Candidate genes linked to QTL regions associated with fatty acid composition in oil palm

The present study searched for candidate genes in five linkage groups (LGs) - T2, T3, OT4, OT6 and T9 hosting the QTLs associated with iodine value (IV) and fatty acid composition (FAC) in an oil palm interspecific hybrid population. Each of the five LGs was successfully anchored to its corresponding chromosomal segment where, a wider repertoire of candidate genes was identified. This study further revealed a total of 19 candidate genes and four transcription factors involved in biosynthesis of fatty acids, lipids (including triacylglycerol) and acetyl-CoA, glycosylation and degradation of fatty acids. Their possible involvement in regulating the levels of saturation are discussed. In addition, 22 candidate genes located outside the QTL intervals were also identified across the interspecific hybrid genome. A total of 92 SSR markers were developed to tag the presence of these candidate genes and 50 were successfully mapped onto their respective positions on the genome. The data obtained here complements the previous studies, and collectively, these QTL-linked candidate gene markers could help breeders in more precisely selecting palms with the desired FAC.


Introduction
The oil palm industry is a major contributor to the global vegetable oils and fats market. The production of palm and palm kernel oil is at about 75.2 million tonnes accounting for almost one-third of the world's vegetable oil production (Kushairi et al. 2018). Interestingly, over 84.0% of production is in South-east Asia (USDA, Oil crops yearbook, world vegetable oils supply and distribution 2012). The high demand for palm oil reflects its position as the most consumed vegetable oil, with India, China and the European Union among the main importing countries (Index Mundi 2016; Kushairi et al. 2018). Despite concerns about sustainable practices, especially in the European Union which the industry is addressing aggressively, the supply of palm oil needs to rise in order to meet increasing demand from the growing population worldwide. Palm oil is produced and accumulates in the fruit mesocarp tissue and is referred to as crude palm oil (CPO). The physical and chemical characteristics (e.g. melting, crystallization and morphology) of CPO are mainly attributed to its fatty acid composition (FAC). In the commercial Elaeis guineensis CPO, FAC comprises a balanced combination of saturated and unsaturated fatty acids (FAs). The saturated FAs consist primarily of palmitic (C16:0;~44.0%) and stearic (C18:0;4 .5%) acids whereas the unsaturated FAs consist mainly of oleic (C18:1;~39.0%) and linoleic (C18:2;~10.0%) acids. The iodine value (IV) which measures the total level of unsaturation is on average about 53.0 in commercial materials. In comparison, CPO from E. oleifera has much higher levels of unsaturated FAs (IV: 70.0-93.0) due to high levels of C18:1 (ranges from 47.0-69.0%) and C18:2 (ranges from 2.0-22.0%). In contrast, saturated FAs in the E. oleifera CPO only range from 15.0-30.0% for C16:0 and 0.2-2.0% for C18:0 Montoya et al. 2014;Corley and Tinker 2016). Increasing the unsaturated FA levels in commercial E. guineensis CPO has advantages, especially for marketing palm oil in temperate countries. As such, conventional breeding programmes have been directed at selecting high IV oil palm. A sure way to achieve this is via interspecific hybrid breeding, where there is a desire for the unsaturated characteristic from E. oleifera to be introgressed into the commercial planting materials. Interspecific breeding crosses have been created using selected high IV E. oleifera palms (> 70.0) and Nigerian E. guineensis palms (~64.0) which appears close to the upper limit that can be achieved in pure E. guineensis materials. Early results showed an additive effect in the interspecific hybrids, which possess an intermediate level of unsaturated FAs compared to both the E. oleifera and E. guineensis parental palms (Rajanaidu et al. 1989;Rajanaidu et al. 2000;Corley and Tinker 2016).
The application of marker-assisted selection (MAS), especially in interspecific hybrid breeding programmes, can expedite the introgression of unsaturated FAs into elite E. guineensis line. In this respect, a number of quantitative trait loci (QTLs) affecting FAC have been previously identified by Singh et al. (2009), Montoya et al. (2013Montoya et al. ( , 2014 and Ting et al. (2016). Markers associated with these QTLs can help breeders in selecting palms with desired FAC, at least in specific genetic backgrounds. In addition to the identification of QTL-linked markers, efforts are also focussed on identifying the genes responsible for the variation in FAC as well as other important economic traits (e.g. yield) in oil palm (Kalyana Babu et al. 2020;Xia et al. 2019;Ting et al. 2018;2016). The availability of the oil palm genome sequence  can facilitate the identification of the genes. In the present study, the E. guineensis genome build 5 (EG5) successfully revealed a number of important genes and transcription factors (TFs) involved in biosynthesis of FAs and triacylglycerols (TAGs) within the QTL intervals, associated with FAC in an E. oleifera x E. guineensis (OxG) interspecific hybrid population (Ting et al. 2016). The authors had identified a total of 12 QTLs distributed across six linkage groups (LGs) -OT1, T2, T3, OT4, OT6 and T9 that were linked to IV, myristic acid (C14:0), C16:0, palmitoleic acid (C16:1), C18:0, C18:1 and C18:2. However, previous search for candidate genes was restricted to the QTL intervals in LGOT1. The method was efficient at revealing potential regulatory genes and as such, a similar approach was extended in the present study to mine for candidate genes from QTL intervals on the five other identified LGs. Interestingly, a number of the genes identified in the QTL intervals were similar to those described in other independent studies as regulating the synthesis of FAs and TAGs in the oil palm mesocarp (Sambanthamurthi et al. 1999;Tranbarger et al. 2011;Bourgis et al. 2011;Dussert et al. 2013;Guerin et al. 2016).

Materials and methods
The OxG mapping population The OxG mapping population is an oil palm interspecific hybrid cross between the maternal Colombian E. oleifera (coded as UP1026) and a paternal Nigerian E. guineensis tenera (coded as T128). The OxG cross consists of 108 F 1 hybrid progenies. F 1 mapping populations have been routinely utilized in genetic linkage and QTL analysis of important economic traits in oil palm as reported by Ong et al. (2019), Bai et al. (2017Bai et al. ( , 2018, Seng et al. (2016), Lee et al. (2015) and Jeennor and Volkaert (2014).
The spear leaves were sampled and stored at −80°C. The frozen leaves were ground into powder in liquid nitrogen and DNA extraction was carried out using the modified cetyltriammonium bromide (CTAB) method (Doyle and Doyle 1990). DNA purity was assessed using a NanoDrop spectrophotometer (NanoDrop Technologies Inc. DE) and an A260/A280 OD ratio of at least 1.8 was considered acceptable. Quality of the extracted DNA was further verified by comparing DNA digested with EcoRI and HaeIII with undigested DNA on a 0.9% agarose gel in 1X TPE buffer (90 mM tris-phosphate buffer and 2 mM EDTA pH 8.0) after electrophoresed at 80 -100 V for 3 h. The DNA was diluted to 50 ng/ uL for genotyping with simple sequence repeats (SSR) markers.

Mining candidate genes and development of candidate markers
Sequence information of the single nucleotide polymorphism (SNP) markers linked to QTLs for IV and FAC (reported by Ting et al. 2016) was downloaded from the publicly accessible Genomsawit database at http://genomsawit.mpob.gov.my. The QTL linked SNP markers were then mapped to the published oil palm reference genome (EG5) ) using BLASTN (Altschul et al. 1997) based on sequence similarity <1e-5. Markers linked to QTLs and candidate genes reported by Bourgis et al. (2011), Montoya et al. (2013) and Jeennor and Volkaert (2014) were also mapped to EG5. Subsequently, the genomic sequences of the entire chromosomal fragment corresponding to each QTL interval were extracted from EG5 and searched for significant homology (BLASTN and BLASTX) to known genes of interest in the National Center for Biotechnology Information (NCBI) databases (https://blast.ncbi.nlm.nih.gov/ Blast.cgi). Relevant information associated with the biological functions of the genes and TFs was obtained from published literature and the Universal Protein Resource (Uniprot) database (http://www.uniprot.org/uniprot/). Genes and TFs involved in regulation of biosynthesis of FAs and TAGs, glycolysis and other possible influential factors were shortlisted as candidates.

Candidate SSR marker analysis
The genotyping of the SSR markers on the 108 F 1 hybrid progenies and two parental palms was carried out as described previously (Ting et al. 2013(Ting et al. , 2014Ting et al. 2016). PCR amplification of each SSR marker was carried out in a 10.0 uL mixture containing genomic DNA (50 ng/uL), Fwd 5′-M13 primer (2.5 uM), Rvs primer (2.5 uM), one fluorescent dye-M13 (2.5 uM), 1X PCR buffer (NEB, USA), 2 mM of each dNTP (NEB, USA) and 0.5 U Taq DNA polymerase (NEB, USA). The PCR conditions were as follows: 95°C for denaturation (3 min); 35 cycles consisting of 95°C (30 s), 52-56°C (depending on primers, 30 s) and 72°C (30 s) and a final extension at 72°C (5 min). Prior to fragment analysis, multiplexing of four to eight PCR products was carried out, depending on the sizes of the expected amplicons. PCR fragments were analysed using capillary electrophoresis and subsequently detected using an ABI3730XL DNA analyser (Applied Biosystems, USA). Sizing and scoring of the SSR alleles were executed using the GeneMapper® 4.1 software (Applied Biosystems, USA).
The genotype profile of the markers was determined as originally described by Billotte et al. (2005). The four segregation profiles that were observed in the OxG mapping population previously (Ting et al. 2016) are illustrated in Online Resource 1: Fig. 1. For profile 1, a polymorphic locus is inherited from one of the parental palms and was scored as ab and aa for the heterozygous and homozygous genotypes, respectively, with an expected ratio of 1:1. For profile 2, polymorphism involved two common segregating alleles (observed as ab in both parents) and was scored as aa, ab and bb in the progenies with the expected ratio of 1:2:1. For polymorphisms that involved three segregating alleles (profile 3), the parental genotypes were scored as ab and ac whereas, the progenies were scored as aa, ab, ac and bc, which are expected to fit a 1:1:1:1 ratio. Finally, for profile 4, the four segregating alleles were scored as ab and cd in the two parents and were expected to segregate as ac, bc, ad and bd in the progenies, also in a 1:1:1:1 ratio.

Mapping candidate SSR markers to the OxG linkage map and QTL analysis
The F 1 interspecific hybrid population was analyzed essentially as a pseudo-testcross (Grattapaglia and Sederoff 1994). The SSR loci coded as 'cross pollinator' (CP) were incorporated into the existing data set (Ting et al. 2016) and linkage phases between the SSR alleles were determined using JoinMap® 4.1 (van Ooijen 2006). Segregation of the SSR marker alleles according to expected Mendelian ratios was evaluated using a built-in chi-square analysis and severely distorted markers (p < 0.0001) were excluded from linkage analysis. The existing OxG linkage map was used as the backbone in the Start Order tabsheet. The new SSR markers were integrated using the maximum likelihood (ML) method and markers were grouped at a recombination frequency (rf) threshold of ≤0.2. The rf between markers was transformed into map distance in centiMorgans (cM) using the Haldane mapping function. Markers with a nearest neighbour stress value >4.0 cM were discarded from each LG and the contribution of each marker to the nearest neighbour fit was also inspected in order to get the best possible map order.
The QTL analysis was performed using three separate methods, namely Interval Mapping (IM), the Multiple-QTL Model (MQM) and the Kruskal-Wallis non-parametric ranking tests where all three methods were implemented via MapQTL 6 (van Ooijen 2009). The logarithm of odds (LOD) thresholds for declaring a significant QTL at genome-wide (GW) and chromosome-wide (CW) in the IM and MQM methods were determined by permutation tests (1000 times) on the phenotypic data, also implemented via MapQTL 6. Only QTLs that were consistently observed in all three methods were considered significant in this study.

Integration of candidate gene-linked SSR markers into existing genetic map
A total of 45 FA and TAG related genes were utilized for development of SSR markers. One to five SSR markers were selected for each candidate gene, resulting in 92 SSR primerpairs being designed (Table 1). Genotyping of these SSR markers in the OxG mapping population resulted in 50 polymorphic SSR markers, of which 47 were scored according to profile 1 (45 inherited from T128 and two inherited from UP1026) and three were scored as having profile 4. The allelic segregation ratios for the 50 SSR markers met the expected Mendelian ratios at p ≥ 0.0005. These 50 markers were then included into the existing marker data set for constructing the genetic map and all were successfully mapped into the existing OxG genetic map ( Fig. 2 and Online Resource 3: Fig. 2).

Mapping of candidate gene markers to the respective QTLs
The candidate gene markers identified in this study were successfully mapped back to the respective QTL intervals in LGs T2, OT3, OT4, OT6 and T9. In LGT2, the 1.0 cM interval (61.2-62.1 cM) related to QTL for C14:0 was mapped to CHR08, but clear candidate genes related to biosynthesis of FAs or TAGs were not detected in the QTL interval. Therefore, the search was extended towards the left and right of the QTL interval and an oleoyl-ACP thioesterase (OTE/ FATA), stearoyl-ACP desaturase (SAD) and hydroxyacyl glutathione hydrolase 2 (HAGH) gene were detected flanking both sides of the interval (Fig. 2). A similar chromosomal region corresponding to the QTL for C14:0 was also reported in an interspecific BC 1 mapping population (Montoya et al. 2013), as determined by the common markers (mEgCIR3649, mEgCIR3282 and mEgCIR0800) mapped on both the studies (Fig. 2).
In LGT9, the QTLs for C14:0, C16:1 and C18:0 were located at regions spanning 17.2-32.6, 2.9-32.6 and 13.4-24.8 cM, respectively (Ting et al. 2016). The QTLs for C14:0 and C18:0 were also found to be located very close to that reported previously in a tenera x dura mapping population (Montoya et al. 2014). This was revealed by two common SSR markers, namely mEgCIR3592 and mEgCIR3787 that were located within/near the similar QTLs reported by Montoya et al. (2014) (Fig. 2). Taking the regions containing all three QTLs, an interval ranging from 2.9-32.6 cM was examined, which identified four potential genes and three TFs. The four candidate genes were beta-ketoacyl-ACP synthases II and III (KASII, KASIII), malate dehydrogenase (MDH) and acetoacetyl-CoA thiolase (AACT) whereas, the TFs were myb family PHL8 (MYB), TCP15 (TCP15) and homeobox-leucine zipper protein ATHB-13 (HD-Zip) ( Table 1). Three SSR markers, namely sPSc00554, sPSc00571 and sPSc00574, associated with the candidate genes AACT, TCP15 and HD-Zip respectively, were successfully mapped within the QTL interval. The TCP15 linked sPSc00571 was mapped closest to the QTL peak (LOD5.2-12.7) and explained 24.5-49.8% the variation for C14:0, C16:1 and C18:0. The LOD score and phenotypic variation explained (PVE) after mapping of the candidate gene markers were higher than that observed before fine-mapping (LOD4.5-10.7 and 21.9-44.2% PVE) with the candidate markers (Ting et al. 2016). Unfortunately, the SSR markers developed      Montoya et al. (2013); NCBI *Non-polymorphic markers; # QTLs reported by Ting et al. (2016) for KASII, KASIII, MDH and MYB, were not polymorphic (*), and thus could not be mapped onto LGT9. They were placed on the LG based on their relative order compared to other markers (and genes) in CHR08, but their exact map positions could not be determined (Fig. 2). On LGOT4 (CHR02), UDP-glycosyltransferase (UGT) was found located underlying the QTL peak for C18:0, defined by SNPM00121 (LOD5.0). UGTs are not involved in FA or TAG biosynthesis. They however, catalyse the covalent addition of sugars to a wide range of lipophilic molecules by transferring the glycosyl group from nucleoside diphosphateactivated sugars (e.g. UDP-sugars), and control the levels of many signalling molecules. The molecules include a broad array of hormones (including phytohormones), secondary metabolites and xenobiotics for maintaining good growth and development in plants (Ross et al. 2001;Barvkar et al. 2012;Ostrowski and Jakubowska 2014). Other genes from the QTL interval were fructose-bisphosphate aldolase (FBA), outer envelope pore protein 16-3 (OEP163), 4-coumarate-CoA ligase 1 (4CLL1) and beta-ketoacyl-ACP reductase (KAR). Among these, KAR is involved in the de novo FA chain elongation cycle and the SSR marker associated with this gene, sPSc00584A (LOD4.5) mapped closest to UGT. The PVE explained by sPSc00584A at 22.5% was similar to that observed for SNPM00121 (23.8%). In fact, another important FA gene namely stearoyl-ACP desaturase (SAD) which The QTL regions overlapping with and close to that reported previously by Montoya et al. 2013;Montoya et al. 2014), are also indicated for LGT2 (CHR08) and LGT9 (CHR13). The QTL chart including the phenotypic variation explained (PVE) at the QTL peak, is also indicated below each LG converts C18:0-to C18:1-ACP was identified on LGOT4 but, at a distance of about 36.0 cM from the QTL interval (Fig. 2).
Two putative QTLs for C18:2 were reported on LGs OT3 and OT6, at intervals 46.9-65.2 and 38.9-54.5 cM, respectively. Three candidate markers -sPSc00664, sPSc00665 and sPSc00666 associated with omega-3 fatty acid desaturase (FAD3/7/8), were developed within the QTL interval at LGOT3 (corresponded to CHR14). All three SSR markers were successfully mapped back to the QTL peak. The second gene within close proximity was acyl-CoA oxidase 4 (ACX4) which is involved in the peroxisomal degradation of shortchain FAs (C4:0-8:0) during beta-oxidation. This process also recycles acetyl-CoA as a carbon and energy source for FA synthesis and plant growth (Poirier et al. 2006;Goepfert and Poirier 2007). Other genes involved in FA and TAG synthesis activities were also detected at the QTL interval in LGOT3. These include lysophosphatidic acid acyltransferase 1 (LPAAT1), acyl-CoA-binding domain-containing protein 4 (acbd4) and glycerol-3-phosphate acyltransferase 3 (GPAT3). A candidate SSR marker, sPSc00694 was developed and mapped close to LPAAT1 whereas, SSR markers for acbd4 and GPAT3 were not polymorphic (*) and could not be mapped (Table 1 and Fig. 2). For QTL-C18:2 on LGOT6 which corresponded to CHR07, the QTL interval hosted a palmitoyl-ACP thioesterase (PATE/FATB) gene. None of the markers developed from the gene were polymorphic (*). However, two important FA genes -SAD and OTE/ FATA were identified at a distance of 4.4 cM from the QTL interval.

Discussion
The present study builds on previous efforts in searching for candidate genes in the QTL intervals on LGs T2, T9, OT3, OT4 and OT6 (Ting et al. 2016). The use of SSR markers common to those utilized in other studies revealed that several of these QTLs were located near or within the genomic regions linked to FAC in previous studies using a BC 1 and a tenera x dura mapping populations (Montoya et al. 2013(Montoya et al. , 2014. This provided confidence to search for candidate genes within the designated QTL intervals. In this study, there were 92 SSR markers developed from FA and TAG related genes, of which 50 (53.0%) were informative. The 50 SSR candidate markers followed two of the four segregation profiles observed in the OxG mapping population previously. All 50 were successfully mapped to the expected LG, corresponding to the genomic region from which they were designed, confirming the appropriateness of the mapping methodology applied in this study.
Each of the QTL intervals of interest in this study was successfully anchored to the corresponding pseudochromosomes and revealed a number of FA and lipid related genes. From the QTL regions associated with C14:0, C16:1 and C18:0 in LGT9 (CHR13), MDH and KASIII were identified. The MDH encoding enzyme can be found in a range of subcellular locations (e.g. cytosol and mitochondria) and it catalyses the interconversion of malate to oxaloacetate which subsequently can be converted to form phosphoenolpyruvate or can be oxidized to form pyruvate (Wedding 1989;Minárik et al. 2002). This provides the pyruvate source to initiate the synthesis of FAs. In vitro experiments in castor bean demonstrated high FA synthesis rate when malate was provided as a precursor (Smith et al. 1992). The enzyme KASIII forms the acetoacetyl-ACP complex from acetyl-CoA and malonyl-ACP in preparation for FA-chain elongation. Both the MDH and KASIII-catalysed reactions take place at a very early stage even before the FA-chain elongation process starts. This suggests that MDH and KASIII activities are important prior to formation of various FAs and could explain the colocalization of the two genes within the same QTL interval associated with C14:0, C16:1 and C18:0 in LGT9.
Another gene, KASII, that plays a critical step in elongating C16:0-ACP to form C18:0-ACP was also detected in LGT9. This is one of the most important enzymatic activities for generating and supplying C18:0 for subsequent desaturation into unsaturated FAs by SAD and FADs. In oil palm, KASII activity was found to be positively correlated with unsaturated FA content. The observed relationship was particularly strong with C18:1 and C18:2, suggesting that increased levels of C18:0-ACP are efficiently converted to C18:1-ACP which subsequently is hydrolysed (Sambanthamurthi et al. 1999). The C18:1 released is activated to C18:1-CoA and channelled to endoplasmic reticulum (ER) for TAG assembly or further desaturated to C18:2 prior to TAG assembly. In contrast, KASII activity was found to be negatively correlated with the saturated FAs (Sambanthamurthi et al. 1999). This was supported by the recent transcriptomic co-expression analysis in oil palm, where lower levels of C16:0 were the result of increased KASII expression (Guerin et al. 2016). It has been suggested that lower rates of KASII activity increase accumulation of shorter FA chains such as C14:0-and C16:0-ACPs. The increased accumulation of C16:0-ACP allows C16:0-ACP to be desaturated to form C16:1-ACP. Increased accumulation of C16:0-ACP also results in increased hydrolysis by PATE/FATB, activation into C16:0-CoA and assembly of higher levels of C16:0 into TAG in the ER. In A. thaliana and cotton seed, silencing or down-regulating the KASII gene has led to two-to six-fold increase in C16:0 (Pidkowich et al. 2007;Liu et al. 2017). In Camelina, suppression of the KASII gene also led to higher accumulation of palmitate and further reduction of unsaturated FAs (Hu et al. 2017). This provides support for the involvement of KASII in the QTL interval linked to C14:0, C16:1 (produced from C16:0) and C18:0. However, the SSR markers designed to the KASII gene did not segregate in the mapping family. It will be interesting to extend the analysis in future to search for polymorphic SNPs, within or flanking the KASII gene. The identification of candidate genes that are required for the initiation of FA synthesis (MDH and KASIII) and in the accumulation of unsaturated FAs (KASII) within the QTL interval in LGT9, suggests that it is an important genomic region influencing FAC in interspecific hybrids.
A number of genes encoding enzymes that show substrate specificity have also been identified in the confidence intervals of QTLs, in accordance with their respective FA preferences. These include FAD3/7/8, acbd4 and LPAAT1 that were associated with QTL for C18:2 in LGOT3 (CHR14). FAD3/7/ 8 encodes desaturase activity to convert C18:2 into C18:3 either in the plastids (by FAD7/8) or in the ER (by FAD3) (Song et al. 2004;Yurchenko et al. 2014). The acbd4 binds oleoyl (C18:1)-CoAs with high affinity and transports them from cytosol to ER for further modification of FAs or synthesis of TAGs (Leung et al. 2004;Xiao et al. 2008). Located next to acbd4 is GPAT which encodes the first step of enzymatic acylation to form TAGs in ER. Generally, GPAT is known to have preference for saturated FAs, especially towards C16:0-CoA (Griffiths et al. 1988;Griffiths and Harwood 1991;Xu et al. 2009). However, Sambanthamurthi et al. (2000), Manaf and Harwood (2000) and Dussert et al. (2013) suggested that oil palm GPAT can use both saturated and unsaturated acyl-CoAs (including C18:1-CoA) as substrates. Interestingly in Brassica napus, GPAT has a wider range of specificity, allowing addition of variety of fatty acyl-CoAs to the stereospecific number 1 (sn-1) position of glycerol-3-phosphate (Gly3P) (Larson et al. 2002). The subsequent acylation is catalysed by lysophosphatidic acid acyltransferase, an enzyme encoded by the LPAAT gene. In the current QTL interval, LPAAT1 was identified and interestingly it has been reported to show high specificity towards unsaturated fatty acyl-CoAs such as C18:1-CoA in humans and most plants (Shindou et al. 2009). In oil palm, LPAAT has also been reported to accept C16:0-CoA as the alternative substrate at the sn-2 position for producing phosphatidate (Sambanthamurthi et al. 2000). The QTL intervals essentially contain genes that regulate both FA synthesis in the plastid and TAG assembly in the ER. As such, the candidate genes and the SSR markers linked to these genes are ideal candidates to further investigate both FA and lipid biosynthesis in independent oil palm populations.
In this study, there were interesting candidate genes identified outside the QTL confidence intervals such as those for C18:0 in LGOT4 (CHR02) and C18:2 in LGOT6 (CHR07). SAD was located at a distance of about 36.0 cM from the QTL for C18:0 whereas on LGOT6, SAD and OTE/FATA were located at a position about 4.4 cM from the QTL for C18:2. SAD and OTE/FATA encode stearoyl-ACP desaturase and oleoyl-ACP thioesterases A, respectively and these two enzymes have high specificity towards C18 FA-ACPs. In the oil palm mesocarp, SAD modifies C18:0-ACP to C18:1-ACP while, OTE/FATA hydrolyses and releases C18:1 from C18:1-ACP. Detection of candidate genes outside the confidence interval has also been reported for QTLs associated with C18:1 and C18:2 in watermelon seeds (Meru and McGregor 2014). A point to consider is the observation by Raghavan and Collard (2012) that for a small mapping population (< 194 samples), there is a possibility that the QTL detected may actually be several cM away from its actual position. As such, even though the candidate genes were located outside the QTL confidence interval, they remain as good candidates for further evaluation.

Conclusion
The increasing availability of information on gene function and genome sequence data of plant species (including oil palm) that are accessible in public databases facilitated the present study to uncover potential candidate genes associated with fatty acid composition. This further facilitated development of markers closely linked to these candidate genes within the QTL confidence intervals. In this study, the candidate gene approach once again proved very efficient and was successfully applied to identify candidate genes and transcription factors from the QTL intervals. More importantly, biological functions of these candidate genes provided potential explanations for their possible involvement in the fatty acid and Kennedy pathways for lipid assembly. Both pathways play an important role in determining the levels of saturation and unsaturation in palm oil. The levels of saturation and unsaturation could possibly be regulated by the expression of these genes. More in-depth evaluation e.g. expression and functional studies will be required to confirm the regulatory effects of these candidate genes. This paper presents an atlas of candidate genes which may be involved in the oil saturation differences between the high IV E. oleifera and lower IV E. guineensis. Introgression of the high IV character into the African oil palm could lead to new markets and applications for palm oil. The current work represents an important step towards realising these objectives.