A skull of Machairodus horribilis and new evidence for gigantism as a mode of mosaic evolution in machairodonts (Felidae, Carnivora)

Abstract

Sabertooth cats were extinct carnivorans that have attracted great attention and controversy because of their unique dental morphology representing an entirely extinct mode of feeding specialization. Some of them are lion-sized or tiger-sized carnivorans who are widely interpreted as hunters of larger and more powerful preys than those of their modern nonsaber-toothed relatives. We report the discovery of a large sabertooth cat skull of Machairodus horribilis from the Late Miocene of northwestern China. It shares some characteristics with derived sabertooth cats, but also is similar to extant pantherines in some cranial characters. A functional morphological analysis suggests that it differed from most other machairodont felids and had a limited gape to hunt smaller preys. Its anatomical features provide new evidence for the diversity of killing bites even within in the largest saber-toothed carnivorans and offer an additional mechanism for the mosaic evolution leading to functional and morphological diversity in sabertooth cats.

Full text

古 脊 椎 动 物 学 报
VERTEBRATA PALASIATICA
A skull of Machairodus horribilis and new evidence for
gigantism as a mode of mosaic evolution in machairodonts
(Felidae, Carnivora)
DENG Tao1,2,3 ZHANG Yun-Xiang3 Zhijie J. TSENG4 HOU Su-Kuan1
(1 Key Laboratory of Vertebrate Evolution and Human Origins of Chinese Academy of Sciences, Institute of Vertebrate
Paleontology and Paleoanthropology, Chinese Academy of Sciences Beijing 100044, China dengtao@ivpp.ac.cn)
(2 CAS Center for Excellence in Tibetan Plateau Earth Sciences Beijing 100101, China)
(3 Department of Geology, Northwest University Xi’an, Shaanxi 710069, China)
(4 Division of Paleontology, American Museum of Natural History New York, NY 10024, USA)
Abstract Sabertooth cats were extinct carnivorans that have attracted great attention and
controversy because of their unique dental morphology representing an entirely extinct mode of
feeding specialization. Some of them are lion-sized or tiger-sized carnivorans who are widely
interpreted as hunters of larger and more powerful preys than those of their modern nonsaber-
toothed relatives. We report the discovery of a large sabertooth cat skull of Machairodus
horribilis from the Late Miocene of northwestern China. It shares some characteristics with
derived sabertooth cats, but also is similar to extant pantherines in some cranial characters. A
functional morphological analysis suggests that it differed from most other machairodont felids
and had a limited gape to hunt smaller preys. Its anatomical features provide new evidence for
the diversity of killing bites even within in the largest saber-toothed carnivorans and offer an
additional mechanism for the mosaic evolution leading to functional and morphological diversity
in sabertooth cats.
Key words Gansu, China; Late Miocene; sabertooth cat; skull; predatory behavior
Citation Deng T, Zhang Y X, Tseng Z J et al., 2016. A skull of Machairodus horribilis and new
evidence for gigantism as a mode of mosaic evolution in machairodonts (Felidae,
Carnivora). Vertebrata PalAsiatica, 54(4): 302−318
1 Introduction
Sabertooth cats (Machairodontinae) were widely distributed in the Neogene and
Quaternary faunas of the Old and New Worlds (Werdelin, 1996; McHenry et al., 2007). They
were a long-living extinct clade among carnivorans, appearing from the Middle Miocene,
and became extinct after a brief period of coexistence with early human beings in the Early
Holocene (for example, at Rancho La Brea in Los Angeles, California, USA) (Turner and
Antón, 1997). Some scenarios for the demise of the largest sabertooth cats include one where
early human beings had gradually developed stronger hunting ability to outcompete the
国家自然科学基金重点项目(批准号：41430102)、国家重点基础研究发展计划项目(编号：2012CB821906)和
中国科学院战略性先导科技专项(编号：XDB03020104)资助。
收稿日期：2015-11-06
第54卷 第4期
2016年10月
pp. 302−318
gs. 1−3

4期 303
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
sabertooth cats, and/or sabertooth cats themselves becoming the prey of human beings (Martin,
1989). Like living big cats (Pantherinae), the large size of sabertooth cats is an advantage
in predatory activity (Turner and Antón, 1997; Antón and Galobart, 1999; Andersson et al.,
2011; Salesa et al., 2005), and some of them had a lion-like or tiger-like size (adult body
mass even more than 400 kg for Smilodon populator) (Christiansen and Harris, 2005; de
Castro and Langer, 2008). On the other hand, the early materials of sabertooth cats were
mostly fragmentary, therefore body size disparity and evolution throughout the machairodont
felids are not well characterized. Only recently, several complete skulls of machaiodonts
were described from Europe and China (Antón et al., 2004; Geraads et al., 2004; Qiu et al.,
2008), among which Machairodus horribilis was considered to be the largest one (Qiu et
al., 2008). The sizes of these machairodonts are judged according to their skulls with basilar
lengths from 285 mm to 299 mm, and within this large-size class there are additional distinct
morphological differences. Among them, some forms are similar to the highly specialized
Smilodon and Homotherium, whereas others are more similar to extant pantherines. These
important materials of sabertooth cats document mosaic evolution and the resulting signicant
morphological and ecological diversity in the evolutionary radiation of these carnivorans
(Antón, 2013).
Here we report on a large sabertooth cat skull found to date from the Late Miocene
Hipparion Red Clay at Longjiagou in Wudu County, Gansu Province, China, with the
geographical coordinates of 33°35′N, 104°50′E (Zhang and Xue, 1995). The Longjiagou Basin
is located on a planation surface of 2400-2600 m, which formed in the Late Miocene, and it is
a small intramontane basin with an area of about 15 km2, with lengths of about 6.5 km in north
and south and 1–2.5 km in east and west. The fossiliferous bed is dark red silty mudstone,
with blue-gray sandstones, lower part of which is mudstones with conglomerates, 20–340 m
in total thickness. A comprehensive study about the Longjiagou Hipparion fauna is currently
being carried out. From the present classication of the specimens, it shows that most elements
of this fauna are typical fossil mammals of Baodean. Therefore biostratigraphically the fauna
is almost certainly of Baodean age. However, the fauna includes younger faunal elements,
for example, Gazella cf. G. blacki, which are found mainly in the Early Pliocene. Hipparion
platyodus from Wudu is more derived than other Turolian Hipparion in other regions of China.
As a result, the age of the Longjiagou fauna is tentatively assigned to late Baodean (Zhang and
Xue, 1995; Qiu et al., 2013; Deng et al., 2013).
Institutional abbreviations Ath. Nr. fossil number of Museum of Palaeontology and
Geology, University of Athens; HMV. vertebrate fossil number of Hezheng Paleozoological
Museum; IVPP. Institute of Vertebrate Paleontology and Paleoanthropology, Chinese Academy
of Sciences; LACM. Natural History Museum of Los Angeles County; NWU. Northwest
University; V. vertebrate fossil number of IVPP.

304 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
2 Systematic paleontology
Order Carnivora Bowdich, 1821
Family Felidae Fischer, 1817
Subfamily Machairodontinae Gill, 1872
Genus Machairodus Kaup, 1833
Machairodus horribilis Schlosser, 1903
(Figs. 1, 2; Tables 1, 2)
Referred material NWU 48Wd0001, an adult skull, laterally compressed (Figs. 1, 2).
Because crushing has in all cases been lateral, measurements along the long axis of the skull
Fig. 1 Skull of Machairodus horribilis (NWU 48Wd0001) from Longjiagou (Wudu, Gansu Province, China)
A. dorsal view; B. reconstruction in dorsal view; C. lateral view; D. reconstruction in lateral view;
E. ventral view; F. reconstruction in ventral view
Abbreviations: ab. auditory bulla; apf. anterior palatine foramen; C. upper canine; coc. central occipital crest;
eam. external auditory meatus; F. frontal bone; s. fossa for lacrimal sac; fm. foramen magnum;
fpp. frontal postorbital process; gf. glenoid fossa; I3. upper third incisor; if. infraorbital foramen; M. maxillary bone;
M1. upper rst molar; mf. maxillary foramen; mp. mastoid process; mrb. median ridge of the basioccipital;
N. nasal bone; nfs. naso-frontal suture; no. nasal opening; oco. occipital condyle; ocr. occipital crest; opf. optic foramen;
orf. orbital foramen; P2. upper second premolar; P3. upper third premolar; P4. upper fourth premolar; pc. parietal crest;
pf. palatine ssure; pg. palatine groove; plf. posterior lacerated foramen; Pm. premaxillary bone; pn. posterior nares;
pog. postglenoid process; pp. paroccipital process; pr. palatine ridge; prg. preglenoid process; sc. sagittal crest;
spf. sphenopalatine foramen; tc. temporal crest; tf. temporal fossa; zpp. zygomatic postorbital process

4期 305
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
Table 1 Selected measurements and comparison of the sabretooth cat specimens (mm)
M. horribilis M. horribilis M. giganteus M. aphanistus S. fatalis H. crenatidens M. nihowanensis
(male) (female)
NWU 48Wd0001 V 15642 Ath.Nr.1967/6 Maximum LACM 67/D817 HMV 1213 HMV 1220
Qiu et al., 2008 Melentis, 1968 Antón et al., 2004 Qiu et al., 2004 Qiu et al., 2004
Vertex L 415 353 355 348* 356.5 330 265
Condylobasal L 370 318 325 305
Basilar L 348 299 285 313.2 300 275 242
Palatal L 163 163 160 146* 151 ~142 118
Width at I3 ~63 60 48* 54 58*
Width at C-C ~100 90 64 71* 98 >65 70
Width behind P4 134 138 104 92* 134 108*
Width at postorbital process ~120 132 118 96* 115
Occipital H 124 120 123* 104.5 94*
Width at mastoid process ~136 127 128 11 3 135
H at mastoid process ~163 142 122 117* 149 124*
I3 L×W 17×15 12.6×11 19×12
C1 L×W 48.4×18.4 39.5×16 35.2×14.3 35.7×14.5 36.7×16.7 30.2×13.5 27.5×14
P3 L×W 26.6×12 24.3×9.6 23.7×10.6 23.7×12 19.8×9 12.2×6
P4 L×W 43×17 42.2×15 43.1×14.8 36.4×17.4 44.4×16.7 44.8×14 33.7×13.7
* Measured from gures.

306 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
are generally reliable, but overall width measurements are impossible to take in most cases
(Table 1). Among its anterior dentition, only lower parts of both the right I3 and canine crowns
are preserved. The zygomatic arches are lost. The right cheek teeth are relatively fragmentary,
with a M1 root. The left P4 lacks its crown apices.
Locality and horizon Longjiagou Town, Wudu County, Gansu Province in northwestern
China; Late Miocene Hipparion Red Clay.
Description and comparisons NWU 48Wd0001 represents an adult individual, given
by the eruption and wear of the teeth and the fusion of the bone sutures. With a cranial vertex
length of 415 cm, this skull obviously exceeds the length of all known Late Miocene skull
specimens of sabertooth cats, including other Machairodus horribilis (353 mm, V 15642) and
also M. giganteus (355 mm, Ath. Nr. 1967/6). This skull most likely represents an adult male
of M. horribilis (Table 2).
In dorsal view (Fig. 1A, B), two nasal bones compose a hexagon that is longer than
wide, and whose antero-posterior mid-points are projected laterally, rather than narrowing
posteriorly as in pantherines (Antón et al., 2004). The posterior half of the naso-frontal suture
is perpendicular to the sagittal axis of the skull, and the sagittal process of the frontal is
absent, but the lateral process, which inserts into the nasal and maxillary bones, is marked,
with a length of 22.5 mm. The temporal fossa is antero-posteriorly elongated, which occurs
consistently in some primitive sabertooth cats (Salesa et al., 2005). The dorsal outline of the
frontal region at the level of the postorbital processes is concave, which is similar to those of
some primitive sabertooth cats. The shape of the nasal opening is intermediate between the
heart-shaped outline observed in pantherines and the rectangular shape typical of more derived
machairodontines like Homotherium and Smilodon (Antón et al., 2004). The frontal bone is
wide and penetrated deeply into the maxilla, which is similar to primitive sabertooth cats; its
postorbital process is strong and thick to form a low and short triangular pyramid, without a
sharp tip, and with a large rough surface that extend both anteriorly and posteriorly, which is
similar to derived sabertooth cats (Qiu et al., 2004). The parietal crests converge into a single
sagittal crest at the level of the external auditory meatus, and anterior to the convergent point is
a weak sagittal groove, unlike the obvious depression at this position in M. aphanistus (Antón
et al., 2004). The muzzle may be narrower at the level of the canine alveoli than that of the
lion, due in part to the much more attened section of the upper canines.
In lateral view (Fig. 1C, D), the dorsal profile is about as convex as in the lion, but
different from the straighter dorsal profiles in more derived machairodontines (Antón et
al., 2004). The occipital plane has a great inclination, which is exhibited in some primitive
sabertooth cats (Salesa et al., 2005). The alveolus region of incisors is not strongly projected
forward so that the antero-dorsal margin of the premaxillary bone is weakly curved and
declined. The infraorbital foramen is large and rounded, which is similar to those of derived
sabertooth cats in shape and size, and its posterior border is located above the parastyle of P4.
The sagittal crest rapidly becomes high as a vertical plate with a maximum depth of about

4期 307
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
58 mm, which is the most striking difference from living pantherine cats. The postglenoid
and mastoid processes are widely separated (8.5 mm), especially on their distal ends, so that
the external auditory meatus is not closed as in pantherines, the external auditory meatus is
not enclosed between the postglenoid and mastoid processes. The mastoid process is robust
and strongly extends inferiorly, covering the auditory bulla laterally. The mastoid and the
paroccipital processes are also widely separated, but connected by a wide curved edge.
In ventral view (Fig. 1E, F), the diastema between I3 and C is 12 mm, slightly shorter
than the length of I3 (15 mm), and its labial margin is weakly concave, so the alveolus margins
of the incisors and canine are obviously separated. The palate is sunken below the cheek
teeth so that their roots are exposed. There is a pair of projecting ridges that extend from the
postero-medial corner of the palatine ssure to the medial side of the anterior palatine foramen
as the medial boundary of the palatine groove, with a premaxillary section which is markedly
crest-like, becoming more blunt in the maxillary and even more so in the palatine. The well-
developed palatine ridge is related to the gripping device (Antón and Galobart, 1999). The
posterior margin of the palate is U-shaped, without a sagittal process and a marked sagittal
indentation seen in pantherines, and its bottom is located at the level of M1. Because the skull
is compressed laterally, the original shape of the posterior nares cannot be known.
The median ridge of the basioccipital extends anteriorly to reach the level of the anterior
end of the auditory bullae, which is similar to those of derived sabertooth cats. The posterior
half of the basioccipital part is a convex triangle, and smoothly connects with the occipital
condyles; the central crest is not very strong, and both sides are deep depressions for the
insertion of the rectus capitis anticus major muscle, whose anterior tip disappears between a
pair of rough swellings in attachment for the rectus capitis ventralis muscle (Fig. 2), which
is more anteriorly than in pantherines. Although the terminal of the postglenoid process is
broken, it is still judged to strongly extend ventrally, so the posterior surface of the glenoid
fossa is almost vertical. The preglenoid process is well developed and laterally located, being
different from the crest-like anterior border in more derived machairodontines (Christiansen,
Table 2 Index of sexual dimorphism (male measurement value: female measurement value) for
Machairodus horribilis and several other felid species
NWU 48Wd0001:
IVPP V 15642
IVPP V 15643:
IVPP V 15642
Machairodus aphanistus
Antón et al., 2004
Smilodon fatalis
Van Valkenburgh and Sacco, 2002
Panthera leo
Antón et al., 2004
Vertex L 1.18
Condylobasal L
1.16 1.24 1.06 1.12
Basilar L 1.16
Palatal L 1.00
I3 L 1.36 1.37
I3 W 1.35 1.29
C L 1.23 1.12 1.22 1.09 1.25
C W 1.15 1.26 1.16 1.12 1.23
P3 L 1.09 0.99
P3 W 1.25 1.00
P4 L 1.02 1.11
P4 W 1.13 1.31

308 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
2013). As a result, the glenoid fossa is very deep as in the modern lion and leopard, suggesting
comparable ranges of jaw motion in M. horribilis. In ventral view, the fossa is elongated as in
pantherines, rather than widening medially as in derived sabertooth cats. The space between
the postglenoid and mastoid processes is broad, which is similar to that of pantherines, but
Fig. 2 Teeth and basioccipital region of Machairodus horribilis (NWU 48Wd0001)
A. upper cheek teeth in occlusal view; B. upper anterior teeth in occlusal view;
C. basioccipital region in ventral view
Abbreviations: acc. anterior cingular cusp; C. upper canine; cmt. canalis musculotubarius; et. entotympanics;
hf. hypoglossal foramen; hjf. hyojugular fossa; I1. upper first incisor; I2. upper second incisor; I3. upper
third incisor; M1. upper rst molar; mcv. attachment area for musculus rectus capitis ventralis; mp. mastoid
process; mrb. median ridge of the basioccipital; ms. metastyle; ob. antero-medial opening of bulla; P2. upper
second premolar; P3. upper third premolar; P4. upper fourth premolar; pa. paracone; pac. posterior accessory
cusp; pcc. posterior cingular cusp; pcf. doubled posterior carotid foramen; pe. posterior edge; pgf. postglenoid
foramen; plf. posterior lacerated foramen; pp. paroccipital process; pps. pre-parastyle; prc. principal cusp;
pro. protocone; ps. parastyle; se. serration; smf. stylomastoid foramen. Scale bars equal 2 cm

4期 309
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
not as in later machairodontines, where both processes tend to come ever closer to each other,
almost touching in Smilodon (Antón et al., 2004). The temporal crest is more developed than
in pantherines and weaker than in later machairodontines. The mastoid process is especially
robust and located at the anterior 3/4 of the auditory bulla, with a laterally aring mastoid crest
as in derived sabertooth cats. The insertion of the atlantomastoid muscles, which is located
under this crest, is thus enlarged, and it becomes oriented more inferiorly and less laterally
than in pantherines (Antón et al., 2004). The paroccipital process is located near the postero-
lateral end of the auditory bulla, with a long distance from the mastoid process, which equals
to the distance from the occipital condyle.
In occipital view, the occipital surface is a high triangle with a pair of laterally projecting
processes near its top, and the occipital crest also surrounds a triangle on the top of the
occipital surface, as in pantherines. The central smooth region is divided into two pairs of
depressions: one is located laterally to the occipital condyle, and the other is located superiorly
to the occipital condyle. Below the top margin and on the each side of the central crest, there
are deep depressions for the attachment of muscles, which are absent in pantherines. Lateral to
the upper margin of the foramen magnum there is a pair of vertical edge-like processes, weaker
than tubercles at the same position in pantherines.
The incisors are arranged in gentle arch, instead of a highly curved arrangement in more
derived machairodontines, but those teeth exhibit several rudimentary characteristics that are
better developed in derived machairodontines (Qiu et al., 2004). Judged from the alveolus, the
incisors are enlarged from I1 to I3 gradually, but I3 is obviously much larger and more robust,
with a weak antero-lingual edge and a sharp postero-labial edge, and without serration when
it is worn. The smaller rst and second incisors and the larger third incisor set in a slight arch,
only slightly more developed than the straight incisor rows of the primitive sabertooth cats
(Christiansen, 2013).
Both the anterior and posterior margins of the huge upper canine of M. horribilis are
serrated, with higher height in the posterior edge (Fig. 2B), similar to the serration seen in
other sabertooth cats, such as Homotherium (Qiu et al., 2004). The upper canine is large and
long, with an antero-posterior basal length of 48.4 mm and a thickness of 18.4 mm. The labial
swelling of the crown is slightly stronger than the lingual one, and the cross section of the
crown is wider in the anterior end than in the posterior end.
P2’s alveolus has a length of 8 mm and a width of 4.5 mm, with a single root, as in
pantherines. The diastema is 7 mm between P2 and C and about 4 mm between P2 and P3. P3
has double roots, a principal cusp, a posterior accessory cusp, and a posterior cingular cusp (Fig.
2A), the latter is absent in both Panthera leo and M. giganteus. The labial border is straight in
the middle and anterior parts, but curved labially at the posterior part, whereas in P. leo and M.
giganteus that border is straight; the lingual border is slightly concave in the middle and curved
labially in the posterior, whereas it is projecting lingually in M. aphanistus (Antón et al., 2004).
There is a broken scar at the lingual base between the principal and posterior accessory cusps,

310 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
which is seem to indicate a prominence in its original state. Both the anterior and posterior
edges of the principal cusp are serrated. The posterior accessory cusp is much lower than the
principal cusp, and the posterior cingular cusp is also much lower than the posterior accessory
cusp.
P4 has a pre-parastyle and a very rudimental protocone supported by an independent root
(Fig. 2A). This carnassial tooth has complete enamel to cover its crown, except for the broken
cusps’ apices. In the crown base, there is a V-shaped groove between the antero-labial root and
the protocone root. There is a marked depression in front of the protocone, and the crown part
of the protocone is a weak swelling on the lingual wall, instead of an isolated cone, similar to
the condition seen in derived sabertooth cats (Qiu et al., 2004). In comparison, M. aphanistus
has a well-developed protocone (Antón et al., 2004). Among the extant felids, the protocone
is well developed in all species except in the cheetah, Acinonyx jubatus (Ficcarelli, 1984). At
the base of the labial wall of the pre-parastyle, there are three tiny granular tubercles. On the
posterior part of the labial wall of the metastyle, there is a projecting edge oblique posteriorly.
The root indicates that M1 is located in the lingual side of the posterior end of P4 (Fig. 1).
It is oval in shape with a labial-lingual diameter of 8.7 mm and an antero-posterior diameter of
7 mm, longer but narrower than that of M. palanderi and other known M. horribilis (Qiu et al.,
2008).
3 Body weight estimation of Machairodus horribilis
In order to determine the body mass of Machairodus horribilis, we utilized allometric
equations relating weight with skull measurements to estimate body mass. Body mass is an
ecologically relevant characteristic like life history traits, diet, population density, population
growth rate, home range size, and behavioral adaptations. In fact, based upon modern
knowledge, a mammal’s body size may be the most useful single predictor of that species’
adaptations (Damuth and MacFadden, 1990).
An entire book (Damuth and MacFadden, 1990) has been written on the challenges
associated with estimating body weights from mammalian osteological remains. Least squares
regression of log10 transformed data is used to model the association between body mass and
skeleton. The regression of log body weight (W) against log condylobasal length (CBL) for
Felidae (Van Valkenburgh, 1990) is chosen to calculate the body weight of M. horribilis:
log W = 3.11log CBL – 5.38
The correlation coefcient (r) of condylobasal length and body weight is high (0.92). The
percent prediction error (%PE) is 38, and the percent standard error of the estimate (%SEE) is 57.
According to the size of the skull (NWU 48Wd0001, Table 1), M. horribilis from
Longjiagou is estimated to have had a body weight of about 405 kg as a living animal.

4期 311
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
4 Phylogenetic analysis
In order to investigate the phylogenetic position of the newly described specimen of
M. horribilis (NWU 48Wd0001), we added this specimen and the material of M. horribilis
described by Qiu et al. (2008) to a recently published comprehensive dataset on sabertooth cat
phylogeny (Christiansen, 2013:appendix 3) to conrm their Machairodontine afnities (Table
3). Scorings for M. horribilis from Longjiagou were based on the skull (NWU 48Wd0001) and
for M. horribilis from Baode were based on skull and mandible (IVPP V 15642). We removed
Dinictis from the data matrix because of the problematic phylogenetic position of nimravids
highlighted by recent cladistics analyses of Carnivoramorpha employing postcranial characters
(Spaulding and Flynn, 2012). However, Canis lupus and Cryptoprocta ferox were retained as
caniform and feliform outgroups, respectively. The data matrix was analysed using the TNT
software package (Goloboff et al., 2008) and PAUP* (Swofford, 1991). Analyses in TNT were
run using both implicit enumeration and traditional search, with default settings apart from the
following: 99999 maximum trees in memory and 1000 replications. The analysis resulted in 19
equally parsimonious trees, each having a length of 116 steps, a CI of 0.621, and a RI of 0.805.
Analyses in PAUP* were run with the heuristic search option with all default settings except
for NREPS set to 1000 replications. The analysis resulted in 65 equally parsimonious trees,
each having a length of 116 steps, a CI of 0.638, and a RI of 0.829. Both TNT and PAUP*
produced cladograms of identical topologies. We next estimated support for the clades present
in the strict consensus tree by running bootstrap and jackknife resampling analyses in both
programs with 1000 replications. In addition, we calculated decay index (or Bremer support)
in both programs. We also calculated support values using the symmetric resampling function
in TNT. Lastly, we analyzed the dataset under a Bayesian framework using MrBayes version
3.2 (Ronquist et al., 2012). The morphology matrix was treated as a single partition assigned
with the Mk model for morphology. The analyses were run for 10 million generations with 8
simultaneous chains, sampling every 1000 trees. The strict consensus tree is shown in Fig. 3,
with support values indicated for the major clades.
The unambiguous synapomorphies for the monophyletic sabertooth cats (Machairodontinae)
:
small c1, very small knob-like M1, and a large P3 parastyle (Christiansen, 2013) are seen in
both specimens of M. horribilis. The analysis of the sabertooth cat dataset place M. horribilis
and M. giganteus as unresolved taxa at the base of a weakly supported Homotherini +
Smilodon clade. The basal position of M. aphanistus and more crownward placement of M.
giganteus and M. horribilis in the consensus tree are consistent with the result of Christiansen
(2013). Because M. aphanistus and M. giganteus do not form a monophyletic group in the
Table 3 Character codes of Machairodus horribilis according to Christiansen (2013, appendix 3)
Character 11111111112 2222222223 3333333334 4444444445
1234567890 1234567890 1234567890 1234567890 1234567890
A. horribilis (IVPP V 15642) 1220121111 2110111121 11100 11111 1101311011 11110 11011
A. horribilis (NWU 48Wd0001) 122?121111 211011112? ?????????1 11013110?1 11?10?????

312 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
Fig. 3 A strict-consensus tree of sabretooth felids
The same topology was recovered in PAUP* (MPT=65, 116 steps, CI=0.638, RI=0.829) and TNT (MPT=19,
116 steps, CI=0.621, RI=0.805) analyses. Clade stability and support values from Bootstrap (BS, n=1000),
jackknife (JK, n=1000), symmetric resampling (SR, n=1000), and decay index (DI) analyses are indicated.
In addition, posterior probabilities from a Bayesian analysis of the morphological characters under the Mk
model are provided. The base matrix was taken from Christiansen (2013:appendix 3) with the addition of
Machairodus horribilis from Baode (female) and Wudu (male). Highly supported (support>90% and DI>=4)
clades are indicated in bold. Position of specimen described in this study is also indicated in bold
consensus tree, Christiansen (2013) used the genus name Amphimachairodus proposed
previously by Kretzoi (1929) for the latter species. On the other hand, we consider that some
scorings of Christiansen (2013) are needed to check personally by ourselves in future, so the
genus name Machairodus is still retained for these species tentatively.
5 Predatory behavior
A phylogenetic analysis places this specimen among the Eumachairodontia clade, and
allied with other homotherins (Fig. 3). In combination with other recent discoveries, such as
those of smaller, presumably female individuals of M. horribilis from the Late Miocene of
Baode in Shanxi Province, China (Table 2; Qiu et al., 2008), the new nd indicates that these
machairodonts likely relied on unspecialized throat bites (Turner and Antón, 1997) in the Late

4期 313
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
Miocene woodland or steppe of the northwestern China to subdue their prey, but nevertheless
with enough clearance between their upper and lower canines to prey on the most common
contemporaneous ungulates, permitted by their gigantic size.
Gigantism affects many aspects of animal structure and function (Xu et al., 2012). Among
living meat eaters, almost all species larger than about 21 kg prey on species as large or larger
than themselves because of energetic demands (Carbone et al., 1999). Because sabertooth cats
were generally strongly built animals, it has been suggested that they specialized in taking
larger prey than extant pantherines (Akersten, 1985; Turner and Antón, 1997). Differences
among carnivoran species in killing and feeding behavior are often reflected in their
craniodental morphology (Biknevicius et al., 1996). It was even once proposed that saber-
like teeth evolved convergently in saber-toothed mammals and allosaurid dinosaurs as an
adaptation for predation on prey species ten or more times heavier than the predators (Bakker,
1998). Larger predators specialize on larger prey (Radloff and du Toit, 2004). Given these
observations and interpretations, does the gigantic body size of M. horribilis indicate that it
must hunt very large preys, or does it account for a unique killing mechanism?
Indeed, there are several features indicating that M. horribilis killed its prey by penetrating
the esh of the throat with its canines and causing massive blood loss, just as has been inferred
for the derived sabertooth cats (Turner and Antón, 1997; Wroe et al., 2005; McHenry et al.,
2007). These features include the high crowned, attened and serrated upper canines, which
are well adapted to penetrate the esh of prey and would be less suitable for either a crushing
nape bite or a suffocating bite (Bryant and Churcher, 1987); the enlarged mastoid crests, slight
anteroposterior projection of the mastoid process, and median crest on the basicranium indicate
locations of muscle insertions consistent with adaptations for a canine shear-bite (Antón et
al., 2004). The elongated and inclined occipital region further allows the movement of the
temporalis to scribe a larger arc as the jaw closes, increasing the force generated at the anterior
teeth (incisors and canines) rather than at the carnassials (Martin et al., 1999).
On the other hand, significant differences that exist among the various taxa of saber-
toothed carnivorous mammals have been suggested to correlate with behavioral and ecological
diversity (Martin, 1980; Bryant and Churcher, 1987, Turner and Antón, 1997; Martin et al.,
1999). If a throat bite was the main killing technique of sabertooth cats, then killing of a larger
prey required a larger gape. For example, Smilodon has an enormous gape of 120° (Antón
et al., 2004; Andersson et al., 2011). On the other hand, M. horribilis has the strong pre- and
postglenoid processes to make its glenoid fossa very deep, which is similar to the pantherine,
such as the lion and leopard, so it has only a moderate gape of about 70° (Antón et al., 1998;
Andersson et al., 2011). Moreover, the inclined occiput of M. horribilis indicates that the
fibers of the temporal muscle are strongly inclined as in primitive cats, while in derived
machairodontines these fibers become more vertically oriented (Antón et al., 2004). The
larger degree of inclination in M. horribilis represents a limitation for dorsal extension of the
head over the atlas, therefore in turn limits the head action required in the canine-shear bite

314 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
(Antón et al., 2004; McHenry et al., 2007; but see recent discussion of Smilodon mechanism
by Brown, 2014). Although the strongly developed mastoid crest of M. horribilis provides a
larger area for insertion of the atlanto-mastoid muscles and indicates that the strength of these
muscles would be greater than in pantherines, contributing to the potential for head depression,
it is less efcient than in more derived machairodontines (Antón et al., 2004).
As a result, the functional morphology suggests that the biting or killing mechanism of
M. horribilis differs from more highly specialized sabertooth cats, but in ways similar to that
of extant lions and leopards and primitive, early felids. Specically, the limited gape and the
intermediate development of musculature arrangements compared to more derived sabertooth
cats would have restricted M. horribilis to somewhat smaller prey sizes than more derived
sabertooth cats with larger gapes. However, given its gigantic size, which effectively increases
the absolute distance between upper and lower canines at maximum gape (assuming canine
length scales proportionally to gape), M. horribilis nevertheless achieved some functionality
of the sabers in biting and killing that is more specialized than less derived machairodontines
such as M. aphanistus but not yet at the stage of derived homotherins such as Homotherium.
The highly mosaic evolution of sabertooth cats exhibits increasingly rened adaptations
for the canine-shear bite in order to kill its prey with more efciency. Efciency is important
because the ability to kill prey faster translates to less struggling time and less opportunity
for accidents involving teeth breakage and/or prey escape (Van Valkenburgh, 1988; Van
Valkenburgh and Ruff, 1987). The canines (length × width = 48.4×18.4 mm, index=2.63)
of M. horribilis are at least as attened as those of M. giganteus and M. aphanistus (Antón
et al., 2004:table 3), and thus equally fragile and breakable (Van Valkenburgh, 1988; Van
Valkenburgh and Ruff, 1987). Like hyaenids, the robust incisors, especially I3, and the less
parabolic incisor arcade of M. horribilis may function to reinforce the adjacent canines during
the killing bites by helping locally to limit the motion of the prey (Biknevicius et al, 1996).
The morphological features discussed above suggest a model of predatory behavior of M.
horribilis that differs not only from that of extant pantherines but also from that hypothesized
for derived machairodontines such as Homotherium and Smilodon. Faunal evidence for a great
number of slow-running three-toed horses in the Longjiagou fauna that belong to a dwarf species
Hipparion platyodus with very short limb bones (Qiu et al., 1987; Zhang and Xue, 1995)
provides at least one candidate species as the possible principal prey of M. horribilis. Although
also present in this fauna (Zhang and Xue, 1995), the deer Eostyloceros and Cervavitus
may have been too small and too fast for M. horribilis, and the giraffes Samotherium and
Honanotherium may have been too large for the felid predators.
The mosaic combination of highly derived upper canines and primitive cranial
morphology in M. horribilis demonstrates that even though high-crowned, flattened sabers
may work best as part of a complex of adaptations for the canine-shear bite, they can also
work, and do work sufciently well, within a different, more feline like mode of biting when
combined with gigantic size. The key advantage of the initial development of saber-like

4期 315
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
teeth would lie in the efficiency of a killing bite that caused massive blood loss instead of
suffocation, rather than in the possibility of taking signicantly larger prey (Turner and Antón,
1997). It is reasonable, however, that more derived sabertooth cats with a larger gape and more
specialized musculoskeletal complex than M. horribilis would have been able to take prey
larger relative to their body size. The extent of gigantism as a benecial addition to a mosaic
mode of improving sabertooth functionality may have been limited in sabertooth cat evolution
given the relatively short stratigraphic range in which very large-bodied Machairodus occur.
This discovery has implications for early predatory behavior evolution of sabertooth cats,
and indicates that at least one sabertooth cat incorporated gigantic size in the functional mosaic
of musculoskeletal features associated with the sabertooth bite. The unique combination of a
restricted gape and very large size in M. horribilis could represent an adaptation to a particular
prey size class, given the multitude of paleoenvironments it has been found in. M. horribilis
lived both in woodlands of Longjiagou and also in completely open grasslands (for example
in Baode). As such M. horribilis likely was sympatric with forest mammals such as primates
(Xue and Delson, 1989), chalicotheres (Xue and Coombs, 1985), and the deer Eostyloceros
(Zhang and Xue, 1995) as well as more typical open grassland forms. The dwarf horses of the
Longjiagou fauna are morphologically different from the tall Hipparion horses in Baode and
other regions, in having a less cursorial postcranial skeleton that may have provided stable
food supply for the population containing the largest known Machairodus. Different from the
rich girafds in other localities, the girafds were rare for individuals and low in taxonomic
diversity at Longjiagou. The predatory behavior of large carnivorans may be interpreted
according to the mass-energetics “law” (Carbone et al., 1999), but the encounter rate for prey
is also an important factor. Moreover, the height of giraffe increased its predator detection
capability and threat of injury to predators from its hooves, thus it is not very suitable as a
regular prey of sabertooth cats (Hayward and Kerley, 2005). Gigantic Late Miocene sabertooth
cats elsewhere, by contrast, lived in an open steppe that was conducive for them to pursue
preys at a burst of high speed and attack preys with a very large gape.
In conclusion, M. horribilis from Longjiagou has the largest skull of any sabertooth
cat, but it likely did not exhibit the predatory behavior of derived taxa such as Smilodon or
Homotherium, and instead hunted comparatively smaller preys. Derived predatory behaviors
apparently have evolved several times independently in sabertooth cats along with changes of
habitats and preys through the whole evolutionary history of sabertooth cats (Antón, 2013),
as has clearly occurred in some ungulates, especially tooth crown changes in the family
Equidae (Mihlbachler et al., 2011). The mixture of primitive and derived morphological
characteristics in the cranium of M. horribilis is consistent with previously observed mosaic
evolutionary patterns in early machairodontines, and additionally provides evidence that
gigantism may be one of several mechanisms to increase gape prior to the evolution of the
full suite of anatomical features associated with more efcient killing bite mechanism (Antón,
2013). Future discoveries of postcranial elements belonging to M. horribilis would allow these

316 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
functional morphological interpretations to be further tested.
Acknowledgments We thank Prof. Xue Xiang-Xu for her contribution to the collection of the
Longjiagou fauna, and Chen Yu for his illustrations of this specimen. This work was supported
by the National Natural Science Foundation of China (41430102), the Strategic Priority
Research Program of the Chinese Academy of Sciences (XDB03020104), and the Ministry of
Science and Technology of China (2012CB821906).
记恐剑齿虎一头骨及剑齿虎镶嵌进化中体型巨大化的新证据
邓 涛1,2,3 张云翔3曾志杰4侯素宽1
(1 中国科学院古脊椎动物与古人类研究所，中国科学院脊椎动物演化与人类起源重点实验室 北京 100044)
(2 中国科学院青藏高原地球科学卓越创新中心 北京 100101)
(3 西北大学地质系 西安 710069)
(4 美国自然历史博物馆古生物部 纽约 NY 10024)
摘要：剑齿虎是一类绝灭的食肉目动物，由于其独特的牙齿形态代表了已完全消失的特化
取食方式而引起了极大的关注和争论。一些剑齿虎是狮子体型或老虎体型的食肉动物，它
们被广泛认为能够比其不具剑形犬齿的现代近亲捕杀更大和更强壮的猎物。本文报道在甘
肃省晚中新世地层中发现的一具属于恐剑齿虎(Machairodus horribilis)的大型头骨。这件标
本的一些特征与进步的剑齿虎相同，但在某些头骨性状上则与现生的豹亚科种类相似。不
同于其他大多数剑齿虎，功能形态分析指示该剑齿虎的口部张开程度受到限制，因此只能
捕猎相对较小的猎物。这具头骨的解剖特征为证明即使在最大的具剑形犬齿的食肉目动物
中也存在捕猎咬杀方式的多样性提供了新的证据，并揭示了在剑齿虎中导致功能和形态多
样性镶嵌进化的另一种机制。
关键词：甘肃，晚中新世，剑齿虎，头骨，捕猎行为
中图法分类号：Q915.874 文献标识码：A 文章编号：1000−3118(2016)04−0302−17
References
Akersten W, 1985. Canine function in Smilodon (Mammalia, Felidae, Machairodontinae). Los Angeles Co Mus Contr Sci,
356: 1–22
Andersson K, Norman D, Werdelin L, 2011. Sabertoothed carnivores and the killing of large prey. Plos One, 6 (10): e24971.
doi:10.1371/journal.pone.0024971
Antón M, 2013. Sabertooth. Bloomington and Indianapolis: Indiana University Press. 1–243
Antón M, Galobart A, 1999. Neck function and predatory behaviour in the scimitar toothed cat Homotherium latidens (Owen).
J Vert Paleont, 19: 771–784
Antón M, García-Perea R, Turner A, 1998. Reconstructed facial appearance of the Sabretoothed felid Smilodon. Zool J Linn

4期 317
Deng et al. - A skull of Machairodus horribilis and new evidence for gigantism
Soc, 124: 369–386
Antón M, Salesa M J, Morales J et al., 2004. First known complete skulls of the scimitar-toothed cat Machairodus
aphanistus (Felidae, Carnivora) from the Spanish Late Miocene site of Batallones-1. J Vert Paleont, 24: 957–
969
Bakker R T, 1998. Brontosaur killers: Late Jurassic allosaurids as sabertooth cat analogues. Gaia, 15: 145–158
Biknevicius A R, Van Valkenburgh B, Walker J, 1996. Incisor size and shape: implications for feeding behaviors in saber-
toothed “cats”. J Vert Paleont, 16: 510–521
Brown J G, 2014. Jaw function in Smilodon fatalis: a reevaluation of the canine shear-bite and a proposal for a new
forelimb-powered class 1 lever model. Plos One, doi: 10.1371/journal.pone.0107456
Bryant H N, Churcher C S, 1987. All sabertoothed carnivores aren’t sharks. Nature, 325: 488
Carbone C, Mace G M, Roberts S C et al., 1999. Energetic constraints on the diet of terrestrial carnivores. Nature, 402: 286-
288
Christiansen P, 2013. Phylogeny of the sabertoothed felids (Carnivora: Felidae: Machairodontinae). Cladistics, 29: 543-
559
Christiansen P, Harris J M, 2005. Body size of Smilodon (Mammalia: Felidae). J Morphol, 266: 369–384
Damuth J, MacFadden B J, 1990. Body Size in Mammalian Paleobiology: Estimation and Biological Implication.
Cambridge: Cambridge University Press. 1–397
de Castro M C, Langer M C, 2008. New postcranial remains of Smilodon populator Lund, 1842 from South-Central Brazil.
Rev Bras Paleont, 11: 199–206
Deng T, Qiu Z X, Wang B Y et al., 2013. Late Cenozoic biostratigraphy of the Linxia Basin, northwestern China. In: Wang
X M, Flynn L J, Fortelius M eds. Fossil Mammals of Asia: Neogene Biostratigraphy and Chronology. New York:
Columbia University Press. 243–273
Ficcarelli G, 1984. The Villafranchian cheetahs from Tuscany and remarks on the dispersal and evolution of the genus
Acinonyx. Palaeont Ital, 73: 94–103
Geraads D, Kaya T, Tuna V, 2004. A skull of Machairodus giganteus (Felidae, Mammalia) from the Late Miocene of Turkey.
Neues Jahrb Geol Paläontol Monatsh, 2: 95–110
Goloboff P A, Farris J S, Nixon K C, 2008. TNT, a free program for phylogenetic analysis. Cladistics, 24: 774–786
Hayward M W, Kerley G I H, 2005. Prey preferences of the lion (Panthera leo). J Zool, 267: 309–322
Kretzoi M, 1929. Materialen zur phylogenetischen Klassikation der Aeluroïden. 10th Int Zool Congr, 2: 1293–1355
Martin L D, 1980. Functional morphology and the evolution of cats. Trans Nebr Acad Sci, 8: 141–154
Martin L D, Babiarz J P, Naples V L et al., 1999. Three ways to be a sabertoothed cat. Naturwissenschaften, 87: 41–44
Martin P S, 1989. Prehistoric overkill: a global model. In: Martin P S, Klein R G eds. Quaternary Extinctions: A Prehistoric
Revolution. Tucson: The University of Arizona Press. 354–404.
McHenry C R, Wroe S, Clausen P D et al., 2007. Supermodeled sabercat, predatory behavior in Smilodon fatalis revealed by
high-resolution 3D computer simulation. Proc Nat Acad Sin, 104: 16010–16015
Melentis J K, 1968. Die Pikermi fauna von Halmyropotamos (Euböa-Griechenland). Ann Geol Pays Hell, 19: 285–
411
Mihlbachler M C, Rivals F, Solounias N et al., 2011. Dietary change and evolution of horses in North America. Science,
331: 1178–1181

318 54卷
Vertebrata PalAsiatica, Vol. 54, No. 4
Qiu Z X, Huang W L, Guo Z H, 1987. The Chinese hipparionine fossils. Palaeont Sin, New Ser C, 25: 1–243
Qiu Z X, Deng T, Wang B Y, 2004. Early Pleistocene mammalian fauna from Longdan, Dongxiang, Gansu, China. Palaeont
Sin, New Ser C, 27: 1–198
Qiu Z X, Shi Q Q, Liu J Y, 2008. Description of skull material of Machairodus horribilis Schlosser, 1903. Vert PalAsiat, 46:
265–283
Qiu Z X, Qiu Z D, Deng T et al., 2013. Neogene land mammal stages/ages of China: toward the goal to establish an Asian
land mammal stage/age scheme. In: Wang X M, Flynn L J, Fortelius M eds. Fossil Mammals of Asia: Neogene
Biostratigraphy and Chronology. New York: Columbia University Press. 29–90
Radloff F G T, du Toit J T, 2004. Large predators and their prey in a southern African savanna: a predator’s size determines
its prey size range. J Anim Ecol, 73: 410–423
Ronquist F, Teslenko M, van der Mark P et al., 2012. MrBayes 3.2: efcient Bayesian phylogenetic inference and model
choice across a large model space. Syst Biol, 61: 539–542
Salesa M J, Antón M, Turner A et al., 2005. Aspects of the functional morphology in the cranial and cervical skeleton of
the saber-toothed cat Paramachairodus ogygia (Kaup 1832) (Felidae, Machairodontinae) from the Late Miocene of
Spain: implications for the origins of the machairodont killing bite. Zool J Linn Soc, 144: 363–377
Spaulding M, Flynn J J, 2012. Phylogeny of the Carnivoramorpha: the impact of postcranial characters. J Syst Palaeont, 10:
653–677
Swofford D L, 1991. PAUP: Phylogenetic Analysis Using. Parsimony, Version 3.1
Turner A, Antón M, 1997. The Big Cats and Their Fossil Relatives. New York: Columbia University Press. 1–234
Van Valkenburgh B, 1988. Incidence of tooth breakage among large, predatory mammals. Am Nat, 131: 291–300
Van Valkenburgh B, 1990. Skeleton and dental predictors of body mass in carnivores. In: Damuth J, MacFadden B J eds.
Body Size in Mammalian Paleobiology: Estimation and Biological Implication. Cambridge: Cambridge University
Press. 181–205
Van Valkenburgh B, Ruff C B, 1987. Canine tooth strength and killing behaviour in large carnivores. J Zool, 212: 379–
397
Van Valkenburgh B, Sacco T, 2002. Sexual dimorphism, social behavior, and intrasexual competition in large Pleistocene
carnivorans. J Vert Paleont, 22: 164–169
Werdelin L, 1996. Carnivoran ecomorphology: a phylogenetic perspective. In: Gittleman J L ed. Carnivore Behavior,
Ecology, and Evolution, Volume 2. Ithaca: Cornell University Press. 582–624
Wroe S, McHenry C, Thomason J, 2005. Bite club: comparative bite force in big biting mammals and the prediction of
predatory behaviour in fossil taxa. Proc R Soc B, 272: 619–625
Xu X, Wang K B, Zhang K et al., 2012. A gigantic feathered dinosaur from the Lower Cretaceous of China. Nature, 484:
92–95
Xue X X, Coombs M C, 1985. A new species of Chalicotherium from the Upper Miocene of Gansu Province, China. J Vert
Paleont, 5: 336–344
Xue X X, Delson E, 1989. A new species of Dryopithecus from Gansu, China. Chin Sci Bull, 34: 223–229
Zhang Y X, Xue X X, 1995. Taphonomy of Longjiagou Hipparionine Fauna (Turolian, Miocene), Wudu County, Gansu
Province, China. Beijing: Geological Publishing House. 1–96