Author for correspondence: Christian LexerTel: +44 (0)20 8332 5341Fax: +44 (0)20 8332 5310Email: email@example.com
• We have studied the origin of salt adaptation in wild sunflower hybrids (Helianthus annuus × H. petiolaris), the precursors of the diploid hybrid species H. paradoxus, at the level of phenotypic traits and quantitative trait loci (QTLs). Here, we review this work and present new results on candidate gene polymorphisms.
• Salt tolerance candidate genes were identified in expressed sequence tag (EST) libraries of sunflower, based on homology to genes with known function, and on previous QTL results. EST polymorphisms were assayed by denaturing HPLC and genetically mapped in an interspecific BC2 for which fitness estimates in the wild were available.
• Out of 11 genes studied, one mapped to a salt tolerance QTL. This EST codes for a Ca-dependent protein kinase (CDPK) and stems from stress-induced root tissue of Helianthus annuus. Two additional stress-induced genes exhibited a significant fitness effect in the wild: an ER-type calcium ATPase, and a transcriptional regulator.
• Our results suggest a possible adaptive role for Ca-dependent salt tolerance genes in wild sunflower hybrids. Also, transgressive segregation appears to be sufficient to explain the origin of adaptive genetic variation in hybrids.
Many evolutionary biologists have suggested that hybridization between divergent populations or species may serve as a source of genetic variation upon which natural selection can act (Anderson, 1949; Stebbins, 1959; Lewontin & Birch, 1966; Grant, 1981; Arnold, 1997; Barton, 2001). In this view, hybridization may play a creative role in adaptation and speciation (even beyond its role in polyploid evolution), because it may provide the genetic ‘raw material’ required for the origin of novel adaptations and for the development of reproductive barriers as a byproduct. This view has been bolstered by the discovery of a number of diploid hybrid species that are divergent ecologically from their parental species (Rieseberg, 1997), and also by the ecological success of clonal hybrid genotypes in many taxa (Benzing, 2000; Schweitzer et al., 2002). Supporting evidence also stems from successful attempts to create new species by experimental crosses – such as Verne Grant's classical hybridization work on Gilia (Grant, 1966).
The wild annual sunflowers (Helianthus spp.) are a highly useful group for studying the role of hybridization in evolution, since three species in this genus are known to be of diploid hybrid origin, stemming from hybridization events between the same two parental species, Helianthus annuus and H. petiolaris (Rieseberg et al., 1990; Rieseberg, 1991). Diploid or ‘homoploid’ hybrid speciation represents a particularly intriguing case of plant speciation from a geneticist's point of view, because the entire speciation process appears to be very rapid (10–60 generations; Ungerer et al., 1998; Buerkle et al., 2000), and because novel adaptations leading to niche divergence are likely to arise very early in the speciation process (Buerkle et al., 2000). This allows geneticists to ‘simulate’ plant speciation events by experimental crosses (Rieseberg et al., 1996; Rieseberg, 2000).
Helianthus paradoxus, the Pecos or ‘Puzzle’ sunflower, is one of the three Helianthus hybrid species. This species is adapted to extremely saline marshes (Heiser et al., 1969; Rogers et al., 1982), whereas both of its parents are salt sensitive (Welch & Rieseberg, 2002a). Although the phenotypic traits responsible for salt adaptation are superficially cryptic (i.e, they do not necessarily manifest themselves in measurable morphological characters), salt tolerance is amenable to the study of adaptation. This is due to the rich body of literature that exists on the physiological and molecular basis of salt tolerance in plants (e.g. reviews by Flowers et al., 1986; Cheeseman, 1988; Yeo, 1998; Hasegawa et al., 2000a,b), which provides the necessary mechanistic background for designing and interpreting experiments involving wild species. Below, we briefly review what is known about the origin of salt adaptation in H. paradoxus both from growth chamber studies and from phenotypic selection and quantitative trait locus (QTL) experiments in the wild.
Early growth chamber studies of ecological divergence in H. paradoxus compared the fitness of the hybrid species and its parents at three different sodium chloride (NaCl) concentrations: 0 mm, 100 mm, and 200 mm (Welch & Rieseberg, 2002a). These studies revealed that H. paradoxus was more than five times as fit, in terms of biomass and survivorship, than either of its parental species when grown under salt stress. The experiments also pointed to greater leaf succulence and leaf sodium sequestration as possible mechanisms for salt tolerance in H. paradoxus (Welch & Rieseberg, 2002a), although the latter explanation proved to be too simplistic when the salt-stress response of hybrids and their parents was studied in the wild (Lexer et al., 2003a).
A field experiment, in which segregating second generation (BC2) backcross hybrids, H. annuus (parent), H. petiolaris (parent), and H. paradoxus were transplanted into the natural salt marsh habitat of the latter, confirmed the greater salt tolerance of H. paradoxus, but the exact relationships between candidate adaptive traits and fitness turned out to be more complex than suspected (Lexer et al., 2003a). While positive directional selection was detected for leaf succulence, the relationship between elemental uptake and fitness was surprisingly different from the growth chamber study, reflecting the mineral ion-richness of the soil in H. paradoxus salt marsh habitat. Increased leaf Ca uptake increased fitness (Fig. 1a), whereas the uptake of Na and other toxic mineral ions all reduced fitness in the salt marsh (Fig. 1b). Interestingly, trait correlations decreased between Ca and Na uptake during the course of the experiment (Fig. 1c), suggesting that increased Ca uptake, coupled with greater net exclusion of Na and related ions, contributed to salt tolerance in early generation hybrids and the natural hybrid species alike (Lexer et al., 2003a). Notably, the ranges of several candidate adaptive traits were greater in the BC2 hybrids than in either parental species, and the phenotypes of many BC2 plants overlapped with those of H. paradoxus (Lexer et al., 2003a). Thus, natural selection should be able to assemble the H. paradoxus phenotype from an ancestral hybrid population within a few generations.
The latest in our series of experiments was a QTL genome scan that extended the field study described above by assaying the transplanted BC2 hybrids for mapped molecular markers that span most of the sunflower genome, followed by searches for correlations between fitness and segregating QTLs (Lexer et al., 2003b). Fourteen QTLs were detected for mineral ion uptake traits and three for survivorship in the salt marsh. Notably, the survivorship QTLs mapped coincident with QTLs for increased Ca uptake and/or for exclusion of Na and related toxic mineral ions, thus confirming an important role of mineral ion uptake traits in wild Helianthus (Lexer et al., 2003a,b). Perhaps even more importantly, QTLs with effects in opposing directions were detected for all mineral ion uptake traits with more than one detected QTL. This suggests that the ancestral hybrid population that gave rise to H. paradoxus likely did so by combining complementary genes from both parental species as suggested by theory (de Vincente & Tanksley, 1993; Rieseberg et al., 1999). Finally, selection coefficients (s) for individual QTLs were sufficiently large (s = 0.09–0.13) to account for the origin of H. paradoxus in the presence of gene flow with its parental species, that is, our estimates for s were far greater than the typical migration rate (m) for sunflower (Lexer et al., 2003b), which is a requirement for divergence in sympatry. Furthermore, the correlation patterns among QTLs expressed in the wild allowed us to define classes of possible candidate genes for salt tolerance in wild Helianthus (Lexer et al., 2003b).
The experimental work presented here extends our phenotypic selection and QTL analyses by assaying the same interspecific BC2 for sequence polymorphism in candidate genes identified in expressed sequence tag (EST) libraries. The resulting data allowed us to determine whether any of the candidate genes maps to a QTL of interest, and to identify additional genomic regions (i.e. QTLs) associated with salt tolerance. In addition, where possible, we compared EST sequences from H. paradoxus with those of H. annuus to determine whether any of the candidate genes have undergone significant adaptive evolution since the origin of H. paradoxus more than 75 000 years bp (Welch & Rieseberg, 2002b).
Materials and Methods
Database searches and primer design
We searched the H. annuus EST sequence database of the Compositae Genome Project (http://cgpdb.ucdavis.edu/) for candidate genes belonging to one of the three classes of genes that might contribute to salt tolerance: genes involved in Ca fluxes across membranes; Ca sensors; and K/Na ion channels as down-stream targets of Ca signals. Our search strategy made use of sequence comparisons with genes of known function (BLAST X searches; Altschul et al., 1997), available from the above website. Primers were designed with the help of sequence contigs from two different H. annuus genotypes that are currently employed in a parallel EST mapping project (Z. Lai & L. H. Rieseberg, unpublished).
Assays of candidate gene polymorphism
Candidate ESTs were PCR amplified using reaction conditions identical to those reported previously for microsatellites in our lab (Burke et al., 2002), employing a standard ‘touch-down’ cycling program with a final annealing temperature (Ta) of 48°C for each locus. Initially, amplification success for each locus was assessed by electrophoresis on 2% agarose gels and ethidium bromide staining. For loci that amplified successfully, PCR products were analyzed by reverse phase ion-pairing liquid chromatography on a ‘WAVE’ dHPLC (denaturing High Performance Liquid Chromatograph) manufactured by Transgenomic, Omaha, NE, USA. This method has the potential to detect single nucleotide polymorphisms (SNPs) in DNA fragments as large as 1 Kb (Oefner & Underhill, 1995) and is therefore well suited for detecting polymorphisms in ESTs. Candidate gene polymorphism was analyzed under partially denaturing conditions, using theoretical melting temperatures for each studied EST as a guide. Partially denaturing conditions allow the detection of sequence polymorphisms that do not result in length differences. In a backcross model like the BC2 employed here, heterozygous genotypes are indicated by the presence of a pattern consisting of homo- and heteroduplex bands (maximally 4 bands; Fig. 2). For one gene, CAT1, analysis under nondenaturing conditions (50°C) was sufficient to detect the presence of an insertion/deletion polymorphism.
Linkage and association analysis
The candidate genes were mapped to an existing QTL framework map (Lexer et al., 2003b) using recombination estimates within a BC2 model in Mapmanager QTX (Manly et al., 2001). EST polymorphisms were placed next to their closest framework marker using the ‘links report’ in Mapmanager QTX, which performs a two-point linkage analysis with each framework marker. A log of odds (LOD) value of 7.00 or more was considered as evidence for linkage. Several loci had affinities to more than one linkage group, due to genome-wide disequilibria typical for crosses between these two species (Rieseberg et al., 1996). Nonetheless, each candidate gene locus clearly mapped to a single linkage group based on at least a 1000-fold likelihood difference.
Associations between candidate gene polymorphisms and fitness in the salt marsh (survivorship on a relative scale; Lexer et al., 2003a,b) were tested by linear marker regressions (Tanksley, 1993). For each regression, three background markers were included in the analysis as cofactors, in order to account for background genetic variation elsewhere in the genome, in analogy to the inclusion of cofactors in Composite Interval Mapping (CIM; Zeng, 1993, 1994). Cofactors were chosen manually on the basis of linear regressions for each framework marker, as described previously (Lexer et al., 2003b). The resulting P-values from linear marker regressions were adjusted for the number of candidate genes tested, using the sequential Bonferroni method (Rice, 1989).
Putative homologues of mapped H. annuus ESTs were identified in the H. paradoxus sequence database by FASTA searches, making use of hit IDs in public sequence databases as additional identifiers (note that the H. paradoxus data set became available only very recently – after the candidate gene mapping was complete). For sequences with > 60% identity, pairwise alignments were attempted between the mapped sequences from H. annuus and their counterparts from H. paradoxus. Sequence editing, pairwise alignments, and analyses of synonymous vs nonsynonymous base substitutions were conducted with the computer program PROSEQ (Filatov, 2002).
Mapping of candidate gene polymorphisms
In total, 36 salt tolerance candidate genes were identified in EST libraries of H. annuus. Fifteen of these were successfully amplified in the interspecific BC2 between H. annuus and H. petiolaris, and 13 were polymorphic in our mapping population. For two genes, the dHPLC patterns were too complex for confident mapping, but the remaining 11 genes were successfully placed on the existing QTL framework map (Table 1). One gene, CDPK3, was duplicated, and both loci were successfully mapped.
Table 1. Salt tolerance candidate genes mapped in an interspecific BC2 population between Helianthus annuus and H. petiolaris
Sequence names, gene description, expressed sequence tag (EST) tissue source, species and E-value of best hit, putative functional role, BC2 map location, and fitness effect in H. paradoxus salt marsh habitat are given for each gene. 1Sequence IDs from the Compositae EST database (http://cgpdb.ucdavis.edu). 2For simplicity, provisional gene names from our laboratory are used throughout the text. 3CDPK3 exhibited two independently segregating dHPLC patterns and was therefore mapped in the form of two independent loci. *, P < 0.1; **, P < 0.05; ***, P < 0.01; ns, not significant after correction for multiple tests.
The mapped candidate genes belong to one of the following classes: genes involved in Ca fluxes across membranes (= Ca uptake/transport); Ca sensors or stress signaling genes; or genes involved in K or Na uptake (Table 1). In addition, a putative transcriptional regulator (candidate gene 2456; Table 1) as well as two homologues of the stress signaling candidate FIERY 1 (genes 2180 and 2238; Table 1) were included because of their potential role in stress response, or because their map location in H. annuus suggested that they might be of interest (Z. Lai & L. H. Rieseberg, unpublished). Consistent with a possible role in salt stress response, seven out of the 11 mapped ESTs originated from stress-induced tissue, while one was isolated from germinating seeds and one from hulls. For two ESTs, no information about the tissue source was available, because their tissue ‘tag’ was not detectable after high-throughput sequencing.
Candidate gene–fitness associations
Three significant candidate gene–fitness associations were detected among the 11 mapped genes. One of these genes mapped to a previously detected salt tolerance QTL on linkage group 4 (CDPK3; Fig. 3; Table 1), while the two remaining ones mapped to linkage groups that had not been associated with salt tolerance previously (Table 1). Low marker densities in these regions may explain why no significant QTLs were found there before. The association between candidate gene 2456 and fitness is only marginally significant when corrected for multiple tests (Table 1), however, one of its adjacent microsatellites also displayed a significant fitness association, hence the results likely indicate the presence of a QTL. By mapping candidate gene polymorphism onto the existing QTL map, we appear to have detected two additional genomic regions involved in fitness differences in the salt marsh. Whether the respective EST polymorphisms play a causal role in salt tolerance remains to be determined.
In order to assess a possible role for transgressive segregation through complementary gene action, we examined the directions of fitness effects of H. annuus-derived candidate gene alleles in the BC2. Out of three candidate gene loci with significant fitness associations, two had H. annuus-derived alleles with a negative fitness effect (CDPK3 and ATP1), whereas for one of them the H. annuus allele had a positive effect on fitness in the salt marsh (2456). If the salt tolerance QTLs detected previously are combined with those detected here, two QTLs derived from H. annuus increase and three decrease fitness. Hence, our results are in concordance with a genetic model for transgressive segregation of fitness-related traits involving complementary gene action (Rieseberg et al., 1999).
Sequence comparisons H. annuus/H. paradoxus
By comparing candidate gene sequences from H. annuus to their putative homologues from H. paradoxus, we attempted to test for a possible role for adaptive evolution at these loci in H. paradoxus following its origin. Unfortunately, only two candidate genes had likely homologues in the H. paradoxus EST libraries. For one of them, CDPK2, a meaningful sequence alignment was not possible because of a high number of indels, and the presence of numerous related sequences in the salt stress-induced H. paradoxus library (not shown), indicative of the presence of multiple expressed members of the CDPK gene family. For the other gene, ATP1, we were able to align 383 basepairs of the respective cDNAs from H. annuus and H. paradoxus (Fig. 4). The alignment revealed six nucleotide substitutions, one of which resulted in an amino acid replacement (Fig. 3, position 83). Hence, the ratio of nonsynonymous vs synonymous substitutions (Ka : Ks) of 0.2 is more consistent with purifying than directional selection.
Hybridization as a source of adaptive genetic variation
There are two main pathways by which hybridization is thought to contribute to adaptive evolution: adaptive trait introgression and hybrid speciation (Barton, 2001). Unfortunately, the former process is difficult to detect because novel mutations tend to spread quickly to fixation if they are strongly advantageous (Slatkin, 1987; Pialek & Barton, 1997), and the latter process appears to be infrequent (Rieseberg, 1997). Thus, examples where hybridization does appear to be associated with ecological divergence, such as in H. paradoxus, represent rare and valuable models for studying this process.
The candidate gene analysis described here illustrates the possible contribution of hybridization to adaptive evolution in wild sunflowers. It is noteworthy that nine of the 11 candidate ESTs mapped in the present study did not exhibit sequence polymorphism in intraspecific crosses of H. annuus (Z. Lai and L. H. Rieseberg, unpublished). Our ability to map these genes in the interspecific BC2 population (Table 1) likely stems from increased levels of heterozygosity at these candidate gene loci, created by hybridizing the two parental species, H. annuus and H. petiolaris. Presumably, the same happened in the ancestral hybrid population that gave rise to H. paradoxus > 75 000 yr ago (Welch & Rieseberg, 2002b), thus furnishing the hybrid neospecies with the genetic variation required for the evolution of salt marsh adaptation by natural selection. The fact that three of the mapped candidate genes in the present study were associated with fitness in the salt marsh (Table 1), and that one of them was linked to a previously detected salt tolerance QTL (Fig. 3), further increases the illustrative value of our study.
Transgressive segregation and the origin of salt adaptation in H. paradoxus
A second result of our study relates to the mechanism by which novel variation may arise via hybridization. A simple and plausible model is transgressive segregation by ‘complementary gene action’, that is, the additive action of adaptively important alleles that are inherited from both parents and recombined in hybrids, assuming these alleles are fixed or nearly fixed in one or the other parental population (de Vincente & Tanksley, 1993; Rieseberg et al., 1999). This mechanism is thought to generate extreme or ‘transgressive’ phenotypes in hybrid populations (Rieseberg et al., 1999), thereby allowing hybrid lineages to colonize new niches and become reproductively isolated from their parents (Rieseberg et al., 1999, Lexer et al., 2003a,b). Transgressive segregation has been observed in many crosses between divergent plant populations or species (Rieseberg et al., 1999), and our own results from H. paradoxus support a role for it in the origin of salt adaptation. That is, the genetic architecture underlying adaptively significant traits in experimental hybrids between the parental species is concordant with the ‘complementary gene action’ hypothesis (Lexer et al., 2003b). This observation is confirmed by our present candidate gene study: out of three H. annuus-derived candidate gene alleles with significant fitness associations, two decrease survivorship in the salt marsh (CDPK3 and ATP1), while one leads to an increase in survival (transcriptional regulator 2456). A ‘two plus/three minus’ pattern becomes apparent when the salt tolerance QTLs detected previously and those identified here are combined. It is easy to envision how hybrids could combine all the plus alleles or all the minus alleles from both parents, thereby allowing them to colonize new and extreme habitats.
It is noteworthy that an analysis of nonsynonymous vs synonymous substitutions in one of the fitness-related candidate gene polymorphisms, ATP1, failed to detect the signature of positive selection between H. annuus and H. paradoxus (Fig. 4). Of course, it may be that selection is acting on other portions of the gene or its regulatory sequences. However, the lack of evidence for selection is concordant with a role for transgressive segregation, which does not require adaptive evolution following hybrid speciation. That is, extreme phenotypes are assumed to have been produced by combining preexisting parental alleles.
Although our results are consistent with a role for transgressive segregation in the origin of novel adaptation in a wild sunflower hybrid species, they do not provide sufficient evidence that it has indeed facilitated the origin of ecological divergence in wild sunflowers. A more powerful approach is required to either accept or reject a role for transgressive segregation. A suitable approach involves phenotypic and genomic comparisons of ancient and synthetic hybrids, and the results of these experiments are described elsewhere (Rieseberg et al., 2003).
A role for Ca-dependent genes in salt adaptation?
A major caveat of our study is that the BC2 population employed here did not allow us to fine-map genomic regions involved in salt tolerance, since the number of recombination events is not sufficient for this purpose (Lynch & Walsh, 1998). Therefore, we do not know at present if the candidate genes studied here are actually causally related to fitness differences in the salt marsh, although the tissue source for most of the studied ESTs (stress-induced root or shoot tissue) leads us to speculate about such causal links. Even in the absence of further experimental evidence, however, some limited functional considerations about the mapped genes are possible.
Most of the genes included in this study were chosen on the basis of our previous QTL results (Lexer et al., 2003b), which suggested that Ca-dependent genes might be responsible for the observed salt tolerance QTLs. These include genes involved in Ca fluxes across membranes, Ca sensors, or K/Na ion channels as down-stream targets of Ca signals (Lexer et al., 2003b). Although it seems obvious that other signaling pathways should be involved in this stress response as well, such as abscisic acid (ABA) related pathways or other second messengers (Hasegawa et al., 2000a; Hunt & Gray, 2001; Xiong et al., 2002), our greenhouse and field data do suggest a possible role for Ca-mediated stress signaling in wild Helianthus (Lexer et al., 2003a,b).
Our observations are further strengthened by soil analyses which indicate that Ca concentrations are high in H. paradoxus habitat (Lexer et al., 2003a). Thus, it seems likely that a stress response of H. paradoxus will make use of this available resource so important to many aspects of abiotic stress tolerance (Volkmar et al., 1998; Sanders et al., 1999; Hasegawa, 2000a,b; Knight, 2000). Also, the osmotic stress induced by high salinity may trigger plant responses similar to those under drought conditions (Flowers et al., 1986; Xiong et al., 2002), and Ca is known to act as a ‘universal’ signal molecule under both salt and drought stress (Knight, 2000; Xiong et al., 2002). Our finding in the present study that two Ca-dependent genes, CDPK3 (a Ca sensor) and ATP1 (a gene involved in Ca transport) – or the chromosomal segments that carry them – are associated with fitness differences in the wild is in agreement with a Ca-mediated stress response in wild sunflowers.
With respect to CDPKs, such as the one associated with a salt tolerance QTL in interspecific BC2 hybrids (Fig. 3), it is noteworthy that an increased number of CDPK gene family members or isoforms was found in a salt-subtracted EST library of H. paradoxus (12 different unigenes) compared to only two different sequences from untreated tissue (data not shown). This agrees well with a role for CDPKs in salt stress response, because each CDPK isoform is thought to respond to a specific set of Ca signals which differ in frequency of oscillation, magnitude and duration (Harmon et al., 2000). Hence, our EST data are consistent with an important role for CDPKs specifically, and stress-induced Ca-signaling in general.
As alluded to above, an open question is whether the candidate genes studied here are indeed causally related to salt tolerance. Research approaches to clarify this may include fine-mapping in later generation crosses or natural hybrid zones, transgenic complementation, RNA interference (RNAi)- mediated gene silencing (Baulcombe, 1999; Chuang & Meyerowitz, 2000), the analysis of allelic sequence variation with respect to the likely consequences on protein structure and function, and/or gene expression studies based on microarrays (Gibson, 2002). With respect to the latter, a challenging task will be to create experimental conditions for expression studies that go beyond traditional NaCl treatments. Experimental settings for studying gene expression in salt marsh species like H. paradoxus should be designed to ‘simulate’ complex salt marsh soils, while at the same time allowing for controlled manipulation of key parameters. Greenhouse experiments involving different Ca/Na treatments as well as microarray-based expression studies are currently underway in our laboratory.
We thank J. Malcom and G. Warrick (US Fish and Wildlife Service) for their help during field work in New Mexico, and Ji Zou (Rieseberg laboratory) for skillful technial assistance during HPLC analysis. This study was supported by Erwin-Schroedinger grant J-2148 of the Austrian Science Foundation to CL, and by NIH award R01 G59065 to LHR.