In contrast to classical kin selection theory, the kinship theory of imprinting takes into account the possible differential expression of maternally and paternally derived genes and thus might explain the non-Mendelian inheritance of phenotypic traits, including behavior toward kin (Haig 2000a). This theory proposes that complex patterns of allele-specific expression have evolved to favor the transmission of identical-by-descent copies of these genes (Haig 2000a; Wilkins and Haig 2003). Differential expression of maternally and paternally derived alleles in an individual or in a litter may lead to conflicts over the gene's expression and treatment of relatives (Burt and Trivers 1998; Haig and Wilkins 2000).

Genomic imprinting has been reported in diverse organisms ranging from plants to humans (Haig 1993, 2004; Herrick and Seger 1999; Messing and Grossniklaus 1999; Spielman et al. 2001; De Jong et al. 2005). In insects, cases of genomic imprinting have been described in Drosophila at the genetic, cytological, and molecular level, but no behavioral consequences have been found (Lloyd 2000; Maggert and Golic 2002). However, some authors predict that imprinting should exist at the behavioral level, especially in social Hymenoptera where the haplo-diploid genetic system ensures that relatedness among cooperating kin is not symmetric along the paternal and maternal lines (Haig 1992, 2000b; Queller and Strassmann 2002; Queller 2003). To date, the only experimental support for this theory comes from the defensive behavior of Africanized honeybees, which might be correlated with parentally imprinted genes (Guzman-Novoa et al. 2005).

In diploid species, relatedness asymmetries can arise because of sex bias in the population, sex-biased dispersal, or the mating system itself (Haig 1999, 2000b; Hurst 1999). The level of relatedness between two offspring found in the same place will be greater on average via the parent that had more partners or dispersed less (for a discussion, see Haig 1999; Hurst 1999; Queller 2003). Therefore, patrilineal and matrilineal relatedness will depend partly on the polygyny and polyandry rates, but also on how females distribute their eggs and how males and females disperse from their natal sites, from mating sites, and between egg clutches. In the case of polygyny, paternal genes will therefore bear a higher selective pressure than maternal genes to diminish competition among offspring, and the same will be true for maternal genes in the case of polyandry. In both cases, if refraining to compete with a kin entails a cost, the parental sex whose genes benefit less from this behavior may bear a selective pressure to stop cooperation. Accordingly, the imprinting (i.e., silencing) of genes favoring such cooperative behaviors among offspring is expected whenever sibs are more related to each other via one parental sex than the other (Queller 2003).

We have searched for evidence of imprinting of kin superparasitism (i.e., more than one larva is present inside the same host) avoidance behavior in larvae of Aleochara bilineata, a diploid coleopteran species whose larvae develop as pupal parasitoids of cyclorrhaphous flies (Colhoun 1953). A. bilineata females lay their eggs near plants infested by their hosts. First instars must then actively search in the soil for a host (Fournet et al. 2001) and only one larva can develop per host, any host intruder being attacked by the resident larva until only one competitor remains or both die (Fuldner 1960; Salt 1961). These larvae can discriminate healthy from parasitized hosts and will avoid the latter when given the choice (Royer et al. 1999), this behavior being common to many parasitoid species (Godfray 1994). Much more strikingly, A. bilineata larvae are able to distinguish hosts occupied by a full sib from those occupied by an unrelated larva, and they will preferentially take the risk of dying without entering a host rather than fight to the death with a sib resident larva (Lizé et al. 2006). In this system, kin competition seems to be the driving force selecting kin recognition ability (Lizé et al. 2006). Since male and female A. bilineata readily mate with several partners during their lifetime, at least under laboratory conditions, an imbalance between polygyny and polyandry rates is possible in this species.

To define the inheritance pattern of kin superparasitism avoidance in A. bilineata larvae, we studied their ability to avoid superparasitism of cousins related through the maternal or paternal lineage. More specifically, we measured in choice experiments the superparasitism rates expressed by larvae toward hosts already parasitized by four types of cousins, according to the patrilineal or matrilineal link between the forager and the resident.

Host and parasitoid strains: Laboratory strains of Delia radicum (fly host) and A. bilineata (parasitoid) were initially established in 1994 and 2002, respectively, using D. radicum pupae collected in cabbage crops in St Méloir des Ondes (Brittany, France). D. radicum and A. bilineata were reared using methods described, respectively, in Neveu et al. (1996) and Hertveldt et al. (1984). Both strains were refreshed yearly with field-captured individuals from the same population.

Preparation of parasitized hosts harboring resident larvae: To obtain parasitized hosts, we placed a healthy D. radicum pupa in moist sand in a small dish (Caubère, Yèbles, France; height: 8 mm; diameter: 26 mm). A 0- to 24-hr-old larva related to one of the resident larvae was then placed on top of the sand, and the dish was closed. The process was stopped 48 hr later by removing the pupa. The pupa was then rapidly cleaned using a wet paintbrush and checked under a dissecting microscope to assess parasitization status. We prepared the two sets of hosts by observing the pupae 48 hr after parasitization and by selecting parasitized hosts where the larvae (now called “resident” larvae) had plugged their entrance holes (Lizé et al. 2006). Only hosts where the parasitoid larvae were still at the first larval instar were used in the experiments.

Experimental setup: The grandparents of the larvae used in our experiments were obtained from parasitized pupae selected at random from our laboratory stock, which harbors ~400 reproductive couples of A. bilineata. The parasitized pupae were stored individually in Eppendorf tubes until emergence. Once emerged, the young virgin adults were placed in small groups of four to five individuals and monitored continuously to observe mating. Ten mating couples were thus formed and each was kept in a separate aerated circular dish (Caubère, Yèbles, France; height: 25 mm; diameter: 80 mm) with food ad libitum and a dampened piece of black cloth at the bottom of the dish as a laying substrate. Eggs were collected twice a week and isolated individually in a Beem capsule (Agar Scientific, Essex, UK; height: 1.4 mm; diameter: 0.7 mm). Upon hatching, each larva was supplied with a host in the setup described in the parasitization procedure. Parasitized pupae were then kept in the setup until adult emergence; virgin adults thereby remained isolated. Upon emergence, each brother or sister was individually placed with an unrelated mate taken at random from another family to observe mating. A total of 50 couples were thus formed, generating the four categories of cousins required in our experiments (Figure 1). After mating, each pair of parents was isolated and each egg was collected and isolated in a Beem capsule according to the family it came from, as described above. Eggs were checked daily for hatched larvae and the larvae remained isolated until the experiment. Only first instar larvae aged <24 hr were used in the experiments. Kin recognition in larvae issued from a single pair was tested with both paternal and maternal cousins (as described below).

Figure 1.— Genealogy of foragers and resident larvae used in choice tests. The arrow connects the forager to the two residents that it will encounter in the choice experiment, i.e., one cousin (K for kin) and one unrelated larva (U). Paternal cousins are encountered (more ...)

For all treatments, two parasitized hosts were placed on a thin layer of moist sand in a small dish. One host was already parasitized by an unrelated larva, and the other was already parasitized by a paternal or a maternal cousin. The two hosts were then covered by moist sand (1 vol water/4 vol sand) up to the upper limit of the dish. A larva aged <24 hr was deposited at the center of the dish at an equal distance from the two buried hosts, and the dish was closed. The experiment was stopped 96 hr later, the pupae cleaned, and the hosts observed to count the number of larvae present in each host as described above. Since the lifetime of A. bilineata first instars is on average 96 hr, any larvae not recovered at the end of the experiment were considered to have starved to death at that time.

To test the parental influence on kin recognition ability, we measured the superparasitism rates of foragers toward cousin larvae, according to the patrilineal or matrilineal link between the forager and the cousin larva. We used dual-choice tests where foragers were presented with a host already parasitized by an unrelated larva and another host already parasitized by a cousin larva. Resident larvae were paternal cousins of the forager when the father of the forager was a full sib of either the father (Figure1A) or the mother (Figure 1C) of the resident. Resident larvae were maternal cousins of the forager when the mother of the forager was a full sib of either the father (Figure 1D) or the mother (Figure 1B) of the resident (total number of tested larvae: A, 105; B, 72; C, 79; and D, 67).

Statistical analysis: In each of the four treatments (A–D) in Figure 1, a G-test was used to compare the relative superparasitism rates of kin (cousin) and unrelated resident larva. The null hypothesis was that, in a choice experiment, kin and unrelated larvae would be superparasitized at similar rates. Analysis was performed using the software R 2.3.1 (Ihaka and Gentleman 1996).

In all treatments, 42–48% of larvae did not superparasitize any hosts. These larvae were not recovered at the end of the experiment and were considered as having starved to death. This phenomenon was expected since A. bilineata larvae typically avoid parasitized hosts (Royer et al. 1999). The proportion of larvae not entering a host did not vary significantly among treatments (G = 0.85, 3 d.f., P = 0.86). Only larvae that superparasitized a host were used to calculate percentages shown in Figure 2.

Figure 2.— Results of superparasitism choice experiments. Percentages of superparasitizing foragers choosing to superparasitize a kin resident (K, shaded bars) or an unrelated resident (U, solid bars) in dual-choice tests. Paternal cousins are encountered in treatments (more ...)

Foraging larvae preferred superparasitizing hosts harboring unrelated larvae rather than hosts harboring paternal cousins. When the forager's father was a sib of the resident's father, of the 55 foragers that entered into a contest with a resident larva, only 18 chose to do so with the paternal cousin (Figure 2A, G = 6.70, 1 d.f., P = 0.009). Comparable results were obtained when the forager's father was a sib of the resident's mother (Figure 2C, 12/41, G = 7.27, 1 d.f., P = 0.007). In contrast, the larvae did not significantly avoid hosts harboring maternal cousins compared to hosts containing unrelated larvae. When the forager's mother was a sib of the resident's mother, 23 of 41 superparasitizing foragers attacked the cousin larva (Figure 2B, G = 0.38, 1 d.f., P = 0.54), and comparable results were obtained when the forager's mother was a sib of the resident's father (Figure 2D, 18/35, G = 0.03, 1 d.f., P = 0.86). Treatments A and B were further repeated and produced similar results (cousin/total superparasitism—treatment A: 19/63, G = 10.20, 1 d.f., P = 0.001; treatment B: 23/46, G = 0, 1 d.f., P = 1). Results are summarized in Table 1.

TABLE 1 Behavior of foraging A. bilineata larvae toward hosts parasitized by the four possible categories of cousins

Our results show that A. bilineata larvae are able to recognize paternal cousins, since they avoided superparasitizing them (treatments A and C) (Table 1). Since those cousins bear a different cytoplasm from that of the foragers, we can conclude that kin recognition in A. bilineata implies nuclear genes and does not result from a maternal effect (the latter could explain recognition only between full or half sibs). In sharp contrast with their response toward paternal cousins, foraging A. bilineata larvae did not avoid superparasitizing their maternal cousins more than unrelated larvae (treatments B and D), demonstrating either that these cousins are not recognized as kin or that the tendency to avoid superparasitizing kin individuals is not expressed in that particular case. Taken together, our results thus show that the transmission of kin superparasitism avoidance is asymmetric between maternally and paternally related individuals, indicating either that the nuclear genes involved in kin recognition in this species are transmitted by sex chromosomes or that they are transmitted by autosomes and show parent-of-origin differential expression (genomic imprinting).

In genetic recognition systems, three different components are needed: (i) a specific signal, emitted by the individual to be recognized as kin; (ii) a specific receptor to match that signal, borne by the individual capable of kin recognition; and (ii) a differential response according to whether the target individual is recognized as kin or not (Hamilton 1964; Dawkins 1976; Keller and Ross 1998). Such a recognition system implies that the genetic determinants of the signal and its specific receptor are cotransmitted to the offspring (Hamilton 1964; Alexander and Borgia 1978; Alexander 1979; Ridley and Grafen 1981). This implies that the kin recognition signal and its specific receptor should be on the same chromosome. We can therefore formulate three hypotheses, depending on whether the signal and its receptor are borne (i) by the Y, (ii) by the X, or (iii) by an imprinted gene located on an autosome.

The transmission of factors on a Y chromosome is a possibility because, unlike haplo-diploid Hymenoptera, the great majority of Coleoptera are diploid with XY sex determination, and the direct observation of A. bilineata chromosomes confirmed that our A. bilineata laboratory stock is diploid, with 2n = 20 chromosomes (B. Dutrillaux, Museum National d'Histoire Naturelle de Paris, personal communication). However, if the kinship signal was borne by the Y sexual chromosome, kin superparasitism avoidance would be expressed only by male cousins toward male cousins bearing the same Y sexual chromosome. Treatment C clearly excludes this hypothesis, since foragers also avoided superparasitizing their cousins although the resident and forager larvae had different Y chromosomes. We can conclude that the genes coding for the kinship recognition are not transmitted through the Y sexual chromosome.

Treatments A and C, moreover, show that genes coding for the kinship signal are expressed regardless of the parent of origin, since resident larvae are recognized by foragers whether they are related to the forager through their father (treatment A) or through their mother (treatment C). The kin recognition genes thus cannot be located on the X chromosome, and the only remaining possibility is that they are located on an autosome. Therefore, the fact that kinship recognition remains asymmetric among paternal and maternal cousins despite an autosomal location can be explained only if genes coding for receptors are imprinted.

Kin recognition in social insects commonly relies on olfactory signals associated with epicuticular lipids (Breed and Bennett 1987). Here, this recognition system would probably not be the most efficient since the resident larva “seals” itself inside the host after obstructing its entrance hole into the puparium with a substance excreted by the anus. However, our previous experiments showed that pupae containing a full sib are not avoided by foragers if the entrance hole is still open (Lizé et al. 2006), indicating that the kinship signal allowing full sibs to avoid superparasitizing each other is precisely associated with this sealing plug (which thus becomes an identification seal). Accordingly, the genes responsible for the production of a chemical signal associated with the sealing plug would be expressed whatever their parental origin, while genes responsible for the chemodetection of this signal would be maternally imprinted (i.e., silent when transmitted by the mother and expressed when transmitted by the father).

Being recognized as kin and subsequently “spared” by a forager is obviously advantageous for the resident that holds the resource (i.e., the host), since it avoids a dangerous fight, which would not increase its fitness anyway. On the other hand, refusing to superparasitize a cousin could be costly for the forager, because it may die without finding another host (Fuldner 1960). Such “altruism” toward a cousin is difficult to interpret for behavioral ecologists since the forager risks death by starvation in favor of a weakly related individual. However, from a gene's point of view, replacing a resident larva bearing a homolog does not increase fitness; it can even decrease fitness if the fight kills both larvae. For an allele, the benefit from avoiding superparasitizing a kin simply depends on the probability that it bears a homolog and this probability may be different for a maternally or a paternally inherited gene, i.e., if larvae foraging in the same patch are on average more related via one parent than the other (kinship asymmetry). As the existence and the direction of kinship asymmetry of A. bilineata larvae has not been investigated in the field, it is not yet possible to say whether maternally imprinted kin recognition in this species is adaptive or not. If maternal relatedness of larvae to their neighbors is equal or greater than paternal relatedness, then imprinted kin recognition in this species is not adaptive. On the contrary, if paternal relatedness between foraging larvae is greater than maternal relatedness, the imprinted kin recognition would be adaptive and would support the kinship theory of genomic imprinting.

Kinship asymmetries can arise because of sex bias in the population, sex-biased dispersal, or because of the mating system itself (Haig 1999, 2000b; Hurst 1999). The level of relatedness between two offspring found in the same place will be greater on average via the parent that had more partners or dispersed less. However, sex-ratio bias has never been reported in A. bilineata, and our personal observations from field or lab studies do not show any significant imbalance between sexes. Thus, relatedness asymmetry could arise either because adult females disperse more than males or because the polyandry and polygyny rates are unbalanced, the polygyny rate being greater. Molecular studies would be necessary to ascertain whether A. bilineata individuals emerging from the same patch are more related via one parental sex than the other.

Even in the absence of asymmetric mating, there could be an asymmetry between the relatedness of larvae within a patch because females control both the location and the number of eggs laid per patch and maybe also their relatedness (full or half sibs). Indeed, choice experiments have shown that females select a patch and adapt their clutch size according to host availability in the patch (Fournet et al. 2001). Therefore, they may directly control the level of competition among their offspring. In contrast, males have no control over the future location of the eggs that they fertilize. Their only way of limiting fatal interactions between their offspring is by the fact that patrilineal kin individuals avoid superparasitizing each other. Patrilineal kin superparasitism avoidance would efficiently limit conflicts, in particular when two females mated by the same male lay eggs in the same patch.

One may wonder why female-transmitted genes avoiding kin superparasitism would be imprinted (not expressed), since females should also benefit—although to a lesser degree—from superparasitism avoidance between maternal cousins. This may be explained if the cost of maintaining the capability of recognizing maternal cousins exceeds the expected benefit. In particular, if the probability of mistakenly recognizing a nonrelative for a cousin is larger than zero, then the cost of these errors will increase when kin interactions are rare (Reeve 1989; Lehmann and Perrin 2002). Therefore, maternal imprinting could allow females to reduce costs for their larvae in terms of errors, risk, and time budgeting, without suppressing full-sib superparasitism avoidance (guaranteed by paternal genes). Females and males of such species may have developed different strategies to limit competition between their offspring.

Comments from B. D. Roitberg, G. Boivin, S. Dugravot, and two anonymous reviewers greatly improved this manuscript. We also thank M. Rault and C. Paty for their technical help. This work was funded by a Ph.D. grant from the French Ministry for Research to A.L.