What Has Natural Variation Taught Us about Plant Development, Physiology, and Adaptation?

Abstract

Nearly 100 genes and functional polymorphisms underlying natural variation in plant development and physiology have been identified. In crop plants, these include genes involved in domestication traits, such as those related to plant architecture, fruit and seed structure and morphology, as well as yield and quality traits improved by subsequent crop breeding. In wild plants, comparable traits have been dissected mainly in Arabidopsis thaliana. In this review, we discuss the major contributions of the analysis of natural variation to our understanding of plant development and physiology, focusing in particular on the timing of germination and flowering, plant growth and morphology, primary metabolism, and mineral accumulation. Overall, functional polymorphisms appear in all types of genes and gene regions, and they may have multiple mutational causes. However, understanding this diversity in relation to adaptation and environmental variation is a challenge for which tools are now available.

INTRODUCTION

Intraspecific natural variation (hereinafter referred to as natural variation) may be broadly defined as the within-species phenotypic variation caused by spontaneously arising mutations that have been maintained in nature by any evolutionary process including, among others, artificial and natural selection. Thus, natural variation embraces the enormous diversity present within wild plant species as well as most of the genetic variants that are found in domesticated plants. Some of the phenotypic differences existing in wild or cultivated plants are due to single-gene (monogenic) allelic variants. However, most of the natural variation is quantitative and determined by molecular polymorphisms at multiple loci and genes (multigenic), which are referred to as quantitative trait loci (QTL) and quantitative trait genes (QTGs).

The natural variation present in crop plants has been exploited since their domestication thousands of years ago by the genetic manipulation of developmental traits and physiological features related to adaptation to agriculture. Methods for genetic analysis and mapping of natural quantitative variation were developed a few decades ago for crop species, in which many more studies have been performed than in wild plants. Currently, genomic resources have been developed for several herbaceous crop plants, such as rice (Oryza sativa), barley (Hordeum vulgare), and tomato (Solanum lycopersicum), enabling the identification of the genes and nucleotide polymorphisms underlying QTL involved in domestication, yield, biotic and abiotic stress, and quality traits (Doebley et al., 2006; Sang, 2009). In addition, the recent availability of the first genome sequences of cultivated perennial woody plants, such as poplar (Populus spp) and grapevine (Vitis vinifera; Tuskan et al., 2006; Jaillon et al., 2007), is increasing the scope of traits and processes whose natural variation can be dissected molecularly.

In addition, the analysis of natural variation in wild species has begun to elucidate the molecular bases of phenotypic differences related to plant adaptation to distinct natural environments and to determine the ecological and evolutionary processes that maintain this variation (Mitchell-Olds et al., 2007). The molecular approach to these questions has been applied mainly to the wild annual crucifer Arabidopsis thaliana, which has become a model plant for the study of natural variation (Mitchell-Olds and Schmitt, 2006). Its wide geographical and environmental distribution, combined with its small genome and the availability of unprecedented genetic and genome resources, have strongly facilitated its molecular analysis in the last decade. As a result, A. thaliana has provided the largest number of genes and nucleotide polymorphisms underlying natural variation of any plant species (Alonso-Blanco et al., 2005). However, the specific ecological niche and life history of A. thaliana limits the plant traits and processes that can be approached in a single species. Therefore, new plant models phylogenetically related to A. thaliana (e.g., Arabidopsis lyrata; Clauss and Koch, 2006) as well as unrelated species (e.g., of the genera Aquilegia [Kramer, 2009], Mimulus [Wu et al., 2007], Ipomoea [Clegg and Durbin, 2003], and Helianthus [Rieseberg et al., 2003]) are beginning to be used in studies of natural variation and speciation.

Our understanding of plant biology has benefited tremendously from work done with artificially induced mutants in model species, as illustrated by A. thaliana mutants currently available for most genes, and mutant screenings continue to be useful in the identification of gene function (Page and Grossniklaus, 2002). Gene functions involved in plant survival and adaptation can be identified partially by induced mutant analyses of different wild genotypes, where mutants with reduced fitness are easily selected. However, current mutant collections have been obtained using a limited number of laboratory strains, which harbor only a small portion of A. thaliana natural variation. Interestingly, Clark et al. (2007) showed that ;9.4% of A. thaliana protein-coding genes are naturally absent or knocked out in wild accessions, limiting the mutant spectra that can be obtained from each accession. Therefore, natural variation provides a relevant complementary resource to discover novel gene functions as well as those allelic variants that specifically interact with the genetic background and/or the environment or alleles showing small effects on phenotype, particularly for traits related to plant adaptation (Benfey and Mitchell-Olds, 2008).

Genetic analyses of natural variation in plants currently are mainly performed by QTL mapping, often called linkage mapping, in which phenotypic variation is associated with allelic variation at molecular markers segregating in experimental mapping populations derived from directed crosses (Doerge, 2002). Thus, genomic regions accounting for trait variation are located in large physical intervals containing the causal QTLs. Further analyses of these regions with a combination of functional strategies allows the final identification of QTGs and nucleotide polymorphisms altering the function of those genes (reviewed in Koornneef et al., 2004; Alonso-Blanco et al., 2005; Weigel and Nordborg, 2005; González-Martínez et al., 2006). Association mapping, which involves looking for phenotype-genotype associations in a general population of individuals whose degree of relatedness or pedigree is unknown, is also becoming more popular and useful in plant systems, owing to improvements in statistical and analytical tools and in gene sequencing technology (see Myles et al., 2009).

The information gleaned from the various types of QTL analysis is extending the current concept of gene function based on a single wild-type allele and provides a more realistic and dynamic view of gene function by estimating the effects of different natural alleles depending on genetic background, environment, and evolutionary temporal scale. In addition, recent genetic analyses of large data sets generated by high-throughput profiling of mapping populations with “-omics” procedures (genomics, transcriptomics, metabolomics, proteomics, ionomics, or epigenomics) is allowing the genetic integration of several levels of molecular regulation of phenotypic trait variation. Thus, QTL and gene networks controlling natural variation are being inferred (Jansen et al., 2009).

In this review, we discuss the major contributions of the analysis of natural variation to our understanding of plant development and physiology. Given the large number of studies performed in the past, we do not aim to completely cover all traits analyzed so far, but to focus on several relevant phenotypes recently or extensively dissected in A. thaliana and major crop species. These traits include the timing of germination and flowering, plant growth and morphology, primary metabolism, and mineral accumulation. Detailed discussion of natural variation for secondary metabolites is covered by Kliebenstein (2009a), and various reviews have specifically described the natural variation involved in crop domestication and diversification (Doebley et al., 2006; Pourkheirandish and Komatsuda, 2007; Purugganan and Fuller, 2009; Sang, 2009), plant development (Alonso-Blanco et al., 2005), or resistance to pathogens (Holub, 2007). Here, we summarize the major genes and allelic variants currently known to participate in plant adaptation to different natural and agricultural environments through some modifications of plant development and physiology. The kind of mutations, molecular mechanisms, and gene networks involved in microevolution and adaptation of different plant species to changing environments are beginning to be elucidated. Consequently, new perspectives in our understanding of phenotype and pan-genome variation and evolution are emerging (Morgante et al., 2007).

DEVELOPMENTAL PROCESSES

Domestication of crop plants mainly has resulted in changes in traits related to floral and seed morphology, such as shattering, free-threshing, and seed coat color in cereals (Salamini et al., 2002, Dubcovsky and Dvorak, 2007), plant architecture-like branching patterns in maize (Zea mays) and rice (Izawa et al., 2009), fruit morphology-like size and shape in Solaneaceae crops (Tanksley, 2004), and seed dormancy, especially in domesticated cereals (Salamini et al., 2002). In addition, the analysis of natural variation has identified novel molecular mechanisms involved in the regulation of life history traits, such as seed dormancy and flowering time, by different environmental and endogenous clues. This section highlights some of the most important advances resulting from these studies.

Seed Dormancy and Germination

Most wild annual plant species, including A. thaliana, show substantial intraspecific natural variation for seed dormancy and germination properties, which is presumably involved in adaptation to different environments. Domestication has often led to a reduction of seed dormancy in crop plants, although considerable variation remains among cultivated varieties. The genetic bases of this variation have been approached mainly in A. thaliana and cereal plants, mostly in relation to environmental factors, such as temperature and light, and to endogenous hormonal signals, such as abscisic acid (ABA) and gibberellin (GA).

In A. thaliana, QTL mapping analyses of germination-related traits have been performed in relation to after-ripening (Alonso-Blanco et al., 2003; Clerkx et al., 2004), light regimes (van der Schaar et al., 1997; Laserna et al., 2008; Meng et al., 2008), low temperature (Meng et al., 2008), or GA inhibitors (van der Schaar et al., 1997). Many of the loci are only detected in specific assays or conditions, indicating strong QTL by environment interactions. However, several QTLs colocate in different crosses, suggesting the presence of moderate frequency natural alleles with pleiotropic effects on dormancy and related traits, such as cold-tolerant germination and seed longevity (Clerkx et al., 2004; Meng et al., 2008). Particularly, the locus DOG1 has been detected in three experimental populations (Alonso-Blanco et al., 2003; Clerkx et al., 2004; Laserna et al., 2008) and has been isolated, identifying a member of a small gene family of unknown molecular function (Bentsink et al., 2006; Table 1). DOG1 induces seed dormancy and is specifically expressed during seed development. In addition, it has been shown that DOG1 functional nucleotide polymorphism(s) affect its expression, dormant accessions having higher RNA levels than nondormant accessions (Bentsink et al., 2006). Furthermore, DOG1 is involved in the ABA-mediated sugar signaling pathway in seedlings since glucose induces the expression of the DOG1/GSQ5 Cvi allele, whereas Landsberg erecta and Columbia alleles do not respond (Teng et al., 2008). Thus, DOG1 has provided the first gene involved in natural variation for seed dormancy and illustrates the value of this variation to identify previously unknown genes for this trait.

Table 1.

Genes and Funtional Polymorphisms Involved in Natural Variation for Different Traits and Species

				Functional Nucleotide Polymorphism		Allelic Dysfunction
Plant Species	Locus/Gene	Phenotypic Effects	Molecular Function	Gene Position	Mutation		References
A. thaliana	DOG1	Seed dormancy, sugar sensing	Unknown	Promoter	Unknown	ME	Bentsink et al. (2006)
Thale cress	FRI	Flowering	Unknown	Promoter, coding	Ins, Del, nonsense Subs	TP, ME	Johanson et al. (2000); Shindo et al. (2005)
	FRL1	Flowering	FRI-like	Coding	Nonsense Subs	TP	Schläppi (2006)
	FRL2	Flowering	FRI-like	Coding	Missense Subs	AP	Schläppi (2006)
	FLC	Flowering	MADS TF	Intron, coding	TE insertion, nonsense Subs	ME, TP	Michaels et al. (2003); Werner et al. (2005a)
	FLW1/FLM	Flowering	MADS TF	Promoter + coding	Gene Del	Null expression	Werner et al. (2005b)
	ART1/HUA2	Flowering, shoot morphology	RNA processing factor	Coding	Nonsense and missense Subs	TP, AP	Doyle et al. (2005); Wang et al. (2007)
	EDI/CRY2	Flowering, seedling growth, fruit size	Photoreceptor	Coding	Missense Subs	AP	El-Assal et al. (2001, 2004)
	PHYD	Flowering, seedling growth	Photoreceptor	Coding	Del	TP	Aukerman et al. (1997)
	PHYC	Flowering, seedling growth	Photoreceptor	Coding	Nonsense Subs	TP	Balasubramanian et al. (2006)
	PHYA	Seedling growth	Photoreceptor	Coding	Missense Subs	AP	Maloof et al. (2001)
	PHYB	Seedling growth	Photoreceptor	Coding	Missense Subs	AP	Filiault et al. (2008)
	TZP/Light5	Seedling growth	Zn knuckle protein	Coding	Ins	TP	Loudet et al. (2008)
	CAL	Flower morphology	MADS TF	Coding	Missense Subs	AP	Kempin et al. (1995)
	IIL1	Leaf morphology	Leu biosynthesis enzyme	Intron	Ins, Del	ME	Sureshkumar et al. (2009)
	GL1	Trichome formation	Myb TF	Coding	Del	ME	Hauser et al. (2001)
	PUB8	Self-incompatibility	U box protein	Promoter	Unknown	ME	Liu et al. (2007)
	MUM2	Seed mucilage	Galactosidase	Coding	Del	ME	Macquet et al. (2007)
	BRX	Root growth	Novel TF	Coding	Nonsense Subs	TP	Mouchel et al. (2004)
	LPR1	Root growth response to P	Multicopper oxidase	Promoter	Del	ME	Svistoonoff et al. (2007)
	INV	Sugar composition, root growth	Invertase	Unknown	Unknown	Unknown	Sergeeva et al. (2006)
	Small effect QTL	Biomass accumulation	Protein kinase	Unknown	Unknown	Unknown	Kroyman and Mitchell-Olds (2005)
	DM1	Hybrid incompatibility, disease resistance	NB-LRR proteins	Promoter + coding	Nonsense Subs, Del	AP	Bomblies et al. (2007)
	QTL3	Hybrid incompatibility, disease resistance	NB-LRR proteins	Unknown	Unknown	Unknown	Alcázar et al. (2009)
	LD1. 1	Genetic incompatiblity, embryogenesis	His biosynthesis enzyme	Promoter, coding	Nonsense Subs	ME, TP	Bikard et al. (2009)
	LD1.5	Genetic incompatiblity, embryogenesis	His biosynthesis enzyme	Promoter + coding	Gene Del	Null expression	Bikard et al. (2009)
	APR2	Sulfate accumulation	Sulfate reductase	Coding	Nonsense Subs	AP	Loudet et al. (2007)
	MOT1	Mo accumulation	Sulfate transporter	Promoter	Del	ME	Baxter et al. (2008)
	HMA5	Cu toxicity	Cu ATPase transporter	Coding	Missense Subs	AP	Kobayashi et al. (2008)
	HKT1	Na accumulation	Na+ transporter	Promoter	Del	ME	Rus et al. (2006)
A. lyrata	FRI	Flowering	FRI-like	Coding	Indel	AP	Kuittinen et al. (2008)
B. rapa	BrFLC1	Flowering	MADS TF	Unknown	Unknown	Unknown	Kole et al. (2001)
Turnip	BrFLC2	Flowering	MADS TF	Unknown	Unknown	Unknown	Schranz et al. (2002)
S. lycopersicum	SUN	Fruit shape	IQ67 domain protein	Coding + promotor	Gene duplication	ME	Xiao et al. (2008)
Tomato	OVATE	Fruit shape	Nuclear protein	Coding	Nonsense Subs	TP	Cong et al. (2002)
	FAS	Fruit shape	Yabby TF	Intron	Ins	ME	Cong et al. (2008)
	Fw2.2	Fruit size	Cell cycle control	Promoter	Subs, indel	ME	Frary et al. (2000)
	Sw4. 1	Seed size	Transporter	Unknown	Unknown	Unknown	Orsi and Tanksley (2009)
	Brix9-2-5/LIN5	Fruit sugar content	Invertase	Coding	Missense Subs	AP	Fridman et al. (2004)
	Cnr	Fruit ripening	SBP TF	Promoter	DNA methylation	ME	Manning et al. (2006)
Pisum sativum	LF/PsTFL1c	Flowering	TFL1-like protein	Promoter, coding	Gene Del	Null expression	Foucher et al. (2003)
Pea	R	Seed composition and morphology	Starch branching enzyme	Coding	TE insertion	TP, ME	Bhattacharyya et al. (1990)
Lynaria vulgaris	Lcyc	Flower simetry	TCP TF	Promoter + coding	DNA methylation	ME	Cubas et al. (1999)
Toadflax
T. aestivum	VRN1	Flowering	AP1-like MADS TF	Promoter, intron	Del, Ins	ME	Yan et al. (2003)
Bread wheat	VRN3	Flowering	FT-like protein	Promoter	TE insertion	ME	Yan et al. (2006)
	Q	Seed free-thressing	AP2-like TF	Coding	Missense Subs	AP	Simons et al. (2006)
	RhtB1, RhtD1	Plant and leaf size	GAI-like protein	Coding	Nonsense Subs	TP	Peng et al. (1999)
	Gpc-B1/NAM-B1	Senescence; Zn, Fe, and protein content	NAC TF	Coding	Del, gen Del	TP	Uauy et al. (2006)
Triticum monococum	VRN1	Flowering	AP1-like MADS TF	Promoter	Del	ME	Yan et al. (2003)
Einkorn wheat	VRN2	Flowering	CCT domain protein	Promoter, coding	Missense Subs	AP	Yan et al. (2004a)
H. vulgare	VRN2	Flowering	CCT domain protein	Promoter + coding	Gene Del	Null expression	Yan et al. (2004a)
Barley	VRN3	Flowering	FT-like protein	Intron	Subs	ME	Yan et al. (2006)
	Ppd1	Flowering	Response regulator	Coding	Missense Subs	AP	Turner et al. (2005)
	Nud	Seed free-threshing	ERF TF	Promoter + coding	Gene Del	Null expression	Taketa et al. (2008)
	Vrs1	Seed number, ear morphology	HD TF	Coding	Ins	TP	Komatsuda et al. (2007)
	Bot1	Boron toxicity tolerance	Boron efflux transporter	Promoter + coding	Gene duplication	ME	Sutton et al. (2007)
O. sativa	qLTG3	Germination	Unknown	Coding	Del	ME	Fujino et al. (2008)
Rice	Rc	Germination, seed color	bHLH TF	Coding	Del	TP	Sweeney et al. (2006)
	Hd1	Flowering	CO-like TF	Coding	Del, TE insertion	TP, ME	Yano et al. (2000); Doi et al. (2004)
	Hd3a	Flowering	FT-like protein	Unknown	Unknown	ME	Kojima et al. (2002)
	Hd6	Flowering	Protein kinase CK2	Coding	Nonsense Subs	TP	Takahashi et al. (2001)
	Ehd1	Flowering	Response regulator	Coding	Missense Subs	AP	Doi et al. (2004)
	Ghd7	Flowering, plant height, seed yield	CCT domain protein	Coding	Gene Del, nonsense Subs	Null expression	Xue et al. (2008)
	sh4	Shattering	Myb TF	Coding	Missense Subs	AP	Li et al. (2006)
	qSH1	Shattering	HD TF	Promoter	Missense Subs	ME	Konishi et al. (2006)
	qSW5	Seed size	Unknown	Coding, promoter	Del	ME	Shomura et al. (2008)
	GW2	Seed size	RING protein	Coding	Del	TP	Song et al. (2007)
	GS3	Seed size	Unknown	Coding	Missense Subs	TP	Fan et al. (2006)
	Gn1a	Seed number	CK oxidase/dehydrogenase	Promoter	Del	ME	Ashikari et al. (2005)
	Wx	Seed starch composition	Starch biosynthesis enzyme	Intron	Splice site Subs	TP	Isshiki et al. (1998)
	SKC1	Salt tolerance	HKT transporter	Coding	Missense Subs	AP	Ren et al. (2005)
Z. mays	Vgt1	Flowering	AP2-like TF	Promoter enhancer	TE insertion, indel	ME	Salvi et al. (2007)
Maize	tb 1	Inflorescence architecture	TCP TF	Promoter enhancer	Unknown	ME	Doebley et al. (1997)
	tga1	Seed morphology	SBP TF	Coding	Missense Subs	AP	Wang et al. (2005)
	Wx	Starch composition	Starch biosynthesis enzyme	Promoter, intron, coding	TE insertion	ME	Varagona et al. (1992)

TF: Transcription factor; TE: Transposable element; Subs: substitution; Ins: insertion; Del: deletion; TP: truncated protein; AP: altered protein; ME: Misexpression

In cereals, several analyses have identified seed dormancy QTLs in crosses between cultivated varieties with low dormancy and between cultivated and dormant weedy accessions or wild relative species (Lin et al., 1998; Gu et al., 2005; Hori et al., 2007; Imtiaz et al., 2008). In particular, the large-effect locus Sdr1 of rice might be orthologous to barley SD4 and wheat (Triticum aestivum) 4AL loci (Lin et al., 1998). Detailed QTL characterizations have shown that some loci may control dormancy through seed maternal tissues, while others function in offspring tissue(s) (Gu et al., 2008). As in wheat, maternally inherited seed dormancy has been associated with the seed pigmentation color variation determined by the rice Rc locus (Gu et al., 2005), which encodes a bHLH transcription factor (Sweeney et al., 2006). In addition, rice QTLs have been found for germination at low temperature, and the first of such loci, qLTG3-1, has been recently isolated (Fujino et al., 2008). qLTG3-1 encodes a protein of unknown function that is expressed in the embryo during seed germination. It has been suggested that qLTG3-1 is involved in tissue weakening during seed germination, which is reduced in the nondormant rice accession bearing a loss-of-function deletion allele.

Genetic analyses of dormancy and germination also have been performed in other crop species, such as lettuce (Lactuca sativa; Argyris et al., 2005). Interestingly, an interspecific cross between L. sativa and L. serriola identified Htg6.1 as a major effect locus involved in germination at high temperature (thermoinhibition), which colocates with germination QTLs for ABA sensitivity and GA response as well as with the ABA biosynthesis gene Ls-CED4. In addition, Ls-CED4 is induced by high temperature only in the thermoinhibited cultivated species, which strongly supports Ls-CED4 as candidate for Htg6.1 (Argyris et al., 2008).

Flowering Time

Molecular studies of the natural variation in the annual species A. thaliana, pea (Pisum sativum), rice, barley, wheat, and maize have identified ;25 different causal genes and their nucleotide polymorphisms of major effect (Table 1). In addition, genetic analyses of flowering time or related traits have been performed in other annual plants, such as tomato (González-Martinez et al., 2006), biennial species like Brassica sp (Schranz et al. 2002), short-lived perennials like Fragaria vesca (Albani et al., 2004), and several long-lived woody trees (Frewen et al., 2000; Casasoli et al., 2006). These studies are identifying multiple candidate genes and polymorphisms, which allow comparative analyses among species with rather different environmental and ecological distributions as well as physiological and developmental behaviors.

A. thaliana is a facultative long-day species widely distributed in temperate regions of the Northern hemisphere. Two extreme flowering behaviors have been described among wild genotypes grown in laboratory conditions, which are considered winter (late) and spring (early) life habits because they show strong and weak vernalization responses, respectively. Nine genes have been isolated that contribute to A. thaliana natural variation for flowering time (Table 1). Among these, FRI and FLC are involved in the regulation of flowering by vernalization and were first identified because of this natural variation. FLC encodes a transcripition factor of the MADS family that represses flowering (Michaels and Amasino, 1999). Its expression is downregulated by the low temperature of winter through epigenetic mechanisms leading to histone methylation and chromatin modifications (reviewed in Sung and Amasino, 2005). FRI encodes a protein of unknown function that also represses flowering initiation by activating the expression of FLC (Figure 1; Johanson et al., 2000). Most winter annual genotypes carry active alleles of both genes, which interact epistatically to delay flowering, while many summer annual accessions bear partial or total loss-of-function alleles in one or both genes (Werner et al., 2005a). More than 25 FRI independent loss-of-function alleles have been described, which mainly encode truncated proteins, indicating that FRI loss-of-function has evolved multiple times from a late flowering ancestor (Johanson et al., 2000; Le Corre et al., 2002; Shindo et al., 2005). By contrast, several natural loss-of-function alleles of FLC with reduced gene expression are caused by independent insertions of transposon elements within the first regulatory intron (Gazzani et al., 2003; Michaels et al., 2003). Other rare alleles are produced by nonsense or splicing mutations predicted to generate truncated proteins (Werner et al., 2005a). Interestingly, it has been suggested that additional allelic variation in FLC cis-regulatory sequences that alters the efficiency of FLC chromatin silencing might contribute to the quantitative variation for vernalization response present among late flowering genotypes (Shindo et al., 2006).

Figure 1.

Gene Networks Involved in Natural Variation for Flowering Responses to Vernalization and Photoperiod signals in A. thaliana, Wheat, and Rice.

Diagrams do not represent full molecular models of flowering regulation, but they show gene network components and branches known to contribute to the natural intraspecific variation existing in these species. Yellow color represents genes that contribute to the variation in a single species, which do not present known homologous genes contributing to natural variation in other species. Green color depicts homologous genes that contribute to the variation in several species. Blue and purple indicate genes accounting for the variation in one species, which show homologous protein domains with genes contributing to the variation in another species. Given their large functional homology, CO, FT, and AP1 homologs of the three species are included in the three diagrams, with small circles depicting genes that do not contribute to natural variation. See text for further details.

QTL mapping studies have also shown that other loci, including homologs of FRI and FLC, affect the vernalization response (Figure 1). Natural loss-of-function alleles in two FRI-like genes (FRL1 and FRL2; Schläppi, 2006) and one FLC-like gene named FLM/FLW1 (Werner et al., 2005b) have been shown to produce early flowering. Additional flowering time QTLs affecting this response have been colocated with other FLC homologs in several crosses (Alonso-Blanco et al., 1998; O'Neill et al., 2008). Furthermore, loss- and gain-of-function alleles of another FLC regulator called HUA2 likely contribute to this variation (Doyle et al., 2005; Wang et al., 2007).

In addition, natural variation for the flowering photoperiod response is caused by three photoreceptor genes, CRY2, PHYC, and PHYD, in A. thaliana (Table 1, Figure 1). Rare loss-of-function alleles of PHYD and PHYC as well as a gain-of-function of CRY2 produce early flowering and reduced photoperiod response (Aukerman et al., 1997; El-Assal et al., 2001; Balasubramanian et al., 2006). Association mapping analyses suggest that haplotypes of CRY2 and PHYC differing in eight amino acid substitutions might further contribute to this variation (Olsen et al., 2004; Balasubramanian et al., 2006).

QTL linkage mapping and association analyses of candidate genes also have been performed in other Brassciaceae species, showing that FLC-like genes probably contribute to the differentiation of annual and biennial habits in Brassica rapa (Schranz et al., 2002; Yuan et al., 2009) and that functional variants of a FRI-like gene of A. lyrata affects its vernalization requirement (Kuittinen et al., 2008). In addition, molecular analyses have identified the LF gene in another dicotyledonous crop species, pea, which is a long-day temperate legume that also responds to vernalization. Pea LF behaves as a flowering repressor that is homologous to A. thaliana TFL1, and a natural loss-of-function LF allele caused by a large deletion shows reduced photoperiod response (Foucher et al., 2003).

In the last few years, comparable progress has been achieved in understanding flowering time variation in several monocotyledonous grass species, including temperate cereals with similar photoperiod and vernalization responses to A. thaliana (Table 1). Two genes largely determine the life habit and vernalization response of barley and wheat. VRN1 encodes a MADS box transcription factor homologous to the meristem identity protein AP1 of A. thaliana (Yan et al., 2003), and VRN2 encodes a grass-specific putative transcription factor with a CCT domain (Yan et al., 2004b). VRN1 promotes flowering and is speculated to be directly downregulated by vernalization, whereas VRN2 is a repressor that reduces the expression of VRN1 mediated by VRN3, a feedback regulatory loop among these genes accounting for their epistatic interactions (Figure 1; Distelfeld et al., 2009a). The early flowering spring habit of wheat and barley has been associated with more than 15 independent insertion and deletions in the first regulatory intron of VRN1 (Yan et al., 2004a; Fu et al., 2005; Szucs et al., 2007) and with several loss-of-function alleles of VRN2 (Distelfeld et al., 2009b). In addition, natural variation for the photoperiod response and its interaction with vernalization is determined by VRN3 and Ppd1. VRN3, which is a homolog of A. thaliana FT encoding a long-distance flowering signal, is upregulated by long-day photoperiod and vernalization and activates VRN1 expression (Yan et al., 2006). Ppd1 encodes a pseudoresponse regulator that mediates the photoperiod upregulation of VRN3 (Turner et al., 2005). Dominant mutations in the promoter and first intron of wheat and barley VRN3 show increased expression and response to vernalization (Yan et al., 2006). By contrast, missense recessive mutations in the CCT domain of barley Ppd-H1 likely cause late flowering under long-day photoperiod (Turner et al., 2005). Moreover, another barley FT homolog has been proposed as a possible candidate for the Ppd-H2 photoperiod QTL (Faure et al., 2007; Kikuchi et al., 2009).

Extensive analyses are also identifying the molecular bases of the natural variation for flowering time of rice and maize. In contrast with A. thaliana and temperate cereals, these are tropical short-day photoperiod species with no significant response to vernalization. To date, five genes have been isolated as determining the large variation for photoperiod response in rice (Table 1). Hd1 and Hd3a are flowering promoters, homologs of A. thaliana CO and FT genes (Yano et al., 2000; Kojima et al., 2002). Several Hd1 loss-of-function alleles reduce the expression of Hd3a and produce late flowering under short days, which indicates a conservation of this photoperiod regulatory module in phylogenetically distant short- and long-day flowering species (Figure 1). In addition, EhdD1 encodes a response regulator that independently upregulates Hd3a under short days (Doi et al., 2004). By contrast, Ghd7 and Hd6 encode a CCT domain protein and a CK2 kinase, respectively, which probably repress Hd3a under long-day photoperiod (Takahashi et al., 2001; Xue et al., 2008). Loss-of-function alleles of both genes produce early flowering, but additional alleles differing in several Ghd7 missense substitutions have been suggested to further contribute to rice flowering natural variation. In maize, the Vgt1 locus has been identified as accounting for the natural variation for flowering photoperiod response (Salvi et al., 2007). Vgt1 is an ;2-kb noncoding cis-element that regulates the expression of an AP2 homolog located 70 kb downstream and behaves as a flowering repressor. Several polymorphisms in the Vgt1 region have been strongly associated with flowering time variation (Salvi et al., 2007; Ducrocq et al., 2008).

The identification of large-effect QTGs has shown that photoreceptor and regulatory genes of different classes, such as those encoding transcription factors or signal transduction components, account for a large portion of the natural variation for flowering time in different species. Comparisons among species with similar physiological and environmental requirements for flowering induction show that some homologous genes contribute to the intraspecific variation in related species of the same plant family (e.g., FLC and VRN3 in Brassicaceae and grasses, respectively). Thus, parallel evolution of flowering time by mutations in homologous conserved genes is apparent in different groups of related species. However, comparisons among distant species from the same or different phylogenetic families show that natural variation for flowering responses to vernalization and photoperiod signals is mainly generated by allelic variation at different genes in each species, which are often highly divergent among species (Figure 1). Current results suggest that such convergent evolution of flowering responses is determined by species-specific genes showing analogous network positions in distant species (e.g., FLC and VRN2) and by homologous genes with partly differentiated functions (e.g., AP1 and VRN1). However, photoperiod responses seem to involve more conserved genes among species (photoreceptors, CO, and FT) than those involved in vernalization (FLC, VRN1, and VRN2), suggesting stronger constrains for the evolution of photoperiod than vernalization responses. Given the well-documented climate changes across evolutionary time, it can be further speculated that these differential constraints have been influenced by a wider temporal prevalence of photoperiod than vernalization signals.

The availability of causal genes and polymorphisms enables the study of the evolutionary processes that generate and maintain flowering time variation within and among species (reviewed in Roux et al., 2006; Mitchell-Olds et al., 2007). In crop species, fitness effects and adaptive values of genetic variants are often documented because segregating populations are grown in their natural agricultural environments (Izawa, 2007). Detailed analyses indicate that some loci affect not only heading date but several other fitness components, as shown with the effects of Ghd7 on plant height, panicle size, and seed production (Xue et al., 2008). Geographical distributions of alleles of Hd1, Ehd1, and Dgh7 from rice (Izawa, 2007; Xue et al., 2008) or Vgt1 and Dwarf8 from maize (Camus-Kulandaivelu et al., 2006; Ducrocq et al., 2008) show latitudinal and/or altitudinal clines, which suggests that they are involved in adaptation to different world regions. Current multilocus analyses of cultivated varieties are elucidating the role of flowering time allelic variation in regional adaptation and crop expansion during the few thousand years after domestication.

Similar studies have been initiated in A. thaliana, where in contrast with crop plants, several of the natural alleles identified in CRY2, FLM, PhyC, PhyD, or HUA2 (Table 1) are found in a single accession; therefore, it is possible that they are rare deleterious alleles. However, analyses of nucleotide diversity of CRY2, PhyC, FRI, and FLC have detected highly differentiated haplotypic groups, suggesting that their allelic variation might be maintained by various types of selection (Olsen et al., 2004; Caicedo et al., 2004; Stinchcombe et al., 2004; Balasubramanian et al., 2006). Most of the large number of FRI loss-of-function alleles appear at very low frequency, suggesting that they have been generated recently, for example, after the last glaciation (Le Corre et al., 2002; Shindo et al., 2005). However, two of these alleles are widely distributed in Central Europe, supporting the notion that their frequency increased due to strong natural selection during the postglacial colonization of this region (Johanson et al., 2000; Toomajian et al., 2006). In addition, a Eurasian latitudinal cline of flowering time and vernalization sensitivity has been shown to be partly determined by interactions among FRI, FLC, and PhyC allelic variation, suggesting that these interactions contribute to latitudinal and local adaptation (Caicedo et al., 2004; Lempe et al., 2005; Balasubramanian et al., 2006). Supporting the adaptive value of this allelic variation, analysis of FRI and FLC alleles in field experiments performed under environments closer to their natural habitat than laboratory conditions show that the FRI/FLC epistatic interaction affects different fitness components depending on the season of germination (Korves et al., 2007). In addition, antagonistic pleiotropic effects of FRI allelic variation on different fitness components, such as flowering time and plant architecture, suggest that complex phenotypic pleiotropy accounts for the adaptive value of FRI alleles (Scarcelli et al., 2007). Similarly, the large pleiotropic effects described for photoreceptor genes and HUA2 (Table 1) probably also contribute to the global adaptive value of their genetic variants. Thus, several demographical and selective processes seem to determine the natural variation for flowering time in crop and wild species, but their relative significance and molecular targets are just beginning to emerge.

Plant Architecture and Morphology

More than 20 loci of major effect have now been cloned for several domestication traits related to plant architecture, revealing that loss of functions of transcription factor genes (sometimes previously unknown) are often involved (Table 1). Examples are the rice shattering locus sh4 encoding a Myb factor (Li et al., 2006), the rice shattering locus qSH1, and the six/two row locus Vrs1 of barley encoding homeodomain transcription factors (Konishi et al., 2006; Komatsuda et al. 2007) and a SBP gene underlying the teosite glume architecture1 (tga1) locus in maize (Wang et al., 2005). Gain-of-function alleles of genes encoding other transcription factors also have been identified. These are illustrated with the classical maize tb1 and wheat Rht1 genes involved in plant architecture, whose selected alleles show mutations in a cis-regulatory region or in the coding region, respectively (Doebley et al., 2006; Peng et al., 1999; Clark et al., 2006). In tomato, fruit size and shape strongly differ from the small and round fruits of ancestral wild species. Detailed QTL analyses (reviewed in Tanksley, 2004) have been followed up with the cloning of four major effect loci (Table 1). Thus, FW2.2 and OVATE identified novel classes of genes (Frary et al., 2000; Liu et al., 2002), the FAS locus has recently found to be a Yabby-type cell fate gene (Cong et al., 2008), and SUN has been shown to encode an IQ67 domain protein (Xiao et al., 2008). An ovate nonsense allele causes fruits with a pear shape (Liu et al., 2002), while different regulatory mutations producing misexpression of FW2.2, FAS, and SUN lead to large fruit phenotypes. Furthermore, seed size loci have been isolated in several species, such as the rice GW2 and GS3 loci encoding an ubiquitin-E3-ligase and a transmembrane protein, respectively (Fan et al., 2006; Song et al., 2007), or the tomato Sw4.1 locus encoding an ABC transporter (Orsi and Tanksley, 2009). On the other hand, a large number of QTL analyses have been described in many crops for other developmental and morphological traits related to yield and plant performance (e.g., http://www.gramene.org/ and http://www.sgn.cornell.edu/index.pl). The first genes underlying those loci have been identified mainly in rice (Sakamoto and Matsuoka, 2008; Weng et al., 2008) as illustrated with the Gn1a locus encoding a cytokinin dehydrogenase that affects grain number (Ashikari et al., 2005) or the pleiotropic Ghd7 locus with effects on various yield components (Xue et al., 2008).

Developmental traits related to plant architecture and organ morphology also have been studied in A. thaliana, where natural variation for seed size (Alonso-Blanco et al.,1999), inflorescence architecture (Juenger et al., 2000), leaf morphology (Pérez-Pérez et al., 2002), and flower size (Juenger et al., 2005) have been described. Although none of the underlying genes have been cloned, microsynteny comparisons between tomato and A. thaliana, combined with functional analyses, suggest that a Sw4.1 ortholog gene might underlie an A. thaliana seed size QTL (Orsi and Tanksley, 2009). Several QTLs have been detected for root growth and architecture (Mouchel et al., 2004; Loudet et al., 2005; Fitz Gerald et al., 2006), and the major effect locus BRX has identified a novel gene with unknown molecular function (Mouchel et al., 2004). In addition, developmental features present as monogenic traits in some A. thaliana accessions have identified non-sense mutations in the Myb transcription factor gene GL1 causing glabrous phenotype (Hauser et al., 2001) and in the β-galactosidase gene corresponding to the MUM2 locus, which reduces seed mucilage liberation after imbibition (Macquet et al., 2007).

Different strategies are revealing cryptic developmental genetic variation of A. thaliana that is only apparent under specific conditions (Table 1). It has been shown that differential expansion of an intronic trinucleotide repeat affects the expression of a Leu biosynthesis enzyme gene, hence causing aberrant leaf morphology defects that are mainly visible under short-day photoperiod and high temperature conditions (Sureshkumar et al., 2009). In addition, A. thaliana transgenic lines carrying the SRK and SCR genes that determine self-incompatibility in A. lyrata have been used to reveal a gene underlying pseudo-self-compatibility. Thus, it has been shown that polymorphisms in the promoter region of the previously uncharacterized gene PLANT U-BOX8 regulating SRK transcript levels reduce its expression and cause such compatibility (Liu et al., 2007). These examples show how the analysis of morphological and developmental traits in crops and wild plants are identifying not only genes with allelic variants increasing fitness under particular conditions, but also cryptic genetic variation with unknown effects on plant adaptation.

VEGETATIVE GROWTH AND PHYSIOLOGY

Natural variation has been described extensively for general growth traits (e.g., biomass accumulation) as well as primary metabolism and mineral uptake and accumulation, and the complex molecular bases underlying these traits are beginning to be elucidated. Such studies are aiding our understanding of hybrid vigor (positive heterosis) and hybrid incompatibility (negative heterosis), which are of primary concern in agronomy, as well as specific metabolic pathways and physiological processes that affect crop growth and/or crop quality.

Growth and Biomass Accumulation

The complexity underlying the regulation of biomass accumulation is illustrated with the precise dissection of a small-effect QTL for this trait in A. thaliana, which includes two closely linked QTLs that have opposing effects on biomass accumulation within a small genomic region of 210 kb (Kroymann and Mitchell-Olds, 2005). In addition, the effect of both loci depends on the genetic background. Thus, a highly polygenic architecture involving genetic interactions has been speculated for biomass accumulation (Kroymann and Mitchell-Olds, 2005). Recently, significant progress has been made in understanding the genetic and molecular bases of growth traits that show positive or negative heterosis, as well as in finding genes contributing to the variation for growth interactions with abiotic environmental factors.

A large number of QTL studies are currently dissecting the genetic bases of positive heterosis or hybrid vigor, which is a growth trait observed in many hybrids of most species and a major plant breeding objective. Numerous QTLs with different levels of dominant, overdominant, and epistatic effects have been mapped for heterosis in maize (Frascaroli et al., 2007), rice (Li et al., 2001), tomato (Semel et al., 2006), rapeseed (Brassica napus) (Radoev et al., 2008), and A. thaliana (Kusterer et al., 2007; Melchinger et al., 2007). Comparisons of genomic sequences among maize inbred lines have shown extensive within-species lost of gene colinearity, with a large portion of genes being absent in some individuals (Fu and Dooner, 2002; Song and Messing, 2003). This lack of colinearity has led to a molecular hypothesis for the dominance model of heterosis and its counterpart inbreeding depression, where the presence of distinct genes in different lines might cause hybrid vigor (Fu and Dooner, 2002). In addition, overdominance frequently has been detected at the level of gene expression in hybrids of maize and A. thaliana (Song and Messing, 2003; Vuylsteke et al., 2005).

On the other hand, analyses of A. thaliana hybrids are identifying the molecular bases of a related fundamental evolutionary phenomenon, namely, negative heterosis or hybrid incompatibility (Table 1). Nearly 2% of A. thaliana intraspecific crosses yield hybrid plants expressing severe growth defects and leaf necrosis, which are due to epistatic interactions between two or three genes fitting the Dobshansky-Muller model of hybrid postzygotic isolation (Bomblies et al., 2007; Alcázar et al., 2009). Alleles of two different tandem repeats of NB-LRR disease resistance genes differing in sequence and gene copy number have been shown to cause these hybrid incompatibilities (Bomblies et al., 2007; Alcázar et al., 2009). Incompatible combinations are characterized by autoimmune-like responses and enhanced disease resistance, which depend on growth temperature and activation of the salicylic acid stress signaling pathway (Alcázar et al., 2009). In addition, embryo lethality segregating in crosses among wild A. thaliana genotypes is determined by an epistatic interaction between two paralogous genes encoding a key enzyme involved in His biosynthesis (Bikard et al., 2009). Thus, divergent evolution of particular duplicated genes appears as the main mechanism causing genetic incompatibilities at different postzygotic levels.

Several QTLs have been isolated that contribute to natural variation for growth responses to light or soil nutrient availability in A. thaliana. Analyses of light responses have shown that amino acid substitutions in photoreceptors, including phytochrome A (Maloof et al., 2001), phytocrome B (Filiault et al., 2008), cryptochome 2 (Botto et al., 2003), and probably phytochrome C (Balasubramanian et al., 2006), affect hypocotyl elongation in different light conditions. In addition, a tandem zinc knuckle/PLU3 (TZP/LIGHT5) domain-encoding gene has been shown to promote growth, associated with an insertion mutation appearing as a rare natural loss-of-function allele (Loudet et al., 2008). Furthermore, it has been shown that natural polymorphisms affecting the expression of LPR1, which encodes a multicopper oxidase, strongly influences the primary root growth response to phosphate starvation (Svistoonoff et al., 2007). Hence, it appears that mainly signal perception and transduction components have been involved in natural variation for growth environmental responses.

Primary Metabolism

Primary metabolism has been defined as “those essential reactions involving compounds that are formed as part of the normal anabolic and catabolic processes, which result in assimilation, respiration, transport, and differentiation processes that take place in most, if not all, cells of an organism” (Fernie and Schauer, 2009). Besides a wide range of intermediate compounds (e.g., compounds from glycolysis or Krebs cycle), primary metabolites include end products of metabolic pathways that accumulate in (harvestable) sink organs (e.g., seeds, fruits, and tubers) determining relevant crop quality traits related to nutritional content and composition. Recent molecular analyses of natural variation have identified a large number of genes encoding enzymes and have associated primary metabolism with phenotypic variation for morphology and growth-related traits (Lisec et al., 2008; Schauer et al., 2008).

Two decades ago, analyses of starch content in seeds of pea and cereals identified the genes underlying classical loci, such as Mendel's pea rugosus (R) and maize Waxy. These pioneering works illustrate the pleiotropic effects of primary metabolism on plant development. The R locus encodes a starch branching enzyme, whose recessive allele carrying a transposon insertion in the coding sequence has been widely spread in European cultivars due to the associated sweetness of the wrinkled pea phenotype (Bhattacharyya et al., 1990). Maize Waxy loci encode granule-bound starch biosynthesis enzymes, where multiple loss-of-function alleles caused by transposon insertions lead to a waxy appearance of kernels that reflect the lack of amylose in endosperm (Varagona et al., 1992). Moreover, a single splice donor site mutation in a homologous Waxy gene is responsible for the absence of amylose in glutinous rice varieties, which has been widely spread in East Asia associated with human cultural behaviors (Olsen and Purugganan, 2002).

Recent quantitative analyses of natural variation for the content of primary metabolites in fruits of tomato interspecific crosses (Schauer et al., 2006) and in potato tubers (Solanum tuberosum; Menéndez et al., 2002) have found a large number of QTLs underlying the content of other carbohydrate metabolites. Isolation of a large-effect QTL for tomato sugar content identified the invertase encoding gene LIN5, where a single amino acid substitution alters its activity (Fridman et al., 2004). Based on colocation and genetic association, two tandemly repeated invertase genes, invGE and invGF, have been also suggested to underlie a cold-sweetening QTL in potato (Li et al., 2005).

Several studies have dissected the variation for primary metabolism of A. thaliana through the analysis of enzymatic activities and metabolite contents in different vegetative organs. QTLs have been found for the activity of multiple carbohydrate metabolism enzymes, some of them colocating with the corresponding structural genes, for example, phosphogluco-mutase (PGM) and hexokinase, which indicates cis-regulatory variation (Mitchell-Olds and Pedersen, 1998; Keurentjes et al., 2008b). Histochemical analysis further shows that certain PGM QTLs are tissue specific (Sergeeva et al., 2004). However, several enzymatic activity QTLs are located in different genomic regions, identifying trans-acting sequences. These loci might simultaneously coregulate several enzymes, hence revealing novel modes of carbohydrate metabolic regulation (Mitchell-Olds and Pedersen, 1998; Sergeeva et al., 2004; Keurentjes et al., 2008b). In addition, a vacuolar invertase has been found that contributes specifically to the natural variation for root elongation but not for hypocotyl length (Sergeeva et al., 2006). Similarly, QTLs have been identified for nitrogen- and sulfur-containing compounds, such as nitrate, sulfate, and free amino acids (Loudet et al., 2003, 2007). A large-effect QTG has been isolated for sulfate content, which corresponds to the APR2 sulfate reductase gene (Loudet et al., 2007). APR2 activity is altered by a single amino acid substitution, and the effects of this polymorphism interact with nitrogen availability, hence illustrating the interaction between sulfate assimilation and nitrogen metabolism. Furthermore, as described in the previous section, genes involved in Leu or His biosynthesis account for natural variation for severe zygotic (embryo) and postzygotic morphological defects (Bikard et al., 2009; Sureshkumar et al., 2009).

A. thaliana natural variation for untargeted metabololites from vegetative tissues has been dissected by quantifying high-throughput metabolic profiles in different mapping populations (Keurentjes et al., 2006; Meyer et al., 2007; Rowe et al., 2008). A large number of metabolite content QTLs (mQTLs) have been mapped, demonstrating that two- and three-way epistatic interactions determine a substantial proportion of this variation (Lisec et al., 2008; Rowe et al., 2008). Candidate genes have been proposed for a large proportion of detected mQTL based on the enzymatic activities involved in different biochemical pathways (Lisec et al., 2008). In addition, comparative analyses have shown colocation of QTLs for biomas accumulation and mQTL, supporting a link between growth and metabolic profiles (Lisec et al., 2008).

Mineral Accumulation

Most nutrients that plants need for growth and development are supplied as minerals to the roots, and they are classified as macronutrients (Ca, K, Mg, N, P, and S) or micronutrients (B, Cl, Fe, Mn, Co, Cu, Mo, Ni, and Zn) depending on the necessary quantities. The composition of mineral nutrients and trace elements (i.e., the inorganic component of an organism) is now referred to as the ionome (Salt et al., 2008). There is substantial natural variation for mineral use efficiency, root uptake, translocation from roots to shoots, and accumulation in the seed as storage and supply for the germinating seedling. This variation has been reported in many species, leading to breeding programs such as those aiming to improve zinc and iron status of cereal grains or tuber crops (www.harvestplus.org). Recent QTL analyses have identified six genes involved in the complex mechanisms controlling the variation for plant nutrition and mineral homeostasis (Table 1; Ghandilyan et al., 2006)

QTL studies in crop species have focused on seed or leaf mineral concentrations, mostly in rice, wheat, barley, Brassica species, and soybean (Glycine max). In cereals, mostly rice, QTLs have been mapped for grain minerals (Joppa et al., 1997; Garcia-Oliveira et al., 2009), flag leaf nitrogen (N) (Ishimaru et al., 2001) or phosphorous (P) uptake (Wissuwa et al., 2002), salt tolerance related to Na/K homeostasis (Lin et al., 2004), and grain accumulation of toxic Cd (Ueno et al., 2009). In Brassica crops, loci have been reported for phosphate concentrations in leaves and seeds of B. rapa (Zhao et al., 2008), for Ca and Mg concentrations in leaves of B. oleracea (Broadley et al., 2008), and for leaf mineral concentrations in B. rapa (Wu et al., 2008). In soybean, QTLs have been found for Ca and S concentrations in seeds (Panthee et al., 2006; Zhang et al., 2009). Three large-effect QTLs have been isolated in crops, which have identified several molecular mechanisms linking mineral content and other physiological traits. The rice SKC gene encodes a Na⁺ selective transporter, whose more active allele determined by missense mutations increases salt tolerance (Ren et al., 2005). A boron tolerance locus of barley corresponds to the boron efflux transporter gene Bot1 that appears duplicated and overexpressed in a B tolerant accession (Sutton et al., 2007). In addition, the Gpc-B1/NAM-B1 locus of wheat encodes a NAC transcription factor that regulates nutrient remobilization from leaves to developing grains (Uauy et al., 2006). It has been shown that nonfunctional alleles caused by different deletions are present in cultivated varieties, leading to delayed senescence and reduced protein, Zn, and Fe grain content.

Detailed analyses of the ionome in A. thaliana have shown considerable variation for leaf mineral concentrations under various mineral/metal supply conditions (Salt et al., 2008). QTLs have been identified for accumulation of different elements (Ca, Cu, Fe, K, Mg, Mn, Na, P, S, and Zn) in seeds, siliques, leaves, and roots under different growth conditions (Vreugdenhil et al., 2004; Waters and Grusak, 2008; Ghandilyan et al., 2009). Overall, these analyses show that mineral concentrations are controlled by a large number of loci with mostly moderate effect, which often interact strongly with growth environment. Interestingly, a strong P QTL has been colocated with a previously identified locus for phytate (myo-inositol-1,2,3,4,5,6-hexakisphosphate or IP6), the storage compound for P in seeds (Bentsink et al., 2003). However, only the ERECTA gene, encoding a receptor-like kinase for which a mutant allele segregates in mapping populations, has been demonstrated to affect mineral concentrations in these studies (Waters and Grusak, 2008; Ghandilyan et al., 2009).

In addition, A. thaliana QTL analyses have been focused on accumulation of specific minerals, including N (as nitrate; Loudet et al., 2003; Harada et al., 2004), K (Harada and Leigh, 2006), Cu (Kobayashi et al., 2008), Mo (Baxter et al., 2008), and Na (Rus et al., 2006). These studies have led to the isolation of three genes underlying large-effect QTLs, which also encode different mineral transport components. A root copper tolerance locus corresponds to the HMA5 gene encoding a Cu-transporting ATPase. Several natural alleles differing in missense mutations in conserved motifs show lower activity and Cu translocation to the shoot (Kobayashi et al., 2008). A mitochondrial molybdenum transporter encoded by the nuclear MOT1 gene underlies a shoot Mo concentration QTL. A deletion in MOT1 promoter region has been associated with low gene expression and low shoot Mo concentration, suggesting that this regulatory mutation is the causal nucleotide polymorphism (Baxter et al., 2008). Finally, HKT1 encodes a Na⁺ transporter for which two loss-of-function alleles associated with promoter deletions produce lower root expression and enhanced shoot Na⁺ levels in two coastal accessions (Rus et al., 2006).

Studies of A. thaliana mineral accumulation have inspired research on natural variation for exceptional zinc and/or cadmium accumulation in leaves of the related metal hyperaccumulator species Arabidopsis halleri and Thlaspi caerulescens. Several loci have been mapped in populations derived from crosses between A. halleri and A. lyrata (Courbot et al., 2007), one of them colocalizing with HMA4, a locus containing three copies of the Ah-HMA4 gene that is essential for Zn hyperaccumulation and Cd tolerance in A. halleri (Hanikenne et al., 2008). In addition, multiple QTLs have been located for root or shoot Zn or Cd accumulation in F3 populations of T. caerulescens (Assunção et al., 2006; Deniau et al., 2006). Thus, comparative analyses with A. thaliana are not only assisting the isolation of genes underlying those loci but will also enable identification of the molecular bases of interspecific variation present in wild crucifers.

REVEALING THE VARIOUS LEVELS OF PHENOTYPIC REGULATION: NATURAL VARIATION FOR MOLECULAR TRAITS

The path from genotype to phenotype includes several molecular intermediate steps in the translation of genetic information, with regulation occurring at each of these levels (Keurentjes et al., 2008a). Extensive natural variation has been described for the amount of molecular components, such as gene transcripts (Vuylsteke et al., 2005; Kliebenstein et al., 2006a), protein abundance and activity (Chevalier et al., 2004; Cross et al., 2006), and metabolites (Kliebenstein et al., 2001; Cross et al., 2006). Quantitative genetic analyses of “-omics” data collected from mapping populations by so-called genetical genomics approaches (Jansen and Nap, 2001) have shown that much of this natural variation is genetically determined by specific expression QTL (eQTL), mQTL, or protein QTL (pQTL) (Keurentjes et al., 2008a). In addition, a portion of this variation is determined by epigenetic mechanisms of regulation, such as methylation and chromatin remodeling (He et al., 2004; Vaughn et al., 2007), which in turn can also result from DNA sequence polymorphisms (Johannes et al., 2008).

The genetic bases of natural variation for genome-wide gene expression have been approached systematically in A. thaliana, barley, and maize (Schadt et al., 2003; Vuylsteke et al., 2006; Keurentjes et al., 2007; West et al., 2007; Potokina et al., 2008b). Expression variation of ;30 to 40% of the segregating genes is regulated by an eQTL colocating with the gene encoding the mRNA, most of those local eQTLs being the result of cis-regulated genes (i.e., allele-specific expression). However, most variably expressed genes are regulated in trans by eQTLs located in distant genomic regions (trans-eQTLs) that often appear concentrated in a small number of regulatory hot spots (Keurentjes et al. 2007; West et al., 2007; Fu et al., 2009). cis-eQTLs show larger average effects than trans-eQTLs probably due to the stronger impact of direct cis-regulation on the target gene. In addition, it has been shown that QTLs for gene expression variation might depend on spatial and temporal regulation (Potokina et al., 2008a). The molecular identification of regulatory eQTLs is currently facilitated by statistical methods that combine the functional annotation of biological relevant gene sets, the genomic position of genes and eQTLs, and the coexpression of candidate regulators and target genes. In this way, complex genetic networks are constructed, which reveal many novel putative regulatory steps together with strong interconnectivity and epistasis (Kliebenstein et al., 2006b; Keurentjes et al., 2007; Druka et al., 2008). As described above, large-scale untargeted analyses of metabolic profiles have also revealed extensive genetic variation for metabolite content in A. thaliana and tomato (Keurentjes et al., 2006; Schauer et al., 2006; Lisec et al., 2008). Large numbers of mQTLs have been mapped, although they show a lower average heritability than eQTL (Rowe et al., 2008).

The genetic dissection of natural variation for developmental and physiological traits is beginning to link phenotypic QTLs for complex traits with eQTL, mQTL, and pQTL (Keurentjes, 2009). Several studies, mostly in A.thaliana, have integrated information from different levels of molecular regulation to improve the elucidation of genetic control of quantitative traits (Hirai et al., 2004; Tohge et al., 2005; Meyer et al., 2007; Wentzell et al., 2007; Druka et al., 2008; Lisec et al., 2008). Thus, it has been shown that plant growth and development are interwoven with primary metabolism, regulation occurring at different levels, including gene expression, enzyme activity, and metabolite abundance (Keurentjes et al., 2008b). Although these studies often reveal direct relationships between the various levels of biological information, such a relationship is not always linear or present. This was illustrated recently with a system-wide analysis, wherein much of the natural genetic variation affecting gene expression or functional protein differences did not to appear to lead to phenotypic variation, a phenomenon known as phenotypic buffering (Fu et al., 2009). Interestingly, several loci have been identified as contributing to natural variation for the ability to buffer developmental processes through environmental canalization (Hall et al., 2007; Sangster et al., 2008). Hence, integrative analyses of multiple phenotypic and molecular regulatory levels are elucidating the mechanisms controlling developmental buffering and canalization, which are presumed to contribute to biological robustness of plant species and facilitate evolvability in changing environments (Queitsch et al., 2002; Kitano, 2004).

MOLECULAR TARGETS OF SELECTION FOR PLANT ADAPTATION: CURRENT AND FUTURE PROSPECTS

Nearly 100 genes underlying natural variation for plant development and growth have been identified in crop plants and A. thaliana. Most of them correspond to QTLs with large effects on trait variation and provide a first inventory of mechanisms, genes, and functional nucleotide polymorphisms that might be involved in plant adaptation to different agricultural or natural environments. They belong to all types of ontological classes, including transcription factors, signal transduction components, enzymes of primary and secondary metabolism, hormone metabolism and signaling, metal transporters, as well as genes with unknown functions. Interestingly, some of these genes had not been found previously in mutant screenings because common laboratory strains carry natural loss-of-function alleles, such as FLC, FRI, or DOG1 in A. thaliana.

Nucleotide polymorphisms affecting the function of these genes are identifying the mutational mechanisms that generate natural variation. A large proportion of natural alleles carry null loss-of-function mutations, which are often produced by structural nonsense or insertion and deletion (indel) mutations (Alonso-Blanco et al., 2005). A particular class of these alleles corresponds to complete gene deletions or duplications caused by retrotransposition, recombination, and other molecular mechanisms, which may play a major role in heterosis of crop and wild species (Fu and Dooner, 2002; Song and Messing, 2003). A second type of common allele is a change-of-function allele produced by a missense mutation altering protein structure and function, such as described for photoreceptor genes (Table 1). In addition, multiple change-of-function alleles carrying regulatory mutations in three types of genomic regions affecting gene expression have been described for many traits and species. Most of these alleles show indels in their nearby promoter regions, in agreement with the large number of cis-eQTLs identified (Keurentjes et al., 2007; West et al., 2007). Second, gene expression changes are caused by unknown mutations located in distant 5′-regions of the genes as illustrated for maize Tb1 (Clark et al., 2006) and Vgt1 (Salvi et al., 2007). Third, expression alterations also appear to be produced by intronic mutations, including insertions of tranposons or differential expansion of microsatellite repeated sequences (Sureshkumar et al., 2009). Finally, although most natural alleles that have been described are caused by variation at the level of nucleotide sequence, natural epigenetic variation also contributes to heritable developmental and physiological variation, as indicated by stably inherited epialleles in the LCyc gene involved in Linaria flower morphology (Cubas et al., 1999) and in an SBP-box gene affecting tomato fruit ripening (Manning et al., 2006). This is supported by the considerable amount of natural variation found for DNA methylation (Riddle and Richards, 2005; Zhang et al., 2006b; Vaughn et al., 2007), which is partly mediated by small interfering RNAs also differing among accessions (Zhai et al., 2008). It is hypothesized that methylation and heterochromatinization mediated by different kinds of small interfering RNAs might cis- and trans-regulate gene expression and function in a more quantitative and reversible manner than sequence variation (Zhai et al., 2008).

All of these studies provide information on the particular classes of genes and alleles that might have been naturally or artificially selected for different traits (reviewed in Alonso-Blanco et al., 2005). For instance, as described by Purugganan and Fuller (2009), it seems that domestication is mainly associated with transcriptional activators and changes in transcriptional networks. By contrast, crop diversification after domestication involves a large proportion of genes encoding enzymes, with loss-of-function alleles differently contributing to crop domestication and diversification. Thus, we are beginning to understand how natural variation is generated and maintained depending on plant species, traits, environments, and selective forces. However, conclusive generalizations on evolutionary processes await systematic extensions of the following aspects of these analyses: (1) the number of species analyzed, particularly of wild plants; (2) the number of genes identified in each species; (3) the spectrum of allelic effects per gene; and (4) the estimates of genetic and environmental interactions.

Understanding the ecological and evolutionary significance of natural variation requires precise evaluation of their role in adaptation to particular environments, which is just beginning to be approached. Based on their presence in a single wild genotype and their relative large effects, some of the A. thaliana natural alleles are probably deleterious variants segregating in natural populations. This has been suggested for loss-of-function alleles of HUA2 and LIGHT5, found only in a subset of lines derived from the same parental accession (Doyle et al., 2005; Loudet et al., 2008). Similarly, the BRX loss-of-function allele has not been found in genotypes collected in the same geographical site where this variant was originally sampled many years before (Shindo et al., 2008). As described above for flowering time, demonstrating that natural variation in specific genes is involved in adaptation is currently addressed by combining population and evolutionary genetic analyses with detailed phenotypic studies of fitness-related traits. Analysis of the amount and pattern of nucleotide diversity may show potential signatures of selection, hence suggesting different evolutionary mechanisms to maintain such variation. This is illustrated with the identification of selective sweeps in several domestication genes of crop plants (Purugganan and Fuller, 2009) and with the highly differentiated haplotypes found in some A. thaliana photoreceptor genes (Olsen et al., 2004; Balasubramanian et al., 2006). However, demographical history also affects the patterns of DNA sequence diversity; therefore, other studies are necessary to test adaptive effects of such alleles. On one hand, detailed phenotypic characterizations under natural and agricultural conditions are detecting fitness effects of particular alleles or genotypes (Tian et al., 2003; Shindo et al., 2008; Wilczek et al., 2009). In addition, comparisons of differentiation among local populations for quantitative traits (Q_ST) and nucleotide polymorphisms (F_ST) might further assist the detection of selection signatures on particular genes (Le Corre, 2005; Stenoien et al., 2005). On the other hand, association analyses of nucleotidic and phenotypic diversity with environmental variation might identify environmental factors driving phenotypic selection. It is expected that new extensive collections of A. thaliana wild genotypes recently developed from several world regions (Jorgensen and Mauricio, 2004; Stenoien et al., 2005; Schmid et al., 2006; Beck et al., 2008; Picó et al., 2008) will facilitate estimations of allele frequencies, geographical distributions, and genetic associations with environmental factors.

In the past decade, the analysis of natural variation in crop plants and A. thaliana has provided an unprecedented amount of information on the genetic and molecular mechanisms that determine intraspecific variation and adaptation. It can be anticipated that this trend will continue in the next decade, especially with the broad implementation of “-omics” technologies for the precise analysis of natural variation at different levels. First, high-throughput genotyping will enable the development of new genetic mapping strategies, such as genome-wide association (Azanzana et al. 2005; Nordborg and Weigel, 2008; Waugh et al., 2009), which is expected to provide higher mapping resolution. Second, comparisons of thousands of whole-genome sequences from genotypes of the same species will provide a dynamic perspective of a species' genome in terms of pan-genomes (Morgante et al., 2007). Third, common integration of QTLs and genes causing phenotypic variation with those affecting the various levels of molecular variation will provide complete (genome-wide and whole organisms) phenotypic regulatory networks. The latter should include not only eQTL, mQTL, and pQTL estimated at particular points but also molecular fluxes in those networks. Finally, extension of the analysis of natural variation to other crop and wild species will allow deep comparisons among species. Such comparative intraspecific and interespecific studies will elucidate the mechanisms involved in the evolution of plant development and physiology at higher levels, including those of speciation and species diversification in relation to earth changing environments.