Oil content in oat, primarily in the form of triacylglycerols (TAGs), is a highly heritable trait. Thro and Frey (1985) observed a polygenic pattern of inheritance with primarily additive gene action and Schipper and Frey (1991) estimated that heritability of the trait ranged from 63 to 93%. The first quantitative trait locus (QTL) analysis for oil content-associated loci in hexaploid oat was conducted by Kianian et al. (1999) using the ‘Kanota’ × ‘Ogle’ mapping population and initial linkage maps (K×O) and the Kanota × ‘Marion’ mapping population and linkage map (K×M). They found there were three or four QTLs in each population, depending on the specific QTL analysis method used. Similar analyses using recombinant inbred line (RIL) mapping populations identified six QTLs for groat oil content in the ‘Terra’ × Marion mapping population and linkage map (De Koeyer et al., 2004) and six QTLs in the Ogle × ‘MAM17-5’ mapping population and linkage map (O×M) (Zhu et al., 2004). Additional genes are probably involved, but their individual effects would have been below the statistical threshold for detection. Kianian et al. (1999) further proposed that acetyl coenzyme A carboxylase (ACCase), which catalyzes the first committed step in fatty acid (FA) synthesis, was a candidate gene for a QTL on linkage group (LG) 11 of K×O. Hybridization experiments using an oat ACCase clone detected the presence of three loci in hexaploid oat (Kianian et al., 1999). One locus (AccaseA or Accase1) was mapped in K×O (Kianian et al., 1999; Wight et al., 2003) and two loci (AccaseA and AccaseB) were mapped in K×M (Kianian et al., 1999; Groh et al., 2001). In a more recent study, Tanhuanpää et al. (2010) identified eight QTLs associated with oil content. These accounted for 50% of the phenotypic variation in a homozygous doubled haploid (DH) population from the cross ‘Aslak’ × ‘Matilda’ (the Aslak × Matilda mapping population and linkage map [A×M]).
In addition to overall oil content, the FA composition of oat oil is also of considerable interest to oat breeders because of the nutritional significance of unsaturated FAs for both humans and animals. Although higher proportions of unsaturated FAs are desired for human consumption, polyunsaturated FAs may adversely affect the flavor and storage quality of oat. Several studies have investigated the FA composition of seeds from different oat varieties grown in various environments (Saastamoinen et al., 1989; Holland et al., 2001) but to our knowledge, genetic studies designed to identify the QTLs affecting these traits have not been conducted. Quantitative trait locus mapping of the genomic regions involved in the variation of oat grain FA content could provide information for the selective breeding of varieties with specific FA profiles.
Many of the reported oil content QTLs in oat show consistency among at least two different studies, but additional studies in other germplasm are required to provide a more complete inventory of these QTL regions and to understand each QTL’s sensitivity to environment and genetic background. Furthermore, there is a need to identify and map more genetic markers that are common among these populations to provide more accurate and systematic comparisons of chromosomal locations of QTLs. To date, most of the maps in oat have been generated using a variety of markers based on sequence characterized amplified region, simple sequence repeat, amplified fragment length polymorphism, random amplified polymorphic DNA, and restriction fragment length polymorphism technologies. The lack of common markers among different experiments has made it difficult to compare the positions of QTLs among oat maps. The Diversity Array Technologies (DArT) marker system is a cost effective hybridization-based marker technology that offers a high level of multiplexing (Mace et al., 2008). A recently developed DArT marker platform for oat and a revised DArT-based reference map in K×O (Tinker et al., 2009) provide new opportunities for rapid map development and comparative mapping in oat.
The objectives of this study were (i) to develop a DArT-based linkage map based on progenies from a cross between oat parents with high vs. low groat oil content, (ii) to measure groat oil contents, FA profiles, and additional agronomic traits in the progenies, (iii) to identify major QTLs affecting these traits, and (iv) to conduct comparative mapping with locations of other QTLs and candidate genes in oat.
Materials and Methods
An F5 oat population consisting of 146 RILs was developed by single seed descent from a cross between the two hexaploid oat varieties Dal (high oil) and Exeter (low oil). The Dal × Exeter mapping population and linkage map (D×E) RILs and the parents were grown in Aberdeen, ID, in 1997 in four-row irrigated field plots seeded with bulked F5:8 grain. Seed from these tests was used to plant two independent nonreplicated field tests in Ottawa in 2010. The first Ottawa trial was planted in hill plots on 19 May 2010. The second trial was planted in four row plots with a spacing of 25 cm between rows on 2 June 2010. The RILs were randomized separately for the two tests.
Samples for genetic analysis were based on 8 to 10 seeds that were selected randomly from the F5:8 Aberdeen-grown bulks. Seeds were planted in Cyg seed germination pouches (Mega Intl.) with one lower corner of each pouch cut so that it could be irrigated from below by capillary action. When the seedlings were approximately 5 cm tall, the leaves were harvested and placed in paper envelopes to which had been added a silica gel desiccant (VWR International). These were left to dry at room temperature and then ground using a TissueLyser (Qiagen). Qiagen DNeasy Plant Mini Kits (Qiagen) were used for DNA extraction from the parents and the mapping population according to manufacturer protocols.
Diversity Arrays Technology marker assays were conducted at Diversity Arrays Technology Pty Ltd based on methods described by Tinker et al. (2009). The standard hexaploid oat marker array available in 2010 was used and supplemented by an expanded array of markers derived from tetraploid oat (Oliver et al., 2011). Molecular mapping was performed using the maximum likelihood algorithm of the mapping program JoinMap 4 (Van Ooijen, 2006). Primary LG assignment was performed at a minimum logarithm of odds (LOD) ratio of 5.0. The grouping threshold was relaxed to LOD 4.0 to identify potential LGs among unassigned loci. Markers that showed high segregation distortion were excluded.
The D×E map was compared to the DArT-based linkage map constructed from K×O (KOD), which includes many marker types, including DArTs (Tinker et al., 2009). Putative homologies between LGs from the two populations were established based on three types of evidence. First, LGs were declared to be homologous if they shared four or more markers. Second, smaller LGs where the proportion of shared markers was high in one or both groups were declared to be homologous. Third, when other supporting evidence was available, either from previous work or the current study, the two groups were declared to be homologous.
Groat oil was measured in 2010 from grain samples grown at Aberdeen in 1997. These samples had been kept at −20°C since harvest. Following extraction using an Accelerated Solvent Extractor (ASE-200, Thermo Fisher Scientific), total lipid content was determined. Samples from each line and the parents were thoroughly cleaned and mixed in a Seedboro quality mixer (Seedboro Equipment). Two sets of subsamples were taken: 100 g for oil extraction and 500 g for protein measurement. The 100-g subsamples were dehulled using a CODEMA laboratory huller (CODEMA LLC) and ground with a cutter mill (Arthur H. Thomas Scientific Apparatus) fitted with a 20 μm mesh screen. Ground samples (flour) weighing between 400 and 500 mg for each line and the parents were then placed in extraction cells. A ratio of 3:2 of hexane:isopropanol was used as solvent. The conditions for extraction were pressure of 6.90 × 103 kPa, temperature of 125°C, heating period of 6 min, and extraction (static) time of 20 min, with three static cycles per sample. The total recovered solvent volume was then evaporated at controlled temperature (80°C). Recovered lipid mass was then weighed on an analytical balance to determine the percentage of oil in the original sample. Oil extraction was performed in two repetitions and the average of the two results was used as the final oil content. Moisture content of each repeated sample was measured at the same time by using the air-oven method (American Association of Cereal Chemists Method 44-15-02 [AACC, 2009]). The oil content of samples was calculated on a dry weight basis.
The lipid oil profile was assayed by analyzing lipid content extracted from oat flour heated at 90°C for 80 min with 2.5 mL of methanolic sulfuric acid (2% v/v) and then cooled down and supplemented with a solution of 1 mL hexane spiked with 0.2 mg methyl heptadecanoate per milliliter of hexane. The mixture was shaken and left overnight for separation of the hexane. The methylated FA samples were analyzed by gas chromatography using a Hewlett-Packard 6890 Series GC system (Hewlett-Packard Technologies) equipped with an Agilent 6890 Series injector (Agilent Technologies) and fitted with a DB-Wax capillary column (Agilent 127-7013; 10 m by 100 μm by 0.20 μm).
Protein levels were determined by near-infrared transmittance on whole groats using an Infratec 1241 system (FOSS). The machine is routinely calibrated using a calibration equation developed in house at the Eastern Cereal and Oilseed Research Centre, Ottawa. The calibration is updated and validated every year with wet chemistry.
Phenotypic data for other agronomic traits were collected on all 146 lines of the mapping population planted in the two Ottawa 2010 field trials. Heading date was defined as the time when 50% of the panicles in a plot were completely emerged. Plant height at maturity was defined as the mean height of five randomly selected plants per plot or three stems per hill. Visual scoring on a scale of 1 to 9 was used to determine lodging.
The program MQTL (Tinker and Mather, 1995) was used to detect and estimate QTL effects. Both simple interval mapping (SIM) and simplified composite interval mapping (sCIM) were used for QTL analysis. The linkage map was scanned at all marker positions and at 5 cM intervals between markers. The QTL with the largest effect on each LG identified by SIM was used as background marker for sCIM to test for additional smaller QTL and also to refine QTL positions. Quantitative trait locus peaks separated by more than 20 cM were generally regarded as distinct QTL. Significant QTL main effects were used as anchors in testing for pairwise epistatic effects. Each anchor “A” was held constant while the genome was scanned at a walking speed of 5 cM for a second QTL “B” that interacted with “A.” Epistasis was declared when the test statistic for the multiplicative interaction term was greater than the average epistatic test statistic in permuted data. Statistical significance thresholds to achieve 5% genomewide type I error were determined by 10,000 random permutations. Final estimates of QTL positions were based on sCIM or on locations of an epistatic model, and final effects were estimated in a multilocus regression model incorporating all reported QTL positions for a given trait.
The DArT assay detected 523 polymorphic markers (Supplemental Table S1), 143 of which showed segregation distortion at the 0.05 level of significance, with Dal alleles overrepresented at 86 loci and Exeter at the remaining 57 (Supplemental Table S2). Of these, 48 were excluded because of extreme segregation distortion (χ2 ≥ 15). The remaining 475 markers formed a linkage map with 40 LGs covering a distance of 1271.8 cM (Supplemental Fig. S1), which is 43% of the 2932 cM genome size estimated by O’Donoughue et al. (1995). Seven markers remained unassigned to any LG. Markers were not evenly distributed along the chromosomes. Some LGs had a high concentration of markers while others were sparsely populated. Thirty-nine percent (185) of the markers mapped to only four LGs (4, 5, 11, and 1 in D×E). Fifteen LGs were made up of 12 to 23 markers each while the smallest groups (25 of 40 LGs) consisted of two to eight markers each. Linkage group 4 in D×E had the highest number of markers, with 75 markers covering a distance of 86.3 cM. In addition, two LGs (7 and 17 in D×E) showed broad regions of skewed segregation ratios favoring the Dal allele (Supplemental Table S3).
The D×E map is considerably shorter than the updated 2028 cM KOD map to which DArT markers had been added (Tinker et al., 2009). This may be a result of inadequate polymorphism in some regions of the parental genomes that correspond to regions of similar ancestry. Dal and Exeter are genetically closer than Kanota and Ogle, with a kinship coefficient (r) of 0.062% (p < 0.05) vs. 0.047% for Kanota and Ogle, as calculated using the program KIN (Tinker and Mather, 1993). Alternately, it can also be due to the fact that the DArT markers were discovered in K×O and therefore biased to being polymorphic in that population (Tinker et al., 2009).
After comparative mapping, 27 LGs in D×E were found to share markers with 20 LGs in K×O (Tinker et al., 2009) (Supplemental Table S4). Of these, four LGs from D×E each shared markers with two different K×O LGs while seven K×O LGs each shared markers with two or three D×E LGs. The number of shared markers ranged from 1 to 32. Twelve of the associations were based on four or more shared markers. These homologs are LG 1 in D×E with KOD 2, LGs 2 and 12 in D×E with KOD 4_12_13, LGs 3 and 26 in D×E with KOD 16_23, LG 4 in D×E KOD 17, LG 6 in D×E with KOD 24_26_34, LG 11 in D×E with KOD 19+25+27, LG 13 in D×E with KOD 11_41_20_45, LG 14 in D×E with KOD 33, LG 17 in D×E with KOD 22_44_18, and LG 27 in D×E with KOD 1_3_38_X1. Of these homolog sets, the strongest evidence for homology was observed between LG 4 in D×E and KOD 17, which shared 32 loci. Although the remaining associations were based on small numbers of shared markers, many of these involve very short LGs with a high proportion of shared markers.
A number of the LGs in K×O were subsequently joined together using evidence from aneuploidy and from other comparative mapping (Kianian et al. 1997; Wight et al., 2003). Some of the LGs in D×E (e.g., 19–23–38, 6–34, 7–13) could potentially be joined, based on their relationships with the KOD LGs. However, doing so could introduce errors in orientation. Since we are aware that comprehensive efforts in developing an oat consensus map are underway, we feel it is preferable to present the current map based solely on evidence from this study.
Several D×E LGs shared markers with more than one K×O LG. While this may indicate groups that could be joined in K×O, it may also indicate homeologous associations (Supplemental Fig. S2d). For example, KOD 4_12_13 and KOD 5_30, which both share markers on LG 2 in D×E, are likely to be homeologues, based on other map comparisons (Wight et al., 2003). Similarly, LGs KOD 11_41_20_45 and KOD 37, which both share markers on LG 7 in D×E, are also likely to be homeologues (Kianian et al. 1999; Tinker et al., 2009).
Quantitative Traits and Quantitative Trait Locus
Phenotypic data was collected for total oil content, the major FA components, and protein as well as agronomic traits such as plant height, heading date, and lodging (Supplemental Table S5). The frequency distributions for all the traits are presented in Fig. 1. Transgressive segregation in both extremes was observed for stearic acid, oleic acid, linoleic acid, and linolenic acid while oil content and palmitic acid exhibited transgressive segregation only in one direction. Transgressive segregation was also observed in both extremes for plant height and heading date, but in only one direction for lodging. The presence of QTLs was initially inferred based on SIM. Further scanning using sCIM confirmed the positions of most of the QTLs discovered by SIM and identified some new QTLs. A few of the QTLs discovered using SIM were not confirmed by sCIM but are being reported as they are statistically highly significant. A summary of QTLs affecting all oil-related traits is presented in Table 1, and those QTLs affecting other traits are summarized in Table 2.
|D×E||Marker or marker interval||Linkage group distance||Distance from marker‡||Oil content §||Palmitic acid§||Stearic acid§||Oleic acid§||Linoleic acid§||Linolenic acid§|
|D×E||Marker or marker interval||Linkage group distance||Distance from marker‡||Protein content§||Plant height
Quantitative Trait Loci Affecting Oil Content
A total of six QTLs associated with oil content were identified in genomic regions belonging to LGs 1, 4, 7, and 13 in D×E (Table 1). Linkage groups 4 and 13 in D×E had two QTLs each and one QTL was observed on each of LGs 1 and 7 in D×E. All alleles for higher oil content were contributed by Dal, and their effects ranged from 0.22 (oPt-13269) to 1.87% (oPt-6135) oil content.
The QTL with the largest effect, which accounts for 32% of the phenotypic variation in the trait, was located on LG 13 in D×E between oPt-17088_A and oPt-6135. The position of this QTL collocates with a QTL near cdo665b found on KOD 11 and on LG 11 in K×M (Kianian et al., 1999) as well as a QTL near rz69 on LG 3 in O×M (Zhu et al., 2004). Both of these were tightly linked to the Accase1 locus. Similarly, a QTL for oil content was identified by Tanhuanpää et al. (2012) closely linked to Accase1 locus on LG 12 on A×M, which is homologous with KOD 11. This QTL has been consistently identified in four different populations. Moreover, in three out of the four studies it was linked to Accase1. This can be considered a validation of this QTL.
The second major QTL, accounting for 19% of the phenotypic variation, was located near oPt-17489 on LG 7 in D×E, which shared markers with two K×O LGs (KOD 11_41_20_45 and KOD 37) on which QTLs for oil have been mapped.
Homology of LG 7 in D×E to either KOD 11_41_20_45 or KOD 37 is ambiguous. The LG 7 in D×E and LG KOD 11_41_20_45 share two markers and both groups contain a major QTL for oil content. In addition, both LGs have regions of segregation distortion (O’Donoughue et al., 1995). The shared markers, the colocation of QTLs for oil content, and the presence of regions of segregation distortion on both LG 7 in D×E and KOD 11_41_20_45 would suggest that these two LGs are homologous. However, LG 7 in D×E has three markers in common with KOD 37, which also contains a major QTL for oil (Kianian et al., 1999). In addition, LG 7 in D×E shares markers with LG 11 on A×M, which is homologous to KOD 37 and on which the candidate gene Accase2 was mapped (Tanhuanapää et al., 2012). More evidence is required to resolve the issue of homology of LG 7 in D×E with either KOD 11_41_20_45 or KOD 37; however, the relationships established here suggest that KOD 37 and KOD 11_41_20_45 are homeologous.
Two other QTLs with minor effects were identified on LG 4 in D×E. The position of one of these loci, oPt-16384, corresponds with a QTL identified on LG 1 of A×M by Tanhuanpää et al. (2012). Our study is the second to detect QTLs for oil content in this region.
The position of the QTL on LG 1 in D×E corresponds to that of the QTL identified on LG 6 in O×M (Zhu et al., 2004). The combined effect of the QTLs in this study explains 76% of the phenotypic variation in oil content among the progenies.
Quantitative Trait Loci Affecting Oil Composition
A total of four to eight QTLs associated with individual FA components were identified on six LGs (Table 1). Four QTLs for palmitic acid, two with major and two with minor effects, were detected on LGs 7, 13, and 30 in D×E. These QTLs explain 54% of the total phenotypic variation of the trait among the progenies. All alleles for higher palmitic acid content were contributed by Exeter. Variation in stearic acid content was also associated with four QTLs located on LGs 7, 13, and 30 in D×E. These QTLs explain 56% of the total phenotypic variation for this trait. Oleic acid content was associated with eight QTLs explaining 76% of the total phenotypic variation. The loci with the highest effects were located on LGs 7 and 13 in D×E. All but one of the alleles for increased oleic acid content were contributed by Dal. Linoleic acid content was associated with seven QTLs on LGs 4, 7, and 14 in D×E, with effects ranging from 0.59 to 2.81%. The alleles for higher levels of this trait were contributed by Exeter, except for one locus, which was derived from Dal. These seven QTLs explain 75% of the phenotypic variation for this trait. The variation in linolenic acid was associated with four QTLs located on LGs 3, 7, 8, and 13 in D×E. The combined effect of these loci explains 42% of the total phenotypic variation for the trait.
A graphical summary of all QTL regions affecting oil components is presented in Fig. 2a. This figure shows multilocus estimates (presented in Supplemental Table S6) for each measured FA regardless of whether the QTL was significant for that specific FA. To visualize possible mechanisms for changes in oil components, we re-estimated QTL effects for each oil component measured as a proportion of total seed weight (Fig. 2b; Supplemental Table S7). When expressed this way, the same QTL regions were significant for at least one of the FA components, although not all QTLs remained significant. When expressed relative to total oil, the general trend for all QTLs was that the Dal allele increased the proportion of oleic acid and decreased the proportion of linoleic acid. However, when expressed as a proportion of seed weight, the Dal allele increased or maintained all FA components, with the greatest increase being in oleic acid.
Quantitative Trait Locus Affecting Agronomic Traits
A total of 12 QTL-associated markers in nine localized regions were associated with plant height, heading date, lodging, and protein content (Table 2). Found between oPt-6441 and oPt-15763 on LG 17 in D×E, only one minor QTL was detected for protein content. This locus accounted for 12% of the total phenotypic variation for the trait. The allele for higher protein content was derived from Exeter.
The variation in plant height for the hill plots was associated with four QTLs. One of these, oPt-10654, was detected only with SIM. When tested with sCIM, the statistic for this locus fell just under threshold. The four loci explain a total of 39% of the total phenotypic variation among the progenies. Both parents contributed alleles for increased height but the alleles from Exeter had the largest effects. Three QTLs were detected with SIM for plant height in the population grown in row plots; however, the test statistics for all three loci fell under the significance threshold on further scanning with sCIM. Of these, two QTLs were in the same position as those identified in hill plots while one, oPt-795082, was located on a different LG.
Similar to the other agronomic traits, observations for lodging were performed on D×E planted in hill plots and in row plots. The hill plots were damaged by extremely strong winds. Therefore, QTL analysis was performed only on measurements taken from the row plots. Three QTLs, found on LGs 14, 15, and 30 in D×E, explained 68% of the phenotypic variation in lodging (Table 2). Both parents contributed alleles that decrease lodging. Two of the alleles were derived from Dal and one was contributed by Exeter. The effects of these loci ranged from 17 to 30%. Overall, reduced lodging was contributed by Dal, which is the shorter parent. All three QTLs were detected with SIM and their positions confirmed with sCIM. The map positions of the loci for lodging in the current study did not correspond to QTLs reported in earlier studies.
Heading date was associated with two QTLs in the hill plots. Both loci (on LGs 14 and 30 in D×E) were identified with SIM and confirmed with sCIM. These same loci were also identified in the row plots as were two more QTLs detected on different LGs (5 and 17 in D×E). One of these, oPt-6446, was detected only with sCIM.
For most QTLs, parental alleles had the expected effects on traits. For example, alleles from the high oil content parent Dal increased oil content while alleles from the low oil content parent Exeter had the opposite effect. However, a few QTLs exhibited effects in the opposite direction from that expected. For example, oPt-17524 from Dal decreased oleic acid content while the rest of the QTLs from the same parent increased the trait.
Quantitative Trait Locus Epistasis
Based on 10,000 permutations for each trait, a consensus threshold statistic of 15 was established for controlling global type I error rate below 5% per full-genome scan. Scans for epistatic interactions conducted for all traits revealed six potential two-way interactions. Of these, only three were accepted as possible interactions (Table 3). The remaining three interactions showed exaggerated genotypic estimates that were likely due to very small sample numbers in one genotypic class. Furthermore, it is recognized that every epistatic scan, of which there were at least several per trait, introduced a new error rate of 5%. Three epistatic loci were associated with plant height. Two of these (oPt-10256 and oPt-14611) interacted with the locus oPt-795758, which is an unlinked marker.
|Locus A†||Locus B†||Test statistic||aabb‡||AAbb‡||aaBB‡||AABB‡|
|oPt-795082 (LG§ 11)||oPt-795525 (LG 38)||17.7||72.29||72.19||66.73||75.3|
|oPt-10256 (LG 14)||oPt-795758 (LG 41)||16.5||76.74||67.8||71.58||71.29|
|oPt-14611 (LG 14)||oPt-795758 (LG 41)||17||76.53||66.81||71.48||71.39|
We have identified eight QTL regions affecting total oil content and/or at least one FA component. The two QTLs with the largest effect on total oil content (near oPt-6135 and oPt-17489) also affected the FA components (palmitic, stearic, oleic, and linoleic acid). Moreover, the multilocus estimates of QTL effects for seven of these eight loci showed the same pattern of effects on all FA, regardless of whether the locus was significant for a specific FA (Fig. 2a and 2b), and the QTL analysis for each FA produced highly parallel QTL scans (Fig. 3). These results suggest that all or most of these QTLs have a similar mode of action, differing only in magnitude and, therefore, in significance. That is, each QTL region may represent a single segregating locus that affects all oil-related traits pleiotropically through a common physiological or biochemical mechanism. Alternative explanations would imply the presence of clusters of linked QTLs with independent effects on each oil component. This seems improbable, because each QTL region behaves in a parallel manner, so the separate QTLs within each cluster would need to have the same linkage phase. In addition to this, there were no significant epistatic effects among loci affecting oil or its components. Therefore, our working hypothesis is that all or most QTLs affecting oil in this population operate additively through a single node of control.
The pattern of effects observed in this study is consistent with the hypothesis of a global upstream mechanism that affects total oil content as well as FA proportions. Given that TAG assembly is a separate downstream process from FA synthesis, this mechanism seems most likely to be operating at the level of de novo FA synthesis in the plastids or even further upstream. Despite our current understanding of the biochemistry of FA and lipid synthesis in plants, the signals and factors that direct the expression of genes in these pathways remain largely unknown, and it is not possible from these results to determine which genes or regulatory elements are responsible. However, previous authors have speculated on the possible role of ACCase, a rate-limiting enzyme that affects the availability of FA precursors, as a primary regulatory node in FA biosynthesis in oat (Kianian et al., 1999). Our current results are consistent with the hypothesis that ACCase is a primary node of control. If, for example, each QTL affects the abundance of ACCase transcripts, then each QTL would have an additive effect on total oil content. The rates of conversion and/or transport among the downstream steps could remain genetically constant, but the relative proportions of FA in the groat would be affected by the abundance of precursors, resulting in our ability to detect QTL for each FA component. Specifically, more precursors could lead to a greater abundance of all FA, but rate-limiting steps could favor the accumulation of oleic acid that we observed. Alternately, White (2007) has suggested that the proportionately higher levels of oleic acid that are observed as total groat oil is increased are a result of increases in the storage of TAGs, where oleic acid is preferred, relative to structural lipids, where it is less dominant. This explanation might be taken to suggest that increased groat oil is a result of increased sink capacity for stored TAGs. Diacylglyserol acyltransferase (DGAT) activity has been proposed as the primary node of control for TAG synthesis. Diacylglyserol acyltransferase, which catalyzes the formation of TAGs from diacylglycerol and acyl coenzyme A (acyl-CoA) in the endoplasmic reticulum, exerts control over oil accumulation (Katavic et al., 1995; Weselake et al., 2008). Increased activity of DGAT may lead to a depletion of acyl-CoAs, which in turn could stimulate de novo fatty acid production to satisfy the demand for acyl-CoA. Metabolic control analysis experiments that showed DGAT has a flux control coefficient of 0.74 in olive (Olea europaea L.) provided some evidence for this (Ramli et al., 2005). However the same experiment in oil palm (Elaeis guineensis Jacq.) showed a coefficient of 0.12 indicating that the control mechanism varies from one species to another. Determining which of these mechanisms (source vs. sink) is responsible for genetic variation in groat oil content will be an important step in elucidating the molecular determinants of oil metabolism in oat.
To our knowledge, this is the first study to investigate QTL effects at the level of FA composition in oat. Because of this, the QTLs associated with FAs in this study cannot be compared directly to previous work. However, comparative mapping has shown that at least three of the eight oil-related QTL regions in D×E are coincident with QTLs affecting total oil content in other oat populations, so we speculate that their mode of action and pleiotropic effects on FA could be similar.
Although oil content in oat is a highly heritable trait that can be increased and decreased by selective breeding (Frey and Holland, 1999), the pleiotropic effects on FA composition observed in the current study suggest that there is limited opportunity for manipulation of the FA composition of the oil independently of the total oil concentration. The exception may be linolenic acid, for which QTLs were detected on LGs 3 and 8 in D×E with no effect on other measured traits. However, the individual effects of these two QTLs were very small, accounting for differences of only 7 and 9% of linolenic acid proportion in the oil or 9 and 13% of the linolenic acid proportion in the total seed by dry weight.
The proportion of markers exhibiting segregation distortion in the current study (27% at p < 0.05) was higher than expected. O’Donoughue et al. (1995) reported that 8% of the markers tested deviated from the 1:1 ratio expected in a RIL population, Zhu and Kaeppler (2003) reported 9%, and Portyanko et al. (2001) reported 13%. A combination of the type of population used and the dominant inheritance pattern of DArT markers may have contributed to the unusually high proportion of markers with distorted segregation. Both DH as well as RIL may exhibit extreme segregation distortion possibly due to the presence of recessive alleles that are lethal or strongly selected (Xian-Liang, 2006). It is also possible that dominant alleles segregating at two or more different genetic loci can masquerade as one overrepresented allele at a single locus because of the nature of the marker assay. This phenomenon is known to affect some DArT markers in oat (Tinker et al., 2009). However, the later phenomenon would only account for extreme (e.g., 3:1) genetic distortions, and such markers would generally be excluded from the map unless they clustered to produce a separate pseudo-LG entirely comprising duplicated loci. It is also possible for segregation distortion to result from the loss of unbalanced gametes caused by translocation heterozygotes. For example there is a major known translocation in oat between chromosomes 7C and 17 (Jellen and Beard, 2000) that has affected mapping in other crosses, but both Dal and Exeter are expected to contain the same version of this translocation and there is no evidence for segregation distortion on groups LGs 27, 28, or 35 in D×E, which correspond to subgroups of K×O 1_3_38 known to be affected by this translocation (Supplemental Fig. S2n; Supplemental Table S3). It is possible that other minor translocations or other chromosomal rearrangements have influenced segregation distortion. Markers with distortions greater than 2:1 were not included in the map, and there were no entire LGs composed of markers distorted to this degree. Rather, the graduated regions of mild to moderate segregation distortion observed in the D×E appear to represent regions where one or the other parental allele has been selected or has drifted to a majority state.
Specific regions of segregation distortion have been reported in other oat populations. O’Donoughue et al. (1995) reported a distorted region on KOD 11. A distorted region was also reported on LG 3 in O×M, which is homologous to KOD 11 (Zhu et al., 2004). The distorted region on LG 7 in D×E is also potentially homologous to KOD 11 (Supplemental Fig. S2h). Genetic control of segregation distortion has been reported in other cereals such as rice (Oryza sativa L.) (Devaux et al., 1995) and barley (Li et al., 2010). The presence of distortion in homologous regions across three different oat populations provides strong evidence for the presence of a heritable factor affecting segregation distortion. Since all three populations were developed through single seed descent, it is unlikely that there is a common mechanism affecting fitness after seed maturation. Possible mechanisms could include gametic or zygotic selection (Liu et al., 2010). However, the loci on both KOD 11 and LG 3 in O×M were skewed toward the male parents whereas the distortion on LG 7 in D×E favored alleles from the female parent, Dal. Therefore, it seems unlikely that there is a common mechanism of gamete selection, but there could be a heritable mechanism affecting the survival or fitness of the zygotes. It is interesting that the same distorted region in all three oat populations contains major QTLs affecting oil content, suggesting that the mechanism responsible for this aberrant segregation pattern may directly or indirectly affect oil content.
Quantitative Trait Loci Affecting Additional Traits
The position of the single QTL identified for protein content in D×E corresponds with the QTL reported by De Koeyer et al. (2004) on LG TM 15 of the Terra × Marion mapping population and linkage map. Previous studies have identified additional QTLs affecting protein content in oat, but the genetic control of this trait does not appear to be consistent, ranging from numerous small effects (Zhu et al., 2004) to relatively large effects (Tanhuanpää et al., 2012). These inconsistencies may be attributed to the large magnitude of environmental influence on protein content. Crop production environments, and N supply in particular, affect protein percentages of oat (Welch and Leggett, 1997).
It is interesting that two QTL regions affecting heading date and height in both trials also coincided with two QTLs affecting lodging in both the hill plots and the row plots. The QTL oPt-11992 on LG 30 in D×E, which had an effect on heading date in both trials, was also associated with lodging. Similarly, oPt-10256, which affected height in both trials, was located close to the QTL that had the largest effect on lodging (oPt-14057). These associations were consistent with the phenotypic correlations observed between the traits (Supplemental Table S8), and the direction of the effects appears to have a developmental basis: Dal alleles that reduced plant height and days to heading also reduced lodging. These correlated responses could be due to linkage of the underlying QTLs or to pleiotropy. Some QTLs affecting plant height and heading date, however, were not consistent between the row plot and hill plot trials. This discrepancy might be attributed to differences in the growing conditions between the two sets of plots. The hill plots were planted 2 wk earlier than the row plots, and they experienced a heavier weed infestation.
In summary, we have identified several chromosome regions and DArT markers associated with oil content, oil quality, and several agronomic traits, some of which have been linked to loci identified in other populations. Moreover, our study has demonstrated the relationship between genetic factors controlling oil content and the concentration of individual FA components of oil in oat. Although QTL analysis lacks the power and resolution to elucidate metabolic pathways, by estimating the effects of all the alleles for each biosynthetic product along the metabolic pathway we have provided an overview of the genetic mechanism of oil biosynthesis in oat. This information will be useful not only for manipulating the composition and concentration of oil but also in ongoing efforts to understand the molecular basis for the control of FA composition and oil accumulation in seeds. The results of this study will be used for further fine mapping and identification of suitable markers for molecular breeding.
Supplemental Information Available
Supplemental material is included with this article:
Supplemental Table S1. Raw Diversity Arrays Technology (DArT) data for the Dal × Exeter recombinant inbred line (RIL) mapping population.
Supplemental Table S2. Chi-square test for segregation ratio in Dal × Exeter recombinant inbred line (RIL) mapping population.
Supplemental Table S3. Segregation ratio and chi-square goodness-of-fit for Diversity Arrays Technology (DArT) loci mapped in the Dal × Exeter recombinant inbred line (RIL) population.
Supplemental Table S4. List of homologous loci between the Kanota × Ogle and Dal × Exeter populations.
Supplemental Table S5. Phenotype data for the Dal × Exeter recombinant inbred line (RIL) mapping population.
Supplemental Table S6. Multilocus estimates of quantitative trait loci (QTLs) for each measured fatty acid (FA) component of the oil when fatty acids were measured as a proportion of the oil content.
Supplemental Table S7. Multilocus effects of genomic regions associated with oil content and fatty acid composition in oat when fatty acids were measured as a proportion of the dry weight of the seed.
Supplemental Table S8. Phenotypic correlations among agronomic traits in the Dal × Exeter recombinant inbred line (RIL) population of oat.
Supplemental Figure S1. Linkage map of the Dal × Exeter population.
Supplemental Figure S2. Comparative map of homologous linkage groups and QTLs in two mapping populations (the Kanota × Ogle mapping population and initial linkage maps [K×O] and the Dal × Exeter mapping population and linkage map [D×E]) of hexaploid oat (A. sativa L.)