Introduction

Rapeseed (Brassica napus L.) is the most important oilseed crop grown in temperate agricultural regions. As the second largest output oil crop in the world, it provides approximately 13 % of the world’s vegetable oil (Amar et al. 2008; Hajduch et al. 2006). Growth of rapeseed as a commercial oil crop in Asia, Europe, North America, and Australia has progressively increased owing to its diversified use in food and feed production, and its growing economic importance as a source of renewable energy, mainly as biodiesel (Mittelbach and Gangl 2001). Unfortunately, it is subject to significant losses of seed as a result of premature dehiscence before and during harvest (Morgan et al. 1998). Seed shedding from pods, commonly referred to as pod shatter, is a major risk for rapeseed production (Liu et al. 2013; Price et al. 1996; Wang et al. 2007).

Fully mature rapeseed pods are extremely susceptible to opening, resulting in seed loss. Significant seed losses occur due to natural shedding and crop disturbances by harvesting machinery or the wind. This loss is typically 8–12 % of total seed yield (Kadkol et al. 1984), and can exceed 20 % if harvesting is delayed and occurred in adverse conditions (Price et al. 1996). Shed seed numbers often reach 10,000 per m2 (Lutman 1993), equivalent to 200–500 g ha−1. Volunteer rapeseed is a common weed, which can emerge in crops following rapeseed. Volunteer seedlings occurs in 11 % of fields in western Canada (Devine and Buth 2001) and up to 25 % of winter wheat crops fields (Orson 1994). Volunteer rapeseed plants may also cause genetic contamination through cross-pollination in following crops that are different cultivars of rapeseed (Morgan et al. 1998). This problem is exacerbated, as rapeseed seeds can remain viable in soil for several years. Therefore, rapeseed is often harvested ahead of maturity to avoid seed loss. However, this results in chlorophyll-contaminated oil extracted from immature seed, and lowering its quality. Genetic improvement of increased pod shatter resistance will promote natural maturing of uniformly ripe seeds with improved oil extraction characteristics; this will help to reduce costs and inconveniences associated with windrowing the crop.

There is little variation in shatter resistance occurring in current commercial breeding lines (Bowman 1984), as assessments of pod shatter susceptibility between lines relied mainly upon observation of the crop in the field, or upon hand tests of pods (Child et al. 2003). However, a testing procedure has been devised by Morgan et al. (1998), and improved by Hu et al. (2012), that exposes pods to random impacts in a similar manner to that occurring in the crop canopy during harvest. This random impact test enables rapid comparison of pod shatter resistance in the fully mature pods of rapeseed plants.

Fully mature rapeseed pods consist of two valves separated by a central replum (Chandler et al. 2005), with the sutures extending on each side and along the whole length of the pod. Early studies of pod structure have aimed at establishing the anatomical basis of shatter (Meakin and Roberts 1990a, b). These investigations revealed that dehiscence zones are important for pod shattering. Dehiscence zones develop at the margins of the valves, adjacent to the replum, and run the length of the silique. The cells of the separation layer eventually begin to degrade, weakening the contact between the valves and the replum. Loss of cellular cohesion is confined to the cells of the dehiscence zone and results from middle lamella breakdown (Meakin and Roberts 1990a, b). Because of its importance, numerous studies have focused on the structure and development of the dehiscence zone and valve margins (Child and Huttly 1999; Child et al. 2003; Meakin and Roberts 1990a; Picart and Morgan 1984), but little research has paid attention to the structure and development of the pod replum, which is also important to pod shattering in rapeseed. This work shows for the first time, how replum structures and dimensions differ in rapeseed lines, and how these differences contribute quantitatively towards pod shatter resistance.

Materials and methods

Plant materials

The 64 accessions (B. napus, listed in Table S1) used in this study were planted at the Institute of Oil Crops Research, Chinese Academy of Agriculture Science in 2009 and 2010 respectively. The field planting followed a randomized complete block design with three replications. Each plot was 3.0 m2 with 30 plants. The seeds were hand sown and the field management followed standard agricultural practice. An F2 population derived from a cross between zy72360 (resistant to pod shattering) and R1 (very susceptible to pod shattering) lines, previously reported by Hu et al. (2012), was also used in this study.

Random impact test

Random impact testing was conducted in accordance with our previously reported method (Hu et al. 2012). Each time 20 intact mature pods were placed in a closed polythene drum of 10 cm diameter, 14 cm height with 50 steel balls of 8 mm diameter. The drum was then shaken mechanically in a rocking bed (HQ45Z, Wuhan Scientific Instrument Factory, China) at 300 rpm for 1 min. Ten replicates with total of 200 pods were conducted. The drum was opened after each replicate and the number of broken and damaged pods was counted. The SSRI was counted according to the expressions described as Hu et al. (2012).

Light microscopy

Flowers were labeled at anthesis, and approximately 5 weeks later siliques that were fully developed, but still green, were collected for microscopy. The basal parts of siliques (from 7 to 10 mm near to pedicel) were fixed and embedded in paraffin wax according to the methods described by Hu et al. (2013) with minor modifications. The material was fixed in 2.5 % glutaraldehyde in a 0.1 M phosphate buffer, pH 6.8, for 24 h. After dehydration using the alcohol series, the material was cleared twice with xylene for 2 h. The material was infiltrated and subsequently embedded in the paraffin wax. Transverse sections of 10 μm thickness were obtained using a Leica RM 2016 microtome. Sections were stained for 5 min in 0.01 % (w/v) aqueous toluidine blue. For lignin analysis, sections were treated for 2 min with 2 % phloroglucinol in 95 % ethanol, and photographed in 50 % hydrochloric acid. All images were acquired using a DP 71 CCD camera (Olympus, Japan).

Morphology observation of intact mature siliques was conducted under a zoom stereo microscope (SZX16, Olympus, Japan), and images acquired using a DP 71 CCD camera.

The replum-valve joint area index was measured using Image Pro Express software (Media Cybernetics Inc., Silver Spring, MD, USA).

Data analysis

Phenotypic trait data and correlation coefficients between traits were analyzed using Statistical Analysis System (SAS) V8 (SAS Institute Inc., Cary, NC, USA).

Results

Screening of resistance to pod shattering

To analyze pod shatter resistance, the silique shattering resistance index (SSRI) of 64 rapeseed lines was investigated through random impact testing method. Two independent experiments were performed with similar results in 2009 and 2010 respectively. The average values of silique traits in 64 rapeseed lines were listed in Tables S1 and S2. The measured silique length didn’t include its peak. The middle of silique was used to measure the silique width, silique wall thickness and replum thickness. According to the investigation in 2009, the accessions displayed normal distribution with a wide variation in SSRI, ranging from 0.05 to 0.94 with a variance coefficient of 41.8 %. As shown in Fig. S1a, 3.13 % of 1ines were very susceptible germplasm to pod shattering (SSRI 0.0–0.1), and 14.06 % were susceptible germplasm (SSRI 0.1–0.3); together, these accounted for 17.19 % of the accessions. Most 1ines (59.38 %, SSRI 0.3–0.7) had medium susceptibility. 21.88 % (SSRI 0.7–0.9) of 1ines showed relative resistance to pod shattering, and one line (zy72360; SSRI 0.94) showed resistance to pod shattering.

Table 1 gives the variance of several silique traits and the correlation coefficients between SSRI and silique traits in the 64 accessions. There was a highly significant positive correlation (P < 0.0001) between SSRI and silique length (the correlation coefficient is 0.6693). However, there were no significant correlations between SSRI and silique width, silique wall thickness, or replum thickness. The repeated investigation in 2010 showed similar results (Tables 2, S2) and the correlation coefficient was 0.6402 between SSRI and silique length.

Table 1 Variance of five silique traits, and correlation coefficients between silique shattering resistance index and five silique traits
Full size table
Table 2 Phenotypic correlations among silique shattering resistance index (SSRI), replum-valve joint area index (RJAI), and silique length in 64 rapeseed lines
Full size table

Morphology of rapeseed pods

To compare and analysis the silique morphology of different resistances to pod shattering in rapeseed, three high SSRI (zy72360, ‘Zhongshuang No.11’, and P7053, SSRI 0.94, 0.89, and 0.79, respectively) and three low SSRI (R1, 9WD321, and 9WD329, SSRI 0.05, 0.06, and 0.10, respectively) lines were chosen from the 64 accessions. Transverse sections of the basal parts of siliques (from 7 to 10 mm near to pedicel) were stained with toluidine blue which stains lignified cell walls intense blue, and with phloroglucinol which stains lignified cell walls dark red, to examine fruit internal structures. All the same cell types can be identified in siliques from these lines. The lignified cells of the dehiscence zone from the vascular bundles and the endocarp inner cell layer were visualized by the staining (Fig. 1). During fruit ripening, the cells of the dehiscence zone became lignified, thus allowing the two valves to separate from the replum after maturation of the fruit. However, the sizes of the inner parts of replum are obviously bigger in the three high SSRI lines than that in the three low SSRI lines (Fig. 1).

Fig. 1
figure 1

Transverse sections of siliques from six different resistance germplasms to pod shattering in rapeseed. Toluidine blue staining of three relative resistant (a zy72360, b ‘Zhongshuang No.11’, c P7053) and three susceptible germplasms (g R1, h 9WD321, i 9WD329). Phloroglucinol staining of the three relative resistant (d–f, the same as a–c) and three susceptible germplasms (j–l, the same as g–i). V valve, R replum, dz dehiscence zone, en endocarp, me mesocarp, ep epicarp, Scale bars 100 μm

Full size image

The morphology of the intact mature siliques was anatomized, revealing obvious differences in the replum structure between the high and low SSRI lines. The replum close to the pedicel was thicker in the high SSRI lines than that in the low SSRI lines (Fig. 2a). However, no distinct differences were observed in the replum near the beak (Fig. 2b).

Fig. 2
figure 2

Structural feature of pod replum in rapeseed. a Replum morphology near to the pedicel of (left) three high SSRI (zy72360, ‘Zhongshuang No.11’, and P7053) and (right) three low SSRI (R1, 9WD321, and 9WD329) lines. b Replum morphology near to the beak of (left) the same three high SSRI and (right) three low SSRI lines. c A replum model defining the replum-valve joint area index = L1 × L2. d Replum morphology near to the pedicel of five high SSRI F2 plants from zy72360 and R1. e Replum morphology near to the pedicel of five low SSRI F2 plants from zy72360 and R1. Scale bars 2 mm

Full size image

Replum-valve joint area index

In a typical Brassica fruit, two carpels enclose two locules and these are separated by a central replum (Chandler et al. 2005). The two carpels stick together throughout the replum, and the replum size reflects the joint area. To study the relationship between replum size and resistance to pod shattering, a replum-valve joint area index (RJAI) was defined as a product of the vertical border (L1) and the level border (L2) in the replum near the pedicel (Fig. 2c).

The RJAI displayed normal distribution with a wide variation (Fig. S1b), ranging from 0.50 to 4.12 mm2, with a variance coefficient of 45.4 %. Twenty accessions (31.3 %), with a RJAI range of 1.0–1.2 mm2, formed a peak, while 76.6 % of accessions were in the 1.0–1.6 mm2 range. Two accessions displayed an RJAI of less than 0.6 mm2, while two accessions were over 3.0 mm2. The RJAI had a highly significant positive correlation with SSRI (P < 0.0001), with the correlation coefficient of 0.6140 and 0.6406 in 2009 and 2010 respectively (Table 2). Concurrently, silique length was significantly positively correlated (P < 0.0001) with RJAI, with correlation coefficients of 0.5603 and 0.5509 (Table 2). These results revealed that a thick replum structure was one of the main reasons for high resistance to pod shattering.

To further validate correlations amongst SSRI, RJAI, and silique length, an F2 population of 276 individuals derived from the cross of zy72360 and R1 previously reported by Hu et al. (2012) were investigated. The F2 population displayed a wide variation in RJAI (Fig. S1c), with a variance coefficient of 40.1 %. There was significant positive correlation (P < 0.0001) between SSRI and RJAI (correlation coefficient 0.5953; Table 3). Concurrently, silique length was significantly positively correlated (P < 0.0001) with SSRI and RJAI, with correlation coefficients of 0.7088 and 0.6223, respectively. The replum near to the pedicel in five high SSRI F2 plants (Fig. 2d) was thicker than that of five low SSRI F2 plants (Fig. 2e). These results were consistent with anterior results, and further validated that the pod replum structure was highly associated with pod shattering.

Table 3 Phenotypic correlations among silique shattering resistance index (SSRI), replum-valve joint area index (RJAI), and silique length in an F2 rapeseed population derived from a cross of zy72360 and R1
Full size table

Discussion

The replum is the site of attachment of the ovules to the placenta, and later develops into a septum or partition separating the seeds within the mature fruit. In past studies, the structure and development of the replum has often been ignored or not reported in B. napus. In Arabidopsis, a key regulator of replum formation was discovered by Roeder et al. (2003). Their work suggest that one role for the outer replum may be to prevent the valve margin lignified layers from fusing together and inhibiting dehiscence. However, one of the obvious differences between Arabidopsis fruits and those of Brassica species is the replum development. Wild-type Arabidopsis produces an outer replum composed of about six cell files at the medial region of the fruits, whereas Brassica fruits lack this tissue (Girin et al. 2010). In the report of Girin et al. (2010), the braA.ind.a-1 mutant and BolC.IND.a RNAi line can form the outer replum tissue and reduce dehiscence in Brassica fruits.

In this study, our results revealed that replum structure displayed wide variation mainly in the inner parts of the replum, and was highly associated with pod shatter resistance. In the high SSRI lines, the replum near to the pedicel was obviously thicker than that observed in low SSRI lines (Fig. 2a). This phenomenon allows a better understanding of pod shatter resistance in rapeseed. The two valves of a mature pod stick together throughout the replum; as the dehiscence zones cells act as glue, the replum size reflected the joint area. Thick replum structures provide larger joint area to the valves, leading to greater pod shatter resistance. To validate the status of the replum in pod shatter resistance, RJAI was designed and measured. In both the 64 rapeseed lines, and in the F2 population of 276 individuals derived from a cross of zy72360 and R1, there were highly significant positive correlations (P < 0.0001) between SSRI and RJAI, with the correlation coefficient reaching approximately 0.6. On the other hand, there are some reports of Arabidopsis gain of function mutants with increase replum sizes and reduced dehiscence, which support the findings of this study. For example, Chung et al. (2013) reported that ntt-3D mutants showed enlarged replum and fruit indehiscence.

In recent years, several transgenic approaches for producing pod shatter resistance rapeseed using Arabidopsis genes have been conducted. Chandler et al. (2005) transformed B. napus using MADSB gene and Østergaard et al. (2006) transformed B. juncea using the Arabidopsis FRUITFULL gene to produce shatter resistance plants. The transformed plants possessed altered dehiscence zone anatomies, wherein valve margin cells were not lignified or lacked valve margin specification, thus leading to a failure in fruit opening. However, the transgenic fruit produced was closed too tightly for a combine harvester to thrash. This makes it necessary to screen pod shatter resistance rapeseed from natural breeding materials. Many evaluation methods for shatter resistance, including the random impact test (Hu et al. 2012), the cantilever test (Kadkol et al. 1984), and the pendulum test (Liu et al. 1994) are performed after fully mature pods are harvested. It is inconvenient to screen pod shatter resistance material during rapeseed breeding. This study reveals that the replum-valve joint area offers a good method to screen breeding material for pod shatter resistance before maturity, which will be much more convenient for breeders. The replum of rapeseed usually developed to its normal size 4–6 weeks after flowering. Therefore, breeders will be able to rapidly screen breeding materials for pod shatter resistance according to replum size before shattering occurs.

Morgan et al. (1998) analyzed statistically significant correlations among rapeseed characteristics and found no correlation between silique length and pod shatter resistance. However, Summers et al. (2003) found that in rapeseed Apex and DK142 populations increased valve length was significantly correlated with increased resistance. However, the shatter susceptible Apex pods were longer than shatter resistant DK142 pods, indicating that the length of valve walls could not account for the differences in shattering between Apex and DK142 (Child et al. 2003). Wang et al. (2007) reported that most plant and pod morphological traits were not correlated with pod shatter resistance, with the exception being pod length. However, their studies revealed that shorter pod lengths had higher pod shatter resistance. Their results were mostly dependent on the B. juncea and Sinapis alba germplasms. Recently, Liu et al. (2013) reported that SSRI had a significant positive correlation with silique length, and the correlation coefficient was 0.69–0.76. Our results are consistent with those of Liu et al. (2013), with correlation coefficients of 0.6693/0.6402 in the 64 rapeseed lines, and 0.7088 in the F2 population. Moreover, silique length had a significant positive correlation with RJAI, which implies that silique length may influence the development of replum. The relationship between silique length and replum development requires further study. Due to both RJAI and silique length are highly significant positive correlated with SSRI, we suggest to combine the replum structure and silique length to screening high resistance materials in rapeseed breeding.