Turiasauria-like teeth from the Upper Jurassic of the Lusitanian Basin, Portugal

Turiasauria is a clade of eusauropods with a wide stratigraphic range that could extend from the Bathonian to the lower Aptian including Turiasaurus, Losillasaurus, Zby and putatively, Galveosaurus, Atlasaurus and isolated remains from Middle Jurassic-to-Lower Cretaceous. Some are characterised by the presence of heart-shaped teeth. Several tooth occurrences from the Portuguese Upper Jurassic with this type of morphology (SI: 1.1–1.8) are reported and discussed. If this morphology is regarded as synapomorphic of Turiasauria, the teeth will be tentatively related to this clade. From a sample of 43 teeth, three main morphotypes are described. Three hypotheses might explain the morphological variation: (1) the range of tooth morphologies indicates variation in the jaw, (2) the range of tooth morphologies indicates taxonomic variation or (3) a combination of both. The general wear pattern in morphotypes I and II starts with a distal facet, then the appearance of mesial/apical facet and finally a ‘V’-shaped facet. In morphotype III, the wear begins with a mesial facet. The variability observed for Portuguese Upper Jurassic specimens is congruent with the morphological variability along the tooth row shown by other sauropods with spatulate/spoon-shaped teeth and it is considered the most parsimonious hypothesis to explain it.


Introduction
The sauropods are one of the vertebrate groups better represented in the last part of the Portuguese Upper Jurassic of the Lusitanian Basin (Kimmeridgian-Tithonian). Their study improves the understanding of vertebrate faunas and their paleobiogeography in this period. Recent works suggest that Iberian sauropods are represented by endemic genera (Dantas et al. 1998;Bonaparte and Mateus 1999;Casanovas et al. 2001;Antunes and Mateus 2003;Sánchez-Hernández 2005;Royo-Torres et al. 2006;Mateus et al. 2014) closely related to groups well represented in other continents during the Upper Jurassic like brachiosaurids (Antunes and Mateus 2003;Mannion et al. 2013), diplodocids (Bonaparte and Mateus 1999;Mannion et al. 2012;Mocho, Royo-Torres, Malafaia, et al. 2014) or camarasaurids ). The supposed close relationship of the Portuguese sauropods with taxa from the North American Upper Jurassic of the Morrison Formation (e.g. Lapparent and Zbyszewski 1957) is less close than it is interpreted in other dinosaur groups. In fact, there are references to genera and even species of theropods, ornithopods and stegosaurs with an amphiatlantic distribution (Galton 1980;Mohr 1989;Pérez-Moreno et al. 1999;Martin 2000;Mateus and Antunes 2000a;Mateus and Antunes 2000b;Ortega et al. 2006;Mateus et al. 2006;Escaso et al. 2007;Malafaia et al. 2007, 2010Ortega et al. 2009;Escaso et al. 2010).
This study analyses a sample of 43 sauropod teeth collected in several Upper Jurassic localities of the Lusitanian Basin. They are tentatively assigned to Turiasauria, a basal eusauropod clade, based on its heartshaped morphology. The meaning of three distinct morphotypes is also discussed. Several other heart-shaped teeth from Europe and Africa are also compared with the present sample. The information provided allows to propose a stratigraphic range for this tooth morphology in the Iberian Upper Jurassic wider than was previously thought.
Turiasauria is defined as a stem-based taxon including all eusauropods more closely related to T. riodevensis than to Saltasaurus loricatus Bonaparte andPowel 1980 (Royo-Torres et al. 2006). Therefore, it includes Turiasaurus, Losillasaurus, Zby, 'Neosodon', 'Cardiodon' andputatively Galveosaurus (Royo-Torres et al. 2006, 2009;Royo-Torres and Upchurch 2012;Mateus et al. 2014) and Atlasaurus (Royo-Torres, Xing et al. 2015). The inclusion in Turiasauria of three Spanish sauropods from the Villar del Arzobispo Formation: Turiasaurus, Losillasaurus (Tithonian-lower Berriasian) and Galveosaurus (Tithonianlower Berriasian) is based on some phylogenetic hypotheses (Royo-Torres et al. 2006, 2012Royo-Torres, Upchurch, et al. 2014;Royo-Torres and Upchurch 2012;Carballido and Sander 2014;. Galveosaurus has been considered in different phylogenetic positions along its history. It was considered as a cetiosaurid (Sánchez-Hernández 2005), a basal eusauropod , after it was proposed as a possible neosauropod , suggesting first its inclusion in Diplodocoidea (Barco 2005) and later in Macronaria? . Galveosaurus was also considered as a non-titanosauriform macronarian (Barco 2009; Barco et al. 2009;Carballido et al. 2011;Carballido and Sander 2014;Mannion et al. 2013). D'Emic (2012) noted the possibility that the holotype material of Galveosaurus might represent more than one individual. For this author, the presence of an elongate cervical vertebra and middle caudal vertebrae with anteriorly set neural arches might relate Galveosaurus to Titanosauriformes. The presence of a rounded proximolateral corner of the humeri suggested affinities to Brachiosauridae (D'Emic 2012), but this feature is also present in basal eusauropods, in particularly, in turiasaurs (e.g. Casanovas et al. 2001;Royo-Torres et al. 2006Royo-Torres and Upchurch 2012). The discovery of more material and an accurate systematic revision will be important to obtain a more precise phylogenetic approach for this taxon.
More recently, the presence of more turiasaurian occurrences in Spain, Portugal, France, the UK, Tanzania and Morocco has been suggested (Mateus 2009;Santos et al. 2009;Ortega et al. 2010;Cobos et al. 2011;Mocho et al. 2012;Royo-Torres and Upchurch 2012;Mateus et al. 2014;Suñer et al. 2014;Xing et al. 2015). From Spain, an unnamed specimen from Riodeva (Teruel) with postcranial material ) and an isolated caudal vertebra from Veguillas de la Sierra (Teruel) were related to Turiasauria . Both specimens come from the Villar del Arzobispo Formation, considered as Tithonian-lower Berriasian in age (Mas et al. 1984(Mas et al. , 2004. Also, Canudo et al. (2010) refer to cf. Turiasaurus riodevensis a fragment of a dentary with some teeth from the Kimmeridgian of Asturias (Spain).
The first evidence of turiasaurian remains in the Lusitanian Basin was based on an isolated tooth (Royo-Torres et al. 2006 from Alcobac a Formation (Kimmeridgian). Subsequently, Mateus (2009) referred to Turiasaurus riodevensis, an incomplete specimen composed by an almost complete hindlimb associated with a scapula, a coracoid, a tooth and a middle chevron, collected in Vale de Pombas (ML 368). This specimen was firstly related to Camarasaurus (Mateus 2005), but a recently systematic revision established a new genus and species, Zby atlanticus, which was tentatively placed within Turiasauria as a closely related form to Turiasaurus riodevensis (Mateus et al. 2014). Ortega et al. (2010) assigned to Turiasaurus some teeth collected on the region of Torres Vedras (Portugal) from the upper Kimmeridgian-Tithonian sediments of the Freixial Fm. and Praia de Amoreira-Porto Novo Fm. Santos et al. (2009) described a new ichnospecies (Polyonyx gomesi Santos et al. 2009) in the Galinha tracksite from the Middle Jurassic of the Macic o Calcário Estremenho (Portugal) that they related to a basal eusauropod, probably a form within the Turiasauria clade. If these footprints belong to Turiasauria, they would be, for now, one of the most ancient occurrences for this clade, together with some teeth from the UK (see Royo-Torres and Upchurch 2012). This ichnospecies might also be present in the Villar del Arzobispo Fm. So far, no trackway has been described, but some isolated manus and pes has been related with Polyonyx Royo-Torres 2009;Santos et al. 2009) in two sites from the El Castellar locality in Teruel (Spain).
More recently, the reassessment of some Gondwanan sauropods from Tendaguru Formation (Tithonian) in Tanzania could provide evidence for the presence of Turiasauria in Gondwana during the Upper Jurassic, suggesting a wider paleobiogeographic range than previously thought. Royo-Torres and  related to Turiasauria a complete right manus (HMN MB.R.2093.1-12), a partial caudal series described by Bonaparte et al. (2000, HMN MB.R.2091, an astragalus (HMN MB.R.2095.6) and a humerus (HMN MB.R.2910). Royo-Torres,  and Xing et al. (2015) also suggested the placement of Atlasaurus from the Middle Jurassic (Bathonian-Callovian) of Morocco inside Turiasauria.

Geological settings
The described teeth were mainly collected northwestern of Lisbon, along the coastal cliffs between the localities of Cambelas and Praia da Gralha in Torres Vedras, Lourinhã, Caldas da Rainha, Alcobac a and Peniche municipalities ( Figure 1). In this area, outcrops an Upper Jurassic to Lower Cretaceous sedimentary sequence, deposited in the Lusitanian Basin. The Upper Jurassic beds are dated from middle Oxfordian to the base of Cretaceous (Schneider et al. 2009), and represent a third rifting episode (Rasmussen et al. 1998, Kullberg et al. 2006) marked by an internal differentiation resulting in the formation of several sub-basins (Turcifal, Arruda and Bombarral subbasins) followed by an important siliciclastic input which progressively filled these sub-basins (Hill 1988;Pena dos Reis et al. 2000;Kullberg et al. 2006). Since the Kimmeridgian, the sedimentary sequence is marked by a strong siliciclastic nature, with a continental signature on Figure 1. Geological map (adapted from Oliveira et al. 1992) showing the Portuguese Mesozoic levels and the localities from where come the studied heart-shaped teeth. The number in parenthesis is the number of teeth found in each locality. the top of the sequence corresponding to the last part of the Upper Jurassic (e.g. Hill 1988;Manuppella et al. 1999;Kullberg et al. 2006). Different stratigraphic approaches have been proposed for the Upper Jurassic sequence of these sub-basins (e.g. Hill 1988;Leinfelder 1993;Manuppella et al. 1999;Kullberg et al. 2006;Schneider et al. 2009;Martinius and Gowland 2011;Taylor et al. 2014; see Figure 2). The teeth described were found in sediments from several Upper Jurassic formations of the Lusitanian Basin, including Montejunto (?), Alcobac a, Praia da Amoreira-Porto Novo, Sobral, Freixial and Bombarral Formations. All of them, except the Montejunto Fm., were included in the Lourinhã Group proposed by Yagüe et al. (2006). One of the first references of this tooth morphology in the Portuguese Upper Jurassic was found near Ourém (MG 16, Sauvage 1897 -98), probably from Montejunto Fm. or Alcobac a Fm, upper Oxfordian or Kimmeridgian-basal Tithonian in age, respectively (Mouterde et al. 1979;Manuppella et al. 1999Manuppella et al. , 2000. Most of the teeth come from the 'Lourinhã Formation' (unit proposed by Hill 1988), which includes the Praia da Amoreira-Porto Novo (included in the Alcobac a Beds sensu Manuppella et al. 1999), Sobral, Freixial andBombarral Formations (Manuppella et al. 1999) (Figure 2). Figure 2 shows the correspondences among some stratigraphic approaches as well as the stratigraphic position of the teeth analysed here. "Lourinhã Formation" sensu Hill (1988) is interpreted as upper Kimmeridgian-tobasal Berriasian in age (Leinfelder 1986;Hill 1988;Leinfelder and Wilson 1989;Mohr 1989;Manuppella et al. 1999). This unity was subdivided in five members by Hill (1988) that in part corresponds to the formations proposed by Manuppella et al. (1999) (see Figure 2). The Sobral Fm. ( ¼ Praia Azul member of Lourinhã Fm. sensu Hill 1988) is a relatively well-dated unit, upper  Kullberg et al. 2006;Schneider et al. 2009), Bombarral sub-basin (based on Manuppella et al. 1999), the coastal sector from Porto da Calada to Salir do Porto (based on Hill 1988) and Alcobac a region (based on Kullberg et al. 2006;Azerêdo et al. 2010), and the respective stratigraphic position of the described teeth by morphotype. Crn, Chronostratigraphy; U, upper; M, middle; L, lower; *sensu Yagüe et al. 2006; **other formations are identified in Lourinhã region by Manuppella et al. 1999 as lateral correlatives of Bombarral Fm.
Kimmeridgian to lower Tithonian in age (Fürsich 1981). It is also important to refer that the coastal sedimentary sequence south of Sizandro river mouth, which corresponds to the transition of Sobral Fm. and Freixial Fm. (Assenta member of Hill 1988), is progressively younger to the south, with a stratigraphic range from lower to upper (?) Tithonian (Leinfelder 1987;Hill 1988;Schneider et al. 2009). Some teeth were found in the coastal cliffs of Salir do Porto and São Martinho do Porto and in Fervenc a locality where outcrops the Alcobac a Fm. (sensu Camarate Franc a and Zbyszewski 1963; Azerêdo et al. 2010). The Alcobac a Fm. in Salir do Porto was dated to the lower Kimmeridgian (Schneider et al. 2009 (Manuppella et al. 1999). Summarising, all the analysing heart-shaped teeth from Portugal are recorded from the upper Oxfordian (?) to the lower-to-upper Tithonian on the Upper Jurassic of the Lusitanian Basin.

Description of general morphology
Each tooth crown has a heart-shaped spoon-like morphology, compressed labiolingually and presenting an enamel with wrinkled texture. In generally, the crown is slightly apicomesially projected and the teeth reach the maximum mesiodistally width near the base of the apex (the apex is considered herein as the apical portion of the tooth, apical to the sagittal deflection of mesial and distal edge). Excluding the most worn teeth, the slenderness index (SI: crown height/maximum crown breadth; sensu Upchurch 1998) ranges between 1.1 and 1.8.
On the labial face, the teeth display an apicobasal bulge bounded by shallow grooves with the same orientation. The lingual face has a low apicobasal ridge, which might extend along the entire apicobasal length. The mesial and distal edges are not parallel and diverge from the base of the tooth. The transition between the row and crown is marked in all the teeth. The teeth exhibit asymmetrical 'D' to lenticular-shaped cross section with a strongly convexity labial face and a flat-to-smooth concave lingual face. The maximum labiolingual width is located near the mesial edge, resulting in steeply angled mesial part on labial surface. The asymmetrical apex deflects distally and could bear mesial, distal and apical wear facets depending of the wear development. The mesial and distal edges of the apex are straight to slightly convex and concave, respectively, in labial/lingual view. Generally, the distal edge of the apex is longer than the mesial one (excluding the morphotype III, see below). Crown-to-crown occlusion produced "V"-shaped wear facets. Some teeth are heavily worn and, in some cases, it is possible to observe the dentine. The wear facets will be commented in detail after the definition and description of the three proposed morphotypes.
The wrinkling pattern of the enamel is similar to those present in several spatulate-or spoon-like teeth (e.g. Camarasaurus, Turiasaurus, "Neosodon") marked by an alternation between apicobasal and waved grooves and ridges. Along its length, these ridges join together forming an anastomosed pattern. This pattern is smoother at the tip of the apex probably because of abrasion, like in other sauropod teeth (e.g. Amygdalodon, Carballido and Pol 2010).

Definition of the morphotypes
Despite a general morphology shared by every studied tooth, a division into three different morphotypes is Historical Biology  Figure 3): The heart-shaped crowns are more apicobasally elongated and labiolingually compressed than the morphotypes II and III. They also bear the higher SI values, ranging between 1.8 and 1.6 (teeth with moderate wear have lower SI values, around 1.5-1.4). The labial face is markedly convex mesiodistally and weakly convex apicobasally. The lingual face is concave apicobasally and mesiodistally and the lingual crest can occupy all the apicobasal extension. In this morphotype, mesial and distal edges in the base of crown are closely parallel and straighter than in the  ? ?
other two morphotypes. The apex is particularly longer in these teeth. It occupies one half or more than the total crown height (apex/crown height ratio . 0.5). In this morphotype, the apex presents a slight distal deflection, which is not so pronounced as in the morphotype II. The distal edge of the apex is concave and the mesial one is convex-to-straight. Both are criteria for the orientation of these teeth.
Morphotype II (SHN (JJS) 130,135,141,147,148;SHN 138,144,145,150,153,503,506,508,509,511,512, see Figure 4): It is more abundant in the sample, with a well-defined heart-shaped outline, crown projected apicomesially and a more pronounced curvature of the apex than in the morphotype I.  Figure 5): This corresponds to heart-to subsquared-shaped teeth more compressed labiolingually and shorter than the other two morphotypes (SI , 1.3). The labial face is not so convex as in morphotypes I and II. The lingual face is concave apicobasally and flat-to-concave mesiodistally with a similar platform at the base of the crown as that occurs in the morphotype II, but less pronounced. The lingual apicobasal crest in this morphotype is incipientto-absent. On the distal edge, in the transition between the apex and the base of the crown, there is a round shoulder lingually projected resulting from the concavity of the distal edge of the apex. Unlike morphotypes I and II, the distal edge is shorter than the mesial edge. The apex is shorter than in morphotypes I and II (apex/crown height ratio , 0.3). SHN (JJS) 139 ( Figure 5(a)) presents a morphology between morphotypes II (SI higher than morphotype III) and III (apex shorter than morphotype II), but it is considered here as morphotype III because of the presence of a short apex.

Wear pattern
The sample presents variable wear patterns. We consider here three main states of wear: (i) absent or weak, without wear facets (e.g. Figure 4(a),(c),(e)), probably non-functional teeth, or teeth with marked mesial wear facets, but sometimes with a slight worn on the distal edge of the apex (e.g. Figures 3(b), 4(d) or 5(c)); (ii) moderate, wear facets in both edges of the apex and an incipient (Figures 3 (d) and 4(b)) or well-marked (e.g. Figures 3(c), 4(h), 6(b), 10(d) -(e)) apical wear facet; (iii) strong, the three wear facets are fused in a unique 'V'-shaped wear facet ( Figure 6(d),(e),(f)). The wear facet on the distal edge is generally longer and more developed than the mesial one, except in the teeth of morphotype III. As far as can be checked, the wearing in morphotypes I and II begin in the distal edge (Figure 3(b) and 4(d)). In less worn teeth, the distal wear is always present and when the mesial one is also present, the distal one is more developed than the mesial one. In a more advanced wear state (moderate), the distal facet becomes more pronounced and the mesial one becomes well defined. An incipient-to-marked apical wear facet is generally associated with moderately worn teeth (e.g. SHN (JJS) 140 or SHN 508; Figures 3(a) and 4(b)).
In more advanced stages of wear (strong wear), the mesial, distal and apical facets produce a unique 'V'-shaped facet (SHN (JJS) 129 or SHN (JJS) 130, Figure 6). The morphotype III teeth show a different condition from the previous morphotypes: the wearing starts in the mesial edge, as in SHN 137 ( Figure 5(c)). Generally, teeth with incipient wear present wear facets with high lingual or labial inclination (almost vertical). More developed wear facets become progressively more subhorizontal (e.g. SHN (JSS) 141). In almost all studied teeth, the wear facets slope lingually, which might suggest a maxillary/premaxillary position (see Nowinski 1971, Calvo 1994, Barrett and Upchurch 1994, Upchurch and Barrett 2000, Carballido and Pol 2010. In SHN (JJS) 140, 141 and 148, the mesial and distal wear facets slope lingually. Nevertheless, on these teeth, the apical wear facet faces labially, what could suggest a mandibular position. In conclusion, the wear pattern in morphotypes I and II begins with the appearance of a distal facet (e.g. SHN (JJS) 142, SHN 138), and then the mesial and apical facets appear (e.g. SHN (JJS) 140, SHN 508) and finally a 'V'-shaped facet develops. In morphotype III, the wear begins with a mesial facet (SHN 137), and then appear a distal and apical facets (SHN (JJS) 139) a finally a 'V'-shaped facet (SHN (JJS) 149). Additionally, in some teeth, it is possible to observe a wear facet on the medial/distal side of the base (e.g. SHN (JJS) 148, Figure 6(d)).

Discussion
Enamel-wrinkled texture, crown overlapping, spoonshaped crowns and "V"-shaped wear facets were traditionally considered as synapomorphic traits of eusauropod teeth (Wilson and Sereno 1998). Nevertheless, recent works show that these features appear earlier than expected in sauropod evolution (e.g. Upchurch, ; Upchurch, Barrett, Xijin, et al. 2007;Yates 2007;Allain and Aquesbi 2008;Carballido and Pol 2010). Regardless of the morphological variability of the discussed teeth, we might tentatively relate these teeth to the Turiasauria clade based on the presence of the following diagnostic features proposed by Royo-Torres et al. (2006): (i) heart-shaped crowns; (ii) a pointed and asymmetrical apex that is strongly compressed labiolingually and (iii) crowns with convex labial surfaces with a bulge extending apicobasally. A pointed and asymmetrical apex that is strongly compressed labiolingually is well developed in Turiasaurus and the rest of the Iberian heartshaped teeth. Giraffatitan also present a similar morphology for the apex (Janensch 1936). Crowns with convex labial surfaces with a bulge extending apicobasally are also shared with Amygdalodon (Carballido and Pol 2010), Patagosaurus (Bonaparte 1986) and Tazoudasaurus (Allain and Aquesbi 2008). Finally, besides the presence of several isolated heart-shaped teeth along the Middle Jurassic to Early Cretaceous, at the moment, the heartshaped tooth morphology is exclusively related with the turiasaurs Turiasaurus Heart-shaped morphology with pointed and asymmetrical apex that is strongly compressed labiolingually results in a so far exclusive combination of Turiasauria. The Turiasaurus riodevensis teeth (Royo-Torres et al. 2006;Royo-Torres and Upchurch 2012) are particularly similar to those of the morphotype II identified in the Portuguese Upper Jurassic, and some of them are virtually indistinguishable. The phylogenetic revision of the Iberian turiasaurs is in progress, which will include some new specimens, and will provide new information about the phylogenetic distribution of these characters along the eusauropod evolutionary history.
The SI values of our sample range between 1.1 and 1.8 and are probably related with the tooth morphology: lower, intermediate and higher SI values correspond to morphotypes III, II and I, respectively. This index was defined by Upchurch (1998) and it have been used to compare taxa and to understand evolutionary trends in teeth morphology and feeding mechanisms (e.g. Salgado and Calvo 1997;Barrett et al. 2002;Barrett and Wang 2007;Wilson and Upchurch 2009;Chure et al. 2010;Mannion 2010;Saegusa and Tomida 2011). Several eusauropods and basal macronarians with spoon-or Modified from Chure et al. (2010), in Figure 7 we plotted the logged SI values found in the Portuguese Upper Jurassic heart-shaped teeth (excluding teeth with strong wear) together with other sauropodomorphs. The average for SI in Turiasaurus teeth (1.25) and in the Portuguese specimens (1.40) is slightly different, probably due to the size of the sample (three teeth) for Turiasaurus, which is composed by teeth with a morphology close to those of the morphotype II. Nevertheless, the SI average of Turiasaurus fits in the range of the Portuguese Upper Jurassic heart-shaped teeth. The SI range of our sample does not show any particular trend and they set within the morphospace occupied by non-neosauropod sauropods (Figure 7) as well as, Turiasaurus teeth. The presence of SI values close to 1 fits the morphotype III outside the morphospace occupied by the non-neosauropod sauropods.
We also plotted the Log 10 SI for Sauropodomorpha, non-neosauropod eusauropods and five sauropod clades: Diplodocoidea, Brachiosauridae, Euhelopodidae and Lithostrotia (using Chure et al. 2010;D'Emic et al. 2013 data; see Appendix 2) versus the number of genera that have teeth included in our data (Figure 8). Turiasaurus riodevensis and the Portuguese Upper Jurassic heartshaped are plotted together and they show a narrow range  (Martínez et al. 2000). Mateus (2009) assigned as Turiasaurus riodevensis a tooth and postcranial material (ML 368) found in Vale das Pombas (Lourinhã), previously related with Camarasaurus (Mateus 2005), and now part of the Zby atlanticus holotype (Mateus et al. 2014). Two other sauropod teeth referred by Sauvage (1897 -98) as 'Pelorosaurus humerocristatus' (Hulke 1874) found in Fervenc a (MG 277, Figure 9(a)) (lower Figure 7. Temporal patterns in sauropodomorph tooth shape modified from Chure et al. (2010). The plot shows the logged tooth SI for sauropodomorph genera throughout the Mesozoic. Orange field indicates non-sauropod sauropodomorphs, brown field indicates basal sauropods, pink field indicates diplodocoids and blue field indicates macronarians. The uncertain phylogenetic position of Jobaria is indicated by cross-hatching; the transparent blue and brown fields indicate that the shape of the tooth space when Jobaria is included within macronarians and basal sauropods, respectively. Time scale based on Gradstein et al. (2004). The grey line is the SI range of the Portuguese Upper Jurassic teeth; the yellow, green and red circles represent the average Log (SI) for morphotypes I, II and III, respectively.
Royo-Torres et al. (2006,2009) also included in Turiasauria the heart-shaped teeth related to 'Neosodon' (invalid taxon according to Upchurch et al. 2004) from the Tithonian of France (figured in Buffetaut and Martín 1993).  distinguished them from Turiasaurus because they exhibit a more concave lingual face and a greater apicobasal development in the terminal part of apex, a feature shared with the herein proposed morphotype I. In fact, it is possible to identify, at least, two of the three morphotypes defined in this study, morphotypes I (BHN2R 113 and BHN2R 1102) and II (BHN2R 1101) (see figured teeth in Buffetaut and Martín 1993). Buffetaut and Martín (1993) also assigned to 'Neosodon' the teeth found in Fervenc a and Ourém, referred earlier. The morphological features present in the heart-shaped teeth of the French Upper Jurassic are not diagnostic and they share their overall morphology with heart-shaped teeth of Iberian Upper Jurassic (e.g. Royo-Torres et al. 2006;Mateus et al. 2014;this work) or the Middle/Upper Jurassic of the UK (Figure 10(a) -(c),(e)); Royo-Torres and Upchurch 2012) and the Lower Cretaceous of France (Néraudeau et al. 2012) and the UK (e.g. Figure 10(d), Lydekker 1888Lydekker , 1889Upchurch et al. 2011). 'Neosodon' must be considered an invalid taxon because of the absence of diagnostic features. The teeth should not be used as infrageneric determinations as noted, for example, Canudo et al. (2002) or García and Cerda (2010). The SI for 'Neosodon' teeth (< 1.26 -1.64) was obtained using the figured specimens on Buffetaut and Martín (1993), and fits in the SI range of Portuguese heartshaped teeth. Buffetaut and Martín (1993) also warn for the similarities among 'Neosodon' teeth and the Cardiodon tooth from the British Bathonian. The specimen of Cardiodon figured in Owen (1844Owen ( , 1875 shows a similar morphology to that found in Turiasaurus and 'Neosodon', which allowed Royo-Torres et al. (2006 to consider it as a Turiasauria member. Cardiodon is included in Eusauropoda because it has teeth with a broad spatulate outline, wrinkled enamel and a groove on the labial surface near the distal margin (Upchurch et al. 2004). The name Cardiodon has been retained because some authors consider that there are is distinct character to all other known spoon-like sauropod teeth: mildly convex lingual face (Upchurch and Martin 2003;Upchurch et al. 2004;Mannion 2010). However, there are other sauropods with  spoon-shaped teeth that also bear a convex lingual face, such as Amygdalodon (Carballido and Pol 2010). Furthermore, the heart-shaped teeth discussed here have a transversely convex lingual face at the base of the crown, which becomes transversely flat to slightly concave apically. This condition is also observed in a tooth referred to 'Cardiodon rugolosus' (Steel 1970;Upchurch and Martin 2003;NHMUK R1527) which has a transversely convex lingual face at the base of the crown and transversely flat to slightly concave lingual face apically. Taking into account the absence of a detailed description and figuration of this tooth, which allows testing the exclusivity of Cardiodon tooth morphology, and the loss of this specimen, we regard Cardiodon as nomina dubia, till the discovery of the type specimen or related information. Three teeth from the Lower Oxford Clay (middle Callovian) of Cambridgeshire (Martill 1988;Barrett 2006) previously assigned to 'Cetiosauriscus leedsi' (NHMUK R3377, Figure 10(b),(c)) also preserve a heart-shaped morphology previously related to Turiasauria (Royo-Torres and Upchurch 2012). Several specimens referred to 'Hoplosaurus' and 'Pelorosaurus' from the Upper Jurassic of the UK (NHMUK R2822, NHMUK R2565, NHMUK R2004-5, Figure 10(a),(e)) also fit in the general morphology here described. Buffetaut and Martín (1993) pointed the similarities between 'Neosodon' and Camarasaurus teeth. Furthermore, Mateus (2005) related preliminary the Zby holotype with a heart-shaped tooth to Camarasaurus genus. However, it is possible to identify some differences in Camarasaurus teeth (e.g. Osborn and Mook 1921;Ostrom and McIntosh 1966;McIntosh et al. 1996): (i) generally higher than heart-shaped teeth (SI: 1.53-2.447 sensu Chure et al. 2010); (ii) distal and mesial edges are straight and almost parallel in the base of the crown, while just the morphotype I of the heart-shaped teeth shows some degree of straightness and parallelism; (iii) the former shows shorter apices; (iv) Camarasaurus bears lingual facets and a marked and complex cingulum on the lingual face; (v) a lingual crest that appears only in the apex sector; (vi) lack of heart-shaped morphology that probably characterises the Turiasauria clade (the heart-shaped morphology fits in the spoon-shaped morphology and Camarasaurus was characterised by a spatulate-shape morphology); (vii) lingual projection of apex more pronounced than heartshaped teeth; (viii) the labial face of the apex is more inflated and globose in Camarasaurus teeth. The most distal teeth of Camarasaurus could share some of these features with heart-shaped teeth.
In Europe, the heart-shaped morphology is not exclusive from the Middle and Upper Jurassic, being identified a few occurrences in the Lower Cretaceous. From Hauterivian-Barremian sediments of Angeac, in France, were found heart-shaped teeth (Néraudeau et al. 2012). Néraudeau et al. (2012) noted the similarities between the morphology of these teeth with Turiasaurus riodevensis and other Jurassic and Cretaceous occurrences. In the UK and Spain, there are also some heartshaped teeth occurrences: (i) the Lower Cretaceous tooth figured in Lydekker (1889, NHMUK R1610, Figure 10 features not yet identified in Upper Jurassic heart-shaped teeth such as the presence of vertically and mesiodistally oriented apical wear facets and the presence of lingual facets and a complex cingulum morphology that is also found in Camarasaurus or Euhelopus (e.g. Ostrom and McIntosh 1966;Wilson and Upchurch 2009).
Three teeth from Africa are related to the heart-shaped morphology. From the Middle Jurassic sediments of Madagascar, more precisely in Ankinganivalaka site (Läng 2008), was found a tooth (MNHN.F MAJ 423, Figure 10(f)) with an almost complete crown. This crown bears the typical heart-shaped morphology present in the morphotype II defined here. Another tooth (MNHN.F 1961-28, Figure 10(g)) from In Gall (Niger) (Lapparent 1960) also bears a heart-shaped morphology shared by the morphotype II. This tooth came from Irhazer Group sediments, probably not younger than upper Middle Jurassic (Rauhut and López-Arbarello 2009). Finally, another tooth (UT-TEN15) from Tendamirah Quarry, Cabao Formation (Hauterivian-Barremian) in Libya (Le Loeuff et al. 2010) has a basal constriction and a crown similar to the morphotype I, and lacks the complex cingulum with associated lingual facets present in Camarasaurus or Euhelopus (Ostrom and McIntosh 1966;Wilson and Upchurch 2009).
Until the end of the twentieth century, neosauropod postcranial references dominate in Portuguese Upper Jurassic. This is incongruent with the relative abundance of these type teeth morphology assigned to Eusauropoda. However, the specimen of Vale das Pombas, Zby atlanticus (Mateus 2009;Mateus et al. 2014), and some new material in study, show that the occurrence of eusauropod postcranial material is not so rare. Therefore, the hypothesis that these teeth belong to a neosauropod form by convergence, or that this morphology corresponds to a more inclusive group than Turiasauria is not ruled out. New discoveries are necessary to confirm the link between these teeth and the occurrences related to Turiasauria.
Royo-Torres et al. (2006,2009) and Royo-Torres and Upchurch (2012) considered that the heart-shaped morphology could be referred to the Turiasauria. At the moment, Turiasaurus and Zby are the only turiasaurs with cranial and postcranial materials. In the light of some recent phylogenetic approaches (Royo-Torres et al. 2006, 2012Royo-Torres, Upchurch, et al. 2014;Royo-Torres and Upchurch 2012;Mocho, Royo-Torres, Malafaia, et al. 2014), the non-neosauropod eusauropods have spoon-shaped teeth. If Turiasaurus, Losillasaurus, Galveosaurus and Zby correspond to a monophyletic clade, their heart-shaped teeth could be considered as a synapomorphy of Turiasauria (the condition is unknown in Galveosaurus and Losillasaurus). However, the presence of this type of tooth morphology in the Middle Jurassic to the Lower Cretaceous of Africa and Europe put in evidence that this particular morphology has a wider stratigraphic and paleogeographic distribution, that could reflect a wider phylogenetic distribution.
Other possibility is to consider that this morphology was acquired by convergence in several sauropod groups. This hypothesis could explain the presence of this tooth morphology in the Middle Jurassic of the UK and in the Lower Cretaceous of France and the UK, where it was not yet found or documented other turiasaurian cranial (non-teeth material) and postcranial remains. The presence of convergences in sauropod tooth morphology has already been identified between diplodocids and titanosaurs (e.g. Salgado and Calvo 1997) or mamenchisaurids and some macronarians (e.g. Suteethorn et al. 2013). Convergence in tooth morphotypes has also been suggested for brachiosaurids and titanosaurs (Chure et al. 2010) and both with euhelopodids (D'Emic et al. 2013).
The differences shown by the three proposed morphotypes can be explained by two different ways (or a combination of both): (i) the three morphotypes represents three distinct taxa (in generic or specific level) inside or outside the Turiasauria clade (this morphology could not be exclusive of the clade); or (ii) the three morphotypes belong to the same taxon and the variability is associated to a distinct position along the tooth row, as occurs in other sauropods such as Giraffatitan (Janensch 1936), Camarasaurus (Gilmore 1925), Abydosaurus (Chure et al. 2010) or Euhelopus (Wiman 1929;Wilson and Upchurch 2009;Poropat and Kear 2013). A slight heterodonty was also suggested for the skull of Turiasaurus riodevensis (Royo-Torres and Upchurch 2012). The distinct wear pattern observed along the morphotypes also could be explained in the same way, or might represent a distinct taxon or are function of the tooth row position.
Observing the variation in teeth morphology along the tooth row for Camarasaurus (Gilmore 1925) or Giraffatitan (Janensch 1936), some remarkable trends to the distal part of the tooth row can be enumerated: (i) decreasing of SI value (SI value varies at least a unit in both taxa: Camarasaurus < 2.7-1.7; and Giraffatitan < 3.5-2.5); (ii) decreasing of apex height; (iii) prominence in distal curvature of the apex; and (iv) progressive medial tooth imbrication. Assuming that the teeth described belong to a single taxon, the defined morphotypes (I, II and III) could represent different positions in tooth row fitting well in the variability observed for Camarasaurus (Gilmore 1925) or Giraffatitan (Janensch 1936). In this case, morphotype I should correspond to an anterior position, morphotype III to a more posterior position, and morphotype II located between morphotypes I and III. It is necessary to have a well-represented in situ tooth sequence of a turiasaurian individual to confirm if the three morphotypes fit in the range of a unique species or otherwise some of the morphotypes represent distinct taxa (inside or outside Turiasauria). Anyway, taking into account the variability present in other taxa, the variation in teeth morphology along the tooth row of a unique taxa seems to be the most parsimonious hypotheses to explain the morphological variability in the Iberian sample.

Conclusions
Forty-three heart-shaped teeth from the Portuguese Upper Jurassic were described and tentatively referred to Turiasauria based on the presence of a heart-shaped crown and a pointed and distally projected apex. Till today, this tooth morphology was the only one found associated with skeletal remains in Turiasaurus and Zby, both considered as members of Turiasauria, suggesting that the heart-shaped morphology could be referred as a possible synapomorphy of this clade. This sample shows a great variability among which can be recognised three different morphotypes: morphotype I, high SI (1.8 -1.6) values and high apex (an half of tooth total height), morphotype II, moderate apex and SI values (1.5 -1.3), with well defined heart-shaped and more strong distal deflection; morphotype III, low SI values (, 1.3) and extremely low apex, with heart-shaped to subsquaredshaped form. Morphotypes I and II set within the morphospace occupied by other non-neosauropod sauropods, with the exception of the morphotype III, with lower SI values.
To explain the present morphological variability, two hypotheses (or combination of both) were mainly discussed: (i) these teeth belong to distinct taxa outside or inside Turiasauria or (ii) these teeth correspond to different positions on the tooth row of unique taxa. The morphological disparity shown by few sauropods (e.g. Turiasaurus or Camarasaurus) along tooth row suggests that this variability could be explained by a slightly heterodonty with morphotypes I, II and III located in mesial, middle and distal position, respectively. The presence of several teeth in different states of wear allows proposing a hypothetical general wear pattern for morphotypes I, II and III. In the former two morphotypes, the wear begins with the appearance of a distal facet, then the appearance of mesial and apical facets and finally a 'V'-shaped facet. In morphotype III, the wear seems to begin with the appearance of a mesial facet.

Supplemental data
Supplemental data for this article can be accessed here.