LEC1 sequentially regulates the transcription of genes involved in diverse developmental processes during seed development

Significance

Seed development is biphasic, consisting of the morphogenesis phase when the basic plant body plan is established and the maturation phase when the embryo accumulates storage reserves and becomes desiccation tolerant. Despite the importance of seeds as human food and animal feed, little is known about the gene-regulatory networks that operate during these phases. We identified genes that are regulated genetically and transcriptionally by a master regulator of seed development, LEAFY COTYLEDON1 (LEC1). We show that LEC1 transcriptionally regulates genes involved in photosynthesis and other developmental processes in early and maturation genes in late seed development. Our results suggest that LEC1 partners with different transcription factors to regulate distinct gene sets and that LEC1 function is conserved in Arabidopsis and soybean seed development.

Abstract

LEAFY COTYLEDON1 (LEC1), an atypical subunit of the nuclear transcription factor Y (NF-Y) CCAAT-binding transcription factor, is a central regulator that controls many aspects of seed development including the maturation phase during which seeds accumulate storage macromolecules and embryos acquire the ability to withstand desiccation. To define the gene networks and developmental processes controlled by LEC1, genes regulated directly by and downstream of LEC1 were identified. We compared the mRNA profiles of wild-type and lec1-null mutant seeds at several stages of development to define genes that are down-regulated or up-regulated by the lec1 mutation. We used ChIP and differential gene-expression analyses in Arabidopsis seedlings overexpressing LEC1 and in developing Arabidopsis and soybean seeds to identify globally the target genes that are transcriptionally regulated by LEC1 in planta. Collectively, our results show that LEC1 controls distinct gene sets at different developmental stages, including those that mediate the temporal transition between photosynthesis and chloroplast biogenesis early in seed development and seed maturation late in development. Analyses of enriched DNA sequence motifs that may act as cis-regulatory elements in the promoters of LEC1 target genes suggest that LEC1 may interact with other transcription factors to regulate distinct gene sets at different stages of seed development. Moreover, our results demonstrate strong conservation in the developmental processes and gene networks regulated by LEC1 in two dicotyledonous plants that diverged ∼92 Mya.

An unusual aspect of seed development is that it is temporally biphasic. After seed development is initiated with the double fertilization of the egg and central cells, giving rise to the zygote and endosperm mother cell, respectively, the embryo and endosperm undergo the morphogenesis phase. During this phase, the basic body plan of the embryo and endosperm are established through morphogenetic events that include cellular and nuclear proliferation, the specification and establishment of subregions and domains, and the differentiation of tissue and cell types (1, 2). Chloroplast biogenesis and photosynthesis are also initiated during this period in many angiosperm taxa (3). The maturation phase partially overlaps but largely follows the morphogenesis phase. During the maturation phase, cell proliferation and morphogenesis become arrested, storage macromolecules, such as lipids and proteins, accumulate to massive amounts and are sequestered in organelles, and the embryo acquires the ability to withstand desiccation (4, 5). At the end of seed development, the embryo and endosperm are arrested developmentally and quiescent metabolically, and they remain so until the seed germinates.

LEAFY COTYLEDON1 (LEC1), an unusual nuclear transcription factor YB (NF-YB) subunit of the NF-Y CCAAT-binding transcription factor (TF), is a central regulator of seed development (6). Loss-of-function lec1 mutations cause defects in storage protein and lipid accumulation, acquisition of desiccation tolerance, and the suppression of germination and leaf primordia initiation (reviewed in refs. 5 and 7). The expression of many maturation genes encoding storage proteins, oil body proteins, and transcriptional regulators of the maturation phase is defective in lec1 mutants. Moreover, ectopic expression of LEC1 induces the activation of genes involved in maturation and in storage protein and lipid accumulation in vegetative organs (6, 8–10). These findings and others implicate LEC1 and the B3 domain TFs ABA INSENSITIVE3 (ABI3), FUSCA3 (FUS3), and LEC2 as master regulators of the maturation phase (reviewed in ref. 11). Analyses of interactions among these TFs suggest that LEC1 acts at the highest level in the regulatory hierarchy controlling the maturation phase (4, 5, 9, 12). Despite its importance, knowledge of the gene-regulatory networks controlled by LEC1 is limited. LEC1 has been shown to bind to genes that are involved in lipid metabolism, hormone responses, and light signaling, and it appears to regulate transcriptionally genes involved in maturation in concert with two other TFs, NF-YC2 and basic LEUCINE ZIPPER TRANSCRIPTION FACTOR 67 (bZIP67) (8, 13, 14).

LEC1 is also required for other aspects of seed development. lec1 mutants are defective in the maintenance of suspensor and cotyledon identity early in seed development, and ectopic LEC1 expression results in somatic embryo formation on vegetative tissues (6, 15, 16). In addition to regulating maturation genes, ectopic LEC1 expression up- and down-regulates genes involved in hormone responses and down-regulates genes that respond to light in seedlings (8). These findings suggest that LEC1 controls other aspects of seed development in addition to the maturation phase. However, the LEC1 gene networks that control these diverse sets of developmental processes remain to be identified.

We present studies that provide unexpected insights into the developmental processes and gene networks that are regulated by LEC1 during seed development. mRNA transcriptome analyses of lec1-null mutants were combined with the identification of genes directly regulated by the LEC1 TF in Arabidopsis seedlings ectopically expressing LEC1 and in developing Arabidopsis seeds and soybean embryos. Together, our studies provide evidence that LEC1 regulates distinct developmental processes in seeds, including photosynthesis/chloroplast biogenesis and seed maturation. Moreover, our studies suggest that LEC1 may regulate distinct gene sets by working combinatorially with different TFs at different stages of seed development.

Results

mRNA Profiling of Developing lec1-Mutant Arabidopsis Seeds Indicates a Role for LEC1 in Several Developmental Processes.

To obtain an overview of the developmental processes that are controlled by LEC1, we profiled mRNA populations in seeds homozygous for the lec1-1–null mutation at five different stages of seed development using Affymetrix ATH1 GeneChips. The 24 h after pollination (24H), globular (GLOB), and linear cotyledon (LCOT) stages and the mature green (MG) and postmature green (PMG) stages represent the morphogenesis and maturation phases, respectively.

Fig. 1A summarizes the number of diverse mRNAs that were detected as present in lec1-1–mutant seeds compared with previously determined values for wild-type seeds at the same stages [Gene Expression Omnibus (GEO) accession GSE680] (17). mRNA numbers in lec1-1 seeds remained relatively constant throughout seed development (P > 0.91, ANOVA) in contrast to wild-type seeds in which mRNA numbers decreased significantly during seed maturation at the MG and PMG stages (P < 0.001) (17). This result is consistent with previous findings that lec1-1–mutant seeds, unlike wild-type seeds, do not become quiescent developmentally or metabolically at late seed-development stages (16).

Fig. 1.

mRNA profiling of lec1 mutant seeds throughout development. (A) The number of diverse mRNAs detected in lec1-1–mutant seeds compared with wild-type seeds (17) at the indicated seed-development stages as determined in ATH1 GeneChip hybridization studies. Representative seeds and MG and PMG embryos as viewed by bright-field (24H), differential interference contrast (GLOB and LCOT), and dark-field whole-mount (MG and PMG) microscopy (Insets, MG and PMG seeds). (B) Numbers of mRNAs differentially expressed between lec1-1 and wild-type seeds at the indicated stages define lec1⁻–down-regulated (Left) and lec1⁻–up-regulated (Right) mRNAs. The green shading and percentages denote lec1⁻–up-regulated mRNAs that also are detected at significantly higher levels in seedlings than in seeds (seedling-enriched). Lists of the mRNAs and their levels that are present in lec1-1 mutants, that are lec1⁻ regulated, and that are seedling specific are given in Dataset S1. (C) Hierarchical clustering of lec1⁻–down-regulated mRNAs. The heatmap shows relative mRNA levels in each subregion at the preglobular, GLOB, heart, LCOT, BCOT, and MG stages (left to right, as indicated by the arrow). SUS mRNAs are shown at the GLOB stage.

We designated mRNAs regulated by LEC1 as those whose levels were at least twofold higher or lower in lec1-1 mutant than in wild-type seeds at the same stage at a statistically significant level [false discovery rate (FDR) < 0.05] (Fig. 2A and Dataset S1). The lec1-1 mutation prominently affected mRNA levels during the maturation phase. Ninety-five percent of the 2,624 lec1⁻–down-regulated mRNAs that were at lower levels in lec1-1 mutant than in wild-type seeds and 99% of the 3,256 lec1⁻–up-regulated mRNAs that were at higher levels in lec1-1 mutant than in wild-type seeds at any stage accumulated differentially at the MG and/or PMG stages (Fig. 1B). Similarly, pairwise comparisons of mRNA populations in wild-type and lec1-1–mutant seeds revealed strong similarities at the 24H, GLOB, and LCOT stages (Pearson correlation coefficients, 0.99, 0.98, and 0.98, respectively) but showed more substantial differences at the MG and PMG stages (Pearson correlation coefficients, 0.81 for both). Thus, relatively few differences in gene activity between WT and lec1-1 seeds were detected early in development, but major differences were observed during the seed-maturation phase. The cause of this biased representation may be that LEC1 and LEC1-regulated genes are expressed in the embryo and endosperm, and these seed regions constitute only a small part of the seed early in development.

Fig. 2.

Design of experiments to identify genes regulated genetically and directly by LEC1 in Arabidopsis and soybean. (A) lec1⁻–down-regulated and lec1⁻–up-regulated mRNAs accumulate to a level that is at least twofold lower or higher (FDR < 0.05), respectively, in lec1-1–mutant than in wild-type seeds. (B) LEC1-bound regions were identified with ChIP-chip (EARLY and LATE) or ChIP-seq (BCOT, GmCOT, GmEM, and GmMM) experiments. (Left) Bound genes have a LEC1-binding site within the 1-kb region upstream of the transcription start site (TSS). (Right) Plant materials and antibodies used for the ChIP experiments. (C, Upper) GeneChip experiments EARLY (1 h) and LATE (4 d) after LEC1 induction identified mRNAs whose levels increased (ACT) or decreased (REP) at least twofold (FDR < 0.05) relative to the controls. (LATE data are from ref. 10.) (C, Lower Left) LEC1 is expressed in embryo subregions but not in seed-coat subregions. (Lower Center) LEC1-coexpressed mRNAs are present at fivefold or higher levels (FDR < 0.05) in embryo subregions than in seed-coat subregions. (Lower Right) The subregions compared at the indicated stages are listed. ent, endothelium; ep, embryo proper; epd, epidermis; es, endosperm; hi, hilum; ii, inner integuments; oi, outer integuments; sc, seed coat; sus, suspensor. (D, Left) LEC1 target genes are bound and regulated by LEC1. (Right) Target gene sets.

We obtained insight into LEC1-regulated processes by using hierarchical clustering to identify when and where lec1⁻–down-regulated and –up-regulated mRNAs normally accumulate, taking advantage of our previously generated dataset of mRNA levels in the embryo proper (EP), suspensor (SUS), micropylar (MCE), peripheral (PEN), and chalazal (CZE) endosperm and the distal seed-coat (SC) and chalazal seed-coat (CZSC) subregions at six different stages of development: preglobular, GLOB, heart, LCOT, bent cotyledon (BCOT), and MG (18). mRNAs affected by the lec1-1 mutation accumulated primarily in embryo and endosperm subregions in spatially and temporally controlled patterns (Fig. 1C and Fig. S1). Consistent with LEC1’s role in controlling the maturation phase, one cluster (D) of lec1⁻–down-regulated mRNAs accumulated late in development in embryo and endosperm subregions, and it was overrepresented (P < 0.001, hypergeometric distribution) for Gene Ontology (GO) terms associated with maturation processes, such as monolayer-surrounded lipid storage body, lipid storage, and seed oilbody biogenesis (Fig. 3 and Dataset S1). This cluster also contained TF mRNAs known to be involved in maturation, including ABI3, bZIP67, and ENHANCED EM LEVEL (EEL) (Dataset S1). Fig. 4 shows that 30 of 50 maturation (MAT) genes were lec1⁻–down-regulated at the LCOT, MG, and/or PMG stages. MAT genes were shown previously in mRNA transcriptome studies to be expressed predominantly during the maturation phase and to encode proteins known or predicted to function in maturation processes (18). We also reanalyzed publically available datasets to identify MAT genes that were down-regulated by mutations in two other maturation-phase regulators, ABI3 and FUS3 (GEO accession no. GSE61686) (Fig. 4) (19).

Fig. S1.

lec1⁻–up-regulated and seedling-enriched genes have similar biological functions. (A) Hierarchical clustering of lec1⁻–up-regulated mRNAs. Heatmap organization of seed subregions and developmental stages is as in Fig. 1. (B) The heatmap shows the P value significance of GO term enrichment of lec1⁻–up-regulated mRNA clusters and seedling-enriched genes. The GO terms listed correspond to the top five most enriched GO terms for each lec1⁻–up-regulated mRNA cluster. Overrepresented GO terms are listed in Dataset S1.

Fig. 3.

Predicted biological functions of lec1⁻–down-regulated and LEC1 target genes. Heatmaps show the P value (Arabidopsis, P ≤ 0.001 cutoff) and q value (soybean, q ≤ 0.05 cutoff) significance of GO terms for lec1⁻–down-regulated gene clusters and LEC1 target gene sets. The GO terms listed represent the five most enriched GO terms for each gene set. The complete GO term lists, the corresponding genes, and their significance levels are given in Datasets S1, S2, and S6.

Fig. 4.

Maturation genes and their regulators. Filled squares indicate Arabidopsis MAT genes and their closest homologs in soybean that are lec1⁻– (black), abi3⁻– (dark gray), or fus3⁻– (light gray) down-regulated at the indicated seed-development stage or LEC1 target genes at the indicated stages (LATE ACT, red; BCOT, dark green; GmCOT, forest green; GmEM, light green; GmMM, gold). Arabidopsis MAT genes with no obvious ortholog in soybean are marked NA.

Analysis of other lec1⁻–down-regulated mRNA clusters suggests that LEC1’s role in seed development is not limited to the maturation phase. For example, one cluster (E) with mRNAs that accumulated in embryo and endosperm subregions primarily at the MG stage was overrepresented for GO terms related to photosynthesis and chloroplast biogenesis (abbreviated “PSN”) (Figs. 1 and 3 and Dataset S1), suggesting that LEC1 regulates these processes directly or indirectly. Another cluster (B) of EP mRNAs was overrepresented for the GO terms organ morphogenesis and regulation of cell proliferation and contained TFs including BBM, PAN, and WOX2 that are known to be involved in embryo development. Other clusters contained mRNAs that accumulated primarily in a single subregion, including the SUS (cluster A), EP (clusters B and C), CZE (cluster G), and MCE (cluster H) (Fig. 1C), and none of these mRNA sets was overrepresented for GO terms related to maturation processes (Fig. 3 and Dataset S1).

By contrast, we found that many lec1⁻–up-regulated mRNAs were normally expressed during seedling development. Approximately 30% and 40% of lec1⁻–up-regulated mRNAs at the MG and PMG stages, respectively, overlapped with seedling-enriched mRNAs, i.e., mRNAs present at fivefold or higher levels in seedlings than in seeds at any stage (FDR < 0.05) (GEO accession no. GSE680) (Fig. 1B), and 20 of 55 and 49 of 86 overrepresented GO terms associated with lec1⁻–up-regulated mRNAs at the MG and PMG stages, respectively, were also associated with seedling-specific mRNAs (Fig. S1 and Dataset S1). This finding is consistent with reports that LEC1 is required to inhibit postgerminative development in seeds (7). Many genes encoding PSN proteins were lec1⁻–down-regulated at the MG stage and lec1⁻–up-regulated at the PMG stage, suggesting that the lec1 mutation compromised the activation of many PSN genes at or before the MG stage and their repression during the transition into metabolic quiescence (Dataset S1). Together, these results suggest that LEC1 directly or indirectly regulates a number of distinct cellular processes during seed development, including seed maturation and photosynthesis.

LEC1 Regulates Different Genes Early and Late After the Induction of LEC1 Activity in Seedlings.

The lec1⁻ mRNA transcriptome analysis suggested that LEC1 directly and/or indirectly activates and represses genes involved in diverse developmental processes. To determine which processes are regulated transcriptionally by LEC1, we identified LEC1 target genes, defined as genes that are both bound and regulated by LEC1. LEC1 target genes were identified in seedlings containing an inducible form of LEC1. Experiments were done with seedlings, because mRNAs encoding LEC1 and other TFs with partially overlapping functions, such as ABI3, FUS3, and LEC2, are not normally present at appreciable levels in seedlings (5). Fig. 2 B and C summarizes the ChIP experiments used to identify LEC1-bound genes in planta and the differential expression analyses used to identify genes regulated following the induction of LEC1 activity. As shown in Fig. 5A, the induction of LEC1 activity caused the formation of embryo-like seedlings similar to those observed in 35S:LEC1 seedlings (6, 8, 10).

Fig. 5.

LEC1 directly regulates different gene sets early and late following induction. (A) Effect of LEC1 activity on seedling development. 35S:LEC1-GR and nontransgenic seedlings were grown on medium containing Dex for 12 d. (Scale bars: 5 mm.) (B) Numbers of genomic regions and genes bound by LEC1 in ChIP-chip experiments at 4 h (EARLY) or 8 d (LATE) after LEC1 induction in seedlings. (C) Numbers of mRNAs activated (ACT) and repressed (REP) following LEC1 induction for 1 h (EARLY) or 4 d (LATE) (10) at a 0.05 FDR significance level. (D) Target genes directly regulated by LEC1. Venn diagrams show the overlap between activated and repressed genes that are bound (BD) by LEC1 EARLY (Left) and LATE (Right) after LEC1 induction. (E) Overlap between activated target genes EARLY and LATE after induction. (F) Pairwise comparisons of the genes regulated and/or bound by LEC1 EARLY or LATE after induction. The number in each row and column intersection indicates the number of genes in both lists, and the heatmap shading represents the statistical significance of the overlap. (G) Comparison of EARLY and LATE LEC1 target genes and lec1⁻-regulated genes. (H) Genome browser view of the chromosomal region surrounding the PEI gene (AT5G07500) showing enrichment of genomic regions bound by LEC1 (ChIP-chip signal relative to control, blue peaks), statistically significant LEC1-bound regions (gray bars), and gene models that are colored to indicate the mRNA fold-change following LEC1 induction (red, activated; green, repressed by LEC1; gray not present on the ATH1 gene chip). LEC1 target genes are highlighted in yellow. The axis is divided into 5-kb segments. Lists of bound genomic regions and genes, activated and repressed mRNAs, and target genes are given in Dataset S2.

Genomic DNA regions bound in planta by LEC1.

Because LEC1 is expressed from the earliest stage of seed development through maturation (18), we conducted two sets of ChIP followed by DNA microarray (ChIP-chip) experiments to identify LEC1-binding sites: one 4 h after the induction of LEC1 activity (EARLY) in 35S:FLAG-LEC1-GR seedlings 8 d after imbibition and another with 35S:FLAG-LEC1-GR seedlings that were induced for 8 d beginning at germination (LATE). Following ChIP with an anti-FLAG antibody, the immunoprecipitated DNA was amplified and hybridized with the GeneChip Arabidopsis Tiling 1.0R Array containing probes for the complete nonrepetitive Arabidopsis genome.

As summarized in Fig. 5B, ChIP-chip analyses showed that LEC1 bound 2,753 genomic regions 4 h after induction of LEC1 activity (EARLY BD), and 4,297 genomic regions 8 d after induction of LEC1 activity (LATE BD). The bound regions were within the 1-kb upstream region of 1,252 and 2,539 genes, respectively, that are represented as single genes (singletons) on the ATH1 GeneChip (Dataset S2). Control experiments shown in Dataset S3 validated the ChIP experiments and provided strong evidence that the anti-FLAG antibody bound FLAG-LEC1-GR specifically. Thus, LEC1 bound genes both early and late following induction.

Identification of genes regulated by LEC1.

Only a small fraction of the genes bound by a TF are regulated by that TF (20). For example, the chromosome browser view in Fig. 5H shows that several genes that were bound by LEC1 at a statistically significant level were not regulated by LEC1 in seedlings. Therefore, we identified genes that were regulated following a 1-h induction of LEC1 activity using ATH1 GeneChip hybridization experiments. We identified 382 EARLY ACTIVATED (EARLY ACT) mRNAs whose levels increased at least twofold and 193 EARLY REPRESSED (EARLY REP) mRNAs whose levels decreased to 50% or less of control levels following induction (FDR <0.05) (Fig. 5C and Dataset S2).

To identify genes regulated after a long-term (4-d) induction of LEC1 activity, we reanalyzed previously published data from ATH1 GeneChip hybridization experiments (10). Fig. 5C shows that 508 LATE ACTIVATED (LATE ACT) and 390 LATE REPRESSED (LATE REP) mRNAs, respectively, were up- and down-regulated relative to controls by this long-term induction of LEC1 activity. We tested mRNAs from 35S:LEC1-GR seedlings grown for 8 d following LEC1 induction using qRT-PCR and showed that 15 of 16 up-regulated mRNAs tested were validated in the 8-d induction experiments (Dataset S3).

Target genes directly regulated by LEC1.

We identified target genes that were bound by LEC1 and activated or repressed by the induction of LEC1 activity. Fig. 5 D and F shows that 16% of EARLY ACT and 14% of EARLY REP mRNAs were associated with genes that were bound by LEC1 at 4 h (Dataset S2). The overlaps between bound and regulated genes were statistically significant (P < 1.4 × 10⁻¹² and P < 7.4 × 10⁻⁵⁴, respectively, hypergeometric distribution). Similarly, 31% of LATE ACT (P < 2.5 × 10⁻²⁹) and 12% of LATE REP (P = 3.8 × 10⁻¹) mRNAs were bound by LEC1 at 8 d. The results suggest that LEC1 is involved in both the transcriptional activation and repression of genes early and late following induction of its activity.

Because the lec1⁻ transcriptome analyses suggested that LEC1 may regulate different developmental processes early and late in seed development, we compared EARLY and LATE target genes. Of the 63 EARLY ACT and 160 LATE ACT target genes, only one overlapped (P = 0.40) (Fig. 5E), and of the 27 EARLY REP and 47 LATE REP target genes only two overlapped (P < 0.02). Direct comparison of genes bound by LEC1 at 4 h and 8 d showed that there was substantial overlap in the genes that were bound at 4 h and 8 d after induction (P < 0) (Fig. 5F). By contrast, there was little overlap in the genes that were regulated by LEC1 early and late following induction. Only eight EARLY ACT and LATE ACT mRNAs overlapped (P = 0.47), and only 11 EARLY REP and LATE REP mRNAs overlapped (P < 0.001) (Fig. 5F). We did find, however, that many genes that were targets only early or late after LEC1 induction remained bound throughout the period tested. For example, 47 of 63 EARLY ACT target genes and 21 of 27 EARLY REP target genes remained bound at 8 d (P < 2.3 × 10⁻²⁹ and P < 1.9 × 10⁻¹⁴, respectively), and 25 of 160 LATE ACT targets and 17 of 47 LATE REP targets were also bound at 4 h (P < 2.1 × 10⁻⁵ and P < 1.4 × 10⁻⁹, respectively). These results suggest that LEC1 binding alone is not sufficient to regulate the expression of these genes, opening the possibility that some other factor(s) contributes to the activation and repression of LEC1 target genes early and late after induction.

We compared the EARLY and LATE targets with genes that were affected by the lec1-1 mutation and found that the most significant overlap occurred between LATE ACT targets and the lec1⁻–down-regulated genes (Fig. 5G). Analysis of overrepresented GO terms showed that the LATE ACT targets had the greatest functional overlap with the lec1⁻–down-regulated cluster D (Fig. 1C), in that they were overrepresented for the GO terms monolayer-surrounded lipid storage body, lipid storage, seed oilbody biogenesis, and seed germination, all of which are characteristic of maturation processes (Fig. 3 and Dataset S2). Moreover, of the 50 MAT genes listed in Fig. 4, 30 were LATE ACT target genes. In addition, genes encoding TFs known to play roles in controlling maturation, including LEC1, FUS3, ABI3, bZIP67, and WRI1, were LEC1 target genes (Dataset S2). LEC2 is also an LEC1 target gene, because it is bound by LEC1 at 4 d and qRT-PCR experiments showed that LEC2 was up-regulated by LEC1 induction at 8 d (Dataset S3). By contrast, EARLY ACT target genes were most significantly enriched for the GO terms positive regulation of translation, kinase activity, response to abscisic acid stimulus, TF activity, trehalose-phosphatase activity, and biosynthetic process.

Together, our results indicate that LEC1 directly activates and represses different target genes at different times after induction. LEC1 binding alone does not appear to be sufficient to regulate gene expression, opening the possibility that other TFs participate in the activation and repression of LEC1 target genes early and late after induction.

LEC1 Transcriptionally Regulates Diverse Gene Sets in Arabidopsis Seeds.

We identified LEC1 target genes in developing Arabidopsis seeds to determine if different target genes are activated at different stages of seed development as they are early and late following LEC1 induction in seedlings. We used transgenic lec1-1–mutant plants containing a LEC1-GFP chimeric gene that was fused with the endogenous LEC1 5′- and 3′-flanking regions (LEC1:LEC1-GFP:LEC1) (Fig. 2). As shown in Fig. 6 A and B, analysis of GFP activity confirmed that the transgene was active in embryo and endosperm subregions, as predicted from our previous analyses of LEC1 mRNA levels (Fig. 6C and ref. 18).

Fig. 6.

LEC1 target genes in Arabidopsis BCOT-stage seeds. (A and B) GFP fluorescence in the embryo and endosperm of a heart stage LEC1:LEC1-GFP:LEC1 seed (A) compared with the autofluorescence signal (B). (C) Relative LEC1 mRNA levels in subregions of the BCOT seed. (D) Numbers of LEC1-bound (singletons), LEC1-coexpressed, and LEC1 target genes at the BCOT stage. (E) Hierarchical clustering of mRNAs corresponding to BCOT target genes. The heatmap organization of seed subregions and developmental stages is as in Fig. 1. (F and G) Venn diagrams showing the overlap between BCOT LEC1 target genes and lec1⁻–down-regulated mRNAs at the LCOT and MG stages (F), LATE ACT targets (G, Left), and EARLY ACT targets (G, Right).

As outlined in Fig. 2B, genes bound by LEC1 in BCOT-stage seeds were identified using ChIP experiments with an anti-GFP antibody followed by DNA sequencing analysis (ChIP-seq). We analyzed BCOT-stage seeds 8–9 d after pollination because the maturation phase is initiated at approximately this stage, and LEC1 mRNA was prevalent in the embryo and endosperm at this stage (Fig. 6C). As summarized in Fig. 6D, we identified 3,703 singleton genes that were bound by LEC1 (Dataset S2). Control experiments validated the analysis and provided strong evidence that the anti-GFP antibody specifically immunoprecipitated LEC1-GFP (Dataset S3).

To identify genes that are activated by LEC1, we reasoned that their expression should be significantly higher in seed subregions containing LEC1 mRNA than in those lacking LEC1 mRNA. We profiled the mRNA transcriptomes of five seed subregions at the BCOT stage: EP, MCE, PEN, CZE, SC, and CZSC, and showed that similar numbers of distinct mRNAs accumulated in each subregion, as observed previously at other stages (Fig. S2, Dataset S4, and ref. 18). Because LEC1 mRNA was present at high levels in the embryo proper and endosperm subregions at the BCOT stage and at extremely low levels in seed-coat subregions (Fig. 6C and Dataset S4), mRNAs coexpressed with LEC1 were defined as those that are present at a fivefold or higher level in the embryo proper and/or peripheral endosperm than in the distal seed coat (FDR <0.05) (Fig. 2). We identified 1,515 genes that were coexpressed and potentially activated by LEC1 (Fig. 6D and Dataset S2).

Fig. S2.

Gene activity in BCOT-stage seeds. (A) Diagrammatic representation of the subregions in Arabidopsis BCOT-stage seeds. (B) Number of distinct mRNAs detected in each subregion and in the whole seed at the BCOT stage. The list of mRNAs and their levels are given in Dataset S4.

We identified 554 LEC1 target genes that represented a significant overlap between LEC1- bound and -coexpressed genes (P < 2.0 × 10⁻⁶⁷) (Fig. 6D and Dataset S2). Of the LEC1 targets, 176 overlapped with the 1,390 genes that were lec1⁻–down-regulated at the LCOT and MG stages (P < 1.1 × 10⁻⁷⁰) (Fig. 6F), confirming their biological significance. Moreover, the BCOT target genes overlapped significantly with the LATE ACT target genes (63 of 554, P < 4.8 × 10⁻⁵⁶) in seedlings but showed little similarity with EARLY ACT target genes (3 of 554, P = 0.25) (Fig. 6G).

We clustered the BCOT target mRNAs to obtain clues about LEC1-regulated processes in seeds and identified at least four mRNA sets with distinct spatial and temporal accumulation patterns (Fig. 6E). One cluster (O) with mRNAs that accumulated primarily in the EP at the earliest stages of seed development was enriched for GO terms related to growth and morphogenesis, including microtubule motor activity, phragmoplast, polarity specification of adaxial/abaxial axis, regulation of meristem structural organization, and asymmetric cell division, and contained TFs that play roles in morphogenetic processes in the embryo such as PHV, PHB, AS1, and SCR (Fig. 3 and Dataset S2). Another cluster (Q) contained mRNAs that accumulated in the EP, MCE, and PEN from middle to late developmental stages and had representatives of most gene families encoding the light-reaction components of photosystems I and II (Fig. S3). The great majority of these PSN target genes were also lec1⁻–down-regulated (Fig. S4). This mRNA set was overrepresented for the GO terms chloroplast thylakoid membrane, chloroplast, chloroplast envelope, thylakoid, and photosynthesis (Fig. 3 and Dataset S2). Additional LEC1 target genes that were both related to chloroplast function and lec1⁻–down-regulated were also identified (Dataset S5, Table S1), suggesting that LEC1 has an integral role in regulating photosynthesis and chloroplast functions in seeds. A maturation cluster (P) of mRNAs that accumulated at the latest stages of development in the EP and all three endosperm domains contained TFs known to regulate maturation processes, including EEL, ABI3, bZIP67, L1L, and 25 of the 50 MAT genes, although the mRNA levels of only 12 of these target genes were significantly affected by the lec1-1 mutation (Fig. 4 and Dataset S2). This maturation mRNA set was overrepresented for the GO terms nutrient reservoir activity, monolayer-surrounded lipid storage body, lipid storage, endomembrane system, and seed oilbody biogenesis. A final cluster (R) contained mRNAs that accumulated primarily in all three endosperm domains and contained TFs known to regulate maturation, including LEC1, FUS3, and WRI1, although the overrepresented GO terms were not typical of maturation (Fig. 3). Together, these results suggest that LEC1 directly regulates distinct gene sets that mediate morphogenetic processes, photosynthesis, and maturation among other cellular processes during seed development.

Fig. S3.

LEC1 target genes in Arabidopsis BCOT seeds and developing soybean seeds that are involved in photosynthesis. (A) Diagrammatic representation of complexes involved in oxygenic photosynthesis located in the thylakoid membrane. LEC1 target genes are circled in the diagram. Printed with permission from John Nield, Queen Mary University of London (London). (B) Filled boxes indicate genes involved in oxygenic photosynthesis that are LEC1 target genes at the indicated stages.

Fig. S4.

LEC1 target genes involved in photosynthesis that are lec1⁻–, abi3⁻–, or fus3⁻–down-regulated. Filled boxes indicate genes encoding components of the photosynthetic apparatus that are LEC1 target genes and/or lec1⁻–, abi3⁻–, or fus3⁻–down-regulated at the indicated stages.

Analyses of LEC1 Target Genes in Developing Soybean Seeds Indicate Different Roles for LEC1 Early and Late in Seed Development.

Our results strongly suggested that LEC1 regulates different gene sets at different stages of seed development. To verify this conclusion and to determine if LEC1’s diverse functions in seed development are conserved, we identified soybean LEC1 (GmLEC1) target genes at several stages of soybean seed development. Four LEC1 paralogs were identified in soybean, GmLEC1-1 (Glyma.07G268100), GmLEC1-2 (Glyma.17G005600), GmLEC1-3, (Glyma.03G080700), and GmLEC1-4 (Glyma.20G000600), with the first two displaying mRNA accumulation patterns most closely related to Arabidopsis LEC1 (Fig. S5).

Fig. S5.

Soybean LEC1 mRNA profiles during seed development. mRNA levels of the four GmLEC1 homologs in soybean seed subregions at the globular (GLOB), heart (HRT), cotyledon (COT), EM, and MM stages. RNA-seq data were taken from GEO series GSE57349, GSE57350, GSE57606, GSE46096, and GSE99109. Subregion abbreviations: abepd, abaxial epidermis; abpy, abaxial parenchyma; adepd, adaxial epidermis; adpy, adaxial parenchyma; ax, axis; cot, cotyledon; ent, endothelium; ep, embryo proper; epd, epidermis; es, endosperm; hg, hourglass; hi, hilum; ii, inner integuments; oi, outer integuments; pl, plumule; ps, palisade; py, parenchyma; rt, root tip; sam, shoot apex; sc, seed coat; st, stele; sus, suspensor; vs., vasculature.

GmLEC1-bound genes were identified in ChIP-seq experiments using anti-GmLEC1 antibodies and embryos at the cotyledon [GmCOT, 15 d after pollination (DAP)], early maturation (GmEM, 23 DAP), and mid-maturation (GmMM, 40–45 DAP) stages that correspond to the morphogenesis phase, transition to maturation, and the maturation phase, respectively (Fig. 2B). As summarized in Fig. 7A, we identified 16,945, 16,657, and 18,749 genes that were bound by GmLEC1 at the GmCOT, GmEM, and GmMM stages, respectively (Dataset S6), and control experiments validated the ChIP-seq results (Dataset S3). We defined genes potentially regulated by GmLEC1 at the three stages using the strategy employed to identify LEC1-coexpressed genes in Arabidopsis BCOT-stage seeds and the Harada–Goldberg Soybean Seed Development LCM RNA-Seq Dataset (GEO accessions GSE57606, GSE46096, and GSE99109; https://www.ncbi.nlm.nih.gov/geo) (Fig. 2C). Potentially regulated genes numbered 3,337, 2,751, and 3,529 at the GmCOT, GmEM, and GmMM stages, respectively, (Fig. 7A and Dataset S6).

Fig. 7.

Soybean LEC1 target genes during seed development. (A) Numbers of GmLEC1-bound, -coexpressed, and target genes at the GmCOT, GmEM, and GmMM stages. Insets show representative embryos at each stage. (Scale bars: GmCOT, 0.5 mm; GmEM and GmMM, 2 mm.) (B) Venn diagrams showing the overlap of the GmLEC1 target gene sets with the most closely related Arabidopsis BCOT target genes. (C) Venn diagram showing the overlap of the GmLEC1 target genes at the three stages. (D) Hierarchical clustering of embryo mRNA levels for the GmLEC1 target genes at the GmCOT, GmEM, and GmMM stages.

We identified 1,699 (P < 2.2 × 10⁻¹⁴⁶), 1,450 (P < 6.5 × 10⁻¹⁵⁴), and 1,983 (P < 1.5 × 10⁻¹⁸⁰) LEC1 target genes that represented a significant overlap between bound and coexpressed genes at the GmCOT, GmEM, and GmMM stages, respectively (Fig. 7A and Dataset S6). The GmLEC1 target genes at the three stages exhibited significant overlap with their orthologous LEC1 target genes identified in BCOT-stage Arabidopsis seeds. Of the 432 Arabidopsis BCOT target genes with annotated soybean homologs, 32% (P < 2.4 × 10⁻⁵⁰), 29% (P < 1.8 × 10⁻⁴⁴), and 28% (P < 2.5 × 10⁻²⁹) corresponded with GmLEC1 target genes at the GmCOT, GmEM, and GmMM stages, respectively (Fig. 7B). The results suggest that LEC1 plays similar roles in Arabidopsis and soybean seed development.

There was significant overlap in the GmLEC1 target genes at the three stages (Fig. 7C). Target genes at the GmEM and GmMM stages displayed the greatest overlap (43 and 58% of GmEM- and GmMM-stage target genes, respectively), followed by GmCOT and GmEM stages (41 and 48%, respectively). The largest numbers of stage-specific target genes were observed at the GmCOT and GmMM stages (814 and 945, respectively), suggesting that GmLEC1 regulates transitions in gene-expression programs from early to late seed development.

Hierarchical clustering of GmLEC1 target mRNA levels in embryos at the three stages (Harada Embryo mRNA-Seq Dataset, GEO accession no. GSE99571; https://www.ncbi.nlm.nih.gov/geo) provided additional support that GmLEC1 regulates distinct gene sets at different stages. Fig. 7D shows that the GmLEC1 target genes clustered into at least four groups. Cluster I mRNAs accumulated at highest levels in GmCOT-stage embryos and were most highly overrepresented for GO terms related to growth and morphogenesis, such as sequence-specific DNA-binding TF activity, nucleosome assembly, polarity specification of adaxial/abaxial axis, and determination of bilateral symmetry (Fig. 3 and Dataset S6). Clusters II and III mRNAs accumulated at highest levels in both the GmCOT and GmEM stages or in the GmEM stage and were primarily overrepresented for PSN GO terms. Cluster IV mRNAs accumulated at highest levels at the GmMM stage and were enriched for GO terms related to maturation such as lipid storage, seed dormancy process, monolayer-surrounded lipid storage body, and nutrient reservoir activity. These results are consistent with the hypothesis generated from the analyses of Arabidopsis LEC1 target genes that GmLEC1 regulates different genes involved in distinct cellular processes at different stages of seed development. The results also suggest that LEC1 function is conserved during seed development in Arabidopsis and soybean.

DNA Sequence Motifs Associated with Bound Genomic Regions Upstream of LEC1 Target Genes.

To obtain clues about the mechanisms that underlie LEC1’s ability to regulate transcriptionally distinct gene sets at different developmental stages, we identified overrepresented DNA sequence motifs in bound regions upstream of LEC1 target genes. LEC1 is an atypical NF-YB subunit of the NF-Y TF that binds the CCAAT DNA motif in association with other NF-Y subunits (21), and it also has been shown to interact with NF-YC and bZIP67, a TF that binds G-box–like motifs (13, 14). Fig. 8A and Dataset S5, Table S2 show the DNA sequence motifs that were enriched in LEC1-bound genomic regions 1 kb upstream of target genes as identified by de novo motif-discovery analyses. These motifs most closely corresponded with the G-box, ABRE-like, CCAAT, RY, and BPC1 cis-regulatory elements that are known to be involved in the control of gene transcription.

Fig. 8.

DNA motifs bound by LEC1. (A) Sequence logos showing DNA motifs identified de novo that are enriched in the bound regions of LEC1 target genes at the indicated stages with their associated E values. Only significantly enriched (E values < 0.01) DNA motifs with homology to the known cis-regulatory elements in (B) are shown. All de novo-identified DNA motifs are shown in Dataset S5, Table S2. (B) Enrichment of annotated cis-regulatory elements homologous to the enriched de novo-identified DNA motifs in the bound promoter regions of LEC1 target genes. Heatmaps show the P value for DNA motif enrichment in LEC1-bound regions relative to randomly selected regions. Motif enrichment frequencies are shown in Fig. S6 and Dataset S5, Table S2.

The motif discovery analysis was validated by quantifying the occurrence of the DNA motifs in the bound regions upstream of LEC1 target genes. Fig. 8B summarizes the P values for motif enrichment in the upstream region of Arabidopsis and soybean target genes, and Fig. S6 shows the frequencies at which these motifs were detected in upstream regions of target genes compared with comparably sized and spaced regions upstream of randomly selected genes. The G-box–like motifs, G-box (CACGTG) and ABRE-like (C/G/T)ACGTG(G/T)(A/C), were the only DNA sequence motifs that were significantly overrepresented in all LEC1 target gene sets identified in Arabidopsis and soybean. The RY motif (CATGCA) that was originally identified in the upstream region of storage protein genes (22) was significantly overrepresented in Arabidopsis and soybean LEC1 target genes except for the EARLY ACT targets. The BPC1 sequence motif (A/G)GA(A/G)AG(A/G)(A/G)A was overrepresented in all target gene sets identified in Arabidopsis and soybean seeds but not in the EARLY ACT and LATE ACT target gene sets. The CCAAT-binding sequence motif bound by the NF-Y transcription complex was significantly overrepresented only in the LEC1 target genes of Arabidopsis BCOT-stage seeds and soybean GmCOT-stage and GmEM-stage embryos. These results suggest that DNA motifs associated with LEC1 function are conserved in Arabidopsis and soybean.

Fig. S6.

DNA sequence motif enrichment in regions upstream of LEC1 target genes. Graphs show the frequency at which the DNA motifs in Fig. 8 are detected upstream of the indicated set of LEC1 target genes. Bars show the percentage of regions bound by LEC1 (colored) and the percentage of equivalently sized and spaced random selected regions (gray) that contain the indicated DNA motif. The statistical significance of the enrichment is presented above the bars. n.s., no statistically significant enrichment.

We asked if GmLEC1 target gene clusters that were differentially expressed temporally during soybean seed development were enriched for distinct DNA motifs (Fig. 7D). Fig. 8B and Fig. S7 show that all four GmLEC1 target gene clusters were enriched for the G-box–like motifs, although the enrichment was most significant for genes expressed at the latest stage (cluster IV). GmLEC1 target genes expressed at the earliest stages of seed development (clusters I and II) were enriched for the CCAAT motif, a known binding site of the LEC1 NF-Y complex. By contrast, genes expressed at the latest stage (cluster IV) were most strongly overrepresented for the RY motif. Similar results were obtained for Arabidopsis LEC1 BCOT target genes, with those expressed at early (cluster Q) and late (cluster P) stages being enriched for the CCAAT and RY motifs, respectively, and all target gene sets being overrepresented for the G-box–like motifs (Fig. S8). To determine if motif enrichment was associated with developmental function, we measured the frequencies with which motifs were linked with genes involved in (i) photosynthesis and chloroplast function (listed in Fig. S4) and (ii) maturation (listed in Fig. 4). PSN genes were significantly enriched for G-box–like and CCAAT motifs, whereas the MAT genes were significantly enriched for G-box–like and RY motifs (Fig. S9). Thus, these two functionally defined gene sets were distinguished by their enrichment for the CCAAT and RY motifs. The differential enrichment of DNA sequence motifs of genes expressed at different stages of development and of those involved in distinct physiological functions opens the possibility that LEC1 may operate in combination with different TFs to regulate distinct target gene sets.

Fig. S7.

Overrepresented DNA sequence motifs in temporally defined clusters of LEC1 target genes. The heatmap shows gene sets derived from hierarchical clustering of soybean LEC1 target genes at the GmCOT, GmEM, and GmMM stages (from Fig. 7D). Graphs show the enrichment of DNA motifs in the LEC1-bound regions for each of the temporally defined gene sets. Graphs are formatted as in Fig. S6.

Fig. S8.

DNA sequence motif enrichment of clustered BCOT target genes. (A) Hierarchical clustering of LEC1 target genes at the BCOT stage as shown in Fig. 6E. (B) Enrichment of the indicated DNA motifs in the bound regions of BCOT target gene clusters. Graphs are formatted as in Fig. S6.

Fig. S9.

DNA sequence motif enrichment of LEC1 target genes involved in photosynthesis/chloroplast functions and maturation in BCOT seeds. Enrichment of the indicated DNA motifs in the bound regions of PSN and MAT target genes is shown. Graphs are formatted as in Fig. S6.

Discussion

We profiled mRNA populations in Arabidopsis lec1-mutant seeds and identified LEC1 target genes in Arabidopsis seedlings ectopically expressing LEC1 and in developing Arabidopsis and soybean seeds to identify genes regulated directly by and downstream of LEC1. Our results demonstrate that LEC1 regulates distinct gene sets at different developmental stages, suggesting that LEC1 plays a more extensive role in controlling diverse aspects of seed development than appreciated previously.

LEC1 Transcriptionally Regulates Genes That Control Several Distinct Aspects of Seed Development.

Our results confirmed a direct role for LEC1 in controlling the maturation phase of seed development. We showed that (i) the great majority of genes differentially expressed in wild-type and lec1 mutant seeds were detected during the MG and PMG stages that encompass the maturation phase (Fig. 1); (ii) lec1⁻–down-regulated genes were overrepresented for GO terms related to maturation (Fig. 3 and Dataset S1); (iii) target genes directly regulated by LEC1 in BCOT Arabidopsis seeds and GmEM and GmMM soybean embryos were overrepresented for maturation GO terms (Fig. 3 and Datasets S2 and S6); and (iv) 26 and 25 of 50 MAT genes were lec1⁻–down-regulated and BCOT target genes, respectively (Fig. 4). These results are consistent with other reports showing that LEC1 is a master regulator of the maturation phase (15, 16, 23).

Comparison of genes that are directly vs. genetically regulated by LEC1 provides insight into the mechanism by which target gene transcription is controlled during seed maturation. Our finding that only 174 of 554 of BCOT LEC1 target genes were identified as lec1⁻–down-regulated, including 13 of 25 MAT target genes, suggests that many target genes are not regulated solely by LEC1 (Figs. 4 and 6). These results implicate the involvement of other TFs in regulating LEC1 target genes. For example, our analyses of the mRNA transcriptomes of abi3 and fus3 mutants showed that of the 12 MAT genes that were LEC1 targets but were not lec1⁻–down-regulated, eight were abi3⁻–down-regulated, and six were fus3⁻–down-regulated (Fig. 4). One interpretation of these results is that LEC1 may not be sufficient to activate some of its target genes completely and that other TFs are required to activate these genes fully. This interpretation is consistent with the findings that many maturation genes are regulated combinatorially by LEC1 and other TFs, including ABI3 and FUS3, which are both LEC1 target genes and are lec1⁻–down-regulated (24–27). Together, these results are consistent with a model in which LEC1 activates ABI3 and FUS3 as well as other target genes (9). ABI3 and/or FUS3 may play major roles in fully activating some of these LEC1 target genes, whereas LEC1 may be predominately responsible for the activation of other target genes. Our results are consistent with the conclusions of other studies showing that LEC1 acts high in the regulatory hierarchy controlling maturation by activating ABI3 and FUS3 and that ABI3 and FUS3 are dominant regulators of many MAT genes (reviewed in refs. 4, 5, 11, and 12).

Several lines of evidence indicate that LEC1 is directly involved in regulating photosynthesis and chloroplast function during seed development. First, 19 of 32 BCOT LEC1 target genes encoding components of photosystem I and II, cyt b6f, and ATP synthase complexes were also lec1⁻–down-regulated (Fig. S4 and Dataset S1). Second, BCOT LEC1 target genes and GmCOT and GmEM LEC1 targets were enriched for PSN genes (Figs. S3 and S4). Third, maturing lec1 mutant embryos are a paler green than wild-type embryos, suggesting that LEC1 is necessary to activate PSN genes fully, although LEC1 must not be absolutely required for their expression, given that lec1 mutants eventually become green (16). Fourth, our results are consistent with other studies that suggest a link between LEC1 and photosynthesis/chloroplast development. For example, LEC1 interacts with pirin to mediate blue light-induced expression of LIGHT-HARVESTING CHLOROPHYLL A/B-BINDING PROTEIN (LHCB) genes (28). Others have shown that LEC1 binds CAB4/LHCA4, LHCB5, and LHCA1 promoters in seedlings ectopically expressing LEC1, although LEC1 binding was concluded to be involved in downregulating these genes (8).

LEC1’s involvement in directly regulating genes required for photosynthesis and chloroplast biogenesis and the maturation phase is consistent with its role as a central regulator of seed development. Functional chloroplasts have been identified in Arabidopsis embryos and endosperm and soybean embryos (18, 29, 30), and we and others showed previously that photosynthesis and maturation are activated sequentially during Arabidopsis embryo and endosperm development (18, 31). Photosynthetic activity in oilseeds, such as Arabidopsis and soybean, serves a primary role in preventing anoxia through the generation of oxygen in internal tissues (29, 32–34) and enhancing carbon conversion efficiency by recycling CO₂ generated from fatty acid biosynthesis (35). Thus, LEC1 promotes photosynthesis and, therefore, fatty acid biosynthesis in oilseeds, the packaging of triacylglycerol into oil bodies, and storage protein accumulation that occurs during the maturation phase. LEC1 was first detected in land plant lineages in the lycophyte Selaginella moellendorffii (36–38). We showed previously that SmLEC1 is expressed in structures that accumulate lipids and speculated that LEC1 may have arisen, in part, in non–seed-bearing land plants to promote fatty acid biosynthesis and storage. The dual role of LEC1 in promoting photosynthetic activity and maturation processes is consistent with this hypothesis.

Analysis of LEC1 target gene clusters suggests that LEC1 regulates several other aspects of seed development. For example, soybean cluster I suggests a role for LEC1 in controlling morphogenesis and cell growth early in seed development (Figs. 3 and 7), whereas Arabidopsis clusters O and R, respectively, suggest that LEC1 controls cell division in the EP and other processes in endosperm domains throughout development (Figs. 3 and 6). Together, these results support previous hypotheses about LEC1 function, based on analyses of mutant phenotypes, that LEC1 is a central regulator of seed development (5, 7).

LEC1 Regulates Transitions in Gene-Regulatory Programs During Seed Development.

How does LEC1 directly activate different genes at different developmental stages? A potential explanation is that LEC1 may interact with different TFs to activate distinct gene sets, and the availability of these interacting TFs may be temporally regulated. LEC1 is a subunit of the NF-Y complex (21), and studies in animals and plants have shown that NF-Y complexes interact with a number of distinct TFs to regulate target gene transcription synergistically (39, 40; reviewed in ref. 41). Moreover, LEC1 has been shown to interact with (i) NF-YC2 and bZIP67 to activate maturation genes (13, 14), (ii) PIF4 to coactivate genes involved in dark-induced hypocotyl elongation (42), (iii) TCL2 to activate genes that inhibit trichome formation (43), and (iv) pirin, a protein that enhances TF binding in mammals, to regulate LHCB genes (28).

We obtained support for this hypothesis by showing that target gene regions bound by LEC1 were enriched for different DNA motifs at different developmental stages. Arabidopsis PSN target genes at the BCOT stage were enriched for the CCAAT and G-box–like motifs, whereas MAT target genes were overrepresented for the RY and G-box–like motifs (Fig. S9). Similarly, GmLEC1 target gene clusters that were enriched for genes involved in photosynthesis and chloroplast function and in maturation, respectively, were overrepresented for the CCAAT and G-box–like DNA motifs and the RY and G-box–like motifs (Fig. 8). Differences in motif enrichment may reflect, in part, the binding specificities of the TFs with which LEC1 interacts. For example, LEC1 may interact with NF-YA and NF-YC subunits to form a NF-Y complex that binds a CCAAT motif to regulate PSN genes. This hypothesis is consistent with the reports that NF-Y complexes regulate genes involved in photosynthesis (28; reviewed in refs. 44 and 45). We also suggest that LEC1 is associated with RY motifs during the maturation phase, because it acts in concert with ABI3, an RY-binding TF, at cis-regulatory modules (25). LEC1 and ABI3 may interact indirectly through their mutual physical association with bZIP TFs (24, 27, 46). Although G-box–like motifs are enriched in both PSN and MAT target genes, it is unclear if the same or different G-box–like binding TFs work with LEC1 to activate these diverse gene sets. For example, bZIP67 interacts with LEC1 and NF-YC to activate genes involved in maturation, and we showed previously that bZIP67 is not detected until after LEC1 PSN target genes are activated, decreasing the possibility that bZIP67 interacts with these genes (18). Thus, it is possible that another bZIP TF that is expressed earlier in seed development than bZIP67, such as HY5, which regulates genes involved in chloroplast function (47), works with LEC1 to activate these target genes during seed development. Alternatively, it is possible that a basic helix–loop-helix (bHLH) TF that also binds G-box–like motifs interacts with LEC1 to regulate photosynthetic genes. For example, LEC1 was shown to interact with the bHLH TF PIF4 and to bind G-box–like motifs, although this combination of TFs represses genes involved in chloroplast development.

How does LEC1 act mechanistically to regulate different target gene sets during seed development? In animals, NF-Y complexes can act as pioneer TFs that facilitate the binding of other TFs (48). For example, NF-Y binds DNA motifs in nucleosomal DNA and promotes nucleosome repositioning and an open chromatin conformation that stabilizes the binding of colocalized master regulator TFs that govern mouse ES cell identity (49). The possibility that LEC1 serves as a pioneer TF could explain, in part, the observation that LEC1 remains bound with many genes early and late following induction in seedlings even though the corresponding genes are expressed at only one stage (Fig. 5). The influence of NF-Y on chromatin conformation may be mediated, in part, by its known effects on posttranslational histone modifications that are correlated with the activation or repression of gene transcription, both in animals (reviewed in ref. 41) and plants (40). Thus, LEC1 may bind DNA and create an open chromatin conformation that allows other TFs to bind and regulate target genes during seed development.

In conclusion, our study of genes regulated genetically and directly by LEC1 has demonstrated its role in regulating distinct gene sets at different stages of seed development. In addition to confirming LEC1’s role in controlling the maturation phase, we revealed a direct role for LEC1 in controlling photosynthesis and chloroplast development and obtained evidence suggesting its involvement in other temporally and spatially regulated developmental processes, such as morphogenesis. Identification of overrepresented DNA motifs in target gene promoters suggests that LEC1 may regulate diverse target gene sets by interacting with different TFs. Moreover, our results provide strong evidence for the conservation of gene-regulatory networks that operate during seed development in two dicotyledonous plants, Arabidopsis and soybean, that diverged ∼92 Mya. The role of LEC1 in controlling two developmental processes, photosynthesis/chloroplast function and maturation, is conserved in the two species, and there are strong similarities, although not complete identity, in the target genes of Arabidopsis and soybean LEC1. We note that similarities and differences are also seen in gene networks that operate in corresponding cell types in humans and mice that also diverged ∼92 Mya (50, 51). Conservation of the developmental processes and gene regulatory networks controlled by LEC1 is consistent with the idea that LEC1 is a major regulator of seed development.

Materials and Methods

Plant Materials.

Arabidopsis and soybean plants were grown as described in SI Materials and Methods.

35S:LEC1-GR and 35S:FLAG-LEC1-GR were constructed using methods similar to those described in ref. 52; the details are provided in SI Materials and Methods. LEC1:LEC1-GFP:LEC1 was created by using PCR to add a C-terminal (Gly)₆ linker to the LEC1 cDNA followed by cloning in frame with sGFP (S65T) (53) and transferring the construct into the LEC1 expression cassette (54). Constructs were transferred into Arabidopsis Ws-0 and lec1-1–mutant plants as described (54).

lec1-1–mutant seeds were staged as described previously (17). Early LEC1-induction experiments with homozygous 35S:LEC1-GR or 35S:FLAG-LEC1-GR transgenic plants were performed as described (52). Shoot apices obtained by removing cotyledons and hypocotyls and whole seedlings were harvested. For the late-induction experiments, 35S:FLAG-LEC1-GR seedlings were grown for 8 d on 30 µM dexamethasone (Dex). Embryos harvested from soybean GmCOT, GmEM, and GmMM seeds were staged as described (55).

RNA Analysis.

Affymetrix Arabidopsis ATH1 GeneChips hybridization experiments were done as described (17). Laser-capture microdissection (LCM) experiments were performed as described (18).

ChIP.

Antibodies used for the ChIP experiments are listed in Fig. 2 and described in SI Materials and Methods. ChIP assays were performed as described (56), with the modifications detailed in SI Materials and Methods. ChIP and input DNAs for ChIP-chip experiments were quantified and prepared as described (57) with modifications listed in SI Materials and Methods and were hybridized to the Arabidopsis GeneChip Tiling 1.0R Array. ChIP-seq libraries were prepared using the NuGEN Ovation Ultralow DR Multiplex System. Libraries were size-selected by electrophoresis, purified, and sequenced at 50-bp single-end reads using an Illumina HiSeq 2000 sequencing system. qPCR validation experiments were done in triplicate, with either 30 pg of unamplified chromatin or 1 ng of amplified DNA. Primers are listed in Dataset S5, Table S3.

Data Analysis.

The mRNA profiling data were analyzed as described in refs. 18, 58, and 59 and as detailed in SI Materials and Methods. Methods used for hierarchical clustering (60) and GO term enrichment (18, 61) are described in SI Materials and Methods. ChIP-chip data were normalized using model-based analysis of tiling array (62), and significantly bound regions were identified using the CisGenome (v1.2) hidden Markov model (HMM) algorithm [posterior probability threshold 0.99999 (63)]. ChIP-seq data were analyzed using Bowtie v0.12.7 (64) and the PeakSeq algorithm of CisGenome (v2.0) as described in SI Materials and Methods. DNA sequence motifs were identified de novo using the MEME-ChIP suite (65) as described in SI Materials and Methods. Data are available at GEO under the following accessions: GSE1051 (lec1-1–mutant seed development), GSE99528 (LEC1-GR induction RNA series), GSE99529 (LEC1-GR ChIP-chip), GSE99587 (Arabidopsis BCOT ChIP-seq), and GSE99882 (soybean GmLEC1 ChIP-seq).

SI Materials and Methods

Plant Materials and Growth.

Arabidopsis plants were grown as described in West et al. (16). Soybean cv Williams 82 plants were grown in 2-gal pots containing Ron’s mix (one part coarse sand, one part compost, one part peat moss, three lb/yd Dolomite) in University of California, Davis greenhouses at temperatures ranging from 22 °C to 30 °C under long-day conditions (16-h light, 8-h dark).

To make the 35S:LEC1-GR construct, LEC1 cDNA was amplified by PCR, cloned in the pART7-GR plasmid, and then transferred to pMLBART. The 35S:FLAG-LEC1-GR construct was created by cloning the LEC1-GR fusion from the 35S:LEC1-GR construct into pART7 into which the FLAG tag had been inserted previously (annealed complementary oligos FLAG-F 5′-CGAAGCCACCATGGACTACAAAGACGATGACGACAAGC-3′ and FLAG-R 5′-TCGAGCTTGTCGTCATCGTCTTTGTAGTCGATGGTGGCT-3′ into the XhoI site), followed by transfer to the binary vector BJ49 (66). Seeds from transgenic lec1-1–mutant plants containing 35S:LEC1-GR or 35S:FLAG-LEC1-GR were not viable, similar to lec1-1–mutant seeds. Treatment of developing siliques with Dex, an inducer of LEC1 activity in these transgenic plants, rescued embryo viability, indicating that LEC1 inducibility is tightly regulated. These experiments showed that LEC1 activity was induced only after Dex treatment and that both LEC1-GR and FLAG-LEC1-GR were functional TFs.

The LEC1:LEC1-GFP:LEC1 transgene genetically suppressed the lec1-1 mutation, as judged by the rescue of embryo lethality (ratio of viable:nonviable seeds 396:82; P < 0.0001), suggesting that LEC1-GFP conferred LEC1 activity in developing seeds.

Tissue Collection.

Seeds from Arabidopsis lec1-1 mutants were collected at the stages described in Le et al. (17), frozen in liquid nitrogen, and stored at −80 °C.

For the EARLY LEC1 induction experiments (1 h and 4 h), homozygous 35S:LEC1-GR or 35S:FLAG-LEC1-GR transgenic plants were grown with or without 30 μM Dex as described by Braybrook et al. (52). Shoot apices were collected by removing cotyledons and hypocotyls and were frozen in liquid nitrogen. For LATE LEC1 induction experiments, 35S:FLAG-LEC1-GR seedlings were grown for 8 d on GM medium with or without 30 μM Dex. For the ChIP experiment, whole seedlings were collected and cross-linked.

GmCOT embryos were dissected from 2.5- to 3.5-mm-long seeds, GmEM embryos were collected from 6- to 7-mm seeds, and GmMM embryos were collected from seeds weighing 200–250 mg. Embryos were cross-linked for ChIP experiments.

RNA Isolation and GeneChip Hybridization Experiments.

Total RNA for the lec1-1 seed series and the 35S:LEC1-GR induction experiments was isolated as described by Le et al. (17). Probe synthesis, GeneChip hybridization to the Affymetrix Arabidopsis ATH1 GeneChip, and scanning were performed as described by Le et al. (17).

qRT-PCR Validations.

Total RNA was treated with the Ambion DNA-free kit. Reverse-transcription, PCR reactions, and data analysis were performed as described by Braybrook et al. (52), except that AT1G13320 was used as the reference gene. Primers used are listed in Dataset S5, Table S3.

ChIP.

ChIP assays were performed mainly as described by Gendrel et al. (56), with the following modifications. Nuclei were resuspended in the lysis buffer described by Johnson et al. (67) and were sonicated to fragment chromatin to 100–500 bp. Chromatin was incubated with the specific antibodies described below. Immune complexes were captured using Dynabeads-Protein A beads or Pierce Protein A/G beads (Thermo Scientific), depending on the antibody type. DNA was eluted, cross-links were reversed, and protein was digested according to Dahl and Collas (68), extracted with phenol:chloroform, and precipitated using standard procedures. ChIP DNA was quantified using the Invitrogen Quant-iT PicoGreen dsDNA assay kit on a NanoDrop 3300 fluorospectrometer (Thermo Fisher Scientific).

Anti-FLAG M2 (F1804; Sigma), anti-GST (Santa Cruz Biotechnology; sc-138, Upstate; 06-332), anti-GFP (75-131; NeuroMab), and anti-mouse rabbit IgGs (M7023; Sigma) were purchased. Antibodies to soybean LEC1 were raised in rabbits against either (i) a recombinant protein containing regions shared by the highly prevalent GmLEC1-1 and GmLEC1-2 but lacking the conserved B domain that is present in non-LEC1 NF-YB subunits (54), coupled with GST with a glycine (Gly)₆ linker and overexpressed in Escherichia coli BL21 cells (University of California, Davis Comparative Pathology Laboratory) or (ii) the CLEGDRTSMRGEPLGKRTV peptide, corresponding to the C domain of GmLEC1-1 and GmLEC1-2, that was conjugated with KLH for antibody production (AnaSpec). The anti-recombinant GmLEC1 antiserum was depleted for IgGs against GST, and both antibodies were affinity-purified using GST-GmLEC1-1 protein immobilized on AminoLink resin (Thermo Fisher Scientific).

ChIP-Chip and ChIP-Seq Experiments.

Preparation of ChIP and input DNAs for GeneChip hybridization experiments was done according to O’Geen et al. (57). ChIP DNA (7.5 μg) was fragmented to an average size of 100 bp using RQ1 DNaseI (Promega), biotin-labeled using the Affymetrix GeneChip WT Double Stranded DNA Terminal Labeling Kit, and hybridized to the Arabidopsis GeneChip Arabidopsis Tiling 1.0R Array at the University of California, Davis School of Medicine Microarray Core Facility.

ChIP-seq libraries were prepared using the NuGEN Ovation Ultralow DR Multiplex System and were enriched by PCR using 15 cycles. Libraries were size-selected by electrophoresis and purified using the Qiagen MinElute gel purification kit. Libraries were multiplexed and sequenced to obtain 50-bp single-end reads with an Illumina HiSeq 2000 sequencing system.

qPCR verifications of ChIP results were done in triplicate, on either 30 pg of unamplified chromatin or 1 ng of WGA2- or WGA4-amplified DNA (GenomePlex Complete Whole Genome Amplification Kit; Sigma-Aldrich) for both ChIP and input DNAs, using Bio-Rad SYBR Green Master Mix. Primers used are listed in Dataset S5, Table S3.

Data Analysis.

GeneChip RNA profiling.

The detection and normalization of GeneChip data were done as described by Belmonte et al. (18). Differentially expressed mRNAs were identified using the limma package (58) with 0.05 as the FDR threshold.

Soybean RNA-seq data analysis.

Reads from GSE57606, GSE46096, GSE99109, and GSE99571 (Harada–Goldberg Soybean Seed Development LCM RNA-Seq Dataset) were mapped to the soybean Wm82.a2.v1 gene models using the algorithm and parameters described for the GEO datasets. We used the EdgeR package (v3.10.5) to obtain normalized expression values using the trimmed mean of M-values (TMM) method and to identify genes differentially expressed among subregions (fivefold, FDR < 0.01) (59).

Hierarchical clustering.

TMM-normalized expression values were averaged across biological replicates, and hierarchical clustering of the LEC1 targets expression during embryo development was performed using dchip 2010_01 (60).

GO term enrichment analysis.

GO term enrichment analyses of Arabidopsis datasets were performed as described by Belmonte et al. (18), using a P value of 0.001 as minimum for statistically significant enrichment. For the soybean RNA-seq datasets, analyses were performed using the Bioconductor package GOSeq, the soybase GO functional annotation, the hypergeometric method, and a q value threshold of 0.05 (61).

ChIP-chip and ChIP-seq data analysis.

ChIP-chip signal was normalized using model-based analysis of tiling array (62), and significantly bound regions were identified using the CisGenome (v1.2) HMM algorithm (posterior probability threshold, 0.99999) (69).

ChIP-Seq reads were quality filtered and uniquely mapped to the genome (TAIR10 or Gmax 275) using bowtie v0.12.7 (64), allowing up to two mismatches. Redundant reads were removed using samtools_v.0.1.19 (70). Data quality was assessed by carrying the strand cross-correlation analysis using the ENCODE tool (https://code.google.com/p/phantompeakqualtools) (71). The biological replicate consistency was evaluated using the irreproducible discovery rate (IDR) pipeline (https://sites.google.com/site/anshulkundaje/projects/idr) (72) and the peak caller MACS2 (https://github.com/taoliu/MACS) (73) and applying a conservative IDR threshold of 0.01. Quality-control data are presented in Dataset S3.

Significantly bound regions in ChIP samples were identified vs. genomic input using the PeakSeq algorithm of CisGenome v2.0 (63). The bound regions were further filtered by applying a minimum FDR threshold of 0.05. Bound genes were identified using the bedtools intersect function (74). A gene is designated as bound if an LEC1-binding region overlaps with the 1-kb region upstream of the gene’s TSS (Dataset S3).

Motif enrichment analysis.

The MEME-ChIP suite was used to identify de novo DNA sequence motifs enriched in the LEC1-bound regions (65). DNA sequences from the 600 most-enriched target genes, trimmed to 100 bp centered on the region maximum (peak summit), were submitted to MEME-ChIP using a custom first-order background Markov model and an E-value cutoff of 1e-2.

The motifs identified by MEME or DREME from the MEME Suite were further screened for enrichment using HOMER (homer.ucsd.edu/homer/motif/index.html) (75) on the complete set of bound regions from target genes (250 bp, centered on the summit) against a negative set of regions, equivalent in number, length, and distribution, relative to the TSS. The cumulative hypergeometric distribution was used to test for motif enrichment significance.

Data availability.

All the data presented in this paper are available at Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under the following accessions: GSE1051 (lec1 mutant seed development), GSE99528 (LEC1-GR induction RNA series), GSE99529 (LEC1-GR ChIP-chip), GSE99587 (Arabidopsis BCOT), and GSE99882 (soybean GmLEC1 ChIP-seq).