Transcriptome Analysis and GC-MS Profiling of Key Fatty Acid Biosynthesis Genes in Akebia trifoliata (Thunb.) Koidz Seeds

Simple Summary Plant oil is an important renewable energy substance, and A. trifoliata seeds are of value in this regard. A. trifoliata fruits have many seeds with high oil content, but research progress on A. trifoliata seed oil is slow. Fatty acid biosynthesis is the most important factor affecting plant oil content. Therefore, analysis of the key genes for fatty acid biosynthesis is beneficial for breeding A. trifoliata varieties with high oil content. Here, we report changes in seed oil and key oil biosynthesis genes in the growth period of A. trifoliata based on transcriptome analysis. We found that the development of A. trifoliata seeds and fruits was not synchronized, and when the fruit was ripe, the seed oil content was not the highest. With the development of A. trifoliata seeds, linoleic and oleic acid content was found to decrease and increase, respectively. Subsequently, several key genes for oil biosynthesis in A. trifoliata were identified. These results further our understanding of the mechanism underlying oil biosynthesis in A. trifoliata seeds. Abstract Akebia trifoliata (Thunb.) Koidz is an important Chinese medicinal and economic crop. Its seeds, which are rich in fatty acids, are usually discarded. As of now, A. trifoliata lipid biosynthesis pathways and genes have not been clearly described. In this work, we found that seed and fruit development of A. trifoliata were not synchronized, and that when the fruit was ripe, seed oil content was not at its highest. As seeds developed, linoleic and oleic acid content was found to decrease and increase, respectively. RNA sequencing yielded 108.45 GB of clean reads from 15 cDNA libraries, containing 8756 differentially expressed genes. We identified 65 unigenes associated with lipid biosynthesis, including fatty acid and triacylglycerol biosynthesis. The 65 unigenes were mapped to the A. trifoliata lipid synthesis pathway. There were 20 AtrFAD family members in A. trifoliata, which could be divided into four sub-groups with the highest number of AtrSADs. Our study revealed the dynamic changes in A. trifoliata seed oil content and composition during its growth period and provides large-scale and comprehensive transcriptome data of A. trifoliata seeds. These findings provide a basis for the improvement of A. trifoliata seed oil yield and quality.


Introduction
With global advancements in industrialization, there is an increasing demand for fossil fuel-derived energy; however, fossil fuels are a non-renewable energy source. Solving the energy crisis and achieving sustainable development are issues that must be urgently addressed [1]. Fatty acids (FAs) are widely distributed in plants and have been considered a renewable energy source to replace petroleum [2]. Plants mostly contain C16-C20 FAs, which can serve as efficient energy sources. Plant FAs have been used in many fields,

Plant Material
A. trifoliata was cultivated in the experimental field of the Institute of Bast Fiber Crops, Chinese Academy of Agricultural Sciences (Yuanjiang, Hunan province). GD-3 was selected for study because of its lower ASO content and higher seed yield compared with other varieties in this area [12]. According to our statistics, the fruit weight of GD-3 is generally less than 200 g, and its peel is purple when ripe. GD3 maintains the biological characteristics of A. trifoliata; the fruit will crack when ripe and produces many seeds (more than 200). GD-3 has been planted for 4 years, and organic fertilizer was applied twice a year (5000 g/per plant) in March and October, respectively. The field management is in accordance with the normal A. trifoliata cultivation and management methods [18]. Flowers were simultaneously marked at the flowering stage, and fruits were harvested at 120, 135, 150, 165, 180, and 195 days after flowering (DAF) (F, S, K, T, U, and I). The seeds were separated from the pulp, cleaned with normal saline, and divided into two parts; one part was immediately frozen at −80 • C, and the other part was weighed and dried at low temperature. Three marked fruits were taken for each period. After all samples were collected, frozen samples were analyzed (F was frozen for 100 days, I was frozen for 5 days). The first sampling time was 6 July 2020, and the last sampling time was 16 September 2020. In July, the average daily temperature in Yuanjiang was 27-35 • C, and the total precipitation was 132 mm. In August and September, the average daily temperature in Yuanjiang was 25-32 • C and 21-28 • C and the total precipitation was 125 and 65 mm, respectively.

Dynamic Changes in ASO Content and FA Composition
Dried seeds (3 g) were selected from the samples in the same growth period. First, a small grinder was used to smash the seeds, and the resulting powder was filtered through a 30-mesh sieve. Then, the Soxhlet extraction method was used to determine ASO content [12]. The conditions for Soxhlet extraction were: temperature 50 • C, solid to liquid ratio 1:60, and extraction time 4 h. After the determination of ASO content, the sample was collected in a 1.5 mL centrifuge tube and stored at 4 • C. Finally, the FA composition of ASO was determined by GC-MS using an Agilent GC 7890 gas chromatograph and an Agilent 5977 mass spectrometer; helium (99.999% purity) was used as the carrier gas. The GC-MS operating conditions were as previously described by Sun et al. [19]. Before GC-MS analysis, FAs were converted to FA methyl esters (FAMEs). As described in previous studies [12,20], 0.06 g of ASO was diluted with diethyl ether/petroleum ether (1:1 v/v, 2 mL) and 0.4 M KOH-CH 3 OH (1 mL), vortexed, and maintained at room temperature (approximately 25 • C) for 2.5 h. Then, redistilled water (2 mL) was added to the mixture, which was then vortexed and centrifuged at 4500 rpm for 2 min. Finally, the organic phase containing FAMEs (100 mL) was collected and diluted with petroleum ether (900 mL).

cDNA Library Construction and Sequence Analysis and Alignment
Total RNA content was extracted from approximately 0.5 g of seed using the RNAprep Pure Plant kit (Tiangen, Beijing, China), and RNA concentration and purity were determined using a NanoDrop 2000 device (Thermo Fisher Scientific, Wilmington, DE, USA). Based on the analysis of oil content in developing A. trifoliata seeds, samples at five crucial stages (F, S, T, U, and I) were selected for transcriptomic analysis, and a total of 15 libraries were constructed for RNA-seq (each stage had three replicates). mRNA was purified from 1 µg of total RNA, fragmented, and then used to prepare a cDNA library using the NEBNext Ultra RNA Library Prep Kit (Illumina; NEB, Ipswich, MA, USA). cDNA library quality was assessed using the Agilent Bioanalyzer 2100 system (Agilent Technologies, Palo Alto, CA, USA). Illumina sequencing was performed using the HiSeq 2500 sequencing system. After the removal of reads containing poly-N and low-quality reads, the remaining clean reads were mapped to the reference A. trifoliata genome using HISAT2 or StringTie, from which unigenes were obtained [21,22].

Bioinformatic Analysis
P-Unigene expression levels were calculated as fragments per kilobase of exon model per million mapped fragments (FPKM) using the Cufflinks software package, and read counts for each gene were obtained using htseq-count. Gene expression levels in various samples were compared using the DESeq method, with p value < 0.05, fold-change > 2, or fold-change < 0.5, as thresholds indicating significant differences in gene expression [23]. The weighted gene co-expression network analysis (WGCNA) package was used to construct DEG co-expression networks [24]. A module containing at least 30 genes was constructed based on the scale-free network model. Then, an association analysis between the co-expression network and ASO, oleic acid (OA), and linoleic acid (LA) content was performed to screen for phenotype-associated modules. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses of DEGs were performed using R, based on a hypergeometric distribution. Using the HMMER 3.1b software, the hidden Markov model was constructed based on the Arabidopsis and walnut FAD gene families and their FAD proteins were downloaded from GenBank (Table S1). A. trifoliata protein sequences were obtained from our genomic database. ProtParam (https://web.expasy.org/protparam/ (accessed on 23 February 2022) was used to predict the physicochemical properties of the proteins, and Plant-mPLoc (http://www.csbio.sjtu.edu.cn/bioinf/plant-multi/# (accessed on 23 February 2022) was used to predict their subcellular localization.

Quantitative Analysis
Total RNA extraction was performed as described in Section 2.3. Approximately 0.5 µg of RNA and the PrimeScript RT Master Mix (Aidlab Biotechnologies, Co., Ltd., Changsha, China) were used to synthesize cDNA. A Bio-Rad CFX96 Touch detection system (Bio-Rad Laboratories, Richmond, CA, USA) and a SYBR Green PCR master mix (Aidlab Biotechnologies, Co., Ltd.) were used to conduct qPCR on each sample. We used EF-1α, which was found to be stably expressed in A. trifoliata [25], as an internal control gene. Primers for the qPCR experiments were designed using the Primer 5.0 software [26], and a total of nine lipid biosynthesis-related genes were analyzed. The qPCR system and procedures were developed based on the SYBR Green PCR Master Mix Kit (Aidlab Biotechnologies, Co., Ltd.) After PCR amplification, the Delta Ct method was used to analyze quantitative variations in each gene.

Statistical Analysis
Data (fruit weight, seed drying rate, ASO content, ASO composition, FPKM value, and gene expression levels) reported in the figures are averages of at least three different measurements. SAS 9.0 was used for one-way ANOVA based on Tukey's test, and different letters represent significance at p ≤ 0.05.

Dynamic Changes in A. trifoliata Fruit Weight and Seed Oil Content
At present, when A. trifoliata fruits crack naturally, they are considered mature (Figure 1A,B). In this study, 195 DAF samples were at this stage. Between periods F and K, A. trifoliata fruit weight increased by only 5.94 g; however, during this period, the seed drying rate increased from 26.81% to 49.82% ( Figure 1C). Therefore, at this stage, fruits developed slowly while seeds developed rapidly. In contrast, from 60 to 90 DAF, fruits developed rapidly (fruit weight increased from 74.55 g to 137.51 g), and seeds developed slowly (no significant change in drying rate) ( Figure 1C). At maturity, fruit weight was 203.63 g and seed-drying rate was 59.66%.  Figure 1D shows the dynamic changes in ASO content, which peaked (37.76% of seed weight) at 180 DAF; however, there was no significant change in its content between 135 and 165 DAF. There was a decreasing trend in its content between 180 and 195 DAF. ASO was found to contain 11 FA types, with the main FAs being palmitic acid (PA), stearic acid (SA), OA, and LA, which accounted for over 97% of ASO (Table 1), while the other seven FAs accounted for less than 1% of ASO. The FA types present in ASO did not change with seed development, and only their relative content changed. Among the four main FAs that constitute ASO, OA and LA exhibited the most significant changes ( Figure 1E). LA  Figure 1D shows the dynamic changes in ASO content, which peaked (37.76% of seed weight) at 180 DAF; however, there was no significant change in its content between 135 and 165 DAF. There was a decreasing trend in its content between 180 and 195 DAF. ASO was found to contain 11 FA types, with the main FAs being palmitic acid (PA), stearic acid (SA), OA, and LA, which accounted for over 97% of ASO (Table 1), while the other seven FAs accounted for less than 1% of ASO. The FA types present in ASO did not change with seed development, and only their relative content changed. Among the four main FAs that constitute ASO, OA and LA exhibited the most significant changes ( Figure 1E). LA content decreased from 37.91% to 28.17%, and with seed development, its relative content gradually decreased. OA content increased from 33.56% to 43.01%, and the trend in its change was opposite to that of LA. There was no significant change in PA content, and SA content increased from 2.62% to 4.75%. In this study, the oil content was reflected by the ratio of oil in dry seeds. Therefore, combined with the dynamic change in the seed dry rate and oil content, the accumulation of seed oil was the highest at 180 DAF.

Transcriptome Sequencing
After the removal of reads containing poly-N and those of low quality, 108.45 GB clean reads were obtained from 15 cDNA libraries. A total of 44.34, 50.44, 50.83, 46.22, and 49.79 million clean reads were generated from the F, S, T, U, and I libraries, respectively. The GC content of the clean reads was 44.00%-44.78%, and 91.91%-94.83% Q30 bases (Table S2). The clean reads were made freely available in the NCBI database (accession number: PRJNA79843).
Between 90.12% and 92.76% of clean reads were mapped to the reference A. trifoliata genome (unpublished data) (accession number: PRJNA750300), and 44,842 unigenes were identified from the transcriptome, 5399 of which were new unigenes that were not mapped to the genome. A total of 5399 new unigenes were annotated using the Basic Local Alignment Search Tool (BLAST). Searches were conducted against the NR, Swiss-Prot, GO, COG, KOG, Pfam, and KEGG databases, and 4143 new unigenes were annotated (Table S3).

Analysis of DEGs
Through pairwise comparison of samples at each time point, 8756 DEGs were identified ( Figure 2A) (Table S4). To clarify the developmental mechanism of A. trifoliata seeds, we focused on DEG trends at different stages of seed development. Through WGCNA, changes in transcriptomic data were examined. Based on the scale-free network model, the soft threshold was set to 12 ( Figure 2B), and 8756 DEGs were categorized into 10 modules ( Figure 2C). The largest module was the light green module (4210 DEGs), and the grey module constituted a collection of genes that were not assigned to other modules (six DEGs) ( Figure 2D). To better understand the relationship between the gene expression patterns of the modules and physiological traits, we conducted an association analysis. The tan ( Figure S1A The FA content in I was lower than that in U, so we also focused on evaluating t 2296 DEGs between U and I. We performed GO and KEGG pathway enrichment analys on DEGs obtained through WGCNA (three modules that were significantly related physiological traits) and those in U and I. We focused on DEGs involved in pathwa associated with the synthesis of plant oils, including FA biosynthesis, FA elongation, a triacylglycerol (TAG) biosynthesis. In these pathways, the DEGs of the three WGCN modules were found to be significantly related to FA biosynthesis ( Figure S2A), wh DEGs in U and I were found to be significantly related to FA elongation ( Figure S2 Several genes involved in FA synthesis were identified. Figure 3 shows the ASO biosy thesis process (Table 2). The FA content in I was lower than that in U, so we also focused on evaluating the 2296 DEGs between U and I. We performed GO and KEGG pathway enrichment analyses on DEGs obtained through WGCNA (three modules that were significantly related to physiological traits) and those in U and I. We focused on DEGs involved in pathways associated with the synthesis of plant oils, including FA biosynthesis, FA elongation, and triacylglycerol (TAG) biosynthesis. In these pathways, the DEGs of the three WGCNA modules were found to be significantly related to FA biosynthesis ( Figure S2A), while DEGs in U and I were found to be significantly related to FA elongation ( Figure S2B). Several genes involved in FA synthesis were identified. Figure 3 shows the ASO biosynthesis process ( Table 2).
Acetyl-CoA carboxylases (ACCases) constitute a group of FA biosynthesis rate-limiting enzymes. ACCases consist of biotin carboxylase, the biotin carboxylase carrier protein (BCCP), α-carboxyl transferase, and β-carboxyl transferase [28], and previous research has shown that any of these subunits can influence lipid content [29]. We identified four genes encoding the ACC-BCCP subunits, and Akebia trifoliata_newGene_7764 did not match the reference genome.
There are three 3-ketoacyl-ACP synthase (KAS) types in plants, and each type has a different function. KASIII catalyzes the synthesis of acetoacetyl-ACP; KASI catalyzes the synthesis of 6-16 carbon compounds; and KASII catalyzes the conversion of C16:0-ACP to C18:0-ACP [30,31]. Four genes encoding KASII and one gene encoding KASIII were identified. The expression levels of KASII and KASIII at the first stage were higher than those at the other four stages ( Table 2).

Identification of Genes Involved in Unsaturated FA Biosynthesis
Over 70% of ASO constitutes unsaturated FAs, and FAD is a key enzyme in their synthesis [32,33]. In this study, we identified four FAD genes, including two SAD and two FAD2 genes. SAD catalyzes the conversion of C18:0-ACP to C18:1-ACP, which is a key enzyme involved in the synthesis of C18:1 FAs. FAD2 catalyzes the conversion of C18:1-ACP to C18:2-ACP, which is a key enzyme involved in the biosynthesis of C18:2 FAs. Twenty-three FAD gene family members were identified in the A. trifoliata genome based on the hidden Markov model. Through Pfam domain analysis, 20 FAD family genes were ultimately obtained and numbered according to their annotation in A. trifoliata (Table 3). Subcellular localization analysis showed that these proteins were mainly located in the endoplasmic reticulum and chromosomes; AtrFAD4, AtrFAD5, and AtrFAD7 were located in the cell membrane. Aside from AtrFAD4 (298), AtrFAD5 (295), and AtrFAD8 (198), AtrFAD proteins consist of 300 to 470 amino acids. Using the neighbor-joining (NJ) method in the MEGA X software, A. trifoliata FAD protein sequences were constructed together with Arabidopsis and walnut FAD protein sequences to build a phylogenetic tree ( Figure 4A). This indicated that these three species have similar FAD gene families. There were four main FAD subfamilies in A. trifoliata, the SAD desaturase subfamily, ∆7/∆9 desaturase subfamily, ∆12/ω-3 desaturase subfamily, and the "front-end" desaturase subfamily, with most of the proteins being members of the SAD desaturase subfamily (seven). Interestingly, chromosome 7 was found to carry most AtrFADs, all of which were SADs. Four AtrFADs (AtrFAD3, AtrFAD17, AtrFAD18, and AtrFAD19) showed high expression levels (FPKM ≥ 100) ( Figure 4B) ( Table 3). The AtrFAD17 and AtrFAD18 levels observed could explain why A. trifoliata seed oil had a higher unsaturated FA content during the early stages of seed development ( Figure 4B). Subcellular localization analysis showed that these proteins were mainly located in the endoplasmic reticulum and chromosomes; AtrFAD4, AtrFAD5, and AtrFAD7 were located in the cell membrane. Aside from AtrFAD4 (298), AtrFAD5 (295), and AtrFAD8 (198), AtrFAD proteins consist of 300 to 470 amino acids. Using the neighbor-joining (NJ) method in the MEGA X software, A. trifoliata FAD protein sequences were constructed together with Arabidopsis and walnut FAD protein sequences to build a phylogenetic tree ( Figure 4A). This indicated that these three species have similar FAD gene families. There were four main FAD subfamilies in A. trifoliata, the SAD desaturase subfamily, Δ7/Δ9 desaturase subfamily, Δ12/ω-3 desaturase subfamily, and the "front-end" desaturase subfamily, with most of the proteins being members of the SAD desaturase subfamily (seven). Interestingly, chromosome 7 was found to carry most AtrFADs, all of which were SADs. Four AtrFADs (AtrFAD3, AtrFAD17, AtrFAD18, and AtrFAD19) showed high expression levels (FPKM ≥ 100) ( Figure 4B) ( Table 3). The AtrFAD17 and AtrFAD18 levels observed could explain why A. trifoliata seed oil had a higher unsaturated FA content during the early stages of seed development ( Figure 4B).

Identification of Genes Involved in TAG Biosynthesis
Glycerol-3-phosphate acyltransferase (GPAT) is the most important key enzyme involved in TAG biosynthesis, and it catalyzes the acylation of glycerol-3-phosphate (G-3-

Identification of Genes Involved in TAG Biosynthesis
Glycerol-3-phosphate acyltransferase (GPAT) is the most important key enzyme involved in TAG biosynthesis, and it catalyzes the acylation of glycerol-3-phosphate (G-3-P) sn1 [34]. In Arabidopsis, GPAT is in the endoplasmic reticulum or the plastid (ATS1), with plastid GPAT being soluble [35]. We identified six GPAT and two ATS1 types in A. trifoliata. As shown in Table 2, the expression levels of GPAT and ATS1 were different, with those of ATS1 being lower than those of GPAT.
Diacylglycerol acyltransferase (DGAT) is the rate-limiting enzyme of the TAG biosynthesis process. DGAT catalyzes the conversion of 1,2-diacylgycerol to TAG; this step is regarded as the key step in TAG synthesis by the Kennedy pathway [36]. Phospholipid:diacylglycerol acyltransferase (PDAT) is another enzyme involved in TAG synthesis [37]. It catalyzes the transfer of the FA in phosphatidylcholine to diphenol glycerol to produce lysophosphatidylcholine and TAG. Only one DGAT1 and two PDAT types were identified in A. trifoliata in this study; DGAT1 and PDAT showed similar expression trends, with their expression levels being higher from U-I (Table 2). From stages U-I, DGAT1 expression levels were higher than those of PDAT. This may also indicate that the Kennedy pathway is the main pathway for TAG biosynthesis.

qPCR Analysis of Lipid-Related Genes
Nine key genes involved in FA biosynthesis were randomly selected and evaluated using the qRT-PCR method. Figure 5A-I shows the expression levels of these genes at five different stages. Each graph shows the changes in the expression levels of each gene as determined by qRT-PCR and RNA-Seq. Figure 5 shows that the trends in expression levels as determined by qRT-PCR and RNA-Seq were highly similar, indicating that the expression data obtained by RNA-Seq was reliable.
Biology 2022, 11, x FOR PEER REVIEW 12 of 16 lysophosphatidylcholine and TAG. Only one DGAT1 and two PDAT types were identified in A. trifoliata in this study; DGAT1 and PDAT showed similar expression trends, with their expression levels being higher from U-I (Table 2). From stages U-I, DGAT1 expression levels were higher than those of PDAT. This may also indicate that the Kennedy pathway is the main pathway for TAG biosynthesis.

qPCR Analysis of Lipid-Related Genes
Nine key genes involved in FA biosynthesis were randomly selected and evaluated using the qRT-PCR method. Figure 5A-I shows the expression levels of these genes at five different stages. Each graph shows the changes in the expression levels of each gene as determined by qRT-PCR and RNA-Seq. Figure 5 shows that the trends in expression levels as determined by qRT-PCR and RNA-Seq were highly similar, indicating that the expression data obtained by RNA-Seq was reliable.

Discussion
Fossil fuels are some of the most important substances in the world but are mainly concentrated in specific geographic areas [38]. Plant oil has been considered a substitute

Discussion
Fossil fuels are some of the most important substances in the world but are mainly concentrated in specific geographic areas [38]. Plant oil has been considered a substitute for fossil fuels. Previous studies have showed that some plant oils could be used to produce biodiesel [39,40]. Although the present study showed that the highest oil content of GD-3 was 37.76%, we previously reported that the highest oil content of 130 A. trifoliata germplasms was 51.27%, with an average of 43.44% [13]. This discrepancy may be due to differences in the varieties, periods, or the main components of ASO. Nonetheless, A. trifoliata seed oil has a high application value as either biodiesel or edible oil. As research on ASO is ongoing, we should first focus on the process for producing biodiesel from ASO in the near future.
When the fruit of A. trifoliata naturally cracks, it is at the mature stage [41]. However, the present study demonstrated that 15 days before natural fruit cracking (180 DAF), the dry rate and oil content of A. trifoliata seeds were significantly higher than those in the cracking period (195 DAF). Therefore, when A. trifoliata is used as an oil crop, it should be harvested earlier. Previous studies have shown that the oil content of plant seeds is due to the dynamics between oil biosynthesis and oil degradation. When the grain is mature, its oil biosynthesis rate decreases and oil degradation increases, which would lead to a certain degree of decline in oil content in the mature and later stages [42]. However, a limitation of the present study was that we only analyzed one A. trifoliata variety and growing environment, so our conclusion that the harvesting period of A. trifoliata seeds should be earlier needs to be verified in follow-up experiments.
Plant seed oil content is a complex quantitative trait that is regulated by multiple genes. Previous reports have shown that genes such as PDHC complex, ACCase complex, KAS, FAD, DGAT, and LPAAT play important roles in the regulation of lipid content and composition [27,29,32,37]. The PDHC complex regulates the biosynthesis of acetyl-CoA, the precursor of FA synthesis. PDHC is composed of four subunits, the activities of which affect the synthesis of acetyl-CoA, which in turn affects the content of plant oil [27]. In the present study, we did not identify the PDH-E3 subunit, but after functional retrieval of all DEGs in the transcriptome, we found the PDH-E3 subunit and subunit-related genes that were not identified as ACCase. The reason for this phenomenon may be that WGCNA compresses the number of DEGs used for candidate gene mining. Fatty acid elongation 1 (FAE1) is a type of 3-ketoacyl-CoA synthase (KCS) found in higher plants. KCS catalyzes the first step of the very long chain FA (VLCFA) biosynthesis process [43,44]. Edible oils with high VLCFA content are regarded as being of poor quality [45,46]. Therefore, in many oil crop breeding programs, VLCFA content is reduced to improve the nutritional value of the oil. We only found approximately 0.7% of C20:0 and C20:1 in ASO, and this was similar to their content in 130 germplasms as determined by Zhong et al. [12]. VLCFA content in ASO was low, but we identified many genes associated with FAE1 in A. trifoliata. However, the expression level of these genes was not high (Table 2), which may explain why, even with the high number of FAE1 genes, VLCFA content was very low.

Conclusions
A. trifoliata seed and fruit development were not synchronized. Between stages A-K and T-U, A. trifoliata seeds developed rapidly but fruits developed slowly. When the fruits were ripe, their average weight was 203.63 g and seed-drying rate was 59.66%. Seeds had the highest oil content during the U period, and with further fruit ripening, the oil content and seed drying rate showed a decreasing trend. Therefore, for A. trifoliata used as an oil crop, its seeds need to be harvested in advance prior to full ripening. Relative LA content was highest during the F period and gradually decreased with seed development, but OA showed the opposite trend. As the seeds developed, relative OA and LA content changed from 33.56% to 43.01% and from 37.91% to 43.01%, respectively. In addition, there was no significant change in PA content (from 23.77% to 22.89%); the relative content of SA was less than 5%, changing from 2.62% to 4.75% (highest) as the seeds developed. RNA-Seq results showed that there were 8756 DEGs in the different comparison groups, and that between F and S, and S and T, there were only 417 and 210 DEGs, respectively. Through WGCNA analysis, the 8756 DEGs were divided into 10 different modules, of which three, which contained 2880 DEGs, were significantly related to phenotype. KEGG and GO analyses showed that these 2880 DEGs were enriched in FA-related processes and pathways. FAD gene family analysis showed that there were 20 AtrFAD family members in A. trifoliata, and these could be divided into four sub-groups. Several specific genes related to FA and TAG biosynthesis, including ACC-BCCP, PDH-E2, FAD2, and SAD were identified. These findings provide a basis for ASO development and A. trifoliata breeding.