SEARCH

SEARCH BY CITATION

Keywords:

  • local adaptation;
  • monkeyflower;
  • osmotic stress;
  • quantitative trait loci (QTLs);
  • reciprocal transplant;
  • salinity;
  • tradeoff;
  • transmission ratio distortion

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • • 
    Local adaptation is a well-established phenomenon whereby habitat-mediated natural selection drives the differentiation of populations. However, little is known about how specific traits and loci combine to cause local adaptation.
  • • 
    Here, we conducted a set of experiments to determine which physiological mechanisms contribute to locally adaptive divergence in salt tolerance between coastal perennial and inland annual ecotypes of Mimulus guttatus. Quantitative trait locus (QTL) mapping was used to discover loci involved in salt spray tolerance and leaf sodium (Na+) concentration. To determine whether these QTLs confer fitness in the field, we examined their effects in reciprocal transplant experiments using recombinant inbred lines (RILs).
  • • 
    Coastal plants had constitutively higher leaf Na+ concentrations and greater levels of tissue tolerance, but no difference in osmotic stress tolerance. Three QTLs contributed to salt spray tolerance and two QTLs to leaf Na+ concentration. All three salt-spray tolerance QTLs had a significant fitness effects at the coastal field site but no effects inland. Leaf Na+ QTLs had no detectable fitness effects in the field.
  • • 
    Physiological results are consistent with adaptation of coastal populations to salt spray and soil salinity. Field results suggest that there may not be trade-offs across habitats for alleles involved in local salt spray adaptations.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The natural landscape contains a heterogeneous array of environments that drive the adaptive differentiation of populations (Linhart & Grant, 1996; Kawecki & Ebert, 2004; Lexer & Fay, 2005; Schemske & Bierzychudek, 2007). Local adaptation to this habitat variation is thought to involve the composite of multiple phenotypic traits, each with a complex genetic basis, that have evolved in response to the mosaic of environmental factors that define habitats (Kawecki & Ebert, 2004; Keurentjes et al., 2008; Karrenberg & Widmer, 2008; Pauwels et al., 2008). Numerous reciprocal transplant experiments have demonstrated that locally adaptive population differentiation is a common phenomenon (Turesson, 1922; Clausen, 1951; Grant, 1981; Linhart & Grant, 1996; Kawecki & Ebert, 2004; Lexer & Fay, 2005; Hereford, 2009). However, most reciprocal transplant studies do not permit the determination of the traits responsible for local adaptation. Further, recent laboratory studies that have discovered loci and even genes putatively involved in particular adaptations (Colosimo et al., 2005; Hoekstra & Coyne, 2007; Stinchcombe & Hoekstra, 2008; Via & West, 2008) rarely test the effects of those loci on fitness under field conditions (Verhoeven et al., 2004, 2008; Gardner & Latta, 2006; Barrett et al., 2008; Stern & Orgogozo, 2008). The relationship between locally adaptive traits, their underlying genetic architecture, and selection in nature remains poorly understood.

Local adaptation is defined as a form of genotype × environment interaction, with genotypes from local populations outperforming foreign transplants (Linhart & Grant, 1996; Kawecki & Ebert, 2004; Lexer & Fay, 2005). How individual loci combine to cause local adaptation is largely unknown (Kawecki & Ebert, 2004; Keurentjes et al., 2008; Karrenberg & Widmer, 2008). One possibility is that local adaptation is mediated by the net effects of loci that perform well in local habitat but are deleterious in foreign habitats (Fry et al., 1998; Kawecki & Ebert, 2004; Gardner & Latta, 2006). Such genetic trade-offs could be caused by linkage to deleterious loci or antagonistic pleiotropy of adaptive loci, especially for traits that incur a physiological cost (Strauss et al., 1999; Kawecki & Ebert, 2004; Roff & Fairbairn, 2007). Alternatively, local adaptation may be caused by combination of loci that individually have fitness effects in one habitat but are effectively neutral in alternative habitats. The few field experiments that have quantified the fitness effects of loci across habitats do not support the hypothesis of local adaptation being caused by trade-offs at individual loci (Weinig et al., 2003; Verhoeven et al., 2004, 2008; Gardner & Latta, 2006).

Coastal perennial and inland annual ecological races of Mimulus guttatus (yellow monkeyflower) occur throughout western North America and are locally adapted to their respective habitats (Hall & Willis, 2006; Lowry et al., 2008). The inland habitat of annual M. guttatus is characterized by the rapid onset of a hot and dry summer drought (Lowry et al., 2008). Inland annual plants escape from this drought through early flowering (Hall & Willis, 2006; Lowry et al., 2008; C. A. Wu et al., unpublished). In contrast to inland habitat, persistent fog maintains low temperatures, high soil moisture, and reduces plant transpiration in coastal habitat during the summer drought (Corbin et al., 2005; Hall & Willis, 2006; Lowry et al., 2008). In this way, coastal habitat favors the later flowering and perennial growth of the coastal race. The genetic basis of this flowering time divergence has now been established through quantitative trait locus (QTL) mapping (Hall et al., 2006). However, while drought is not a major factor for coastal plant populations, they are inundated by persistent salt spray from the Pacific Ocean (Boyce, 1954; Barbour, 1978).

In a reciprocal transplant field experiment, inland annual M. guttatus plants transplanted into coastal habitat were observed to have much higher rates of leaf necrosis than coastal perennial plants, presumably owing to salt from soil and/or oceanic spray (Lowry et al., 2008). Subsequent laboratory experiments confirmed that salt water spray causes leaf necrosis and found that coastal populations of M. guttatus are genetically more tolerant to this salt stress than inland populations (Lowry et al., 2008).

While the degree of salt-spray tolerance of vegetation differs between coastal perennial and inland annual M. guttatus, salt spray also contributes to higher concentrations of salt in the soil (Lowry et al., 2008). Further physiological assays are necessary to determine the contribution of such soil salinity to overall salt tolerance. Salt spray can enter the aboveground portion of a plant through the cuticle or stomata, which means that shoot tissue must tolerate high concentrations of toxic sodium (Na+) ions (Boyce, 1954; Bukovac, 1973; Zobel & Nighswander, 1990; Griffiths & Orians, 2003). Similarly, plant tolerance to soil salinity is often mediated by the Na+ tolerance of shoot tissue but, alternatively, can involve tolerance to osmotic stress or the exclusion of Na+ from the shoot (Munns & Tester, 2008). Osmotic stress acts to inhibit transpiration and the growth of the aboveground portion of a plant (Greenway & Munns, 1980; Fricke & Peters, 2002). One way that plants cope with this osmotic stress is to accumulate salt to arrive at an osmotic balance with the surrounding environment (Munns & Tester, 2008). However, accumulation of too much Na+ can lead to toxic concentrations in the shoot (Berthomieu et al., 2003; Munns et al., 2006; Rus et al., 2006). Therefore, soil salinity tolerance often involves a delicate balancing act of Na+ accumulation and exclusion (Zhu, 2001; Munns & Tester, 2008).

Despite the importance of salt tolerance to the local adaptation of coastal populations, nothing is currently known about the genetic basis of salt tolerance in M. guttatus. Since salt tolerance can involve multiple mechanisms, it is necessary to first determine which physiological traits contribute to salt tolerance before proceeding with genetic analysis. While there has been extensive work on salt tolerance across many other plant taxa, especially agricultural crops (Flowers, 2004; Yamaguchi & Blumwald, 2005; Bhatnagar-Mathur et al., 2008; Munns & Tester, 2008), very little is known about the genetic basis of adaptation to salt stress in coastal habitat (but see Rus et al., 2006), which is characterized by salt spray. Furthermore, it is currently unknown whether loci involved in salt tolerance adaptations to coastal habitat have effects on fitness in inland habitat.

In this paper, we examine the physiological basis and genetic architecture of local adaptation of M. guttatus to salt stress in the coastal habitat. First, we determine which mechanisms contribute to salt tolerance through multiple physiological assays. To discover loci involved in salt tolerance, we mapped salt spray tolerance and leaf Na+ concentration QTLs using recombinant inbred lines (RILs) derived from a cross between a pair of coastal and inland populations (Hall & Willis, 2006). We then determined whether these same loci play a role in local adaptation through a combined analysis of this new genotypic data with fitness data from a previously published reciprocal transplant experiment (Hall & Willis, 2006) to determine whether there are genetic trade-offs across habitats for salt tolerance QTLs.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Mechanisms of soil salinity tolerance

To determine the mechanisms of physiological salt tolerance differences between a pair of coastal perennial and inland annual populations of M. guttatus we conducted a hydroponic experiment using various concentrations of NaCl and evaluated osmotic stress responses, Na+ accumulation, as well as tissue tolerance. The coastal population (DUN) is located in coastal sand dune habitat in the Oregon Dune National Recreation Area, USA (43°53′35″ N 124°08′16″ W). The inland population (IM) is located in montane habitat on Iron Mountain in the Oregon Cascade Mountains, USA (44°24′03″ N 122°08′57″ W). Seeds of the inbred lines IM62 and DUN10 were planted in Fafard 4P soil and stratified at 4°C for 1 wk. Seeds were then moved to the Duke University glasshouses for germination. Seven days after germination, seedlings were transplanted into 2.5-inch (approx. 6.4 cm) square pots that contained Perlite. Transplanted seedlings (75 IM62 and 75 DUN10) were moved to a growth chamber with 8-h periods of light at 22°C and 16-h periods of dark at 18°C. Plants were randomized into 30 × 18 × 10 cm plastic bins, with five IM62 and five DUN10 plants per bin (bin = block). Half-strength Hoaglands solution (pH 6.0) was added to each block as a growth media, and solution was changed every third day to maintain a consistent concentration of nutrients in the solution. Salt treatment was initiated 14 d after transplantation. Salt (NaCl) was added to the half-strength Hoaglands solution to produce treatment solutions. Sets of three blocks were randomly assigned treatments consisting of 0 mm, 25 mm, 50 mm, 100 mm, or 150 mm NaCl solution for a total of 15 blocks (three blocks per each of the five treatments). Salinity concentrations were selected based on a recent review (Munns & Tester, 2008).

To examine genotypic differences in the effects of osmotic stress on plants, we measured growth of one newly emerged leaf every 2–3 d after the initiation of the salt treatment, as suggested by Munns & Tester (2008). These measurements were conducted for 14 d, but were subsequently terminated because many plants began to senesce in the higher concentration treatments (100 mm, 150 mm). To test for differences in the growth of young leaves between genotypes (DUN vs IM) across treatments we conducted a two-way repeated measures manova of leaf length data from all time points during the experiment.

To test for differential shoot accumulation of Na+ ions, the entire aboveground biomass of randomly selected plants in the 0 mm and 100 mm NaCl bins was harvested 15 d after the initiation of the salt treatment. Two plants of each type from each block were harvested giving a total of 24 plants. We also collected 10 DUN10 and 10 IM62 plants, which were grown in 0 mm NaCl half strength Hoaglands solution for mapping of leaf sodium concentration QTLs (below). The tissue collected from the 44 samples was briefly submerged in 0.05% Triton, followed by a rinse in deionized water, placed into 15 ml tubes (VWR International, West Chester, PA, USA), and dried in an oven for 24 h at 90°C. Dried samples were shipped to Purdue University for ionomic analysis (Baxter et al., 2007). To determine if DUN10 and IM62 constitutively differ in concentrations of Na+, and to test whether there is interaction between genotypes across treatments for Na+ concentration, we conducted a two-way anova with shoot ion concentration data from the 0 mm and 100 mm treatments. Because the concentration of potassium (K+) across treatments is often associated with salt tolerance (Chen et al., 2007), we also conducted the same two-way analysis on the shoot concentration of K+ ions.

To determine if genotypes differ in tissue tolerance to NaCl, we photographed blocks every 2–3 d for 50 d after the initiation of the salt treatment. Subsequent analysis of photos was used to establish the date of death (100% leaf necrosis) for each plant. According to Munns & Tester (2008), measurement of the time to senescence of leaves is a good assay of shoot tissue tolerance to Na+. We tested for differences between genotypes within each of the five treatments by survival analysis, where data was censored for plants that survived longer than 50 d. All analyses of salt tolerance mechanisms were preformed in jmp 7.0.1 (SAS, Cary, NC, USA).

RIL genetic map

To map salt tolerance QTLs and study fitness effects of those QTLs in the field, we genotyped previously constructed RILs. These RILs were made through reciprocal crosses between an inbred inland montane (IM62) line and a field-collected coastal dune (DUN) line and inbred for six to eight generations (Hall & Willis, 2006). To select markers for genetic mapping of RILs, we screened the inbred IM62 and DUN10 lines for polymorphism in hundreds of PCR-based markers. Markers used in this study are exon-primed intron-crossing (EPIC) markers derived from expressed sequence tags (ESTs). Polymorphism was evaluated in terms of variation in the length of PCR products, which is typically caused by indel variation in the introns. The development of these markers is outlined elsewhere (Fishman et al., 2008) and primers can be found at the website (http://www.mimulusevolution.org). Each primer pair included a forward primer fluorescently labeled with VIC, HEX or FAM (Invitrogen). Polymorphic markers were then tested in multiplex PCR reactions with three to five other markers. The PCR products were subjected to capillary electrophoresis and fragment analysis on an ABI 3730xl DNA Analyser (Applied Biosystems, Foster City, CA, USA). The size of the amplified fragments was scored using the programs genemapper (Applied Biosystems) and genemarker (SoftGenetics, State College, PA, USA).

We used an iterative process to assemble the linkage map for this study. Through regenotyping markers, genotyping additional markers with known locations in gaps between markers, and additional map assembly attempts we arrived at a final set of markers. Over the course of all iterations, we identified 239 markers that were polymorphic and amplified successfully in test multiplexes. These multiplex sets were used to genotype 186 RILs (113 with the DUN cytoplasmic background and 73 with the IM background), which had been used in the previous field experiment (Hall & Willis, 2006). Assembly of the linkage map was conducted with the program joinmap (Stam, 1993) using the Haldane mapping function with the default Maximum Likelihood settings. In addition, we used joinmap to identify markers with non-Mendelian segregation ratios.

QTL mapping of salt spray tolerance and leaf Na+ concentration

To identify loci involved in salt-spray tolerance, the 185 RILs (one RIL failed to germinate) were each tested for their respective tolerance to salt spray. Seeds of RILs and parental genotypes were stratified for 2 wks at 4°C, before being transferred to a growth chamber. In total, five replicates of each RIL and 31 replicates of each parental type (DUN and IM) were potted individually in 2.5-inch square pots in Fafard 4P soil (Conrad Fafard, Agawam, MA, USA). Pots were then fully randomized, placed in flats, and grown in a growth chamber under the same conditions as the physiological experiments described earlier. This short-day treatment prevented flowering of RILs, so that salt tolerance could be assessed across all plants at the same rosette stage of development. Plants were watered and flat positions were haphazardly rotated every other day. A regime of salt spray was initiated 25 d after germination. All plants were sprayed every other day with 5 ml of 500 mm NaCl, following Lowry et al. (2008). The number of days of survival following the initiation of salt spray was recorded immediately before the application of each spray treatment, with death defined as no remaining green tissue. We statistically controlled for variation among flats in our analysis by fitting a single-factor anova, with flat as a fixed effect. Using the residuals of this model, we calculated the mean survival time for each RIL line. A QTL analysis was then carried out using these centered line means.

To determine the loci involved in leaf Na+concentration, 169 of the RILs were grown in half-strength Hoagland's solution (pH 6.0). Three replicates of each RIL were grown in a fully randomized design under the same growth chamber conditions as in the physiological experiments. The second set of true leaves were collected from plants 30 d after germination and processed for ionomic analysis as described earlier. Leaves from all three replicate plants of each RIL were bulked into a single tube for this analysis. The resultant Na+ concentration data was used for QTL mapping. To test for the effect of cytoplasmic background on salt spray tolerance and leaf Na+ concentration, we used one-way anovas to compare RILs with DUN versus IM cytoplasmic backgrounds.

To map QTLs for survivorship in the growth chamber and leaf Na+ concentration, we implemented the standard model, forward and backward composite interval mapping method in QTL cartographer 2.5 (Wang et al., 2007). The parameters for both analyses included seven control markers, a 15 cM window size, and a 2 cM walk speed. The data were permuted 1000 times to estimate a significant experiment-wise likelihood ratio threshold for each trait (Churchill & Doerge, 1994). We followed the initial mapping with single marker analysis (anova) of loci located closest to the peak of significant QTLs. To determine the amount of variation between parental lines explained by each QTL, we calculated the additive effect of each QTL (2a) and divided it by the parental divergence of that trait. All analyses except genome-wide QTL mapping were conducted in jmp 7.0.1.

Effects of salt spray tolerance and leaf Na+ QTLs on fitness in the field

To determine if salt spray tolerance or leaf Na+ concentration QTLs had an effect on fitness under field conditions, we incorporated the genotypic data into a reanalysis of fitness data of the RILs from a previous reciprocal transplant experiment (Hall & Willis, 2006). In that experiment, the same RILs used for QTL mapping in the growth chamber were backcrossed as the female parent to independent inbred lines from both parental populations (IM494 and DUN10) to eliminate the effects of inbreeding depression (Hall & Willis, 2006). The progeny of the RIL backcross lines are referred to as BC-IM and BC-DUN. Three replicates of each BC-RIL type and 150 replicates of each parental type were planted at the seedling stage at the DUN field site (June 1, 2003) and at a field site on Browder Ridge (May 31, 2003) in the Oregon Cascades (3.2 km from the IM population site; see Hall & Willis, 2006 for details).

We analysed the effect of QTLs on total lifetime fitness, lambda (λ), which incorporated both survival and seed production for each plant (Hall & Willis, 2006). However, salt stress may be more extreme in later developmental stages of a plant owing to the accumulation of Na+ ions over time (Munns & Tester, 2008). Further, salt spray declines with proximity to the ground (Martin, 1959; Randall, 1970; Barbour, 1978), which is consistent with the observation that M. guttatus plants incur more necrosis in coastal habitat when they grow tall and flower (D. B. Lowry, pers. obs.). Therefore, we also divided the field fitness data from 2003 into two components for each backcross RIL line: mean survival to flowering and mean seed production per lines with surviving plants. These two fitness components were analysed separately for QTL analysis.

Our QTL analysis of the field data was restricted to an a priori determined set of loci based on the salt tolerance QTLs detected in the growth chamber. For each previously mapped QTL, we conducted a single marker analysis using only the marker within each QTL interval that was centered closest to the QTL peak in the original mapping experiment. All BC individuals were homozygous for coast (DUN) or inland (IM) alleles, depending on backcross direction, for 50% of alleles across the genome and heterozygous for the remaining 50% of alleles. Because of this difference in genetic composition backcrosses to DUN were analysed separately from backcrosses to IM. One-way anovas were used to test for associations between genotype and λ as well as the two components of fitness at both field sites. All analyses were implemented in jmp 7.0.1.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Mechanisms of soil salinity tolerance

Of the potential mechanisms of soil salinity tolerance, we found no evidence for the evolution of differences in osmotic stress tolerance between DUN10 and IM62, as the growth rates of both ecotypes were affected similarly by the treatment with salt solution (Table 1, Fig. 1a). For DUN10 plants, growth was reduced by 24, 30, 47 and 62% relative to the control in the 25 mm, 50 mm, 100 mm and 150 mm treatments, respectively. Similarly, IM growth was reduced relative to the control by 23, 23, 36 and 60% in the same treatments, but there was no interaction between genotypes and treatments.

Table 1.  Comparison of tolerance to osmotic stress between DUN10 (coast) and IM62 (inland) genotypes of Mimulus guttatus
Source of variationNum dfDen dfFP
  1. Osmotic stress tolerance was assessed through the growth rate of young leaves after initiation of salt stress.

Leaf growth (Osmotic stress)
Genotype11071.460.2298
Treatment41078.76< 0.0001
Genotype × treatment41072.050.0926
image

Figure 1. Physiological responses of DUN10 (coast) and IM62 (inland) Mimulus guttatus plants grown in salt (NaCl) solution. (a) Comparison of young leaf growth (osmotic stress tolerance) differences between DUN10 (open circles) and IM62 (closed circles) in the 0 mm NaCl treatment, as well as between DUN10 (open squares) and IM62 (closed squares) in 150 mm NaCl. (b) Comparison of difference of the shoot concentration of Na+ (closed) and K+ (tinted) ions in DUN10 and IM62 plants in 0 and 100 mm NaCl treatments. Error bars denote one standard error. (c) Comparison of tissue tolerance differences (survival) of DUN10 (open circles) and IM62 (closed circles) in the 150 mm NaCl treatment over the 50-d period following initiation of salt treatment.

Download figure to PowerPoint

We found evidence consistent with differential accumulation of Na+ between the ecotypes. Leaf Na+ concentration was constitutively greater for DUN10 than IM62 plants and there was a significant interaction between genotypes and treatments (Table 2a, Fig. 1b). Leaf Na+ concentration was 49% greater in DUN10 plants (Mean ± SE, 1205 ± 144 ppm) than IM62 plants (809 ± 164 ppm) in the 0 mm treatment. Leaf Na+ concentrations were much greater for both ecotypes in the 100 mm treatment. The DUN10 plants (27 074 ± 3572 ppm) had a 44% greater concentration of Na+ than IM plants (18 789 ± 1976 ppm) in the 100 mm treatment. While DUN10 and IM62 plants did not differ significantly in their leaf K+ concentrations, there was a significant genotype × treatment interaction of K+ (Table 2b, Fig. 1b). Concentration of K+ decreased for both DUN10 and IM62 in the 100 mm treatment, but this decrease was more pronounced for DUN10 (66% reduction in concentration) than for IM62 (28% reduction in concentration; Fig. 1b).

Table 2.  Analysis of shoot ion concentration differences between DUN10 (coast) and IM62 (inland) genotypes of Mimulus guttatus
Source of variationdfFP
  1. Two-way anova of (a) Na+ and (b) K+ shoot ion concentration of plants in 0 mm (control) and 100 mm NaCl treatments.

(a) Concentration of Na+ ions
Genotype 113.060.0008
Treatment 1333.28< 0.0001
Genotype × treatment 110.780.0021
Error40  
(b) Concentration of K+ ions
Genotype 10.130.7243
Treatment 196.12< 0.0001
Genotype × treatment 115.940.0003
Error40  

Consistent with a substantial difference in tissue tolerance, the survival of DUN10 was significantly greater than IM62 in the 25 mm (Wilcoxon inline image = 5.98, P = 0.0144), 50 mm (inline image = 21.98, P < 0.0001), 100 mm (inline image = 4.80, P = 0.0285) and 150 mm treatments (inline image = 11.54, P = 0.0007; Fig. 1c). The mortality after 50 d for IM was 0, 62, 100, 100 and 100% in the 0, 25, 50, 100 and 150 mm treatments, respectively. For DUN, the mortality rate was 0, 0, 0, 67 and 100% in the same treatments.

RIL genetic map

Following multiple iterations of genotyping and map assembly attempts we were able to construct a linkage map that was largely consistent with other previous and ongoing mapping projects (Fishman et al., 2008; C. A. Wu, unpublished; Y. W. Lee, unpublished). Many markers were difficult to score when genotyped on all of the RILs primarily because of poor amplification. In total, 50 markers were removed from the data set before arriving at a final number of 189 markers. Heterozygotes were removed from the data set for map assembly (Fig. 2) and QTL mapping, which when combined with other sources of missing data resulted in a high level of missing individual data points per marker (Mean ± SD, 17.85 ± 11.83%).

image

Figure 2. Linkage map of Mimulus guttatus for recombinant inbred lines (RILs) generated from a cross between a coastal perennial (DUN) and inland annual (IM) population. Regions with non-Mendelian inheritance (segregation distortion) are indicated at the right of each linkage group. The direction of segregation distortion (more DUN alleles = D/more IM alleles = I) and the level of significance are reported for each locus (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001). Segregation distortion calculated with both cytoplasmic types combined as one group.

Download figure to PowerPoint

Even with sufficient marker coverage, the assembly of linkage group 2 (Lg 2) was not initially possible because of extreme transmission ratio distortion in multiple regions of the linkage group. Assembly of Lg 2 with all of the marker data led to a map that was highly inconsistent with maps of Lg 2 from other linkage studies (Fishman et al., 2008; C. A. Wu, unpublished; Y. W. Lee, unpublished). Transmission ratio distortion on Lg 2 was especially strong in the DUN cytoplasmic background, with nearly complete distortion towards DUN alleles of some markers (Fig. 3). There was also strong distortion toward IM alleles in the IM cytoplasmic background at other nearby markers along Lg2, but this was not as severe as in the DUN cytoplasmic background (Fig. 3). To assemble the map for Lg 2, we restricted our data set only to genotypes with the IM cytoplasmic background. The final assembly of Lg 2 had a consistent marker order with other mapping studies, where the IM population was used in crosses (Fishman et al., 2008; C. A. Wu, unpublished; Y. W. Lee, unpublished).

image

Figure 3. Non-Mendelian inheritance of markers on linkage group 2 when divided into (a) recombinant inbred lines (RILs) with IM (inland) cytoplasmic background versus (b) RILs with DUN (coast) cytoplasmic background. All markers used in the assembly of the linkage map were included in this figure. The expectation for normal Mendelian inheritance is 50% IM alleles and 50% DUN alleles across RILs.

Download figure to PowerPoint

Significant transmission ratio distortion (P < 0.05) was also observed on a portion of all other linkage groups, except for Lg 10 (Fig. 2). None of these other distortion locations included obvious cytonuclear incompatibilities, as on Lg 2. Of the 189 markers included in the framework map, 91 (48%) were distorted at P < 0.05 and 33 (17%) were distorted at P < 0.001. Of the significantly distorted markers (P < 0.05), 59% had an excess of DUN alleles while 41% had an excess of IM alleles.

A total of 189 markers were used for the construction of the linkage map (Fig. 2). These markers formed 14 distinct linkage groups, which is consistent with previous mapping and cytogenetic studies of M. guttatus (Fishman et al., 2001, 2008; Hall & Willis, 2005). The total map length was 1394.4 cM Haldane, which is marginally shorter than map lengths in other studies in the M. guttatus species complex (Fishman et al., 2001; Hall & Willis, 2005). Recombination rate was extremely suppressed on a large portion of Lg 8 in comparison with ongoing mapping studies (C. A. Wu, unpublished; Y. W. Lee, unpublished). In other crosses involving the IM population, the distance between markers e299 and e278 on Lg 8 ranged from 23.3 to 32.0 cM (Y. W. Lee, unpublished). In this study, the distance between e299 and e278 was 2.5 cM. Additional studies have determined that a large chromosomal inversion is the cause of suppressed recombination here (D. B. Lowry & J. H. Willis, unpublished) and this may at least partially account for the short genome-wide map length.

QTL mapping of salt-spray tolerance and leaf Na+ concentration

Consistent with previous research (Lowry et al., 2008), our new experiments showed that salt spray tolerance significantly differed between DUN (Mean ± SE time of mortality, 20.32 ± 1.87 d) and IM (13.24 ± 0.64 d) parentals (t-test; df = 58, t = 3.49, P < 0.001, Fig. 4a). Cytoplasmic background had a significant effect on the salt-spray tolerance (F = 17.52, P < 0.0001) but not on leaf Na+ concentration (P > 0.05). Counterintuitively, RILs with IM cytoplasmic background had significantly higher salt-spray tolerance than RILs with the DUN cytoplasmic background (Fig. 4b). Because cytoplasm had an effect on salt-spray tolerance, we controlled for its effect in our QTL analysis.

image

Figure 4. Effects of genotypes on salt spray tolerance (survival) in growth chamber experiment. (a) Difference in survival between IM (inland) and DUN (coast) parental Mimulus guttatus plants. (b) Difference in survival of recombinant inbred lines (RILs) with DUN or IM cytoplasmic background. Effect of (c) SST1, (d) SST2, and (e) SST3 loci on survival, where AA are RILs homozygous for IM alleles and BB are RILs homozygous for DUN alleles. Note that the y-axis is not set to zero.

Download figure to PowerPoint

We identified three significant SALT SPRAY TOLERANCE (SST) (Fig. 5a, Table 3) and two significant LEAF SODIUM CONCENTRATION (LSC) QTL (Fig. 5b). The significance threshold for the SST loci was LR = 12.22 and LR = 13.22 for the LSC loci (P < 0.05, 1000 permutations). SST1 was located on Lg 1 centered closest to e543, SST2 on Lg 1 closest to e757, and SST3 on Lg 12 closest to e510. For two out of the three SST QTL (SST2 and SST3), RILs homozygous for the DUN allele had significantly greater survival in our experimental assay than RILs homozygous for IM (Fig. 4, Table 3a). For the third QTL, SST1, RILs homozygous for the IM allele outperformed RILs homozygous for the DUN allele (Fig. 4, Table 3). LSC1 was located on Lg 2 between e761 and e249 and RILs with the DUN allele had higher leaf Na+ concentrations (Fig. 5b). LSC2 had the opposite effect on leaf Na+ concentration and was located on Lg 14 closest to e583 (Fig. 5b). It should be noted that there were multiple sharp nonsignificant peaks detected in the QTL analyses of both traits (Fig. 5).

image

Figure 5. Significant quantitative trait loci (QTLs) were mapped (a) for Mimulus guttatus for salt spray tolerance (survival) and (b) for leaf Na+ concentration in growth chamber conditions. Composite interval mapping significance threshold (P < 0.05) of LR = 12.2 for SST QTLs and LR = 13.2 for LSC QTLs was established by 1000 permutations in QTL cartographer 2.5. Linkage group number and additive effects are displayed below the QTL maps.

Download figure to PowerPoint

Table 3.  The effect of three significant salt-spray tolerance quantitative trait loci (QTLs) on recombinant inbred line (RIL) means for (a) tolerance to salt spray (survival) under controlled growth chamber conditions and fitness (λ) of BC-IM RILs at the (b) IM (inland) and (c) DUN (coast) field sites
Source of variationN2a2a/diff (r2)FP
  1. Seed production of BC-IM RILs that survived to flower at the (d) IM and (e) DUN field sites. The number of lines (N), the divergence of alternative homozygous alleles (2a), the proportion of the parental divergence (2a/diff), as well as F and P values are provided for the growth chamber study. Comparisons in the field were between individuals homozygous for IM alleles or heterozygous for DUN and IM alleles and thus, only the additive effect ‘a’ is given. Because RILs were backcrossed for field experiments, the proportion of the line mean variance explained (r2) is provided instead of parental divergence.

(a) Growth chamber
SST1 (e543)179−1.70−0.23957.330.0075
SST2 (e757)1641.930.27268.570.0039
SST3 (e510)1791.820.25708.870.0033
(b) IM field site (λ)
SST1 (e543)178−0.230.00130.230.6311
SST2 (e757)162−0.080.00010.030.8715
SST3 (e510)175−0.200.00110.200.6568
(c) DUN field site (λ)
SST1 (e543)1770.060.02865.150.0245
SST2 (e757)1610.060.02804.580.0338
SST3 (e510)1740.060.02804.950.0274
(d) IM field site (seeds)
SST1 (e543)1451.49< 0.00010.010.9138
SST2 (e757)129−3.930.00050.070.7988
SST3 (e510)1415.090.00100.150.7037
(e) DUN field site (seeds)
SST1 (e543)15555.640.03345.290.0228
SST2 (e757)14161.970.03615.210.0239
SST3 (e510)15257.980.03705.760.0177

Effects of salt-spray tolerance and leaf Na+ QTLs on fitness in the field

To determine if QTLs identified in the growth chamber had an effect on fitness in the field, we conducted single-marker analysis with the markers e543 (SST1), e757 (SST2), e510 (SST3), e249 (LSC1) and e583 (LSC2). At the DUN field site, all three SST QTLs had a significant effect on lifetime fitness (λ) in the BC-IM lines (Table 3). None of the loci had an effect on λ in the BC-DUN lines (P > 0.05). Cytoplasm also had no effect on λ for either the BC-IM or BC-DUN lines. Separation of fitness components revealed that none of the QTL had a significant effect on survival to flowering. However, the three SST QTLs had a significant effect on seed set of surviving plants for the BC-IM lines at the DUN site (Table 3). For each of the three SST QTLs, heterozygous lines produced almost threefold more seeds per plant that survived to flower than lines homozygous for the inland allele (Fig. 6). Interestingly, eight out of the ten top seed-producing lines had a least one copy of the DUN allele at all three QTLs. The other two highest fitness lines had a copy of the DUN allele at two of those three loci. Neither of the LSC QTLs had a significant effect (P > 0.05) on fitness in either genetic background at either field site.

image

Figure 6. Effects of three salt-spray tolerance (SST) quantitative trait loci (QTLs) on seed production of Mimulus guttatus plants that survived to flower for recombinant inbred lines (RILs) backcrossed to IM (inland) at the DUN (coast) and IM field sites. Effects of (a) SST1, (b) SST2 and (c) SST3 loci on mean seed production. Comparisons made between backcross RILs homozygous (AA) for IM allele (closed circles) or heterozygous (AB) for DUN and IM alleles (open circles). Error bars indicate one standard error.

Download figure to PowerPoint

We found no evidence of genetic trade-offs for any of the three SST QTLs across the DUN and IM field sites. While all three SST QTLs affected λ and seed set of surviving plants at the DUN site, none of these QTLs had a significant effect on fitness at the IM (Browder Ridge) field site (P > 0.05; Fig. 6).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

In this study, we sought to understand the differential response of coastal perennial and inland annual populations of M. guttatus to salt stress, its genetic architecture, and the fitness effects of any QTLs involved in salt tolerance in native field habitats. The physiological divergence of the DUN10 and IM62 genotypes appears to involve differential accumulation of Na+ ions and tissue tolerance of the shoot, but not osmotic stress tolerance. These physiological results are consistent with adaptation of the DUN population to both soil salinity and salt spray. Genetic differences between DUN and IM in salt spray tolerance are caused by at least three QTLs of moderate effect, while two other QTLs affect leaf Na+ concentration. All three of the salt-spray QTLs contribute to fitness in coastal habitat but have no detectable fitness effects in inland habitat. Leaf Na+ concentrations QTLs had no fitness effects at either field site.

The physiology of salt tolerance in coastal populations

Although we had previously determined that coastal perennial and inland annual populations differ in salt spray tolerance (Lowry et al., 2008), we did not know whether soil salinity tolerance mechanisms contributed to this divergence. Physiological assays in this study suggest a major role for shoot tissue tolerance to Na+ions. Tissue tolerance in plants is thought to involve cellular processes such as the sequestration of toxic Na+ ions in vacuoles (Zhu, 2001), and is consistent with adaptation to soil salinity (Munns & Tester, 2008) or oceanic salt spray (Boyce, 1954).

Both DUN and IM had over an order of magnitude more Na+ in their leaves in the 100 mm treatment, which suggests that the roots of both ecotypes cannot exclude Na+ ions under saline conditions. The higher concentration of Na+ in DUN than IM plants suggests that DUN may be actively accumulating Na+ ions to achieve osmotic balance with the saline coastal soils (Barbour, 1978; Rus et al., 2006; Munns & Tester, 2008). Even so, the leaf growth assays suggest that there is no difference in osmotic stress tolerance between DUN and IM. One possible reason for this finding is that both populations are adapted to osmotic stress, but by different mechanisms. Inland annual populations may be adapted to osmotic stress from rapidly drying soils during the summer drought while coast perennial populations are adapted to osmotic stress caused by soil salinity (Hall & Willis, 2006; Lowry et al., 2008). This hypothesis is supported by studies in other coast and inland ecotypes of plants such as the salt brush (Atriplex halimus), where tolerance to the osmotic stress of drought and soil salinity differ in their underlying physiological mechanisms (Hu et al., 2007; Teixeira & Pereira, 2007; Ben Hassine et al., 2008).

Retention of K+ when subjected to saline conditions is thought to be crucial for salt tolerance of plants and has been found to be predictive of grain yield in crops such as barley and wheat (Wu et al., 1996; Zhu et al., 1998; Ren et al., 2005; Chen et al., 2007). Unexpectedly, K+ ion loss was significantly greater in the more salt tolerant DUN plants. Thus, it is possible that K+ shoot concentration is not important for the adaptation of M. guttatus to salt stress in coastal habitats.

Genetic basis of salt-spray tolerance

Very little is known about the genetic basis of the adaptation of coastal ecotypes to salt spray or soil salinity (but see Rus et al., 2006). While RILs homozygous for the DUN allele at SST2 and SST3 survived longer in the salt-spray treatment, this was not the case for SST1. The longer survival of RILs homozygous for IM at the SST1 locus is consistent with Lexer et al. (2003), who found salt tolerance QTLs to act in opposing directions in Helianthus hybrids. However, direction of effect of SST1 could have been influenced by the nature of the salt-spray assay. The NaCl solution used in our experiment was concentrated enough to eventually kill all of the plants, making it an easily measured assay. However, the overall dose is likely greater than that experienced by plants that were exposed to salt spray in coastal habitat. At this high level of salt stress, salt tolerance alleles that are neutral or beneficial under natural field conditions could have negative consequences in the laboratory. This would be especially true for a locus involved in the accumulation of Na+ ions in order to come into osmotic balance with the environment. Under low soil salinity levels, an allele that elevates Na+ accumulation would be beneficial, but at higher salinity levels the shoot concentration of Na+ would become toxic (Zhu, 2001; Munns & Tester, 2008). The QTL LSC1 and, to a lesser extent, SST1 appear to show such a trade-off pattern in a comparison of the direction of effect of these loci on salt-spray tolerance and leaf Na+ concentrations. In other words, these QTLs co-localize with peaks that have opposing effects in the salt spray tolerance and leaf sodium accumulation assays, as seen in Fig. 5. Alternatively, the negative effect of SST1 in the growth chamber may be caused by an inbreeding depression or hybrid inviability allele. The RILs in the growth chamber experiment were not outcrossed and thus any recessive deleterious alleles would be homozygous across many lines.

Beyond the significant salt spray tolerance and leaf Na+ concentration QTLs, we detected many sharp nonsignificant peaks (Fig. 5) that may also influence these traits. Our power to significantly detect these other potential QTLs was likely diminished by three major factors. First, there was a large amount of missing genotypic data in our QTL analysis. Second, the precision of the assays used for trait measurement may have influenced detection of QTLs. Finally, transmission ratio distortion may have played a role in QTL detection. Nearly 50% of loci were significantly distorted in this study and in a previous study that involved a F2 mapping population generated from a cross between the same DUN and IM populations (Hall & Willis, 2005).

While the cause of transmission ratio distortion is unclear for many of the linkage groups in this cross, cytonuclear incompatibilities appear to play a role on Lg 2. Strong distortion occurs in at least one place (markers: e340, e617, e624, e153) and possibly two (markers: e761, e294) in RILs with the DUN cytoplasmic background (Fig. 2). Since the LSC1 QTL is located between e761 and e249, the distortion in this region may have had consequences for the estimation of the effect of this locus in the glasshouse and the field. Interestingly, distortion favoring IM alleles appears in the IM cytoplasmic background (Fig. 2). However, this distortion is less likely to be caused by cytonuclear incompatibilities in the IM background because similar distortion occurs in the DUN background. The distortion toward IM alleles, present in both cytoplasmic backgrounds, appears to be tempered by the cytonuclear effects in the DUN background.

Genetic basis of local adaptation

Local adaptation is often assumed to be caused by alleles that perform well in local habitats but have negative consequences in foreign environments (Hawthorne & Via, 2001; Kawecki & Ebert, 2004). In our study, however, the three SST QTLs that had effects on fitness at the DUN field site did not have significant effects on fitness at the IM site. Ongoing analysis has also revealed that there are no negative fitness consequences at the DUN site for QTLs that affect fitness at the IM field site (M. C. Hall & D. B. Lowry, unpublished). Therefore, while there is a genotype × environment interaction for QTLs involved in local adaptation, we found no evidence of negative consequences of QTLs across habitats. This finding is consistent with the handful of other studies that have assessed the fitness effect of a locus between environments in reciprocal transplant experiments (Weinig et al., 2003; Verhoeven et al., 2004, 2008; Gardner & Latta, 2006).

The collective implication of these few reciprocal transplant QTL studies is that locally adaptive alleles may not have deleterious fitness consequences in other habitats. If these alleles are truly neutral under other environmental conditions, then they could diffuse unidirectionally into other habitats, since selection acts on them in only one habitat (Gardner & Latta, 2006). Alternatively, these QTLs could have slight deleterious effects in other habitats that were not detected owing to statistical power. Even so, the question arises as to whether the different levels of selection across habitats found in this and other studies (Weinig et al., 2003; Verhoeven et al., 2004, 2008; Gardner & Latta, 2006; Hereford, 2009) implies that local adaptation mostly involves the action of nonoverlapping sets of loci for each habitat. Answering this question will require detailed genetic analysis and field experimentation, but is crucial to the determination of the ultimate causes of local adaptation (Fry et al., 1998; Schemske, 2000; Keurentjes et al., 2008; Stinchcombe & Hoekstra, 2008).

Different but tightly linked genes may underlie QTLs that appear to affect both traits in the laboratory and fitness in the field, causing spurious associations (Kawecki & Ebert, 2004; Stinchcombe & Hoekstra, 2008). This could be the case for SST1, where the IM allele performed better under artificial salt-spray conditions, but the DUN allele preformed better at the DUN field site. Cloning of genes that underlie QTLs will help to better understand QTL effects across experiments. Further, as genomic resources and advanced molecular techniques are applied to reciprocal transplant field experiments, the mechanisms of local adaptation should come into focus (Stinchcombe & Hoekstra, 2008; Wu et al., 2008).

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The authors thank Louise Cooley, Ruth McDevitt, Nettie McMiller, Jason Morich and Calvin Sheng for field and laboratory assistance. Lisa Bukovnik, Beverly Calhoun, Brian King, Brett Lahner, Shengchu Wang, the Duke University glasshouse staff, and the Purdue University Ionomics Center made our experiments and analyses possible. Discussion with and suggestions from Arielle Cooley, Kathleen Donohue, Katie Ferris, Sheril Kirshenbaum, Young Wha Lee, Thomas Mitchell-Olds, Jennifer Modliszewski, William Morris, Mohamed Noor, Laura Nutter, Mark Rausher, Rafael Rubio de Casas, Jessica Selby, Jenny Tung, Greg Wray, Kevin Wright, Carrie Wu, and two anonymous reviewers, greatly improved this manuscript. Funding was provided by the National Science Foundation, through a FIBR grant (EF-0328636), an Environmental Genomics Grant (EF-0723814), and two Doctoral Dissertation Improvement Grants to D.L. (DEB-0710094) and to M.H. (DEB-010577). Funding was also provided by a Sigma Xi Grants-In-Aid of Research, a National Institute of Health Graduate Student Fellowship, and a Duke University Travel Grant.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References