Multi-omics Analysis Reveals Sequential Roles for ABA during Seed Maturation

Hormone and omics analyses during seed development in ABA metabolism mutants highlighted the ABA sequential repression of germination-related processes and its contribution to dormancy depth.

Abstract

Abscisic acid (ABA) is an important hormone for seed development and germination whose physiological action is modulated by its endogenous levels. Cleavage of carotenoid precursors by 9-cis epoxycarotenoid dioxygenase (NCED) and inactivation of ABA by ABA 8′-hydroxylase (CYP707A) are key regulatory metabolic steps. In Arabidopsis (Arabidopsis thaliana), both enzymes are encoded by multigene families, having distinctive expression patterns. To evaluate the genome-wide impact of ABA deficiency in developing seeds at the maturation stage when dormancy is induced, we used a nced2569 quadruple mutant in which ABA deficiency is mostly restricted to seeds, thus limiting the impact of maternal defects on seed physiology. ABA content was very low in nced2569 seeds, similar to the severe mutant aba2; unexpectedly, ABA Glc ester was detected in aba2 seeds, suggesting the existence of an alternative metabolic route. Hormone content in nced2569 seeds compared with nced259 and wild type strongly suggested that specific expression of NCED6 in the endosperm is mainly responsible for ABA production. In accordance, transcriptome analyses revealed broad similarities in gene expression between nced2569 and either wild-type or nced259 developing seeds. Gene ontology enrichments revealed a large spectrum of ABA activation targets involved in reserve storage and desiccation tolerance, and repression of photosynthesis and cell cycle. Proteome and metabolome profiles in dry nced2569 seeds, compared with wild-type and cyp707a1a2 seeds, also highlighted an inhibitory role of ABA on remobilization of reserves, reactive oxygen species production, and protein oxidation. Down-regulation of these oxidative processes by ABA may have an essential role in dormancy control.

Biological processes taking place during seed development are controlled by endogenous and exogenous signals and, after dispersal or harvest, determine seed longevity, germination vigor, and successful establishment of seedlings in a variable range of environmental conditions (Penfield and MacGregor, 2017). In many dicots, including Arabidopsis (Arabidopsis thaliana), embryogenesis starts after the double fertilization that gives rise to the diploid embryo and the triploid endosperm. At the heart stage, when morphogenesis is completed, the embryo grows at the expense of the endosperm, and reserve storage takes place. At the end of this maturation phase, the Arabidopsis embryo is surrounded by a single layer of endosperm and a seed coat (or testa) of maternal origin, which is composed of dead dry tissues (North et al., 2010). Abscisic acid (ABA) has been extensively reported to regulate many aspects of seed development, mainly during maturation. Its roles at earlier stages are poorly documented; nevertheless, reduced synthesis in the abscisic acid deficient2 (aba2) mutant has been shown to delay embryogenesis and endosperm cellularization (Cheng et al., 2014). After embryogenesis, ABA inhibits embryo growth and positively regulates reserve accumulation and at later stages induces primary dormancy and desiccation tolerance. Analysis of seed phenotypes of ABA metabolism or signaling mutants in many species pointed out its crucial action in preventing vivipary and allowing seed dispersal in a dormant state (Finkelstein et al., 2008; Nambara et al., 2010; Graeber et al., 2012). Dormancy is an adaptive trait limiting germination under environmental conditions that would promote optimal germination in nondormant seeds. The dormancy depth determines the timing of germination under seasonally varying conditions at which seedling survival and growth are the most favorable (Burghardt et al., 2016). In Arabidopsis, seed dormancy of the most commonly used accession, Columbia-0, is relatively low and can be released by a few weeks of after-ripening (dry storage) or by stratification (cold imbibition). Genetic studies demonstrated that dormancy depth correlates well with ABA levels; it is increased in mutants defective for ABA catabolism and reduced in biosynthesis mutants (Finkelstein et al., 2008).

The ABA metabolism pathway has been largely uncovered (Nambara and Marion-Poll, 2005). The first step specific to ABA biosynthesis is the cleavage of cis-isomers of the oxygenated carotenoids, violaxanthin and neoxanthin, by a 9-cis epoxycarotenoid dioxygenase (NCED) that is encoded in Arabidopsis by a gene family of five members, namely, NCED2, NCED3, NCED5, NCED6, and NCED9. The 15-carbon product, xanthoxin, is then transported from plastids to cytosol and converted into abscisic aldehyde by a short-chain alcohol dehydrogenase, encoded by the ABA2 gene. The final biosynthesis step is catalyzed by an abscisic aldehyde oxidase, requiring activation of its molybdenum cofactor for activity. ABA inactivation occurs either by oxidation or conjugation. The major route is 8′ hydroxylation by the CYP707A subfamily of P450 monooxygenases, which is followed by the spontaneous isomerization of 8′-hydroxy-ABA into phaseic acid (PA). PA has been recently shown to retain some biological activity, and complete inactivation occurs after its conversion into dihydrophaseic acid (DPA) by a PA reductase (Weng et al., 2016). ABA conjugation into ABA Glc ester (ABA-GE) is catalyzed by glucosyl-transferases, and subsequent ABA-GE hydrolysis by two β-glucosidases, BG1 and BG2, contributes to ABA production (Lee et al., 2006; Xu et al., 2012). In developing and imbibed seeds of Arabidopsis, spatiotemporal regulation of specific members of NCED and CYP707A gene families has a major role in modulating ABA levels and regulating dormancy depth and germination timing, as deduced from phenotypes of multiple nced or cyp707a mutants (Lefebvre et al., 2006; Millar et al., 2006; Okamoto et al., 2006; Frey et al., 2012).

There is now strong evidence that members of a multigene family encoding pyrabactin resistance (PYR)/PYR-like (PYL)/regulatory components of ABA receptor (RCAR) are ABA receptors, which sequester and inhibit protein phosphatases 2C (PP2C) when ABA is present. PP2C inactivation allows phosphorylation of Sucrose non-fermenting1-related kinases2, which in turn phosphorylate Basic Leucine Zipper Domain transcription factors of the ABA-INSENSITIVE5 (ABI5)/ABA-responsive element-binding protein/ABA-responsive element-binding factor family, which bind to ABA response elements in promoter sequences of ABA-inducible genes (Cutler et al., 2010). The germination phenotypes of either PP2C or SnRK2 multiple mutants suggest that the ABA signaling pathway involving the PYR receptor and PP2C/SnRK2 phosphorylation cascade operates in dormancy regulation of Arabidopsis seeds (Fujii and Zhu, 2009; Nakashima et al., 2009). Moreover, recent studies reported the interaction of DELAY OF DORMANCY1 (DOG1), a major regulator of seed dormancy whose molecular function remains elusive, with a subset of clade A PP2C phosphatases, resulting in their inhibition (Née et al., 2017; Nishimura et al., 2018).

In this study we exploited the distinctive expression patterns of NCED genes in developing seeds to assess the tissue-specific origin of ABA production. Our results highlight the importance of endospermic ABA and also suggest the existence of alternative pathways for ABA-GE synthesis in seeds. Previous transcriptome studies have focused on seed imbibition and dormancy release. Here, we evaluated the genome-wide impact of ABA deficiency during seed development, in relation with dormancy induction. Comparative analysis of the dry seed proteome and metabolome of the deficient nced2569 mutant with those of the wild type and overproducing cyp707a1a2 correlated well with differential gene expression in developing seeds. The data reported here strongly support the positive effect of ABA on reserve storage and desiccation tolerance and interestingly also reveal its importance in the repression of photosynthesis, cell cycle, reserve remobilization, and oxidative processes to prevent premature germination and promote dormancy.

RESULTS

ABA Levels Are Similarly Reduced in nced2569 and aba2 Seeds

The ABA-deficient aba2 has been often used to study the role of ABA-regulated biological processes. ABA levels are severely reduced in both vegetative and reproductive tissues, and aba2 exhibits typical phenotypes associated with ABA deficiency, i.e. increased plant water loss and reduced seed dormancy (Léon-Kloosterziel et al., 1996; Nambara et al., 1998). However, the aba2 mutation also has major effects on seed development and plant growth, resulting in reduced plant size and poor seed yield (Cheng et al., 2002, 2014). We previously described a triple mutant nced569 whose dormancy was strongly reduced, without visible impact on plant development and seed production (Frey et al., 2012). Because AtNCED3 has been described to have the major contribution to increased ABA accumulation in response to water stress (Iuchi et al., 2001; Tan et al., 2003), we reasoned that the quadruple mutant nced2569 might also exhibit a reduction in dormancy depth and seed ABA content, without deleterious effects on plant development and seed yield. Comparison of plant size and flowering time of nced2569, aba2-2, and wild-type Arabidopsis indicated that NCED3 activity was sufficient for normal development of nced2569 plants (Supplemental Fig. S1).

In nced2569 developing siliques harvested from 6 to 18 d after pollination (DAP) and in dry seeds, ABA content was strongly decreased compared to that in the wild type and, at all developmental stages, ABA levels in nced2569 siliques were as low as those in aba2-2 (Fig. 1A). This suggested a minor contribution of NCED3 activity to ABA accumulation in these tissues, either from on-site synthesis or transported from vegetative tissues. In our previous work, NCED6 was shown to be specifically expressed in the endosperm during seed development and nced6 mutation reduced ABA accumulation in dry seeds (Lefebvre et al., 2006). Therefore, the contribution of NCED6 expression, as a proxy of ABA production in the endosperm, was assessed here by comparing ABA content in nced2569 and nced259 siliques. Surprisingly, ABA levels in nced259 siliques were similar to those in the wild type, suggesting that NCED6 activity in the endosperm was mainly responsible for the observed ABA accumulation. In developing siliques, ABA was shown to be mostly accumulated in seeds (Kanno et al., 2010). To ascertain that ABA measured here in siliques well reflected seed contents, we dissected wild-type developing seeds from silique envelopes and confirmed this previous observation (Supplemental Fig. S2A).

Figure 1.

ABA biosynthesis and catabolism in nced mutants. A–C, ABA (A), DPA (B), and ABA-GE (C) levels in developing siliques at 6, 10, 14, and 18 DAP and dry seeds of aba2-2, nced2569, and nced259 mutants, compared to wild type. Means of three biological replicates are shown with se (n = 3). Two independent experiments were performed with similar results. DW, dry weight; WT, wild type.

Alternative Catabolic Routes Are Used in Mutant Seeds

It is well established that rates of both synthesis and degradation modulate ABA accumulation; therefore, the content of the most abundant catabolites, DPA and ABA-GE, was assessed in nced mutants compared to wild type and aba2-2. In accordance with their impaired ABA synthesis and low seed ABA content, DPA levels were strongly reduced in nced2569 and aba2-2, compared to wild type and nced259 (Fig. 1B), but unexpected differences in ABA-GE accumulation were observed between nced2569 and aba2-2 (Fig. 1C). In contrast to nced2569 siliques that did not accumulate significant ABA-GE amounts, in aba2 siliques, this conjugate was detected at similar levels to wild type at 10 DAP and at higher levels at 14 and 18 DAP. Therefore, it suggested that this conjugate could be produced in this mutant that is unable to convert xanthoxin into abscisic aldehyde.

A more expected observation was the respective similarity of ABA, DPA, and ABA-GE profiles in nced259 siliques compared to wild type, which indicated that ABA catabolism was unaffected in this mutant. Interestingly, in both genotypes, DPA and ABA-GE levels were maximal at 10 DAP, in contrast to ABA levels that peaked at 14 DAP. Thus, at early developmental stages, active ABA hydroxylation and conjugation would likely limit ABA accumulation (Fig. 1). DPA accumulation was also observed in envelopes of wild-type siliques, although at lower levels compared to dissected seeds. The higher amounts of catabolites, compared to ABA, at 10 DAP in both seeds and siliques suggested either active ABA production in these tissues at this stage or/and translocation from vegetative tissues, before subsequent hydroxylation/conjugation (Supplemental Fig. S2). In contrast to seeds, low ABA-GE levels were detected in wild-type silique envelope (Supplemental Fig. S2C), indicating that ABA hydroxylation was predominant.

In cyp707a1a2 siliques that are defective for ABA 8′-hydroxylase activity, ABA levels were increased and DPA levels reduced compared to wild type (Fig. 2, A and B). Interestingly, in contrast to wild type, high contents of ABA-GE and 7′-hydroxy ABA were measured, with maximal levels at 10 DAP for ABA-GE, and from 10 to 18 DAP for 7′-hydroxy ABA (Fig. 2, C and D). Moreover, the detection of high amounts of 7′-hydroxy ABA in cyp707a1a2 siliques suggested that its production did not require CYP707A1 or CYP707A2 activity. The severe reduction in 8′-hydroxylation in cyp707a1a2 siliques was therefore partly compensated by the alternative production of other metabolites that were barely detected in wild-type siliques.

Figure 2.

ABA biosynthesis and catabolism in cyp707a1a2 mutants. A–D, ABA (A), DPA (B), ABA-GE (C), and 7′-hydroxy ABA (D) levels in developing siliques at 6, 10, 14, and 18 DAP and dry seeds of cyp707a1a2 mutants, compared to wild type. Means of three biological replicates are shown with se (n = 3). Two independent experiments were performed with similar results. DW, dry weight; WT, wild type.

Seeds of nced2569 and aba2 Mutants Exhibit Similar Dormancy Levels

Dormancy of the quadruple mutant nced2569 was compared to that of nced259, aba2-2, cyp707a1a2, and wild type. In accordance with their decreased ABA accumulation, nced2569 seeds exhibited reduced dormancy levels, because germination rates at harvest were 15% to 30% and reached 90% after 2 weeks of dry storage, as also observed in aba2 seeds (Fig. 3). Oppositely and in good agreement with the very high ABA contents observed in developing siliques and dry seeds, cyp707a1a2 seed dormancy was strongly enhanced and was not released after two months of dry storage. Despite ABA, DPA, and ABA-GE levels in nced259 seeds being very similar to those of the wild type, dormancy depth was lower. At harvest, germination rates were very similar to wild type (<10%); however, dormancy release was faster (Figs. 1 and 3), thus suggesting that minor ABA pools may fine-tune seed dormancy. Furthermore, comparison of germination rates of nced2569 seeds with those observed in the four combinations of triple mutants confirmed the contribution of ABA production from NCED5, NCED6, and NCED9 activity to dormancy depth (Supplemental Fig. S3).

Figure 3.

Dormancy is reduced in nced2569 and aba2-2. Germination of nced269, nced2569, aba2-2, cyp707a1a2, and wild-type seeds, sown after 2 weeks of dry storage (A). Germination 4 d after sowing of dry seeds stored at room temperature during 0–8 weeks (B). Means of three biological replicates are shown with se (n = 3). Two independent experiments were performed with similar results. WT, wild type.

Transcriptome Variations in Developing Seeds Correlate with ABA Contents

The regulation of seed transcriptome by endogenous ABA has been previously studied in dry and imbibed seeds of aba2 and cyp707a1a2a3 mutants in relation with dormancy release (Okamoto et al., 2010). However, ABA has also a very prominent role in dormancy induction, which we were interested to investigate here by analyzing the impact of ABA deficiency on the global transcriptome in developing nced2569 seeds, isolated from dissected siliques. Because NCED6 activity in the endosperm severely reduced ABA content, we decided to evaluate the specific impact of nced6 mutation, by comparing nced2569 not only to wild type, but also to nced259. We focused on two mid-developmental stages, 10 and 14 DAP, when ABA increase was observed (Fig. 1) and dormancy establishment is reported to occur in Arabidopsis seeds (Karssen et al., 1983). In parallel, ABA levels were analyzed in seed samples at the two chosen stages, with relative ABA contents in isolated seeds corroborating those measured in whole siliques (Supplemental Fig. S4; Fig. 1).

At 10 DAP, transcriptome analysis revealed that 288 transcripts were significantly down-regulated (>1.5-fold, false discovery rate [FDR], P < 0.05) in nced2569 compared to either wild type or nced259. Most of these were down-regulated to a similar extent in both comparisons, and only 41 of them were down-regulated by <1.3-fold in either of the comparisons (Supplemental Table S1). A smaller number of transcripts (98) were significantly up-regulated (1.5-fold or greater) in nced2569 compared to either wild type or nced259. Again, gene regulation was very similar in both comparisons, and only 13 transcripts were <1.3-fold more abundant in one or other comparison (Supplemental Table S2). A large overlap between differentially regulated transcripts was therefore observed in nced2569 compared to either wild type or nced259, in accordance with similar relative differences in ABA levels (Supplemental Tables S1 and S2; Fig. 1).

At 14 DAP, a larger number of transcripts were differentially regulated. In nced2569, 853 transcripts were significantly down-regulated (>1.5-fold, FDR, P < 0.05) compared to either wild type or nced259. A large fraction of these (542) were down-regulated to a similar extent in both comparisons, >1.3-fold (Supplemental Table S3). Like at 10 DAP, a smaller number of transcripts (441) were significantly up-regulated (>1.5-fold) in nced2569 compared to either wild type or nced259. Again, gene regulation was very similar in both comparisons, with an overaccumulation (>1.3-fold) of 334 transcripts (Supplemental Table S4), suggesting that transcriptome variations mainly correlated with ABA levels.

ABA Deficiency Negatively Impacts on Reserve Storage and Desiccation Transcript Accumulation

Gene ontology (GO) enrichment was analyzed using ThaleMine tools (https://apps.araport.org/thalemine/begin.do). Among the down-regulated transcripts in nced2569 at 10 DAP, enrichment was found in GO terms related to lipid and protein storage (Table 1; Supplemental Table S5). Accumulation of transcripts encoding seed storage albumins (SESA), SESA1, SESA3, SESA4, and SESA5, was strongly reduced, from 2- to 10-fold. Seven genes of the deoxythymidine diphosphates-4-dehydrorhamnose 3,5-epimerase (RmlC)-like cupins superfamily were also down-regulated; this family notably encodes the cruciferins, CRU1/CRA1 and CRU2, which transcript accumulation was strongly decreased (4- to 5-fold). Among lipid storage proteins, transcripts encoding oil-body associated proteins, five encoding oleosins, two caleosins (CLO), and two hydroxysteroid dehydrogenases, were down-regulated at variable levels, up to 7-fold for hydroxysteroid dehydrogenase1 and CLO2/Peroxygenase2. Enrichment was also observed for genes related to seed development, among which those encoding late embryogenesis abundant (LEA) genes were strongly down-regulated, i.e. LEA28 (5-fold), LEA29 (5-fold), LEA48 (4-fold), and LEA42 (2.5-fold). Accumulation of other transcripts associated with desiccation tolerance was also reduced, such as AT1G48130/PER1 encoding 1-Cys peroxiredoxin1 and several transcripts encoding aquaporins (Supplemental Table S5).

Table 1. Differentially expressed transcripts encoding proteins involved in reserve storage, carbohydrate transport, and LEA.

The proteins were mostly down-regulated in nced2569 compared to either wild type or nced259 (>1.5-fold; Supplemental Tables S1 and S3). Gene and protein names are indicated according to https://apps.araport.org/thalemine and LEA are numbered according to Candat et al. (2014).

AGI	Symbol	Variation		Protein Encoded
		10 DAP	14 DAP
AT4G27140	SESA1	DOWN	DOWN	SEED STORAGE ALBUMIN1
AT4G27160	SESA3	DOWN	—	SEED STORAGE ALBUMIN3
At4G27170	SESA4	DOWN	—	SEED STORAGE ALBUMIN4
AT5G54740	SESA5	DOWN	DOWN	SEED STORAGE ALBUMIN5
AT5G44120	CRA1	DOWN	—	CRUCIFERINA; RmlC-like cupin superfamily protein
AT1G03880	CRU2	DOWN	—	CRUCIFERIN2; RmlC-like cupin superfamily protein
AT1G03890	AT1G03890	DOWN	DOWN	RmlC-like cupin superfamily protein
AT2G28490	AT2G28490	DOWN	—	RmlC-like cupin superfamily protein
AT2G43120	AT2G43120	—	DOWN	RmlC-like cupin superfamily protein
AT3G04150	AT3G04150	—	DOWN	RmlC-like cupin superfamily protein
AT3G22640	PAP85	DOWN	—	RmlC-like cupin superfamily protein
AT4G36700	AT4G36700	DOWN	—	RmlC-like cupin superfamily protein
AT5G15120	AT5G15120	DOWN	—	RmlC-like cupin superfamily protein
AT5G39110	AT5G39110	—	UP	RmlC-like cupin superfamily protein
AT5G40420	OLEO2	DOWN	—	OLEOSIN2
AT2G25890	OLEO6	DOWN	DOWN	OLEOSIN6
AT3G01570	OLEO7	DOWN	—	OLEOSIN7
AT5G61610	OLEO11	—	DOWN	OLEOSIN11
AT5G07510	GRP14	DOWN	DOWN	GLY-RICH PROTEIN14
AT5G07520	GRP18	DOWN	—	GLY-RICH PROTEIN18
AT4G26740	ATS1	DOWN	—	SEED GENE1/PEROXYGENASE1/CALEOSIN1
AT5G55240	PGX2	DOWN	DOWN	PEROXYGENASE2/CALEOSIN2
AT2G33380	RD20	—	DOWN	RESPONSIVE TO DESICCATION20/CALEOSIN3
AT1G70670	CLO4	—	DOWN	CALEOSIN4
AT1G70680	AT1G70680	—	DOWN	Caleosin-related family protein
AT5G50600	HSD1	DOWN	DOWN	HYDROXYSTEROID DEHYDROGENASE1
AT5G50770	HSD6	DOWN	—	HYDROXYSTEROID DEHYDROGENASE6
AT1G05510	AT1G05510	—	DOWN	Naphthalene 1,2-dioxygenase subunit alpha (duf1264)
AT2G18370	AT2G18370	—	DOWN	Bifunctional inhibitor/lipid-transferprotein/seed storage 2S albumin superfamily protein
AT4G22610	AT4G22610	—	DOWN	Bifunctional inhibitor/lipid-transfer protein/seed storage 2S albumin superfamily protein
AT5G22500	FAR1	—	DOWN	FATTY ACID REDUCTASE1
AT3G44540	FAR4	—	DOWN	FATTY ACID REDUCTASE4
AT3G44550	FAR5	—	DOWN	FATTY ACID REDUCTASE5
AT5G49900	AT5G49900	—	DOWN	Beta-glucosidase, GBA2 type family protein
AT5G57800	CER3	—	DOWN	Fatty acid hydroxylase superfamily
AT5G58860	CYP86A1	—	DOWN	Cytochrome P450, family 86, subfamily A, polypeptide 1
AT4G00360	CYP86A2	—	DOWN	Cytochrome P450, family 86, subfamily A, polypeptide 2
AT1G01600	CYP86A4	—	DOWN	Cytochrome P450, family 86, subfamily A, polypeptide 4
AT5G23190	CYP86B1	—	DOWN	Cytochrome P450, family 86, subfamily B, polypeptide 1
AT5G08250	AT5G08250	—	DOWN	Cytochrome P450 superfamily protein
AT1G08920	ESL1	—	DOWN	ERD (EARLY RESPONSE TO DEHYDRATION) SIX-LIKE 1
AT1G22710	SUC2	DOWN	—	SUC-PROTON SYMPORTER2
AT1G71890	SUC5	—	UP	Major facilitator superfamily protein
AT1G73220	OCT1	—	DOWN	ORGANIC CATION/CARNITINE TRANSPORTER1
AT1G77210	STP14	—	DOWN	SUGAR TRANSPORTER14
AT3G05160	AT3G05160	DOWN	—	Major facilitator superfamily protein
AT3G05165	AT3G05165	—	DOWN	Major facilitator superfamily protein
AT3G05400	AT3G05400	—	DOWN	Major facilitator superfamily protein
AT3G19930	STP4	—	DOWN	SUGAR TRANSPORTER4
AT3G47420	G3Pp1	DOWN	DOWN	Putative glycerol-3-P transporter1
AT5G26340	MSS1	—	DOWN	Major facilitator superfamily protein
AT3G48740	SWEET11	—	DOWN	Nodulin MtN3 family protein
AT4G25010	SWEET14	—	DOWN	Nodulin MtN3 family protein
AT5G13170	SAG29	DOWN	DOWN	SENESCENCE-ASSOCIATED GENE2
AT5G23660	SWEET12	—	DOWN	Bidirectional sugar transporter SWEET12-like protein
AT4G23010	UTR2	—	DOWN	UDP-GAL TRANSPORTER2
AT1G17810	BETA-TIP	DOWN	DOWN	BETA-TONOPLAST INTRINSIC PROTEIN
AT4G19030	NIP1;1	—	DOWN	NOD26-LIKE MAJOR INTRINSIC PROTEIN1;1
AT4G18910	NIP1;2	—	DOWN	NOD26-LIKE INTRINSIC PROTEIN1;2
AT2G39010	PIP2;6	—	DOWN	PLASMA MEMBRANE INTRINSIC PROTEIN2;6
AT4G35100	PIP3	—	UP	PLASMA MEMBRANE INTRINSIC PROTEIN3
AT3G16240	TIP2;1	DOWN	DOWN	DELTA TONOPLAST INTEGRAL PROTEIN
AT1G73190	TIP3;1	DOWN	—	Aquaporin-like superfamily protein
AT1G52690	LEA7	—	DOWN	LATE EMBRYOGENESIS ABUNDANT7
AT1G72100	LEA9	—	DOWN	Late embryogenesis abundant domain-containing protein
AT2G21490	LEA/LEA14	—	DOWN	DEHYDRIN LEA
AT2G23110	LEA15	—	DOWN	Late embryogenesis abundant protein, group 6
AT3G02480	LEA28	DOWN	DOWN	Late embryogenesis abundant protein (LEA) family protein
AT3G15670	LEA29	DOWN	DOWN	Late embryogenesis abundant protein (LEA) family protein
AT3G17520	LEA30	—	DOWN	Late embryogenesis abundant protein (LEA) family protein
AT3G50970	LTI30/LEA33	—	DOWN	LOW TEMPERATURE-INDUCED30
AT3G50980	XERO1/ LEA34	—	DOWN	DEHYDRIN XERO1
AT3G51810	EM1/LEA35	—	DOWN	LATE EMBRYOGENESIS ABUNDANT1
AT4G21020	LEA42	DOWN	DOWN	Late embryogenesis abundant protein (LEA) family protein
AT5G06760	LEA4-5/LEA46	—	DOWN	LATE EMBRYOGENESIS ABUNDANT4-5
AT5G44310	LEA48	DOWN	DOWN	Late embryogenesis abundant protein (LEA) family protein
AT5G53270	LEA50	—	DOWN	Seed maturation protein
AT5G66400	RAB18/LEA51	—	DOWN	RESPONSIVE TO ABA18
AT1G17620	AT1G17620	—	DOWN	Late embryogenesis abundant (LEA) Hyp-rich glycoprotein family
AT4G01410	AT4G01410	—	DOWN	Late embryogenesis abundant (LEA) Hyp-rich glycoprotein family
AT5G45320	AT5G45320	—	DOWN	Late embryogenesis abundant protein
AT1G54540	AT1G54540	—	DOWN	Late embryogenesis abundant (LEA) Hyp-rich glycoprotein family
AT5G66780	AT5G66780	—	DOWN	Late embryogenesis abundant protein

Among the down-regulated transcripts at 14 DAP, enrichment was found in GO terms related to lipid synthesis and accumulation (Table 1; Supplemental Table S5). Accumulation of transcripts related to lipid storage, encoding the oil-body associated protein OBAP1a, the four caleosin-related proteins CLO2, CLO3, CLO4, and AT1G70680, and the three oleosin-related proteins OLEO6, OLEO11, and GRP14, were reduced ∼2-fold in nced2569. In addition, a number of genes belonging to the GO terms cutin, suberine, and wax biosynthesis (KEGG pathway 00073), were concomitantly down-regulated. These included the three fatty acid reductase genes (FAR), FAR1, FAR4, and FAR5, and five genes related to the cytochrome P450 family 86, which expression was reduced up to 2-fold. Compared to 10 DAP, fewer genes encoding storage proteins were differentially expressed at 14 DAP in nced2659 compared to the other genotypes; however, SESA1 and SESA5 transcripts were down-regulated at both stages. In contrast, a higher number of LEA transcripts were down-regulated at 14 DAP than at 10 DAP, which included LEA35/EM1, LEA46, LEA30, LEA51/RAB18, LEA29, LEA33/XERO2/LTI30, LEA28, LEA50, LEA34/XERO1, LEA15, AT5G66780, AT4G01410, LEA9, LEA7, LEA42, AT1G17620, AT1G54540, AT5G45320, LEA48, and LEA14, listed from the most strongly down-regulated LEA35/EM1 (18-fold) to LEA14 (1.5-fold). Interestingly, the accumulation of transcripts encoding two heat shock transcription factors, AT3G02990/HSFA1E and AT3G24520/HSFC1, was also reduced (Supplemental Table S3).

At this stage, an enrichment was also found in GO terms related to carbohydrate transport and, among the transcripts for which accumulation was reduced in nced2569, we found four members of the SWEET (a nodulin MtN3 family protein) Suc efflux transporter family (SWEET11, SWEET12, SWEET14, and SENESCENCE-ASSOCIATED GENE29), eight members of the major facilitator superfamily (EARLY RESPONSE TO DEHYDRATION SIX-LIKE 1, MSS1 (a major facilitator superfamily protein), SUGAR TRANSPORTER4, SUGAR TRANSPORTER14, AT3G05165, AT3G05400, AT3G47420, and ORGANIC CATION/CARNITINE TRANSPORTER1), and six members of the aquaporin family (BETA-TONOPLAST INTRINSIC PROTEIN (TIP), PLASMA MEMBRANE INTRINSIC PROTEIN2;6, NOD26-'LIKE MAJOR INTRINSIC PROTEIN1;1, NOD26-LIKE MAJOR INTRINSIC PROTEIN1;2, TIP2;1, and TIP3;1). Several of these transcripts were already reduced at 10 DAP (Table 1). In accordance with a possible reduction of sugar transport, a number of transcripts encoding sugar metabolism enzymes were less accumulated, notably Suc synthases, SUS2 at 10 DAP, and SUS3, SUS5, and stachyose synthase, and trehalose-6-P synthase (TPS1, TPS8, and TPS10) at 14 DAP. Taken together, these observations were in favor of a positive role of ABA in reserve storage and desiccation tolerance.

ABA Deficiency Positively Impacts on Photosynthesis and Cell Cycle

In green seeds, such as in Arabidopsis, photosynthesis is active during seed maturation and has an important contribution to germination vigor (Allorent et al., 2015). Interestingly, among the up-regulated transcripts at 14 DAP, a strong enrichment in GO terms related to photosynthesis was observed in nced2569 seeds compared to the other two genotypes, whereas a single transcript related to these GOs was found differentially regulated at 10 DAP (Table 2; Supplemental Table S5). Moreover, an important enrichment was also found in GO terms related to plastid/chloroplast localization (113 transcripts in GO:0009507 and GO:0009536). In good agreement with increased accumulation of transcripts involved in light harvesting and photosynthesis, the transcript encoding STAY GREEN2/NON-YELLOWING2, an enzyme of the chlorophyll degradation pathway, was down-regulated.

Table 2. Differentially expressed transcripts encoding proteins involved in photosynthesis and cell cycle.

The proteins were mostly up-regulated at 14 DAP in nced2569 compared to either wild type or nced259 (>1.5-fold; Supplemental Tables S2 and S4). Gene and protein names are indicated according to https://apps.araport.org/thalemine.

AGI	Symbol	Variation		Protein Encoded
		10 DAP	14 DAP
AT1G31330	PSAF	—	UP	PHOTOSYSTEM I SUBUNIT F
AT1G45474	LHCA5	—	UP	PHOTOSYSTEM I LIGHT HARVESTING COMPLEX PROTEIN5
AT1G48510	AT1G48510	—	UP	Surfeit locus 1 cytochrome c oxidase biogenesis protein
AT1G50900	GDC1	—	UP	GRANA DEFICIENT CHLOROPLAST1
AT1G52220	AT1G52220	—	UP	Curvature thylakoid protein
AT1G52230	PSAH2	—	UP	PHOTOSYSTEM I SUBUNIT H2
AT1G54500	AT1G54500	—	UP	Rubredoxin-like superfamily protein
AT1G54780	TLP18.3	—	UP	THYLAKOID LUMEN PROTEIN18.3
AT1G60600	ABC4	—	UP	ABERRANT CHLOROPLAST DEVELOPMENT4
AT1G60950	FED A	—	UP	2Fe-2S ferredoxin-like superfamily protein
AT1G76100	PETE1	—	UP	PLASTOCYANIN1
AT1G76570	LHCB7	—	DOWN	LIGHT-HARVESTING COMPLEX B7
AT1G80480	PTAC17	—	UP	PLASTID TRANSCRIPTIONALLY ACTIVE17
AT2G05100	LHCB2.1	—	UP	PHOTOSYSTEM II LIGHT HARVESTING COMPLEX PROTEIN2.1
AT2G26670	TED4	—	DOWN	REVERSAL OF THE DET PHENOTYPE4
AT2G28605	AT2G28605	—	UP	Photosystem II reaction center PsbP family protein
AT2G40100	LHCB4.3	—	UP	LIGHT HARVESTING COMPLEX PHOTOSYSTEM II
AT2G42310	AT2G42310	—	UP	ESSS subunit of NADH:ubiquinone oxidoreductase (complex I)
AT2G46820	PSI-P	—	UP	PHOTOSYSTEM I P SUBUNIT
AT2G47450	CAO	—	UP	CHAOS
AT2G47940	DEG2	—	UP	DEGRADATION OF PERIPLASMIC PROTEINS2
AT3G04550	AT3G04550	—	UP	Rubisco accumulation factor-like protein
AT3G22840	ELIP1	—	DOWN	EARLY LIGHT-INDUCIBLE PROTEIN
AT3G27690	LHCB2.3	—	UP	PHOTOSYSTEM II LIGHT HARVESTING COMPLEX PROTEIN2.3
AT3G47470	LHCA4	—	UP	LIGHT-HARVESTING CHLOROPHYLL-PROTEIN COMPLEX I SUBUNIT A4
AT3G50820	PSBO2	—	UP	PHOTOSYSTEM II SUBUNIT o-2
AT3G56090	FER3	—	UP	FERRITIN3
AT3G59400	GUN4	—	UP	GENOMES UNCOUPLED4
AT4G01150	AT4G01150	—	UP	Curvature thylakoid 1A-like protein
AT4G04640	ATPC1	—	UP	ATPase, F1 complex, gamma subunit protein
AT4G05180	PSBQ-2	—	UP	PHOTOSYSTEM II SUBUNIT Q-2
AT4G10340	LHCB5	—	UP	LIGHT HARVESTING COMPLEX OF PHOTOSYSTEM II 5
AT4G11910	AT4G11910	—	DOWN	STAY-GREEN-like protein
AT4G21280	PSBQA	—	UP	PHOTOSYSTEM II SUBUNIT QA
AT4G27440	PORB	—	UP	PROTOCHLOROPHYLLIDE OXIDOREDUCTASE B
AT4G37925	NDHM	UP	—	NADH DEHYDROGENASE-LIKE COMPLEX M
AT5G01530	LHCB4.1	—	UP	LIGHT HARVESTING COMPLEX PHOTOSYSTEM II
AT5G28500	AT5G28500	—	UP	Rubisco accumulation factor-like protein
AT5G45040	CYTC6A	—	UP	CYTOCHROME C6A
AT5G51010	AT5G51010	—	UP	Rubredoxin-like superfamily protein
AT5G52970	AT5G52970	—	UP	Thylakoid lumen 15.0 kD protein
AT5G54270	LHCB3	—	UP	LIGHT-HARVESTING CHLOROPHYLL B-BINDING PROTEIN3
AT5G55710	TIC20-V	—	UP	TRANSLOCON AT THE INNER ENVELOPE MEMBRANE OF CHLOROPLASTS 20-V
AT5G61410	RPE	—	UP	d-RIBULOSE-5-PHOSPHATE-3-EPIMERASE
AT5G66570	PSBO1	—	UP	PS II OXYGEN-EVOLVING COMPLEX1
AT1G15570	CYCA2;3	—	UP	CYCLIN A2;3
AT1G20610	CYCB2;3	—	UP	CYCLIN B2;3
AT1G20930	CDKB2;2	—	UP	CYCLIN-DEPENDENT KINASE B2;2
AT1G26870	FEZ	—	UP	NAC (No Apical Meristem) domain transcriptional regulator superfamily protein
AT1G34460	CYCB1;5	—	UP	CYCLIN B1;5
AT1G44110	CYCA1;1	—	UP	CYCLIN A1;1
AT1G50490	UBC20	—	UP	UBIQUITIN-CONJUGATING ENZYME20
AT1G65470	FAS1	—	UP	FASCIATA1
AT1G76310	CYCB2;4	—	UP	CYCLIN B2;4
AT1G78770	APC6	—	UP	ANAPHASE PROMOTING COMPLEX6
AT2G26760	CYCB1;4	—	UP	CYCLIN B1;4
AT2G45490	AUR3	—	UP	ATAURORA3
AT3G19050	POK2	—	UP	PHRAGMOPLAST ORIENTING KINESIN2
AT3G19590	BUB3.1	—	UP	BUB (BUDDING UNINHIBITED BY BENZYMIDAZOL)3.1
AT3G57060	AT3G57060	—	UP	Binding protein
AT3G57860	UVI4-LIKE	—	UP	UV-B-INSENSITIVE4-LIKE PROTEIN
AT4G05190	ATK5	—	UP	KINESIN5
AT4G31805	AT4G31805	—	UP	WRKY family transcription factor
AT4G33260	CDC20.2	—	UP	CELL DIVISION CYCLE20.2
AT4G34160	CYCD3;1	—	UP	CYCLIN D3;1
AT4G37490	CYCB1;1	—	UP	CYCLIN B1;1
AT5G13840	FZR3	—	UP	FIZZY-RELATED3
AT5G18700	AT5G18700	—	UP	Kinase family with ARM repeat domain-containing protein
AT5G48600	SMC3	—	UP	STRUCTURAL MAINTENANCE OF CHROMOSOME3
AT5G50375	CPI1	—	UP	CYCLOPROPYL ISOMERASE

At the same developmental stage, an enrichment was also found in several GO terms related to cell division (GO:0051301) and cell cycle (GO:0007049), the list of transcripts partially overlapped with transcripts related to the GO term tissue development (GO:0009888), for which an enrichment was also observed. Among these up-regulated transcripts, the most represented belonged to the cyclin family (Table 2; Supplemental Table S5).

ABA Deficiency Differentially Alters Signaling Pathways

At 10 DAP, among the small number of up-regulated genes, we found that three members of the PYR/PYL/RCAR family of ABA receptors, PYL2, PYL4, and PYL6, were up-regulated by ∼2-, 2.4-, and 1.5-fold, respectively, in comparisons of nced2569 with either wild type or nced259 (Supplemental Table S2). At 14 DAP, only PYL4 transcript was found significantly up-regulated (∼2-fold; Supplemental Table S4). This suggested that reduction in ABA levels may be compensated to some extent by an increased sensitivity, thanks to overexpression of these three PYR/PYL/RCAR receptor genes. Moreover, the transcript encoding BG1, the β-glucosidase (BGLU18) converting ABA-GE into ABA, was also 2-fold up-regulated (Supplemental Table S4), suggesting that ABA deficiency may activate ABA release from stored conjugated forms. Expression of these four genes at 14 DAP was tested by quantitative PCR (qPCR), which confirmed that PYL4 was the most up-regulated gene (Fig. 4). Despite Complete Arabidopsis Transcriptome Micro Array (CATMA) analysis detecting a significantly increased accumulation of PYL2 and PYL6 transcripts only at 10 DAP (FDR-adjusted P value < 0.05), increased expression of these genes was also observed at 14 DAP by qPCR. Interestingly, accumulation of transcripts encoding several PP2C proteins, including ABI2, PP2CA/ABA HYPERSENSITIVE GERMINATION3, and HIGHLY ABA-INDUCED3, was reduced up to 2-fold in nced2569 at 14 DAP (Supplemental Table S3). Concomitantly, expression of DOG1, which protein has been shown to interact with PP2C, was similarly reduced, together with its related DOG-like genes (DOGL2, DOGL3, and DOGL4). Furthermore, a lower accumulation of ABI5 transcripts, which encode a major ABA-regulated transcription factor downstream of SnRK2 kinase, was also observed, in good correlation with that of its well-known targets, such LEA35/EM1 (Table 1; Carles et al., 2002).

Figure 4.

ABA deficiency increases expression of ABA receptors and BG1. Relative expression levels of PYL2, PYL4, PYL6, and BG1 in developing seeds dissected from mutant and wild-type siliques at 14 DAP. The values are means of three biological replicates, presented with se values. Expression levels were normalized with those of EF1α4 (At5g60390) and At4g12590 reference genes. Significant differences compared to wild type were analyzed by Student’s t test (*P < 0.05, **P < 0.01, ***P < 0.001). WT, wild type.

A number of other transcription factors were also differentially regulated at 10 and 14 DAP in nced2569 seeds compared to wild type (Supplemental Tables S1–S4). At 10 DAP, transcripts encoding six transcription factors were up-regulated in nced2569 (MYB13/AT1G06180, RELATED TO ABI3/VIVIPAROUS1 2/AT1G68840, AGAMOUS-like22/SHORT VEGETATIVE PHASE/AT2G22540, CYTOKININ RESPONSE FACTOR5/AT2G46310, ABERRANT TESTA SHAPE/AT5G42630, and MADS AFFECTING FLOWERING 5/AT5G65080). In contrast, nine transcripts encoding proteins involved in transcriptional regulation were down-regulated (Ethylene-responsive transcription factor ERF023/AT1G01250, CRYPTOCHROME-INTERACTING BASIC-HELIX-LOOP-HELIX 5/AT1G26260, No apical meristem domain containing protein NAC019/ANAC019/AT1G52890, WRKY1/AT2G04880, FERTILIZATION INDEPENDENT SEED 2/AT2G35670, REPRESSOR OF SILENCING 4/AT3G14980, NAC2/ANAC056/AT3G15510, OXIDATION-RELATED ZINC FINGER 2/AT4G29190, and RELATED TO AP2 10/AT4G36900). Previous works showed that ANAC019 acts as a positive regulator of ABA signaling (Jensen et al., 2010). The oilseed rape (Brassica napus) BnaNAC56 transcription factor, the ortholog gene of ANAC056, has been shown to be induced by ABA and jasmonic acid (Chen et al., 2017). Another transcription factor, the CCCH zinc finger protein gene, OXIDATION-RELATED ZINC FINGER 2/AT4G29190, is also involved in ABA and jasmonic acid responses (Lee et al., 2012). The ABA deficiency appears to negatively affect defense pathways. Indeed, WRKY1/AT2G04880 involved in the salicylic acid signaling pathway was also down-regulated in nced2569 developing seeds.

Among 64 transcription factors up- or down-regulated at 14 DAP, at least 29 of them were described to be involved in hormonal responses (GO:0009725). In accordance with the tight interaction between the ABA and ethylene signaling pathways (Corbineau et al., 2014), transcription factors involved in responses to these hormones were the most represented. Nevertheless, ABA deficiency broadly affected the expression of other hormone signaling components (Table 3).

Table 3. Differentially expressed transcripts encoding proteins involved in hormone response pathways.

The proteins were either up- or down-regulated in nced2569 compared to either wild type or nced259 (>1.5-fold at 14 DAP; Supplemental Tables S3 and S4).

AGI	Symbol	Variation	GO Terms
AT1G04250	AXR3	UP	Response to auxin; auxin-activated signaling pathway
AT1G26870	FEZ	UP	Response to auxin
AT4G16780	HB-2	UP	Response to auxin and to cytokinin
AT5G39860	PRE1	UP	Gibberellic acid-mediated signaling pathway; response to brassinosteroid; brassinosteroid-mediated signaling pathway
AT5G17490	RGL3	UP	Gibberellic acid-mediated signaling pathway; response to gibberellin; negative regulation of gibberellic acid-mediated signaling pathway; response to ethylene; jasmonic acid-mediated signaling pathway; response to abscisic acid
AT5G25190	ESE3	UP	Ethylene-activated signaling pathway
AT2G44840	ERF13	UP	Ethylene-activated signaling pathway
AT5G65510	AIL7	UP	Ethylene-activated signaling pathway; auxin-mediated signaling pathway involved in phyllotactic patterning
AT3G23050	IAA7	DOWN	Response to auxin; auxin-activated signaling pathway; response to jasmonic acid
AT5G17300	RVE1	DOWN	Auxin-activated signaling pathway
AT1G79700	WRI4	DOWN	Ethylene-activated signaling pathway
AT3G16280	AT3G16280	DOWN	Ethylene-activated signaling pathway
AT3G50260	CEJ1	DOWN	Ethylene-activated signaling pathway
AT3G54990	SMZ	DOWN	Ethylene-activated signaling pathway
AT4G06746	RAP2.9	DOWN	Ethylene-activated signaling pathway
AT4G39780	AT4G39780	DOWN	Ethylene-activated signaling pathway
AT3G61630	CRF6	DOWN	Ethylene-activated signaling pathway; cytokinin-activated signaling pathway
AT1G13960	WRKY4	DOWN	Response to ethylene; response to jasmonic acid
AT3G15500	NAC3	DOWN	Jasmonic acid mediated signaling pathway
AT1G61660	AT1G61660	DOWN	Cellular response to abscisic acid stimulus
AT1G01720	ATAF1	DOWN	Negative regulation of abscisic acid-activated signaling pathway
AT1G17950	MYB52	DOWN	Response to abscisic acid
AT2G35940	BLH1	DOWN	Response to abscisic acid
AT2G46270	GBF3	DOWN	Response to abscisic acid
AT4G28110	MYB41	DOWN	Response to abscisic acid
AT5G65310	HB5	DOWN	Response to abscisic acid; abscisic acid-activated signaling pathway
AT2G36270	ABI5	DOWN	Response to abscisic acid; abscisic acid-activated signaling pathway; response to gibberellin
AT5G37260	RVE2	DOWN	Response to abscisic acid; response to gibberellin; response to ethylene; response to auxin; response to jasmonic acid
AT2G36890	RAX2	DOWN	Response to abscisic acid; response to gibberellin; response to jasmonic acid

Protein Accumulation in Dry Seeds Correlates Well with Transcriptome Variations in Developing Seeds

To get a deeper insight into the biological functions affected by ABA content in developing seeds, proteomics and metabolomics were carried out on dry seeds. Because the transcriptome of nced569 appeared similar to that of the wild type, the more contrasting genotypes, nced2569, cyp707a1a2, and wild type, were chosen to assess the impact of altered ABA levels on the global proteome and metabolome in dry seeds.

Shotgun liquid chromatography-tandem mass spectrometry (LC-MS/MS) proteomics led to the identification of >1,500 proteins, among which 838 were accurately quantified based on extracted ion currents (Valot et al., 2011). Statistical analysis revealed the differential accumulation of 120 proteins, among which 68 displayed a higher abundance in nced2569 seeds compared to either wild type or cyp707a1a2 (>1.3-fold, one-wayANOVA, P < 0.05; Supplemental Table S6). A large number of them (63) were more abundant in nced2569 compared to cyp707a1a2, with 48 more abundant in nced2569 versus wild type. Moreover, most of them were also more abundant in wild type compared to cyp707a1a2. On the other hand, 46 proteins were more abundant in cyp707a1a2 compared to nced2569 mutant and 15 were more abundant in cyp707a1a2 compared to wild type (>1.3-fold, one-way ANOVA, P < 0.05; Supplemental Table S7). Interestingly, 25 of these proteins were significantly less abundant in nced2569 than in wild type.

Proteome modifications observed in mature dry seeds were in good accordance with transcriptome variations described above in developing seeds. Indeed, similar enrichments were found in GO terms, notably those associated with biological functions and metabolic pathways (Fig. 5; Supplemental Table S5). Among up-regulated proteins in nced2569, enrichments were found for GO terms related to metabolic processes. The strong enrichment in several GO terms associated with carbon fixation and related carbohydrate metabolism was in good correlation with that observed for Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways related to Calvin cycle and photosynthesis (KEGG:00710), pentose P pathway (KEGG:00030), glycolysis/gluconeogenesis (KEGG:00010), and Fru and Man metabolism (KEGG:00051). Moreover, among proteins significantly more abundant in nced2569, enrichment was observed in GO terms related to organic acid metabolic processes and FA metabolism. Interestingly, three enzymes involved in fatty acid (FA) β-oxidation showed a higher abundance in nced2569, such as the peroxisomal acyl-CoA oxidase3 (ACX3/AT1G06290), enoyl-CoA hydratase (ECHIA/AT4G16210), multifunctional protein2 (MFP2/AT3G06860), and NAD(P)-binding Rossmann-fold superfamily protein (AT1G24360), which could be associated with the induction of key enzymes involved in reserve remobilization, namely isocitrate lyase (ICL)/AT3G21720 and phosphoenolpyruvate carboxykinase 1/AT4G37870, suggesting that the oil reserve remobilization previously reported during late maturation in oilseed rape and Arabidopsis seeds (Eastmond and Rawsthorne, 2000; Baud et al., 2002) was activated in nced2569. A parallel conclusion can be drawn from the differential accumulation of proteins involved in glycolysis and/or gluconeogenesis pathways, such as glyceraldehyde-3-P dehydrogenase B subunit (GAPB/AT1G42970), triosephosphate isomerase (TIM/AT2G21170), Fru-bisphosphate aldolase2 (FBA2/AT4G38970), Fru-1,6-bisphosphatase (FBP/AT1G43670), and high cyclic electron flow1 (HCEF1/AT3G54050).

Figure 5.

Differentially accumulated proteins and metabolites in dry seeds of ABA metabolism mutants. Proteins and metabolites involved in photosynthesis, glycolysis/gluconeogenese, energy, and amino acid and glucosinolate metabolism in cyp707a1a2, wild type, and nced2569. ABA-regulated proteins are displayed within rectangles; metabolites are shown in blue. 1,3-BPG, Glycerate-1,3-bisphosphate; 1,6 FBP, Fru-1,6-bisphosphate; 2-PGA, Glycerate-2-P; 3-PGA, glycerate-3-P; 4MTB, 4-methylthiobutyl glucosinolate; 5MTP, 5-methylthiopentyl glucosinolate; 5MTPN, 5-methylthiopentanenitrile glucosinolate breakdown product; 6MTHN, 6-methylthiohexanenitrile glucosinolate breakdown product; α-KG, α-ketoglutarate (2-oxoglutarate); ACX3, Acyl-CoA oxidase3; ASP2, Asp aminotransferase2; AT1G24360, NAD(P)-binding Rossmann-fold superfamily protein; AT2G22230, Thioesterase superfamily protein; AT3G58610, Ketol-acid reductoisomerase; AT3G60750, Transketolase; CAC3, acetyl Coenzyme a carboxylase carboxyltransferase alpha subunit; CAT3, Catalase3; DHA, Dehydroascorbate; DHAP, dihydroxyacetone P; E-4P, Erythrose-4-P; ECHIA, Enoyl-CoA hydratase/isomerase A; FBA2, Fru-bisphosphate aldolase2; FBP, Fru-1,6-bisphosphatase; Fru-6P, Fru-6-P; GABA, γ-aminobutyric acid; Glc-6P, Glc-6-P; Glycerol-3P, Glycerol-3-P; GSH, Glutathione; GSL, Glucosinolate; GSSG, Glutathione disulfide (oxidized glutathione); HCEF1, High cyclic electron flow1 (chloroplastic Fru 1,6-bisphosphate phosphatase); MAML-4, Methylthioalkylmalate synthase-like4; MDAR1/6, Monodehydroascorbate reductase1/6; MDHA, Monodehydroascorbate; MPF2, Multifunctional protein2; NSP2, nitrile specifier protein2; OAA, Oxaloacetate; OPHS, o-phospho-l-homo-Ser; PCK1, Phosphoenolpyruvate carboxykinase1; PEP, Phosphoenolpyruvate; PGAL, Glyceraldehyde 3-P; PSBO1, PS II oxygen-evolving complex1; R-5P, Ribose-5-P; RBCL, Ribulose-bisphosphate carboxylase large chain; RBCS1A, Ribulose bisphosphate carboxylase small chain1A; Ru-5P, Ribulose-5-P; Ru-BP, Ribulose-1,5-bisphosphate; S-7P, Sedoheptulose-7-P; SBP, Sedoheptulose-1, 7-bisphosphate; SBPASE, Sedoheptulose-bisphosphatase; TAG, Triacylglyceride; TGG1, Thioglucoside glucohydrolase1; TIM, Triosephosphate isomerase; Xu-5P, Xylulose 5-P AT1G24360, NAD(P)-binding Rossmann-fold superfamily protein.

Among proteins up-regulated in cyp707a1a2 compared to the other genotypes, enrichment was found for GO terms related to seed development, reserve storage, and stress responses. These functions confirmed the central role of ABA in the control of late maturation programs. In particular, abundance of four members of RmlC-like cupins superfamily (AT5G44120/CRU1/CRA1, AT1G03880/CRU2, AT2G18540, and AT2G28490) and albumin (AT5G54740/SESA5) was higher in cyp707a1a2 compared to nced2569 seeds. Moreover, the abundance of cruciferin CRU1/CRA1 was also higher in cyp707a1a2 seeds compared to the wild type. Furthermore, 10 LEA proteins were more abundant in cyp707a1a2 mutants (Supplemental Table S7). AT1G47980, annotated as a desiccation-like protein, was similarly differentially accumulated (2.7-fold) in the two mutants.

Glucosinolate catabolism enzymes, nitrile specifier protein2 (NSP2/AT2G33070) and thioglucoside glucohydrolase1 (TGG1/AT5G26000), were also found differentially accumulated. Strikingly, these two proteins were respectively more abundant in nced2569 and cyp707a1a2 (Fig. 5; Supplemental Tables S6 and S7). TGG1 transcript was also overexpressed in nced2569 seeds at 14 DAP (Supplemental Table S4). Glucosinolates are sulfur- and nitrogen-containing secondary metabolites. Their hydrolysis by myrosinases promote the release of defense-related compounds, such as thiocyanate, cyanate, or nitrile. However, it has been also suggested that glucosinolates could act as sulfur- and nitrogen-sources, notably upon seed imbibition (Galland et al., 2014). Thus, the accumulation of these proteins may indicate a premature resumption of metabolism in nced2569 seeds, which would contribute to germination activation upon seed imbibition.

Metabolite Analysis Reveals the Impact of ABA Levels on Amino Acid, Sugar, and Glucosinolate Contents

The metabolomic analysis of nced2569, cyp707a1a2, and wild-type dry seeds was carried out using gas chromatography-MS analysis. In total, 274 metabolites were identified based on their retention times and their mass spectra and a differential accumulation of 74 metabolites was detected among the three genotypes (one-way ANOVA, P < 0.05, Supplemental Tables S8 and S9).

Metabolomic analysis revealed a positive effect of ABA-deficiency on the accumulation of free primary metabolites during late maturation (Fig. 5). Indeed, 44 metabolites showed a higher level in nced2569 compared to cyp707a1a2, among which 15 metabolites were also significantly more abundant compared to wild type (Supplemental Table S8). Amino acids, organic acids, and sugars were predominantly represented among these metabolites. Indeed, 11 of the 16 quantified amino acids, notably branched-chain amino acids (Val, Leu, and Ile) and shikimate pathway-related amino acids (Trp, Phe, and Tyr), displayed higher abundance in nced2569 and wild type than in cyp707a1a2. Organic acids, as key intermediates of tricarboxylic acid (TCA)/glyoxylate cycles (namely citrate, aconitate, fumarate, and malate), were also more abundant in nced2569. In agreement with the up-regulation of beta-oxidation and glyoxylate cycle related enzymes, this observation suggested an increase of FA-derived carbon skeleton fluxes into energetic and remobilization pathways.

In good correlation with transcriptomics data, a number of metabolites related to carbohydrate metabolism were negatively impacted by ABA-deficiency, notably Gal metabolism-related sugars (e.g. Gal, galactinol, galactiosylglycerol, Fru, Suc, and myo-inositol). Several of these metabolites belong to the raffinose family oligosaccharides (RFO; namely stachyose, raffinose, myo-inositol, galactinol, Suc, and Gal). RFO synthesis is closely linked to the acquisition of desiccation tolerance during seed maturation (Bailly et al., 2001). Whereas galactinol, myo-inositol, and Gal displayed higher abundance in nced2569, raffinose levels were higher in cyp707a1a2 seeds, suggesting a differential fine-tuning of RFO biosynthesis by ABA content.

In contrast, 21 metabolites showed a higher level in cyp707a1a2 compared to both nced2569 and wild type (Supplemental Table S9). Interestingly, these included Met-derived aliphatic glucosinolates and some of their breakdown products, e.g. 5-methylthio-pentanenitrile, 6-methylthio-hexanenitrile, 4-methylthiobutylglucosinolate (4-MTB), and 5-methylthiopentylglucosinolate (5-MTP). Together with the differential accumulation of proteins related to glucosinolate breakdown (Fig. 5; Supplemental Tables S5 and S6), these results further suggested a regulatory role of ABA on glucosinolate metabolism.

Seed ABA Content Influences Oxidation Events

Reactive oxygen species (ROS) and nitric oxide (NO) have been reported to play pivotal roles in the regulation of seed dormancy and seed germination (Bailly et al., 2008; Arc et al., 2013). Some clues indicated that nced2569 mutant dry seeds had to cope with oxidative stress because antioxidant compounds namely, alpha- and gamma-tocopherols, were more abundant (Supplemental Table S8). In the vitamin E group, alpha-tocopherol was reported as the most efficient free radical trap. In addition, gamma-tocopherol was described to react with nitrogen dioxide (NO₂), resulting in NO production (Cooney et al., 1993). Consistently, transcriptome data showed that the gene expression of HAEMOGLOBIN 2/AT3G10520 encoding the nonsymbiotic hemoglobin 2 was promoted in nced2569 developing seeds at 10 and 14 DAP, indicating an amplified NO response. Thus, hydrogen peroxide (H₂O₂), superoxide anion (O₂°⁻), and NO released by imbibed seeds were assayed and our results supported the assumption that ABA-deficient seeds produced more radicals than overaccumulating seeds (Fig. 6). Indeed, nced2569 seeds released more H₂O₂, O₂°⁻, and NO than cyp707a1a2 seeds, whereas wild type was intermediate. Dry seed protein oxidation profiles were also assessed by detection of carbonyl groups on alpha 12S-cruciferin subunits using Anti-2,4-dinitrophenylhydrazone (DNP) immunoassay (Fig. 6A). Alpha 12S-cruciferin subunits have been previously described as the most sensitive proteins to carbonylation in Arabidopsis dry seeds (Job et al., 2005). Our results showed a higher protein oxidation in nced2569 than in wild-type seeds. In contrast, the rate of protein oxidation was very low in cyp707a1a2 seeds, suggesting a negative correlation between ABA content and protein oxidation.

Figure 6.

Seed ABA content influences oxidation events. A, Detection of carbonylated proteins in dry seeds: Coomassie blue protein staining of alpha 12S-cruciferin subunits (top) and anti-DNP immunoassay of alpha 12S-cruciferin subunits (bottom). Three biological replicates were analyzed with similar results. B–D, O₂°⁻ (B), H₂O₂ (C), and NO (D) released by imbibed seeds. Three biological replicates were analyzed and measurements were performed using five seed samples per biological replicate. Box plots were drawn for groups of ROS and NO measurements for each genotype. The middle “grayed box” represents the middle 50% of values for the group. The median marks the midpoint of the data and is indicated by the black line. Box plots display the data distribution through their quartiles and whiskers are used to indicate variability outside the upper and lower quartiles. DSW, dry seed weight before imbibition; WT, wild type.

Furthermore, it is worth noting that our proteomic results also showed that several proteins involved in antioxidant processes were more abundant in nced2569 dry seeds (Fig. 5), highlighting the induction of protective mechanisms and influence of ABA levels on ROS production and scavenging. ABA deficiency had a positive impact on the abundance of catalase3 (CAT3/AT1G20620) and monodehydroascorbate reductases1 and 6 (MDAR1/AT3G52880 and MDAR6/AT1G63940). MDARs function as key enzymes of monodehydroascorbate recycling into ascorbic acid. Ascorbic acid plays an important role in the protection against ROS and is involved in the reduction of tocopheroxyl radical into alpha-tocopherol to prevent oxidative damage of cellular membranes.

DISCUSSION

NCED6 Expression in the Endosperm Has a Major Contribution to ABA Accumulation in Developing Seeds

ABA accumulation in Arabidopsis developing seeds and siliques has been previously shown to peak at midmaturation, as observed here (Karssen et al., 1983; Okamoto et al., 2006; Kanno et al., 2010). However, its origin has not been fully established. From crosses between an ABA-deficient aba1 female and wild-type male, genetic studies concluded that at early maturation stages ABA was synthesized in maternal tissues and at later stages in zygotic tissues (Karssen et al., 1983). More recently, an in-depth analysis of ABA content in seeds dissected from siliques of F1 (aba2-2 female and wild-type male) plants indicated that ABA was produced by zygotic tissues. However, measurement of ABA contents in single F2 seeds from this cross suggested dual maternal and zygotic origins (Okamoto et al., 2010). Maternal ABA can be provided by either vegetative or reproductive (silique/testa) tissues. As observed here and previously reported (Okamoto et al., 2010), ABA accumulation in silique envelopes is low. However, the presence of DPA in these tissues suggests active ABA synthesis and catabolism, and ABA translocation to seeds cannot be excluded.

NCED3 has been described to contribute to ABA synthesis in vegetative tissues, and be essential in water stress responses (Tan et al., 2003; Urano et al., 2009). Upon drought induction, both transcript and protein were specifically detected in vascular tissues (Endo et al., 2008), suggesting a key role for NCED3 in ABA supply to other tissues. However, the very low levels of ABA in isolated seeds at 10 and 14 DAP and both ABA and catabolites in nced2569 siliques suggest that ABA production from NCED3 activity and transport from vegetative to reproductive tissues seeds is marginal. Furthermore, in agreement with its low expression, NCED3 does not significantly contribute to ABA synthesis inside seeds. Among the five NCEDs, NCED6 transcripts are the most abundant in seeds and specifically detected in the endosperm from fertilization to maturation stages (Lefebvre et al., 2006) and NCED6 together with NCED5 and NCED9 were shown to have a major role in dormancy regulation by ABA. Unexpectedly, comparison of ABA levels in nced2569, nced259, and wild type strongly suggests that NCED6 activity in endosperm is responsible for most of the xanthoxin production that will give rise to ABA in developing seeds. It also provides further evidence that zygotic tissues, and in particular endosperm, are likely the major source of ABA during seed maturation. Nevertheless, because nced6 mutants exhibit only mild dormancy phenotypes (Lefebvre et al., 2006), ABA synthesis in restricted embryo tissues is certainly essential for dormancy induction.

ABA Catabolite Profiles Suggest New Routes in the ABA Metabolism Pathway

A previous study demonstrated that the CYP707A enzyme is responsible for ABA hydroxylation at the C-8′ and C-9′ positions (Okamoto et al., 2011). Indeed, a decrease in both DPA and neoPA, the respective end-products of the 8′- and 9′-hydroxylation pathways, was observed in cyp707a1a3-developing siliques and water-stressed plants compared to wild type. Furthermore, ABA-GE and 7′-OH-ABA levels were higher in cyp707a1a3 seeds than in wild type, as observed here in cyp707a1a2 seeds. ABA-GE has been described as a storage form of ABA, the remobilization of which is under environmental control in vegetative tissues (Lee et al., 2006; Xu et al., 2012; Ondzighi-Assoume et al., 2016). Because the expression of the β-glucosidase gene BG1 was detected in developing seeds, this conjugate could be a reusable form of ABA also in seeds. In contrast to ABA-GE and 9′-OH-ABA, 7′-OH-ABA synthesis remains obscure. Because its accumulation was observed here in cyp707a1a2, as previously in cyp707a1a3 (Okamoto et al., 2011), it is unlikely to be a side product of 8′-OH-hydroxylase activity. Nevertheless, it may have, together with ABA-GE, a role in removal of ABA excess.

A very surprising observation was the accumulation of ABA-GE in aba2 mutant. Several studies suggested the existence of minor routes for ABA production from xanthoxin, as reviewed by Endo et al. (2014). In tomato (Solanum lycopersicum) and Tex-Mex tobacco (Nicotiana plumbaginifolia) mutants defective for the conversion of abscisic aldehyde into ABA, production of abscisic alcohol has been observed, the conversion of which into ABA results in less severe phenotypes. In aba2 mutants, residual ABA levels were found and the involvement of nonspecific short-chain dehydrogenases/reductases has been hypothesized. Finally, the production of ABA from xanthoxin via xanthoxic acid has also been envisaged as a possible minor route. The biosynthesis steps that would form ABA-GE from xanthoxin remain to be investigated and the subsequent question is whether this conjugate can be a source of ABA in this mutant. Because BG1 expression is detected at high levels in developing seeds, starting from preglobular stage (http://www.bar.utoronto.ca), the existence of this alternative pathway may favor seed development and explain the residual dormancy of aba2 seeds.

ABA Differentially Regulates Specific Components of Its Own Signaling Pathway

ABA treatment has been described to down-regulate expression of a majority of PYR/PYL/RCAR genes and oppositely up-regulate most PP2C (Santiago et al., 2009; Szostkiewicz et al., 2010; Gonzalez-Guzman et al., 2012). In accordance, we found that PYL2, PYL4, and PYL6 expression was lower in wild type than in nced2569 seeds at 10 and 14 DAP. Moreover, seed transcriptome available from globular embryo to mature green stages for PYR1 and PYL1 to PYL9 (http://www.bar.utoronto.ca) and here for PYR1 and PYL1 to PYL13, at 10 and 14 DAP, showed detectable expression levels for most of them. However, only three were significantly up-regulated in nced2569 seeds, suggesting specific transcriptional regulation of the receptor gene family. Similar observations were made for the PP2C family, among which three members were down-regulated in nced2569 seeds. Therefore, during seed development, combinatorial interactions between PP2C and PYR/PYL proteins may be modulated, in a tissue-specific manner, by their transcriptional regulation by endogenous ABA. Two PP2Cs, AHG1 and AHG3, were recently identified as DOG1-interacting proteins (Née et al., 2017; Nishimura et al., 2018). Here we found that expression of DOG1 and DOG1-like genes was differentially affected by ABA levels. This observation suggests that interactions between DOG1 and ABA signaling do not only exist at the protein level, but also at the transcriptional level through the up-regulation of DOG1 by ABA, which would hence reinforce dormancy induction.

ABA Is a Repressor of ROS Production and Protein Oxidation in Seeds

Transcriptome analysis during seed development, in good accordance with dry proteome data, provided clear evidence that ABA is a broad-spectrum regulator of seed maturation, because ABA deficiency results in the down-regulation of many storage reserve and desiccation tolerance genes. Furthermore, ABA has been previously shown to repress photosynthesis during late-maturation, notably by triggering chlorophyll degradation through the activation of ABA-related transcription factors, such as ABI3 and ABI5 (Nambara et al., 1994; Delmas et al., 2013). Here the overaccumulation of a large number of transcripts and proteins related to Calvin cycle and photosynthesis was observed in nced2569 seeds, thus further extending the inhibitory role of ABA on chloroplast activity.

Chloroplasts have been described as the major generation site of ROS (Asada, 2006). Therefore, the activation of chloroplast activity in nced2569 and conversely its reduction in cyp707a1a2, compared to wild type, well correlate with the observed differences in ROS production and protein oxidation. Thus, it suggests that ABA reduces ROS levels in both developing and imbibed seeds to prevent germination. The repressive effect of ABA on oxidative processes in seeds contrast with observations in guard cells, in which ABA has been reported to induce ROS and NO production to promote stomatal closure, through the activation of NADPH oxidases by SnRK2 (Waszczak et al., 2018). Recently, it has been shown that ABA also promotes the accumulation of ROS in Arabidopsis seedlings through an antagonist crosstalk with ethylene (Yu et al., 2019). The production of radicals may be required for the transition from heterotrophic (seed) to autotrophic (seedling) development and closely related to acquisition of photosynthetic competence (Ha et al., 2017). Because seeds, seedlings, and guard cells share upstream core ABA signaling components, distinct ROS signaling pathways may operate in these tissues.

In seeds, ROS play a key role in the regulation of germination performance through specific oxidation phenomena, notably targeting proteins (Arc et al., 2011). The role of protein oxidation has been investigated in the context of seed physiology, showing that dormancy breaking is associated with an increased level in the carbonylation of specific proteins (Oracz et al., 2007; Arc et al., 2011). Furthermore, during germination, oxidized storage molecules would be more easily accessible to degradation and used as fuel for radicle protrusion and seedling establishment. Thus, the higher rate of protein oxidation in nced2569 seeds, which may result from increased metabolic activity during seed development, would facilitate seed storage protein mobilization and boost germination upon imbibition. In contrast, the very low content of oxidized proteins observed in cyp707a1a2 seeds would prevent the use of storage compounds and maintain molecular locks imposing dormancy.

Metabolism Activation in ABA-Deficient Dry Seeds Supports the Premature Initiation of Germination Programs

Seed dormancy reduction in ABA-deficient seeds might be a consequence of the decreased repression by ABA of cell-cycle, photosynthesis, and oxidative processes. In good accordance, the concomitant accumulation of free metabolites, such as sugars, amino acids, and organic acids and enzymes involved in these metabolic pathways, suggested a premature metabolism resumption state in ABA-deficient dry seeds. Indeed, high free amino acid content, as observed here for branched-chain and shikimate pathway amino acids, may contribute to seed vigor by supporting protein neosynthesis during early imbibition and fueling the TCA cycle for energy production (Fait et al., 2006). Furthermore, higher levels in intermediates of TCA and glyoxylate cycles (aconitate, citrate, fumarate, and malate) and the overaccumulation of the key enzyme of glyoxylate cycle ICL likely implies an absence of repression by ABA of energetic metabolism that would be required for dormancy induction. In accordance, ICL accumulation has been shown to occur upon nondormant seed imbibition and also during dormancy release (Chibani et al., 2006; Arc et al., 2012). Globally, our results showed an overaccumulation of free metabolites and enzymes involved in reserve remobilization, in nced2569 dry seeds compared to wild type and cyp707a1a2, before imbibition, thus likely facilitating a quick restart of primary metabolism and anabolic processes during seed germination.

ABA Influences Nitrogen and Sulfur Allocation through Glucosinolate Metabolism

The role of ABA in glucosinolate metabolism remains obscure. Here we observed that a strong metabolic signature of high ABA content in seeds was an increase in glucosinolate content. These active secondary metabolites, mainly found in Brassicaceae, derive from either Met or Trp. Their synthesis requires coordination of several enzymes involved in amino acid, sulfur, and primary metabolisms, which are subjected to a redox regulation (He et al., 2009). Hydrolysis of the Glc moiety by the myrosinase leads to the production of defense compounds (thiocyanate, isothiocyanate, and nitrile). Glucosinolate synthesis is regulated bymethyl jasmonate (Mikkelsen et al., 2003) and various abiotic stresses such as salinity, drought, and temperature (Del Carmen Martínez-Ballesta et al., 2013). Furthermore, the myrosinases TGG1 and TGG2 activities have been reported to participate in ABA and methyl jasmonate signaling pathways in guard cells (Islam et al., 2009), suggesting possible auxiliary signaling roles in stress responses. During germination of cabbage sprouts, exogenous ABA triggers glucosinolate breakdown by increasing myrosinase activity and subsequent accumulation of hydrolysis products (Wang et al., 2015). In contrast, our results indicated that, during seed development, ABA promoted both glucosinolate accumulation and catabolism. It can be hypothesized that, in the deeply dormant cyp707a1a2 seeds, the high content of glucosinolates may contribute to sulfur and nitrogen storage for the production of defense compounds as long as dormancy is maintained.

In summary, the combination of hormone profiling and omics approaches on ABA metabolism mutants gave new insights on the role of ABA during seed development and highlighted the prominent role of endosperm to provide 15-carbon precursors of ABA. ABA has been shown here to sequentially regulate multiple biological processes that globally contribute to dormancy establishment. This study also emphasized the importance of ABA in fine-tuning the protein oxidation processes and ROS production associated with dormancy release.

MATERIALS AND METHODS

Plant Material and Growth Conditions

Arabidopsis (Arabidopsis thaliana [Columbia-0 accession]) wild-type and mutant seeds were surface-sterilized, sown in petri dishes containing Arabidopsis Gamborg B5 medium (Duchefa; http://www.duchefa.com) supplemented with 30-mm Suc, and stratified at 4°C in the dark for 3 d. Petri dishes were then placed for 4 d in a growth chamber (16-h photoperiod, 50-μmol m⁻² s⁻¹ light intensity, 18°C, 60% relative humidity). Germinated seedlings were transferred to soil (Tref Substrates; http://www.trefgroup.com) and, unless otherwise stated, grown in a growth chamber (16-h light/21°C, 8-h dark/19°C, 200 μE.m⁻².s⁻¹; 65% relative humidity). In each experiment, all genotypes were grown together.

As previously described, the nced6-1 mutant was identified among the Sainsbury Laboratory Arabidopsis Transposant lines (Tissier et al., 1999), nced9-1 (SALK_033388) was obtained from the Salk database (Alonso et al., 2003; http://signal.salk.edu/cgi-bin/tdnaexpress), and nced5-2 (GK_328D05) was obtained from the GABI-Kat mutant collection (http://www.gabi-kat.de; Li et al., 2007). These were used to generate the triple mutant nced569 (Frey et al., 2012). The mutant nced2-3 (SALK_090937) was obtained from the Salk database and provided by the Nottingham Arabidopsis Stock Centre (http://arabidopsis.info). The quadruple mutant nced2569 was identified in the F2 progeny after crossing nced2-3 with nced569. The mutants aba2-2 and cyp707a1a2 were kindly provided by Dr E. Nambara. Phenotypic comparisons of allelic series of nced2, nced5, nced6, nced9, cyp707a1, cyp707a2, and some combinations of double and triple mutants have been previously reported (Lefebvre et al., 2006; Okamoto et al., 2006; Toh et al., 2008; Frey et al., 2012).

Germination Experiments

For dormancy assays, freshly harvested seeds were sown in triplicate in petri dishes containing 0.5% (w/v) agarose and then placed in a growth chamber (continuous light, 25°C, 70% relative humidity). Germination was scored each day based on radicle protrusion. After few weeks of dry seed storage at room temperature, dormancy release was analyzed on the same seed lots by scoring germination 4 d after sowing. All genotypes were grown in a glasshouse with a minimum photoperiod of 13 h, assured by supplementary lighting, and three independent seed lots were harvested for each genotype; each seed lot was obtained by pooling seeds from three to four plants.

Determination of ABA and Catabolite Contents

Siliques, isolated seeds, or envelopes and dry seeds were frozen in liquid nitrogen and then ground and freeze-dried. Samples were then extracted in acetone/water/acetic acid (80:19:1, v/v/v) with addition of 1 ng of d4-ABA, 2 ng d3-PA, 8 ng d3-DPA, 10 ng d5-ABA-GE, and 5 ng d4-7′OH-ABA isotopically labeled standards (National Research Council Canada Plant Biotechnology Institute). Their ABA content was measured using LC-MS/MS (I-Class UPLC system coupled with Xevo TQ-S, Waters; http://www.waters.com), as described in Li-Marchetti et al. (2015). Multiple Reaction Monitoring transitions for ABA catabolites were analyzed as described by Chiwocha et al. (2003).

RNA Extraction and CATMA Analysis

Developing seeds were dissected from siliques, 10 and 14 d after tagging of flowers at fertilization stage, and immediately frozen in liquid nitrogen. Total RNAs of three biological replicates were extracted using the RNeasy Plant Mini Kit (Qiagen; http://www.qiagen.com). Transcriptome profiling was carried out using Arabidopsis CATMA 6.2 arrays containing 30,834 probes corresponding to DNA coding sequences from The Arabidopsis Information Resource (v8 annotation; including 476 probes of mitochondrial and chloroplast genes) and 1,289 probes corresponding to EUGENE software predictions. Moreover, it included 5,352 probes corresponding to repeat elements, 658 probes for miRNA, 342 probes for other RNAs (ribosomal RNA, transfer RNA, small nuclear RNA, and small nucleolar RNA), and finally 36 controls (IPS2 POPS platform; http://ips2.u-psud.fr/en/platforms/spomics-interactomics-metabolomics-transcriptomics/pops-transcriptomic-platform.html). One technical replicate with fluorochrome reversal was performed for each biological replicate (i.e. four hybridizations per comparison). The labeling of complementary RNAs with Cy3-dUTP or Cy5-dUTP (Perkin-Elmer-NEN Life Science Products; http://www.perkinelmer.com) was performed as described by Lurin et al. (2004). The hybridization and washing were performed according to NimbleGen Arrays User’s Guide (v5.1) instructions. Two-micron scanning was performed with an InnoScan 900 scanner (Arrayit; http://www.arrayit.com) and raw data were extracted using the software MapixR (InnopsysR; https://www.innopsys.com).

CATMA Array Data Analysis

For each array, the raw data comprised the logarithm of median feature pixel intensity at wavelengths 635 nm (red) and 532 nm (green). For each array, a global intensity-dependent normalization using the “loess” procedure (Yang et al., 2002) was performed to correct the dye bias. The dataframes were analyzed with the software R (R Development Core Team, 2005). The “lm()” and “Anova()” functions from, respectively, “{stats}” and “{car}” packages were used to calculate global P values. Adjusted P values were calculated by a FDR Benjamini-Hochberg correction, and a threshold of 0.05 was used to select candidates. Paired comparisons were done for each variable (probe) to determine the sources of variation by the “TukeyHSD” function in the package “R {stats}.” Along with the fold change ratios, these paired comparisons results were used to determine gene expression profile variations among the three genotypes.

Microarray data were submitted to the international repository Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo; Edgar et al., 2002) with accession numbers GSE68088 and GSE125573. All steps of the experiment, from growth conditions to bioinformatic analyses, are detailed in the CATdb database (Gagnot et al., 2008; http://tools.ips2.u-psud.fr.fr/CATdb/) with Project IDs RA12-02_ABA-seed and RA14-04_nced according to the “Minimum Information About a Microarray Experiment” standards.

RNA Analysis

Total RNA was prepared from frozen dissected developing seeds using RNeasy Plant Mini Kit (Qiagen). Total RNA (1 µg) was used as a template to synthesize cDNA. Reverse transcription and qPCR reactions were performed as described in Plessis et al. (2011). Gene-specific primers are listed in Supplemental Table S10.

Protein Extraction and LC-MS/MS Analysis

Dry seeds (25 mg) were ground in liquid nitrogen and total proteins extracted in TCA-acetone, as described by Méchin et al. (2007). Proteins were solubilized in 400 μL of ZUT buffer (6 m of urea, 2 m of thiourea, 10 mm of dithiothreitol, 30 mm of Tris-HCl at pH 8.8, and 0.1% (w/v) zwitterionic acid labile surfactant I [Protea Biosciences; https://proteabio.com]). Protein concentrations were measured using the 2DQuant Kit (GE Healthcare; https://www.gehealthcare.com) and adjusted to 4 µg.µL⁻¹ before digestion. Diluted proteins (10 µL) were reduced (10 mm of dithiothreitol present in ZUT buffer) for 30 min at 24°C and then alkylated in 55 mm of iodoacetamide in 50 mm of ammonium bicarbonate for 1 h at room temperature in the dark. Digestion was performed overnight at 37°C with 600 ng of modified trypsin (Promega V5111; https://france.promega.com) dissolved in 50 mm of ammonium bicarbonate. Digestion was stopped by adding 1% (v/v) trifluoroacetic acid. Protein samples were then speed-vacuum–dried and solubilized in 20 μL of loading buffer (0.1% [v/v] formic acid in water, and 2% [v/v] acetonitrile).

Tryptic peptides were separated on a NanoLC-Ultra system (Eksigent; https://sciex.com) coupled to a Q-Exactive Mass Spectrometer (Thermo Electron; https://www.thermofisher.com), as described by Balliau et al. (2018). Eluted peptides were analyzed on-line with a Q-Exactive Mass Spectrometer (Thermo Electron) using a nano-electrospray interface. Data were acquired with the software Xcalibur (v2.1; Thermo Fisher Scientific) with the following data-dependent parameters: a full MS scan covering 300–1,400 m/z range of mass-to-charge ratio (m/z) with a resolution of 70,000 and a MS/MS scan with a resolution of 17,500 and normalized collision energy = 30%. The MS/MS scan was reiterated for the eight most abundant ions detected in the MS scan with dynamic exclusion set to 45 s.

Proteome Data Analysis

Data search was performed with the software X!Tandem (v. 2015.04.01.1; Craig and Beavis, 2004) against the Arabidopsis TAIR database (https://www.arabidopsis.org) and a homemade database containing contaminants (trypsin, keratins). Specific protein cleavage was fixed as tryptic cleavage with one authorized missed cleavage. Cys carbamidomethylation was set as a fixed modification, and Met oxidation was set as a potential modification. Additional peptide identification was performed by submitting samples to the refine mode of X!Tandem. In this second round, Ser, Thr, and Tyr phosphorylation, Gln and Asn deamidation, and Trp oxidation were set as possible modifications. Protein inference was performed by using X!TandemPipeline (v. 3.4.2; Langella et al., 2017). Selection of peptides and proteins was performed following the settings: peptide E value < 0.01, protein E value <10⁻⁵, one peptide identified per protein. Peptide quantification was performed on extracted ion currents by using the software MassChroQ (v. 2.2.1; Valot et al., 2011; http://pappso.inra.fr/bioinfo/masschroq/).

Data analysis was performed using the package MassChroQ for R. Peptides showing too much variation of their retention time and associated with too-large chromatographic peaks were removed. Normalization of peptide intensities was performed taking into account peptide median-retention time (Lyutvinskiy et al., 2013). Only protein-specific peptides present in at least two of the three biological replicates were kept for further protein quantification. Proteins quantified using fewer than two peptides were eliminated. Normalized peptide intensities were summed to obtain protein relative abundances. The final number of quantified proteins obtained was 838. A differential analysis by combining a one-way ANOVA and a multiple comparison procedure (Tukey’s honestly significant difference test) was performed on log₁₀-transformed protein abundances. Proteins with adjusted P values < 0.05 were considered as showing significant differential abundance between genotypes.

Metabolite Profiling Using Gas Chromatography-MS

Dry seeds were ground in liquid nitrogen and lyophilized. Extraction, derivatization, analysis, and data processing were performed as described by Fiehn (2006) and Clément et al. (2018). Lyophilized samples (20 mg) were extracted in 1 mL of cold water/acetonitrile/isopropanol (2:2:3, v/v/v) containing ribitol (4 µg.mL^–1) in tubes placed in a Thermomixer (10 min, 4°C; Eppendorf) before centrifugation (20,000g, 5 min). Supernatants were collected and speed-vacuum–dried for 4 h. Then, 10 µL of 20 mg.mL^–1 methoxyamine in pyridine were added to the samples, which were incubated for 90 min at 28°C under continuous shaking. Silylation of metabolites was performed for 30 min at 37°C with n-methyl-n-trimethylsilyl-trifluoroacetamide (Sigma-Aldrich). Metabolites were analyzed on a model no. 7890A Gas Chromatograph (Agilent; https://www.agilent.com) coupled to a model no. 5975C Mass Spectrometer (Agilent) with the parameters described by Clément et al. (2018). Metabolites were annotated and their levels were normalized with respect to the ribitol internal standard.

Metabolome Data Analysis

Raw Agilent datafiles were converted to NetCDF format and analyzed with Automated Mass Deconvolution and Identification System (http://chemdata.nist.gov/dokuwiki/doku.php?id=chemdata:amdis). A home retention index/mass spectra library, built from the National Institute of Standards and Technology, Golm, and Fiehn databases and standard compounds, was used for metabolite identification. Peak areas were then integrated using the software QuanLynx (Waters; http://www.waters.com) after conversion of the NetCDF file to MassLynx format.

Statistical analysis was performed with a home-made R script using the packages Car (v3.0-2) and Multicomp (v1.4-8; http://cran.r-project.org). Metabolites detected in at least 80% of the samples (i.e. 173) were selected for a subsequent quantitative analysis. One-way ANOVA and a multiple comparison procedure (Tukey Honestly Significant Difference test) were carried out on log₁₀-transformed metabolite abundances. Metabolites showing adjusted P values < 0.05 were considered.

Detection of NO Emission in Seeds

The cell membrane nonpermeable fluorescent probe 4-amino-5-methylamino-2′,7′-difluorescein (D1821; Sigma-Aldrich; https://www.sigmaaldrich.com) was used to measure NO production by imbibed seeds. This probe reacts with N₂O₃, which is the main NO autooxidation derivative and a robust indicator of NO production (Planchet and Kaiser, 2006; Liu et al., 2016; Nagel et al., 2019). The amount of N₂O₃ was determined in five dry seed samples per biological replicate using sodium nitroprusside (S0501, Sigma-Aldrich; Supplemental Fig. S5) as standard, as described by Sechet et al. (2015), and three biological replicates were analyzed for each genotype. Fluorescence intensity was measured after incubation of seeds (5 mg) during 24 h at 25°C in the dark.

Hydrogen Peroxide and Superoxide Anion Production In Seeds

Dry seeds (20 mg) were sown in petri dishes containing water-imbibed filter paper (VWR) and placed at 25°C in the light during 24 h. No germination (radicle protrusion) was observed before seeds were collected and incubated in 500 mL of 10 mm P buffer at pH 6.0, containing 5 mm of scopoletin (Sigma-Aldrich) and 10 U/mL of horseradish peroxidase (Sigma-Aldrich), and vigorously shaken for 1 h. Fluorescence measurements were performed using a Cary Eclipse fluorometer (Agilent; excitation, 350 nm; emission, 420–500 nm). A linear calibration curve was determined using increasing concentrations of H₂O₂ (0 to 20 μM). Three biological replicates were analyzed, and measurements were performed using five seed samples per biological replicate.

For the determination of superoxide anion production measurements, seeds imbibed for 24 h were incubated for 3 h at room temperature on a shaker in assay solution containing 10 mm of P buffer at pH 6.0 and 500 µM of sodium 3′-1-[phenylamino-carbonyl]-3,4-tetrazolium-bis(4-methoxy-6-nitro)benzenesulfonic acid hydrate (Sigma-Aldrich). Sodium 3′-1-[phenylamino-carbonyl]-3,4-tetrazolium-bis(4-methoxy-6-nitro)benzenesulfonic acid hydrate reduction was measured at 470 nm (ε₄₇₀ = 24.2 mM⁻¹ cm⁻¹). The blanks were done without seeds and were used to correct for unspecific absorbance changes. Measurements were performed using three biological replicates and three technical replicates for each of them.

Protein Oxidation

Carbonylated protein profiles of dry mature seeds were determined by 1D PAGE of total protein extract followed by derivatization with 2,4-dinitro-phenylhydrazine and immunological detection of the DNP adducts with monoclonal anti-DNP antibody (OxyBlot Oxidized Protein Detection Kit; Chemicon; www.merckmillipore.com) as described in Job et al. (2005).

Accession Numbers

Microarray data were submitted to the international repository GEO (http://www.ncbi.nlm.nih.gov/geo) with accession numbers GSE68088 and GSE125573.

Supplemental Data

The following materials are available.

• Supplemental Figure S1. Stature of aba2, nced2569, and wild-type plants.

• Supplemental Figure S2. ABA biosynthesis and catabolism in wild-type seeds.

• Supplemental Figure S3. Germination of nced256, nced259, nced269, nced569, nced2569, aba2-2, and wild-type seeds.

• Supplemental Figure S4. ABA levels in wild-type, nced2569, and nced259 seeds.

• Supplemental Figure S5. Nitric oxide (NO) determination using sodium nitroprusside (SNP) for calibration.

• Supplemental Table S1. Transcriptome analysis at 10 DAP revealed 285 transcripts that were significantly down-regulated (>1.5-fold, FDR, P < 0.05) in nced2569 compared to either wild type or nced259.

• Supplemental Table S2. Transcriptome analysis at 10 DAP revealed 98 transcripts that were significantly up-regulated (>1.5-fold, FDR, P < 0.05) in nced2569 compared to either wild type or nced259.

• Supplemental Table S3. Transcriptome analysis at 14 DAP revealed 853 transcripts that were significantly down-regulated (>1.5-fold, FDR, P < 0.05) in nced2569 compared to either wild type or nced259.

• Supplemental Table S4. Transcriptome analysis at 14 DAP revealed 441 transcripts that were significantly up-regulated (>1.5-fold, FDR, P < 0.05) in nced2569 compared to either wild type or nced259.

• Supplemental Table S5. GO enrichment (P < 0.05, Holm-Bonferroni) among differentially expressed genes (>1.5-fold) in nced2569 compared to wild type and nced259, proteins (>1.3-fold) in nced2569 compared to wild type and cyp707a1a2, and proteins (>1.3-fold) in cyp707a1a2 compared to wild type and nced2569.

• Supplemental Table S6. Proteome analysis in dry mature seeds revealed 68 proteins that were significantly up-regulated (one-way ANOVA, P < 0.05) in nced2569 compared to either wild type or cyp707a1a2. Ratios < 1.3 are indicated in gray.

• Supplemental Table S7. Proteome analysis in dry mature seeds revealed 52 proteins that were significantly up-regulated (one-way ANOVA, P < 0.05) in cyp707a1a2 compared to either wild type or nced2569. Ratios < 1.3 are indicated in gray.

• Supplemental Table S8. Metabolome analysis in dry mature seeds revealed the overaccumulation of 44 metabolites (one-way ANOVA, P < 0.05) in nced2569 (15 metabolites) or in both nced2569 and wild type (29 metabolites), compared to cyp707a1a2. Ratios < 1.3 are indicated in gray.

• Supplemental Table S9. Metabolome analysis in dry mature seeds revealed the overaccumulation of 21 metabolites (one-way ANOVA, P < 0.05) in cyp707a1a2 compared to both wild type and nced2569. Ratios < 1.3 are indicated in gray.

• Supplemental Table S10. Gene-specific primers used for expression analysis.