Introduction

Palaeogene marsupials from Europe and North America were initially thought to be closely related to opossums and considered to represent the ancestral stock for the didelphids, and classified in the Didelphidae Gray, 1821 (i.e. von Koenigswald 1970; Crochet 1979, 1980a). Phylogenetic analysis by Sánchez-Villagra et al. (2007a), including skeletal data, removes Herpetotherium Cope, 1873 from the Didelphidae, and Herpetotherium is now classified in the Herpetotheriidae Trouessart, 1879 (originally as Herpetoherinae). The taxonomy above family level also remains subject to change. There is no consensus in the literature and mentioned are Didelphimorphia Gill, 1872 (Crochet et al. 2007), ‘Didelphimorphia’ (Hooker et al. 2008), Marsupialiformes (Vullo et al. 2009; Ladevèze et al. 2012; Selva and Ladevèze 2016), Marsupialia (Horovitz et al. 2008, 2009, Wilson et al. 2016) and Marsupialiformes incertae sedis (Eldridge et al. 2019). We use the clade Marsupialiformes as defined by Vullo et al. (2009) and consider the Herpetotheriidae closely related to the Marsupialia Illiger, 1811 (Sánchez-Villagra et al. 2007a). All these classifications are based on morphology of dentition and jaws. Ladevèze et al. (2020) included the anatomy of the petrosal bone and divided the Herpetotheriidae in the subfamilies Peratheriinae (including Peratherium Aymard, 1850 and Amphiperatherium Filhol, 1879) and Herpetotheriinae (including all American herpetotheriid genera). As in Ladevèze et al. (2012) the informal name marsupial is used here to refer to any metatherian taxon.

Herpetotheriids are known from North America since the Late Cretaceous by many genera, and the dental similarity of the American and European marsupials suggests at least two dispersals from North America to Western Europe, one in the Late Cretaceous (Martin et al. 2005), and one in or before the early Eocene. Hypotheses on origin and dispersal routes of marsupials are discussed by many authors, such as in the extensive reviews of Vullo et al. (2009) and Crespo and Goin (2021).

Marsupials known from Cenozoic fossils in Europe are Peratherium, Amphiperatherium and Peradectes Matthew and Granger, 1921. The first two genera are classified in the Herpetotheriidae and the latter in the Peradectidae Crochet, 1979 (Peradectinae in McKenna and Bell 1997). Peradectidae are represented in the Eocene of Europe by three Peradectes species and are reported from the Miocene of Thailand (Ducrocq et al. 1992, Mein and Ginsburg 1997) and China (Storch and Qiu 2002). Herpetotheriidae are represented by twelve species of Peratherium in Eocene and Oligocene assemblages, and nine species of Amphiperatherium in Eocene to Early Miocene faunas. They are well known from sites in Belgium, Bosnia and Herzegovina, Czech Republic, France, Germany, United Kingdom, Portugal, and Switzerland (i.e. Crochet 1979, 1980a, 1988, 1995, Fejfar 1987, Hooker 1986, 2010, Smith 2003, Ziegler and Storch 2008, van der Sar et al. 2017, Hooker and Weidmann 2000). Amphiperatherium frequens (v. Meyer, 1846), the last occurring species in Europe, is known from many Early Miocene localities (Czech Republic, France, Germany, Spain; Ziegler 1999, Klietmann 2013, Crespo et al. 2020).

A few isolated occurrences of marsupials are known from the early Eocene of Turkey (Kappelman et al. 1996, Maga and Beck 2017, Métais et al. 2018), the early Eocene of Tunisia (Crochet 1986), the early Oligocene of Egypt and Oman (Bown and Simons 1984, Crochet et al. 1992, Sánchez-Villagra et al. 2007b, Hooker et al. 2008) and the early Miocene of Uganda (Crespo and Goin 2021). Most of these marsupials are morphologically rather different from the European species and are considered to represent different lineages. Only Peratherium africanum Bown and Simons, 1984 from Egypt and Oman and Kasserinotherium tunisense Crochet, 1986 from Tunesia seem to be related to European marsupials (Hooker et al. 2008). Asiadidelphis Gabunia, Shevyreva and Gabunia, 1984 is known by several species from the late Eocene and early Oligocene of Kazakhstan, the early Oligocene of Mongolia, and the Oligocene of Pakistan (Ziegler et al. 2006, Crochet et al. 2007). These species show some morphological similarities with Amphiperatherium minutum (Crochet et al. 2007), but whether these species are related to North American Herpetotheriinae or to European Peratheriinae is not clear.

Most European marsupial findings are restricted to Western and Central Europe, thus the occurrence of marsupial species in the Palaeogene of south-eastern Serbia enlarges the geographical distribution of this group south- and eastwards, and the concept of faunal exchange between Europe, southern Asia, and northern Africa becomes more likely.

This paper is an addition to the eight papers on the Paleogene rodent faunas from the Pčinja and Babušnica-Koritnica basins of South-east Serbia. The geological setting and composition of the faunas from south-eastern Serbia (consisting of rodents) have been described by de Bruijn et al. (2017). Marsupials, insectivores, bats, small artiodactyls and carnivorans, are relatively rare in the assemblages. The rodents in these assemblages are quite different from those of Western and Central Europe, containing many Asian elements, such as dipodid and a diatomyid species. Only the murid genera Paracricetodon Schaub, 1925, Pseudocricetodon Thaler, 1969 and Heterocricetodon Schaub, 1925 occur in both areas during the early Oligocene (van de Weerd et al. 2018, Marković et al. 2019). The Serbian rodent assemblages are divers and show a clear evolution through time, interpreted as ‘local evolution’ and immigration events. The assumed age of these assemblages is based on the evolutionary level of the rodents, so far other age indications are lacking. The rodent association of Buštranje is quite different and is considered late Eocene (de Bruijn et al. 2017, 2018). The faunas from Strelac-1, Strelac-2, Strelac-3, Valniš-1 and Raljin-B are early Oligocene in age (de Bruijn et al. 2017, 2018; Marković et al. 2017, 2019; van de Weerd et al. 2018; Wessels et al. 2018, 2020).

Material and Methods

The collecting and treatment of sediment and fossils of seven localities is described in de Bruijn et al. (2017), since that time mammals have been collected at two more sites (Raljin-B and Valniš-2). Six of the nine localities from the Pčinja and Babušnica-Koritnica basins in south-east Serbia contain marsupial remains. Marsupials have not yet been found in Zvonce, the oldest assemblage, Raljin-A and Valniš-2, and only one specimen is known from Strelac-2. Jaws and jaw-fragments are not preserved, only isolated dental elements are available. In total 168 specimens (3 in Buštranje; 15 in Strelac-1; 1 in Strelac-2; 4 in Strelac-3; 132 in Valniš-1 and 13 in Raljin-B). Many teeth are incomplete or heavily worn, 104 are sufficiently complete to be used for measurements.

For a description and location of the sites we refer to de Bruijn et al. (2017). The Raljin-B site is in the same location as Raljin-A, but stratigraphically two meters higher. Valniš-1 was originally described as Valniš, but due to the discovery of other sites in the village of Valniš it is renamed Valniš-1. The new site of Valniš-2 is located one kilometre west of Valniš-1.

The fossils are stored in the Natural History Museum in Belgrade, where a numerical code is used to identify the localities: 024 for Strelac-1 (STR1), 025 for Strelac-2 (STR2), 026 for Strelac-3 (STR3), 027 for Valniš-1 (VA1), 031 for Buštranje (BUS) and 041 for Raljin-B (RA-B).

Nomenclature

The dental nomenclature in Fig. 1 follows Crochet (1980a), who based it on Van Valen (1966) and, Simpson (1929). Following Williamson et al. (2014) we have added: preparacrista, postparacrista, premetacrista and postmetacrista (subdivision of the centrocrista), and the paracristid is subdivided in postparacristid and preprotocristid. The postmetacristid and preentocristid are on the labial edge of the molars between metaconid and the entoconid. The stylar shelf is the labial part of the molar, including the stylar cusps (StA, StB, StC, StD, StE), the area labial of the paracone, metacone, preparacrista, postparacrista, premetacrista and postmetacrista. The metastylar area is the most posterior portion of this surface, including StE. The metaconule can have a postmetaconule crista and premetaconule crista, the paraconule a postparaconule crista and preparaconule crista. In a few upper molars, a narrow and low transverse crista or a minute cusp is present on the labial side of paracone or metacone.

Fig. 1
figure 1

The nomenclature of cusps and crests of the upper and lower molars used in the descriptions of the marsupials

Tooth position

Only isolated dental elements are available and our allocation to a specific tooth position is not always certain due to the large variability in size and morphology of the dental elements. Identification of the tooth position is based on characteristics described by Crochet (1980a) and on differences in size. Characteristic for the M1 is the relative short preparacrista, a more rounded base of the protocone, and a well pronounced stylar shelf, which extends posteriorly. The M3, which is more symmetrical than the M2, has a more pronounced ectoflexus. The DP3 is narrower than the M1 and has a low-lying protocone. The M1-3 are similar in morphology and size and are thus difficult to assign to a certain tooth position. The M4 is short and its talonid is narrower than the trigonid. To distinguish the M1, a combination of characteristics is used: the prefossid in M1 is wider than in M2 and M3; the paracristid is more straight in the M1 than in the other molars, where it is slightly bent towards the lingual edge. The M1 is shorter and slightly narrower compared to M2 and M3, and M3 is distinguished by the slightly narrower talonid.

Measurements

All measurements are in millimetres (mm). Figure 2a illustrates the followed method of measurement (used by e.g. Crochet 1979, 1980a, Ziegler 1990, Ladevèze et al. 2012), it is based on Clemens (1966). The orientation of the main axis of the upper molars is defined by the apices of the paracone and metacone and for the lower molars, the main axis is defined by the apices of the paraconid and entoconid. Length is the largest dimension parallel to these axes, width the greatest dimension measured perpendicular to it. However, in worn upper specimens the reference axis is difficult to determine, and we use the antero-lingual base of paracone and metacone as axis for orientation.

Fig. 2
figure 2

Measurement methods of the upper and lower marsupial molars. a The measurement method of Clemens (1966). b Our modified measurement method of von Koenigswald (1970)

Figure 2b shows our modified version of the measuring method of von Koenigswald (1970) of late Oligocene and Miocene marsupials from Europe. The length of the upper molars is the measured along the labial edge, and width is measured along the anterior edge of the molar (von Koenigswald measured over the StB and protocone), length and width are often not at right angles to each other. For the lower molars, the two measuring methods are the same. The length of the lower molars is measured along the lingual side of the molar, and the width of trigonid and talonid is measured perpendicular to it. The modified von Koenigswald method has been used by us to determine the tooth position of incomplete upper molars, such as labial fragments.

The outline of the upper molars is triangular, and each dental element is characterised by a different L/W ratio. The molars become shorter but broader from M1 to M4, and this is more pronounced when using the modified method. This clearly shows the change in shape: the L/W ratio decreases towards the posterior molars (Table 1 a). The measurements between quotation marks in the text indicate specimens that are almost complete, either worn, damaged or eroded. Such measurements are not used for the scatters diagrams and not used in Tables 2 and 3.

Table 1 Length and width ratios for the upper and lower molars of Amphiperatherium minutum (Aymard, 1846) and Amphiperatherium exile (Gervais, 1848–52) from Serbia. The results are based on the measuring method of Clemens (1966; depicted in Crochet 1980a) and our modified measuring method of von Koenigswald (1970). a Length-Width ratios for the upper molars of A. minutum from Valniš-1, and A. exile from Valniš-1, Strelac-1 and Strelac-3. b Length-Width ratios for the lower molars of Amphiperatherium minutum and Amphiperatherium exile from Valniš-1 and the differences in width of trigonid and talonid

Abbreviations

BE - Belgium, FR- France, GER- Germany, SW-Switzerland, UK-United Kingdom; UU - Utrecht University; Sin - sinistral, Dex - dextral; L - length, W - Width; N - Number, SE - Standard Error, SD - Standard Deviation

Systematic Palaeontology

Class Mammalia Linnaeus, 1758

Subclass Theria Parker and Haswell, 1897

Order Metatheria Huxley, 1880

Clade Marsupialiformes Vullo, Gheerbrant, de Muizon and Néraudeau, 2009

Family Herpetotheriidae Trouessart, 1879

Subfamily Peratheriinae Ladevèze, Selva and de Muizon, 2020

European genera included in the family Herpetotheriidae are Peratherium and Amphiperatherium. Both genera are included by Ladevèze et al. (2020) in the subfamily Peratheriinae, to distinguish them from other herpetotheriids. Ladevèze et al. (2020) state that the most important features to separate these two genera is the morphology of the petrosal bone. Petrosal bones, however, are rarely preserved, and usually only dental remains are available. The main difference in dentition between Peratherium and Amphiperatherium is the relative enlargement of molars from M1 to M3 in Peratherium (Crochet 1980a; Hooker et al. 2008; Ladevèze et al. 2012), Peratherium has a distinct dilambdodont pattern with upper molars widening from M1 to M3, cusps generally slender, StB dominates the stylar shelf, StC is generally present but often incorporated with the crista posterior to StB, and paracone and metacone about the same height. The protoconid increases in height from M1 to M4, entoconid and hypoconulid are about the same size and M4 has a reduced talonid (Crochet 1980a). In Peratherium the postmetacrista is inconspicuous (Ladevèze et al. 2020). In Amphiperatherium the cusps are massive, with metacone much higher than paracone, and a dominating stylar cusp is absent. The crista connecting the StC to StD, tends to disappear on the M1 and M2 (with StB and StD of sub-equal size), or tends to be reduced together with the StD on the M3 (StB then becomes dominant) (Crochet 1980a).

Amphiperatherium Filhol, 1879

Type species: Amphiperatherium frequens (v. Meyer, 1846)

Type locality: Weisenau (Germany)

Type level: Early Miocene (MN1)

Included species: Amphiperatherium minutum (Aymard, 1846); Amphiperatherium exile (Gervais, 1848-1852); Amphiperatherium lamandi (Filhol, 1876); Amphiperatherium ambiguum (Filhol, 1877); Amphiperatherium bastbergense Crochet, 1979; Amphiperatherium bourdellense Crochet, 1979; Amphiperatherium fontense Crochet, 1979; Amphiperatherium goethei Crochet, 1979; Amphiperatherium maximum Crochet, 1979

Amphiperatherium minutum (Aymard, 1846)

(Figs. 3, 4a-l, 6a-f, 8)

Type locality: Ronzon (France)

Type level: MP21

Geographical distribution: Belgium, Germany, France, United Kingdom and Spain; Serbia (this study)

Stratigraphical distribution: upper Eocene – lower Oligocene (MP16–MP25)

Emended diagnosis: Ladevèze et al. (2020) “Small-sized Amphiperatherium, with upper molars (M2 and M3) relatively larger than in A. giselense. Molars of A. minutum are characterised by: very slight size increase from M1 to M3, ectoflexus present but moderate and asymmetrical, stylar cusp C almost always present, stylar cusp B smaller than or subequal to the paracone, metastylar area enlarges on M2 and M3, M2 and M3 subequal in length, M4 of same length or slightly shorter than M3. Petrosals are characterised by: anterior extension of caudal tympanic process absent; mastoid process absent.”

Localities: Strelac-1, Strelac-2, Strelac-3, Valniš-1; Raljin-B (Serbia)

Age: early Oligocene (de Bruijn et al. 2017)

Material and measurements: In total 124 specimens (Figs. 3, 4a-l, 6a-f, 8); overview measurements in Table 2 and scatters in Figure 5.

Fig. 3
figure 3

Upper dentition of Amphiperatherium minutum (Aymard, 1846) from Strelac-3 and Valniš-1 (Serbia). a M4 dex (VA1-1054). b M3 dex (VA1-1034). c M2 dex (VA1-1022). d M1 dex (VA1-1023). e DP3 sin (VA1-1001); f M4 dex (VA1-1055). g M2 dex (STR3-1002). h M2 sin (VA1-1012). i M1 dex (VA1-1033). j M1 dex (VA1-1031). All specimens are figured as right ones, specimens indicated by an underlined letter are depicted in reverse. Rectangles in b and i, see Fig. 6

Fig. 4
figure 4

Lower dentition of Amphiperatherium minutum (Aymard, 1846) from Strelac-1 and Valniš-1 (Serbia). a M4 dex (VA1-1126). b M3 dex (VA1-1106). c M2 sin (VA1-1071). d M1 sin (VA1-1078). e M4 dex (VA1-1127). f M3 dex (VA1-1117). g M2 sin (VA1-1076). h M1 dex (VA1-1114). i M4 dex (VA1-1128). j M3 dex (VA1-1119). k M2 dex (STR1-1011). l M1 dex (VA1-1107). Peratherium sp. from Valniš-1. m M4 dex (VA1-1096). All specimens are figured as right ones, specimens indicated by an underlined letter are depicted in reverse. Rectangle in c, see Fig. 6

Table 2 Length and width measurements of Amphiperatherium minutum (Aymard, 1846) from Strelac-1, Strelac-3, Valniš-1 and Raljin-B (Serbia)

Strelac-1, 5 specimens: M2 dex STR1-1011 (1.59 x 1.00), STR1-1012 (— x 1.00); M3 sin STR1-1007 (1.72 x 1.11), STR1-1008 (1.69 x 1.09), fragment STR1-1009.

Strelac-2, 1 specimen: fragment, upper molar (STR2-1001).

Strelac-3, 1 specimen: M2 dex STR3-1002 (1.57 x 1.98).

Valniš-1, 108 specimens: DP3 sin VA1-1001 (1.27 x 1.07), VA1-1002 (1.16 x 1.18), VA1-1061 (1.33 x 1.35), VA1-1062 (1.18 x 1.33), fragment VA1-1003; M1 sin VA1-1004 (1.33 x —), VA1-1013 (1.62 x 1.64), VA1-1014 (1.59 x1.74), VA1-1015 (1.51 x 1.81), M1 sin VA1-1018 (1.60 x 1.98); VA1-1019 (1.59 x 1.60), VA1-1020 (1.48 x 1.66); M1 dex VA1-1023 (1.58 x 1.75), VA1-1031 (1.51 x 1.64), VA1-1033 (1.57 x 1.69), fragments VA1-1017, 1029, 1030; M2 sin VA1-1011 (1.59 x 1.75), VA1-1012 (1.46 x 1.92); M2 dex VA1-1022 (1.51 x 1.84), VA1-1032 (1.41 x 1.86), VA1-1036 (1.41 x —), VA1-1037 (1.52 x —), fragment VA1-1041; M3 sin VA1-1016 (1.39 x 2.01), M3 dex VA1-1021 (1.53 x 1.96), VA1-1034 (1.42 x 1.94), VA1-1036 (1.46 x 1.91), VA1-1038 (1.50 x —), VA1-1039 (1.481 x —); M4 sin VA1-1051 (1.04 x 1.43); M4 dex VA1-1054 (1.13 x 1.55), VA1-1055 (1.10 x 1.40), VA1-1056 (1.13 x 1.50), VA1-1059 (1.19 x —); M1 sin VA1-1078 (1.51 x 0.92), VA1-1080 (‘1.37’ x ‘0.88’), VA1-1084 (1.50 x 0.84), VA1-1087 (‘1.39’ x ‘0.83’), fragments VA1-1133, 1135, 1137, 1140, 1144, 1146; M1 dex VA1-1105 (1.48 x 0.91), VA1-1107 (1.59 x 0.87), VA1-1114 (1.61 x 0.90), fragments VA1-1149, 1152, 1158, 1163, 1168; M2 sin VA1-1071 (1.57 x 1.04), VA1-1073 (1.54 x 0.93), VA1-1074 (1.52 x 0.92), VA1-1075 (1.56 x 0.91), VA1-1076 (1.58 x 1.00), VA1-1082 (— x 0.97), VA1-1083 (1.62 x 0.98), VA1-1085 (— x 1.01), VA1-1094 (1.41 x 0.83), fragments VA1-1131, 1132, 1136, 1139, 1141, 1142, 1143, 1145, 1147; M2 dex VA1-1104 (1.52 x 0.88), VA1-1109 (1.51 x 0.89), VA1-1113 (1.52 x 0.95), VA1-1115 (1.53 x 1.00), VA1-1118 (‘1.42’ x ‘1.01’), fragments VA1-1148, 1150, 1151, 1153, 1154, 1155, 1156, 1157, 1159, 1160, 1161, 1165, 1170; M3 sin VA1-1072 (1.64 x 1.05), VA1-1086, (1.68 x 1.01); M3 dex VA1-1106 (1.56 x 1.03), VA1-1117 (1.60 x 1.05), VA1-1119 (1.61 x 1.07), VA1-1120 (1.56 x 1.08), fragments VA1-1164, 1166, 1167, 1173; M4 sin VA1-1091 (1.47 x 0.85), VA1-1092 (1.41 x 0.80), VA1-1093 (1.46 x 0.92); M4 dex VA1-1108 (1.48 x 0.78), VA1-1126 (1.44 x 0.77), VA1-1127 (1.36 x 0.94), VA1-1128 (1.47 x 0.88), fragment VA1-1172.

Raljin-B, 9 specimens: M1 dex RA-B-1001 (— x 1.59); M1 dex, fragment RA-B-1023; M2 sin, fragments RA-B-1012, 1013, 1014; M2 dex RA-B-1021 (1.63 x 0.98), RA-B-1022 (1.60 x 0.99), fragment RA-B-1026; M3 sin RA-B-1011 (1.56 x 1.05).

Description of the upper dentition

DP3: The outline of the DP3 is triangular, resembling the M1, but is smaller and especially narrower. A dilambdodont pattern is absent. The anterior outline of the DP3 is sinusoidal. The paracone and metacone are about the same size and are connected by a low and short, slightly curved centrocrista. The lingual base of the paracone is not, or weakly, connected to the preprotocrista and the lingual base of the metacone is not connected to the postprotocrista. A minute paraconule and a minute metaconule are present. The paracingulum is absent. The stylar shelf is absent labial the paracone and StB is absent. StA is isolated on the antero-labial base of the paracone. StC and StD are small, low, elongated, and positioned on the labial edge of the stylar shelf, anterior to the metastylar area.

M1-3: The M1 is less broad than M2 and M3, M3 and M2 are in the same width range. M1 is slightly longer and M4 has the same width-range as the other molars but is considerably shorter (Fig. 5).

Fig. 5
figure 5

Length and width scatters of the dental elements from Strelac-1, Strelac-3, Valniš-1 and Raljin-B (Serbia) of Amphiperatherium minutum (Aymard, 1846), ◊ DP3 and DP3, ○ M1 and M1, □ M2 and M2, △ M3 and M3, ▽ M4 and M4; Amphiperatherium exile (Gervais, 1848–52), ◆ DP3 and DP3, ● M1 and M1,■ M2 and M2, ▲ M3 and M3, ▼ M4 and M4; and Peratherium Aymard, 1850 M4

The outline of the upper molars is triangular. The molars are slightly shorter and broader from M1 to M4, resulting in different triangular shapes for each dental element. M1, M2 and M3 are dilambdodont. The cusps are high, with protocone as the lowest cusp and metacone the highest, they surround a deep trigon basin. The lingual base of the protocone is usually rounded in M1 and more angular in M2 and M3. The protocone is in line with the paracone and slightly asymmetrical with a stronger postprotocrista than preprotocrista. The paraconule is slightly smaller than the metaconule. The premetaconule crista is short and connected to the lingual base of the metacone, the postmetaconule crista is very short. The preparaconule crista continues into the paracingulum, the postparaconule crista is very short and connected to the lingual base of the paracone. The paracingulum, broad on its labial part, connects the paraconule to the base of StA. In M3 the paracingulum is very narrow or absent anterior to the paracone. The preparacrista is connected to the StB in 4 M1 (Fig. 3j), in five M1 and two M2 to the crista between StA and StB (Fig. 3i), in four M2 and five M3 to StA (Fig. 3b). The centrocrista is V-shaped and asymmetrical due to the wider angle of the postparacrista.

The stylar shelf, the area labial paracone and metacone, shows wide range of variations. The stylar shelf is shallow and not broad. In early wear the central part of the stylar shelf, together with the postparacrista and premetacrista, are strongly affected.

Stylar cusps A, B, C and D are present in all molars. StA is smaller than StB and StD, StB is mostly rounded and StD is elongated with sharp anterior and posterior sides. StB and StD are about the same size, although StB is usually slightly higher. StC is small or minute, connected to StD in most M1. It is isolated from StB and StD in M2 and M3 and positioned in most specimens on the edge of the ectoflexus. StE is absent, instead a large metastylar area is present. This metastylar area is the largest in M1 and in worn specimens quite flat. The most labial part of the centrocrista is very near to the ectoflexus and in a few not very worn molars almost connected to StC. In M1 and M2 the distance between paracone and StB is shorter compared to the distance between metacone and StD, in M3 these distances are equal. The labial side of paracone and metacone is often angulated, and in worn specimens a crista-like structure becomes clearly visible, in these molars the labial sides of the paracone and metacone are less steep and more voluminous. In nine molars (VA1) the labial angular side of the metacone connects to the lingual side of StD (in four M1, three M2 and two M3) (Fig. 3h-i). In two M1 and one M2 the angular labial part of the paracone is connected to the angular lingual part of StB (Fig. 3h), also small cuspules or narrow, low crista can be seen on the stylar shelf (Fig. 3j).

M4: The M4 is short. The prominent paracone is connected by a strong preparacrista to a small StA on the labial edge of the molar and by a short and low slightly curved centrocrista to the very small metacone. The postprotocrista connects protocone and metacone and bears a minute metaconule. The stylar shelf is narrow, short, and labial to the paracone. It has one small elongated stylar cusp. The paracingulum, with a broad labial part, connects protocone to StA.

Remarks: The differences between M1, M2 and M3, beside size and shape of the outline, are the small StC connected to StD and a large metastylar area of M1; the deeper ectoflexus of M3; the weaker or absent paracingulum anterior to the paracone of M3 and the more angular base of the protocone in M2 and M3. The M1 from RA-B and the M2 from STR3 fit in the same morphological variation of those from Valniš. The M2 from STR3 is slightly broader than the molars from Valniš. The width of the M1 from RA-B is in the size range of the Valniš-1 specimens.

Description of the lower dentition

M1-3 : The M1 and M2 show a large overlap in size. M3 is, in general, longer and broader, M4 is shorter and narrower. The first three lower molars are very similar in morphology and have slender cusps. The protoconid is the tallest cusp, metaconid is transverse or slightly posterior to protoconid. The lingual border is slightly undulating, due to a prominent metaconid and bulbous entoconid. A post metacristid is present in most specimens and connects to the entocristid. The hypoconulid, hook-shaped, is posterior to the entoconid and separated from it by a narrow valley. The slightly oblique postcristid connects hypoconulid and hypoconid. The cristid obliqua is low and connected to the posterior base of the protoconid. An anterior cingulum is always present on the antero-labial base of the protoconid and has a small cusp on its anterior end. The posterior cingulum connects the postero-labial base to the lowest part of the hypoconulid. The hypoflexid is very shallow, and a short cingulum can be present in most molars on the base of the molar labially this hypoflexid. The M1 has a more open prefossid, a long and straight paracristid, and a slightly broader trigonid than talonid.

M4: The M4 has a comparable morphology to M2 and M3, but its talonid is narrower and the crista obliqua is more lingually bent, connected to the more central part of the protolophid. The entoconid is smaller and the hypoconulid is farther from the lingual edge than in the other molars. The postcingulum is absent in most specimens.

Remarks: Most of the specimens are from Valniš; in the description the specimens from STR1, STR3 and RA-B are not separately mentioned since they fit in with the morphological variations of those from Valniš. As for size, M2 of Strelac-3 is slightly larger than those of Valniš-1, the M2 of RA-B is slightly larger than those of VA1 and the M3 of STR1 is larger than those of VA1 and larger than the M3 of RA-B.

The DP3 of Amphiperatherium minutum is not very well known and descriptions are almost absent in the literature. Our specimens are allocated to this species because of the presence of slender cusps and their smaller size compared to the DP3 of A. exile. The presence of a short transverse crista, on the stylar shelf of the upper molars connecting the paracone to StB and the metacone to StD is not described from A. minutum specimens from elsewhere. The morphology and sizes are comparable to Amphiperatherium minutum from France, Belgium (Crochet 1979, 1980a) and United Kingdom (Hooker 2010).

In our assemblages, the length decreases slightly and the width increases from M1 to M3, and in the lower molars, the length and width increase from M1 to M3. Length and width ranges for Amphiperatherium minutum molars show change through time (MP18 to MP23, Crochet 1980a, figs. 68, 69). The molars from younger sites are slightly larger, but an overlap between the oldest and younger sites is present. The size range of A. minutum from Valniš-1 is similar to the size range of the sites of Mas de Got and La Plante2 (FR, MP22), and similar to the size ranges of the lower molars of Ronzon and Aubrelong2 (FR, MP21; however upper molars have not been found at these two sites).

Amphiperatherium exile (Gervais, 1848–52)

(Figs. 6g, 7)

Type locality: Cournon (France)

Type level: MP28

Geographical distribution: Belgium, Bosnia-Herzegovina, France, Germany, Switzerland, United Kingdom; Serbia (this study)

Stratigraphical distribution: upper Eocene - upper Oligocene

Diagnosis: Crochet (1980a)

Localities: Buštranje, Strelac-1, Strelac-3, Valniš-1, Raljin-B (Serbia)

Age: early Oligocene (de Bruijn et al. 2017)

Material and measurements: In total 43 specimens (Figs. 6g, 7); measurements in Table 3 and scatters in Figure 5.

Fig. 6
figure 6

Detail of the dentition of Amphiperatherium minutum (Aymard, 1846) and Amphiperatherium exile (Gervais, 1848–52) from Strelac-1 and Valniš (Serbia). a Amphiperatherium minutum M1 dex (VA1-1033), detail of the postero-labial part of the molar, with a transverse crista on the stylar shelf. b M1 dex (VA1-1029), striae on the postmetacrista. c M1 dex (VA1-1034), striae on the postparacrista. d M2 dex (VA1-1032), a heavily worn stylar shelf and details of the paracone and metacone. e M2 sin (VA1-1071), striae in the trigonid. f M4 dex (VA1-1126), striae on the protoconid. g Amphiperatherium exile M1 sin (STR1-1002), striae on the metacone. All specimens are figured as right ones, specimens indicated by an underlined letter are depicted in reverse

Fig. 7
figure 7

Dentition of Amphiperatherium exile (Gervais, 1848–52) from Strelac-1, Strelac-3 and Valniš-1 (Serbia). a DP3 dex (STR1-1005). b DP3 dex (VA1-1062). c DP3 dex (STR1-1006). d M1 dex (VA1-1035). e M4 sin (STR3-1057). f M3 sin (STR3-1001). g M2 sin (STR1-1001). h M1 sin (STR1-1002). i DP3 sin (VA1-1064). j M3 dex (VA1-1101). k M2 sin (VA1-1112). l M1 dex (VA1-1103). m M2 dex (VA1-1077). n M1 dex (VA1-1110). All specimens are figured as right ones, specimens indicated by an underlined letter are depicted in reverse. Rectangle in h, see Fig. 6

Table 3 Length and width measurements of Amphiperatherium exile (Gervais, 1848–52) from Buštranje, Strelac-1, Strelac-3 and Valniš-1 (Serbia)

Strelac-1, 10 specimens: DP3 dex STR1-1005 (1.51 x 1.45), STR1-1006 (1.45 x 1.48); M1 sin STR1-1002 (1.95 x 2.05); M2 sin STR1-1001 (1.89 x 2.24); M1 dex STR1-1013 (— x 1.22), STR1-1015 (— x 1.22); M2 dex STR1-1014 (— x 1.17); M3 dex STR1-1016 (— x 1.17). M3 sin STR1-1010 (— x 1.15), STR1-1017 (— x 1.24).

Strelac-3, 3 specimens: M3 sin STR3-1001 (1.84 x 2.42); M3 sin, fragments STR3-1004, STR3-1005;

Valniš-1, 23 specimens: DP3 sin VA1-1063 (1.50 x 1.31); VA1-1065 (1.45 x 1.45); DP3 dex VA1-1064 (1.45 x 1.55); M1 dex VA1-1035 (1.65 x 2.01); M2 dex, fragment VA1-1040; M3 sin VA1-1027 (‘1.62’ x ’2.38’; not complete, protocone and paracone are damaged; StA with paracingulum dislocated); M4 sin VA1-1052 (1.01 x 1.83); M4 dex VA1-1057 (1.23 x 1.81); M1 sin VA1-1079, (1.59 x 0.97); M1 dex VA1-1103 (1.64 x 0.99), VA1-1110 (1.65 x 1.10), VA1-1121 (1.70 x —), fragments VA1-1162, VA1-1169; M2 sin VA1-1077 (1.70 x 1.22), M2 dex VA1-1111 (‘1.75’ x 1.05), M2 dex VA1-1112 (1.77 x 1.03); fragments VA1-1122, VA1-1171; M3 sin VA1-1081 (1.71 x 1.29); M3 dex VA1-1101 (1.77 x 1.24), VA1-1102 (1.77 x 1.00), VA1-1116 (1.80 x 1.20).

Raljin-B, 4 specimens: M1 sin, fragments RA-B-1015, RA-B-1017; M1 dex, fragment RA-B-1024; M3 dex, fragment RA-B-1025.

Buštranje, 3 specimens: M2 dex BUS-1001 (1.61 x —); M3 dex, fragment BUS-1002; M4 sin, fragment BUS-1005.

Description upper dentition

DP3: The DP3 resembles M1 but is smaller, especially narrower, and without a dilambdodont pattern. The anterior outline of the DP3 is sinusoidal. The metacone is connected to the smaller paracone by a low and curved centrocrista. StA is isolated on the antero-labial base of the paracone. The stylar shelf is absent labial of the paracone. The small StC, on the anterior side of the large StD has a strong connection to the centrocrista. The metastylar area is small. The preprotocrista and postprotocrista are not connected to the paracone and metacone, respectively. A minute paraconule and metaconule are present. The paracingulum is present and weak in all but 1 specimen, where it is absent anterior to the paracone.

M1-3: The M1, M2 and M3 are dilambdodont. The outline of the upper molars is triangular. The molars become broader from M1 to M3 (Fig. 5), resulting in different triangular shapes for each dental element, the cusps are robust. The specimens from Strelac are better preserved and used for the description because the upper molars from VA1 (M1, M2) are strongly worn. The lingual base of the protocone in M1 is more rounded compared to the more angular shape in M2 and M3. The stylar shelf is well-developed and StA, StB, StC and StD are present in all upper molars. StA is smaller than StB, and StD is the largest. StB is rounded and, in the M2, separated from StC and StD by the ectoflexus. In M3, the small StC is on the edge of the ectoflexus. StA is on its posterior side connected to the anterior edge of StB, and anteriorly connected to the short and broad paracingulum. StE is absent, instead a large metastylar area is present. This area is the largest in M1 and in worn specimens quite flat (Fig. 7d). The centrocrista is not connected to the stylar shelf, nor to one of the stylar cusps. The short transverse preparacrista connects paracone to StB in one M1 (Fig. 7h) and four M2 (Fig. 7g), to the stylar shelf between StA and SB in one M1. In the four M3 ‘s the preparacrista is either connected to StA, to STB or to the stylar shelf in between these cusps. The paracingulum is absent anterior of the paracone. The paraconule and metaconule are of about the same size, the preparaconule crista is short and ends just before the anterior base of the paracone, the postparaconule crista connects to the lingual base of the paracone. The premetaconule crista is connected to the lingual base of the metacone, the postmetaconule crista is short and ends at the posterior edge of the molar and is separated by the metacone by a small sinus (Fig. 7f).

M4: The M4 is short with a prominent paracone connected by the strong preparacrista to StA on the labial edge of the molar, and by a low centrocrista to the very small metacone. The postprotocrista connects protocone and metacone and bears a small metaconule. The stylar shelf is short, labial the paracone, and has one small elongated stylar cusp. A paracingulum is only present at the labial part of the paracone.

Description lower dentition

The M1, M2 and M3 have robust trigonid cusps. The first three lower molars are very similar in morphology, the M1 has usually the widest trigonid with a more open prefossid and a long and straight paracristid, combined with a slightly narrower talonid. The protoconid is the tallest cusp and the metaconid is transverse to the protoconid. The lingual border is slightly undulating, due to the large metaconid and bulbous entoconid. The posterior part of the metaconid (post metacristid) is present or short in M1 and M2, and very short or absent in the M3 (in most specimens ridgelike). When present, connected at its base to a slightly elongated entoconid (entocristid). The hook-shaped hypoconulid is posterior to the entoconid, separated from it by a narrow valley. The slightly oblique postcristid connects hypoconulid and hypoconid. The cristid obliqua is low and connected to the posterior base of the protoconid. An anterior cingulum is always present on the antero-labial part of the protoconid base and has a small cusp on its anterior end. The posterior cingulum connects the postero-labial base of the molar to just below the hypoconulid. The hypoflexid is very shallow. In most molars, a short cingulum can be present on the base of the molar labial to the hypoflexid.

Remarks: The morphology, especially the robust cusps, is as in Amphiperatherium exile. The size of the molars is comparable to A. exile specimens from Hoogbutsel (BE, MP21) and Pech Crabit (FR, MP23; Crochet 1980a, table 19a,b), they are smaller than the ones from Entreroche (SW, MP19; Hooker and Weidmann 2010) and Hampstead (GB, MP21; Hooker 2010).

In the upper molars, the width increases from M1 to M3 and an increase in length and width is observed from M1 to M3. The length and width from the many assemblages of A. exile do not show a clear trend, although a small decrease in size through time seems to be present (levels MP19 to MP30; Crochet 1980a, table 19a,b).

A short transverse crista on the labial side of paracone and metacone, as in A. minutum, is not present in the Serbian A. exile molars, nor referred to in other A. exile occurrences.

The DP3 ‘s of Amphiperatherium exile are only known from a few late Oligocene French localities (Aubenas-les-Alpes MP25, Maridet et al. 2019; Pech du Fraysse MP28 and Les Chaffours MP29, Crochet 1980a).

The dental elements of Amphiperatherium minutum differ from Amphiperatherium exile by their smaller size and the leaner cusps. In the upper dentition the dental elements (M4 excepted) increase in width towards the back-end of the toothrow, while the length remains the same. The lower dentition (M4 excluded) increases in length and width towards the back. The centrocrista in upper molars is closer to the stylar shelf, in M2 and M3 the preparacrista is only connected to StA, and the paracingulum is complete in most specimens.

Peratherium Aymard, 1850

Type species: Peratherium elegans (Aymard, 1846)

Type locality: Ronzon (France)

Geographical distribution: Western and Central Europe, Egypt

Stratigraphical distribution: upper Eocene – lower and middle Oligocene

Included species: Peratherium cuvieri (Fisher, 1829); Peratherium antiquum (Blainville, 1840); Peratherium elegans (Aymard, 1846); Peratherium cayluxi Filhol, 1877; Peratherium constans (Teilhard de Chardin, 1927); Peratherium bretouense Crochet, 1979; Peratherium lavergnense Crochet, 1979; Peratherium matronense Crochet, 1979; Peratherium monspeliense Crochet, 1979; Peratherium perrierense Crochet, 1979; Peratherium sudrei Crochet, 1979; Peratherium africanum Bown and Simons, 1984

Peratherium sp

(Fig. 4m)

Localities: Valniš-1 (Serbia)

Age: early Oligocene (de Bruijn et al. 2017)

Material and measurement: M4 dex VA1-1096 (1.75 x 0.98; not complete; the metaconid is broken off).

Description of the M4: The trigonid has high cusps, with a high steep anterior side of the paracristid. The metaconid is a little posterior to protoconid, and the protocristid is a high straight wall. The anterior cingulum connects the anterior base of protoconid to the central base beneath the paracristid. The talonid is low and shorter than trigonid, with a small entoconid, hypoconulid and hypoconid. These three cusps are well connected, of almost the same size, symmetrically placed on the posterior edge of the molar, with the hypoconulid on its most distal part. A cristid obliqua is absent and the entoconid is separate from the metaconid. The talonid is much narrower than trigonid, its lingual side is not in line with the labial edge of the trigonid, it is slightly lingually bent. The posterior cingulum is absent.

Remarks: This M4 differs from the Valniš-1 Amphiperatherium minutum specimens by its lager size, the prominent high trigonid cups, the absence of a connection between the small entoconid and metaconid, the hypoconulid is connected to the entoconid and is equally far from to the hypoconid. The crista obliqua is absent.

The M4‘s of Amphiperatherium minutum from the Eocene and Oligocene of Western and Central Europe (Crochet, 1980a) are slightly smaller, have an entoconid well separated from the hypoconulid and a short crista obliqua. The M4 of A. minutum from the early Oligocene of Wight (UK, MP21) is smaller, has a relatively larger talonid and a strong crista obliqua, less symmetrically placed cuspids, but has the same outline (Ham4, Hooker 2010, fig. 4d).

The M4 from Valniš-1 fits in the size range of Amphiperatherium exile from Pech du Fraysse (MP28, FR; Crochet 1980a), but the latter M4’s have a relatively longer talonid, a more lingually positioned hypoconulid and is close to the entoconid and further away from the hypoconid, and a prominent cristid obliqua is present. The M4 of A. exile from Entreroches (SW, MP19; Hooker and Weidmann 2010, fig. 15a) is larger, has a relatively longer talonid, and the hypoconulid is almost on the lingual edge of the molar near the entoconid and has a crista obliqua.

This Valniš-1 M4 resembles the M4 of Peratherium cuvieri from Entreroches (SW, MP19; Hooker and Weidmann 2010; fig. 15i), which has a short talonid, with its cuspids of almost the same size and centrally placed on the posterior edge of the molar. However, the Entreroches specimen is smaller and the connection between entoconid and hypoconulid is stronger. The M4 from the type-locality of this species (Crochet 1979) is larger, it displays a clear separation of entoconid and hypoconulid, and has a cristid obliqua. The Valniš-1 M4 falls in the size range of Peratherium lavergnense from several French localities (Crochet 1980a), but entoconid and hypoconulid are not connected and a cristid obliqua is always present.

Although we have an isolated specimen and cannot confirm the, for Peratherium, typical character of a distally enlargement of the molars along the tooth row, we assign this molar to this genus because of its similar morphology to Peratherium cuvieri from Entreroches (Hooker and Weidmann 2010). The diagnosis of Peratherium by Crochet (1980a) notes that the reduction of the talonid size of M4 is associated with the disappearance of the postcristid and the hypoconulid has the same distance to hypoconid as to entoconid. These features are prominent in the Valniš-1 M4.

Tooth eruption and occlusion in marsupials

The sequence of tooth eruption and pattern of tooth replacement is similar between early marsupials and now living didelphids (Cifelly and de Muizon 1998) and the only replacement is the DP3 by the P3 (Luckett and Woolley 1996). The premolars and the first two molars erupt at about the same time in young individuals, the third molar lags a little and the fourth molar is the last molar to erupt. The DP4 and DP4 are replaced by the P4 and P4, at about the same time as the eruption of the fourth molar. The DP4 is thus an important part of the dentition during the life of a young marsupial (Luckett and Woolley 1996). Figures 8a and 8b show a composed upper and lower dentition of Amphiperatherium minutum and shows the orientation and interlocking of these teeth (DP4-M4 and M1-M4). Figure 8b illustrates the occlusion of lower and upper jaw dentition.

Fig. 8
figure 8

Dentition of Amphiperatherium minutum (Aymard, 1846) composed of isolated dental elements from Valniš-1 (Serbia). a right upper jaw, M4 (VA1-1126), M3 (VA1-1034), M2 (VA1-1012), M1 (VA1-1023), DP3 (VA1-1001). b the lower molars (thin line) in occlusion with the upper molars (thick line). c right lower jaw, M4 (VA1-1127), M3 (VA1-1106), M2 (VA1-1071), M1 (VA1-1078)

Tooth occlusion of Amphiperatherium is probably the same as in extant insectivorous and faunivorous marsupials such as Didelphis Linnaeus, 1758 (Crochet 1980b, Butler 1996, Hooker 2010). Figure 7 shows the wear facets and striae on the teeth interpreted as the result of chewing. The lower jaw moves up so that food comes into contact with the lower and upper teeth (the preparatory stroke of Crompton and Hiiemae 1970). While the food is crushed and punctured, the teeth do not come into occlusion. The food is pressed between protocristid and preparacrista (Fig. 6e-f), and between paracristid and postmetacrista (Fig. 6b) and the hypoconid moves laterally from the top of the centrocrista towards the protocone, during which pressure occurs between the crista obliqua and postparacrista (Fig. 6d), and between postcristid and premetacrista (Fig. 6d, g). Thus, only the area around the centrocrista, the anterior edge of the molar and the postmetacrista are included in the initial crushing and puncturing of food (Butler 1996). In the subsequent power stroke, the food is sheared and cut and occlusal contact between the teeth is possible (Crompton and Hiiemae 1970). At first, the metastylar area (Fig. 6a) is affected, and subsequently the main cusps (Fig. 7d) (Butler 1996). During the preparatory phase, when the food is pushed upwards, and during the power stroke when the food is pressed between the stylar shelf the cheek, the stylar shelf comes into contact with food, causing wear (Figs. 3c, 6d). The striations on the anterior and posterior sides of upper molar cusps (Fig. 6b-d, g) are shallow, mainly oriented vertically, while the surface of the stylar shelf is smooth. Striations on the lower molar cusps (Fig. 6e-f) are also shallow and vertical, and in the deeper part of the trigonid basin only a few striations are present. The stronger abrasion observed in Peradectes species, compared to that in Amphiperatherium and Peratherium species, is interpreted by Crochet (1980a, b) as an indication of a diet of harder and less carnivorous food.

Transverse crista and small cusps on the stylar shelf present in Serbian Amphiperatherium specimens (Figs. 3h-j, 6a-b) are also present in other extinct marsupials. Such extra crista or cusps on the stylar shelf are visible in figured specimens such as Amphiperatherium frequens (Ziegler and Mörs 2000, pl. 1,4), Peratherium antiquum (Hooker et al. 2008, fig. 1D), Amphiperatherium exile and Peratherium cf. perrierense (Hooker 2010, fig. 5a,b,h), Peratherium constans (Ladevèze et al. 2012, fig. 4B,5A,B) and in the collection of Headon Hill (Wight, GB) collection of the UU, and in Asiadidelphis tjutkovae (Emry et al. 1995, fig. 2). The wide range of morphological variation of the stylar shelf is probably caused by the absence of occlusal contact (Crompton and Hiiemae 1970), i.e. absence of constraints on morphological change (Ladevèze et al. 2012; Ladevèze et al. 2020) and can be interpreted as intraspecific variations (Ladevèze et al. 2012). In our opinion, it is therefore better not to use the morphological variations of the stylar shelf as a distinguishing feature of a new species.

Discussion and conclusions

Peradectidae and Herpetotheriidae are common in Western and Central European Eocene and Oligocene faunas. These families are diverse and represented by many species (Crochet 1980a), they are not known from the Palaeocene. The Peradectidae disappear from the European fossil record in the early Eocene and the Herpetotheriidae become gradually less divers during the Oligocene but remain part of the fossil assemblages until the early Neogene with Amphiperatherium frequens, and disappear during the late Early Miocene (Furio et al. 2012, Klietmann et al. 2014, Crespo et al. 2020). Records of herpetotheriids east of Germany are from the early Oligocene of the Czech Republic (MP21, Detan; Fejfar and Kaiser 2005) and from the latest Oligocene of Bosnia and Herzegovina (MP 30, Banovići; van der Sar et al. 2017).

In many Western and Central European late Palaeogene faunas two or more marsupial species are present and the presence of three herpetotheriids, Amphiperatherium minutum, A. exile, and Peratherium sp. in the Palaeogene of Serbia is thus comparable. The joint presence of these species in Western and Central European assemblages in the late Eocene to early Oligocene, is in accordance with the age range of the Serbian rodent faunas as suggested by de Bruijn et al. (2017).

The occurrence of Amphiperatherium minutum, A. exile and Peratherium sp. in the Late Palaeogene of south-eastern Serbia enlarges the known biogeographical distribution of the herpetotheriids towards the Balkans. The presence of Amphiperatherium exile in the oldest Serbian assemblage shows that the eastward expansion of herpetotheriids was already established in the latest Eocene.

The difference in composition between the Serbian and Western and Central European rodent faunas indicates limited faunal exchange between the two areas. Only Paracricetodon dehmi Hrubesch, 1957 and Pseudocricetodon montalbanensis Thaler, 1969 occur in the Balkans as well as in West and Central Europe (van de Weerd et al. 2018, Marković et al. 2019). The presence of the same three marsupials in Serbia and Western and Central Europe is therefore surprising. But it shows their ability to cross barriers and disperse over large areas. An example of this ability is Peratherium which crossed the Neotethys as it occurs in the lower Oligocene sediments of the Fayum area in Egypt.

The Eocene-Oligocene boundary is clearly visible in the mammalian faunas of Western and Central Europe as a distinct and sharp break, the ‘Grande Coupure’. In contrast, the Eocene-Oligocene boundary shows no distinct break in the rodent faunas of the Balkans and Anatolia. Similarly, the marsupial assemblages do not how a break across this time boundary, neither in the Balkans nor in Western and Central Europe.