Specimens from the following eight localities were analyzed, presented in chronological order from the earliest to the most recent.

Saint-Vallier (France, Auvergne-Rhône-Alpes, MNQ17) is dated at ca. 2.5 Ma (Nomade et al., 2014). Because of its significant biodiversity, it is considered a reference for the middle Villafranchian (Azzaroli et al., 1988, Guérin, 1990, Guérin et al., 2004 and Heintz et al., 1974). The habitat of Saint-Vallier has been described as a mosaic landscape with the presence of steppe, punctuated by a deciduous tree cover typical of a temperate or warm temperate climate and associated with a watering place based on a palynologic study (Argant, 2004). Based on the mammal assemblage, the humidity in the habitat is considered significant and the temperature warm (Argant and Philippe, 2011 and Guérin et al., 2004). Cervids are the most abundant family, representing 60% of the faunal assemblage (Valli, 2004). The typical deer assemblage of this period is present in Saint-Vallier, with Eucladoceros ctenoides vireti (= E. senezensis vireti) together with Metacervoceros rhenanus (= “Cervus” philisi valliensis) and Croizetoceros ramosus medius (Heintz, 1970, Valli, 2004 and Valli and Palombo, 2008).

Chilhac (France, Auvergne-Rhône-Alpes, MNQ17) is radiometrically dated at 2.36 Ma (Nomade et al., 2014) and is a reference site in Europe. Eucladoceros ctenoides, Metacervoceros rhenanus and Croizetoceros ramosus are present in this locality (Boivin et al., 2010 and Heintz, 1970). It has been described as a humid site (Boeuf and Gilbert, 1997) and the high proportion of cervids is considered indicative of the significant tree cover in the locality (Guérin et al., 2004).

Varshets (Bulgaria, Province of Montana, MNQ17) is a rich middle Villafranchian karstic locality with similarities to the reference locality of Saint-Vallier (Spassov and Crégut-Bonnoure, 1999). The faunal assemblage is dominated by cervids, represented by Eucladoceros ctenoides cf. vireti, Metacervoceros rhenanus (= “Cervus” philisi valliensis) and Croizetoceros ramosus medius (Guérin et al., 2004). The habitat has been described as a mosaic forest-steppe type of landscape (Bechev and Georgiev, 2016 and Spassov and Crégut-Bonnoure, 1999).

Dafnero (Greece, Western Macedonia, MNQ17) is a middle Villafranchian assemblage of three localities with important faunal similarities to Saint-Vallier (Guérin et al., 2004, Koufos, 1993 and Koufos, 2016). Given the balanced ratio of bovids to cervids, the locality is interpreted as a mosaic habitat with forest and savannah areas (Kahlke et al., 2011). Cervids include E. ctenoides (= Eucladoceros senezensis) and Metacervoceros rhenanus (= “Cervus” philisi).

Senèze (France, Auvergne-Rhône-Alpes, MNQ18), situated only 10 km from Chilhac, is a reference MNQ18 locality for western Europe (Pastre et al., 2015). It corresponds to a maar deposit. The cervids present in the locality are Eucladoceros senezensis senezensis (synonym of E. ctenoides according to De Vos et al., 1995), Metacervoceros rhenanus (= “Cervus” philisi philisi), Croizetoceros ramosus minor and Cervalces gallicus (Azzaroli et al., 1988, Geraads, 1990 and Heintz, 1970). Eucladoceros ctenoides is among the dominant species of the locality. The environmental context has been described as similar to Saint-Vallier, i.e., mosaic with the presence of open forest and steppe areas. A pollen analysis highlighted the presence of a mesothermophilic vegetation indicative of a temperate climate (Azzaroli et al., 1988, Delson et al., 2006, Guérin et al., 2004 and Nomade et al., 2014).

Slivnitsa (Bulgaria, Province of Sofia, MNQ18) is a late Villafranchian cave locality north-west of Sofia. The faunal assemblage presents similarities with the Senèze fauna. The habitat seems to have been more open and the climate more arid than that of Varshets, which would explain the significant proportion of bovids in this locality. Deer are represented by Eucladoceros cf. ctenoides (= E. cf. senezensis) and Metacervoceros rhenanus (Spassov, 1998 and Spassov and Crégut-Bonnoure, 1999). Despite the significant abundance of fossils in the locality, Eucladoceros dental remains are scarce.

Gerakarou (Greece, Central Macedonia, MNQ19) is a rich locality from the beginning of the late Villafranchian (Konidaris et al., 2015 and Koufos, 2016), south-east of the Mygdonian Basin, Greece (Koufos et al., 1995). Deer from Gerakarou are represented by Eucladoceros ctenoides, Metacervoceros rhenanus (“Cervus” philisi) and Croizetoceros ramosus (Kostopoulos and Athanassiou, 2005 and Kostopoulos and Koufos, 1994). When considering faunal composition, no noticeable climatic changes have been observed between Gerakarou and earlier middle Villafranchian localities. Among herbivores, intermediate feeders are considered to be dominant (Kahlke et al., 2011).

Krimni (Greece, Central Macedonia, MNQ19) is a late Villafranchian locality (Koufos, 2016) situated in the Mygdonian Basin, Greece. The material from Krimni is scarce. The faunal assemblage presents similarities with the locality of Gerakarou. There are few fossils from Eucladoceros sp. (Kostopoulos, 1996 and Koufos et al., 1995) that may not be conspecific with the Eucladoceros ctenoides from Gerakarou (Kostopoulos and Koufos, 1994).

The wide range of variation in the dental microwear texture results of Eucladoceros illustrates the diversity of the food that this taxon is able to consume and therefore its significant dietary plasticity (Fig. 3).

When comparing the dietary plasticity of Eucladoceros with that of the two contrasted modern red deer populations (Fig. 3), it is clear that the dental microwear signal of fossil deer overlaps the signal of both populations and occupies the entire range between the two. This similarity in the dietary plasticity of Cervus elaphus and Eucladoceros ctenoides is not surprising as it is well known for extant herbivores that a large body mass permits an increased retention time of the digesta in the gut and thereby implies a higher digestive efficiency (Müller et al., 2013). This characteristic means that unlike smaller herbivores, Eucladoceros is able to feed on a large range of what is available in its habitat in order to fulfill its energy and nutrient requirements, rather than to select only high-quality food items. This selective behavior is classically found in smaller herbivores such as the extant concentrate feeder Capreolus capreolus (Hofmann, 1985) or the intermediate feeder Dama dama (Chapman and Chapman, 1980 and Hofmann, 1989), which feed on selected foods not necessarily abundant in their habitat (Jackson, 1977, Nugent, 1990 and Tixier et al., 1997).

Guérin et al. (2004) consider deer from Saint-Vallier forest inhabitants. Valli and Palombo (2005) concluded, based on a dental microwear, mesowear and skull anatomy analysis of five specimens that the large-sized deer from Saint-Vallier was engaged in browsing. The pollen analysis by Argant and Philippe (2011) highlighted the presence of temperate deciduous tree cover in the locality.

The diatoms and pollen analyses of Senèze (Ehrlich, 1968 and Elhaï, 1969) show climatic oscillations between a warm temperate climate with deciduous tree cover and a cool to cold climate with herbaceous monocots and conifers. Pastre et al. (2015) consider the deer from Senèze to be forest inhabitants. Valli and Palombo (2005), based on the analysis of dental microwear, mesowear and skull anatomy of three specimens, concluded that E. ctenoides from Senèze was a browser. The abundance of deer in this locality has been interpreted by Bonifay and Brugal (1996) as indicative of “a fresh temperate climate with a constant hygrometry and short cold winters”.

With a grazing signal similar to that of the modern reference population from the Dehesa Lugar Nuevo, the feeding ecology of fossil deer from Saint-Vallier and Senèze contrasts highly with interpretations of Eucladoceros as being forest inhabitants. This signal somewhat implies the exploitation by these deer of an open habitat with an abundant herbaceous vegetal layer, highlighting the consumption of herbaceous monocots by Eucladoceros. Even confronted with mesothermophilic vegetation as in Senèze (Nomade et al., 2014), Eucladoceros could have exploited the herbaceous stratum.

Thus, the occurrence of Eucladoceros in itself as a proxy for paleoenvironmental reconstructions is flawed.

One could be surprised by the contrasting results on Eucladoceros ctenoides from Saint-Vallier and Senèze between the present study and the analysis conducted by Valli and Palombo (2005). In our opinion, such a discrepancy could be the result of a bias linked to a too small sample size (N = 5 and 3) in their study (2005). Such sample sizes are not large enough to draw conclusions in terms of diet reconstructions, more specifically when considering dental microwear patterns which may vary on a week to month scale basis. A fact that, by the way, the authors themselves do mention. In their study of the skull anatomy, mesowear and microwear of Metacervoceros rhenanus and Croizetoceros ramosus from Saint-Vallier, Valli and Palombo (2008) underscored the fact that some taxa, i.e. Cervus elaphus, present a huge versatility in their diet while their skull anatomy remains relatively unchanged (See also Berlioz et al., in press and Gebert and Verheyden-Tixier, 2001). In our study, we show that E. ctenoides displays a dietary plasticity similar to that seen in modern Cervus elaphus throughout the European continent.

Despite strong similarities in faunal assemblages (Spassov and Crégut-Bonnoure, 1999), Eucladoceros from Saint-Vallier differs from the deer from Varshets in its dental microwear signal, the latter being more engaged in browsing. Based on the high proportion and diversity of deer in the faunal assemblage, similar conclusions have been drawn regarding the paleoenvironment of Chilhac (Guérin et al., 2004). Those assumptions are confirmed by dental microwear. Therefore, in a habitat characterized by significant tree cover, Eucladoceros present a browsing signal. The same kind of results is supported by dental mesowear for Eucladoceros from the Italian locality of Coste San Giacomo. There, the large-sized deer has been found to browse in an environment characterized by a heterogeneous vegetation providing a large spectrum of food resources (Strani et al., 2015).

While both fossil populations present a browsing signal, Eucladoceros from Chilhac and Varshets differ significantly by their textural fill volume results. Ramdarshan et al. (2016) showed that among browsing species, seedeaters tend to have a higher textural fill volume than leaf browsers. Indeed, high values of this parameter are classically interpreted as resulting from the inclusion of hard food items in the diet, leading to the formation of large pits and deep scratches on the dental facet. In addition, the textural fill volume has been shown to be especially high for frugivorous herbivores (Scott, 2012) and higher values for Eucladoceros from Chilhac may reflect the ingestion of a larger proportion of hard items such as seeds.

The absence of significative difference between the browsing signal in Chilhac, Varshets and Gerakarou and deer with a more grazing signal results from the fact that deer populations more engaged in browsing have a large spectrum of texture values. This possibly underlines a certain diversity of available food categories in the habitat they occupy either on a spatial range or linked to annual (seasonal) variations in the vegetal phenology. Schulz et al. (2013) have shown that the dispersion of the dental microwear textures is greater when the silica proportion in the diet is low, which is the case for browsing populations like the ones in Chilhac, Varshets and Gerakarou. One could consider these differences in the dental microwear textures between localities as reflecting differences in the season of death. The wide spectrum in Chilhac, Varshets and Gerakarou could alternatively be interpreted as a homogeneous annual mortality rate in comparison with other localities. Biplots considering inter-individual variations in anisotropy and complexity (Supplementary data; Fig. S1) do not highlight any seasonality in the dental microwear textures. Alternatively, no information relative to seasonality based on other proxies (antler shedding, mortality profiles) does allow us to go further into such interpretations for these three localities.

Conversely, Senèze and Saint-Vallier fossil deer are characterized by a low dispersion in their microwear signal that illustrates a vegetal resource availability being more restricted to abrasive food items.

These results concerning fossil deer diet plasticity are of particular interest. The plasticity in dietary habits of extant deer is well understood (Gebert and Verheyden-Tixier, 2001), but the presence of fossil deer in a locality is still too often considered a sufficient paleoenvironmental proxy for drawing conclusions when it comes to the existence of tree cover (e.g., Guérin et al., 2004). This may result in the inaccuracy of paleoenvironmental reconstructions (Alcalde and van den Hoek Ostende, 2015). Yet several studies focusing on early and middle Miocene cervids (DeMiguel et al., 2008, DeMiguel et al., 2010, DeMiguel et al., 2011, Merceron et al., 2007, Merceron et al., 2012 and Solounias and Moelleken, 1994) stressed the fact that some Miocene deer are already engaged in mixed feeding or even grazing strategies despite presenting a brachydont to mesodont molar morphology. Procervulus has even been shown to be highly plastic in its diet, modulating its feeding habits in response to the important Miocene climatic and environmental changes (DeMiguel et al., 2010). Such results are not unexpected when looking at the most recent large ecological studies on modern cervid populations. Indeed, several modern deer taxa also include an important portion of herbaceous monocotyledons in their diet (Berlioz et al., in press, Bugalho and Milne, 2003, Gebert and Verheyden-Tixier, 2001 and Geist, 1998 and references therein). For example, the diet of Cervus elaphus is composed of up to 90% of herbaceous monocots from March to August in Lugar Nuevo (Azorit et al., 2012).

By being plastic in its diet, the dental microwear textures of Eucladoceros might be an adequate proxy to assess paleo-climatic variations during its period of abundance. The picture of vegetal resource availability in the habitat given by such a plastic taxon is much more complete than what could be observed with a deer presenting a selective behavior (e.g. Capreolus capreolus).

The dental microwear textures of Eucladoceros oscillate over time between a grazing and browsing dietary pole depending on localities. It could reflect an alternation between wooded and open habitats, resulting from climatic fluctuations from glacial to interglacial events following the 41 ky periodicity that occurred during the middle and late Villafranchian (Kahlke et al., 2011 and Lisiecki and Raymo, 2005). The microwear signal varies in the same way in western Europe between the deer from Chilhac and Senèze and in eastern Europe between Eucladoceros from Varshets and Dafnero. This result may illustrate a certain homogeneity of the environments of southern Europe during the early Pleistocene (Bonifay and Brugal, 1996) (Fig. 4). It is not possible to associate each locality with one specific glacial or interglacial event identified by Lisiecki and Raymo (2005), as most of the fossil sites are only biochronologically dated. However, it is possible to go further into interpretations.

Despite the fact that the differences between eastern European localities are not significant because of a low sample size (Table 1) and the wide dispersion of dental microwear results in woodland habitats, tendencies were identified between 2.3 and 1.6 Ma. As referred to by Bonifay and Brugal (1996), glacial events during the early Pleistocene of southern Europe are characterized by the spread of steppe environments, progressively leading to the reduction of the temperate, mild and humid climate. Hence, observed fluctuations of the dental microwear signal of Eucladoceros from browsing in Varshets to grazing in Dafnero and browsing again in the more recent locality of Gerakarou might illustrate climatic oscillations during this period (Fig. 4). Therefore, Dafnero may correspond to a glacial event characterized by a more open habitat compared to Varshets and Gerakarou that may indicate interglacial events. DMTA tends to indicate an increased involvement in browsing for Eucladoceros from Gerakarou when compared to earlier representatives, while no noticeable climatic change was observed between Gerakarou and older localities based on the faunal assemblage (Kahlke et al., 2011).

In the western region between 2.5 and 2.2–2.1 Ma, Eucladoceros also experiences several significant shifts in its diet that could be superimposed on climatic oscillations occurring during this period. While there is an important abrasive component in the diet of deer from Saint-Vallier that reflects the occupation by Eucladoceros of an open habitat, fossil deer from Chilhac present a woodland signal. In Senèze, Eucladoceros display dental microwear corresponding to grazing habits. The locality of Saint-Vallier has been described by Guérin et al. (2004) as a mostly open habitat. This could correspond to the description Bonifay and Brugal (1996) made of the Artemisia-Ephedra steppe that developed in southern Europe during glacial recurrences. Conversely, the browsing signal results that are observed in Chilhac may correspond to an interglacial event, with a higher vegetal diversity and tree cover. The signal of Senèze would correspond again to a glacial locality characterized by an open habitat.

The dental microwear signal of Eucladoceros from Chilhac and Senèze underlines a good example of climatic variation through time, as the two localities are geographically close to each other (10 km) but are separated by several hundred thousand years if referring to recent age estimations (Nomade et al., 2014 and Pastre et al., 2015). This significant shift in the diet of Eucladoceros remains in the range of environmental variations that the deer is able to handle.