Samples used in this study were collected by multiple researchers from 1989 to 2018 using a variety of collection and preservation techniques. Table 1 gives locality and collection information, and sample IDs corresponding to those used in Figure 1 . Most specimens were collected directly from colonies, though some bees were collected while they were foraging. Most specimens are adult worker bees, while a few are pupae collected directly from nests. Individual bees or bee thoraces were preserved in the field in liquid nitrogen (1988–1990) or in 95% ethanol (1991 onwards). Frozen specimens were later stored at −80°C. Ethanol-preserved specimens were stored at 4° to −20°C.

Genomic DNA was extracted from the mitochondrion-rich thoracic flight muscle tissue using DNA spin-columns, primarily the Qiagen DNEasy Blood and Tissue kit (www.Qiagen.com, Ann Arbor, MI USA) and the GenElute Mammalian Genomic DNA Miniprep kit (www.sigmaaldrich.com, St. Louis, MO USA) following the manufacturers’ recommendations. Extracted DNAs were stored at −20°C. Portions of the mitochondrial genome were amplified using the primers shown in Table 2 . These sequences included a large portion of cox1, leucine tRNA_UUR, a short non-coding sequence, and a portion of cox2. Figure 2 shows the relative position of the primers on the honey bee cox1 to cox2 sequences. Sanger sequencing was carried out at the Idaho State University Molecular Research Core Facility, Pocatello, ID. As only protein-coding sequence was included in the phylogenetic analysis, the tRNA and non-coding sequences were removed after alignment (see below) and the cox1 and cox2 sequences were concatenated. Some of the sequences were also obtained from Genbank ( Table 1 ). The total number of sequences for each taxon and their geographic origins are summarized in Table 3 .

Sequences were aligned manually with cox1 and cox2 sequences from Apis mellifera ligustica (Crozier and Crozier, 1993; Genbank accession L06178) in MEGA7 (Kumar et al., 2016). Sequences were screened for missing bases and correct reading frame by translating DNA sequences to amino acid sequences. In total, 46 sequences were used in the phylogenetic analysis and another four Genbank sequences of A. dorsata dorsata from India (Mizoram) and Myanmar ( Table 1 ) that were too short to include in the phylogenetic analysis were aligned with the larger data set to determine which sequences they matched most closely.

The best model of sequence evolution was selected using MEGA “Model Selection” analysis and the following conditions: maximum likelihood statistical methods, partial deletion of sites with missing data, coverage cutoff of 75%, all codon positions used, moderate branch swapping filter. The model of sequence evolution with the lowest Bayesian Information Criteria (BIC) score was selected for use in the phylogenetic analysis. This model (BIC score 13302.38) was a general time reversible model with non-uniform rates of evolution among sites (gamma distributed) and a fraction of sites seemingly invariable (GTR+G+I).

Phylogenetic trees were constructed using Maximum Likelihood methods in MEGA7 with the following settings: model of evolution gamma distributed with invariant sites (GTR+G+I) with 5 gamma categories, partial deletion of sites with missing data, 75% site coverage cutoff, all codon positions used, maximum likelihood heuristic method Subtree-Pruning-Regrafting-Fast, initial tree generated by Neighbor-Joining, moderate branch swap filter, 3 threads. Support for the branching patterns was evaluated with 1000 bootstrap replicates. Branches with less than 95% bootstrap support were collapsed. A coverage cutoff of 75% was chosen during model choice and tree-building to ensure that inclusion of shorter sequences did not result in elimination of informative data.

Although color, morphometric and morphological differences among A. d. dorsata, A. d. binghami and A. d. breviligula have been reported (e.g., Maa, 1953, but see Kitnya et al., 2024), there are no obvious morphological differences between A. d. SouthIndia and the widespread A. d. dorsata (Kitnya et al., 2024). In their study of A. laboriosa and A. dorsata in India, Kitnya et al. (2022) examined specimens of A. dorsata from Arunachal Pradesh in the extreme northeast of India and Karnataka (specifically Bangalore) in south India. In a dendogram displaying morphometric similarity of the samples, the south India specimens did not form a discrete cluster but were mixed in among the specimens from northeast India. However, in their phylogenetic analysis of the same collections (using a 500 bp fragment of the mitochondrial cox1 gene), the samples from Arunachal Pradesh and south India formed two separate clades with 99% and 100% bootstrap support, respectively. To the best of our knowledge, there is no information on the geographical distributions of A. d. dorsata and A. d. SouthIndia. Although giant honeybees have been collected at points between southern and northern India, at the moment the only way to tell the two apart is by genetic testing.

The distinctive nature of A. d. breviligula and A. d. binghami compared to the more widespread A. d. dorsata has long been recognized (e.g., Maa, 1953; Ruttner, 1988). A. d. breviligula and A. d. binghami are primarily black, with white stripes on metasomal tergites 3,4, and 5 formed by short white hairs, while the metasomal terga 1-3 (and sometimes tergites 4-5)and sterna 1-2 are yellow to brown in A. dorsata dorsata (Kitnya et al., 2024). In addition to differences in coloration, A. dorsata dorsata of mainland Asia differs from the two island taxa based on ocellus size and the spacing of the compound eyes and ocelli (Kitnya et al., 2024). Whether the two island forms constitute separate species remains to be determined. Kitnya et al. (2024) found no morphological basis for separating the two, and considered them a single species, distinct from A. d. dorsata, with two subspecies: A. binghami binghami and A. b. breviligula. Evidence from mitochondrial gene sequences presented here retrieves A. d. binghami and A. d. breviligula as genetically distinct but is insufficient determine species status.

The fact that isolated island populations show traits distinct from those of mainland populations is not surprising. What is more surprising is the presence of a genetically distinct giant honey bee in southern India, along with the more widespread A. dorsata dorsata in northern India. However, a broader view of Indian Apis shows that this pattern has appeared more than once. India is also home to two cavity-nesting bees, the yellow or plains bee, and the hill or black bee (Ruttner, 1988 and references cited therein; Bhatta et al., 2020). According to Engel (2002) the yellow or plains bee corresponds to A. cerana indica Fabricius, 1798 while the black or hill bee corresponds to A. cerana cerana Fabricius, 1793. Genetic evidence collected over the past three decades (e.g., mitochondrial cox1, cox2 and non-coding sequences, Smith, 1991; Smith and Hagen, 1996; mitochondrial and nuclear gene sequences, Lo et al., 2010; and genomic SNPs, Su et al., 2023) support species status of the yellow Indian bee, as proposed by Lo et al., 2010; Smith, 2011, and Su et al., 2023. The dwarf honey bee, Apis florea, also consists of two distinct groups revealed by mitochondrial cox1-cox2 sequences and nuclear SNPs (Smith, 2011; Su et al., 2023, and Figure 1 of this study). These are an eastern lineage including populations from Thailand eastwards, and a western lineage including populations from India westward, including the invasive dwarf honey bee populations in Jordan and Israel, and probably those in east Africa as well. The “switchover” from East to West is apparently in the poorly sampled region from northeastern India through Bangladesh and Myanmar.

Why does India have a distinct variety of cavity-nesting bee, A. cerana indica, along with A. cerana cerana, and a distinct south Indian variety of giant honey bee along with A. dorsata dorsata in northern India? And why does it have a variety of A. florea different from that in eastern Asia? Answering these questions requires (1) information on the ranges of the species and putative species of Apis in India, particularly the distributions of the yellow and black cavity-nesting bees, and A. d. SouthIndia and A. dorsata dorsata, and (2) a phylogeny that resolves the branching patterns of the four lineages within A. dorsata in the broad sense. A robust phylogeny would provide information on the order and timing of diversification events. A time calibrated phylogeny could suggest specific geological and climatic events that could have promoted diversification, and help us determine if the dwarf, giant, and cavity nesting lineages responded to historical events with similar patterns of diversification.

At least three of the four lineages within A. dorsata in the broad sense exhibit migratory behavior. The vast majority of giant honey bee migration research has been carried out on populations that our study would place in A. d. dorsata (for example, Dyer and Seeley, 1994; Kahono et al., 1999; Neumann et al., 2000; Paar et al., 2000; Itioka et al., 2001; Rattanawannee et al., 2013), which is not surprising, as it is the most widespread. Koeniger and Koeniger (1980) investigated giant honey bee migration in Sri Lanka; though we have not sampled any giant bees from Sri Lanka, we predict that they are part of the A. d. SouthIndia clade, based on the fact that the south Indian plains bee, A. c. indica, is also found in Sri Lanka. At least one set of observations has been made on migration by A. d. binghami in Sulawesi (Nagir et al., 2016). Morse and Laigo (1969, cited in Robinson, 2021) reported that A. d. breviligula in the Philippines does not migrate.

Migration is typically a predictable annual response to seasonal patterns of rainfall and resource availability (e.g., Dyer and Seeley, 1994) or a response to erratically occurring masting events in which forest trees produce a superabundance of blossoms and resources (Itioka et al., 2001). Migrating bees appear to show fidelity to their nesting sites at either end of the migratory route (Neumann et al., 2000; Paar et al., 2000). This alone means that protecting a giant honey bee’s nesting and foraging habitat means protecting more than one location. Migrating colonies of giant honey bees may travel distances that require “rest stops” to forage. Recent work by Robinson, (2012, 2021) has shown that migrating A. d. dorsata in Thailand make use of “traditional” rest stops, where they forage for food and water for variable lengths of time before continuing their journey. These rest stops are likely to be crucial for successful migration. To maintain the genetic diversity represented by linages within A. dorsata in the broad sense, it will be necessary to maintain not only the endpoints of their migratory routes, but quite probably sufficient rest stops along the route too, in conditions that provide the bees with the forage, resting sites and nesting sites they need. This would require tracking migration routes and their timing and noting changes in migration timing or route due to climate change or habitat destruction.

Warrit et al. (2024) discuss the challenges facing bee research and bee conservation in Asia, noting, “If we do not know the species present, their distribution and threats, we cannot protect them.” They point to the eusocial bees as “flagship species” for bee conservation measures, as their economic value to humans—through pollination services and honey production—is generally known to the public. In particular, “the honey bee” is likely the only bee most people know, especially in urban populations. Although the giant bees are large, conspicuous, and widespread across the Asian continent, we are still discovering new diversity (at the species or subspecies level) and still lack basic information on the ranges of some lineages such as the south Indian giant bee.

Giant honey bees are not just major pollinators in Asian ecosystems. With their large, conspicuous open-air combs, large aggregations of nests, and migratory behavior, plus the well-publicized harvesting of cliff-side A. laboriosa nests, they are arguably the most charismatic of the Asian social bees. Public support for protection of giant honey bees would also have the effect of protecting habitat for the many other social and solitary Asian bee species.

Our results suggest avenues for additional research, particularly regarding Indian populations. What is the range of the South Indian giant honey bee, and what are its migration patterns? Is the range of the southern Indian cavity-nesting “plains bee” (currently A. cerana indica) congruent with the range of the southern Indian giant honey bee, suggesting similar biogeographic history? Does the south Indian giant honey bee differ in behavior or ecology from A. d. dorsata? And of course, will behavioral and genetic study of giant honey bees from a greater portion of their ranges (e.g., as in Cao et al., 2012b) reveal more diversity?