REDUCED DROUGHT TOLERANCE DURING DOMESTICATION AND THE EVOLUTION OF WEEDINESS RESULTS FROM TOLERANCE–GROWTH TRADE-OFFS

Authors


Abstract

The increased reproductive potential, size, shoot allocation, and growth rate of weedy plants may result from reduced resource allocation to other aspects of plant growth and defense. To investigate whether changes in resource allocation occurred during domestication or the evolution of weediness, we compared the mycorrhizal responsiveness, growth, and drought tolerance of nine native ruderal, nine agriculturally weedy (four U.S. weedy and five Australian weedy), and 14 domesticated populations (eight ancient landraces and six improved cultivars) of the common sunflower (Helianthus annuus). Domesticated sunflower cultivars were less drought tolerant, but had higher plant growth and fecundity and coarser roots than wild populations. There were no changes in level of drought tolerance between improved cultivars and ancient landrace plants, but there was an increase in allocation to flowers with recent selection. Weedy populations were intermediate between domesticated cultivars and native ruderal populations for plant growth rate, root architecture, and drought tolerance. Weedy populations benefited most from mycorrhizal inoculation by having fewer wilted leaves and wetter soil. Overall, we found that trade-offs between drought tolerance and several aspects of plant growth, including growth rate, allocation to flowering, and root architecture, govern evolution during sunflower domestication and the invasion of disturbed habitat.

Trade-offs are thought to play a fundamental role in the evolution of life histories. Ruderal, weedy, and invasive plants, for example, grow fast, mature early, and allocate heavily to reproduction (Grime 1977; Van Kleunen et al. 2010) and are thereby able to rapidly colonize disturbed lands. The traits commonly found in weedy plants often mirror those that occur in domesticated crops, such as increased fecundity, changes in allocation patterns, and altered physical and chemical defenses (Harlan 1992). Natural and artificial selection for increased reproduction in weedy and domesticated plants may result in reduced allocation to competitive ability, to biotic environmental interactions, or to tolerance of abiotic stress.

Most studies of life-history evolution in weeds and domesticated plants have focused on trade-offs between reproduction and costly structural or chemical defenses (Rosenthal and Dirzo 1997; Franks et al. 2008; Mayrose et al. 2011), perhaps due to the popularity of the evolution of increased competitive ability (EICA) hypothesis. However, in some instances, the expected trade-offs with defense have not been observed, suggesting that other nondefense-related traits should be examined. For example, faster growing invasive plants have been shown to be less shade tolerant (Martin et al. 2010) and have stronger reactions to water and nutrient stress (He et al. 2010; Mayrose et al. 2011). However, the extent to which increased allocation to growth and fecundity in weedy and domesticated plants trades off with tolerance of abiotic stress remains poorly understood.

Plant tolerance to drought may be mediated by their symbioses with arbuscular mycorrhizal (AM) fungi. Associations with AM fungi increase plant access to soil resources through higher surface area and enzymatic activity (Smith and Read 2008). AM fungi also have been shown to increase tolerance to environmental stressors such as herbivory (Bennett et al. 2009) and drought stress (Davies et al. 1993; Marulanda et al. 2003). AM fungi alleviate drought stress, for example, both through positive effects on water uptake due to their increased surface area, and indirectly through improved nutrient uptake in dry conditions (Augé 2001). Given the potential role of these symbioses in plant tolerance to stress, measures of evolutionary trade-offs need to be made within the context of this symbiosis. However, we are unaware of previous studies that have investigated the potential for AM fungi to ameliorate trade-offs between reproduction and tolerance to abiotic stress.

Although the role of AM fungi in mediating trade-offs with stress has not been investigated, the interaction of plants with nutritional symbionts such as AM fungi creates an additional axis for trade-offs with growth and reproduction. Support of AM fungi is energetically expensive and plants would be better off not investing in AM fungi under conditions in which plant roots can directly access sufficient soil resources such as highly fertile soils (Schultz et al. 2001), or under conditions where the abundance and infectivity of these symbionts are reduced. As disturbed environments, such as the agricultural fields in which crops are bred and weeds proliferate, can have both high available resources and reduced infectivity of AM fungi (Abbott and Robson 1991; Jasper et al. 1991), it may be expected and has been observed that populations have evolved reduced investment in, and dependency on, AM fungi during domestication (Hetrick et al. 1993) and invasion (Seifert et al. 2009). However, the potential adaptation to stress to modify expectations for the shifts in dependence on AM fungal mutualisms has not been investigated.

Identifying the costs and constraints of evolution of rapid growth rate and high yield requires investigating multiple dimensions that may trade-off with growth and fecundity traits. We present the first study that tests for potential interactive effects of shifts in drought tolerance and dependence on AM fungi during the evolution of weediness and domestication. We specifically test the extent to which weedy and domesticated sunflowers exhibit growth and fecundity trade-offs with drought tolerance, and the role that AM fungi may play in the expression of traits on both sides of these trade-offs. Helianthus annuus is an ideal study organism to investigate changes in allocation during the evolution of weediness and domestication because there are many geographically distinct populations of H. annuus adapted to a wide range of environments. For example, native populations are typically ruderals colonizing recently disturbed patches within tallgrass prairies in the central U.S. that are subjected to drought conditions. Native Americans domesticated sunflowers, generating numerous ancient landraces, which formed the genetic foundation of improved cultivars that were further selected for high yield in fertilized and well-watered agricultural fields. In addition, weedy populations persist within agricultural areas of North America and are thought to be derived from native ruderals (Kane and Rieseberg 2008). Invasively weedy populations in Australia (Seiler et al. 2008), Europe (Rehorek 1997), and South America (Cantamutto et al. 2010) likely originated from crop-wild hybrids (Muller et al. 2011; Lai et al. 2012). Our study contrasts growth, drought tolerance, and response to mycorrhizal fungi across five categories of H. annuus populations: improved cultivars, ancient landraces, U.S. weedy, Australian weedy, and native ruderal populations. We decomposed variation among these population categories into four orthogonal contrasts: the effects of domestication (contrasting domesticated lines vs. nondomesticated populations), the effects of artificial selection in intensive agricultural environments (contrasting improved cultivar lines vs. ancient landraces lines), the evolution of weediness (contrasting native ruderal populations vs. weedy populations), and the differences between weedy populations from two continents (contrasting U.S. weedy populations vs. Australian weedy populations).

Methods

STUDY SYSTEM

To investigate whether native ruderals, U.S. weedy, Australian weedy, and domesticated ancient landraces or improved cultivars responded differently to drought or AM fungi, we selected 32 populations of H. annuus consisting of nine U.S. native ruderal, nine weedy (four U.S. and five Australian), and 14 domesticated lines (eight improved cultivars and six ancient landrace) (Table S1). Domesticated lines were maintained in Ames, Iowa by the USDA and included eight modern improved cultivars and six Native American ancient landraces. Native ruderal populations were collected from the native grassland habitat of the common sunflower in Denver, Colorado; Colby, Kansas; Roswell, New Mexico; Sylvia, Kansas; Trout Creek, Utah; Fish Springs National Wildlife Refuge, Utah; Provo, Utah; Rozel, Utah; and Stafford, Kansas. The U.S. weedy populations were collected from cornfields in Townsend, Indiana; Norton, Kansas; Delta, Utah; and Davis, California. The Australian weedy populations we used were collected by Gerald Seiler and Tom Gulya (USDA) from roadsides in or near croplands in southern Australia (Seiler et al. 2008).

EXPERIMENTAL DESIGN

Replicates from each population were grouped into eight drought treatment plots, within which all other treatments were arranged within a randomized design. Eight maternal lineages were used as replicates for each of the weedy and native ruderal populations. The seedlings for all populations were planted into pots with both sterilized soil and AM fungi inoculated soil and both drought and nondrought water treatments, totaling 128 plants per block and 1024 plants in entire experiment.

The soil was a 1:1:1 mixture of Indiana topsoil, Indiana river sand, and calcined clay. The soil mixture was steam sterilized twice for 4 h with a one-day rest period in between sterilizations. The Indiana soil had a pH of 6.3, 2.0% organic matter, and nutrient concentrations of 9 ppm N, 6 ppm P, and 113 ppm K. Noninoculated pots received 2000 cc of the sterilized soil mixture.

The AM fungi we used were isolated from The Nature Conservancy's Kankakee Sands Prairie in northern Indiana. Pure cultures were grown with sorghum grass on mixture of sand and Indiana soil (Vogelsang et al. 2006). An AM fungi mixture was made after cutting the pure cultures into small fragments. The mixture included the following species listed according to decreasing volume; Glomus claroideum, Scutellospora fuldiga, G. mosseae, S. pellucida, G. lamellosum, Acaulospora spinosa, Paraglomus sp., and Entrophospora infrequens. Sterilized calcined clay was added to the AM fungi mixture to match the background soil mix. The AM fungi pots were filled with 830 cc of sterile soil on the bottom. Of the additional 1170 cc of soil, 8.1% was inoculated with the AM fungi mixture at the center depth. Each noninoculated pot received 20 mL of a microbial filtrate prepared from the AM fungi mixture (Seifert et al. 2009).

Seedlings were germinated by removing seeds from the seed coat and placing seeds into sterile petri dishes for several days before being transplanted into pots in the Indiana University greenhouse at the start of summer of 2009. Seedlings were fertilized with 200 ppm N and K (15-0-15) every 10 days. After 33 days, we initiated the drought treatment and discontinued fertilization. The experiment was harvested when plants began senescence, approximately 90 days after planting.

Control plants were watered twice daily throughout the experiment, at approximately 900 h and 1800 h, whereas drought-stressed plants received no water during an entire drought period. Drought periods for all drought-stressed plants were one, two, three, nine, twelve, and seventeen days, with one full day of watering in between each drought period. The length of a drought was increased after drought period three to enhance plant response. Drought response measurements included the numbers of dead and wilted leaves as well as the degree and height at which the main stem showed wilting (i.e., fell downwards). A wilted leaf was defined as: (1) The edges of a leaf curling up more than 90°. (2) A petiole that was angled 45° or less from the stem of the plant. (3) Any yellowing leaves. After each drought period, the numbers of healthy, dead, and damaged leaves were measured for each stressed plant. Damaged leaves were defined as partially dead, whereas dead leaves were defined as entirely brown or completely inflexible. Drought response measurements were taken daily during drought periods one through four and once during the fifth drought period. Drought response measurements were not collected frequently during the fifth and sixth drought period because drought responses became skewed by leaf death. As leaves that were significantly wilted early in the experiment died, drought tolerance, as measured by the proportion of wilted leaves, became confounded by the fact that those plants that were most severely affected by drought had fewer leaves and were therefore less likely to experience drought stress as they had early in the experiment.

The Theta Probe ML2x was used to measure soil moisture content of the drought-stressed plants on day 2 of drought period two, on day 4 of drought period four, and on day 7 of drought period five. The ThetaProbe ML2x was calibrated specifically to our soil mixture according to the Delta-T manufacturer's instructions (Delta-T 1999). Each pot received three mV readings of soil moisture.

The total above ground biomass was collected by cutting the stem at the soil line, and measuring the dry mass of the stem and inflorescences. During leaf and inflorescence measurements, any leaf or inflorescence larger than 1.0 cm was counted.

A subsample of roots from each well-watered plant was washed and stained with Trypan Blue to confirm AM fungi colonization (Vogelsang and Bever 2009). The results are recorded as the percentage of intercepted roots that are successfully colonized by AM fungi as scored at 20× in a compound microscope.

ROOT ARCHITECTURE EXPERIMENT

To investigate whether native ruderals, U.S. weedy, Australian weedy, and domesticated populations plants differed in their root architecture, we selected 19 populations of H. annuus consisting of the same nine native ruderal and nine weedy (four U.S. and five Australian) populations as used above, and the PI 432514 improved cultivar domesticated line. Five replicates of each population were germinated and as described above with the following exceptions: five replicates were used instead of eight, plants were grown in the sterile soil mixture, the plants received no drought treatments, and only one domesticated line was used due to logistical restrictions. After five weeks of growth, we harvested the plants by carefully washing the roots free of debris. From each plant, we selected two strands of root for analysis that were representative of the entire root length. We analyzed root architecture using WhinRhizo image analysis software in accordance with the methods outlined in Seifert et al. (2009). Specific root length was calculated by calculating the length of every segment in a scanned image and dividing it by the dry mass.

STATISTICAL ANALYSIS

We analyzed plant response using a mixed model, with populations within categories and all of their interactions as random effects using Proc Mixed in SAS (SAS Institute 2008). We first analyzed total biomass, inflorescence biomass, and allocation to flowering across both well watered and drought treatments and tested for variation in drought tolerance in the interaction term of drought by population categories. We then analyzed the well-watered and drought treatments separately to get independent measures of population growth under optimal conditions (in the well-watered treatment) and specific phenotypic measures of drought tolerance such as proportion of leaves wilted in the drought treatment. As large plants may be more likely to experience drought in pots by virtue of their increased ratio of plant size to pot volume, we included measures of above ground plant size as covariates to remove the effects of plant size from analyses of drought tolerance and thereby isolated variation in physiological resistance to drought.

For analysis of above-ground biomass and response to drought and inoculation, we decomposed differences between population categories into four a priori chosen orthogonal contrasts corresponding to domesticated lines (both types) versus nondomesticated populations (all three types), ancient landraces lines versus improved cultivars, native ruderal populations versus agricultural weedy populations (both types), and Australian weedy populations versus U.S. weedy populations. Equivalent contrasts were used within the interaction terms of population category with drought, AM fungi, and drought × AM fungi. For analysis of root characters, we decomposed differences between population categories into two a priori chosen orthogonal contrasts corresponding to native ruderal populations versus weedy populations and Australian weedy populations versus U.S. weedy populations.

We observed significant variance components between populations within categories for plant size and allocation patterns, drought tolerance, and root characters. As plant growth, drought tolerance, and root characters were independently measured on separate plants and therefore unconfounded, we were able to test for interpopulation genetic covariance without problems commonly associated with tests of trade-offs of growth and tolerance (Brett 2004). We calculated the best linear unbiased predictor (BLUP) of these measurements for each population from the mixed models described above.

Direct or indirect selection on characters that have significant interpopulation genetic covariance with drought tolerance may mediate observed differences in drought tolerance between the population categories. We tested this possibility using analysis of covariance (ANCOVA) of the mean population drought tolerance using Proc Mixed in SAS. Covariates that mediate the differences in drought tolerance between population categories will reduce the sums of squares explained by category and by the four orthogonal contrasts between domesticated lines verses other populations, ancient landrace lineages versus improved cultivars, native ruderal populations verses weedy populations, and Australian weedy populations verses U.S. weedy populations. We constructed an F test for this mediation from the difference in sums of squared deviations due to the four comparisons of categories without covariates and the sums of squared deviations for these same comparisons with covariates.

Results

PLANT GROWTH AND FLOWERING

When comparing drought-stressed versus well-watered plants, drought stress reduced the shoot mass, total inflorescence mass, and average inflorescence mass across all populations (Fig. 1, F1,27= 82.2, P≤ 0.001; F1,27= 61.6, P≤ 0.0001; F1,20= 49.9, P≤ 0.0001, respectively). Drought stress had smaller effects on shoot mass of domesticated plants than on the shoot mass of native ruderal or weedy populations (Fig. 1a, F1,27= 16.0, P= 0.0004) and drought stress had smaller effects on the shoot mass of improved cultivars than on ancient landraces (Fig. 1a, F1,27= 9.5, P= 0.005). The effects of drought on the total inflorescence mass (Fig. 1b, F1,27= 16.2 P= 0.0004) and the average inflorescence mass (Fig. S1, F1,20= 38.9, P≤ 0.0001) of domesticated plants were larger than for the other three population categories, as domesticated plants produced the largest inflorescences in well-watered conditions and proportionally the smallest inflorescences under drought-stressed conditions.

Figure 1.

The effects of drought on shoot mass, allocation to flowering, and inflorescence mass, where D stands for drought-stressed plants, and C stands for well-watered control plants. The bars represent mean natural log of shoot mass (a), the mean natural log of total inflorescence mass (b), and allocation to flowering (c) for each population category measured at the end of the experiment. Error bars represent variation between populations in each category.

In an analysis of well-watered plants, allocation to flowering, the inflorescence mass divided by shoot mass, of well-watered plants varied by population type (Fig. 1c; Table 1, P= 0.002). Under well-watered conditions, domesticated plants had greater allocation to flowering (Fig. 1c; Table 1, P= 0.002), with improved cultivar lineages having greater allocation to flowering than ancient landrace lines (Fig. 1c; Table 1, P= 0.007). Domesticated plants also had the greatest allocation to flowering under drought-stressed conditions (Fig. 1c, F1,27= 19.5, P= 0.0001), which was driven by the improved cultivars having more allocation to flowering than the ancient landrace lines (Fig. 1c, F1,27= 5.0, P= 0.03).

Table 1.  Growth measurement analysis of well-watered plants. Our experiment included 32 sunflower populations consisting of nine native ruderal, nine weedy (four U.S. and five Australian), and 14 domesticated populations (eight improved cultivars and six ancient landrace). Population types are fixed effects with populations within population types being random effects. Differences between population types are decomposed into four orthogonal contrasts.
 dfShoot MassTotal inflorescence massAllocation to flowering
F P F P F P
  1. 1No significant contrasts.

Block72.130.048.64<0.0017.24<0.0001
Population type 4 4.52 0.006 4.64  0.006 5.64  0.002
Domesticated vs. others14.540.0414.9 0.000611.3 0.002
Improved vs. landrace 1 10.6 0.003 1.54  0.2 8.59  0.007
Native ruderal vs. weedy14.240.050.06 0.80.1 0.7
U.S. weedy vs. Australian weedy 1 0.09 0.8 0.55  0.5 0.1  0.8
Soil13.800.063.46 0.071.55 0.2
Soil×population type1 4 0.64 0.6 0.82  0.5 1.11  0.4
 Est. (SE) PEst. (SE) PEst. (SE) P
Population type×population 0.1 (0.04) 0.0004 0.2 (0.06) 0.0003 0.008  (0.002) 0.0002
Soil×population type×population0.0006 (0.006)0.50 (.).0 (.).
Residual 0.1 (0.01)<0.0001 0.2 (0.01)<0.0001 0.005  (0.0003)<0.0001

Inoculation with mycorrhizal fungi decreased shoot mass by 2.8% compared to noninoculated control (Fig. S2a, F1,27= 9.8, P= 0.004). Mycorrhizal fungi decreased the average inflorescence mass of domesticated plants under well-watered conditions (Fig. 2, F1,20= 6.7, P= 0.02). The fecundity of well-watered U.S. weedy, Australian weedy, and native ruderal populations was not strongly affected by mycorrhizal inoculation. However, inoculation of native ruderal plants reduced the total inflorescence mass (Fig. S2b, F1,27,= 3.79, P= 0.06) and allocation to flowering (Fig. S2c, F1,27= 4.05, P= 0.05). Overall, inoculated plants were 64% colonized whereas noninoculated plants were 7% colonized (F1,27= 260.9, P < 0.0001).

Figure 2.

The effects of inoculation with arbuscular mycorrhizal (AM) fungi on the average inflorescence mass of well-watered control plants, where AMF represents mycorrhizal inoculated plants and STER represents plants grown in nonmycorrhizal soil. The bars represent mean inflorescence mass of each population in a category as measured during the second day of drought period two. Error bars represent variation between populations in each category.

LEAF WILTING AND SOIL MOISTURE RESPONSES TO DROUGHT

The five population categories varied in their tolerance to drought as measured by proportion of wilted leaves, degree of stem wilting, and soil moisture. Domesticated populations were the most susceptible to drought stress when considering the proportion of wilted leaves (Fig. 3a; Table 3, P= 0.0004) and stem wilting (Fig. S3, F1,27= 10.73, P= 0.003), even though they had the wettest pots (Fig. 3b; Table 3, P= 0.0004). Native ruderal plants had fewer wilted leaves than weedy plants (Fig. 3a; Table 3, P= 0.03).

Figure 3.

The effects of inoculation with arbuscular mycorrhizal (AM) fungi on the proportion of wilted leaves (a) and soil moisture (b) for drought-stressed plants, where AMF represents mycorrhizal inoculated plants and STER represents plants grown in nonmycorrhizal soil. The bars represent the mean proportion of wilted leaves and mean soil moisture for each population in a category as measured during the second day of drought period two. Error bars represent variation between populations in each category.

Table 3.  Growth response of drought-stressed plants during the second day of drought period two. Our experiment included 32 sunflower populations consisting of nine native ruderal, nine weedy (four U.S. and five Australian), and 14 domesticated populations (eight improved cultivars and six ancient landrace). Population types are fixed effects with populations within population types being random effects. Differences between population types are decomposed into four orthogonal contrasts.
 dfWilted leaves1Soil moisture1
F P F P
  1. 1Leaf number after drought period two, stem diameter and leaf size before drought period one were used as covariates.

Block74.60<0.00014.51<0.0001
Leaf number 1 0.02  0.9 8.09  0.005
Stem diameter13.62 0.060.12 0.7
Leaf size 1 38.1 <0.0001 107 <0.0001
Population type45.14 0.0034.59 0.006
Domesticated vs. others 1 16.1  0.0004 16.5  0.004
Improved vs. landrace10.29 0.60.74 0.4
Native ruderal vs. weedy 1 5.64  0.03 3.71  0.06
U.S. weedy vs. Australian weedy10.23 0.60.31 0.6
Soil 1 5.37  0.03 6.07  0.02
Soil×population type41.49 0.20.88 0.2
Domesticated vs. others 1 1.30  0.3 2.57  0.1
Improved vs. landrace10.35 0.60.93 0.3
Native ruderal vs. weedy 1 4.14  0.05 0.02  0.9
U.S. weedy vs. Australian weedy10.01 0.90.14 0.7
  Est. (SE) P Est. (SE) P
Population type×population0.004 (0.004) 0.154.4 (116.8) 0.3
Soil×population type×population 0.002 (0.004) 0.03 183.8 (151.7) 0.1
Residual0.09 (0.007)<0.00012009.8 (147.7)<0.0001

Inoculation with AM fungi increased drought tolerance in terms of the proportion of wilted leaves (Fig. 3a; Table 3, P= 0.03) and pot soil moisture (Fig. 3b; Table 3, P= 0.02). Inoculated weedy plants had fewer wilted leaves than noninoculated plants, in contrast to native ruderal and domesticated plants which did not respond to mycorrhizal inoculation by having fewer wilted leaves (Fig. 3a; Table 3, P= 0.05). Mycorrhizal inoculation increased the soil moisture of native ruderal and weedy plants, but not domesticated plants (Fig. 3b; Table 3, P= 0.1).

The patterns we found during drought period two of day 2 continued throughout the experiment. At the end of drought period five, mycorrhizal inoculated plants had fewer dead or damaged leaves than noninoculated plants (Fig. S4, F1,27= 6.86, P= 0.01), and domesticated plants had more dead and damaged leaves than the other population categories (Fig. S4, F1,27= 258.83, P= 0.001), with improved cultivars having more dead and damaged leaves than ancient landrace lines (Fig. S4, F1,27= 12.39, P= 0.002). Inoculated soil marginally decreased leaf death and damage for native ruderal and weedy populations but not domesticated lines (Fig. S4, F1,27= 3.09, P= 0.09).

ROOT ARCHITECTURE EXPERIMENT

The average root diameter and specific root length of populations were strongly negatively correlated with each other (r=−0.96, df = 18, P < 0.0001) and varied among population categories (Fig. S5; Table 2, P= 0.06 and 0.02, respectively). Domesticated plants had the thickest roots and smallest specific root lengths, whereas native ruderals had the thinnest roots and longest specific root lengths of the population types (Fig. S5). The roots of weedy populations differed from native ruderal populations by being thicker (Table 2, P= 0.02) and by having smaller specific root lengths (Table 2, P= 0.006). The root architecture of U.S. weedy populations and Australian weedy populations did not differ from each other (Table 2).

Table 2.  Root architecture analysis consisted of nine native ruderal and nine weedy (four U.S. and five Australian) populations and the PI 432514 improved cultivar line. Plants were grown under well-watered conditions in sterilized soil.
 dfAverage diameterSpecific root length
F P F P
Population type33.140.05664.330.0219
Native ruderal vs. weedy 1 7.05 0.018 9.97 0.0065
Australian weedy vs. U.S. weedy10.410.53250.000.967
  Est. (SE) P Est. (SE) P
Population type×population0.00014 (0.00019) 0.21.34×106 (2.7×106)  0.3
Residual 0.0013 (0.00024)<0.0001 20.8×106 (4.0×106)<0.0001

ANALYSES OF GENETIC CORRELATIONS AND COVARIANCE

Analysis of correlations of population means (best linear unbiased predictors) demonstrated that drought tolerance, as measured by the resistance to wilting leaves after drought period two of day 2, was negatively correlated with total above-ground biomass (Fig. 4a; r=−0.825, df = 32, P≤ 0.0001), allocation to flowering (Fig. 4b; r=−0.317, df = 32, P= 0.08), and average root diameter (Fig. 4c; r=−0.744, df = 18, P= 0.0004).

Figure 4.

Best linear unbiased predictor correlations of proportion of wilted leaves on the second day of drought period two against the (a) plant size (r=−0.825, df = 32, P≤ 0.0001), (b) allocation to flowering (r=−0.317, df = 32, P= 0.08), and (c) average root diameter (r=−0.744, df = 18, P= 0.0004) for each population.

ANCOVA provided evidence that selection on growth rate, allocation to flowering, or root architecture contributed to the reduction in drought tolerance with domestication (plant growth rate: F1,27= 136.1, P= 0.0003; allocation to flowering: F1,27= 32.8, P= 0.005; and average root diameter: F1,15= 4.5, P= 0.1; Tables S2 and S3). Including the individual measures of plant growth rate, allocation to flowering, and average root diameter can explain 78%, 19%, and 36%, respectively, of the difference in the drought tolerance variation among domesticated cultivars and other population types (Tables S2 and S3). ANCOVA also suggested that direct or indirect selection on plant growth rate and root architecture could contribute to the reduction in drought tolerance with the evolution of weediness (plant growth rate: F1,27= 11.7, P= 0.03; average root diameter: F1,15= 11.1, P= 0.04; Tables S2 and S3), with plant growth rate and root architecture explaining 39% and 60%, respectively, of the difference in drought tolerance among weedy populations and native ruderal populations (Tables S2 and S3).

Although U.S. weedy and Australian weedy populations were not different in overall drought tolerance (Table 3), significant differences in drought tolerance are revealed in analyses of population means once growth rate was included as a covariate (F1,27= 7.06, P= 0.01, Table S2). Inclusion of a biomass × population category interaction within the model reveals a significant difference in the slope of the relationship of drought tolerance to growth rate between Australian weedy and U.S. weedy populations (F1,22= 6.02, P= 0.02). This result suggests that the underlying basis for overall similar levels of drought tolerance differs between these two types of weedy populations.

Discussion

EVOLUTION DURING DOMESTICATION

We found evidence that trade-offs between drought tolerance and several aspects of plant growth, including growth rate, allocation to flowering, and root architecture govern the evolution of H. annuus during domestication. Our experiment is consistent with and extends the results found by Mayrose et al. (2011), which suggested that direct selection on the growth of domesticated sunflowers in benign growing conditions can lead to a reduction in drought tolerance. Domesticated sunflower cultivars were less drought tolerant, but had higher plant growth rates than native ruderal populations. Additionally, domesticated populations exhibited greater allocation to flowering and, based on a single population, coarser roots than wild populations.

We also found significant negative interpopulation genetic correlations between drought tolerance as measured by resistance to leaf wilting and several aspects of plant growth and allocation, including plant size, allocation to flowering, and root architecture (Fig. 4). It is important to note that this measure of drought tolerance was made on different plants than those used for measures of growth rate, allocation to flowering, and root architecture, thereby eliminating sources of spurious correlations found in some attempts to evaluate genetic correlations of tolerance (Brett 2004). Moreover, our measure of drought tolerance also removed potential artifacts of plant size, as measures of leaf number and height at the time of drought were used as covariates in the analysis of proportion of leaves wilted after drought.

The interpopulation genetic correlations could result from correlated selection for drought tolerance and plant growth characteristics if specific combinations of traits had highest fitness. Alternatively, the genetic correlations could result from underlying trade-offs such as physiological constraints. Should the genetic correlations reflect underlying trade-offs, the decline in drought tolerance in domesticated sunflowers could result from direct selection on plant growth and allocation attributes. ANCOVA provided support for this possibility as plant growth rate, allocation to flowering, and root architecture individually explained 78%, 19%, and 36%, respectively, of the difference between domesticated cultivars and wild populations (Tables S2 and S3).

As average root diameter and specific root length were strongly negatively correlated (Fig. S5), the smaller specific root length may also significantly contribute to the variation in drought tolerance of sunflower populations. Root architecture is known to play a role in the drought tolerance of domesticated plants, as cultivars with greater specific root length should have more effective access to soil water reserves (McKersie and Leshem 1994). Although some cultivars are being bred for increased root production (McKersie and Leshem 1994), there is evidence that finer roots are more costly to produce and defend (Schultz et al. 2001), which may contribute to the decreased allocation to reproductive fitness in plants with finer roots, as seen in the native ruderal populations relative to domesticated lineages.

There was little difference between domesticated ancient landrace populations and improved cultivar populations in the proportion of wilted leaves (Fig. 3), however improved cultivar populations tended to have less resistance to stem wilting with drought, perhaps due to heavier inflorescences in the improved cultivar varieties (Fig. 1). Besides larger inflorescences, improved cultivars also had smaller shoots than ancient landrace populations. Both ancient landrace and improved cultivar lineages exhibited greater antagonistic effects of inoculation with AM fungi on average inflorescence mass than nondomesticated plants. The common sunflower generally has low dependence on mycorrhizal fungi, as is commonly true for annual plants (Hoeksema et al. 2010). However during domestication, they evolved further reductions in benefits from mycorrhizal fungi, as has been found in domestication of wheat (Hetrick et al. 1993).

It has been suggested that traits that increase the competitive ability of plants can become disadvantageous when plants are exposed to environmental stress (Grime 1977). We found evidence of this phenomenon, as the largest domesticated populations were also the least drought tolerant. Additionally, fast growing domesticated plants continued to grow with each watering under drought conditions, resulting in high mortality of leaves and increased vulnerability to stem wilting with drought.

EVOLUTION OF WEEDINESS

Weedy populations were intermediate between native ruderal populations and domesticated cultivars in terms of drought response, plant size, and root structure (Figs. 1, 3, and S5). Weedy populations differed from native ruderal populations by having coarser roots, reduced drought tolerance, larger shoot growth, and increased drought tolerance when inoculated with AM fungi. This intermediate pattern of weedy sunflowers has been previously demonstrated for resistance to fungal pathogens (Mayrose et al. 2011), suggesting that growth-drought tolerance and growth-defense trade-offs govern this intermediate phenotype.

These declines in drought tolerance in weedy sunflowers could result from direct or indirect selection on other plant characteristics, as suggested by the negative interpopulation genetic correlations between drought tolerance and plant growth rate, allocation to flowering, and root architecture (Fig. 4). Using an ANCOVA, the growth rate and root architecture explained 39% and 60%, respectively, of the variation in drought tolerance exhibited by native ruderal plants and weedy plants, whereas allocation to flowering was not found to be a significant contributor to differences in drought tolerance. The consistency of evidence that plant growth rate and root architecture mediated changes in drought tolerance during both domestication and the evolution of weediness suggests an underlying trade-off, whereas the inconsistency of effects of allocation to flowering suggests that the negative genetic correlation between drought tolerance and allocation to flowering is spurious and driven by correlated selection.

Trade-offs between relative growth rate and tolerance to drought and/or nutrient limitations have been studied extensively in wild sunflower species, especially in relation to leaf traits (Gross et al. 2004; Ludwig et al. 2004; Donovan et al. 2007; Howard and Donovan 2007; Donovan et al. 2009; Rosenthal et al. 2010; Brouillette and Donovan 2011). In general, wild sunflower species with lower relative growth rates exhibited a reduced leaf area ratio, reduced specific leaf area, and increased nitrogen use efficiency. We suspect that similar mechanisms may account for trade-offs observed in the present study, because domesticated, weedy, and invasive sunflowers have larger and thinner leaves than native ruderal plants (Mayrose et al. 2011). Interestingly, increased water use efficiency was not typically associated with lower relative growth rates in wild species (Donovan et al. 2007), but whether it mediates changes in drought tolerance during domestication and the evolution of weediness remains unclear.

Although we have found little difference in growth patterns, mycorrhizal responsiveness, or overall drought tolerance between U.S. and Australian weedy populations, we have found differences in the relationship between drought tolerance and plant size between these two categories of weeds (Table S2). Interestingly, the U.S. weedy populations share their overall relationship between drought tolerance and plant growth rate with the native ruderal populations, whereas the slope of the relationship for Australian weedy populations is closer to that of domesticated cultivars (Fig. 4a). This pattern is consistent with what is known about the origins of the U.S. and Australian weedy populations. Weedy populations in North America are likely to be derived from native ruderals, whereas invasively weedy sunflower populations in Australia likely originated as crop-wild hybrids (Kane and Rieseberg 2008; Muller et al. 2011).

Association with AM fungi was found to contribute to the drought tolerance of weedy sunflower populations, as weedy populations had fewer wilted leaves and maintained higher soil moisture content with the addition of AM fungi. This result suggests that sunflowers may have evolved greater dependence on mycorrhizal fungi during the evolution of weediness, which is in contrast to Hypericum perferatum, which was found to evolve reduced dependence on mycorrhizal fungi during invasion of North America (Seifert et al. 2009). This difference may result from the highly ruderal nature of native sunflower populations, as early successional species often have low dependencies on mycorrhizal fungi (Reynolds et al. 2003). The increased mycorrhizal dependency of the weedy sunflower is also consistent differences in mycorrhizal responsiveness observed between pairs of native and weedy plants (Callaway et al. 2004; Niu et al. 2007). The higher mycorrhizal response of weedy plants, combined with their intermediate levels of drought tolerance and root architecture, may reflect that U.S. weedy and Australian weedy H. annuus populations have adapted to grow in environments such as to the margins of agricultural lands. These environments may harbor mycorrhizal fungi, but may also have increased access to water. Alternatively, the intermediate levels of drought tolerance and root architecture could reflect crop to wild hybridization, which is common in H. annuus, with introgression persisting in populations for many generations after a single hybridization event (Whitton et al. 1997; Linder et al. 1998).

Overall, our experiments substantiates trade-offs in abiotic stress tolerance along multiple axes of plant traits. Moreover, we find evidence that these trade-offs contribute to phenotypic differences in drought tolerance between native ruderals, weedy, and domesticated populations. Specifically, we find that drought tolerance trades off with plant growth and root architecture, and that these trade-offs can mediate the reductions in drought tolerance with the evolution of weediness and domestication. The evidence of a trade-off between drought tolerance and allocation to flowering is not consistent across population categories, and therefore the negative correlation may be spuriously caused by correlated selection for high growth rates and high allocation to flowering during domestication. Interestingly, we found that AM fungi contributed to increased drought tolerance for agriculturally weedy sunflower populations, but not domesticated or native ruderal populations. Together this work more precisely identifies the inadvertent costs of selection for high growth rates in agricultural settings, as our results suggests that high-yield crops under well-watered conditions will necessarily be more vulnerable to drought stress.


Associate Editor: J. Kelly

ACKNOWLEDGMENTS

We thank the Bever-Schultz laboratory group for comments and encouragement. This study was supported with funding from IU Metacyt, National Science Foundation grants DEB-1050237 and DEB-0919434 and NIH-5 R01 GM092660.

Ancillary