The first dinosaurs from the Early Cretaceous Hami Pterosaur Fauna, China

The Early Cretaceous Hami Pterosaur Fauna in Northwest China preserves a large number of specimens of the sexually dimorphic pteranodontoid pterosaur Hamipterus tianshanensis, including 3D eggs and embryos. During the last decade, several more fossils have been collected in this area, including three somphospondylan sauropod specimens. The first is Silutitan sinensis gen. et sp. nov., which consists of an articulated middle to posterior cervical vertebrae series. The second, Hamititan xinjiangensis gen. et sp. nov., consists of an incomplete articulated caudal sequence that could be assigned to lithostrotian titanosaurs based on the strongly procoelous caudal vertebrae with lateral concave surface, as well as marked ventrolateral ridges. The third specimen consists of four sacral vertebral elements, apparently unfused, with exposed camellate internal bone and regarded as somphospondylan. Cladistic analyses based on different datasets recovered Silutitan sinensis as an euhelopodid closely related to Euhelopus and Hamititan xinjiangensis as a titanosaur. Besides the pterosaur Hamipterus and one theropod tooth, these dinosaurs are the first vertebrates reported in this region, increasing the diversity of the fauna as well as the information on Chinese sauropods, further supporting a widespread diversification of somphospondylans during the Early Cretaceous of Asia.

Geological settings. The Tugulu Group is composed of medium-to-fine grained grey-green sandstones alternating with red to brown-red mudstones beds. In Turpan-Hami Basin, this stratigraphic unit includes, from bottom to top, the Sanshilidadun, Shengjinkou, and Lianmuqin formations 38,45 . The Hami Pterosaur Fauna comes from the Shengjinkou Formation whose lacustrine sedimentary sequence is mainly composed by the gray-white sandstones, within some tempestite interlayers made up of brown mudstone breccias 38,40,45 . While almost Hamipterus specimens and their eggs were found in tempestite interlayers 40,41 , the new sauropod specimens were discovered in lacustrine sandstones. The three sauropod specimens were collected from different sites which are 2-5 km away from each other, all showing the presence of Hamipterus. The horizon where the cervical vertebrae sequence (IVPP V27874) was collected is particularly rich in pterosaur specimens. The layers from which the other two sauropod specimens (HM V22 and IVPP V27875) were recovered are positioned about 2 to 3 m higher than the latter. Occasionally, isolated bones of both, sauropods and theropods (undescribed) are also found.

Background
Other fossil sites close to the Hami region are the strata from the Junggar Basin, especially the outcrops near the Mazong Mountain 46 . Among the sauropod remains known from the Mazong Mountain (also called the "Mazongshan area" 1 ), are two well-known titanosauriform taxa 46 Zhao, 1993, sensu 47 ). Fushanosaurus qitaiensis Wang et al., 2019, a putative titanosauriform, was recovered from the Shishugou Formation and is based on a right femur 48 . From all listed somphospondylans recovered from the Mazong Mountain, the only one that shows comparable elements with one of the specimens described here (IVPP V27874-Silutitan) is Qiaowanlong.
Recently, during the redescription of Klamelisaurus by Moore et al. 47 , the phylogenetic analyses conducted using the dataset from Carballido et al. 49 and González-Riga et al. 50 found the "core Mamenchisaurus-like taxa" (Klamelisaurus and Mamenchisaurus sinocanadorum included), and some taxa (e.g., Euhelopus) thought to represent somphospondylans outside of Neosauropoda 47 , which is inconsistent with most sauropod cladistics analysis literature [49][50][51][52][53][54] . It is important to note to the authors highlighted a need for further redescriptions due to conflicting phylogenetic results 47 , as well as revisions of these sauropods (particularly Mamenchisaurus and Omeisaurus). The comparisons with this "core-Mamenchisaurus-like taxa" and consequently full revision of the anatomy and systematics of mamenchisaurids is beyond the scope of this paper.
The specimens described here (IVPP V27874, IVPP V27875 and HM V22) are compared with the following somphospondylans: Abdarainurus barsboldi Averianov & Lopatin, 2020 55 , Baotianmansaurus henanensis Zhang www.nature.com/scientificreports/ zdanskyi neck, one of the most complete Somphospondyli known to date 7 , Silutitan sinensis gen. et sp. nov. (IVPP V27874) represents the cervical sequence 10 to 15, and we will refer as such. The body length is estimated as >20 m by comparison with length of the cervicals of Euhelopus 7 . All elements are strongly opisthocoelous and decrease gradually in length posteriorly, a common condition for sauropods (Table 1). The articular surfaces seem to be mediolaterally wider than dorsoventrally tall (since its still covered by matrix), similar to Qiaowanlong 2 but differing from Euhelopus 7 and Erketu 28,29 . As expected for titanosauriform sauropods and some mamenchisaurids 47 , the vertebrae show camellate pneumatic structure. Silutitan has the lateral margin of the centra slightly excavated, with the pneumatic fossae (pleurocoel) restricted anteriorly and thus differing from Erketu and Euhelopus (Fig. 3). The fossae of the new species are further placed ventral to the diapophysis, which is best observed in cervical vertebra 12. The parapophysis presents its dorsal surface excavated and deflected ventrally as in Qiaowanlong 2 , Euhelopus 7 and Erketu 28 . The ventral surfaces are concave and show some sharp ridges formed by the parapodiapophyseal laminae (PPDL), that extend along the ventrolateral edges of the centrum, as in Euhelopus 7 but less transversely developed in the latter. The PPDL are more ventrally developed than in Euhelopus, forming a flange-like structure, which do not reach the posterior articulation of the centrum (Fig. 3). As in many somphospondylans, Silutitan presents low and anteroposteriolly short neural arch in almost all preserved cervicals, with exception of the last one. The low and anteroposteriolly short neural arches are observed for example, in the euhelopodids Euhelopus 7 , Qiaowanlong 2 and Erketu 28,29 ; in the somphospondylan Phuwiangosaurus 23 and in the titanosaurs Arrudatitan 58 , Bonatitan 61 and Trigonosaurus 78 .
The prezygapophyses are relatively large but with thin prezygodiapophyseal laminae (PRDL). The PRDL is anteriorly projected forming a developed flange ("pre-epipophysis", sensu 7 ), which reaches the articular facet of the prezygapophyses. The PRDL forming a developed flange is observed in the somphospondylans Euhelopus 7 , Erketu 28,29 , Phuwiangosaurus 23 , Huabeisaurus 22 and on the eusauropod turiasaurian Moabosaurus 81 . The epipophyseal-prezygapophyseal lamina (EPRL) is present in all recovered cervical vertebrae (Fig. 3). The EPRL is also observed in Klamelisaurus 47 , in the somphospondylans Euhelopus 7,47 , Qiaowanlong 2 , Phuwiangosaurus 23 and in the titanosaur Kaijutitan 67 . As in the taxa cited, the EPRL divides the spinodiaphophyseal fossa (sdf) into two subfossae, but more similar to the condition observed in Euhelopus 7 , two subfossae present, one located dorsally (sdf1) and the other ventrally (sdf2). The EPRL is absent in Erketu and only present on the posteriormost cervical vertebrae of Kaijutitan 67 and in Qiaowanlong 2 , as in some middle to posterior cervical elements of Phuwiangosaurus 23 . Moore et al. 47 presents an extensive comparative anatomy of the EPRL among Klamelisaurus, Euhelopus, Kaijutitan, and Phuwiangosaurus, as well as other sauropod taxa. But in Kaijutitan and Phuwiangosaurus, the EPRL is almost vertically oriented, while this structure is diagonal in Qiaowanlong, and horizontal in Euhelopus and Silutitan.
The diapophyses are relatively short, directed laterally and curved ventrally, as in several somphospondylans (e.g. 2,7,28,29 ). Also, the diapophyses-tuberculum contact surface is constricted anteroposteriorly in the middle and posterior cervical vertebrae. Albeit this feature is presented in 8th cervical of Erketu ( 29 : Fig. 1C) and in the posterior-most cervical of Euhelopus ( 7 : Fig. 11) and Daxiatitan ( 3 : Fig. 1a), we note that is not the same condition as observed in Silutitan. In Euhelopus, Daxiatitan and Erketu, the "constriction" is presented solely on anterior www.nature.com/scientificreports/ face at the contact surface of diapophysis and tuberculum, while in Silutitan it is presented on the anterior and posterior faces (Fig. 3).
The new taxon shows a developed posterior centrodiapophyseal lamina (PCDL) that is directed ventroposteriorly and bifurcated in cervical 11 to 15. Except for the lognkosaurian titanosaur from Brazil, Austroposeidon, (MCT 1628-R, 52), the bifurcation of the PCDL (Fig. 3) in Silutitan is unique. The bifurcation of the PCDL is placed close to the diapophysis and not as posterior as in Austroposeidon, where it originates on the centrum and is not as deep 60 . Also, since Austroposeidon has this feature observable in the posterior-most preserved cervical vertebra, it is unknown if the bifid PCDL was presented along the four last cervical vertebrae (as in Silutitan), what clearly differentiates Silutitan from the Brazilian species. Still regarding the PCDL, this structure differs in the new species from Qiawanlong where it is more horizontal 2 . The postzygodiapophyseal fossae (PODF) in all vertebrae are not so deep as in other somphospondylians but are well delimited by the inclined  47 in Euhelopus, Klamelisaurus and several "core Mamenchisaurus-like taxa". However, in Euhelopus and Klamelisaurus the ventrolaterally bifurcated PODL is observed solely on the posterior-most cervical vertebrae, while in Silutitan this feature is observed in the 9th, 10th, 12th, 14th and 15th element (Fig. 3), being less developed in the anterior-most cervical vertebrae. This persistence of the ventrally bifid PODL along most of the cervical vertebrae of Silutitan is considered an autapomorphy since, to our knowledge, this feature is not observed in any other taxa.
Silutitan also presents a variation of the development of the ventrally bifid PODL along the cervical vertebrae sequence. On the anterior-most vertebrae preserved, the bifurcation of this lamina is restricted more anteriorly and positioned almost exclusively ventrally; but on the posterior-most cervical, the bifurcation extends in length and becomes more ventrolaterally than the other cervical vertebrae. We regard the development of the ventrolaterally bifurcated PODL through the cervical series unique to Silutitan.
Silutitan presents, as other sauropods, developed epipophyses on the dorsal surface of the postzygapophyses, but unlikelythe euhelopodids Euhelopus zdanskyi 7 , Erketu ellisoni 28,29 , Phuwiangosaurus 23 and Qiaowanlonq 2 , the epipophyses of Silutitan are elongated. The neural spines are well preserved in cervical 14 and 15, showing that they are low and reduced anteroporsteriorly.
An incomplete lower jaw of a pterosaur was recovered associated with this specimen (Fig. 3F). Despite its incompleteness, this specimen shows the same anatomy of the sole pterosaur collected in this region, Hamipterus tianshanensis 38 , and is therefore referred to this taxon. Etymology. "Hami" refers to Hami city where the specimen was found, "titan", from the giants of the Greek myths and commonly used to name titanosaur taxa.

Diagnosis. The same for the species.
Hamititan xinjiangensis new species.
Holotype. An articulated series of seven anterior to middle caudal (HM V22), including the proximal portions of three chevrons, housed at Hami Museum (Figs. 2, 4; Table 2).    It should be noted that strongly procoelous caudal vertebrae are known for Daxiatitan 3 , Dongbeititan 9 and Xianshanosaurus 14 . Hamititan shares with these taxa the lack of pleurocoels and the prezygapophyses positioned close to the proximal margin of the centrum. All four taxa also show the neural spine oriented posterodorsally. Hamititan differs from Daxiatitan 3 and Dongbeititan 9 by showing well-marked ventrolateral ridges. Although such ridges are also recorded in Xianshanosaurus 54 , the latter differs from Hamititan by having longer transverse processes that are also more horizontal 14 . Lastly, Hamititan differs from these three taxa by having stouter prezygapophyses 3,9,14 . Furthermore, Hamititan shows an abrupt change of the orientantion of the transverse processes throughout the caudal series.
The new taxon lacks internal spongy bony tissue as many titanosaurs 84 . The procoelous caudal vertebrae is not exclusively present in titanosaurs and has been recorded in several eusauropods such as Mamenchisaurus and Wamweracaudia keranjei 53 , the turiasaurian Moabosaurus utahensis 81 . However, in lithostrotian titanosaurs the centra presents at the condylar convex a distinct rim, which separates the condyle from the lateral surface of the main body of the centrum 82 . This feature is also observed in the anterior caudal and middle vertebrae of some derived titanosaurs, such as Trigonosaurus and Baurutitan and the unnamed titanosaur NHMUK R5333 82 . The new Chinese species has the ventral surface of the centrum slightly concave anteroposteriorly (Fig. 4) 75 and Xianshanosaurus (based on the scorings of 54 ). The 4th, 5th and 6th elements have a smooth excavation on the lateral surface of the centrum that does not form a pneumatopore. The chevron facets are poorly developed, being better developed on 8th and 9th vertebrae.
The neural arches are remarkably high when compared with the total height of the vertebrae (at least on 5th and 6th caudal vertebrae). The neural arches are located on the half of the centrum length on more anterior caudal vertebrae but become closer to the anterior half along the caudal series (Table 2), as in Abdarainurus 47 . In the most anterior caudal vertebrae, the neural arch is anteroposteriorly short at its base but broader dorsally at its end, showing a sagittal expanded neural spine, similar to Lirainosaurus 68,69 and Tengrinsaurus 76,77 . However, Hamititan differs from this Spanish titanosaur as the neural arch that does not reach the anterior border of the centrum as in the former (Fig. 4). In lateral view, the neural arch of the caudal vertebra 5th shows a deep postzygapophyseal spinodiapophyseal fossa [POSDF] delimited by equally thick posterior spinodiapophyseal lamina (SPDL), and the postzygodiapophyseal lamina (PODL), which delimit the are much more robust than the normally presented in other sauropods-a feature so far only observed in Hamititan xinjiangensis (Fig. 4).
The prezygapophyses and the postzygapophyses are not preserved except for the 5th and 6th. In lateral view, the zygapophyseal pedicels on the 7th, 8th and 9th are strongly curved and directed upwards (Fig. 4), like the anterior to middle caudal vertebrae of Opisthocoelicaudia (plate 4, Fig. 1b 25 ). In caudal elements 5 and 6, the prezygapophyses are relatively long and project mainly anterodorsally. In the new species, it is more vertically oriented similar to Phuwiangosaurus 23 76,77 ], Abdarainurus barsboldi 55 , as well as in the saltasaurines such as Neuquensaurus (e.g. 71 ).
The transverse processes are placed ventral to the neural arch-centrum contact and have a triangular base, presenting a ridge-like rugosity on the ventral surface. In the 4th element they are short and become longer in the subsequent caudal vertebrae. From caudal vertebra 4th to the 9th, the transverse processes are laterally and with slight upward deflection abruptly changing to a downward deflection on the 10th and 11th. To our knowledge, such an odd and abrupt change of the deflection of transverse processes is reported for the first time in sauropods Three proximal ends of chevrons are preserved and found articulated with the 8th, 9th and 10th anterior caudal vertebrae. The proximal process is laterally compressed and curves gently backwards. Since there is no taphonomic evidence of any deformation, we regard the changes of orientation of the transverse process as an anatomical feature characteristic of this species. It should be noted that the morphology of the laminae and the transverse processes are consistent throughout the caudal series, corroborating with our interpretation.
A small theropod tooth was found associated with this caudal sequence (Fig. 4). It is very curved and has no root. The crown is strongly compressed. The subquadrangular denticles are only preserved on the middle of the distal carina.

SOMPHOSPONDYLI Wilson & Sereno, 1998
Specimen. One incomplete sacrum (IVPP V27875) consisting of 4 fragmentary elements with co-ossified centra and some sacral ribs (Figs. 2, 5 www.nature.com/scientificreports/ Description of IVPP V27875 and comparisons. The specimen (IVPP V27875) consists of the remains of at least four incompletes sacral centra with incomplete sacral ribs. They are not completely fused, with clearly marked sutures between sacral 4 and 5. Compared with complete somphospondylan sacra (e.g. 22,64,72,75,[78][79][80] ), we regard them tentatively to represent the sacral 2 to 5. The most complete centra are of the sacral 2 and 5, which are short and opisthocoelic (Table 3). In ventral view, the centra are transversely convex (Fig. 5), such as in Diamantinasaurus 64 . The dorsal surface is completely eroded with the camellate internal bone exposed. On ventral view, neither centra possess external pneumatic fossae, despite the evidence of closed foramina in sacral 4 and 5 (Fig. 5). The absence of pneumatic fossa distinguishes IVPP V27875 from some somphospondylan like Phuwiangosaurus 23 , MLP 46-VIII-21-2 80 , the more derivate titanosaurs Saltasaurus 75 and Neuquensaurus 72 . This specimen also differs from Rapetosaurus 60 that shows deep lateral pneumatic foramina. The ventral surface of the centra is concave differing from MLP 46-VIII-21-2 80 and Diamantinasaurus 64 . The sacral ribs are robust, especially in the second sacral vertebra). The best-preserved rib is long, directed laterally and has an extensive and mediolaterally deep fossa on the ventral face (Fig. 5) that is reported for the first time in somphospondylan sauropods. Three well-developed pneumatic foramina are observed on the anteromedial surface of this rib. Phylogenetic analysis. The phylogenetic relationships of the somphospondylans Silutitan sinensis, Hamititan xinjiangensis and IVPP V27875 (Figs. 6, 7, 8) were evaluated using the data matrices focused on Titanosauriformes published by Filippi et al. 67 and Mannion et al. 53,54 , since those datasets are focused on somphospondylans. While the study of Filippi et al. 67 , focuses on titanosaur interrelationships, Mannion et al. 53,54 is more concerned with the interrelationships of somphospondylans, especially the basal ones. The data matrix was edited with Mesquite version 3.6 85 and cladistic analyses were conducted using the software T.N.T. 1.5 86 . Unstable taxa were detected a priori, using the 'iterpcr' method in TNT 87 .
Since the materials studied here do not show overlapping elements, we performed three different combinations for each dataset in order to better understand the phylogenetical positioning of each specimen. First, we coded all three specimens as a single taxon. Secondly, we combined Silutitan and Hamititan as one taxon and  were found as the sister-taxon of Euhelopus, sustained only by cervical characters (Fig. 6A). Scoring Silutitan (IVPP V27874) and Hamititan (HM V22) as separate taxa resulted in 128 MPTs of 1331 steps. The strict consensus tree is less resolved than the one of the original study. Silutitan had no effect on the topology and is recovered as the sister-taxon of Euhelopus. This relationship is supported by four characters: the shape and orientation of the parapophysis along the cervical series (122: 0 > 2); the parapophysis shape on middle and posterior cervical vertebrae (147: 0 > 1); the epipophyses shape (129: 0 > 1) and the lateral profile of the neural spine of the posterior cervical vertebrae (149: 0 > 0).

Results on Mannion et al. dataset.
As in the dataset before, when all three specimens were considered as the same taxon, a large polytomy for Titanosauriformes is recovered (Supplementary Information).
Mannion et al. 54 used two procedures running their dataset. First, the used equal weighting and after they applied extended implied weighting with different value of k (see 53,54 for details). Scoring IVPP V27874 (Silutitan), HM V22 (Hamititan) and IVPP V27875 as a single taxon with equal weighting resulted in 54,450 MPTs with 2672 steps. In this analysis, the composite taxon is also recovered a sister-taxon of Euhelopus but retrieved in a large polytomy with other somphospondylans. The inclusion of the two news specimens as a single taxon resulted in a much less resolved consensus tree than the one recovered in the original study 54 . The same results were achieved when eliminating the sacral elements (IVPP V27875) and scoring IVPP V27874 (Silutitan), HM V22 (Hamititan) as representing the same taxon (Fig. 7A). Running the dataset with the extended implied  www.nature.com/scientificreports/ weighting (k = 9, Fig. 7B), has not changed the relation of Silutitan + Euhelopus but brought a better resolution within Euhelopodidae, being similar to the topology of the original study by Mannion et al. 53,54 . Scoring Silutitan (IVPP V27874) and Hamititan (HM V22) as separate taxa without extended implied weighting resulted in 4148 MPTs with 2616 steps. The strict consensus is less resolved than the one published by Mannion et al. 54 with only a few clades in Somphospondyli recovered. After pruning the unstable taxa, we find 2615 MPTs and most of the topology of the original study recovered (Fig. 8A). Silutitan falls as sister-taxon of Euhelopus in the Euhelopodidae, with the remaining taxa of this clade collapsed. Hamititan was recovered in a small polytomy with basal titanosaurians, outside Colossosauria (Fig. 8A). The clade formed by Silutitan + Euhelopus is supported by three characters (Char. 118: 0 → 1; Char. 121: 1 → 0, and Char. 128: 0 → 1). Hamititan is nestedad as a basal titanosaurian, what is supported by two synapomorphies (Char. 182: 0 → 1 and Char. 489: 1 → 0).
Running the matrix using extended implied weighting with k = 9 (Fig. 8B) recovered Hamititan as a derived titanosaur, as the sister-taxon of (Aeolosaurus + Rapetosaurus). Comparing the topologies of the consensus trees using both matrices showed the same results regarding Silutitan sinensis, always recovered as the sister-taxon of Euhelopus, suggesting that they form a separate clade within Euhelopodidae (Fig. 8B).
The position of Hamititan xinjiangensis, however, is more instable. Regarding the dataset of Filippi et al. 67 , this taxon is recovered as a Colossossauria, whose position regarding Rincosauria, Lithostrotia and Lognkosauria cannot be established at the time being.
Regarding the dataset of Mannion et al. 54 , Hamititan xinjiangensis is recovered in rather extreme positions. When equal weighting is applied, Hamititan xinjiangensis is found on the base of Titanosauria while if implied weighing is applied (k = 9), this taxon moves to the top, as the sister-taxon of (Aeolosaurus + Rapetosaurus).
Nonetheless, both datasets, shows that Hamititan xinjiangensis is well nested within Titanosauria, distant from Silutitan sinensis, corroborating that both represent quite distinct taxa.

Discussion
Comments about other East Asian sauropods. The diversity of somphospondylan sauropod genera from the Cretaceous of East Asia increased vastly in the last decades 47,[52][53][54] . Several taxa, however, a lack comparable elements with Silutitan and Hamititan: Zhuchengtitan is represented by a single humerus 19 , Liubangosaurus consists of a set of dorsal vertebrae 10 , Borealosaurus known by two distal caudals, a humerus and one tooth 16 , and Gannansaurus erected based on one middle caudal and one posterior dorsal vertebra 20 .
Silutitan sinensis can be assigned to the Euhelopodidae based on the presence of a thick EPRL dividing the spinodiaphophyseal fossa into two subfossae, and the pendant cervical ribs. This taxon cannot be compared with the euhelopodids Gobititan, Liubangosaurus, and "Huanghetitan" ruyangensis 1,10,15 that do not show comparable elements.
Silutitan differs from Qiaowanlong 2 , that shows the EPRL diagonally oriented, the neural spine bifid, and two fossae on the lateral surface of the centrum of the cervical vertebrae.
The cervical vertebrae of Yongjinglong differ from Silutitan by having large pleurocoels that almost occupy the entire lateral side of the cervical vertebrae 4 .
Silutitan can be distinguished from Euhelopus, the first discovered euhelopodid 6,7 by the absence of a median turbercle on the cervical vertebrate and by having the PODL elongated and ventrolaterally bifurcated in all preserved cervical vertebrae.
Huabeisaurus differs from Silutitan by the lack of bifurcated PRDL and vertically oriented EPRL, present in the new species.
The cervical vertebrae of Ruyangosaurus are not well preserved 11,12 . As far as comparisons are possible, this taxon differs from Silutitan by lacking pleurocoels in the anteriormost cervical vertebrae (but having a deep and large pleurocoel in the posteriomost cervical element) and having longer centra. The two posterior-most cervical vertebrae of Ruyangosaurus are stouter than the Silutitan and show a thick and not bifurcated posterior centrodiapophyseal lamina (PCDL).
Hamititan xinjiangensis was recovered as a member of the Titanosauria. Several Early Cretaceous Chinese sauropods had been originally classified in this clade (e.g. 4,17 ), but recent studies regarded some as representing other lineages of Titanosauriformes, such as Gannansaurus 20 , Borealosaurus 16 , Yongjinglong 53,54 , Dongyangosaurus and Jiangshanosaurus 54 . In any case, the caudal vertebrae present in Gannansaurus 20 are from the posterior region of the tail, being amphycoelous and therefore differing from Hamititan. The same difference can be observed in the anterior and posterior caudal elements of Jiangshanosaurus 54 . Borealosaurus also has only posterior caudals 16 , but opisthocoelous. Huabeisaurus shows a fairly complete tail 22 , with the anterior being opisthocoelic while the middle and posterior are amphycoelic.
Other titanosaurian taxa recovered from China also differ from Hamititan. The putative titanosaur Dongyangosaurus has two anterior caudal vertebrae, that shows the anterior and posterior surfaces of the two caudal centra gently concave 18,54 and proportionally shorter than the ones of Hamititan. Daxiatitan 3 , Xianshanosaurus 14 and Dongbeititan 9 show strongly procoelous anterior caudal elements and share with Hamititan the lateral surface of the centra lacking pleurocoels, prezygapophyses positioned beyond the proximal margin of the centrum, and the neural spine oriented posterodorsally. Hamititan presents well-marked ventrolateral ridges that are absent in Dongbeititan and Daxiatitan 3,9 . Xianshanosaurus further differs from Hamititan by having longer and horizontally placed transverse processes 14 , as well as the presence of lateral openings. www.nature.com/scientificreports/ Although our phylogenetic analyses did not recover Hamititan as an euhelopodid, we have also compared this species with members of this clade. It should be noted that most taxa referred to the Euhelopodidae lack caudal elements, including Euhelopus 7 . Depending on the dataset, some authors do regard Tangvayosaurus, Phuwiangosaurus, Ruyangosaurus and Gobititan, originally described as titanosaurs 1,11,12,23 , as euhelopodids 53,54,82 . They differ from Hamititan by having proportionally shorter caudal vertebrae with nearly flat articular ends, presenting an amphiplatyan condition. Gobititan shows variation in some of the posterior-most caudals 1 , that can be slightly procoelous, but the anterior elements, as pointed out, differ from Hamititan. Hamititan further differs from all the above-mentioned taxa by having stouter prezygapophyses, taller neural arches, and presents the unique morphology of the transverse processes that show an abrupt change from being directed upward in the anterior elements to being directed downward in the posterior ones.
Comments about Euhelopodidae. The Euhelopodidae is a rather problematic clade of sauropod dinosaurs. This name was first proposed by Romer 88 containing five genera: Mamenchisaurus, Chiayusaurus, Omeisaurus, Tienshanosaurus, and Euhelopus. The original taxa of Romer have not been recovered as a clade by most of the recent phylogenetic analysis, with Euhelopus mostly recovered as Somphospondyli (e.g. [49][50][51][52][53][54] ). The first phylogenetic definition of Euhelopodidae was formulated by D'Emic 89 as the clade containing "neosauropods more closely related to Euhelopus zdanskyi than to Neuquensaurus australis" ( 89 : pg. 626). Some authors recovered Euhelopodidae as paraphyletic with Euhelopus nested far from other euhelopodids, such as Erketu and Qiaowanlong 49,63 . A more extreme result was obtained by Moore et al. 47 that recovered Euhelopus outside of Macronaria, highlighting the necessity of reviewing this and closely related taxa. In any case, in the phylogenetic analyses presented here, we did consistently recover we recovered Silutitan sinensis consistently as the sistertaxon of Euhelopus.
Other associated taxa. Two of the new specimens described here were found associated with elements of other taxa. Close to the 10th cervical vertebrae of Silutitan sinensis (IVPP V27874), an incomplete lower jaw attributed to the pterosaur Hamipterus tianshanensis is preserved. The association of pterosaur with sauropods have not commonly been reported in the literature (e.g. 90 ). It is not clear, however, if there were any more specific palaeoecological interactions between these taxa and this association is likely due to taphonomy.
Regarding Hamititan xinjiangensis, a small theropod tooth was observed above the neural arch of the 6th caudal vertebrae. It is the first report of theropod dinosaur discovered in this area. Theropod teeth are commonly found associated with the sauropod remains, generally suggesting that theropods could have fed on their carcasses (e.g. 91,92 ). Although this might also have been possible here, no evidence of tooth marks has been observed in this specimen, or on the other sauropod material described here.

Conclusions
The discovery of Silutitan sinensis and Hamititan xinjiangensis increased the sauropod diversity of Asia, particularly from an area where these vertebrates are not common. Silutitan sinensis is closely related to Euhelopus. The existence of a more inclusive clade of similar sauropods (Euhelopodidae) is still a matter of debate and pends on more detailed description of some putative euhelopodid.
Hamititan xinjiangensis is one of the few titanosaurian sauropod recovered from Asia, which shows an unusual combination of sauropod features. The presence of two somphospondylan species in the Tugulu Group novel information on somphospondylan evolution and provides further support for a widespread diversification of these sauropods during the Early Cretaceous of Asia.

Materials and methods
Anatomical terminology. We used the traditional "Romerian" terminology as proposed by Wilson 93,94 , using for example "anterior" rather than "cranial", as directional terms. For the identification and designation of vertebral laminae and fossae for Sauropoda we follow the landmark-based scheme proposed by Wilson 93,94 and Wilson et al. 95 respectively.
Heuristic tree search. The datasets of Filippi et al. 67 and Mannion et al. 53,54 were analyzed using the "New Technology Search". The algorithms ("Sectorial Search", "Ratchet", "Drift" and "Tree Fusing") are applied together with the traditional search procedures, such as Wagner Trees, Tree Branch Reconnection (TBR) and Subtree-Pruning-Regrafting (SPR) algorithms, to find the Minimum Length Trees (MLTs). A final round of TBR branch swapping was applied to the best trees obtained at the end of the replicates to find all of the Most Parsimonious Trees (MPTs).  55 . The characters (11,14,15,27,40,51,104,122,147,148,195,205,259,297,426,435,472 and 510) were ordered, as in the original analysis and first we use equal weighting of characters. Six unstable taxa were removed after the method of 87  www.nature.com/scientificreports/