Reproductive consequences of fl ower damage in two contrasting habitats : The case of Viola portalesia ( Violaceae ) in Chile

The indirect impact of fl ower herbivory on plant reproduction depends on the pollination environment, particularly on the presence or absence of pollinator species with the ability to discriminate damaged from undamaged fl owers. The change in pollinator assemblages, due to habitat modifi cation, may modify the impact of fl ower herbivory on plant reproductive success. In this work, we evaluate the effect of fl ower herbivory on the seed production of Viola portalesia (Gay) in two contrasting environments, a native and low-disturbed habitat and an extensively transformed habitat characterized by Pinus radiata plantations. Even though the two habitats dif fered substantially in the composition of pollinator assemblages and visitation rate, the fl ower damage performed on different petals had no impact on seed production neither within nor between habitats, indicating that change in pollinator assemblages have no indirect reproductive impact via discrimination of damaged fl owers. There was a strong habitat effect, however, for seed production, being higher in the pine plantation than in the native habitat. Seed production on a per fl ower basis increased at a higher rate with the number of fl owers in the pine plantation, which suggests a numerical response of pollinators to changes in food availability in this habitat.


INTRODUCTION
The impact of flower herbivor y on plant reproductive success may var y from strong negative ef fects (Mothershead & Marquis 2000) to neutral ef fects (Malo et al. 2001), depending on the pollination environment (McCall & Ir win 2006).In general, flower h e r b i v o r y m a y h a v e a d i r e c t i m p a c t on reproductive success by damaging the reproductive structure of fl owers and also an indirect ef fect on reproduction by altering the flower characters involved in pollinator attraction (e.g., Herrera et al. 2002, McCall & Irwin 2006, Pohl et al. 2006).For example, fl ower herbivory has been shown to infl uence the quality and quantity of diverse flower traits such as petal size (Mothershead & Marquis 2000), nectar production (Krupnick & Weis 1999), fl ower production (Lohman et al. 1996, McCall & Irwin 2006) and fl owering time (Frazee & Marquis 1994), all characters that influence pollinator attraction and the quality and quantity of pollination ser vice (Cunningham 1995, Krupnick & Weis 1999, Mothershead & Marquis 2000).The most obvious consequence of a reduction or loss of pollination service is the reduction in fruit set and seed set (e.g., Aguilar & Galetto 2004, Aguilar et al. 2006).Notwithstanding, pollinator species often differ in their ability to distinguish damaged from undamaged fl owers (Schaefer et al. 2004), which suggests the fi nal outcome of fl ower herbivory may be contingent to the composition of the pollinator assemblage (Pohl et al. 2006).One of the most conspicuous effects of habitat loss is the alteration in the composition and abundance of pollinator assemblages, as well as the behavioral change of par ticular pollinator species (e.g., Didham et al. 1996, Thomas et al. 2001, Goverde et al. 2002).Less obvious are effects on pollination service caused by habitat modifi cation that propagate indirectly to plant reproduction through changes in the response of pollinator assemblages to fl ower damage.If habitat change modifies the composition and abundance of pollinator assemblages it is likely that the effect of fl ower damage in reducing pollinator ser vice depends on the presence or absence of particular pollinator species sensitive to habitat alteration.While many studies have examined the ef fect of habitat change on the composition of pollinator assemblages (e.g., Rathcke and Jules 1993, Murcia 1996, Cunningham 2000, Aizen & Feinsinger 2004, Murúa et al. 2010), to our knowledge evidence for an indirect role of habitat change on fl ower herbivory and plant reproductive success is lacking.
The aim of this study is to examine the effect of fl ower damage on the reproductive success of Viola por talesia in contrasting habitats of the Maulino forest.We will also evaluate the impact of floral damage as a function of fl ower density, which will provide information on the potential density-dependent effects on plant fi tness.More specifi cally, we will attempt to answer the following questions: 1) Do native and pine plantation habitats differ in the composition of pollinator assemblages?2) Does fl ower damage affect seed production in native and pine plantation habitats?3) To what extent, variation in seed production can be ascribed to changes in the pollination environment between habitats?These questions relate to a more general, 4) Do differences exist in the indirect impact of fl ower damage on plant reproductive success between the native and modifi ed habitat?METHODS Viola portalesia (Violaceae) is an endemic selfincompatible herb species that distributes along the Costal Range in central Chile.Its fl owering season spans from September -November, after which a trivalve fruit is produced.Viola portalesia inhabits indistinctly natural and pine plantation habitats in the Maulino forest, a natural forest that has been extensively subject to human-induced alteration, mainly through deforestation and subsequent replacement of the native forest by plantations of Pinus radiata D. Don (Bustamante et al. 2003).This practice has produced a highly fragmented landscape of native habitat embedded in a matrix of Pinus that leads to an ample variation in the abundance and species richness of diverse functional groups (Grez et al. 2006).
Fieldwork was performed in two contrasting populations of Los Ruiles National Reserve and surrounding areas from October 2007 to January 2008, thereby covering the complete flowering season of V. portalesia.The fi rst population was located within the National Reserve (35°83' S; 72°50' W).The most representative species at the Reserve are Nothofagus alessandri, Nothofagus glauca, Ugni molinae, Oxalis articulata, Teline monspessulana, and Viola portalesia.The second population was embedded in a matrix of Pinus radiata located 3.6 km from the fi rst population (35°85' S; 72°46' W).The Pinus plantation was 20year old, and had an understory with a low herb and shrub species richness that included Azara integrifolia, Lithraea caustica, Ugni molinae and Viola portalesia as the most representative species.
We performed experimental damage on 63 plants in the National Reserve and 59 plants in the Pinus plantation.Only plants with at least four fl ower buds were chosen.As pollinators may respond to the fl ower number in the neighborhood of the focal plant of V. portalesia, we recorded the number of flowers in a circular area 2 m diameter around the focal plant.The number of fl owers per patch ranged between 1-110 in the native forest and 1-115 fl owers in the pine plantation habitat.We allocated 1-day flowers to one of four treatments per plant: 1) removal of the landing petal, 2) removal of the lower petals, 3) removal of the upper petals, and 4) control fl ower without damage (Fig. 1).The total area of the petals removed was kept as constant as possible in all treatments.In addition to experimental treatments, we estimated the frequency of natural fl ower herbivory by recording the number of fl owers damaged in 50 replicated plants per habitat, one randomly chosen fl ower per plant.
To ensure that potential response of pollinators is due to damage and not to a loss of symmetry, the removal of petals was performed on the left and right petals both in the lower and upper petal removal treatment, keeping constant the original symmetry of the fl ower.We checked for potential withering or additional damage of fl owers on a daily basis.Flowers with natural damage were not considered in subsequent analyses.After four weeks of the experimental damage we collected the capsules of 56 and 54 plants in the Reserve and Pinus plantation, respectively.All capsules were analyzed for seed counting in the laboratory.To test the importance of treatment and population on the seed production per fl ower, we performed one-way analysis of covariance using the number of flowers in the neighborhood as covariate.All data were logtransfomed to fulfi ll the assumptions of normality and homoscedasticity.Because all the four treatments were performed within replicated plants, data were also analyzed in the context of mixed models to remove the within-plant variance component.
The pollinator assemblage was recorded at each population by quantifying the pollinator species richness and visitation rate, measured as the number of visits per fl ower per hour using the plant focal method (Dafni et al. 2005).Visitation rate was estimated by recording the number of pollinator visits to all fl owers within patches during 15 minutes focal censuses (see also Murúa et al. 2010, Rivera-Hutinel et al 2012).We performed a total of 190 censuses (47.5 h of observation) in the native habitat and 108 censuses (27 h of observation) in the Pinus plantation All samplings were performed between 0900 and 1700 hours, always in sunny days.Only insects that contacted the reproductive structures of fl owers were considered as legitimate visitors.Visitation rate to V. portalesia was compared between populations by using a generalized linear model with Poisson distribution.All statistical analyses were performed in R package (2007).

RESULTS
The composition and species richness of pollinator assemblages dif fered between  List of species involved in the pollination process of Viola portalesia in native and pine plantation habitats.Mean visitation rate and standard error on a per species basis are depicted.† NF: not found in the habitat.
Visitation rate (Visits fl ower  1).Finally, the overall visitation rate taking all insects as a whole, regardless of its species identity, was also higher in the pine plantation than in the native habitat (mean ± SE, native = 0.27 ± 0.06; Pinus plantation = 0.44 ± 0.13; Generalized Linear Model, χ 2 1 = 4.27; P < 0.038), confi rming that fl owers of V. portalesia receive a higher visitation rate at the Pinus plantation than at the National Reserve.
Regarding fl ower herbivory, the frequency of natural fl oral damage was similar between populations.Eleven out of fi fty fl owers from dif ferent plants (22 %) presented flower herbivor y in the Reser ve and four teen out of fifty flowers (28 %) were herbivorized in the Pinus habitat (Chi-square test, χ 2 1 = 0.48; P = 0.488).It is likely that similarity in natural fl orivory levels between habitats does not convey dif ferences in seed production as suggested by experimental results.For instance, artifi cial damage treatments revealed that while habitat and number of flowers per patch had an impor tant ef fect on seed production, the experimental fl ower damage imposed on corollas had no signifi cant effect (Table 2, Fig. 2).Flowers of V. por talesia inhabiting the native forest produced 40 % less seeds than fl owers in the Pinus plantation (mean ± SE; 15.8 ± 0.24, n = 224 versus 22.1 ± 0.37, n = 216, respectively; P < 0.0001).The statistical interaction between habitat-type and fl ower number per patch was also signifi cant (P = 0.0006), indicating that habitats differed in seed production after keeping constant the number of fl owers per patch (Table 2, Fig. 3).Seed production increased at a higher rate in the pine plantation than in the native habitat (Fig. 3).The interaction term including habitat x treatment, and treatment x fl ower number was unimportant in accounting for variation in seed production (Table 2).This result did not change substantially after removal of withinplant variation in Mixed Linear Models (F 3,318 = 1.50;P = 0.214).TABLE 2 Summary of effects of ANCOVA of habitat (H) and fl ower treatment (T) impact on seed production in Viola portalesia using fl ower number per patch (F) as covariate.Degrees of freedom (df), mean squares (MS), F-values, and P-level are listed.
Resumen de los efectos de ANCOVA para impacto del hábitat (H) y tratamiento fl oral (T) sobre la producción de semillas en Viola portalesia usando el número de fl ores por parche (F) como covariable.Grados de libertad (df), cuadrados medios (MS), valores de F, y nivel de P son indicados.Producción de semillas en cada uno de los cuatro tratamientos de remoción de pétalos aplicados a Viola portalesia en los hábitats nativo y plantación de pino.Las cajas representan los dos quartiles alrededor de la mediana de los datos.La línea horizontal dentro de las cajas indica la mediana de la producción de semillas.Las barras representan los percentiles 5 y 95 de la distribución de los datos.Valores atípicos no son ilustrados para mayor claridad.

DISCUSSION
We did not fi nd signifi cant effects of fl ower damage on seed production among treatments within habitats, indicating that pollinators were largely insensitive to the location of flower damage.Three lines of reasoning may help to understand these results.One possibility is that the quality and quantity of damage that we applied to petals did not emulate natural herbivor y.Several studies have reported that insects often use fl ower symmetry and spatial pattern to guide their foraging activities and fl ower preferences (Bell 1985, Johnson et al. 1995, Moller & Eriksson 1995, Giurfa & Lehrer 2001).The location of damage may infl uence the impact of fl ower herbivory on reproductive success.For example, Pohl et al. (2006) reported that fl owers damaged on the landing petal in Mimulus luteus produced fewer seeds than fl owers damaged in other petals.Likewise, Sánchez-Lafuente ( 2007) reported that removal of lower petals alone or in combination with the removal of upper petals had a strong negative impact on pollinator visitation and seed production in Linaria lilacina.Because our treatments consisted on damage on all petals (landing petal, lower petals, and upper petals), keeping as constant as possible the flower symmetr y (Fig. 1), it is unlikely this factor accounts for the absence of effect of fl ower herbivory on seed production within habitats.Regardless of location, however, it is likely that the amount of damage imposed to fl owers, that is, the area removed with respect to the total corolla area, was not suffi cient for pollinators to reject damaged fl owers.Even though we removed about 50 % of the petal area at each treatment (25 % of each lower and upper petals), we can not assure this amount is suffi cient to elicit rejection of damaged fl owers by insect pollinators.
Second, in addition to the potential limitation of the experimental procedure, factors unrelated to fl ower herbivory may be important for fl ower discrimination by pollinators.For example, regardless of changes in flower morphology, pollinators may discriminate fl owers on the base of additional clues such as the level of fl oral reward (Krupnick et al. 1999), the presence of chemical signals (Galen et al. 1987, Ashman et al. 2005), and the presence of damaged fl owers in the neighborhood of focal plants (Krupnick et al. 1999).
Third, we measured the reproductive impact of flower damage on seed production only, assuming that female and male reproductive success are correlated variables.However, it is known that herbivor y may not only infl uence the female but also the male fi tness component (e.g., Mutikainen & Delph 1996, Krupnick & Weis 1999, Lehtila & Strauss 1999).Because we did not measure pollen export, pollen tube growth rates, or siring success, we cannot r ule out that flower herbivores reduced male rather than female fitness in this species.The absence of ef fect of a modifi ed corolla on seed production suggests that pollinators have a low impact as agents of natural selection in these species.For instance, complementar y information has revealed weak pollinator-mediated selection coef ficients on the flower phenotype of V. por talesia in the two populations (Murúa et al. 2010).These results together indicate that no clear function related with the pollination process can be ascribed to petals in this species.Similar results have been previously reported in other studies.For example, Herrera (1993) rejected the hypothesis of adaptation of corolla spur to hawk moth pollination in Viola cazorlensis (Violaceae).Similarly, Malo et al. (2001) obser ved that experimental corolla modifi cation had no effect on pollination and reproductive success in Myrmecophila tibicinis (Orchidaceae).Future studies need to explicitly address the ecological conditions under which fl oral damage is expected to have a negative impact on plant reproductive success.
Even though herbivor y treatments were largely irrelevant in accounting for variation in seed production, there was a strong habitat effect (Table 2).Seed production was lower in the native habitat than in the Pinus plantation, which raises the question on the factors that contribute to variation in seed production between habitats.There are several factors potentially relevant to explain this pattern.In principle, the higher visitation rate detected in the pine plantation might be suffi cient to account for the higher seed production in this habitat.However, only 13.3 % of pollinator species were shared between habitats, which suggests that visitation rate may not be necessarily comparable between native and pine habitats.Variation in the ef ficiency of pollinators between habitats has been suggested as an important factor to explain variation in seed production among populations (e.g., Prada et al. 1998, Sanchez-Lafuente et al. 1999).In our study, the pollinator species with the highest visitation rate dif fered between habitats (the bee Manuelia gayatina in the native habitat and the beetle Ar throbrachus nigromaculatus in the pine plantation, see Table 1).Even though the effi ciency of such species is unknown at present, it is likely that dif ferences in the composition of pollinator assemblages imply a completely dif ferent spectrum of species with var ying ef ficiency at pollen deposition.A second, non-mutually exclusive explanation bases on plant species richness and pollen interference.It is known that inter ference associated to increased foreign pollen deposition in rich-species plant communities may reduce successful fer tilization in comparison to poor-species plant communities (Waser & Fugate 1986, Murphy & Aarsen 1995, Totland 2001, Hegland & Totland 2008).The two habitats compared in this study dif fered substantially in plant species richness.While 63 fl owering species occur red in the natural Reser ve, only 17 fl owering species accompanied V. portalesia in the pine plantation.This notorious asymmetry suggests that seed production is more likely to be pollinator-limited in the native habitat.In this way, the reduction in fl owering plant diversity observed in the Pinus plantation may lead to a lower foreign pollen deposition in stigmas of V. portalesia.Third, the high metabolic rates of fl ying insects create a high demand for food, and the patterns of occurrence of pollinators tend to track the spatial and temporal patterns of food availability (Bronstein 1995).In doing so, numerical responses of pollinator species to flower number might be more prevalent in habitats with low food availability, that is, Resultados de ANCOVA del número de fl ores por parche sobre la producción de semillas por fl or por tipo de hábitat.Círculos llenos representan fl ores de V. portalesia en la plantación de pino y círculos abiertos representan fl ores en el hábitat nativo.Las líneas continuas y segmentadas representan el mejor ajuste lineal en los hábitats de Pinus radiata (pendiente ± EE: 3.01 ± 0.84, r 2 = 0.06, n = 214, P < 0.001) y nativo (0.41 ± 0.54, r 2 = 0.01, n = 224, P = 0.448), respectivamente.
in poor habitats where any increase in fl ower number should result in a higher visitation rate and seed production per fl ower.Indirect evidence for this hypothesis comes from the observation that seed production increased at a higher rate with fl ower number at the Pinus plantation than native habitat (Fig. 3).The extent to which the higher seed production obser ved in the pine plantation results as a consequence of a change in pollinator assemblages and species ef ficiency, pollen interference at stigmas, or from a numerical response of pollinator species in poor habitats needs to be assessed in future studies.
In conclusion, our results indicate that the strong dif ferences in the composition of pollinator assemblages and visitation rate between habitats do not translate into important reproductive effects through fl ower herbivor y.As pollinators do not discriminate against damaged flowers, they were largely inconsequential for indirect effects of fl ower herbivor y on plant reproduction.The higher seed production of V. portalesia obser ved in the transformed habitat cannot be ascribed to variation in flower herbivor y.Rather, factors unrelated to fl ower herbivory, such as variation in the composition and effi ciency of the pollination assemblage, variation in the accompanying fl ora, and numerical responses of pollinators to changes in food availability seem to be important to understand the higher seed production observed in the P. radiata plantation as compared to the native habitat.

Fig. 2 :
Fig. 2: Box and whisker plot for seed production in each of the four petal removal treatments applied to Viola portalesia in native and pine plantation habitats.Treatments as in Figure 1.Boxes represent the two quartiles around the median of data.Horizontal line within boxes indicates the median of seed production.Whiskers represent the 5 th and 95 th percentile of data distribution.Outliers are no depicted for clarity purposes.

Fig. 3 :
Fig. 3: Results of ANCOVA of fl ower number per patch on seed production per fl ower by habitat type.Filled dots are fl owers of V. portalesia in the pine plantation and open dots are fl owers in the native habitat.Continuous and dashed lines represent the best linear fi t in the pine plantation (slope ± SE: 3.01 ± 0.84, r 2 = 0.06, n = 2, P < 0.001) and native habitat (0.41 ± 0.54, r 2 = 0.01, n = 224, P = 0.448), respectively.