Alterations in seed development gene expression affect size and oil content of Arabidopsis seeds.

Seed endosperm development in Arabidopsis (Arabidopsis thaliana) is under control of the polycomb group complex, which includes Fertilization Independent Endosperm (FIE). The polycomb group complex regulates downstream factors, e.g. Pheres1 (PHE1), by genomic imprinting. In heterozygous fie mutants, an endosperm develops in ovules carrying a maternal fie allele without fertilization, finally leading to abortion. Another endosperm development pathway depends on MINISEED3 (a WRKY10 transcription factor) and HAIKU2 (a leucine-rich repeat kinase). While the role of seed development genes in the embryo and endosperm establishment has been studied in detail, their impact on metabolism and oil accumulation remained unclear. Analysis of oil, protein, and sucrose accumulation in mutants and overexpression plants of the four seed development genes revealed that (1) seeds carrying a maternal fie allele accumulate low oil with an altered composition of triacylglycerol molecular species; (2) homozygous mutant seeds of phe1, mini3, and iku2, which are smaller, accumulate less oil and slightly less protein, and starch, which accumulates early during seed development, remains elevated in mutant seeds; (3) embryo-specific overexpression of FIE, PHE1, and MINI3 has no influence on seed size and weight, nor on oil, protein, or sucrose content; and (4) overexpression of IKU2 results in seeds with increased size and weight, and oil content of overexpressed IKU2 seeds is increased by 35%. Thus, IKU2 overexpression represents a novel strategy for the genetic manipulation of the oil content in seeds.

In Arabidopsis (Arabidopsis thaliana), the seed endosperm is derived from the fusion of the two maternal polar nuclei with the other sperm nucleus and thus is triploid with two maternal and one paternal genomes (Reiser and Fischer, 1993). Cell division of the fertilized egg cell results in embryo development with different morphological stages (globular, heart, torpedo, and cotyledon; Sun et al., 2010). Seed development in Arabidopsis from anthesis to maturity takes about 20 d and can be organized into two phases. The early phase encompasses the proliferation of the endosperm while the embryo develops into the heart stage and remains small. In the second phase, the embryo grows in size until it fills the cavity provided by the endosperm (Sun et al., 2010). In mature Arabidopsis seeds, the embryo fills almost the entire seed space, while the endosperm represents a thin layer around the embryo. The outermost layer of the seed is the seed coat, which is of maternal origin and is derived from the inner and outer integument cells (Haughn and Chaudhury, 2005). Embryo proliferation in the second phase is accompanied with accumulation of the seed storage compounds, in particular of proteins and triacylglycerol (TAG), which are crucial to nourish the seedling before the onset of photosynthesis (Focks and Benning, 1998). The seed accumulates large amounts of starch in the early phase but this starch is later degraded. The endosperm in the mature seed contains only low amounts of TAG that however contribute to the feeding of the seedling after germination (Penfield et al., 2004).
Embryo and endosperm development are under epigenetic control of the polycomb group complex (PgC). In Arabidopsis, the four proteins MEDEA (MEA), FERILIZATION INDEPENDENT ENDOSPERM (FIE), FERTILIZATION INDEPENDENT SEED2, and MULTI-COPY OF IRA1 are components of the PgC, which structurally and functionally resembles the Polycomb Repressor Complex2 complex of animals (Baroux et al., 2007). Control of endosperm development by the PgC is mediated via imprinting. One of the downstream factors regulated by the PgC is PHERES1 (PHE1), a type 1 MADS box transcription factor (Köhler et al., 2003). PHE1 is transiently expressed during the early phase of embryo and endosperm development. Mutations in the MEA or FIE genes result in elevated PHE1 expression during seed development, e.g. by increased methylation of 39 region of the PHE1 gene (Köhler et al., 2005;Makarevich et al., 2008). Ovules of heterozygous fie mutants carrying a mutant allele undergo endosperm development without fertilization. The unfertilized embryo reaches early heart stage before it aborts (Ohad et al., 1996). It is believed that the derepression of the maternal PHE1 allele in polycomb group (PcG) mutants causes endosperm overproliferation and seed abortion (Köhler et al., 2005;Makarevich et al., 2008). Homozygous phe1-1 mutants plant carrying a transposon insertion showed no difference in vegetative and reproductive development (Köhler et al., 2005).
In addition to the PcG pathway, further genes, e.g. MINISEED3 (MINI3) and HAIKU2 (IKU2), are involved in the regulation of seed development (Garcia et al., 2003;Luo et al., 2005). MINI3 and IKU2 encode a WRKY10 transcription factor and a Leucine-rich repeat (LRR) kinase, respectively. MINI3 and IKU2 are expressed in developing endosperm during the early stage, and MINI3 is additionally expressed in the embryo at the globular stage (Luo et al., 2005). Homozygous mutant plants of mini3 and iku2 produce smaller seeds. The exact mechanism or downstream factors of the MINI3/ IKU2 pathway remain unknown. MINI3 and IKU2 are under control of the SHORT HYPOCOTYL UNDER BLUE1 (SHB1) gene (Zhou et al., 2009). SHB1 harbors SPX (for SYG1, PHOSPHATE TRANSPORTER81 and XENOTROPIC AND POLYTROPIC MURINE LEU-KEMIA VIRUSES RECEPTOR1 proteins) and EXS (for Endoplasmic Reticulum RETENSION-DEFECTIVE1, XENOTROPIC AND POLYTROPIC MURINE LEU-KEMIA VIRUSES RECEPTOR1 proteins) domains, which are found in proteins such as yeast (Saccharomyces cerevisiae) SYG1 (for suppressor of yeast gpa1 [Guanine nucleotide-binding protein a-1 subunit]) or mouse XPR1 (for XENOTROPIC AND POLYTROPIC MU-RINE LEUKEMIA VIRUSES RECEPTOR1), but the functions of these domains remain unclear (Kang and Ni, 2006). The SHB1 protein associates with the promoters of MINI3 and IKU2, thereby inducing their expression and triggering endosperm proliferation and seed growth. In line with this scenario, the shb1-D mutant, which confers SHB1 overexpression, produces larger seeds, while homozygous mutant seeds of a loss of function allele, shb1, are smaller (Zhou et al., 2009).
While the role of the PgC genes and of MINI3 and IKU2 during endosperm and embryo development has been studied, only little is known about the role of these genes in seed storage compound allocation. We selected four genes, FIE and PHE1 of the PcG regulative pathway, MINI3, and IKU2 to study the changes in carbohydrate, protein, and TAG accumulation in the respective loss-of-function mutants. From these studies, it became clear that a reduced seed size in the knockout mutants is primarily associated with reduced TAG content in the embryo. Furthermore, transgenic lines with strong seed-specific overexpression of the four genes were generated. The results demonstrate that overexpression of specific seed development genes can be employed to increase embryo size and TAG content in transgenic Arabidopsis seeds.

Isolation of Mutants of Arabidopsis Seed Development Genes
Embryo and endosperm development in seeds are strictly controlled and depend on the concerted action of different genetic factors. To study the influence of alterations in the seed developmental program on metabolism and on the accumulation of the storage components, Arabidopsis genes known to be involved in seed development were selected for analysis: FIE (At3g20740), PHE1 (At1g65330), MINI3 (At1g55600), and IKU2 (At3g19700). Mutant lines were derived from the original alleles (fie-11, phe1-1, mini3-1, mini3-2, and iku2-3; Guitton et al., 2004;Köhler et al., 2005;Luo et al., 2005). Furthermore, new alleles carrying transfer DNA (T-DNA) or transposon insertions were obtained from stock centers (fie-13, fie-14, phe1-6, phe1-7, mini3-3, and iku2-4; Fig. 1; Supplemental Table S1). Two alleles with T-DNA insertions between start and stop codon (fie-13 and fie13) were obtained. Two insertion lines were collected for phe1, carrying insertions between the start and stop codon, with the insertion site of phe1-7 very close to the 39 end of the open reading frame. For MINI3, two insertion lines were isolated harboring a T-DNA between start and stop codon. Furthermore, a T-DNA line, iku2-4, was obtained. The T-DNA insertion in line iku2-4 is located at the 39 site of the start codon.
Homozygous mutant lines for phe1, mini3, and iku2 were obtained after PCR screening of genomic DNA from plants obtained after selfing of heterozygous lines. For the fie mutant lines, homozygous plants cannot be isolated, because ovules carrying the fie allele cannot develop into fertile seeds (Ohad et al., 1996). Expression of the genes was studied by reverse transcription (RT)-PCR of RNA extracted from young siliques (5 d post pollination; Fig. 1). The RT-PCR signal of the PHE1, MINI3, and IKU2 genes was strongly suppressed in the corresponding homozygous mutants compared with the wild-type lines. A faint RT-PCR signal was detected in the phe1-7 line, indicating that there is residual expression. For the other mutant lines, no RT-PCR signal could be detected, suggesting that they represent null alleles.
In agreement with previous results that showed that loss-of-function mutations in the genes MINI3 and IKU2 result in reduced seed size (Luo et al., 2005), the seed length and width and the seed weight of the lines mini3-2, mini3-3, iku2-4, mini3-1, and iku2-3 were reduced as compared to their wild-type controls, ecotypes Columbia (Col-0) and Landsberg erecta (Fig. 1, C and D). Seeds sizes (length and width) of phe-1, phe1-7, and phe1-6 were only slightly reduced compared with the respective controls, ecotypes Landsberg erecta and Nossen, while their weight was strongly reduced (Fig. 1,C and D). Close inspection of the seeds under the microscope revealed that the surfaces of the mutant alleles of mini3, iku2, and phe1 were shriveled, indicating a reduced accumulation of storage compounds during seed maturation (Supplemental Fig. S1). The fact that the surface of phe1 seeds was shriveled can explain why the weight of the seeds was reduced to a much stronger extent than seed size.

Oil, Protein, and Sugars in Mutants of Arabidopsis Seed Development Genes
The changes observed in seed size and weight of the phe1, mini3, and iku2 mutants prompted us to analyze the contents of seed storage compounds. TAG and proteins are the most abundant forms for carbon deposition in Arabidopsis seeds (Focks and Benning, 1998). Total fatty acids were quantified in the mutant seeds by gas chromatography (GC) of methyl esters as a measure for the content of storage oil. Total fatty acids amount to approximately 7 to 8 mg per seed in the Arabidopsis wild type (Baud et al., 2002;Li et al., 2006; Fig. 2A). This amount was reduced in all mutant lines by approximately 15% to 40%. The fatty acid composition revealed no major changes between the mutants and the controls (Supplemental Fig. S2). There was a tendency that the amount of linoleic acid (18:2) was reduced while the a-linolenic acid (18:3) content was increased in the lines mini3-2, mini3-3, phe1-1, phe1-6, and phe1-7.
Furthermore, seed storage protein was measured. The total protein content was slightly reduced in the mutant lines, in particular in min3-1, mini3-2, phe1-1, and iku2-3 (Fig. 2B). Furthermore, Suc was determined photometrically using an enzyme-based assay. Suc is known to be imported into the embryo and to represent the main source for acetyl-CoA production and oil biosynthesis in the seed (Baud et al., 2005). The amount of Suc in mature Arabidopsis wild-type seeds is around 0.5 mg per seed (Focks and Benning, 1998;Baud et al., 2002;Fig. 2C). The Suc content in mature mini3 and iku2 seeds was similar to corresponding controls (Fig. 2C). The seeds of phe1 lines contained an increased amount of Suc. These results demonstrate that the reduction in weight observed in the seed development mutants is mainly based on a reduction in oil, while the impact on storage protein is smaller.

Accumulation of Storage Compounds during Seed Development
Storage compound accumulation was recorded in developing seeds of the mini3, phe1, and iku2 mutants to study the metabolic changes associated with the reduced oil and protein contents. Seeds were harvested at 7, 11, 15, and 19 d after flowering (DAF). The seeds of two lines of each mutant were subjected to the  Table S1). B, Expression analysis of seed development genes in the phe1, mini3, and iku2 mutants. The RT-PCR products obtained with RNA isolated from developing siliques (5 DAF) of homozygous mutant plants (phe1, mini3, and iku2) were separated in agarose gels and stained with ethidium bromide. Ubiquitin-specific primers were used as control. C, The seed size (length, black bars; width, gray bars) was determined after scanning the seeds using the Evaluator software. Data represent mean and SD of a minimum of 100 seeds. D, Seed weight obtained by measuring 100 seeds (n = 3, mean and SD). Mutants are organized according to their respective ecotype. Student's t test, *P , 0.05, **P , 0.01. determination of fatty acids, protein, starch, Suc, and Glc contents (Fig. 3). Seed fatty acid content increased from approximately 0 to approximately 5 mg per seed in the wild type, and it was reduced in all mutant lines from day 11 on. At 19 DAF, fatty acid content in the wild type was still lower than 7 to 8 mg per seed as measured in mature seeds (Fig. 2), indicating that oil accumulation and seed development were not finished. Protein content in the wild type was 2 to 3 mg per seed at 7 DAF, and it increased to approximately 5 mg per seed at 19 DAF. The mutant seed protein content was lower in some mutant lines. Starch is known to accumulate in the early phase of seed development. At 7 DAF, the starch content was highest, and it declined during the later phase of seed development to reach values of 0 to 1 mg per seed at 19 DAF in the wild type. The starch degradation in all mutant seeds was retarded. Therefore, the starch content in the seeds of mini3, phe1, and iku2 was elevated at 19 DAF. Suc accumulated in the wild type from approximately 0.2 mg per seed at 7 DAF to approximately 0.4 mg per seed at 19 DAF in a linear way. The Suc contents of the iku2 Figure 2. Oil, protein, and Suc content in mutants of Arabidopsis seed development genes. A, Total fatty acids were measured in seeds of the phe1, mini3, and iku2 mutants after transmethylation by GC. B, Total protein content of seeds was determined photometrically. C, Suc in seeds was measured enzymatically. Data show mean and SD of at least three measurements of five seeds (fatty acids), 20 seeds (protein), and 50 seeds (Suc) each. Mutants of phe1, mini3, and iku2 are organized according to their respective ecotype. Student's t test, *P , 0.05, **P , 0.01. Figure 3. Accumulation of fatty acids, protein, starch, Suc, and Glc in phe1, mini3, and iku2 mutants during seed development. Flowers were tagged at anthesis, and seeds were collected after 7, 11, 15, and 19 DAF. The isolated seeds were employed for the determination of total fatty acids (A), protein (B), starch (C), Suc (D), and Glc (E). Data represent mean 6 SD of five measurements. Similar results were obtained in an independent biological experiment. and mini3 mutant seeds were similar to the wild type. In the phe1 seeds, the amount of Suc was comparable to approximately at 7, 11, and 15 DAF, but it was elevated at 19 DAF. The Glc content declined during development of wild-type seeds. Glc content in the mutant seeds was, in general, slightly lower, and it was similar to the wild type at 19 DAF, except in mini3-1 seeds, which contained less Glc. Taken together, major changes in storage compound synthesis were associated with a reduction of oil and protein accumulation, while starch degradation in the mutant seeds was retarded.

Changes in Oil Content and Composition in fie Mutant Seeds
After selfing, endosperm and embryo development in mutant ovules of heterozygous fie plants is initiated, but the embryo is aborted at the early heart stage (Ohad et al., 1996;Vinkenoog et al., 2000). The endosperm finally dries out and forms extremely small, infertile seeds (Ohad et al., 1996;Supplemental Fig. S1). To study the oil accumulation in the fie mutant, a large number of aborted seeds was collected from a heterozygous fie plant, and total fatty acids were measured by GC. Figure 4A shows that the oil content of the aborted fie seeds is extremely low, at around 0.05 mg per seed, compared with 8 mg in Col-0 or C24 seeds. Analysis of the fatty acid composition of the aborted seeds of three fie mutant alleles revealed characteristic differences to the wild-type controls. The mutant lines contained an increase in the saturated fatty acids palmitic acid (16:0) and stearic acid (18:0), a decrease in linoleic acid (18:2), a-linolenic acid (18:3), and eicosenoic acid (20:1), and an increase in erucic acid (22:1; Fig. 4B). TAG molecular species were measured in whole wild-type C24 and aborted fie-11 seeds by quadrupole time-of-flight (Q-TOF) mass spectrometry. On a per seed basis, the amounts of all molecular TAG species were drastically reduced in the fie-11 mutant. Compared on a percentage basis, the molecular species containing saturated fatty acids (16:0 and 18:0) were increased, while those rich in unsaturated fatty acids, in particular 18:3 and 20:1, were decreased (Fig. 4C).

Overexpression of IKU2 Leads to an Increased Seed Size and Weight
To study the effect of an increased expression level on seed development, the four genes FIE, PHE1, MINI3, Figure 4. Lipid composition of the fie mutant. Lipids were measured in aborted seeds derived from mutant ovules of heterozygous fie mutant plants after selfing. A, Total fatty acids were determined by transmethylation and GC. B, Fatty acid composition measured by GC of fatty acid methyl esters. C, TAG molecular species were quantified by Q-TOF mass spectrometry. Data represent mean and SD of three (GC) or five (Q-TOF) measurements of five seeds (the Col-0 and C24 wild type) or 25 seeds (fie-11 seeds). Student's t test, *P , 0.05, **P , 0.01. and IKU2 were overexpressed in transgenic Arabidopsis plants under control of the seed-specific glycinin promoter from soybean (Glycine max). For each construct, at least 15 independent transformants were isolated. The expression level of the transgenes was measured by RT-PCR of RNA isolated from siliques harvested approximately 13 DAF. Three independent lines were selected for each construct that showed strong expression of FIE, PHE1, MINI3, and IKU2 as compared with an empty vector control (Fig. 5A).
The seeds were first visually inspected under a fluorescence microscope. Because the overexpression constructs harbor a DsRed marker gene, transgenic seeds can easily be identified by their red fluorescence under green light. For the T2 seeds of the different T1 lines, a three-to-one segregation of fluorescent to nonfluorescent seeds was observed, indicating a heterozygous integration of the T-DNA. Figure 5B shows transgenic (top) and nontransgenic seeds (bottom) for one representative line for each construct. The seed sizes for the overexpressing lines FIE-OE, PHE1-OE, and MINI3-OE were very similar as the Col-0-empty vector (EV) control and the nontransgenic segregant seeds. Overexpression of IKU2, however, resulted in seeds that were larger than Col-0-EV. Seed length and width of the fluorescent (transgenic) seeds was determined after scanning the seeds. In agreement with the microscopic analysis, seed length and width of FIE-OE, PHE1-OE, and MINI3-OE lines were unchanged. For the IKU2-OE seeds, the length was increased by approximately 20%, while the width was similar to the control (Fig. 5C). In agreement with the seed size measurement, the weight of the transgenic seeds of the lines FIE-OE, PHE1-OE, and MINI3-OE was not altered as compared with Col-0-EV. However, the weight of the transgenic seeds of the three IKU2-OE lines was increased by approximately 30% (Fig. 5D).

Overexpression of IKU2 Results in an Increase in Oil
To study the impact of overexpression of seed development genes on the allocation of storage compounds, total fatty acids, total protein, and Suc were measured in the seeds of the lines FIE-OE, PHE1-OE, MINI3-OE, and IKU2-OE. Total fatty acid content was very similar to the control in the lines FIE-OE, PHE1-OE, and MINI3-OE, but it was increased by approximately 35% in the seeds of IKU2-OE plants (Fig. 6A). The fatty acid composition of the seeds of all lines as determined by GC was not altered (not shown). Total protein was similar to the wild type in mature seeds of the transgenic lines (Fig. 6B). The content of Suc was similar to the control in the seeds of the overexpressing lines (Fig. 6C). Therefore, only the overexpression of IKU2 resulted in an increase of storage oil, Figure 5. Overexpression of seed development genes in Arabidopsis. The cDNAs of the genes FIE, PHE1, MINI3, and IKU2 were cloned behind the seed-specific glycinin promoter and transferred into transgenic Arabidopsis plants. A, Expression of FIE, PHE1, MINI3, and IKU2 in overexpression lines of Arabidopsis. Developing siliques were harvested (approximately 13 DAF) from three independent plants each and used to isolate RNA for RT-PCR. Col-0 plants transformed with an empty vector (Col-0-EV) were used as control. The four lanes (1-4) of Col-0-EV represent RT-PCR reactions using primers for FIE, PHE1, MINI3, and IKU2, respectively. The RT-PCR products were separated in agarose gels and stained with ethidium bromide. Ubiquitin-specific primers were used as a reference. Note that expression of the four genes in the control lanes (Col-0-EV) is much lower compared with Figure 1 because the RNA was isolated from older siliques. B, Size of T2 seeds harvested from transgenic plants overexpressing FIE, PHE1, MINI3, or IKU2. Transgenic T2 seeds were selected based on their red fluorescence (DsRed marker). The photo shows transgenic seeds (top) and wild-type segregants (bottom) of the same plant. Bar = 0.5 mm. C, The size (length and width) of transgenic T2 seeds. Data show mean and SD of approximately 100 seeds. D, The weight of transgenic T2 seeds was determined by weighing three times 100 seeds each. Data show mean and SD. Student's t test, *P , 0.05, **P , 0.01. [See online article for color version of this figure.] in agreement with the increase in seed size and weight of IKU2-OE lines (Fig. 5).

Accumulation of Storage Compounds in IKU2-OE Seeds during Development
Two IKU2-OE lines were selected for the analysis of seed storage compound accumulation during seed development (Fig. 7). Total fatty acid content was indistinguishable from the wild type at 7 DAF, but already at 11 DAF, it was higher in IKU2-OE1 and IKU2-OE2 seeds, and it stayed elevated at 15 and 19 DAF. The protein content of the IKU2-OE seeds was also slightly Figure 6. Content of oil, protein, and sugars in seeds of transgenic overexpression plants. T2 seeds from heterozygous plants overexpressing FIE, PHE1, MINI3, or IKU2 were selected based on their red fluorescence. A, Total fatty acid content was measured by GC after transmethylation. B, Protein content was measured photometrically. C, The content of Suc was determined enzymatically. Data show mean and SD of at least three measurements of five seeds (fatty acids), 20 seeds (protein), and 50 seeds (Suc) each. Student's t test, *P , 0.05, **P , 0.01. higher at all time points. Interestingly, the amount of starch at 7 DAF was strongly increased in IKU2-OE seeds to 8 to 10 mg per seed, about two times higher as compared with the wild type. Then, the starch was degraded such that it was similar to the wild type at 11, 15, and 19 DAF, when it amounted to approximately 1 mg per seed. The pattern of Suc accumulation showed an opposite effect, as it was lower than the wild type at 7 DAF in the two IKU2-OE lines, but it reached values similar to the wild type at 19 DAF. The Glc content was similar to the wild type at all time points.

Seed Yield in Transgenic IKU2 Overexpressing Plants
The overexpression of IKU2 in the seeds of transgenic plants resulted in the production of larger seeds with increased seed weight and oil content. To study the impact of IKU2 overexpression on seed yield, the silique length, number of seeds per silique, and the total amount of seeds per plant were determined for three independent heterozygous transgenic IKU2-OE plants. As shown in Figure 8, A and B, the silique length and number of seeds per silique were not altered in the IKU2-OE plants. Figure 8C shows that the seed yield for all three IKU2-OE lines was reduced by approximately 25% to 50% compared with the Col-0-EV control. Therefore, IKU2 overexpression has no impact on the number of seeds per silique, but the total weight of seeds per plant is decreased.

The Increase in Seed Size of IKU2-OE Plants Is Suppressed in Homozygous Transgenic Plants
Overexpression of IKU2 resulted in an increase in seed size, weight, and oil content when transgenic seeds of a heterozygous line were analyzed (Fig. 5). We obtained T2 lines homozygous for the IKU2 overexpression construct from each of the three independent T1 plants, as judged by the segregation of the DsRed marker. Figure 9 shows the results for homozygous and heterozygous seeds obtained from the IKU2-OE1 line. The siliques from heterozygous IKU2-OE1 plants where thicker compared with siliques from a homozygous IKU2-OE1 line or the controls (Col-0-EV1 or Col-0; Fig. 9A). Developing seeds of the late cotyledon stage from the different siliques were microscopically analyzed after thin sectioning and toluidine blue staining. The embryos of developing seeds of homozygous IKU2-OE1 plants had a size similar to Col-0-EV1, while seeds of a heterozygous IKU2-OE plant contained embryos of different sizes, i.e. small embryos with a size comparable to Col-0-EV1 (Fig. 9B, left) and large embryos (Fig. 9B,  right). This result was confirmed by the analysis of mature seeds under the fluorescent microscope. Seeds of a homozygous IKU2-OE1 plant were of similar size as Col-0-EV seeds. Seeds from the heterozygous IKU2-OE1 line contained different types of seeds, i.e. large fluorescent (transgenic, heterozygous) seeds, small fluorescent (transgenic, homozygous) seeds, and large nonfluorescent seeds (wild-type segregants), in a ratio of about 1:2:1. The measurement of length and width of the seeds confirmed these results. The length of the seeds of a homozygous IKU2-OE1 line was similar to that of a Col-0-EV line (Fig. 9D). Seeds of a heterozygous IKU2-OE1 line were separated into two classes, i.e. fluorescent (transgenic, homozygous/heterozygous) and nonfluorescent (nontransgenic segregant) seeds. Figure 9D shows that the mean of the length of the transgenic seeds was larger compared with seeds of a Col-0-EV or homozygous IKU2-OE line (Fig. 9D).
Interestingly, the length of the nontransgenic segregants of the IKU2-OE1 line was also larger than that of Col-0-EV1, and it was similar to the large fluorescent (heterozygous) seeds (Fig. 9D). This increase in the size Siliques and seeds of Col-0-EV and of three independent heterozygous IKU2-OE plants were analyzed. A, Silique length (n = 10, mean and SD). B, Numbers of seeds per silique (n = 10, mean and SD). C, Total weight of seeds per plant (mean and SD, n = 7-10). Student's t test, *P , 0.05, **P , 0.01. of nontransgenic segregant seeds from a heterozygous IKU2-OE line can already be observed after visual inspection of the seeds (Fig. 9C) and is indicative for a maternal effect on seed size of the heterozygous IKU2-OE plant.
A reciprocal cross of a homozygous IKU2-OE line with wild-type Col-0 was performed to unravel whether the increased seed phenotype, which is suppressed in the homozygous stage, can be derepressed in the F1 generation. Figure 9E shows F1 seeds from the reciprocal crosses of Col-0 with a homozygous IKU2-OE plant. Furthermore, wild-type Col-0 seeds were mixed to the F1 seeds of the reciprocal crosses. The fluorescent (heterozygous) IKU2-OE seeds are clearly larger compared with the small nonfluorescent Col-0 seeds. This result was independent of the direction of the cross. The finding was confirmed by measuring the size of the seeds, because the length of the heterozygous IKU2-OE F1 seeds derived from the reciprocal crosses (0.572 6 0.039 and 0.550 6 0.032 mm for Col-0 3 IKU2-OE and IKU2-OE 3 Col-0, respectively) was larger than that of the control Col-0 (0.495 6 0.035 mm). Therefore, the increase in seed size was only observed in heterozygous IKU2-OE seeds, but not in homozygous IKU2-OE seeds.
To study the molecular origin for the suppression of the increased seed size in homozygous IKU2-OE lines, IKU2 expression was measured by RT-PCR in siliques harvested at 13 d after pollination of Col-0-EV plants, heterozygous IKU2-OE plants, and homozygous IKU2-OE plants. Figure 9F shows that IKU2 expression is low in the Col-0-EV siliques and is strongly increased in the siliques of the heterozygous IKU2-OE line. However, expression of IKU2 was also low in seeds of the homozygous IKU2 plant, similar to the Col-0-EV control. Therefore, expression of IKU2 is silenced in homozygous IKU2-OE seeds, presumably by cosuppression, due to the combination of two transgenic IKU2-OE alleles in a single seed. This result demonstrates that the Figure 9. Variation in sizes of seeds from homozygous and heterozygous IKU2-OE plants. A, Siliques from a heterozygous IKU2-OE1 plant are thicker than those from a homozygous IKU2-OE1 plant or from the controls (Col-0, Col-0-EV1). Bar = 5 mm. B, Light microscopy of immature seeds of the late cotyledon stage (14 DAF) after thin sectioning and toluidine blue staining. Representative seeds from siliques from Col-0-EV and a heterozygous (two adjacent seeds) and homozygous IKU2-OE1 plant are shown. Note that siliques of a heterozygous IKU2-OE plant contain large (heterozygous, nontransgenic segregants) or small seeds (homozygous). Bar = 200 mm. C, Mature seeds from a Col-0-EV and heterozygous and homozygous IKU2-OE plant. The photo shows red fluorescence of the DsRed marker. Seeds of a homozygous IKU2-O1 plant are small, similar to Col-0-EV1. Seeds of the heterozygous IKU2-OE1 plant segregate into three classes, large fluorescent seeds (heterozygous, left), small fluorescent seeds (homozygous, right), and large nonfluorescent seeds (nontransgenic segregants, center). Bar = 500 mm. D, Size measurement of transgenic seeds from a heterozygous Col-0-EV1 plant, fluorescent/transgenic (homozygous/heterozygous) seeds of a heterozygous IKU2-OE1 plant, wildtype segregant seeds of a heterozygous IKU2-OE plant, and seeds of a homozygous IKU2-OE1 plant. The length of the transgenic and wildtype segregant seeds of a heterozygous IKU2-OE plant is larger than that of Col-0-EV or homozygous IKU2-OE seeds. E, F1 seeds from reciprocal crosses of wild-type Col-0 with a homozygous IKU2-OE1 plant. The photo shows fluorescent seeds of Col-0-EV and F1 seeds obtained after reciprocal crosses (Col-0 3 IKU2-OE1, IKU2-OE1 3 Col-0). Nontransgenic, nonfluorescent Col-0 seeds were mixed to the F1 seeds as an internal size control. F, Expression analysis of IKU2 in seeds of heterozygous and homozygous IKU2-OE lines. Total RNA was isolated from immature siliques (approximately 13 DAF). The RT-PCR products were separated in agarose gels and stained with ethium bromide. Expression in seeds of a homozygous plant is severely repressed compared with seeds from the heterozygous plant and is even lower than in seeds of a Col-0-EV1 plant. Student's t test, *P , 0.05, **P , 0.01. [See online article for color version of this figure.] expression level of IKU2 is critical for the regulation of seed development.

Storage Compounds in phe1, mini3, and iku2 Mutant Seeds
In the phe1, mini3, and iku2 lines, the amount of total fatty acids per seed was reduced by approximately 25% to 40% in mature seeds, while protein content was also slightly reduced. The reduction in oil content became already visible at 11 DAF, i.e. when oil synthesis strongly increased. Therefore, the disturbance in seed development primarily becomes visible at 7 to 11 DAF, when storage compound production begins. In the phe1, mini3, and iku2 lines, a slight increase in 18:3 and a decrease in 18:2 content were observed (Supplemental Fig. S2). This increase in the degree of desaturation by an enhanced conversion of 18:2 to 18:3 has previously been observed in Arabidopsis mutants affected in seed oil deposition. It can be explained by the reduction in the pool size of fatty acids available for desaturases in the mutants resulting in an enhanced conversion to the unsaturated fatty acids (Focks and Benning, 1998;Moreno-Pérez et al., 2012).
Starch degradation was retarded during mutant seed development, and therefore, the starch content remained elevated at 19 DAF compared with the wild type. These results point toward a block in starch degradation in the mutants during the later phase of seed development. Seed starch synthesis and degradation are crucial to sustain high oil accumulation (Vigeolas et al., 2004). A reduction in oil accumulation accompanied with a retardation in starch degradation was already observed in seeds of the wrinkled1 mutant (Focks and Benning, 1998). Baud et al. (2002) showed that only about 10% of the transient starch accumulating during Arabidopsis seed development localizes to the embryo, but that larger amounts of starch are associated with the integuments.
It is possible that the altered endosperm functions in the mini3, phe1, and iku2 mutants affect transient starch degradation during seed development. The reduced oil accumulation in the embryo might in turn be a consequence of the decrease in starch mobilization. Suc in the mature seeds of mini3 and iku2 remains similar to the wild type, but in the mature phe1 seeds, Suc content was increased. In line with this result, the Suc content in developing mini3 and iku2 seeds was similar to the wild type, but it was elevated at 19 DAF in phe1 seeds. The increased Suc content in phe1 seeds can be explained by a decreased sink strength associated with a reduced conversion of Suc into fatty acids as also observed for other seed mutants, e.g. wrinkled1 (Focks and Benning, 1998).

Oil Composition in fie Seeds
All three fie lines produce aborted seeds ( Fig. 5;  Supplemental Fig. S1). Concomitantly, the oil content of fie seeds is strongly reduced below 1% of the wild type ( Fig. 5; Ohad et al., 1996;Luo et al., 2000;Vinkenoog et al., 2000). The fatty acid pattern of aborted fie seeds showed characteristic differences to the wild type, because the amounts of saturated fatty acids (16:0, 18:0) were increased, while the contents of unsaturated fatty acids, particularly 18:3 and 20:1, were decreased. This result was confirmed by the analysis of molecular species of TAG by Q-TOF mass spectrometry. The TAG fatty acid composition of fie-11 resembles that of heart stage wild-type seeds, which contain high amounts of 16:0 and 18:0 and low amounts of 18:3 and 20:1 compared with mature seeds (Baud et al., 2002).

Seed-Specific Overexpression of FIE and PHE Has No Effect on Seed Development
To study the impact of the increased expression level on seed size and storage compound accumulation, the FIE and PHE1 genes were overexpressed. We used the promoter of the soybean seed storage protein glycinin1, which is embryo specific (Nielsen et al., 1989;Thomas, 1993). Expression in nonseed organs (leaves, stems, or roots) is very low (Nielsen et al., 1989). The glycinin promoter is active during the mid to late phase of seed development. By contrast, FIE and PHE1 are expressed early (approximately 1-4 d post pollination) during seed development, and their expression pattern is restricted to different tissues of the seed. FIE expression was detected in the prepollination embryo sac and in embryo and endosperm around 4 d after pollination (Luo et al., 2000). The fact that, in FIE cosuppression plants, plant development and organ establishment was severely affected indicates that FIE also exerts its function throughout sporophyte development (Katz et al., 2004). Therefore, FIE must also be expressed in other plant organs, at least to a low extent. The PHE1 gene is expressed in the preglobular embryo and in the endosperm during early development. Later, expression is restricted to the chalazal endosperm (Köhler et al., 2003). This difference in expression has to be taken into account for the interpretation of the results obtained after overexpression of the respective genes. Vinkenoog et al. (2000) showed that lethality of seeds carrying a maternal fie allele could be rescued by pollination with pollen from hypomethylated plants. These results are in line with the finding that overexpression of FIE or PHE1 per se has no strong effect on seed development.

Seed-Specific Overexpression of IKU2 and MINI3
The MINI3 gene is expressed in pollen and in the ovules during the first 4 d after pollination. MINI3 expression was detected in the endosperm and embryo until the globular stage, but was absent from heart stage embryos (Luo et al., 2005). IKU2 expression was found in young ovules after fertilization and in the endosperm of developing seeds, but it was not expressed in the embryo or pollen (Luo et al., 2005). Therefore, the temporal and spatial gene expression pattern in the overexpression lines under control of the glycinin promoter differs from the expression pattern of the authentic genes. IKU2-OE plants produced seeds with an increase in size, weight, and oil content. The increase in oil became visible as early as 7 to 11 DAF (Fig. 7). At this time, oil accumulation begins in the embryo. It is tempting to speculate that the elevated starch content at 7 DAF, accompanied with an increased starch degradation rate at later time points, is associated with the increase in oil accumulation. In line with this scenario, the Suc content in IKU2-OE seeds is reduced at 7 DAF but steadily increases to reach values similar to the wild type at 19 DAF. The protein content was also slightly higher in developing IKU2-OE seeds until 19 DAF but was similar to the wild type in mature seeds (Figs. 6 and 7). This result indicates that seed developing has not finished at 19 DAF and that protein accumulation in IKU2-OE and wild-type seeds reaches similar values in mature seeds.
The overexpression of IKU2 resulted in a decrease in the total amount of seeds per plant. Therefore, the increase in carbon deposition per seed is associated with a seed yield penalty. As the number of seeds per silique remained unchanged, the number of siliques per plant must be reduced. Furthermore, the fact that the total weight of seeds per plant is reduced indicates that in IKU2-OE plants, less carbon is deposited in the seeds compared with the control (Col-0-EV). Therefore, the reduction in total seed weight cannot be caused by a general limitation in photosynthetic capacity. However, it is possible that the expression of IKU2 under the control of the strong glycinin promoter has negative effects on overall plant development, e.g. flower initiation, which would explain the reduced number of siliques per plant. In this regard, it is interesting to note that overexpression of LEAFY COTYLEDON1 (LEC1) and LEC1-LIKE in Arabidopsis and rapeseed (Brassica napus) under control of the strong seed-specific napin promoter also led to an increase in seed size, but this was accompanied with an abnormal development of the seedlings (Tan et al., 2011). Using truncated versions of the napin promoter, it was possible to separate these two effects, i.e. to obtain transgenic LEC1overexpressing rapeseed plants that produce larger seeds without any detrimental effects on plant development. Thus, it is possible that the use of truncated versions of the glycinin promoter can result in the generation of transgenic IKU2-OE plants, which produce larger seeds, but without reduction of the total seed yield per plant.
Overexpression of IKU2 resulted in an increase in seed size of heterozygous IKU2-OE seeds, and the nontransgenic segregant seeds in siliques of a heterozygous IKU2-OE plant were also larger than those of an empty vector control (Fig. 9, C and D). This finding points to a maternal effect of IKU2 overexpression in heterozygous siliques on the nontransgenic progeny. At present, the basis for this effect remains unknown, but it is possible that a small increase in the expression of IKU2 in maternal tissue, e.g. seed coat or siliques, caused by side activity of the glycinin promoter has a positive effect on the development of the seeds.
Furthermore, a strong cosuppression effect was observed in homozygous IKU2-OE plants, because the IKU2 transcript abundance was reduced to control (Col-0-EV) levels, whereas it was strongly increased in heterozygous IKU2-OE seeds. In line with this result, the increase in seed size, weight, and oil content was only observed in transgenic seeds of a heterozygous plant, but not in homozygous seeds. In Figures 5 and 9, the size, weight, and oil content of fluorescent (transgenic seeds) of a heterozygous IKU2-OE line were measured. These plants produce heterozygous and homozygous IKU2-OE seeds in a predicted ratio of 2:1. The increase in seed size, weight, and oil for a collection of heterozygous IKU2-OE seeds free of homozygous seeds would be expected to be even higher. Therefore, the seed size is correlated with the abundance of IKU2 transcript during seed development. Interestingly, expression of the authentic IKU2 mRNA is restricted to the endosperm, while in the transgenic IKU2-OE lines, expression was redirected to the embryo. This result indicates that IKU2 activity for seed development can also be exerted in the embryo.
Overexpression of the SHB1 gene was found to result in an increase in seed size, weight, and oil content presumably by an increased expression of MINI3 or IKU2 (Zhou et al., 2009). In line with these data, our results show that overexpression of IKU2 resulted in an increase in seed size. IKU2 encodes a LRR kinase involved in endosperm proliferation. LRR kinase-like proteins are involved in a large number of regulative processes, in particular in hormone functions (brassinosteroids, abscisic acid) and in the regulation of organ shape, reproductive cell shapes, and plant pathogen interactions. Our results demonstrate that increased expression of IKU2 in embryos exerts a positive effect on seed development.

Plant Material
Plants were grown in the phytochamber at 55% relative humidity and 22°C with 16 h of fluorescent light per day (150 mmol s -1 m -2 ). The origin of the different Arabidopsis (Arabidopsis thaliana) mutants and control lines is indicated in Supplemental Table S1. The lines Col-0, Landsberg erecta, and C24 and a Feodoroff transposon line (Nottingham Arabidopsis Stock Centre [NASC] stock N8511, ecotype Nossen) were used as controls. Homozygous mutant plants were selected by PCR using oligonucleotides specific to the flanking genomic sequence or to the insertion (Supplemental Table S2).

Overexpression of Seed Development Genes in Arabidopsis Seeds
The expressed sequence tags PYAT1G65330, U60089, and DQ446362 carrying the full-length complementary DNAs (cDNAs) for PHE1, FIE, and MINI3, respectively, were obtained from the Arabidopsis Biological Resource Center. The coding sequence of PHE1 was amplified from PYAT1G65330 by PCR using the oligonucleotides Bn155/Bn156, adding EcoRI/MluI and XhoI sites at the 59 and 39 end of the PCR product, respectively. The PHE1 PCR fragment was cloned into pGEM-T Easy and then ligated into the EcoRI, XhoI sites of the binary vector pBinGlyRed1 (Edgar Cahoon) for expression under control of the strong embryo-specific glycinin promoter. This vector harbors a DsRed marker for selection of transgenic seeds under the fluorescence microscope. This pBinGlyRed1-PHE1 construct contains an extra MluI site derived from the 59 primer Bn155 that was used for the cloning of the other constructs (see below).
The coding region of FIE was amplified from the EST U60089 using the primers Bn157/Bn158 (Supplemental Table S3), subcloned into pGEM-T Easy, and then ligated into the MluI, XhoI sites of pBinGlyRed1-PHE1. Similarly, the MINI3 coding region was amplified from DQ446362 with the primers Bn159/ Bn160, cloned into pGEM-T Easy, and ligated into the MluI, XhoI sites of pBinGlyRed1-PHE1.
No full-length cDNA for IKU2 was available in the stock centers. Therefore, the 39 part (726 bp) of the coding sequence was amplified by PCR from the EST clone X3G19700EFK (Arabidopsis Biological Resource Center) using the primers Bn553/Bn551, thereby introducing an XhoI site at the 39 end, and the PCR product cloned into pGEM-T Easy. The 59 part of the coding sequence (2,248 bp) was amplified from genomic DNA using the primers Bn550/Bn669, introducing an MluI site at the 59 end, and the fragment cloned into pGEM-T Easy. This PCR product encompasses part of exon 1 of the IKU2 gene and therefore contains no introns. The two PCR fragments contain a partially overlapping sequence with an authentic PstI site. Finally, the 59 PCR fragment (released with MluI, PstI from pGEM-T Easy) and the 39 PCR fragment (released with PstI, XhoI) were ligated into the MluI, XhoI sites of pBinGlyRed1-PHE1 in one step.
Binary constructs were transferred into Agrobacterium tumefaciens and used for Arabidopsis Col-0 infiltration using floral dip (Clough and Bent, 1998). Transformed seeds were selected by their red fluorescence under green light using a fluorescent microscope (Olympus SZX 16).

Determination of Seed Weight and Seed Size
Seeds were dried in the vacuum for 48 h prior to the determination of the 100 seed weight. For size determination, approximately 100 Arabidopsis seeds were spread on a microscope slide and fixed with transparent tape. The slide and a scale bar were scanned with high resolution and saved as a TIFF file, which was subsequently exported into bitmap format. The measurement of seed length and width was performed using the Evaluator software (developed by Dmitry Peschansky, IPK Gatersleben) according to the software instructions.

Fatty Acid and TAG Measurements
Total fatty acid content and fatty acid composition in seeds were measured by converting the lipids into fatty acid methyl esters in 1 N methanolic HCl for 2 h at 80°C. Measurements are based on three replicas of five seeds each. Fatty acid methyl esters were quantified by GC in the presence of internal standard (pentadecanoic acid, 15:0) as described (Browse et al., 1986). TAG molecular species were determined by Q-TOF mass spectrometry (Lippold et al., 2012).

Sugar, Starch, and Protein Measurements
Soluble sugars were extracted from the seeds with ethanol as described (Focks and Benning, 1998). For sugar quantification, the extract was added to reaction buffer containing Glc-6-P dehydrogenase. Glc and Suc were measured after addition of hexokinase, phosphoglucoisomerase, and fructosidase (invertase), respectively. The production of NADPH was followed in the photometer at a wavelength of 340 nm (Stitt et al., 1989). Starch was measured in the ethanol-insoluble pellet after hydrolysis with KOH using a starch quantification kit (Roche/R-Biopharm). Total seed proteins were extracted as described by Focks and Benning (1998). Protein was quantified photometrically with the bicinchoninic acid method (Bio-Rad).

RT-PCR
RNA was isolated from green siliques as described Kay et al. (1987). Firststrand cDNA was synthesized from total RNA using the SuperScript III kit (Invitrogen). Transcripts were amplified by RT-PCR and separated in agarose gels, and the bands were visualized with ethidium bromide. A RT-PCR reaction with oligonucleotides to ubiquitin were used as control. Primers and expected RT-PCR product sizes are given in Supplemental Table S4.

Light Microscopy
Green seeds were removed from siliques and placed into fixative solution (4% [w/v] paraformaldehyde, 0.2% [w/v] glutaraldehyde in 0.1 M sodium phosphate buffer, pH 8.0). The samples were incubated in the vacuum for 1 h and then subjected to dehydration using a series of buffers with increasing content of ethanol (from 30%-100% [v/v]). Specimens were preinfiltrated with 100% (v/v) ethanol/Technovit 700 solution (1:1) and then incubated in solution A (1 g of Hardener added to 100 mL Technovit 7100) for 24 h prior to transfer to embedding solution (solution A/Hardener II, 15:1) for polymerization (Heraeus Kulzer). Thin sections (3-4 mm) were prepared with a microtome and stained with 0.05% toluidine blue followed by four to five washes with water and then dried for 10 min at 42°C. The embryo structure was observed using a light microscope (Olympus BH-2).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Morphological phenotype of Arabidopsis seeds affected in seed development.
Supplemental Figure S2. Fatty acid composition in Arabidopsis seeds of mutants affected in seed development.
Supplemental Table S1. Mutant lines used in this study.
Supplemental Table S2. Oligonucleotides for PCR selection of Arabidopsis T-DNA and transposon mutants.
Supplemental Table S3. Oligonucleotides used for generating overexpression constructs.
Supplemental Table S4. Oligonucleotides used for RT-PCR.