We re-evaluated the proposition based on the present-day distribution of honey bees (genus Apis) that their centre of origin resides in Asia, with subsequent migration and diversification into Europe and Asia. In our research, we considered the so-far largely ignored fossils distributed through a variety of late Palaeogene (Oligocene) and early Neogene (Miocene) deposits, aiming at a synthesis of both present-day and past honey bee distribution.
Northern Hemisphere, Africa, Southeast Asia.
We examined the morphological diversity (also termed ‘disparity’) and affinities of the full living and fossil diversity of honey bees ranging from their earliest origins to the present day. Based on the fossil record and present-day distribution of species, considering continental drift from the Oligocene until today, we have established an evolution and migration scenario for the genus Apis.
The fossil record of Apis demonstrates a diversity that is predominantly European in origin, particularly among the most basal species of the genus. Honey bees exhibited a greater morphological disparity during the Oligocene and, particularly, the Miocene epochs, a time when the principal lineages were established. Contradicting earlier propositions, the geological models allowed a migration via western Europe to north-western Africa during the Miocene, and the fossil record corroborates such a migration.
From the full complement of available data, Apis apparently originated in Europe, spreading from there into Asia, Africa and North America, with subsequent diversification in the former two regions and extinction in the latter. The western honey bee, Apis mellifera, descended from European ancestors that probably migrated from western Europe to Africa during the late Miocene and re-immigrated into Europe during the Holocene and possibly preceding interglacials.
No other insect has such an intimate relationship with Homo sapiens as the honey bee (genus Apis). The industriousness of the hive and the honey these species produce have fascinated and nourished human populations and civilization for millennia, and the honey bee has been intricately woven into most cultures and mythologies. The domestication of hives has been recorded from as long ago as the Minoan and Egyptian civilizations (c. 2500 bc). Naturally, once domesticated, honey bees were actively moved as part of human migrations, and eventually their pollination services became recognized and held in equal importance to the honey harvested as a food and medicine. Today, honey bees, principally the western honey bee, Apis mellifera Linnaeus, 1758, represent a multi-billion dollar agricultural industry. Through the efforts of humans, honey bees have been introduced multiple times into the Americas, and also into Australia, New Zealand, New Caledonia and many areas of Oceania that were not part of their original distribution. In addition, colonies of European and other strains of honey bees are regularly imported throughout the world for managed apiaries, but often such varieties intermix with feral populations rendering the world honey bee fauna quite heterogeneous.
The native (i.e. non-human influenced) distribution and migration of honey bee species and populations has been a matter of serious and continued debate. Traditional studies posited that the centre of origin for Apis was in Asia, the region in which the greatest number of living species occurs (e.g. Ruttner, 1988; Han et al., 2012). From here it was suggested that A. mellifera similarly originated and then migrated, either once or twice, into Europe and Africa. More recent genetic studies implied that A. mellifera originated in tropical Africa and then expanded its distribution into Eurasia at least twice (Whitfield et al., 2006; Han et al., 2012). Examining modern honey bee diversity certainly seems to support such conclusions. Today there are at least seven valid species of Apis, although at least two other Asiatic populations are debated to be sufficiently distinct to warrant species status. All of these species occur as natives in Asia, with the sole exception A. mellifera, which was endemic to western Asia and throughout Africa and Europe. Phylogenetic analyses based on morphology of adults and immatures, as well as molecular data, support the placement of A. mellifera as sister to a clade of Asian species comprising A. cerana Fabricius, 1793, A. nigrocincta F. Smith, 1861, and A. koschevnikovi Enderlein, 1906, and together these form the monophyletic subgenus Apis s. str. The giant honey bees, A. dorsata Fabricius, 1793, are then sister to Apis s. str. as subgenus Megapis. Basal to all of these is the lineage of dwarf honey bees, or the subgenus Micrapis, comprising A. florea Fabricius, 1787 and A. andreniformis F. Smith, 1858.
All of these scenarios, while attempting to address historical biogeographical patterns, fail to take into consideration the significant fossil record available for honey bee species. This is a considerable oversight given that the fossil record demonstrates that past honey bee distributions were significantly different and included regions where they do not occur as natives today, such as North America (Engel et al., 2009). The fossil record, revised extensively over the last 10 years, documents significant diversity of species and morphological varieties, particularly in the late Paleogene and early Neogene of Europe (Engel, 1998; Nel et al., 1999; Ohl & Engel, 2007; Michez et al., 2012). Indeed, phylogenetic investigations into the relationships among living and fossil honey bees indicate that basal species of Apis occur not in Asia, where the centre of modern diversity can be found, but instead within Europe. This fact alone highlights a European centre of origin for Apis. Herein we attempt to address the morphological diversity, also termed ‘disparity’ to distinguish it from taxonomic measures (Foote, 1991), present among living and fossil honey bees, and to bring a truly historical perspective to the question of ancient honey bee migrations and diversification.
Materials and Methods
Fossil material originated from the Staatliches Museum für Naturkunde, Stuttgart, the Heimatmuseum Göppingen Jebenhausen, and the Paläontologisches Museum Nierstein, and was examined with a stereomicroscope. Analysis Pro 3.1 software (Olympus Soft Imaging Solutions GmbH, Münster, Germany) was used for distance and angle measurements. The software past 1.75b (Hammer et al., 2001) was used for cluster analyses. Terminology of forewing veins and cells followed standard apoid nomenclature (Engel, 2001), while landmarks and angles for forewing venational analysis (FWVA) followed established methods used for living and fossil honey bee species (Ruttner, 1988; Rinderer et al., 1989, 1995; Wedmann, 2000; Kotthoff et al., 2011; see Appendix S1 in Supporting Information); these methods represent the principal means of examining population/subspecific variation in honey bees, thereby making our analysis comparable with extensive work on modern Apis populations (e.g. Barour et al., 2011; Miguel et al., 2011). FWVA measurements of all examined specimens with well-preserved forewings were subjected to a cluster analysis together with measurements of representative Recent specimens of honey bees from Germany (A. mellifera) and Asia (A. florea, A. dorsata, A. cerana). Additionally, Miocene and Oligocene specimens of honey bees from deposits that had sufficiently well-preserved forewings were included to permit meaningful measurement and comparison (Cockerell, 1907; Théobald, 1937; Arillo et al., 1996; Nel et al., 1999; Wedmann, 2000; Engel, 2006; Engel et al., 2009). We used FWVA measurements from Eocene Apidae, representing the genera Electrapis, Electrobombus, Succinapis, Thaumastobombus, Melikertes and Pygomelissa, along with other tribes of Recent corbiculate bees, as well as bumble bees (Bombus, a Miocene and a present-day species) and orchid bees (Euglossa and Eufriesea); and with representative non-corbiculate Apinae (Centris, Epicharis, Xeromelecta, Zacosmia) as outgroup taxa because all these taxa provide a sufficiently rich venation to permit meaningful comparison. We also added drones of A. cerana and a queen of A. mellifera to the analysis in order to test whether these are placed on the same branches as related workers to determine whether caste-related wing venation differences may influence the results. For the cluster analysis, we used the paired group algorithm, and the Euclidean distance as similarity measure. Further details of the approach are provided elsewhere (Kotthoff et al., 2011).
Results and Discussion
Morphometry of living and fossil honey bees
In the morphometric analysis all taxa other than Apis were grouped together (Fig. 1), demonstrating the gulf in morphological disparity between honey bees and other Apidae, including the other corbiculate apines. Within the range of Apis variation, the giant honey bees (A. dorsata), dwarf honey bees (A. florea) and medium-sized honey bees (A. mellifera and A. cerana) all formed individual recognizable units. The variation within Apis seems to fall into two major types: a ‘CM morphotype’ (cerana/mellifera-like venation), which also at its extreme encompasses the dwarf honey bees (Micrapis); and a ‘D morphotype’ (dorsata-like venation), which includes the giant honey bees (A. dorsata and A. lithohermaea), as well as a great diversity of venationally similar but smaller bees. In other words, the D form of venation observed in the dendrogram covers bees of the usual cerana/mellifera-like body size as well as the distinctly larger species of Megapis.
Those honey bees from the Oligocene and Miocene are, independent of their geographical origin, distributed widely across the main clusters, although many of the earliest Apis (A. henshawi, subgenus Synapis) did form their own cluster within the broader range of variation allied to A. dorsata, perhaps reflecting the overall plesiomorphic nature of giant honey bee wing venation. Subsets of the various fossil populations exhibited a forewing venation more similar to that of CM bees, while others from the same populations showed wings of a more D-like form. While this was most dramatically shown by the numerous representatives of A. armbrusteri from the Miocene Randeck Maar of south-western Germany (Kotthoff et al., 2011), the same pattern was observed for other deposits such as the honey bees from Montagne d'Andance and Sainte-Reine, both in France. The clusters demonstrate that within a given population the honey bees exhibit a range of venation greater than that observed within a given species today. Honey bees in the Oligocene and Miocene had a wider range of variability in comparison to modern populations, despite the perceived dramatic variability within widespread modern species such as A. mellifera and A. cerana (Ruttner, 1988). This pattern suggests a diminishing disparity through time, with individual species today having tighter variation than their more distant ancestors. Moreover, molecular genetic studies have demonstrated that the rate of evolution in Apis is slow compared with other insects (Honeybee Genome Sequencing Consortium, 2006). This perhaps explains how European populations of Apis maintained such hypervariability within an otherwise single evolutionary species for such a considerable time throughout the late Oligocene and Miocene. Consequently, the various morphotypes observed across these European populations would perhaps all represent a single, widespread species, much like the modern widespread species such as A. mellifera and A. cerana. Accordingly, late Oligocene and early Miocene honey bee species encompassed a hypervariable morphotype which encompassed the full range across the CM and D morphotypes of today, representing an extinct ‘CMD hybrid morphotype’ series of populations.
Honey bee historical biogeography
Based on the available fossil record for Apis, it appears as though the centre of origin for the genus was in Europe and during the earliest Oligocene or perhaps the latest Eocene. This is in stark contrast to the central dogma of apicultural literature, namely that Apis originated and diversified in south-eastern Asia. The earliest records of definitive Apis are found in the Oligocene of Germany and France, and these are also the most basal species for the genus (Engel, 1999, 2001). These honey bees persisted into the Miocene and exhibited a wide range of morphological variation, encompassing and exceeding the range of variability known from modern species. These hypervariable (CMD) honey bees invaded Asia and North America in the Miocene, giving rise to taxa such as A. lithohermaea in the middle Miocene of Japan (among the D morphotype) (Engel, 2006), and various CM morphotype bees in China (12) and North America (Engel et al., 2009). It is likely that once in Asia, diversification and specialization took place subsequently, giving rise to the three living lineages of modern honey bees, particularly given that these together form a monophyletic group relative to the Oligocene and Miocene populations. The dwarf honey bees descended from a CM-like ancestor. The giant honey bees diverged from a D morphotype population relative to their sister lineage, the medium-sized honey bees (Apis s. str.), which also evolved from a probably widespread CM morphotype. Higher wing venational plasticity, as is observed in the primitive CMD honey bees, is also well known among other basal branches of eusocial insects, such as the basal living and fossil termites that also exhibit high variability in wing venation (e.g. Grimaldi et al., 2008).
The biogeographical origin of A. mellifera remains a controversial issue. It has been suggested that it was not possible for Oligocene/Miocene honey bees to disperse to Africa (Ruttner, 1988), leading to the conclusion that the only migration route during this period was via the Middle East and south-western Asia to Southeast Asia. This assertion was certainly correct during the Oligocene (Fig. 2), and Ruttner's view that honey bees travelled to eastern Asia during the late Oligocene/early Miocene is supported by the presence of Miocene Apis in China and Japan (e.g. Hong, 1983; Zhang, 1989, 1990; Engel, 2006). A factor contributing to this migration could have been the relatively cooler conditions in western/central Europe during the Oligocene (e.g. Mosbrugger et al., 2005) and that those early honey bees, such as modern Micrapis and Megapis, may have nested in exposed locations rather than within cavities as do modern populations of Apis s. str. (compare, e.g., Engel, 1998 and Koeniger, 1976 for different views). However, newer climate and palaeogeography reconstructions (Rögl, 1999; Kouwenhoven & van der Zwaan, 2006) imply that during the Miocene a possible migration route was present via the Iberian Peninsula and the Strait of Gibraltar into Africa (Fig. 2). Honey bees would not have had to travel more than a few kilometres over sea to reach Africa during the middle and late Miocene, for example at the onset of the Tortonian, when the European climate became relatively cooler once again (Mosbrugger et al., 2005). The presence of Apis at Bellver de Cerdanya (cf. Fig. 1) and Rubielos de Mora (Arillo et al., 1996; Nel et al., 1999) during the subtropical, dry-seasonal climate of the early/middle Miocene in eastern Spain (Jiménez-Moreno et al., 2007) demonstrates that honey bees thrived in this region at that time. Interestingly, Nel et al. (1999) describe the Apis specimen from Rubielos de Mora as of the CM morphotype. A possible counter-argument for Apis reaching Africa during the Miocene is the presence of the desert-like conditions in North Africa (Ruttner, 1988). However, these conditions probably started with the appearance of the Sahara Desert around 7 Ma (Schuster et al., 2006; Micheels et al., 2009), whereas humid conditions still prevailed in North Africa as late as 9 Ma (Köhler et al., 2010).
Genetic analyses revealed that the split between A. mellifera and the lineage of other cavity-nesting species (the ‘cerana’ species group, or Maa's ‘Sigmatapis’) may have occurred already by 8 Ma (Whitfield et al., 2006; Han et al., 2012), and at a time when CMD morphotypes of average-sized honey bees were recorded from the late Miocene in France (Nel et al., 1999; Kotthoff et al., 2011). Unfortunately, there are no or insufficient Pliocene–Pleistocene records for Apini, with the exception of a single subfossil of A. mellifera in East African copal (Zeuner & Manning, 1976). Such records could definitely help to conclusively rule out any of the major hypotheses for the origin of Apis s. str. (recently reviewed by Han et al., 2012).
The absence of fossil honey bees from the Miocene/Pliocene of Africa was used as an argument by Ruttner (1988) for his assumption that an Apis migration to Africa did not occur. However, caution must be used when relying on absence data as one cannot prove a negative and all it takes is a single record to overturn long-held beliefs, for example the historical assertion of Apis not having once been native to North America (Engel et al., 2009). Indeed, the absence of fossil honey bees from Africa is more conservatively explained by the general scarcity of fossil insects from the African Miocene and Pliocene sediments (Schlüter, 2003). Furthermore, there are no fossil honey bees from the Pliocene of Asia, either, and yet they are well documented to have been present there in the Miocene (Hong, 1983; Zhang, 1989, 1990; Engel, 2006). Absence data for fossil bees is probably unreliable as their fossil record is often scant relative to that of other groups (Engel, 2004, 2011).
All of these data suggest the following scenario for honey bee historical biogeography.
During the Oligocene, honey bees of the CMD morphotypes used the only possible southward migration route from Europe to the south-east and invaded Asia. One early branch of this group in Asia much later evolved an autapomorphic reduction in body size, as well as other unique morphological and biological adaptations (e.g. male metatibial process), giving rise to the lineage that eventually developed into the Micrapis clade (i.e. A. florea and A. andreniformis), either in the latest Oligocene or Miocene.
Asian Miocene Apini such as A. lithohermaea were descendants of bees of the D morphotype. Apis dorsata probably descended from these forms, which is supported by the similarity of wing venation and further morphological characteristics, apomorphically increasing in size some time during the Miocene. In addition, some populations of the CM morphotype in Asia perhaps entered western North America via Miocene connections across Beringia (Hopkins, 1967; Kontrimavichus, 1985), giving rise to A. nearctica (Engel et al., 2009).
The CMD bees remained widespread during the Miocene as evidenced by their abundance and distribution across Europe and even in Asia (Zhang, 1989, 1990). Towards the close of the Miocene, bees of the CM morphotype expanded into Africa, most probably via migration towards the Iberian Peninsula, as indicated by the presence of Miocene CM morphotype bees (Nel et al., 1999). The descendants of the African immigrants eventually evolved into A. mellifera (Whitfield et al., 2006), while Asian descendants of the CM morphotype gave rise to the common ancestor of the ‘cerana’ group of species (i.e. the ancestor of A. cerana, A. koschevnikovi and A. nigrocincta).
During the cooler conditions of the Pliocene and the cold conditions of the Pleistocene in western/central Europe (e.g. Fauquette et al., 2006), Apini were not present in Europe. Apis mellifera probably used Africa as a refugium during these intervals, and re-immigrated into Europe during the Holocene (and perhaps also during earlier interglacials) (Fauquette et al., 2006). During the various Pliocene and Pleistocene climatic changes and alterations in sea levels, population fragmentation in the ‘cerana’ group resulted in the diversity of cerana-like morphologies seen today (e.g. Radloff et al., 2011; Han et al., 2012), as well as its close relatives, A. koschevnikovi and A. nigrocincta.
This pattern of migration and diversification differs greatly from those models established solely on living species and populations alone, highlighting the critical importance of using palaeontological data in any historical biogeographical endeavour, particularly for a lineage as hypervariable and as old as the honey bees.
We are grateful to the personnel of the Landesanstalt für Bienenkunde, Universität Hohenheim; the Fachgruppe Biologie, Universität Bonn; the Institut für Bienenkunde, Universität Frankfurt; and the Zoologisches Museum, Universität Hamburg, as well as to several apiculturists for providing extant honey bee specimens. We are further thankful to Sonja Wedmann for discussions and to the Staatliches Museum für Naturkunde Stuttgart, the Paläontologisches Museum, Nierstein, the Urweltmuseum Hauff, Holzmaden, and the Städtisches Naturkundliches Museum, Göppingen, for access to fossils from the Randeck Maar. We are particularly grateful to three anonymous referees for their helpful comments on an earlier version of the manuscript. Partial support was provided by grants DEB-0542909 and DBI-1057366 from the U.S. National Science Foundation (to M.S.E.).
Ulrich Kotthoff is a lecturer, scientist and curator at the Institute for Geology, Center for Earth System Research and Sustainability at Hamburg University. In addition to his palynology-based works on ecosystem and climate development during the Cenozoic, he is interested in the evolution of Diptera and Hymenoptera.
Torsten Wappler is a senior lecturer at the Steinmann Institute at University of Bonn. He is a palaeoentomologist with interests in host–parasite interactions, the inscriptions of insectivory in the fossil record, and the differential climatic sensitivity/synchroneity of plant and arthropod community responses.
Michael S. Engel is Professor of Ecology & Evolutionary Biology and Senior Curator of Entomology at the University of Kansas, and is a specialist in both palaeoentomology and the systematics and biology of bees.
Author contributions: U.K. had the original idea and collected most of the data; U.K., T.W. and M.S.E. designed and performed the analyses; and U.K. and M.S.E. wrote the paper with substantial contributions by T.W.