A new tyrannosaur with evidence for anagenesis and crocodile-like facial sensory system

A new species of tyrannosaurid from the upper Two Medicine Formation of Montana supports the presence of a Laramidian anagenetic (ancestor-descendant) lineage of Late Cretaceous tyrannosaurids. In concert with other anagenetic lineages of dinosaurs from the same time and place, this suggests that anagenesis could have been a widespread mechanism generating species diversity amongst dinosaurs, and perhaps beyond. We studied the excellent fossil record of the tyrannosaurid to test that hypothesis. Phylogenetic analysis places this new taxon as the sister species to Daspletosaurus torosus. However, given their close phylogenetic relationship, geographic proximity, and temporal succession, where D. torosus (~76.7–75.2 Ma) precedes the younger new species (~75.1–74.4 Ma), we argue that the two forms most likely represent a single anagenetic lineage. Daspletosaurus was an important apex predator in the late Campanian dinosaur faunas of Laramidia; its absence from later units indicates it was extinct before Tyrannosaurus rex dispersed into Laramidia from Asia. In addition to its evolutionary implications, the texture of the facial bones of the new taxon, and other derived tyrannosauroids, indicates a scaly integument with high tactile sensitivity. Most significantly, the lower jaw shows evidence for neurovasculature that is also seen in birds.

1. Figure S1. Results of U-Pb LA-ICPMS dating of zircon from a volcanic ash bed (DVTM-1) from the upper Two Medicine Formation. The (ordered) weighted mean age for sample DVTM-1 is calculated and shown in (A) and the lower intercept solution for the sample is shown on the Terra-Wasserberg plot in (B).
3. Figure S3. Topology of the strict consensus tree that summarizes 18 equally parsimonious trees; a 50% majority rule consensus tree results in the same topology. Numbers three and less indicate Bremer support values, numbers greater than three are bootstrap values (1000 replicates in TNT, Goloboff et al., 2008); nodes without bootstrap numbers have less than 50% support. 4. Figure S4. Ontogram for Daspletosaurus horneri sp.nov.

Discussion S1. U-Pb geochronology for the MOR 1130 locality
The MOR 1130 Daspletosaurus locality is located in a relatively well-exposed stratigraphic section along the Sun River in Teton County, Montana. Although a relatively continuous stratigraphic section is exposed along the Sun River and the nearby Pishkun Canal, this area is located in part of the Montana foreland basin known as the disturbed belt, and complex folding and faulting makes stratigraphic assessments challenging in this area. However, based on geologic mapping and measuring of sections in this area (e.g., Scherzer and Varricchio, 2010), MOR 1130 is placed near the top of the Two Medicine Formation. In order to confirm this assertion, a previously undated volcanic ashbed (DVTM-1) located 5.9 m below MOR 1130 was dated herein.

Sample Preparation
The sample was carefully collected to avoid contamination, and repeatedly washed and decanted in the lab to remove the clay-sized fraction. For this particular sample, the purity of the sample and lack of reworking of this bentonite was demonstrated by the lack of a detrital sand and silt component as part of the coarse-grained fraction. The coarse-grained (sand and coarse-silt sized) mineral separate was washed a final time in an ultrasonic bath, followed by density separation using lithium adjusted to a specific gravity of 2.85-2.87. The heavy mineral fraction was then washed, dried, and run through a Frantz magnetic separator at progressively higher magnetic currents of 0.5, 0.8, 1.3, and 1.5, set at a constant 10° side slope. The non-magnetic (>1.5) heavy mineral separates were then handpicking for zircon and only the clearest, most euhedral acicular zircons were selected for U-Pb dating. Roughly 60 zircons were mounted in a 25 mm epoxy resin puck, polished to expose their mid-sections and imaged using a Jeol JSM5410LV scanning electron microscope with attached cathodoluminescence detector in order to document microstructures, cracks, inclusions and other complexities.

U-Pb dating of zircon via laser ablation inductively coupled mass-spectrometry
All work was done at the Advanced Analytical Centre of James Cook University, using a Coherent GeolasPro 193nm ArF Excimer laser ablation system connected to a Bruker 820-MS (formerly Varian 820-MS). The ablation cell was connected to the Bruker 820-MS via Tygon tubing and a 3-way mixing bulb (volume ~5 cm 3 ). The standard cylindrical sample cell was used througout the study, but with a custom-designed polycarbonate insert to reduce the effective volume to 4 cm 3 (see Tucker et al., 2013). This insert combined with the mixing bulb provides both a very stable time-resolved signal and rapid signal washout. The Bruker 820-MS employs an ion mirror design, which reflects the ion beam exiting the skimmer cone by 90° and focusses this into the mass analyzer. Non-ionized large particles and neutrals, as well as partially ionized particles, are not reflected and extracted by a pump located behind the mirror. In this way, the electrostatic mirror acts as a particle size filter to admit only fully atomized and ionized particles into the quardupole mass filter and detector. The advantage of this unique configuration is that it facilitates tuning of the ICP-MS to minimise instrumental mass fractionation focussing on the key ratio of Pb/U, as described below. The instrumental parameters and operating conditions are provided in Tucker et al. (2013). All instrument tuning was performed using a 5 Hz repetition rate, 44 µm beam aperture and 6 J/cm 2 energy density, as determined by energy meter at the ablation site. Under these conditions, the ablation rate for NIST 610 and zircon was about 0.1 µm per laser pulse and 0.06 µm per laser pulse respectively. Tuning was achieved by iteratively adjusting the He carrier gas, Ar sampling gas, sheath gas flow rate, RF Power, 1 st , 2 nd , 3 rd Extraction lens and corner lens voltage to achieve 238 U/ 232 Th ratio ~1, ThO/Th < 1% typically 0.5% and 206 Pb/ 238 U ~ 0.22 in NIST610. Tuning the instrument towards the 'true' 206 Pb/ 238 U ratio thus minimizes the magnitude of the total Pb/U fractionation correction applied zircon analyses, thus reducing the inherent uncertainties in this correction procedure where there are large age differences between standard and sample zircons. Using this technique improved the accuracy and reproducibility of zircon U-Pb isotope analysis in our laboratory (See Tucker et al., 2013). For sample analysis, the total measurement time was set at 65 seconds. The first 30 seconds was for gas blank measurement (laser firing but with the shutter closed), with the shutter opened to allow sample ablation for the final 35 seconds, standard bracketing was used throughout the study to correct for remaining elemental fractionation and mass bias. Forty-eight grains were chosen for analysis using the optimized LA-ICP-MS tuning method outlined above, using a 32 µm beam diameter was used for sample analysis. The primary zircon standard GJ-1 (608. 5  Following tuning of the instrument, each set of five unknown analyses were packaged between paired analyses of the both the primary and secondary zircons. All standard analyses were within 2% of the expected ages, and most were within 1% of the expected age. NIST 610 or 612 was analyzed at the beginning and end of each session, and at least once in between, for the purpose of calibrating Th and U concentrations.

Results
Data processing and reduction was performed using the Glitter software package (Van Achterbergh et al., 2001). All time-resolved isotope signals were filtered for signal spikes related to inclusions and fractures, and then background and sample intervals were evaluated to assess fractionation patterns and to identify the most stable and representative isotope ratios. U-Pb LA-ICP-MS zircon analyses were performed and all inherited older zircons and highly discordant analyses were filtered out prior to age calculation. Age calculations based on measured isotope ratios were completed using Isoplot/Ex version 4.15 and the Age7 function in isoplot for correction of common Pb was performed (Ludwig, 2012) (Table S1). 206 Pb/ 238 U ages are utilized because of the young age of the sample. A weighted mean age of 74.38 +-0.72 Ma (1σ) with an MSWD of 0.55 was calculated for DVTM-1 ( Figure S1A). This is comparable with the lower intercept solution of 74.44 +-0.74 Ma (MSWD=0.56) for this sample ( Figure S1B). The age of DVTM-1 fits the stratigraphic superposition of the site and is consistent with published Ar/Ar dates for the Two Medicine in the type section (Rogers et al., 1993). Moreover, U-Pb detrital zircon geochronology of the Two Medicine Formation is ongoing in the study area (Roberts et al., 2011), and the maximum depositional ages documented for samples from the upper Two Medicine Formation in the study area are closely consistent with the age reported here for DVTM-1. ligament scar (B. sealeyi, Albertosaurus libratus, A. sarcophagus, Teratophoneus curriei, R. kriegsteini). Entire circumference of the pneumatic recess of the squamosal is undercut and clearly defined-The recess is entirely undercut in D. horneri and Tyrannosaurus bataar, whereas in other taxa at least one corner of the recess is not undercut, where the recess and the ceiling of the bone are joined together (D. torosus, R. kriegsteni, T. rex). Sinuous rostral edge in dorsal view of the dorsotemporal fossa on the frontal-Although weathered in both mature specimens (MOR 590, MOR 113), the rostral margin of the dorsotemporal fossa in D. horneri is sinuous as it winds its way from the sagittal midline to the orbital notch. This condition is seen in other derived tyrannosauroids (Albertosaurus libratus, L. argestes, R. kriegsteini). In contrast, the margin is straightened in D. torosus. The margin is variable in T. rex, where it is might be sinuous, straightened, or rostrally convex, and it may extend rostrolaterally or caudolaterally from the midline. The rostrally convex condition is seen in T. bataar. Joint surface for the squamosal on the parietal covers the base of the caudolateral process-In D. horneri the joint surface for the squamosal covers the base of the caudolateral process. This condition is also seen in Albertosaurus libratus, A. sarcophagus, and T. rex. In contrast, the joint surface covers only the ventral half of the base in other derived tyrannosauroids (A. remotus, D. torosus, T. bataar). Tympanic ridge extends onto the prootic-In D. horneri the tympanic ridge that bounds the lateral tympanic space dorsally extends caudally onto the prootic such that it also bulges laterally above the caudal tympanic recess. This condition is also seen in Albertosaurus libratus, A. sarcophagus, T. bataar and T. rex. In contrast, the ridge is stout in Alioramus remotus and D. torosus, where it fades nearly immediately and does not continue onto the prootic as an eave-like structure. Pneumatic foramen penetrating the lateral surface of the quadratojugal-This feature is seen in adult D. horneri, subadult T. rex, and and an unidentified taxon from the Dinosaur Park succession from southern Alberta. This foramen is not seen in Iren Dabasu taxon, Bistahiversor sealeyi, Albertosaurus libratus, A. sarcophagus, Alioramus remotus, D. torosus, or T. bataar. Shallow notch between the basal tubera-The notch is relatively shallow in D. horneri (less than 40% the total height of the basioccipital below the occipital condyle. A shallow notch is also seen in Albertosaurus libratus. The tall condition, where the notch is greater than 40%, is seen in D. torosus. Variation is present in T. rex where both extremes in height are seen, which ranges from 21% to 45%). A tall notch is considered present in Bistahieversor sealeyi, A. sarcophagus, Alioramus remotus, and T. bataar (Bever et al., 2013). Short epipophyses of the anterior cervicals-In D. horneri, the epipophyses of the anterior and mid cervicals are stout and project only a short distance past the postzygapophyses, if at all. This condition is also seen in T. bataar (Maleev, 1974). In contrast, the epipophyses are long and extend past the postzygapophyses in the anteriormost cervicals of Albertosaurus libratus, Teratophoneus curriei, and Tyrannosaurus rex. The humerus is ~34% the length of the femur-In D. horneri, the humerus is one third (MOR 590) the length of the femur. A similar ratio is seen in A. libratus (TMP 1986.144.0001: 36%; CMN 2120: 31%; Lambe, 1917), A. sarcophagus (CMN 11315: 34%; Russell, 1970), and Teratophoneus curriei (BYU 8120/9396: 32%). The humerus to femur ratio of D. torosus is unknown owing to the absence of a femur from the holotype; so using the skull and ilium as a proxy for femur length, the ratio was in the range of 34% to 32%, respectively. Therefore, it may 4. Frontals are separated caudally on the midline by a prominent wedge of the parietal (also seen in A. libratus, A. sarcophagus, Nanuqsaurus, Raptorex, Alioramus altai, and in some T. rex specimens; wedge is not as rostrally long in D. torosus).
2. Lateral flange of the palatine is positioned below the rostral end of the caudal pneumatic recess (also seen in Alioramus altai and in some specimens of T. rex; positioned caudal to the rostral end of the recess in D. torosus). 3. Rostral pneumatic recess is deeply excavated (also seen in Albertosaurus libratus, A. sarcophagus, Alioramus altai, some specimens of T. rex; shallow in adults of D. torosus). 4. Pneumatic recesses are widely separated from each other (also seen in Alioramus altai, some T. rex specimens; close together or overlapping in D. torosus). 5. Caudal pneumatic recess exceeds the height of the maxillary ramus (also seen in Bistahieversor, and some specimens of Albertosaurus sarcophagus and T. rex; at most is as tall as the ramus in D. torosus). 6. The rostral and caudal pneumatic recesses of the palatine are connected to each other internally (also seen in Alioramus altai and in some specimens of T. rex; separated by a partition in D. torosus). 7. Pneumatic recesses that penetrate the medial surface of the palatine are absent (also seen in Albertosaurus libratus, A. sarcophagus, Alioramus altai, some specimens of T. rex; present in D. torosus). 8. Joint surface on the caudolateral process of the palatine extends rostrally below the caudal pneumatic recess (also seen in Albertosaurus sarcophagus, Alioramus altai; stops far caudal to the recess in D. torosus). 9. One pneumatic recess is present in the ectopterygoid (also seen in Alioramus altai, T. bataar, and in some specimens of T. rex; two are present in D. torosus). 10. Caudal process of the ectopterygoid is penetrated by an accessory pneumatic foramen (seen in some specimens of T. rex; absent from D. torosus).
Mandibular ramus -unique D. horneri characters: 1. A wide shelf separates the joint surface for the angular from the caudal surangular foramen (shelf is absent from D. torosus). Mandibular characters seen in D. horneri and other tyrannosaurids, except D. torosus: 1. Surangular shelf does not overhang the caudal surangular foramen (also seen in Albertosaurus libratus, Alioramus altai, Raptorex, seen in some T. rex specimens; overhangs the foramen in D. torosus). 2. Caudal surangular foramen does not extend to the ventral surface of the surangular shelf (also seen in Albertosaurus libratus; foramen reaches the shelf in D. torosus). 3. Dorsal margin of the joint surface for the intercoronoid reaches the ventral margin of the medial bar (also seen in A. libratus, Raptorex, T. bataar, and in some specimens of T. rex; dorsal margin fades before reaching the ventral margin in D. torosus). 4. The joint surface for the angular extends behind the level of the caudal surangular foramen (also seen in Bistahieversor, A. libratus, Lythronax, Alioramus remotus, A. altai, Raptorex, T. rex; stops at the caudal edge of the foramen in D. torosus). 5. Seventeen (17) dental alveoli (also seen in subadult T. rex; 16 in D. torosus).

Discussion S3. Discussion of the Superficial Craniomandibular Soft Tissues of Tyrannosaurids
Goals & Methods. The excellent quality of preservation of these specimens (MOR 553S/7.19.0.97, MOR 590, MOR 1130) permits us to assess the type of soft tissue that covered the face (premaxilla, maxilla, nasal, lacrimal, jugal, postorbital, squamosal, dentary). In D. horneri, and in all derived tyrannosauroids (Bistahieversor + Tyrannosauridae), the subcutaneous texture is coarse and shows a hierarchy of textures. In order to identify the soft tissue covering that produced this complex surface, we compared the condition of tyrannosaurids with that of crocodylians (Alligator) and birds (Struthio, Anser, Anas, Cygnus), and we applied the results of Hieronymus et al. (2009) and Hieronymus and Witmer (2010) to deduce the soft tissues that produced the osteological features. Snout & orbital region. In tyrannosaurids, large neurovascular foramina penetrate the subcutaneous surface from which deep, sharply inset sulci extend. Based on comparisons with crocodylians and birds, the foramina that emerge onto the lateral surface of the snout and jaws correspond to: the sensory dorsal and ventral branches of the medial ophthalmic nerve (N. V 1 ), internal maxillary artery, maxillary division of trigeminal nerve (N. V 2 ), dorsal alveolar vessels and nerve, ventral subnarial vessels and nerve, lateral paravestibular neurovascular bundle, lateral premaxillary neurovascular bundle, rostral premaxillary neurovascular bundle, and median palatine artery (Papp, 2000;Sedlmayr, 2002). The openings that emerge through the dorsum of the snout correspond to the lateral nasal and medial nasal vessels and twigs of the ophthalmic ramus of the trigeminal nerve (N. V 1 ) (Sedlmayr, 2002).
In form, the foramina and their sulci correspond to a tangential rugosity profile (Hieronymus et al., 2009), where the foramina are oriented oblique to the external surface of the bone. The grooves usually bifurcate several times as they extend across the bone, producing wide swaths and series of V-shaped grooves, which produce a hummocky rugosity profile between them (Hieronymus et al., 2009). The grooves tend to be ordered and anastomosing in orientation, and they are numerous and closely spaced.
On top of the primary texture there is an overprint of fine grooves and ridges and, finally, a tertiary texture coarsens the fine ridges. These correspond to a projecting rugosity profile (Hieronymus et al., 2009). In some areas of the skull (nasals, lacrimals, maxillae) the secondary and tertiary textures are elaborated into large, coarse ridges, cusps, and papillae.
The texture has a sharp boundary with the smooth temporal region on the skull that is marked by a raised lip on the postorbital, whereas the transition from coarse to smooth is gradational and without relief on the lower jaw. As such, the coarse texture is limited to the rostral part of the facial skeleton ahead of the temporal region. Tyrannosaurids bear three or four pairs of ornamental horns, which occur within the boundaries of the coarse region. The surface of the temporal region is generally smooth or textured by fine ridges and grooves.
Although tyrannosaurids, crocodylians, and birds share neurovascular foramina that are densely clustered at the front of the jaws and form rows along the oral margins, a sharp difference in texture separates birds from their closest relatives. The texture in crocodylians is identical to that of tyrannosaurids, except that the entire face of crocodylians is coarse in texture. In contrast, the texture deep to the beak in birds is smooth, aside from sharply incised neurovascular sulci. The osteological texture between the primary neurovascular sulci of tyrannosaurids compares best with the hummocky texture that is consistent with overlying scales as is seen in living crocodylians (Hieronymus et al., 2009).

Postorbital bone.
In tyrannosaurids a lip-like transition is seen on the postorbital between the textured and nontextured surfaces; this abrupt, stepped transition indicates the presence of a cornified sheath (Hieronymus et al., 2009) on at least the postorbital that graded rostrally into the scaly covering on the rest of the face. A localized sheath is supported by the observation in Tyrannosaurus (AMNH FARB 5027, MPC-D 107/2) where a pair of osteoderms are situated above the orbit that fuse to the underlying bones; osteoderms are associated with overlying scales (Hieronymus et al., 2009) and therefore the caudal margin of the caudal osteoderm marks the rostral limit of the cornified sheath. Should similar osteoderms be seen in other tyrannosaurids then the rostral limit of the sheath can be identified.
The cornual process of the postorbital in Daspletosaurus and other tyrannosaurids often has a coarse and warty dorsal and caudal surface, and a smooth lateral surface. This arrangement indicates the presence of a projecting skin structure such as a horn (Hieronymus et al., 2009). Indeed, the undercut caudal and dorsal edge of the cornual process (which extends rostrally along the frontal process) corresponds to the lip-like edge of bony structures that support cornified sheaths (Hieronymus et al., 2009). Based on these lines of evidence, we conclude that at least part of the postorbital cornual process of tyrannosaurids supported a cornified sheath. Nasal bone. Tyrannosaurids bear a series of low bumps along the dorsal midline of the nasal bones; these bumps almost certainly correspond to individual overlying scales, as is seen in centrosaurine ceratopsians (Hieronymus et al., 2009). In some tyrannosaurids (Alioramus remotus), the bumps are quite tall, but they do not possess the basal rims that indicate a cornified sheath; therefore, those bumps also mark the position of overlying scales. In D. horneri and other tyrannosaurids, the nasal rugosity increases with maturity (see 'Armor-like dermis' below), but then it becomes smooth in the most mature specimens. A similar ontogenetic process is seen in centrosaurines, from supraorbital horns to supraorbital pits, but that change reflects the transition from a cornified sheath to a cornified boss; the process in tyrannosaurids does not show evidence for a change in the overlying epidermal tissue since the difference is only a decrease in relief. Gingiva. Texture is also informative regarding the soft tissues that covered the oral margins of tyrannosaurids. The foramina along the margin of the jaws emit stout (upper jaw) or long (lower jaw) sulci that often fade before reaching the oral margin. This arrangement produces a smooth zone between the alveolar foramina and the edge of the jaw, which is also seen in crocodylians. In tyrannosaurids vertical rows of minute foramina extend perpendicular to the tooth row in the smooth zone. These subordinate rows of foramina are also seen in crocodylians and in mammals (Equus); which may be an osteological correlate of gingiva for Amniota. Rhamphotheca. Birds are similar to tyrannosaurids and crocodylians in that the surface deep to the beak is penetrated by large foramina that emit sharp-edged, long and branching sulci, but an overprint of a secondary or tertiary texture is not seen and the beak-bearing surface is otherwise smooth (Papp, 2000;Hieronymus et al., 2009). Also, externally the foramina are low in number and are concentrated at the rostral and caudal extremes of the upper jaw (Papp, 2000). Therefore, we reject the hypothesis that tyrannosaurids bore beaks, and we fail to reject the hypothesis that their faces had a scaly covering comparable to crocodylians, a cornified sheath on the postorbital, and armor-like dermis on the lacrimal, nasal, and dentary (see below).
In addition to a coarse subcutaneous texture, tyrannosaurids and crocodylians share other features that are not seen in birds; these include neurovascular foramina that penetrate through the top of the snout, and the caudalmost foramen of the alveolar row of the maxilla is large and emits a long, deeply incised, and caudally extending sulcus that excavates the side of the caudalmost tip of the bone. These differences from birds almost certainly reflect a lack of specialized blood supply to the snout, the large size of the jaws, and presence of teeth.
Osteological correlates of rhamphothecal plates, such as the bony groove deep to the nasolabial groove along the premaxillomaxillary suture that is seen in modern birds and the extinct Cretaceous birds Hesperornis and Ichthyornis (Hieronymus and Witmer, 2010), are not seen in tyrannosaurids or crocodylians. On the lower jaw, the mentolabial groove is a correlate in birds that marks the separation between the plates of the rhamphotheca of the lower jaw; this correlate is not seen in tyrannosaurids, crocodylians, or Velociraptor (AMNH FARB 6515), but it is seen in the extinct birds Hesperornis and Ichthyornis (Hieronymus and Witmer, 2010). Papp (2000) identified osteological correlates for the presence of a rhamphotheca, namely numerous neurovascular foramina and sulci, and sulci, ridges (e.g., tomial ridges), or knobs within the beak region. Although numerous neurovascular foramina are seen in crocodylians and nonavian theropods, the associated sulci, ridges, and knobs are not. We emphasize here that the rugose sculptured texture on the ventral surface of the dentary by itself is not evidence for a rhamphotheca since a beak is absent from crocodylians and some nonavian dinosaurs, whereas beaks are seen in birds and some nonavian dinosaurs (Papp, 2000).
We therefore conclude that rhamphothecae were absent from nonavian theropods, if a compound rhamphotheca is the primitive condition (Hieronymus and Witmer, 2010). Also, the groove that unites the primary alveolar row of foramina in the dentary of tyrannosaurids, Velociraptor, and modern birds, is not a correlate of the infralabial groove, which is a novel feature limited to cassowaries, anatoideans, and some procellariiformeans (Hieronymus and Witmer, 2010). It therefore appears that the compound rhamphotheca evolved between the divergence of deinonychosaurians and the appearance of Hesperornis. Alveolar groove. In crocodylians, nonavian theropods, and birds, a primary row of foramina extends along the lateral surface of the dentary parallel to the dorsal edge of the bone. These foramina correspond to the sensory intramandibular branch of the mandibular nerve (N. V 3 ), branches of the external cutaneous vessels, and the ventral alveolar artery (Papp, 2000;Sedlmayr, 2002). As in crocodylians, multiple rows of foramina are seen on the lateral surface of the dentary, including the ventrolateral surface; this is unlike birds, in which only a single lateral row is seen, aside from the rostral tip of the lower jaw (Papp, 2000).
Tyrannosaurids are similar to birds in that the alveolar row of foramina in the dentary occurs in a common groove that extends for much of the length of the bone; this common groove is not seen in crocodylians, although long, caudodorsally extending sulci extend from the foramina. In birds this groove is covered by the rhamphotheca, and the ventral rictal vessels and the external branch of the mandibular nerve lie in the groove (Sedlmayr, 2002); we regard the groove as a possible osteological correlate for those structures.
In tyrannosaurids the groove is not tangential, whereas it is undercut in more highly derived, bird-like theropods such as Velociraptor and birds (Struthio, Anser, Anas); we therefore suggest that the dentary of tyrannosaurids was covered by epidermal scales such as is seen in crocodylians. Ergo, the groove is an osteological correlate of the neurovasculature and not of the overlying skin, and indicates that this change to the jaw supply preceded the evolution of the rhamphotheca. The presence of a deep groove in Velociraptor suggests that the rhamphotheca had first evolved on the mandible before the upper jaw, and beaked jaws arose long before tooth loss. In texture the subcutaneous surface of the dentary in tyrannosaurids is identical to that of the upper jaw (but see below), which is consistent with the hypothesis of a scaly epidermis.
Armor-like dermis. In subadult and adult tyrannosaurids, the rostral surface of the premaxilla and the lateral surface of the dentary at the rostral half of the bone is coarsened by small pustulelike bumps. This texture is associated with armor-like dermis (Hieronymus et al., 2009), which would have protected against abrasion and scrapes while eating a carcass. This papillate texture is also seen on the dorsolateral surface of the lacrimal, which often bears a horn or swelling and indicates the presence of armor-like dermis and not a superimposed scale. Also, rugose 'burrs' and papillae are not infrequently seen on the coarse nasal bones of tyrannosaurids, which might mark localized 'warts' of armor-like dermis on the top of the snout. Summary. In conclusion, we hypothesize that the coarse region of the face (upper and lower jaws) in tyrannosaurids was covered by flat epidermal scales as is seen in crocodylians. Unlike crocodylians, tyrannosaurids have a variety of other epidermal structures that include armor-like dermis on the cornual processes of the jugal and lacrimal, the top of the snout, and side of the lower jaw, and cornified epidermis on the postorbital cornual process. These hypotheses of epidermal identity can be tested with a microscopy-based examination for their corresponding histological correlates (Hieronymus et al. 2009) and by the discovery of fossils that preserve craniofacial integument either as fossilized skin, or, as natural molds or casts.

Discussion S4. Growth Series
We compared the growth series with the hierarchy of tyrannosaurine synapomorphies to assess the presence of recapitulation of phylogenetic characters during growth. We found that characters at the level of Tyrannosaurinae and the Daspletosaurus + (Zhuchengtyrannus + Tyrannosaurus) clade appear without congruence throughout the growth series. This noncongruence between growth and phylogeny almost certainly reflects the incomplete condition of the subadult specimens (represented by single bones) and the significantly more complete skulls of the mature specimens.

Phylogenetic Character List
Listed here are the phylogenetic characters, with descriptions of their states, which were used to resolve the ingroup phylogenetic relationships of Tyrannosauroidea. Much of the wording is unaltered from , but only essential notes have been retained. Characters that are new in this study are identified in boldface and start with the word "new"; the abbreviation "var." indicates character descriptions that are modified from their original source. Red text indicates departures from the wording of . 3) Skull, general shape, lateral view: long and low, length to height ratio greater than 3.2 (0); deep, length to height ratio less than 3.2 (1). Note: Length is premaxilla to quadrate condyle length; height is maximum height of the upper jaw, not counting any cranial crests. 7) Orbital fenestra, shape, lateral view: round (0); more than 10% taller dorsoventrally than long rostrocaudally, but less than twice as tall as long (1); dorsoventrally tall, more than twice as tall as long (2)   18) Premaxilla, rostral margin, shape, lateral view: smoothly curved and extending caudodorsally, angle between ventral margin of premaxilla and rostral margin is less than 90 degrees (0); smoothly curved and extending vertically or slightly rostrodorsally, angle between ventral margin of bone and rostral margin is equal to or greater than 90 degrees (1); oriented vertically or slightly rostrodorsally, with a distinct inflection point between nearly vertical rostral region and more horizontal dorsal region (2)  24) Maxilla, maxillary fenestra, rostrocaudal length compared to the distance between the rostral margins of the antorbital fossa and fenestra, lateral view: less than half (0), greater than half (1), greater than half and also greater than half of the length of the eyeball-bearing portion of the orbit (2)

QUADRATOJUGAL (9 characters)
132) Quadratojugal and squamosal, constriction of laterotemporal fenestra: absent, rostral margin of both bones are approximately vertical (0); present, convex kink along the suture between the two bones that projects into the fenestra, constricting it to approximately one half of its maximum rostrocaudal length (1); present, dorsal region of quadratojugal moderately expanded rostrocaudally relative to the remainder of the bone, constricting fenestra to approximately one half of its maximum rostrocaudal length (2); present, dorsal region of quadratojugal expanded rostrocaudally by at least twice the minimum rostrocaudal dimension of the bone, forming a flange that meets the ventral ramus of the squamosal to nearly divide the fenestra (3)

PREFRONTAL (2 characters)
148) Prefrontal, contacts nasal, dorsal view: yes (0); no, excluded by frontal-lacrimal contact (1) (Brusatte et al., 2010). Note: this character was discarded because the derived state is only coded for Guanlong; an autapomorphy is uninformative regarding ingroup relationships. 149) Prefrontal, exposure, dorsal view: widely exposed, forms much of orbital rim and usually separates or nearly separates frontal and lacrimal (0); reduced, not exposed along the orbital rim and allows for wide contact between frontal and lacrimal (1)   189) Palatine, maxillary articulation brace, form, lateral view: projects ventrally due to a jugal process that extends further ventrally than the maxillary process, such that there is a discrete corner between the two processes in lateral view (0); projects laterally, with no discrete corner between the smoothly confluent jugal and maxillary processes in lateral view (1)
48. Maxilla, palatal process, width next to last alveolus: narrower than alveolus (0)  72. Lacrimal, strut that extends between the ventral ramus and rostral ramus, inflation, lateral view: not inflated such that it is positioned medial to the external surface of the ventral ramus and dorsal part of the rostral ramus (0), inflated such that it is nearly level with the external surface of each ramus (1). 73. Lacrimal, antorbital fossa, margin along the dorsal edge of the strut that bounds the lacrimal pneumatic recess rostrally: the fossa and the subcutaneous surface do not merge together (0)