Linking nutrition and sexual selection across life stages in a model butterfly system

Authors


Correspondence author. E-mail: natasha.tigreros@tufts.edu

Summary

  1. Early nutrition plays an important role in determining adult fitness. Theory proposes that in organisms with complex life cycles, the fitness effects of larval nutritional constraints are mainly captured by two developmental traits: time and size at metamorphosis. However, recent evidence suggests that latent effects, which are independent of these developmental traits, must be included to fully understand how larval nutrition impacts fitness.
  2. In this study, I used the cabbage butterfly (Pieris rapae) as a model system to investigate how larval nutrition influences male fitness through development time, adult size and latent effects. Specifically, I examined how variation in dietary nitrogen impacts male fitness by measuring a comprehensive suite of pre- and postcopulatory traits.
  3. Results indicate a complex role for larval nutrition in determining adult fitness. Larval nutritional constraints influence adult fitness through three different pathways: development time, adult size and latent effects. Longer development times were associated with shorter adult life span and reduced male mating success. Body size was positively correlated with traits related to postcopulatory fitness, such as spermatophore size and protein content. Larval nitrogen availability, independent of the developmental traits, also affected traits associated with male mating success such as wing coloration and latency to first mating.
  4. These results provide new insights into how larval nitrogen availability alters adult fitness by revealing novel links between larval dietary nitrogen and various adult fitness components.

Introduction

Animals and plants grow and reproduce surrounded by nutritional variation, where food is often scarce or key nutrients are lacking. Because the juvenile nutritional environment has major effects on the adult phenotype, linking nutrition and fitness is an increasingly important aspect of ecology, evolution and life-history theories (e.g. Boggs 2009; Raubenheimer, Simpson & Mayntz 2009; Morehouse et al. 2010).

Developing under nutritional constraints can be particularly challenging for animals with a complex life cycle, such as amphibians, holometabolous insects and many marine invertebrates. Such creatures are characterized by discrete larval and adult stages that often live in distinct habitats with different nutritional conditions. Thus, individuals have limited time to acquire the resources that are necessary to form the adult and to reproduce. Traditionally, life-history models propose that the effects of larval nutrition on adult fitness are mainly captured by two developmental traits: size and timing of transition out of the larval stage (e.g. Rowe & Ludwig 1991; Moran 1994; Awmack & Leather 2002; Roff 2002). Typically, nutrient limitation produces longer development time and/or smaller body size at metamorphosis, and these changes in developmental traits are correlated with reduced adult fitness. However, recent studies challenge these models by showing that larval nutrition also impacts adult fitness independently of these developmental traits (e.g. De Block & Stoks 2005; Stoks, De Block & McPeek 2006; Block & Stoks 2008). This occurs through ‘carry over’ or latent effects, in which traits originate from the larval nutritional experience and yet are only expressed in adulthood (Pechenick, Wendt & Jarrett 1998). Examples of latent effects due to restricted larval diets include decreased adult immune function (Fellous & Lazzaro 2010), adult shortage of energy storage molecules (Stoks, De Block & McPeek 2006) and increased oxidative stress (Block & Stoks 2008). Thus, by taking into account these additional latent effects, we can achieve a more complete understanding of how early nutrition might affect various components of adult fitness.

A second important issue when examining how juvenile nutrition alters adult fitness concerns the methods used to evaluate fitness. Measuring fitness is a challenging enterprise, especially because male fitness often depends not only on mating success, but also on different pre- and postcopulatory fitness components (Hughes 1998; Fedina & Lewis 2008). When resources are limited, fitness variation among individuals arises through different resource allocation patterns among various fitness components (Stearns 1992). However, the vast majority of studies looking at the fitness effects of larval nutrition have only focused on traits related to either precopulatory or postcopulatory fitness; remarkably few studies have considered both (but see Lewis, Sasaki & Miyatake 2011; Lewis et al. 2012). A more complete understanding of how nutrition affects fitness requires taking into account both pre- and postcopulatory fitness. An additional challenge is that fitness is ideally measured under natural conditions where keeping track of individuals is often a difficult task. Therefore, it is by considering both pre- and postcopulatory fitness components and by measuring these components in settings reflecting natural conditions that we obtain more precise estimates of fitness.

Herbivorous insects are an appropriate study system for examining the linkages between nutrition and fitness. Nitrogen is a key nutrient for all animal species, because it is required to build proteins, nucleic acids and many essential body structures (Mattson 1980; Bernays & Chapman 1994). However, because plant tissue contains only a small fraction of the nitrogen contained in animal tissue, nitrogen becomes a limiting element for most herbivores (Mattson 1980; Scriber & Slansky 1981; White 1984; Slansky & Rodriguez 1987; Bernays 1998; Awmack & Leather 2002). This results in a fundamental nutritional mismatch between herbivores and their food plants; caught in what has been called the herbivore's dilemma (Pierce & Berry 2011), these creatures must reconcile their nitrogen-rich lifestyle with their nitrogen-poor diet. The cabbage butterfly, Pieris rapae, is a model organism for studying the role of nitrogen limitation because of its particularly high demands for nitrogen, with adult bodies consisting of ~13% nitrogen at eclosion (Morehouse & Rutowski 2010a). This butterfly is also useful for testing how larval nutrition affects adult fitness because several male traits related to pre- and postcopulatory fitness components have previously been identified (e.g. Suzuki et al. 1977; Bissoondath & Wiklund 1996; Wedell & Cook 1999; Morehouse & Rutowski 2010b). Finally, Morehouse & Rutowski (2010a) found that larval nitrogen availability strongly affected larval growth and development. These authors also suggested that dietary nitrogen influences key male adult traits including male ornaments such as wing coloration and nutritive nuptial gifts that are passed to the females during copulation (Morehouse 2009; Morehouse & Rutowski 2010a). In this species, females prefer to mate with more colourful males (Morehouse & Rutowski 2010b) and wing coloration is based on pterins, a group of nitrogen-rich pigments (Kayser 1985). Pieris rapae males' nuptial gift consists of a protein-rich spermatophore (Bissoondath & Wiklund 1995) that increases male reproductive success by increasing female fecundity (Watanabe & Ando 1993), remating latency (Sugawara 1979; Kandori & Ohsaki 1996) and male paternity share (Wedell & Cook 1998).

In this study, I adopt a holistic approach to address the long-standing question of how adult fitness of an herbivorous insect is affected by the larval nutritional environment. I manipulated larval dietary nitrogen availability and measured its effects on adult fitness through three different pathways: development time, adult size and latent effects. In addition, I measured a suite of adult male fitness components in a setting that approximates natural conditions for the cabbage butterfly. Using this approach, I expected to gain a more comprehensive picture of how dietary nitrogen influenced male adult fitness in an organism with a complex life cycle.

Materials and methods

Dietary nitrogen manipulation

Nitrogen is a key limiting nutrient in herbivores, such as butterflies, and variation in the availability of this nutrient is expected to be tightly linked with variation in organismal fitness. Host plants of the cabbage butterfly naturally vary in nitrogen content ranging from 1·9% to 4·8% nitrogen (Slansky & Feeny 1977; Morehouse & Rutowski 2010a). In this experiment, I manipulated the level of dietary nitrogen using a semi-synthetic diet following methods used in Morehouse & Rutowski (2010a). Briefly, I manipulated the levels of nitrogen by replacing vitamin-free casein, the primary source of protein in this diet, with equivalent amounts of cellulose, which is an non-nutritive and nontoxic filler. This manipulation altered nitrogen concentration, while leaving all the other nutrients constant. Seven semi-synthetic diets were prepared with different nitrogen percentages that represented the natural nitrogen range, including one ‘standard’ diet (3·5%) and six ‘treatment’ diets with percentage nitrogen by dry mass of 2·0, 2·4, 2·8, 3·2, 3·7 and 4·1%.

All larvae were reared from hatching to 10 days (late 2 or early 3 instar) on the standard diet (3·5% nitrogen), at which time experimental males were switched to one of the six treatment diets. Females used in this experiment were reared on diets with similar nitrogen levels (3·2, 3·5 or 3·7%).

Experimental individuals

Butterflies in this experiment were the F1 offspring of nine singly mated females reared from eggs obtained commercially (Carolina Biological Supply Company, Burlington, NC, USA). For the goals of this experiment, I needed to rear larvae from different nitrogen diets and yet get all adults to emerge at approximately the same time. Because decreasing nitrogen increases larval development time (Morehouse & Rutowski 2010a; Tigreros unpublished), I synchronized adult emergence by starting individuals assigned to the lowest nitrogen diets first. The offspring from at least three different females were assigned to each diet treatment and each female contributed eggs to at least two diet treatments. Females' first batch of eggs was assigned to several different (low- and high-nitrogen) diets.

Individuals were kept from egg to adult in individual petri dishes (Fisher Scientific, Houston, Texas) stored in an incubator with an 18-L/6-D photoperiod and 25 °C:20 °C at 60% relative humidity. Larvae were provided with cubes of diet which were replaced every 4 days. All adults were fed with a 25% honey solution and stored at ~8 °C until the mating trial.

Measured variables

Developmental traits

I measured development time as the number of days from egg hatching to pupation. Pupal size was quantified as wet mass measured during the 3 day after pupation.

Adult phenotypic traits

I measured aspects of the adult phenotype known to correlate with different fitness components. These included wing size and coloration, both important in courtship and mate acquisition and testis size, important in sperm production. Because these traits may change over an individual's lifetime, I measured these in a subset of adults that were frozen immediately after emergence. To measure wing size, I photographed the right fore- and hind wings and measured their area using ImageJ 1.45 software. Pterin-based wing coloration was measured according to Morehouse & Rutowski (2010b) protocols. Briefly, I measured the spectral reflectance of a 2-mm area on the dorsal left forewing using a JAZ Ocean Optics spectrophotometer. From the measured spectra, I calculated four parameters that have been shown to be relevant to both pterin coloration and female vision (Morehouse & Rutowski 2010b): R300–375, R550–650, λR50, βR50. These four parameters were reduced using a principal component analysis into one single variable, PC1: this first principal component explained 99·9% of the variation in pterin-based wing coloration.

To measure testis size, both testes were dissected, dried and weighed to the nearest 0·1 μg using a Mettler Toledo MT5 microbalance (Columbus, OH, USA).

Male mating success

One important male fitness component is the ability to attract and successfully mate with females. Male mating success should ideally be measured in natural or seminatural conditions where male–male competition and female rejection and acceptance behaviours can be freely displayed. I measured male mating success in such conditions by releasing adult butterflies into an outdoor insectary (20 m long, 3 m height, 3 m wide) large enough to keep a natural adult density (~1·7 individuals m−3). Within this enclosure, adults could fly, court and reject potential mates. A similar approach used in other studies to measure adult fitness in both butterflies (Bergman, Gotthard & Wiklund 2011) and damselflies (De Block & Stoks 2005). The insectary was located on the Tufts University campus where flying adults of P. rapae are normally seen. During the trial, temperatures ranged from 27 °C to 35 °C, and butterflies were provided with cabbage (host plant), damp sponges and flower bouquets.

A total of 37 females and 68 males (6 to 17 males per diet treatment) were simultaneously released in the outdoor insectary on July 2010. Prior to their release, males were individually marked by placing four coloured dots (using Sharpie Extra Fine markers) near the base of the forewing on the costal margin area. These marks were very small (~0·8 mm diameter), and microscopic examination showed they did not damage wing scales. Although the number of males decreased over the course of the experiment, the number of females in the insectary was also changed to maintain a 2 : 1 sex ratio. This sex ratio matches that typically found in the beginning of the mating season, when males are most competitive and when most virgin females are found (Wiklund, Wickman & Nylin 1992).

The insectary was continuously patrolled from 06:30 h to 17:30 h, these butterflies' active mating period, for a total of 5 days. During this period of continuous observation, I measured latency to first mating and mating success for each male by recording the date and time when each male mated. Mating pairs were collected in plastic cups, male ID was determined, and the male was released back into the insectary after the end of copulation. Mated females were sacrificed for later dissection and measurement of the male's spermatophore; each female removed was replaced with a new virgin female.

Spermatophore quality

Like other Lepidoptera, male P. rapae transfer a spermatophore during each mating (Wedell 2005). This spermatophore is a package that contains sperm as well as significant amounts of protein and other nutrients (Marshall 1982). Spermatophore quality, in terms of size, protein content and sperm number, is an important determinant of a male's postcopulatory fitness in pierid butterflies (e.g. Boggs & Watt 1981; Bissoondath & Wiklund 1995; Wedell & Cook 1998). In P. rapae, large spermatophores increase female egg laying (Watanabe & Ando 1993), female remating latency (Sugawara 1979; Kandori & Ohsaki 1996) and increase male paternity share in competitive mating situations (Wedell & Cook 1998). In this study, I measured wet mass, protein per cent and sperm number contained within a male's first and second spermatophores. To do this, mated females were dissected within 20 min of finishing copulation, before sperm began to migrate out of the spermatophore (Cook & Wedell 1996); female bursas containing the spermatophore were stored individually in 80% ethanol for later analysis. Later, male spermatophores were carefully dissected out, and their size determined as wet mass measured to 1.0-ug accuracy (Mettler Toledo MT5 microbalance). Within the spermatophore the sperm ampulla consists of a small compartment containing two types of sperm: eupyrene (nucleated) and apyrene (anucleate sperm) sperm. In this study, I counted eupyrene sperm by first carefully separating the sperm ampulla from the rest of the spermatophore, opening the ampulla and counting eupyrene sperm bundles at 40× magnification; the number of bundles was multiplied by 256, the total number of sperm contained in a bundle, to obtain the total number of eupyrene sperm (Cook & Wedell 1996). The remaining part of the spermatophore (without sperm) was used for protein analysis using the Bradford method (Bradford 1976) with a BioRad protein assay (Bio-Rad Laboratories, Hercules, CA, USA) following methods in Bissoondath & Wiklund (1995).

Adult life span

Pieris rapae adults live for about 2–3 weeks. Adult life span was measured for a period of 12 days, starting on the day when males were first released into the insectary. Prior to this time, emerging adults were kept at ~8 °C to slow metabolism. Males that remained alive after the observation period in the insectary were kept in an incubator (18-L/6-D photoperiod and 25 °C: 20 °C at 60% relative humidity) to record survival over seven additional days. After that time, remaining males were sacrificed and their life spans included as truncated data.

Data analysis

To examine the effects of dietary nitrogen on development time and pupal size, I used two separate linear regressions for development time and pupal size. Then, I used a series of models to estimate the independent effects of dietary nitrogen, developmental time and pupal size on each adult trait.

The analysis of each model consisted of first using a generalized additive model (Proc GAM, SAS ver. 9.1, Cary, NC, USA) to identify the appropriate form of dependence by applying nonparametric regression and smoothing techniques. Second, each model was analysed using a generalized mixed linear model (GLIMMIX macro in SAS ver. 9.1) using the particular form (linear, quadratic, etc.) suggested by the GAM procedure. Results are presented from the GLIMMIX analysis that controls for possible nonindependence of males within families by including individual males as the random factor blocked by family.

The distribution specified in the GLIMMIX models for wing size, wing colour, testis size, life span and latency to first mating was normal. A Poisson distribution was used in GLIMMIX for mating success. Because life span included truncated data, I used the Cox proportional hazards model (PHREG procedure, SAS ver. 9.1) that allows inclusion of censored data.

Results

Effect of dietary nitrogen on developmental traits

Larval dietary nitrogen level affected both development time and pupal size of P. rapae males. Increased nitrogen concentration resulted in significantly shorter development time (Fig. 1a; linear regression, F = 23·28 P < 0·0001, N = 220) and larger pupal size (Fig. 1b; F = 19·5 P < 0·0001, N = 220). The effect of dietary nitrogen was stronger for development time (R² = 0·41) than for pupal mass (R² = 0·17).

Figure 1.

Effect of manipulating nitrogen concentration in the larval semisynthetic diet of male P. rapae butterflies on developmental traits: (a) Development time (days from egg hatching to pupation). (b) Pupal size (mass at 3rd day after pupation).

Adult morphological traits

Dietary nitrogen concentration, independent of pupal size and development time did not affect either male fore-wing area (GLIMMIX, F1, 44 = 3·01, P = 0·09) or hind-wing area (F1, 54 = 0·91, P = 0·34). Also, testes size was not affected by larval nitrogen concentrations (F1, 42 = 2·31, P = 0·14). In contrast, adult wing coloration was significantly altered by nitrogen concentration (Figs 2 and 3a; F1, 61 = 30·10, P < 0·0001) independent of the developmental traits. Adult morphological traits covaried with variation in pupal mass (Fig. 2); males with greater pupal mass had larger fore-wing areas (F1, 44 = 178·17, P < 0·0001), hind-wing areas (F1, 52 = 181·93, P < 0·0001) and larger testes sizes (F1, 42 = 24·76, P < 0·0001). Male pupal mass was also positively related to wing coloration (F1, 59 = 14·79, P = 0·0003).

Figure 2.

Summary of how dietary nitrogen during the larval stage of P. rapae males influenced adult phenotypic traits and different components of adult fitness both through and independent of development time and pupal size. Line types indicate significant positive (–––), negative (---) or quadratic (………) relationships.

Figure 3.

Effects of dietary nitrogen on different fitness-related traits of P. rapae males. (a) Partial regression plot showing the effect of nitrogen, independent of development time and pupal size, on wing colour (first principal component explaining 99% of variation). (b) Partial regression plot showing independent effect of nitrogen on latency to first mating. (c) Quadratic effect of nitrogen on mating success achieved by each male. Bars represent mean (±1 SE) number of matings for males reared on each nitrogen level. Numbers at the top of bars represent sample size.

On the other hand, variation in development time did not influence adult phenotypic traits (fore-wing area F1, 44 = 0·01, P = 0·91; hind-wing area: F1, 54 = 2·71, P = 0·10; testis size: F1, 42 = 0·62, P = 0·44; wing colour F1, 59 = 1·18, P = 0·28).

Adult fitness traits

Male mating success

During the 5 days of continuous observation, 50 of 68 males mated at least once. For these males, the time between when they were released into the mating arena and the first mating (latency to first mating) ranged from 15 min to 1·5 days, and this variation was significantly related to dietary nitrogen concentration (Fig. 2, F1,38 = 5·41, P = 0·02). Surprisingly, males reared on lower nitrogen diets achieved their first mating sooner than those reared on higher nitrogen diets (Fig. 3b). Yet, male latency to first mating was not affected by either of the developmental traits (pupal mass: F1,38 = 1·08, P = 0·31; development time: F1,38 = 0·01, P = 0·94).

Overall, P. rapae males averaged 1·8 ± 0·21 matings (median = 1·5) during the 5 days of observation, with a range from 0 to 7. Interestingly, there appeared to be an almost significant quadratic effect of dietary nitrogen concentration on total number of matings, with males reared on the low- and high-nitrogen diets both averaging more matings per male (Figs 2 and 3c; F1, 30·13 = 4·05, P = 0·0502). Total number of matings was also influenced by variation in development time (Fig. 2): males developing more quickly generally mated more often than those with longer development times (Fig. 4a; F1, 42·35 = 11·25, P = 0·002). However, there was no effect of pupal mass on total mating number (F1, 56·63 = 0·81, P = 0·37).

Figure 4.

Partial regression plots illustrating the independent effect of development time (days from egg hatching to pupation) on male adult fitness-related traits in P. rapae. (a) Effect of development time on total number of matings (b) Effect of development time on first spermatophore protein content (per cent of total spermatophore wet mass) (c) Effect of development time on adult life span.

Spermatophore quality

During their first mating, P. rapae males transferred spermatophores that weighed 4·39 ± 0·13 mg (N = 48, range 1·73–5·85 mg wet mass). During second matings, spermatophore weight declined significantly to 1·73 ± 0·17 mg (n = 30, range 0·4–3·9 mg) (Paired t-test, t = 16·15, P < 0·0005, N = 27). Spermatophore protein content expressed as per cent of total wet mass also decreased from 1·65 ± 0·16% in first spermatophores to 0·3 ± 0·07% in second spermatophores (Paired t-test, t = 6·96, P < 0·0005, N = 27). However, no significant difference was detected in number of eupyrene sperm transferred in first (11 620 ± 1071) compared with second spermatophores (12 604 ± 1058) (Paired t-test, t = −0·3, P = 0·7, N = 12).

Male spermatophore mass was not affected by either dietary nitrogen (first spermatophore: F1,36 = 0·77, P = 0·39; second spermatophore: F1, 21 = 1·19, P = 0·29) or development time (first spermatophore: F1,36 = 0·61, P = 0·44; second spermatophore: F1,21 = 2·83, P = 0·11). However, males with large pupal size produced significantly larger spermatophores (Figs 2 and 5a; first spermatophore: F1,36 = 39·96, P < 0·0001; second spermatophore: F1,21 = 9·41, P = 0·006).

Figure 5.

Partial regression plots showing independent effect (a) of pupal size (mass at 3rd day of pupation) on size of first and second spermatophore produced by P. rapae males and (b) dietary nitrogen on number of sperm contained within second spermatophore.

The protein concentration of male spermatophores was not altered by either dietary nitrogen (first spermatophore: F1,36 = 0·81, P = 0·37; second spermatophore: F1, 20 = 1·09, P = 0·31) or by pupal mass (first spermatophore: F1,36 = 3·24, P = 0·08; second spermatophore: F1,20 = 0·78, P = 0·39). Surprisingly, males with longer development times produced first spermatophores with higher protein content than did males with shorter development times (Figs 2 and 4b; F1,36 = 5·17, P = 0·03). However, development time was not significantly related to protein concentration of males' second spermatophores (F1,20 = 0·85, P = 0·37).

In addition, there was a nonsignificant trend for males reared in low-nitrogen diets to produce second spermatophores containing fewer sperm (Figs 2 and 5b; F1,13 = 4·39, P = 0·056). Nevertheless, even with a larger sample size, no effect of nitrogen was detected for sperm content of first spermatophores (F1,23 = 1·46, P = 0·24). Neither of the developmental traits was correlated with number of sperm contained within males' first or second spermatophores. Development time did not correlate with sperm number within male's first (F1,13 = 0·35, P = 0·5) or second spermatophores (F1,23 = 0·01, P = 0·97) nor was pupal size correlated with sperm numbers (first spermatophore: F1,13 = 2·17, P = 0·16; second spermatophore: F1,23 = 2·11, P = 0·16).

Adult life span

Interestingly, development time influenced adult male life span (Fig. 4c; PHREG: x² = 8·68, d.f. = 1 P = 0·003): males with longer larval development times had significantly shorter adult life spans. However, adult life span was not affected by either larval dietary nitrogen (x² = 2·31, d.f. = 1, P = 0·13) or pupal size (x² = 0·74, d.f. = 1, P = 0·39).

Discussion

Nutritional constraints early in life are predicted to have important consequences for adult fitness. This study offers one of the most complete views of how larval nutrition affects fitness by including a comprehensive suite of pre- and postcopulatory traits and by examining the effects of larval diet acting not only through key developmental traits but also independently of them. The main findings are discussed below.

Effect of nitrogen on developmental traits

As expected based on previous studies (Wolfson 1982; Chen et al. 2004; Hwang, Liu & Shen 2008; Morehouse & Rutowski 2010a), decreasing dietary nitrogen significantly lengthened development time. Nitrogen also affected pupal mass but to a lesser degree, which may explain why pupal mass has been often found to be unaffected by nitrogen (Chen et al. 2004; Morehouse & Rutowski 2010a). Thus, these results suggest that when larval nitrogen availability is reduced, P. rapae males tend to increase their development time and maintain their body size.

Independent effects of nitrogen on adult fitness

Larval dietary nitrogen, independent of development time and pupal mass, had significant effects on adult phenotypic traits and fitness of P. rapae males. First, dietary nitrogen was positively related to a male's sexual signal, wing coloration, supporting the hypothesis that dietary nitrogen constrains male allocation to nitrogen-rich pterin pigments (Morehouse & Rutowski 2010b). Morehouse (2009) used a similar diet manipulation to examine the effects of genotype and dietary nitrogen in P. rapae but did not detect a main effect of nitrogen on male wing coloration. Instead, they showed that wing coloration was affected by the interaction between diet and genotype. Thus, these two studies illustrate the importance of dietary nitrogen availability in enhancing male wing coloration in some genotypes of P. rapae butterflies.

Results of this study showing that dietary nitrogen alters male wing coloration suggest that this male trait may provide information on the quality of a male's nutritional environment and therefore signal male phenotypic condition. Previous studies have proposed that male wing coloration in the Pieridae may be an honest signal of direct benefits such as spermatophore quality (Kemp & Rutowski 2007; Morehouse 2009). If this is true, both the signal and the benefit would be predicted to vary as a function of diet. Although I found that wing colour reflected larval nutritional status, most measured aspects of spermatophore quality were unaltered by larval diet. In this study, wing coloration and spermatophore quality were assessed in different individuals (sib-related); thus, I cannot directly assess the relationship between these two traits. Studies directly assessing the relationship between male wing coloration and spermatophore quality are necessary to gain more insight into the evolution of wing coloration as an honest signal of direct benefits.

These results also revealed that larval dietary nitrogen influenced male mating success in two distinct ways. First, males reared on low-nitrogen diet tended to mate sooner than those in the nitrogen-rich diets. For species such as P. rapae with a short adult life span (Moreau, Benrey & Thiéry 2006) and a high risk of sperm competition (Simmons et al. 1994), how soon a male begins mating has important fitness consequences. A short latency to first mating might increase both a male's chance of mating at all before death and his chance of mating with a virgin female, thus avoiding sperm competition with a female's previous mates. Low-nitrogen diet fed males might be able to achieve a short latency to first mating by being less choosy or having a faster sperm production rate than males reared in high-nitrogen diets. Thus, by decreasing their mating latency, low-quality males might compensate for disadvantages in other fitness-related traits such as being less attractive or producing few sperm. This study also suggests that dietary nitrogen increases mating success of both low- and high-nitrogen diet fed males, which highlights the importance of examining diet effects using a diet gradient that allows detection of nonlinear effects. High mating success is expected for males in high-nitrogen diets, which are likely to be the most attractive males (e.g. in terms of wing coloration). For example, Morehouse & Rutowski (2010b) showed that P. rapae females prefer to mate with more colourful males. Conversely, the high mating success of males in low-nitrogen diets is a surprising result, indicating that males reared under nitrogen constraints somehow buffer the effects of limited nitrogen. A possible explanation for this result is that these low-nitrogen diet fed males achieve high mating success by being less choosy. In P. rapae, not only females but also males exhibit mate choice (Obara, Koshitaka & Arikawa 2008; Obara et al. 2008; Tigreros unpublished). In some species, the degree of choosiness is variable, with low-quality individuals being less choosy than high-quality individuals (e.g. Holveck, Geberzahn & Riebel 2011). This idea that males reared on low-nitrogen diets are less choosy is also supported by the finding that these males also initiated mating sooner than did males reared on high-nitrogen diets. In short, while high-nitrogen diet fed males may increase their mating success by being more attractive to females, low-quality males may increase their mating success by being less choosy.

Effects of developmental traits on adult fitness

Development time, which was strongly affected by variation in dietary nitrogen, influenced several components of male fitness: total number of matings, protein content of a male's first spermatophore and adult life span. Conventionally, short development times have been thought to provide a fitness advantage by reducing the risk of death before reproduction (Sibly & Calow 1986; Stearns 1992). Nevertheless, in this study, I detected additional fitness effects of development time that have not been traditionally considered. First, I found that shorter larval development times were associated with longer male adult life span and higher mating success. Although a connection between time spent as larvae and adult life span has been rarely considered, this result is congruent with the genetic trade-off between larval development time and adult life span shown in another lepidopteran, the butterfly Bicyclus anynana (Pijpe et al. 2006). The relationship between larval development time and mating success is probably a consequence of males living longer as adults and therefore having more chances to mate. On the other hand, males that took a long time to develop allocated greater amounts of protein into their first spermatophore than those males with shorter development. In the cabbage butterfly, as in many other species, males can increase their own reproductive fitness by transferring a spermatophore that enhances their mate's reproductive output (Watanabe & Ando 1993). As protein is an essential component of oogenesis (Chapman 1998), a high investment in spermatophore protein might increase reproductive fitness. Taken together, these results suggest two alternative strategies linked with larval development time through which males could increase their reproductive success. On the one hand, males with rapid larval development will live longer as adults and thus will acquire more matings. On the other hand, males with slow larval development, although they have short adult life spans, will produce high-quality spermatophores (high protein concentrations) in their first mating.

Male body size is often considered to impact precopulatory success, with larger males benefitting through female choice or male-male competition. Although several aspects of male phenotype which could potentially influence mate acquisition ability (e.g. wing area and wing colour) were found in this study to be influenced by pupal size, pupal size did not directly alter male mating success (latency to first mating or number of matings). Nevertheless, variation in pupal size did influence traits that are linked with a male's postcopulatory success. As found in other studies (Bissoondath & Wiklund 1996; Wedell & Cook 1998), males with larger pupal size produced larger spermatophores that contained more protein. These two components are important in the male's postcopulatory performance in several ways. First, larger spermatophores increase female refractory periods before remating (Sugawara 1979; Kandori & Ohsaki 1996). Second, large and protein-rich spermatophores might increase female reproductive output (Watanabe & Ando 1993) and third, males with large spermatophore are able to sire more offspring even when their sperm is competing with other males' sperm (Wedell & Cook 1998). This study supports the contention that variation in body size due to nutritional constraints might play an important role not only in the outcome of precopulatory but also postcopulatory processes (e.g. McGraw et al. 2007).

In summary, while a large body of literature shows that larval nutrition is an important determinant of adult fitness, here I show that the role of nutrition is considerably more complex than what has been commonly thought. First, as pointed out by De Block & Stoks (2005), larval constraints are not entirely captured by the developmental traits of development time and size at pupation. Larval nutrition influenced adult fitness through three different pathways: by altering time to metamorphosis, by altering pupal size (size at transition) and through nutritional latent effects carried over into the adult independently of time and size at metamorphosis. Second, the impact of larval diet on adult fitness (through any of the mentioned pathways) was neither entirely positive nor negative. Instead, some fitness components were positively and others were negatively related to larval nutrition, which is consistent with the idea that organisms can buffer the effects of nutritional constraints by changing allocation to the different fitness components (Stearns 1992). Although this study focused on nitrogen, long considered the key limiting nutrient for herbivores (Mattson 1980; Scriber & Slansky 1981; White 1984; Slansky & Rodriguez 1987; Bernays 1998; Awmack & Leather 2002), other nutrients, like phosphorus, might also influence herbivorous fitness (e.g. Huberty & Denno 2006; Apple et al. 2009). Thus, further progress on understanding the link between larval nutrition with adult fitness may benefit from empirical studies considering latent effects when multiple nutrients are limited.

Acknowledgements

I would like to thank Sara Lewis for her invaluable advice on the design and performance of this work as well as her insightful comments on the present manuscript. I would also like to thank Nathan Morehouse for his guidance through the wing colour spectrometry measurements and comments on this manuscript. Thanks to Francie Chew for helping me to understand the biology and rearing of Pieris rapae. Thanks to Durwood Marshal for statistical advice. Thanks to Emma Sass and Wilson Acuna for all their hard work collecting some of the data presented in this paper. Finally, I would like to thank Sue Bertram and an anonymous reviewer for helpful comments on an earlier draft. This work was also possible thanks to a Sigma Xi and Tufts Graduate Student Research Awards.

Ancillary