Catch-All No More: Integrative Systematic Revision of the Genus Allolobophora Eisen, 1874 (Crassiclitellata, Lumbricidae) with the Description of Two New Relict Earthworm Genera

,


Introduction
The taxonomy of earthworms has proven to be very unstable over the last two centuries. This is due to a lack of systematically useful morphological characters and the fact that most of them are homoplasious or symplesiomorphic [1]. Within Lumbricidae Rafinesque-Schmaltz 1805, the most speciose family of the Palaearctic, some genera have been inadvertently used as taxonomic wastebaskets due to excessively loose diagnoses. As shown by the molecular phylogenetic analyses of Dominguez et al. [2], such catch-all taxa, e.g., Aporrectodea Örley, 1885, Allolobophora Eisen, 1874, or Helodrilus Hoffmeister, 1885, contain independent evolutionary lineages which should be described as different genera.
Allolobophora was established by Eisen in 1874 based on external morphological characters and included seven species, but without selecting a type species. Further species were progressively added to a somewhat refined Allolobophora (as internal characters were added to its diagnosis), but after the revision of Pop [3], it became a catch-all genus [1]. Later works suggested that Allolobophora should be divided into subgenera or even different genera [4][5][6][7][8], but with little consensus.
Eophila Rosa, 1893 is in strong conflict with Allolobophora, as numerous species have been transferred between the two genera in successive taxonomic revisions. It has also been considered synonymous with Heraclescolex Qiu and Bouché, 1998, a genus consisting mainly of green pigmented earthworms with little distinction from the other two genera. The inclusion of the type species of the genus Eophila, Eophila tellinii Rosa, 1888, in molecular phylogenetic analysis by de Sosa et al. [11] has eliminated the confusion: Eophila should be restricted to Eo. tellinii, Eophila gestroi (Cognetti de Martiis, 1905), and Eophila crodabepis [12]. Furthermore, the genotype of Heraclescolex (All. moebii) was nested with All. chlorotica, thus suggesting the synonymy of the latter genera.
Koinodrilus  was created to accommodate species previously assigned to Allolobophora and/or Aporrectodea. Its type species, Aporrectodea georgii Michaelsen, 1890, and a few other representatives (Aporrectodea jassyensis (Michaelsen, 1891), Aporrectodea limicola (Michaelsen, 1890), Aporrectodea rosea (Savigny, 1826), and Allolobophora oliveirae (Rosa, 1894)) were included in phylogenetic trees by Domínguez et al. [2], but the results did not strongly support this genus. On the contrary, Ap. georgii and Ap. jassyensis formed a clade, but the rest of the species appeared scattered within other well-supported clades or behaved as rogue taxa (with unstable positions).
This paper places the aforementioned species in a molecular phylogenetic context to (i) place the last few incerta sedis French endemic species into their appropriate genera, (ii) interrogate whether Allolobophora and Heraclescolex are synonymous, and (iii) test the validity of Koinodrilus.

Samplings and Studied
Specimens. This study follows the methodology of Marchán et al. [9,15], and the method description partially reproduces their wording. 15 specimens of six species of Lumbricidae (genera Allolobophora and Aporrectodea) were collected during two sampling surveys in southeastern and northeastern France in spring 2021 and autumn 2021, respectively. The list of species and the sampling localities are given in Table 1. Individuals were collected by digging up the soil and sorting by hand, rinsing with water, and fixing in pure ethanol to allow further molecular analyses. Sampling and handling of specimens were done ethically and in accordance with Directive 2010/ 63/EU. Species classification and morphological diagnoses were made under a binocular stereomicroscope using the set of external and internal morphological characters used by Qiu and Bouché [16] and following the format established by Domínguez et al. [17]. The following main external morphological characters were considered: mean length, mean number of segments, mean weight, pigmentation, type of prostomium, setal arrangement, position of papillae, position of first dorsal pore, nephridial pore arrangement, position and development of male pores, position and development of female pores, position of spermathecal pores, position of clitellum, and position of tubercula pubertatis. The main internal anatomical features were position of oesophageal hearts, position and morphology of calciferous glands, position of crop, position of gizzard, type of typhlosole, shape of nephridial bladders, number and position of seminal vesicles, and number and position of spermathecae.
The best-fitting evolutionary model for each partition was selected with jModelTest v. 2.1.3 [27] following the Akaike Information Criterion (AIC; [28]) and the Bayesian Information Criterion (BIC; [29]). GTR+I+G was selected as the best evolutionary model for COI, 28S, and ND1, whereas GTR+G and HKY+I+G were chosen for the molecular markers 12S and 16S, respectively. Phylogenetic relationships were inferred by Bayesian inference analysis (BI) using MrBayes v.3.2.6 [30] implemented in CIPRES Science Gateway V. 3.3 [31]. Parameters were set to 50 million generations and sampled every 5,000th generation (10,000 trees). Two independent runs of four chains each were performed, and 20% of the trees were discarded as burn-in. The remaining trees were combined and summarized on a 50% majority-rule consensus tree. Maximum likelihood phylogenetic inference was performed using RAxML-NG [32], starting from 10 random initial trees and 1000 fast bootstrap replicates. Clade support values over 70% and 90% (for bootstrap and posterior probability, respectively) were considered high.
Uncorrected average pairwise distances between the studied species for the molecular markers COI and 16S were calculated in MEGA 11 [33] to support their status as separate species and to investigate genetic distances within and between genera.

Morphological Study.
The studied specimens fitted the diagnoses of their respective species [4,13,14].

Molecular Phylogenetic Analysis.
Phylogenetic relationships determined by Bayesian inference (Figure 1) and maximum likelihood (Supplementary Figure 1) were congruent with previous phylogenetic analyses and between them, with the main internal branches resolved receiving high statistical support. The results do not support the monophyly of Allolobophora as currently defined, as the studied species appeared scattered throughout different clades. However, All. burgondiae and Ap. icterica were recovered in a clade with All. chlorotica and the species formerly assigned to Heraclescolex (All. moebii-designated as its type-All. molleri, Ap. icterica, and Ap. dubiosa). In contrast, the four remaining species were recovered as more distantly related taxa, not sharing the most recent common ancestor with Allolobophora. Although with low support, All. zicsii was recovered in both phylogenetic analysis as a sister species of Avelona ligra (Bouché, 1969), the only known species of the genus Avelona Qiu and Bouché, 2000; both species formed a clade sister to the strongly supported genus-level clade Gatesona. A similar scenario was found for All. satchelli, who was recovered as a sister taxon to Panoniona leoni Michaelsen, 1891. With respect to Allolobophora tiginosa, this taxon was placed in a large clade comprising Castellodrilus Qiu and Bouché, 1998; Zophoscolex Qiu and Bouché, 1998; Ethnodrilus Bouché, 1972;Prosellodrilus Bouché, 1972; and Cataladrilus Qiu and Bouché, 1998, forming a well-supported clade with the last two mentioned genera. However, the precise phylogenetic relationship between them remained unsolved. Finally, the results for Aporrectodea pseudoantipai show that it belongs in a clade with the representatives of Aporrectodea (except for Aporrectodea rosea, which is considered a rogue taxon), with Aporrectodea handlirschi (Rosa, 1897) as its closest relative. Aporrectodea georgii (type species of Koinodrilus) showed no affinity with the species once assigned to Koinodrilus (zicsii, pseudoantipai, rosea, jassyensis, limicola, and oliveirae). Allolobophora bartolii (Bouché, 1970), although with minor support, does not appear to be related to Allolobophora sensu stricto. Instead, it is more closely related to Eophila Rosa, 1893 and Imetescolex Szederjesi et al., 2022.
The four taxa not placed within the Allolobophora sensu stricto clade showed high distance values for both markers with representatives of that clade (including All. burgondiae and All. icterica), with All. satchelli showing the smallest distances with the species of the genus (Table 2). However, All. satchelli displayed its lowest distance values for both markers with P. leoni (14.71% for COI, the lowest value for this marker, and 6.00% for 16S). For All. zicsii, UAPG distances for the molecular markers COI and 16S were high between the species and every taxon included in the analysis, even with its closest relative Av. ligra (19.17% for COI and 14.02% for 16S). All. tiginosa showed its lowest distance with the type species of the genus Cataladrilus, C. monticola (10.61% for 16S), while Ap. pseudoantipai showed generally lower distances with Ap. handlirschi than with any other species of Allolobophora.

Discussion
4.1. Systematic Implications. The inclusion of All. burgondiae and Ap. icterica within a well-supported clade with All. chlorotica, All. molleri, and All. dubiosa and the low UAPG distance values among them suggest that both species should be placed within Allolobophora. The phylogenetic position of Ap. pseudoantipai with the representatives of Aporrectodea, On the other hand, the nonmonophyly of All. satchelli, All. zicsii, and All. tiginosa with the species of Allolobophora sensu stricto implies these species should be assigned to other genera. For All. satchelli, the Bayesian inference and the UAPG distances support the inclusion of the species within the genus Panoniona Mršić and Šapkarev, 1988, changing its status to Panoniona satchelli. Interestingly, All. bartolii was placed by Qiu and Bouché [8] within Panoniona, which was not supported by molecular phylogenetics results. This highly polymorphic species (probably being a species complex) requires further study before a new genus can be proposed for it. The high distance values between All. zicsii and all the studied species and the long branches that separate this taxon from its closest relative suggest the designation of a new genus for this species. Similarly, All. tiginosa has been shown to be closely related to the genera Prosellodrilus and Cataladrilus but without belonging to the same genus-level clade, thus justifying the creation of its own genus.
The nonmonophyly of Aporrectodea georgii with the other species included in Koinodrilus suggests that it would be a monospecific genus if retained. Also, Aporrectodea georgii does not appear to be related to Aporrectodea sensu stricto, the closest relatives of Aporrectodea trapezoides (Dugès 1828), so it should not be classified within Aporrectodea. Therefore, Koinodrilus would be a valid name, and a new one would not be necessary. Interestingly, Marchán et al. [34] recovered a close phylogenetic relationship between Ap. georgii and All. chlorotica when using the novel phylogenomic technique AHE (Anchored Hybrid Enrichment, [35]): this relationship was recovered with low support in the maximum likelihood analysis based on the traditional Sanger-sequenced markers. As the taxonomic sampling was limited (43 species), an increased number of species could confirm or deny such an evolutionary affinity.
Allolobophora differs from Panoniona by the more posterior position of the clitellum, frequently green pigmentation vs. absence of pigmentation, repetition of spermathecae (simple vs. double our multiple), number of seminal vesicles (four pairs in 9, 10, 11, and 12 vs. two pairs in 11 and 12), and shape of nephridial bladders (U-shaped vs. J-shaped).
Allolobophora can be separated from Eophila by its generally smaller size, its pigmentation (if present: green vs. purplish-brown in bands or brown), and the position of the spermathecae (variable in 7/8, 8/9, 9/10, and 10/11 vs. 9/10 and 10/11). The comparison of the mentioned genera is shown in Table 3.
Species included the following: (i) Allolobophora chlorotica (Savigny, 1826) (ii) Allolobophora burgondiae Bouché, 1972 (iii) Allolobophora icterica (Savigny, 1826) (iv) Allolobophora molleri Rosa, 1889 (v) Allolobophora moebii Michaelsen, 1895 (vi) Allolobophora dubiosa Örley, 1881 Remarks: the species assigned to Heraclescolex by Qiu and Bouché [13] form a remarkably homogenous group in terms of morphology and anatomy. Since four species formerly assigned to Heraclescolex (including the type species) have been recovered within Allolobophora by phylogenetic analyses, it is likely that the rest will also belong to this genus. The taxonomic status of the Allolobophora molleri- Journal of Zoological Systematics and Evolutionary Research Table  2: Uncorrected average pairwise genetic (UAPG) distances between the species of study and representatives of the genus Allolobophora, Cataladrilus, Prosellodrilus, Avelona, Panoniona, and Aporrectodea, for 16S (above the diagonal) and COI (below the diagonal). The cell could be shown in green (low genetic distance between the compared taxa), yellow (medium), or red (high).   Differential diagnosis:

Journal of Zoological Systematics and Evolutionary Research
Panoniona can be differentiated from Allolobophora by the more anterior position of the clitellum and absence of pigmentation vs. frequently green pigmentation, repetition of spermathecae (double our multiple vs simple), number of seminal vesicles (two pairs in 11 and 12 vs. four pairs in 9, 10, 11, and 12), and shape of nephridial bladders (Jshaped vs. U-shaped).
Panoniona is clearly separated from Gatesona by the absence of pigmentary dots in cephalic segments, absence of very conspicuous genital papillae between segment 9 and the end of the clitellum, shape of nephridial bladders (J-shaped vs. sigmoid), repetition of spermathecae (double our multiple vs. simple), and number of seminal vesicles (two pairs in 11 and 12 vs. three to four pairs in (9, 10), 11, and 12).
Panoniona differs from Eophila by its generally smaller size, its pigmentation (absent vs. purplish-brown in bands or brown), and repetition of spermathecae (double our multiple vs. simple; Table 3).
Species included the following: Differential diagnosis: Vosgesia shares with Avelona (their closest relatives) the position of spermathecae (9/10 and 10/11) and seminal vesicles (11 and 12), typhlosole shape (bifid), and several external character states, but it is clearly separated by the position of the dorsal pore (4/5 vs. 8/9), the position of the clitellum (26-34 vs. 29-36), the nephridial pore disposition (aligned vs. irregular), the type of spermathecae (simple or double vs. simple), and the shape of the nephridial bladders (U-shaped vs. digitoid). The complete diagnoses of both genera are listed in Table 3.
Other species included: none.

Evolutionary and Biogeographic
Implications. Historically, the genus Allolobophora includes a large number of species that are often taxonomic unrelated, resulting in a genus with a wide distribution and high diversity (the second largest one within the Lumbricidae), but with weak evolutionary ties. Unfortunately, this has complicated evolutionary and biogeographic interpretations. The gradual implementation of molecular techniques has revolutionized the systematics of the genus, with the elevation of redefined Cernosvitovia [10] and Gatesona [9], previously considered subgenera of Allolobophora, to generic status. The present work has continued this trend by supporting the elevation of Panoniona to genus status and the assignment of two of the studied species to newly described genera. The resulting monophyletic clade comprising the former Allolobophora sensu stricto (with All. chlorotica and All. burgondiae) and Heraclescolex (comprising All. moebii and All. molleri, All. icterica, and All. dubiosa) appears more coherent in its morphological variability and distribution than the loosely defined Allolobophora of old. Qiu and Bouché [16] already indicated the strong morphological similarity between Allolobophora sensu stricto and Heraclescolex; in spite of that, surprisingly, they proposed to place them in different tribes. Domínguez et al. [2] showed for the first time that All. molleri, All. moebii, and All. dubiosa are closely related taxa to All. chlorotica by molecular phylogenetic methods. The presence of All. burgondiae and All. icterica within that clade reinforces the idea that Allolobophora and Heraclescolex are synonyms. An alternative interpretation of the topology of the phylogenetic trees might be that Allolobophora sensu stricto (with All. chlorotica and All. burgondiae) and Heraclescolex do not constitute the same genus but are very closely related, independent genera. Even though both scenarios are likely, their extremely similar morphology (including the widespread green pigmentation) advises to synonymize them. Their main difference, the orientation of the nephridial bladders, is a rather feeble taxonomic character within a strongly variable structure within genera [9,24]. Examination of the remaining species of Heraclescolex and their inclusion in the molecular phylogenetic analysis could recover them within Allolobophora, which would further support our hypothesis.
Interestingly, the genotypes of both merged genera (All. chlorotica and All. moebii) constitute species complexes on opposite poles of the morphological variability spectrum.
Allolobophora chlorotica has been shown to comprise several cryptic lineages [39,40], some of which are widespread or even cosmopolitan, while a few are endemic to Provence and the Alps [39,41]. Moreover, several subspecies have been described, but their association with those genetic lineages is still unknown. The integration of both sources of information with the new phylogenetic frame is essential to understand their phylogeography and the geographic origin of the genus.
The molleri-moebii species complex is a particularly controversial group that has been assigned to the genera Allolobophora, Eophila, Aporrectodea, or Heraclescolex. Trigo et al. [37,38] found specimens with intermediate character states with respect to the position and extension of the clitellum in a continuum between the previously described species, which blurred the boundary between them. Barros [36] concluded that All. molleri and All. moebii constitute a single species named All. molleri. However, these results were not accepted by Qiu and Bouché [8,13,14,16], who described additional subspecies and closely related species (the Maghrebian Heraclescolex kionionus Bouché, 1998, Heraclescolex rifanus Qiu and, and Heraclescolex postsellis . Our results appear to support the former view, as All. molleri and All. moebii do not have sufficient genetic divergence to be considered separate species. In order to solve the decade-old riddle of the taxonomy of this species complex, a systematic sampling of different populations of All. molleri, All. moebii, and the other species within this complex (Allolobophora fernandae Graff, 1957, Allolobophora monchicana Trigo, Mascates, Briones and Diaz Cosin, 1990, and Allolobophora opisthosellata Graff, 1961) will be performed in the near future; this will allow integrative species delimitation and will shed light on the evolutionary significance of the extraordinary clitellar variability of this group.
Although several questions remain to be solved, our current knowledge allows us to propose a rough scenario for the evolutionary origin and historical biogeography of Allolobophora. All the species assigned to the newly defined Allolobophora (plus the currently unstudied Heraclescolex species) are native to the western Iberian Peninsula, France, Switzerland, Italy, and central-eastern Europe. The center of origin could be near the western Alps, as suggested by the presence of "endemic" lineages of All. chlorotica as well as All. burgondiae and All. icterica (Figure 2). The less intuitive distribution corresponds to the All. moebii-molleri complex and Heraclescolex species, which show a remarkable disjunction. However, the highest diversity of species and subspecies in southwest of the Iberian Peninsula, Morocco, and Algeria suggests an origin of this species group in the AlKaPeCa (Alboran-Kabylian-Peloritan-Calabrian) terrane, whose fragments migrated to their current positions in the Iberian Peninsula and northern Africa [42]. This terrane showed continuity with alpine Corsica and the developing Alps, allowing the dispersal and diversification of the ancestors of the extant Allolobophora species to their current range. An increased taxonomic and phylogeographic sampling of the areas and species involved would allow to refine and test this hypothesis, offering valuable insights into the link between the paleogeography of the western Mediterranean and its diversity.
The small, isolated genera Heraultia and Vosgesia and the occurrence of Panoniona in northeastern France and Switzerland also provide interesting biogeographic and evolutionary insights (Figure 2). Heraultia tiginosa represents a phylogenetic and biogeographic link between the French Zophoscolex and Ethnodrilus and the mostly Pyrenean Prosellodrilus and Cataladrilus. While the latter two are remarkably diverse genera, Heraultia comprises a single species known from a single location: this suggests that Heraultia is a relict genus whose diversity and range might have been reduced by environmental and faunistic changes (as inferred for Galiciandrilus, Compostelandrilus, and Vindoboscolex [15]). Although new species belonging to Heraultia might be discovered in the future extending its distribution area, its presence highlights the high conservation value of Herault (and the area around Montpellier) as a hotspot of Mediterranean earthworm diversity.
Similarly, V. zicsii appears as an isolated distant relative of Av. ligra and Gatesona and the most northeastern representative of their clade (Figure 1). Its presence attests to a wider distribution of such lineage in the past and putatively to a greater diversity that is no longer present in France. Panoniona satchelli is isolated from its eastern relative P. leoni (present in Italy, Slovenia, Croatia, and further east) by the Alps (and about 400 km in a straight line) ( Figure 2). In the case of Panoniona, the cause of such disjunction could indeed be the progressive elevation of the mountain chain, with an associated change in climate and vegetation, which could have led to a large unsuitable area between the current species. However, there is another cause that could explain the present range of Vosgesia zicsii and P. leoni. As stated by Bouché [4], endemic species in France appear to be restricted to the south of the Loire River: towards the North, they would have been eliminated by the presence of permafrost during the Last Glacial Maximum (LGM, ca. 33,000-15,000 years ago [43]). Even though most recent models suggest that permafrost would have not been present south of 49°N in France [44], the dominant 11 Journal of Zoological Systematics and Evolutionary Research biome in most of France and the rest of Europe would have been steppe-tundra [45] which is unsuitable for most lumbricid earthworms. The almost complete absence of endemic earthworm species (and low diversity) in Germany, Czech Republic, Poland, Slovakia, and countries further north supports this hypothesis. Nevertheless, Bouché [4] noted the presence of a few strictly endemic species in a small region (Val de Saône-Vosges) within the "glaciated" area. Allolobophora burgondiae, P. satchelli, V. zicsii, and Ap. pseudoantipai (as well as Ap. gogna Bouché, 1972 and Ap. velox (Bouché 1967)) occur there. Thus, the isolation of these species relative to their closest southern or eastern relatives would be explained by the disappearance of intermediate species during the LGM or other unsuitable climatic events. Interestingly, some models show that the aforementioned region roughly corresponds to the presumed distribution of different vegetation types (such as ombrotrophic bogs) during the LGM [46], which supports the role of the Val de Saône-Vosges area as a glacial refuge for endemic Lumbricidae. Further efforts should be devoted to biodiversity assessment and conservation of this unique hotspot, which has been generally neglected in favour of the southern regions.
The biogeographic inferences drawn from Allolobophora and the relict species from northeastern France illustrate the relevance of the Alpine region for the evolution of several lineages of Lumbricidae, with the effects of the LGM likely disrupting the remaining evidence. This is probably also true for other genera such as Aporrectodea sensu stricto, Lumbricus Linnaeus, 1758 or Octodrilus Omodeo, 1956. A greater focus on the implementation of phylogeography and molecular phylogenetics to the earthworm faunas of Switzerland, Italy, Austria, Slovenia, Croatia, and Serbia would greatly enhance our understanding of the evolution and historical biogeography of Lumbricidae.

Conclusions
Molecular phylogenetic analyses revealed that the representatives of Allolobophora sensu stricto and Heraclescolex belong to a clearly delimited clade, which suggest their synonymy. On the other hand, other species formerly assigned to Allolobophora were recovered within Panoniona, or as isolated, genus-level branches. Therefore, the revised definitions for Allolobophora and Panoniona were provided, and the new genera Heraultia and Vosgesia were described.