Introduction

The European hornet, Vespa crabro, is one of the most widely recognized species of the genus Vespa, with a distribution that stretches from western Europe eastward to southern Siberia, China, Korea, and Japan1. Despite its prominence, the taxonomic history of V. crabro has been characterized by significant complexity and ongoing debates, particularly regarding the delineation and validity of its subspecies2. Historically, morphological characteristics, particularly variations in color patterns, have been extensively utilized to differentiate subspecies of hornets1,3. However, recent revisions in Vespa taxonomy have questioned the reliability and consistency of these morphological characters. Smith-Pardo et al.4 critically reviewed Vespa taxonomy, concluding that previously recognized subspecies names should be considered synonyms, thereby substantially undermining the taxonomic significance previously attributed to morphological variations. Instead, these variations have been reduced to informal labels reflective of regional diversity rather than definitive taxonomic characters4. This pivotal reassessment underscores a broader, longstanding controversy in evolutionary biology regarding the utility and empirical grounding of the subspecies concept itself.

One of the most influential modern definitions of subspecies is provided by Braby5, who stated: "subspecies comprise evolving populations that represent partially-isolated lineages of a species that are allopatric, phenotypically distinct, possess at least one fixed diagnosable character state, and for which those character differences are (or are assumed to be) correlated with evolutionary independence as revealed by population-genetic structure." When substantiated by rigorous genetic and ecological evidence, subspecies classifications can be informative, capturing meaningful evolutionary divergence and guiding effective conservation strategies. For example, several Apis mellifera subspecies—and, albeit more controversially, some within A. cerana—exhibit parallel genetic and morphometric divergence that aligns with differences in climate tolerance, pathogen resistance, and colony productivity, underscoring the practical value of subspecies recognition6. Conversely, the subspecies concept has been criticized as arbitrary and subjective, frequently reliant on superficial morphological features that may obscure true genetic variation. This perspective has been notably supported by researchers such as Wilson and Brown7, and later Carpenter and Kojima2, who argued against the empirical objectivity of subspecies. These contrasting views highlight the critical necessity of integrating robust genetic analyses alongside morphological assessments when evaluating subspecies validity.

In the context of V. crabro, earlier classifications by researchers like Matsuura and Yamane3 delineated multiple subspecies primarily based on coloration patterns, a method later refined by Archer1 into a smaller subset of recognized forms: V. c. altaica, V. c. crabro, V. c. germana, V. c. gribodoi, V. c. crabroniformis, and V. c. flavofasciata. In Korea, two of these—V. c. flavofasciata and V. c. crabroniformis—have been diagnosed primarily by mesoscutal and 2nd tergum coloration8,9,10. However, many studies have demonstrated that coloration characters can be plastic and taxonomically misleading11,12,13. Accordingly, recent studies advocate supplementing morphology with molecular markers to test the validity of such color-based classifications.

Mitochondrial markers—particularly the cytochrome-oxidase subunit 1 (CO1) gene—are valuable for delineating subspecies and inferring population structure because they are maternally inherited, lack recombination, and have a small effective population size, which promotes rapid lineage sorting and low within-population diversity. CO1 is now the standard locus for DNA barcoding and phylogeographic studies across a wide range of taxa, offering clear signals of species-level differentiation and historical dispersal routes14,15. In the genus Vespa, CO1 sequences have revealed marked interspecific divergences16,17,18,19; however, subspecies-level comparisons remain limited.

Moreover, mitochondrial DNA markers are invaluable for tracing the geographic origins and dispersal pathways of invasive species because they are maternally inherited and essentially free of recombination20. For example, CO1 barcodes showed that Vespa velutina specimens intercepted in the United Kingdom shared the single haplotype dominant in French populations, pinpointing France as the source of the British incursion21. By contrast, comparable phylogeographic analyses of V. crabro—now established beyond its native Eurasian range—are scant22, leaving key questions about its introduction routes and subsequent spread unresolved.

To fill these knowledge gaps, we conducted an integrated survey combining detailed color-pattern analysis with CO1 mitochondrial barcoding across V. crabro populations. Specifically, we (i) re-examine the historical color-based subspecies scheme and (ii) test its congruence with mitochondrial lineage divergence. By linking morphology and genetics, our study seeks to resolve taxonomic ambiguities surrounding Korean V. crabro subspecies and to clarify the species’ regional invasion pathways and evolutionary diversification.

Results

Color pattern variations in Korean Vespa crabro

Previous studies classified Korean V. crabro as either V. c. crabroniformis or V. c. flavofasciata based on thoracic and abdominal color patterns (Fig. 1A,B)8,9. In this study, we analyzed the color patterns of 163 V. crabro specimens collected from 2009 to 2023 (Supplementary Fig. S1 and Table S1) and identified 27 distinct variants of Korean V. crabro (Table 1). Among these, 42 individuals matched the criteria8,9 for V. c. crabroniformis (group C), displaying seven thoracic and two abdominal patterns (Fig. 1C,D), while 59 individuals classified as V. c. flavofasciata (group F), exhibited three thoracic and one abdominal color patterns (Fig. 1E,F). Notably, 62 specimens exhibited color patterns limited to either the abdomen (tergum 2, Fig. 2A) or thorax (mesoscutum, Fig. 2B) alone, not fitting the criteria of either subspecies as defined Kim et al.8,9 and were therefore assigned to a new category, designated as group X.

Fig. 1
figure 1

Taxonomic classification of Vespa crabro subspecies in Korea based on traditional taxonomic characters, including their color patterns. (A, B) Reference color patterns of V. crabro crabroniformis (A) and V. crabro flavofasciata (B), as described by Kim et al.8,9. Labeled regions: a, pronotum; b, mesoscutum; c, scutellum; d, metanotum; e, basal area of tergum 2. (C–D) Thoracic (C, patterns 1–7) and abdominal (D, patterns i–ii) color patterns observed in individuals assigned to V. c. crabroniformis in this study. (E, F) Thoracic (E, patterns 8–10) and abdominal (F, pattern iii) color patterns of individuals assigned to V. c. flavofasciata. Scale bars: 3 mm.

Full size image
Table 1 Classification of Korean Vespa crabro using color pattern phenotypes.
Full size table
Fig. 2
figure 2

Previously unclassified color pattern combinations in Korean Vespa crabro. (A) Individual exhibiting a T2 abdominal pattern but lacking a mesoscutum pattern. (B) Individual exhibiting a mesoscutum pattern but no visible abdominal (T2) pattern. These individuals were grouped into the novel category “group X”. Black arrow indicates T2 color pattern; Red arrow indicates mesoscutum color pattern. Scale bars: 6 mm.

Full size image

To assess the consistency of coloration across anatomical regions, we conducted Cramér’s V correlation analyses23 involving mesoscutum, scutellum, metanotum, and basal area of T2 coloration. Correlation coefficients between thoracic (b–d) and abdominal (e) color patterns were notably low (0.22–0.26; Fig. 3), underscoring significant independence in coloration across these anatomical regions. Because extremely low-frequency categories may distort correlation coefficients and reduce statistical power24,25,26, pronotum color pattern was excluded from correlation analyses. Conversely, 36 of the 95 individuals (37.9%) lacking mesoscutum coloration (thorax patterns 8–10 in Table 1) still exhibited clear T2 color patterning (abdominal patterns i and ii in Table 1), reinforcing the weak correlation. Within thoracic regions, however, correlations were stronger: mesoscutum-scutellum (0.57), mesoscutum-metanotum (0.53), and notably scutellum-metanotum (0.95) (Fig. 3). These findings challenge the diagnostic utility of mesoscutum and T2 color patterns for subspecies classification.

Fig. 3
figure 3

Correlations analysis of color patterns expression across body regions. Cramér’s V correlation coefficients for color pattern presence across mesoscutum, scutellum, metanotum, and basal area of T2 regions. Values near 0 indicate weak correlation, whereas those near 1 indicate strong correlation. Cross-tabulation of color pattern expression across body regions was performed using the Pandas library27, and the Cramér’s V correlation coefficient was calculated with NumPy28 and SciPy29. The resulting coefficients were visualized using Seaborn30.

Full size image

CO1 sequence analysis of Korean Vespa crabro

To genetically evaluate the subspecies classification of V. crabro based on color patterns, we analyzed mitochondrial CO1 sequences from representative individuals of three groups (C, F, and X; Table 2) using V. binghami as an outgroup (Fig. 4A). The analysis of 29 specimens revealed nine haplotypes (KOR I–IX), with KOR I being most common (n = 14) (Fig. 4A). Previous reported Korean sequences16 (MN716838 and MN716840) matched KOR V and VI, respectively, while 10 GenBank sequences (MN609218–MN609227) aligned with KOR I. Despite overall similarity, six unique single nucleotide polymorphisms (SNPs) (G229, T292, T346, T364, G430, and G466) defined KOR VII (C14 and C15) (Fig. 4A).

Table 2 Information on Vespa crabro species used in CO1 sequence analysis.
Full size table
Fig. 4
figure 4

CO1 haplotype analysis and genetic divergence of Korean Vespa crabro. (A) CO1 sequence alignment of Korean V. crabro individuals. C1–15 correspond to specimens classified as V. c. crabroniformis, whereas F1–6 as V. c. flavofasciata according to the taxonomy of Kim et al.8,9. X1–8 represent individuals with previously unreported color pattern variation. The I–IX designation indicates the specific type of CO1 haplotype. (B) CO1 divergence pattern of Korean V. crabro individuals. The cladogram was constructed using the maximum likelihood method, and the TN93 + I model was applied as the substitution model. No clustering corresponding to color-based subspecies was observed. Arabic numerals enclosed in parentheses indicate the thoracic pattern; Roman numerals represent the abdominal pattern; Regional abbreviations denote collection sites. (C) Collection sites for sequenced individuals. GP Gapyeong; HS Hoengseong; JJ Jinju; NH Namhae; PC Pocheon; PJ Paju; UB Uijeongbu.

Full size image

Contrary to expectations under the color-based subspecies hypothesis, CO1 haplotypes did not correlate with the previously defined groups (C, F, and X). For instance, six individuals classified as V. c. crabroniformis (C1–C6), three individuals classified as V. c. flavofasciata (F1–F3), and five individuals with mixed patterns (X1–X5) shared an identical KOR I sequence (Fig. 4A). Additionally, individuals with identical thoracic patterns but differing abdominal patterns showed identical CO1 sequences (C12, C13), while specimens with identical thoracic and abdominal patterns exhibited sequence divergence (F4–F6) (Fig. 4A,B, and Table 2). Moreover, analysis of genetic divergence patterns using CO1 sequences did not reveal any clustering into discernible groups based on pattern variation (Fig. 4B). The genetic distances measured using the TN93 model ranged from 0.000 to 0.014 among individuals classified as V. c. crabroniformis and from 0.000 to 0.008 among those classified as V. c. flavofasciata (Table 3). Inter-group distances were 0.000 to 0.017, overlapping considerably with intra-group distances. Group X individuals displayed similarly low divergence from both group C and F (0.000 to 0.012 and 0.000 to 0.008, respectively; Table 3), providing no genetic support for subspecies separation. In contrast, the divergence between Korean V. crabro and V. binghami was substantial (0.145–0.150).

Table 3 Genetic distances of Vespa crabro by subspecies and habitat region, calculated using the Tamura-Nei 93 model.
Full size table

Individuals from Jinju (JJ: C2, C10), Paju (PJ: X1, X5), Pocheon (PC: C1, C3-6, F1-2), Uijeongbu (UB: F3), and Namhae (NH: F4) exhibited no consistent genetic clustering (Fig. 4B,C). Furthermore, the sequence data demonstrated that one individual from Namhae (NH: F4) and two from Pocheon (PC: C9, X8) shared identical sequences (Fig. 4B), indicating no discernible regional differences.

Comparative analysis of haplogroups of Korean V. crabro and other countries

To contextualize Korean V. crabro, we compared CO1 sequences with those from Japan (n = 9), Europe (n = 16), and North America (n = 8). According to color-based assignments1, these included V. c. crabro, germana, gribodoi, and flavofasciata (Table S2). Comparative sequence analysis identified two characteristic SNPs, C292 and G592, exclusive to Korean haplotypes (Fig. 5A). All Korean haplotypes had C292 except KOR VII, and G592 was found in all haplotypes except KOR III and VII. Haplotype KOR VII further displayed unique SNP profiles, clearly distinguishing it from the other Korean haplotypes (Fig. 5A).

Fig. 5
figure 5

Comparative CO1 sequence analysis of Vespa crabro across global populations. (A) SNP comparisons between Korean V. crabro haplotypes and those from Japan, Europe, and North America. The colors of the bars indicate SNPs from specific regions (blue = Korea; light pink: Europe A; dark pink = Europe B; medium pink = Europe A and B; green = Japan). KOR Korea; UK United Kingdom; HUN Hungary; EST Estonia; NOR Norway; BLR Belarus; DNK Denmark; GER Germany; FRA France; JPN Japan. (B) CO1 divergence pattern showing the evolutionary relationships of V. crabro across global regions. (C) CO1 sequence comparison showing a Canadian individual (CAN 1) matching Korean haplotype KOR I. (D) SNP profile comparison of North American individuals (US1–4, CAN2–4) with a United Kingdom (UK 1) sample. (E) CO1 haplotype network of global V. crabro populations. Red arrows highlight individuals from North American.

Full size image

Divergence pattern analysis with the CO1 sequences available to us delineated five geographic haplogroups: one Japanese, two Korean (Korea A and B), and two European (Europe A and B) (Fig. 5B). The Japanese group exhibited three distinct haplotypes (JPN 1–4, JPN 5–7, and JPN 8–9) defined by ten SNPs (T38, C103, T136, C205, T268, C334, T397, G544, T556, and T637) (Fig. 5A). The European group was distinguished from the other groups at G229 and was further divided into two groups: a group with T and C at nucleotide positions 61 and 400, respectively, and a group with T at position 277 (Fig. 5A). Overall, our sequence analyses suggested that the Japanese (JPN) V. crabro population is genetically distinct from the Korean and European populations. Moreover, contrary to expectations, a Danish individual (DNK) classified as V. c. germana based on its color pattern clustered genetically with individuals from Estonia (EST) and Norway (NOR), which were classified as V. c. crabro within Europe A (Table S2). Similarly, a British individual (UK) classified as V. c. gribodoi based on its color pattern31 was grouped genetically with individuals from France (FRA) and Germany (GER), classified as V. c. germana, in the Europe B (Fig. 5A, Table S2). The genetic distances between Korean and European groups were moderate (0.008 to 0.017), whereas Japanese populations, historically grouped with Korean populations5, exhibited greater divergence (0.023–0.028) (Table 3).

Analysis of CO1 sequences of North American individuals known to have invaded the continent32 revealed that one Canadian individual (CAN 1, British Columbia), which was identified as V. c. flavofasciata33, exhibited a perfect match with the Korean haplotype KOR I (Fig. 5C). Additionally, four individuals from the United States (US 1–4) and three from Canada (CAN 2–4, Ontario) exhibited SNP profiles matching European populations, clustering most closely with Europe A (Fig. 5D,E, Table S2).

Discussion

The color pattern of V. crabro has historically been regarded as the most important and practical taxonomic character for subspecies identification in both Korea and Europe1,8,9. Indeed, subspecies classification of V. crabro has conventionally been based on their geographic distributions coupled with these color patterns1. However, the reliability of color patterns as stable diagnostic markers has increasingly been challenged due to frequent occurrences of intermediate patterns at the boundaries between subspecies’ geographic ranges34. Therefore, the current study reassessed the validity of color patterns as a taxonomic character for the two previously recognized Korean subspecies: V. c. crabroniformis and V. c. flavofasciata. Kim et al.8,9 suggested that simultaneous presence of specific mesoscutum and tergum 2 (T2) color patterns could distinguish V. c. crabroniformis from V. c. flavofasciata. Contrary to this hypothesis, our analysis identified 27 distinct color patterns among Korean specimens, significantly exceeding previous descriptions. Notably, mesoscutum and T2 color patterns, used as diagnostic characters by Kim et al.8,9, demonstrated a remarkably low correlation (r = 0.22), indicating independent variation between these characters. Furthermore, 38% (62/163) of specimens examined displayed intermediate color patterns combining characters of both putative subspecies. Additionally, we found no evidence of geographic structuring corresponding to color patterns, nor did we detect any genetic differentiation matching these patterns based on analyses of several nuclear and mitochondrial markers. Collectively, these findings clearly demonstrate the inherent variability and limited reliability of color patterns for subspecies delimitation within Korean V. crabro.

Kosyakova et al.35 posited that color pattern variations in social wasps could be influenced by differences in colony rank and initiation timing. Ferrari and Polidori36 reported that Polistes dominula exhibited diminished abdominal pigmentation in regions with higher temperatures and fewer yellow T2 spots with increasing altitude and latitude. These patterns suggest that variations in abdominal pigmentation may reflect differential expression of xanthopterin, a yellow pterin pigment known to provide photoprotection against ultraviolet radiation and hypothesized to function as a biological heat sink in social wasps36. Archer1 classified V. crabro subspecies based on variations in color pattern, assigning the Japanese population and some Korean populations to V. c. flavofasciata. However, the present study demonstrated significant genetic divergence between Korean and Japanese V. crabro populations despite their similar color patterns. This discordance suggests that convergent responses to comparable climates, rather than shared ancestry, underlie the observed pattern similarity.

Our results inevitably cast doubt on the validity of the current subspecies scheme for V. crabro, because the only diagnostic character used to delimit subspecies—the external color pattern—proves unreliable. In the Korean population, we detected (i) no geographically discrete genetic clusters, (ii) no population-specific phenotypes, and (iii) no fixed diagnostic characters satisfying Braby’s widely cited biological-subspecies criteria5. Hence, at least within Korea, V. crabro cannot be partitioned into the two nominal subspecies V. c. crabroniformis and V. c. flavofasciata.

Our CO1 survey of V. crabro across Korea, Europe and Japan likewise challenges the current color-based subspecies scheme. As in Korea, haplogroups defined by CO1 polymorphisms do not align with the traditional color morphs elsewhere, indicating that the latter do not track evolutionary lineages. Although we found several well-supported CO1 haplogroups, their partial geographic clustering is better explained by restricted gene flow over large distances than by historical subspeciation. Recent advances in integrative taxonomy demonstrate that such fine-scale phylogeographic structure routinely emerges—even in taxa long treated as monotypic—when genomic, morphological, and ecological data are jointly analyzed37. Recognizing these micro-lineages is essential for evidence-based taxonomy and precision conservation, as they refine diversification models and inform management decisions37.

Historical accounts place the first North-American record of V. crabro in New York around 1840 and, based on color-pattern comparisons, attribute it to a Western European source1,32. Among the eight North American specimens analyzed, four were assigned morphologically to V. c. germana. However, the CO1 sequences already available show that seven of the eight North-American specimens—including all four identified as V. c. germana carry the T277 SNP characteristic of Northern and Eastern European lineages, challenging the previously assumed Western-European origin. Thus, most North American individuals genetically clustered with Northern and Eastern European groups, implying introductions from these regions. Additionally, a specimen from British Columbia, Canada, reported by Bass et al.33, matched the Korean haplotype KOR I, suggesting an Asian introduction route, specifically from Korea. To definitively determine the precise introduction pathways, broader geographic sampling (e.g., China, Russian Far East) is necessary.

Even so, the present study—like many molecular studies on hornets—relies primarily on the maternally inherited CO1 locus. CO1 data alone cannot justify formal taxonomic splits without support from nuclear markers, morphology and ecology38; accordingly, we treat the haplogroups detected here as provisional mitochondrial lineages. Robust validation of their evolutionary independence will require genome-wide markers, morphometric analyses, and ecological data collected throughout the species’ Eurasian range. Because mtDNA captures only the maternal lineage, it cannot reveal nuclear-level processes such as male-mediated gene flow or sex-biased dispersal39. Identical or divergent CO1 haplotypes may therefore reflect recent introgression, genetic drift or selection on mitochondrial variants, potentially biasing inferences about population history38. Comprehensive genomic approaches—such as the whole-genome study that traced the East-Asian source for the V. mandarinia invasion of North America40—are essential to resolve these uncertainties and to reconstruct V. crabro’s true differentiation and introduction routes.

Nevertheless, the mitochondrial CO1 segment remains valuable due to globally standardized amplification primers and relatively high substitution rates, facilitating rapid and cost-effective detection of recent divergence and low-level gene flow. Additionally, comparing sequences against the extensive international barcode databases (BOLD, GenBank) enables a global-scale quantitative and qualitative understanding of maternal lineages. This approach can identify unexpected maternal introgression signals early on and highlight potential conservation management units. Thus, our study highlights the practical utility of CO1 analysis by uncovering five previously unrecognized provisional mitochondrial lineages within V. crabro and tracing at least two distinct introduction events into North America, thereby providing a solid baseline for forthcoming genome-wide studies of the species’ complex evolutionary dynamics.

In conclusion, our findings clearly indicate that traditional color pattern-based subspecies classifications of V. crabro lack both morphological and genetic support, corroborating previous taxonomic critiques by Carpenter and Kojima2. Nevertheless, we detected five provisional mitochondrial lineages that are geographically structured and can serve as initial genetic markers for reconstructing hornet invasion pathways. By supplementing unstable morphological characters with molecular evidence, this study refines the biogeography of V. crabro and underscores the need for genome-wide data to resolve persistent taxonomic and invasion–ecology questions in social wasps.

Methods

Samples

A total of 163 individuals were collected from 11 regions [Gapyeong (1), Uijeongbu (3), Hoengseong (1), Pocheon (140), Paju (2), Seoul (3), Osan (5), Cheongju (2), Jinju (2), Ulsan (1) and Namhae (3)] between 2009 and 2023, using malaise and sugar-based traps (Table S1). The specimens were stored at − 20 °C in 100% ethanol and were identified following a previously described taxonomic approach for V. crabro subspecies8,9. Only the color patterns of the worker bees were recorded. Specimens were photographed using a Stemi 508 microscope (Carl Zeiss, Oberkochen, Germany) and an Axiocam 506 color camera (Carl Zeiss).

DNA extraction and PCR

DNA was extracted from 29 individuals, including 15 from group C (previously classified as V. c. crabroniformis), 6 from group F (previously classified as V. c. flavofasciata), and 8 from group X (previously unclassified) (Table 1). For each individual, DNA was extracted from the legs using the Drosophila protocol41, and LCO1490 and HCO2198 primers42 were used to amplify the barcoding region of CO1. Using 45 ng of sample DNA, PCR was performed for 35 cycles in a Tubocycler 2 (TCST-9622, Blue-Ray Biotech, Taipei, Taiwan), with denaturation at 94 °C for 30 s, primer annealing at 52 °C for 30 s, and extension at 72 °C for 30 s. The PCR products were sequenced by Macrogen (Seoul, South Korea). Sequences from both directions were edited and combined using BIOEDIT v 7.2.543, and each sequence was aligned using MUSCLE44 with all sequences aligned to a length of 658 bp.

Genetic divergence pattern analysis

Genetic divergence pattern analysis of CO1 sequences was performed using the maximum likelihood method via Mega 1145, with the TN93 + I model46 selected as the substitution model. The analysis was conducted using 1,000 bootstrap iterations. The most appropriate substitution model was selected using the JMODEL test47 and haplotype networks were constructed using Popart48. Genetic distance was determined using the TN93 model46 and marker information for the outgroups was obtained from GenBank (GenBank accession number: MN716843). The analysis included 19 V. crabro accessions from GenBank and 14 accessions from BOLD: 2 from the United Kingdom (GenBank no. LC510543 and OU342425), 5 from Germany (BOLD ID. GBACU1942-12, GBACU1943-12, GBACU212-12, GBACU1944-12, and GBACU210-12), 1 from Denmark (GenBank no. MT862429), 1 from Belarus (GenBank no. MF416105), 1 from Hungary (BOLD ID. GBMIX563-14), 2 from Norway (BOLD ID. NOVES042-14 and NOVES043-14), 2 from Estonia (BOLD ID. ACUFI813-13 and ACUFI814-13), 2 from France (BOLD ID. LPRCI1430-21 and LPRCW149-19), 4 from the United States (GenBank no. PP736946, MH587765, KF933087; BOLD ID. WASPS020-11), 4 from Canada (GenBank no. OL702713; BOLD ID. TZBCA227-06, OPPEM2370-17 and OPPQQ093-17), and 9 from Japan (GenBank no. PP737000, PP737001, PP736997–PP736999, PP737024, PP737025, PP737028, and LC619074).