Skip to main content
NCBI home page
GigaScience logoLink to GigaScience
. 2023 Mar 27;12:giad019. doi: 10.1093/gigascience/giad019

Honey bee (Apis mellifera) wing images: a tool for identification and conservation

Andrzej Oleksa 1, Eliza Căuia 2, Adrian Siceanu 3, Zlatko Puškadija 4, Marin Kovačić 5, M Alice Pinto 6,7, Pedro João Rodrigues 8,9, Fani Hatjina 10, Leonidas Charistos 11, Maria Bouga 12, Janez Prešern 13, İrfan Kandemir 14, Slađan Rašić 15, Szilvia Kusza 16, Adam Tofilski 17,✉
PMCID: PMC10041535  PMID: 36971293

Abstract

Background

The honey bee (Apis mellifera) is an ecologically and economically important species that provides pollination services to natural and agricultural systems. The biodiversity of the honey bee in parts of its native range is endangered by migratory beekeeping and commercial breeding. In consequence, some honey bee populations that are well adapted to the local environment are threatened with extinction. A crucial step for the protection of honey bee biodiversity is reliable differentiation between native and nonnative bees. One of the methods that can be used for this is the geometric morphometrics of wings. This method is fast, is low cost, and does not require expensive equipment. Therefore, it can be easily used by both scientists and beekeepers. However, wing geometric morphometrics is challenging due to the lack of reference data that can be reliably used for comparisons between different geographic regions.

Findings

Here, we provide an unprecedented collection of 26,481 honey bee wing images representing 1,725 samples from 13 European countries. The wing images are accompanied by the coordinates of 19 landmarks and the geographic coordinates of the sampling locations. We present an R script that describes the workflow for analyzing the data and identifying an unknown sample. We compared the data with available reference samples for lineage and found general agreement with them.

Conclusions

The extensive collection of wing images available on the Zenodo website can be used to identify the geographic origin of unknown samples and therefore assist in the monitoring and conservation of honey bee biodiversity in Europe.

Keywords: honey bee, Apis mellifera, biodiversity, conservation, wing, geometric morphometrics

Data Description

We provide 26,481 forewing images of honey bee workers. They represent 1,725 samples from 13 European countries (Table 1, Fig. 1). The shape of the wings was described using the coordinates for 19 landmarks at wing veins’ intersections (Fig. 2). The whole dataset, including the wing images, landmark coordinates, geographic coordinates of sampling locations, and other data, is available on the Zenodo website [1] under a public domain license.

Table 1:

Sample size of honey bee wings used in this study

Country Country abbreviation Number of wings Number of samples
Austria AT 198 10
Croatia HR 6,103 160
Greece GR 1,444 244
Hungary HU 426 22
Moldova MD 263 10
Montenegro ME 300 20
Poland PL 5,955 253
Portugal PT 960 192
Romania RO 6,498 197
Serbia RS 299 20
Slovenia SI 835 21
Spain ES 2,563 516
Turkey TR 637 60
Open in a new tab

Figure 1:

Figure 1:

Open in a new tab

Locations from which honey bee samples were collected. Jitter was used to show multiple samples from the same or similar location.

Figure 2:

Figure 2:

Open in a new tab

Position of landmarks on the forewing of a honey bee worker. The landmarks are indicated with red dots. The blue circles around them should be tangent to the venation outline in 3 points.

Introduction

Honey bees (Apis mellifera, NCBI:txid7460) are ecologically and economically important. Their value as pollinators of wild plants and crops is much greater than the economical return from honey or other products of the beekeeping industry [2]. At the same time, in the United States [3] and some European countries [4, 5] there has been a decline in the number of managed honey bee colonies. In addition to pathogens, pesticides [6], and socioeconomic factors [7], loss of genetic variability is considered one of the possible causes of such decline [8–10].

The honey bee native distribution covers Europe, up to a latitude of about 60 degrees north, Africa, the Middle East [11], and Central Asia [12, 13]. Within this wide range, the environment varies markedly. Such diverse environmental conditions, as well as the history of range expansion and isolation of populations, have resulted in notable variation of morphologic and behavioral traits, as represented by more than 24 subspecies (geographical races) [11–18]. The subspecies were initially grouped into 4 evolutionary lineages (A, C, M, and O [11]) from morphologic data, but recently, 3 additional lineages (Y, L, and U) were identified from molecular data [19, 20].

The biodiversity of the honey bee is becoming increasingly endangered by the mass introduction of queens produced by breeding. This process began in the 19th century [21] and has intensified recently [22, 23]. The honey bee queens used by many beekeepers are the daughters of relatively few selected breeder queens [24], which are often hybrids [25]. In some European countries, more than 15% of all colonies are re-queened every year [26]. In consequence, some populations, which are well adapted to the local environment [27], are threatened by introgressive hybridization [22]. In this context, it is important to distinguish native honey bees, which occurred in a particular area before the intensification of beekeeping, from nonnative ones, introduced into the area by human intervention.

The conservation of honey bee subspecies requires their identification, which may be based on molecular markers. Recently, identification methods have been developed from single-nucleotide polymorphisms (SNPs) [28–30], which are more accurate than microsatellites [31]. The problem is that, despite the decreasing costs of SNP genotyping [29], molecular identification is still expensive and not easily and quickly accessible to beekeepers. A cheaper alternative is the identification of subspecies based on wing venation measurements [11, 18, 32]. The wings can be measured using various methodologies based on distances and angles [33], landmark coordinates [34–36], outlines [37], or image pixels [38]. While it has been demonstrated that morphologic and molecular markers might provide similar results [39, 40, but see also 41], identification always requires a reference dataset, which is often inadequate.

Public data suitable for the identification of honey bee subspecies or evolutionary lineages using wing morphometry are scarce. Early studies based on multiple measurements of wings provided averages and standard deviations for all measured distances and angles, as well as details related to linear discriminant analysis (LDA) [42, 43]. Unfortunately, in later studies, including those covering global honey bee diversity, averages were not provided [33] or were provided only for a few selected variables [11]. For example, in a study of Apis mellifera mellifera, only 9 out of 36 characteristics were reported [11: Table 13.1]; among them, there are no data for the venation angles used for discrimination of this subspecies [11: p. 229]. More important, the details of LDA were often not presented in those studies, preventing readers from using them to identify unknown samples. Later, the LDA details were provided in a few studies [34, 36, 44–46]. However, despite this progress, it would be even more useful to give readers access to all of the raw data used in the analysis instead of providing LDA details, as has been done in some studies related to wing measurements in Diptera [47–49]. In the case of honey bees, only 1 study is known to us in which the landmark data were made available [50]. Providing landmark coordinates solves the problem of data availability only partially because different studies may use different configurations of landmarks [18]. Not only can the order of landmarks differ between studies, but their positions may also vary. This makes comparisons between studies difficult or even impossible. On the other hand, if a wing image is available, it can be reanalyzed, and missing or incompatible landmarks can be determined. In addition, the wing images can be used to determine the landmarks automatically [35, 51, 52].

Currently, while wing measurements are usually based on images, they are rarely made available after publication. This applies not only to honey bees but also to other insects. Usually, only 1 image is presented with information about the position of the landmarks [47]. There are only a few examples in which wing images have been made publicly available. This is the case of 1 study on Vespidae [53] and 1 study on Drosophila melanogaster, for which a large repository of wing images was recently provided [54]. Honey bee wing images have never been made available in such significant numbers.

The lack of reference data on the morphologic variation of honey bees can be alleviated by data sharing, which is a common practice in some scientific fields, including genomics [55] and neurosciences [56]. It is one of the factors that has facilitated the rapid growth of those fields in recent years. In contrast to molecular biology, data sharing is relatively rare in ecological studies [57]. The benefits of data sharing within the scientific community are well known. However, individual authors often resist making their data available [58–60]. Even if some data are provided, they are often incomplete [61]. Large-scale investigations require large datasets [62], which are difficult for a single researcher to obtain. Individual studies on honey bee biogeography often focus on a relatively small area of 1 or a few countries, and large-scale comparisons are rare [32, 39, 45]. Data sharing would allow the combining of datasets from multiple studies to obtain better knowledge about large-scale geographic variation and the conservation status of honey bee subspecies.

In an attempt to begin building a global reference dataset for the honey bee, here we provide an extensive repository of wing images representative of its diversity in a large tract of Europe. The wing images are accompanied by the coordinates of 19 landmarks, which can easily be used in future comparisons. We also present an R script in which we analyze the coordinates and show how they can be used to identify the origin of an unknown sample. Among other applications, the repository can be used to identify native honey bees, which is essential for their conservation.

Methods

Material

In this study, we used 26,481 forewing images of honey bee workers (nonreproductive diploid females). They represent 1,725 samples from 13 European countries (Table 1, Fig. 1). A sample consists of workers that were collected either from 1 colony or from flowers in 1 location. In the case of samples collected from colonies, the workers were swept with a brush from a middle comb. In the case of samples collected from flowers, the workers were captured using an entomological net. After collection, the workers were stored in alcohol until their preparation. The number of workers per sample ranged from 5 to 20. In some cases (Poland and Hungary), when only 1 or 2 workers were collected from 1 location, neighboring locations were treated together to obtain at least 10 wings per sample. The samples from Austria, Montenegro, and Serbia were obtained from queen breeders who use artificial selection. While breeding lines in Austria were maintained through instrumental insemination of queens, queen bees in Montenegro and Serbia were open-mated. Country names are abbreviated according to ISO 3166–1 (Table 1). The geographic coordinates of the samples, the year of their collection, and other information are provided in CSV files for each country separately.

Some of the samples were analyzed in earlier studies addressing other goals: Croatia and Slovenia [63], Greece [64, 65], Hungary and Poland in part [66, 67], Portugal and Spain [39, 41], Romania [68], Serbia and Montenegro [69], and Turkey [70]. Those studies used various methods, and their results could not be directly compared. Therefore, most of the wings were remeasured for the analyses performed herein. In none of the earlier studies were the wing images made publicly available. The data from Austria and Moldova had never been analyzed or published before. Wing preparation and image acquisition differed between the studies. In most cases, the wings were detached from the bees’ bodies and then mounted between 2 microscopic slides. For image acquisition, different types of cameras combined with stereo microscopes [39, 64, 65, 69, 70] or macro lenses [63, 66–68, 68] were used.

Each wing from the dataset was saved as a separate PNG file. The wing image file name begins with the 2-letter country code (Table 1), followed by a hyphen (identical to the minus sign), a 4-digit sample code, another hyphen, and finally the original file name. The original file names vary as they originate from various studies. They usually consist of strings separated by hyphens. In some cases, the name ends with a letter L or R indicating left or right wing. The samples were independently numbered in each country; for this reason, the unique sample name has to include the 2-letter country code. The wing images were sorted by country and compressed into 13 ZIP files.

In each wing image, the coordinates of 19 landmarks were determined (Fig. 2) [45]. The landmarks are compatible with the “standard honey bee morphometry” approach used in earlier studies [33], and the landmark coordinates can be converted to distances and angles [as in 63]. The landmarks were saved within each wing image file and can be viewed and edited in the IdentiFly software application [45]. The raw coordinates of the landmarks were saved in CSV files for each country separately. The whole dataset, including the 26,481 forewing images, landmark coordinates, geographic coordinates of sampling locations, and other data, is available on the Zenodo website [1] under a public domain license.

Statistical analysis

The statistical analysis was performed in R (v. 4.0.3) [71] using RStudio (v. 2022.12.0, RRID:SCR_000432). All details of the statistical analysis are available at the WorkflowHub website [72]. Landmark coordinates from all wings were superimposed using generalized Procrustes analysis in the geomorph package (v. 4.0.4) [73]. The aligned coordinates were averaged within samples, and the averages were used in the subsequent analysis. Principal component analysis was used to extract the first 2 principal components, which were used to describe how the wing shape varied geographically. The association between the principal components and geographic coordinates (latitude and longitude) was analyzed using generalized additive model (GAM) regression in the mgcv package (v. 1.8–33) [74]. Canonical variate analysis (CVA) and the differences between countries and regions were calculated using the Morpho package (v. 2.9) [75]. The wing shape was clustered using the unweighted pair group method with arithmetic mean (UPGMA) in the phangorn package (v. 2.5.5) [76]. The differences in wing shape between countries or regions were described using Mahalanobis distance. The correlation between the Mahalanobis distances and geographical distances was analyzed using the Mantel test. The coordinates of the unknown samples obtained from Nawrocka et al. [45, 77] were aligned using generalized Procrustes analysis and averaged within samples. Next, the samples were aligned with a consensus of the reference sample using ordinary Procrustes analysis in the shapes package (v. 1.2.6) [78]. Finally, the CVA scores of unknown samples were obtained and compared with the CVA scores of reference samples to calculate their probabilities of belonging to each country or region.

Results

Exploratory data analysis

The wing shape varied significantly according to principal component analysis. In the graph of the first 2 principal components (PCs), which account for 51.2% and 8.6% of the variance, respectively, at least 2 clear clusters of points are visible (Fig. 3). One of the clusters represents the Iberian Peninsula, and the other represents central and southeastern Europe. The second principal component differentiated Greece from the countries of central Europe, particularly Austria. The wing shape variation was strongly correlated with geographic location. Latitude and longitude correlated significantly with both the first and second principal components (GAM regression, PC1: the effective degrees of freedom (EDF) = 26.18, F = 720.7, P < 10−15; PC2: EDF = 27.86, F = 58.47, P < 10−15). The first principal component was much lower in the Iberian Peninsula than in the Balkans (Fig. 4A). Additionally, it decreased in Poland from south to north and in Greece from west to east (Fig. 4A). The second principal component was lowest in southeastern Greece, increasing toward the north and west, with some intricate patterns in Romania and Poland (Fig. 4B).

Figure 3:

Figure 3:

Open in a new tab

The first 2 principal components of wing shape (A). Ellipses indicate 95% confidence regions assuming multivariate t-distribution. The 2 bottom wireframe graphs illustrate change of wing shape along the first (B) and the second (C) principal components. The blue and red lines and dots indicate samples with minimum and maximum values, respectively.

Figure 4:

Figure 4:

Open in a new tab

First (A) and second (B) principal components interpolated over sampling locations using a generalized additive model.

As expected, canonical variate analysis revealed deeper differences between countries and showed a similar pattern to principal component analysis (Fig. 5). The shape of honey bee wings (represented by 34 principal components) differed significantly among countries (multivariate analysis of variance: F = 22.1, P < 10−15). In pairwise comparisons, most countries differed markedly from each other. Only Romania did not differ significantly from Moldova, Serbia from Montenegro, and Slovenia from Croatia and Hungary (Table 2). The largest Mahalanobis distance was found between samples from Portugal and Greece and the smallest between samples from Portugal and Spain. The UPGMA tree shows more details about the similarities between the wings collected from different countries (Fig. 6). Most neighboring countries cluster together: Portugal with Spain, Greece with Turkey, Moldova with Romania, and Slovenia with Croatia. Isolation by distance was confirmed by a significant positive correlation between geographic distances and Mahalanobis distances of wing shape between countries (Mantel test: r = 0.7046, P = 0.0015). The Austrian samples did not fit well into this relationship. Despite their close geographic proximity to Slovenia, Croatia, and Hungary, they had unexpectedly different wing shapes (Fig. 7).

Figure 5:

Figure 5:

Open in a new tab

Discrimination between countries based on the first 2 canonical variates. Ellipses indicate 95% confidence regions assuming multivariate t-distribution.

Table 2:

Differences between countries in wing shape (expressed as Mahalanobis distances, lower triangle) and significance of pairwise comparisons (upper triangle). For country abbreviations, see Table 1.

Country AT ES GR HR HU MD ME PL PT RO RS SI TR
AT — 0.0001 0.0001 0.0001 0.0026 0.0009 0.0002 0.0001 0.0001 0.0001 0.0005 0.0001 0.0001
ES 10.3602 — 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
GR 9.6709 11.3111 — 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
HR 7.4676 11.1342 5.2491 — 0.0250 0.0083 0.0001 0.0001 0.0001 0.0001 0.0001 0.0933 0.0001
HU 6.3694 10.9745 6.6387 2.9550 — 0.0155 0.0003 0.0011 0.0001 0.0025 0.0017 0.0524 0.0002
MD 7.4600 10.5178 6.7506 4.6274 5.1587 — 0.0059 0.0072 0.0001 0.0669 0.0160 0.0404 0.0012
ME 6.9289 10.5056 6.4158 5.2746 5.7927 5.8540 — 0.0001 0.0001 0.0001 0.8670 0.0001 0.0001
PL 6.9864 9.2408 6.2533 3.9108 3.9095 4.6505 6.2075 — 0.0001 0.0001 0.0001 0.0004 0.0001
PT 10.2711 1.8918 11.4833 11.2249 10.9754 10.3770 10.4285 9.1929 — 0.0001 0.0001 0.0001 0.0001
RO 6.3672 10.5980 5.8505 3.7678 3.8011 3.6076 4.8826 4.4288 10.5648 — 0.0003 0.0003 0.0001
RS 7.0790 10.7933 6.1816 4.7181 5.3638 5.2334 1.8720 5.8124 10.7111 4.5827 — 0.0002 0.0001
SI 8.0962 11.4743 5.9944 2.4516 3.5920 4.6343 6.5067 4.1484 11.4397 4.2440 5.8863 — 0.0001
TR 7.7594 10.2036 4.9768 4.3165 5.2857 5.7750 6.0269 5.1654 10.3717 4.3490 6.0440 5.1704 —
Open in a new tab

Figure 6:

Figure 6:

Open in a new tab

UPGMA tree illustrating similarities between the shape of the wings collected from different countries.

Figure 7:

Figure 7:

Open in a new tab

Relationship between geographical distance and Mahalanobis distance among countries.

Classification of samples as lineages

When the samples were classified as lineages (Supplementary Figs. S1, S2), using the data from Nawrocka et al. [45], many of them (n = 844, 48.9%) were classified as lineage C, which occurred in all samples from 6 countries: Austria, Croatia, Hungary, Montenegro, Serbia, and Slovenia. Lineage C was also the most frequent one in Greece, Moldova, Poland, Romania, and Turkey (Supplementary Fig. S2). As expected, the samples most similar to lineage C occurred in southeastern Europe, except in southeastern Greece (Fig. 8B).

Figure 8:

Figure 8:

Open in a new tab

Mahalanobis distance to lineages A (A), C (B), M (C), and O (D) interpolated over sampling locations using a generalized additive model.

The second most frequent was lineage M. It occurred in 652 samples (37.8%). It was dominant in Portugal and Spain but also occurred in Poland (Supplementary Fig. S2). A clear similarity to lineage M was observed in the Iberian Peninsula. Moreover, similarity to this lineage increased in Poland from south to north (Fig. 8C).

Unexpectedly, a relatively large fraction of the samples (n = 179, 10.4%) was classified as lineage A. It was detected mainly in Poland but also in the Iberian Peninsula, Greece, Turkey, and Moldova at lower proportions (Supplementary Fig. S2). Similarity to lineage A increased in Poland from south to north and in Greece from northwest to southeast (Fig. 8A).

Lineage O was the least frequent (n = 50, 2.9%). It occurred in Greece and sporadically in Moldova, Poland, Romania, and the European part of Turkey (Supplementary Fig. S2). Similarity to lineage O increased in Greece from the northwest to the southeast (Fig. 8D).

Identification of unknown samples

The data provided here can be used for the identification of an unknown sample of honey bee workers. To test the accuracy of such identification, the leave-one-out cross-validation approach was used. This was based on samples and not on single wing measurements, in order to increase accuracy (for more information, see the Discussion). One sample was temporally removed from the dataset, which was subject to canonical variate analysis, and the obtained data were used to classify the newly removed sample. In this procedure, the removed sample was treated as unknown and the remaining data as reference. The cross-validation was repeated for all samples, and the percentage of correctly classified samples was calculated. When the samples were classified (with cross-validation) according to their country of origin, 86.26% of them were assigned to the correct group. Misclassifications most often occurred between neighboring countries. For example, the correct classification rate for Portugal was only 79.69%, with all the cases of misclassification occurring with the neighboring Spain. Many misclassifications can be attributed to the small sample size for some countries; therefore, the second classification was based on regions. Samples from some smaller countries were combined with those from their large neighbors (Portugal with Spain, Moldova with Romania, and Slovenia with Croatia), and other countries with sample sizes below 25 were excluded (Austria, Hungary, Montenegro, and Serbia). In the case of classification to regions, the correct classification rate (with cross-validation) increased to 98.31%.

Additionally, we classified to regions an independent dataset of historical data from Nawrocka et al. [45, 77]. In the analysis, we used a subset consisting of 53 colonies originating from Europe (lineages C and M). Within this subset, there was 18 colonies that originated from countries represented in the regions’ reference data. Among them, a single sample from Croatia was correctly classified as belonging to region HR-SI; out of 11 samples from Greece, 3 were correctly classified, and 8 were classified as belonging to the neighboring Turkey and the region HR-SI; 2 samples from Romania were incorrectly classified as belonging to region HR-SI; 1 sample from Slovenia was correctly classified as belonging to region HR-SI, and the other was incorrectly classified as belonging to Poland; and finally, 2 samples from Spain were correctly classified.

The reference samples provided here cover only part of Europe; therefore, identification of samples from other parts of Europe and the world can give unexpected results. Often, such samples can be detected as outliers, which have low identification probabilities for all groups. Here, we used an arbitrary threshold value of 0.001; if the maximum probability of identification for a sample is lower, we assume it is an outlier. Unfortunately, many of the identified samples, which originated from countries not covered by reference data, were classified as coming from one of the regions with a relatively high probability. For example, most samples from Italy (representing Apis mellifera ligustica) were classified as belonging to either Poland or HR-SI with a probability above 0.001.

Discussion

Comparison with earlier studies

The data presented here show that the geographic variation of honey bee wing shape in Europe is still large. This variation is most likely an effect of natural selection and not of beekeepers’ mass introduction of nonnative bees. When compared with historical reference samples from the Morphometric Bee Data Bank in Oberursel [11], the bees analyzed here fit well into the pattern, which is believed to have been shaped by natural processes. In particular, there is a high similarity to lineage M in the Iberian Peninsula and in the north of Poland, and there is a high similarity to lineage C in most of central and southeastern Europe, as expected from the seminal work on honey bee taxonomy and biogeography of Ruttner [11]. In the eastern part of the Aegean Sea, there is a high similarity to lineage O, which occurred naturally in Turkey and the Middle East. In general, the distribution of the lineages presented in this study is in line with earlier studies based on morphometry [33], mitochondrial DNA [79, 80], microsatellites [80–82], and SNPs [30, 83]. On the other hand, there are some discrepancies, which are discussed below.

It can be expected that the introduction of nonnative bees will reduce geographic variation because beekeepers prefer certain honey bee subspecies (Apis mellifera carnica, A. m. ligustica, Apis melliferacaucasia) or their hybrids [22]. Beekeepers’ preference for a limited number of breeding lines [24] may lead to a significant homogenization of the population structure across Europe. It is worth noting, however, that despite the increased beekeeper-mediated gene flow in the past decades [22], our results indicate that native honey bee diversity is still not lost, at least in eastern and southern Europe. The population size of honey bees is very large (approximately 19 million colonies) [84]; hence, it may take many generations to change the genetic structure and thus the phenotype of the European population. Free trade in breeding material has intensified only in the past several decades; hence, homogenization of population structure may occur in the future, unless remedial measures are taken to protect local genetic variability.

Some of the results from our study do not agree with the patterns reported by Ruttner [11]. Many colonies in northern Poland and some other countries were classified as belonging to lineage A. This lineage is expected on the African continent and does not occur naturally in central and northern Europe [11]. One possible explanation for the presence of this lineage in Europe is human-induced introgression. A recent mitochondrial DNA survey in central Europe detected haplotypes of African ancestry, although with a frequency of only 1.64% [66], which is much lower than that reported here for the forewing samples from Poland classified according to Nawrocka et al. [45] as lineage A (38.3%). The high proportion of samples assigned to lineage A may also be related to the fact that hybrids between lineages are more likely to be classified as lineage A. Those hybrids have an intermediate phenotype [85], similar to the mean shape of all lineages.

While hybridization may be caused by the introduction of nonnative bees by beekeepers, hybrids between lineages can also occur naturally. Aside from the Alps, there is no physical barrier separating lineages M and C. In such a situation, a wide hybrid zone can be expected. Earlier studies reported that in Poland, there is a wide transition zone with a clinal change in both morphologic [86, 87] and molecular [66] markers. This spatial pattern was most likely due to a natural phenomenon because it was already present in the 1960s [86], when the importation and rearing of nonnative bees were less common.

Hybrids between lineages can be identified to some degree using wing measurements [41, 85]. However, this requires adequate reference samples that are not currently available. The reference sample from the Morphometric Bee Data Bank [11] for lineage M consists of only 16 colonies, whereas those for lineages A, C, and O are larger, consisting of 85, 37, and 49 colonies, respectively [45]. The M reference sample is clearly small, especially when considering the very large native distribution of this lineage, which extends from Iberia to western China [12]. Thus, a large portion of M lineage variation is inevitably underrepresented in the reference dataset. The intriguing detection of lineage A in Europe needs further investigation to determine whether hybrids between lineages C and M are being incorrectly classified as lineage A or the African wing phenotypes are present in Europe more often than was previously expected.

Identification of unknown samples

When using the data provided here for the identification of an unknown sample, it is important to understand the limitations of the described methodology. If the unknown samples originate from one of the regions covered in the reference dataset, the results should be relatively accurate (i.e., the fraction of samples assigned to the correct region should be 98%). However, if the unknown samples originate from a region not covered in the reference dataset (e.g., Italy), they might be incorrectly classified as belonging to one of the regions or countries included in the identification model. In order to detect such misidentifications, the user should examine the probability of identification. If the probability is low, the unknown samples can be classified as outliers, as they will not match any of the geographical regions covered in the reference data. The threshold probability below which the samples are classified as outliers is arbitrary (here, we established a value of 0.001). Unfortunately, detection of outliers failed in many cases (e.g., in the case of samples from Italy). The lower identification rate of historic samples can be related to introgressive hybridization, which occurred in recent years in some parts of Europe [68]. The problem of false positives can be alleviated to some degree by adding contemporary samples from a wider geographic range to the identification model. The geographical coverage of Europe presented in this study is far from complete, as there are many countries without any data or with incomplete data. Among the countries included in this study, a better geographical coverage is warranted for Austria, Serbia, and Montenegro. The Austrian samples, which were obtained from a queen breeder, do not agree with the samples collected from the neighboring countries. This is particularly evident in the isolation-by-distance plot, where they are clear outliers (Fig. 7). This discrepancy may be related to artificial selection and drift, with the colonies being kept in genetic isolation from the surrounding population as a result of instrumental insemination. Ideally, reference samples should be collected from colonies in which queens are not sourced from breeding programs and that are widespread in the study area to ensure coverage of the genetic variation in the population. Another possibility is collecting bees from flowers, in which case they represent multiple nearby colonies.

Despite the problems related to introgression and the presence of false positives that were mentioned above, the wing data presented in this study can be used as a reference for future studies aimed at the identification and monitoring of nonnative honey bees. The wing shape of an unknown sample can be compared with a reference sample from a particular country or region. The identification results can be sorted by similarity to the reference, and colonies with the smallest similarity can be classified as nonnative. These outlier colonies can be re-queened or removed from the population. In this context, the presence of false positives is not a big problem because this procedure is focused on the detection of true negatives. In fact, beekeepers should re-queen only a fraction of their colonies, and the procedure developed here is focused on the detection of the most extreme outliers. Moreover, it is not essential that the reference sample perfectly represents the native phenotype for a particular geographical region. The full range of original variation, present in Europe before large-scale movements of nonnative honey bees, may be irreversibly lost. Fortunately, there is still considerable variation, and this variation deserves to be protected from further genetic erosion.

A large sample size of reference data should not be confused with the sample size of an unknown sample to be classified. In the latter case, one sample consisting of 10 to 20 workers from 1 colony or location should be sufficient. The identification may be based on only a single wing. However, in such cases, the results are inaccurate [36, 88]. By averaging multiple wings within a sample, the measurement error and influence of environmental (and not genetic) factors are minimized.

The conservation of honey bee biodiversity is often focused on the protection of certain subspecies. For example, there are efforts to preserve A. m. mellifera [89]. These efforts are justified because in some parts of western and northern Europe, native honey bees are threatened with extinction due to introgressive hybridization with nonnative bees [90, 91]. However, this approach can overlook some intraspecific variation. The subspecies concept oversimplifies the problem and attempts to classify a continuous variation into a categorical one. For example, different populations of A. m. mellifera within its wide range differ from each other much more than some subspecies do (e.g., A. m. carnica and A. m. ligustica) [30]. Relatively large intra-subspecific variation has been observed in A. m. carnica, Apis mellifera macedonica [92], Apis mellifera iberiensis [79, 83, 93], and A. m. mellifera [94]. Also, in Africa, there is clinal variation, which makes the discrimination of subspecies difficult [95, 96]. Continuous spatial variation is present even on the American continent, where the honey bee has been introduced [97]. In this study, we observed variation not only between countries but also within them. In general, there is isolation by distance, with smaller differences between bees from neighboring regions than between bees that are far apart [80, 94]. We do not advocate abandoning the well-established concept of subspecies but stress the importance of continuous variation, especially in the protection of honey bee diversity.

Conclusion

The problem of a lack of reference samples can be solved by data sharing. Here, we provide for the first time a large collection of honey bee forewing images, accompanied by geographic coordinates as well as measurements and some additional data [1]. The collection and metadata are easy to review, reuse, and update. Moreover, we show how the data can be analyzed, for example, to predict the origin of an unknown sample. The dataset can be used, among other things, as a reference for future studies on biogeography and conservation of honey bees. We hope that future studies will also make wing images freely available in order to expand the dataset and improve our knowledge of honey bee geographic variation.

Supplementary Material

giad019_GIGA-D-22-00297_Original_Submission
giad019_giga-d-22-00297_original_submission.pdf (2.8MB, pdf)
giad019_GIGA-D-22-00297_Revision_1
giad019_giga-d-22-00297_revision_1.pdf (3.6MB, pdf)
giad019_Response_to_Reviewer_Comments_Original_Submission
giad019_response_to_reviewer_comments_original_submission.pdf (67.9KB, pdf)
giad019_Reviewer_1_Report_Original_Submission

Chris Armit -- 11/17/2022 Reviewed

giad019_reviewer_1_report_original_submission.pdf (125.4KB, pdf)
giad019_Reviewer_1_Report_Revision_1

Chris Armit -- 2/10/2023 Reviewed

giad019_reviewer_1_report_revision_1.pdf (110.2KB, pdf)
giad019_Reviewer_2_Report_Original_Submission

Alexis Matamoro Vidal -- 11/29/2022 Reviewed

giad019_reviewer_2_report_original_submission.pdf (111KB, pdf)
giad019_Reviewer_2_Report_Revision_1

Alexis Matamoro Vidal -- 2/10/2023 Reviewed

giad019_reviewer_2_report_revision_1.pdf (109.5KB, pdf)
giad019_Supplemental_Figures
giad019_supplemental_figures.zip (280.5KB, zip)

Acknowledgement

Some samples from Poland were obtained within the framework of a project funded by the State Forests, “Pszczoły wracają do lasu” (Bees Return to the Forest), coordinated by Dr. Kazimierz Szabla. Samples from Croatia were obtained during the project “Biodiversity of the Honey Bee (Apis mellifera carnica) Population in the Republic of Croatia (BioBeeCro),” funded by the Paying Agency for Agriculture, Fisheries and Rural Development. Samples from Portugal were obtained in the framework of the research project PTDC/BIA-BEC/099640/2008, funded by FCT and COMPETE/QREN/EU. Samples from Moldova were obtained thanks to Dr. Valentina Cebotari from the Institute of Zoology, Academy of Sciences of Moldova, during Smartbees project (grant 613960 funded by Program EC FP7-KBBE). Bernadeta Rzeźnicka assisted with the wing measurements.

Contributor Information

Andrzej Oleksa, Department of Genetics, Faculty of Biological Sciences, Kazimierz Wielki University, Bydgoszcz 85-090, Poland.

Eliza Căuia, Honeybee Genetics and Breeding Laboratory, Institute for Beekeeping Research and Development, Bucharest 013975, Romania.

Adrian Siceanu, Honeybee Genetics and Breeding Laboratory, Institute for Beekeeping Research and Development, Bucharest 013975, Romania.

Zlatko Puškadija, Faculty of Agrobiotechnical Sciences, Josip Juraj Strossmayer University of Osijek, Osijek 31000, Croatia.

Marin Kovačić, Faculty of Agrobiotechnical Sciences, Josip Juraj Strossmayer University of Osijek, Osijek 31000, Croatia.

M Alice Pinto, Centro de Investigação de Montanha, Instituto Politécnico de Bragança, Campus de Santa Apolónia, Bragança 5300-253, Portugal; Laboratório Associado para a Sustentabilidade e Tecnologia em Regiões de Montanha (SusTEC), Instituto Politécnico de Bragança, Campus de Santa Apolónia, Bragança 5300-253, Portugal.

Pedro João Rodrigues, Centre in Digitalization and Intelligent Robotics, Instituto Politécnico de Bragança, Campus de Santa Apolónia, Bragança 5300-253, Portugal; Laboratório Associado para a Sustentabilidade e Tecnologia em Regiões de Montanha (SusTEC), Instituto Politécnico de Bragança, Campus de Santa Apolónia, Bragança 5300-253, Portugal.

Fani Hatjina, Department of Apiculture, Institute of Animal Science–Ellinikos Georgikos Organismos ‘DIMITRA’, Nea Moudania 63200, Greece.

Leonidas Charistos, Department of Apiculture, Institute of Animal Science–Ellinikos Georgikos Organismos ‘DIMITRA’, Nea Moudania 63200, Greece.

Maria Bouga, Lab of Agricultural Zoology and Entomology, Agricultural University of Athens, Athens 11855, Greece.

Janez Prešern, Agricultural Institute of Slovenia, Ljubljana SI-1000, Slovenia.

İrfan Kandemir, Ankara University, Department of Biology, Faculty of Science, Ankara University, Beşevler-Ankara 06100, Turkey.

Slađan Rašić, Faculty of Ecological Agriculture, EDUCONS University, Sremska Kamenica 21208, Serbia.

Szilvia Kusza, Centre for Agricultural Genomics and Biotechnology, University of Debrecen, Debrecen 4032, Hungary.

Adam Tofilski, Department of Zoology and Animal Welfare, University of Agriculture in Krakow, Krakow 31-425, Poland.

Data Availability

The whole dataset, including the wing images, landmark coordinates, geographic coordinates of sampling locations, and other data, is available on the Zenodo website [1] under a public domain license. All details of the statistical analysis, including the identification of an unknown sample, are available at the WorkflowHub website [72]. All supporting data and materials are available in the GigaScience GigaDB database [98].

Availability of Source Code and Requirements

Additional Files

Supplementary Fig. S1. Projection of the samples from different countries to canonical variate space of honey bee lineages published in Nawrocka et al. [45]. Ellipses indicate 95% confidence regions assuming multivariate t-distribution. Black ellipses indicated lineages; other ellipses indicate countries.

Supplementary Fig. S2. Samples from different countries classified as honey bee lineages. Jitter was used to show multiple samples from the same or similar location.

Abbreviations

CVA: canonical variates analysis; GAM: generalized additive model; LDA: linear discriminant analysis; SNP: single-nucleotide polymorphism; UPGMA: unweighted pair group method with arithmetic mean.

Competing Interests

The authors declare that they have no competing interests.

Funding

This research was funded in part by the National Science Centre, Poland, grant numbers 2021/41/B/NZ9/03153 and 2015/19/B/NZ9/03718, and by the National Centre for Research and Development, grant number TANGO-V-A/0042/2021. M. Alice Pinto acknowledges Fundação para a Ciência e a Tecnologia for providing financial support from national funds (FCT/MCTES) to CIMO (UIDB/00690/2020 and UIDP/00690/2020) and SusTEC (LA/P/0007/2021). J. Prešern acknowledges the Slovenian Research Agency for providing funding from ARRS Research Program P4-0431 NextGenAgri.

Authors’ Contributions

A.T. and A.O. planned this study. All authors provided honey bee wing images and metadata from their countries. A.T. wrote the first version of the manuscript, and all authors revised and contributed to its final version. All authors have read and approved the final manuscript.

References

  • 1. Oleksa  A, Căuia  E, Siceanu  A  et al.  Collection of wing images for conservation of honey bees (Apis mellifera) biodiversity in Europe. Zenodo. 2022. 10.5281/zenodo.7244070. [DOI] [Google Scholar]
  • 2. Gallai  N, Salles  J-M, Settele  J  et al.  Economic valuation of the vulnerability of world agriculture confronted with pollinator decline. Ecol Econ. 2009;68:810–21. [Google Scholar]
  • 3. Ellis  JD, Evans  JD, Pettis  J. Colony losses, managed colony population decline, and colony collapse disorder in the United States. J Apic Res. 2010;49:134–6. [Google Scholar]
  • 4. Gray  A, Brodschneider  R, Adjlane  N  et al.  Loss rates of honey bee colonies during winter 2017/18 in 36 countries participating in the COLOSS survey, including effects of forage sources. J Apic Res. 2019;58:479–85. [Google Scholar]
  • 5. Potts  SG, Roberts  SPM, Dean  R  et al.  Declines of managed honey bees and beekeepers in Europe. J Apic Res. 2010;49:15–22. [Google Scholar]
  • 6. Steinhauer  N, Kulhanek  K, Antúnez  K  et al.  Drivers of colony losses. Curr Opin Insect Sci. 2018;26:142–8. [DOI] [PubMed] [Google Scholar]
  • 7. Moritz  RFA, Erler  S. Lost colonies found in a data mine: global honey trade but not pests or pesticides as a major cause of regional honeybee colony declines. Agriculture Ecosystems Environ. 2016;216:44–50. [Google Scholar]
  • 8. Panziera  D, Requier  F, Chantawannakul  P  et al.  The diversity decline in wild and managed honey bee populations urges for an integrated conservation approach. Front Ecol Evol. 2022;10:767950. [Google Scholar]
  • 9. Themudo  GE, Rey-Iglesia  A, Robles Tascón  L, et al.  Declining genetic diversity of European honeybees along the twentieth century. Sci Rep. 2020;10:10520. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. vanEngelsdorp  D, Meixner  MD. A historical review of managed honey bee populations in Europe and the United States and the factors that may affect them. J Invertebr Pathol. 2010;103:S80–95. [DOI] [PubMed] [Google Scholar]
  • 11. Ruttner  F. Biogeography and taxonomy of honeybees. Berlin: Springer, 1988.; [Google Scholar]
  • 12. Chen  C, Liu  Z, Pan  Q, 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. 2016;33:1337–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Sheppard  WS, Meixner  M. Apis mellifera pomonella, a new honey bee subspecies from Central Asia. Apidologie. 2003;34:367–75. [Google Scholar]
  • 14. Bouga  M, Alaux  C, Bienkowska  M  et al.  A review of methods for discrimination of honey bee populations as applied to European beekeeping. J Apic Res. 2011;50:51–84. [Google Scholar]
  • 15. Engel  MS. The taxonomy of recent and fossil honey bees (Hymenoptera : apidae ; Apis). J Hymenoptera Res. 1999;8:165–96. [Google Scholar]
  • 16. Meixner  MD, Costa  C, Kryger  P, et al.  Conserving diversity and vitality for honey bee breeding. J Apic Res. 2010;49:85–92. [Google Scholar]
  • 17. Meixner  MD, Leta  MA, Koeniger  N, et al.  The honey bees of Ethiopia represent a new subspecies of Apis mellifera—Apis mellifera simensis n. Ssp. Apidologie. 2011;42:425–37. [Google Scholar]
  • 18. Meixner  MD, Pinto  MA, Bouga  M  et al.  Standard methods for characterising subspecies and ecotypes of Apis mellifera. J Apic Res. 2013;52:1–28. [Google Scholar]
  • 19. Dogantzis  KA, Tiwari  T, Conflitti  IM, et al.  Thrice out of Asia and the adaptive radiation of the western honey bee. Sci Adv. 2021;7:2151. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Franck  P, Garnery  L, Loiseau  A  et al.  Genetic diversity of the honeybee in Africa: microsatellite and mitochondrial data. Heredity. 2001;86:420–30. [DOI] [PubMed] [Google Scholar]
  • 21. Crane  E. The world history of beekeeping and honey hunting. London: Routledge, 1999. [Google Scholar]
  • 22. De la Rúa  P, Jaffé  R, Dall'Olio  R  et al.  Biodiversity, conservation and current threats to European honeybees. Apidologie. 2009;40:263–84. [Google Scholar]
  • 23. Nielsdatter  MG, Larsen  M, Nielsen  LG, et al.  History of the displacement of the European dark bee (Apis mellifera mellifera) in Denmark. J Apic Res. 2021;60:13–18. [Google Scholar]
  • 24. Schiff  NM, Sheppard  WS. Genetic analysis of commercial honey bees (Hymenoptera: apidae) from the southern United States. J Econ Entomol. 1995;88:1216–20. [Google Scholar]
  • 25. Lodesani  M, Costa  C. Bee breeding and genetics in Europe. Bee World. 2003;84:69–85. [Google Scholar]
  • 26. Chauzat  M-P, Cauquil  L, Roy  L  et al.  Demographics of the European apicultural industry. PLoS One. 2013;8:e79018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Büchler  R, Costa  C, Hatjina  F, et al.  The influence of genetic origin and its interaction with environmental effects on the survival of Apis mellifera L. colonies in Europe. J Apic Res. 2014;53:205–14. [Google Scholar]
  • 28. Chen  C, Parejo  M, Momeni  J  et al.  Population structure and diversity in European honey bees (Apis mellifera L.)—an empirical comparison of pool and individual whole-genome sequencing. Genes. 2022;13:182. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29. Henriques  D, Browne  KA, Barnett  MW  et al.  High sample throughput genotyping for estimating C-lineage introgression in the dark honeybee: an accurate and cost-effective SNP-based tool. Sci Rep. 2018;8:8552. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Momeni  J, Parejo  M, Nielsen  RO  et al.  Authoritative subspecies diagnosis tool for European honey bees based on ancestry informative SNPs. Bmc Genomics. 2021;22:101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Muñoz  I, Henriques  D, Jara  L, et al.  SNPs selected by information content outperform randomly selected microsatellite loci for delineating genetic identification and introgression in the endangered dark European honeybee (Apis mellifera mellifera). Mol Ecol Resour. 2017;17:783–95. [DOI] [PubMed] [Google Scholar]
  • 32. Kandemir  İ, Özkan  A, Fuchs  S. Reevaluation of honeybee (Apis mellifera) microtaxonomy: a geometric morphometric approach. Apidologie. 2011;42:618–27. [Google Scholar]
  • 33. Ruttner  F, Tassencourt  L, Louveaux  J. Biometrical-statistical analysis of the geographic variability of Apis mellifera L. Apidologie. 1978;9:363–81. [Google Scholar]
  • 34. Francoy  TM, Wittmann  D, Drauschke  M, et al.  Identification of Africanized honey bees through wing morphometrics: two fast and efficient procedures. Apidologie. 2008;39:488–94. [Google Scholar]
  • 35. Rodrigues  PJ, Gomes  W, Pinto  MA. DeepWings©: automatic wing geometric morphometrics classification of honey bee (Apis mellifera) subspecies using deep learning for detecting landmarks. BDCC. 2022;6:70. [Google Scholar]
  • 36. Tofilski  A. Using geometric morphometrics and standard morphometry to discriminate three honeybee subspecies. Apidologie. 2008;39:558–63. [Google Scholar]
  • 37. Francoy  TM, de Faria Franco  F, Roubik  DW. Integrated landmark and outline-based morphometric methods efficiently distinguish species of Euglossa (Hymenoptera, Apidae, Euglossini). Apidologie. 2012;43:609–17. [Google Scholar]
  • 38. De Nart  D, Costa  C, Di Prisco  G, et al.  Image recognition using convolutional neural networks for classification of honey bee subspecies. Apidologie. 2022;53:5. [Google Scholar]
  • 39. Henriques  D, Chávez-Galarza  J, Teixeira  J  et al.  Wing geometric morphometrics of workers and drones and single nucleotide polymorphisms provide similar genetic structure in the Iberian honey bee (Apis mellifera iberiensis). Insects. 2020;11:89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Oleksa  A, Tofilski  A. Wing geometric morphometrics and microsatellite analysis provide similar discrimination of honey bee subspecies. Apidologie. 2015;46:49–60. [Google Scholar]
  • 41. García  CAY, Rodrigues  PJ, Tofilski  A, et al.  Using the software DeepWings© to classify honey bees across Europe through wing geometric morphometrics. Insects. 2022;13:1132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. DuPraw  EJ. The recognition and handling of honeybee specimens in non-Linnean taxonomy. J Apic Res. 1965;4:71–84. [Google Scholar]
  • 43. DuPraw  EJ. Non-Linnean taxonomy and the systematics of honeybees. Syst Zool. 1965;14:1–24. [Google Scholar]
  • 44. Bustamante  T, Fuchs  S, Grünewald  B  et al.  A geometric morphometric method and web application for identifying honey bee species (Apis spp.) using only forewings. Apidologie. 2021;52:697–706. [Google Scholar]
  • 45. Nawrocka  A, Kandemir  İ, Fuchs  S, et al.  Computer software for identification of honey bee subspecies and evolutionary lineages. Apidologie. 2018;49:172–84. [Google Scholar]
  • 46. Rinderer  TE, Buco  SM, Rubink  WL  et al.  Morphometric identification of Africanized and European honey bees using large reference populations. Apidologie. 1993;24:569–85. [Google Scholar]
  • 47. López-García  J, Angell  C, Martín-Vega  D. Wing morphometrics for the identification of Nearctic and palaearctic Piophilidae (Diptera) of forensic relevance. Forensic Sci Int. 2020;309:110192. [DOI] [PubMed] [Google Scholar]
  • 48. Szpila  K, Żmuda  A, Akbarzadeh  K  et al.  Wing measurement can be used to identify European blow flies (Diptera: calliphoridae) of forensic importance. Forensic Sci Int. 2019;296:1–8. [DOI] [PubMed] [Google Scholar]
  • 49. Szpila  K, Johnston  NP, Akbarzadeh  K  et al.  Wing measurements are a possible tool for the identification of European forensically important Sarcophagidae. Forensic Sci Int. 2022;340:111451. [DOI] [PubMed] [Google Scholar]
  • 50. Węgrzynowicz  P, Łoś  A. Dataset of wing venation measurements for Apis mellifera caucasica, A. Mellifera carnica and A. Mellifera mellifera (Hymenoptera: apidae), their hybrids and backcrosses. BDJ. 2020;8:e53724. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51. Ángel-Beamonte  E, Martín-Ramos  P, Santolaria  P, et al.  Automatic determination of landmark coordinates for honey bee forewing venation using a new MATLAB-based tool. J Apic Res. 2018;57:605–10. [Google Scholar]
  • 52. Batra  SWT. Automatic image analysis for rapid identification of Africanized honey bees. In: Needham  GR, editor. Africanized honey bees and bee mites. Chichester: Ellis Horwood, 1988, p. 260–3. [Google Scholar]
  • 53. Baracchi  D, Dapporto  L, Turillazzi  S. Relevance of wing morphology in distinguishing and classifying genera and species of Stenogastrinae wasps. CTOZ. 2011;80:191–9. [Google Scholar]
  • 54. Sonnenschein  A, VanderZee  D, Pitchers  WR  et al.  An image database of Drosophila melanogaster wings for phenomic and biometric analysis. Gigascience. 2015;4:25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55. Kaye  J, Heeney  C, Hawkins  N, et al.  Data sharing in genomics—re-shaping scientific practice. Nat Rev Genet. 2009;10:331–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56. Eglen  SJ, Marwick  B, Halchenko  YO, et al.  Toward standard practices for sharing computer code and programs in neuroscience. Nat Neurosci. 2017;20:770–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57. Parr  CS, Cummings  MP. Data sharing in ecology and evolution. Trends Ecol Evol. 2005;20:362–3. [DOI] [PubMed] [Google Scholar]
  • 58. Huang  X, Hawkins  BA, Lei  F  et al.  Willing or unwilling to share primary biodiversity data: results and implications of an international survey. Conserv Lett. 2012;5:399–406. [Google Scholar]
  • 59. Moles  A, Dickie  JB, Flores-Moreno  H. A response to Poisot et al.: publishing your dataset is not always virtuous. IEE. 2013;6:20–2. [Google Scholar]
  • 60. Soranno  PA, Cheruvelil  KS, Elliott  KC  et al.  It's good to share: why environmental scientists’ ethics are out of date. Bioscience. 2015;65:69. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61. Roche  DG, Kruuk  LEB, Lanfear  R  et al.  Public data archiving in ecology and evolution: how well are we doing?. PLoS Biol. 2015;13:e1002295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62. Hampton  SE, Strasser  CA, Tewksbury  JJ  et al.  Big data and the future of ecology. Front Ecol Environ. 2013;11:156–62. [Google Scholar]
  • 63. Puškadija  Z, Kovačić  M, Raguž  N  et al.  Morphological diversity of carniolan honey bee (Apis mellifera carnica) in Croatia and Slovenia. J Apic Res. 2021;60:326–36. [Google Scholar]
  • 64. Bouga  M, Hatjina  F. Genetic variability in Greek honey bee (A. mellifera L.) populations using geometric morphometrics analysis. In: Gruev B, Nikolova  M, Donev  A, editor. Proceedings of the Balkan Scientific Conference of Biology in Plovdiv (Bulgaria). Plovdiv: Plovdiv University Press, 2005, 598–602. [Google Scholar]
  • 65. Charistos  L, Hatjina  F, Bouga  M  et al.  Morphological discrimination of Greek honey bee populations based on geometric morphometrics analysis of wing shape. J Apicultural Sci. 2014;58:75–84. [Google Scholar]
  • 66. Oleksa  A, Kusza  S, Tofilski  A. Mitochondrial DNA suggests the introduction of honeybees of African ancestry to east-central Europe. Insects. 2021;12:410. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67. Oleksa  A, Tofilski  A. Podgatunki pszczoły miodnej i rola lasów w ochronie ich różnorodności. In: Czekońska  K, Szabla  K, editors. Ochrona owadów zapylających w ekosystemach leśnych. Krakow: Wydawnictwo Uniwersytetu Rolniczego w Krakowie, 2020. [Google Scholar]
  • 68. Tofilski  A, Căuia  E, Siceanu  A  et al.  Historical changes in honey bee wing venation in Romania. Insects. 2021;12:542. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69. Rašić  S, Mladenović  M, Stanisavljević  L. Use of geometric morphometrics to differentiate selected lines of carniolan honeybees (Apis mellifera carnica) in Serbia and Montenegro. Arch Biol Sci (Beogr). 2015;67:929–34. [Google Scholar]
  • 70. Çakmak  İ, Fuchs  S, Çakmak  SS  et al.  Morphometric analysis of honeybees distributed in northern Turkey along the Black Sea coast. Uludag Bee J. 2014;14:59–68. [Google Scholar]
  • 71. R Core Team . R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, 2018. [Google Scholar]
  • 72. Oleksa  A, Căuia  E, Siceanu  A  et al.  Apis-wings-EU: a workflow for morphometric identification of honey bees from Europe. WorkflowHub.  2023. 10.48546/WORKFLOWHUB.WORKFLOW.422.1. [DOI] [Google Scholar]
  • 73. Baken  EK, Collyer  ML, Kaliontzopoulou  A  et al.  geomorph v4.0 and gmShiny: enhanced analytics and a new graphical interface for a comprehensive morphometric experience. Methods Ecol Evol. 2021;12:2355–63. [Google Scholar]
  • 74. Wood  SN. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J R Stat Soc Ser B Stat Methodol. 2011;73:3–36. [Google Scholar]
  • 75. Schlager  S. Morpho and Rvcg—shape analysis in R: r-packages for geometric morphometrics, shape analysis and surface manipulations. In: Zheng  G, Li  S, Székely  G, editors. Statistical shape and deformation analysis. London: Academic Press, 2017, p. 217–56. [Google Scholar]
  • 76. Schliep  KP. phangorn: phylogenetic analysis in R. Bioinformatics. 2011;27:592–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 77. Nawrocka  A, Kandemir  İ, Fuchs  S  et al.  Dataset: computer software for identification of honey bee subspecies and evolutionary lineages. Apidologie. 2018;49:172–84. [Google Scholar]
  • 78. Dryden  IL. shapes package. Vienna, Austria: R Foundation for Statistical Computing, 2021. [Google Scholar]
  • 79. Chávez-Galarza  J, Garnery  L, Henriques  D, et al.  Mitochondrial DNA variation of Apis mellifera iberiensis: further insights from a large-scale study using sequence data of the tRNAleu-cox2 intergenic region. Apidologie. 2017;48:533–44. [Google Scholar]
  • 80. Franck  P, Garnery  L, Solignac  M, et al.  The origin of West European subspecies of honeybees (Apis mellifera): new insights from microsatellite and mitochondrial data. Evolution. 1998;52:1119–34. [DOI] [PubMed] [Google Scholar]
  • 81. Garnery  L, Franck  P, Baudry  E, et al.  Genetic diversity of the west European honey bee (Apis mellifera mellifera and A. m. iberia). II. Microsatellite loci. Genet Sel Evol. 1998;30:S49. [Google Scholar]
  • 82. Kandemir  İ, Meixner  MD, Ozkan  A, et al.  Genetic characterization of honey bee (Apis mellifera cypria) populations in northern Cyprus. Apidologie. 2006;37:547–55. [Google Scholar]
  • 83. Chávez-Galarza  J, Henriques  D, Johnston  JS  et al.  Revisiting the Iberian honey bee (Apis mellifera iberiensis) contact zone: maternal and genome-wide nuclear variations provide support for secondary contact from historical refugia. Mol Ecol. 2015;24:2973–92. [DOI] [PubMed] [Google Scholar]
  • 84. Beekeeping sector: results of the pilot study on honey bee selection. https://agriculture.ec.europa.eu/news/beekeeping-sector-results-pilot-study-honey-bee-selection-2022-03-15_en. Accessed 2 February 2023.
  • 85. Węgrzynowicz  P, Gerula  D, Tofilski  A  et al.  Maternal inheritance in hybrids of three honey bee subspecies. J Apicultural Sci. 2019;63:131–8. [Google Scholar]
  • 86. Bornus  L, Demianowicz  A, Gromisz  M. Morfologiczne badania krajowej pszczoły miodnej. Pszczelnicze Zeszyty Naukowe. 1966;10:1–46. [Google Scholar]
  • 87. Meixner  MD, Worobik  M, Wilde  J, et al.  Apis mellifera mellifera in eastern Europe—morphometric variation and determination of its range limits. Apidologie. 2007;38:191–7. [Google Scholar]
  • 88. Oleksa  A, Wilde  J, Tofilski  A  et al.  Partial reproductive isolation between European subspecies of honey bees. Apidologie. 2013;44:611–9. [Google Scholar]
  • 89. Parejo  M, Wragg  D, Gauthier  L  et al.  Using whole-genome sequence information to foster conservation efforts for the European dark honey bee, Apis mellifera mellifera. Front Ecol Evol. 2016;4. 10.3389/fevo.2016.00140. [DOI] [Google Scholar]
  • 90. Groeneveld  LF, Kirkerud  LA, Dahle  B  et al.  Conservation of the dark bee (Apis mellifera mellifera): estimating C-lineage introgression in Nordic breeding stocks. Acta Agric Scand Sect A Anim Sci. 2020;69:157–68. [Google Scholar]
  • 91. Jensen  AB, Palmer  KA, Boomsma  JJ, et al.  Varying degrees of Apis mellifera ligustica introgression in protected populations of the black honeybee, Apis mellifera mellifera, in northwest Europe. Mol Ecol. 2005;14:93–106. [DOI] [PubMed] [Google Scholar]
  • 92. Muñoz  I, De la Rúa  P. Wide genetic diversity in Old World honey bees threaten by introgression. Apidologie. 2021;52:200–17. [Google Scholar]
  • 93. Cánovas  F, De la Rúa  P, Serrano  J, et al.  Geographical patterns of mitochondrial DNA variation in Apis mellifera iberiensis (Hymenoptera: apidae). J Zool System. 2008;46:24–30. [Google Scholar]
  • 94. Miguel  I, Iriondo  M, Garnery  L  et al.  Gene flow within the M evolutionary lineage of Apis mellifera: role of the Pyrenees, isolation by distance and post-glacial re-colonization routes in the western Europe. Apidologie. 2007;38:141–55. [Google Scholar]
  • 95. Diniz-Filho  JA, Hepburn  HR, Radloff  S  et al.  Spatial analysis of morphological variation in African honeybees (Apis mellifera L.) on a continental scale. Apidologie. 2000;31:191–204. [Google Scholar]
  • 96. Hepburn  HR, Radloff  SE. Honeybees of Africa. Berlin: Springer Science & Business Media, 2013. [Google Scholar]
  • 97. Daly  HV, Hoelmer  K, Gambino  P. Clinal geographic variation in feral honey bees in California, USA. Apidologie. 1991;22:591–609. [Google Scholar]
  • 98. Oleksa  A, Căuia  E, Siceanu  A, et al.  GigaScience Database.  2023. 10.5524/102360. [DOI] [PMC free article] [PubMed]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Citations

  1. Oleksa  A, Căuia  E, Siceanu  A, et al.  GigaScience Database.  2023. 10.5524/102360. [DOI] [PMC free article] [PubMed]

Supplementary Materials

giad019_GIGA-D-22-00297_Original_Submission
giad019_giga-d-22-00297_original_submission.pdf (2.8MB, pdf)
giad019_GIGA-D-22-00297_Revision_1
giad019_giga-d-22-00297_revision_1.pdf (3.6MB, pdf)
giad019_Response_to_Reviewer_Comments_Original_Submission
giad019_response_to_reviewer_comments_original_submission.pdf (67.9KB, pdf)
giad019_Reviewer_1_Report_Original_Submission

Chris Armit -- 11/17/2022 Reviewed

giad019_reviewer_1_report_original_submission.pdf (125.4KB, pdf)
giad019_Reviewer_1_Report_Revision_1

Chris Armit -- 2/10/2023 Reviewed

giad019_reviewer_1_report_revision_1.pdf (110.2KB, pdf)
giad019_Reviewer_2_Report_Original_Submission

Alexis Matamoro Vidal -- 11/29/2022 Reviewed

giad019_reviewer_2_report_original_submission.pdf (111KB, pdf)
giad019_Reviewer_2_Report_Revision_1

Alexis Matamoro Vidal -- 2/10/2023 Reviewed

giad019_reviewer_2_report_revision_1.pdf (109.5KB, pdf)
giad019_Supplemental_Figures
giad019_supplemental_figures.zip (280.5KB, zip)

Data Availability Statement

The whole dataset, including the wing images, landmark coordinates, geographic coordinates of sampling locations, and other data, is available on the Zenodo website [1] under a public domain license. All details of the statistical analysis, including the identification of an unknown sample, are available at the WorkflowHub website [72]. All supporting data and materials are available in the GigaScience GigaDB database [98].

Articles from GigaScience are provided here courtesy of Oxford University Press

RESOURCES