The Kiklades (Cyclades) island group is located in the central Aegean Archipelago in Greece and comprises 22 large islands and numerous smaller islands and islets. The floristic region of the Kiklades hosts approximately 1,750 vascular plant taxa, 162 of which are Greek endemics (Dimopoulos et al. 2016). This relatively high proportion of endemism reflects a long history of geological activity, isolation, and ecological heterogeneity, and is further enhanced by genetic drift in small populations, which has repeatedly promoted local diversification and insular speciation (Runemark 1969, 1970; Strid 1996; Kougioumoutzis et al. 2021).

Within this regional context, the genus Allium L. represents one of the most diverse and evolutionarily dynamic components. Globally, Allium comprises more than 1,000 species, currently classified into 15 subgenera and 72 sections (Friesen et al. 2006; POWO 2026). Greece is widely regarded as a major diversification centre for the genus (Stearn 1981; Tzanoudakis and Vosa 1988), hosting more than 110 species, over 50% of which are endemic (Dimopoulos et al. 2013; Brullo et al. 2015; Tzanoudakis and Trigas 2015; Trigas et al. 2017). In the Kiklades alone, approximately 26 species and subspecies of Allium are currently recorded, including three narrow local endemics (A. luteolum Halácsy, A. apolloniense Biel., Kit Tan & Tzanoud., and A. brulloi Salmeri), underlining the importance of this region as a diversification centre for the genus within the Aegean Archipelago.

All endemic Allium taxa in the Kiklades belong to A. sect. Codonoprasum Rchb. (subg. Allium), one of the most species-rich and taxonomically intricate sections of the genus, currently comprising more than 150 accepted species (Kobrlová et al. 2024). Allium sect. Codonoprasum is morphologically coherent yet highly polymorphic, and is characterized by the presence of two unequal, persistent spathe valves that both exceed the length of the pedicels and typically consist of a lanceolate base that tapers into a slender, caudate (tail-like) appendage. The flowers are cylindrical-campanulate or cup-shaped, with simple stamens. Nectaries are absent or inconspicuous (Stearn 1980; Brullo and Salmeri 2021). The section shows a marked concentration of narrow endemics in the Eastern Mediterranean, particularly on isolated and specialized habitats, highlighting the strong tendency toward local differentiation, micro-endemism, and insular radiation. This pronounced pattern of fine-scale diversification, combined with extensive morphological overlap in closely related taxa, and frequent polyploidy, renders A. sect. Codonoprasum one of the most taxonomically challenging groups within Allium.

In May 2025, two groups of botanists visited the Milos island group independently. Kit Tan and Gert Vold from the University of Copenhagen completed the sampling for the preparation of the Flora of Milos (Biel and Tan 2026). The second group focused on the small offshore islets as part of a biodiversity research project initiated following the Greek government’s declaration of the South Aegean Marine Park 1 – Southern Cyclades. Noteworthy Allium populations, which do not match the descriptions of any currently known Allium species from the Milos island group, were collected by the first group at the north-western cape of Milos and by the second group on the island of Antimilos, 8 km northwest of Milos. These populations were morphologically similar to each other; they superficially resembled A. occultum Tzanoud. & Trigas, a species described from Skyros Island in the Western Aegean (Tzanoudakis and Trigas 2015). Using an integrative biosystematic approach, it was demonstrated that the populations from Milos and Antimilos differ markedly from A. occultum and other Mediterranean species of A. sect. Codonoprasum. Thus, we decided to describe them as representing A. erimomeli, a species new to science.

Milos island, the fifth largest island in the Kiklades, covers an area of 151 km² and reaches a maximum elevation of 751 m a.s.l. Antimilos island lies northwest of Milos, covering an area of ca 8.5 km², and reaching an elevation of ca 650 m a.s.l. Both islands are volcanic in origin and form part of the South Aegean Volcanic Arc (Xydous et al. 2023), a geologically young and environmentally diverse region, which has repeatedly promoted local plant diversification. Although the flora of Milos has been the subject of numerous floristic studies (e.g. Biel and Tan 2022, 2024, 2026), and an ethnobotanical study (Perouli and Bareka 2022), the flora of Antimilos remains poorly documented.

The morphological study and description of the new species were based on herbarium and living material that was collected and cultivated, comprising six individuals from Antimilos Island and ten from Milos Island. Comparative analyses, focusing on A. occultum, which shares some key morphological features with the new species, were conducted using ten individuals collected from the type locality on Skyros Island. Additionally, one accession of A. pilosum Sm., another species of A. sect. Codonoprasum occurring on Milos Island, was incorporated into selected analyses. Field collections of the new species were made in May 2025, and the living material was cultivated in the experimental garden of the Laboratory of Systematic Botany at the Agricultural University of Athens. All morphological observations were conducted using a Zeiss Stemi 508 stereomicroscope.

Chromosome numbers and morphology of A. erimomeli and A. occultum were determined from metaphase plates of root tip meristems prepared using the squash technique of Östergren and Heneen (1962), with minor modifications following Kriemadi et al. (2025). Chromosome preparations were examined under a Zeiss Axiophot photomicroscope equipped with a Jenoptik Gryphax digital camera.

Leaf cross-sections were fixed in glutaraldehyde, post-fixed in osmium tetroxide, dehydrated in an acetone series, and embedded in Spurr’s resin. Transverse sections were prepared using a ULTROTOME III (LKB) ultramicrotome, and structural observations were carried out using a transmission electron microscope (TEM).

DNA-ploidy and absolute genome size (2C value in pg as per Greilhuber et al. 2005) were estimated for the new species (4 individuals/1 population), A. occultum (4/1) and A. pilosum (3/1), respectively (Suppl. material 1). Following the protocols of Duchoslav et al. (2010) and Vojtěchová et al. (2023), measurements were conducted on a CytoFLEX flow cytometer (Beckman Coulter Inc., U.S.A.). Secale cereale L. ‘Daňkovské’ (2C = 16.19 pg; Doležel et al. 1998) served as a primary internal standard; Triticum aestivum ‘Saxana’ (2C = 34.24 pg) was used for the 2C estimation of the new species and it was calibrated against S. cereale. Propidium iodide served as a fluorescent stain. To obtain the absolute monoploid genome size (1Cx DNA), the 2C values were divided by the corresponding ploidy level of each individual.

For the molecular part of the study, accessions of the new species (4 individuals/1 population), A. occultum (3/1) and A. pilosum (3/1) were processed (Suppl. material 1). Total genomic DNA was isolated from fresh leaf tissue using a high-salt CTAB extraction buffer combined with a sorbitol pre-wash (Inglis et al. 2018). Amplification and sequencing of the nuclear ribosomal internal transcribed spacer (nrITS) region followed the methodology of Vojtěchová et al. (2024). All generated sequences were uploaded to GenBank (PZ356864–PZ356875). Manual alignment of the sequences was performed in Geneious v.7.1.7 (Kearse et al. 2012). To broaden the analysis, we added 64 sequences from 20 species of A. sect. Codonoprasum and two species of closely related A. sect. Pallasia (Tzag.) F.O.Khass., R.M.Fritsch & N.Friesen sourced from earlier datasets (Friesen et al. 2022; Vojtěchová et al. 2024; Duchoslav et al. 2026a, 2026b; Nikolopoulos et al. 2026). Allium vineale L. and A. ampeloprasum L. (sect. Allium) were used as outgroups. We identified the optimal model for sequence evolution through jModelTest (Posada 2008). Maximum likelihood (ML) trees were generated in W-IQ-TREE (Trifinopoulos et al. 2016) using the TIM2 + G model, 1,000 bootstrap replicates, and pairwise deletion. Bayesian inference (BI) was conducted in MrBayes v.3.2.6 (Ronquist and Huelsenbeck 2003) under the GTR + G model. The MCMC algorithm involved two independent runs of four chains for 5 million generations, with sampling every 1,000th generation. Chain convergence was confirmed before discarding the first 10% of samples as burn-in.

An assessment of conservation status was conducted following the IUCN Red List Categories and Criteria (IUCN 2012) and the Guidelines for Using the IUCN Red List Categories and Criteria (IUCN Standards and Petitions Committee 2024). Extent of Occurrence (EOO) and Area of Occupancy (AOO) were estimated using GeoCAT (Bachman et al. 2011), with AOO calculated using the standard 2 × 2 km grid.

Allium erimomeli belongs to A. sect. Codonoprasum, as it exhibits all diagnostic morphological features of the section. Among the other members of the section occurring in the Kiklades island group, the new species displays distinct morphological characters. Even when compared to A. pilosum, a species occurring on Milos, Kimolos, Astypalea, and Psara, A. erimomeli is clearly different. Although A. pilosum may appear similar at first glance, its densely hairy leaves, the shape of perianth segments, and its subglobose to subglobose-ovoid ovary distinguish it from the new species. These observed morphological differences are further supported by our phylogenetic analysis, which demonstrated that the two species are not closely related. Instead, A. pilosum belongs to a distinct clade, where it is sister to A. rhodopeum Velen, which has hairy leaves and is distributed on the mainland Balkan Peninsula, the East Aegean Islands, and Anatolia. This phylogenetic position indicates that the presence of hairs on vegetative parts is likely a synapomorphy within that lineage, supporting previous hypothesis (Duchoslav et al. 2026b).

However, morphological observations and measurements revealed that A. erimomeli shares several characteristics with A. occultum (Tzanoudakis and Trigas 2015), a species native to Skyros Island in the West Aegean floristic region. Both species possess ovoid to narrowly ovoid bulbs with bulblets present, short spathe valves, and similar perianth segments. However, they are clearly distinguished by differences in leaf number, shape, and dimensions, pedicel length, and by ovary morphology (Table 1; Fig. 6).

Despite the overall similarity in chromosome morphology, A. erimomeli and A. occultum differ markedly in the morphology of their satellited chromosomes (Fig. 3). In A. erimomeli, the SAT chromosomes bear small, spherical satellites that are often difficult to detect. By contrast, A. occultum has large, readily detachable satellites, which may give the impression of B chromosomes, as reported by Tzanoudakis and Trigas (2015). In both species, the presence of acrocentric chromosomes suggests that, in addition to polyploidization, structural chromosomal rearrangements have contributed substantially to karyotype evolution. These acrocentric homologues also increase karyotype asymmetry, which is more pronounced in A. erimomeli.

The close morphological affinity between A. erimomeli and A. occultum is further corroborated by our phylogenetic analysis, recovering them as strongly supported separate sister clades in the ITS tree (Fig. 5). Karyo-cytogenetic analysis further supports their separation, by revealing clear differences in ploidy levels and genome size: A. erimomeli is hexaploid (2n = 6x = 48), whereas A. occultum is tetraploid (2n = 4x = 32) with a correspondingly lower 2C DNA value. In contrast to the relatively common tetraploidy, hexaploidy is exceptionally rare in A. sect. Codonoprasum (Kobrlová et al. 2024) and was previously documented in only three taxa: the strictly hexaploid A. exaltatum (Meikle) Brullo, Pavone, Salmeri & Venora (Brullo et al. 2004), the mixed-ploidy A. flavum subsp. tauricum (Besser ex Rchb.) K.Richt. (Kobrlová et al. 2024 and reference therein), and the widespread polyploid A. oleraceum L. (Duchoslav et al. 2010, 2013, 2020). Notably, our new AGS estimates represent the lowest 2C values recorded to date for hexaploid representatives of A. sect. Codonoprasum and are among the lowest for tetraploids in this section. In fact, they closely resemble the 2C values typically observed in diploids/triploids and tetraploids within this section (Kobrlová et al. 2024). Similarly low AGS estimates, contrasting with the predominantly mainland pattern in A. sect. Codonoprasum (Kobrlová et al. 2024), have recently been reported for the recently described A. lesbiacum (Nikolopoulos et al. 2026) from Lesvos Island and the Balkan A. nebulosum Duchoslav, Vojtěch. & Kobrlová (Duchoslav et al. 2026a), both of which are closely related to A. erimomeli and A. occultum.

Recent large-scale studies have demonstrated that insular endemics tend to have lower 2C values compared to their mainland relatives due to DNA loss as an adaptive response to nutrient-limited soils and more climatically stressful environmental conditions on islands (Suda et al. 2003, 2005; Knight et al. 2005; Pisarenco et al. 2025). Often this is combined with chromosome doubling in insular lineages, which together likely provides the necessary genomic flexibility to trigger rapid speciation and better adaptation to vacant ecological niches (Meudt et al. 2021; Sobhanian et al. 2026). While this scenario might also seem attractive for A. sect. Codonoprasum, the pattern of 2C values discussed above might be more likely explained by shared ancestry within this specific lineage.

The leaf anatomy of A. erimomeli provides additional diagnostic characters that support its distinction from A. occultum (Fig. 4). In a transverse section, the leaf of A. erimomeli is semicylindrical, with five prominent abaxial ribs and a flat to slightly concave adaxial surface, whereas A. occultum has a subcylindrical to terete outline with six ribs. The leaf of A. occultum is also smaller in cross-sectional diameter, indicating a more reduced leaf structure. Internal anatomy differs accordingly: A. erimomeli shows two layers of elongated, cylindrical palisade cells and 14 vascular bundles (ten abaxial and four adaxial), whereas A. occultum has a single palisade layer and only seven vascular bundles. Together, these differences in leaf size, rib number, palisade organization, and vascular arrangement provide additional evidence for recognizing A. erimomeli as a distinct species.

The origin of the Aegean polyploids A. erimomeli and A. occultum warrants further investigation, particularly towards the Anatolian Peninsula. Currently, the insufficient taxon sampling of Anatolian representatives of A. sect. Codonoprasum in published phylogenetic studies precludes a robust assessment of their evolutionary relationships with potential Anatolian relatives. Consequently, the phylogenetic relationships of these two Aegean species to eastern lineages remain unknown. Nevertheless, comparative morphological evidence suggests that the Anatolian endemic A. sibthorpianum Schult. & Schult.f. exhibits a close phenotypic resemblance to both A. erimomeli and A. occultum. The three species share a suite of characters, including low plant stature, strongly coriaceous outer bulb tunics, a relatively few-flowered inflorescence with short spathe valves, and pinkish tepals. Despite these similarities, A. sibthorpianum can be clearly distinguished from A. erimomeli by several diagnostic traits. Notably, A. sibthorpianum is a high-altitude species that differ from A. erimomeli in its leaf number and shape, longer leaf sheaths that envelop the scape for 1/2–1/3 of its length, larger tepals, mucronate anthers, and its oblong ovary that is conspicuously constricted in its median portion (Özhatay et al. 2010). Moreover, the leaf cross-section of A. sibthorpianum is semicylindrical, bearing six abaxial undulations (costae), and has a palisade tissue composed of a single layer of cylindrical cells (Özhatay et al. 2010), clearly differing from that of A. erimomeli.

Allium rumelicum Koçyigit & Özhatay is a tetraploid species known only from its type locality in European Turkey and has been considered related to A. sibthorpianum (Özhatay et al. 2010). It is a robust species resembling A. occultum, particularly in leaf morphology, but it clearly differs from A. erimomeli in several diagnostic characters. These include ribbed bulb tunics with pinkish-purple inner tunics, the absence of bulblets, leaf sheaths enclosing the scape for 1/2–2/3 of its length, differences in leaf number and shape, mucronate outer tepals, and a larger, ellipsoid-oblong ovary that is distinctly narrowed toward the apex (Özhatay et al. 2010; Koçyiğit et al. 2024).

In addition to morphological differentiation, cytogenetic evidence further supports the distinctiveness of A. erimomeli from both A. sibthorpianum and A. rumelicum. Allium sibthorpianum is diploid (2n = 2x = 16), while A. rumelicum is tetraploid (2n = 4x = 32) (Özhatay et al. 2010). While A. erimomeli is characterized by the presence of two sets of acrocentric chromosomes, the karyotypes of A. rumelicum and A. sibthorpianum lack acrocentric chromosomes but are distinguished by the presence of satellite (SAT) chromosomes (Özhatay et al. 2010). In A. erimomeli, only two satellited acrocentric chromosomes were detected. By comparison, A. rumelicum exhibits two sets of SAT homologues, whereas in the diploid A. sibthorpianum possesses two pairs of SAT chromosomes, mainly associated with metacentric and submetacentric chromosomes. This configuration indicates more symmetrical karyotypes than those of species bearing acrocentric chromosomes. Additionally, both A. erimomeli and A. occultum have chromosomes that are considerably smaller in size than those of A. rumelicum and A. sibthorpianum.

Overall, A. erimomeli appears to have undergone both polyploidization and structural chromosomal rearrangements, resulting in an increased number of acrocentric chromosomes. In contrast, A. rumelicum and A. sibthorpianum retain more conserved karyotypes. These observations support that karyotypic differentiation and polyploidy play a crucial role in species delimitation and evolutionary divergence within this group.

Broader phylogenetic analyses with expanded Anatolian sampling are required to clarify the relationships between the two Aegean polyploids and eastern representatives of A. sect. Codonoprasum. In particular, the inclusion of A. sibthorpianum and A. rumelicum is crucial, as these morphologically allied taxa may be part of an eastern lineage complex related to A. erimomeli and A. occultum. Such data will be critical for assessing whether their similarities reflect shared ancestry or convergent evolution.