Genetic dissection of harvest index and related traits through genome-wide quantitative trait locus mapping in Brassica napus L.

The harvest index (HI) is the ratio of grain yield to the total biomass and represents the harvestable yield of crops. In Brassica napus, the HI is lower than that of other economically important crops, and limited relevant studies have been carried out regarding this issue. In this study, phenotypic analyses of 11 related traits showed the complexity of HI and the feasibility of cultivating desirable varieties with high HI. Quantitative trait loci (QTL) mapping based on a high-density genetic map identified 160 QTL, 163 epistatic loci pairs for HI and three closely related traits: seed yield (SY), biomass yield (BY) and plant height (PH), including two, five and three major QTL for HI, SY and PH, respectively. The related candidate genes underlying the QTL and epistatic loci with coding region variation were identified and investigated, including BnaA02g14010D, homologous to OsTB1, which functions as a negative regulator for lateral branching, and BnaA02g18890D, homologous to OsGW2, which controls grain width and weight. The complex correlation of HI with related traits, numerous QTL and epistatic loci and the candidate genes identified here provide new insights into the genetic architecture of HI, which might further enhance effective breeding strategies for yield improvement in rapeseed.

Introduction

In the study of modern crop cultivation and physiology, especially in the study of super-high-yield cultivation theory, the “source”, “flow”, “sink” theory has been used to explain the mechanisms underlying crop yield. Above-ground green tissues with photosynthetic capability, including the functional leaves, green stems, leaf sheath (e.g., in rice and wheat) and peel (e.g., in soybean and canola), are designated as “source” organs, the translocation of assimilates to storage organs is considered to be “flow”, while the seeds, which are storage organs, serve as the “sink” for photosynthesis. The balance among the photosynthetic “source”, “flow” and “sink” is a critical factor for yield improvement. The criterion to evaluate this balance, termed the harvest index (HI), is the ratio of economic yield (grain yield) to the total biomass and reflects the partitioning of assimilated photosynthates to the harvestable product; it is often measured and evaluated in breeding programs (Craufurd , Hay 2010, Mao , Yang and Zhang 2010).

The HI emphasizes the importance of carbon allocation in grain production and has been verified to be highly correlated with plant height (PH) and yield-related traits in important crop species (Lorenz , Sabouri ). Brassica napus L. (B. napus, AACC, 2n = 4x = 38) is one of the most important oilseed crops. In B. napus, HI was reported to range from 0.04 to 0.42, with an average value of 0.28 (Hirth , Lu ), which was lower than that of the most intensively cultivated grain crops, such as rice, wheat, barley and maize, ranging from 0.4 to 0.6 (Craufurd , Hay 2010, Mao , Yang and Zhang 2010) and other oilseed crops (such as soybean, 0.47 (Hay 2010)). The higher heterosis percentage for vegetative biomass than for seeds per silique (SPS) or thousand seed weight (TSW) indicates that the translocation of assimilates is the main limiting factor for seed yield in B. napus (Fu and Zhou 2013, Luo ). So improvement of HI through promoting the assimilates which flow to the seed, instead of being retained in vegetative organs, is a critical issue to improve SY of B. napus.

HI is a complex quantitative trait that is highly heritable and easily influenced by the transport and distribution of photosynthates, organ development, plant architecture and various environmental factors such as nitrogen nutrition levels (Sinclair 1998, Zhang and Yang 2004). In cereal crops, HI exhibits a negative correlation with PH (Lorenz , Sabouri ), and it has been significantly improved through the reduction of PH between 1900 and 1980 (Evans and Fischer 1999, Khush 2001). Additionally, improving HI by controlling crop height and side shoot production has also been reported in tomato (Gianfagna ). These results indicated that a low PH is more favorable for the transport of photosynthates to the storage organs. In recent years, HI has gained wide attention of researchers studying B. napus. Luo identified nine single nucleotide polymorphisms (SNPs) that were significantly associated with HI in the C genome by using genome-wide association studies (GWAS) and these SNPs explained 3.42% of the phenotypic variance in HI. Furthermore, by combining GWAS and transcriptome analysis, Lu detected 294 SNPs that were significantly associated with HI and HI-related traits, and 33 functional candidate genes located within the confidence intervals (CIs) of significant SNPs associated with these traits were revealed.

QTL mapping based on phenotype and genotype linkage analysis is a powerful approach to dissect complex quantitative traits. QTL for some complex quantitative traits, such as seed oil content, yield, plant architecture and boron efficiency, have been successfully identified in B. napus (Cai , Shi , Wang , Xu ). In addition, the extensive application of the Brassica Illumina Infinium 60K SNP array, based on next-generation sequencing technology, has greatly accelerated the process of genetic architecture dissection, and high-density genetic linkage maps based on SNPs have greatly improved the efficiency and precision of detected QTL (Cai , Chao , Trick ). Furthermore, the release of the B. napus reference genome has enabled researchers to identify potential candidate genes underlying the CI of associated QTL (Cai , Chalhoub ).

As yet, few QTL studies have been conducted for “HI=SY/BY” based on linkage mapping, although many QTL for SY and BY have been reported in B. napus. In this study, to better understand the genetic architecture of HI and its related traits in B. napus, we report (1) the correlation among HI and 11 other yield-related traits, with new insights for the breeding strategy of rapeseed resulting from phenotypic analysis and the dissection of the association pattern of 11 traits related to HI; (2) QTL mapping of HI and the related characteristics (SY, BY and PH) across multiple environments, based on a high-density SNP-based genetic linkage map; and (3) the identification and analysis of candidate genes based on the alignment of the linkage map to the B. napus reference genome. The results presented in this paper provide novel information about the genetic mechanism of HI and will help breeders to improve the yield potential of B. napus according to a new perspective.

Materials and Methods

Plant material and genetic linkage map

In this study, a segregating DH population named KN from the cross between ‘KenC-8’ and ‘N53-2’ was used to detect QTL for HI, SY, BY and PH. This population, containing 348 lines, was first developed by Wang and was used for the linkage map construction, of which the average marker interval was 4.4 cM. To improve the precision of QTL identification, a high-density SNP-based linkage map that included 3106 SNP-bins (including 17978 SNPs) and 101 non-SNP markers (SSR and STS) with an average genetic distance of 0.96 cM between adjacent loci was reconstructed (Chao ). The high-density SNP-based map was used for QTL mapping of HI and related traits in the current study.

Field experiment and data collection

The field experiments were carried out in 3 macroenvironments, as described by Wang : Wuhan of Hubei Province (a semi-winter-type rapeseed growing area, coded WH), Zhangye of Gansu province (a spring-type rapeseed growing area, coded GS) and Dali of Shannxi Province (a winter-type rapeseed growing area, coded DL). These experiments were conducted in 11 environments, including five consecutive years in DL (2008–2012, named as 08DL-12DL, respectively), three consecutive years in GS (2010–2012, 10GS-12GS) and three consecutive years in WH (2013–2015, 13WH-15WH). Zhao and Wang reported parts of the phenotypic data, prior to 2013, and detected QTL by using a genetic linkage map with 403 markers, but they obtained QTL with large CIs that were difficult to apply in practical breeding programs or in map-based cloning. However, the HI was not involved in the previous studies. In addition, the data from 13WH, 14WH and 15WH are new and are used for the first time in this study. All 348 lines, together with their parents ‘KenC-8’ and ‘N53-2’, were grown in a completely random block design, with two replications in WH and GS and three replications in DL. Each line was grown in two rows for each replication, every row contained approximately 12 plants, and the distance between rows and between plants was 0.4 m and 0.2 m, respectively. The field management followed normal agricultural practices.

At the maturity stage, five representative samples (selected randomly from plants with no damage and uniform growth) from each DH line and two parental lines were selected from each plot for measurement of related traits: SY (g), BY (g), PH (cm), SPS, TSW (g), silique length (SL, cm), first branch number (FBN), branch height (BH, cm), main inflorescence length (MIL, cm) and silique number of main inflorescence (SMI), as previously described in Zhao and Wang . The HI was calculated as the ratio of SY to BY. To perform a more in-depth genetic dissection of HI, the data of BY, SY and PH (including part of the data prior to 2013) (Wang , Zhao ) were used for the QTL analysis based on the high-density map mentioned above.

QTL analysis, meta-analysis and identification of epistatic loci

The QTL analyses for HI, BY, SY and PH were performed separately in each environment by composite interval mapping (CIM) using QTL Cartographer 2.5 software (Zeng 1994) based on the above described high-density genetic linkage map. The walking speed was set to 2 cM, with a window size of 10 cM and 5 background cofactors. The LOD threshold was determined by a 1000-permutation test based upon a 5% error rate (Doerge and Churchill 1996). Thus, an LOD of 2.9–3.2 was used to identify the existence of QTL in each environment; these QTL were termed ‘identified QTL’. The method for QTL nomenclature was as in Wang . A serial number was added if more than one QTL was located on the same chromosome, for instance qHI-A1-2.

Identified QTL repeatedly detected in different microenvironments and had overlapping CIs for the same trait were integrated into consensus QTL by meta-analysis using BioMercator2.1 software (Arcade , Goffinet and Gerber 2000) and were then named with initial letters “cq”, e.g., cqSY-A1-2. Additionally, if an identified QTL had no overlapping CIs with others, it was also regarded as a consensus QTL. QTL with a phenotypic variation (PV)>20% or a PV>10% and detected in at least two environments were considered to be major QTL. The consensus QTL with overlapping CIs for different traits were then further integrated into unique QTL and designated with the initial “uq”, e.g., uq-A1-1. The unique QTL integrated from two or more consensus QTL were considered to be pleiotropic-unique QTL and were named with an extra initial letter “P”, such as PuqA2-1. The consensus QTL for PH, SY and BY based on this high-density genetic linkage map were then compared with previous QTL based on a low-density map with 403 markers (Wang , Zhao ) using BioMercator2.1 software.

An associative network for QTL and traits was constructed by Cytoscape_V3.2.1 software and was based on the PV and additive effect (AE) of the QTL and the correlation between traits (Shannon ).

Epistasis analysis for HI, BY, SY and PH was performed separately in each environment by inclusive composite interval mapping (ICIM) using QTL IciMapping software (Meng ). The walking speed was set to 1 cM, and the LOD threshold followed the default parameter (5.0).

Comparative mapping and identification of candidate genes

Due to the release of the genome information of B. napus, the genetic linkage map can be aligned to the B. napus reference genome (http://www.genoscope.cns.fr/blat-server/cgi-bin/colza/webBlat) by BLAST, as described by Altschul and Chao ; the SNP-probe sequence was used as a query sequence. The CDS sequences of SY and PH-related genes in Arabidopsis and rice were used as queries to search for homologous genes using the BLASTn program in the B. napus reference genome, and these homologous genes were then mapped onto the B. napus genetic linkage map according to the relationship of the genetic linkage map and the reference genome. SY- and PH-related homologues falling within the CI of QTL or within 2.0 cM of the epistatic loci were selected for further analysis and were annotated according to the genome resequencing of the two parents from bulk segregant analysis (BSA) (Chao ). In addition, comparing the QTL to previous reported significant SNPs was implemented through mapping the QTL onto the reference genome based on an alignment of the linkage map and genome.

Results

Phenotype variation and correlation analysis among different traits

The HI phenotypic performance and frequency distribution were analyzed. The variation showed a “continuous distribution” as expected, which suggested the polygenetic effect of HI (Table 1, Fig. 1; the phenotypic analysis of BY, SY and PH prior to 2013 can be found in our previous studies). The high phenotypic variation and transgressive segregation of these traits suggested that favorable alleles for the three traits were from both two parents (Table 1). For instance, the HI values of the two parents were 0.27 and 0.35, whereas the phenotype value of the KN DH population varied from 0.13 to 0.41 in 11DL. The HI values of the two parents and HI frequency distribution of DH population showed significant difference in 11 environments (Table 1, Fig. 1). These results indicated the complex genetic control of HI, while at the same time it was influenced easily by the environment.

Table 1

Phenotypic statistics of four traits in different environments

	Material		08DL-HI	09DL-HI	10DL-HI	10GS-HI	11DL-HI	11GS-HI	12DL-HI	12GS-HI
HI	KenC-8	Mean	0.34	0.33	0.34	0.23	0.27	0.24	0.48	0.41
	N53-2	Mean	0.31	0.34	0.32	–	0.35	–	0.64	–
	DH	Mean±SD	0.31±0.06	0.32±0.05	0.32±0.04	0.22±0.08	0.31±0.03	0.32±0.1	0.35±0.13	0.16±0.09
	Range	Min–Max	0.16–0.54	0.18–0.51	0.14–0.52	0.1–0.46	0.13–0.41	0.08–0.59	0.13–0.54	0.01–0.33
	13WH-HI	14WH-HI	15WH-HI				13DL-PH	14DL-PH	13WH-PH	14WH-PH	15WH-PH
HI	0.33	0.25	0.25	PH	KenC-8	Mean	135	153	154.06	149.5	154
	0.31	0.25	0.28		N53-2	Mean	142	174.5	168.55	139.8	124
	0.25±0.1	0.24±0.16	0.26±0.05		DH	Mean±SD	143.01±14.19	164.12±17.35	160.77±20.71	135.72±14.61	134.39±16.02
	0.02–0.31	0.03–0.31	0.03–0.55		Range	Min–Max	74–174.33	100–206.5	90.5–230.13	90.88–183	86–173
			13WH-SY	14WH-SY	15WH-SY				13WH-BY	14WH-BY	15WH-BY
SY	KenC-8	Mean	14.4	15.6	7.6	BY	KenC-8	Mean	44.2	63	30.1
	N53-2	Mean	14.8	6.6	8.3		N53-2	Mean	47.4	26.2	29.6
	DH	Mean±SD	10.19±4.52	5.21±3.93	5.76±2.37		DH	Mean±SD	41.31±13.33	21.74±10.1	22.12±8.07
	Range	Min–Max	1–45.2	0.4–51.6	0.8–19.2		Range	Min–Max	15.5–100.3	4.6–102.6	5.6–74.7

Fig. 1

Frequency distribution of HI, BY, PH and SY in the KNDH population. In each plot, a marker denotes the median of the data, a box indicates the interquartile range, and spikes extend to the upper and lower adjacent values. The distribution density is overlaid.

Because improving the HI depended on both BY and SY, the relationship of HI, BY and SY, and the relativity of these traits with other eight traits involved in plant architecture and yield, including PH, SPS, TSW, SL, FBN, BH, MIL and SMI, were surveyed (Fig. 2, Supplemental Table 1). For instance, HI showed a positive correlation with SPS (0.34), SL (0.15) and SY (0.42) in a winter-type environment (DL), but showed a significant negative relationship with BY (−0.24). In a spring-type environment (GS), HI showed a positive correlation with SY (0.73) and a negative correlation with PH (−0.46), BH (−0.48) and BY (−0.33). In a semi-winter-type environment (WH), a significant positive correlation was also detected between HI and SY (0.49), SMI (0.30), and SL (0.20); in contrast, a significant negative correlation was detected between HI and PH (−0.14). In conclusion, HI showed a significant positive correlation with SY across all three macroenvironments, and negative relationships between HI and BY and between HY and PH were observed across two macroenvironments. The negative correlation of HI with BH was detected in only one macroenvironment, and the positive correlation of HI with TSW, SMI, SPS and SL were detected in two or one macroenvironment (Fig. 2). These results revealed HI to be a complex characteristic that is determined by several components which have positive or negative effects upon it. In addition, SY showed a significant correlation with all the other traits; BY and PH were also significantly correlative with most traits except for SPS and SL.

Fig. 2

Correlation of four traits in a winter-type (DL), semi-winter-type (WH), and spring-type (WH) rapeseed area, respectively. * and ** represent statistical significance at P < 0.05 and 0.01, respectively.

To further dissect the association patterns of the eight traits with HI and to provide reasonable breeding indices, the average HI of the DH lines with a certain scope of each trait was calculated, and the HI variation under the different phenotypic values of each trait was evaluated (Supplemental Fig. 1). Notably, when some traits (MIL, BH, BY, and PH) changed over a certain range, the HI was relatively high and basically remained the same, and values of these traits that were either too low or too high were unfavorable to HI. For example, when MIL varied between 40 and 65, HI remained at its highest value and was nearly unchanged across all environments. When PH varied between 100 and 180, HI was also not affected. These results suggested that these traits could be reasonably optimized to increase yield and cultivate optimal varieties, with traits such as lodging-resistant, excellent plant architecture, and suitable for machine harvesting, without changing HI.

QTL mapping for HI, SY, BY and PH in the KN population

In addition to being significantly correlated with HI, PH has been proven to have a significant effect on HI in breeding practices. Therefore, together with SY, BY and HI, these four traits were used for QTL mapping in the present study. In total, 160 QTL were identified in 11 environments and could explain 1.69–44.54% of the total PV, with an average confidence interval (CI) of 4.77 cM (Fig. 3, Supplemental Table 1). These QTL were distributed on 18 of 19 chromosomes (except for C7), 31, 37, 50, and 42 of which were identified for HI, SY, BY and PH, respectively (Fig. 4). Among them, 143 (89.38%) QTL could be detected only in a single microenvironment and were considered to be environment-specific QTL. Of the QTL for HI and SY, 78.57% (11 QTL) and 70.58% (12 QTL), respectively, were detected in GS with a negative AE, and 70.59% (12 QTL) and 73.68% (15 QTL), respectively, were detected in DL with a positive AE. These results indicated that the positive alleles underlying HI and SY from Ken-C8 were more inclined to take effect in a spring-type environment; in contrast, those from N53-2 were more inclined to express in winter-type environments. In addition, 22 QTL (84.62%) for BY with a positive AE and 19 QTL (95.00%) for PH with a negative AE were identified in DL and WH, suggesting that more alleles from N53-2 would express to increase BY and more alleles from KenC-8 would express to increase PH in winter- or semi-winter-type environments.

Fig. 3

Distribution of identified QTL, consensus QTL, unique related-gene QTL and candidate genes on each linkage group. From inside to outside, the 11 inner circles with background color represent the 11 different environments, and short bars with color within the 11 inner circles represent the identified QTL detected in each environment. The blocks at the outermost circle represent the 19 genetic linkage groups. The second circle is a line. Inside of the line shows consensus QTL and unique QTL. The red short bars on the line represent SY- and PH-related homologous gene positions, and the gene label shows candidate genes in QTL regions and their position.

Fig. 4

The number of identified and consensus QTL for four traits in each chromosome. There are two columns for each chromosome, the numbers of identified QTL are in the left, and the numbers of consensus QTL are in the right.

After the first round of meta-analysis, 160 QTL were integrated into 135 consensus QTL, of which 17 consensus QTL could be detected in at least two environments. There were more consensus QTL detected in the A genome (88 QTL) than in the C genome (47 QTL), and the number of consensus QTL varied from 1 (on C2 and C5) to 17 (A3) on each chromosome (Fig. 4, Supplemental Table 2). Of the consensus QTL, 30 QTL were for HI with AE and ranged from −0.12 to 0.05; these explained 1.69–44.54% of PV. Two QTL, cqHI-A2-2 and cqHI-A4-4, which could explain more than 20% of PV, are designated as major QTL. The QTL cqHI-A5-1 could be repeatedly detected in two microenvironments (two winter-type environments: 10DL and 11DL). Thirty-one SY consensus QTL were integrated from 37 identified QTL with AE and ranged from −17.88 to 13.75; these could explain 4.04–22.57% of PV. Five major QTL were identified for SY, four of which (cqSY-A2-6, cqSY-A6-1, cqSY-C6-5 and cqSY-C6-7) were detected in at least two microenvironments, and displayed a relatively large PV (>10%). An additional QTL (cqSY-A2-4) could explain more than 20% of PV despite that it was detected in only one microenvironment. The major QTL cqSY-A2-6 could be detected in all three spring-type environments (10GS, 11GS and 12GS), and cqSY-C6-7 could be detected in three winter- or semi-winter-type environments (11DL, 12DL and 13WH). Forty-three BY consensus QTL were integrated from 50 identified QTL with AE and ranged from −46.76 to 33.23; these could explain 4.59–18.16% of PV. Five consensus QTL could be detected in at least two microenvironments, but no major QTL were identified for BY. For PH, 31 consensus QTL were identified with AE, and they ranged from −7.11 to 10.19, while the PV ranged from 3.78% to 15.77%. Seven consensus QTL could be expressed in at least two microenvironments, and three of them (with PV>10%) were designated as major QTL. The major QTL cqPH-C9-6 was found to express in three winter-type environments (08DL, 09DL and 10DL).

Identification of pleiotropic-unique QTL (Pu-QTL)

Through the second round of meta-analysis, the 135 consensus QTL of different traits were pooled into 112 unique QTL, of which 20 unique QTL were pleiotropic, affecting at least two traits, with 8, 17, 10 and 8 related to HI, SY, BY and PH, respectively (Table 2). An associative network was also constructed based on QTL and the correlation between traits to assess the genetic control between traits (Fig. 5).

Table 2

The Pu-QTL controlling at least two traits

Pu-QTL	Chr.a	Peak (cM)	CI (cM)	Physical region (Mb)b	Related traitsc	Consensus QTL	LOD	PV(%)	Additived	Peak (cM)	CI (cM)
PuqA2-1	A2	53.37	52.73–54	4.81~4.95	SY, HI	cqSY-A2-3	3.32	9.24	−9.12	53.41	52.6–54.3
						cqHI-A2-1	5	15.03	−0.044	53.41	52.5–54.3
PuqA2-2	A2	57.81	57.51–58.1	6.9~6.49	SY, HI	cqHI-A2-2	19.5	26.82	−0.0534	57.81	57.6–58.4
						cqSY-A2-4	15.53	22.57	−17.8759	57.81	57.5–58.4
PuqA2-3	A2	58.63	58.06–59.2	6.49~6.7	SY, PH	cqSY-A2-5	5.21	18.88	−17.5073	58.51	58.1–59.8
						cqPH-A2-2	5.74–8.64	8.41–12.68	7.76–8.27	58.74	57.96–59.52
PuqA2-4	A2	63.85	63.05–64.65	9.57~9.63	SY, PH	cqSY-A2-6	3.35–13.24	9.32–19.64	8.86	63.02	62.4–63.65
						cqPH-A2-3	5.4	7.89	8.128	63.81	62.5–64.2
PuqA2-5	A2	64.21	64–70.58	9.68~2.49	BY, PH	cqBY-A2-1	3.38	5.99	−21.7586	64.21	64–70.4
						cqPH-A2-4	6.99	10.43	7.1086	64.21	64–71
PuqA3-1	A3	31.11	30.43–31.78	3.82~4.1	SY, HI	cqHI-A3-2	6.7	7.93	−0.0255	31.11	30.5–33.7
						cqSY-A3-1	4.77	6.09	−8.2701	31.11	30.8–32.3
PuqA3-2	A3	55.67	53.28–58.06	9.34~12.22	SY, BY	cqBY-A3-4	4.00–4.50	6.35–9.71	21.95–32.06	55.18	52.39–57.98
						cqSY-A3-2	4.98	7.6	8.2158	57.01	51.8–61
PuqA3-3	A3	62.71	61.96–63.24	12.23~12.38	SY, PH	cqPH-A3-1	4.31	9.15	5.3882	62.71	57–63.9
						cqSY-A3-3	4.55	8.74	10.3468	62.71	62.3–63
PuqA3-4	A3	138.81	137.25–140.36	25.48~26.66	BY, PH	cqBY-A3-7	3.35	5.89	21.8089	138.81	135.6–141.3
						cqPH-A3-4	7.36	10.41	9.2729	138.81	138.4–142.1
PuqA4	A4	24.58	21.5–28.97	12.97~13.4	SY, HI	cqHI-A4-1	2.9	4.85	−0.0222	23.51	21.5–26.1
						cqSY-A4-3	4.09	7.22	−3.8201	25.51	21.5–29.3
PuqA6-1	A6	0.01	0–0.37	0.8~0.81	SY, BY, PH	cqBY-A6-1	3.4	5.38	15.8476	0.01	0–4.7
						cqPH-A6-1	4.77	6.65	4.0949	0.01	0–3.2
						cqSY-A6-1	3.79–6.12	6.17–10.14	5.65–7.91	0.01	0–0.77
PuqA6-2	A6	5.75	5.24–6.27	1.47~1.75	SY, BY, PH	cqSY-A6-2	6.8	13.39	10.4543	5.71	4.7–5.8
						cqPH-A6-2	4.42	6.16	3.99	5.81	4.7–8.6
						cqBY-A6-4	3.10–3.70	4.94–6.70	15.09–23.16	6.53	4.21–8.86
PuqA6-3	A6	9.52	7.49–11.55	2.21~2.43	SY, BY	cqSY-A6-3	8.03	12.99	8.9214	9.11	7.1–11.7
						cqBY-A6-5	4	6.12	12.2009	11.01	5.8–14.5
PuqA8	A8	18.6	17.94–19.25	2.2~8.17	SY, BY	cqSY-A8-1	5	17.62	−13.4602	18.51	17.8–19.2
						cqBY-A8-2	3.71	11.64	−35.3212	19.21	17.3–20.9
PuqA9	A9	34.21	32.49–35.92	26.81~27.92	SY, HI	cqHI-A9	3.6	12.33	−0.036	34.21	31.4–35.8
						cqSY-A9	3.78	10.56	−8.4804	34.21	31.4–36.9
PuqA10	A10	22.31	20.74–23.88	12.44~12.74	SY, HI	cqSY-A10	5.93	7.68	−11.0106	21.51	19.5–26.6
						cqHI-A10-1	5	5.69	−0.0254	22.51	21.4–24.9
PuqC1	C1	73.85	72.51–75.85	34.5~35.93	BY, PH	cqBY-C1-6	3.3	5.6	−16.7359	72.61	72.4–75.3
						cqPH-C1-1	4.55	6.71	−3.3719	74.91	72.6–75.9
PuqC6-1	C6	59.73	59–60.45	22.42~26.9	SY, BY	cqBY-C6-1	3.6	6.71	25.6847	58.71	54.8–60.2
						cqSY-C6-2	3.83	7.31	8.5777	59.81	58.7–60.2
PuqC6-2	C6	62.96	62.42–63.3	32.25~34.22	SY, HI	cqHI-C6-3	9.3	14.98	0.0144	62.71	62.5–63.3
						cqSY-C6-3	10.28	16.59	13.7509	63.01	62.7–63.3
PuqC6-3	C6	65.48	63.12–68.5	34.23~35.42	SY, BY, HI	cqHI-C6-4	3.5	6.77	0.0354	64.11	63.7–69.1
						cqBY-C6-2	4.6	8.41	29.1197	65.01	62.1–68.2
						cqSY-C6-5	4.04–6.58	8.21–10.51	8.82–9.38	67.01	63.7–67.4

Chromosome that QTL located in.

The approximate physical interval corresponding to the CI of Pu-QTL.

Traits controlled by Pu-QTL.

Additive affect.

Fig. 5

Associative network of QTL and four traits. The larger yellow nodes represent four traits. Blue nodes represent consensus QTL that control only one trait, while red nodes represent pleiotropic-unique QTL. The size of the nodes that represent QTL show the level of PV value that they can explain. The dashed and solid lines between the QTL and the trait represent a negative or positive AE of the QTL, respectively. The dashed and solid lines between traits represent the correlation of traits in three macroenvironments (DL, GS and WH), and the color shows the degree of correlation, while a grey color represents no significant correlation.

Most of the pleiotropic-unique QTL affected different traits with the same direction, except for the two Pu-QTL (PuqA2-3 and PuqA2-4) controlling SY and PH. The significant positive correlation of SY with HI and SY with BY was coupled to a number of Pu-QTL detected for the two pairs of traits with an AE in the same direction. For instance, eight Pu-QTL related to HI affected SY, all with the same direction of AE. Of these, PuqA2-2 was integrated from two major QTL (qHI-A2-2 and qSY-A2-4), while uqC6-3 was a co-localized QTL effecting three traits, including HI, SY and BY, simultaneously. These results also indicated that HI was determined by SY to a large extent and was hardly effected by BY and PH, which was consistent with the correlation results. Six pleiotropic-unique QTL were identified to control SY and BY with AE in the same direction; of these there were two QTL (PuqA6-2 and PuqA6-3) that also affected PH, with AE in the same direction. In addition, three SY-QTL and three BY-QTL were found to co-localize with PH-QTL.

Identification of an epistatic interaction of four traits

An epistasis interaction was detected for HI, SY, BY and PH using IciMapping software. Seventy, thirty-three, fifty and ten epistatic loci pairs were identified for HI, SY, BY and PH, respectively (Fig. 6, Supplemental Table 3). Three epistatic loci pairs (EpA4-60 and EpA4-64, EpA9-35 and EpA9-40, EpC9-75 and EpC9-77) were identified to control SY and BY, simultaneously. One epistatic loci pair (EpC7-72 and EpC7-74) controlled HI, SY and BY simultaneously, and one pair (EpA5-46 and EpA5-49) controlled BY and PH simultaneously. In addition, a total of 51 epistatic loci were found to be located within QTL (Fig. 6, Supplemental Table 4). EpC4-117 was located in cqHI-C4-1 and interacted with five loci (EpA5-103, EpA6-22, EpA7-96, EpC1-53, and EpC8-90) and still influenced HI. EpA4-26 and EpA4-29 located in SY-QTL (cqSY-A4-3) affected BY and HI, respectively, through interactions with other loci.

Fig. 6

Epistatic loci and related candidate genes for HI, SY, BY and PH. The blocks at the outermost circle represent the 19 genetic linkage groups. The second circle is a line. Inside of the line shows consensus QTL for four traits and unique QTL. The red short bars on the line represent SY- and PH-related homologous gene positions, and the gene label shows candidate genes near epistatic loci and their position. Internal connections represent epistatic interactions between loci.

Identification of candidates underlying HI and related trait QTL

To identify candidate genes, 37 genes involved in SY and 65 genes involved in PH in rice and A. thaliana were collected (Supplemental Table 5). A total of 163 and 202 homologous genes for SY and PH, respectively, were identified in the B. napus genome and were evenly distributed across all B. napus chromosomes (Fig. 3, Supplemental Table 6). Nearly all of these genes had multiple copies except for ATDDF2 and CED1, and even SRS3 was found to have up to 10 copies.

According to the relationship between the B. napus genetic map and the physical map, 122 and 46 homologous genes involved in SY and PH, respectively, were located in the CIs of QTL of four traits (Fig. 3, Supplemental Table 7). BnaA02g14010D homologous to OsTB1, which functions as a negative regulator for lateral branching and modulates tillering in rice, was identified in the CI of the major QTL cqSY-A2-6 and near the CIs of cqHI-A2-3 (Choi , Takeda ). Resequencing and annotation of the two parents showed that N53-2, the parent with a relatively higher SY and HI, harbored a 9-bp deletion in the exon corresponding to three amino acid insertions (Phe-Pro-Ser) after 44-Ser in the BnaA2.TB1 protein. GW2, encoding a RING-type ubiquitin E3 ligase and identified from a QTL controlling grain width and weight in rice (Song ), was homologous to BnaA02g18880D and BnaA02g18890D; these two genes were found to be located in PuqA2-5 controlling BY and PH and near the CI of cqHI-A2-3. According to differences in sequence annotation, BnaA02g18890D had no variation between the two parents, but many differences were found in BnaA02g18880D. In addition to six SNPs in the exon, BnaA02g18880D has a 9-bp insert/deletion and two frameshift mutants caused by a single base insertion and a single base deletion between the two parents, respectively. Therefore, BnaA02g18880D is thought to be a more credible candidate for the region. IPA1, which controls tillering and regulates grain number in rice, was homologous to BnaC04g02520D, which was identified to underlie cqPH-C4-2 and is close to cqHI-C4-1b (Jiao , Miura ). BnaC04g02520D had three SNPs in the exon region, including two nonsynonymous. Interestingly, multiple copies with SNPs or/and InDels of some genes could be identified in QTL genomic regions. In addition to the two copies in PuqA2-5, GW2 was also located underlying cqSY-C6-1 (BnaC06g20390D) and PuqC6-1 (BnaC06g20400D). Genome resequencing of N53-2 identified two SNPs of BnaC06g20400D compared to Ken-C8, which directed a nonsynonymous mutation and a stop-loss mutation in the predicted protein. Furthermore, in some QTL it was found that multiple genes related to SY or PH could be identified in the CI of the individual QTL. For instance, four related homologues in rice (PAP2, FZP, DEP1 and OsPIN2) and two Arabidopsis-related homologues (BUD2 and STA1) were located in the CI of cqHI-C3-3; however, BnaBUD2, with four SNPs predicted to be nonsynonymous, was thought to be a more likely candidate gene. In the CI of PuqC6-3, four related genes were identified, but only SBH2 was identified as related to PH in Arabidopsis (Chen , Lan ). It had an SNP predicted to cause nonsynonymous mutations between two parents and was considered to be a candidate.

In addition, 106 related genes were identified underlying the epistatic loci of four traits (Fig. 6, Supplemental Table 8). For example, qSW5/GW5, which exerts a strong effect on rice grain width and weight (Liu ), was identified near EpA3-79 and EpA5-62 that affected HI and SY, respectively, through epistasis. PIN2, a gene affecting tiller numbers, angle and shorter plant height in rice (Chen ) and SRS5, encoding alpha-tubulin, which regulates seed cell elongation in rice (Segami ), were located near the two endpoints of one epistatic loci pair, EpA10-11 and EpC6-11, respectively. BnaA10. PIN2 had eight SNPs annotated to be nonsynonymous mutations and was considered to be a candidate gene of EpA10-11, but BnaC06. PIN2 had no difference in the promoter region and no variation causing amino acid changes in the coding region. This result might indicate that the interaction of BnaA10. PIN2 with other genes, rather than the copy of BnaC06. PIN2, underlies the epistatic loci pair.

Discussion

In the present study, HI showed a high sensitivity to environmental factors, consistent with earlier reports (Li , Luo ). In winter-type environments, the average HI of the DH population ranged from 0.31 to 0.35, which indicates the strong potential for HI improvement in B. napus. Nevertheless, in semi-winter-type environments, the average value was just 0.24–0.26, and the highest value of any individual line was only 0.31 (in 13WH and 14WH).

HI showed a significant positive correlation with SY across all three macroenvironments but exhibited complex correlations with BY, PH and the other traits in different growing areas. These results were consistent with those of a previous study (Luo ), which might suggest that the growth of tissues and organs in different environments might affect the synthesis of photosynthetic products and the transport efficiency of photosynthetic tissues to the seeds, thereby influencing seed filling. Therefore, as breeders select desirable cultivars, it is necessary to improve HI through the optimization of the related traits to improve the transport efficiency from the photosynthetic tissues to the seeds in specific plant environments. In addition, there was significant negative correlations between HI and FT only in spring-type environments. Three QTL detected in only spring-type environment were found to co-localize with flowering time QTL detected in spring-type environment. For instance, qHI-A2-2, with biggest PV value, co-localized with a major flowering time QTL which could be detected in 12GS, 13GS and 14GS (Li ). These QTL will be a very useful for breeding in specific spring-type environment.

The interaction of genotypes and the environment significantly affect HI and its related traits (Karinae , Li , Luo ). In the present study, to dissect the genetic architecture of HI, the traits of SY, BY and PH were employed for QTL analysis. Eleven main effect QTL (R2 >10%) were detected for HI, most of which were detected in a single environment, except for cqHI-A5-1. The results suggested that HI depends on the interaction between the environment and genotype, and this was also consistent with the phenotypic results.

Interestingly, the alleles from two parents tended to have different effects in different environments. Most QTL for HI and SY with a negative AE were detected in GS, with a positive AE detected in DL. The positive alleles underlying both HI and SY from Ken-C8 were more inclined to take effect in spring-type environments, and those from N53-2 were more inclined to express in winter-type environments. This might be attributed to their cultivated characteristics (KenC-8 is a spring-type B. napus, and N53-2 is a winter-type B. napus). Moreover, more alleles from N53-2 expressed to increase BY, and more alleles from KenC-8 expressed to increase PH, in winter- and semi-winter-type environments. These results suggested that HI and its associated traits were significantly affected by the interaction between the environment and the genotype.

Wang mapped 20 QTL for PH, and Zhao mapped 13 and 9 QTL for BY and SY, respectively, based on a previous map that included only 403 markers. Using the same environmental data, in our present study, more QTL (39, 26 and 29 for BY, SY and PH, respectively) were detected by using the high-density genetic linkage map. In addition to the QTL detected in a new environment, some previous QTL were mapped in greater detail. cqSY-A2-1, detected by Zhao , was finely subdivided into 4 QTL with smaller CIs, and cqBY-A3 was subdivided into up to 5 QTL (Supplemental Fig. 2). In addition, the CI of some QTL was obviously narrowed, such as in the major QTL cqSY-A2-1, detected by Zhao (from 20.2 cM to 1.15 cM, corresponding to the major QTL cqSY-A2-6 in this study) (Supplemental Fig. 2). Therefore, the high-density genetic linkage map enhances mapping resolution and allows us to more accurately localize QTL.

We next compared the QTL identified in this study with the significant SNPs associated with HI and the related traits, as detected by GWAS (Lu , Luo ). Forty-two QTL involved in HI, SY, BY and PH were validated by GWAS (Supplemental Table 9), of which two major HI-QTL were validated by two SNPs (Bn-A02-p10067386 and Bn-A04-p15108913) that had a strong association signal underlying HI and BY, respectively. The SNP Bn-A02-p10067386, which Lu had previously reported to associate with both HI and SY, was closely linked to cqHI-A2-2 (a major HI QTL), cqSY-A2-6 (a major BY QTL) and cqPH-A2-2. Therefore, we infer that the differentially expressed gene BnaCYP735A2 (encoding an ortholog of a cytochrome P450 monooxygenase of Arabidopsis (Kiba )) within the LD block of Bn-A02-p10067386 is quite valuable for the future function analysis. Additionally, due to some QTL having large CIs, there were many significant SNPs found in individual QTL. For example, 16 significant SNPs were found in the region consistent with a CI of 6 M in cqBY-A8-2, which signifies that further identification of variation loci might require a denser genetic linkage region near the CI of QTL and fine mapping using secondary mapping populations, such as near-isogenic lines (NILs). In addition, we noticed that nine SNPs on C8 could explain only 3.42 % of the total PV in the research reported by Luo . However, 11 main effects QTL (R2 >10%) detected for HI in this study was worth paying attention to, especially, major QTL cqHI-A2-2 with PV of 26.82% was co-localized with Bn-A02-p10067386 with 5.34% of PV reported by Lu to associate with HI.

As it is known that there have been two prominent breakthroughs in rice breeding. The first breakthrough was a sharp increase in rice yield potential coupled with an increase in HI from 0.3 to 0.5; the second breakthrough was an increase in yield potential that resulted from an enhancement of BY. Currently, the super-rice program is ongoing in China, Japan and other countries, which attempts to further improve rice yield through a continuous increase of both BY and HI. Based on the present study, this strategy is also likely suitable for B. napus. Seven and four consensus QTL controlling SY were co-localized with consensus QTL for HI and BY, respectively. This result coincided with the strong phenotype correlation of SY with both HI and BY, and the basis of the formula “HI=SY/BY”. It is exciting that all of these pleiotropic QTL control both SY and HI or both SY and BY, with an AE in the same direction. Therefore, breeding super B. napus by combining positive alleles for HI and BY to improve SY through MAS is a significant future goal. On the other hand, PuqC6-3 was found to affect HI, SY and BY simultaneously, and the direction of AE was the same. This result indicates that there might be an elite allele that further increases “source”, “flow” (dredge) and “sink” (fill), and would be quite valuable in gene cloning and breeding practices for the cultivation of high-yield B. napus in the future.

PH is one of the most important characteristics that can directly or indirectly influence the B. napus yield. PH plays an important role in lodging in B. napus, which can significantly decrease SY, final dry weight and HI (Islam and Evans 1994). Li reported that PH-QTL were co-localized with QTL that controlled height of the primary effective branch, length of main inflorescence, number of siliques per plant, and number of siliques on the first branch. Recently, Yuan (2014) proposed the theory of “new type morphology”, the main concept of which was to increase PH, and subsequently plant biomass, to raise the rice yield ceiling further and to break through the yield limitation imposed by HI. In the current study, pleiotropic QTL controlling PH and BY or SY were detected with AEs of the same or reverse direction to prove their associations, but none of the QTL for PH were found to be co-located with HI. In addition, there were two pleiotropic QTL (PuqA6-2 and PuqA6-3) that affected SY, BY and PH at the same time. Therefore, discriminating and utilizing favorable loci could be critical in breaking the undesired associations through MAS to obtain a high yield potential of B. napus with a suitable PH.

The epistatic interaction analysis provided more information to explain the complex variation of HI. More epistatic interaction pairs for HI were detected than for the other three traits (Fig. 6). Due to interactions with other loci, some loci no longer affected only the traits they originally controlled. For example, EpA3-98, located in cqPH-A3-2 and affecting PH, interacted with EpA9-17, located in cqBY-A9-2 and affecting BY, which in turn affected HI. EpA3-56, located in PuqA3-2, which controlled BY and SY, shifted to controlling PH by interacting with EpC6-47. These results provided us with new insights into the complex genetic architecture of HI. Besides, these epistatic loci co-localized with QTL provided more available information for increasing yield and improving plant structure at the same time through MAS strategy.

Mapping the genes that underlie QTL and epistatic loci could be a good method for rapidly identifying candidate genes. The available reference genome information for B. napus has accelerated this study. Because HI is a derived trait, no genes directly related to HI could be utilized. However, many genes related to SY identified in rice (one of the most important alimentary crops and a model species) and genes regulating PH from the database of Arabidopsis (a model species and a homologous species of B. napus) could provide useful information for identifying candidates that underlie these four traits. In the current study, the homologues of SY- and PH-related genes were first collected from rice and Arabidopsis and were identified in the reference genome. The candidate genes underlying QTL and epistatic loci were identified through mapping the QTL from linkage maps onto the reference genome, and 168 and 106 homologous genes involved in SY and PH, respectively, were mapped to QTL and epistatic loci of HI and related traits. The resequencing results of two parents provided additional information for the identification of candidate genes. The candidate genes with coding region variations, such as BnaC06g20400D, BnaA02g18880D and BnaC06g20400D require further validation through functional analysis. These results demonstrated the usefulness of strategies for the rapid prediction of candidate genes associated with complex traits.