Heterosis occurs in individuals when genetic diversity, i.e., heterozygosity, increases fitness. Many advanced eusocial insects evolved mating behaviors, including polyandry and polygyny, which increase inter-individual genetic diversity within colonies. The possibility of this structure of diversity to improve group fitness has been termed social heterosis. Neither the independence of individual and social heterosis nor their relative effect sizes have been explicitly measured. Through controlled breeding between pairs of Western honey bee queens (Apis mellifera L.; n=3 pairs) from two distinct populations, we created inbred colonies with low genetic diversity, hybrid colonies with high heterozygosity, and mixed colonies (combining inbred workers from each population) with low heterozygosity and high social diversity. We then quantified two independent traits in colonies: survival against bacterial challenge and maintenance of brood nest temperature. For both traits, we found hybrid and mixed colonies outperformed inbred colonies but did not perform differently from each other. During immune challenge assays, hybrid and mixed colonies experienced hazard ratios of 0.49 (95% CI [0.37, 0.65]) and 0.69 (95% CI [0.50, 0.96]) compared to inbred colonies. For nest temperatures, hybrid and mixed colonies experienced 1.94±0.97°C and 2.82±2.46°C less thermal error and 0.14±0.11°C² and 0.16±0.06°C² less thermal variance per hour than inbred lines. This suggests social and individual heterosis operate independently and may have similar effect sizes. These results highlight the importance of both inter- and intra-individual diversity to fitness, which may help explain the emergence of polyandry/polygyny in eusocial insects and inform breeding efforts in these systems.

Breeding Design

Mated honey bee “founder” queens were sourced from the US states of Indiana (population A) and Texas (population B). On-going whole-genome scans indicate these populations represent the maximum possible genetic variance in naturalized honey bee populations in the United States (Ryals, Fikere et al. 2024, in prep.). All full-size colonies were kept at the Purdue University Research Apiary in West Lafayette, Indiana. We produced daughter queens and drones from a pair of founder queens A and B, one from each population. Using instrumental insemination, inbred crosses were performed for both A and B by mating daughter queens to a single drone from their own mother (AA and BB crosses). Accounting for haplodiploidy, their offspring have an expected inbreeding coefficient of 0.25. Reciprocal outbred crosses were also performed by mating a daughter queen of A to a single drone of B and vice versa (AB and BA crosses; Figure 1). Semen from a different, single drone was used in each cross to decrease unwanted genetic variation within worker groups and control for effects due to multiple mating.

This design partitions the average genetic divergence existing between the chosen founder pair within individuals in the hybrid offspring of AB or BA and between individuals when combining inbred offspring of AA and BB in a single mixed colony. These groups were compared to the mean performance of inbred colonies to measure the effect of individual and social heterosis respectively. Hybrid, mixed, and inbred workers represent our three experimental groups. Because there is natural phenotypic variation in each population, the performance of all experimental groups and corresponding effect sizes of heterosis likely depend on the choice of founder queens. To achieve independent measurements of heterosis, the breeding design was repeated using three separate pairs of founder queens, resulting in 12 instrumentally inseminated daughter queens and 15 worker groups.

Apiculture

To produce drones, uniquely paint-marked founder queens were introduced into separate queenless colonies and presented with a frame of wide-diameter “drone comb”, inducing them to lay unfertilized eggs that develop into drones. Each drone comb was cleared of bees and placed in a cage prior to drone eclosion. Newly emerged, caged drones were paint-marked according to parentage and returned to natal colonies. Daughter queens were produced following standard methods (Büchler et al., 2024). The presence of each founder queen was verified before grafting their larvae into paint-marked JZBZ cups (Mann Lake ltd. Z350). Grafting was scheduled so queens and drones would be ready for insemination around the same date. All grafted larvae were raised into queens in a single queenless “cell-building” colony. Prior to eclosion, all daughter queens were introduced into separate queenless, nucleus colonies sized for five standard Langstroth frames. Queen excluders were placed over entrances to prevent mating flights.

Daughter queens in each replicate were instrumentally inseminated on the same day, 5-10 days after queen emergence and 10-20 days after drone emergence, following standard procedures (Büchler et al., 2024; Laidlaw, 1977). Daughter queens were captured from their nucleus colonies in small, marked cages and anesthetized with CO₂ gas. Marked drones were captured from natal colonies into large cages. Each queen was inseminated immediately after semen collection from the intended drone. All queens were marked with numbered disks. Queens were reintroduced to respective nucleus colonies (with excluders in place) and treated with CO₂ the following day to induce oviposition. All nucleus colonies were then fed one gallon sucrose solution to stimulate growth. After verifying the presense of capped worker brood (indicating successful insemination), colonies were expanded into 10-frame deep boxes and fed an additional gallon. Colonies were routinely inspected to remove “supersedure” queen cells and prevent replacement of daughter queens by workers.

Immune Challenge Assay

The immune challenge assay was performed in “micro-colonies,” each made from ~1.5 L plastic cups with mesh-covered windows for ventilation, a ~20 cm² piece of comb suspended from the top, absorptive paper towel covering the bottom, and two feeding tubes (15 ml Falcon tubes with 1 mm holes in the lids) entering from the top and in contact with the comb. Each micro-colony was filled with 20 marked, day-old worker offspring from a single experimental queen in hybrid and inbred trials and 10 from each queen in mixed trials. Because these groups are much smaller than natural colonies, more replication is allowed and bees likely experience increased stress leading to greater hazard. Taken together, this likely results in greater sensitivity to differences between treatments. A culture of Serratia marcences (Thomas Scientific laboratory stock) was combined with equal parts Pro-Sweet (Mann Lake ltd. Z320) and dispersed to micro-colonies in the treatment group. Bacterial stock was brought to a standard viscosity by achieving a reading at A600 nm of 0.11 on a NanoDrop spectrophotometer (Thermo Scientific ND-2000), corresponding to roughly 3x10⁸ cells per mL (estimated via hemocytometer), after incubating for 3 days at 23 °C in LB broth (Thermo Scientific H26760.36). The control group was fed equal parts Pro-Sweet and filtered water. Micro-colonies were kept in a dark incubator (Percival I-36NL) at 34°C and checked at ~12-hr intervals to remove and record dead workers until all workers died. Within each replicate, mixed groups of all possible pairs of crosses were created, but for direct comparison to thermoregulation results, only the AA+BB mixed group was included in analysis (see supplemental figure 3 for summary of all mixed groups).

Results from three micro-colonies were removed from analysis due to holes in cages allowing workers to escape or enter, potentially compromising their results. In total, 22 trials and 13 controls were included for analysis. The survival of each micro-colony was estimated using the Kaplan-Meier survivor function (Kaplan & Meier, 1958) considering replicate as a stratification variable to allow for differences in baseline hazard function between family groups. Because stress likely increases as individuals are removed from trials, differences between treatments were tested using logrank tests (Peto & Peto, 1972) and estimated using a Cox proportional hazard model (Cox, 1972), neither of which assume a constant hazard function.

Thermoregulation Assay

The thermoregulation assay was performed in the field, following Jones et al., 2004, using newly-made queenless nucleus colonies sized for three langstroth frames. Within each replicate, five separate colonies were created using the offspring of experimental crosses: one for each inbred and outbred cross and one for the mixed group. Due to the premature death of an outbred queen (replicate 2), this gave a total of 14 colonies. After allowing inseminated queens to produce workers for at least 51 days to replace any unrelated workers in their (full-sized) home colonies, a total of 500 ml of workers were moved at night into each nucleus colony. For mixed nucleus colonies, one volume of 250 ml was added from each source, spraying bees with sugar water to minimize potential fighting during introduction. Each colony was provided a full frame of honey, a full frame of unrelated larvae, TempQueen pheromone (to simulate the presence of a queen; Mann Lake ltd. Z133), and a frame of empty comb. All nucleus colonies were moved to Purdue’s Lee Apiary in Americus, Indiana, roughly 20 km from natal colonies. They were placed randomly in full sun roughly 5 m apart with entrance directions varying randomly between SSE and SSW, and allowed to freely forage for seven days before temperature recordings began.

To measure thermoregulation ability, iButton (Thermocron DS1921H) temperature sensors were placed in the brood nest recording every five minutes for four days between August 30th and September 3rd, 2023 (Figure 3A). An additional sensor was placed in a nucleus box without bees to record ambient temperature. At the end of the period, all sensors and data were recovered. For each colony, we subset data into one-hour windows (12 readings per window, 96 total windows) and calculated two response variables: thermal variance as the variance in temperature readings and thermal error as the absolute difference between recorded and optimal nest temperature (supplemental figure 1). To determine significance of individual and social heterosis, hybrid and mixed treatments were compared to the average of inbred controls within each replicate using a paired t-test for each thermal variable (for explicit paired data, see supplemental figure 2). All tests were single tailed because the directions of inequalities were explicitly predicted (see introduction).