*Correspondence and current address: Aaron Gassmann, Department of Entomology, 410 Forbes Building, University of Arizona, PO Box 210036, Tucson, AZ 85721, USA (tel. +1 520 621 1151; fax +1 520 621 1150; e-mail firstname.lastname@example.org).
1Although substantial genetic variation in photosynthetic rate exists within natural populations, it is unclear how such variation might affect the capacity of plants to tolerate herbivore damage.
2I tested if decreased efficiency of electron transport, caused by a mutation that confers resistance to the herbicide triazine, affected the ability of Amaranthus hybridus to tolerate leaf removal, and if this genotypic effect interacted with water availability.
3Tolerance levels were compared with the root-to-shoot ratio and with compensatory photosynthesis, to investigate the mechanisms of tolerance.
4Triazine-resistant A. hybridus was found to have lower tolerance to herbivory suggesting that herbivores should drive selection for higher photosynthetic capacity.
5Although interactive effects between water availability and photosynthetic genotype were not detected, reduced water availability did cause an overall increase in the ability of plants to tolerate damage.
6Greater tolerance of plants under low water availability and of the TS genotype was associated positively with root-to-shoot ratios, but there was a negative association between greater tolerance under low water availability and compensatory photosynthesis.
7These results suggest that photosynthetic variation is more likely to alter tolerance indirectly though changes in resource allocation, and that direct associations between physiological variables and tolerance may be uncommon in natural systems.
Compensatory photosynthesis is typically correlated with higher stomatal conductance (Kolb et al. 1999), which can elevate the concentration of carbon dioxide within a leaf (Lambers et al. 1998). Increased leaf-level nitrogen concentration often accompany compensatory photosynthesis, although the extent to which this results in higher concentrations of the carbon-fixing protein ribulose-1,5-bisphosphate carboxylase/oxygenase (RUBISCO) is less clear (Ovaska et al. 1993; Morrison & Reekie 1995). Although compensatory photosynthesis is often observed following damage (reviewed in Welter 1989; also see Senock et al. 1991; Ovaska et al. 1993; Morrison & Reekie 1995; Meyer 1997; Meyer 1998) its role as a mechanism of tolerance has not been well established (Stowe et al. 2000). The only evidence explicitly linking compensatory photosynthesis and tolerance comes from a study of two species of Agropyron in which Caldwell et al. (1981) found that interspecific differences in tolerance were mirrored by differences in compensatory photosynthesis.
Increased root-to-shoot ratios may influence tolerance by either increasing a plant's stored carbon reserves, which can be reallocated to above-ground biomass following damage (Prud’homme et al. 1992), or elevating the level of compensatory photosynthesis by increasing the supply of water and nitrogen to leaf tissue (McNaughton 1983). Current evidence that greater tolerance is indeed conferred by higher root-to-shoot ratios comes from interspecific and intraspecific comparisons for both annual and perennial species. Abutilon theophrasti grown at high density had decreased root-to-shoot ratios and lower tolerance (Lee & Bazzaz 1980; Mabry & Wayne 1997). Hochwender et al. (2000) found a positive genetic correlation in Asclepias syriaca between root-to-shoot ratio and tolerance for plants grown at low fertilization. In a comparison of five species of biennial forbs, van der Meijden et al. (1988) found a significant positive association between each species’ level of tolerance and its root-to-shoot ratio. Tolerance in domesticated tomatoes was lower than a wild variety, as was the investment in roots relative to above-ground biomass (Welter & Steggall 1993). These studies point to a general association between greater tolerance of damage and higher root-to-shoot ratios, although the relative importance of compensatory photosynthesis vs. reallocation of stored reserves from root tissue in mediating this effect is less clear.
I tested how tolerance in the annual Amaranthus hybridus L. (Amaranthaceae) is affected by a genetically based reduction in photosynthetic efficiency and by water availability, and addressed the underlying mechanism, by measuring the association of root-to-shoot ratio and compensatory photosynthesis with tolerance. The herbicide triazine acts by blocking electron transport thereby inhibiting photosynthesis (Wolber & Steinback 1984), and resistance is conferred by a point mutation in the psbA gene of the chloroplast genome, which alters the shape of the electron-transport protein that binds plastoquinone (Hirschberg & McIntosh 1983) and greatly diminishes binding by triazine. In the absence of triazine, however, triazine-resistant (TR) plants have slower rates of electron transport and typically display a 10–35% reduction in leaf-level photosynthesis compared to triazine-susceptible (TS) plants (Arntz et al. 1998; Arntz et al. 2000a), although this reduction in photosynthetic rate can be absent at later phenological stages (Dekker & Burmester 1992; Arntz et al. 2000b). Additionally, TR genotypes tend to display lower root-to-shoot ratios (Arntz et al. 2000b). Although the greater allocation of resources to above-ground biomass at the expense of root tissue may compensate for the decreased photosynthetic efficiency of TR plants by increasing their total leaf area (Körner 1991), lower root-to-shoot ratios also can decrease a plant's capacity to tolerate folivory.
While the effect of nutrient availability on tolerance has been considered in several studies (reviewed in Strauss & Agrawal 1999), the effect of water stress has received less attention (but see Maschinski & Whitham 1989; Nowatzki & Weiss 1997). Because water stress often increases the root-to-shoot ratio of plants (Lambers et al. 1998), it might act to mitigate the effects of herbivory by increasing tolerance. Arntz et al. (2000b) showed that the reduction in root-to-shoot ratio in TR vs. TS A. hybridus is more pronounced under lower water availability. Consequently, although lower water availability could cause an overall increase in tolerance for both genotypes, the effect might be greater for the TS A. hybridus.
By establishing the extent to which photosynthetic variation affects tolerance and the potential contingencies of this effect on water availability, this study provides an assessment of how environmental and physiological variation might alter the impact of herbivores on plants, and the mechanisms underlying changes in tolerance.
Research was conducted with triazine-susceptible (TS) and triazine-resistant (TR) genotypes of the pigweed Amaranthus hybridus L.; an herbaceous annual native to the Americas and widely distributed there as an agricultural weed (Weaver & McWilliams 1980). TR and TS genotypes were collected from populations in Beltsville, Maryland and Blacksburg, Virginia. Maternal lines of A. hybridus were generated by reciprocally crossing TR and TS plants collected from the same population (i.e. Maryland or Virginia) to produce F1 individuals. Because the triazine resistance is a maternally inherited trait conferred by a mutation in the chloroplast genome, reciprocally crossing plants randomized any pre-existing differences in the nuclear genome between TR and TS lines (Jordan 1996). F1 lines were self-pollinated for two generations to produce F2 and F3 maternal lines. The resistance or susceptibility of lines to triazine was confirmed through fluorometry and bioassays with triazine (see Jordan (1996) for additional details on the collection and crossing of these lines).
tolerance across a gradient of damage
This experiment used seven TR and six TS F2 lines of A. hybridus originating from Maryland. Seedlings were started in a Percival growth chamber (8/16 L/D, 35 °C/25 °C) following Jordan (1996). Seedlings from each line were started en masse within 460 cm3 pots using Pro-Mix BX (Premier Horticultural Ltd, Dorval, Canada). After 13 days, pots were removed from the growth chamber and placed in a glasshouse under supplemental incandescent and fluorescent lights (16/8 L/D), and were treated with Benomyl 50 W Systemic Fungicide (Gabriel Chemicals, Paterson, New Jersey) to arrest damping off. The following day, seedlings were transplanted singly into 1.8 L pots filled with Pro-Mix BX.
Plants were grown in a glasshouse under 16-hour days for the first 76 days after transplanting, and then placed under 12-hour days to stimulate flower and seed production (Weaver & McWilliams 1980). A canvas suspended from cables, which completely covered both the top and sides of the glasshouse bench, was used to cover and uncover plants daily in order to shorten day length. Plants were randomized at 3-week intervals and watered as needed. Sixty-one days after transplanting, plants were fertilized to flow-through with Miracle-Gro Professional Excel Cal Mag 15 : 5 : 15 (N : P : K) (The Scotts Company, Marysville, Ohio) at a concentration of 300 p.p.m. nitrogen. Ninety-six days after transplanting, most of the flowering was completed and brown pericarps, indicating mature seeds, were common on the infructescences. Reproductive biomass and above-ground vegetative biomass were harvested 108 days after transplanting. In A. hybridus, reproductive biomass is highly correlated with seed production (r = 0.98) (Jordan 1996). All plant parts were dried at 60 °C and allowed to equilibrate with ambient laboratory conditions before they were weighed to the nearest hundredth of a gram.
Four vegetative clipping treatments (removing 0, 30, 60 or 90% of leaf area from all previously unclipped leaves over 2 cm in length) were crossed with two floral clipping treatments (0 or 60% removal). Vegetative clipping was conducted 36, 58 and 80 days after transplanting, and floral clipping 80 days after transplanting. Leaf tissue was removed with scissors by making a single cut, perpendicular to the mid-vein. Because A. hybridus has simple ovate leaves, visual inspection was used to determine the amount of leaf tissue to remove, following practice trials with digitally scanned images of randomly selected A. hybridus leaves. The image analysis program Scion Image Beta 3b (Scion Corporation, Fredrick, Maryland) showed that the accuracy possible with visual inspection was: intended 30% vs. actual 31.8% ± 3.16%, intended 60% vs. actual 59.4% ± 3.48%, intended 90% vs. actual 89.3% ± 2.52% (mean ± SD, n = 17 in all comparisons). Only plants with a terminal inflorescence that extended beyond the upper-most leaves after 80 days were included in the experiment and the upper 60% of this was removed in the floral clipping treatment. At the end of the experiment, the average sample size in each combination of genotype (2) by floral clipping (2) by vegetative clipping (4) was 9.31 ± 1.45, for a total of 149 plants. Within each genotype (TR vs. TS), plants were assigned at random to combinations of floral and vegetative clipping, with 4.75 ± 0.86 maternal lines from each genotype in each floral by vegetative damage combination.
All analyses were based on a mixed-model analysis of variance (anova) using PROC MIXED in SAS 8.02 (SAS Institute 1999). All means were calculated as least-squares means. Reproductive biomass and vegetative biomass were log transformed to ensure normality of the data. Genotype and floral clipping were coded as categorical variables and vegetative clipping was treated as a continuous variable. An interaction between plant genotype and vegetative clipping would indicate that the slope of the linear regression for a response variable (e.g. reproductive biomass) differed between TR and TS plants, and consequently, that the genotypes differed in tolerance. Maternal line and its interactions with floral and vegetative clipping were included as random factors in the anova. Consequently, the anova followed a split-plot design because maternal line was nested within plant genotype but crossed with floral and vegetative damage. The significance of random effects was tested based on a log-likelihood ratio statistic (−2 RES Log Likelihood), which follows a χ2 distribution with a P-value equal to one-half the probability of obtaining a greater χ2-value assuming one degree of freedom (Littell et al. 1996:44). Where maternal line or its interactions were not significant at a level of α < 0.25, these factors were excluded from the model following Quinn & Keough (2002).
interaction between water availability and photosynthetic genotype
Plants from six TR and six TS F3 lines, originally collected in Blacksburg, Virginia, were assigned to a fully crossed design with two levels of water availability (high and low) crossed with two levels of damage (undamaged and 30% clipping). Eighteen plants per genotype (TR or TS) were assigned at random to each damage-by-water treatment, and each combination of genotype by damage by watering level contained 5.38 ± 0.52 maternal lines.
Seedlings were started in a growth chamber (8/16 L/D, 25 °C/35 °C) in individual pots (volume = 143 cm3) filled with Pro-Mix BX and then moved to a glasshouse, with natural light augmented by fluorescent and incandescent lighting (16/8 L/D). After 20 days, seedlings were transplanted into 6 L pots filled with a potting medium composed of equal parts Pro-Mix BX and Supercourse Perlite (The Schandler Company, Metuchen, New Jersey). From 4 days after transplanting, plants were either watered every third day or every sixth day and from 28 days, were fertilized every sixth day (on the same day as watering) with Miracle-Gro Professional Excel Cal Mag 15 : 5 : 15 (N : P : K) at a concentration of 100 p.p.m. nitrogen. The position of plants in the glasshouse was randomized weekly.
Clipped plants had 30% of available leaf tissue removed (from all previously undamaged leaves over 4 cm in length) between 51 and 55 days after transplanting, and again after 78–83 days. Clipping was administered with a paper punch that removed 0.32 cm2 disks. The quantity of leaf tissue to remove at first clipping was estimated using a linear regression of leaf area onto mid-vein length obtained from five leaves sampled at random for each of seven plants in each treatment. No difference among treatments was found in regression slopes (d.f. = 3,132; F = 1.437; P = 0.2348) or intercepts (d.f. = 3135; F = 0.588; P = 0.6242) and a single regression equation was therefore used for all treatments (r2 = 0.90). For the second round of clipping, regressions were based on six undamaged leaves from each of seven previously clipped plants in each treatment. Because these data displayed a non-linear relationship, a second-order regression was used. No significant difference was found among treatments for either the first (d.f. = 3,159; F = 1.25; P = 0.2923) or the second (d.f. = 3,159; F = 1.3; P = 0.2753) order terms, and groups did not differ in their intercepts (d.f. = 3,165; F = 1.18; P = 0.3188). All data were pooled into a single, second-order regression (r2 = 0.97).
Measurements of photosynthetic rate and transpiration were taken over a period of 8 days, starting seven days after first clipping treatment. At that time, plants were still growing vegetatively, although early signs of a terminal floral bud were present on some plants. Measurements were taken with a LI-6400 portable infra-red gas analyser with a temperature- and light-controlled cuvette (Licor Inc, Lincoln, Nebraska). PAR was set at 1500 (a light intensity at which photosynthesis was close to the maximum rate), relative humidity was 55–65%, and the concentration of CO2 was at the average atmospheric level of 360 p.p.m. Measurements were taken for TR and TS plants in all treatments, and approximately equal numbers of plants from each treatment-by-genotype combination were measured each day. At the first clipping, the sixth well-differentiated leaf below the apical meristem was marked and left intact for measurements of photosynthesis.
After all photosynthetic measurements were completed (71 days after transplanting), the total area of new leaves over 4 cm in length was measured for all plants on which photosynthetic data were collected. Leaf area was estimated as before the second clipping. The second order regression terms for leaf area onto mid-vein length were found to differ between damaged and undamaged plants. Consequently, two separate regression equations were used (r2 = 0.92 and 0.95 for undamaged and damaged plants, respectively).
Plants were harvested 112–118 days after transplanting. Reproductive biomass and above-ground vegetative biomass were measured as above, and root biomass harvested by gently washed root tissue to remove the potting medium before drying.
All analyses were based on either a mixed-model anova or mixed model analysis of covariance (ancova) using PROC MIXED in SAS 8.02, and all mean values were calculated as least-squares means. Reproductive biomass and total vegetative biomass (above-ground plus root biomass) were analysed with a mixed-model anova. Plant genotype, damage from clipping, and watering treatment were analysed as fixed factors, with a difference in tolerance indicated by a significant interaction between damage and the model's other main effects (i.e. water treatment and plant genotype). Maternal line was analysed as a random factor nested within genotype and where this factor or its interactions were not significant at a level of α < 0.25, terms were excluded from the final model. Because there was a significant interaction between clipping and water availability for reproductive biomass (see Results), a pairwise comparison was made between damaged and undamaged plants in both the high and low water treatments using the PDIFF option in SAS, with the significance level set at 0.025 based on the Dunn-Šidák correction assuming two pairwise comparisons (Sokal & Rohlf 1995). Pairwise comparisons also were made between TR and TS plants in the high and low water treatments where there was a marginally significant interaction (see Results).
The ratio of root biomass to above-ground biomass (root-to-shoot ratio) was log transformed to ensure normality and analysed by anova considering only undamaged plants. Measurements of new leaf area produced were analysed with an ancova using the factors of genotype, watering treatment and clipping damage. Time that plants grew was included as a covariate because this affected total leaf area.
Photosynthetic rate was analysed using a similar ancova. Least-squares means indicate the mean expected if all leaves had been the same age. A significant interaction was present between clipping and water availability (see Results), and consequently, a pairwise comparison was made between damaged and undamaged plants in both the high and low water treatments following the same procedure used with reproductive biomass. Water use efficiency was calculated as photosynthetic rate divided by transpiration (mol m−2 s−1), and analysed with an ancova using the factors of plant genotype, watering treatment, clipping damage, and the covariate of leaf age.
A tolerance score was assigned to each damaged plant by dividing its reproductive biomass by the mean reproductive biomass of undamaged plants from the same maternal line and watering treatment. In cases where data on undamaged plants were not available (3 out of the 24 instances), the mean reproductive biomass mass from undamaged plants from the same watering treatment and photosynthetic genotype (TR or TS) was used. To ensure normality, these tolerance scores were log transform. Data were analysed first with an anova using the factors of genotype and water availability, to test if the differences in tolerance were consistent with those in the previous analysis. ancova with root-to-shoot ratio as a covariate was then used to test if root-to-shoot ratios could explain differences in the level of tolerance observed among treatments.
tolerance across a gradient of damage
A significant interaction was present between genotype and vegetative clipping for reproductive biomass (Table 1), which indicates that the genotypes differed in their level of tolerance. Although both genotypes had similar levels of reproductive biomass when undamaged, the reduction in reproductive biomass under clipping was greater for TR than TS plants (Fig. 1). For vegetative biomass, there was no significant effect of genotype and none of the interactions between genotype and floral or vegetative damage were significant (Table 2), although vegetative biomass did decrease significantly with clipping.
Table 1. Analysis of variance for reproductive biomass of Amaranthus hybridus across vegetative and floral clipping treatments. Factors in the analysis include triazine-resistant and triazine-susceptible genotypes (Genotype), the presence or absence of floral clipping at a level of 60% (Floral Clipping), and the removal 0, 30, 60 or 90% of leaf tissue (Vegetative Clipping). Random factors included in this model were Line (Genotype) (χ2 = 1.3, d.f. = 1, P = 0.13) and Floral Clipping × Line(Genotype) (χ2 = 1.5, d.f. = 1, P = 0.11). *P < 0.05, ***P < 0.001
Table 2. Analysis of variance for above-ground vegetative biomass of Amaranthus hybridus. Random factors included in this model were Line(Genotype) (χ2 = 2.8, d.f. = 1, P < 0.05) and Vegetative Clipping × Floral Clipping × Line(Genotype) (χ2 = 0.8, d.f. = 1, P = 0.19). ***P < 0.001
interaction between water availability and photosynthetic genotype
This experiment found no evidence for a difference in tolerance between TR and TS plants, as indicated by the lack of a significant genotype-by-damage interaction or a significant three-way interaction for either reproductive or vegetative biomass (Tables 3 and 4). A significant interaction between watering treatment and clipping was, however, present for reproductive biomass (Table 3). Although reproductive biomass was similar between damaged and undamaged plants in the low water treatment, clipping significantly reduced reproductive biomass of both genotypes in the high water treatment (Fig. 2), showing that plants grown under low water availability had higher tolerance. There was a marginally significant interaction between plant genotype and water level (Table 3), because, although reproductive biomass was higher for TS than TR A. hybridus across all four treatments, this difference was greater at high water availability (Fig. 2). When damaged and undamaged plants were grouped within watering treatment, a significant pairwise difference was present between TR and TS plants grown at high water availability but not at low water availability (Fig. 2). Total vegetative biomass was significantly affected by plant genotype and clipping (Table 4); TS plants had higher vegetative biomass across all treatments and clipping decreased vegetative biomass.
Table 3. Analysis of variance for reproductive biomass of Amaranthus hybridus with two levels of leaf removal (0 or 30%). Line(Genotype) was included as a random factor in the model (χ2 = 22.3, d.f. = 1, P < 0.001). *P < 0.05, **P < 0.01, ***P < 0.001, (*)P = 0.07
Among undamaged plants, the root-to-shoot ratio was significantly greater in TS plants and at lower water availability (Table 5, Fig. 3), but a significant interaction between plant genotype and water availability was not detected.
Table 5. Analysis of variance for the ratio of root biomass to above-ground vegetative biomass of undamaged Amaranthus hybridus. *P < 0.05, **P < 0.01
The photosynthetic rate of early reproductive plants was affected by plant genotype and by the interaction between damage and water availability (Table 6). Generally, TR plants displayed higher rates of photosynthesis than TS plants (Fig. 4). For plants grown at low water availability, photosynthetic rates were almost identical between undamaged and clipped plants. However, for plants in the high water treatment, photosynthetic rate was significantly higher for clipped plants (Fig. 4), indicating that compensatory photosynthesis occurred only in the high water treatment. No significant main effects or interactions were found for water use efficiency (Table 7) or for the total leaf area produced following damage (Table 8).
Table 6. Analysis of covariance for photosynthetic rate of Amaranthus hybridus. The age of a leaf at the time photosynthetic measurements were taken was included as a covariate. *P < 0.05, ***P < 0.001
Table 8. Analysis of covariance for production of new leaf area following damage. Water × Clipping × Line(Genotype) was included as a random factor in the model (χ2 = 0.9, d.f. = 1, P = 0.17). ***P < 0.001
An anova of ratio-based scores for tolerance revealed a marginally significant effect of water availability on tolerance (Table 9), with plants in the low water treatment displaying higher tolerance (Fig. 5, cf. reproductive biomass in Fig. 2). In the ancova, the factor of water was no longer significant, but root-to-shoot ratio explained a significant portion of the total variation in tolerance (Table 9).
Table 9. Analysis of tolerance scores for Amaranthus hybridus. Tolerance was calculated as the ratio of reproductive biomass for damaged plants divided by the mean value of undamaged plants from the same photosynthetic genotype, water treatment, and maternal line. Root-to-shoot ratio was either excluded or included as a covariate. (*)P = 0.08; *P = 0.05
Although genotypes within populations can differ by approximately 25–55% in maximum photosynthetic rate (calculated as (largest value – smallest value)/(largest value)) (Zangerl & Bazzaz 1983; Scheiner et al. 1984; Geber & Dawson 1990; Sultan & Bazzaz 1993; Geber & Dawson 1997), little is known about how such variation might affect levels of tolerance, and consequently, how herbivores might impose selection on the photosynthetic capacity of plants. Even in the absence of additional ecological factors such as herbivory, understanding the relationship between photosynthesis and fitness remains a challenge. This arises because variation in the partitioning of photosynthate can obscure the direct relationship between photosynthesis and fitness (Lambers 1987; Nelson 1988; Körner 1991) and because correlations between photosynthesis and other physiological traits affecting carbon acquisition (e.g. water use efficiency) may have a stronger direct effect on fitness (reviewed in Arntz & Delph 2001; Geber & Griffen 2003).
The TR and TS genotypes of A. hybridus present an unusual opportunity to consider how photosynthetic variation affects fitness in the context of plant–herbivore interactions. Because TR and TS lines were reciprocally crossed, genotypic differences in tolerance can be attributed directly to differences in photosynthetic efficiency or to accompanying pleiotropic effects that this variation engenders. Based on the comparison of reproductive biomass across levels of vegetative damage in the first experiment, tolerance was lower for TR than TS plants (Fig. 1, Table 1), suggesting that herbivores should select for higher photosynthetic rate because of the capacity of this trait to increase tolerance. By contrast, Mooney & Gulmon (1982) have suggested that herbivorous insects should select for lower photosynthetic rates because the associated decrease in leaf-level nitrogen concentrations might reduce herbivory. The extent to which herbivorous insects might shape the evolution of photosynthetic capacity most likely represents a balance between the benefit of greater tolerance and the cost of higher damage.
Although some have suggested that a broad spectrum of damage levels should be used to quantify tolerance, others favour the use of only naturally occurring levels of damage (reviewed in Stowe et al. 2000). Amaranthus spp. can be subject to high rates of infestation by herbivorous insects (Clark-Harris & Fleischer 2003), and have been found to suffer complete defoliation under field conditions (Mitchell 1984). Consequently, the range of damage used in the first experiment falls within the levels of damage expected to arise in populations of this agricultural weed. Although controversy still exists over the relative merits of natural vs. simulated damage (Tiffin & Inouye 2000; Inouye & Tiffin 2003; Lehtilä 2003), the use of simulated damage has several advantages when quantifying tolerance. For example, the level of damage will not be biased by genotypic differences in defensive chemistry and nutritional quality (Tiffin & Inouye 2000) and variation in fitness costs associated with resistance to herbivores will not confound the assessment of tolerance (for a review see Stowe et al. 2000). Although factors such as inducible chemical defences are expected to alter the magnitude of genotypic differences in tolerance between natural and simulated herbivory (see Strauss & Agrawal 1999 for a review), artificial damage allows for a more accurate assessment of whether genotypic differences in tolerance actually exist (Tiffin & Inouye 2000).
In the second experiment, there was not a significant interaction between genotype and clipping or a significant three-way interaction, and F-ratios were close to unity indicating that the effect size of any interaction was small (Sokal & Rohlf 1995, p. 195). Consequently, the failure to detect a significant effect of genotype on tolerance appears not to have resulted from lack of statistical power, but rather from the absence of genotypic differences in tolerance to 30% damage. This interpretation is consistent with the first experiment in which values of reproductive biomass only showed clear patterns of segregation at 60% and 90% leaf removal, with reproductive biomass higher in TS than TR lines.
However, the higher tolerance of plants in the low than in the high water treatment (Fig. 2, Table 3) provides some insight into the mechanistic basis of tolerance in A. hybridus. This experiment found no evidence that compensatory photosynthesis increases the level of tolerance. Only plants in the high water treatment showed a significant change (an increase) in photosynthesis following damage (Fig. 4, Table 6), despite their tolerance being significantly less than that of plants in the low water treatment (Fig. 2, Table 3). It is possible that the mechanisms of tolerance might differ with environment, and compensatory photosynthesis could have contributed to the ability of plants in the high water treatment to tolerate damage, albeit less effectively than in the low water environment. Alternatively, the increased carbon acquisition conferred by higher photosynthetic rate may have gone to the production of defensive agents. Amaranthus is defended by phenolics (Teutonico & Knorr 1985; Wesche-Ebeling et al. 1995), which are carbon-based compounds. The induction of higher levels of defences in plants following damage is increasingly recognized as an important aspect of interactions between plants and herbivorous insects (Karban & Baldwin 1997) and past work has documented elevated levels of phenolic compounds following damage (Keinänen et al. 1999).
Root-to-shoot ratio was positively related to patterns of tolerance, with the low water treatment increasing both tolerance and root-to-shoot ratio (Table 5, Fig. 3). The role of a higher root-to-shoot ratio in conferring greater tolerance was further supported by the analysis of ratio-based scores of tolerance in which the marginally significant difference in tolerance between watering treatments (Fig. 5) was explained by including the covariate of root-to-shoot ratio in the analysis (Table 9). The lack of concordance between compensatory photosynthesis and root-to-shoot ratio suggests that reallocation of carbon from root tissue to above-ground biomass, rather than an effect on photosynthesis, was responsible for greater tolerance. Several studies have documented the mobilization of stored carbon from stem and root tissue following damage (Danckwerts & Gordon 1987; Prud’homme et al. 1992; Morvan-Bertrand et al. 1999; Lawson et al. 2000). Amaranthus hybridus possesses a tap root (Weaver & McWilliams 1980), which was found to account for 42.5% ± 3.4% of the total root biomass 39–46 days after transplanting in the second experiment (n = 52). Given the role of taproots in the storage of carbohydrates and nitrogen in other species (Lambers et al. 1998), it is likely that pre-flowering allocation of resources to the taproot is the primary mechanism by which A. hybridus can tolerate folivory.
Although root-to-shoot ratio was not measured in the first experiment, it appears that the higher tolerance of TS plants was associated with higher root-to-shoot ratios. TS plants had a significantly higher root-to-shoot ratio than TR plants (Table 5) in both watering treatments (Fig. 3), as well as in other environments, including higher availability of water and shading (Arntz et al. 2000b). This indicates that higher photosynthetic efficiency leads to enhanced tolerance by elevating root-to-shoot ratio, rather than through direct effects on photosynthetic rate, as suggested by the data of Strauss et al. (2003) with Raphanus raphanistrum where tolerance had a strong positive correlation with root biomass and only a weak negative correlation with photosynthetic rate. Consequently, it may be difficult to demonstrate the role of variation in physiological parameters, such as photosynthetic rate, in affecting tolerance in natural systems because changes in allocation patterns by plants in response to physiological variation may in itself affect patters of tolerance.
Although the triazine-resistance trait is typically associated with lower photosynthetic rates in A. hybridus (Ahrens & Stoller 1983; Ort et al. 1983; Arntz et al. 1998; Arntz et al. 2000a) and in other species (Sunby et al. 1993; Holt 1988; Jursinic & Pearcy 1988) exceptions do exist (Dekker & Burmester 1992; Arntz et al. 2000b). Arntz et al. (2000b) found that TR plants typically have a lower photosynthetic rate early in development (not measured here), but, at the pre-reproductive stage (the stage at which photosynthesis was measured in this study), TR genotypes can have higher photosynthetic rates than TS plants. A similar pattern was found for TR and TS Brassica napus (Dekker & Burmester 1992). In general it appears that TR plants delay the onset of reproduction (Beversdorf et al. 1988; Kremer & Kropff 1999) and that the reversal of photosynthetic rates between genotypes at later phenological stages might arise from a delay in the reallocation of nitrogen from photosynthetic proteins, such as RUBISCO, to reproductive tissue. Such a strategy could enable TR plants to offset the fitness cost of decreased photosynthetic rate during earlier phenological stages. Although the TR and TS genotypes differed only in electron transport efficiency, this single genetic change appears to have multiple phenotypic consequences, including shifts in root-to-shoot ratio and contrasting effects on photosynthetic rate between early and late phenological stages.
Arntz et al. (2000b) found that TS plants had significantly greater floral biomass when grown at a level of moderate water stress (Ψ = −0.71) approximating field conditions, but that under very low water stress (Ψ =−0.21 MPa) the TR genotype had higher fitness (assuming a permanent wilting point of 1.5 MPa, Lambers et al. 1998). The authors attributed this effect to the higher water use efficiency of TS compared to TR plants. In this experiment, there was a marginally significant interaction between genotype and water availability (Table 4) and TS plants had significantly greater floral biomass than TR plants only in the high water treatment. The level of water availability was, however, similar in high (Ψ = −0.648 ± 0.255 Mpa, n = 8) and low (Ψ =−0.638 ± 0.256 Mpa, n = 5) water treatments, presumably because the shift toward greater root production in the low water treatment offset decreased water availability. In contrast to Arntz et al. (2000b), no differences in water use efficiency were found between genotypes (Table 7), indicating that other physiological factors can mediate fitness differences between these genotypes and that there is an intermediate range of water availability over which decreased efficiency in electron transport will affect fitness.
The reduction in photosynthetic efficiency imposed by the triazine-resistance trait caused plants to suffer lower tolerance in the presence of artificial herbivory. This effect was not the direct result of changes in photosynthetic rate but rather the consequences of a shift in biomass allocation toward lower root-to-shoot ratio on the part of TR plants. Although greater allocation of resources to above-ground vs. below-ground tissue may compensate for the decreased photosynthetic efficiency of TR plants by increasing their total amount of photosynthetic tissue, this shift in allocation reduced tolerance. As is the case with other traits, such as resistance to herbivores (Strauss et al. 2002), this study points to the presence of a strong component of ecological dependence in the strength of natural selection on photosynthetic traits.
I thank D. Futuyma, C. Janson, M. Lerdau and two anonymous referees for helpful comments on this manuscript. N. Jordan generously provided the Amaranthus lines used in this study and H. Throop provided assistance with photosynthetic measurements. My research was supported by a Doctoral Dissertation Improvement Grant (DEB 0206448) from the National Science Foundation and a GAANN Fellowship from the US Department of Education. This is Contribution No. 1124 in Ecology and Evolution from the State University of New York at Stony Brook.