Mapping Quantitative Trait Loci Associated with Aluminum Toxin Tolerance in NJRIKY Recombinant Inbred Line Population of Soybean (Glycine max)

To investigate the genetic mechanism of Al‐tolerance in soybean, a recombinant inbred line population (RIL) with 184 F2:7:11 lines derived from the cross of Kefeng No.1 × Nannong 1138‐2 (Al‐tolerant × Al‐sensitive) were tested in pot experiment with sand culture medium in net room in Nanjing. Four traits, i.e. plant height, number of leaves, shoot dry weight and root dry weight at seedling stage, were evaluated and used to calculate the average membership index (FAi) as the indicator of Al‐tolerance. The composite interval mapping (CIM) under WinQTL Cartographer v. 2.5 detected five QTLs (i.e. qFAi‐1, qFAi‐2, qFAi‐3, qFAi‐4 and qFAi‐5), explaining 5.20%–9.07% of the total phenotypic variation individually. While with the multiple interval mapping (MIM) of the same software, five QTLs (qFAi‐1, qFAi‐5, qFAi‐6, qFAi‐7, and qFAi‐8) explaining 5.7%–24.60% of the total phenotypic variation individually were mapped. Here qFAi‐1 and qFAi‐5 were detected by both CIM and MIM with the locations in a same flanking marker region, GMKF046‐GMKF080 on B1 and satt278‐sat_95 on L, respectively. While qFAi‐2 under CIM and qFAi‐6 under MIM both on D1b2 were located in neighboring regions with their confidence intervals overlapped and might be the same locus. Segregation analysis under major gene plus polygene inheritance model showed that Al‐tolerance was controlled by two major genes (h2mg = 33.05%) plus polygenes (h2pg = 52.73%). Both QTL mapping and segregation analysis confirmed two QTLs responsible for Al‐tolerance with relatively low heritability, and there might be a third QTL, confounded with the polygenes in segregation analysis.

Keywords: aluminum toxin tolerance; membership index; quantitative trait loci; recombinant inbred line population; soybean

Aluminum (Al) toxicity is a major constraint to plant growth and crop productivity in 40% of the world's arable soil ([21]). Al‐toxicity is caused by the uptake of toxic Al3+ and H+ ions in the acid medium, leading to a marked reduction in root growth and consequently in crop productivity ([12]). It is not sufficient to develop a sustainable production system in acid soils by application of lime and other fertilizers because of its un‐sustainability and extra expense, especially in developing countries in the tropics and sub‐tropics ([18]). A cost effective alternative is the fitting of Al‐tolerant cultivars to these problem soils ([6]).

Since aluminum tolerance in soybean was reported, studies have been undertaken to elucidate the genetics and the physiological processes underlying the trait. The effect of Al3+ on root growth of soybean has been investigated in nutrient solution ([2]) and sand culture ([16]) and apparent discrepancies between methods of evaluation of Al‐tolerance have been. However, sand culture is a potentially viable alternative as a substitute for hydroponics in screening for Al‐tolerance. It simulates field conditions and allows for the quantification of root characters that may be especially important to breeders in comparing multiple genetic sources of Al‐tolerance.

Advances in molecular marker technology have added a new dimension to the study of inheritance of genetic traits, offering opportunities for marker‐assisted selection in breeding programs and map‐based cloning of some favorable genes. DNA markers have proved to be efficient tools to detect important quantitative trait loci (QTLs) controlling quantitative traits in various crops. Since several well‐developed soybean molecular‐marker linkage maps have been constructed ([4]; [1]; [24]; [26]), QTL mapping has been reported for a number of agronomic and quality traits in soybean ([10]; [26]; [27]; [8]). Molecular marker analyses have identified linkage group regions with genes for pest resistance ([15]), salt tolerance ([14]) and phosphorus deficiency ([26]). It is rarely reported that genes controlling Al‐tolerance have been located in the soybean genome. However, [1] identified up to five QTLs for Al‐tolerance with the restriction fragment length polymorphism (RFLP) marker technology in soybean. [16] used a recombinant inbred line (RIL) population derived from the cross of Bogao and NG94‐156 to analyze the inheritance of five relative root traits by using the segregation analysis for quantitative traits under the major gene plus polygene mixed inheritance model ([9]) and demonstrated that the relative values of lateral root number, tap root length, total root length, and dry root weight between two parents were controlled by three major genes plus polygenes. Despite these, information on putative QTLs for Al‐tolerance is scant and therefore more investigations are needed. The objectives of the present study were to detect QTLs for Al‐tolerance and to determine the allelic effects on Al‐tolerance in a RIL population derived from the cross of Kefeng No.1 × Nannong 1138‐2 (Al‐tolerant × Al‐sensitive). Knowledge of the genetic basis for Al‐tolerance should facilitate the transfer of Al‐tolerance from exotic germplasm to economically important cultivars.

Table 1 shows the membership index values (Fij) of the four component traits, namely, those of plant height, number of leaves, shoot dry weight and root dry weight, which were computed from each replication and then averaged over three replications for each trait and each genotype, including parents and recombinant inbred lines. The Fij values of the four traits of Kefeng No. 1 were larger than those of Nannong 1138‐2, therefore, the average over four traits (FAi) for Kefeng No. 1 was larger than that of Nannong 1138‐2 (0.725 vs. 0.446). It confirmed the former being tolerant and the latter being sensitive to Al‐toxin and was consistent with earlier reports. The Fij values of the RILs varied among the lines and the coefficient of variation (CV), SD and range values of Fij showed that among the four traits, shoot dry weight and root dry weight were of higher sensitivity than that of a number of leaves and in turn, especially that of plant height. For a balance among the traits, the Fij values were averaged over the four traits for each material to obtain FAi. The segregation data of FAi for the RIL population and the related statistics are summarized in Table 2. The results showed that the FAi values of lines were significantly different (P < 0.01) and almost normally distributed with mean 0.426, range 0.191–0.732 and genotypic coefficient of variation (GCV) 24.16% (only a little less than the phenotypic CV), implying that the experiment was effective in detecting the wide segregation of FAi in the population. This observation was also supported by the high heritability value (88.36%) under the experimental conditions. A number of sensitive transgressions over the sensitive parent Nannong 1138‐2 were observed, but no other side transgression over the tolerant parent Kefeng No.1 was found.

1 The membership index values of component traits of parents and recombinant inbreeding lines (RILs) (F ij )

1 CV, coefficient of variation.

2 Frequency distribution of aluminum tolerance in soybean (FAi)

2 GCV, genotypic coefficient of variation; RIL, recombinant inbred line. **Significant at P = 0.01 (or very significant) level.

Composite interval mapping (CIM) and multiple interval mapping (MIM) were used to map QTLs for Al‐tolerance using FAi

(Table 3, Figure 1). Both methods indicated the existence of QTLs. With the CIM method, five QTLs (i.e. qFAi‐1, qFAi‐2, qFAi‐3, qFAi‐4, qFAi‐5) conferring Al‐tolerance in soybean were mapped on four linkage groups, explaining 5.20%–9.07% of the total phenotypic variation individually, whereas with the MIM method, five QTLs, (i.e. qFAi‐1, qFAi‐5, qFAi‐6, qFAi‐7 and qFAi‐8) explaining 5.70%–24.60% of the total phenotypic variation individually were detected on four linkage groups. Here, an independent QTL was defined if located between two adjacent markers without any other candidate QTL(s). Thus two QTLs, qFAi‐1 and qFAi‐5, were recognized as independent ones due to their detection by both methods and being located in the same flanking marker region, GMKF046‐GMKF080 on linkage group B1 and satt278‐sat_195 on linkage group L, respectively. Each of the two QTLs showed a negative additive effect explaining less than 10% of phenotypic variability under both methods. Accordingly, their existence may be genuinely associated with the respective markers. The other two QTLs, qFAi‐2 under CIM and qFAi‐6 under MIM, were located on adjacent regions of satt703 on D1b2 with respective distances of 2.3 cM and 7.0 cM on either side of the marker and their confidence intervals overlapped and therefore, might be a same locus. However, if the two QTLs are different, then qFAi‐6 is more likely to lie within the overlap region (logarithm of odds (LOD) threshold 3.7) and might be a major QTL (R2 = 24.6%) of Al‐tolerance, with qFAi‐2 (R2 = 5.2%) being a minor QTL.

3 Quantitative trait locus (QTL) mapping of aluminum toxin tolerance in soybean

3 aDis.‐1 and‐2 indicates the distance from QTL to the left and right markers, respectively. bThe distance from the top marker to the confidence interval. LOD, logarithm of odds.

Graph: 1 Quantitative trait locus (QTL) position on linkage groups in this experiment (composite interval mapping (CIM) and multiple interval mapping (MIM)).

Another two QTLs detected with MIM (i.e. qFAi‐7 and qFAi‐8) on linkage groups D1b2 and I, respectively, were quite different from the other two, qFAi‐3 and qFAi‐4, detected with CIM on linkage group D2. Since qFAi‐3 and qFAi‐4 located on nearby regions around satt397 with a narrow overlap of confidence intervals and both showed approximate additive effects (0.033 6 vs 0.029 3) and R2 values (9.07% vs 6.83%), it was postulated that the two QTLs might be the same. Anyway, further confirmatory investigation is needed since it was only detected by CIM.

Genetic analysis was carried out under a major gene plus polygene mixed inheritance model in the P1, P2, and F2:7:11 of the cross Kefeng No.1 × Nannong 1138‐2 (Table 4). The best‐fitted model was selected in two steps. In the first step, one or several elite models were selected under the principle of entropy maximization and AIC minimization. In the next step, the selected models were further tested for goodness‐of‐fit to find the best‐fitted model ([9]). The results showed that the best‐fitted model was E‐1‐9 and therefore, Al‐tolerance was controlled by two interacted major genes (one major gene inhibitory to other one) plus polygenes with a major gene heritability value of 33.05% and polygene heritability value of 52.73%, respectively. This indicated that Al‐tolerance was involved with both major genes and polygenes, with polygene heritability being more than the two major genes. This result might suggest greater genetic effect of polygenes on Al‐tolerance and further identity investigation is required.

4 Estimates of genetic parameters in segregation analysis

4 m, mean; i*, sum of additive and additive × additive epistasis effect; σ2mg, major gene variance; σ2pg, polygene variance; h2mg, major gene heritability; h2pg, polygene heritability.

Aluminum tolerance is considered as a quantitative trait controlled by many genes. Modes of gene action accounting for the variability may be due to dominance, additive and epistasis. Different Al‐tolerance mechanisms are identified with different screening methods ([17]) on different genotypes at different growth stages with different indicators ([20]). Relative root elongation has been considered as the most reliable indicator of Al‐tolerance evaluation in rice seedlings in solution culture ([25]). However, the hematoxylin staining and root re‐growth procedures have been used to evaluate Al‐tolerance in barley ([5]). In soybean, the physiological and biochemical mechanisms of the toxic effects of aluminum on root elongation have been extensively investigated ([11]; [13]; [3]). Sand culture method has been considered as a reproducible standard that mimics the actual growth environment and allows for phenotypic quantification of roots from older plants ([19]). This may be especially important to breeders in identifying multiple genetic sources of Al‐tolerance that might not be detected by hydroponics or hematoxylin staining procedures. In soybean, the inheritance of Al‐tolerance has been reported to be controlled by certain genes ([1]; [16]). The genetic system controlling Al‐tolerance appears to be complex, involving genes of major and minor effects and the use of only one trait (e.g. length or weight) may not be sufficient for in‐depth inquisition into the genetics of Al‐tolerance. In the present study, one parametric function (FAi) derived from measurements of shoot and roots traits was used to study the genetics of Al‐tolerance in soybean in sand culture medium, and according to our experiences it made the results rather stable and could be used for future studies.

With CIM, five Al‐tolerance QTLs were detected on the four linkage groups, explaining 33.25% of the total variation. The MIM method identified QTLs qFAi‐1 and qFAi‐5 located in the identical region on the same linkage groups. This phenomenon showed indeed the existence of the two QTLs (i.e. qFAi‐1, qFAi‐5) that enhanced aluminum tolerance in soybean. For a further check of the results, genetic analysis was carried out under a major gene plus polygene mixed inheritance model in the P1, P2, and F2:7:11 of the cross between Kefeng No. 1 × Nannong 1138‐2. The results showed that the Al‐toxin tolerance of the cross was controlled by two major genes plus polygenes. Accordingly, it validated further the real existence of the two QTLs and implied for research in genetic systems of quantitative traits in crops to adopt the use of alternative methods to validate one another to enhance the credibility and reliability of the results.

In soybeans, [2] found the cultivars Young and a soybean introduction from Japan, PI 416937, to be Al‐sensitive and Al‐tolerant respectively, and constructed a 120 F4‐derived progeny population from the cross of Young by PI 416937. Using RFLP analysis of this F4‐derived population, [1] detected up to five QTLs, each with minor effects, to condition varying levels of Al‐tolerance in soybean, indicating a multi‐genic level of control for Al‐tolerance.

In the present study, five QTLs, explaining about 5.20%–9.07% individually, were detected and mapped on four different linkage groups. The results were different from the earlier research work above probably because of two major factors in our experiment. First, the plant material was different between the two trials. In this study, a sample of 184 RILs derived from the cross between Nannong 1138‐2 and Kefeng No.1 were used to detect the QTLs. However, [1] used a set of 120 F4‐derived lines from the cross between Young and PI 416937. Different genetic backgrounds of the two sets of parents may partly explain the difference in QTL mapping for Al‐tolerance. Second, different screening methods may have caused differences in results. In this study, sand culture was used to identify Al‐tolerance, whereas most other experiments in Al‐tolerance have used hydroponics. Sand culture is supposedly close to the natural growth environment, so the result in sand culture (when sources of errors are carefully controlled) is expected to be more credible and reliable than that in hydroponics.

The inheritance and genetic expression of Al‐tolerance in soybean is complex and differs with growth stage. In this study, the number of QTLs for Al‐tolerance at the seedling stage in soybean was found, each explaining a minor effect on the total variation. This phenomenon indicated that the hereditary mechanism for Al‐tolerance at seedling stage was quite complex. Many QTLs for Al‐tolerance in soybean should exist and need to be identified; therefore, the potential of marker‐facilitated selection as a viable approach in breeding for Al‐tolerant soybeans should be expected.

In summary, the present results imply that not only the major genes but also the minor genes should be used towards a better usage of the genetic resources in breeding for Al‐tolerance, and the information might be useful for marker‐assisted selection (MAS) and map‐based cloning of Al‐tolerance genes.

A recombinant inbred line population (designated as NJRIKY) with 184 lines derived from a cross between a highly productive southern China Al‐sensitive variety Nannong 1138‐2 (male parent) and a black seeded Al‐tolerant variety Kefeng No.1 (female parent) was used in this research. The original population consisting of 206 F2:7:10 lines developed by single‐seed descent, was reduced to the 184 lines through deletion of some seriously biased lines by using the method of Simulated Population Sampling Criteria (SPSC) ([23]). The population was used for phenotyping and for linkage map construction.

An Al‐toxicity dose response using aluminum concentrations of 0, 7, 14, 21, 28 and 35 mg/L was generated with a standard Al‐tolerant soybean PI 416937 ([19]) to define the level of aluminum concentration to be used in the genetic studies ([16]). The concentration 28 mg/L detected a wide range of variation for most of the traits, and thus was used to test the population.

The RILs with both parents were tested for their Al‐tolerance under Al‐stress (designated as HIAL, 28 mg/L) versus Al‐free (NOAL, 0 mg/L) treatment. The seeds were soaked in pure water overnight and transferred to 1.5‐L plastic pots containing washed sand‐media culture. Plants were thinned to three per pot and then treated with the two treatments at the seventh day after planting. During the 2‐week treatment period, all pots were supplied, on a 2‐d interval (seven times), with 400 mL modified 1/5 strength Steinburg nutrient solution ([7]) with Al3+ (HIAL) and without Al3+ (NOAL) adjusted to pH 4.1 with 1 M HCl. To maintain Al3+ activity and to minimize interactions between Al3+ and other nutrients in the solution, Al3+ without nutrient solution was applied to the HIAL pots for the next 2 d following an Al3+‐nutrient treatment. This procedure is thought to achieve some temporal separation of Al3+ from nutrient solution and therefore masks the effects of cationic interactions ([19]). At the 14th day of treatment, four agronomic traits (i.e. plant heights, number of leaves, shoot dry weights and root dry weights) were evaluated to assess Al‐tolerance. Plant height was measured with a ruler before harvest. The number of fully opened leaves was counted also at the time of harvest. The shoot of each plant (above the cotyledon node) was removed and the roots (below the cotyledon node) were carefully separated from the sand by washing gently to minimize damage. The shoots and the roots were separately dried to constant weight at 65 °C for 48 h then weighed.

The experiment was in a split‐plot design with three replications. The main plots were assigned to the two Al3+ levels and the subplots were genotypes containing the 184 RILs and two parents. The entire experiment was conducted in June 2006 at Jiangpu Station, Nanjing Agricultural University, China. The mean temperature during the experimental period in the greenhouse was about 35 °C.

For each trait per genotype, data were transformed and mean membership index value (FAi) for Al‐tolerance was estimated according to the following formulae:

Graph

Where Fijk represents the value of ith genotype, jth trait, kth replication, X is defined as the ratio of trait value in HIAL to that under normal growth (NOAL). Xjk.max and Xjk.min represents, respectively, the maximum and minimum transformed value for trait j in replication k. The membership index value (Fij) of Al‐tolerance for the four traits was then computed for each genotype, thus:

Graph

The average membership index value (FAi) for Al‐tolerance was used for genetic analysis.

Graph

FAi index ranks the genotypic Al‐tolerance on certain criteria as that described by [16]. Here the membership index was calculated based on the ratio of stressed value to non‐stressed value, and then to take a ratio of the difference between the tested line and the minimum in the population relative to the whole range of the population to make the index comparable among the lines (even among different populations and environments). The four traits used were both above‐ and underground growth traits. Taking the average over the four traits was to make the index stable and comparable among the recombinant inbred lines.

The original linkage map of the population used in this study was constructed by [26]. The map consisted of 189 RFLP, 219 simple sequence repeats (SSR), 40 expressed sequence tag (EST), three resistance genes and one phenotype markers that were mapped onto 21 linkage groups covering 3 595.9 cM. All of the linkage groups except F (divided into F1 and F2) were consistent with those of the Cregan's consensus map. An improvement on this map has been made by adding new SSR markers and removing some data‐incomplete markers. The resultant new genetic linkage map, used in this study, covered 4 151.2 cM and consisted of 25 linkage groups. Each linkage group contained 19.5 evenly spaced markers with an average 8.6 cM between adjacent markers.

The anova and calculation of FAi were carried out by using the program SAS 9.0. In anova, the RIL effect was treated as a random model. The broad sense heritability was estimated as h2=σ2g/((σ2e/n) +σ2g), where σ2g, σ2e, and n represent genetic variance, error variance, and number of replications, respectively. The inheritance of FAi was analyzed by using the segregation analysis of quantitative trait under major gene plus the polygene mixed inheritance model given by [9] for the parents P1 and P2, and the 184 RILs.

To identify the QTLs for Al‐tolerance, the composite interval mapping (CIM) method was used and then compared with the multiple interval mapping (MIM) method under WinQTL Cartographer Version 2.5 ([22]) based on the phenotypic FAi measured across the population. The threshold for declaring a putative QTL for Al‐tolerance in CIM was LOD > 2.5, corresponding to a significance probability level 0.003 2, which was determined under a comprehensive consideration on the LOD value of 1 000 times of permutation test in CIM and the results of CIM, MIM and segregation analysis.

(Handling editor: Ai‐Min Zhang)