Genome-wide association study for candidate genes controlling seed yield and its components in rapeseed (Brassica napus subsp. napus)

Abstract

Genetic improvement of seed yield per plant (SY) is one of the major objectives in Brassica napus breeding programme. SY, being a complex quantitative trait is directly and indirectly influenced by yield-component traits such as siliqua length (SL), number of seeds per siliqua (NSS), and thousand seed weight (TSW). Therefore, concurrent improvement in SL, NSS and TSW can lead to higher SY in B. napus. This study was conducted to identify significant SNPs and putative candidate genes governing SY and its component traits (SL, NSS, TSW). All these traits were evaluated in a diverse set of 200 genotypes representing diversity from wide geographical locations. Of these, a set of 125 genotypes were chosen based on pedigree diversity and multi-location trait variation for genotyping by sequencing (GBS). Best linear unbiased predictors (BLUPs) of all the traits were used for genome-wide association study (GWAS) with 85,126 SNPs obtained from GBS. A total of 16, 18, 27 and 18 SNPs were found to be significantly associated for SL, NSS, TSW and SY respectively. Based on linkage disequilibrium decay analysis, 150 kb genomic region flanking the SNP was used for the identification of underlying candidate genes for each test trait. Important candidate genes involved in phytohormone signaling (WAT1, OSR1, ARR8, CKX1, REM7, REM9, BG1) and seed storage proteins (Cruciferin) were found to have significant influence on seed weight and yield. Genes involved in sexual reproduction and fertilization (PERK7, PERK13, PRK3, GATA15, NFD6) were found to determine the number of seeds per siliqua. Several genes found in this study namely ATS3A, CKX1, SPL2, SPL6, SPL9, WAT1 showed pleiotropic effect with yield component traits. Significant SNPs and putative candidate genes identified for SL, NSS, TSW and SY could be used in marker-assisted breeding for improvement of crop yield in B. napus. Genotypes identified with high SL, NSS, TSW and SY could serve as donors in crop improvement programs in B. napus.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12298-021-01060-9.

Introduction

Rapeseed (Brassica napus subsp. napus) is one of the most important oilseed crops. It is the third largest source of vegetable oil accounting for approximately 15% of the total vegetable oil used for human consumption (Carruthers et al. 2017; Kaur et al. 2020). It is an allotetraploid oilseed plant species (AACC, 2n = 38), originated through the hybridization of two diploid species namely, B. rapa (AA, 2n = 20) and B. oleracea (CC, 2n = 18) about 7,500 years ago (Nagaharu 1935). It is being grown mainly under sub-humid areas with irrigation. In India, it is largely cultivated from October to April (winter season) in north-western parts comprising states of Punjab, Himachal Pradesh, Haryana and Delhi under assured irrigated conditions (Priyamedha et al. 2015).

Breeding for genetic improvement of seed yield in B. napus is an important objective to increase productivity of rapeseed. Seed yield is a complex quantitative trait governed by various components, and highly influenced by genotype-environment interaction (Yang et al. 2016; Zhu et al. 2020). Siliqua number, number of seeds per siliqua and seed weight have been well reported as important determinants of seed yield (Yang et al. 2012). Seeds per siliqua and seed weight have further been found to be significantly correlated with siliqua length (Chay and Thurling 1989; Lebowitz 1989; Diepenbrock 2000) that contributed towards higher productivity in B. napus. X-ray studies suggested failure of double fertilization as the main factor leading to fewer seeds per siliqua (Pechan and Morgan 1985). Therefore, concurrent improvement in siliqua length, number of seeds per siliqua and seed weight can lead to higher seed yield in B. napus. Hence, there is a need for marker-assisted approach to accelerate commercial breeding program for enhanced crop yield.

In B. napus, number of QTLs for these traits have been identified using bi-parental mapping population and genome wide association studies (GWAS). More than 100 QTLs were reported for number of seeds per siliqua (NSS), out of which, three major QTLs located on chromosomes A6, A7 and C9 have been fine mapped (Shi et al. 2015; Yang et al. 2016; Zhu et al. 2020). Based on trait-by-trait meta-analysis, Zhang et al. (2011) reported 26 QTLs for siliqua-traits, out of which, five for siliqua length, two for seeds per siliqua and three for seed weight were consensus QTL. Liu et al. (2015a) reported co-located genes for seed weight and siliqua length in a major QTL on chromosome A9 of B. napus and increase in these traits due to a 165-bp deletion in the auxin-response factor 18 (ARF18).

Li et al. (2019) reported significant differential expression of genes related to pod development, seed development, cell division, nutrient reservoir and ribosomal proteins in the pod wall and seeds of the large-seed pool and cloned two pleiotropic major QTLs which indirectly influenced seed weight via their effects on siliqua length. Recently, one QTL (cqSW.A03-2) explaining 8.46–13.7% of the phenotypic variation was reported for seed weight (Wang et al. 2020). Zhu et al. (2020) validated and characterized seed number per siliqua QTL qSN.A7 and demonstrated that the difference in single nucleotide polymorphisms (SNPs) between the two homologous near-isogenic lines was determined by the embryonic genotypic effect. However, still more efforts are required to unfold genetic variation for these traits particularly in the diverse population, identify significant QTLs and associated markers that can be used in marker-assisted breeding programme. With availability of whole genome sequence of important crop plants, next-generation sequencing technologies based high-throughput genotyping platforms, GWAS has emerged as an important approach to identify putative QTLs for complex quantitative traits in diverse set of populations. In this study, a collection of geographically diverse germplasm, which was maintained through sowing and continuous selfing in alternate years, was evaluated in multi-environment across locations and years to identify putative genetic loci influencing seed yield and its associated traits using GWAS.

Materials and methods

Plant material

A diverse set of 200 genotypes comprised of 113 lines from different countries namely, Australia (26), Canada (7), China (37), France (23), Poland (2), South Korea (1), Sweden (6), United States (11), India (85 advanced breeding lines and two commercial cultivars developed at Punjab Agricultural University (PAU), Ludhiana) (Table 1, Table S1, Fig. 1). The stock is being maintained through continuous selfing. The germplasm set was evaluated for test traits from October to April during 2017–18 (Y1) and 2018–19 (Y2) at two locations in India viz. Punjab Agricultural University (PAU, Latitude: 30.90° N, Longitude: 75.81° E, Altitude: 247 m) marked as LDH_Y1 and LDH_Y2, respectively and at Regional Research Station, PAU, Bathinda (Latitude: 30.21° N, Longitude: 74.95° E, Altitude: 210 m) marked as BTI_Y1 and BTI_Y2, respectively. These two locations represent the different agroclimatic zones of Punjab state in India- Ludhiana located in the central plain zone and Bathinda in western zone. The soil is sandy loam to clayey with normal reaction (pH from 7.8 to 8.5). Bathinda (BTI) receives relatively low average annual rainfall and higher maximum and minimum temperatures than Ludhiana (LDH) (Kaur and Hundal 2008; Kaur and Kaur 2015). Variation of minimum, maximum temperature and rainfall pattern between Ludhiana and Bathinda from October to April in year 2018–19 was presented in Fig. S1.

Table 1.

Number of genotypes collected from different countries for developing a diverse set for GWAS

S. no.	Country	No. of genotypes
1	Australia	26
2	Canada	7
3	China	37
4	France	23
5	Poland	2
6	South Korea	1
7	Sweden	6
8	United states	11
9	India	87
Total		200

Fig. 1.

Geographical distribution of germplasm stock of 200 genotypes

Phenotyping of germplasm stock for test traits

In all the four environments (LDH_Y1, LDH_Y2, BTI_Y1 and BTI_Y2), the field experiments were conducted. In each experiment, each genotype of 200 germplasm stock was sown in paired rows of two meters row length at a row to row and plant to plant spacing of 45 cm and 10 cm, respectively in alpha lattice design with two replications (Anonymous 2020). At maturity, the crop was harvested manually. Five siliquae were taken from the middle of the main shoot of five plants of each genotype and siliqua length (SL, cm) was measured in centimeteres and averaged. Number of seeds per siliqua (NSS) was counted as an average number of seeds from ten siliqua borne on the middle of the main shoot of five plants. Thousand seed weight (TSW, g) was measured as weight in grams of thousand fully developed seeds from the main shoot. The seed weight values for five plants were averaged. Seed yield was recorded on a plot basis. Seed yield per plant (SY, g) was calculated by dividing total plot yield by the number of plants in the plot. TSW and SY were recorded after proper drying of seeds (Sandhu et al. 2019).

Statistical analysis of phenotypic data

Analysis of variance (ANOVA), coefficient of variance (CV) and BLUPs (best linear and unbiased predictors) were computed using META-R version 6.0 statistical software (Alvarado et al. 2020). BLUPs for each genotype were calculated for each environment viz. LDH_Y1, LDH_Y2, BTI_Y1 and BTI_Y2 as well as pooled over years for each location (E1: Ludhiana, E2: Bathinda) and used for GWAS analysis. Following linear model for analysing individual environments was used:

$$Y_{\mathit{ijk}}=\mu +{\mathit{Rep}}_i+{\mathit{Block}}_j\left({\mathit{Rep}}_i\right)+{\mathit{Gen}}_k+{\epsilon }_{\mathit{ijk}}$$

where Y_ijk is the trait of interest, μ is the overall mean effect, Rep_i is the effect of the ith replicate, Block_j(Rep_i) is the effect of the j^th incomplete block within the i^th replicate, Gen_k is the effect of the kth genotype and ε_ijk is the effect of the error associated with the i^th replication, jth incomplete block, and kth genotype, which is assumed to be iid normal with mean zero and variance ${\sigma }_{\epsilon }^2$.

For, a combined analysis across years, following linear model was used:

$$Y_{\mathit{ijk}}=\mu +{\mathit{Yr}}_i+{\mathit{Rep}}_j\left({\mathit{Yr}}_i\right)+{\mathit{Block}}_k\left({\mathit{Yr}}_i,{\mathit{Rep}}_j\right)+{\mathit{Gen}}_l+{\mathit{Yr}}_i\times {\mathit{Gen}}_l+{\epsilon }_{\mathit{ijk}}$$

where, the new terms ${\mathit{Yr}}_i$ and ${\mathit{Yr}}_i\times {\mathit{Gen}}_l$ are the effects of the ith environment and the environment × genotype (G × E) interaction, respectively. In both models, all effects, except the overall mean, are declared to be random and iid normal with mean zero and effect-specific variances. The random assumption for the genotype effects allowed to calculate BLUPs.

Genotyping of GWAS panel

DNA extraction, GBS and SNP calling

A representative diverse panel of 125 lines out of stock of 200 genotypes was selected based on geographic origin, pedigree and variation in recorded values of test traits. Genomic DNA was isolated from young leaf tissue of 125 genotypes using CTAB (cetyl trimethyl ammonium bromide) method (Doyle et al. 1990). Agarose gel electrophoresis (0.8%) and NanoDrop 8000 spectrophotometer (Thermo Scientific Waltham, Massachusetts, United States) were used to examine the quantity and quality of DNA. Genotyping by sequencing (GBS) was outsourced from “Nucleome Informatics Pvt Ltd, India” and pair-end sequencing was performed on Illumina HiSeq™ 2500 platform. BWA software was used for alignment of quality clean sequencing data of each accession with B. napus ‘ZS11’ v2.0 reference genome {chromosome assembly was downloaded from NCBI (http://www.ncbi.nlm.nih.gov/genome/term=txid3708orgn)} to build the consensus genomic sequence of each genotype (Li and Durbin 2009). SAMtools were used for handling alignment files and variants were called based on the discrepancies between the consensus sequence and the reference genome using BCFtools (Li et al. 2009). SNPs having MAF < 0.05, missing data > 10%, and multiple alleles were filtered out using VCFtools (Danecek et al. 2011). Further, SNPs having up to 50% heterozygosity were only used for the GWAS analysis.

Genome-wide association analyses

A total of 85,126 filtered SNPs as genotypic data and BLUPs datasets as phenotypic data of test traits were used for GWAS using FarmCPU (Fixed and Random Model Circulating Unification) statistical method implemented through GAPIT version 3 (Genomic Association and Prediction Integrated Tool) (Lipka et al. 2012; Liu et al. 2016). FarmCPU is a multi-locus model approach and addresses the confounding problems of mixed linear models (MLM) by using both fixed effect model and the random effect model iteratively. R version 3.6.3 was implemented in RStudio software (R Core Team 2018; RStudio Team 2020). Principal components (PCs) and kinship computed from genotypic SNP data were incorporated in the FarmCPU model. Uniform threshold (−log10(p) = 3.74) was set to control the genome-wide type 1 error rate and to identify significant SNPs associated with test traits in all environment datasets (Duggal et al. 2008; Yang et al. 2014).

Identification of candidate genes

Linkage disequilibrium (LD) analysis was performed across the genome using PopLDdecay version 3.41 (Zhang et al. 2019) and a cut off of r² = 0.1 was used for determining LD decay. DNA sequence from 150 kb genomic region on either side of significant SNPs was considered as confidence interval for comparison with previous studies. Information about genes present in the confidence interval of SNPs was obtained from NCBI (https://www.ncbi.nlm.nih.gov/genome/browse/#!/proteins/203/335272%7CBrassica%20napus/) and functions of the predicted candidate genes were reviewed to establish their importance for improving test traits in B. napus. The list of softwares/tools used in this study was given in Table S2.

Results

Variation for test traits in germplasm stock of B. napus in multiple environments

• (a) Siliqua length (SL): Genotypic variance and genotype-by environment interaction variance was significant (p < 0.001) for SL in E1, E2 and across all environments (Table 2). Comparative variation of SL between two locations (Ludhiana and Bathinda) was presented in Fig. 2a. Top ten genotypes with highest SL were given in Table 3. Maximum SL was 8.26 cm for genotype “BLN-33–51” from Australia origin.

• (b) Number of seeds per siliqua (NSS): Genotypic variance and genotype-by environment interaction variance was significant (p < 0.001) for NSS in E1, E2 and across all environments (Table 2). Comparative variation of NSS between two locations (Ludhiana and Bathinda) was presented in Fig. 2b. Top ten genotypes with highest NSS were given in Table 3. Maximum NSS was 27 for genotype “PN-78–7-2” from advanced breeding lines.

• (c) Thousand seed weight (TSW): Genotypic variance and genotype-by environment interaction variance were significant (p < 0.001) for TSW in E1, E2 and across all environments (Table 2). Comparative variation of TSW between two locations (Ludhiana and Bathinda) was presented in Fig. 2c. Top ten genotypes with highest TSW were given in Table 3. Maximum TSW was 4.61 g for genotype “CHARLTON” from Australia origin.

• (d) Seed yield per plant (SY): Genotypic variance and genotype-by environment interaction variance were significant (p < 0.001) for TSW in E1, E2 and across all environments (Table 2). Comparative variation of SY between two locations (Ludhiana and Bathinda) was presented in Fig. 2d. Top ten genotypes with the highest SY were given in Table 3. Maximum SY was 15.52 g for genotype “BOOMER” from Canada origin.

Table 2.

Analysis of variance (ANOVA) of seed yield and component traits in multi-environments

Trait†	Env§	${\sigma }_G^2$	${\sigma }_{\mathit{GXE}}^2$	Mean	LSD (5%)	CV
SL	E1	0.20***	0.07***	6.76	0.48	6.39
	E2	0.21***	0.09***	6.93	0.48	5.12
	Across	0.21***	0.07***	6.85	0.36	5.8
NSS	E1	4.03***	5.46***	22.28	2.84	11.14
	E2	4.89***	6.7***	21.87	3.05	10.56
	Across	4.89***	5.64***	22.08	2.4	10.9
TSW	E1	0.15***	0.02***	3.44	0.23	3.96
	E2	0.28***	0.12***	3.43	0.47	6.21
	Across	0.11***	0.17***	3.43	0.36	5.21
SY	E1	5.97***	0.09	12.58	1.12	8.23
	E2	3.73***	1.12***	11.71	1.64	8.63
	Across	3.75***	1.72***	12.14	1.4	8.49

${\sigma }_G^2$ = Genotype variance, ${\sigma }_{\mathit{GxE}}^2$ = G X E variance, LSD = Least significant difference, CV = Coefficient of variance

^*** = Significance at p < 0.001

^†SL: Siliqua length (cm), NSS: Number of seeds per siliqua, TSW: Thousand-seed weight (g), SY: Seed yield per plant (g)

^§ENV: Environment, E1: Ludhiana pooled over two years, E2: Bathinda pooled over two years, Across: Pooled across all environments

Fig. 2.

a Boxplots presenting variation between Ludhiana (E1) and Bathinda (E2) pooled over two years for siliqua length (SL). b Boxplots presenting variation between Ludhiana (E1) and Bathinda (E2) pooled over two years for number of seeds per siliqua (NSS). c Boxplots presenting variation between Ludhiana (E1) and Bathinda (E2) pooled over two years for thousand-seed weight (TSW). d Boxplots presenting variation between Ludhiana (E1) and Bathinda (E2) pooled over two years for seed yield per plant (SY)

Table 3.

Top ten genotypes with highest observations for test traits based on BLUPs (across four environments: two years at two locations)

S. no.	Genotype	SL†	Genotype	NSS†	Genotype	TSW†	Genotype	SY†
1	BLN-33–51	8.26*	PN-78–7-2	27*	CHARLTON	4.61*	BOOMER	15.52*
2	ZING-821	8.03*	MYSTIC	26*	ZY-0–14	4.28*	P-617	15.07*
3	P-617	7.75*	03-P-74–6	26*	USCN-56	4.13*	HYOLA-75	14.82
4	ACN-39	7.63*	06–06-3737	26*	P-617	4.06*	CHARLTON	14.76
5	BLN-33–44	7.57	EC-28960	26*	ACN-50	4.06*	LCN-10	14.73
6	PN-47–1	7.56	EC-609308	25*	PN-84–2	4.03*	ACN-17	14.45
7	DING-474	7.54	FM-24	25*	ZY-0–12-4	3.95*	VCN-62	14.41
8	FM-7	7.49	FM-35	25*	BCN-58	3.92*	ZING-SHU-ANG-4	14.39
9	MYSTIC	7.49	BLN-33–50	25*	PN-87–3	3.92*	NE-371	14.39
10	RR-009	7.47	P-617	25*	FAN-023	3.89*	FAN-028	14.09
Check	GSC-7	7.22	GSC-7	20	GSC-7	3.49	GSC-7	13.54
	LSD (5%)	0.36	LSD (5%)	2.40	LSD (5%)	0.36	LSD (5%)	1.40

^*Significantly different from check with α = 0.05

^†SL, Siliqua length (cm); NSS, Number of seeds per siliqua; TSW, Thousand seed weight (g); SY, Seed yield per plant (g)

Performance of important productivity component traits (i.e., SL, NSS, TSW) for ten high productivity (SY) elite genotypes across the environments was given in Table 4. These genotypes were from diverse origins i.e., Canada (BOOMER), China (P-617 and ZING-SHU-ANG-4), Australia (HYOLA-75 and CHARLTON), France (FAN-028), United States (NE-371) and India (LCN-10, ACN-17, VCN-62). Pearson correlation coefficient was presented in Table 5. SYs have significant positive correlation with NSS (r = 0.15) and TSW (r = 0.23). SLs have weak correlation with SY. However, SLs have significant positive correlation with NSS (r = 0.38) and TSW (r = 0.16).

Table 4.

Trait performance of elite genotypes across four environments (two years at two locations)

S. no.	Genotype	Origin	Traits†
			SY	NSS	SL	TSW
1	BOOMER	Canada	15.52*	22.48*	7.19	3.00
2	P-617	China	15.07*	25.17*	7.75*	4.06*
3	HYOLA-75	Australia	14.82	22.45*	6.84	3.16
4	CHARLTON	Australia	14.76	16.00	5.89	4.61*
5	LCN-10	India	14.73	20.62	6.54	3.33
6	ACN-17	India	14.45	22.40*	6.90	3.50
7	VCN-62	India	14.41	19.47	6.38	3.02
8	ZING-SHU-ANG-4	China	14.39	21.12	6.94	3.61
9	NE-371	United states	14.39	22.96*	6.90	3.51
10	FAN-028	France	14.09	22.91*	7.00	3.45
Check	GSC-7	India	13.54	19.91	7.22	3.49
	LSD (5%)		1.40	2.40	0.36	0.36

^*Significantly different from check with α = 0.05

^†SY: Seed yield per plant (g), SL: Siliqua length (cm), NSS: Number of seeds per siliqua, TSW: Thousand seed weight (g)

Table 5.

Pearson correlation coefficients of test traits in B. napus

Traits†	SL	NSS	TSW
NSS	0.38*
TSW	0.16*	− 0.08
SY	0.05	0.15*	0.23*

^*Significance at α = 0.05

^†SL, Siliqua length (cm); NSS, Number of seeds per siliqua, TSW, Thousand seed weight (g); SY, Seed yield per plant (g)

Genotype-by-sequencing of B. napus genotypes

A diverse association mapping panel of 125 genotypes was used for GWAS studies (Fig. 3a). A total of 245.6 to 964.78 million raw reads were obtained through the GBS of association mapping panel (AMP). GC percent ranged from 35.13% to 38.89% with Q₂₀ and Q₃₀ values ranging from 94.75 to 98.33 and 87.37 to 95.64, respectively. After alignment with B. napus reference genome and variant calling, a total 18,41,123 variants (SNPs and Indels) from the whole genome of B. napus were obtained. Out of which, 1,03,274 bi-allelic SNPs having < 10% missing data and MAF ≥ 5% were obtained after filtering. These SNPs were located on 19 chromosomes of B. napus. Out of these, 18,148 SNPs were excluded due to more than 50% heterozygosity and 85,126 SNPs were finally used for GWAS. SNPs were found to be almost equally distributed over the two genomes (A and C) of B. napus (Table 6). Total 42,252 SNPs were obtained on ten chromosomes of genome A with the highest number on chromosome A03 and the least number on chromosome A04. Similarly, on genome C, total 42,874 SNPs were found on nine chromosomes with maximum SNPs obtained on chromosome C03, whereas the minimum number of SNPs found on chromosome C06. Linkage disequilibrium (LD) decay in population was estimated to be ~ 150 kb where r² value drops below 0.1 (Fig. 3b).

Fig. 3.

a Distribution of 125 genotypes from different geographical locations based on Principle Component Analysis (PCA) plot of 85,126 SNPs. b Linkage disequilibrium (LD) decay plot in the genome of B. napus based on 85,126 SNPs in 125 B. napus diverse lines

Table 6.

Genome-wide distribution of SNPs on all chromosomes of B. napus

Chromosome	Size (Mb)§	Number of markers
A01	35.82	4287
A02	35.33	4791
A03	49.14	5727
A04	23.56	2834
A05	31.47	4267
A06	36.08	4867
A07	27.43	4021
A08	27.74	3006
A09	45.94	5298
A10	22.22	3154
C01	50.78	4682
C02	68.4	4782
C03	80.38	7214
C04	70.55	6402
C05	44.14	3336
C06	45.51	3256
C07	62.44	5691
C08	46.35	4086
C09	51.7	3425
Total	854.98	85,126

^§Size of B. napus reference genome in this study

Marker-trait associations for test traits

Siliqua length (SL)

Total 20 significant SNPs were identified for SL, out of which six were found in E1, eight in E2 and six in pooled across environment BLUPs (Table 7). Out of 20 SNPs, one SNP (A05.1643587) from E1 and three SNPs (C03.56028089, C06.29644578, C07.36219105) from E2 were consistently identified in pooled across environment BLUP. For 12 SNPs minor allele had favourable effect (allelic effect: 0.18 to 0.28) and for eight SNPs major allele had favourable effect (allelic effect: 0.18 to 0.4). Manhattan plots depicting marker-traits associations of SL were presented in Fig. 4a.

Table 7.

Significant marker-trait associations for siliqua length based on BLUPs (E1, E2, Pooled across four environments: two years at two locations)

Env†	SNPs associated	Chromosome	Position	MAF	Effect§	−log10(p)
E1	A09.18413957	A09	18,413,957	0.39	0.19	4.09
	A05.16435872	A05	1,643,587	0.38	0.18	4.04
	A01.1223282	A01	1,223,282	0.47	− 0.18	4.01
	A08.12100271	A08	12,100,271	0.20	0.22	3.89
	C04.17855316	C04	17,855,316	0.19	0.27	3.81
	A01.1108249	A01	1,108,249	0.28	− 0.18	3.80
E2	C02.46888810	C02	46,888,810	0.15	− 0.29	4.24
	C03.560280893	C03	56,028,089	0.10	− 0.38	4.21
	A05.31053566	A05	31,053,566	0.37	− 0.22	4.16
	C07.43257273	C07	43,257,273	0.27	0.27	4.09
	C07.48993159	C07	48,993,159	0.26	0.24	3.90
	C07.362191053	C07	36,219,105	0.13	− 0.32	3.86
	C06.296445783	C06	29,644,578	0.28	0.18	3.85
	C09.50218707	C09	50,218,707	0.16	0.28	3.78
Pooled	C07.362191053	C07	36,219,105	0.13	-0.35	3.98
	C06.296445783	C06	29,644,578	0.28	0.20	3.97
	C03.560280893	C03	56,028,089	0.10	− 0.40	3.95
	A08.24703686	A08	24,703,686	0.41	0.20	3.82
	A05.16435872	A05	1,643,587	0.38	0.20	3.79
	A01.14870209	A01	14,870,209	0.28	0.21	3.75

^†ENV, Environment; E1, Ludhiana pooled over two years, E2, Bathinda pooled over two years; Across, Pooled across all environments

²Common SNP between E1 and pooled, ³Common SNP between E2 and pooled

^§A positive allelic effect estimate indicates that the minor allele is favorable and a negative allelic effect estimate indicates that the major allele is favourable

Fig. 4.

Manhattan plot showing significant marker-trait associations (MTAs) for E1, E2 and pooled across environments: a Siliqua length (SL), b Number of seeds per siliqua (NSS), c Thousand-seed weight (TSW), d Seed yield per plant (SY). *Chromosome 1 to 10 represents A genome mentioned as A01 to A10 and chromosome 11 to 19 represents C genome mentioned as C01 to C09 in manuscript

Number of seeds per siliqua (NSS)

There were a total of 23 significant SNPs for NSS, out of which seven were found in both E1 and E2 and nine in pooled (Table 8). Out of 23, four SNPs were consistent across environments. one SNP (C03.496706371) was present across all environments (E1, E2, pooled). Two SNPs (A04.5535255, C02.24437477) were common between E1 and pooled datasets. One SNP (A08.18254593) was common between E2 and pooled datasets. For 12 SNPs minor allele had favourable effect (allelic effect: 0.61 to 1.79) and for 11 SNPs major allele had favourable effect (allelic effect 0.6 to 1.25). Manhattan plots depicting marker-traits associations of NSS were presented in Fig. 4b.

Table 8.

Significant marker-trait associations for number of seeds per silique based on BLUPs (E1, E2, Pooled across four environments: two years at two locations)

Env†	SNPs associated	Chromosome	Position	MAF	Effect§	−log10(p)
E1	C02.244374772	C02	24,437,477	0.36	0.64	4.07
	A04.55352552	A04	5,535,255	0.30	–0.66	4.04
	A10.22094031	A10	22,094,031	0.34	0.63	4.02
	C01.10935066	C01	10,935,066	0.32	0.65	3.98
	C03.496706371,2	C03	49,670,637	0.32	–0.60	3.87
	C02.24309423	C02	24,309,423	0.36	-0.60	3.86
	A06.2046187	A06	2,046,187	0.43	0.67	3.76
E2	A04.19663920	A04	19,663,920	0.19	–0.89	4.31
	C03.496706371,3	C03	49,670,637	0.32	–0.67	4.05
	C04.38665294	C04	38,665,294	0.17	–0.97	4.04
	A02.29797443	A02	29,797,443	0.09	1.24	3.93
	A08.182545933	A08	18,254,593	0.40	0.62	3.90
	A03.4833681	A03	4,833,681	0.12	1.14	3.89
	A09.18412313	A09	18,412,313	0.49	0.61	3.76
Pooled	C03.496706372.3	C03	49,670,637	0.32	–0.89	5.37
	A08.182545933	A08	18,254,593	0.40	0.78	4.61
	C01.36201849	C01	36,201,849	0.08	1.79	4.38
	C02.244374772	C02	24,437,477	0.36	0.81	4.21
	C04.52890124	C04	52,890,124	0.19	–1.25	4.08
	A04.55352552	A04	5,535,255	0.30	–0.83	4.08
	A08.18028056	A08	18,028,056	0.36	–0.77	3.94
	C08.38958037	C08	38,958,037	0.12	–1.22	3.75
	A04.5535337	A04	5,535,337	0.30	0.79	3.75

^†ENV, Environment; E1: Ludhiana pooled over two years; E2: Bathinda pooled over two years; Across, Pooled across all environments

¹Common SNP between E1 and E2,

²Common SNP between E1 and pooled,

³Common SNP between E2 and pooled

^§A positive allelic effect estimate indicates that the minor allele is favorable and a negative allelic effect estimate indicates that the major allele is favourable

Thousand seed weight (TSW)

Total 30 SNPs were identified for TSW, out of which 11 were found each in both E1 and pooled while eight were in E2 (Table 9). Out of 30 SNPs, three (A03.42737659, A04.3187095, A10.91859) were common between E2 and pooled datasets. Major allele was favourable for 14 SNPs (allelic effect: 0.16 to 0.37) and 16 SNPs have favourable minor allele (allelic effect: 0.14 to 0.37). Manhattan plots depicting marker-traits associations for TSW were presented in Fig. 4c.

Table 9.

Significant marker-trait associations for thousand seed weight based on BLUPs (E1, E2, Pooled across four environments: two years at two locations)

Env†	SNPs associated	Chromosome	Position	MAF	Effect§	−log10(p)
E1	C02.41257489	C02	41,257,489	0.18	0.21	5.10
	C05.13774545	C05	13,774,545	0.09	− 0.29	4.58
	C02.42717465	C02	42,717,465	0.17	0.19	4.52
	A03.13569864	A03	13,569,864	0.20	− 0.17	4.47
	C02.43603089	C02	43,603,089	0.12	0.23	4.36
	C08.24798665	C08	24,798,665	0.29	0.16	4.35
	A03.13570024	A03	13,570,024	0.21	0.16	4.26
	C02.39280275	C02	39,280,275	0.13	− 0.20	4.21
	C02.43441429	C02	43,441,429	0.13	− 0.20	4.01
	C02.43540356	C02	43,540,356	0.16	0.19	3.95
	C02.41059840	C02	41,059,840	0.14	− 0.20	3.85
E2	A03.427376593	A03	42,737,659	0.27	0.33	5.68
	A04.31870953	A04	3,187,095	0.17	0.37	4.88
	A10.918593	A10	91,859	0.29	0.36	4.48
	A03.40403702	A03	40,403,702	0.28	0.30	4.23
	A10.9206788	A10	9,206,788	0.20	0.35	4.20
	A09.18695626	A09	18,695,626	0.30	− 0.26	4.13
	C03.32913880	C03	32,913,880	0.19	− 0.37	3.99
	C05.13684652	C05	13,684,652	0.23	− 0.35	3.98
Pooled	A10.918593	A10	91,859	0.29	0.20	4.70
	A03.427376593	A03	42,737,659	0.27	0.17	4.58
	A04.14864106	A04	14,864,106	0.14	− 0.20	4.23
	A03.34381273	A03	34,381,273	0.12	− 0.20	4.09
	A03.40590580	A03	40,590,580	0.22	− 0.16	4.02
	A04.31870953	A04	3,187,095	0.17	0.18	4.00
	C07.43724006	C07	43,724,006	0.10	− 0.20	3.99
	A10.630419	A10	630,419	0.40	0.14	3.91
	C02.41257479	C02	41,257,479	0.18	− 0.17	3.84
	A03.34381456	A03	34,381,456	0.12	− 0.19	3.82
	C06.20084594	C06	20,084,594	0.08	0.27	3.80

^†ENV, Environment; E1, Ludhiana pooled over two years; E2, Bathinda pooled over two years; Across, Pooled across all environments

³Common SNP between E2 and pooled

^§A positive allelic effect estimate indicates that the minor allele is favourable and a negative allelic effect estimate indicates that the major allele is favourable

Seed yield per plant (SY)

There are a total 18 SNPs that were found to be significant in this case. Out of the total of 18 significant SNPs, eight were identified in E1 and E2 each, while two SNPs were identified in the pooled dataset (Table 10). Major allele was favourable for eight SNPs (allelic effect: 0.91 to 1.34) and minor allele was favourable for ten SNPs (allelic effect: 0.14 to 0.37). Manhattan plots depicting marker-traits associations for SY were presented in Fig. 4d. No common SNPs were found across any environment data set in this case.

Table 10.

Significant marker-trait associations for seed yield per plant based on BLUPs (E1, E2, Pooled across four environments: two years at two locations)

Env†	SNPs associated	Chromosome	Position	MAF	Effect§	−log10(p)
E1	A02.19604813	A02	19,604,813	0.14	1.72	4.13
	C09.6453403	C09	6,453,403	0.23	− 1.34	4.05
	C07.59073470	C07	59,073,470	0.39	− 1.19	3.98
	A10.15563864	A10	15,563,864	0.06	2.52	3.97
	C02.67336115	C02	67,336,115	0.19	− 1.08	3.94
	C03.27718223	C03	27,718,223	0.50	1.01	3.88
	A02.19263168	A02	19,263,168	0.18	1.47	3.88
	C02.17346965	C02	17,346,965	0.43	− 1.12	3.79
E2	C01.47142794	C01	47,142,794	0.15	1.34	4.57
	A08.27703652	A08	27,703,652	0.11	1.13	4.32
	A03.46280715	A03	46,280,715	0.06	1.86	4.28
	A10.11535749	A10	11,535,749	0.24	0.73	4.08
	A03.43617362	A03	43,617,362	0.06	1.47	4.04
	A10.19925371	A10	19,925,371	0.29	− 0.96	3.93
	A02.23510581	A02	23,510,581	0.28	− 0.91	3.92
	C08.45226147	C08	45,226,147	0.17	− 1.07	3.81
Pooled	C02.22109942	C02	22,109,942	0.14	− 1.11	4.09
	C01.41279341	C01	41,279,341	0.14	0.99	3.84

^†ENV, Environment; E1, Ludhiana pooled over two years; E2, Bathinda pooled over two years; Across, Pooled across all environments

^§A positive allelic effect estimate indicates that the minor allele is favourable and a negative allelic effect estimate indicates that the major allele is favourable

Identification of candidate genes

LD decay tends to be stable when the distance is 150 kb, therefore, the candidate region is defined as the genomic region located within 150 kb upstream and downstream of the genome-wide significant SNPs. The candidate genomic region of significant SNPs was inspected for putative candidate genes for each test trait and presented in Table 11. For SL, four putative candidate genes viz. SPL2, ATS3A, CKX1 and SPL9 were identified on two chromosomes A09 (SNP: A09.18413957) and A05 (SNP: A05.1643587). For NSS, five candidate genes were identified located on chromosomes A06 and C02. Genes PERK7 and GATA15 were identified on chromosome A06 in the candidate region of SNP “A06.2046187”. Chromosome C02 has two significant SNPs, which have overlapping candidate regions. Genes NFD6 and PERK13 were identified in the candidate region of SNP “C02. 24,309,423”, while gene PRK3 was identified in the candidate region of SNP “C02. 24,437,477”.

Table 11.

Gene annotation in the confidence interval of significant SNPs associated with test traits

Traits†	Chr§	Position		Locus	Gene	Protein Name
		Start	End
SL	A09	18,412,161	18,413,643	LOC106447058	SPL2	squamosa promoter-binding-like protein 2
SL	A05	1,497,493	1,498,487	LOC106450668	ATS3A	embryo-specific protein ATS3A-like
SL	A05	1,508,609	1,510,590	LOC106450662	CKX1	cytokinin dehydrogenase 1-like
SL	A05	1,693,622	1,695,604	LOC106450633	SPL9	squamosa promoter-binding-like protein 9
NSS	A06	1,946,277	1,949,436	LOC106351125	PERK7	proline-rich receptor-like protein kinase PERK7
NSS	A06	2,075,018	2,075,610	LOC106346114	GATA15	GATA transcription factor 15-like
NSS	C02	24,188,687	24,189,457	LOC106404610	NFD6	protein NUCLEAR FUSION DEFECTIVE 6
NSS	C02	24,245,382	24,250,169	LOC106402973	PERK13	proline-rich receptor-like protein kinase PERK13
NSS	C02	24,467,574	24,469,776	LOC106357053	PRK3	pollen receptor-like kinase 3
TSW	A03	13,420,280	13,421,490	LOC106438015	SRC2	protein SRC2 homolog
TSW	A03	13,451,242	13,452,029	LOC106438007	WAT1	WAT1-related protein At2g40900-like
TSW	A03	13,489,152	13,492,022	LOC106438000	AVT1A	amino acid transporter AVT1A
TSW	A03	13,524,554	13,524,829	LOC106443562	OSR1	protein ORGAN SIZE RELATED 1-like
TSW	A03	13,550,794	13,552,034	LOC106443560	ARR8	two-component response regulator ARR8-like
TSW	A03	13,609,966	13,612,015	LOC106437984	CKX1	cytokinin dehydrogenase 1-like
TSW	A03	13,652,828	13,655,279	LOC106428105	At2g41710	AP2-like ethylene-responsive transcription factor At2g41710
TSW	A03	42,703,613	42,704,499	LOC106444685	REM9	B3 domain-containing protein REM9-like isoform X2
TSW	A03	42,703,613	42,704,474	LOC106444685	REM7	B3 domain-containing protein REM7-like isoform X1
TSW	C02	40,921,964	40,923,997	LOC106383582	BnC1	cruciferin BnC1-like
TSW	C02	40,924,993	40,930,838	LOC111203134	BnC2	cruciferin BnC2-like
SY	A02	19,489,729	19,490,450	LOC106411812	VPS45	vacuolar protein sorting-associated protein 45 homolog
SY	A02	19,727,985	19,730,528	LOC106397017	WAT1	WAT1-related protein At1g44800-like
SY	C02	21,998,836	21,999,801	LOC111202902	BG1	protein BIG GRAIN 1-like E
SY	C02	22,027,303	22,028,455	LOC111202903	SPL6	squamosa promoter-binding-like protein 6
SY	C02	22,037,557	22,038,378	LOC106441608	AGL82	agamous-like MADS-box protein AGL82
SY	C02	22,192,124	22,195,820	LOC106372743	LHT1	lysine histidine transporter 1

^†NSS = number of seeds per silique, SL: silique length, TSW: thousand seed weight, SY: seed yield per plant

^§Chr: Chromosome

For seed weight, presented as TSW, 13 candidate genes were identified on three chromosomes viz. A03, A09 and C02. Total nine candidate genes were identified on chromosome A03, out of which, seven genes viz. SRC2, WAT1, AVT1A, OSR1, ARR8, CKX1 and At2g41710 were located in the candidate region of SNP “A03.13569864”, while two genes REM9 and REM7 were located in the candidate region of SNP “A03.42737659”. Both chromosomes A09 and C02 have two candidate genes. LTP1 and PAH2 were present in the candidate region of SNP “A09.18695626”, while BnC1 and BnC2 were located in the candidate region of SNP “C02.41059840”. For SY, six genes were identified on two chromosomes viz. A02 and C02. Out of these, two genes viz. VPS45 and WAT1 were located on chromosome A02 in the candidate region of SNP “A02.19604813” and four genes viz. BG1, SPL6, AGL82 and LHT1 were identified in the candidate region of SNP C02.22109942 located on chromosome C02.

Discussion

B. napus being domesticated only during the recent 400–500 years in Brassicaceae family (Prakash et al. 2012), has narrow genetic base and hence, a hurdle for extensive genetic enhancement. The dual bottlenecks of polyploidy and domestication, and participation of a small number of progenitor plants in the ancient hybridization events are the probable reasons put forward. Intensive plant breeding activities for canola quality may have further eroded the variation for traits of high breeding interest (Hasan et al. 2006) particularly seed yield. Siliqua length (SL), number of seeds per siliqua (NSS), thousand-seed weight (TSW) are important components of seed yield per plant (SY) in B. napus which were investigated in a set of diverse germplasm stock to identify potential donors for these traits, associated SNPs and candidate genes. Out of 200, ten elite genotypes having the highest yield were presented in Table 4. Out of these, genotype P-617 from China has the highest NSS (~ 25), SL (7.75 cm) and second-highest TSW (4.06 g) and SY (15.07 g). Based on traits studied P-617 was recommended for commercial breeding programme. Based on variation for test traits across environments, a set of diverse germplasm stock collected from different geographical locations was used to constitute AMP, based on variation for test traits across environments. The presence of G X E interaction for test traits was found to be highly significant in ANOVA. Therefore, BLUPs were calculated and were used for GWAS.

GWAS and candidate gene identification

Significant variation and higher rate of LD decay (150 kb) presented in this panel made it highly suitable for GWAS analysis. FarmCPU uses the most significant markers as covariates in the GWAS model (Liu et al. 2016), therefore, SNPs are rarely identified within the same LD block for an environment-specific dataset (Steketee et al. 2019). Comparative studies of various models for GWAS analysis have demonstrated better performance and statistical power of FarmCPU in controlling both false positives and false negatives and indicated consistent results in identifying significant SNPs (Kaler et al. 2020). This study identified 11 loci (four for SL and NSS each, three for TSW), which were consistently identified both in individual environments (E1, E2) as well as in pooled datasets (Table 7a-d). For SL, SNP “C03.56028089” had the largest allelic effect (−0.40) with favourable major allele and was consistently identified in E2 and pooled datasets. For NSS, the largest allelic effect was 1.79 and minor allele was favourable for SNP “C01.36201849”. For TSW, SNPs “A04.3187095” and “C03.32913880” had largest allelic effect (0.37 each) with favourable minor and major alleles, respectively. SNP “A04.3187095” had significant association in both E2 and pooled datasets. SNP “A10.15563864” had the largest allelic effect (2.52) with favourable minor allele for SY.

The development of seeds in flowering plants is placed under complex interactions between maternal tissues, the embryo, and the endosperm. Functional annotation of the associated SNPs and surrounding genome space(s) helped to predict candidate genes for test traits. Important among these were the genes encoding auxin response factors, regulators of cytokinin signalling, seed storage proteins, transports related proteins and transcription factors and protein involved in successful fertilization. Important candidate genes were discussed as follows:

Candidate genes for SL

Total four candidate genes were identified for siliqua length, one of them was located on chromosome A09 and the other three were located on chromosome A05. Out of which, two genes encoding Squamosa promoter binding protein-like (SPL) proteins were identified in this study i.e., SPL2 located on chromosome A09 (SNP: A09.18413957) and SPL9 located on chromosome A05 (SNP: A05.1643587).

SPL proteins are plant-specific transcription factors and play critical roles in plant growth and development. The functions of many SPL gene family members were well characterized in various plant species, such as Arabidopsis (Cardon et al. 1999), rice (Xie et al. 2006; Wang et al. 2012) and wheat (Zhang et al. 2014). Cao et al. (2019) investigated ectopic expression of TaSPL16 gene from wheat in Arabidopsis thaliana and reported significantly longer siliquae, more seeds per siliqua and larger seed size. In Arabidopsis, siliquae of agl15, hda3, tub6, spl15, at5g10625, and at3g01323 mutants showed a significant reduction in length (Mizzotti et al. 2018).

On chromosome A05, gene ATS3A was an embryo-specific gene and expressed in a pattern similar to the Arabidopsis seed storage protein genes (Nuccio and Thomas 1999). Another gene CKX1 (LOC106450662) codes for cytokinin dehydrogenase, which catalyses the oxidation of cytokinin. Liu et al. (2018) reported higher expression of cytokinin dehydrogenase in reproductive organs e.g., buds, flowers or siliquae and their role in siliqua length, siliqua development and stress responses in B. napus.

Bennett et al. (2011) reported that siliqua is not only a photosynthetic source organ, but also serves to coordinate seed filling, regulates the reallocation of reserves, and protects seeds against biotic and abiotic stresses. Li et al. (2019) cloned two pleiotropic major QTLs, that acted indirectly on seed weight via their effects on siliqua length. Previous studies have reported pleiotropic effect of candidate genes, identified in this study, for SL viz., SPL gene family (Cao et al. 2019), ATS3A (Nuccio and Thomas 1999), CKX1 (Liu et al. 2018). Pleiotropic effect of identified candidate genes and significant positive correlation of SL with yield-component traits such as NSS (r = 0.38) and TSW (r = 0.16) suggested the importance of SL for SY improvement through marker assisted selection (MAS).

Candidate genes for TSW

Seed size that determines the seed weight, is one of the key traits for seed yield. The growth regulator auxin is known to have an important role in the regulation of different cellular processes involved in seed development (Cao et al. 2020). Auxin response factors (ARFs) are the transcription factors having the capability of recognizing AUXIN RESPONSE ELEMENTs (AuxREs) within the promoter regions of downstream genes and regulating their expression (Ulmasov 1997, 1999). Previous studies reported the role of ARFs in influencing seed size in various crops e.g., Camelina sativa (Na et al. 2019), rice (Liu et al. 2015b) and B. Napus (Liu et al. 2015a). In the previous study, Liu et al. (2015a) identified the role of ARF18 in determining final seed weight by regulating the classic auxin signalling pathway during siliqua development. Our study identified two genes REM9 and REM7 coding for B3 domain containing proteins ARF23 and ARF25 in the candidate region of SNP “A03.42737659” associated with TSW.

This study identified a gene (At2g40900-like) coding for WAT1-related protein in the candidate region of SNPs “A03.13569864” and “A03.13570024” associated with TSW. Ranocha et al. (2013) have established that Arabidopsis thaliana WAT1 (WALLS ARE THIN1) encodes a novel auxin transporter in plants and showed that WAT1 exports IAA from the vacuole to the cytoplasm in plant cells.

The high level of cytokinin activity found in endosperm suggested that the plant hormone cytokinin regulates growth of seed components and controls seed mass or yield (Letham 1973; Mok and Mok 2001; Day et al. 2008). In our study, two genes related to cytokinin signalling were identified in candidate regions of SNPs “A03.13569864” and “A03.13570024”. Gene ARR8-like (LOC106443560) codes for protein “two-component response regulator” which are primary cytokinin response targets and possibly involved in cytokinin signalling (Li et al. 2013). Siliqua pericarp provides photosynthates to the developing seeds and therefore, genes regulating siliqua length may lead to increased seed weight in B. napus (Liu et al. 2015a). In our study, CKX1 gene encoding cytokinin dehydrogenase was identified for TSW on chromosome A03 (LOC106437984) as well as for SL on chromosome A05 (LOC106450662).

Feng et al. (2011) suggested the role of phytohormones in regulating organ growth and final organ size by modifying the expression of different ORGAN SIZE RELATED (OSR) genes such as OSR1, ARGOS and ARL. ARGOS is induced by auxin and cytokinin, ARL is induced by brassinosteroid (BR), while OSR1 is induced by ethylene but repressed by abscisic acid and brassinosteroid. Our study identified gene (LOC106443562) located on chromosome A03 encoding protein ORGAN SIZE RELATED 1-like in the candidate region of SNPs “A03.13569864” and “A03.13570024” associated with TSW. Another gene (LOC106428105) encoding “AP2-like ethylene-responsive transcription factor At2g41710 '' was identified 128 kb downstream of OSR1.

The seeds of B. napus contain 25% seed storage proteins (SSPs), out of which, cruciferin (60%) is major storage protein followed by napin (20%) and other minor proteins (Gehrig et al. 1996). Our study identified two genes viz. BnC1 and BnC2 coding for seed storage protein “Cruciferin” in the confidence interval of SNP “C02.41059840” associated with TSW. Seed storage proteins are synthesized in the endoplasmic reticulum (ER) and transported to protein storage vacuoles (PSV) where they are primarily accumulated (Hohl et al. 1996; Hara-Nishimura et al. 1998). Our study identified gene SRC2 (LOC106438015) coding for “Protein SRC2 homolog” in the candidate region of co-localized SNPs “A03.13569864” and “A03.13570024” associated with TSW. Oufattole et al. (2005) identified a type II endoplasmic reticulum (ER) membrane protein (At)SRC2 in the protein storage vacuole (PSV) of mature seeds of Arabidopsis thaliana and identified its preferential interaction with a targeting motif specific for the ER-to-vacuole pathway. In the same candidate region, gene AVT1A coding for amino acid transporter was also identified, its role was reported in vacuolar amino acid transport in Arabidopsis thaliana (Fujiki et al. 2017).

Candidate genes for NSS

Five important candidate genes viz. PERK7, GATA transcription factor 15-like, NFD6, PERK13 were identified for NSS. Out of these, PERK7 and GATA transcription factor 15-like” were located in the candidate region of SNP “A06.2046187” and PERK13 and NFD6 in the candidate region of SNP “C02.24437477″. Gene PERK7 encodes proline-rich receptor-like kinase protein and NFD6 encodes NUCLEAR FUSION DEFECTIVE 6. Haffani et al. (2006) have studied ectopic expression and antisense suppression using the BnPERK1 cDNA and reported significant increase in the number of ovules per pistil and seeds per siliqua in Arabidopsis thaliana. Wang et al. (2009) cloned the rice gene NECK LEAF 1 (NL1) which encodes a GATA-type transcription factor and reported smaller panicles with overgrown bracts due to mutation in NL1.

In plant sexual reproduction, karyogamy or nuclear fusion is essential for sexual reproduction. Drews and Yadegari (2002) suggested segregation distortion and reduced seed set due to mutation affecting karyogamy. In Arabidopsis, Portereiko et al. (2006) identified nine female gametophyte mutants viz. nuclear fusion defective1 (nfd1) to nfd9, that are defective in fusion of the polar nuclei. In the mutants nfd1 to nfd6, failure of fusion of the polar nuclei is the only defect detected during mega-gametogenesis.

In A. thaliana, Takeuchi and Higashiyama (2016) have reported function of AtPRK3, along with AtPRK1, AtPRK6, and AtPRK8, as the receptors that sense the AtLURE1.2 peptide and discovered vital role of AtPRK3 in pollen tube growth, plant fertility, and plant reproduction. Chakraborty et al. (2018) have reported structural similarity of extracellular domain of pollen receptor kinase 3 to the SERK (somatic embryogenesis receptor kinase) family of co-receptors. Recently, Lee and Goring (2020) reported reduced seed set due to synergistic effect of mutation in SERKs and Leucine rich repeat receptor kinases (LRR-VIII-2 RKs) in A. thaliana.

Candidate genes for SY

This study identified “vacuolar protein sorting-associated protein 45 homolog” in the candidate region of SNP “A02.19604813” associated with SY. Bassham and Raikhel (1998) have demonstrated the role of AtVPS45p homolog in protein transport to the vacuole in Arabidopsis thaliana. Zouhar et al. (2009) suggested the role of AtVPS45 in pollen growth and cell expansion. Also, WAT1 and SPL6 (squamosa promoter-binding-like protein 6) were found to be associated with SY. Role of TaSPL3 and TaSPL6 was reported in regulating flowering time and promoting biomass accumulation in Arabidopsis (Cao et al. 2019). Role of similar genes in governing seed weight (WAT1) and siliqua length (SPL2, SPL9) was already discussed above.

Another gene (BG1) encoding protein “BIG GRAIN 1-like E” was identified on chromosome C02 in the candidate region of SNP “C02.22109942” associated with SY. In both Arabidopsis and rice, BG1 improved plant biomass, seed weight, and yield through its involvement in auxin transport and auxin response (Liu et al. 2015b). Gene AGL82 encoding agamous-like MADS-box proteins have a role in maintaining the proper function of the central cell in the pollen tube attraction (Li et al. 2015). Another gene (LHT1) encoding protein “lysine histidine transporter 1” was identified on chromosome C02. Hirner et al. (2006) showed that mutation of a single transporter gene (LHT1) is sufficient to inhibit plant growth and interfere with both the uptake and distribution of amino acids in Arabidopsis. LHT1 predominantly expresses in nonvascular tissues, including root surface and leaf mesophyll.

Conclusion

This study identified SNPs having significant association with key yield component traits viz. SL, NSS, TSW and SY in Brassica napus. Candidate gene analysis in the genomic region around SNPs (within 150 kb LD region) identified important candidate genes having significant role in phytohormone signaling, seed storage proteins, sexual reproduction and fertilization. Genotypes identified with high SL, NSS, TSW and SY suggested to serve as donors in crop improvement programs. As environments have significant influence on trait performance, multi-environment GWAS is to be preferred to capture consistent SNPs across environments with high allelic effect. The candidate regions will help to develop KASP (kompetitive allele specific polymerase chain reaction) based markers which can be used in marker assisted breeding. It is suggested to carry further research work on validation of SNPs, identified in present study, to develop gene-based markers for their utilization in marker assisted selection (MAS) in crop improvement.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

SL

Siliqua length

NSS

Number of seeds per siliqua

TSW

Thousand seed weight

SY

Seed yield per plant

BLUPs

Best linear unbiased predictors

GWAS

Genome wide association study

SNPs

Single nucleotide polymorphisms

PAU

Punjab Agricultural University

BTI

Bathinda

LDH

Ludhiana

ANOVA

Analysis of variance

CV

Coefficient of variance

CTAB

Cetyl trimethyl ammonium bromide

GBS

Genotyping by sequencing

FarmCPU

Fixed and Random Model Circulating Unification

GAPIT

Genomic Association and Prediction Integrated Tool

PCs

Principal Components

LD

Linkage disequilibrium

ARF

Auxin response factor

AuxREs

Auxin response elements

OSR

ORGAN SIZE RELATED

BR

Brassinosteroid

SSP

Seed storage proteins

ER

Endoplasmic reticulum

PSV

Protein storage vacuoles

SERK

Somatic embryogenesis receptor kinase

LRR-VII-2 RKs

Leucine rich repeat receptor kinases

SPL

Squamosa promoter binding like protein

Declarations

Conflict of interest

The authors declare that there is no conflict of interest.