Postcranial Osteology of the Basally Branching Hadrosauroid Dinosaur Tanius sinensis from the Upper Cretaceous Wangshi Group of Shandong, China

ABSTRACT Tanius sinensis was one of the first dinosaur species to be named from China. It was established on a partial skeleton recovered by a joint Sino-Swedish expedition in 1923. The fossils were excavated from Upper Cretaceous strata of the Jiangjunding Formation (Wangshi Group) in Shandong Province, and although their discovery dates back almost 100 years, they have not been reassessed in detail since their initial description in 1929. This omission is critical because T. sinensis is now recognized as one of the stratigraphically youngest non-hadrosaurid hadrosauroid taxa. Here, we reevaluate the postcranial osteology of T. sinensis as a prelude to an anatomical and phylogenetic revision of the species. We examined the holotype and all currently referred specimens of T. sinensis first-hand, and identified a unique postcranial character state combination incorporating tall dorsal neural spines, a reduced postacetabular ridge on the ilium, a fully enclosed flexor tunnel formed by the distal condyles of the femur, and a lunate proximal end on metatarsal III. Comparisons with other species of Tanius confirm that: (1) T. chingkankouensis is a nomen dubium erected on non-diagnostic composite material; (2) T. laiyangensis was established on indeterminate hadrosaurid remains that are not attributable to Tanius; and (3) the anecdotal assignments of Bactrosaurus prynadai and Tsintaosaurus spinorhinus to Tanius cannot be substantiated. Close inspection of the holotype caudal vertebra further reveals a possible healed bite trace consistent with a prey–predator interaction. Lastly, our calculated average body mass estimate for T. sinensis of between 2091–3533 kg suggests that it was one of the largest non-hadrosaurid hadrosauroids.


INTRODUCTION
The first dinosaur species named from China was 'Trachodon' amurense Riabinin, 1925, which was later referred to the genus Mandschurosaurus Riabinin, 1930, but is now regarded as a nomen dubium (sensu Horner et al., 2004). The otherwise earliest documented valid Chinese dinosaur taxa were erected by the eminent Swedish paleontologist Carl Wiman in 1929. These included the titanosauriform sauropod Euhelopus zdanskyi (Wiman, 1929) (see Mateer and McIntosh, 1985;Wilson and Upchurch, 2009;Poropat, 2013;Poropat and Kear, 2013a), and the hadrosauroid Tanius sinensis Wiman, 1929. Tanius sinensis was coined in honor of its discoverer, the Chinese geologist Tan Xichou, who was also often referred to (even by Tan himself) using the westernized version of his name, H. C. T'an. The remains of T. sinensis were recovered from Upper Cretaceous (upper Campanian-lower Maastrichtian) sediments in the Jiangjunding Formation, which is a subunit of the Wangshi Group cropping out in the Shandong peninsular region of eastern China (Hu et al., 2001;Poropat and Kear, 2013b: fig. 1; Fig. 1).
Since its initial description, the systematic relationships of T. sinensis have been debated. Wiman (1929) initially assigned T. sinensis to Hadrosauridae, a classification reiterated by Young (1958). Conversely, von Huene (1956) suggested that T. sinensis might constitute a separate family 'Prohadrosauridae', comprising both T. sinensis and Telmatosaurus transsylvanicus (Nopcsa, 1900). More recently, Weishampel and Horner (1990:555) listed T. sinensis as: "known from poor material that has yet to yield phylogenetically significant information". Nonetheless, these authors interpreted the taxon as a basally branching member of Hadrosauridae, and probably positioned amongst the sister lineages of Euhadrosauria (Lambeosaurinae + Hadrosaurinae). This proposal was predicated upon the identification of various plesiomorphic cranial traits (see discussion in Weishampel and Horner, 1990), yet the presence of tall neural spines on the dorsal vertebrae and a robust, angular deltopectoral crest on the humerus were both cited as features shared with lambeosaurines. By contrast, the phylogenetic analyses of Sues and Averianov (2009) and Prieto-Márquez and Wagner (2009) both returned T. sinensis as a non-hadrosaurid hadrosauroid, an affinity that has since been advocated by multiple later studies (e.g., McDonald et al., 2010;Prieto-Márquez, 2010a;McDonald, 2012;Xing et al., 2014a). This topological placement is significant because T. sinensis represents one of the geologically youngest non-hadrosaurid hadrosauroids. Moreover, its fossils occur in close stratigraphic proximity to those of the lambeosaurine (sensu Prieto-Márquez and Wagner, 2009) Tsintaosaurus spinorhinus Young, 1958, as well as the gigantic hadrosaurine (sensu Xing et al., 2014aXing et al., , 2014bXing et al., , 2017 Shantungosaurus giganteus Hu, 1973, indicating that the replacement of nonhadrosaurid hadrosauroids by hadrosaurid lineages at around the Santonian-early Campanian interval might have been exclusive to Laramidia (Norman, 2004;Evans et al., 2012;Campione et al., 2013). As a notable aside, though, Brownstein (2020) recently reported the presence of a non-hadrosaurid hadrosauroid (= 'hadrosauromorph' sensu Brownstein, 2020) from the Maastrichtian of eastern North America based on a fragmentary tibia.
Two additional species of Tanius Wiman, 1929 have been described since 1929: Tanius chingkankouensis Young, 1958;and Tanius laiyangensis Zhen, 1976. Nevertheless, these were variously synonymized with other taxa, or treated as nomina dubia (Buffetaut, 1995;Horner et al., 2004;Lund and Gates, 2006;Prieto-Márquez and Wagner, 2013;Zhang et al., 2020). Yet, Zhang et al. (2017) argued that T. chingkankouensis should be reinstated as a valid species because its scapular morphology was distinguished from that of T. sinensis. Moreover, the remains assigned to T. sinensis have received little research attention beyond literature-based comparative interpretations (Gilmore, 1933;Young, 1958;Buffetaut and Tong-Buffetaut, 1993) and phylogenies (e.g., Sues and Averianov, 2009;Prieto-Márquez and Wagner, 2009;McDonald et al., 2010;Prieto-Márquez, 2010a;McDonald, 2012;Xing et al., 2014a;McDonald et al., 2017;Zhang et al., 2020). Therefore, it is the purpose of this article to redress the shortfall in knowledge via a comprehensive reassessment of the holotype and all referred specimens of T. sinensis. Given the scale of such a study, we have elected to present our data as a series of papers, the first of which focuses on the postcranial skeleton and the species-level taxonomy of Tanius. This will be followed by a synopsis of the historically associated Wangshi Group vertebrate assemblage, and lastly a reappraisal of the cranial osteology, together with a phylogenetic revision and contextual evaluation of nonhadrosaurid hadrosauroid interrelationships.
Locality and Horizon-PMU 24720 was reportedly discovered approximately 500 m (= one 'li': Buffetaut, 1995) southwest of the village of Jiangjundingcun (alternately recorded as "Chiang Chün Ting" or "Chiang-Chün-Ting" on the original field labels). Additional documentation states that Jiangjundingcun was around 8 km ('16 li') southwest of the city of Laiyang (Fig. 1). This area has extensive outcrops of the Jiangjunding Formation (Wangshi Group), which are the most likely source unit for PMU 24720, and stratigraphically range from the upper Campanian to lower Maastrichtian (Xing et al., 2014b). A radiometric date of 72.9-73.5 Ma was obtained from the uppermost section of the underlying Hongtuya Formation (Yan et al., 2005), providing a maximum lower age limit for the succession.
Diagnosis-The postcranial skeleton of Tanius sinensis can be uniquely distinguished among non-hadrosaurid hadrosauroids by its tall dorsal neural spines (height/centrum height ratio = 3.2; proportional measurements in Supplemental Data 2, Table S2), a reduced postacetabular ridge (sensu Campione, 2014; = "caudal protuberance" sensu Prieto-Márquez, 2010a:495) on the ilium; posterior fusion of the medial and lateral distal femoral condyles to form an enclosed flexor tunnel, and a lunate profile on the proximal end of metatarsal III. The taxon also manifests a distinctive combination of other character states including: a recurved and anterodorsally directed anterior end on the scapular acromion process; maximum dorsoventral width of the proximal end of the scapula being 1.5 times that of the distal end (proportional measurements in Supplemental Data 2, Table S3); a strongly curved dorsal margin on the scapula; a deltopectoral crest less than 50% of the maximum length of the humerus; an expanded 'club-like' distal end on the fibula; central plate of the ilium with a depth/length ratio = 0.91; and the proximodistal length of metatarsal III being 4.5 times greater than its mid-shaft width.
Comments- Wiman (1929) mentioned that the PMU 24720 skeleton was largely complete when found, but that many bones (including the rostral section of the skull) disintegrated during excavation and could not be retrieved. Indeed, the remaining elements are extremely fragile and require continual conservation. PMU archive documents report that Tan Xichou initially collected the cervical vertebrae and some other postcranial remains on April 20, 1923 (see also Wiman, 1929). Additional components of the skeleton were then recovered throughout October 1923 by the Austrian paleontologist, Otto Zdansky, who was commissioned to explore for fossils in China on behalf of Uppsala University. Wiman (1929:49) further noted that the caudal vertebra PMU 24720/12 was separated from the rest of the skeleton, and occurred in conjunction with various theropod and sauropod bones (Poropat, 2013;Poropat and Kear, 2013b). However, its morphological and proportional compatibility supports attribution to the type series. Wiman (1929) also described an isolated dorsal rib (PMU 22481) discovered around 2.5 km ('five li') southwest of Jiangjundingcun, together with a right humerus (PMU 24721), radius (PMU 22483) and metatarsal (PMU 22482), all collected northeast of the village of Tianqiaocun (transcribed as 'T'ien Ch'iao T'un'), 1.5 km to the northwest of Jiangjundingcun. Nevertheless, these specimens are not diagnostic beyond Hadrosauroidea indet., and are thus excluded from the holotype material herein. Lastly, Wiman (1929) documented the distal section of a right ischium (PMU 22485), which was later recognized as part of a theropod left pubis (Gilmore, 1933;Buffetaut and FIGURE 2. Diagrammatic reconstruction of the Tanius sinensis holotype skeleton (PMU 24720/1-32) in right lateral view showing life position of the recovered (white fill) and restored (darker grey fill) elements. The cranium and ilium are mirrored from the left side for completeness. Maximum body length/height as restored equals 7.84/2.83 m. Scale bar equals 1 m. Artwork © Tomas Wigren (Uppsala).
Tong -Buffetaut, 1993). We also identified a corresponding right pubis (PMU 22484) that, together with PMU 22485, are likely attributable to a tyrannosauroid (Supplemental Data 1, Table S1), but will be documented in more detail elsewhere.
A wedge-shaped intercentrum is fused to the anteroventral articulation surface of the axial pleurocentrum. The posterior articular surface of the axis is concave and oval, with a dorsal notch for the neural canal. The parapophysis is robust and dorsoventrally elongate. The posterolateral centrum surface is smooth with nutrient foramen that is 5 mm in maximum diameter. The neural arch and centrum are fused without trace of a neurocentral suture. The projecting right diapophysis is laterally tapered with an oval distal outline; it is situated at the same level as the prezygapophyses, unlike B. johnsoni where the diapophysis is more posteroventrally oriented (Godefroit et al., 1998). The prezygapophyses extend anteriorly in front of the dorsolaterally inclined articular facets.
The axial neural spine is transversely compressed (minimum width = 11 mm) and has smooth lateral surfaces. The anterior margin is broken but likely projected forwards. The dorsal edge has been restored, but was apparently tallest above the postzygapophyses, and lacks any obvious embayment like the axial neural spines of many hadrosaurines (Campione et al., 2007;Prieto-Márquez, 2010a).
The axial postzygapophyses are ventrolaterally oriented and diverge from a common base that overhangs the posterior articulation surface; this roofs the neural canal and forms the ventral margin of the post-spinal fossa. The post-spinal fossa is defined laterally by tall epipophyses that extend dorsally from the postzygapophyses. A small post-spinal ridge is situated within the fossa and extends about one-third the height of the epipophyses, as preserved. The post-spinal fossa imparts a Y-shape to the dorsal margin of the axial neural spine. Gasca et al. (2015) recorded an axial spine height/length ratio of 43%; however, we consider this ambiguous because the specimen is incomplete.
The post-axial cervicals C3-C10 (Figs. 3D-I, 4A-I, 5A-I) are opisthocoelous with extremely reduced neural spines. The parapophyses are located on the lateral sides of the centra and border the posterior edges of the anterior articular surfaces. Where intact, the parapophyseal apices are circular to oval in outline and laterally projecting. The left parapophysis of vertebra C4 is situated along the mid-line of the centrum, slightly more posteriorly than that of C3. The parapophyses of C6 and C7 are positioned lower on the centra than those of the other cervicals, but this may be a result of distortion.
An anteroposteriorly elongate lateral ridge is present on C4-C10 and buttresses the parapophysis anteriorly to delimit paired dorsal and ventral fossae. The lateral ridges become more pronounced posteriorly, although this is exaggerated by crushing. An anteroposterior keel is also present on the ventral surface of the centrum in C3. In C4, the ventral fossa is flanked by two anteroposteriorly oriented ridges. In C8 (but possibly also in C5-7), the ventral fossa is reduced, and the ventral surface becomes planar.
The posterior articular surfaces of the centra in C3-C8 are transversely broad and dorsoventrally tall, imparting a heartshaped profile. By contrast, C9 and C10 are ovate in outline. The centra of C3-C7 become increasingly compact in lateral view, unlike those of C8 and C10, which are rhombic. The anterior articulation surface of C10 is angled posteriorly. Nutrient foramina are visible on the ventral surfaces of C3 and C6, but were probably present on all of the cervical vertebra in life.
The post-axial cervical neural arches are tall and enclose neural canals that increase in diameter posteriorly; in C6 and FIGURE 4. Tanius sinensis holotype cervical vertebrae. Cervical 5 (PMU 24720/5) in A, lateral, B, anterior, and C, posterior view. Cervical 6 (PMU 24720/6) in D, lateral, E, anterior, and F, posterior view. Cervical 7 (PMU 24720/ 7) in G, lateral, H, anterior, and I, posterior view. Scale bar equals 100 mm. Abbreviations: lr, lateral ridge; ns, neural spine. C7, the anterior opening is transversely wider than that of C8, but this is likely due to distortion. The neural spines form low ridges and are knob-like in C6.
The prezygapophyses are broader than long and have dorsomedially angled articular facets; this changes to a more horizontal orientation with the prezygapophyses becoming increasingly divergent in the posterior cervicals. The prezygapophyses contact the diapophyses at their base, and together support the transverse processes. Unusually, the posterior margin of the right prezygapophysis on C3 bears a deep excavation, which does not appear to be pathological.
The postzygapophyseal processes extend posterolaterally and are triangular in cross-section. Their articular facets are ventrolaterally oriented, but more horizontally facing on C10. A distally sloping ridge extends along the dorsal surface of each postzygapophyseal process and becomes medially oriented throughout the cervical column. A post spinal fossa is present between the postzygapophyseal processes on C2 but is less evident in subsequent vertebrae.
Where preserved, the transverse processes on the post-axial cervicals are long and robust, resembling those of E. caroljonesa (McDonald et al., 2012a) and Edmontosaurus regalis Lambe, 1917(Campione, 2014. The diapophyses contact the base of the neural arch via a buttress, situated near the posteroventral margin of each transverse process in C3-C6, but shifted towards the anteroventral margin in C7-C10. The diapophyseal articular surfaces are circular to oval in outline, and ventrolaterally to laterally oriented along the column, but nearly horizontal in C10. Cervical Ribs-Of the four recovered cervical ribs, Wiman (1929) identified PMU 24720/13 as being from the right side of C6 and PMU 24720/14 from the left side of C9 ( Fig. 5J-L). All of the ribs possess a posteriorly directed prong-like shaft, which, together with the capitulum and tuberculum, diverge at about a 90°angle from the proximal end. Broken remnants of the tubercula are also preserved in articulation with the diapophyses on C4 and C6. A short spur-like process projects anteriorly from between the capitulum and tuberculum on each rib. The posterior section of the shaft is otherwise straight to slightly curved in lateral view and flat to concave in medial view. A ridge extends from the anterior process along the shaft for around one-third of its total length. The tuberculum and the capitulum terminate in anteroposteriorly expanded convex contacts for the diapophysis and parapophysis, respectively.
Dorsal Vertebra-The only recovered dorsal vertebra, PMU 24720/11, is virtually complete except for the distal-most extremity of the left transverse process ( Fig. 6A-C). The centrum is amphiplatyan with a flat anterior articular surface and a shallowly concave posterior articular surface. The anterior articular surface is 96/99 mm in maximum height/width and has a circular profile. The posterior articular surface is 110/96 mm in maximum height/width and more heart-shaped in outline because of its inset dorsal margin bordering the neural canal. The lateral surfaces of the centrum are pierced by nutrient foramina and converge medially to form a weak ventral keel.
parapophyseal articular surface is flush with the lateral surface of the neural arch and situated adjacent to the anterior base of the transverse process between the prezygapophysis and diapophysis. A thin ridge connects the lateral margin of the prezygapophyseal facet with the dorsal margin of the transverse process.
The transverse processes project perpendicularly relative to the vertical axis of the dorsal vertebra. Their anterior surfaces are flat, but become shallowly concave dorsally. A stout buttress extends along the ventral surface to the base of the neural arch and delimits a shallow triangular fossa that can be observed in posterior view. The maximum mediolateral width of the less distorted right transverse process is 130 mm when measured from the vertical midline of the vertebra.
The dorsal prezygapophyses are situated at the base of the neural spine. The prezygapophyseal facets are dorsomedially oriented and separated along their mid-line. The corresponding postzygapophyseal facets are ventrolaterally oriented and likewise medially separated. Stout buttresses extend ventrally from the postzygapophyses to diverge around the neural canal. Narrow ridges also ascend from the postzygapophysis along the posterolateral margins of the neural spine, and define a small fossa that terminates in a ridge extending along approximately two-thirds of the vertical height of the posterior edge towards the neural spine apex.
The distal extremity of the dorsal neural spine is expanded into a 'club-like' apex, which is 74/43 mm in maximum transverse width/ anteroposterior length, respectively. Similar 'club-like' neural spine apices occur in Bactrosaurus johnsoni, Gilmoreosaurus mongoliensis (Gilmore, 1933), and B. sicinskii (Godefroit et al., 1998;Prieto-Márquez and Norell, 2010;Prieto-Márquez, 2011b). The dorsal-most apical surface of the neural spine is rugose and was probably covered by cartilage in life. The lateral apical surfaces are shallowly concave and ornamented by striations. The anterior apical surface is transversely convex but becomes progressively flattened distally.
The right prezygapophysis of PMU 24720/12 has broken off, and the left prezygapophysis is incomplete. Nonetheless, both were evidently small, with a maximum mediolateral width equivalent to that of the adjacent neural canal. The corresponding postzygapophyses are similarly reduced and do not appear to have projected over the succeeding vertebra. Low ridges extend from the prezygapophyses and merge along the anterior edge of the neural spine, which is posteriorly inclined like those of I. bernissartensis, B. johnsoni, and E. caroljonesa (Norman, 1980;Godefroit et al., 1998;McDonald et al., 2012a); this inclination becomes more extreme in the posterior-most caudal vertebrae of E. regalis (Campione, 2014).
Hemal Arch-The recovered hemal arch PMU 24720/19 ( Fig. 6G-I) has a maximum proximodistal length of 152 mm, a maximum depth of the hemal canal at 74 mm, and a maximum width between the proximal facets of 43 mm. In general form, PMU 24720/19 resembles the more anterior hemal arches of Iguanodon bernissartensis, Eolambia caroljonesa, and Edmontosaurus regalis (Norman, 1980;McDonald et al., 2012a;Campione, 2014). The hemal spine of PMU 24720/19 is incomplete but the anterior and posterior surfaces are concave with a thin vertical midline ridge visible posteriorly. The proximal facets are borne on transversely narrow hemal processes, which enclose the hemal canal. Each facet surface is anteroposteriorly elongate and oval in outline with a posterodorsal inflection, suggesting that the hemal arch was probably posteroventrally angled when articulated with the corresponding caudal centra.
The proximal articular surface of PMU 24720/20 supports both the glenoid and coracoid contact. Its maximum dorsoventral height exceeds that of the distal extremity of the scapula, which is similar to most non-hadrosaurid hadrosauroids, except for B. johnsoni (Godefroit et al., 1998). The deltoid ridge of PMU 24720/20 is less robust than that of B. johnsoni (Prieto-Márquez, 2011a), but this trait is influenced by ontogeny (Brett-Surman and Wagner, 2007) and thus might reflect the relative osteological maturity of the T. sinensis holotype individual.
Sternal Plate-The right sternal plate, PMU 24720/21, lacks its anterior-most end (Fig. 7C-F) but clearly possesses a classic 'hatchet-shaped' outline typical of hadrosauroids (Norman, 2004). Its shaft is short and stocky with a D-shaped crosssection formed by the rounded ventral and flattened dorsal surfaces. The dorsal surface becomes progressively more concave towards the expanded anterior end. The distal end is rugose and elliptical in outline with a medial deflection relative to the long axis of the bone. Conversely, the distal extremity of the sternal plate is dorsoventrally flattened in Edmontosaurus regalis (Campione, 2014), or distinctly faceted in Probactrosaurus gobiensis  (see also Rozhdestvensky, 1967 for an English translation) and Bactrosaurus johnsoni (Godefroit et al., 1998;Norman, 2002). In comparison, the shaft of the sternal plate in PMU 24720/21 is shorter and broader than that of B. johnsoni (Godefroit et al., 1998), but this can vary intraspecifically in ornithopods (Brett-Surman and Wagner, 2007;Verdú et al., 2017;Prieto-Márquez and Guenther, 2018).
Humerus-The left humerus, PMU 24720/22 (Fig. 8A-D), has a sinuous profile and a mediolaterally broad proximal end with a medially offset humeral condyle. The lateral and proximal margins of the humerus form an obtuse angle, as evident in Eolambia caroljonesa and Edmontosaurus regalis (McDonald, 2012;Campione, 2014), but unlike the more perpendicular angle of the proximal humerus in Iguanodon bernissartensis (Norman, 1980). The prominent deltopectoral crest of PMU 24720/22 extends about 40% of the maximum distance along the humeral shaft and has a flattened lateral margin. Corresponding deltopectoral crest lengths of less than 50% are typical of non-hadrosaurid hadrosauroids, although this may be affected by ontogeny (Dilkes, 2001;Egi and Weishampel, 2002;Verdú et al., 2017). The shape of the deltopectoral crest in PMU 24720/22 is otherwise similar to those of Bactrosaurus johnsoni (Godefroit et al., 1998) and Nanyangosaurus zhugeii Xu, Zhao, Lü, Huang, Li, and Dong, 2000, yet is not as ventrally constricted or apically tapered as that of N. zhugeii (Xu et al., 2000).
The humeral shaft of PMU 24720/22 bears a small posterior prominence for attachment of the m. latissimus dorsi (see Campione, 2014). In addition, despite being partially encased in matrix, the medial and lateral condyles on the distal extremity are visibly sub-equal in size and separated by deep intercondylar grooves.
Ulna-The right ulna is broken into proximal (PMU 24720/23) and distal (PMU 24720/24) sections ( Fig. 8E-G), and lacks most of the olecranon process. Nonetheless, its general morphology is consistent with the ulnae of other hadrosauroids (Norman, 2004;Horner et al., 2004). A distally tapering trough is present on the anterior and posterior surfaces, and the proximal end of the shaft has an irregular 'T-shaped' cross-section similar to that reported in Edmontosaurus regalis (Campione, 2014). By contrast, the distal end of the shaft is elliptical in cross-section with a maximum diameter of 43 mm. The distal-most articulation surface is convex and highly rugose indicating the presence of cartilage.
Radius-Like the ulna, the right radius is broken into proximal (PMU 24720/25) and distal (PMU 24720/26) sections ( Fig. 8H-J) and has been mediolaterally compressed during diagenesis. Nonetheless, the shaft is visibly sub-trapezoidal in crosssection, and has a concave laterally sloping proximal articular surface that is bordered by a low posterior flange. The shaft of the radius is straight and anteroposteriorly flattened with a tapering anterior trough that expands distally. The distal articular surface is convex and rugose with a triangular outline in distal view.
The preacetabular process of PMU 24720/17 curves ventrally but is less deflected than those of I. bernissartensis and B. johnsoni (Norman, 1980;Godefroit et al., 1998;Verdú et al., 2017); this is alternatively reminiscent of Edmontosaurus regalis (Campione, 2014) the preacetabular process of which is similarly elongate relative to the maximum length of the ilium. An anteroposteriorly elongate ridge extends from the preacetabular process to around the midpoint of the dorsal third of the central plate (contra Prieto-Márquez and Norell, 2010), and corresponds to the attachment area for the sacral ribs.
The pubic peduncle of PMU 24720/17 is blunt, triangular, and does not align with the ischial peduncle when the ilium is oriented in life position. The pubic peduncle is mediolaterally thickened, like that of E. caroljonesa (McDonald et al., 2012a), but lacks the projecting pubic peduncle evident in Z. huangi .
Femur-The right femur, PMU 24720/27 (Fig. 10A-E), is slightly posteriorly bowed. Its proximal end was diagenetically compacted but the femoral head was clearly medially directed. The anterior trochanter is reduced and situated anterior to the greater trochanter, which has a more pronounced dorsal extension. A proximodistal groove separates the anterior and greater trochanters. The fourth trochanter was reconstructed in plaster but is mediolaterally flattened and posteriorly projecting. Attachment scars for the m. puboischiofemoralis internus and m. caudofemoralis longus border the fourth trochanter anteromedially (see Campione, 2014). The insertion for the m. caudofemoralis longus is delimited by a proximodistal ridge that borders a rugose area for the m. puboischiofemoralis internus (Campione, 2014).
The distal end of the femur is divided into lateral and medial condyles, which are anteroposteriorly expanded and approximately equivalent in size. These fully enclose the extensor tunnel anteriorly, as in other hadrosauroids (e.g., Hadrosaurus foulkii Leidy, 1858;Prieto-Márquez et al., 2006); however, the posterior flexor tunnel is also fully enclosed, which is a unique condition of Tanius sinensis. Tibia-The right tibia, PMU 24720/28 (Fig. 11A-D), is massive with perpendicularly offset proximal and distal articular surfaces. The proximal articular surface is anteroposteriorly expanded like those of Iguanodon bernissartensis, Bactrosaurus johnsoni, and Eolambia caroljonesa (Norman, 1980;Godefroit et al., 1998;McDonald et al., 2012a). The cnemial crest is laterally demarcated by a broad depression that extends along the proximal half of the shaft, as in many hadrosauroids (Godefroit et al., 1998). A deep cleft separates the medial and lateral condyles, whose combined width is broader than those of B. johnsoni, E. caroljonesa, and Edmontosaurus regalis (Godefroit et al., 1998;McDonald et al., 2012a;Campione, 2014).
The shaft of the tibia is mediolaterally flattened in crosssection, and the distal articular surface is sub-divided into lateral and medial malleoli separated by a deep groove.
Fibula-The right fibula, PMU 24720/29 (Fig. 11E-H), is slender and straight. A medial trough extends along the proximal third of the shaft, which is shorter than that documented in Edmontosaurus regalis where the medial trough continues until about the mid-length of the bone (Campione, 2014). The fibular shaft gradually tapers distally but expands again into a 'club-like' distal end. The distal articular extremity is 'teardropshaped' in outline and longest along is anteroposterior axis.
The proximomedial surface of PMU 24720/31 is shallowly concave to accommodate metatarsal II; a corresponding trough on the distolateral surface represents the contact for metatarsal IV. The distal articulation is rugose and noticeably less robust than those of B. johnsoni and Gi. mongoliensis (Godefroit et al., 1998;Prieto-Márquez and Norell, 2010), but is similar to that of Nanyangosaurus zhugeii (Xu et al., 2000). Wiman (1929) did not describe the right metatarsal II, PMU 24720/30 (Fig. 12D, E), which is poorly preserved and partly reconstructed in plaster. Nonetheless, the bone is mediolaterally compressed like the metatarsals of other hadrosauroids (Horner et al., 2004). The medial surface of the shaft is smooth, but the lateral surface is rugose for contact with metatarsal III. The exposed distal extremity is 'squared-off' with a shallow posterior trough that creates an irregular triangular outline.
Pedal Ungual-The pedal ungual phalanx, PMU 24720/31 (Fig. 12F, G), probably derives from the right digit II based on its asymmetrical shape and overall proportions (Godefroit et al., 1998;Zheng et al., 2011). The ungual process is 'hoof-like', similar to those of Bactrosaurus johnsoni, but more elongate in dorsal view (Godefroit et al., 1998;McDonald et al., 2012a) and lacking the tapered shape seen in Gilmoreosaurus mongoliensis (Prieto-Márquez and Norell, 2010). The dorsal surface of the ungual process in PMU 24720/31 is convex, while the ventral surface is concave. The tip of the ungual process is rugose but has been damaged by weathering. The proximal articulation is concave and oval in outline.

Pathological Bone Modifications
The caudal vertebra PMU 24720/12 bears three conspicuous and closely arranged pits on its left lateral surface (Supplemental Data 3, Fig. S1A, B). These show evidence of peripheral bone fiber remodeling consistent with callus formation and healing after a puncture injury (based on the pathological bone modification categories listed by Pardo-Pérez et al., 2018). The dorsal-most of these traces is situated close to the neurocentral suture and was deep enough to penetrate into the spongiosa. The immediately adjacent trace is drawn out into an elliptical gouge. We interpret these bone modifications as probable bite marks that compare with similar pathologies reported on hadrosauroid caudal vertebrae found elsewhere (e.g., Carpenter, 2000;Canudo et al., 2005;Rivera-Sylva et al., 2009;DePalma et al., 2013;Murphy et al., 2013). The only currently identified terrestrial carnivores from the Jiangjunding Formation faunal assemblage that would have been large enough to inflict such bite traumas are tyrannosauroid theropods. Tyrannosauroid remains were recovered in association with PMU 24720 and occur throughout other strata in the Wangshi Group succession (e.g., Hone et al., 2011;Brusatte et al., 2013;Poropat and Kear, 2013b).

Estimated Body Mass of PMU 24720
We calculated an average body mass range for the PMU 24720 holotype individual of Tanius sinensis at between 2091 kg and 3533 kg based on its humeral and femoral circumferences and the scaling equation presented in Campione and Evans (2012). This estimate includes an associated % prediction error and spans the mass ranges typical for hadrosaurids, including Prosaurolophus maximus Brown, 1916Brown, (2315, Lambeosaurus lambei Parks, 1923Parks, (2454, and Corythosaurus casuarius FIGURE 12. Tanius sinensis holotype pedal elements. Left metatarsal III (PMU 24720/31) in A, posterior, B, medial, and C, proximal view. Right metatarsal II (PMU 24720/30) in D, posterior and E, medial view. Right pedal ungual phalanx (PMU 24720/32) in F, dorsal and G, ventral view. Scale bar equals 100 mm for A-E; 50 mm for F, G. Abbreviations: tha, metatarsal III articulation; twa, metatarsal II articulation. Brown, 1914Brown, (2570. Tanius sinensis also appears to have been larger, on average, than other non-hadrosaurid hadrosauroids, such as Bactrosaurus johnsoni (1337-2258 kg) and Gilmoreosaurus mongoliensis (966-1631 kg) using the limb circumference data from Benson et al. (2014). We emphasize that T. sinensis was significantly smaller than the coeval Shantungosaurus giganteus, which reached masses of between 12,918 kg and 21,822 kg, and presumably occupied a very different ecological niche (Benson et al., 2014;Hone et al., 2014).

Other Skeletal Remains
Dorsal Rib-Wiman (1929) attributed a dorsal rib, PMU 22481 (Supplemental Data 3, Fig. S2A), to the Tanius sinensis hypodigm. This element is virtually complete, and includes the capitulum, tuberculum, and an entire shaft that lacks only its distalmost extremity. Comparisons with the articulated skeletons of Edmontosaurus regalis (Campione, 2014) suggest that PMU 22481 probably derived from the anterior to mid-dorsal region of the axial skeleton, based on the overall length of the rib and the ventrally deflected capitulum relative to the shaft.
The rib head on PMU 22481 has a dorsoventrally broad capitulum, which is acutely offset relative to the shaft. The tuberculum is 'teardrop-shaped' in articular aspect, and thus resembles those on the dorsal ribs of E. regalis (Campione, 2014), but differs from the more elliptical tubercula in Bactrosaurus johnsoni (Godefroit et al., 1998). The rib shaft of PMU 22481 is mediolaterally flattened.
Humerus-The referred left humerus, PMU 24721 (Supplemental Data 3, Fig. S2B, C) closely resembles the holotype specimen PMU 24720/22, although it is stouter and more medially bowed with an angular distal extremity on the deltopectoral crest. The humeral head of PMU 24721 is also slightly lunate in proximal view, which is similar to that of Edmontosaurus regalis (Campione, 2014), but distinguished from the distinctly rounded humeral head of Eolambia caroljonesa (McDonald et al., 2012a). In addition, the proximal articular surface on the humeral head of PMU 24721 is bordered by pronounced lateral and medial ledges and has less distinct muscle scarring for the m. latissimus dorsi. Distally, the radial and ulnar condyles of PMU 24721 are separated by an intercondylar groove, although this is demonstrably shorter than that of PMU 24720/22.
Radius-The referred radius, PMU 22483 (Supplemental Data 3, Fig. S2D-G), is closely compatible with the holotype specimens PMU 24720/25-26. Its slender shaft is transversely compressed and expands towards the proximal and distal articular ends. The shaft is also gently curved, like the radii of Bactrosaurus johnsoni and Eolambia caroljonesa (Godefroit et al., 1998;McDonald et al., 2012a). The proximal articular extremity of PMU 22483 is 'leafshaped' in outline, while the distal extremity is quadrangular.
Metatarsal-The referred metatarsal, PMU 22482 (Supplemental Data 3, Fig. S2H, I), has not been fully prepared. However, it appears to be straight and has a prominent 'teardrop-shaped' depression on the medial side of the distal articular surface. The anteroposterior width of the bone indicates that it is probably a metatarsal II based on comparisons with other hadrosauroids (see Zheng et al., 2011).
Prieto-Márquez (2011a) likewise reported that the deltoid ridge on the scapula of Bactrosaurus johnsoni becomes more robust and elongate with osteological maturity. Relative ossification must also have affected the proximal articular surface, which is dorsoventrally expanded relative to the distal end of the scapula in T. sinensis, and resembles other non-hadrosaurid hadrosauroids (Prieto-Márquez and Norell, 2010). Unusually, however, the anterodorsally curved acromion process in T. sinensis is equivalent to the derived condition characterizing lambeosaurines, but may be homoplastic given its occurrence elsewhere in iguanodontians (e.g., Iguanodon bernissartensis) and other non-hadrosaurid hadrosauroids, such as Eolambia caroljonesa (Prieto-Márquez, 2010a).
The pubic peduncle of T. sinensis is reduced relative to the ischial peduncle, and the central plate of the ilium is almost symmetrical with a depth/length ratio of 0.91; this is a plesiomorphic condition among hadrosauroids (Prieto-Márquez, 2010a). Similarly, positioning of the ventral apex of the supraacetabular process posterodorsal to the ischial peduncle approaches the placement in I. bernissartensis, as well as other non-hadrosaurid hadrosauroids (Prieto-Márquez and Norell, 2010). Lastly, reduction of the supraacetabular process to a narrow shelf in T. sinensis (limiting its lateroventral projection: sensu Prieto-Márquez, 2010a), together with a straight dorsal ilial profile are typical in hadrosauroids (Prieto-Márquez, 2007;Campione, 2014).

Taxonomic Implications
Although sometimes considered ambiguously diagnostic (e.g., Weishampel and Horner, 1990), the taxonomic validity of the Tanius sinensis type material (PMU 24720/1-32) has never been questioned, and, indeed, is reinforced by our identification of multiple potential autapomorphies in the postcranial skeleton. Unfortunately, however, the other proposed species of Tanius-Tanius chingkankouensis and Tanius laiyangensis-were established on postcranial remains from Upper Cretaceous (probable upper Campanian) strata of the Jingangkou Formation (Young, 1958;Zhen, 1976) that have questionable character state utility.
Tanius chingkankouensis- Young (1958) erected this species on the basis of an aggregated series of cervical and dorsal vertebrae, a sacrum, scapula, ilium, and ischium. Young (1958) referred this material to the genus Tanius because of overall similarity with the cervical vertebrae of PMU 24720. Young (1958) further designated these remains as the composite holotype of a separate species because the posterior sacral vertebrae bore a furrow along their ventral surfaces. Subsequent studies have, nevertheless, shown this particular state to be broadly distributed throughout Hadrosauroidea (Godefroit et al., 1998;Norman, 1998;Horner et al., 2004;Prieto-Márquez, 2008), and possibly also intraspecifically variable (Hone et al., 2014). The widespread occurrence of a ventral furrow on the cervical vertebrae was also acknowledged by Zhang et al. (2017), who reaffirmed affinity between T. chingkankouensis and PMU 24720 based on their shared possession of parallel dorsal and ventral margins on the distal scapular blade. Zhang et al. (2017) additionally argued that T. chingkankouensis must be valid because its scapular blade is straight, as opposed to the distinctively curved dorsal margin on the scapula of PMU 24720/20. Yet from our observations, PMU 24720/20 differs markedly in its narrower distal end on the scapular blade, strongly curved acromion process, and relatively more robust glenoid. The potential for intraspecific variability should also be considered since it affects the scapular blade in other iguanodontians (Guenther, 2009(Guenther, , 2014Verdú et al. 2017). Moreover, Zhang et al. (2017:195, 196) concluded that the ilium associated with T. chingkankouensis could not be referred to Tanius because the "apex of the supraacetabular process is located anterodorsal to the caudal tuberosity of the ischial peduncle", which is "a characteristic of saurolophines (Prieto-Márquez, 2008)". While we agree with this conclusion, we also point out that the PMU 24720/18 ilium differs in its straighter dorsal margin and more prominent supraacetabular processes. Weishampel and Horner (1990), Buffetaut and Tong-Buffetaut (1993), and Buffetaut (1995) all initially considered the T. chingkankouensis composite material to be an invalid taxon, which has since been regarded as either a junior synonym of Tanius sinensis (Horner et al., 2004;Lund and Gates, 2006), or a nomen dubium (Prieto-Márquez and Wagner, 2013). However, given that Zhang et al. (2017) excluded the ilium, and, as we show, generic referral of the scapula to Tanius is problematic, we deem a general similarity of the cervicals to be insufficient grounds for unambiguously asserting a taxonomic relationship. In our opinion, therefore, T. chingkankouensis should be regarded as a nomen dubium (sensu Prieto-Márquez and Wagner, 2013) that incorporates hadrosaurid skeletal elements (sensu Zhang et al., 2020), together with an ischium that "has a slightly inflated distal end" "characteristic of basal hadrosauroids (Prieto-Márquez, 2008)" Zhang et al. (2017:196), and also probably various other non-hadrosaurid hadrosauroid bones.
Tanius laiyangensis-Zhen (1976) named T. laiyangensis from a sacrum and right ilium, which Zhang et al. (2020) recently interpreted as representing a single individual. Zhen (1976) did not provide specific locality information for these remains, but Zhang et al. (2020) confirmed their recovery from the same horizon as the composite material of T. chingkankouensis. Tanius laiyangensis was distinguished by a combination of features, including: "taller neural spines, more robust sacral bars, elongate ventral sulcus and horizontally oriented transverse processes of the sacrum, and the proportionally longer postacetabular process of the ilium" (see Zhang et al., 2020:793). Conversely, a detailed re-examination of T. laiyangensis by Zhang et al. (2020) concluded that all these features were consistent with advanced hadrosaurids, and thus proposed reassignment as an indeterminate kritosaurin. This concurs with previous assertions that T. laiyangensis is incompatible with Tanius (Weishampel and Horner, 1990;Buffetaut, 1995;Horner et al., 2004;Zhang et al., 2017), and supports our own observations that PMU 24720/18 can be distinguished from the T. laiyangensis remains by its shorter and deeper iliac profile, more prominent supraacetabular process, shallower embayment between the preacetabular ridge and pubic peduncle, and relatively more ventrally deflected preacetabular process. In any case, we agree with Zhang et al. (2020) that T. laiyangensis is an invalid taxon erected on indeterminate hadrosaurid material.
Tsintaosaurus spinorhinus-Rozhdestvensky (1964) considered Tsintaosaurus spinorhinus to be a junior synonym of Tanius sinensis, suggesting that the missing rostral section of the skull in PMU 24720/1a might have supported a crest. Rozhdestvensky (1974Rozhdestvensky ( , 1977 subsequently revised this interpretation, claiming that T. spinorhinus and T. sinensis could be congeneric but not conspecific. Weishampel and Horner (1990) further argued that T. spinorhinus was based on composite lambeosaurine and hadrosaurine remains, and that its famous spike-like crest was a displaced nasal. This reinterpretation was justified via a personal communication that was later elaborated by Taquet (1991), who proposed that T. spinorhinus was probably a crestless hadrosauroid and a junior synonym of Tanius sinensis. Buffetaut and Tong- Buffetaut (1993) refuted this suggestion, citing key differences in cranial osteology. Prieto-Márquez and Wagner (2013) and Zhang et al. (2017) also both conclusively demonstrated that a cranial crest was present in T. spinorhinus, as well as the lack of a comparable structure in T. sinensis (see Zhang et al., 2017:196, fig. 4A), a conclusion that we reaffirm from our first-hand inspections of PMU 24720.

CONCLUSIONS
Here, we undertake a detailed re-description of the diagnostic holotype postcranial skeleton (PMU 24720/2-32) and other referred elements (PMU 22481, PMU 22482, PMU 22483, PMU 24721) of one of the first dinosaur taxa identified from China-Tanius sinensis. These remains constitute some of the many historically important dinosaur fossils recovered by the Sino-Swedish expeditions of the 1920s, and also form part of an ongoing project to restore and re-evaluate the significant collections of this material held in perpetuity at the PMU in Sweden (Poropat, 2013;Poropat andKear, 2013a, 2013b;Borinder et al. 2016).
Based on our firsthand examination of these remains we identify a unique postcranial character state combination that defines T. sinensis among recognized hadrosauroid taxa. These include a scapula with strongly curved dorsal margin, recurved and anterodorsally directed anterior end on the acromion process, and dorsoventrally expanded proximal versus distal extremity, a reduced deltopectoral crest on the humerus, a deep central plate on the ilium, a 'club-like' distal end on the fibula, and a proximodistally elongate metatarsal III. In addition, at least four postcranial apomorphies serve to establish T. sinensis as a distinct non-hadrosaurid hadrosauroid species: (1) the presence of a tall neural spine on the dorsal vertebra that equals up to three times the height of the corresponding centrum; (2) a reduced postacetabular ridge on the ilium; (3) uniquely complete fusion of the medial and lateral condyles on the femur to form a fully enclosed flexor tunnel; and (4) a lunate profile on the proximal articular surface of metatarsal III. Furthermore, we report a series of associated pathological bone modifications consistent with healed bite traces found on the holotype caudal vertebra (PMU 24720/12), which might evince an encounter with a predatory tyrannosauroid.
The recognized fossils of T. sinensis are thus of significance for elucidating hadrosauroid diversity and distribution across the Laurasian landmasses towards the end of the Mesozoic. In particular, the faunal context of T. sinensis as part of a rich succession of dinosaur assemblage spanning the Jiangjunding and Jingangkou formations (see Zhang et al., 2017) offers the potential to reconstruct ecological segregation between coeval species and will provide a focus for the next of our forthcoming studies on the PMU vertebrate fossil collection from the Wangshi Group of China.