Endosperm cell death promoted by NAC transcription factors facilitates embryo invasion in Arabidopsis

Summary

In flowering plants, two fertilisation products develop within the limited space of the seed: the embryo and the surrounding nutritive endosperm. The final size of the endosperm is modulated by the degree of embryo growth. In Arabidopsis thaliana, the endosperm expands rapidly after fertilization, but later gets invaded by the embryo that occupies most of the seed volume at maturity, surrounded by a single remaining aleurone-like endosperm layer^1–4. Embryo invasion is facilitated by the endosperm-expressed bHLH-type transcription factor ZHOUPI that promotes weakening of endosperm cell walls^5,6. Endosperm elimination in zou mutants is delayed and embryo growth is severely affected; the endosperm finally collapses around the dwarf embryo causing the shrivelled appearance of mature zou seeds^5–7. However, whether ZOUPI facilitates a mechanical endosperm destruction by the invading embryo, or whether an active programmed cell death (PCD) process causes endosperm elimination has been subject to debate^2,8. Here we show that developmental PCD controlled by multiple NAC transcription factors in the embryo-adjacent endosperm promotes gradual endosperm elimination. Misexpressing the NAC transcription factor KIRA1 in the entire endosperm caused total endosperm elimination, generating aleurone-less mature seeds. Conversely, dominant and recessive higher-order NAC mutants led to delayed endosperm elimination and impaired cell corpse clearance. Promoting PCD in the zhoupi mutant partially rescued its embryo growth defects, while the endosperm in a zhoupi nac higher-order mutant persisted until seed desiccation. These data suggest that a combination of cell wall weakening and PCD jointly facilitates embryo invasion by an active auto-elimination of endosperm cells.

Results and Discussion

ESR cells display cellular and genetic features of developmental PCD

Developmentally controlled PCD (dPCD) in plants is a highly organised, actively controlled process that involves specific genes, different from those regulating animal PCD^9,10. Different NAC-type TFs promote dPCD in specific tissues, such as VND6 and VND7 in the xylem^11,12, SOMBRERO in the lateral root cap¹³, ANAC087 and ANAC046 in the root cap and senescing leaves^14–17, KIRA1 (KIR1) in the floral stigma¹⁸, and ORESARA1 (ORE1), in senescing leaves, the stigma, and the endodermis overlying emerging lateral roots^18–20. Based on the expression of dPCD-associated genes such as PUTATIVE ASPARTIC PROTEINASE A3 (PASPA3) and BIFUNCTIONAL NUCLEASE 1 (BFN1) in the terminally differentiated endosperm^2,3, we hypothesised that an active dPCD process is responsible for endosperm elimination in Arabidopsis. We first analysed endosperm elimination dynamics by quantifying endosperm nuclei by flow cytometry (Figure 1A-C). Endosperm nuclei numbers increase until the bent-cotyledon stage before decreasing in later stages. This suggests the bent-cotyledon sage, at 8 days after pollination (8DAP), as tipping point when endosperm elimination outpaces endosperm proliferation (Figure 1C). We thus examined the spatial and subcellular features of endosperm cell death by transmission electron microscopy (TEM) at this stage (Figure 1D). The central endosperm cells distant from the embryo appear as regular differentiated cells with a large central vacuole and thin cell walls (Figure 1E). Closer to the embryo, cells show cytoplasmic ingrowths into the central vacuole, indicating a start of vacuole shrinking (Figure 1F). The cells directly adjacent to the embryo have a strongly reduced central vacuole indicating advanced vacuolar degradation, and high organellar density indicating cell volume reduction. Cell walls are barely visible, suggesting advanced cell wall degradation, and contain branched extracellular structures, possibly the remains of plasmodesmata (Figure 1G,H). In some cells, the plasma membrane has fragmented into several distinct portions that contain different kinds of apparently intact organelles (Figure 1H). At the interface with the embryo, degenerating organelles float in an extracellular matrix of fibrous material of similar aspect as the embryo sheath, suggesting a complete endosperm cell corpse clearance and incorporation into the embryo sheath (Figure 1I). DAPI staining of nuclei in vibratome-cut sections show that living central endosperm and aleurone cells have regularly shaped round nuclei, while cells adjacent to the embryo display highly irregular fragmenting nuclei (Figure 1J-N). These observations indicate that endosperm cell death execution is restricted to the ESR directly adjacent to the embryo.

Figure 1. ESR cells present molecular and morphological features of dPCD.

A) Ploidy profile of nuclei extracted from developing seeds at torpedo stage and analysed by flow cytometry. Diploid peaks (2C, 4C, 8C, 16C) correspond to embryo and seed coat nuclei. Triploid peaks (3C, 6C, 12C) correspond to endosperm nuclei. B) Number of nuclei per seed at different stage based on flow cytometry counting. 3<n<6, error bars = SD. C) Number of endosperm nuclei at different stages based on flow cytometry. The different ploidy levels are indicated. 3<n<6. D) Semi-thin section of an 8 DAP seed giving an overview on the regions shown in panels B-F. E-I) TEM images from an 8 DAP seed showing E) the center of the central endosperm; F) a region two cell layers away from the embryo, arrows indicating cytoplasmic invagination into the central vacuole: G-H) the cell layer closest to the embryo: I) the embryo-endosperm interface. J) Vibratome section of an 8 DAP seed stained with DAPI, giving an overview on the nuclei shown in detail in panels K-N. White = DAPI, red = tissue autofluorescence. K-N) Details of DAPI-stained nuclei located K) in the aleurone; L) in the central endosperm; M-N) in the ESR. O) Scheme showing in purple the different tissues analysed by RNAseq in²¹. EE = early endosperm (4 DAP); DAL = developing aleurone (7 DAP); ESR = embryo surrounding region (7DAP); TE = total endosperm (7DAP). P-Q) Heatmaps displaying gene expression levels determined by RNAseq ²¹. Expression levels are normalized by row for each gene, the legend indicates relative expression levels. P) Expression of dPCD-associated genes. Q) Expression of established dPCD-promoting NAC TFs expressed in the endosperm. R-U) Confocal images of 8 DAP vibratome-sectioned seeds. Blue = DAPI counter staining, green = GFP. R) pORE1:ORE1-GFP, S) pANAC046:NLS-GFP, arrow indicates expression in the inner integument, arrowheads in the ESR. T) pANAC087:NLS-GFP, U) pKIR1:KIR1-GFP. Abb.: en = endosperm, em = embryo, sc = seed coat, cw = cell wall, ep = embryo epidermis, ecv = extra cellular vesicles, n = nucleus, p = plast, v = vacuole, es = embryo sheath, mx = fibrous matrix, dec = dead endosperm cell. Scale bars: A,O-R = 100 μm B-C,H-K = 5 μm, D-F = 1 μm, G = 50 μm, S-T = 20 μm. See also Data S1 and Data S2.

We next analysed the expression pattern of established dPCD-associated genes³ using a recently published dataset of different endosperm domains²¹ (Figure 1O, Data S1). Strikingly, all dPCD-associated genes are strongly and specifically upregulated in the ESR (Figure 1P). We thus analysed the expression of TFs known to promote dPCD in other tissues by directly activating dPCD-associated genes^16,18. Four NAC TFs known to promote dPCD are strongly and specifically upregulated in the ESR, namely KIR1, ANAC087, ANAC046 and ORE1 (Figure 1Q, Data S2). Promoter-reporter lines confirm their ESR-specific expression in the endosperm (Figure 1R-U).

Together these results suggest that a NAC-regulated canonical dPCD process occurs in the ESR and is responsible for a controlled and gradual breakdown of endosperm cells at the endosperm-embryo interface.

KIR1 mis-expression accelerates endosperm cell death and elimination

We tested the importance of restricting putative dPCD to the ESR by expanding the KIR1 expression domain from the ESR to the entire endosperm. We used the TE2 promoter, driving endosperm-wide expression from the torpedo stage onwards (Figure S1A,B). In pTE2:KIR1 lines, embryo development appears to occur at the same rate as in the wild-type (Figure S1C). However, the entire endosperm, including the aleurone, has fully disappeared at 9 DAP, indicating both a precocious central endosperm elimination and an extension of cell death to the aleurone (Figure 2A,B). TEM on 8 DAP seeds suggest that the whole endosperm in pTE2:KIR1, including the normally surviving aleurone layer, has died. While aleurone cells adjacent to the embryo have been completely cleared, there are still remnants of more distant endosperm cells at this stage (Figure 2C-F, Fig. S1D-F), suggesting that the proximity of the embryo is important for corpse clearance. Notably, the embryo sheath seems unaffected in these lines (Figure 2F).

Figure 2. KIR1 mis-expression expands dPCD, generating aleurone-less seeds.

A) Semi-thin sections of 9 DAP seeds from Col-0 and two independent pTE2:KIR1 lines (#8, #9). B) Details of panels in A, centered on the aleurone region. Asterisks indicate the location of the fully eliminated endosperm. C) TEM image from a 8 DAP Col-0 seed centered on the endosperm. D) Semi thin section of a 8 DAP pTE2:KIR1 #8 seed giving an overview on the regions shown in panels F-G. E-F) TEM images from a 8 DAP pTE2:KIR1 #8 seed. E) Region where the embryo did not reach the integument yet. F) Region where the embryo reached the integument. G-J) Confocal images of 8 DAP vibratome-sectioned seeds. Blue = DAPI staining, red = tdTOMATO. G-H) Seed from the cross Col0 x pPASPA3:NLS-tdTOMATO. H) Detail of G. I-J) Seed from the cross pTE2:KIR1 #9 x pPASPA3:NLS-tdTOMATO. J) Detail of I. K-Q) Phenotypic analysis of two independent pTE2:KIR1 lines (#8, #9) compared to Col-0. K) Percentage of germination at different times after stratification (n=6). L) Images of pTE2:KIR1 #8 seedlings displaying regular and shoot first germination. M) Dormancy quantification : percentage of germination for non-stratified seeds growing in darkness at different times after ripening, n = 6. N) Images of seedlings at different times post-germination. O-P) Etiolated seedlings 9 days after stratification O) grown without sucrose, P) grown with 0,5% of sucrose. Q) 7 days after stratification etiolated seedlings stained with toluidine blue, as a proxy of cuticular defects. Abb.: Oi = outer integument, ii = inner integument, al = developing aleurone, emb = embryo, ce = central endosperm, es = embryo sheath. Scale bars: A,D,I,G = 100 μm, B,H = 20 μm, C,E,J = 10 μm, E = 2 μm, L,Q = 200 μm, N = 1mm, O,P = 5 mm. See also Figure S1.

Next, we investigated the expression patterns of BFN1 and PASPA3 by crossing pTE2:KIR1 with pPASPA3:NLS-tdTOMATO and pBFN1:NLS-tdTOMATO. In wild-type, BFN1 and PASPA3 are ESR-specific and never expressed in the surviving aleurone, while in pTE2:KIR1, both promoters were active in aleurone cells (Figure 2G-J, Figure S1G-J).

We conclude that in contrast to the central endosperm, the aleurone survives endosperm elimination by avoiding to activate dPCD gene regulatory networks, and that KIR1 misexpression is sufficient to eliminate the aleurone.

Aleurone-less seeds display deficiencies during germination and seedling establishment

pTE2:KIR1 misexpression generates aleurone-less mature seeds, a phenotype that has to our knowledge not been reported before in Arabidopsis. Up to now, aleurone functions have been assessed by analysing mutants in aleurone-expressed genes or by manual dissection after seed imbibition^22,23. These experiments have implicated the aleurone in the maintenance of seed dormancy^24–26, embryo cuticle remodelling²⁷, and seedling sustenance during germination^28,29. To characterize the consequences of aleurone absence in an otherwise morphologically intact seed, we selected two independent pTE2:KIR1 lines that were totally or almost totally devoid of aleurone cells in mature seeds (Figure S1K,L). We observed a delay in germination of up to 4 days indicating that the aleurone promotes germination (Figure 2K). In a few cases aleurone-less seeds germinated with the shoot emerging before the root, which we never observed in the wild-type (Figure 2L, Figure S2N). Aleurone-less seeds also show reduced dormancy (Figure 2M), and after germination, aleurone-less seedlings have a slower pace of development, likely caused by the absence of nutrients from the aleurone (Figure 2N). When grown in darkness, aleurone-less seedlings stayed smaller, a defect that was only partially compensated by the addition of sucrose (Figure 2O-P, Figure S2O). This indicates deficiencies other than reduced carbon supply caused by the absence of the aleurone, in line with observations on embryos that have been germinated after manual dissection of the aleurone²⁷. We cannot exclude, however, that precocious cell death of the endosperm compromised embryo filling prior to germination. Finally, we found that aleurone-less seeds have a more permeable cuticle (Figure 2Q, Figure S2P). These results demonstrate that the entirety of aleurone functions are crucial for a rapid and orderly germination, as well as for seedling establishment after germination.

ANAC046, KIR1, ANAC087, ORE1 regulate the pace of endosperm elimination

To test the hypothesis that a NAC-regulated dPCD process is responsible for endosperm elimination, we first expressed a KIR1 dominant negative mutant under a strong promoter expressed ubiquitously in the pre-cellularization endosperm (pEE:KIR1-SRDX)^18,21. Several independent lines displayed shrivelled seeds, although to a lesser extent than zou, indicating a reduction of embryo growth (Figure 3A, Figure S2A). Despite high activity of the pEE promotor during early endosperm development²¹, pEE:KIR1-SRDX seeds are undistinguishable from the wild-type until heart stage. However, from cell death onset at torpedo stage on, most seeds presented a strong embryo growth delay. Embryos had shorter sometimes twisted cotyledons, indicating an increased invasion resistance of the endosperm (Figure 3B,C, Figure S2B). Phenotype penetrance was variable between sibling seeds, some appearing almost like the wild-type, while others showed a very reduced embryo (Figure 3B,C). In seeds at 9 DAP, significantly more endosperm cells persisted in comparison to the wild-type, indicating delayed endosperm elimination, although endosperm cells displayed the same mature morphology as in the wild type (Figure 3D-G). In addition, TEM analyses revealed that pEE:KIR1-SRDX embryos are surrounded by a mass of undigested cell corpse debris floating in a matrix that expands over several micrometres, indicating a substantial decrease in post-mortem corpse clearance (Figure 3G-I, Figure S2C-G). The embryo sheath appeared as a clear layer distinct from the surrounding cell debris, suggesting its formation does not depend on KIR1 (Figure S2E,F,G). When assessing the expression of dPCD-associated genes in KIR1-SRDX siliques by qPCR, we found both BFN1 and PASPA3 downregulated (Figure S2H,I). In crosses of pEE:KIR1-SRDX with the dPCD reporter line pPASPA3:NLS-tdTOMATO, we observed that seeds with a strong delay of embryo growth had the most clearly reduced pPASPA3 activity (Figure 3J). These results support that dPCD inhibition by KIR1-SRDX causes an impaired endosperm elimination with a negative impact on embryo growth.

Figure 3. dPCD-promoting NAC TFs promote endosperm elimination facilitating embryo expansion.

A) Dry seeds from Col-0 and two independent pEE:KIR1-SRDX lines (#18 and #14). B) Images of chemically cleared pEE:KIR1-SRDX seeds at 8,5 DAP representing the four main phenotypes observed. C) Proportion of seeds corresponding to the phenotype classes showed in B. D-E) Semi thin sections of 9 DAP seeds giving an overview on the region shown in detail in panels F-I. D) Col-0, E) pEE:KIR1-SRDX #18. F-I) TEM picture of 9 DAP seeds. F,H) Col-0. G,I) pEE:KIR1-SRDX #18. J) Vibratome sections of seeds from the cross: pEE:KIR1-SRDX #18 x pPASPA3:TdTOMATO. The three seeds shown are from the same silique and are representative of the expression variation observed in 10 independent siliques. Blue = DAPI counter staining, Red = tdTOMATO. K-L) Dry seeds of K) Col-0 and L) anac046-1 anac087-1 kir1-1 ore1-1. M-N) SEM images of dry seeds. M) Col-0, N) nac_4x. O-P) Semi-thin sections of 9 DAP seeds. O) Col-0, P) nac_4x. Q) RT-qPCR on developing siliques 8 DAP for WT and nac_4x for BFN1. BFN1 expression is normalized by the expression of housekeeping gene EiF4a. *** indicates a statistical difference by t-test with pvalue < 0,001. R) Ratio of the endosperm-embryo area measured on the longitudinal plane of chemically cleared seeds for Col-0 and nac_4x at different developmental stages (6, 8, 9, 10 and 12 DAS). Statistical differences by t-test are indicated by * for pvalue < 0,05, *** for pvalue < 10^-4, **** for pvalue < 10^-9. S) Ratio of endosperm-embryo area measured on the longitudinal plan of chemically cleared seeds for Col-0, nac_4x, nac_4x_cp and nac_8x mutants. a, b and c indicate statistical differences revealed by one-way ANOVA followed by post-hoc Tukey test. Abb.: ii = inner integument, al = developing aleurone, emb = embryo, ce = central endosperm, es = embryo sheath, ucc = mass of undigested cell corpses. Scale bars: A,K,L = 500 μm, B,D,E,J,M-P = 100 μm, F,G = 10 μm, H,J = 1 μm. See also Figure S2 and S3.

Next, we investigated the consequence of knocking out NACs expressed in the ESR. As mature dry seeds of kir1-1 ore1-1¹⁸ and anac087-1 anac046-1¹⁶ appear normal, we generated a anac046-1 anac087-1 kir1-1 ore1-1 quadruple (nac_4x) mutant by crossing. Mature dry seeds of nac_4x appeared less plump, having more pronounced ridges, indicating a smaller embryo and reduced endosperm elimination (Figure 3K-N). Section of seeds at 9 DAP confirmed longer persisting central endosperm cells (Figure 3O-P). Quantifying the ratio between endosperm and embryo area in the longitudinal axis of chemically cleared seeds, we found that nac_4x started to display a significantly higher endosperm/embryo ratio than the wild-type from 8 DAP on (Figure 3R). Embryos have also relatively shorter cotyledons, reminiscent of pEE:KIL1-SRDX embryos (Figure S3A), suggesting a higher endosperm invasion resistance. We measured similar phenotypes in an independent quadruple mutant generated by CRISPR/Cas9 (nac_4x_cp) (Figure 3S, Figure S3A,F). As quadruple mutants show less reduction of endosperm elimination than most pEE:KIR1-SRDX seeds, we wondered whether there was more extensive genetic redundancy among NAC TFs. Hence, we mutated an additional four NACs upregulated in the ESR (ANAC003, ANAC010, ANAC019, ANAC079) (Figure S3B,F). However, this nac_8x mutant showed only a slight increase in endosperm/embryo ratios and a similar cotyledon/hypocotyl ratio compared to nac_4x (Figure 3S, Figure S3A). As in the KIR1-SRDX lines, the expression level of dPCD associated genes was reduced (Figure 3Q, Figure S3E), and a promoter-reporter line indicated a strong downregulation of pBFN1 in nac_4x endosperm (Figure S3C,D). These data show that NAC TFs, including but not limited to KIR1, ORE1, ANAC046, and ANAC087 control a dPCD pathway that regulates the pace of endosperm elimination and cell corpse clearance.

Endosperm dPCD is not absent, but delayed in zou mutants

In comparison with these NAC loss-of-function lines, zou mutants have been described to show a pronounced block of endosperm elimination^2,5. To test if ZOU acts upstream of ESR-expressed NACs, we investigated endosperm development in zou-4 mutants in DAPI-stained vibratome sections Until 13 DAP, DAPI-stained nuclei were clearly discernible in the endosperm, though they started disappearing from 14 DAP on, indicating central endosperm cell death without subsequent cell elimination (Figure 4A). The dead endosperm finally collapsed, resulting in visually shrivelled seeds from 17 DAP on (Figure 4I). However, the zou embryo remained small, possibly because endosperm cell death happens too late for embryo growth to resume (Figure 4A,I). Interestingly, dPCD promoter-reporters show activation from 11 DAP (pPASPA3) and 13 DAP (pBFN1) onwards (Figure 4B, Figure S4A). In line with these findings, ESR-expressed NAC TFs show delayed upregulation in the zou mutant (Figure 1R-U). pANAC087 activity starts at 9 DAP in the ESR and spreads throughout the central endosperm between 11 and 13 DAP. pANAC046 only shows a strong activity from 13 DAP on. However, neither pKIR1 nor pORE1 activity was detectable at any stages of zou seed development (Figure 4C). In conclusion, these results suggest that endosperm dPCD is not blocked, but rather delayed, in zou mutant seeds.

Figure 4. NACs and ZOU regulate two partially independent pathways contributing to endosperm elimination.

A-C) Confocal images of vibratome-sectioned seeds. A) zou-4 mutant seeds at different stages (7, 10, 13, and 14 DAP from left to right) and stained by DAPI. B) zou-4 mutant seeds expressing pPASPA3:NLS-tdTOMATO at different stages (8, 9, 11, and 13 DAP from left to right). Red = tdTOMATO, blue = DAPI. C) Confocal images of zou-4 mutant seeds expressing pANAC046:NLS-GFP, pANAC087:NLS-GFP, pORE1:ORE1-GFP, pKIR1:KIR1-GFP at different stages (8, 9, 11, and 13 DAP from left to right). Green = GFP. Blue = DAPI. D) SEM images of dry seeds of Col-0, zou-4 and two independent lines of pTE2:KIR1 and pTE2:DT-A in zou-4. The underneath embryo is highlighted in green. E-H) Confocal images of vibratome-sectioned seeds stained by DAPI. E) pTE2:KIR1 #4 in zou-4. F) pTE2:DT-A #5 in zou-4. Arrowheads indicate undigested endosperm cell corpses. G-H) Zoom on the embryo surface area. Asterisks indicate the location of the former endosperm. G) pTE2:KIR1 #4 in zou-4 H) pTE2:DT-A #5 in zou-4. I) Confocal images of vibratome-sectioned seeds of zou-4 and zou_nac at different stages after pollination and stained by DAPI. J) SEM images of dry zou-4 and zou_nac seeds at different stages after pollination. Location of the embryo and collapsed endosperm under the seed coat is indicated. K-L) Detail of zou_nac endosperm cells by confocal imaging of vibratome-sectioned seeds stained with DAPI. K) 21 DAP. L) 25 DAP. M) Germination quantification of freshly harvested Col-0, zou-4 and zou_nac mutant seeds. Scale bars: A-F,H = 100 μm, G,H,K,L = 20 μm. Abb.: al = developing aleurone, ce = central endosperm, emb = embryo, ii = inner integument, ucc = undigested cell corpses from the endosperm. See also Figure S4.

Early endosperm cell death rescues the zou embryo growth phenotype

To test if restoration of an earlier cell death was sufficient to rescue embryo growth in zou, we expressed KIR1 and a toxic protein used for genetic ablation, Diphtheria Toxin A (DT-A)³⁰, under the pTE2 promoter. In several independent lines of both constructs, the entire endosperm including the aleurone was killed at around 9 DAP. Interestingly, while endosperm death was followed by corpse clearance in KIR1 lines, cell corpses remained unprocessed in DT-A lines (Figure 4E-H, Figure S4C,D). This difference indicates that KIR1 restores a regular dPCD process including genetically controlled corpse clearance, while DT-A kills endosperm cells without subsequent corpse degradation.

Intriguingly, in lines of both constructs, mature dry seeds were substantially bigger than in zou, indicating a larger mature embryo (Figure 4D, Figure S4B). This partial rescue of zou embryo growth suggests that even in the absence of endosperm cell wall weakening by ZOU², endosperm cell death is sufficient to allow a large degree of embryo invasion. However, the early endosperm cell death did not restore the defective cuticle integrity of zou mutants^31,32 (Figure S4C,E, Figure 4E). These results demonstrate that the zou embryo growth defect is largely due to a delayed dPCD process in the zou mutant endosperm.

Combinations of zou and nac mutations cause a persistent endosperm

To test if NAC- and ZOU-controlled dPCD networks depend on each other in the ESR, we generated a quintuple zou-cp1 anac046-1 kir1-1 anac087-1 ore1-1 (zou_nac) mutant (Figure S4I). While early endosperm development in zou and zou_nac was indistinguishable (Figure S4G), the central endosperm in zou_nac mutants remained alive even after 14 DAP. When siliques started to visibly senesce at 21 DAP³³, zou nac seeds were still plump with an apparently viable endosperm (Figure 4I,J, Figure S4F). This persistent endosperm cells displayed large nuclei surrounded by a dense cytoplasm, a feature never seen in zou-4 or Col-0 (Figure 4K), suggesting a change in endosperm nature. Seeds only started to collapse around 25 DAP, at a stage of advanced desiccation marked by browning siliques³³. At this stage, central endosperm cells in zou_nac appeared dead, but cell corpses were not at all cleared, suggesting a failure to initiate endosperm dPCD (Figure 4I,L, Figure S4F). This persistent endosperm caused also a much higher water retention at 21 DAP when compared with wild-type or zou mutant seeds that likely interferes with seed desiccation (Figure S4H). Additionally, seeds of zou_nac displayed a delayed and uncomplete germination, indicating that the persistent endosperm decreases embryo fitness (Figure 4M).

In conclusion, as NAC expression and dPCD are delayed in the zou mutant, it appears that dPCD is controlled developmentally downstream of ZOU. However, the additive endosperm elimination phenotype in the zou_nac mutant indicates that ZOU and NAC-regulated dPCD pathways act at least partially in parallel pathways. The partial rescue of zou embryo growth by KIL1 or DT-A expression shows that timely endosperm cell death is crucial for embryo invasion. When dPCD in zou mutants finally starts at 14 DAP, embryos appear unable to resume growth and remain dwarfed. The mechanical strength of the endosperm and most other primary plant tissues is based on the principle of the hydrostatic skeleton, in which the turgor pressure of living cells is counteracted by the tensile strength of the plant cell wall³⁴. Or data show that both ZOU-controlled cell wall weakening as well as ZOU- and NAC-promoted dPCD are necessary to remove the mechanical resistance of the endosperm to facilitate rapid and effective invasive growth of the embryo. This information does not only shed a new light on the altruistic nature of the endosperm in the competition with the embryo for space in the confines of seed coat³⁵, but also paves the way for novel approaches to optimize the coordination of embryo and endosperm development in crops

Star Methods

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Moritz Nowack (Moritz.Nowack@psb.vib-ugent.be).

Materials availability

All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Experimental Model and Subject Details

Arabidopsis plants were grown on soil (JIFFy-7) in long-day growth chambers at 21°C (16h light / 8h dark). For in vitro culture, Arabidopsis seed were seeds gas-sterilised with chlorine gas, sown on ½ MS media (2.2g/L MS basal salt mixture, 0.1g/L MES, pH = 5,7, Plant agar = 8g/L), stratified for 2 days at 4°C in darkness and grown in a growth chamber under continuous light at 21°C.

Nac_4x corresponds to anac046-1 anac087-1 kir1-1 ore1-1 quadruple mutant. It was generated by crossing the anac87-1 anac46-1 double mutant reported by ¹⁹ with nac2-1 kir1-1 double mutant reported by ²¹. We have renamed nac2-1 into ore1-1 in the manuscript to be consistent with the gene name. Nac_4x_cp corresponds to anac046-cp1 anac087-cp1 kir1-cp1 ore1-cp1 quadruple mutant. It was generated by introducing a CRISPR destination vectors containing the guides against ANAC046 ANAC087 KIR1 ORE1 into Col-0 plants by floral dip transformation³⁶.

Nac_8x mutant corresponds to anac46-1 anac87-1 kir1-1 ore1-1 anac003-1, anac010-1, anac019-1 anac079-1. A CRISPR destination vectors containing the guides against ANAC003, ANAC010, ANAC19 and ANAC079 was transformed into the anac46-1 anac87-1 kir1-1 ore1-1 quadruple mutant

Zou-4 allele was reported by ⁵. The reporter lines pKIR1:KIR1-GFP and pORE1:ORE1-GFP were reported by ¹⁸, pANAC046:NLS-GFP and pANAC087:NLS-GFP by ¹⁶, pPASPA3:NLS-tdTOMATO by ³⁷, and pBFN1:NLS-tdTOMATO ³⁸. pKIR1:KIR1-GFP and pORE1:ORE1-GFP were introduced in zou-4 by crossing, while pANAC046:NLS-GFP and pANAC087:NLS-GFP by floral dip transformation. pPASPA3:NLS-tdTOMATO and pBFN1:NLS-tdTOMATO were transformed into nac_4x. pTE2:KIR1, pTE2:DT-A and pEE:KIR1-SRDX were transformed into Col-0 and zou-4. Zou nac_4x was created by transforming a CRISPR/Cas9 cassette against ZOU into nac_4x.

Methods

Seed staging

Seed staging was carried out by marking stage 14 flowers, corresponding to the stage when flowers undergo self-pollination ³⁹. The day after pollination (DAP) refers to number of days after marking the stage 14 flower.

In vitro culture of seeds and seedlings

For in vitro culture, seeds were gas-sterilised with chlorine gas overnight and sown on ½ MS media (2.2g/L MS basal salt mixture, 0.1g/L MES, pH = 5,7, Plant agar = 8g/L). Stratification was performed for 2 days at 4°C in darkness. Plants were then transferred into a growth chamber under continuous light at 21°C. For each experiment, variation from this standard protocol is indicated below.

Germination speed

The percentage of germination in each line was assessed for every timepoint by counting germinated seeds (visibly emerged radicle or cotyledons) and non-germinated seeds under a binocular dissection microscope. Additionally, pTE2:KIR1 lines, the percentage of seedlings with cotyledons emerging before the radicle was determined Each replicate corresponds to an individual seed lot coming from an independent mother plant.

Etiolated seedling length

For hypocotyl length measurement on etiolated seedlings, ½ MS plates were either supplemented with 1g/L of sucrose or not. After stratification, seeds were transferred into a growth chamber but kept in darkness by wrapping plates with an aluminium foil. To buffer the variation in germination speed between wild-type and pTE2:KIR1 lines and to let enough time for all the seedlings to reach their maximal size, we carried out the measurements 9 days after stratification. Plates were scanned on an EXPR. 12000XL EPSON scanner and seedling measurements were carried out using ImageJ software.

Dormancy

As Col-0 ecotype is weakly dormant, we assessed seed dormancy by using a specific protocol that exacerbate dormancy⁴⁰. Freshly matured seeds were harvested (week 0) and stored at room temperature. At different times post-harvest, seeds were sown on plates wrapped in aluminium foil to avoid light-triggered dormancy release. Plants were directly transferred into growth chambers, without any stratification. Germination percentages were assessed after 7 days to let the time for all the genotypes to fully germinate. Plates were scanned on an EXPR. 12000XL EPSON scanner and germinated and non-germinated seeds were counted using the cell counter plug-in using of ImageJ⁴¹. After counting, plates were transferred to light where a full germination rate was observed for all genotypes and post-ripening time points. Each replicate corresponds to an individual seed lot coming from an independent mother plant.

Cuticle integrity by toluidine blue

A toluidine blue assay was used as a proxy of seedling cuticle integrity, according to published protocols ⁴². Basically, young seedlings were stained during 2 minutes in a 0,05% aqueous solution of Toluidine blue O (EMS #22050). For imaging (Figure 2Q), seedlings were grown 7 days in darkness and images were taken with a LEICA M80 light microscope. Quantitative analyses were performed according to published protocols by extracting toluidine blue in an 80 % ethanol solution, and measuring light absorbance at 626nm and normalizing it by the chlorophyll absorbance at 430nm with a Genesys 10S UV-Vis spectrophotometer ⁴³. Each replicate corresponds to an individual seed lot coming from an independent mother plant.

Chemical clearing of developing seeds

Seeds were dissected from siliques coming from at least 5 independent plants and mounted in a clearing solution (8g of chloral hydrate, 3mL of water and 1mL of pure glycerol) to be cleared for at least 5 days at 4°C in darkness. For seedlings after bent-cotyledon stage, the testa was carefully perforated with a small needle (0.45x1.3mm) to allow penetration by the clearing solution. Brightfield images of cleared seeds were then taken with a BX51 Olympus microscope.

Estimation of embryo and endosperm size

Embryo and endosperm areas on the optical longitudinal mid-section plane of a seed were measured as proxies of their respective volumes⁴⁴. We selected chemically cleared seeds showing hypocotyl and cotyledons on the same plane. The focus of the microscope was adjusted to show the longitudinal mid-section of the seed, where the endosperm area on the Z axis is maximal. In the brightfield image, image J was used to measure the embryo and the endosperm areas (see Figure S3G). The size of the hypocotyl was measured by drawing a line from the root apex to the shoot apical meristem. The size of the cotyledons was estimating by drawing a straight line between the shoot apex and the most distant tip of the cotyledons (see Figure S3G).

Seed weight loss during desiccation

To estimate the water content of seeds during desiccation, siliques were selected at 21 DAP. Siliques at this stage were yellow to brown indicating the start of fruit desiccation. Seeds from 10 siliques from the same mother plant were harvested in 2mL Eppendorf tubes and weighed on an ABJ 120-4M microscale from Kern and then dried 1 week at 28 C. Seeds were then weighed again. The weight lost measured in the drying in the process was used as a proxy of water content at 21 DAP. Each replicate corresponds to an individual seed lot coming from an independent mother plant.

LR white sectioning

Developing seeds from different genotypes were fixed in ice-cold PEM buffer (50 mm PIPES, 5 mm EGTA and 5 mm MgSO4, pH 6.9) with 4% w/v of paraformaldehyde for 1h under vacuum. The samples were then dehydrated in a gradient pf ethanol and subsequently included in a gradient of LR white resin according to published protocols⁴⁵. 1 μm sections were then cut with a ultra-microtome (Leica EM UC6) and mounted on Superfrost Plus slides from Thermo Scientific. Sections were then stained 5 minutes in an aqueous solution containing 1% w/v of Toluidine blue O (EMS #22050) and 1 % w/v of Sodium Borate Tetra (EMS #21130). Images of the sections were taken with a BX51 Olympus microscope.

Transmission Electron Microscope (TEM)

Developing seeds of a 8 and 9 days after pollination were immersed in 20% (w/v) BSA and frozen immediately in a high-pressure freezer (Leica EM ICE; Leica Microsystems, Vienna, Austria). Freeze substitution was carried out using a Leica EM AFS (Leica Microsystems) in dry acetone containing 1% (w/v) OsO4 and 0.5% glutaraldehyde over a 4-days period as follows: -90°C for 54 hours, 2°C per hour increase for 15 hours, -60°C for 8 hours, 2°C per hour increase for 15 hours, and -30°C for 8 hours. Samples were then slowly warmed up to 4°C, rinsed 3 times with acetone for 20 min each time and infiltrated stepwise over 3 days at 4°C in Spurr’s resin and embedded in capsules. The polymerization was performed at 70 °C for 16 h. Semi-thin and ultrathin sections were made using an ultra-microtome (Leica EM UC6). Semi-thin sections were stained 5 minutes in an aqueous solution containing 1% w/v of Toludine blue O (EMS #22050) and 1 % w/v of Sodium Borate Tetra (EMS #21130) and imaged on a Zeiss (axiolab E) with a Axiocam ICC1 camera. Ultra-thin sections were post-stained in a Leica EM AC20 for 40 min in uranyl acetate at 20 °C and for 10 min in lead stain at 20 °C. Ultra-tin sections were collected on formvar-coated copper slot grids. Grids were viewed with a JEM 1400plus transmission electron microscope (JEOL, Tokyo, Japan) operating at 60 kV.

Imaging of seeds and young seedlings

Seeds were imaged either with a LEICA M80 light microscope or with a TM-1000 scanning electron microscope from Hitachi. Young seedlings were imaged with a LEICA M80 light microscope.

Cloning

pTE2:KIR1, pTE2:DT-A and pEE:KIR1-SRDX were cloned in a pB7 gateway destination plasmid that carry a FAST selection marker, by Gateway reaction using pTE2 and pEE in a pENTR P1 P4R entry vector, KIR1, KIR1-SRDX and DT-A in a pENTR P1 P2 entry vector. DT-A comes from ³⁰, KIR1 and KIR1-SRDX from ¹⁸, pEE from ²¹, pTE2, which consists of a 1,9kb fragment from the SAP collection⁴⁶. For CRISPR/Cas9 vectors, guides were designed on the CRISPR-P v2 software (http://crispr.hzau.edu.cn/CRISPR2/)⁴⁷ and were chosen among the most specific for the target gene. For guides against ANAC03, ANAC10, ANAC19, ANAC079, and guides against KIR1, ORE1, ANAC087, and ANAC046, 2 guides per genes were designed and combined in one vector for higher efficiency. For CRISPR against ZOU, 4 guides were designed and combined in one vector. Guides were combined into a pFASTRK-HMGP-AtCas9-NLS-P2A-mCherry-G7T-A-CmR-ccdB-G destination plasmid according to published protocols⁴⁸. In short, guides were first cloned by Gibson assembly into Green Gate entry plasmids that contain an AtU6 promoter. Primer sequences are described in Table S1. Entry plasmids with promoter and guides were combined into the destination plasmid pFASTRK-HMGP-AtCas9-NLS-P2A-mCherry-G7T-A-CmR-ccdB-G by Green Gate cloning⁴⁹. The final plasmids contain the AtU6 promoters driving each one guide, an AtCas9-NLS-P2A-mCherry driven by the ubiquitous HMG promoter and a FAST RED selection marker. Final plasmids were eventually transformed into Arabidopsis plants by floral dipping as previously described³⁶.

Generation of higher-order mutants

Nac_4x corresponds to anac046-1 anac087-1 kir1-1 ore1-1 quadruple mutant. It was generated by crossing the anac87-1 anac46-1 double mutant with nac2-1 kir1-1 double mutants^16,18. We have renamed nac2-1 into ore1-1 in the manuscript to be consistent with the gene name.

Nac_4x_cp corresponds to anac046-cp1 anac087-cp1 kir1-cp1 ore1-cp1 quadruple mutant. It was generated by transforming a CRISPR destination vectors containing the guides against ANAC046 ANAC087 KIR1 ORE1 into Col-0 plants. T1 seeds were selected with the FAST red marker. Mutations in the genes of interest were identified by PCR and sequencing using primers described in Table S1. T1 plants with the highest proportion of mutations were selected and homozygous multiple mutants were screened in the T2 and T3 progenies after a selection of FAST negative seeds to remove the CRISPR/CAS9 cassette. The mutant alleles generated are described in Figure S3F.

Nac_8x mutant corresponds to anac46-1 anac87-1 kir1-1 ore1-1 anac003-1, anac010-1, anac019-1 anac079-1. A CRISPR destination vectors containing the guides against ANAC003, ANAC010, ANAC19 and ANAC079 was transformed into the anac46-1 anac87-1 kir1-1 ore1-1 quadruple mutant and the multiple mutant generated as described previous. The mutant alleles generated are described in Figure S3F.

Zou_nac corresponds to the zou-cp1 anac46-1 anac87-1 kir1-1 ore1-1 quintuple mutant. It was generated by transforming a CRISPR destination vectors containing four guides against ZOU into nac_4x. zou-cp1 allele is described in Figure S4I.

Other transgenic lines used

The zou single mutant used is the zou-4 allele⁵. The reporter lines used were published previously: pKIR1:KIR1-GFP, pORE1:ORE1-GFP¹⁸, pANAC046:NLS-GFP and pANAC087:NLS-GFP¹⁶, pPASPA3:NLS-tdTOMATO³⁷, pBFN1:NLS-tdTOMATO³⁸.

Several homozygous transgenic lines were generated for pTE2:KIR1 and pTE2:DT-A and pEE:KIR1-SRDX after transformation in Col-0 or zou-4 plants. pKIR1:KIR1-GFP, pORE1:ORE1-GFP in zou-4 were generated by crossing with established reporter lines in the Col-0 background, while pANAC046:NLS-GFP and pANAC087:NLS-GFP construct were transformed into zou-4 plants.

Gene identifiers

The Arabidopsis Genome Initiative (AGI) codes of the genes mentioned in this paper are: ZOU (AT1G49770), ANAC03 (AT1G02220), ANAC010 (AT1G28470), ANAC019 (AT1G28470), ANAC046 (AT1G52890), ANAC074 (AT4G28530), ANAC079 (AT4G28530), ANAC087 (AT5G18270), ANAC092 (AT5G39610), PASPA3 (AT4G04460), BFN1 (AT1G11190), CEP1 (AT5G50260), SCPL48 (AT3G45010), RNS3 (AT1G26820), DMP4 (AT4G18425), EXI1 (AT2G14095), MC9 (AT5G04200), CAN1 (AT3G56170), TE2 (AT2G27380) and EE (AT5G09370).

Confocal observation

Before confocal observation, developing seeds were cut by a vibratome device. First, the two valves of the silique were manually removed and the seeds attached to the replum were fixed in 4% w/v of paraformaldehyde in ice-cold PEM buffer (50 mm PIPES, 5 mm EGTA and 5 mm MgSO4, pH 6.9) with 4% w/v for 1 hours under vacuum. Samples were then washed in PEM buffer and mounted in a 5 % agarose gel. Rectangles of agarose (Ultrapure agarose, Invitrogen, #16500-500) containing the samples were cut, glued on a carrier and sectioned with a Leica VT1200 S Vibratome. Parameters used for sectioning were a speed of 1.40 mm/s an amplitude of 0.6 mm and a section thickness of 120μm. Sections were mounted in PBS with or without 3mg/L of 4’,6-diamidino-2-phénylindole (DAPI). Sections were eventually observed either using a Zeiss LSM710 or a Leica SP8 confocal using 20X objectives. Excitation wavelengths were 405nm for DAPI, 514nm for tdTOMATO, 488nm for GFP and FDA. Emissions were captured between 418nm and 450nm for DAPI, 565nm and 612nm for tdTOMATO, 488nm and 554nm for GFP and FDA, and 595nm and 699nm for tissue autofluorescence.

For confocal observation of aleurone post-germination, seeds at 18 hours after imbibition were incubated in a solution of fluorescein diacetate (FDA). FDA solution was done by mixing a stock solution of FDA at 2 mg/ml in acetone into an aqueous solution with 10% sucrose until the solution gets trouble.

RNA-seq analysis

Data from RNA-seq of distinct endosperm domains²¹ were mapped on the Col-0 genome. Quality verification of raw RNA sequencing data was performed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were pre-processed using Trimmomatic⁵⁰ with 4 bases to average across and a required average quality of 20. Mapping was achieved using Salmon⁵¹ annotating to the latest Araport11⁵² reference genome. Differentially Expressed Genes (DEGs) were identified using the BioConductor database package EdgeR⁵³.

RT-qPCR

At least 4 siliques containing seeds at 8 DAP were harvested in 2mL Eppendorf tubes containing two 2mm beads. Samples were directly frozen in liquid nitrogen and subsequently homogenized with a MM40 Retsch device at 20 Hz. Total RNA was extracted with a ReliaPrep RNA tissue kit from PRomega that includes a step of on-column DNAse treatment. 1 μg of total RNA was used to perform reverse transcription with the qScript cDNA SuperMix 526 (Quantabio). The qPCR reaction was then performed in technical triplicates with a LightCycler 480 (Roche) using SYBR green for DNA detection. Gene expression was calculated according to published methods and normalized by the expression of EiF4a housekeeping gene⁴⁵.

Flow cytometry for seed ploidy analysis

For flow cytometry analysis, around 20 seeds were dissected from siliques, counted and chopped finely on ice in 100 μl of fresh Galbraith buffer (20mM MOPS, 45mM MgCl2 and 30mM sodium citrate; pH 7) with 0.5% triton x-100 using a razor blade. Another 1400ul of fresh Galbraith buffer with 0.5% triton x-100 was added to resuspend and dilute the nuclei and cell debris. The resulting preparations were filtered through a 50 μm nylon mesh and stained with 1,5 μl of DAPI (1 mg/ml). After a five minute incubation on ice, flow cytometry was performed using a Cyflow flow cytometer (Partec, http://www.partec.com) using the UV or violet diode laser for excitation and a bandpass filter for analyzing DAPI fluorescence. GFP was excited with a 488 nm blue solid state laser. The results were analyzed using Cyflogic (http://www.cyflogic.com). Absolute measurements of nuclei were calculated by using the particle concentration provided by the Cyflow flow cytometer and the “gating” function in Cyflogic. Number of nuclei per seed were then estimated based on the number on initial seed chopped.

Quantification and Statistical Analysis

All the details of statistical tests can be found in the corresponding figure legends. Statistical tests were carried out using the Excel software for Windows. Bars on graphs correspond to the standard deviation (SD).

Data S1: Mapping the RNA-seq reads²¹ to the Col-0 genome. Number of reads per gene per replicate. Relates to Figure 1.

Data S2: TFs differentially expressed between developing aleurone DAL (living endosperm cells) and ESR (dying endosperm cells). Only genes with False discovery rate < 0,05 are displayed. Mean values for the two sample, Pvalue and LogFC = Log2(fold-change) are indicated for each comparison. Relates to Figure 1.

Key Resources Table

Reagent or Resource	Source	Identifier
Bacterial and virus strains
E. coli strain: DH5a	VIB	N/A
A. bacterium: C58C1	VIB	N/A
Chemicals, peptides, and recombinant proteins
MS basal salts	Duchefa Biochemie	Cat#M0221.0050
Agar No.4 - Plant Tissue Culture Agar	NEOGEN	Cat#NCM0250A
Fluorescein diacetate (FDA)	ThermoFisher Scientific	Cat#F1303
4',6-diamidino-2-phénylindole (DAPI)	Sigma Aldrich	Cat#D8417-5MG
MES
Toluidine blue O	Electron Microscopy sciences	Cat#22050
Chloral hydrate	Sigma Aldrich	Cat#23100-1KG
Glycerol	MP Biomedicals	Cat#800688
PIPES	Sigma Aldrich	Cat#P3768-100G
EGTA	Sigma Aldrich	Cat# E4378-100G
MgSO4	Sigma Aldrich	Cat# M2643-500G
Sodium Borate Tetra	Electron Microscopy sciences	Cat #21130
Bovine serum Albumin (BSA)	Sigma Aldrich	Cat #A7906-50G
Glutaraldehyde	EMS	Cat #16220
Spurr’s resin (homemade mix of four components)	NSA, ERL, DMAE, DER	Cat # 19050 Cat #15004 Cat # 13300
Cat #13000
Uranyl acetate	SPI-Chem	Cat #02624-AB
OsO4	EMS	Cat #19170
Acetone	VWR	Cat # 83683.269
Ultrapure agarose	Invitrogen	Cat #16500-500
Sucrose	Sigma Aldrich	Cat #S0389-1KG
MOPS	MERCK LIFE _Sigma	Cat #M3183-500G
MgCl2	Sigma Aldrich	Cat #M8266-100G
Triton x-100	Sigma Aldrich	Cat #X100-500ML
Sodium citrate	Sigma Aldrich	Cat #W302600-1KG-K
Critical commercial assays
ReliaPrep RNA	Promega	Cat #Z6010
qScript cDNA SuperMix	Quantabio	Cat #95048-100
Gateway LR clonase II	ThermoFisher Scientific	Cat #11791020
T4 DNA Ligase (400.000 units/ml)	New England BioLabs	Cat #M0202L_37612
BsaI-HFv2 restriction enzyme	New England BioLabs	Cat #R3733
Experimental models: Organisms/strains
Arabidopsis thaliana: wild type Col-0	PSB-VIB	N/A
Arabidopsis thaliana: mutant nac2-1 kir1-1	18
Arabidopsis thaliana: mutant anac046-1 anac087-1	16
Arabidopsis thaliana: mutant nac_4x	This paper	N/A
Arabidopsis thaliana: mutant nac_4x_cp	This paper	N/A
Arabidopsis thaliana: mutant zou-4	5	N/A
Arabidopsis thaliana: mutant zou-4 nac_4x	This paper	N/A
Arabidopsis thaliana: pPASPA3:NLS-tdTOM in Col-0	37	N/A
Arabidopsis thaliana: pPASPA3:NLS-tdTOM in zou-4	This paper	N/A
Arabidopsis thaliana: pBFN1:NLS-tdTOM in Col-0	38	N/A
Arabidopsis thaliana: pBFN1:NLS-tdTOM in zou-4	This paper	N/A
Arabidopsis thaliana: pBFN1:NLS-tdTOM in nac_4x	This paper	N/A
Arabidopsis thaliana: pTE2:DT-A in Col-0	This paper	N/A
Arabidopsis thaliana: pTE2:DT-A in zou-4	This paper	N/A
Arabidopsis thaliana: pTE2:KIR1 in Col-0	This paper	N/A
Arabidopsis thaliana: pTE2:KIR1 in zou-4	This paper	N/A
Arabidopsis thaliana: pEE:KIR1-SRDX in Col-0	This paper	N/A
Arabidopsis thaliana: pTE2:NLS-GFP	This paper	N/A
Arabidopsis thaliana: pANAC046:NLS-GFP	16	N/A
Arabidopsis thaliana: pANAC087:NLS-GFP	16	N/A
Arabidopsis thaliana: pANAC046:NLS-GFP in zou-4	This paper	N/A
Arabidopsis thaliana: pANAC087:NLS-GFP in zou-4	This paper	N/A
Arabidopsis thaliana: pORE1:ORE1-GFP	18	N/A
Arabidopsis thaliana: pKIR1:KIR1-GFP	18	N/A
Arabidopsis thaliana: pORE1:ORE1-GFP in zou-4	This paper	N/A
Arabidopsis thaliana: pKIR1:KIR1-GFP in zou-4	This paper	N/A
Oligonucleotides
See Table S1	This paper	N/A
Recombinant DNA		N/A
sgRNAs against ANAC003 ANAC010 ANAC019
ANAC079 in pFASTRK-HMGP-AtCas9-NLS-P2A-mCherry-G7T	This paper	N/A
sgRNAs against ANAC046 ANAC087 KIR1 ORE1 in pFASTRK-HMGP-AtCas9-NLS-P2A-mCherry-G7T	This paper	N/A
sgRNAs against ZOU in pFASTRK-HMGP-AtCas9-NLS-P2A-mCherry-G7T	This paper	N/A
pTE2:KIR1 in pB7m34GW	This paper	N/A
pTE2:DT-A in pB7m34GW	This paper	N/A
pANAC087:NLS-GFP in pK7m34GW	16	N/A
pANAC046:NLS-GFP in pK7m34GW	16	N/A
pEE:KIR1-SRDX in pB7m34GW	This paper	N/A
pBFN1:NLS-tdTOM in pB7m34GW	This paper	N/A
pPASPA3:NLS-tdTOM in pB7m34GW	This paper	N/A
DT-A in pDONR221	30	N/A
KIR1 in pDONR221	18	N/A
KIR1-SRDX in pDONR221	18	N/A
pTE2 in pDONRP1RP4	46	N/A
Software and algorithms
CRISPR-P v2	47	http://crispr.hzau.edu.cn/CRISPR2/
FastQC		http://www.bioinformatics.babraham.ac.uk/projects/fastqc/
Trimmomatic	50	N/A
Salmon	51	N/A
EdgeR	53	https://bioconductor.org/packages/release/bioc/html/edgeR.html
ImageJ	4U	https://fiji.sc/

Highlights.

• Arabidopsis endosperm is actively eliminated by a programmed cell death process

• Several redundantly acting NAC transcription factors promote programmed cell death

• Ectopic NAC expression causes total endosperm elimination and aleurone-less seeds

• Combining zhoupi and nac mutations causes endosperm persistence until desiccation

eTOC blurb.

Embryo and endosperm have to develop alongside each other in the confined space of the plant seed. Doll et al. reveal that invasive embryo growth during seed development is facilitated by an actively controlled endosperm cell death program in the model plant Arabidopsis thaliana.

Data and code availability

• Plasmids generated in this study have been deposited to the VIB-UGent plasmid repository (https://gatewayvectors.vib.be) and are publicly available as of the date of publication. Accession numbers are listed in the key resources table. All data generated or analyzed during this study are included in this published article (and its supplementary information files). All data reported in this paper will be shared by the lead contact upon request.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.