QTLs and epistasis for drought-tolerant physiological index in soybean (Glycine max L.) across different environments

Abstract

The additive and epistatic effects and their genotype × environment (GE) effects on percentage gain with physiological traits under different environments were analyzed in soybean. The genetic characters of physiological traits in soybean and characters of quantitative trait loci (QTL) detected in different environments are discussed. A backcross inbred line (BIL) population of soybean derived from the cross of strong drought tolerant wild soybean, “SNWS0048” as a recipient parent and drought-sensitive variety “Jinda73” as a donor parent was examined to identify the QTL and epistatic QTL and GE effects on drought-tolerant physiological traits in Shanxi, China, in 2008 and 2010. Six QTLs with additive (a) effects and/or additive × environment interaction (ae) effects and three pairs of QTLs with additive × epistatic main (aa) and/or epistasis × environment interaction (aae) effects were detected in different water environment over two years. Six QTLs with a effects and/or ae effects were mapped to linkage groups D2, G, M and N. Among them, there were two QTLs without ae effects, one with ae effects in all four environments, two QTLs with ae effects in two environments and one QTL with ae effects in only one environment. Among three pairs of QTLs, one pair was identified with aa effect and aae effect, one pair only exhibited aa effect, and one pair only had aae effect without aa effect. These indicated that different water environments could affect the expression of genes relevant to drought tolerant traits.

Keywords:

• soybean
• quantitative trait loci (QTLs)
• epistasis
• drought tolerance
• physiological index

Introduction

As one of the five major crops in the world, soybean is an important source of high-quality protein and edible oil (Wang et al. 2012). Soybean is usually grown in arid and semi-arid regions, and there have been many studies seeking to identify drought tolerant soybean cultivars. However, genetic improvement of drought tolerance is very difficult, because human selection has reduced the genomic diversity of cultivated soybeans (Hyten et al. 2006; Lam et al. 2010). Wild soybeans (Glycine soja) have a close genetic relationship with the cultivated soybeans (G. max) that were domesticated about 5000 years ago, and they have many advantages including strong resistance to different adverse surroundings. Since the two types of soybeans can interbreed, introgressions of genes from wild soybeans to cultivated soybeans have been reported, which may influence the genomic composition of cultivated soybeans (Lam et al. 2010).

Indicators of drought tolerance include leaf relative water content (RWC), relative electrical conductivity (REC) malondialdehyde content (MDA), chlorophyll content (CC) and the concentration of soluble sugars (CSS); these criteria could be used as the basis of selection for breeding programs (Sairam et al. 1997; Porcel and Ruiz 2004; Mohammadkhani and Heidari 2008; Zhu et al. 2008; Zheng et al. 2010). However, measuring these indicators has proven difficult. Advances in DNA marker technology offered hope that marker technology can begin to clarify the genetics of drought tolerance as an aid to practical breeding. To date, a series of genetic maps have been established to study quantitative trait loci (QTLs) underlying drought tolerance in soybean, including QTLs for water use efficiency (WUE), leaf ash (LASH), water retention curve (WRC), leaf pubescence density and water status traits, relative root traits, and yield under water-stressed conditions (Mian et al. 1996, 1998; Specht et al. 2001; Liu et al. 2005; Du, Fu, et al. 2009; Du, Wang, et al. 2009; Li et al. 2011). However, identification of QTLs for physiological indices including RWC, REC, MDA, CC and CSS is rare in the literature.

Lu et al. (2011) estimate that F₂ and backcross inbred line (BIL) populations have the ability to detect relatively more QTLs or the information than others. Thus, in this study, a BIL population derived from a cross between a cultivated and a wild soybean was used to identify and map QTLs associated with drought tolerance involved in physiological traits.

Materials and methods

Plant materials

A mapping population with 200 BILs derived from a backcross of JD73 / SNWS0048 //JD73 by the single-seed descent method was employed to test the presence of drought tolerance over the whole genome of soybean. All seeds were harvested from each individual BC₁F₁ plant and one seed from each plant was chosen randomly and planted in the field to obtain BC₁F₂ plants. This procedure was continued in the following generations until BC₁F₄ plants were obtained. The recipient SNWS0048 was an accession of common wild soybean from Shanxi Province, China, with strong drought tolerance; the donor JD73 was a drought-sensitive variety with superior agronomic traits.

Solution pot culture experiment

The entire experiment was conducted in a greenhouse at a temperature of 27–32°C and relative air humidity range from 57 to 67%, in 2008 and 2010 at Shanxi Agricultural University, Taigu of Shanxi province, China. The BIL population and the parents were arranged in a split block design with two replicates and each replication consisted of two treatments (control and water-stress). Nine seedlings were planted into a 10-liter pot (35 cm in diameter and 35 cm in height) filled with 9.5 kg of mixed soil (soil:sand = 1:1). At 7 days after emergence, plants were thinned to three similar-sized plants per pot. For soybean, 70% and 30% of available water content (AWC) in the soil were considered to be control check (CK) and water stress (WS) conditions, respectively. The soil moisture was maintained with required amounts of water for the pots of CK and WS conditions by weighing pots and watering the plants every day.

Drought resistance screening

RWC was measured according to a previously described method (Larbi and Mekliche 2004). About 0.5 g of fresh leaves were cut into fragments (squares of 2 × 2 cm²) and weighed for the fresh weight (FW), then saturated in water for 8 h at 4°C and weighed for the turgid weight (TW). To obtain dry weight (DW), the leaf samples were oven-dried at 80°C for 24 h and then weighed. Leaf RWC (%) was calculated as: (FW – DW)/(TW – DW) ×100%.

REC was assessed according to a previously described method (Liu et al. 1997) with a few modifications. Fresh leaves were punched into 10 fragments (1 cm in diameter) and were saturated in water for 3 h, then electrical conductivity (Ec1) was monitored with a DDS-11A hand electronic conductivity meter. Then, the samples were boiled in a water bath for 10 min and cooled to room temperature: the electrical conductivity (Ec2) was measured, and the REC was calculated as: Ec1/Ec2 × 100%.

MDA content was measured by the thiobarbituric acid test according to a previously described method with a few modifications (Sairam et al. 1997). About 0.1 g of fresh leaves (FW) were homogenized in 5 ml of 10% trichloroacetic acid (TCA). The homogenate was centrifuged at 4000 g for 10 min. To 2 ml aliquot of the supernatant, 2 ml of 0.6% thiobarbituric acid (TBA) in 20% TCA was added. The mixture was heated at 95°C for 15 min and then quickly cooled in an ice bath. After centrifugation at 10000 g for 10 min, the absorbance of the supernatant was recoded at 532 nm (D₅₃₂), 600 nm (D₆₀₀) and 450 nm (D₄₅₀). The concentration of MDA was calculated as: C (μmol l⁻¹) = 6.45(D₅₃₂ – D₆₀₀) – 0.56D₄₅₀. The MDA content was calculated as: MDA content (μmol g⁻¹) = C (μmol l⁻¹) × 5 ml/FW (g).

The content of soluble sugars was determined based on the method for measuring MDA content. CSS was calculated as: C (mmol l⁻¹) = 11.71D_450. CC was measured using a SPAD-502, and it was calculated as: Y = 0.0996X – 0.152, where X is the SPAD-502 reading, and Y is the chlorophyll content (mg dm⁻²).

Construction of genetic linkage map and QTL mapping

The genetic map of this population was constructed using the Mapmaker/EXP 3.0b software (Lander et al. 1987). The map contained 122 markers distributed among 24 linkage groups covering 1655.4 cM with an average distance of 13.6 cM between markers (Wang et al. 2012). The means of the traits were used to identify QTLs by composite interval mapping using QTL Network 2.0 (Yang et al. 2007, 2008) software. The QTLs were detected according to the mixed linear model, and the test standard is p < 0.005 significance (Zhu 1997).

Results

Phenotypic variation and correlation between traits

The phenotypic values of physiological traits in the BIL population in water stress and control are shown in Table 1. The value of RWC was lower in water stress conditions, while REC, MDA, CC and CSS values were higher in water stress than in the control. Water stress increased MDA, CC and CSS and decreased RWC in both years, and significant differences between the two water regimes were found under such conditions according to analysis of variance (ANOVA). These indicated that the relative performance of genotypes changed depending on the water regime.

Table 1. The performances of physiological traits in BIL population in water stress and control.

Physiological trait			BIL population				Parents		T effect
			Mean	SD	Range	CV(%)	JD 73	SNWS0048
RWC	2008	CK	78.52	6.23	61.03–97.21	7.94	72.39	83.22	**
(%)		WS	66.86	4.62	53.09–84.38	6.91	64.08	69.95
	2010	CK	82.15	6.22	64.41–97.45	7.57	75.83	87.71	**
		WS	74.44	6.31	51.19–88.29	8.47	66.28	79.44
REC	2008	CK	0.29	0.08	0.14–0.61	26.29	0.41	0.22	**
		WS	0.43	0.11	0.18–0.90	26.59	0.54	0.32
	2010	CK	0.26	0.10	0.11–0.61	37.97	0.38	0.21	**
		WS	0.42	0.16	0.18–0.92	37.93	0.55	0.30
MDA	2008	CK	35.35	9.14	11.90–78.53	25.85	44.86	32.32	**
(μmol g^−1)		WS	50.50	13.05	18.89–94.85	25.85	53.39	45.24
	2010	CK	27.33	7.05	9.16–50.65	25.78	33.16	19.02	**
		WS	42.02	14.94	17.59–85.44	35.56	48.63	39.99
CC (mg dm^−2)	2008	CK	3.04	0.21	2.49–3.83	6.93	2.95	3.26	**
		WS	3.51	0.22	2.95–4.26	6.20	3.32	3.76
	2010	CK	2.83	0.23	2.09–3.54	8.11	2.58	3.01	**
		WS	3.08	0.19	2.61–3.71	6.28	3.01	3.26
CSS (mmol l^−1)	2008	CK	6.02	2.58	2.68–14.58	42.83	8.33	5.13	**
		WS	10.59	1.60	4.20–16.51	15.09	12.20	8.47
	2010	CK	4.42	1.35	2.29–10.49	30.48	5.08	3.61	**
		WS	8.02	2.02	3.90–14.01	25.23	10.19	6.77

The RWC and CC in water stress were greater for the non-recurrent parent SNWS0048 than for recurrent JD73, with RWC of 69.95% and 64.08% in 2008 and 79.44% and 66.28% in 2009; and CC of 3.76 and 3.32 mg dm⁻² in 2008 and 3.26 and 3.01 mg dm⁻² in 2009. Meanwhile, the REC, MDA and CSS in water stress were lower for SNWS0048 than for JD73. REC of the two parents in water stress was 0.32 and 0.54 in 2008, and 0.30 and 0.55 in 2010; MDA was 39.99 and 48.63 μmol g⁻¹in 2008, and 45.24 and 53.39 μmol g⁻¹ in 2009; and CSS was 6.77 and 10.19 mmol l⁻¹ in 2008, and 8.47 and 12.20 mmol l⁻¹ in 2010, respectively. These suggested that the two parents differ in the genes controlling these traits. The values of SNWS0048 indicated that it experienced a higher level of drought tolerance compared to recurrent JD73.

For all traits, the mean values of the BIL plants were intermediate to those of the parents. The coefficient of variation (CV) ranged from 6.20 (in 2008 WS for CC) to 42.83 (in 2008 CK for CSS). Values of BIL plants outside of the parental range were observed to different extents for the majority of traits, indicating that the alleles increasing the phenotypic values were dispersed in both parents. No matter how large or small the trait difference between the two parental lines, the variation of each trait in the BIL population was large and continuous. The distributions of phenotypic data were suitable for QTL analysis.

In water stress, drought tolerance index (DTI) had a positive correlation with RWC, a significant negative correlation with other traits (bottom row in Table 2). This indicated that REC, MDA, CC and CSS were important indicators appraising the drought tolerance of soybean.

Table 2. Correlations between physiological traits and drought tolerance index (DTI) of BIL lines in water stress.

	RWC	REC	MDA	CC	CSS
REC	–0.2591**
MDA	–0.1939*	0.1913*
CC	–0.0139	0.2995**	0.0332
CSS	–0.1061	0.2602**	0.1502	0.3571**
DTI	0.1130	–0.5299**	–0.3771**	–0.5974**	–0.5090**

QTLs with additive effects and additive × environment interaction effects

An additive main (a) effect is the accumulated effect expressed in the same way across different environments, while an additive × environment interaction (ae) effect is the deviation due to a specific environment. Under a specific environment, the total effect of a QTL could include the main effects plus QTL × environment (QE) interaction effects in that environment.

QTLs detected with a effect and/or ae effect associated with drought tolerance of soybean in BIL population at flowering stage were identified to specific chromosomal regions in the linkage map (Table 3). Six QTLs with a effects and/or ae effects involved in drought tolerance of soybean were mapped to linkage groups D2, G, M and N. All QTLs had significant a effects at the 0.005 level. The drought tolerance parent SNWS0048 contributed alleles for increasing drought tolerance at QTLs qREC-D2-3 and qREC-N-4, but for decreasing drought tolerance at qRWC-M-2, qMDA-G-2, qCC-D2-4 and qCSS-G-5. This suggested that alleles for drought tolerance were dispersed within the two parents, resulting in small differences of phenotypic values between parents and transgressive segregants among the BIL population.

Table 3. The QTL locations and estimated effects (additive a and additive by environment ae) associated with physiology traits of soybean in the BIL population.

Traits	QTL	MLG	Marker interval	Location/cM	a effects	h^2(a)/%	ae in 2008CK	ae in 2008WS	ae in 2009CK	ae in 2009WS	h^2(ae)/%
RWC	qRWC-M-2	G21–M	Satt567–Satt551	43.5	–0.7543	0.62		–1.8086	1.5280		1.73
REC	qREC-D2-3	G10–D2	Satt311– Satt528	12.4	–0.0450	11.69
	qREC-N-4	G22–N	Satt237–Sat_241	69.6	–0.0542	11.57
MDA	qMDA-G-2	G14–G	Sat_223–Satt594	21.6	1.8086	0.99		3.5149			1.79
CC	qCC-D2-4	G10–D2	Satt528–Sat_365	14.0	0.0691	1.90	–0.1043			0.0662	2.04
CSS	qCSS-G-5	G14–G	Satt217–Satt352	76.2	0.5528	2.83	–0.6049	0.8139	–0.6052	0.3964	5.21

Of the six QTLs, there were two QTLs without ae effects, one with ae effects in all four environments, two QTLs with ae effects in two environments and one QTL with ae effects in only one environment. One QTL detected to be responsible for RWC showed significant a and ae effects; it was located within the interval Satt567–Satt551on the G21-M linkage group named as qRWC-M-2. It decreased RWC by 0.7543%, with corresponding contribution of 0.62%. It reduced RWC by 1.8086% in 2008WS, while enhancing RWC by 1.5280% in 2009CK, with a corresponding contribution of 1.73%. Two QTLs affecting REC were identified as being significant in a effects; they were located on linkage group G10-D2 and G22-N, and were named qREC-D2-3 and qREC-N-4. The QTL of qREC-D2-3 located within the interval Satt311–Satt528 reduced REC by 0.0450, with a corresponding contribution of 11.69%. The other QTL of qREC-N-4 located within the interval Satt237–Sat_241 reduced REC by 0.0542, with corresponding contribution of 11.57%.

One QTL mapped for MDA was significant in a and ae effects; this was located within the interval Sat_223-Satt594 on linkage group G14-G, and was named qMDA-G-2. It increased MDA by 1.8086 μmol g⁻¹, with a corresponding contribution of 0.99%. It enhanced MDA by 3.5149 μmol g⁻¹ in 2008WS, with a corresponding contribution of 1.79%. One QTL for CC was identified as being significant in a and ae effects; this was located within the interval Satt528–Sat_365 on linkage group G10-D2, and was named qCC-D2-4. It increased CC by 0.0691 mg dm⁻², with a corresponding contribution of 1.90%. It reduced CC by 0.1043 mg dm⁻² in 2008CK, while enhancing CC by 0.0662 mg dm⁻² in 2010WS, with a corresponding contribution of 2.04%. One QTL controlling CSS was significant in a and ae effects; this was located within the interval Satt217–Satt352 on linkage group G14-G, and was named qCSS-G-5. It increased CSS by 0.5528 mmol l⁻¹, with a corresponding contribution of 2.83%. In 2008, it reduced the CSS by 0.6049 mmol l⁻¹, while enhancing the CSS by 0.8139 mmol l⁻¹. In 2010, it decreased the CSS by 0.6052 mmol l⁻¹, while increasing the CSS by 0.3964 mmol l⁻¹, with a corresponding contribution of 5.21%.

Most QTLs had ae effects, indicating that environments could affect the expression of genes for relevant drought tolerant traits, especially different water environments.

Epistatic effects on drought tolerance

Additive × additive epistatic main (aa) effects and epistasis × environment interaction (aae) effects could be analyzed along with a effects and ae effects. Three pairs of QTLs had interactions with each other. Among these epistatic interactions, one pair was identified with aa and aae effects, one pair only exhibited aa effect, and one pair only had aae effect without aa effect (Table 4).

Table 4. The QTL locations and estimated effects (epistasis aa and epistasis by environment aae) associated with physiology traits of soybean in the BIL population.

Traits	QTL	Marker interval	Location/cM	aa effects	h^2(aa)/%	aae in 2008CK	aae in 2008WS	aae in 2009CK	aae in 2009WS	h^2(aae)/%
REC	qREC-3-1	Satt509–Satt251	10.0	–0.0424	2.03			–0.0532	0.0348	2.10
	qREC-17-4	Satt132–Satt215	88.3
	qREC-6-1	Satt640–Satt281	3.0			–0.0214			0.0360	2.12
	qREC-15-3	Satt469–Satt541	53.7
CC	qCC-14-2	Sat_223–Satt594	22.6	–0.0937	0.85
	qCC-22-3	Satt257–Satt237	63.2

For one pair of QTLs, qREC-3-1 and qREC-17-4, located between the marker intervals Satt509–Satt251 and Satt132–Satt215, respectively, the aa effect in parent type decreased REC by 0.0422, with a corresponding contribution of 2.03%. In 2010, the aae effect decreased REC by 0.0532 in CK, while increasing REC by 0.0348 in WS, with a corresponding contribution of 2.10%. The other pairs of QTLs, qREC-6-1 and qREC-15-3, had no aa effect and took part in the aae effect. The aae effect decreased REC by 0.0214 in 2008CK, while increasing REC by 0.0360 in 2010WS, with a corresponding contribution of 2.12%. For one pair of QTLs, qCC-14-2 and qCC-22-3, located between the marker intervals Satt509–Satt251 and Satt132–Satt215, the distances from the left marker were 22.6 and 63.2 cM, respectively; the aa effect in parent type decreased CC by 0.0937 mg dm⁻², with a corresponding contribution of 0.85%.

These QTLs did not show a effects but showed aa and/or aae effects, and these QTLs involved in 6 loci probably took effects through modifying other gene loci.

Pleiotropy and tightly linkage

Table 3 shows that the same locus could associate with two traits with its alleles performed in their own way in direction and size. For example, on the locus of Satt528, the allele has a negative effect on REC but a positive effect on CC. The same allele conferring two related traits, or the pleiotropy of an allele, might be the genetic basis of their phenotypic correlation.

Discussion

Some of the QTLs mapped in the present study were mapped to regions where drought tolerant traits were previously reported. Du, Wang, et al. (2009) used a recombinant line population of 184 F_2:7:11 lines from a cross of Kefeng1 and Nannong1138-2 to map QTLs for the seed yield and drought susceptibility index, and a QTL for yield under water stress in the greenhouse associated with satt_223 on MLG G was detected. We mapped a QTL for MDA by satt_223. Du, Wang, et al. (2009) also reported QTLs for water stress index in the greenhouse, located on MLG D2. QTLs for REC and CC in present paper were also detected on linkage group D2. These QTLs may be related to soybean drought tolerance in the greenhouse. We mapped a QTL for CC marked by Satt528–Sat_365 on MLG G10-D2, coinciding with the same region of a QTL involving germination potential, germination rate, radicle length, root hair number and germination index in germination stage reported by Wang et al. (2012). This QTL had a genetic effect on more than one trait, and seems to act as a major gene. Some other QTLs were mapped to regions where other traits were previously reported. QTLs associated with Satt567 were detected controlling insect resistance, seed protein content, pod maturity date, reproductive stage, seed abortion, seed glutamine content, etc. (Terry et al. 2000; Specht et al. 2001; Csanadi et al. 2001; Orf et al. 1999; Wang et al. 2004; Tischner et al. 2003; Komatsu et al. 2005; Panthee et al. 2006). In this paper the QTLs for RWC were detected near Satt567. The locus of Satt311 was found to control lodging (Reinprecht et al. 2006), content of oil (Wen et al. 2008), yield and biomass (Zhang et al. 2008), yield (kg hm⁻²) and 100-seed weight (Zhang et al. 2009), and we detected a QTL for REC in the vicinity of Satt311. Alcivar et al. (2007) identified a QTL for internode length near Satt594 and we identified a QTL for MDA near the marker Satt594. Zhang et al. (2010) detected a QTL controlling flower number in the vicinity of Satt352, and we detected a QTL for CSS near Satt352. The above results imply that the same locus could associate with multiple traits.

Additive effects, additive × additive epistatic effects, and their QE interaction effects were considered as important genetic effect for complex traits (Zhao et al. 2005; Jiang et al. 2011; Korir et al. 2011; Ha et al. 2012). The mapping procedure QTLNetwork2.0 can integrate multiple QTLs, epistasis effects (not limited for main-effect QTLs), and QTL × environment and epistasis × environment interactions into one mapping system, and therefore the additive and epistatic effects, and their interactions with environments, can all be identified simultaneously (Gai et al. 2012). In the trials, additive effects, epistatic effects and the QTL-by-environment interactions associated with five physiological traits related to drought resistance across various years and water conditions were studied. Among four QTLs with a effects and ae effects, a high genotype × environment interaction variance was observed that was higher than the genetic variance, which was reflected by the high heredity frequency of QTL showing ae effects. This implies that these QTLs would be affected by different environments, especially by different water environments. The other two QTLs controlling REC with only a effects would be not affected by different environments. The QTL mapping also demonstrated a substantial contribution to the variation in drought tolerance by aa epistatic effects. Three types of epistatic interactions can be distinguished: interactions between two QTLs with additive effects (type 1), interactions between a QTL with additive effect and a “background” locus without additive effect (type 2), and interactions between two loci showing epistatic effects only (type 3) (Zhao et al. 2005). In this study, all three pairs of QTLs with aa effect and/or aae effects have identified as type 3, which is in agreement with earlier studies (Li et al. 2001; Luo et al. 2001). These loci without additive effects involved in interactions may represent genes with additive effects that were too small to be significant. Of them, two pairs of QTLs have aae effects, indicating that they interact with two loci under different environments.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (Grant No. 31171580), Programme of Shanxi science infrastructure platform (NO.: 2011091004-0103), the R & D Infrastructure and Facility Development Program of Shanxi Province, China (Grant No. 20110910040103), the Specialized Research Fund for the Doctoral Program of Higher Education (Grant No. 20111403120001), Shanxi Province Science Foundation for Youths (Grant No. 2012021023-3), the Shanxi Agricultural University Breeding Fund of China (2011001), the Shanxi Agricultural University Introduced the Talented Person Scientific Research Start Funds Subsidization Project of China (XB2010010), and plans to support the youth of top-notch innovative personnel in Shanxi Agricultural University (Grant No.201203).