One of the most enduring research problems in evolutionary biology is understanding how organisms invest resources into reproduction, and the dynamic selective forces and tradeoffs that bear upon this. In an evolutionary context, a tradeoff incurs a fitness cost when a beneficial change in a trait is coupled with a detrimental change in another (Stearns, 1989). Natural selection has given rise to an incredible diversity of reproductive strategies, yet a tradeoff between offspring size and number (fecundity) is universal (Harper et al., 1970; Fox and Czesak, 2000). A large body of theoretical and empirical work has addressed this tradeoff, but emphasis has been either on plants (by botanists) or on animals (by zoologists); rarely has it been inclusive of both. It is only the rare biologist who studies both plants and animals (e.g., Wallace, 1878; Bradshaw, 1972), and rarer are those who study both systems with an intention to decode ecological and evolutionary tradeoffs (e.g., Darwin, 1859). The fact that work so far has looked at either plants or animals exclusively is understandable given the profound differences in ecological and life-history attributes of the two lineages, which diverged at least 1.5 billion years ago (Wang et al., 1999). Although both plants and animals have diversified tremendously in their reproductive strategies, their fitness is inevitably determined by both quality and quantity of offspring, and is limited by resource availability. Thus, the reproductive strategies of these deeply divergent lineages may have converged due to the underlying similarity in evolutionary pressures and constraints. Indeed, the cardinal theoretical construct for the offspring size-vs.-number tradeoff (Smith and Fretwell, 1974) was mostly agnostic with respect to the two taxonomic groups. Following this seminal paper, there has been a substantial body of work dedicated to understanding variation in offspring size and number, but we still do not fully understand why and how this diversity evolves.

Several studies have established a strong negative correlation between seed size and seed number within and across plants species (Greene and Johnson, 1994; Turnbull et al., 1999; Jakobsson and Eriksson, 2000). In recent decades, a few comprehensive reviews have evaluated the differences in seed size in the context of variation (a) within species (Harper et al., 1970), (b) amongst plant functional groups (Westoby et al., 1992; Coomes and Grubb, 2003), (c) in the productivity of crop plants (Sadras, 2007), (d) in mechanisms of dormancy and germination (Rees, 1996; Finch-Savage and Leubner-Metzger, 2006), and the consequences of these differences for the evolution and diversification of seed plants (Leishman et al., 2000; Moles et al., 2005; Linkies et al., 2010). Similarly, among animals, a negative correlation between offspring size and number (fecundity) is widespread across evolutionary scales—(a) within individuals, (b) among individuals of a population, (c) across populations, and (d) across species and higher-level taxonomic groups—thus there is robust evidence for a tradeoff between offspring size and number (Lawlor, 1976; Berrigan, 1991; Sinervo and Licht, 1991; Roff, 1992; Stearns, 1992; Clarke, 1993; Carrière and Roff, 1995; Mckee and Ebert, 1996; Guntrip et al., 1997; Schwarzkopf et al., 1999; Ernsting and Isaaks, 2000; Fox and Czesak, 2000; García-Barros, 2000a,b; Fischer and Fiedler, 2001; Kolm et al., 2006b).

The quest to explain the changing shapes of the offspring size-vs.-number tradeoff (hereafter size-number tradeoff) within species has led to arguments with mutually exclusive and opposing emphases on the roles played by environmental variation (Janzen, 1969; Kaplan and Cooper, 1984) and genetic/developmental constraints (McGinley et al., 1987). However, strong empirical evidence supporting either argument is scarce. This review revolves around the size-number tradeoff to highlight various unanswered facets of the evolutionary ecology of plant and animal reproductive strategies. We begin by providing an overview of theoretical models of progeny size and fitness relationships, applicable equally to plants and animals. From that point onwards, we develop the discussion on plants and animals separately. After discussing theory, we summarize important life-history constraints on offspring size and number and then explore evidence for parent-offspring conflict at the genetic level in model organisms (Arabidopsis and Drosophila). Finally, we evaluate how the size-number tradeoff is affected by environmental variables. Wherever feasible, we juxtapose available evidence with the predictions of theoretical models.

Reviews focused on animal reproductive strategies already exist (e.g., Fox and Czesak, 2000 for arthropods; Godfray et al., 1991; Monaghan and Nager, 1997 for birds; Einum et al., 2004 for fishes; Bernardo, 1996; Roff, 2002 for animals in general). The animal section aims to distill information from these reviews and from more recent literature, and synthesize the salient points. On the other hand, detailed reviews on plant reproductive strategies are lacking. Therefore, the plant section endeavors to provide a fairly comprehensive review of the literature related to this theme. We aim to exploit commonalities between animals and seed plants to identify broad trends in reproductive strategies of the two lineages. In all of the following, the term “offspring” may refer to seeds in plants, eggs in oviparous animals, and neonates in viviparous animals.

The theoretical underpinnings of the offspring size-number tradeoff have their origins in ornithology. Pioneering work by David Lack in the 1940s on the evolution of clutch size in birds ignited a rich series of studies on vertebrate systems, and later also on invertebrates and plants. Initially, clutch size models for vertebrates and invertebrates were developed independently, both inspired by optimal foraging models (Charnov, 1976; Parker and Stuart, 1976; Stephens et al., 2007), but converged over time (Wilson and Lessells, 1994).

The ground-breaking model by Smith and Fretwell (1974) (hereafter the SF model) assumes that offspring fitness increases with parental investment but experiences diminishing returns (Figure 1A). As a consequence, parental fitness is maximized by an intermediate (optimal) investment in individual offspring. The offspring size corresponding to the maternal parent's optimum (m^*) is inferred by drawing a tangent through the origin to the SF curve (Smith and Fretwell, 1974). Investing more than m^* in individual offspring decreases parental fitness. In contrast, the offspring benefits by procuring as many resources from the maternal parent as possible (pushing m^* toward o^* in Figure 1A). We note that the SF model closely parallels the optimal foraging model of Charnov (1976; The Marginal Value Theorem or MVT). In the MVT model, energy gain per unit time decreases as a function of time spent in a single patch, and the forager is predicted to leave the patch when the foraging rate (the slope of the energy gain curve) equals the mean foraging rate in the environment. The SF model describes parental investment into offspring and profit in terms of parental fitness (through offspring survival), whereas the MVT model describes investment in terms of time as well as energy spent foraging, with energy gained as profit.

Maternal- and offspring-specific optimal sizes, as predicted by the SF model (Figure 1A), create a conflict between parent and offspring, which raises the important issue of maternal vs. zygotic control over reproductive resource allocation (see below; Trivers, 1974; Uma Shaanker and Ganeshaiah, 1997). This parent-offspring conflict becomes even more complex when there is “sibling rivalry” or competition among many offspring of the same mother for maternal resources (Ganeshaiah and Uma Shaanker, 1988), as can happen during development of the offspring on the parent or after birth in animals with parental care. While competition between animal offspring may be active and even include direct fratricide, competition between sibling seeds in plants is more likely to be passive, via the exhaustion the maternal resource pool (but see the section on Endosperms). Importantly, the nutrient drawing ability (Uma Shaanker et al., 1995) of a seed is a function of its size: bigger seeds can invest more resources into resource drawing, and by doing so may overcome maternal controls and create a resource shortage for smaller seeds. When the offspring demands become overwhelming, the parent plant may selectively abort “inferior” embryos to ensure survival of a few offspring (Uma Shaanker et al., 1988; Pélabon et al., 2015; Figure 1B) and thus increase parental fitness (Haig, 1990; Vaughton and Carthew, 1993). Similarly, some animals selectively starve the weakest offspring when resources become limited after oviposition or birth (Lack, 1954; Ricklefs, 1965; Klopfer and Klopfer, 1973).

By selectively aborting or starving inferior offspring (by exerting direct control over offspring number), the parent will also sacrifice some of its control over the size of the remaining offspring. This is because the surviving offspring will receive more resources, grow bigger, and thereby (in the case of seeds) increase their nutrient drawing ability, meaning that zygotic control becomes stronger. On the other hand, shifting resources from inferior to superior offspring is also in the parent's own best interest. Theoretical models show that, when offspring vary in genetic quality, the maternal optimum for the size of high-quality offspring (s^* in Figure 1B) is greater than the classical optimum in the SF model (ms*), and the maternal optimum for low-quality offspring is lower (i^* vs. mi*; Temme, 1986; Haig, 1990). This argument requires some mechanism of post-zygotic recognition of genetic quality by parents. Little is known about such mechanisms, but offspring size is a very reasonable proxy for genetic quality, and indeed, all known instances of selective abortion (in plants) and starvation (in animals) involve elimination of the smallest among competing siblings (e.g., Melser and Klinkhamer, 2001; Mock, 2004). Furthermore, there is evidence that seeds generated from cross-fertilization (presumably of higher quality) tend to be larger than those from self-fertilization (e.g., Wingard, 1927; Shi et al., 2005; Petit et al., 2009). There is also some evidence to suggest that maternal genetic quality determines the extent of selective abortion in plants (e.g., Kärkkäinen et al., 1999). Yet, we do know what role paternal genetic quality plays in determining offspring fitness. More specifically, we do not know how the maternal genome discriminates between offspring that are superior due to better quality eggs and those that are superior due to better quality pollen. The relevance of “genomic imprinting” and the relative influence of paternal and maternal genomes in zygotes is beyond the scope of this discussion.

The SF model depends on two critical assumptions- (a) an optimum fraction of available total resources should be invested in reproduction such that it maximizes lifetime reproduction of a parent and (b) the resource pool is a homogenous entity, which parents invest to determine offspring size. While point (a) is intuitively true, resources are often heterogeneous. For example, plants require both carbon and nitrogen, and the availability of these two components may vary over the course of a season. McGinley and Charnov (1988) developed an extension of the SF model that predicts optimal seed size as a function of the carbon-to-nitrogen (C/N) ratio. In Figure 1C, we provide a slightly modified interpretation of this model, with the aim of highlighting the role of changing C/N ratio through the seasons on seed size. While the distinction between carbon and nitrogen acknowledges resource heterogeneity, the changing C/N ratio identifies resource availability, and the two are inextricably linked. For a given C/N ratio (which the maternal plant cannot alter), there is an optimum allocation to individual seeds, just as in the classical SF model. However, different C/N ratios lead to different optima. More precisely, increasing the C/N ratio is predicted to increase seed carbon content (and hence size) and decrease seed nitrogen content, whereas decreasing the C/N ratio has the opposite effect of reducing seed size while increasing nitrogen content.

While some empirical studies show limited or no effect of C/N ratio on seed sizes (Uma Shaanker et al., 1988; Lalonde, 1991), the most convincing evidence supporting a role of C/N comes from elevated CO₂ experiments (atmospheric [CO₂]> 700 ppm) on crop plants. Both seed size and numbers generally increase after CO₂ fertilization (e.g., Jablonski et al., 2002). Increases in seed size are attributed to greater carbon fixation through photosynthesis under elevated CO₂, while increases in seed number are attributed to increased mobilization of nitrogen from leaves to seeds. This is surprising given that nitrogen is often a limiting factor for plant growth, and this limitation is expected to be enhanced under elevated CO₂. But, the results are not that surprising when we take into account greater water use efficiency and greater photosynthetic rates in plants under elevated CO₂, and this is despite decreases in leaf nitrogen content (Jablonski et al., 2002; Hikosaka et al., 2011). The second and third crop of flowers and seeds may become small since carbon fixed through photosynthesis declines with plant age (because of senescence), but seed numbers may be maintained since nitrogen supply can be ensured through continued remobilization of nitrogen from senescing leaves to seeds. A decreasing C/N ratio over the course of a season may thus explain why, in many species, seed size decreases over the course of a season (e.g., Salisbury, 1942; Cavers and Steel, 1984; McGinley and Charnov, 1988).

It needs to be emphasized that resource or nutrient limitation is one of many different types of sub-optimal (harsh) environments experienced by plants, and plant response may be different for a different type of harsh condition. A key prediction of the SF model is that the maternal optimum m^* is independent of the total amount of resources invested in offspring (i.e., reproductive effort; Venable, 1992) but is likely to vary between different environments. Hence, offspring size should be adaptively adjusted by the mother due to G × E interactions. A simple example is shown in Figure 1D (after Einum et al., 2004), where an unfavorable environment (environment 1 in Figure 1D) is characterized by a higher minimum viable size and lower overall offspring fitness, which selects for larger offspring (increased m^*).

In the classical SF model, the offspring size-vs.-fitness function is a horizontal line below the viability threshold, followed by a unipolar convex curve representing diminishing returns. However, it seems unrealistic that diminishing returns should set in immediately after the size viability threshold, especially since increasing seed size increases nutrient drawing ability (which suggests that the curve should have an early exponential—or at least concave—part). Thus, the SF curve is more appropriately represented as a sigmoid (Figure 1E). Therefore, the offspring size-vs.-fitness relationship might be more appropriately represented by a sigmoid function (Figure 1E). Like the classical SF function, a sigmoid predicts an intermediate maternal optimum situated in the convex part of the function. Yet, exponential decrease or increase in fitness of seeds belonging to size classes close to a maternal optimum (as seen in a sigmoid) suggests strong selection pressures on seed size optimization. The sigmoid accommodates the idea that maternal control increases above a certain offspring size (for animal examples see Levitan, 2000; Reeve et al., 2000; Jørgensen et al., 2011). The sigmoid is consistent with the concept of sibling rivalry (depressing the fitness of small offspring) and parental preference for superior offspring (maternal optimum s* is greater than m* in Figure 1E). However, the assumption that big offspring always show greater fitness compared to small offspring may not hold in some special cases where developmental and habitat constraints play a role. Hendry et al. (2001) evaluated the impact of developmental stage (pre- vs. post-hatching) on offspring fitness as a function of size. They proposed a version of the SF model where eggs larger than an optimum size (o^*) experience diminishing fitness (Figure 1F). Theoretically, larger hatchlings from bigger eggs are expected to show higher fitness in favorable environments. Nonetheless, big eggs may not survive to hatchling stage either due to space constraints for incubation (lack of parental care) or suffer from asphyxia in sub-optimal aquatic habitats. In such cases, optimum offspring size for maximizing maternal and offspring fitness are likely identical (contrast Figures 1D,F). We will reassess shifts in maternal optima in greater detail under the empirical section dealing with environmental responses of offspring size and number tradeoff.

The origin(s), evolution and diversity of seeds themselves are thought to be critically important in the domination of seed plants in terrestrial ecosystems. Seeds ensured the safety of the stationary female gametophyte (ovule), and led to novel mechanisms of long-distance dispersal (Nathan et al., 2008). In conjunction with breakthrough adaptations such as efficient xylem transport and stomatal mechanisms (e.g., Crepet et al., 1991), diverse mechanisms of seed tolerance to desiccation (Baskin and Baskin, 1998) allowed seed plants to overcome dispersal limitations on land and colonize drier geographic territories away from the tropical wet forests. Seed diversity is dramatic across the plant kingdom, and principal aspects of this variation are size and number.

Plant form, habitat, ecophysiology, and reproductive modes are some of the higher-order life- history determinants of seed size and number, and these factors have both restricted variation and led to common trends within and between plant species and families. While these constraints are not directly related to the theoretical models presented earlier, they have a central place in our discussion, since it may turn out that explanations for observed trends in seed plants may lie in these deep life history constraints on reproductive strategies.

At a macrogenomic level, a positive correlation between genome size and seed size is observed both within and between phylogenetic clades, especially in angiosperm lineages with small genomes and within the pine clade in gymnosperms (Thompson, 1990; Beaulieu et al., 2007). Existing arguments to explain co-divergence of genome size and seed size are inadequate and inconclusive (Beaulieu et al., 2007), despite a broad allometric agreement between genome size and other phenotypic properties, stronger at the cellular level and weaker at higher bodily organization in plants (Knight and Beaulieu, 2008).

While it is important to understand genome-phenome associations across plant families, answering such questions may not address the diversity in seed size and numbers at the level of species and ecotypes, which are subtler and likely regulated by microgenomic interactions. We do not yet have enough information, especially in non-model plant systems, for an overarching genetic paradigm of seed size, shape and number regulation (Orsi and Tanksley, 2009; Gegas et al., 2010; Fang et al., 2012). However, some insight is offered by model systems such as Arabidopsis thaliana (hereafter simply Arabidopsis). Arabidopsis has a fast generation time, and possesses a vastly under-appreciated natural genetic and phenotypic variation (Koornneef et al., 2004; Singh et al., 2015). However, the microscopic size of Arabidopsis seeds poses methodological difficulties in characterizing seed phenotype, which prevented it from being a model to investigate evolutionary ecology of plant reproductive strategies. The advent of digital techniques to precisely quantify seed characteristics has led to closer examinations of seeds in Arabidopsis mutants (e.g., Herridge et al., 2011). In their study of 64 different mutant lines, Van Daele et al. (2012) reaffirmed a negative correlation between seed size and seed number. A re-examination of the same data shows that the mean seed size in Arabidopsis has limited variation (0.1 ± 0.02 mm) across all mutant lines (Figure 2B). Total seed number per plant showed a 10-fold range (400–4,000, including outliers) while total seed weight per plant showed a 5-fold range (50–250 mg). However, there was no correlation between total seed weight and total seed number or mean seed size (Figures 2B,C): plants with high total seed weight had heavier (denser), not more or larger, seeds than those with low total seed weight (probably due to differences in oil content, see section on nutrient quality and quantity above). We also found that there was zero correlation between total rosette (leaf) area and total seed weight and similarly there was no relationship between rosette area and total seed number per plant (Figures 2A,D). While photosynthesis and general metabolism in leaves may have been compromised in some Arabidopsis mutants leading to differential seed production, the complete lack of any relationship between leaf mass and seed mass or number remains to be explained.

Research on genetic controls on seed size has been reviewed extensively in recent years (Sundaresan, 2005; Linkies et al., 2010; Sun et al., 2010; Li and Berger, 2012). Instead of providing a descriptive account of genes regulating seed size, we present an illustrative summary of seed size variation in Arabidopsis mutants (Figure 3; Table 1). The data suggest that genes acting in maternal tissues have less impact on seed size than genes operating directly in zygotes (embryos), which can be partly due to the bias in the data set, since we know more about genes acting in maternal tissues. It is equally significant to note that the percent increase in seed size due to loss-of-function mutations in genes whose normal function is to restrict seed size (irrespective of whether they act in maternal or zygotic tissue) is significantly greater than the percent decrease in seed size due to loss-of-function in genes that are actively promoting larger seeds.

Maternal genes involved in restricting the size of inferior (small) seeds will certainly be different from those involved in diverting more resources toward superior (large) seeds. Furthermore, it is well established that during plant development (also in animals), maternal genetic controls wean and are successively replaced by zygotic controls (e.g., Baroux et al., 2008). What will be important to establish, however, is whether increases in the size of superior seeds are due to a weakening control of maternal genes restricting seed size or due to overexpression of zygotic genes promoting seed size (Figure 3).

Arabidopsis mutants have provided valuable insights into specific genes and gene-networks that control seed size (North et al., 2010). In this context, it is important to distinguish between genes that exclusively control seed size and those that indirectly influence seed development through their effects on global resource allocation and organ size within a plant (Mizukami and Fischer, 2000; Adamski et al., 2009; Van Daele et al., 2012). Analysis of genes directly involved in seed-size control in Arabidopsis mutants invariably show that variation in seed size is not linked to seed number (Zhou et al., 2009); In other words, mutations altering seed size do not pleiotropically alter seed numbers. QTL analysis of natural variation also suggests that genetic factors controlling seed size and seed numbers are situated in different parts of the Arabidopsis genome (on different chromosomes) and can evolve independently (Gnan et al., 2014).

We know a lot less about the genes involved in regulating seed number than about genes regulating seed size. Some of the genes regulating seed numbers are associated with regulation of sizes of floral meristems and inflorescences. Maize mutants producing more seeds (kernels) per row do not produce seeds that are smaller than those from wild, control groups (Bommert et al., 2013), suggesting that the tradeoff between seed size and number within a species need not hold always. It is also possible that genes involved in initial floral differentiation predetermine the number of ovules per ovary and number of flowers per inflorescence, thus impacting seed number through pleiotropic effects (Huang et al., 2013), without causing an evolutionary tradeoff with size. Some other genes such as APETALA2 and AHKs (cytokinin signal receptors) have the most dramatic impact on not only seed size but also seed number through their global impact on resource reallocation within a plant (Ohto et al., 2005; Riefler et al., 2006).

Overall, our findings in Arabidopsis are consistent with the notion that diversity in seed size is not as great as that in seed number per plant within a species (Harper, 1967), suggesting seed number to be less responsive to selection. This finding also suggests that the SF model may be right in assuming that the optimization of offspring size is independent of selection pressures on reproductive effort (Venable, 1992; but see the section on environmental factors). Thus, mechanisms to restrict seed size (and not numbers) might have become prominent and diversified through the course of seed plant evolution.

Many functional traits affecting the size-number tradeoff are strongly influenced by environmental factors. These traits can respond to environmental changes either through phenotypic plasticity, or through rapid evolution, and they do not have the evolutionary inertia that is typical of most life history traits (discussed earlier). While short-term physiological or phenotypic adjustments can often can be classified as stress- responses, they also provide new raw material for natural selection through epigenomic changes inherited by offspring produced while the sub-optimal conditions prevail. If the same environmental conditions persist over a longer term (several thousand generations), for example due to a shift in climate, then in an evolutionary sense what was stressful to the ancestors of the past becomes the new norm for those progenies that manage to survive until the present. Therefore, analysis of stress responses is relevant also in the light of larger-scale evolutionary trends. In the following section, we discuss various environmental factors that influence plant reproductive strategies.

Animals show a general trend toward reduction of fecundity and concomitant increase in offspring size associated with increasing organismal complexity (Brockelman, 1975). Both fecundity and offspring size generally increase with the size of the mother (Parsons, 1964; Roff, 1992; Honěk, 1993; Fox and Czesak, 2000; Shanbhag et al., 2003; Kolm et al., 2006a; Rollinson and Rowe, 2016), although exceptions exist (e.g., Fischer and Fiedler, 2001). Species for which no size-number tradeoff has been found typically show considerable variation in reproductive effort (Fox and Czesak, 2000). Many studies have reported conservatism in egg size, with greater variation in clutch size under varying environmental conditions (birds: Lack, 1954, 1968; fish: Scott, 1962; Hester, 1964; Fleming and Gross, 1990; mammals: Jordan et al., 1967; Stevenson-Hamilton, 2012; readers are referred to other focused reviews for more detailed discussions of the literature; vertebrates: Brockelman, 1975; Winkler and Walters, 1983; invertebrates: Wilson and Lessells, 1994; Fox and Czesak, 2000). These results suggest that in animals, selection primarily optimizes offspring size, again in accordance with optimality models, rather than fecundity. However, exceptions exist where offspring number, rather than size, is optimized. For instance, in reptiles, although large clutch sizes are ancestral, invariant clutch size with a small number of eggs has evolved independently in some lineages (Kratochvíl and Kubička, 2007; see also Sinervo and Licht, 1991; Goodman, 2006), and egg size can vary in such lineages (Shanbhag et al., 2003). The evolutionary advantages of invariant clutch size are not well-understood.

As in the case of plants, morphological and other constraints may limit fecundity or offspring size in animals. Our analysis focuses on data rich groups (reptiles and birds).

Egg size has generally been shown to have high heritability in animals, for instance, in birds such as the great tit (Parus major; Ojanen et al., 1979) and turkey (Meleagris gallopavo; Nestor et al., 1972 and references therein), as well as lizards (Sinervo and Doughty, 1996). Economically important egg traits in chicken (yolk volume, egg weight) are known to have high heritability (Olsen and Knox, 1940; Shultz, 1953; Zhang et al., 2005). We know little about the genetic control of egg size and number in non-model animal taxa, and most of what we know comes from the common fruit fly Drosophila melanogaster, for which the egg size-number tradeoff is well-known (Mackay et al., 2012; Jha et al., 2015).

The first attempt to understand the genetic regulation of egg size in D. melanogaster was by Warren (1924), who found that egg size varied considerably across mutant lines. Based on data from controlled crosses, he concluded that egg size is polygenically controlled by genes on all four chromosomes. Jha et al. (2015) adopted a whole-genome resequencing approach on replicated experimental-evolution lines selected for large and small egg sizes, respectively. Corroborating Warren's findings, Jha and colleagues report that egg size is polygenically controlled by at least 200 genes. Their analyses detected signatures of selection on specific genomic regions linked to egg size, as well as directional selection on known genes, including those involved in regulation of cell size, egg growth and development. They identified 193 candidate genes, and highlight the role of two genes, Plod and Iola, using crosses of mutant lines. RNA-i knockdown experiments have previously shown that Plod controls egg length (Lerner et al., 2013). The list of candidate genes identified by Jha and colleagues promises to lay the foundation for further studies that can elucidate the genetic basis of egg-size variation in D. melanogaster. Schwarzkopf et al. (1999) selected D. melanogaster for large and small egg size relative to female body size, and found that egg size responded strongly in both directions. Surprisingly, a negative correlation between absolute egg size and fecundity was observed in the large-egg selected line but not in the small-selected lines or the control (without selection). There was no genetic correlation between relative egg size and relative fecundity in any selection line, suggesting that there was no negative genetic correlation between egg size and number (a trend indicating similarity with Arabidopsis). However, as the authors discuss, it is possible that such a correlation may be obscured by complex phenotypic interactions with other life history traits. Furthermore, total reproductive allocation decreased in the small-egg selected line, which suggests that resource allocation to reproduction may be dependent on offspring size. The intriguing results in this study further underscore the potential of experimental evolution to understand genetic controls of offspring size variation.

Differences in environments can select for variation in offspring size, and the fitness difference between larger and smaller offspring is generally greatest in poorest quality or stressful environments (see Fox and Czesak, 2000). Generally, more stressful environments tend to select for larger offspring size in animals. We first discuss selection pressures exerted by components of the abiotic environment. In arthropods, these include soil depth (Carrière et al., 1997), temperature (Perrin, 1988; Avelar, 1993; Azevedo et al., 1996; Crill et al., 1996; Ernsting and Isaaks, 1997, 2000; Fischer et al., 2003) and photoperiod (Ernsting and Isaaks, 2000). A given environmental variable can have a positive or negative correlation with offspring size in different species. For instance, populations of Sceleporus lizards show a pattern of increased egg size with decreasing temperature, which manifests across both altitudinal and latitudinal gradients (Sinervo, 1990). Similar results were reported in various other species, including Drosophila (Azevedo et al., 1997), where wild and experimental laboratory populations evolving at lower temperatures produced larger eggs. The same pattern has been repeatedly found in various other species, while a converse pattern of larger eggs at higher temperatures has also been reported, albeit less commonly (see Fox and Czesak, 2000). One argument for the prevalence of larger eggs at lower temperatures is that survival over longer developmental periods at lower temperature necessitates more resources (Kolding and Fenchel, 1981; Yampolsky and Scheiner, 1996). Thus far, experimental evidence for the adaptive significance of variation in egg size across different conditions is limited, possibly because experimental conditions do not adequately mimic the complexity of natural environments. Other biotic factors that strongly select for large or small differential offspring size include food quality (e.g., Brody and Lawlor, 1984; Braby, 1994), competition (e.g., Parker and Begon, 1986), inbreeding (e.g., Duthie et al., 2016), limited dispersal (e.g., Kuijper and Johnstone, 2012) and predation (e.g., Kerfoot, 1974; Ernsting and Isaaks, 1997) (see also Table 2 in Rollinson and Rowe, 2016). In particular, increased competition generally selects for larger offspring size, while size-dependent predation may select for either smaller or larger offspring.

Seeds (in plants) and eggs (in animals) have evolved independently, but serve the same function. Consequently, strikingly similar adaptive strategies have emerged in parallel in plants and animals. The current review brings together theoretical and empirical work on plants and animals, at both micro- and macro-evolutionary scales, focusing on one of the most universal evolutionary tradeoffs. Although the diversity in reproductive strategies is almost infinite, it is striking that limitation of resources available for reproduction is ubiquitous and selection has resulted in surprisingly similar solutions in animals and plants. In particular, we identify the following similarities, which enable an all-encompassing view of reproductive strategies in plants and animals alike:

Offspring size is generally less variable than offspring number, suggesting that optimization of size is generally under stronger selection than optimization of numbers. Costs of increasing offspring size are likely greater than increasing offspring number in both plants and animals.

Despite these similarities, there are also significant idiosyncratic differences between reproductive strategies of animals and plants, which demand further exploration.

While there is overwhelming experimental evidence supporting the assumption that offspring born from larger eggs or seeds have higher fitness, there are various scenarios where this assumption may not hold true due to both intrinsic (genetic) and extrinsic (environmental) regulatory factors. We still do not understand the genetic architecture of parent-offspring conflict during instances of selective abortion in both animals and plants. Exciting mechanistic and field investigations of offspring size-number tradeoffs under sub-optimal environmental conditions are anticipated ahead.

UK proposed the review idea to KGSD. KGSD synthesized and analyzed the literature on plant reproductive strategies and optimization models, designed the illustrations, and developed arguments concerning genetic controls on offspring size. UK analyzed the animal literature and conceived the syntheses. Both the authors contributed to the manuscript writing process.

This work was supported by the grant BT/PR12720/COE/34/21/2015 (Project A3) from the Department of Biotechnology, Government of India.