Is the extant southern short-tailed opossum a pigmy sabretooth predator?

Authors


  • Editor: Andrew Kitchener

Correspondence

Rudemar Ernesto Blanco, Instituto de Física, Facultad de Ciencias, Universidad de la República, Iguá 4225, Montevideo 11400, Uruguay. Tel: 598 2 5258624; Fax: 598 2 5250580

Email: ernesto@fisica.edu.uy

Abstract

Hypertrophied canines evolved several times among mammalian carnivores. Several palaeobiological hypotheses related to sabretooth evolution and killing behaviours have been suggested based on biomechanical and functional considerations. However, the lack of well-studied extant analogues makes it difficult to test these hypotheses. Here we propose the South American short-tailed opossum Monodelphis dimidiata as a living analogue of extinct sabretooth predators. Our morphological analysis shows that M. dimidiata not only has relatively the largest canines among extant marsupial carnivores, but they are also within the range of those of sabretooth predators. It also has cranial adaptations for a wide gape typical of sabretooth carnivores. The small body size of this species allows further biological studies that can provide useful information to understand the evolution, behaviour and physiology of extinct sabretooth carnivores.

Introduction

The sabretooth morphology originated independently at least four times in mammalian predators (Emerson & Radinsky, 1980; Radinsky & Emerson, 1982; Turner & Antón, 1997) or five times if the nimravids are split in two separate groups (Peigné, 2003; Peigné & de Bonis, 2003; Morlo, Peigné & Nagel, 2004). There have been many functional studies of the sabretooth condition (Christiansen, 2011 and references therein). Several hypotheses about the adaptive advantages, ecological role, biomechanical constraints and the palaeobiology of sabretooth mammalian predators have been suggested (Turner & Antón, 1997; Martin et al., 2000; Therrien, 2005; Wroe, McHenry & Thomason, 2005; Christiansen, 2006; Slater & Van Valkenburgh, 2008; Meloro & Slater, 2012). However, it is difficult to evaluate these hypotheses without a living analogue. The clouded leopards, Neofelis spp., seem to show skull features considered to be characteristic of the primitive sabretooth condition (Christiansen, 2006, 2008). Unfortunately, little is known of their ecology and hunting behaviour (Nowak, 1991; Sunquist & Sunquist, 2002; Grassman et al., 2005; Christiansen, 2006, 2008). Moreover, other morphometric analyses failed to find much similarity between extant Neofelis nebulosa and sabretoothed carnivores (Slater & Van Valkenburgh, 2008). In another study (Goswami, Milne & Wroe, 2010), N. nebulosa clustered with the nimravids Dinictis and Hoplophoenus, but not the other sabretooths.

Therefore, the status of N. nebulosa is controversial, but still it is one of the very few living analogues of the primitive sabretooth previously proposed. To speculate about the hunting behaviour of primitive sabretooth cats, Christiansen (2006) used N. nebulosa and considered available evidence of killing large prey (Rabinowitz, Andau & Chai, 1987; Grassman et al., 2005) and each other (Seager & Demorest, 1978) with a powerful nape bite and suggested the following: ‘It may be that its enlarged gape and hypertrophied canines are an adaptation for nape killing of large prey, but this is, at present, speculation’. Christiansen (2011), based on a dynamic model, speculated about mandibular adductor histochemistry and morphology in sabrecats. But all these ideas would remain speculations ‘… until a Pleistocene sabrecat is unearthed from the permafrost, as have been numerous proboscideans and other megaherbivores’ (Christiansen, 2011). Until a frozen Pleistocene sabrecat is found, a strategy to test ideas about killing behaviour, mandibular adductor histochemistry and morphology is to identify a living primitive sabretooth analogue that allows further study.

The sabretooth ecomorphology originated not only in the order Carnivora, but also among predatory marsupials such as the borhyaenids (see, e.g. Blanco, Jones & Grinspan, 2011 and references therein). The living predatory marsupials are the didelphids and dasyurids; among them we found the southern short-tailed opossum Monodelphis dimidiata, a very small species. Monodelphis dimidiata is a grassland-dwelling opossum from Uruguay, Argentina and Brazil. The species presents sexual dimorphism, adult male body mass is between 100 and 150 g and adult female body mass is between 30 and 70 g (González, 2001). The diet in the wild includes plants, insects, arachnids and small rodents. Analysis of stomach contents found mammal hairs and/or rodent bones in 33% of the individuals (Busch & Kravetz, 1991). The rodent species found in the stomachs were: Bolomys obscurus, Oligoryzomys flavescens, Calomys laucha and Oxymycterus rutilans with body mass ranges of 30–80, 18–39, 9–15.5 and 50–120 g, respectively (data from González, 2001).

Monodelphis dimidiata is one of the best examples of a semelparous marsupial. Some dasyurid marsupials are semelparous, although females may live a second year (Lee & Cockburn, 1985). Adult male M. dimidiata disappears from the population in March, 2 months earlier than females, and thus exhibits a male mortality syndrome after mating (Pine, Dalby & Matson, 1985). Males have only one opportunity for reproductive success and there may be severe competition for access to females, with the larger, more aggressive and more canine-enhanced males having a competitive advantage (González & Claramunt, 2000).

Monodelphis dimidiata shows a broad repertoire for dealing with various kinds of prey, such as dehairing hairy caterpillars, crunching the heads of arthropods and killing mice by means of a neck bite (González & Claramunt, 2000). The authors described the following: ‘Laboratory mice are quickly and continually attacked until the opossum can grasp the mouse by the throat. The mouse is then held in that way until it stops moving’ (González & Claramunt, 2000). Generally, carnivorous marsupials use crushing bites directed to the anterior of the prey's body and often strike the head, neck or even chest (Eisenberg, 1985; Croft, 2003; Jones, 2003). The reported killing behaviour of M. dimidiata, which avoids biting bones, could be analogous to the killing technique proposed for several extinct sabretooth predators (Biknevicius & Van Valkenburgh, 1996; Antón & Galobart, 1999; Salesa et al., 2005; Turner & Antón, 1997).

Emerson & Radinsky (1980) described cranial features that distinguish sabretooths from living felids and marsupial predators. They concluded that sabretooth predators have modifications for a wider gape with the retention of a powerful bite force at the carnassial.

Here we make morphological studies, using methods already used in the study of the sabretooth condition, in order to determine how suitable M. dimidiata is as a living analogue of primitive sabretooth predators.

Methods and materials

We worked with an osteological sample of 44 individuals of living marsupials from South America (didelphids, 14 species) and Australia (dasyurids, 18 species). The sample includes four specimens of M. dimidiata, three males and one female. The specimens are housed in the collections of the Museo Nacional de Historia Natural in Montevideo and the Western Australian Museum. For details of the specimens, see Supporting Information Appendix S1.

Using dial calipers, we took 15 linear measurements on each skull based on those of Emerson & Radinsky (1980) (see Figs 1 and 2). For comparing our data with those of Emerson & Radinsky (1980), we calculated 14 indices. The indices are the ratio between the actual measured value and the expected value from allometric equations based on our whole sample (see Supporting Information Appendix S1). The results in M. dimidiata are compared with the range in the whole living marsupial sample (except M. dimidiata) and the published data in Emerson & Radinsky (1980). We compare these indices with those considered as indicative of the sabretooth condition in Emerson & Radinsky (1980), and we will test if any of the indices for M. dimidiata lie outside the ranges of those of other marsupial predators.

Figure 1.

Lateral, frontal and dorsal views of opossum Monodelphis dimidiata cranium showing the measurements used in the present work.

Figure 2.

Lateral view of the left mandible and the humerus of opossum Monodelphis dimidiata showing the measurements used in the present work.

We calculated a separate series of 14 indices in order to perform principal component analyses (PCAs) to identify combinations of features that distinguish M. dimidiata from other marsupial predators. Each cranial measurement and jaw length (JL) were divided by the skull length (SL), while each mandibular measurement was divided by the JL of the same specimen. For temporal fossa width index (TFW/SL) the numerator is the difference between zygomatic arch width (ZAW) and post-orbital constriction. The purpose of this study is to determine if a combination of indices can distinguish M. dimidiata from other marsupial predators, and to compare those features with the features that distinguish sabretooths.

We performed a PCA using all 14 indices and examined the principal components (PCs) to identify one that separated M. dimidiata from the other marsupial predators. Then we excluded, one by one, indices that contributed least to the separation and repeated the PCA until we had a significant PC (eigenvalue larger than Jolliffe cut-off) that separated M. dimidiata male specimens from the whole sample. We considered that those remaining indices correspond to the morphological features that characterize the peculiarities of M. dimidiata as a carnivorous marsupial.

We took three measurements on the humerus of our marsupial specimens (see Fig. 2). From these measurements, we derived two indices that would give an indication of the robusticity of the glenohumeral joint and the development of forearm musculature.

Results

Comparing the indices used by Emerson & Radinsky (1980), M. dimidiata males have larger values for canine height (C1Hi), and the outlever for the M3 bite (COM3i) than the other predatory marsupials in our sample. Canine length (C1Li) is also significantly larger in M. dimidiata males than those of other marsupials (t-test P < 0.05). Comparing the indices of M. dimidiata with the data in Emerson & Radinsky (1980), M. dimidiata canine height and length scores are above the ranges of those for living felids and within the ranges of those for the sabretooth condition. Among sabretooths and modern felids, Emerson & Radinsky (1980) only provide COM3 data for Thylacosmilus and Machaeroides and they have values well below the ranges for M. dimidiata of either sex. For all the other indices, the scores for M. dimidiata overlap with both modern felids and the sabretooth condition. However, the sabretooth condition generally has lower values for the lever arms of masseter and temporalis (MFLi and MATi).

In the PCA using all 14 indices, PC5 separates M. dimidiata males from all other marsupial predators (Fig. 3). The loadings presented in Table 1 show the indices that contribute most to PC5. However, PC5 only accounts for 3.5% of the total variation and its eigenvalue is below the Jolliffe cut-off point. We excluded, one by one, indices that contribute least to the PC that separates M. dimidiata from the other marsupial predators. The exclusion of indices was made in the following order: TFL/SL, JH/JL, C1W/SL, TRL/SL, OCPH/SL, OCPW/SL, ZAW/SL and TFW/SL.

Figure 3.

Plots of (a) PC1 versus PC5 from principal component analysis with 14 indices, and (b) PC1 versus PC3 from principal component analysis with 6 indices, for opossum Monodelphis dimidiata, Didelphidae and Dasyuridae.

Table 1. PC loadings (eigenvectors) from an analysis using all 14 indices (Table 4). The percentage variation for the first five PCs is given in the first row. PC5, which distinguished opossum Monodelphis dimidiata from the other marsupials, is presented in bold font, but this PC is not significant (below the Jolliffe cut-off value)
IndicesPC1PC2PC3PC4PC5
53.44%15.51%14.18%5.93%3.50%
  1. C1H = upper canine height; C1L = upper canine anteroposterior length; C1W = upper canine mediolateral width; COM3 = condyle to M3; JL = jaw length; OCPH = occipital height; OCPW = occipital width; SL = skull length; TFL = temporal fossa length; TFW = temporal fossa width; TRL = tooth row length; ZAW = zygomatic arch width; JH = jaw height; MFL = masseteric fossa length; MAT = moment arm of temporalis; PC, principal component.
C1H/SL0.077540.18330.14930.025940.4713
C1L/SL0.031220.09271−0.00319−0.020180.1398
C1W/SL0.038710.035360.05755−0.009670.0615
JL/SL0.27790.4016−0.48770.2326−0.1597
OCPH/SL0.1364−0.1444−0.16460.13360.1056
OCPW/SL0.3151−0.7264−0.02643−0.060110.1044
TFL/SL0.1574−0.217−0.077120.81150.02709
TFW/SL0.60230.31890.2552−0.15280.122
TRL/SL0.12050.1193−0.551−0.06483−0.06862
ZAW/SL0.6165−0.10110.03461−0.1765−0.1265
COM3/JL−0.014160.14980.09780.28080.63
JH/JL0.058140.16690.18950.02343−0.0401
MFL/JL0.032230.14760.41730.3478−0.5116
MAT/JL0.09814−0.038560.33650.07534−0.1015

With the removal of the eighth index, PC3, which accounted for 13.8% of the total variation, was above the Jolliffe cut-off point, and separated M. dimidiata males from the other marsupial predators (Fig. 3). Table 2 shows the loadings of the remaining six indices in the PC analysis. The height and length of the canine (C1H/SL and C1L/SL) and the outlever for the M3 bite (COM3/SL) have high positive loadings, and the lever arms of masseter and temporalis (MFL/JL and MAT/JL) and the mandibular length index (JL/SL) have high negative loadings.

Table 2. PC loadings (eigenvectors) from an analysis using six indices. The percentage variation for all six PCs is given in the first row. PC3, which distinguishes opossum Monodelphis dimidiata from the other marsupials, is significant and presented in bold font
IndicesPC1PC2PC3PC4PC5PC6
46.16%28.06%13.84%7.40%4.18%0.38%
  1. C1H = upper canine height; C1L = upper canine anteroposterior length; COM3 = condyle to M3; JL = jaw length; MFL = masseteric fossa length; MAT = moment arm of temporalis; PC, principal component; SL = skull length.
C1H/SL0.13030.3060.47610.4134−0.65−0.2633
C1L/SL0.096070.062270.15390.07186−0.18260.9616
JL/SL0.97150.1123−0.1503−0.061880.1197−0.05294
COM3/JL−0.000270.29540.7514−0.33430.4857−0.02213
MFL/JL−0.14120.7508−0.3517−0.4826−0.24410.01148
MAT/JL−0.099940.4889−0.19690.68950.48410.05022

For both humeral indices, M. dimidiata has values beyond the range of any studied marsupial predator, and the difference in the means is highly significant (t-test P < 0.0001). Monodelphis dimidiata has a relatively large inter-epicondylar width, and humeral head compared to those of other predatory marsupials (Table 3).

Table 3. Humeral indices for the study species
SpeciesHumeral head index (HHD/HL)Inter- epicondylar index (IEW/HL)
  1. HHD = diameter humeral head; HL = humeral length; IEW = Inter-epicondylar width; MNHN = Museo Nacional de Historia Natural in Montevideo; WAM = Western Australian Museum. 
Opossum Monodelphis dimidiata males mean (n = 3)0.2420.343
Phylander opossum MNHN 43150.1780.273
Chironectes minimus MNHN 38060.1960.253
Didelphis marsupialis MNHN 55300.1970.239
Didelphis albiventris MNHN 43470.2010.274
Dasycercus cristacaudalis WAM 418910.1610.261
Dasyurus geofferoii mean (n = 3)0.1440.228
Dasyurus hallucatus mean (n = 2)0.1620.254
Dasyurus viverrinus WAM 212300.1410.206
Sarcophilus harrisii mean (n = 2)0.1620.243
Pseudoantechinus macdonnellensis WAM 490860.1650.226
Thylacinus cynocephalus WAM 3318-0.188

Discussion

The indices defined in Emerson & Radinsky (1980) show that M. dimidiata has hypertrophied canines in comparison with those of modern carnivorous marsupials. The C1Hi index of all three M. dimidiata and the C1Li index for two of the three males are outside of the ranges of those of other carnivorous marsupials. The C1Hi and C1Li indices of the males of M. dimidiata calculated here are relatively larger or comparable to those of sabretoothed felids and nimravids such as Dinictis, Hoplophoenus, Machairodus, Homotherium and Ischyrosmilus (see Table 4). Only the more extreme sabretoothed felids and nimravids such as Eusmilus, Barbourofelis and Smilodon have relatively larger canines than those of M. dimidiata. Among the marsupials and creodont sabretooths, Thylacosmilus has relatively much larger canines than those of M. dimidiata, but C1Li measured in the incomplete skull of Machaeroides is lower. The indices C1Hi and C1Li of M. dimidiata are even larger than those in N. nebulosa. Therefore, the canines of M. dimidiata are hypertrophied in comparison with those of living carnivorous marsupials in a way that is similar to the canines of sabretoothed felids, which are hypertrophied in comparison with those of living felids. The other indices that characterize sabretooth crania and mandibles are not clearly outside of their ranges of variation in modern carnivorous marsupials (see Supporting Information Appendix S1). A similar condition can be seen, for example, in the extinct sabretooth nimravid Dinictis, and could be an indication of a primitive sabretooth condition where the canines are hypertrophied, but the rest of the skull has not evolved all the changes towards a more derived sabretooth condition. This is consistent with the generally accepted idea that sabrecat evolution was mosaic, not pleiotropic, with enlarged blade-like canines not appearing in concert with other specialized craniomandibular morphologies (Salesa et al., 2005; Slater & Van Valkenburgh, 2008; Christiansen, 2011, but see Meloro & Slater, 2012, who suggest covariation between canine dimensions and skull shape). However, the PCA made by us provides signs of morphological modifications for a sabretoothed condition in the skull of M. dimidiata. According to the loadings for the variables in the PCA (see Table 2), there are four key anatomical features that distinguish M. dimidiata cranial morphology from that of living carnivorous marsupials (see Fig. 4):

  • (1) Upper canine height and anteroposterior length are very large, but the mediolateral width of the canines (C1W) is within the marsupial range. This implies that the canines have a large anteroposterior length and normal mediolateral width, a sabre-like condition observed in fossil sabretooth predators (Biknevicius & Van Valkenburgh, ).
  • (2) Short masseteric fossa length and lever arm of temporalis. This implies short lever arms for the main adductor muscles that would imply low bite force. But short lever arms allow a wide gape without overstretching the masseter and temporalis. Cranial adaptations for wide gape are a distinctive feature of fossil sabretooth predators and generally, the consequence is a low bite force (Wroe et al., ; Christiansen, ; McHenry et al., ). Therefore, M. dimidiata also has some cranial modifications to allow a wide gape such as short lever arm of temporalis and short masseteric fossa length.
  • (3) The distance from the mandibular condyle to M3 is long. This condition improves the separation of molar teeth for any angle of gape and is probably another modification to maintain mandible functionality with hypertrophied upper canines. For the same angular gape, M. dimidiata could process larger pieces of food using its molar tooth. This would have the biomechanical effect of lowering bite force at M3. The lower index of COM3i of Thylacosmilus atrox compared to that of the M. dimidiata and the lower values for COM1i in sabretooth cats compared to those in living felids could reflect mandibular and cervical muscular adaptations of the former to maintain a large bite force in posterior molars with a wide gape (Turnbull, ). This large molar/carnassial bite force may be required for bone crushing in larger specimens, but would be less important for a small sabretooth predator.
  • (4) Short JL. This could be an adaptation to maintain a powerful canine bite, even with the constraints requiring adaptations for a wide gape. The short jaw with long COM3 also suggests that the mandibular tooth row is short and that the canine bite is not weak relative to the molar bite force.
Figure 4.

Plots of skull length to (a) upper canine height/skull length ratio; (b) upper canine anteroposterior length/skull length ratio; (c) jaw length/skull length ratio; and plots of jaw length to (d) condyle to M3/jaw length ratio; (e) masseteric fossa length/jaw length ratio; and (f) moment arm of temporalis/jaw length ratio, for opossum Monodelphis dimidiata, Didelphidae and Dasyuridae. The smallest convex polygonal areas of Didelphidae and Dasyuridae are shown where relevant.

Table 4. Cranial and mandibular indices based on marsupial species, and comparisons with indices from Emerson & Radinsky (1980)
SpeciesIndices based on skull lengthIndices based on jaw length
C1HiC1LiCOM3iJLiOCPHiOCPWiTFLiTFWiTRLiZAWiCOM3iJHiMFLiMATi
  1. aDiscontinuously high index for Neofelis nebulosa.
  2. bData from Emerson & Radinsky (1980).
  3. cDiscontinuously high index.
  4. C1H = upper canine height; C1L = upper canine anteroposterior length; C1W = upper canine mediolateral width; COM3 = condyle to M3; JL = jaw length; OCPH = occipital height; OCPW = occipital width; TFL = temporal fossa length; TFW = temporal fossa width; TRL = tooth row length; ZAW = zygomatic arch width; JH = jaw height; MFL = masseteric fossa length; MAT = moment arm of temporalis.
Opossum Monodelphis dimidiata (three males)1.75–1.871.40–1.551.12–1.171.00–1.031.02–1.121.02–1.030.97–1.071.11–1.140.94–1.001.01–1.031.12–1.171.11–1.170.97–1.051.02–1.09
M. dimidiata female1.181.421.050.981.040.980.880.931.050.911.071.030.970.87
Range marsupials (our sample without M. dimidiata)0.62–1.610.71–1.500.82–1.160.86–1.140.76–1.210.72–1.470.79–1.230.56–1.430.78–1.160.74–1.380.84–1.110.67–1.530.80–1.300.79–1.25
Maximum value of didelphids (our sample)1.611.501.161.071.030.971.151.081.141.001.111.331.111.02
Minimum value of didelphids (our sample)0.650.790.860.860.850.720.790.760.870.830.960.670.800.78
Eleven species of modern didelphidsb0.88–1.380.80–1.180.91–1.060.93–1.020.93–1.070.91–1.080.89–1.070.85–1.130.96–1.060.89–1.100.95–1.050.81–1.11(1.30c)0.90–1.100.81– 1.10(1.21c)
Fourteen species of modern felidb0.87–1.07(1.33a)0.89–1.120.95–1.140.89–1.120.88–1.130.89–1.080.85–1.160.86–1.090.92–1.100.91–1.080.94–1.070.90–1.09
Dinictisb1.521.31.041.040.930.991.151.160.991.010.970.89
Hoplophoenusb1.91.331.031.140.910.861.041.250.91.090.860.85
Eusmilusb2.682.261.111.030.830.620.831.390.791.040.930.81
Barbourofelis morrisib2.11.461.031.080.970.750.841.280.841.30.870.95
Barbourofelis frickib2.852.061.151.1810.580.831.310.891.330.710.9
Homotherium1.521.40.970.890.810.761.060.790.990.79
Machairodusb1.491.290.990.791.10.840.850.80.68
Ischyrosmilusb1.581.520.931.070.950.81.110.82
Smilodonb2.41.670.911.050.980.820.81.130.871.080.690.73
Thylacosmilusb2.261.940.780.861.561.420.680.720.921.020.911.040.551
Machaeroidesb0.980.640.791.351.171.190.830.880.821.270.91.11

The values for MAT/JL, MFL/JL and JL/SL are not outside the ranges of these indices for other marsupials, but the PCA indicates that M. dimidiata is unusual in that it has a combination of large canines with smaller JL, MAT and MFL, whereas other marsupials with large canines have larger values for JL, MAT and MFL. Therefore, M. dimidiata has a combination of features that is shared with sabretooth predators.

It is interesting to note that OCPH/SL and OCHW/SL were among the last indices to be excluded and were large in M. dimidiata. Therefore, this species has a relatively large occiput, suggesting that it may have strong neck muscles to position and stabilize the head while biting.

We conclude that M. dimidiata males have hypertrophied canines, some adaptations for a wider gape and probably a lower bite force in comparison with those of other living marsupials. This morphological pattern is similar to that observed in primitive sabretoothed fossil species (Emerson & Radinsky, 1980; Christiansen, 2006; Slater & Van Valkenburgh, 2008). Therefore, M. dimidiata seems to be a living analogue of the primitive sabretooth condition, such as that found in the nimravid Dinictis and the creodont Machaeroides but not of the more specialized sabretooth predators.

Several studies show that sabretoothed predators had substantially lower bite forces than those of similar-sized predators (Wroe et al., 2005; Christiansen, 2007, Christiansen & Wroe, 2007). Studies of cranial morphology suggest that sabretoothed predators augmented their relatively lower bite forces with a combination of mandibular adduction and head depression using their hypertrophied cranial depressor musculature (Riggs, 1934; Marshall, 1976; Turnbull, 1976; Akersten, 1985; Turner & Anton, 1997; Salesa et al., 2005; Christiansen & Adolfssen, 2007). Here we found that the cranial morphology of M. dimidiata shows the same traits: adaptation for a wide gape that probably reduces bite force, and occipital morphology suggesting powerful neck muscles. Christiansen (2011) suggests that there could be several histochemical and anatomical adaptations to increase the force of mandibular adductor muscles to compensate for reduced lever arms. However, there are no experimental comparative studies of the bite mechanics of M. dimidiata, and the anatomy and physiology of this didelphid are poorly known. Further experimental studies of the bite force of M. dimidiata in comparison with that of other marsupials, and studies of neck and mandible adductor muscles could provide relevant information to improve our understanding of the bite mechanics of fossil sabretoothed predators.

The evolutionary sequence and selective forces resulting in the extreme sabretooth condition remain unclear. The hypertrophied canines of M. dimidiata seem to be strongly selected for agonistic behaviour between males, owing to the strong selective pressure derived from the semelparous condition. The robust forearms (as the inter-epicondylar index indicates, see Table 3) may also be related to intraspecific fighting that involves energetic forearm movements (González & Claramunt, 2000). However, the canines are fully functional for killing large prey as several studies have shown (Busch & Kravetz, 1991; González & Claramunt, 2000), and the canines are relatively the largest among living marsupial predators (see Table 4). When competing for mates and during aggressive encounters, canine display is very important in M. dimidiata and several other carnivorous marsupials (González & Claramunt, 2000; Croft, 2003). Therefore, hypertrophied canines can improve the reproductive success of M. dimidiata males, providing the selective pressures towards a primitive sabretooth condition. Gittleman & Van Valkenburgh (1997) claimed that dimorphic canines are associated with sexual selection pressures. A low level of sexual dimorphism was identified in Smilodon fatalis (Van Valkenburgh & Sacco, 2002; Christiansen & Harris, 2012), but Antón et al. (2004) identified sexual dimorphism in the size of the upper canines of Machairodus aphanistus. Similar studies on more basal sabretooth groups have not been conducted, but a recent study debated the question of sexual dimorphism in sabretooth cats, and claimed that sexual dimorphism could have been important in extinct Felidae (Turner et al., 2011). Perhaps, hypertrophied canines were exaptations for functions other than those related to the ability to kill large prey (see, e.g. Turner & Antón, 1997; Turner et al., 2011 and references therein). Pine et al. (1985) consider that intersexual differences in craniodental dimensions in M. dimidiata could be a response to trophic niche segregation that affected the kind of prey taken by each sex. Christiansen & Harris (2012) proposed a similar explanation for the craniodental sexual dimorphism in Smilodon and Panthera genera.

A general consensus on killing behaviours of fossil sabretoothed predators is that the canines were used to deliver a throat bite that severed the main blood vessels to kill the prey quickly (Turner & Antón, 1997). This behaviour is thought to reduce the likelihood of tooth breakage, while still bringing about the rapid death of large prey (Biknevicius & Van Valkenburgh, 1996; Turner & Antón, 1997; Antón & Galobart, 1999; Salesa et al., 2005). It was suggested that the strong forelimbs of sabretooth predators were needed to restrain the prey before the delivery of the killing bite, thus reducing the probability of canine breakage (Gonyea, 1976; Van Valkenburgh, 1987; Meachen-Samuels & Van Valkenburgh, 2010; Meachen-Samuels, 2012). This could be the case in M. dimidiata when killing prey larger than itself. The humeral head is relatively larger than that of any other marsupial predator, indicating an ability to transmit greater forces through the shoulder. Similarly, the epicondyles are relatively wider, indicating that M. dimidiata has more powerful forearm musculature than that of the other marsupial predators (see Table 3, Supporting Information Appendix S1 and Fig. 5 for a visual comparison with the robust humerus of Didelphis albiventris). Other authors observed similar humeral robusticity in large-prey specialists (Meachen-Samuels & Van Valkenburgh, 2009). The epicondyles are the origin of carpal and digital muscles that facilitate grasping of large prey during capture. Further studies on the forearm of M. dimidiata would allow comparison with other predators, but the constraints on the morphological evolution of the marsupial forelimb and its precocial development for accessing the mother's pouch immediately after birth must be taken into account (Sears, 2004).

Figure 5.

Humeri of (a) opossum Monodelphis dimidiata, specimen MNHN 1331, and (b) Didelphis albiventris, specimen MNHN 4347, for comparison. Both specimens are shown at the same size. Scale bars equal 1 cm. MNHN = Museo Nacional de Historia Natural in Montevideo.

The difficulties of small carnivores to catch prey of their own size or larger were recently analysed theoretically by Carbone, Teacher & Rowcliffe (2007). The observed killing behaviour of M. dimidiata involves extensive manipulation with forelimbs before the bite. In the case of killing mice larger than itself, the bite is described as a single ‘neck bite’ delivered after a long struggle with the forelimbs (González & Claramunt, 2000). This ‘neck bite’ actually refers to a bite to the throat with a low probability of biting the cervical vertebrae (E. González, pers. comm.). This behaviour could be an alternative explanation for the convergent morphological features that M. dimidiata shares with sabretooth predators of the past.

Further evolutionary, behavioural and ecological studies of M. dimidiata and Neofelis spp. will provide a better understanding of these species and of the origin and behaviour of sabretooths in the past. In the case of M. dimidiata its manageable size makes it a very suitable species to be studied further in captivity.

Acknowledgements

We are grateful to PEDECIBA and ANII for financial support and Enrique González for useful comments on M. dimidiata behaviour.

Ancillary