Genomic imprinting drives eusociality

The origin of eusociality, altruistically foregoing personal reproduction to help others, has been a long-standing paradox ever since Darwin. Most eusocial insects and rodents likely evolved from subsocial precursors, in which older offspring “helpers” contribute to the development of younger siblings without a permanent sterile caste. The driving mechanism for the transition from subsociality (with helpers) to eusociality (with lifelong sterile workers) remains an enigma because individuals in subsocial groups are subject to direct natural selection rather than kin selection. Our genomic imprinting theory demonstrates that natural selection generates eusociality in subsocial groups when parental reproductive capacity is linked to a delay in the sexual development of offspring due to sex-antagonistic action of transgenerational epigenetic marks. Focusing on termites, our theory provides the missing evolutionary link to explain the evolution of eusociality from their subsocial wood-feeding cockroach ancestors, and provides a novel framework for understanding the origin of eusociality.


Introduction 39
The origin of eusociality poses a major challenge to Darwin's theory of natural selection 40 for altruistic behavior toward close relatives. It has long been studied within the paradigm 44 of inclusive fitness and kin selection (Hamilton, 1964). However, the role of kin selection 45 in the origin of eusociality remains controversial (Wilson and Hölldobler, 2005). 46 Explanations of the origin of new traits are faced with the ''selection paradox,'' the notion 47 that selection cannot act on traits that do not yet exist, and therefore cannot directly 48 cause novelty (Jacob, 1977;Müller and Wagner, 1991). While biologists have made 49 great progress in the last 150 years in understanding how existing traits diversify, we 50 have made less progress in understanding how novel traits arise in the first place. What 51 generates the novelty-here the first appearance of sterile workers? 52 In colonies of eusocial insect (ants, bees, wasps and termites), workers take on non-53 reproductive tasks and contribute to their parents' reproduction (Wilson, 1971). The 54 colony essentially behaves as a superorganism operating as a functionally integrated 55 that is, the amount of resources acquired by a parent. Both the male and female founder 122 contribute to the production of eggs equally. The available resources for egg production 123 of a t-year-old colony is . In the first year (0-year-old colony), the parents produce 124 eggs by using the available resources 0 (J) (= 2 ) (Fig. 2B). The number of first brood 125 eggs 0 is given by 126 where is the metabolic efficiency of the parents to convert the resources into eggs, is 128 the cost for producing one egg. For simplicity, we set / = 1, i.e., 0 = 2 . The final 129 family composition of the 0-year-old colony is: a founding male, a founding female and 130 first brood ( 0 eggs). 131 Second year (1-year-old colony): The 1st brood develop into 1-year old larvae, which 132 are too small and undeveloped to contribute to social labor. In the second year, the 133 available resources 1 (J) are still 2 . Parents produce 0 eggs of the 2nd brood in the 134 second year. Therefore, the final family composition of the 1-year-old colony is: a 135 founding male, a founding female, first brood ( 0 1-year-old larva) and 2nd brood ( 0 136 eggs). 137 In the third year, the 2nd brood are 1-year-old, which do not contribute to social labor. 138 The 1st brood are 2-years-old, at which time they begin to contribute to social labor as 139 helpers. where is the reproductive capacity of the parents at colony age (Fig. 3A). 158 Reproductive capacity of the parents increases with their sexual development. 159 Reproductive capacity of the parents is given by 160 = • tanh ( + 1) , 161 where is the maximum fecundity of the parents (Fig. 3A). 162 The maximum reproductive capacity is determined by sexual development of the 163 parents, which is acquired through canalization of the expression of sexual development 164 genes through epigenetic modification (epimark level) (Fig. 3B). 165 Epigenetic marks are partially inherited by offspring, which leads to a genomic imprinting 167 effect on the sexual development of the offspring. Therefore, imprinting level at colony 168 age is given by 169 where is the transmission rate of parental epimarks to offspring (Fig. 3B). 171 Antagonistic action of paternal imprinting and maternal imprinting delay the initiation 172 of sexual development of offspring, resulting in prolongation of the offspring's time to 173 become alates (Fig. 3C). For simplicity, we assume 174 = = . 175 176 Genomic imprinting and maturation delay of offspring. The time required to become 177 alates, numerically calculated from and its boundary conditions, is given by 178 . 179 The offspring that have spent after hatching in year become alates, contributing to the 180 founders' fitness ( Fig. 2A). We set = 4 (years) based on the subsocial sister group 181 Nalepa, 1984). Note that when an offspring stays longer in the colony as a helper, its direct 183 contribution to the yearly fitness is not only delayed, but also its residual life span as a 184 colony founder is shortened. However, its contribution as a helper will increase the 185 production of future offspring (younger brothers and sisters) in the current colony (Fig. 2B). 186 This trade-off between the contribution by staying as helpers and leaving success as 187 alates is key for the evolution of eusociality. where , is the number of alates at age produced in year and , is the reproductive 193 value of an alate at age . In exchange for a prolonged helper period, the reproductive 194 value of an alate , declines with the shortening of its residual life span ( − ) after 195 colony foundation. This discounted reproductive value of an alate at age is numerically 196 calculated from the fitness function −1, of the previous generation − 1, such that 197 In order to calculate the fitness of the first generation 1, , we set 1, = 1 because there 199 is no imprinting effects on offspring (all offspring reaching age become alates without 200 delay). Total fitness of generation is given by 201 When the resources exceed the maximum reproductive capacity of the parents , 204 there is no benefit to the colony since they cannot be used to produce additional eggs. 205 The maximum amount of excess resources ∆ due to the limitation of reproductive 206 capacity is given by 207 where , is the maximum yearly resources of a colony at generation . Mutation from 209 to + 1 takes place if ∆ ≥ (at least one egg can be produced by the surplus 210 resource) (Fig. 2C imprinting further prolongs the helper period resulting in in increased resources in the next 215 generation + 1, which again produces surplus resources ∆ . This whole process may 216 be repeated to increase reproductive capacity as long as the two above-mentioned 217 conditions are satisfied (Figs. S5 and S6). We consider that eusociality arises at the time 218 when the first-brood offspring lose all direct fitness. We also note the time at which the 219 first-brood helpers spend their whole life in the colony, i.e., the appearance of lifelong 220 helpers. 221 222

223
We show that genomic imprinting can explain the origin of eusociality, i.e., the first 224 occurrence of a lifelong neuter caste. To consider the evolutionary transition from 225 subsocial to eusocial in the ancestor of termites, we suppose a diploid subsocial 226 organism in which a male and a female cooperatively found a new colony in an enclosed 227 habitat (Figs. 1F, 1G and 2). We developed a genomic imprinting model that predicts the 228 evolution of eusociality via natural selection when parental reproductive capacity is 229 linked to a delay in the sexual development of offspring through transgenerational 230 epigenetic marks that act in a sex-antagonistic manner. We performed evolutionary 231 simulations by running colony generation, where each generation includes a colony life 232 cycle as a part of the simulation (Fig. 2, Fig. S6). Presence of helpers contribute to social 233 labor, which increases the resources available for egg production (Fig. 2B). Selection 234 favors higher reproductive capacity of the parents so as to fully utilize the available 235 resources (Fig. 2C). Acquisition of greater reproductive potential is derived from higher 236 epigenetic modification (canalization) of the genes for sexual development (Fig. 3 A and  237 B). This leads to higher genomic imprinting and thus to the prolonged helper period of 238 the offspring (Fig. 3C). 239 Our numerical algorithm demonstrates that genomic imprinting can drive eusociality via 240 natural selection on colony founders (Fig. 4). The evolutionary transition from subsocial to 241 eusocial systems readily occurs when the epimark transmission rate and/or the helpers' 242 work performance are high (Fig. 4A). Because helpers increase the resources available 243 for egg production (Figs. S3 and S4), natural selection acts on the parents to increase 9 production of alates. Greater reproductive potential requires higher levels of epigenetic 245 modification. Trans-generational carry-over of sex-specific epimarks leads to higher 246 genomic imprinting and thus to a prolonged helper period for offspring. This evolutionary 247 feedback between natural selection and genomic imprinting facilitates the transition from 248 subsociality to eusociality (Fig. 4B). At some generation, the first-brood offspring lose their 249 direct fitness (the origin of eusociality, red arrowheads in Fig. 4 B and C), and then spend 250 their entire life in the colony as helpers (open arrowheads in Fig. 4 B and C). Here the 251 offspring are essentially the extended phenotype of the parents via genomic imprinting. 252 This feedback process generating eusociality creates parent-offspring conflict 253 frequently because the first brood decreases its direct (and even inclusive) fitness by 254 prolonged helper periods (zigzag black and red lines in Fig 4B), but eventually increases 255 its inclusive fitness. Eusociality can still be achieved even if the helpers' work 256 performance is reduced by one-half ( = 0.8 → 0.4), although increase in the rate of 257 parental direct fitness is slowed down (Fig. 4C). When the helpers' work performance is 258 further reduced ( = 0.1), natural selection does not drive eusociality (Fig. 4D), because 259 helpers cannot produce sufficient resources for increased egg production. Most 260 importantly, the level of genomic imprinting is a key factor driving eusociality; the number 261 of generations required for eusociality (or lifelong helper) to evolve sharply decreases 262 with imprinting strength (i.e., epimark transmission rate for a given ) (Fig. 4E). 263 Eusociality cannot be achieved if imprinting strength is too low (gray area in Fig. 4E). 264 Also, the development of eusociality is impossible if colony longevity is too short (Fig.  265 S7), because the contribution of the first brood helpers cannot be used for the production 266 of alates in the colony. It should be noted that when is larger, it takes more time for 267 eusociality to evolve. This is because it takes more generations for the first brood to 268 become lifelong helpers when they have a longer lifespan ( = + ). 269

Discussion 271
The theoretical framework of the present model assumes iteroparity of the founder 272 parents within the nest. This assumption is reasonable because all eusocial insects are 273 iteroparous, which is essential to produce staggered age classes within colonies, The developmental time to reach maturity should have been long in the subsocial 284 termite ancestor due to a nitrogen-poor wood diet as in wood-feeding cockroaches in 285 which developmental time ranges from three to seven years (Nalepa, 2015). Among broods within a colony providing the opportunity for older brood to help younger sibs 291 (Thorne, 1997). We can reasonably expect that in the termite ancestor the presence of 292 older brood was advantageous for the parents and younger sibs because they provided 293 nest excavation, trophallaxis, allogrooming, defense and other forms of brood care. 294 Therefore, we can expect high work performance of the helpers in the subsocial 295 ancestral state. In our model, a certain level of work performance is necessary for the 296 origin of eusociality. Interestingly, however, can be as low as 0.1 when other 297 conditions are favorable (e.g., ≥ 15, ≥ 0.2, 0 ≥ 20, Fig. 4A). 298 As we discussed at the beginning, the evolutionary mechanisms driving the origin of 299 eusociality and selection for the maintenance and later elaboration of eusociality should 300 be discussed separately. We demonstrated that the acquisition of eusociality promotes 301 the production of alates by a colony and also increases the inclusive fitness of the first 302 brood compared with the initial subsocial state (Fig. 4 B and C). In this respect, our 303 findings are consistent with inclusive fitness theory. Most importantly, however, we found 304 that inclusive fitness theory itself cannot explain the origin of eusociality. During the 305 transition from the subsocial to the eusocial state, inclusive fitness of the first brood 306 results in a zigzag line alternating up and down for levels of fitness ( Fig. 4 B and C), 307 because there is a time lag between the origin of a prolonged helper period in the first 308 brood offspring and an increase in the reproductive capacity of the parents to utilize the 309 additional resources. This indicates that selection for higher inclusive fitness of the 310 offspring cannot get over the selection valley (it goes down and then up again) and thus 311 cannot generate lifelong helpers. 312 The earlier hypothesis that tried to overcome this selection valley is parental 313 manipulation (Alexander, 1974): parents restrict the reproductive options of some 314 offspring so that they assist in the production of fully fertile offspring. A serious problem 315 with this hypothesis is that counter-selection on the offspring would make them 316 overcome parental manipulation. Moreover, this hypothesis also cannot explain the 317 origin of eusociality, although parental manipulation may increase the parental fitness  the level of DNA methylation is known to be considerably higher than in any other social 365 insect studied to date (Glastad et al., 2016). Future studies are needed to determine the 366 exact molecular mechanism underlying the genomic imprinting of termites, even though 367 our conclusion holds irrespective of the exact agents. 368 The current theory to explain the origin of eusociality has the potential for broad

627
Therefore, it is apparent that biology is in a major transitional period that will take it forward from 628 the previously narrow perspective.

629
Over the past decade there has been a rapid accumulation of definitive evidence for the

656
The degree of sexual dimorphism will be associated with the degree of sex-specific 657 epigenetic modifications (Deegan and Engel, 2019). Selection favoring a higher degree of sexual 658 dimorphism would lead to a higher imprinting level, if the efficiency of erasure is not changed.

659
This carryover effect of sex-specific parental epimarks on the offspring's development can

676
After the first report of AQS in the Japanese subterranean termite R. speratus (Matsuura et

711
As shown in the scheme (Fig. S2)  colony, which provides the opportunity for older brood to contribute to the production and growth of 745 younger sibs (Thorne, 1997). The presence of older brood is advantageous for the parents and 746 younger sibs rather than imposing a deleterious effect because they provide nest excavation, 747 trophallaxis, allogrooming, defense and other forms of brood care. Therefore, we can reasonably 748 expect a high work performance of the helpers in the subsocial ancestral state.

749
Work performance (J/year) of a helper offspring increases with age . Eggs (0-year-old) and

804
Hence, the work performance of a helper at age (year) is given by Equation (S1).

Rt
The available resources for egg production of a t-year-old colony