A total of 70 oil palm individuals were selected from the oil palm germplasm resources of the Coconut Research Institute of Chinese Academy of Tropical Agricultural Sciences, Wenchang town, Hainan province, China. Detailed information for these oil palm individuals is listed in Supplementary Table 1. DNA samples were prepared from spear leaves using the mini-CTAB method (Murray and Thompson, 1980).

Three fruits per oil palm individual (three biological replicates) were harvested, and fatty acid extraction and analyses for each mesocarp tissue were performed in triplicate (three different

extractions as technical replicates). Approximately 60 mg mesocarp was used for extracting fatty acids according to methods described by Li-Beisson et al. (2013). Fatty acid composition was subsequently examined and measured using 7890A gas chromatograph equipped with a HP-5MS column (30 m by 250 μm, 0.25 μm). The heating procedure was as follows: initial temperature 180°C, followed by a temperature increase to 220°C at a rate of 10°C per min. The contents of decanoic acid (C10:0), lauric acid (12:0), myristic acid (C14:0), tripalmitelaidin acid (16:1), palmitic acid (16:0), stearic acid (18:0), oleic acid (18:1), and linoleic acid (18:2) were determined with reference to an internal standard using the following calculation: (1) total fatty acid = (total area of all measured fatty acid peaks × quantity of heptadecanoic acid-methyl ester)/(peak area of heptadecanoic acid-methyl ester × quantity of the sample); (2) relative fatty acid percentage = peak area of a specific fatty acid /total area of all measured fatty acid peaks. The peak area was calculated using Agilent software.

The whole-genome sequence of E. guineensis was downloaded from the National Center for Biotechnology Information (NCBI). Protein sequences involved in fatty acid biosynthesis were downloaded from the Arabidopsis Information Resource (TAIR), available at http://www.arabidopsis.org. The protein sequences from Arabidopsis were used as queries for BLASTx searches against the coding sequence (CDS) database from E. guineensis to identify candidate genes involved in fatty acid biosynthesis. Conserved motifs were also predicted by alignment with Conserved Domains Database (CDD)¹ and PFAM databases.²

CDSs of candidate genes involving in fatty acid biosynthesis were used as queries for BLASTn searches against the genome sequence of E. guineensis to identify the boundary between exon and intron. The MEME program³ was used to identify the gene structures of candidate genes. Primers were designed based on the intron sequences using Primer3 software (Rozen and Skaletsky, 2000). The list of PCR primers used in this study is provided in Supplementary Table 2.

PCR amplification was performed in 10 μl reaction mixtures containing 100 ng genomic DNA, 1× PCR buffer, 2 mM MgCl₂, 1 U Taq DNA polymerase (TaKaRa, China), 0.5 μM of each primer, and 0.2 mM dNTP mix. The PCR program comprised a denaturation step for 4 min at 94°C, followed by 30 cycles of 30 s at 94°C, 30 s at 54.7°C, and 30 s at 72°C. PCR products were electrophoretically separated on 8% polyacrylamide denaturing gels and visualized by silver staining. Product sizes were determined by comparison with a 100 bp DNA ladder.

Bayesian clustering was employed to analyze the population structure of 70 oil palm individuals using the software STRUCTURE (Pritchard et al., 2000). Ten independent calculations were performed for K value (K set from 2 to 7). Burn-in time and replication number were both set to 100,000 in each run. The maximum-likelihood method was applied to assign every individual to a cluster, and the cut-off probability was set to 0.6. The most probable number of populations (K) was identified by plotting ΔK values of K from 1 to 10 in replicate runs for each K and corresponded to the peak of the ΔK graph. Association analysis was conducted using the software Tassel⁴; statistical significance (value of p) was determined by 100,000 permutations (Bradbury et al., 2007).

The protein sequences involved in de novo synthesis of fatty acids from Arabidopsis thaliana were used as queries to align against the protein database of E. guineensis. A total of 52 candidate genes that might participate in the de novo synthesis of fatty acids, carbon termination or dehydration were identified. These candidate genes were distributed across 16 different chromosomes. Among them, the maximum number of candidate genes were detected in chromosome 7, followed by chromosome 5, and then chromosome 9. Chromosomes 4 and 11 each contained only one candidate gene involved in fatty acid biosynthesis (Figure 1).

CDSs of 52 candidate genes involved in fatty acid biosynthesis were used as queries to align with the E. guineensis genome to identify the sequence boundary between intron and exon. Almost all candidate genes contained at least two intron sequences. Among all the candidate genes, the EgPLA2-2 gene contained the maximum number of introns (22). A large proportion of candidate genes contained approximately 6–7 introns (Figure 2).

A total of 205 primer pairs were designed based on intron region of candidate genes and were used to amplify the DNA templates of different 70 oil palm individuals (Figure 3). Amplification products from 64 primer pairs showed polymorphisms between oil palm individuals; the observed heterozygosity varied from 0.0286 to 0.9714, with an average of 0.4514. Meanwhile, a total of 81 alleles were identified, with an average of 2.7 alleles per locus. Sizes of the amplification products varied from 120 bp to 1,300 bp, with an average of 424 bp per PCR product. The 81 alleles were located on the intron region of 18 candidate genes involved in fatty acid biosynthesis, were ACP4, CPT, EAR, FAD3, KASII-1, LACS4-1, LACS4-2, LPAATA, DGAT1-1, PDH-E2, PDH-E1, WRI1, WRI2, FatA, LACS9, LACS4-2, LPCAT-1, and LPAATB (Table 1).

Fatty acid components of the 70 oil palm individuals were analyzed. Myristic acid content varied from 0.54 to 1.99%, with an average of 1.03%; palmitic acid content varied from 32.7 to 44.32%, with an average of 39.21%; oleic acid content varied from 35.86 to 53.42%, with an average of 46.13%; linoleic acid content varied from 7.17 to 18.07%, with an average of 11.61%; stearic acid content varied from 0.24 to 1.26%, with an average of 0.65%, and lauric acid content varied from 0.03 to 1.97%, with an average of 0.44% (Table 2; Figure 4).

The relationships between different fatty acid components were analyzed using SPSS software. Oleic acid content was negatively correlated with myristic acid content (r = −0.408^** and p = 0.00). Meanwhile, palmitic acid content showed significant negative correlation with oleic acid content (r = −0.53^** and p = 0.00), indicating that decreasing palmitic acid content in oil palm may enhance oleic acid content. Palmitic acid content and oleic acid content both showed significant negative correlation with linoleic acid content (r = −0.25^* and p = 0.04, and r = −0.549^** and p = 0.00 for palmitic acid and oleic acid, respectively). Oleic acid content showed significant negative correlation with stearic acid content (r = −0.369^** and p = 0.002), while linoleic acid content showed significant positive correlation with stearic acid content (r = 0.345^** and p = 0.003; Table 3).

The developed IP markers were used to evaluate the population structure of 70 individuals of E. guineensis. When the method of Evanno et al. (2005) was applied to identify the most likely number of ‘true population’, K = 3 genetic groups were found. The STRUCTURE assignment procedure revealed that the largest genetic group comprised 34 oil palm individuals, the second group comprised 19 oil palm individuals, and the remaining group only included 17 oil palm individuals (Figure 5).

A simple linear model (SLM) was used to identify the association between IP markers and fatty acid components. Five IP markers were significantly associated with the variation in myristic acid content, including EgLACS9-2 (p = 0.0356 and r² = 0.0643), EgPDH-E2-1-3 (p = 0.0193 and r² = 0.079), EgLACS4-2-3 (p = 0.0103 and r² = 0.0943), EgWRI1-1-2 (p = 0.0475 and r² = 0.0574) and EgLPAATB-2 (p = 0.0329 and r² = 0.0662). Three IP markers showed significant association with linoleic acid, including EgLACS4-1-2 (p = 0.0386 and r² = 0.053), EgPDH-E2-2-2 (p = 0.0286 and r² = 0.0591), and EgFATA-1(p = 0.0108 and r² = 0.0791). One IP marker, EgPDH-E2-2, showed significant association with stearic acid content (p = 0.0237 and r² = 0.0726). Furthermore, four IP markers were significantly associated with palmitic acid content, including EgPDH-E2-2 -4 (p = 0.0134 and r² = 0.0856), EgPDH-E2-2 -5 (p = 0.0261 and r² = 0.0699), EgKASII-1-2 (p = 0.0242 and r² = 0.0717), and EgKASII-1-3 (p = 0.0493 and r² = 0.055; Table 4).

To analyze the expression patterns of the candidate genes in different tissues, transcriptomic raw reads data were downloaded from the SRA (Short Read Archive) database of the NCBI website, including SSR851069 (mesocarp 10 weeks after anthesis), SRR190698 (mesocarp 15 weeks after anthesis), SRR190699 (mesocarp 17 weeks after anthesis), SRR190700 (mesocarp 19 weeks after anthesis), SRR190701 (mesocarp 21 weeks after anthesis), SRR190702 (mesocarp 23 weeks after anthesis), SSR851068 (kernel 10 weeks after anthesis), SSR851068 (kernel 15 weeks after anthesis), SSR851099 (pollen), SRR851103 (shoot), and SRR851110 (root). Results of the analysis revealed that almost all candidate genes had a high level of expression in mesocarp and kernel tissues compared with other tissues, except for the gene EgLACS4-2. Among them, EgLPAATB had higher expression in oil palm kernel compared with that in other tissues. However, seven other candidate genes had a higher expression level in mesocarp and kernel compared with the expression in three other tissues; these genes were EgKASII-1, EgPDH-E2-2, EgLACS9, EgFATA, Eglacs4-1, EgWRI1-1, and EgPDH-E2-1 (Figure 6).