Brooke L. W. Miller, Department of Ecology and Evolutionary Biology, University of California at Santa Cruz, Santa Cruz, CA 95064, USA. Tel.: +1 831 227 7770; fax: +1 831 459 5353; e-mail: email@example.com
Physiological trade-offs between life-history traits can constrain natural selection and maintain genetic variation in the face of selection, thereby shaping evolutionary trajectories. This study examines physiological trade-offs in simultaneously hermaphroditic banana slugs, Ariolimax dolichophallus. These slugs have high heritable variation in body size, which strongly predicts the number of clutches laid, hatching success and progeny growth rate. These fitness components were associated, but only when examined in correlation with body size. Body size mediated these apparent trade-offs in a continuum where small animals produced rapidly growing progeny, intermediate-sized animals laid many clutches and large animals had high hatching success. This study uses a novel statistical method in which the components of fitness are analysed in a mancova and related to a common covariate, body size, which has high heritability. The mancova reveals physiological trade-offs among the components of fitness that were previously masked by high variation in body size.
Physiological trade-offs between life-history traits can constrain natural selection and thereby strongly shape evolutionary trajectories (Zera & Harshman, 2001). Life-history traits, such as fecundity, egg mass, age of first reproduction or longevity, are frequently negatively correlated with each other (Fisher, 1930). When this occurs, positive selection for one trait will result in negative selection for another (Smith & Fretwell, 1974; Reznick, 1985; Sinervo & Licht, 1991). These physiological trade-offs often result from resource limitation, where energetic restrictions mean that an increase in allocation to one trait must come at the expense of another (Daan & Tinbergen, 1997; Zera & Harshman, 2001). Physiological trade-offs create a framework which constrains natural selection, as selection can only act within the limits of the trade-off (Reznick, 1985; Stearns, 1992; Roff, 2002). While constraining natural selection in this respect, physiological trade-offs may also act to maintain the genetic variability that natural selection depends upon through antagonistic pleiotropy. Under this model, genetic variation is maintained because selection on one trait is balanced by antagonistic selection on another trait, preventing fixation in either direction. Thus, physiological trade-offs are important in evolutionary biology, as they have been implicated as a possible mechanism to maintain genetic variation (Roff, 1997).
Roff (2002) illustrates the importance of considering multiple traits in the study of life-history evolution, by noting that a trade-off can be obscured if important covariates are ignored. For example, he notes that if one were to look for a trade-off between egg size and fecundity, the trade-off might be missed if both are positively correlated with body size, and body size has large variance. Because of difficulties arising from confounding factors, such as covariance with other traits or environmental variability, studying physiological trade-offs can be difficult in field studies without carefully controlled manipulations (Van Noordwijk & De Jong, 1986; Sinervo et al., 1992; Doughty & Shine, 1998; Zera & Harshman, 2001). Laboratory studies have been powerful tools for studying trade-offs because of their ability to reduce confounding factors and control for variation in the traits involved in a particular trade-off (Chippindale et al., 1997; Zera & Harshman, 2001).
This paper reports on a laboratory study of physiological trade-offs of life-history traits in a simultaneous hermaphroditic banana slug, Ariolimax dolichophallus. Simultaneous hermaphrodites are capable of producing both eggs and sperm at the same time. Although numerous studies have examined trade-offs between allocation to male and female function in hermaphrodites (e.g. Greeff & Michiels, 1999; Locher & Baur, 2000; Scharer & Wedekind, 2001; Angeloni, 2003; Scharer et al., 2005), few studies have examined trade-offs in other important life-history traits, such as those involved in egg production (but see, Locher & Baur, 2002). Ariolimax dolichophallus exhibit large variation in maximum body size, up to a fivefold difference among individuals in the same population when reared under identical laboratory conditions. This study finds that body size is also highly heritable. Body size is often under strong natural selection (Bumpus, 1899), which erodes genetic variability, yet, in this system, large amounts of variation in body size remains, when assayed in a laboratory environment. This paper explores physiological trade-offs of life-history traits involved in egg production that could act as a mechanism for the maintenance of variation in body size.
Materials and methods
Banana slugs in the genus Ariolimax (Stylommotophora: Arionidae) are simultaneous hermaphrodites that engage in both sexual roles during mating (Leonard et al., 2002) and produce eggs and sperm from a single gonad (Mead, 1942). They are the largest native terrestrial slug in North America (Mead, 1942; Gordon, 1994). The genus is widely distributed from the southern end of Alaska through San Diego, CA, and is also found on some of the coastal islands off southern California (Pilsbry, 1948; Groves, 1992).
Ariolimax dolichophallus Mead, is restricted to the southern San Francisco Peninsula in California. These slugs use their spermatheca for long-term sperm storage (Mead, 1942). Although capable of self-fertilization, isolated virgins rarely produce viable clutches in the laboratory (B.L.W. Miller, unpublished data). Laboratory studies have found that A. dolichophallus reach sexual maturity in 13–18 months (Mead, 1942; Leonard et al., 2002). Under controlled laboratory conditions, they exhibit a broad range of body sizes, with some individuals reaching maximum body masses of over 170 g and others never growing larger than 30–40 g, yet attaining sexual maturity and producing fertilized eggs (B.L.W. Miller, unpublished data).
Animal collection and husbandry
In March 2002, we collected 100 slugs with various body masses along Bonny Doon Road in Bonny Doon, CA (n = 100, mean body mass 34.02 g, median 34.05, SD 15.23, SE 1.52). Animals were taken to the laboratory, weighed and placed in individual terraria (5.7-L Sterilite® plastic shoeboxes; Sterilite Corp., Townsend, MA, USA) with potting soil and food ad libitum. Slugs were maintained in a temperature-controlled room between 13 and 17 °C (closely approximating the winter activity temperature of coastal California), and fed a rotating diet of cornmeal, powdered milk, peanut butter, spirulena algae, dog food, oatmeal and paper. Every 1–2 weeks, terraria were cleaned and all animals were weighed and provided with fresh food. Individual animals were isolated during this experiment, so hatchlings were the product of field mating history and sperm storage or self-fertilization. Animals were reared for 22 months.
When clutches were laid, eggs were weighed and transferred to an incubation chamber consisting of an 8-oz. plastic deli container with moistened Kimwipe® tissue as substrate (Kimberly-Clark Corp., Neenah, WI, USA). To specifically test for maternal effects of egg size on progeny growth parameters, we removed a random subset of 5–10 eggs from each clutch and reared them in individual egg cups (1-oz. plastic cups) with Kimwipe® tissue substrate. This test was conducted so that we could definitively know which animals came from which eggs and test for maternal effects arising from egg mass. Additionally, this allowed us to separate out environmental variables because of a common clutch environment. All progeny (n = 216, both from communal incubation chambers and individual egg cups) used in the study were immediately transferred to individual terraria upon hatching and reared in isolation, to minimize common environment effects on post-hatching growth. All egg-to-adult slugs were maintained in the same conditions as the parental slugs and cleaned, weighed and given fresh food ad libitum every 1–2 weeks.
Between November 2002 and September 2003, we collected 99 clutches of eggs from 38 different field-collected animals, which produced one to five viable clutches each. This produced a total of 3123 eggs from which 1288 slugs hatched. Of those hatchlings, 216 progeny from 18 different maternal lineages were reared in isolation for over 22 months and used for heritability estimates and growth data.
All analyses used the maximum body mass achieved during the 22-month study as a measure of size (Mead, 1942). Unless otherwise stated, the data were not transformed, as they did not deviate significantly from a normal distribution, and all analyses were performed using JMP-SAS 5.1.1® statistical software package (SAS Institute Inc., Cary, NC, USA). All tests were two-tailed and critical P-values were set to 0.05. One of the field-collected animals was excluded from the analysis as it was an outlier at more than two standard deviations away from the mean body mass (mean 89.29 g, SD 19.05, mass of outlier 47.75 g), and the animal appeared to be unhealthy.
To determine the heritability (h2) of body size (in the broad sense), we first performed a parent–offspring regression on the average maximum body mass of all adult progeny (averaged across all clutches for the 216 progeny reared for 22 months) against the maximum body mass of the mother. If the progeny result from sexual reproduction, then this regression is a standard single-parent regression and the h2 is estimated from twice the slope of the regression line (Roff, 1997). However, because these slugs are capable of self-fertilization, we also estimated h2 from the slope of the regression line, which would be synonymous with a standard mid-parent regression (Roff, 1997), as the value of the mother would actually represent the values of both parents. These two estimates of heritability gave us a bounded upper and lower values for h2, depending on the frequency of self-fertilization. When self-fertilization is low, h2 is closer to the upper estimate, and when self-fertilization is high, h2 is closer to the lower estimate. We report h2 as a bounded range, as we currently do not definitively know self-fertilization rates in this system, which could be used to weigh heritability estimates for a more accurate estimate of the actual heritability. These estimates of broad sense heritability are confounded by salient maternal effects. However, we tested for the explicit maternal effects of egg mass on the body mass of mature progeny, which is the most common type of maternal effect to influence growth rate and thus body mass (Sinervo, 1991; Bernardo, 1996; Fox et al., 1999). This test does not preclude that unmeasured yolk or egg constituents (e.g. hormones) do not still confound the estimate of broad sense heritability. Genotype × environmental effects can confound estimates of heritability; however, in this study, they were reduced as all individuals were reared under nearly identical conditions, but in isolation from each other.
To test for the effects of maternal body size on fitness components, we regressed variables representing female fitness components against maternal maximum body mass using a simple linear regression model. These included: (1) average egg size; (2) number of eggs per clutch; (3) total clutch mass; (4) the hatching success as the proportion of a clutch that hatched; (5) the total fitness, as measured by the total number of hatchlings produced; and (6) the total number of viable clutches laid in the laboratory over the course of the study (clutches that hatched at least one egg). Total fitness, defined as the total number of hatchlings produced, was not normally distributed; so, we used a square root transformation to normality and then regressed it against maternal maximum body mass.
where t1 and t2 are successive time intervals, measured in days. Growth rates are reported as ln(grams) per day and we plotted them against the log body mass to determine mass-specific growth trajectories (Sinervo & Adolph, 1989). These growth trajectories were negative linear plots, where the Y-intercept estimates the initial growth rate upon hatching and the X-intercept estimates the asymptotic body mass, the body size where net growth is zero. A preliminary analysis revealed that asymptotic body mass strongly correlates with maximum body mass (P < 0.0001, r2 = 0.3512). We also plotted the growth rates against the age of the progeny to construct age-specific growth trajectories, where the X-intercept estimates the age to asymptotic body mass. Because sexual maturity in A. dolichophallus cannot be determined without dissection (Mead, 1942), and these animals were not dissected in this study, we used age to asymptotic body mass as a proxy for time to sexual maturity. We averaged these variables (initial growth rate and age to asymptotic body mass) for all progeny from a single mother and regressed them against maternal maximum body mass. We also regressed all progeny growth parameters against egg mass to test for explicit maternal effects on body size or growth that might be mediated through egg mass (Sinervo, 1990a).
To directly test for trade-offs between physiological traits, we first did regressions between all traits to look for negative relationships. However, correlation with other traits can obscure trade-offs, so we developed a new multivariate test to examine relationships that can be masked by other associations.
When two components of fitness that are traded-off are correlated with a third trait, and there is high variability in that third trait, trade-offs can be masked by the nature of their relationships to the third trait. To reveal trade-offs, we performed a mancova to test if the relationship between each component of fitness and body size was statistically different, which would reveal a trade-off between those traits. In the mancova, the covariate (maternal body mass) was regressed against pairs of traits that were considered to be components of fitness (hatchling growth rate, the number of clutches laid in the laboratory and hatching success). This reveals trade-offs that are masked by covariation with a third trait. For example, if two fitness components are positively correlated with fitness, but vary with body size in an inverse relationship, then increasing body size increases fitness through one trait but decreasing body size increases fitness through the other trait. That inverse relationship means that there is a trade-off between those two components of fitness in that both cannot be maximized simultaneously to increase fitness. Increasing one trait will necessarily lead to a decrease in the other trait, as they are inversely correlated to each other through body size.
The mancova tests the relationships between two traits and how they covary with a third trait. In this study, it was used to ask if the relationships between different fitness components and maternal body size were different, where increasing maternal body size would necessarily increase one trait (and thus, fitness through that trait) and decreasing body size would necessarily increase the other trait (and thus fitness through that other trait). This reveals a trade-off between the two traits, as they cannot be simultaneously increased because of their inverse relationships with maternal body size.
The relationship between the covariate and fitness components need not be simple and linear, rather the covariate body mass might be involved in a quadratic relationship as well. We therefore also included body mass as a linear and quadratic covariate in the mancova when the simple regression analyses (above) indicated it was appropriate.
Of the 216 progeny reared to sexual maturity, there was large variation in maximum body size, ranging from 17 to 145 g (n = 216, mean 80.79, median 77.28, SD 28.85, SE 1.96). Body size was highly heritable, with low-end estimates of = 0.64 if all progeny are the product of self-fertilization, and = 1.28 if all progeny are the product of sexual reproduction (Fig. 1: P = 0.0227, n = 18, r2 = 0.2842). Eight of 216 sexually mature, laboratory-reared, isolated virgins laid clutches during a 9-month period (3.7%), compared with 38 of 100 field-collected animals laying clutches (38%). The extremely low percentage of unmated laboratory animals laying eggs is suggestive of low self-fertilization rates, and additionally, the relatively low percentage of field animals laying eggs is suggestive that many were virgins when originally collected.
There were no significant effects or trends of maternal body size on egg mass (Fig. 2a), number of eggs per clutch (Fig. 2b) or the total number of hatchlings produced (total fitness) (Fig. 2c). Maternal body mass was, however, a strong predictor of hatch success (Fig. 3a: P < 0.0001, n = 37, r2 = 0.3872). The higher hatching success for larger mothers resulted in more total hatchlings per clutch (P = 0.0172, n = 37, r2 = 0.1517, y = −2.5 + 0.18x). Maternal body mass also was a strong predictor of the total number of clutches that a slug laid in the laboratory, but the relationship was quadratic, with intermediate-sized animals laying the most clutches (Fig. 3b: P = 0.0191, n = 37, r2 = 0.1607).
Maternal body mass was a strong predictor of progeny growth trajectories, with progeny from small slugs growing faster initially after hatching (Fig. 3c: P = 0.0386, n = 18, r2 = 0.2408) and taking less time to reach asymptotic body size (Fig. 4: P = 0.0258, n = 18, r2 = 0.2741). There was over 30% difference in growth rate between the progeny from the largest and smallest animals. Progeny from the smallest animals reached asymptotic body size at around 360 days, whereas progeny from the largest animals took approximately 470 days to reach asymptotic body size (Fig. 4).
Egg size and clutch size did not have any effects on any of the growth parameters, and revealed no suggestive trends. There were also no significant relationships or trends among any of the traits when regressed against each other, without considering body mass.
The mancova revealed, however, that there were trade-offs between initial progeny growth rate, the number of viable clutches laid in the laboratory and hatching success. A mancova that simultaneously tested for the difference in slope also revealed a significant interaction among the three components of fitness and maternal maximum body mass (F2,14 = 5.49, P < 0.02), although this test cannot resolve which of the traits is responsible for the trade-off. Instead, we focus on the results from the pairwise analyses, which suggest all three traits are involved (Fig. 3). Figure 3 shows that statistical result for the mancova test between each of those traits and their relationship to maternal maximum body size (P-values reported there are derived from pairwise tests with each of the components paired into a mancova with body mass as the heritable covariate).
The mancova test is best understood by visual inspection. Figure 3 shows that there are three different relationships between the three traits and maternal body size (the heritable covariate). Intuitively, this figure shows that fitness can be increased by maximizing any of the three components of fitness (as all three traits are presumed to increase fitness). However, maximizing one trait would prevent maximization of another. The mancova tells us that the relationships between body mass and the three components of fitness (Fig. 3) are statistically different from one another.
This test was done in a pairwise fashion because a mancova with all three traits is more difficult to visualize, as two of the traits had linear relationships and the third trait had a quadratic one. The mancova revealed that hatchling growth rate and hatching success both covary with maternal body size in opposite, linear ways and both those traits covary with the number of clutches in a significantly different manner (the curvature of the relationships is different between those traits and the number of clutches laid). These results shows that all three traits are traded off with each other, but only when examined in the context of their correlations with maternal body mass.
In this study, A. dolichophallus displayed large variation in body size, with maximum body sizes varying over threefold among individuals of the same population. This variation is highly heritable and is unlikely to be the product of an obvious maternal effect, as the most common maternal effects are mediated through variation in egg size (Sinervo, 1991; Bernardo, 1996; Fox et al., 1999). We found very strong evidence of high heritability through our parent–offspring regression, although we acknowledge that half-sib assessments of heritability are more robust, because of constraints of this study system, we were unable to conduct half-sib analyses at this time. Because there were no effects of egg size on any of the growth parameters and no association between maternal body size and egg size, the observed variation is most likely genetically based and subject to natural selection.
This study reveals an unexpected result that maternal body size does not correlate with egg size, number or total clutch mass, as these are typical relationships found in a variety of taxa, including hermaphrodites (e.g. Kaplan & Salthe, 1979; Reznick & Endler, 1982; DeWitt, 1996; Scharer et al., 2001). Yet, there are strong relationships between maternal body mass and (1) progeny growth rate; (2) number of viable clutches laid; and (3) hatching success. These three traits or putative components of fitness are negatively associated with each other, but only when correlated through body size. The mancova test demonstrates that all the three traits have statistically different relationships to maternal body size, and thus reveals a trade-off between each of these traits. The slope of the relationship, in the case of progeny growth rate and hatching success (the repeated measures of components of fitness) was involved in a significant interaction within individuals (negative in one trait but positive in the other), which is indicative of a trade-off, as these two components presumably enhance lifetime reproductive success. In the case of the quadratic effect on the fitness component, the number of viable clutches, the quadratic was involved in a significant interaction with each of the other traits (tested separately in each case). This is also related to a trade-off, as the number of viable clutches has an intermediate optimum, whereas the other two traits have the highest value at either extreme of body size. A simple interpretation of these mancovas is that there exist three distinct optima for body size, one for each of the traits seen in Fig. 3.
Life-history trade-offs involving progeny production are often seen as ‘quality vs. quantity’ trade-offs (e.g. Sinervo & Svensson, 1998). A typical quality vs. quantity relationship is between egg size and egg number (e.g. Fleming & Gross, 1990; Sinervo, 1990b; Schwarzkopf et al., 1999; Malmqvist et al., 2004). In this study, the quantity trait is not manifest in egg number, but rather in the total number of viable clutches laid during the course of the study. There are two quality trade-offs, but instead of egg size, they come in the form of increased progeny growth rate or increased hatching success. In this study, we did not measure survival effects on progeny, as the goal of this paper was to establish progeny production trade-offs in a controlled laboratory setting.
Maternal body mass is highly heritable and mediates the apparent trade-offs among the number of viable clutches, hatch success and progeny growth rate. Mothers with large body masses have a higher proportion of their clutches hatch and produce more hatchlings per clutch, which translates into higher efficiency of reproduction. This quality trait, however, appears to come at a cost, as they lay fewer viable clutches and their progeny are slow growing. This study did not determine the mechanism for higher hatching success, but it could be an energetic or hormonal supplementation to the eggs that only larger animals are able to provide. Perhaps larger mothers have better sperm storage capabilities, and the effects we see are because of sperm limitation.
Intermediate-sized slugs have lower hatching success and produce fewer hatchlings per clutch, but compensate by producing more clutches of viable eggs. The apparent trade-off associated with enhancing hatch success might be a high energetic or hormonal demand that intermediate-sized animals are not able to fulfil in a single reproductive episode. By laying multiply, animals may be able to recover energy depleted on individual clutches and allocate their energy over subsequent reproductive bouts.
The smallest animals exhibit another apparent trade-off, where these mothers do not lay many clutches and have low hatching success, but their progeny grow more quickly and reach asymptotic body size earlier than slugs from larger mothers. If asymptotic growth is used as a proxy for sexual maturity, then progeny from the smallest mothers reach sexual maturity up to 30% faster than those from the largest mothers. Rapid growth of progeny leads to a shorter generation time and a larger intrinsic rate of increase (r) (Gotelli, 2001). These small animals may lack the energetic stores for either multiple reproductive bouts or increased hatching success, but may have higher metabolic rates that translate into faster progeny growth.
The actual effects of these results on fitness in nature might, in fact, be more extreme than that reported here. For example, the effect of time to sexual maturity could be exaggerated if growth were not constant throughout the year. If animals only grow on average for 6 months of the year, then 270 days would actually translate into 2 years to sexual maturity, while 470 days would actually translate into 3 years to sexual maturity. The same effect would be observed if they grew continuously, but at a lower rate in the field than in the laboratory. This difference in r would be much more extreme than just the directly measured 30% difference in maturity times found in this study. Unfortunately, there are no studies thus far that have examined growth and time to sexual maturity in a field setting for these animals.
Likewise, the actual effect of body size on hatching success might underestimate fitness effects, as they were measured in a benign laboratory environment. In the field, quality effects like these might be exaggerated, as marginal eggs might have an even lower hatching success. Furthermore, this study did not examine hatchling survival, which could be correlated to the higher hatching success if they are linked through increased energetic allocation to the eggs, which would mean that progeny from larger mothers might have higher survival as well.
When each of the traits is regressed against the other traits without consideration of body size, there is no indication of a trade-off. It is only when placed into the context of body mass that the negative associations among the traits are revealed. Because these traits are linked through maternal body mass, the apparent trade-offs can act as a mechanism that maintains variation in body size. This system appears to consist of a continuum of traits and associated body sizes, with small animals and rapid growth rate, intermediate-sized animals and many clutches of viable eggs and large animals with high hatch success. However, the mancova suggests that there may exist three different optima involving the components of fitness that act on additive genetic variation in body size.
Although this study did not examine sex allocation, it could prove useful for future studies in sex allocation theory. Most models of sex change and sex allocation theory in hermaphrodites are based on the assumption that there are size or age advantages for one sex over the other (Ghiselin, 1969; Charnov, 1982). The empirical work in this field has typically considered only sperm production or egg production as a measure of sex allocation and have failed to look at trade-offs between traits related to progeny production. The results of this study find that there is no size advantage in fecundity, egg size or the total number of hatchlings produced. Essentially, there is no discernable advantage of body size in these standard measures of female fitness, and instead, slugs trade-off other fitness components based on body size. This suggests that future studies of trade-offs between male and female functions in simultaneous or sequential hermaphrodites may benefit from consideration of trade-offs between functions within a single sex.
Understanding life-history evolution requires identifying and understanding the life-history trade-offs that form the framework that structures evolutionary trajectories. Trade-offs can lead to the maintenance of genetic variability in a population. This study identifies a new set of apparent trade-offs among multiple traits in a previously unexplored context, the simultaneous hermaphrodite, which is mediated through body size. This species exhibits large variation in body mass, which is likely to be maintained by constraints on selection caused by these antagonistic interactions.
We gratefully acknowledge C. Fernandez, E. Vogel, K. Horjus, J. Weaver, J. Thompson, A. Davis, A. Corl and J. Pearse for help with various drafts of the manuscript. We would also like to gratefully thank M. Mangel and N. Dominy for excellent and thought-provoking discussions. We could not have done this work without the help of some truly excellent laboratory assistants, in particular, we would like to thank S. Dwiggins and J. Reed for their tireless hard work and devotion. I was also greatly benefited by the wonderful assistance of K. Butcher, C. McLaughlin, A. Hughes, M. Weigle, G. Tarbill and E. Klein. Finally, we would like to thank AZM, PK, SC and P. Beckman for their support. We appreciatively acknowledge funding for this research provided by the Western Society of Malacologists (with support from the Santa Barbara Malacological Society, the Southwest Shell Club, the San Diego Shell Club, and the Northern California Malacological Club), Sigma Xi, the Society of Integrative and Comparative Biology, and the National Science Foundation Doctoral Dissertation Improvement Grant no. 0408060 and ongoing NSF grants to B. Sinervo.