Correspondence and present address: Department of Biology, Bucknell University, Lewisburg, PA 17837, USA (tel. +1 570 577 1142; fax +1 570 577 3537; e-mail firstname.lastname@example.org).
1The series of steps used to regulate a plant's investment in reproduction in response to environmental stresses is a life-history strategy critical to maximizing fitness. We investigated how the andromonoecious herb Solanum carolinense regulates its maternal investment in response to stress from foliar and floral herbivory.
2Most of the variation among S. carolinense individuals (ramets) in maternal investment occurred during the initiation of flower-bud primordia, with ramets initiating between 9 and 167 flower buds.
3In response to simulated floral herbivory, S. carolinense regulated maternal investment by decreasing the abortion rate of flower buds, increasing the ratio of perfect to male flowers and decreasing the rate of fruit abortion.
4In response to foliar herbivory, the plants increased the rate of fruit abortion and decreased allocation to perennial root growth.
5Some individuals specialized at regulating during early phenological stages, others specialized at later regulation, and there appeared to be trade-offs between these strategies.
6Because most plants must cope with multiple stresses that occur at different times in their phenology, such trade-offs suggest the presence of adaptive constraints. In particular, a plant's ability to tolerate damage by one species of herbivore may be constrained by the cost of lower tolerance of the damage caused by the other herbivores that feed on the plant.
Because plants are sessile organisms, they are at the mercy of the stresses of the environment in which they find themselves. The most successful plants will be those with sufficient flexibility, or plasticity, to cope with stress and acquire resources in an optimal manner. Of crucial importance are strategies by which they control maternal investment so that resources are not wasted (e.g. by making more flowers than can be matured into fruits), and so that opportunities are not missed (e.g. by making too few flowers to take advantage of a resource boon) (Lloyd et al. 1980; Stephenson 1981, 1984).
Following an explicit designation of the series of stages at which plants may hypothetically adjust their investment in maternal reproduction (Lloyd 1979, 1980), many empirical studies have been performed to identify the particular stages that are actually used by different plants (e.g. Lloyd et al. 1980; Morrison & Myerscough 1989; Stephenson 1992; Corbet 1998). The implication of much of this work is that plants tend to rely on a single, species-specific strategy to regulate maternal investment. However, every population of plants faces a wide variety of stresses, and the degree to which a plant species may employ different regulation strategies for different kinds of stresses is largely untested (but see Poveda et al. 2003). It is also possible that different individuals within a population specialize in regulation at different stages. If different types of stresses are best handled by controlling regulation of investment at different stages, then such specialization for a strategy to cope with one type of stress may limit a plant's ability to handle other stresses.
A stress with which virtually all plants have to cope is damage caused by herbivores. In fact, most plants are subjected to multiple species of herbivores, and a plant's response may differ depending on the phenology of the herbivores and the types of tissues they damage. For instance, the options available to respond to an early season herbivore (e.g. initiating more flower buds) may not be available when herbivory occurs later in the season, such as when fruits are maturing. In addition, the optimal response to a folivore, which removes a source of photoassimilates, is likely to be different from the optimal response to a herbivore that removes a sink of photoassimilates (e.g. flowers or fruits). Studying how plants respond to different herbivores will not only sharpen our understanding of the strategies by which maternal investment is regulated, but it will also provide insight into important evolutionary questions, such as whether the evolutionary response to selection for tolerance of one type of herbivore is likely to be constrained by correlated costs of lower tolerance of other herbivores (Tiffin & Rausher 1999; Mauricio 2000; Pilson 2000).
This paper reports an experiment in which we measured the regulation of maternal investment in the herb Solanum carolinense L. (Solanaceae) in response to leaf feeding (folivory) by Gargaphia solani Heidemann (Tingidae) and to simulated flower feeding (florivory) (mimicking Anthonomus nigrinus Boheman (Curculionidae)). Specifically, we addressed two main questions. (i) At which stages do the plants regulate maternal investment in response to folivory and florivory? A particular focus is whether herbivory can alter floral-sex ratio in this andromonoecious plant. (ii) Do all individuals use the same stage of regulation, or do they specialize at different stages? If the latter, we then asked whether such specialization suggests trade-offs between tolerance of the two kinds of herbivory.
Materials and methods
This study took place at the University of Virginia's Blandy Experimental Farm in Clarke County, Virginia, USA (78° W, 39° N; elevation of 190 m). Solanum carolinense, or horsenettle, is a perennial herb native to the south-eastern United States. A typical horsenettle stem (ramet) flowers for about 1 month between June and September. The flowers are borne on racemes that contain from one to more than 20 flowers (average c. 7), and the flower buds mature acropetally (from base to tip) within a raceme (Wise & Cummins 2002). Horsenettle flowers are self-incompatible and are generally pollinated by large-bodied bees. Horsenettle is weakly andromonoecious, with mostly hermaphroditic (perfect) flowers but a small, variable number of staminate (male) flowers, usually concentrated on the distal ends of the racemes (Solomon 1985; Whalen & Costich 1986). Often, one or more bud primordia on the ends of the racemes do not expand. These tiny buds, called ‘betsies’, dry up and fall off at a very small size, but their number can be inferred even after they are gone by counting the scars they leave on the racemes. The mature fruits of horsenettle are yellow berries that average c. 1.5 cm in diameter (Wise & Sacchi 1996; Wise 2003).
Gargaphia solani (the eggplant lace bug) is one of the most abundant leaf-feeding herbivores of horsenettle. A female lays eggs in a cluster on the underside of a leaf, and the nymphs feed communally until they disperse as adults. The lace bugs feed by piercing and sucking leaf parenchyma cells, which leads to yellowing and early abscission of leaves (Loeb 2003). Nymphal development is rapid, and the lace bugs may go through as many as eight generations per year in Virginia (Tallamy & Denno 1982). Horsenettle is also subjected to high levels of florivory. In a field experiment with 960 horsenettle ramets at this study site, just over half of the total flower crop was destroyed by herbivores (Wise 2003). The species that generally causes the most damage is Anthonomus nigrinus (the potato bud weevil). A female weevil will oviposit in a nearly mature flower bud, before immediately chewing through the bud's pedicel, severing the bud from the plant. The egg hatches and the larva completes its development within the unopened flower bud.
The plants used in this experiment were a subset of plants originally collected as roots from an old-field population at Blandy Farm in the spring of 1997. Roots were excavated from 24 newly emerging horsenettle ramets that were at least 2 m apart, and thus were likely to be from separate genetic individuals (i.e. genets). The roots were cut into segments weighing an average of 4 g and planted individually into 3.8-litre (1-gallon) pots in a commercial potting mix (WESCO growing media III‘; Wetsel Seed Company, Harrisonburg, Virginia). Ramets of each of these genets were used in an experiment on weevil herbivory in 1997: roots from these ramets were collected and refrigerated over the winter, and their new root growth served as the source of the plants for the current experiment.
In the spring of 1998 (29–30 May), roots from nine randomly chosen genets were washed, cut into segments averaging 9.0 ± 3.5 cm in length, weighed, planted individually into 3.8-litre pots in growing media, and allowed to grow for 1 month in a glasshouse. If multiple ramets emerged from a single pot, all but the first to emerge were removed by cutting them as close to the root as possible. At the end of June, 12 healthy ramets of each of the nine genets (108 pots total) were chosen for the experiment and were transplanted into 7.6-litre pots to which 44 cm tall tomato cages were attached. The pots were placed in two rows on wooden pallets outdoors in direct sunlight, and a fine-mesh bag was placed over each cage, attached to the rim of the pot with string, and closed at the top with a twist-tie to allow experimenter access but prevent insect escape or colonization. Three ramets of each genet were randomly assigned to each of four herbivory treatments: zero-herbivory control, simulated weevil florivory, lace bug folivory and both simulated weevil florivory and lace bug folivory.
Newly eclosed Gargaphia solani adults were collected from a horsenettle population near Blandy Farm. One male and one female lace bug were placed on each lace bug-treatment ramet on 1 July, and all the females had laid a clutch of eggs by 6 July. By the end of the month, adults were beginning to eclose, and the bags were periodically opened to allow dispersal from the experimental ramets. In the second week of August, the bags were removed on two successive days to ensure that all of the lace bugs could disperse. Each ramet received one or two additional clutches of eggs during this dispersal period. The bags were reattached to the pots and the second-generation lace bugs were allowed to complete their development. These bugs began eclosing to adulthood by the last week of August, and the bags were again opened periodically to allow their dispersal from the plants. From 31 August through to 4 September, each ramet was closely examined to remove all remaining bugs and prevent a third generation from developing.
Floral herbivory by Anthonomus nigrinus was simulated manually to ensure that each treatment ramet would receive the same proportional amount and spatial pattern of florivory. Each ramet was examined every 2–3 days, and buds were cut at their pedicels with bonsai scissors when they reached a size at which weevils generally oviposit in them. In each of the weevil-treatment ramets, the flower buds on the first (basal) half of each raceme were clipped. This pattern does not necessarily mimic the natural spatial pattern of weevil herbivory, but it allowed for standardization among the treatments and it ensured that the damage would be spread out both spatially and temporally, as it is in the field (M.J. Wise personal observation). This pattern also allowed more power in observing changes in floral sex because the distal flowers on racemes tend to be more variable in sex expression than the basal flowers (Solomon 1985; Whalen & Costich 1986; Elle 1998).
In general, simulated herbivory does not always trigger the same responses in plants as natural herbivory, particularly when feeding damage accrues slowly and when chemicals from the herbivores’ saliva induce a cascade of chemical responses in the plant tissue (Baldwin 1990; Hjältén 2004; Lehtilä & Boalt 2004). However, potato bud weevils clip the buds very rapidly, and the feeding by the larvae occurs within buds after they are detached from the plant. Thus, the method of bud clipping employed in this experiment is likely to be a fair representation of what happens to the plant when weevils actually do the clipping.
Because the mesh bags prevented pollinator access to the plants, the flowers had to be hand pollinated in order to set fruit. Horsenettle flowers remain open for about 3–4 days. To ensure that no flowers were missed, each ramet was pollinated every 2–3 days. Early each morning, fresh flowers were collected from a mixture of horsenettle genets in field populations at Blandy Farm. Using a mixture of genets for the pollen ensured that the majority of the pollen grains each flower received were of compatible genotypes (i.e. had different S-alleles) (Young & Young 1992; Richman et al. 1995). The anthers were vibrated with a battery-powered pollinator to shake the pollen into a glass vial. Pollen was then applied to the stigmas of the open flowers of each of the plants with a camel-hair paint brush. Pollen was applied regardless of the length of the pistil, but those with short pistils (i.e. the male flowers) never set fruit. The last flower opened and was pollinated on 29 August.
To keep track of all the flowers, each raceme was numbered with a small plastic twist-tie as it began to expand. The plants were examined every 2–3 days through the end of August to record the fate of the flower buds. By the end of October, the above-ground parts of the plants had senesced and each fruit had either ripened or aborted. From 11 to 16 November, the roots were excavated from the pots, rinsed, oven-dried at 60 °C for 48 hours (until achieving constant mass) and weighed. The following stages of regulation of maternal investment were thus available from this experiment: (i) number of flower buds initiated (including betsies); (ii) proportion of buds that expanded beyond the betsy stage (expanded buds); (iii) proportion of expanded buds that reached anthesis (successfully opened); (iv) proportion of open flowers that were perfect (completely developed gynoecium); and (v) proportion of perfect flowers that successfully developed into fruits. In addition, the root mass at the end of the season provided a measure of investment in potential future reproduction.
As a first step in elucidating which stages are important in the control of maternal investment in horsenettle, we calculated coefficients of variation (CV) for each stage. The higher the CV, the more influential a stage is in general in determining maternal investment differences among plants.
Analyses of covariance (ancova) were then run to determine how the ramets regulated maternal investment at each stage in response to stress from lace bug folivory and simulated weevil florivory. The proportional response measures were arcsine-square-root transformed to improve the normality of the distributions (except for the proportion of buds that expanded beyond the betsy stage, which was very nearly normal without a transformation). The explanatory factors in the ancovas were plant genet (which was treated as random), lace bug treatment, weevil treatment and all pairwise interactions. The three-way interaction did not approach significance in any of the ancovas, so it was dropped. Initial-root mass and length were included as potential covariates, but they were dropped if they did not explain a significant (P < 0.05) amount of the variation in the response variable. (Initial root mass was never a significant predictor.)
We then tested for differences among least-squares means (corrected for genet and initial-root-length variation) for all pairwise contrasts of herbivory treatments using Tukey's honestly significant difference method (Sokal & Rohlf 1995). All statistical analyses presented in this paper were performed using JMPIN® 4.0.4 software (SAS Institute 2001).
The number of flower-buds initiated was the most variable stage of regulation of maternal investment, with individual ramets initiating from 9 to 167 buds (CV = 49.2%; Table 1). The succeeding stages were less variable, with survival from open perfect flower through mature fruit being slightly more variable than the other stages. The mass allocated to root production was nearly as variable among ramets as flower-bud initiation. There was significant variation among the genets for all of the stages except for the proportion of expanded buds that reached anthesis, with clonal repeatabilities (% variation attributable to genet differences, Falconer 1989) for the other stages ranging from 0.42 to 0.50 (Table 1).
Table 1. Means and variations of ramets at various maternal-investment stages. The sample size was 107 ramets for percentage open flowers perfect and percentage fruits matured, and 108 ramets for all other measures. The minimum and maximum of the nine genet means for each stage, and the percentage of the variance that was attributable to genet differences (i.e. clonal repeatabilities) are also provided. Clonal repeatabilities shown in boldface indicate a significant genet effect at P < 0.0001
Maternal investment stage
Mean ± 1 SD
Coefficient of variation (%)
Range in genet means
67.8 ± 33.3
% buds expanded
74.2 ± 10.1
% expanded buds reaching anthesis
96.9 ± 7.4
% open flowers perfect
88.9 ± 13.7
% fruits matured
85.4 ± 14.6
Root mass (g)
17.2 ± 7.2
The ramets responded to simulated weevil florivory by upwardly regulating maternal investment at several stages. There was a statistically non-significant trend to increase the number of flower-buds initiated (Fig. 1) and then a strong increase in the proportion of buds expanded beyond the betsy stage (Fig. 2, Table 2). The ramets also responded to simulated weevil florivory by maturing the ovaries on a larger proportion of the uncut flower buds, resulting in an increase in the proportion of perfect flowers (from 86% perfect flowers on ramets without florivory to 92% on ramets with florivory). This increase in perfect flowers is remarkable because all the open flowers on the simulated weevil florivory ramets were on the distal half of the racemes, which tended to have a lower proportion of perfect flowers. Only 79% of the flowers on the distal half of the racemes of the ramets without florivory were perfect. Thus, while simulated weevil florivory increased the overall floral-sex ratio by 6%, it increased the likelihood of distal buds being perfect by 13%. Ramets also responded to simulated weevil florivory by reducing fruit-abortion rates by about 9% (Fig. 2, Table 2).
Table 2. F ratios from the ancovas for the maternal-investment stages, final-root mass and total-fruit production. Three-way interactions (Genet × Lace bug folivory × Simulated weevil florivory) were not significant and were dropped from the models. The covariate, initial-root length, was dropped when not significant (P > 0.05). Values in boldface are significant at P < 0.05. Other significance levels: * P < 0.01; ** P < 0.0005; *** P < 0.0001. LB = lace bug; SW = simulated weevil
Responses to lace bug folivory tended to be phenologically later and in the opposite direction to responses to simulated weevil florivory. In response to lace bugs, the ramets showed a non-significant trend towards decreasing the number of flower-buds initiated (Fig. 1, Table 2), but there were no changes in either the proportion of buds expanded beyond the betsy stage, the abortion rate of expanded buds or the sex ratio of opened flowers (Fig. 2, Table 2). The first large response was an increase in fruit-abortion rate by about 9%. The ramets exposed to lace bugs also allocated about 28% less mass to root production.
There were seven statistically significant correlations among the ramets in their values at the different maternal-investment regulation stages (Table 3). Four of these correlations were significant at an experiment-wise error rate of 0.05 using the Dunn-Šidák method of correcting for multiple tests (Sokal & Rohlf 1995). Note that correlations between early stages tended to be positive and correlations between late stages (including root mass) tended to be positive, but correlations between early and late stages tended to be negative.
Table 3. Product-moment correlations between values at each maternal-investment regulation stage. Values in boldface are significant at a pairwise error rate of 0.05, and underlined values are significant after correcting for multiple comparisons using the Dunn-Šidák method (Sokal & Rohlf 1995)
Maternal investment stage
% buds expanded
% expanded buds reaching anthesis
% open flowers perfect
% fruits matured
% buds expanded
% expanded buds reaching anthesis
% open flowers perfect
% fruits matured
regulation of maternal investment in response to herbivory
The investment-regulation strategies used in response to the two types of herbivory differed in both direction and timing. Maternal investment decreased following lace bug folivory, but increased following simulated florivory. The differences in direction can be explained as strategies to maintain or regenerate a source-sink balance. Because lace bugs decreased photosynthetic area by sucking out the contents of leaf parenchyma cells and by causing earlier leaf abscission, damaged plants would be more likely to be source limited. They therefore responded by decreasing allocation to sinks, for example by increasing the abortion of ovaries during fruit maturation and decreasing allocation to root growth.
In contrast, simulated weevil florivory decreased the number of sinks available, and the plant responses were all in the direction of increasing the number of ovaries that matured into fruits. For instance, the plants responded to flower-bud clipping by increasing the proportion of bud primordia that expanded, increasing the proportion of flowers with functional pistils and decreasing fruit-abortion rate. These responses enabled substantial compensation: although half the buds were cut off, the number of fruits the plants matured only decreased by one-quarter.
Although the lace bug and weevil treatments occurred over the same 2-month period, the responses to folivory (no effect until the fruit-maturation stage) were phenologically delayed relative to simulated florivory, where effects started at bud-primordia expansion. The more immediate response to flower loss is likely to be due to direct interactions among the buds within a raceme. If the basal flower buds on a raceme are fertilized and their ovaries begin to swell into fruits, the more distal buds may suffer from competition for resources, or their development may be suppressed by hormones produced by the developing fruits (Whalen & Costich 1986; Solomon 1988; Medrano et al. 2000). When some of the flower buds are lost to herbivory, then the other buds will experience reduced competition and suppression, and they are thus more likely to expand past the betsy stage, their pistils are more likely to develop, and the fruits are less likely to abort.
This degree of fine-tuning of maternal investment did not happen when the plants were exposed to lace bugs. For simulated weevil florivory, both the stress and responses involved the reproductive structures. In contrast, the stress from lace bugs involved the leaves, and the greater physical distance between stress and response would tend to make immediate regulatory responses more difficult. The greatest stress from lace bug herbivory was probably not felt until the end of the growing season, when the demand for photosynthates was greatest, i.e. for the filling of fruits and the expansion of perennial roots.
As an andromonoecious plant, horsenettle allows convenient testing of sex-allocation hypotheses because changes in allocation can be readily observed in the phenotype of floral sex (Lloyd & Bawa 1984). A basic premise of sex-allocation theory is that reproduction through the maternal route (fruit and seed production) is more costly than through the paternal route (pollen production) (Lloyd 1979; Charnov 1982). It follows that relatively stressed plants will tend to allocate a greater proportion of their reproductive effort to male function (Freeman et al. 1980; Korpelainen 1998). This prediction has been borne out in numerous plant species responding to such stresses as low levels of nutrients, light and water (Heslop-Harrison 1957; Freeman et al. 1980; Solomon 1985; Korpelainen 1998). While herbivory might also be expected to cause plants to concentrate on male function, empirical evidence of such a response is very limited (Spears & May 1988).
The hypothesis of greater allocation to male function in response to stress was not supported by the response to either type of herbivory. Because the plants exposed to lace bugs were probably not particularly photosynthate-stressed until after most of the flowers had opened, it is not surprising that lace bug damage had no discernible effect on floral-sex ratio. In contrast, while simulated weevil florivory did affect floral-sex ratio, it did so in a direction of opposite to that predicted. Similar increases in the proportion of perfect flowers (i.e. increased female allocation) have also been observed in a few other studies of florivory in andromonoecious plants (Hendrix & Trapp 1981; Hendrix 1984; Krupnick & Weis 1998). Unlike most stresses, florivory does not reduce resource availability or acquisition, but removes reproductive sinks, and the change in production from male to perfect flowers can serve as a mechanism to compensate for this loss (Hendrix 1979; Krupnick & Weis 1999).
It should be emphasized that the increase in the ratio of perfect:male flowers in horsenettle was not an artifact of the method of florivory simulation. The method of cutting the basal buds actually worked against finding an increase in the proportion of perfect flowers in response to clipping because the distal buds are generally more likely to produce male flowers than the basal buds (Solomon 1985; Elle 1998). If the buds were clipped randomly with respect to position, then it is likely that there would have been an even greater perfect:male flower ratio in the simulated weevil florivory treatment.
specialization and trade-offs
The primary control of maternal investment in horsenettle occurred at the initiation of flower-bud primordia, as this was the stage with the most variability among the plants. This stage is extremely important because it sets the absolute limit on the number of fruits a plant can mature in a growing season. The root mass at the end of the season was similarly variable among plants, indicating that allocation to roots is also an important means of regulating investment, as it determines both how far a horsenettle genet can spread in an established field and the potential for sexual reproduction in the succeeding year.
While both early and late regulation stages are important for horsenettle, there were some indications that individual plants tended to specialize in either one or the other. For instance, there were positive correlations among the earliest stages of control (initiation of flower-bud primordia, expansion of bud, and anthesis) and between the last stage (fruit maturation) and root mass at the end of the growing season. One might expect competition for resources between fruits and roots to lead to a negative correlation between fruit maturation and root mass. The results suggest instead that a relative resource surplus late in the season will lead to greater investment in fruits and roots, while a resource shortage will lead to less investment in both.
Despite these positive correlations between stages, there appeared to be trade-offs between investing in early vs. late stages. In particular, plants that invested heavily in initiating and expanding flower-bud primordia were not able to invest as much in perennial root growth, and vice versa. There was also a trade-off between the number of flower-buds initiated and the proportion of the flowers that were perfect. Such trade-offs may limit a plant's ability to regulate maternal investment optimally in response to different types of stresses. In particular, such trade-offs may be important constraints on the evolution of plant tolerance of herbivory in multiple-herbivore communities.
We have previously shown that the ability to initiate more flower-bud primordia and to mature a greater proportion of these buds in response to florivory are important compensation mechanisms for horsenettle. However, specific tolerance mechanisms for lace bug folivory have not yet been identified. Nevertheless, because lace bug damage stressed horsenettle relatively late in the season, and because the primary effect of this stress is reduced root growth, the ability to maintain root growth in spite of the lace bug damage is likely to be an important tolerance mechanism. If so, then the negative correlations between early and late maternal investments are likely to act as constraints on simultaneous tolerance of weevil florivory and lace bug folivory. Like horsenettle, most plant species are damaged by multiple species of herbivores, feeding at different times and on different organs. Thus, trade-offs in maternal investment strategies may constitute a widespread constraint on the evolution of plant tolerance of herbivory in nature.
We thank M.L.G. Loeb for collecting and sorting the lace bugs used in this experiment, and J. Peachey Schaefer and H.F. Sahli for assistance with root washing. We are also grateful to W.G. Abrahamson and C.P. Blair for constructive comments on the manuscript. This work was supported financially by undergraduate and graduate student summer research fellowships to J.J.C. and M.J.W., respectively, from the University of Virginia's Blandy Experimental Farm. M.J.W. was also supported through the David Burpee Endowment of Bucknell University and a National Science Foundation Grant (DEB-0515483) to W.G. Abrahamson and M.J.W. during the writing of the manuscript.