Changes in the content of pollen total lipid and TAG in Arabidopsis thaliana DGAT1 mutant as11

Abstract In mature pollen grains, lipids are primarily stored in the form of lipid droplets that provide energy and act as a carbon source for normal pollen development and germination. Triacylglycerol (TAG) is the major form of stored plant lipids. Diacylglycerol transferase, which is encoded by DGAT1 in Arabidopsis thaliana, is an important enzyme regulating triacylglycerol synthesis. Within the seeds of the DGAT1 mutant as11, the content of TAG is significantly decreased and the fatty acid composition also differs from the wild type. Transcriptome data of mature anthers showed that the genes involved in the TAG synthesis pathway were downregulated in as11. Analysis of gene expression patterns via transcriptome data also revealed that the expression of PDAT1, which functions in a manner complementary to the DGAT1 gene, was significantly decreased in as11, whereas the amylopectin synthase genes SS1 and SS2 were upregulated in mutant as11. We also detected lower total lipid, TAG and fatty acid contents in mature as11 pollen, with palmitic acid (C16:0) and linolenic acid (C18:3) being the major fatty acids in mature pollen. The cytological results showed that the lipid droplet content was reduced in mature as11 pollen. In the binuclear pollen grain II stage, WT pollen contained lipid droplets that were primarily accumulated around the generative nucleus, whereas the pollen in the mutant as11 was rich in starch grains that were primarily distributed around the vegetative nucleus. Ultrastructural analysis indicated that during pollen development in as11, the amount of endoplasmic reticulum in tapetal cells and pollen grains decreased, whereas the Golgi body content increased, which directly or indirectly led to a decrease in the levels of lipidosomes and an increase in the starch content in as11. Changes in the lipid content and fatty acid composition of the mutant as11 differ from those in the wild type during pollen development.


Introduction
Pollen development in higher plants is a complicated process that involves the synthesis and secretion of nutriments such as carbohydrates, lipids, and others to provide nutrition for subsequent pollen development (Liu et al., 2011). Tapetal cells can also synthesize and secrete different nutriments at various stages throughout development to allow for the nourishment of pollen grains Chen et al., 2013). Typically, a large number of nutriments are accumulated within microspores at the binuclear pollen grain stage, and these nutriments are primarily lipids, storage proteins, and/or carbohydrates (Calzoni et al., 1990). For example, during pollen development in Lilium, a large number of starch grains are present within the vegetative cells at the early binuclear pollen grain stage. However, they begin to store lipid droplets as the primary nutriment in mature pollen (Clément et al., 1994). Interestingly, during pollen development in tomato and Asian lotus, the opposite phenomenon is observed, with more starch grains than lipid droplets in mature pollen (Zhu and Tian, 2015;Zhang et al., 2019). However, during pollen development in the Hong Kong orchid tree, lipid droplets are present at the early stage, the lipid droplet and starch grain contents increase at the late binuclear pollen stage, and mature pollen grains store both lipid droplets and starch grains, but with the lipid content being dominant (Zheng et al., 2017). At the binuclear pollen grain stage in Chinese cabbage, lipid droplets are abundant in both the tapetum and the pollen (Xie et al., 2005). Additionally, the transport and transformation of nutriments are both directly related to pollen fertility. The innermost layer of anther parietal cells (the tapetum) will select, transform and transport nutriments entering the anther locule. Therefore, the abnormal accumulation of nutriments within the pollen and tapetum may result in pollen abortion (Xie et al., 2005;Xu et al., 2009;Zhang et al., 2009;Yang et al., 2017).
Lipids and lipid derivatives are important components of pollen grains, particularly the mature pollen of most entomophilous flowers, and the lipid content can reach up to 25% of their dry weight (Murphy, 2001). Lipids that are accumulated in mature pollen can be divided into two categories. In the first category, the main components are phenolic substances and lipids. These are primarily composed of fatty acids and carotenoids that are synthesized and secreted by the tapetum and released into the anther locule to cover and deposit onto the extine to form the tryphine (Ischebeck, 2016). The second category includes intracellular self-synthesized lipids that are formed during pollen development and is primarily comprised of polar lipids that are related to membrane synthesis and neutral lipids that are stored in lipid droplets (Rodriguez-García et al., 2003;Ischebeck, 2016). Lipid droplets in mature pollen possess one unique and relatively simple structure and are similar to oil bodies that are stored in plant seeds. Triacylglycerol (TAG) and sterol ester are the main components, and the outside is covered by a phospholipid monolayer membrane that is inlaid with a variety of proteins (Thiam et al., 2013;Ohsaki et al., 2014). There are two sites in plants where lipid droplets are synthesized (chloroplasts and endoplasmic reticulum), and lipid droplets in mature pollen are primarily synthesized by the endoplasmic reticulum (Pyc et al., 2017). During late pollen development, lipid droplets are surrounded by a large number of endoplasmic reticulum structures to ensure that the lipid droplets are prevented from fusing with each other during pollen rehydration without the protection of oleosin (Murphy, 2001). TAG is stored in large amounts within lipid droplets and acts as the primary source for providing energy and as an inert intermediate that regulates lipid synthesis (Ziekiewicz et al., 2013). Therefore, TAG synthesis is essential for pollen development Ischebeck, 2016).
In addition to its content within pollen, TAG exists extensively in plant seeds and fruits as the most primary storage form of plant oil (Lu et al., 2003;Saha et al., 2006), where it provides a carbon source and the nutriments necessary for seed germination and development (Routaboul et al., 1999;Zou et al., 1999;Baud et al., 2002;O'Neill et al., 2003;Zheng et al., 2003;Lung and Weselake, 2006). Diacylglycerol transacylase (DGAT) is encoded by the DGAT1 gene in Arabidopsis thaliana and is an important enzyme that catalyzes the synthesis of TAG. Compared to the wild type (WT), seeds from the DGAT1 gene mutant as11 exhibit a clear delay in development (Katavic et al., 1995;Hobbs et al., 1999;Routaboul et al., 1999;Zou et al., 1999;Jako et al., 2001;Lu and Hills, 2002). The DGAT1 gene mutant as11 slows DAG accumulation, which lowers the TAG/DAG ratio (Katavic et al., 1995;Hobbs et al., 1999;Routaboul et al., 1999;Zou et al., 1999;Jako et al., 2001). The seed oil content of as11 is decreased by 25-35% relative to the WT, and the fatty acid composition is altered with an increased linolenic acid (C18:3) content (Katavic et al., 1995;Zou et al., 1999;Lu and Hills, 2002). Although the oil content in mature as11 seeds differs from the WT, there is no significant difference between as11 and WT in regard to morphology (Katavic et al., 1995;Zou et al., 1999;Lu and Hills, 2002), dry weight, fresh weight and protein levels (Katavic et al., 1995). However, as11 seeds show delayed germination, which can be recovered by treatment with a high concentration of mannite (Routaboul et al., 1999;Lu and Hills, 2002). Additionally, the sucrose content in dry seeds of as11 was higher than in the WT, and increased and then decreased after absorbing water and finally fell to the same level as that of the WT (Lu and Hills, 2002).
In recent decades, the majority of studies examining the DGAT1 mutant as11 have primarily focused on genetics, molecular biology, physiology and biochemistry (Katavic et al., 1995;Hobbs et al., 1999;Routaboul et al., 1999;Zou et al., 1999;Jako et al., 2001;Lu et al., 2003;Zhang et al., 2009;Xu et al., 2012). Studies on the total lipid and TAG content during pollen development in as11 are rare. Our transcriptional analysis found that the expression levels of key genes in the TAG synthesis pathway were downregulated in mature pollen, and the amount of endoplasmic reticulum (which indirectly influences the lipidosome content in the tapetum) was different in the mutant, ultimately resulting in a decrease in the total lipid and TAG contents in the mature pollen and in changes in fatty acid composition and expression of genes related to lipid synthesis during pollen development.

Plant material
Seeds of A. thaliana ecotype Columbia and DGAT1 mutant as11 were sown onto damp nutrient soil and cold treated for at least 2 days at 4 °C. Then seedlings were placed into an incubator with illumination for 16 h and darkness for 8 h per 24-h cycle. Plants with normal growth and development were used for the experiments.

Thin section and histochemical staining
Flower buds at different developmental stages were harvested and processed according to a previously described method, with a few modifications (Zhang et al., 2002). Semithin sections (1 μm thick) were cut using a Leica RM2255 microtome, stained with PAS and Sudan Black B, and then observed and photographed under a Leica DMLB. Ultrathin sections (60-70 nm) were cut using a Leica UC6 ultramicrotome, stained with uranyl acetate and lead citrate, and observed and photographed under a Philips FEI-TECNAI 12 transmission electron microscope.

Total lipid and TAG quantification and total fatty acid profiles
Mature anthers were harvested in glass tubes on dry ice and weighed after freeze drying. Total lipid was extracted using the chloroform/methanol method described in the Acyl-Lipid Metabolism Chapter of The Arabidopsis Book (2013). TAG was fractionated by TLC (Silica gel 60, Merck, USA) in developing solvent (hexane: diethylether: acetic acid = 70: 30: 1 (v/v/v)) and visualized by iodine vapor. Spots were scraped for following fatty acid methyl esters (FAME) with TAG standard (C15:0) to quantify the content of TAG. FAME of TAG and anthers at the VM (vacuolate microspore), BM (binuclear pollen grain), SMD (second mitotic division) and MP (mature pollen grain) stages were produced by a direct acid-catalyzed transmethylation protocol presented in the Acyl-Lipid Metabolism Chapter of The Arabidopsis Book (2013), and analyzed by GC-MS (7890A-5975C, Agilent Technologies, USA). GC conditions were as follows-splitless mode injection, injector, and flame ionization detector (FID) temperature, 250 °C; oven temperature program-100 °C for 1 min, 15 °C min -1 to 175 °C for 12 min, then 2 °C min -1 to 245 °C, holding this temperature for 2 min.

RNA isolation and qRT-PCR
Total RNA of anthers from the VM, BM, SMD and MP stages was extracted by RNA FAST200 (Fastagen, Shanghai, China). Reverse transcription was completed using a PrimeScript™ 1st Strand cDNA Synthesis Kit (Takara, Dalian, China). Expression of genes selected according to TAG synthesis (http://aralip.plantbiology.msu.edu/pathways/triacylglycerol_ biosynthesis) was determined by 2× SYBR Premix Ex Taq with a CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Inc., USA). Arabidopsis thaliana Actin2 gene was used as the housekeeping gene. Three biological replicates were done for each gene. The primers are listed in Supporting Information- Table S1.

Library preparation for transcriptome sequencing
Total mature anther RNA for each sample (3 biological replicates) was extracted using the TRIzol method (TIANGEN BIOTECH, Beijing) and treated with RNase-free DNase I (TaKaRa). RNA degradation and contamination was monitored on 1% agarose gels. RNA was quantified using an Agilent 2100 Bioanalyzer (Agilent Technologies, CA, USA), and its quality and integrity were assessed using a NanoDrop spectrophotometer (IMPLEN, CA, USA). A total of 1.5 μg RNA per sample was used as input material for the RNA sample preparations. Sequencing libraries were generated using the NEBNext ® Ultra™ RNA Library Prep Kit for Illumina® (NEB, USA) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in NEBNext First Strand Synthesis Reaction Buffer (5×). First-strand cDNA was synthesized using random hexamer primers and M-MuLV Reverse Transcriptase (RNase H). Second-strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activity. After adenylation of the 3ʹ ends of DNA fragments, NEBNext Adaptor with a hairpin loop structure was ligated to prepare for hybridization. In order to select cDNA fragments of preferentially 200-250 bp in length, the library fragments were purified with the AMPure XP system (Beckman Coulter, Beverly, USA). Then 3 μL USER Enzyme (NEB, USA) was used with size-selected, adaptor-ligated cDNA at 37 °C for 15 min followed by 5 min at 95 °C before PCR. Then PCR was performed with Phusion High-Fidelity DNA polymerase, universal PCR primers and Index (X) Primer. Finally, PCR products were purified (AMPure XP system) and library quality was assessed on the Agilent Bioanalyzer 2100 system. The library preparations were sequenced on an Illumina Hiseq 4000 platform by the Beijing Allwegene Technology Company Limited (Beijing, China) and paired-end 150 bp reads were generated.

Quality control
Raw data (raw reads) of fastq format were firstly processed through in-house perl scripts. In this step, clean data (clean reads) were obtained by removing reads containing adapter, reads containing ploy-N and low-quality reads from raw data. At the same time, Q20, Q30, GC-content and sequence duplication level of the clean data were calculated. All the downstream analyses were based on clean data with high quality.

Quantification of gene expression levels
HTSeq v 0.5.4 p3 was used to count the read numbers mapped to each gene. Gene expression levels were estimated by fragments per kilobase of transcript per million fragments mapped (FPKM).

Differential expression analysis
Differential expression analysis of two groups was performed using the DESeq R package (1.10.1). DESeq provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting P values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted P-value < 0.05 found by DESeq were classified as differentially expressed.

GO enrichment analysis
Gene Ontology (GO) enrichment analysis of the differentially expressed genes (DEGs) was implemented by the GOseq R packages based Wallenius non-central hyper-geometric distribution (Young et al., 2010), which can adjust for gene length bias in DEGs.

KEGG pathway enrichment analysis
KEGG (Kanehisa et al., 2008) is a database resource for understanding high-level functions and utilities of biological systems, such as the cell, the organism and the ecosystem, based on molecular-level information, especially large-scale molecular datasets generated by genome sequencing and other high-throughput experimental technologies (http:// www.genome.jp/kegg/). We used KOBAS (Mao et al., 2005) software to test the statistical enrichment of differentially expressed genes in KEGG pathways.

Data analysis
Figures were plotted by Origin2018 (Microcal Software Inc., Northampton, MA). All histograms were processed with Excel (Microsoft, Seattle, WA). The data were statistically analyzed using SPSS version 20.0 (SPSS Inc., Chicago, IL). One-way ANOVA (Duncan's multiple range test) was performed and differences were considered significant at p < 0.05.

Transcriptome analysis revealed candidate genes involved in TAG biosynthesis
Using the Illumina HiSeq4000 sequencing platform and PE150 sequencing strategy, 141,560,394 raw read pairings were measured and 139,274,010 clean read pairings were obtained after quality control. A total of 10,988 differential genes were identified, of which 5,175 were upregulated and 5,813 were downregulated in as11 (Fig. 1). GO enrichment analysis of differentially expressed genes resulted in 10,562 genes being annotated into 30 subclasses of two GO classes: biological processes and cellular components. A total of 6573 DEGs were identified as being involved in cellular processes in the biological process domain, and 7910 and 7920 DEGs were identified as being involved in cell parts and cells in the cellular component morphogenesis domain respectively. KEGG enrichment analysis of differentially expressed genes identified 2109 DEGs. In as11, 2105 DEGs were upregulated and 704 DEGs were downregulated. They were mainly enriched in metabolic pathways (383 upregulated, 460 downregulated) such as biosynthesis of secondary metabolism, carbon metabolism, biosynthesis of amino acids, fatty acid metabolism and fatty acid degradation (fatty acid metabolism) [see Supporting Information- Fig. S1]. Forty-eight of the 69 related genes identified in the fatty acid metabolism pathway were differentially expressed, and 41 of the 60 genes identified in the glycerolipid metabolism pathway were differentially expressed. Among the 43 genes identified in the fatty acid biosynthesis pathway, 28 were differentially expressed. According to the TAG synthesis pathway (Fig. 2), 8 key genes with changes in expression were identified (Table 1), and subsequent studies of these were carried out.

The lower expression of genes related to lipid synthesis and content of total lipid and TAG in mature as11 pollen
Based on transcriptome data, we next tested the expression of lipid synthesis related genes of wild type and mutant as11 in mature pollen grains. Our results indicated that at the mature pollen grain stage, the expression of GPAT (GPAT1, GPAT6, and GPAT9) and PDAT1, which are involved in seed TAG synthesis, was downregulated in as11 (Fig. 3). The GPDHc1 gene, which regulates the transition from glycometabolism to lipid metabolism, was expressed at a higher level in the WT (Fig. 3). The expression analysis of the key genes related to fatty acid synthesis revealed that the expression of FAB1, which catalyzes the conversion of C16:0-ACP to C18:0-ACP, and the key enzyme gene FAD2, which can catalyze the production of linoleic acid (C18:2) was distinctly upregulated in the WT (Fig. 3). However, it is interesting that the expression of starch synthetase genes, particularly the SS1 and SS2 genes that are primarily responsible for the regulation of amylopectin synthesis, was higher in as11 (Fig. 3).
Based on the lower expression of lipid synthesis related genes in as11, we further determined the content of total lipid and TAG in mature pollen grains of the WT and as11. The total lipid in mature pollen grains of as11 was 26% less than that in the WT (Fig. 4A). The TAG content of the total lipid was also significantly lower in as11 (Fig. 4B). As a result, the total content of TAG in mature as11 pollen grains was distinctly lower than that in the WT (Fig. 4C). We next measured the fatty acid composition of the mature WT and as11 pollen grains, and found no significant difference in content (Fig.  4D). Palmitic acid (C16:0) and linolenic acid (C18:3) were the main fatty acids in mature pollen in both the WT and as11. The content of linoleic acid (C18:2) and linolenic acid (C18:3) was a little higher in the WT. The stearic acid (C18:0) content followed the inverse pattern.
We also measured the fatty acid composition of wild type and mutant as11 pollen grains at the VM (vacuolate microspore), BN (binuclear pollen grain), and SMD (second mitotic division) stages, and found that the content of linolenic acid (C18:3) was much higher in as11 than in the WT at the SMD stage, presenting a 'trade-off' with linoleic acid (C18:2), palmitic acid (C16:0) and stearic acid (C18:0). The relative content of fatty acids at the VM stage was used as the standard to obtain the relative ratio of palmitic acid (C16:0), stearic acid (C18:0) and linolenic acid (C18:3) at four developmental stages. We found that at the SMD stage, the ratios of palmitic acid (C16:0) and stearic acid (C18:0) were lower in as11 than in the WT (Fig. S3). Accordingly, the ratio of linolenic acid (C18:3) was significantly higher in as11 than in the WT at the same stage. When the grains were mature, the ratio of the three fatty acids in both recovered the same between WT and as11.

The pollen ultrastructure changed during as11 pollen development
Based on the developmental characteristics of the tapetum and pollen grains in the WT (Col ecotype) and according to the 13 growth stages of the entire developmental period in the wild-type Arabidopsis thaliana ecotype Wassilewskija (Owen and Makaroff, 1995), we divided the mature pollen grain stages of Arabidopsis thaliana ecotype Columbia into mature I and mature II stages, giving a total of 14 growth stages. The growth stages of mutant as11 were divided on the basis of pollen development.
Histochemical staining results indicated that there was no significant difference in the accumulation and distribution of nutrients during pollen development in either the WT or as11 from the pre-meiosis stage to the released microspore stage (Fig. S2). At the vacuolate microspore stage, black granular lipid could be observed only in the tapetum of the WT (Fig.  5A). At the binuclear pollen grain II stage, starch grains surrounded the vegetative nucleus in both the WT and as11 according to PAS staining ( Fig. 5B and b); however, the mutant exhibited more starch accumulation. After counterstaining with Sudan Black B, there was obvious lipid accumulation in both the tapetum and the pollen grains of the WT and as11, and lipid droplets within the pollen grains were predominantly distributed around the generative nucleus ( Fig.   Fig. 1. Differentially expressed genes in WT and as11. Volcano plot of DEGs between WT and as11. Each dot in the volcano plot represents a gene; the abscissa represents the logarithm of the difference multiple of gene expression between the WT and as11 lines, and the ordinate represents the negative logarithm of statistically significant changes in gene expression. The larger the absolute value of the abscissa, the greater the multiple difference of expression quantity between the two lines; the larger the vertical coordinate value, the more significant the differential expression level, and the more reliable the selected DEGs. The green dots represent downregulated DEGs, the red dots represent upregulated DEGs, and the blue dots represent no DEGs. In comparing the gene expression in the WT and mutant, 10988 DEGs were consistently expressed in both the WT and mutant, 5175 were upregulated and 5813 were downregulated in the mutant. 5C and c). Furthermore, the content of lipid droplets within WT pollen was significantly higher than that in as11. When pollen grains developed to the second mitosis I stage, the number of starch grains within the pollen grains decreased significantly ( Fig. 5D and d). Tapetal cells began to degrade, and lipid accumulation within the pollen grains continued to increase, with more lipid droplets in the WT (Fig. 5E and e). At the mature stage WT pollen grains were rich in lipid droplets (Fig. 5F), and the lipid droplet content was significantly higher (Fig. 5f). Both mature WT and as11 pollen only contained a small amount of starch grains ( Fig. 6F and f).
To explore the results at the cytological level, the ultrastructure of both the WT and as11 was observed at different developmental stages. At the released microspore II stage, the microspores had a centrally located nucleus that was surrounded by a small amount of endoplasmic reticulum, and these microspores began to secrete substances to form the pollen intine ( Fig. 6A and a). By this stage, small lipid droplets had already begun to form in the WT (Fig. 6A), while more Golgi bodies were present in as11 (Fig. 6a). At the binuclear pollen grain I stage, a lenticular-shaped generative cell produced by an asymmetric mitosis was located on one side of the   6B and b). For further development to the binuclear pollen grain II stage, generative cells were free in the cytoplast of the vegetative cells, and these cells contained abundant vesicles and organelles such as mitochondria, plastids, and others. Lipid droplets precipitated at the surface of the generative nucleus, and plastids containing starch grains were distributed around the vegetative nuclei in both the WT and as11 (Fig. 6C and c). Compared to the WT, the content of lipid droplets was reduced but plastids that were rich in starch grains were more numerous in mutant as11 at this stage. When pollen grains matured, the cytoplasm of WT vegetative cells was rich in lipid droplets and endoplasmic reticulum and contained numerous electrontransparent vesicles (Fig. 6D). In contrast, lipid droplets and endoplasmic reticulum were lacking in as11 pollen (Fig. 6d); however, electron-transparent vesicles that were covered with endoplasmic reticulum (Fig. 6d and d-1) and electron-dense vesicles ( Fig. 6d-1) were both abundant in the cytoplasm.

The pollen ultrastructural changes were related to ultrastructural changes in the elaiosomes and the lipidosomes in the tapetum
Most of the nutrients in pollen grains come from the tapetum, so we also observed and compared the ultrastructure of associated organelles in the WT and mutant as11 tapetal cells at different developmental stages, with particular focus on the developmental processes of the elaiosomes and the lipidosomes. At the post-meiosis and cytokinesis stage in WT tapetum, a small number of Golgi bodies were present in the cytoplasm, but plastids were abundant, and osmiophilic deposits appeared to be internal (Fig. 7A). Osmiophilic deposits were not obvious within plastids in the mutant tapetal cells, and the Golgi bodies were present (Fig. 7a). At the tetrad stage, plastids in WT tapetum began to accumulate internal osmiophilic and electron-transparent deposits. The endoplasmic reticulum was abundant and primarily located between the nucleus and the cytomembrane. The Golgi bodies were highly developed (Fig. 7B). The tapetum of mutant as11 exhibited an increase in the content of endoplasmic reticulum and these structures were also rich in Golgi bodies actively secreting vesicles. Osmiophilic deposits were also present in plastids (Fig. 7b). When microspores were released, in the WT tapetum numerous vesicles were stacked along the edge of plasmalemma, and the swollen endoplasmic reticulum was more obvious. The plastids were not significantly altered (Fig. 7C). In as11, the Golgi bodies were very abundant and were swollen with rich secretory vesicles. Osmiophilic droplets appeared to be increased within plastids (Fig. 7c). At the stage of vacuolate microspores in WT tapetum, the endoplasmic reticulum was still plentiful, and a large number of plastoglobules that exhibited a centered rectangular chip structure began to accumulate in plastids. Additionally, numerous expanded Golgi bodies and exuberant secretory vesicles developed into a relatively stable elaiosomes. New lipid-like structures called lipidosomes also became visible, which was surrounded by extensive endoplasmic reticulum (Fig. 7D). At the vacuolate microspore stage in as11 tapetum, vesicles harboring electron-dense materials increased in number, and plastoglobules began to accumulate within plastids. At this stage, lipidosomes were not observed in mutant tapetal cells (Fig. 7d). At the binuclear pollen grain I stage, the plastoglobules formed a centered rectangular chip structure in plastids that subsequently became relatively stable elaiosomes. Additionally, a small number of lipidosomes appeared (Fig. 7e). As pollen grains continued to develop, the Golgi bodies were reduced and vesicles possessing electrondense inclusions were still present in the WT tapetum. The elaiosomes remained stabilized, and the lipidosomes increased and fused with each other to form sheeting structures that were surrounded by plentiful endoplasmic reticulum (Fig. 7E). At the binuclear pollen grain II stage, the tapetal cells were primarily occupied by elaiosomess and lipidosomes, and the content of both structures reached its maximum (Fig. 7F). The development of mutant tapetal cells during the late binuclear pollen grain stage was not significantly different from that of the WT (Fig. 7f). The content of elaiosomes and liposomes, which occupied most of the volume in tapetal cells, also reached its peak at the binuclear pollen grain II stage; however, the lipidosome content was not as rich as that of the WT (Fig. 7f). At the second mitotic division I stage, the tapetal cells of both the WT and as11 were degraded. The electron density of the lipidosomes increased, and the structure of the elaiosomes remained stable. The structures were both released into the anther loculus after the plasma membrane had degraded, and they were deposited as a covering for the extine of pollen grains to form the tryphine (Fig. 7G and g). At the second mitotic division II stage, the tapetum of the WT and as11 continued to release degraded cellular inclusions, and ultimately, these cells became vacuolate sheeting lipid structures that were fused by the lipidosomes. It should be noted that the lipid structure in the mutant as11 was less electron-dense and did not fill the entire tapetal residual structure compared to the observations in the WT (Fig. 7H and h).
We randomly calculated the area of the lipidosomes and elaiosomes in 20 tapetal cells from the WT and mutant at the binuclear pollen grain stage, and found that the area of lipidosomes in WT tapetal cells was 3.49 times that of the mutant, while the area of the elaiosomes in mutant tapetal cells was 2.26 times that of the WT (Fig. 8).

Fatty acid metabolism-related DEGs involved in as11
Transcriptome sequencing is now a routine experimental method for differential gene expression, quantitative gene expression, and transcriptional identification (Mao et al., 2021). Kim et al. (2016) analyzed and identified genes related to fatty Fig. 4. Comparison of total lipid, TAG levels and relative content of total fatty acid levels between anthers of the WT and as11. A. The content of total lipid. Based on 100 mg freeze-dried mature anthers, the content of total fat in WT mature anthers was 25.71 mg, and that of as11 was 18.88 mg, 26.57% lower than that of the WT. B. The TAG of total lipid. 1 mg of the total lipid was separated by TLC, and GC-MS was used to calculate the TAG content: 6.60 μg/mg (TAG/total lipid) in WT mature anthers and 3.48 μg/mg in as11 anthers. Compared with the WT, the content was 47.21% lower in the mutant. C. The total TAG in the anthers of the WT and as11. The content of TAG in the WT was 169.60 μg per 100 mg of freeze-dried mature anthers, while the content of TAG in as11 was 65.32 μg, a decrease of 61.48% compared with that in the WT. D. The relative content of total fatty acid levels at the mature pollen grain stage. Significant differences are indicated by *P < 0.05 and **P < 0.01. Three biological replicates were performed. acid biosynthesis in Perilla frutescens (L.) var frutescens by transcriptomic analysis. Studies on changes in the fatty acid metabolism in Arabidopsis DGAT1 mutant as11 seed have been relatively comprehensive, and the transcript profiling of developing WT and as11 seed showed differences in the fatty acid metabolism gene expression (Zou et al., 1999;Aulakh and Durrett, 2019), but the genes related to fatty acid biosynthesis during pollen development of the mutant have not yet been clarified. Zhang et al. (2021) found fatty acid metabolism genes in Ophiocordyceps sinensis by transcriptomics.
In the current study, a total of 10,988 differential genes were identified, of which 5,175 were upregulated and 5,813 were downregulated according to the transcriptome analysis (Fig.  1).
DGAT is an important enzyme catalyzing TAG synthesis in the Kennedy pathway, which is of great significance for TAG synthesis and accumulation (Zou et al., 1999;Lu and Hills, 2002;Lu et al., 2003). In addition to the Kennedy pathway, Dahlqvist et al. (2000) found that phospholipid:diacylglycerol acyltransferase (PDAT) in Arabidopsis thaliana can catalyze the synthesis of TAG. The GPDHc1 gene expression product is glycerol-3-phosphate dehydrogenase with NAD dependence, which can directly catalyze the formation of dihydroxyacetone phosphate (DHAP) in the process of sugar metabolism to generate glycerol-3-phosphate (G3P), and then directly enter the Kennedy pathway to provide the glycerol framework for TAG synthesis. It plays an important role in the mutual conversion of sugars and lipids (Shen et al., 2006;To et al., 2012). The FAB1 gene expression product β-Ketoacyl-ACP synthase II (KAS II) catalyzed the elongation of fatty acids from C16:0-ACP to C18:0-ACP (Hirano et al., 2001;Pidkowich et al., 2007).
Phosphatidylcholine:diacylglycerol choline phosphotransferase PDCT can transfer choline from C18:2/3-PC to C18:1-DAG and generate C18:2/3-DAG and C18:1-PC. The former can continue to synthesize C18:2/3-TAG. The latter, as a substrate for fatty acid dehydrogenase (FAD), continues to be desaturated to produce C18:2/3-PC (Lu et al., 2009). PAH2 can encode phosphatidic acid phosphatase (PAP) and catalyze the production of diacylglycerol DAG (Eastmond et al., 2010). The DGAT activity of as11 seeds decreased by 40-70% relative to the WT, which affected the amount of oil accumulation in seeds, especially the synthesis of TAG (Katavic et al., 1995;Zou et al., 1999), whereas the amount of α-linolenic acid (18:3) almost doubled (Zou et al., 1999;Jako et al., 2001;Lu and Hills, 2002), this being the main fatty acid for storing TAG in mature seeds (Zou et al., 1999). In the GO enrichment analysis of differentially expressed genes, 10,562 genes were annotated into 30 subclasses of two GO classes: biological processes and cellular components. Following a KEGG enrichment analysis of differentially expressed genes in fatty acid metabolism and fatty acid degradation and based on the results of functional enrichment analysis of DEGs, we selected fatty acids as the focus of research on the differences between WT and as11. The transcriptomic data of as11 revealed that the expression levels of GPGHC1, GPAT1, GPAT6, GPAT9, PDAT1, PAH2 and FAB1 all decreased in as11. So, 7 genes related to TAG biosynthesis were selected for further analysis (Table 1).

Lipid accumulation and fatty acid composition change during pollen development in as11
Previous studies demonstrated that the content of oil and TAG in mature seeds of the Arabidopsis DGAT1 gene mutant as11 was distinctly lower than that in the WT (Katavic et al., 1995;Zou et al., 1999;Lu and Hills, 2002). The DGAT1 gene can also reduce TAG accumulation in other nutritional organs of Arabidopsis (Slocombe et al., 2009). Additionally, the TAG content in mature seeds, roots, stems, leaves, and petals was decreased in tobacco after silencing of the DGAT1 gene (Zhang et al., 2005). Antisense suppression of the DGAT1 gene in Brassica napus not only decreased the oil content of mature seeds, but also reduced the seed yield and germination rate and resulted in highly abnormal development (Lock et al., 2009). Although plant pollen is very small, it is extremely abundant in nutrients such as lipid droplets and/or starch, which are needed to ensure that there is sufficient energy and material to promote the normal germination of pollen. Any abnormal accumulation of nutrients during pollen development may cause pollen sterility (Zhang et al., 2002Nashilevitz et al., 2009;Li et al., 2012;Yang et al., 2017;Gao et al., 2019). As the main component of lipid stored in pollen grains, triacylglycerol (TAG) not only serves as an energy and carbon source, but it also participates in important metabolic processes such as membrane lipid remodeling and has great significance in regard to normal development and successful pollination (Yang and Benning, 2018). DGAT, the enzyme catalyzing the synthesis of TAG, is important for TAG synthesis and accumulation. It not only has a high content and activity in developing seeds, but also widely exists in different organs such as leaves, petals, fruits, and anthers (Hobbs et al., 1999;Zou et al., 1999;Lu and Hills, 2002;Lu et al., 2003). It was reported that during pollen development in Arabidopsis thaliana (Owen and Makaroff, 1995;Zajac, 1997;Zhang et al., 2002) and Brassica napus (Platt et al., 1998), lipid droplets first begin to accumulate at the binuclear pollen grain stage where they are mainly distributed around the generative nucleus, and these lipid droplets gradually increase as pollen develops until the pollen has matured. This phenomenon is not limited to cruciferous plants, but can also be observed in the pollen development of other plants such as Lycium barbarum (Xu et al., 2009) and Pancratium maritimum (Konyar, 2018), indicating that the binuclear pollen stage of pollen development is a critical period for the accumulation of important nutrients. The current study found that although DGAT1 gene mutation did not cause pollen abortion, it seriously affected the lipid accumulation characteristics and content during pollen development. At the binuclear pollen grain II stage, although lipid droplets that accumulated in vegetative cells of mutant as11 increased compared to the previous stage and were mostly distributed around the generative cell nucleus, the content was significantly less than that of the WT. In contrast, the number of plastids containing starch grains surrounding the vegetative nucleus was significantly higher in the mutant than in the WT (Fig. 5C and c; Fig. 6C and c). As the pollen developed, lipid continued to accumulate in pollen grains. Mature as11 pollen was not as rich in lipid droplets as WT pollen (Fig. 5F  and f; Fig. 6D and d). Furthermore there was less total lipid and TAG in mature as11 pollen than in the WT ( Fig. 4A and  C). Previous studies revealed that seed oil in mutant as11 was 25-35% lower than in the WT (Zou et al., 1999) and oil accumulated in seeds to approximately 70% of that in the WT (Xu et al., 2012). Our results revealed that the DGAT1 gene also affected lipid and TAG synthesis and accumulation in pollen grains, and that mutating the DGAT1 gene reduced the lipid and TAG content in pollen compared to that in seeds.
Fatty acids are important components of lipids. Heyl (1923) found that the pollen of ragweed contained saturated fatty acids such as lauric acid (C12:0), carmine acid (C14:0), and palmitic acid (C16:0) and unsaturated fatty acids such as oleic acid (C18:1) and linoleic acid (C18:2). Ching (1962) and Standifer (1966) detected fatty acids in pollen of five pine species and dandelion species respectively, and found that the contents of palmitic acid (C16:0), stearic acid (C18:0), oleic acid (C18:1), linoleic acid (C18:2) and linolenic acid (C18:3) were higher. Thus the fatty acids of 16-carbon palmitic acid (C16:0) and 18-carbon fatty acids are the main types in mature pollen. Zheng et al. (2003) measured fatty acids in mature pollen of WT and GPAT1-1 mutants in Arabidopsis thaliana. They found that mutation of GPAT1 affected the synthesis of linolenic acid (C18:3) in pollen. The fatty acids that accumulated in mature pollen of the WT and as11 mutant were similar to those that accumulated in Arabidopsis seeds (Lu et al., 2009;Napier and Graham, 2010), and, consistent with the types of fatty acids in pollen measured by Zheng et al. (2003), palmitic acid (C16:0) and linolenic acid (C18:3) were the main fatty acids accumulated in the current study (Fig. 4D). However, we also measured the pollen fatty acid composition at the VM, BN, and SMD stages, and observed dynamic changes in fatty acid content during pollen development. The results showed that linolenic acid (C18:3) was the main fatty acid in pollen of the WT and as11 during the second mitosis stage, with higher levels in the mutant (Fig.  S3). Moreover, changes in the relative content ratio of linolenic acid in the anthers at each developmental stage (Fig. S3) indicated that the content and synthesis rate of linolenic acid (C18:3) in as11 pollen grains were highest from the BN stage to the SMD stage. It has been reported that the DGAT1 gene can affect the composition of fatty acids accumulated in seeds of Arabidopsis thaliana, in which the level of α-linolenic acid (C18:3) accumulation in mature seeds of as11 was nearly doubled compared with that of the WT (Katavic et al., 1995;Hobbs et al., 1999), being the main storage fatty acid for TAG in mature seeds (Zou et al., 1999). In conclusion, the DGAT1 gene of Arabidopsis thaliana does not affect the type and content of fatty acids accumulated in pollen, but can affect the synthesis rate of linolenic acid. Therefore, we speculate that DGAT1 affects the expression of genes related to linolenic acid (C18:3) synthesis.

Lipid accumulation and fatty acid composition may be related to the changes in TAG biosynthesis
In addition to the Kennedy pathway, TAG was found to be available through phospholipid:diacylglycerol acyltransferase (PDAT)-mediated catalysis of DAG using phospholipids in Arabidopsis thaliana (Dahlqvist et al., 2000). In Arabidopsis seeds, PDAT is not the primary determinant of TAG synthesis, as the low transcript levels of PDAT1 in WT seeds were unchanged in DGAT1 mutant seed (Aulakh and Durrett, 2019). However, after mutating the DGAT1 gene, the expression of the PDAT1 and LPCAT2 genes was upregulated, and these pathways became the primary pathways for TAG synthesis and resulted in approximately 70% more oil accumulation in seeds than that in the WT (Xu et al., 2012). The failure to obtain a double mutant after crossing the DGAT1 gene mutant and the PDAT1 gene mutant indicated that the functions of DGAT1 and PDAT1 are complementary during pollen development . However, our research found that the expression levels of PDAT1 were higher in mature WT pollen grains. The expression of GPAT9, which is involved in seed TAG synthesis (Shockey et al., 2016), was also not clearly upregulated in as11. Zheng et al. (2003) found that the TAG content in flower buds of GPAT1 gene mutants was decreased by approximately 10%, indicating that the GPAT1 gene is involved in TAG synthesis in floral organs. The GPAT6 gene, which is mainly expressed in the floral organs including the tapetum and microspores, could cause abnormal accumulation of nutrients in pollen and defects in pollen wall development by affecting the lipid metabolism of the tapetum, and this gene can also cause premature anther senility (Li et al., 2012). It was reported that the GPAT1 and GPAT6 genes exerted important effects on the normal development of Arabidopsis pollen (Zheng et al., 2003;Li et al., 2012). Additionally, the expression of the PAH2 gene promotes the transformation of sugar to lipids and provides the glycerin skeleton for TAG synthesis, which catalyzes DAG synthesis and the GPDHc1 gene (Shen et al., 2006;Eastmond et al., 2010;To et al., 2012). Recent research shows that the soybean GPDHc1 gene is involved in lipid synthesis in the endoplasmic reticulum, and increases the content of C18:1 fatty acid in TAG by increasing the diacylglycerol (de novoDAG) produced by the G3P pathway (Zhao et al., 2021). So, we suggest that the high expression levels of GPAT1 and GPAT6 genes in the WT, in accordance with the trend for increased GPDHc1 and PAH2 expression in the same pathway of DAG synthesis, promotes TAG synthesis (Fig. 4) and the synthesis of endoplasmic reticulum in the tapetum. The synthesis of lipidosomes that are rich in TAG is closely related to that of the endoplasmic reticulum. Therefore, the high expression of GPAT1 and GPAT6 in the WT could promote the synthesis of endoplasmic reticulum in the tapetum. As a result, the obvious lipidosomes that were observed in the tapetal cells could provide raw materials for lipid synthesis in pollen.
DGAT1 gene expression not only affects the content of oil in seeds, but can also alter the type of fatty acid accumulation. Compared to that in the WT, the accumulation of linolenic acid (C18:3) was twice as high in as11 mature seeds (Katavic et al., 1995;Hobbs et al., 1999;Routaboul et al., 1999;Zou et al., 1999;Jako et al., 2001;Lu and Hills, 2002), and it is the main fatty acid stored in the TAG of mature seeds (Zou et al., 1999). Our study indicated that the main fatty acids in mature mutant as11 pollen grains did not differ from that in the WT (Fig. 2D). However, during the SMD stage, linolenic acid (C18:3) became the main fatty acid in as11, based on the change trend of relative content of linolenic acid (C18:3) during development (Fig. S3). It was concluded that DGAT1 increases the content of linolenic acid (C18:3) by regulating the expression of FAD2 during pollen development. In DGAT1 mutant seed, the transcript levels of FAD2 encoding the ER-localized desaturases responsible for the synthesis of 18:2 and 18:3 were higher than that of the WT, and the desaturase genes that introduce double bonds in various plastidial glycerolipids were unaffected, like FAB2 (Aulakh and Durrett, 2019). It was reported that the linolenic acid (C18:3) content was decreased in mutant GPAT1-1 mature pollen grains and seeds (Zheng et al., 2003), and increased in mutant GPAT9 mature seeds, compared to the WT (Shockey et al., 2016). Consistent with previous work, GPAT1 was slightly upregulated in mutant seed (Zou et al., 1999;Aulakh and Durrett, 2019). Therefore, the high expression of GPAT1 and slight upregulation of GPAT9 gene expression in the WT may lead to a slight increase in linolenic acid (C18:3) content, and we speculated that this could be one of the reasons for the upregulation of FAD2 gene expression. Furthermore, scanning electron microscopy indicated that there were few mature pollen grains in the mutant GPAT1-1, and this significantly reduced pollen fertility (Zheng et al., 2003). Based on this, the changes in pollen fatty acid composition may also exert effects on pollen fertility. The observed normal pollen morphology and self-pollination of as11 (Katavic et al., 1995), however, indicate that to guarantee pollen sterility, the fatty acid composition of as11 mature pollen grains should be significantly different. When the pollens were mature, the composition of the main fatty acids in as11 tended to be similar to that in the WT (Fig. 5D). It was shown that in order to guarantee pollen sterility, FAD2 expression decreased, which prevented the accumulation of linolenic acid, with the composition of fatty acids in the as11 mature pollen gradually tending to be the same as the composition of the main fatty acids in WT pollen grains. Thus, expression of the DGAT1 gene affects lipid accumulation and fatty acid composition during pollen development, but does not affect pollen fertility.
In conclusion, the DGAT1 gene can affect the accumulation of TAG in pollen grains by affecting the expression levels of GPDH, GPAT1, GPAT6, GPAT9 and PAH2, and affects the composition of fatty acids in pollen by affecting the expression levels of FAB1 and FAD2. 4.4 The lower lipid content in as11 may be indirectly related to the development of elaiosomes and lipidosomes in tapetal cells As the innermost anther parietal cells, tapetal cells possess the function of absorbing and transporting polysaccharides or lipids to provide nutrients for the development and germination of pollen. Therefore, abnormal metabolism in tapetal cells will directly affect the development of microspores in the anther loculus and the macromolecular nutrient synthesis and accumulation that occurs in these cells. In addition to pollen grains, the tapetum also secretes lipids to nourish microspores during anther development Chen et al., 2013). Studies examining gene expression related to TAG synthesis using in situ labeling have shown that the anther is the second most active tissue besides the seed (Piffanelli et al., 1997(Piffanelli et al., , 1998Murphy, 2001). In regard to the lipid cytological changes in the tapetum and pollen grains, Hernandez-Pinzón et al., (1999) isolated two organelles, the elaiosomes and the tapetosome, that were related to lipid synthesis by the tapetum of Brassica napus. The former is a type of oil-containing organelle that functions in the process of plant pollen development that involves sterol esters, TAG, and protein (Hernandez-Pinzón et al., 1999;Hsieh and Huang, 2004). The latter is a specific organelle in tapetal cells (Wu et al., 1997), and is rich in TAG and some proteins similar to seed oleosin. The tapetosome is closely related to the endoplasmic reticulum during anther development and is generally believed to be formed by rough endoplasmic reticulum budding (Hsieh and Huang, 2005). Similar structures have been observed in the tapetum of the Arabidopsis thaliana ecotype Columbia (Zhang et al., 2002;Suzuki et al., 2013), and these structures are called lipidosomes. The proteins in lipidosomes and sterol esters in elaiosomes are primarily used to synthesize the pollen coat. The specific function of TAG located within these structures is unclear, but it may be transported through the pollen coat by esterase or β-oxidation to the inner pollen for further metabolism and utilization (Zhang et al., 1994;Hernandez-Pinzón et al., 1999). In this study, we found that despite the fact that the lipidosomes and elaiosomes were both present in the tapetal cells of WT and as11 mutants, there were differences in timing of occurrence and their characteristics.
First, plastids in the WT began to exhibit osmiophilic deposits at the post-meiosis and pre-cytokinesis stage (Fig. 7A), and reached a steady state during the vacuolate microspore stage when the lipidosomes formed (Fig. 7E). Plastids of mutant as11 began to exhibit osmiophilic deposits at the tetrad stage (Fig. 7e), and reached a relatively steady state at the binuclear pollen grain I stage when the lipidosomes formed (Fig. 7f). Both organelles filled almost the entire tapetum in the WT and mutant at the binuclear pollen grain II stage ( Fig. 7G and g), when the content reached its maximum. At the second mitosis stage, the tapetum was degraded and these structures were released into the anther loculus where they covered the surface of pollen grains. The area ratio of lipidosomes and elaiosomes showed that the area of lipidosomes was higher in the WT tapetal cells, while the area of elaiosomes was higher in mutant tapetal cells. Throughout the entire process of pollen development, the content of endoplasmic reticulum in WT tapetal cells was generally richer than that in as11, and the lipidosomes were primarily formed by rough endoplasmic reticulum budding (Platt et al., 1998;Ting et al., 1998;Huang, 2004, 2005). The MAGL gene (a homolog of AtMAGL8) which encodes the monoacylglycerol lipase (MAGL) in Arabidopsis thaliana, results in male sterility, and the lipidosome content in the tapetum was higher than that in the mutant (Gao et al., 2019). Based on this, the lipidosomes that were rich in TAG in the mutants appeared later and were less abundant than those observed in the WT. Zheng et al. (2003) examined the ultrastructure of GPAT1 mutant anthers and found that the fusion of endoplasmic reticulum with plasma membranes occurred rarely and endoplasmic reticulum's dilation was generally decreased. The ultrastructure of the GPAT6 gene mutant anther also indicated that the content of the endoplasmic reticulum in tapetal cells was reduced and was difficult to detect in mature pollen grains (Li et al., 2012). The above results indicate that the relevant genes within the TAG synthesis pathway may also exert a certain impact on the development of the pollen. It has been reported that DGAT, an important enzyme encoded by the DGAT1 gene, catalyzes the synthesis of TAG and is located on the endoplasmic reticulum (Routaboul et al., 1999;Zou et al., 1999;Jako et al., 2001;Kaup et al., 2002;Lu and Hills, 2002;Ramli et al., 2005). Thus, it is justifiable that in mutant as11, the decreased DGAT enzyme activity resulted in a lower content of endoplasmic reticulum in tapetal cells and pollen grains.
The elaiosomes contain only small amounts of TAG, and the reduction in DGAT activity should have no significant effect on its content. However, our study found that the content of elaiosomes in the mutant was higher than that in the WT. We speculated that the mutant may supplement TAG accumulation by increasing the number of elaiosomes. Compared to the WT, the content of Golgi bodies, which is closely related to the position of the elaiosomes, was relatively rich in mutant tapetum development. Studies have shown that the Golgi bodies are closely associated with the secretion of polysaccharides in plant cells (Buvat, 1989), as their membranes possess a variety of enzymes involved in polysaccharide synthesis. We considered that in the process of as11 pollen development, due to the reduced TAG content, the SS1 and SS2 genes related to starch synthesis were clearly upregulated at the mature pollen grain stages to increase polysaccharide substances (Fig. 5B). As a result, the content of Golgi bodies that were associated with polysaccharide secretion also increased. Additionally, lipids are involved in the processing of secretory vesicles of the Golgi bodies (Kirchhausen, 2000;Shemesh et al., 2003;Weiss and Nilsson, 2003;Bethune et al., 2006). In particular, DAG is involved in the production of COP vesicles during the transport process facilitated by the Golgi-endoplasmic reticulum (Asp et al., 2009;Martinez-Alonso et al., 2013), suggesting that DAG has a specific relationship with the transport of Golgi bodies and endoplasmic reticulum. The content of endoplasmic reticulum systems in tapetal cells of the mutants was less than that in the WT, and the reduction in TAG synthesis led to an increase in DAG accumulation. The Golgi body secreted vesicles increased instead of decreased endoplasmic reticulum vesicles.
Based on our findings, we concluded that the DGAT1 gene regulates tapetum cell structure during pollen development. DGAT1 indirectly regulates the activity characteristics of the lipidosomes and elaiosomes by affecting the contents of the endoplasmic reticulum and Golgi bodies and their secretory vesicle activity, and DGAT1 ultimately affects the lipid accumulation in mature pollen grains.

Supporting Information
The following additional information is available in the online version of this article - Fig. S1. General bioinformatics analysis of all unigenes. A. GO classifications of the Venn diagram of annotated unigenes. B. Histogram of the KOG categories. Fig. S2. Anther cross sections of WT (A-F) and mutant as11 (a-f) with histochemical staining of PAS and Sudan Black at different developmental stages. A, a: premeiosis I stage, the anther wall composed of four layers, respectively epidermis, endothecium, middle layer and tapetum. The microspore mother cell has a large and distinct nucleus (arrow); B, b: premeiosis II stage, the microspore mother cells begin to deposit callose, the tapetum cells contract, and cytoplasm contains numerous vesicles (arrow). C, c: meiosis stage; C, the microspore mother cell was surrounded by callose, and the tapetum cells contracted with dense cytoplasm and a larger vesicle (arrow); c, the mother cell of the spore surrounded the callose, the tapetum cells contracted, with many small vesicles (arrow); D, d: Tetrad stage, after cytokinesis, the microspore mother cell formed four mononuclear microspores, which were enclosed and separated by callose (arrow). E, e: Released microspore I stage, callose wall was degraded and an irregularshaped microspore was released into the pollen sac; only one nucleus was located in the microspore (arrow). F, f: Released microspore I stage, the microspore became more rounded and had thick walls. Ep, epidermis; En, endothecium; ML, middle layer; T, tapetum. Fig. S3. Relative content of total fatty acids (A-C) and rate of relative content of three fatty acids (D-G) in WT and mutant as11 at different developmental stages. VM, ringvacuolate microspores; BN, binuclear pollen grains; SMD, second mitotic division; MP. Significant differences are indicated by *P < 0.05 and **P < 0.01. Three biological replicates were performed. Table S1. The primer sequences for qRT-PCR.