Seeds of wild trees of the African oil palm (Elaeis guineensis Jacq.) growing on different sites of the natural palm grove in Cameroon were collected and germinated at the nursery in La Dibamba (Institute of Agricultural Research for Development, Center for Oil Palm Research, IRAD-CEREPAH; Supplementary Table 1 ). Wild trees and control trees of the Deli x La Mé elite hybrid were planted in the field in 2011 and grown under the same conditions. Fruits of ripe bunches from these trees were harvested, flash frozen and the exocarp pealed. Leaves were also harvested and flash frozen in liquid nitrogen. The frozen mesocarp and leaf material were lyophilized.

Arabidopsis thaliana wild type Col-0 and vte4-1 mutant plants (Bergmüller et al., 2003) were germinated on Murashige and Skoog medium containing 2% sucrose (Murashige and Skoog, 1962). After two weeks, plants were transferred to fresh plates, or to pots containing soil/vermiculite (2:1). Plants were grown under a 16 h light regime with light intensity of 120 µmol m^-2 s^-1 and 60% humidity.

The lyophilized mesocarp or leaf material were homogenized in 300 mM ammonium acetate with a Precellys homogenizer (Bertin, Frankfurt). After addition of internal standards (1 µg tocol, 3 µg canthaxanthin), tocochromanols and carotenes were extracted with hexane. Tocochromanols were separated on a LiChrospher diol column (Knauer, Berlin) by isocratic elution with hexane/tertiary butylmethylether (96:4, v/v) and measured with a fluorescence detector on an Agilent 1200 HPLC system (Agilent, Waldbronn) (Zbierzak et al., 2010). After a 1:10 dilution with methanol/tertiary butylmethylether (89:11), carotenes were separated on an EC Nucleoshell column (Macherey & Nagel, Düren) by isocratic elution with methanol and measured with an diode array detector using an Agilent 1100 HPLC system (Thayer and Björkman, 1990; Jeffrey et al., 1997).

Genomic DNA was extracted from lyophilzed mesocarp of Deli x La Mé and of palm C59. Exons and flanking intron sequences of the locus LOC105033221 which shows sequence similarity to Arabidopsis AtVTE4 (At1g64970) were amplified by PCR using the oligonucleotides Bn3839, Bn3840, Bn3841, Bn3842, Bn3843, Bn3844, Bn3845 and Bn3846 ( Supplementary Table 2 ), and the PCR products were sequenced by Sanger sequencing.

The plastid transit peptide of EgVTE4 was predicted using ChloroP1.1 and TargetP 2.0 (cbs.dtu.dk/services/TargetP/) (Emanuelsson et al., 1999; Almagro Armenteros et al., 2019). VTE4 protein sequences from different plant species were retrieved at UniProt (uniprot.org). Sequences were aligned using T-coffee (Notredame et al., 2000) and illustrated with boxshade (embnet.vital-it.ch/software/BOX_form.html). The domain structure of AtVTE4 was analyzed using pfam (pfam.xfam.org) (Mistry et al., 2021). The AtVTE4 sequence (At1g64970) was used for homology-modeling of the three-dimensional structure at Swiss-Model (https://swissmodel.expasy.org/) (Waterhouse et al., 2018).

For expression in E. coli, mature EgVTE4 lacking the transit peptide was amplified with Bn3902 including a 6xHis tag, and Bn3896 ( Supplementary Table 1 ) from the construct pBinGG-EgVTE4 (see below). The G/C mutation was introduced into the mature EgVTE4-W290S sequence by site directed mutagenesis. Bn3902 and Bn3898 were used to amplify the 5´ fragment, and Bn3899 and Bn3896 for the 3´ fragment from pBinGG-EgVTE4. The cDNA of mature EgVTE4 and the two fragments representing mature EgVTE4-W290S were cloned into the E. coli expression vector pTVGG by Golden Gate assembly. E. coli BL21(AI) cells harboring the empty vector (EV, pTVGG), pTV-EgVTE4 or pTV-EgVTE4-W290S constructs were grown in Terrific Broth containing kanamycin. Protein expression was induced with isopropyl β-D-1-thiogalactopyranoside (IPTG) and 0.1% arabinose, and cells cultivated at 16°C overnight.

For tocochromanol supplementation, E. coli cells were collected by centrifugation and re-suspended in Terrific Broth with kanamycin. Tocochromanols (γ-tocopherol, δ-tocopherol, γ-tocotrienol or δ-tocotrienol) dissolved in ethanol were added and the cells incubated for 3 hours at 30°C. The cultures were diluted to the same OD₆₀₀ values, and cells from the same culture volumes harvested by centrifugation. After washing with water, tocochromanols were extracted with chloroform/methanol (2:1) in the presence of tocol. After addition of 300 mM ammonium acetate and vortexing, the chloroform phase was harvested. The chloroform was evaporated under a stream of nitrogen and the tocochromanols dissolved in hexane for HPLC measurements.

Protein was extracted from E. coli cells expressing EgVTE4 or EgVTE4-W290S after re-suspension in 1.5 ml lysis buffer (10 mM HEPES pH 7.8, 5 mM dithiothreitol, 240 mM sorbitol) with 150 µl 1.5 mM phenylmethylsulfonyl fluoride (PMSF). Cells were homogenized in a Precellys homogenizer. After addition of Triton X-100 (final concentration, 0.1%), samples were homogenized and protein collected in the supernatant after centrifugation at 14,000 x g. Proteins were separated by SDS polyacrylamide gel electrophoresis and stained with Coomassie Brilliant Blue R-250, or SDS gels were blotted onto nitrocellulose membranes. The EgVTE4 and EgVTE4-W290S proteins were detected with Ni-horseradish peroxidase conjugate (Kirkegaard & Perry, Gaithersburg) and the signals visualized by chemiluminescence using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific).

The VTE4 enzyme assay contained 200 µl 125 mM Tricine-NaOH (pH 7.6), 100 µl 1.25 mM sorbitol, 20 µl 250 mM ascorbate, 15 µl 4.6 mM S-adenosyl-methionine (SAM), 10 µl 50 mM MgCl₂, 20 µl of different tocochromanols (1 mg/ml in ethanol) and 300 µg of E. coli protein (Bergmüller et al., 2003). The assays were incubated for 4 h at 25°C, and tocochromanols extracted with chloroform/methanol (2:1) in the presence of tocol. After addition of 300 mM ammonium acetate and mixing, the chloroform phase was harvested and evaporated under nitrogen. Tocochromanols were dissolved in hexane and quantified by HPLC.

The full length EgVTE4 cDNA was amplified by rt-PCR from mesocarp of Deli x La Mé using primers Bn3895 and Bn3896. The amplicon was cloned into the binary vector pTVGG carrying a 35S promoter by Golden Gate assembly. The mutation G869C at position 899 of the full-length open reading frame was introduced by site directed mutagenesis to generate EgVTE4-W290S. To this end, the oligonucleotides Bn3895 and Bn3898 were used to amplify the 5’ fragment of EgVTE4-W290S and Bn3899 and Bn3896 for the 3’ fragment. The two fragments were assembled into the vector pBinGG by Golden Gate cloning. The constructs eV (empty vector, pBinGG), pBinGG-EgVTE4 and pBinGG-EgVTE4-W290S were transferred into Agrobacterium tumefaciens GV3101-pMP90. Arabidopsis vte4-1 plants were transformed by floral dip with Agrobacterium harboring one of the constructs. Transformed seeds were selected based on their red fluorescence derived from the DsRed marker and transgenic plants grown on soil. Frozen leaves were homogenized and tocochromanols extracted with diethyl ether in the presence of tocol. After addition of 300 mM ammonium acetate, the diethyl ether phase was collected and dried under nitrogen. Tocochromanols were dissolved in hexane and measured by HPLC.

Lyophilized mesocarp slices of Deli x La Mé (palm C10) or of the wild palms C52, C57 and C59 were ground to a powder under liquid nitrogen. The samples were placed into open tubes in an incubator at 37°C under air (oxidative conditions) for 1 to 6 weeks. Control samples (T₀) were immediately collected. To terminate the incubation, the sample tubes were flushed with nitrogen gas to remove oxygen, closed with a cap, frozen in liquid nitrogen and stored at -80°C.

Aliquots of commercial palm oil (100% pure unrefined palm oil; country of origin, Sri Lanka; KTC Ltd., UK), (20 mg each) were dissolved in hexane and distributed to several microfuge tubes. Different tocochromanols (α-tocotrienol or γ-tocotrienol, dissolved in ethanol) were supplemented to each tube at a final concentration of 2 µg/mg of oil. All tubes were mixed and the solvents evaporated under nitrogen gas. T₀ samples were immediately collected. The open tubes with the oil were placed at 37°C for 3 and 6 weeks as described above. Carotenes were extracted from the mesocarp powder or from the palm oil with hexane in the presence of canthaxanthin and measured by HPLC.

Statistical analyses were carried out using Excel (Microsoft) or SigmaPlot (Systat). Tukey HSD tests or Student t-tests were performed for statistical significance, where significant differences are denoted by different letters, or by asterisks (*, P < 0.05; **, P < 0.01).

Based on the previous findings that tocochromanol composition in the CPO can considerably vary, we measured the tocochromanol composition of fruits from wild trees originating from different regions in Cameroon by HPLC ( Figure 1 ). One palm (C59) was identified with a unique tocochramanol profile. Palm C59 is deficient in α-tocopherol and α-tocotrienol, while it contains elevated amounts of γ-tocotrienol and δ-tocotrienol, compared with Deli x La Mé control and two other wild trees from Cameroon ( Figure 1 ). The total tocochromanol contents of the mesocarp of the Deli x La Mé hybrid (C10) and of the wild palms (C52, C57 and C59) were similar. In addition, tocochromanols were measured in leaves of the four palm trees. In contrast to mesocarp, leaves of Deli x La Mé are devoid of tocotrienols and accumulate mostly α-tocopherol. However, the α-tocopehrol content in leaves of plant C59 was very low, while at the same time, γ-tocopherol accumulated, in line with the results observed in the mesocarp.

Because of the accumulation of γ forms and the lack of α forms of tocochromanols, the palm tree C59 presumably harbors decreased VTE4 activity, and it was possible that the VTE4 gene in C59 carries a SNP compared with Deli x La Mé. The BLAST search with the genomic sequence of Arabidopsis AtVTE4 (At1g64970) in the genome of African oil palm at NCBI resulted in the identification of one locus, LOC105033221 (Ong et al., 2020). Two further loci, LOC105042568 and LOC10504611, that were previously designated EgTMT-2 and EgTMT-3 (Dou et al., 2021), displayed lower similarity with AtVTE4 and are likely not involved in tocopherol synthesis. Two transcripts have been predicted for the locus LOC105033221 which originate from alternative splicing (www.ncbi.nlm.nih.gov). Transcript X1 (XM_010907927) is derived from exons 1a, 2, 3, 4, 5 and 6, while transcript X2 (XM_010907928) is derived from exons 1b, 2, 3, 4, 5 and 6 ( Figure 2A ). The intron between exons 1a and 2 is very large (> 18 kbp), while the intron between exon 1b and 2 is ~3.5 kbp.

The subcellular localization of the two polypeptide sequences X1 and X2 was studied using the prediction tools ChloroP and TargetP. While the polypeptide X1 is predicted to be localized in the chloroplast with a chloroplast targeting sequence of 48 amino acids, the polypeptide X2 was not predicted to be chloroplast localized. Because the EgVTE4 protein likely resides in the chloroplast, the sequence X1 was considered to be the most relevant VTE4 transcript in oil palm.

Exons and flanking intron regions of locus LOC105033221 of the Deli x La Mé hybrid and of palm C59 were amplified by PCR and sequenced. We identified four single nucleotide polymorphisms (SNPs) in the coding region between palm C59 and the Deli x La Mé/published sequence. The SNPs G352A and C407T in exon 1a result in amino acid exchanges of A31T and A49V of the polypeptide X1, while the SNP G20044A in exon 1b causes an R28Q amino acid exchange in the polypeptide X2. In addition, a G25212C exchange was detected in exon 5, which gives rise to an amino acid exchange of W290S in polypeptide X1 (or W274S in polypeptide X2) ( Figure 2A ). The amino acid exchanges of A31T and A49V in polypeptide X1, as well as R28Q in polypeptide X2 reside within the chloroplast transit sequence of X1, or at the N-terminal region of X2, and therefore presumably do not affect enzymatic activity. In addition, these amino acid exchanges do not affect the prediction for subcellular localization of X1.

The published VTE4 sequence of African oil palm, of Deli x La Mé and palm C59 were aligned with VTE4 sequences from other monocotyledoneous and dicotyledoneous plants, algae and cyanobacteria ( Figure 2B , Supplementary Figure 1 ). The comparison of amino acid sequences revealed that the tryptophan at position 290 of EgVTE4 is absolutely conserved in the VTE4 sequences from oil palm and all other species. Therefore, the amino acid exchange W290S was a strong candidate for the cause of the deficiency in VTE4 activity in palm tree C59.

Analysis of the domain structures of the Arabidopsis AtVTE4 protein (Q9ZSK1) at EMBL (http://pfam.xfam.org/) revealed the presence of a conserved γ-tocopherol methyltransferase region (accession PLN02244) which includes the S-adenosyl-methionine (SAM)-dependent methyltransferase domain (pfam08241) at positions 131-229. For three-dimensional modeling of the structure, we used the AtVTE4 sequence to search for related sequences at the Swiss-Model web page. The geranyl-diphosphate 2-C-methyltransferase (GPP-MT) was retrieved as the most similar sequence with identity of 23.5%. This enzyme catalyzes the methylation of geranyl-diphosphate at the C2 position and is involved in the synthesis of the terpenoid 2-methylisoborneol in Streptomyces coelicolor (Köksal et al., 2012). The structure of GPP-MT has been elucidated with its substrate and cofactors, i.e. Mg²⁺, S-adenosyl-L-homocysteine (methyl donor) and geranyl-diphosphate at 2.05 Å resolution (PDB file 3vc2.1.pdb). The GPP-MT structure was used for the modeling of the AtVTE4 sequence using Swiss-Model ( Figure 3A ). The green and turquoise sequence parts correspond to the methyltransferase domain (pfam08241). The position of W295 of AtVTE4 (equivalent to W290 in EgVTE4) is located in another α-helix outside of this domain. The two structures of GPP-MT (3vc2.1.pdb) and AtVTE4 (model_01.pdb) were overlaid using the structure comparison module of Swiss-Model. Figure 3B shows that the tryptophan at position 295 of AtVTE4 and the corresponding tryptophan of GPP-MT localize in the same α-helix, and that they extend into a cavity close to the substrate of GPP-MT, i.e. geranyl-diphosphate. Therefore, it is likely that the tryptophan 295 in AtVTE4 is crucial to mediate substrate binding for methylation.

To unravel whether the polymorphism of EgVTE4-W290S in palm C59 results in a loss of VTE4 activity, the cDNA of the mature form of the EgVTE4 sequence X1 lacking the predicted transit peptide was obtained from Deli x La Mé. A mutant version of EgVTE4, EgVTE4-W290S, carrying the G-to-C polymorphism was created by site directed mutagenesis. The two sequences were expressed in E. coli cells. Separation by SDS polyacrylamide gel electrophoresis and Western blotting revealed that the two proteins EgVTE4 and EgVTE4-W290S were expressed at similar levels with sizes of ~34 kDa close to the calculated sizes (33.1 kDa) ( Supplementary Figure 2 ). Therefore, the amino acid exchange W290S does not interfere with polypeptide accumulation in E. coli. First, E. coli cells harboring the EgVTE4 or EgVTE4-W290S constructs were supplemented with δ-tocopherol, γ-tocopherol, δ-tocotrienol or γ-tocotrienol after induction of expression. Tocochromanols were extracted and measured by HPLC. While a large proportion of the γ- and δ-tocochromanols were methylated and thereby converted into the corresponding α- and β-tocochromanols in the EgVTE4 expressing cells, no conversion was observed with the EgVTE4-W290S expressing cells ( Figure 4A ). Next, proteins from E. coli cells expressing EgVTE4 or EgVTE4-W290S were employed for in vitro enzyme assays with SAM and different tocochromanols. The methylation activity of EgVTE4 was highest with γ-tocotrienol and γ-tocopherol, and it was similar for δ-tocopherol and δ-tocotrienol ( Figure 4B ). The EgVTE4-W290S protein was inactive with the different tocochromanols. Therefore, EgVTE4 shows high activity with γ-tocotrienol and γ-tocopherol, but EgVTE4-W290S was inactive, although the two proteins were expressed in E. coli at similar levels.

To study the impact of the W to S exchange on the VTE4 activity in planta, the full length EgVTE4 and EgVTE4-W290S sequences including the transit peptides were introduced into the Arabidopsis vte4-1 mutant that is deficient in α-tocopherol but accumulates γ-tocopherol in the leaves (Bergmüller et al., 2003) ( Figure 4C ). The tocochromanol composition in the leaves of the empty vector control (vte4+eV) and vte4+EgVTE4-W290S plants was similar to the non-transformed vte4-1 plants. Introduction of EgVTE4 into the vte4-1 plants resulted in the complementation of α-tocopherol deficiency in the leaves. Therefore, the EgVTE4 protein from Deli x La Mé was capable of converting γ-tocopherol into α-tocopherol in the Arabidopsis vte4-1 mutant, but the EgVTE4-W290S protein was inactive. Taken together, expression in E. coli and in the Arabidopsis vte4-1 mutant revealed that the W290S substitution in EgVTE4-W290S leads to a non-functional enzyme and therefore is causal for the loss of α-tocopherol and α-tocotrienol in palm tree C59.

To unravel whether the over-accumulation of γ-tocochromanols in palm tree C59 compared with Deli x La Mé contributes to higher oxidative stability of carotenes, the dynamics of carotene degradation were studied. The oxidative stability in mesocarp from palm C59 was compared with other palms (C10, Deli x La Mé; wild lines, C52, C57) which contain similar amounts of tocochromanols, but accumulate α-tocochromanols instead of γ-tocochromanols ( Figure 1 ). Homogenized mesocarp samples were incubated under oxidative conditions in open tubes under air for 0, 3 or 6 weeks, followed by carotene analysis by HPLC. After incubation, the contents of α-carotene and β-carotene were declined in all samples ( Figure 5A ). However, the contents of carotenes were more strongly decreased to only 5 to 30% in the three control samples (C10, C52, C57) while carotene amounts stayed elevated in palm C59 (~40% of the content at day 0), indicating that carotene oxidation was mitigated in the mesocarp of palm tree C59.

Tocochromanols have been described to provide prooxidant activities under certain conditions, in addition to antioxidant activities. Thus, the attenuation of carotene degradation in palm C59 could be due to an increased antioxidant activity of γ-tocotrienol or a reduced prooxidant activity caused by the loss of α-tocotrienol. Therefore, it was important to study the effects of α-tocochromanols and γ-tocochromanols on oxidative stability of the carotenes in palm oil. Commercial palm oil was supplemented with α-tocotrienol or γ-tocotrienol and incubated under oxidative conditions as described above. The supplementation with either of the two tocochromanols efficiently attenuated the degradation of carotene ( Figure 5B ). The γ-tocotrienol supplementation was even more effective than α-tocotrienol in mitigating carotene degradation. Therefore, γ-tocotrienol provides higher antioxidant activity to protect carotenes from oxidation compared with α-tocotrienol. Taken together, these results indicate that the increased oxidative stability of carotenes in palm C59 is due to the superior antioxidant capacity of γ-tocotrienol.