Early Miocene remains of Melissiodon from Mokrá-Quarry (Moravia, Czech Republic) shed light on the evolutionary history of the rare cricetid genus

Remarks on the systematics

The formal definition of Melissiodontidae at the family level was published by Schaub (1925), after the author earlier indicated that the clade deserved a family rank (Schaub, 1920). However, many studies hesitated to include the genus Melissiodon in the family Cricetidae (e.g., Freudenthal, Lacomba & Malagón, 1992; Kristkoiz, 1992) because of its aberrant skull features and the derived morphology of the cheek teeth. Therefore, this genus was the only member of the subfamily Melissiodontinae for most of the last century. Later, Ünay-Bayraktar (1989) erected the genus Edirnella based on a few isolated upper cheek teeth from Kocayarma MP25 Middle Oligocene locality in the Turkish Thrace Basin. This genus was mainly accommodated into the subfamily Melissiodontinae because it shares a cusp between the anterocone complex and the protocone with Melissiodon. This cusp is referred to as the Melissiodon cusp by Ünay-Bayraktar (1989) and Wessels et al. (2018) and as the protostyle by Jovells-Vaqué & Casanovas-Vilar (2018). Ünay-Bayraktar (1989) also grouped the Melissiodontinae subfamily containing Melissiodon and Edirnella and the Paracricetodontinae with its Paracricetodon and Trakymys into the family Melissiodontidae. This was strongly criticized and avoided in subsequent works by Freudenthal, Lacomba & Malagón (1992); Kristkoiz (1992) and Kalthoff (2006). New discovered remains from Süngülü Late Eocene deposits in the Lesser Caucasus of Turkey were referred to Edirnella and included in the Melissiodontinae subfamily (De Bruijn et al., 2003). Mogilia represents the third genus belonging in the subfamily Melissiodontinae, which contains the two M. miloshi and M. lautus species of the Eocene-Oligocene of Serbia (Wessels et al., 2018).

The Cricetopsinae subfamily was originally established by De Bruijn & von Koenigswald (1994), but McKenna & Bell (1997) amended it to Cricetopinae, and this is followed in our work. The phylogenetic analyses performed by Maridet & Ni (2013) recovered Melissiodon into Cricetopinae subfamily, being Melissiodon the sister species of Selenomys, and the two would constitute the sister group of Mirrabella. The monophyletic group including three genera was called Melissiodon clade. Maridet & Ni (2013) included Cricetops, Meteamys, and Selenomys genera into Cricetopinae. However, the most remarkable reallocation was that Edirnella was recovered into subfamily Pseudocricetodontinae, and consequently, more closely related to the genera Pseudocricetodon, Adelomyarion and Raricricetodon. This proposal was not followed by Wessels et al. (2018), as they placed all the three genera, i.e., Edirnella, Melissiodon and Mogilia, into subfamily Melissiodontinae mainly based on the enamel microstructure, a key character omitted by Maridet & Ni (2013) in their phylogenetic study. Despite the debate regarding its phylogenetic position, Melissiodontinae represent, together with Atavocricetodon and Pseudocricetodon, one of the first muroids that have been recorded outside of Asia, and yet, these are still poorly known.

Measurements: see Table 1

(Figs. 3A–3AB)

Studied material:

MWQ 2/2003: fragment of left maxilla with M1-M2 (Pal. 3381); fragment of right maxilla with M1-M2 (Pal. 3382); fragment of left maxilla with M1 (Pal. 3490); fragment of right maxilla with M1 (Pal. 3488); four left M1 (Pal. 3371, Pal. 3491, Pal. 3492, Pal. 3493); two right M1 (Pal. 3380, Pal. 3489); one left M2 (Pal. 3496); four right M2 (Pal. 3372, Pal. 3383, Pal. 3495); one left M3 (Pal. 3377); one left m1 (Pal. 3373); two right m1 (Pal. 3374, Pal. 3494); one left m2 (Pal. 3375); two right m2 (Pal. 3497, Pal. 3498); two left m3 (Pal. 3501, Pal. 3502); three right m3 (Pal. 3376, Pal. 3499, Pal. 3500).

MCQ 3/2005: one right M1 (Pal. 3931); one fragment left m1 (Pal. 3903); one left m2 (Pal. 3930).

Description

M1 (Figs. 3A–3K): The teeth have five cylindrical roots, with the two larger ones below the anterocone complex and the protocone. Two slimmer roots are below the metacone and the hypocone and the fifth root is conspicuously thinner and situated below the paracone. The anterocone complex is massive and protrudes anteriorly. It presents two cusps, the largest one placed in a centro-lingual position. Both cusps are separated by a deep and narrow notch, which is almost enclosed anteriorly in Pal. 3492 (Fig. 3I). The lingual anterocone is connected to the anterolophule. This cusp also has a lingual spur that merges with a distolabial crest running from the protostyle. Furthermore, it connects to the anterolophule, dividing the protosinus in 8 out of 11 teeth (Pal. 3491, Fig. 3H); in Pal 3492 (Fig. 3I), this lingual spur merges with the anterolophule before the anterolophule cusp, leaving the protostyle isolated and the protosinus complete; in Pal. 3489 (Fig. 3G), it ends before reaching the spur from the protostyle, and in Pal. 3382 (Fig. 3B), it reaches the anterolophule cusp without connecting to the protostyle. In addition, Pal. 3493 (Fig. 3J) has a large spur from the protostyle that reaches the labial spur of the lingual anterocone. The anterior spur of the protostyle connects to a spur running lingually from the lingual anterocone enclosing a small sinus in 7/11 teeth (Pal. 3371, Fig. 3E). This crest is specially developed in Pal. 3490 (Fig. 3C) and the rest of the specimens show a different degree of development of the spurs without fully enclosing the small sinus (i.e., Pal. 3491, Fig. 3H). The labial anterocone is also large: it connects to the lingual anterocone through a lingual spur and to the two anterior paracone spurs, enclosing the anterosinus. The anterior lingual paracone spur has a short crest running labially in Pal. 3380 (Fig. 3F). In Pal. 3489, 3491 and 3493 (Figs. 3G, 3H, 3J) there is a short crest running lingually towards a short labial spur from anterolophule but without connecting to it. A small anterostyle is present in Pal. 3381, 3382, 3371 and 3493 (Figs. 3A, 3B, 3E, 3J); it is very strong and isolated from the other cusps in Pal. 3492 (Fig. 3I). Pal. 3931 has a short transverse crest between the paracone and the labial anterocone (Fig. 3H). The protolophule is defined by a labial spur of the protocone, which is slightly longer than the lingual one of the paracone. The mesocone is well-developed and connected to the protolophule through a short anterior arm (10/11) or isolated from it (1/11; Pal. 3490, Fig. 3C). The posterior arm of the mesocone runs labially and connects to the entoloph, except for Pal. 3490 (Fig. 3C), in which seems to be isolated.

Although the morphology of the mesoloph is also variable, it always runs obliquely to the labial end of the tooth. It connects to the posterior arm of the mesocone, and in three specimens, it is long enough to reach the anterior arm of the mesocone (Figs. 3E, 3F, 3G). The entoloph runs labially and joins the metalophule. The mesoloph area is highly variable. In Pal. 3382 and 3488 there is a single connection to the mesoloph, but closer to the mesocone (Figs. 3B, 3D). In Pal. 3488, the extra metalophule spur runs distolabially and connects to the metacone (Fig. 3D); however, this crest is much shorter in Pal. 3491 and Pal. 3931 (Figs. 3H, 3K) in anteroposterior direction. The remaining teeth (7/11) have an extra metalophule spur that is joined to the mesoloph, enclosing a small pit that can be narrow (Pal. 3381, Fig. 3A) or wide (Pal. 3371, Fig. 3E). Pal. 3382 (Fig. 3B) is the only tooth that does not show an extra metalophule spur. Both the paracone and the metacone show labial spurs that connect with the mesoloph. This connection protrudes labially and is specially marked in Pal. 3382 (Fig. 3B). The sinus is partially open. The posteroloph is low and connects to the posterior spur of the metacone. In 7/11 teeth it runs labially surpassing the posterior spur of the metacone (Fig. 3F). In 9/11 teeth a hypoconule is present (Fig. 3F). All specimens show a short lingual posteroloph from the hypocone.

M2 (Figs. 3L–3O): All molars have a rectangular outline. The anterolabial is the thinnest of its four roots. The lingual anteroloph is short but very robust and anterocone-like in 6/7 teeth, and very thin in Pal 3383 (Fig. 3M). The anterolophule reaches the anterior lingual spur of the paracone. There is a distinct anterolophule cusp in half the molars. Additionally, Pal. 3381 shows a small bump where the anterolophule meets the anterior lingual spur of the paracone (Fig. 3A). The labial anteroloph is connected to the labial anterior spur of the paracone in 6/7 teeth, but Pal. 3381 shows an incomplete labial anteroloph, leaving the anterosinus open (Fig. 3A). The protolophule is straight. The posterior lingual spur of the paracone is long in all teeth except for Pal. 3381 (Fig. 3A). It reaches the protolophule (3/6; Fig. 3L) or ends immediately before it (3/6; Fig. 3N). The protolophule merges with the mesoloph in Pal. 3381, 3382, and 3495 (Figs. 3A, 3B, 3N). The mesoloph and mesocone show high variability within the assemblage of Mokrá-Quarry. The mesocone is round. The anterior arm of the mesocone turns anteriorly and connects to the protolophule in two specimens (Figs. 3L, 3O). It runs labially and connects to the mesoloph in Pal. 3383 (Fig. 3M). However, in Pal. 3381 and 3495 the anterior arm of the mesocone runs labially and splits in two, connecting to the protolophule and to the mesoloph (Figs. 3A, 3N). A further deviation is the reduction of the anterior arm of the mesocone to a short spur that does not reach the mesoloph or the protolophule. This, however, is apparent only in Pal. 3382 (Fig. 3B). The posterior arm of the mesocone connects to the entoloph in all specimens except for Pal. 3372 where it is connected to the metalophule (Fig. 3L). Pal. 3495-3496 (Figs. 3N, 3O) highlight the extra spur from the posterior arm of the mesocone which reaches the mesoloph. This extra spur has a double connection to the entoloph and does not fully connect to the mesoloph in Pal. 3381 (Fig. 3A). The metalophule is always connected to the mesoloph. A mesostyle is absent and the ectoloph connects to the metalophule (Fig. 3L). The sinus is open in all teeth and the posteroloph reaches the posterior spur of the metacone. All teeth show a lingual posteroloph running from the hypocone (Fig. 3M).

M3 (Fig. 3P): This molar has two labial roots and a thicker lingual root. It is narrow and has a round outline. The labial anteroloph is thin and does not reach the paracone labial (Fig. 3P). There is a preserved very low lingual anteroloph that does not reach the base of the protocone, and both protocone and paracone are high and well-developed. The anterolophule is straight and reaches the anterior lingual spur of the paracone. The protolophule is fused with the metalophule, which runs from a very small metacone. The lingual paracone spur does not reach the protolophule. The entoloph is somewhat short and therefore does not reach the metalophule. The posterior arm of the hypocone connects with the metacone, and all the sinuses are open.

m1 (Figs. 3Q–3T): Both lingual and labial anteroconids are well-developed and separated by a deep anterior groove which is clearly visible in Pal. 3373 (Fig. 3Q). Despite its fragmentary preservation, Pal. 3494 has a short cristid running lingually from the labial anteroconid towards the lingual one, without fully attaching to it (Fig. 3S). The lingual anteroconid is massive and located slightly posterior to the labial anteroconid. It is connected to the metaconid through a short posterior spur. The metaconid lacks a well-developed anterolingual cristid, which leaves the anterosinusid open. The anterolophulid connects to both anteroconids and to the protoconid in all teeth. The anterolophulid connects to the labial anteroconid posteriorly. The anterolophulid cuspid is large and there is a short spur towards the metaconid, but without connecting to it. The labial anteroconid is large and bulge-like, and protrudes anteriorly. It is isolated from the protoconid as the protosinusid is open. The protoconid is large and its hind arm connects to the labial posterior spur from the metaconid, and reaches the lingual edge of the tooth (Fig. 3R). The anterior arm of the mesoconid connects to the protoconid hind arm in all specimens, showing a small bulge in Pal. 3374 (Fig. 3R). The mesolophid always starts at the anterior arm of the mesoconid, showing different degrees of development: in Pal. 3373 it is divided and does not reach the anterior arm of the entoconid (Fig. 3Q); in Pal. 3374 it turns slightly anterior, connecting to the protoconid hind arm and enclosing a small pit (Fig. 3R); in Pal. 3494 (Fig. 3S) it runs parallel to the protoconid hind arm towards the labial anterior spur of entoconid; in Pal 3903, the anterior arm of the mesoconid appears to run labially and the mesolophid is reduced to a short spur that ends next to the anterior arm of the entoconid without attaching to it (Fig. 3T). The ectolophid is short and attached to the mesoconid. The entoconid has the labial cristids much better developed than the lingual ones, which are almost absent. The labial posterior cristid reaches the posterolophid. There is a well-developed posteriorly oriented spur running from the posterolophid (Fig. 3S). The sinusid is open labially. A well-developed ectomesolophid that almost reaches the mesoconid is present in Pal 3494 (Fig. 3S). However, in Pal 3903 there is a small bump instead (Fig. 3T).

m2 (Figs. 3U–3X): Pal. 3498 is damaged anteriorly and lingually (Fig. 3W). The four teeth preserve a very short labial anterolophulid. The protosinusid is open, and the lingual anterolophid is well-developed and connects to an anterior labial spur running from the metaconid. There is no anterior lingual cristid from the metaconid, and this leaves the anterosinusid open. The protoconid hind arm is long and runs distolingually reaching the labial edge. The posterior spurs of the metaconid and the anterior spur from the entoconid attach to the protoconid hind arm. The mesoconid is well-developed and its short anterior arm runs anterolabially to connect to the protoconid hind arm. The mesolophid preserves different configurations: in Pal. 3375, a very short spurs starts from the protoconid hind arm but does not reach the mesoconid (Fig. 3U); in Pal. 3497, a short mesolophid starts from the centre of the mesoconid and merges with the protoconid hind arm (Fig. 3V); in Pal. 3498, a long mesolophid, starting from the centre of the mesoconid runs parallel to the protoconid hind arm and connects to the anterior labial spur of the entoconid (Fig. 3W); in Pal. 3930 (Fig. 3X), the short mesolophid merges with the protoconid hind arm as seen in Pal. 3497. The ectolophid is short and preserves a short spur in 3375 (Fig. 3U), but it is long in Pal. 3390 (Fig. 3X). The posterolophid is very low and connects to the two posterior spurs of the entoconid. The sinusid is open.

m3 (Figs. 3Y–3AB): The molars are elongated with a rounded posterior side. There are two roots, with the posterior root much broader than the anterior. The protoconid connects to an anterior labial spur which runs from the metaconid through a well-developed anterolophulid that runs straight. Although some teeth, such as Pal. 3376 and 3502, have a thickening on the anterolophulid there is no distinct anteroconid (Figs. 3Z, 3AB). The protosinus is open, and the protoconid hind arm reaches the labial edge and connects to the lingual posterior spur of the metaconid and the anterior one from the entoconid. In addition, Pal. 3376 and 3501 also have a posterior labial spur of the metaconid (Figs. 3Z, 3AA). The most variability occurs in the mesoconid, as in m2. While Pal. 3376 and 3502 present a small but distinct mesoconid (Figs. 3Z, 3AB), the remainder have no distinguishable bulge. The hypolophulid is short and directed distolingually (Fig. 3Y). Pal. 3501 has a hypolophulid divided into two very short spurs (Fig. 3AA). The posterolophid becomes progressively lower and attaches to the base of the entoconid. There is no posterior spur running from this cuspid. The sinusid is open.

Evolutionary history of Melissiodon during the Early Miocene

The first phylogenetic hypothesis of the Melissiodon genus was proposed by Hrubesch (1957: fig. 125), in which he suggested the existence of various lineages during the Oligocene, while during the Early Miocene only two branches persisted. Although he also considered Melissiodon arambourgi a completely different lineage, the species has been proven synonymous with M. dominans (Agustí, 1981; Jovells-Vaqué & Casanovas-Vilar, 2018). Melissiodon dominans became the most abundant species during the Early Miocene MN3 and MN4 localities and reached the Burdigalian before its extinction during the latest Early Miocene. The second lineage present in the Early Miocene led to M. schlosseri. The recovered Haslach material was quite sparse, but Hrubesch (1957) considered it different to any other species of the genus. He based this on the specialization of the m1 anterior half, and especially on its stronger cuspids and well-developed bulge-like anteroconids. These differences were enough for him to suggest a different lineage with no connection to M. dominans. Since only Haslach’s discovery of these remains existed, this species was considered a completely isolated branch in European localities, with unknown origin and no continuation throughout the Early Miocene.

Werner (1994) later described a partial left mandible with m1-m3 from Ulm-Uniklinik (MN2), a slightly younger locality than Haslach, and referred it to Melissiodon aff. schlosseri. The author indicated a small difference in the labial anteroconid size. Although it was large, it was only somewhat smaller than the type species. Since the type species included only one m1, the smaller size could be explained by intraspecific variability.

Engesser & Mödden (1997) then reported two different Melissiodon taxa from La Chaux 7, which is an MN2 Switzerland site with similar age to Haslach. The remains were attributed to M. aff. schlosseri and M. cf. dominans, but no descriptions or measurements of either taxon were provided. La Chaux 7 is the only current site where two different taxa of the genus have been recorded, and no other remains of Melissiodon related to M. schlosseri or this lineage of the genus (sensu Hrubesch, 1957) have been discovered at MN3 or MN4 localities.

The Mokrá-Quarry Melissiodon aff. schlosseri has highly developed m1 lingual and labial anteroconids, and it has greater development than M. dominans from the MN3 type locality. It is also arranged differently than in M. dominans from other MN4 Central European localities. In addition, the poorly known Melissiodon schlosseri has m1 anteroconids that resemble the size and development of Mokrá-Quarry M. aff. schlosseri.

The overall differences between the Mokrá-Quarry assemblage and M. dominans (see Discussion section above) imply that this material could be a younger species derived from M. schlosseri. However, our M. schlosseri knowledge depends on a very few teeth, and we consider that type material revision is required before it is possible to recognize a new species.

Although the high diachrony between European regions prevents a more straightforward correlation (e.g., Steininger, 1999; Gozhyk et al., 2015), the following figure herein depicts the biochronologic distribution of Melissiodon taxa during the Early Miocene (Fig. 5). If the Mokrá-Quarry Melissiodon aff. schlosseri is closely related to M. schlosseri, this would imply that the lineage survived in MN4 localities. However, the evolution and distribution of this lineage at MN3 localities still remains unclear.

The single m1 from the MN3 Sant Andreu de la Barca has a similar anteroconid developmental pattern (Jovells-Vaqué & Casanovas-Vilar, 2018: fig. 2c), despite being larger than M. aff. schlosseri and M. schlosseri (see Fig. 4). The species-level taxonomical assignment of this single molar in the evolution of the genus is still uncertain, but it could indicate the presence of more taxa at MN3 localities besides M. dominans.

The m1s in MN4 Melissiodon assemblages are also smaller than in the type population. In addition to being smaller, the Melissiodon population from the German sites have a completely different m1 anteroconid area arrangement (Ziegler & Falhbusch, 1986), as noted in the Discussion section. The size increase in the anteroconid area is a noted common trend in the M. dominans lineage (Hrubesch, 1957). That, together with the complete lack of a labial anteroconid could be indicative of a different taxon. In addition, our recent inspection of the material from Dolnice and Ořechov (IBS 2021, personal observation) attributed to M. dominans by Fejfar (1990) reveals two different m1 morphotypes in those MN4 populations. These were the Mokrá-Quarry Melissiodon aff. schlosseri cuspid-like labial anteroconid and the German sites’ M. dominans cristid-like labial anteroconid. There was also clear predominance of the latter morphotype.

It is noteworthy that the m1 morphotype in the Mokrá-Quarry and other Czech localities is not restricted to Central Europe or MN4 sites. Mein & Freudenthal (1981) recovered a single m1 from Vieux-Collonges (France; MN5) which clearly resembles the Mokrá-Quarry M. aff. schlosseri. The authors considered the morphological differences in Melissiodon assemblages to be sufficiently distinct. Despite the insuficient material prevented them from erecting a new species, the morphological variability of the genus during the latest Early Miocene, prior to its extinction appears to be higher than previously acknowledged.

Besides, the alpha-taxonomy of Melissiodon populations in MN4 sites is complex, and the lineage leading to Melissiodon dominans remains poorly known. The existence of a tendency towards size increase between the type locality (Wintershof-West) and younger MN3 sites has been confirmed by several publications: Ramblar, Bañon and Turó de les Forques 1 sites in Spain (Sesé, 1987; Jovells-Vaqué & Casanovas-Vilar, 2018) and Beaulieu 2B and Jauquet in France (Aguilar et al., 2003; Bulot, Antoine & Duranthon, 2009). However, M. dominans from MN4 assemblages are usually smaller than those of the type locality (Fig. 4), which has also been noticed by other authors (Ziegler & Falhbusch, 1986; Bulot & Ginsburg, 1996; Bulot, Antoine & Duranthon, 2009). Overall, there are only a few exceptions in MN4 localities where Melissiodon dominans is larger than in other coeval sites. One of them is Montalvos 2, where M. dominans is larger than those from most MN4 localities where Democricetodon and Megacricetodon are also found (Hordijk et al., 2015). While these Montalvos 2 remains accord with the Wintershof-West M. dominans size-range, their m1 morphotype is similar to those of the Forsthart and Rembach MN4 sites. In addition, while the Dolnice 1 site contains two m1s similar to Montalvos 2 size, and are clearly larger than the remaining assemblage, the m1s have a cuspid-like labial anteroconid rather than the cristid-like configuration found in Montalvos 2.

The above evidence confirms Melissiodon dominans’ complex evolution, and therefore, the coexistence of two different lineages in the latest Early Miocene cannot be discounted. These comprise one that slightly increases in size throughout the MN3 (Aguilar et al., 2003); and a smaller lineage that replaced the larger one in most Central and Western Europe MN4 localities. In any case, the assessment of whether these represent different Melissiodon species or different populations within M.dominans is beyond the scope of this work. The small M. dominans populations and the scarcity of European fossil localities in the Early Miocene have made Melissiodon dominans into a catch-all species. Therefore, the complete revision of all Miocene localities with Melissiodon remains should elucidate the relationships between the different forms and species that existed in the European Early Miocene.

Paleoecological implications

Historically, the presence of Melissiodon has been linked to wooden areas and somewhat humid conditions (Van der Weerd & Daams, 1978). According to Sesé (1987), Melissiodon enters the Iberian Peninsula during a change from a relatively dry to a relatively wet climate (zone Z; Daams & Van der Meulen, 1984). In addition, it is no longer recorded in the Calamocha area beyond zone A, when the climate is presumed to have regained aridity.

The Mokrá-Quarry sites are characterized by its karstic conditions, together with patches of forest and swampy areas (Ivanov, 2008; Ivanov, Musil & Brzobohatý, 2006; Ivanov et al., 2017; Ivanov et al., 2020; Sabol et al., 2007; Luján et al., 2017; Luján et al., 2021; Bonilla-Salomón et al., 2021a). This concurs with conditions that presumably favored Melissiodon presence. Jovells-Vaqué & Casanovas-Vilar (2018) add that rich Melissiodon collections from the Vallès-Penedès Basin come from localities with diverse and abundant forest-dwelling rodents, such as dormice and tree-squirrels. We concur with this because several genera of sciurids adapted to arboreal lifestyle have also been recovered in the Mokrá-Quarry (Bonilla-Salomón et al., 2021b).

The existence of different Melissiodon lineages in the Early Miocene could also imply preference for specific palaeoecological conditions. However, there is no known locality during the Early Miocene where two different species have certainly been proven. La Chaux 7 (Switzerland, MN2), provides the only record of Melissiodon aff. schlosseri and M cf. dominans (Engesser & Mödden, 1997). However, without detailed descriptions, these attributions remain dubious. The high degree of specialization in the first upper and lower molars and clear difference between morphotypes are most likely related to slightly different feeding behaviors and consequent preference for those specific habitats.

Mein & Freudenthal (1981) suggested frugivorous diet for Melissiodon and ventured that the genus may be arboreal. However, Hordijk et al. (2015) hypothesized that Melissiodon was a ground-dweller genus that fed on invertebrates (mainly earthworms). They based this assumption on the similarity of teeth morphology to extant shrew rats from Sulawesi and the Philippines. This assumption was also supported by Wessels et al. (2018), suggesting that subfamily Melissiodontinae (sensu Wessels et al., 2018) would feed on small invertebrates. However, the main obstacle to interpreting feeding behavior and paleoecological preferences is the paucity of Melissiodon localities.