Genetic Interactions Underlying the Biosynthesis and Inhibition of β-Diketones in Wheat and Their Impact on Glaucousness and Cuticle Permeability

Cuticular wax composition greatly impacts plant responses to dehydration. Two parallel pathways exist in Triticeae for manipulating wax composition: the acyl elongation, reduction, and decarbonylation pathway that is active at the vegetative stage and yields primary alcohols and alkanes, and the β-diketone pathway that predominates at the reproductive stage and synthesizes β-diketones. Variation in glaucousness during the reproductive stage of wheat is mainly controlled by the wax production genes, W1 and W2, and wax inhibitor genes, Iw1 and Iw2. Little is known about the metabolic and physiological effects of the genetic interactions among these genes and their roles in shifting wax composition during plant development. We characterized the effect of W1, W2, Iw1, and Iw2 and analyzed their interaction using a set of six near-isogenic lines (NILs) by metabolic, molecular and physiological approaches. Loss of functional alleles of both W genes or the presence of either Iw gene depletes β-diketones and results in the nonglaucous phenotype. Elimination of β-diketones is compensated for by an increase in aldehydes and primary alcohols in the Iw NILs. Accordingly, transcription of CER4-6, which encodes an alcohol-forming fatty acyl-CoA reductase, was elevated 120-fold in iw1Iw2. CER4-6 was transcribed at much higher levels in seedlings than in adult plants, and showed little difference between the glaucous and nonglaucous NILs, suggesting that Iw2 counteracts the developmental repression of CER4-6 at the reproductive stage. While W1 and W2 redundantly function in β-diketone biosynthesis, a combination of both functional alleles led to the β-diketone hydroxylation. Consistent with this, transcription of MAH1-9, which encodes a mid-chain alkane hydroxylase, increased seven-fold only in W1W2. In parallel with the hydroxyl-β-diketone production patterns, glaucousness was intensified and cuticle permeability was reduced significantly in W1W2 compared to the other NILs. This suggests that both W1 and W2 are required for enhancing drought tolerance.


Introduction
The aerial organs of terrestrial plants are coated with an extracellular layer of hydrophobic lipids, termed cuticle. Produced by epidermal cells, the cuticle plays important roles in plant growth and development and, as the interface between sessile plants and the environments they live in, in the interaction with abiotic and biotic elements [1,2]. Based on solubility in organic solvents, the cuticle is composed of insoluble cutin and soluble cuticular wax. Cutin, a cell wall-bound ester polymer of modified fatty acids and glycerol, serves as the backbone of the cuticle [3,4,5]. Intracuticular wax is embedded in the underlying cutin framework and epicuticular wax is overlaid on the cutin matrix and intracuticular wax [2,6]. Wax composition varies with developmental stage, between organs, and with genetic and environmental conditions [1,4,6], causing the plant to be bluishwhite (glaucous) or nonglaucous. Glaucousness is the visible form of densely distributed epicuticular wax crystallites.
The resulting VLCFAs can be released from the FAE complex as free fatty acids, or converted either to primary alcohols by acyl reduction [16] or to alkanes by acyl decarbonylation through an aldehyde intermediate [6,17] and further to secondary alcohols and ketones by hydroxylation [18]. In the acyl reduction branch, fatty acyl-CoA reductase CER4 exclusively reduces VLCFAs to the corresponding primary alcohols [19], and wax ester synthase/ acyl-CoA:diacylglycerol acyltransferase 1 (WSD1) uses the longchain and very-long-chain primary alcohols and C 16 fatty acid for wax ester production [20]. In the decarbonylation branch, CER1 physically interacts with the wax-associated protein CER3 and ER-localized cytochrome b5 isomers to catalyze the redoxdependent biosynthesis of alkanes [21]. Subsequently, midchain alkane hydroxylase 1 (MAH1), a cytochrome P450 enzyme, oxidizes the alkanes to generate secondary alcohols and ketones [18].
All of the cuticular lipid species synthesized in the ER need to be deposited extracellularly. The ABC transporters ABCG12/CER5 [22] and ABCG11/WBC11 [23,24] are involved in exporting wax through the plasma membrane to the apoplastic space. A glycosylphosphatidylinositol-anchored lipid transfer protein, LTPG, functions as part of the machinery, either as a regulatory component or by creating the appropriate conditions for cuticular lipid exportation [25]. Recent efforts have started to shed light on the regulatory network underlying wax production by identifying transcription factors involved in this process. Several transcription factors of the AP2/EREBP [26,27,28,29] and MYB family [30,31,32,33,34] regulate genes involved in cuticle biosynthesis. CER7, a 39-to-59 exoribonuclease, conditions wax synthesis by degrading a specific mRNA species of a CER3 repressor [35,36].
In the tribe Triticeae, including wheat and barley, the acyl elongation, reduction, and decarbonylation pathway for the synthesis of alcohol-rich waxes is active in the vegetative stages. A parallel pathway, which predominates in the reproductive stages, is responsible for biosynthesis of b-diketones, i.e., hentriacontane-14,16-dione and its hydroxyl derivatives [48,49,50]. The b-diketone pathway differs from the acyl elongation, reduction, and decarbonylation pathway in substrate and inhibitor specificity [49] and in pathway-specific genes [51]. Because b-diketones are not detected in wax extracts from the model plants Arabidopsis and rice, the b-diketone biosynthetic pathway draws much less attention.
Common wheat or bread wheat (Triticum aestivum L) is an important, widely adapted crop that is cultivated in many arid or semiarid areas. The inheritance of wheat glaucousness is mainly governed by two sets of dominant genes, W1 and W2, which promote glaucous formation, and Iw1 and Iw2, which inhibit it. W1 and Iw1 are located on the short arm of chromosome 2B (2BS), and W2 and Iw2 on 2DS [52]. Genetic mapping using molecular markers has localized the Iw1 [53] and Iw2 locus [54,55] to the distal regions of the chromosome arms. While the non-glaucous phenotype prevails in the wild diploid ancestors [52,56], the glaucous type predominates in cultivated polyploid wheat, suggesting that glaucousness was favored during wheat domestication. Field physiology studies showed that glaucousness is positively related to wheat yield, especially under droughtstressed conditions [57,58,59]. Despite its adaptive importance, the molecular mechanisms underlying variation in glaucousness and its association with drought tolerance remain largely un-known. We characterized a set of six near isogenic lines (NILs) by scanning electron microscopy (SEM), metabolite profiling of wax extracts, measuring the rate of water loss and chlorophyll efflux, and transcriptional profiling of 72 cuticle genes. Here, we report the results of this study and their implications in our understanding of cuticular wax pathways and drought tolerance.

Genetic Background Analysis of Wax NILs
Six wax NILs were developed by ten backcrosses to wheat cultivar S-615 [52] and their genotypes and glaucousness on flag leaf sheaths and peduncles are shown in Figure 1. The NILs W1W2, W1w2, and w1W2 are glaucous, whereas w1w2, Iw1iw2 and iw1Iw2 are not, indicating that W1 and W2 are functionally redundant in forming glaucousness, but that a single Iw gene is sufficient to suppress glaucousness formation. Before physiological, metabolic, and expressional characterization, we inspected the genetic background by single nucleotide polymorphism (SNP) and simple sequence repeat (SSR) genotyping. Of the 9,000 SNP probes assayed, over 5,200 produced positive signals in each line and 18 probes (0.35% of 5,200) or fewer detected polymorphisms among the NILs, suggesting that the NILs are identical over 99.5% of the genome. The lowest level of polymorphism was found between NILs W1W2 and iw1Iw2, where only one probe detected a polymorphism. We also genotyped the six NILs using SSR markers previously mapped at the tips of 2BS and 2DS arms, where the W and Iw genes are located. All of the NILs were positive for all SSR markers, indicating that the mutations were not caused by chromosome deletions. These results indicate that the NILs differ from each other only at a very small fraction of the genome and suggest that there are probably no polymorphisms in wax genes other than at the W and Iw loci.

Wax Morphology
When flowering begins at Fakes' stage 10.5.1 (F10.5.1), glaucousness is fully developed in the upper part of the wheat plant. To gain insight into the organization of epicuticular wax, we examined the wax crystallites deposited on the abaxial surface of the flag leaf blades and sheaths, the peduncles, and the glumes of the NILs at this stage under SEM. The morphology of wax crystallites in glaucous NILs was clearly different from that of nonglaucous ones ( Figure 2). In all tissue types, the cuticle surfaces of glaucous NILs were covered with a meshwork of wax tubes, and the density was higher in W1W2 (Figure 2a, 2b, 2c, and 2d) and W1w2 (Figure 2e, 2f, 2g, and 2h) than in w1W2 (Figure 2i, 2j, 2k, and 2l). Besides wax tubes, large wax sheets were also observed above the wax meshwork in the sheath and peduncle of the W1W2 plants (Figure 2b and 2c). On the leaf cuticles of the w1W2 plants, sparse wax tubes were seen over the ground cells, whereas fine wax granules densely covered the guard cells (Figure 2i).
In contrast to the glaucous NILs, the cuticles of the nonglaucous NILs were either smooth (Figure 2p) or carried fine wax particles. On the leaf cuticles, wax crystallites were mainly found over the guard cells (Figure 2m Wax morphology is determined by wax composition. The dramatic difference in wax morphology suggests possible differences in wax content, composition, and cuticle permeability between the glaucous and nonglaucous NILs. Considering that sheath wax morphology was a good indicator of the differences among the NILs, we concentrated our research on the sheath wax.

Wax Composition
We extracted cuticular wax from the flag leaf sheaths of the wax NILs at stage F10.5.1 and profiled it using gas chromatographymass spectrometry (GC-MS). The results showed that ,90% of the wax extract consisted of known wax species and the remaining ,10% consisted of unknown species. As expected, the NILs varied greatly in total wax load and composition. All glaucous NILs carried the same wax load (P.0.82980; Figure 3a) and had a similar wax composition, with b-diketones accounting for .60%, fatty acids and alkanes for ,15% each, primary alcohols for 2 to 4%, and cyclic lipids for trace amounts of total wax (Figures 3b and 3c). Wax load varied by ,60% among the nonglaucous NILs. Compared to W1W2, no significant reduction was found in iw1Iw2 (P = 0.42065); however, a 50% and 68% reduction was detected in Iw1iw2 (P = 0.0412) and w1w2 (P = 0.01577), respectively ( Figure 3a). This indicates that wax composition, instead of wax load, determines glaucousness. In contrast to the glaucous NILs, b-diketones were depleted in the nonglaucous NILs, being reduced to ,8% of total wax species in w1w2 (P = 0.00925) and completely eliminated in Iw1iw2 and iw1Iw2 ( Figure 3c). There was little net increase in w1w2 for the remaining components (P.0.12834), which elevated the alkane proportion to ,45% of the total wax species (Figure 3c), indicating that W genes are specific for b-diketone synthesis. In contrast to the situation in w1w2, loss of b-diketones in Iw1iw2 and iw1Iw2 was compensated by an increase of aldehydes and primary alcohols. Compared to W1W2, aldehydes and primary alcohols were increased ,250-fold and ,10-fold in Iw1iw2, and ,600-fold and 20-fold in iw1Iw2, respectively (P,0.00254; Figure 3b). This indicates that the Iw genes inhibit the synthesis of b-diketones and shunt their substrates to the fatty acyl reduction pathway. In this respect, the action of Iw2 is much stronger than that of Iw1. As a result, three different types of waxes are recognized in this NIL set: b-diketone-rich wax in the glaucous NILs, alkane-rich wax in w1w2, and primary alcohol-rich wax in Iw1iw2 and iw1Iw2 ( Figure 3c). In contrast to the aliphatic wax species, cyclic wax species, such as sterols and triterpenes (ST&TP) were detected at much lower levels. Compared to W1W2, ST&TP, which mainly consisted of b-sitosterol and b-amyrin, were increased nine-fold in w1w2 (P = 0.0171; Figure 3b).
A single carbon chain length, C 31 , was detected for b-diketones; however, three substituted derivatives were identified: 8-and 9hydroxyl isomers, and enolic isomer. Considering keto-enol tautomerism, we combined enolic b-diketone with b-diketone. No difference in the unsubstituted b-diketone was found among the glaucous NILs; however, that amount was reduced .10-fold in w1w2 compared with in the glaucous NILs (P,0.00985; Figure 4c). Hydroxyl-b-diketones were only detected in W1W2, and not in W1w2 and w1W2 (Figure 4c), indicating that an interaction between W1 and W2 is required for b-diketone hydroxylation at C 8 or C 9 . Depending on the presence of hydroxyl b-diketones, the b-diketone-rich wax can be further divided into two types, OHb wax in W1W2 and b-D wax in W1w2 and w1W2.

Transcriptional Profiling of Cuticle Genes
When flag leaves are fully emerged from the whorl (stage F9.0), flag leaf sheaths elongate rapidly. Because cuticle genes are highly expressed in the elongating epidermal cells [60,61], we selected flag leaf sheaths at this stage for transcription profiling. We chose nine candidate reference genes (Table S1) from previous publications [62,63,64] and validated their expression stability in 18 cDNA samples from the NIL set using programs qBase Plus (Biogazelle, Belgium) and NormFinder [65]. Both programs b-Diketones, Glaucousness and Cuticle Permeability demonstrated that TaRPII36 was the best reference gene for quantifying cuticle gene expression ( Figure S3).
The increase of aldehydes and primary alcohols in Iw1iw2 and iw1Iw2 suggests that the Iw genes activate CER4 homologs; the abundance of b-diketones in the glaucous NILs suggests that the W genes up-regulate special KCS homologs and the decarbonylation components and the Iw genes suppress them; and the presence of hydroxyl-b-diketones only in W1W2 suggests that the interaction between W1 and W2 activates MAH1 homologs. To test these hypotheses, we profiled the transcription of wheat wax candidate genes with an emphasis on the CER1, CER3, CER4, KCS, and MAH1 gene families. To test if cutin is also involved in the phenotype variation, we selected five cutin biosynthetic genes. We designed primers for 64 unigenes and adopted primers for eight additional genes from a previous publication [66]. Thus, we evaluated the transcription of 72 genes involved in cutin and wax biosynthesis, transport, and transcription regulation (Table S1). Compared with W1W2, the expression of 11 genes in Iw1iw2 and 29 genes in iw1Iw2 was significantly altered, of which the expression patterns of seven genes were shared by Iw1iw2 and iw1Iw2 (Table S2). Of these 72 genes examined, nine, including four CER4 members, four MAH1 members, and LTP4, showed over a two-fold difference in expression. The most dramatic change was observed in CER4-6, which was up-regulated ,120fold in iw1Iw2 and 2.8-fold in Iw1iw2 ( Figure 5). Searching the Dgenome physical mapping database (http://probes.pw.usda.gov/ WheatDMarker) revealed that CER4-6 is located on chromosome 5D. This indicates that CER4-6 is not Iw2 itself, which is located on chromosome 2D, but an Iw2 target that possibly plays a role in increasing aldehyde and primary alcohol content. CER4-11 was also up-regulated two-fold in both Iw1iw2 and iw1Iw2. In addition, four MAH1 members, MAH1-4, MAH1-7, MAH1-8, and MAH1-9, were differentially expressed in Iw1iw2 and iw1Iw2 ( Figure 5 and Table S2). Together, these results indicate that Iw1 differs from Iw2 in terms of the molecular mechanisms for increasing aldehyde and primary alcohol content.
We further asked if the up-regulation of CER4-6 was the result of activation or de-repression by Iw2. To test this, we examined the transcription of CER4-6 in sheaths of W1W2, Iw1iw2, and iw1Iw2 plants at the seedling stage (F4.0). Wax composition changed dramatically during wheat development, from alcoholrich wax in the vegetative stage to b-diketone-rich wax in the reproductive stage, particularly in leaf sheaths [48]. We found that CER4-6 transcription was increased by less than two-fold in Iw1iw2 and iw1Iw2 compared to W1W2 at stage F4.0 (P,0.0002; Figure 6a), which was much smaller than the difference detected at stage F9.0 ( Figure 5). Compared to stage F4.0, CER4-6 was downregulated 815-fold in W1W2 (P = 1610 27 ) and 450-fold in Iw1iw2 (P = 2610 27 ), but only 17-fold in iw1Iw2 at stage F9.0 (P = 4610 27 ; Figure 6b). This indicates that expression of CER4-6 is under developmental regulation and is suppressed at the reproductive stages, and that Iw2 counteracts this developmental suppression.
Compared to the situation in w1w2, expression of four genes in W1w2, 19 genes in w1W2, and 26 genes in W1W2 was significantly altered (Table S3). The expression patterns of three genes overlapped between W1w2 and W1W2, 15 between w1W2 and W1W2, and two among all three glaucous NILs. While transcriptional changes of five genes relative to w1w2 were detected in W1w2 or w1W2 but not in W1W2, changes of another 10 genes were detected in W1W2 but not in W1w2 or w1W2. This indicates that W1 and W2 had a non-additive effect on the expression of these cuticular wax-related genes. The expression of four genes in w1W2 and six genes in W1W2 was altered by twofold or more, including four CER4 members, five MAH1 members, and CER1-8 (Table S3; Figure 7). The expression of MAH1-8 matches the production pattern of hydroxyl-b-diketones: no change in W1w2 and w1W2 were observed, but the expression increased seven-fold in W1W2. MAH1 is responsible for alkane hydroxylation in Arabidopsis [18], suggesting a role for this gene in generating hydroxyl-b-diketones in wheat.

Cuticle Permeability
Nonstomal transpiration is directly correlated with cuticle permeability. We evaluated the cuticle permeability of this NIL set by measuring the rates of water loss and chlorophyll efflux of the flag leaf sheaths at stage F10.5.1. Simultaneously, we inspected stomatal density and aperture size under a light microscope. All of the NILs had a similar stomal density, i.e., ,60 stomata in a field of view at a magnification of 10620, and most stomata were closed 1 h after detachment at room temperature and under lab conditions. Therefore, the differences observed in water-loss rate were presumably attributed to cuticle permeability.
Compared to W1W2, Iw1iw2 and iw1Iw2 started showing higher rates of water loss 1 h after detachment (P,0.03694) and the differences remained and even increased thereafter (Figure 8a). The Iw NILs also showed significantly higher chlorophyll leaching rates than W1W2 after 4 h of treatment (P,0.04285; Figure 8b). Although Iw1iw2 and iw1Iw2 exhibited similar chlorophyll efflux rates (P.0.05503; Figure 8b), Iw1iw2 showed a higher water-loss rate after 4 h of treatment (P,0.03272; Figure 8a). These findings echo the wax data and suggest that the increased amount of primary alcohols and aldehydes in iw1Iw2 have some effect on blocking nonstomatal transpiration, but not on preventing chlorophyll efflux.
Similar to iw1Iw2, w1w2 showed a significantly higher rate of both water loss and chlorophyll leaching compared to W1W2 (Figure 8a and 8b). The difference in water loss between w1w2 and W1W2 was detectable as early as 1 h after detachment (P = 0.03607; Figure 8c) and that in chlorophyll leaching at 6 h of treatment (P = 0.04527; Figure 8d). In both experiments, the performance of w1W2 was similar to that of w1w2 (P.0.13101) and significantly different from W1W2 after 3 h of desiccation and 6 h of chlorophyll-leaching (P,0.0494). The findings for W1w2 were not consistent in the two experiments. The water-loss rate of this NIL was intermediate, being higher than that of W1W2 (P,0.04174), but lower than that of w1w2 (P,0.0356 (Figure 8c). In the chlorophyll-leaching experiment, W1w2 showed high similarity to w1w2 (P.0.11128) and significant difference from W1W2 after 6 h of treatment (P,0.02202; Figure 8d). Collectively, W1W2 showed significantly lower rates of water loss and chlorophyll efflux than the nonglaucous and other glaucous NILs, suggesting that cuticle permeability is not inversely proportional to wax load, but rather that it is closely related to wax composition, and that W1 and W2 function epistatically to reduce cuticular transpiration and chlorophyll leaching.

Discussion
Glaucousness is a classic genetic marker trait of wheat. Recent drought threat renewed interest in this trait. Despite several reports on the molecular mapping of Iw genes [53,54,55] and quantitative trait loci that affect glaucousness [67], this is the first systematic characterization of wheat glaucousness. In this study, we combined SEM, metabolite profiling, gene expression profiling, and physiologic approaches in a near isogenic background to gain insight into the genetic regulation of cuticular wax biosynthesis and its perturbation during drought tolerance.

Wax Composition and Glaucousness
Numerous cuticular wax genes were identified based on the lossor reduction-of-glaucousness phenotypes, such as the cer mutants of Arabidopsis. The CER genes participate in almost all steps of cuticle biosynthesis, indicating that glaucousness formation requires multiple wax components. Comparative analyses of the  Table S2. The bars represent standard deviation of the mean foldchange of mRNA levels. Asterisks indicate that the difference is significant at P,0.05 (*) or at P,0.01 (**). doi:10.1371/journal.pone.0054129.g005 b-Diketones, Glaucousness and Cuticle Permeability cuticular wax profiles of glaucous and nonglaucous durum lines and of wheat aneuploids found that b-diketone was critical for glaucousness formation [68,69]. Our research showed that wax composition is more important than wax load in glaucousness determination and that the nonglaucous NILs differ from the glaucous ones in terms of b-diketone content. Iw1iw2 and iw1Iw2 completely lacked b-diketones and w1w2 exhibited a ,90% reduction in b-diketones. These findings confirm that b-diketones are essential for glaucousness formation. SEM observation also confirmed that wax composition was directly related to wax morphology: b-diketones form wax tubes in the glaucous NILs and primary alcohols form wax chips in the Iw NILs. The glaucous NILs and w1w2 carried trace aldehydes and low levels of primary alcohols, suggesting that these wax species play a limited role in glaucousness formation.
Compared to W1w2 and w1W2, W1W2 showed slightly higher levels of glaucousness (Figure 1), but no difference in total bdiketone load. The major difference among the glaucous NILs is the presence of 8-and 9-hydroxyl b-diketones in W1W2 and the absence of these hydroxy isomers in W1w2 and w1W2. It is believed that hydroxyl-b-diketone is derived from b-diketone through mid-chain hydroxylation. In barley, this hydroxylation is defined by the cer-u mutants [70]. In the genus Triticum, 25-isomer is the typical hydroxyl-b-diketone in durum wax, but is substituted by 8-and 9-isomers in hexaploid wheat [71], suggesting that the generation of these substituted b-diketones involves the interaction of several genes. The absence of hydroxyl b-diketones in W1w2 and w1W2 suggests that the W1-W2 interaction is required for bdiketone hydroxylation and that these substituted b-diketones are important for intensifying glaucousness and forming wax crystallite sheets in W1W2 (Figure 2b and 2c).

Wax Pathways
During plant development, multiple wax pathways operate in parallel to produce different wax types [49,51,72], and wax composition changes as plants develop and grow. In the Triticeae, primary alcohols and alkanes are the predominant wax species at the seedling stage [61,66] and b-diketones are abundant at the reproductive stage, particularly in the sheaths and spike [48,61,70]. Genetic and biochemical studies in barley supported the existence of two parallel pathways, the acyl elongation, reduction, and decarbonylation pathway and the b-diketone pathway [50]. Mutations in CER-cqu of barley [51] and in the W genes of wheat ( Figure 3) specifically affect b-diketone synthesis. However, we know little about the cross-talk between these two pathways. In the present research, we identified three major types of waxes in a set of six NILs: alkane-rich wax in w1w2, alcohol-rich wax in the Iw NILs, and b-diketone-rich wax in the glaucous NILs (Figure 3c). A survey of Triticeae species found that many wild species produce alcohol-rich wax at the reproductive stage [73], implying that the Iw alleles prevail in nature and that the iw mutations led to the production of b-diketone-rich wax. The similar wax loads in W1W2 and iw1Iw2 (Figure 3a) suggest that precursors are shunted between the acyl reduction pathway and the b-diketone pathway by Iw2. Consistent with this, we found that expression of CER4-6 was elevated 120-fold in iw1Iw2. The wax composition in the Iw NILs at the adult stage (F10.5.1) resembles that at the seedling stage [48], being rich in primary alcohols and lacking b-diketones. Transcription quantification further demonstrated that CER4-6 is active in the seedling stage (F4.0) in W1W2, Iw1iw2, and iw1Iw2 (Figure 6a). Expression of CER4-6 was reduced in all three NILs at stage F9.0; however, the reduction was most pronounced in W1W2 (Figure 6b). Similarly, expression of four FAE genes showed no difference between iw1Iw2 and W1W2 in the seedling stage, but was reduced at the adult plant stage to a much greater extent in W1W2 than in iw1Iw2 ( Figure  S4). This suggests that the genes involved in the acyl elongation and reduction pathway are repressed by reproductive development and de-repressed by the Iw genes, which act as cross-talkers that regulate the acyl reduction and b-diketone pathway. Iw2 maintains an active CER4-6 at the reproductive stage and suppresses b-diketone production. It would be fascinating to determine how these two opposite activities are coupled by one gene. This is probably achieved by an interaction between Iw2 and different sets of genes. Molecular cloning of Iw2 and an in-depth analysis of CER4-6 expression will provide insight into the developmental regulation of the wax composition shift.  Table S3. The bars represent standard deviation of the mean fold-change of mRNA levels. Asterisks indicate that the difference is significant at P,0.05 (*) or at P,0.01 (**). doi:10.1371/journal.pone.0054129.g007 Much work has focused on the biosynthesis of b-diketones in barley by genetic and metabolic analyses of the cer mutants. Wax profiling found that cer-q, cer-c, and cer-u mutants define chain elongation, decarbonylation, and hydroxylation reactions in bdiketone synthesis, respectively [50,70]. However, genetic analyses of these mutants in barley support the hypothesis that b-ketoacyl elongation, decarbonylation, and hydroxylation are carried out by one gene, cer-cqu [50,51]. It is difficult to imagine how one gene might govern three different reactions. cer-cqu has been redesignated as gsh1 and was mapped to the subtelomeric region of the barley chromosome arm 2HS (http://wheat.pw.usda.gov/ ggpages/bgn/26/BGS351), which is colinear to the W1 locus in wheat [52]. In the present research, we found that the W1 and W2 genes in wheat are each sufficient for the deposition of b-diketone and complement each other, but that both are required for the synthesis of hydroxyl-b-diketones. One scenario would be that the W genes encode transcription factors belonging to the same family of proteins that activate the transcription of the b-diketone biosynthetic genes. We profiled 18 FAE genes; however, none of them were upregulated in the glaucous NILs. This is probably because b-diketones are synthesized by a separate enzyme system [50], but that the genes tested were chosen based on their function in model plants, which do not synthesize b-diketones. The proposed b-diketone biosynthetic pathway includes a decarbonyla- tion reaction [50]. We examined the expression of eight CER1 and five CER3 members of the decarbonylation pathway, and found that only the expression of CER1-8 was increased in w1W2 (2.6fold). To explore the possibility that MAH1 participates in bdiketone hydroxylation, we evaluated the expression of eight MAH1 members and found that MAH1-8 matched the production pattern of hydroxyl-b-diketones (Figures 4c and 7). Compared to W1W2, transcription of MAH1-8 was further elevated (,10-fold) in iw1Iw2, but hydroxyl-b-diketone was not detected, because Iw2 inhibits the biosynthesis of b-diketone, the substrate of hydroxyl-bdiketones.
The wheat W1 gene resembles the barley cer-cqu locus in terms of chromosomal location and regulation of b-diketone synthesis and hydroxylation. Considering the close phylogenetic relationship between barley and wheat, cer-cqu is probably orthologous to W1 and may also need to interact with other genes for b-diketone hydroxylation.

b-Diketones and Drought Tolerance
Previous physiological studies in wheat found that glaucousness significantly increased grain and biomass yield in irrigated and rainfed field experiments [74], and increased the photosynthesis to transpiration ratio and reduced the photosynthetic surface temperature in greenhouse experiments [58]. Glaucousness significantly reduced transpiration at night, which caused a relatively greater reduction in transpiration than photosynthesis and the increase of water-use efficiency [58]. In the present research, we measured cuticle permeability in terms of water loss and chlorophyll leaching in six NILs of four wax genes, and profiled their wax composition and inspected their wax morphologies. This allowed us to compare the effect of the individual genes and to analyze their interaction. We found that cuticle traits are closely associated with wax composition, mainly with respect to the bdiketones. The nonglaucous NILs had little or no b-diketones and showed significantly higher water-loss and chlorophyll-leaching rates (Figures 8a and 8b). A small but significant effect was also seen for other wax species. Although Iw1iw2 and iw1Iw2 had similar chlorophyll-leaching rates ( fig. 8b), the former had a higher water-loss rate (Figure 8a), probably because iw1Iw2 wax had a significantly higher content of primary alcohols, aldehydes, and alkanes ( Figure 3). Cuticle trait differences were also observed among the glaucous NILs. W1w2 and w1W2 had higher chlorophyll-leaching and water-loss rates than W1W2 (Figures 8c  and 8d). The most significant difference in the glaucous NILs in terms of wax composition is the presence of 8-and 9-hydroxyl-bdiketones in W1W2, suggesting a role of these hydroxyl isomers in reducing cuticle permeability and improving drought tolerance. Compared to b-diketone, the hydroxyl-b-diketones were much less abundant in the sheath wax. It is hard to imagine how hydroxyl-bdiketones reduce cuticle permeability. One explanation is that the addition of hydroxyl-b-diketones may underlie the changes in wax crystal organization of W1W2 (e.g., the formation of wax crystal sheets), which result in reduced water loss and chlorophyll efflux. We hypothesize that the hydroxyl-b-diketones function as the ''glue'' that cross-links the b-diketone tubes via the formation of hydrogen bonds between the hydroxyl and keto groups. Application of matrix-assisted laser desorption/ionization-MS imaging to the NILs and re-crystallization analysis of combinations of bdiketone and substituted b-diketone at different ratios may shed light on this possibility.
Glaucousness is an adaptive trait to dry cultivation conditions and will play an important role in developing cuticle-based strategies to improve drought tolerance. The Iw genes have a negative impact on drought tolerance and need to be eliminated from wheat breeding programs. Due to the dominant mode of inheritance, Iw-mediated nonglaucousness can be eradicated by one selection in early generations. The manipulation of the W genes is more challenging. Our results indicate that one W gene is sufficient to restore glaucousness, but not to prevent nonstomal transpiration. Therefore, both W genes are required for enhancing drought tolerance. In this respect, marker-assisted selection will help improve breeding efficiency. To this end, user-friendly molecular markers tightly linked to the W genes need to be developed. Molecular mapping and cloning of these W genes will open novel routes to manipulate cuticle permeability for drought tolerance.

Conclusions
In summary, characterization of a set of six NILs demonstrated that b-diketones contribute to glaucousness formation in the reproductive stage. A single W gene is sufficient for b-diketone synthesis, but both W1 and W2 are required for b-diketone hydroxylation. The Iw genes suppress b-diketone synthesis, but promote the production of aldehydes and primary alcohols. Consistent with the wax profiles, CER4-6 was de-repressed by Iw2, and MAH1-8 was activated by the interaction between W1 and W2. W1W2 showed the lowest cuticle permeability, suggesting that hydroxyl-b-diketones play a role in drought tolerance.

Plant Materials and Growth Conditions
One set of six NILs in the S-615 background was developed by Tsunewaki and Ebana [52] and the seeds were obtained from the corresponding author. The NILs were planted in 464 inch pots containing SunshineH Container Potting Mix 3 (Sun Gro Horticulture) supplemented with MulticoteH 8 Controlled-Release Fertilizer (Haifa) and grown in a greenhouse at a temperature of 20uC in the day and 18uC at night and with a day length of 16 h. Total genomic DNA was isolated from the NIL set using the Plant DNeasy Kit (Qiagen), following the manufacturer's instructions, and used for SNP and SSR genotyping. SNP genotyping was conducted commercially by Infinium Assay (Illumina, CA).

Microscopic Observation
Stoma counting and aperture observations were performed using an imprinted slide. Briefly, both sides of flag leaf sheaths were coated with 10% cellulose acetate dissolved in acetone using a paint brush. When dried, the cellulose acetate film (,261 cm) was carefully peeled, mounted on a slide, covered with a cover glass, and observed using a light microscope at a magnification of 10620. Two imprints were taken from each side of the sheath and three independent inspections were carried out on each of the imprints.
For SEM imaging of cuticle surfaces, a 0.5-cm tissue fragment was harvested from the base of the flag leaf, the uppermost part of the flag leaf sheath, and the peduncle, and glumes were collected from the spikelets in the middle of spikes at anthesis (F10.5.1). The tissue samples went through 10 glycerol gradients, from 10% to 100%, with 2 h in each glycerol solution, to replace the cellular water. The pretreated samples were sputtered with gold powder using the CrC-150 Sputtering System, and inspected with a Hitachi S-3400N SEM (Hitachi). Images were captured with the voltage set at 5 kV.

GC-MS Profiling of Wax Composition
Cuticular wax was extracted from two flag leaf sheaths of similar age from the same plant by submerging tissues in a glass tube containing 10 ml of HPLC grade chloroform (Fisher Scientific) and 4 mg of tetracosane (Sigma-Aldrich) as an internal standard and agitating manually for 1 min. The tissue was rinsed with an additional 5 ml of chloroform, and the two extracts were combined. The wax extract was filtered through a polytetrafluoroethylene filter with a 10-mL SGE syringe (Supelco) into a new glass tube and dried under a nitrogen stream. The dried wax was resuspended in 500 ml of acetonitrile and silylation was performed in 6% bis-trimethylsilyl-trifluoroacetamide and 10% trimethylchlorosilane at 80uC for 30 minutes to transform the hydroxyl and carboxylic groups into trimethylsilyl derivatives. The suspension was concentrated to 200 ml, and 1 ml was used for GC-MS analysis. Six biological replicates were included for each genotype. Wax silylation, GC-MS profiling, and substance identification were performed at the W.M. Keck Metabolomics Research Laboratory of Iowa State University (Ames, IA) on a fee-forservice basis.

Quantification of Cuticle Traits
To evaluate the effect of cuticular wax on water loss rate, flag leaf sheaths were excised from wheat plants at stage F10.5.1, dehydrated for 12 h at room temperature with a relative humidity of 44%, and weighed every hour using an analytical balance with a readability of 0.001 mg. The dry weight of tissues was measured after incubation at 37uC for 72 h. For the chlorophyll efflux assay, a flag leaf sheath was placed in a 50-ml tube containing 30 ml of 80% ethanol and the tube was gently agitated on a rotator at 50 rpm. A 150-ml aliquot of chlorophyll extract was transferred to a well in a microplate for quantification and then returned to the same tube. Absorbance was measured at 647 and 664 nm using a Synergy 2 Multi-Mode Microplate Reader (Biotek) and total chlorophyll micromoles were calculated as described [75]. Two measurements were adopted as technical replicates and four biological replicates were included for both experiments.

Transcription Quantification
Wheat cuticle gene homologs were identified by BLASTn searches of the wheat gene index database (http://compbio.dfci. harvard.edu/tgi) using known wax gene sequences as queries. The maximal cut-off value was set at E-20. qPCR primers were designed using the Primer3 program (http://frodo.wi.mit.edu/), with the PCR product size ranging from 80 to 200 bp (Table S1). Total RNA was isolated from flag leaf sheaths using TrizolH reagent (Invitrogen), following the manufacturer's instructions. After RNA integrity evaluation by agarose gel electrophoresis and quantification using Nanodrop ND-1000 (Thermo Scientific), 1 mg of total RNA was used for cDNA synthesis in a 20-ml reaction using the QuantiTect Rev Transcription Kit (Qiagen). After dilution, ,5 ng of cDNA was used as template for qPCR in a volume of 20 ml. qPCR was performed in 96-well plates with an ABI 7900HT High-Throughput Real-Time Thermocycler (Life Tech) using the iTaq TM Fast SYBRH Green Supermix with ROX (Bio-Rad). Two technical and four biological replicates were included for each NIL. The comparative DDCT method was used to evaluate the relative quantities of each amplified product using TaRPII36 as an internal reference in the same run. PCR specificity was determined by melt curve analysis of the amplified products.

Data Analysis and Statistics
Measured values from replicates were averaged and their standard deviations (SD) were estimated using MicrosoftH Excel functions. Student's t-test was performed using pooled SDs to evaluate the statistical significance of the differences among isogenic lines. The cut-off for statistical significance was set to a P-value of 0.05 or less.