1. Introduction
Social insects are characterized by large numbers of individuals living in close proximity in a nest. Within this group, eusocial insects are defined by the presence of overlapping generations, cooperative brood care, and division of labor into reproductive and non-reproductive individuals. Despite their multiple advantages, these traits make eusocial insects particularly vulnerable to diseases [
1]. In response to this increased susceptibility, a wide range of collective and individual defenses against parasites and pathogens have evolved in insect societies [
2,
3]. Within an eusocial insect colony, the diversity of resistance and tolerance traits can be very high, even more so if queens mate with several unrelated males [
4,
5,
6].
Honeybees (genus
Apis) stand amongst the most polyandrous eusocial insects [
7]. In this taxon, diploid queens achieve extreme levels of multiple mating with males through highly diverse and dynamic lek-like mating systems called Drone Congregation Areas (DCAs) [
8,
9,
10,
11]. Colonies of honeybees are formed by a single polyandrous queen, which produces haploid males and diploid females that either turn into new queens or into workers. Within a nest, workers can be full-sisters if they share the same parents, or half-sisters if their fathers differ. In the Western honeybee,
A. mellifera, about twelve half-sister groups (called “subfamilies” or “patrilines”) are found in average per colony [
7]. In addition to these extreme levels of polyandry, very high rates of recombination have been documented across the genome of this species [
12]. Altogether, this enhanced genetic diversity affects colonies development [
13,
14] and can increase chances of the colonies to survive diseases [
15].
A. mellifera suffers from a great number of diseases, of which the varroosis is undoubtedly the most harmful at the global scale [
16,
17]. This syndrome is caused by the ectoparasitic mite
Varroa destructor and generally leads to the rapid collapse of the western honeybee colonies if no treatments are applied. The native parasite of
Apis cerana in Asia is a mite that has managed to spillover to
A. mellifera after colonies of the western honeybee were introduced in its distribution range and subsequently spread across most regions of the globe [
18]. Acting as a vector for highly virulent honeybee viruses [
19,
20,
21],
V. destructor particularly affects the brood of its host, where its reproduction takes place [
22].
The life cycle of
V. destructor is composed of two phases: reproduction within brood cells and dispersal on the adult workers and drones. Reproduction starts with the invasion of a 5th instar larvae brood cell by a mature female mite (the foundress) shortly before capping by the workers. Approximately three days after the cell is closed, the foundress will lay a first haploid egg, which will develop into a male. She will then lay several diploid female eggs in 30 h interval. Her offspring will take about six days to reach maturity. At this stage, reproduction will occur between the mature offspring in the cell, resulting in incestuous mating if the cell was infested by a single foundress, or in the possible admixture of mite lineages if several foundresses initially infested the same cell [
23,
24]. Mating will occur until the host is fully developed and emerges, after nine to twelve days post-capping. Once the fully developed bee exits the cell, the mated female mites will enter a dispersal phase. They will crawl on the combs, climb onto adult bees, and hide between the sclerites of their host until an opportunity to infest new brood cells emerges.
V. destructor dispersal phase finishes with the detection and infestation of a new 5th instar larva cell, where a new reproductive cycle can start. A wide variety of factors are believed to trigger host finding, and chemical cues from the host seem to play a crucial role in this important step [
25].
A single
V. destructor foundress will typically perform several reproductive cycles during its life [
26], leading to a rapid buildup of parasite populations within a honeybee colony [
27,
28]. However, the reproduction of
V. destructor depends on the availability of bee brood, which fluctuates greatly during the season, across environments and among
A. mellifera populations [
29]. In addition, the type of brood (i.e., worker vs. drone) also affects mite population dynamics. In fact, the honeybee drone brood, which takes more time to develop and leads to the production of more offspring per foundress than the worker brood, is more attractive to
V. destructor [
30,
31,
32]. However, little knowledge exists on the links between the invasion behavior of the mite and the biology of the brood. More specifically, whether mites consider certain biological traits of the individual larva they infest is currently unknown.
The rapid growth of
V. destructor populations in
A. mellifera generally induces the collapse of colonies within a few years in the absence of beekeepers’ intervention. However, several resistance and tolerance traits against the parasite have arisen and some Western honeybee populations can now survive without treatments [
33,
34]. One of these adaptations, the Suppression of Mite Reproduction (SMR), is of particular interest for beekeepers, breeders and scientists [
35,
36]. This trait is highly heritable [
37,
38] and consists of the absence or the delay of mite foundresses’ egg laying in the host brood cells and results in strong diminution of the parasite population dynamics. However, the biological mechanisms behind SMR are currently not fully understood. More specifically, whether the reproduction failure of the parasite is solely due to the action of adult workers, of brood, or both simultaneously remains unclear. In fact, adult worker bees may interfere with the mite reproduction through a diverse range of mechanisms, including the detection, unsealing, and resealing of the infested cells (the “recapping behavior”), or the selective removal of the infested pupae (the “Varroa-Sensitive Hygiene”, or “VSH behavior”) [
39,
40,
41,
42]. Yet, the brood may also play a role by directly altering the reproduction of the mite with kairomones [
43] or by signaling the workers that it is infested [
44].
We herein used behavioral genetics to investigate the interactions between
A. mellifera and
V. destructor. This discipline aims at unraveling the links between behaviors and genotypes and has been used extensively to study
A. mellifera [
45]. In this study, we first investigated four traits of the mite or of its host that take place during the parasite reproductive cycle: the infestation of brood cells, the number of foundresses infesting brood cells, the reproduction of foundresses, and the recapping of cells by adult bee workers. We then compared these different phenotypes to the genotype of the brood on which they were observed, using sibship reconstruction analyses with microsatellite markers to reconstruct the subfamilies of the pupae. The comparison of the phenotypic traits across patrilines allowed us to investigate whether specific bee subfamilies (i) are more frequently targeted by varroa infestations, (ii) are able to block the reproduction of the mites and (iii) are more likely to be opened by workers performing the recapping behavior.
4. Discussion
In this study, we used behavioral genetics to investigate the interactions between A. mellifera and V. destructor, focusing on several crucial aspects of the reproduction of the mite. To do so, we compared the status and level of infestation of the mite, its fertility, and the recapping behavior of workers to the subfamilies of the pupae where these phenotypes were observed. Our results show that the traits varied in time, but revealed no significant associations between these phenotypic observations and the most prevalent patrilines found in the colonies we investigated, suggesting that these traits are not strictly determined by the genotype of the drones siring the honeybee worker brood.
Although not all colonies could be sampled at each time point, our data provide interesting insights into the temporal evolution of the expression of the traits we observed. In fact, the different phenotypes we investigated in this study varied significantly across the different sampling dates. First, the infestation level (number of infested cells and multiple infestations) of
V. destructor over all colonies showed variation that reflect
A. mellifera colony dynamics. In regions with a hot climate like Provence, where this study took place,
A. mellifera queens typically stop producing brood during the hottest days of summer (mid-July to mid-August) and restart laying eggs again once the nectar flow restarts (end of August). In parallel, mite numbers generally increase in an exponential fashion throughout the brood season when no treatments are performed [
28]. Our data reflect these population dynamics, since we found higher levels of infestation and multiple infestations at the end of August (i.e., moderate number of mites but few brood cells) and October (i.e., more mites and fewer brood cells) than in September (i.e., moderate number of mites but more brood). These findings are in line with previous studies documenting correlated population dynamics between brood and mites in
A. mellifera colonies [
24,
27,
28]. The next phenotype, the reproduction of
V. destructor, varied greatly across the sampling dates and colonies and tended to decrease towards the end of the season. Temporal variation in
V. destructor reproduction has been documented in the past [
54,
55], but these patterns may vary across years, and factors governing this variation remain currently unknown. Finally, the recapping phenotype varied highly across colonies, with a general temporal increase from the end of August to the end of October. In the recent past, this behavior has been proposed as a key mechanism explaining survival of mite infested colonies in the same honeybee population as we studied here [
39]. However, although recapping significantly correlated with the level of infestation at the colony level, our data show that the expression of this trait on infested cells was not systematically greater than on non-infested cells. This result suggests that recapping does not take place in response to
V. destructor infestation, and stress the need for more investigations on the mechanisms behind this behavioral trait. Overall, the important variations of the phenotypes we observed at the colony level may also be explained by the finite number of cells analyzed in this study. However, our aim here was not to provide precise colony-level parameters, but to look at individual bee phenotypes, an aim that was not perturbed by this phenotypic variation.
The genotyping revealed marked variation in genetic diversity across the three colonies. The number of markers and their polymorphism level, as well as the sample sizes used, permitted to study accurately the dominant subfamilies in the colonies, as reflected by the low
NDE and
NSE estimates. Notably, the distribution of the patrilines was homogenous across the collection dates, which is in accordance to former results on sperm admixture in the queen spermatheca [
56]. Curiously, a consequent amount of rare subfamilies (<5 individuals per patriline) were sampled in the colonies. Unfortunately, these patrilines could not be included in our statistical analyses for methodological issues. Such rare patrilines have been documented in the past, and may have specific functions in the colony such as developing into emergency queens [
57,
58]. Although these subfamilies may possess increased resistance towards
V. destructor, their low prevalence in the colony would not affect significantly the population dynamics of the mite, and the parasite populations could quickly build up on the brood of more frequent, sensitive patrilines. In contrast, some patrilines were very common in the colonies (e.g., 41.66% of individuals from the colony B belonged to a single patriline). Although this finding could be due to chance alone, it could also be caused by the fact that queens mated with several drones with identical genotypes (e.g., brothers from the same colony). Altogether, these observations stress the need for more studies on the colony-level behavioral genetics of
A. mellifera, as little is currently known on the exact prevalence and specialization of the honeybee worker subfamilies.
These phenotypic and genotypic analyses allowed us to study the links between
A. mellifera pupae subfamilies and several reproductive traits of
V. destructor. First, the invasion behavior of
V. destructor was not affected by its host subfamilies, since the presence/absence of mites and the number of foundresses did not vary significantly across bee patrilines. In the past, physical properties of the cells and the position of the larva have been shown to affect the mite invasion behavior [
59]. In addition, mites use specific chemical cues of the larvae to infest cells [
60] and mite infestation levels were shown to vary significantly between different bee brood race [
59]. However, the distribution of mites in the brood cell of
A. mellifera does not seem to reflect specific aggregation patterns [
61]. Here, the absence of significant association between subfamilies and mite infestation bring new insights into the invasion behavior of
V. destructor and adds to other recent findings showing that foundresses do not co-infest a cell based on genetic cues [
23].
In parallel, the absence of significant association between the reproductive status of mites and the pupae subfamilies brings new knowledge on the expression mechanisms of SMR in diploid workers. The heritability of the main
V. destructor resistance traits has been known for decades [
37], and traits such as VSH have been used in selection programs with promising results. Notably, honeybee colonies selected for this trait also expressed lower mite reproduction levels [
42]. SMR can be transmitted by queens to their progeny, and expressed in colonies even if the founding females are mated with unselected drones [
62]. Interestingly, when performing crosses between colonies expressing high and low SMR levels, Locke [
38] found that colonies formed with susceptible queen and resistant drones had low levels of mite reproduction, suggesting that SMR had a strong dominant genetic component that can be passed across generations by males. Our results do not match the predictions of that study. The variation of SMR was substantial in the colonies we genotyped (colony-level proportions of non-reproducing mites ranging from 0.46 to 0.20), but this trait did not vary significantly across subfamilies. These discrepancies may be due to the fact that SMR has a different genetic component in the colonies we studied compared to the Swedish resistant colonies analyzed by Locke [
38]. Indeed, genomics studies performed in these two populations have found distinct genetic bases for this trait in Sweden [
63] and in France [
64]. However, the French
A. mellifera population used in [
64] is not the same as the population used here. Thus, conclusions from the latter study cannot be applied to our findings.
Finally, the recapping of cells by the adult workers was also not significantly affected by the subfamilies of the brood within the cells. Honeybee workers may be able to discriminate between brood genotypes, since behavior such as the rearing of emergency queens has been shown to be affected by the pupae’s subfamilies [
57,
58]. Our GLM also showed that recapping varied significantly in time and across colonies. While the reason for the temporal pattern we detected here remains unknown, high variability of this trait across colonies of Avignon and other populations was previously documented [
39]. Here, irrespective of the fact that recapping has evolved in response to
V. destructor in the population we studied (see above), our result indicate that the adult workers do not perform this behavior according to the brood subfamily found in the cell.
The level of
A. mellifera genetic diversity has been linked to the level of resistance to
V. destructor [
65] and other pathogens [
15,
66] at the colony level. However, in this study we did not detect a significant link between the dominant honeybee worker subfamilies and the invasion behavior of the mite, or with the expression of two honeybee resistance traits (SMR and recapping). However, with our study design, we may have missed rare worker subfamilies that specialize in
V. destructor resistance behavior. Although the expression of SMR in these rare patrilines would only poorly disturb the mite population dynamics at the colony level, workers from these patrilines could affect mite populations by specializing in recapping or other behavior such as Varroa-Sensitive Hygiene. The potential role of these rare subfamilies and the interactions between the different resistance traits at the colony level need to be further examined.