FITNESS IMPACTS OF HERBIVORY THROUGH INDIRECT EFFECTS ON PLANT-POLLINATOR INTERACTIONS IN OENOTHERA MACROCARPA.
University of Missouri, 8001 Natural Bridge Road, St. Louis, Missouri 63121-4499 USA
Abstract. The negative impacts of herbivores on plant fitness may include both direct and indirect effects. Direct effects on female plant fitness occur when decreased seed production is due to decreased resource availability from loss of leaf area and its attendant photosynthesis or through consumption of reproductive structures. Indirect effects occur when folivore- and florivore-mediated changes in floral traits influence pollinator preference and/or pollinator efficiency, thus reducing pollen receipt. We examined the effects of both leaf and floral herbivory (bud damage) and determined the relative contribution of direct and indirect effects of damage on female fitness through changes in floral traits for Oenothera macrocarpa (Onagraceae). The experiment was a two-factorial design that manipulated both leaf damage ([tilde]25% leaf area removed by hand) and pollen receipt (supplemental hand pollination). We then measured effects of leaf damage on floral traits (corolla diameter, floral tube length, and fl ower number) and fruit and seed set. Because herbivores damaged corollas in the bud stage in this experiment, we were also able to determine the effect of florivores on subsequent female reproduction.
Plants in the increased leaf damage treatment (33.4% leaf area loss) produced fewer flowers than natural leaf damage plants (6.5% leaf area loss), and their flowers had smaller corolla diameters and shorter floral tube lengths. Experimentally damaged plants also had 18% lower fruit set and produced 33% fewer seeds compared with natural leaf damage plants. Hand pollination increased fruit set by 60% and total seed number by 38% above that of naturally pollinated plants, demonstrating that female fitness was pollen-limited. Bud damage significantly decreased corolla diameter and floral tube length, and it led to a 68% reduction in fruit set. Hawk moths were observed preferentially visiting flowers with larger corollas, and study flowers with larger corollas had significantly increased probability of setting fruit. Additionally, seed number per fruit was positively related to floral tube length. In path analysis, we found no significant path between percentage leaf damage and female reproduction, suggesting tha t herbivores did not reduce seed production directly through decreased resource availability. Instead, we found a significant indirect path from percentage leaf damage to seed number through corolla diameter. We conclude that decreased female reproduction was due to changes in floral traits from both leaf and floral bud damage, which affected hawk moth preference (flowers with smaller corollas received fewer visits) and hawk moth efficiency of pollen delivery (flowers with shorter floral tubes had fewer seeds). This study demonstrates that herbivores can mitigate the mutualistic relationship between plants and pollinators through their effects on floral traits.
Key words: direct and indirect effects; floral herbivory; hawk moths; herbivory; leaf herbivory; Missouri Ozark glade communities; Oenothera macrocarpa; path analysis; plant-pollinator interactions; pollen limitation; pollinator efficiency; pollinator preference.
Both herbivores and pollinators have the potential to select for changes in plant traits. Herbivores do so by decreasing plant growth (Kinsman and Platt 1984, Marquis 1984, Strauss 1991), and both female (Stephenson 1982, Marquis 1984, Edwards 1985, Meyer and Root 1993, Wise and Sacchi 1996, Juenger and Bergelson 1997) and male reproduction (Allison 1990, Quesada et al. 1995, Mutikainen and Delph 1996, Strauss et al. 1996). The plant traits that mediate a plant's interactions with its herbivores are those that confer defense (i.e., leaf trichomes, secondary chemicals, and extra-floral nectaries) and tolerance (the ability to grow, survive, and reproduce in the presence of herbivore damage). In turn, pollinators influence plant traits that mediate pollen removal and receipt within and among plants. These traits include flower shape, size, color, scent, and reward (e.g., Schemske and Horvitz 1984, Nilsson 1988, Young and Stanton 1990, Herrera 1993b, Campbell et al. 1996, Conner et al. 1996, Galen 1996, Johnson and Steiner 1997). Perhaps because separate traits are involved, as well as different animal species (in most cases), plant herbivore ecology and pollination ecology traditionally have developed as separate disciplines.
The negative impacts of herbivores on plant fitness are likely to include both the direct effects of folivory on seed production through decreased resource availability and indirect effects on mating success. Most studies of herbivore impacts on plant fitness have implicitly assumed that decreases in seed production are due to decreases in resource availability for seed production resulting from loss of photosynthetic leaf area. However, because folivores and florivores can affect floral traits and flowering phenology, some of these presumed direct effects may be due to the indirect effects of herbivory on plant--pollinator interactions. To date little attention has been paid to these potential indirect effects of leaf and floral damage (but see Karban and Strauss 1993, Lohman et al. 1996, Strauss et al. 1996, Strauss 1997).
Results from studies that have measured effects of natural and experimental leaf damage on floral traits suggest the potential for indirect effects through delays in flowering time (Marquis 1988, Meyer and Root 1993, Frazee and Marquis 1994, Juenger and Bergelson 1997), reduced flower number (Karban and Strauss 1993, Quesada et al. 1995, Juenger and Bergelson 1997), decreased flower size (Michaud 1991, Frazee and Marquis 1994, Strauss et al. 1996, Strauss 1997), and a reduction in quality or quantity of pollinator reward (Quesada et al. 1995, Mutikianen and Delph 1996, Strauss et al. 1996). Similarly, floral herbivores can indirectly affect plant fitness by changing floral morphology (petal or inflorescence damage), and in so doing influence the probability of pollination (Karban and Strauss 1993, Cunningham 1995, Lohman et al. 1996).
These indirect effects of both floral and foliar herbivory come through changes in pollinator behavior, and can be manifested in two ways: (1) changes in pollinator preference (i.e., whether a flower receives a visit) and (2) changes in pollinator efficiency (i.e., during a visit, how much pollen is transferred). First, decreased attractiveness of the flower to pollinators may decrease the number of visits. Several studies have shown that pollinator visitation is affected by variation in specific plant and floral traits, including inflorescence height and corolla size (Campbell 1989, Young and Stanton 1990, Johnston 1991, Johnson et al. 1995, Strauss et al. 1996). Second, during a visit, changes in floral characters may decrease effective pollen receipt and/or removal (pollinator efficiency). Here, traits such as floral tube length may determine whether the pollinator contacts sexual parts effectively (Nilsson 1988, Johnson and Steiner 1997), and variation in pollinator rewards (nectar and/or pollen) may inf luence the duration of each visit (Heinrich and Raven 1972, Galen and Plowright 1984). One study has linked herbivore-induced changes in floral morphology to pollinator preference and efficiency: in Raphanus raphanistrum, leaf damage by Pieris rapae larvae decreased flower size and pollen and nectar production, and in so doing, reduced the number of visits to the plant and the time spent on the plant during visitation (Strauss et al. 1996).
The potential for herbivores to indirectly influence female fitness depends on seed production being pollen-limited at some stage of reproduction. If pollen is not limiting, then additional pollen delivery is not likely to influence seed production. For leaf and floral herbivores to influence female reproduction indirectly via pollinators, three conditions should be met: (1) some aspect of reproduction must be pollen-limited, (2) damage must affect floral traits, and (3) pollinators must discriminate against flowers on damaged plants (pollinator preference), or be less effective pollinators during visitation (pollinator efficiency) to those flowers.
We investigated the reproductive consequences of herbivore damage on female (and not male) fitness in the hawk moth pollinated species Oenothera macrocarpa (Onagraceae) by manipulating both the level of leaf damage and the amount of pollen to the plant. Additionally, the existence of natural floral herbivory (bud damage) allowed us to examine the indirect effects of leaf and floral damage through their impact on pollinator performance and pollinator efficiency. We are not aware of previous studies that have manipulated leaf damage and pollen receipt and measured changes in floral traits to determine the relative contribution of direct and indirect effects of damage to female plant fitness. Specifically, this study addressed the following questions: (1) Is plant female fitness (fruit and seed number) limited by pollen availability in O. macrocarpa? (2) What are the effects of increased leaf damage and natural florivory on O. macrocarpa floral traits and plant female fitness? (3) How does variation in floral t raits affect pollinator preference and efficiency and plant female fitness?
Oenothera macrocarpa (Onagraceae) is a native, perennial herb in Missouri Ozark glade communities. Glades are treeless and rocky barrens, characterized by thin soil cover (Nelson and Ladd 1980). Oenothera macrocarpa is a low, sprawling plant with a large taproot. Leaves are entire, linear in shape, and are produced continuously throughout the growing season. Plants bloom from late May through mid-June (in 1996, 21 May-22 June at the study site), and go dormant in July after seed ripening. Hermaphroditic flowers develop from leaf axils and have eight stamens (with pollen connected by viscin threads), and a cross-shaped stigma (Eisendrath 1978). The large yellow flowers (7-10 cm corolla diameter) open at night, are self-incompatible (Crowe 1955), and last until the following morning. Four sepals are fused to form a long slender floral tube (8-15 cm) with an inferior ovary. Each plant typically produces only one flower per night (see Plate 1), with a range of 1-25 flowers per plant for the entire flowering seas on (mean in 1996 was 3.9 flowers per plant). The fruit is a four-winged capsule typically 6 cm in length.
Flowers are pollinated solely by night-flying hawk moths (Sphingidae) which collect nectar from the base of a long floral tube with their long tongues. Previous studies of hawk moth pollination suggest this system is ideal for a study of indirect effects because female reproduction has the tendency to be pollen-limited (Haber and Frankie 1989, Herrera 1993a, Willmot and Burquez 1996) and hawk moths discriminate among floral characters (Miller 1981, Hererra 1993b). Furthermore, because O. macrocarpa plants generally produce only one flower per night, effects of herbivory on pollination can be tested by examining individual floral traits without the potentially confounding effects of flower number.
Research was conducted from May to August of 1996 at a restored glade of the Shaw Arboretum (of the Missouri Botanical Garden) in Gray Summit, Missouri, located 35 miles southwest of St. Louis, Missouri, USA. This dolomite glade has recently been restored (1996) through cutting of Juniperus virginiana and prescribed burns. Oenothera macrocarpa grows primarily on the south-facing slope of this glade. The glade covers [tilde]2.5 ha, while the south-facing slope is [tilde]40 m x 35 m. Only two species of hawk moths (Dolba hylceus (Dru.) and Paratraea plebeja (F.)) were observed visiting flowers of O. macrocarpa at the study site. Leaf herbivores seen feeding on O. macrocarpa at the study site included unidentified scarab and chrysomelid beetles, and grasshoppers.
In this experiment, we manipulated leaf damage and pollen receipt, and subsequently measured changes in floral traits and plant female fitness. Our goal was to distinguish between the contribution of direct and indirect effects of herbivores on O. macrocarpa. In May 1996, 160 plants were tagged along eight 20-m transects on the south-facing slope of the dolomite glade (20 plants per transect). Transects ran perpendicular to the slope; distance between transects was three meters. Every 4 m along a transect, four tagged plants on the north side of the transect were randomly assigned to one of four treatments: control (natural pollination and natural leaf damage); leaf damage (natural pollination and natural leaf damage + 25% experimental leaf damage); hand pollination (hand pollination and natural leaf damage); hand + damage (hand pollination and natural leaf damage + 25% experimental leaf damage). Each treatment had 40 plants.
Plants of all treatments received natural leaf damage (on average 9% of leaf area for the entire season, measured in July). However, plants in leaf damage and hand + damage treatments (hereafter = increased leaf damage treatment), in addition to natural leaf damage, had [tilde]25% area removed on each leaf (avoiding the primary vein) using a paper punch to mimic natural insect damage. Application of leaf damage was staggered: during 14-16 May, 25% of every other leaf was damaged on plants in the increased leaf damage treatments, and the following week, 21-23 May, the remaining leaves received 25% leaf damage. Late season new leaf growth was damaged 25% on 21 June. Natural and increased leaf damage both occurred after early spring floral bud formation and continued throughout the growth season. Twenty-five percent leaf damage was chosen to maximize the likelihood of treatment effects, by increasing leaf damage levels above natural leaf damage levels, but to fall within the natural range of leaf damage based on plant observations prior to the 1996 field season.
Plants in treatments hand pollination and hand + damage (hereafter = hand pollination treatment) received nightly supplemental hand pollination to all flowers during the flowering season. Hand pollination was done by clipping the anthers of 2-3 male donors from flowers of nonstudy plants (2-3 m from recipient plants) and then rubbing the anthers with pollen grains across the stigmas of recipient flowers.
Measurement and analysis of plant leaf area and leaf damage
Total plant leaf area was measured on all study plants three times: at beginning of the flowering season (20-22 May 1996) (initial leaf area), at the end of the flowering season (14-19 June), and at the end of the growing season (16-19 July). At each of these censuses total leaf number was counted and the length and width of every fifth leaf (minimum of five leaves per plant) was measured (to the nearest mm) for each plant. Leaf area missing (natural and experimental) was also measured on every fifth leaf (same leaf as growth measurement) during the June and July censuses using a clear plastic grid (square equal to 0.25 [cm.sup.2]).
Total potential area of each leaf was estimated by regression analysis (potential leaf area = 0.528 + 0.457 (L X W) [[R.sup.2] = 0.947, P [less than] 0.001, N = 223]). Potential leaf area per plant was estimated for each census by multiplying the mean leaf area by the number of leaves present at that census. Percentage leaf damage was calculated as the sum of leaf area missing/potential leaf area for that census (June or July). Percentage leaf damage was transformed by using arcsine(square root[percentage damage + 1]) to improve normality (Sokal and Rohlf 1995). Treatments did not differ in initial plant size (=potential leaf area in May; ANOVA, [F.sub.3,141] = 0.18, P = 0.90).
Measurement and analysis of floral traits
We measured corolla diameter and floral tube length of all flowers produced by study plants. Corolla diameter was estimated as the mean of two measurements across the flower from petal tip to petal tip (to the nearest mm). Measurements were made nightly, starting [tilde]1 h after dusk when flowers were fully open. Floral tube length (to the base of the sepal lobes) was measured on individual flowers the following day (before wilting) to avoid damaging flowers. A colored embroidery thread was then tied around the base of each ovary to follow the fate of individual flowers.
To examine whether increased leaf damage and/or flower order (successive seasonal flowering within a plant) affected floral traits (corolla diameter, floral tube length), flowers from natural and increased leaf damage plants were compared using analysis of covariance (ANCOVA). The response variable was the individual floral trait and flower order was the covariate. Flower order was used as a covariate to separate potential effects of a plant's seasonal change in flower size from leaf damage effects on flower size. A significant interaction term (leaf damage X flower order) indicates a difference in the slope of the relationship between floral traits and order for the two levels of leaf damage. The median test (Sokal and Rohlf 1995) was used to determine whether median flowering phenology differed for plants in natural and increased leaf damage treatments, by comparing treatment proportions to each side of the overall median flowering date (28 May for 1996). The effect of increased leaf damage and hand pollin ation on mean number of flowers per plant was analyzed by two-way ANOVA. The relationship between initial leaf area and corolla diameter was analyzed by linear regression.
Measurement and analysis of female plant fitness
Flowers in which no ovules developed were classified as uninitiated fruit. Flowers that became fully formed fruits and had at least one seed developed inside were classified as mature fruits. Mature fruits were collected from plants prior to seed dispersal (29 June-18 July) and number of seeds, undeveloped ovules, and aborted seeds were counted (sum of these variables = total ovule number per fruit). Seed mass was calculated as total seed mass (mg) per fruit per number of seeds per fruit. Plant means were then calculated for analysis for total ovule number, aborted seeds, and seed mass. Fruit set per plant was calculated as the ratio of mature fruit/total number of flowers. Total seed number per plant equals the sum of seeds over all mature fruits. Fifteen plants did not flower during the experiment and therefore were not included in any analyses. The distribution of these nonflowering plants was independent of treatment (df = 3, G = 2.0, P [greater than] 0.5).
Fruit set was transformed [arcsine(square root[fruit set + 1])] to meet assumptions of normality. To determine if mature and uninitiated fruits differed in floral characters, logistic regression was performed with corolla diameter and floral tube length as explanatory variables and fruit as the response variable. The relationship between floral tube length and seed number per fruit was analyzed by linear-regression using one randomly chosen floral tube per naturally pollinated plant.
The effects of hand pollination and increased leaf damage (June census) on mean total ovule number, aborted seeds, seed mass, fruit set, and seed number were analyzed using two-way ANCOVA, with initial leaf area as a covariate and leaf damage and pollination as the main effects. The relationships between initial leaf area and number of flowers, initial leaf area and seeds per plant, and number of flowers and seeds per plant were each analyzed by linear regression.
Natural floral herbivory
Flower petals were naturally damaged both at the bud stage (whose impact was studied here; assessed after corolla opening) and following corolla opening (not studied) by scarab beetles and other insect herbivores (unidentified). To determine the effect of bud stage floral herbivory on subsequent fruit set, flowers were categorized as either damaged or not damaged and fruit was categorized as either matured or uninitiated for analysis. To determine if floral herbivory was randomly distributed among the treatments, data were analyzed by contingency table (2 X 4). Two-sample t tests were performed individually on the floral traits measured to compare flowers with and without natural floral herbivory. The probability of fruit maturation was compared between damaged and undamaged flowers in a 2 X 2 contingency table to determine if fruit development was independent of floral damage. Two-sample t tests were performed to compare total ovule number and seed set of damaged and undamaged flowers.
Hawk moth visitation was quantified on three separate nights, totaling 16 person hours of observation. Patches of at least 10 flowers, from naturally occurring plants on the glade (located outside of experimental transects), were observed by one person. Each flower was marked with a numbered flag prior to opening so that pollinators were presented with a range of flower sizes to visit, all from unmanipulated plants. Observations began [tilde]1 h after twilight and lasted for 1.5-2 h, during peak hawk moth activity. All flowers were measured at the end of the observation period for corolla diameter and floral tube length. Logistic regression was used to determine if floral traits explained whether a flower received a visit (response variable) from a hawk moth.
We used path analysis to determine whether the observed impact of experimental leaf damage on seed production was due to a direct effect on plant resources, or instead an indirect effect on floral traits and subsequent plant pollinator interactions. Path analysis was performed using PROC REG (SAS Institute 1991; see Schemske and Horvitz  and Mitchell ). Individual path analyses were performed for plants in the control (natural pollination + natural leaf damage) and leaf damage (natural pollination + 25% leaf damage) treatments. Corolla diameter was used in path analysis because this trait is potentially an important visual cue for pollinator attraction (Herrera 1993b, Johnson et al., 1995), whereas O. macrocarpa floral tubes are not visible to pollinators. Initial leaf area was the independent variable in the diagram. For each dependent plant variable (mean percentage damage at end of flowering [June census], corolla diameter, and seed number), standardized partial regression coefficients (=path coefficients) were calculated using the STB option in PROC REG (SAS Institute 1991). In path diagrams, one-headed arrows represent a causal effect of one variable on another and U represents unexplained causes. The variable U was calculated as the square root of (1 - [R.sup.2]).
Flowers with larger corolla diameters were more likely to receive a visit than those with smaller corolla diameter (logistic regression, P [less than] 0.05, Table 1). Floral tube length, however, did not significantly explain variation in flower visitation (logistic regression, P = 0.81, Table 1).
Plant leaf area and leaf damage
Mean initial leaf area ([plus or minus] 1 SE) across all plants was 131.8 [plus or minus] 7.3 [cm.sup.2] (with a range of 29.9-720 [cm.sup.2]). For natural leaf damage plants there was no relationship between initial leaf area and percent damage at the 14-19 June census (df = 72, F = 2.3, P = 0.14) or the 16-19 July census (df = 72, F = 0.04, P = 0.84). The mean [plus or minus] 1 SE percentage leaf damage (June) was 6.5% [plus or minus] 0.01 for natural leaf damage plants and 33.4% [plus or minus] 0.01 for increased leaf damage plants (ANOVA, [F.sub.1,143] = 1149, P [less than] 0.0001). The seasonal range of leaf damage to plants was 0-59% for natural leaf damage plants and 21-44% for increased leaf damage plants. Thus, experimental levels of leaf damage fell well within measured natural levels of damage.
Female plant fitness
Hand pollination significantly increased fruit set per plant (ANCOVA, P [less than] 0.0001, Table 2), while damage significantly decreased fruit set (ANCOVA, P [less than] 0.05, Table 2). Comparisons of least-square means (Fig. la) revealed the difference in fruit set between control and damage treatment was marginally significant (P = 0.063); hand pollination and hand + damage fruit set did not differ significantly (P = 0.142). Number of seeds per plant was significantly affected by pollination, damage, and initial leaf area, with hand pollination plants producing more seeds than natural pollination plants, and natural leaf damage plants producing more seeds than increased leaf damage plants (Table 2, Fig. 1b). Hand pollination did not affect the number of flowers per plant (ANOVA, [F.sub.1,144] = 1.04, P = 0.30). There were no significant treatment effects on total ovule number, aborted seeds, or seed mass per plant (ANCOVA, [F.sub.3,130] [less than] 2.9, P [greater than] 0.15).
Data from naturally pollinated study plants provide further information about plant-pollinator interactions in the system. For these plants, pollinators apparently visited with greater frequency flowers with larger corolla diameters, as corolla diameter was a significant positive predictor of fruit maturation in these plants (logistic regression, P [less than] 0.05, Table 1). In contrast, floral tube length did not significantly predict flowers that produced a mature fruit (logistic regression, P = 0.25, Table 1). However, there was a significant positive relationship between floral tube length and the mean number of seeds per fruit (df = 63, [R.sup.2] = 0.24, F 19.35, P [less than] 0.0001) in naturally pollinated flowers. This latter result suggests that once a flower is visited, pollinator efficiency increased with longer floral tube lengths.
The number of flowers per plant was positively related to initial leaf area (df = 144, [R.sup.2] = 0.45, F = 72.8, P [less than] 0.001). However, this relationship between plant size and flower number did not translate into an effect on plant fitness: neither naturally pollinated plants with greater initial leaf area (df = 64, F = 0.96, P = 0.33), nor greater flower number (df = 64, F = 0.36, P = 0.55) produced more seeds per plant. Additionally, there was no relationship between plant size (initial leaf area) and corolla diameter (df = 143, F = 0.15, P =0.70).
Effects of increased leaf damage on floral traits
Median flowering phenology per plant did not differ significantly between natural leaf damage and increased leaf damage treatments (median test, df 1, [X.sup.2] = 0.90, P = 0.34), or between natural pollination and hand pollination treatments (median test, df = 1, [X.sup.2] = 1.29, P = 0.26). However, mean number of flowers produced per plant was significantly lower in increased leaf damage plants compared with the natural leaf damage plants (ANOVA, [F.sub.1,144] = 12.69, P [less than]] 0.001). Mean [plus or minus] 1 SE number of flowers was 3.0 [plus or minus] 0.1 and 4.3 [plus or minus] 0.2 for increased leaf damage and natural leaf damage plants, respectively. Both leaf damage and flower order (successive flowering within a plant) affected floral traits throughout the flowering season (Table 3). Corolla diameter significantly decreased (ANCOVA, P [less than] 0.05) and floral tube significantly increased (ANCOVA, P [less than] 0.0001) as subsequent flower order increased (Table 3). Increased leaf damage significantly decreased mean corolla diameter and floral tube length compared with natural leaf damage plants (Table 3, Fig. 2). In summary, increased lea f damage reduced the total number of flowers produced, and decreased corolla diameter and floral tube length. However, increased leaf damage did not change flowering phenology.
Our hypothesized path diagram (Fig. 3A) shows the relationship between initial plant leaf area, percentage leaf damage at end of flowering time (June census), corolla diameter, and seed number. We included a direct effect of leaf damage on seed number (female fitness) and an indirect effect of damage on seed number through the path from corolla diameter to seed number. We also hypothesized that leaf area might have a potential direct effect on total plant seed number (i.e., larger plants would have more resources for seed production) and an indirect effect through the paths percentage leaf damage and corolla diameter (i.e., larger plants might attract more herbivores and/or produce larger flowers).
Results of path analysis of control plants (N = 34) are shown in Fig. 3B (only path coefficients and variables P [less than] 0.05 are shown). Percentage leaf damage was negatively correlated with corolla diameter (P = 0.01). However, the paths from leaf area to percentage leaf damage (P = 0.88) and seed number (P = 0.26), percentage leaf damage to seed number (P = 0.75), and corolla diameter to seed number (P = 0.34) were not significant.
Results of the path analysis of experimental leaf damage plants with natural pollination (N = 35) are shown in Fig. 3C. Percentage leaf damage reduced corolla diameter (P = 0.01), and there was a marginally significant path (P = 0.06) between corolla diameter and seed number. The paths from leaf area to percentage leaf damage (P = 0.80) and seed number (P = 0.70), and percentage leaf damage to seed number (P = 0.15) were not significant. Thus, the observed effect of leaf damage on seed production was due to its indirect effect on corolla diameter.
Effects of natural floral herbivory on floral traits and female fitness
Floral herbivory was randomly distributed throughout the four treatments ([X.sup.2] = 1.09, df = 3, P [greater than] 0.77). Floral herbivory was relatively rare, recorded in 9% of marked flowers (44 of 514 flowers). Floral herbivory, however, significantly affected floral traits and fruit maturation when it occurred. Corolla diameter (t test, df = 30.4, t = 3.30, P [less than] 0.005) and floral tube length (t test, df = 43.3, t = 2.92, P [less than] 0.01) were both significantly reduced in damaged flowers (Fig. 4). The probability that a flower would set fruit was also significantly reduced for damaged flowers ([X.sup.2] = 32.04, df = 1, P [less than] 0.0001), with a mean fruit set ([plus or minus] 1 SE) of 0.23 [plus or minus] 0.10 for damaged flowers and 0.70 [plus or minus] 0.05 for undamaged flowers. No difference was found in total ovule number (t test, df = 146, t = 0.65, P = 0.52) or seed set between fruits produced by damaged and undamaged flowers (t test, df = 143, t = 1.5, P = 0.13). Together, thes e results demonstrate that flowers damaged by herbivores in the bud stage were smaller in corolla diameter, shorter in floral tube length, and much less likely to set fruit.
As in other systems (e.g., Marquis 1992, Meyer and Root 1993, Cunningham 1995, Juenger and Bergelson 1997), herbivores reduced female fitness in Oenothera macrocarpa. In this study, natural bud damaged flowers had a 68% reduction in fruit set compared with undamaged flowers, and experimental leaf damaged plants had 33% fewer total seeds than natural leaf damaged plants. To determine whether these observed declines in female fitness were due to indirect effects of herbivores through pollination, three necessary conditions were examined: (1) pollen limitation, (2) damage impact on floral traits, and (3) pollinator discrimination against flowers of damaged plants. Pollen limitation for O. macrocarpa was tested by supplemental hand pollination, and resulted in a 60% increase in fruit set and a 38% increase in total seeds compared with natural pollination plants. Because this species is self-incompatible and no apomixis occurred (K. Mothershead and R. Marquis, unpublished data), naturally pollinated plants can on ly initiate a fruit after they receive a visit. Therefore, we conclude that pollen was indeed limiting female reproduction during this study year.
Effects of increased leaf damage on female plant fitness: indirect or direct?
Increased leaf damage decreased both fruit set and seed number per plant. Initially, leaf damage reduced both flower number and flower size (corolla diameter and floral tube length). We consider the reduction in flower number and floral traits to be an initial direct effect of leaf damage through decreased resource availability. However, evidence suggests that ultimate female fitness of O. macrocarpa plants declined following experimental leaf damage due to an indirect effect of damage on plant pollinator interactions rather than through the effects of resource loss. Specifically, pollen availability in natural pollination plants was limited to such a degree that plants producing more flowers did not produce more fruits and seeds. In contrast, pollinator observations demonstrated that hawk moths visited flowers with larger corollas, and that the probability of producing a fruit for natural pollination plants increased with corolla diameter. Thus, differential female fitness between plants of increased and na tural leaf damage treatments appears to be due to discrimination by hawk moth pollinators between flowers of those plants, based on corolla diameter.
Path analysis results support the idea that herbivory had an indirect effect on female plant fitness. For increased leaf damage plants, there was a significant negative path from percentage leaf damage to corolla diameter and a positive path from corolla diameter to seed number, but no significant direct path from damage to seed number. Additionally, there was no direct path from initial leaf area to total seeds, indicating that plant size was not a factor affecting corolla diameter or plant fitness. Because the range of leaf damage to natural damage plants overlapped that found in increased damage plants, we believe path results apply to plants in the natural population that experienced higher levels of leaf damage.
Reproductive output is often positively correlated with plant size. Such a relationship would be expected in 0. macrocarpa if direct effects of herbivores were occurring because plants more heavily attacked (smaller plants) would produce fewer flowers and have fewer total ovules per plant to mature into seeds. However, this correlation may not exist when reproduction is limited by pollinator availability. Juenger and Bergelson (1997) also found no fitness benefit of larger plant size. If pollinator availability varies from year to year in 0. macrocarpa, as it does in ipomopsis aggregata (Juenger and Bergelson 1997), then we predict that the impact of leaf herbivores on plant fitness would shift from being an indirect effect to a direct effect with increasing pollinator abundance. In low pollination years, hawk moths discriminate among flowers of different sizes. Damaged plants in those years have lower fitness because their flowers are smaller. In high pollinator years, most flowers may be pollinated, but da maged plants would produce fewer seeds because they produce fewer flowers. Therefore, while flower number has the potential to influence female fitness greatly, the effect of leaf herbivory on flower number observed in our study may be realized only when pollinators are abundant.
Qenothera macrocarpa flower buds were formed early in the spring before the leaf damage treatment. The fact that loss of leaf area in the spring before flowering caused plants to produce fewer and smaller flowers suggests that carbohydrate resources necessary for floral development come in part from the current season's photosynthesis in this species. However, this direct effect of leaf damage early in the season did not carry over to a direct effect on later female plant fitness (path diagram). This lack of direct effect was especially evident for plants from the hand pollination + damage treatment in which fruit set and seed production were still high. Why then might resource limitation occur early during flower production but not later during fruit and seed production? A possible explanation is that because fruits are large (4-8 cm length), winged, and green during development, the direct effects of damage are offset through photosynthesis by the fruits themselves (Bazzaz et al. 1979, Reekie and Bazzaz 19 87). In fact, the area of an average fruit represents 44% of the average plant's leaf area (K. Mothershead and R. Marquis, unpublished data). Therefore, indirect effects of leaf damage in the absence of direct effects may occur due to a combination of pollinator limitation, floral discrimination by hawk moth pollinators, and fruit morphology and size.
It is important to note that the magnitude of change in floral traits due to damage was not great: only a 5% reduction in corolla diameter and a 4% reduction in floral tube length occurred in increased leaf damaged compared with natural leaf damaged flowers. Observations of hawk moth visitation at nonexperimental plants showed that flowers with smaller corolla diameters (9% difference) had significantly reduced visitation. In addition, experimental flowers that matured a fruit (and therefore received a visit) had significantly larger corollas compared with corollas of flowers that did not initiate a fruit. Because plants are self-incompatible, this is consistent with our view that hawk moth pollinators discriminate between flowers that have very small differences in corolla size (pollinator preference). Furthermore, the positive relationship between floral tube length and seeds per fruit suggests that reductions in floral tube length, caused by leaf damage, may decrease efficiency of hawk moth visitation (po llinator efficiency). When pollinator tongues are longer than the floral tube (which is more likely for plants with greater leaf area lost to herbivores), hawk moths can reach the nectar reward without coming into complete contact with reproductive organs. Both Nilsson (1988) and Johnson and Steiner (1997) found that spur length was important for proper pollinia receipt in longtongued orchid pollinators. Thus, small changes in floral traits coupled with very efficient pollinators (one visit is enough) suggests that hawk moths can potentially exert a strong selective force on floral traits. Several other studies have documented the importance of small changes in flower size affecting pollination (Rodriguez-Robles et al. 1992, Johnson et al. 1995, Conner and Rush 1996, Strauss et al. 1996).
An alternate hypothesis to explain patterns of hawk moth visitation might be that hawk moths were actually responding to plant density, which in turn was correlated with flower size. In this case, a good microsite might have a higher density of healthier plants (with larger corollas). While density of plants was not measured during pollinator observations, we found there was no relationship between plant size and corolla diameter in experimental plants (healthier plants did not produce bigger flowers), suggesting that hawk moths were attracted to plants based on flower size rather than on overall density of the microsite.
This study did not measure male fitness of plants; however, other studies have shown that both pollen quantity (Frazee and Marquis 1994, Strauss et al. 1996) and pollen quality (Quesada et al. 1995, Mutikainen and Delph 1996) can be reduced by herbivore damage. In 0. macrocarpa (where pollen is not a reward), unstudied effects on male fitness may have occurred through: (1) a decrease in flower number that in turn reduced the total amount of pollen and its quality for siring seeds (direct effects) and (2) a reduction in both corolla diameter and floral tube length that potentially affected the amount of pollen removed by hawk moths (indirect effects).
Effects of natural floral herbivory on female plant fitness: indirect or direct?
The overall effect of floral damage at the bud stage was to alter floral development, reducing both corolla diameter and floral tube length. Unfortunately, we did not observe damaged flowers for pollination. Our evidence suggests, however, that floral damage affected pollinator preference, such that hawk moths avoided damaged flowers due to changes in visual cues (see also Karban and Strauss 1993, Lohman et al. 1996). The result was a drastically reduced fruit set (68%) by damaged versus undamaged flowers. The decreased fruit set due to floral herbivory was not a direct effect of herbivores, as no flowers were observed in which stamen or pistils were damaged and there was no reduction in ovule number or seed set per fruit in cases where damaged flowers actually set fruit. This suggests that on the rare occasion that a damaged flower received a visit, it could set seed.
In summary, herbivory is likely to affect plant fitness by constraining resources. However, because plants interact with both herbivores and pollinators, this resource constraint can have both direct and indirect effects on plants. The potential for indirect effects of herbivory occurs when female fitness is pollen-limited and pollinators are known to discriminate among floral characters. Thus, foliar and floral herbivores can mitigate the mutualistic relationship between plants and pollinators. As a result, predicting the long-term selective impact of these interactions will be difficult. In general, we are most likely to detect indirect effects in plants with specialist pollinators, such as in this study of O. macrocarpa, because of the tight relationship between visitation and reproduction. Even in a specialized system, however, variation in pollinator and herbivore abundance may result in variation in the relative contribution of direct and indirect effects. Thus, the evolution of plants traits in respon se to both herbivores and pollinators will likely depend on spatial and temporal variation in abundance of each with respect to the other. Future studies should continue to explore indirect effects of herbivore damage on pollination, considering both male and female components of fitness.
We gratefully acknowledge C. Kelly, V. Sork, J. Le Corff and, J. Lill for valuable suggestions on earlier drafts, and C. Hochwender, S. Marchini, J. Morisaki, K. Stowe, and E. Wold for helpful comments in the planning stages. Many thanks are extended to L. Jones, T. Kunza, J. Le Corff, J. Lill, N. Maldonado, J. Mothershead, L. Slyman, J. Trager, and P. Vaheb who all stumbled around in the dark to help with field research. We are especially thankful to J. Trager and the Shaw Arboretum staff for permission to study at this Missouri Botanical Garden glade. Financial support was provided by a 1996 Trans World Airlines Scholarship.
(1.) E-mail: email@example.com
Allison, T. D. 1990 The influence of deer browsing on the reproductive biology of Canada yew (Taxus canadensis). Oecologia 83:523-529.
Bazzaz, F. A., R. W. Carlson, and J. L. Harper. 1979. Contribution to reproductive effort by photosynthesis of flowers and fruits. Nature 279:554-555.
Campbell, D. R. 1989. Measurements of selection in a hermaphroditic plant: variation in male and female pollination success. Evolution 43:318-334.
Campbell, D. R., N. M. Waser, and M. V. Price. 1996. Mechanisms of hummingbird-mediated selection for flower width in Ipomopsis aggregata. Ecology 77:1463-1472.
Conner, J. K., and S. Rush. 1996. Effects of flower size and number on pollinator visitation to wild radish, Raphanus raphanistrum, Oecologia 105:509-516.
Conner, J. K., S. Rush, and P. Jennetten. 1996. Measurements of natural selection on floral traits in wild radish (Raphanus raphanistrum). I. Selection through lifetime female fitness. Evolution 50:1127-1136.
Crowe, L. C. 1955. The evolution of incompatibility in species of Oenothera. Heredity 9:293-05.
Cunningham, S. A. 1995. Ecological constraints on fruit initiation by Calyptrogyne ghiesbreghtiana (Arecaceae): floral herbivory, pollen availability, and visitation by pollinating bats. American Journal of Botany 82:1527-1536.
Edwards, J. 1985. Effects of herbivory by moose on flower and fruit production of Aralia nudicaulis. Journal of Ecology 73:861-868.
Eisendrath, E. R. 1978. Missouri wildflowers of the St. Louis area. Monographs in Systematic Botany, Missouri Botanical Garden. St. Louis, Missouri, USA.
Frazee, J. E., and R. J. Marquis. 1994. Environmental contribution to floral trait variation in Chamaecrista fasciculata (Fabaceae: Caesalpinoideae). American Journal of Botany 81:206-215.
Galen, C. 1996. Rates of floral evolution: adaptation to bumblebee pollination in an alpine wildflower, Polemonium viscosum. Evolution 50:120-125.
Galen, C., and R. C. Plowright. 1984. The effects of nectar level and flower development of pollen carry-over in inflorescences of fireweed (Epilobium angustifolium) (Onagraceae). Canadian Journal of Botany 63:488-491.
Haber, W. A., and G. W. Frankie, 1989. A tropical hawkmoth community: Costa Rican dry forest Sphingidae. Biotropica 21:155-172.
Heinrich, B., and P. Raven. 1972. Energetics and pollination ecology: the energetics of pollinators may have wide implications in floral biology and community ecology. Science 176:597-602.
Herrera, C. M. 1993a. Selection of floral morphology and environmental determinants of fecundity in a hawk moth-pollinated violet. Ecological Monographs 63:251-275.
Herrera, C. M. 1993b. Selection on complexity of corolla outline in a hawkmoth-pollinated violet. Evolutionary Trends in Plants 7:9-13.
Johnston, M. O. 1991. Natural selection on floral traits in two species of Lobelia with different pollinators. Evolution 45:1468-1479.
Johnson, S. G., L. F. Delph, and C. L. Elderkin. 1995. The effect of petal-size manipulation on pollen removal, seed set, and insect-visitor behavior in Campanula americana. Oecologia 102:174-179.
Johnson, S. D., and K. E. Steiner. 1997. Long-tongued fly pollination and evolution of floral spur length in the Disa draconis complex (Orchidaceae). Evolution 51:45-53.
Juenger, T., and J. Bergelson. 1997. Pollen and resource limitation of compensation to herbivory in scarlet gilia, Ipomopsis aggregata. Ecology 78:1684-1695.
Karban, R., and S. Y. Strauss. 1993. Effects of herbivores on growth and reproduction of their perennial host, Erigeron glaucus. Ecology 74:39-46.
Kinsman, S., and W. J. Platt. 1984. The impact of a herbivore upon Mirabilis hirsuta, a fugitive prairie plant. Oecologia (Berlin) 65:2-6.
Lohman, D. J., A. R. Zangerl, and M. R.. Berenbaum. 1996. Impact of floral hcrbivory by parsnip webworm (Oecophoridae: Depressaria pastinacella Duponchel) on pollination and fitness of wild parsnip (Apiaceae: Pastinaca sativa L.). American Midland Naturalist 136:407-412.
Marquis, R. J. 1984. Leaf herbivores decrease fitness of a tropical plant. Science 226:537-539.
Marquis, R. J. 1988. Phenological variation in the neotropical understory shrub Piper arieianum: causes and consequences. Ecology 69:1552-1565.
Marquis, R. J. 1992. The selective impact of herbivores. Pages 301-325 in R. S. Fritz and E. L. Simms, editors. Plant resistance to herbivores and pathogens: ecology, evolution, and genetics. University of Chicago Press, Chicago, Illinois, USA.
Meyer, G. A., and R. B. Root. 1993. Effects of herbivorous insects and soil fertility on reproduction of goldenrod. Ecology 74:1117-1128.
Michaud, J. P. 1991. Biomass allocation in fireweed Epilobium angustiolium L. (Onagraceae) in response to simulated defoliation. Botanical Gazette 152:208-213.
Miller, R. B. 1981. Hawkmoths and the geographic patterns of floral variation in Aquilegia caerulea. Evolution 35:763-774.
Mitchell, R. J. 1993. Path analysis: pollination. Pages 211-231 in S. M. Scheiner and J. Gurevitch, editors. Design and analysis of ecological experiments. Chapman and Hall, New York, New York, USA.
Mutikainen, P., and L. D. Delph. 1996. Effects of herbivory on male reproductive success in plants. Oikos 75:353-358.
Nelson, P., and D. Ladd. 1980. Preliminary report on the identification, distribution, and classification of Missouri glades. Pages 59-69 in C. L. Kucera, editor. Proceedings of the Seventh North American Prairie Conference. Springfield, Missouri, USA.
Nilsson, L. A. 1988. The evolution of flowers with deep corolla tubes. Nature 334:147-149.
Quesada, M., K. Bollman, and A. G. Stephenson. 1995. Leaf damage decreases pollen production and hinders pollen performance in Cucurbita texana. Ecology 76:437-443.
Reekie, E. G., and F. A. Bazzaz 1987. Reproductive effort in plants: 1. Carbon allocation to reproduction. American Naturalist 129:876-896.
Rodriguez-Robles, J. A., E. J. Melendez, and J. D. Ackerman. 1992. Effects of display size, flowering phenology, and nectar availability on effective visitation frequency in Comparettia falcata (Orchidaceae). American Journal of Botany 79:1009-1017.
SAS Institute. 1991. SAS system for linear models. Third edition. SAS Institute, Cary North Carolina, USA.
Schemske, D. W., and C. C. Horvitz. 1984. Variation among floral visitors in pollination ability: a precondition for mutualism specialization. Science 255:519-521.
Schemske, D. W., and C. C. Horvitz. 1988. Plant-animal interactions and fruit production in a neotropical herb: a path analysis. Ecology 69:1128-1137.
Sokal, R. R., and F. J. Rohlf. 1995. Biometry: the principles and practice of statistics in biological research. State University of New York at Stony Brook, Stony Brook, New York, USA.
Stephenson, A. G. 1982. The role of extrafloral nectaries of Catalpa speciosa in limiting herbivory and increasing fruit production. Ecology 63:663-669.
Strauss, S. Y. 1991. Direct, indirect, and cumulative effects of three native herbivores on a shared host plant. Ecology 72:543-558.
Strauss, S. Y 1997. Floral characters link herbivores, pollinators, and plant fitness. Ecology 78:1640-1645.
Strauss, S. Y., J. K. Conner, and S. L. Rush. 1996. Foliar herbivory affects floral characters and plant attractiveness to pollinators: implications for male and female plant fitness. American Naturalist 147:1098-1107.
Willmot, A. P., and A. Burquez. 1996. The pollination of Merremia palmeri (Convolvulaceae): Can hawkmoths be trusted? American Journal of Botany 83:1050-1056.
Wise, M. J., and C. E Sacchi. 1996. Impact of two specialist insect herbivores on reproduction of horse nettle, Solanum carolinense. Oecologia 108:328-337.
Young, H. J., and M. L. Stanton. 1990. Influences of floral variation on pollen removal and seed production in wild radish. Ecology 7:536-547.
Logistic regression analysis of corrolla diameter (cm) and floral tube length (cm) for visited flowers (N = 18) and unvisited flowers (N = 48) based onobservations of nonexperimental plants, and for maturefruits (N = 136) anduninitiated fruits (N 84) from naturally pollinated experimental plants (df = 2 for each logisticregression). Means [plusor minus] 1 SE are shown.
Nonexperimental plants Floral trait Visit No visit Corolla diameter 9.7 [plus or minus] 0.3 8.7 [plus or minus] 0.2 Floral tube length 11.6 [plus or minus] 0.3 11.2 [plus or minus] 0.2 Floral trait Wald [X.sup.2] P value Mature Corolla diameter 4.79 0.026 8.5 [plus or minus] 0.1 Floral tube length 0.05 0.817 11.6 [plus or minus] 0.1 Experimental plants Floral trait Uninitiated Wald [X.sup.2] P value Corolla diameter 8.0 [plus or minus] 0.2 4.55 0.032 Floral tube length 11.3 [plus or minus] 0.1 1.31 0.252 Two-way ANCOVA of the effects of leaf damage and hand pollination on mean fruit set per plant and number of seeds per plant with initial leaf area as the covariate. Fruit Set Seed number Source df MS F MS F Damage 1 0.461 5.48 [*] 23 905 9.0 [**] Pollination 1 2.195 26.11 [***] 29 843 11.3 [**] Damage X pollination 1 0.004 0.05 256 0.1 Leaf area 1 0.098 1.17 170 532 64.2 [***] Error [++] 141 0.084 4332 (*.)P [less than] 0.05, (**.)P [less than] 0.005, (***.)P [less than]0.0001. (++.)Error df for seed number = 137. Univariate ANCOVAs for the effects of increased leaf damage on mean corolla diameter (cm) and floral tube length (cm) with flower order as the covariate. Corolla diameter Floral tube length Source df MS F P MS F P Damage 1 15.7 8.2 0.004 29.3 15.4 0.0001 Flower order 1 7.9 4.1 0.04 41.4 21.7 0.0001 Damage X order 1 3.1 1.6 0.21 8.4 4.4 0.034 Error 514
|Printer friendly Cite/link Email Feedback|
|Author:||MOTHERSHEAD, KRISTINE; MARQUIS, ROBERT J.|
|Date:||Jan 1, 2000|
|Previous Article:||FLOWER-TO-SEEDLING CONSEQUENCES OF DIFFERENT POLLINATION REGIMES IN AN INSECT-POLLINATED SHRUB.|
|Next Article:||DOES EARLY SEASON BROWSING INFLUENCE THE EFFECT OF SELF-POLLINATION IN SCARLET GILIA?|