Quantitative trait locus mapping for seed mineral concentrations in two Arabidopsis thaliana recombinant inbred populations

Authors

  • Brian M. Waters,

    1. USDA/ARS Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, 1100 Bates Street, Houston, TX 77030, USA
    Search for more papers by this author
  • Michael A. Grusak

    1. USDA/ARS Children's Nutrition Research Center, Department of Pediatrics, Baylor College of Medicine, 1100 Bates Street, Houston, TX 77030, USA
    Search for more papers by this author

  • The contents of this publication do not necessarily reflect the views or policies of the US Department of Agriculture, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government.

Author for correspondence:
Michael A. Grusak
Tel:+1 713 798 7044
Fax:+1 713 798 7078
Email: mgrusak@bcm.edu

Summary

  • • Biofortification of foods, achieved by increasing the concentrations of minerals such as iron (Fe) and zinc (Zn), is a goal of plant scientists. Understanding genes that influence seed mineral concentration in a model plant such as Arabidopsis could help in the development of nutritionally enhanced crop cultivars.
  • • Quantitative trait locus (QTL) mapping for seed concentrations of calcium (Ca), copper (Cu), Fe, potassium (K), magnesium (Mg), manganese (Mn), phosphorus (P), sulfur (S), and Zn was performed using two recombinant inbred line (RIL) populations, Columbia (Col) × Landsberg erecta (Ler) and Cape Verde Islands (Cvi) × Ler, grown on multiple occasions. QTL mapping was also performed using data from silique hulls and the ratio of seed:hull mineral concentration of the Cvi × Ler population.
  • • Over 100 QTLs that affected seed mineral concentration were identified. Twenty-nine seed QTLs were found in more than one experiment, and several QTLs were found for both seed and hull mineral traits. A number of candidate genes affecting seed mineral concentration are discussed.
  • • These results indicate that A. thaliana is a suitable and convenient model for discovery of genes that affect seed mineral concentration. Some strong QTLs had no obvious candidate genes, offering the possibility of identifying unknown genes that affect mineral uptake and translocation to seeds.

Introduction

On a worldwide basis, plants are an important source of human food. Plant-based foods often have low mineral density and, as a result, a large proportion of the world's population suffers from mineral malnutrition, especially for iron (Fe) and zinc (Zn). In recent years, plant scientists have adopted a strategy known as biofortification in order to address this problem (Grusak & DellaPenna, 1999; White & Broadley, 2005). The goal of biofortification is to increase nutrient density in the edible portions of crop plants, which for many important staple crops, such as rice (Oryza sativa), wheat (Triticum aestivum), maize (Zea mays), bean (Phaseolus vulgaris) and other legumes, are seeds.

Achieving biofortification of crops is a major challenge. The physiology and regulation of mineral uptake and translocation to seeds in plants are not well understood (Briat et al., 2007; Krämer et al., 2007; Zhang et al., 2007). Thus, it is unclear which genes should be targeted in breeding programs or in transgenic efforts to accomplish crop improvement. Additionally, it is unknown what other modifications may be needed directly in seeds to allow accumulation of those minerals that are potentially toxic to the plant at elevated concentrations. Use of a small, fast-growing model plant, such as Arabidopsis thaliana, for discovery of genes that affect seed mineral concentration could save considerable time and effort as compared to working directly in crops that require more time, space, and labor to grow. Arabidopsis thaliana is closely related to the Brassica genus, which includes several important crop species. Seeds of some of these crops, such as broccoli (Brassica oleracea), are consumed as sprouts, and rapeseed (Brassica napus) meal is commonly used in animal feed or as an oil source. Additionally, genetic and genomic resources for A. thaliana are highly developed and available to the plant science community.

Initial biofortification efforts have focused on overexpression of single genes to increase mineral uptake or storage (Goto et al., 1999; Vasconcelos et al., 2003; Ramesh et al., 2004; Vasconcelos et al., 2004, 2006). Analysis of mineral overaccumulation mutants indicates that translocation of minerals to seeds is tightly regulated, and that simply increasing uptake into the plant will probably not result in seeds with higher mineral contents or concentrations. It is likely that multiple genes will need to be overexpressed in tandem or at appropriate developmental stages in order to increase both mineral intake into the plant and mineral translocation to the target tissues. These targets may include not only genes that increase mineral uptake, but also genes for internal transporters, such as for vascular tissue loading or unloading, or for organelle influx or efflux. Another category of potential target genes includes metal chaperones or chelators that are necessary for metal transport, storage, or detoxification. Regulatory genes that alter expression of entire pathways are another potential category of target genes.

Molecular genetics or genomic approaches have been used in preliminary steps to identify genes that allow certain plants to highly accumulate minerals such as cesium (Payne et al., 2004), selenium (Zhang et al., 2006) and Zn (Filatov et al., 2007). Microarray comparisons of hyperaccumulators and nonhyperaccumulators have revealed many genes that are differentially expressed in these plants (Becher et al., 2004; Weber et al., 2004; Filatov et al., 2006; Hammond et al., 2006; Talke et al., 2006; van de Mortel et al., 2006). Quantitative trait locus (QTL) mapping has been performed in bean (Guzmán-Maldonado et al., 2003; Gelin et al., 2007) and rice (Stangoulis et al., 2007), and in an A. thaliana recombinant inbred line (RIL) population to identify loci that influence seed phosphorus (P) (Bentsink et al., 2003) and other mineral characteristics (Vreugdenhil et al., 2004). In this work, we extend the use of the model plant A. thaliana as a source of discovery of genes that alter seed mineral concentrations. We present QTL mapping data for two growth cycles of the Cape Verde Islands (Cvi) × Landsberg erecta (Ler) population and three cycles of the Columbia (Col) × Ler population to examine the reproducibility of QTL results. Additionally, QTLs were mapped for mineral concentrations in silique hulls and the ratios of mineral concentrations in mature seeds:hulls.

Materials and methods

Plant material and growth conditions

Seeds of Arabidopsis thaliana (L.) Heynh. were obtained from the Arabidopsis Biological Resource Center at The Ohio State University, USA. In addition to RIL populations Col × Ler (CS1899) and Cvi × Ler (CS22000), erecta mutant lines er-114 (CS3918), er-116 (CS3920), er-117 (CS3921), and er-123 (CS3927), and T-DNA lines heavy metal P1b-ATPase 5 (hma5; SALK_040252C), yellow stripe-like 8 (ysl8; CS859713), ferritin 2 (fer2; SALK_002947), and zinc regulated transporter, iron regulated transporter-like protein 5 (zip5; SALK_009007) were grown. Seeds were placed in 0.1% agar at 4°C for 3–5 d before sowing on commercial potting mix (MetroMix 300; Sun Gro Horticulture, Bellevue, WA, USA). For all QTL studies and experiments with T-DNA lines, plants were grown in an air-conditioned glasshouse under shadecloth (to reduce sunlight intensity) during winter months for the 2005 and 2006 experiments, and during late winter/early spring months for the 2003 and 2007 experiments. Fluorescent lighting was supplied at approx. 100 µmol photons m−2 s−1 for a 16-h photoperiod. Water and nutrients were provided by subirrigation as needed (usually twice per week) as a solution of the following composition: 1.2 mM KNO3, 0.8 mM Ca(NO3)2, 0.8 mM NH4H2PO4, 0.3 mM KH2PO4, 0.2 mM MgSO4, 25 µM CaCl2, 25 µM H3BO3, 2 µM MnSO4, 2 µM ZnSO4, 0.5 µM CuSO4, 0.5 µM H2MoO4, 0.1 µM NiSO4 and 10 µM Fe-EDDHA as Sprint 138 (Becker-Underwood, Ames, IA, USA). For low-nutrient treatments in erecta mutant experiments, plants received nutrient solution once, at 21 d after sowing, and deionized water at all other times. For high-light treatments in erecta mutant experiments, plants were removed from the shadecloth area after 2 wk and grown under ambient sunlight with supplemental lighting supplied by metal halide lamps on a 15-h photoperiod. Plants of each RIL were sown in three pots at a density of 3–5 plants per pot. At maturity, seeds from all plants were bulked, and a minimum of two replicate subsamples were used for inductively coupled plasma-optical emission spectrometry (ICP-OES) analysis. Silique valves or hulls (when collected) were collected at the same time as seeds collected for mineral analysis. Silique hulls were carefully cleaned and inspected to remove seed, floral, and leaf tissue.

Mineral analysis

Plant tissues were oven-dried at 60°C for 48 h before determination of dry weight (DW). Samples of 0.1–0.25 g were digested in nitric-perchloric acid (4 : 1) using a ramped heating protocol going from 100 to 220°C, and remaining at 220°C until samples were taken to dryness. Residues were re-suspended in 15 ml of 2% nitric acid. All acids were trace metal grade (Fisher Scientific, Pittsburgh, PA, USA) and water was filtered through a MilliQ system (Millipore, Billerica, MA, USA) to at least 18 MΩ resistivity. Concentrations of calcium (Ca), copper (Cu), Fe, potassium (K), magnesium (Mg), manganese (Mn), P, sulfur (S), and Zn were determined by ICP-OES (CIROS ICP Model FCE12; Spectro, Kleve, Germany).

QTL mapping and candidate gene selection

The Col × Ler (Lister & Dean, 1993) and Cvi × Ler (Alonso-Blanco et al., 1998) RIL populations were previously mapped. For seed mineral concentrations, in 2003 and 2007, 100 lines of the Col × Ler population were analyzed, and in 2005, 97 lines were analyzed. In 2003, 159 lines of the Cvi × Ler population were analyzed for seed mineral concentrations, and in 2006 146 lines were analyzed for seed mineral concentrations and 84 lines were analyzed for silique hull mineral concentrations. Genetic markers and comparisons of these genetic maps are available on the Natural-EU project website (http://www.dpw.wau.nl/natural/). QTLs were mapped by composite interval mapping using WinQTL Cartographer (Wang et al., 2007). A likelihood ratio (LR) significance threshold of P = 0.05 was determined for each trait by performing 1000 permutations before mapping (Supporting Information Table S1). Genetic markers that have been anchored to the physical map for Cvi × Ler (Peters et al., 2001) and Col × Ler (http://www.arabidopsis.org) were used to estimate the boundaries of confidence intervals. Annotated genes known to be involved with mineral uptake or mineral homeostasis (or family members of such genes) that fell within these confidence intervals were considered to be candidate genes.

Results

The Col × Ler RIL (CL) population was grown on three occasions (2003, 2005 and 2007), and the Cvi × Ler RIL (CVL) population was grown on two occasions (2003 and 2006). Seed mineral concentrations of both populations and all experiments exhibited wide ranges from low to high (Table 1). The ranges were lowest for Mg and highest for Cu, and were generally consistent between occasions. Silique hulls were collected and analyzed for CVL in 2006. Silique hulls had wider mineral concentration ranges than did seeds, except for Ca. Hull values were much higher than seed values for Ca, K, and Mg, as expected (Waters & Grusak, 2008).

Table 1.  Mineral concentrations of seeds and hulls of Arabidopsis thaliana Columbia (Col) × Ler (CL) and Cvi × Ler (CVL) recombinant inbred line (RIL) populations
Mineral (µg g−1)Experiment
CVL 2003CVL 2006CVL 2006 hullsCL 2003CL 2005CL 2007
  1. Data are for seeds unless otherwise stated.

Ca2630–82962475–606121 009–40 6193455–78722882–64183489–6634
Cu3.98–13.393.69–13.52.2–9.43.4–10.72.9–8.62.9–8.4
Fe57.2–158.050.7–144.714.9–122.655.1–117.748.0–148.166.2–151.9
K6122–20 7917841–18 45524 629–82 3397662–17 4506482–16 4096917–14 034
Mg2596–41432559–42083225–12 1372855–40292805–40463242–4773
Mn19.2–40.911.9–28.23.5–14.917.3–34.323.9–65.713.9–38.9
P6253–12 4445955–12 2802430–13 0396914–11 2626606–10 3706738–11 616
S8048–14 3606824–12 5752614–13 1729230–14 8354378–12 9396854–11 846
Zn30.2–86.146.5–111.922.4–81.533.3–64.128.4–80.338.0–71.7

Although mineral trait ranges for populations were sometimes shifted upward in some experiments, the distribution of trait values among individual lines fell into the expected normal distribution pattern. For example, the frequency distribution of Fe concentration in both RIL populations was quite similar between experiments (Fig. 1a,c), while most lines had a lower Zn concentration in 2003 in both populations than in other experiments (Fig. 1b,d). Frequency distributions for the remaining seed minerals are shown in Supporting Information Figs S1 and S2. Silique hull mineral concentrations also had normal frequency distributions, although a few lines had substantially higher Cu and Fe concentrations than the majority of lines (Supporting Information Fig. S3). Correlations of seed mineral concentrations of individual RILs from each growth cycle are presented in Supporting Information Figs S4–S6.

Figure 1.

Histograms of seed iron (Fe) and zinc (Zn) concentrations in Arabidopsis thaliana Columbia (Col) × Landsberg erecta (Ler) (CL) and Cape Verde Islands (Cvi) × Ler (CVL) recombinant inbred line (RIL) populations. (a) Frequency distribution of seed Fe concentration in the CL population. (b) Frequency distribution of seed Zn concentration in the CL population. (c) Frequency distribution of seed Fe concentration in the CVL population. (d) Frequency distribution of seed Zn concentration in the CVL population.

Within each population and experiment, some seed mineral concentrations were consistently highly correlated, for example, the minerals Fe and Zn (Fig. 2). Cu and Zn were also consistently highly correlated, as were Mg and P (Supporting Information Table S2), but most minerals had weak positive correlations with other minerals, with a few exhibiting weak negative correlations. Correlations of mineral concentrations in silique hulls were weaker, with Mn and Zn exhibiting the strongest correlation.

Figure 2.

Correlations of seed iron (Fe) and zinc (Zn) concentrations in Arabidopsis thaliana Columbia (Col) × Ler (CL) and Cvi × Ler (CVL) recombinant inbred line (RIL) populations. (a) CL population correlations. (b) CVL population correlations.

QTL mapping results for the CL population are presented in Table 2 and Supporting Information Fig. S7. Several QTLs were mapped in all three experiments. These include QTLs on chromosome 2 for Ca, Cu, and P, and a QTL for seed S on chromosome 4. All of these QTLs had additive effects large enough, explaining at least 15% of the total variation, to potentially allow fine mapping of the quantitative gene in the QTL region. Several other QTLs for Ca, Cu, K, Mg, and Mn were mapped in two of the three experiments. The majority of QTLs mapped were found in only one experiment. For the seed traits of the CVL population (Table 3 and Supporting Information Fig. S8), we compared the results of our two experiments with each other and with a previous publication on this population (Vreugdenhil et al., 2004). Several QTLs were mapped in all three experiments, including QTLs for Zn on chromosomes 1 and 2, one for K on chromosome 2, one for Mn on chromosome 1, and a P QTL on chromosome 3. The P QTL was quite strong, explaining 43 and 54% of trait variability in our two experiments. Twelve QTLs were mapped in two out of the three experiments; on chromosome 1, QTLs for Cu, S, and P; on chromosome 2, QTLs for Mg and P; on chromosome 3, QTLs for Mg, Mn, and S; and on chromosome 5, QTLs for Ca, S, and Zn. It should be noted that seed S was not studied by Vreugdenhil et al. (2004).

Table 2.  Significant quantitative trait loci (QTLs) for seed mineral concentration traits in the Arabidopsis thaliana Columbia (Col) × Ler (CL) recombinant inbred line (RIL) population
QTL no.Chrom.Trait200320052007
CI (peak)a% expl.bAdd.cCI (peak)a% expl.bAdd.cCI (peak)a% expl.bAdd.c
  • a

    Confidence interval (CI) of QTL (P < 0.05) in cM. The position of the peak logarithm of odds (LOD) score is in parentheses or underlined. Some QTLs contain multiple peaks.

  • b

    Percentage of variability explained by this trait (r2 at peak position).

  • c

    Additive effect. A negative value indicates that the Col allele decreases the trait value.Chrom., chromosome.

CL12Ca      0–13 (11)   9
CL22Ca40–52 (50.6)1746.5–56.8 (50.6)3044.5–56.8 (50.6)  25
CL32Ca60 960126010
CL44Ca44.5–64.2 (54.6)21+   44.5–64.2 (54.6)  25+
CL54Ca   63.7–74.2 (67.9) 9+   
CL65Ca141 9+      
CL72Cu38.2–54.6 (48.8)2444.4–60 (50.6)2240.4–60 (50.6)  27
CL82Cu   69.9–71.412+   
CL93Cu   8.4–11 9+   
CL103Cu   15.7–18.6 8+   
CL113Cu      55.3–66.6 (55.9)   9+
CL125Cu   90.6–111.7 (104.6)26100.6–120.4 (111.7)  15
CL131Fe   0–110   
CL143Fe   8.4–22.6 (11)18+   
CL154Fe0–8.3 (5.1)14      
CL165Fe      25.7–35.9 (29.6)  11
CL175Fe67.2–70.4 (68.4) 9+      
CL185Fe   90–9110+   
CL195Fe   95.1–107.4 (98.9)14   
CL201K   22.6 8   
CL212K   33.1–35.1 9+   
CL222K   38.2–44.5 (42.4)11+   
CL232K40.4–60 (50.6)53      
CL242K65.2–86.5 (69.9, 75.8, 84.5)25+65.2–73.8 (71.4)10+   
    27+      
    18+   76.1–94 (86.1)  24+
CL255K80.8–88.2 (84.4)10      
CL261Mg   36.6–43.1 (39.6)17   
CL271Mg   69.3–72.9 (70.9)10+   
CL282Mg40.4–60 (50.6)22      
CL293Mg      10.4–17.7 (15.5)  12+
CL303Mg26.6–41.2 (36.3)10+   20.6–30  13+
CL315Mg117.5–134.4 (127.1)11      
CL321Mn16.7–1710      
CL331Mn55.710+      
CL342Mn   40.4–58 (50.6)1540.4–60 (50.6)  17
CL351P69.3–78.7 (70.9) 8+      
CL361P   78.7–91.2 (87.2)11+   
CL371P      118.3–120.3  11+
CL382P40.4–60 (50.6)3746.5–56.8 (50.6)1948.8–52.6 (50.6)  15
CL391S   0–4 (1.9)11   
CL402S46.5–52.6 (50.6)10      
CL413S      20.6–30 (22.6)   9
CL423S34.9–36.3 7      
CL433S40–41.210      
CL444S3.1–23.4 (15.5)25+8.3–30.1 (17.5)22+13.5–30.1 (15.5)  31+
CL455S14.3–23.7 (16.5) 9+      
CL461Zn69.3–70.9 9+      
CL472Zn44.5–52.6 (50.6)16   44.5–60 (50.6)  33
CL485Zn      25.7–40.8 (35.9)  11
CL495Zn      44.2–46.6 (44.6)   7
CL505Zn      100.6   6
Table 3.  Significant quantitative trait loci (QTLs) for seed mineral concentration traits in the Arabidopsis thaliana Cvi × Ler (CVL) recombinant inbred line (RIL) population
QTL no.Chrom.Trait20032006
CI (peak)a% expl.bAdd.cCI (peak)a% expl.bAdd.cVr.?d
  • a

    Confidence interval (CI) of QTL (P < 0.05) in cM. The position of the peak logarithm of odds (LOD) score is in parentheses or underlined. Some QTLs contain multiple peaks.

  • b

    Percentage of variability explained by this trait (r2 at peak position).

  • c

    Additive effect. A negative value indicates that the Col allele decreases the trait value.

  • d

    ‘Yes’ indicates that a similar QTL was identified by Vreugdenhil et al. (2004) (‘Vr.?’).Chrom., chromosome.

CVL11Ca   16–21 (18)10 
CVL21Ca45–50 (47)6+    
CVL32Ca42–45 (43)7+    
CVL42Ca   61–69 (67)12+ 
CVL53Ca0–17 (9)193–9 (8) 6Yes
CVL65Ca0–10 (2)10+0 7+ 
CVL71Cu83–103 (96)1690–96 8 
CVL82Cu37–58 (49)24+    
CVL92Cu   67–69 8 
CVL104Cu   47–55 (53) 7 
CVL114Cu   63 8 
CVL125Cu   37–40 7+ 
CVL131Fe98–106 (103)6    
CVL142Fe35–58 (49)18+    
CVL153Fe9–29 (19)16    
CVL165Fe   35–46 (40) 6+ 
CVL171K75–82 (79)5    
CVL182K45–5815+    
CVL192K65–691965–6910Yes
CVL205K   0–1210 
CVL211Mg16–25 (23)6    
CVL222Mg35–58 (51)27+42–5117+ 
CVL233Mg0–550–1114 
CVL245Mg30–45 (32, 40)6+    
    8+    
CVL255Mg   68–71 (69) 5+ 
CVL261Mn104–124 (115)17122–124 9Yes
CVL273Mn0–11 (3)11   Yes
CVL285Mn5–65+    
CVL295Mn90–107 (103)11+    
CVL301P0–11 (4)44–8 (6) 4 
CVL311P16–204    
CVL321P512    
CVL332P35–58 (49)13+37–58 (49) 7+ 
CVL343P0–11430–11 (3)54Yes
CVL353P172    
CVL364P46–50 (48)3    
CVL371S14–32 (23)7+8–20 (11)10+ 
CVL381S110–122 (115)12.5    
CVL393S0–1160–11 9 
CVL403S72–796+    
CVL414S   40–53 (47) 8 
CVL425S77–96 (89)1287–89 8 
CVL435S   92–107 (99)12 
CVL441Zn0–20 (8)130–8 (4)12Yes
CVL452Zn35–58 (42, 49)10+39–42 (40) 5+Yes
    15+47–54 (49) 7+ 
CVL463Zn0–1110    
CVL475Zn20–39 (30, 37)11+    
    10+    
CVL485Zn   45–48 (46) 5+Yes
CVL495Zn   80–105 (92)14+ 

We also mapped QTLs for silique hull mineral concentration and seed:hull concentration ratio (Table 4 and Supporting Information Fig. S9) for CVL in 2006. One QTL, for P on chromosome 3, was mapped in all experiments (seeds, hulls, and seed:hull concentration ratio), while another QTL for Zn on chromosome 2 was mapped in all three seed mineral experiments and in silique hulls. QTLs that were found in both our experiments and mapped previously include Mn and Zn on chromosome 1, K on chromosome 2, and Ca on chromosome 3. Several other QTLs were mapped for both seed and hull traits.

Table 4.  Significant quantitative trait loci (QTLs) for silique hull mineral concentrations and seed:hull mineral concentration ratio traits in the Arabidopsis thaliana Cvi × Ler (CVL) recombinant inbred line (RIL) population grown in 2006
QTL no.Chrom.TraitHullsSeed:hull
CI (peak)a% expl.bAdd.cCI (peak)a% expl.bAdd.cSeed QTL?d
  • a

    Confidence interval (CI) of QTL (P < 0.05) in cM. The position of the peak logarithm of odds (LOD) score is in parentheses or underlined. Some QTLs contain multiple peaks.

  • b

    Percentage of variability explained by this trait (r2 at peak position).

  • c

    Additive effect. A negative value indicates that the Col allele decreases the trait value.

  • d

    Overlapping QTLs identified in Table 3. Chrom., chromosome; Vr., Vreugdenhil et al. (2004).

CVL501Cu92–98 (93) 8   2003, 2006
CVL512Cu39–58 (49)28+   2003
CVL522Cu67–6910   2006
CVL533Cu70–78 7    
CVL545Cu32–3412+   2006
CVL552K32–51 (47)12+34–54 (40)202003
CVL565K64–661464 9+ 
CVL575Mg   30–3213+2003
CVL582P   67–6912+ 
CVL593P0–11 (2)430–1118+2003, 2006, Vr.
CVL601S   32–47 (38)19+ 
CVL613S0 9   2003, 2006
CVL623S7014   2003
CVL632Zn32–43 (34)16+   2003, 2006, Vr.
CVL645Zn55–61 (59)16+    

By anchoring the genetic markers used for map construction with known positions on the physical map, we were able to estimate which known or predicted genes fall within the QTL confidence intervals. Genes with known or predicted functions that could influence seed mineral traits were designated as candidate genes (Table 5). Four single-gene mutants with T-DNA insertions in candidate genes were analyzed for seed mineral concentrations. There were no significant differences in hma5, ysl8, or zip5, but fer2 had approx. 10% lower seed Fe concentration.

Table 5.  Candidate genes for quantitative trait loci (QTLs), categorized by predicted function
TransportersChelators/storage
QTL no.Mineral traitGeneLocus IDQTL no.Mineral traitGeneLocus ID
  1. AKT, Arabidopsis potassium transporter; APR, APS reductase; APS, ATP sulfurylase; ATX, anti-oxidant; CAX, cation exchanger; CCH, copper chaperone; COPT, copper transporter; CSD, Cu, Zn superoxide dismutase; CXIP, CAX-interacting protein; FER, ferritin; FRD, ferric reductase defective; FRO, ferric reductase oxidase; HMA, heavy metal P1b-ATPase; KT, potassium transporter; KUP, potassium uptake permease; MGT, magnesium transporter; MRS, mitochondrial RNA splicing; MT, metallothionine; MTP, metal tolerance protein; NFU, nitrogen-fixation-specific-like; Nramp, natural resistance-associated macrophage protein; PAA, P-type ATPase of Arabidopsis; PHO, phosphate overaccumulator; Pht, phosphate transporter; RAN, responsive to antagonist; Sultr, sulfate transporter; YSL, yellow stripe-like; ZIF, zinc-induced facilitator; ZIFL, ZIF-like; ZIP, zinc regulated transporter, iron regulated transporter-like protein.

CL23, CVL18, CVL55KAKT1At2g26650CVL7, CVL50CuATX1At1g66240
CVL6CaCAX4At5g01490CVL53CuCCHAt3g56240
CL12CuCOPT1At5g59030CL8, CVL9, CVL52CuCCH-likeAt2g37390
CL7, CVL8, CVL51CuCOPT1 familyAt2g26975CL12CuCCH-likeAt5g63530
CL11CuCOPT2At3g46900CL12CuCCH-likeAt5g66110
CL12CuCOPT3At5g59040CL14FeFER2At3g11050
CL8, CVL9, CVL52CuCOPT4At2g37925CL7, CVL8, CVL51CuCu-binding familyAt2g28660
CVL12, CVL54CuCOPT5At5g20650CVL44ZnMT1aAt1g07600
CL17, CVL16FeFerroportin2At5g26820CVL44ZnMT1cAt1g07610
CL14FeFRD3At3g08040    
CVL45, CVL63ZnHMA4At2g19110Regulators
CVL7, CVL50CuHMA5At1g63440QTL no.Mineral traitGeneLocus ID
CL12CuHMA7/RAN1At5g44790CVL39, CVL61SAPS kinaseAt3g03900
CVL12, CVL54CuHMA8/PAA2At5g21930CL3CaCXIP4At2g28910
CL22, CVL18, CVL55KKT6At2g25600CL1CaCXIP4-likeAt2g01100
CL23, CVL18, CVL55KKUP1At2g30070CVL33PPHO2At2g33770
CL24KKUP2At2g40540CVL46ZnZn-binding TFAt3g07780
CVL20KKUP7At5g09400    
CL23, CVL19KKUP11At2g35060Metabolism
CVL24, CVL57MgMGT1-likeAt5g22830QTL no.Mineral traitGeneLocus ID
CL30MgMRS2 familyAt3g19640CL42SAPR1At4g04610
CL31MgMRS2 familyAt5g64560CL44SAPS1At3g22890
CL47ZnMTPb1At2g29410CL7, CVL8, CVL51CuCSD2At2g28190
CVL14FeNramp3At2g23150CL49, CVL47ZnCSD3At5g18100
CVL58PPht1;4At2g38940CL13FeFRO1At1g01590
CVL33PPht1;5At2g32830CL13FeFRO2At1g01580
CL37PPht1;9At1g76430CVL16FeFRO4At5g23980
CVL38SSultr1;2At1g78000CVL16FeFRO5At5g23990
CVL38SSultr2;2At1g77990CL19FeFRO6At5g49730
CL40SSultr3.4At1g23090CL19FeFRO7At5g49740
CVL60SSultr3;3At3g15990CL19FeFRO8At5g50160
CVL39, CVL61SSultr4;2At3g12520CL15FeNFU1At4g01940
CVL16, CVL47FeYSL2At5g24380    
CL12CuYSL3At5g53550    
CVL7, CVL50Cu, ZnYSL7At1g65730    
CL46ZnYSL8At1g48370    
CL48, CVL47ZnZIF1At5g13740    
CL48, CVL47ZnZIFL1At5g13750    
CVL46ZnZIP familyAt3g08650    
CL14ZnZIP familyAt3g08650    
CVL46ZnZIP1At3g12750    
CVL49ZnZIP2At5g59520    
CVL45ZnZIP3At2g32270    
CVL44ZnZIP4At1g10970    
CVL44ZnZIP5At1g05300    
CL47, CVL45ZnZIP6At2g30080    
CVL49ZnZIP12At5g62160    

Several of the QTLs mapped to chromosome 2 coincided with the Erecta (ER) locus (50.6 cM in CL; 49 cM in CVL). To test whether the ER locus was a quantitative trait gene, we grew four er mutant lines (er-114, er-116, and er-117 in the Col-0 background, and er-123 in the Wassilewskija-2 (Ws-2) background) and the appropriate wild-type lines under different conditions and quantified seed mineral concentrations (Table 6). Under high-light, high-nutrient conditions, Ca, Cu, Mg, P, and S were unchanged or significantly lower in the er mutants in the Col background, while Cu, Mg, P, and Zn were significantly higher in er-123. K was higher in two of the er lines, while Mn was lower in three of the four er lines. By contrast, when grown in low-light conditions (as the RIL populations were grown), with either low or high nutrients, in the majority of lines and plant growth cycles, Ca, Cu, K, and Zn were significantly higher in the er mutant lines relative to their wild-type parent. Seed Fe concentration was higher in six instances and lower in one, while Mg was higher in five of 12 instances. Seed Mn, P, and S were not consistently different from wild-type parent lines.

Table 6.  Relative mineral concentrations of wild-type and erecta (er) mutant seeds of Arabidopsis thaliana grown under high light, high nutrients (HLHN), low light, low nutrients (LLLN), or low light, high nutrients (LLHN; two experiments shown)
TreatmentLineCaCuFeKMgMnPSZn
  • *

    Significantly different than wild type (P < 0.05).

  • a

    Normalized to Columbia (Col).

  • b

    Normalized to Wassilewskija-2 (Ws-2).

HLHNCol100100100100100100100100100
Lera7211913895961027585119
er-114a87*80*8911091*79*87*90*98
er-116a86*84*100111*94*939580*99
er-117a9174*911029588*949197
Ws-2100100100100100100100100100
er-123b97176*106170*115*92*146*92134*
LLLNCol100100100100100100100100100
Lera10320113481901328399138
er-114a143*210*1391009986115124131
er-116a119*114*81*1089992*11395106*
er-117a140*155*115130*10078*115111113*
Ws-2100100100100100100100100100
er-123b113*155*72147*121*71*139*112114*
LLHN (Expt 1)Col100100100100100100100100100
Lera8414613866931596999155
er-114a121*144*123*91*104125*108113140*
er-116a102102959198125*9499117*
er-117a111*124*97106*103100104108124*
Ws-2100100100100100100100100100
er-123b117*146*109*11610373*115*115*107*
LLHN (Expt 2)Col100100100100100100100100100
Lera85165186929914092107184
er-114a89*105172*106*118*97100122*158*
er-116a91109147*136*121*98110*116*166*
er-117a117*93143*110*107*82*97111*144*
Ws-2100100100100100100100100100
er-123b110124*111*138*120*85*118*102121*

Discussion

QTL mapping is a powerful genetic technique that can be used to identify markers or genes associated with a quantitative trait, such as seed mineral concentration. A gene within QTL confidence intervals is unlikely to be the sole determinant of the trait, as there are often multiple genes that affect the trait of interest through individual additive effects or interactions (Flint & Mott, 2001; Tonsor et al., 2005). Movement of minerals from the soil into and through the plant to seeds requires an unknown number of membrane transport processes through multiple tissue types. As such, QTLs for increased seed mineral concentration may indicate genes that are important at any limiting step, such as uptake at the root surface, translocation (xylem loading/unloading or phloem loading/unloading), storage capacity (in source tissue or seed), and remobilization processes, or genes that encode regulatory proteins. Genes that encode proteins vital to these processes are very likely to also be important for the orthologous processes in crop plants. Thus, discovery of genes that increase A. thaliana seed mineral concentrations could help to determine the most effective genes to target for biofortification of crops.

Robustness of traits

One of the main objectives of this research was to determine whether A. thaliana is a reliable model plant for QTL mapping of seed mineral concentrations for discovery of target genes for biofortification. To meet this criterion, significant QTLs should be robust enough to be mapped in multiple growth cycles and, ideally, in similar but differing environments, such as growth facilities of different laboratories or different seasons within a given growth facility. Thus, we thought it important to compare our results with those from another research group (Vreugdenhil et al., 2004). On each occasion on which the RIL populations were grown, there was a wide range of values (2–3-fold) observed for each trait (Table 1), and a normal distribution of the traits within these ranges (Fig. 1, Supporting Information Figs S1–S3), indicating that adequate trait diversity accompanies the natural genetic variation. A wide range of mineral concentration values has been observed in edible portions among accessions of a number of crop plants (reviewed by White & Broadley, 2005).

Between occasions of RIL population growth, most minerals that were correlated on one occasion were similarly correlated on other occasions (Fig. 2, Supporting Information Table S1), further indicating that A. thaliana seed mineral concentrations are reliable traits. The most important test, however, is repetition of the QTL mapped for each trait. If seed mineral QTLs are repeated between different research groups and different experiments within a group, that is a good indication that the quantitative gene effect may be sufficiently robust to pursue by fine mapping or a candidate gene approach. Several seed mineral QTLs met this criteria; four QTLs in the CL population and five QTLs in the CVL population were identified in three separate experiments, while another 15 QTLs were identified in two of three experiments. It would be useful to grow the RIL populations in differing environments, such as with low nutrient supply or higher light intensity, to see if the seed mineral concentration QTLs are maintained over more varied environments.

Linking genetics to physiology

While the bulk of mineral nutrients in A. thaliana seeds come into the plant during seed fill (Waters & Grusak, 2008), differences in translocation through the plant or remobilization from source tissues may account for the small differences between RILs. Because some minerals have been shown to be mobilized from A. thaliana silique walls (hulls) to seeds during fruit maturation, and the mineral concentration of mature hulls is reflective of the mineral content of hulls (Waters & Grusak, 2008), the seed:hull concentration ratio can be considered a quantitative trait that reflects fruit mineral partitioning to seeds. Genes important for remobilization or translocation of minerals through hull tissue may be specific to this tissue, but could also be important for these processes in other seed mineral source tissues, such as leaves. Seven hull QTLs were also mapped from seed mineral concentrations, and four of the six QTLs for seed:hull concentration ratios were also found in hull or seed mineral concentration QTLs, indicating that differences in mineral partitioning between lines can also be used as a quantitative trait.

The strongest seed mineral QTL identified in this study, for P at the top of chromosome 3 in the CVL population, was previously identified (Bentsink et al., 2003; Vreugdenhil et al., 2004). In our study, this QTL was found not only for seed P in both CVL experiments, but also for P in silique hulls and the seed:hull ratio. This indicates that the quantitative trait gene is robust enough to affect and be detected for multiple traits, highlighting the importance of comparing multiple experiments, and the potential for performing QTL analysis on certain source tissues that might affect the ultimate trait of interest. Bentsink et al. (2003) have initiated fine mapping of this region of chromosome 3.

Although a number of QTLs were identified in multiple experiments, most of the QTLs mapped were found in only one experiment. It is unclear why this was the case, but a number of explanations can be offered. First, most of the single-occurrence QTLs had small effects on seed mineral traits and significant but low LR scores. Thus, these loci may in fact have a minor effect on seed mineral concentration, but, as a result of slightly varying environmental conditions between experiments, may not have been as important in each experiment. If this is the case, it would be impractical to fine map these weak QTLs. Secondly, these weak QTLs may reflect the quantitative nature of seed mineral traits, where dozens of gene products may have small additive effects on the processes involved in mineral movement to and accumulation in seeds, and in a given replicate experiment these weak QTLs were not always statistically significant. Thirdly, another reason why individual QTLs may be weak or not mapped in replicate experiments could be that many genes involved in plant mineral nutrition are members of gene families with individual members overlapping in function and expression patterns. It is not uncommon for single-gene knockouts to have no detectable mutant phenotype (as we observed with hma5, ysl8, and zip5), whereas disruption of two gene family members results in a severe mutant phenotype (Hussain et al., 2004; Waters et al., 2006).

Candidate genes

Genes related to mineral transport or homeostasis (or annotated as such) that fell within QTL confidence intervals are listed in Table 5. While we realize that the majority of mineral-related genes in the QTL region will not be the quantitative trait gene detected within that confidence interval, it is logical to consider the potential of these genes to have an effect on seed mineral concentration. Most of these genes were categorized as transporters, while others fell into the categories of chelators and storage molecules, regulators of transporters or homeostasis, or genes involved in metabolism. Many of the candidate genes are members of gene families. The ultimate confirmation of genes responsible for each of the seed traits will be challenging, but may include evidence such as that obtained from positional cloning by construction of near isogenic lines (NILs), association mapping of specific polymorphisms, transgenic or deficiency complementation, gene expression analysis, or mutational analysis (Koornneef et al., 2004; Weigel & Nordborg, 2005).

On chromosome 2, several of the QTLs mapped to a region that coincided with the ER locus, with additive effects indicating that the er mutant lines had higher concentrations of Ca, Cu, K, Mg, Mn, P, S, and Zn in the CL population, and Cu, Fe Mg, P, and Zn in the CVL population. Recently, the ER locus was shown to affect water use efficiency (Masle et al., 2005), which may in turn differentially affect mineral translocation in the xylem to aerial parts of the plant. We tested the idea that er might affect seed mineral concentration using single er mutants, and based on the results (Table 6), it appears that Ca, Cu, and Zn are consistently increased in most of the er mutant lines in low-light conditions, although penetrance was not 100%. Results were mixed for Fe, K, and Mg, while no consistent increase in seed mineral concentration was observed in er mutants for Mn, P, and S. These results suggest that for Mn, P and S QTLs in this region of chromosome 2, some gene other than er is responsible for the seed mineral variation. However, it appears likely that the er locus itself contributes to increasing seed mineral Ca, Cu and Zn concentrations under low-light conditions, but not under high-light conditions where increased stomatal conductance to support higher photosynthetic rates would result in increased transpiration both in er and in ER plants, thereby minimizing water flux differences between mutant and wild-type lines. Conversely, when environmental conditions lead to differential rates of transpiration in ER versus er lines and less overall transpiration in ER lines (e.g. lower light), reduced xylem flow, especially to low-transpiring organs such as siliques (Jensen et al., 1998), may lead to lowered delivery of minerals to these structures. Because silique hulls are an important source tissue for seed minerals (Waters & Grusak, 2008), a reduced delivery of minerals to the silique hulls (ER relative to er lines) could explain the reduced seed concentrations of certain minerals. However, as the ER gene encodes a serine/threonine protein kinase and the er mutation has pleiotropic effects (Lease et al., 2001), reasons other than water use efficiency, including interactions with other genes, could affect seed mineral concentrations. In the RIL populations, the effects of the QTLs containing er could be explained by other nearby genes; there are Cu- and Zn-related genes in this region that could be involved in metal translocation within the plant.

Candidate genes: Cu, Fe, and Zn

Several members of the copper transporter (COPT) family of Cu transporters are found in seed and hull Cu-concentration QTL regions (Table 5). COPT family genes are expressed in roots, leaves, stems, and flower tissues (Sancenón et al., 2003). COPT1 is the most well-characterized family member. COPT1 antisense lines had 40–60% lower Cu concentrations in rosette leaves and defective pollen development (Sancenón et al., 2004).

Four members of the heavy metal P1b-ATPase (HMA) family are located within seed and hull mineral QTL confidence intervals. HMA4 was in Zn seed and hull concentration QTL confidence intervals. Rosette Zn concentration is low in hma4 mutants (Baxter et al., 2007), and was important for translocation of Zn from roots to shoots (Hussain et al., 2004; Verret et al., 2004). HMA5, HMA7 and HMA8 are monovalent Cu transporters. HMA5 is found in two Cu QTLs, one for seed concentration and one for hull concentration. Mutant hma5 plants grown on high Cu had increased Cu concentration mainly in the roots (Andrés-Colás et al., 2006). HMA7/response to antagonist 1 (RAN1) is necessary for loading Cu into intracellular compartments of the secretory pathway (Hirayama et al., 1999). HMA8/P-type ATPase of Arabidopsis 2 (PAA2) is required for Cu transport into thylakoids and delivery to plastocyanin (Abdel-Ghany et al., 2005); hma8/paa2 mutants had increased expression of stroma-localized Cu, Zn superoxide dismutase CSD2 (Abdel-Ghany et al., 2005). CSD2 and CSD3 are both localized to QTLs mapped for seed Zn concentration and for seed and hull Cu concentration (Table 5).

Within cells, Cu is trafficked bound to chaperone proteins, such as copper chaperone (CCH) and anti-oxidant 1 (ATX1). CCH, ATX1, and three genes annotated as CCH-like are all within Cu QTL regions. ATX1 interacts with HMA5 (Andrés-Colás et al., 2006) and with HMA7/RAN1 (Puig et al., 2007). CCH and HMA7/RAN1 are both up-regulated during leaf senescence (Himelblau & Amasino, 2000; Mira et al., 2001) and could be important for removing Cu from leaves to be translocated to seeds. Up to 29% of total seed Cu content was derived from vegetative tissue remobilization in A. thaliana (Waters & Grusak, 2008). Several genes related to Cu homeostasis were differentially regulated in A. thaliana lines differing in Cu tolerance and Cu accumulation, including Col and Ler (Schiavon et al., 2007), lending credence to the idea that differences in the expression or sequence of these genes could alter quantitative traits such as seed mineral concentration.

YSL family genes encode metal-nicotianamine transporters (DiDonato et al., 2004; Koike et al., 2004; Schaaf et al., 2005). Several members of the YSL family are located within seed mineral QTL regions for Fe and Zn. Mutations in YSL1 and YSL3 resulted in decreased seed concentrations of Fe, Zn, and Cu (Le Jean et al., 2005; Waters et al., 2006; Waters & Grusak, 2008), while ysl1ysl3 mutants overaccumulated Cu and Zn in leaves (Waters & Grusak, 2008).

An uncharacterized cation diffusion facilitator metal tolerance protein (MTP)-family gene, MTPB1, is located within a seed Zn-concentration QTL. Arabidopsis thaliana mutants of the vacuole-localized mtp1 had lower accumulation of Zn in stem and leaf tissue, but normal Zn concentration in flowers and siliques (Desbrosses-Fonrouge et al., 2005), whereas MTP3 moves Zn into vacuoles in Fe-deficient plants and is important for regulating Zn partitioning between roots and shoots (Arrivault et al., 2006). Another family of Zn transporters, the ZIP proteins, are responsible for Zn uptake from the soil, and some members that are expressed in shoot tissues (Grotz et al., 1998) may be involved in translocation of Zn throughout the plant (Colangelo & Guerinot, 2006). Several ZIP family members are localized in QTL regions for Fe and Zn (Table 5).

The Fe-citrate transporter ferric reductase defective 3 (FRD3) (Durrett et al., 2007) controls root-to-shoot Fe translocation (Green & Rogers, 2004), and localizes to a seed Fe QTL, as does an annotated gene similar to zebrafish ferroportin (Donovan et al., 2000), which is an Fe exporter in the gut. An analogous function in plants may be to move Fe into the apoplasm for loading into the xylem pathway. Before internal translocation, Fe uptake from the soil depends on reduction by ferric reductase oxidase 2 (FRO2; Robinson et al., 1999). The FRO2 gene and several other FRO family members that are expressed in leaf vascular and flower tissues (Wu et al., 2005; Mukherjee et al., 2006) are found in three seed Fe QTL regions.

Mineral hyperaccumulating plants can have mineral concentrations several orders of magnitude greater than crop plants and often have higher expression of mineral transporter genes than their nonhyperaccumulating relatives (Broadley et al., 2007). The widely studied Zn hyperaccumulator Thlaspi caerulescens can accumulate high concentrations of Zn not only in vegetative tissues, but also in seeds (Ernst, 1996). Identification of genes that contribute to this trait may offer clues as to which genes would make effective biofortification targets in crop plants. Several microarray studies identified genes more highly expressed in the Zn hyperaccumulator Arabidopsis halleri than in A. thaliana, and revealed several genes identical to or in the same families as our candidate genes, including genes of the HMA family, the MTP family, natural resistance-associated macrophage protein 3 (Nramp3), YSL6, FRD3, and several ZIP family members (Becher et al., 2004; Weber et al., 2004; Filatov et al., 2006; Talke et al., 2006; Broadley et al., 2007). Similarly, genes more highly expressed in T. caerulescens than in A. thaliana included Nramp3, HMA family genes, ZIP family genes, YSL7, MTP1, FRD3, and FRO5 (van de Mortel et al., 2006).

Candidate genes: P and S

The phosphate transporter 1 (Pht1) family encodes phosphate transporter proteins, and several Pht1 genes are localized in QTL regions. Pht1;4 is one of these, and has been demonstrated to be important for uptake of inorganic phosphate (Pi) into the root (Muchhal et al., 1996; Misson et al., 2004; Shin et al., 2004), whereas work using a Pht1;5 promoter-reporter construct indicated that expression was primarily in leaves (Mudge et al., 2002). The finding that phosphate overaccumulator 2 (pho2) mutants had 3-fold higher P concentrations in shoots, and higher P in stems, siliques, and seeds suggests that PHO2 is a regulator of P accumulation (Delhaize & Randall, 1995).

Five members of the sulfate transporter (Sultr) family of sulfate transporters are localized in QTL regions for seed and hull S concentration, three in plasma membrane-localized subgroups 1–3, and one in vacuole- localized subgroup 4 (Hawkesford & De Kok, 2006). Of these candidate genes, Sultr1;2 has been indicated to be important for root sulfate uptake (Shibagaki et al., 2002), while Sultr2;2 is a low-affinity transporter expressed near vascular tissue in roots and leaves (Takahashi et al., 2000). Other family members have been implicated in redistribution of S from source tissues (Yoshimoto et al., 2003), and loss of Sultr2;1 function resulted in decreased seed S concentration (Awazuhara et al., 2005). Other candidate genes are involved in S metabolism (Hawkesford & De Kok, 2006), including ATP sulfurylase 1 (APS1), APS kinase, and APS reductase 1 (APR1).

Conclusions

Candidate genes were identified for many of the QTLs based on known or predicted gene function, but other QTLs have no obvious candidates, suggesting that new genes affecting mineral uptake or translocation are yet to be identified. Many QTLs were identified for seed mineral concentration traits in two A. thaliana RIL populations, with several of these QTLs found in multiple experiments, and thus are sufficiently robust to allow future fine mapping of the quantitative genes. This suggests that A. thaliana is a reliable model for discovery of genes that could be targeted for use in biofortification efforts to increase mineral concentrations in seeds of crop plants. The fact that the majority of QTLs were identified in only one experiment suggests that seed mineral concentration is a truly quantitative trait; many genes have small additive effects on the final trait value. This indicates that, to develop biofortified crops, several genes may need to be targeted simultaneously, highlighting the importance of developing our understanding of the molecular and physiological steps required to accumulate minerals in seeds.

Acknowledgements

This work was funded in part by funds from USDA-ARS under Agreement No. 58-6250-6-003 and from the Harvest Plus Project under Agreement No. 58-6250-4-F029 to MAG.

Ancillary