PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of elifeeLifeRecent contentAbout eLifeFor authorsSign up for alerts
 
eLife. 2017; 6: e18770.
PMCID: PMC5340529

A fitness trade-off between seasons causes multigenerational cycles in phenotype and population size

Lutz Becks, Reviewing editor
Lutz Becks, Max Planck Institute for Evolutionary Biology, Germany;

Abstract

Although seasonality is widespread and can cause fluctuations in the intensity and direction of natural selection, we have little information about the consequences of seasonal fitness trade-offs for population dynamics. Here we exposed populations of Drosophila melanogaster to repeated seasonal changes in resources across 58 generations and used experimental and mathematical approaches to investigate how viability selection on body size in the non-breeding season could affect demography. We show that opposing seasonal episodes of natural selection on body size interacted with both direct and delayed density dependence to cause populations to undergo predictable multigenerational density cycles. Our results provide evidence that seasonality can set the conditions for life-history trade-offs and density dependence, which can, in turn, interact to cause multigenerational population cycles.

DOI: http://dx.doi.org/10.7554/eLife.18770.001

Research Organism: D. melanogaster

eLife digest

Many wild populations go through long cycles in abundance that span several generations. The traditional explanation for such “multigenerational” cycles is that they are driven by predator/prey relationships, the classic example being oscillations between the numbers of lynx and snowshoe hares.

Population cycles could also be driven by seasonal changes. For example, traits that help animals to produce large numbers of offspring during the breeding season may reduce the ability of the animal to survive the non-breeding season. Body size is one such trait. Large individuals tend to produce more offspring, but their larger body size means that they find it harder to survive when food is scarce. As a consequence, large individuals should have an advantage and be more common when the population size is low and there are enough resources for all individuals. However, small individuals should be more abundant when population size is high. This trade-off caused by seasonality could set the population in motion towards predictable, multigenerational cycles.

To test this idea, Betini et al. established populations of fruit flies that went through ‘breeding’ and ‘non-breeding’ seasons. This was achieved by periodically altering the flies’ food to prevent the females from laying eggs (in the lab, fruit flies do not normally have non-breeding seasons). Over 58 generations, the number of flies in each population cycled between peaks of high and low numbers.

When the population contained relatively few flies, there was strong selection for large flies because they have high reproductive success. Hence, the population grew. When the population was large, meaning that the flies had to compete for a limited amount of food, there was strong selection for small flies because they are better able to survive on limited resources. However, small flies also produce fewer offspring on average, resulting in a decrease in population size. When the flies all had sufficient food during the non-breeding season, these regular cycles completely disappeared.

A major challenge will be to understand how common this phenomenon is in the wild. Virtually all organisms live in seasonal environments but whether they face strong trade-offs in the expression of traits is not well understood. This is primarily because of the difficulty in following individuals throughout the year.

DOI: http://dx.doi.org/10.7554/eLife.18770.002

Introduction

In many organisms, reproduction is confined to seasonal fluctuation in periods of high resource, in which both fecundity (reproduction) and viability (survival) selection can occur, and periods of low resources, when reproduction stops and natural selection occurs only through viability selection. Consequently, sequential episodes of reproduction and survival caused by seasonality could be a major source of fluctuations in the strength and direction of natural selection (Darwin, 1859; Lack, 1954; Fretwell, 1972; Schluter et al., 1991; Bell, 2010; Bergland et al., 2014), giving rise to classic life-history trade-offs (Lack, 1947; Roff, 1992; Stearns, 1992; Garland, 2014). More specifically, traits that confer a fecundity advantage, but which are associated with a survival cost, will experience natural selection in one season that is opposed by selection in the subsequent season (Levins, 1968; Michod, 2006; Bell, 2010; Bergland et al., 2014). The sequential rather than simultaneous nature of trade-offs driven by seasonality could have important consequence for the trait distribution within and across generations (Levins, 1968; Grafen, 1988; Michod, 2006) and population dynamics (Ozgul et al., 2010).

One way by which life history trade-offs might arise from seasonal variation in resources is via body size (Ozgul et al., 2010, 2014). Large individuals usually have higher fecundity (Mueller and Joshi, 2000; Schulte-Hostedde and Millar, 2004), but they could also have lower survival during the non-breeding season when resources are scarce (Stockhoff, 1991; Reznick et al., 2000; Munch et al., 2003; Monaghan, 2008). In addition, large individuals might take more time to grow and require more resources for maintenance (Munch et al., 2003), which could negatively impact their survival probability (Kingsolver and Huey, 2008).

This association between fitness and body size in seasonal environments could have important consequences for population dynamics, particularly when selection on body size is density-dependent (Mueller, 1997; Sinervo et al., 2000; Travis et al., 2013). Differences in the selective advantage of body size across seasons could also shed light on the long-standing question about why population densities of many species fluctuate periodically over time (Elton and Nicholson, 1942; Kendall et al., 1999; McCauley et al., 2008; Yan et al., 2013). For example, when body size is positively related to fecundity, but small individuals survive better in the non-breeding season (Stockhoff, 1991; Munch et al., 2003; Monaghan, 2008; Betini et al., 2014), these opposing patterns of selection could cause population cycles if selection is density-dependent. Specifically, if smaller offspring have higher survival when density is high, then the population will be composed of smaller than average individuals with lower average fecundity in the following breeding season. This lower mean fecundity will reduce population growth rates even though larger individuals will be favoured through fecundity selection. As population size declines, the strength of density-dependent viability selection on body size will also decline, which could cause net selection to reverse and favour larger individuals due to the fecundity benefit of being larger. As large individuals increase in frequency, population size should also increase via an improvement of reproductive output, returning populations to high densities. Although changes in the intensity and direction of natural selection caused by environmental variation are widespread (Schluter et al., 1991; Bell, 2010; Thompson, 2013; Bergland et al., 2014), we have little information about whether opposing episodes of natural selection could arise from seasonality and indirectly affect population dynamics through the feedback loop between ecological (density dependence) and evolutionary (selection and evolution) processes (Chitty, 1960; Krebs, 1978; Hairston et al., 2005; Pelletier et al., 2009; Ozgul et al., 2010; Schoener, 2011).

Here, we investigated how a seasonal fitness trade-off related to body-size could affect changes in population size and body size over time using replicate populations of Drosophila melanogaster exposed to repeated changes in food resources. In addition to standard breeding conditions for Drosophila, we also created a ‘non-breeding season’ by manipulating the food medium to prevent females from laying eggs during this period. Thus, in this system, breeding and survival were restricted to two sequential and distinct seasons (hereafter 'breeding' and 'non-breeding'). The number of days and amount of food in each season was determined so that both fecundity and non-breeding survival were density-dependent (Betini et al., 2013a, 2014, 2015), which is an important feature of many populations. In Drosophila, as in many other species, the positive correlation between body size and fecundity is well known (Mueller and Joshi, 2000; Appendix 1) and we previously demonstrated that small individuals have higher survival during the non-breeding season when abundance is high (Betini et al. 2013a, 2014). In addition, populations of D. melanogaster do not show evidence of multigenerational cycles (Mueller and Joshi, 2000), even when kept under the same conditions as our breeding season (i.e. ‘aseasonal populations’; Appendix 2). We, therefore, hypothesized that density dependence and opposing episodes of fecundity and viability selection on body resulting from seasonality could cause predictable and repeatable fluctuations in both population and body size. Specifically, we predicted that seasonal fitness trade-offs would cause population size and body size to undergo multigenerational cycles between periods of high abundance, when small individuals predominate, and periods of low abundance, when large individuals are more frequent. Furthermore, we predicted that populations not exposed to viability selection in the non-breeding season would lack periodic fluctuations in population size and body size.

We tested whether seasonality could result in multigenerational cycles in population size and body size using three experiments (Figure 1). In the first experiment, we submitted 45 replicate populations to the seasonal treatment described above and tracked the total number of individuals and body size at the end of the breeding and non-breeding season for 58 generations (the ‘long-term control’ treatment; Figure 1). In the second experiment, we tested the role of viability selection during the non-breeding season by tracking 13 additional populations over 31 generations using a similar protocol to the ‘long-term control’, but in which we experimentally prevented viability selection in the ‘non-breeding’ season by providing high levels of food during this season (the ‘stop-selection’ treatment; Figure 1). This protocol also maintained direct density effects on fecundity and survival, similar to the ones observed in the ‘long-term control’. In order to address potential environmental changes in the lab, we conducted a third experiment using the same protocol as in the 'long-term control', but under the same initial conditions and at the same time as the 'stop selection' treatment. This 'short-term control' experiment also had 13 replicate populations tracked over 31 generations (Figure 1).

Figure 1.
A schematic of the three experiments conducted in this study with accompanying duration, number of replicates and brief summary of their purpose.

In addition to these experiments, we also developed a mathematical model to investigate the contributions of both viability selection and delayed density dependence to population dynamics. The ‘stop-selection’ experiment was designed to eliminate viability selection, but might have also reduced potential effects of past densities on fecundity and survival. Such delayed density-dependent effects can also cause populations to cycle (Stenseth et al., 2003; Yan et al., 2013), or lead to more complex dynamics, such as chaos (May 1973). One potential mechanism for delayed density dependence are carry-over effects, which we have previously identified in this seasonal system (Betini et al., 2013a). Thus, we first investigated if lag effects were present in all three experiments, and then used the mathematical model to understand whether they played a role in the dynamics of their populations.

Results

Long-term control

Over 58 generations, the 45 replicated seasonal populations showed a predictable increase in abundance during the breeding season, where the food medium allowed females to lay eggs, and a decline in the subsequent non-breeding season (Figure 2a), as is typical of many natural seasonal systems. However, the autocorrelation functions (ACF) also revealed that these short, seasonal cycles were embedded within longer multigenerational cycles where average population size fluctuated 3-fold (insert in Figure 2a). In these populations, the ACF function was characterized by stationary periodic dynamics, which resulted in an oscillatory decay to zero (Figure 2a inset).

Figure 2.
Population size, changes in body size and selection differentials for body size in the 'long-term control' experiment.

To investigate the presence of viability selection for small body size and whether this selection was density-dependent, we measured female dry weight in 38 generations from 25 different populations. As expected, there is a negative correlation between population size at the end of the non-breeding season and body size at the end of the non-breeding season (Pearson's product-moment correlation = −0.64; t = −4.48, p<0.001), suggesting that density negatively impact body size in the non-breeding season. Mean survival during the non-breeding season was 71% (±0.21 SD) and survival was density-dependent (βsurvival = −0.001, t = −27.73, p<0.001). Mean female dry weight was significantly lower after the non-breeding season (0.279 mg, n = 3620 females) than before the non-breeding season (0.381 mg; n = 5258 females; standardized values = 0.577 before and −0.566 after the non-breeding season; Welch t-test: t = −35.90, df = 1,589.240.24, p<0.001; Figure 2b) and this viability selection was density-dependent (Figure 2c; Table 1, Appendix 3). That is, when population size was high at the start of the non-breeding season, there was stronger selection for smaller flies and this was driven by changes in mean dry weight after the non-breeding season rather than changes in the mean dry weight before the non-breeding season (Appendix 3). Average dry weight measured after the non-breeding season also showed multigenerational cycles (Figure 2d), as indicated by the autocorrelation function (Figure 2d, inset), varying between average peaks of 0.32 mg and lows of 0.23 mg.

Table 1.
Parameter estimates obtained from linear mixed effect models to investigate viability selection on body size as a function of thenumber of individuals at the beginning of the non-breeding season. In the ‘long-term control’, R2LMM(m)=0.18 ...

'Stop selection' experiment

In order to test the role of viability selection in these multigenerational cycles, we experimentally eliminated viability selection during the non-breeding season in 13 additional populations. Unlike the ‘long-term controls’ (Figure 2a), there was no evidence of multigenerational population cycles in these ‘stop selection’ populations (Figure 3a inset). In addition, body size did not significantly decline after the non-breeding season (Figure 3b; average body size was 0.337 mg before and 0.331 after the non-breeding season; n = 1290 and n = 689 females, respectively; standardized values: 0.028 before and −0.052, respectively; Welch t-test: t = −1.733, df = 1,479.400.40, p=0.081). There was also no evidence of density-dependent selection (Figure 3c; Table 1) and no evidence of cycles in body size (Figure 3d inset).

Figure 3.
Population size, changes in body size and selection differential for body size in the 'stop-selection' experiment.

'Short-term control'

Over 31 generations, and similar to the ‘long-term control’ (Figure 2a), the ‘short-term control' exhibited evidence for multigenerational cycles (Figure 4a). These 'short-term control' populations also experienced viability selection that was density-dependent. Overall, female dry weight significantly decreased from an average of 0.371 mg before the non-breeding season (n = 968 females) to 0.276 mg after the non-breeding season (n = 623 females; standardized values = 0.487 before and −0.761 after the non-breeding season; Welch t-test: t = −30.601, p<0.001; Figure 4b), but the magnitude of this viability selection was stronger when densities were higher at the start of the non-breeding season (Table 1, Figure 4c).

Figure 4.
Population size, changes in body size and selection differentials for body size in the 'short-term control' experiment.

Delayed density dependence

We statistically investigated whether fecundity and survival were influenced by density in past seasons (i.e. delayed density dependence) in all three experiments: ‘long-term control’, ‘short-term control’ and ‘stop-selection’. We used linear mixed effect models with vial (population) as a random effect and densities going back up to two generations as fixed effects. In the ‘long-term control’ and ‘short-term control’, fecundity and survival were influenced by density in the current and past season (Table 2 and Table 3). In contrast, the ‘stop-selection’ experiment showed evidence of the delayed effects on fecundity (Table 2), but not on survival (Table 3), meaning that the ‘stop-selection’ treatment eliminated both viability selection on body size as well as delayed effects of density on survival.

Table 2.
Parameter estimates obtained from linear mixed effect models to investigate the effects of current and past density on fecundity in the ‘long-term control’, ‘stop-selection’ and ‘short-term control’ experiments. ...
Table 3.
Parameter estimates obtained from linear mixed effect models to investigate the effects of current and past density on survival in the ‘long-term control’, ‘stop-selection’ and ‘short-term control’ experiments. ...

Mathematical model

A mathematical model including delayed density dependence as well as the effects of body size on survival (viability selection) and fecundity resulted in multigenerational cycles in population size (before red line in Figure 5a), similar to those observed in our ‘long-term control’ and ‘short-term control’ (Figure 2a and Figure 4a, respectively). The model without viability selection on body size and delayed density dependence (i.e. including only effects of current abundance on fecundity and survival), resulted in the elimination of multigenerational cycles (after red line in Figure 5a), as observed in our 'stop selection' populations (Figure 3a). The exclusion of only viability selection eliminated the fitness trade-off in body size and allowed larger flies to survive to breed. This led to unstable population dynamics (i.e. the population crashed after 10 generations; Figure 5b) because larger flies have greater fecundity and there was a negative interaction between body size and abundance. The exclusion of delayed density dependence alone eliminated the cycles (Figure 5c), suggesting that viability selection and delayed density dependence are necessary for these persistent multigenerational cycles to occur. The model with low heritability (h2 = 10−5) also generated multigenerational population cycles (Figure 5d).

Figure 5.
Predicted population size according to the integral projection model.

Discussion

Our results provide empirical and mathematical evidence that the interplay between density-dependence and evolutionary trade-offs caused by seasonality can have important consequences for population dynamics. In our experimental system, seasonal variation in resources resulted in a fitness trade-off and multigenerational population cycles. These cycles were observed in the absence of predation, long-term fluctuations in resources, or any explicit negative-frequency dependence, all of which are known to cause regular fluctuations in population size (Lindström et al., 2001; Yan et al., 2013). Thus, the emergence of population cycles caused by three common characteristics of natural populations (seasonality, a life-history trade-off and density-dependence), suggests that these ecological and evolutionary processes could contribute to fluctuations in population size over a wider range of taxa and environments than has previously been considered. Our experimental elimination of viability selection and delayed density-dependence and our mathematical model confirmed the importance of both of these evolutionary and ecological processes in the persistence of multigenerational population cycles resulting from seasonal environments.

Recently, it has been proposed that eco-evolutionary dynamics are essential for understanding oscillations in population size (Hairston et al., 2005; Hiltunen et al., 2014), but evidence for evolutionary change (i.e. genetically based change in ecologically relevant traits) to feedback and influence demography and population dynamics is rare (Schoener, 2011; Hendry, 2013); but see (Cameron et al., 2013). Our results suggest that the evolutionary response to selection might not be required to observe feedback between ecological and evolutionary processes, as long as selection is strong. In flies, as in many other species, body size is highly heritable (Prout and Barker, 1989), and it is reasonable to suppose that, in our populations, offspring from smaller flies tend to be smaller. However, even when genetic variation is low, strong selection can still affect trait distributions within generations (Grafen, 1988), potentially affecting population size. In our experimental system, selection for body size during the non-breeding season was strong enough to drive observed cycles, potentially in the absence of any genetic change across generations. Simulations with our mathematical model with essentially zero heritability (h2 = 10−5 as opposed to 0.3) also generated multigenerational population cycles because viability selection changed the distribution of body size, which subsequently affected fecundity. Thus, changes in the trait distribution within a generation caused by seasonal fitness trade-offs could generate feedback loops between ecological and evolutionary process in traits across generations.

We showed that variation in non-breeding abundance can cause density-dependent selection, but our mathematical model suggested that long-lasting effects of past density on survival were essential to generate the multigenerational cycles. These effects were strong in both the ‘long-term control’ and ‘short-term control’, but not in the ‘stop-selection’ treatment. One mechanism that can generate delayed effects is a density-mediated carry-over effect. We have shown how competition during the non-breeding season can negatively impact the physiological conditions of the survivors, reducing their breeding output in the following breeding season (Ratikainen et al., 2008; Betini et al., 2013a). This carry-over effect could also negatively impact survival if individuals during the non-breeding season were in lower body condition because of high abundance at the beginning of the season. Delayed density dependence can also arise from maternal effects, whereby offspring attributes are affected by parental abundance, which have previously been shown to generate population cycles (Plaistow and Benton, 2009). Both carry-over effects and maternal effects have the potential to interact with viability and fecundity selection because they can affect individual fecundity and survival via changes in body size and condition.

Our mathematical model also suggested that the delayed density dependence alone would cause short cycles and population crashes after only a few generations. It is clear from the model that the addition of viability selection allowed for sustained multigenerational cycles, as was observed in our empirical results. These results are in general agreement with previous studies. For example, in a predator-prey system, evolutionary changes in the prey population extended density cycles compared to the typical quarter-period phase lags between prey and predator densities (Yoshida et al., 2003). This evolutionary effect is now believed to be widespread in lab systems (Becks et al., 2012; Hiltunen et al., 2014). Thus, it is possible that evolutionary processes affecting trait distributions within and between generations can buffer strong negative feedback caused by ecological processes that are typical in many natural systems (e.g. overcompensation caused by density dependence), representing a different form of evolutionary rescue (Bell, 2013; Carlson et al., 2014).

Opposing viability and fecundity selection on body size was necessary for the persistence of multigenerational population cycles. Such fitness trade-offs between seasons are likely more common than currently appreciated (Schluter et al., 1991; Schmidt et al., 2005; Schmidt and Conde, 2006, Schmidt and Paaby 2008; Kingsolver and Diamond, 2011; Behrman et al., 2015), especially given that many organisms use vastly different habitats over the annual cycle. Although body size caused fitness trade-offs between seasons in our experimental system, many other traits might also exhibit seasonal trade-offs. For example, sexually selected traits enhance mating success, but might also reduce survival during the non-breeding season (Emlen, 2001). Alternatively, increased development rates can accelerate age of first reproduction, but are often associated with reduced survival during the adult stage (Stearns, 1992; Kingsolver and Huey, 2008).

Seasonal fitness trade-offs also represent a form of balancing selection, which can maintain increased levels of genetic variation at evolutionary equilibrium (Barton and Keightley, 2002). Seasonal fluctuations in selection might allow for the maintenance of increased adaptive potential in populations (Shaw and Shaw, 2014), facilitating more rapid responses to directional environmental change (Bradshaw and Holzapfel, 2006; Bell, 2010; Huang et al., 2016). This temporal variation in selective pressures might also help to explain why some species are more successful at colonizing new areas, as suggested by studies of seasonal changes in genetic variation in wild populations of D. melanogaster (Schmidt and Conde, 2006; Bergland et al., 2014; Behrman et al., 2015). Thus, the consequences of seasonal fluctuations in resources are, therefore, not restricted to purely ecological pathways. Seasonality can also alter the strength and direction of natural selection and might maintain a population’s adaptive potential over longer evolutionary timescales.

Examples of delayed density dependence causing cycles have been documented in natural populations subjected to strong seasonality (Merritt et al., 2001; Stenseth et al., 2003; Yan et al., 2013). We showed how opposing episodes of selection driven by a life-history trade-off arise in seasonal environments, which could interact with delayed effects to influence population size. Recently, it has been proposed that results from lab studies on consumer-resource dynamics were likely confounded by fast evolutionary changes (Hiltunen et al., 2014). The same could be true for lab and field studies on population cycles, if changes in trait distributions are linked to population size. To understand whether this is a common phenomenon in natural populations, one needs to tease apart the evolutionary from the ecological consequences of variation in density. As in our lab system, this is a constraint that needs to be overcome with either improved controlled experiments or mathematical approaches. Nevertheless, given that variation in density is common in many natural systems, measurements of selection and population density in both seasons will provide a better understanding of the dynamics of natural populations and how environmental change might alter such dynamics.

Materials and methods

‘Long-term control’ experiment

To simulate seasonality in populations with non-overlapping generations, we changed food composition to generate two distinct ‘seasons’ (n = 45 populations). During the breeding season, we placed adults in vials (28 × 95 mm) with a dead yeast-sugar medium to lay for 24 hr (day 0), after which adults were discarded and larvae were allowed to mature to adults (16 days). Individuals were then transferred (day 17) to a non-breeding environment for 4 days. The non-breeding season consisted of an empty vial of the same size as the breeding vials, but food was provided by a pipette tip filled with 0.200 ml of 5% water–sugar solution per day from the top of the vial. This solution was sufficient for many flies to survive (~95% survival at low population size) but did not allow females to lay eggs (Bownes and Blair, 1986; Betini et al., 2013a). In both seasonal treatments, a consistent amount of food was provided regardless of population size, which caused reproduction and survival to be density-dependent (Betini et al., 2013a). Each of 45 replicate populations was repeatedly transferred between breeding and non-breeding environments for 58 generations and together they are referred to as ‘long-term controls’. The quality and amount of the medium during the breeding season mimics the scenario well explored in other Drosophila systems (Mueller and Joshi, 2000), where the food is of lower quality and more limited for adults compared to larvae. Dynamics of populations experiencing only this medium (i.e. only the breeding season) do not show evidence of cycles (Mueller and Joshi, 2000); (Appendix 2—figure 1).

'Stop-selection' experiment

To investigate whether opposing selective pressures caused multigenerational cycles in seasonal populations, we experimentally stopped viability selection for a smaller body size in the non-breeding season while preserving density-dependent survival. To do this, we exposed 13 new replicate populations to the same seasonal change in food resources (described above) over 31 generations, but provided unlimited access to food during the non-breeding season (0.8 ml/day instead of 0.2 ml/day; initial population size was 5 males and 5 females). Average mortality was reduced from 28% (±0.4; mean ± s.e.) in the 'long-term control' populations to 0.5% (±1) in these new 'stop selection' treatments. After four days in the non-breeding season, we haphazardly selected the survivors that moved to the breeding season. To preserve the density-dependent survival process during the non-breeding period, the number of survivors for each population was calculated based on a logistic non-breeding survival function parameterized from our 'long-term control' populations (Wilson, 1994; Figure 6)

Su=11+(Nv)w
Figure 6.
The relationship between survival and population size at the beginning of the non-breeding season.

where Su is survival (number of survivors at the end of the non-breeding season divided by the total number of individuals at the beginning of the non-breeding season), N is the population size at the beginning of the non-breeding season, and v and w are constant to be estimated from the data. The function was parameterized with our 'long-term control' populations (45 replicates over 43 generations) using the non-linear function nls in R (R Core Team, 2015). Individuals that moved to the following breeding season were haphazardly selected.

‘Short-term control’ experiment

Our ‘long-term control’ and ‘stop-selection’ populations were initiated at different periods and with different population sizes. Thus, differences in the lab environment and initial conditions could have influenced the dynamics of these populations. For these reasons, we also initiated an additional 13 replicate populations at the same time and same initial population size as the ‘stop-selection’ populations. In these populations, we used the same protocol as in the ‘long-term control’, but initial population size (5 males and 5 females) and number of generations (n = 31) was the same as in the ‘stop-selection’ experiment.

Measuring viability selection on body size

We measured linear selection differentials (S) for female body weight (i.e. population mean dry

weight after – population mean dry weight before the non-breeding season) in all three experiments (‘long-term control’, ‘stop-selection’ and ‘short-term control’). We used female dry body weight as a proxy for body size. We experimentally confirmed that the mean body size declined along with mean dry weight in surviving females after the non-breeding season by placing individuals from the stock population in the non-breeding season in either high (n = 300) or low (n = 20) abundance. We measured the thorax size in 10 individual females sampled before the non-breeding season and 10 females sampled from the population after the non-breeding season. Mean thorax size after the non-breeding season was significantly lower than mean thorax size prior to the non-breeding season (Welch t-test: t = −2.87, df = 37.66, p=0.006), whereas there was no significant differences in average thorax in the low abundance treatment (Welch t-test: t = −0.55, df = 36.96, p=0.586; average size before = 1.318 mm; average size after the non-breeding season for the high abundance treatment was 1.225 and 1.297 mm for the low abundance treatment).

We obtained female dry weight from 5% of total population size before and after the non-breeding season. We then sexed, dried and weighed individual females to the nearest 0.001 mg using a microbalance. To calculate the linear selection differentials (S) for female body weight, we standardized dry weight measures to a mean of zero and unit variance prior to calculating selection (Brodie et al., 1995). The significance of our estimates of viability selection during the non-breeding season was assessed using a Welch t-test that tested for differences between mean female dry body weight before and after the non-breeding season. We then used a linear mixed effect model (LMM) to investigate whether S was density-dependent. In the LMM, the selection differential calculated for each population in each generation was the response variable, number of individuals at the beginning of the non-breeding season was fitted as a fixed effect and population was fitted as a random effect.

In the ‘long-term control’, we measured female dry weight in 38 generations from an arbitrary number of replicate populations (16 to 25 population per generation) in generations 9, 10, 15, 16, 21, 22, 25, 26, 28–58. Total number of individuals measured was 5258 before and 3620 after the non-breeding season. In the ‘stop-selection’ and ‘short-term control’, we sampled females in seven populations in generations 1 to 25 (n = 1591 females and n = 1979 females in the 'short-term control' and 'stop selection' treatments, respectively), with the exception of generation 13 in the 'short term control'.

Documenting multigenerational cycles

To investigate whether population size cycled in our experimental populations, we used the autocorrelation function method (Turchin and Taylor, 1992; Box et al., 2013). Cycles can be inferred by investigating the shape of the estimated ACF: population cycles are characterized by stationary periodic dynamics, which result in an oscillatory decay to zero of the ACF estimates. We calculated the ACF for each treatment using the mean population size across all replicates at the start of the breeding and non-breeding season in each generation. For all-time series, we excluded the first three generations to avoid any transient effects caused by the initial population size. We also detrended the time series to eliminate any overall temporal trends in abundance by subtracting the fitted values from a linear regression of average population size on generation (‘long-term control’: β = 0.468, s.e.=0.177, t = −2.65, p=0.011; ‘short-term control’: β = −1.855, s.e.=0.686, t = −2.71, p=0.01).

To investigate temporal changes in body size in the 'long-term control' and ‘stop-selection’ populations (measured as female dry weight), we used the same autocorrelation function method described above for the population size, including detrending the data.

Documenting delayed density dependence

We investigated the effects of past abundance on fecundity and survival with LMM with vial (population) as a random effect in all three experiments: ‘long-term control’ and ‘short-term control’ and ‘stop-selection’ treatment. In previous work, we have shown that fecundity in this system is influenced by an interaction between current abundance at the beginning of the breeding season and at the beginning of the previous non-breeding season (Betini et al., 2013a, 2013b). Thus, in this study, we also investigated if this interaction significantly explained variation in fecundity in all three experiments. Fecundity was calculated as the number of individuals at the end of the breeding season (i.e. number of offspring) divided by the number of individuals at the beginning of the same breeding season (number of parents). Because we did not know how environmental effects (e.g. physiological condition) could influence survival in our system, we incorporated the breeding and non-breeding abundance going back two generations as explanatory variables in the model to explain survival. Lag effects have been documented in natural populations subjected to strong seasonality (Merritt et al., 2001, 2001; Stenseth et al., 2003). Survival was calculated as the number of individuals at the end of the non-breeding season divided by the number of individuals at the beginning of the non-breeding season. Explanatory variables were standardized before analysis by subtracting the sample mean from each observation and dividing each value by the sample standard deviation and response variables were log transformed.

Mathematical model

We developed a mathematical model to investigate the contributions of both viability selection and delayed density dependence to population dynamics. The model linked quantitative trait evolution with demography and was similar to integral projection modelling approaches that are commonly used to study ecological and evolutionary processes concurrently (Slatkin, 1979, 1980; Ellner and Rees, 2006). In particular, the number of individuals with body size y at the next time step t + 1, n (y, t + 1), was a function of a projection function k (y) and n (y, t), such that n (y, t + 1) = k(y) * n (y, t). The projection function k (y) accounted for the additive genetic and environmental contributions to body size, as well as both density-independent and density-dependent differences in fecundity or survivourship among body sizes, depending on season.

Model description

Two population sizes were modeled in discrete time: population size at the beginning of the breeding season, and population size at the beginning of the non-breeding season within each generation. For each season, the mean breeding value for body size and the mean phenotypic value for body size were modelled, where the phenotypic value of an individual was its breeding value plus a random environmental effect. We assumed that breeding and phenotypic values were normally distributed and that the variance in breeding values and the variance in environmental effects remained constant across generations. The assumption of a constant variance in breeding values assumed that mutational variance and changes in genetic architecture (including linkage and epistatic interactions) could restore variance that was initially depleted by selection. Slatkin’s model provides for changes in the variance of breeding values due to selection and considers the effect of genetic architecture, such as linkage (Slatkin, 1979, 1980). In our experiment, we do not have information on the genetic architecture for body size, nor mutational variance. We did not see a decline in the magnitudes of the rates of change in body size in the experiment, which is consistent with the maintenance of heritability.

The fecundity of an individual during the breeding season was a function of its body size and current and past effects of density. Past density influenced fecundity via changes in the physiological conditions of individuals that spent the previous non-breeding season at high densities but survived to breed at low densities (i.e. a carry-over effect that was modelled as the interaction between density at the end of the breeding season in the previous generation and current breeding density (Betini et al., 2013a, 2014). Current density could also negatively influence fecundity via density-dependent effects. There was a stronger negative effect of density on fecundity for large versus small individuals, although large individuals have higher intrinsic fecundity (Figure 3 in the main text). During the 'stop selection' phase of the simulations, we decreased carry-over effects (parameters φ5and φ6) by 25% given that flies had unlimited amount of food in the 'stop selection' treatments and were likely to be in better physiological condition than those in the 'short-term control' treatments. Parameters used in the fecundity function were chosen such that fecundities as a function of body size and densities were within the normal range observed in D. melanogaster.

The survival of an individual during the non-breeding season was also a function of its body size (when viability selection occurs) and current and past effects of density. As indicated by our previous experiment, larger flies had lower survival and this effect was magnified by increase density (Betini et al., 2014). We included the effects of past density beyond the ones described above, going back two generations. During the 'stop selection' phase of the simulations, survival was not a function of body size nor past density, but was instead only affected by current density, following the logistic equation Su for survivorship and the experimental design. To accomplish this and to ensure proper scaling of survival, we assigned all individuals the same phenotype in the survivorship function. Parameters for the survival function were based on values obtained in the long-term control (Figure 2a in the main text).

In the context of an individual's breeding and phenotypic values, an individual with breeding value x had an environmental effect y added with a mean effect that was a negative linear function of density and was normally distributed with variance V[e]. Consequently, each breeding value expressed a distribution of phenotypes as a function of the current environment.

In the context of survivorship, since each breeding value expressed a distribution of phenotypes, each breeding value expressed a distribution of survivorship and the model integrated across environmental effects to get the mean survivorship associated with a breeding value. This integration of survivorship across environmental effects gave the value of the projection function for a given breeding value. The projection function in the context of fecundity was modelled similarly; breeding values were transformed into phenotypes via a distribution of environmental effects, which when integrated gave the mean fecundity of individuals with a particular breeding value for body size.

Population size across seasons was the product of the current population size times the average fecundity (breeding season) or average survivorship (non breeding season) at the phenotypic level. Below is a list of variables, parameters and functions used in the model.

Variables

Xi - population size at the beginning of the breeding season i generations ago

Yi - population size at the beginning of the non-breeding season i generations ago

bX¯ - mean breeding value for body size at the beginning of the breeding season

bY¯ - mean breeding value for body size at the beginning of the non-breeding season

pX¯ - mean phenotype for body size at the beginning of the breeding season

pY¯ - mean phenotype for body size at the beginning of the non-breeding season

z - body size

Note, a generation consists of a breeding season and then a non-breeding season. At the beginning of the breeding season the mean breeding value(bX¯)is a function of densities in previous generations, such that i > 0.

Parameters

VA=0.003 - additive genetic variance for body size

VE=0.007 - environmental variance for body size

such that h2=0.30 (Prout and Barker, 1989)

Fecundity function:

f(z,X,Y)=(φ1+(φ2z)4)exp((φ3+φ4z4)(X0+φ5Y1+φ6X0Y1))

Survival function:

s(z,X,Y)=1z(υ1(X1+Y1)+υ2(X2+Y2))1+(z(Y0+(υ3Y0z)exp(z)υ4)υ6)υ5

Integral projection model:

b¯x=zminzmaxzminxzmaxxxs(x+y,X,Y)N(x,b¯Y,VB)+N(y,eX(X1),VE)dydxzminzmaxzminxzmaxxs(x+y,X,Y)N(x,b¯Y,VB)+N(y,eX(X1),VE)dydx
X0=Y1zminzmaxs(y,X,Y)N(y,p¯Y,VB+VE)dyzminzmaxN(y,p¯Y,VB+VE)dy
p¯x=zminzmaxzminxzmaxx(x+y)N(x,b¯Y,VB)N(y,eY(Y1),VE)dydxzminzmaxzminxzmaxxN(x,b¯X,VB)N(y,eY(Y1),VE)dydx
b¯y=zminzmaxzminxzmaxxxf(x+y,X,Y)N(x,b¯X,VB)N(y,eY(Y1),VE)dydxzminzmaxzminxzmaxxf(x+y,X,Y)N(x,b¯X,VB)N(y,eY(Y1),VE)dydx
Y0=X0zminzmaxf(y,X,Y)N(y,p¯X,VB+VE)dyzminzmaxN(y,p¯X,VB+VE)dy
p¯y=zminzmaxzminxzmaxx(x+y)N(x,b¯Y,VB)N(y,eX(X1),VE)dydxzminzmaxzminxzmaxxN(x,b¯Y,VB)N(y,eX(X1),VE)dydx

In the equations above N(z,m,v) is the probability density of a normally distributed random variable with a mean of m and variance v. Functions eX(N)=λXNand eY(N)=λYNgive the average environmental effect on the phenotype.

Other parameters and their values

φ1=1.5 (intrinsic fecundity for small flies)

φ2=2.8 (rate of increase in fecundity with body size)

φ3=0.00007 (baseline rate of decline in fecundity with density for small flies)

φ4=0.04588 (rate of increase in the magnitude of the decline in fecundity as body size increases)

φ5=1 (constant characterizing the effects of population size at the end of the breeding season one generation ago; i.e. carry-over effects)

φ6=0.001 (constant characterizing the interaction between current population size and population size at the end of the breeding season one generation ago)

υ1=0.005 (rate of decline in survivorship due to population sizes one generation ago)

υ2=0.0026 (rate of decline in survivorship due to population sizes two generations ago)

υ3=0.7 (constant governing negative effect of the interaction between the current population size and body size)

υ4=0.7 (constant governing the shape of the negative effect of the interaction between current population size and body size on survivorship)

υ5=6 (a second constant governing the shape of the negative effect of the interaction between current population size and body size on survivorship)

υ6=350 (a third constant governing the shape of the negative effect of the interaction between current population size and body size on survivorship)

λX=0.00001 (rate of decline in body size [with density], i.e. the environmental effect of density on body size due to density at the beginning of the breeding season)

λY=0.00025 (rate of decline in body size [with density], i.e. the environmental effect of density

on body size due to density at the beginning of the non-breeding season)

All LMM were performed using the lmer function from the lme4 package (Bates, 2010) and p-values were obtained using the lmerTest package (Kuznetsova et al., 2014). All analyses were conducted in R (R Core Team, 2015). Marginal (R2LMM(m)) and conditional (R2L<MM(c)) variance for the LMMs were calculated with MuMIn package according to (Nakagawa and Schielzeth, 2013). R2LMM(m) is the variance on the response variable that is explained only by the fixed effects and R2LMM(c) is the variance that is explained by both fixed and random effects (Nakagawa and Schielzeth, 2013).

Acknowledgements

We thank E Hentsch, RJ Kilgour, RS Cruz and AN Pessoa for assistance in the lab and three anonymous reviewers for helpful suggestions. The research was supported by NSERC Discovery grants to DRN, CKG, AGM, a University Research Chair from the University of Guelph and an Early Researcher Award to DRN and an Ontario Graduate Scholarship to GSB.

Appendix 1

Experimental system

We used Drosophila melanogaster from an outbred population collected in Dahomey (now Benin) in 1970, which has since been maintained in cage culture at 25°C. To simulate seasonality in populations with non-overlapping generations, we changed food composition to create two distinct ‘seasons’. During the ‘breeding season’, flies were allowed to lay eggs for 24 hr (day 0) in a dead yeast–sugar medium (1000 ml H2O, 100 g sucrose, 50 g Fleischmann’s yeast, 16 g agar, 8 g C4H4KNaO6, 1 g KH2PO4, 0.5 g NaCl, 0.5 g MgCl2, 0.5 g CaCl2, 0.5 g Fe2(SO4)3), after which adults were discarded and larvae were allowed to mature to adults. During the breeding season, emerging offspring were moved from old to new vials containing fresh food on days 10, 12 and 14 to avoid high offspring mortality (Dey and Joshi, 2006). During this period, densities of offspring were not manipulated, that is, their densities were a function of the number of flies that emerged during the breeding season. Flies were marked with a fluorescent dust (as part of the protocol of another experiment) and let rest overnight in larger bottles (177.4 ml) with 20 ml of food so that they could easily remove the excess of dust. On day 17, flies were lightly anesthetized with CO2, counted and placed into the non-breeding vials. The non-breeding season consisted of an empty vial of the same size as the breeding vials (28 × 95 mm) and a pipette tip filled with 0.200 ml of 5% water–sugar solution per day. This medium prevented females from producing eggs (Terashima et al., 2005) but also resulted in density-dependent mortality, as would be expected from wild populations (Betini et al., 2013a). Oviposition resumed within <12 hr when flies were placed back on a protein-rich food (Terashima et al., 2005). After 4 days in the non-breeding season, surviving flies were counted and the cycle was repeated. Note that since adults were removed after laying at the start of the breeding season, there were non-overlapping generations and each generation experienced only a single breeding and non-breeding season. We also randomly removed 5% of each population each time they were moved between seasons to mimic migratory mortality and dispersal. This procedure was repeated for 58 generations in each of 45 replicate populations. During all experiments, flies were kept at 25°C, 12 hr light/dark cycles and humidity was between 30% to 50%.

As in many other animals, body size was positively related to fecundity in our experimental system according to a linear mixed effect model with per capita breeding output as the response variable, female dry weight as a explanatory variable (average for each replicate, over all 38 generations) and population as a random effects (Appendix 1—figure 1 and Appendix 1—table 1).

Appendix 1—figure 1.

An external file that holds a picture, illustration, etc.
Object name is elife-18770-app1-fig1.jpg
Fecundity as a function of average female dry weight at the beginning of the breeding season.

Larger individuals tend to have higher fecundity. The solid line represents the linear mixed effect model line for female dry weight (Appendix 1—table 1).

DOI: http://dx.doi.org/10.7554/eLife.18770.012

Appendix 1—table 1.

Parameter estimates obtained from a linear mixed effect model with fecundity as a response variable and average female dry weight at the beginning of the breeding season as explanatory variable. R2LMM(m)=0.04 and R2LMM(c)=0.14.

DOI: http://dx.doi.org/10.7554/eLife.18770.013

ParametersSEdftp

Fixed effects estimate





Intercept

0.072

0.043

260

1.69

0.101

Female dry weight

1.178

0.092

3601

12.82

<0.001


Random effect variance





Population

0.015





Residual

0.125





Appendix 2

Multigenerational cycles

To understand the overall effect of seasonality on changes in population size over time, we compared the time series generated by our 'long-term control' with another treatment where 30 replicate populations of flies were maintained in breeding vials (28 × 95 mm) with standard medium across 30 generations. These populations lacked a non-breeding period, so we refer to these as the 'aseasonal' populations (Appendix 2—figure 1a,b). In these populations, all offspring were moved to a new fresh vials with breeding medium on day 16.

To investigate whether population size cycled in the ‘aseasonal’ treatment, we used the same autocorrelation function method described in the main text, after detrending the data ('aseasonal': β = 0.403, s.e.=0.579, t = 0.69, p=0.49; Appendix 2—figure 1).

Appendix 2—figure 1.

An external file that holds a picture, illustration, etc.
Object name is elife-18770-app2-fig1.jpg
Overall effect of seasonality in multigenerational cycles.

(a) Population size of ‘aseasonal’ populations over 30 generations do not present evidence of multigenerational cycles according to the (b) autocorrelation plots. In a solid blue line denotes mean population size for each generation from all replicates and dotted lines denote ±1 s.d.

DOI: http://dx.doi.org/10.7554/eLife.18770.014

Appendix 3

Viability selection

We found that viability selection on body size (S) became more negative (i.e. stronger selection for reduce size) with increasing non-breeding density (Figure 2c in the main text; Appendix 3—table 1) and this was driven by changes in mean dry weight after the non-breeding season rather than changes in the mean dry weight before the non-breeding season (Appendix 3—table 1; Appendix 3—figure 1).

Appendix 3—table 1.

Parameter estimates obtained from a linear mixed effect models to investigate variation in female dry weight as a function of the interaction between density at the beginning of the non-breeding season and whether weight was measured before or after the non-breeding season. Status refers to a dummy variable indicating the time when body size was measured (before or after the non-breeding season). Reference value was after the non-breeding season. Marginal (only taking account the fixed effects) and conditional (taking into account both fixed and random effects) coefficient of determination were R2LMM(m)=0.42 and R2LMM(c)=0.50, respectively. In all models, population and generation were entered as random effects.

DOI: http://dx.doi.org/10.7554/eLife.18770.015

ParametersSEdfDftTpP

Fixed effects estimate





Intercept

0.680

0.06

266

10.31

<0.001

Non-breeding density

0.005

0.0002

900

−29.38

<0.001

Status Before

0.430

0.009

896

7.55

<0.001

Non-breeding density:Status

0.003

0.0001

892

14.11

<0.001


Random effect variance

SD




Generation

0.045

0.212




Population

0.037

0.193




Residual

0.474

0.688




Appendix 3—figure 1.

An external file that holds a picture, illustration, etc.
Object name is elife-18770-app3-fig1.jpg
The strength of selection on body weight in seasonal environments.

As non-breeding density increased, the magnitude of directional selection for reduced dry weight also increased (Figure 2c in the main text). This relationship was driven by changes in the weight after (red points) rather than before (blue points) the non-breeding season. Solid lines indicate the linear mixed effect model line for non-breeding density before (blue line) and after (red line) the non-breeding season.

DOI: http://dx.doi.org/10.7554/eLife.18770.016

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Funding Information

This paper was supported by the following grants:

  • Ontario Graduate Scholarship to Gustavo S Betini.
  • http://dx.doi.org/10.13039/501100000038Natural Sciences and Engineering Research Council of Canada to Andrew G McAdam, Cortland K Griswold, D Ryan Norris.

Additional information

Competing interests

The authors declare that no competing interests exist.

Author contributions

GSB, Conceptualization, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing.

AGM, Conceptualization, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing.

CKG, Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing.

DRN, Conceptualization, Resources, Supervision, Funding acquisition, Writing—original draft, Writing—review and editing.

References

  • Barton NH, Keightley PD. Understanding quantitative genetic variation. Nature Reviews Genetics. 2002;3:11–21. doi: 10.1038/nrg700. [PubMed] [Cross Ref]
  • Bates DM. Lme4: Mixed-Effects Modeling with R. New York, USA: Springer; 2010.
  • Becks L, Ellner SP, Jones LE, Hairston NG. The functional genomics of an eco-evolutionary feedback loop: linking gene expression, trait evolution, and community dynamics. Ecology Letters. 2012;15:492–501. doi: 10.1111/j.1461-0248.2012.01763.x. [PubMed] [Cross Ref]
  • Behrman EL, Watson SS, O'Brien KR, Heschel MS, Schmidt PS. Seasonal variation in life history traits in two Drosophila species. Journal of Evolutionary Biology. 2015;28:1691–1704. doi: 10.1111/jeb.12690. [PMC free article] [PubMed] [Cross Ref]
  • Bell G. Fluctuating selection: the perpetual renewal of adaptation in variable environments. Philosophical Transactions of the Royal Society B: Biological Sciences. 2010;365:87–97. doi: 10.1098/rstb.2009.0150. [PMC free article] [PubMed] [Cross Ref]
  • Bell G. Evolutionary rescue and the limits of adaptation. Philosophical Transactions of the Royal Society B: Biological Sciences. 2013;368:20120080 doi: 10.1098/rstb.2012.0080. [PMC free article] [PubMed] [Cross Ref]
  • Bergland AO, Behrman EL, O'Brien KR, Schmidt PS, Petrov DA. Genomic evidence of rapid and stable adaptive oscillations over seasonal time scales in Drosophila. PLoS Genetics. 2014;10:e1004775 doi: 10.1371/journal.pgen.1004775. [PMC free article] [PubMed] [Cross Ref]
  • Betini GS, Fitzpatrick MJ, Norris DR. Experimental evidence for the effect of habitat loss on the dynamics of migratory networks. Ecology Letters. 2015;18:526–534. doi: 10.1111/ele.12432. [PubMed] [Cross Ref]
  • Betini GS, Griswold CK, Norris DR. Carry-over effects, sequential density dependence and the dynamics of populations in a seasonal environment. Proceedings of the Royal Society B: Biological Sciences. 2013a;280:20130110 doi: 10.1098/rspb.2013.0110. [PMC free article] [PubMed] [Cross Ref]
  • Betini GS, Griswold CK, Norris DR. Density-mediated carry-over effects explain variation in breeding output across time in a seasonal population. Biology Letters. 2013b;9:20130582 doi: 10.1098/rsbl.2013.0582. [PMC free article] [PubMed] [Cross Ref]
  • Betini GS, Griswold CK, Prodan L, Norris DR. Body size, carry-over effects and survival in a seasonal environment: consequences for population dynamics. Journal of Animal Ecology. 2014;83:1313–1321. doi: 10.1111/1365-2656.12225. [PubMed] [Cross Ref]
  • Bownes M, Blair M. The effects of a sugar diet and hormones on the expression of the Drosophila yolk-protein genes. Journal of Insect Physiology. 1986;32:493–501. doi: 10.1016/0022-1910(86)90011-9. [Cross Ref]
  • Box GEP, Jenkins GM, Reinsel GC. Time Series Analysis: Forecasting and Control. John Wiley & Sons, Inc; 2013.
  • Bradshaw WE, Holzapfel CM. Evolutionary response to rapid climate change. Science. 2006;312:1477–1478. doi: 10.1126/science.1127000. [PubMed] [Cross Ref]
  • Brodie ED, Moore AJ, Janzen FJ. Visualizing and quantifying natural selection. Trends in Ecology & Evolution. 1995;10:313–318. doi: 10.1016/S0169-5347(00)89117-X. [PubMed] [Cross Ref]
  • Cameron TC, O'Sullivan D, Reynolds A, Piertney SB, Benton TG. Eco-evolutionary dynamics in response to selection on life-history. Ecology Letters. 2013;16:754–763. doi: 10.1111/ele.12107. [PMC free article] [PubMed] [Cross Ref]
  • Carlson SM, Cunningham CJ, Westley PA. Evolutionary rescue in a changing world. Trends in Ecology & Evolution. 2014;29:521–530. doi: 10.1016/j.tree.2014.06.005. [PubMed] [Cross Ref]
  • Chitty D. Population processes in the vole and their relevance to general theory. Canadian Journal of Zoology. 1960;38:99–113. doi: 10.1139/z60-011. [Cross Ref]
  • Darwin C. On the Origin of Species. London: Murray; 1859.
  • Dey S, Joshi A. Stability via asynchrony in Drosophila metapopulations with low migration rates. Science. 2006;312:434–436. doi: 10.1126/science.1125317. [PubMed] [Cross Ref]
  • Ellner SP, Rees M. Integral projection models for species with complex demography. The American Naturalist. 2006;167:410–428. doi: 10.1086/499438. [PubMed] [Cross Ref]
  • Elton C, Nicholson M. The Ten-Year cycle in numbers of the Lynx in Canada. The Journal of Animal Ecology. 1942;11:215–244. doi: 10.2307/1358. [Cross Ref]
  • Emlen DJ. Costs and the diversification of exaggerated animal structures. Science. 2001;291:1534–1536. doi: 10.1126/science.1056607. [PubMed] [Cross Ref]
  • Fretwell SD. Populations in a Seasonal Environment. Princeton: Princeton University Press; 1972.
  • Garland T. Trade-offs. Current Biology. 2014;24:R60–R61. doi: 10.1016/j.cub.2013.11.036. [PubMed] [Cross Ref]
  • Grafen A. Reproductive Success. Chicago: University of Chicago Press; 1988. On the uses of data on lifetime reproductive success; pp. 454–471.
  • Hairston NG, Ellner SP, Geber MA, Yoshida T, Fox JA. Rapid evolution and the convergence of ecological and evolutionary time. Ecology Letters. 2005;8:1114–1127. doi: 10.1111/j.1461-0248.2005.00812.x. [Cross Ref]
  • Hendry AP. Key questions in the genetics and genomics of eco-evolutionary dynamics. Heredity. 2013;111:456–466. doi: 10.1038/hdy.2013.75. [PMC free article] [PubMed] [Cross Ref]
  • Hiltunen T, Hairston NG, Hooker G, Jones LE, Ellner SP. A newly discovered role of evolution in previously published consumer-resource dynamics. Ecology Letters. 2014;17:915–923. doi: 10.1111/ele.12291. [PubMed] [Cross Ref]
  • Huang Y, Tran I, Agrawal AF. Does genetic variation maintained by environmental heterogeneity facilitate adaptation to novel selection? The American Naturalist. 2016;188:27–37. doi: 10.1086/686889. [PubMed] [Cross Ref]
  • Kendall BE, Briggs CJ, Murdoch WW, Turchin P, Ellner SP, McCauley E, Nisbet RM, Wood SN. Why do populations cycle? a synthesis of statistical and mechanistic modeling approaches. Ecology. 1999;80:1789–1805. doi: 10.1890/0012-9658(1999)080[1789:WDPCAS]2.0.CO;2. [Cross Ref]
  • Kingsolver JG, Diamond SE. Phenotypic selection in natural populations: what limits directional selection? The American Naturalist. 2011;177:346–357. doi: 10.1086/658341. [PubMed] [Cross Ref]
  • Kingsolver JG, Huey RB. Size, temperature, and fitness: three rules. Evolutionary Ecology Research. 2008;10:251–268.
  • Krebs CJ. A review of the chitty hypothesis of population regulation. Canadian Journal of Zoology. 1978;56:2463–2480. doi: 10.1139/z78-335. [Cross Ref]
  • Kuznetsova A, Brockhoff PB, Christensen RHB. lmerTest: Tests in Linear Mixed Effects ModelsR package version 2.0-20http://CRAN.R-project.org/package=lmerTest 2014
  • Lack D. The significance of Clutch-size. Ibis. 1947;89:302–352. doi: 10.1111/j.1474-919X.1947.tb04155.x. [Cross Ref]
  • Lack DL. The Natural Regulation of Animal Numbers. London: Oxford University Press; 1954.
  • Levins R. Evolution in Changing Environments: Some Theoretical Explorations. Princeton: Princeton University Press; 1968.
  • Lindström J, Ranta E, Kokko H, Lundberg P, Kaitala V. From arctic lemmings to adaptive dynamics: charles Elton's legacy in population ecology. Biological Reviews of the Cambridge Philosophical Society. 2001;76:129–158. doi: 10.1017/S1464793100005637. [PubMed] [Cross Ref]
  • McCauley E, Nelson WA, Nisbet RM. Small-amplitude cycles emerge from stage-structured interactions in Daphnia-algal systems. Nature. 2008;455:1240–1243. doi: 10.1038/nature07220. [PubMed] [Cross Ref]
  • Merritt JF, Lima M, Bozinovic F. Seasonal regulation in fluctuating small mammal populations: feedback structure and climate. Oikos. 2001;94:505–514. doi: 10.1034/j.1600-0706.2001.940312.x. [Cross Ref]
  • Michod RE. The group covariance effect and fitness trade-offs during evolutionary transitions in individuality. PNAS. 2006;103:9113–9117. doi: 10.1073/pnas.0601080103. [PubMed] [Cross Ref]
  • Monaghan P. Early growth conditions, phenotypic development and environmental change. Philosophical Transactions of the Royal Society B: Biological Sciences. 2008;363:1635–1645. doi: 10.1098/rstb.2007.0011. [PMC free article] [PubMed] [Cross Ref]
  • Mueller LD, Joshi A. Stability in Model Populations. Princeton, NJ, USA: Princeton University Press; 2000.
  • Mueller LD. Theoretical and empirical examination of density-dependent selection. Annual Review of Ecology and Systematics. 1997;28:269–288. doi: 10.1146/annurev.ecolsys.28.1.269. [Cross Ref]
  • Munch SB, Mangel M, Conover DO. Quantifying natural selection on body size from field data: winter mortality in menidia menidia. Ecology. 2003;84:2168–2177. doi: 10.1890/02-0137. [Cross Ref]
  • Nakagawa S, Schielzeth H. A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods in Ecology and Evolution. 2013;4:133–142. doi: 10.1111/j.2041-210x.2012.00261.x. [Cross Ref]
  • Ozgul A, Bateman AW, English S, Coulson T, Clutton-Brock TH. Linking body mass and group dynamics in an obligate cooperative breeder. Journal of Animal Ecology. 2014;83:1357–1366. doi: 10.1111/1365-2656.12239. [PubMed] [Cross Ref]
  • Ozgul A, Childs DZ, Oli MK, Armitage KB, Blumstein DT, Olson LE, Tuljapurkar S, Coulson T. Coupled dynamics of body mass and population growth in response to environmental change. Nature. 2010;466:482–485. doi: 10.1038/nature09210. [PubMed] [Cross Ref]
  • Pelletier F, Garant D, Hendry AP. Eco-evolutionary dynamics. Philosophical Transactions of the Royal Society B: Biological Sciences. 2009;364:1483–1489. doi: 10.1098/rstb.2009.0027. [PMC free article] [PubMed] [Cross Ref]
  • Plaistow SJ, Benton TG. The influence of context-dependent maternal effects on population dynamics: an experimental test. Philosophical Transactions of the Royal Society B: Biological Sciences. 2009;364:1049–1058. doi: 10.1098/rstb.2008.0251. [PMC free article] [PubMed] [Cross Ref]
  • Prout T, Barker JS. Ecological aspects of the heritability of body size in Drosophila buzzatii. Genetics. 1989;123:803–813. [PubMed]
  • R Core Team . Vienna, Austria: R Foundation for Statistical Computing; 2015.
  • Ratikainen II, Gill JA, Gunnarsson TG, Sutherland WJ, Kokko H. When density dependence is not instantaneous: theoretical developments and management implications. Ecology Letters. 2008;11:184–198. doi: 10.1111/j.1461-0248.2007.01122.x. [PubMed] [Cross Ref]
  • Reznick D, Nunney L, Tessier A. Big houses, big cars, superfleas and the costs of reproduction. Trends in Ecology & Evolution. 2000;15:421–425. doi: 10.1016/S0169-5347(00)01941-8. [PubMed] [Cross Ref]
  • Roff DA. Evolution of Life Histories: Theory and Analysis. Springer-Verlag, USA: Springer; 1992.
  • Schluter D, Price TD, Rowe L. Conflicting selection pressures and life history Trade-Offs. Proceedings of the Royal Society B: Biological Sciences. 1991;246:11–17. doi: 10.1098/rspb.1991.0118. [Cross Ref]
  • Schmidt PS, Conde DR. Environmental heterogeneity and the maintenance of genetic variation for reproductive diapause in Drosophila Melanogaster. Evolution. 2006;60:1602–1611. doi: 10.1554/05-430.1. [PubMed] [Cross Ref]
  • Schmidt PS, Paaby AB, Heschel MS. Genetic variance for diapause expression and associated life histories in Drosophila Melanogaster. Evolution. 2005;59:2616–2625. doi: 10.1111/j.0014-3820.2005.tb00974.x. [PubMed] [Cross Ref]
  • Schmidt PS, Paaby AB. Reproductive diapause and life-history clines in north american populations of Drosophila Melanogaster. Evolution. 2008;62:1204–1215. doi: 10.1111/j.1558-5646.2008.00351.x. [PubMed] [Cross Ref]
  • Schoener TW. The newest synthesis: understanding the interplay of evolutionary and ecological dynamics. Science. 2011;331:426–429. doi: 10.1126/science.1193954. [PubMed] [Cross Ref]
  • Schulte-Hostedde AI, Millar JS. Intraspecific variation of testis size and sperm length in the yellow-pine chipmunk (Tamias amoenus): implications for sperm competition and reproductive success. Behavioral Ecology and Sociobiology. 2004;55:272–277. doi: 10.1007/s00265-003-0707-z. [Cross Ref]
  • Shaw RG, Shaw FH. Quantitative genetic study of the adaptive process. Heredity. 2014;112:13–20. doi: 10.1038/hdy.2013.42. [PMC free article] [PubMed] [Cross Ref]
  • Sinervo B, Svensson E, Comendant T. Density cycles and an offspring quantity and quality game driven by natural selection. Nature. 2000;406:985–988. doi: 10.1038/35023149. [PubMed] [Cross Ref]
  • Slatkin M. The evolutionary response to frequency- and density-dependent interactions. The American Naturalist. 1979;114:384–398. doi: 10.1086/283487. [Cross Ref]
  • Slatkin M. Ecological character displacement. Ecology. 1980;61:163–177. doi: 10.2307/1937166. [Cross Ref]
  • Stearns SC. The Evolution of Life Histories. Oxford: Oxford University Press; 1992.
  • Stenseth NC, Viljugrein H, Saitoh T, Hansen TF, Kittilsen MO, Bølviken E, Glöckner F. Seasonality, density dependence, and population cycles in hokkaido voles. PNAS. 2003;100:11478–11483. doi: 10.1073/pnas.1935306100. [PubMed] [Cross Ref]
  • Stockhoff BA. Starvation resistance of gypsy moth, Lymantria dispar (L.) (Lepidoptera: lymantriidae): tradeoffs among growth, body size, and survival. Oecologia. 1991;88:422–429. doi: 10.1007/BF00317588. [Cross Ref]
  • Terashima J, Takaki K, Sakurai S, Bownes M. Nutritional status affects 20-hydroxyecdysone concentration and progression of oogenesis in Drosophila Melanogaster. Journal of Endocrinology. 2005;187:69–79. doi: 10.1677/joe.1.06220. [PubMed] [Cross Ref]
  • Thompson JN. Relentless Evolution. University of Chicago Press; 2013.
  • Travis J, Leips J, Rodd FH. Evolution in population parameters: density-dependent selection or density-dependent fitness? The American Naturalist. 2013;181 Suppl 1:S9–S20. doi: 10.1086/669970. [PubMed] [Cross Ref]
  • Turchin P, Taylor AD. Complex dynamics in ecological time series. Ecology. 1992;73:289–305. doi: 10.2307/1938740. [Cross Ref]
  • Wilson DL. The analysis of survival (mortality) data: fitting gompertz, Weibull, and logistic functions. Mechanisms of Ageing and Development. 1994;74:15–33. doi: 10.1016/0047-6374(94)90095-7. [PubMed] [Cross Ref]
  • Yan C, Stenseth NC, Krebs CJ, Zhang Z. Linking climate change to population cycles of hares and Lynx. Global Change Biology. 2013;19:3263–3271. doi: 10.1111/gcb.12321. [PubMed] [Cross Ref]
  • Yoshida T, Jones LE, Ellner SP, Fussmann GF, Hairston NG. Rapid evolution drives ecological dynamics in a predator-prey system. Nature. 2003;424:303–306. doi: 10.1038/nature01767. [PubMed] [Cross Ref]
2017; 6: e18770.

Decision letter

Lutz Becks, Reviewing editor
Lutz Becks, Max Planck Institute for Evolutionary Biology, Germany;

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Fitness trade-off between seasons causes multigenerational cycles in phenotype and population size" for consideration by eLife. Your article has been reviewed by two peer reviewers, one of whom, Lutz Becks (Reviewer #1), is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

While the reviewers agreed that this is an interesting manuscript with a timely topic and interesting results, the reviewers also felt that the experiments reported in the manuscript don't provide sufficiently strong support for the conclusions. Some of the interpretation of the results are not supported by statistical tests and the study lacks an effort to establish lines at different time points to determine whether genetic changes did occur. The overall structure of the manuscript was confusing. In summary, we didn't feel that the manuscript provides a decisive answer to the question raised.

Reviewer #1:

This is an interesting study examining the joined effects of ecological and evolutionary processes on population dynamics in seasonally fluctuating environments. Using a series of experimental evolution studies with Drosophila populations, Betini and colleagues show how density dependency and viability selection leads to multi-generational cycles of the populations. They make use of the fact that body size affects fecundity (positively) and survival (negatively) and that these two traits trade off between breeding and non-breeding seasons (limited food supply). Specifically, they compare the evolutionary (body size) and ecological changes (density) within populations when manipulating seasonality (absence, presence with very low and low food supply). They found that seasonality leads to multi-generational cycles of 7-9 generations and that these cycles increase with lower food supply in the non-breeding season. An additional mathematical model shows that the multi-generational cycles depend on the trade-off and differential selection between seasons and the density-dependency of growth. Overall, this is an important study showing in an elegant way the underlying mechanism producing multi-generational cycles in populations without interspecific species interactions. It adds to the growing evidence that ecological and (rapid) evolutionary dynamics interact at one timescale. I have however several concerns that need to be addressed:

1) The Results section is not clearly structured and the differences between the different treatments were only clear after reading it several times. I suggest to restructure this section and provide the reader with a short overview of what has been tested in which experiment at the beginning of this section, which should also mention the predictions.

2) Some statements are not supported by statistical analyses and it is difficult to judge the results:

– The fold changes in population size differences within the multi-generational cycles are not supported by any statistical analyses and for the 'long-term-control' not easy to see in the data. Could you provide some statistical test for the differences in amplitude?

Figure 7: the ACF plot does not provide evidence for a statistical period in the time series.

– Could you provide some statistical evidence for the model results (e.g., AFC plots similar to the experimental ones) so the reader can compare better?)?

- Subsection “Mathematical Model”: how do you judge 'unstable population dynamics' in your model?

– It is interesting to see that the period of the densities (Figure 2) and female dry weight (Figure 6) are very similar (as expected). Is there a correlation between these two time series? And what would one expect?

3) The model should be introduced in the main text for the reader to be able to follow better what was manipulated and to test for the effects of viability selection and delayed density effects.

4) ACF analyses: please provide more explanation for the ACF plots, so that the readers know how to interpret the maximum in ACF and the significance line you show in the plots.

Reviewer #2:

I am deeply interested in the question and I found the work to be done well. However, I cannot fully judge the novelty of the work within the ecological literature. The eco-evolutionary part of the work is less well developed as it is not fully clear whether the evolutionary changes were really important in the generation of the longer cycles as the authors claim even though they have some suggestive data. At the end, lack of effort to establish lines at different times of the cycles and determine whether genetic changes did occur I think limits the work. My hunch is that it does not rise to the level of an eLife paper.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for choosing to send your work entitled "Fitness trade-off between seasons causes multigenerational cycles in phenotype and population size" for consideration at eLife. Your letter of appeal has been considered by a Senior Editor, a Reviewing editor, and one reviewer and we are prepared to consider a revised submission with no guarantees of acceptance.

The Reviewing Editor has drafted this decision to help you prepare a revised submission.

Reviewer #3:

The authors present an impressive set of long lasting and well replicated experiments testing important eco-evolutionary consequences of seasonality. This shows the power of model lab systems! I think the questions addressed were well introduced building up from more targeted mechanisms to more complex eco-evo interactions that could possibly explain well known ecological patterns (multigenerational populations cycles).

My main concern is that the manuscript is overly complicated mostly due to a lack of proper organization. It was very confusing reading this for the first and even the second time. I have made a number of suggestions below that reorder the methods and improve clarity. The interesting manipulation to compare the control to the stop-selection experiment is lost in the confusion that is the methods sections.

Main Concerns:

1) There is a lot of importance given to delayed density dependence in the Discussion (but none in the Introduction). However, there was no formal test of it, which I could recognize, in the actual empirical data even though the model suggests it is very important. I am not an expert in this topic so why wasn't the experimental data tested for it?

2) The manuscript was very difficult to follow and unnecessarily confusing. I think this stems from two many issues.

a) The number, naming, and ordering of experiments is confusing. There are a lot of similar experiments being presented and it is difficult to follow which is being discussed or how they differ. This is caused firstly by the having their descriptions only appear after the results (see point 2). In addition the order of the experiments doesn't seem to match the figures or results. Given the number of experiments it might be helpful to replace one of the many results graphs with a simple illustration showing the various experiments. b) Because the journal format is to put methods after the results (I am not a fan of this) the authors need to adjust their sections to reflect this or else the manuscript is unintelligible if read the way the journal intends. This means that results and figure captions need to be understandable before getting to the methods. Currently, both of these sections refer to experiments that have not been described or even mentioned yet.

I think the most important thing to add to alleviate both issues a and b is to add a paragraph in the Introduction that says: To address question 1 we compared the following experiments: long-term seasonal control (brief description) versus stop-selection (brief description). We then test the prediction that if resources are reduced, the trade-off should magnify the effect leading to.…. We tested this by.…. and so on. In this summary use the same terminology as the rest of the paper. Currently the Introduction fails to introduce half of the experiments.

The Introduction should also mention the importance of delayed density dependence in a bit more detail than in half a sentence.

Methods: The methods seem to be in a random order.

I would restructure the methods to help readability. Start with 5 small paragraphs one for each experiment subtitled by the experiment name. Please include the short-term control in the methods also. The order of the experiments should match the figures, results, and questions. Then discuss data collection and the analyses (ACF… stats…).

Subsection “Multigenerational cycles”, paragraph two, his comes out of nowhere. I though the prediction is being tested by a treatment with a non-breeding season but no selection.

Where all the 5 experiments conducted at the same time? If not there could be differences in the lab environment among experiments that could possibly (although probably unlikely) impact the multigenerational dynamics.

Results: The results jump among experiments and it is unclear which experiments are being discussed. Maybe it is best to explain what happened in the Control Experiment (all results). Then compare it to the other experiments.

At all the times the manuscript should refer to the experiments by their established names, e.g subsection “Long-term control” should say the aseasonal experiment. Why is this result presented before the no-selection experiment? Before Figure 2B?

In the same paragraph there are experiments that were never introduced

Paragraph two? For which experiment? the long or short term?

Subsection “Short-term control”. I don't think initial population size of the experiments is mentioned anywhere except here in the results. Why present a complication concerning the stop-selection experiment before presenting the stop-selection experiment or their results?

Subsection “’Stop selection’ experiment following the order of these figures is exhausting and confusing.

Figures:

Titles don't match content (see Figures 3 and and66 for example).

Consider a large figure that includes all experiments (combines Figures 2 and and33 and and7).7). If the 4 other experiments are all compared to the long-term control then they should all be in the same figure.

Discussion

The Discussion is only big picture and skips any discussion of many of the results (no mention of the reduced food experiment results at all….) This disconnects the work from the bigger interpretation and feels unbalanced.

Discussion paragraph two, I am not sure what 'at ecological scales' means Do you mean timescales? If so there are numerous reviews with hundreds of examples of rapid evolution of ecologically relevant traits (Hendry, Kinison, Reznick, Thompson…). Whether these have been robustly shown to impact ecological dynamics is less clear but see good experimental examples by Willliams Levine 2016, Turcotte et al. 2011 or 2103, Becks multiple, Yoshida et al. 2003 and numerous others by Hairston).

A lot of the Discussion concerns delayed density dependence but this was barely mentioned in the Introduction.

Immediately when I saw the title of the manuscript, and especially the Abstract, I thought of Paul Schmidt's work (U Penn) on seasonal adaptation in Drosophila. Although the authors cited one of their studies I think the links to this body of work should be more prominent as they have discussed evolutionary changes over multiple generations in a lab experiment that are driven by life-history trade-offs (Schmidt Conde 2006, Evolution).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "A fitness trade-off between seasons causes multigenerational cycles in phenotype and population size" for further consideration at eLife. Your revised article has been favorably evaluated by Ian Baldwin (Senior editor) and a Reviewing editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The revised manuscript includes comments and changes in response to reviewer 3, but the comments by reviewer 1 and 2 have not been taken into consideration when revising the manuscript (although commented on when appealing the first decision). The authors should make those changes as well.

2) In addition, the revised version needs further restructuring and clarification as it is still not clear from reading what experiments have been done and why (e.g., add an overview figure).

3) Also, in the paragraph 'Delayed density dependence' of the Results section, the authors compare the results of independent correlations between selection differentials and population sizes. A formal statistical test such as ANCOVA or mixed effect model is need for this comparison and the authors have all the data.

[Editors’ note: the author responses to the first round of peer review follow.]

While the reviewers agreed that this is an interesting manuscript with a timely topic and interesting results, the reviewers also felt that the experiments reported in the manuscript don't provide sufficiently strong support for the conclusions.

We understand the point made by the Editor here but we don’t see any specific comments by the reviewers that indicate that our experiments do not support our conclusions. It seems like reviewer 1 has questioned some of the statistical approaches and addressing these comments would require minor clarifications.

Some of the interpretation of the results are not supported by statistical tests and the study lacks an effort to establish lines at different time points to determine whether genetic changes did occur.

The first point here, that the results are not supported by the statistical tests, arises from several minor comments provided by reviewer 1. Our interpretations of the results are well supported by all of our statistical tests. The second point, made by reviewer 2, has been addressed in detail below but to summarize here: our study shows that cycles can be generated by fluctuating episodes of natural selection and that genetic change (i.e. an evolutionary response to this selection) is not required for population cycles to emerge. This is, in fact, one of the novel aspects of our work.

The overall structure of the manuscript was confusing.

We have made several changes in the manuscript to address this issue (see specific points in the response to the reviewers).

In summary, we didn't feel that the manuscript provides a decisive answer to the question raised.

We respectfully disagree with this point. We asked whether a seasonal fitness trade-off could affect changes in population size. We showed empirically that both aseasonal populations and populations with no or weak viability selection in the non-breeding season had very different population dynamics and phenotypic variation compared to controls. In addition, our mathematical model supports our empirical results. As reported in the paper, when heritability is

0.3 for body size (equal to published estimates) and natural selection occurs in the non-breeding season (differential survival), the model generates cycles in population size. But if we ‘turn-off’ natural selection in the non-breeding season, then the population crashes. Furthermore, the model also simulates a small average decline in body size with time, as seen in the long-term control treatments. Importantly, it is also the case that if heritability is nearly zero (h2=10-5) and natural selection occurs in the non-breeding season, the model generates cycles in population size. This result clearly shows that an evolutionary response to selection is not required for population cycles to emerge. Likewise, when we turn off natural selection in the non-breeding season, there are no cycles and the population crashes. When the heritability of body size is nearly zero there is no average decline in body size, which is expected when there is no between generation response to selection.

For simplicity, we focused on results when h2=0.3 because this reproduced all aspects of the trajectories of population size with time. We have now added the predicted population size based on the mathematical model when heritability is close to zero (Figure 5D).

Reviewer #1:

This is an interesting study examining the joined effects of ecological and evolutionary processes on population dynamics in seasonally fluctuating environments. Using a series of experimental evolution studies with Drosophila populations, Betini and colleagues show how density dependency and viability selection leads to multi-generational cycles of the populations. They make use of the fact that body size affects fecundity (positively) and survival (negatively) and that these two traits trade off between breeding and non-breeding seasons (limited food supply). Specifically, they compare the evolutionary (body size) and ecological changes (density) within populations when manipulating seasonality (absence, presence with very low and low food supply). They found that seasonality leads to multi-generational cycles of 7-9 generations and that these cycles increase with lower food supply in the non-breeding season. An additional mathematical model shows that the multi-generational cycles depend on the trade-off and differential selection between seasons and the density-dependency of growth. Overall, this is an important study showing in an elegant way the underlying mechanism producing multi-generational cycles in populations without interspecific species interactions. It adds to the growing evidence that ecological and (rapid) evolutionary dynamics interact at one timescale. I have however several concerns that need to be addressed:

1) The Results section is not clearly structured and the differences between the different treatments were only clear after reading it several times. I suggest to restructure this section and provide the reader with a short overview of what has been tested in which experiment at the beginning of this section, which should also mention the predictions.

We have made several changes in the manuscript to improve clarity. For example, we have now added a new paragraph in the Introduction to describe the experiments and link them with our predictions. Given that the methods section comes after the presentation of the results, we have also clearly explained the goal of each analysis when presenting the results. For example, on we wrote: “To investigate the presence of viability selection for small body size and whether this selection was density-dependent, we measured female dry weight in 38 generations from 25 different populations.”. Further on we wrote: “We statistically investigated whether fecundity and survival were influenced by densities in past seasons (i.e. delayed density dependence) in all three experiments: ‘long-term control’ and ‘short-term control’ and ‘stop-selection’”.

Furthermore, following the suggestion of the editor, we have also included a new figure that consists of a schematic of the experiments (now Figure 1), as well as the length and the purpose of each experiment.

2) Some statements are not supported by statistical analyses and it is difficult to judge the results:

– The fold changes in population size differences within the multi-generational cycles are not supported by any statistical analyses and for the 'long-term-control' not easy to see in the data. Could you provide some statistical test for the differences in amplitude?

As suggested by reviewer 1, we have modified the way we present the time series for all three experiments (please see Figures 2A, ,3A3A and and4A4A in the main text) to help the reader evaluate these differences in the data. However, we are not aware of a formal statistical test to compare amplitudes. We could use an arbitrary cut-off value and apply a simple t-test. For example, we could take the 5% highest (>281, mean = 299) and lowest (<132, mean = 122) values in the time series and run a simple Welch Two Sample t-test. In the case of the long-term control treatment, there is a statistical difference between the two means (t = -21.152, df = 9.87, p < 0.001). This result is robust for a wide range of cut-off values. However, as we indicated, these are arbitrary cut-offs and we would rather avoid using them.

Figure 7: the ACF plot does not provide evidence for a statistical period in the time series.

We understand the concerns raised by reviewer 1 here, but we believe the important point about the ACF plots is the overall shape of the ACF estimates. In the long-term control, the ACF plot is characterized by an oscillatory decay to zero of the ACF estimates, which is an evidence of a stationary and periodic process (inset plot in Figure 2A in the main text; see our methods section and Turchin and Taylor 1992). This characterization of cyclic dynamic is lacking in the ACF estimates for the stop-selection treatment (inset plot in Figure 1C in the main text). In addition, there is no consensus in the literature about the relevance of statistical tests for periodicity. Different methods or different choices within methods could give different results. For example, in our manuscript, we have eliminated the first 3 generations to avoid transient dynamics. However, we could have eliminated the first 5 or 6 generations, which would result in statistical evidence for a 16 generation cycle in our long-term treatment, but still no evidence of cycle in the stop-selection treatment (see Author response images 1 and and22).

Author response image 1.
ACF plot for the long-term control after eliminating the first 6 generations to avoid transient dynamics.
Author response image 2.
ACF plot for the stop-selection treatment after eliminating the first 6 generations to avoid transient dynamics.

– Could you provide some statistical evidence for the model results (e.g., AFC plots similar to the experimental ones) so the reader can compare better?)?

We have now added the time series for the model results (Figure 5A and D).

– Subsection “Mathematical Model”: how do you judge 'unstable population dynamics' in your model?

When the population crashes to zero during simulations. We have now added this explanation to the main text: “This led to unstable population dynamics (i.e. the population crashed after 10 generations; Figure 5B)”.

– It is interesting to see that the period of the densities (Figure 2) and female dry weight (Figure 6) are very similar (as expected). Is there a correlation between these two time series? And what would one expect?

The correlation between population size at the beginning of the non-breeding season at time t and female dry weight after the non-breeding season at time t is -0.64 (Pearson's product-moment correlation = -0.64; t = -4.48, df = 29, p-value < 0.001). This result supports the prediction from our hypothesis that high population size in the non-breeding causes an overall decline in body size. We have added this result to the main text (subsection “Long-term Control”).

3) The model should be introduced in the main text for the reader to be able to follow better what was manipulated and to test for the effects of viability selection and delayed density effects.

We have now moved the model description to the main text (subsection “Mathematical model”)

4) ACF analyses: please provide more explanation for the ACF plots, so that the readers know how to interpret the maximum in ACF and the significance line you show in the plots.

We have now provided more explanation for the ACF plots in both the Results and the analysis sections (”Documenting multigenerational cycles”). In the analysis section, we wrote: “Cycles can be inferred by investigating the shape of the estimated ACF: population cycles are characterized by stationary periodic dynamics, which result in an oscillatory decay to zero of the ACF estimates.”

Reviewer #2:

I am deeply interested in the question and I found the work to be done well. However, I cannot fully judge the novelty of the work within the ecological literature. The eco-evolutionary part of the work is less well developed as it is not fully clear whether the evolutionary changes were really important in the generation of the longer cycles as the authors claim even though they have some suggestive data. At the end, lack of effort to establish lines at different times of the cycles and determine whether genetic changes did occur I think limits the work. My hunch is that it does not rise to the level of an eLife paper.

From what we can infer from the overall nature of the comments, this is likely the primary reason why our paper was rejected. The reviewer’s feeling is that our paper does not rise to the level of an eLife paper because we have not shown genetic changes, which does not support the eco-evolutionary aspect of the study. We respectfully disagree with this viewpoint. We have shown that cycles can occur by internal processes through a combination of opposing episodes of natural selection between seasons (an evolutionary process) and delayed density dependence (an ecological process). Furthermore, our model shows that accompanying genetic changes are not necessary for these cycles to occur. From an ecological perspective, population cycles are typically thought to be driven by external processes, such as climate or predators, which, from an ecological perspective make our results very significant. Not only that, but we show that these cycles are driven by 3 very common elements found in almost all natural systems: seasonality, density-dependence, and natural selection, which makes our laboratory-based results applicable to wild species. Understandably, the reviewer admits that she/he is not well versed on the ecological significance of these results. However, we do not think that the paper should be rejected based on the grounds that we have not shown concurrent genetic changes because we have provided evidence that genetic changes are not necessary for cycles to occur. This represents a fundamental misunderstanding of our main finding, for which we take responsibility.

To address some of these issues, we have now provided the results with the model with low heritability in the main text, suggesting that cycles can occur in the presence of strong selection and in the absence of genetic changes across generations (subsection “Mathematical model”). We have also emphasized this in the Discussion.

[Editors' note: the author responses to the re-review follow.]

Reviewer #3:

The authors present an impressive set of long lasting and well replicated experiments testing important eco-evolutionary consequences of seasonality. This shows the power of model lab systems! I think the questions addressed were well introduced building up from more targeted mechanisms to more complex eco-evo interactions that could possibly explain well known ecological patterns (multigenerational populations cycles).

My main concern is that the manuscript is overly complicated mostly due to a lack of proper organization. It was very confusing reading this for the first and even the second time. I have made a number of suggestions below that reorder the methods and improve clarity. The interesting manipulation to compare the control to the stop-selection experiment is lost in the confusion that is the methods sections.

As suggested, we have now re-arranged the Introduction, Results and Materials and methods (see details below) and believe the clarity of the manuscript has improved substantially.

Main Concerns:

1) There is a lot of importance given to delayed density dependence in the Discussion (but none in the Introduction). However, there was no formal test of it, which I could recognize, in the actual empirical data even though the model suggests it is very important. I am not an expert in this topic so why wasn't the experimental data tested for it?

As suggested, we have now used mixed effect models to investigate the effect of past densities on fecundity and survival (“Short-term control” and “Documenting multigenerational cycles), and have reported the results in Tables 2 and 3. In the Introduction, we have also clearly explained how delayed density dependence can influence population fluctuations (Last paragraph).

2) The manuscript was very difficult to follow and unnecessarily confusing. I think this stems from two many issues.

a) The number, naming, and ordering of experiments is confusing. There are a lot of similar experiments being presented and it is difficult to follow which is being discussed or how they differ. This is caused firstly by the having their descriptions only appear after the results (see point 2). In addition the order of the experiments doesn't seem to match the figures or results. Given the number of experiments it might be helpful to replace one of the many results graphs with a simple illustration showing the various experiments. b) Because the journal format is to put methods after the results (I am not a fan of this) the authors need to adjust their sections to reflect this or else the manuscript is unintelligible if read the way the journal intends. This means that results and figure captions need to be understandable before getting to the methods. Currently, both of these sections refer to experiments that have not been described or even mentioned yet.

I think the most important thing to add to alleviate both issues a and b is to add a paragraph in the Introduction that says: To address question 1 we compared the following experiments: long-term seasonal control (brief description) versus stop-selection (brief description). We then test the prediction that if resources are reduced, the trade-off should magnify the effect leading to.…. We tested this by.…. and so on. In this summary use the same terminology as the rest of the paper. Currently the Introduction fails to introduce half of the experiments.

We thank reviewer 3 for the helpful comments. We have now added a new paragraph in the Introduction to describe the experiments and link them with our predictions. Given that the methods section comes after the presentation of the results, we have also clearly explained the goal of each analysis when presenting the results.

The Introduction should also mention the importance of delayed density dependence in a bit more detail than in half a sentence.

We agree with reviewer 3 and have now explained how delayed density dependence could influence population fluctuations.

Methods: The methods seem to be in a random order.

I would restructure the methods to help readability. Start with 5 small paragraphs one for each experiment subtitled by the experiment name. Please include the short-term control in the methods also.

As suggested, we have modified the Materials and methods and included the description for the short-term control (subsection “’Stop-selection’ experiment”). We now introduced each experiment in the same order as presented in the Introduction and Results: long-term control, stop-selection and short-term control.

The order of the experiments should match the figures, results, and questions. Then discuss data collection and the analyses (ACF… stats…).

This is now fixed in the manuscript. In the Introduction, Results and Materials and methods, we first present the long-term control, then the stop-selection and finally short-term control.

Subsection “Multigenerational cycles”, paragraph two, this comes out of nowhere. I though the prediction is being tested by a treatment with a non-breeding season but no selection.

We are sorry for the confusion. These experiments were used to show that populations of Drosophila do not show multigenerational cycle under standard rearing conditions (aseasonal experiment). However, as reviewer 3 pointed out, the crucial comparison is between the long-term and the stop-selection. Thus, to improve clarity, we have moved the aseasonal experiment to the appendix. In addition, after re-organizing the results and methods based on the reviewer’s comments, we realized that the reduced-food experiment did not add any new information to our study, so we have decided to remove it from the manuscript.

Where all the 5 experiments conducted at the same time? If not there could be differences in the lab environment among experiments that could possibly (although probably unlikely) impact the multigenerational dynamics.

The stop-selection and short-term control experiments were conducted at the same time. The critical comparison in our study is between the long-term control and the stop-selection. However, because they were conducted at different times there may have been unknown differences in lab conditions, which is why we started the short-term control at the same time as the stop selection treatment. The short-term control produced similar results compared to the long-term control. We have now explained this in the manuscript (Introduction section and subsection “’Stop-selection’ experiment”).

Results: The results jump among experiments and it is unclear which experiments are being discussed. Maybe it is best to explain what happened in the Control Experiment (all results). Then compare it to the other experiments.

As suggested, we have now modified both the results and the figures and presented all the results from each experiment in the following sequence: long-term control, stop-selection and short-term control.

At all the times the manuscript should refer to the experiments by their established names, e.g subsection “Long-term control” should say the aseasonal experiment. Why is this result presented before the no-selection experiment? Before Figure 2B?

As explained above, we have now moved the aseasonal experiment to the appendix.

In the same paragraph there are experiments that were never introduced

Sorry for the mistake. We have now removed this experiment from the manuscript.

Paragraph two? For which experiment? the long or short term?

For the long-term control above, as indicated by the subheading. We believe this is not an issue anymore because now all the results from each experiment are presented separately under different subheadings.

Subsection “Short-term control”. I don't think initial population size of the experiments is mentioned anywhere except here in the results. Why present a complication concerning the stop-selection experiment before presenting the stop-selection experiment or their results?

As suggested by the reviewer, we have explained the differences in initial population size and presented the stop-selection before the short-term control.

Subsection “’Stop selection’ experiment following the order of these figures is exhausting and confusing.

We are sorry for the confusion. Now, each figure presents the main results for each experiment, and the figures match the order they are presented in the Introduction, Results and Materials and methods.

Figures:

Titles don't match content (see Figures 3 and and66 for example).

We have now fixed this in the manuscript.

Consider a large figure that includes all experiments (combines Figures 2 and and33 and and7).7). If the 4 other experiments are all compared to the long-term control then they should all be in the same figure.

As indicated before, each figure presents all results for each experiment, but we have moved the aseasonal experiment to the appendix.

Discussion

The Discussion is only big picture and skips any discussion of many of the results (no mention of the reduced food experiment results at all….) This disconnects the work from the bigger interpretation and feels unbalanced.

As explained above, we have removed the reduced-food experiment from the manuscript. However, we have now discussed some of our results in more detail (Results section).

Discussion paragraph two, I am not sure what 'at ecological scales' means Do you mean timescales? If so there are numerous reviews with hundreds of examples of rapid evolution of ecologically relevant traits (Hendry, Kinison, Reznick, Thompson…). Whether these have been robustly shown to impact ecological dynamics is less clear but see good experimental examples by Willliams Levine 2016, Turcotte et al. 2011 or 2103, Becks multiple, Yoshida et al. 2003 and numerous others by Hairston).

What we meant is that there are few studies showing rapid evolutionary changes that, at the same time, feedback to affect population dynamics. We have now clarified this in the manuscript (Discussion section, paragraph one).

A lot of the Discussion concerns delayed density dependence but this was barely mentioned in the Introduction.

As suggested, we have now explained why delayed density dependence is important in our system (Introduction section).

Immediately when I saw the title of the manuscript, and especially the Abstract, I thought of Paul Schmidt's work (U Penn) on seasonal adaptation in Drosophila. Although the authors cited one of their studies I think the links to this body of work should be more prominent as they have discussed evolutionary changes over multiple generations in a lab experiment that are driven by life-history trade-offs (Schmidt Conde 2006, Evolution).

Thank you for the suggestion. We have now included and discussed this and other studies from the Schmidt’s group in the manuscript (Discussion section).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

1) The revised manuscript includes comments and changes in response to reviewer 3, but the comments by reviewer 1 and 2 have not been taken into consideration when revising the manuscript (although commented on when appealing the first decision). The authors should make those changes as well.

We have now made the changes requested by reviewers 1 and 2 as well as the changes we suggested to the editor in our appeal letter.

2) In addition, the revised version needs further restructuring and clarification as it is still not clear from reading what experiments have been done and why (e.g., add an overview figure).

We have now added an overview figure (now Figure 1) of all three experiments to highlight what experiments have been done and their purpose.

We also want to note that, as requested by reviewer 3, we have made extensive changes in the manuscript to improve clarity. Specifically, we have added a new paragraph in the Introduction to describe the experiments and link them with our predictions. For example, we wrote: “In the second experiment, we tested the role of viability selection during the nonbreeding season by tracking 13 additional populations over 31 generations using a similar protocol to the ‘long-term control’…(the ‘stop-selection’ treatment).” and wrote: “In order to address potential environmental changes in the lab, we conducted a third experiment(…) This 'short-term control' experiment also had 13 replicate populations tracked over 31 generations.”

We have also made changes in the results and analysis to remind the reader about the goal of each experiment. For example, in the results, we wrote: “To investigate the presence of viability selection for small body size and whether this selection was density-dependent, we measured female dry weight in 38 generations from 25 different populations.”

3) Also, in the paragraph 'Delayed density dependence' of the Results section, the authors compare the results of independent correlations between selection differentials and population sizes. A formal statistical test such as ANCOVA or mixed effect model is need for this comparison and the authors have all the data.

We believe there is a misunderstanding here. The delayed density dependence analysis only used population size as a response variable. We have clarified this point in the text (subsection “Delayed density dependence”). The editor might be referring to the analysis in the Appendix 3—table 1 and Appendix 3—figure 1, where we used two linear mixed effect modes to compare the effects of density on body size before and after the non-breeding season. As suggested, we have now used a single linear mixed effect model with body size as response variable and the interaction between density and whether body size was measured before or after the non-breeding season as fixed effect (Subsection “Long-term control”). Generation and population were entered as random effects. The results are consistent with what we presented before: changes in body size were driven by changes in mean dry weight after the non-breeding season rather than changes in the mean dry weight before the non-breeding season.


Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd