Insect egg size and shape evolve with ecology, not developmental rate

The evolution of organism size is hypothesized to be predicted by a combination of development, morphological constraints, and ecological pressures. However, tests of these predictions using phylogenetic methods have been limited by taxon sampling. To overcome this limitation, we generated a database of more than ten thousand observations of insect egg size and shape from the entomological literature and combined them with published genetic and novel life-history datasets. This enabled us to perform phylogenetic tests of long-standing predictions in size evolution across hexapods. Here we show that across eight orders of magnitude in egg volume variation, the relationship between egg shape and size itself evolves, such that predicted universal patterns of scaling do not adequately explain egg shape diversity. We test the hypothesized relationship between size and development, and show that egg size is not correlated with developmental rate across insects, and that for many insects egg size is not correlated with adult body size either. Finally, we show that the evolution of parasitism and aquatic oviposition both help to explain the diversification of egg size and shape across the insect evolutionary tree. Our study challenges assumptions about the evolutionary constraints on egg morphology, suggesting that where eggs are laid, rather than universal mathematical allometric constants, underlies egg size and shape evolution.


24
The evolution of organism size is hypothesized to be predicted by a combination of development, morphological 25 constraints, and ecological pressures. However, tests of these predictions using phylogenetic methods have been 26 limited by taxon sampling. To overcome this limitation, we generated a database of more than ten thousand 27 observations of insect egg size and shape from the entomological literature and combined them with published 28 genetic and novel life-history datasets. This enabled us to perform phylogenetic tests of long-standing predictions 29 in size evolution across hexapods. Here we show that across eight orders of magnitude in egg volume variation, 30 the relationship between egg shape and size itself evolves, such that predicted universal patterns of scaling do not 31 adequately explain egg shape diversity. We test the hypothesized relationship between size and development, and 32 show that egg size is not correlated with developmental rate across insects, and that for many insects egg size is not 33 correlated with adult body size either. Finally, we show that the evolution of parasitism and aquatic oviposition both 34 help to explain the diversification of egg size and shape across the insect evolutionary tree. Our study challenges 35 assumptions about the evolutionary constraints on egg morphology, suggesting that where eggs are laid, rather than 36 universal mathematical allometric constants, underlies egg size and shape evolution.  There are 10,449 published egg descriptions in the database; the plot shows the 7,935 descriptions with both volume and aspect ratio data. B, Egg shape is described in the database with three parameters (aspect ratio, asymmetry, and angle of curvature) calculated from the measurements shown in purple (see text for details). C, Images of six example eggs, arranged from largest to smallest and oriented with the axis of rotational symmetry vertical: 1. Earth-borer beetle Bolboleaus hiaticollis 2 , 2. Tropical carpenter bee Xylocopa latipes 3 , 3. Kissing bug Rhodnius robustus 4 , 4. Many-banded daggerwing butterfly Marpesia chiron 5 , 5. Tephritid fruit fly Anastrepha distincta 6 , 6. Parasitoid wasp Platygaster vernalis 7 . Hypothetical eggs for each hypothesis are shown as schematics with length axis vertical. D, Each hypothesis (solid line = A, dotted line = B, dashed line = C) predicts a different value of the scaling exponent-that is, the slope of the regression between egg length and width in log-log space. E, The distribution of egg length and width in log-log space. The dashed black line represents a hypothetical 1:1 relationship (isometry, hypothesis C). Solid colored lines are the phylogenetic regressions for seven clades, and each colored point is a randomly selected representative egg for a genus. F, Distribution of the scaling exponents for the seven monophyletic insect clades included in this analysis, calculated over the posterior distribution of trees. We resampled tree topology and genus-level representatives from the egg morphology dataset 100 times, calculating a scaling exponent each time. White lines, boxes, bars, and dots represent median, 25-to-75th percentiles, 5-to-95th percentiles, and outliers, respectively. Asterisks indicate that the relationship between length and width is significant (p-value threshold <0.01, exact values in Table S6), and double-dagger indicates the relationship is not statistically distinguishable from isometry (p-value threshold >0.01, exact values in Table S7). In E and F the colors correspond to the clades shown in Fig. 1A.

Evolutionary patterns of egg shape and size
Two opposing hypotheses based on predicted geometric constraints have been proposed to explain the evolutionary 75 relationship between propagule shape and size. Hypothesis A: When eggs evolve to be larger, they get wider (increases 76 in egg size are associated with decreases in aspect ratio) 24,25 . This hypothesis predicts a reduction in relative surface 77 area as size increases, which has been proposed as a solution to the presumed cost of making eggshell material 25 .

78
Hypothesis B: When eggs evolve to be larger, they get longer (increases in egg size are associated with increases in 79 aspect ratio) 19,20,25 . This hypothesis predicts a reduction in relative cross sectional area as eggs get larger, which has 80 been proposed as a solution to the need for eggs to pass through a narrow opening during oviposition 19,20 .

81
To test these hypotheses about the physical scaling of size and shape, we first modeled the individual evolutionary   Figure 3: Rate of embryogenesis does not co-vary predictably with egg size. A, Mature eggs undergo embryonic development, hatch, and grow into adult insects. We define duration of embryogenesis as the time elapsed from egg-laying to the point at which the nymph (in Hemimetabola) or larva (in Holometabola) emerges from the egg. B, Duration of embryogenesis, adjusted for incubation temperature (hours, plotted on a log scale), compared to egg volume (mm 3 , plotted on a log scale). Each colored point represents an insect species for which duration of time-to-hatching has been reported (see Supplemental Information for sources), and egg morphological data are contained in our database 31 . When phylogenetic relationships are taken into account, there is no significant relationship between egg volume and duration of embryogenesis. C, Egg volume (mm 3 , plotted on a log scale) compared to adult body volume, calculated as body length cubed (mm 3 , plotted on a log scale). The dashed black line shows an example 1:1 relationship (isometry), solid colored lines are the phylogenetic regression for the seven clades included in this analysis, and colored points are family-and order-level averages for egg size and median adult insect size. D, The distributions of the allometric exponents for the seven monophyletic insect clades included in this analysis. Asterisks indicate that the relationship between length and width is significant (p-value threshold <0.01, exact values in Table  S12), and double-dagger indicates the relationship is not statistically distinguishable from isometry (p-value threshold >0.01, exact values in Table S13). Colors and labels in B-E correspond to the clades shown in Fig. 1A.
volume and duration of embryogenesis yields the previously reported positive relationship 29 (Fig. S22). However, a 120 linear regression that does not account for phylogenetic relationships is inappropriate for this analysis due to the 121 covariance of traits on an evolutionary tree 40 . When we accounted for such potential phylogenetic covariance of 122 data, we found that there is no significant relationship between egg size and duration of embryogenesis across insects, 123 such that eggs of very different sizes can develop at a similar rate and vice versa (0.08 < p-value < 0.26; Fig. 3B, S21, 124 Table S21). These results suggest that the often-invoked trade-off between size and development 26-29 does not hold 125 across insect species. 126 We also tested the hypothesis that the size of the egg has a positive relationship with adult body size. Previous work 127 reported this relationship in subsets of insects, and moreover suggested that smaller insects lay proportionally larger  Table S23). In general, the predictive power of the relationship between 136 body size and egg size is low: average egg volume can vary by up to four orders of magnitude among species with 137 similar body size (Fig. 3C). This decoupling of both body size and duration of embryogenesis from egg size evolution 138 suggests that egg diversification has not been universally constrained by development across insects. and asymmetrical eggs 19 . We asked whether an analogous relationship exists between insect flight capability and egg 143 shape. Unlike birds, insects have undergone many hundreds of evolutionary shifts to flightless and even wingless 144 forms 43 . We focused on two clades in which the patterns of flight evolution have been extensively studied. Stick 145 insects (Phasmatodea) have flightless and wingless species 44,45 (Fig. S17), and many butterflies (Lepidoptera) show 146 migratory behavior 46 , which could be considered a proxy for increased flight capability relative to non-migratory 147 taxa (Fig. S17). We found that, in contrast to birds, evolutionary changes in flight ability in these two insect clades 148 were not associated with changes in egg shape (OU model with multiple optima per regime; ∆AICc < 2, exact 149 values in Tables S17, S18).

150
Like flight capacity, the microenvironment that eggs experience varies widely across insects, including being exposed 151 to air, submerged or floating in water, or contained within a host animal 15 (Fig. 4A). Each of these microenvi-152 ronments places different demands on the egg, such as variable access to oxygen and water during development 22 . 153 Preliminary studies in small groups of insects suggested that evolutionary changes in oviposition ecology and life 154 history may drive the evolution of egg size and shape 17,30 . To test this prediction robustly across all insects, we  Pol.
Ant. < * < * < * < * Figure 4: Shifts in oviposition ecology are associated with morphological changes in insect eggs. A, Laying eggs within an animal host (orange; e.g parasitoid wasps) and in water (blue; e.g. mosquitoes) are two modes of insect oviposition ecology. Other oviposition substrates (e.g. terrestrial) are shown in gray (see SI for further details on oviposition substrate classification). B, Ancestral state reconstruction of each oviposition ecology trait reveals that they have arisen multiple times in distantly related insect lineages (see Fig. S17 and S18 for full phylogeny). C-F, The distribution of four egg morphology parameters across insect clades, colored by oviposition ecology and plotted by phylogenetic group: C, volume (mm 3 , plotted on log scale); D, aspect ratio (unitless, plotted on log scale); E, asymmetry (unitless); and F, angle of curvature (degrees). The x-axes of D-F include theoretical egg silhouettes shown with the length axis vertical. Asterisks indicate that the model accounting for oviposition ecology (Ornstein-Uhlenbeck process with multiple regimes) fits the data better (∆AICc >2, exact values in Table S14-S19) than a non-ecological model. Clade labels in B-F are abbreviations of clades shown in Fig. 1A.
compiled records on two specific oviposition ecology modes that have been extensively studied: oviposition within 156 an animal host (which can be in the host adult body or in the host egg, called internal parasitic oviposition) and 157 oviposition in or on water. For each mode we reconstructed ancestral changes along the insect phylogeny, and found 158 that both aquatic and internal parasitic oviposition have been gained and lost multiple times independently ( Fig.   159 4A, B, S15, S16). This extensive convergent evolution allowed us to perform a strong test of whether egg size and 160 shape evolution is predicted by the evolution of oviposition ecology. 161 We found that the evolution of new oviposition environments was linked to changes in egg size and shape. Models 162 that accounted for shifts to new oviposition environments better explained egg size and shape distributions than 163 models that did not (OU multistate model, ∆AICc > 2, exact values in Tables S14-S19). Specifically, shifts to 164 aquatic oviposition were significantly associated with the evolution of smaller, wider eggs ( Fig. 4C-D, Tables S11, 165 S14), and shifts to internal parasitic oviposition were significantly associated with smaller, more asymmetric eggs Table S11). Moreover, we note the smallest eggs in the dataset are from wasps with internal parasitic 167 oviposition that develop polyembryonically (i.e. multiple embryos form from a single egg 47 ; Fig. S7). Neither 168 ecological change was associated significantly with consistent changes in the scaling relationship between size and 169 shape (Fig. S18). These results were robust to uncertainty in how taxa had been classified for oviposition ecology; 170 using broader ecological definitions, such as combining endo-and ectoparasites or semi-aquatic and riparian insects, 171 did not change these results (Tables S12, S15, S13, S16). Taken together, these convergent evolutionary events are 172 evidence that the microenvironment experienced by the egg following oviposition plays an important in role in the 173 evolution of egg size and shape. 174 We have shown that insect eggs present an ideal example case for testing the predictability of macroevolutionary 175 patterns in size and shape. By comparing egg size and shape across taxa, we find that prevalent assumptions about macroevolutionary studies, establishing computational tools and methods that can be followed in future work.

181
Finally, we provide evidence that the ecology of oviposition drives the evolution of egg size and shape. the egg is a flattened ellipsoid 7 ), we defined width as the wider of the two diameters, and breadth as the diameter 305 perpendicular to both the width and length.

306
Egg volume: Volume (mm 3 ) was calculated using the equation for the volume of an ellipsoid: 1 6 lw 2 , following 307 previous workers 8, 9 .

308
Egg aspect ratio: Aspect ratio was calculated as the ratio of length to width.

309
Egg asymmetry: Asymmetry was calculated as the ratio between the two egg diameters at the first and third quartile 310 of the length axis. The first quartile was always defined as the larger of the two diameters, such that asymmetry is 311 always between zero and one.

312
Angle of egg curvature: The angle of curvature was measured as the angle (degrees) of the arc created by the endpoints 313 and midpoint of the length axis.  The ancestral state of volume, aspect ratio, and angle of curvature were mapped on the summary phylogeny using 345 the R package phytools 23 (version 0.6-44, function contMap). Evolutionary rate regimes of volume, aspect ratio, 346 and the angle of curvature were fit on the summary phylogeny using the program BAMM 24,25 (version 2.5.0, R 347 package BAMMtools verison 2.1.6, setBAMMpriors, prior for expected number of shifts set to 10, for 10,000,000 348 generations).

349
Following ancestral state reconstruction for ecological regimes (see above), for each trait (volume, aspect ratio, were compared using the R package OUwie 26 (version 1.50).

353
All evolutionary regression analyses were performed using a phylogenetic generalized least squares (PGLS) approach 354 in the R packages ape 27 (version 5.0, correlation structure = corBrownian) and nlme 28 (version 3.1-131.1). Given 355 that the EB models best fit the data, we also tested a corBlomberg correlation structure, which invokes an accelerating- Antliophora). In addition, the scaling exponent between egg length and width was calculated for each monophyletic 368 group of taxa with more than 20 tips but fewer than 50.