Mapping quantitative trait loci (QTLs) for fatty acid composition in an interspecific cross of oil palm

The oil palm is a perennial crop that belongs to the genus Elaeis and to the botanical family Palmae. Within the genus Elaeis, two species are distinguished, the economically important oil palm (Elaeis guineensis) originally native to Africa and the economically less important South American relative, Elaeis oleifera (which inherently has lower oil yield potential). The oil palm produces two distinct types of oil based on the fatty acid composition. The mesocarp of the fruit produces an oil (crude palm oil or CPO) which has a predominantly higher palmitic (C16:0) and oleic acid (C18:1) profile. In contrast, the endosperm (enclosed in a nut) produces oil (crude palm kernel oil or CPKO) in which the lauric fatty acids (C12:0) are predominant.

The main feature of the E. oleifera palm that distinguishes it morphologically from the commercial species E. guineensis is its procumbent trunk, distinctly smaller sized fruits and smaller canopy. Moreover, the angle of insertion of its leaflets is in a single plane as compared to a double plane for E. guineensis [1, 2]. In E. oleifera, up to 65% of the fruits tend to be parthenocarpic [1] and have a much lower oil content [3]. As such, the oil yield of E. oleifera is much lower, with oil to bunch ratio of 5%, as compared to the E. guineensis (tenera) with oil to bunch ratio of more than 25% [4]. Nevertheless, E. oleifera possess certain important characteristics that are of significant interest to oil palm breeders. This includes the low annual stem height increment (between 5 and 10 cm per year as compared to between 45 to 65 cm per year for E. guineensis) [1, 2]. The fatty acid composition of its CPO is especially of interest since its iodine value (IV, which is a measure of the degree of unsaturation of the oil) can be as high as 90 as compared to the average of 53 of E. guineensis [4]. The CPO derived from the E. oleifera oil has high levels of oleic and linoleic acid and lower levels of the palmitic acid and other saturated fatty acids, thus imparting a property quite akin to olive oil in composition. In South America, interest in the E. oleifera was driven by the fact that it shows resistance to bud rot disease [5].

In view of the apparent lack of variability for traits associated with high oil yield within E. oleifera and because E. guineensis has all the desired attributes for high oil yield, the only viable proposition (using conventional plant breeding approach) is to carry out interspecific hybridization between the two species. Fortunately, the E. guineensis and E. oleifera hybridize readily, producing fertile offspring in spite of their different areas of origin, which implies that they share a common ancestry before the two continents (Africa and South America) drifted apart some 110 million years ago. The fact that the two species can still hybridize to produce viable offspring itself suggests that the species isolation barrier is incomplete [1] despite the millions of years of separation.

The interspecific hybridization approach is viewed as a viable method to introgress the traits of interest i.e. namely higher oil unsaturation (to obtain a more liquid olein) [1, 6]. This is a long term breeding strategy, with results obtained thus far showing that oil quality, taken as unsaturated fatty acid content, is better in the hybrids and in their backcrosses than in the commercial E. guineensis [1, 7, 8]. However, the conventional breeding approach is severely hampered by the fact that being a perennial crop, the oil palm has a long selection cycle of between 10 and 12 years [9]. Furthermore, it requires enormous resources in terms of land (usually one can only plant between 136 and 148 palms per hectare), labour and field management in breeding trials. The development of marker-assisted selection (MAS) techniques would greatly facilitate hybrid-breeding programmes as well as speed up the development of planting materials with an oil composition high in unsaturated fatty acids (especially oleic fatty acid). With MAS, selection can be carried out in segregating generations of interspecific hybrids and their backcrosses more discriminately using molecular markers linked to the specific fatty acids.

For the purpose of MAS, a number of DNA marker systems have been applied to genetic mapping in oil palm. Restriction Fragment Length Polymorphism (RFLP) markers from genomic libraries have been applied to oil palm linkage mapping [10]. This map harbouring 97 RFLP markers in 24 groups of two or more was generated using a selfed guineensis cross. Moretzsohn et al. [11] reported genetic linkage mapping for a single controlled cross of oil palm using random amplified polymorphic DNA (RAPD) markers and the pseudo-testcross mapping strategy described by Grattapaglia et al. [12]. More recently, Billotte et al. [13] reported a simple sequence repeat (SSR)-based high density linkage map for oil palm, involving a cross between a thin shelled E. guineensis (tenera) palm and a thick shelled E. guineensis (dura) palm. The map consisting of 255 SSR markers and 688 amplified fragment length polymorphism (AFLP) markers represents the first linkage map for oil palm to have 16 independent linkage groups corresponding to the haploid chromosome number of 16 in oil palm [14]. Mayes et al. [10], Moretzsohn et al. [11] and Billotte et al. [13] reported the identification of RFLP, RAPD and AFLP markers respectively, linked to the shell thickness locus, an important economic trait which exhibits monofactorial inheritance. However, most of the traits of economic interest in oil palm exhibit quantitative inheritance. In this area, Rance et al. [15], expanding on the genetic map developed by Mayes et al. [10], reported the detection of QTLs associated with vegetative and yield components of oil palm. The work reported above represents important developments in the application of MAS in oil palm breeding programmes. Despite the advances being made and the progress achieved in genetic mapping of oil palm, only a limited number of economically important traits have been tagged to date. Furthermore, none has been reported for fatty acid composition. This is probably because of the lower variability for most fatty acids within the E. guineensis populations.

In this study, we hoped to exploit the use of complementary DNA (cDNA) probes as RFLP markers for linkage map construction. The cDNA clones represent gene fragments that occur in the expressed regions of the genome. Their identity can be determined via sequencing and such sequences are known as expressed sequence tags (ESTs). The usefulness of ESTs as markers has been demonstrated in several plant species [16, 17]. ESTs help to map known genes apart from providing anchor probes for comparative mapping. Furthermore, mapping ESTs closely linked to or co-segregating with a trait allows the gene for that trait to be identified by the candidate gene approach. This could eventually expedite the application of MAS in oil palm breeding programmes.

The strategy adopted in this research was to capitalize on the differences between the two species of oil palm and use an interspecific hybrid for the analysis of QTLs associated with palm oil fatty acid composition. This study employed both dominant (AFLP) and co-dominant (RFLP and SSR) markers to generate a linkage map. The map was subsequently used to locate QTLs associated with the fatty acid composition.

The FACs of palm oil and palm kernel oil render these oils applicable to both edible and non-edible uses. However, to venture into new markets, at least in the Malaysian palm oil industry, the focus is to change the oil towards a higher unsaturated FAC content particularly oleic acid (C18:1), at the expense of saturated fatty acids such as palmitic acid (C16:0) [31]. An oil with such properties has the potential to open up wider markets for palm oil in the salad oil sector, especially in the cooler climatic regions of the world where currently rapeseed, sunflower and soyabean are preferred [1]. In addition, it is envisaged that such an oil could be industrially useful for producing chemical derivatives, which could serve as alternatives to petrochemical feedstock [31] as well as be a potentially cheaper feedstock for production of biofuels by virtue of having a lower cloud point. The ultimate objective is to try to produce breeding lines that can produce oil with IV content of above 72, palmitic acid content of below 25% and oleic acid content of 60%, without sacrificing the palm oil yield per unit area. This will ensure that maximum benefit could be achieved from diversifying away from the present commercial planting material that has a higher saturated fatty acid profile and into a more liquid oil without sacrificing the inherent high oil yield potential of the crop [1, 26].

Two approaches are being taken to achieve this objective: i) genetic engineering of oil palm [31, 32] and ii) using the more conventional breeding approach of interspecific hybrid breeding. The work carried out in this study was also intended to develop probes to help expedite the latter approach, which is not complicated by issues of bio-safety and bioethics. The mapping population chosen for this purpose met two important criteria; segregating for the trait of interest and is relevant in the long term breeding scheme or strategy to improve the oil quality trait. Although the female parent (E. oleifera) was mostly homozygous for the loci analyzed, the male parent, E. guineensis fortunately was highly heterozygous, hence contributing to a significant level of genetic variability that was exploited for QTL analysis. In fact, it has been reported that the range of fatty acid composition observed in Nigerian based materials such as the male parent palm T128 used in this study, extends beyond that of the breeding materials currently in use [33]. This suggests that the Nigerian based E. guineensis materials are more suited for breeding oil palms with improved fatty acid composition [33]. The variation captured in the male parent palm T128 can also prove useful for selection within E. guineensis, which can directly affect desirable changes in fatty acid composition in hybrids created subsequently. Nevertheless, it is also acknowledged that to fully exploit the value of oil palm interspecific hybrids and to capture the variation between the two parents, backcross populations have to be analyzed in the future.

The genetic map constructed had an excess of linkage groups in relation to the haploid chromosome number despite the relatively high number of markers used. Failure to obtain the basic chromosome number despite applying high numbers of markers has also been reported for other species [34, 35]. The reason for this could perhaps be due to the relatively small sample size of the F₁ progeny used in this study. Another possible explanation is the lack of polymorphic markers in particular chromosomal regions, which could be due to the marker systems being employed and/or presence of large homozygous regions in the genome of the female E. oleifera parental palm used to create the interspecific hybrid population used in this study. Furthermore, it has to be stressed that very strict criteria were used to carry out map construction in this study. Only markers that fit extremely well in a linkage group were retained. Markers that caused even a slight friction were discarded in order not to compromise the subsequent QTL analysis. This also explains why only 252 markers (56%) were successfully ordered in the genetic map. The genetic map reported in this study depicts the mapping of expressed genes. The sequences of the RFLP markers mapped in this study have been submitted to GenBank. Since the RFLP markers were well distributed across the linkage groups, they can be used as potential anchor markers for integration or comparison of maps of different populations. As the oil palm EST database is growing rapidly [36, 37], additional probes either as RFLP markers or EST-SSR markers will be placed on the genetic map concerned. More importantly, the growing oil palm EST database will allow the selective mapping of genes associated with the fatty acid composition pathway. The use of allele specific markers linked to genes underlying the synthesis of seed oils has been demonstrated in Brassica [38].

In this study, 11 QTLs were detected for IV and the six components of the fatty acid composition (C14:0. C16:0, C16:1, C18:0, C18:1 and C18:2) in four different linkage groups. For C18:1, two QTLs were detected, one major QTL in Group 1 and a minor QTL in Group 15, which collectively explained 45.9% of the total phenotypic variation. Two QTLs were detected for C14:0 and three for C18:0, explaining 33.6% and 56.7% of the total phenotypic variation observed respectively. One QTL each was detected for IV, C16:0 and C18:2. For the first time, this study has revealed QTLs associated with FAC in oil palm. The traits were largely controlled by a limited number of genomic regions with large effects. QTLs for five traits (IV, C14:0, C16:0, C18:0 and C18:1) were located in Group 1. All traits showed similar shaped LOD profiles suggesting that the same QTL is influencing all five traits. The fact that four of the traits are significantly correlated further supports this assumption. Furthermore, looking at the pathway for fatty acid biosynthesis where C16:0 is in fact elongated to C18:0 by the enzyme β-ketoacyl ACP synthase II (KASII), and C18:0 is subsequently desaturated by Δ9-stearoyl ACP desaturase to form C18:1, supports the fact that the same locus could be influencing these traits. Also considering that IV is a measure of unsaturation of oils and fats, C18:1 is the most abundant unsaturated fatty acid while C16:0 is the most abundant saturated fatty acid in palm oil, it is only logical to assume that the same genomic region is influencing these traits in oil palm.

QTLs for C14:0, C16:1, C18:0 and C18:1 were located on Group 15. The stearoyl ACP desaturase enzyme, although highly specific to the conversion of C18:0 to C18:1, is also known to sometimes act on C16:0 as a poor substrate and convert it to C16:1 [27]. This probably explains the strong negative correlation (r = -0.734) between C18:0 (stearic acid) and C16:1 (palmitoleic acid) and why the same QTL may be influencing the traits. As expected, the effect of the QTL for C16:1 and C18:0 is also in the opposite direction. The likelihood profile for the QTLs affecting the two traits in Group 15 (Figure 4) is also very similar, adding further strength to the argument that the same locus is influencing both traits.

Previously, a single preliminary QTL was reported for IV in oil palm in a similar population consisting of only 77 palms [39]. The LOD peak of 3.1 reported is not significant at the threshold level employed in the present study. Since no other similar work especially for FAC has been reported for oil palm, it was not possible to carry out a direct comparison with findings from other research groups. However, a comparison with other crops (mainly annual crops) is possible. For example, in maize, Alrefai et al. [40] detected 15 QTLs (in eight groups) associated with C16:0 only. Similarly, Mangolin et al. [41] detected 13 QTLs distributed in eight chromosomes for kernel oil content in maize. The low number of QTLs detected in this study, were however in agreement with the work by Somers et al. [42] and Jourdren et al. [43], who found that a few QTL loci could explain a large proportion of the phenotypic variation associated with one of the fatty acid components, C18:3 (linoleic acid) in Brassica napus. Furthermore, the same genomic region influencing two or more fatty acid components have also been reported for sunflower [44] and Brassica napus [45]. However, the experience in soybean was different where Panthee et al. [46] reported lack of common markers associated with C16:0, C18:0 and C18:1, although the same genomic region appears to be influencing the three 18-carbon unsaturated fatty acids (C18:1, C18:2 and C18:3). Nevertheless, it is important to note that the differences in QTLs mapped in this research cannot be directly compared to those reported above because of the different crop, type of markers, mapping population structure and the density of the genetic maps used in the analysis. The population size employed is another major factor that may explain differences in studies on QTL analysis, as the population size can affect the power to detect QTLs. The population size for QTL analysis in this study was 81 palms, smaller than that reported for some annual crops [40, 46], which also makes direct comparison with other studies more difficult.

It is also noted that QTLs could not be detected for C18:3. The small population size employed limited power to detect QTLs of smaller effect. Analysis of further populations, particularly backcrosses derived from the same cross, may yield QTLs for this trait.

The QTLs identified in this study will provide breeders with a valuable tool to manipulate the FAC content in oil palm. For example, absence of the RFLP probe CB75A could be indicative of palms having oil with higher unsaturation level. The absence of the RFLP marker resulted in an increase of about 2.6% above the family mean for IV (level of unsaturation), and a decrease of about 6.5% below the family mean for C16:0 content (saturated fatty acid). If the marker/QTL linkage holds across different pedigrees, this RFLP marker could be used to enrich for genotypes with higher levels of unsaturation.

The association reported here was found only in a particular mapping population and as such may not yet be applicable for molecular breeding. Many researchers have pointed out that associations established in one cross may not hold true in other crosses [47, 48]. However, Grattapaglia et al. [12] were of the opinion that substantial linkage disequilibrium can be maintained for marker/traits associations established in a single cross. The linkages established however can only be defined as "confirmed linkages" once they have been confirmed in a further sample, preferably by an independent group of investigators [49]. Nevertheless, it is heartening to note that QTLs for fatty acid composition have generally been validated across populations, even those associated with minor QTLs [29].

Although the efforts in Malaysia are largely directed towards decreasing levels of saturation, increasing levels of certain saturated fatty acids can also have some economic benefits. In this respect, there is interest in increasing the stearic acid content (C18:0), which can give rise to new applications such as cocoa butter substitution and personal care products (lotions, shaving creams and rubbing oils) [32]. This is also partly motivated by the substantial price differential between cocoa butter and commodity oils [50]. Like most plant oils, the oil palm has low stearate content of less than 5% [27]. The SSR marker P4T8 could play an important role in MAS for high stearate palms. The presence of the SSR alleles resulted in an increase of 6.8% above the family mean for C18:0 content.

An important point to note is that the saturated fatty acids, e.g. C16:0 and C18:0 are negatively correlated with total unsaturation (C18:1, C18:2 and C18:3) (data not shown). Furthermore, the QTLs for saturated and unsaturated fatty acids are largely in the opposite direction. As such, it is unlikely that a particular palm for high unsaturation and C18:0 can be bred. It may be more practical to select separately palms for high saturated and unsaturated oils.

Rajanaidu et al. [26] reported that repeatability of measurements for FAC is high indicating that a single measurement is sufficient to describe the fatty acid content of a bunch. Rajanaidu et al. [26] also predicted high heritability for most of the fatty acid traits in oil palm. Arasu et al. [33] reported that genotype × environment (G × E) interaction was not detected for any of the fatty acid traits in the 40 E. guineensis Nigerian germplasm populations analyzed. As such, good repeatability, high heritability and minimum G × E interaction suggest that FAC content is actually amenable to improvement with simple selection procedures. FAC can be improved rapidly using the strategy pointed out by Hospital et al. [51]. After having established the linkages, the genetic gain can be accelerated by scoring for markers associated with the QTLs for two generations without phenotypic observation. If the marker/QTL linkage holds true, Rance et al. [15] predicted that such a strategy could reduce the generation time by almost half for oil palm as the crosses can be made right after flowering (about 3 years), without having to wait for the fruits to be formed and analyzed (which can take up to 5 years).

In this study, the QTLs were only detected for the male E. guineensis parent, T128. As such, we cannot conclude if the marker/QTL linkages will hold true for E. oleifera. Nevertheless, we believe that the linkages established between the marker and QTLs could be followed in backcross populations, which usually involve backcrossing the F₁ hybrid to the E. guineensis parent. The QTLs identified in this study would also be potentially useful in exploiting the huge E. guineensis germplasm that Malaysia (through MPOB) has accumulated. QTLs with favorable alleles can be identified in the germplasm collection for incorporation into the existing breeding programmes. The high phenotypic variation explained by most of the QTL improves confidence in their application for MAS. Nevertheless, certain drawbacks should also be pointed out. There is always a possibility of linkage drag occurring, especially when involving germplasm collections and in oil palm interspecific hybrids, where unfavorable alleles such as that responsible for low yield are also incorporated together with the favorable alleles for higher unsaturation. However, as pointed out by Rance et al. [15], this can be minimized by selecting for QTLs with small confidence interval that defines a very narrow region.