| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Agroenvironmental Science and Technology, Viale Fanin 44, 40127 Bologna, Italy
* Corresponding author (roberto.tuberosa{at}unibo.it).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONConclusionsREFERENCES |
|---|
Abbreviations: ABA, abscisic acid • AB-QTL, advanced backcross quantitative trait locus • ASI, anthesis–silking interval • BC, backcross • BDL, backcrossed-derived line • CWP, cell wall protein • eQTL, quantitative trait loci that influences the level of expression • G x E, genotype x environment • L-ABA, leaf abscisic acid concentration • LD, linkage disequilibrium • MARS, marker-assisted recurrent selection • MAS, marker-assisted selection • MAYG, Mapping As You Go • NIH, near-isogenic hybrid • NIL, near-isogenic line • QTLs, quantitative trait loci • RIL, recombinant inbred line • SAGE, serial analysis of gene expression
Received for publication July 9, 2007.
Department of Agroenvironmental Science and Technology, Viale Fanin 44, 40127 Bologna, Italy
* Corresponding author (roberto.tuberosa{at}unibo.it).
Genome-wide approaches offer new, unprecedented opportunities to identify, clone, and manipulate the plethora of genes affecting drought tolerance in model species and crops. Compared to conventional breeding approaches, the dissection of the genetic basis of quantitative traits into their single components (i.e., quantitative trait loci [QTLs]) provides a more direct access to valuable genetic diversity of the morpho-physiological processes regulating the adaptive response to drought. This, in turn, enables us to utilize marker-assisted selection (MAS) for enhancing crops' performance. However, despite the impressive progress in molecular techniques and the large number of QTLs shown to influence yield in drought-stressed crops, the overall impact of MAS on the release of drought-tolerant cultivars has so far been marginal. It is foreseeable that QTL cloning will be facilitated by sequence information and the profiling of the transcriptome, proteome, and metabolome, all of which will improve the identification of plausible candidate genes. The cloning of major QTLs will offer additional opportunities for a more effective exploitation of the allelic richness present in germplasm collections. Allele mining in germplasm and mutant collections through forward- and reverse-genetics approaches, coupled with marker-assisted backcrossing and/or genetic engineering, will further expand the possibilities to introgress novel genetic variation in elite materials. New QTL-based modeling approaches, while improving our capacity to understand the genetic and molecular bases of genotype x environment interaction at varying water regimes, will contribute to singling out the most promising "molecular" ideotypes. This notwithstanding, a sizeable impact of MAS and other genomics approaches on the release of cultivars more resilient to drought will only be possible through (i) a deeper integration with conventional breeding methodologies, (ii) the capacity to accurately phenotype on a large scale, and (iii) a sound multidisciplinary knowledge of the biochemical and physiological processes determining crops' yield and its stability under a broad range of water regimes.
Abbreviations: ABA, abscisic acid • AB-QTL, advanced backcross quantitative trait locus • ASI, anthesis–silking interval • BC, backcross • BDL, backcrossed-derived line • CWP, cell wall protein • eQTL, quantitative trait loci that influences the level of expression • G x E, genotype x environment • L-ABA, leaf abscisic acid concentration • LD, linkage disequilibrium • MARS, marker-assisted recurrent selection • MAS, marker-assisted selection • MAYG, Mapping As You Go • NIH, near-isogenic hybrid • NIL, near-isogenic line • QTLs, quantitative trait loci • RIL, recombinant inbred line • SAGE, serial analysis of gene expression
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONConclusionsREFERENCES |
|---|
Although appropriate irrigation and/or other agronomic practices may to a certain extent mitigate the reduction in yield caused by drought (Bacon, 2004; Chaves and Oliveira, 2004; Kang and Zhang, 2004; Turner, 2004), their effects largely depend on the genetic make-up of the crop. In maize (Zea mays L.), a landmark review by Duvick (2005) has recently presented a critical assessment of the factors that have contributed to improve drought tolerance. Maize is now the third most important source of calories for humankind after rice (Oryza sativa L.) and wheat (Triticum aestivum L.). Due to the growing demand for dairy and meat products in developing countries and the decline in rice production in China and India, maize has been projected to become the most important crop by 2030. Future projections indicate that maize production will also face a reduction in irrigation volumes, even in regions where supplemental water is essential for securing a profitable harvest (Rosegrant et al., 2002). Based on these facts, Duvick (2005:135) indicated that "...breeders in regions of water shortage will need to increase their emphasis on breeding for drought tolerance, either by moving to direct selection or by increasing the emphasis on already existing direct selection for drought tolerance." He also speculated that "...evidence now accumulating indicates that these breeding efforts may someday be made even more efficient because of new insights provided by molecular biology investigations."
The linear increase in maize yield recorded during the past 70 yr indicates that conventional breeding has successfully improved maize performance under a broad range of environments, including those characterized by drought conditions. However, increasing financial resources are required to sustain this linear gain (Duvick and Cassman, 1999; Duvick, 2005). Yield improvements related to more upright leaves, tassels of smaller size, and lower grain protein content, all factors that have contributed during the past decades to sustaining the gains in maize yield and to improving yield stability under unfavorable conditions, might have gone as far as they can go because the traits themselves cannot be selected any further in the same direction (Duvick, 2005). Conversely, increasing attention should be devoted to the understanding and improvement of other morpho-physiological traits that have so far been neglected, for good reasons, by breeders. Clear examples include root architecture and size, efficiency of CO2 fixation, and other physiological traits that influence reproductive fertility (e.g., availability of photosynthates), particularly under water-limited conditions. The low heritability of these quantitative traits coupled with the intrinsic difficulty in their accurate phenotyping provide an ideal ground for deploying genome-wide approaches to better characterize their genetic basis and more effectively improve their value in elite germplasm.
Compared to conventional breeding, the advent of molecular markers has enabled us to dissect quantitative traits into their single genetic components (i.e., the quantitative trait loci [QTLs]) (Tanksley, 1993; Lee, 1995; Quarrie, 1996; Prioul et al., 1997; Bernardo, 2001; Tuberosa et al., 2002b; Morgante and Salamini, 2003; Parisseaux and Bernardo, 2004; Pelleschi et al., 2006), and to assist the selection and pyramiding of the beneficial QTL alleles through marker-assisted selection (MAS) (Ribaut et al., 2002a, 2004; Ribaut and Ragot, 2007). Importantly, MAS reduces or eliminates altogether the reliance on environmental conditions during the selection phase, a major hindrance to conventional breeding when dealing with traits highly influenced by drought (e.g., anthesis–silking interval [ASI] in maize; Ribaut et al., 2004). It has now been almost 20 yr since the first QTL study was documented in maize (Stuber et al., 1987). Despite the increasing number of studies that have striven to map QTLs in maize during the past two decades, surprisingly the number of drought-related studies has not increased (Fig. 1 ). More recently, bioinformatics (Bray, 2002) and the deluge of information generated by sequencing and postgenomics platforms (Hazen and Kay, 2003; Tuberosa and Salvi, 2006) have added new dimensions for deciphering the role and function of genes governing the response to drought. Despite all of these impressive technological breakthroughs, the overall impact of MAS and other genomics approaches on the release of drought-resilient cultivars has been marginal.
|
Adaptive vs. Constitutive QTLs
A critical factor for the success of any breeding program aimed at improving tolerance to drought is the possibility for accurately phenotyping the materials under conditions of managed stress levels. This is best achieved in nurseries located in a rain-free environment, which would allow the breeder to manage accurately and at will the irrigation volumes. This, in turn, will increase the heritability of yield under drought and will provide more accurate insights on the factors limiting yield at varying levels of water deprivation (Crosbie, 1982). This issue is particularly crucial for the identification of QTLs of traits categorized as purely adaptive (e.g., accumulation of osmolytes, abscisic acid [ABA], and/or other metabolites in response to cellular dehydration) as compared to constitutive traits (e.g., root elongation rate per se). Notwithstanding that all traits, even those constitutively expressed, show a certain degree of plasticity in response to environmental cues and thus influence crops' adaptation to drought (Blum, 2002, 2005), the ranking of the genotypes of a mapping population scored for a constitutive trait under different water regimes will be little affected by environmental factors. As compared to QTLs characterized by a strong adaptive component, QTLs expressed in a prevailingly constitutive fashion will thus be easier to characterize and more predictable as to the outcome of MAS. Hence, a major advantage of constitutive traits is that their phenotyping does not require conditions of programmed drought stress as needed for adaptive traits. Additionally, the outcome of MAS will on average be more predictable for constitutively expressed traits. Unfortunately, the majority of QTLs in trials conducted under varying water regimes and/or during different seasons characterized by different rainfall patterns show a high QTL x environment interaction (Pelleschi et al., 1999, 2006; Ribaut et al., 2002a; Tuberosa et al., 2002c).
Regardless of the mode of expression (adaptive vs. constitutive) of the gene or genes underlying a drought-related QTL, the unpredictability of the outcome that MAS for a morpho-physiological trait might have on grain yield is to a large extent related to the factors that limit yield according to the dynamics (i.e., timing, frequency, and/or intensity) of the drought episodes prevailing in the target environment. Further complexity is added by the presence of other abiotic stresses (e.g., heat, salinity, aluminum) which might amplify the adverse effects of water deficit. As an example, a high concentration of aluminum in the soil causes poor root development, thus making the plant more vulnerable to drought. Similarly, widespread damage to the root system caused by infestation of root worms will substantially increase the negative effects of drought on the final yield of maize hybrids more susceptible to this pest.
Experimental evidence from trials in which the same maize mapping population was tested under different water regimes indicates that most QTLs for grain yield and drought-related traits are stress adaptive, hence detected prevailingly under water-limited conditions (Ribaut et al., 1997; Tuberosa et al., 2002c; Pelleschi et al., 2006), a result that is in keeping with the sizeable genotype x environment (G x E) interaction usually detected for the same traits when phenotyped across a wide range of water regimes. When 120 recombinant inbred lines (RILs) of the Io x F2 population were tested under well-watered and drought conditions, only 14% of the 84 QTLs were common to both treatments, thus indicating that different sets of specific loci influenced phenotypic variability amongst RILs subjected to the two water regimes (Pelleschi et al., 2006). These findings are somehow counterintuitive when compared to the widespread notion that modern breeding has improved maize yield across a broad range of water regimes (Duvick, 2005). These apparent contradictions can be in part reconciled considering that mapping populations have mainly been derived from nonelite lines which, as compared to elite materials, likely have greater allelic diversity at drought-adaptive QTLs. Conversely, elite lines used for commercial purposes have probably undergone a very strong selection for the best possible alleles at such loci. A detailed haplotype analysis of the drought-adaptive QTL regions performed in the historical series of the Era Hybrids developed in the past 70 yr (Duvick, 2005) would provide useful insight in this direction.
Optimizing the Quest for Drought-Related QTLs
The first critical step is to identify genes or QTLs playing a major role in governing genetic variability for tolerance to drought. In this context, priority should be given to QTLs characterized by a limited interaction with the water-regime, or other environmental variables, and by a consistent, associated effect on yield (Vargas et al., 2006). The second, and perhaps more critical step from an applicative perspective, is to verify to what extent the effect of the beneficial QTL allele is consistent in the genetic backgrounds to be improved.
A major shortcoming of any QTL study is the low accuracy in detecting the real number of QTLs affecting the genetic variation of the investigated traits (Xu, 2003). In a simulation study applied to experimental data, Beavis (1994) clearly showed that with populations of approximately 100 to 200 progenies, only a modest fraction of QTLs were identified; furthermore, the effect of each single QTL was usually overestimated. Another study showed that with fewer than 500 progenies, and independently from marker density, it is very difficult to detect QTLs of small effect (Beavis, 1998). These predictions were substantiated in experiments performed with maize mapping populations obtained from elite materials sufficiently large (>400 progenies) to allow for meaningful subsamplings (Openshaw and Frascaroli, 1997; Melchinger et al., 1998; Schön et al., 2004). Additionally, most of the current methods of QTL analysis, in terms of experimental design, population dimension, and statistical approach, are inadequate for an effective detection of epistatic QTL interactions, a strong source of variation for traits that affect tolerance to drought (Yue et al., 2006). Given the potential effect of epistasis on response to selection (Wade, 2002), empirical investigations to quantify its importance for trait phenotypes are an important component of the design and optimization of any MAS strategy (Podlich et al., 2004). Statistical methods to detect epistatic QTLs that have been proposed (Kao and Zeng, 2002; Yi and Xu, 2002; Stich et al., 2006) are waiting for field-experiment validation.
With maize as a model species, computer simulation performed by Bernardo (2001) suggested that gene information is most useful in selection when few loci (e.g., 
10) control the trait. With many loci (e.g., >50), the least squares estimates of gene effects become imprecise. Overall, the results of the simulation indicated that the typical reductionist approach pursued through QTL discovery has only limited potential for enhancing selection for quantitative traits in hybrid crops (Bernardo, 2001).
One reason for the limited applicative impact of the QTL approach is because the choice of the parental lines used for QTL discovery has prevalently disregarded their agronomic value, while being mainly based on the differences for the target traits. While this approach maximizes the opportunities for identifying major QTLs, it does not guarantee any real progress when the beneficial QTL alleles are introgressed via MAS, because such alleles, or those with even more beneficial effects, may have already been fixed in the elite germplasm.
A problem frequently encountered in correctly interpreting QTL effects under drought conditions in the field is due to the confounding effect that differences in the water status of the plants may have on the value of the investigated traits. Therefore, a correct interpretation of QTL effects on morpho-physiological traits influenced by water status (e.g., concentration of ABA and osmolytes) and, more importantly, the interpretation of their associated effects on yield, require careful consideration of this aspect. This is easier said than done, since an accurate measurement of the water status in the large number of plants typically required by a QTL study is a rather daunting task. For this category of traits, a more accurate evaluation of a set of genotypes can be obtained under controlled conditions that allow for a better control of daily fluctuations in the water status of the plants (Yue et al., 2006; Jones, 2007). Additionally, exposure of plants to a given concentration of osmolytes (e.g., polyethylene glycol) provides an opportunity to evaluate "constitutive" traits whose value is influenced by the water status. As an example, the capacity to accumulate ABA in response to a water deficit is best measured under such controlled conditions (Sanguineti et al., 2006).
The small plot size usually adopted to evaluate mapping populations could represent an additional source of inaccurate estimates of QTL effects. A small plot size exacerbates competition for water, nutrients, and light between adjacent plots, which may lead to overestimating the effect on yield of the QTLs for morpho-physiological traits (e.g., early vigor, root architecture, leaf angle, anthesis date) affecting the level of competition and/or affected by such competition. In this case, even if a QTL is consistently detected in several field trials conducted with a mapping population, once the beneficial QTL allele is introgressed and evaluated in larger plots of genetically homogeneous material, its overall effects on yield may disappear or even become negative. As an example, although a QTL allele for a more vigorous root mass might provide a yield advantage in a mapping population segregating for such QTL alleles, when fixed in homozygosity it could prove detrimental at the high plant densities commonly adopted by maize farmers.
In view of all the above-mentioned points, the negligible impact that MAS has made so far on the release of drought-tolerant hybrids should thus not be a surprise.
Target Traits for Improving Drought Tolerance in Maize
Yield improvement through conventional approaches has been mainly achieved through a direct selection for yield, with limited or no knowledge of the morpho-physiological determinants of drought tolerance (Edmeades et al., 1997; Duvick, 2005). A careful choice of the target traits and a critical analysis of the literature will facilitate the identification of major QTLs characterized by a more consistent effect across water regimes. Indeed, the widespread and accepted notion that drought is inevitably a complex trait has been challenged. According to Blum (2002, 2005), the complexity of drought tolerance can be greatly reduced if two major points are considered: (i) a number of plant traits crucial for the control of plant water status and yield under drought (e.g., root depth, root mass) are to a large extent constitutively expressed; (ii) plant water status, more than plant function, controls crops' performance under drought. It is only under rather severe drought conditions threatening the survival of the crop (i.e., conditions that are only rarely present in most farmers' fields) that the targeted manipulation of adaptively expressed genes or QTLs of molecular pathways specifically regulated under such extreme conditions will improve the fine-tuning of crops' survival.
The limited contribution that molecular approaches (including genetic engineering) have so far provided in improving drought resistance somehow relates also to the difficulty in identifying the key determinants of yield under drought conditions in the field (Blum, 1988, 2005; Ludlow and Muchow, 1990; Boyer, 1996; Turner, 1997; Passioura, 2002; Nguyen and Blum, 2004; Tuberosa, 2004) as also illustrated in the previous section. As an example, a greater capacity of the root meristem to adjust osmotically at a given water potential will have a positive impact on final yield only if deeper roots can extract additional moisture from the soil. Conversely, when moisture is unavailable in deep soil layers, growing deeper roots might negatively influence final yield due to an excessive partitioning of photosynthates to the root. Additionally, an excessively vigorous root mass will inevitably entail a higher metabolic cost, hence reducing the amount of photosynthates available for the reproductive organs. Roots show a high degree of plasticity in terms of response to environmental factors, especially to the availability of water and nutrients. To a varying degree, this plasticity is under genetic control (O'Toole and Bland, 1987; Landi et al., 2002; Tuberosa et al., 2002c, 2003), with each genotype being characterized by its own response to different environmental cues. This is the case for maize and most cereals where the phasic development of adventitious roots (Hochholdinger et al., 2004) determines the root system profile and water extraction from the soil (Blum and Arkin, 1984).
In maize, flowering is the most crucial stage in terms of negative effects of drought on yield. During this stage, one single day of drought can potentially decrease yield up to 8% (Shaw, 1977). Therefore, due to the high frequency of drought episodes at flowering, reproductive failure is the most important drought-related factor contributing to yield losses in maize. The work performed by Boyer and coworkers (Boyle et al., 1991; Boyer, 1996; Boyer and Westgate, 2004; McLaughlin and Boyer, 2004) as well as by Zinselmeier and coworkers (Zinselmeier et al., 1995, 1999, 2002) has clearly demonstrated the key role of sucrose level on kernel abortion. The importance of flowering time in terms of yield reduction under drought conditions is further substantiated by the higher correlation of grain yield with kernel number per plant rather than kernel weight (Duvick, 2005; Landi et al., 2005, 2007). Biochemical factors other than sucrose supply appear to be involved in kernel abortion during the early stages of kernel growth (Boyle et al., 1991; Zinselmeier et al., 2002; McLaughlin and Boyer, 2004; Boyer and McLaughlin, 2007). The growth regulator ABA has been implicated as one of such factors that, either per se or through more complex interactions, may allow the plant to gauge the availability of environmental factors (e.g., water) with a key role in determining the final reproductive potential. Interestingly, the evaluation of the historical series of the Era Hybrids has indicated a significant decrease in the capacity to accumulate ABA of modern maize hybrids when exposed to a given level of water stress (Sanguineti et al., 2006).
As illustrated and discussed in the remaining paragraphs, genomics and postgenomics approaches offer new opportunities to unravel the physiological and genetic basis of the factors curtailing maize yield under such adverse conditions and to harness beneficial allelic variation.
Harnessing Allelic Diversity
Domestication has drastically restricted the level of genetic variability present in crops compared to their wild counterparts (Tanksley and McCouch, 1997), an aspect particularly relevant for those drought-related traits that play a substantial role in survival of the plant under natural conditions. This limitation can be overcome through advanced backcross QTL (AB-QTL) analysis, an approach that offers the opportunity to quickly discover and exploit beneficial QTL alleles present in wild germplasm (Tanksley and Nelson, 1996). The AB-QTL approach relies on the evaluation of backcross (BC) families between an elite variety used as recurrent parent and a donor accession, usually a wild species sexually compatible with the crop. Preferably, QTL analysis is delayed until the BC2 generation and after selecting in BC1 against characteristics with a negative effect on yield (e.g., ear shattering in small-grain cereals, tillering in maize). The AB-QTL approach appears particularly valuable for the identification of beneficial wild alleles improving plant survival under extreme drought. Ideally, introgression of such beneficial alleles should bear no negative consequences under more favorable conditions, a circumstance that might only seldom occur when the donor is a wild progenitor.
Ample experimental evidence in a number of crop species underlines the validity of the AB-QTL approach based on the exploitation of wild relatives (Talamè et al., 2004; Grandillo and Tanksley, 2005). In maize AB-QTL analysis has been extended to crosses between elite, heterotic inbreds (Ho et al., 2002). If high-order epistasis plays a major role in hybrid maize performance, MAS in advanced backcross progeny may be superior to such selection in the F2. As compared to an F2 population, favorable epistatic gene combinations accumulated by traditional breeding are less likely to be disrupted in an advanced backcross population (Allard, 1996). Epistasis has been established as a major mechanism of adaptation in various plant species (Allard, 1996), particularly in rice (Yu et al., 1997; Li et al., 2001; Yue et al., 2006). The work of Ho et al. (2002) identified four QTL regions where the donor parent allele significantly improved grain yield, hence demonstrating that AB-QTL analysis can be deployed to identify and manipulate useful QTLs in heterotic inbreds of elite maize. The production of introgression libraries of near-isogenic lines (NILs) obtained following more extensive backcrossing (5–6 cycles) further extends the concept of AB-QTL analysis and provides interesting opportunities for a detailed analysis of QTL effects on target traits (Salvi et al., 2004).
QTL Characterization and Validation
The accurate characterization of a QTL requires its isogenization, usually performed through MAS (Landi et al., 2005). The evaluation of testcrosses obtained by crossing a number of tester lines with pairs of NILs contrasted at a QTL region allows us to more accurately determine the consistency of the QTL effect and its applicative value (Landi et al., 2007). Although the derivation of congenic strains through MAS does not lead to short-term applications, it is an essential step required to "Mendelize" single QTLs and for their positional cloning.
Near-isogenic lines contrasted for the parental chromosome regions at the target QTL can be obtained through repeated selfings (at least 5–6) of one or more individuals heterozygous at the QTL region followed by the identification of the homozygotes for each one of the two parental segments. Alternatively, the parental lines of the original mapping population evaluated for QTL discovery can be used as recurrent parent in a backcross scheme in which a single plant heterozygous at the QTL in question is utilized as donor of the alternative QTL regions; in this case, the congenic lines are identified as backcrossed-derived lines (BDLs). For a cross-pollinated species like maize, the evaluation of the effects of a particular QTL on yield or other highly heterotic traits should preferably be performed using near-isogenic hybrids (NIHs), which can be obtained by crossing pairs of BDLs homozygous at the same target region but different as to the genetic backgrounds (Landi et al., 2005). Depending on the BDLs used as parents, NIHs are either homozygous or heterozygous at the target QTL region, while being heterozygous for most of the remaining portion of the genome.
In maize, MAS has been applied to derive pairs of BDLs differing for the parental alleles (Os420 and IABO78) at a major QTL (root-ABA1) on bin 2.04 near csu133 that was originally shown to consistently affect leaf ABA concentration (L-ABA; Tuberosa et al., 1998; Sanguineti et al., 1999) and, following its isogenization, also root architecture and other drought-related traits (Giuliani et al., 2005b; Landi et al., 2005, 2007). Figure 2 reports the effects of this QTL on root architecture in plants at the flowering stage. Interestingly, similar experimental evidence on the influence of the region near csu133 on root architecture and L-ABA was gathered also from Polj17 x F2 (Lebreton et al., 1995). A field evaluation conducted under well-watered and water-stressed conditions during two consecutive seasons indicated that each pair of root-ABA1 BDLs differed significantly and markedly for L-ABA, thus confirming the effectiveness of MAS (Landi et al., 2005). Furthermore, the BDLs and derived NIHs showed significant differences for root lodging, root mass and brace root angle, and grain yield (Giuliani et al., 2005b; Landi et al., 2005), supporting the hypothesis that some of the QTLs for L-ABA may derive from primary QTL effects on root architecture as originally postulated by Tuberosa et al. (1998). These results suggest a primary effect of the QTL on root architecture followed by a secondary effect on the ABA accumulated in the leaf because of an effect on the amount of xylem ABA flowing from the roots in the more superficial soil layers (usually subjected to quick dehydration), rather than an effect on the water status of the leaf.
|
|
The first large-scale attempt to apply MAS to ameliorate drought tolerance in maize occurred at CIMMYT (Mexico), where in 1994 two parallel projects were initiated to improve the performance of elite lines and open-pollinated populations. In this case, MAS was utilized to introgress QTL alleles for reducing the ASI, that is, the interval between the extrusion of the anthers and silks. Under conditions of water deficit, ASI is negatively associated with grain yield (Bolaños and Edmeades, 1996; Ribaut et al., 1997, 2002a). The availability of molecular markers linked to the QTLs for ASI allows for a more effective selection under drought as well as when drought fails to occur at flowering (Ribaut et al., 2002a, 2004). Using the line CML247 as the recurrent parent and Ac7643 as the drought-tolerant donor, Ribaut and coworkers started a backcross-MAS (BC-MAS) project based on the manipulation of five QTLs affecting ASI. CML247, an elite line with high yield per se under well-watered conditions, is drought susceptible and shows long ASI under drought. The QTL regions carrying alleles for short ASI were introgressed through MAS from Ac7643 into CML247. A number of lines (ca. 70) derived through BC-MAS were crossed with two testers and were evaluated for three consecutive years under several water regimes. Under severe stress conditions that reduced yield by at least 80%, the selected lines were superior to the unselected control. However, this advantage decreased at lower stress intensity, and disappeared for a stress that reduced yield less than 40%. Across the water-limited trials, a few genotypes consistently outperformed the controls. Under well-watered conditions the selected lines did not show any yield penalty when compared to the control lines. Notwithstanding the success of this BC-MAS experiment, Ribaut et al. (2002b) pointed out that QTL manipulation to improve germplasm for polygenic traits has a number of limitations, the most distinct being the inability to predict the phenotype of any given genotype based on its allelic composition, a constraint particularly important when epistatic interactions influence the target trait (Podlich et al., 2004).
A refinement of MAS is offered by marker-assisted recurrent selection (MARS), a scheme based on successive generations of crossing individuals based on their molecular profile, the goal being to attain a somehow ideal genotype at the different target QTL region (Peleman and van der Voort, 2003). When compared to BC-MAS, MARS allows for the selection of additional favorable alleles besides those targeted by BC-MAS. As an example, when the goal is to improve drought tolerance, MARS would enable the breeder to select for favorable alleles at yield QTLs in addition to the QTL alleles targeted for enhancing drought tolerance. Resources required for MARS would probably be somewhat more than those required by a partial backcross approach, although the difference might not be very large if background markers are used to select for recurrent parent recovery in the BC-MAS approach. Simulation studies have shown that the relative advantage of MARS over phenotypic selection vanishes rapidly when the portion of the total genotypic variance explained by the selected QTLs decreases (Van Berloo and Stam, 2001). As to the applications of MARS, while Moreau et al. (2004) and Openshaw and Frascaroli (1997) showed limited advantage for MARS when compared to conventional selection, others (Ragot et al., 2000; Johnson, 2004; Eathington, 2005; Crosbie et al., 2006) have reported more successful applications of MARS in maize breeding programs.
A major factor that will greatly affect the extent to which MAS will be more routinely exploited in breeding programs relates to its cost-effectiveness as compared to conventional breeding practices (Ribaut and Betran, 1999; Morris et al., 2003) The application of high-throughput genotyping platforms based on the scoring of markers that do not require the use of gels (Salvi et al., 2001; Hardenbol et al., 2003; West et al., 2006) coupled with quick DNA extraction protocols are needed to streamline MAS and make it more cost-effective and widely applicable.
Cloning QTLs for Traits Affecting Drought Tolerance
The identification of major QTLs should be viewed as the first step of a longer process aimed at identifying and isolating the underlying molecular cause at the sequence level (i.e., polymorphism) of the functional variation revealed by QTL analysis. The cloning of the sequence underpinning the QTL will allow us to characterize and manipulate more precisely the available germplasm for the type of alleles present at the target QTL (Salvi and Tuberosa, 2005). QTL cloning should be viewed as an ideal entry point toward a more effective and informed exploitation of sequence variability at selected loci and to unlock the allelic richness present in germplasm collections (e.g., by means of EcoTILLING; Stemple, 2004). Clearly, the number of drought-related QTLs suitable for a positional cloning approach is very limited, and such QTLs are and will remain the exception rather than the rule. A number of options are available to clone QTLs. While positional cloning has been the procedure most widely adopted, more recently association mapping has received growing attention.
Positional Cloning
Following the identification of a major QTL, its positional cloning typically relies on the production of a large secondary cross in a nearly isogenic background where only the target QTL segregates. The major drawback of this approach is the long time required to obtain the NILs. The large number of plants (usually >1000) in the segregating population obtained from the cross of the NILs allows for the recovery of a sufficiently high number of recombination events in the target region, an essential prerequisite to achieve the desired level of map resolution required to more accurately map the QTL compared to the primary analysis conducted before the isogenization of the QTL. Recently, the notion that QTL peaks have low mapping accuracy has been challenged: Price (2006) compared the position of the gene or sequence underlining each one of 20 cloned QTLs with that of the LOD peak of the QTL in the original primary analysis conducted before the isogenization of the QTL. Most of the primary analyses were conducted with populations with <200 families and even with <100 families. Interestingly, the causative gene/sequence was in most cases within 1 to 2 cM from the map position of the original LOD peak. These results indicate that the primary mapping of a major QTL might be sufficiently accurate to avoid the time-consuming procedure of its fine mapping before the anchoring of the genetic map with the physical map to identify suitable candidates. Future cloning efforts based on this approach will verify its suitability for speeding up QTL cloning.
Another important prerequisite for the positional cloning of a QTL is the availability of markers in the target region. Bulk segregant analysis (Salvi et al., 2002b) and comparative mapping based on synteny with model species and related crops (Paterson et al., 1996; Bennetzen and Ma, 2003; Sorrells et al., 2003) are valuable sources of additional markers. Microarray analysis can provide clues toward the identification of polymorphisms influencing the level of expression of genes in the target region and possibly, of the gene or genes underlying the QTL (Hazen and Kay, 2003). Following the fine mapping, the markers more tightly linked to the QTL are used to anchor the genetic region to a physical chromosome region. Once the maize genome sequence becomes available, the anchoring will be extended to the complete genome sequence, thus making positional cloning of Mendelian loci more of a routine practice and strongly facilitating QTL cloning (Borewitz and Chory, 2004). An element of complexity inherent to the organization of the maize genome that should be duly considered for its practical implications is the lack of colinearity between the sequence of different lines (Fu and Dooner, 2002). In the meantime, due to the fact that the genome sequence is not yet available, contigued genomic libraries (prevalently based on BAC clones) are screened to anchor specific markers onto the physical map. Polymorphic genes or genomic sequences that are found to be completely linked with the QTL are then functionally tested with a number of different approaches (e.g., genetic engineering, identification of knockouts). Other procedures such as RNAi (Waterhouse and Helliwell, 2003; Kusaba, 2004) and TILLING (Targeted Induced Local Lesions IN Genomes; McCallum et al., 2000; Till et al., 2004) allow for a genome-wide functional screening. The TILLING platform that has been established in maize (see http://genome.purdue.edu/maizetilling/) represents a valuable tool to validate the role of drought-related candidates genes.
Association Mapping
The discovery and/or cloning of QTLs through association mapping relies on the molecular profiling and phenotypic characterization of unrelated accessions. The analysis evaluates the difference in allele frequency in case–control samples, or, preferably when dealing with complex traits, the change in the mean of the investigated traits caused by allele substitution (Flint-Garcia et al., 2003). In plants, the interest lies in the possibility of performing QTL analysis and cloning without the costly and time-consuming production of large experimental populations (Breseghello and Sorrells, 2006; Ersoz et al., 2007). Strategies of association mapping are strongly influenced by the level of linkage disequilibrium (LD) present in the target population. Populations characterized by a higher level of LD are more suitable for genome-wide gene or QTL discovery, particularly when the panel of accessions has been profiled at a limited number of loci. This condition is typical of elite collections of autogamous crops (Maccaferri et al., 2005). In maize, even with elite materials the level of LD decays quickly (Rafalski and Morgante, 2004); this finding implies that a rather large number of markers is required to identify a significant association. The validation of the role of a candidate gene or sequence, requires the utilization of panels with a very low level of LD, which in turn allows for a high level of genetic resolution. A major point to consider is that a particular QTL may be underlined by a noncoding sequence (Clark et al., 2006; Konishi et al., 2006), in some cases quite far (up to 70 kb) from the effector gene modulating the QTL effect, as recently shown in maize for Vgt1 (Vegetative to generative transition 1) (Salvi et al., 2007), a QTL for flowering time located on bin 8.05 (Vladutu et al., 1999; Salvi et al., 2002a, 2002b).
Both positional cloning and/or association mapping seek an association between polymorphisms at marker loci and variability in the target quantitative trait and exploit LD to identify the most promising candidate gene or genes for the subsequent validation phase. Functional maps assembled with drought-induced genes provide valuable clues for the identification of candidate genes for drought-related QTLs (Andersen and Lubberstedt, 2003). Specific efforts toward enriching linkage maps with function-specific genes have also been undertaken in maize to identify QTL candidates. A large-scale effort has been recently completed by Davis and coworkers (2006): over 6000 root-specific expressed sequence tags have been mapped in silico on the consensus map. This map provides an excellent framework for investigating the role of root-specific genes in controlling variability in root growth rate and architecture under different water regimes.
Among the QTLs cloned in plants (for a review see Salvi and Tuberosa, 2005), those controlling flowering time are relevant for the release of early cultivars that, due to their earlier flowering, are able to escape or at least limit the negative consequences of summer drought on reproductive fertility. However, it should be noted that flowering time, due to its high heritability, can be manipulated very effectively through conventional breeding (Hallauer and Miranda Fo, 1989). In maize, although several QTLs for flowering time have been mapped (reviewed in Chardon et al., 2004), their molecular identity is mostly unknown. One maize flowering time gene, id1, has been cloned following its identification as mutant (Colasanti et al., 1998); genetic polymorphisms at a second gene, Dwarf8, have been correlated with quantitative effects on flowering time (Thornsberry et al., 2001). Recently, Salvi et al. (2007) have shown that allelic variation responsible for the maize flowering time QTL Vgt1 is confined to an apparently noncoding region located approximately 70 kb upstream of a transcription factor gene belonging to the AP2 family known to regulate flowering time in Arabidopsis (Aukerman and Sakai, 2003). Additionally, Vgt1 is centered at a sequence polymorphism caused by a transposon insertion that disrupts a maize–rice conserved noncoding sequence (Kaplinsky et al., 2002). The chromosome region corresponding to bin 8.05 has been recognized as a "hotspot" for flowering time QTLs and genes (Chardon et al., 2004, 2005) responsible for an important part of variation for flowering time in maize. Vgt1 adds to the scanty and poorly characterized list of plant long-range enhancers (Stam et al., 2002; Clark et al., 2006; Konishi et al., 2006). The observation that three out of four long-range enhancers characterized so far in plants underlie QTLs could indicate that this type of regulatory control is particularly relevant for quantitative natural genetic variation. As recognized by Rafalski and Morgante (2004), the identification of regulatory regions often quite distant from the effector genes indicates that the selection of a candidate sequence to be tested for association mapping with a phenotype is not a trivial undertaking if the genomic scan aims to be comprehensive.
A few more QTLs with a putative role in the adaptive response of maize to drought have been functionally linked to specific candidate genes (Pelleschi et al., 1999, 2006). Additionally, efforts are underway in maize to positionally clone root-ABA1, a QTL on bin 2.04 which affects root architecture and leaf ABA concentration (Giuliani et al., 2005b; Landi et al., 2005). With the only exception of flowering time, the major hindrance to the cloning of major QTLs for most drought-related traits is due to their usually low to moderate heritability which requires an adequately large number of replicates to obtain accurate estimates of the phenotypic value of each segmental isoline.
Postgenomics Approaches
The next logical step following the identification of a major QTL is to identify the most suitable candidate sequence, validate its role, and proceed accordingly with a more direct manipulation of the target trait. The identification of candidate genes and the elucidation of their role can be greatly facilitated by combining QTL analysis with different sources of information and technological platforms (Pflieger et al., 2001; Tuberosa et al., 2002a; Wayne and McIntyre, 2002; Sharp et al., 2004). In this respect, the recent progress in high-throughput profiling of the transcriptome, proteome, and metabolome enables us to investigate the concerted expression of thousands of genes and measure the level of their products.
Transcriptomics
High-throughput profiling of mRNAs has been applied to investigate the changes in gene expression in response to dehydration in maize (Zinselmeier et al., 2002; Yu and Setter, 2003; Sharp et al., 2004; Poroyko et al., 2007). Additionally, it should be noted that the functional basis of a number of cloned plant QTLs relates to differences in the level of expression (Salvi and Tuberosa, 2005). Therefore, QTL discovery may in some cases be possible through a direct mRNA profiling approach (Hazen and Kay, 2003), such as in the identification of the so-called eQTLs, namely QTLs that influence the level of expression (hence the "e") of a particular gene. In this case, the analysis of the level of gene expression performed on each progeny of a mapping population will identify eQTLs influencing the observed variability among progenies in mRNA level of the profiled genes. Circumstantial evidence regarding the importance of each open reading frame in governing variability for yield under conditions of drought can be obtained by comparing the map position of QTLs for yield with the map position of the open reading frame itself and the corresponding eQTLs. Although plant eQTLs were first reported in maize (Schadt et al., 2003) to date this approach has not been applied to identify eQTLs for drought-related traits. Clearly, the cost associated with the profiling of the large number of RNA samples required to identify eQTLs is still too high to conceive a more routine application of this approach. Instead, transcriptome profiling is better suited for studies involving a limited number of samples extracted from congenic strains differing at key genomic regions (e.g., NILs) and/or bulked samples obtained from the opposite tails of mapping populations. Besides providing clues for the identification of candidate genes, transcriptome profiling of congenic
QTL strains offers the opportunity of enriching the region with additional functional markers (for an example in maize see Giuliani et al., 2005a), provided that the different levels of expression of the investigated genes are caused by a sequence polymorphism that can be utilized for producing a useful marker. Under this respect, single nucleotide polymorphism markers are particularly valuable because their profiling can be automated.
The interpretation of the results obtained from profiling experiments performed under controlled environments should duly consider the experimental conditions utilized to cause the water loss in the targeted plant tissue (Talamè et al., 2007). In a number of studies, a rather severe dehydration has been artificially imposed in a very short time, commonly in a few hours (reviewed in Hazen et al., 2003). These experimental conditions will certainly be more shocking and damaging at the cellular level if compared to similar intensities of water deficit that plant tissues usually experience in the field, where dehydration unfolds over a prolonged time (commonly days or weeks), thus allowing for a more proper activation of those molecular mechanisms leading to long-term adaptive responses (e.g., osmotic adjustment, early flowering, thickening of leaf cuticles) that allow the plant to counteract the negative effects of drought. Consequently, molecular results obtained under artificial conditions should be considered with caution and duly validated before their utilization in more applicative terms. Another factor that may greatly reduce the effectiveness of profiling experiments to capture the key events triggering important adaptive responses is the timing of the sampling in relation to the dynamic of the drought episode or episodes. In this respect, Boyer and Westgate (2004) have indicated the difficulty in reconciling the results on the role of invertase activities in ovary abortion in maize obtained with microarrays (Zinselmeier et al., 2002) with the pivotal role played by assimilate supply in preventing maize ovaries from aborting under drought conditions (Boyle et al., 1991; Zinselmeier et al., 1995, 1999; McLaughlin and Boyer, 2004).
A remarkable example on how transcriptome analysis can advance our understanding of the physiology underpinning traits which play an important role under conditions of water deficit is provided by the recent work on the expression profiling of primary root apices in maize seedlings (Sharp et al., 2004; Poroyko et al., 2007). This research has taken advantage of a kinematic approach based on a detailed and refined study of the spatial and temporal patterns of cell expansion within the growth zone of the root apex, which has highlighted an important role of ABA and ethylene in sustaining growth of the root meristem (reviewed in Sharp et al., 2004). Based on the extensive work conducted by Sharp and coworkers (Spollen et al., 2000; Sharp et al., 2004), four regions of the maize primary root tip have been identified based on their differential elongation rates under well-watered and water-stressed conditions. The spatial distinction of regions that showed cell expansion under both conditions from regions with expansion growth only in well-watered roots provided information to begin an analysis of transcriptional events that influenced root growth under water stress in a region-specific manner.
By generating cDNA libraries from the four regions of the maize root tip, Poroyko et al. (2005) were able to detect more than 6500 expressed genes. Spatial resolution and the distribution of transcripts of different nature highlight expression profiles and changes that cannot be obtained by an analysis of the entire root. The comparative analysis of more than 6500 unigenes in different regions of the primary root tip showed a distribution of transcripts in different categories that were partly expected, such as the large amount of transcripts for functions in cell cycle, cell division, and chromatin structure that were present in the most distal segment which includes the quiescent center (Poroyko et al., 2007). In addition, functions in cell expansion, cell wall biosynthesis, cell maturation, and hormone biosynthesis were assigned to particular regions. However, a novel and unexpected finding was the wide range of transcript steady-state levels in different regions. Many transcripts were detected with more than a 10-fold difference in expression intensity in cells separated by a short distance along the root tip. This result underlines the importance of a very detailed spatial analysis of the transcriptome when different cell types are involved. A similar approach to defining genes active in the elongation zone of the maize primary root has been reported (Bassani et al., 2004). Using suppression subtractive hybridization, and Northern-type and in situ hybridizations, 150 nonredundant transcripts were documented in regions of the root tip that coincided with the regions investigated by Poroyko et al. (2005). An analysis of the root tip transcriptome by serial analysis of gene expression (SAGE) documented the presence of almost 16,000 different transcripts while the analysis and extrapolation of SAGE tags that are found in single copy (and thus excluded from the transcript count) suggests that at least 23,000 genes are transcribed in the maize root (Poroyko et al., 2005). The root transcriptome has been sampled to an estimated copy number of approximately five transcripts per cell. Frequency ranged from low copy number (2–5; 68.8% of transcripts) to highly abundant transcripts (100–1,200; 1% of transcripts). Quantitative reverse transcription-PCR for selected transcripts indicated high correlation with tag frequency. Computational analysis compared this set with known maize transcripts and other root transcriptome models. Among the 14,850 tags, 7010 (47%) were found for which no maize cDNA or gene model existed. The value of this tag collection will increase as more maize sequence information becomes available.
These multidisciplinary approaches are expected to contribute novel information toward a more comprehensive understanding of the regulation of root growth during water deficits and the characterization of the role of particular gene families (e.g., aquaporins; Hachez et al., 2006) involved in the adaptive response of the root to a drying soil.
Proteomics and Metabolomics
Deciphering gene function can also be facilitated by information gathered through the profiling of the proteome and metabolome which, as compared to the transcriptome, are functionally "closer" to the phenotypic traits selected for improving drought tolerance and can thus account also for the effects due to posttranscriptional and posttranslational regulation, a factor that remains unaccounted for in transcriptomics studies.
The pioneering work of de Vienne and coworkers has led to the identification of QTLs influencing protein quantity (protein quantity loci; Damerval et al., 1994) under drought conditions (de Vienne et al., 1999; Pelleschi et al., 1999; Zivy and de Vienne, 2000; Consoli et al., 2002). In maize, Jeanneau et al. (2002) have shown that under conditions of mild water stress the Asr1 gene, a putative transcription factor, co-localizes with a protein quantity locus for its protein (ASR1) and a QTL for ASI and leaf senescence. Based on these findings, it was hypothesized that the Asr1 polymorphism is responsible for the presence or absence of the ASR1 protein, which would pleiotropically affect the other responsive traits; the validity of the hypothesis was confirmed by means of genetic engineering (Jeanneau et al., 2002).
A detailed study on proteome profiling is in progress to ascertain the role of cell wall proteins (CWPs) in the elongation of the primary root in maize (Zhu et al., 2006). Although many of the CWPs identified in this study have previously been shown to be involved in cell wall metabolism and cell elongation, a number of CWPs (e.g., endo-1,3;1,4-β-D-glucanase and 
-L-arabinofuranosidase) were not described in previous cell wall proteomic studies. More recently, the two-dimensional profiling of the maize xylem proteome was reported (Alvarez et al., 2006). Although the xylem in plants has mainly been described as a conduit for the flow of water, minerals, and signaling molecules, emerging evidence indicates the presence of proteins in the xylem sap. In maize, the composition of the xylem sap proteome revealed proteins related to different phases of xylem differentiation including cell wall metabolism, secondary cell wall synthesis, and programmed cell death, all of which are likely to play relevant roles under conditions of water deficit. Many proteins were found to be present as multiple isoforms and some of these isoforms are affected by posttranslational modification, such as glycosylation. Proteins involved in defense mechanisms were also present in xylem sap and the sap proteins were shown to have antifungal activity in bioassays.
Metabolome profiling attempts to identify and quantify as many metabolites as possible in a given biological sample (Fiehn, 2002). Metabolic databases provide a valuable framework to predict biochemical pathways and to reveal the phenotype of specific mutations, thus offering a more comprehensive view of the functional characteristics under investigation. When applied to a mapping population, metabolome profiling can be used to identify QTLs regulating the level of a particular metabolite and verify its coincidence with QTLs for yield and/or genes involved in metabolic pathways. In maize, the changes which occurred during a drought episode in the level of sugars and starch and other important metabolites in the reproductive organs and in the growing kernel have been investigated (Zinselmeier et al., 1999), and QTLs for invertase activity were first described in a population subjected to drought stress (Pelleschi et al., 1999). The number of QTLs for invertase activity detected under drought (nine in total) was more than twice the number detected under well-watered conditions (four in total), an indirect indication of the important role of this enzyme under drought conditions. One QTL common to both treatments was located near Ivr2, an invertase-encoding gene on chromosome bin 5.03. Drought produced an early stimulation of acid-soluble invertase activity in adult leaves, whereas the activity of the cell wall invertase was found to be unaffected. This response was closely related to the mRNA level for only one (Ivr2) of the invertase genes. In a more recent study, the activity of three enzymes (invertase, sucrose-P synthase, and ADP glucose pyrophosphorylase) as well as the content of hexose, sucrose, and starch were investigated in the mature fourth leaf of maize plants grown in the greenhouse under well-watered and water-stressed conditions. The materials evaluated in this study comprised 120 RILs and were the same investigated by Pelleschi et al. (1999). The results showed that QTLs for traits related to carbohydrate metabolism were more frequently colocated with growth trait QTLs than photosynthesis-related ones, particularly under well-watered conditions; one colocation was observed between the three enzyme activities for sucrose and starch metabolism and a relevant structural gene, which can thus be considered as a candidate accounting for part of the variability of each enzymatic trait. Based on these findings, the authors concluded that QTL analysis for carbohydrate metabolism provides valuable insights for understanding and improving maize response to water stress. Collectively, the above-mentioned studies indicate that invertase activity is an important limiting factor for grain yield in maize exposed to drought during the reproductive phase, in keeping with what suggested by the work of Boyer and coworkers (Boyer and Westgate, 2004; McLaughlin and Boyer, 2004). Clearly, the compelling evidence on the role of invertase warrants further investigation.
From a technical standpoint, it is worth emphasizing that the combination of the above-mentioned "omics" platforms with laser-capture microdissection or other ablation techniques allows for unprecedented levels of functional resolution at the anatomical level, thus replacing classical cell fractionation approaches that tend to alter the transcript profiles. In maize, a combination of laser capture microdissection and subsequent microarray analyses applied to the root pericycle of wild-type and rum1 mutant allowed Woll et al. (2005) to identify 19 genes involved in signal transduction, transcription, and the cell cycle that are active before lateral root initiation; these findings will contribute to the identification of the developmental checkpoints involved in lateral root formation in maize downstream of rum1. More recently, the identification of a novel marker (ZmGrp3) for root initiation in maize has allowed for the development of a robust assay to quantify allele-specific contributions to gene expression in hybrids (Woll et al., 2006).
Searching for Candidate Genes in Species Other than Maize
The vast amount of sequence information and/or mutants available for drought-related genes mapped and/or cloned in species other than maize provides additional opportunities for identifying candidate genes also of interest for maize breeding. This is particularly true for traits that have been more extensively investigated in species whose genome has been sequenced such as Arabidopsis (Flavell, 2005; Maggio et al., 2006) and rice. An example is provided by candidate genes for root architecture and size, a category of traits that has been largely neglected in maize until recently.
Although root development in Arabidopsis differs from maize in both overall architecture and the anatomy of individual roots, genes cloned in Arabidopsis could in some cases provide interesting leads for the identification of candidate genes for root QTLs in maize, particularly for those functional and morphological features of root development that may have been conserved to a greater extent from an evolutionary standpoint. A vast amount of information is available on the root transcriptome of Arabidopsis (Birnbaum et al., 2003; Birnbaum and Benfey, 2004; Fizames et al., 2004). One aspect of root architecture that might have been conserved to a greater extent across species is the development of root hairs. These specialized epidermal cells play an important role in the uptake of water and nutrients from the soil and are the primary interaction site with soil microorganisms and mycorrhiza. While the process of root hair formation is well characterized in Arabidopsis, little information is available on the genetic basis of root hair development in monocots. In maize, a single recessive gene controlled the phenotype of three root hair mutants (rth1, rth2, and rth3) isolated using the Mutator transposon system. The chromosomal location of each of these three genes has been determined and the protein product for RHT1 was shown to be involved in polar exocytosis (Wen et al., 2005). Recent work performed in barley (Hordeum vulgare L.) using a hairless mutant has led to the isolation and cloning of the β-expansin (EXPB) gene HvEXP1 involved in root hair initiation (Kwasniewski and Szarejko, 2006). Expansins are a large family of plant proteins endowed with a unique cell wall–loosening activity that have been shown to be involved in a number of processes related to plant growth and development (Li et al., 2003) which are likely to play an important role in cellular and organ elongation under different water availabilities. The redundancy of expansin-coding genes (58 in rice; Sampedro and Cosgrove, 2005) points out the specificity of their role in growth and developmental processes, including root growth in maize (Wu et al., 1996, 2001). Based on these observations, a comparative analysis with the location of the QTLs reported for root hair length in maize (Zhu et al., 2005) with the map position of root-specific expansins will provide an opportunity to verify their role in controlling variability in root hair elongation.
An interesting example of the power of combining QTL analysis for root morphology and target metabolites with fine-mapping and mutant analysis was recently shown in Arabidopsis (Sergeeva et al., 2006), where the possible role in root elongation of the sucrose-splitting enzymes sucrose synthase and invertase was tested. Several QTLs affected both invertase activity and root length. The fine mapping of a major QTL for root length revealed its consistent colocation with a QTL for invertase activity and a gene coding for a vacuolar invertase, whose role in root elongation was confirmed by the analysis of a functional knockout line.
Future Challenges and Opportunities
Improving maize tolerance to drought will increasingly rely on the exploitation of MAS and a better understanding of the molecular and physiological bases of such tolerance. Among the traits which appear more amenable to MAS, root architecture and photosynthetic efficiency are interesting targets, in view of the difficulties of manipulating these traits through conventional approaches. An example is provided by the identification of QTLs regulating root penetration through soil hardpans, a circumstance which can be frequently encountered in maize-growing areas. The presence of a hardpan limits root growth, hence the amount of soil that roots can explore for extracting moisture and nutrients. Additional traits whose selection might be facilitated by MAS are osmotic adjustment, relocation of stem reserves, and stay green (Tuberosa and Salvi, 2004).
A recent study in which QTL information was linked to crop modeling has shown that QTL analysis removes part of the random errors of measured model input parameters and that this information can successfully be coupled with crop models to replace measured parameters (Tardieu, 2003). The QTL-based modeling relies on the estimation of parameters of response curves to environmental variables and allows for a more effective selection on the basis of QTL information. On the other hand, crop modeling has the potential to resolve G x E interactions as well as the genetic basis of trait plasticity (Chapman et al., 2003; Reymond et al., 2004; Cooper and Hammer, 2005; Cooper et al., 2005, 2006; Malamy, 2005). For this approach to be effective, crop models that are capable of predicting yield differences among genotypes in a population under various environmental conditions are needed (Tardieu, 2003; Cooper et al., 2005; van Eeuwijk et al., 2005; Hammer et al., 2006). A valuable example on how an ecophysiological model and QTL analysis can be integrated to investigate the genetic basis of leaf growth as related to drought and other environmental factors has been provided by Reymond et al. (2003), who have identified QTLs affecting leaf elongation rate in maize as a function of water vapor pressure difference, soil water status, and meristem temperature. The responses of leaf elongation rate to evaporative demand and to soil water status were common to several experiments for each genotype, implying that the ranking of genotypes across experiments was consistent. Differences in ranking of RILs and QTL instability of leaf length were linked to intrinsic differences in the sensitivities of RILs to evaporative demand or soil water deficit (Reymond et al., 2004). An unexpected outcome of this study was the identification of a strong and stable QTL for leaf width which was unrelated to QTLs for leaf length or for parameters of the model, hence suggesting that leaf length and width are governed by different genetic factors. The main advantage in identifying QTLs responsive to an environmental variable is that this method provides results which characterize a genotype per se, rather than its performance in a particular environment, an important feature when searching for drought-adaptive QTLs. The ultimate goal of the modeling approach is the possibility to implement an in silico selection able to identify the combinations of the desirable alleles at the target QTLs. Remarkably, a model based on the combined QTL effects predicted 74% of the variability for maize leaf elongation rate measured among a random sample of RILs of the mapping population used for QTL detection (Reymond et al., 2003). Of the several QTLs identified, most were specific for their response to only one variable. Further work is required to test the validity of models when more of such variables vary simultaneously, the condition typically encountered by crops in the field. Therefore, crop modeling based on QTLs for morpho-physiological traits responding to drought holds promise to help us to more appropriately resolve G x E interactions and to identify the genetic basis of growth plasticity. In turn, this will provide novel opportunities for collaboration between breeders and modelers to define the best ideotypes and may assist dissecting yield into characters under more simple genetic control (Yin et al., 2003; Cooper et al., 2006).
Genomics and bioinformatics allow us to investigate sequence colinearity in the main crops (Paterson et al., 1996) and compare their gene order and content with those of model species whose genomes have been sequenced, such as rice and Arabidopsis. The colinearity between Arabidopsis and maize (Van Buuren et al., 2002) has been eroded to such an extent that the Arabidopsis sequence does not appear to be of much help for the identification of related genes in maize. Conversely, comparative mapping between rice and maize (Ware et al., 2002) as well as other cereals (Liu et al., 2006) provides valuable opportunities to exploit high resolution collinear maps to facilitate the positional cloning of maize QTLs and/or identify candidate genes and to establish whether sequences with high homology are so because they represent orthologous loci.
The combined analysis of genome and transcript sequences suggests the presence of genes with a paralogous relationship in most plants (Bohnert et al., 2005). As compared to their orthologous sequences, paralogs, resulting from genome rearrangements (e.g., duplications, genome fusions, and/or chromosome or individual gene duplications) would appear to be under less evolutionary constraint to evolve new connections to stress-signaling networks (Kramer et al., 2004). Under this respect, the sequencing of entire genomes provides a vital framework to better understand and interpret the relationships and functions amongst members of gene families. However, comparing highly expressed transcripts of the maize root transcriptome with that in Arabidopsis indicated that <5% of the most abundant transcripts were common to both species (Poroyko et al., 2005), possibly suggesting that different regulatory pathways are deployed during root ontogeny of these two species.
Once the sequence of the entire maize genome is made available, the interest in deploying sequence from rice and other cereal species {e.g., sorghum [Sorghum bicolor (L.) Moench]} will decrease. Supplemental information useful for MAS or assigning function to genes can be obtained through the construction of comparative consensus maps that integrate the information of anchor markers and the results of different mapping populations within the same species (Khavkin and Coe, 1997; Tuberosa et al., 2002b; Sawkins et al., 2004; Pelleschi et al., 2006). Preliminary evidence suggested the possibility that clusters of QTLs may govern developmental features (Khavkin and Coe, 1997), some of which influence adaptation to drought conditions. Similar observations have been reported more recently in a mapping population that segregates for a number of morpho-physiological features (e.g., growth, photosynthesis, water status, carbohydrate metabolism, and ABA concentration) related to drought tolerance (Pelleschi et al., 2006). In maize, a clear example in this direction is provided by the region on chromosome bin 2.04 which has been shown to influence a number of traits important for tolerance to drought in different backgrounds (Lebreton et al., 1995; Tuberosa et al., 1998; Quarrie et al., 1999; Sawkins et al., 2004; Landi et al., 2005). Bin 1.06 is another chromosome region that has consistently shown the presence of QTLs for grain yield and drought-related traits (e.g., root traits) in a number of genetic backgrounds (Lebreton et al., 1995; Tuberosa et al., 1998, 2002b, 2003; Pelleschi et al., 2006). Determining the extent to which transcript profiles that record the influence of developmental and environmental changes on hormonal homeostasis are transferred into changes at the protein level will be one of the future challenges (Zhu et al., 2006).
Further challenges and opportunities have been ushered in by the detection of regulatory systems that depend on small noncoding RNAs in plants (Sunkar and Zhu, 2004; Jones-Rhoades and Bartel, 2004). Some small interfering RNAs (siRNAs) have been shown to be stress-inducible (Sunkar and Zhu, 2004). In addition to affecting the translation process, these siRNAs might participate in the alternative splicing (Jen et al., 2005) of mRNAs and represent components of an additional level of regulation that so far has gone undetected and could partially account for the low level of particular transcripts and their final product.
|
Conclusions |
|---|
|
TOPABSTRACTINTRODUCTIONConclusionsREFERENCES |
|---|
For its implications, QTL cloning should be viewed as one of the most innovative and potentially important contributions of genomics to our future capacity to more effectively understand and manipulate traits influencing drought tolerance. Additionally, high-throughput transcriptome, proteome, and metabolome profiling performed at unprecedented levels of cellular resolution (e.g., laser-capture microdissection; Nakazono et al., 2003) will further enhance our dissecting capacity of the phenotype, hence our ability to identify the causative traits and underlining mechanisms contributing to drought adaptation or to increase yield potential per se. However, the deluge of data originated by the "omics" platforms does not automatically translate into knowledge on how to improve drought tolerance. Under this respect, it is unlikely that a single profiling experiment or a single QTL study will provide clear indications as to which molecular targets should be pursued to improve drought tolerance.
The success and effectiveness of MAS in assembling cultivars more tolerant to drought will rely on the identification of the relevant QTL alleles and their pyramiding in the correct combinations, in a sort of molecular jigsaw puzzle. This new concept of "breeding by design" (Peleman and van der Voort, 2003), although already applicable from a purely technical standpoint in maize and other crops, in the case of drought tolerance it is still a long way from being routinely applicable. This is mainly due to our poor understanding of the molecular basis of drought tolerance and, most importantly, the difficulty in predicting the phenotypic value of a new genotype tailored through MAS.
Although it is difficult to predict to what extent MAS, genomics, and postgenomics approaches will eventually impact conventional breeding practices, future progress toward more sustainable agricultural practices will be accelerated through a more systematic discovery of the gene functions influencing yield under water-limited conditions and a deeper understanding of their interactions with the environment. In conclusion, the successful exploitation of MAS and other genome-wide approaches to enhance drought tolerance will only be possible within a multidisciplinary context able to provide an accurate understanding of the factors curtailing yield potential under water-limited conditions and relying on a thorough field assessment of the selected materials in the target environments.
Received for publication July 9, 2007.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONConclusionsREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
