The TRAF Mediated Gametogenesis Progression (TRAMGaP) Gene Is Required for Megaspore Mother Cell Specification and Gametophyte Development

TRAMGaP gene coordinates the expression of diverse sets of genes to modulate MMC specification and gametophyte development in Arabidopsis.

Abstract

In plants, the role of TRAF-like proteins with meprin and the TRAF homology (MATH) domain is far from clear. In animals, these proteins serve as adapter molecules to mediate signal transduction from Tumor Necrosis Factor Receptor to downstream effector molecules. A seed-sterile mutant with a disrupted TRAF-like gene (At5g26290) exhibiting aberrant gametogenesis led us to investigate the developmental role of this gene in Arabidopsis (Arabidopsis thaliana). The mutation was semidominant and resulted in pleiotropic phenotypes with such features as short siliques with fewer ovules, pollen and seed sterility, altered Megaspore Mother Cell (MMC) specification, and delayed programmed cell death in megaspores and the tapetum, features that overlapped those in other well-characterized mutants. Seed sterility and reduced transmission frequency of the mutant alleles pointed to a dual role, sporophytic and gametophytic, for the gene on the male side. The mutant also showed altered expression of various genes involved in such cellular and developmental pathways as regulation of transcription, biosynthesis and transport of lipids, hormone-mediated signaling, and gametophyte development. The diverse phenotypes of the mutant and the altered expression of key genes related to gametophyte and seed development could be explained based on the functional similarly between At5g26290 and MATH-BTB domain proteins that modulate gene expression through the ubiquitin-mediated proteasome system. These results show a novel link between a TRAF-like gene and reproductive development in plants.

The life cycle of higher plants comprises a short, haploid, gametophyte phase and a long, diploid, multicellular, sporophyte phase. The transition to the gametophyte phase is initiated when a diploid spore mother cell differentiates and undergoes meiosis to give rise to four haploid megaspores. In the ovule, of the four spores, the three that are closest to the micropylar end of the ovule degenerate while the fourth, closest to the chalazal end, becomes a functional megaspore and differentiates into a megagametophyte after typically three cycles of free nuclear mitotic divisions followed by highly polarized cellularization (Yang et al., 2010; Sprunck and Gross-Hardt, 2011). The resulting embryo sac is the mature female gametophyte (FG). A typical FG comprises a seven-celled embryo sac containing six haploid cells (an egg cell, two synergids, and three antipodals) with one diploid central cell. Likewise, following meiosis, the male microspore undergoes an asymmetric cell division resulting in a male gametophyte comprising one vegetative cell and two sperm cells (Twell, 2011).

The haploid gametes develop in the midst of diploid maternal tissue, which requires constant communication between the two types of cells. After fertilization, intercommunications among the embryo, the endosperm, and the seed coat, each with a unique genetic identity, is essential for proper seed development (Figueiredo and Köhler, 2016). During the alternation of generations, cells have to undergo two transition phases: the first transition occurs during the specification of gametophytic cells from sporophytic cells (archesporial cell formation), which is designated as a mitotic-to-meiotic transition, and the second occurs during the development of the gametophyte, which is designated as a meiotic-to-mitotic transition (Van Durme and Nowack, 2016). A number of mutations affecting various stages of male gametophyte or FG development have been identified and reviewed (Liu and Qu, 2008), and the role of phytohormones and lipid molecules in controlling these developmental stages through complex signaling has been documented (Nakamura, 2015; Schuster et al., 2015). Cloning and characterization of such genes (Liu and Qu, 2008; Drews and Koltunow, 2011) have provided valuable insights into the molecular events underlying these critical stages in the life cycle of a plant. However, the molecular mechanisms that coordinate the germline, zygote, and seed development in plants are as yet only poorly understood (Juranić et al., 2012).

Convincing evidence now available with animal cells suggests that the transition and progression of the cell cycle and cell fate determination are temporally and spatially controlled by targeting key regulators through the ubiquitin/26S proteasome pathway (McCarthy Campbell et al., 2009). Studies on mammalian cells and Caenorhabditis elegans have revealed that multisubunit Cullin3 (CUL-3)-RING E3 ligases (CRL) regulate different cellular and signaling processes during the meiosis-to-mitosis transition (Bowerman and Kurz, 2006; Sawin and Tran, 2006; Sumara et al., 2008). Rapid degradation of meiotic proteins is a prerequisite for the switch to mitotic division (the oocyte-to-embryo transition) in nematodes. The earliest mitotic spindle differs from its meiotic counterpart in that the latter is labeled with meiosis-specific proteins that are earmarked for degradation so as to avoid any interference from the meiotic spindle during subsequent mitotic divisions (Lu and Mains, 2007). A female germline-specific CUL-3 substrate adaptor, namely the Maternal Effect Lethal26 (MEL-26) consisting of a MATH (Meprin-Associated Traf Homology) domain and a Broad-complex, Tramtrack, Bric-a-brac (BTB) domain, is responsible for the spatial and temporal targeting and degradation of MEI-1/katanin (Defective in Meiosis1) during the meiosis-to-mitosis transition to allow the cells to enter the mitotic stage. Similarly, MEL-26 targets the microtubule-interacting protein Fidgetin-Like1 for degradation during mitosis (Luke-Glaser et al., 2007). Weber et al. (2005) showed that a MATH-BTB protein functions in a similar fashion in Arabidopsis (Arabidopsis thaliana), and Juranić et al. (2012) characterized similar proteins in maize (Zea mays).

Cosson et al. (2010) identified 71 proteins in Arabidopsis that contain the MATH domain and classified them into four families. Of these, two belong to the ubiquitin-specific protease 7 family, six to the MATHd/BTB family, and one to the MATHd/filament protein family. The remaining 62 proteins, which possess up to four MATH domains without any other associated domains, belong to the MATHd-only protein family and are classified as Tumor Necrosis Factor Receptor (TNF-R) Associated Factors like (TRAF-like) genes. TRAF proteins are widely found in metazoans and serve as adapter proteins that help in the transmission of external signals received by the TNF-R to downstream effector molecules. TRAFs are E-3 ubiquitin ligase proteins that mediate the interactions between TRAF members and receptors and also those between TRAF members and several intracellular signaling molecules (Zapata, 2003; Alvarez et al., 2010) by virtue of their MATH domain. In animals, TRAF proteins regulate diverse cell processes, including immune response, inflammatory response, apoptosis, cancer, embryogenesis, and the survival of the cell itself (Kedinger et al., 2005). In plants, TRAF-like genes have been reported to play a role in pathogenesis (RTM3; Cosson et al., 2010), in abscisic acid-mediated drought stress signaling (SINA2; Bao et al., 2014), and in insect herbivory (At5g26260 and At3g28220; Schweizer et al., 2013). However, no role has been reported so far in reproductive development; to our knowledge, this is the first such report. We found that a mutation in the TRAF-like gene (At5g26290) affected normal reproductive development in Arabidopsis. The At5g26290 gene is expressed specifically during the development of the male and female germline in gametophytic and the surrounding sporophytic tissue. Detailed expression profiling of this gene revealed its role in modulating key genes at various stages of gametogenesis and during the progression of embryo sac differentiation. Therefore, we have named the At5g26290 gene as the TRAF Mediated Gametogenesis Progression (TRAMGaP) gene.

RESULTS

Mutation in TRAMGaP Causes Seed Sterility

While screening a T-DNA promoter trap population of Arabidopsis generated in house, we noticed a mutant with short siliques and high (∼50%) seed sterility (Fig. 1; Table I). Using the genome-walking approach, the T-DNA insertion site was identified in the mutant (Supplemental Fig. S1A). A BLAST search of the sequence amplified in the genome walk against the Arabidopsis genome database showed a complete match with the last exon of the At5g26290 gene between the coordinates 9,227,747 and 9,228,082 of the fifth chromosome of Arabidopsis (Fig. 1E; Supplemental Fig. S1B), indicating that insertion of T-DNA had disrupted the At5g26290 gene. This gene is known to code for a TRAF-like protein. The site of insertion of the T-DNA was further confirmed by PCR using forward and reverse primers specific to the T-DNA flanking region (Supplemental Fig. S1C). In kanamycin-positive T2 progeny of this mutant, two amplicons were detected: a 1.8-kb amplicon corresponding to the wild-type allele and an ∼7-kb amplicon coming from the mutant allele (Supplemental Fig. S1C). Sequencing of the ∼7-kb amplicon revealed the presence of two tandem copies of T-DNA. Thus, PCR not only confirmed the insertion of T-DNA into the At5g26290 locus but also helped to separate plants that were homozygous for the mutant from those that were heterozygous for it.

Figure 1.

TRAMGaP mutant phenotypes and location of the T-DNA insertion in the mutants. A, Dissected silique from a wild-type plant showing well-filled, viable seeds. B, Homozygous tramgap1-1 silique at the stage comparable to that in the wild type showing normal and aborted ovules (asterisks). C, Homozygous tramgap1-2 silique at the stage comparable to that in the wild type bearing normal and aborted ovules (asterisks). D, Silique from a genetically complemented homozygous tramgap1-1 mutant plant showing full seed set. E, T-DNA insertion in tramgap1-1 and tramgap1-2 mutant lines. Bars = 500 µm.

Table I. Silique characteristics of tramgap mutant lines of Arabidopsis.

Genotype	Mean Silique Length	Mean No. of Ovules in Each Silique
	mm
TRAMGaP/TRAMGaP (wild type)	21.34 ± 0.2	45.0 ± 1.4
TRAMGaP/tramgap1-1	9.99 ± 0.2a	30.7 ± 1.7a
tramgap1-1/tramgap1-1	7.47 ± 0.3a	27.7 ± 1.1a
tramgap1-2/tramgap1-2	6.98 ± 0.4a	23.8 ± 1.3a

^aMean values differ from the wild type at P = 0.01.

For more detailed characterization of the At5g26290 gene, another mutant line (SALK_0146328; Fig. 1C) with the site of T-DNA insertion in the fourth intron was procured from the Arabidopsis Biological Resource Center and the insertion site was confirmed as above. Of the two allelic variants, the mutant developed in house was designated as tramgap1-1 and the SALK mutant was designated as tramgap1-2 (Fig. 1E). Unless stated otherwise, all statements in this article pertaining to the mutants refer to homozygous mutant plants.

A gene complementation study was undertaken to confirm that the failure to set seed in tramgap1-1 was due to loss of the TRAMGaP gene. Plants with tramgap1-1 were transformed with the gene cassette 35S::TRAMGaP. The transgenic plants selected on hygromycin set seeds normally, confirming the functional complementation of the mutant phenotype by the transgene (Fig. 1D).

At5g26290 Is a Member of the TRAF-Like Gene Family

At5g26290 (TRAMGaP) is 1.75 kb long, contains seven exons, and is capable of coding for a polypeptide consisting of 322 amino acids (Supplemental Fig. S2). Phylogenetic analysis of TRAF proteins of both the Animalia and Plantae kingdoms, including human, mouse, Drosophila spp., Populus spp., rice (Oryza sativa), and Arabidopsis, using MEGA6 software (Tamura et al., 2013) showed that the sequences were highly conserved over long evolutionary periods (Supplemental Fig. S3). However, proteins from the two kingdoms fell into separate groups, indicating a clear divergence between the two. In the phylogenetic tree, five closely related Arabidopsis TRAFs fall into a single clade (At5g26260, At5g26280, At5g26290, At5g26300, and At5g26320). Cosson et al. (2010) grouped all 69 MATH domain-containing proteins of Arabidopsis into four families. Of these, 62 proteins, including At5g26290, have been assigned to the MATHd-only protein family, which contains up to four MATH domains without any other associated domains. Most of the genes of this family are organized into clusters (Cosson et al., 2010). At5g26290 belongs to the third largest cluster with four other genes arranged in tandem (At5g26260, At5g26280, At5g26300, and At5g26320; Supplemental Fig. S4), is 64% to 71% identical to At5g26290 at the level of amino acids (Supplemental Fig. S5), and grouped separately from all other members of the family (Supplemental Fig. S4). Thus, homology analysis clearly showed that At5g26290 is a TRAF-like protein.

Going by TAIR annotation, At5g26290 is denoted as a MATH domain-containing TRAF-like protein. Structurally, TRAF proteins share a conserved region of about 180 residues with meprins. At5g26290 contains two MATH domains, and the Simple Modular Architecture Research Tool (http://smart.embl-heidelberg.de) predicted two MATH domains at amino acid positions 57 to 162 and 203 to 303 of the polypeptide encoded by At5g26290 (Supplemental Fig. S2). The MATH domains contain eight β-strands, as shown in the secondary structure prediction analysis (Combet et al., 2000), that also are found in At5g26290 (Supplemental Fig. S6). Quaternary structure analysis of TRAMGaP also revealed that one of the MATH domains was nearly identical (99.86%) to a TRAF-like protein, whereas the other MATH domain was nearly identical (99.94%) to a speckle-type POZ protein (SPOP; Supplemental Fig. S7). The alignment of At5g26290 MATH domains with TRAFs reported from human samples and of RTM3, a MATH protein reported in Arabidopsis by Cosson et al. (2010), showed conservation at the level of amino acids for all eight TRAF-2 β-strands (Fig. 2). In tramgap1-1, T-DNA insertion disrupted the C-terminal MATH domain, whereas in tramgap1-2, it was the N-terminal MATH domain that was disrupted (Supplemental Fig. S8).

Figure 2.

Amino acid sequence alignment of the MATH domain of TRAMGaP (At5g26290) with TRAF1, TRAF2, TRAF3, and TRAF5 of Human and RTM3 from Arabidopsis. The eight β-strands identified in TRAF proteins are indicated by horizontal arrows. Numbers in parentheses are amino acid positions corresponding to the start and end of each MATH domain. Residues conserved between At5g26290 and at least one TRAF protein are shaded. At5g26290.1 and At5g26290.2 represent the first and second MATH domains of TRAMGaP, respectively.

TRAMGaP Is Expressed Mainly in Seedlings and Reproductive Organs

Reverse transcription-PCR analysis showed TRAMGaP transcripts in roots and aerial parts of 7-d-old seedlings and in the inflorescence (Supplemental Fig. S9A). Quantitative expression analysis using quantitative reverse transcription (qRT)-PCR also revealed the highest levels of expression of this gene in 7-d-old seedlings and in flower buds (stages 5–9). TRAMGaP transcript levels decreased 3- to 7-log fold in various tissues such as stem, leaf, axillary branches, and cauline leaves during different floral developmental stages, namely stages 10 to 12 and 13 to 14, and were undetectable after fertilization (stages 15–18; Supplemental Table S1). These results indicate that TRAMGaP is expressed specifically in seedlings and in floral tissues, results that are in agreement with the expression pattern of TRAMGaP reported in the Genevestigator microarray database (www.genevestigator.ethz.ch; Supplemental Fig. S9B).

To further understand the tissue-specific expression pattern of TRAMGaP, a 1.5-kb fragment upstream of the TRAMGaP coding sequence was amplified and used to assemble a pTRAMGaP::uidA construct. T2 transgenic Arabidopsis plants carrying this gene cassette showed strong GUS expression in both microgametophytes and megagametophytes (Fig. 3). GUS expression in ovules was found in nucellar tissue as well as in all cells of the embryo sac (Fig. 3, A–E), whereas in anthers, GUS expression was localized to microspores or pollen and to tapetal cells (Fig. 3, F–J). Thus, TRAMGaP is expressed mainly in sporophytic and gametophytic tissues of reproductive structures and within flowers, and the expression is limited to the nucellus and the tapetum.

Figure 3.

GUS expression in ovules (A–E) and anthers (F–J) of a transgenic (T1) Arabidopsis plant carrying the pTRAMGaP::uidA construct. A, GUS expression in the nucellus and MMCs in the premeiotic stage ovule. B, GUS expression in the functional megaspore and surrounding sporophytic tissue in the ovule at stage FG1. C to E, Ovule (stage FG6/7) showing GUS expression in the gametophyte cells (synergids, egg cell, and central cell) at the micropylar end (C), at antipodals (D), and at the chalazal end (E). F, Anther from flower at stage 5 showing GUS expression in pollen mother cells. G, Anther from flower at stage 7 showing GUS expression in tetrads and tapetal cells. H, Anther from flower at stage 9 showing GUS expression in developing microspores and in the tapetum. I, Anther from flower at stage 11 showing GUS expression in mature pollen and tapetum. J, Anther from flower at stage 13 showing GUS expression restricted to mature pollen. AP, Antipodals; CC, central cell; EC, egg cell; S, synergid; T, tapetum.

Mutation in TRAMGaP Causes Seed Sterility, and Mutant Alleles Are Transmitted through Both Male and Female Gametes

Both the allelic variants of the tramgap mutant had shorter siliques and fewer ovules in each silique compared with those of wild-type plants (Table I). Plants heterozygous for the mutation also had significantly fewer (∼31) ovules in each silique and showed ∼50% seed set (Table II). In plants homozygous for the mutation, seed sterility increased to 65% to 70% and the number of ovules in each silique decreased further to 24 to 28 (Table I; Fig. 1). In general, phenotypes of tramgap1-2 plants were affected more severely than those of tramgap1-1. Seed sterility ranging from 34% to 37% also was recorded in the crosses between wild-type plants and the heterozygous (tramgap/+) mutant plants (Table II). Seed sterility in wild-type plants crossed with pollen from heterozygous mutant plants suggests that heterozygosity results in seed abortion. However, the recovery of homozygous mutant plants upon selfing shows that the tramgap mutation is not completely lethal. These results indicate that haploinsufficiency of TRAMGaP in heterozygotes leads to seed sterility. In order to better understand in greater detail the transmission efficiency of the mutant alleles through male and female gametes, progeny of the reciprocal crosses between the wild type and tramgap1-1/+ were screened for kanamycin resistance, which is linked to the mutant allele. Nearly one-half of the progeny of tramgap1-1/+ (♀) × wild type (♂) was kanamycin positive (Table III), indicating very high female transmission efficiency of the mutant allele. On the other hand, the male transmission efficiency was very low (16.7%). Nearly normal female transmission of the mutant allele and 37% seed sterility in the cross between heterozygous mutants and the wild type (with pollen from the wild type) suggest that the mutation affects wild-type female gametes in heterozygous mutant plants through some sporophytic effect. In contrast, low male transmission frequency and comparable seed sterility in reciprocal crosses indicate both sporophytic and gametophytic effects of the mutation on the male side.

Table II. Seed sterility in tramgap mutants from selfed plants and from various crosses.

Cross (Female × Male)	Normal Seeds	Aborted Ovules or Seeds	Sterility
			%
Wild type (selfed)	845	11	1.3 ± 0.22
tramgap1-1/+ (selfed)	235	226	49.0 ± 1.3
tramgap1-1/tramgap1-1 (selfed)	182	346	65.5 ± 0.90
Wild type × tramgap1-1/+	255	133	34.2 ± 1.47
tramgap1-1/+ × wild type	191	114	37.3 ± 2.2
tramgap1-2/tramgap1-2 (selfed)	86	200	69.9 ± 1.47
Wild type × tramgap1-2/tramgap1-2	49	97	66.4 ± 4.5
tramgap1-2/ tramgap1-2 × wild type	120	122	50.4 ± 3.3
tramgap1-1/TRAMGaP (transgene complemented)	425	40	8.6 ± 0.8

Table III. Transmission efficiency of the tramgap1-1 allele.

Cross (Female × Male)	Total Seedlings Tested	KanR	KanS	χ2 (1:1)	tramgap1-1	TEa
					%	%
tramgap1-1/+ × wild type	59	27	32	0.42	46	84.4
Wild type × tramgap1-1/+	97	14	83	49.08b	15	16.7

^aTransmission efficiency (TE) = kanamycin resistant (Kan^R)/kanamycin susceptible (Kan^S) × 100.

^bSignificant at P < 0.01.

tramgap Mutants Display a Wide Range of Abnormalities during Gametophyte Development by Hindering the Specification of MMC and the Progression of Gametogenesis

To understand the role of TRAMGaP in gametophyte development, we compared the development of the ovule and of the anther in the wild type with that in the two mutants. In the wild type, most of the premeiotic ovules showed only one MMC (Fig. 4A), and although a few (5.8%) showed more than one MMC (Fig. 4B), only one of them differentiated into a functional embryo sac. In contrast, in tramgap mutants, 40% of the ovule primordia showed several abnormally enlarged subepidermal cells at the premeiotic stage (Fig. 4, C–E). Ovule development at anthesis also was examined to ascertain the synchronization of ovule maturation with anthesis. Surprisingly, only 47.2% of ovules from the tramgap1-1/+ plants, 34.7% from tramgap1-1/tramgap1-1 plants, and 29% from tramgap1-2/tramgap1-2 plants had attained maturity (stage FG7/8), whereas in wild-type plants, the corresponding value was as high as 86.3% (Table IV; Fig. 4F). Approximately 31.3% of ovules were at the FG1 stage, displaying a teardrop-like cell (the functional megaspore) accompanied by three nuclei (Fig. 4J), a typical FG1 stage phenotype as described by Sundaresan and Alandete-Saez (2010). In some ovules from tramgap1-1 and tramgap1-2, four nuclei arranged in a row were visible at the micropylar end even at the FG4 stage (Fig. 4, H and I; Supplemental Fig. S10, C and D). Such persistence of all four meiotic products indicates impaired developmental programmed cell death (dPCD) during ovule development. In about 8% of ovules, a degenerating embryo sac was observed, and about 44.8% of ovules lagged behind the FG7/8 stage at the time of anthesis in the tramgap1-1/+ plants (Table IV): in tramgap1-1/tramgap1-1 and tramgap1-2/tramgap1-2 plants, such ovules accounted for 60.7% and 63.3% of the total, respectively. Also, the differentiation of the chalazal megaspore into a functional megaspore was affected: instead of a typical seven-celled wild-type embryo sac (Fig. 4F), we found multinucleate megagametophytes with irregularly distributed nuclei (Fig. 4G). In tramgap1-1, embryo sacs with a variable number of nuclei (two to six nuclei) at the center also were observed, suggesting failure in the progression of mitosis after the commitment of the functional megaspore (Fig. 4G). The arrest of mitosis occurred at any of the three mitotic divisions (Supplemental Fig. S10, A and B). Postfertilization defects also were evident in the mutant: some developing seeds contained only the endosperm (Supplemental Fig. S10F) or only the embryo (Supplemental Fig. S10G), whereas others had defective embryo and endosperm (Supplemental Fig. S10H). Such abnormalities could result from defects related to fertilization by defective pollen or of defective ovules.

Figure 4.

Female gametogenesis in the tramgap mutant. A, Wild-type (WT) ovule primordium showing a proper single MMC at the premeiotic stage. B, Wild-type ovule primordium showing two MMCs at the premeiotic stage. C, Homozygous tramgap1-1 ovule primordium showing two MMCs at the premeiotic stage. D, Homozygous tramgap1-1 ovule primordium showing six MMCs at the premeiotic stage (white arrows). In one of them, nuclear division has been initiated (red arrow). E, Homozygous tramgap1-2 ovule primordium showing multiple abnormal MMCs at the premeiotic stage (white arrows). F, Wild-type ovule at anthesis (FG7/8) showing a fully differentiated proper embryo sac. G, Heterozygous tramgap1-1 ovule showing a multinucleate megagametophyte with irregularly distributed nuclei (asterisks) at the center of the abnormal embryo sac. H, Homozygous tramgap1-1 ovule showing four linearly arranged doughnut-shaped (arrows) nuclei at the micropylar end of the embryo sac. I, Homozygous tramgap1-2 ovule with four nuclei in a rectangular cluster in the micropylar region. J, Homozygous tramgap1-1 ovule arrested at stage FG1 showing a pear-shaped functional megaspore. K, Expression of a synergid cell marker (ET-884) in the wild type. L, Absence of expression of the synergid cell marker in the nuclei of the tramgap1-1 mutant embryo sac. M, Normal ovule from a heterozygous mutant plant showing expression of a synergid cell marker. N, Differential interference contrast (DIC) image of a wild-type ovule carrying a central cell marker (DD65). O, Fluorescence image of the ovule in N showing expression of a central cell marker (DD65). P, DIC image of a tramgap1-1 ovule carrying a central cell marker (DD65). Q, Fluorescence image of the ovule in P lacking expression of the central cell marker (DD65). Defective cells/nuclei are marked with asterisks and arrows. CC, Central cell; EC, egg cell; ES, endosperm; S, synergid cell.

Table IV. Frequency of FGs at different stages of development at anthesis in wild-type, tramgap1-1, and tramgap1-2 mutant lines.

Values shown are percentages. DE, Degenerating; N, total number of ovules examined.

Genotype	Stage of FG Development
	FG1	FG2	FG4	FG5/6	FG7/8	DE	N
+/+ (wild type)	0.0	0.0	5.7	6.3	86.3	1.7	175
tramgap1-1/+	31.3	3.6	9.9	0.0	47.2	8.0	336
tramgap1-1/tramgap1-1	50.3	3.4	7.0	0.0	34.7	4.6	328
tramgap1-2/tramgap1-2	46.4	0.8	13.7	2.4	29.03	7.7	243

To ascertain cell specifications in defective embryo sacs, tramgap1-1 homozygous mutants were crossed with lines expressing different cell-specific markers, and ovules from the F1 plants were examined for the expression of the marker genes. The marker specific to synergid cells (ET884) showed expression in two micropylar cells of the embryo sac of wild-type ovules and in normal ovules of the mutant siliques (Fig. 4, K and M), but no GUS expression was observed in the embryo sac of defective ovules (Fig. 4L). Likewise, the marker specific to the central cell (DD65) also failed to show expression in defective ovules (Fig. 4, P and Q). These results indicate that the nuclei observed in the micropylar region of the embryo sacs from defective ovules are either haploid megaspores, the development of which is arrested after meiosis because of the loss of dPCD, or mitotic products of megagametophytes, the development of which is arrested after two mitotic divisions. Thus, the tramgap mutation affects gametophyte development by hindering the progression of mitosis and cell specification in functional megaspores.

The tramgap mutants showed reduced pollen viability and low-frequency transmission of the mutant allele through the male line. Therefore, we investigated pollen development in wild-type and mutant plants. The microspore mother cells underwent meiosis, giving rise to four haploid cells in wild-type and mutant plants (Fig. 5). However, morphological defects became visible at the microspore stage corresponding to flower development stage 9. At this stage, compared with the wild type, the microspore wall was not well formed in the tramgap mutants (Fig. 5, A–C), and degeneration of microspores became evident at flower development stage 11 (Fig. 5, D–F). In the wild type, the tapetum underwent dPCD and pollen grains were fully developed, with sporopollenin in their outer walls (Fig. 5D). In contrast, anthers in the mutant plants showed a mix of sterile, empty, and degenerating pollen grains and normal mature pollen grains (Fig. 5, E and F). Furthermore, the tapetal cells were not fully degenerated, once again pointing to the loss of dPCD. Anther dehiscence in the mutant was comparable to that in the wild type (Fig. 5, G–I) and occurred around flower development stage 13. No other defects were observed in other cells of the anther, suggesting that the mutant phenotype in anthers is confined to the microspores, the tapetum, or both. Staining mature anthers with Alexander’s stain showed that nearly 100% of pollen grains were viable in the wild type (Fig. 6, A and C), whereas in the mutants, the proportion was only about 30% (Fig. 6, B and D). Scanning electron microscopy showed uniform, elliptical pollen in the wild type (Fig. 6E), whereas defects in the pollen wall were observed in both heterozygous and homozygous tramgap1-1 mutant plants (data not shown). Likewise, a mix of normal and irregularly shaped, shriveled pollen was seen in tramgap1-1 mutants (Fig. 6F). Staining of mature pollen from the tramgap1-1/+ plants with 4′,6-diamidino-phenylindole (DAPI) showed uninucleate, binucleate, and trinucleate pollen as well as dead pollen that lacked nuclei (Fig. 6H). In contrast, only trinucleate pollen was observed in the wild type (Fig. 6G; Table V). Both stains showed that approximately, 38.7% of pollen was apparently normal and trinucleate (Table V). Thus, the defects observed in the embryo and the endosperm in this mutant probably were due to the failure of double fertilization: pollen carrying only a single sperm can fertilize either the central cell or the egg cell but not both. As a result, further development of the seed would be arrested owing to the lack of endosperm or of embryo. In all cases of the failure to set seed, embryos never progressed beyond the globular stage. Thus, even the pollen carrying the wild-type TRAMGaP allele produced on heterozygous mutant plants appear to be somewhat defective, which supports the inference that the mutation also has a sporophytic effect.

Figure 5.

Comparative analysis of anther development in the wild type and homozygous tramgap mutants. Anther development is shown in the wild type (top row), in homozygous tramgap1-1 (middle row), and in homozygous tramgap1-2 (bottom row). A, Cross section of wild-type anther at stage 9 showing proper development of microspores and tapetum. B, Cross section of homozygous tramgap1-1 anther at stage 9 showing defective microspores and tapetum. C, Cross section of homozygous tramgap1-2 anther at stage 9 showing defective microspores and tapetum. D, Wild type anther at stage 11 showing mature pollen and degenerated tapetum. E, Sterile and fertile pollen and partly consumed tapetum in homozygous tramgap1-1. F, Sterile and fertile pollen and partly consumed tapetum in homozygous tramgap1-2. G, Wild-type anther at stage 13 showing dehiscence and release of fertile pollen. H, Fertile and sterile pollen in homozygous tramgap1-1 at stage 13. I, Fertile and sterile pollen in homozygous tramgap1-1 at stage 13. DP, Defective pollen; MP, mature pollen; T, tapetum.

Figure 6.

Analysis of pollen viability, integrity, and structure in the wild type and homozygous tramgap1-1 and tramgap1-2 mutants. A, DIC image of anther from the wild type showing normal pollen. B, DIC image of anther from a homozygous tramgap1 mutant showing defective pollen. C, Wild type anther stained with Alexander’s stain showing fully viable pollen. D, Homozygous tramgap1-1 anther stained with Alexander’s stain showing empty and nonviable pollen (arrows) along with viable pollen. E, Scanning electron micrograph showing normal, elliptical wild-type pollen with proper reticulate exine. F, Scanning electron micrograph showing abnormal, shriveled, irregularly shaped tramgap1 pollen with smooth exine. G, DAPI-stained wild-type pollen showing trinucleate pollen. H, Pollen from a homozygous tramgap1-1 mutant with zero, one, two, or three nuclei. Insets in E and F are closeup views of a single pollen grain showing defective pollen wall. Bars = 20 µm.

Table V. DAPI analysis of pollen of tramgap mutants.

Genotype	No. of Pollen Grains Examined	Percentage of Pollen Showing
		Three Nuclei	Two Nuclei	One Nucleus	No Nucleus
+/+ (wild type)	219	91.7 ± 2.17	5.7 ± 2.1	0.5 ± 0.4	2.0 ± 0.9
+/tramgap1-1	230	38.7 ± 3.7a	30.9 ± 2.7a	5.4 ± 1.9a	24.9 ± 4.1a
tramgap1-1/tramgap1-1	466	29.9 ± 2.6a	16.6 ± 1.9a	3.5 ± 1.1a	49.9 ± 2.8a
tramgap1-2/tramgap1-2	589	34.1 ± 1.9a	35.2 ± 2.2a	10.7 ± 1.5a	20.0 ± 2.06a

^aMeans of samples differ from the wild type at P < 0.01.

Mutation in TRAMGaP Alters the Expression of Its Closely Related Homologs

As shown earlier, in the 69-member TRAF family, At5g26290 is grouped with its adjacent four tandemly arranged members and placed in a separate clade. To test whether these members share any redundancy with TRAMGaP, the expression pattern of its close homologs, namely At5g26260, At5g26280, At5g26300, and At5g26320 (63%–70% homology at the level of amino acids) was studied using reverse transcription-PCR. Transcripts of all four of these genes were less abundant than those of TRAMGaP in wild-type flower buds of stages 5 to 9 but were 3- to 5-log fold higher in flower buds of stages 10 to 12. At the postfertilization stage, siliques of wild-type plants contained transcripts of At5g26260, At5g26280, At5g26300, and At5g26320, but those of TRAMGaP were undetectable (Table VI). qRT-PCR failed to detect the transcripts of TRAMGaP in the tramgap1-2 mutant at any stage of flower development, nor was the expression of the At5g26300 gene detectable in flower buds at stages 5 to 9 and 10 to 14, whereas transcripts of At5g26280 showed more than 2-log fold up-regulation at all the flower stages tested. Likewise, the expression of At5g26320 was up-regulated 3-log fold during flower stages 5 to 9. Thus, At5g26260, At5g26280, At5g26300, and At5g26320 showed overlapping expression with TRAMGaP, especially during flower stages 12 to 14, and loss of At5g26290 activity changed the expression of other closely related homologs, suggesting some cross talk or redundancy among these genes.

Table VI. qRT-PCR analysis of TRAMGaP (At5g26290) and its close paralogs at various stages of flower development in the wild type and the SALK mutant line Numbers indicate log fold differences.

Gene	Wild Type				SALK
	5–9a	10–12	13–14	15–18	5–9	10–12	13–14	15–18
At5g26260	−4.24	−4.74	−5.87	−8.54	−4.46	−4.84	−6.45	−7.08
At5g26280	−6.08	−4.02	−4.26	−9.00	−4.61	−2.11	−2.35	−6.97
At5g26290b	0.00	−7.05	−4.19	No CT	No CT	No CT	No CT	No CT
At5g26300	−8.42	−3.05	−2.54	−4.17	No CT	−3.64	No CT	−3.24
At5g26320	−7.14	−2.63	−0.23	−2.14	−4.17	−3.65	−3.70	−2.69

^aFlower stages 5–9, 10–12, 13–14, and 15–18 correspond to microsporogenesis, megasporogenesis, anthesis, and postfertilization stages, respectively.

^bAt5g26290 (i.e. TRAMGaP) is taken as the calibrator. No CT means RT-PCR amplification did not reach the threshold level

Mutation in TRAMGaP Alters the Expression of Kinases, Transcription Factors, Proteases, and Genes Involved in Lipid Biosynthesis

Considering that TRAFs are adaptor molecules, which assist signal transduction in different biological processes, the pleiotropic effect of the tramgap mutation is not unexpected. Microarray analysis to ascertain the possible involvement of TRAMGaP in various developmental processes showed a total of 3,150 differentially expressed genes in flower buds between tramgap1-2 homozygous plants and the wild type. Of these, 1,707 genes were down-regulated and 1,443 were up-regulated. We found differential regulation of 132 kinases involved in developmental cell growth, protein autophosphorylation, and male gamete generation and 19 genes involved in ubiquitination and 18 F-box proteins in the tramgap1-2 mutant (Supplemental Table S2). A number of transcription factors also were found to be differentially expressed. These included those that contained the AP2 domain (10), those of the bHLH family (nine), zinc finger (55), CONSTANS-LIKE1 (four), MYB (16), those that contained the NAC domain (12), Arabidopsis Response Regulators (eight), those of the PHD family (four), SCARECROW-LIKE (three), AGAMOUS-LIKE (four), BTB (three), bZIP (three), and 45 others. In addition, down-regulation was observed in 58 genes involved in lipid biosynthesis and in 37 genes of the phenylpropanoid pathway. Furthermore, seven genes related to the metabolism of very-long-chain fatty acids, 15 genes related to lipid transfer proteins, six genes related to elongase β-ketoacyl-CoA synthase and fatty acid desaturases, and one, namely CYP86A1 (which catalyzes the ω-hydroxylation of fatty acids involved in lipid biosynthesis), also were found to be differentially expressed in the homozygous tramgap1-2 mutants, along with a number of tapetal oleosin and APG-like genes.

The protein-protein interaction study confirmed that all of these genes form a major cluster (comprising 570 genes), indicating their probable function in different molecular and biological processes leading to reproductive success. Ontological classification clearly indicated that the differentially expressed genes are involved in regulating transcription, lipid biosynthesis, transport, hormone-mediated signaling, and gametophyte development (Supplemental Fig. S11). Phenotypic as well as microarray analyses suggested that loss of TRAMGaP function affects some key genes related to the transition phase of plant growth and development.

For further confirmation of differential expression, genes such as katanin, AGOs, DICER-like, and SGS, which are crucial for sporophyte-to-gametophyte transition signaling, were examined using qRT-PCR. The expression of several key genes related to hormonal signaling and of some well-characterized genes playing an important role in gametogenesis and in lipid biosynthesis also was examined in homozygous tramgap1-2 plants using qRT-PCR (Fig. 7). A total of 56 genes were analyzed to validate and extend the microarray results (Supplemental Table S3). The expression levels of these genes ranged from 2-fold up-regulation to 36-fold down-regulation at floral stages 5 to 9, whereas at later stages, the expression ranged from 4-fold up-regulation to 9-fold down-regulation. This is consistent with the fact that TRAMGaP expression peaks at flower stages 5 to 9 (Table VI), and the consequences of its loss would be most evident around these stages. The changes observed at later stages might exert cascading downstream effects. The results of qRT-PCR were in agreement with those from the microarray for most of the genes. For example, genes CS5, ACS5, MS2, DICER2, AGO9, BAM1, AN3, ERD1, VANGUARD, QRT2, GA2oX, AGAMOUS-like, MYB24, and Jasmonate were down-regulated at stages 5 to 9 (preanthesis) in both microarray and qRT-PCR analyses. The level of down-regulation was comparable for genes such as argonautes and katanin involved in gene repression mediated through the small interfering RNA (siRNA) pathway. The dynamics of argonautes are known to be regulated by genes such as RDR2, RDR6, DCL3, SGS3, AGAMOUS, APETALA1, and LEAFY, which also were down-regulated in homozygous tramgap1-2 plants (Supplemental Table S3).

Figure 7.

qRT-PCR analysis of selected reproductive pathway genes in the wild type (WT) and the tramgap1-2 mutant. Mutation in TRAMGaP leads to down-regulation of many genes involved in different development-related pathways. qRT-PCR analysis for key genes involved in reproductive development, including pathway genes related to RNA-mediated development, CRL complex, ovule and pollen development, and lipid and hormones, showed differential expression in the homozygous tramgap1-2 mutant. Most of the genes were down-regulated in the mutant.

TRAMGaP Regulates the Specification of MMC through Interaction with AGO Proteins

qRT-PCR analysis showed down-regulation of argonaute, dicer-like, and SGS genes, which play a crucial role in RNA interference-mediated signaling to control the differentiation of MMC progenitor cells. This observation prompted us to examine whether the tramgap mutants also display phenotypes similar to those of the ago9 mutation at the premeiotic stage. Indeed, out of 135 ovules examined, about 40% displayed the ago9-like phenotype (i.e. the presence of multiple MMCs in the nucellar region; Fig. 4, C–E), whereas the remaining 60% of ovules showed the wild-type-like phenotype (Fig. 4A). Postmeiotic ovules also showed ago9-like phenotypes (Fig. 8, A and B) in which abnormal gametic cells as well as degenerating and functional megaspores were visible. The pattern of callose deposition was assessed in wild-type and tramgap mutant ovules to determine whether two or more enlarged cells in tramgap ovules show MMC-specific callose deposition. Ovules of the wild type showed uniform callose deposition in the MMC before the initiation of meiosis (Fig. 8, C and D); after meiosis, callose deposition was observed in transverse walls between the functional megaspore and its degenerated sister cells (Fig. 8, E and F). In tramgap1-2 ovules, faint patches of callose deposition were seen in enlarged abnormal gametic cells (Fig. 8, G and H). However, in tramgap mutant ovules at postmeiotic stages, callose deposition was observed in MMC as well as in the nucellar region (Fig. 8, I and J). Ovules with several enlarged cells in the tramgap mutant showed callose deposition in multiple cells (Fig. 8, K and L). These results indicate that all four meiotic products had survived in some of the ovules and that TRAMGaP is essential for restricting the identity and differentiation of MMC to a single subepidermal cell in premeiotic ovules, which may be achieved in combination with AGO-9.

Figure 8.

Callose deposition in developing ovules of the wild type and homozygous tramgap mutants of Arabidopsis. A, Postmeiotic tramgap1-1 ovule showing abnormal gametic cells (AGC) adjacent to a degenerated megaspore (arrows) and a functional megaspore (FM). B, Postmeiotic tramgap1-2 ovule showing abnormal gametic cells adjacent to a degenerated megaspore (asterisk) and a functional megaspore. C, DIC image of a premeiotic wild-type ovule showing a single MMC. D, Fluorescent image of the ovule in C showing uniform callose deposition on the MMC. Callose is seen on transverse walls between the functional megaspore and degenerated sister cells. E, DIC image of a premeiotic tramgap1-2 ovule showing two MMCs. F, Fluorescent image of the ovule in E showing faint callose patches in both MMCs. G, DIC image of a postmeiotic tramgap1-2 ovule showing MMC and surrounding tissue. H, Fluorescent image of the ovule in G showing callose patches in MMC and also in surrounding cells. I, DIC image of a postmeiotic tramgap1-2 ovule showing multiple megaspores. J, Fluorescent image of the ovule in I showing patches of callose in a number of cells. K, DIC image of a postmeiotic tramgap1-2 ovule showing multiple megaspores. L, Fluorescent image of the ovule in K showing callose patches in a number of cells. Percentages are the frequencies that occur in the respective phenotypes of the mutant.

AGOs and katanin-like proteins participate in the translational repression of genes guided by siRNA/microRNA (miRNA; Brodersen et al., 2008). An in silico search at AthaMap (www.athamap.de/search.php?restriction=0&chromosome=5&pos=9226100) revealed that TRAMGaP is a small RNA-regulated gene with one miRNA (MIR777) targeting the 5′ untranslated region and 30 siRNAs matching the TRAMGaP gene. A majority of these siRNAs had overlapping sequences that corresponded to the fourth intron of TRAMGaP (Supplemental Fig. S12). Furthermore, two-thirds of these siRNAs were associated with AGO4. Thus, TRAMGaP mutation is expected to affect siRNA and target mRNA stoichiometry, leading to pleiotropic effects.

DISCUSSION

Intercellular communications between sporophytic and gametophytic cells are critical for cell specification and proper development during the alternation of generations (Chevalier et al., 2011). Although a number of genes involved in embryo sac development have been identified in Arabidopsis, the signaling pathways and gene interactions are far from clear. In animals, TRAF proteins serve as adaptors in signal transduction involving membrane-bound TNF-R. Very few of the large number of TRAF-like genes that have been annotated in Arabidopsis have been investigated so far. Considering that TRAF-like genes in Arabidopsis occur in a cluster and share a high degree of homology (Cosson et al., 2010), functional redundancy might prevent the ready manifestation of mutant phenotypes. To our knowledge, this is the first report to demonstrate that TRAF-like genes are critical to reproductive development in plants.

We studied two T-DNA insertion mutants each disrupting different MATH domains of the TRAMGaP gene. Both showed comparable altered phenotypes in heterozygous plants, probably owing to haploinsufficiency. The expression pattern of TRAMGaP assessed by reporter gene assay showed GUS expression in the developing gametophytes and in the surrounding sporophytic tissue. Furthermore, crossing experiments and pollen viability studies clearly indicated that the development of wild-type gametes also was affected adversely in heterozygous mutants. Thus, the tramgap mutation was found to have both sporophytic and gametophytic effects. Nevertheless, the tramgap mutation was not completely lethal and some homozygous plants were obtained, probably because the overexpression of some closely related paralogs compensated to some extent for the loss of tramgap expression.

TRAFs regulate several functions of the TNF-R superfamily, apparently by linking the cytosolic domain of the receptors to downstream protein kinases or ubiquitin ligases, and the trimerization of the MATH domain of TRAFs is essential for the binding of other downstream signaling proteins (Arch et al., 1998; Wallach et al., 1999; Deng et al., 2000). In animals, the TNF-R superfamily is involved in a wide range of biological functions, such as adaptive and innate immunity, embryonic development, stress response, and bone metabolism, through the induction of cell activation, cell survival, and antiapoptotic functions mostly mediated by the family of TRAFs (Park et al., 2000; Chung et al., 2002). Therefore, disruption of this domain is expected to abolish gene function. This expectation was strengthened by the observation that the tramgap1-2 mutation that led to the loss of both MATH domains produced a more pronounced phenotype. The MATH domain of MEL-26 (a substrate adaptor) interacts with specific substrates such as MEI-1 or katanin to associate with the CUL-3 complex (Srayko et al., 2000; Kurz et al., 2002; Pintard et al., 2003; Bowerman and Kurz, 2006). These findings indicate that the MATH domain interacts directly with the KATANIN protein in the CRL complex. Burk et al. (2001) reported that a mutation in AtKTN1, the Arabidopsis ortholog of katanin, results in siliques that are 30% to 80% shorter than normal siliques, a feature that was also observed in the tramgap mutants. These results strongly implicate MATH as an interacting partner for the katanin-like protein. The BTB/SPOP adaptor proteins of the CUL-3 complex have been shown to bind to the substrate at one end through the MATH domain and to the CRL complex at the other end through the BTB domain (Pintard et al., 2003; Hua and Vierstra, 2011; Zhang et al., 2014). The two MATH domains in At5g26290 might behave as the BTB/SPOP-like protein of the CRL complex. Quaternary structure analysis of TRAMGaP also showed that one of the MATH domains was a TRAF-like protein whereas the other was similar to the SPOP. Therefore, we speculate that TRAMGaP works as a substrate adaptor protein of the CRL complex.

In C. elegans, KATANIN localizes to the spindle and chromosomes during meiosis, and ubiquitin-ligase activity of the CUL-3 complex is required to degrade KATANIN after meiosis, which is essential for assembling a functional mitotic spindle leading to the meiosis-to-mitosis transition (Pintard et al., 2003). The persistence of meiotic products in ovules at anthesis clearly indicates that meiosis-to-mitosis transition signaling is disrupted in tramgap mutants. Moreover, katanin disruption has been shown to increase the polypeptide levels of several small RNA-regulated genes, including argonautes, without a corresponding change in the abundance of their mRNAs (Ma et al., 2013). A recent study by Gao et al. (2014) showed that miRNAs play an important role in regulating the TRAF-mediated NF-κB signaling pathway. Classification of At5g26290 as a siRNA target gene (AthaMap), the association of siRNAs corresponding to the TRAMGaP gene with AGO4, and the down-regulation of different argonaute members in tramgap mutants suggest that disturbance of one gene would affect the other. Olmedo-Monfil et al. (2010) and Tucker et al. (2012) showed that argonaute proteins are critical to signaling during the development of gametogenic progenitor cells (mitotic-to-meiotic transition) and during gametogenesis, wherein AGO9 is confined mainly to the L1 layer of the developing ovule but passes on its signal to gametogenic progenitor cells to form functional megaspores; however, an AGO gene from rice (MEL1) was reported to express itself in MMC (Nonomura et al., 2007). Both AGO5 and AGO104 were expressed in the L1 layer and in the nucellus, suggesting signaling from the sporophytic tissue of the nucellus to the developing germline (Singh et al., 2011; Tucker et al., 2012). TRAMGaP shares the expression pattern of the above-mentioned argonautes in that it is expressed in the L1 layer, in the nucellus, and in MMC (Figs. 3 and 8). As an adaptor molecule, TRAMGaP probably acts as a linker or as a carrier to transmit signals from sporophytic tissues to gametophytic tissues, thereby affecting gametophyte development. The occurrence of AGO9-like and related phenotypes (Figs. 4 and 8) at the stage of gametogenic progenitor cell differentiation points to TRAMGaP as an important member of this signaling pathway. This assumption is further supported by the altered expression of genes involved in siRNA biogenesis in conjunction with argonautes (Supplemental Table S3).

The overrepresentation of hormone-related genes in microarray (Supplemental Table S2), their down-regulation at the transcription level (Fig. 7; Supplemental Table S3), and the report that At5g26290 (TRAMGaP) is a cytokinin-responsive gene (Bhargava et al., 2013) lend further support to the assumption that TRAMGaP acts by direct or indirect binding of katanin-like proteins to control homeostasis in hormonal signals. GA signaling is routed through the CRL complex, leading to the ubiquitination of DELLA proteins (Plackett et al., 2011). These proteins are negatively regulated by MYB-like genes, which, in turn, enlist AGAMOUS-like genes to control downstream genes involved in male gametogenesis (Tang et al., 2010). Cytokinins are crucial to female gametogenesis and induce Arabidopsis Response Regulators (Cheng et al., 2013), which are involved in programmed cell death and the specification of functional megaspores in the FG in Arabidopsis (Vescovi et al., 2012; Cheng et al., 2013). Similarly, auxins are critical to the sporophytic-to-gametophytic cell transition as well as at later stages of ovule development, namely cellularization and speciation of the seven-celled embryo sac (Schmidt et al., 2015). Furthermore, F-box proteins act through the CRL complex and are known to regulate the cell cycle machinery to promote male germ cell division in Arabidopsis (Borg et al., 2009). The progression to reproductive phase, acquisition, and the restriction of reproductive cell fate during the somatic-to-gametic transition involve many different pathways such as hormonal pathways, siRNA-mediated regulation, posttranscriptional control mechanisms, and regulation by transcription factors (Schmidt et al., 2015). The overrepresentation of genes related to different hormones, kinases, and components of the CRL complex in microarray analysis (Supplemental Table S2), the altered expression of representative key genes recorded in qRT-PCR (Fig. 7; Supplemental Table S3), and the phenotypes observed at different stages of reproductive development indicate TRAMGaP to be an important molecular link that coordinates the development of male and female gametes in Arabidopsis.

In animals, TRAF proteins mediate apoptosis and cell division (Arch et al., 1998), and the observed phenotypes during the development of the ovule and pollen in Arabidopsis suggest conservation of this function in plants as well. Programmed cell death of the tapetum and the formation of exine are closely interlinked in terms of the production and accumulation of materials required for the formation of sporopollenin and its correct deposition on the pollen surface (Vizcay-Barrena and Wilson, 2006; Zhou et al., 2012). Defects in exine formation and the delayed programmed cell death of tapetal cells in tramgap mutants (Figs. 3, 5, and 6) and TRAMGaP gene expression in tapetum cells of the wild type (Fig. 3) clearly link programmed cell death to TRAMGaP. Moreover, according to eFP browser (http://bbc.botany.utoronto.ca/efp/cgi-bin/efpWeb.cgi), most of the phenylpropanoid and lipid synthesis genes are expressed during flower developmental stages 10 and 11 (prefertilization stages). The results of microarray analysis and qRT-PCR in these experiments showed that a number of genes related to the above pathways were down-regulated in the tramgap mutant and may account for the defects in programmed cell death of the tapetum.

Furthermore, we also observed several defects in megaspore and microspore mitoses. Among the genes involved in the development of the gametophyte or in the early stages of seed development, the expression of SWA1 and PHE1 was down-regulated in tramgap mutants as compared with that in the wild type. Although SWA1 is expressed in several types of tissue, it is strongly expressed in anthers, in functional megaspores, and in the embryo sac (Shi et al., 2005). Furthermore, SWA1 mutation leads to the development of the embryo sac being arrested at the two-, four-, or eight-cell stage. The differential expression of key genes regulating cell division during the male and female mitotic phase of growth, as shown by the microarray and qRT-PCR analyses, further suggests the orchestration of key genes controlling male and female gametogenesis by TRAMGaP through a complex network of interacting partners. Such key genes include ACT8, PROLIFERA, AN3, ARGONAUTES, FIE, PHE1, LIS, RBR1, and MADS-BOX (Holding and Springer, 2002; Anastasiou et al., 2007; Kapoor et al., 2008; Drews and Koltunow, 2011; Tucker et al., 2012). The protein-protein interaction network of differentially expressed genes in microarray and their gene ontology (Supplemental Fig. S11) also suggest a similar role for TRAMGaP.

Based on the phenotypes, microarray, and qRT-PCR analysis, we propose a TRAF-mediated signal transduction pathway model (Supplemental Fig. S13) to explain the probable role of TRAMGaP in reproductive development. Ubiquitination coupled with kinase-mediated signaling is a prerequisite to activation of the ubiquitin-proteasome system for hormone-mediated plant growth and development (Sadanandom et al., 2012). The interaction of katanin with the MATH domain of the CRL complex signals the regulation of different transcription factors mediated through hormonal homeostasis, which, in turn, involves precisely controlled orchestration of the differential expression of genes. Chen et al. (2013) showed that TRAFs are negatively regulated by E3 ligase components involving the F-box protein. The hormonal signals were subjected to controlled synthesis and degradation through SCF-like F-box proteins, which, in turn, are regulated by DELLA proteins. Our results support the involvement of TRAMGaP in gametogenesis-related gene expression and regulation through ubiquitin-mediated signal transduction. The SCF complexes are involved in many cellular and developmental processes in plants (Choi et al., 2014). A complete loss of their function results in an embryo-lethal phenotype, whereas the hypomorphic mutants show sterility and pleiotropic defects affecting flower development; auxin, cytokinin, GA₃, and jasmonate signaling; progression of meiosis in male gametes; and embryogenesis (Choi et al., 2014). Furthermore, the defects in the signaling complex change the expression of key genes that regulate different developmental stages in male and female gametogenesis.

Thus, we conclude that the TRAMGaP protein plays an important role in reproductive development by acting as an adaptor molecule of a complex signaling hub that integrates various transcription factors, hormonal signals, and some key regulatory genes of male and female gametogenesis. Our findings open the way to test various possible interactions between TRAMGaP and other proteins to establish a clear molecular pathway of the complex process of gametogenesis in plants.

MATERIALS AND METHODS

Plant Growth and Hybridization

While screening a population of Arabidopsis (Arabidopsis thaliana) T-DNA promoter trap (Columbia-0) developed in house, we noticed a line showing only partial seed set. Surface-sterilized seeds of that line and of the SALK line (SALK_0146328) were sown on Murashige and Skoog (1962) medium containing kanamycin (50 mg L⁻¹), and 16-d-old seedlings were transferred to pots filled with a mixture of vermiculite, peat moss, and perlite in equal proportions by volume. The plants were raised in a growth chamber (Conviron; model Adaptis A-1000) with 16 h of light alternating with 8 h of darkness at 21°C ± 1°C and 70% relative humidity. For transmission studies of the mutant alleles, flower buds were emasculated manually 1 d before anthesis and covered with bags made of waxed paper. The emasculated buds were pollinated the next day with fresh pollen from selected plants and covered again. Siliques were harvested at maturity, hybrid seeds were collected, and the progeny were raised as above. The marker line ET884 was a gift from Ueli Grossniklaus, and DD65:GFP was obtained from G.C. Pagnussat’s laboratory and analyzed as described by Pagnussat et al. (2007).

Phylogenetic Analysis

The third largest cluster of TRAF-like genes of Arabidopsis was aligned using ClustalW (MEGA 6.0; Tamura et al., 2013; http://www.ebi.ac.uk/Tools/msa/clustalw2/). The maximum parsimony program was used for tree construction with the 1,000 bootstrap test to judge the robustness of branches being clustered. The protein AML2 (At2g42890) was used as an outgroup because it is paralogous to TRAMGaP, a meiotic gene involved in a similar function, and showed a minimal level of homology (28%) throughout the protein sequence of TRAMGaP (Kaur et al., 2006).

DNA/RNA Isolation and Identification of the T-DNA Insertion Site

DNA and RNA were isolated from leaf or inflorescence samples as described earlier (Pratibha et al., 2013, 2017). The T-DNA insertion site in the mutant line was identified using genome walking (Pratibha et al., 2013). To identify the homozygous mutant lines, primers were designed from the T-DNA region and the flanking Arabidopsis chromosomal region. Homozygous and heterozygous mutant plants were separated based on the presence or absence of specific PCR fragments (Supplemental Fig. S1). Nucleotide sequences of the primers used in this study are given in Supplemental Table S4.

Mutant Complementation

The coding sequence of the At5g26290 gene was amplified with specific primers (Supplemental Table S4) having NcoI and BglII restriction sites at the 5′ ends of the forward and reverse primers, respectively, and cloned into a pCAMBIA1302 vector downstream of the cauliflower mosaic virus 35S promoter. Homozygous mutant plants carrying the T-DNA insertion in the At5g26290 gene were transformed with Agrobacterium tumefaciens containing the above gene cassette, and T1 plants were selected on Murashige and Skoog agar plates supplemented with hygromycin (20 mg mL⁻¹). Hygromycin-resistant seedlings were transferred to pots and raised to maturity.

Promoter Characterization

A 1.5-kb fragment upstream of the At5g26290 coding sequence was amplified by PCR using specific primers (Supplemental Table S4) and cloned into the pORE plant binary vector (Coutu et al., 2007) between HindIII and SacI restriction sites to drive the uidA reporter gene. Wild-type Arabidopsis plants were transformed with A. tumefaciens containing the above promoter-reporter construct, and T1 plants were analyzed for GUS expression using a histochemical assay (Jefferson et al., 1987). Tissues cleared with 70% (v/v) ethanol were observed directly with a microscope (Axio Imager 1000; Carl Zeiss).

Alexander’s Staining

Flower buds were incubated overnight in ethanol at 4°C; the anthers were dissected on a slide and incubated in Alexander’s stain (Alexander, 1969), mounted with a coverslip, kept for 30 min at room temperature, and observed with the microscope with DIC optics through a 40× objective.

Whole-Mount Preparation of Ovules and Anthers

The development of the ovule was studied using whole mounts after clearing the inflorescence in methyl benzoate as described by Siddiqi et al. (2000). Briefly, flowers were fixed in 3.7% (v/v) formalin, 5% glacial acetic acid, and 50% (v/v) ethanol (v/v) overnight at 4°C and dehydrated by passing through a series of acetone solutions. The dehydrated tissues were cleared for 2 h in methyl benzoate. Ovules were dissected on a slide using a stereo dissecting microscope, mounted with a coverslip, and observed with the microscope as described above.

The development of pollen was studied by fixing the inflorescence in formalin, glacial acetic acid, and 50% (v/v) ethanol (1:1:18) for 2 d at room temperature. The samples were dehydrated subsequently in a tertiary butyl alcohol series (Jensen, 1962), embedded in paraffin (melting point, 58°C–60°C), and 8- to 10-μm-thick sections were cut using a Finesee microtome. The sections were dewaxed and stained with 0.1% (w/v) Toluidine Blue in water followed by rinsing in clove oil to remove haziness, mounted in Canada balsam, and examined using conventional bright-field microscopy.

For DAPI staining, samples were stained with 0.1% (w/v) DAPI in 0.1 m phosphate buffer, pH 7, and photomicrographs were taken under UV illumination. To examine the architecture of the pollen wall and the patterns on the exine, pollen grains were sputtered on a double-sided rubber tape, coated with carbon using a sputter coater (E1010; Hitachi), and viewed with a Hitachi S-3400N field emission scanning electron microscope using an accelerating voltage of 30 kV.

Total RNA Isolation

Total RNA (50–100 mg) was isolated from the wild type and from the homozygous SALK mutant line using the Spectrum Plant Total RNA Kit (Sigma-Aldrich) and subjected to on-column DNase (Sigma-Aldrich) treatment to remove any contaminating DNA. The quality of RNA samples was analyzed on a 1.2% (w/v) denaturing agarose gel and quantified using NanoDrop 2000 (Thermo Scientific), and the samples were stored at −80°C until required.

Quantitative Real-Time PCR Analysis

Total RNA (1 µg) was converted to single-stranded cDNA using Invitrogen SuperScript III Reverse Transcriptase (Thermo Fisher) by following the manufacturer’s protocol, and qRT-PCR analysis was performed in an Mx3005P system (Stratagene and Agilent Technologies) using the KAPA SYBR FAST qPCR Kit (KAPA Biosystems). Total cDNA was diluted to ∼25 ng µL⁻¹, and a total of 100 ng was used in a 10-µL reaction mixture. For each reaction, three technical replicates were used along with a no-template control to check for contaminants. The following thermal cycling program was used for all qRT-PCRs: 3 min at 95°C (enzyme activation), 3 s at 95°C (denaturation), and 30 s at 60°C (annealing/extension) for 40 cycles, which includes data acquisition. Finally, a dissociation curve analysis was performed from 65°C to 95°C in increments of 0.5°C, each lasting for 5 s, to confirm the presence of a specific product. qRT-PCR primers were designed using standard parameters available at http://eu.idtdna.com/scitools/Applications/RealTimePCR (Supplemental Table S4). The concentration of ACT2 (At3g18780) was used to normalize gene expression in different samples, and the expression of TRAMGaP (At5g26290) in wild-type flower buds (stages 5–9) was used to calibrate the data in other samples. Log fold changes in expression values were calculated using the 2^−ΔΔCT method (Livak and Schmittgen, 2001).

Microarray Analysis

Global gene expression analysis and identification of the key genes showing differential regulation in the tramgap1-2 mutant were carried out using GeneChip Arabidopsis ATH1 Genome Array (Affymetrix) representing ∼24,000 genes. The homozygous tramgap1-2 mutant line was used for comparison with the wild type. A total of three technical replicates were used for each sample analyzed. From each wild-type and mutant plant, samples were collected from whole inflorescence, which included floral buds at stages 5 to 18. Total RNA was isolated as described above. For labeling, total RNA (100–250 ng) was amplified and labeled in three independent reactions using the 3′ IVT express labeling kit (Affymetrix). The biotinylated RNA samples were fragmented and hybridized to the GeneChip Arabidopsis Genome Array. The arrays were washed using an Affymetrix GeneChip fluidic station 400 and scanned using a Hewlett-Packard Gene Array Scanner G2500A. CEL files were generated using the combined console software (Affymetrix). Data from the chips showing 0.9 or greater Pearson correlation coefficient between replicates were included in the analysis.

The CEL files were loaded into GeneSpring GX11 (Agilent Technologies), and the robust multi-array average (RMA) algorithm was applied to normalize the data. Quality control (principal component and correlation analyses) was performed among the replicates to detect batch effects or other random effects. To identify the differentially expressed genes, a two-way ANOVA was used, and the computed P values were corrected by the Benjamini-Hochberg (false discovery rate) multiple testing correction method. Furthermore, probe identifiers satisfying two conditions, namely P ≤ 0.05 and a 2-log fold or greater change in expression level, were shortlisted for detailed analysis.

The differentially expressed genes were used to prepare the protein-protein interaction network of Arabidopsis by using Cytoscape, and putative functions of the clustered genes were investigated using ClueGO and REVIGO (Supek et al., 2011).

Accession Number

The Arabidopsis Information Resource accession number for TRAMGaP gene is At5g26290.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Identification of the T-DNA insertion in the tramgap1-1 mutant by genome walking.

• Supplemental Figure S2. Nucleotide and amino acid sequences of TRAMGaP.

• Supplemental Figure S3. Phylogenetic tree generated with known TRAF genes from plants and animals.

• Supplemental Figure S4. Phylogenetic analysis of TRAF genes of Arabidopsis.

• Supplemental Figure S5. Cluster alignment of TRAF-like genes of Arabidopsis showing conservation of the At5g26290 protein with its neighboring genes of the family.

• Supplemental Figure S6. Prediction of the secondary structure of the At5g26290 protein.

• Supplemental Figure S7. Prediction of the quaternary structure of the At5g26290 protein.

• Supplemental Figure S8. TRAMGaP gene, amino acid sequence, and T-DNA insertions.

• Supplemental Figure S9. Expression profile of the TRAMGaP gene.

• Supplemental Figure S10. Abnormal ovule phenotypes in tramgap mutants.

• Supplemental Figure S11. REVIGO of differentially expressed genes in microarray analysis of tramgap1-2 during flower development corresponding to stages 5 to 9.

• Supplemental Figure S12. Base pairing between siRNAs or miRNA sequences with the At5g26290 gene.

• Supplemental Figure S13. Possible TRAMGaP-mediated signal transduction during male gametophyte and FG development in Arabidopsis.

• Supplemental Table S1. qRT-PCR of TRAMGaP in various tissues.

• Supplemental Table S2. Microarray analysis of the tramgap1-2/tramgap1-2 mutant.

• Supplemental Table S3. qRT-PCR of key genes with altered expression in tramgap1-2/tramgap1-2 at preanthesis and postanthesis.

• Supplemental Table S4. Details of primers used in this study.