier Science Program (HFSP). S.R. was supported by a Ph.D. fellowship from the Carl-Zeiss Foundation. Author contributions: G.I. led the study. G.I. and N.G. obtained funding for the study. G.I., N.G., A.Sc., M.H., T.W., and A.St. supervised the work. N.M.D., S.R., S.F., S.O., and S.C. carried out the experiments. All authors were involved in the analysis of the results. G.I., A.Sc., and N.M.D. wrote the paper with input from all authors. Competing interests: The authors declare no competing interests. Data and materials availability: All lines used in the study will be provided upon signature of an appropriate material transfer agreement. All data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article