Background

Acer truncatum, which belongs to the Acer genus in the Aceaceae family, is a woody tree species native to northern China, Japan, and Korea. Today, it is cultivated globally for its ornamental and oil value [1]. Furthermore, up to 90% of the seed oil content consists of unsaturated fatty acids (UFAs) including 25.19% oleic acid (C18:1, OA), 32.97% linoleic acid (C18:2, LA), 2.76% α-linolenic acid (C18:3, ALA), 7.90% eicosenoic acid (C20:1, ESA), 16.49% erucic acid (C22:1, EA), and 5.76% nervonic acid (C24:1, NA) [2, 3]. ALA, a polyunsaturated fatty acid, is an essential human dietary nutrient, but it cannot be synthesized independently by the human body [4]. NA, a long-chain monounsaturated fatty acid, is a core component of the brain’s nervous system and has a healing effect on damaged nerve fibers [

Results

Metabolic diversity during A. truncatum seeds development

To test the FAs accumulation pattern during A. truncatum seed development, seeds at 95, 110, 125, 140, 155, and 170 days after flowering (DAF) were selected for oil content and composition measurements (Fig. 1A). A total of eight major FAs were obtained using GC-MS analysis: PA, SA, OA, LA, ALA, ESA, EA, and NA. During seed development, the total FAs content increased from 95 DAF (60.29 mg/g FW) to 140 DAF (103.45 mg/g FW) rapidly, increased slightly from 140 DAF to 170 DAF (111.22 mg/g FW), and reached the highest content at 170 DAF (Fig. 1B). Among UFAs, the contents of OA, LA, ALA, ESA, EA and NA rapidly increased from 95 DAF to 140 DAF, which were from 17.15 to 26.04, 16.99 to 27.40, 1.13 to 2.15, 5.86 to 10.45, 10.84 to 20.96 and 3.19 to 8.11 mg/g FW and slightly increased from 140 DAF to 170 DAF (29.23, 28.48, 2.29, 11.03, 22.98, and 8.46 mg/g FW), respectively. Based on the above results, we selected 95 (stage 1), 140 (stage 2) and 170 DAF (stage 3) as three key stages for subsequent omics analysis.

Fig. 1
figure 1

The FAs accumulation pattern of A. truncatum seeds at six time points during their development. (A) The developmental time points of A. truncatum seed. Six time points during A. truncatum seeds development were selected: 95, 110, 125, 140, 155, and 170 DAF. (B) Content of accumulation pattern of eight main FAs in A. truncatum seed using GC-MS during seed development. The mean values ± SD from three biological replicates (n = 3) are shown. PA, palmitic acid; SA, stearic acid; OA, oleic acid; LA, linoleic acid; ALA, α-linolenic acid; ESA, eicosenoic acid; EA, erucic acid; NA, nervonic acid. DAF, days after flowering

Full size image

In order to further understand the diversity of metabolite accumulation during seed development of A. truncatum, we conducted widely targeted metabolomic analysis from stage 1–3 seeds. A total of 966 metabolites were detected, including 40 alkaloids, 92 amino acids and derivatives, 217 flavonoids, 43 lignans and coumarins, 135 lipids, 54 nucleotides and derivatives, 78 organic acids, 164 phenolic acids, 23 tannins, 15 terpenoids, and 105 others (Fig. S1). Principal component analysis (PCA) results showed that the first and second principal components can explain 50.8% and 28.3% of the total variance, respectively, and they can distinctly distinguish stage 1–3 seed sample (Fig. S2A). Differential metabolites (DMs) between groups were selected by VIP ≥ 1 and absolute log2 fold change ≥ 1. There were 334 (84 up- and 250 down-regulated), 441 (213 up- and 228 down-regulated) and 314 (244 up- and 70 down-regulated) DMs being compared in stage 1 vs. 2, stage 1 vs. 3, and stage 2 vs. 3 seeds, respectively (Table S1). The results of kyoto encyclopedia of genes and genomes (KEGG) pathway enrichment showed that DMs were all enriched in purine metabolism and linoleic acid metabolism in the above compared groups. Moreover, DMs were both enriched in ALA metabolism in stage 1 vs. 3 and stage 2 vs. 3 (Fig. S3).

Transcriptome analysis of the develo** A. truncatum seeds

To further explore the gene expression changes involved in the biosynthesis of FAs during seed development, the transcriptome analysis was carried out at three developmental stages. The average raw and clean reads were 46,844,390 and 45,117,354, respectively. The clean Q30 base rate was higher than 94.58%, and the mapped ratio was 83.74% from transcriptome, which indicated the high-quality sequences for further analysis (Table S2). In total, 87,343 unigenes, with an average length of 1,022 bp, were obtained after filtering out low quality reads. PCA results showed that the first and second principal components can explain 32.4% and 17.1% of the total variance, respectively, and they can distinguish stage 1–3 seeds samples clearly (Fig. S2B). Differential expression genes (DEGs) between compared groups were selected by FDR < 1 and absolute log2 fold change ≥ 1. There were 6,846 (2,868 up- and 3,978 down-regulated), 16,129 (7,428 up- and 8,701 down-regulated) and 11,937 (5,997 up- and 5,940 down-regulated) DEGs in stage 1 vs. 2, stage 1 vs. 3, and stage 2 vs. 3 seeds, respectively (Table S1). The results of KEGG pathway enrichment showed that DEGs were all enriched in plant hormone signal transduction, phenylpropanoid biosynthesis, metabolic pathways, flavonoid biosynthesis, fatty acid metabolism and biosynthesis of secondary metabolites in the above comparison group. Moreover, DEGs were enriched in fatty acid degradation, fatty acid biosynthesis and biosynthesis of unsaturated fatty acids in stage 1 vs. 2 and stage 2 vs. 3 seeds (Fig. S4), which suggests that important changes in genes related to fatty acids regulation occurred during seed development of A. truncatum.

Key enzyme genes of FAD and KCS families indicated various expression profile in A. truncatum during seed development

Acetyl-CoA undergoes a series of reactions to produce C18:0-ACP, followed by desaturation to produce C18:1, C18:2 and C18:3, among which FADs played an important role [12, 31]. OA can be further extended to produce VLCFA through the action of KCS, KCR, HCD and ECR in turn. Among them, KCS is a key rate-limiting enzyme that determines the tissue and substrate specificity for fatty acid elongation (Fig. 2A).

Fig. 2
figure 2

Transcriptomic profiles of genes putatively involved in the FAs biosynthetic pathway. (A) Sketch of FAs biosynthetic pathway. Acetyl-CoA undergoes a series of reactions to produce C18:0-ACP, followed by desaturation to produce C18:1, C18:2, and C18:3. C18:1 can be further extended to produce VLCFA. Copyright permission of KEGG pathway maps (map00061) was obtained from Kanehisa laboratory by using the copyright permission request form in www.kegg.jp/kegg/kegg1.html. (B & C) Expression of (B) FADs and (C) KCSs involved in the FAs biosynthesis. The scales refer to log2FPKM. (D & E) The phylogenetic trees of (D) AtruFADs and AtFADs, and (E) AtruKCSs and AtKCSs. The phylogenetic trees of AtruFADs and AtFADs, and AtruKCSs and AtKCSs were constructed using the Maximum Likelihood method based on WAG + G and LG + G + I model respectively, within MEGA software (version X) and Evolview (http://www.evolgenius.info/evolview.html). The triangle marks the protein from A. truncatum

Full size image

In the transcriptome, we screened 13 FADs, including 3 FAD2, 3 FAD3, 1 FAD4, 4 FAD6, 1 FAD7, and 1 FAD8 (Fig. 2B & D; Table S3). Among these FADs, FAD2 had the highest expression levels, especially Cluster-7222.41384, whose FPKM increased during seed development and up to 1845.99 at stage 3. While the other FAD2 (Cluster-7222.27320 and Cluster-7222.33081) were both highly expressed at stage 1–2, they decreased at stage 3, which is consistent with the metabolic result finding LA (C18:2) increased rapidly at stage 1–2 and slowly at stage 2–3 (Figs. 1B and 2B). Moreover, the expression level of FAD6 (Cluster-7222.21727) also increased during seed development and reached the highest level at stage 3. Compared with ω6-FADs, ω3-FADs had differing expression patterns. Generally, FAD3 had higher expression levels than that of FAD7/8, in which the transcript of Cluster-7222.41455 increased from stage 1 to 2, but decreased at stage 3, which is consistent with the metabolic result finding that ALA (C18:3) raised rapidly at stage 1–2 and slowly at stage 2–3. The expression level of another FAD3 (Cluster-7222.15845) was decreased from stage 1 to 2 but increased to the highest at stage 3. However, the ω3-FADs in plastid (FAD7/8) had low expression levels during seed development and reached the highest at stage 1 and gradually decreased with seed development (Fig. 2B).

A total of 20 KCSs were identified from the transcriptome, including 1 KCS1, 1 KCS2, 3 KCS4, 2 KCS7, 1 KCS10, 6 KCS11 and 6 KCS20 (Fig. 2C & E, Table S4). Among these KCSs, KCS20 (Cluster-7222.40643) had the highest expression levels, which was highly expressed at stage 1–2, but decreased sharply at stage 3. This is consistent with the metabolic result that found ESA (C20:1), EA (C22:1) and NA (C24:1) raised rapidly at stage 1–2 and slowly at stage 2–3 (Figs. 1B and 2C). The results suggested that KCS20 (Cluster-7222.40643) is critically important in VLCFA metabolism in A. truncatum. Moreover, the expression pattern of four selected genes determined by quantitative RT-PCR (RT-qPCR) was similar to the RNA-seq results, thus indicating its reliability (Fig. S5).

DEGs demonstrated high correlation with DMs during seed development

The common enrichment pathway of DMs and DEGs was obtained by KEGG enrichment, which showed that DMs and DEGs were all enriched in ALA metabolism, biosynthesis of unsaturated fatty acids, glycerolipid metabolism, plant hormone signal transduction, zeatin biosynthesis and starch and sucrose metabolism in stage 1 vs. 2, stage 1 vs. 3, and stage 2 vs. 3 seeds (Fig. 3B & S6). The correlation between genes and metabolites was analyzed using cor function in R-analysis to calculate Pearson correlation coefficient. There were 71, 79 and 106 differential lipids in stage 1 vs. 2, stage 1 vs. 3 and stage 2 vs. 3, respectively (Fig. 3A & S7). And the number of differential lipids accounted for 21.26%, 17.91%, and 33.76% of all differentially expressed metabolites, respectively, which reflected important changes in lipids accumulation during A. truncatum seed development. Furthermore, the correlation coefficient greater than 0.80 and p-value less than 0.05 were selected for further analysis. ESA, ALA, and LA can be found with their correlated genes in stage 1 vs. 3 and stage 2 vs. 3, but just LA was correlated with genes between stage 1 and 2. The results suggested that DEGs were responsible for DMs, which is important for seed development and metabolites accumulation through the previously mentioned pathways in A. truncatum.

Fig. 3
figure 3

Correlation and KEGG pathways analysis of DMs and DEGs in stage 2 vs. 3 of A. truncatum seed. (A) Correlation heatmap of all the DMs (top) and DEGs (left) in stage 2 vs. 3 of A. truncatum seed. The red indicates the positive correlation and green indicates the negative correlation. (B) KEGG pathway enrichment showed DMs and DEGs in stage 2 vs. 3 of A. truncatum seeds

Full size image

The origin affected diversity of morphological characteristics and FAs content in A. truncatum seeds from 17 various locations

Inspecting transcriptional control on seed metabolites- especially FAs accumulation- showed that various factors, including environment, play an important role in seed traits such as phenotypes and the content of FAs. We have uncovered that the total FAs content was the highest at 170 DAF (stage 3), when the seeds are fully mature. Therefore, the seeds at stage 3 were collected to investigate the effect of environmental factors. In this study, different origins of seeds from 17 locations (L1–17, Table S5) demonstrated large diversity. Our data showed that the wing length and width of samara varied from 23.72 (L6) and 8.87 (L11), to 34.61 (L14) and 11.71 mm (L15), with an average of 27.79 and 10.02 mm, respectively. Moreover, the seed length, width, and thickness ranged from 7.39 (L5), 5.80 (L5), 2.43 (L5), to 10.36 (L14), 7.86 (L7) and 4.12 mm (L9), with an average of 9.19, 6.83 and 3.54 mm, respectively. Further, the 100-seed weight ranged from 4.87 (L4) to 14.80 g (L7), and the average weight was 10.01 g. The seed coat rate ranged from 28.41% (L9) to 49.20% (L5), and the average rate was 37.39%, indicating that the coat accounted for between 1/3 and 1/2 of the seed (Table S6, Fig. 4A).

Fig. 4
figure 4

Morphological characteristics and eight main fatty acids of the A. truncatum stage 3 seeds (170 DAF) from sample locations. (A) Representative morphology of A. truncatum seeds sample from 8 sampling locations. (B) Relative content of 8 main FAs in A. truncatum seed from 17 different sampling locations. The scale bar refers to row scale

Full size image

Similarly, the content of FAs in the A. truncatum seeds from 17 sample locations also demonstrated a wide variation range: the average content of UFAs was up to 91.42%, and all FAs were significant differences (p < 0.05). Among these FAs, ALA and NA varied from 1.39% (L6) and 4.48% (L8) to 4.03% (L10) and 8.09% (L6), with an average of 2.33% and 6.47%, respectively. LA is the most abundant form of FAs. Its content ranged from 25.49% (L1) to 32.74% (L2), with an average of 28.32%. LA is followed by OA from 23.34% (L2) to 29.42% (L14), EA from 16.26% (L8) to 20.68% (L11), and ESA from 7.77% (L2) to 10.74% (L13), with an average of 25.73%, 18.98%, and 9.53%, respectively (Table S7, Fig. 4B).

Bioclimatic factors showed significant influence on FAs content in A. truncatum seeds from 17 locations

To further explore the effect of different bioclimatic factors on seed FAs content, we selected 19 bioclimatic variables for further study (Bio1–19, Table S8). After determination of bioclimatic variables using MaxEnt model, ArcGIS and SPSS analysis, 7 bioclimatic variables were selected to build the final MaxEnt model: Bio2, 3, 10, 11, 13, 15 and 19 (Table 1). According to the results of correlation analysis, the influence of each bioclimatic factor on the content of FAs in A. truncatum seeds varied greatly (Fig. 5A). The contents of OA and LA, which are the most abundant FAs in A. truncatum seeds, were significantly influenced by Bio2 (mean diurnal range, r = -0.50) and Bio10 (mean temperature of the warmest quarter, r = -0.50), respectively. In addition, the content of ESA was also negatively correlated with Bio2 (r = -0.57) and positively correlated with Bio10 (r = 0.58). The content of PA was also positively correlated with Bio2 (r = 0.54). Moreover, Bio3 (isothermality) was the only key influence factor on the content of NA (r = -0.48), and there were no bioclimatic factors significantly correlated with SA and ALA.

Fig. 5
figure 5

The correlation analysis between bioclimatic factors and relative contents of seed FAs and potential habitats of A. truncatum. (A) The relationship between bioclimatic factors and relative contents of A. trunactum seed FAs collected from different locations. The scale bar refers to row scale. (B) Modeling environmentally suitable areas for the potential habitats of A. truncatum. The figure is based on the prediction of the maximum entropy model using MaxEnt software for species habitat modeling (v3.4.4), and the map was generated by ArcMap 10.7 (ESRI, Redlands, CA, USA, http://www.esri.com/)

Full size image

Modeling environmentally suitable areas for the potential habitats of A. truncatum

Considering A. truncatum’s importance as a diet vegetable oil, screening the potential habitats for planting and industrial production is vital. Therefore, we selected 7 bioclimatic variables based on the above analysis, combining 116 distribution locations data to construct a model for potentially ecological suitability regions using MaxEnt model (Fig. 5B). The suitable habitat areas were then divided into four classes: (i) highly suitable habitats (0.4986–0.9081), (ii) moderately suitable habitats (0.2742–0.4986), (iii) poorly suitable habitats (0.0962–0.2742), (iv) unsuitable habitats (0–0.0962). According to the map, the potential distribution area of A. truncatum is mainly located in 18 provinces/regions in China (probability > 0.2742): Inner Mongolia, Liaoning, Hebei, Bei**g, Tian**, Shandong, Henan, Anhui, Jiangsu, Shanxi, Shaanxi, Ningxia, Gansu, Hubei, Sichuan, Chongqing, Guizhou and Yunnan. Among these, Shandong, southern Hebei, Bei**g, Tian**, southern Liaoning, southern Shanxi, northern Shaanxi, Ningxia, southern Inner Mongolia, northern Henan, northern Jiangsu, northern Anhui, northern Hubei, middle Sichuan and northern Yunnan were most suitable habitat areas for A. truncatum.

The accuracy of the simulation results is shown in Fig. S8. MaxEnt model predictions performed better than that of the random one [the area under ROC curve (AUC) = 0.5] and had high accuracy with a high mean training and testing AUC value, which was 0.902 and 0.879, respectively. The relative contribution of 7 bioclimatic variables is shown in Table 1. The four main bioclimatic variables affecting the potential distribution of A. truncatum are mean temperature of the coldest quarter (Bio11, 37.5%), precipitation of the wettest month (Bio13, 26.6%), mean temperature of the warmest quarter (Bio10, 18.1%) and precipitation of the coldest quarter (Bio19, 7.9%), with cumulative contribution as high as 90.1% (Fig. 6; Table 1).

Fig. 6
figure 6

Response curves for selected 7 bioclimatic variables in the distribution model for A. truncatum. The curves show the mean response of the 10 replicate Maxent runs (red) and the mean ± SD (blue)

Full size image

Discussion

Our study firstly monitored the accumulation pattern of FAs during seed development of A. truncatum and identified 95 (stage 1), 140 (stage 2) and 170 DAF (stage 3) as three key stages, which were the important time points for FAs accumulation. Moreover, after correlation analysis between transcriptome and metabolome in A. truncatum developmental seeds, we found FAs and their regulatory genes played a key role in seed development, especially FAD3 (Cluster-7222.41455) expression level was increased from stage 1 to 2, but decreased at stage 3, which was consistent with the metabolic result of ALA. KCS20 (Cluster-7222.40643) had the highest expression levels in KCSs family, and also highly expressed at stage 1–2, but decreased sharply at stage 3, which was consistent with the metabolic result of ESA, EA and NA. Furthermore, after evaluating regional differences of morphological characteristics and FAs content in A. truncatum seeds from 17 main production areas, we found that the seed from L5 was the smallest one among 17 locations’ seeds, and ALA and NA varied from 1.39% (L6) and 4.48% (L8) to 4.03% (L10) and 8.09% (L6). After analyzing the influences of 7 bioclimatic factors on FAs content of ATSO, we found Bio3 was the only key influence factors on the content of NA, but no bioclimatic factor had significant correlation with ALA. Using MaxEnt model based on 116 distribution locations data and 7 bioclimatic variables, we found the areas of 18 provinces were more suitable habitat for A. truncatum in China and Bio11, 13, 10 and 19 were the main four bioclimatic variables affecting its potential distribution .

Correlation analysis between transcriptome and metabolome has become a new effective way in the study of color, quality, growth and development, biotic or abiotic stress and other aspects. Recently, an integrative analysis of the transcriptome and metabolome was conducted in five Camellia japonica cultivars with different petal colors, and 187 flavonoids and 634 overlap** DEGs were identified, which provides new insights for elucidating the function of anthocyanins in C. japonica petal color demonstration [24]. ETHYLENE-INSENSITIVE 3/ETHYLENE-INSENSITIVE 3-LIKEs (EIN3/EILs) are important ethylene response factors during fruit ripening. Correlation analysis of transcriptome and metabolome data revealed that EIL2 is involved in β-carotene and ascorbic acid accumulation in tomato (Solanum lycopersicum), which could provide a future in strategy genetic engineering to improve the nutritional value and quality of tomato fruit [25]. Combined transcriptomic and metabolomic analysis between the staminate and pistillate flowers of mulberry revealed that DEGs and DMs may be associated with sex differentiation of mulberry through the regulation of the enrichment pathways, such as the MAPK pathway, which provides a rich source for the analysis of the molecular mechanism of mulberry sex differentiation processes [26]. In regards to stress tolerance, Guo et al. [27]. found the combination of drought and cold stress induced transcription-associated metabolomic alterations, in which, raffinose, trehalose-6-phosphate, and proline accumulated, and monosaccharide content was increased, and meanwhile, the increased abscisic acid (ABA) content and genes involved in ABA signaling pathway were upregulated, indicating the transcriptional regulation for the metabolic changes in maize. In this study, we carried out the correlation analysis between transcriptome and widely targeted metabolome in A. truncatum for the first time using developmental seed, the DMs and DEGs were all enriched in alpha-linolenic acid metabolism and biosynthesis of unsaturated fatty acids, etc. during seeds development, which provides a potential strategy for further study to improve the nutritional value and quality of ATSO. Furthermore, other methods of associative omics analysis have been developed, such as the integrated genome, transcriptome and metabolome analysis providing important resources for the research of kiwifruit functional genes, and molecular breeding [28]. Similar method was used to reveal the molecular mechanism of russeting in sand pear (Pyrus pyrifolia Nakai.) [29], which motivates us to pursue further research in plant development, metabolism, and stress with associative omics data.

ALA is the most abundant fatty acid in the chloroplast membrane of higher plants, which can capture photons in plant thylakoid membranes and plays an essential role in plant photosynthesis. In addition, ALA is also important in the plant’s ability to withstand adverse external factors such as cold temperatures, pests, and mechanical damage. It is catalyzed by ω-3 FADs, which are mainly divided into FAD3 located in ER and FAD7/FAD8 in plastid. It is widely believed that FAD3 is mainly responsible for the desaturation of stored oil in seeds and that it is also involved in the desaturation of membrane lipids in non-photosynthetic tissues (roots and flowers). FAD7 and FAD8 are mainly involved in the desaturation of chloroplast membrane lipids, which closely relate to plant tolerance of low and high temperatures [30]. In oil crops, the results are similar, since FAD3A is mainly responsible for the ALA accumulation in the seed, while FAD7-1 and FAD7-2 contribute mostly to the ALA present in the mesocarp by gene expression measurement and lipid analysis in Olea europaea [15]. In a new type of oil crop Paeonia ostii, PoFAD3, rather than PoFAD7 or PoFAD8, plays a major role in the biosynthesis of ALA in seed [31]. In this study, we also found that FAD3 overall had higher expression levels compared to that of FAD7/8, while, AtFAD3 is closely related to A. truncatum seed ALA accumulation. However, we found 3 FAD3 in seeds with different expression patterns, but needs further research in order to differentiate them. In plants, the expression of KCS determines whether VLCFAs are synthesized in specific cells. The expression of KCS genes, strict substrate specificity, and tissue specificity are directly and closely related to the biosynthesis of NA. KCS is the rate-limiting enzyme of the NA biosynthesis pathway [32, 33]. KCS11 from Malania oleifera has a substrate preference for ESA, but not for EA to produce NA [34]. In A. truncatum, Ma et al. [17]. found that expression levels of most KCS genes were high during the early seed development and then gradually decreased. Chr4.2307.KCS, Chr4.2308.KCS and Chr4.2311.KCS, which all belong to KCS2/20 subfamily, might be mainly responsible for regulating NA biosynthesis. In this study, we also found that KCS20 (Cluster-7222.40643) had the highest expression levels, which was highly expressed at stage 1–2, but decreased sharply at stage 3. This is consistent with the VLCFA metabolism, including NA accumulation. However, the specific role of KCSs including KCS20 in the NA biosynthesis pathway in A. truncatum are still unclear and therefore needed to be further studied through enzyme activity detection and further experimentation.

Lipids in the seed of plant are crucial in controlling seed dormancy and growth, dispersal and defenses against biotic and abiotic stress [35]. Its biosynthesis and accumulation are affected by various natural environmental conditions. The kinds and contents of fatty acids are especially sensitive in colder climates [36]. In turn, the fatty acid composition may restrict the geographical distributions of plants to a significant extent [37]. Typically, higher proportions of unsaturated fatty acids help plants distribute in higher latitudes and altitudes, due to the promotion of germination earlier, and the growth being more rapid at lower temperatures [37]. Similar results were obtained in this study; we found that the content of UFAs including OA, ESA and NA had a significantly negative correlation with Bio2 or Bio3, which are related to temperature. The results provide very strong evidence to explain the content of the UFAs correlated with at least one pair of climatic condition groups. Since lipids or FAs of seeds are important for both plants and human needs, it is important to seek potential production areas for seed yields.

Species distribution models (SDMs) are ecological models that study the statistical relationship between species occurrence records and environmental factors and help determine the set of environmental conditions that are suitable for a species distribution [38, 39]. In recent years, SDMs have become an effective tool to predict the distribution patterns and potential habitat suitability assessments in many species [40,41,43]. The A. truncatum seeds at three key developmental stages were used to extract metabolites. The seed samples were freeze-dried by a vacuum freeze-dryer (Scientz-100 F) and then crushed using a mixer mill (MM 400, Retsch). The lyophilized powder sample of 100 mg was dissolved with 1.2 mL of 70% methanol/water (v/v) solution, then vortexed for 30 s every 30 min for 6 times, and finally incubated at 4°C overnight. The extracts were filtered by 0.22 μm pore size (SCAA-104, ANPEL, Shanghai, China) before Ultra-high performance liquid chromatography (UPLC)-MS/MS analysis (UPLC, SHIMADZU Nexera X2; MS, Applied Biosystems 4500 Q TRAP). The metabolites were identified through their mass-to-charge (m/z) values, retention times, and fragmentation patterns by Metware Biotechnology Co., Ltd. (Wuhan, China).

DMs and KEGG pathway enrichment analysis

DMs between groups were selected by VIP ≥ 1 and absolute log2 fold change ≥ 1. After annotating metabolites by KEGG database, the pathways with significantly varied metabolites underwent enrichment analysis, and their significance was determined by hypergeometric test’s p-values.

RNA extraction, cDNA library preparation and transcriptome assembly

The cDNA libraries were sequenced on the Illumina sequencing platform by Metware Biotechnology Co., Ltd. (Wuhan, China). Seeds of A. truncatum at three key developmental stages were used to extract RNA, and RNA integrity was evaluated using the RNA Nano 6000 Assay Kit of the Agilent Bioanalyzer 2100 system (Agilent Technologies, CA, USA). Sequencing libraries were generated using the NEBNext® UltraTM RNA Library Prep Kit for Illumina® (NEB, MA, USA), and the quality of the libraries was assessed using the Agilent Bioanalyzer 2100 system. The clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina). To obtain clean data, reads with low quality or containing adapter and poly-N were filtered out. Then, Trinity (v2.11.0) was used for transcriptome assembly.

RT-qPCR analysis

After extracting total RNA by E.Z.N.A.® Plant RNA Kit (Omega Bio-Tek, GA, USA), the cDNA templates for RT-qPCR were made with HiScript II reverse transcriptase kit (Vazyme, Nan**g, China). Then, StepOne Real-Time PCR System (Applied Biosystems, Carlsbad, USA) and the 2× M5 HiPer Realtime PCR Super mix (Mei5bio, Bei**g, China) were used for following reactions. AtruActin (ACT) was used as an internal control. The primers used in RT-qPCR are listed in Table S9.

DEGs, gene ontology (GO), and KEGG pathway enrichment analysis

The DESeq2R package DESeq2 (1.10.1) was used for determining the DEGs. All DEGs were analyzed based on the GO term; on KEGG pathways, a hypergeometric distribution test was conducted with the pathway as the unit.

Phylogenetic analyses

Phylogenetic analyses were performed as previously described [12, 17]. The phylogenetic trees of AtruFADs and AtFADs, and AtruKCSs and AtKCSs were constructed using the Maximum Likelihood method based on WAG + G and LG + G + I models, using MEGA software (vX) and Evolview (http://www.evolgenius.info/evolview.html), respectively. The analysis of AtruFADs and AtFADs, and AtruKCSs and AtKCSs was conducted with 20 and 41 amino acid sequences, respectively, and all positions with less than 80% site coverage were eliminated.

Correlation analysis between genes expression levels and metabolites content

Quantitative values of genes and metabolites in all seed samples were used for correlation analysis. Using the cor function in R, we calculated the Pearson correlation coefficient of genes and metabolites. The results of correlation coefficient greater than 0.80 and p-value less than 0.05 were selected.

Seed morphological characteristics analysis

Sixty samaras were randomly selected from each collection site. The wing length (the straight distance between the two wing tips) and wing width (the distance of parallel lines at the widest part of one wing) were measured with a vernier caliper (Ningbo Deli Tools Co., Ltd.). After the samara dried, the exocarp was removed, and the transverse diameter, longitudinal diameter and thickness of the seed were measured with vernier caliper to an accuracy of 0.01 mm. The 100-seed weight was investigated according to Chinese standards of GB/T 3543.7–1995. The seed coat rate was calculated by the following equation:

$$\eqalign{{\text{The}}\,{\text{seed}}\,{\text{coat}}\,{\text{rate}}(\% ) & = [{\text{the}}\,{\text{weight}}\,{\text{of}}\,{\text{the}}\,{\text{seed}}\,{\text{coat}}({\text{g}})] \cr & \quad /[{\text{the}}\,{\text{weight}}\,{\text{of}}\,{\text{the}}\,{\text{whole}}\,{\text{seed}}({\text{g}})] \times 100 \cr}$$

Occurrence and distribution of A. truncatum

Three types of data were used to determine the occurrence and distribution of A. truncatum: (i) the 17 sampling locations previously mentioned; (ii) the online specimen information of A. truncatum in Chinese Digital Herbarium (https://www.cvh.ac.cn/); (iii) the distribution location data of A. truncatum provided by researcher Lin Zhang from Taishan Academy of Forestry Sciences. A total of 145 distribution locations were obtained, and the latitude and longitude details were obtained from Google Map (Table S10). The spatial analysis function of ArcGIS (v10.7) was used to determine the distance between the records and the center. This ensured that each censored grid only contained the distribution point closest to the center, therefore minimizing the effects of spatial autocorrelation. Ultimately, 116 distribution locations were selected (Table S10).

Determination of bioclimatic variables and evaluation of potentially optimal habitat of A. truncatum

Nineteen bioclimatic variables (Bio1-Bio19) at spatial resolutions of 1 km were derived from the latest climate data provided by WorldClim (http://www.worldclim.org/) (Table S8). To further select the most important bioclimatic factors, the following three steps were taken, inspired by previous studies [44]. Firstly, all the 19 bioclimatic variables and 116 distribution locations were imported into the MaxEnt model (v3.4.4), which obtained the contribution rate of every bioclimatic variable (Table S8). Secondly, the attribute values of the bioclimatic variables corresponding to 116 distribution locations were extracted using ArcGIS, and the correlation between bioclimatic variables was calculated by SPSS (v26) using Pearson correlation analysis (Table S11). When r > 0.8, the contribution rates of two bioclimatic variables were compared, and variables with lower contribution rates were eliminated. After the above steps, 7 bioclimatic variables were retained to build the final MaxEnt model, which were Bio2, 3, 10, 11, 13, 15, and 19 (Table 1).

Based on the selected 116 distribution locations and 7 bioclimatic variables, MaxEnt model was used as the algorithm to calculate the percent contribution rate and construct the model for current habitat distribution of A. truncatum in China. To construct a prediction model, drawing on a previous study [44], the proportion of test data was set as ‘Random seed’, the output format was logistic, the maximum iterations was 500, and the importance of bioclimatic variables was measured by the ‘Jackknife test’. The model ran ten times, and the receiver operating characteristic curve by ROC analysis was used to confirm the accuracy of MaxEnt model’ s prediction.

Statistical analysis

Statistical significance and clustering analysis were both conducted by GraphPad Prism (v7) and SPSS. PCA was performed by statistics function prcomp within R, with the data being scaled to unit variance before PCA. Pearson correlation analysis was used to assess the relationship between fatty acids and bioclimatic factors based on the seed FAs content and 7 selected bioclimatic variables using SPSS. The correlation analysis between expression levels of four genes from transcriptome and RT-qPCR data was conducted using SPSS with the default arguments and Pearson correlation coefficient to evaluate statistical significance (**, P < 0.01).

Table 1 Selected bioclimatic variables used to predict potential optimal habitat of A. truncatum
Full size table