New insights into the genetic networks affecting seed fatty acid concentrations in Brassica napus

Rapeseed (B. napus, AACC, 2n = 38) is one of the most important oil seed crops in the world, it is also one of the most common oil for production of biodiesel. Its oil is a mixture of various fatty acids and dissection of the genetic network for fatty acids biosynthesis is of great importance for improving seed quality. The genetic basis of fatty acid biosynthesis in B. napus was investigated via quantitative trail locus (QTL) analysis using a doubled haploid (DH) population with 202 lines. A total of 72 individual QTLs and a large number pairs of epistatic interactions associated with the content of 10 different fatty acids were detected. A total of 234 homologous genes of Arabidopsis thaliana that are involved in fatty acid metabolism were found within the confidence intervals (CIs) of 47 QTLs. Among them, 47 and 15 genes homologous to those of B. rapa and B. oleracea were detected, respectively. After the QTL mapping, the epistatic and the candidate gene interaction analysis, a potential regulatory pathway controlling fatty acid biosynthesis in B. napus was constructed, including 50 enzymes encoded genes and five regulatory factors (LEC1, LEC2, FUS3, WRI1 and ABI3). Subsequently, the interaction between these five regulatory factors and the genes involved in fatty acid metabolism were analyzed. In this study, a potential regulatory pathway controlling the fatty acid was constructed by QTL analysis and in silico mapping analysis. These results enriched our knowledge of QTLs for fatty acids metabolism and provided a new clue for genetic engineering fatty acids composition in B. napus.


Background
Oilseed rape (Brassica napus L., AACC, 2n = 38) is one of the most important oil crops producing multi-purpose oil for food and biofuel in many parts of the world. In 2007, biodiesel production accounted for 7% of the global vegetable oil supplies, in which 68% were used for biofuels in the EU [1]. As the global requirements for rapeseed oil are growing rapidly, increasing the oil content and improving the oil composition are important ways to meet the demands of agricultural feed stocks.
The fatty acid composition of rapeseed oil is considered to be genetically more variable than any other major vegetable oils [2]. Rapeseed oil is a mixture of seven main fatty acids [3]. Fatty acid biosynthetic pathways are generally controlled by multiple genes and considered as quantitative traits regulated by QTLs. So far, a number of QTLs controlling oil composition were identified in B. napus. Ecke et al. identified two QTLs for erucic acid distributed on chromosomes A6 and C2 [4]. Four QTLs for erucic acid distributed across chromosomes A1, A2, A8 and C3 were reported and three of these coincided with QTLs for the accumulation of oil content [5]. Burns et al. observed 13 QTLs affecting composition of 10 fatty acids, and seven also affected oil content [6]. Hu et al. identified two QTLs for oleic (on A1 and A5) and linolenic acids (on A4 and C4), respectively [7]. One to eight QTLs were detected for seven individual fatty acids by Zhao et al., and eight of these also affected oil content [8]. Recently, Smooker et al. identified 34 QTLs for five major fatty acids [9], and Yan et al. detected a total of 40 QTLs for six fatty acids, which were most clustered on chromosomes A8, A9 and C3 [10].
The allotetraploid B. napus has two progenitor species, B. rapa and B. oleracea, which shared their last common ancestor with A. thaliana about 20 million years ago [11,12]. Both a high degree of sequence similarities and chromosomal colinearities between Brassica species and Arabidopsis were reported [13][14][15]. Parkin et al. reported 21 conserved blocks within the Arabidopsis genome shared with B. napus [16], and Schranz et al. proposed a set of 24 conserved chromosomal blocks in B. napus [17]. Furthermore, all the genome sequence of B. rapa, B. oleracea and B. napus have been released [18][19][20]. It is feasible to predict the Arabidopsis orthologous genes for specific agronomic traits within the Brassica genome. For example, a number of candidate genes were mapped to CIs of QTLs for flowering time by in silico mapping [21], and a total of 14 lipid-related candidate gene loci were located in the CIs of six QTLs for seed oil content [22]. In fact, many important genes involved in fatty acid metabolism were identified in Arabidopsis, such as FAB2, FAD2, FAD3 and FAE1 [23][24][25][26], and the orthologs of these genes in B. napus were also reported and mapped. Bna-FAD2 was mapped on A1, A5, C1 and C5 chromosomes [27][28][29], and one major QTL BnaA.FAD2.a located on A5 was responsible for high C18:1 [27]. BnaFAD3 was mapped on A3, A4, A5, C3 and C4 [9,27], and two major QTLs BnaA.FAD3.b and BnaC.FAD3.b were both responsible for low C18:3 [27]. BnaFAE1 was mapped on both A8 and C3 [9], and two FAE1 homologous genes on A8 and C3 linkage groups were also found by Qiu et al. [5] and Fourmann et al. [30]. Collectively, although genes or QTLs for fatty acid biosynthesis have been identified, the genetic network for all these metabolic pathways in B. napus needs to be elucidated.
In Arabidopsis, more than 120 enzymatic reactions and at least 600 genes are involved in acyl-lipid metabolism [31]. Li-Beisson et al. gave metabolic pathways associated with the biosynthesis and degradation of acyl-lipids in Arabidopsis [31]. The genome of polyploid B. napus may typically contain six distinct alleles for each gene present in Arabidopsis [32], the fatty acid biosynthesis and the gene regulation in B. napus might have a more complex pathway than that in Arabidopsis. Though much attention was given to genes and regulatory factors involved in acyl-lipid metabolism in Arabidopsis [31,[33][34][35][36][37], similar questions concerning the genetic basis of fatty acid biosynthesis in B. napus remain open, mainly due to the lack of integrative studies at a population scale. Moreover, the interaction of genes involved in acyl-lipid metabolism has not yet been studied based on co-location of mapped candidate genes with QTLs in B. napus. To determine these key steps in relevant complex metabolic pathways of acyl-lipids in B. napus, it is first necessary to identify QTLs or genes associated with fatty acids composition.
In this paper, we describe the genetic bases of seed fatty acid composition through QTL mapping in B. napus. The aims of this study were as follows: (1) to add knowledge concerning QTL mapping of the fatty acid composition in B. napus; (2) to predict candidate genes of major QTLs for different fatty acids' biosynthesis by comparative genome analysis; and (3) to construct a regulatory pathway for fatty acids metabolism in B. napus.

Results
Variation and single QTL analysis of fatty acid composition in the 'Tapidor' × 'Ningyou7' cross (TN) DH population Means of all traits measured from the TN DH population over six environments were close to the mid-parent values (Table 1). There was a wide range of variations and transgressive segregations for the concentration of each fatty acid ( Figure 1). The population appeared to have a normal or near-normal distribution for C16:0, C18:0, C18:2, C18:3, C20:0, C22:0 and FAS (Saturated Fatty Acid), suggesting complexity of their genetic networks. However, C18:1, C20:1 and C22:1 showed bi-modal distributions, indicating that they might be controlled by few major genes with a relatively large effect. The distribution patterns of 10 fatty acids' compositions showed that they were genetically stable but also affected by environment. The correlation between different fatty acid compositions showed great differences (Table 2). Erucic acid (C22:1) content was highly and positively correlated with the level of C20:0, C20:1 and C22:0 (Coefficients 0.30-0.63), but was negatively correlated with other fatty acids (−0.92 to −0.03), especially C18:1 (−0.92) ( Table 2). C18:1 showed a high positive correlation with C16:0 and C18:0 (0.70-0.75) and moderate positive correlation with C18:2 and C18:3 (Table 2), but showed a high negative correlation with other fatty acids (−0.92 to −0.33).
For the other four unsaturated fatty acid components, phenotypic data were obtained from six different environments, except for C20:1 from three environments (07D, 07 W1 and 08 W2). Twelve QTLs for C18:1 were distributed across seven chromosomes and their additive effect ranged from −10.96 to 3.18 and explained 1.88-42.43% of PV (Table 3, Additional files 1 and 2). Nine QTLs on six chromosomes were associated with C18:2, with the additive effect ranging from −1.69 to 0.56 and explaining 3.10-36.26% of PV (Table 3, Additional files 1 and 2). Twenty-one QTLs on 10 chromosomes were significantly associated with C18:3, which had −1.64 to 0.42 additive effect and explained 5.70-22.29% of PV (Table 3, Additional files 1 and 2). For C20:1, 10 QTLs were detected across A1, A8, A9, C1, C3 and C8 in three environments ( Table 3). Six of them were found in two or more environments, and the remaining four were only in one environment. Nine QTLs were observed for C22:1 distributed across A4, A7, A8, C3, C6 and C9, and singly explained 1.27-45.79% of PV (Table 3, Additional files 1 and 2). Thirty-four synteny blocks (28 for A genome, 6 for C genome) and 149 insertion fragment islands (95 for A genome, 54 for C genome) were identified between Arabidopsis pseudochromosomes and TN DH genetic linkage groups by the in silico mapping approach (Additional file 3). More than twenty-nine thousand homologous genes were found to underline the CIs of 61 QTLs associated with the concentrations of 10 different fatty acids, by comparative analysis between the linkage map of TN DH and the genome of Arabidopsis (Additional file 4). Among them, a total of 111 key genes involving fatty acid metabolism were used as candidate genes in the present study (Additional file 4). A number of important genes were found, such as FAB2, FAD2, FAD3 and FAE1, but also five regulatory factors, LEC1, LEC2, FUS3, WRI1 and ABI3. As the polyploid B. napus genome may typically contain six loci for each gene present in Arabidopsis, these 111 genes were found to have 824 homologous genes mapped on the TN DH linkage map in total, including 97 key genes of 234 homologs underlying the CIs of 47 QTLs (Additional file 4). All of the QTLs with CIs containing homologous genes were separately compared to the physical genomic regions of B. rapa (A genome) and B. oleracea (C genome). To compare with the B. rapa genome, 32 QTLs containing candidate gene(s) distributed on the A genome of the TN DH linkage map were used for analysis, accounting for 65.3% of the total QTLs on the A genome. In total, 47 genes in B. rapa matched those in Arabidopsis underlying 24 QTL CIs (Additional file 4). For the C genome of the TN DH linkage map, 23 QTLs were detected and 15 (65.2%) of them contained candidate gene(s) in Arabidopsis. Comparison of the candidate genes showed 15 genes in B. oleracea matched those in Arabidopsis underlying 7 QTL CIs (Additional file 4). For example, on the C3 linkage group, 33 candidate genes (40 homologous genes) of Arabidopsis were located in the CIs of eight QTLs, and 10 candidate genes of B. oleracea were found based on the candidate genes of Arabidopsis, including the regulatory factor ABI3 of previous studies ( Figure 3).

Epistatic QTLs and interaction analysis of candidate genes for fatty acid compositions
QTLNetwork_2.0 and Genotype Matrix Mapping ver2.1 (GMM) software were used to identify the epistatic        were underlying the CI of qC9-3 (Additional file 4). A total of 34 significant loci interactions were identified for C16:0 by GMM, containing one pair with digenic and 33 pairs with trigenic interactions, most of these trigenic interactions have loci underlying three QTLs of CIs, including QTLs qA8-5, qA8-6 and qC3-3 (Additional file 7). A number of homologous genes were mapped to the CI of qA8-6, including LEC1, LEC2, ACC2, KASIII, FAD3, GPAT7, BCCP2, ACBP, PDAT, and ATS1, but only one homologous gene CYCA1 for qC3-3 was found (Additional file 7). All the results above suggested that the potential of these gene interactions increased the level of C16:0. Similar to C16:0, a series of pairs of epistatic QTLs and loci interactions for another nine traits were identified by QTLNet-work_2.0 (Additional file 6) and GMM (Additional file 7), and numerous important genes underlying QTL CIs were identified (Additional file 4). For example, for C18:0, C18:1 and C18:2, besides the trigenic interaction of QTLs qA8-5, qA8-6 and qC3-3, the interactions between QTLs qC3-2 and qA8-5 were also identified, and FAE1 was underlying the CI of qC3-2.
Cytoscape_V2.6.3 software was used to investigate the interaction of candidate genes that were observed from the single and epistatic QTL results and the gene interaction network was constructed. The results revealed that the whole network incorporated 167 nodes and 416 edges, that could be divided into three sub-clusters: the five regulatory factors (FUS3, ABI3, WRI1, LEC1 and LEC2) and the genes that were directly affected by only one regulatory factor (A cluster, Figure 4a), the genes affected by two or more regulatory factors (B cluster, Figure 4b) and the genes indirectly affected by regulatory factors (C cluster, Figure 4c). The A cluster consisted of 32 nodes and 105 edges in total. The five regulatory factors formed a tightly intra-linked group, with each regulatory factor under the influence of at least two other regulatory factors, and LEC1 and LEC2 were especially affected by all other four regulatory factors. In addition to the five regulatory factors, BCCP2 and CAC2 were underlying the QTL CIs. BCCP2 was associated with QTLs qA3-1 and qA8-6, which affected C16:0, C18:2 and C20:0; CAC2 was associated with QTL qC9-4 and affected C18:0. These two genes were both regulated by WRI1. The B cluster was composed of seven nodes and 40 edges, in which the genes were affected by at least two regulatory factors. SEP2 and LFY in this group were directly affected by all five regulatory factors. LFY was linked to other eight genes (Figure 4), and two of them (FAD2 and FAE1 in cluster C) were well-known important genes involved in fatty acid synthesis. The C cluster included 128 nodes and 319 edges, and 27 genes were detected to underline the QTL CIs. FATB associated with QTL qA8-2 was notably affected by both FAE1 and FAD2, and FATB also affected another eight genes (ACP1, FAD3, FATA, ACT1, TES, TAG1, CYP86A2 and GPAT5) that were underlying the QTL CIs. The network mentioned above indicated that the genes involved in fatty acid synthesis were directly or indirectly affected by regulatory factors.

Construct of a potential controlling pathway of fatty acids in B. napus
Among the 97 genes of Arabidopsis located in the CIs of 47 QTLs in B. napus, 57 were shown to have roles in 32 different pathways (Additional file 8). Among them, 31 genes were involved in at least one of the five pathways, including biosynthesis of unsaturated fatty acids, fatty acid metabolism, glycerophospholipid metabolism, fatty acid biosynthesis and alpha-linolenic acid metabolism, with 13, nine, nine, eight and six candidate genes, respectively (Additional file 9). The potential pathway for fatty acid synthesis in B. napus, which included 50 genes and the five regulatory factors mentioned above, was constructed by combining the main pathways that involved interaction of candidate genes as well as knowledge of fatty acid regulatory pathways in Arabidopsis ( Figure 5). The candidate genes in the deduced pathways could be divided into three types: genes associated with the fatty acid biosynthetic pathway (Figure 5a), the phospholipid synthesis and other pathways (Figure 5b) and the triacylglycerol and fatty acid degradation (Figure 5c).
Step 16 was the final step of TAG synthesis and TAG1 (qC8-1 and qC9-3), AT3G44830 (qC3-4) and PDA (qA3-1 and qA8-6) were observed in this step. Many genes or QTLs involved in one step indicated that the fatty acid synthesis in B. napus was very complicated.
Besides those genes that had direct activities in catalyzing fatty acid synthesis, some regulators (LEC1, LEC2, FUS3, WRI1 and ABI3) underlying QTL CIs were also found ( Figure 5). The interaction analysis indicated that WRI1 directly affected BCCP2 and CAC2, and indirectly affected FAB1 and FATB in early stages of fatty acid synthesis regulation, while WRI1, LEC1/2, FUS3 and ABI3 all indirectly affected FAE1, FAD2 and FAD3, which played important roles in later modifying and transporting processes. LEC1 and LEC2 were underlying the CIs of five QTLs (qA3-3, qA8-4, qA8-6, qA9-6 and qC9-3) and six QTLs (qA3-3, qA5-2, qA8-4, qA8-6, qA9-4 and qC9-3), respectively. Both LEC1 and LEC2 affected nine fatty acid compositions, except for C18:0, in different environments. FUS3 located in the CI of the QTL qC7-1 that was only associated with C18:3. WRI1 located in the CIs of QTLs qA4-3, qA5-2 and qA10-1 was associated with C16:0, FAS and C18:3, respectively. ABI3 was associated with three QTLs (qA5-3, qC3-8 and qC5-1) and affected the levels of C18:3, FAS, C18:1, C18:2 and C20:0 ( Figure 5). Moreover, the epistatic effects of major genegene interactions were also identified in fatty acid synthesis ( Figure 5). For example, the major genes FATB, FAE1, FAD2 and FAD3 were all under the influence of each other and all could affect the gene TAG1 in the final step of TAG synthesis, and then affected the level of different fatty acids ( Figure 5). All these findings indicated that the regulatory pathway of fatty acid synthesis was a complex network, controlled by a large number of genes and also affected by gene-gene interactions.
Potential phospholipid synthesis (Figure 5b), and triacylglycerol and fatty acid degradation (Figure 5c), were also predicted in B. napus based on knowledge of acyllipid metabolism in Arabidopsis [31]. G3P was converted to DAG3P by sequential acylation reactions in steps 17 and 18, which was regulated by GPAT and LPAT, and the homologous genes involved in the two steps were very similar to the genes in steps 13 and 14, respectively. Then DAG3P was converted to PC or PE by a series of enzyme reactions. In the pathway of triacylglycerol and fatty acid degradation, a number of homologous genes were also found, and the six genes in step 26 were the same as the genes in step 20. Moreover, gene LPI1, which functioned in degradation of TAG to MAG, affected the level of nine traits but not C20:0.
In addition to the 47 QTLs with candidate genes mapped in the CIs, the remaining 25 QTLs also affected the concentrations of one or two fatty acids. QTL qA4-4 on chromosome A4 with additive effect on traits C16:0 and C18:0 might be assigned to step 5 ( Figure 5) with potential ketoacyl ACP synthetase activity. QTL qC5-2 might be assigned to step 11 with a similar function to FAD3, as the QTL had a consistent additive effect on traits C18:2 and C18:3. QTL qC9-5 might be a pleiotropic locus with a similar function to FAE1 as it had an additive effect on traits C18:0 and C22:1. The remaining 22 QTLs only had additive effects on one trait. Based on the premise that plastidic synthesis and modification should result in consistent QTL effects on products in any of the individual sets of saturated fatty acids [42], qA7-3, qA7-4 and qC9-1 affected C16:0, and qA1-5 and qA6-2 affected C18:0, which allows us to assign these QTLs tentatively to early stages of synthesis. The other 17 QTLs may be assigned to desaturation and elongase steps, as they affected C18:3, C20:0, C20:1, C22:0, C22:1 and FAS. Generally, it is important to take into account all the preceding synthetic and modification steps where a QTL could be identified for a particular fatty acid [42]; however, assigning a specific enzyme activity to these QTLs is relatively complex.

Discussion
A total of 72 QTLs and a large number of pairs of epistatic loci for 10 fatty acid compositions were identified using QTL analysis. Then 234 homologous genes were mapped in the CIs of 47 QTLs. QTL mapping and interaction analysis of candidate genes enabled the construction of a regulatory pathway that controlled the different concentrations of fatty acids in B. napus.
Single QTL mapping of each fatty acid concentration showed that QTLs on A8 and C3 were highly correlated. QTL qA8-5 on A8 affected concentrations of all 10 types of fatty acids, and explained PV of individual fatty acids ranging from 15.89% for C20:1 to 47.55% for C20:0, and QTL qA8-6 affected seven fatty acids; meanwhile, both QTLs qC3-2 and qC3-3 on C3 affected nine fatty acids but not C18:3. All these results revealed that these QTLs were major QTLs, consistent with previous studies [6,9,10]. For the two QTLs on C3, the alleles from 'Ningyou 7' conferred higher levels of C20:0, C20:1, C22:0 and C22:1, but had negative effects on levels of C16:0, C18:0, C18:1 and C18:2. The group C16:0, C18:0, C18:1 and C18:2 were positively correlated with each other, and likewise for the group C20:0, C20:1, C22:0 and C22:1; however, the two groups showed a negative correlation ( Table 2). The fact that QTLs clustered in certain linkage groups and controlled different traits with opposite effects, corroborated strongly with the significant positive and negative correlations among the fatty acids analyzed. Besides the two major QTLs on A8 and C3, another 15 QTLs also affected more than one (See figure on previous page.) Figure 5 Potential regulatory pathways and candidate genes associated with fatty acid synthesis in B. napus. a, the fatty acid biosynthetic pathway; b, the phospholipid synthesis and other pathways; c, the triacylglycerol and fatty acid degradation. The black characters indicate genes detected in previous studies and also in this study, the red characters indicate genes not found in this study but in previous studies. The colored arrows indicate genes underlying a QTL which affects traits with additive effect, up-arrows indicate QTLs with positive additive and down-arrows a negative additive effect. The double-headed arrows indicate genes underlying different QTLs which affect the same trait with opposite additive effect at the same time based on the WinQTLcart2.5 and QTLNetwork_2.0 results (not combined results of the two). Different colors denote different traits as the bar shown at the lower right corner of the picture. The abbreviations of gene name are same with the following reference papers [30,37,38]. fatty acid. Moreover, the number of QTLs associated with a single fatty acid in the present study ranged from nine (C18:2, C20:0 and C22:1) to 21 (C18:3), and a total of 72 QTLs for all 10 traits detected were more than that from previous studies [6][7][8][9][10], with a number of new QTLs for different fatty acids. No QTLs for C16:0 were found on A4, A7 and C9 previously [6,8,10], suggesting that the 10 QTLs on these chromosomes in the TN population were all new QTLs. The unsaturated fatty acids were synthesized as a result of the fatty acid desaturation pathway, starting from C18:0; however, only a few QTLs for C18:0 were detected in previous studies. Zhao [6]. Herein, 12 QTLs across seven chromosomes were detected, and the QTLs on A4 (one QTL), A10 (two QTLs), C3 (four QTLs) and C9 (two QTLs) were new. QTLs for C18:1 were identified on 14 chromosomes except A4, A6, C2, C5 and C7 [6][7][8][9][10]. In the present study, there were 12 QTLs associated with C18:1 across seven chromosomes, nine of which also affected other fatty acids. QTLs for C18:2 were identified on 15 chromosomes except A4, A10, C5 and C9 [6,[8][9][10], indicating that the two QTLs (qC5-1 and qC5-2) for C18:2 on C5 were new. There were 21 QTLs for C18:3 identified here, and three QTLs (qA2-1, qA2-2 and qA2-3) mapped on A2 and one QTL (qC7-1) on C7 were new QTLs, but the former 18 QTLs were distributed across 14 chromosomes excluding A2, A3, A9, C6 and C7 [6][7][8][9][10]. Many QTLs for C22:1 were identified on 14 chromosomes excluding A4, A5, A7, C1 and C5 [4][5][6][8][9][10]43]. Two major QTLs associated with C22:1 on A8 and C3 chromosomes were observed in the present study, which was consistent with previous studies [5,9], and two new QTLs (qA4-1 and qA7-1) were mapped on A4 and A7 chromosomes (Table 3).
Generally, additive effects are considered as the main factors contributing to variations in quantitative traits [44,45]. However, epistatic effects can also play an important role in complex traits in B. napus [46,47]. Epistatic QTLs was found to explain PV ranging from 0.01% for C18:2 to 25.47% for C20:1 (Additional file 6), showing similar trends to previous studies [48]. Moreover, interactions among 395 loci were identified by GMM, including 34 pairs of digenic and 361 pairs of trigenic interactions (Additional file 7). A number of candidate genes were mapped to the CIs of these QTLs. These results showed that epistasis may substantially contribute to variation in oil concentration in different crops.
The regulatory pathway constructed in this study was based on QTL information and the interaction of candidate genes associated with the single and epistatic QTLs of 10 fatty acids, as well as previous knowledge concerning Brassica species and Arabidopsis [31,42]. A large number of important genes were found to be candidate genes for the processes of fatty acid metabolism. Among the 50 genes and five regulatory factors detected, 31 genes (62.0%) affected more than one trait, and FAE1 and ATS1 affected nine (but not C20:0) and 10 traits, respectively. Previous studies found that FAE1 affected oil composition in several steps of seed fatty acid synthesis and modification pathways in B. oleracea [42] and B. napus [30,49]. In B. napus, the orthologs of FAD2 and FAD3 were mapped in different linkage groups of the genome. BnaFAD2 loci were mapped in linkage groups A1, A5, C1 and C5 [28], and BnaFAD2 located in different linkage groups was found to be responsible for the high C18:1 level [29] and erucic acid content [9]. It seemed clear that a major functional variant of FAD2 orthologs in the A genome was located on linkage group A5, but unclear whether FAD2 loci were present in the Brassica C genome [8,29]. The ortholog of FAD3 was identified in the control of C18:3 level in B. napus, and a specific marker of fad3 was located in the CI of the QTL for C18:3 in linkage group A1 [50]. In the present study, FAD2 was located in three QTLs CIs: qA3-5 by C16:0 and C18:1, qC3-7 by C20:0, and qC5-1 by C18:2 and C20:0 using in silico mapping. The homologous genes of FAD3 were mapped in the CIs of seven QTLs: qA3-5 by C16:0 and C18:1; qA5-3 by C18:3 and FAS; qA5-4 by C18:1 and C18:3; qA8-6 by C18:1, C18:2, C20:0, C20:1, C22:0, C22:1 and FAS; qA10-3 by C18:0; qC3-7 by C20:0; and qC5-1 by C18:2 and C20:0. Taken together, these findings increased the understanding of the important genes affecting the processes of fatty acid metabolism in B. napus.
Despite the multiple studies on the function of several important genes in fatty acid biosynthesis [23][24][25][26], the regulatory mechanism of the pathway is still not well understood in B. napus. Five regulatory factors (LEC1, LEC2, WRI1, FUS3 and ABI3) were involved in the process concerning the fatty acid biosynthetic pathways, which might correspond to new important findings in this study. Both LEC1 and LEC2 could act as positive regulators upstream of ABI3 and FUS3 in the control of seed maturation [51][52][53], and several studies revealed that these four regulatory factors might have a function in regulation of fatty acid metabolism [37,[51][52][53][54]. WRI1, which acted downstream of LEC2, controlled expression of a subset of genes involved in fatty acid biosynthesis [33], and overexpression of WRI1 caused an increased TAG level in both seeds and leaves of Arabidopsis [34]. In this study, each of the five regulatory factors was under the influence of at least two other regulatory factors, and LEC1 and LEC2 were especially affected by all four other regulatory factors (Figures 4 and 5). Moreover, all five regulatory factors were associated with C18:3, but none affected C18:0. Conversion of acetyl-CoA to malonyl-CoA by acetyl carboxylase (ACCase) is one of the most committed steps in fatty acid biosynthesis [39,40], and reduction of ACCase activity caused a decrease of fatty acid content in the range of 1.5-16.0% in transgenic seeds of oilseed crops [55]. Most of the enzymes involved in de novo fatty acid synthesis seemed to be affected by WRI1 [33], and only WRI1 of the five regulatory factors had a direct impact on transcription levels of genes in the plastidial fatty acid synthetic pathway [38]. As a result of the interaction analysis of candidate genes, two genes CAC2 and BCCP2, which encoded subunits of ACCase, were found to be under the influence of WRI1. In the process of fatty acid elongation and desaturation, no candidate genes were shown to be under the influence of the five regulatory factors in this study ( Figure 5). Among the targets of the five regulatory factors, LFY might represent a key factor mediating the regulatory action of elongation and desaturation processes, as it was directly affected by all five regulatory factors and also regulated important downstream genes, including FAD2, FAE1 and FATB, and these three genes regulated a series of downstream genes ( Figure 4). LFY was arguably the most important floral meristem identity gene, which encoded a plant-specific transcription factor and controlled multiple aspects of floral morphogenesis [56,57]. It was striking that LFY played an important role in the regulatory pathway of fatty acid metabolism, and is worthy of future study.
Overall, these results provided important new insights into the regulatory model for the control of oil synthesis in B. napus and enhanced our understanding of the fatty acid synthesis pathways.

Conclusions
In this study, 72 individual QTLs and a large number pairs of epistatic interactions contributing to fatty acid biosynthesis were identified. By using in silico mapping analysis, 234 homologous genes of A. thaliana that could be involved in fatty acid metabolism within the CIs of 47 QTLs were found. After QTL mapping and candidate gene analysis, a potential regulatory network controlling fatty acid metabolism was revealed in B. napus. Our results provided new insights into the regulatory model for the control of oil synthesis in B. napus.

Plant materials and field experiments
The segregating DH population (the TN DH population), with 202 DH lines derived from the cross 'Tapidor' × 'Ningyou7' was constructed by Qiu et al. [5]. The DH lines as well as their parents were grown in six independent environments in China. The  No specific permissions were required for the field trials. For each field trial, all lines were planted in a randomized complete-block designed with three replicates, and each plot contained 30 plants per genotype [58].

Measurements of fatty acid composition
Bulked seed samples (200 mg) from each of replicates in the three environments (07D, 07 W1 and 08 W2) were analyzed for their fatty acid compositions by gas liquid chromatography according to Rücker and Röbbelen [59]. In three other environments (09D, 09 W2 and 09 W3), fatty acid compositions were determined by near-infrared reflectance spectroscopy (NIRS) using standard methods [60] and all saturated fatty acid compositions in these three environments were considered as one trait named FAS.
Linkage map construction and map alignment with Arabidopsis, B. rapa and B. oleracea The TN genetic linkage map of B. napus, which was introduced by Jiang et al. was updated in the present study [61]. A total of 932 markers were mapped to the new linkage map generated with the TN DH population using Join-Map_4.0 [62]. Detailed information of TN genetic linkage map was summarized in Additional file 3. Of the 932 linked markers, 429 with known sequence information or corresponding to Arabidopsis genes were used as anchored markers to carry out map alignment between B. napus and Arabidopsis according to the method of Long et al. [21]. If more than three homologous sequence-informative markers in the TN DH population were closely linked within one conserved block of Arabidopsis as described by Schranz et al. [17], a synteny block was considered to exist. If there were only one or two sequence-informative marker(s), it was recognized as an insertion segment. The homologous genes underlying the block or island were considered to be candidate genes if they were just underlying the QTL CI. There were 111 key genes (824 homologous genes) (Additional file 4) involved in fatty acid metabolism of Arabidopsis collected from the TAIR website [63] and in silico mapping in the genetic map as described by Long et al. [21]. The comparative mapping analysis between B. napus and B. rapa/B. oleracea was done using BLAST analysis on the Brassica Database [64] or B. oleracea Database [65]. The sequence of markers in the TN DH linkage map was used as a query sequence. All available sequence-informative markers on B. napus were subjected to BLASTn analysis to identify their physical positions on Arabidopsis, B. rapa and B. oleracea chromosomes. Then the Brassicaceae building blocks that corresponded to each informative marker were identified. The candidate gene(s) in B. napus, B. rapa and B. oleracea for each informative marker were also identified.
QTL and epistasis analysis QTL analysis was done using the QTLNetwork_2.0 software described by Yang et al. [66,67] and WinQTLCart_2.5 [68], and epistasis QTL mapping analyses were performed using the QTLNetwork_2.0 and GMM [69]. For QTLNet-work_2.0, a mixed linear model was used to identify QTLs at 2-cM intervals with a window size of 10 cM. Twodimensional (2D) genome scans were used to search for multiple interacting QTLs. For each trait, a genomewide threshold value of the F-statistic (P = 0.05) for declaring the presence of a QTL was estimated by 1000 random permutations [70]. A Bayesian method with Gibbs sampling was used to estimate QTL effects [71]. The sum of individual PV explained by each QTL was calculated as the total PV explained by all QTLs for each trait. Based on genetic effects, mapped epistatic QTLs comprised three types: interactions between two QTLs with additive effects (AA), interactions between a QTL with additive effect and a locus without significant additive effect (AN or NA), and interactions between two loci within non-QTL (NN) [45]. For WinQTL-Cart_2.5, composite interval mapping (CIM) was used with a significance threshold of P = 0.05 using the permutation test method [68], based on 1000 runs of randomly shuffling the trait values [72]. QTL CIs were determined by 2-LOD intervals surrounding the QTL peak. QTLs for different traits that mapped to the same region with overlapping CIs were assumed to be the same [58,73]. For GMM, two and three loci interactions were detected with default parameters. For the QTL comparison between different populations, QTLs detected in this study were preliminarily compared with those in other publications, because few common markers were available between TN and other populations. If one or more QTLs were identified on a chromosome in this study but not in previous studies, these were considered as new QTLs.