Physiological and molecular analysis of polyethylene glycol-induced reduction of aluminium accumulation in the root tips of common bean (Phaseolus vulgaris)

Authors


Author for correspondence:
Walter Johannes Horst
Tel: +49 511 7622627
Email: horst@pflern.uni-hannover.de

Summary

  • Aluminium (Al) toxicity and drought are two major stress factors limiting common bean (Phaseolus vulgaris) production on tropical acid soils. Polyethylene glycol (PEG) treatment reduces Al uptake and Al toxicity.
  • The effect of PEG 6000-induced osmotic stress on the expression of genes was studied using SuperSAGE combined with next-generation sequencing and quantitative reverse transcription-polymerase chain reaction (qRT-PCR) for selected genes.
  • Less Al stress in PEG-treated roots was confirmed by decreased Al-induced up-regulation of MATE and ACCO genes. The withdrawal of PEG from the Al treatment solution restored the Al accumulation and reversed the expression of MATE and ACCO genes to the level of the treatment with Al alone. Using SuperSAGE, we identified 611 up- and 728 down-regulated genes in PEG-treated root tips, and the results were confirmed by qRT-PCR using 46 differentially expressed genes. Among the 12 genes studied in more detail, XTHa and BEG (down-regulated by PEG) and HRGP, bZIP, MYB and P5CS (up-regulated by PEG) recovered completely within 2 h after removal of PEG stress.
  • The results suggest that genes related to cell wall assembly and modification, such as XTHs, BEG and HRGP, play important roles in the PEG-induced decrease in cell wall porosity, leading to reduced Al accumulation in root tips.

Introduction

Common bean (Phaseolus vulgaris) is the major food legume for human nutrition in the world, and a major source of calories and protein, particularly in many Latin American and African countries, where middle- and low-income families are often unable to produce, or afford, sufficient animal protein (Graham, 1978; Rao, 2001). However, in these tropical countries, the production of common bean is often limited by the adverse acidic soil conditions, particularly with aluminium (Al), proton and manganese toxicity, together with nutrient deficiencies and seasonal dry spells (Graham & Ranalli, 1997; Thung & Rao, 1999; Yang et al., 2010).

Common bean is generally poorly adapted to acidic soil environments and is also a drought-sensitive crop (Rao, 2001; Beebe et al., 2008). The crop yield on acidic soils is mainly limited by Al toxicity. Aluminium resistance in common bean is attributed to the release of citrate by the root apex (Rangel et al., 2010). Generally, in citrate-releasing plant species, the multidrug and toxin extrusion (MATE) family protein, as an Al-activated citrate transporter, has been suggested to be responsible for Al resistance. For example, in sorghum (Sorghum bicolor), SbMATE is expressed only in the root tips of the Al-resistant genotype in an Al-inducible way (Magalhaes et al., 2007). Similarly, in barley (Hordeum vulgare), HvMATE expression in the root apices correlates with Al-activated citrate exudation and Al resistance in a set of barley cultivars (Furukawa et al., 2007). However, in contrast with these plant species, in common bean, the MATE gene is highly expressed by Al within 4 h of treatment in both Al-resistant (Quimbaya) and Al-sensitive (VAX 1) genotypes. The expression of MATE is a prerequisite for citrate exudation, but the build-up of Al resistance within 24 h in Quimbaya is exclusively dependent on the capacity to sustain the synthesis of citrate for the maintenance of the cytosolic citrate pool that enables continuous exudation (Eticha et al., 2010; Rangel et al., 2010). In addition, the genotype-independent initial Al-induced inhibition of root elongation and subsequent recovery in the Al-resistant genotype are closely correlated with the expression of the 1-aminocyclopropane-1-carboxylic acid oxidase (ACCO) gene (Eticha et al., 2010). It has been speculated that the Al-induced inhibition of root growth is a result of enhanced gene expression and enzyme activity of ACCO, leading to increased ethylene production in Lotus japonicus and Medicago truncatula (Sun et al., 2007). Thus, it appears that the MATE gene behaves as an Al sensor in common bean, independent of the Al resistance of the genotype, and the ACCO gene as an indicator of Al-induced inhibition of root elongation.

The combined effect of Al toxicity and drought stress on root growth, with special emphasis on the Al–drought interaction in the root apex of common bean, has been well studied. Using polyethylene glycol (PEG) to simulate osmotic stress (OS) or drought stress, we found that OS enhances Al resistance by inhibiting Al accumulation in the root apices of the Al-sensitive genotype VAX 1. This alleviation of Al toxicity was related to an alteration in cell wall (CW) porosity resulting from PEG 6000-induced dehydration of the root apoplast. A biochemical and molecular regulation of the OS-induced change in CW porosity has been proposed (Yang et al., 2010). The plant CW is a composite structure consisting of a cellulose–hemicellulose framework embedded within a pectic polysaccharides and proteins matrix (Carpita & Gibeaut, 1993). This viscoelastic semi-solid component is a decisive factor for its resistance to external stress. Water loss from the wall matrix can result in serious disruption to polymer organization. One obvious effect is that polymers usually well separated in the hydrated wall are brought into close proximity to each other, thus causing polymer adhesion or cross-linking under water stress (Moore et al., 2008). It has been reported that several proteins play key roles in the adjustment of CW structure, such as expansin, xyloglucan endotransglycosylase (XET) and glucanase (Wu & Cosgrove, 2000; Cosgrove, 2005). Therefore, the identification of the genes particularly involved in CW modification appears to be necessary for a better understanding of PEG-induced reduction of root tip Al accumulation.

Common bean is a molecularly under-researched crop; extensive microarray-based transcriptomic studies are not yet possible owing to the lack of available gene and expressed sequence tag (EST) information. Thus, less comprehensive approaches, such as suppression subtractive hybridization (SSH) libraries, need to be taken, which do not per se allow the quantification of the expression of differentially expressed genes (Molina et al., 2008). One powerful technique for gene expression analysis is serial analysis of gene expression (SAGE) developed by Velculescu et al. (1995). However, the short tag sequence of only 13–15 bp generated from SAGE is not always sufficient to unequivocally identify the gene from which the tag is derived. A single tag sequence usually corresponds to several different ESTs and genomic sequences, which require further analysis (Matsumura et al., 2003). SuperSAGE is an improved version of SAGE which overcomes the limitations of SAGE by producing 26-bp-long fragments from defined positions in cDNAs, providing sufficient sequence information to unambiguously characterize the mRNAs (Matsumura et al., 2003; Molina et al., 2008). This technique has been applied successfully in several gene expression studies (Matsumura et al., 2003; Hamada et al., 2008; Molina et al., 2008; Gilardoni et al., 2010).

The main objectives of this study were to describe the interaction of Al toxicity and OS (PEG) in the Al-sensitive common bean genotype VAX 1 at the molecular level and to identify OS-induced genes in the bean root tips using SuperSAGE with particular emphasis on genes related to CW modification, in order to better understand the OS-induced changes in CW structure and thus the reduction in Al accumulation in the root tips at the transcriptional level.

Materials and Methods

Plant materials and growing conditions

Seeds of the common bean (Phaseolus vulgaris L.) genotype VAX 1 (Al-sensitive) were germinated on filter paper sandwiched between sponges. After 3 d, uniform seedlings were transferred to a continuously aerated simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 (Rangel et al., 2007). Plants were cultured in a growth chamber under controlled environmental conditions of a 16 h : 8 h, light : dark cycle, 27°C : 25°C day : night temperature, 70% relative air humidity and a photon flux density of 230 μmol m−2 s−1 of photosynthetically active radiation at plant height. The pH of the solution was gradually lowered to 4.5 within 2 d. The plants were then transferred into the simplified nutrient solution without or with AlCl3 (25 μM), PEG 6000 (150 g l−1) and PEG 1000 (115 g l−1) (Sigma-Aldrich Chemie GmbH, Steinheim, Germany), pH 4.5. Root tips (1 cm) were harvested for Al analysis or immediately frozen in liquid nitrogen (N) in diethylpyrocarbonate (DEPC)–H2O-treated Eppendorf vials for RNA isolation. The osmotic potentials (OPs) of both PEG 6000 (150 g l−1) and PEG 1000 (115 g l−1) solutions were −0.60 MPa, measured with a cryoscopic osmometer (Osmomat 030; Gonotec GmbH, Berlin, Germany).

Measurement of root elongation rate

Two hours before the treatment was initiated, tap roots were marked 30 mm behind the root tip using a fine point permanent marker (Sharpie blue, Stanford Corporation, Oak Brook, USA) which did not affect root growth during the experimental period. Afterwards, the plants were transferred to the simplified nutrient solution without or with PEG in the absence or presence of Al. Root elongation was measured after the 24-h treatment period using a millimetre scale.

RNA isolation and construction of the SuperSAGE library

For construction of the SuperSAGE library, only PEG 6000 was used. After treating the plants with PEG 6000 for 24 h in the simplified nutrient solution, the roots were rinsed with distilled water and five to six root tips (length, 1 cm) from each plant were harvested and shock-frozen in liquid N. Root tips of 10 plants per treatment were bulked and ground to powder in liquid N. Total RNA was isolated using the NucleoSpin RNA plant kit (MACHEREY-NAGEL GmbH and Co., KG, Düren, Germany) following the manufacturer’s protocol. From total RNA, poly(A)-RNA was purified with the Oligotex mRNA mini kit (Qiagen) according to the manufacturer’s protocol.

SuperSAGE libraries were constructed by GenXPro GmbH (Frankfurt am Main, Germany) essentially as described by Matsumura et al. (2010). In order to avoid PCR bias during the amplification steps, GenXPro’s ‘PCR-bias-proof technology’ was employed to distinguish PCR copies from original tags. Sequencing was performed on an Illumina GA II machine (Illumina, Inc., San Diego, CA, USA). For each library, 26-bp-long tags were extracted from the sequences using the GXP-Tag sorter software provided by GenXPro GmbH. Sequencing artefacts were reduced according to Akmaev & Wang (2004).

Library comparisons were carried out using the DiscoverySpace 4.01 software (Canada’s Michael Smith Genome Sciences Centre, available at http://www.bcgsc.ca/discoveryspace). Statistical analysis of differentially expressed tags was conducted using the probability (P) value according to the description of Audic & Claverie (1997). The expression ratios of the 26-bp tags from control versus PEG 6000-treated roots were calculated as: ratio (R) = log2(PEG 6000/control), after normalizing to 1 million. Tags that were present zero times were replaced by 0.05 to allow calculation of the ratio.

Sequence homology alignments

Tag sequences were BLASTed (Altschul et al., 1990) against different public databases (Phaseolus_TIGR_PHVGI.release_3/PHVGI.052909; TIGR_Phaseolus_cocc_TiGR_PCGI.release_1/PCGR.052909; Glycine_MAX_TIGR_GMGI.release_14/GMGI.052909; Medicago_TIGR_MTGI.release_9/MTGI.071708; Lotus_TIGR_LJGI.release_5/LJGI.052909; Refseq_plant_June09/refseqPlantJune09.fna; All-plant-EST.fasta (plantGDB)).

Primer design for quantitative reverse transcription-polymerase chain reaction (qRT-PCR)

The ESTs from different organisms with high similarity to the sequences of candidate genes obtained from the SuperSAGE library were aligned, and the conserved regions were BLASTed against the P. vulgaris database. Finally, the ESTs of P. vulgaris were aligned and the conserved region was used for primer design. Primers were designed using Primer3 software (Rozen & Skaletsky, 2000). The primers of the β-tubulin, MATE family protein and ACCO genes were obtained from Eticha et al. (2010). The specifications of the primers of the genes studied in more detail are given in Table 1. For a complete list of all primers used, see Supporting Information Table S1. The PCR efficiencies of the primer pairs were in the range 90–110% as determined by dilution series of the cDNA template. Primer pairs with PCR efficiencies deviating from this range were discarded and new primers of the genes were designed to obtain more reliable quantification.

Table 1.   List of main genes and specific primer pairs used for quantitative gene expression analysis
Candidate genesPrimer pairs (5′→3′)aAmplicon size (bp)TC/GBb Accession No.
  1. a(+) and (−) indicate forward and reverse primers, respectively.

  2. bTC/GB, Tentative Consensus/GenBank®.

LTP (protease inhibitor/seed storage/lipid transfer protein family protein)(+) CCTCAGCAGCACAAGATGAG147CV542382
(−) TGACAGCAATCTGAGGGTTG
HRGP (hydroxyproline-rich glycoprotein)(+) CCTGTCTTGATGGTGAAGCA114CV543261
(−) TTCATTTGTTGCAGGCTGAC
SUS (sucrose synthase 2)(+) GCATGGCCTCATGAAAGAGT133TC11609
(−) GAAAGCAGGCTGAACGAAAG
XTHa (xyloglucan endotransglycosylase precursor)(+) ATATGTCATCGGAGGGTCCA151TC12227
(−) TTGGTAGGGTCGAACCAAAG
AQP (aquaporin)(+) CCACATCACCATCCTCACTG102TC14630
(−) ATTGCCAAACCTCCTGTGAC
BEG (glucan endo-1,3-β-glucosidase precursor)(+) ATGGAAGACTTGGCAACGAC122TC11172
(−) GCCTCTCAAAGCTCCAAGAA
PRP (proline-rich protein)(+) GCAAGTGTTGTGCATTGCTT160TC12228
(−) TGGAAGCCAGAAGGAACTGT
XTHb (xyloglucan endotransglucosylase/hydrolase)(+) TTTGACCAACCCATGAAGGT153CV542742
(−) GCATTCACTGAGGCTTCACA
CYP701A (cytochrome P450 monooxygenase CYP701A16)(+) GGATGCAACATGGACAAGAA136TC18728
(−) AACCTGCACACACCCTCTTC
MYB (MYB transcription factor MYB134)(+) CCGATTCCGACAAAATGAAC136TC13287
(−) GCATCAGGTGTGTTCAGCTC
P5CS (VuP5CS protein)(+) GACAGTGCTGCTGTTTTCCA128TC14708
(−) AAACCCTCTACTCCCACAGGA
bZIP (bZip transcription factor)(+) AAACTGCCACTTCCCTCCTT127TC17978
(−) TCTCCTGTGCTTCCTTTCGT
ACCO (1-aminocyclopropane-1-carboxylic acid oxidase) (Eticha et al., 2010)(+) GAAGATGGCGCAAGAAGAAG105AB002667
(−) TGGAGCAAAGGTTCAAGGAG
MATE (multidrug and toxin extrusion; citrate transporter family) (Eticha et al., 2010)(+) CTGGATGCAGTTTCAAGAGAG138CV535133
(−) ACTCCAGCAGCTGCAAGTTC
β-tubulin (Eticha et al., 2010)(+) CCGTTGTGGAGCCTTACAAT117CV530631
(−) GCTTGAGGGTCCTGAAACAA

First-strand cDNA synthesis and qRT-PCR

After isolating the RNA from 1-cm root tips of bean genotype VAX 1 (see above, RNA isolation), first-strand cDNA was synthesized using a RevertAid H-Minus first strand cDNA synthesis kit (Fermentas, http://www.fermentas.com) following the manufacturer’s protocol. qRT-PCR was performed using the CFX96™ Real Time System plus C1000™ Thermal Cycler (http://www.bio-rad.com). The SYBR Green detection system was used with self-prepared SYBR Green master mix. The qRT-PCR mix was composed of 1 × hot-start PCR buffer (DNA Cloning Service, Hamburg, Germany), 3.6 mM MgCl2 (DNA Cloning Service), 200 μM each deoxynucleoside triphosphate (dNTP: dATP, dTTP dCTP dGTP) (Fermentas), 0.1 × SYBR Green-I (Invitrogen), 0.75 U μl−1 DCSHot DNA Polymerase (DNA Cloning Service), 252 nM each forward and reverse primer (Biolegio, Nijmegen, The Netherlands), 2 ng μl−1 cDNA template and ultra-pure DNase/RNase-free distilled water (Invitrogen) in a final volume of 25 μl. The qRT-PCR cycling stages consisted of initial denaturation at 95°C (10 min), followed by 45 cycles of 95°C (15 s), 60°C (30 s), 72°C (30 s), and a final melting curve stage of 95°C (15 s), 60°C (15 s) and 95°C (15 s). Samples for qRT-PCR were run in three biological replicates and two technical replicates. Relative gene expression was calculated using the comparative ΔΔCT method according to Livak & Schmittgen (2001). For the normalization of gene expression, β-tubulin was used as an internal standard according to Eticha et al. (2010), and the control (nontreated) plants of bean genotype VAX 1 were used as reference sample.

Confirmation of SuperSAGE expression profiles via qRT-PCR

Parallel RNA extractions to the SuperSAGE library construction in 1-cm root tips from control (−PEG) and PEG 6000-treated plants were used. Forty-six differentially expressed genes according to SuperSAGE with a putative role in the regulation of CW properties and response to OS were selected, and the primers were designed according to the method described above (Table S1). The expression of these genes in control and PEG 6000-treated 1-cm root tips of bean was confirmed by SYBR Green-based qRT-PCR.

Determination of Al

For the determination of Al, 1-cm root tips were digested in 500 μl ultra-pure HNO3 (65%, v/v) by overnight shaking on a rotary shaker. The digestion was completed by heating the samples in a water bath at 80°C for 20 min. Then, 1.5 ml ultra-pure deionized water was added after cooling the samples in an ice–water bath. Aluminium was measured with a Unicam 939 QZ graphite furnace atomic absorption spectrophotometer (GFAAS; Analytical Technologies Inc., Cambridge, UK) at a wavelength of 308.2 nm after appropriate dilution, with an injection volume of 20 μl.

Statistical analysis

A completely randomized design was used with 4–12 replicates in each experiment. Statistical analysis was carried out using SAS 9.2 (SAS Institute, Cary, North Carolina, USA). Means were compared using t or Tukey test, depending on the number of treatments being compared; *, **, *** and ns denote significant differences at < 0.05, < 0.01, < 0.001 and not significant, respectively.

Results

Both PEG 6000 and PEG 1000 treatment induced the same extent of inhibition of root elongation (30%) of common bean at the same OP treatment level (Fig. 1a). Aluminium treatment strongly reduced root elongation (80%). The addition of PEG in the presence of Al reduced significantly the Al-induced inhibition of root elongation to 50% and 40% with PEG 1000 and PEG 6000, respectively (Fig. 1a). The PEG-induced decrease in Al toxicity was related to a significant reduction in Al accumulation in the 1-cm root tips (PEG 6000 > PEG 1000, Fig. 1b).

Figure 1.

Root elongation rate (a) and aluminium (Al) content (b) of 1-cm root tips of the common bean (Phaseolus vulgaris) genotype VAX 1 under osmotic (0, −0.60 MPa osmotic potential (OP)) and Al (0, 25 μM Al) stresses. Plants were precultured in a simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 for 48 h for acclimation and pH adaptation, and then treated without or with 25 μM Al in the absence or presence of polyethylene glycol (PEG) (150 g l−1 PEG 6000, 115 g l−1 PEG 1000) in the simplified nutrient solution for 24 h, pH 4.5. The background value (dashed line) in (b) represents the mean Al content of the root tips without Al treatment. Bars represent mean ± SD, n = 12 for (a) and = 4 for (b). Means with different capital letters are significantly different at < 0.05 (Tukey test) for the comparison of treatments.

The citrate transporter MATE gene and the ACCO gene clearly showed enhanced expression with increasing Al supply (Fig. 2a,b). The relative expression levels were significantly negatively correlated with root elongation as affected by Al supply (Fig. 2c,d) and positively correlated with the Al concentrations of the roots tips (Fig. 2e,f). This confirms the decisive role of root tip Al accumulation in Al-induced inhibition of root elongation.

Figure 2.

The effect of aluminium (Al) (0–30 μM) on the relative expression level of MATE (multidrug and toxin extrusion) (a) and ACCO (1-aminocyclopropane-1-carboxylic acid oxidase) (b) genes in 1-cm root tips of the common bean (Phaseolus vulgaris) genotype VAX 1, and the correlation of MATE and ACCO gene expression with root elongation rate (c, d) and Al content (e, f). Plants were precultured in a simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 for 48 h for acclimation and pH adaptation, and then treated with different concentrations of Al (0–30 μM) in the simplified nutrient solution for 24 h, pH 4.5. The background value (dashed line) in (a) and (b) represents the calibrated reference using untreated plants. Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed using the β-tubulin gene as internal standard. Bars represent mean ± SD, = 4 for root elongation rate and Al content, = 3 for gene expression. Means with different capital letters are significantly different at < 0.05 (Tukey test) for the comparison of Al treatments. ***Significant at < 0.001.

It thus appears that the expression of MATE and ACCO genes is a sensitive indicator of Al toxicity and Al accumulation, which could provide opportunities to further clarify the PEG-induced reduction in Al accumulation in the root tips. In contrast with Al stress, OS induced by PEG 1000 or PEG 6000 did not affect the regulation of either gene (Fig. 3). However, the Al treatment-induced up-regulation was decreased significantly by PEG treatment (PEG 6000 > PEG 1000) in agreement with the greater suppression of Al accumulation and Al stress by PEG 6000 (see Fig. 1).

Figure 3.

Relative expression level of MATE (multidrug and toxin extrusion) (a) and ACCO (1-aminocyclopropane-1-carboxylic acid oxidase) (b) genes in 1-cm root tips of common bean (Phaseolus vulgaris) genotype VAX 1 under osmotic (0, −0.60 MPa osmotic potential (OP)) and aluminium (Al) (0, 25 μM Al) stress. Plants were precultured in a simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 for 48 h for acclimation and pH adaptation, and then treated without or with 25 μM Al in the absence or presence of polyethylene glycol (PEG) (150 g l−1 PEG 6000, 115 g l−1 PEG 1000) in the simplified nutrient solution for 24 h, pH 4.5. The background value (dashed line) represents the calibrated reference using untreated plants. Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed using the β-tubulin gene as internal standard. Bars represent mean ± SD, = 3. Means with different capital letters are significantly different at < 0.05 (Tukey test) for the comparison of treatments.

The removal of PEG from the treatment solution rapidly allowed Al to accumulate in the root tips (Fig. 4a), and the expression of MATE and ACCO genes was restored close to the expression levels of the Al treatment alone, with the exception of ACCO in PEG 1000-treated root tips (Fig. 4b,c).

Figure 4.

Aluminium (Al) content (a) and relative expression level of MATE (multidrug and toxin extrusion) (b) and ACCO (1-aminocyclopropane-1-carboxylic acid oxidase) (c) genes in 1-cm root tips of common bean (Phaseolus vulgaris) genotype VAX 1 under osmotic (−0.60 MPa osmotic potential (OP)) and Al (0, 25 μM Al) stress. Plants were precultured in a simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 for 48 h for acclimation and pH adaptation, and then subjected to a series of Al (0, 25 μM), polyethylene glycol (PEG) 6000 (0, 150 g l−1) and PEG 1000 (0, 115 g l−1) treatment solutions. A detailed description of the different treatments is shown in the table below the figure. Different numbers (1–7) represent different Al and PEG treatment combinations, and asterisks indicate the presence of treatment. Briefly, the plants were pretreated with either PEG 1000 or PEG 6000 for 8 h. They were then transferred for 2 h (8–10 h) to a solution containing Al and/or PEG or not. This treatment was then followed by a 2-h (10–12 h) recovery period from PEG stress in the presence or absence of Al. The background value (dashed line) represents the calibrated reference using untreated plants. Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed using the β-tubulin gene as internal standard. Bars represent mean ± SD, n = 4 for (a) and = 3 for (b, c). Means with different capital letters are significantly different at < 0.05 (Tukey test) for the comparison of treatments.

The results clearly showed that PEG-6000, in particular, reduced the Al accumulation in root tips, and Al accumulation in the same roots tips was rapidly restored on withdrawal of PEG. In order to better understand the molecular basis of the PEG effect, two libraries from control (−PEG 6000) and PEG 6000-treated root tips were constructed using SuperSAGE. After excluding the singletons from the total sequenced 9 015 356 tags of 26 bp in length in both libraries (data not shown), we analysed in total 8 960 486 tags, 3 913 099 (44%) from the root tips of the control and 5 047 387 (56%) from the root tips of PEG-treated plants (Table 2). These tags represented 75 867 unique transcripts (UniTags) overall, 67 185 (89%) from the PEG 6000-treated and 68 969 (91%) from the control roots; among these, 9810 UniTags were up- and 8019 UniTags were down-regulated at P < 0.05 (Fig. 5a, Table 2). UniTags present at < 100, 100–1000 and > 1000 copies per million (copies × million−1) were considered as low-, mid- and high-abundant tags, respectively. The frequency distribution of the 68 969 UniTags in the control library showed 0.2% low-, 2.3% mid- and 97.6% high-abundant tags, and the 67 185 UniTags in the PEG 6000 library showed 0.2% low-, 2.4% mid- and 97.5% high-abundant tags (Table 2). The annotation of the 75 867 UniTags matched 39 314 previously well-characterized sequences from the public databases with a maximum of five mismatches (scores ≥ 42.1; c. 52%). Of these, 28% matched to sequences from P. vulgaris, 6% to Phaseolus coccineus, 3% to Glycine max, 0.3% to Medicago truncatula, 0.2% to Lotus japonicus and 15% to other species (Table 3).

Table 2.   Features of SuperSAGE libraries from control and polyethylene glycol (PEG) 6000-treated root tips of common bean (Phaseolus vulgaris) genotype VAX 1
LibraryControl (%)PEG 6000 (%)Total
  1. *Values normalized to 1 million tags.

Sequenced tags5 047 387 (56)3 913 099 (44)8 960 486 (100)
Number of unique transcripts (UniTags)68 969 (91)67 185 (89)75 867 (100)
Abundance classes of UniTags*
 High-abundant: > 1000 copies million−1104 (0.2)98 (0.2)
 Mid-abundant: 100–1000 copies million−11578 (2.3)1594 (2.4)
 Low-abundant: < 100 copies million−167 287 (97.6)65 493 (97.5)
Total68 96967 185
Figure 5.

(a) Scatter plot showing the UniTags expressed in control and polyethylene glycol (PEG) 6000-treated libraries. (b) Venn diagram showing down- and up-regulated UniTags using rigorous threshold parameters (|R| > 1; < 0.05) in root tips of common bean (Phaseolus vulgaris) genotype VAX 1. In (a), lines represented as 99.9% (dark blue), 99% (yellow) and 95% (turquoise) demarcate UniTags that are significantly up- or down-regulated at P values lower than 0.001, 0.01 and 0.05, respectively. In (b): A/A′, UniTags exclusively expressed in control/PEG 6000 libraries; B, B′, UniTags expressed in both libraries; a, d, a′, d′, no hit; b, c, b′, c′, hit to previously known public databases.

Table 3.   BLAST search results of the SuperSAGE Unitags in different expressed sequence tag (EST) databases
EST databaseControl (%)PEG 6000 (%)Total (%)
Phaseolus vulgaris19 147 (28)18 291 (27)21 121 (28)
Glycine max2116 (3)1993 (3)2223 (3)
Phaseolus coccineus4066 (6)3914 (6)4338 (6)
Medicago truncatula195 (0.3)248 (0.4)248 (0.3)
Lotus japonicus125 (0.2)136 (0.2)136 (0.2)
Other species9449 (14)9542 (14)11 248 (15)
Not matched33 871 (49)33 061 (49)36 553 (48)
Total68 96967 18575 867

For further analysis of the PEG-induced differentially expressed genes, the significant differences in the abundances of UniTags in control and PEG-subjected roots were rigorously limited to a threshold |R| ≥ 1 (≥ two-fold change) and P < 0.05. On this basis, a total of 12 624 UniTags (7492 up-regulated; 5132 down-regulated) were differentially expressed (Fig. 5b) and, of these, 3259 down-regulated UniTags (2153 exclusively expressed in control and 1106 expressed in both control and PEG-treated libraries) and 4673 up-regulated UniTags (2982 exclusively expressed in control and 1691 expressed in both control and PEG-treated libraries) hit previously known sequences from P. vulgaris, P. coccineus, G. max, M. truncatula, L. japonicus and other organisms (Fig. 5b). By assembling the same ESTs from the differentially expressed UniTag annotation in control and PEG 6000-treated roots, and the exclusion of the UniTags with more than six 3′-poly(A) in the 26-bp tags, a total of 611 up- and 728 down-regulated transcripts was obtained for the gene functional categorization (Fig. 6, Table S2). The gene functional categories of the differentially expressed transcripts were BLASTed against the nonredundant GenBank and UniProt protein databases (http://www.uniprot.org/) by the gene ontology (GO) annotation (Table S2).

Figure 6.

Functional categories of differentially expressed transcripts in control and polyethylene glycol (PEG) 6000-treated root tips of common bean (Phaseolus vulgaris) genotype VAX 1.

The unique transcripts were categorized and identified according to the functional category defined by the KEGG PATHWAY and UniProt protein database (Fig. 6). About 47.5% of the up- and 42.0% of the down-regulated transcripts had unknown functions or were unclassified. The metabolism categories (12.4% up; 11.8% down) were subclassified into carbohydrate, energy, amino acid, lipid, nucleotide, secondary and other metabolic pathways. Functional categories equally up- and down-regulated were transcription regulation (5.89% up; 5.22% down) and cytoskeleton (0.33% up; 0.41% down). Predominantly down-regulated transcripts by PEG were found in the categories of signal transduction, transport, stress/defence, CW synthesis and organization, and protein post-translational modification, whereas transcripts in the categories of protein translation, processing and degradation, RNA processing and modification, replication and repair were more often up-regulated (Fig. 6). The differentially regulated UniTags with different functional categories are listed in Table S2.

To validate the results generated from SuperSAGE, 46 differentially expressed genes (Table S1) according to SuperSAGE with a putative role in the regulation of CW properties and response to OS were selected, and their expression was tested by qRT-PCR. A highly significant correlation (R2 = 0.71, P < 0.0001) between SuperSAGE and qRT-PCR was found (Fig. 7).

Figure 7.

Validation of the polyethylene glycol (PEG)-induced expression of 46 genes selected from SuperSAGE in root tips of common bean (Phaseolus vulgaris) genotype VAX 1 by quantitative reverse transcription-polymerase chain reaction (qRT-PCR). ***Significant at P < 0.001.

Among the 46 genes, six genes each with suspected functions in PEG-induced CW modification (XTHa, XTHb, BEG, HRGP, PRP and LTP) and the OS response (bZIP, MYB, AQP, P5CS, SUS and CYP701A) were selected for a more detailed gene expression study using qRT-PCR (Fig. 8). The results showed that, among the CW-associated genes, only LTP and HRGP were significantly up-regulated by PEG treatment, whereas XTHa, XTHb, BEG and PRP were down-regulated. All OS-associated genes were significantly up-regulated with the exception of CYP701A. Removal of the PEG stress for 2 h, which allowed the root elongation rate to recover (data not shown) and partly restored the Al accumulation capacity (see Fig. 4), reversed the gene expression to the control level for XTHa, BEG and HRGP among the CW-associated genes and bZIP, MYB and P5CS among the OS-associated genes. The LTP, SUS and AQP genes remained up-regulated and PRP, XTHb, CYP701A down-regulated compared with the controls. These results were highly reproducible, as similar observations were found in another experiment in which root tips were treated for only 10 h compared with 24 h as in the current experiment (data not shown).

Figure 8.

Relative expression level of cell wall (CW)- and osmotic stress (OS)-associated genes in 1-cm root tips of common bean (Phaseolus vulgaris) genotype VAX 1 under OS (−0.60 MPa osmotic potential (OP)). Plants were precultured in a simplified nutrient solution containing 5 mM CaCl2, 1 mM KCl and 8 μM H3BO3 for 48 h for acclimation and pH adaptation, and then transferred into a simplified nutrient solution containing polyethylene glycol (PEG) 6000 (150 g l−1) for 24 h; the plants were then allowed to recover in a simplified nutrient solution for 2 h. The background value (dashed line) represents the calibrated reference using untreated plants. Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed using the β-tubulin gene as internal standard. Bars represent mean ± SD, = 3. Means with different capital letters are significantly different at < 0.05 (Tukey test) for the comparison of treatment periods.

Discussion

In the CW of the root apex, Al is primarily bound to the negatively charged carboxylic groups (COO) provided by demethylated pectin (Blamey et al., 1990; Horst, 1995; Horst et al., 2010). The exclusion of Al from the root tip apoplast is a prerequisite for the reduction of Al-induced inhibition of root elongation, and thus Al resistance, in common bean, which is conferred by citrate exudation (Rangel et al., 2009; Horst et al., 2010). The role of Al exclusion from the root tip in Al resistance was corroborated by the current study which demonstrated that the PEG (OS)-induced improvement of root growth under Al stress was related to a reduction in Al accumulation in the root tip (Fig. 1). The PEG (OS)-induced reduction of Al accumulation was not a result of enhanced citrate exudation, precipitation or complexation of Al3+ in the PEG treatment solution or of a reduction in CW negativity, but rather of the reduction in CW porosity which limits Al flux into the apoplast (Yang et al., 2010). This conclusion is based on the specificity of exclusion for Al compared with lanthanum (La), strontium (Sr) and rubidium (Rb), and is consistent with their hydrated ionic radii (Al3+ > La3+ > Sr2+ > Rb+) (Yang et al., 2010). In this study, a greater reduction in Al accumulation in PEG 6000- than in PEG 1000-treated plants was observed (Fig. 1). This differential change in Al accumulation is related to the molecular size and estimated hydrodynamic radii (PEG 6000 > PEG 1000) of the applied PEGs. The higher the hydrodynamic radius of the osmotic solute, the better the exclusion from the apoplast, and thus the higher level of dehydration of the apoplast (Kuga, 1981; Yang et al., 2010). The removal of PEG from the pretreatment solution quickly allowed the accumulation of Al in the root tip (Fig. 4a), indicating that the CW can rapidly recover from the shrinkage and structural alteration caused by OS independent of the osmotic solute used.

In agreement with the highly sensitive reaction of MATE and ACCO gene expression in response to Al in common bean, reported by Eticha et al. (2010), a significant correlation between MATE and ACCO gene expression, root elongation and Al concentration in the root tips of common bean was found (Fig. 2). PEG treatment alone had no effect on the expression of the two genes (Fig. 3), providing opportunities to further clarify the PEG-induced reduction in Al accumulation in the root tips and Al-induced inhibition of root elongation using the expression of MATE and ACCO genes as a sensitive indicator of Al toxicity. As expected, the OS-induced exclusion of Al from the root apex and the improvement in root growth were accompanied by a decrease in gene expression of MATE and ACCO (Fig. 2), confirming that OS reduces Al injury.

In this study, the results generated from SuperSAGE were well confirmed by qRT-PCR in OS-treated bean root tips (Fig. 7), indicating the reliability of SuperSAGE. However, approx. 49% of the UniTags generated from SuperSAGE did not match with previously known genomic and EST sequences found in public databases (Table 3). This may have hampered the identification of genes responding to OS in the root tips. Among the differentially expressed genes, c. 55% of the UniTags matched with P. vulgaris EST databases (Table 3). The observation that nearly one-half of the tags generated by SuperSAGE did not match with previously known sequences could be attributed to the nature of the 26-bp tags. The 26-bp tag fragments were isolated from the NlaIII recognition site (5′-CATG-3′) closest to the poly-A tail of the cDNA, which in most cases lies in the 3′-untranslated region (3′-UTR). Therefore, it appears that the 3′-UTRs of common bean transcripts are very specific and do not match with previously known sequences.

Although SuperSAGE is a powerful tool for quantitative gene expression analysis, as well as for the discovery of novel genes, the method may miss out some transcripts. In particular, transcripts which do not have the NlaIII restriction site (5′-CATG-3′), and those which have the restriction site extremely close to the poly-A tail, cannot be recognized. Based on in silico sequence data analysis of the Arabidopsis RefSeq database, Matsumura et al. (2010) reported that, within 35 286 genes, 2000 genes (5.7%) did not have the NlaIII restriction site. Similarly, in common bean, some genes which lack the recognition site for the anchoring enzyme, NlaIII, might have been missed out.

Drought-induced genes were classified into two groups according to microarray analysis in Arabidopsis (Shinozaki et al., 2003). The first group codes for proteins which probably function in stress tolerance, such as late embryogenesis abundant proteins, osmotin, key enzymes for osmolyte biosynthesis, such as proline, water channels, sugar and proline transporters, and lipid transfer proteins. It has been reported that dehydration induces the expression of osmoregulation-related genes, such as Δ1-pyrroline-5-carboxylate synthase (P5CS) and sucrose synthase (SUS) in the resurrection plant Craterostigma plantagineum (Kleines et al., 1999; Rodriguez et al., 2010), maize (Zea mays) (Zheng et al., 2004; Spollen et al., 2008) and Arabidopsis (Oono et al., 2003). Overexpression of the P5CS gene in various plants results in elevated proline production and improved OS tolerance (Bartels & Sunkar, 2005). In addition, it has been found that some water transport-related AQP (aquaporin family protein) genes are up-regulated by drought in Arabidopsis, upland rice (Oryza sativa) and grapevine (Vitis vinifera) (Alexandersson et al., 2005, 2010; Lian et al., 2006; Vandeleur et al., 2009), which may trigger greater membrane water permeability, facilitating water flux (Bartels & Sunkar, 2005).

The second group comprises genes coding for regulatory proteins involved in signal transduction and transcription factors, such as bZIP, MYB, MYC and DREB (Shinozaki & Yamaguchi-Shinozaki, 2007). The bZIP and MYB genes are transcription factors involved in an abscisic acid (ABA)-dependent pathway mediating gene expression in plants during OS (Shinozaki & Yamaguchi-Shinozaki, 2007). Rodriguez-Uribe & O’Connell (2006) found that a root-specific bZIP transcription factor is responsive to water deficit in tepary bean (Phaseolus acutifolius) and common bean, which may allow the plant to maintain root elongation. More recently, the OsbZIP23 gene in rice and the MYB96 gene in Arabidopsis were found to confer drought resistance (Xiang et al., 2008; Seo et al., 2009). In addition, the cytochrome P450s superfamily (CYP) may serve as mono-oxygenases involved in the biosynthesis of metabolites conferring abiotic stress tolerance (Schuler & Werck-Reichhart, 2003).

On the basis of this information, the OS-associated genes SUS, P5CS, AQP, CYP701A, bZIP and MYB were selected to underpin the PEG-induced dehydration and rehydration in the root tips in the present study. All of these genes were significantly up-regulated by OS stress, with the exception of CYP701A, and the removal of PEG stress for only 2 h rapidly reversed the gene expression to the control level for P5CS, bZIP and MYB (Fig. 8), suggesting a differential response of gene expression to dehydration and rehydration in common bean. The rapid reversal of bZIP and MYB gene expression after the removal of PEG stress may support the recovery of root elongation within 2 h (data not shown) by mediating the expression of ABA-regulated genes, as it was reported that, in maize, the accumulation of ABA in the root tips is required for the maintenance of primary root elongation at low water potentials (Sharp et al., 2004). P5CS catalyses the first committed and rate-limiting step for proline biosynthesis in plants (Kavi Kishor et al., 2005). In addition to osmotic adjustment, proline also functions as a major constituent of CW structural proteins in plants (Nanjo et al., 1999). In P5CS antisense transgenic Arabidopsis thaliana plants, proline and hydroxyproline contents in hydrolysates of a purified CW fraction were specifically and significantly reduced, indicating that proline deficiency affects the biosynthesis of CW matrix proteins, such as proline-rich proteins (PRPs) and hydroxyproline-rich glycoproteins (HRGPs) (Nanjo et al., 1999), which could provide mechanical support for cells under stressed conditions (Cosgrove, 1997). Therefore, the rapid reversal of P5CS gene expression after the removal of PEG stress may allow the recovery of CW porosity by modifying the CW structural properties. A further discussion of the OS-induced expression of PRP and HRGP genes is given later in this section.

Among the OS-regulated genes, CW synthesis and organization-related genes were mostly down-regulated (Fig. 6), providing an opportunity to identify the genes involved in the OS-induced alteration of CW porosity. Water loss in plant tissues reduces turgor pressure and so affects directly the extensibility of the plant CW. In maize, Sharp et al. (2004) reported that the extent of osmotic adjustment in the root tip is insufficient to maintain turgor under severe water deficit. The maintenance of root elongation requires the enhancement of longitudinal CW extensibility under water stress (Sharp et al., 2004; Yamaguchi & Sharp, 2010). Some genes may play a role in CW extension during dehydration. For example, it has been reported that LTPs (lipid transfer proteins) are associated with hydrophobic wall compounds, causing nonhydrolytic disruption of the CW and subsequently facilitating wall extension in tobacco (Nieuwland et al., 2005). Our study showed that the expression of an LTP gene in the root tips of bean was enhanced significantly by OS (Fig. 8), suggesting that this gene may contribute to the maintenance of the root elongation of common bean under OS (Fig. 1).

Structural proteins are one of the main components of the growing plant CW. The CW structural proteins have been classified according to their predominant amino acid composition, for example, HRGP, glycine-rich protein and PRP (Cosgrove, 1997). In our studies, we observed the differential expression of genes encoding HPRG and PRP in the root tips of common bean by OS (Fig. 8, Table S2). These proteins can be insolubilized rapidly in the CW during stress conditions, such as on wounding (Showalter, 1993; Cosgrove, 1997). By contrast, two proline-rich glycoproteins of 33 and 36 kDa (p33 and p36), similar to soybean PRP2, were highly accumulated in the soluble fraction of the CWs in common bean in response to water deficit (Covarrubias et al., 1995; Battaglia et al., 2007). However, this does not exclude the possibility that larger amounts of proline-rich glycoproteins are tightly bound with the CW polymers in immobilized form, as only very small amounts of p33/p36 were detected in the soluble fraction of CWs of well-watered bean hypocotyls, and immunolocalization indicated that these proteins were abundantly localized in the cell corners of the cortex, epidermis, pith, vascular cells and phloem. However, we found that the expression of one PRP gene was suppressed significantly by OS and the recovery of roots from OS could not rapidly restore the expression of this gene, which was in contrast with the HRGP gene (Fig. 8). The expression of an HRGP gene was enhanced significantly by OS, and withdrawal of OS rapidly restored the expression of this gene (Fig. 8). The HRGPs are particularly abundant in dicots compared with other structural proteins (Showalter, 1993); thus, the role of the HRGP gene appears to be important in the OS-induced modification of the CW in common bean. Once HRGP is secreted into the wall, it will be rapidly insolubilized. The insolubility of HRGP may be mediated by the water deficit-induced enhancement of hydrogen peroxide and catalysed by a CW peroxidase. This response is thought to be an ultra-rapid stress response reaction that serves to further strengthen the CW (Showalter, 1993; Zhu et al., 2007).

It has been reported that the pectin matrix is the decisive factor in CW porosity (Baron-Epel et al., 1988). Although the pectin content in the root tips of bean was reduced by approx. 25% as a result of PEG 6000-induced OS (Yang et al., 2010), this level of decrease in pectin content may not drastically alter CW porosity. It thus appears that the reduction in CW porosity may be largely a result of the physical shrinkage of the CW, and further enhanced by the deposition of other wall components, such as structural proteins, as depicted schematically in Fig. 9(a). The increased deposition of HRGPs in the wall may increase the cross-linking between HRGPs and other wall components, such as pectin (Showalter, 1993), and further reinforce the CW barrier, thus impeding Al uptake (Fig. 9b).

Figure 9.

(a) A model representing the effect of osmotic stress (OS) on the cell wall (CW) structure of common bean (Phaseolus vulgaris) plants. (b) Hypothesis for the possible role of CW modification-related genes or structural proteins in the OS-induced change in CW porosity, and thus aluminium (Al) binding to the CW. In (b), the blue and red arrows denote up- and down-regulated changes, respectively; the solid and dashed arrows indicate the effect of OS and recovery (RC), respectively. ?, uncertain; …, unidentified genes. Green line frame represents CW. BEG, β-1,3-glucanase; HRGP, hydroxyproline-rich glycoprotein; XTH, xyloglucan endotransglucosylase/hydrolase.

The shrinkage of the CW resulting from PEG 6000-induced dehydration of the apoplast causes adhesion and cross-linking of wall polymers through hydrogen bonding. This bonding will be enhanced by the removal of water from the apoplast, and is likely to cause an irreversible bonding between polymers, resulting in altered biophysical CW properties (Fig. 9a; Moore et al., 2008; Yang et al., 2010), unless some CW loosening or modifying genes/enzymes are re-induced/activated. In Arabidopsis, water deficit consistently down-regulated the expression of the genes involved in CW synthesis and modification (Bray, 2004). Similarly, in common bean, the number of down-regulated genes related to CW synthesis and organization under OS was two-fold higher than that of up-regulated genes (Fig. 6). Several CW proteins/enzymes are believed to play key roles in modifying the wall structure and controlling wall extension. These include expansin, xyloglucan endotransglucosylase/hydrolase (XTH) and glucanases (Wu & Cosgrove, 2000; Bray, 2004; Sharp et al., 2004; Moore et al., 2008).

The XTH proteins are a large family of CW proteins with 33 members known in the Arabidopsis genome; they are involved in controlling CW extensibility through the cleavage and re-formation of bonds between xyloglucan chains (Rose et al., 2002; Bray, 2004). Under soil moisture deficit, the genes encoding XTH were among the commonly down-regulated genes when 23 genes annotated in the XTH family were analysed (Bray, 2004). In chickpea (Cicer arietinum), Romo et al. (2005) found that the expression of the CaXTH1 gene encoding the XTH protein was repressed by PEG treatment, which inhibited epicotyl growth. The removal of PEG resulted in the restoration of the normal expression level of this gene, suggesting the involvement of XTH encoded by CaXTH1 in cell expansion. Spatial analysis of CW proteomics in maize primary roots indicated that the abundance of XTH proteins was significantly reduced by water deficit in the first 3 mm of the root apex (Zhu et al., 2007). In the present study, qRT-PCR results showed that the expression of XTHa and XTHb genes in the root tips of common bean was reduced significantly by PEG-induced dehydration of the apoplast (Fig. 8). Withdrawal of PEG from the pretreatment solution rapidly allowed the recovery of the expression level of the XTHa gene (Fig. 8), supporting the view that the XTHa gene may be involved in the CW modification during the recovery period of the apoplast from dehydration (Fig. 9b).

In addition, glucan endo-1,3-β-glucosidase (or β-1,3-glucanase) (BEG) may also play an important role in OS-induced wall modification and thus influence the Al accumulation that was observed in the present study (Fig. 9b). BEGs are abundant proteins found in all higher plants. They are known to be involved in pathogen defence as well as in a wide range of normal developmental processes, and can hydrolytically cleave the 1,3-β-linked glucans, a major component of the fungal CW (Minic & Jouanin, 2006). BEG belongs to the family of 17 plant glycoside hydrolases, and molecular studies have suggested that this enzyme shares a common ancestry with β-1,3-1,4-glucanase (Minic & Jouanin, 2006; Borad & Sriram, 2008). Although the function of many CW enzymes has yet to be determined, a role for BEG in the OS-induced CW modification of common bean root tips in the present study is likely as PEG-induced OS reduced significantly the BEG gene, but removal of PEG led to a rapid recovery of its expression (Fig. 8). Wu et al. (2001) reported that the induction of BEG in the micropylar tissues of imbibed tomato (Lycopersicon esculentum) seeds can be inhibited by ABA. Therefore, the expression of the BEG gene may be repressed by the OS-induced accumulation of ABA in the root apex of plants. This is corroborated by the observation that water deficit increased the accumulation of ABA in the root apex of maize (Sharp et al., 2004; Yamaguchi & Sharp, 2010), which consequently reduced the abundance of BEG proteins in the root tips (Zhu et al., 2007).

In the present study, PEG treatment by itself inhibited root growth by 30% compared with the control, possibly by reducing CW expansion. Although sole Al treatment inhibited root growth by 80%, combined Al and PEG treatment only resulted in 40% inhibition (see Fig. 1a). Therefore, the PEG-induced recovery from Al injury resulted from the reduced Al accumulation in the root tips of common bean as a consequence of the PEG-induced reduction in CW porosity (Yang et al., 2010). The enhanced expression of the HRGP gene and the reduced expression of the XTH and BEG genes may contribute to the reduction in CW porosity, but the main cause of the reduction in CW porosity by PEG is the OS-induced physical collapse of CW structure (presented in Fig. 9a). The main role of HRGP, XTH and BEG is involvement in the recovery of CW porosity (Fig. 9b).

In conclusion, our results suggest that several CW-modifying- and CW-assembling-related genes, such as XTHs, BEG and HRGP, may play important roles in the PEG (OS)-induced changes in CW porosity leading to reduced Al accumulation in the root tips. There is a need for further research to determine the role of functional genes related to CW modification under conditions of water deficit.

Acknowledgements

This research was supported by a restricted core project from the Bundesministerium für Wirtschaftliche Zusammenarbeit/Gesellschaft fur Technische Zusammenarbeit (BMZ/GTZ) (No. 05.7860.9-001.00) granted to the International Center for Tropical Agriculture (CIAT). We thank Dr Steve Beebe, leader of the Bean Program of CIAT, for the supply of seeds of the common bean genotypes, and the China Scholarship Council for providing a scholarship to the first author.

Ancillary