Oil palm interspecific hybrid breeding worldwide aims to develop compact palms with higher unsaturated oil without sacrificing yield. Higher yield can then be obtained by planting more of the palms per unit land area [29]. The breeding efficiency can be improved by linking markers to the desired traits and accumulating favorable alleles for them. This is a far better approach than directly targeting the yield parameters which are generally under polygenic control and strongly influenced by the environment, making it quite a difficult forest to navigate through with the current technology [30].
As expected, all the traits examined had values between E. oleifera and E. guineensis, similar to the observations in other studies [16,31]. This adds confidence to the phenotypes observed. The wide distribution for all the traits measured suggests that both the BC2 populations are ideal for QTL mapping and selection and improvement in oil palm. Rachis length and height increment in both populations were considerably lower than in commercial DxP [27], signifying their potential in the development of compact palms.
Correlations between the traits are important information for breeding. The correlations between the different FAC parameters corroborate those from other studies [14,16,20], and suggest, as may be expected, that any increase in the oil unsaturation (IV) will reduce the saturated fatty acids, namely, C16:0 and C18:0 content. The inverse relationship between the two most abundant unsaturated fatty acids C18:1 and C18:2 content further implies that any increase in C18:1 content will likely overflow to C18:2 content. Similarly, the positive correlations of petiole cross section with rachis length and height increment indicate that reducing petiole cross section will give shorter and more compact palms. This fits with the assumption that a larger petiole cross section supports a larger/longer rachis. Interestingly, and as also observed in a previous study [32], there was generally no correlation between the vegetative measurements and FAC components. The exceptions were the positive correlations between height increment and petiole cross section with C18:2 content, and between height increment and iodine value. The latter correlation between height increment and iodine value was likely caused by C18:2 content which strongly impacts iodine value, especially in interspecific hybrids. [32] did not observe any correlation between iodine value and C18:2 content with height in E. guineensis, although [15] had a positive correlation between iodine value and percentage pulp (a yield parameter) in selected E. guineensis families. Interestingly, in Jatorpha, [33] found a positive correlation between C18:2 content and seed weight. The results in this study generally indicate that breeding for compactness in oil palm interspecific hybrids will not affect most of the fatty acid contents, although the relationships between C18:2 content and some of the compactness parameters need to be further investigated. Interestingly, C18:2 differs from the other fatty acids in that it is assembled in the endoplasmic reticulum, while the rest (C16:0, C18:0 and C18:1) are assembled in the plastid [34].
The number of palms in an individual breeding trial are almost always smaller than in other crops, such being the quid pro quo for the size of the palm. The small number could have caused some of the QTLs to be missed and only allowed those with the most prominent effects to be detected. [35] found that the number of QTLs detected increased with the population size. Other factors, such as the phenotypic measurement accuracy and marker density, also contribute to QTL detection and localization [36]. To obviate the limitation of the small population, the map resolution and, hence, QTL detection power was improved by integrating the two individual BC2 genetic maps. This strategy was recommended by [16] for increasing the detection of QTLs in oil palm interspecific hybrids, building on the successful approach by [15] for E. guineensis. A total 149 SSR and 1814 SNP markers (1,963 altogether) that met the expected segregation ratios and had a near-neighbor stress value below 3cM, generated 16 LGs in the consensus map, consistent with the 16 chromosome-pairs in oil palm [37]. The genome length (1,793 cM) was close to those reported by [14,17] of 1,815 – 1,867 cM for E. guineensis. The average length of the LGs was 112 cM, in the range of most agricultural crops [38]. More importantly, there was generally a high collinearity among the common markers in the individual and consensus maps (Additional file 5). Some minor discrepancies in the marker order were to be expected due to differences in calculating the independent genetic maps [39]. The genome coverage of the integrated consensus map was much better than those by the independent maps in terms of the number of markers mapped. The average gap observed on the consensus map, 0.91 cM, was also much smaller than those previously reported on oil palm interspecific hybrids of 1.2 - 7.2 cM [14,16,17]. This was as expected, since the use of multiple parents increases the chances of identifying polymorphic markers at a specific genomic region, which, in turn, increases the effectiveness of uncovering the marker-trait association. The fact that the largest gap between markers did not exceed 20 cM further suggests that the markers were well distributed across the 16 linkage groups. The selfing of palm T128 to generate one of the two female parents used in this study (Figure 4) also likely did not result in large homozygous blocks. The reason is because oil palm, being an outcrossing species, has high heterozygosity in its genome, especially palms like T128, which was from the wild in Nigeria [40].
To ensure the robustness of the genomic region linked to the traits, two independent QTL analyses (Genstat and MapQTL) were done, and only the QTLs detected by both were considered significant. Generally, a majority of the QTLs were detected by both methods, and they could be classified as major or minor. In the vegetative traits, the QTLs associated with petiole cross section and height increment were located at the same genomic region on LG4, in the independent genetic maps of both populations and consensus map, likely representing a major and stable locus influencing compactness. A major QTL peak associated with C18:2 content was located very close to the QTLs for petiole cross section and height increment (~1 cm away). Co-localization of the QTLs for different traits was not surprising, especially if the traits are correlated, and suggest pleiotropy with the genomic region containing a number of genes that influence several traits [41]. Population-specific QTLs were also identified, but only those reproducible in the consensus map by both the detection methods were considered significant. A case in point are the QTLs for petiole cross section and rachis length which were in close proximity in LG8 for population 2.6-1 and in the consensus map. Similarly, a minor QTL for rachis length was also found in LG4 of population 2.6-1 and the consensus map, close to the QTLs for height and petiole cross section in both populations. This provides further supporting evidence that the QTLs in LG4 are in a major region influencing all these three traits for compactness. The markers in the QTL intervals are useful for identifying favorable alleles for the characteristics desired. The population-specific QTLs were most likely due to the female parent, where in 2.6-1 it was a cross between an advanced breeding line (Serdang pisifera) and a germplasm from Nigeria (tenera palm, T128). The female parent in 2.6-5 was a self of T128. The slight difference in the genetic background of the female parents likely caused the phenotypic variability for the three traits in both populations, which contributed to some of the differences in QTL results from both populations. This explanation is also likely for the variability in the other population-specific QTLs, such as those for carotene content and FAC (C18:1, C18:2). Nevertheless, the population-specific QTLs can still be used to accumulate favorable alleles for oil unsaturation and compactness in crosses involving the populations. Similarly, population-specific QTLs were reported for pod-dehiscence in two families of soybean linked by a common parent [42]. Such specific QTLs are generally common in QTL analyses of bi-parental populations [43]. This suggests that the QTLs detected consistently across separate populations, as is the case in this study for petiole cross section, height increment, C18:2 content and iodine value, can be used with high confidence in marker-assisted breeding.
The marker SNPM00563 is significantly linked to two traits - petiole cross section and height increment - in LG4. From Table 6 and Figure 3, it is obvious that it was inherited from the male interspecific parent and the polymorphic allele gave larger petiole cross section. This is expected as oleifera has larger fronds and petioles. Since the male parent also has variability in height increment, the marker also points to a genomic region contributing to increased height increment. Furthermore, since both traits are positively correlated, it is to be expected that the genomic region that influences petiole cross section will also influence height increment. The marker can thus potentially identify the individuals (in this case, those not carrying the heterozygous alleles) for optimizing the crosses to be made. Interestingly, the marker linked to the second major QTL for height increment in LG7 (SNPM00010) was also inherited from the male parent, but the polymorphic allele resulted, on average, in lower height. As lower height is also expected from the hybrid (due to the oleifera), the marker could be used to accumulate favorable genes for height increment.
To date, there are no QTL analysis of vegetative traits in interspecific hybrids. The QTLs detected in this study were compared to those described previously [15] for a segregating E. guineensis population. Most of the QTLs we detected were on different chromosomes from those by [15], with the exception of those for rachis length on LG11. This suggests that separate genomic regions influence compactness in E. guineensis and interspecific hybrids. [18] reported two QTLs related to rachis length and petiole cross section in E. guineensis. However, a comparison for similarity between the linkage groups could not be made as the sequence for restriction fragment length polymorphism (RFLP) markers linked to the traits were not published. More specifically on height increment, recent reports revealed QTLs and candidate genes influencing it in E. guineensis [10,11]. However, the genomic region linked to height increment in this study was different from these two reports. Our results are consistent with an earlier study that compared the genetic architecture of FAC in both species of oil palm, where apart from some limited QTL loci that were common, the major QTL regions influencing FAC did not overlap and were species specific [32]. This suggests that the compactness and FAC traits are under the influence of different genomic regions in the two species.
Eight significant QTLs (seven majors, one minor) were detected for iodine value as well as C16:0, C18:1 and C18:2 content in four LGs in this study. Fewer were found than in previous reports on interspecific hybrids [16,20] with 19 and 12, respectively. Five of the identified QTLs by us were similar to those in both the previous reports. The QTL for iodine value on LG15 agrees with that by [16]. The major QTLs for iodine value and C16:0 content on LG1 overlap with those by [20] which shows their potential to be used in breeding, at least on their genetic backgrounds. The consistency of our QTLs detected with those reported previously is proof of the quality of the FAC phenotypic data in this study. The detection of strong QTLs linked to C16:0 content, the most abundant saturated fatty acid, and iodine value, similar to in previous studies [16,20], further supports the observation that the same genomic region has a major influence on the unsaturation (IV) and saturation (C16:0 content) of palm oil.
Three of the seven major QTLs associated with FAC resulted from markers informative in the male hybrid parent. Since higher unsaturation is to be expected from oleifera, it is interesting that the parental QTL marker alleles from the hybrid male parent (which has an oleifera component) - SNPM02507 (LG4) and SNPM03285 (LG15) - predicted on average for higher C18:1 content and iodine value, respectively. The third QTL peak denoted by marker SNPM00796 (LG1), also inherited from the male parent, predicted for lower C16:0 content (Figure 3). Generally, the higher unsaturation of E. oleifera oil is also due to its alleviated C18:2 content, almost twice that in E. guineensis [38] oil. Interestingly, a major QTL for C18:2 content was located about 165 cM away from the C18:1 content QTL in LG4, suggesting that they are separate QTLs. The marker closest to the QTL peak for C18:2 content in LG4 (SNPM00249) was contributed by the female E. guineensis parent and resulted in lower C18:2 content. As the female parental lines only involved E. guineensis, it is fairly obvious that this particular QTL was responsible for the phenoytpic variation of the E. guineensis palms used (intra-phenotypic variation). The directional effect of the QTL is also consistent with the dominant effect observed for the E. guineensis alleles for C18:2 content [32]. As such, enriching the hybrid parental allele for markers related to C18:1 content and iodine value, and with the alternative allele for C18:2 content, will give palms with higher oil unsaturation.
Interestingly, markers for three of the major QTLs linked to FAC were polymorphic in both the female (E. guineensis) and hybrid male parent. These QTLs were specifically linked to the unsaturated fatty acids (C18:1 and C18:2) and iodine value (Table 6 and Figure 3). It is thus impossible for these markers to accurately determine the source of the parental alleles in the offspring. However, as both parental lines (the female E. guineensis and hybrid male parent) had inherited alleles from palm T128, a Nigerian E. guineensis with less saturated oil (because of higher C18:1 content) [40], it is possible that even the E. guineensis contributed to the variability in FAC. As such, these QTLs could also likely account for the intra E. guineensis variation.
This study also detected the QTLs for carotene content. As such work has not been done before, no comparison could be made with other studies. Higher carotene is desirable because its pro-vitamin A and antioxidant activities enhance the nutritional attributes of palm oil [44]. The current palm oil has 500 - 700 ppm carotene content, while oleifera oil can have as much as 3000 ppm [45]. The individual BC2 palms had carotene content as high as 2600 ppm (Table 1). Of the three QTLs we detected, two were from the female parent (LG7 and 10), and the third heterozygous in both parents (LG3). The markers from the female parent predicted both for increased (SNPM03921, LG10) and reduced carotene content (SNPM03960, LG7), reflecting the variability of carotene content in the E. guinnensis oil. These markers can help select favorable alleles in future breeding for higher carotene content, apart from for compactness and more desirable FAC. Carotene content in the interspecific hybrids also appears to be under independent genetic regulation as there was a general lack of correlation with the compactness traits and FAC. Thus, breeders can select for higher carotene content and unsaturation without compromising compactness.
The phenotypic variation explained by the QTLs were corrected for the small population size as described by [32]. After the correction, the phenotypic variation explained by the major QTLs linked to individual fatty acids was 30 - 60%, similar to that reported by [32] who also adjusted for their limited population size. The highest PVE was for iodine value (89%), higher than in other studies [16,32], but in line with [16] who found that the QTLs linked to iodine value gave the highest PVE among the traits evaluated. The generally medium to high PVEs for the QTLs associated with the individual fatty acids and iodine value, respectively, is also consistent with the high broad sense heritabilities reported for these traits in a BC1 hybrid, which ranged from 80% (C18:0, C18:1 and C18:2 content) to 90% (C16:0 content and iodine value) [32]. The high PVE for carotene content (85%) is similar to in the potato, where the QTLs had 71% PVE [46]. This is also consistent with the high broad sense heritability (78%) for carotene content in E. guineensis oil [47]. Since the heritabilities of several traits are similar in both species of oil palm, it is likely that the heritabilities for carotene content in E. oleifera and the hybrid oils are also high.
Although broad sense heritability can vary depending on the breeding material and environment in which it is evaluated, high values are consistently reported for the individual fatty acids and iodine value in both species of oil palm [48, 49]. Similarly, the PVE for markers linked to FAC in maize was as high as 83% [50], and the cumulative evidence thus far clearly suggest that the FAC traits in oil palm are under simpler genetic control and less polygenic than the yield traits. Similarly, high broad sense heritabilities of 30 - 80% have also been reported for the compactness traits (petiole cross section, height increment and rachis length) in both interspecific hybrids [51] and E. guineensis [52] oil palm. As such, the compactness traits are also likely influenced by fewer genes than the more complex yield traits. Basically, there is clear support for our observation that a few loci with large effects control FAC, compactness and carotene content in oil palm. As natural selection tends to fix the alleles of large effect during the adaptation process, presumably the traits evaluated will assist oil palm in its reproductive success [53,54].
The medium to high heritabilities and PVEs for the QTLs in this study (for oil quality and compactness) are proof of their attractiveness for QTL analysis, as well as their amenability to selection via conventional and molecular breeding. However, PVE is often overestimated in most studies and may not be true in any breeding programme [55]. In fact, [56] cautioned that a small population can over- or underestimate PVE by as much as 19%, even with the correction for small populations (see Methods), as there is still the Beavis’ effect, which remains unadjusted [16, 57]. As such, caution must be exercised in expecting the gains from breeding.
Our small populations also explain why only the QTLs with medium to high effects were detected. Further, as demonstrated by [56], the chances of false negatives, i.e., not detecting the QTLs, increases the smaller the population below than 194, with the added risk that the QTLs detected may be as much as ~20 centimorgans (cM) from their actual positions. Another point of caution is that false positives increase with smaller populations, although they are more likely detected as minor QTLs, significant only chromosome-wide [56]. As such, the number of minor QTLs detected have to be treated with caution until validated in larger or other populations. The minor QTL detected for C18:2 content in LG1 was less than10cM away from the major QTLs associated with C16:0 content and iodine value, and it corresponded to almost the same genomic region previously linked to FAC in other studies [17], and so is likely a true QTL. As this study has demonstrated it possible to combine two families for marker-trait analysis, the integration can be easily extended to more BC2 families to increase the population size further for yet better detection of the QTLs and their positions. More importantly, as simulated by [58], increasing the number of palms will increase the detection of significant QTLs, especially those with small effects with PVEs less than 5%, which were missed in this study.
Knowledge of the oil palm genome [28] has allowed the underlying QTL intervals in the integrated map to be positioned on the EG5 physical map to identify potential candidate genes influencing the traits of interest. Even a 10cM region in a genetic linkage map can contain hundreds of genes [28,59], and this study only focussed on the selected genes (21of them) that had high similarity to known genes that impact fatty acid synthesis or have been associated in the literature with vegetative development.
The auxin transport protein, BIG, located in the QTL region for height, is required for auxin efflux and polar auxin transport (PAT) and could influence auxin-mediated developmental responses, e.g., cell elongation, apical dominance, lateral root production, inflorescence architecture, general growth and development [60]. Generally, BIG controls elongation of the pedicel and stem internodes through auxin action, which supports its role in regulating height in oil palm. In addition, BIG also plays a role in Arabidopsis in regulating its response to phytohormones, such as auxins, cytokinins, ethylene and gibberellic acid (GA), particularly during light-mediated stimuli, e.g., shade avoidance and etiolation [61,62]. BAM1, found in the QTL region for rachis length, encodes a leucine-rich repeat receptor-like serine/threonine-protein kinase and is known to regulate cell division and differentiation, such as in the formation of shape, size and symmetry of leaves [63], suggesting its possible influence on rachis length. ERECTA, linked to petiole cross section on LG4, regulates the aerial architecture (including inflorescence), e.g. shoot apical meristem-originating organ shape, elongation of internodes and pedicels, adaxial-abaxial polarity and stomatal patterning, probably by tuning cell division and expansion [64], which explains how it may control the petiole size in oil palm.
In FAC, the KCS11 gene on LG1 is associated with the QTLs for iodine value and C18:2 content. It is involved in fatty acid biosynthesis in both the saturated and mono-unsaturated acyl chains, C16 - C20 [65]. As such, it is interesting that the gene is located in the QTL interval regulating both saturated and unsaturated fatty acids in oil palm. The genomic region corresponding to the QTL intervals associated with carotene content did not contain any candidate genes associated with the carotenoid pathway, which is well characterized in plants [66]. A similar observation was made by [67] and [46], where the QTL intervals associated with carotene content did not contain any known carotenoid gene, suggesting the complexity of the mechanism regulating carotenoid biosynthesis in plants.
Although the candidate genes in the QTL intervals are interesting, it is, however, important to keep in mind that their involvement and influence in controlling compactness and FAC of oil palm are still speculative. Further studies are necessary to determine their functions in regulating the traits.