Strong selective pressure by parasitoids mitigates the costs of consuming chemically defended plants

Co-evolutionary interactions between plants and herbivores have led to a range of plant defenses that minimize insect damage and a suite of counter-adaptations that allow herbivores to feed on defended plants. Consumption of plant secondary compounds results in herbivore growth and developmental costs but can be beneficial if eating these secondary compounds results in deterrence or harm to natural enemies. To test the role of secondary compounds on herbivore fitness in the context of natural enemies, I combined field measurements of the prevalence of a parasitoid wasp (Cotesia congregata) with detailed measurements of the costs of plant secondary compounds on growth, immune, and fitness traits across developmental stages in the herbivore Manduca sexta. When M. sexta larvae consume defended plants, Cotesia congregata are known to have reduced success. However, this anti-enemy benefit to the M. sexta host must be considered in relationship to parasitoid abundance and the type and strength of the fitness costs M. sexta incurs feeding on plant secondary compounds. I found that Cotesia congregata parasitoids exert large negative selective pressures, killing 31-57% of M. sexta larvae in the field. Manduca sexta developed fastest during the instars most at risk for parasitoid oviposition but growth was slowed by consumption of plant secondary compounds (nicotine and rutin). These negative size effects at the larval stage carried over to influence adult traits associated with flight and mating but there were no immune, survival, or fecundity costs of consuming plant defensive compounds as larvae. Synthesis. These results suggest that the developmental costs experienced by M. sexta herbivores consuming defensive compounds may be outweighed by a survival benefit in the face of abundant enemy pressures.


Introduction
Coevolution between herbivorous insects and their host plants often mitigates their reciprocal (negative) fitness effects, resulting in rapid evolution of plant anti-herbivore defense and insect counter-adaptations to these defenses (Ehrlich andRaven 1964, Maron et al. 2019).
Plants produce combinations of physical and chemical defenses that may lower herbivore fitness by reducing growth, disrupting development, decreasing survival, and/or attracting natural enemies of herbivores (Price et al. 1980, Howe and Jander 2008, Furstenberg-Hagg et al. 2013. Herbivores feeding on defended plants may experience immediate or delayed effects of consuming secondary compounds and these consequences may manifest differently across herbivore life stages (Fellous andLazarro 2010, van Dam et al. 2011). Understanding how insect herbivores evolve to feed on defended plants requires quantifying the fitness effects of plant defenses across herbivore life stages and examining the conditions under which consumption of plant secondary compounds is beneficial to herbivores. An important component of estimating these fitness consequences is to determine how additional stressors, such as natural enemies, affect herbivore fitness in conjunction with plant defense.
One way that herbivores mitigate the negative consequences of plant defense is if consumption of these secondary compounds directly or indirectly harms natural enemies. Insects are predicted to consume defended plants despite the apparent negative effects if the anti-enemy benefit of consuming plant secondary compounds outweighs the costs. Enemies can exert a large negative selective pressure on their targets when the enemies are at high abundance and/or when they drastically decrease herbivore fitness (Hassell and Waage 1984). Parasitoids, for instance, have an especially negative fitness effect because they kill their host at an immature (prereproductive) stage and reduce host fitness to zero (Godfray 1994). Endoparasitoids-those that develop inside the body of their hosts--are common enemies of lepidopteran species and endoparasitoid fitness may be particularly affected by host quality (Eggleton and Belshaw 1992).
In insect hosts that sequester secondary compounds, there is clear co-option of plant toxins for herbivore anti-enemy defense but sequestration is considered more effective against predators rather than parasitoids (Gault et al. 1992, Smilanich et al. 2009).
In herbivore hosts that do not sequester plant secondary compounds, such as Manduca sexta, the reduced success of endoparasitoids on hosts fed secondary compounds may be either the result of direct toxicity to parasitoids or the indirect effects on host quality (Beckage and Riddiford 1978, Thorpe and Barbosa 1986, Barbosa et al. 1991Harvey et al. 2007).
Endoparasitoids may experience exposure to the toxic compounds their herbivore hosts consume as these compounds are detoxified or excreted from the host (Wink andTheile 2002, Kumar et al. 2014). Endoparasitoids are also sensitive to the indirect effects of secondary compounds on host growth, survival, and immune function (Parr and Thurston 1972, Price et al. 1980, Barbosa et al. 1991, Appel and Martin 1992, Stamp 1993, Alleyne and Beckage 1997, Ode 2006, Bukovinsky et al. 2009, Thaler et al. 2012, D'Incao et al. 2012. One host immune response in particular, the encapsulation and melanization of parasitoid eggs, can decrease parasitoid egg hatching success but whether this response is increased or hindered by secondary compounds is hard to predict given current evidence (Kraaijeveld et al. 2001, Bukovinszky et al. 2009, Smilanich et al. 2009).
Disentangling the effects of secondary compounds and their impacts on herbivore immune function and growth is necessary to determine how secondary compounds alter herbivore health and interactions between herbivores and parasitoids. Secondary compounds may prime the insect's immune response to allow the insect to better respond to subsequent stress, such as parasitoid attack, or (alternatively) insect immunity may suffer as a result of eating defended plant tissues (Fellous and Lazzaro 2010). The impact of secondary compounds on herbivore immune function must be interpreted in the context of larval growth since limited resources are predicted to result in a trade-off between growth and immunity (Ode 2006, Bascunan-Garcia et al. 2010van der Most et al. 2011). Secondary compounds may also have direct effects on specific herbivore developmental stages that increase exposure to parasitoids.
Studies that measure overall increases in time from hatching to pupation or short-term decreases in growth rate do not fully capture whether these developmental changes alter herbivore fitness or exposure to parasitoids (e.g. Parr and Thurston 1972, Granzow et al. 1985, Barbosa et al. 1991, Harvey et al. 2007, but see Van Dam et al. 2001).
Whether larval consumption of secondary compounds influences herbivore fitness depends in part on if stress experienced at the larval stage impacts adult mating and reproductive traits (Bessin andReagan 1990, Spurgeon et al. 1995). The effect of larval experience on reproductive fitness may differ based on insect life histories and developmental patterns. Larval experience has been shown to impact adult traits in non-holometabolous insects (Hopkins 1917, Corbet 1985, but has received less attention in insects that undergo metamorphosis because pupation is often thought of as a re-setting period that can erase larval experiences (Barron 2001, Fellous andLazzaro 2010). However, pupal size and adult fecundity are correlated in some lepidopteran species, indicating that larval experiences are not always negated by the restructuring that occurs during pupation (Bessin and Reagan 1990, Spurgeon et al. 1995, Kariyat and Portman 2016. Therefore, quantifying the impacts of larval stress on adult morphological and behavioral traits is necessary to determine if there are lasting effects of plant defense on adult mating and fitness traits or if larval consumption of secondary compounds simply alters survival to adulthood but the surviving adults are unaffected by larval stress.
In this study, I use the herbivore Manduca sexta (Lepidoptera: Sphingidae) to determine the fitness costs posed by natural enemies and the costs of larval consumption of chemically defended plants on herbivore immunity, development, survival, and adult fitness traits. By collecting and monitoring M. sexta from a field population, I establish that Cotesia congregata parasitoids exert large negative survival costs on larval M. sexta, creating a scenario where the protective effects of specializing on host plants high in secondary compounds could outweigh the negative developmental effects on herbivores. Using field-collected and lab colonies of M. sexta, I show that while two different types of secondary compounds (inducible nicotine and constitutive rutin) affect M. sexta larval and adult size and morphological traits, they do not have strong negative effects on survival to adulthood, immune responses to artificial parasitoids, or adult fecundity. Because nicotine is known to have protective effects against C. congregata parasitoids (Beckage and Riddiford 1978, Thorpe and Barbosa 1986, Barbosa et al. 1991, Harvey et al. 2007, these results suggest that the developmental costs experienced by M. sexta consuming defensive compounds may be outweighed by a survival benefit in the face of abundant enemy pressures.

Materials and Methods
Study system: Manduca sexta and Cotesia congregata Manduca sexta are ecologically and economically important pollinators and herbivores of Solanaceous plants. While feeding on host plants, M. sexta larvae are targeted by natural enemies, including Braconid wasp and Tachinid fly parasitoids that lay eggs inside their hosts (Yamamoto andFraenkel 1960, Stireman et al. 2006). Cotesia congregata parasitoid eggs hatch and feed inside the M. sexta host larvae before they emerge from the host larval cuticle to pupate, ultimately killing the host (Alleyne and Beckage 1997). Prior studies have shown that consumption of plant secondary compounds by larvae of M. sexta can be protective against parasitoids (Beckage and Riddiford 1978, Thorpe and Barbosa 1986, Barbosa et al. 1991Harvey et al. 2007). 28 July 2014 N = 141). These dates were timed to occur after the residual insecticide used in transplanting (late May-early June) wore off and before application of additional pesticides.

Measurements of parasitoid abundance on field M. sexta
After each of the three field collections, M. sexta larvae were brought back to the lab and monitored twice daily for parasitoid emergence. Because M. sexta consume a large amount of leaf tissue, field-collected larvae were transitioned to an artificial wheat germ-based diet with 10-20% wet volume of Solanaceous leaf tissue added to facilitate diet acceptance (SI Table 1).
Parasitoid development takes a predictable number of days, meaning that the time between field collection and parasitoid emergence can be used to estimate the instar at which parasitoid oviposition occurred (Gilmore 1938). In July 2014, I recorded the approximate instar at field collection and determined the time it took parasitoids to emergence from different host instars.
Chi-squared tests were used to test for variation in the proportion of M. sexta larvae with parasitoids among the three field collection dates. Parasitoid load and number of days post-field collection until C. congregata emergence were compared for larvae collected from the field at different instars using Poisson general linear models (GLM) (R v. 3.2.2) (R Core Team 2014).
Robust standard errors were used as Breusch-Pagan tests showed heteroskedasticity (bptest() in LMTEST; Zeileis and Hothorn 2002). Instar five was excluded from instar-specific models because I collected only two fifth instars. Preparation of M. sexta diets with secondary compounds Using artificial diets containing either nicotine or rutin (SI Table 1), I tested the effects of secondary compounds on M. sexta development and fitness traits. As a specialist herbivore, M. sexta often feed on leaves containing nicotine, a pyridine alkaloid found only in the Solanaceae plant family, which serves as a defensive chemical against herbivory and can be induced via the jasmonic acid pathway (Keinanen et al. 2001, Steppuhn et al. 2004. I used 0.5% wet weight nicotine, which represents a high but relevant concentration that larvae feeding on tobacco would encounter (Parr and Thurston 1972, Saitoh et al. 1985, Sisson and Saunders 1982, Sisson and Saunders 1983, Thompson and Redak 2007. To test whether herbivore responses to plant defensive chemicals are consistent across different compounds, I also tested the effects of 0.5% rutin (quercetin 3-rhamnoglucoside) on the same herbivore traits. Rutin is found in 32 plant families and is constitutively present at 0.008-0.61% wet mass in tobacco (Krewson and Naghski 1953, Keinanen et al. 2001, Kessler and Baldwin 2004. Manduca sexta used for the immunity, growth, and adult measurements were fed artificial diet (control, 0.5% nicotine, or 0.5% rutin depending on treatment) ad libitum.

M. sexta larval immune responses to artificial parasitoids
Injections of artificial parasitoid eggs into M. sexta larvae were used to test whether secondary compounds alter host immune responses and if growth and immunity trade off.
Manduca sexta from both colonies were collected concurrently as neonate larvae and reared individually on 0.5% nicotine, 0.5% rutin, or control diets until the fourth instar. Forceps were used to insert an artificial parasitoid egg (a 2 mm-long piece of roughened nylon filament) through a needle hole pricked behind the fourth proleg (as in Piesk et al. 2013). Larvae were returned to their respective diets and fed readily after egg insertion. Pre-challenge growth rate was calculated as ln(larval mass at fourth instar)/number of days from hatching to fourth instar.
Post-challenge growth rate was calculated as ln(larval mass 24 hours post egg insertion/larval mass at time of egg insertion) (Diamond and Kingsolver 2011).
After the final weighing, larvae were frozen at -20°C for dissections to quantify the strength of the immune response to the artificial parasitoid. Melanization (dark buildup by hemocyte immune cells) was photographed using a Leica M205FA Stereo microscope and the percent melanized was calculated using ImageJ (Diamond and Kingsolver 2011). Percent melanized was used for GLM with robust standard errors to determine whether immune responses differed based on secondary compounds, prior condition (pre-challenge growth rate), or trade-offs between post-challenge growth rate and melanization. Pairwise interactions between diet-colony and diet-growth rates were non-significant (P > 0.1 for all) and were removed from the model. Area melanized was transformed for non-normality using Box-Cox

M. sexta larval and pupal traits on nicotine and rutin diets
Because of the large numbers of M. sexta needed, the effects of nicotine and rutin on growth and fitness traits were tested and analyzed at separate generations. Larvae fed control or experimental diets (0.5% nicotine or 0.5% rutin) were monitored daily for molting and the number of days per larval instar. The total number of larval instars was recorded because larvae undergo either five or six instars depending on size (Kingsolver 2007) (SI Table 2). Poisson GLM and Wald tests were used to test for variability in the number of days per instar and whether any instar was more sensitive to the effects of nicotine or rutin (wald.test() in AOD; Lesnoff and Lancelot, 2012). Non-significant interactions (P > 0.1) between instar and nicotine or rutin were removed. I recorded the number of days to pupation and pupal mass for M. sexta males and females to test whether larval consumption of nicotine or rutin altered pupal traits. Poisson GLM was used to determine if nicotine or rutin extended the number of days to reach pupation. ANOVA with type III sums-of-squares was used to test for differences in pupal size (mass) based on secondary compounds or sex and whether the effects of nicotine or rutin were stronger for either sex (diet*sex interaction) (Anova() in CAR; Fox and Weisberg 2011). Pupal mass for lab moths in the rutin experiment was transformed by BC = 2. One-sided Fisher tests (fisher.test()) were used to test if secondary compounds increased larval and pupal deformities.

M. sexta adult size and fitness traits
To test if larval consumption of secondary compounds resulted in size differences postpupation, adults were frozen at -20°C the morning post-eclosion to measure adult body and wing length. Kaplan-Meier survival analyses were used to determine if either secondary compound reduced moth survival to eclosion (survdiff() in SURVIVAL; Therneau and Grambsch 2000, Therneau 2015). ANOVA models for adult body and wing length included diet, sex, and a diet* sex interaction. Wing length was transformed by BC = 6 for moths in the lab colony.
Fecundity estimates were obtained by dissecting ovarioles from adult females and counting follicle numbers under a dissecting scope (as in Diamond et al. 2010). Because larger moths may produce more eggs, the ratio of follicles to body area was used as the dependent variable in ANOVA models testing for an effect of larval consumption of secondary compounds on fecundity. Follicles/body area was calculated as: (number of follicles in a female moth) / (½ body length x ½ body width x 3.14). Correlations among adult traits are shown in SI Table 3.

M. sexta larval dietary choice trials
Binary choice trials were used to determine if neonate M. sexta exhibit a preference for artificial diets with or without secondary compounds. Neonate larvae from both colonies were collected within three hours of hatching and placed in the center of individual petri dishes (nine cm diameter) with 1 cm 2 pieces of control diet and experimental diet (0.5% nicotine or 0.5% rutin) placed on opposite sides. Dishes were oriented haphazardly under 14:10 dark:light conditions and monitored at 1, 6, and 24 hours before scoring contact with either diet at 48 hours as a choice. Larvae did not leave or switch once choosing a diet. Chi-Squared analyses were used to test if control or experimental diets were chosen significantly more than half the time. Larvae that did not choose in 48 hours (field N = 11/60 and lab N = 3/76) were excluded.

Results
Cotesia congregata parasitoids are common on M. sexta larvae in the field Surveys of a tobacco plot at three timepoints revealed that a high but variable proportion of M. sexta larvae were parasitized. The highest proportion of larvae parasitized by C.
The timing of C. congregata parasitoid emergence from M. sexta collected in the field at different instars was consistent with parasitoids ovipositing in younger larvae. Because parasitoids take a predictable amount of time to emerge from the host cuticle after oviposition (Gilmore 1938), the length of time between field collection and parasitoid emergence for larvae of different instar stages can be used to determine the age at parasitism. The number of days post-M. sexta field collection to C. congregata emergence was higher for host M. sexta collected as younger instars (GLM; instar 2 b = 2.599, P < 0.01; instar 3 b = -0.629, P < 0.001; instar 4 b = -1.100, P < 0.001). There were no significant increases in the number of parasitoids emerging from third and fourth instar host M. sexta larvae compared with second instar host larvae (GLM; instar 2 b = 2.99, P < 0.01; instar 3 b= 0.255, P = 0.236; instar 4 b = 0.340, P = 0.094) ( Table 1).
Immune responses to an artificial parasitoid do not trade-off with larval growth Following implantation of an artificial parasitoid egg, Manduca sexta immune response (melanization) was higher in larvae with faster growth rates, regardless of diet. There was a positive relationship between growth rate post-challenge and the melanization of the artificial parasitoid (GLM; z = 0.20, P < 0.001). Growth rate prior to the artificial parasitoid did not affect melanization (z = -0.03, P = 0.85). Neither nicotine (z = -0.02, P = 0.37) nor rutin (z = 0.04, P = 0.10) affected melanization. Larvae from the field-collected colony had higher melanization levels than larvae from the lab colony (z = 0.080, P < 0.001).

Secondary compounds increase developmental time for each larval instar
Developmental assays of M. sexta larvae revealed that the number of days needed to complete each instar is variable and secondary compounds extend the length of each instar.
Nicotine and rutin increased the number of days needed to complete each of the first four instars but specific instars were not more sensitive to the effects of the secondary compounds (Table 2).
Larvae spent the fewest number of days in the second instar but there was variation in development times between the nicotine and rutin experiments. In the M. sexta generation used to test the effects of nicotine, the number of days taken to complete the second and third instars was shorter than the number of days taken to complete the first and fourth instars (Table 2A). In the generation used to test the effects of rutin, only the second instar was shorter (Table 2B).
The overall effect of secondary compounds on larval development was to increase the number of days from hatching to pupation. In both colonies, the number of days from hatching to pupation was higher on the nicotine diet (GLM; field nicotine z = 2.183, P = 0.029; lab nicotine z = 4.039, P < 0.001) and on the rutin diet (field rutin z = 4.149, P < 0.001; lab rutin z = 5.65, P < 0.001) compared to control diets (Table 3). Larvae normally complete five instars, but a small percent of larvae went through an additional sixth instar prior to pupation and this was more common in M. sexta in the lab colony fed nicotine and for M. sexta in lab and field-collected colonies fed rutin (SI Table 2).

M. sexta survival to adulthood is not decreased by larval consumption of secondary compounds
The proportion of M. sexta surviving to adult eclosion was not significantly reduced by either secondary compound. Larvae from the lab colony had only marginally significantly reduced survival to adult eclosion when reared on nicotine compared to those fed the control diet (X 2 1 = 3.6, N = 149, P = 0.057) and there were no differences in survival for moths from the field-collected colony when fed nicotine versus control diets (X 2 1 = 0, N = 152, P = 0.886) ( Fig.   2A). Rutin did not significantly decrease survival of moths from either colony (lab: X 2 1 = 0.7, N = 158, P = 0.408; field X 2 1 = 0, N = 189, P = 0.956) (Fig. 2B).
Adult body size was smaller when moths had consumed secondary compounds as larvae Measurements of body length on newly eclosed adults showed that consumption of secondary compounds at the larval stage decreased adult M. sexta size, but the effects differed for female and male moths. For both colonies, nicotine decreased adult length (ANOVA: field nicotine F1, 72 = 7.978, P = 0.006; lab nicotine F1, 105 = 19.896, P < 0.001). The negative effect of nicotine on adult length was stronger on females than males from the field-collected colony (field sex*nicotine F1, 72 = 4.072, P = 0.047) but there was no sex difference in the effect of nicotine in the lab colony (lab sex*nicotine F1, 105 = 0.366, P = 0.546).

Female fecundity was unaffected by larval consumption of secondary compounds
Female fecundity (follicle number) did not differ between M. sexta reared on control diets versus those reared on diets containing secondary compounds, even when taking adult size differences into account. Although pupal weight and adult size traits were positively correlated, correlations between adult body size and follicle numbers were not present in most treatments (SI Table 3). Larval consumption of nicotine did not reduce adult follicle-body area ratios (ANOVA; field nicotine F1,39 = 0.113, P = 0.738; lab nicotine F1, 57 = 0.663, P = 0.419).
Neonate M. sexta do not show behavioral avoidance of diets with secondary compounds.
In the behavioral experiments testing for a preference for diets with or without secondary compounds, neonate larvae did not display significant avoidance of either nicotine or rutin diets.

Discussion
Plant-insect coevolution depends not only on plant defenses and herbivore counteradaptions to these defenses, but also on tri-trophic interactions that include top-down effects  (Table 4). These findings fit with the prediction that herbivore consumption of plant secondary compounds is beneficial when these compounds harm natural enemies that exert a large negative selective pressure on herbivores (Gauld et al. 1992).
Previous studies have established that a defended diet is protective against parasitoids (Thorpe and Barbosa 1986, Barbosa et al. 1991, Harvey et al. 2007) but the ecological importance of this benefit is highly dependent on parasitoid prevalence. Prior studies using controlled parasitoid oviposition on predetermined hosts or studies that introduce laboratory M. sexta into a field setting may not reflect interactions in the field. In this study, I provide important evidence that that parasitoids kill a large proportion of M. sexta larvae in the field, representing a total loss of fitness for infested hosts.
The M. sexta larval growth patterns seen in this study minimize exposure to parasitoids during the timeframe that larvae are most likely to be parasitized. I found relatively faster M. sexta development time during the second and third instars, which aligns with the instars preferred for C. congregata oviposition (Gilmore 1938, Beckage and Riddiford 1978, Barbosa et al. 1991, Kingsolver et al. 2012). In the field survey, similar numbers of parasitoids emerged from larvae removed from the field as second to fourth instars, suggesting that larvae remaining in the field as fourth instars did not result in additional parasitoids. Parasitoids also emerged quickly from larvae removed from the field as fourth instars, indicating the parasitoids had been laid prior to the fourth instar based on a 12-16 day oviposition-to-emergence time (Gilmore 1938).
Rapid development that reduces exposure time to parasitoids is expected to be beneficial, as fast growth did not come at the cost of reducing immune responses to parasitoid eggs. In contrast to the predicted energetic trade-off between growth and immunity (Smilanich et al. 2009), I found that larvae with higher growth rates following injection of the artificial parasitoid egg actually had higher levels of melanization regardless of control or defended diets. The lack of a growth-immune trade-off in Manduca sexta is likely because they do not actively sequester plant compounds and therefore do not have this energetic cost (Wink andTheile 2002, Smilanich et al. 2009). Because nicotine and rutin did not increase melanization of parasitoid eggs, host immune responses to secondary compounds are unlikely to be responsible for the reduced parasitoid success on hosts fed nicotine seen in other studies (Beckage and Riddiford 1978, Barbosa et al. 1986, Barbosa et al. 1991, Harvey et al. 2007). The protective effects of secondary compounds against M. sexta parasitoids probably result from toxicity of nicotine to the parasitoids or the indirect effects of slowed M. sexta growth on C. congregata development (Barbosa et al. 1986, Appel andMartin 1992). Interestingly, I observed higher melanization rates in the field-collected colony than in the lab colony, which may reflect that the lab colony has reduced consumption of defended diets (Voelckel 2001), or disruptions in juvenile hormone (Lee et al. 2015). The low levels of incomplete molting I observed in M. sexta larvae indicate any changes in hormone levels due to secondary compounds were not enough to fully disrupt molting. Regardless of mechanism, an extended development time means herbivores are exposed to parasitoids for longer when feeding on defended tissues, but this may be offset by the smaller size of these larvae making them harder for parasitoids to locate (Clancy and Price 1987, Benrey andDenno 1997).
These effects of exposure to secondary compounds at the larval stage contribute to M. sexta fitness either by altering the probability of surviving to reproduce as adults or through correlations with adult reproductive traits. In the absence of parasitoid pressures, larval consumption of secondary compounds did not affect survival to adult eclosion or fecundity but had negative effects on adult body size and wing size. Females with smaller bodies have been shown to have reduced pheromone production in other Lepidoptera and may be less attractive to males (Harari et al. 2011). Wing size is positively correlated with increased flight time and distance which may be important for finding mates or host plants for egg laying (Shirai 1993, Berwaerts et al. 2002, Cahenzli et al. 2015. Although I found that fecundity (female follicle numbers) did not differ based on consumption of secondary compounds, actual fertility may be lower if these eggs are not fertilized because of reduced mating success. Body and wing traits may also impact an adult's ability to disperse offspring and/or choose appropriate oviposition sites. However, the impact of maternal oviposition choice depends on whether offspring can behaviorally select their own feeding sites or if early diet is determined by hatching location (Jaenike 1978, Soler et al. 2012. The lack of neonate differentiation I observed between defended and non-defended diets is evolutionarily important because it indicates that maternal oviposition choices rather than offspring choices likely determine whether offspring experience early exposure to secondary compounds (Kester et al. 2002). Behavioral responses to plant compounds may vary based on herbivore age. In other studies that have used older larvae, nicotine has shown to be deterrent (Kester et al. 2002;Parr and Thurston 1972) while rutin has not been shown to deter feeding and has even been seen to stimulate feeding (De Boer andHanson 1987, Stamp andSkrobola 1993).
The lack of an effect of secondary compounds on neonate choice could indicate that mechanisms needed to recognize chemical cues are not completely developed until later instars, that there are additional important leaf cues that are not present in an artificial diet, or that deterrence may occur via post-ingestive mechanisms rather than pre-ingestive mechanisms (Glendinning 2002).

Conclusions
Overall, the results of this study indicate that although nicotine and rutin differ in their chemical composition and prevalence in plants, both have negative effects on M. sexta that extend past the larval stage at which the compounds are consumed. Despite effects on adult body and wing size that may influence mating and offspring dispersal, there were no strong effects on survival or fecundity. Therefore, the negative effects of secondary compounds on M. sexta development and fitness are likely outweighed by the protection that ingestion of these compounds offers against C. congregata parasitoids, which exerted large negative survival costs on M. sexta in the field. At a larger scale, coevolutionary and tri-trophic interactions can maintain a balance between the costs and benefits of secondary compounds on herbivore fitness.    follicle number ---- Figure 1. Manduca sexta pupae were smaller when larvae were fed secondary compounds. A) M. sexta larvae fed nicotine (red) weighed less at pupation than those fed control diets (grey) (ANOVA, P < 0.01 for both colonies) B) M. sexta larvae fed rutin (green) weighed less at pupation than those fed control diets (grey) (ANOVA, P < 0.01 for both colonies). Separate controls are presented for nicotine and rutin because the effects of these secondary compounds were tested at different generations. versus control diets (grey) but there was a slight, non-significant decrease in survival for lab colony fed nicotine (Kaplan-Meier survival curves; field P = 0.9, lab P = 0.06). B) Neither colony had reduced survival on rutin (green) versus control diets (grey) (field P > 0.01, lab P > 0.1). Tick marks represent censored data (insects that pupated prior to end of time period).

Figure 3.
Manduca sexta adults had shorter wings when reared on larval diets containing secondary compounds. Males had shorter wings than females but there were no sex-specific effects of either compound (ANOVA P > 0.1 for all sex-diet interactions). A) Nicotine (red) decreased wing length compared to control diets (grey) in both the lab (left pane, ANOVA P < 0.01) and field-collected colonies (right pane, P< 0.01). B) Rutin (green) decreased wing length compared to control diets (grey) in both the lab (left pane, P < 0.01) and field-collected colonies (right pane, P < 0.01). Separate controls are presented for nicotine and rutin because the effects of these secondary compounds were tested at different generations.