Multiple mitogenomes indicate Things Fall Apart with Out of Africa or Asia hypotheses for the phylogeographic evolution of Honey Bees (Apis mellifera)

Oct 19, 2023

Scientific Reports volume 13, Article number: 9386 (2023) Cite this article

Metrics details

Previous morpho-molecular studies of evolutionary relationships within the economically important genus of honey bees (Apis), including the Western Honey Bee (A. mellifera L.), have suggested Out of Africa or Asia origins and subsequent spread to Europe. I test these hypotheses by a meta-analysis of complete mitochondrial DNA coding regions (11.0 kbp) from 22 nominal subspecies represented by 78 individual sequences in A. mellifera. Parsimony, distance, and likelihood analyses identify six nested clades: Things Fall Apart with Out of Africa or Asia hypotheses. Molecular clock-calibrated phylogeographic analysis shows instead a basal origin of A. m. mellifera in Europe ~ 780 Kya, and expansion to Southeast Europe and Asia Minor ~ 720 Kya. Eurasian bees spread southward via a Levantine/Nilotic/Arabian corridor into Africa ~ 540 Kya. An African clade re-established in Iberia ~ 100 Kya spread thereafter to westerly Mediterranean islands and back into North Africa. Nominal subspecies within the Asia Minor and Mediterranean clades are less differentiated than are individuals within other subspecies. Names matter: paraphyletic anomalies are artefacts of mis-referral in GenBank of sequences to the wrong subspecies, or use of faulty sequences, which are clarified by inclusion of multiple sequences from available subspecies.

The genus Apis comprises at least a dozen nominal species of Honey Bees found throughout Asia, Africa, the Middle East, and Europe, including the type species A. mellifera L., the Western Honey Bee (Table 1)1,2. Although several species are important as traditional local sources of honey, only two cavity-nesting species, A. cerana in India and A. mellifera, have historically been domesticated for this purpose. Apiculture of A. mellifera is documented in Egypt as far back as 2600 BCE3. (References to the Levant as "a land of milk and honey" (Exodus 3:8) may refer instead to syrup from wind- and (or) hand-pollinated oasis dates (Phoenix dactylifera) of even greater antiquity3). A. mellifera has been transported by humans around the world as the principal commercial source of honey and beeswax, and is also of particular agricultural importance in the Americas as an introduced pollinator of certain vegetable and fruit crops that are themselves introduced4. Competition between introduced generalist honey bees and specialist native insect pollinator (including other non-apine bees) for pollen ("pollen theft") may be detrimental to native plant species5. Geographic variation among different local strains or subspecies of A. mellifera is known to contribute to this impact, as it does quality and quantity of honey and other behaviors6. Notably, so-called "Africanized" bees have resulted from the escape of African subspecies crossed into domesticated South American hives, where the hybrids combine increased yield of honey with greater aggressiveness, and pose a hazard to humans7.

Morpho-behavioral taxonomy has recognized three species groups within the genus: Dwarf Honey Bees (A. florea and A. andreniformis), Giant Honey Bees (A. laboriosa, A. dorsata, and A. breviligula), and Cavity-Nesting Honey Bees (A. mellifera together with A. nuluensis, A. nigrocincta, A. cerana (including A. indica), and A. koschevnikovi). A. mellifera is the only species with a native range in Africa and Europe, and includes more than 30 nominal subspecies2. These have historically been clustered as four continental groups, designated ACMO for African, Continental, Mellifera, and Oriental distributions8. The geographic origin and evolutionary spread of these groups remains controversial, as has their resolution into ancestor–descendant phylogenetic lineages. Molecular studies based on analyses of various components of the nuclear genome (see Discussion) agree broadly on rearrangement of these as an MAOC backbone, rooted so as to offer alternative theories of origin, either "Thrice Out of Africa"9 [root within A] or "Thrice Out of Asia"10 [root between A & O]. These studies are based on fewer than a dozen of the available subspecies of A. mellifera. Other phylogeographic conclusions have been reached from analyses of complete mitochondrial DNA (mtDNA) genomes from single sequences per subspecies11,12. In particular, Tihelka and co-workers in this journal12 have recently provided a meta-analysis of mtDNA, based on a substantial body of data from Boardman et al.11 and new methods of phylogenetic inference. They reached yet another alternative hypothesis, a Middle Eastern/North African origin of A. mellifera.

Tihelka and co-workers emphasize the need for a reliable backbone phylogeny for A. mellifera L. to understand the evolution of the subspecies, including its geographic origin and the development of adaptative differences among subspecies that contribute to their ecological and commercial success6. My editorial review of the use of an mtDNA spacer region between the CO1 and CO2 genes as a tool for discrimination of individual bees within and among subspecies of A. mellifera [cf. Ref.13] found no comprehensive phylogenetic evaluation that included multiple mtDNA genome variants within multiple subspecies11,12,14,15,16,17,18,19,20,21,22,23,24,25,26. Unexpectedly, my preliminary assembly of multiple sequences assigned to A. m. mellifera in GenBank, taken at face value, seemed to indicate extensive para- and (or) polyphyly and (or) extraordinary genetic diversity within this subspecies and with respect to other taxa.

In the twenty-first century CE and the third century L., the importance of classical "alpha" taxonomy27 in finding, describing, and naming taxa remains critical. I provide here a phylogeographic re-evaluation of the evolution of Apis, based on a meta-analysis of the 13 coding regions (11,006 bp) from available mitogenomes in nine species of the genus Apis, including 78 individuals from 22 subspecies of the type species A. mellifera L. This meta-analysis indicates that Things Fall Apart with Out of Africa, Asia, and Middle East hypotheses, in favor of a European origin.

Complete mitochondrial DNA genomes of taxa within the genus Apis were identified in the GenBank Taxonomic library by selection of "Genome"-flagged sequences within the display of species and subspecies. This was supplemented by a search of the "Nucleotide" library with the search term "Apis mitochondrion" through the end of October 2022. Table 1 lists GenBank accession numbers for sequences from nine species of Apis and 22 subspecies of A. m. mellifera as used here, together with the subspecies’ geographic provenance.

The mitogenomic sequence of a curated Norwegian specimen of A. m. mellifera (KY926884) was used as the alignment reference. Alignment was done by eye with the help the MEGA X program28. The Apis mitochondrial coding genome comprises 13 genes over 11,043 bps (Table 2). The GenBank annotations of various authors delimit coding regions with slightly different 5′ and 3′ endpoints, especially among species: spaces were inserted between coding regions so as to preserve open reading frames. The light-strand coding regions of the ND5, ND4, ND4L, and ND1 genes were included in their sense-strand equivalent 5′ → 3′ coding directions.

The 3′ region of the ND4 heavy-strand equivalent is difficult to align across subspecies, and a 37 bp region was excluded from all analyses. Ambiguous base calls (mostly W for A/T, flanked by AT bases) in several sequences have been tacitly resolved, such that these positions are invariant over all sequences. Several autapomorphic insertions of single triplets among species and subspecies have been tacitly removed. The coding region mitogenome sequence for a Bumble Bee, Bombus ignitus (GenBank NC010967), was used as the outgroup for inter-species comparisons. Honey bees and Bumble Bees occur in the tribes Apini and Bombini, respectively, of the subfamily Apinae. The alignment of coding regions between Bombus and Apis spp. is in some areas speculative: experimental inclusion or exclusion of ambiguous regions does not affect the inferred branching order of Apis species, and does not materially affect its statistical support. The consensus alignment for analysis comprises 11,006 bases within A. mellifera subspecies, and 11,070 bases across Apis and Bombus.

I performed three forms of phylogenetic analyses with MEGA X28. Maximum Parsimony (MP) analyses was performed with all nucleotide positions equally weighted, and SPR search. Maximum Parsimony analysis is foregrounded to identify inter-subspecific SNPs and to provide intra-specific patristic differences within Apis, and intra- versus inter-subspecific differences within A. mellifera. Maximum Likelihood analyses (ML) was performed with the general time reversible (GTR) model allowing for invariant sites and nearest-neighbor interchange (NNI) search. For molecular clock calculations, I used a fixed rate of 0.0115 substitutions/site/Myr on node distances calculated in a linearized Maximum Likelihood tree RelTime-ML model in MEGA (see "Discussion"). Neighbor Joining (NJ) analysis was performed on counts of differences, with the maximum composite likelihood model, and SPR search. Statistical confidence in all three methods was estimated from 3000 bootstrap replications each, under the same conditions as the main search. Concern has been expressed12 as to the effect of inclusion of third triplet-position bases on phylogenetic inference. I examined the effect of exclusion of these data (model P12 and modifications thereof12) in all three analyses.

Initial assessment of the available mitogenomes for A. mellifera identified 16 Arabian bees in four monophyletic clades assigned to A. m. mellifera GenBank accessions MT745901–MT74591514 that comprise sets of 5, 4, 4, and 3 identical sequences each: one sequence was included from each set, and the rest tacitly. A set of 11 mitogenomes from Kenyan bees (KJ396181–KJ396191)15 assigned in GenBank to A. m. mellifera are used here with their correct subspecies identifications as provided by Z. Fuller (pers. comm.). A series of 20 sequences attributed to A. m. capensis (MG552683–692) or A. m. scutellata (MG552693–702) was retained in full. GenBank accessions KY926882 and KY926883 are attributed to A. m. syriaca and A. m. intermissa, respectively. Inspection of both sequences indicates numerous anomalies in coding region that produce branch attraction between the two and large phylogenetic separations from other GenBank accessions assigned to the same subspecies (Supplementary Fig. S1). Both are excluded from primary analyses here.

All figures were drawn with Corel PaintShop Pro 2023 (version 25).

Figure 1a shows the MP analysis for nine species of Apis (Apini) with Bombus ignitus (Bombini) as outgroup. Out of seven interspecies nodes, six are supported by > 94% of bootstraps. Analyses with the NJ and ML methods give identical branching orders and substantially similar bootstrap support. Within Apis, the Dwarf Honey Bee pair [A. florea + A. andreniformis] is the outgroup to the remaining taxa. Within the Cavity-Nesting species, A. mellifera is the sister species to the others, and its phenetic difference from other cavity-nesting species is greater than that between the Dwarf and Giant Honey Bee pairs. Figure 1b shows the MP analysis of Apis species alone including ten subspecies of Apis mellifera: A. m. mellifera is sister to the remaining subspecies. Note that A. florea and A. mellifera, the only species whose native ranges overlap in the Middle East, are in separate morpho-behavioral and molecular clades.

Maximum Parsimony analysis of phylogenetic relationships among mtDNA genomes sequences of nine species of Apis honey bees. (a) The tree is rooted with a bumblebee Bombus ignitus as outgroup. A. m. mellifera (KY926884) is the basal-most member of that species and the alignment reference. Numbers above branches are inferred numbers of nucleotide substitutions; numbers in bold below branches are percent support in 3000 bootstrap replicates. Identical branching order and substantially similar bootstrap support are given by Maximum Likelihood and Neighbor Joining methods. (b) As above, with removal of Bombus and addition of key subspecies of A. mellifera (cf. Fig. 2).

Figure 2 shows the schematic MP analysis of 66 (+ 12 tacit) sequences from 22 subspecies of A. mellifera, with multiple sequences included as described in Methods. The tree is rooted with A. m. mellifera as obtained from the analysis shown in Figs. 1b. MP, ML and NJ analyses (Supplementary Figs. S1–S3, with complete GenBank accession numbers) give the same branching order among subspecies and most multiple individuals within subspecies, and substantially similar bootstrap support.

Schematic maximum parsimony analysis of phylogenetic relationships among mtDNA genome sequences of 66 individual A. mellifera honey bees from 22 subspecies. Rooting as indicated by Fig. 1. Numbers above branches are inferred numbers of nucleotide substitutions; numbers in bold below branches are percent support in 3000 bootstrap replicates, with SPR branching swapping. The tree shown is one of nine minimum length trees that differ only by rearrangements at unresolved nodes. Sequences in the Sub-Saharan clade that make that subspecies paraphyletic are tagged in Roman font, as are two sequences referred to A. m. scutellata that are outside that clade. An additional 12 sequences from the Arabian series that are identical to the four shown are not included. Sequences curated as A. m. mellifera in GenBank15 are re-assigned their proper names in parentheses (J. M. Fuller, pers. comm). Subspecies represented by single sequences are indicated by (*). Named phylogeographic clades discussed in the text are indicated in color. The complete MP tree with GenBank accession numbers is given in Supplementary Fig. S1, along with those for Maximum Likelihood and Neighbor Joining methods (Supplementary Figs. S2 and S3, respectively).

The consensus branching order identifies a series of well-defined, successively nested phylogeographic clades, with European A. m. mellifera sensu lato ([[KY926884 + A. m. sinisxinyuan] + Arabian MT745913]; MT745912, 914, and 915 not shown) basal to the remainder. Their geographic provenances are as follows (Fig. 3):

Southeast European = [A. m. ligustica + [A. m. carnica + A. m. carpatica]]. The pairwise patristic difference between A. m. carnica + A. m. carpatica (14) is less than the maximum between variant A. m. ligustica sequences (32).

Asia Minor = [A. m. meda + [A. m. caucasia + [A. m. anatoliaca]]], with both nodes supported > 95%. Maximum pairwise patristic differences between subspecies sequences (5–10) are also less than those within A. m. ligustica.

Levantine/Nilotic/Arabian = [A. m. lamarckii + [A. m. jemenitica + [A. m. syriaca]]] as three separate subspecies lineages. The pairwise patristic difference between A.m. lamarckii and A. m. syriaca (44) is less than the maximum within A. m. jemenitica (49).

Mediterranean = [A. m. iberiensis + [[A. m. intermissa + A. m. sahariensis] + [A. m. ruttneri + A. m. siciliana]]]], with all nodes supported > 99%. The mainland European variant is basal to the North African variants, and the two island variants are paired. Maximum pairwise differences between subspecies sequences (3 ~ 25) are less than those within A. m. ligustica or A. m. jemenitica.

Sub-Saharan = paraphyletic assemblage of sequences assigned to A. m. scutellata and A. m. capensis. A. m. scutellata occurs in two clades, one of which is more closely related to A. m. capensis. Monophyly of either subspecies is further confounded by scattered referral of individual sequences to the other subspecies across these three clades. The single A. m. adansonii type is paired with an A. m. scutellata sequence. The two A. m. monticola variants encompass a third assigned to A. m. capensis.

Phylogeographic evolution in context of geographic distribution2 of subspecies of A. mellifera as inferred from mitogenomic data. Numbered symbols indicate five clades described in the text and in Fig. 2. Dark and light green circles indicate respectively subspecies in the Southeast European and Asia Minor clades included within the Eurasian superclade. Blue symbols indicate the Levantine (circles), Nilotic (squares), and Arabian (A. m. jemenitica) (diamonds) clades. Light and dark purple circles indicate independent A. m. simensis and A. m. unicolor lineages, respectively. Light orange symbols indicate subspecies in the Mediterranean clade. Red circles indicate the paraphyletic assemblage of A. m. scutellata and A. m. capensis, including A. m. adansonii (light red) and A. m. monticola (brown). Base map modified from [https://commons.wikimedia.org/wiki/File:BlankMap-World.svg].

The (SE European + Asia Minor) clades constitute an inclusive Eurasian superclade, and an African clade can be recognized as [Ethiopian A. m. simensis + [Malagasy A. m. unicolor + [Mediterranean + Sub-Saharan]]], allowing for the mainland European and insular taxa in the Mediterranean clade.

Pairwise differences between all nominal subspecies represented by single sequences within Clades I, II, and IV are smaller than those between replicate sequences in the same or sister lineages (cf. A. m. ligustica, A. m. jemenitica, A. m. scutellata, and A. m. capensis). The same is true for a single sequence from A. m. adansonii (from Niger) with respect to A. m. scutellata, and two sequences from A. m. monticola (from the east African mountains) with respect to A. m. capensis in the Sub-Saharan lineage.

Small but consistent majorities of bootstraps in all phylogenetic methods supports the Levantine/Nilotic/Arabian clade as sister to the African clade, and the Mediterranean clade as sister to the Sub-Saharan clade. A pair of near-identical Kenyan bee sequences (KJ396184 and KJ396190)15 referred to A. m. scutellata by Z. Fuller (pers. comm.) is cladisitically separate from other scutellata and both the Mediterranean and Sub-Saharan clades; its exact position varies slightly among methods.

With a fixed clock rate of 0.0115 subs/site/Myr (see "Discussion"), all node distances in linearized ML trees can be converted directly to their age of divergence in a molecular clock (Fig. 4). For example, given a calculated distance to the basal node in Fig. 4 of 0.008234 subs/site (Relative Time), the node is dated at (0.008234 subs/site)/(0.0000115 subs/site/Kyr) = 716 Ka. For Bombus and other Apis species included as in Fig. 1, the molecular clock indicates a late Miocene radiation (6–11 Mya) of morpho-behavioral groups within Apis (Supplementary Fig. S4), and a European origin of A. mellifera 780 Kya in the late Pleistocene. The molecular clock for subspecies of A. mellifera is shown in Fig. 4 for a subset of 37 sequences included in Fig. 2a, with single representative sequences for A. m. scutellata and A. m. capensis.

An mtDNA-based molecular clock for within- and among-subspecies divergences of A. mellifera. Sequences are coded as in Fig. 2: only one representative each of A. m. scutellata and A. m. capensis is included (n = 37). Divergence times are calculated from a linearized ML model with A. m. mellifera sequences as the designated outgroup (cf. Supplementary Fig. S2). The clock is calibrated from the mean linearized nucleotide subs/site distances to each node (Relative Time) at 0.0115 subs/site/Myr (see text for sample calculation). See Supplementary Fig. S4 for the clock of A. mellifera within Apis, with Bombus as the designated outgroup.

"I have just been thinking, and I have come to a very important decision. These are the wrong sort of bees."—Winnie the Pooh.

The mitogenomic analysis confirms with high confidence previous inferences as to morpho-behavioral relationships among nine species as Dwarf, Giant, and Cavity-Nesting Honey Bees (Figs. 1a,b). Within the Apini, rooting with Bombus (Bombini) confirms Dwarf Honey Bees [A. florea + A. andreniformis] as outgroup to the Giant Honey Bees [A. dorsata + A. laboriosa] and the remaining, Cavity-Nesting Honey Bees. A. mellifera is the sister group to these [A. koschevnikovi + [A. nigrocincta + [A. nuluensis + A. cerana]]], at a greater phenetic distances than among any other Apis species pairs.

Mitogenomic sequences assigned to the taxon A. m. mellifera in GenBank occur in nine lineages collectively made paraphyletic by the placement of other subspecies. The phylogenetic analysis indicates that most of these apparent anomalies are artefacts of miss-assignment of individual sequences to the type subspecies. As noted, the Kenyan bee series when correctly referred to subspecies accords with the cladistic arrangement described here. The Arabian bee series referred to A. m. mellifera in GenBank falls phylogenetically into two distinct clades, one as part of the basal A. m. mellifera sensu lato and the other A. m. jemenitica sensu lato, along with Kenyan bees reassigned to the same subspecies. Another two Kenyan sequences assigned to A. m. mellifera are more closely related to A. m. simensis and A. m. unicolor, respectively, which suggests they are in fact members of those two subspecies. The latter extends the subspecies range beyond insular Malagasy.

The received picture of the evolution of A. mellifera has been an "Out of Africa" model9, with as many as three Recent excursions, originally to Europe, more recently to Iberia, and in historic times to South America as so-called "Killer" or "Africanized" Bees, which resulted from accidental crossing of imported A. m. scutellata queen with local A. m. ligustica drones7. This has been challenged more recently by models of "Out of Asia"10 or North African/Middle Eastern12 origins. The alternative phylogeographic clade structure presented here indicates that Things Fall Apart with any of these hypotheses. The mitogenomic phylogeography (Fig. 3) indicates instead that A. mellifera evolved as a North-to-South expansion from Europe to Africa by way of Asia Minor and the Levantine. The type form A. m. mellifera sensu stricto originated in northern Europe, and has diversified as A. m. ligustica in southeastern Europe, including expansion into Asia Minor (A. m. caucasia). European bees then spread southward via the Levant (A. m. syriaca) into Nilotic East Africa (A. m. lamarckii), and across the Red Sea into the Yemeni coast of the Arabian Peninsula (A. m. jemenitica). Southward expansion has left Ethiopian (A. m. simensis) and Malagasy (A. m. unicolor) lineages as earlier offshoots. Sub-Saharan mainland forms constitute a single clade that comprises individuals referred to A. m. scutellata and A. m. capensis, as well as A. m. adansonii and A. m. monticola both closely related to sequences assigned to the former and latter subspecies, respectively. A. m. adansonii occurs throughout central Africa: the single available sequence from Niger may not be representative. The structure of the Mediterranean clade indicates a secondary return to Europe during a refugial period, and a more recent tertiary return to North Africa via the western Mediterranean islands.

Divergence times within and among animal species, including Insecta, are routinely estimated from measured molecular divergences, including those made from mtDNA genomes29, including Apis30. Calibration of a molecular clock requires reliably dated external events, often geographic31, and accurate measurements of rates of substitution or divergence (2 × the former). Reliable geographic events are unavailable for A. mellifera30; substitutions per bp measured here over complete coding regions avoid gene-specific variance. Based on a fixed rate of 0.0115 substitutions/site/Myr (Brower 1994 in Ref.31) diversification of major lineages within A. mellifera can be dated to the Chibanian Age (late Pleistocene), 770–126 Kya. This coincides with and may be influenced by the European Günz glacial cycles. Separation of more southerly lineages from continental European subspecies occurred ca. 720 Kya, separation of African-endemic from Levantine/Nilotic/Arabian lineages ca. 660 Kya, establishment in Africa ca. 540 Kya, and separation of Euro-African Mediterranean from Sub-Saharan lineages ca. 250 Kya. Diversity within several subspecies dates to > 100 Kya, notably among multiple replicates of A. m. ligustica and A. m. jemenitica, 120 Ka and 150 Ka, respectively. Differences among other nominal subspecies originate only a few 10 s of Kya, notably the Asia Minor forms (A. m. meda, A. m. caucasia, and A. m. anatoliaca, 10–20 Ka) and Mediterranean species including insular forms (A. m. ruttneri and A. m. siciliana, 40 Kya).

The standard classificatory and evolutionary system of Apis is based on meristic analysis of four continental groups, African (A), European (C), Mellifera (M), and Asian (O)8. In Fig. 5 (redrawn from Fig. 10.8 in Ref.8), the placement of individual bees in Principal Components space corresponds roughly to the four quadrants of the first two axes, clockwise from the upper left as A, M, C, and O, respectively. Re-tagged and grouped genetically, the Arabian/Nilotic clade overlies the Sub-Saharan clade, including local variants. However, the Asia Minor A. m. caucasia clade is distributed along the M → O axis, and is bisected by its Southeast Europe sister clade A. m. ligustica. Although the African and European/Asian clades are essentially non-overlapping, the Mediterranean clade overlies all three, as well as the basal A. m. mellifera clade. Despite their dispersion in PC space, A. m. caucasia and A. m. iberiensis are the least genetically diverse clades; individuals in the latter plot are almost entirely African. Consolidation of all "African" subspecies as a single group does not recognize the origins of the Euro-African-Mediterranean subspecies, nor does a single "Asian" group recognize the complex biogeographic connection of Eurasian subspecies in southeast Europe and Asia Minor, or of these with Africa via the Levant.

Molecular clades of A. mellifera mapped onto the first two PCA axes of morphometric ACMO space, redrawn after Fig. 10.8 in Ref.8. Individual bees are identified to subspecies by the numeric color codes, and re-grouped to genetic clades as in Fig. 3, with additional color variants for Sub-Saharan taxa. Ruttner's four groups [African (A), European (C [Continental]), Mellifera (M), and Asian (O [Oriental]] correspond roughly to the four quadrants, clockwise from the upper left as A, M, C, and O. The Levantine/Nilotic/Arabian clade overlays the Sub-Saharan clade , including local variants. The Asia Minor A. m. caucasia clade is distributed along the M → O axis, and is bisected by its Southeast European sister clade A. m. ligustica. The Afro-European Mediterranean clade overlies all three, as well as the basal A. m. mellifera clade.

The ACMO model is taken as support for an argument against an Oligocene/early Miocene dispersal from Europe to Africa. More recent morphological analysis of wing-venation patterns including fossil material32,33 suggests a European origin of A. mellifera antecedents from an A. cerana/dorsata type in the Miocene, in agreement with the molecular clock estimate here. However, the hypothesized evolution of the A. mellifera type in Sub-Saharan Africa late in the epoch is according to the clock too far back, and return to Europe in the Holocene is both too far forward and in the opposite direction. Despite European antecedents, this remains a "(Long) Out of Africa" model. The similarity dendrogram in Fig. 1 of Ref.32 shows [Bombus + [[A. dorsata + [A. florea + [A. cerana / A. mellifera]]]]] with cerana & mellifera forms co-mingled. This arrangement transposes A. florea and A. dorsata and therefore the Dwarf and Giant Honey Bee morphotypes (cf. Fig. 1a), and fails to resolve the deep phylogenetic separation of A. cerana / A. mellifera (cf. Fig. 1b and Supplementary Fig. S4).

Previous meta-analyses by Boardman and Eimanifar et al.11,16,17,18,19,20,21,22,25,26 and Tihelka et al.12 are compared with the present analysis in Fig. 6. The base phylogram is a Maximum Parsimony analysis calculated as in Fig. 2, with the addition of two problematic sequences noted in the text, KY926882 and KY926883, attributed to A. m. syriaca and A. m. intermissa, respectively. KY926882 "" falls outside the jemenitica cluster that includes A. m. syriaca KP163643, which is not used in either the "" or "" sets. KY936883 "" is more closely related to KY936882 than to A. m. intermissa KM458618, which occurs where predicted biogeographically. There are indications of faulty data in KY936882 and KY936883: inspection of the last quarter of the aligned sequences shows two unbroken runs of seven shared phylogenetically informative sites31 that unite KY936882 and KY936883 to the exclusion of their nominal subspecific sisters. In the Sub-Saharan clade, the A. m. capensis/A. m. scutellata pair (KX870183/KY614238) in the "" set occurs in A. m. capensis + A. m. scutellata clade I (but not II), and in the "" set MG552698 appears to be mis-referred to A. m. scutellata instead of capensis. A. m. adansonii (MN585109) "" is more closely related to the other, larger clade of A. m. scutellata. A. m. monticola (MF678581) "" is more closely related to A. m. capensis. Thus, reliance on single, atypical sequences in previous meta-analyses has led to faulty phylogenetic inferences.

Comparison of Boardman et al.11 and Tihelka et al.20 meta-analyses with Fig. 2. The base phylogram is a Maximum Parsimony analysis calculated as in Fig. 2, with the addition of two problematic sequences mentioned in the text, KY926882 and KY926883, attributed to A. m. syriaca and A. m. intermissa, respectively. The 17 sequences used by Boardman et al. are marked "" and the 16 sequences used by Tihelka et al. "", with the 11 sequences common to both sets "". Note the anomalous pairwise placements of two sequences attributed to KY026882 and KY926883 with respect to their nominal sister subspecies ( and , respectively).

Boardman, Eimanifar et al. [11 et seq.] root their species-inclusive trees at the midpoint, rather than by an outgroup, which as drawn implies that the Dwarf and Giant Honey Bee species are sister groups, rather than that the former are the outgroup to the remaining species (cf. Fig. 1b). Their alignments include rDNA along with Coding Region sequences, which may contribute to the relatively weaker bootstrap support for some nodes: among subspecies of A. mellifera, my trial rDNA alignments were problematic at best, and these regions were excluded from the analysis here.

Tihelka et al.12 inferred their phylogeny by a BI site-heterogeneous mixture model. The root with respect to A. cerana falls between A. m. intermissa and other African (A) subspecies (notably A. m. scutellata), in support of an African origin, in contrast to Figs. 2 and 3. A. m. mellifera and A. m. iberiensis (M) are on the same side of the root as A. m. intermissa. The Asian (O) subspecies are further along the backbone, followed by the European (C) subspecies. The Asian form indicated by "?" is unidentified: the likeliest candidate seems to be A. m. caucasia. Of 11 nodes in their tree, only the C group is found in Fig. 2 here, and none corresponding to groups A, M, O, or any of the three-member clades. Exclusion of third-positions SNPs in their P12 model yields essentially the inverse of the trees in Fig. 2, with African lineages first derived, and A. m. mellifera as outgroup to the Eurasian lineages. Their model transposes the African and Eurasian lineages to either side of A. m. mellifera. At least two nearest-neighbor subspecies pairs are markedly different: East African A. m. simensis/A. m. unicolor, A. m. intermissa with respect to A. m. sahariensis/A. m. iberiensis. Relationships among subspecies lineages are similar to those in Fig. 2 (equivalent to the P123 model), but with much lower bootstrap support. However, within the A. m. simensis-inclusive clade, structure and support for the African and Mediterranean lineages collapse in the P12 with respect to the P123 model, and bootstrap support for key branches is < 50%. Especially, MN119925 A. m. unicolor is transposed as a long branch from outside to inside these clades. Of course, not all first-position SNPs are substitutions, nor all third-position SNPs silent. A Maximum Parsimony analysis of inferred amino acid substitutions at 125 sites across representatives of 22 subspecies (Supplementary Fig. S5) gives a clade structure is similar to Fig. 2 here or that from their P12 model (except as to rooting), with bootstrap support for African and Mediterranean lineages again < 50%. Bootstrap support for other clades in Fig. 2 is strong.

As shown by the molecular clock in Fig. 4, pairwise differences among nominal subspecies within the Southeast European, Asia Minor, and Mediterranean clades are less than those between individuals within other subspecies, even where within-clade relationships are well-defined. Where these local subspecies were originally defined by perceived morphological differences, review of their alpha and beta taxonomy may be indicated, with a view towards synonymizing local forms in gamma taxonomy: names matter27.

Besides the 22 subspecies of A. mellifera with complete mitogenomes compared in the main MS, GenBank includes six additional subspecies with partial sequences from the 5′ end of the ND2 region. Their names and provenance are A. m. adami (Crete), A. m. cecropia (Greece), A. m. cypria (Cyprus), A. m. macedonica (Greece into Romania), A. m. pomonella (Kazakhstan), and A. m. sicula (Sicily). Over the first 574 bp common to all these subspecies, in combination with the 22 subspecies’ mitogenomes examined in the main MS, there are 39 variable sites of which 19 are phylogenetically informative sensu Nei34 (Supplementary Fig. S6).

Sequences from subspecies A. m. adami, A. m. cecropia, and A. m. macedonica are identical to those from A. m. ligustica and carnica in the Southeast Europe clade, as is the sequence from A. m. cypria except for a single autapomorphic Y/G SNP. The sequence from A. m. pomonella is identical to that of A. m. meda in the Asia Minor clade. The sequence from A. m. sicula is identical with that of the reference A. m. mellifera, except for a T SNP otherwise found only in the African superclade.

Inclusion of these short sequences tests the phylogeographic hypothesis. The similarity of the Central Asian A. m. pomonella and Middle Eastern A. m. meda extends the Asia Minor clade further eastward, separately from A. m. sinisxinyuan. A. m. sicula is cladisitically distinct from A. m. siciliana in the Mediterranean clade, and the co-occurrence of the two forms on Sicily suggests a secondary colonization from the westerly islands after in initial occupation from the European mainland. Likewise, the continental association of A. m. adami and A. m. cypria from Crete and Cyprus, respectively, is consistent with a dispersal corridor from southeastern Europe through the eastern Mediterranean and into the Levant, and suggests an eastern limit of the re-colonization of the insular Mediterranean at Sicily. The phylogeography proposed here accommodates these data well: further insight may be expected from complete mitogenomes.

The "Thrice Out of Africa" hypothesis9 follows a phylogenetic scheme that retrieved the same four groups (ACMO) as the then current continental geographic scheme based on meristics and morphology8. Their molecular network transposes the alphabetical order to an MAOC backbone, rooted with respect to a composite outgroup within the (A)frican subspecies, which along with Mellifera (M) was separated from the European (Continental) and Asian (Oriental) subspecies. Subsequent analyses of various nucDNA markers and combinations of A. mellifera subspecies broadly identify the MAOC groups9,10,23,24,32,33,35, however placement of the root on various branches according to internal and (or) external criteria varies, so as to suggest alternative geographic origins of A. mellifera. These include alternative interpretations of the Thrice Out of Africa data29 [root indeterminate between M + A and O + C], as well as Thrice Out of Asia9 and North African/Middle Eastern origin models10,32,33,35, for example with A. m. jemenitica in the Y lineage as outgroup to other subspecies of A.

There are fundamental differences between the MAOC backbone and mtDNA genome phylogenies (Supplementary Fig. S7), both with respect to placement of the root and allocation of subspecies to clades. No MAOC model places A. m. mellifera as outgroup to other subspecies: the mtDNA data place A. m. scutellata distal rather than proximal to the root of A. mellifera evolution. Closely related subspecies and geographically contiguous subspecies within the Mediterranean mtDNA clade here, including A. m. iberiensis and A. m. intermissa9,32, and A. m. ruttneri24, are dispersed over the M, A, and C clusters, respectively. The A. m. jemenitica/A. m. lamarckii/A. m. syriaca clade here is also dispersed over the A(Y), A(L), and O clusters. On the other hand, pairings of A. m. ligustica and A. m. carnica (C) and A. m. anatoliaca and A. m. caucasia (O) are consistent between nuc- and mtDNA data. I note that allocation of subspecies among rooted MAOC clusters conforms closely to the original continental model8, which was not cladistic in approach or form. The mtDNA-based model explains a more complex continental distribution, where European, Asian, and African subspecies are of multiple origin.

Classification of nucDNA SNP markers from eight of the subspecies examined here24 inter alia overlays A. m. anatoliaca and A. m. caucasia in the O group, and separates A. m. iberiensis from A. m. mellifera (three clusters) in the M lineage. A. m. ruttneri, the only representative of the African A group, is placed midway between A. m. mellifera and the C group, rather than close to A. m. intermissa and A. m. iberiensis in the Mediterranean clade as here. That is, lineages paired in the two subspecies of the Eurasian clade and variants within the Mediterranean clade here are separated by SNP data in contrast to the clade hierarchy here.

Contrasts between nucDNA, mtDNA, and even morphological phylogenies are not unknown29. Maternally-inherited mtDNA phylogenies have the virtue of tracing maternal evolutionary lineages36, and may thus be particularly reliable for phylogeographic inferences about the origins and spread of "queen"-dispersed eusocial insect. A phylogeographic origin Out of Europe inferred from prior affinity with central Asian cavity-nesting bees seems more parsimonious that disjunct Sub-Saharan or Asian origins.

Rooting of complete mtDNA Coding Region sequences from eight species of Apis (Apini) with that of a bumble bee (Bombus ignitus: Bombini) confirms a Miocene origin of Dwarf Honey Bees as outgroup to the Giant and Cavity-Nesting species groups, the latter including the Western Honey Bee (A. mellifera).

Phylogenetic analysis that includes multiple mtDNA genomes sequences per subspecies indicates that the Western Honey Bee originated in Northern Europe during the late Pleistocene. The phylogeographic distribution of its component subspecies clades indicates that Things Fall Apart with either the Out of Africa or Out of Asia hypotheses, when those are based on single mitogenomes per subspecies.

From a basal origin in Northern Europe, A. mellifera dispersed to Southeast Europe (A. m. ligustica/A. m. carnica/carpatica) and extended to Asia Minor (A. m. meda/A. m. caucasia/A. m. anatoliaca). The European lineage then spread southward via the Levant into the Nile Valley, East Africa and Arabia (A. m. syriaca/A. m. lamarckii/A. m. jemenitica), and thence into sub-Saharan Africa (A. m. simensis/A. m. unicolor; paraphyletic A. m. scutellata & A. m. capensis including A. m. adansonii and A. m. monticola). The Mediterranean lineage was re-established from an African lineage in Iberia and the western Mediterranean islands (A. m. iberiensis/A. m. ruttneri + A. m. siciliana/…), and thence spread back into North Africa (…/A. m. sahariensis + A. m. intermissa). Partial sequences from more easterly insular Mediterranean subspecies suggest affinity with the Mellifera (A. m. sicula, Sicily) or Southeast European lineages (A. m. cyprii, Cyprus; A. m. adami, Crete).

A molecular clock estimates the European origin of A. mellifera ca. 780 Kya, separation of more southerly lineages ca. 720 Kya, separation of African from Levantine/Nilotic/Arabian lineages ca. 660 Kya, spread to Africa ca. 550 Kya, and separation of Mediterranean from Sub-Saharan lineages ca. 250 Kya. Diversities within several subspecies date to > 100 Kya, whereas difference between Asia Minor and insular Mediterranean subspecies date to only a few 10 s of Kya.

Inclusion of multiple sequences referred to A. m. mellifera in GenBank shows that apparent paraphyletic anomalies within and among subspecies are in several cases artefacts of mis-assignment to the type subspecies. Incorrect phylogenetic inferences may also proceed from use of non-representative or faulty sequences.

Morphology- and (or) nucDNA-based biogeographic models are not consistent with this multi-mitogenome per taxon model. Genetic differences between nominal southeast European, Asia Minor, and Mediterranean subspecies are typically much less than those within other subspecies for which multiple individual sequences are available. Re-consideration of the alpha and gamma taxonomy of these taxa is indicated. Names matter.

All mtDNA genome sequences used were obtained from and are available through NCBI GenBank. The accession numbers of the 66 Apis mellifera ssp sequences analyzed in detail are given in Table 1 and Supplementary Figs. S1–S3; 12 additional accessions referred to in Fig. 2 are also given there, and two questionable sequences are given in Fig. 6. Accession numbers of the sequences of nine species of Apis examined and the Bombus ignitus outgroup are given in Fig. 1a.

Lo, N., Gloag, R. S., Anderson, D. L. & Odroyd, B. P. A molecular phylogeny of the genus Apis suggests that the Giant Honey Bee of the Philippines, A. breviligula Maa, and the Plains Honey Bee of southern India, A. indica Fabricius, are valid species. Syst. Entomol. 35, 226–233 (2010).

Article Google Scholar

Engel, M. S. The taxonomy of recent and fossil honey bees (Hymenoptera: Apidae; Apis). J. Hymenoptera Res. 8(2), 165–196 (1999).

MathSciNet Google Scholar

Goor, A. The history of the date through the ages in the holy land. Econ. Bot. 21, 320–334 (1967).

Article Google Scholar

Requier, F. et al. The conservation of native honey bees is crucial. Trends Ecol. Evol. 34, 789–798 (2019).

Article PubMed Google Scholar

Geldmann, J. & González-Varo, J. P. Conserving honey bees does not help wildlife. Science 359, 392–393 (2018).

Article ADS PubMed Google Scholar

Wallberg, A. et al. A worldwide survey of genomes sequence variation provides insight into the evolutionary history of the honeybee Apis mellifera. Nat. Genet. 46, 1081–1088 (2014).

Article CAS PubMed Google Scholar

Collins, A., Rinderer, T., Harbo, J. & Bolten, A. Colony defense by africanized and European honey bees. Science 72, 218–220 (1982).

Google Scholar

Ruttner, F. Biogeography and Taxonomy of Honeybees (Springer, 1988).

Book Google Scholar

Whitfield, C. W. et al. Thrice out of Africa: Ancient and recent expansions of the honey bee, Apis mellifera. Science 314(5799), 642–645 (2006).

Article ADS CAS PubMed Google Scholar

Dogantzis, K. A. et al. Thrice out of Asia and the adaptive radiation of the western honey bee. Sci. Adv. 7, 2151 (2021).

Article ADS Google Scholar

Boardman, L. et al. The mitochondrial genome of the Maltese honey bee, Apis mellifera ruttneri (Insecta: Hymenoptera: Apidae). Mitochondrial DNA B Resour. 5(1), 877–878 (2020).

Article PubMed PubMed Central Google Scholar

Tihelka, E., Cai, C., Pisani, D. & Donoghue, P. C. J. Mitochondrial genomes illuminate the evolutionary history of the Western Honey bee (Apis mellifera). Sci. Rep. 10, 14515 (2020).

Article ADS CAS PubMed PubMed Central Google Scholar

Rortais, A., Arnold, G., Alburaki, M., Legout, H. & Garnery, L. Review of the DraI COI-COII test for the conservation of the black honeybee (Apis mellifera mellifera). Conserv. Genet. Resour. 3, 383–391 (2011).

Article Google Scholar

About-Shaara, H. F., Al-Ghamdi, A. A., Ali Shan, K. & Al-Kahtani, A. N. Genetic network analysis between Apis mellifera subspecies based on mtDNA argues the purity of specimens from North Africa, the Levant and Saudi Arabia. Saudi J. Biol. Sci. 28, 2718–2725 (2021).

Article Google Scholar

Fuller, Z. L. et al. Genome-wide analysis of signatures of selection in populations of African honey bees (Apis mellifera) using new web-based tools. BMC Genom. 16, 518 (2015).

Article Google Scholar

Eimanifar, A. et al. The complete mitochondrial genome of Apis mellifera meda (Insecta: Hymenoptera: Apidae). Mitochondrial DNA B Resour. 2(1), 268–269 (2017).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The complete mitochondrial genome of Apis mellifera unicolor (Insecta: Hymenoptera: Apidae), the Malagasy honey bee. Mitochondrial DNA B Resour. 4(2), 3286–3287 (2019).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The mitochondrial genome of the Carniolan honey bee, Apis mellifera carnica (Insecta: Hymenoptera: Apidae). Mitochondrial DNA B Resour. 4(2), 3288–3290 (2019).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The mitochondrial genome of Apis mellifera simensis (Hymenoptera: Apidae), an Ethiopian honey bee. Mitochondrial DNA B Resour. 5(1), 9–10 (2019).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The complete mitochondrial genome of the West African honey bee Apis mellifera adansonii (Insecta: Hymenoptera: Apidae). Mitochondrial DNA B Resour. 5(1), 11–12 (2019).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The mitochondrial genome of the Spanish honey bee, Apis mellifera iberiensis (Insecta: Hymenoptera: Apidae), from Portugal. Mitochondrial DNA B Resour. 5(1), 17–18 (2019).

Article PubMed PubMed Central Google Scholar

Boardman, L. et al. The complete mitochondrial genome of Apis mellifera jemenitica (Insecta: Hymenoptera: Apidae), the Arabian honey bee. Mitochondrial DNA B Resour. 5(1), 875–876 (2020).

Article PubMed PubMed Central Google Scholar

Chen, C. et al. Genomic analyses reveal demographic history and temperate adaptation of the newly discovered honey bee subspecies Apis mellifera sinisxinyuan n. ssp. Mol. Biol. Evol. 33(5), 1337–1348 (2016).

Article CAS PubMed PubMed Central Google Scholar

Momeni, J. et al. Authoritative subspecies diagnosis tool for European honey bees based on ancestry informative SNPs. BMC Genom. 22, 101 (2021).

Article CAS Google Scholar

Eimanifar, A. et al. The complete mitochondrial genome of the Cape honey bee, Apis mellifera capensis (Insecta: Hymenoptera: Apidae). Mitochondr. DNA B 1, 817–819 (2016).

Article Google Scholar

Eimanifar, A., Kimball, R. T., Braun, E. L. & Ellis, J. D. Mitochondrial genome diversity and population structure of two western honey bee subspecies in the Republic of South Africa. Sci. Rep. 8, 1333 (2018).

Article ADS PubMed PubMed Central Google Scholar

Mayr, E. Principles of Systematic Zoology (McGraw-Hill, 1963).

Google Scholar

Kumar, S., Stecher, G., Li, M., Knyaz, C. & Tamura, K. MEGA X: Molecular evolutionary genetics analysis across computing platforms. Mol. Biol. Evol. 35, 1547–1549 (2018).

Article CAS PubMed PubMed Central Google Scholar

Dong, A., Wang, Y., Li, C., Li, L. & Men, X. Mitochondrial DNA as a molecular marker in insect ecology: Current status and future prospects. Ann. Entomol. Soc. Am. 114, 470–476 (2021).

Article CAS Google Scholar

Aris, M. C. & Shephard, W. S. Phylogenetic relationships of honey bees (Hymenoptera: Apinae: Apini) inferred from nuclear and mitochondrial DNA sequence data. Mol. Phylogenet. Evol. 37, 25–35 (2005).

Article Google Scholar

Papadopoulou, A., Anastasiou, I. & Vogler, A. P. Revisiting the insect mitochondrial molecular clock: The Mid-Aegean trench calibration. Mol. Biol. Evol. 27, 1659–1672 (2010).

Article CAS PubMed Google Scholar

Kotthoff, U., Wappler, T. & Engel, M. S. Greater past disparity and diversity hints at ancient migrations of European honey bee lineages into Africa and Asia. J. Biogeogr. 40, 1832–1838 (2013).

Article Google Scholar

Han, F., Wallberg, A. & Webster, M. T. From where did the Western honeybee (Apis mellifera) originate?. Ecol. Evol. 2, 1949–1957 (2012).

Article PubMed PubMed Central Google Scholar

Nei, M. Molecular Evolutionary Genetics (Columbia University Press, 1975).

Google Scholar

Cridland, J. M., Tsutsui, N. D. & Ramírez, S. R. The complex demographic history and evolutionary origin of the Western Honey Bee, Apis mellifera. Genome Biol. Evol. 9, 457–472 (2017).

Article PubMed PubMed Central Google Scholar

Wilson, A. C. et al. Mitochondrial DNA and two perspectives on evolutionary genetics. Biol. J. Linnean Soc. 26, 375–400 (1985).

Article Google Scholar

Download references

I dedicate this MS to the life and memory of Professor Dave Wake (1936–2021) for his inspired teaching and many contributions to evolutionary biology. I thank Dr Zack Fuller for providing corrected subspecies assignments to GenBank acquisitions cited in Ref.9, Professor Gard Otis for comments on previous drafts of the MS, an anonymous reviewer for major improvement to previous drafts, and my daughter Matilda for instruction in the use of PaintShop layers. The final draft was prepared while in residence at the Atlanta Soto Zen Center: gassho to Michael Elliston Roshi and all dharma friends.

Genetics, Evolution, and Molecular Systematics Laboratory, Department of Biology, Memorial University of Newfoundland, St John's, NL, A1C 5S9, Canada

Steven M. Carr

You can also search for this author in PubMed Google Scholar

S.M.C wrote the manuscript text and prepared all figures, tables, and supplementary materials.

Correspondence to Steven M. Carr.

The author declares no competing interests.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

Carr, S.M. Multiple mitogenomes indicate Things Fall Apart with Out of Africa or Asia hypotheses for the phylogeographic evolution of Honey Bees (Apis mellifera). Sci Rep 13, 9386 (2023). https://doi.org/10.1038/s41598-023-35937-4

Download citation

Received: 28 April 2023

Accepted: 26 May 2023

Published: 09 June 2023

DOI: https://doi.org/10.1038/s41598-023-35937-4

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.