Introduction

During the Paleogene and most of the Neogene, the guild of carnivorous mammals in South America was occupied by metatherians. Placental carnivores arrived later, by the late Miocene, to South American landscapes (Soibelzon and Prevosti 2007; Prevosti et al. 2013). By the early Paleogene, several metatherian lineages occupied the niche of carnivory. All of these were mesocarnivores (Zimicz 2012) with a diet composed of vertebrate meat complemented with invertebrates and fruits. Some members of the Peradectidae, Mayulestidae, and the first Sparassodonta (Allqokirus australis) composed this stock of early carnivores (Muizon 1991, 1998; Zimicz 2012). By the late Paleocene-early Eocene, sparassodonts developed the first hypercarnivore type, e.g., Nemolestes spalacotherinus (a medium-sized species; Zimicz 2012). By the late Eocene the Sparassodonta radiated successfully into the niches of medium- and large-sized hypercarnivores (Zimicz 2012). The first specialized proborhyaenids are well represented at this point in time (Forasiepi 2009).

The oldest hathliacynids are recorded in the early Oligocene (Goin et al. 2010) but the age predicted for the origin of the family is middle-late Eocene (Forasiepi 2009). As other sparassodonts, the main history of the group took place during the Neogene when they reached their acme and posteriorly declined and went extinct. From an ecological point, the Hathliacynidae have been considered as omnivores (Marshall 1978, 1981; Argot 2004) as well as hypercarnivores (Prevosti et al. 2012, 2013) of small and medium size forms (Prevosti et al. 2013). The hathliacynids were probably terrestrial and scansorial (Marshall 1977a; Muizon 1998; Argot 2003; Ercoli 2010; Ercoli et al. 2012; Prevosti et al. 2012).

Didelphimorphia developed different carnivorous types, especially from the middle Neogene until the Recent. All Neogene taxa are included in the Didelphoidea, and are actually considered a natural group including the Didelphidae, Sparassocynidae, and Caluromyidae (Goin 1991). The family Didelphidae includes the monophyletic group of large carnivorous-omnivorous opossums (Jansa and Voss 2005; Jansa et al. 2006; Flores 2009). The Sparassocynidae is an extinct family of carnivorous didelphimorphians known from the Miocene and Pliocene of Argentina (Reig and Simpson 1972; Simpson 1974; Forasiepi et al. 2009). Finally, the Caluromyidae include the more strictly frugivorous opossums represented in the fossil record since the early Miocene (Goin et al. 2007). Didelphoids carnivores filled the size range of small and medium sizes from 0 to 10 kg and they were adapted to a terrestrial and scansorial lifestyle (Szalay 1994; Muizon and Argot 2003; Viera and Astúa de Moraes 2003).

The late Cenozoic history of didelphoids is mostly known from fossil specimens recovered at numerous localities in the Pampean Region, central Argentina (Goin 1995). Carnivorous didelphoids are well represented since the Huayquerian SALMA (Hyperdidelphys, Thylophorops, Didelphis, Sparassocynus, Hesperocynus, and Thylatheridium), reaching their largest diversity by the Chapadmalalan SALMA (Goin and Pardiñas 1996; Goin and Montalvo 1988; Forasiepi et al. 2009; Goin et al. 2009).

From the late Miocene to the early Pliocene, the Sparassodonta coexisted in South America with placental carnivores (Soibelzon and Prevosti 2007; Prevosti et al. 2013). Several authors have suggested that the arrival of placental carnivores was the cause of the extinction of the sparassodonts (Simpson 1950, 1969, 1971, 1980; Patterson and Pascual 1972; Savage 1977). Other authors question this idea (Marshall 1977b, 1978; Reig 1981; Bond 1986; Pascual and Bond 1986; Goin 1989, 1995; Ortiz Jaureguizar 1989, 2001; Marshall and Cifelli 1990; Alberdi et al. 1995; Forasiepi 2009; Forasiepi et al. 2009; Prevosti et al. 2009). A close examination of the diversity curve of both groups demonstrates that no active or passive replacement occurred (Prevosti et al. 2013). Sparassodonta coexisted also with Didelphoidea. The sparassodonts were more specialized in the predator niche than Didelphoidea. The temporal and the body size overlap, the terrestrial/scansorial habits, and the decided carnivorous trend of several extinct didelphoids led several authors to postulate a possible competitive interaction between the small- and medium-sized Sparassodonta (Hathliacynidae) and the carnivorous didelphoids (Marshall 1977b, 1978; Goin 1989; Goin and Pardiñas 1996; Goin et al. 2009; Forasiepi 2009). On the contrary, Prevosti et al. (2013) pointed out the existence of a potential niche separation between these two groups of South American metatherians.

Different models have been used to understand the evolution of two clades occupying the same adaptive zone at the same time in a given geographic area. The active displacement that implies competition (Benton 1983) or the passive replacement that implies the action of external factors on the species populations (Benton 1983) commonly shows “double-wedge” geometries on the diversity patterns (Van Valkenburgh 1999). In general, one clade begins to decline in diversity while a second begins to increase. When the rate of displacement is gradual, a competitive interaction is inferred but, when the rate of replacement is rapid and associated with environmental changes, an opportunistic process is suspected (Benton 1983). The simplest model to analyze the competitive interaction in the fossil record is that which verifies a linear decrease in the diversity of the first clade concomitantly with a linear increase in the diversity of the second clade (Benton 1983; Krause 1986; Van Valkenburgh 1999). Other models consider the joined action of biotic and physical factors (Rosenzweig and McCord 1991; Sepkoski 1996, 2001; Sepkoski et al. 2000).

The aim of the present review is to clarify the ecological relationships between Hathliacynidae and the carnivorous Didelphoidea in order to determine the potential existence of competitive interactions between them.

Methods

In order to understand the ecological interactions, the diversity of Sparassodonta and carnivorous Didelphoidea was calculated for each age, following current chronostratigraphic and biogeographic charts (Flynn and Swisher 1995; Cione and Tonni 2005, modified by Woodburne et al. 2006; Gelfo et al. 2009; Madden et al. 2010; Ré et al. 2010; see Prevosti et al. 2013: Fig. 1). The geographical and chronological information of the fossil taxa included here was drawn from the literature (Marshall 1978, 1979, 1981; Patterson and Marshall 1978; Marshall et al. 1983; Goin and Montalvo 1988; Pascual and Ortiz Jaureguizar 1990; Goin 1991; Goin and Pardiñas 1996; Goin et al. 2009; Forasiepi 2009; Forasiepi et al. 2009). The taxonomy of Hathliacynidae follows Forasiepi (2009). The taxonomy of Didelphoidea mainly follows Goin and Montalvo (1988), Goin (1991), Goin and Pardiñas (1996), Goin et al. (2009), and Forasiepi et al. (2009).

Fig. 1
figure 1

a, Diversity through time for Hathliacynidae (gray line) and Didelphoidea (black line); b, Bivariate plot of body mass versus time for Hathliacynidae (triangle) and Didelphoidea (oval). Refs: SALMAs South American Land Mammal Ages; Col Colhuehuapian; San Santacrucian; Fri Friasian; Coll Colloncuran; Lav Laventan; May Mayoan; Chas Chasicoan; Hua Huayquerian; Mon Montehermosan; Chap Chapadmalalan; Blob Barrancalobian; Vor Vorohuean; Sand Sanandresian; Ens Ensenadan; Bon Bonaerian; Luj Lujanian; Pla Platan; Rec Recent

Full size image

The diversity was measured as the number of species in each period considering South American Land Mammal Ages (SALMAs) as time unit. A taxon was included in counts of intermediate ages when it was found in an older and younger age. The diversity pattern of Hathliacynidae and the carnivorous Didelphoidea were compared in order to detect ecological interactions. The terms “competitive displacement” and “opportunistic replacement” as well as the condition to infer each process in the fossil record follows Benton (1983) and Krause (1986).

To evaluate the ecological overlap, body mass data of Hathliacynidae were compiled from Prevosti et al. (2013) and the body mass of extinct Didelphoidea was then estimated. Data of body mass of living didelphids were taken from Smith et al. (2003). The least square regressions used to estimate the weight of extinct species are based on dental measurements (Gordon 2003). Size categories mentioned in the text are mostly based in Carbone et al. (2007): small (< 2 kg); medium (2–10 kg); and large (> 10 kg).

Following Van Valkenburgh (1991) and Friscia et al. (2007), a set of morphometric variables were analyzed to evaluate the potential superposition in the trophic habits. These variables are: relative grinding area (RGA); relative blade length (RBL); relative size (RPS) and shape (PS) of the biggest lower premolar. Additionally, the relative length of the premolar blade with respect to the carnassials (LP/LM) was taken following Palmqvist et al. (2011). The lower fourth molar was considered as a carnassial tooth following Werdelin (1987). The RGA was used to discriminate between hypercarnivores (with RGA < 0.5), mesocarnivores (with 0.5 < RGA < 0.8), and hypocarnivores (RGA > 0.8). A principal component analysis (PCA) on the variance-covariance matrix was performed with all morphometric variables in order to visualize the distribution of species in the morphospace. The potential overlap in the feeding requirements of Hathliacynidae and Didelphoidea was evaluated through the superposition in the multivariate space.

Results

Diversity

During most of their record, the Hathliacynidae had a low diversity of six taxa or less. The diversity rose to a maximum in the Santacrucian SALMA (Fig. 1a). The Colhuehuapian SALMA provided four taxa and two in the Friasian. The Colloncuran and the Laventan SALMAs were represented by one and two taxa respectively. During the Chasicoan there was one species and four in the Huayquerian (Fig. 1a). The number of species fell to one in the Montehermosan and the Chapadmalalan SALMAs. By the Barrancalobian the Hathliacynidae seem definitely extinct. The carnivorous Didelphoidea appear in the fossil record by the Huayquerian with five taxa (Fig. 1a). During the Montehermosan there were six species present and nine in the Chapadmalalan. After that, a strong decrease in the species richness was recorded with two species in the Barrancalobian, one in the Vorohuean and the Sanandresian. From the Ensenadan to the Platan, non-carnivorous didelphoids were recorded. In the Recent, a new rise in the species richness is recorded with nine taxa (Fig. 1a).

In general, the species of the Hathliacynidae were of medium size and ranged between 2 kg and 10 kg in body mass (Fig. 1b). On the contrary, the majority of the Didelphoidea weighed below 2 kg in body weight (see Electronic Supplementary Information). However, during the period of coexistence, both groups overlaped in the small size range (Fig. 1b). The hathliacynids involved in that overlap are Borhyaenidium musteloides, B. altiplanicus, B. riggsi, Notyctis ortizi, and Notocynus hermosicus (Table 1). The didelphoids are Thylophorops perplanus, Hyperdidelphys pattersoni, and Hyperdidelphys dimartinoi (Table 1).

Table 1 Raw data used in the morphospace analysis
Full size table

Geographic Distribution

The geographic distribution of the coeval species is sympatric in few taxa (see Elecronic Supplementary Information). Borhyaenidium musteloides is known from Huayquerian levels of the Epecuén Formation in La Pampa Province, Argentina (Marshall 1981), while B. riggsi comes from the Montehermosan levels of the Corral Quemado Formation, Catamarca Province (Marshall 1981). The specimens of B. altiplanicus were exhumed from the Huayquerian levels of the Mauri VI Formation, La Paz, Bolivia (Villarroel and Marshall 1983). Notocynus hermosicus is known from Monte Hermoso locality in Buenos Aires Province, while Noticitis ortizi come from the Ituzaingó Formation (Entre Ríos Province). Regarding the geographic distribution of didelphoids, Hyperdidelphys pattersoni is known from the Huayquerian and the Montehermosan beds of Catamarca and La Pampa provinces (Goin and Pardiñas 1996). Hyperdidelphys dimartinoi comes from Chapadmalalan beds of Buenos Aires Province. Thylophorops perplanus is known from the Montehermosan levels of Buenos Aires Province. In summary, only in Buenos Aires Province were sympatric species recorded for the Huayquerian (B. musteloides and H. pattersonii) and the Montehermosan (N. hermosicus and T. perplanus) stages.

Diet

A PCA were performed using the variance-covariance matrix of all ratios (Table 1). Three components was extracted that explain 98.71 % of the variance among all taxa (Table 2). PC1 accounts for the size effect. Most of the size-independent morphological variation is explained by PC2 and to a lesser degree by PC3. The variable loadings show a strong correlation between the RPS and the PC1, which is expected because it is a size dependent ratio. The RGA is highly correlated with the PC2 (Table 2) as well as RBL. These variables are opposite in sign and explain the major variance along the PC2 (Fig. 2). The PS and the LP/LM are slightly correlated with the PC2; however, they are strongly correlated with the PC3 (Table 2). The morphospaces defined by PC1-2 and PC2-3 show a clear segregation of the hathliacynids from the didelphoids (Fig. 2a–b). In the former, the Hathliacynidae are confined to the left lower corner of the graph. An exception is Acyon myctoderos, which is located in the right lower corner. This fact reflects the high body mass of this species, which is the largest hathliacynid (Table 1). The Didelphoidea occupy the left and right upper corners of the graph (Fig. 2). On the space defined by PC2-3, the hathliacynids occupy the left part of the graph while didelphoids are located on the right side with the exception of Lutreolina crassicaudata (Fig. 2b). PC2 again separates the two groups while PC3 accounts for much of the within group variance. In relation to the contribution of ratios to the variance, RGA and RBL clearly define the taxonomic groups. Hathliacynids are characterized by large relative blades and small grinding areas in their carnassials teeth while the didelphoids are characterized by short relative blades and moderate to large grinding areas (Fig. 2a–b). The morphological variation within Hathliacynidae is low and only A. myctoderos constitutes an outlier in the morphospace. Contrary to this, the variation within the Didelphoidea is large. The species of Thylophorops are distanced from the remaining didelphoids by their large relative grinding areas, short relative blades lengths, and short and massive lower premolars. The species of Didelphis and Philander are characterized by the presence of moderate grinding areas, short relative blades lengths, and large premolars. The species of Hyperdidelphys, Thylatheridium, Sparassocynus, and Hesperocynus are characterized by having moderate grinding areas, short relative blades, and short lower premolars.

Table 2 Variable loadings for principal components analysis (PCA). Under the heading for each component, in parentheses, is the percentage of the total variance it explains. For explanation of abbreviations, see Table 1
Full size table
Fig. 2
figure 2

Plot of principal component 1 versus principal component 2 (a) and principal component 2 versus principal component 3 (b) for the principal components analysis performed using all ratios, with taxa distinguished by taxonomic group. Ref: Hathliacynidae (triangle) and Didelphoidea (oval)

Full size image

The dietary classification based on the RGA ratio allowed identifying the Hathliacynidae as hypercarnivores (Table 1). The Didelphoidea result in being mainly classified as mesocarnivorous with the exception of the genus Thylophorops, a hypocarnivorous taxon (Table 1). These results are interesting in light of the PCA because the hathliacynids are well segregated from didelphoids and, the species of Thylophorops are clearly separated from the remaining didelphoids. The mesocarnivorous didelphoids are a group highly variable in the PCA with at least two subcategories.

In short, the PCA and the RGA suggest that the Hathliacynidae and the Didelphoidea constitute two clearly different morphotypes. These differences in the morphofunctional traits probably represent different dietary specializations.

Despite the overlap in the range of small sizes recorded in the Huayquerian and the Montehermosan (Fig. 1a), any dietary superposition was evident between sympatric species. The results of PCA with all morphofunctional variables show a complete separation in the hathliacynid and the didelphoid morphospaces (Fig. 2a–b). All Hathliacynidae showed the morphological characteristics of the hypercarnivores, while Didelphoidea showed those of the meso- and hypocarnivores.

Discussion

There is a positive correlation between the number of taxa and the number of localities per age that affects the sample size in the diversity analysis of the Cenozoic South American fossil record (Prevosti et al. 2013). Several mammalian assemblages are only partially known (e.g., Mayoan Age) and others have mostly regional distributions (Patagonian assemblages of the Miocene). Such biases contribute to the incompleteness of the South American fossil record and obviously affect the interpretation of the diversity patterns.

Taken into account the biases in the fossil record, the origin of Hathliacynidae can be traced back to the middle-late Eocene (Forasiepi 2009). The species richness of the family was low (less than two species per age: Zimicz 2012) until the early-middle Miocene (Santacrucian) when the Hathliacynidae rose to their acme. The low diversity of the family recorded from the Colloncuran to the Chasicoan is probably due to biases in the fossil record. The number of sites of the Santacrucian and the Huayquerian is twice or more than that of any age of the mentioned span (Prevosti et al. 2013: Table 1). Based on this, the definite decline of Hathliacynidae is considered to begin after the Huayquerian. Although the last record of the family is in the Chapadmalalan (early Pliocene; Fig. 1a), a well-sampled land mammal-bearing age, the span from Barrancalobian to Sanandresian is represented by few localities, so the diversity of hathliacynids could increase with new collections (expeditions or prospections) from this time interval. The first carnivorous didelphoids were recorded in the Huayquerian SALMA. However, the high initial diversity (five species and five genera) is suggestive of a bias in the previous record of the group. The group increases their diversity until the Chapadmalalan SALMA (Fig. 1a). Between the Barrancalobian and the Platan, the diversity of carnivorous didelphoids falls drastically from nine species to less than two. As it occurs with the hathliacynids, this drop could well be exaggerated by a biases in the fossil record. A new rise in the Recent is recorded with nine species of carnivorous didelphoids. Despite the bias recorded for the Barrancalobian to Sanandresian span, the low diversity of both metatherian groups could be related to a strong environmental change promoted by the onset of the Northern Hemisphere Glaciation (see below). The joined diversity curves of the hathliacynids and didelphoids shows double wedge geometry. Such a pattern could be the result of negative interactions between clades (competitive displacement or predation) or the result of chance or environmental changes (Van Valkenburgh 1999).

From a methodological point of view, the results of PCA corroborate previous studies that emphasize the utility of simple linear measures on the dietary inference of small- to medium-sized carnivores (Van Valkenburgh 1991; Friscia et al. 2007). This study shows that the relative grinding areas and the relative blade lengths are well discriminant ratios in extinct taxa as suggested by Friscia et al. (2007) and can be used to determine the diet of extinct carnivores. Other variables such as premolar shape and relative premolar length are less correlated with dietary preferences but also contribute to the total morphological variance (Fig. 2a). The analysis of RGA shows that the Hathliacynidae were restricted to hypercarnivore niches while the didelphoids occupied the niches of meso- and hypocarnivores. Although the PCA does not discriminate between all dietary categories, it shows a clear separation between hathliacynids and didelphoids (Fig. 2a–b). Additionally, a clear separation is evident between the species of Thylophorops (the only hypocarnivores from the RGA analysis) and the remaining didelphoids. Some didelphoids fall in the intermediate zone between the two taxonomic groups. Lutreolina crassicaudata is closer to the hathliacynids zone (Fig. 2b). The RGA value for Lutreolina is in the boundary between meso- and hypercarnivory (Table 1), so probably their classification as mesocarnivores responds to a inability of RGA to discriminate in the boundaries of the categories. Despite the dietary consideration of the species, the multivariate analysis shows that the two metatherian groups clearly diverged in their morphologies and any overlap is recorded in their morphofunctional traits.

Previous studies on the decline and extinction of Hathliacynidae conferred a major role to the ecological interplay with medium-sized didelphids, mustelids, and canids (Marshall 1978; Goin 1989; Goin and Pardiñas 1996). Recently, Prevosti et al. (2013) demonstrated that probably none ecological displacement did take place between Sparassodonta and Carnivora. In regard to the interaction between the hathliacynids and the didelphoids, the results of PCA do not support any competitive displacement or passive replacement. Diet and body mass played a crucial role in the niche differentiation of sympatric species of carnivorous metatherians in South America. The same pattern was observed in the case of the interaction between the placental carnivores and the sparassodonts (Prevosti et al. 2013). Contrary to what occurred with the placental carnivores, the didelphoids never occupied the niche of hypercarnivores even after the Chapadmalalan SALMA when the last hathliacynid was recorded. Probably the didelphoids were poor competitors with the successful placental carnivores that occupied the hypercarnivore niche after the extinction of the sparassodonts (Prevosti et al. 2013). This fact probably reflects a metabolic constraint of the metatherians to survive in cold environments (see below). With respect to the potential intra-guild predation mentioned earlier, although the body mass of hathliacynids allows them to prey on didelphoids of similar and lower weight, a closer examination of the craniodental variables of both groups (Van Valkenburgh and Koepfli 1993; Meachen-Samuels and Van Valkenburgh 2009) is needed to clarify this aspect.

The hathliacynids originated in the Paleogene (probably middle Eocene: Forasiepi 2009) when greenhouse condition prevailed in the global climates. A minor component of the carnivoran mammalian fauna persisted (Forasiepi 2009; Zimicz 2012; Prevosti et al. 2013) and survived the global climatic crash of the Eocene-Oligocene boundary that caused several extinction and turnovers events around the world (Stehlin 1909; Prothero 1994; Meng and McKenna 1998; Goin et al. 2010). The Neogene climate represents the transition from the greenhouse climate of the Paleogene to the icehouse climate of the Quaternary (Bruch et al. 2007). After the Mi1 glaciation, the Miocene climates were distinguished by the extreme climatic optima (Middle Miocene Climatic Optimum) alternating with major long-term climatic cooling (Middle Miocene Transition). Both events determine the evolution of hathliacynid diversity promoting the acme of the clade and probably the gradual decrease in the species richness, respectively. The final step in the evolution of the hathliacynids occurred in the Mio-Pliocene Transition. The cooling trend established with the Middle Miocene Transition was sustained throughout the remaining of the Miocene until the early Pliocene (Flower and Kennett 1994; Lewis et al. 2008; Filippelli and Flores 2009). The onset of the Northern Hemisphere Glaciation at the middle-late Pliocene imposed a new and strong fall in global temperatures (Filippelli and Flores 2009). Such freezing conditions probably caused the final extinction of Hathliacynidae; their last record occurs during the Chapadmalalan SALMA with Borhyaenidium riggsi. Suggestively, the extinction of the hathliacynids was coeval with a local faunal turnover that took place at the mid-Pliocene. Thirty-seven percent of all genera and 53 % of all species became extinct between the older Chapadmalalan and the younger Barrancalobian chronostratigraphic units (Vizcaíno et al. 2004). One of the most striking features of the Barrancalobian fauna is the absence of medium and large carnivores (Vizcaíno et al. 2004). With the exception of Sparassocynus derivatus, a small mesocarnivorous didelphoid, no other vertebrate carnivore is recorded. The virtual absence of carnivorous taxa and especially the absence of the hypercarnivores shows the strong influence that climatic deterioration would cause on the guild of the highly specialized carnivores (Van Valkenburgh 2007). In this sense, it is important to note that carnivory is a food strategy related with high metabolism in eutherians. Marsupials have low basal metabolic rates (McNab 2005); consequently, the high cost of thermoregulation has important consequences on their overall biology including their resistance to cold environments and the potential food resources that can be exploited by them (McNab 2005, 2008). If all metatherians have a similar energetic metabolism compared to living marsupials, then the extinction of the Hathliacynidae seems to have occurred as the result of the sustained drop in ambient temperatures during most of the 10 Ma, from the middle Miocene to middle Pliocene (Zachos et al. 2001). The climatic changes involved in the mid-Pliocene faunal turnover probably accelerated an extinction process that was triggered by the Middle Miocene Transition. The impact of cold climates on the carnivorous metatherian is evident looking at the Paleogene faunas. In South American landscapes, where metatherians coexisted with placentals at least since the early Cenozoic (Bond et al. 1995; Gelfo et al. 2009; Woodburne et al. 2014), hypercarnivory was successfully exploited during the greenhouse period (Zimicz 2012). After the Middle Miocene Transition, when temperatures decline progressively, hypercarnivorous metatherian were no longer present and the only carnivorous marsupials recorded were the meso- and hypocarnivorous didelphoids.

The carnivorous Didelphoidea appeared in the Huayquerian stage with a diversity of five species that represent five genera. This high initial diversity suggests a bias in the fossil record probably related to the scarce number of sites between the Colloncuran and Chasicoan ages. Considering that the origin of the Didelphoidea has been estimated as a late Eocene event (Steiner et al. 2005), it is probable that new paleontological prospecting in the early-middle Miocene beds of South America will result in an increase in the diversity of carnivorous didelphoids. Taking into account the known record of carnivorous didelphoids, the success of this group is strongly evident being represented by nine species and five genera. As mentioned above, the didelphoids had meso- and hypocarnivorous diets even after the extinction of Hathliacynidae. The generalist nature of their feeding requirements probably allows them to survive in the changing world of the Pliocene-Pleistocene Transition, characterized by a glacial-interglacial cyclic event (Tiedemann et al. 1994). The onset of the Northern Hemisphere Glaciation and the Chapadmalalan-Barrancalobian extinction event (Vizcaíno et al. 2004) probably had a strong impact on the carnivorous didelphoid fauna, promoting a taxonomic and functional turnover (sensu Blois and Hadly 2009). The genera Hyperdidelphys and Thylatheridium became extinct after the Chapadmalalan SALMA. The only Chapadmalalan genera that survive into Barrancalobian were Thylophorops and Sparassocynus, each represented by one species. However, none of them survived into the Pleistocene. The last Sparassocynus (S. derivatus) was recorded in the Vorohuean, while the last Thylophorops (T. lorenzinii) was recovered from the Sanandresian subage of the Marplatan Age (Goin et al. 2009). The extinction of Thylophorops may be understood as a competitive interaction with placental hypocarnivores present in South America since the Chapadmalalan (see Prevosti et al. 2013). The post-Chapadmalalan didelphoids were all mesocarnivorous species (Table 1). The post-Pleistocene record of carnivorous didelphoids corresponds exclusively to members of this modern fauna. The genera Didelphis and Lutreolina have their oldest record in the Huayquerian and Montehermosan, respectively (Goin and Pardiñas 1996), and both survived into the present. In this sense, it is important to note that the recent radiation of carnivorous didelphoids has its origin in the late Miocene and the main components of the living fauna (e.g., Didelphis and Lutreolina) survived through the climatic crash of mid-late Pliocene.

In summary, the extinction of Hathliacynidae, a clade that originated during a greenhouse world, is related it seems with the sustained drop in global temperatures since the middle Miocene. Contrary to this, didelphoid carnivores survived and radiated in an icehouse world. Interesting, the generalist Didelphimorphia that coexisted with placental carnivores, presumably were at no physiological disadvantage, having similar metabolic rates to generalist placental species (McNab 2005; Cooper et al. 2009). As suggested by the fossil record, the extinction was the cost of the highly specialized carnivory for metatherians in South America (Forasiepi et al. 2009). The global cooling trend starting in the middle Miocene seems to be the major selective pressure acting on the carnivorous metatherian faunas. The application of modern techniques for estimating the physiology of fossil vertebrates (e.g., Chinsamy et al. 1995; Eagle et al. 2009; Köhler and Moyá-Solá 2009) can provide more clues to the ecological evolution of metatherians in South America.

Conclusions

The actual fossil record of the Hathliacynidae and the carnivorous Didelphoidea suggests that no competitive displacement or passive replacement took place between these groups of carnivorous metatherians during the Neogene. The Hathliacynidae were exclusively hypercarnivores of small- and medium-sized, while the didelphoids were restricted to meso- and hypocarnivorous diets even after the extinction of the hathliacynids. This dietary segregation contradicts the hypothesis of passive replacement of Hathliacynidae by the carnivorous Didelphoidea. The extinction of the first (Hathliacynidae), seems to be triggered by the Middle Miocene Transition that promoted a sustained decline in the global temperatures during more than 10 Ma. The onset of Northern Hemisphere Glaciation was probably the “death blow” for the Hathliacynidae. The climatic crash involved in the Chapadmalalan-Barrancalobian event caused a taxonomical turnover within the carnivorous Didelphoidea. The ancient genera Hyperdidelphys, Thylophorops, Sparassocynus, Hesperocynus, and Thylateridium became extinct before the Plio-Pleistocene transition. The late fossil record of this group corresponds entirely to members of the modern fauna of didelphoids.