P. L. Valverde, Departamento de Biologia, Universidad Autónoma Metropolitana, Unidad Iztapalapa, Apartado Postal 55–535, México 09340, Distrito Federal, México. e-mail: firstname.lastname@example.org
This study assessed the role of leaf trichome density as a component of resistance to herbivores, in six populations of Datura stramonium. Phenotypic selection on plant resistance was estimated for each population. A common garden experiment was carried out to determine if population differences in leaf trichome density are genetically based. Among population differences in leaf trichome density, relative resistance and fitness were found. Leaf trichome density was strongly positively correlated to resistance across populations. In 5 out of 6 populations, trichome density was related to resistance, and positive directional selection on resistance to herbivores was detected in three populations. Differences among populations in mean leaf trichome density in the common garden suggest genetic differentiation for this character in Datura stramonium. The results are considered in the light of the adaptive role of leaf trichomes as a component of defence to herbivores, and variable selection among populations.
Besides defence, leaf trichomes may serve other physiological functions, hence selection on the antiherbivory role of leaf trichome density can either be constrained or synergistically favoured by selection imposed by other environmental stresses (Bell, 1997; Roy et al., 1999). For instance, if leaf trichome density is correlated with other leaf characteristics, selection on those traits can produce changes in trichome density (Björkman & Anderson, 1990; Roy et al., 1999) without being the target of selection. In addition, as for other plant traits, phenotypic variation in leaf trichome density may have both genetic and environmental (and their interaction) causal factors (Falconer & Mackay, 1996). Within-population genetic variation in leaf trichome density will lead to an evolutionary change provided the trait is under selection. However, environmentally determined phenotypic variation in leaf trichomes (i.e. phenotypic plasticity; Ågren & Schemske, 1994; Schlichting & Pigliucci, 1998), may limit response to selection.
The present study aimed specifically to address the following questions: (1) is individual variation in leaf trichome density related to resistance to herbivores? and if so (2) is resistance selectively favoured within populations? Given that populations of Datura stramonium occur in a broad variety of plant communities (Núñez-Farfán, 1991), phenotypic differences among populations in leaf trichome density might be both environmentally and genetically based. Thus, we also asked if (3) the relationship between leaf trichome density and resistance differs across populations of Datura stramonium, and (4) to what extent phenotypic differences are the result of phenotypic plasticity or genetic differentiation among populations. Under the null hypothesis that variation in leaf trichome density in Datura stramonium occurs for other reasons (e.g. temperature regulation), it is not expected to be related to resistance to herbivores.
In six populations of Datura stramonium from the central part of Mexico (four States) all natural growing individual plants (16–46) were marked and, at reproduction, foliar damage produced by herbivorous insects and the number of mature fruits were recorded. The six populations occurred in different plant communities: one in a tropical dry forest, two in pine–oak forests, and three in xerophytic shrub communities (Table 1). Distances between pairs of populations ranged from 20 to 300 km.
Table 1. Environmental characteristics of six populations of Datura stramonium in central Mexico.
In each population the following data were taken for each individual plant: (1) total number of branches, (2) total number of fruits, and (3) average seed number per fruit (seed-set), from a sample of 10 fruits per plant.
Relative resistance, trichome density and fitness
A large random sample (mean=31.62 standard error=0.66) of fully expanded leaves was collected from each individual plant and measured with a leaf-area meter (Delta-T Devices, Cambridge, UK) to obtain standing leaf area (i.e. remnant undamaged leaf area). For each plant, relative damage was obtained by dividing consumed leaf area (CLAi) by total leaf area (TLAi). Original total leaf area was estimated using a regression analysis of leaf area as a function of leaf length following Núñez-Farfán & Dirzo (1994),. Since leaf shape (hence, leaf area) varied slightly among populations, four different equations were applied to estimate original total leaf area (R2 ranging from 0.964 to 0.987, P < 0.001, n=30–120). Relative resistance to herbivores was estimated for each plant as 1 – (CLAi/TLAi) following previous studies (Berenbaum et al., 1986; Fritz & Price, 1988; Simms & Rausher, 1989; Núñez-Farfán & Dirzo, 1994; Núñez-Farfán et al., 1996; Tiffin & Rausher, 1999). For statistical analyses, resistance was arcsine-transformed to normalize its error distribution (Sokal & Rohlf, 1995).
Trichome density was measured as the total number of trichomes within an area of 2.5 mm2 on the basal central area of the adaxial side of the leaf (see Mauricio et al., 1997), using a dissecting microscope. This sampled area of the leaf gives a good estimate of the whole-leaf average trichome density: for 30 randomly chosen leaves from 15 different plants, trichome density in 2.5 mm2 was highly correlated with the average trichome density of nine other 2.5-mm2 areas within the same leaf (R2=0.81, F1,13=60.1, P < 0.0001). Thus, in each population, average trichome density of 16–20 plants was calculated on a sample of 10 randomly chosen fully expanded mature leaves for each plant. For statistical analyses, trichome density was square-root transformed to normalize its error distribution (Sokal & Rohlf, 1995).
Maternal plant fitness was estimated as the average seed-set per fruit times total fruit number per plant. Because absolute maternal fitness varies with plant size (Núñez-Farfán, 1991), the analysis of fitness as a function of plant resistance to herbivores was made in each population using the residuals from the regression analysis (see Sinervo, 2000) of absolute maternal fitness (total seed number per plant) on plant size, as estimated by the total number of branches. Total number of seeds and branches were square-root transformed before statistical analyses (Sokal & Rohlf, 1995). Hereafter, residuals for maternal fitness will be referred to simply as fitness.
The effect of trichome density on plant resistance to herbivores among and within populations was analysed using covariance analysis (ANCOVA), under the null hypothesis that trichome density, the covariate, is not a plant resistance component. In the same way, the effect of leaf area on trichome density was analysed (see Roy et al., 1999). Differences in average values among populations in plant resistance and trichome density were obtained through Tukey–Kramer HSD tests (Sokal & Rohlf, 1995). The relationship between fitness and plant resistance, within and among populations, was analysed by means of ANCOVA, where plant resistance was the covariate. The relationship between average resistance and average trichome density per population was assessed by means of a Spearman rank correlation (RS) (Sokal & Rohlf, 1995, p. 598). The analyses were carried out using the JMP® statistical package (SAS Institute, 1995).
In order to estimate phenotypic directional selection gradients (βi) for each population, linear regression analysis of individual fitness (wi) as a function of the standardized resistance to herbivores (X=0 and s2=1) were performed (Lande & Arnold, 1983; Mitchell-Olds & Shaw, 1987; Nagy, 1997). Selection coefficients and standard errors were estimated using FREE-STAT (version 1.10; Mitchell-Olds, 1989). Jackknife estimates of the standard errors of the selection coefficients were also obtained (FREE-STAT). The Jackknife procedure permits approximate t-tests of significance which are robust to deviations from normality and to heterogeneity of residual variances (Mitchell-Olds, 1989). We estimated the selection coefficients only on resistance to herbivores following the reasoning that leaf trichome density is a putative component of resistance and correlated with it (Mauricio et al., 1997; van Dam & Hare, 1998b; Elle et al., 1999). Data of this study indicated no relationship between trichome density and fitness. Then a covariance analysis was performed to assess if trichome density is related to resistance (see Results). Because trichome density is correlated with resistance in 5 of 6 populations, this validates our criterion for not including trichome density in the selection analyses given the lack of independence between both traits (Mitchell-Olds & Shaw, 1987).
Population differentiation in trichome density
To determine possible genetic differences among populations in leaf trichome density, natural progenies (sensuLawrence, 1984; hereafter families) from three populations were collected and grown in a common garden. Plants of a given family were derived from a single fruit and related as half- or full-sibs. Given the size of individual plants and the number of populations sampled in the field, three randomly selected populations were grown in the common garden due to space limitations. The three populations grown were: Population I, 15 families and 133 plants; Population II, 14 families and 118 plants; Population III, 10 families and 81 plants (cf. Table 1). Total sample size was 332. The common garden (59 × 13 m) was located in an area within the Pedregal de San Angel Ecological Preserve (National Autonomous University of Mexico; UNAM) where Datura stramonium grows naturally (Núñez-Farfán & Dirzo, 1994). The seeds were germinated in the greenhouse (protocol in Fornoni & Núñez-Farfán, 2000), and then transplanted to the common garden under a complete randomized design once the first pair of true leaves appeared. Plants were spaced 1 m apart in a regular grid. When plants reached maturity (reproduction), trichome density was estimated for all plants, following the same methodology employed for field collected plants (see above). A nested-analysis of variance was performed to test differences due to population and family (within population) (Sokal & Rohlf, 1995), using the JMP® statistical package (SAS Institute, 1995).
Trichome density and resistance to herbivores
Populations experienced different average levels of damage (10–50% of total leaf area) (Table 2). In each population, all individual plants had some degree of foliar damage. In all populations, leaf damage was caused mainly by tobacco flea beetles (Epitrix spp., Coleoptera: Chysomelidae). Trichome density varied from 3.041 to 15.429 trichomes × mm–2 (Table 2). ANCOVA detected statistically significant differences among populations in plant resistance to herbivores, and a significant effect of trichome density on plant resistance (Tables 2 and 3a). Furthermore, the significant trichome density × population interaction indicated that the slope for the relationship between trichome density and resistance varied among populations (Table 3a). In contrast, differences among populations in trichome density were not related with leaf area (Table 3b). In five out of six populations, a significant relationship between trichome density and plant resistance was detected and the explained variance (R2) ranged from 0.50 to 0.68. Populations I, II, IV and V showed positive relationships, whereas population VI had a concave downward relationship between leaf trichome density and resistance (Table 2). Multiple comparisons also showed differences in trichome density and resistance among populations (Table 2). Finally, population mean resistance and trichome density were highly positively correlated across populations (Fig. 1, RS=0.83, n=6, P=0.0416).
Table 2. Mean leaf trichome density (trichomes × mm–2) per plant (SE), plant resistance to herbivores (1 – relative damage) (±1 SE) and regression analysis of relative resistance (y) on trichome density (x) in six populations of Datura stramonium from central Mexico. Different letters for each character indicate significant among-population differences at P < 0.01 (see Materials and methods).
Table 3. Analyses of covariance for plant resistance to (a) herbivores, (b) trichome density (trichome × mm–2) and (c) fitness in Datura stramonium. All F-ratios were based on type-III sums of squares.
Phenotypic selection of resistance across populations
ANCOVA revealed significant differences among populations in fitness (Table 3c). Also, resistance to herbivores had a significant effect on plant fitness. The significant interaction between plant resistance and population on fitness indicates that the slope of the relationship between fitness and plant resistance differed among populations (Table 3c and Fig. 2). These results suggest differences among populations in the effectiveness of resistance against herbivorous insects, and imply that a similar amount of damage had different consequences on plant fitness among populations.
Significant directional selection coefficients were detected in three out of six populations. In Populations I, II and IV, resistance was positively favoured, indicating that higher fitness was attained by those plants with higher levels of resistance (Table 4 and Fig. 2). After Jackknife procedure, selection coefficients for these populations remained significant (Table 4). In contrast, no selection on resistance was detected in populations III, V and VI (Table 4 and Fig. 2).
Table 4. Directional selection coefficients (β) and standard errors (SE) of resistance to insect damage of six populations of Datura stramonium. R2 for the lineal models and Jackknife estimates for the significant selection coefficients are provided. Sample sizes correspond to those of Table 2.
Population differentiation in trichome density
Nested ANOVA revealed significant differences among populations in leaf trichome density, whereas no significant differences among families within population were found (Table 5). Plants from population I had a significant higher mean leaf trichome density than plants from populations II and III, which did not differ from each other (Fig. 3). Thus, differences found in the field were maintained in a common garden suggesting genetic differentiation between populations for this character. However, the same experiment revealed the plastic nature of trichome density: the novel environment represented by the common garden had distinctive effects on the plants of the different populations (i.e. they tended to converge phenotypically; cf. Fig. 3).
Table 5. Nested ANOVA of leaf trichome density (trichome × mm–2) for families of three populations of Datura stramonium in a common garden (see Materials and methods).
Significant among-population variation in both leaf trichome density and plant resistance to herbivores coupled with the association of trichome density with resistance in most populations of Datura stramonium, support the expectation of a defensive role of trichomes within populations. In addition, trichome density affected plant fitness through its association with plant resistance. However, the effectiveness of leaf trichome density varied among populations. Directional selection of phenotypes with higher resistance to herbivores was significant in only three populations of Datura stramonium. Thus, these results support the adaptive hypothesis of trichome density as a defensive trait against herbivory. Even though leaf trichome density is a phenotypically plastic character, our evidence indicated significant population differences in trichome density, highlighting the potential for genetic differences among populations on this defensive trait.
The relative effectiveness of trichome density as a defensive trait differed among populations (significant population × trichome density interaction; cf. Table 3a). In addition, the result that resistance may or may not be selectively advantageous in a given population is reflected in the interaction between population and resistance (Table 3c and Fig. 2). In fact, for three populations no evidence of selection on resistance was detected suggesting that natural levels of damage did not exert significant negative effects on individual plant fitness, and that other factors besides trichome density might determine resistance. Recently, it has been proposed that compensation after damage constitutes an alternative strategy of plant defence besides resistance (Maschinski & Whitham, 1989; Belsky et al., 1993; Simms & Triplett, 1994; Fineblum & Rausher, 1995; Mauricio et al., 1997). If some populations of D. stramonium compensate following damage, selection on resistance might not be expected (Herms & Mattson, 1992). A recent study in D. stramonium indicates that this species can compensate for foliar damage (Fornoni & Núñez-Farfán, 2000); however, it is not possible to establish, at present, if those populations do not have selection on resistance.
Although resistance could be the best strategy under certain environments, lack of genetic variation brought about by genetic drift (i.e, founder effects) will constrain selection. Similarly, phenotypic plasticity in defensive traits may limit selection despite the presence of genetic variation is present in certain environments (see Núñez-Farfán & Dirzo, 1994; Fornoni & Núñez-Farfán, 2000). Also, the capacity of plants to produce inducible defences (see Zangerl & Berenbaum, 1990; Underwood et al., 2000) once damage has occurred may prevent the detection of selection on resistance. We did not examine whether there are inducible defences in Datura stramonium, or if genetic variation for induction occurs in natural populations.
It must be stressed that population differences in leaf trichome density may occur even if it is not a component of plant resistance. For instance, trichome number might be positively or negatively selected in different stressful environments because it is correlated with other characters (e.g. leaf size; see Roy et al., 1999). However, if trichome density were not a resistance component in Datura stramonium, no relationship between trichomes and resistance would be expected either among or within populations. In this study, the results for Datura stramonium show that variation in trichome density is independent on leaf size (cf. Table 3b). Furthermore, no relationship between leaf size and trichome density was found for two populations (I and III) of this species in the greenhouse (P. L. Valverde, unpublished data). Leaf trichomes have been proposed to reduce water loss in water-limited environments (Turner & Kramer, 1980; Fitter & Hay, 1987). Still, this does not exclude the possibility that leaf trichomes function as a component of plant resistance to herbivores (Woodman & Fernandes, 1991). The present data support leaf trichome density as a component of resistance regardless of selection imposed by other environmental factors.
Response to selection within populations is expected only if part of the phenotypic variation in leaf trichome density is genetic in origin (Falconer & Mackay, 1996). Leaf trichome density is a highly variable plant character (Ågren & Schemske, 1994; Roy et al., 1999), and evidence of environmental induction (e.g. phenotypic plasticity) has been documented (Sharma & Dunn, 1969; Conklin, 1976; Wilkens et al., 1996; Elle et al., 1999). Several studies have detected heritable variation for leaf trichomes (van Dam & Hare, 1998a; Elle et al., 1999; van Dam et al., 1999). In Datura stramonium, the common garden experiment revealed the plastic nature of leaf trichome density since the population averages tended to converge (i.e. their change was in opposite directions; cf. Fig. 3). Yet, the populations analysed maintained their differences, suggesting genetic differentiation. The analysis did not reveal within-population differences among families and thus no potential for selection to change genetic frequencies at loci determining leaf trichomes. However, the common garden experiment involved only a small number of families per population. Thus, genetic variation for trichome density may exist in natural populations of Datura stramonium but a quantitative genetics study of this character in natural conditions is needed.
Selection on traits involved in plant–animal interactions is not expected to act in the same magnitude and direction across populations of a species. Due to the relevance of the environment in modulating genetic variation, the study of natural variation is important to estimate selection in characters of putative adaptive value and in guiding experiments aimed to establish causality (Mousseau, 2000; Sinervo, 2000). As the present results show, the analysis of variation in defensive traits in only one population might result in misleading conclusions when evolutionary inferences are made above the level of populations (Thompson, 1994, 1999).
We are very grateful to Phyllis D. Coley, Rodolfo Dirzo, Luis Eguiarte, Miguel Franco, José G. García-Franco, Carl D. Schlichting, and Stephen Weller for valuable comments and suggestions to the manuscript. R. Torres, J. Vargas and J. Zamudio offered valuable technical support during the field and laboratory work. This study is part of the Doctoral dissertation of P.L.V. who is grateful to the Department of Biology, Universidad Autónoma Metropolitana-Iztapalapa and CONACyT for the scholarship granted. J.F. thanks the Secretaría de Intercambio Académico, UNAM, and to the UNC, Argentina, for the scholarship for graduate studies. This study was supported by project CONACyT no.25662-N granted to J.N-F. Financial support for a sabbatical year (J.N-F.) at the University of Connecticut, Storrs, provided by CONACyT and DGAPA, UNAM, is greatly appreciated.