Introduction

Phosphorus (P) is an essential mineral nutrient for plants. It plays important roles in plant growth and metabolism, including nucleic acid synthesis, photosynthesis, glycolysis, respiration, redox reactions, and carbohydrate metabolism (Marschner 1995). The total amount of P in soils is sufficiently high to support plant growth. However, it is largely unavailable for uptake by plants because it readily forms complexes with some metal cations, in particular, aluminum and iron in acid soils and calcium and magnesium in alkaline soils, which results in P deficiency in soils (Holford 1997). P deficiency is a major limitation to plant growth worldwide (Lynch 2007; Vance et al. 2003). It has been estimated that 5.7 billion hectares of land worldwide are deficient in P, that is, the concentration of available P does not exceed 10 μM (Batjes 1997). In China, most cultivated soils are either acidic or calcareous, which gives them a strong capability to fix P; hence, the majority of soils in China are P deficient (Yan et al. 2006). The application of P fertilizers can alleviate P deficiency, but it is not an entirely satisfactory solution, due to limited bioavailability, potential environmental threats such as eutrophication, and the exhaustion of resources of phosphate rock (Vance et al. 2003). Therefore, it has become attractive to use an alternative approach that exploits germplasms from P-efficient crops or involves the breeding of P-efficient cultivars (Fageria et al. 2008).

Plants with an extensive root system can absorb more nutrients under conditions of nutrient stress than can plants with a less extensive system. To survive during low P (LP) stress, plants modify their root system to increase the acquisition of P by increasing the root/shoot ratio, altering root morphology and architecture, stimulating lateral root initiation, and increasing the density and elongation of root hairs (Hammond and White 2008; Hermans et al. 2006; Jain et al. 2007; Lynch 2007). However, cultivars and species show wide genetic variation with respect to changes in root morphology in response to LP stress. Thus, analysis of quantitative trait loci (QTL) has been used to elucidate the genetic mechanisms that control modification of the root system in response to P deficiency. A QTL (qREP-6) for root elongation under condition of P deficiency has been detected in rice that improved the adaptability of rice to LP stress (Shimizu et al. 2004, 2008; Ismail et al. 2007). Main-effect QTL that affect root hairs and lateral roots were identified in maize under P deficiency (Zhu et al. 2005a, b). In common bean, QTL for root traits, acid exudation and P uptake were shown to overlap in four linkage groups (Beebe et al. 2006; Yan et al. 2004). In Arabidopsis, a QTL (Atlg23010) for primary root elongation in response to LP stress has been cloned (Reymond et al. 2006; Svistoonoff et al. 2007). These studies have provided a solid basis for understanding the genetic control of tolerance to P deficiency through specific root traits.

QTL for the same or related traits that are detected in different experiments with the same population are usually located in the same regions. To identify the candidate gene, it is essential to determine whether or not these QTL represent a single locus. Appropriate statistical tools, such as QTL meta-analysis, can evaluate information about a QTL from multiple environments and refine the position of the QTL, with a reduction in the confidence interval (Arcade et al. 2004). This method has been applied successfully to the analysis of QTL for flowering time in maize and seed yield in rapeseed. Meta-analysis of these QTL revealed the genetic architecture of both flowering time and yield, and candidate genes were proposed (Chardon et al. 2004; Shi et al. 2009). However, this approach has not been used to analyse QTL for root system traits.

Brassica napus is one of main oil crops in the world and also a potential bioenergy crop. Because of a high P requirement for optimal seed yield and quality, the low concentration of available P in soils seriously limits the production of B. napus. Genotypic variation with respect to P efficiency has been investigated among Brassica species (Akhtar et al. 2008; Hammond et al. 2009; Solaiman et al. 2007; Zhang et al. 2009; Zhao et al. 2008). Compared with P-inefficient cultivars, P-efficient cultivars absorb more P and produce more biomass under LP stress due to alterations in root morphology and secretion of organic acid (Solaiman et al. 2007; Zhang et al. 2009). These results indicate that rapeseed germplasms show significant genotypic differences with respect to P uptake and utilization, and that it is possible to breed P-efficient cultivars. However, the genetic control of tolerance to P deficiency in rapeseed remains poorly understood.

The critical period for P nutrition in plants is the early growing stage. Thus, an understanding of the genetic control of P efficiency at the seedling stage should help to reveal the mechanisms of P efficiency. The aims of this study were to elucidate the genetic basis of P efficiency in B. napus and to provide a foundation for breeding P-efficient rapeseed cultivars by marker-assisted selection (MAS). For this purpose, we identified QTL for plant total dry weight (DW), total root length (RL), root surface area (RSA), root volume (RVol), and plant P uptake (PU), under conditions of LP and high P (HP) at the seedling stage of a recombinant inbred line (RIL) population of B. napus in three independent paper culture experiments. In addition, we analyzed the genetic basis of root system modification in response to P deficiency by QTL meta-analysis, and developed functional markers from genes in the P metabolic pathway in Arabidopsis.

Materials and methods

Plant materials

An F10 RIL population that consisted of 124 lines was developed from a cross between the P-efficient rapeseed cv. ‘Eyou Changjia’ and the P-inefficient cv. ‘B104-2’ using a single seed descent method. The parents, ‘Eyou Changjia’ and ‘B104-2’, were selected on the basis of their P efficiency coefficients (the ratio of shoot dry weight or seed yield under LP to that under HP) (Duan et al. 2009). In our previous studies, it was found that ‘Eyou Changjia’ had a larger root system and acquired more P than ‘B104-2’ under LP condition in pot culture, root-soil compartments, and solution culture.

Paper culture system

The paper culture experiments were conducted to investigate the root morphology of the RILs. Uniform seeds were selected from the 124 lines and the two parents, surface-sterilized for 15 min in a 0.5% (w/v) sodium hypochlorite solution, and then washed three times with deionized water. The surface-sterilized seeds were germinated on moistened gauze that was fixed to a porcelain tray filled with deionized water. Then they were grown in an illuminated culture room for 6 days at 22–24°C until cotyledons were fully developed, but lateral roots were not differentiated from primary roots. Finally, the seedlings were transferred carefully to the paper culture system.

The paper culture system was designed according to the system described by Liao et al. (2001). It consisted of a sheet of 15.6 × 22.0-cm P-free blue germination paper (Anchor Paper Company, St Paul, MN, USA) placed inside a polyethylene bag that was punctured evenly at 1.5-cm intervals for aeration. Two seedlings were transplanted onto the top of the paper. The sheets of paper with the seedlings were suspended upright in nutrient solution pool of dimensions 200 × 40 × 22 cm (length × width × height), and 50 l of quarter-strength modified Hoagland’s solution were added to submerge the bottom of the sheets (Hoagland and Arnon 1950). Two P treatments, LP (5 μM) and HP (1 mM), were designed with three replicates. The experiment was conducted in a randomized complete block design.

The quarter-strength nutrient solution was replaced with half-strength solution after 5 days, which was replaced in turn with full-strength solution after a further 5 days. The full-strength modified Hoagland’s solution consisted of 4 mM KNO3, 1 mM KH2PO4, 2 mM MgSO4·7H2O, 4.5 mM Ca(NO3)2·4H2O, 46 μM H3BO3, 0.32 μM CuSO4·5H2O, 0.77 μM ZnSO4·7H2O, 9.14 μM MnCl2·4H2O, 0.37 μM NaMoO4·2H2O, and 50 μM EDTA-Fe. In the LP nutrient solution, 1 mM KH2PO4 was replaced by 0.5 mM K2SO4. The pH value of the nutrient solution was adjusted to 5.7 ± 0.2 with 2 M NaOH or HCl every 2 days. The three independent paper culture experiments were all conducted in the same greenhouse at Huazhong Agricultural University, Wuhan, China. The first experiment (Exp. 1) was carried out from February 1 to March 6, 2008, the second (Exp. 2) from March 4 to 31, 2008, and the third (Exp. 3) from November 13 to December 15, 2008. During the three experiments, the minimum temperature in the greenhouse ranged from 5 to 8°C, 10 to 14°C, and 8 to 12°C, respectively, and the maximum temperature ranged from 18 to 21°C, 25 to 28°C, and 22 to 26°C, respectively.

Measurement of phenotypic traits

The intact root systems grown on the germination paper were scanned as a digital image with a scanner after harvesting. The RL, RSA and RVol were determined from the root images using the root image analysis software WinRHIZO Pro (Regent Instruments, QC, Canada).

After they had been scanned, the plants were detached from the paper sheets. They were then oven-dried at 105°C for 30 min and then at 60°C to constant weight to determine the plant DW. The dried plants were ground to fine powder and digested by adding 5 ml of a mixture of concentrated H2SO4:HClO4 (19:1, v:v). P concentration in plant dry matter in the digested solution was analyzed by P vanadate–molybdate yellow colorimetric method using a flow injection analysis instrument (FIAstar 5000 analyzer; FOSS, Hilleroed, Denmark). Plant PU was calculated as plant DW × P concentration.

Analysis of molecular markers and construction of a linkage map

Genomic DNA was extracted from plant leaf tissues by the Cetyltrimethylammonium Bromide (CTAB) method (Doyle and Doyle 1990). The genotypes of the RILs were analyzed using simple sequence repeat (SSR), amplified fragment length polymorphism (AFLP), sequence related amplified polymorphism (SRAP), and functional markers (FM) to enable the construction of a genetic linkage map.

Primer sequences for SSR markers were obtained from various sources as follows: UK (http://www.brassica.bbsrc.ac.uk/BrassicaDB, prefixed by OL and Na), Australia (http://www.hornbill.cspp.latrobe.edu.au, prefixed by sA), Canada (http://www.brassica.agr.gc.ca/index_e.shtml, prefixed by sR and sN), Japan (Suwabe et al. 2002, prefixed by BRMS), and France (Piquemal et al. 2005, prefixed by BRAS, CB and MR). The remainder of SSR primer pairs were obtained from private communications (prefixed by CNU, niab and HBr). The protocol for SSR analysis followed the method described by Lowe et al. (2004).

The AFLP analysis was performed according to the protocol developed by Vos et al. (1995). Polymorphisms were detected using genomic DNA digested with the restriction enzymes EcoRI and MseI. One hundred primer combinations, which involved 10 EcoRI and 10 MseI primers, were used. The AFLP markers were named using codes for each EcoRI and MseI primer, for example e2m4.

The analysis of SRAP markers followed the protocol of Li and Quiros (2001), and the polymorphic primer pairs were named by combining the names of forward and reverse primers (e.g., em5me3).

FM were obtained from the functional genes that are related to P transporters and from regulatory factors that are induced in response to P starvation in Arabidopsis (http://www.arabidopsis.org/). The primers were designed from sequences that are conserved between the homologous genes from Arabidopsis and Brassica. The FM were named using the prefix Bn and the gene name in Arabidopsis, for example BnIPS2.

All PCR products were separated by PAGE and stained with AgNO3. When a primer pair detected more than one polymorphic locus, the different loci were distinguished by small letters after the name of the marker. For example, the primer pair BRMS056 generated three polymorphic loci, which were named BRMS056a, BRMS056b and BRMS056c, respectively.

The genetic linkage map was constructed using the software program Joinmap4.0 (Van Ooijen 2006), which is based on the Kosambi mapping function. The mapping procedure followed the method of Qiu et al. (2006). Firstly, a framework linkage map was constructed using 269 SSR markers. The threshold for goodness of fit was set to ≤5.0 with logarithm of the odds ratio (LOD) scores >1.0 and a recombination frequency <0.4. Secondly, using the order of the SSR markers in the framework map as a basis, the AFLP, SRAP, and FM markers were joined to construct a map with a recombination level equal to 0.45 and minimum LOD scores equal to 0. Markers with a χ2 value > 3.0 were removed in every genetic group.

Statistical analysis and QTL detection

Statistical analysis of the traits was conducted using the ANOVA procedure of SAS 8.1 (SAS Institute, Cary, NC, USA). Pearson’s correlation analysis was performed using the CORR procedure of SAS.

QTL were detected by the composite interval method (CIM) using WinQTL Cartographer 2.5 software (Wang et al. 2006; Zeng 1994). CIM was used to scan the genetic map and estimate the likelihood of a QTL and its corresponding effect at every 2 cM. For each trait, the threshold for the detection of a significant QTL (P < 0.05) was estimated by 1,000 permutations (Churchill and Doerge 1994).

The two-round strategy of QTL meta-analysis was performed according to the method described by Shi et al. (2009). The coincidental significant QTL for the same traits from the three experiments were first integrated into consensus QTL using the BioMercator2.1 software (Arcade et al. 2004). The consensus QTL for different traits that overlapped were integrated into unique QTL in the second round of QTL meta-analysis.

For the consensus QTL, the QTL nomenclature corresponded to “q” (abbreviation of QTL), followed by “HP” or “LP” (to indicate whether the QTL was identified under HP or LP), then an abbreviation of the trait (e.g., RL), the linkage group number (A1–A10 and C1–C9), and finally the serial letter (a, b, c…), for example qHP-RL-A3a and qLP-RVol-C3b. For the unique QTL, the nomenclature corresponded to “uq.”, followed by the linkage group number, and then a small serial letter, for example uq.C3b.

Results

Genetic linkage map

Four types of molecular marker, SSR, AFLP, SRAP and FM, were used to construct a genetic linkage map. Firstly, SSR markers were used to construct a framework map that was based on the anchored markers reported by Long et al. (2007) and Piquemal et al. (2005). The framework map contained 176 SSR markers (Supplementary Table S1a). These SSR markers were assigned to 23 linkage groups, which corresponded to 19 chromosomes, named as A1–A10 and C1–C9. Chromosomes A4, A5, A9 and C7 were divided into two smaller linkage groups, which were named A4a and A4b, A5a and A5b, A9a and A9b, and C7a and C7b, respectively.

Secondly, a new linkage map was constructed by adding the AFLP, SRAP and FM markers to the framework map. This linkage map consisted of 553 markers: 202 SSR, 62 AFLP and 234 SRAP markers and 55 FM (Supplementary Table S1b). The total length of the 19 linkage groups was 1592.7 cM, with an average distance of 2.9 cM between adjacent markers. In general, the order of the SSR markers in the map agreed with that reported previously (Long et al. 2007; Piquemal et al. 2005; Qiu et al. 2006). This map was employed in subsequent QTL analysis.

Phenotypic variation and correlation analysis among traits

Plant DW, root traits and PU were investigated in the two parents under conditions of HP and LP. Under HP, the two parents showed no significant difference in plant DW or root traits, including RL, RSA, and RVol. However, they consistently showed significant differences under LP condition. The P-efficient parent ‘Eyou Changjia’ yielded a higher plant DW and developed a larger root system (longer RL, and larger RSA and RVol) than ‘B104-2’ (Tables 1, 2). Moreover, ‘Eyou Changjia’ acquired more P than ‘B104-2’ under LP condition in all three experiments (Table 1). ‘Eyou Changjia’ accumulated 41.3% more P than ‘B104-2’ in Exp. 1, 37.6% more in Exp. 2, and 38.4% more in Exp. 3. ANOVA revealed that significant effects of environment, P level, genotype, and interactions of E × P and G × P were detected for most traits (Table 2).

Table 1 Mean values and ranges of the five traits in the parents and the RIL population
Full size table
Table 2 Significance of three-way ANOVA analysis for the five traits among the parents and RILs in HP and LP treatments in three experiments
Full size table

In all three experiments, all five traits showed a continuous distribution in the RIL population (Table 1; Fig. 1). The coefficients of variation (CVs) for these traits ranged from 20.5 to 40.6%. Significant transgressive segregation was observed for each trait under both P conditions. The transgressive lines showed more extreme phenotypes than ‘B104-2’ or ‘Eyou Changjia’. For example, in the three individual experiments, 21, 11 and 10 lines, respectively, showed a significantly lower RSA than ‘B104-2’ under LP condition, whereas 13, 23 and 24 lines, respectively, had a significantly higher RSA than ‘Eyou Changjia’ (data not shown). The results of ANOVA showed that environment, P level, genotype and the interactions between these factors had highly significant effects on the four traits (except for PU) in the RIL population (Table 2).

Fig. 1
figure 1

Frequency distribution of plant DW, PU, RL, RSA and RVol in the RIL population under LP condition in the three experiments. The letters ‘B’ and ‘E’ indicate the values for the parents ‘B104-2’ and ‘Eyou Changjia’, respectively, in the three experiments

Full size image

Pearson’s correlation coefficients between traits were calculated (Supplementary Table S2). Significant positive correlations (P < 0.01) were found among the five traits under either HP or LP condition in every experiment. In any two of the three independent experiments, the same traits showed highly significant positive correlations at the same P level.

QTL detection and the analysis of consensus QTL

From the three experiments, a total of 62 significant QTL was identified for the five traits under the two P conditions. These QTL were distributed on 10 linkage groups (Fig. 2; Supplementary Table S3). In the three independent experiments, 12, 28 and 22 QTL were detected, respectively. Of the 62 significant QTL, 26 were detected specifically in one experiment under one P condition, 24 were detected in two experiments, and 12 were identified in all three experiments. The majority of the QTL (72.6%) were clustered on A1, A3, C2 and C3 (Fig. 2). So QTL meta-analysis was performed to integrate the QTL clusters. Significant QTL that were identified for the same traits on the four linkage groups were integrated into consensus QTL by first round meta-analysis. As a result, 26 consensus QTL were identified on A1, A3, C2 and C3 (Table 3).

Fig. 2
figure 2

Distribution of total QTL for plant DW, RL, RSA, RVol, PU on ten linkage groups under HP (above abscissa) and LP (below abscissa) conditions in the three experiments. The three experiments are differentiated by the two long dotted lines

Full size image
Table 3 Overview of consensus QTL and unique QTL obtained from QTL meta-analysis on linkage groups A1, A3, C2, and C3 of B. napus
Full size table

Sixteen consensus QTL were detected under LP condition. Three consensus QTL for plant DW were detected in the three experiments. qLP-DW-A3a was detected only in Exp. 2 and accounted for 13.7% of the phenotypic variation. qLP-DW-C3a was identified in Exps. 2 and 3, and accounted for 13.8 and 19.7% of the phenotypic variation, respectively. qLP-DW-C3b was identified in all three experiments, and accounted for 13.3% in Exp. 1, 9.9% in Exp. 2 and 10.1% in Exp. 3 of the phenotypic variation, respectively. Two consensus QTL for RL were identified in Exp. 3. qLP-RL-C2a explained 12.4% of the phenotypic variation with a positive additive effect, whereas qLP-RL-C3a contributed 11.3% of the variation with a negative effect. A total of four consensus QTL for RSA were identified. Three of them, qLP-RSA-A1a, qLP-RSA-A3a and qLP-RSA-C3a, were found in Exps. 2 and 3, each explaining 12.9, 16.1 and 10.1% of the phenotypic variation in Exp. 2, and 10.7, 12.6 and 20.3% in Exp. 3, respectively. An additional QTL, qLP-RSA-C3b, was identified only in Exp. 3 and contributed 9.4% of the phenotypic variation. Four consensus QTL for RVol were identified on A1, A3 and C3. qLP-RVol-A1a and qLP-RVol-C3a were identified in Exps. 2 and 3, each explaining 8.0 and 9.6% of the phenotypic variation in Exp. 2, and 10.9% and 14.0% in Exp. 3, respectively. qLP-RVol-A3a was identified only in Exp. 2 and contributed 12.2% of the phenotypic variation. qLP-RVol-C3b was detected in all three experiments, and explained 17.0% in Exp. 1, 8.1% in Exp. 2 and 12.5% in Exp. 3 of the phenotypic variation, respectively. Three consensus QTL for PU were detected in Exp. 2 and Exp. 3. qLP-PU-A3a was detected in Exp. 2 and accounted for 9.5% of the phenotypic variation. qLP-PU-C3a explained 14.6% of the phenotypic variation in Exp. 2 and 18.9% in Exp. 3. qLP-PU-C3b was detected in Exp. 3 and accounted for 11.3% of the variation.

Ten consensus QTL were detected under HP condition. Two consensus QTL for plant DW were identified on A3 and C2, respectively. qHP-DW-A3a explained 11.0% of the phenotypic variation in Exp. 2 and 10.6% in Exp. 3. qHP-DW-C2a accounted for 12.7% of the phenotypic variation in Exp. 1 and 10.5% in Exp. 3. Two consensus QTL for RL, qHP-RL-A3a and qHP-RL-C2a, were detected in Exp. 2 and Exp. 3, respectively. Two consensus QTL for RSA, qHP-RSA-A3a and qHP-RSA-A3b, were detected in Exps. 2 and 3. The consensus QTL for RVol, qHP-RVol-A3a, was identified in all three experiments. Three consensus QTL were identified for PU. qHP-PU-A3a and qHP-PU-A3b were detected only in Exp. 2, whereas qHP-PU-C2a was detected in all three experiments.

Analysis of unique QTL identified under two P conditions

Most of the consensus QTL for each trait overlapped with those for other traits. Under LP condition, the consensus QTL for the different traits were clustered on A1, A3 and C3. For example, consensus QTL were detected for all five traits in the 6.9–15.4 cM region of C3. Similarly, the consensus QTL identified under HP condition clustered on A3 and C2 (Table 3; Fig. 3). Therefore, the consensus QTL were subjected to a second round of QTL meta-analysis and were integrated into eight unique QTL (Fig. 3). Among these eight QTL, four and two unique QTL were specific for LP and HP conditions, respectively. The remaining two unique QTL were detected under both P conditions. Thus, the eight unique QTL could be classified into three categories, namely, LP-specific, HP-specific, and constitutive QTL.

Fig. 3
figure 3

Unique QTL integrated from consensus QTL on linkage groups A1, A3, C2 and C3 of B. napus by the second round of QTL meta-analysis. The symbols ‘┤’ in gray to the left of the linkage groups represent the positions of consensus QTL for different traits, and those in black represent the positions of unique QTL. The length of the vertical bar in the ‘┤’ symbol indicates the confidence interval of the QTL, and the position of the horizontal bar indicates the peak position of the QTL. The names of the QTL are labeled above the ‘┤’ symbol. The black arrows to the right of the linkage groups represent the positions of functional markers (FM) developed from genes in the P metabolic pathway in Arabidopsis

Full size image

The four LP-specific unique QTL were uq.A1, uq.A3a, uq.C3a and uq.C3b. uq.A1 was integrated from two consensus QTL, qLP-RSA-A1a and qLP-RVol-A1a that were identified under LP condition. uq.A3a corresponded to one consensus QTL qLP-PU-A3a, which could therefore be considered as a single unique QTL. uq.C3a and uq.C3b were detected for all five traits, and were located at intervals of 9.3–11.4 cM and 43.5–46.0 cM, respectively. uq.C3a was integrated from five consensus QTL, qLP-DW-C3a, qLP-RL-C3a, qLP-RSA-C3a, qLP-RVol-C3a and qLP-PU-C3a, whereas uq.C3b was integrated from four consensus QTL, qLP-DW-C3b, qLP-RSA-C3b, qLP-RVol-C3b and qLP-PU-C3b.

uq.A3c and uq.A3d were HP-specific unique QTL. uq.A3c was related to qHP-RL-A3a and was considered as a single unique QTL. uq.A3d was integrated from two consensus QTL, qHP-RSA-A3b and qHP-PU-A3b.

The remaining two unique QTL, uq.A3b and uq.C2, were constitutive QTL. uq.A3b was integrated from seven consensus QTL, qLP-DW-A3a, qLP-RSA-A3a and qLP-RVol-A3a, which were obtained under LP condition, and qHP-DW-A3a, qHP-RSA-A3a, qHP-RVol-A3a, qHP-PU-A3a, which were obtained under HP condition. uq.C2 was integrated from one consensus QTL obtained under LP (qLP-RL-C2a) and three obtained under HP (qHP-DW-C2a, qHP-RL-C2a, and qHP-PU-C2a).

Functional markers and unique QTL

As shown in Fig. 3, six functional markers (FM) were distributed in or near the confidence intervals of unique QTL. Three of these markers, BnGPT1-C2, BnIPS2-C3 and BnGPT1-C3, were located in the confidence intervals of the unique QTL, uq.C2, uq.C3a and uq.C3b, respectively. BnGPT1-C2 and BnGPT1-C3, which were developed from AtGPT1, were located in the region of the constitutive QTL uq.C2 and the LP-specific QTL uq.C3b, respectively. BnIPS2-C3, which was developed from AtIPS2, mapped within the interval of the LP-specific QTL uq.C3a. BnPHR1-A1, which was developed from AtPHR1, was located near to the region of the LP-specific QTL uq.A1. BnGPT1-A3, which was developed from AtGPT1, mapped near to the interval of the LP-specific QTL uq.A3a. The final marker, BnMGD2-C2, which was developed from AtMGD2, mapped near to the region of the constitutive QTL uq.C2.

Discussion

Root system traits are very important for the adaptation of plant to P deficiency (Ismail et al. 2007; Jain et al. 2007; Lynch 2007). Given that root traits are influenced readily by the growth environment and are difficult to measure accurately, the selection of root traits as a phenotype for QTL analysis presents certain challenges (De Dorlodot et al. 2007). Many techniques that mimic field conditions have been developed to investigate root morphology and root architecture, for example, the gel chamber (Bengough et al. 2004; Reymond et al. 2006), the root-compartment chamber (Zhang et al. 2009), the stratified sand culture system, and the paper culture growth system (Hammond et al. 2009; Liao et al. 2001; Zhu et al. 2005a, b). In the present study, the root system, which was grown naturally on the surface of P-free blue paper, could be investigated visually and was not destroyed when harvested. This should have improved the accuracy of the data obtained, which was demonstrated by the highly significant correlation (P < 0.001) of the data for the same root traits from the three independent experiments (Supplementary Table S2). The lower correlation coefficients between Exp. 1 and Exp. 2 or Exp. 3 might have been due to the lower environmental temperature in the greenhouse when Exp. 1 was conducted (minimum temperature: 5–8°C in Exp. 1, 10–14°C in Exp. 2 and 8–12°C in Exp. 3). This finding also demonstrates the plasticity of the root traits.

Although plants grown on paper do differ from plants grown in soil, they can reflect the responses of plants in soil to a certain extent. Research on common bean has indicated that the effects of P on basal roots are consistent in paper, sand, and soil culture experiments (Liao et al. 2001). Previously, we showed that similar results were obtained in paper culture, pot culture and root-soil compartment chamber with respect to the root system and PU efficiency of P-inefficient and P-efficient lines from the RIL population, together with those of the two parents, the P-inefficient cultivar ‘B104-2’ and P-efficient cultivar ‘Eyou Changjia’. The seed yield of RILs from 2-year field trials showed a highly significant correlation with RL (R = 0.45–0.56, P < 0.001, n = 124), RSA (R = 0.60–0.72, P < 0.001, n = 124), and RVol (R = 0.52–0.63, P < 0.001, n = 124) when grown on paper under LP condition (unpublished data). Furthermore, three root traits (RL, RSA and RVol) investigated in this study were mathematical interdependent, but were mathematical independent with PU. The following results from two parents, ‘B104-2’ and ‘Eyou Changjia’, and RILs proved that root traits with PU were physiologic correlated.

In general, plants develop three main mechanisms to adapt to LP stress: root interception of P, P acquisition efficiency, and internal P use efficiency (Ismail et al. 2007). Modifications in root architecture that are related to plant PU have been identified as an efficient adaptation to P deficiency (Lynch 2007). Therefore, we used root morphology and PU to assess tolerance to P deficiency in B. napus. The cultivar ‘Eyou Changjia’ absorbed more P and had a higher uptake efficiency than ‘B104-2’ (Table 1), but had a lower P use efficiency than ‘B104-2’ (data not shown). This suggested that the high P efficiency of ‘Eyou Changjia’ could be attributed mainly to uptake efficiency, which was consistent with our previous results on the two cultivars (Duan et al. 2009). The data obtained in the present study showed that ‘Eyou Changjia’ was superior to ‘B104-2’ in root traits under the condition of P deficiency (Table 1). This indicated that the high PU efficiency of ‘Eyou Changjia’ was related closely to the development of an extensive root system. Thus, ‘Eyou Changjia’ could acquire more P from an LP environment and produce more biomass than ‘B104-2’, which represented higher tolerance to LP stress. Similar findings from earlier studies have demonstrated that the P-efficient cultivars of B. napus intercept more available P and have a higher phosphatase activity and greater total RL than P-inefficient ones (Solaiman et al. 2007; Zhang et al. 2009).

QTL mapping using the data from multi-environment experiments enables the identification of reliable QTL. Sixty-two significant QTL were identified in the three experiments performed (Fig. 2; Supplementary Table S3). Twenty-six of these QTL were identified in only one experiment, which suggested that these QTL were influenced readily by environmental factors. This was consistent with the significant effect of environment on the phenotype of the RIL population (Table 2). The remaining 36 QTL were detected repeatedly in at least two experiments, which suggested that these QTL were reliable in the RIL population.

The majority of the significant QTL clustered within six regions of four linkage groups (Table 3; Fig. 3). Under LP condition, two QTL clusters, which included QTL for plant DW, root traits and PU, were identified on C3. These clusters were then integrated into two unique QTL, uq.C3a and uq.C3b, respectively. Another QTL cluster for LP condition including QTL for RSA and RVol was located on A1 and was integrated into uq.A1. QTL for plant DW and root traits, which were identified under HP condition, coincided with QTL for PU on linkage groups A3 and C2, and were integrated into further two unique QTL, uq.A3d and uq.C2. These indicated that different mechanisms were involved in the response of B. napus to different P environments. In addition, uq.A3b was detected consistently for root traits and P content under both P conditions; hence it corresponded to a constitutively expressed QTL. Both uq.C3a and uq.C3b were specific QTL for root traits and PU under LP stress, which suggested that the two unique QTL corresponded to a single gene or gene clusters. These genes or gene clusters are likely to be involved in the metabolic pathway that is involved in the response of B. napus to P deficiency.

Arabidopsis has been exploited as a standard in comparative genomics analyses between Arabidopsis and Brassicaceae species. A set of 24 conserved chromosomal blocks (labeled A–X) has been identified among the ancestral karyotype, A.thaliana and B. rapa (Schranz et al. 2006). In the present study, we mapped the alignment between the B. napus linkage groups and Arabidopsis genome using the comparative mapping approach described by Long et al. (2007). One syntenic island (R) and one syntenic block (W) were identified between C3 of B. napus and chromosome 5 of Arabidopsis. uq.C3a and uq.C3b were located within the regions R and W, respectively (Fig. 4a). In B. oleracea, QTL for shoot DW, shoot P content and P use efficiency under LP condition have been identified in two regions of chromosome C3. The positions of these QTL aligned with blocks 5A, and 3C, 1B and 4B of Arabidopsis (Hammond et al. 2009). Comparison of the positions of uq.C3a with QTL on C3 in B. oleracea revealed that they were located within the same genomic block of Arabidopsis. We speculated the QTL on C3 of B. napus and B. oleracea co-localized. Furthermore, in silico mapping revealed that a number of orthologous genes for root development, auxin transport, and P metabolism in Arabidopsis were located in the confidence intervals of uq.C3a and uq.C3b; six genes corresponded to uq.C3a and eight to uq.C3b (Fig. 4b). Near-isogenic lines should be developed to allow further fine-mapping of these QTL and the cloning of potential candidate genes.

Fig. 4
figure 4

The alignment between the linkage group C3 of B. napus and chromosome 5 of Arabidopsis. a The distribution of two unique QTL, uq.C3a and uq.C3b, on C3 and the alignment between C3 and Arabidopsis genome. C3 is shown as a black thick line with vertical lines to indicate the positions of the molecular markers. The unique QTL, uq.C3a and uq.C3b, are drawn with horizontal bars where the length indicates the confidence interval and the circle indicates the peak position. The two blocks with “R” and “W” represent the two genomic regions of Arabidopsis that aligned to C3 of B. napus by comparative mapping approach. b Genes for root development, auxin transport, and P metabolism in Arabidopsis located in the confidence intervals of uq.C3a and uq.C3b by in silico mapping approach

Full size image

Two functional markers, BnIPS2-C3 and BnGPT1-C3, were located in the confidence intervals of two unique QTL (Table 3; Fig. 3). BnIPS2-C3, which was developed from AtIPS2, was located in the interval of the LP-specific QTL uq.C3a. AtIPS2 is up-regulated in roots in response to LP stress in Arabidopsis (Shin et al. 2006), and OsIPS2 in rice exhibits a similar pattern of expression (Hou et al. 2005). Therefore, BnIPS2 was a possible candidate gene for uq.C3a and might be involved in the P-deficiency-response pathway in B. napus. BnGPT1-C3, which was developed from AtGPT1, was located in the interval of the LP-specific QTL uq.C3b. AtGPT1 belongs to the P translocator family and apparently functions as an importer of glucose 6-phosphate (Niewiadomski et al. 2005). BnGPT1-C2 was also developed from AtGPT1 but was located in the interval of the constitutive unique QTL uq.C2. The results of in silico mapping demonstrated that AtGPT1 was located in the intervals of uq.C2 and uq.C3b. Therefore, AtGPT1 might be one of the candidate genes that underlie uq.C2 and uq.C3b (Fig. 4). From the information about the possible candidate genes, we assumed that the gene(s) that underlie uq.C3a might promote lateral root development under P starvation, resulting in improved PU, whereas the gene(s) that underlie uq.C3b might be responsible for P translocation from roots to shoots and P utilization in shoots under LP stress. This information could assist in identifying and cloning of the genes that underlie uq.C3a and uq.C3b, and that contribute to adaptability to P deficiency.

MAS is an efficient system for the selection of traits of interest than traditional phenotypic-based selection techniques. It relies on the availability of markers that are linked closely for traits or target QTL. For example, molecular markers that are linked tightly with QTL, Sub1, for submergence tolerance and with QTL, Pup1, for PU were developed in rice, and could be used to breed cultivars with high tolerance to submergence and P deficiency, respectively, by MAS (Neeraja et al. 2007; Heuer et al. 2009; Wissuwa et al. 2002). These previous studies have demonstrated that MAS can be used effectively in a molecular breeding program for abiotic stress. Therefore, another major goal of this study was to develop markers linked closely to target traits or QTL for MAS. In the study, two LP-specific unique QTL, uq.C3a and uq.C3b, were detected, which accounted for 9.4–16.8% of the phenotypic variation for plant DW, PU, RSA and RVol (Table 3). In addition, QTL for seed yield co-localized with uq.C3b under P deficiency (unpublished data). Therefore, uq.C3a and uq.C3b were represented important targets for MAS. Two FM, BnIPS2-C3 and BnGPT1-C3, which were located in the confidence intervals of these two QTL were linked significantly by single-marker analysis with plant DW, PU, RSA and RVol (P < 0.005) under LP conditions in at least two experiments (data not shown). The RI lines with favorable alleles related to BnIPS2 and BnGPT1 could develop larger RSA, produce higher DW, and acquire more P under LP condition, and yield more seed in the field trials (Supplementary Table S4). These suggested that two functional markers may be potential markers for MAS to breed P-efficient rapeseed cultivars.