Bottom-Up and Top-Down Effects Influence Bruchid Beetle Individual Performance but Not Population Densities in the Field

Plant quality (bottom-up) and natural enemies (top-down) can influence the individual performance of herbivorous insects on their host plants, but few studies measured at the same time the influence on population densities in the field. We investigated if plant quality of different wild common bean populations, Phaseolus vulgaris L. (Fabaceae), affects the performance of the bean weevil, Acanthoscelides obtectus (Say) (Coleoptera: Bruchidae), and one of its enemies, the ectoparasitoid Dinarmus basalis (Rondani) (Hymenoptera: Pteromalidae), in controlled laboratory experiments. Additionally, we examined if parasitoids influence the beetles' development and if increased individual beetle and parasitoid fitness lead to higher field population densities. We show that bean quality and parasitoids affected individual bean weevil performance under laboratory and field conditions. In the presence of parasitoids, fewer and smaller beetles emerged. However, beetle and parasitoid performance were not correlated. Increased individual performance was not leading to higher population densities; we found no correlations between measured performance components and beetle field infestation levels or parasitism rates. We conclude that bottom-up or top-down effects measured at the individual level do not always translate into population effects; therefore it is important to discriminate between effects acting on individual insects and those acting on populations.


Introduction
The importance of plant-mediated effects (bottom-up) and natural enemies (top-down) in determining insect herbivore abundance on plants has been and is still controversially discussed [1][2][3][4][5]. Recently it has become more evident that those factors do not act individually in shaping insect communities, but are integrated with other factors like the herbivores' life-history strategies, location and time [2,3,6]. Plant secondary metabolites can affect herbivore performance via e.g. survival and growth rates, allocation of resources to eggs and egg quality, which can lead to bottom-up regulation of herbivores [7][8][9][10]. Differences in food plant quality can be found among different plant genotypes or even among plant individuals, thus herbivores are expected to choose high quality hosts in order to maximise their fitness [11][12][13][14].
The quality of host plants as food does not only affect herbivores, but can also influence their natural enemies either directly (if enemies are directly exposed to chemical compounds) or indirectly, e.g. via altered host size [15][16][17]. Changes in host plant quality can thus cascade up to higher trophic levels and influence performance and abundance of parasitoids or even hyperparasitoids of herbivores [18][19][20]. Hunter [7] has reviewed indirect effects of host plants on parasitoids through altered host size, herbivore growth rate (apparency) and herbivore chemistry or vigour. He found that many studies show that preference or performance of parasitoids is linked with the size of herbivores, but few consider the role of plant quality in generating the variation in herbivore traits. More and larger herbivore hosts on high quality plants can produce more parasitoids of larger size with greater longevity and egg loads [21], which should lead to higher parasitism rates. Therefore plant nutritional quality could also affect population densities by increasing individual fitness of herbivores and their natural enemies, leading to larger parasitoid populations or higher parasitism rates on high quality host plants [9,20,22,23].
Some studies have investigated the influence of variation in plant nutritional quality on the individual performance of herbivores or natural enemies in the laboratory, but few considered simultaneously the impact of plant quality on populations of herbivores and parasitoids in the field [24,25]. Plant genotype has been shown to influence population dynamics or abundances of herbivores; whereas the influence on natural enemies is less pronounced. Kos et al. [20] have found an effect of chemistry and morphology in cultivated Brassica plants on the abundance of herbivores and natural enemies. In contrast, Newton et al. [9] reported no direct effect of plant genotype on numbers of natural enemies. Since evidence of host plant effects on field abundances of herbivores and natural enemies remain scarce, we wanted to assess those effects in common beans, Phaseolus vulgaris L. (Fabaceae).
Here, we first investigated in controlled laboratory experiments the effect of host plant nutritional quality on the performance of a herbivore and its parasitoid. In a second step, we examined if the benefit of nutritionally superior plants for the individual insects is also transferred to the population levels, i.e. if individual performance is correlated with field population densities of herbivores and parasitoids. We studied these multi-trophic effects in wild common beans, Phaseolus vulgaris L. (Fabaceae) and their main herbivore, bean weevils from the genus Acanthoscelides (Coleoptera: Bruchidae) and their parasitoids.
Apart from the main storage protein phaseolin, common bean seeds contain a family of closely related seed proteins (lectins and lectin-related proteins), who are considered to play a role in the plant's defence against attacks from herbivores [26]. In the intestinal tracts of herbivores, these defence proteins can disrupt cell walls and inhibit nutrient absorption [26]. It has been shown that seeds from wild P. vulgaris populations differ in their defence protein composition [27][28][29]. One of these defence proteins, arcelin, causes sub lethal effects in A. obtectus larvae, prolonging developmental time of beetles and reducing their weight [27,28]. Velten et al. [30] have shown that arcelin does not have any direct effects on the development of the parasitoid Dinarmus basalis since parasitoids feed on the hemolymph of beetle larvae and therefore do not come into direct contact with arcelin. However indirect effects of the protein, like reduced quality of their larval hosts can influence the parasitoid performance [30].
In addition to direct lethal effects, predators or parasitoids can have indirect non lethal effects, which are expressed as changes of the prey's behaviour in order to avoid being killed (Lima [31]), for example a reduction of the feeding activity [32][33][34][35][36]. Such nonconsumptive predator effects have been shown to influence herbivore population dynamics, and their effects on structuring predator-prey interactions may be as strong as or even stronger than the effects of direct consumption [31,[36][37][38]. In laboratory experiments, we investigated if the performance of A. obtectus is directly and indirectly affected by the presence of natural enemies by estimating the effect of Dinarmus basalis (Rondani) (Hymenoptera: Pteromalidae) on beetle weights and developmental times.
We examined several hypotheses: 1. Bean populations vary in their nutritional quality for bean weevils and therefore they should influence the performance of individual beetles. 2. We further expect that increases in individual beetle performance on good quality hosts translate into increased performance at the population level, i.e. higher field infestation rates. 3. Additionally we predict that not only lethal, but also non-lethal effects of the presence of parasitoids influence the performance/fitness of their hosts (trait-mediated predator effects). This should result in smaller beetles and longer developmental times when parasitoids are present. 4. Since the development of parasitoids should depend on the quality of their hosts, we expect wasps to grow larger on larger hosts. We thus predict to find correlations between sizes and developmental times of A. obtectus and their parasitoids. 5. Finally, we tested if parasitism rates (as an indicator for population densities) are higher on larger hosts and/or high quality bean populations in the field.

Bean collection and insect determination
Wild common bean, P. vulgaris, seeds were collected at different locations in Mexico, in the States of Colima, Jalisco, México, Michoacán, Morelos and Puebla from December till March 2009/ 2010. Bean populations were found in ruderal habitats, close to cultivated crops, along highways or in rocky areas at elevations from 1319 m.a.s.l. (meters above sea level) to 2159 m.a.s.l. 50-100 bean pods were collected from each of totally 26 populations (Table 1). Pods were shelled and seeds from each population were kept in separate containers at 27uC and 70% R.H. until bruchids and/or parasitoids emerged. Every second to third day, containers were checked for emerging insects.
We determined the sex of the emerged beetles and differentiated between the genera Zabrotes and Acanthoscelides. Two Acanthoscelides species are commonly found on wild P. vulgaris in Mexico, namely A. obtectus (Say) and A. obvelatus (Bridwell) [39]. Those sister species are closely related, co-occur on the same host plants and are difficult to distinguish from each other by their external morphology; thus they have been confused for a long time [40,41]. Apart from differences in voltinism (A. obtectus is multivoltine, A. obvelatus is univoltine), A. obtectus has found to be more frequent at lower altitudes (,1800 m.a.s.l.); whereas A. obvelatus prefers higher altitudes (.1800 m.a.s.l). However, the range of both species overlaps [41]. Due to the species' high ecological similarity and our sampling sites being at altitudes suitable for both (Table 1), we did not differentiate between the two Acanthoscelides species emerging from field collected seeds. Additionally we did not differentiate between Zabrotes subfasciatus (Boheman) and its close relative Z. sylvestris [42]. Parasitoids of the genus Horismenus were determined to species level with the help of a determination key and previously collected and determined specimen [43]. All other parasitoids emerged were determined to family level.
To measure variation in beetle sizes among different bean populations, we averaged for each population the weights of about each 8-15 randomly selected male and female Acanthoscelides spp., the main bruchid genus found in the field. Tibia lengths of female parasitoids of Horismenus missouriensis, the most dominant parasitoid species, were measured as a surrogate for their size [30]. Field infestation levels of Acanthoscelides spp. were calculated for each bean population by dividing the number of insects emerged by the number of seeds collected (wild bean seeds are very small compared to commercial beans, and usually are used by only one beetle; pers. obs.). Field parasitism rates were estimated by dividing the sum of parasitoids emerged by the sum of potential hosts, which was calculated as sum of emerged parasitoids + sum of emerged bruchids. Gregariousness and hyperparasitism were not considered.

Performance experiment with Acanthoscelides obtectus and Dinarmus basalis
Seeds of all 26 wild bean populations and two commercial Mexican bean cultivars (''Pinto'', ''San Lanzano'') were used for performance experiments with A. obtectus and the ectoparasitoid Dinarmus basalis, a solitary idiobiont ectoparasitoid on larvae and pupae of several species of grain and bean weevils [44]. Females of D. basalis are synovigenic, i.e. females mature eggs throughout adult life. Prior to the experiments, beans were stored at 220uC for two days in order to kill all potential insects inside seeds [45]. Beetles originated from beans collected at local Mexican markets and they have been reared in climate chambers on red kidney beans for several generations prior to the experiments. Parasitoids  For each bean population, the State of origin, GPS coordinates (latitude and longitude) and the altitude were recorded. Additionally field infestation levels of Acanthoscelides spp., the most dominant bruchid genus, were computed for each population by dividing the number of emerged beetles by the number of collected seeds. Parasitism rates were calculated for each bean population by dividing the number of emerged parasitoids by the number of potential hosts (sum of bruchids + sum of parasitoids). The mean tibia length 6 SD of H. missouriensis females, the most dominant species emerging from field collected seeds, and the mean weight 6

SD of
Acanthoscelides spp. were measured for each bean population. doi:10.1371/journal.pone.0055317.t001 were provided by Christoph Lüthi of the Research Station Agroscope Reckenholz-Tä nikon, Switzerland. Performance experiments were carried out according to a modified protocol of Velten et al. [30]. A. obtectus eggs were obtained by carefully sieving beans on which newly emerged adult beetles have been depositing eggs for 1 day. A. obtectus females do not attach their eggs to the surface of beans, but scatter them singly or in clusters among seeds [46]. Hatching larvae can move freely among beans and choose their hosts. Small plastic containers (height: 4 cm, diameter: 2.5 cm) were filled with 5 g of undamaged seeds of similar size. 50 eggs were added to each container using a moistened brush. We conducted 8 replicates for each of the bean populations tested. After 21 days at 27uC and 70% R.H., when bruchid larvae had reached the third to fourth instar, two males and two females of 2-6 day old D. basalis were introduced to half of the containers (4 replicates per bean population). Prior to the experiments, freshly emerged parasitoids were kept in small plastic containers without hosts and provided with a drop of honey to enhance oogenesis [47]. Parasitoids were left on experimental seeds throughout their lifespan, assuring sufficient time for host handling and mating [48]. Containers with beetles were checked daily for newly emerged insects. Emerged insects were kept singly in Eppendorf tubes, killed by deep-freezing at 280uC and beetles were immediately weighed. Bruchid sex was determined by dissecting the genitalia [39].
Parameters recorded for each beetle individual were its weight as an indicator for its size, sex and its developmental time, measured as number of days until adult emergence. Beetle survival was calculated as the percentage of A. obtectus emergence for each container by dividing the number of emerged bruchids by the number of eggs. For each parasitoid we recorded the developmental time and determined its sex. As a surrogate for parasitoid progeny size, we measured the length of the left hind tibia of each female parasitoid [30].

Data analysis
Performance experiments with A. obtectus and D.
basalis. Statistical analyses on the performance of beetles were carried out in R, version 2.14.0 [49]. For analyses with the dependent variables ''size'' and ''developmental time'', the individual beetles were the experimental units. For the analysis on ''survival'' (coded as binomial variable: survivors over dead, separate for males and females), the individual containers were the experimental units. Since data on individual beetles that emerged from the same container cannot be regarded as independent, we included the container as random variable in the analyses. The explanatory variables of all models were the bruchids' sex, if parasitoids were added or not and the bean population.
Data on A. obtectus weight showed no departure from the assumptions of normality and heteroscedasticity; thus they were modelled using a linear mixed effects model (function lme from the package nlme; [50]). An analysis of variance table (function anova) was used to investigate the effects of the fixed factors on beetles' weight. Data on developmental time were assumed to follow a Poisson distribution (first emergence counted as day zero) and were analysed with a generalized linear mixed effects model with a logarithmic link function (function lmer from the package lme4, version 0.999375-42; [51]). Overdispersion was accounted for by including beetle individuals as random variable in the model. An analysis of deviance table with type II error (function Anova from the package car, version 2.0-11; [52]), which uses a Wald Chisquare test, was used to investigate the effect of the explanatory variables on beetles' developmental time. A generalized linear model (function glm), assuming a binomial distribution of the response variable and a logit link function, was used to analyse the number of successfully emerged beetles over the failures per container. Results were displayed as analysis of deviance table with type II error (function Anova from the package car).
Since parasitoids are depending in their development on the bruchid hosts, we investigated if the performance of bruchid beetle individuals from the experiments showed an association with parasitoid performance using Spearman rank correlation tests (function cor.test). For this test, we computed for each bean population the mean bruchid weight (separately for males and females) and the mean bruchid and parasitoid developmental times from the containers to which parasitoids were added. The bruchid weight was tested for correlation with the size of parasitoid females and the mean beetle and parasitoid developmental times were tested for correlations.
Correlations between performance of insects in the laboratory experiments and the field. To test for an association between performance of bruchids on the same bean populations in the laboratory experiments and in the field, we used Spearman rank correlation tests. For each bean population we calculated the mean weight of A. obtectus males and females, the mean developmental time and the mean survival in the laboratory experiments. As surrogate for beetle performance in the field, we used for each bean population the mean weight of emerged Acanthoscelides spp. males and females. For parasitoids, we tested for correlations between mean female tibia sizes of D. basalis in the laboratory experiment and those of H. missouriensis, the most dominant parasitoid species in the field, on the same bean populations.
Correlations between host and parasitoid performance in the field. We hypothesized that parasitoids in the field should depend in their development on the quality of their beetle hosts; therefore we tested for a correlation between the mean size of H. missouriensis and the mean size of Acanthoscelides spp. males and females per bean population. Additionally we used Spearman correlations to investigate if parasitism rates were higher in bean populations with higher bruchid densities. To test if performance of Acanthoscelides spp. in the field is affected by parasitoid densities, we correlated the mean weight of beetle males and females per bean population with parasitism rates. Infestation rates (bruchids per bean, wasps per bruchid) are interpreted as indicators of population densities in the field (numbers per host, not per area).
Correlations between insect performance and population densities. Bean populations, which are of higher nutritional quality for bruchids and thus produce fitter individuals, could also allow larger insect population densities in the field. Therefore we tested for associations between performance of individual bruchids (laboratory and field) and their respective field infestation levels. We also investigated if there are correlations between parasitoid size in the field and parasitism rates. Correlations were carried out with Spearman's rank correlation test. We excluded populations where no insects emerged from the field-collected beans.
Bruchid females needed on average 33.560.05 SE days to develop to adults; whereas males needed on average 32.560.05 SE days (p,0.001; Table 3). When parasitoids were added to containers, beetles emerged earlier (p = 0.005; Table 3; estimate: 20.0260.01 SE). Developmental time of bruchids was varying among different bean populations (p,0.001; Table 3).
We found no significant correlation between performance of bruchid hosts and parasitoids in the experiment. The mean tibia length of parasitoid females in the experiment was not significantly correlated to the mean weight of bruchid hosts (Spearman's rank test: p = 0.26, rho = 0.255). We also found no significant association between D. basalis and A. obtectus developmental times (Spearman's rank test: p = 0.41, rho = 20.18).

Correlations between performance of insects in the laboratory experiments and in the field
In both bruchid sexes, we found significant positive correlations between mean field weights and weights from the performance experiment (Table 4). We found no correlation between parasitoid sizes measured in the experiment (D. basalis) and in the field (H. missouriensis) (Spearman's rank test: p = 0.62, rho = 0.143).

Correlation of host and parasitoid performance in the field
In the field, the size of H. missouriensis parasitoids was not determined by the size of their hosts (Spearman's rank test: bruchid females: p = 0.72, rho = 0.097; males: p = 0.19, rho = 0.346). Likewise, parasitism rates were not higher at higher host densities (Spearman's rank test: p = 0.38, rho = 20.192). Bruchid females developing in the field were smaller at high parasitoid population densities (Spearman's rank test females: p = 0.01, rho = 20.503). In contrast, male weight was not significantly reduced at high parasitoid densities (Spearman's rank test: p = 0.08, rho = 20.372).

Correlations between insect performance and population densities
We found no evidence that bean populations, from which larger beetles or parasitoids emerged, supported higher population densities in the field. No performance parameter of A. obtectus measured in the laboratory experiment correlated significantly with Acanthoscelides spp. field infestation levels ( Table 4). Additionally, the mean weight of Acanthoscelides spp. in the field did not correlate with field infestation levels (females: p = 0.98, rho = 0.006; males: p = 0.73, rho = 0.08). Parasitoid sizes measured in the experiments showed no correlation with parasitoid population densities in the field (Spearman's rank test: p = 0.93, rho = 0.02) and bean populations with larger H. missouriensis females did not have higher parasitoid population densities (Spearman's rank test: p = 0.33, rho = 20.259).

Performance of bruchid beetles and parasitoids in the experiment and field
Our results provide evidence that bottom-up (bean population) and top-down factors (presence of parasitoids) both influence the performance of A. obtectus at the individual level in the laboratory and in the field. Consistent with our hypotheses, beetle fitness components like weights, developmental time and survival were strongly affected by both the bean population and the presence of the parasitoid D. basalis. Our results confirm previous studies showing that those factors can influence herbivore performance simultaneously [3,5,53,54]. Varying nutritional quality of seeds is a likely explanation for the observed variation in A. obtectus performance among bean populations. The correlation of bruchid weights across bean populations between laboratory and field is a strong indication that nutritional quality of beans is also important in determining the performance and fitness of bruchids in the field. Other studies have confirmed that differences in nutritional quality (mostly due to allelochemicals) can affect the development of herbivores and also their natural enemies [9,20,24,55]. The domestication status of wild beans has been shown in laboratory experiments to be important in determining performance of herbivores and natural enemies [55]. Bruchid beetles of the genus Zabrotes and their parasitoids performed better on cultivated than on wild Phaseolus plants presumably because cultivars have lower concentrations of toxic allelochemicals and thus were easier to digest. Wild populations of common beans, P. vulgaris, vary in their seed defence protein contents and mainly arcelin has been found to affect bruchid beetle performance [27,28,30,56]. In another study [57] we have analysed the seed protein contents of some of the bean populations used in the performance experiments and found that arcelin is present and might be responsible for the low bruchid beetle performance in the bean population ''QUES''. However, we found no indications for the presence of arcelin in any other bean population. Therefore we conclude that there are components in bean seeds, other than arcelin, which are responsible for differences in plant quality among populations. For example, the concentration of phaseolin, which is the main seed storage protein and easily digestible, has been shown to be important for larval development [58,59]. The seed coat is another bean characteristic which can affect beetle larval performance. As a chemical (lignin content, bitingdeterrents) or mechanical barrier (hardness) it might influence the penetration success of first instar larvae and thus affect A. obtectus performance and adult emergence rates [60,61,62]. Stamopoulos et al. [61] have demonstrated that lignin of P. vulgaris teguments, which have been incorporated in artificial diets, can have negative effects on weights of emerging A. obtectus.
However, in contrast to other findings, we found no evidence that the influence of plant nutritional quality on the herbivores' performance was passed on to the third trophic level, since we did not find larger parasitoids on larger hosts, neither in the laboratory nor in the field. Likewise, no indications were found for correlations between host and parasitoid developmental times. This is in contrast to other study systems in which qualitatively superior host plants supported larger herbivores and thus larger parasitoids [18,19,63,64] and where the parasitoids developmental time was correlated with its hosts' developmental time [65]. Apparently differences in nutritional quality between bean populations, which affected bruchid beetle performance, did not alter the quality of beetles as hosts for parasitoids. The finding that parasitoid sizes in the experiment did not correlate with those in the field further indicates that parasitoid performance is governed  Table 4. Correlation between laboratory and field performance of Acanthoscelides beetles. more by other factors than the observed variation in host performance. These results indicate that although plant quality might affect herbivore performance in some cases, this effect is not necessarily passed on to higher trophic levels. However, some of the correlations might have a low statistical power due to the low numbers of parasitoids that emerged from some field samples (Table 1). Variation in plant quality can affect different trophic levels to different extents and the strength of these effects can decrease along the food chain [13,55,66]. Velten et al. [30] concluded that arcelin in common bean seeds, which affected A. obtectus performance, caused no direct effects on D. basalis progeny fitness and size (tibia length, head width) and parasitoids were able to develop on all tested bean lines containing different arcelin concentrations. Also, size might not always be the best indicator of parasitoid fitness. For example, under bean storage conditions, D. basalis shows a very good capacity to move through the seed column and therefore to locate hosts [67]; here, being too large may actually be disadvantageous for females. It should also be mentioned that increases in body size need sometimes to be traded-off with other parameters affecting fitness (e.g. suitability of habitat conditions, survival) or may simply be physiologically constrained [68].
In the laboratory, apart from parasitoids directly killing their hosts, we found indirect, non-lethal top-down effects on bruchids, which resulted in reduced beetle weights and shorter developmental times in the presence of parasitoids. These findings were partly confirmed in the field, where female beetles were remaining smaller at high parasitoid densities (indicated by high parasitism rates); for males this trend was also present, but just not significant. Reduced feeding activity of beetle larvae in the presence of natural enemies could explain smaller weights of adult beetles. Vibrations, that are unavoidable during host searching by parasitoids, can be used by insect larvae living in the substrate to detect the presence of their enemies [69], and arresting feeding themselves prevents emitting vibrations that would give away their presence to foraging wasps [70]. Bruchid larvae also started pupation earlier in the presence of parasitoids, which shortens the time they are exposed to parasitoid attack. The benefit of such anti-predatory behaviours is a lower risk of being killed; the cost however is a usually lower energy intake rate, which results in smaller sizes and reduced fecundity [31]. Such indirect trait-mediated predator effects can be as strong as direct lethal effects in influencing herbivore communities [37,38]. Skelly and Werner [33] have shown that tadpoles metamorphosed at smaller sizes, when predators were present and they have argued that predators are important in structuring the behaviour and life-historical attributes of prey, even without considering lethal effects.
Reduced A. obtectus weights in the presence of D. basalis could also be explained by oviposition preferences of parasitoids for larger hosts, because in our experiments we didn't prevent parasitoids from laying eggs. Larger hosts are a larger resource for the developing parasitoids and could thus be more profitable than small ones [71,72] or easier to locate for parasitoid females [73]. However, we find it unlikely that D. basalis females in the experiment showed a strong preference for larger hosts because of several reasons. First, we did not find a correlation between host size and parasitoid size, indicating that parasitoids do not actually grow larger on large hosts, and therefore there is no obvious reason to prefer larger hosts. Second, D. basalis females anesthetize the host larvae before depositing the eggs thereby preventing further larval growth after parasitism [74]. Thus, if parasitoids would preferentially lay eggs on large hosts and there would be no non-lethal negative effects on bruchid larvae, the fast-developing larvae would be parasitized first and the remaining slow-growing larvae, that escaped parasitism, would develop to their final size, which is reached at a later time. This would actually increase larval developmental times while having no strong effect on final host size. Our results suggest the opposite: reduced developmental times and smaller final size in the presence of parasitoids. Thus, we conclude that we found strong indications for the presence of traitmediated indirect top-down effects. A decisive experiment would be to use parasitoids which can search for hosts, but are not allowed to oviposit; however, this is technically difficult to reach.

Correlations between performance of insects and densities in the field
We found no evidence that the better performance of bruchids or parasitoids on certain bean populations would lead to an increase in field abundances. Measured field weights, parasitoid sizes and bruchid experimental performance components (weight, developmental time and emergence rates) showed no correlation with bruchid field infestation levels or parasitism rates. In contrast, other studies have shown that herbivore population sizes or population growth rates are bottom-up regulated and differ between plant genotypes [9,24,25]. Aphid colony sizes in field systems were found to be more bottom-up regulated by plant genotype, while natural enemy abundance was unaffected by the plant secondary metabolites [9]. Johnson [25] found evidence that plant genotype had a significant direct effect on the abundance of natural enemies, irrespective of the herbivore density and an indirect effect, mediated through herbivore density. However in our study, parasitism rates did not depend on host densities (bruchid infestation levels). This result is consistent with other study systems in which the percentage of parasitized hosts was density-independent [23].
It appears that other factors, for example environmental stochasticity or the surrounding landscape, have a more important effect on population densities in this system than top-down or bottom-up effects. It has been shown that Mexican bean weevil, Z. subfasciatus, females show behavioural and physiological plasticity in oviposition behaviour according to host availability [75]. When beans were scarce and competition was high, beetles laid more eggs onto the same seed and fewer, less fecund adults emerged. This could have strong impacts on beetle population densities when plant resources vary in their availability from season to season, as it seems to be the case in wild beans (personal observation). Bowler and Benton [76] have shown that soil mite populations, which experienced variation in daily food supply (variable food availability versus constant food supply), had lower and more variable population densities than populations in a constant environment. Environmental variation has been shown to be important in influencing population dynamics [77][78][79]. Mutshinda et al. [79] have found, when analysing community time series among different taxa that population dynamics were dominated by environmental stochasticity. This accounted for 40-95% of the temporal variances in individual species abundances. However, the variability of an environment, expressed as for example in resource stochasticity, is an important factor in determining insect population abundances [80]. One limitation of our study is that we collected field data from only one year. In order to measure the effect of environmental stochasticity on population dynamics over time, it would be necessary to analyse insect abundance data from several years.
Parasitoids of the genus Horismenus are important natural enemies of bruchid beetles, but little is known about their biology and most species of the Neotropical region remain undescribed [43,81]. Horismenus missouriensis is a generalist parasitoid of Coleoptera, Lepidoptera and Diptera [82]. Since those parasitoids are generalists and accept a wide range of hosts, they do not depend upon the density of one particular host species. This could be another explanation why parasitoids were unaffected in their densities of Acanthoscelides spp. field infestation levels.

Conclusions
Our study shows that plant quality (bottom-up) and natural enemies (top-down) act together in influencing performance of a herbivore, confirming previous studies. We also showed that enemies can have direct and indirect, non-lethal effects on individual host performance. However, the bottom-up effects of plant quality on individual bruchid beetles was not passed on to the third-trophic level. Up to now, few laboratory experiments and field studies have investigated and provided evidence for plant quality effects cascading up to higher trophic levels [13,25,64,65,83]. Although in some systems, plants might affect natural enemies; apparently in our study, factors other than plant quality are more important in determining parasitoid performance. Furthermore, we found that increased individual performance does not necessarily translate into increased densities at population levels. Our results are in contrast to other studies, which provided evidence for bottom-up regulation of herbivore populations by plant genotype [25]. A plausible explanation is that in our system environmental variation can have a larger impact on insect communities than biotic interactions. We conclude therefore that it is important to differentiate between effects acting on individual insects and those acting on insect population levels. Factors influencing lower trophic levels do no necessarily cascadeup to higher trophic levels and increased individual fitness does not necessarily result in increased population densities. However, to elucidate the effect of environmental stochasticity on abundances of bruchid beetles and their natural enemies, long-term studies are needed.