Grooming against Varroa: mandible shape and mite damage in feral honey bees
Annual losses of 30–40% have been reported in managed honey bee colonies (Kulhanek et al. 2017). The ectoparasitic mite Varroa destructor has been a major threat to European honey bees since its introduction into the United States (Akimov et al. 2004; Nazzi and Le Conte 2016). The mite feeds on bee fat body tissue and transmits viruses. Infestation can impair bee development, alter behaviour and shorten lifespan (Ramsey et al. 2019; Wilfert et al. 2016). The Asian honey bee (Apis cerana) is the original host of Varroa mites (Oldroyd 1999).Apis cerana has evolved several defences against Varroa, including grooming, removal of infested brood, other hygienic behaviours and suppression of mite reproduction (Grindrod and Martin 2023; Morfin et al. 2023). Some feral populations of Apis mellifera have also developed mite resistance.
Grooming is one of several behaviours associated with mite resistance. Bees can dislodge mites and may bite them in the process (Arechavaleta-Velasco et al. 2012; Büchler et al. 1992; Diao et al. 2018; Moritz et al. 2007; Peng et al. 1987; Ruttner and Hanel 1992; Seeley 2007; Morfin et al. 2020; Smith et al. 2021). Feral honey bees in Kentucky have been identified as Apis mellifera mellifera (Carpenter and Harpur 2021). A study of Africanised honey bees reported mite removal at a rate of 38.5% (Moretto et al. 1991). Puerto Rican gentle Africanised honey bees (gAHB, hybrids with A. m. scutellata ancestry) have also shown mite resistance in the context of local evolution (Avalos et al. 2017).
X-ray microcomputed tomography (micro-CT or µCT) produces three-dimensional images at fine spatial resolution with relatively little sample preparation (Palci et al. 2016). It has been used to study the anatomy and evolution of bees, ants and other insects (Coty et al. 2014; Larabee et al. 2017; Li et al. 2011; Smith et al. 2021). The authors used micro-CT to scan bee mandibles and compare their three-dimensional measurements. Researchers traditionally assessed mite damage associated with grooming by examining specimens under a microscope (Grindrod and Martin 2023). Mobile apps and image classification may make this assessment easier. A convolutional neural network (CNN) can learn visual features with relatively little manual preprocessing (Jogin et al. 2018). Transfer learning can adapt a pretrained network to classify new objects using a relatively small dataset (Shin et al. 2016).
The authors expected feral colonies in Ohio (A. m. ligustica) and Kentucky (A. m. mellifera) to show more mite-biting behaviour than commercial colonies. They compared mandible morphology in selected A. m. ligustica populations, gentle Africanised hybrids and commercial bees. They also developed a mobile app that uses CNNs to identify damaged mites in images.
Results
Grooming behaviour in feral Apis mellifera
Regression analysis found no significant difference associated with the total number of mites collected per colony. The authors compared the three populations using the proportion of damaged mites in each colony as an indicator of mite-biting behaviour. They collected 2,518 mites from 58 commercial colonies and 4,981 from 97 feral colonies. Kentucky feral A. m. mellifera colonies had a proportion of damaged mites about 34% higher than that of commercial colonies.Ohio feral A. m. ligustica colonies had a 49% higher proportion of damaged mites than commercial colonies. Both the Ohio and Kentucky feral groups showed evidence of grooming. Population was a significant predictor of the proportion of damaged mites.
The authors used ordinary least-squares regression to test whether the sampling year predicted grooming. They sampled 20, 28, 57 and 50 colonies in 2020, 2021, 2022 and 2023, respectively. Year was not a significant predictor.
Comparison of mandibles in Apis mellifera and Apis cerana
The authors measured four features of worker mandibles: long edge, short edge, height and the span of the spined area. The groups were A. cerana, gentle Africanised hybrids with A. m. scutellata ancestry, feral Ohio A. m. ligustica and commercial A. m. ligustica package bees. The long and short edges were significantly longer in the commercial bees than in the other three groups. Mandible height was greatest in A. cerana and lowest in feral Ohio A. m. ligustica. The span of the spined area did not differ significantly among groups.
Conclusions
The proportion of damaged mites was higher in feral Ohio A. m. ligustica and Kentucky A. m. mellifera colonies than in commercial colonies, consistent with more mite-biting during grooming. The authors also identified differences in mandible shape among A. cerana, feral Ohio A. m. ligustica, Puerto Rican gentle Africanised hybrids and commercial A. m. ligustica. They developed a mobile application that uses machine learning to identify damaged mites in images.
The total number of mites collected did not differ significantly between commercial and feral colonies, suggesting a similar distribution of mite counts across the colonies sampled. Grooming is one of several traits that may help reduce infestation in adult bees. Its effectiveness and heritability have been debated (Büchler 2010). More recent studies reported survival in colonies with high grooming activity even at infestation levels similar to those of control colonies (Nganso et al. 2017; Smith et al. 2021). Infestation also depends on factors such as genetic differences among mites, pathogens transmitted by Varroa and other interactions between host and parasite (Mondet et al. 2020; Le Conte et al. 2020; Moro et al. 2021; Zheng et al. 2023).
Further research could use population genetics to determine whether different Varroa lineages are present in the colonies as selection for grooming, one component of mite resistance, continues.Earlier research showed that workers can amputate the legs of Varroa mites (Ruttner and Hanel 1992). Ruttner and Hanel (1992) described visible white tissue at the cut ends of legs on damaged mites collected alive or dead.
Counting damaged mites revealed a significantly higher proportion in the feral Ohio and Kentucky colonies than in commercial colonies. The finding supports earlier evidence that grooming or mite-biting is associated with mite resistance in feral colonies (Russo et al. 2020; Smith et al. 2021).The proportions of damaged mites in feral Kentucky and Ohio colonies were 34% and 49% higher, respectively, than in commercial colonies. Grooming that involves removing and biting parasitic mites has been reported since 1929 (Haydak 1929; Frisch 1967; Smith et al. 2021).
The findings support considering grooming alongside other traits in breeding programmes for mite resistance (Danka et al. 2011; Rinderer et al. 2010; Spivak 1996). Van Alphen and Fernhout (2020) also reported high levels of grooming in feral colonies in Europe and South America.
Worker bees use their mandibles for several tasks, including defence against parasites (Papachristoforou et al. 2012; Ruttner and Hanel 1992). The authors found differences in the long edge, short edge and height among the studied populations of A. cerana and A. mellifera. Earlier scanning electron microscopy images of amputated mite legs led researchers to suggest that the two cutting edges of the mandible act like scissor blades (Ruttner and Hanel 1992). The cutting edges in feral and gentle Africanised bees tended to resemble those of A. cerana and were significantly shorter than those of commercial bees. The authors suggest that shorter edges might improve flexibility or efficiency when biting mites. The comparison does not establish that these differences evolved in response to Varroa.
The authors propose that natural selection since the introduction of Varroa in 1987 may have favoured more efficient mite removal in feral colonies (Hunt et al. 2016; Smith et al. 2021). This is consistent with earlier work reporting differences in the long cutting edge in Indiana colonies selected for mite-biting (Smith et al. 2021). A. cerana is generally slightly smaller than A. mellifera (Ruttner and Hanel 1992; Yue et al. 2018). In this study, the long and short cutting edges of feral bees were shorter than those of commercial bees and longer than or similar to those of A. cerana. Earlier work also found that heatwaves and temperature changes can influence worker body size (Poot-Báez et al. 2020).
Further comparisons of mandible size in feral and commercial colonies are needed. This study used the proportion of damaged mites as an indirect measure of grooming and assumed that the damage was caused by workers in the colony. Earlier work supports using damaged mites as an indicator of grooming (Büchler et al. 1992; Ruttner and Hanel 1992).
Direct observations of bees grooming live mites are difficult to obtain in colonies (Peng et al. 1987). Automated systems have been used to analyse aggressive behaviour in insects (Li-Byarlay et al. 2014; Lim et al. 2014). The authors call for field observations of bee grooming. Earlier work described a camera and machine-learning system for detecting Varroa in bee colonies (Bilik et al. 2024). In this study, the authors developed a CNN-based application to identify mite damage in images. It could reduce the time needed to assess damage, and a future version could run in the cloud.
Overall, the study found a higher proportion of damaged mites in feral A. m. ligustica and A. m. mellifera colonies than in commercial colonies, consistent with more mite-biting during grooming. The researchers found differences in the long edge, short edge and height of worker mandibles in their comparisons involving gentle Africanised hybrids and feral Ohio A. m. ligustica. They also developed a mobile app that uses a convolutional neural network to identify damage to mites. The findings help characterise grooming in feral colonies and point to questions for future research and breeding.
Apidologie (2025) © The Author(s), 2024 · Hongmei Li-Byarlay (Department of Agricultural and Life Sciences, Central State University, Wilberforce, OH 45384, USA) Kaila Young (Department of Agricultural and Life Sciences, Central State University, Wilberforce, OH 45384, USA) Xaryn Cleare (Department of Agricultural and Life Sciences, Central State University, Wilberforce, OH 45384, USA) Deng Cao (Department of Computer Science and Mathematics, Central State University, Wilberforce, OH 45384, USA) Shudong Luo (Institute of Apicultural Research, Chinese Academy of Agricultural Sciences, Beijing 102600, China)
Full article https://doi.org/10.1007/s13592-024-01126-z
Comments